Datasets:

Hendrik-a

/

arxiv

like 0

the ordering of the classical heisenberg model on the pyrochlore lattice with the antiferromagnetic nearest - neighbor interaction and the ferromagnetic next - nearest - neighbour interaction is investigated by means of a mean - field analysis and a monte carlo simulation . for a moderate-value , the model exhibits a first - order transition into an incommensurate multiple- ordered state where multiple bragg peaks coexist in the spin structure factor . we show that there are two types of metastable multiple- states , a cubic symmetric sextuple- state and a non - cubic symmetric quadruple- state . based on a monte carlo simulation , we find that the cubic sextuple- state appears just below the first - order transition temperature , while another transition from the cubic sextuple- state to the non - cubic quadruple- state occurs at a lower temperature .

1107.5440

recently , geometrically frustrated magnets have attracted much interest due to their unconventional ordering behaviors@xcite . spin systems on the pyrochlore lattice , which consists of a three - dimensional network of corner - sharing tetrahedra ( fig.[fig_pyro ] ) , are typical examples of such geometrically frustrated magnets . the classical heisenberg magnet with the antiferromagnetic nearest - neighbor ( nn ) interaction is known to exhibit no magnetic long - range order even at zero temperature @xcite . this is due to the macroscopic degeneracy of the ground state induced by geometrical frustration . since such a high degeneracy is realized via a fine balance among frustrated interactions , it might be lifted by introducing small perturbations , _ e.g. _ the further - neighbor interactions@xcite , the quenched randomness@xcite , or the lattice distortion@xcite . the lifting of the degeneracy as a result of small perturbations might lead to an exotic magnetic state peculiar to geometrical frustration . and the next - nearest - neighbor interaction @xmath1 are indicated.,width=226 ] in this paper , we focus on the effects of the next - nearest neighbor ( nnn ) interaction on the pyrochlore - lattice classical heisenberg model . effects of the further neighbor interactions were investigated within a mean - field approximation by reimers _ up to the fourth neighbors @xcite . in the case of the nnn interaction @xmath1 only , they showed that a @xmath4 order was stabilized for the antiferromagnetic nnn interaction @xmath5 , while an incommensurate @xmath3 order was stabilized for the ferromagnetic nnn @xmath6 . since then , several monte carlo ( mc ) simulations on this @xmath0-@xmath1 pyrochlore heisenberg model were performed @xcite . for the antiferromagnetic @xmath5 , these works revealed that the system exhibited a first - order transition from the paramagnetic phase to the @xmath4 ordered phase with a collinear up - up - down - down spin structure @xcite . selection of such a collinear structure among possible @xmath4 states might be due to `` order - by - disorder '' mechanism , where thermal fluctuations lift the degeneracy through an entropic contribution to the free energy @xcite . for the ferromagnetic @xmath7 , on the other hand , the situation seemed to be more subtle . _ observed that the model with @xmath8 exhibited a first - order transition into a peculiar ordered state , where enhanced spin fluctuations apparently coexisted with sharp bragg peaks associated with a multiple-@xmath3 structure @xcite . _ studied the case of smaller @xmath9 and showed that there occurred successive phase transitions with an intermediate phase characterized by a finite nematic order parameter and a layered structure , which was not predicted in the mean - field analysis @xcite . et al . _ also pointed out that the low - temperature phase was a multiple-@xmath3 state which were basically the same as the one observed by tsuneishi _ et al_. in any case , the explicit spin configuration of the multiple-@xmath3 state was not identified so far . in the present paper , we investigate the nature of the multiple-@xmath3 state observed in the @xmath0-@xmath1 pyrochlore - lattice heisenberg antiferromagnet with the ferromagnetic nnn interaction on the basis of a mean - field analysis and an extensive mc simulation . particular attention is paid to the explicit spin configuration of the multiple-@xmath3 state and the nature of spin fluctuations in such state . we see that there are mainly two stable multiple-@xmath3 structures , the sextuple-@xmath3 state and the quadruple-@xmath3 state , each of which is characterized by whether they keep the cubic lattice symmetry or not . a first - order transition observed in earlier studies corresponds to the transition from the paramagnetic state to the cubic - symmetric multiple-@xmath3 ( sextuple-@xmath3 ) state . we predict that another phase transition might occur at a lower temperature from the cubic - symmetric multiple-@xmath3 state to the non - cubic multiple-@xmath3 ( quadruple-@xmath3 ) state . reflecting their characteristic spin fluctuations , internal fields in such multiple-@xmath3 states exhibit a broad distribution . the rest of the paper is organized as follows . in sec . [ model ] , we describe our model and briefly review the previous results on the model . in sec . [ mean ] , we present the result of our mean - field calculation determining the explicit spin configurations of the possible multiple-@xmath3 states . our mc results are presented in sec . the results are discussed in conjunction with the mean - field result , including the explicit spin configurations of the multiple-@xmath3 states . finally , we summarize our results in sec . [ conclusion ] . the model considered is the classical heisenberg antiferromagnet on the pyrochlore lattice , whose hamiltonian is given by @xmath10 where the first and the second sums are taken over all nn and nnn pairs @xmath0 and @xmath1 ( see fig.[fig_pyro ] ) , respectively . we suppose the antiferromagnetic nn interaction @xmath11 and the ferromagnetic nnn interaction @xmath6 ( @xmath0-@xmath1 model ) , and @xmath12 to be the length of the cubic unit cell . note that the pyrochlore lattice , which can be viewed as an fcc lattice formed by regular tetrahedra of a fixed orientation , is a non - bravais lattice . the ordering of this @xmath0-@xmath1 model was first studied by reimers _ based on a mean - field approximation . there , the full density matrix of the system was approximated by a product of single spin density matrices with local effective fields which were determined so as to minimize the free energy . _ found that the model exhibited a phase transition from the paramagnetic phase to a magnetically ordered phase , where the magnetic long range order was characterized by twelve incommensurate wavevectors of the fcc bravais lattice : @xmath13,@xmath14 , @xmath15,@xmath16 , @xmath17,@xmath18 and their minus , with @xmath19 @xcite . since the pyrochlore lattice consists of four fcc sublattices , each wavevector has four independent eigenmodes and only a certain combination of them becomes unstable at the transition point . in fact , there are infinitely many ways to mix such twelve critical modes . thus , just the determination of unstable modes is not enough to specify the explicit spin configuration of the ordered phase . indeed , reimers _ _ did not determine the ordered spin configuration , nor specified even whether it was a single-@xmath3 state or a multiple-@xmath3 state . multiple-@xmath3 state is generally incompatible with the fixed spin - length condition @xmath20 so that it is not favored in the classical spin system at low enough temperatures where the fixed spin - length condition needs to be observed . meanwhile , the unstable critical mode of the present model entails the magnitudes of frozen spin moments differ from one sublattice to the other so that even a simple single-@xmath3 state is incompatible with the fixed spin - length condition . it means that the multiple-@xmath3 state might have a higher chance to be stabilized , particularly at moderate temperatures where thermal fluctuations play a role . as mentioned , a multiple-@xmath3 ordered state characterized by multiple bragg peaks in the spin structure factor was reported in previous mc simulations on the model @xcite . interestingly , tsuneishi _ et al . _ also reported that this multiple-@xmath3 ordered state accompanies only small amount of spin freezing , and some sort of spin fluctuations apparently coexisted with sharp bragg peaks . the detailed spin structure of the multiple-@xmath3 state , however , has not been clarified so far . especially , the origin of the spin fluctuation reported by tsuneishi _ _ remains unsolved . chern _ et al . _ reported that yet another phase , a partially ordered collinear phase , might be realized between the paramagnetic phase and the multiple-@xmath3 phase for sufficiently small values of @xmath21 @xcite . this partially ordered collinear phase is characterized by a finite nematic order parameter and a layered spin structure . note that this phase is different from the multiple-@xmath3 ordered phase discussed above where the system retains a magnetic long range - order characterized by sharp bragg peaks . although the partially ordered collinear phase as discussed by chern _ et al . _ is also interesting , we focus in the present paper on the nature of the multiple-@xmath3 ordered state and of the associated fluctuations in the multiple-@xmath3 ordered phase . in studying the nature of the multiple-@xmath3 ordered state , we begin by revisiting a mean - field analysis done earlier by reimers _ constructed a landau - type free energy @xmath22 of the pyrochlore heisenberg antiferromagnet within a mean - field approximation up to the quartic order , @xmath23 where @xmath24 is the order parameter corresponding to the fourier magnetization of sublattice @xmath25 ( @xmath26 ) given by @xmath27 where @xmath28 is the spin at the site @xmath29 belonging to sublattice @xmath25 , @xmath30 is the position vector of that site and @xmath31 is the number of spins belonging to sublattice @xmath25 . the sum @xmath32 runs over all @xmath33s which satisfy @xmath34 , and @xmath35 is the fourier transform of the exchange interaction between sublattices @xmath25 and @xmath36 . the quadratic term can be diagonalized by a unitary matrix @xmath37 with the eigenvalue @xmath38 , @xmath39 where @xmath29 indicates each eigenmode . by transforming the order parameter to normal modes @xmath40 , @xmath41 the landau free energy @xmath22 is reduced to @xmath42 from the quadratic term of the free - energy expansion , one sees that the normal mode corresponding to the maximum eigenvalue @xmath43 becomes unstable at @xmath44 where @xmath45 is the critical wavevector , leading to a phase transition to the ordered state characterized by the wavevector @xmath45 . when the maximum eigenvalue is degenerate as in the present case , the ordered - state spin configuration still remains largely undetermined at the quadratic level . in such a case , there are infinitely many ways of mixing @xmath46s with keeping the order parameter @xmath47 constant . in order to specify the explicit ordered - state spin configuration , one needs to go to the quartic term . _ made such an analysis only for the special case of a simple @xmath48 ordered state , whereas such an analysis has not been made for more general cases of an incommensurate ordered state , which is the target of our following analysis . now , we wish to go beyond the analysis of reimers _ et al . _ to derive the explicit ordered - state spin configuration of the @xmath0-@xmath1 model within a mean - field approximation . for the case of @xmath11 and @xmath6 of our interest , the maximum eigenvalue of @xmath49 appears in a symmetric direction @xmath50 and its cubic - symmetry counterparts . along this direction , @xmath49 has a form @xmath51 with @xmath52 the eigenvalues of this matrix are calculated as @xmath53 ^ 2},\end{aligned}\ ] ] with the corresponding eigenvectors given by @xmath54 and @xmath55 with @xmath56}. \label{alpha}\ ] ] the maximum engenvalue lies in the @xmath57 branch at a wavevector @xmath58 . note that , since @xmath59 generally in the corresponding eigenvector @xmath60 , a single-@xmath3 state associated with such an eigenvector can not satisfy the fixed spin - length condition @xmath20 for all sublattices . hence , even a pure single-@xmath3 state can not be realized at @xmath61 as in multiple-@xmath3 states , though it might be realized at finite temperatures due to thermal fluctuations . due to the cubic symmetry of the pyrochlore lattice , @xmath62 is twelvefold degenerate . hence , a variety of multiple-@xmath3 states might be possible in principle . even if one is to identify two eigenmodes corresponding to @xmath63 and @xmath64 , there are still six independent eigenmodes . within the landau - type free - energy expansion , the free - energy difference among possible multiple-@xmath3 states arises from the quartic term , @xmath65 where the sum over @xmath66 is taken over eigenmodes ( @xmath67 ) . just below the transition temperature , one might neglect in the summation @xmath68 the contributions from all modes other than the twelve critical modes . let us represent independent critical wavevectors as @xmath69 when one gives the number of critical modes to be mixed , there remain degrees of freedom associated with the amplitudes and the directions of the critical modes @xmath70 , which should be determined so as to minimize @xmath71 . in case of the single-@xmath3 state , @xmath71 takes a minimum for a simple spiral state @xmath72 where @xmath73 with @xmath74 and @xmath75 , while the minimized @xmath71 is given by @xmath76 where @xmath77 . in addition to the simple single-@xmath3 state , various types of multiple-@xmath3 states are possible depending on the number of mixed critical modes . the multiple-@xmath3 states with odd number of critical modes , _ i.e. _ , the states with three or five mixed eigenmodes , turn out to have higher @xmath71 , and we consider here only the multiple-@xmath3 states with even number of critical modes , _ i.e. _ , the states with two , four and six mixed eigenmodes , which are described as the double-@xmath3 , the quadruple-@xmath3 , and the sextuple-@xmath3 states , respectively . for a given even number of mixed critical modes , one needs to minimize @xmath71 for a fixed order parameter @xmath78 with respect to the amplitudes and the directions of the critical modes @xmath70 . some of the details of this minimization procedure are given in appendix . in case of the double-@xmath3 state , the optimized spin configuration turns out to be a superposition of two _ distorted _ spirals of @xmath79 which have mutually orthogonal spiral planes as illustrated in fig.[fig_multiple ] ( a ) ( the d1 state in appendix ) . the minimized @xmath71 is calculated to be @xmath80m^4.\ ] ] in case of the quadruple-@xmath3 state , the optimized spin configuration turns out to be a superposition of two spirals @xmath79 sharing a common spiral plane and two linearly - polarized spin - density waves ( sdw ) , @xmath81 , with its spin polarization perpendicular to the spiral plane as illustrated in fig.[fig_multiple ] ( b ) . the minimized @xmath71 is calculated to be @xmath82m^4.\end{gathered}\ ] ] finally , in case of the sextuple-@xmath3 state , the optimized spin configuration turns out to be a superposition of six linearly - polarized sdws where each pair of @xmath83 shares a common axis @xmath84 , and axes of different @xmath33s are mutually orthogonal @xmath85 as illustrated in fig.[fig_multiple ] ( c ) . note that this sextuple-@xmath3 state retains a cubic symmetry of the pyrochlore lattice , in sharp contrast to the other multiple-@xmath3 states or the single-@xmath3 state which break the cubic lattice symmetry . the minimized @xmath71 is calculated to be @xmath86 in fig.[fig_free_ene ] , we show @xmath87 of various states discussed above , including the single-@xmath3 state , the double-@xmath3 state , the quadruple-@xmath3 state and the sextuple-@xmath3 state , as a function of the parameter @xmath88 . in fact , @xmath88 depends on @xmath2 only weakly . for the value of @xmath89 , @xmath88 takes values around @xmath90 : see the inset of fig.[fig_free_ene ] . around this value , the free energy of the single-@xmath3 state turns out to be much higher than those of the multiple-@xmath3 states . this result is consistent with the observation of the multiple-@xmath3 states in previous mc simulations @xcite . within the present mean - field approximation , the double-@xmath3 state becomes most stable for @xmath90 . the free energy difference among various multiple-@xmath3 states , however , turns out to be rather small , and fluctuations neglected here might eventually select the multiple-@xmath3 state other than the double-@xmath3 state . in order to determine the true ordered state of the model , we should carefully examine the effect of fluctuations . for this purpose , we perform extensive mc simulations of the model in the next section . of the single-@xmath3 state and of various types of multiple-@xmath3 states , including the double-@xmath3 state ( the @xmath91 state in appendix ) , the quadruple-@xmath3 state ( the @xmath92 state in appendix ) and the sextuple-@xmath3 state , are plotted versus the parameter @xmath93 : see eq .. inset exhibits a relation between @xmath88 and @xmath94 . for @xmath95 the double-@xmath3 state and the quadruple-@xmath3 state become locally unstable . such unstable regions are indicated by thin dotted lines . for further details of the stability , see appendix.,width=302 ] in this section , we present the results of our mc simulation on the @xmath0-@xmath1 pyrochlore - lattice classical heisenberg model with @xmath96 and @xmath97 . as a typical example , we deal with the case of @xmath98 . in the phase diagram reported by chern _ et al._@xcite , this @xmath94 ratio is sufficiently large so that the partially ordered collinear phase should not appear . a direct transition from the paramagnetic phase to the multiple-@xmath3 ordered phase is then expected . the pyrochlore lattice contains 16 spins in its cubic unit cell . the system size we deal with is of linear size @xmath99 in units of cubic unit cell , _ i.e. _ the system contains @xmath100 spins in total . since the ordered state is generally incommensurate with the underlying lattice , the system under periodic boundary conditions ( bc ) would be subject to severe finite size effects . in order to examine such finite - size effects , we also consider free bc , and carefully compare the results between these two bc . in the case of free bc , we extend the lattice with a half of the unit cell in all three directions so that it keeps the cubic symmetry of the lattice . then , the system contains @xmath101 spins in total . mc simulations are performed based on the standard heat - bath method combined with the over - relaxation method . the lattice size is of @xmath102 for periodic bc , and of @xmath103 for free bc . the system is gradually cooled from the high temperature . a run at each temperature contains typically @xmath104 mc steps per spin ( mcs ) and about a half of mcs are discarded for equilibration . our 1 mcs consists of 1 heat - bath sweep and subsequent 10 over - relaxation sweeps . in fig.[fig_ene ] , we show the temperature and size dependence of the energy per spin . for both cases of periodic and free bc , an almost discontinuous change of the energy , a characteristics of a first - order transition , is observed . such a first - order nature of the transition is consistent with the results of previous calculations @xcite . we estimate the bulk transition temperature to be @xmath105 . .,width=264 ] at temperatures below @xmath106 , several types of metastable states are observed in our simulations , depending on the spin initial conditions and the random - number sequences . in order to probe the spin configurations in these metastable states , we calculate the spin structure factor @xmath107 defined by @xmath108 where @xmath109 represents a thermal average . from the spin structure factor , we find that none of metastable states is a single-@xmath3 state . all of them are multiple-@xmath3 states where the multiple magnetic bragg peaks coexist in the spin structure factor . in fig.[fig_sq ] we show the typical spin structure factors of these metastable states calculated at a temperature @xmath110 under periodic bc . these metastable states might be classified into two types according as whether they keep the cubic symmetry or not . in the cubic - symmetric state , there are six independent main bragg peaks with the same intensity at @xmath111 and at its cubic - symmetry counterparts with @xmath112 ( see fig.[fig_sq ] ( a ) ) . the peak position @xmath113 is close to the corresponding mean - field value @xmath114 . there are also other bragg peaks related to the main peaks via the reciprocal lattice vectors of the fcc lattice . since the observed cubic - symmetric state involves six critical wavevectors , we identify this state as a sextuple-@xmath3 state . in case of periodic bc , we sometimes observe `` almost cubic symmetric '' states where the position of the bragg peak shifts by @xmath115 , or one of the bragg peaks splits into two , due to the finite size effects associated with periodic bc . we regard them as modified forms of the sextuple-@xmath3 state . in the observed non - cubic state , two out of six main bragg peaks with wavevectors on a common plane , _ e.g. _ , @xmath116 and @xmath117 , vanish , while other four remain . since there are four critical wavevectors in this non - cubic state , we identify the state as a quadruple-@xmath3 state . two different types of the non - cubic states are observed , depending on the intensity ratio among the remaining four main bragg peaks . in one type ( type 1 ) , two out of four remaining bragg peaks with wavevectors on a common plane , _ e.g. _ , @xmath118 and @xmath119 , have stronger intensity than the other two with wavevectors lying in the other plane , _ e.g. _ , @xmath120 and @xmath121 : see fig.[fig_sq ] ( b ) . in the other type ( type 2 ) , four main bragg peaks have equal intensities ( see fig.[fig_sq ] ( c ) ) . in this type 2 structure , all bragg peaks including main and submain are actually split into two , one at @xmath122 and the other at @xmath123 . defined by eq . , calculated at @xmath124 for a @xmath125 lattice with free boundary conditions . the interaction parameter is @xmath98 . initially , a half side of the system is prepared to be a sextuple-@xmath3 state , while the other half a quadruple-@xmath3 state . the two curves represent the subsequent time evolution of the cubic - symmetry - breaking order parameter @xmath126 , each calculated at each half side of the system . the data indicates that the cubic state expands as the simulation goes on.,width=302 ] state and the quadruple-@xmath3 state plotted versus the temperature for both cases of periodic and free boundary conditions . the interaction parameter is @xmath98 . in case of periodic boundary conditions , we choose a pair of metastable states which have the lowest energy among several `` almost cubic - symmetric '' states or `` almost quadruple-@xmath3 ( type 1 ) '' states where peak shift or peak splitting occur due to finite size effects.,width=302 ] in fig.[fig_sq_free ] , we show the typical spin structure factors calculated at the same temperature @xmath127 under free bc . similarly to the case of periodic bc , we find a cubic - symmetric state ( fig.[fig_sq_free ] ( a ) ) and a non - cubic state of type 1 ( fig.[fig_sq_free ] ( b ) ) . by contrast , the non - cubic state of type 2 is never observed in contrast to the case of periodic bc . ( in the non - cubic state observed under free bc , the main and sub peaks appear along the direction of @xmath128 slightly off the high symmetric direction @xmath129 reflecting the non - cubic character of the ordered states . however , the magnitude of the shift @xmath130 is always of order of @xmath131 , apparently vanishing in the thermodynamic limit . ) we deduce that the type-1 state rather than the type-2 state is a stable quadruple-@xmath3 state due to the following two reasons . first , the type-1 quadruple-@xmath3 state appear as a metastable state in both periodic and free bc , while the type-2 quadruple-@xmath3 state is never realized in free bc . second , within our mean - field analysis , the type-1-like state with the spin structure factor of unequal peak heights , which corresponds to the @xmath92 state in appendix , has a lower free energy than the type-2-like state with the spin structure factor of equal peak heights , which corresponds to the @xmath132 state in appendix . in contrast to the mean - field result for @xmath133 , we do not observe in our mc simulation the double-@xmath3 metastable state below @xmath106 . since the double-@xmath3 state can continuously be changed into the quadruple-@xmath3 state , the double-@xmath3 state is probably unstable toward the quadruple-@xmath3 state in this parameter region . next , we wish to determine either the cubic sextuple-@xmath3 state or the non - cubic quadruple-@xmath3 state is stable below @xmath106 . since both states are metastable states surviving for a very long time in the course of mc simulation , it is extremely difficult to identify which state is thermodynamically stable by performing fully thermalized simulation below @xmath106 . in order to identify the truly stable state below @xmath106 , we then employ the mixed - phase method @xcite . in this method , one prepares as an initial state a two - phase coexisting state , where the sextuple-@xmath3 state or the quadruple-@xmath3 state occupies each half of the lattice . by thermalizing such a state and by monitoring which ordered state expands during the course of the subsequent simulation , one can determine which phase is more stable . in fig.[fig_mix ] , we show an example of the mc time evolution of such a two - phase coexisting initial state in our mc simulations . we look at the time evolution of the cubic - symmetry - breaking order parameter @xmath126 for each half of the lattice , defined by @xmath134 where @xmath135 and @xmath136 represent the maximum and the minimum values of the sum of the inner product @xmath137 among three nearest - neighbor bonds , denoted by @xmath130 , emanating from a given site @xmath29 . this quantity measures the extent of the cubic - symmetry breaking in the spin configuration : it tends to be large for a non - cubic quadruple-@xmath3 state , while it tends to be small for a cubic sextuple-@xmath3 state . as can be seen from the mc time dependence of @xmath126 shown from fig.[fig_mix ] , at the measuring temperature @xmath124 , the sextuple-@xmath3 state expands its area , and the system finally settles in the sextuple-@xmath3 state with smaller @xmath126 . by performing such a mixed - phase method for various system size up to @xmath138 for periodic bc , and up to @xmath125 for free bc , we find that the sextuple-@xmath3 state is stable at least in the temperature range @xmath139 . unfortunately , the dynamics of the system becomes so slow with further lowering the temperature that the domain wall between the two ordered states hardly moves , which hampers to determine the stable state at lower temperatures @xmath140 . in the low temperature limit , the entropy contribution to the free energy becomes smaller , while the energy contribution becomes dominant . thus , the state which has the lower energy should be realized . in fig.[fig_ene_diff ] , we show the energy difference between the sextuple-@xmath3 state and the quadruple-@xmath3 state versus the temperature . for both cases of periodic and free bcs , the energy of the quadruple-@xmath3 state becomes lower than that of the sextuple-@xmath3 state at low enough temperatures . it means that the quadruple-@xmath3 state should be realized as a stable state at sufficiently low temperatures . by combining the results of the energy analysis with those of the mixed - phase method , we conclude the ordering behavior of the @xmath0-@xmath1 model with @xmath98 as follows . with decreasing the temperature , the system first exhibits a phase transition from the paramagnetic phase to the sextuple-@xmath3 ordered state at a temperature @xmath141 . the transition is of first order . note that the cubic symmetry of the lattice is still fully respected even below @xmath142 in spite of the existence of the magnetic long - range order characterized by sharp bragg peaks . with further decreasing the temperature , another phase transition occurs at @xmath143 into the quadruple-@xmath3 state . it accompanies the breaking of the cubic symmetry of the lattice . this transition is probably of weakly first order . now , we turn to the determination of the detailed spin configurations in the multiple-@xmath3 states identified above . we begin with the cubic - symmetric sextuple-@xmath3 state . guided by our mean - field result and by inspecting the spin configurations obtained by mc simulations , we observe that the ordered spin configuration in the sextuple-@xmath3 state can be represented by the superposition of six linearly - polarized sdws as @xmath144 \notag \\ s_y^{(\mu)}(\bm{r } ) & = i \bigl [ u^\mu_{\bm{q}_2^+}\cos(\bm{q}_2^+\cdot\bm{r}+\theta_2^+)\notag \\ & \qquad\qquad+u^\mu_{\bm{q}_2 ^ - } \cos(\bm{q}_2 ^ -\cdot\bm{r}+\theta_2 ^ -)\bigr ] \notag \\ s_z^{(\mu)}(\bm{r } ) & = i \bigl [ u^\mu_{\bm{q}_3^+}\cos(\bm{q}_3^+\cdot\bm{r}+\theta_3^+)\notag\\ & \qquad\qquad+u^\mu_{\bm{q}_3 ^ - } \cos(\bm{q}_3 ^ -\cdot\bm{r}+\theta_3 ^ -)\bigr ] , \label{model_sextuple}\end{aligned}\ ] ] where @xmath145 represent the spin at the position @xmath146 belonging to sublattice @xmath25 . any spin configuration generated from the one given by via global spin rotation in spin space is equally allowed . the coefficient @xmath147 takes a value unity or @xmath148 , depending on whether the site on the sublattice @xmath25 has a nn bond in the direction of the wavevector @xmath149 ( @xmath150 ) or not ( @xmath151 ) . indeed , the spin configuration given by eq . is fully consistent with the mean - field result shown in sec . iii and appendix . in the the case of the quadruple-@xmath3 state , inspection of the mc data lead us to the ordered - state spin configuration represented by the superposition of two spirals and two linearly - polarized sdws as @xmath152 \notag \\ s_y^{(\mu)}(\bm{r } ) & = i_{xy } \bigl [ \tilde{u}^\mu_{\bm{q}_1^+}\sin(\bm{q}_1^+\cdot\bm{r}+\theta_1^+ ) \notag \\ & \qquad\qquad+\tilde{u}^\mu_{\bm{q}_1 ^ - } \sin(\bm{q}_1 ^ -\cdot\bm{r}+\theta_1 ^ -)\bigr ] \notag \\ s_z^{(\mu)}(\bm{r } ) & = i_z \bigl [ \tilde{u}^\mu_{\bm{q}_2^+}\cos(\bm{q}_2^+\cdot\bm{r}+\theta_2^+)\notag \\ & \qquad\qquad + \tilde{u}^\mu_{\bm{q}_2 ^ - } \cos(\bm{q}_2 ^ -\cdot\bm{r}+\theta_2 ^ -)\bigr ] , \label{model_quadruple}\end{aligned}\ ] ] where the two wavevectors with stronger intensity in the spin structure factor are associated with @xmath153 and those with weaker intensity with @xmath154 . any spin configuration generated from the one given by via global spin rotation in spin space is equally allowed . the same rule given above for the sextuple-@xmath3 state is understood also for the coefficient @xmath155 . this spin configuration is also fully consistent with the mean - field result of sec . iii and appendix . in order to investigate the nature of fluctuation in the ordered state , we compute the time - dependent spin overlap @xmath156 defined by @xmath157 ^ 2 , \label{eq_q2_freeze}\ ] ] which measures an overlap between the spin configurations at time @xmath158 and at time @xmath159 . note that @xmath156 is defined so as to be invariant under any global spin rotation in spin space . we also employ this quantity to check the validity of the proposed spin structures given by eqs . and . for this purpose , we compute @xmath156 in the following two ways : the two spin configurations at time @xmath158 and at subsequent time @xmath159 are taken , either ( i ) from the raw spin configuration data of our equilibrium mc simulations , or ( ii ) from the the proposed model spin structure described by eq . or which is evaluated by fourier transforming the raw mc data and by extracting the amplitudes ( @xmath160 , @xmath161 and @xmath162 ) and the phases ( @xmath163 ) associated with the critical mode at @xmath164 . in the procedure ( ii ) , the @xmath88-factor associated with @xmath165 or @xmath166 is also taken as a fitting parameter , not a given constant , while the contribution of other non - critical modes are simply neglected . if the proposed model structures properly represent the ordered - state spin configuration , @xmath156s calculated in the above two ways ( i ) and ( ii ) should agree . in fig.[fig_freeze ] , we show the mc - time dependence of @xmath156 calculated in the two ways ( i ) and ( ii ) for the sextuple-@xmath3 state ( a ) , and for the quadruple-@xmath3 state ( b ) . indeed , in both cases of the sextuple-@xmath3 and the quadruple-@xmath3 states , @xmath156 calculated in the two ways agree with high precision , indicating that our model spin structures and properly represent the actual ordered - state spin configurations . second , the computed @xmath156 exhibit a significant time variation indicating that the spin configuration changes considerably in the course of our mc simulation . indeed , as can be seen from fig.[fig_int_phase ] , this time dependence comes from the time dependence of the phase factor @xmath167 , whereas the amplitude turns out to be nearly time - independent . establishing the spin configuration of the ordered state , we wish to further investigate the nature of spin fluctuations in the ordered state . two kinds of spin fluctuations associated with the identified spin structures are observed . the first one is a fluctuation of the phase factors @xmath167 as mentioned above . while the amplitudes @xmath160 , @xmath161 , @xmath162 , which contribute to the bragg intensity , turn out to be nearly constant in time ( see fig.[fig_int_phase ] ( a ) and ( c ) ) , the phase factors @xmath167 fluctuate a lot even when the system exhibits sharp bragg peaks ( see fig.[fig_int_phase ] ( b ) and ( d ) ) . the second type of spin fluctuation is a necessary outcome of the distribution of ordered spin moments in the multiple-@xmath3 state , where some spins should reduce their frozen moments exhibiting large fluctuations . first , we examine the effect of phase fluctuations . as we have seen in the dynamics of @xmath168 shown in fig.[fig_int_phase ] ( b ) and ( d ) , phase fluctuations do occur in finite - size systems . the question to be addressed is whether such phase fluctuations disappear or not in the thermodynamic limit . for this purpose , we define the phase auto - correlation function by @xmath169 where @xmath170 represents @xmath168 . in fig.[fig_phase_auto ] ( a ) , we show the phase autocorrelation function of @xmath171 in the quadruple-@xmath3 state . as can be seen from the figure , the relaxation of the phase becomes slower with increasing the system size , suggesting that the phase motion might eventually be locked in the thermodynamic limit . the computed phase relaxation follows a simple exponential decay characterized by the phase correlation time @xmath172 . in fig.[fig_phase_auto ] ( b ) , we plot @xmath173 as a function of the inverse system size @xmath174 . as can be seen from the figure , the phase relaxation time diverges almost linearly with @xmath175 , and we conclude that the phase motion will eventually be locked in the thermodynamic limit . hence , spin fluctuations arising from phase fluctuations are finite - size effects and are expected to vanish in the thermodynamic limit . autocorrelation function in the quadruple-@xmath3 state at @xmath176 . the lattices are @xmath177 and @xmath178 with periodic boundary conditions . ( b ) the phase autocorrelation time obtained as in ( a ) is plotted versus the inverse system size @xmath174 . dotted line exhibits a liner fit of the data.,width=302 ] next , we discuss the second source of spin fluctuations . even when phase fluctuations vanish in the thermodynamic limit , spin fluctuations associated with the multiple-@xmath3 nature of the ordering should still remain . remember that the multiple-@xmath3 order entails the the existence of certain fraction of spins with reduced frozen moments , which should be caused by significant local spin fluctuations . such spin fluctuations intrinsic to the multiple-@xmath3 order might be observable experimentally via local probes like nmr . in fig.[fig_nmr ] , we exhibit the distribution of internal fields in the sextuple-@xmath3 ( eqs . ) and the quadruple-@xmath3 ( eq . ) states probed at an @xmath179 site of a pyrochlore oxide @xmath180 with the oxygen - location @xmath181 parameter @xmath182 @xcite . we assume here that the @xmath183 site is magnetic , and internal fields are borne by the dipolar interaction . as can be seen from the figure , both the sextuple-@xmath3 and the quadruple-@xmath3 states exhibit broad internal - field distributions , in contrast to the one associated with a simple all - in / all - out structure shown in the inset . the latter exhibits a sharp delta - function - like distribution . the observed broad distributions of internal fields mean that the magnitude of frozen spin moments are spatially distributed , as expected from the multiple-@xmath3 nature of the ordering . site of a pyrochlore oxide @xmath180 with the oxygen - location @xmath181 parameter @xmath182 where @xmath184 site is magnetic , calculated for an ideal sextuple-@xmath3 state and quadruple-@xmath3 state , respectively . dipolar interactions are assumed as an origin of internal fields . inset exhibits an internal field distribution corresponding to a commensurate all - in / all - out structure on the pyrochlore lattice.,width=302 ] we studied the nature of the multiple-@xmath3 ordered states realized in the @xmath0-@xmath1 pyrochlore - lattice heisenberg model by means of a mean - field analysis and a monte carlo simulation . performing a mean - field analysis beyond the previous analysis by reimers _ et al._@xcite , we could explicitly determine the possible multiple-@xmath3 spin structures of the model , which include a cubic - symmetric sextuple-@xmath3 state , a non - cubic quadruple-@xmath3 state and a non - cubic double-@xmath3 state . we have found that these multiple-@xmath3 states have considerably lower free energy than that of the single-@xmath3 spiral , while the free energy difference between these different multiple-@xmath3 states are rather small . with reference to the mean - field results , we also performed an extensive mc simulations of the model , mainly for the case of @xmath98 , to determine which state is really stabilized as an ordered phase . as expected from the mean - field analysis , we found that the system exhibited a phase transition from the paramagnetic phase to the multiple-@xmath3 ordered state characterized by the multiple peaks in the spin structure factor . recent studies have revealed that such multiple-@xmath3 states are also stabilized in other frustrated heisenberg magnets , _ e.g. _ the triangular lattice heisenberg antiferromagnet with the next - nearest or the third neighbor interactions under magnetic fields @xcite . with the help of the mixed - phase method , we found that the cubic - symmetric sextuple-@xmath3 state is stabilized just blow the transition temperature @xmath141 down to at least @xmath185 , while , at sufficiently low temperatures , the non - cubic quadruple-@xmath3 state becomes stable . hence , another phase transition from the sextuple-@xmath3 state to the quadruple-@xmath3 state should occur at a temperature @xmath186 . the nature of each transition was also examined . the transition at @xmath187 is first order , as was already indicated by previous studies . note that , at this transition the cubic symmetry of the lattice is still fully preserved . it is remarkable that , in spite of the incommensurate and rather complex nature of the spin order , the cubic symmetry of the lattice is fully respected in the ordered state . the second transition at @xmath188 is weakly first order , and it accompanies a spontaneous breaking of the cubic symmetry of the lattice . it should be noticed that the phase transition between the sextuple-@xmath3 and the quadruple-@xmath3 states analyzed in the present paper is distinct from the transition between the paramagnetic state and the nematic state ( or the one between the nematic state and the multiple-@xmath3 state ) as discussed by chern _ et al._@xcite . we also performed mc simulations for several smaller values of @xmath189 where chern _ et al . _ observed the nematic phase . in such smaller-@xmath94 region , we found as metastable states both the sextuple-@xmath3 and the quadruple-@xmath3 states in the parameter region where chern _ et al . _ reported the multiple-@xmath3 state ( chern _ et al . _ did not specify the type of the multiple-@xmath3 state ) . thus , there remains a possibility that , even for smaller @xmath94-values where the nematic phase appears at higher temperature region , another phase transition from the sextuple-@xmath3 state to the quadruple-@xmath3 state occurs at a lower temperature . we also determined the explicit spin configuration of the multiple-@xmath3 ordered state . the sextuple-@xmath3 state is a superposition of six sdws running along the wavevector @xmath190 and its cubic - symmetry counterparts . in this sextuple-@xmath3 state , the system fully retains a cubic symmetry of the lattice . by contrast , the quadruple-@xmath3 state is a superposition of two helices and two sdws . this state spontaneously breaks the cubic symmetry of the lattice as one can easily see from the fact that two out of six critical wavevectors should vanish in this state . both the sextuple-@xmath3 and the quadruple-@xmath3 spin configurations are consistent with the ones obtained in our mean - field analysis . in both cases of the sextuple-@xmath3 and the quadruple-@xmath3 states , a broad distribution of internal fields was observed . such broad distributions reflect the distribution of frozen spin moments which arises from the multiple-@xmath3 nature of the ordering . hence , if the pyrochlore magnets are in the multiple-@xmath3 ordered phase we proposed , one should observe two apparently conflicting features , _ i.e. _ , the coexistence of sharp bragg peaks measured by neutron diffraction and of enhanced spin fluctuations measured by local probes such as nmr . experimental observation of the multiple-@xmath3 states as revealed here remains most interesting . the authors are thankful to z. hiroi , s. maegawa and t. arima for useful discussion . this work is supported by grand - in - aid for scientific research on priority areas `` novel state of matter induced by frustration '' ( 19052006 ) . we thank the supercomputer center , institute for solid state physics , university of tokyo , and the cyber media center , osaka university for providing us with the cpu time . in this appendix , we present the details of our mean - field analysis . we consider here twelve critical eigenvectors only , corresponding to six wavevectors of . as mentioned in section iii , the free - energy difference between different types of multiple-@xmath3 states arises at the quartic term of the free energy . thus , we calculate here the quartic term of the free energy @xmath71 for several multiple-@xmath3 structures . for simplicity , we abbreviate @xmath191 as @xmath192 , where @xmath193 and @xmath194 . the quartic term @xmath71 consists of following four terms , @xmath195 , @xmath196 , @xmath197 as and @xmath198 as @xmath199 , \label{eq_f4_app}\end{gathered}\ ] ] where @xmath200 may be regarded to represent the interaction among @xmath29 distinct wavevectors , and are given by @xmath201,\ ] ] @xmath202,\ ] ] @xmath203,\ ] ] and @xmath204 , \label{a4_eq}\end{gathered}\ ] ] where @xmath205 and @xmath206 $ ] means a real part , while @xmath207 means the summation over @xmath208 . the @xmath198 term represents interactions of four wavevectors such as @xmath209 , @xmath210 , @xmath211,@xmath212 , which satisfy @xmath213 . in the case of the single-@xmath3 state , only the @xmath195 term contributes to @xmath71 . if we fix the amplitude @xmath214 associated with the wavevector @xmath215 , @xmath71 takes a minimum when @xmath192 satisfy @xmath216 this condition means that @xmath192 is given by @xmath217 where @xmath218 and @xmath219 are orthogonal unit vectors in spin space . in the real space , this @xmath192 means a spiral , @xmath220 by substituting this condition into eq . , we get the minimized free energy for the single-@xmath3 state as @xmath221 in mixing two wavevectors , there are two possible ways : one is to mix a pair of @xmath222 and @xmath223 , and the other is to mix @xmath79 and @xmath224 with @xmath225 . for @xmath226 , the former always minimizes the free energy , since the coefficients of @xmath196 is smaller than that of @xmath227 . hence , we consider below a mixture of @xmath222 and @xmath223 . in this situation , the @xmath227-term vanishes identically . the @xmath196-term is minimized for @xmath228 if we assume that @xmath229 forms a spiral given by eq . , these conditions are satisfied when @xmath230 is perpendicular to the spiral plane formed by @xmath229 , _ i.e. _ , @xmath231 where @xmath232 is a unit vector perpendicular to the spiral plane formed by @xmath229,_i.e . _ , @xmath233 . in this situation , the spin configuration corresponding to @xmath230 forms a spin - density wave ( sdw ) @xmath234 hence , the @xmath196-term favors a mixture of a spiral and a sdw . as we have seen for the case of the single-@xmath3 state , @xmath195 is minimized when @xmath192 forms a spiral . the competition between the @xmath195- and the @xmath196-terms often causes a distortion of a spiral structure , _ i.e. _ , @xmath235 where @xmath236 and @xmath237 are orthogonal real vectors @xmath238 while the amplitudes @xmath239 and @xmath240 are generally different from each other . note that such a distorted ( or elliptical ) spiral can continuously be transformed to an isotropic spiral ( @xmath241 ) and a sdw ( @xmath242 or @xmath243 ) . thus , we assume both @xmath229 and @xmath230 form distorted spirals eq . , and minimize @xmath71 with respect to the directions of spiral planes , the phase factors @xmath244 , and the amplitudes @xmath245 and @xmath246 . under this situation , the @xmath196-term is minimized when the two spiral planes are mutually orthogonal , _ i.e. _ , @xmath247 then , @xmath229 and @xmath230 are given by @xmath248 where @xmath249 represents the chirality of each spiral . in this structure , the two spirals share one of the basis vectors @xmath218 while the other two basis vectors @xmath219 and @xmath232 are mutually orthogonal . note that this @xmath71 is independent of the phase factor @xmath244 and the chirality @xmath251 . by minimizing @xmath71 with keeping @xmath252 fixed , we find that there are two types of double-@xmath3 structure , _ i.e. _ , a mixture of two spirals ( the @xmath91 state ) and a mixture of a spiral and a sdw ( the @xmath253 state ) . in the case of a mixture of two spirals ( the @xmath91 state ) , @xmath71 takes an extremum @xmath254m^4 , \label{eq_fd1}\ ] ] for @xmath255 in the @xmath91 state , two spirals are distorted ( @xmath256 , @xmath257 ) to lower the @xmath196-term . for @xmath258 ( @xmath259 ) , this @xmath91 state becomes a minimum of @xmath71 , while for @xmath260 it becomes locally unstable . in the case of a mixture of a spiral with @xmath261 and a sdw with @xmath262 ( the @xmath253 state ) , @xmath71 takes an extremum when the spiral is isotropic and the sdw is orthogonal to the spiral as @xmath263 where @xmath71 is calculated to be @xmath264m^4 . \label{f4_d2}\end{gathered}\ ] ] for @xmath265 ( @xmath259 ) , this @xmath253 state becomes a minimum of @xmath71 , while for @xmath266 it becomes locally unstable . in fig.[fig_freeenergy_ap ] , we plot @xmath87 of the single-@xmath3 spiral state ( the @xmath267 state ) and of the two kinds of double-@xmath3 states ( the @xmath91 and @xmath253 states ) as a function of @xmath88 . note that the critical value @xmath268 ( @xmath269 ) of the stability is in common between the @xmath91 state and the @xmath253 state . thus , the mixture of two spirals ( the @xmath91 state ) is stable for @xmath270 , while the mixture of a spiral and a sdw ( the @xmath253 state ) is stable for @xmath271 . since @xmath88 is in the range of @xmath90 for the realistic parameter value of @xmath2 , the mean - field approximation favors the @xmath91 state over the @xmath253 state in the @xmath0-@xmath1 model . of the single-@xmath3 spiral state ( the @xmath272 state ) , of the two kinds of double-@xmath3 states ( the @xmath91 and @xmath253 states ) and of the two kinds of quadruple-@xmath3 states ( the @xmath132 and @xmath92 states ) are plotted versus the parameter @xmath88 . for @xmath95 , the @xmath91 state and @xmath92 state become unstable , while for @xmath273 the @xmath253 and the @xmath132 states become unstable . the unstable regions are indicated by the thin dotted lines . , width=264 ] when four wavevectors coexist , the @xmath198-term contributes to the free energy . since the increase of the number of mixed wavevectors tends to enhance the contribution of the @xmath196- and the @xmath227-terms , it is necessary to lower the @xmath198-term to stabilize the quadruple-@xmath3 state . as an example , we consider here a mixture of @xmath153 and @xmath274 . since spiral structures tend to lower the @xmath195-term while sdw structures tend to lower the @xmath196- and the @xmath227-terms , we again assume that four @xmath192s form distorted spirals described by eq . , and minimize @xmath71 with respect to the directions of spiral planes , the phase factors @xmath244 , and the amplitudes @xmath245 and @xmath246 . first , we optimize the spiral plane . in order to lower the @xmath196- and the @xmath227-terms , the ideal situation would be that four spiral planes are mutually orthogonal . however , this is simply impossible for the heisenberg spin . since the coefficient of @xmath196 is smaller than that of @xmath227 , next way might be that a pair of @xmath153 ( or @xmath274 ) shares a common spiral axis which is orthogonal to the others like , @xmath275 remaining spiral planes of @xmath274 prefer to be orthogonal to each other if we consider only the @xmath196-term . however , if we consider also the @xmath198-term , the sum of the @xmath196- and the @xmath198-terms takes a minimum when the pair of @xmath274 also shares a common spiral plane orthogonal to that of @xmath153 like @xmath276 as long as they are not reduced to the double-@xmath3 or the single-@xmath3 states . thus , four @xmath192s are given by @xmath277 where @xmath249 represents the chirality of the spiral . next , we optimize the phase factor , @xmath244 , and the chirality , @xmath251 . although the @xmath195- , the @xmath196- , and the @xmath197-terms are independent of the phase factor and the chirality , the @xmath198-term depends on them . by substituting eq . into eq . , the @xmath198-term is calculated as @xmath278 . \end{gathered}\ ] ] then , the @xmath198-term is minimized for @xmath279 where @xmath198 is given by @xmath280 . \end{gathered}\ ] ] by substituting eqs . and into eq . , the optimized @xmath71 is given by @xmath281 + 32\alpha^2 \sum_{i=1,3}\left[\phi_{i+}^2\phi_{i-}^2 + 2\left(r_{i+}^2r_{i-}^2 + i_{i+}^2i_{i-}^2\right)\right]\\ + 8 ( 1 + 2\alpha^2+\alpha^4 ) \sum_{\sigma=\pm}\sum_{\sigma'=\pm}\left[\phi_{1\sigma}^2\phi_{3\sigma'}^2 + 2r_{1\sigma}^2i_{3\sigma'}^2\right]\\ -64\alpha^2 \bigl[\left(r_{1+}r_{1-}+i_{1+}i_{1-}\right)\left(r_{3+}r_{3-}+i_{3+}i_{3-}\right ) + 2r_{1+}r_{1-}i_{3+}i_{3-}\bigr ] \biggr\}.\end{gathered}\ ] ] by minimizing @xmath71 for a fixed @xmath282 , we find that there are two types of quadruple-@xmath3 structure , _ i.e. _ , a mixture of four distorted spirals ( the @xmath132 state ) and a mixture of two spirals and two sdws ( the @xmath92 state ) . in the case of a mixture of four spirals ( the @xmath132 state ) , @xmath71 takes an extremum , @xmath283m^4,\\ \label{eq_fq1}\end{gathered}\ ] ] with @xmath284 in this @xmath132 state , similarly to the @xmath91 state , the four spirals are distorted ( @xmath285 , @xmath286 ) to lower the @xmath197-term . for @xmath287 ( @xmath259 ) , this @xmath132 state becomes a minimum of @xmath71 , while for @xmath266 it becomes locally unstable . next , we consider a mixture of two spirals and two sdws ( the @xmath92 state ) . in the @xmath92 state , a set of @xmath229 and @xmath230 ( @xmath288 or @xmath289 ) form spirals , while the other set of @xmath290 and @xmath291 ( @xmath292 ) form sdws perpendicular the spiral plane . as an example , we choose here a mixture of two spirals characterized by @xmath293 and two sdws characterized by @xmath294 . in this situation , @xmath71 takes an extremum when two spirals are isotropic , @xmath295 and two sdws are perpendicular to the spiral plane , @xmath296 in addition , their amplitudes satisfy @xmath297 the optimized @xmath71 is then calculated to be @xmath298m^4 . \label{eq_fq2}\end{gathered}\ ] ] for @xmath299 ( @xmath259 ) , this @xmath92 state becomes a minimum of @xmath71 , while for @xmath260 it becomes locally unstable . thus , for the quadruple-@xmath3 state , we find two metastable structures , @xmath132 and @xmath92 . in fig.[fig_freeenergy_ap ] , we plot @xmath87 of these two metastable quadruple-@xmath3 structures , given by and , respectively , as a function of @xmath88 . note that the critical value @xmath300 ( @xmath269 ) of the stability is common between the @xmath132 state and the @xmath92 state . since @xmath88 is in the range of @xmath301 for the realistic parameter value of @xmath2 , the structure characterized by a mixture of two spirals and two sdws ( the @xmath92 state ) is the stable one between the two metastable quadruple-@xmath3 structures . we finally discuss the case where all six wavevectors coexist , _ i.e. _ , the sextuple-@xmath3 state . in this situation , although almost all types of combinations of spirals and sdws turn out to be unstable , a mixture of six sdws can be locally stable . for a mixture of six sdws , @xmath71 is minimized when sdws of @xmath302 and @xmath303 run along a common axis while sdws of @xmath304 and @xmath305 ( @xmath225 ) are mutually orthogonal , _ i.e. _ , @xmath306 as in the case of the quadruple-@xmath3 state , the phase factors are optimized to minimize the @xmath198-term . by substituting eq . into eq . , the @xmath198-term is calculated as @xmath307 then , the @xmath198-term is minimized for @xmath308 where @xmath198 is given by @xmath309 from eqs . and , the optimized @xmath71 is given by if we fix the total amplitude @xmath311 , @xmath71 is minimized when the amplitudes of six sdws are in common , _ i.e. _ , @xmath312 in this situation , @xmath71 of the sextuple-@xmath3 state is reduced to @xmath86 99 edited by h.t . diep , ( world scientific publishing , singapore , 2004 ) . edited by c. lacroix , p. mendels , and f. mila , ( springer , berlin , 2011 ) . special topics on novel states of matter induced by frustration , edited by h. kawamura , j. phys . * 79 * , 011001 - 011012 ( 2010 ) . reimers , phys . b * 45 * , 7287 ( 1992 ) . c. l. henley , phys . b * 71 * , 014424 ( 2005 ) . isakov , k. gregor , r. moessner , and s.l . sondhi , phys . * 93 * , 167204 ( 2004 ) . conlon , and j.t . chalker , phys . * 102 * , 237206 ( 2009 ) . reimers , a.j . berlinsky , and a .- c . shi , phys . b * 43 * , 865 ( 1991 ) . d. tsuneishi , and h. kawamura , j. phys . : * 19 * , 145273 ( 2007 ) . t. nakamura , d. hirashima , j. magn . , * 310 * , 1297 ( 2007 ) . chern , r. moessner , and o. tchernyshyov , phys . b * 78 * , 144418 ( 2008 ) . conlon , and j.t . chalker , phys . b * 81 * , 224413 ( 2010 ) . l. bellier - castella , m.j.p . gingras , p.c.w . holdsworth , and r. moessner , can . j. phys . * 79 * , 1365 ( 2001 ) . saunders , and j.t . chalker , phys . 98 * , 157201 ( 2007 ) . a. andreanov , j.t . chalker , t.e . saunders , and d. sherrington , phys . b * 81 * , 014406 ( 2010 ) . tam , a.j . hitchcock , and m.j.p . gingras , arxiv:1009.1272 . h. shinaoka , y. tomita , and y. motome , phys . lett . * 107 * , 047204 ( 2011 ) . k. penc , n. shannon , and h. shiba , phys . lett . * 93 * , 197203 ( 2004 ) . bergman , r. shindou , g.a . fiete , and l. balents , phys . b * 74 * , 134409 ( 2006 ) . n. shannon , k. penc , and y. motome , phys . b * 81 * , 184409 ( 2010 ) . j. villain , r. bidaux , j.p . carton , and r. conte , j. phys . ( france ) * 41 * , 1263 ( 1980 ) . m. creutz , l. jacobs , and c. rebbi , phys . d , * 20 * , 1915 ( 1979 ) .

a ( sub-)millimeter line and continuum study of the class i protostar elias 29 in the ophiuchi molecular cloud is presented , whose goals are to understand the nature of this source , and to locate the ices that are abundantly present along this line of sight . within 1560 beams , several different components contribute to the line emission . two different foreground clouds are detected , an envelope / disk system and a dense ridge ofrich material . the latter two components are spatially separated in millimeter interferometer maps . we analyze the envelope / disk system by using inside - out collapse and flared disk models . the disk is in a relatively face - on orientation ( ) , which explains many of the remarkable observational features of elias 29 , such as its flat sed , its brightness in the near infrared , the extended components found in speckle interferometry observations , and its high velocity molecular outflow . it can not account for the ices seen along the line of sight , however . a small fraction of the ices is present in a ( remnant ) envelope of mass 0.120.33 , but most of the ices (% ) are present in cool ( k ) quiescent foreground clouds . this explains the observed absence of thermally processed ices ( crystallized ho ) toward elias 29 . nevertheless , the temperatures could be sufficiently high to account for the low abundance of apolar ( co , n , o ) ices . this work shows that it is crucial to obtain spectrally and spatially resolved information from single - dish and interferometric molecular gas observations in order to determine the nature of protostars and to interpret infrared iso satellite observations of ices and silicates along a pencil beam .

astro-ph0201317

[ se29:intro ] a rich chemical and physical interplay exists between gas and grains in which molecules are formed on grains , creating ice mantles that are preserved in environments ranging from quiescent dense molecular clouds to envelopes and disks around protostars . various processes , among which are bombardment by cosmic rays , ultraviolet irradiation , heating , and shocks , can physically or chemically alter the icy mantles , or return molecules into the gas phase ( see @xcite , and references therein ) . a study of the chemical evolution of dense clouds to planet forming disks would ideally involve observations of molecular gas and ices in a range of environments , from quiescent clouds to disk dominated protostars . the most pristine , initial conditions are presumably well sampled by field stars behind clouds tracing quiescent molecular cloud material ( e.g. @xcite ) . lines of sight to protostars are more difficult to characterize , however , since they may trace quiescent foreground material , in addition to the gas and ices in their envelopes and disks ( e.g. @xcite ) . it is thus crucial to characterize the line of sight conditions in order to locate the ices and derive physical conditions and eventually the evolution of the molecular gas and solid state in the interstellar medium . in this paper , we study the line of sight of the class i protostar elias 29 in the @xmath0 oph molecular cloud , using ( sub)millimeter single dish and interferometer gas phase observations . this object is one of the most luminous protostars ( @xmath8 ; @xcite ) in the nearby @xmath0 oph complex ( @xmath9 pc ; @xcite ) , yet little is known about its nature and line of sight conditions . abundant ice has been detected in its direction @xcite . a detailed analysis of ice band profiles indicates that the ices are not strongly thermally processed ( i.e. the ices are not crystallized or segregated ) , despite the presence of abundant warm molecular gas toward the object @xcite . this contrasts with high mass , luminous ( @xmath10 ) protostars , where significant thermal processing of the ices accompanies the presence of abundant warm molecular gas @xcite . the result obtained for elias 29 can only be understood once the location of the ices and the physical conditions of the various gas components are known . therefore , in this paper we try to identify any foreground material , the presence of a circumstellar envelope as well as the presence and orientation of a circumstellar disk , and the column density of each component . we will then address the question where the ices are located , and what their relation is to the young star . this study will , as a consequence , reveal important information on the nature and evolutionary stage of elias 29 , which has many interesting and unique properties @xcite . details of the single dish and interferometer observations are presented in 2 , and the maps and spectra are decomposed and interpreted in 3 . the physical conditions are determined for the different components along the line of sight , among which are two foreground clouds ( 4.1 ) , a remnant envelope and face - on disk ( 4.2 ) , as well as a dense ridge from which elias 29 probably formed ( 4.3 ) . the gas phase conditions and abundances are linked to the ice observations in 4.4 . the depletion of gas phase species is compared to young class 0 objects and quiescent clouds in 4.5 . the results are summarized in 5 . single dish and interferometer millimeter wave spectral line and continuum observations were made toward the infrared position of elias 29 ( @xmath11 = 16@xmath1227@xmath1309@xmath14.5 , @xmath15 = @xmath1637@xmath1718@xmath1 ; j2000 ) . rotational lines of co , @xmath2 , cs , @xmath18 , @xmath19 , and isotopes were selected in the 70 - 400 ghz spectral range based on their sensitivity to column density , temperature , and density @xcite . below we discuss the technical details of these observations . not discussed is a map of the co 6 - 5 line ( 692 ghz ) , for which we refer to c. ceccarelli et al . ( in preparation ) . also , the technical details of the 1.3 mm iram-30 m continuum map at 15 spatial resolution that is used in this paper are discussed elsewhere ( motte , andr , & neri 1998 ) . finally , the infrared spectral energy distribution ( sed ) that we use consists of 2 - 45 , and 45 - 200 @xmath20 spectra that were obtained with the iso sws and lws instruments in apertures of @xmath525 and @xmath580respectively , and were extensively discussed in @xcite . a summary of all the observations used in this paper is given in table [ t : obssum ] . single dish nrao-12 m , jcmt , and cso millimeter wave observations were made with a single pointing toward elias 29 during a number of observing runs in the 1995 - 2001 period ( table [ t : obssum ] ) . in some lines , small maps were made of the elias 29 environment with the cso . at the jcmt and cso , we used the das and aos spectrometer back - ends in the highest available spectral resolution mode ( @xmath50.1 @xmath21 ) , in the 200 - 400 ghz spectral range . at the nrao-12 m , we used the autocorrelator back - end , with a channel width of 0.049 mhz , resulting in nyquist sampled spectra with resolutions of 0.40 - 0.20 @xmath21 at the frequencies 70 - 145 ghz . for low frequency transitions , that trace extended cloud structure , we took @xmath11 = 16@xmath1223@xmath1301@xmath14.5 , @xmath15 = @xmath1636@xmath1758@xmath1 ( j2000 ) as an off position , which is found to have very little or no @xmath22 emission @xcite . for higher frequency co transitions , we took an off position of 2700 in azimuth , and for other molecules the azimuth offset was 900 . no contamination by emission in the off position is evident in our spectra . our line observations were corrected for atmospheric attenuation and telescope losses , using the standard chopping wheel technique @xcite . the nrao-12 m data retrieved from the telescope are on a @xmath23 scale , and the cso and jcmt data in @xmath24 . to convert to the main beam brightness temperature ( @xmath25 ) , we divide by the beam efficiencies : @xmath26 this corrects for losses in the beam side lobes . for the jcmt and cso , losses due to forward spillover and scattering are included in @xmath27 . the beam efficiencies applied to our data are summarized in table [ t : eta ] . for the cso telescope , @xmath27 and the main beam size ( table [ t : sdish ] ) were determined by observing mars during observing runs in march 1999 and january 2001 . for the nrao-12 m , we used the theoretical values of @xmath28 from the ruze equation , increased by a small factor ( @xmath29 ) , needed to reproduce the experimental values @xcite . as a check , we determined @xmath28 from our observation of jupiter , yielding @xmath30=0.80 at 145 ghz , in excellent agreement with the assumed value . for the jcmt , the efficiency factors are taken from regularly measured values reported on the internet . the calibration accuracy between different observing runs is not expected to be better than 25% ( e.g. @xcite ) . as a check on the calibration accuracy , we observed the very bright nearby source iras 162932422 during our jcmt and cso runs . we find that line intensities , observed with the same telescope during the night and in separate observing runs , indeed scatter with 20 - 30% variations around the mean . similar differences are found when comparing jcmt observations to the standard spectra available at the jcmt internet page . in this paper , we will assume an uncertainty of 25% in the line intensity , unless noted otherwise . the data were reduced with the gildas / class reduction package , applying low order baselines and using weighted means in averaging individual spectra . the reduced single dish spectra and maps are presented in figs . [ f : hcop] [ f : sdish ] and table [ t : sdish ] , and analyzed in 3.1 . elias 29 was observed with the ovro millimeter array in the 1999/2000 season ( table [ t : obssum ] ) . the low elevation of the source allowed for only short tracks ( @xmath57 hours ) . the digital correlator system was used in the `` 1 8 2 '' mode , providing simultaneous observations of the @xmath31 1 - 0 and @xmath22 1 - 0 lines in one local oscillator setting as well as the @xmath2 1 - 0 and sio 2 - 1 lines in another . the spectral resolution is 0.35 @xmath21 , with a velocity coverage of 20 @xmath21 . the telescope configurations l , h , and e were used in all lines , as well as the continuum . the weather conditions were best during the l and h observations , resulting in @xmath32 500 k at 110 ghz , and 800 k at 87 ghz . system temperatures were worse in the e track : 700 and 1300 k respectively . the data were reduced in a standard way in the ovro / mma reduction package . the nearby source nrao 530 was used as a gain calibrator , and flux calibration was performed on the planets uranus and neptune . the calibrated tracks were combined , and deconvolved images were constructed with the miriad software package using uniform weighting . the resulting synthesized elliptical beam sizes are 4@xmath338at 87 ghz and 3@xmath336 at 110 ghz ; the noise level achieved per channel in the final data is 0.26 jy / beam and 0.13 jy / beam respectively . strong and extended @xmath22 and @xmath2 1 - 0 emission is detected , with perhaps a weak detection of @xmath311 - 0 ( figs . [ f : ovromap][f : ovrochan ] ) . sio 2 - 1 is undetected toward elias 29 . no obvious strong continuum source was seen in the ovro 2.7 and 3.4 mm maps . a weak 3 @xmath34 peak ( 7 mjy ) is detected at 3.4 mm . this is the strongest peak in the map , and it is centered on the infrared position of elias 29 . at 2.7 mm a peak of only 2 @xmath34 ( 5 mjy ) significance appears at the same position . for our analysis , we take a weighted mean of these values , which we will refer to as the continuum flux at 3.0 mm : @xmath35(3 mm)=6.1@xmath361.7 mjy . this is only a 3.5 @xmath34 result , and needs confirmation by additional observations . the single dish maps show that elias 29 is not a particularly prominent center of molecular line emission . within a radius of @xmath37 , the @xmath2 3 - 2 and @xmath31 2 - 1 emission is highly structured , and quite differently distributed ( figs . [ f : hcop ] and [ f : csochan ] ) . [ f : hcop ] [ f : csochan ] the @xmath2 3 - 2 emission , tracing high densities ( @xmath38 @xmath39 ) , is concentrated in a remarkable ridge - like structure , oriented in the southeast - northwest direction ( fig . [ f : hcop ] ) , at a velocity of @xmath40@xmath55.0 @xmath21 ( fig . [ f : csochan ] ) . it is likely no coincidence that the protostars elias 29 , wl 20 , lfam 26 , gy 210 as well as the 1.3 mm continuum protostellar condensations e - mm3 and e - mm5 are all located along this dense ridge ( fig . [ f : hcop ] ) . the ridge is also particularly prominent in the 800 @xmath20 continuum @xcite . star formation along dense filamentary structures is common in the @xmath0 oph cloud , and has been explained by the presence of magnetic field tubes or , more likely , by externally induced shocks ( see @xcite for a short discussion ) . the @xmath31 2 - 1 emission , a column rather than volume density tracer , shows that at least three clouds are present along the line of sight of elias 29 ( fig . [ f : csochan ] ) . the channel maps show a cloud at @xmath40@xmath41 @xmath21 that peaks to the northeast of elias 29 , and a cloud at @xmath40@xmath42 @xmath21 spread rather evenly over the map . the brightest cloud at @xmath55 @xmath21 peaks prominently near the south - southwest of the map , and is probably that in which the dense @xmath2 ridge resides , given the similar velocities . all these clouds are likely present in the foreground , since absorption in the @xmath43 emission lines is seen at these velocities ( fig . [ f : sdish ] ) . [ f : sdish ] llrcccccll & transition & frequency & @xmath44 & @xmath45 & fwhm & @xmath40 & beam @xmath46 & telescope & date + & & mhz & k & k.@xmath21 & @xmath21 & @xmath21 & arcsec & & + co & 2 - 1 & 230538.0 & 15.9 & 19.4 & 6 & 2/6.5 & 21 & jcmt & 03/1997 + & 3 - 2 & 345796.0 & 26 & 98 & 7 & 2/6.5 & 14 & jcmt & 02/1996 + & & & 9 & 46 & 6 & 2/6.5 & 21 & cso & 07/2000 + & 6 - 5 & 691473.0 & 17 & 124 & 12 & 1.8/6.5 & 7 & jcmt & 04/1995 + @xmath47co & 6 - 5 & 661067.4 & 10 & 39 & 3.6 & 4.78 & 7 & jcmt & 04/1995 + c@xmath48o & 1 - 0 triplet & 112358.7 & 0.40 & 0.99 & 2.3 & & 56 & nrao & 05/1995 + & & 112359.0 & 0.94 & 1.00 & 1.0 & 3.58 & & nrao & 05/1995 + & & 112360.0 & 0.68 & 1.05 & 1.5 & & & nrao & 05/1995 + & 2 - 1 multiplet & 224714.3 & 1.76 & 4.95 & 2.63 & 4.08 & 22 & jcmt & 03/1995 + c@xmath49o & 1 - 0 & 109782.2 & 6.34 & 11.0 & 2.0 & 3.6 & 57 & nrao & 05/1995 + & 2 - 1 & 219560.4 & 5.81 & 16.2 & 2.62 & 4.19 & 22 & jcmt & 03/1995 + & & & 9.6 & 21.3 & 2.41 & 4.05 & 35 & cso & 01/2001 + & 3 - 2 & 329330.6 & 4.3 & 10.4 & 2.28 & 4.24 & 15 & jcmt & 02/1996 + cs & 2 - 1 & 97981.0 & 1.28 & 3.55 & 2.61 & 3.96 & 64 & nrao & 05/1995 + & 5 - 4 & 244935.6 & 0.56 & 0.94 & 1.56 & 4.92 & 20 & jcmt & 03/1995 + & 7 - 6 & 342883.0 & @xmath500.09 & @xmath51 & & & 14 & jcmt & 02/1996 + c@xmath52s & 2 - 1 & 96412.9 & 0.14 & 0.23 & 1.62 & 3.29 & 65 & nrao & 05/1995 + h@xmath7co para & @xmath53 & 72838.0 & 0.62 & 1.44 & 2.19 & 3.80/5.39 & 86 & nrao & 05/1995 + & @xmath54 & 145603.0 & 0.56 & 0.89 & 2.0 & 3.67/4.10 & 43 & nrao & 05/1995 + & @xmath55 & 218222.2 & 0.39 & 0.66 & 1.89 & 4.60 & 22 & jcmt & 03/1995 + & @xmath56 & 218475.6 & @xmath500.06 & & & & 22 & jcmt & 03/1995 + h@xmath7co ortho & @xmath57 & 140839.5 & 0.80 & 1.85 & 2.17 & 4.20 & 45 & nrao & 05/1995 + & @xmath58 & 150498.4 & 0.80 & 1.85 & 2.17 & 4.20 & 42 & nrao & 05/1995 + & @xmath59 & 225697.8 & 0.41 & 0.76 & 1.76 & 4.93 & 21 & jcmt & 03/1995 + & & & 0.26 & 0.48 & 1.70 & 4.51 & 32 & cso & 07/2000 + ch@xmath60oh & 2 - 1 triplet & 96739.4 & 0.15 & 0.19 & 1.24 & & 65 & nrao & 05/1995 + & & 96741.4 & 0.18 & 0.33 & 1.68 & 3.59 & & nrao & 05/1995 + & & 96744.6 & @xmath500.03 & & & & & nrao & 05/1995 + & 5 - 4 multiplet & 241802.0 & @xmath500.05 & & & & 20 & jcmt & 03/1997 + & & & @xmath500.02 & & & & 31 & cso & 06/2000 + @xmath2 & 1 - 0 & 89188.5 & 3.57 & 5.69 & 3.3/1.0 & 4.3/4.7 & 70 & nrao & 05/1995 + & 3 - 2 & 267557.6 & 1.06 & 2.7 & 2.28 & 4.77 & 18 & jcmt & 05/1996 + & & & 4.34 & 4.9 & 1.05 & 4.63 & 28 & cso & 06/2000 + & 4 - 3 & 356734.3 & 0.30 & 0.80 & 2.5 & 4.5 & 21 & cso & 03/1999 + h@xmath47co@xmath61 & 1 - 0 & 86754.3 & 0.38 & 0.36 & 0.90 & 4.66 & 72 & nrao & 05/1995 + & 3 - 2 & 260255.5 & 0.04 ( 0.02 ) & 0.09 ( 0.02 ) & 2.15 & 4.89 & 29 & cso & 03/1999 + in various other single dish lines these clouds show up as discrete emission components ( fig . [ f : sdish ] ) . the 2.7 and 3.8 @xmath21components are primarily seen in the low - lying transitions of the cs , @xmath2 , @xmath18 , and @xmath19 molecules , indicating low densities and temperatures of these extended clouds . the emission at 5.0 @xmath21 , identified with the elias 29 core and ridge , shows up in the high excitation lines . given the fact that these emission components have different spatial distributions , their relative intensity also depends on beam size . the @xmath2 3 - 2 line is a factor of 4 stronger and a factor of 2 narrower in the cso data compared to jcmt : the larger cso beam ( 28@xmath1 versus 18@xmath1 ; table [ t : sdish ] ) picks up more emission from the ridge , which is bright in the west ( 3.2 ) . the wide beam nrao-12 m @xmath2 1 - 0 and h@xmath47co@xmath61 1 - 0 spectra have remarkably narrow 5.0 @xmath21 components , originating from the extended ridge . in contrast , the @xmath18 @xmath59 line may peak on the elias 29 core , rather than the ridge . it has the same width in the cso and jcmt beams , but the cso spectrum is weaker ( table [ t : sdish ] ) . we have performed a gaussian decomposition of the emission lines to separate the foreground clouds from the dense material around elias 29 . the derived relative strength of these blended lines depends sensitively on the assumed line width . the most reasonable solution , based on the @xmath18 and cs lines and the @xmath31 map , is to simultaneously fit the peak intensity with three gaussians , centered on @xmath40= 2.7 , 3.8 , and 5.0 @xmath21 , with fwhm=0.5 , 0.8 , and 1.5 @xmath21 . when needed , we allow a 30% variation in the fwhm , and 0.20 @xmath21 in @xmath40 . some of these small variations may be real , but given the complexity of the line profiles and spatial distribution of the various components , we will not seek a physical interpretation for this . the integrated intensities are summarized in table [ t : decomp ] , and the physical conditions are derived from these in 4 . we conclude that when studying individual protostars in the @xmath0 oph cloud , it must be realized that many different physical components are present within single dish beams of @xmath62 or larger . the spatial information of interferometers is essential here . [ f : ovromap ] [ f : ovrospec ] [ f : ovrochan ] in the ovro interferometer maps , bright @xmath22 and @xmath2 1 - 0 emission is detected in the direct neighbourhood of elias 29 ( fig . [ f : ovromap ] ) . the @xmath22 emission is most strongly peaked on the infrared position , and with a spatial fwhm@xmath63 ( 900 au ) it is resolved along the ew direction , where the ovro beam is smallest ( 3@xmath1 ) . this is the elias 29 core , which likely consists of a disk / envelope system ( 4.2 ) . an extension of @xmath64 toward the southwest is visible in both lines , but most prominently in @xmath21 - 0 , which must be attributed to the ridge discussed above . indeed , on this small scale the @xmath2 emission is also parallel to the iram-30 m 1.3 mm continuum emission ( fig . [ f : ovromap ] ) . the ovro @xmath22 and @xmath2 1 - 0 spectra peak between @xmath40=57 @xmath21 , and show no evidence for the strong , extended 2.7 and 3.8 @xmath21foreground emission seen in the single dish maps ( fig . [ f : ovrospec ] ) . the @xmath31 1 - 0 emission is _ spectacularly _ absent in the interferometer spectrum ; it is at least a factor of 10 weaker compared to the single dish data . after dilution to the nrao-12 m beam ( taking the emitting area from the ovro @xmath22 image ) , the ovro @xmath31 signal is even a factor of 60 weaker than the bright detection with the nrao-12 m telescope . this shows again that the bulk of the molecular material in the elias 29 line of sight is not associated with elias 29 , but instead is present in extended foreground clouds , which are resolved out by the ovro interferometer . it is interesting to note that the ovro line profiles are double peaked ( fig . [ f : ovrospec ] ) , with a 5.0 @xmath21 component emitting in the dense ridge ( fig . [ f : ovrochan ] ) . emission at the central infrared position is strongest at 6 @xmath21 . whether this is a true dynamical difference between the elias 29 core and the ridge , or artificially created by self - absorption at @xmath65 @xmath21 , is difficult to answer at present . in this paper we will keep referring to this as the 5.0 @xmath21 component . a disk / envelope / ridge decomposition is attempted in 4.2 . it is our aim to derive the physical structure of the surroundings of elias 29 , and to locate the origin of the ices , seen abundantly along this line of sight @xcite . in 4.1 the extended clouds at 2.7 and 3.8 km s@xmath66 identified above are discussed ; 4.2 and 4.3 describe the more immediate circumstellar environment of elias 29 in terms of a near face - on disk , a remnant envelope , and the dense ridge from which the star may have formed ( see fig . [ f : schem ] ) . the physical conditions in each of the components are constrained using the intensities and intensity ratios of the single dish and interferometer line emission , the 1.3 mm continuum emission @xcite , and the infrared sed @xcite . this is linked to the ice observations in 4.4 . the c@xmath49o line emission at 2.7 and 3.8 km s@xmath66 is extended over several arcminutes ( fig . [ f : csochan ] ) , and thus is associated with the overall @xmath0 oph cloud complex . assuming that the clouds are sufficiently homogeneous , we can use the ratios of the decomposed line intensities ( table [ t : decomp ] ) to find the density and temperature , and subsequently the column density , in each of the clouds . we use the escape - probability method described by @xcite to calculate the molecular excitation . the most useful constraints on the density and temperature are given by the intensity ratios of @xmath31 10/32 and @xmath210/32 . [ f : phys ] visualizes the density and temperature values allowed by the observed ratios , taking into account line - opacity effects . the 2.7 @xmath21 component has a temperature @xmath67 k , and a density @xmath68(h@xmath7)=(@xmath69 ) @xmath70 @xmath39 . for the 3.8 @xmath21 component , the density is less than @xmath71 @xmath39 , but the temperature can not be significantly constrained from the line ratios . here we use the far infrared sed to find that @xmath72 k ( 4.2 ) . taking a c@xmath49o abundance with respect to h@xmath7 of @xmath73 ( @xmath74[co]/@xmath74[c@xmath49o]=560 , @xcite ; @xmath74[h@xmath7]/n[co]=5000 , @xcite ) , the h@xmath7 column densities of the 2.7 and 3.8 km s@xmath66 clouds are @xmath76 @xmath77 and @xmath78 @xmath77 , respectively ( table [ t : phys ] ) . this also assumes that co is not strongly depleted in these clouds , which is validated in 4.5 . these parameters fit the optically thin c@xmath48o lines also , indicating that a correct line opacity was used with these physical parameters at hand , we calculate that the optical depth of the clouds in the co 5 - 6 transition is @xmath54 . this is sufficiently high to explain the absorption in the co 6 - 5 lines at 2.7 and 3.8 @xmath21 ( fig . [ f : sdish ] ) . the bright co 6 - 5 emission is closely associated with elias 29 ( c. ceccarelli et al . , in prep . ) , and therefore the 2.7 and 3.8 @xmath21 clouds are located in front of this source . independent information on the physical conditions in these extended foreground clouds is obtained from the sed longward of 100 @xmath20 ( 4.2 ) . the columns and temperatures compare also favorably with the extended cold dust found by sensitive 1801100 @xmath79 m balloon - based measurements ( @xcite ; table [ t : colden ] ) . the observed lines of cs , h@xmath7co , and ch@xmath60oh do not further constrain the physical conditions , but they can be used to derive abundances . these are listed in table [ t : phys ] , and further discussed in 4.5 . understanding the circumstellar environment of elias 29 requires the combination of several pieces of crucial information : the single - dish @xmath2 32 map and 1.3 mm dust - continuum distribution ( fig . [ f : hcop ] ) , the spatially resolved emission in @xmath47co 10 and @xmath2 10 observed by ovro ( fig . [ f : ovromap ] ) , and the infrared sed obtained by the sws and lws spectrometers of the iso satellite ( @xcite ; fig . [ f : sed ] ) . the @xmath2 32 and 1.3 mm continuum maps show that elias 29 is located in a narrow , dense ridge , @xmath80 wide and several arcminutes long . the strongest @xmath2 32 peak in the ridge is situated near elias 29 . the ovro images and channel maps ( figs . [ f : ovromap ] and [ f : ovrochan ] ) show that part of the @xmath47co and @xmath2 emission is centered on elias 29 , while a significant fraction follows the crest of the ridge , @xmath81@xmath82 offset to the west / southwest . this shows that multiple components are present in the immediate vicinity of elias 29 ( i.e. , within a typical single - dish beam ) , in addition to the multiple velocity components along the line of sight ( 4.1 ) . the velocity coincidence between the ridge and elias 29 further supports that the star formed from the ridge . the most important clues about the nature of elias 29 are offered by the infrared sed . because of the large beam with which these measurements were taken ( 2 ) , the sed reflects not only elias 29 but also many of the other components identified above . our ability to derive accurate parameters for each component therefore proved essential in helping to understand the nature of elias 29 . the sed ( fig . [ f : sed ] ) is remarkably flat between 10 and 200 @xmath79 m . at 10 @xmath79 m there is a prominent silicate absorption band , and at 5 @xmath79 m the sed shows a broad peak on top of which a number of sharp ice absorption features are present . below 2 @xmath79 m the emission drops off sharply . flat protostellar seds are generally explained by a circumstellar disk . compared to other circumstellar material distributions , a disk , especially a flared one , has the surface - area vs. temperature distribution required to create a flat sed . for our analysis , we adopt the flared disk model of @xcite with the parameters listed in table [ t : model ] . the flatness of the sed is explained in this model by a superheated surface layer . the continuum emission of the disk is modeled using the radiative transfer part of the monte - carlo code of @xcite ; this part of the code simply calculates the expected continuum emission from a given distribution of density and temperature by building up a grid of points on the sky , for each of which the radiative transfer is solved along straight lines . the resulting grid is then convolved with the appropriate beam sizes . because of the large number of parameters involved , their values should be considered as representative only . we believe that the general characteristics of the model are firmly established , however . the flatness of the sed over a large wavelength range ( 4 - 100 @xmath20 ; fig . [ f : sed ] ) limits the inclination of the system to @xmath83 ( @xmath84 being edge - on ; @xcite ) . such a low inclination can not explain the deep ice and silicate absorption features that are visible in the observed sed ; these must originate in the envelope and foreground clouds . the low disk inclination is compatible with the very high velocity ( @xmath85 @xmath21 ) , variable outflowing hot co gas seen in this object ( a.c.a . boogert , g.a . blake , & m.r . hogerheijde , in prep . ) . a fully face on orientation is however not likely , given the presence of low surface brightness scattered k band light out to distances of 15@xmath1 ( 2400 au ) from the central object ( zinnecker , perrier , & chelli 1988 ) . this light presumably traces the outflow lobes , adjacent to the remnant envelope and the outer edges of the disk . the emission is extended along the southwest / northeast direction , perpendicular to the high velocity outflowing gas ( c. ceccarelli et al . , in prep . ) , with an axis ratio compatible with an inclination of roughly @xmath86 . flattening on a similar scale is observed in the ovro images and in the 1.3 mm continuum iram-30 m single dish map . the flux between 20 and 40 @xmath79 m indicates a temperature scaled upward by a factor of 2.25 with respect to the values used by @xcite . the central star is therefore approximately 50 times brighter than their standard model , presumably because it has a higher mass ( using the scaling relation @xmath87 ; @xcite ) . the disk midplane dominates the emission beyond 25 @xmath79 m , and the shape of the sed at these wavelengths limits the disk radius to @xmath5 500 au . the warmer surface layer dominates the shorter wavelengths and generate a silicate _ emission _ feature at 10 @xmath79 m ( fig . [ f : sed ] ) . the disk model can explain only @xmath520% of the emission maximum around 5 @xmath79 m . indeed , speckle observations of elias 29 reveal the presence of a @xmath5400 au radius thermally emitting region ( @xmath88 a few 100 k ; i.e. not scattered light ) responsible for @xmath520% of the m band flux @xcite , which , in our model , is explained by the warm disk surface layer . the observed remaining 80% of the m band emission originates from hot dust within a few au from the central object . perhaps this is a puffed up hot rim at the inner disk edge , just outside the region where dust has evaporated ( @xmath89 k ; @xcite ; @xcite ) . the innermost , hottest , part of the rim may be related to the 0.25 au radius structure found in lunar occultation k band observations @xcite . we do not attempt to fit the inner regions and stellar photosphere with that level of detail , but instead use an 800 k blackbody with an effective radius of 1.4 au to explain the remaining 80% of m band emission , and roughly the shape of the sed at shorter wavelengths ( fig . [ f : sed ] ) . the disk alone is also insufficient to explain the emission beyond @xmath90 @xmath79 m . the single - dish line emission clearly reveals the presence of appreciable columns of material at 2.7 and 3.8 km s@xmath66 ( 4.1 ) . when filling the iso beam with these columns , the emission between 55 and 180 @xmath79 m can be fit at a dust temperature of 19 k. this material provides sufficient opacity in the 10 @xmath20 silicate band to turn the disk s emission feature into the observed absorption feature . the opacity shortward of 2 @xmath79 m is also sufficient to explain the observed steep drop in emission at short wavelengths . this strongly suggests that a large fraction of the 2.7 and 3.8 km s@xmath66 material is located in front of elias 29 , as was found from the absorption in the @xmath43 emission lines as well ( 4.1 ) . while the model of disk and foreground clouds fits the sed well , it produces a 1.3 mm continuum emission that is a factor of 3 larger than observed with the iram-30 m telescope , at large distances from the object ( @xmath91 mjy ; fig . [ f:13mmprof ] ) . this may result directly from the adopted dual - beam mapping procedure of the iram-30 m observations , which filters out emission on the chopping scale @xcite . additionally , the model of disk and foreground predicts a source size at 1.3 mm that is too small ( fig . [ f:13mmprof ] ) . the observed profile indicates that the source extends to a radius of at least 3800 au ( 25 ) , much larger than is realistic for a circumstellar disk . instead , it suggests that elias 29 and its disk are embedded in a residual cloud condensation from which the system formed . for lack of stronger constraints , we model this envelope ( as we will refer to this condensation from now on ) with the inside - out collapse model of @xcite . earlier work ( e.g. , @xcite ; @xcite ) has shown that this model provides an adequate description of the density and temperature of envelopes around ysos . the two parameters of this model are the sound speed and the time since the collapse started , for which we take , rather arbitrarily , 0.13 @xmath21 and @xmath92 yr respectively . this gives an envelope mass of 0.12 @xmath93 and a radius of the ` collapse expansion wave ' of 6000 au . additional parameters are listed in table [ t : model ] . this model , combined with the disk , produces a radial emission profile that fits the measured 1.3 mm continuum distribution much better than the disk alone ( fig . [ f:13mmprof ] ) . the sed below @xmath94 @xmath20 is remarkably little affected by the envelope , even though with 0.12 m@xmath95 it is much more massive than the disk ( 0.012 m@xmath95 ; table [ t : model ] ) . the material in the disk , of course , has a somewhat higher temperature range than the envelope ( 20033 k vs. 4417 k , respectively ) , yielding significantly more infrared flux . the envelope produces only a small amount of additional absorption in the 10 @xmath79 m silicate band . the parameters for the disk and envelope derived above ( table [ t : model ] ) maximize the contribution to the sed from the disk and minimize the mass of the envelope . there are several pieces of evidence that favor , but not proof , this maximum disk model above models with smaller disks . first , as for sources with directly detected disks @xcite , the disk contributes about half of the 1.3 mm continuum flux ( @xmath5 120 mjy/15 beam ; foreground subtracted ) . indeed , assuming that the weak 3 mm continuum flux within the small ovro beam ( 2.2 ) orginates from the disk , then the expected 1.3 mm disk flux would be half the single dish flux ( @xmath35[1.3 mm]=49@xmath3614 mjy , taking spectral index 2.5 ; @xcite ) . second , the abovementioned 400 au radius structure , emitting thermally at 5 @xmath20 , may well be the surface of a large , warm disk . and third , a large face - on flared disk is the simplest way to explain the observed flat infrared sed . nevertheless , the evidence for a large 500 au radius disk is not conclusive and therefore we also try to maximize the envelope s contribution in our model . we find that its mass can be as large as 0.33 @xmath93 , while the corresponding disk mass needs to be reduced to 0.002 m@xmath95 , in part by reducing the disk s radius to 30 au to preferentially decrease the amount of cool material . the disk s temperature scaling does not change appreciably with respect to that used before . it must be emphasized that in this minimum disk scenario , the flatness of the infrared sed is the result of a `` disk / envelope conspiracy '' , rather than a direct consequence of the flared disk s properties . although the disk and envelope masses are different from those listed in table [ t : model ] , they are comparable to within the same order of magnitude and this level of uncertainty does not affect our conclusions . in the remainder of this section we will adopt the slightly preferred values from table [ t : model ] , i.e. the maximum disk case . we use the accelerated monte - carlo method of @xcite to calculate the molecular excitation and the resulting line emission in the various beams , including a realistic sampling of spatial scales recovered by the interferometer observations . because the disk and envelope exhibit a range of densities and temperatures , we can not use the much simpler escape probability approach here . we calculate the line intensities for three different cases ( table [ t : predict ] ) : for the disk alone , for the envelope alone , and for the combination of disk+envelope taking into account optical depth effects . the predicted line intensities are meant to be illustrative only . the general trends are robust , however . adopting ` fiducial ' molecular abundances derived for the two foreground clouds ( table [ t : phys ] ) , we find that the emission of c@xmath49o observed with ovro is overestimated by at least a factor of 5 by the disk+envelope model . this can be accounted for by adopting a depletion factor of 5 - 10 for co in the disk . a large co depletion factor has been seen in many other disks as well @xcite . from the single - dish lines it can also be concluded , in particular from cs 7 - 6 , that a significant depletion in the disk is needed ( table [ t : phys ] ) . furthermore , we find that the higher excitation transitions are dominated by the disk+envelope , while the lower excitation lines can only be partially explained by this model . a fair fraction of these lines must therefore originate in the ridge . as mentioned above , elias 29 appears to be embedded in a ridge traced by the 1.3 mm continuum and @xmath2 3 - 2 line emission , but slightly offset from the crest of the ridge . how much of the line emission measured in the single - dish and interferometer beams at 5.0 km s@xmath66 can be attributed to the elias 29 disk+envelope system , and how much originates in this dense ridge ? this question is relevant , because it addresses the issue of how much additional column density resides in the ridge that can harbor the ice - absorption bands . when we subtract the predicted emission of elias 29 s disk and envelope from the single - dish line fluxes at 5.0 km s@xmath66 of table [ t : decomp ] , we can use the ratios of the remaining intensities to roughly constrain the conditions in the ridge . [ f : phys]c shows the results for the c@xmath49o 1 - 0/3 - 2 and @xmath2 1 - 0/3 - 2 lines . the density is constrained at @xmath96 @xmath97 , significantly higher than that of the foreground clouds , while the temperature is found to be similarly low at @xmath98 k. these parameters are consistent with the appearance of the ridge in the 1.3 mm and @xmath2 3 - 2 maps ( cold and dense ) . the corresponding column density along the ridge is @xmath99 @xmath77 , comparable to that of the largest column density foreground cloud at 3.8 km s@xmath66 . elias 29 is offset from the maximum column density in the ridge , however , and the pencil beam traced in the infrared absorption features may contain a significantly smaller column density . the fact that the foreground clouds and , to a lesser extent , the envelope can already fit the 10 @xmath79 m absorption would suggest that the ridge does not contribute significantly . this conclusion does not change if we assume the higher envelope mass of 0.3 @xmath93 . we have identified four regions of significant column density in the line of sight of elias 29 : the 2.7 and 3.8 @xmath21 foreground clouds , the dense ridge , and the envelope . all of these are cold environments , in which icy mantles can exist , and together they must account for the rich spectrum of ices seen with the iso satellite @xcite . their relative contribution to the visual extinction and ice bands depends on which fraction of the column is actually in front of elias 29 . the 2.7 and 3.8 @xmath21 clouds are likely in front ( 4.1 ) with columns corresponding to visual extinctions @xmath100=2.9 and 8.2 . the visual extinction of the dense ridge , @xmath1015.9 , is an upper limit beause the ridge is somewhat displaced from elias 29 . for the envelope , @xmath102 is 2.5 magn . for the minimum , and 6 magn . for the maximum envelope case . these are upper limits because the system is seen at low inclination and much of the envelope material is likely not in front of elias 29 . the extinction from either ridge or envelope is however not negligible , given the absorption seen at 5.0 @xmath21 in the co 6 - 5 emission line ( fig . [ f : sdish ] ) . finally , extinction by the circumstellar disk is not significant due to its low inclination . there is observational evidence that ices in molecular clouds form only above a certain threshold extinction value ( @xmath103 ; @xcite ) , due to photodesorption at the cloud edges @xcite . above @xmath103 , the ice abundances grow linearly with @xmath102 . the visual extinction can thus , in principle , be used as an indicator for h@xmath7o ice growth and destructions mechanisms . unfortunately , the usefulness of this relation is limited by the uncertainty in @xmath103 . measurements of a variety of sight - lines in @xmath0 oph indicate @xmath103=12 @xcite , but @xmath104 toward ob stars tracing ices in foreground clouds only @xcite . the @xmath102 values used to calibrate the aforementioned relation are highly uncertain as well @xcite . while our emission line method has its own uncertainties ( beam dilution , disk / envelope ratio ) , the generally used h k color method may be uncertain by a factor of 2 , for example due to an unknown color of the intrinsic light source . indeed , for elias 29 a much lower @xmath105 was found from j h color ( th.p . greene , priv . comm . ) , compared to the h k color ( @xmath106 ; @xcite ) . nevertheless , using the relation in @xcite , and assuming that the threshold extinction for h@xmath7o formation ( @xmath103=12 ) applies only once for all clouds combined , we find that the total @xmath107 corresponds to @xmath108 @xmath109 , which is @xmath5030% of the observed ice column density @xcite . in this picture , the observed ice column is thus much higher than that expected from the total visual extinction along the line of sight . given the abovementioned uncertainties in the ice column versus @xmath102 and @xmath103 relation , the most reliable conclusion that can be drawn from this is that h@xmath7o ice sublimation does not play an important role in the elias 29 line of sight . from a different perspective , the low temperature of the clouds along the line of sight ( table [ t : phys ] ) indicates that indeed no sublimation or other forms of thermal processing ( e.g. crystallization ) of polar , h@xmath7o rich ices can take place . a similar conclusion was previously drawn from the ( low ) far infrared color index , which is found to be a good indicator of thermal processing of ices @xcite . even the temperature of the warmest absorbing component , the inner part of the envelope ( @xmath545 k ) , is less than the h@xmath7o crystallization ( @xmath110 k ) and sublimation temperatures ( @xmath111 k ) . this explains the observed lack of crystallization signatures in the absorption band profiles of solid h@xmath7o and co@xmath7 @xcite . thermal processing must have played a role for apolar , co rich ices toward elias 29 , however . the solid co / h@xmath7o ratio is 5% , which is a factor of five less than in several other lines of sight within the @xmath0 oph cloud @xcite . we find that elias 29 has heated its envelope sufficiently ( 1744 k ; 4.2 ) to evaporate co rich ices ( @xmath1121520 k ; @xcite ) , but preserve h@xmath7o rich ices . the envelope , however , harbors only a small fraction of the ices , and it is unlikely that this 38 @xmath113 object can heat ices located beyond the envelope , such as the dense ridge and the two foreground clouds . nevertheless , the temperatures of the gas and dust in the quiescent foreground clouds are similar to , or slightly higher ( table [ t : phys ] ) than the sublimation temperature of apolar ices . several known luminous b stars and the sco ob2 association play an important role in heating the dust in the @xmath0 oph cloud @xcite . the differences of the solid co / h@xmath7o ratio within @xmath0 oph may thus be explained by the location of the ices with respect to these luminous heating sources . bright external heating sources are absent in the taurus molecular cloud , and indeed the co / h@xmath7o ratios are significantly larger in that cloud @xcite . the depletion factor of co ( defined as solid/[gas+solid ] ) amounts to a few percent , assuming the ices are present in several of the clouds toward elias 29 ( using @xmath74[co - ice]=@xmath114 @xmath109 ; @xcite ) . direct measurements of solid co yield larger depletion factors in the quiescent taurus molecular cloud ( 8 - 40% ; @xcite ) . indirect measurements , that rely on gas phase abundances only , found co depletion factors as high as 90% toward deeply embedded class 0 sources such as ngc 1333 : iras 4a and 4b ( @xcite ) . we will compare gas phase abundances of other molecules toward elias 29 with those of class 0 sources , as well as the well studied dense cloud tmc 1 ( table [ t : phys ] ) . even when taking the uncertainty of gas phase abundances into account , we find that co , cs , @xmath18 , and possibly @xmath2 are an order of magnitude more depleted in the class 0 sources compared to the clouds in front of elias 29 . this can be ascribed to the much larger densities and low dust temperatures in the outer regions ( @xmath115 700 au ) of these class 0 objects , and thus shorter depletion times scales @xcite . depletion is much lower in the tmc 1 cloud compared to class 0 objects ( table [ t : phys ] ) . in fact , the @xmath18 and @xmath19 abundances are an order of magnitude larger than toward elias 29 . whether these species are depleted toward elias 29 can not be answered directly , because the upper limit to the ice abundances are typically a few@xmath116 ( @xcite ; taking @xmath117 @xmath109 ) . this is four and two orders of magnitude larger than the gas phase @xmath19 and @xmath18 abundances respectively , and demonstrates that a direct determination of depletion factors by using infrared ice bands suffers from a lack of sensitivity . alternatively , the difference in gas phase @xmath19 and @xmath18 abundances between these clouds may have its origin in different chemical evolutionary states , such was found within the tmc 1 cloud itself for several carbon bearing species @xcite . indeed , @xmath18 abundances of several other quiescent and star forming cores ( 210@xmath118 ; @xcite ) are in much better agreement with elias 29 . in some class 0 sources , the @xmath19 and @xmath18 abundances and temperatures are larger than in our comparison source ngc 1333 : iras 4 , possibly because they are more evolved @xcite . this warm gas originates from evaporated ices , that are enriched in hydrogenated molecules by grain surface reactions during the collapse of the cold envelope . shocks @xcite and thermal evaporation @xcite in the inner envelope have been proposed as ice desorption mechanisms , which can enhance the @xmath18 abundance at radii @xmath119 au by 2 orders of magnitude . the single dish observations of elias 29 do not show evidence of such warm gas , as is illustrated in fig . [ f : meth ] by comparing the @xmath19 spectrum with the class 0 source iras 162932422 . this may well be an effect of the larger column and accretion rate of the younger class 0 objects . infrared absorption line studies do reveal , however , that significant amounts of warm co and @xmath120 are present on small scales near the protostar , partly related to the outflow , and probably partly present in the accreting disk s atmosphere ( @xcite ; a.c.a . boogert et al . , in preparation ) . emission from this gas is severely diluted in the large single dish milimeter wave beams , and thus difficult to detect . finally , the @xmath2 abundance toward elias 29 is comparable to that found toward class 0 objects , but an order of magnitude lower compared to tmc 1 . the importance of depletion is more difficult to establish , because the @xmath2 abundance is highly dependent on local cosmic ray flux and electron density . even in the direction of elias 29 the @xmath2 abundance varies dramatically . although the @xmath2 10 and cs 2 - 1 lines have similar critical densities , the ridge ( at 5.0 @xmath21 ) is much more prominent in @xmath2 than in cs , with respect to the 2.7 and 3.8 @xmath21 foreground clouds ( fig . [ f : sdish ] ) . we stress however that the prominent appearance of the ridge in the @xmath2 3 - 2 map ( fig . [ f : hcop ] ) , is _ both _ an effect of increased @xmath2abundance and increased density : the ridge is weaker in the column density tracer @xmath31 , and stronger in the density tracer cs , relative to the foreground clouds . we have analyzed infrared and millimeter wave line and continuum observations to construct a model of the class i protostar elias 29 and its environment . this model has to contain a number of different components ( summarized in fig . [ f : schem ] ) : a disk to account for the @xmath5250 @xmath20 sed , an envelope contributing to the emission at 1.3 mm ( in particular its size ) , a dense ridge from which elias 29 may have condensed , and foreground material which provides most of the extinction . elias 29 can then be well described by a 500 au radius face - on flared disk with a mass of 0.012 @xmath93 , embedded in a 6000 au radius , 0.12 @xmath93 envelope . this large disk provides the simplest explanation for the observed flat sed , weakly detected 3 mm continuum emission , and 400 au radius 5 @xmath20 thermal continuum emission . the present data does however not fully exclude models with smaller disks . the minimum possible disk has a 30 au radius and a mass of 0.002 @xmath93 surrounded by an envelope of 0.33 @xmath93 . in this case the combination of disk and envelope emission produces a flat sed . the entire system is embedded in a long , dense , cold , and @xmath2rich ridge . elias 29 is slightly offset from the crest of this ridge . in front of the disk , envelope , and ridge system are two foreground clouds at a few @xmath21 lower radial velocities that cover the entire field of view . the large column of the foreground clouds , corresponding to @xmath12111 , may be responsible for the ` class i ' appearance of elias 29 , which would otherwise appear as a t tauri or herbig ae / be star ( i.e. , optically visible ) . these same foreground clouds are also the most likely repository of most of the ices seen along the line of sight ( @xmath122% ) . the low temperature of the foreground clouds explains the observed absence of crystallized ices , and the presence of large abundances of polar , h@xmath7o rich ices , i.e. thermal processing did not play a major role for the ices toward elias 29 . the foreground cloud temperature ( 25@xmath3615 k ) could , however , be high enough to explain the low abundance of apolar , volatile co rich ices , presumably due to the proximity of a number of luminous b type stars . the important question as to whether the ices in the disk or envelope have experienced thermal processing , as is seen in the envelopes of massive objects @xcite , can not be addressed given the large column of foreground material and the face - on nature of the system . this work shows the value of spectrally and spatially resolved information offered by single - dish and interferometric molecular gas observations in interpreting infrared iso satellite observations of ices along a pencil beam . it shows that , at least for the @xmath0 oph cloud , it is crucial to disentangle the different physical components along the line of sight , since otherwise incorrect conclusions may be derived , for example , on the origin and evolution of interstellar and circumstellar ices . follow up observations with sensitive submillimeter interferometers ( e.g. alma ) or high frequency ( @xmath123 ghz ) single dish telescopes with small beams will clearly provide essential information on the structure , abundances , and depletion factor of species in the disk and outflow of elias 29 , and its relationship to the envelope and ridge . in turn , such studies would provide invaluable information on the initial conditions of planet formation . we thank remo tilanus , goeran sandell , and fred baas for carrying out part of the jcmt observations in service mode . the research of fm and acab at the caltech submillimeter observatory is funded by the nsf through contract ast-9980846 . the research of mrh is supported by the miller institute for basic research in science . andr , ph . , & montmerle , th . 1994 , , 420 , 837 blake , g.a . , sandell , g. , van dishoeck , e.f . , groesbeck , t.d . , mundy , l.g . , & aspin , c. 1995 , , 441 , 689 bohlin , r.c . , savage , b.d . , & drake , j.f . 1978 , , 224 , 132 boogert , a.c.a . , ehrenfreund , p. , gerakines , p.a . , tielens , a.g.g.m . , whittet , d.c.b . , et al . 2000a , , 353 , 349 boogert , a.c.a . , tielens , a.g.g.m . , ceccarelli , c. , boonman , a.m.s , van dishoeck , e.f . , et al . 2000b , , 360 , 683 ceccarelli , c. , castets , a. , caux , e. , hollenbach , d. , loinard , l. , molinari , s. , & tielens , a.g.g.m . 2000 , , 355 , 1129 ceccarelli , c. , loinard , l. , castets , a. , tielens , a.g.g.m . , caux , e. , lefloch , b. , & vastel , c. 2000 , , 372 , 998 chandler , c.j . , & richer , j.s . 2000 , , 530 , 851 chen , h. , myers , p.c . , ladd , e.f . , & wood , d.o.s . 1995 , , 445 , 377 chiang , e.i . , & goldreich , p. 1997 , , 490 , 368 chiang , e.i . , & goldreich , p. 1999 , , 519 , 279 chiar , j.e . , adamson , a.j . , kerr , t.h . , & whittet , d.c.b . 1995 , , 455 , 234 dickens , j.e . , & irvine , w.m . 1999 , , 518 , 733 dullemond , c.p . , dominik , c. , & natta , a. 2001 , , in press dutrey , a. , guilloteau , s. , & guelin , m. 1997 , , 317 , l55 elias , j.h . 1978 , , 224 , 857 greene , th.p . , & young , e.t . 1989 , , 339 , 258 hogerheijde , m.r . 1998 , ph . d. thesis , rijksuniversiteit leiden hogerheijde , m.r . , & sandell , g. 2000 , , 534 , 880 hogerheijde , m.r . , & van der tak , f.f.s . 2000 , , 362 , 697 imanishi , k. , koyama , k. , & tsuboi , y. 2001 , preprint ( astro - ph/0104190 ) jansen , d.j . 1995 , ph . d. thesis , rijksuniversiteit leiden johnstone , d. , wilson , c.d . , moriarty - schieven , g. , giannakopoulou - creighton , j. , & gregersen , e. 2000a , , 131 , 505 johnstone , d. , wilson , c.d . , moriarty - schieven , g. , joncas , g. , smith , g. , gregersen , e. , & fich , m. 2000b , , 545 , 327 kutner , m.l . , ulich , b.l . 1981 , , 250 , 341 lacy , j.h . , knacke , r. , geballe , t.r . , & tokunaga , a.t . 1994 , , 428 , 69 loren , r.b . 1989 , , 338 , 902 mangum , j.g . 1993 , , 105 , 117 mangum , j.g . 1999 , user s manual for the nrao-12 m millimeter - wave telescope , 131 and 132 motte , f. , andr , ph . , & neri , r. 1998 , , 336 , 150 ohishi , m. , & kaifu , n. 1998 , in : chemistry and physics of molecules and grains in space , faraday discussions ( london ) , 109 , 205 ossenkopf , v. , & henning , th . 1994 , , 291 , 943 pratap , p. , dickens , j.e . , snell , r.l . , miralles , m.p . , bergin , e.a . , irvine , w.m . , & schloerb , f.p . 1996 , , 486 , 862 ristorcelli , i. , serra , g. , lamarre , j.m . , giard , m. , bernard , j.p . , et al . 1999 , in : solid interstellar matter : the iso revolution . edp sciences , les ulis , eds : l. dhendecourt , c. joblin , & a. jones , 49 shu , f. 1977 , , 214 , 488 shuping , r.y . , chiar , j.e . , snow , th . , kerr , th . 2001 , , 547 , 161 shuping , r.y . , snow , th . , chiar , j.e . , kerr , th . 2000 , , 529 , 932 simon , m. , howell , r.r . , longmore , a.j . , wilking , b.a . , peterson , d.m . , & chen , w.p . 1987 , , 320 , 344 tanaka , m. , sato , s. , nagata , t. , & yamamoto , t. 1990 , , 352 , 724 teixeira , t.c . , & emerson , j.p . 1999a , , 351 , 303 teixeira , t.c . , & emerson , j.p . 1999b , , 351 , 292 tielens , a.g.g.m . , & charnley , s.b . 1997 , origins of life and evolution of the biosphere , 27 , 23 tielens , a.g.g.m . , & hagen , w. 1982 , , 114 , 245 tielens , a.g.g.m . , tokunaga , a.t . , geballe , t.r . , & baas , f. 1991 , , 381 , 181 van der tak , f.f.s . , van dishoeck , e.f . , evans , n.j . ii , & blake , g.a . 2000 , , 537 , 283 van dishoeck , e.f . , blake , g.a . , jansen , d.j . , & groesbeck , t.d . 1995 , , 447 , 760 van dishoeck , e.f . , & blake , g.a . 1998 , , 36 , 317 whittet , d.c.b . 1974 , , 168 , 371 whittet , d.c.b . , bode , m.f . , longmore , a.j . , adamson , a.j . , mcfadzean , a.d . , aitken , d.k . , & roche , p.f . 1988 , , 233 , 321 whittet , d.c.b . , gerakines , p.a . , tielens , a.g.g.m . , adamson , a.j . , boogert , a.c.a . , 1998 , , 498 , 159 wilking , b.a . , & lada , c.j . 1983 , , 274 , 698 wilson , t.l . , & rood , r.t . 1994 , , 32 , 191 zinnecker , h. , webster , a.s . , & geballe , t.r . 1985 , in : nearby molecular clouds , springer - verlag , berlin , ed . g. serra , 81 zinnecker , h. , perrier , c. , & chelli , a. 1988 , in : eso conference and workshop proceedings no . 29 , noao - eso conference on high - resolution imaging by interferometry , ed . f. merkle , 505

the ab initio quasirelativistic hartree - fock method developed specifically for the calculation of spectral parameters of heavy atoms and highly charged ions was applied to determine atomic data for tungsten ions . the correlation effects were included by adopting configuration interaction method . the breit - pauli approximation for quasirelativistic hartree - fock radial orbitals was employed to take into account relativistic effects . the energy level spectra , radiative lifetimes , lande factors were calculated for the , and configurations of w ion . the atomic data , namely , the transition wavelengths , spontaneous emission rates and oscillator strengths for the electric dipole , electric quadrupole and magnetic dipole transitions among and within the levels of these configurations are tabulated .

1504.02055

due to its unique features , the metallic tungsten is widely employed in devices which operate at extremely high temperatures . these properties facilitate adoption of tungsten in fusion reactors including internationally developed iter tokamak @xcite . high temperatures within fusion plasma cause tungsten to evaporate . furthermore , the tungsten atoms in plasma are ionized to very high degrees . this process has a negative consequences on plasma since it makes plasma temperatures to drop significantly . to perform plasma processes modeling and diagnostics one needs huge amount of spectroscopic data describing various ionization stages of tungsten and other atoms @xcite . atomic spectroscopy can help to determine important properties of plasma @xcite . nevertheless , the spectra of the tungsten ions present in fusion plasma devices are still not investigated properly . this conclusion is evident from review @xcite where all available experimental wavelengths and energy level spectra for ions from w iii to w lxxiv are provided . it is evident from the analysis of @xcite and references within , that only a few data for the spectra of tungsten ions with open 4d- shells are available . in present work we investigate a relatively simple system , namely w@xmath4 . this ion has only one valence 4d- electron in its ground state . it follows from @xcite that only 11 excited levels from the possible 48 ones for the configurations under consideration have been determined experimentally . on the other hand , a theoretical study of energy level spectra was presented in @xcite . however , this work suffers from the fact that the data are provided just for some fifty percent of existing levels representing the configurations under consideration . furthermore , the transition data were determined for the highly excited levels only . in present work we employ the quasirelativistic method @xcite developed specifically for the _ ab initio _ calculation of spectral parameters of heavy atoms and highly charged ions . a preliminary analysis of the suitability of this method for the tungsten ions with open 4d- shell was presented in @xcite . it was demonstrated that our quasirelativistic approach is appropriate for the inclusion of the relativistic effects , and the further improvement of accuracy of the calculated data can be achieved with inclusion of the correlation effects . in section [ calc ] we provide a short description of our calculation method . the more detailed description of the applied approach and the analysis of the convergence within the configuration interaction ( ci ) approximation applied for the inclusion of correlation effects is presented in @xcite . obtained results are discussed in section [ result ] . [ calc ] [ qrhf ] we use a quasirelativistic approximation for _ ab initio _ calculations of ion energy spectra . this approach significantly differs from widely used method described in @xcite . in our method , the radial orbitals ( ro ) are obtained by solving the quasirelativistic equations of the following form : @xmath5 the first line of eq.([eq - qrhf ] ) represents the traditional hartree - fock equations , where @xmath6 denotes the exchange part of the potential and @xmath7 represents the direct part of the potential including the interaction of an electron with nucleus @xmath8 and with other electrons . we take into account the finite size of a nucleus within the nucleus potential @xmath8 @xcite . this allow us to represent the radial orbitals in powers of a radial variable in the nucleus region . next two terms with the multiplier @xmath9 describe the relativistic correction of the mass - velocity dependence . the last term of eq.([eq - qrhf ] ) represents the potential of the electron contact interaction with the nucleus . in our approach we include the contact interaction with the nucleus not only for the s- electrons but also some part of it for the p- electrons : @xmath10 a detailed discussion of the particular features of eq.([eq - qrhf ] ) is given in @xcite , whereas their solution techniques are described in @xcite . concluding the description of applied approximation , we want to articulate the unique features of our quasirelativistic method which significantly differs from widely used approach described in @xcite . the main differences arise from our adopted set of hartree - fock quasirelativistic equations ( qrhf ) featuring several distinctive properties , namely : 1 . no statistical potentials are used . there are only conventional self - consistent field direct and exchange potentials in qrhf . the finite size of the nucleus is taken into account to determine the potential while solving the quasirelativistic equations @xcite . the mass - velocity term is divided into two parts - the direct potential and the exchange one . the contact interaction term contains only the nucleus potential derivative in the numerator . there are no two - electron potentials in the numerator . only the direct part of the potential is included into the denominator of the contact interaction with a nucleus term . 6 . the contact interaction with a nucleus is taken into account not only for s - electrons , but also for p - electrons with some additional corrections made @xcite . one must look in @xcite for the most complete description of the way to determine qrhf equations employed in this work . the methods to calculate the energy level spectra were discussed extensively in @xcite . for the energy level spectra calculation , we include all two - electron interactions in the same way as it is done in conventional breit - pauli approximation . this similarity makes it possible to apply widely used code mchf breit @xcite for the angular integration of the breit - pauli hamiltonian matrix elements . we employ the computer program mchf mltpol @xcite to determine the matrix elements of transition operators along with the code mchf lsjtr @xcite which has been adopted for use with the quasirelativistic radial orbitals . [ tro ] we use configuration interaction ( ci ) approximation to include electron correlation effects . in present work , depending on the type of one - electron state , two sorts of radial orbitals are exploited in the ci expansion of wavefunctions . for the configurations under consideration ( adjusted configurations ) , we employ one - electron radial orbitals @xmath11 determined from solutions of eq.([eq - qrhf ] ) . this type of ro was applied for all electrons with @xmath12 and @xmath13 . we adopt the transformed radial orbitals ( tro ) to describe virtually excited electrons included in ci expansion . the method of transformed radial orbitals to account for the configuration interaction effects was developed in @xcite and successfully applied for the non - relativistic radial orbitals ( see , e.g. @xcite ) . the comparison of tro properties with those of the solutions of the multiconfigurational hartree - fock - jucys equations @xcite demonstrates their close similarity @xcite . such a way to include correlation effects within our quasirelativistic approach was successfully applied for the studies of the tungsten spectral parameters in @xcite . in current work the tro were employed to describe the admixed configurations , having virtually excited electrons with the principal quantum number @xmath14 and with all allowed values of orbital quantum number @xmath15 . these orbitals are determined from the ro of the configurations under consideration by ensuring the effective inclusion of the correlation effects . for the present calculations , we employ the form of tro with two free variational parameters @xmath16 and @xmath17 : @xmath18 here the multiplier @xmath19 ensures the normalization of determined tro , the first term in parenthesis performs the necessary transformation , and the second term provides the orthogonality of all employed radial orbitals . the parameters @xmath16 and @xmath17 are varied in order to ensure the maximum of averaged energy correction @xmath20 to the energy of the adjusted configuration in the second order of perturbation theory : @xmath21 we apply the same set of radial orbitals to describe both the even and odd configurations . this approach enables us to avoid any problems caused by the non - orthogonality of ro in calculation of the electron transition parameters . as a consequence of adopted tro , the basis of radial orbitals and the number of possible admixed configurations increases rapidly . therefore one needs to perform a selection of the admixed configurations by including only those which have the greatest influence on the configurations being adjusted . as a selection criteria we apply the mean weight of the admixed configuration within the ci expansion of the adjusted configuration wavefunction determined in the second order of perturbation theory : @xmath22 we include only those configurations which have @xmath23 larger than the specified small parameter @xmath24 in our calculation of energy levels and transition parameters . the methods and computer codes to calculate the mean weight from eq . ( [ eq - w ] ) have been described in @xcite . such a way to select the interacting configurations in the non - relativistic approximation was sucessfully exploited ( see , e.g. , earlier mentioned @xcite and many others ) . the configuration state functions ( csfs ) of the selected configurations are used to form the energy operator matrices . by diagonalizing these hamiltonian matrices , we can determine the energy eigenvalues and eigenfunctions for the investigated configurations and exploit them for calculation of transition parameters . more details of employed approach one can find in @xcite and references therein , whereas some examples of application of our method are presented in @xcite . in present work we have selected only those admixed configurations which have the mean weights in the wave functions of the configurations under consideration larger than @xmath25 . this procedure has allowed us to reduce the number of admixed configurations in the ci expansion of the wavefunctions to 102 for even configurations and to 217 for odd configurations . so here we use an approximation denoted as * f * in @xcite . it was demonstrated in @xcite that any further expansion of the admixed configurations basis does not lead to significant improvement in the calculated energy level spectra accuracy . we have determined energy eigenvalues and eigenfunctions by diagonalizing hamiltonian matrices obtained in the @xmath26-coupling scheme in breit - pauli approximation . using these results , we have calculated the transition parameters . we have computed data for the electric dipole ( e1 ) , electric quadrupole ( e2 ) and magnetic dipole ( m1 ) operators for the transitions both between the configurations under consideration and within these configurations . the emission transition rates of all above mentioned types were used to determine the radiative lifetimes of excited states . [ result ] we present the energy level spectra with the percentage contribution of eigenfunctions together with lande factors @xmath0 and the radiative lifetimes @xmath27 for the investigated configurations in table [ tab1 ] . it was demonstrated in @xcite that deviations of our theoretical level energy values from the experimental spectra data do not exceed @xmath28 in most cases . consequently , we can conclude that these discrepancies in transition energies do not affect significantly the accuracy of calculated values of the transition rates or oscillator strengths . in table [ taben ] we compare our calculated energy levels with data obtained from the completely relativistic calculations relac @xcite and with available experimental data @xcite presented in compilation @xcite . this evaluation favorably demonstrates a better agreement of our data with the experimental level energies , especially for highly - excited states , comparing to the data from relativistic calculations relac . this is caused mainly by considerably improved inclusion of correlation effects . such a good agreement with experimental data leads to conclusion that our quasirelativistic approach adequately accounts for the relativistic effects even for highly - ionized heavy atoms . it is worth to mention a very good agreement of theoretical fine - structure level splitting for the ground term 4p@xmath294d @xmath30d pointing to close spin - orbit interaction values obtained in our work and in fully relativistic calculations @xcite . as it was mentioned before , all our calculations were performed in @xmath31-coupling scheme . it is well established fact that the spin - orbit interaction induces a very strong mixing of @xmath31-terms for highly - charged ions . therefore in table [ taben ] we present only the level numbers ( indices ) @xmath19 taken from table [ tab1 ] where one can find a corresponding ci wavefunction expansion . it is easy to notice from the csfs contributions presented in table [ tab1 ] that the strong mixing occurs both for the @xmath31-terms of the same configuration and for terms of different configurations . a well - established case of a strong interaction between the @xmath32 and @xmath33 configurations ( see @xcite ) can be confirmed from the analysis of the wave function percentage contributions . for example , a contribution from the @xmath34 configuration for the @xmath35 level @xmath36 in table [ tab1 ] is merely @xmath37 . moreover , the largest contribution for the level @xmath38 comes from the @xmath2 configuration although the level itself clearly must be attributed to the configuration @xmath3 . in table [ tab2 ] we present the transition wavelengths @xmath39 , the emission transition rates @xmath40 and the weighted oscillator strength values @xmath41 . we use only the level indices @xmath19 and the total angular momentum @xmath42 from table [ tab1 ] to describe both the initial and final levels along with the type of transition operator ( e1 , e2 or m1 ) applied to calculate the transition parameters . we present the transitions from the excited level to lower - lying levels in decreasing order of their rates . we limit the number of presented transitions by removing those having the transition rate values by 3 orders of magnitude lower comparing to the strongest transition from the particular energy level . this condition would remove completely the e2 transitions . nonetheless , we have placed some most important e2 transitions in table [ tab2 ] . the magnetic dipole m1 transitions within configuration @xmath1 were computed using a completely relativistic approximation in @xcite . as one can see from table [ tab2 ] , their value of transition rate @xmath43 is very close to the value of @xmath44 from our calculation . from analysis of data presented in table [ tab2 ] , we can notice that the electric dipole transition rates are the most strong ones for all excited configuration levels when the total angular momentum @xmath45 . furthermore , the strong e1 transitions define the values of radiative lifetimes @xmath27 for these levels presented in table [ tab1 ] . the only exemption from this rule demonstrates the level @xmath46 with total angular momentum @xmath47 . even if the e1 transitions from this level have the largest rates @xmath40 , the values of two m1 transition rates are of the same order of magnitude as for the e1 transitions . this causes the significant decrease of the radiative lifetime value @xmath27 for this energy level . the electric dipole transitions from the excited levels with @xmath48 are forbidden because the largest value of total angular momentum of the ground state @xmath35 . consequently , the radiative lifetimes @xmath27 for these levels are determined by the magnetic dipole transitions since the e2 transition rates are significantly smaller . this work , partially supported by the european communities under the contract of association between euratom / lei fu07-ct-2007 - 00063 , was carried out within the framework of the european fusion development agreement . 99 j. reader , phys . scripta t134 ( 2009 ) 014023 . c. skinner , phys . scripta t134 ( 2009 ) 014022 . h. p. summers , m. g. omullane , am . proc . 1344 ( 2011 ) 179 . c. biedermann , r. radtke , r. seidel , t. p " utterich , phys . scripta t134 ( 2009 ) 014026 . r. hutton , y. zou , m. anderssin , t. brage , i. martinson , j. phys . b : at . mol . 43 ( 2010 ) 144026 . j. clementson , p. beiersdorfer , e. w. magee , h. s. mclean , r. d. wood , j. phys . b : at . mol . opt . phys . 43 ( 2010 ) 144009 . s. b. utter , p. beiersdorfer , e. trabert , can . 80 ( 2002 ) 1503 . a. e. kramida , t. shirai , at . data nucl . data tables 95 ( 2009 ) 305 . fournier , at . data nucl . data tables 68 ( 1998 ) 1 . p. bogdanovich , o. rancova , phys . a 74 ( 2006 ) 052501 . bogdanovich , o. rancova , phys . a 76 ( 2007 ) 012507 . p. bogdanovich , o. rancova , phys . scripta 78 ( 2008 ) 045301 . p. bogdanovich , v. jonauskas , r. karpukien . e , o. rancova , nuclear instr . meth . a 619 ( 2010 ) 15 . p. bogdanovich , o. rancova , a. tikonas , phys . scripta 83 ( 2011 ) 065302 . r. d. cowan , _ the theory of atomic structure and spectra _ university of california press , los angeles , 1981 p. bogdanovich , o. rancova , lithuanian j. phys . 42 ( 2002 ) 257 . p. bogdanovich , o. rancova , lithuanian j. phys . 43 ( 2003 ) 177 . p. bogdanovich , v. jonauskas , o. rancova , nuclear instr . b 235 ( 2005 ) 145 . a. hibbert , r. glass , c. f. fischer , comput . phys . commun . 64 ( 1991 ) 445 . c. f. fischer , m. r. godefroid , a. hibbert , comput . 64 ( 1991 ) 486 . c. f. fischer , m. r. godefroid , comput . 64 ( 1991 ) 501 . p. bogdanovich , r. karpukien . e , lithuanian j. phys . 39 ( 1999 ) 193 . p. bogdanovich , r. karpukien . e , i. martinson , phys . scripta 67 ( 2003 ) 44 . r. karpukien . e , p. bogdanovich , j. phys . b : at . mol . 36 ( 2003 ) 2145 . p. bogdanovich , r. karpukien . e , a. udris , j. phys . b : at . mol . opt . phys . 37 ( 2004 ) 2067 . p. bogdanovich , r. karpukien . data nucl . data tables 95 ( 2009 ) 533 . a. yutsis , jetp 23 , 2 ( 1952 ) 129 ( in russian ) . p. bogdanovich , lithuanian j. phys . 44 ( 2004 ) 135 . p. bogdanovich , nuclear instr . b 235 ( 2005 ) 92 . o. rancova , p. bogdanovich , r. karpukien . e , j. phys . series 163 ( 2009 ) 012011 . p. bogdanovich , r. karpukien . e , comput . commun . 134 ( 2001 ) 321 . p. bogdanovich , r. karpukien . e , a. momkauskait . e , comput . commun . 172 ( 2005 ) 13 . r. karpukien . e , o. rancova , p. bogdanovich , j. phys . b : at . mol . opt . 43 ( 2010 ) 085002 . v. jonauskas , r. kisielius , s. kuas , a. kynien . a 81 ( 2010 ) 012506 . v. jonauskas , s. kuas , r. karazija , j. phys . b : at . mol . 40 ( 2007 ) 2179 . llrrrrr + @xmath19 & term @xcite & level name @xcite & @xmath42 & exp @xcite & qrhf & relac @xcite + 1 & 4p@xmath294d @xmath30d & ( 4p)@xmath29(4d@xmath49)@xmath50 & 1.5 & 0 & 0 & 0 + 2 & 4p@xmath294d @xmath30d & ( 4p)@xmath29(4d@xmath51)@xmath50 & 2.5 & 1546 & 1538 & 1537 + 3 & 4p@xmath524d@xmath30 @xmath53f ( 3/2,2 ) & & 0.5 & 12276 & 12311 & + 4 & 4p@xmath524d@xmath30 @xmath53f ( 3/2,2 ) & & 1.5 & 12276 & 12312 & + 5 & & ( 4p@xmath49)@xmath30(4p@xmath51)@xmath53(4d@xmath49)@xmath30 & 2.5 & & 12591 & 12569 + 6 & & ( 4p@xmath49)@xmath30(4p@xmath51)@xmath53(4d@xmath49)@xmath30 & 3.5 & & 12752 & 12720 + 10 & & ( 4p@xmath49)@xmath30(4p@xmath51)@xmath53(4d@xmath49)@xmath50(4d@xmath51)@xmath50 & 4.5 & & 13922 & 13861 + 13 & & ( 4p@xmath49)@xmath30(4p@xmath51)@xmath53(4d@xmath49)@xmath50(4d@xmath51)@xmath50 & 5.5 & & 14255 & 14201 + 16 & & ( 4p@xmath49)@xmath30(4p@xmath51)@xmath53(4d@xmath49)@xmath50(4d@xmath51)@xmath50 & 3.5 & & 14631 & 14646 + 17 & & ( 4p@xmath49)@xmath30(4p@xmath51)@xmath53(4d@xmath49)@xmath50(4d@xmath51)@xmath50 & 2.5 & & 14680 & 14691 + 18 & & ( 4p@xmath49)@xmath30(4p@xmath51)@xmath53(4d@xmath49)@xmath50(4d@xmath51)@xmath50 & 1.5 & & 14756 & 14794 + 19 & & ( 4p@xmath49)@xmath30(4p@xmath51)@xmath53(4d@xmath49)@xmath50(4d@xmath51)@xmath50 & 4.5 & & 14853 & 14861 + 20 & 4p@xmath524d@xmath30 @xmath50 g ( 3/2,4 ) & ( 4p@xmath49)@xmath30(4p@xmath51)@xmath53(4d@xmath49)@xmath50(4d@xmath51)@xmath50 & 2.5 & 15085 & 15220 & 15220 + 21 & & ( 4p@xmath49)@xmath30(4p@xmath51)@xmath53(4d@xmath51)@xmath30 & 5.5 & & 15500 & 15443 + 22 & 4p@xmath524d@xmath30 @xmath53f ( 3/2,3 ) & ( 4p@xmath49)@xmath30(4p@xmath51)@xmath53(4d@xmath49)@xmath50(4d@xmath51)@xmath50 & 1.5 & 15426 & 15618 & 15680 + 24 & & ( 4p@xmath49)@xmath30(4p@xmath51)@xmath53(4d@xmath51)@xmath30 & 3.5 & & 15671 & 15649 + 25 & & ( 4p@xmath49)@xmath30(4p@xmath51)@xmath53(4d@xmath51)@xmath30 & 4.5 & & 15854 & 15851 + 28 & & ( 4p@xmath49)@xmath30(4p@xmath51)@xmath53(4d@xmath51)@xmath30 & 3.5 & & 16561 & 16567 + 30 & 4p@xmath524d@xmath30 @xmath53f ( 3/2,4 ) & ( 4p@xmath49)@xmath30(4p@xmath51)@xmath53(4d@xmath51)@xmath30 & 2.5 & 17315 & 17535 & 17630 + 32 & 4p@xmath294f @xmath30f & ( 4p)@xmath29(4f@xmath49)@xmath50 & 2.5 & 17581 & 17752 & 17840 + 33 & 4p@xmath294f @xmath30f & ( 4p)@xmath29(4f@xmath51)@xmath50 & 3.5 & 17696 & 17889 & 17951 + 34 & 4p@xmath524d@xmath30 @xmath53f ( 1/2,2 ) & ( 4p@xmath49)@xmath50(4p@xmath51)@xmath54(4d@xmath49)@xmath30 & 2.5 & 20145 & 20319 & 20396 + 44 & 4p@xmath524d@xmath30 @xmath50d ( 1/2,2 ) & & 1.5 & 21843 & 22069 & + 46 & 4p@xmath524d@xmath30 @xmath50 g ( 1/2,4 ) & ( 4p@xmath49)@xmath50(4p@xmath51)@xmath54(4d@xmath49)@xmath50(4d@xmath51)@xmath50 & 3.5 & 23255 & 23274 & 23488 + 47 & & ( 4p@xmath49)@xmath50(4p@xmath51)@xmath54(4d@xmath49)@xmath50(4d@xmath51)@xmath50 & 1.5 & & 23309 & 23525 + 48 & & ( 4p@xmath49)@xmath50(4p@xmath51)@xmath54(4d@xmath49)@xmath50(4d@xmath51)@xmath50 & 2.5 & & 23369 & 23347 + + throughout the table we present all the eigenfunction components which have their percentage contributions higher than @xmath58 @xmath19 & & & & & + + + 1 & 0 & 0.80 & & 1.5 & 98 4d @xmath30d & & & & + 2 & 1538 & 1.20 & 2.55(@xmath594 ) & 2.5 & 98 4d @xmath30d & & & & + 3 & 12311 & 1.00 & 8.44(@xmath601 ) & 1.5 & 42 4p@xmath524d@xmath30(@xmath53f@xmath61 ) @xmath54d & + 23 4p@xmath524d@xmath30(@xmath53f@xmath61 ) @xmath54f & + 10 4p@xmath524d@xmath30(@xmath50d@xmath61 ) @xmath30d & & + 4 & 12312 & 0.12 & 2.44(@xmath601 ) & 0.5 & 84 4p@xmath524d@xmath30(@xmath53f@xmath61 ) @xmath54d & + 11 4p@xmath524d@xmath30(@xmath50d@xmath61 ) @xmath30p & & & + 5 & 12591 & 1.03 & 3.83(@xmath601 ) & 2.5 & 36 4p@xmath524d@xmath30(@xmath53f@xmath61 ) @xmath54f & + 18 4p@xmath524d@xmath30(@xmath50d@xmath61 ) @xmath30d & + 12 4p@xmath524d@xmath30(@xmath53f@xmath61 ) @xmath54 g & + 10 4p@xmath524d@xmath30(@xmath53f@xmath30 ) @xmath54d & + 6 & 12752 & 1.03 & 1.75(@xmath591 ) & 3.5 & 32 4p@xmath524d@xmath30(@xmath53f@xmath61 ) @xmath54 g & + 31 4p@xmath524d@xmath30(@xmath53f@xmath61 ) @xmath30 g & + 23 4p@xmath524d@xmath30(@xmath50d@xmath61 ) @xmath30f & + 7 4p@xmath524d@xmath30(@xmath53f@xmath61 ) @xmath54f & + 7 & 13284 & 1.36 & 5.37(@xmath590 ) & 1.5 & 28 4p@xmath524d@xmath30(@xmath53f@xmath61 ) @xmath54p & + 21 4p@xmath524d@xmath30(@xmath50s@xmath62 ) @xmath30p & + 15 4p@xmath524d@xmath30(@xmath53p@xmath61 ) @xmath54d & & + 8 & 13636 & 1.30 & 1.95(@xmath591 ) & 2.5 & 32 4p@xmath524d@xmath30(@xmath53f@xmath61 ) @xmath54d & + 22 4p@xmath524d@xmath30(@xmath53p@xmath61 ) @xmath54p & + 15 4p@xmath524d@xmath30(@xmath53p@xmath61 ) @xmath54d & & + 9 & 13915 & 1.29 & 2.88(@xmath601 ) & 3.5 & 35 4p@xmath524d@xmath30(@xmath53f@xmath61 ) @xmath54f & + 27 4p@xmath524d@xmath30(@xmath53p@xmath61 ) @xmath54d & + 16 4p@xmath524d@xmath30(@xmath53f@xmath61 ) @xmath54d & + 10 4p@xmath524d@xmath30(@xmath50d@xmath61 ) @xmath30f & + 10 & 13922 & 1.18 & 3.04(@xmath594 ) & 4.5 & 70 4p@xmath524d@xmath30(@xmath53f@xmath61 ) @xmath54 g & + 15 4p@xmath524d@xmath30(@xmath53f@xmath61 ) @xmath30 g & + 13 4p@xmath524d@xmath30(@xmath53f@xmath61 ) @xmath54f & & + 11 & 14052 & 1.09 & 1.29(@xmath601 ) & 1.5 & 40 4p@xmath524d@xmath30(@xmath53f@xmath61 ) @xmath54d & + 14 4p@xmath524d@xmath30(@xmath53f@xmath61 ) @xmath30d & + 14 4p@xmath524d@xmath30(@xmath50d@xmath61 ) @xmath30d & + 11 4p@xmath524d@xmath30(@xmath50d@xmath61 ) @xmath30p & + 12 & 14104 & 1.23 & 4.29(@xmath601 ) & 2.5 & 40 4p@xmath524d@xmath30(@xmath53p@xmath61 ) @xmath54p & + 17 4p@xmath524d@xmath30(@xmath53f@xmath61 ) @xmath30f & + 15 4p@xmath524d@xmath30(@xmath50g@xmath61 ) @xmath30f & & + 13 & 14255 & 1.17 & 2.68(@xmath596 ) & 5.5 & 57 4p@xmath524d@xmath30(@xmath50g@xmath61 ) @xmath30h & + 42 4p@xmath524d@xmath30(@xmath53f@xmath61 ) @xmath54 g & & & + 14 & 14290 & 1.06 & 1.15(@xmath594 ) & 3.5 & 29 4p@xmath524d@xmath30(@xmath53f@xmath61 ) @xmath30 g & + 15 4p@xmath524d@xmath30(@xmath53f@xmath61 ) @xmath30f & + 15 4p@xmath524d@xmath30(@xmath53f@xmath61 ) @xmath54f & + 15 4p@xmath524d@xmath30(@xmath50g@xmath61 ) @xmath30 g & + 14 4p@xmath524d@xmath30(@xmath50g@xmath61 ) @xmath30f + 15 & 14408 & 1.66 & 4.37(@xmath592 ) & 0.5 & 33 4p@xmath524d@xmath30(@xmath53p@xmath61 ) @xmath54p & + 29 4p@xmath524d@xmath30(@xmath53p@xmath61 ) @xmath30s & + 19 4p@xmath524d@xmath30(@xmath50d@xmath61 ) @xmath30p & & + 16 & 14631 & 1.17 & 3.20(@xmath591 ) & 3.5 & 20 4p@xmath524d@xmath30(@xmath53p@xmath61 ) @xmath54d & + 13 4p@xmath524d@xmath30(@xmath50d@xmath61 ) @xmath30f & + 13 4p@xmath524d@xmath30(@xmath50g@xmath61 ) @xmath30 g & + 12 4p@xmath524d@xmath30(@xmath50g@xmath61 ) @xmath30f & + 11 4p@xmath524d@xmath30(@xmath53f@xmath61 ) @xmath54d + 17 & 14680 & 1.22 & 1.60(@xmath601 ) & 2.5 & 36 4p@xmath524d@xmath30(@xmath53p@xmath61 ) @xmath54d & + 25 4p@xmath524d@xmath30(@xmath53p@xmath61 ) @xmath30d & + 15 4p@xmath524d@xmath30(@xmath50d@xmath61 ) @xmath30d & & + 18 & 14756 & 1.39 & 2.77(@xmath590 ) & 1.5 & 30 4p@xmath524d@xmath30(@xmath53p@xmath61 ) @xmath54s & + 25 4p@xmath524d@xmath30(@xmath53p@xmath61 ) @xmath30d & + 17 4p@xmath524d@xmath30(@xmath53p@xmath61 ) @xmath54p & & + 19 & 14853 & 1.13 & 1.47(@xmath595 ) & 4.5 & 38 4p@xmath524d@xmath30(@xmath50g@xmath61 ) @xmath30 g & + 24 4p@xmath524d@xmath30(@xmath53f@xmath61 ) @xmath54f & + 18 4p@xmath524d@xmath30(@xmath53f@xmath61 ) @xmath30 g & + 17 4p@xmath524d@xmath30(@xmath50g@xmath61 ) @xmath30h & + 20 & 15220 & 0.88 & 6.50(@xmath603 ) & 2.5 & 41 4p@xmath524d@xmath30(@xmath50g@xmath61 ) @xmath30f & + 28 4f @xmath30f & + 13 4p@xmath524d@xmath30(@xmath53f@xmath61 ) @xmath30f & & + 21 & 15500 & 1.20 & 1.70(@xmath594 ) & 5.5 & 57 4p@xmath524d@xmath30(@xmath53f@xmath61 ) @xmath54 g & + 42 4p@xmath524d@xmath30(@xmath50g@xmath61 ) @xmath30h & & & + 22 & 15618 & 0.85 & 2.16(@xmath603 ) & 1.5 & 32 4p@xmath524d@xmath30(@xmath53f@xmath61 ) @xmath54f & + 22 4p@xmath524d@xmath30(@xmath53p@xmath61 ) @xmath30d & + 12 4p@xmath524d@xmath30(@xmath53f@xmath61 ) @xmath30d & & + 23 & 15619 & 1.21 & 1.60(@xmath590 ) & 2.5 & 24 4p@xmath524d@xmath30(@xmath50d@xmath61 ) @xmath30d & + 18 4p@xmath524d@xmath30(@xmath53p@xmath61 ) @xmath54p & + 15 4p@xmath524d@xmath30(@xmath53f@xmath61 ) @xmath30d & + 13 4p@xmath524d@xmath30(@xmath53p@xmath61 ) @xmath30d & + 24 & 15671 & 1.22 & 5.32(@xmath602 ) & 3.5 & 34 4p@xmath524d@xmath30(@xmath50d@xmath61 ) @xmath30f & + 29 4p@xmath524d@xmath30(@xmath53p@xmath61 ) @xmath54d & + 13 4f@xmath30f & & + 25 & 15854 & 1.17 & 1.20(@xmath594 ) & 4.5 & 38 4p@xmath524d@xmath30(@xmath53f@xmath61 ) @xmath54f & + 30 4p@xmath524d@xmath30(@xmath50g@xmath61 ) @xmath30 g & + 18 4p@xmath524d@xmath30(@xmath53f@xmath61 ) @xmath30 g & + 13 4p@xmath524d@xmath30(@xmath50g@xmath61 ) @xmath30h & + 26 & 15972 & 1.08 & 6.58(@xmath603 ) & 0.5 & 65 4p@xmath524d@xmath30(@xmath50d@xmath61 ) @xmath30p & + 14 4p@xmath524d@xmath30(@xmath53p@xmath61 ) @xmath54p & + 12 4p@xmath524d@xmath30(@xmath53p@xmath61 ) @xmath30s & & + 27 & 16350 & 0.76 & 1.49(@xmath603 ) & 0.5 & 52 4p@xmath524d@xmath30(@xmath53p@xmath61 ) @xmath30p & + 24 4p@xmath524d@xmath30(@xmath53p@xmath61 ) @xmath54d & + 18 4p@xmath524d@xmath30(@xmath53p@xmath61 ) @xmath30s & & + 28 & 16551 & 1.18 & 4.48(@xmath602 ) & 3.5 & 17 4f @xmath30f & + 16 4p@xmath524d@xmath30(@xmath53f@xmath61 ) @xmath30f & + 15 4p@xmath524d@xmath30(@xmath50d@xmath61 ) @xmath30f & + 14 4p@xmath524d@xmath30(@xmath53f@xmath61 ) @xmath54d & + 12 4p@xmath524d@xmath30(@xmath53p@xmath61 ) @xmath54d + 29 & 16897 & 1.27 & 5.82(@xmath601 ) & 1.5 & 50 4p@xmath524d@xmath30(@xmath50s@xmath62 ) @xmath30p & + 14 4p@xmath524d@xmath30(@xmath53p@xmath61 ) @xmath30d & & & + 30 & 17535 & 1.09 & 1.02(@xmath603 ) & 2.5 & 29 4p@xmath524d@xmath30(@xmath53f@xmath61 ) @xmath30d & + 15 4p@xmath524d@xmath30(@xmath53f@xmath61 ) @xmath54f & + 13 4f @xmath30f & & + 31 & 17553 & 1.36 & 1.58(@xmath603 ) & 1.5 & 20 4p@xmath524d@xmath30(@xmath50s@xmath62 ) @xmath30p & + 17 4p@xmath524d@xmath30(@xmath53p@xmath61 ) @xmath54s & + 16 4p@xmath524d@xmath30(@xmath53p@xmath61 ) @xmath30p & + 13 4p@xmath524d@xmath30(@xmath50d@xmath61 ) @xmath30d & + 12 4p@xmath524d@xmath30(@xmath53p@xmath61 ) @xmath54d + 32 & 17752 & 0.93 & 1.21(@xmath603 ) & 2.5 & 49 4f @xmath30f & + 18 4p@xmath524d@xmath30(@xmath50g@xmath61 ) @xmath30f & & & + 33 & 17889 & 1.12 & 1.72(@xmath603 ) & 3.5 & 63 4f @xmath30f & + 18 4p@xmath524d@xmath30(@xmath50g@xmath61 ) @xmath30 g & & & + 34 & 20319 & 0.77 & 8.80(@xmath604 ) & 2.5 & 39 4p@xmath524d@xmath30(@xmath53f@xmath61 ) @xmath54 g & + 28 4p@xmath524d@xmath30(@xmath53f@xmath61 ) @xmath30f & + 15 4p@xmath524d@xmath30(@xmath50d@xmath61 ) @xmath30f & & + 35 & 21314 & 1.02 & 7.13(@xmath601 ) & 3.5 & 46 4p@xmath524d@xmath30(@xmath53f@xmath61 ) @xmath54 g & + 19 4p@xmath524d@xmath30(@xmath53f@xmath61 ) @xmath30 g & + 14 4p@xmath524d@xmath30(@xmath53f@xmath61 ) @xmath54f & & + 36 & 21387 & 1.07 & 1.70(@xmath602 ) & 2.5 & 18 4p@xmath524d@xmath30(@xmath53f@xmath61 ) @xmath54f & + 16 4p@xmath524d@xmath30(@xmath50d@xmath61 ) @xmath30f & + 14 4p@xmath524d@xmath30(@xmath53f@xmath61 ) @xmath54 g & + 14 4p@xmath524d@xmath30(@xmath53p@xmath61 ) @xmath30d & + 12 4p@xmath524d@xmath30(@xmath53f@xmath61 ) @xmath30d + 37 & 21478 & 1.07 & 9.10(@xmath603 ) & 1.5 & 37 4p@xmath524d@xmath30(@xmath53p@xmath61 ) @xmath54d & + 16 4p@xmath524d@xmath30(@xmath53f@xmath61 ) @xmath30d & & & + 38 & 21547 & 1.96 & 1.39(@xmath603 ) & 0.5 & 41 4p@xmath524d@xmath30(@xmath53p@xmath61 ) @xmath54p & + 33 4p@xmath524d@xmath30(@xmath53p@xmath61 ) @xmath30s & + 13 4p@xmath524d@xmath30(@xmath53p@xmath61 ) @xmath30p & & + 39 & 21555 & 0.43 & 9.08(@xmath604 ) & 0.5 & 56 4p@xmath524d@xmath30(@xmath53p@xmath61 ) @xmath54d & + 21 4p@xmath524d@xmath30(@xmath53p@xmath61 ) @xmath30p & + 13 4p@xmath524d@xmath30(@xmath50s@xmath62 ) @xmath30p & & + 40 & 21754 & 1.04 & 1.39(@xmath592 ) & 4.5 & 45 4p@xmath524d@xmath30(@xmath50g@xmath61 ) @xmath30h & + 19 4p@xmath524d@xmath30(@xmath50g@xmath61 ) @xmath30 g & + 16 4p@xmath524d@xmath30(@xmath53f@xmath61 ) @xmath30 g & & + 41 & 22069 & 0.88 & 3.03(@xmath604 ) & 1.5 & 43 4p@xmath524d@xmath30(@xmath53f@xmath61 ) @xmath30d & + 17 4p@xmath524d@xmath30(@xmath50g@xmath61 ) @xmath30d & + 12 4p@xmath524d@xmath30(@xmath53f@xmath61 ) @xmath54f & & + 42 & 22794 & 1.29 & 2.73(@xmath603 ) & 3.5 & 45 4p@xmath524d@xmath30(@xmath53f@xmath61 ) @xmath54d & + 32 4p@xmath524d@xmath30(@xmath53f@xmath61 ) @xmath30f & + 18 4p@xmath524d@xmath30(@xmath53f@xmath61 ) @xmath54f & & + 43 & 23116 & 1.15 & 3.88(@xmath604 ) & 2.5 & 34 4p@xmath524d@xmath30(@xmath53f@xmath61 ) @xmath30d & + 12 4p@xmath524d@xmath30(@xmath53p@xmath61 ) @xmath30d & + 10 4p@xmath524d@xmath30(@xmath53f@xmath61 ) @xmath54d & & + 44 & 23159 & 1.20 & 4.30(@xmath603 ) & 1.5 & 33 4p@xmath524d@xmath30(@xmath50d@xmath61 ) @xmath30p & + 26 4p@xmath524d@xmath30(@xmath50d@xmath61 ) @xmath30d & + 12 4p@xmath524d@xmath30(@xmath53p@xmath61 ) @xmath54d & + 11 4p@xmath524d@xmath30(@xmath53p@xmath61 ) @xmath54p & + 45 & 23221 & 1.11 & 1.30(@xmath592 ) & 4.5 & 31 4p@xmath524d@xmath30(@xmath53f@xmath61 ) @xmath30 g & + 23 4p@xmath524d@xmath30(@xmath50g@xmath61 ) @xmath30h & + 17 4p@xmath524d@xmath30(@xmath53f@xmath61 ) @xmath54 g & + 15 4p@xmath524d@xmath30(@xmath53f@xmath61 ) @xmath54f & + 11 4p@xmath524d@xmath30(@xmath50g@xmath61 ) @xmath30 g + 46 & 23274 & 1.06 & 5.17(@xmath604 ) & 3.5 & 53 4p@xmath524d@xmath30(@xmath50g@xmath61 ) @xmath30f & + 29 4p@xmath524d@xmath30(@xmath50g@xmath61 ) @xmath30 g & & & + 47 & 23309 & 1.48 & 4.69(@xmath604 ) & 1.5 & 52 4p@xmath524d@xmath30(@xmath53p@xmath61 ) @xmath30p & + 20 4p@xmath524d@xmath30(@xmath53p@xmath61 ) @xmath54s & + 11 4p@xmath524d@xmath30(@xmath53p@xmath61 ) @xmath54p & & + 48 & 23369 & 1.11 & 4.68(@xmath601 ) & 2.5 & 35 4p@xmath524d@xmath30(@xmath50d@xmath61 ) @xmath30f & + 24 4p@xmath524d@xmath30(@xmath53p@xmath61 ) @xmath30d & + 17 4p@xmath524d@xmath30(@xmath53p@xmath61 ) @xmath54d & & + 49 & 24058 & 0.67 & 3.13(@xmath602 ) & 0.5 & 75 4p@xmath524d@xmath30(@xmath50s@xmath62 ) @xmath30p & + 13 4p@xmath524d@xmath30(@xmath53p@xmath61 ) @xmath54d & & & + + @xmath63 & @xmath64 & @xmath65 & @xmath66 & & @xmath39 ( ) & & + + + 2.5 & 2 & 1.5 & 1 & m1 & 650.2 & 3.93e@xmath5904 & 1.49e@xmath6005 + & & 1.5 & 1 & e2 & 650.2 & 1.89e@xmath5901 & 7.19e@xmath6009 + 1.5 & 3 & 1.5 & 1 & e1 & 81.23 & 1.15e@xmath5909 & 4.56e@xmath6003 + & & 2.5 & 2 & e1 & 92.83 & 3.38e@xmath5907 & 1.75e@xmath6004 + 0.5 & 4 & 1.5 & 1 & e1 & 81.22 & 4.09e@xmath5909 & 8.09e@xmath6003 + 2.5 & 5 & 1.5 & 1 & e1 & 79.42 & 2.37e@xmath5909 & 1.34e@xmath6002 + & & 2.5 & 2 & e1 & 90.47 & 2.47e@xmath5908 & 1.82e@xmath6003 + 3.5 & 6 & 2.5 & 2 & e1 & 89.18 & 5.72e@xmath5907 & 5.45e@xmath6004 + 1.5 & 7 & 2.5 & 2 & e1 & 85.14 & 1.78e@xmath5908 & 7.74e@xmath6004 + & & 1.5 & 1 & e1 & 75.28 & 8.25e@xmath5906 & 2.80e@xmath6005 + 2.5 & 8 & 1.5 & 1 & e1 & 73.34 & 5.13e@xmath5907 & 2.48e@xmath6004 + & & 2.5 & 2 & e1 & 82.66 & 3.11e@xmath5904 & 1.91e@xmath6007 + 3.5 & 9 & 2.5 & 2 & e1 & 80.80 & 3.47e@xmath5909 & 2.72e@xmath6002 + 4.5 & 10 & 3.5 & 6 & m1 & 854.6 & 3.29e@xmath5904 & 3.60e@xmath6005 + & & 3.5 & 6 & e2 & 854.6 & 2.13e@xmath5900 & 2.33e@xmath6009 + 1.5 & 11 & 1.5 & 1 & e1 & 71.17 & 5.76e@xmath5909 & 1.75e@xmath6002 + & & 2.5 & 2 & e1 & 79.91 & 1.97e@xmath5909 & 7.56e@xmath6003 + 2.5 & 12 & 2.5 & 2 & e1 & 79.58 & 2.31e@xmath5909 & 1.31e@xmath6002 + & & 1.5 & 1 & e1 & 70.90 & 2.44e@xmath5907 & 1.11e@xmath6004 + 5.5 & 13 & 4.5 & 10 & m1 & 2999 & 3.68e@xmath5902 & 5.95e@xmath6006 + & & 3.5 & 6 & e2 & 665.1 & 5.94e@xmath5900 & 4.73e@xmath6009 + 3.5 & 14 & 2.5 & 2 & e1 & 78.42 & 4.94e@xmath5904 & 3.65e@xmath6007 + & & 2.5 & 5 & m1 & 588.5 & 2.32e@xmath5904 & 9.64e@xmath6006 + & & 3.5 & 6 & m1 & 650.0 & 1.42e@xmath5904 & 7.21e@xmath6006 + & & 2.5 & 8 & m1 & 1528 & 2.66e@xmath5902 & 7.44e@xmath6007 + 0.5 & 15 & 1.5 & 1 & e1 & 69.41 & 2.25e@xmath5906 & 3.25e@xmath6006 + 3.5 & 16 & 2.5 & 2 & e1 & 76.38 & 3.12e@xmath5907 & 2.18e@xmath6004 + 2.5 & 17 & 1.5 & 1 & e1 & 68.12 & 6.25e@xmath5909 & 2.61e@xmath6002 + & & 2.5 & 2 & e1 & 76.09 & 2.40e@xmath5904 & 1.25e@xmath6007 + 1.5 & 18 & 2.5 & 2 & e1 & 75.65 & 2.47e@xmath5908 & 8.49e@xmath6004 + & & 1.5 & 1 & e1 & 67.77 & 1.13e@xmath5908 & 3.12e@xmath6004 + 4.5 & 19 & 3.5 & 6 & m1 & 475.9 & 3.48e@xmath5903 & 1.18e@xmath6006 + & & 3.5 & 9 & m1 & 1066 & 2.55e@xmath5903 & 4.34e@xmath6006 + & & 4.5 & 10 & m1 & 1074 & 4.18e@xmath5902 & 7.23e@xmath6007 + & & 3.5 & 14 & m1 & 1777 & 1.63e@xmath5902 & 7.72e@xmath6007 + & & 5.5 & 13 & m1 & 1673 & 1.07e@xmath5902 & 4.49e@xmath6007 + & & 3.5 & 16 & m1 & 4510 & 4.65e@xmath5901 & 1.42e@xmath6006 + 2.5 & 20 & 1.5 & 1 & e1 & 65.70 & 1.54e@xmath5911 & 5.96e@xmath6001 + & & 2.5 & 2 & e1 & 73.09 & 1.94e@xmath5908 & 9.32e@xmath6004 + 5.5 & 21 & 4.5 & 10 & m1 & 633.7 & 4.30e@xmath5904 & 3.11e@xmath6005 + & & 5.5 & 13 & m1 & 803.5 & 1.50e@xmath5904 & 1.74e@xmath6005 + & & 4.5 & 19 & m1 & 1546 & 5.80e@xmath5902 & 2.49e@xmath6006 + 1.5 & 22 & 1.5 & 1 & e1 & 64.03 & 4.61e@xmath5911 & 1.13e@xmath5900 + & & 2.5 & 2 & e1 & 71.02 & 1.64e@xmath5909 & 4.97e@xmath6003 + 2.5 & 23 & 2.5 & 2 & e1 & 71.02 & 4.59e@xmath5908 & 2.08e@xmath6003 + & & 1.5 & 1 & e1 & 64.02 & 1.66e@xmath5908 & 6.13e@xmath6004 + 3.5 & 24 & 2.5 & 2 & e1 & 70.76 & 1.88e@xmath5910 & 1.13e@xmath6001 + 4.5 & 25 & 3.5 & 14 & m1 & 639.4 & 3.06e@xmath5904 & 1.87e@xmath6005 + & & 3.5 & 9 & m1 & 515.6 & 2.63e@xmath5904 & 1.05e@xmath6005 + & & 4.5 & 10 & m1 & 517.5 & 1.50e@xmath5904 & 6.01e@xmath6006 + & & 4.5 & 19 & m1 & 998.8 & 6.50e@xmath5903 & 9.73e@xmath6006 + & & 5.5 & 13 & m1 & 625.5 & 3.67e@xmath5903 & 2.15e@xmath6006 + & & 3.5 & 6 & m1 & 322.3 & 7.60e@xmath5902 & 1.18e@xmath6007 + & & 3.5 & 16 & m1 & 817.7 & 5.13e@xmath5902 & 5.14e@xmath6007 + 0.5 & 26 & 1.5 & 1 & e1 & 62.61 & 1.52e@xmath5911 & 1.79e@xmath6001 + 0.5 & 27 & 1.5 & 1 & e1 & 61.16 & 6.71e@xmath5911 & 7.53e@xmath6001 + 3.5 & 28 & 2.5 & 2 & e1 & 66.61 & 2.23e@xmath5910 & 1.19e@xmath6001 + 1.5 & 29 & 2.5 & 2 & e1 & 65.11 & 1.23e@xmath5909 & 3.13e@xmath6003 + & & 1.5 & 1 & e1 & 59.18 & 4.87e@xmath5908 & 1.02e@xmath6003 + 2.5 & 30 & 1.5 & 1 & e1 & 57.03 & 5.33e@xmath5911 & 1.56e@xmath5900 + & & 2.5 & 2 & e1 & 62.51 & 4.46e@xmath5911 & 1.57e@xmath5900 + 1.5 & 31 & 2.5 & 2 & e1 & 62.44 & 6.34e@xmath5911 & 1.48e@xmath5900 + & & 1.5 & 1 & e1 & 56.97 & 1.14e@xmath5908 & 2.21e@xmath6004 + 2.5 & 32 & 1.5 & 1 & e1 & 56.33 & 5.39e@xmath5911 & 1.54e@xmath5900 + & & 2.5 & 2 & e1 & 61.67 & 2.89e@xmath5911 & 9.89e@xmath6001 + 3.5 & 33 & 2.5 & 2 & e1 & 61.16 & 5.81e@xmath5911 & 2.61e@xmath5900 + 2.5 & 34 & 1.5 & 1 & e1 & 49.22 & 1.14e@xmath5912 & 2.47e@xmath5900 + & & 2.5 & 2 & e1 & 53.25 & 7.48e@xmath5908 & 1.91e@xmath6003 + 3.5 & 35 & 2.5 & 2 & e1 & 50.57 & 1.40e@xmath5909 & 4.28e@xmath6003 + 2.5 & 36 & 2.5 & 2 & e1 & 50.38 & 5.28e@xmath5910 & 1.20e@xmath6001 + & & 1.5 & 1 & e1 & 46.76 & 6.11e@xmath5909 & 1.20e@xmath6002 + 1.5 & 37 & 1.5 & 1 & e1 & 46.56 & 9.49e@xmath5910 & 1.23e@xmath6001 + & & 2.5 & 2 & e1 & 50.15 & 1.50e@xmath5910 & 2.26e@xmath6002 + 0.5 & 38 & 1.5 & 1 & e1 & 46.41 & 7.20e@xmath5911 & 4.65e@xmath6001 + 0.5 & 39 & 1.5 & 1 & e1 & 46.39 & 1.10e@xmath5912 & 7.11e@xmath6001 + 4.5 & 40 & 5.5 & 13 & m1 & 133.4 & 4.27e@xmath5906 & 1.14e@xmath6004 + & & 4.5 & 19 & m1 & 144.9 & 1.81e@xmath5906 & 5.68e@xmath6005 + & & 3.5 & 16 & m1 & 140.4 & 3.26e@xmath5905 & 9.63e@xmath6006 + & & 3.5 & 14 & m1 & 134.0 & 3.09e@xmath5905 & 8.32e@xmath6006 + & & 3.5 & 9 & m1 & 127.6 & 8.37e@xmath5904 & 2.04e@xmath6006 + & & 4.5 & 10 & m1 & 127.7 & 8.22e@xmath5904 & 2.01e@xmath6006 + & & 5.5 & 13 & e2 & 133.4 & 6.84e@xmath5904 & 1.82e@xmath6006 + & & 4.5 & 19 & e2 & 144.9 & 6.03e@xmath5904 & 1.90e@xmath6006 + 1.5 & 41 & 1.5 & 1 & e1 & 45.31 & 3.29e@xmath5912 & 4.06e@xmath5900 + & & 2.5 & 2 & e1 & 48.71 & 1.15e@xmath5910 & 1.63e@xmath6002 + 3.5 & 42 & 2.5 & 2 & e1 & 47.05 & 3.67e@xmath5911 & 9.74e@xmath6001 + 2.5 & 43 & 2.5 & 2 & e1 & 46.34 & 2.58e@xmath5912 & 4.98e@xmath5900 + & & 1.5 & 1 & e1 & 43.26 & 6.71e@xmath5908 & 1.13e@xmath6003 + 1.5 & 44 & 2.5 & 2 & e1 & 46.25 & 2.17e@xmath5911 & 2.78e@xmath6001 + & & 1.5 & 1 & e1 & 43.18 & 1.57e@xmath5910 & 1.76e@xmath6002 + 4.5 & 45 & 5.5 & 21 & m1 & 129.5 & 4.47e@xmath5906 & 1.13e@xmath6004 + & & 4.5 & 25 & m1 & 135.7 & 2.10e@xmath5906 & 5.81e@xmath6005 + & & 3.5 & 28 & m1 & 149.9 & 3.53e@xmath5905 & 1.19e@xmath6005 + & & 3.5 & 33 & m1 & 187.5 & 1.57e@xmath5905 & 8.29e@xmath6006 + & & 3.5 & 24 & m1 & 132.5 & 1.34e@xmath5905 & 3.52e@xmath6006 + & & 4.5 & 25 & e2 & 135.7 & 8.96e@xmath5904 & 2.48e@xmath6006 + & & 5.5 & 21 & e2 & 129.5 & 8.02e@xmath5904 & 2.02e@xmath6006 + & & 3.5 & 35 & m1 & 524.5 & 7.74e@xmath5904 & 3.19e@xmath6005 + 3.5 & 46 & 2.5 & 2 & e1 & 46.01 & 1.94e@xmath5912 & 4.91e@xmath5900 + 1.5 & 47 & 2.5 & 2 & e1 & 45.93 & 2.13e@xmath5912 & 2.70e@xmath5900 + & & 1.5 & 1 & e1 & 42.90 & 2.80e@xmath5909 & 3.09e@xmath6003 + 2.5 & 48 & 2.5 & 2 & e1 & 45.81 & 2.12e@xmath5909 & 4.00e@xmath6003 + & & 1.5 & 1 & e1 & 42.79 & 9.56e@xmath5906 & 1.57e@xmath6005 + 0.5 & 49 & 1.5 & 1 & e1 & 41.57 & 3.20e@xmath5910 & 1.66e@xmath6002 + +

we use an adiabatic approximation in terms of instantaneous resonances to study the steady - state and time - dependent transport properties of interacting electrons in biased resonant tunneling heterostructures . this approach leads , in a natural way , to a transport model of large applicability consisting of reservoirs coupled to regions where the system is described by a nonlinear schrdinger equation . from the mathematical point of view , this work is non - rigorous but may offer some fresh and interesting problems involving semiclassical approximation , adiabatic theory , non - linear schrdinger equations and dynamical systems .

cond-mat9607088

man - tailored semiconductor heterostructures @xcite offer , for the first time , the possibility to test quantum mechanics at a mesoscopic level @xcite . the scenario of systems which can be investigated is so rich that the art of their realization deserves the name of quantum design . in the simplest case , a quantum designer can grow sandwiches of different semiconductor alloys by choosing the number of atomic layers for each kind of alloy . in the resulting heterostructure , the conduction band profile along the growth direction forms steps whose height can be continuously varied by a proper choice of the alloy composition . typical widths and heights are of the order of tens of and tenths of ev , respectively . at low - temperature , the mean free path of carriers for scattering from crystal impurities is of the order of @xmath0 and for heterostructures smaller than this size the electric transport along the growth direction is a phase coherent quantum scattering from the conduction band discontinuities @xcite . due to the translational invariance in the plane orthogonal to the growth direction , the problem is one - dimensional . moreover , the carriers are described by an effective mass which accounts for the microscopic scattering with the periodic crystal sites and their wave function is an envelope wave function @xcite . in a homogeneous neutral conductor , the electron - electron interaction can be taken into account by a renormalization of the carrier effective masses @xcite and one deals with a transport problem like in a noninteracting case . in a heterostructure , even as simple as that described above , the breaking of translational invariance in the transport direction allows the electric neutrality to be locally violated . the corresponding interaction potential , obtained , at hartree level , by solving a proper poisson equation , can strongly modify the transport properties . the example of a double barrier heterostructure with the exterior regions doped with donors is illuminating @xcite . due to tunneling , electrons populate the resonance(s ) created by the double barrier and the region between the barriers becomes negatively charged . this generates an electric potential which decreases the tunneling probability of electrons in the double barrier region . as a consequence , current oscillations on the picosecond scale @xcite and chaotic behavior without classical counterpart @xcite have been predicted in a ballistic configuration in which electrons are injected at some chosen energy . experiments with ballistic electrons are difficult and measurements became available only recently @xcite . technologically simpler is the case of biased heterostructures where transport is due to the presence of reservoirs at thermal equilibrium with different chemical potentials . manifestations of the electron - electron interaction are known also in this configuration . for example , hysteresis in the current - voltage characteristics of double barrier heterostructures have been observed @xcite and recognized as a consequence of the accumulation of electrons in the resonance @xcite . in this case , however , one has the theoretical problem of attaching reservoirs at thermal equilibrium to a piece of conductor where quantum coherent transport takes place . in the recent paper @xcite we proposed an approach to this problem based on a mathematical method earlier applied in the framework of ballistic transport @xcite . we showed that for heterostructures with a single resonance our approach allows one _ i ) _ to obtain steady - state voltage - current characteristics having hysteresis or not in agreement with the experimental results @xcite and _ ii ) _ to predict time - dependent properties analogous to those studied in optically bistable systems @xcite . here , we develop the general mathematical scheme of this approach and discuss the case with several resonances where multistability phenomena can take place as in superlattices @xcite . for simplicity , consider the one - dimensional double barrier heterostructure discussed above . the idea is that due to the presence of resonances the corresponding schrdinger problem can be divided in two parts : a schrdinger equation for the barrier region and one for the exterior space , the two being weakly coupled by tunneling . this decomposition corresponds to the schematization of the transport process as a coherent process fed by reservoirs . in the exterior space ( reservoirs ) homogeneous and neutral , the electron - electron interaction is neglected and thermal equilibrium is taken into account by considering a continuous set of energy eigenstates distributed according to the fermi statistics . in the barrier region ( coherent conductor ) , the coulomb interaction is included in a self - consistent potential obtained by solving the poisson equation associated to the local charge density . under the assumption that the barriers are wide enough , the corresponding nonlinear schrdinger problem is discussed in two steps . in the first step we eliminate the potential well between the two barriers , by artificially increasing the potential there , and we solve the schrdinger equation asymptotically for the new potential by means of wkb - expansions . the resulting solution is then very small near the ( filled ) potential well , so we get only a small error in the schrdinger equation when we go back to the true potential . in the second step we correct for this small error by adding a wave function concentrated near the potential well . assuming a priori that the charge in the well changes slowly with time , the correcting wave function can be expected to be large only at energies close to the resonances , and be well approximated by some linear combination of the resonant states . in most of the paper we discuss the case in which only one resonance participates . the validity of this one - mode approximation has been tested numerically with excellent results in the ballistic configuation of @xcite . here , the coefficient of the one - mode approximation obeys an ordinary differential equation with respect to time in the infinite dimensional space of square integrable functions of energy . we study the stationary points of the corresponding vector field and their nature , whether they are attractive or not , and arrive at quite neat answers . for solutions of the dynamical system which have existed as bounded solutions for a long time and in a suitable asymptotic limit ( of wide barriers ) we derive a simplified scalar differential equation for the evolution of the sheet density of electrons trapped in well which gives good global understanding of the more complete dynamical system . using these results , we are able to discuss the phenomenon of hysteresis and we support and illustrate the discussion with several numerical results . the discussion includes the evolution of solutions away from fixed points which necessarily appears when there is hysteresis . we also discuss the case of several resonances , and get analogous results . from the mathematical point of view , the present paper could be a starting point for rigorous work on some fresh problems , involving semiclassical analysis , adiabatic theory , non - linear schrdinger equations and dynamical systems . a strong motivation for such an enterprise is the fact that the theory of electric transport in semiconductor devices offers many problems similar to that one we illustrate here @xcite . the plan of the paper is as follows . in section [ 2 ] we define the model . in section [ 3 ] we review the wkb expansion for slowly varying potentials . in sections [ 4 ] and [ 5 ] we determine the driving term and the ground resonant state , respectively , within the wkb approximation . the central equation of our paper is derived in section [ 6 ] and the general properties of the associated fixed points and linearizations are discussed in section [ 7 ] . in section [ 8 ] we introduce an approximation valid in the limit of small resonance width and discuss the corresponding fixed point solutions and linearizations . in section [ 9 ] we obtain a simplified differential equation describing the dynamics of the electron density in the well . a qualitative discussion of the hysteresis phenomenon in comparison with numerical results is given in section [ 10 ] . in section [ 11 ] we finally consider the case with several resonances . let us consider a heterostructure whose conduction band profile consists of two barriers of height @xmath1 located in @xmath2 $ ] and @xmath3 $ ] @xmath4 with @xmath5 along the growth direction @xmath6 . we wish to evaluate the transport properties of this device when a bias energy @xmath7 is applied between the emitter ( @xmath8 ) and collector ( @xmath9 ) regions uniformly doped . due to doping , the band of conduction electrons formed in the emitter and collector regions is characterized by a fermi energy @xmath10 , where @xmath11 is the net donor concentration . we will use everywhere effective atomic units @xmath12 and @xmath13 , where @xmath14 is the electron effective mass and @xmath15 the dielectric constant . in these units , every physical quantity is expressed in terms of the effective bohr radius @xmath16 . assuming an ideal heterostructure homogeneous in the plane @xmath17 parallel to the junctions ( and orthogonal to the growth direction @xmath6 ) , the single - electron momenta @xmath18 and @xmath19 are conserved quantities . as a consequence , the single - electron wavefunction at energy @xmath20 , where @xmath21 , can be factorized as @xmath22 with @xmath23 we will assume periodic boundary conditions in a two - dimensional region @xmath24 so that the momenta @xmath18 and @xmath19 are quantized as in a real device having finite lateral area of size @xmath24 . the time dependent schrdinger equation for the single - electron wavefunction at energy @xmath25 along the @xmath6 direction is @xmath26 \phi(x , t , e ) = 0 , \label{1dnseu}\ ] ] where @xmath27 takes into account the applied bias and , at hartree level , the electron - electron interaction . assuming ideal metallic behavior in the emitter and collector regions , i.e. , neglecting the formation of accumulation and depletion layers , @xmath27 can be obtained as solution of the poisson equation @xmath28 with dirichlet boundary conditions @xmath29 and @xmath30 . the density @xmath31 takes into account all the electrons in the occupied energy states and depends only on the wavefunction component @xmath32 . indeed , if the emitter and collector regions are at thermal equilibrium with temperature @xmath33 we have @xmath34 where the factor 2 takes into account the spin degeneracy . energies are measured from the bottom of the emitter conduction band and the lower integration bound @xmath35 in the first line of ( [ density ] ) stems from the fact that for @xmath36 , as we will assume , only electrons from the emitter conduction band can penetrate the region @xmath37 $ ] where the electron density is of interest . in the second line of ( [ density ] ) this lower bound is absorbed in the definition of @xmath38 by a heaviside function @xmath39 . the function @xmath38 can be explicitly evaluated by approximating the sum over the parallel degrees of freedom with an integral @xmath40 . \label{g}\end{aligned}\ ] ] note that the chemical potential at temperature @xmath33 in the fermi function has been approximated with its value at @xmath41 , i.e. , the fermi energy determined by the net donor concentration . we will try to solve the nonlinear partial differential equation @xmath58 \phi(x , t , e ) = 0 , \label{1dnse}\ ] ] where @xmath54 is given by ( [ s ] ) , in two steps . let @xmath59}(x)$ ] be the potential obtained by filling the well @xmath43 $ ] . here @xmath60}(x)$ ] is the characteristic function of the interval @xmath43 $ ] . first we solve @xmath61 \tilde{\mu}(x , t , e ) = 0 \label{muse}\ ] ] and then we look for @xmath32 in the form @xmath62 where @xmath63 should solve @xmath64 \tilde{\nu}(x , t , e ) = v_0~1_{[b , c]}(x ) \tilde{\mu}(x , t , e ) . \label{nuse}\ ] ] the wave function @xmath65 describes an electron at energy @xmath25 which is delocalized in the emitter and collector regions and has an exponentially small probability to be found in the forbidden region @xmath37 $ ] . the wave function @xmath63 describes the localization , driven by @xmath65 , of the same electron in the well @xmath43 $ ] . the wave function @xmath32 of the original problem ( [ 1dnse ] ) can be approximated by @xmath63 or @xmath65 inside or outside the two barriers , respectively , with an error which is exponentially small in the limit of wide barriers @xcite . to evaluate @xmath65 we will use a wkb approximation in the forbidden region @xmath37 $ ] . equation ( [ nuse ] ) will be treated with a one mode approximation in which @xmath63 is assumed proportional to a resonant state corresponding to the potential @xmath51 . to evaluate this resonant state and the corresponding resonance , we will again use a wkb approximation . in both cases , the justification of using a wkb approximation stems from the fact that @xmath66 and @xmath51 are slowly varying potentials in the barriers regions if @xmath67 and @xmath68 are large while @xmath7 and @xmath69 remain bounded . let @xmath70 be a real valued potential on some interval , with @xmath71 and @xmath72 , where @xmath73 is a parameter . let @xmath74 be a real energy and assume that @xmath75 is bounded from above and from below by some strictly positive constants that are independent of @xmath76 . this means that we are in the classically forbidden region . then @xmath77 \left ( { \cal u } - { \cal e } \right)^{-1/4 } e^{- \int^x dx'~ \left ( { \cal u } - { \cal e } \right)^{1/2 } } \nonumber \\ & & = \left [ -{5\over16 } \left ( { \cal u } - { \cal e } \right)^{-9/4 } \left ( \partial_x { \cal u } \right)^2 + { 1\over4 } \left ( { \cal u } - { \cal e } \right)^{-5/4 } \partial_x^2 { \cal u } \right ] e^{- \int^x dx'~ \left ( { \cal u } - { \cal e } \right)^{1/2 } } \nonumber \\ & & = e^{- \int^x dx'~ \left ( { \cal u } - { \cal e } \right)^{1/2 } } { \cal o } ( h^2),\end{aligned}\ ] ] and therefore @xmath78 is a good approximation to a corresponding exact eigenfunction , even over intervals of length @xmath79 . in the following sections , we will apply the above approximation in the barrier regions @xmath2 $ ] and @xmath3 $ ] with @xmath76 equal to the @xmath6 derivative of @xmath51 in these intervals . equation ( [ muse ] ) can be solved by evaluating the instantaneous eigenstates of the potential @xmath80 . we put @xmath81 and suppose that @xmath7 and @xmath69 are slowly varying functions of time so that also @xmath82 is slowly varying in time . thus in the equation @xmath83 \mu(x , t , e ) = 0,\ ] ] we make a very small error if we neglect the term @xmath84 , as we shall do in the following . in the emitter region @xmath8 , we take @xmath82 as the sum of a left and a right - going plane wave at energy @xmath25 @xmath85 where @xmath86 is a reflection amplitude to be computed . note that the normalization factor in ( [ muemitter ] ) is chosen in order to have @xmath87 in agreement with the expression of the electron density ( [ density ] ) in terms of an integral over the energy @xmath25 . we propagate the expression ( [ muemitter ] ) to the adjacent regions by requiring @xmath88 to be of class @xmath89 and applying the wkb approximation described in section iii . in the interval @xmath2 $ ] the potential is @xmath90 , where @xmath91 plays the role of the small parameter @xmath76 of section iii . for @xmath92 we can then use the wkb approximation @xmath93 where @xmath94 is a transmission amplitude to be determined with @xmath86 from the @xmath89 condition at @xmath95 @xmath96.\end{aligned}\ ] ] neglecting the last term in the bracket , which is @xmath97 , we get @xmath98 note that the neglected term would give correction factors @xmath99 to @xmath86 and @xmath94 . at @xmath48 we can set up a similar transition problem but here @xmath80 is continuous and the corresponding transmission amplitude is @xmath99 . neglecting again a factor @xmath99 , for @xmath46 we get @xmath100 where @xmath101 { 2/3\alpha } } \over 1 + i ( v_0/e -1)^{1/2}}. \label{mu0}\end{aligned}\ ] ] only this expression of @xmath88 in the region @xmath43 $ ] will be used in the following as driving term of eq . ( [ nuse ] ) . in this section we will obtain a wkb approximate expression for the ground state resonance @xmath102 and the corresponding resonant state @xmath103 for the potential @xmath51 . we will assume that @xmath104 is bounded from below and from above by positive constants , while @xmath67 and @xmath68 are large enough . to start with , we recall the construction of the ground state eigenvalue @xmath105 of the potential @xmath106-\infty , b]}(x ) + 1_{[c,+\infty[}(x ) \right]$ ] which coincides , up to the constant shift @xmath107 with the potential @xmath51 in the well region @xmath43 $ ] . the corresponding ground eigenstate is @xmath108 where @xmath109 is determined by the requirement that @xmath110 is of class @xmath89 @xmath111 and the normalization constant is @xmath112^{-1/2}\ ] ] where we used the identities @xmath113 , @xmath114 . in the following we will assume that @xmath115 . next we look at the ground state of the potential @xmath116 which coincides , up to the constant shift @xmath117 , with @xmath51 on the larger region @xmath37 $ ] which includes the barriers . in eq . ( [ vb ] ) @xmath118 is given by ( [ alpha ] ) and @xmath119 note that the potential @xmath120 has been obtained by bending the barriers of @xmath121 in the intervals @xmath2 $ ] and @xmath3 $ ] proportionally to @xmath118 and @xmath122 , respectively . let @xmath123 be the ground state of @xmath120 and @xmath124 the corresponding eigenfunction . since @xmath125 and @xmath126 are small , from the same wkb considerations of section iii we have @xmath127 and @xmath128 . to get the leading asymptotics of the resonance width , we need to determine the linear contribution to @xmath129 in @xmath123 . by differentiating the eigenvalue equation for the potential @xmath120 , we have @xmath130 @xmath131 and using ( [ ew ] ) we get @xmath132 observing that @xmath133 , we finally get @xmath134 the real part @xmath135 of the shape resonance of @xmath136 which is close to the ground state eigenvalue of @xmath137 is very well approximated by the above calculated @xmath138 which can be rewritten as @xmath139 where @xmath140 and @xmath141 now we discuss the determination of the imaginary part @xmath142 of the resonance . in the interval @xmath37 $ ] the ground state of @xmath120 is @xmath143 where @xmath144 . in the interval @xmath37 $ ] , the resonant state @xmath103 can be approximated by adding to ( [ eb ] ) terms due to reflections at @xmath95 and @xmath145 . for @xmath146 we try with @xmath147 where we have also replaced the exponent with its linear approximation at @xmath145 . for @xmath148 we try the right - going plane wave @xmath149 the @xmath89 condition at @xmath145 gives , up to terms @xmath150 , @xmath151 which determines @xmath152 and @xmath153 so that for @xmath148 we have @xmath154^{-1 } \nonumber \\ & & \times \exp\left\ { { 2\over 3\beta } \left [ ( v_0-e_0^b)^{3/2 } - ( v_0 + \beta(d - c ) -e_0^b)^{3/2 } \right ] + i ( e_0^b - \beta(d - c))^{1/2 } ( x - d ) \right\}. \label{esgd}\end{aligned}\ ] ] in these calculations we have assumed that @xmath155 , @xmath156 . the first inequality is always fulfilled in experimentally relevant situations , while the second one , equivalent to @xmath157 may be more critical and , possibly , one should replace ( [ esgd ] ) by a more complicated formula . the same calculation can be repeated for @xmath95 . for @xmath158 we try with @xmath159 with a new reflection amplitude @xmath152 . for @xmath160 we try the left - going plane wave @xmath161 with a new transmission amplitude @xmath153 . the @xmath89 condition at @xmath95 gives , up to terms @xmath162 , @xmath163 which determines @xmath152 and @xmath153 so that for @xmath160 we have @xmath164^{-1 } \nonumber \\ & & \times \exp\left\ { { 2\over 3\alpha } \left [ ( v_0 - \alpha(b - a ) -e_0^b)^{3/2 } - ( v_0-e_0^b)^{3/2 } \right ] - i ( e_0^b + \alpha(b - a))^{1/2 } ( x - a ) \right\}. \label{esma}\end{aligned}\ ] ] note that for @xmath160 , @xmath103 is a true left - going plane wave only for @xmath7 not too large when @xmath165 . if @xmath166 , eq . ( [ esma ] ) becomes an exponentially decaying function whose corresponding probability current density vanishes . since @xmath167 , this case corresponds to @xmath168 . in eq . ( [ esma ] ) , we also assumed that @xmath169 , i.e. , @xmath170 . the resonance width can be now computed by means of the green formula @xmath171 where @xmath172 and @xmath173 . the integral in the l.h.s . of ( [ green ] ) is @xmath174 and using ( [ esgd ] ) and ( [ esma ] ) we get , up to such a factor , @xmath175 { 4/3\beta } } \nonumber \\ & & + ( v_0 - \alpha ( b - a ) - e_0^b)^{1/2 } ~ ( e_0^b + \alpha ( b - a))_+^{1/2 } ~ e^ { \left [ ( v_0 - \alpha(b - a ) -e_0^b)^{3/2 } - ( v_0-e_0^b)^{3/2 } \right ] { 4/3\alpha } } \bigg ] , \label{gamma}\end{aligned}\ ] ] where we used @xmath176 . equation ( [ nuse ] ) can be simplified by developing @xmath63 into the instantaneous eigenstates of the potential @xmath51 and keeping only the contributions from the discrete resonant states , i.e. , neglecting the contributions from the continuous spectrum @xcite . for the moment , we will suppose there is only one resonant state and put @xmath177 where @xmath103 is the ( ground ) resonant state of the potential @xmath51 @xmath178 e(s , x ) = 0 \label{es}\ ] ] with complex eigenvalue @xmath179 . the eigenfunction @xmath103 is of class @xmath180 on the contour @xmath181-\infty,0]+a \right ) \bigcup ~[a , d ] \bigcup ~\left ( d+e^{i\theta}[0,+\infty [ \right)$ ] for @xmath182 conveniently chosen @xcite and satisfies @xmath183 multiplying ( [ nuse ] ) with @xmath103 and integrating over @xmath184 , we get @xmath185 z(t , e ) + { \cal b}(t , s , e ) \label{zdot}\ ] ] with the driving term given by @xmath186 and the sheet density ( [ s ] ) reduced , with small error , to @xmath187 we consider the vector field in the l.h.s . of ( [ zdot ] ) , @xmath188 where @xmath189 is a non - vanishing function . for simplicity , we assume that @xmath190 is independent of @xmath153 . when @xmath190 varies slowly with @xmath153 , the discussion below should be applied to each such fixed value of @xmath153 . we first look for fixed points of @xmath191 , i.e. , functions @xmath192 in @xmath193 with @xmath194 . clearly @xmath192 is a fixed point iff @xmath195 so the @xmath180-norm @xmath196 has to satisfy @xmath197 conversely , if @xmath198 is a solution of ( [ 6.4 ] ) , then @xmath199 gives the unique fixed point of @xmath191 with @xmath200 . assuming that we have found a fixed point @xmath192 , we look for the linearization of the vector field @xmath191 at that point . by giving an infinitesimal increment @xmath201 to @xmath202 , the corresponding increment @xmath203 to @xmath191 is @xmath204 where @xmath196 is the corresponding solution of ( [ 6.4 ] ) and @xmath205 . @xmath206 so with @xmath207 and @xmath208 , we get the complexification of the linearization , @xmath209 the matrix in the first term of the r.h.s . has continuous spectrum contained in @xmath210 and the second term appears as a rank one perturbation . if @xmath211 is an eigenvalue of @xmath212 with real part different from @xmath213 , we get @xmath214 @xmath215 we must then have @xmath216 @xmath217 where @xmath218 . in order to have a non - trivial solution @xmath219 , it is necessary and sufficient that @xmath220 here , the l.h.s . is real for real @xmath221 , and tends to 1 , when @xmath222 . on the other hand , the @xmath69-derivative of the l.h.s . minus the r.h.s . of ( [ 6.4 ] ) is @xmath223 which coincides with the l.h.s . of ( [ 6.12 ] ) for @xmath224 . so if the expression ( [ 6.13 ] ) is @xmath225 , we see that ( [ 6.12 ] ) must have a solution @xmath226 . let us say that the fixed point is stable if the spectrum of the linearization @xmath212 is contained in the open left half - plane and unstable otherwise . the discussion above then gives : [ section ] let @xmath227 be a fixed point of @xmath191 so that ( [ 6.4 ] ) and ( [ 6.5 ] ) hold . if the @xmath69-derivative of the l.h.s . minus the r.h.s . of ( [ 6.4 ] ) is @xmath225 , then @xmath227 is an unstable fixed point . more precisely , the linearization @xmath212 then has an eigenvalue which is @xmath228 . in this section we assume that the driving term @xmath230 is a sufficiently smooth function of @xmath25 , at least near the point @xmath231 , where @xmath69 solves ( [ 6.4 ] ) . when the barriers are very wide , @xmath232 will be very small and @xmath233 is a function of @xmath25 which is sharply peaked at @xmath234 . in ( [ 6.4 ] ) it is therefore justified to replace @xmath235 by the constant value @xmath236 . then ( [ 6.4 ] ) is well approximated by @xmath237 we shall next apply a similar argument to the equation ( [ 6.12 ] ) for the eigenvalues of the linearization @xmath212 and , for more transparency , we start with a simplified case , in which @xmath238 in this case , ( [ 6.12 ] ) reduces to @xmath239 [ ( { \gamma / 2})^2+(e - e_r(0)-\eta s)^2 ] } = 0 . \label{7.3}\ ] ] we shall use , @xmath240 @xmath241 if we replace @xmath242 in the integral in ( [ 7.3 ] ) with its value at @xmath243 , that integral vanishes since the integrand becomes an odd function of @xmath244 . instead , we get an approximation of the integral in ( [ 7.3 ] ) by replacing @xmath242 with the linear term in its taylor expansion at @xmath243 . using ( [ 7.4 ] ) , we then get from ( [ 7.3 ] ) @xmath245 @xmath246 where @xmath247 . the solution of ( [ 7.6 ] ) is @xmath248 and this is an eigenvalue of the linearization @xmath212 as long as @xmath249 the solution of ( [ 7.7 ] ) is @xmath250 and describes an eigenvalue of @xmath212 precisely when ( [ 7.9 ] ) is fulfilled . we then have the following conclusion under the simplifying assumption ( [ 7.2 ] ) and in the small-@xmath229 limit . + when @xmath251 : no eigenvalues of @xmath212 and hence an attractive fixed point . + when @xmath252 : two eigenvalues of @xmath212 and still an attractive fixed point . + when @xmath253 : two eigenvalues and a non - attractive fixed point . the main conclusion under the same assumptions is then : [ section ] we have an attractive fixed point precisely when the @xmath69-derivative of the difference of the l.h.s . and the r.h.s . in ( [ 7.1 ] ) is @xmath228 . now we drop the simplifying assumption ( [ 7.2 ] ) and see that the preceding proposition still holds in the small-@xmath229 limit . let @xmath227 be a fixed point , so that @xmath196 ( approximately ) solves ( [ 7.1 ] ) . in view of ( [ 6.2 ] ) , eq . ( [ 6.12 ] ) can be written as @xmath254 [ ( { \gamma ( s)/ 2})^2+(e - e_r(0)-\eta s)^2 ] } = 0\ ] ] here , the numerator of the integrand can be simplified to @xmath255 \nonumber\\ & & + [ ( e - e_r(0)-\eta s)(-2\eta \vert { \cal b}\vert ^2+i\lambda ( \partial_s { \cal b } \overline{{\cal b}}-{\cal b } \partial_s \overline{{\cal b } } ) ] -[(e - e_r(0)-\eta s)^2 \partial_s \vert { \cal b}\vert ^2].\end{aligned}\ ] ] accordingly , we split the integral into three pieces and apply the small-@xmath256 approximation to each one . if we assume , for simplicity , that @xmath257 ( which is necessarily the case if the eigenvalue @xmath258 is to ruin attractiveness ) we get @xmath259\pi g \over ( \gamma /2)({\gamma /2}+\lambda ) ( \gamma + \lambda ) } \nonumber\\ & & + { [ { \partial_e } ( g ( -2\eta \vert { \cal b}\vert ^2+i\lambda ( \partial_s { \cal b } \overline{{\cal b}}-{\cal b } \partial_s \overline{{\cal b } } ) ) ) -g \partial_s \vert { \cal b}\vert ^2 ] \pi \over \gamma + \lambda } = 0\end{aligned}\ ] ] at @xmath243 . this can be rewritten as @xmath260 again at @xmath243 . noticing that @xmath261 and multiplying with @xmath262 , we get the following approximation of ( [ 6.12 ] ) @xmath263 \nonumber \\ & & = -\gamma(s ) -2\pi { ( g\vert { \cal b}\vert ^2)(s , e_r(0)+\eta s ) \over \gamma(s ) } \partial_s \gamma(s ) + 2\pi { d\over ds}((g\vert { \cal b}\vert ^2)(s , e_r(0)+\eta s ) ) . \label{7.12}\end{aligned}\ ] ] we assume that @xmath264 , so that the solution @xmath221 of ( [ 7.12 ] ) is real and has the same sign as the r.h.s . of ( [ 7.12 ] ) . on the other hand , the @xmath69-derivative of the l.h.s . minus the r.h.s . of ( [ 7.1 ] ) is @xmath265 which is of the opposite sign to the r.h.s . in ( [ 7.12 ] ) . we then have : [ p8.1]proposition under the weaker assumptions above and in the small-@xmath229 limit , we still have an attractive fixed point precisely when the @xmath69-derivative of the l.h.s . minus the r.h.s . of ( [ 7.1 ] ) is @xmath228 . consider the differential equation ( [ zdot ] ) @xmath266~z(t , e ) + { \cal b}(s(t),e ) , \label{9.1}\ ] ] where @xmath267 , and where we could also let @xmath190 be a slowly varying function of @xmath153 through @xmath268 . assuming @xmath268 to be a known function , the solution of ( [ 9.1 ] ) with a prescribed initial value at time @xmath269 is @xmath270 assuming that the solution has existed as a bounded solution for a very long time , say from the time @xmath271 , we can let @xmath269 tend to @xmath271 in the formula above and get @xmath272 taking the scalar product of ( [ 9.1 ] ) and @xmath227 gives the following equation for the derivative of the sheet density @xmath273 where @xmath274 we now assume that @xmath268 varies slowly with @xmath153 and replace @xmath275 in the above integral by @xmath276 . making the @xmath25-integration first , we get @xmath277 where @xmath278 denotes the fourier transform with respect to @xmath25 . assuming @xmath279 sufficiently smooth as a function of @xmath25 , we see that @xmath280 is sufficiently rapidly decreasing as a function of @xmath281 for the following approximations to be made : _ i ) _ since @xmath142 is small , we may assume that @xmath282 _ ii ) _ since @xmath283 varies slowly , we may replace @xmath284 by @xmath285 . we then get @xmath286 using the property @xmath287 , valid for any real valued function @xmath288 , we obtain @xmath289 inserting this in ( [ 9.2 ] ) , we get the approximate differential equation for the sheet density @xmath290 @xmath291 . \label{simplediff}\ ] ] this equation is valid for slowly varying solutions which have evolved for a time much longer than @xmath292 . we start by examining the simplified fixed point equation ( [ 7.1 ] ) . for @xmath293 with @xmath294 , we have @xmath295 ( of the same order of magnitude ) . by evaluating the integral in ( [ b ] ) with @xmath103 approximated by ( [ ew ] ) and the driving term given by ( [ mu ] ) , we have @xmath296 4/3\alpha } .\end{aligned}\ ] ] assuming for simplicity zero temperature , so that @xmath297 , we get @xmath298 4/3\alpha}.\end{aligned}\ ] ] recalling that @xmath299 , where @xmath300 and @xmath301 are given by ( [ alpha ] ) and ( [ beta ] ) , respectively , from ( [ gamma ] ) we get @xmath302 4/3\alpha } \nonumber \\ & & ~+ ( v_0-\delta v - e_r(s))^{1/2}~(e_r(s)+\delta v)^{1/2 } e^{\left [ ( v_0- \delta v - e_r(s)-\beta ( d - c))^{3/2}- ( v_0-\delta v - e_r(s))^{3/2 } \right ] 4/3\beta } \bigg].\end{aligned}\ ] ] we will consider the following two cases : \1 ) the barrier @xmath303 is more opaque than @xmath304 in the sense that the exponential factor in the second term of the above expression for @xmath232 is much smaller than the exponential factor in the first term . \2 ) the barrier @xmath304 is more opaque than @xmath303 . in the intermediate case when the two barriers have opacity of the same order , the discussion of case 1 ) will roughly apply . notice that opacity depends not only on the relative sizes of @xmath67 and @xmath68 , but also on @xmath69 and @xmath7 . therefore , we may have transitions between the two cases when these parameters vary . interesting phenomena appear when the case 1 ) is possible and we start with that case , recalling that @xmath305 . in this case ( and neglecting , to start with , the possibility of a transition to the case 2 ) ) the first term in the expression for @xmath232 dominates , except when @xmath135 is negative or very small and positive . the function @xmath306 vanishes for @xmath307 and rises very sharply ( with a square root singularity at @xmath308 ) from @xmath309 to @xmath310 when @xmath135 is increased from 0 to a small positive value . when @xmath135 is further increased , the function @xmath311 decreases roughly linearly and vanishes for @xmath312 . the values @xmath308 , @xmath313 correspond to @xmath314 and describe the boundary points of the support of the function ( [ 10.1 ] ) . when @xmath7 is increased , these two points move to the right with the same speed as shown in the example of fig . 2 . in fig . 2 we also see the graphical solution of eq . ( [ 7.1 ] ) , @xmath315 , for different values of @xmath7 . it is clear that ( [ 7.1 ] ) will first have only one solution when @xmath316 , then three solutions for @xmath7 in some interval , until @xmath317 , and again only one solution for even larger values of @xmath7 . according to the results of section [ 8 ] , we see that in the case in which ( [ 7.1 ] ) has only one solution , this solution corresponds to an attractive fixed point , and when there are three solutions , the smallest and the largest of these correspond to attractive fixed points , while the intermediate solution corresponds to an unstable fixed point . for many experimentally relevant situations the resonance width is much smaller than the other energy scales ( essentially @xmath318 ) . in this case we may expect the simplified fixed point equation ( [ 7.1 ] ) to be a very good approximation of the more correct equation ( [ 6.4 ] ) , except near the boundary points of the support of the function ( [ 10.1 ] ) . this is confirmed by fig . 3 where we show the numerical solutions ( stable and unstable ) of both ( [ 6.4 ] ) and ( [ 7.1 ] ) for a system having @xmath319 at @xmath320 ev . in the case of eq . ( [ 6.4 ] ) , the corresponding energy integral has been evaluated on a uniform energy mesh having a density of points @xmath321 . the solutions of the simplified differential equation ( [ simplediff ] ) converge to one of the solutions of ( [ 7.1 ] ) , associated to an attractive fixed point . the phenomenon of hysteresis then becomes clear . we let @xmath7 increase very slowly from some sufficiently small value up to some sufficiently large positive value and subsequently decrease it very slowly , back to its initial value . consider a corresponding solution of the time dependent schrdinger equation ( [ 1dnseu ] ) so that we expect the corresponding evolution of the sheet density to be well approximated by the solution of ( [ simplediff ] ) , where @xmath190 varies slowly with time . first , there is only one ( attractive ) fixed point and the time dependent solution has to stay close to that fixed point . then we have creation of a pair of fixed points ( one stable and one unstable ) away from the solution , but the solution continues to stay close to the old ( stable ) fixed point . when @xmath7 reaches a sufficiently large value , the unstable fixed point runs into the old stable one and they both disappear . at this point , the time dependent solution has no other choice than to converge to the only remaining fixed point ( which is stable ) . when decreasing @xmath7 back to its initial value , we have the same behaviour , in the sense that the solution stays close to the initially unique fixed point as long as it exists and converges to the new unique fixed point after the old one has collapsed with the unstable one . the bias energy @xmath7 at which this collapse happens depends on the value of the time dependent solution and therefore is different when @xmath7 is increased or decreased . the phenomenon of hysteresis is clearly seen in fig . 3 , where the collapse points for @xmath7 decreased from large values and @xmath7 increased from small values have been marked with @xmath24 and @xmath322 , respectively . we have @xmath323 . we can estimate the order of magnitude of the hysteresis width @xmath324 by considering that @xmath325 is determined by the condition @xmath326 and @xmath327 by the condition @xmath328 . we have @xmath329 in figs . 4 and 5 we show the time dependent evolution of the sheet density @xmath268 when we start from a fixed point solution corresponding to the point @xmath24 or @xmath322 and give an instantaneous small decrement or increment @xmath330 to @xmath325 or @xmath327 , respectively . in these figures , the thick lines are the solutions of the full schrdinger equation ( [ zdot ] ) and ( [ nz2 ] ) and the thin lines the solution of the simplified differential equation ( [ simplediff ] ) . in fig . 4 the solutions corresponding to the small-@xmath229 limit and the full schrdinger equation start , as shown in fig . 3 , from different fixed point values , @xmath331 , and converge to the same ( approximatively ) values . on the other hand , when the starting point is @xmath322 ( fig . 5 ) the small-@xmath229 approximation is close to the solution of the full equation except for the value which @xmath268 has to converge to , again in agreement with fig . 3 . as a third example of time evolution of the sheet density of electrons in the well , in fig . 6 we show the behavior of @xmath268 solution of the full schrdinger equation ( [ zdot ] ) and ( [ nz2 ] ) after an instantaneous change @xmath330 of the bias energy corresponding to the point @xmath332 of fig . 3 well inside the hysteresis region . if @xmath333 is chosen sufficiently large , we observe oscillations of @xmath268 on the picosecond time scale . contrary to the claim of @xcite , these oscillations are damped since @xmath268 has to converge to the fixed point solution corresponding the bias energy @xmath334 . in the case 2 ) , when the barrier @xmath2 $ ] is more opaque than the barrier @xmath3 $ ] , the function ( [ 10.1 ] ) is very small , and for solutions of ( [ 7.1 ] ) we can observe only a microscopical hysteresis effect , due to the square root singularity at @xmath308 , which is likely to be completely absent in the more correct equation ( [ 6.4 ] ) . the absence of the hysteresis effect in this case is in agreement with the experimental results of @xcite and is discussed in @xcite . let us finally consider the case of very wide barriers and see that a transition between the cases 1 ) and 2 ) has to take place in the hysteresis region . let @xmath335 , @xmath336 , and @xmath337 . in this limit , @xmath338 , and the values in ( [ 10.2 ] ) are the endpoints of a short interval of length @xmath339 . let us consider ( [ 7.1 ] ) when @xmath7 is increased from the initial value @xmath340 for which the right end point in ( [ 10.2 ] ) is 0 . if the constant @xmath341 is sufficiently small , we are in the case 1 ) . for @xmath342 , we remain in the case 1 ) , provided that @xmath341 is sufficiently small , and ( [ 7.1 ] ) has a unique solution . at @xmath343 we have the creation of two new fixed points . if we follow the old fixed point , we can not remain in the case 1 ) until it disappears . indeed , if we did , the disappearance would take place when @xmath344 and at a bias energy @xmath345 obtained by the condition @xmath346 . since @xmath347 is between @xmath309 and @xmath318 , the inclination @xmath300 of the first barrier would have to be very small and we get a finite inclination @xmath348 for the barrier @xmath3 $ ] . therefore , when @xmath337 only the opacity of the first barrier would diverge and , at some point , we would be no more in the case 1 ) . what will actually happen is that when @xmath7 reaches some value which is bounded independently of @xmath67 , we have a transition from the case 1 ) to the case 2 ) and @xmath349 decreases to some value which is much smaller than the r.h.s . in ( [ fmax ] ) . this will cause the disappearance of the fixed point for a much smaller value of @xmath69 . when a transition from case 1 ) to case 2 ) happens , we still observe a hysteresis phenomenon , but this is now caused not only by the _ translation _ of @xmath311 as a function of @xmath7 but also by a _ variation _ of its height . this effect is already apparent in fig . 3 where we see a decreasing of the height of @xmath311 when increasing @xmath7 . in this section we discuss very briefly the case with several shape resonances . much of the discussion is similar to the case of one resonance and we shall assume that we are in a parameter range where all the wkb considerations apply . first we review the approximation for the shape resonances . we start with the potential @xmath121 and consider its eigenstates @xmath350 , @xmath351 and the corresponding eigenvalues @xmath352 . since @xmath350 is even as a function of @xmath353 for even @xmath354 and odd for odd @xmath354 , we have @xmath355 the @xmath89 condition at @xmath48 , or equivalently at @xmath49 , gives the quantization condition @xmath356 which can also be written as @xmath357 representing this equation graphically , we see that @xmath358 is the largest integer @xmath359 with @xmath360 . the functions @xmath350 are normalized , if we choose @xmath361 the eigenvalues @xmath362 associated to the potential @xmath120 in ( [ vb ] ) can be studied as before , and we get @xmath363 in the following , we neglect the error @xmath364 . the shape resonances @xmath365 for the potential @xmath51 are then given by @xmath366 where @xmath367 and @xmath3684/3\beta } \nonumber\\ & & + ( v_0-\alpha ( b - a)-e_j^b)^{1/2 } ( e_j^b+\alpha(b - a))_+^{1/2 } e^{[(v_0-\alpha ( b - a)-e_j^b)^{3/2}-(v_0-e_j^b)^{3/2}]4/3\alpha } \bigg ] . \label{11.5}\end{aligned}\ ] ] the corresponding resonant state @xmath369 , satisfying ( [ n&o ] ) , can be described as in section [ 5 ] . we still try to solve ( [ 1dnse ] ) in two steps . equation ( [ muse ] ) is treated as before , while the eq . ( [ nuse ] ) is now handled by letting @xmath370 be a linear combination of the @xmath371 resonant states @xmath372 . more precisely , we write @xmath373 and @xmath374 , so that ( [ nuse ] ) becomes @xmath375 assume , @xmath376 where @xmath69 is defined in ( [ s ] ) and hence will be time dependent . the functions @xmath377 approximately form an orthonormal family in @xmath378)$ ] , and if we assume that @xmath379 dominates over @xmath380 in @xmath381 then , with a small error , we have @xmath382 where the norms are in @xmath193 and in @xmath383 , respectively . substituting ( [ 11.7 ] ) into ( [ 11.6 ] ) , multiplying by @xmath369 and integrating over the contour @xmath384 , we get @xmath385 ( z_k(t , e ) ~e_k(s , x))~ e_j(s , x ) = v_0~\int_b^c dx~ \mu ( x , t , e ) e_j(s , x ) . \label{11.9}\ ] ] from the relations @xmath386 , we conclude that @xmath387 is an anti - symmetric matrix , and since @xmath388 are approximately real functions near @xmath389 , this matrix is also very close to a real one . equation ( [ 11.9 ] ) can be written as @xmath390 z_j(t , e ) - i\partial _ t ( s(t ) ) \sum_{k=0}^{n-1 } \int_\gamma dx~ ( \partial_se_k(s , x ) ) e_j(s , x ) \nonumber \\ = v_0 \int_b^c dx \mu ( x , t , e ) e_j(s , x ) . \label{11.10}\end{aligned}\ ] ] due to the facts that _ i ) _ @xmath391 can be expected to be very small and _ ii ) _ @xmath388 is roughly independent of @xmath69 near @xmath392 so that the integral @xmath393 can be expected to be very small , we will neglect the sum in the l.h.s . of ( [ 11.10 ] ) . in this case , we have @xmath394 z_j(t , e ) + { \cal b}_j(t , s , e ) , \label{11.11}\ ] ] where @xmath395 . we assume that @xmath396 vary slowly with @xmath153 , so it is meaningful to look at instantaneous fixed points of the vector field defined by the r.h.s . of ( [ 11.11 ] ) in @xmath383 . assuming , for simplicity , that @xmath396 are independent of @xmath153 we see that @xmath397 is a fixed point precisely when : @xmath398 from which we get the compatibility condition for @xmath196 @xmath399 conversely , if @xmath69 is a solution of ( [ 11.13 ] ) , then ( [ 11.12 ] ) defines the unique fixed point with @xmath200 . in the small-@xmath256 limit , as in section [ 8 ] we get the simplified fixed point equation @xmath400 in view of ( [ 11.4 ] ) , the term of index @xmath354 in ( [ 11.14 ] ) is a function of @xmath69 with support in the interval @xmath401 and when @xmath7 increases this interval moves to the right with speed @xmath402 as shown in the example of fig . 7 . in fig . 8 we compare the corresponding fixed point solutions obtained by solving ( [ 11.13 ] ) with those obtained in the small-@xmath229 limit ( [ 11.14 ] ) as a function of the bias energy @xmath7 . between the points marked as @xmath24 and @xmath322 we observe five fixed points . below we give some results about the nature of fixed points , which are more complicated than in the case of a single resonance and it is not clear that those results are applicable in the situation of fig . 8 . if we assume that they are applicable , then three fixed points are stable and two unstable . the existence of more than three fixed points , i.e. , the maximum number allowed for @xmath403 , is related to the possibility that the intervals ( [ 11.15 ] ) are not disjoint , as clearly understood by fig . it is interesting to study the evolution of the sheet density @xmath268 away from a point like @xmath322 in fig . 8 where a ( presumably ) stable fixed point and an unstable one collapse while two other fixed points survive . in fig . 9 we show the behavior of @xmath268 obtained by numerically integrating eq . ( [ 11.11 ] ) after an instantaneous increase @xmath330 of the initial bias @xmath327 . if the total bias @xmath404 , where @xmath332 is the next point where a new couple of fixed points collapse , @xmath268 converges to the fixed point closest to its initial value @xmath331 . when @xmath405 , @xmath268 first approaches the value corresponding to the collapse point @xmath332 but finally has to converge to the lower unique fixed point corresponding to the chosen bias . next we study the linearization of the vector field defined by the r.h.s . of ( [ 11.11 ] ) at a fixed point under the following simplifying assumption : @xmath406 then ( [ 11.11 ] ) becomes , @xmath407 the same calculations as in section [ 8 ] show that the complexification @xmath212 of the linearization of the vector field defined by the r.h.s . of ( [ 11.17 ] ) at a fixed point , is given by @xmath408 u_0 - i \eta ( \langle u | z \rangle + \langle v | \overline{z } \rangle ) z_0 \\ \vdots \\ \protect { [ - \gamma_{n-1 } /2 + i(e - e_{r , n-1}(0)- \eta s ) ] u_{n-1 } - i \eta ( \langle u | z \rangle + \langle v | \overline{z } \rangle ) z_{n-1 } } \\ \protect { [ - \gamma_0 /2-i(e - e_{r,0}(0)- \eta s ) ] v_0 + i \eta ( \langle u | z \rangle + \langle v | \overline{z } \rangle ) \overline{z}_0 } \\ \vdots \\ \protect { [ - \gamma_{n-1 } /2-i(e - e_{r , n-1}(0)- \eta s ) ] v_{n-1 } + i \eta ( \langle u | z \rangle + \langle v | \overline{z } \rangle ) \overline{z}_{n-1 } } \end{array } \right ) . \label{11.18}\ ] ] here , @xmath409 . the operator @xmath212 is a rank one perturbation of an operator with essential spectrum contained in @xmath410 . we look for eigenvalues @xmath211 with @xmath411 for all @xmath354 . if @xmath412 is a corresponding eigenvector , we get as in section [ 8 ] @xmath413 where @xmath414 using ( [ 11.12 ] ) , ( [ 11.20 ] ) in ( [ 11.21 ] ) , we see that @xmath221 is an eigenvector precisely when @xmath415 [ ( { \gamma_k/ 2})^2+(e - e_{r , k}(0)-\eta s)^2]}=0 . \label{11.22}\ ] ] as in the case @xmath403 , we observe that the l.h.s . of ( [ 11.22 ] ) for @xmath224 is equal to the @xmath69-derivative of the l.h.s . of ( [ 11.13 ] ) . moreover , when @xmath222 , the l.h.s . of ( [ 11.22 ] ) converges to 1 , so if it is @xmath225 for @xmath224 , it has to vanish for some @xmath226 . hence , as in the case @xmath403 , we get : [ section ] assume that the intervals ( [ 11.15 ] ) are disjoint and let @xmath227 be a fixed point of ( [ 11.17 ] ) . then @xmath227 is attractive precisely when the @xmath69-derivative of the l.h.s . of ( [ 11.14 ] ) is @xmath228 . * proof . * the @xmath69-derivative of the l.h.s . of ( [ 11.14 ] ) is @xmath416 where @xmath417 . on the other hand , in the small-@xmath229 limit , the equation ( [ 11.22 ] ) for the eigenvalues of the linearization becomes as in section [ 8 ] @xmath418 we are only interested in the possible existence of solutions to this equation with @xmath419 , and for such @xmath221 ( [ 11.24 ] ) reduces to @xmath420 if @xmath221 is a solution , then by the condition that the intervals ( [ 11.15 ] ) are disjoint , only one term in the last sum , say for @xmath421 , is @xmath422 , so that ( [ 11.25 ] ) becomes @xmath423 while the expression ( [ 11.23 ] ) becomes @xmath424 it is then easy to see that the solution of ( [ 11.26 ] ) has a negative real part precisely when the expression ( [ 11.27 ] ) is positive , and this concludes the proof of the last proposition . when the intervals ( [ 11.15 ] ) have non - empty intersections , the situation is more complicated , and the following example is an indication that the last proposition may be false . + _ example . _ there exist @xmath425 , @xmath426 , such that @xmath427 , while @xmath428 for some positive @xmath221 . indeed , choose @xmath429 , @xmath430 , @xmath431 very small , @xmath432 . then @xmath433 . if @xmath434 , we have @xmath435 hence @xmath436 must vanish for some @xmath221 between @xmath309 and @xmath437 . as in section [ 9 ] , we can derive a simplified differential equation for @xmath438 , where @xmath439 , so that @xmath440 . we drop the simplifying assumption ( [ 11.16 ] ) , but keep , for simplicity , the assumption that @xmath396 are independent of @xmath153 . assume that @xmath441 is a solution of ( [ 11.11 ] ) which has existed for a long time with a uniformly bounded norm . as in section [ 9 ] , we take the scalar product of ( [ 11.11 ] ) with @xmath442 and get @xmath443 using @xmath444 and , under the assumption that @xmath445 is slowly varying , we get as in section [ 9 ] @xmath446 and the simplified equations @xmath447 , \label{11.31}\ ] ] for @xmath351 , and @xmath448 . we notice that the region defined by @xmath449 for @xmath450 is stable under the forward flow associated to the system ( [ 11.31 ] ) . moreover , if @xmath451 is a fixed point of this system , then we get precisely ( [ 11.14 ] ) . conversely , if @xmath69 is a solution of ( [ 11.14 ] ) , then @xmath452 defines the corresponding unique fixed point solution with @xmath448 . we end this section by investigating the linearization of ( [ 11.31 ] ) at a fixed point solution , under the simplifying assumption ( [ 11.16 ] ) . an easy calculation shows that the linearization is given by @xmath453 if @xmath221 is an eigenvalue of @xmath454 with @xmath455 for every @xmath354 , and @xmath456 a corresponding non - trivial eigenvector , we have @xmath457 then necessarily the sum is @xmath422 , and by summing these @xmath371 relations , we see that @xmath221 is an eigenvalue precisely when ( [ 11.25 ] ) holds . we finally get : [ p11.1]proposition under the simplifying assumption ( [ 11.16 ] ) and in the small-@xmath229 limit , let @xmath227 be a fixed point of ( [ 11.17 ] ) and let @xmath451 be the corresponding fixed point solution of ( [ 11.31 ] ) . then the linearizations of ( [ 11.17 ] ) and ( [ 11.31 ] ) at the corresponding fixed points have the same eigenvalues in the right half plane @xmath419 ( given by ( [ 11.25 ] ) ) . in particular , @xmath227 is an attractive fixed point for ( [ 11.17 ] ) precisely when @xmath458 is an attractive fixed point for ( [ 11.31 ] ) . we thank g. jona - lasinio for interesting discussions on the conceptual aspects of the physical model . we also thank g. perelman for interesting discussions on the mathematical aspects and acknowledge that she independently found the basic differential equation ( [ simplediff ] ) . partial support of infn , iniziativa specifica rm6 , is acknowledged . the crystals which compose the heterostructure may have different periodic potentials giving rise to a layer - dependent effective mass as discussed , for example , by g. t. einevoll , p. c. hemmer , and j. thomsen , phys . b * 42 * , 3485 ( 1990 ) . here , we will suppose to have a constant effective mass which is a reasonable assumption for gaas - algaas heterostructures .

as is well known , mixing offers a profound probe into the effects of physics beyond the standard model . the data obtained at the factories have already provided valuable insights into the-meson system , and very recently also the oscillation frequency has been measured at the tevatron . we give a critical discussion of the interpretation of these data in terms of model - independent new - physics parameters . we address in particular the impact of the uncertainties of the relevant input parameters , set benchmarks for their accuracies as required by future precision measurements at the lhc , and explore the prospects for new cp - violating effects in the system . to complement our model - independent analysis , we also discuss the constraints imposed by the cdf measurement of on popular models of new physics , namely scenarios with an extra boson and supersymmetry . we find that the new data still leave sizeable room for new - physics contributions to mixing , which could be detected at the lhc . ippp/06/23 + dcpt/06/46 + cern ph th/2006065 + hep - ph/0604249 * patricia ball and robert fleischer * _ ippp , department of physics , university of durham , durham dh1 3le , uk _ _ theory division , department of physics , cern , ch-1211 geneva 23 , switzerland _ april 2006

hep-ph0604249

one of the most promising ways to detect the effects of new physics ( np ) on @xmath3 decays is to look for deviations of flavour - changing neutral - current ( fcnc ) processes from their standard model ( sm ) predictions ; fcnc processes only occur at the loop - level in the sm and hence are particularly sensitive to np virtual particles and interactions . a prominent example that has received extensive experimental and theoretical attention is @xmath15@xmath16 mixing ( @xmath17 ) , which , in the sm , is due to box diagrams with @xmath18-boson and up - type quark exchange . in the language of effective field theory , these diagrams induce an effective hamiltonian , which causes @xmath15 and @xmath16 mesons to mix and generates a @xmath19 transition : @xmath20 where @xmath21 is the @xmath22-meson mass . thanks to @xmath15@xmath16 mixing , an initially present @xmath15 state evolves into a time - dependent linear combination of @xmath15 and @xmath16 flavour states . the oscillation frequency of this phenomenon is characterized by the mass difference of the heavy " and light " mass eigenstates , @xmath23 and the cp - violating mixing phase @xmath24 which enters mixing - induced " cp violation . the mass difference @xmath25 can be and has been measured from the proper - time distribution of @xmath15 candidates identified through their decays into ( mostly ) flavour - specific modes , after having been tagged as mixed or unmixed . the current experimental results are @xmath26\,{\rm ps}^{-1}\,,\ ] ] where the value of @xmath27 is the world average quoted by the heavy flavour averaging group " ( hfag ) @xcite . concerning @xmath7 , only lower bounds were available for many years from the lep experiments at cern and sld at slac @xcite . since the currently operating @xmath2 @xmath3 factories run at the @xmath28 resonance , which decays into @xmath29 , but not into @xmath8 mesons , the @xmath8 system can not be explored by the babar and belle experiments . however , plenty of @xmath8 mesons are produced at the tevatron ( and later on will be at the lhc ) , which very recently allowed the cdf collaboration to measure @xmath7 with the result given above @xcite ; the d0 collaboration has provided , also very recently , a two - sided bound on @xmath7 at the 90% c.l . @xcite : @xmath30 which is compatible with the cdf measurement and corresponds to a @xmath31 signal at @xmath32 . these new results from the tevatron have already triggered a couple of phenomenological papers @xcite@xcite . in the sm , @xmath33 is given by @xmath34 where @xmath35 is fermi s constant , @xmath36 the mass of the @xmath18 boson , @xmath37 a short - distance qcd correction ( which is the same for the @xmath38 and @xmath5 systems ) @xcite , whereas the bag " parameter @xmath39 and the decay constant @xmath40 are non - perturbative quantities . @xmath41 and @xmath42 are elements of the cabibbo kobayashi maskawa ( ckm ) matrix @xcite , and @xmath43 with @xmath44 @xcite is one of the inami lim " functions @xcite , describing the @xmath45-quark mass dependence of the box diagram with internal @xmath45-quark exchange ; the contributions of internal @xmath46 and @xmath47 quarks are , by virtue of the glashow iliopoulos maiani ( gim ) mechanism @xcite , suppressed by @xmath48 . the mixing phases @xmath49 can be measured from mixing - induced " cp asymmetries . in the sm , one has @xmath50 where @xmath51 is the usual angle of the conventional " unitarity triangle ( ut ) of the ckm matrix , while @xmath52 characterizes another unitarity triangle @xcite that differs from the ut through @xmath53 corrections in the wolfenstein expansion @xcite . processes carry essentially no cp - violating weak phase . physical cp asymmetries are of course independent of the applied ckm phase convention , as shown explicitly in ref . @xcite . ] the purpose of this paper is to explore the possibility that @xmath15@xmath16 mixing is modified by np contributions at the tree level and/or new particles in the loops . we shall find in particular that despite the apparently strong constraints posed by the precise measurements of @xmath25 in eq . ( [ exp ] ) these results can contain potentially large np contributions , which presently can not be detected . the outline of our paper is as follows : in section [ sec : input ] , we collect the input parameters of our analysis and discuss the status of the relevant hadronic uncertainties . in sections [ sec : bd ] and [ sec : bs ] , we then focus on the @xmath4- and @xmath8-meson systems , respectively , and investigate , in a model - independent way , the size of possible np contributions to @xmath25 and @xmath49 in the light of present and future experimental measurements and hadronic uncertainties . in this analysis , we consider also a scenario for the experimental and theoretical situation in the year 2010 , and set benchmarks for the required accuracy of the relevant hadronic parameters . it turns out that the situation in the @xmath8 system is more favourable than in the @xmath4 system , and that still ample space for np effects in @xmath5@xmath6 mixing is left , which could be detected at the lhc . in section [ sec : np - models ] , we complement the model - independent discussion of sections [ sec : bd ] and [ sec : bs ] by analyses of two specific scenarios for np : models with an extra @xmath9 boson and supersymmetry ( susy ) with an approximate alignment of quark and squark masses . we summarize our conclusions in section [ sec : concl ] . before going into the details of @xmath15@xmath16 mixing and possible np effects , let us first have a closer look at the relevant input parameters and their uncertainties . throughout our analysis , we assume that the ckm matrix is unitary , and shall use this feature to express the ckm elements entering @xmath15@xmath16 mixing in terms of quantities that can be determined through tree - level processes of the sm . the key rle is then played by @xmath54 and @xmath55 . the former quantity is presently known with 2% precision from semileptonic @xmath3 decays ; in this paper we shall use the value obtained in ref . @xcite from the analysis of leptonic and hadronic moments in inclusive @xmath56 transitions @xcite : latexmath:[\[\label{vcb } from exclusive decays . the situation is less favourable with @xmath55 : there is a 1@xmath58 discrepancy between the values from inclusive and exclusive @xmath59 transitions @xcite : @xmath60 is dominated by the theoretical uncertainty of lattice and light - cone sum rule calculations of @xmath61 and @xmath62 transition form factors @xcite , whereas for @xmath63 experimental and theoretical errors are at par . we will use both results in our analysis . whereas any improvement of the error of @xmath54 will have only marginal impact on the analysis of @xmath3 mixing , a reduction of the uncertainty of @xmath55 will be very relevant . as a benchmark scenario for the situation in 2010 , we will assume that the central value of @xmath63 gets confirmed and that its uncertainty will shrink to @xmath64 , i.e. 5% , thanks to better statistics and an increased precision of theoretical predictions , for instance from further developments in the dressed gluon exponentiation @xcite . the next ingredient in the sm prediction for @xmath66 are the non - perturbative matrix elements @xmath67 . these parameters have been the subject of numerous lattice calculations , both quenched and unquenched , using various lattice actions and implementations of both heavy and light quarks . the current front runners are unquenched calculations with 2 and 3 dynamical quarks , respectively , and wilson or staggered light quarks . despite tremendous progress in recent years , the results still suffer from a variety of uncertainties which is important to keep in mind when interpreting and using lattice results . one particular difficulty in determining @xmath68 is the chiral extrapolation needed to go to the physical @xmath69-quark mass.- and @xmath69-quark masses and use @xmath70 . the physical value of the light - quark mass ratio is then @xmath71 from chiral perturbation theory @xcite . ] lattice calculations are usually performed at unphysically large @xmath47- and @xmath69-quark masses , as the simulation of dynamical fermions involves many inversions of the fermions functional determinant in the path integral and is very dear in terms of cpu time . therefore , an extrapolation , called the chiral extrapolation , in the light - quark masses from feasible to physical masses is necessary , which is done using the functional form predicted by chiral perturbation theory . based on these arguments , the chiral extrapolation of @xmath72 to the physical limit is expected to be smooth , whereas that of @xmath68 is potentially prone to logarithms @xcite , which leads to a considerable increase in the uncertainty . the most recent ( unquenched ) simulation by the jlqcd collaboration @xcite , with non - relativistic @xmath73 quarks and two flavours of dynamical light ( wilson ) quarks , yields @xmath74gev and @xmath75 where the first error includes uncertainties from statistics and various systematics , whereas the second , asymmetric error comes from the chiral extrapolation . note that part of the systematic errors cancel in the ratio @xmath76 . in this calculation , the ratio @xmath77 was varied between 0.7 and 2.9 . more recently , ( unquenched ) simulations with three dynamical flavours have become possible using staggered quark actions . the hpqcd collaboration obtains @xmath78mev @xcite , where a ratio of @xmath77 as small as 0.125 could be achieved , due to the good chiral properties of the staggered action . this implies that the chiral extrapolation is less critical and the corresponding error much smaller . the quoted error on @xmath40 is now dominated by yet uncalculated higher - order matching terms which are needed to match the ( effective theory ) lattice calculations to continuum qcd . lacking any direct calculation of @xmath39 with three dynamical flavours , and in view of the fact that the bag parameter is likely to be less sensitive to chiral extrapolation , it has been suggested to combine the results of @xmath40 from hpqcd with that of @xmath39 from jlqcd , yielding @xcite : @xmath79 where all errors are added in quadrature . although we shall use both ( [ jlqcd ] ) and ( [ hpqcd ] ) in our analysis , we would like to stress that the errors are likely to be optimistic . apart from the issue of the chiral extrapolation discussed above , there is also the question of discretisation effects ( jlqcd uses data obtained at only one lattice spacing ) and the renormalisation of matrix elements ( for lattice actions without chiral symmetry , the axial vector current is not conserved and @xmath40 needs to be renormalised ) , which some argue should be done in a non - perturbative way @xcite . simulations with staggered quarks also face potential problems with unitarity , locality and an odd number of flavours ( see , for instance , ref . . a confirmation of the hpqcd results by simulations using the ( theoretically better understood ) wilson action with small quark masses will certainly be highly welcome . given this situation , we consider it not very likely that the errors on @xmath40 , @xmath80 and @xmath76 will come down considerably in the near future . for our benchmark 2010 scenario , we hence will assume the values of hadronic parameters and uncertainties given in ( [ hpqcd ] ) . we are now well prepared for the discussion of @xmath15@xmath16 mixing . let us first have a closer look at the @xmath38@xmath81 mixing parameters . in the presence of np , the matrix element @xmath82 can be written , in a model - independent way , as @xmath83 where the real parameter @xmath84 measures the `` strength '' of the np contribution with respect to the sm , whereas @xmath85 is a new cp - violating phase ; analogous formulae apply to the @xmath8 system . the @xmath4 mixing parameters then read @xmath86,\label{m - d}\\ \phi_d & = & \phi_d^{\rm sm}+\phi_d^{\rm np}= \phi_d^{\rm sm } + \arg ( 1+\kappa_d e^{i\sigma_d})\,.\label{phi - d}\end{aligned}\ ] ] the experimental result for @xmath27 and the theoretical prediction @xmath87 provide the following constraint on @xmath88 and @xmath85 : @xmath89 which determines , for instance , @xmath88 as function of @xmath85 : @xmath90 in fig . [ fig : kappa - rho - d ] , we illustrate the corresponding contours in the @xmath85@xmath88 plane for values of @xmath91 between 0.6 and 1.4 , varied in steps of 0.1 . interestingly enough , a value of @xmath91 smaller than 1 imposes a constraint on the weak np phase @xmath85 : @xmath92 = 0.47 in order to make use of these constraints , one needs to know the sm prediction @xmath93 . in particular , one has to make sure that the parameters entering @xmath94 , eq . ( [ m12sm ] ) , are free from np . this can be achieved , to very good accuracy , by expressing the relevant ckm factor in @xmath93 in terms of parameters measured in tree - level processes . to this end , we use the wolfenstein parametrization @xcite , as generalized in ref . @xcite , and the unitarity of the ckm matrix to write @xmath95 is given by @xmath96 with @xmath97 @xmath98 measures one side of the ut , and @xmath99 denotes the usual ut angle . as we saw in section [ sec : input ] , @xmath54 and @xmath55 can be determined from semileptonic @xmath3 decays , which arise at tree level in the sm and hence are very robust with respect to np effects . a similar comment applies to the wolfenstein parameter @xmath100 @xcite , which can be determined , for instance , from @xmath101 decays . using the values of @xmath54 and @xmath55 discussed in section [ sec : input ] and @xmath102 @xcite , we obtain @xmath103 where the labels incl " and excl " refer to the determinations of @xmath55 through inclusive and exclusive @xmath104 transitions , respectively . the angle @xmath99 can be determined in a variety of ways through cp - violating effects in pure tree decays of type @xmath105 @xcite . using the present @xmath3-factory data , the following results were obtained through a combination of various methods : @xmath106 ( 65\pm 20)^\circ & \mbox{(utfit collaboration \cite{utfit})}. \end{array } \right.\ ] ] a more precise value for @xmath99 was obtained in ref . @xcite , from the @xmath3-factory data on cp asymmetries in @xmath107 and @xmath108 decays , which receive both tree and penguin contributions : @xmath109 within the np scenario of modified electroweak penguins considered in ref . @xcite , ( [ gam - pik ] ) is not affected by np effects . the central value of ( [ gam - pik ] ) is higher than that of ( [ gam - dk ] ) , but both results are perfectly consistent because of the large errors of the @xmath105 determinations . an even larger value of @xmath99 in the ballpark of @xmath110 was recently extracted from @xmath111 data with the help of soft collinear effective theory " ( scet ) @xcite . in our analysis , we use the utfit value @xmath112 with the help of ( [ vcb ] ) , ( [ ckm - bd ] ) and ( [ rb ] ) , we then obtain latexmath:[\[|v_{td}^ * v_{tb}|_{\rm incl } = ( 8.6\pm 1.5)\cdot 10^{-3}\,,\qquad uncertainty is dominated by that of the angle @xmath99 . for our 2010 benchmark scenario , we assume that the central value of @xmath99 will settle at @xmath114 , and that the error will shrink to @xmath115 thanks to strategies using pure tree decays of @xmath29 and @xmath8 mesons for the determination of @xmath99 , which can be implemented at the lhc . in fact , a statistical accuracy of @xmath116 is expected at lhcb after 5 years of taking data @xcite . for the convenience of the reader , we summarise all ckm input parameters , as well as their counterparts for the @xmath8 system to be discussed in section [ sec : bs ] , in tab . [ tab : ckm ] ; in tab . [ tab:2010 ] , we give the input data for our 2010 scenario . @xmath117 @xmath118 @xmath119 in fig . [ fig : dmd ] , we illustrate the dependence of @xmath91 defined in ( [ rhod - def ] ) on @xmath99 , @xmath98 and @xmath120 . it is evident that @xmath91 depends rather strongly on @xmath99 and @xmath120 , but less so on @xmath98 . for the two different lattice results , we obtain @xmath121\,{\rm ps}^{-1}\,,\nonumber\\ \left.\rho_d\right|_{\rm jlqcd } & = & 0.97\pm0.33(\gamma , r_b)^{-0.17}_{+0.26}(f_{b_d } \hat b_{b_d}^{1/2})\,,\nonumber\\ \left.\delta m_d^{\rm sm}\right|_{\rm ( hp+jl)qcd } & = & \left[0.69\pm 0.13(\gamma , r_b)\pm 0.08(f_{b_d } \hat b_{b_d}^{1/2})\right]\,{\rm ps}^{-1}\,,\nonumber\\ \left.\rho_d\right|_{\rm ( hp+jl)qcd } & = & 0.75\pm0.25(\gamma , r_b)\pm0.16(f_{b_d } \hat b_{b_d}^{1/2})\,,\label{dmd - sm - numbers}\end{aligned}\ ] ] where we made explicit the errors arising from the uncertainties of @xmath122 , @xmath123 and @xmath124 . these results are compatible with the sm value @xmath125 , but suffer from considerable uncertainties , which presently leave sizeable room for np contributions to @xmath27 ; we shall quantify below the allowed values of @xmath88 and @xmath85 following from the contours in fig . [ fig : kappa - rho - d ] . @xmath126 @xmath127 the second constraint on the allowed values of @xmath88 and @xmath85 is provided by the experimental value of the @xmath4 mixing phase @xmath129 . using ( [ phi - d ] ) , a given value of @xmath130 allows one to determine @xmath88 as a function of @xmath85 with the help of the following expressions , which hold again in the general case @xmath17 : @xmath131 @xmath132 in fig . [ fig : kappa - sigma ] , we illustrate the corresponding contours for various values of @xmath133 . note in particular that @xmath134 is bounded from below for any given value of @xmath135 . the relation between the allowed values of @xmath136 and @xmath134 is given by @xmath137 i.e. for any non - zero value of @xmath133 , @xmath134 must be larger than the minimum value plotted in fig . [ fig : kappamin ] . in order to make use of these theoretically clean contours , one needs to determine the np phase @xmath133 . as is well known , @xmath128 can be experimentally accessed in the mixing - induced cp asymmetry of the golden " decay @xmath138 ( and similar @xmath139 charmonium modes ) @xcite . the most recent average of the @xmath3-factory data for such transitions obtained by hfag is @xcite @xmath140 in principle , this quantity could be affected by np contributions to both @xmath38@xmath81 mixing and @xmath141 decay amplitudes @xcite . a probe of the latter effects is provided by decays like @xmath142 , which are pure tree decays and do not receive any penguin contributions . if the neutral @xmath143 mesons are observed through their decays into cp eigenstates @xmath144 , these decays allow an extremely clean determination of the true " value of @xmath145 @xcite . a possible discrepancy with @xmath146 would be attributed to np contributions to the @xmath141 decay amplitudes . consequently , detailed feasibility studies for the exploration of @xmath147 modes at a super-@xmath3 factory are strongly encouraged . in this paper , however , we assume that np effects entering decay amplitudes are negligible . ( [ phi - d ] ) then gives the following expression : @xmath148 the experimental value ( [ s2b - exp ] ) yields the twofold solution @xmath149 where the latter result is in dramatic conflict with global ckm fits and would require a large np contribution to @xmath38@xmath81 mixing @xcite . however , experimental information on the sign of @xmath150 rules out a negative value of this quantity at greater than 95% c.l.@xcite , so that we are left with @xmath151 . the sm prediction of the mixing phase , @xmath152 , eq . ( [ phi - sm ] ) , can easily be obtained in terms of the tree - level quantities @xmath98 and @xmath99 , as @xmath153 using eq . ( [ phi - d ] ) , the experimental value of @xmath128 can then immediately be converted into a result for the np phase @xmath154 , which depends on @xmath99 and @xmath98 as illustrated in fig . [ fig : phi - np - d ] . @xmath155 it is evident that the dependence of @xmath156 on @xmath99 is very small and that @xmath98 plays actually the key rle for its determination . hence , we have a situation complementary to that shown in fig . [ fig : dmd ] , where the main dependence was on @xmath99 . the parameters collected in tab . [ tab : ckm ] yield @xmath157 corresponding to @xmath158 results of @xmath159 were also recently obtained in refs . note that the emergence of a non - zero value of @xmath130 is caused by the large value of @xmath55 from inclusive semileptonic decays , but that @xmath130 is compatible with zero for @xmath55 from exclusive decays . the consequences of the presence of a small np phase @xmath159 are rather dramatic : from fig . [ fig : kappamin ] , one reads off the sizeable lower bound @xmath160 . although this result hinges on the value of @xmath63 , and hence presently is not conclusive , the underlying reasoning also applies to the @xmath8 system : even a small np phase @xmath161 implies considerable np contributions to the mixing matrix element @xmath162 . @xmath163 @xmath164 we are now finally in a position to combine the constraints from both @xmath27 and @xmath128 to constrain the allowed region in the @xmath85@xmath88 plane . the corresponding results are shown in fig . [ fig : res - k - sig - d ] , demonstrating the power of the contours described in the previous subsections for a transparent determination of @xmath85 and @xmath88 . we see that a non - vanishing value of @xmath130 , even as small as @xmath165 , has a strong impact on the allowed space in the @xmath85@xmath88 plane . in both scenarios with different lattice results and different values for @xmath55 , the upper bounds of @xmath166 on the np contributions following from the experimental value of @xmath27 are reduced to @xmath167 . values of this order of magnitude are expected , for instance , on the basis of generic field - theoretical considerations @xcite , as well as in a recently proposed framework for next - to - minimal flavour violation " @xcite . in order to determine @xmath88 more precisely , it is mandatory to reduce the errors of @xmath91 , which come from both @xmath99 and lattice calculations . as we noted above , the value of @xmath99 can be determined with impressive accuracy at the lhc @xcite , whereas progress on the lattice side is much harder to predict , but will hopefully be made . assuming our benchmark scenario of tab . [ tab:2010 ] , which corresponds to the lattice results of eq . ( [ hpqcd ] ) , the @xmath85@xmath88 plane in 2010 looks like shown in fig . [ fig : future - d ] and actually implies @xmath168 evidence for np from @xmath169 . although there is only a small allowed region left , @xmath88 is still only badly constrained ; for an extraction with 10% uncertainty , @xmath124 is required to 5% accuracy , i.e. the corresponding error in ( [ hpqcd ] ) has to be reduced by a factor of 2 , which is the benchmark lattice theorists should strive for . let us now have a closer look at the @xmath8-meson system . in order to describe np effects in a model - independent way , we parametrize them analogously to ( [ m - d ] ) and ( [ phi - d ] ) . the relevant ckm factor is @xmath170 . using once again the unitarity of the ckm matrix and including next - to - leading order terms in the wolfenstein expansion as given in ref . @xcite , we have @xmath171 consequently , apart from the tiny correction in @xmath172 , the ckm factor for @xmath7 is independent of @xmath99 and @xmath98 , which is an important advantage in comparison with the @xmath4-meson system . the accuracy of the sm prediction of @xmath7 is hence limited by the hadronic mixing parameter @xmath173 . using the numerical values discussed in section [ sec : input ] , we obtain @xmath174 where we made the experimental and theoretical errors explicit . the values of @xmath175 , which is defined in analogy to ( [ rhod - def ] ) , refer to the cdf measurement of @xmath7 in ( [ exp ] ) . these numbers are consistent with the sm case @xmath176 , but suffer from significant theoretical uncertainties , which are much larger than the experimental errors . nevertheless , it is interesting to note that the ( hp+jl)qcd result is @xmath177 below the sm ; a similar pattern arises in ( [ dmd - sm - numbers ] ) , though at the @xmath178 level . any more precise statement about the presence or absence of np requires the reduction of theoretical uncertainties . in fig . [ fig : mds - np ] , we show the constraints in the @xmath179@xmath180 plane , which can be obtained from @xmath175 with the help of the @xmath8 counterpart of ( [ kappa - sig - rho ] ) . we see that upper bounds of @xmath181 arise from the measurement of @xmath7 . in the case of ( [ 36 ] ) , the bound on @xmath179 following from ( [ sigma - bound ] ) would interestingly be effective , and imply @xmath182 . consequently , the cdf measurement of @xmath7 leaves ample space for the np parameters @xmath179 and @xmath180 . this situation will change significantly as soon as information about cp violation in the @xmath8-meson system becomes available . we shall return to this topic in subsection [ ssec : bs - cp ] . @xmath183 it is interesting to consider the ratio of @xmath7 and @xmath27 , which can be written as follows : @xmath184 where the hadronic @xmath185-breaking parameter @xmath76 is defined in subsection [ ssec : hadr - uncert ] . in the class of np models with minimal flavour violation " @xcite , measurement . ] which contains also the sm , we have @xmath186 , so that ( [ dms - dmd - rat ] ) allows the extraction of the ckm factor @xmath187 , and hence @xmath188 , as @xmath189 is known to excellent accuracy from ( [ vts ] ) . the advantage of this determination lies in the reduced theoretical uncertainty of @xmath76 as compared to @xmath124 . in this paper , however , we turn the tables and constrain the ratio @xmath190 through @xmath191 . to this end , we express in analogy to ( [ ckm - bd ] ) the ut side @xmath192\ ] ] in terms of @xmath98 and @xmath99 : @xmath193 allowing the determination of @xmath194 through processes that are essentially unaffected by np . the resulting value of @xmath194 depends rather strongly on @xmath99 , which is the main source of uncertainty . another determination of @xmath194 that is independent of @xmath99 and @xmath98 can , in principle , be obtained from radiative decays , in particular the ratio of branching ratios @xmath195 , but is presently limited by experimental statistics ; see ref . @xcite for a recent analysis . @xmath196 combining ( [ dms - dmd - rat ] ) and ( [ rt - def ] ) , we obtain the following expression for @xmath190 : @xmath197 \left[1+(1 - 2r_b\cos\gamma)\lambda^2+{\cal o}(\lambda^4)\right ] \frac{1}{\xi^2}\frac{m_{b_d}}{m_{b_s } } \frac{\delta m_s}{\delta m_d}\,.\ ] ] in fig . [ fig : rhos - rhod ] , we plot this ratio for the central values of @xmath27 and @xmath7 in ( [ exp ] ) , as a function of the ut angle @xmath99 for the values of @xmath76 given in ( [ jlqcd ] ) and ( [ hpqcd ] ) . we find that the corresponding curves are nearly independent of @xmath98 and that @xmath99 is actually the key ckm parameter for the determination of @xmath190 . the corresponding numerical values are given by : @xmath198 because of the large range of allowed values of @xmath99 , eq . ( [ gamma ] ) , this ratio is currently not very stringently constrained . this situation should , however , improve significantly in the lhc era thanks to the impressive determination of @xmath99 to be obtained at the lhcb experiment . for our 2010 scenario as specified in tab . [ tab:2010 ] , which corresponds to the right panel of fig . [ fig : rhos - rhod ] with @xmath199 , we find : @xmath200 where we made the errors arising from the uncertainties of @xmath99 and @xmath76 explicit , and , in the last step , added them in quadrature . consequently , the hadronic uncertainties and those induced by @xmath99 would now be of the same size , which should provide additional motivation for the lattice community to reduce the error of @xmath76 even further . despite the impressive reduction of uncertainty compared to the 2006 values in ( [ 2006 ] ) , the numerical value in ( [ rhos - rhod-2010 ] ) would still not allow a stringent test of whether @xmath190 equals one : to establish a @xmath201 deviation from 1 , central values of @xmath202 or 0.7 would be needed . the assumed uncertainty of @xmath99 of @xmath203 could also turn out to be too pessimistic , in which case even more progress would be needed from the lattice side to match the experimental accuracy . the result in ( [ rhos - rhod-2010 ] ) would not necessarily suggest that there is no physics beyond the sm . in fact , and as can be seen from tab . [ tab:2010 ] , the central values of @xmath91 and @xmath175 would both be smaller than one , i.e. would both deviate from the sm picture , although the hadronic uncertainties would again not allow us to draw definite conclusions . in order to shed further light on these possible np contributions , the exploration of cp - violating effects in the @xmath8-meson system is essential . to date , the cp - violating phase associated with @xmath5@xmath6 mixing is completely unconstrained . in the sm , it is doubly cabibbo - suppressed , and can be written as follows : @xmath204 here we used again ( [ rho - bar - eta - bar ] ) to express the wolfenstein parameter @xmath205 in terms of @xmath98 and @xmath99 . because of the small sm phase in ( [ phis - sm ] ) , @xmath5@xmath6 mixing is particularly well suited to search for np effects , which may well lead to a sizeable value of @xmath206 @xcite . in order to test the sm and probe cp - violating np contributions to @xmath5@xmath6 mixing , the decay @xmath207 , which is very accessible at the lhc @xcite , plays a key rle . thanks to mixing - induced cp violation in the time - dependent angular distribution of the @xmath208\phi[\to k^+k^-]$ ] decay products , the quantity @xmath209 can be measured @xcite , in analogy to the determination of @xmath145 through @xmath138 . after one year of data taking ( which corresponds to 2 @xmath210 ) , lhcb expects a measurement with the statistical accuracy @xmath211 ; adding modes such as @xmath212 and @xmath213 , @xmath214 is expected after five years @xcite . also atlas and cms will contribute to the measurement of @xmath215 , expecting uncertainties at the 0.1 level after one year of data taking , which corresponds to 10 @xmath210 @xcite . [ cols="^,^ " , ] in order to illustrate the impact of np effects , let us assume that the np parameters satisfy the simple relation @xmath216 i.e. that in particular @xmath217 . this scenario would be supported by ( [ rhos - rhod-2010 ] ) , although it would _ not _ belong to the class of models with mfv , as new sources of cp violation would be required . as we have seen in the previous section , the analysis of the @xmath38 data for @xmath218 indicates a small np phase around @xmath219 in the @xmath4-system . in the above scenario , that would imply the presence of the same phase in the @xmath8-system , which would interfere constructively with the small sm phase and result in cp asymmetries at the level of @xmath220 . cp - violating effects of that size can easily be detected at the lhc . this exercise demonstrates again the great power of the @xmath8-meson system to reveal cp - violating np contributions to @xmath15@xmath16 mixing . the presence of a small np phase could actually be considerably magnified , as illustrated in fig . [ fig : sinphis ] . in specific np scenarios , also large cp - violating phases can still arise , and are in no way excluded by the cdf measurement of @xmath7 in ( [ exp ] ) . let us finally also discuss the impact of cp violation measurements on the allowed region in the @xmath179@xmath180 plane in our 2010 scenario . to this end , we consider two cases : * @xmath222 , in accordance with the sm ; * @xmath223 , in accordance with the np scenario of fig . [ fig : sinphis ] . the measurement of @xmath215 implies a twofold solution for @xmath206 and , therefore , also for @xmath161 . however , this ambiguity can be resolved through the determination of the sign of @xmath224 , which can be fixed through the strategies proposed in ref . @xcite . in fig . [ fig : sis - kas - cp ] , we show the situation in the @xmath179@xmath180 plane . , as our 2010 scenario is based on the ( hp+jl)qcd lattice results . ] the dotted lines refer to negative values of @xmath224 . assuming that these are experimentally excluded , we are left with strongly restricted regions , although @xmath180 could still take sizeable ranges , with upper bounds @xmath225 . in the sm - like scenario , values of @xmath179 around @xmath226 would arise , i.e. a np contribution with a sign opposite to the sm . however , due to the absence of new cp - violating effects , the accuracy of lattice results would have to be considerably improved in order to allow the extraction of a value of @xmath180 incompatible with 0 . on the other hand , a measurement of @xmath223 would give a np signal at the @xmath227 level , with @xmath228 from eq . ( [ phimaxmin ] ) . in analogy to the discussion in subsection [ ssec : comb - bd ] , a determination of @xmath180 with 10% uncertainty requires @xmath229 with 5% accuracy , i.e. the corresponding error in ( [ hpqcd ] ) has to be reduced by a factor of 2 . since the discussion given so far does not refer to a specific model of np , the question arises whether there are actually extensions of the sm that still allow large cp - violating np phases in @xmath5@xmath6 mixing . in this section , we address the impact of the cdf measurement of @xmath7 on two popular scenarios of np , to wit * an extra @xmath9 boson with flavour non - diagonal couplings ; * generic effects in the minimal supersymmetric extension of the sm ( mssm ) in the `` mass insertion approximation '' . we would like to stress that our examples for np scenarios should be viewed as illustrative rather than comprehensive and are not intended to compete with more dedicated analyses . let us start with the effect of an extra @xmath230 gauge boson @xmath9 , which is the most simple application of the model - independent method discussed in sections [ sec : bd ] and [ sec : bs ] . the existence of a new @xmath9 gauge boson can induce fcnc processes at tree - level if the @xmath9 coupling to physical fermions is non - diagonal . such @xmath9 bosons often occur , for instance , in the context of grand unified theories ( guts ) , superstring theories , and theories with large extra dimensions , see , for instance , refs . @xcite . in this paper , we illustrate the constraints on an extra @xmath9 under the conditions that * the @xmath231 couplings stay flavour diagonal , i.e.@xmath231@xmath9 mixing is negligible and the @xmath231 does not contribute to @xmath3 mixing ; * the @xmath9 has flavour non - diagonal couplings only to left - handed quarks , which means that its effect is described by only one complex parameter . note that the @xmath9 contribution to @xmath8 mixing is related to that for hadronic , leptonic and semileptonic decays in specific models where the @xmath9 coupling to light quarks and leptons is known ; in this paper , however , we treat the @xmath9 in a model - independent way and assume its couplings to the @xmath232 and @xmath233 quark fields as independent . we only discuss the @xmath8-system and closely follow the notations of ref . @xcite , where an earlier analysis of this scenario was given . a purely left - handed off - diagonal @xmath9 coupling to @xmath73 and @xmath234 quarks gives the following contribution to @xmath235 : should be taken at lo accuracy ; here , we effectively absorb the ( small ) difference between lo and nlo expressions into the definition of @xmath236 . ] @xmath237 where @xmath238 is defined in terms of the sm @xmath239 gauge coupling @xmath240 , the @xmath230 coupling @xmath241 , the respective gauge boson masses @xmath242 and the fcnc coupling @xmath243 of the @xmath9 to @xmath232 and @xmath233 . generically , one would expect @xmath244 , if both @xmath245 groups have the same origin , for instance in a gut framework , and @xmath246 for a tev - scale @xmath9 . if in addition the size of the @xmath9 couplings @xmath247 is set by the quarks yukawa couplings , one also expects @xmath248 and @xmath249 . the impact of the cdf measurement of @xmath7 on this model can be directly read off fig . [ fig : mds - np ] through the identifications @xmath250 with @xmath251 presently , values of @xmath180 as large as 2.5 are still allowed , see fig . [ fig : mds - np ] , which corresponds to @xmath252 if a non - zero value of the np phase @xmath161 should be measured at the lhc , this value can be immediately translated into a lower bound on @xmath236 , using ( [ phimaxmin ] ) . assuming @xmath253 , one has @xmath254 and @xmath255 . any more precise constraint on @xmath236 will depend on the progress in lattice determinations of @xmath173 . the upper bound on @xmath236 given in eq . ( [ 47 ] ) can be converted into a lower bound on the @xmath9 mass : @xmath256 in the scenario of ( [ 48 ] ) , there is also an upper bound and the lower bound is raised : @xmath257 we would like to stress again that these bounds apply to a model where the @xmath9 has flavour non - diagonal couplings only to left - handed quarks . ( [ zprimebound ] ) can be compared to the existing lower bounds on the @xmath9 mass from direct searches , as for instance quoted by cdf @xcite ; these limits are model - dependent , but in the ballpark of @xmath258gev , which is perfectly compatible with ( [ zprimebound ] ) . on the other hand , if a @xmath9 was found in direct searches at the tevatron or the lhc , the bounds on @xmath236 would constrain its couplings . this is particularly interesting in a framework with nearly family - universal couplings and illustrates the potential synergy between direct searches for np and constraints from flavour physics . note that ( [ 47 ] ) can also be translated into an upper bound on the branching ratio of @xmath259 , at least if the coupling of the @xmath9 to @xmath260 is known . the relevance of such a bound is not quite clear , however , since we have set the coupling of the @xmath9 to right - handed fermions to 0 . let us now discuss @xmath3 mixing in supersymmetry . whereas in the sm flavour violation is parametrized by the ckm matrix , in susy there are many more possible ways in which both lepton and quark flavours can change . this is because scalar quarks and leptons carry the flavour quantum numbers of their susy partners , which implies that flavour violation in the scalar sector can lead to flavour violation in the observed fermionic sector of the theory . the parameters controlling flavour violation in the mssm are quite numerous there are about 100 soft susy breaking parameters which could give rise to huge and unobserved flavour violation . one way to defuse this so - called susy flavour problem is to assume that the squark ( and slepton ) masses are approximately aligned with the quark ( and lepton ) masses . `` alignment '' means that , in the basis of physical states , where the fermion masses are diagonal , the scalar mass matrices are approximately diagonal as well . in this case , one can treat the off - diagonal terms in the sfermion mass matrices , @xmath261 as perturbations . here @xmath262 are family indices , @xmath263 refers to the `` chirality '' of the sfermions and @xmath264 refer to the fact that they are the susy partners of left- and right - handed quark fields , respectively . ] and @xmath265 is the average sfermion mass . this so - called mass insertion approximation ( mia ) has been first introduced in ref . @xcite , and was extensively applied to fcnc and cp - violating phenomena in ref . its strength is the fact that it is independent of specific model assumptions on the values of soft susy - breaking parameters , but its weakness is that there are many free parameters , so there is a certain loss of predictive power . in this paper , we do not attempt a sophisticated analysis , which will only be possible once a full nlo calculation of the corresponding short - distance functions has become available , which is in preparation , see ref . @xcite . rather , we would like to illustrate the impact of the constraints from @xmath7 on the dominant mass insertions , along the lines of , for instance , refs . @xcite . bounds on mass insertions from @xmath4 mixing have been investigated in ref . @xcite . in supersymmetric theories the effective hamiltonian @xmath266 responsible for @xmath3 mixing , ( [ eq1 ] ) , is generated by the sm box diagrams with @xmath18 exchange and box diagrams mediated by charged higgs , neutralino , photino , gluino and chargino exchange . for small values of @xmath267 , which is the ratio of vacuum expectation values of the two mssm higgs doublets , the higgs contributions are suppressed by the quark masses and can be neglected . photino and neutralino diagrams are also heavily suppressed compared to those from gluino and chargino exchange , due to the smallness of the electroweak couplings compared to @xmath268 . the gluino contribution was calculated in ref . @xcite , the chargino one in ref . @xcite . the analysis of ref . @xcite has shown that the chargino contributions are also very small , so that the @xmath5@xmath6 transition matrix element is given , to good accuracy , by @xmath269 where @xmath270 and @xmath271 indicate the sm and gluino contributions , respectively . it turns out that the largest contribution to @xmath271 comes from terms in @xmath272 , whereas chirality flipping @xmath273 and @xmath274 mass insertions are only poorly constrained from @xmath8 mixing , but dominantly enter @xmath275 decays . the bounds on @xmath276 and @xmath277 posed by the corresponding branching ratio have been investigated in ref . @xcite , a recent update can be found in ref . @xcite . as for the chirality - conserving mass insertions , the impact of the d0 bound ( [ d0-bound ] ) has been studied in refs . @xcite . here we set all but one mass insertion to 0 and restrict ourselves to bounds on @xmath278 and the impact of a future measurement of @xmath206 on these bounds . the effective @xmath19 hamiltonian in the mssm contains a total of eight operators as compared to only one in the sm . the corresponding hadronic matrix elements ( bag parameters ) have been calculated , in quenched approximation , in ref . the evolution of the wilson coefficients from @xmath279 , the scale where the susy particles are integrated out , to @xmath280 is known to next - to - leading order @xcite . the expression for @xmath7 in the mssm then depends on @xmath279 , @xmath281 , the average sfermion mass , and @xmath282 , the gluonino mass . we take @xmath283 and also @xmath284 as illustrative values . we then obtain the constraints on @xmath285 and @xmath286 shown in fig . [ fig : mia ] . @xmath287 the closed curves in the centre of the plots correspond to the allowed values of the real and imaginary part of @xmath288 after the measurement of @xmath7 ; note that the experimental value of @xmath7 is incompatible with the sm prediction at 1.5@xmath58 level when the ( hp+jl)qcd lattice data are used , eq . ( [ 36 ] ) , hence the origin is excluded in the right panel . the open lines correspond to constraints imposed by a measurement of the mixing phase @xmath206 , as explained in the caption . it is obvious that at present no value of @xmath206 is excluded and that the precise measurement of the mixing phase , expected to take place at the lhc , will considerably restrict the parameter space of susy mass insertions . if susy is found at the lhc , and the gluino and average squark masses are measured , the results from mia analyses of flavour processes will help to constrain the soft susy breaking terms and hence the yet to be understood mechanism of susy breaking . given the sheer number of these terms ( about 100 ) , it will be very difficult to resolve the richness of susy breaking from direct susy searches alone , which will have to be complemented by constraints ( or measurements ) from flavour physics which , in turn , will become more expressive , once the direct searches will have provided the relevant mass scales . the fcnc processes of @xmath38@xmath81 and @xmath5@xmath6 mixing offer interesting probes to search for signals of physics beyond the sm . although the former phenomenon is well established since many years , the latter has only just been observed at the tevatron , thereby raising in particular the question of the implications for the parameter space of np . the current situation can be summarized as follows : the experimental value of the mass difference @xmath27 and the recently measured @xmath7 agree with the sm . however , the sm predictions of these quantities suffer from large uncertainties . in particular , some lattice calculations ( ( hp+jl)qcd ) indicate a value of @xmath289 that is 1.5@xmath58 larger than the experimental cdf value , whereas the jlqcd results show no such effect . a similar pattern arises at the 1@xmath58 level in the @xmath4-meson system . in view of these uncertainties , values of @xmath290 , the strength of the np contributions to @xmath291 mixing , as large as 2.5 are still allowed by the experimental values of @xmath292 , and the new cp - violating phases @xmath293 are essentially unconstrained . complementary information is provided by cp violation . interestingly , the impressive measurement of mixing - induced cp violation in @xmath138 ( and similar modes ) at the @xmath3 factories may indicate a small but noticeable cp - violating np phase @xmath130 around @xmath219 , which would have a drastic impact on the allowed region in the @xmath294@xmath295 plane and would result in a lower bound on @xmath88 of @xmath296 . in any case , the experimentally excluded large values of @xmath130 reduce the upper bound @xmath297 significantly to @xmath298 . on the other hand , no information about @xmath161 is currently available , so that we are left with the large range of @xmath299 . the following quantities play a key rle for these studies : mixing : * the ckm parameters @xmath99 and @xmath300 , which enter the analysis of @xmath38@xmath81 mixing in a complementary manner . whereas the ut angle @xmath99 is currently a significant source of uncertainty for the sm prediction of @xmath27 ( and @xmath191 ) , @xmath98 is crucial for the detection of a np phase @xmath130 . thanks to the lhcb experiment , the situation for @xmath99 will improve dramatically in the future , where we assumed @xmath199 in our 2010 benchmark scenario . concerning @xmath98 , the error of @xmath54 has already a marginal impact . however , there is currently a @xmath178 discrepancy between the inclusive and exclusive determinations of @xmath55 , pointing towards @xmath159 and @xmath301 , respectively . consequently , it is crucial to clarify this situation and to reduce the uncertainty of @xmath55 . in our benchmark scenario , we assume that the central value of @xmath63 will be confirmed , and that its uncertainty shrinks to @xmath302 due to experimental and theoretical progress . it is an advantage of the @xmath8-meson system that the sm analysis of its mixing parameters is essentially unaffected by ckm uncertainties . * the hadronic parameters @xmath65 , which enter the sm predictions of @xmath25 . for a determination of @xmath134 with 10% uncertainty , the errors of the ( hp+jl)qcd lattice results have to be reduced by a factor of 2 . the hadronic uncertainties are smaller if one considers the ratio @xmath191 , involving the @xmath185-breaking parameter @xmath76 . presently , there is no indication of this ratio to deviate from its sm prediction , but there is still a large uncertainty . in our 2010 benchmark scenario , the error from @xmath76 would match that from @xmath99 . nevertheless , it will probably be challenging to detect np through deviations of @xmath190 from 1 . moreover , a result in agreement with 1 does not allow any conclusion about the presence or absence of np , as @xmath175 and @xmath91 may both deviate similarly from 1 , except for excluding certain np scenarios , like for instance higgs penguins enhanced by large values of @xmath267 . concerning the prospects for the search for np through @xmath5@xmath6 mixing at the lhc , it will be very challenging if essentially no cp - violating effects will be found in @xmath303 ( and similar decays ) . on the other hand , as we demonstrated in our analysis , even a small phase @xmath304 ( inspired by the @xmath4 data ) would lead to cp asymmetries at the @xmath220 level , which could be unambiguously detected after a couple of years of data taking , and would not be affected by hadronic uncertainties . conversely , the measurement of such an asymmetry would allow one to establish lower bounds on the strength of np contribution even if hadronic uncertainties still preclude a direct extraction of this contribution from @xmath7 and to dramatically reduce the allowed region in the np parameter space . in fact , the situation may be even more promising , as specific scenarios of np still allow large new phases in @xmath5@xmath6 mixing , also after the measurement of @xmath7 . we have illustrated this exciting feature by considering models with an extra @xmath9 boson and susy scenarios with an approximate alignment of quark and squark masses . in essence , the lesson to be learnt from the cdf measurement of @xmath7 is that np may actually be hiding in @xmath5@xmath6 mixing , but is still obscured by parameter uncertainties , some of which will be reduced by improved statistics at the lhc , whereas others require dedicated work of , in particular , lattice theorists . the smoking gun for the presence of np in @xmath5@xmath6 mixing will be the detection of a non - vanishing value of @xmath161 through cp violation in @xmath207 . let us finally emphasize that the current @xmath3-factory data may show in addition to @xmath159 other first indications of new sources of cp violation through measurements of @xmath305 and @xmath306 decays , which may point towards a modified electroweak penguin sector . all these examples are yet another demonstration that flavour physics is not an optional extra , but an indispensable ingredient in the pursuit of np , also and in particular in the era of the lhc . we would like to thank martin lscher for a discussion of the current status of lattice calculations of @xmath3 mixing parameters . v. abazov _ et al . _ [ d0 coll . ] , hep - ex/0603029 . m. carena _ et al . _ , hep - ph/0603106 . m. ciuchini and l. silvestrini , hep - ph/0603114 . l. velasco - sevilla , hep - ph/0603115 . m. endo and s. mishima , hep - ph/0603251 . m. blanke , a.j . buras , d. guadagnoli and c. tarantino , hep - ph/0604057 . z. ligeti , m. papucci and g. perez , hep - ph/0604112 . j. foster , k.i . okumura and l. roszkowski , hep - ph/0604121 . g. buchalla , a.j . buras and m.e . lautenbacher , _ rev . mod . * 68 * ( 1996 ) 1125 . n. cabibbo , _ phys . lett . _ * 10 * ( 1963 ) 531 . m. kobayashi and t. maskawa , _ prog . phys . _ * 49 * ( 1973 ) 652 . tevatron electroweak working group , hep - ex/0507091 . t. inami and c.s . lim , _ prog . phys . _ * 65 * ( 1981 ) 297 [ e : * 65 * ( 1981 ) 1772 ] . glashow , j. iliopoulos and l. maiani , _ phys . _ * d2 * ( 1970 ) 1285 . p. ball _ et al . _ , hep - ph/0003238 . l. wolfenstein , _ phys . rev . lett . _ * 51 * ( 1983 ) 1945 . s. eidelman _ et al . _ [ particle data group ] , _ phys . _ * b592 * ( 2004 ) 1 . r. fleischer , _ phys . _ * 370 * ( 2002 ) 537 . o. buchmller and h. flcher , hep - ph/0507253 . gambino and n. uraltsev , _ eur . j. _ * c34 * ( 2004 ) 181 . m. okamoto _ _ , _ nucl . * 140 * ( 2005 ) 461 ; + e. gulez _ et al . _ , _ phys . _ * d73 * ( 2006 ) 074502 . a. khodjamirian _ et al . _ , _ phys . rev . _ * d62 * ( 2000 ) 114002 ; + p. ball and r. zwicky , _ jhep _ * 0110 * ( 2001 ) 019 ; _ phys . _ * d71 * ( 2005 ) 014015 ; _ phys . _ * d71 * ( 2005 ) 014029 ; _ phys . lett . _ * b625 * ( 2005 ) 225 . andersen and e. gardi , _ jhep _ * 0601 * ( 2006 ) 097 . h. leutwyler , _ phys . _ * b378 * ( 1996 ) 313 . kronfeld and s.m . ryan , _ phys . * b543 * ( 2002 ) 59 . s. aoki _ et al . _ [ jlqcd coll . ] , _ phys . * 91 * ( 2003 ) 212001 . a. gray _ et al . _ [ hpqcd coll . ] , _ phys . * 95 * ( 2005 ) 212001 . m. okamoto , _ pos * lat2005 * _ ( 2005 ) 013 . f. palombi , m. papinutto , c. pena and h. wittig [ alpha coll . ] , _ pos _ * lat2005 * ( 2005 ) 214 . m. creutz , hep - lat/0603020 ; + c. bernard , m. golterman , y. shamir and s. sharpe , hep - lat/0603027 ; + c. bernard , m. golterman and y. shamir , hep - lat/0604017 . buras , m.e . lautenbacher and g. ostermaier , _ phys . _ * d50 * ( 1994 ) 3433 . e. blucher _ et al . _ , hep - ph/0512039 . g. cavoto _ et al . _ , hep - ph/0603019 . j. charles _ et al . _ [ ckmfitter group ] , _ eur . j. _ * c41 * ( 2005 ) 1 ; for the most recent updates , see http://ckmfitter.in2p3.fr/. m. bona _ et al . _ [ utfit collaboration ] , _ jhep _ * 0507 * ( 2005 ) 028 ; for the most recent updates , see http://utfit.roma1.infn.it/. a.j . buras , r. fleischer , s. recksiegel and f. schwab , _ eur . j. _ * c45 * ( 2006 ) 701 . bauer , i.z . rothstein and i.w . stewart , _ phys . lett . _ * 94 * ( 2005 ) 231802 ; hep - ph/0510241 . o. schneider , talk at the flavour in the era of the lhc " workshop , cern , november 2005 , http://cern.ch/flavlhc . a.b . carter and a.i . sanda , _ phys . _ * 45 * ( 1980 ) 952 ; _ phys _ * d23 * ( 1981 ) 1567 ; + i.i . bigi and a.i . sanda , _ nucl . _ * b193 * ( 1981 ) 85 . r. fleischer and t. mannel , _ phys . _ * b506 * ( 2001 ) 311 . r. fleischer , _ j. phys . _ * g32 * ( 2006 ) r71 . r. fleischer , _ phys . _ * b562 * ( 2003 ) 234 ; _ nucl . _ * b659 * ( 2003 ) 321 . r. fleischer and j. matias , _ phys . _ * d66 * ( 2002 ) 054009 . r. fleischer , g. isidori and j. matias , _ jhep _ * 0305 * ( 2003 ) 053 . m. bona _ et al . _ [ utfit collaboration ] , _ jhep _ * 0603 * ( 2006 ) 080 . k. agashe , m. papucci , g. perez and d. pirjol , hep - ph/0509117 . g. dambrosio , g.f . giudice , g. isidori and a. strumia , _ nucl . _ * b645 * ( 2002 ) 155 ; + a.j . buras _ et al . _ , _ phys . _ * b500 * ( 2001 ) 161 . buras , _ acta phys . polon . _ * b34 * ( 2003 ) 5615 . p. ball and r. zwicky , _ jhep _ * 0604 * ( 2006 ) 046 . y. nir and d.j . silverman , _ nucl . _ * b345 * ( 1990 ) 301 . branco , t. morozumi , p.a . parada and m.n . rebelo , _ phys . rev . _ * d48 * ( 1993 ) 1167 . i. dunietz , r. fleischer and u. nierste , _ phys . rev . _ * d63 * ( 2001 ) 114015 . dighe , i. dunietz and r. fleischer , _ eur . j. _ * c6 * ( 1999 ) 647 . m. smizanska ( atlas collaboration ) , private communication . langacker and m. plmacher , _ phys . _ * d62 * ( 2000 ) 013006 . m. cvetic , g. shiu and a.m. uranga , _ phys . _ * 87 * ( 2001 ) 201801 ; _ nucl . * b615 * ( 2001 ) 3 ; + m. cvetic , p. langacker and g. shiu , _ phys . rev . _ * d66 * ( 2002 ) 066004 . v. barger , c.w . chiang , j. jiang and p. langacker , _ phys . _ * b596 * ( 2004 ) 229 . a. abulencia _ et al . _ [ cdf coll . ] , hep - ex/0602045 ; hep - ex/0603006 . hall , v.a . kostelecky and s. raby , _ nucl . _ * b267 * ( 1986 ) 415 . f. gabbiani _ et al . _ , _ nucl . _ * b477 * ( 1996 ) 321 . m. ciuchini _ et al . _ , _ pos * hep2005 * _ ( 2006 ) 221 . p. ball , s. khalil and e. kou , _ phys . * d69 * ( 2004 ) 115011 . d. becirevic _ et al . _ , _ nucl . _ * b634 * ( 2002 ) 105 . e. gabrielli and s. khalil , _ phys . * d67 * ( 2003 ) 015008 . m. ciuchini , e. franco , a. masiero and l. silvestrini , _ phys . rev . _ * d67 * ( 2003 ) 075016 [ e : * d68 * ( 2003 ) 079901 ] . l. silvestrini , hep - ph/0510077 . d. becirevic _ et al . _ , _ jhep _ * 0204 * ( 2002 ) 025 . buras , s. jger and j. urban , _ nucl . _ * b605 * ( 2001 ) 600 .

a direct search for doubly - charged higgs bosons is one of the most important probe in the higgs triplet model , which is motivated by generation mechanisms of tiny neutrino masses . there are two major decay modes of ; i.e. , the same - sign dilepton decay and the same - sign diboson decay . for the case where the former decay mode is dominant , the lower limit on the mass of has been set at about 400 gev by atlas and cms collaborations . on the other hand , for the case where the latter decay mode is dominant , no dedicated search has been performed in the past . by taking into account characteristic signals of in the diboson decay scenario at lep and the lhc experiments , we find that the lower mass bound of 60 - 68 gev can be obtained using the same - sign dilepton search performed by atlas collaboration with 4.7 fb data at the collision energy of 7 tev . we also show that the limit can be extended up to about 85 - 90 gev , assuming the integrated luminosity of 20 fb and 7 tev for the collision energy . we give detailed explanations on the decay properties of for relatively small mass cases and also on production cross sections of at the next - to - leading order of qcd at the lhc .

1407.6547

in 2012 , a higgs boson was discovered at the cern large hadron collider ( lhc ) @xcite . its observed properties are consistent with the prediction in the standard model ( sm ) within the current experimental uncertainties @xcite . in addition , so far , no report has been delivered to us for discovery of the other new particles . therefore , it has been found that the sm is a good description for particle physics at the scale of hundred gev , not only in the gauge interactions but also in the sector of electroweak symmetry breaking . although the higgs boson has been discovered and its property has turned out to be sm - like , we know nothing about the structure of the higgs sector . in fact , the minimal higgs sector with one isospin doublet scalar field is just an assumption without any theoretical principle . thus , it is natural to consider a possibility that the higgs sector takes a non - minimal form with additional isospin multiplet scalar fields , such as extra singlet , doublet , triplet and so on . most of these non - minimal higgs sectors can explain current experimental data as well . furthermore , these extended higgs sectors are often introduced in the context of new physics models which try to explain the phenomena beyond the sm ; i.e. , neutrino masses , dark matter , and baryogenesis . therefore , it is very important to experimentally explore the possibility of extended higgs sectors . we then may be able to discriminate new physics models from the property of the higgs sector . for example , extended higgs sectors with multi - doublet scalar fields are introduced in supersymmetric extensions of the sm . they are also motivated to introduce an additional source of cp violation @xcite , and to realize the strong 1st - order phase transition @xcite , both of which are required to have successful electroweak baryogenesis @xcite . singlet scalar fields are often introduced in models with the spontaneously broken @xmath4 gauge symmetry @xcite . the higgs sector with a complex triplet scalar field appears in models that can explain neutrino masses via the seesaw mechanism @xcite . tiny neutrino masses can also be explained via the loop - induced effects of extended scalar sectors @xcite . extended scalar sectors with a discrete symmetry such as @xmath5 can provide a candidate of dark matter @xcite . we here focus on the higgs triplet model ( htm ) @xcite . its higgs sector is composed of an isospin doublet higgs field with a hypercharge as @xmath6 , where @xmath7 is the electric charge and @xmath8 is the third component of the isospin . ] @xmath9 and an isospin triplet higgs field with @xmath10 . in this model , majorana masses of neutrinos are generated via new yukawa interactions among the left - handed lepton doublets and the higgs triplet field ; @xmath11 , where @xmath12 is the vacuum expectation value ( vev ) of the triplet field , and @xmath13 is a matrix in the lagrangian for the yukawa interactions . one of the most characteristic features of the htm is the existence of doubly - charged higgs bosons @xmath0 , in addition to the other additional higgs bosons ; i.e. , singly - charged @xmath14 , cp - even @xmath15 and cp - odd @xmath16 higgs bosons . the discovery of @xmath0 at collider experiments is the direct evidence of the htm . production of these bosons at collider experiments has been studied in refs . . for the decay of @xmath0 , there are three sources ; i.e. , the yukawa interactions with left - handed lepton doublets , electroweak gauge interactions in the gauge - gauge - scalar type , and those in the gauge - scalar - scalar type . they cause the same - sign dilepton decay @xmath17 , the same - sign diboson decay @xmath18 and the cascade decay @xmath19 , respectively decay occurs via the scalar triple couplings , if there is a large mass difference between @xmath0 and @xmath20 . however , such a situation is severely constrained by electroweak precision measurements @xcite . ] . although the dominant decay mode of @xmath0 is determined by parameters in the model , the dilepton decay scenario has been considered as the most promising one for discovery @xcite , because of its cleanness for the detection at colliders . it is also quite appealing that the structure of neutrino mass matrix can be directly tested by measuring the dileptonic branching ratios of @xmath0 @xcite and @xmath20 @xcite , because the branching ratios are predominantly determined by the neutrino yukawa couplings . in this scenario , a sharp peak in the invariant mass distribution of the same - sign dilepton is the characteristic signal of @xmath0 . the experimental searches for @xmath0 in the same - sign dilepton events have been performed at lep @xcite , hera @xcite , tevatron @xcite and the lhc @xcite . assuming that the branching ratio of @xmath0 decay into @xmath21 is 100% , the strongest lower bound on the mass of doubly - charged higgs bosons has been obtained as 459 gev at the lhc @xcite . current bounds have also been set at around 400 gev in several benchmark points for the structure of the neutrino mass matrix @xcite . in this paper , we discuss the direct searches for @xmath0 in the diboson decay scenario , where @xmath0 predominantly decay into same - sign w bosons , at the past , current and future collider experiments , such as lep , the lhc with 7 - 8 tev run and 13 - 14 tev run . the same - sign diboson decay scenario is equally important to the same - sign dilepton decay scenario in the htm . collider phenomenology for this decay mode has been studied in refs . @xcite . in ref . @xcite , the lower limit on the mass of @xmath0 has been derived by using the same - sign dilepton events collected by atlas collaboration at the lhc with 7 tev and 4.7 fb@xmath3 data @xcite . up to our knowledge , this is the first analysis for the constraints on @xmath0 in the diboson scenario . the aim of the present paper is to explain details of the analysis done in ref . @xcite , and to make update on the results by including the qcd correction to the production cross sections . this paper is organized as follows . in section ii , we briefly review the htm . after we define the mass eigenstates for the higgs bosons , we derive the yukawa interaction and the gauge interaction for the triplet - like higgs bosons at the tree level . in section iii , we give expressions for the decay rates of @xmath0 in the all three decay modes . partial decay widths of @xmath0 are evaluated with particular attention to the case of relatively small masses where one or both of the w bosons are forced off - shell . we then show the phase diagram indicating the main decay mode of @xmath0 . next , we evaluate the cross section of @xmath0 productions at the lhc in the leading order ( lo ) and the next - to - leading order ( nlo ) of qcd . in section iv , we exhibit constraints on the mass of @xmath0 in the diboson decay scenario at the lep experiments and also at the lhc . at the lep i experiment , the lower limit on the mass of @xmath0 can be obtained from the total width of the z boson . we also evaluate the expected number of events for the various final states in the process of @xmath22 at the lep ii experiment . we then discuss the mass bound on @xmath0 by using the current lhc limit on the cross section for anomalous production of same - sign dileptons . section v is devoted to our conclusion . in appendix , the cross sections for @xmath0 production at the lhc with various collision energies are collected for reader s convenience . the scalar sector of the htm is composed of the isospin doublet field @xmath23 with hypercharge @xmath9 and the triplet field @xmath24 with @xmath10 . the most general form of the higgs potential under the gauge symmetry is written as @xmath25\notag\\ & + \lambda_1(\phi^\dagger\phi)^2 + \lambda_2\left[\text{tr}(\delta^\dagger\delta)\right]^2 + \lambda_3\text{tr}\left[(\delta^\dagger\delta)^2\right ] + \lambda_4(\phi^\dagger\phi)\text{tr}(\delta^\dagger\delta ) + \lambda_5\phi^\dagger\delta\delta^\dagger\phi , \label{pot_htm}\end{aligned}\ ] ] where all the parameters are taken to be real without loss of generality @xcite . the higgs fields can be parameterized as @xmath26 where the neutral components are expressed as @xmath27 the vevs of the doublet and triplet higgs fields are denoted by @xmath28 and @xmath12 , respectively . they are related to the fermi constant @xmath29 by @xmath30 . the non - zero @xmath12 deviates the electroweak rho parameter from unity at the tree level ; @xmath31 where @xmath32 , @xmath33 and @xmath34 are the w boson mass , the z boson mass and the weak mixing angle , respectively . since the experimental value of the rho parameter is close to unity ; i.e. , @xmath35 @xcite , @xmath12 has to be less than about 3.5 gev at the 95% confidence level ( cl ) . mass eigenstates in the doubly - charged states ( @xmath0 ) purely come from @xmath24 ; i.e. , @xmath36 . for the other scalar bosons , mass eigenstates are defined by introducing the following orthogonal transformations ; @xmath37 where mixing angles , @xmath38 , @xmath39 and @xmath40 are given by @xmath41 with @xmath42 in eq . ( [ mixing1 ] ) , @xmath43 and @xmath44 are the nambu - goldstone bosons which are absorbed into the longitudinal component of w and z bosons , respectively . because all the mixing angles given in eq . ( [ eq : mixing_angles ] ) are quite small due to @xmath45 , @xmath20 , @xmath16 and @xmath15 are mostly composed of the triplet higgs field . we thus call these scalars ( @xmath20 , @xmath16 , @xmath15 and @xmath0 ) as the triplet - like higgs bosons . on the other hand , by the same reason , @xmath46 can be regarded as the sm - like higgs boson , because it mainly comes from the doublet higgs field . by neglecting @xmath47 terms , the masses of these physical higgs bosons are given in a good approximation by @xmath48 thus , it can be observed that there are relationships among the masses of triplet - like higgs bosons @xcite ; i.e. , @xmath49 and @xmath50 . from these relations , three patterns of the mass spectrum first two patterns are @xmath51 in the case with @xmath52 and @xmath53 in the case with @xmath54 . in the special case with @xmath55 , all the triplet - like higgs bosons degenerate in mass . the kinetic term of the lagrangian for the higgs fields is given by @xmath56 , \end{aligned}\ ] ] where the covariant derivatives are defined as @xmath57 -ig'b_\mu\delta . \end{aligned}\ ] ] from the above lagrangian , higgs - gauge - gauge type vertices are derived . coefficients of the vertices for the triplet - like higgs bosons are given as follows : @xmath58 we note that according to eq . ( [ eq : mixing_angles ] ) , all the couplings are proportional to @xmath59 . next , we introduce the yukawa interaction terms with the triplet field . left - handed lepton doublet fields @xmath60 can couple to the triplet higgs field by , @xmath61 if we extract the vev in the neutral component of the triplet field , we find a majorana mass term for neutrinos @xcite , @xmath62 couplings of the yukawa interactions among the triplet - like higgs bosons and leptons are expressed in terms of @xmath12 and the neutrino mass matrix , @xmath63 with the use of eq . ( [ eq : mn ] ) as follows : @xmath64 where @xmath65 is the left - handed projection operator @xmath66 . from eqs . ( [ eq : gauge_coupling ] ) and ( [ eq : lepton_coupling ] ) , we see that the gauge ( yukawa ) coupling constants are enhanced ( suppressed ) as @xmath12 gets increased . these features are important to understand the decay property of the triplet - like higgs bosons which is discussed in the next section . we note that the interaction terms between quarks and triplet - like higgs bosons except @xmath0 are induced from the yukawa interaction for the doublet higgs field @xmath23 via the small mixing denoted by @xmath38 , @xmath39 and @xmath40 @xcite . in this section , we discuss the decay and production of @xmath0 . for the decay of @xmath0 , we present the decay rates for all the three decay modes . especially , we discuss the diboson decay mode in detail , focusing on the cases where one or both of the w bosons are forced off - shell . for the production of @xmath0 at the lhc , we evaluate the cross sections in the lo and the nlo in qcd . we estimate the uncertainties of theoretical calculations by taking into account the scale ambiguity and the uncertainty from parton distribution functions ( pdfs ) . the decay properties of @xmath0 strongly depend on @xmath12 and the mass spectrum of the triplet - like higgs bosons . for the case where @xmath0 are the lightest among all the triplet - like higgs bosons ; i.e. , @xmath67 , the same sign dilepton decay @xmath68 and the same sign diboson decay @xmath69 are possible . on the other hand , for the case where @xmath0 are the heaviest , @xmath70 , another cascade - type decay @xmath71 is also possible . for the diboson decay , in the case with @xmath72 , the tree level decay rate is given by @xmath73 furthermore , the branching ratio for four - fermion final - states is simply given by multiplying the decay branching ratio of the w bosons ; i.e. , @xmath74 on the other hand , in the case with @xmath75 , at least one of the w bosons is forced off - shell , and the decay rate given in eq . ( [ eq : diboson ] ) is no longer valid . thus , the branching ratio of @xmath0 into four - fermion final - states is not simply described by eq . ( [ eq : on - shell ] ) . in order to clarify how the difference in the decay rate of @xmath0 appears in the case with the off - shell w boson(s ) , we first consider the decay process of @xmath0 into the four - lepton final - states ; @xmath76 we can divide the decay modes into two cases ; one is the same - flavour ( s.f . ) dilepton mode such as @xmath77 , @xmath78 , and the other is the different - flavour ( d.f . ) dilepton mode such as @xmath79 . both the two diagrams contribute for the same - flavour dilepton cases , while only one diagram contributes for the different - flavour dilepton cases . , scaledwidth=50.0% ] in the s.f . dilepton decay , two feynman diagrams drawn in fig . [ fig : llvv ] contribute , while only one feynman diagram contributes in the d.f . dilepton decay . for each case , the partial decay width is calculated as @xmath80 where @xmath81 with @xmath82 are the four momenta of the final - state leptons in the order of the last term in eq . ( [ eq : llvv ] ) , @xmath83^{-1}$ ] , and @xmath84 denotes full phase - space integration over the four - body final - state . we neglect the mass of leptons . the difference between the two widths exists only in the interference term in eq . ( [ sfdecay ] ) . similarly , the partial decay width for the @xmath85 channel is given by @xmath86 for each lepton flavour , where a jet @xmath87 includes @xmath88 and @xmath89 quarks and those anti - particles . in addition , the partial decay width for the @xmath90 channel is given by @xmath91 . in total , the sum of the decay width through @xmath92 is given by @xmath93 . we note that in the case with @xmath94 where both the w bosons can be on - shell , @xmath95 is a good approximation by neglecting the interference term , and the branching ratios reduce to the product of branching ratios of the w bosons as in eq . ( [ eq : on - shell ] ) . into @xmath90 , @xmath96 , same - flavour and different - flavour @xmath97 modes as a function of @xmath98 . in this plot , only the @xmath2 mode is taken into account . , scaledwidth=50.0% ] in fig . [ fig : br4l ] , we plot the branching ratios for the decay of @xmath0 into various four - fermion final states , such as @xmath90 , @xmath96 and @xmath97 with s.f . or d.f.leptons , as a function of @xmath98 . notice that we neglect the dilepton decay and cascade decay channels here . it is found that the branching ratio of the s.f.@xmath97 decay mode is enhanced by 80% for @xmath99 gev , while by 10 - 20% for 100 gev @xmath100 gev . the ratio of all hadronic decay mode is also enhanced for @xmath75 by 5% , while the ratio of @xmath96 and d.f . @xmath97 decay modes is suppressed by 10% and 5% , respectively . therefore , for @xmath75 , the interference term can have sizable and constructive contribution to the decay rate , and consequently the s.f . @xmath97 decay becomes relatively important . the tree level formula for the dilepton decay rate of @xmath0 is given by @xmath101 where @xmath102 for @xmath103 ( @xmath104 ) . for the cascade decay , taking into account the off - shellness of the w boson , the tree - level formula is given by @xmath105 where the phase space functions are defined as @xmath106y\log x\notag\\ & + 6\left(1+x - y\right)y\sqrt{-\lambda(x , y ) } \big[\tan^{-1}\big(\frac{x - y-1}{\sqrt{-\lambda(x , y)}}\big)+\tan^{-1 } \big(\frac{x+y-1}{\sqrt{-\lambda(x , y)}}\big)\big]\bigg\},\\ \lambda(x , y)&=1+x^2+y^2 - 2xy-2x-2y . \end{aligned}\ ] ] we note that the decay rate given in eq . ( [ eq : cascade ] ) is valid for the case with @xmath107 . in ref . @xcite , it is shown that the mass difference larger than about 60 gev is excluded by the electroweak precision data . therefore , the on - shell decay mode of @xmath108 is disfavored . . the left ( right ) panel shows the @xmath12 ( @xmath98 ) dependence in the case with @xmath109 , 150 and 300 gev ( @xmath110 kev , 1 mev and 1 gev ) . @xmath111 is assumed , so that the cascade decay is absent . , title="fig:",width=283 ] . the left ( right ) panel shows the @xmath12 ( @xmath98 ) dependence in the case with @xmath109 , 150 and 300 gev ( @xmath110 kev , 1 mev and 1 gev ) . @xmath111 is assumed , so that the cascade decay is absent . , title="fig:",width=283 ] we then evaluate the decay of @xmath0 for several values of @xmath12 and @xmath98 by taking into account the all three decay channels ; dilepton , diboson and cascade decays . for the dilepton decay mode , we take all the elements of the neutrino mass matrix @xmath112 0.1 ev . in fig . [ width1 ] , we show the total decay width of @xmath0 as a function of @xmath12 for fixed values of @xmath113100 , 150 and 300 gev ( left panel ) , and as a function of @xmath98 for fixed values of @xmath114 1 kev , 1 mev and 1 gev ( right panel ) . the mass of @xmath14 is taken to be the same as that of @xmath0 so that the cascade decay mode is absent . as seen in the left panel , the total decay width takes its minimum at around @xmath115 mev , 1 mev and 0.2 mev in the case with @xmath109 , 150 and 300 gev , respectively . at these minima , the decay rates into the dilepton mode and the diboson mode are almost the same order . in the right panel , in the case where @xmath12 is as small as 1 kev , the decay width increases linearly with @xmath98 , because the decay width is calculated dominantly from the dilepton decay rate given in eq . ( [ eq : dilepton ] ) . on the other hand , in the case where @xmath12 is as large as 1 mev or 1 gev , the decay rate rapidly increases at around @xmath116 gev , because of the threshold of the on - shell w boson pair . we note that the decay rate of @xmath117 gev corresponds to the decay length of about 1 meter , so that @xmath0 produced at colliders would decay inside a detector . on the @xmath12-@xmath118 plane in the case of @xmath119 gev ( left panel ) and @xmath120 gev ( right panel ) . the parameter regions on the black and red curves give the 50% and 90% branching ratio for @xmath0 , where the decay mode is indicated inside the curves.,title="fig:",width=283 ] on the @xmath12-@xmath118 plane in the case of @xmath119 gev ( left panel ) and @xmath120 gev ( right panel ) . the parameter regions on the black and red curves give the 50% and 90% branching ratio for @xmath0 , where the decay mode is indicated inside the curves.,title="fig:",width=283 ] in fig . [ contour ] , we show contour plots for the decay branching ratio of @xmath0 in the @xmath12-@xmath121 plane , where @xmath122 , in the case with @xmath119 gev ( left panel ) and 300 gev ( right panel ) . in each block bordered by the black ( red ) contours , the branching ratio for the decay mode indicated inside the block is greater than 50% ( 90% ) . we can observe that by increasing @xmath12 with fixing @xmath121 smaller than about 1 gev , the main decay mode is replaced from the dilepton mode to the diboson mode at @xmath123 - 1 mev . the regions where the diboson decay mode dominates are enlarged by increasing @xmath98 from 150 gev to 300 gev , due to the cubic power dependence of the diboson decay rate on @xmath98 as expressed in eq . ( [ eq : diboson ] ) . in the case where @xmath0 are the lightest among the triplet - like higgs bosons , the regions where the cascade decay dominates disappear . we here comment on the decays of the other triplet - like higgs bosons @xcite . when @xmath12 is smaller than about 1 mev and @xmath124 is enough small , @xmath20 , @xmath16 and @xmath15 mainly decay into @xmath125 , @xmath126 and @xmath126 , respectively , similarly to the decay of @xmath68 . when @xmath12 is large ; i.e. , @xmath127 mev , @xmath20 mainly decay into @xmath128 , @xmath129 , @xmath130 and/or @xmath131 , while @xmath16 mainly decays into @xmath132 , @xmath133 , @xmath134 and/or @xmath135 . the decay of @xmath15 depends on the mixing angle @xmath38 in addition to @xmath12 and @xmath121 . as seen in eq . ( [ eq : mixing_angles ] ) , @xmath136 is proportional to @xmath137 , so that small but non - zero @xmath38 is typically provided at the same order as @xmath39 and @xmath40 , unless a large value of @xmath138 couplings is introduced . in such a case , the dominant decay mode of @xmath15 can be @xmath139 , @xmath140 , @xmath141 , @xmath133 , @xmath134 and/or @xmath135 . in the case with non - zero mass difference , a cascade decay @xmath142 can take place . in the following studies , we focus on the same - sign diboson decay scenario where @xmath143 is assumed to be almost 100% . this scenario can be realized in the case with rather large @xmath12 with @xmath144 or @xmath145 gev as discussed in this subsection . the leading production processes of @xmath0 at the lhc are @xmath146 in perturbative qcd , the total cross sections for these processes are expressed as @xmath147 where @xmath148 for eq . ( [ xs_pair ] ) and @xmath149 for eq . ( [ xs_ass ] ) . @xmath150 is the factorization scale . the partonic cross sections are given at the lo as @xmath151,\\ & \hat\sigma_{q\bar{q}'\to h^{\pm\pm}h^{\mp}}(\hat s ) = \frac{\pi\alpha^2\cos^2\beta}{36\hat ss_w^4}\left|\frac{1}{1-x_w(1+i\gamma_w / m_w ) } \right|^2 \lambda^{3/2}(x_{h^{\pm\pm}},x_{h^\pm } ) , \label{eq : cs_lhc}\end{aligned}\ ] ] where @xmath152 ( @xmath153 or @xmath20 ) , @xmath154 , @xmath155 , @xmath156 is the electric charge of @xmath157 , @xmath158 , and @xmath159 , @xmath160 . the electric charge and the third component of the isospin for a fermion @xmath161 are denoted by @xmath162 and @xmath163 , respectively . the partonic luminosity functions are defined as @xmath164 the nlo qcd corrections to the total cross sections are calculated in ref . we evaluate the lo and nlo total cross sections for the processes in eq . ( [ eq : pair ] ) and eq . ( [ eq : asso ] ) at the lhc . ( left ) and @xmath165 ( right ) processes at the lhc with @xmath166 tev evaluated at lo and nlo with cteq6l1 and cteq6 m pdfs , respectively . the @xmath167-factors are also plotted , which are defined as the cross sections evaluated at the lo and the nlo with varying the scales @xmath168 for @xmath169 divided by the lo cross section evaluated with @xmath170 . for the second process , @xmath111 is assumed . , title="fig:",scaledwidth=45.0% ] ( left ) and @xmath165 ( right ) processes at the lhc with @xmath166 tev evaluated at lo and nlo with cteq6l1 and cteq6 m pdfs , respectively . the @xmath167-factors are also plotted , which are defined as the cross sections evaluated at the lo and the nlo with varying the scales @xmath168 for @xmath169 divided by the lo cross section evaluated with @xmath170 . for the second process , @xmath111 is assumed . , title="fig:",scaledwidth=45.0% ] in fig . [ fig : lhc ] , we show the cross sections and @xmath167-factors for @xmath171 ( left ) and @xmath172 ( right ) production at the lhc with @xmath173 tev . we use cteq6l1 and cteq6 m pdfs @xcite for lo and nlo calculations , respectively , and vary the factorization scale @xmath150 and the renormalization scale @xmath174 , where the latter enters at the nlo , to see uncertainties of the cross section calculations . in the top panel , nlo ( lo ) cross sections are plotted in solid ( dashed ) lines as a function of @xmath98 . for each order , two lines are drawn which correspond to the maximum and minimal values by varying @xmath175 from @xmath176 to @xmath177 , where @xmath7 is the invariant mass of the final - state scalar pair . thus , the difference of the two lines indicates the uncertainty of the calculation by the choice of the scales . in the bottom panel , the corresponding @xmath167-factors are plotted , which are defined as the ratios of those cross sections to the lo cross section evaluated with @xmath170 . for both processes , the @xmath167-factors are about 1.2 . the scale uncertainties are typically 5% ( 10% ) level for the nlo ( lo ) calculation , while these are suppressed accidentally at around @xmath178 gev . the uncertainties from pdfs are found to be about @xmath179 for the lower mass regions but about @xmath180 for the higher mass regions . for reader s convenience , in appendix [ sec : tab ] , we present tables for the cross sections and their uncertainties for all the processes in eq . ( [ eq : pair ] ) and ( [ eq : asso ] ) for various values of @xmath98 and various collision energies at the lhc . the other @xmath0 production processes ; i.e. , the vector boson fusion @xmath181 and the weak boson associated production @xmath182 , are induced by the @xmath183 coupling which is proportional to @xmath12 as shown in eq . ( [ eq : gauge_coupling ] ) . therefore , these production cross sections are suppressed due to @xmath184 . in this section , we discuss the collider signals for the searches for @xmath0 at lep and the lhc in the diboson decay scenario . since there have been no dedicated studies in past experiments , up to our knowledge , we start to consider the experimental constraint from relatively small mass regions by using the precise measurement on the z boson width at the lep i experiment . expected signal events for @xmath0 in the @xmath22 process are studied for the lep ii energy and luminosity . after that , we study the constraint on @xmath0 in the inclusive same - sign dilepton events at the lhc . the lep experiment was operated with the electron - positron collision at the center - of - mass energy on the z boson mass ( lep i ) and up to about 209 gev ( lep ii ) . at the lep i experiment , the total decay width of the z boson has been precisely measured @xcite . the measurement can be used to constrain @xmath0 whose mass is smaller than a half of @xmath33 independently of the decay modes of @xmath0 . for @xmath98 smaller than a half of @xmath33 , the total decay width of the z boson receives a sizable correction from the partial width for the @xmath185 decay as @xmath186 using the current experimental data and the sm prediction for the z boson width @xcite , @xmath187 gev and @xmath188 gev , respectively , we obtain the lower bound @xmath189 gev at the 95% cl . for @xmath190 , a pair production process of @xmath0 , @xmath191 , is utilized to search for @xmath0 at the lep ii experiment . the total cross section for this process is given by @xmath192,\end{aligned}\ ] ] where @xmath193 and @xmath194 . the searches for @xmath0 in the dilepton decay mode have been performed at the lep experiment @xcite . we consider the searches for @xmath0 in the diboson decay scenario . through the decays of @xmath0 into the ( off - shell ) w bosons , it subsequently leads to various exotic signals , such as 8-jets , lepton plus 6-jets plus missing energy , same - sign or opposite - sign dilepton plus 4-jets plus missing energy , trilepton plus 2-jets plus missing energy , and tetralepton plus missing energy . produced numbers of events for these signals are estimated to be @xmath195 where @xmath196 , but the signals with @xmath197 s are neglected for simplicity . process in the diboson decay scenario at the lep ii experiments as a function of @xmath98 . total number of events for the @xmath22 production is also plotted . the collision energies and the integrated luminosities collected at the lep ii experiments are listed in table [ tab : lep2 ] . , scaledwidth=60.0% ] in fig . [ fig : lep2 ] , we plot the expected number of events for these signals as a function of @xmath98 at the lep ii experiment . we calculate the expected number of events by collecting the cross sections for various collision energies and integrated luminosities listed in table [ tab : lep2 ] @xcite . .collision energies @xmath198 and integrated luminosities @xmath199 at the lep ii experiments @xcite . [ cols=">,>,>,>,>,>,>,>,>",options="header " , ] the efficiencies of detector acceptance and kinematical cuts are separately estimated by using the monte - carlo simulation at the parton level . in order to generate the signal events , we use madgraph5 @xcite and cteq6l pdfs @xcite . in fig . [ fig : dist ] , we show the distributions for the signal events in the transverse momentum of a muon , the missing transverse momentum and @xmath200 for @xmath201 gev to 100 gev to check the shape of the distributions and their mass dependence . according to ref . @xcite , the kinematical cuts by detector acceptance are taken as @xmath202 where @xmath203 and @xmath204 represent the transverse momentum and the pseudorapidity of a muon , respectively . in addition , a cut on @xmath205 gev is applied . in the fourth and fifth rows in table [ tab : lhc ] , the efficiencies by the kinematical acceptance for muons and that by the kinematical cut on the dimuon invariant - mass are listed , respectively . because only the muon momenta are measured , our parton - level analysis is expected to be a good approximation to the realistic detector - level observation . finally , in the last row , we list the fiducial cross section as a function of @xmath98 , calculated by using the numbers in the upper rows . the fiducial cross section takes its maximum value at around @xmath201 gev , because that for lower mass is significantly reduced by the acceptance cut . of muons , missing transverse momentum , and the invariant mass of the same - sign dimuon for the inclusive @xmath206 process at the lhc with @xmath207 tev . distributions are evaluated by using madgraph @xcite with @xmath201 gev , 60 gev , 80 gev and 100 gev . ] in fig . [ fig2 ] , the fiducial cross section for the @xmath208 events is plotted as a function of @xmath98 by a dark - green band , where its width indicates 5% uncertainty for the total cross section at the nlo . for the comparison , the lo results previously obtained in ref . @xcite is also shown in the light green band where 10% theory uncertainty is taken into account . an orange shaded band gives the 95% cl upper limit for the fiducial cross section obtained in ref . @xcite using 4.7 fb@xmath3 data . the width of the data band comes from the uncertainty of @xmath209 for the @xmath210 system between 59% and 72% @xcite . taking a conservative examination , @xmath0 is excluded for @xmath211 to 68 gev in the diboson decay scenario , depending on the value of @xmath212 in ref . we emphasize again that this is the first verification by using the collider data on the searches for @xmath0 in the diboson scenario . we find that a stronger mass bound is obtained by using the lhc data more than the bound obtained via the @xmath213 measurement at the lep experiment . the red shaded band is drawn by extrapolating the atlas results to those for 20 fb@xmath3 by assuming that the upper limit of the cross section @xmath214 becomes small by a factor 2 . by comparing the extrapolated band with the theoretical cross section , we obtain that the regions of @xmath215 to 90 gev can be surveyed by using the existing lhc data with 20 fb@xmath3 . we note that the difference of the signal cross - sections from 7 tev to 8 tev is not taken into account for this extrapolation , since we do nt know how the background cross sections scale at the same time . we remark that the lower bound obtained in this analysis can be improved by taking into account the followings : ( i ) the other source of extra @xmath0 from the decay of @xmath14 is not considered here , for simplicity , since the decay rate of @xmath216 depends on the other parameters in the model @xcite . to count the @xmath0 production from the decay of @xmath14 , all the processes of @xmath20 production have to be also taken into account @xcite ; such as @xmath217 , @xmath218 and @xmath219 . ( ii ) although we have studied the same - sign dileptons only from the decay of @xmath0 , they can also appear in the decay of @xmath14 ; e.g. , @xmath220 @xcite . however , the decay of @xmath20 also strongly depends on the mass difference among the triplet higgs bosons , @xmath221 , so that we here neglect these contributions as a conservative assumption . ( iii ) it is studied that a requirement of relatively hard jets in addition to the same - sign dilepton in the event can enhance the significance for discovering @xmath0 @xcite . in the future lhc run with @xmath222 to 14 tev , the mass bound can be further improved by such an optimized analysis . events at the lhc with @xmath173 tev as a function of @xmath98 . dark - green and light - green bands show the estimated fiducial cross - sections at the nlo and lo , respectively . the widths of the band come from the 5% ( 10% ) uncertainty for the production cross - sections at the nlo ( lo ) . the horizontal thick ( thin ) band shows the ( expected ) 95% cl upper limit from the data with the integrated luminosity of 4.7 fb@xmath3 ( 20 fb@xmath3).,scaledwidth=60.0% ] finally , we comment on further observations after the discovery of @xmath0 , such as the determination of its properties and searches for the other triplet - like higgs bosons . if @xmath0 are discovered by the @xmath97 events , the observations of the other signals which come from the hadronic decays of the ( off - shell ) w bosons are important to indeed conclude the diboson decay of @xmath0 . the electric charge of @xmath0 shall be determined from the charges of the same - sign dilepton as discovery signals in either the dilepton decay scenario or the diboson decay scenario . in the dilepton decay scenario , the mass of @xmath0 can be easily determined by the sharp peak in the invariant - mass distribution of the same - sign dilepton . angular distributions of leptons discriminate the spin of @xmath0 . in addition , if @xmath223 decays are available , the spin of @xmath0 can be directly observed by using the spin correlation of the two reconstructed @xmath197 leptons @xcite . on the other hand , in the diboson decay scenario , the determination of the mass of @xmath0 is not straightforward , but still possible at hadron colliders by using the endpoint behavior of the transverse mass distribution constructed from the dilepton momenta and the missing transverse momentum @xcite . moreover , the method using the lepton energy distribution @xcite may be also applicable since clean signal events can be extracted in the sense of the kinematical cuts for leptons and the sm background contributions . observation of the spin of @xmath0 may be performed in the same method as the observation of the spin-0 nature of the sm - like higgs bosons at the lhc @xcite . the searches for the other triplet - like higgs bosons will be performed at the future lhc run and also at the future lepton colliders , such as the international linear collider ( ilc ) @xcite and compact linear collider ( clic ) @xcite . since the heavier triplet - like bosons would decay in the cascade - type , the searches at the lhc may be difficult and there can be an advantage for the searches at the future lepton colliders @xcite . searches for the triplet - like higgs bosons at photon colliders are also discussed in refs . @xcite . we close this section with comments on the diboson decay of @xmath0 in the other su(2)@xmath224 multiplet . in general , an su(2)@xmath225 scalar multiplet @xmath226 which contains both the doubly - charged @xmath227 and neutral @xmath228 components can have the @xmath229 vertex at the tree level , when @xmath228 acquires a non - zero vev vertex for the singlet scalar boson @xmath230 is recently discussed in ref . @xcite . ] . although the bound obtained in this paper is limited for the triplet scalar , the same searches can be applied to them by adopting the appropriate production cross section and branching ratio . we have investigated the collider phenomenology of @xmath0 in the htm , focusing on the scenario where @xmath0 mainly decay into the same - sign diboson . such a diboson decay scenario can be realized in the case with @xmath231 - 1 mev and @xmath232 gev as shown in fig . [ contour ] . we have shown that the decay branching ratio for the @xmath233 decay with a s.f . dilepton is enhanced by up to 80% for @xmath75 due to the interference effect . total production cross sections of @xmath234 and @xmath235 are calculated up to the nlo in qcd . the predicted cross sections are enhanced by about 20% from those at the lo . these arguments are found to be important to search for @xmath0 in the diboson decay scenario in relatively small mass regions . since there have been no dedicated studies for the search for @xmath0 in the diboson scenario , we have discussed the constraints at the past collider experiments . at the lep experiment , by comparing the total decay width of the z boson with the partial decay rate of @xmath185 , we have found that @xmath236 gev is excluded at the 95% cl . we also have calculated the number of events for various final states deduced from @xmath22 at the lep ii . although the signal with tetra leptons plus missing momentum can not be used to derive a constraint on @xmath98 , due to the reduction of the signal cross section by a square of the branching fraction of @xmath237 decay , the other signals which include a same - sign dilepton can be useful for the search for @xmath0 in the diboson decay scenario . we have finished our discussion with emphasizing a need of dedicated analysis for these signals by using the data from the lep ii experiment . we then have discussed the searches for @xmath0 in the diboson decay scenario at the lhc , and also have discussed the bound on the mass of @xmath0 from the current data . in order to find an evidence of @xmath0 in relatively lower mass regions , we treat the theoretical framework for the inclusive same - sign dilepton signal , which consists of total cross sections for @xmath0 production at the nlo , the decay branching ratio into the same - flavour dilepton decay with interference effects , and efficiencies for detector acceptance and the kinematical cuts . by combining them , we have evaluated the theoretical prediction for the fiducial cross section for the same - sign dimuon events . by comparing it with the upper limit reported by atlas collaboration using the 4.7 fb@xmath3 data at the 7 tev run , we find that the lower limit of @xmath98 in the diboson decay scenario is revised to 60 - 68 gev depending on the estimation of the signal efficiency in the search . we have estimated by naive extrapolation that the limit can be extended up to 85 - 90 gev , if full analysis with the available 20 fb@xmath3 data set at the 8 tev run is performed . our analysis shows that relatively light @xmath0 with @xmath238 gev are still allowed if @xmath0 dominantly decay into a ( off - shell ) dibosons . in the near future at the lhc run with 13 to 14 tev , the searches for @xmath0 in the diboson decay scenario will be performed to push the limit toward a few hundreds gev @xcite . at the future lepton colliders , such as the ilc @xcite and clic @xcite , we also have a chance to study the properties of not only @xmath0 but also the other triplet - like higgs bosons @xcite as long as the masses of them are within the reach of these colliders . we thank koji terashi for useful discussions . this work was supported in part by grant - in - aid for scientific research , nos . 22244031 , 23104006 and 24340046 , jsps , no . 25@xmath23910031 , and the national science council of r.o.c . under grant no.nsc-101-2811-m-008-014 in this appendix , we present the cross sections for @xmath0 production at the lhc . we consider the three processes , @xmath234 , @xmath172 and @xmath240 at the lhc with @xmath173 , 8 , 13 and 14 tev . we evaluate the total cross sections at the nlo in qcd @xcite with ct10 pdfs @xcite , and also their uncertainties by taking into account the scale ambiguity and the pdf uncertainty . for the latter two processes , the mass of @xmath14 is taken to be @xmath111 for simplicity . the scale ambiguity is estimated by seeking the maximum and minimum cross sections by varying the factorization and renormalization scales @xmath241 in the range @xmath242 where @xmath7 is the invariant mass of the final - state scalar pair . the pdf uncertainties are calculated according to the hessian method with 26 eigenvector set provided in ref . @xcite . in table [ tab : cs7 ] , the total cross section at the nlo with @xmath243 , its uncertainties from the scale choice and the pdf , in order , are presented for the three processes for various values of @xmath98 at the lhc with @xmath173 tev . the same results but for @xmath207 tev , 13 tev and 14 tev are also presented in table [ tab : cs8 ] , [ tab : cs13 ] and [ tab : cs14 ] , respectively . mass [ gev ] & & & + 50 & @xmath244 & @xmath245 & @xmath246 & @xmath247 & @xmath248 & @xmath249 + 60 & @xmath250 & @xmath251 & @xmath252 & @xmath253 & @xmath254 & @xmath255 + 70 & @xmath256 & @xmath257 & @xmath258 & @xmath259 & @xmath260 & @xmath261 + 80 & @xmath262 & @xmath263 & @xmath264 & @xmath265 & @xmath266 & @xmath267 + 90 & @xmath268 & @xmath269 & @xmath270 & @xmath271 & @xmath272 & @xmath273 + 100 & @xmath274 & @xmath275 & @xmath276 & @xmath277 & @xmath278 & @xmath279 + 120 & @xmath280 & @xmath281 & @xmath282 & @xmath283 & @xmath284 & @xmath285 + 140 & @xmath286 & @xmath287 & @xmath288 & @xmath289 & @xmath290 & @xmath291 + 160 & @xmath292 & @xmath293 & @xmath294 & @xmath295 & @xmath296 & @xmath297 + 180 & @xmath298 & @xmath299 & @xmath300 & @xmath301 & @xmath302 & @xmath303 + 200 & @xmath304 & @xmath305 & @xmath306 & @xmath307 & @xmath308 & @xmath309 + 250 & @xmath310 & @xmath311 & @xmath312 & @xmath313 & @xmath314 & @xmath315 + 300 & @xmath316 & @xmath317 & @xmath318 & @xmath319 & @xmath320 & @xmath321 + 350 & @xmath322 & @xmath323 & @xmath324 & @xmath325 & @xmath326 & @xmath327 + 400 & @xmath328 & @xmath329 & @xmath330 & @xmath331 & @xmath332 & @xmath333 + mass [ gev ] & & & + 50 & @xmath334 & @xmath335 & @xmath336 & @xmath337 & @xmath338 & @xmath339 + 60 & @xmath340 & @xmath341 & @xmath342 & @xmath343 & @xmath344 & @xmath345 + 70 & @xmath346 & @xmath347 & @xmath348 & @xmath349 & @xmath350 & @xmath351 + 80 & @xmath264 & @xmath352 & @xmath350 & @xmath353 & @xmath354 & @xmath355 + 90 & @xmath356 & @xmath357 & @xmath358 & @xmath359 & @xmath360 & @xmath361 + 100 & @xmath362 & @xmath363 & @xmath364 & @xmath365 & @xmath366 & @xmath367 + 120 & @xmath368 & @xmath369 & @xmath370 & @xmath371 & @xmath372 & @xmath373 + 140 & @xmath374 & @xmath375 & @xmath376 & @xmath377 & @xmath378 & @xmath379 + 160 & @xmath380 & @xmath381 & @xmath382 & @xmath383 & @xmath384 & @xmath385 + 180 & @xmath386 & @xmath387 & @xmath388 & @xmath389 & @xmath390 & @xmath391 + 200 & @xmath392 & @xmath393 & @xmath394 & @xmath395 & @xmath396 & @xmath397 + 250 & @xmath398 & @xmath399 & @xmath400 & @xmath401 & @xmath402 & @xmath403 + 300 & @xmath404 & @xmath405 & @xmath406 & @xmath407 & @xmath408 & @xmath409 + 350 & @xmath410 & @xmath411 & @xmath412 & @xmath413 & @xmath414 & @xmath415 + 400 & @xmath416 & @xmath417 & @xmath418 & @xmath419 & @xmath420 & @xmath421 + mass [ gev ] & & & + 80 & @xmath254 & @xmath422 & @xmath423 & @xmath424 & @xmath425 & @xmath426 + 90 & @xmath427 & @xmath428 & @xmath429 & @xmath430 & @xmath431 & @xmath432 + 100 & @xmath433 & @xmath434 & @xmath264 & @xmath435 & @xmath436 & @xmath437 + 120 & @xmath438 & @xmath439 & @xmath268 & @xmath440 & @xmath441 & @xmath442 + 140 & @xmath443 & @xmath444 & @xmath445 & @xmath446 & @xmath447 & @xmath448 + 160 & @xmath449 & @xmath450 & @xmath451 & @xmath452 & @xmath453 & @xmath454 + 180 & @xmath455 & @xmath456 & @xmath457 & @xmath458 & @xmath459 & @xmath460 + 200 & @xmath461 & @xmath462 & @xmath463 & @xmath464 & @xmath465 & @xmath466 + 250 & @xmath467 & @xmath468 & @xmath469 & @xmath470 & @xmath471 & @xmath472 + 300 & @xmath473 & @xmath474 & @xmath475 & @xmath476 & @xmath477 & @xmath478 + 350 & @xmath479 & @xmath480 & @xmath481 & @xmath482 & @xmath483 & @xmath484 + 400 & @xmath485 & @xmath486 & @xmath487 & @xmath488 & @xmath489 & @xmath490 + 450 & @xmath491 & @xmath492 & @xmath493 & @xmath494 & @xmath495 & @xmath496 + 500 & @xmath497 & @xmath498 & @xmath499 & @xmath500 & @xmath501 & @xmath502 + 550 & @xmath503 & @xmath504 & @xmath505 & @xmath506 & @xmath507 & @xmath508 + mass [ gev ] & & & + 80 & @xmath509 & @xmath510 & @xmath511 & @xmath512 & @xmath513 & @xmath514 + 90 & @xmath515 & @xmath516 & @xmath517 & @xmath518 & @xmath519 & @xmath520 + 100 & @xmath521 & @xmath522 & @xmath523 & @xmath524 & @xmath525 & @xmath526 + 120 & @xmath527 & @xmath528 & @xmath529 & @xmath530 & @xmath531 & @xmath532 + 140 & @xmath533 & @xmath534 & @xmath535 & @xmath536 & @xmath537 & @xmath538 + 160 & @xmath539 & @xmath540 & @xmath541 & @xmath542 & @xmath288 & @xmath543 + 180 & @xmath544 & @xmath545 & @xmath546 & @xmath547 & @xmath548 & @xmath549 + 200 & @xmath550 & @xmath551 & @xmath552 & @xmath553 & @xmath554 & @xmath555 + 250 & @xmath556 & @xmath557 & @xmath558 & @xmath559 & @xmath560 & @xmath561 + 300 & @xmath562 & @xmath563 & @xmath560 & @xmath564 & @xmath565 & @xmath566 + 350 & @xmath567 & @xmath568 & @xmath569 & @xmath570 & @xmath571 & @xmath572 + 400 & @xmath573 & @xmath574 & @xmath575 & @xmath576 & @xmath577 & @xmath578 + 450 & @xmath579 & @xmath580 & @xmath581 & @xmath582 & @xmath322 & @xmath583 + 500 & @xmath584 & @xmath585 & @xmath586 & @xmath587 & @xmath588 & @xmath589 + 550 & @xmath590 & @xmath591 & @xmath497 & @xmath592 & @xmath593 & @xmath594 + g. aad _ et al . _ [ atlas collaboration ] , phys . b * 726 * , 88 ( 2013 ) ; g. aad _ et al . _ [ atlas colaboration ] , phys . b * 726 * , 120 ( 2013 ) ; s. chatrchyan _ et al . _ [ cms collaboration ] , jhep * 1401 * , 096 ( 2014 ) ; s. chatrchyan _ et al . _ [ cms collaboration ] , arxiv:1312.5353 [ hep - ex ] . s. weinberg , phys . lett . * 37 * , 657 ( 1976 ) ; j. liu and l. wolfenstein , nucl . b * 289 * , 1 ( 1987 ) . j. schechter and j. w. f. valle , phys . d * 22 * , 2227 ( 1980 ) ; g. lazarides , q. shafi and c. wetterich , nucl . b * 181 * , 287 ( 1981 ) ; r. n. mohapatra and g. senjanovic , phys . rev . d * 23 * , 165 ( 1981 ) ; m. magg and c. wetterich , phys . b * 94 * , 61 ( 1980 ) . a. zee , phys . b * 93 * , 389 ( 1980 ) [ erratum - ibid . b * 95 * , 461 ( 1980 ) ] ; phys . lett . b * 161 * , 141 ( 1985 ) ; nucl . b * 264 * , 99 ( 1986 ) ; d. schmidt , t. schwetz and t. toma , phys . rev . d * 85 * , 073009 ( 2012 ) . m. aoki , s. kanemura and o. seto , phys . lett . * 102 * , 051805 ( 2009 ) ; phys . d * 80 * , 033007 ( 2009 ) ; m. aoki , s. kanemura and k. yagyu , phys . d * 83 * , 075016 ( 2011 ) . n. g. deshpande and e. ma , phys . d * 18 * ( 1978 ) 2574 ; r. barbieri , l. j. hall and v. s. rychkov , phys . d * 74 * ( 2006 ) 015007 . j. mcdonald , phys . d * 50 * , 3637 ( 1994 ) ; c. p. burgess , m. pospelov and t. ter veldhuis , nucl . b * 619 * , 709 ( 2001 ) . e. j. chun , k. y. lee and s. c. park , phys . b * 566 * , 142 ( 2003 ) . g. azuelos _ et al . _ , eur . j. c * 39s2 * , 13 ( 2005 ) ; g. azuelos , k. benslama and j. ferland , j. phys . g * 32 * , no . 2 , 73 ( 2006 ) . a. hektor , m. kadastik , m. muntel , m. raidal and l. rebane , nucl . b * 787 * , 198 ( 2007 ) ; j. garayoa and t. schwetz , jhep * 0803 * , 009 ( 2008 ) ; m. kadastik , m. raidal and l. rebane , phys . rev . d * 77 * , 115023 ( 2008 ) ; a. g. akeroyd , m. aoki and h. sugiyama , phys . d * 77 * , 075010 ( 2008 ) . f. del guila and j. a. aguilar - saavedra , nucl . b * 813 * , 22 ( 2009 ) ; a. g. akeroyd and c. -w . chiang , phys . d * 80 * , 113010 ( 2009 ) ; a. g. akeroyd , c. w. chiang and n. gaur , jhep * 1011 * , 005 ( 2010 ) . a. melfo , m. nemevsek , f. nesti , g. senjanovic and y. zhang , phys . d * 85 * , 055018 ( 2012 ) . m. aoki , s. kanemura and k. yagyu , phys . d * 85 * , 055007 ( 2012 ) . a. g. akeroyd , s. moretti and h. sugiyama , phys . d * 85 * , 055026 ( 2012 ) . c. -w . chiang , t. nomura and k. tsumura , phys . d * 85 * , 095023 ( 2012 ) . e. j. chun and p. sharma , jhep * 1208 * , 162 ( 2012 ) ; phys . b * 728 * , 256 ( 2014 ) . h. sugiyama , k. tsumura and h. yokoya , phys . b * 717 * , 229 ( 2012 ) . s. kanemura , k. yagyu and h. yokoya , phys . b * 726 * , 316 ( 2013 ) . f. del guila and m. chala , jhep * 1403 * , 027 ( 2014 ) . b. dutta , r. eusebi , y. gao , t. ghosh and t. kamon , arxiv:1404.0685 [ hep - ph ] . z. kang , j. li , t. li , y. liu and g. -z . ning , arxiv:1404.5207 [ hep - ph ] . j. abdallah _ et al . _ , [ delphi collaboration ] , phys . b * 552 * , 127 ( 2003 ) ; + p. achard _ et al . _ , [ l3 collaboration ] , phys . b * 576 * , 18 ( 2003 ) ; + g. abbiendi _ et al . _ , [ opal collaboration ] , phys . b * 526 * , 221 ( 2002 ) . a. aktas _ et al . _ [ h1 collaboration ] , phys . b * 638 * , 432 ( 2006 ) . d. acosta _ et al . _ [ cdf collaboration ] , phys . rev . lett . * 93 * , 221802 ( 2004 ) ; + d. acosta _ et al . _ [ cdf collaboration ] , phys . rev . * 95 * , 071801 ( 2005 ) ; + t. aaltonen _ et al . _ [ cdf collaboration ] , phys . * 101 * , 121801 ( 2008 ) . v. m. abazov _ et al . _ [ d0 collaboration ] , phys . * 93 * , 141801 ( 2004 ) ; + v. m. abazov _ et al . _ [ d0 collaboration ] , phys . * 101 * , 071803 ( 2008 ) ; + v. m. abazov _ et al . _ [ d0 collaboration ] , phys . lett . * 108 * , 021801 ( 2012 ) . achard _ et al . _ [ l3 collaboration ] , phys . b * 600 * , 22 ( 2004 ) . g. abbiendi _ et al . _ [ opal collaboration ] , phys . b * 440 * , 393 ( 1998 ) . g. abbiendi _ et al . _ [ opal collaboration ] , eur . j. c * 40 * , 287 ( 2005 ) . m. herquet , f. maltoni , o. mattelaer and t. stelzer , jhep * 1106 * , 128 ( 2011 ) . b. k. bullock , k. hagiwara and a. d. martin , nucl . b * 395 * , 499 ( 1993 ) . s. kawabata , y. shimizu , y. sumino and h. yokoya , phys . b * 710 * , 658 ( 2012 ) ; jhep * 1308 * , 129 ( 2013 ) . h. baer , t. barklow , k. fujii , y. gao , a. hoang , s. kanemura , j. list and h. e . logan _ et al . _ , arxiv:1306.6352 [ hep - ph ] . e. accomando _ et al . _ [ clic physics working group collaboration ] , hep - ph/0412251 . l. linssen , a. miyamoto , m. stanitzki and h. weerts , arxiv:1202.5940 [ physics.ins-det ] . k. yagyu , arxiv:1405.5149 [ hep - ph ] . j. cao and j. -f . shen , mod . a * 29 * , 1450041 ( 2014 ) . s. f. king , a. merle and l. panizzi , arxiv:1406.4137 [ hep - ph ] . h. -l . lai _ et al . _ , phys . d * 82 * , 074024 ( 2010 ) .

in this paper , we prove two results . first , there is a family of sequences of embedded quarters of the hyperbolic plane such that any sequence converges to a limit which is an end of the hyperbolic plane . second , there is no algorithm which would allow us to check whether two given ends are equal or not . * keywords * : hyperbolic plane , pentagrid , sequence of quarters , ends of the hyperbolic plane

1507.08495

this study takes place in hyperbolic geometry , in a specific tiling of the hyperbolic plane , the tessellation @xmath0 which i called the * pentagrid * , see @xcite . fix such a tessellation . denote by @xmath1 the length of a side of a tile of the tessellation . in this tiling , we call * quarter * , a subset of the tiling which is the intersection of two half - planes whose lines support consecutive edges of a pentagon @xmath2 of the tessellation . this pentagon is called the * head * of the quarter and the common point of the lines delimiting the half - planes is called the * vertex * of the quarter . note that the quarter is delimited by two rays issued from the vertex and supported by the above mentioned lines . these rays are also the * border * of the quarter . in this paper we are interested by sequences of quarters such that each term of the sequence is included in the next one . we shall show that the vertices of such quarters tend to a limit . to this aim , section [ convinf ] fixes the notion of neighbourhood for a point at infinity . section [ prelim ] studies simple properties of included quarters . but before , we had to establish specific projection properties of the pentagrid in section [ cornucopia ] . section [ sequences ] proves that a sequence of embedded quarters has a limit and section [ nocomput ] shows two results of undecidability concerning points at infinity . we fix @xmath3 a point of the hyperbolic plane and two orthogonal rays issued from @xmath3 : @xmath4 and @xmath5 . we may assume that , counter - clockwise turning around @xmath3 , @xmath4 comes before @xmath5 . we say that @xmath4 is * horizontal * and that @xmath5 is * vertical*. the rays @xmath4 and @xmath5 constitute the border of a * quarter * of the plane , @xmath6 . such a quarter can also be viewed as the intersection of two half - planes whose borders are perpendicular . first , let us fix notations . consider a pentagon @xmath2 . counter - clockwise and consecutively number the sides of @xmath2 by @xmath7 with . denote by @xmath8 be the line which supports the side @xmath7 and let @xmath9 , @xmath10 , @xmath11 , @xmath12 and @xmath13 be the vertices of @xmath2 , counter - clockwise labelled in this way , with @xmath9 , @xmath13 belonging to both sides 5 and 1 , sides 5 and 4 respectively . each line @xmath8 defines two half - planes @xmath14 and @xmath15 . let @xmath14 denote the half - plane which contains @xmath2 . call * lower strip * of @xmath2 the region which is defined by . in the lemmas of the paper , we shall speak of the side @xmath7 of a pentagon , having in mind a numbering as the one we already considered for @xmath2 , and we shall always remind which side is side 1 in order to avoid ambiguities . note that sides 1 and 4 are opposite and that @xmath16 is the common perpendicular of @xmath17 and @xmath18 . let denote a sequence of pentagons lying in @xmath6 such that any @xmath19 has an edge contained in @xmath4 , such that @xmath20 has @xmath3 as a vertex and that its edges meeting at @xmath3 are contained in @xmath4 and @xmath5 , see figure [ bij ] , and such that for any @xmath21 , @xmath19 and @xmath22 have a common side : it is both the side 1 of @xmath19 and the side 4 of @xmath22 . the complement of the @xmath19 s in @xmath6 can be represented as a union of quarters @xmath23 as illustrated by the left - hand side picture of figure [ bij ] . we call * cornucopia of @xmath6 * the union of the @xmath19 s . the @xmath23 s are defined as follows : @xmath24 is bordered by @xmath5 and by the line which support the side 3 of @xmath20 , @xmath25 is bordered by the line which supports the side 3 of @xmath19 and by the line which supports the side 2 of @xmath19 . remember that in a pentagon , sides 2 and 3 are perpendicular at the point where they meet . the quarters @xmath23 can be defined in another way : @xmath24 is the image of @xmath6 by the shift @xmath26 along @xmath5 of amplitude @xmath1 . note that @xmath26 transforms @xmath3 in the other vertex of the side 4 of @xmath20 which lies in @xmath5 . note that the shift @xmath27 along @xmath4 of amplitude @xmath1 transforms the side 1 of @xmath19 into the side 1 of @xmath22 . now , let @xmath28 be the image of @xmath29 by @xmath27 , putting . then , @xmath25 is the image of @xmath23 under @xmath27 . note that @xmath25 is also the image of @xmath29 by the shift @xmath30 along the side 1 of @xmath19 of amplitude @xmath1 . note that @xmath30 translates this decomposition of @xmath6 into each quarter @xmath25 . consider the recursive iteration of this decomposition in all new quarters generated in this way . we say that the regions @xmath31 belong to the first generation , so that the shift of the decomposition of @xmath32 in each of them by @xmath33 defines the second generation . in a similar way , the generation @xmath21 + 1 is obtained from the generation @xmath21 . the decomposition of each region into the cornucopia and its complement constitute the * cornucopia decomposition * of @xmath6 . presently , we wish to give a better algorithmic representation of the cornucopia decomposition of @xmath6 which will allow us to prove interesting properties . [ visilow ] let @xmath34 and @xmath35 be the pentagons obtained from @xmath2 by refection in its sides @xmath36 and @xmath37 respectively . define the side @xmath38 of @xmath34 , @xmath35 to be the side @xmath36 , @xmath37 of @xmath2 respectively . then the lower strip of @xmath34 , @xmath35 respectively , contains the lower strip of @xmath2 . remember that the lower strip @xmath39 of @xmath2 is defined as . note that @xmath34 , @xmath35 is also the shift @xmath40,@xmath41 respectively of @xmath2 along the side 1 , 4 , respectively , of @xmath2 of amplitude @xmath1 , see figure [ lower_strip ] where @xmath20 plays the role of @xmath2 . denote by @xmath42 , @xmath43 the strip of @xmath34 , @xmath35 respectively . then , , with , and . we have that . now , @xmath40 can be decomposed into the reflection @xmath44 in the bisector of side 1 followed by the reflection @xmath45 in side 2 . now , @xmath44 transforms @xmath18 into @xmath46 and @xmath45 leaves @xmath46 globally invariant , so that . we have too that @xmath44 transforms @xmath16 into @xmath47 and @xmath47 is invariant under @xmath45 . consequently , . now , a product of two reflections in axes which are perpendicular to @xmath17 shows that . similarly , , and , so that we obtain that . [ visi_shift ] consider the pentagon @xmath2 . let @xmath27 be the shift along the side @xmath38 of @xmath2 of amplitude @xmath1 , transforming @xmath18 into @xmath17 . let @xmath48 be the image of @xmath2 under @xmath27 . let @xmath49 , @xmath50 be the image of @xmath48 by reflection in its sides @xmath36 , @xmath37 respectively . let the side @xmath38 of @xmath49 , @xmath50 be the side @xmath36 , @xmath37 of @xmath48 respectively . then the lower strip of @xmath49 contains that of @xmath2 , but the strip of @xmath50 does not meet that of @xmath2 , except on the line @xmath18 of @xmath2 . proof . denote by @xmath51 , @xmath52 the shift of amplitude @xmath1 along the line 1 , 4 of @xmath48 respectively , which transforms the side 5 of @xmath48 into its side 2 , 3 respectively . denote by @xmath53 the side @xmath7 of @xmath48 . denote by @xmath54 , @xmath55 the lower strip of @xmath49 , @xmath50 respectively . repeating the proof of lemma [ visilow ] , we obtain that , with and and . repeating the same argument , we get that . now , as . this is obtained by decomposing @xmath27 into the reflection @xmath56 in the bisector of side 5 followed by a the reflection @xmath57 in the side 1 . similarly , we have that . at last , note that , so that and . from this we get that . for @xmath50 , note that @xmath58 . now , . accordingly , as , @xmath55 can not meet @xmath39 . [ projinstrip ] consider a pentagon @xmath2 with its sides , their support and its vertices labelled as above indicated . let @xmath59 be a point in @xmath39 , the lower strip of @xmath2 . let @xmath60 be the orthogonal projection of @xmath59 on @xmath16 . then @xmath60 is in side @xmath38 . if @xmath59 is in @xmath17 or in @xmath18 , then or respectively . let @xmath61 , @xmath62 be the half - plane defined by @xmath17 , @xmath18 respectively which contains @xmath2 . if , then , by construction of @xmath61 and @xmath62 we have or . assume that . then , @xmath63 cuts @xmath18 in @xmath64 . whether or , from @xmath64 we have two distinct perpendiculars to @xmath16 which is impossible . a similar argument proves that @xmath60 can not be in @xmath61 . [ projpenta ] let @xmath2 be a pentagon with the same labelling as in lemma [ projinstrip ] . let @xmath59 belong to the lower strip of @xmath2 . let @xmath60 , @xmath65 and @xmath66 be the orthogonal projection of @xmath59 on @xmath16 , @xmath47 and @xmath46 respectively . then , @xmath65 belongs to side @xmath36 , @xmath66 belongs to side @xmath37 , @xmath67 cuts @xmath16 in the open segment @xmath68ak[$ ] and @xmath69 cuts @xmath16 in the open segment @xmath68ke[$ ] . note that if @xmath59 belongs to @xmath17,@xmath18 , then @xmath60 and @xmath65,@xmath66 respectively also belong to @xmath17,@xmath18 respectively , and the conclusion for @xmath66,@xmath65 respectively still holds . proof . from lemma [ projinstrip ] , @xmath60 belongs to the side @xmath38 of @xmath2 . let @xmath70 and @xmath71 be the reflections of @xmath2 in @xmath47 and @xmath46 respectively . from lemma [ visilow ] , the lower strip of @xmath2 is both contained in the lower strip of @xmath70 and in that of @xmath71 . accordingly , @xmath65 belongs to side @xmath36 and @xmath66 belongs to side @xmath37 . as @xmath59 is not in the same side of @xmath16 as @xmath2 , @xmath67 , @xmath69 cuts side @xmath38 in @xmath72 , @xmath73 respectively : note that , as @xmath18 is the common perpendicular to @xmath16 and @xmath46 , and as @xmath74 is perpendicular to the side 5 of @xmath2 , , , and . also note that and . otherwise , if @xmath72 or @xmath73 would coincide with @xmath60 , @xmath75 or @xmath76 respectively would be a rectangle , which is impossible . now , by construction , @xmath77 is a lambert quadrangle , so that must be acute . clearly , . as @xmath74 is perpendicular to the side 5 of @xmath2 , is an acute angle , so that we must have @xmath78 $ ] : @xmath72 is inside @xmath79 . a similar argument with the lambert quadrangle @xmath80 shows us that @xmath73 is in @xmath68ke[$ ] . the case when @xmath59 is on @xmath17 or on @xmath18 is obvious . let us go back to the cornucopia decomposition of @xmath6 . [ heads ] consider the cornucopia decomposition of @xmath6 . consider a region @xmath72 of the generation @xmath21 : let @xmath81 s be the pentagons of the cornucopia of @xmath72 , and let @xmath82 s be the regions of the generation @xmath21@xmath83@xmath84 inside @xmath72 , both sequences of objects being numbered as in the cornucopia of @xmath6 . then , the head of the region @xmath24 inside @xmath72 is the image of the head of @xmath20 by the shift along the side @xmath85 of @xmath20 with an amplitude of @xmath1 , and the head of the region @xmath82 inside @xmath72 with is the image of @xmath86 under the shift along the side @xmath84 of @xmath86 with an amplitude of @xmath1 . under these shifts , the correspondence between the sides / lines of @xmath81 and those of the head of @xmath87 as well as between the sides / lines of @xmath20 and those of the head of @xmath24 is given by table [ corressides ] . -10pt # 1 # 2 # 3 # 4 # 5 # 6 110= proof . the line in table [ corressides ] associated to @xmath24 is a corollary of lemma [ visilow ] . for the regions @xmath82 with , this is a corollary of lemma [ visi_shift ] . remember that the shifts described in the statement of the lemma keep the orientation of the numbering invariant and that due to the definition of the shifts , a side @xmath7 is transformed into a side @xmath7 under a shift along the support of the former side @xmath7 for . [ ovisible ] let @xmath72 be a region in the cornucopia decomposition of @xmath6 . let @xmath64 be the head of @xmath72 and @xmath88 be the line which supports the side @xmath38 of @xmath64 . then , the half - plane defined by @xmath88 which does not contain @xmath64 contains @xmath3 . proof . this is a corollary of lemma [ heads ] . we know that the head of the region is delimited by its side @xmath38 . from lemma [ projinstrip ] , @xmath3 belongs to the lower strip of @xmath89 and of @xmath90 , the heads of the region @xmath24 and @xmath91 of generation 1 . lemma [ visi_shift ] extends this property to all the other regions @xmath82 of generation 1 . assume that the property is true for the generation @xmath21 . consider a region @xmath92 of the generation @xmath21 . let @xmath64 be its head and let @xmath93 be the half - plane defined by the support of the side 5 of @xmath64 which does not contain @xmath3 . then , the heads of the regions @xmath24 and @xmath91 of the generation @xmath21 + 1 are contained in @xmath93 , so that lemma [ heads ] applied to @xmath64 says that @xmath3 is also in the lower strip of the heads of the regions @xmath94 and @xmath95 of the generation @xmath21 + 1 : consequently , the property also holds for these two regions . the shift along the side 5 of @xmath64 of amplitude @xmath1 which transforms the side 4 of @xmath64 into its side 1 satisfies the hypothesis of lemma [ visi_shift ] . by induction , the lemma allows us to extend the property from the region @xmath96 with to the regions @xmath97 . accordingly , the property is true for all regions of the generation @xmath21 + 1 . this completes the proof of the corollary . [ oposit ] consider a region @xmath72 of the cornucopia decomposition of @xmath6 . then @xmath3 is in the lower strip of the head of @xmath72 . for another pentagon @xmath48 of the cornucopia of @xmath72 , @xmath3 is in the lower strip of the pentagon which is the image of @xmath48 under the shift along its side @xmath84 , going from the border of @xmath72 to the side @xmath36 of @xmath48 . this is also a consequence of the proof given for corollary [ ovisible ] . we arrive to the key property of this section . [ cornudist ] in @xmath6 , the distance from @xmath3 to a region of the generation @xmath21 is at least @xmath98 . we need a preliminary result : [ projsofo ] for each region @xmath72 in the cornucopia decomposition of @xmath6 , the orthogonal projection of @xmath3 on the border of @xmath72 occurs on the side @xmath38 of its head , ends of the side excepted when @xmath5 is not a border of the region . proof . this is a corollary of corollary [ oposit ] and of lemma [ projpenta ] and of the fact that the projection of @xmath99 on a region @xmath72 is the same as its projection on the head @xmath93 of @xmath72 : the projection is also the projection of @xmath3 on the line @xmath88 supporting the side 5 of @xmath93 . accordingly , all points in the half - plane defined by @xmath88 which does not contain @xmath3 are further from @xmath3 than its projection on @xmath88 . proof of lemma [ cornudist ] . note that the result is true for generation 1 . the cornucopia of @xmath6 has a complex border : it is @xmath4 which contains the side 5 of all pentagons contained in the cornucopia . another infinite part of the border consists of the sides 5 of the heads of the regions of generation 1 . as the pentagons @xmath81 with are outside the half - plane defined by the side 1 of @xmath20 which does not contain @xmath3 , the distance of each @xmath81 to @xmath3 is at least @xmath1 . in particular , this is the case for @xmath100 where @xmath101 is the orthogonal projection of @xmath3 on the border of @xmath23 , . now , as @xmath82 is contained in the half - plane defined by the side 5 of its head containing its head , the distance from @xmath3 to @xmath23 is at least @xmath100 , so that it is at least @xmath1 . assume that the result is true for the generation @xmath21 . consider a region @xmath72 of the generation @xmath21 and consider a region @xmath102 of the generation @xmath21 + 1 contained in @xmath102 . the head of @xmath102 is obtained from a pentagon @xmath81 of the cornucopia of @xmath72 . now , from lemma [ projsofo ] , the orthogonal projection @xmath103 of @xmath3 on @xmath102 occurs on the head of @xmath102 . let @xmath60 be the orthogonal projection of @xmath3 on@xmath72 . unless @xmath102 is the region 0 of @xmath72 , the head of @xmath102 is obtained from the head of @xmath72 by a shift along the side 1 of @xmath72 . from lemma [ projpenta ] , we have that @xmath103 is in the side 2 of the head of @xmath60 : informally , @xmath104 is to the left of @xmath105 . let @xmath104 cuts the side 5 of the head of @xmath72 at @xmath106 . as the quadrangle @xmath107 is a lambert quadrangle , remember that @xmath108 is the side 1 of the head of @xmath72 , the angle @xmath109 is acute , so that . on another hand , as @xmath106 is on the side 5 of the head of @xmath72 and as . accordingly , . if @xmath102 is the region 0 , then @xmath103 is on the side 3 of the head of @xmath72 . now , we consider the quadrangle @xmath110 which is also a lambert quadrangle , so that the same estimates can be performed , leading us to the same conclusion . and so , the property is true for the regions of the generation @xmath21 + 1 . lemma [ cornudist ] has a very important corollary which we establish now , although it is not tightly connected to the topic of this paper . the left - hand side picture of figure [ tree_bij ] illustrates the bijection between the restriction of the pentagrid to @xmath6 with a tree we called the fibonacci tree , see @xcite . the name of the tree comes from the fact that the number of nodes of the tree which are at the same distance @xmath111 from its root in term of crossed tiles is @xmath112 where is the fibonacci sequence with . in @xcite , we remember the proof of the property already mentioned in @xcite that the restriction of the pentagrid to @xmath6 is in bijection with a tree which we called the fibonacci tree : the tree can be constructed by the infinite iteration of two rules we can formulate as @xmath113 and @xmath114 , @xmath10 denoting the nodes which have two sons and @xmath115 denoting those which have three of them , the root of the tree being a @xmath115-node . we can state the following result : [ thmbij ] ( see @xcite ) the fibonacci tree is in bijection with the restriction of the pentagrid to @xmath6 . proof . the proof of the injection is easy : it is enough to note that the sons of a node @xmath116 are obtained by the reflection of the tile @xmath64 associated to @xmath116 in two or three different sides of @xmath64 . for the surjection , we have to prove that any point of @xmath6 belongs to a tile of the pentagrid restricted to @xmath6 . using the cornucopia decomposition , it is rather easy . let @xmath59 be a point of @xmath6 . if @xmath59 belongs to the cornucopia of @xmath6 , it belongs to some @xmath81 and we are done . if this is not the case , it belongs to some @xmath102 of generation 1 . in @xmath102 we repeat the same argument : either @xmath59 belongs to the cornucopia of @xmath102 and we find a pentagon of the tiling containing @xmath59 , or we find that @xmath59 belongs to some region @xmath117 of generation 2 . as from lemma [ cornudist ] , the distance from @xmath3 to a region of the generation @xmath21 is at least @xmath98 , we can find an @xmath118 such that @xmath119 , so that necessarily , @xmath59 belongs to the cornucopia of a region @xmath120 of the generation @xmath121 with @xmath122 . eventually , @xmath59 belongs to some pentagon of the tiling . it is not difficult to see that the pentagons of the cornucopia decomposition are those of the fibonacci tree : the cornucopia of @xmath6 corresponds to the leftmost branch of the fibonacci tree . its regions @xmath24 and @xmath91 have the @xmath115-sons of the root for their respective heads . now , in each region , the cornucopia is the leftmost branch of the sub - tree rooted at the node corresponding to the head . note that the heads of the regions are the white nodes of the tree , the root being the head of @xmath6 . this is illustrated by the right - hand side picture of figure [ tree_bij ] . in the following sections , we shall have to deal with sequences of points which are converging to infinity . convergence in the hyperbolic plane is easy and we can rely on poincar s disc model as far as topology only is concerned . to study points at infinity , we have to resist to the use of poincar s disc model : it fairly represents what hilbert called ends in the hyperbolic plane , but there is always the danger that the euclidean intuition plays some bad trick on us . in order to define convergence to infinity , we have to justify that the notion of convergence to a point of the border in poincar s disc model turns out to be valid . consider the same fixed point @xmath3 of the hyperbolic plane and the same quarter @xmath6 whose vertex is @xmath3 which were defined in section [ cornucopia ] and consider a sequence @xmath123 of points of the hyperbolic plane such that for all @xmath21 and that @xmath124 tends to infinity as @xmath21 tends to infinity . say that a * neighbourhood of infinity * for @xmath123 is a half - plane @xmath93 defined by a line @xmath88 such that contains finitely many points of @xmath123 only . consider a point at infinity @xmath125 and a line @xmath88 which does not pass through @xmath125 . the line defines two half - planes : @xmath61 and @xmath126 . in one of them , say @xmath126 , any line contained in the half - plane does not pass through @xmath125 . in the other , there are such lines : for any point @xmath59 of @xmath88 there is a unique line which passes through @xmath59 and through @xmath125 . in our sequel we say that @xmath61 is the half - plane defined by @xmath88 which * touches * @xmath125 and that @xmath126 is the one which does * not touch * @xmath125 . [ faraway ] let @xmath88 be a line of the hyperbolic plane and let @xmath93 be the half - plane delimited by @xmath88 which does not contain @xmath3 . let @xmath60 be the orthogonal projection of @xmath3 on @xmath88 . let @xmath127 and @xmath128 be the ray issued from @xmath3 which are parallel to @xmath88 . then tends to zero as @xmath105 tends to infinity and conversely . proof . by construction , as , is the angle of parallelism of @xmath105 for @xmath88 . the conclusion of the lemma is a well known property already established by lobachevsky . [ infneigh ] let @xmath125 be a point at infinity . let @xmath129 and @xmath130 be two rays issued from @xmath3 such that @xmath131 is the bisector of the angle and . then , there is a unique line @xmath132 of the hyperbolic plane such that @xmath132 is parallel to both @xmath129 and @xmath130 . let @xmath133 be the half - plane defined by @xmath132 which does not contain @xmath3 . then the @xmath133 s constitute a basis of neighbourhoods for @xmath125 . proof . the existence of @xmath132 is a well known property : it comes from the fact that @xmath132 is the unique line of the hyperbolic plane which is parallel to @xmath127 and which is perpendicular to @xmath131 . in order to prove that the @xmath133 s constitute a basis of neighbourhoods for @xmath125 , we first note that a neighbourhood of @xmath125 is a subset of @xmath134 which contains a half - plane @xmath93 which touches @xmath125 . of course , we may assume that @xmath93 does not contain @xmath3 . now , let @xmath88 be the border of @xmath93 . consider the ray @xmath131 : it cuts @xmath88 at @xmath60 , otherwise , @xmath93 can not touch @xmath125 . if it is perpendicular to @xmath88 , there is a point @xmath106 on @xmath131 with @xmath135 $ ] such that the parallel @xmath132 issued from @xmath106 to @xmath129 is perpendicular to @xmath131 . then , as @xmath131 is the bisector of , @xmath132 is also parallel to @xmath130 . if @xmath131 is not perpendicular to @xmath88 , then there is a point @xmath106 on @xmath131 with @xmath135 $ ] such that the perpendicular @xmath136 to @xmath131 passing through @xmath106 is non - secant with @xmath88 . we repeat with @xmath136 the just above argument . we remain to prove that if @xmath44 is another point at infinity , so that , there is a @xmath133 so that @xmath133 does not touch @xmath44 : we may even construct @xmath133 so that its border @xmath132 does not pass through @xmath44 . indeed , we take @xmath21 so that and we repeat the above construction . it is then plain that @xmath133 is contained in the half - plane delimited by @xmath137 which contains @xmath125 . by the construction , this latter half - plane does not touch @xmath44 as @xmath137 does not pass through @xmath44 and as we may assume that @xmath125 and @xmath44 are not on the same side of @xmath137 . say that a line of the hyperbolic plane is a * line of the pentagrid * if it supports at least an edge of a pentagon of the tessellation . we wish to prove that in lemma [ infneigh ] we can replace the lines @xmath132 by lines of the pentagrid . to this aim we prove the following result . [ linepenta ] let @xmath125 be a point at infinity of the hyperbolic plane and let @xmath88 be a line which does not pass through @xmath125 . then there is a line of the pentagrid @xmath138 such that @xmath138 is completely contained in the half - plane defined by @xmath88 which touches @xmath125 . proof let @xmath139 be the diameter of the regular rectangular pentagon . it is plain that : take any picture in figure [ bij ] to check the latter inequality as @xmath139 is the distance from a vertex of the pentagon to the midpoint of the opposite side . this means that for any point @xmath2 of the hyperbolic plane , within a disc of radius @xmath139 centered at @xmath2 we can find a vertex of the pentagrid . consider @xmath88 , a line of the hyperbolic plane which does not pass through @xmath125 . denote by @xmath61 the half - plane defined by @xmath88 which touches @xmath125 and by @xmath126 the other half - plane : that which does not touch @xmath125 . take @xmath9 a point on @xmath88 and let @xmath73 be a vertex of the pentagrid such that , which is in @xmath61 and which is the closest to @xmath73 . let @xmath140 and @xmath141 be the rays issued from @xmath73 which are supported by the lines of the pentagrid which meet at @xmath73 and which delimit a quarter whose head @xmath2 cuts @xmath88 . if both @xmath140 and @xmath141 do not meet @xmath88 , we are done . if we require @xmath140 and @xmath141 to be non - secant with @xmath88 , we take @xmath72 on the continuation of @xmath140 in @xmath61 , at the distance @xmath1 and then we take @xmath64 on the next side of the pentagon @xmath48 which contains @xmath72 and @xmath73 and which has a common side of @xmath2 . then the rays issued from @xmath64 and supporting the edges of @xmath48 abutting @xmath64 are non - secant with @xmath140 and @xmath141 as having a common perpendicular with these rays . at least one of the half - planes delimited by @xmath140 and @xmath141 touches @xmath125 . we take the line corresponding to this half - plane . if @xmath140 and @xmath141 are not in this case , at least one of them , say @xmath140 cuts @xmath88 . continue the ray @xmath140 by the other ray @xmath142 on the same line until we meet a vertex @xmath72 of the pentagrid for which the other ray @xmath143 abutting @xmath72 and which is on the same side of @xmath140 as @xmath2 , is non - secant with @xmath88 . indeed , let @xmath60 be the orthogonal projection of @xmath72 on @xmath88 . as @xmath72 tends to infinity on @xmath142 , @xmath144 also tends to infinity and the angle of @xmath142 with @xmath144 tends to zero so that we can find such an @xmath72 that the angle of parallelism for @xmath144 with @xmath88 is less than @xmath145 . then the angle @xmath146 of @xmath142 with @xmath144 satisfies as @xmath140 cuts @xmath88 . accordingly @xmath143 makes an angle which is bigger than @xmath145 so that @xmath143 and its continuation in a line is non - secant with @xmath88 and it clearly lies in @xmath61 . let @xmath147 be the continuation of @xmath143 after @xmath72 . then , we can find on @xmath147 a vertex @xmath64 of the pentagrid so that the perpendicular @xmath148 to @xmath147 passing through @xmath64 is non - secant with @xmath88 . then at least one of the half - planes delimited by @xmath147 and @xmath148 and which does not contain @xmath88 touches @xmath125 . we take the line defined by this half - plane . [ neighgrid ] the lines of the pentagrid define neighbourhoods for the points at infinity . proof . it is a direct consequence of lemmas [ infneigh ] and [ linepenta ] . figure [ decomp ] indicates two ways to decompose a quarter into other quarters . consider two quarters @xmath149 and @xmath150 whose vertices are @xmath151 and @xmath49 respectively and whose heads are @xmath61 and @xmath126 respectively . we say that @xmath149 is * embedded * , * strictly embedded * in @xmath150 , denoted by , respectively , if , respectively , where @xmath152 is the interior of @xmath150 . from the definition , strictly embedded quarters are embedded but embedded quarters may be not strictly embedded . denote by @xmath153 the border of the quarter @xmath65 . in the left - hand side of figure [ decomp ] , we can see that the orange quarter is embedded in the quarter @xmath48 whose head is the red tile . we also can see on the same picture that the blue quarter is strictly embedded in @xmath48 . on the right - hand side of figure [ decomp ] , the blue quarter and the quarter which extends the light orange zone are both embedded in @xmath48 , but not strictly . in the situation when the head of @xmath149 shares an edge with the head of @xmath150 , there are three possible cases . in two of them , @xmath149 is embedded in @xmath150 but not strictly , while in the third case , @xmath149 is strictly embedded in @xmath150 . we shall denote these cases by when the embedding is not strict and when the embedding is strict . the index 0 reminds us that the heads share an edge . in both cases we speak of a * one step * embedding . note that when , it is not possible to find a quarter @xmath66 such that and . now , we can prove the property indicated in lemma [ dist_embedded ] . [ dist_embedded ] let @xmath149 and @xmath150 be two embedded quarters whose vertices are @xmath151 and @xmath49 respectively . there is a finite sequence of quarters such that , and or for . moreover , the distance from @xmath151 to @xmath49 is @xmath121@xmath1541 in number of tiles . 1=0 proof . identifying the head of @xmath150 as the tile in bijection with the root of the fibonacci tree , see the left - hand side picture of figure [ tree_bij ] , it is easy to find a finite sequence of tiles @xmath155 , with , with @xmath90 being the head of @xmath150 and @xmath156 that of @xmath149 . each tile is in correspondence with the nodes of the tree which are on the branch which leads from the root to the node in bijection with the head of @xmath149 . by construction , @xmath90 is the head of @xmath150 and by construction . for each @xmath155 with @xmath157 , we look at the place of @xmath155 with respect to @xmath158 which is the head of @xmath159 . there are three possible cases only as indicated by figure [ decomp ] . if the edge shared with @xmath158 has a vertex on the border of @xmath160 , then we define @xmath159 as indicated by figure [ decomp ] : there is a single possibility which yields . if the edge shared with @xmath158 has no vertex on the border of @xmath160 , there is again a single possibility given by the left - hand side decomposition and we have . the distance in number of tiles from @xmath151 to @xmath49 is the number of tiles on the branch , the last tile being excepted , so it is @xmath121@xmath1541 . [ dist_vertex ] let @xmath149 and @xmath150 be two quarters whose vertices are @xmath151 and @xmath49 respectively . then . clearly , the distance is bigger if the embedding is not in one step . for a one step embedding , figure [ decomp ] clearly proves corollary [ dist_vertex ] . from what we have seen in section [ prelim ] , when we are dealing with a sequence of quarters such that , we may assume that each embedding of consecutive terms of the sequence is a one step embedding . say that such a sequence is * stepwise*. now , consider a stepwise sequence @xmath161 of embedded quarters . let @xmath162 be the vertex of @xmath163 . the sequence can not converge in the hyperbolic plane as the distance between two consecutive terms is at least @xmath1 . note that the topologies induced in poincar s disc by the euclidean metric and by the hyperbolic one coincide despite the fact that the metrics are very different . this is a well known feature , coming from the property that hyperbolic circles are euclidean circles contained in the open disc . now , the closure of the disc is compact , so that the sequence has at least one limit point @xmath125 which is a point of the border of the poincar s disc , which corresponds to an end of the hyperbolic plane . consider three consecutive terms of the sequence : @xmath163 , @xmath164 and @xmath165 . consider the one - step relations between consecutive terms . if we have both and , we have two cases : we have either or , but in that latter case , we also have , see the first two pictures of figure [ onesteplarge ] . the figure shows us that starting from @xmath164 , there are two possibilities to construct @xmath165 and only them : those which are illustrated by the pictures of the figure . indeed , we have only two possibilities for choosing the new head . once the new head is chosen , we have _ a priori _ two possibilities for choosing the vertex in order to obtain a quarter which contains @xmath164 . but one of them defines @xmath165 as strictly embedding @xmath164 . so that a single vertex remains to obtain @xmath165 as embedding @xmath164 but not strictly , see the first two pictures of figure [ onesteplarge ] . and so , we remain with the two cases which are illustrated by the first two pictures of figure [ onesteplarge ] . now , the rightmost picture of figure [ onesteplarge ] show us the following property : [ no_altern ] consider three quarters , @xmath149 , @xmath150 and @xmath166 such that and . then , there is a quarter @xmath167 such that : and . conversely , if we have , and , we may find @xmath167 such that and . [ large_and_strict ] let be a stepwise sequence of consecutively embedded quarters . we may assume that if , and , then we have . consider a stepwise sequence of embedded quarters . say that @xmath164 presents an * alternation * if and only if , and . from corollary [ large_and_strict ] , we may assume that a stepwise sequence does not contain any alternation . this necessarily means that if , then . we say that a stepwise sequence of embedded quarters with no - alternation is * ultimately direct * if there is an integer @xmath168 such that for all positive @xmath121 we have . if in an ultimately direct sequence we may have , we say that the sequence is * direct*. we can state : [ non_alternate ] let be a stepwise sequence of embedded quarters with no alternation and assume the sequence to be ultimately direct . let @xmath162 be the vertex of @xmath163 . then @xmath162 tends to the point at infinity @xmath125 which is on a line @xmath88 which supports one border of all @xmath163 s starting from a certain rank . moreover , all @xmath163 s are contained in the same half - plane defined by @xmath88 . proof . from the assumption , we have an integer @xmath168 such that for all @xmath169 and such that for any positive @xmath121 . and so , there is a line @xmath88 issued from @xmath170 such that @xmath88 contains a part of the border of @xmath171 and such that for all positive @xmath121 , there is a ray issued from @xmath172 which is in the border of @xmath173 and which is contained in @xmath88 . clearly , @xmath172 converges to a point at infinity which is on @xmath88 as @xmath121 tends to infinity . also clearly , as there is no alternation , all @xmath173 s are on the same side of @xmath88 . due to the consecutive embedding of all terms of the sequence , all @xmath163 s are also in the same side . now , what can be said for stepwise sequences of embedded quarters with no alternation which are not ultimately direct ? [ close ] consider a stepwise sequence @xmath161 of embedded quarters and assume it to be with no alternation and assume that the sequence is not ultimately direct . let @xmath162 be the vertex of @xmath163 . assume that the sequence converges to an end @xmath125 . let @xmath88 be a line which does not pass through @xmath125 . then , there is an @xmath168 such that for all @xmath21 , , @xmath163 contains the half - plane delimited by @xmath88 which does not touch @xmath125 . assuming lemma [ close ] , we can prove : [ to_end ] consider a sequence @xmath161 . let @xmath162 be the vertex of @xmath163 . then there is an end @xmath125 such that @xmath162 converges to @xmath125 when @xmath21 tends to infinity . proof of theorem [ to_end ] . from what we have already noticed , the sequence has at least one limit point , and any limit point is an end . assume that the sequence has at least two distinct limit points @xmath174 and @xmath175 . then we can find lines @xmath17 and @xmath47 such that if @xmath176 and @xmath177 respectively are the half - planes defined by @xmath17 and @xmath47 and which touches @xmath174 and @xmath175 respectively , then . indeed , consider two lines @xmath178 and @xmath179 which pass by @xmath174 and @xmath175 respectively . as , the lines are distinct . we may assume that they meet at some point @xmath9 of the hyperbolic plane . if not , the lines are non secant . then replace @xmath178 and @xmath179 by the lines which are parallel to @xmath178 and @xmath179 and which are issued from the mid - point of the segment of the common perpendicular to @xmath178 and @xmath179 which joins @xmath178 to @xmath179 . from @xmath9 , consider the bisector of the angle . it defines a point at infinity @xmath44 . then take the bisector of and of . these new bisectors define two new points at infinity @xmath180 and @xmath181 . now , define @xmath17 , @xmath47 as the perpendicular to @xmath178 , @xmath179 respectively , issued from @xmath180 , @xmath181 respectively . consider two sub - sequences of the @xmath163 s , and such the vertices of the @xmath182 converge to @xmath174 and those of the @xmath183 converge to @xmath175 . we have and where and are distinct sub - sequences of @xmath184 . assume that and are both ultimately direct . there is an integer @xmath168 such that for all positive @xmath121 we have both and together with both and . there is a line @xmath185 and a line @xmath186 such that @xmath185 passes through @xmath174 and @xmath186 passes through @xmath175 and , from lemma [ non_alternate ] , all @xmath182 s are on the same side of @xmath185 and all @xmath183 s are on the same side of @xmath186 . these sides define half - planes @xmath187 , @xmath188 delimited by @xmath185 , @xmath186 respectively . assume that contains when we get close to @xmath175 . consider some @xmath118 with . we can find such that we have for instance that @xmath189 contains @xmath190 : but then , @xmath189 contains points of the hyperbolic plane which are not in , a contradiction . so that we now assume that @xmath188 and @xmath187 do not meet when we get close to @xmath175 . but the same inclusion as above immediately shows that @xmath189 contains points which are not in @xmath187 . accordingly , in that case , . consider now that one is ultimately direct , and that is not . from lemma [ close ] there is @xmath168 such that when , @xmath191 contains . now , assume that . take a @xmath192 such that . then , @xmath191 contains as well as a few points of @xmath177 . as , @xmath182 contains also points in @xmath193 , a contradiction as . now , assume that . again , take @xmath192 such that . as contains @xmath191 , it also contains points which are on the complement of @xmath187 , again a contradiction . and so , in that case too , . now , we remain with the case when both sequences and are not ultimately direct . from lemma [ close ] there is @xmath168 such that when , @xmath133 contains both and . but , so that @xmath194 contains @xmath134 for a certain @xmath118 , which is impossible . and so , we again conclude that . this proves that there is a unique limit point , hence the convergence of the sequence . we can now turn to the proof of lemma [ close ] . we already know that the sequence is stepwise , that it has no alternation , that it is not ultimately direct and that it has at least one limit point , say @xmath125 . from lemma [ linepenta ] , we may replace @xmath88 by a line @xmath138 of the pentagrid which does not pass through @xmath125 . let @xmath61 be the half - plane delimited by @xmath138 which touches @xmath125 and @xmath126 be its complement in @xmath134 . it is also plain that if we find a quarter @xmath163 satisfying the conclusion of the lemma , this will also be the case for all @xmath194 s with . there is a first @xmath21 such the vertex @xmath162 of the quarter @xmath163 is in the interior of @xmath61 . accordingly , the head @xmath2 of @xmath163 has a side on @xmath138 and @xmath162 is either one of its two vertices at the distance @xmath1 from @xmath138 or the single one at the distance @xmath139 . in the latter case we are done : the rays @xmath140 and @xmath141 issued from @xmath162 have both a common perpendicular with @xmath138 so that @xmath163 contains @xmath126 . now , assume that @xmath162 is at the distance @xmath1 from @xmath138 . let @xmath118 be the first integer not smaller than @xmath21 such that . as there is no alternation , @xmath195 is on the same line @xmath17 passing through @xmath195 and @xmath162 which is perpendicular to @xmath138 . as the sequence is not ultimately direct , there is such an @xmath118 . from the non - alternation assumption , the head @xmath196 of @xmath197 has one ray @xmath140 of its border which is perpendicular to @xmath17 and the other ray @xmath141 is perpendicular to @xmath140 and it lies in the same side of @xmath17 as @xmath198 . then @xmath199 is the vertex which is opposite to the side @xmath200 of @xmath196 shared with @xmath198 . now , @xmath17 is a common perpendicular to @xmath140 and to @xmath138 , so that @xmath140 lies in @xmath61 . now , @xmath141 has a common perpendicular with the line @xmath136 which supports @xmath200 . now , @xmath136 itself is perpendicular to @xmath17 , so that it is contained in @xmath61 . accordingly , @xmath141 is also contained in @xmath61 , as it is on the other side of @xmath136 with respect to @xmath138 . this proves that @xmath197 contains @xmath126 . theorem [ to_end ] makes use of the compacity theorems which are not algorithmically true . we shall use the tools used in the proof of theorem [ to_end ] to prove that it is algorithmically impossible to say whether two given ends are equal or not . the theorem says that in a sequence of embedded quarters , their vertices converge to a limit which is a point at infinity of the hyperbolic plane . we can easily be convinced that a quarter can be clearly identified by three vertices of its head @xmath2 : the vertex @xmath73 of the quarter and the two vertices @xmath9 and @xmath10 of @xmath2 which are joined to @xmath73 by an edge of @xmath2 . call * hat of the quarter * the triple @xmath201 or @xmath202 . the rays defining the quarter are defined by @xmath203 and @xmath204 with @xmath73 being the point from which the ray is issued and the second point being a point on the ray . as each vertex can be identified by a coordinate , see for instance @xcite , the hat of a quarter is a piece of information which can easily be encoded for an algorithm . in an algorithmic approach , a sequence of embedded quarters is an algorithm , which , in principle , can also finitely be encoded . the embedding condition can also be encoded , much more easily if we assume the sequence to be stepwise with no - alternation . however , the fact that there is no alternation can not algorithmically be checked and the stepwise condition also can not algorithmically be checked : intuitively , this would require an infinite time . the algorithmic translation of theorem [ to_end ] translates the sentence _ to each sequence of embedded quarters , we can define an end to which the sequence of their vertices converge_. this notion of convergence means that it is possible to assign to each line @xmath138 of the pentagrid a rank @xmath168 which ensures that the quarters with a higher rank are beyond @xmath138 and that the sequence of the @xmath138 s define an end . we may assume that the lines of the pentagrid can also be encoded , for example , by a pair of vertices of the pentagrid . the convergence of this sequence of lines to an end can not be checked but it nonetheless can be defined . indeed , from lemma [ linepenta ] , if a line @xmath88 defines a half - plane containing an end @xmath125 , there is a line of the pentagrid @xmath138 which defines a half - plane also containing @xmath125 . this allows us to consider the ends which can be defined by a sequence of lines of the pentagrid . and so , to each sequence of quarters , we associate a sequence of lines of the pentagrid which defines the end and this translation from a sequence of quarters to a sequence of lines of the pentagrid must be algorithmic . assume also that two lines of the pentagrid being given , it is possible to decide whether they define non - intersecting half - planes or not . now we show that it is not possible to algorithmically distinguish given ends . [ noalgo ] there is no algorithm which would for any sequence of lines of the pentagrid defining ends whether these ends are equal or not . proof . the proof consists in constructing a sequence of sequences of quarters for which there is no algorithm defining an end . we define the sequence of sequences as follows . first , we need an algorithmic ingredient : it is the kleene function , which takes value 1 if the @xmath205 step of computation of the @xmath206 turing machine halted on the data encoded by @xmath21 and it takes value 0 if this is not the case . note that if , then . our algorithm works as follows . fix a line of the pentagrid , say @xmath207 which passes through @xmath3 , a vertex of the pentagrid which we fixed once and for all . fix @xmath208 the other line of the pentagrid which passes through @xmath3 . define @xmath20 to be a pentagon with vertex @xmath3 . call @xmath209 the end of @xmath207 which is not in the same side as @xmath20 with respect to @xmath208 . define @xmath210 to be the other vertex of @xmath20 on @xmath208 and @xmath211 to be the other vertex of @xmath20 on @xmath207 . the hat of @xmath212 is then defined by the triple @xmath213 . for each @xmath21 and @xmath121 , we define a quarter of @xmath214 by its head @xmath215 and its hat : @xmath216 . define , , , and . we define a flag by . it is clear that for each @xmath21 , the sequence is a sequence of embedded quarters . the sequence is stepwise and it has no alternation by construction . if the turing machine numbered by @xmath21 does not halt on the data @xmath21 , then for all @xmath121 and so , the sequence is ultimately direct , it is even direct , so that , by lemma [ non_alternate ] , the sequence @xmath217 converges to @xmath209 . if the turing machine numbered by @xmath21 halts on the data @xmath21 , there is an integer @xmath118 such that and . accordingly , but , afterwards , the sequence satisfies and . the sequence is again ultimately alternate but , this time , it converges to the end @xmath174 of @xmath127 which is contained in the other side of @xmath208 with respect to @xmath20 . now , by construction , as @xmath218 is a common perpendicular to @xmath207 and @xmath127 , these lines are non - secant , in particular , they can not be parallel . accordingly , . now , if , among the lines of the pentagrid which defines these ends , we can find two of them @xmath219 and @xmath220 such that denoting by @xmath176 , @xmath221 the half - plane defined by @xmath219 , @xmath220 respectively and which touches @xmath174 , @xmath175 respectively , we get . and this can be performed algorithmically if . now , if we had an algorithm which could tell us whether these limits are the same or not , this algorithm could be used to decide the halting problem for turing machines , which is known to be impossible . now , we can prove another result of the same flavor . [ noconv ] we can construct a sequence of quarters whose vertices are @xmath222 such that for each @xmath21 the sequence converges to a point at infinity @xmath223 and such that the sequence can not algorithmically converge to any point at infinity . note that if the sequence converges , it must converge to a point at infinity . proof of theorem [ noconv ] . consider the same function as previously . we change the construction as follows . we construct a sequence of sequences , again defining a quarter by its head and its hat . in what follows , the hat will be given as previously but the order of the vertices is important . in a tile @xmath224 , we consider that the hat is @xmath225 . let @xmath226 be the side which is opposite to @xmath222 . we consider that @xmath226 is at the * bottom * of the tile , that @xmath222 is at its top , that @xmath227 is at its left - hand side and @xmath228 at its right hand - side : we can consider that starting from @xmath227 and clockwise turning around the tile we meet @xmath222 , then @xmath228 and then the ends of @xmath229 . the side @xmath230 will be called side 0 and the side @xmath231 will be called side 1 . for each fixed @xmath21 , we start with a fixed once and for all tile with bottom @xmath232 and hat @xmath233 which will be denoted by @xmath234 . at the beginning . then we construct the sequence as follows . @xmath235 is the quarter defined by @xmath224 and its hat @xmath225 . by construction , it is plaint that for each @xmath21 and @xmath121 we have . accordingly , the sequence is stepwise , with no alternation and it is not ultimately direct : when @xmath21 is fixed , there is always a turing machine numbered with @xmath121 such that its computation on @xmath121 is completed at the @xmath236 step . from theorem [ to_end ] , the sequence tends to a point at infinity @xmath223 . by construction of the quarters , we can notice that for each @xmath121 we the rest of the sequence evolves in . fix @xmath121 and let when and when . it is not difficult to see that . more than that , the line defined by @xmath231 for @xmath237 and the line defined by @xmath230 for @xmath238 are non - secant . this means that the distance between @xmath237 and @xmath238 tends to infinity when we go to infinity on both these borders . from this remark , assume that the sequence tends to a limit @xmath239 which is also a point at infinity . then there is a half - plane @xmath93 delimited by a line @xmath138 which may be assumed to belong to the pentagrid such that there is @xmath168 such that for , all @xmath223 s are touched by @xmath93 . by the remark we made about @xmath237 and @xmath238 , we can see that , necessarily , for any , @xmath240 , otherwise , @xmath223 and @xmath241 can not be both in @xmath93 . accordingly , if we have an algorithm @xmath242 which , for each @xmath21 , gives an integer @xmath243 such that all @xmath244 with are in @xmath93 which is at distance @xmath21 from @xmath20 , then , looking at the value of @xmath245 , we know whether for ever or not . and this decides the halting problem , which is impossible . probably , other undecidability results of analysis can be transported into the hyperbolic plane in similar way . this might open a new area .

this paper presents predictions for jet pseudorapidity ( ) and transverse momentum ( ) distributions for the production of the standard model higgs boson in association with a high- hadronic jet . we discuss the contributions of electroweak loops and of bottom - quark parton processes to the cross section . the latter arise in the five - flavour scheme . predictions for the tevatron and the large hadron collider with collision energy are presented . for higgs boson masses of , and , we find the maximal effects of the electroweak contributions to the higgs plus jet and distribution to be and , respectively , for the tevatron , and and , respectively , for the lhc . for the maximal contribution of bottom - quark parton processes to the and distribution , we find and , respectively , for the tevatron , and and , respectively , for the lhc . a separate study of the higgs +-jet cross section demonstrates that a calculational approach which respects the hierarchies of yukawa couplings yields a leading order cross section prediction which is more accurate in the high- regime than conventional approaches . fr - pheno-2010 - 016 * electroweak and bottom quark contributions + to higgs boson plus jet production + * + _ physikalisches institut , albert - ludwigs - universitt freiburg + hermann - herder - str . 3 , d-79104 freiburg i m breisgau , germany + oliver.brein@physik.uni-freiburg.de_ +

1003.4438

since the discovery of the massive electroweak gauge bosons @xmath14 and @xmath15 @xcite , we are sure to observe electroweak symmetry breaking in nature . however , the underlying dynamics of this symmetry breaking has not yet been established experimentally . spontaneous symmetry breaking via the higgs mechanism @xcite is an appealing theoretical suggestion for such a dynamics . depending on the specific model considered , the higgs mechanism implies the existence of one or more scalar particles ( higgs bosons ) . therefore , the search for higgs bosons is an important task in the quest to unravel the nature of electroweak symmetry breaking at high energy collision experiments . besides inclusive single higgs production , higgs boson production in association with a high-@xmath1 hadronic jet provides a useful channel for higgs searches at hadron colliders , which allows for refined cuts increasing the signal - to - background ratio . specifically , for the standard model ( sm ) higgs plus jet process @xcite , early simulations considering the decay channels @xmath16 @xcite and @xmath17 @xcite have demonstrated promising signal - to - background ratios . while the promising findings of the parton - level study @xcite were confirmed by a simulation using the fast ( and less detailed ) simulation for the atlas detector response @xcite , an up - to - date simulation of the detailed atlas detector performance appeared in @xcite . from the cms collaboration , only a simulation of the inclusive @xmath16 channel is currently available @xcite . considerable progress has been made in improving the sm cross section predictions : the fully differential distribution for higgs production is available at next - to - next - to - leading order ( nnlo ) qcd accuracy in the large top - mass limit @xcite , improved by the resummation of logarithmically enhanced terms for low @xmath1 @xcite . dedicated calculations of higher - order corrections to differential cross sections for higgs boson production associated with a high-@xmath1 jet have been performed : the next - to - leading order ( nlo ) qcd corrections in the large top - mass limit @xcite and the corresponding resummation of large threshold logarithms @xcite . furthermore , the soft - plus - virtual gluon approximation to the nlo qcd prediction of the higgs @xmath1 distribution @xcite goes beyond the large top - mass limit . for the @xmath13-quark process @xmath18 , the nlo qcd corrections are also known @xcite , and very recently a calculation of the weak corrections to this process appeared @xcite . the remaining scale uncertainty of the nlo qcd result for the higgs plus high-@xmath1 jet production cross section via the light quark parton processes is at the level of 10% @xcite . the achievement of this level of accuracy in the qcd corrections , albeit in the large top - mass limit , motivates to study other leading order contributions which could potentially affect the higgs plus jet cross section prediction at the 10% level . specifically , there are to consider : loop - induced electroweak contributions to the processes @xmath19 and @xmath20 for the light quark flavours ( @xmath21 ) and the contributions from the processes @xmath22 and @xmath23 . all these contributions may change the differential distributions of the process , which are therefore of particular interest . parts of these contributions to the higgs plus jet cross section have been considered elsewhere . electroweak and non - zero quark - mass effects on the higgs @xmath1 distribution using 5-flavour parton distribution functions ( pdfs ) neglecting the contributions from @xmath13-quark parton processes have been studied for the lhc and the tevatron in @xcite . leading order qcd and electroweak contributions to the @xmath24 and @xmath25 processes , including electroweak contributions to the @xmath13-quark processes enhanced by the top - higgs yukawa coupling , have been studied for the lhc in @xcite . however , the impact of those contributions on the total higgs plus jet cross section , which involves also the dominant gluon fusion channel , has not been studied in @xcite . to our knowledge , a combination of the contributions of the light quark flavours ( @xmath26 ) and the bottom quarks in order to get a cross section prediction for higgs plus ( any ) jet in a 5-flavour scheme has so far only been done for the leading - order qcd predictions @xcite . in the present paper , we study the combined effect of contributions of leading order ( lo ) electroweak loop graphs and bottom - quark parton processes on the jet pseudorapidity ( @xmath0 ) and transverse momentum ( @xmath1 ) distribution of higgs + high-@xmath1 jet production . in order to obtain the best lo prediction , we employ a somewhat different scheme than usual , which particularly affects the bottom - quark processes . instead of discarding higher order contributions in the cross section predictions according to mere power counting of qcd and electroweak couplings ( @xmath27 ) , regardless of the appearing type of yukawa coupling ( @xmath28 , @xmath29 ) , we argue that keeping the lo contributions to each monomial in @xmath28 and @xmath29 and discarding the higher order contributions to each of them separately , gives a more accurate approximation to the full result . this approach is particularly well - suited for the bottom - quark processes , because in this case tree - level graphs @xmath30 and one - loop graphs @xmath31 contribute and just discarding the loop - contributions because of the higher power of @xmath32 does not do justice to the hierarchy of yukawa couplings @xmath33 in the standard model . as the higgs plus @xmath13-jet final state is also of separate interest , in particular in models beyond the sm , where the top and bottom yukawa couplings can be dramatically different , we discuss this process also separately in appropriate places below . section [ sect : partonic processes ] briefly reviews the contributing parton processes and section [ sect : calculational approach ] gives details on our calculational approach . in section [ sect : numerical results for hadronic cross sections ] we present numerical results for the distributions at the lhc and the tevatron and section [ sect : conclusions ] contains our conclusions . at the partonic level , production of a higgs boson @xmath34 together with a jet at hadron colliders is mediated by three classes of processes ( see fig . [ hjet - qcd ] ) : @xmath35 for the gluon fusion channel , the leading order contribution to the cross section prediction is of order @xmath36 in the strong and electromagnetic coupling constants @xmath37 and @xmath38 . the corresponding feynman graphs have one - loop topologies and are displayed in fig . [ hjet - qcd](a ) , where the shaded blobs represent triangle- and box - type loops of virtual quarks . the loop contributions can be further subdivided according to their scaling with the quark yukawa couplings @xmath39 , or equivalently @xmath40 , appearing in the mathematical expression for the amplitude . naturally , the dominant cross section contribution ( @xmath41 ) comes from top - quark loops in the sm . however , as has been demonstrated e.g. in @xcite , including bottom - quark loops ( @xmath42 ) changes the gluon fusion cross section by several percent . there is no electroweak contribution to gluon fusion at one - loop order . for quark gluon scattering and quark anti - quark annihilation of light quarks ( @xmath26 ) , the leading order qcd cross section prediction is also loop - induced and of order @xmath43 , where only the @xmath44 and @xmath45 terms are relevant . the prediction is calculated from the graph topologies displayed in figs . [ hjet - qcd](b ) and [ hjet - qcd](c ) with triangle - type loops of heavy quarks as subgraphs . the feynman graphs for the process @xmath46 is obtained by crossing the appropriate quark lines in fig . [ hjet - qcd](b ) . in contrast to gluon fusion , these two reactions are also mediated by electroweak one - loop graphs @xmath47 . [ ughu - ew ] exemplarily shows the full set of feynman graphs for the scattering of an up - type quark with a gluon . the corresponding feynman graphs for the processes @xmath48 and @xmath49 can be obtained by crossing the appropriate external lines in fig . [ ughu - ew ] . in the five - flavour scheme , also @xmath13-quark initiated processes for quark gluon scattering and quark anti - quark annihilation contribute to the higgs plus jet cross section . for bottom gluon scattering and @xmath50 annihilation , the qcd leading order cross section prediction is @xmath51 and comes from tree - level feynman graphs shown in fig . [ hjet - qcd](d ) . because of the hierarchy of yukawa couplings , loop - induced qcd and electroweak contributions to the @xmath13-quark parton processes , which contain a factor @xmath44 , can be as important as the tree level contribution in the sm @xcite . the qcd contributions to @xmath52 scattering and @xmath50 annihilation are shown in fig . [ hjet - qcd](b ) and fig . [ hjet - qcd](c ) , respectively , where the external quarks are @xmath13 quarks and the shaded blob represents triangle - type loops of virtual top - quarks . to the cross section prediction and can not be considered separately . while the interference of this particular graph with the tree - level qcd amplitude in our approach ( @xmath28 non - zero and @xmath53 otherwise , to be explained in the next section ) vanishes , there are more one - loop graphs which contribute to qcd correction of that order . ] the feynman graphs are @xmath31 . the electroweak contributions to @xmath52 scattering which do not vanish for @xmath54 are shown in fig . [ bghb - ew ] . the contributions to the @xmath55 and @xmath56 amplitudes can be obtained by crossing the appropriate external lines in fig . [ bghb - ew ] . the first two rows of feynman graphs in fig . [ bghb - ew ] are proportional to the top - quark yukawa coupling @xmath29 . we do not consider graphs @xmath57 here , as they only form a part of the subleading electroweak contribution @xmath58 to the cross section prediction and can not be considered separately . the @xmath13-quark initiated processes are usually not considered a relevant contribution to the higgs plus jet final state in the standard model ( sm ) , but may even become dominant in models beyond the sm with a strongly enhanced bottom - higgs yukawa coupling . this is for instance the case in the minimal supersymmetric standard model ( mssm ) for the lightest mssm higgs boson for low values of the @xmath59-boson mass @xmath60 ( @xmath61 ) @xcite . however , in our calculation , we include the @xmath13-quark parton processes in the higgs plus jet cross section calculation in order to assess the size of their contribution in comparison with taking effects of a non - vanishing bottom - higgs yukawa - coupling into account in the loop - induced light - quark parton processes . for perturbative predictions involving higgs bosons , the higgs couplings to other particles , which are proportional to their mass , create further hierarchies within loop - contributions in addition to the power series in the qcd and the electromagnetic coupling constant @xmath37 and @xmath38 . therefore , it is useful to judge the importance of contributions also according to the hierarchy of yukawa coupling constants @xmath62 ( @xmath63 ) which is induced by the hierarchy of quark masses @xmath64 . for the two heaviest quarks , this hierarchy is even more pronounced than the hierarchy between radiative corrections of consecutive perturbative qed orders . in fact , for our choice of mass parameters ( see below ) we get @xmath65 and @xmath66 and @xmath67 . in order to provide an overview of the leading order contributions to the higgs plus jet production channels subdivided by the hierarchy of yukawa couplings , we provide symbolic equations for the scattering amplitudes and the squared matrix elements which highlight the dependence on the coupling constants . the scattering amplitudes we consider have the form : @xmath68 with @xmath69 and @xmath70 . the squared matrix elements , only keeping non - vanishing contributions for @xmath53 while retaining @xmath45 non - zero ( explanation follows below ) , read : @xmath71 \alpha_s^3 \,,\end{aligned}\ ] ] @xmath72 \alpha_s^3 \big\ } \\\nonumber & + \sqrt{\alpha_t } \big\ { 2 { \text{re}}\big[({b_x^t}(q))^\star { c_x}(q)\big ] \alpha_s^2\alpha\sqrt{\alpha } \big\}\\\nonumber & + \sqrt{\alpha_b } \big\ { 2 { \text{re}}\big[({b_x^b}(q))^\star { c_x}(q)\big ] \alpha_s^2\alpha\sqrt{\alpha } \big\ } \\\nonumber & + |{c_x}(q)|^2 \alpha_s\alpha^3 \,,\ ] ] @xmath73 \alpha_s^2\alpha + |{e_x^t}(b)|^2 \alpha_s \alpha^2 \big\ } \\\nonumber & + \alpha_b \big\ { |d_x(b)|^2\alpha_s \big\ } ( 4\pi)^{-2}\\\nonumber & + \sqrt{\alpha_t } \big\ { 2{\text{re}}\big[({b_x^t}(b ) ) ^\star ( { c_x}(b ) + { f_x}(b ) ) \big ] \alpha_s^2\alpha \sqrt{\alpha } \big.\\\nonumber & \big . \hspace{1.5 cm } + 2{\text{re}}\big[({e_x^t}(b))^\star ( { c_x}(b ) + { f_x}(b))\big ] \alpha_s \alpha^2 \sqrt{\alpha } \big\}\\\nonumber & + |{c_x}(b ) + { f_x}(b)|^2 \alpha_s \alpha^3 \,.\end{aligned}\ ] ] for the @xmath13-quark processes , the interference between the tree level graphs ( see fig . [ hjet - qcd](d ) ) and one - loop qcd ( see figs . [ hjet - qcd](b),(c ) ) and electroweak graphs ( see fig . [ ughu - ew ] ) vanishes for @xmath54 . in order to provide an accurate prediction of all qcd and electroweak leading order contributions to the higgs plus jet cross section , respecting the yukawa hierarchies , we choose the following approach . in the squared amplitude of each process we take into account the full polynomial in @xmath74 and @xmath75 , i.e. we do not discard terms according to mere qcd power counting . beyond leading order , this approach would amount to a reordering of the perturbative expansion , where corrections to each of the appearing monomials in @xmath76 ( mainly @xmath77 ) has a separate power series in the qcd and electromagnetic coupling constant . furthermore , e.g. for processes with external @xmath13-quarks , additional positive powers of @xmath78 ( @xmath79 ) can appear in squared matrix elements from interference terms . in such cases , a consistent expansion of the massive - quark squared matrix element in @xmath78 , regardless whether it occurs in the coupling constant @xmath75 or directly , has to be performed . and thus further `` yukawa - suppressed '' than all other terms we take into account . this is the formal justification for retaining bottom - mass dependence only in the yukawa coupling and using @xmath54 otherwise , which makes the interference contribution vanish . ] that this approach is adequate for a study of all relevant leading contributions is confirmed by the comparably large relative size of the squared and interference contributions of the different terms of the amplitude in eqs . ( [ q - amp - sq ] ) and ( [ b - amp - sq ] ) to the cross section prediction . as an example , fig . [ fig : mesquared - relative - contributions ] demonstrates this behaviour for bottom - gluon parton scattering for @xmath80 and a cut on the scattering angle @xmath81 . while the top - loop squared contribution is quite large throughout the displayed centre - of - mass energy @xmath82 , also the contribution of the squared electroweak loops reach more than 25% for @xmath83 , which corresponds to @xmath84 . in this range , all other subleading terms still contribute more than 7% . for @xmath85 ( i.e. @xmath86 ) , the contribution of the tree - level squared amplitude alone falls below 50% . in order to represent effects of @xmath13 quarks as complete as possible , we always take them into account in quark loop graphs contributing to gluon fusion and all light quark parton processes using the pole mass @xmath78 . furthermore , we adopt the five - flavour scheme for the description of @xmath13 quark parton process . working in the five - flavour scheme , we consider a set of parton distribution functions ( pdfs ) which uses a perturbatively defined @xmath13-quark pdf where , in the procedure of fitting free parameters in the light - quark and gluon pdfs to experimental results , also @xmath13-sensitive observables have been considered . virtually all modern pdf sets are five - flavour pdf sets . using @xmath13-quark pdfs in the five - flavour scheme for predicting the @xmath34 + 1 @xmath13-jet production cross section is well justified by the good agreement of the nlo qcd result in this scheme with the corresponding nlo qcd result calculated in a four - flavour scheme @xcite . in the latter approach , among the quark flavours only @xmath26 are considered to be distributed in the proton while final state bottom quarks can only appear through gluon - initiated processes . we consider the @xmath13-quark to be a massless parton to be consistent with the parton model but retain a non - zero yukawa coupling @xmath28 . specifically for the @xmath13 quark parton process , we use the running @xmath87 mass at nlo qcd for the bottom quark , @xmath88\right\ } \,,\ ] ] as the mass parameter in @xmath28 and we choose the factorisation scale @xmath89 . in this way , the prediction for the tree level bottom - gluon scattering cross section gives a good approximation to the nlo qcd result @xcite . hadron colliders like the lhc ( tevatron ) collide protons with protons ( anti - protons ) with a total energy @xmath90 in the laboratory frame . hadronic cross sections are obtained via convolution of the parton - level cross sections with the parton distribution functions ( pdfs ) and summation over the various contributing partons . experimental restrictions to the detectability of the produced particles are conventionally realized by imposing specific cuts to the kinematically allowed phase space . typically , cuts are imposed on the final - state transverse momentum and/or the pseudorapidity in order to have high-@xmath1 jets not too close to the beam axis . in our case , we choose the selection criteria , latexmath:[\[\begin{aligned } \label{the - cuts } p_t & > p_t^{{\text{min}}}\ , , & @xmath1 and @xmath92 denote transverse momentum and pseudorapidity of the final state parton . we evaluate the differential hadronic cross sections @xmath93 and @xmath94 ( with @xmath95 or @xmath96 ) in the presence of those cuts as described in detail in @xcite . the numerical evaluation has been carried out with the nlo mstw2008 pdfs @xcite and a consistently chosen strong coupling constant @xmath97 , i.e. using the formula including the nlo qcd corrections ( see e.g. ref . @xcite ) for @xmath98 with @xmath99 . the renormalisation scale @xmath100 and factorisation scale for the gluon and the light quarks @xmath101 are chosen both equal to @xmath102 . for the bottom - quark factorisation scale @xmath103 we choose @xmath104 . in the calculation , the pole masses for the top and bottom quark are set to the values @xmath105 while the values of the electroweak parameters are chosen as @xmath106 the ckm matrix is assumed to be diagonal . we have evaluated the partonic cross sections with the help of the computer programs feynarts 3.2 and formcalc 3.2 @xcite and further convoluted them with pdfs according to formulae given in @xcite . for the evaluation of the nlo mstw2008 pdfs @xcite and the corresponding @xmath97 we used lhapdf @xcite . for the numerical evaluation of all cross sections we use @xmath107 and the cuts in eq . ( [ the - cuts ] ) with @xmath108 ( as in ref . @xcite ) and @xmath109 , i.e. we use the former ( latter ) in the calculation of @xmath0 ( @xmath1 ) distributions . at the tevatron , the contributions of quark - gluon scattering and gluon fusion to the higgs plus jet cross section are comparable in size , e.g. for @xmath110 and the given cuts , quark - gluon scattering contributes in total 41.5% and bottom - gluon scattering alone 1.4% . furthermore , @xmath25 annihilation contributes in total 4.5% and @xmath50 annihilation alone 0.23% . we start by presenting the impact of electroweak loops and @xmath13-quark processes on the differential cross section predictions for quark - gluon scattering and quark anti - quark annihilation separately . [ fig : pt.mh120.qg.tev](a ) shows the jet transverse momentum ( @xmath1 ) distribution of higgs plus jet production via quark - gluon scattering calculated in four approximations : either including or neglecting electroweak loop contributions combined with either including or neglecting the @xmath13-quark parton contributions in the prediction , represented by the identifiers `` all '' , `` all , no b '' , `` qcd '' and `` qcd , no b '' in the figure . for comparison , the contributions of gluon fusion ( including top- and bottom - loop graphs in the amplitude ) and of bottom - gluon scattering to the higgs plus jet cross section are also shown in fig . [ fig : pt.mh120.qg.tev](a ) . among the parton contributions to the higgs plus jet @xmath1 distribution , quark gluon scattering is the largest contribution for @xmath1 above about @xmath111 , while gluon fusion dominates at lower @xmath1 . the relative contribution of electroweak loops is very small at low @xmath1 and increases with rising @xmath1 , while it is the other way around with the contributions of @xmath13-quark parton processes . [ fig : pt.mh120.qg.tev](b ) shows relative differences @xmath112 between the four approximations to the quark - gluon scattering cross section displayed in fig . [ fig : pt.mh120.qg.tev](a ) . the effect of including electroweak contributions depends on @xmath1 ( dashed line ) and the overall effect ranges from + 5% to -12% for @xmath113 . the effect of the electroweak loops contributing to the @xmath13-quark processes amounts to a slight rise of the higgs plus jet cross section by at most 2% . it is approximately given by the difference between the dashed and dotted lines in fig . [ fig : pt.mh120.qg.tev](b ) , where the latter represents the relative effect of electroweak loops in the light - quark parton processes only which corresponds to ( and agrees with ) results of ref . @xcite . [ fig : pt.mh120.qg.tev](c ) shows the bottom - gluon scattering cross section alone calculated in four approximations . while the identifier `` bg , all '' in fig . [ fig : pt.mh120.qg.tev](c ) corresponds to retaining all contributions to the amplitude non - zero in eq . ( [ b - amp ] ) , `` bg , qcd '' corresponds to setting @xmath114 to zero , i.e. the amplitude is approximated by the tree - level graphs in fig . [ hjet - qcd](d ) and the top - loop graphs in fig . [ hjet - qcd](b ) . the identifier `` bg , tree '' indicates that only the tree - level amplitude is considered , i.e. only @xmath115 in eq . ( [ b - amp ] ) is non - zero , and `` bg , loops only '' corresponds to setting only @xmath115 to zero . in the @xmath1 range displayed , none of the contributions to bottom - gluon scattering is negligible . for the prediction of the @xmath1 distribution of @xmath52 scattering , it turns out that discarding the squared top - loop graphs in the cross section prediction ( @xmath116 in eq . ( [ b - amp - sq ] ) ) , because their contribution is formally of higher order in @xmath37 , as has often been done in the past , is a bad approximation for larger @xmath1 values . although this contribution to the partonic @xmath52 cross section is smaller than the tree - level contribution at low centre - of - mass energy @xmath82 , it has a flatter energy dependence . thus , it becomes larger than the tree - level contribution beyond a certain energy @xmath117 . for instance , for @xmath80 and a cut on the scattering angle @xmath81 , we get @xmath118 and for @xmath119 the top - loop contribution is already more than @xmath120 of the tree - level contribution ( see fig . [ fig : mesquared - relative - contributions ] ) . the situation is similar for the contribution of electroweak loop graphs . taking those into account as well , the turnover point where the loop - squared becomes larger than the tree - squared is @xmath121 and for @xmath122 the loop contribution is @xmath123 of the tree - level contribution . as final state jets with a certain @xmath1 introduce the energy threshold @xmath124 the above behaviour implies that for @xmath125\,{\text{gev}}$ ] top - quark [ electroweak and top - quark ] loop contributions dominate the @xmath1 distribution . therefore , we argue that these loop contributions , which do not interfere with the tree - level cross section for @xmath54 , need to be taken into account for an accurate leading order prediction of the bottom gluon scattering contribution to higgs plus jet . while a mere qcd power counting would suggest otherwise , clearly , the hierarchy of yukawa couplings in the sm also suggests not to discard loop - squared parts @xmath126 in the cross section prediction . [ fig : eta.mh120.qg.tev ] shows the same breakdown of contributions as the previous figure , now for the pseudorapidity distribution @xmath127 of the final state jet . electroweak and @xmath13-quark parton contributions depend strongly on @xmath0 ( dashed and solid line in fig . [ fig : eta.mh120.qg.tev](b ) ) and are roughly of the same order ( but opposite sign ) with extrema of about @xmath128 and 9% , respectively , at @xmath129 , i.e. for a jet radiated into the central part of the detector . the explanation of the strong @xmath0 dependence of these contributions has two elements . one element is the appearance of feynman graph topologies in the amplitudes with different angular dependence than in the dominant light - quark qcd amplitudes . the other is the different angular dependence in the laboratory frame caused by parton collisions which have preferentially boosted final states due to unbalanced valence quark contributions ( double - peak in @xmath127 for light - quark gluon scattering ) as compared to parton collisions with no valence quarks involved ( central peak in @xmath127 for bottom gluon scattering ) . the @xmath52 scattering contribution to the @xmath0 distribution in fig . [ fig : eta.mh120.qg.tev](c ) illustrates how much the large contribution of loop - squared terms ( qcd top - loop and electroweak ) to the @xmath1 distribution in fig . [ fig : pt.mh120.qg.tev](c ) contribute significantly to @xmath127 , i.e. after @xmath1 integration for given @xmath0 . the bottom - gluon scattering cross section prediction including electroweak and top - loop graphs in the amplitude ( solid line ) is about 40% larger than the tree - level prediction ( dotted line ) at @xmath130 . the bulk of the contributions comes from including the top - loop graphs ( dot - dashed line ) which leads to a prediction that deviates from the full result for @xmath52 scattering only by about @xmath131 . [ fig : pt.mh120.qq.tev](a ) shows the jet @xmath1 distribution via the @xmath25 annihilation channel calculated in the same four approximations as described in the previous section ( [ sect : quark - gluon scattering ] ) . also here , the hadronic cross section via gluon fusion and @xmath50 annihilation are shown for comparison . bottom quarks dominate the @xmath25 contribution to higgs+jet below @xmath132 . the reason for this dominance is that @xmath50 annihilation is mediated at tree - level by @xmath133- and @xmath134-channel virtual @xmath13-quark exchange graphs ( see fig . [ hjet - qcd](d ) ) which make the cross section large at small non - zero @xmath1 and it would rise indefinitely for @xmath135 . in contrast to that , the contributions from light - quark parton processes vanish in this limit . this rise of the @xmath50 annihilation cross section for small @xmath1 is analogous to the situation for the gluon - fusion and quark - gluon scattering process , where @xmath133- or @xmath134-channel virtual gluon exchange graphs cause a similar behaviour . in the low @xmath1 region , soft gluon resummation is called for in order to obtain a reliable prediction . however , we do not expect that the relative sizes of different contributions to the higgs plus jet cross section , which is our main concern here , are altered greatly by resummation effects . the contribution of @xmath25 annihilation reaches the size of gluon fusion above @xmath136 . remarkably , this contribution rises with @xmath1 up to about @xmath137 despite the fact that the parton luminosity is falling at the same time . this behaviour is caused by a pronounced threshold peak in the energy dependence of the partonic cross sections for @xmath138 ( @xmath70 ) which enters the phase space integration over jet final states with a given @xmath1 . the threshold peak is located around @xmath139 and is due to the top - loop graphs of the qcd part of the amplitude @xcite . the influence of this peak on the hadronic @xmath1 distributions is strong if it occurs near the energy threshold for jets with given @xmath1 : @xmath140 in eq . ( [ jet - energy - threshold ] ) . equating @xmath141 , we get the transverse momentum where the peak exactly occurs at the boundary of the integration over jet phase space : @xmath142 approaching @xmath143 from below , the double effect of the pdfs falling with rising @xmath140 and the cross section for @xmath144 rising because of the peak moving closer towards @xmath140 is an overall rise of the @xmath1 distribution . for @xmath145 , the peak region falls outside of the region of jet phase space integration and thus @xmath146 drops sharply . [ fig : pt.mh120.qq.tev](b ) and [ fig : pt.mh120.qq.tev](c ) show relative differences between results of the four approximations . in particular , fig . [ fig : pt.mh120.qq.tev](b ) zooms in on the special area @xmath147 . naturally , as the quark annihilation contribution of @xmath13-quarks rises and of light quarks vanishes with @xmath148 , respectively , the relative contribution of the @xmath13-quark processes ( solid line ) reaches almost a factor of 100 at @xmath149 . while the electroweak contributions to the light - quark processes lead to large and negative interference ( dotted line ) at low @xmath1 ( e.g. @xmath150 for @xmath151 ) , the electroweak contributions to the @xmath13-quark processes are much smaller in the same @xmath1 region ( e.g. @xmath152 for @xmath151 ) . hence , with increasing dominance of the @xmath13-quark processes towards lower @xmath1 , the relative contribution of electroweak loops in the sum of all quarks ( dashed line ) is diminished . we confirm the results for the electroweak loop effects in the light - quark @xmath25 annihilation processes @xcite but also demonstrate the importance of including the @xmath13-quark initiated processes as well . in fig . [ fig : pt.mh120.qq.tev](d ) , a breakdown of different contributions to the @xmath1 distribution of the @xmath50 annihilation process analogous to the one in fig . [ fig : pt.mh120.qg.tev](c ) , described in section [ sect : quark - gluon scattering ] , demonstrates the smallness of all loop contributions below @xmath153 . from this information , one can already conclude that the @xmath0 distribution for the given cut of @xmath154 will be practically independent of the loop contributions . the hadronic cross section via @xmath50 using only loop graphs in the amplitude ( dashed line ) , shows qualitatively similar behaviour with @xmath1 as the light - quark annihilation cross section . this is also due to a threshold peak caused by top - loop graphs in the partonic scattering amplitude , as explained above . in fig . [ fig : eta.mh120.qq.tev ] , the @xmath0 distribution of the @xmath25 annihilation channel is shown split in separate contributions as in previous figures . the hadronic cross section via @xmath25 calculated taking electroweak and @xmath13-quark contributions into account ( solid line in fig . [ fig : eta.mh120.qq.tev](a ) ) and neglecting them both ( @xmath155-dashed line ) are roughly equal . this shows that for the given kinematical cuts , electroweak and @xmath13-quark contributions roughly compensate each other , as can be seen more quantitatively in fig . [ fig : eta.mh120.qq.tev](b ) . fig . [ fig : eta.mh120.qq.tev](c ) demonstrates the expected smallness of the loop contributions to the @xmath50 channel . we discuss the impact of contributions from electroweak loops and @xmath13-quark parton processes on the total higgs plus high-@xmath1 jet cross section , i.e. the sum of the gluon fusion , quark - gluon scattering and @xmath25 annihilation contribution . [ fig : reldiff.tev](a ) and [ fig : reldiff.tev](b ) show the relative size of the contributions to the @xmath1 and @xmath0 distribution , respectively , for three different higgs masses : @xmath80 , @xmath4 and @xmath5 . a feature , both present in the @xmath1 and the @xmath0 distribution , is that the @xmath13-quark contributions decrease and electroweak contributions increase with rising higgs mass . in general , effects on the @xmath1 distribution ( see fig . [ fig : reldiff.tev](a ) ) by @xmath13-quark contributions ( black lines ) decrease and electroweak contributions ( green lines ) increase with @xmath1 . the largest effect of @xmath13-quark processes amounts to an increase of about 3% at low @xmath1 for @xmath80 . the largest effect of the electroweak contributions is a decrease of about @xmath6 for @xmath1-values in the range 110 to @xmath4 and @xmath156 . [ fig : reldiff.tev](b ) shows the impact of electroweak and @xmath13-quark contributions on the @xmath0 distribution . both contributions have their largest magnitude at @xmath129 : @xmath11 for @xmath110 for the @xmath13-quark contribution and @xmath7 for @xmath157 for the electroweak contribution . for the numerical evaluation of all cross sections we use @xmath158 and the cuts in eq . ( [ the - cuts ] ) with @xmath159 and @xmath160 which have been used in previous sm studies for the lhc @xcite . at the end of section [ sect : lhc : corrections to higgs + jet distributions ] we make some comments on the @xmath161 dependence of the results . at the lhc , gluon fusion dominates largely over quark - gluon scattering and the @xmath25 annihilation contribution is quite small , e.g. for @xmath110 and the given cuts , quark - gluon scattering and @xmath25 annihilation contribute 27.8% and 0.8% , respectively , to the total higgs plus jet cross section . the contribution of @xmath52scattering and @xmath50 annihilation alone is 2% and 0.3% , respectively . note that detailed descriptions of certain common features of the results which have been already given for the tevatron case in section [ sect : tevatron ] will not be given for the lhc case . [ fig : pt.mh120.qg.lhc](a ) shows the @xmath1 distribution of higgs plus jet production via quark - gluon scattering calculated in the four approximations ( `` all '' , `` all , no b '' , `` qcd '' and `` qcd , no b '' ) described in section [ sect : quark - gluon scattering ] . the hadronic cross section via gluon fusion and bottom - gluon scattering are also shown in fig . [ fig : pt.mh120.qg.lhc](a ) for comparison . similar to the tevatron case , the relative contribution of electroweak loops , shown in fig . [ fig : pt.mh120.qg.lhc](b ) , is very small at low @xmath1 and increases with rising @xmath1 , while it is the other way around for the @xmath13-quark contributions . the electroweak contribution ( dashed line ) ranges from @xmath162 to @xmath163 in the displayed @xmath1 interval from 10 gev to 200 gev . the relative @xmath13-quark contribution ( solid line ) ranges from 20% at low to 4.7% at high @xmath1 . the effect of the electroweak loops contributing to the @xmath13-quark processes , which cause an increase of about @xmath164 of the cross section for these subprocesses for @xmath165 ( see fig . [ fig : pt.mh120.qg.lhc](c ) ) , amounts to a slight rise of the total higgs+jet cross section by at most 2.5% for @xmath1 above @xmath166 . [ fig : pt.mh120.qg.lhc](c ) shows the bottom - gluon scattering cross section alone calculated in four approximations ( `` bg , all '' , `` bg , qcd '' , `` bg , tree '' , and `` bg , loops only '' ) described in section [ sect : quark - gluon scattering ] . in the @xmath1 range displayed , similar to the tevatron case , none of the contributions is negligible , confirming results of ref . @xcite . [ fig : eta.mh120.qg.lhc ] shows the same breakdown of contributions as the previous figure , for the jet pseudorapidity distribution @xmath127 . electroweak and @xmath13-quark parton contributions depend strongly on @xmath0 ( dashed and solid line in fig . [ fig : eta.mh120.qg.lhc](b ) ) and have extrema of about @xmath128 and @xmath167 , respectively , at @xmath129 . fig . [ fig : pt.mh120.qq.lhc](a ) shows the jet @xmath1 distribution via the @xmath25 annihilation channel calculated in the same four approximations as described in section [ sect : quark - gluon scattering ] and of the gluon fusion and @xmath50 annihilation channel for comparison . as the qualitative behaviour of this channel is similar to the tevatron case , we refer to section [ sect : tevatron quark anti - quark annihilation ] for their explanation . the @xmath50 contribution dominates the @xmath25 contribution to the higgs+jet cross section below @xmath168 . this dominance is even more pronounced than in the tevatron prediction ( see section [ sect : tevatron quark anti - quark annihilation ] ) and is again due to tree - level @xmath133- and @xmath134-channel virtual @xmath13-quark exchange graphs ( see fig . [ hjet - qcd](d ) ) . this contribution would rise indefinitely for @xmath135 while the contribution from light - quark parton processes vanish in this limit . [ fig : pt.mh120.qq.lhc](b ) and [ fig : pt.mh120.qq.lhc](c ) show relative differences between results of the four approximations . in particular , fig . [ fig : pt.mh120.qq.lhc](b ) zooms in on the special area @xmath169 . naturally , as the annihilation contribution of @xmath13-quarks rises and of light quarks vanishes with @xmath148 , respectively , the relative contribution of the @xmath13-quark processes ( solid line ) reaches a factor of about 370 at @xmath149 . however , the @xmath13-quark contribution is not only relevant in the low @xmath1 region , in which soft gluon resummation is called for in order to obtain a reliable prediction . it stays above 20% up to @xmath170 which is where it first falls behind the electroweak contributions in significance . while the electroweak contributions to the light - quark processes lead to large and negative interference ( dotted line ) at low @xmath1 ( e.g. @xmath171 for @xmath151 ) , confirming results of @xcite , the electroweak contributions to the @xmath13-quark processes alone , in the same @xmath1 region , are very small ( e.g. @xmath172 for @xmath151 ) . hence , the increasing dominance of the @xmath13-quark processes towards lower @xmath1 leads to a decrease of the relative contribution of electroweak loops in the sum of all quark parton contributions ( dashed line ) . in fig . [ fig : pt.mh120.qq.lhc](d ) , a breakdown of different contributions to the @xmath1 distribution of the @xmath50 annihilation process analogous to fig . [ fig : pt.mh120.qg.lhc](c ) demonstrates that sizeable loop contributions to this process only set in above @xmath153 . in fig . [ fig : eta.mh120.qq.lhc ] , the @xmath0 distribution of the @xmath25 annihilation channel is shown split in separate contributions as in previous figures . the fact that electroweak and @xmath13-quark contributions are sizeable is reflected by the clear separation of all lines in fig . [ fig : eta.mh120.qq.lhc](a ) . for the given cut of @xmath173 , the @xmath13-quark contribution ( solid line in fig . [ fig : eta.mh120.qq.lhc](b ) ) still reaches 77% at @xmath129 , while the electroweak contribution ( dashed line ) varies between @xmath174 and @xmath175 with a minimum at @xmath129 . similar to the tevatron case , loop contributions do not affect the @xmath0 distribution of the @xmath50 channel for the given cuts ( see fig . [ fig : eta.mh120.qq.lhc](c ) ) . having studied the effects of electroweak loops and @xmath13-quark processes on quark - gluon scattering and @xmath25 annihilation separately , we discuss here the impact of those contributions on the total higgs plus high-@xmath1 jet cross section . [ fig : reldiff.lhc](a ) and [ fig : reldiff.lhc](b ) show the relative size of the contributions to the @xmath1 and @xmath0 distribution , respectively , for three different higgs masses : @xmath80 , @xmath4 and @xmath5 . the qualitative features of the @xmath13-quark and electroweak contributions to the @xmath1 and @xmath0 distributions are similar to the tevatron case , e.g. @xmath13-quark decrease and electroweak contributions increase with rising higgs mass , respectively . on the quantitative level , the maximal effect of the electroweak loops on the @xmath1 distribution ( green lines in fig . [ fig : reldiff.lhc](a ) ) is diminished to a @xmath8 change of the cross section in the range @xmath176 and for @xmath177 , compared to @xmath6 for the tevatron case . the reason for this difference is that the @xmath24 and @xmath25 channels , which are affected at leading order by electroweak loops , dominate the higgs + jet cross section for large @xmath1 at the tevatron , while at the lhc ( with @xmath158 ) gluon fusion dominates over the full @xmath1 range . contrary to that , the @xmath13-quark contributions to the @xmath1 distribution ( black lines ) are slightly higher throughout the displayed @xmath1 range , reaching maximally @xmath12 at low @xmath1 for @xmath80 . [ fig : reldiff.lhc](b ) shows the impact of electroweak and @xmath13-quark contributions on the @xmath0 distribution . both contributions have their largest magnitude at @xmath129 : @xmath178 for @xmath110 for the @xmath13-quark contribution and @xmath179 for @xmath157 for the electroweak contribution . the large difference between the electroweak contributions for larger @xmath1 values in the tevatron and lhc case is not seen in the @xmath0 distributions . this is to be expected , as the @xmath1 range where the difference in @xmath146 is substantial contributes only little to the cross section @xmath127 . as the collision energy of the lhc will probably be changed a few times before the design goal of @xmath180 is reached , we have also studied the @xmath90 dependence of the electroweak and @xmath13-quark contributions to the total higgs plus jet cross section ( with the given cuts : @xmath159 and @xmath160 ) for the three higgs masses @xmath3 , @xmath4 and @xmath5 . we find that the relative contributions depend only mildly on the collider energy . for instance , for @xmath90 from @xmath181 to @xmath180 , the electroweak contributions for @xmath157 vary from -2.2% to -1.5% and the @xmath13-quark contributions for @xmath110 vary from + 2.3% to + 2.45% . as is clear from figs . [ fig : reldiff.lhc](a ) and [ fig : reldiff.lhc](b ) , a tighter @xmath1-cut will increase the relative electroweak contribution while the relative @xmath13 quark contribution will stay almost the same . a tighter @xmath0-cut will increase the relative size of both contributions . we have calculated specific contributions to jet pseudorapidity and transverse momentum distributions for the standard model higgs plus high-@xmath1 jet production cross section at the lhc and the tevatron . the remaining scale uncertainty of the nlo qcd prediction ( in the large top - mass limit ) for this higgs production mode via light quark parton processes is at the level of 10% @xcite . motivated by this , we discussed here the contributions of electroweak loops and of bottom - quark parton processes ( in the five - flavour scheme ) to cross section predictions for the tevatron and the large hadron collider with @xmath2 collision energy . for higgs bosons with a mass of @xmath3 , @xmath4 and @xmath5 , we find the maximal effects of the electroweak contributions to the total higgs plus jet @xmath1 and @xmath0 distribution to be @xmath6 and @xmath7 , respectively , for the tevatron , and @xmath8 and @xmath9 , respectively , for the lhc . for the maximal contribution of bottom - quark parton processes to the @xmath1 and @xmath0 distribution , we find @xmath10 and @xmath11 , respectively , for the tevatron , and @xmath12 and @xmath10 , respectively , for the lhc . those contributions are smaller but still comparable in size to the numerical impact of including bottom quarks of non - zero mass in quark - loop mediated contributions to the higgs plus jet cross section prediction . for both colliders , the magnitude of the electroweak contribution rises with higgs mass , while the bottom parton contribution falls with higgs mass . a separate study of the higgs + @xmath13-jet cross section demonstrates that a calculational approach which respects the hierarchies of yukawa couplings yields a leading order cross section prediction which is more accurate in the high-@xmath1 regime than conventional approaches . i thank frank petriello for the efficient comparison of our numerical results . i thank stefan dittmaier , alberto guffanti , karl jakobs and christian schwinn for valuable comments and useful discussions . this work was supported by the helmholtz alliance ha-101 physics at the terascale and in part by the european community s marie - curie research training network under contract mrtn - ct-2006 - 035505 ` tools and precision calculations for physics discoveries at colliders ' ( heptools ) . g. arnison _ et al . _ [ ua1 collaboration ] , phys . b * 122 * ( 1983 ) 103 ; m. banner _ [ ua2 collaboration ] , phys . b * 122 * ( 1983 ) 476 ; g. arnison _ et al . _ [ ua1 collaboration ] , phys . b * 126 * ( 1983 ) 398 ; p. bagnaia _ et al . _ [ ua2 collaboration ] , phys . b * 129 * ( 1983 ) 130 . p. w. higgs , phys . * 12 * ( 1964 ) 132 ; phys . * 13 * ( 1964 ) 508 ; phys . rev . * 145 * ( 1966 ) 1156 ; f. englert and r. brout , phys . rev . * 13 * ( 1964 ) 321 ; g. s. guralnik , c. r. hagen and t. w. b. kibble , phys . * 13 * ( 1964 ) 585 . r. k. ellis , i. hinchliffe , m. soldate and j. j. van der bij , nucl . b * 297 * ( 1988 ) 221 ; m. chaichian , i. liede , j. lindfors and d. p. roy , phys . b * 198 * ( 1987 ) 416 [ erratum - ibid . b * 205 * ( 1987 ) 595 ] . u. baur and e. w. glover , nucl . b * 339 * ( 1990 ) 38 . b. mellado , w. quayle and s. l. wu , phys . b * 611 * ( 2005 ) 60 . g. aad _ et al . _ [ the atlas collaboration ] , cern report cern - open-2008 - 020 , arxiv:0901.0512 [ hep - ex ] . g. l. bayatian _ et al . _ [ cms collaboration ] , j. phys . g * 34 * ( 2007 ) 995 . g. bozzi , s. catani , d. de florian and m. grazzini , phys . b * 564 * ( 2003 ) 65 ; s. catani , d. de florian , m. grazzini and p. nason , jhep * 0307 * ( 2003 ) 028 . d. de florian , m. grazzini and z. kunszt , phys . rev . lett . * 82 * ( 1999 ) 5209 ; v. ravindran , j. smith and w. l. van neerven , nucl . b * 634 * ( 2002 ) 247 ; c. j. glosser and c. r. schmidt , jhep * 0212 * ( 2002 ) 016 . j. campbell , r. k. ellis , f. maltoni and s. willenbrock , phys . d * 67 * ( 2003 ) 095002 ; s. dittmaier , m. krmer and m. spira , phys . d * 70 * ( 2004 ) 074010 . s. dawson , c. b. jackson , l. reina and d. wackeroth , phys . * 94 * ( 2005 ) 031802 and mod . phys . lett . a * 21 * ( 2006 ) 89 . s. dawson and p. jaiswal , arxiv:1002.2672 [ hep - ph ] . w. y. keung and f. j. petriello , phys . d * 80 * ( 2009 ) 013007 . j. m. campbell _ et al . _ , arxiv : hep - ph/0405302 ; k. a. assamagan _ et al . _ [ higgs working group collaboration ] , arxiv : hep - ph/0406152 ; c. buttar _ [ standard model and higgs working group ] , arxiv : hep - ph/0604120 . d. rainwater , m. spira and d. zeppenfeld , hep - ph/0203187 ; t. plehn , phys . rev . d * 67 * ( 2003 ) 014018 ; f. maltoni , z. sullivan and s. willenbrock , phys . rev . d * 67 * ( 2003 ) 093005 ; e. boos and t. plehn , phys . rev . d * 69 * ( 2004 ) 094005 ; r. v. harlander and w. b. kilgore , phys . d * 68 * ( 2003 ) 013001 . a. d. martin , w. j. stirling , r. s. thorne and g. watt , arxiv:0901.0002 [ hep - ph ] . w. m. yao _ et al . _ [ particle data group ] , j. phys . g * 33 * ( 2006 ) 1 . j. kblbeck , m. bhm , a. denner , comput . * 60 * ( 1990 ) 165 ; h. eck , ph.d . thesis , university of wrzburg ( 1995 ) ; t. hahn and m. perez - victoria , comput . phys . commun . * 118 * ( 1999 ) 153 ; t. hahn , comput . * 140 * ( 2001 ) 418 . m. r. whalley , d. bourilkov and r. c. group , arxiv : hep - ph/0508110 ; + http://hepforge.cedar.ac.uk/lhapdf/ qcd contributions to the scattering amplitudes of the partonic subprocesses ( a ) @xmath182 , ( b ) @xmath183 , ( c ) @xmath138 ( @xmath184 ) and ( d ) @xmath22 at leading order . the shaded blob represents a quark loop ( only top- and bottom loops contribute significantly ) . the symbol below each graph indicates to which coefficient in the mathematical expressions for the amplitudes ( eqs . ( [ gg - amp ] ) and ( [ q - amp ] ) ) this graph contributes . ] electroweak loop contributions to the @xmath185 scattering amplitude at leading order , assuming no up - quark higgs yukawa coupling . the depicted graphs contribute all to the coefficient @xmath186 in the mathematical expression for the amplitude ( eq . ( [ q - amp ] ) ) . the contributions look similar for the scattering of other quark flavours . ] electroweak loop contributions to the @xmath18 scattering amplitude at leading order which do not vanish for @xmath53 . the symbol below each graph indicates to which coefficient in the mathematical expression for the amplitude ( eq . ( [ b - amp ] ) ) this graph contributes . ] relative contributions of different parts of the squared matrix element @xmath187 ( see eq . ( [ b - amp - sq ] ) ) to the integrated partonic cross section of @xmath188 as a function of centre - of - mass energy @xmath82 for @xmath80 and a cut on the scattering angle @xmath81 . ] @xmath1 distribution for quark - gluon scattering at the tevatron : ( a ) quark parton processes with and without the @xmath13 quark contributions , with and without electroweak contributions ; ( b ) relative differences to the left panel ; ( c ) contributions to the @xmath13 quark parton processes . the depicted approximations are described in the main text . ] @xmath0 distribution for quark - gluon scattering at the tevatron : ( a ) quark parton processes with and without the @xmath13 quark contributions , with and without electroweak contributions ; ( b ) relative differences to the left panel ; ( c ) contributions to the @xmath13 quark parton processes . the depicted approximations are described in the main text . ] @xmath1 distribution for quark anti - quark annihilation at the tevatron : ( a ) quark parton processes with and without the @xmath13 quark contributions , with and without electroweak contributions ; ( b),(c ) relative differences to the left panel ; ( d ) contributions to the @xmath13 quark parton processes . the depicted approximations are described in the main text . ] @xmath0 distribution for quark anti - quark annihilation at the tevatron : ( a ) quark parton processes with and without the @xmath13 quark contributions , with and without electroweak contributions ; ( b ) relative differences to the left panel ; ( c ) contributions to the @xmath13 quark parton processes . the depicted approximations are described in the main text . ] relative size of the effects of including @xmath13-quark parton processes ( black lines ) and electroweak contributions ( green lines ) on the tevatron ( @xmath189 ) differential distributions for higgs + jet production : ( a ) @xmath1 distribution and ( b ) @xmath0 distribution of the recoiling jet . ] @xmath1 distribution for quark - gluon scattering at the lhc ( @xmath190 ) : ( a ) quark parton processes with and without the @xmath13 quark contributions , with and without electroweak contributions ; ( b ) relative differences to the left panel ; ( c ) contributions to the @xmath13 quark parton processes . the depicted approximations are described in the main text . ] @xmath0 distribution for quark - gluon scattering at the lhc ( @xmath190 ) : ( a ) quark parton processes with and without the @xmath13 quark contributions , with and without electroweak contributions ; ( b ) relative differences to the left panel ; ( c ) contributions to the @xmath13 quark parton processes . the depicted approximations are described in the main text . ] @xmath1 distribution for quark anti - quark annihilation at the lhc ( @xmath190 ) : ( a ) quark parton processes with and without the @xmath13 quark contributions , with and without electroweak contributions ; ( b ) relative differences to the left panel ; ( c ) contributions to the @xmath13 quark parton processes . the depicted approximations are described in the main text . ] @xmath0 distribution for quark anti - quark annihilation at the lhc ( @xmath190 ) : ( a ) quark parton processes with and without the @xmath13 quark contributions , with and without electroweak contributions ; ( b ) relative differences to the left panel ; ( c ) contributions to the @xmath13 quark parton processes . the depicted approximations are described in the main text . ] relative size of the effects of including @xmath13-quark parton processes ( black lines ) and electroweak contributions ( green lines ) on the lhc ( @xmath190 ) differential distributions for higgs + jet production : ( a ) @xmath1 distribution and ( b ) @xmath0 distribution of the recoiling jet . ]

motivated by future collider proposals that aim to measure the higgs properties precisely , we study the partial decay widths of the lightest higgs boson in the minimal supersymmetric standard model with an emphasis on the parameter region where all superparticles and heavy higgs bosons are not accessible at the lhc . taking account of phenomenological constraints such as the higgs mass , flavor constraints , vacuum stability , and perturbativity of coupling constants up to the grand unification scale , we discuss how large the deviations of the partial decay widths from the standard model predictions can be . these constraints exclude large fraction of the parameter region where the higgs widths show significant deviation from the standard model predictions . nevertheless , even if superparticles and the heavy higgses are out of the reach of lhc , the deviation may be large enough to be observed at future collider experiments . ut-15 - 02 + ipmu-15 - 0018 + kek - th-1795 + february , 2015 + .75 in * footprints of supersymmetry on higgs decay * .75 in motoi endo , takeo moroi , and mihoko m. nojiri 0.25 in _ department of physics , university of tokyo , tokyo 113 - 0033 , japan _ 0.1 in _ kavli ipmu ( wpi ) , university of tokyo , kashiwa , chiba 277 - 8583 , japan _ 0.1 in _ kek theory center , ipns , kek , tsukuba , ibaraki 305 - 0801 , japan _ 0.1 in _ the graduate university of advanced studies ( sokendai ) , + tsukuba , ibaraki 305 - 0801 , japan _ .5 in

1502.03959

the discovery of the higgs boson at atlas and cms experiments @xcite made a revolutionary impact on the field of particle physics . it not only confirmed the so - called higgs mechanism for the electroweak symmetry breaking , but also opened a new possibility to perform a precise test of the standard model ( sm ) by studying the properties of the higgs boson . in the sm , the coupling constants of the higgs boson with other particles are well understood using the fact that the masses of quarks , leptons , and weak bosons originate in the vacuum expectation value ( vev ) of the higgs field , resulting in the prediction of the partial decay widths of the higgs boson into various particles . in models with physics beyond the sm ( bsm ) , measurements of the higgs couplings provide even exciting possibilities . in large class of bsm models , there exist new particles at the electroweak to tev scale , which affect the properties of the higgs boson . thus , with the detailed study of the higgs properties at collider experiments , we have a chance to observe a signal of bsm physics . such a study will be one of the major subjects in forthcoming collider experiments , i.e. , the lhc and future @xmath1 colliders like ilc and tlep @xcite . low energy supersymmetry ( susy ) is a well - motivated candidate of bsm physics . compared to the sm , the particle content is enlarged in susy models . even in the minimal setup , i.e. , in the minimal susy sm ( mssm ) , there exist two higgs doublets , @xmath8 and @xmath9 , as well as superparticles . the lightest higgs boson @xmath10 , which plays the role of the `` higgs boson '' discovered by atlas and cms , is a linear combination of the neutral components of @xmath8 and @xmath9 , while there exist other heavier higgses . in the case where the mass scales of the heavier higgses and the superparticles are high enough , the properties of @xmath10 are close to those of the sm higgs boson . on the contrary , if the heavier higgses or superparticles are relatively light , deviations of the higgs properties from the sm predictions may be observed by future collider experiments . with the precise measurement of the partial decay widths ( or branching ratios ) of the higgs boson , information about the heavy higgses and/or superparticles may be obtained even if those heavy particles can not be directly discovered . in this paper , we discuss how low the mass scales of the heavier higgs bosons and superparticles should be to observe a deviation . we evaluate the partial decay widths of the lightest higgs boson in the mssm , taking account of the following phenomenological constraints : higgs mass , flavor constraints of the @xmath11 mesons , stability of the electroweak ( sm - like ) vacuum against the transition to charge and color breaking ( ccb ) vacua , and perturbativity of coupling constants up to a high scale . these constraints exclude large fraction of the parameter region giving rise to a significant deviation . even so , we will see that the deviations of the partial widths from the sm predictions can be of @xmath12 for some of the decay modes in the parameter region allowed by the above - mentioned constraints . in particular , the deviations may be large enough to be observed by future future @xmath1 colliders like ilc and tlep even if superparticles are so heavy that they would not be observed at the lhc . the organization of this paper is as follows . in sec . [ sec : mssm ] , we briefly overview the properties of the higgs bosons in the mssm . we also summarize the phenomenological constraints that are taken into account in our analysis . then , in sec . [ sec : width ] , we calculate the partial decay widths of the lightest higgs boson in the mssm and discuss how large the deviation from the sm prediction can be . [ sec : discussion ] is devoted for conclusions and discussion . we review some of the important properties of the higgs sector in the mssm . there are two higgs doublets , @xmath8 and @xmath9 . as the neutral components acquire vevs , the electroweak symmetry breaking ( ewsb ) occurs . the ratio of the two higgs vevs is parameterized by @xmath13 . assuming no cp violation in the higgs potential , the mass eigenstates are classified as lighter and heavier cp - even higgs bosons ( denoted as @xmath10 and @xmath14 , respectively ) , cp - odd ( pseudo - scalar ) higgs @xmath15 , and charged higgs @xmath16 . in the following , we concentrate on the case where the masses of the heavier higgses ( @xmath14 , @xmath15 , and @xmath16 ) are much larger than the electroweak scale . then , the lightest higgs boson @xmath10 should be identified as the one observed by the lhc . on the other hand , the masses of the heavier higgses are almost degenerate . we parameterize the heavier higgs masses using the pseudo - scalar mass @xmath17 . at the tree level , the lightest higgs mass is predicted to be smaller than the @xmath18-boson mass , while it is significantly pushed up by radiative corrections @xcite . the mass matrix of the neutral cp - even higgs bosons is denoted as @xmath19 where @xmath20 represents radiative corrections . at the one - loop level , the top - stop contribution dominates the radiative correction to the lightest higgs mass , and is approximated as @xmath21,\end{aligned}\ ] ] where @xmath22 is the sm higgs vev , @xmath23 is the top - quark mass , @xmath24 ( with @xmath25 and @xmath26 being the lighter and heavier stop masses , respectively ) , and @xmath27 ( with @xmath28 and @xmath29 being the tri - linear scalar couplings for stop and the susy invariant higgsino mass parameter , respectively ) . in this paper , we adopt the convention of the slha format @xcite . the top - stop contribution can significantly enhance the lightest higgs mass . on the other hand , the bottom - sbottom contribution to the lightest higgs mass becomes sizable when the bottom yukawa coupling is large . it is likely to decrease the lightest higgs mass . when stop masses are @xmath30 , there are up to four solutions for @xmath28 to satisfy the observed value of the higgs mass @xmath31 , for which we use @xmath32 @xcite . let us call these four solutions as * ns : negative @xmath28 with smaller @xmath33 , * nl : negative @xmath28 with larger @xmath33 , * ps : positive @xmath28 with smaller @xmath33 , * pl : positive @xmath28 with larger @xmath33 . assuming universal sfermion masses at the susy scale , the value of @xmath33 is typically a few times larger than the stop mass for nl and pl cases . such a large value of @xmath33 has significant phenomenological implications , as we will discuss in the next section . since the one - loop correction to the higgs mass is comparable to the tree - level value , higher order corrections are necessary to obtain reliable results . in particular , qcd correction , which appears at the two - loop level , and a large hierarchy between the susy scale and the electroweak scale require the resummation of the leading and sub - leading logarithms . we use feynhiggs 2.10.2 @xcite for the precise evaluation of the higgs masses ( as well as the mixing parameters and the partial decay widths of @xmath10 ) . at the tree level , @xmath8 ( @xmath9 ) couples only to up - type quarks ( down - type quarks as well as leptons ) . however , this is not the case once radiative corrections due to superparticles are taken into account . the higgs couplings to bottom quark and tau lepton can be subject to sizable corrections even when susy breaking scale is very large . let us parameterize the effective @xmath34 and @xmath35 vertices including radiative corrections as @xcite @xmath36 where @xmath37 , @xmath38 , and @xmath39 are right - handed bottom , right - handed top , and third - generation quark - doublets , respectively . in addition , @xmath40 , @xmath41 and @xmath42 are @xmath43 indices , while the color indices are omitted for simplicity . here , @xmath44 and @xmath45 are non - holomorphic radiative corrections to the yukawa coupling constants . for a detailed treatment of the non - holomorphic corrections , see refs . @xcite . after the electroweak symmetry breaking , the yukawa couplings are related to the quark masses as these relations hold at the susy breaking scale . thus , the quark masses and yukawa couplings in the formula should be understood as the running parameters at the scale . @xmath46 where , at the leading order in the mass - insertion approximation , @xmath47 is given by @xmath48 \tan\beta , \\ \delta_t & \simeq \left [ \frac{2\alpha_s}{3\pi } m_3 \mu\ , i(m_{\tilde t_1}^2,m_{\tilde t_2}^2,m_3 ^ 2 ) + \frac{y_b^2}{16\pi^2 } \mu a_b\ , i(m_{\tilde b_1}^2,m_{\tilde b_2}^2,\mu^2 ) \right ] \cot\beta,\end{aligned}\ ] ] with @xmath49 being the gluino mass . the loop integral is defined as @xmath50 notice that @xmath51 is enhanced when @xmath52 is large , while @xmath53 is suppressed by @xmath54 . the mass eigenstates of the higgs bosons are given by linear combinations of @xmath8 and @xmath9 . the cp - even parts of their neutral components are related to the mass eigenstates as @xmath55 the mixing angle @xmath56 depends on the pseudo - scalar mass @xmath17 , and shows a decoupling behaviour , i.e. , @xmath57 as @xmath58 . in this limit , @xmath10 behaves as the sm higgs boson . using eq . , we obtain @xcite @xmath59 in fig . [ fig : cosba ] , we show the behavior of @xmath60 as a function of @xmath17 with the masses of superparticles being fixed . here , all the sfermion mass parameters , @xmath61 , @xmath62 , @xmath63 , @xmath64 , and @xmath65 , are taken to be universal at the susy scale @xmath66 , where these mass parameters are soft susy breaking masses of sfermions with gauge quantum numbers of @xmath67 , @xmath68 , @xmath69 , @xmath70 , and @xmath71 , respectively , with the numbers in the parenthesis being quantum numbers for @xmath72 , @xmath43 , and @xmath73 . throughout our study , we take @xmath74 . in addition , the sfermion masses are assumed to be universal in generation indices . we can see that radiative corrections can enhance @xmath60 by an order of magnitude when @xmath28 is large , while it is comparable to the tree - level value with the smaller @xmath33 solutions . is plotted as a function of @xmath17 with @xmath75 , @xmath76 , and the approximate gut relation among gaugino masses ( see eq . ) . all the sfermion masses are assumed to be universal at @xmath66 , and are also taken to be @xmath77 . the red and blue lines correspond to the cases of the pl and ps solutions of @xmath28 , respectively . the black dashed line is the tree - level value . ] denoting @xmath78 coupling ( with @xmath79 being the sm fermions ) as @xmath80 we obtain the @xmath34 coupling constant as @xmath81 g_{h\bar{b}b}^{\rm ( sm ) } , \label{g_hbb}\end{aligned}\ ] ] where the superscript `` ( sm ) '' is used for the sm prediction . when @xmath17 is relatively large , @xmath82 is almost unity , while the second term proportional to @xmath60 induces a sizable deviation from the sm value . similar relation holds for @xmath83 vertex , and @xmath84 is approximately given by @xmath85 with @xmath86 being the wino mass . quantitatively , @xmath84 is smaller than @xmath51 in the parameter space of our study . as we will see later , @xmath87 and @xmath88 may show sizable deviations from the sm predictions even if @xmath17 is above tev . the @xmath35 coupling constant is obtained as @xmath89 g_{h\bar{t}t}^{\rm ( sm)}. \label{g_htt}\end{aligned}\ ] ] the deviation from the sm value mainly comes from the second term in the bracket and is not enhanced by @xmath52 , since @xmath53 is proportional to @xmath54 . the gauge - boson final states are also important . in order to calculate the partial decay widths of the higgs to gauge bosons , we have modified feynhiggs 2.10.2 package to properly take account of the effect of non - holomorphic correction to the higgs interaction . we use eq . for the @xmath34 coupling to calculate the partial decay widths of @xmath90 , @xmath91 and @xmath92 . we found that the factor of @xmath93 was missing in feynhiggs 2.10.2 package ( see eq . ) . we have also modified the package to take the @xmath94 approximation @xcite for calculating the partial widths , in which the renormalization scale of the higgs wave functions is set to be @xmath95 , and the effects of radiative corrections are included in the mixing between the light and heavy higgs bosons . then , as we will see below , the partial decay widths show proper decoupling behavior in the large @xmath17 limit . in particular , the partial width @xmath96 converges to the sm value in the limit of large @xmath17 and squark masses . this behavior looks inconsistent with the result shown in ref . the processes induced by triangle loops , @xmath97 and @xmath98 , have been important for the higgs discovery at the lhc . they are also important in studying new particles that couple to the higgs boson . in susy models , the stop and sbottom loops contribute to the @xmath99 coupling . it is expressed by an approximate formula ( cf . @xcite ) @xmath100 up to @xmath101-term and bottom - loop contributions . here , @xmath102 . in the right - hand side , the first term comes from the top loop , while the second term is given by the stop and sbottom loops . the correction is positive in the non - mixing limit ( i.e. , @xmath103 ) , whereas it becomes negative when the mixing terms are sizable . we also note here that @xmath104 couplings ( with @xmath105 and @xmath106 ) are approximately given by @xmath107 . this ratio is very close to unity , and hence the deviations in these modes are very small . before discussing the possibility of observing a deviation of the higgs partial widths from the sm prediction at future colliders , we summarize phenomenological constraints on the mssm parameter space which are adopted in our analysis . in the sm , the flavor - changing decay , @xmath108 , proceeds by virtual exchanges of the @xmath18 and @xmath109 bosons . they are suppressed by the final - state helicity . in contrast , susy contributions can be enhanced considerably by large @xmath52 , when virtual exchanges of the heavy higgs boson contribute to the decay @xcite . the branching ratio is expressed as @xcite @xmath110,\end{aligned}\ ] ] where @xmath111 is the fermi constant , @xmath112 is the @xmath113-meson mass , @xmath114 is the decay constant of @xmath113 , @xmath115 is the muon mass , @xmath116 is the lifetime of @xmath113 , and @xmath117 is cabbino - kobayashi - maskawa matrix element . in the above expression , @xmath118 , @xmath119 and @xmath120 are the wilson coefficients of the effective operators , @xmath121 , @xmath122 , and @xmath123 , respectively , while @xmath124 are obtained by flipping chiralities , @xmath125 . among them , the sm contribution appears only in @xmath118 . including higher order contributions and taking account of effects of the @xmath126 oscillation , the sm prediction becomes @xcite @xmath127 this can be compared with the lhcb measurements @xcite @xmath128 defining @xmath129 , where the first term in the right - hand side includes both the susy and sm contributions , the 95% c.l . bound is estimated as @xmath130 we will adopt this constraint in our numerical analysis . even when superparticles are heavy , they affect the branching ratio through non - holomorphic contributions to the heavy higgs couplings . including radiative corrections and diagonalizing the quark mass matrices , effective couplings of the heavy higgs bosons to the down - type fermions become @xcite @xmath131 where @xmath132 . the flavor - changing coupling is induced by @xmath133 as @xmath134 here , soft scalar masses are assumed to be universal in generation , and @xmath135 is obtained by substituting @xmath136 in eq . . then , the wilson coefficients receive higgs - mediated contributions , @xmath137 where @xmath138 is the bottom - quark mass , @xmath139 is the weinberg angle , and @xmath140 . the non - holomorphic correction @xmath133 as well as @xmath51 does not decouple even for very heavy superparticles . we will see later that the corrections to @xmath141 can be sizable and that the constraint excludes some part of the parameter space of our interest . the branching ratio of the inclusive decay of @xmath142 may also be sensitive to the non - decoupling contributions to the ( charged ) higgs boson . in the numerical analysis , defining @xmath143 , we adopt the 95% c.l . bound , @xmath144 where the experimental value @xmath145 @xcite and the sm prediction @xmath146 @xcite are combined . at the current accuracies , @xmath147 imposes more stringent bound on the parameter space than @xmath142 except when superparticles are light . in the numerical analysis , superiso 3.4 @xcite is used for evaluating the susy contributions to the branching ratios as well as the sm predictions . in our analysis , we assume that the squark masses are universal in generations . if the squark masses are non - universal , there are extra contributions to @xmath133 , and the flavor constraints are affected ( see e.g. , ref . @xcite ) . such a non - universality is expected even in the model with the universal scalar masses at the gut scale . thus , it should be noted that the flavor constraints that we will show below are just for a particular choice of the squark - mass parameters and may change if the squark mass matrices have non - universal structures . with sufficiently large @xmath33 , ccb vacua arise , and the minimum of the scalar potential with the correct ewsb becomes a false vacuum @xcite . when @xmath148 , stop and the up - type higgs fields acquire large vevs at the ccb vacua , while the vevs of other fields are relatively small . recently , the decay rate of the sm - like vacuum has been studied in detail for such a case @xcite . on the other hand , if @xmath29 is as large as the stop masses , the down - type higgs boson also has a large vev at the ccb vacua due to the tri - linear scalar coupling among stops and the down - type higgs , which is proportional to @xmath149 . the vacuum stability condition is important in such a case because significant deviations of the higgs partial widths from the sm prediction may occur . in order to study the sm - like vacuum stability , we consider the tree - level scalar potential in the field space involving @xmath150 , @xmath151 , @xmath152 and @xmath153 ( which are canonically normalized scalar fields embedded in the left - handed stop , right - handed stop , the up - type higgs and the down - type higgs , respectively ) . the relevant part of the potential is given by @xmath154 where @xmath155 when the ccb vacua become deeper than the sm - like vacuum , the latter is not stable . the vacuum decay rate per unit volume is calculated in the semi - classical approximation , and is expressed as @xmath156 where @xmath157 is the euclidean action of the so - called bounce solution @xcite . for the decay of the sm - like vacuum , @xmath158 is estimated to be @xmath159 . in order for the lifetime of the sm - like vacuum to be longer than the present age of the universe , the euclidean action is required to satisfy @xmath160 in our numerical analysis , cosmotransition 2.0a1 @xcite is used to find the bounce solution in the four - dimensional field space parameterized by @xmath153 , @xmath152 , @xmath161 and @xmath162 , and to calculate the euclidean bounce action @xmath157 . the calculation is done at zero temperature , and therefore thermal effects are not taken into account . the model parameters such as the tri - linear and top yukawa couplings are evaluated at the susy scale @xmath66 . in fig . [ fig : se ] we show the contours of constant @xmath157 for the ps and pl solutions on @xmath17 vs. @xmath29 plane . here , @xmath163 , while sfermion masses , @xmath63 , @xmath64 and @xmath65 , are taken to be @xmath164 ( see sec . [ sec : width ] ) . in each plot , the region between the solid ( dashed ) green lines satisfies @xmath165 ( @xmath166 ) . for the ps solution , these lines appear only for @xmath167 . on the other hand , the upper bound on @xmath168 is comparable to the stop masses for the pl solution . and @xmath169 on @xmath17 vs. @xmath29 plane for the ps and pl solutions of @xmath28 . here , @xmath163 , @xmath170 , and @xmath171.,title="fig : " ] and @xmath169 on @xmath17 vs. @xmath29 plane for the ps and pl solutions of @xmath28 . here , @xmath163 , @xmath170 , and @xmath171.,title="fig : " ] + the tree - level potential eq . is used for our numerical analysis . radiative corrections may change the scalar potential particularly around the sm - like vacuum . in order to study their effects , we also estimated the decay rate of the sm - like vacuum by including the top - stop and bottom - sbottom one - loop correction to the higgs potential . a better treatment would be to introduce full one - loop radiative corrections to the potential involving stop , sbottom and the higgs boson so that the potential is stable against the renormalization scale at least at the one - loop level . we found that the euclidean action tends to increase by about 10 20% for @xmath172 from that of the tree - level potential . in order to see the sensitivity of @xmath157 to @xmath29 , we show the contours of @xmath173 in the same plot . when both @xmath29 and @xmath52 are large , we might have to consider other ccb vacua in the sbottom - higgs direction , which is driven by the tri - linear coupling of @xmath174 ( where @xmath175 and @xmath176 are left- and right - handed sbottoms , respectively ) ( cf . if both @xmath29 and @xmath177 are large , the bottom yukawa coupling can be enhanced not only by @xmath52 but also by @xmath178 ( see eq . ) . however , when the squark masses are universal , such a parameter region is already excluded by the other constraints discussed in this section . therefore we do not consider the constraint coming from the ccb vacuum involving the sbottom sector in this paper . when @xmath52 is very large , the bottom yukawa coupling is sizable . we can impose an upper bound on @xmath52 by requiring perturbativity of the yukawa coupling constants up to , for instance , the gut scale . in the mssm , the bottom yukawa coupling constant @xmath179 is proportional to @xmath178 , and hence , @xmath179 is enhanced when @xmath51 is negative ( see eq . ) . consequently , the bound on @xmath52 is more severe when @xmath180 . in our numerical analysis , we estimate the bottom yukawa coupling constant using the following relation : in discussing the perturbative bound , we neglect holomorphic corrections to the yukawa coupling constant , which are orders - of - magnitude smaller than the tree - level value of the yukawa coupling constant for @xmath181 . @xmath182 then , we follow the evolutions of coupling constants by solving the renormalization group equations at the one - loop level and impose a condition that the bottom yukawa coupling is perturbative up to the grand unified theory ( gut ) scale @xmath183 . numerically , we require @xmath184 where we take @xmath185 . this constraint excludes a large @xmath52 region especially when @xmath186 . notice that @xmath51 is a non - decoupling parameter , and hence , this constraint is important even in the limit of heavy superparticles . in this section , we discuss the partial decay width of the lightest higgs boson @xmath10 . we define the ratio of the partial decay width of the lightest higgs boson to that of the sm prediction : @xmath187 where @xmath188 denotes a specific final state . in the following , we show how much @xmath189 can deviate from the sm prediction ( @xmath190 ) for various final states . as we have mentioned , feynhiggs 2.10.2 is used to calculate the partial decay widths of the higgs boson , in which full one - loop contributions are taken into account for the fermionic final states . in the package , a resummation of the @xmath51 corrections is also included in calculating the partial decay width for @xmath191 @xcite . in our numerical calculation , we adopt ( approximate ) gut relation among the @xmath72 , @xmath43 , and @xmath73 gaugino masses : @xmath192 all the phases in the mssm parameters are assumed to be negligible , and we adopt the convention of @xmath193 . for simplicity , we also assume that the sfermion masses are universal with respect to the generation indices . ( left ) and @xmath194 ( right ) are shown as functions of @xmath195 ; here all the fermion masses , @xmath49 , and @xmath168 are equal to @xmath195 , and @xmath196 . ( the sign of @xmath29 is taken to be negative , while wino and bino masses are given by using the approximate gut relation . ) in the left ( right ) plot , the black , red , and blue lines correspond to @xmath197 ( @xmath198 ) , @xmath199 ( @xmath200 ) , and @xmath97 , respectively . the solid ( dashed ) lines correspond to the pl ( ps ) solutions of @xmath28 . note that the constraints in sec . [ sec : constraint ] are not taken into account . , title="fig : " ] ( left ) and @xmath194 ( right ) are shown as functions of @xmath195 ; here all the fermion masses , @xmath49 , and @xmath168 are equal to @xmath195 , and @xmath196 . ( the sign of @xmath29 is taken to be negative , while wino and bino masses are given by using the approximate gut relation . ) in the left ( right ) plot , the black , red , and blue lines correspond to @xmath197 ( @xmath198 ) , @xmath199 ( @xmath200 ) , and @xmath97 , respectively . the solid ( dashed ) lines correspond to the pl ( ps ) solutions of @xmath28 . note that the constraints in sec . [ sec : constraint ] are not taken into account . , title="fig : " ] first , we show the soft mass dependence of @xmath201 without taking the phenomenological constraints into account . in fig . [ fig : msoft ] , @xmath201 are shown for @xmath202 , @xmath203 , @xmath92 , @xmath198 , and @xmath200 as functions of @xmath195 , taking @xmath204 and @xmath196 . we have checked that @xmath205 holds in the parameter region of our study . by choosing negative @xmath29 with @xmath206 , the correction to the bottom yukawa coupling becomes significant . remarkably , even though @xmath195 is around several tev , the partial decay widths for @xmath207 , @xmath200 and @xmath92 deviate from the sm prediction by more than @xmath208% for the pl solution . however , susy contributions to the other widths are smaller . note that @xmath209 is as about twice as @xmath210 for the pl solution , which is due to the difference between @xmath51 and @xmath211 . we also note here that the partial decay widths have an appropriate decoupling behaviour , i.e. , @xmath212 as the masses of superparticles and the heavy higgs bosons become infinitely large . are shown for the ps , ns , pl , and nl solutions of @xmath28 . here , all the sfermion masses , @xmath49 , and @xmath168 are taken to be @xmath77 , and the sign of @xmath29 is set to be negative . the left regions of the blue lines are excluded by @xmath213 , while those of the green solid ( dashed ) lines are constrained by the vacuum stability condition , @xmath214 ( 360 ) . the bottom yukawa coupling becomes non - perturbative below the gut scale in the region above the red line.,title="fig : " ] are shown for the ps , ns , pl , and nl solutions of @xmath28 . here , all the sfermion masses , @xmath49 , and @xmath168 are taken to be @xmath77 , and the sign of @xmath29 is set to be negative . the left regions of the blue lines are excluded by @xmath213 , while those of the green solid ( dashed ) lines are constrained by the vacuum stability condition , @xmath214 ( 360 ) . the bottom yukawa coupling becomes non - perturbative below the gut scale in the region above the red line.,title="fig : " ] + are shown for the ps , ns , pl , and nl solutions of @xmath28 . here , all the sfermion masses , @xmath49 , and @xmath168 are taken to be @xmath77 , and the sign of @xmath29 is set to be negative . the left regions of the blue lines are excluded by @xmath213 , while those of the green solid ( dashed ) lines are constrained by the vacuum stability condition , @xmath214 ( 360 ) . the bottom yukawa coupling becomes non - perturbative below the gut scale in the region above the red line.,title="fig : " ] are shown for the ps , ns , pl , and nl solutions of @xmath28 . here , all the sfermion masses , @xmath49 , and @xmath168 are taken to be @xmath77 , and the sign of @xmath29 is set to be negative . the left regions of the blue lines are excluded by @xmath213 , while those of the green solid ( dashed ) lines are constrained by the vacuum stability condition , @xmath214 ( 360 ) . the bottom yukawa coupling becomes non - perturbative below the gut scale in the region above the red line.,title="fig : " ] next , let us include the phenomenological constraints discussed in sec . [ sec : constraint ] . in fig . [ fig : matanb ] , contours of @xmath209 are shown on @xmath17 vs. @xmath52 plane . here , all the sfermion masses , @xmath49 , and @xmath168 are set to be @xmath77 . in the cases of the ps and pl solutions , the contours end at around @xmath215 . this is because , when @xmath52 is large , there is no solution for @xmath28 to satisfy @xmath32 . the bottom - sbottom loop contribution interferes destructively with the top - stop contribution , and thus , larger ( smaller ) @xmath28 is required for the ps ( pl ) solution . then , @xmath216 and @xmath217 merge into a single solution at certain @xmath52 , which is the value where the top - stop contributions can not raise the higgs mass anymore for fixed @xmath195 . note that , @xmath51 is negative when @xmath218 and @xmath219 . in such a case , @xmath179 is significantly larger than the tree - level value , and thus , the bottom - sbottom loop contributions are enhanced . in fig . [ fig : matanb ] , the constraints from @xmath220 , the vacuum stability , and the perturbativity of the bottom yukawa couplings are shown . wide parameter region is excluded when @xmath28 is large , i.e. , in the pl and nl panels . the constraints from @xmath141 and the vacuum stability become weaker for large @xmath17 , while that from the perturbativity is not . for the pl solution , the constraints from @xmath141 and the vacuum stability change drastically as we vary @xmath52 for @xmath221 , where the bottom yukawa coupling is much larger than the tree - level value . .expected accuracies of the determinations of the partial decay widths of the higgs boson in units of percents @xcite . the accuracies of the widths are assumed to be twice the accuracies of the determinations of couplings . [ cols="<,^,^,^,^,^,^,^,^ " , ] even if the masses of superparticles are relatively large ( i.e. , @xmath77 ) , @xmath209 can be as large as @xmath208% . such a large deviation may be within the reach of future collider experiments . expected accuracies at the future experiments have been discussed ( see table 1 - 16 of ref . @xcite and ref . @xcite ) ; the numbers are summarized in table [ table : experiments ] . the accuracies of @xmath222 , @xmath223 , and @xmath224 are claimed to be achievable at @xmath1 colliders ultimately . therefore , it is found that , even if the superparticles are kinematically unaccessible at the lhc , we may observe the mssm signal by studying the partial decay widths of @xmath10 in detail . for the pl solution , it is also found that the partial width of @xmath191 can deviate from the sm prediction by about 2% even for @xmath225 . on the other hand , we checked that @xmath226 is smaller by about 1% than @xmath227 at the same parameter point , because the non - holomorphic correction @xmath228 is bigger than @xmath229 at this model point . when @xmath52 is smaller , the difference between @xmath227 and @xmath226 decreases , since @xmath51 is approximately proportional to @xmath52 ; in such a case both @xmath227 and @xmath226 are well approximated by @xmath230 . ( blue ) and @xmath226 ( red ) with small @xmath33 solutions ( i.e , the ps and ns solutions for @xmath28 ) . we take @xmath231 , 3 , 4 and @xmath77 . other sfermion masses are set to be @xmath232 . the partial widths are constrained by @xmath233 , the vacuum stability condition ( @xmath214 ) , and the perturbativity of the bottom yukawa coupling ( @xmath234 ) . the black region corresponds to the tree - level prediction for @xmath235@xmath236 . the black dashed line is the sm prediction . , title="fig : " ] ( blue ) and @xmath226 ( red ) with small @xmath33 solutions ( i.e , the ps and ns solutions for @xmath28 ) . we take @xmath231 , 3 , 4 and @xmath77 . other sfermion masses are set to be @xmath232 . the partial widths are constrained by @xmath233 , the vacuum stability condition ( @xmath214 ) , and the perturbativity of the bottom yukawa coupling ( @xmath234 ) . the black region corresponds to the tree - level prediction for @xmath235@xmath236 . the black dashed line is the sm prediction . , title="fig : " ] + ( blue ) and @xmath226 ( red ) with small @xmath33 solutions ( i.e , the ps and ns solutions for @xmath28 ) . we take @xmath231 , 3 , 4 and @xmath77 . other sfermion masses are set to be @xmath232 . the partial widths are constrained by @xmath233 , the vacuum stability condition ( @xmath214 ) , and the perturbativity of the bottom yukawa coupling ( @xmath234 ) . the black region corresponds to the tree - level prediction for @xmath235@xmath236 . the black dashed line is the sm prediction . , title="fig : " ] ( blue ) and @xmath226 ( red ) with small @xmath33 solutions ( i.e , the ps and ns solutions for @xmath28 ) . we take @xmath231 , 3 , 4 and @xmath77 . other sfermion masses are set to be @xmath232 . the partial widths are constrained by @xmath233 , the vacuum stability condition ( @xmath214 ) , and the perturbativity of the bottom yukawa coupling ( @xmath234 ) . the black region corresponds to the tree - level prediction for @xmath235@xmath236 . the black dashed line is the sm prediction . , title="fig : " ] , but with all the solutions of @xmath28.,title="fig : " ] , but with all the solutions of @xmath28.,title="fig : " ] + , but with all the solutions of @xmath28.,title="fig : " ] , but with all the solutions of @xmath28.,title="fig : " ] let us see how much @xmath227 and @xmath226 can change in the parameter space consistent with the phenomenological constraints . we have performed the scan in the following parameter space of the mssm : * @xmath231 , @xmath237 , @xmath238 , and @xmath77 , * @xmath170 , * @xmath239 , @xmath240 , @xmath216 , @xmath217 , * @xmath241 , * @xmath242 , or @xmath243 , * @xmath244 . here , @xmath28 is determined to satisfy @xmath245 . the other tri - linear couplings ( e.g. , @xmath177 and @xmath246 ) are assumed to be equal to @xmath28 ; we checked that our numerical results are insensitive to this assumption . we take @xmath247 , while sfermion masses other than @xmath61 and @xmath62 are set to be equal to @xmath248 . we have checked that @xmath227 and @xmath226 are almost insensitive to these scalar masses unless they are very small . however , when @xmath168 is much larger than @xmath63 , @xmath64 , and @xmath65 , bottom and stau mixings become sizable which causes additional complexity to our parameter scan , and/or lighter stau become tachyonic . the values of @xmath63 , @xmath64 , and @xmath65 are chosen to avoid this problem . if @xmath168 would become much larger than squark masses ( e.g. , @xmath249 ) , the higgs partial widths could deviate from the sm prediction sizably , since the higgs mixing angle would be enhanced by radiative corrections . however , the vacuum stability condition excludes significant amount of the parameter region with large @xmath168 , as we have shown in fig . [ fig : se ] . in our scan , we set @xmath250 ; this upper bound is large enough to cover the whole region allowed by the vacuum stability . in figs . [ fig : rbbtaus ] and [ fig : rbbtau ] , we show the minimal and maximal values of @xmath227 and @xmath226 as functions of @xmath17 and @xmath195 , taking into account the phenomenological constraints discussed in the previous section . fig . [ fig : rbbtaus ] corresponds to the small @xmath28 solutions ( ns and ps ) . the largest and smallest values of @xmath227 and @xmath226 are achieved when @xmath168 and @xmath52 are large and are marginally allowed by the phenomenological constraints . the deviations mainly come from radiative corrections to the higgs mixing angle and @xmath51 . in particular , the maximal value of @xmath227 is larger than that of @xmath226 for large @xmath251 . as discussed earlier , @xmath51 becomes negative and sizable with @xmath252 , which results in a significant enhancement of @xmath227 . [ fig : rbbtau ] shows the minimal and maximal values of @xmath227 and @xmath226 , taking all the solutions of ns , ps , nl , and pl into consideration . they change significantly compared to fig . [ fig : rbbtaus ] . the partial widths become extremum when @xmath28 takes the nl or pl solution except the maximal value for small @xmath17 . for @xmath253 , the ns or ps solutions give the maximal value , since the phenomenological constraints , especially the vacuum stability condition , are too severe for the nl and pl solutions ( see fig . [ fig : matanb ] ) . as in the case of fig . [ fig : rbbtaus ] , the maximal value of @xmath227 is much larger than that of @xmath226 because @xmath254 . the maximal value of @xmath227 has a non - trivial bump - like structure . when @xmath17 is small , medium ( i.e. , around the peak of the bump ) , and large , the partial decay width is bounded by the vacuum stability , flavor constraint , and the perturbativity of @xmath179 , respectively . the measurement of the higgs couplings may provide an evidence of bsm , because the values of @xmath227 and @xmath226 may show significant deviation from the sm prediction . in future experiments , the partial decay widths of @xmath255 and @xmath256 may be measured at the @xmath257 level ( see table [ table : experiments ] ) . in the case of the ps and ns solutions , @xmath209 can be as large as 2% ( 3% ) if @xmath17 is @xmath258 ( @xmath259 ) for @xmath260 , and @xmath261 ( @xmath262 ) for @xmath263 . on the other hand , @xmath210 becomes larger than 2% ( 3% ) when @xmath17 is smaller than @xmath264 ( @xmath265 ) for @xmath260 , and @xmath262 ( @xmath266 ) for @xmath263 . including the pl and nl solutions , deviations of @xmath267 @xmath237% level are achieved with larger value of @xmath17 . for example , with @xmath263 , @xmath268 can be achieved with @xmath269 . ( left ) and @xmath270 ( right ) are shown as functions of @xmath62 . the parameters are scanned in the same manner as fig . [ fig : rbbtau ] , but @xmath17 is fixed to be @xmath77 while @xmath62 is varied . here , all the solutions of @xmath28 are included in the scan with imposing the phenomenological constraints discussed in sec . [ sec : constraint ] . the black dashed line is the sm prediction.,title="fig : " ] ( left ) and @xmath270 ( right ) are shown as functions of @xmath62 . the parameters are scanned in the same manner as fig . [ fig : rbbtau ] , but @xmath17 is fixed to be @xmath77 while @xmath62 is varied . here , all the solutions of @xmath28 are included in the scan with imposing the phenomenological constraints discussed in sec . [ sec : constraint ] . the black dashed line is the sm prediction.,title="fig : " ] in fig . [ fig : msoft2 ] , we show the sfermion mass dependence of @xmath201 with taking the phenomenological constraints into account . the parameter scan is performed in the same setup as fig . [ fig : rbbtau ] , but @xmath62 is varied . here , @xmath17 is fixed to be @xmath77 . in the left plot , the minimal and maximal values of @xmath227 and @xmath226 are shown . they are achieved by the pl or nl solution . since @xmath17 is fixed , the decoupling behaviour is not observed . rather , the non - decoupling contribution to @xmath51 enhances @xmath227 as @xmath62 increases , since larger @xmath33 is required to satisfy @xmath245 for the solutions . on the other hand , it is found that radiative corrections to the higgs mixing angle do not change so much even if @xmath62 increases . in the right panel of fig . [ fig : msoft2 ] , the minimal and maximal values of @xmath270 are displayed . the maximal value occurs for the ps or ns solution , while the minimal value is achieved by the pl or nl solution . we can see the decoupling behavior , i.e. , @xmath270 approaches to unity as @xmath62 increases . the partial decay width of @xmath97 can deviate from the sm prediction by about 3% for @xmath271 , while it decreases rapidly and becomes 1% for @xmath272 . we have also checked that these values do not change so much for @xmath273 . ( however , they change significantly if @xmath17 is smaller , since the phenomenological constraints exclude the parameter space severely . ) according to table [ table : experiments ] , the partial width of @xmath97 is expected to be measured at the @xmath274 accuracy in future experiments . thus , as far as superparticles are relatively light , we may observe a signal of the mssm in the measurements of this partial width even if the heavier higgses are out of the reach of the lhc . as a function of @xmath17 ( green histogram ) , as well as that of model points allowed by the phenomenological constraints ( dotted histogram ) . upper - right : the number of model points allowed by the phenomenological constraints on @xmath17 vs. @xmath52 plane . lower - left : the number of model points accessible with ilc by @xmath275 . lower - right : the number of model points accessible with ilc by @xmath276 . , title="fig : " ] as a function of @xmath17 ( green histogram ) , as well as that of model points allowed by the phenomenological constraints ( dotted histogram ) . upper - right : the number of model points allowed by the phenomenological constraints on @xmath17 vs. @xmath52 plane . lower - left : the number of model points accessible with ilc by @xmath275 . lower - right : the number of model points accessible with ilc by @xmath276 . , title="fig : " ] + as a function of @xmath17 ( green histogram ) , as well as that of model points allowed by the phenomenological constraints ( dotted histogram ) . upper - right : the number of model points allowed by the phenomenological constraints on @xmath17 vs. @xmath52 plane . lower - left : the number of model points accessible with ilc by @xmath275 . lower - right : the number of model points accessible with ilc by @xmath276 . , title="fig : " ] as a function of @xmath17 ( green histogram ) , as well as that of model points allowed by the phenomenological constraints ( dotted histogram ) . upper - right : the number of model points allowed by the phenomenological constraints on @xmath17 vs. @xmath52 plane . lower - left : the number of model points accessible with ilc by @xmath275 . lower - right : the number of model points accessible with ilc by @xmath276 . , title="fig : " ] finally , we show how large fraction of the parameter space can be covered by future @xmath1 colliders . for this purpose , we define the @xmath277 variable as @xmath278 ^ 2 } { [ \delta\gamma(h\rightarrow f)]^2},\end{aligned}\ ] ] where @xmath279 is the expected accuracies of the determinations of the higgs partial decay widths at ilc with @xmath280 and @xmath281 ( see table [ table : experiments ] ) . based on this quantity , we define the parameter region which is accessible with ilc at @xmath282 . we perform a parameter scan and study if each model point is accessible with ilc and satisfies the phenomenological constraints . the mssm parameters are scanned in the ranges of @xmath283 ( with the step size of @xmath284 ) , @xmath285 ( with the step size of @xmath286 ) , and @xmath287 ( with the step size of @xmath288 ) . thus , for each set of @xmath289 , 384 model points are studied , taking account of positive and negative values of @xmath29 as well as all four solutions of @xmath28 . in addition , we adopt the relations @xmath290 , and @xmath291 . concentrating on the parameter space where the lhc will have a difficulty in finding superparticles , @xmath62 is taken to be @xmath237 , @xmath238 , and @xmath77 . in fig . [ fig : scan ] , the ilc coverage of the parameter space of our scan is displayed . in the upper - left panel , we show the distribution of the number of model points accessible with ilc by any of the decay modes of @xmath10 as a function of @xmath17 . the green histogram is a distribution of the number of model points which satisfy @xmath292 and the phenomenological constraints , while the dotted one is that satisfying the phenomenological constraints without imposing @xmath293 . we find that the number reduces drastically at @xmath294 . in the upper - right panel , we show the number of model points which survive the phenomenological constraints on @xmath17 vs. @xmath52 plane . here , @xmath282 is not imposed . then , in the lower panels , we show the numbers of model points which can be accessed by the decay modes of @xmath275 ( lower - left ) and @xmath276 ( lower - right ) with taking account of the phenomenological constraints . they reduce significantly for @xmath295 irrespective of @xmath52 . the accessible points for @xmath295 are mostly with pl or nl solution . in this paper , we have studied the partial decay widths of the lightest higgs boson in the mssm . taking account of relevant phenomenological constraints , i.e. , higgs mass , flavor constraints , vacuum stability , and the perturbativity of coupling constants up to the gut scale , we have calculated the expected deviations of the partial decay widths from the sm predictions . the partial decay widths are enhanced if the @xmath29-parameter is relatively large . however , such a choice may conflict with some of the phenomenological constraints . in particular , the vacuum - stability condition imposes a stringent constraint on the parameter space . we have found that , with too large @xmath168 , there show up ccb vacua where the down - type higgs field as well as the up - type higgs and stop fields acquire large vevs ; existence of such ccb vacua was not seriously considered in the previous studies . in addition , when @xmath296 is large , non - holomorphic correction to the bottom yukawa interaction becomes so large that the bottom yukawa coupling constant becomes non - perturbative below the gut scale . large value of @xmath296 may also cause too large flavor - violating decay of @xmath11-mesons . by taking these constraints into account , the maximal and minimal possible values of the higgs partial widths are restricted . we found that the deviations of the partial decay widths from the sm predictions can be of @xmath208% for some of the decay modes . in particular , those of @xmath87 and @xmath88 may show significant deviations even if the superparticles are out of the reach of 14tev lhc . in addition , the deviation of @xmath96 may also be sizable if the superparticles are relatively light . we emphasize that , although our scan is limited to some part of the mssm parameter space , we have found the regions where the deviations from the sm predictions are within the reach of proposed @xmath1 colliders even if superparticles would not be observed at the lhc . _ acknowledgements _ : the authors would like to thank yasuhiro shimizu for the collaboration in the early stage of this project . the authors also acknowledge yitp for their hospitality , at which this work was initiated . this work is supported by jsps kakenhi no . 23740172 ( m.e . ) , no . 26400239 ( t.m . ) , and no . 26287039 ( m.m.n . ) . the work is supported by grant - in - aid for scientific research from the ministry of education , science , sports , and culture ( mext ) , japan , no . 23104008 ( t.m . ) and no . 23104006 ( m.m.n . ) , and also by world premier international research center initiative ( wpi initiative ) , mext , japan . g. aad _ et al . _ [ atlas collaboration ] , phys . b * 716 * ( 2012 ) 1 [ arxiv:1207.7214 [ hep - ex ] ] . s. chatrchyan _ et al . _ [ cms collaboration ] , phys . b * 716 * ( 2012 ) 30 [ arxiv:1207.7235 [ hep - ex ] ] . s. dawson , a. gritsan , h. logan , j. qian , c. tully , r. van kooten , a. ajaib and a. anastassov _ et al . _ , arxiv:1310.8361 [ hep - ex ] . y. okada , m. yamaguchi and t. yanagida , prog . theor . phys . * 85 * ( 1991 ) 1 . y. okada , m. yamaguchi and t. yanagida , phys . b * 262 * , 54 ( 1991 ) . j. r. ellis , g. ridolfi and f. zwirner , phys . b * 257 * ( 1991 ) 83 . j. r. ellis , g. ridolfi and f. zwirner , phys . b * 262 * , 477 ( 1991 ) . h. e. haber and r. hempfling , phys . * 66 * ( 1991 ) 1815 . z. skands , b. c. allanach , h. baer , c. balazs , g. belanger , f. boudjema , a. djouadi and r. godbole _ et al . _ , jhep * 0407 * ( 2004 ) 036 [ hep - ph/0311123 ] . k. a. olive et al . ( particle data group ) , chin . c , * 38 * , 090001 ( 2014 ) . s. heinemeyer , w. hollik and g. weiglein , comput . commun . * 124 * , 76 ( 2000 ) [ hep - ph/9812320 ] . s. heinemeyer , w. hollik and g. weiglein , eur . j. c * 9 * , 343 ( 1999 ) [ hep - ph/9812472 ] . g. degrassi , s. heinemeyer , w. hollik , p. slavich and g. weiglein , eur . j. c * 28 * , 133 ( 2003 ) [ hep - ph/0212020 ] . m. frank , t. hahn , s. heinemeyer , w. hollik , h. rzehak and g. weiglein , jhep * 0702 * , 047 ( 2007 ) [ hep - ph/0611326 ] . t. hahn , s. heinemeyer , w. hollik , h. rzehak and g. weiglein , phys . * 112 * ( 2014 ) 14 , 141801 [ arxiv:1312.4937 [ hep - ph ] ] . l. j. hall , r. rattazzi and u. sarid , phys . rev . d * 50 * , 7048 ( 1994 ) [ hep - ph/9306309 , hep - ph/9306309 ] . r. hempfling , phys . d * 49 * , 6168 ( 1994 ) . m. s. carena , m. olechowski , s. pokorski and c. e. m. wagner , nucl . b * 426 * , 269 ( 1994 ) [ hep - ph/9402253 ] . m. s. carena , s. mrenna and c. e. m. wagner , phys . d * 60 * , 075010 ( 1999 ) [ hep - ph/9808312 ] . h. eberl , k. hidaka , s. kraml , w. majerotto and y. yamada , phys . rev . d * 62 * , 055006 ( 2000 ) [ hep - ph/9912463 ] . m. s. carena , d. garcia , u. nierste and c. e. m. wagner , nucl . b * 577 * , 88 ( 2000 ) [ hep - ph/9912516 ] . l. hofer , u. nierste and d. scherer , jhep * 0910 * , 081 ( 2009 ) [ arxiv:0907.5408 [ hep - ph ] ] . d. noth and m. spira , phys . rev . * 101 * , 181801 ( 2008 ) [ arxiv:0808.0087 [ hep - ph ] ] . m. s. carena , h. e. haber , h. e. logan and s. mrenna , phys . d * 65 * ( 2002 ) 055005 [ erratum - ibid . d * 65 * ( 2002 ) 099902 ] [ hep - ph/0106116 ] . a. arvanitaki and g. villadoro , jhep * 1202 * , 144 ( 2012 ) [ arxiv:1112.4835 [ hep - ph ] ] . m. reece , new j. phys . * 15 * , 043003 ( 2013 ) [ arxiv:1208.1765 [ hep - ph ] ] . e. bagnaschi , r. v. harlander , s. liebler , h. mantler , p. slavich and a. vicini , jhep * 1406 * , 167 ( 2014 ) [ arxiv:1404.0327 [ hep - ph ] ] . s. heinemeyer , w. hollik and g. weiglein , eur . j. c * 16 * ( 2000 ) 139 [ hep - ph/0003022 ] . m. cahill - rowley , j. hewett , a. ismail and t. rizzo , phys . d * 90 * ( 2014 ) 9 , 095017 [ arxiv:1407.7021 [ hep - ph ] ] . k. s. babu and c. f. kolda , phys . * 84 * , 228 ( 2000 ) [ hep - ph/9909476 ] . s. r. choudhury and n. gaur , phys . b * 451 * , 86 ( 1999 ) [ hep - ph/9810307 ] . g. buchalla and a. j. buras , nucl . b * 400 * , 225 ( 1993 ) . m. misiak and j. urban , phys . b * 451 * , 161 ( 1999 ) [ hep - ph/9901278 ] . c. bobeth , t. ewerth , f. kruger and j. urban , phys . d * 64 * , 074014 ( 2001 ) [ hep - ph/0104284 ] . c. bobeth , a. j. buras , f. kruger and j. urban , nucl . b * 630 * , 87 ( 2002 ) [ hep - ph/0112305 ] . c. bobeth , m. gorbahn , t. hermann , m. misiak , e. stamou and m. steinhauser , phys . * 112 * ( 2014 ) 101801 [ arxiv:1311.0903 [ hep - ph ] ] . v. khachatryan _ et al . _ [ cms and lhcb collaborations ] , arxiv:1411.4413 [ hep - ex ] . y. amhis _ et al . _ [ heavy flavor averaging group ( hfag ) collaboration ] , arxiv:1412.7515 [ hep - ex ] . m. misiak , h. m. asatrian , k. bieri , m. czakon , a. czarnecki , t. ewerth , a. ferroglia and p. gambino _ et al . _ , phys . lett . * 98 * , 022002 ( 2007 ) [ hep - ph/0609232 ] . f. mahmoudi , comput . commun . * 180 * ( 2009 ) 1579 [ arxiv:0808.3144 [ hep - ph ] ] . w. altmannshofer , m. carena , n. r. shah and f. yu , jhep * 1301 * ( 2013 ) 160 [ arxiv:1211.1976 [ hep - ph ] ] . j. m. frere , d. r. t. jones and s. raby , nucl . phys . b * 222 * , 11 ( 1983 ) . j. f. gunion , h. e. haber and m. sher , nucl . b * 306 * , 1 ( 1988 ) . j. a. casas , a. lleyda and c. munoz , nucl . phys . b * 471 * , 3 ( 1996 ) [ hep - ph/9507294 ] . a. kusenko , p. langacker and g. segre , phys . d * 54 * , 5824 ( 1996 ) [ hep - ph/9602414 ] . j. e. camargo - molina , b. oleary , w. porod and f. staub , jhep * 1312 * , 103 ( 2013 ) [ arxiv:1309.7212 [ hep - ph ] ] . d. chowdhury , r. m. godbole , k. a. mohan and s. k. vempati , jhep * 1402 * , 110 ( 2014 ) [ arxiv:1310.1932 [ hep - ph ] ] . n. blinov and d. e. morrissey , jhep * 1403 * , 106 ( 2014 ) [ arxiv:1310.4174 [ hep - ph ] ] . j. e. camargo - molina , b. garbrecht , b. oleary , w. porod and f. staub , phys . b * 737 * , 156 ( 2014 ) [ arxiv:1405.7376 [ hep - ph ] ] . s. r. coleman , phys . d * 15 * ( 1977 ) 2929 [ erratum - ibid . d * 16 * ( 1977 ) 1248 ] . c. g. callan , jr . and s. r. coleman , phys . d * 16 * ( 1977 ) 1762 . c. l. wainwright , comput . commun . * 183 * ( 2012 ) 2006 [ arxiv:1109.4189 [ hep - ph ] ] . k. e. williams , h. rzehak and g. weiglein , eur . j. c * 71 * , 1669 ( 2011 ) [ arxiv:1103.1335 [ hep - ph ] ] . m. e. peskin , arxiv:1312.4974 [ hep - ph ] .

we study the dynamics and thermalization of strongly correlated fermions in finite one - dimensional lattices after a quantum quench . our calculations are performed using exact diagonalization . we focus on one- and two - body observables such as the momentum distribution function [ and the density - density structure factor [ , respectively , and study the effects of approaching an integrable point . we show that while the relaxation dynamics and thermalization of for fermions is very similar to the one of hardcore bosons , the behavior of is distinctively different . the latter observable exhibits a slower relaxation dynamics in fermionic systems . we identify the origin of this behavior , which is related to the off - diagonal matrix elements of in the basis of the eigenstates of the hamiltonian . more generally , we find that thermalization occurs far away from integrability and that it breaks down as one approaches the integrable point .

0908.3188

in recent years , the study of the nonequilibrium dynamics of isolated quantum systems has attracted a great deal of attention . the main motivation for these studies lies behind the spectacular success that experimentalists have achieved in trapping and manipulating ultracold quantum gases . this has allowed them to , for example , load ultracold bosons in optical lattices and study the collapse and revival of the matter wave interference after quenching the interaction strength from deep in the superfluid regime into deep in the mott insulating regime @xcite , and to observe the damping of the dipole oscillations when the center of mass of the gas was displaced away from the center of the trap @xcite . since these atomic gases are trapped , cooled , and manipulated in a very high vacuum , i.e. , in no contact with any kind of thermal or particle reservoir , a fundamental question that arises is whether after a sudden perturbation the system will relax to thermal equilibrium ( an equilibrium in which observables are described by standard statistical mechanical ensembles ) . the usual assumption is that thermalization occurs in general . however , recent experiments in one - dimensional ( 1d ) geometries ( created by a deep two - dimensional optical lattice ) have failed to observe relaxation to a thermal distribution after a quench @xcite . the absence of thermalization can be understood in the very special regime in which the system is at an integrable point @xcite . interestingly , in ref . @xcite the gas was perturbed away from integrability and thermalization did not occur . other experiments in 1d geometries ( generated by an atom chip ) have reported the indirect observation of thermalization @xcite . in the latter case , the transverse confinement was not as strong as the one induced by the optical lattice in ref . . the question of whether strict one dimensionality could affect the outcome of the relaxation dynamics has not been fully addressed experimentally . the particular case in which the system is at an integrable point lends itself to combined analytical and numerical studies , which have allowed theorists to show that in that case the relaxation dynamics after a quench lead to nonthermal equilibrium distributions of few - body observables @xcite . in addition , a generalization of the gibbs ensemble , which takes into account the conserved quantities that make the system integrable , has been shown to successfully describe experimental observables after relaxation @xcite . further analytical and numerical works have analyzed the relevance of the generalized gibbs ensemble to different models and observables and addressed its limits of applicability @xcite . when the systems are far away from integrability , for example in higher dimensions , the expectation is that they should thermalize . recent numerical work has shown this to be the case for bosons in two dimensions , studied with exact diagonalization @xcite , and for fermions studied within the dynamical mean - field theory ( dmft ) approximation @xcite . in addition , thermalization in the isolated two - dimensional bosonic case could be understood within the eigenstate thermalization hypothesis ( eth ) proposed by deutsch @xcite and srednicki @xcite , in which the individual eigenstates of the many - body hamiltonian exhibit thermal behavior @xcite . in one - dimensional nonintegrable systems , the situation is a bit more subtle as many models can be tuned ( by changing some hamiltonian parameters ) to be arbitrarily close to integrable points . an early study of the relaxation dynamics in fermionic systems when breaking integrability , using time - dependent density renormalization group ( tdmrg ) @xcite , concluded that thermalization does not occur even if the system is perturbed away from the integrable point . another early work in 1d , also using tdmrg , studied the relaxation dynamics of the ( nonintegrable ) bose - hubbard model when quenching the system across the superfluid - to - mott - insulator transition @xcite . in that case the authors concluded that thermalization occurs in some regimes but not in others . more recently , we have performed a systematic study of thermalization in finite 1d systems of hardcore bosons . our results indicate that far away from integrability the system does thermalize @xcite . however , thermalization breaks down as one approaches the integrable point . an important question that still needs to be answered is what happens with the point at which thermalization breaks down as one approaches the thermodynamic limit . it may either move toward the integrable point or toward some point away from integrability . more powerful numerical techniques or analytical approaches may be required to answer that question , which will not be addressed here . in this work , we extend the analysis in ref . @xcite to the fermionic case , which allows us to discuss some of the issues that remained open in ref . in particular , the time scales required for different observables to relax to a stationary `` equilibrium '' distribution , and the description of the observables after relaxation . we also study how the closeness to an integrable point affects the dynamics and thermalization after a quantum quench . as in earlier works @xcite , we show that the breakdown of thermalization for our observables is directly linked to the breakdown of eth . the presentation is organized as follows . in sec . [ sec : hamiltf ] , we introduce the lattice model to describe 1d fermions with nearest- and next - nearest - neighbor hopping and repulsive interactions . we define our observables of interest and briefly discuss our numerical approach . the nonequilibrium dynamics of this model is presented in sec . [ sec : dynamicsf ] . section [ sec : thermodynamicsf ] is devoted to the analysis of statistical mechanical approaches to describe observables after relaxation . in sec . [ sec : ethf ] , we justify these statistical mechanical approaches , or their failure , within eth . an explanation of why the momentum distribution function of the fermions has a distinctively slow relaxation dynamics is presented in sec . [ sec : offdiagonal ] in terms of the off - diagonal elements of this observable in the basis of the eigenstates of the hamiltonian . finally , the summary and outlook are presented in sec . [ sec : summary ] . the hamiltonian of spin - polarized fermions in a 1d lattice with periodic boundary conditions can be written as @xmath4 where the fermionic creation and annihilation operators at site @xmath5 are denoted by @xmath6 and @xmath7 , respectively , and the local - density operator by @xmath8 . in eq . ( [ eq : hamiltonianf ] ) , the nearest- and next - nearest - neighbor hopping parameters are denoted by @xmath9 and @xmath10 , respectively , and the nearest- and next - nearest - neighbor interactions are denoted by @xmath11 and @xmath12 , respectively . in our study , we only consider repulsive interactions ( @xmath13 ) , and the number of lattice sites is denoted by @xmath14 . in ref . @xcite , we have already studied a very similar hamiltonian for hardcore bosons ( bosons with an infinite on - site repulsion ) , in which case @xmath15 where the hardcore boson creation and annihilation operators at site @xmath5 are denoted by @xmath16 and @xmath17 , respectively , and the local - density operator by @xmath18 . the parameters @xmath19 and @xmath12 in eq . ( [ eq : hamiltonianhcb ] ) have exactly the same meaning for hardcore bosons as for the fermions in eq . ( [ eq : hamiltonianf ] ) . for hardcore bosons , the creation and annihilation operators commute as usual for bosons @xmath20= [ \hat{b}^{}_{i},\hat{b}^{}_{j}]= [ \hat{b}^{\dagger}_{i},\hat{b}^{\dagger}_{j}]=0,\quad \mathrm{for } \quad i\neq j,\ ] ] but on the same site , the hardcore bosons operators satisfy anticommutation relations typical of fermions , @xmath21 these constraints avoid double or higher occupancy of the lattice sites . similar to the hardcore boson case studied in ref . @xcite , the hamiltonian ( [ eq : hamiltonianf ] ) is integrable for @xmath22 . in this section , we restrict our analysis to systems with 1/3 filling , and study the effects of finite , but small , values of @xmath23 when one departs from the integrable point . in this case , the ground state of our hamiltonian is always a luttinger liquid and no metal - insulator transition occurs in the system @xcite . for @xmath24 , the fermionic hamiltonian ( [ eq : hamiltonianf ] ) can be mapped onto the identical hamiltonian for hardcore bosons ( [ eq : hamiltonianhcb ] ) ( up to a boundary effect ) , i.e. , the fermionic operators in eq . ( [ eq : hamiltonianf ] ) just need to be replaced by operators describing hardcore bosons . this can be done using the jordan - wigner transformation @xcite , and implies that in the thermodynamic limit both systems have the same spectrum and identical diagonal correlations . however , off - diagonal correlations and related observables , such as the momentum distribution function , are very different for fermions and bosons . a finite value of @xmath10 still allows for a mapping of the fermionic hamiltonian ( [ eq : hamiltonianf ] ) onto a hamiltonian of hardcore bosons . however , the new hardcore boson hamiltonian for @xmath25 will be different to the one in eq . ( [ eq : hamiltonianhcb ] ) . the mapping introduces additional operators in the term proportional to @xmath10 . in addition , in the fermionic case , a finite value of @xmath10 breaks the particle - hole symmetry present for @xmath24 and present for any value of @xmath10 in the hardcore boson case . the study of the nonequilibrium dynamics and thermodynamics of these systems is performed using full exact diagonalization of the hamiltonian ( [ eq : hamiltonianf ] ) . for our largest system sizes , eight fermions in 24 sites , the total hilbert space has dimension @xmath26 . we take advantage of the translational symmetry of the lattice , which allows us to block diagonalize the full hamiltonian while the dimension of the largest block ( in @xmath27 space ) is @xmath28 . in this work , we focus our analysis on two observables . the first one in the momentum distribution function @xmath29 which is the fourier transform of the one - body density matrix ( @xmath30 ) . as mentioned before , this observable behaves very differently for fermions as compared with the hardcore bosons to which they can be mapped . we should add that @xmath3 is usually measured in experiments with ultracold quantum gases via time - of - flight expansion . the second observable of interest is the density - density structure factor @xmath31 which is the fourier transform of the density - density correlation matrix . since we work at fixed number of fermions @xmath32 , the expectation value of @xmath33 is always @xmath34 and , as usual , we set it to zero by subtracting that trivial constant . the structure factor can be measured in cold gases experiments by means of noise correlations @xcite . we will restrict our study of the nonequilibrium dynamics to the case in which the system is taken out of equilibrium by means of a quench . quenches are a special way to induce dynamics by starting with an eigenstate of some initial hamiltonian @xmath35 , and then instantaneously changing the hamiltonian ( at time @xmath36 ) to some final time - independent hamiltonian @xmath37 . as mentioned previously , our model is integrable for @xmath22 . since we are interested in studying the effect that approaching an integrable point has on the dynamics and thermalization of the system , we take the initial state to be an eigenstate of a system with @xmath38 , @xmath39 , @xmath10 , @xmath12 , and then quench the nearest - neighbor parameters @xmath9 and @xmath11 to @xmath40 , @xmath41 without changing @xmath10 , @xmath12 , i.e. , we only change @xmath42 . this quench is performed for different values of @xmath10 , @xmath12 as one approaches the integrable point @xmath22 . notice that our quenches do not break translational symmetry . for that reason , we always select the initial state to be one of the eigenstates of the initial hamiltonian in the total @xmath43 sector . since translational symmetry is preserved , only states with zero total momentum are required to calculate the exact time evolution of the system @xmath44 where @xmath45 is the time - evolving state , @xmath46 is the initial state , @xmath47 are the eigenstates of the final hamiltonian with zero total momentum , energy @xmath48 , and @xmath49 . the sum runs over all the eigenstates of the total @xmath43 sector . in this section , we perform the exact time evolution of up to eight fermions in lattices with up to 24 sites . this means that the largest total @xmath43 sector diagonalized had 30666 states . in contrast to classical systems , where one has to perform the time evolution in order to compute the long - time average of any observable , in isolated quantum systems ( where the time evolution is unitary ) one can predict such time averages without the need of calculating the dynamics . since the wave function of any initial state can be written in terms of the eigenstates of the final hamiltonian @xmath50 as @xmath51 , one finds that the time evolution of the quantum expectation value of an observable @xmath52 can be written as @xmath53 where @xmath54 are the matrix elements of @xmath52 in the basis of the final hamiltonian . this in turn implies that the infinite time average of the observable can be written as @xmath55 we have assumed that , as in the case of generic ( nonintegrable ) systems , the spectrum is nondegenerate and incommensurate . this means that if the expectation value of @xmath52 relaxes to some kind of equilibrium value ( up to the recurrences that must occur if the system is isolated ) , that value should be the one predicted by eq . ( [ eq : diagonal ] ) . following previous work , we prefer to think of this exact result as the prediction of a `` diagonal ensemble , '' where @xmath56 is the weight of each state within this ensemble @xcite . one objection that may arise at this point is that even if the system is nondegenate and incommensurate , the spectrum may have arbitrarily close levels and it may take an unrealistically long time for the prediction in eq . ( [ eq : diagonal ] ) to apply , i.e. , that it may not be relevant to describe experiments . in order to study how the dynamics drives our observables of interest [ @xmath3 and @xmath1 toward the prediction of eq . ( [ eq : diagonal ] ) , we follow the scheme in ref . @xcite . we study the normalized area between @xmath57 and @xmath58 , during the time evolution , and their infinite time average , i.e. , at different times we compute @xmath59 and @xmath60 in fig . [ fig : timeevolution_l24t2.0 ] , we show results for @xmath61 and @xmath62 as a function of time @xmath63 for four different quenches as one approaches the integrable point . besides undergoing the same change @xmath64 , @xmath65 @xmath66 @xmath67 , @xmath68 , these quenches have another very important property in common . the initial state for each was selected to be one eigenstate of the initial hamiltonian in such a way that the effective temperature @xmath69 of the system @xcite is always the same ( @xmath70 in fig . [ fig : timeevolution_l24t2.0 ] ) . given the energy of the time - evolving state in the final hamiltonian , which is conserved , @xmath71 the effective temperature @xmath69 @xcite is defined by the expression @xmath72 where @xmath73 is the partition function , and we have set the boltzmann constant @xmath74 to unity , and @xmath75 sets the energy scale in our system . in figs . [ fig : timeevolution_l24t2.0](a ) and [ fig : timeevolution_l24t2.0](f ) , we compare the initial momentum distribution and structure factors with the predictions of eq . ( [ eq : diagonal ] ) away from integrability ( @xmath76 ) . ( the results for other values of @xmath77 are similar and not shown here . ) the dynamics of those observables for four different values of @xmath77 is presented in figs . [ fig : timeevolution_l24t2.0](b)[fig : timeevolution_l24t2.0](e ) for @xmath61 and in figs . [ fig : timeevolution_l24t2.0](g)[fig : timeevolution_l24t2.0](j ) for @xmath62 . they show that the dynamics of @xmath3 and @xmath2 are different . away from integrability ( @xmath23 ) , @xmath62 quickly relaxes [ in a time scale @xmath78 to a value @xmath790.02 and then fluctuates around that value . @xmath61 , on the other hand , slowly drifts toward @xmath80@xmath79@xmath81@xmath82 in a much longer time scale , orders of magnitude longer than the one seen for the relaxation of @xmath62 . ( color online ) quantum quench @xmath64 , @xmath65 @xmath66 @xmath67 , @xmath68 , with @xmath83 and @xmath84 in a system with @xmath85 and @xmath86 . the initial state was chosen in such a way that after the quench the system has an effective temperature @xmath70 @xcite in all cases . given the energy of the initial state in the final hamiltonian @xmath87 , the effective temperature is computed following eq . ( [ eq : temperature ] ) . ( a ) initial and diagonal ensemble results for @xmath3 when @xmath76 . ( b)(e ) time evolution of @xmath88 for @xmath76 , 0.12 , 0.04 , and 0.0 , respectively . ( f ) initial and diagonal ensemble results for @xmath2 when @xmath76 . ( g)(j ) time evolution of @xmath89 for @xmath76 , 0.12 , 0.04 , and 0.0 , respectively.,scaledwidth=48.0% ] the behavior of @xmath62 in these spin - polarized systems is qualitatively ( and quantitatively ) very similar to the one observed for hardcore bosons in ref . the time evolution of @xmath88 for the fermions , on the other hand , is different from the one seen for the hardcore bosons in ref . @xcite . for the latter systems , @xmath3 quickly relaxed toward the predictions of the diagonal ensemble , even at integrability . in fig . [ fig : timeevolution_l24t2.0](j ) , one can see that at integrability @xmath62 for the fermions also relaxes quickly toward the predictions of the diagonal ensemble and then exhibits fluctuations between @xmath90 and @xmath91 , while @xmath61 in fig . [ fig : timeevolution_l24t2.0](e ) exhibits very large oscillations between @xmath92 and @xmath93 . additionally , we studied the dynamics of @xmath88 for a time scale ten times longer than the one depicted in fig . [ fig : timeevolution_l24t2.0](e ) and observed exactly the same behavior . for hardcore bosons @xcite , we have shown that increasing the effective temperature @xcite ( or the final energy @xmath94 ) decreases the mean value of @xmath61 and @xmath62 after relaxation and also decreases the temporal fluctuations of those quantities . in fig . [ fig : timeevolution_l24t3.0 ] , we show results for quenches similar to the ones in fig . [ fig : timeevolution_l24t2.0 ] , but with a higher effective temperature ( @xmath95 ) . the comparison between figs . [ fig : timeevolution_l24t3.0 ] and [ fig : timeevolution_l24t2.0 ] also reveals a reduction in the mean values of @xmath61 and @xmath62 and their fluctuations , which is particularly obvious at integrability . still , differences remain between the relaxation dynamics of @xmath61 and @xmath62 . also , one can notice that at integrability the mean values of @xmath61 and @xmath62 and their fluctuations are in general larger than away from integrability , and they all decrease the further away one moves from the integrable point . ( color online ) quantum quench @xmath64 , @xmath65 @xmath66 @xmath67 , @xmath68 , with @xmath83 and @xmath84 in a system with @xmath85 and @xmath86 . the initial state was chosen in such a way that after the quench the system has an effective temperature @xmath95 @xcite in all cases ( see the caption of fig . [ fig : timeevolution_l24t2.0 ] ) . ( a ) initial and diagonal ensemble results for @xmath3 and @xmath76 . ( b)(e ) time evolution of @xmath88 for @xmath76 , 0.12 , 0.04 , and 0.0 , respectively . ( f ) initial and diagonal ensemble results for @xmath2 and @xmath76 . ( g)(j ) time evolution of @xmath89 for @xmath76 , 0.12 , 0.04 , and 0.0 , respectively.,scaledwidth=48.0% ] the improvement of the relaxation dynamics with increasing the effective temperature ( energy ) of the isolated state @xcite , discussed above , can be related in part to the increase in the density of states in the final system with increasing energy . this increases the number of eigenstates of the hamiltonian that participate in the dynamics making dephasing in eq . ( [ eq : timeevolution ] ) more effective , and hence reducing temporal fluctuations after relaxation . this can be better seen in fig . [ fig : statecount ] , which clearly shows that the number of states with the largest values of @xmath96 increases as the temperature decreases . since @xmath97 , this means that the lower the temperature the smaller the total number of states that participate in the dynamics . this can be also seen in fig . [ fig : statecount ] , where the number of states with @xmath98 and @xmath99 is larger for @xmath95 [ fig . [ fig : statecount](b ) ] than for @xmath70 [ fig . [ fig : statecount](a ) ] ( notice the logarithmic scale in both axes ) . overall , one can conclude that dephasing becomes less effective with decreasing temperature and this leads to an enhancement of temporal fluctuations . a similar argument , based on the number of states participating in the dynamics , can help us understand why the average values and temporal fluctuations of @xmath61 and @xmath62 are in general larger after relaxation when one is at integrability , or close to it . figure [ fig : statecount ] also shows that for any given temperature , the total number of states that participate in the dynamics decreases as one approaches the integrable point . at integrability , this may be related to the presence of additional conserved quantities , which restrict the number of eigenstates of the final hamiltonian that can have a significant overlap with the initial state . ( color online ) number of states with @xmath96 greater than the value presented in the @xmath100 axis , for an effective temperature ( a ) @xmath70 and ( b ) @xmath95 , and for the same quenches studied in figs . [ fig : timeevolution_l24t2.0 ] and [ fig : timeevolution_l24t3.0 ] , respectively . here , @xmath85 and @xmath86.,scaledwidth=48.0% ] it would be desirable to study how the mean values and temporal fluctuations of @xmath61 and @xmath62 behave after relaxation as one increases the system size . unfortunately , the computational requirements of our approach increase exponentially with system size and a rigorous finite - size scaling is not possible . as a step in understanding how finite - size effects affect our results here , we show in fig . [ fig : timeevolution_l21t3.0 ] results for a smaller system with seven fermions in 21 lattice sites ( with @xmath95 ) . by comparing those results with the ones in fig . [ fig : timeevolution_l24t3.0 ] one can see that , as expected , decreasing the system size increases the mean values of @xmath61 and @xmath62 after relaxation and the temporal fluctuations of both quantities . this supports the expectation that with increasing the system size , after relaxation , the mean values of @xmath61 and @xmath62 should become arbitrarily small and so should the temporal fluctuations . ( color online ) quantum quench @xmath64 , @xmath65 @xmath66 @xmath67 , @xmath68 , with @xmath83 and @xmath84 in a system with @xmath101 and @xmath102 . the initial states were selected in the same way as the ones in fig . [ fig : timeevolution_l24t3.0 ] ( @xmath95 ) . ( a ) initial and diagonal ensemble results for @xmath3 and @xmath76 . ( b)(e ) time evolution of @xmath88 for @xmath76 , 0.12 , 0.04 , and 0.0 , respectively . ( f ) initial and diagonal ensemble results for @xmath2 and @xmath76 . ( g)(j ) time evolution of @xmath89 for @xmath76 , 0.12 , 0.04 , and 0.0 , respectively.,scaledwidth=48.0% ] a different question is what happens to the time scale for @xmath3 to relax to the diagonal ensemble prediction as the system size increases . previous work on that direction @xcite has suggested a possible intermediate quasisteady regime that occurs before full relaxation . this has been seen in recent numerical work @xcite , and our results for the behavior of @xmath3 in smaller systems may be an indication in that direction . something that is important to keep in mind from our results in this section is that different observables may exhibit different relaxation times . in particular , and in contrast to the hardcore boson case , the momentum distribution function of fermions is an observable that takes a long time to relax to an equilibrium distribution . this may have influenced the failure to observe thermalization in the momentum distribution function of a nonintegrable fermionic system in ref . @xcite . it is interesting to note that if @xmath103 ( noninteracting case ) the momentum distribution function of fermions in a periodic system is a conserved quantity , i.e. , for any initial state it will not change in time . however , the @xmath3 of hardcore bosons , to which the fermions can be mapped , can have a nontrivial dynamics and relaxes to the predictions of a generalized gibbs ensemble , as shown in refs . @xcite and @xcite . having @xmath104 ( in our case @xmath105 and @xmath12 varies between 0 and 0.32 ) allows the @xmath3 of fermions to change with time ( since now the fermions are interacting ) , but the time scale for relaxation to the equilibrium distribution still seems to be affected by the fact that in @xmath27 space the scattering between fermions is very special . we find the relaxation time for @xmath3 to be much longer than the corresponding time scale for other observables such as @xmath2 . the differences between the time scales for the relaxation of different observables ( and their temporal fluctuations ) can be understood in terms of the off - diagonal matrix elements of the observables , which will be discussed in sec . [ sec : offdiagonal ] . in sec . [ sec : dynamicsf ] , we have shown that , for our systems of interest , the diagonal ensemble provides an accurate description of observables after relaxation . some observables may take longer to relax , and under some circumstances they may not even relax at all [ see , _ e.g. _ , @xmath3 in fig . [ fig : timeevolution_l24t2.0](e ) ] , but whenever we find relaxation to an equilibrium value , it is well described by the predictions of eq . ( [ eq : diagonal ] ) . this was expected and we refer the reader to refs . @xcite for similar results in other quantum systems . in what follows , we study how the predictions of the diagonal ensemble compare to standard statistical mechanical ensembles . a great advantage of the infinite time average [ eq . ( [ eq : diagonal ] ) ] is that all time dependence has been removed from the time evolution of the quantum - mechanical problem and one has a unique result to test statistical mechanics . since the systems on which we have performed the dynamics are isolated , the most appropriate statistical ensemble to compare observables after relaxation is the microcanonical ensemble . as usual , the computations in the microcanonical ensemble are performed averaging over all eigenstates ( from all momentum sectors ) that lie within a window @xmath106 $ ] , where @xmath107 , and we have taken @xmath108 in all cases . similarly to what was done in refs . @xcite and @xcite , we have checked that our results are independent of the exact value of @xmath109 in the neighborhood of @xmath108 . ( color online ) ( main panels ) comparison between the prediction of the microcanonical and diagonal ensembles , for @xmath3 and @xmath2 , as a function of increasing @xmath77 for @xmath70 and @xmath95 . results are shown for ( a ) @xmath86 , @xmath85 , and ( b ) @xmath102 , @xmath101 . the diagonal ensembles correspond to the quenches in figs . [ fig : timeevolution_l24t2.0 ] , [ fig : timeevolution_l24t3.0 ] , and [ fig : timeevolution_l21t3.0 ] . ( insets ) comparison between the prediction of diagonal ensembles generated by two different initial states , for @xmath70 and @xmath95 . the results shown are for ( a ) @xmath86 , @xmath85 , and ( b ) @xmath102 , @xmath101 . as in the main panels , one of the diagonal ensembles is generated by initial states selected from the eigenstates of a hamiltonian with @xmath64 , @xmath65 , the other diagonal ensemble is generated by initial states selected from the eigenstates of a hamiltonian with @xmath110 , @xmath111 . the final hamiltonian ( with @xmath67 , @xmath68 ) and the effective temperature @xcite are identical for both initial states . by relative differences in this figure we mean the normalized area between the different ensemble predictions for @xmath3 and @xmath2 . relative differences are computed in exactly the same way as @xmath61 and @xmath62 are computed in eqs . ( [ eq : errorn ] ) and ( [ eq : errorn ] ) , respectively.,scaledwidth=36.0% ] the main panel in fig . [ fig : thermodynamics](a ) depicts how the difference between the diagonal and the microcanonical ensembles behaves as one moves away from integrability . once again , we find a different behavior for @xmath3 and @xmath2 . for @xmath2 , we find that for both effective temperatures @xcite considered , the difference between the diagonal and microcanonical ensemble is always smaller than 1% for @xmath112 , and one can say that the system exhibits thermal behavior . this is very similar to the results for the same quantity obtained in the hardcore boson systems analyzed in ref . @xcite . on the other hand , @xmath3 exhibits a larger difference between the diagonal and the microcanonical ensemble for the same values of @xmath77 , in particular at lower temperatures . for both observables , one can still see that as one approaches the integrable point the difference between both ensembles increases , signaling the breakdown of thermalization in all cases . in the inset in fig . [ fig : thermodynamics](a ) , we also compare the predictions of two diagonal ensembles generated by different initial states , which are chosen in such a way that the effective temperature of both system after the quench @xcite are the same . the behavior is qualitatively similar to the one seen in the main panel of the same figure and shows that the breakdown of thermalization for both observables is accompanied by a dependence on the initial state . results for a smaller system size , with the same density and effective temperatures , are shown in the main panel and inset in fig . [ fig : thermodynamics](b ) . they show that , as expected , the differences between the two ensembles for any given value of @xmath77 increase with decreasing system size . notice that the @xmath113 axis in figs . [ fig : thermodynamics](a ) and [ fig : thermodynamics](b ) has a different scale , so that the contrast between the results in both panels is stronger than may appear at first sight . for @xmath112 , the comparison between the results for different systems sizes led us to expect that the predictions of the microcanonical ensemble , for both observables , will coincide with the ones of the diagonal ensemble as the system size is increased , i.e. , that thermalization takes place if one is sufficiently far from integrability . we do note that @xmath3 for these fermionic systems is more sensitive to finite - size effects than @xmath2 , something that is in contrast to the behavior seen for hardcore bosons @xcite . for values of @xmath114 closer to the @xmath22 integrable point , the outcome of increasing system size is less clear , at least for the system sizes that we are able to analyze here . two possibilities emerge . ( i ) as the system size increases the point at which the results for the diagonal an microcanonical ensembles start to differ will move toward @xmath22 . ( ii ) the very same point will move toward a nonzero value of @xmath77 . here , we are not able to discriminate between those two scenarios , and further studies will be needed to address that question . it is relevant to notice that , for the small systems sizes studied in this work , it is important to select the microcanonical ensemble when comparing with the outcome of the dynamics . this is because finite - size effects can make the predictions of different standard statistical mechanical ensembles different from each other . in fig . [ fig : microvscanonical ] , we compare the results of the microcanonical ensemble , which we have been using up to this point , with the ones obtained with the canonical ensemble . one can see there that depending on the observable under consideration the results of both ensembles can differ up to @xmath115 2.5% . the differences are not very much affected by the proximity to integrability , and , as expected , they can be seen to decrease with increasing system size . one can also see in fig . [ fig : microvscanonical ] that the differences between the two ensembles are always larger for @xmath3 . interestingly , they are not as large as the ones observed for the same quantity in the case of hardcore bosons @xcite . ( color online ) relative differences ( normalized area ) between the predictions of the microcanonical and canonical ensembles for @xmath3 [ upper four curves ] and @xmath2 [ lower four curves ] in systems with @xmath102 , @xmath101 , and @xmath86 , @xmath85 . the relative differences are computed in exactly the same way as @xmath61 and @xmath62 are computed in eqs . ( [ eq : errorn ] ) and ( [ eq : errorn ] ) , respectively.,scaledwidth=36.0% ] in what follows , we will analyze the reason underlying thermalization in these isolated quantum systems , and the cause for the differences seen between @xmath3 and @xmath2 when comparing the diagonal and microcanonical ensembles . in sec . [ sec : thermodynamicsf ] , we have shown that away from integrability there are regimes in which the system thermalizes , i.e. , in which the predictions of the diagonal ensemble and the microcanonical ensemble coincide . this means that @xmath116 where @xmath117 is the number of energy eigenstates with energies in the window @xmath118 $ ] . this result is certainly surprising because the left - hand side of eq . ( [ eq : paradox ] ) depends on the initial conditions through the overlaps of the initial state with the eigenstates of the final hamiltonian ( @xmath119 ) , while the right - hand side of eq . ( [ eq : paradox ] ) only depends on the energy @xmath94 ( as mentioned before , our results do not depend on the specific value of @xmath109 ) . a possible solution to this paradox was advanced by deutsch @xcite and srednicki @xcite in terms of the eigenstate thermalization hypothesis ( eth ) . within eth , the difference between the eigenstate expectation values of generic few - body observables [ @xmath120 are presumed to be small between eigenstates that are close in energy . this implies that if one takes a small window of energy @xmath109 , as it is done in the microcanonical ensemble , all the eigenstate expectation values within that window will be very similar to each other . hence , the microcanonical average and a single eigenstate will provide essentially the same answer , i.e. , the eigenstates already exhibit thermal behavior . the same will happen with the prediction of the diagonal ensemble as long as the @xmath121 s are strongly peaked around the energy @xmath94 . the latter has been shown to be the case for quenches in hardcore boson systems in two dimensions @xcite , hardcore bosons in one dimension @xcite , softcore bosons in one dimension @xcite , and generic quenches in the thermodynamic limit @xcite . ( color online ) ( a),(d ) distribution of @xmath56 for two of the quenches depicted in fig . [ fig : timeevolution_l24t3.0 ] , for ( a ) @xmath122 and ( d ) @xmath123 . in both cases , the final effective temperature of the system is @xmath95 @xcite . the vertical dashed lines signal the energy @xmath94 of the time - evolving state . ( b),(e ) eigenstate expectation values of @xmath124 in the full spectrum ( including all momentum sectors ) of the hamiltonian ( [ eq : hamiltonianf ] ) with @xmath125 and , ( b ) @xmath76 and ( e ) @xmath126 . ( c),(f ) eigenstate expectation values of @xmath127 in the full spectrum ( including all momentum sectors ) of the hamiltonian ( [ eq : hamiltonianf ] ) with @xmath125 and , ( c ) @xmath76 and ( f ) @xmath126 . these results were obtained for systems with @xmath86 and @xmath85 , i.e. , the total hilbert space consists of 735471 states.,scaledwidth=47.0% ] in figs . [ fig : eth](a ) and [ fig : eth](d ) , we show the distributions of @xmath128 for two of the quenches studied in fig . [ fig : timeevolution_l24t3.0 ] . one quench is far away from integrability and the other one at integrability . there are some features of the distributions of @xmath128 that are important to mention here . ( i ) they are neither similar to the microcanonical nor to the canonical distributions . ( ii ) they are strongly peaked around the energy @xmath94 of the time - evolving state . actually , one can see in the figures that the values of @xmath128 decay exponentially as one moves away from @xmath94 ( signaled by the vertical line in the figures ) . ( iii ) they are not related to the effective temperature of the system , which is exactly the same in both cases ( @xmath95 ) while the distributions of @xmath128 are clearly different . the specific values of @xmath128 , and the exponent of their decay , depend on the initial state @xcite . because of these properties of the @xmath128 distributions , they alone can not explain why thermalization takes place , for example , for @xmath2 when @xmath76 in fig . [ fig : thermodynamics ] . in the bottom four panels of fig . [ fig : eth ] , we show the eigenstate expectation values of @xmath124 [ ( b),(e ) ] and @xmath127 [ ( c),(f ) ] for a system far from integrability , with @xmath125 and @xmath76 [ ( b),(c ) ] , and for a system at integrability , with @xmath125 and @xmath126 [ ( e),(f ) ] . those results , together with the @xmath128 distributions in figs . [ fig : eth](a ) and [ fig : eth](d ) , can help us understand the differences between the diagonal and the microcanonical ensembles in fig . [ fig : thermodynamics](a ) . the most striking feature in the bottom panels in fig . [ fig : eth ] is the contrast between the eigenstate expectation values at integrability and far away from it . the former ones exhibit a very wide distribution of eigenstate expectation values . this is true even if one selects a narrow window at the center of the spectrum ( equivalent to very high temperatures ) . as one moves away from integrability , one can clearly see that the fluctuations between eigenstate expectation values of nearby eigenstates reduces dramatically and eth starts to be valid . for @xmath76 , a comparison between the expectation values of @xmath124 and @xmath127 shows that for the energies of the final states created in our quenches , the distribution of the eigenstate expectation values of @xmath124 is relatively broader than the one seen for @xmath127 , which explains why the differences between the diagonal and microcanonical ensembles in fig . [ fig : thermodynamics](a ) are larger for the former one . it also helps in understanding why @xmath129 is more sensitive to the selection of the initial state . in order to gain a more quantitative understanding of how eth breaks down as one approaches integrability , we have computed the average relative deviation of the eigenstate expectation values with respect to the microcanonical prediction , @xmath130 and @xmath131 . for any given energy of the microcanonical ensemble , these quantities are computed as @xmath132 and @xmath133 are the eigenstate expectation values of @xmath124 and @xmath127 , respectively , and @xmath134 and @xmath135 are the microcanonical expectation values at any given energy @xmath94 . the sum over @xmath136 contains all states with energies in the window @xmath106 $ ] . as discussed before , we have taken @xmath108 . clearly , for different values of the hamiltonian parameters , the lowest and highest - energy states as well as the level spacing are different . hence , in order to make a meaningful comparison between different systems as one moves away from the integrable point , we study the behaviors of @xmath130 and @xmath131 as a function of the effective temperature @xmath69 of a canonical ensemble that has the same energy @xmath94 as the microcanonical ensemble . given an energy @xmath94 within the microcanonical ensemble , the effective temperature can be computed by means of eq . ( [ eq : temperature ] ) . we should stress that the effective temperature is only used here as an auxiliary tool for comparing different systems , i.e. , it just provides a unique energy scale for comparing observables independently of the hamiltonian parameters , which change the ground - state energy and the level spacing . ( color online ) average relative deviation of eigenstate expectation values with respect to the microcanonical prediction vs @xmath69 ( see text ) . results are presented for : ( a),(c ) @xmath130 and ( b),(d ) @xmath131 in systems with : ( a),(b ) @xmath86 , @xmath85 and ( c),(d ) @xmath102 , @xmath101 . the inset in ( d ) shows a direct comparison of @xmath131 between the systems with @xmath86 , @xmath85 ( thin lines ) and the systems with @xmath102 , @xmath101 ( thick lines ) . the values of @xmath77 in the three cases depicted in the inset are , from top to bottom , @xmath22 , @xmath137 , and @xmath138 , and follow the same legends of the main panel shown in ( b ) . in all cases @xmath125.,scaledwidth=48.0% ] in fig . [ fig : deviationfrometh ] , we present results for @xmath130 [ ( a),(c ) ] and @xmath131 [ ( b),(d ) ] vs @xmath69 for six different values of @xmath114 in systems with 24 lattice sites and eight fermions [ ( a),(b ) ] and 21 lattice sites and seven fermions [ ( c),(d ) ] . the average relative deviations for both observables are presented in the same scale , which immediately allows one to see what was already evident in fig . [ fig : eth ] , namely , that the average relative deviations for @xmath130 are larger ( more than two times larger ) than the corresponding ones for @xmath131 for any given value of @xmath77 . for temperatures @xmath139 , one can see that those deviations for both observables saturate with increasing @xmath77 for @xmath140 . comparing the results for the same observables but for different systems sizes , one can see that with increasing system size the average relative deviations are decreasing . this can be better seen in the inset in fig . [ fig : deviationfrometh](d ) , where we have presented a direct comparison of @xmath131 for three different values of @xmath77 in the systems with 21 ( thick lines ) and 24 ( thin lines ) sites . at integrability , one does not see much of a change with changing system size for @xmath131 , but some reduction can be seen for @xmath130 [ figs . [ fig : deviationfrometh](a ) and [ fig : deviationfrometh](c ) ] . for @xmath141 , we do see a clear reduction in all cases . at low temperatures @xmath142 , the density of states is low and in many cases the use of the microcanonical ensemble is not justified . from fig . [ fig : eth ] , one can see that in that regime ( of low energies ) the fluctuations of the observables are very large and thermalization is not expected to occur . an important question that needs to be addressed in the future is what will happen with that window of temperatures ( energies ) with increasing the system size . for our small systems , we do not see a noticeable change with increasing system size . ( color online ) average relative deviation of eigenstate expectation values with respect to the microcanonical prediction at two fixed temperatures ( @xmath70 and @xmath95 ) and for two system sizes ( @xmath102 , @xmath101 , and @xmath86 , @xmath85 ) vs @xmath77 . results are presented for : ( a ) @xmath130 and ( b ) @xmath131 . in all cases @xmath125.,scaledwidth=36.0% ] an alternative way to present some of the results depicted in fig . [ fig : deviationfrometh ] , which can help us make direct contact with the results discussed in fig . [ fig : thermodynamics ] , is to take two values of the effective temperature within the microcanonical ensemble and plot the average relative deviation of eigenstate expectation values with respect to the microcanonical prediction for those temperatures as a function of the integrability breaking parameters @xmath77 . this is done in fig . [ fig : deviationfrometh_tpvp ] for @xmath70 and @xmath95 , and @xmath102 and @xmath86 . figure [ fig : deviationfrometh_tpvp ] clearly shows that the breakdown of thermalization seen in fig . [ fig : thermodynamics ] as one approaches integrability is directly related to the increase in the relative deviation of eigenstate expectation values with respect to the microcanonical prediction , i.e. , to the increase in the fluctuations of the eigenstate to eigenstate expectation values of @xmath129 and @xmath143 , or what is the same to the failure of eth . a comparison between figs . [ fig : deviationfrometh_tpvp](a ) and [ fig : deviationfrometh_tpvp](b ) also helps in understanding why in fig . [ fig : thermodynamics ] the difference between the diagonal and microcanonical ensembles is larger for @xmath3 than for @xmath2 , even when one is far from integrability . figure [ fig : deviationfrometh_tpvp](a ) shows that for the effective temperatures selected in fig . [ fig : thermodynamics ] , @xmath130 is more than two times larger than @xmath131 . the reduction in @xmath130 and @xmath131 , with increasing system size , can also be seen in fig . [ fig : deviationfrometh_tpvp ] when comparing the results for @xmath102 and @xmath86 . while the diagonal elements of the observables , in the basis of the eigenstates of the hamiltonian , helped us understand whether after relaxation the expectation values of generic few - body observables can be described by standard statistical mechanical approaches , it follows from eqs . ( [ eq : timeevolution ] ) and ( [ eq : diagonal ] ) that the off - diagonal elements of the observables can help us understand the relaxation dynamics and temporal fluctuations of observables after relaxation , @xmath144 ( color online ) off - diagonal elements of @xmath124 [ ( a),(c ) ] and @xmath127 [ ( b),(d ) ] for fermions [ ( a),(b ) ] and hardcore bosons [ ( c),(d ) ] far from integrability . results are shown for the 100 eigenstates closest to the one with energy @xmath145 ( @xmath95 , see text ) for the fermions [ ( a),(b ) ] and @xmath146 ( @xmath95 , see text ) for the hardcore bosons [ ( c),(d ) ] . in all cases @xmath125 , @xmath76 , @xmath86 , and @xmath147.,scaledwidth=48.0% ] as we pointed out in sec . [ sec : dynamicsf ] , one of our findings in this work is that the momentum distribution function of fermions exhibits a slower relaxation dynamics than the one seen for the same observable in an identical system of hardcore bosons @xcite . these differences were seen far from the integrable point as well as at the integrable point . in figs . [ fig : offdiagonaltpvp0.32_l24t3.0](a ) and [ fig : offdiagonaltpvp0.32_l24t3.0](c ) , we compare the off - diagonal elements of @xmath124 between a system of spinless fermions [ fig . [ fig : offdiagonaltpvp0.32_l24t3.0](a ) ] and an identical system with hardcore bosons [ fig . [ fig : offdiagonaltpvp0.32_l24t3.0](c ) ] , both systems are far from integrability [ with @xmath76 ] and the central eigenstate has an energy for which the temperature of a canonical ensemble with the same energy would be @xmath95 . it is apparent in these two figures that the off - diagonal elements of @xmath124 are larger for the fermions than for the hardcore bosons , in particular close to the diagonal . the relaxation dynamics of our other observable of interest @xmath2 , on the other hand , was seen to be very similar between the fermions studied in this work and the hardcore bosons studied in ref . @xcite . in both systems , this observable relaxed very quickly ( in a time scale @xmath149 ) . figures [ fig : offdiagonaltpvp0.32_l24t3.0](b ) and [ fig : offdiagonaltpvp0.32_l24t3.0](d ) depict results for the off - diagonal elements of @xmath127 . the results for fermions and hardcore bosons are almost indistinguishable in this case , with all off - diagonal elements being much smaller than the diagonal ones . the contrast between the magnitude of the off - diagonal elements of @xmath124 in fig . [ fig : offdiagonaltpvp0.32_l24t3.0](a ) and @xmath127 in fig . [ fig : offdiagonaltpvp0.32_l24t3.0](b ) also explains the difference between the relaxation dynamics of both observables in fermionic systems depicted in fig . [ fig : timeevolution_l24t2.0 ] . ( color online ) off - diagonal elements of @xmath124 [ ( a),(c ) ] and @xmath127 [ ( b),(d ) ] for fermions [ ( a),(b ) ] and hardcore bosons [ ( c),(d ) ] at integrability . results are shown for the 100 eigenstates closest to the one with energy @xmath150 ( @xmath95 , see text ) for the fermions [ ( a),(b ) ] and @xmath150 ( @xmath95 , see text ) for the hardcore bosons [ ( c),(d ) ] . in all cases @xmath125 , @xmath151 , @xmath86 , and @xmath147.,scaledwidth=48.0% ] results for the same quantities for fermions and hardcore bosons at the integrable point are shown in fig . [ fig : offdiagonaltpvp0.00_l24t3.0 ] . we find the results at integrability in stark contrast with those far away from the integrable point . at integrability , the off - diagonal elements of both observables are in most cases very small or zero , but then there are some states between which the off - diagonal elements can be very large ( much larger than any off - diagonal element seen in the nonintegrable case ) . in addition , very large off - diagonal elements can be seen between states that have quite different energies . in fig . [ fig : offdiagonaltpvp0.32_l24t3.0 ] , the largest off - diagonal elements in the nonintegrable case always occur between states that are close in energies , and they are seen to reduce as the energy of the states depart from each other . this is not the case at integrability . a more quantitative understanding of these issues can be gained by taking one state from figs . [ fig : offdiagonaltpvp0.32_l24t3.0 ] and [ fig : offdiagonaltpvp0.00_l24t3.0 ] and plotting the off - diagonal elements between that state ( which we take to be the one at the center ) and all the closest energy eigenstates for our two observables of interest . these are depicted in fig . [ fig : offdiagonal_all ] for the nonintegrable case with @xmath76 [ ( a)(c ) ] and at integrability @xmath22 [ ( d)(f ) ] . ( color online ) off - diagonal elements of @xmath124 [ ( a),(c ) ] and @xmath127 [ ( b),(d ) ] for fermions [ ( a),(b ) ] and hardcore bosons [ ( c),(d ) ] at integrability . results are shown for the 400 eigenstates closest to the one with energy @xmath150 ( @xmath95 , see text ) for the fermions [ ( a),(b ) ] and @xmath150 ( @xmath95 , see text ) for the hardcore bosons [ ( c),(d ) ] . in all cases @xmath125 , @xmath151 , @xmath86 , and @xmath147.,scaledwidth=48.0% ] first , let us focus on the behavior of @xmath152 . results for that quantity in the nonintegrable fermionic and hardcore boson cases are shown in panels ( a ) and ( b ) for two lattices with @xmath102 and @xmath86 , respectively . for each lattice size , one can clearly see that the off - diagonal elements of @xmath124 of the fermions are always larger than the ones of the bosons , and the differences are largest close to the diagonal . as the energies between the states depart from each other @xmath152 for bosons and fermions become similar . the comparison between the two lattices also shows that with increasing system size the off - diagonal elements of the observables decrease , but the fact that @xmath152 for the fermions is larger close to the diagonal remains . similarly , we find that with decreasing the energy of the state the off - diagonal elements of @xmath152 also become larger , although for temperatures @xmath153 this effect is smaller than the finite - size effect difference between panels ( a ) and ( b ) in fig . [ fig : offdiagonal_all ] . results for @xmath152 at integrability are presented in panels ( d ) and ( e ) of fig . [ fig : offdiagonal_all ] . there one can also see that the values of @xmath152 are in general larger for fermions than for hardcore bosons and that they decrease with increasing the system size . however , here fermions and bosons share a common feature , namely , large off - diagonal elements can be seen between states that have energies that are increasingly different from each other . this , in conjunction with the fewer number of states that have a significant overlap with the initial state , can lead to larger temporal fluctuations during the time evolution after a quench at integrability . this effect is enhanced at lower temperatures where we see some increase in the magnitude of the off - diagonal elements and a decrease in the number of states that overlap with the initial state ( see fig . [ fig : statecount ] ) . finally , the off - diagonal elements for @xmath127 far from integrability and at integrability are presented in fig . [ fig : offdiagonal_all ] , panels ( c ) and ( f ) , respectively . those panels show that , for the density - density structure factor , off - diagonal elements are always much smaller than the diagonal ones , and they are relatively smaller than the ones of @xmath129 . at integrability [ fig . [ fig : offdiagonal_all](f ) ] , one can see a few values of @xmath154 that are clearly larger than the rest , but they are still less and relatively smaller than the ones seen for @xmath152 in fig . [ fig : offdiagonal_all](e ) . these results explain why @xmath143 takes less time to relax to the diagonal ensemble prediction and why the temporal fluctuations after relaxation are also smaller for this observable . we have presented a detailed study of the relaxation dynamics after a quantum quench and the description after relaxation of the momentum distribution function @xmath3 ( a one - particle observable ) and the density - density structure factor @xmath2 ( a two - particle observable ) of spinless fermions with nearest and next - nearest hoppings and interactions in one - dimensional periodic lattices . we have also compared some results for fermions with those of a similar hardcore boson system . we should stress that those models are typical for describing the physics of one - dimensional systems . for example , they can be mapped onto the spin-1/2 @xmath155 linear chain with next - nearest - neighbor interactions . ( the observables analyzed here are related to the different spin - spin correlation functions . ) hence , we expect the results reported in this manuscript to be generic and apply to other gapless models and observables in one - dimensional systems . we have shown that , in general , @xmath3 for fermions exhibits a slower relaxation dynamics , i.e. , it takes longer to relax to an equilibrium distribution than other observables for fermions [ such as @xmath1 and than @xmath3 for a similar one - dimensional system of hardcore bosons ( studied in ref . @xcite ) . close to and at integrability , we have also found that @xmath3 may even fail to relax to an equilibrium distribution while @xmath2 , and @xmath3 for a similar bosonic system , do relax to equilibrium . we have shown that this behavior of the dynamics of @xmath3 is related to the particularly large off - diagonal elements of @xmath129 in the basis of the eigenstates of the hamiltonian . those off - diagonal elements are larger than the ones for the same observable in hardcore boson systems , and much larger than the ones for @xmath143 in the same fermionic system . from the contrast between the dynamics of @xmath3 and @xmath2 emerges a general picture in which some few - body observables in isolated quantum systems may quickly relax to an equilibrium distribution while other observables may take longer to relax , or even fail to relax , to an equilibrium distribution . we have shown that far from integrability observables after relaxation are well described by standard ensembles from statistical mechanics , such as the microcanonical ensemble , which is particularly relevant to our small and isolated systems . the ability of the microcanonical ensemble to predict the expectation value of few - body observables after relaxation was shown to be related to the validity of the eigenstate thermalization hypothesis , within which the expectation values of observables in the eigenstates of the hamiltonian already exhibit thermal behavior . we have also shown that as one approaches the integrable point thermalization breaks down , with the differences between the microcanonical predictions and the results of the relaxation dynamics increasing continuously as one converges toward the integrable point . we have established that there is a clear correlation between the failure of the system to thermalize and the failure of the eigenstate thermalization hypothesis , i.e. , as one approaches integrability the differences between the eigenstate expectation values of few - body observables increases between eigenstates that are close in energy , and this happens over the full spectrum of the hamiltonian . our results here have been obtained for small one - dimensional lattices , which are relevant to current experiments in one - dimensional geometries @xcite . however , several important questions remain to be addressed in future works . for example , what happens to the point at which thermalization breaks down ( at all temperatures ) as one increases the system size . far from integrability , we find that for our small systems thermalization occurs whenever the effective temperature is @xmath139 @xcite . another question that needs to be addressed is what happens with the window of energies where thermalization fails to occur far from integrability as one increases the system size . finally , there is the question of the time scale that @xmath3 of the fermions takes to relax to the thermal distribution , and the emergence of intermediate quasistationary distributions @xcite , as the size of the system is increased . experiments with ultracold gases will generate many new questions and help address the current ones . this work was supported by the u.s . office of naval research under award no . n000140910966 and by a summer academic grant from georgetown university . we thank amy cassidy and itay hen for useful comments on the manuscript . we are grateful to the aspen center for physics for hospitality . by the effective temperature ( @xmath69 ) , we mean the temperature of a canonical ensemble with the same energy as the ( isolated ) time - evolving state . this temperature help us gauge the excitation energy of the time - evolving state with respect to the ground state . in addition , in thermodynamics systems where observables are expected to thermalize , @xmath69 allows one to predict the expectation values of observables of interest after relaxation .

we have studied properties and shell structure of the superheavy elements from to within the framework of the rmf theory . the region of study spans nuclides with neutron numbers . the lagrangian model nl - sv1 with the inclusion of the vector self - coupling of the-meson has been employed in this work . we have performed rmf + bcs calculations for an axially deformed configuration of nuclei . the ground - state binding energies , single - particle properties and quadrupole deformation of nuclei have been obtained from the mean - field minimizations . two - neutron separation energies , values and-decay half - life have been evaluated . it is shown that a large number of nuclides exhibit the phenomenon of shape - coexistence over a significant region of the superheavy elements . shape coexistence of a prolate and an oblate shape is prevalent in nuclides far below , whilst nuclei in the vicinity of tend to show a shape coexistence between a spherical and an oblate shape . the shell structure and 2-neutron separation energies obtained with the rmf theory reinforce the neutron number as a major magic number . it is shown that the neutron number acts akin to a magic number in the deformed region . it is suggested that the combination and has the potential of being a doubly magic number in the superheavy region .

0908.2678

significant progress has been made in the discovery of new superheavy nuclei in the last decade @xcite . superheavy elements at the extreme end of the periodic table have been synthesized in the laboratory . the synthesis of superheavy nuclei with @xmath8 at gsi , darmstadt @xcite and jinr , dubna @xcite has stimulated the interest in the properties of superheavy nuclei . discovery of isotopes of @xmath9 and @xmath10 was reported in jinr , dubna and synthesis of @xmath11 at berkeley was reported some time ago @xcite . earlier superheavy nuclei could be unambiguously identified by their @xmath5-decay chains leading to already known nuclei . however , the decay chains of the newly found superheavy nuclei can not be linked to any nuclei known due to their relatively extreme positions in the periodic table . therefore , their identification depends much upon a comparison with theoretical models . much progress has also been made in theoretical investigations of the properties of superheavy nuclei . theoretical models such as macroscopic - microscopic ( mm ) models have been employed to discern and predict properties of superheavy nuclei . at the same time , various mean - field approaches have also been employed to investigate basic properties in this domain . the prediction of @xmath12114 was confirmed by mm models such as frdm with folded yukawa single - particle potential @xcite . for the next doubly - magic nucleus , the available self - consistent models predict this to the more proton rich @xmath13120 or the even heavier one at @xmath14126 . predictions in various models tend to diverge primarily due differences in the form of the single - particle potentials . possibly different extrapolations arise in self - consistent relativistic mean - field ( rmf ) theory . here , the spin - orbit coupling is naturally built - in as a consequence of the dirac - lorentz structure of nucleons , whereas in all non - rmf models , it has to be introduced by hand . the non - rmf models tend to overestimate spin - orbit splitting with increase in mass number , thus affecting actual predictions of shell closures in the superheavy region . superheavy nuclei with inherently large density of single - particle states can be a sensitive probe for models of nuclear structure . various microscopic approaches such as non - relativistic density - dependent skyrme hartree - fock ( shf ) theory @xcite and that of mm type @xcite are used extensively to investigate the properties and structure of superheavy nuclei . in the mm approach , a sum of smooth energy based upon liquid - drop type formula on which shell correction is imposed using the method of strutinsky is obtained as the total binding energy of nuclei . the finite - range droplet model ( frdm ) has achieved a significant success amongst the mm models . the mass formula frdm is used to calculate the shell correction energies in order to identify the major magic numbers in the region of superheavy nuclei . a major proton magic number @xmath10 and a neutron magic number @xmath6 have been predicted by the frdm . the mm approach has also been performed by sobiczewski group @xcite , where the strutinsky shell correction @xcite is used for the microscopic component , whereas yukawa + exponential model is used for the macroscopic part . superheavy nuclei have also been studied extensively within the self - consistent mean - field models of skyrme hartree - fock ( shf ) type . comparative studies of shf and rmf models have been made . one of the first studies @xcite performed within rmf theory earlier on was using the force nl - sh @xcite . it was shown that neutron number @xmath6 appears to be magic for lighter superheavy nuclei whereas its importance is reduced for heavier superheavies with @xmath15 . there was also an indication of a small deformed shell - closure at @xmath10 for the proton number . in comparison , the frdm predicted a strong shell gap at @xmath10 in the deformed region . extensive shell correction calculations were performed by kruppa et al . @xcite for sets of the skyrme and rmf forces in order to identify the magic numbers in the region of superheavy elements theoretically . the skyrme models predict the strongest shell effects for @xmath6 and @xmath16 124 , 126 and not at @xmath10 , whereas rmf theory did not show a clear shell gap at @xmath6 especially for heavier she . this was attributed to the difference in the shell structure in the above two approaches . this may have been due to a different isospin dependence in the skyrme model as compared to the rmf theory @xcite . in comparison , all the skyrme forces have predicted a broad valley of shell stabilization around @xmath1 and @xmath17 172 and 184 . on the other hand , cwiok _ et al . @xcite have suggested a shift in the island of stability towards @xmath14126 . in spite of an impressive agreement with experimental data for the heaviest elements , the theoretical uncertainities are large when extrapolating to unknown regions of the nuclear chart . in particular , there is no consensus among theories with regard to the regions of the shell stability for superheavy nuclei . since in these nuclei the single - particle level density is relatively large , small shifts in the position of single - particle levels ( e.g.,due to the coulomb or spin - orbit interaction ) can be crucial for determining the shell stability of a nucleus . while most macroscopic - microscopic ( non - self - consistent ) models predict @xmath10 to be magic , most self - consistent calculations suggest that the center of the proton shell stability should be moved up to higher proton numbers , i.e. , @xmath18 , or @xmath19 . for neutrons most non - relativistic calculations predict magic gap at @xmath6 while the relativistic mean - field theory yields @xmath7 - due to a slightly different spin - orbit interaction @xcite . the experimental determination of superheavy nuclei from this region will thus be very important for pinning down the fundamental question of the spin - orbit force . a number of theoretical studies have been carried out within the framework of the rmf theory @xcite to investigate properties of superheavy nuclei , in particular , @xmath5-decay properties . in ref . @xcite the nature of possible magic numbers within the framework of spherical relativistic hartree - bogoliubov theory using various interactions has been studied . a recent theoretical work @xcite sheds some light on the question of magic superheavy nuclei . according to the calculations , the patterns of single - particle levels are significantly modified in superheavy elements . firstly , the overall level density grows with mass number a , as @xmath5 @xmath20 . secondly , no pronounced and uniquely preferred energy gaps appear in the spectrum . this shows that shell closures which are to be associated with large gaps in the spectrum are not robust in superheavy nuclei . the theory predicts that beyond @xmath21 and @xmath22 the usual localization of shell effects at magic numbers is diminished . instead , theory predicts fairly wide areas of large shell stabilization without magic gaps . the relativistic mean - field ( rmf ) theory @xcite has proved to be successful as a framework for description of various facets of nuclear properties . in the rmf theory , the nuclear force is produced by a virtual exchange of various mesons . the nuclear saturation is achieved by a balance between an attractive @xmath23- and a repulsive @xmath3-field . the relativistic lorentz covariance of the theory allows an intrinsic spin - orbit interaction based upon exchange of @xmath23- and @xmath3-mesons . this has been shown to be advantageous for properties which depend upon spin - orbit potential . an immediate advantage of the proper spin - orbit potential has been the success of the rmf theory to be able to describe the anomalous kink in the isotope shifts of pb nuclei @xcite . the isotopic shifts of pb have been measured with a high precision using atomic beam laser spectroscopy and are known to show a pronounced kink about the magic number @xmath22 . the non - relativistic approaches based upon the skyrme and gogny forces have , on the other hand , been unable to reproduce this kink . it was shown that this difference in the predictions of the skyrme model and the rmf theory is due to the isospin dependence of the spin - orbit term @xcite . the isospin dependence in the rmf theory is provided mainly by the coupling constant @xmath24 of the @xmath25-meson . however , as the strength of the spin - orbit term derives from a large sum of the absolute values of the scalar and vector fields , the contribution of @xmath25-field is much weaker in comparison . consequently , the isospin dependence of the spin - orbit potential in the rmf theory is relatively weak @xcite . with the success of the rmf theory in properties of nuclei related to the shell effects , we have carried out a study of the superheavy nuclei with an rmf lagrangian model which has been shown to have improved shell effects in nuclei . in the present work , we have explored the region of superheavy nuclei within the framework of the relativistic mean - field theory using the lagrangian model with the vector self - coupling of the @xmath3 meson . the region encompasses the superheavy elements from @xmath0 up to @xmath1 . nuclides in these isotopic chains with neutron number spanning @xmath26 to @xmath27 have been considered . we have used the force nl - sv1 due to sharma et al . @xcite with the vector self - coupling of the @xmath3 meson . our focus is primarily to look for possible signatures of neutron shell gaps or magic numbers in the region of superheavy nuclei . in the present work , we have employed the rmf theory to study properties of superheavy nuclei . the starting point of the rmf theory is the model due to walecka @xcite , which describes the nucleons as dirac spinors interacting by the exchange of several mesons : scalar mesons ( @xmath23 ) couple to the nucleons ( @xmath28 ) through a yukawa term @xmath29 and produce strong attraction , isoscalar vector mesons ( @xmath3 ) couple to the conserved nucleon current @xmath30 and cause almost as strong a repulsion . in addition , there are isovector @xmath25 mesons which couple to the isovector current and photons to produce the well - known electromagnetic interaction . we restrict ourselves to the hartree treatment , where the @xmath31 independent nucleons with the single - particle spinors @xmath32 ( i= 1, .. ,a ) form a slater determinant and move independently in the meson fields . the rmf theory starts with the standard lagrangian density @xmath33 where the summation convention is used and the sum over @xmath34 runs over all nucleons . isovector quantities are indicated by the arrow bars . as proposed by boguta and bodmer @xcite , the @xmath23 meson moves in a nonlinear potential of the form : @xmath35 here , @xmath36 , @xmath37 , @xmath38 and @xmath39 are the nucleon , the @xmath23- , the @xmath3- , and the @xmath25- meson masses , respectively . @xmath40 , @xmath41,@xmath42,and @xmath43/4@xmath44=1/137 are the coupling constants for the @xmath23-,the @xmath3-,the @xmath25-mesons and for the photon . the coupling constant @xmath45 represents the vector self - coupling of @xmath3-meson as introduced in ref . @xcite . the classical variation principle provides the klein - gordon equations for the mesons : @xmath46 whereby the sources are determined by the corresponding density and current distributions in the static nucleus . variation for the nucleonic field gives the dirac equation for nucleons : @xmath47 the effective mass @xmath48(r ) is determined by the scalar field @xmath23(r ) as @xmath49 and the potential vector potential @xmath50 is obtained as : @xmath51 bold faced letters indicate vectors in the 3-dimensional space . these fields are the solutions of the inhomogeneous klein - gordon equations . as is the case in the majority of applications , the contributions of antiparticles are neglected , i.e. , the _ no - sea _ approximation is employed . equation ( [ eqn:2.3 ] ) together with eq . ( [ eqn:2.4 ] ) provides a closed set of equations . in order to describe the ground - state properties of nuclei , static solutions are obtained from the equations of motion eqs . ( [ eqn:2.3]-[eqn:2.4 ] ) . in this case , the nucleon spinors are the eigenvectors of the stationary dirac equation , which yields the single - particle energies @xmath52 as eigenvalues . the solution of these equations is obtained iteratively . using these solutions , one calculates physical quantities such as the total energy , charge radii and quadrupole moments amomg others . for nuclei in the superheavy region exhibit significant deformations , we have employed the rmf theory for axially deformed configuration of nuclei . we solve the dirac equation ( [ eqn:2.3 ] ) as well as the klein - gordon equations ( [ eqn:2.4 ] ) by expansion of the wavefunctions into a complete set of eigen solutions of an harmonic oscillator potential @xcite . in the axially symmetric case the spinors @xmath53 and @xmath54 are expanded in terms of the eigenfunctions of a deformed axially symmetric oscillator potential @xmath55 taking volume conservation into account , the two oscillator frequencies @xmath56 and @xmath57 can be expressed in terms of a deformation parameter @xmath58 . @xmath59 @xmath60 the corresponding oscillator length parameters are given by @xmath61 the volume conservation gives @xmath62 . the basis is now determined by the two constants @xmath63 and @xmath58 , which are chosen optimally . the deformation parameter of the oscillator basis @xmath58 is chosen to be identical for the dirac spinors and the meson fields . the deformation parameter @xmath64 is obtained from the calculated quadrupole moments for protons and neutrons through @xmath65 with @xmath66 ( fm ) . the quadrupole moments are calculated by using the expression : @xmath67 shell effects play an important role in identifying and consequently synthesizing superheavy elements due to additional stability provided by the shell effects . hence few comments on shell effects would be in order . the shell effects manifest strongly in nuclei with the existence of magic numbers . the origin of shell closure has long been understood due to spin - orbit coupling and to an ensuing splitting of levels . the spin - orbit potential and a bunching of levels create shell closures ( magic numbers ) which are predicted correctly by most of the models . in the non - relativistic density - dependent theory of the skyrme type , the spin - orbit interaction is added phenomenologically and its strength is adjusted to reproduce the spin - orbit splitting in @xmath68o . in the rmf theory , on the other hand , the spin - orbit interaction arises naturally due to exchange of @xmath23 and @xmath3 mesons between the nucleons . the strength of the interaction is determined by the spin - orbit splitting in @xmath68o and other nuclei , which is uniquely decided by the effective mass . as mentioned in the introduction , the shell effects in the rmf theory have been advantageous in a few aspects such as isotope shifts in nuclei . in view of this , application of the rmf theory to superheavy nuclei may have more realistic predictions . in this work , we have employed the lagrangian model with the quartic coupling of @xmath3 meson . the non - linear vector self - coupling of @xmath3-meson was introduced by bodmer @xcite and properties of nuclear matter were discussed on adding a quartic term in @xmath3-meson potential . the inclusion of the vector self - coupling of @xmath3 meson in addition to the non - linear scalar self - coupling of @xmath23 meson was found to have the effect of softening the high - density equation of state ( eos ) of the nuclear matter @xcite . subsequently , the force nl - sv1 @xcite with the inclusion of the vector self - coupling of @xmath3 meson was developed with a view to improve the predictions of the ground - state properties of nuclei , such as binding energies , charge radii and isotopes shifts of nuclei along the stability line . it was shown that with the lagrangian model nl - sv1 , the shell gaps in nuclei along the stability line such as ni and sn isotopic chains were improved as compared to lagrangian model with the nonlinear scalar coupling alone @xcite . with the improved shell structure along the line of @xmath69-stability with nl - sv1 , it is expected that it shall have improved predictions in the region far away from the stability line . in view of this , we have used the parameter set nl - sv1 with the vector self - coupling of @xmath3 meson in our investigations . the parameters of the parameter set nl - sv1 @xcite are given in table i. 0.7 cm .[table:]the lagrangian parameters of the force nl - sv1 @xcite used in the rmf calculations . [ cols="<,>",options="header " , ] the input parameters required to carry out explicit numerical calculations are : neutron pairing gap @xmath70 , the proton pairing gap @xmath71 and the number of oscillator shells n@xmath72 and n@xmath73 of the fermionic wavefunctions and meson fields , respectively . both the fermionic and bosonic wavefunctions have been expanded in a basis of 20 harmonic oscillator shells in this work . we have carried out a study of superheavy nuclei from no ( @xmath0 ) to @xmath1 . the neutron number ranges from @xmath26 to @xmath27 . this region encompasses the neutron number @xmath6 which has been predicted to be a magic number in various theories . a possible deformed shell has also been predicted in this region as discussed in section i. we have carried out deformed rmf+bcs calculations for superheavy nuclei in the above region . for pairing gaps , we have used the formula due to mller and nix @xcite as given by : @xmath74 @xmath75 an axially symmetric deformed configuration with reflection symmetry has been assumed for nuclei . for each nucleus an initial basis deformation of prolate as well as oblate type has been taken in order to seek a minimum both in the prolate as well as in the oblate region of the deformation space . consequently , we have obtained a prolate and an oblate minimum for most of the nuclei . however , nuclei close to a potential magic number assume a spherical shape . the rmf+bcs minimizations with nl - sv1 give rise to a prolate and an oblate solution for most of the nuclei . we show the quadrupole deformation @xmath64 for the lowest energy state ( ground state ) for all the isotopic chains in figures [ fig : fig1](a)- [ fig : fig1](j ) . a large number of nuclei in these chains assume a prolate shape in the ground state . this is especially the case in the region of neutron number @xmath76 . typically , these nuclei exhibit a prolate deformation of @xmath77 . for the isotopic chains of no @xmath78 , rf @xmath79 and sg @xmath80 , there is a transition from a prolate shape to an oblate shape at @xmath81 . beyond @xmath7 , nuclei assume an oblate shape in the ground state for these isotopic chains ( see figs . [ fig : fig1](a)@xmath82 [ fig : fig1](d ) ) . the magnitude of the oblate deformation decreases as the neutron number increases above @xmath7 , which approaches a vanishing value as the neutron number @xmath6 arrives . in the vicinity of @xmath6 , nuclei exhibit a spherical shape . such a feature has been commonly predicted by several relativistic as well as non - relativistic theories , whereby @xmath6 constitutes a major magic number in the region of superheavy nuclei . values for isotopes of the elements @xmath83 obtained with nl - sv1 . the results with frdm are shown for comparison.,scaledwidth=100.0% ] as one goes to the heavier elements above sg @xmath80 , one can visualize that the tendency of prolate shapes in the range of neutron numbers @xmath84 continues for heavier superheavy elements as well . the point of shape transition from prolate to oblate seems to be shifting to a higher neutron number . for instance , for hs ( @xmath85 ) the prolate to oblate transition occurs at @xmath86 . for the isotopic chain ds ( @xmath87 ) the shape transition point advances to @xmath88 . however , for @xmath9 and above , the transition point falls back to @xmath86 and below it . for instance , for @xmath89 and @xmath90 , the shape transition occurs at @xmath91 and @xmath7 , respectively . similarly , the shape transition point for @xmath1 falls back to @xmath92 . thus , the number of nuclei which exhibit an oblate shape below the neutron number @xmath6 increases steadily in going from @xmath9 to @xmath1 . as far as the spherical shape for @xmath6 nuclei is concerned , it is exhibited by all the isotopic chains from @xmath0 to @xmath1 . this is reminiscent of a magic shell character for the neutron number @xmath93 . the general feature of shape transition from prolate to oblate in the region of @xmath94 prevails for all the isotopic chains considered in this study . we have made a comparison of nl - sv1 deformations with those predicted by the macroscopic - microscopic mass formula frdm . for the isotopic chains @xmath95 , the frdm exhibits a predominantly prolate shape for the neutron numbers @xmath84 . this feature is very similar to that predicted by nl - sv1 . the frdm shows a shape transition from prolate to an oblate shape at the neutron number @xmath96 for the chains @xmath97 . there is only a single nuclide exhibiting an oblate shape in frdm at @xmath96 for these chains , beyond which all the nuclei assume a spherical shape for the isotopic chains of @xmath95 in the frdm predictions , in contrast to nl - sv1 . the oblate deformation for @xmath86 nuclei in this region amounts to @xmath98 in the frdm . this is unlike nl - sv1 , where a significant number of nuclei exhibit a moderately deformed oblate shape until the supposedly magic number @xmath6 arrived at . for the isotopic chains @xmath99 , the behaviour of frdm for deformation properties is rather uneven . for the region of the lower neutron numbers , some nuclei in these chains show a moderately prolate shape , whereas a few others exhibit a small prolate deformation . this is followed by a shape transition to an oblate deformation at @xmath7 , followed by a spherical shape in going to higher neutron numbers . that nuclei in the vicinity of @xmath6 are predominantly spherical in frdm points to the hypothesis that @xmath6 maintains its magicity even for heavier superheavy elements in frdm . the mass formula extended thomas - fermi with strutinsky integral ( etf - si ) @xcite was obtained to fit experimental masses of more than 1600 nuclei from all over the periodic table . the etf - si , due to its partly microscopic structure , has found broader acceptance for nuclear physics applications . it predicts binding energies and deformations of nuclei over a broad range of masses quite well . here we compare the ground state deformations obtained with nl - sv1 with those predicted by etf - si . the quadrupole deformations @xmath64 are compared for the given isotopic chains in figs . [ fig : fig2](a)@xmath82[fig : fig2](j ) . figs . [ fig : fig2](a)@xmath82[fig : fig2](d ) show that etf - si deformations are in good agreement with the nl - sv1 values from @xmath26 to @xmath100 for the elements from ds ( @xmath0 ) to hs ( @xmath85 ) . for ds ( @xmath87 ) , the agreement between nl - sv1 and etf - si is good up to @xmath101 ( see fig . [ fig : fig2](e ) ) . a constrasting picture with the etf - si is that for neutron numbers @xmath100 and above , it shows a strong divergence from the nl - sv1 predictions . etf - si does not exhibit a prolate to oblate transition as do nl - sv1 and frdm predictions . in constrast , etf - si predicts a constant and a large value of @xmath102 for nuclei in the range of @xmath103 for the isotopic chains of @xmath97 . for the heavier elements @xmath104 , etf - si predicts a nearly constant value @xmath102 for all the neutron numbers @xmath105 . values for the isotopes of the elements @xmath83 obtained with nl - sv1 are compared with etf - si predictions.,scaledwidth=100.0% ] etf - si predicts a spherical shape of nuclei in the vicinity of @xmath6 for the elements @xmath106 . this feature of etf - si is similar to that of frdm . however , for @xmath89 , etf - si bypasses the spherical shape for @xmath6 . on the other hand , for several isotopic chains etf - si predicts a shape transition from spherical to a strongly prolate shape at @xmath107 ( see figs.[fig : fig2](b)@xmath82 [ fig : fig2](g ) ) . for @xmath87 and @xmath9 , nuclei with two neutrons below and two neutrons above @xmath6 exhibit a spherical shape , whereas for @xmath10 there is only one nuclide i.e , @xmath6 that shows a spherical shape . it s neighbours on both sides , assume a prolate shape in etf - si . this is indicative of a decrease in the shell strength at @xmath6 with etf - si . this point is supported by a deformed shape taken by the @xmath6 nuclide for @xmath89 ( see fig . [ fig : fig2](h ) . ) with etf - si . whilst etf - si values for @xmath90 and @xmath1 are not known , the aforesaid diminishing of @xmath6 shell strength points to an erosion of a possible @xmath6 magicity for the heaviest superheavy elements from @xmath89 to @xmath1 with etf - si . obtained with nl - sv1 for @xmath83 and @xmath2.,scaledwidth=90.0% ] a summary of quadrupole deformation @xmath64 in the ground state of superheavy nuclei with nl - sv1 obtained in this work is shown in the fig [ fig : fig3 ] . the nuclei denoted by the red colour indicate a prolate ground state . an oblate ground state is shown by the green colour . spherical nuclei are indicated by yellow . as seen in fig . [ fig : fig3 ] , nuclides with @xmath108 for almost all the isotopic chains exhibit a prolate shape . on the other hand , nuclides with @xmath109 exhibit an oblate shape until a spherical shape appears for nuclei close to @xmath6 . there is a little exception here in that for she with @xmath85 to @xmath9 , a few nuclides above @xmath7 continue to exhibit a prolate shape . this indicates an intrusion of prolate shape into the domain of oblate shapes as indicated in the middle of the figure . nuclei in the vicinity of @xmath6 are predominantly spherical as shown by the yellow blocks . in a few cases and especially for @xmath0 , nuclei two neutrons below and two neutrons above @xmath6 assume a spherical shape . as shown by the picture , @xmath6 acts as a major magic number in the neutron - rich region for all the she with @xmath83 . as we will see in the single - particle levels , the neutron number @xmath6 remains as a major magic number . also we will see that in the rmf theory with nl - sv1 , the proton number @xmath1 is also a magic number . the combination @xmath1 and @xmath6 seems to act as double closed shell in the rmf theory . nuclides above @xmath6 show a prolate ( red ) region for a few lighter she . however , for heavy she with @xmath85 , nuclides above @xmath6 are predominantly oblate ( green ) as shown in the fig . [ fig : fig3 ] . thus , a significant region below as well as above @xmath6 lends itself to an oblate deformation . the phenomenon of shape co - existence in the region of deformed nuclei arises due to interplay of deformed single - particle levels . splittings and alignment of deformed single - particle levels leads often to ground states with a prolate and an oblate shape nearly degenerate in energy . for practical purposes , when the two minimum energy states are within @xmath1101 mev , the two states are said to be shape co - existent . exhibiting the shape - coexistence in the lagrangian model nl - sv1.,scaledwidth=95.0% ] we show in figs . [ fig : fig4](a)@xmath82[fig : fig4](j ) nuclides which exhibit a shape co - existence with nl - sv1 . looking at the figures one can see that a large number of nuclides exhibit the shape co - existence for the superheavy chains . the nuclides in isotopic chains from @xmath0 ( no ) to @xmath9 show a common pattern of shape co - existence between prolate and oblate shapes for the region of neutron numbers @xmath111 to @xmath112 . whilst for @xmath0 and @xmath113 shape co - existence is limited to neutron numbers @xmath114 to @xmath96 , for @xmath115 to @xmath9 the region of shape co - existence extends to neutron number @xmath116 . whereas for the lighter neutron numbers there is a prolate - oblate shape co - existence for elements @xmath0 to @xmath9 , there is a shape co - existence of spherical - oblate shapes for @xmath116 isotones from @xmath117 . this indicates a softer nature of @xmath6 shell closure even for the lighter superheavy elements . several nuclides with @xmath118 , also exhibit a spherical - oblate shape co - existence for the isotopic chains @xmath119 . figures [ fig : fig4](g)- [ fig : fig4](j ) show the shape co - existence for heavier superheavy elements from @xmath10 to @xmath1 . unlike the picture for superheavy elements @xmath95 , the shape co - existence in the lighter neutron region is reduced significantly for @xmath120 . only a few nuclides below @xmath88 exhibit a prolate - oblate shape co - existence . for @xmath10 and @xmath89 , several nuclides exhibit a spherical - oblate shape co - existence for neutron numbers @xmath121 . interestingly , @xmath122 nuclide for @xmath89 shows a spherical - oblate shape co - existence . this indicates a softening of the shell effects for @xmath6 for @xmath89 . the superheavy elements @xmath90 and @xmath1 exhibit the shape co - existence rather scarcely i.e , only two isotopes @xmath92 and @xmath7 show a prolate - oblate shape co - existence . in the rest of the neutron region , there is no shape co - existence for the heaviest superheavy elements with @xmath90 and @xmath1 this may imply that the shell strength at @xmath6 has picked up for these chains and that a readjustment of the deformed single - particle levels does not lend to degenerate shapes in the ground state . a summary of the shape co - existence in superheavy nuclei is illustrated in fig . [ fig : fig5 ] . the brown squares depict a shape co - existence between a prolate and an oblate shape , whereas the blue squares represent a shape co - existence between a spherical and an oblate shape . as the figure shows , the phenomenon of prolate - oblate shape co - existence pervades the region from @xmath92 to @xmath88 . for the lighter elements this shape co - existence lies in the neutron region 170 - 178 . the lower boundary of this region gradually shifts to higher neutron numbers @xmath87 is reached . as one moves above @xmath87 , the region shifts gradually towards lower neutron numbers again . from @xmath10 to @xmath1 , however , there are fewer cases of the prolate - oblate shape co - existence . apparently , the region of prolate - oblate shape co - existence as shown by the brown colour is symmetric about @xmath87 . in addition , there are a few more cases of prolate - oblate shape co - existence for higher neutron numbers near n@xmath123 . the region of spherical - oblate shape co - existence lies near @xmath6 . the neutron number @xmath6 being a major magic number , the ground - state of nuclides with @xmath6 is spherical as seen earlier . for almost all the isotopic chains there is no shape co - existence ( empty squares ) at @xmath6 , with the exception of @xmath89 . only for @xmath89 , there is a spherical - oblate shape co - existence at @xmath6 . this is due to an erosion of the magicity of @xmath6 for @xmath89 . nuclides with @xmath116 show a spherical - oblate shape co - existence as shown by blue squares . nuclides with @xmath124 and @xmath125 also exhibit a spherical - oblate shape co - existence for several isotopic chains . the fact that nuclides just below and above @xmath6 display a spherical shape co - existence indicates that @xmath6 is not such a strong magic number . this can be constrasted with other magic numbers in the periodic table which exhibit robustness . there are rarely occasions of a shape - coexistence or even a deformation in the direct vicinity of a major magic number especially when one is not very far from the line of @xmath69-stability . superheavy nuclei especially those associated to @xmath6 do not fall in this category and are deemed to be in a region far from the stability line . therefore , mellowing of the shell strength of @xmath6 is not unexpected . 0.1 cm magic numbers in nuclei are characterized by a large shell gap between the last single - particle level of the magic number and the next single - particle level above it . this results in a larger value of 2-neutron separation energy @xmath126 for a nucleus with 2-neutrons more than the magic number . consequently , the major magic numbers exhibit a characteristic kink in @xmath126 values all over the periodic table . thus , the difference @xmath127 at the magic neutron number @xmath128 reflects its shell gap . its magnitude is a measure of the shell strength . for the isotopes of @xmath0 to @xmath1 with neutron numbers ranging from @xmath26 to @xmath27.,scaledwidth=62.0% ] with a view to visualize as to whether there are possible shell closures in neutron numbers for superheavy nuclei , we have computed @xmath126 values for all the isotopic chains . using the binding energies of neighbouring isotopes , s@xmath129 is defined as , @xmath130 the @xmath126 values for the isotopic chains considered in this work are shown in fig . [ fig : fig6 ] . the values correspond to the lowest energy state ( ground state ) obtained with rmf+bcs minimization using the force nl - sv1 . as expected , the @xmath126 values show a decreasing trend in going to nuclei with higher neutrons . the @xmath126 values show a rather monotonous decrease in going from @xmath131 to @xmath132 with a slight kink - like structure in the neighbourhood of @xmath133 and @xmath134 . due to significant changes in deformation properties of nuclei near @xmath135 and @xmath136 , there is much more structure in the @xmath126 values . for the isotopes of elements @xmath0 to @xmath1 , with neutron numbers ranging from @xmath131 to @xmath92 . ( b ) s@xmath129 values for @xmath92 to @xmath27.,scaledwidth=80.0% ] in order to visualize the structure in the @xmath126 values , we have split fig . [ fig : fig16 ] into two parts which are shown in fig . [ fig : fig7 ] . in fig . [ fig : fig17 ] ( a ) the @xmath126 values for @xmath137 are shown . the higher neutron part with @xmath138 is shown in fig . [ fig : fig7 ] ( b ) . one can notice a slight kink at @xmath139 for several isotopic chains . a kink like structure is visible for @xmath140 . this structure diminishes for @xmath90 and @xmath1 . the structure at @xmath139 is compatible with a deformed shell gap predicted by non - relativistic macroscopic - microscopic approaches @xcite . one can also notice a slight kink at @xmath134 . thus , there seems to exist a few pockets of presumably deformed shell gaps at @xmath133 and @xmath134 . however , there are no signatures of any strong magic number in the region @xmath137 . fig . [ fig : fig7 ] ( b ) shows structures in @xmath126 values at @xmath141 172 , 178 and 184 . the behaviour of @xmath126 values in the neutron region @xmath138 is far from monotonous . however , for the lighter superheavy elements , there is very little structure at @xmath142 except for @xmath0 and 104 . for the heavier elements with @xmath143 , the kink at @xmath135 shows a slightly upward trend . a kink like structure in this region indicates a possible @xmath135 deformed shell gap for superheavy elements . a combination of @xmath142 and @xmath144 might lend itself to a structure akin to a double shell closure . the region near @xmath145 is full of structural effects as can be seen in fig . [ fig : fig7 ] ( b ) . a kink in the @xmath126 values can be seen clearly at @xmath6 in several isotopic chains with @xmath146 . thus , for the lighter superheavy elements nl - sv1 predicts a shell closure at @xmath6 as indicated by the kink . for the deformed rmf+bcs calculations with nl - sv1 , a sharp decrease in the s@xmath129 values at @xmath6 is indicative of a shell closure which is consistent with that with nl - sh @xcite . for she above @xmath147112 the kink and the structure at @xmath6 is reduced significantly . thus , for the heaviest she , @xmath126 values do not give an unambiguous indication of a shell closure at @xmath6 . in the pursuit of superheavy heavy elements , it is one of aims of research to find out regions of extra - stability with a view to synthesize superheavy elements with the most appropriate projectile and target combinations . with this in mind , it will be helpful if one can explore the regions of higher stability theoretically . depending upon the region of an extra - stability which would arise from shell gaps and possible shell closures , the half - life of @xmath5-decay is a potential indicator of a possible value of stability for an area of enhanced stability . it is expected that for nuclei near a shell closure the @xmath5-decay half - lives would be larger than their neighbours . nuclei in the transuranium and superheavy region are known to be @xmath5 emitters . the rate of @xmath5-decay depends strongly on @xmath148 value of a nucleus . we have calculated @xmath148 values for nuclei in all the isotopic chains we have considered . the q@xmath149 values thus obtained are shown in fig . [ fig : fig8 ] . as one moves from @xmath113 to @xmath1 , @xmath4 values show a continuous increase with increase in the charge number of a nucleus . for all the isotopic chains one can see a decrease in the @xmath4 values with an increase in neutron number . a decreasing @xmath5-value implies a higher @xmath5-decay half - life . consequently , in moving from the lower neutron number @xmath26 towards @xmath6 , one has a stabilizing factor with a decrease in the @xmath4 values . for nuclides with @xmath124 , a slight peak in @xmath4-values for several isotopic chains with @xmath150 can be seen . in terms of @xmath5-decay half - life ( as we will see below ) it will mean that nuclides with @xmath124 become more unstable towards @xmath5-decay . values for the isotopic chains with @xmath83 and @xmath2 obtained with nl - sv1.,scaledwidth=60.0% ] in going above @xmath6 one sees a little peak at @xmath124 for the isotopic chains with @xmath151 , though for the chain with @xmath113 , @xmath4 shows an increasing trend from @xmath6 to @xmath124 . in terms of @xmath5-decay half - life this would imply a decreased half - life for nuclides with @xmath124 . for the isotopic chains above @xmath10 such a behaviour is not seen , but for only @xmath1 . above @xmath124 , there is a decrease in @xmath4 values for all the isotopic chains . a large amount of experimental effort to the synthesis of superheavy elements as well as to explore the nuclear properties of superheavy nuclei by the lawrence berkeley national laboratory , berkeley , gsi , darmstadt and the joint institute for nuclear research , dubna . binding energies ( masses ) , @xmath4 values and @xmath5-decay half - lives of many nuclei have been determined or estimated . naturally , such data is accessible only for light mass she especially with @xmath152 . only few data exist for @xmath9 and @xmath10 . values obtained with nl - sv1 for the isotopic chains with @xmath153 and 108 for the neutron numbers ranging from @xmath2 . the experimental values taken from the systematics of audi - wapstra @xcite are shown for comparison . , scaledwidth=85.0% ] we show in figs . [ fig : fig9 ] and [ fig : fig10 ] , @xmath4 values obtained with nl - sv1 . these are compared with @xmath4 values taken from 2003 compilation of systematics by audi - wapstra @xcite . as seen in fig . [ fig : fig9 ] , nl - sv1 values underestimate the empirical @xmath4 values for @xmath154 slightly . the disagreement was a bit more pronounced for @xmath113 . for @xmath115 the agreement of nl - sv1 values with the data is reasonably good . for the case of @xmath113 and @xmath115 , the kink in @xmath4 values at @xmath155 in the experimental data is reproduced well by nl - sv1 . for the chain of @xmath85 , though nl - sv1 values are close to the experimental data , the kink at @xmath155 in the experimental data is not reproduced by nl - sv1 . this may be due to complex shell structure around @xmath85 than that presented by the mean field in the rmf . in fig . [ fig : fig10 ] the nl - sv1 values show agreement with several data points , whereas there is a discrepancy of more than 1 mev for the data points of @xmath156 and @xmath157 . such differences may arise from a different evolution of shape in reality than predicted by the theory . in contrast , the nl - sv1 values show a very good agreement with experimental data for she chain with @xmath9 . it may however be mentioned that for @xmath9 the empirical data available is limited to four data points . the number of data points available for @xmath10 is limited to two points only . the trend of these two data points is represented quite well by the theoretical values , though the discrepancy for data point at @xmath91 is little more than 1 mev . given the complex picture of superheavy region that evolves due to interplay between deformation and single - particle states , the theoretical description of q@xmath149 values as seen in figs . [ fig : fig9 ] and [ fig : fig10 ] can be considered as satisfactory . values obtained with nl - sv1 for the isotopic chains of @xmath158 and 114 for the neutron numbers @xmath2 . the experimental values taken from the systematics compiled by audi - wapstra @xcite are shown for comparison.,scaledwidth=85.0% ] using the @xmath4-values , @xmath5-decay half - lives can be calculated using the phenomenological formula of viola and seaborg as given by , @xmath159 where @xmath160 is the atomic number of the parent nucleus and @xmath4 is the @xmath5-decay energy in mev . then , @xmath161 is obtained in seconds . the parameters @xmath162 and @xmath163 have been fitted to @xmath5-decay half - lives of known nuclei over a large range . this includes new experimental data obtained in the last decade . the values of @xmath164 , @xmath165 , @xmath166 and @xmath163 have been obtained by sobiczewski et al . @xcite . the values of these parameter are @xmath167 , @xmath168 , @xmath169 , and @xmath170 . using the viola and seaborg formula of eq . ( [ eqn:3.5 ] ) , we have calculated the @xmath5-decay half - lives @xmath161 for nuclides in all the isotopic chains . the results are shown in fig . [ fig : fig11 ] . as can be seen from eq . ( [ eqn:3.5 ] ) , @xmath171 is inversely proportional to @xmath172 . the decreasing trend of @xmath4 with neutron number as shown in fig . [ fig : fig8 ] translates into an increasing trend for @xmath171 . as seen in fig . [ fig : fig11 ] , @xmath171 values show a rather linear increase with neutron number in the region @xmath26 to @xmath88 for all the isotopic chains with the exception of the heaviest she @xmath143 . the latter show a linear behaviour in the curves in the neutron region @xmath173 . above @xmath92 the slope exhibits a further increase in the value of @xmath171 . a nearly linear increase of @xmath171 with neutron number implies an exponential increase in @xmath161 in going to higher neutron numbers . the half - life curve shows the highest values for the isotopic chain @xmath0 . as the @xmath160 value of an isotopic chain increases , the half - life curves show a gradual decrease . this shows that in going to heavier she , nuclides become less and less unstable . values for the isotopic chains with @xmath83 and @xmath2 obtained with nl - sv1.,scaledwidth=60.0% ] commensurate with q@xmath149 values in fig . [ fig : fig8 ] , one notices a few peak - like structures in half - lives at certain neutron numbers . a small peak - like structure can be seen at the neutron number @xmath139 in nearly all the isotopic chains with the exception of the heaviest chains @xmath90 and @xmath1 . the increased @xmath161 at @xmath139 can be understood in terms of a deformed shell gap at this neutron number . a similar behaviour ( feature ) has also been predicted by non - relativistic macroscopic - microscopic calculations @xcite . another minor enhancement in @xmath161 can be seen in the vicinity of @xmath174 for the isotopic chains for the lightest she @xmath95 . the region of @xmath175 has also been predicted to be a deformed shell @xcite . a major enhancement in @xmath161 values can be conspicuously seen near @xmath176 . a peak like structure in @xmath171 curves can be seen near @xmath145 for the isotopic chains of she with @xmath95 . apparently , @xmath116 exhibits a higher value of @xmath161 as compared to @xmath6 . in going to @xmath124 , there is a sudden decrease in @xmath161 values by a few orders of magnitude especially for lighter she with @xmath154 . for the chains with @xmath8 , there is also a decrease in @xmath161 at @xmath124 . this decrease is , however , smaller than for lighter she . thus , for she with @xmath106 , the region of neutron number @xmath6 denotes a region of an enhanced stability against @xmath5-decay . as one goes above @xmath10 to heavier she , the @xmath161 values show a significant decrease by several orders of magnitude . nonetheless , she with @xmath89 to @xmath1 still show a maximum near @xmath177 . this implies that even for the heavier she the region of @xmath145 shows a enhanced stability as compared to their lighter neutron counterparts . thus , the region with @xmath6 would be suitable for synthesizing heavier superheavy elements . for the case of @xmath1 isotopic chain , there is also a peak in the @xmath161 curve in the fig . [ fig : fig11 ] at @xmath91 . this indicates that the combination @xmath1 and @xmath91 may be suitable for synthesizing the @xmath1 superheavy element . a potential energy landscape represents the total energy of a nucleus as a function of deformation . it reflects how a nucleus rearranges its single - particle structure under the influence of deformation . we have carried out constrained rmf calculations with the lagrangian set nl - sv1 . the quadrupole deformation @xmath64 is used as a constraint parameter i.e , the binding energy of a nucleus is calculated for a given deformation . the rmf calculations are carried out for each value of @xmath64 in the deformation plane with a small grid in @xmath64 . rf@xmath178 obtained with nl - sv1 . , scaledwidth=65.0% ] obtained with nl - sv1 . , scaledwidth=65.0% ] obtained with nl - sv1 . , scaledwidth=65.0% ] we show in fig . [ fig : fig12 ] the potential energy surface ( pes ) for a nuclide of the lighter superheavy element @xmath179rf ( @xmath113 ) with the neutron number @xmath91 . two valleys can be seen , one in an oblate region and another in the prolate region . this picture is a clear reflection of a prolate - oblate shape co - existence whereby the two minima are nearly degenerate in energy . the actual ground state in this case is , however , prolate as shown by the slightly lower valley in the prolate region . the pes for a heavier she for the nuclide @xmath180 ( @xmath89 ) with the neutron number @xmath6 is shown in fig . [ fig : fig13 ] . this figure illustrates a shape co - existence between a spherical and an oblate shape , as has been indicated by a blue square in fig . [ fig : fig5 ] . however , here the ground state is spherical in shape . this is consistent with @xmath6 still retaining some magicity in going to @xmath89 . we show in fig . [ fig : fig14 ] the potential energy landscape for another nuclide with @xmath90 and @xmath7 . it shows a clear minimum for a spherical shape ( @xmath64=0 ) . there is no shape co - existence with any other shape for this nucleus . the spherical shape at @xmath7 reflects the character of a shell closure in the midst of deformed nuclei in a very heavy superheavy region . a spherical shape for @xmath7 in the deformed region is accentuated by the proximity of its proton number to the presumed magic number @xmath1 . the single - particle structure of nuclei is a good reflection of magic numbers , whereby larger shell gaps at a magic number can be seen in single - particle levels . in order to visualize some of the potential magic numbers in the region of superheavy nuclei , here we show the single - particle spectrum for a few key nuclei . isotones for @xmath160 = 112 , 114 , 116 , 118 , 120 with nl - sv1.,scaledwidth=50.0% ] as mentioned before , the neutron number @xmath7 is perceived as akin to a magic number . we show in fig . [ fig : fig15 ] neutron single - particle levels calculated for nuclides with @xmath7 for she s with @xmath104 . in all these nuclei the shell gaps at @xmath7 and at @xmath6 can be seen clearly . in these nuclei the shell gap at @xmath7 is maintained in going from @xmath9 to @xmath1 . quantitatively , there is no change as such in the shell gap at @xmath7 in going from @xmath9 to @xmath1 . however , the gap itself is pushed down in energy in reaching @xmath1 . thus , the neutron number @xmath7 would be helpful in synthesizing the superheavy elements with @xmath90 and @xmath1 . = 112 , 114 , 116 , 118 , 120 and @xmath6 with nl - sv1.,scaledwidth=50.0% ] in fig . [ fig : fig16 ] we show the neutron single - particle spectrum for nuclides with @xmath6 for she s with @xmath104 . the shell gap at @xmath6 can be seen clearly . thus , @xmath6 manifests as a major magic number . this is consistent with several other theoretical predictions . in the rmf theory with nl - sv1 , the shell gap at @xmath6 shows a decrease in going from @xmath9 to @xmath1 . we also see a major shell gap at @xmath181 . however , it may be of little practical importance in synthesis of superheavy elements due to very large number of neutrons involved . with @xmath128 = 172 , 184 and 198 using nl - sv1.,scaledwidth=50.0% ] in order to visualize the single - particle structure of she @xmath1 , we show in fig . [ fig : fig17 ] , the proton single - particle levels obtained with nl - sv1 for nuclides with @xmath7 , @xmath6 and @xmath181 . the nuclide with @xmath1 and @xmath7 is very close to being unbound . the proton shell gap at @xmath1 can be seen for @xmath7 , though it couples to continuum . from a practical point of view , the nuclide with @xmath1 and @xmath7 , which shows the character of a doubly magic nucleus can be useful in synthesizing she with @xmath90 and @xmath1 . the combination @xmath1 and @xmath6 also comes out as a doubly magic configuration . it may be possibly of use in synthesizing the heaviest superheavy nuclei . we have studied the structure of superheavy nuclei within the framework of the relativistic mean - field theory . the region of superheavy elements from @xmath0 to @xmath1 has been explored . the isotopic chains of the superheavy elements encompassing the neutron number from @xmath26 to @xmath27 have been investigated . the lagrangian model nl - sv1 with the inclusion of the vector self - coupling of the @xmath3-meson has been employed . rmf calculations for a large number of nuclides in the isotopic chains of superheavy elements have been performed by taking an axially symmetric deformed configuration . the binding energies , deformation properties and single - particle levels have been obtained from rmf+bcs minimizations . nuclides in general are found to possess a moderate oblate or a prolate deformation . on the other hand , nuclei in the vicinity of the neutron number @xmath6 display a spherical shape . display of a spherical shape for all the isotopic chains indicates an inclination of @xmath6 being toward a potential magicity . it is shown that nuclei in the region of superheavy elements exhibit the phenomenon of shape co - existence . a large number of nuclei much below @xmath6 exhibit a shape co - existence between a prolate and an oblate shape . this is specially the case for the isotopic chains from @xmath0 to @xmath10 . some isotopic chains exhibit a shape - coexistence between a spherical and an oblate shape especially those in the vicinity of @xmath6 . the shape - coexistence in nuclei is demonstrated by the potential energy landscapes obtained for a few nuclei . the @xmath4 values and the corresponding @xmath5-decay half - lives were obtained using the ground - state binding energies . the @xmath5-decay half - lives @xmath161 demonstrate clearly that in the vicinity of @xmath6 nuclides have a significantly larger @xmath161 as compared to their neighbours . this is again a strong indicator of the magicity of the neutron number @xmath6 . two - neutron separation energies of nuclides were calculated using the results with nl - sv1 . the @xmath126 values exhibit a small kink near @xmath182 and @xmath7 . this suggests that the neutron number @xmath182 and @xmath7 have a larger shell gap than their neighbours . in comparison , a strong kink in @xmath126 values at @xmath6 is a clearest indication of a major shell gap at @xmath6 . the single - particle levels obtained with the lagrangian model nl - sv1 show that there exist shell gaps at neutron numbers @xmath7 and @xmath6 . whilst @xmath7 can not be construed as a major magic number as displayed by a mild kink in @xmath126 values , the neutron number @xmath6 does behave as a major magic number in the neutron - rich region . evidently , the magicity of @xmath6 is demonstrated succintly by a large shell gap at @xmath6 seen in the single - particle levels . this picture of the magicity of n=184 in the rmf theory is consistent with the predictions in the density - dependent skyrme hartree - fock approach and also with the previous studies performed within the rmf theory using the conventional model of the nonlinear scalar self - couplings . thus , our study with the lagrangian model nl - sv1 with the vector self - coupling of @xmath3 meson reinforces the magic character of the neutron number @xmath6 . albeit , a decrease in the shell gap at @xmath6 in going towards @xmath1 implies a weakening of the shell strength in approaching the extreme end of the periodic table . whilst a combination of @xmath6 in conjuction with @xmath160 in the vicinity of 120 produces nuclei which have impractically large number of neutrons , the combination of the neutron number @xmath7 and @xmath1 does lie within the domain of feasibility . our study shows that this combination is suggestive of a double magicity . thus , synthesis of superheavy elements heavier than @xmath89 with neutron number close to @xmath7 should be susceptible to an extra stability rendered by the apparent magic nature of @xmath1 and @xmath7 . this work is supported by the graduate research project y01/05 of the research administration , kuwait university . useful discussions with professor g. mnzenberg are thankfully acknowledged . hofmann s , hessberger f p , ackermann d , mnzenberg g , antalic s , cagarda p , kindle , b , kojuharova j , leino m , lommel b , mann r , popeko a g , reshitko s , saro s , uusitalo j and jeremin a v 2002 _ eur . phys . j. _ * a14 * 147 ginter t n , gregorich k e , loveland w lee , d m kirbach , u w , sudowe r , folden iii c w , patin j b , seward n , wilk p a , zielinski p m , aleklett k , eichler r , nitsche h and hofmann d c 2003 _ phys . c _ * 67 * 064609 oganessian yu ts , utyonkov v k , lobanov yu a , abdullin f sh , polyakov a n , shirokovski i v , tsyganov yu s , gulbekian g g , bogomolov s l , gikal b n , mezentsev a n , iliev s , subbotin , v g , sukhov , a m , buklanov , g v , subotic , k , itkis , m g , moody , k j , wild , j f , stoyer , n j , stoyer , m a , and lougheed r w 1999 _ phys . * 83 * 3144 oganessian yu ts , utyonkov vk , lobanov , yuv , abdullin , f sh , polyakov , a n , shirokovski , i v , tsyganov , yu s , gulbekian , g.g . , bogomolov , s.l . , gikal , b.n . , mezentsev , a.n . , iliev , s , subbotin , v g , sukhov , a m , ivanov , o v , buklanov , g v , subotic , k , itkis , m g , moody , k j , wild , j f , stoyer , n j , stoyer , m a , and lougheed , r w 2000 _ phys . c _ * 6 * ( r ) 041604 oganessian , yu ts , utyokoy , v k , lobanov , yu v abdullin , f sh , polyakov , a n , shirokosvsky , i v , tsyganov , yu s , gulbekian , g g , bogomolov , s l , gikal , b n , mezentsev , a n , iliev , s , subbotin , v g , sukhov , a m , ivanov , o v , buklanov , g v , subotic , k , itkis , m g , moody , k j , wild , j f , stoyer , n j , stoyer , m a , lougheed , r w , laue , c a , karelin , ye e and tatarinov , a n 2000 _ phys . c _ * 63 * ( r ) 011301 oganessian , yu ts utyonkov , v k , lobanov , yu a , abdullin , f sh polyakov , a n , shirokovski , i v , tsyganov , yu s , gulbekian , g g , bogomolov , s l , gikal , b n , mezentsev , a n , iliev , s , subbotin , v g , sukhov , a m , voinov , a a , buklanov , g v , subotic , k , zagrebaev , v i , itkis , m g , patin , j b , moody , k j , wild , j f , stoyer , m a , stoyer , n j , shaughnessy , d a , kennally , j m and lougheed , r w 2004 _ phys . c _ * 69 * 054607 ogasessian , yu ts , utyokoy , v k , lobanov , yu v , abdullin , f sh , polyakov , a n , shirokosvsky , i v , tsyganov , yu s , gulbekian , g g , bogomolov , s l , mezentsev , a n , iliev , s , subbotin , v g , sukhov , a m , voinov , a a , buklanov , g v , subotic , k , zagrabaev , v i , itkis , m g , patin , j b , moody , k j , wild , j f , stoyer , m a , stoyer , n j , shaughnesy , d a , kenneally , j m , and lougheed , r w 2004 _ phys . c _ * 69 * ( r ) 021601 wilk p a , gregorich k e , terler a , laue c a , eicher r , ninov v , adams j l , kirbach u w , lane m r , lee d m , patin j b , shaughnessy d a , strellis d a , nitsche h , hoffman d c 2000 _ phys . rev . * 85 * 2697

we report molecular line and dust continuum observations , made with the sest telescope , towards four young high - mass star forming regions associated with highly luminous ( ) iras sources ( 15290 - 5546 , 15502 - 5302 , 15567 - 5236 and 16060 - 5146 ) . molecular emission was mapped in three lines of cs ( j=2 , 3 and 5 ) , two lines of sio ( j=2 and 3 ) , two rotational transitions of choh ( j=3 and ) , and in the cs(j=3 ) line . in addition , single spectra at the peak position were taken in the co(j=1 ) ,co(j=1 ) and co(j=1 ) lines . we find that the luminous star forming regions are associated with molecular gas and dust structures with radii of typically 0.5 pc , masses of m , column densities of , molecular hydrogen densities of typically and dust temperatures of k. the 1.2 mm dust continuum observations further indicate that the cores are centrally condensed , having radial density profiles with power - law indices in the range . we find that under these conditions dynamical friction by the gas plays an important role in the migration of high - mass stars towards the central core region , providing an explanation for the observed stellar mass segregation within the cores . the cs profiles show two distinct emission components : a bright component , with line widths of typically 5 ( fwhm ) , and a weaker and wider velocity component , which extends up to typically from the ambient cloud velocity . the sio profiles also show emission from both components , but the intensity of the pedestal feature relative to that of the bright component is stronger than for cs . the narrow sio component is likely to trace warm ambient gas close to the recently formed massive stars , whereas the high velocity emission indicates mass outflows produced by either the expansion of the hii regions , stellar winds , and/or collimated outflows . we find that the abundances of cs , choh and sio , relative to h , in the warm ambient gas of the massive cores are typically , , and , respectively .

0912.2655

during the last decades several observational efforts have been carried out to determine the characteristics of the molecular gas and dust associated with massive star forming regions . surveys of molecular emission in high density tracers ( plume et al . 1992 ; juvela 1996 ; plume et al . 1997 ; shirley et al . 2003 ) and of dust continuum emission ( beuther et al . 2002 ; mueller et al . 2002 ; fandez et al . 2004 ; williams et al . 2004 ; garay et al . 2007 ) , made with single dish telescopes , show that the maternities of high - mass stars are regions of molecular gas with distinctive physical parameters , with typical radii of 0.4 pc , densities of @xmath22 @xmath14 and masses of @xmath23 @xmath24 . the dust observations also show that these regions have dust temperatures of typically @xmath25 k , indicating that a luminous energy source has been already formed inside them . these structures , also refereed as massive and dense cores , have typically column densities of @xmath26 @xmath12 ( garay et al . 2007 ) which make them very dark at optical wavelengths and even dark at infrared wavelengths ( menten et al . 2005 , rathborne et al . the iras luminosity of most regions in these studies are in the range @xmath27 @xmath28 . in order to investigate possible differences , in the physical and chemical characteristics as well as in the stellar population , between massive and dense cores of different luminosities , we have carried out a multi - wavelength study of four highly luminous iras sources in the southern hemisphere selected from the list of iras point sources with colors of uc hii regions and detected in cs(2@xmath11 ) emission compiled by bronfman et al . the observed objects are listed in table [ tbl - obssources ] 2 , 3 , and 4 give , respectively , the name , right ascension , and declination of the associated iras source . the kinematic distance to the objects is given in col . the four objects are located in the tangent region of the norma spiral arm and therefore there is little ambiguity in the kinematic determination of their distances ; their @xmath29 s are nearly the maxima allowed by pure circular motion in their lines of sight . the total far infrared luminosities , @xmath30 , are given in col.6 . the objects have @xmath30 greater than @xmath31 @xmath28 and their cs(2@xmath11 ) line profiles show broad molecular line wing emission suggesting the presence of powerful molecular outflows . here we report molecular line and dust continuum observations towards these four regions . the aim of these observations is to undertake a detailed study of the molecular gas and dust content of luminous massive and dense cores in order to determine their physical conditions , kinematics , and chemistry . molecular emission was mapped in a number of species that are commonly used as diagnostics of the conditions of massive and dense cores : carbon monosulfide , silicon monoxide , and methanol ( e.g. , zinchenko et al . 1995 ; juvela 1996 ) . thermal emission of sio is both a good probe of high velocity gas in regions of massive star formation ( downes et al . 1982 ) and a tracer of high - temperature chemistry ( ziurys et al . the methanol abundance is highly enhanced in the hot gas surrounding infrared sources ( menten et al . 1988 ) as well as in shocked gas ( liechti & walmsley 1997 ) , thus thermal methanol emission is an excellent probe of the environments of high - mass yso s . the cs and dust continuum observations are intended to determine the density structure of the cores . in a subsequent paper ( chavarra et al . 2009 ; paper ii ) we will present the characteristics of the stellar content within these regions . the molecular line and dust continuum observations were carried out using the 15-m swedish - eso submillimetre telescope ( sest ) located on la silla , chile . the molecules , transitions , and frequencies observed and the instrumental parameters are summarized in table [ tbl - obspar ] . the cs(@xmath32 ) observations were made between december 1990 and january 1992 . during this period the telescope was equipped with a cooled schottky mixer and an acousto - optical spectrometer with 2000@xmath3343 khz channels . the typical system temperature above the atmosphere was 500 k. the cs(@xmath32 ) line emission was mapped with spacings of 45 within regions of at least 135@xmath33135 . the observations were done in the frequency switch mode with a frequency throw of 15 mhz . the sio(@xmath32 ) , sio(@xmath34 ) , ch@xmath2oh(@xmath5 ) , ch@xmath2oh(@xmath35 ) , cs(@xmath34 ) and c@xmath6s(@xmath34 ) observations were made during march of 1997 . in this epoch the telescope was equipped with sis receivers , and it was possible to simultaneously observe lines at 2 and 3 mm wavelengths . single - sideband receiver temperatures were typically 120 k for both receivers . as backend we used high resolution acousto - optical spectrometers providing a channel separation of 43 khz and a total bandwidth of 43 mhz . the emission in the above lines was mapped within regions of at least 90@xmath33 90 , with 30 spacings . the observations were performed in dual beam - switching mode , with a beam separation of 1147 in azimuth . the integration times on source per position and the resulting rms noise in antenna temperature are given in table [ tbl - tinrms ] . within the available bandwidths , three rotational lines of ch@xmath2oh could be observed at 2 mm ( a@xmath36 , e , and e lines ) and four rotational lines at 3 mm ( e , e , a@xmath36 , and e lines ) . during march 2001 we further observed the sio ( 217104.935 ghz ) and ch@xmath2oh(@xmath37 ) ( 84521.21 ghz ) lines toward the center of the g330.949 - 0.174 core . the receivers and backends were the same as described above . single - sideband receiver temperatures were 1200 k at 217 ghz and 190 k at 85 ghz . the cs(@xmath38 ) observations of g324.201 , g329.337 , and g330.949 were made between september 1991 and january 1992 . in this epoch the telescope was equipped with a cooled schottky mixer , and an acousto - optical spectrometer with 2000@xmath3343 khz channels ; however , a setup of 1000@xmath3386 khz resolution elements was used . the typical system temperature above the atmosphere was 1850 k. the cs(@xmath38 ) line emission was mapped with spacings of 15 within regions of at least 60@xmath3345 . the observations were done in the frequency switch mode with a frequency throw of 15 mhz . the cs(@xmath38 ) observations of g328.307 were made in may 2001 , with an sis receiver and the acousto - optical spectrometer with 2000@xmath3343 khz channels . typical system temperature above the atmosphere was of 450 k. observations were done in dual beam switching mode , with a beam separation of 1147 in azimuth , in a grid of 9 positions with spacings of 45 . observations of the co , @xmath7co , and c@xmath8o emission in the j=1@xmath10 lines were made during may , 2000 and june , 2001 . spectra were obtained only toward the peak position of the cs emission in each region . these observations were performed in the position switched mode , with off positions taken from bronfman et al . the goal of these observations was to obtain the data needed to determine co column densities , which in turn will permit to derive abundance enhancements relative to co. the 1.2-mm continuum observations were made using the 37-channel sest imaging bolometer array ( simba ) during october , 2001 and august , 2003 . the passband of the bolometers has an equivalent width of @xmath39 ghz and is centered at @xmath40 ghz . the hpbw of a single element is 24 and the separation between elements on the sky is 44 . we observed in the fast mapping mode , using a scan speed of 80 s@xmath41 . each observing block consisted of 50 scan lines in azimuth of length 800and separated in elevation by 8 , giving a map size of 400 in elevation . this block required @xmath42 15 minutes of observing time . we observed @xmath43 blocks per source , achieving rms noise levels of typically @xmath44 mjy beam@xmath41 . the data were reduced according to a standard procedure using the software package mopsi , which included baseline subtraction and rejection of correlated sky - noise . flux calibration was performed using a sky - opacity correction and a counts - to - flux conversion factor derived from maps of uranus . uncertainties in the absolute calibration and pointing accuracy are estimated at @xmath45 and 3 , respectively . figs . [ fig - specmapg324 ] to [ fig - specmapg330 ] show spectral maps of the cs(2@xmath11 ) , cs(3@xmath12 ) , c@xmath6s(3@xmath12 ) , sio(2@xmath11 ) , sio(3@xmath12 ) , ch@xmath2oh(3@xmath4 ) , and ch@xmath2oh(2@xmath46 ) line emission observed toward g324.201 + 0.119 , g328.307 + 0.423 , g329.337 + 0.147 and g330.949 - 0.174 , respectively . the spacing is 30 for all transitions , except cs(2@xmath11 ) for which the spacing is 45 . offsets are from the reference position given in columns 3 and 4 of table [ tbl - obssources ] . table [ tbl - linepar ] gives the parameters of the line emission observed at the peak position ( cols . 2 to 4 ) and of the spatially averaged emission ( cols . 5 to 7 ) , determined from gaussian fits to the respective line profiles . carbon monosulfide was detected in all observed transitions towards all sources . the cs(2@xmath11 ) and cs(3@xmath12 ) line profiles indicate the presence of two emission components : ( i ) a bright , spatially extended , and relatively narrow velocity feature ; and ( ii ) a weaker and broader velocity feature , which is usually spatially confined to the central region . the line widths of the narrow component emission , which most likely arises from the dense ambient gas , are typically about 5 , a value considerably larger than the thermal width expected for kinetic temperatures of @xmath4240 k , of @xmath47 , implying that the line widths are dominated by non - thermal supersonic motions . whether the non - thermal motions are due to the effect of stellar winds or are intrinsic to the initial conditions of the formation process of massive stars is still an open question . the velocity range of the broad wing emission ( full width at zero power ) in different transitions are given in table [ tbl - wingpar ] . the broad emission , which has typically full widths at zero power of @xmath48 , is mainly detected toward the peak position of the massive cores . although this emission most likely arises from outflowing gas , the coarse angular resolution ( @xmath49 ) of our observations does not allow to determine its spatial distribution and degree of collimation . the nature of the high velocity gas giving rise to the pedestal feature is thus difficult to assess . higher angular resolution observations are needed to investigate whether the high velocity emission arises from a collimated bipolar outflow driven by the recently formed massive stars or from an spherically symmetric structure possible driven either by stellar winds or by the expansion of the hii region onto the ambient medium . silicon monoxide was detected toward the four observed high - mass star forming regions , the strongest emission being observed toward the g330.949 - 0.174 core . sio lines are tracers of both the warm ambient gas surrounding young luminous stars and of molecular outflows associated with newly formed massive stars . harju et al . ( 1998 ) have already shown that sio emission is relatively easy to detect toward regions of high - mass star formation , with the detection rate increasing with the fir luminosity . the profiles of the sio emission show a broad component of high velocity gas superposed on a narrower component , whose central velocity is close to the velocity of the ambient gas . the full widths at zero power ( emission above a @xmath50 level ) of the broad component range from 17.4 to 36.6 , with a median of 26 . the relative contribution to the observed intensity is different from source to source . in g324.201 the sio emission from the broad component dominates the spectrum , whereas in g330.949 the narrower emission is stronger . we suggest that , on the one hand , the narrow sio emission arises from the warm ambient gas surrounding the recently formed massive stars , and that this component results from the evaporation of silicon compounds from grain mantles produced by the stellar radiation . on the other hand , the broad sio emission most likely indicates mass outflows from a star or stars embedded within the massive core , and that this component results from grain disruption by shock waves . hence , the sio emission traces both high kinetic temperature ambient gas and outflowing shocked gas . the shape of the profiles are different from molecule to molecule depending on the relative intensities between the narrow and broad components . to illustrate in more detail the differences between the line profiles , figs . [ fig - specpeakg324 ] to [ fig - specpeakg330 ] show the spectra of all molecular lines , except methanol , observed at the peak position of each region . the cs(5@xmath14 ) spectra corresponds to the average spectra over the 30@xmath3330 mapped region , and therefore it is directly comparable to the peak spectra observed in the other transitions with @xmath4230 beams . [ fig - csmaps ] shows contour maps of the velocity integrated ambient gas emission in the cs(2@xmath11 ) line toward all four regions . the 50% contour level of the emission is marked with dark lines . in all cases the emission , at the resolution of 52 , arises form a single source exhibiting a simple morphology . the deconvolved major and minor fwhm angular sizes determined by fitting a gaussian profile to the emission are given in 3.3 . [ fig - dustsimba ] presents maps of the 1.2-mm continuum emission , within regions of typically @xmath51 in size , towards the four observed sources . emission was detected toward all four regions . the extent and morphology of the 1.2-mm dust continuum emission is similar to that of the cs line emission , indicating that these two probes trace the same physical conditions . the observed parameters of the 1.2-mm sources are given in table [ tbl - obspardust ] . cols.(2 ) and ( 3 ) give their peak position . cols.(4 ) and ( 5 ) give , respectively , the peak flux density and the total flux density , the later measured directly from the maps using the aips tasks imean . col.(6 ) gives the deconvolved major and minor fwhm angular sizes determined by fitting a single gaussian profile to the whole spatial distribution , except for g329.337 for which two gaussian components were fitted . in three of the four regions the morphology of the dust emission can be described as arising from a bright compact peak surrounded by an extended envelope of weaker emission , which is characteristic of sources with centrally condensed density profiles . this is illustrated in fig . [ fig - intcuts ] which shows slices of the observed 1.2-mm intensity across the two sources with nearly circular morphology ( g324.201 and g330.949 ) . we find that the observed radial intensity profiles can be well fitted with single power - law intensity profiles of the form @xmath52 , where @xmath53 is the distance from the center , properly convolved with the beam of 24 . we derive @xmath54 indices of 1.6 and 1.9 for the g324.201 and g330.949 cores , respectively . for the g328.307 core an @xmath54 index of 1.5 has been already derived by garay et al . ( 2007 ) . three of the high - mass star forming regions investigated here , g324.201 , g329.337 and g330.949 , are in the list of @xmath422000 massive young stellar objects ( myso ) candidates compiled by the rms survey based on a colour selection criteria of msx point sources ( hoare et al . radio continuum observations have , however , shown that all four norma cores are associated with compact radio sources , indicating that they are in a more advanced stage of evolution in which an uchii region has already developed ( walsh et al . 1998 ; garay et al . 2006 ; urquhart et al . 2007 ) . in all cases the compact radio sources are found projected at the peak position of the 1.2-mm dust continuum emission , suggesting that massive stars are located at the center of the massive and dense cores . in what follows we discuss the characteristics of the molecular gas and dust emission , as well as of the radio emission , toward each of the iras sources , individually . _ g324.201 + 0.119_.@xmath55 the g324.201 + 0.119 massive core is associated with the iras source 15290 - 5546 , which has a total luminosity of @xmath56 . the cs profiles across the core show emission from a nearly gaussian component , with central velocities of @xmath57 and fwhm widths of @xmath58 . the spatial distribution of the emission in the cs(@xmath32 ) line from the ambient gas is slightly elongated , with fwhm deconvolved major and minor axis of 43 and 30 , respectively . the morphology of the 1.2-mm dust emission also shows the presence of a single structure with fwhm deconvolved major and minor axis of 35 and 28 , respectively . a power - law fit to the 1.2-mm intensity profile gives an index of 1.6 ( see fig . [ fig - intcuts ] ) . the total flux density of the region at 1.2 mm is 17.3 jy . toward the center of the core the gaussian component is superposed on a broader and weaker component . in the cs(@xmath34 ) line the wing emission extends up to a flow velocity of @xmath59 in the blue side and 11.4 in the red side . the flow velocity , @xmath60 , is defined as @xmath61 , where @xmath62 is the ambient cloud velocity . the sio line profiles are , on the other hand , dominated by the emission from the broad component , covering a flow velocity range from @xmath63 to @xmath64 . radio continuum observations towards g324.201 + 0.119 show the presence of two compact radio sources : a cometary - like component , with an angular diameter of @xmath424 `` , located at the center of the core and an unresolved component located @xmath426 '' north of the former ( urquhart et al . if excited by single stars , the rate of uv photons needed to produce the cometary and unresolved regions of ionized gas implies the presence of o7 and o8.5 zams stars , respectively . _ g328.307 + 0.423_.@xmath55 the g328.307 + 0.423 massive core is associated with the iras source 15502 - 5302 , which has a total luminosity of @xmath65 . the spatial distribution of the cs(@xmath32 ) emission from the ambient gas shows an elongated morphology , with fwhm deconvolved major and minor axis of 65 and 37 , respectively . the elongated morphology is more clearly seen in the observed spatial distribution of the 1.2-mm emission ( see fig . [ fig - dustsimba ] ) , which has fwhm deconvolved major and minor axis of 43 and 26 , respectively . the total flux density of the region at 1.2-mm is 24.1 jy . the co(@xmath66 ) spectra observed at the center of the core show a double peaked profile , with a strong blueshifted peak at the velocity of @xmath67 and a weaker redshifted peak at the velocity of @xmath68 . on the other hand , the spectra of the optically thin c@xmath8o(@xmath66 ) line show a single gaussian profile with a peak velocity of -92.6 . these spectral characteristics indicate the presence of infall motions ( leung & brown 1977 , walker , narayanan , & boss 1994 , myers et al . 1996 ) . using expression ( 9 ) of myers et al . ( 1996 ) we derive an infall velocity of @xmath69 . the profiles of the cs and c@xmath6s emission observed across the core are generally asymmetric , showing a peak at velocities of @xmath70 and a shoulder toward redshifted velocities . the observed shapes are consistent with the infall motions suggested by the co(@xmath66 ) emission but for transitions with smaller optical depths . in addition to infalling gas , outflowing gas is clearly detected toward the central region . in the cs(3@xmath12 ) line , blueshifted and redshifted emissions are detected up to flow velocities of @xmath71 and 12.2 , respectively . the sio emission , which is detected in the velocity range from @xmath72 to @xmath73 , arises mainly from the outflow component , but a narrower component from the ambient core gas is also present . radio continuum observations towards g328.307 + 0.423 show the presence of a complex region of ionized gas , consisting of a handful of low brightness , extended components and a bright , compact component located at the center of the core ( garay et al . the flux density of the later component increases with frequency whereas its angular size decreases with frequency . these characteristics suggest the presence of a steep gradient in the electron density . _ g329.337 + 0.147_.@xmath55 the g329.337 + 0.147 massive core is associated with the iras source 15567 - 5236 , which has a total luminosity of @xmath74 . the cs profiles across the core show emission from a bright component with a central velocity of @xmath75 and a fwhm width of @xmath76 . strong high velocity wing emission is detected toward the central region in all observed species ( see fig . [ fig - specpeakg329 ] ) . the blueshifted and redshifted emissions extend up to flow velocities of @xmath77 and @xmath78 in the co(1@xmath10 ) line , @xmath79 and @xmath64 in the cs(3@xmath12 ) line , and @xmath80 and 17.4 in the sio(3@xmath12 ) line . the cs(@xmath32 ) emission from the ambient gas arises from a structure that is slightly elongated in the se direction , with deconvolved fwhm major and minor axis of 44 and 39 , respectively . the peak of the cs emission is coincident with a methanol maser ( walsh et al . the spatial distribution of the silicon monoxide and methanol emission show notable differences with respect to that of the emission in carbon monosulfide . the morphology of the sio and ch@xmath2oh emission shows a double peaked structure , with the main peak being displaced from the peak of the cs emission by @xmath42 60 toward the southeast ( see fig . [ fig - mapsg329 ] ) . the secondary sio peak coincides with the peak of the cs emission . the double peaked morphology is also clearly seen in the map of the dust continuum emission . radio continuum observations towards g329.337 + 0.147 show the presence of a single source , with an angular diameter of @xmath81 " , exhibiting an irregular multi - peaked morphology ( walsh et al . 1998 ; urquhart et al . 2007 ) . if excited by a single star , the rate of uv photons required to ionize this region implies the presence of an o6.5 zams star . _ g330.949 - 0.174_.@xmath55 the g330.949 - 0.174 massive core is associated with the iras source 16060 - 5146 , which has a total luminosity of @xmath82 . the cs(2@xmath11 ) line exhibit an asymmetric profile , with a peak at the velocity of @xmath83 and a shoulder toward redshifted velocities . the cs(3@xmath12 ) and cs(5@xmath14 ) profiles are double peaked , with peaks at @xmath84 and @xmath85 . the blue peak is stronger than the red peak by a factor of @xmath42 1.5 . the profiles of the c@xmath6s and sio emission are , on the other hand , more symmetric , with peak velocities lying between the two velocity peaks of the cs emission . this result indicates that the complex cs profiles are produced by optical depth effects rather than being due to two velocity components along the line of sight . the presence of a high velocity gas toward the central region is clearly indicated by the profiles of the cs and sio emission . in co lines the outflow emission is blended with emission from less dense molecular clouds in the line of sight . the emission in the cs(@xmath34 ) line extends up to a velocity of @xmath86 in the red side and up to a velocity of @xmath87 in the blue side . the spatial distribution of the cs(@xmath32 ) emission from the ambient gas is nearly circular , with fwhm deconvolved major and minor axis of 43 . the circular morphology is also seen in the 1.2-mm dust emission , which exhibits a single structure with fwhm deconvolved major and minor axis of 33 . the total flux density of the region at 1.2 mm is 47.2 jy . a power - law fit to the 1.2-mm intensity profile gives an index of @xmath88 ( see fig . [ fig - intcuts ] ) . the peak of the molecular and dust emissions are displaced about 30 east of the iras position , but are coincident with the position of a methanol maser and a compact radio continuum source ( walsh et al . 1997 , 1998 ) . the radio continuum observations of urquhart et al . ( 2007 ) towards g330.949 - 0.174 show the presence of two compact sources : a bright cometary - like component , with an angular diameter of @xmath423 `` , located near the center of the core and an unresolved , weak component , located @xmath4212 '' northeast of the former . if excited by single stars , the rate of uv photons needed to produce the cometary and unresolved regions of ionized gas implies the presence of o7 and b0 zams stars , respectively . [ fig - sed ] shows , for each of the 4 sources , the spectral energy distribution ( sed ) from 12 @xmath89 m to 1.2 mm . in this wavelength range the emission is mainly due to thermal dust emission . the flux densities at 12 , 25 , 60 and 100 correspond to iras data , and those at 8.3 , 12.1 , 14.7 and 21.3 correspond to _ midcourse space experiment ( msx ) _ data ( price et al . the iras and msx fluxes were obtained from the ipac database . we fitted the sed with modified blackbody functions of the form @xmath90\omega_s~ , $ ] where @xmath91 is the dust optical depth , @xmath92 is the planck function at the dust temperature @xmath93 , and @xmath94 is the solid angle subtended by the dust emitting region . the opacity was assumed to vary with frequency as @xmath95 , i.e. @xmath96 , where @xmath97 is the frequency at which the optical depth is unity . a single temperature model produced poor fits , underestimating the emission observed at wavelengths smaller than 25@xmath89 m , and therefore we used a model with two temperature components . since the regions are likely to present temperature gradients , this is indeed a coarse simplification . however , this model allows us to determine the average dust temperatures representative of each wavelength range . the parameters derived from the fits , namely dust temperature , opacity power - law index , wavelength at which the opacity is unity , and angular size ( assuming a gaussian flux distribution ) , for the colder dust component are given in fig . [ fig - sed ] . the parameters derived from the 1.2-mm observations are summarized in table [ tbl - deriveddust ] . the dust temperatures given in col . ( 2 ) correspond to the colder temperature determined from fits to the spectral energy distribution ( sed ) . the radius given in col . ( 3 ) were computed from the geometric mean of the deconvolved major and minor angular sizes obtained from single gaussian fits to the observed spatial structure . the masses of the cores , given in col.(4 ) , were derived using the expression , ( e.g. , chini , krgel , & wargau 1987 ) @xmath98 where @xmath99 is the observed flux density at 1.2 mm , @xmath100 is the mass absorption coefficient of dust with a value of 1 @xmath101 g@xmath41 ( ossenkopf & henning 1994 ) , @xmath102 is the dust - to - gas mass ratio ( assuming 10% he ) with a value of @xmath103 , and @xmath104 is the planck function at the dust temperature @xmath93 . this expression implicitly assumes that the observed emission at 1.2 mm corresponds to optically thin thermal dust emission and that the source is isothermal . the masses derived in this way range from @xmath105 to @xmath106 @xmath24 . cols . ( 5 ) and ( 6 ) give , respectively , the average molecular densities and average column densities derived from the masses and radius assuming that the cores have uniform densities . clearly , this is a rough simplification , since as discussed in 3.2 the massive and dense cores are likely to have steep density gradients . finally , the continuum optical depth at 1.2-mm is given in col . ( 7 ) . the optical depth in the cs(3@xmath12 ) line , @xmath107 , can be derived from the ratio of observed main beam brightness temperatures in the cs(3@xmath12 ) and c@xmath6s(3@xmath12 ) lines using the expression , @xmath108 } \over{[j_{32}^{\small c^{34}s } ( t_{ex})-j_{32}^{\small c^{34}s } ( t_{bg } ) ] } } { { t_{mb}(c^{34}s)}\over{t_{mb}(cs ) } } , ~ \label{eqn - optdepth}\ ] ] where the subscript 32 refers to the @xmath34 transition , @xmath109 is the excitation temperature of the transition , @xmath110 is the background temperature , @xmath111 and @xmath53 is the @xmath112 optical depth ratio , given by @xmath113 where @xmath114 is the [ c@xmath115s / c@xmath6s ] isotopic abundance ratio , and @xmath89 and @xmath116 are the permanent dipole moment and the rotational constant of the molecule , respectively . if the excitation temperature is known , the above expressions allow the optical depth to be determined . we avoid the usual approximation that @xmath117 , and solve equation ( [ eqn - optdepth ] ) using an interpolation procedure , assuming a [ c@xmath115s / c@xmath6s ] abundance ratio of 22.5 ( blake et al . 1994 ) . the optical depth in the co(@xmath66 ) line , @xmath118 , can be derived from the ratio of observed main beam brightness temperatures in the co(@xmath66 ) and @xmath7co(@xmath66 ) lines in a similar manner as described above ( see bourke et al . 1997 , for the appropriate expressions ) . we assumed a [ co/@xmath7co ] abundance ratio of 55 , corresponding to the average value across the galactic disk ( wannier et al . 1982 ) . further , the observations of the j=@xmath66 line emission in the three isotopic species of carbon monoxide ( co , @xmath7co and c@xmath8o ) allow to determine the [ @xmath7co]/[c@xmath8o ] isotopic abundance ratio . using a [ co]/[@xmath7co ] abundance ratio of 55 , we derive [ @xmath7co]/[c@xmath8o ] abundance ratios in the range 8 to 10 . the peak optical depths at the peak position of the massive cores in the cs(3@xmath12 ) , c@xmath6s(3@xmath12 ) , co(1@xmath10 ) and @xmath7co(1@xmath10 ) lines , are given in cols . 2 to 5 of table [ tbl - opacities ] . they were computed assuming @xmath109 = 40 k. we find that the emission is optically thick in the cs lines and optically thin in the c@xmath6s line for all cores . we note that the dependence of the optical depth with excitation temperature is weak . for instance , assuming @xmath109 = 100 k the derived optical depths are only 0.4% larger . the total column density @xmath119 of a linear , rigid rotor , molecule can be derived from the optical depth and excitation temperature of a rotational transition at frequency @xmath120 , using the expression ( e.g. garden et al . 1991 ) @xmath121 where @xmath122 is the rotational quantum number of the lower state . this expression assumes that all energy levels are populated according to local thermodynamic equilibrium ( lte ) at the temperature @xmath109 . in particular the total column density of the cs molecule is given , in terms of the opacity and excitation temperature of the @xmath34 transition , by @xmath123 where @xmath124 is measured in . 2 of table [ tbl - colden ] gives the derived cs column density through the center of the cores . the co column density through the center of the cores are given in col . 3 of table [ tbl - colden ] . they were determined in the same way as the cs column densities using the [ co,@xmath7co ] pair of observations . from the column density in cs , n(cs ) , we can estimate the mass of the cores using the expression , @xmath125 \mu_m~ m_h~ \pi r^2~~,\ ] ] where @xmath126 $ ] is the h@xmath18 to cs abundance ratio , @xmath127 is the mean molecular weight , and @xmath128 is the core radius . the radii , given in column 2 of table [ tbl - derivedpar ] , were computed from the geometric mean of the deconvolved cs(2@xmath11 ) angular sizes using the distances given in table [ tbl - obssources ] . the core masses computed from expression ( 6 ) , using @xmath126= 3\times10 ^ 7 $ ] ( see 4.3 ) , are given in col . 3 of table [ tbl - derivedpar ] . they range from @xmath129 to @xmath130 @xmath24 . these masses are in good agreement with those derived from the dust continuum observations . the rotational temperature , @xmath131 , and the total column density , @xmath132 , of methanol and silicon monoxide can be computed using a rotational diagram analysis ( see e.g. , linke , frerking & thaddeus 1978 ; blake et al . 1987 ) , with the assumptions of optically thin conditions and local thermodynamical equilibrium ( lte ) , which relates the integrated line intensity , rotational temperature , and column density via @xmath133 where @xmath89 , @xmath120 , and @xmath134 are the transition dipole moment , frequency , and line strength of the transition , respectively , @xmath135 is the velocity integrated main beam brightness , obtained directly from the observations , @xmath136 is the upper state energy , and @xmath137 is the rotational partition function . [ fig - trotmet ] shows rotational diagrams for the ch@xmath2oh emission observed at the peak position of the massive cores . the column densities and rotational temperatures , derived from a linear least squares fit to the data , are given in cols . 4 and 6 of table [ tbl - colden ] . the rotational temperatures of methanol range from 12.0 k to 22.6 k. these temperatures are significantly smaller than the temperatures derived from the dust observations , of @xmath42 40k , indicating that the methanol populations are sub - thermally excited ( bachiller et al . 1995 ; avery & chiao 1996 ) . from rotational diagram analysis of the sio emission observed at the peak position of the massive cores we derived column densities and rotational temperatures of both the ambient gas , given in cols . 5 and 7 of table [ tbl - colden ] , and flowing molecular gas , given in cols . 8 and 9 the rotational temperatures of the flowing gas are typically two times larger than those of the ambient gas . this is illustrated in fig . [ fig - trotsio ] , which presents a rotational diagram for the sio emission observed at the peak position of the g330.949 massive core . in this case the gas giving rise to the narrow emission has a rotational temperature of 11.2@xmath138 k and a column density of @xmath139 @xmath12 , whereas the outflowing gas producing the broad emission has a rotational temperature of 24.9@xmath140 k and a column density of @xmath141 @xmath12 . the chemical composition of a gas cloud is usually characterized by the fractional abundance of molecules relative to molecular hydrogen , the main constituent of the interstellar gas . the abundance of h@xmath18 is however not directly measured . our observations allow us , however , to directly determine abundances relative to co. we compute the abundance of species x relative to co , [ x / co ] ( where x = cs , sio , or ch@xmath2oh ) , as the ratio of the column density of species x and the co column density in the same velocity range . for sio and ch@xmath2oh we use the column densities derived from rotational analysis ( see x.x ) whereas for cs and co we use the column densities derived from the observations of their main and isotopic species ( see 4.3.1 ) . table [ tbl - abundances ] summarizes the abundances of cs , sio and ch@xmath2oh relative to co. to convert [ x / co ] to fractional abundances relative to h@xmath18 we will assume [ co / h@xmath18]=@xmath143 . * silicon monoxide*. we find that the [ sio / co ] abundance ratio in the ambient gas range from @xmath144 to @xmath145 . assuming that [ co / h@xmath18 ] is @xmath143 , then the [ sio / h@xmath18 ] abundance ratio in the ambient gas ranges between @xmath146 and @xmath147 . chemical models predict that both high temperature and shocks can produce enhancements in the gas phase sio abundance . the abundance of sio in the ambient gas is likely due to the gas - phase high temperature chemistry that takes place in the hot regions surrounding the recently formed massive stars . models of gas - phase chemistry of hot molecular cores that takes into account grain mantle evaporation by stellar and/or shock radiation ( mackay 1995 ) , predict [ sio / h@xmath18 ] abundances of @xmath146 for gas kinetic temperatures of 100 k and densities of @xmath148 @xmath14 . we also find that the [ sio / co ] abundance ratios in the outflowing gas are similar to those derived in the ambient gas . the abundance of sio in the high velocity gas is likely to be due to gas phase chemistry behind a shock ( iglesias & silk 1978 ; hartquist et al . 1980 ; mitchell & deveau 1983 ) . the enhancement of the sio abundance in the high velocity material is probably the result of grain destruction by shocks which releases silicon into the gas phase , allowing sio to form ( seab & shull 1983 ; schilke et al . 1997 ; caselli et al . 1997 ) . * methanol*. we find that the [ ch@xmath2oh / co ] abundance ratio in the ambient gas is similar for all observed cores , having a mean value of @xmath149 . assuming an [ h@xmath18/co ] of @xmath150 , it implies that the [ ch@xmath2oh / h@xmath18 ] abundance ratio in the ambient gas is typically @xmath20 . at temperatures smaller than 100 k , chemistry models predict that the production of ch@xmath2oh in the gas phase yield abundances of only 10@xmath151 relative to h@xmath18 ( lee et al . 1996 ) . the high abundance of methanol in the cores suggests that the radiation from their embedded luminous stars are producing a substantial evaporation of icy grain mantles . van der tak et al . ( 2000a ) found a [ ch@xmath2oh / h@xmath18 ] ratio of a few @xmath152 for hot cores and @xmath153 for the cores with the lower temperatures . the values obtained for the cores investigated here are in between the values found for dark clouds and hot cores , and may correspond to an average of the values appropriate for the cold envelope mass and hot core mass . * carbon monosulfide*. we find that the [ cs / co ] abundance ratio in the ambient gas is similar for all observed cores , having a mean value of @xmath154 . assuming an [ h@xmath18/co ] of @xmath150 , it implies that the [ h@xmath18/cs ] abundance ratio in the ambient gas is typically @xmath155 . this value is typical of molecular cores associated with uc hii regions and hot cores ( van der tak et al . 2000b ) . in the previous analysis we have implicitly assumed that the density and temperature are uniform within the massive cores . this is obviously a simplification , since the physical conditions are likely to vary with distance from the center of the core . in fact , the intensity profiles of the dust continuum emission indicate that the cores are highly centrally condensed . assuming that cores have density and temperature radial distributions following power laws , then for optically thin dust emission the intensity index @xmath156 is related to the density index @xmath157 ( @xmath158 ) and the temperature index @xmath159 ( @xmath160 ) , by the expression @xmath161 ( adams 1991 ; motte & andr 2001 ) , where q is a temperature and frequency correction factor with a value of @xmath162 at 1.2-mm and 30 k ( beuther et al . 2002 ) and @xmath163 is a correction factor to take into account the finite size of the cores . the massive and dense cores investigated here are heated by a cluster of luminous sources located at their central position ( chavarra et al . 2009 ; paper ii ) . for this type of objects van der tak et al . ( 2000b ) found that the temperature decreases with distance following a power - law with an index of 0.4 . adopting this value for @xmath159 and a value of 0.1 for @xmath163 , we then infer , using the above expression , that massive and dense cores have density distributions with power - law indices in the range @xmath164 . several dust continuum studies have already shown the presence of density gradients within massive and dense cores ( e.g. , van der tak et al . 2000b , mueller et al . 2002 , beuther et al . 2002 , williams et al . in particular , beuther et al . ( 2002 ) observed a large sample of high - mass protostellar candidates and found that the steeper indices are associated with the more luminous and more massive objects . the steep indices derived for the norma cores are characteristics of the subgroup of _ strong molecular sources _ within beuther s sample , which have a mean density index of 1.9 . the steep profiles are thought to indicate objects in the collapse and accretion phase . given the large mass of the massive and dense cores investigated here , of typically @xmath105 m@xmath10 , their overall collapse process is likely to produce a protostar cluster . the questions of how an individual massive star forms and how the cluster forms are therefore closely related . the derived properties of massive and dense cores are then important boundary conditions to be taken into account for models and simulations . for dynamical considerations , the relevant mass of the cores is their total mass , including gas and stars . we have already shown that the mass of gas in the cores determined from two independent methods , one based on dust observations and the other on molecular line observations , are in very good agreement . the determination of the total mass of the cores is not straightforward and is usually estimated assuming that the cores are in virial equilibrium . masses computed under this assumption , using the observed velocity dispersion and size of the cs(2@xmath11 ) emission , are given in column 4 of table [ tbl - derivedpar ] . the virial masses are similar , within the errors , to the gas masses , implying that the bulk of the core mass is in the form of molecular gas . we note that the virial masses are in average a factor of 1.5 smaller than the dust or molecular masses which might also indicate that the cores do not have enough support and may be contracting . in paper ii we show that the massive cores investigated here are associated with clusters of embedded stars which exhibit clear mass segregation . the high - mass stars are found concentrated near the center of the core while intermediate mass stars are spread over much larger distances from the core center . garay et al . ( 2006 ) have already reported that massive stars are typically located near the center of the massive and dense cores . the question arises as to whether the massive stars are formed in situ or if their spatial location is a result of dynamical effects . the characteristics of the cores derived here , namely their high densities , steep density profiles and sizes , together with the relatively low fraction of the total mass in the form of stars , suggest that dynamical friction ( also called gravitational drag ) produced by the gaseous background onto the stars may play an important role in the orbital evolution of stars ( ostriker 1999 , sanchez - salcedo & brandenburg 2001 , kim & kim 2007 ) . the role of gravitational drag on a gaseous background in the evolution and migration of stars within dense gaseous star forming clouds has been investigated using numerical smooth particle hydrodynamics simulations ( escala et al . 2003 , 2004 ) . simulations have been carried out with initial conditions similar to those of the dense cores investigated here , namely steep cloud density profile ( @xmath165 ) and perturber having masses of 1 % of the total gas mass ( @xmath166 @xmath24 ) . for a typical cloud mass of @xmath167 @xmath24 and initial distance of the stars to the center of the cloud of 0.4 pc , the migration timescale for high - mass stars is @xmath168 yr . this value is smaller than the estimated ages of the young clusters within the cores of @xmath42 1 myr ( see paper ii ) . since the migration timescales for dynamical friction are proportional to the inverse of the perturber s mass ( binney & tremaine 1987 ) , intermediate mass stars have migration timescales typically a factor @xmath42 10 longer than those of high mass stars and greater than the estimated age of the clusters . therefore , gravitational drag is considerably less effective for intermediate mass stars than for high mass stars . this suggest that in the young embedded stellar clusters only the massive stars have been able to efficiently migrate towards the center due to gravitational drag . we conclude that dynamical interactions of cluster members with the gaseous component of massive and dense cores can explain the observed mass segregation , indicating that the gas play an important role in the dynamic of young cluster members and originating the observed mass segregation . we can not rule out , however , the hypothesis that the cradles of massive stars are formed by direct accretion at the center of centrally condensed massive and dense cores ( mckee & tan 2003 ) . in their turbulent and pressurized dense core accretion model , the collapse of a massive and dense core is likely to produce the birth of a stellar cluster , with most of the mass going into relatively low - mass stars . the high - mass stars are formed preferentially at the center of the core , where the pressure is the highest , and in short time scales of @xmath169 yrs ( osorio et al . 1999 ; mckee & tan 2002 ) . we undertook molecular line observations in several species and transitions and dust continuum observations , both made using sest , towards four young high - mass star forming regions associated with highly luminous iras point sources with colors of uc hii regions and cs(2@xmath11 ) emission . these are thought to be massive star forming regions in early stages of evolution . the objectives were to determine the characteristics and physical properties of the dust clouds in which high - mass stars form . our main results and conclusions are summarized as follows . we find that the luminous star forming regions in norma are associated with molecular gas and dust cores with radii of typically 0.5 pc , masses of @xmath9 m@xmath10 , molecular hydrogen densities of typically @xmath13 @xmath14 , column densities of @xmath11 @xmath12 , and dust temperatures of @xmath15 k. all the derived physical parameters of the norma cores , except the last , are similar to those of cores harboring uc hii regions ( hunter et al . 2000 ; fandez et al . 2004 ) . for a sample of about 150 iras point sources with colours of uc hii regions and average luminosity of @xmath170 @xmath28 , fandez et al . ( 2004 ) derived an average mass of @xmath105 m@xmath10 , an average size of 0.4 pc , and an average density of @xmath171 @xmath14 . the dust temperature of the norma cores are however larger than of the fandez et al . cores , for which they find an average temperature of 32 k. the larger values derived for the norma cores are consistent with them being centrally heated by more luminous objects ( average luminosity of @xmath172 @xmath28 ) . we find that the observed radial intensity profiles of the dust continuum emission from the norma cores are well fitted with single power - law profiles , with intensity indices in the range 1.5 to 1.9 . this in turn indicates that the cores are highly centrally condensed , having radial density profiles with power - law density indices in the range @xmath173 . these steep profiles are thought to indicate objects in the collapse and accretion phase of evolution . we find that under the physical conditions of the norma cores , high - mass stars can migrate towards the central region due to gravitational drag in time scales of @xmath174 yr , smaller than the estimated ages of the clusters , providing an explanation for the observed stellar mass segregation within the cores ( paper ii ) . lccccc g324.201 + 0.119 & 15290 - 5546 & @xmath175 & @xmath176 & 6.0 & @xmath177 + g328.307 + 0.423 & 15502 - 5302 & 15 54 06.3 & @xmath5553 11 38 & 5.6 & @xmath178 + g329.337 + 0.147 & 15567 - 5236 & 16 00 33.4 & @xmath5552 44 45 & 7.0 & @xmath179 + g330.949 - 0.174 & 16060 - 5146 & 16 09 49.1 & @xmath5551 54 45 & 5.4 & @xmath180 + lcrccc cs & & 97980.968 & 52 & 0.73 & 0.132 + & & 146969.049 & 34 & 0.66 & 0.085 + & & 244935.606 & 22 & 0.45 & 0.053 + c@xmath6s & & 144617.147 & 34 & 0.66 & 0.087 + sio & & 86846.998 & 57 & 0.75 & 0.144 + & & 130268.702 & 40 & 0.68 & 0.098 + ch@xmath2oh & a@xmath36 & 96741.42 & 52 & 0.73 & 0.129 + & e & 145097.470 & 34 & 0.66 & 0.086 + co & & 115271.204 & 45 & 0.70 & 0.111 + @xmath7co & & 110201.353 & 47 & 0.71 & 0.116 + c@xmath8o & & 109782.160 & 47 & 0.71 & 0.116 + lccccccccccc cs & 3 & 0.12 & & 3 & 0.14 & & 6 & 0.064 & & 3 & 0.094 + cs & 4 & 0.065 & & 4 & 0.071 & & 4 & 0.065 & & 4 & 0.057 + cs & 5 & 0.47 & & 3 & 0.11 & & 6 & 0.41 & & 5 & 0.44 + c@xmath6s & 4 & 0.074 & & 4 & 0.068 & & 4 & 0.068 & & 4 & 0.049 + sio & 20 & 0.018 & & 12 & 0.028 & & 12 & 0.027 & & 8 & 0.026 + sio & 12 & 0.033 & & 22 & 0.016 & & 22 & 0.021 & & 4 & 0.035 + ch@xmath2oh & 8 & 0.037 & & & & & 4 & 0.054 & & 4 & 0.038 + ch@xmath2oh & 8 & 0.043 & & 8 & 0.045 & & 8 & 0.044 & & 4 & 0.049 + co & 1 & 0.276 & & 1 & 0.276 & & 1 & 0.254 & & 1 & 0.254 + @xmath7co & 3 & 0.072 & & 3 & 0.071 & & 3 & 0.068 & & 3 & 0.076 + c@xmath8o & 6 & 0.057 & & 6 & 0.055 & & 5 & 0.060 & & 5 & 0.063 + lccccccc cs & @xmath1810.12 & @xmath182.04 & @xmath1830.3 & & @xmath184.02 & @xmath185 & @xmath186 + cs & @xmath1870.07 & @xmath188.01 & @xmath1830.1 & & @xmath189.02 & @xmath190 & @xmath191 + cs & @xmath1920.47 & @xmath193.05 & @xmath1940.1 & & @xmath1950.16 & @xmath196 & @xmath197 + c@xmath6s & @xmath1980.074 & @xmath199.05 & @xmath2000.1 & & @xmath201 & @xmath202 & @xmath203 + sio & @xmath2040.018 & @xmath2050.3 & @xmath2061.1 & & @xmath207 & @xmath208 & @xmath209 + sio & & & & & @xmath210 & @xmath211 & @xmath212 + ch@xmath2oha@xmath36 & @xmath2130.037 & @xmath214.05 & @xmath2150.2 & & @xmath216 & @xmath217 & @xmath218 + ch@xmath2oha@xmath36 & @xmath2190.043 & @xmath220.03 & @xmath2210.1 & & @xmath222 & @xmath223 & @xmath191 + cs & & @xmath224 & & & @xmath225 & @xmath226 & @xmath227 + cs & @xmath228 & @xmath229.02 & @xmath2300.1 & & @xmath2310.02 & @xmath232 & @xmath233 + cs & @xmath2340.11 & @xmath235.04 & @xmath2000.2 & & & & + c@xmath6s & @xmath236 & @xmath237.09 & @xmath1940.2 & & @xmath2380.02 & @xmath239 & @xmath240 + sio & @xmath2410.028 & @xmath242.13 & @xmath2430.4 & & & @xmath244 & + sio & @xmath245 & @xmath246.10 & @xmath2060.3 & & & & + ch@xmath2oha@xmath36 & @xmath2470.045 & @xmath248.05 & @xmath2490.1 & & @xmath250 & @xmath251 & @xmath252 + cs & @xmath2530.06 & @xmath254.02 & @xmath2550.1 & & @xmath2560.01 & @xmath257 & @xmath258 + cs & @xmath2590.07 & @xmath260.01 & @xmath2610.1 & & @xmath2620.02 & @xmath263 & @xmath264 + cs & @xmath2650.41 & @xmath266.08 & @xmath2670.3 & & @xmath2680.14 & @xmath269 & @xmath270 + c@xmath6s & @xmath2710.07 & @xmath272.03 & @xmath2000.1 & & @xmath2730.02 & @xmath274 & @xmath275 + sio & @xmath2760.027 & @xmath277.10 & @xmath2780.2 & & @xmath2790.007 & @xmath280 & @xmath281 + sio & @xmath282 & @xmath283.13 & @xmath2780.3 & & & & + ch@xmath2oha@xmath36 & @xmath2840.054 & @xmath285.04 & @xmath2860.1 & & @xmath2870.014 & @xmath288 & @xmath289 + ch@xmath2oha@xmath36 & @xmath2900.044 & @xmath291.02 & @xmath2300.1 & & @xmath292 & @xmath293 & @xmath264 + cs & & @xmath224 & & & @xmath294 & @xmath295 & @xmath296 + cs & & @xmath224 & & & @xmath297 & @xmath298 & @xmath296 + cs & & @xmath224 & & & @xmath299 & @xmath300 & @xmath301 + c@xmath6s & @xmath3020.049 & @xmath303.06 & @xmath2300.5 & & @xmath304 & @xmath305 & @xmath306 + sio & @xmath3070.026 & @xmath308.03 & @xmath3090.2 & & @xmath310 & @xmath311 & @xmath312 + sio & @xmath313 & @xmath314.01 & @xmath2060.2 & & & & + ch@xmath2oha@xmath36 & @xmath315 & @xmath316.03 & @xmath2060.1 & & @xmath317 & @xmath311 & @xmath318 + ch@xmath2oha@xmath36 & @xmath319 & @xmath320.02 & @xmath3210.1 & & @xmath322 & @xmath323 & @xmath324 + llccc g324.201 + 0.119 ... & cs & @xmath55101.5 & @xmath5577.0 & 24.5 + & sio & @xmath55101.0 & @xmath5571.6 & 29.4 + & sio & @xmath55101.4 & @xmath5572.0 & 29.4 + g328.307 + 0.423 ... & cs & @xmath55101.0 & @xmath5580.6 & 20.4 + & sio & @xmath55100.0 & @xmath5582.6 & 17.4 + g329.337 + 0.147 ... & cs & @xmath55122.5 & @xmath5590.9 & 31.6 + & sio & @xmath55114.0 & @xmath5592.2 & 21.8 + & sio & @xmath325 & @xmath5590.3 & 36.6 + g330.949 - 0.174 & cs & @xmath55106.0 & @xmath5575.7 & 30.3 + & sio & @xmath55107.0 & @xmath5570.4 & 36.6 + & sio & @xmath55103.4 & @xmath5575.2 & 28.2 + lcccccc g324.201 + 0.119 & 15 32 52.91 & -55 56 10.8 & & 5.28 & 17.3 & @xmath326 + g328.307 + 0.423 & 15 54 06.33 & -53 11 43.1 & & 6.35 & 24.1 & @xmath327 + g329.337 + 0.147a & 16 00 33.10 & -52 44 46.6 & & 4.35 & 11.7 & @xmath328 + g329.337 + 0.147b & 16 00 38.68 & -52 45 31.7 & & 1.45 & 4.58 & @xmath329 + g330.949 - 0.174 & 16 09 52.41 & -51 54 58.5 & & 21.75 & 47.2 & @xmath330 + lcccccc g324.201 + 0.119 & 37 . & 0.45 & @xmath331 & @xmath332 & @xmath333 & 0.016 + g328.307 + 0.423 & 42 . & 0.46 & @xmath334 & @xmath332 & @xmath335 & 0.017 + g329.337 + 0.147a & 40 . & 0.53 & @xmath336 & @xmath337 & @xmath338 & 0.010 + g329.337 + 0.147b & 40 . & 0.76 & @xmath339 & @xmath340 & @xmath341 & 0.002 + g330.949 - 0.174 & 37 . & 0.44 & @xmath342 & @xmath343 & @xmath344 & 0.038 + lccccccccc g324.201 + 0.119 & @xmath347 & @xmath348 & @xmath349 & & 12.0 & & & @xmath350 & 31.0 + g328.307 + 0.423 & @xmath351 & @xmath352 & @xmath353 & @xmath354 & 10.3 & 6.3 & & @xmath355 & 24.3 + g329.337 + 0.147 & @xmath356 & @xmath352 & @xmath357 & @xmath358 & 17.0 & 16.7 & & @xmath359 & 45.6 + g330.949 - 0.174 & @xmath360 & @xmath361 & @xmath362 & @xmath139 & 22.6 & 11.2 & & @xmath141 & 24.9 + lcccc g324.201 + 0.119 & @xmath369 & & @xmath370 + g328.307 + 0.423 & @xmath371 & @xmath372 & @xmath373 + g329.337 + 0.147 & @xmath374 & @xmath375 & @xmath376 + g330.949 - 0.174 & @xmath374 & @xmath377 & @xmath378 +

we present the results of a meta - analysis of keck spectra of extragalactic globular clusters ( gcs ) in a sample of eight galaxies , ranging from dwarfs to massive ellipticals . we infer ages for the metal - poor and metal - rich gcs in these galaxies through comparisons to galactic gcs . both subpopulations appear to be no younger than their galactic counterparts , with ages gyr . this is the largest sample of galaxies for which ages have been constrained spectroscopically . our results support the formation of most gcs in massive galaxies at high redshift . we propose a scenario for the formation of gc subpopulations that synthesizes aspects of both accretion and in situ approaches in the context of galaxy formation through hierarchical merging .

astro-ph0506289

bimodal color distributions were first discovered in extragalactic globular cluster ( gc ) systems in the early nineties ( ostrov , geisler , & forte 1993 ; zepf & ashman 1993 ) and are now known to be commonplace among luminous galaxies ( larsen 2001 ; kundu & whitmore 2001 ) . it was realized early on that this finding had far - reaching implications for our understanding of the formation histories of gc systems . since gc formation is thought to accompany all major star forming episodes in a galaxy s history ( schweizer 2000 ) , an understanding of the origin of gc bimodality implicitly constrains the formation histories of their host galaxies . several scenarios have been developed to explain these gc subpopulations , which are generally termed metal - poor ( mp ) and metal - rich ( mr ) . the major merger model ( schweizer 1987 ; ashman & zepf 1992 ) predicts the formation of mr gcs out of enriched gas during spiral spiral mergers that form elliptical galaxies . the mp gcs are associated with the progenitor spirals . forbes , brodie , & grillmair ( 1997 ) pointed out that bimodality could also arise in a multiphase in situ scenario : the mp gcs are formed in gaseous clumps within the early protogalactic potential , while the mr gcs are formed after a dormant period in a second phase of star formation . it is during this second phase that the bulk of the galaxy starlight is thought to be formed . finally , c^ ot ' e et al . ( 1998 , 2000 , 2002 ) suggested an accretion scenario in which a massive seed galaxy forms mr gcs in situ , while mp gcs are acquired via the accretion of dwarf galaxies and their accompanying low - metallicity gc systems . gc formation has also recently been simulated with semi - analytic models ( sams ; beasley 2002 ) and n - body+gasdynamic codes ( kravtsov & gnedin 2005 ) . the beasley model , which followed gc systems to the present day , found it necessary to truncate mp gc formation at @xmath1 to produce bimodality . candidates for this truncation process include feedback in low - mass halos ( forbes 1997 ) and cosmic reionization ( santos 2003 ) . some of the scenarios mentioned above were developed before the clear success of @xmath2cdm cosmology and the hierarchical merging paradigm . however , they are relevant for understanding structure formation . these scenarios describe the behavior of baryons within dark matter halos , which is dominated by processes that are relatively unconstrained at present . all of the scenarios predict old ages for mp gcs , but not necessarily for mr gcs . in the merger model , the mergers ( and hence the mr gcs ) will have a range of ages . in the multiphase scenario both gc subpopulations are expected to be old ( @xmath0 gyr ) with the mr gcs slightly younger than the mp gcs . some variation with environment and galaxy mass may be expected . in the accretion scenario , both subpopulations are @xmath3 coeval and very old . of course , reality is likely to be some hybrid of the these alternatives , and in a very real sense the distinctions between them blur if the processes occur at high enough redshift . nonetheless , it is important to determine whether one ( or more ) of these ideas represents a dominant mechanism . thus , the goal of age - dating gc subpopulations and , in particular , detecting or ruling out age differences between mr and mp gcs , will set constraints on the processes of galaxy formation and limits on the epochs of galaxy assembly . the goal of this work is to determine relative ages of gc subpopulations and set limits on their absolute ages ( modulo modeling and observational uncertainties ) . our data are among the best spectra currently available for extragalactic gcs , and it is unlikely that significantly tighter age constraints for gcs will be forthcoming using lick index diagnostics ( trager 1998 and references therein ) in the era of @xmath4 m class telescopes . since the lick system is not optimized for studying stellar systems with low velocity dispersion , this current short study is intended as a summary of our spectroscopic work in age - dating gcs and as a starting point for our future investigations in which new , higher resolution ( 3 ) index definitions will be employed . not all galaxies for which our group has previously obtained spectra are included in this study . here we require all lick indices ( see details below ) from h@xmath5 to fe5335 to be measurable , and that at least two gcs per subpopulation have @xmath6 and h@xmath5 errors @xmath7 . this left a sample of eight galaxies , which are listed in table 1 . the sample is dominated by massive early - type galaxies in group and cluster environments , but includes two dwarfs . the gc spectra were obtained at keck with lris ( oke 1995 ) over a variety of runs , from dec . 1995 to apr . 2004 . spectra were taken in multislit mode ( except for the gcs of the fornax dwarf spheroidal ) , with resolution ranging from 3.38 . the spectral coverage was typically @xmath8 , but varied with slit placement and from run to run , sometimes going further to the red or blue . we refer the reader to the respective papers noted in table 1 for details on the reduction of each data set ; however , a short summary follows . gc spectra extracted from individual exposures ( typically 30 min in length ) were wavelength calibrated using arc lamps and then co - added . small zero - point shifts to the wavelength calibration were made using bright night sky lines , and a standard flux calibration was carried out using one or two standards from massey ( 1988 ) . lick indices were measured using the definitions in trager ( 1998 ) and worthey & ottaviani ( 1997 ) . the spectra were smoothed to the wavelength - dependent resolution ( 811 ) of the lick system before the measurement of indices . an unfortunate aspect of the present study is that we can not fully calibrate our index measurements onto the lick system . the number and type of lick standard stars varied from run to run , rendering consistent corrections impossible . we note that many of our previous studies of gcs in individual galaxies ( e.g. , brodie 2005 ; strader , brodie , & forbes 2004a ) have found such corrections unnecessary : they are generally smaller than the error bars on individual index measurements , and there appear to be no large systematic effects . however , our lack of lick corrections could still result in small offsets when comparing to lick index predictions of simple stellar population ( ssp ) models . for this reason , we have decided to concentrate on a _ differential _ analysis between our data and integrated spectra of galactic gcs . these galactic data are from gregg ( 1994 ) and puzia ( 2002b ) . in the former case we measured indices directly from the digital spectra , as described above . in the latter case we de - corrected " the published indices using lick offsets given in the same paper . the resulting datasets should be amenable to consistent comparisons . even with our best individual spectra , estimating accurate ages and metallicities is difficult . an error in h@xmath9 of only @xmath10 corresponds to an age error of @xmath114 gyr for a 13 gyr old gc . thus , to facilitate comparison to the galactic gcs , we constructed _ mean _ mp and mr ( and in ngc 4365 and ngc 1407 , intermediate - metallicity ) subpopulations for all galaxies , except for the two dwarfs fornax and vcc 1087 . fornax has only mp gcs ; vcc 1087 may have a small number of mr gcs , but they were sparsely sampled in our spectroscopy and do not meet the criteria outlined above ( beasley 2005 ) . photometric studies of the remainder of our sample galaxies show clear evidence for subpopulations , so this division is justified ( see individual papers for details ) . gcs were sorted into subpopulations primarily on the basis of this photometry ( with typical separation at @xmath12 , @xmath13 , or @xmath14 ) . these criteria were augmented by principle components analysis ( pca)-based metallicities when ambiguous ( strader & brodie 2004 ) . the number of gcs per subpopulation are listed in table 1 . lick indices for these subpopulations are weighted means of the individual gc indices with @xmath15 rejection . the variance is calculated as the sum of the weights and is equivalent to a weighted standard variance of the mean . the mean subpopulation indices and standard errors of the mean are given in tables 2 and 3 . the only gcs excluded outright were two suspected young ( @xmath162 gyr ) metal - rich gcs in the dynamically young elliptical ngc 3610 ( strader 2003a ; strader 2004a ) , since our focus is on deriving the properties of the old " subpopulations . for each subpopulation with at least three gcs , weighted standard deviations of the indices are given in table 4 , and represent the combined effects of observational error and intrinsic differences among the gcs within each subpopulation . the majority of our gc samples are biased . first there is the necessary luminosity bias : nearly all of our gcs have @xmath17 . adequate s / n can not be achieved for gc fainter than this , even in @xmath18 hours at keck . this apparent magnitude corresponds to @xmath19 at the distance of virgo , 0.5 mag brighter than the turnover of the gc luminosity function ( lf ) . however , neither metallicity or age correlates strongly with gc mass in the milky way , suggesting that this introduces minimal bias in derived properties . more importantly , the gcs were not selected uniformly in color . this makes it impossible to construct `` true '' composite subpopulations that accurately reflect the underlying metallicities . however , we _ can _ constrain mean ages , since younger gc subpopulations with a standard log - normal lf will be , on average , brighter than their old counterparts , even taking into account their probable higher metallicities . thus magnitude - selected samples will give lower limits to mean gc subpopulation ages , unless the lf of the younger gcs is biased to low masses , either at birth or because of differing gc destruction . the traditional method of estimating ages and metallicities using lick indices involves constructing index - index plots . [ mgfe ] , the harmonic mean of mg@xmath20 and ( fe5270+fe5335)/2 , is often used as a metallicity indicator largely insensitive to variations in [ @xmath21/fe ] . however , the use of [ mgfe ] does not take advantage of the metallicity information contained in the remainder of the lick indices . we have developed a pca - based method for estimating gc metallicities using 11 lick indices ( strader & brodie 2004 ) , and choose to use these more robust metallicities instead of the standard [ mgfe ] ( or the slightly different [ mgfe]@xmath22 , in which @xmath23fe@xmath24 is replaced by ( @xmath25 fe5270 + @xmath26 fe5335 ) ; thomas , maraston , & bender 2003 ) . puzia ( 2004 ) have argued that the best lick balmer index for estimating gc ages is h@xmath27 . however , the blue pseudocontinuum bandpass of h@xmath27 encompasses most of the ch absorption of the g band centered at @xmath28 . this makes the index extremely sensitive to changes in the carbon abundance ( and overall metallicity , though its sensitivity decreases toward high metallicities ) . thus , we have chosen not to use either of the lick h@xmath29 indices in this study . as noted above , we are currently engaged in a redefinition of the lick system that is optimized for gcs . instead , we will use the average of h@xmath9 and @xmath30 as our age - sensitive balmer indices ( in fact , using h@xmath27 in the average as well gives consistent results ) . in figure 1 we plot pca [ m / h ] vs. ( h@xmath31+@xmath30)/2 . the formal random errors for pca [ m / h ] are negligible , but we assign a systematic calibration error of 0.12 dex to all points ( see strader & brodie 2004 ) . our gc subpopulations fall at or below the envelope defined by the galactic gcs , suggesting that they are no younger than galactic gcs at all metallicities . 14 gyr model lines for [ @xmath32/h ] = @xmath33 to 0 and [ @xmath21/fe ] = 0 and + 0.3 from thomas , maraston , & korn ( 2004 , tmk ) are superposed . these include a blue horizontal branch below [ @xmath32/h ] = @xmath34 . the models are plotted only to show that our index combinations are relatively insensitive to [ @xmath21/fe ] variations ; we can not make any direct model comparisons since our data are not on the lick system . indeed , ages derived using pca metallicities are even less sensitive to [ @xmath21/fe ] than those derived from [ mgfe]@xmath22 . thus , the use of pca metallicities is preferable when deriving ssp ages for gcs , and this is likely to be true even for more metal - rich stellar systems , up to the current pca calibration limit of solar metallicity . the mr galactic gc with high balmer line strength is ngc 6553 ; its location is due not to youth or warm horizontal branch stars but probably poor sampling in the crowded bulge ( puzia 2002b ) . let us consider the scatter among the mean subpopulations in figure 1 more quantitatively . the relation between [ m / h ] and balmer line strength appears to be linear to solar metallicity ; a least - squares fit to all points with [ m / h ] @xmath35 gives : ( h@xmath9+@xmath30)/2 = @xmath36 [ m / h ] + 0.632 , with a residual standard error of 0.17 . the mean balmer error for these subpopulations is 0.16 . thus , the extragalactic points are entirely consistent with expected scatter in the mean values around a single line . this need not necessarily be a single isochrone ( as there could be a shift to younger ages with increasing metallicity ) , but it does indicate that there is not a large age spread from galaxy to galaxy at fixed metallicity . what can we conclude about the absolute ages of our gc subpopulations ? our results are directly tied to knowledge of the ages of mp and mr gcs in the milky way ( mw ) . unfortunately , there is considerable uncertainty in this subject as well . using @xmath21-enhanced isochrones with diffusion , gratton ( 2003 ) find that the oldest mp gcs have formal ages of 13.4 @xmath37 gyr ( random and systematic errors ) . krauss & chaboyer ( 2002 ) used a monte carlo technique to perform isochrone fits to simulated metal - poor color - magnitude diagrams with varying physical parameters , and found best - fit ages of 12.6 gyr with @xmath38 errors of @xmath16 gyr . a recent revision of the @xmath39n(@xmath40)@xmath41o reaction rate raises these ages by 0.71.0 gyr ( imbriani 2004 ) ; they are indistinguishable from the age of the universe within the errors . similar ages are found through integrated spectroscopy ( proctor , forbes , & beasley 2004 ) . few mr gcs have robust ( determined through direct main - sequence fitting ) age determinations . even for the best - studied of these gcs , 47 tuc , there are still significant discrepancies , depending on the choice of subdwarfs and colors used for fitting . 47 tuc could be coeval with mp gcs , or @xmath42 gyr younger at 11 gyr ( gratton 2003 , grundahl , stetson , & andersen 2002 ; or @xmath43 gyr with the updated @xmath39n(@xmath40)@xmath41o rate ) . the main source of error is the distance modulus , and a convincing resolution of the issue may require direct parallax measurements from sim or gaia . in addition , though the bulk of the mr gcs are probably associated with the bulge ( minniti 1995 ; c^ ot ' e 1999 ; forbes , larsen , & brodie 2001 ) , a subset may belong to the thick disk including 47 tuc . such thick disk clusters , if they exist , probably have few counterparts in early - type galaxies . secure turnoff ages for more mr gcs will be needed before a more certain estimate of their mean properties can be made , but for the purposes of this paper we assume a minimum age of 10 gyr for mr gcs . in conclusion , the gc subpopulations in our sample all appear to have mean ages @xmath0 gyr . however , relative _ mean _ subpopulation ages more accurate than @xmath44 gyr ( @xmath38 ) can not be determined at present , and age errors on individual clusters are @xmath45 gyr . the old gc ages derived in this paper are consistent with an emerging observational and theoretical consensus that very massive galaxies , especially in dense environments , formed the bulk of their stars at high redshift ( mccarthy 2004 ; chen & marzke 2004 ; somerville 2004 ; nagamine 2001 ) . there is no sharp cutoff for the classification of a galaxy as `` very massive '' , but a number often adopted is 3@xmath46 = @xmath47 , which corresponds to @xmath48 for an old metal - rich stellar population . most spectroscopic studies of gc systems beyond the local group have focused on such galaxies , since they generally have the largest gc systems and thus are easiest to study . our findings in this paper , coupled with the enormous number of photometric studies of gcs in such galaxies over the last several decades , are entirely consistent with the formation at high redshift of most gcs in massive galaxies in dense environments . these ages are consistent with previous spectroscopic studies of gc systems in individual galaxies by our group and by other workers ( e.g. , kuntschner 2002 ; cohen , blakeslee , & ryzhov 1998 ) . old ages for the majority of gcs in galaxies of all types might seem , on the face of it , to conflict with the discovery of young `` proto - gc '' candidates in gas - rich mergers ( e.g. , ngc 1275 , holtzman 1992 ; ngc 7252 , whitmore 1993 ; the antennae , whitmore & schweizer 1995 ; ngc 3921 , schweizer 1996 ) and intermediate - age ( @xmath165 gyr ) gcs in some merger remnants . these latter systems have been targeted because they have already suffered the bulk of their dynamical evolution and can thus be more readily compared to old gc systems . there are only two galaxies for which intermediate - age gcs have been spectroscopically confirmed : ngc 1316 ( goudfrooij 2001 ) and ngc 3610 ( strader 2003a ) , though the existence of such gcs has been suggested on the basis of lower s / n spectra in ngc 5128 ( peng , ford , & freeman 2004 ) and ngc 1399 ( forbes 2001 ) . ngc 1316 appears to have a significant population of intermediate - age clusters , as revealed by variations of the gclf with galactocentric radius ( goudfrooij 2004 ) . however , these clusters make up only a minor component of the total gc system in ngc 3610 ( strader 2004a ) . the 12 gyr merger remnant ngc 1052 shows no evidence for intermediate - age gcs ( pierce 2005 ) . on the basis of optical / near - ir photometry , it has been suggested that intermediate - age subpopulations exist in several other galaxies ( e.g. , ngc 4365 , puzia 2002a ; ngc 5846 , hempel 2003 ) . recent high s / n , follow - up spectroscopy of prime intermediate - age candidates in ngc 4365 found only old gcs ( brodie 2005 ) , showing the need for high s / n spectroscopic confirmation . the intermediate - age merger remnants studied thus far are a mixed bag ; at least some remnants have smaller populations of intermediate - age gcs than might be expected if the major merger model is correct . in addition to gc ages , a successful formation scenario must also reproduce the gc metallicity galaxy mass relations for _ both _ mp and mr gcs . the slopes of these relations are now well - determined ( larsen 2001 , strader , brodie , & forbes 2004b ) , and can offer important quantitative constraints on gc formation models , although they have not yet been used in this respect . the relations are also useful for defining qualitative scenarios for the formation of massive ellipticals , as discussed below . the in situ scenario ( forbes , brodie & grillmair 1997 ) is sometimes considered together with the sam of beasley ( 2002 ) under the rubric hierarchical " . in some sense all modern scenarios must be hierarchical , given the strength of evidence for the assembly of present - day dm halos through hierarchical merging in a @xmath2cdm cosmology . indeed , the accretion models of c^ ot ' e et al . ( 1998 , 2000 , 2002 ) were explicitly inspired by the convergence toward hierarchical galaxy assembly . while this specific scenario is difficult to reconcile with recent observations ( since the mean metallicities of gcs in dwarf galaxies are lower than those of mp gcs in massive galaxies ; strader 2004b ) , the hierarchical nature of structure formation implies that all models are accretion models to a certain extent . a key problem with the accretion model ( and a point not considered in strader 2004b ) is that local dwarfs can not be directly compared to their counterparts at high redshift ( see , e.g. , santos 2003 as it relates to mp gcs , and robertson 2005 on the formation of the galactic halo ) . dwarf galaxies sitting `` on top '' of a large overdensity destined to become a massive galaxy will collapse before those on the periphery . if mp gc formation was terminated by reionization , the dwarfs that collapsed first might have produced mp gcs of a higher metallicity than those dwarfs that collapsed later . this could happen in a variety of ways : the dwarfs collapsing first could accrete a larger quantity of pristine gas ( thereby enhancing star formation ) , have more time to process the gas they accreted , or capture more enriched , outflowing gas from the higher density of nearby starforming halos . these more centrally concentrated halos would generally be accreted into their parent halo , with their gcs forming its mp subpopulation , while some of the outlying halos would survive to be present - day dwarfs . this scenario could qualitatively reproduce the mp gc - metallicity galaxy mass relation , since larger overdensities collapse first and thus would have mp gcs of higher metallicity . since the mp gcs are accreted from low - mass halos that are enriched in place " within a dm overdensity , our picture can be seen as a synthesis of the accretion and in situ approaches . we have begun work to test this hypothesis using sam+n - body halo merger trees and simple chemical evolution models . recent numerical simulations of the assembly of milky way - sized halos ( moore 2005 , priv . communication ) indicate that the radial distribution of mp galactic gcs they predict is consistent with our scenario . we also note here the continuing efforts to place gc formation in a cosmological context . given the lack of observational evidence for dm halos around gcs ( moore 1996 ; odenkirchen 2003 ) , a popular view among cosmologists is that mp gcs formed with halos that were later stripped away ( e.g. , cen 2001 ; bromm & clarke 2002 ; scannapieco et al . 2004 ) or are still present but difficult to detect ( mashchenko & sills 2005 ) . however , any model which requires different formation mechanisms for mp and mr gcs must contend with the strong similarities between the two subpopulations , including sizes and the shapes of the mass and metallicity distributions ( see also harris 2003 ) . what can we infer about the masses of halos in which mp gcs formed ? a number of galactic satellites with total masses @xmath49@xmath50 lack gcs entirely ( forbes 2000 ) , while the fornax and sagittarius dsphs , with masses @xmath51 ( walcher 2003 ; law 2005 ) , have at least five gcs each . the dsph ddo 78 in the m81 group is slightly less massive and has one gc ( sharina 2003 ) . grebel ( 2000 ) find a candidate gc of the m31 satellite and i ( @xmath52 ; @xmath53 a few @xmath54 for an @xmath55 similar to fornax ) , but at a projected distance of @xmath56 kpc from m31 the gc could be a contaminant . these results might suggest that mp gcs formed in halos with a minimum mass of @xmath57@xmath58 . this `` intermediate '' regime could be better probed with further study of and i ( and the similar mass satellite and ii ) . despite a variety of evidence suggesting that massive , compact star clusters form in most ( possibly all ) major episodes of star formation ( e.g. , schweizer 2000 ) , and that a subset of these clusters may survive to become gcs , the relation between gcs and field stars is poorly understood . for lack of data , arguments for the universality of gc formation efficiency ( e.g. , mclaughlin 1999 ) have generally not been based on the separation of gc and field star subpopulations . in the mw , modern interpretations associate mp and mr gcs with the halo ( @xmath59 ; carney , latham , & laird 1990 ) and bulge ( @xmath60 ) , respectively there are @xmath61 mp and 50 mr gcs in the mw . the subpopulations have mass - normalized specific frequencies ( @xmath62 = the number of gcs per @xmath63 @xmath64 ; zepf & ashman 1993 ) of 67 and 5 , respectively . the `` universal '' gc formation efficiency of mclaughlin corresponds roughly to @xmath65 ( assuming all gas in a system has been converted to stars ) , and is rather similar to the mr value . however , the mp value is more than a factor of 10 higher than the mr one . a strong link between mr gcs and field stars is predicted in all the gc formation scenarios discussed to date and is consistent with observations of several extragalactic gc systems ( e.g. , m31 : jablonka 2000 ; ngc 5128 : harris 1999 , harris & harris 2002 ; ngc 1399 : forte , faifer , & geisler 2005 ) and the nearly - constant bulge specific frequency " in elliptical galaxies and many spiral bulges ( forbes , brodie & larsen 2001 ; see goudfrooij 2003 for a slightly different view ) . in none of these galaxies have strong constraints been placed on the existence of a population of metal - poor `` halo '' field stars ; existing surveys in ngc 5128 and m31 ( though see guhathakurta 2005 ) have found few metal - poor stars , suggesting that the mp @xmath62 value may generically be higher than the mr . in our scenario , truncation by reionization offers a natural explanation for these high @xmath62 values of mp gc systems . the intimate connection between mr gcs and the underlying galaxy starlight has implications for the present study , as the old ages of mr gcs in our sample of galaxies suggests old ages for the bulk of their field stars . however , with our small samples of gcs , we can not constrain at present whether they were formed in a short - lived `` monolithic '' dissipational event nor whether a small age spread ( @xmath66 gyr ) might exist . both in situ and accretion scenarios predict old ages for red gcs ( but not the formation mechanism ) . the classical major merger picture ( ashman & zepf 1992 ) , however , would predict younger ages , unless the main epoch of major mergers is pushed to higher redshift ( say @xmath67 ) ; these could be visible as scuba sources ( smail 2002 ) . indeed , massive disks have recently been found at these redshifts ( e.g. , labb ' e 2003 ; stockton , canalizo , & maihara 2004 ) . however , whether such disks had sufficient gas or were present in the requisite numbers to account for the formation of most massive ellipticals is still an open question . our point here is not that gas - rich starbursts were uncommon in the early universe , but that these starbursts may have been driven by stochastic , ongoing merging of objects over a range of mass scales rather than simple " major mergers of fully - formed massive disks . other issues with the major merger scenario , such as specific frequencies , relative numbers of mp and mr gcs , and metallicities of the mr gcs , have been discussed in detail elsewhere ( e.g. , ashman & zepf 1998 ; harris 2001 ) . it is clear that much larger samples of high s / n gc spectra will be needed to investigate the details of mr gc formation . because of downsizing , the most massive galaxies should have smaller age differences among their mr gcs but larger numbers of bright gcs to probe these differences ; less massive galaxies have fewer gcs but could have more age structure . due to the steep central concentration of gc systems , multi - object slit spectrographs , even if highly multiplexing , will continue to be limited in their ability to observe large numbers of gcs . fiber instruments on large telescopes or natural - seeing ifus may be better suited to this topic . we have undertaken a meta - analysis of high signal - to - noise keck spectra of gcs in eight galaxies , ranging from dwarfs to massive ellipticals . we have established the best constraints on gc ages that can realistically be achieved using current methods and technology . ages have been inferred from a direct comparison to galactic gcs to avoid any dependence on stellar evolutionary synthesis models . we conclude that both the mp and mr subpopulations in these galaxies are no younger than their galactic counterparts ( @xmath0 gyr ) and , within the uncertainties ( @xmath68 gyr ) , both subpopulations are coeval . these results indicate that the vast majority of gcs , at least in dense environments , were formed at high redshift ( @xmath67 ) . given the expectation that gc formation has accompanied all major star forming episodes in a galaxy s history , the old ages of gcs implies an old formation epoch for the stars in their host galaxies . this is consistent with results emerging from large photometric surveys of galaxies , which are increasingly revealing massive galaxies already in place at similar redshifts . we have synthesized a new gc formation scenario which contains elements of the in situ and accretion models that have been invoked in recent years to explain the existence of gc subpopulations . the scenario suggests that mp gcs formed in low - mass halos in the early universe , with their formation truncated by reionization . mr gcs formed along with the bulk of the field stars in massive early - type galaxies and spiral bulges . this scenario qualitatively reproduces the mp and mr gc metallicity - galaxy mass relations and can explain the relative formation efficiencies of mp and mr gcs in massive galaxies . we thank graeme smith for useful comments and discussion . we acknowledge support by the national science foundation through grant ast-0206139 and a graduate research fellowship ( j.s . ) . a. j. c. acknowledges financial support from a ucm fundacin del amo fellowship . this research has made use of the nasa / ipac extragalactic database ( ned ) which is operated by the jet propulsion laboratory , california institute of technology , under contract with the national aeronautics and space administration . 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 the national aeronautics and space administration . the observatory was made possible by the generous financial support of the w.m . keck foundation . 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 . ashman , k. m. , & zepf , s. e. 1992 , , 384 , 50 ashman , k. m. , & zepf , s. e. 1998 , globular cluster systems / cambridge : cambridge university press beasley , m. a. , baugh , c. m. , forbes , d. a. , sharples , r. m. , & frenk , c. s. 2002 , , 333 , 383 beasley , m. a. , strader , j. , cenarro , a. j. , brodie , j. p. , & geha , m. 2005 , in preparation brodie , j. p. & huchra , j. p. 1990 , , 362 , 503 brodie , j. p. , strader , j. , denicol , g. , beasley , m. a. , cenarro , a. j. , larsen , s. s. , kuntschner , h. , & forbes , d. a. 2005 , , submitted bromm , v. , & clarke , c. j. 2002 , , 566 , l1 burstein , d. , faber , s. m. , gaskell , c. m. , & krumm , n. 1984 , , 287 , 586 carney , b. w. , latham , d. w. , & laird , j. b. 1990 , , 99 , 57 cen , r. 2001 , , 560 , 592 cenarro , a. j. , strader , j. , brodie , j. p. , beasley , m. a. , 2005 , in preparation chen , h. , & marzke , r. o. 2004 , , 615 , 603 chomiuk , l. b. , strader , j. , & brodie , j. p. 2005 , in preparation cohen , j. g. , blakeslee , j. p. , & ryzhov , a. 1998 , , 496 , 808 c^ ot ' e , p. , marzke , r. o. , & west , m. j. 1998 , , 501 , 554 c^ ot ' e , p. 1999 , , 118 , 406 c^ ot ' e , p. , marzke , r. o. , west , m. j. , & minniti , d. 2000 , , 533 , 869 c^ ot ' e , p. , west , m. j. , & marzke , r. o. 2002 , , 567 , 853 durrell , p. r. , harris , w. e. , & pritchet , c. j. 2001 , , 121 , 2557 forbes , d. a. , brodie , j. p. , & grillmair , c. j. 1997 , , 113 , 1652 forbes , d. a. , masters , k. l. , minniti , d. , & barmby , p. 2000 , , 358 , 471 forbes , d. a. , brodie , j. p. , & larsen , s. s. 2001 , , 556 , l83 forbes , d. a. , beasley , m. a. , brodie , j. p. , & kissler - patig , m. 2001 , , 563 , l143 forte , j. c. , faifer , f. , & geisler , d. 2005 , , 357 , 56 goudfrooij , p. , mack , j. , kissler - patig , m. , meylan , g. , & minniti , d. 2001 , , 322 , 643 gratton , r. g. , bragaglia , a. , carretta , e. , clementini , g. , desidera , s. , grundahl , f. , & lucatello , s. 2003 , , 408 , 529 grebel , e. k. , dolphin , a. e. , & guhathakurta , p. 2000 , astronomische gesellschaft meeting abstracts , 17 , 61 gregg , m. d. 1994 , , 108 , 2164 grundahl , f. , stetson , p. b. , & andersen , m. i. 2002 , , 395 , 481 guhathakurta , p. , et al . 2005 , astro - ph/0502366 harris , w. e. 1996 , , 112 , 1487 harris , g. l. h. , harris , w. e. , & poole , g. b. 1999 , , 117 , 855 harris , w. e. 2001 , saas - fee advanced course 28 : star clusters , 223 harris , w. e. , & harris , g. l. h. 2002 , , 123 , 3108 harris , w. e. 2003 , extragalactic globular cluster systems , 317 hempel , m. , hilker , m. , kissler - patig , m. , puzia , t. h. , minniti , d. , & goudfrooij , p. 2003 , , 405 , 487 holtzman , j. a. , et al . 1992 , , 103 , 691 imbriani , g. , et al . 2004 , , 420 , 625 jensen , j. b. , tonry , j. l. , & luppino , g. a. 1998 , , 505 , 111 jerjen , h. , binggeli , b. , & barazza , f. d. 2004 , , 127 , 771 krauss , l. m. , & chaboyer , b. 2003 , science , 299 , 65 kravtsov , a. & gnedin , o. 2005 , , in press kundu , a. , & whitmore , b. c. 2001 , , 121 , 2950 kuntschner , h. , ziegler , b. l. , sharples , r. m. , worthey , g. , & fricke , k. j. 2002 , , 395 , 761 labb ' e , i. , et al . 2003 , , 591 , l95 larsen , s. s. , brodie , j. p. , huchra , j. p. , forbes , d. a. , & grillmair , c. j. 2001 , , 121 , 2974 larsen , s. s. , brodie , j. p. , beasley , m. a. , & forbes , d. a. 2002 , , 124 , 828 law , d. r. , johnston , k. v. , & majewski , s. r. 2005 , , 619 , 807 mashchenko , s. , & sills , a. 2005 , , 619 , 243 massey , p. , strobel , k. , barnes , j. v. , & anderson , e. 1988 , , 328 , 315 mccarthy , p. j. , et al . 2004 , , 614 , l9 mclaughlin , d. e. 1999 , , 117 , 2398 minniti , d. 1995 , , 109 , 1663 moore , b. 1996 , , 461 , l13 nagamine , k. , fukugita , m. , cen , r. , & ostriker , j. p. 2001 , , 558 , 497 odenkirchen , m. , et al . 2003 , , 126 , 2385 oke , j. , cohen , j. , carr , m. , cromer , j. , dingizian , a. , harris , f. , labrecque , s. , lucinio , r. , schaal , w. , epps , h. , & miller , j. 1995 , , 107 , 375 ostrov , p. , geisler , d. , & forte , j. c. 1993 , , 105 , 1762 peng , e. w. , ford , h. c. , & freeman , k. c. 2004 , , 602 , 705 perrett , k. m. , hanes , d. a. , butterworth , s. t. , kavelaars , j. , geisler , d. , & harris , w. e. 1997 , , 113 , 895 pierce , m. , forbes , d. a. , proctor , r. , beasley , m. a. , brodie , j. p. , & strader , j. 2005 , , in press pritchet , c. j. , & van den bergh , s. 1987 , , 316 , 517 proctor , r. p. , forbes , d. a. , & beasley , m. a. 2004 , , in press puzia , t. h. , zepf , s. e. , kissler - patig , m. , hilker , m. , minniti , d. , & goudfrooij , p. 2002 , , 391 , 453 puzia , t. h. , saglia , r. p. , kissler - patig , m. , maraston , c. , greggio , l. , renzini , a. , & ortolani , s. 2002 , , 395 , 45 puzia , t. h. , et al . 2004 , , 415 , 123 rose , j. a. & tripicco , m. j. 1986 , , 92 , 610 santos , m. r. 2003 , extragalactic globular cluster systems , 348 scannapieco , e. , weisheit , j. , & harlow , f. 2004 , , 615 , 29 schlegel , d. j. , finkbeiner , d. p. , & davis , m. 1998 , , 500 , 525 schweizer , f. 1987 , nearly normal galaxies . from the planck time to the present , 18 schweizer , f. 2000 , phil . london a , 358 , 2063 schweizer , f. , miller , b. w. , whitmore , b. c. , & fall , s. m. 1996 , , 112 , 1839 sharina , m. e. , silchenko , o. k. , & burenkov , a. n. 2003 , , 397 , 831 smail , i. , ivison , r. j. , blain , a. w. , & kneib , j .- p . 2002 , , 331 , 495 somerville , r. 2004 , astro - ph/0401570 stockton , a. , canalizo , g. , & maihara , t. 2004 , , 605 , 37 strader , j. , brodie , j. p. , schweizer , f. , larsen , s. s. , & seitzer , p. 2003 , , 125 , 626 strader , j. , brodie , j. p. , forbes , d. a. , beasley , m. a. , & huchra , j. p. 2003 , , 125 , 1291 strader , j. , brodie , j. p. , & forbes , d. a. 2004 , , 127 , 295 strader , j. , brodie , j. p. , & forbes , d. a. 2004 , , 127 , 3431 strader , j. & brodie , j. p. 2004 , , 128 , 1671 thomas , d. , maraston , c. , & bender , r. 2003 , , 339 , 897 tikhonov , n. a. , galazutdinova , o. a. , & drozdovsk , i. o. 2005 , , 431 , 127 thomas , d. , maraston , c. , & korn , a. 2004 , , 351 , l19 tonry , j. l. , dressler , a. , blakeslee , j. p. , ajhar , e. a. , fletcher , a. b. , luppino , g. a. , metzger , m. r. , & moore , c. b. 2001 , , 546 , 681 trager , s. c. , worthey , g. , faber , s. m. , burstein , d. , & gonzalez , j. j. 1998 , , 116 , 1 trager , s. c. , faber , s. m. , worthey , g. , & gonz ' alez , j. j. 2000 , , 120 , 165 tully , r. b. 1988 , nearby galaxies catalog ( cambridge : cambridge univ . press ) walcher , c. j. , fried , j. w. , burkert , a. , & klessen , r. s. 2003 , , 406 , 847 whitmore , b. c. , & schweizer , f. 1995 , , 109 , 960 whitmore , b. c. , schweizer , f. , leitherer , c. , borne , k. , & robert , c. 1993 , , 106 , 1354 worthey , g. , & ottaviani , d. l. 1997 , , 111 , 377 zepf , s. e. & ashman , k. m. 1993 , , 264 , 611 zinn , r. & west , m. j. 1984 , , 55 , 45 ngc 1407 & e0 & @xmath69 & mp , i m , mr & 4 , 9 , 8 & jensen , tonry , & luppino ( 1998 ) & cenarro ( 2005 ) + ngc 4365 & e3 & @xmath70 & mp , i m , mr & 8 , 9 , 5 & tonry ( 2001 ) & brodie ( 2005 ) + ngc 4594 & sa & @xmath70 & mp , mr & 7 , 7 & tonry ( 2001 ) & larsen ( 2002 ) + ngc 3610 & e5 & @xmath71 & mp , mr & 3 , 7 & tully ( 1988 ) & strader , brodie , & forbes ( 2004a ) + ngc 1052 & e4 & @xmath72 & mp , mr & 7 , 2 & tonry ( 2001 ) & pierce ( 2005 ) + ngc 7457 & s0 & @xmath73 & mp , mr & 9 , 3 & tonry ( 2001 ) & chomiuk , strader , & brodie ( 2005 ) + vcc 1087 & de , n & @xmath74 & mp & 9 & & beasley ( 2005 ) + fornax & dsph & @xmath75 & mp & 4 & & strader ( 2003b ) + ngc 1407 & mp & 2.53 & 2.10 & @xmath76 & @xmath77 & 0.59 & 3.05 & @xmath78 & 1.34 & 1.54 & 0.60 & 1.84 & 0.50 & 2.41 & 2.65 & 0.024 & 0.082 & 1.27 & 1.25 & 1.15 & 0.67 & @xmath79 + ngc 4365 & mp & 1.31 & 1.46 & @xmath80 & 0.017 & 0.47 & 3.49 & @xmath81 & 0.80 & 2.02 & 0.62 & 1.90 & 1.34 & 2.22 & 2.86 & 0.042 & 0.111 & 1.75 & 1.30 & 1.33 & 0.89 & @xmath82 + ngc 4594 & mp & 2.35 & 2.04 & @xmath83 & @xmath84 & 0.57 & 3.40 & @xmath85 & 1.27 & 1.08 & 0.43 & 1.82 & 0.86 & 2.15 & 2.39 & @xmath86 & 0.037 & 1.18 & 1.15 & 1.25 & 0.62 & @xmath87 + ngc 3610 & mp & 0.27 & 1.51 & @xmath86 & 0.018 & 0.14 & 3.09 & @xmath88 & 1.55 & 1.66 & 0.17 & 1.81 & 1.30 & 2.29 & 2.02 & 0.017 & 0.051 & 0.70 & 0.82 & 1.33 & 0.46 & @xmath89 + ngc 1052 & mp & 2.06 & 2.13 & @xmath90 & @xmath91 & 0.09 & 3.11 & @xmath92 & 0.82 & 0.92 & 0.87 & 1.86 & 0.74 & 2.39 & 2.80 & 0.051 & 0.094 & 1.68 & 1.61 & 1.07 & 0.95 & @xmath93 + ngc 7457 & mp & 2.20 & 1.95 & @xmath94 & @xmath95 & 0.59 & 3.00 & @xmath78 & 1.03 & 0.82 & 0.62 & 1.25 & 0.59 & 2.17 & 2.14 & 0.036 & 0.083 & 0.93 & 1.47 & 1.16 & 0.69 & @xmath87 + vcc 1087 & mp & 2.66 & 2.31 & @xmath96 & @xmath97 & 0.25 & 2.46 & @xmath98 & 1.57 & 0.95 & 0.42 & 1.12 & 0.44 & 2.35 & 1.86 & 0.029 & 0.077 & 0.91 & 1.24 & 1.10 & 0.43 & @xmath34 + fornax & mp & 4.00 & 2.80 & @xmath99 & @xmath100 & 0.19 & 1.23 & 2.14 & 2.35 & 0.41 & 0.15 & 0.87 & 0.01 & 2.56 & 0.85 & @xmath101 & 0.020 & 0.54 & 0.62 & 0.61 & 0.34 & @xmath102 + ngc 1407 & mr & @xmath103 & 0.36 & 0.117 & 0.158 & 1.07 & 5.16 & @xmath104 & @xmath105 & 4.17 & 1.29 & 3.06 & 3.08 & 1.72 & 4.31 & 0.091 & 0.235 & 4.21 & 2.39 & 2.28 & 1.49 & @xmath106 + ngc 4365 & mr & @xmath107 & 0.18 & 0.095 & 0.131 & 0.97 & 4.85 & @xmath108 & @xmath109 & 4.32 & 1.13 & 3.02 & 3.72 & 1.67 & 4.93 & 0.091 & 0.236 & 4.05 & 2.73 & 1.87 & 1.50 & @xmath110 + ngc 4594 & mr & @xmath111 & @xmath98 & 0.118 & 0.155 & 1.26 & 5.00 & @xmath112 & @xmath113 & 4.66 & 1.19 & 3.43 & 3.79 & 1.69 & 3.93 & 0.033 & 0.159 & 3.69 & 2.62 & 2.53 & 1.66 & @xmath114 + ngc 3610 & mr & 1.03 & 0.34 & 0.007 & 0.031 & 0.82 & 3.77 & @xmath115 & 0.32 & 3.82 & 1.64 & 3.82 & 3.19 & 1.32 & 5.38 & 0.055 & 0.158 & 3.00 & 2.58 & 2.65 & 1.46 & @xmath116 + ngc 1052 & mr & @xmath117 & 0.06 & 0.007 & 0.024 & 0.98 & 3.14 & @xmath87 & 0.49 & 2.48 & 0.79 & 2.23 & 2.12 & 2.43 & 3.50 & 0.069 & 0.203 & 2.80 & 2.71 & 1.60 & 0.86 & @xmath118 + ngc 7457 & mr & @xmath119 & 0.23 & 0.016 & 0.028 & 0.87 & 4.12 & @xmath120 & @xmath121 & 2.68 & 1.13 & 1.87 & 1.96 & 1.95 & 3.22 & 0.071 & 0.156 & 2.27 & 1.72 & 1.72 & 1.21 & @xmath122 + ngc 1407 & i m & 0.63 & 1.27 & 0.020 & 0.056 & 0.61 & 4.24 & @xmath123 & 0.18 & 2.97 & 1.04 & 2.48 & 1.60 & 1.97 & 3.97 & 0.047 & 0.143 & 2.52 & 1.80 & 1.59 & 1.05 & @xmath124 + ngc 4365 & i m & @xmath125 & 0.83 & 0.054 & 0.086 & 0.70 & 4.23 & @xmath126 & 0.07 & 3.32 & 0.91 & 2.57 & 2.48 & 1.99 & 3.86 & 0.067 & 0.177 & 2.96 & 1.96 & 1.64 & 1.20 & @xmath127 + ngc 1407 & mp & 0.18 & 0.13 & 0.005 & 0.007 & 0.10 & 0.17 & 0.18 & 0.11 & 0.26 & 0.13 & 0.20 & 0.30 & 0.12 & 0.26 & 0.003 & 0.003 & 0.13 & 0.15 & 0.17 & 0.13 + ngc 4365 & mp & 0.15 & 0.12 & 0.006 & 0.007 & 0.08 & 0.20 & 0.44 & 0.17 & 0.21 & 0.10 & 0.16 & 0.23 & 0.14 & 0.21 & 0.003 & 0.003 & 0.08 & 0.05 & 0.12 & 0.11 + ngc 4594 & mp & 0.14 & 0.09 & 0.004 & 0.005 & 0.08 & 0.13 & 0.14 & 0.09 & 0.21 & 0.10 & 0.20 & 0.31 & 0.12 & 0.25 & 0.003 & 0.003 & 0.14 & 0.13 & 0.16 & 0.12 + ngc 3610 & mp & 0.29 & 0.18 & 0.008 & 0.010 & 0.16 & 0.25 & 0.26 & 0.15 & 0.41 & 0.19 & 0.28 & 0.40 & 0.15 & 0.34 & 0.003 & 0.004 & 0.16 & 0.17 & 0.21 & 0.15 + ngc 1052 & mp & 0.14 & 0.09 & 0.004 & 0.005 & 0.08 & 0.14 & 0.14 & 0.09 & 0.22 & 0.11 & 0.17 & 0.27 & 0.10 & 0.23 & 0.002 & 0.003 & 0.11 & 0.12 & 0.13 & 0.10 + ngc 7457 & mp & 0.13 & 0.09 & 0.004 & 0.005 & 0.07 & 0.13 & 0.13 & 0.08 & 0.19 & 0.10 & 0.15 & 0.23 & 0.09 & 0.20 & 0.002 & 0.002 & 0.10 & 0.11 & 0.13 & 0.09 + vcc 1087 & mp & 0.09 & 0.06 & 0.003 & 0.003 & 0.05 & 0.09 & 0.10 & 0.06 & 0.16 & 0.07 & 0.11 & 0.18 & 0.06 & 0.14 & 0.002 & 0.002 & 0.07 & 0.08 & 0.09 & 0.07 + fornax & mp & 0.11 & 0.07 & 0.003 & 0.004 & 0.06 & 0.10 & 0.10 & 0.06 & 0.15 & 0.07 & 0.11 & 0.17 & 0.06 & 0.13 & 0.001 & 0.002 & 0.06 & 0.07 & 0.08 & 0.06 + ngc 1407 & mr & 0.22 & 0.14 & 0.006 & 0.007 & 0.10 & 0.16 & 0.19 & 0.11 & 0.22 & 0.12 & 0.17 & 0.26 & 0.10 & 0.21 & 0.002 & 0.003 & 0.11 & 0.12 & 0.13 & 0.10 + ngc 4365 & mr & 0.17 & 0.12 & 0.006 & 0.006 & 0.07 & 0.14 & 0.17 & 0.10 & 0.18 & 0.08 & 0.13 & 0.25 & 0.09 & 0.19 & 0.003 & 0.004 & 0.10 & 0.11 & 0.15 & 0.09 + ngc 4594 & mr & 0.18 & 0.13 & 0.005 & 0.006 & 0.09 & 0.16 & 0.18 & 0.10 & 0.23 & 0.13 & 0.21 & 0.34 & 0.15 & 0.29 & 0.003 & 0.004 & 0.15 & 0.16 & 0.20 & 0.13 + ngc 3610 & mr & 0.26 & 0.19 & 0.008 & 0.009 & 0.15 & 0.27 & 0.27 & 0.16 & 0.35 & 0.17 & 0.22 & 0.41 & 0.15 & 0.30 & 0.003 & 0.005 & 0.17 & 0.17 & 0.20 & 0.15 + ngc 1052 & mr & 0.31 & 0.21 & 0.009 & 0.010 & 0.16 & 0.29 & 0.32 & 0.20 & 0.45 & 0.23 & 0.36 & 0.54 & 0.21 & 0.45 & 0.005 & 0.005 & 0.21 & 0.23 & 0.26 & 0.19 + ngc 7457 & mr & 0.25 & 0.17 & 0.007 & 0.008 & 0.12 & 0.21 & 0.24 & 0.14 & 0.31 & 0.15 & 0.25 & 0.36 & 0.13 & 0.29 & 0.003 & 0.004 & 0.15 & 0.17 & 0.19 & 0.14 + ngc 1407 & i m & 0.09 & 0.06 & 0.003 & 0.003 & 0.04 & 0.08 & 0.08 & 0.05 & 0.11 & 0.06 & 0.08 & 0.13 & 0.05 & 0.11 & 0.001 & 0.001 & 0.06 & 0.06 & 0.07 & 0.05 + ngc 4365 & i m & 0.11 & 0.08 & 0.003 & 0.003 & 0.05 & 0.10 & 0.13 & 0.08 & 0.14 & 0.07 & 0.12 & 0.18 & 0.08 & 0.17 & 0.003 & 0.003 & 0.08 & 0.10 & 0.10 & 0.07 + ngc 1407 & mp & 0.67 & 0.33 & 0.028 & 0.017 & 0.39 & 0.65 & 1.15 & 0.53 & 1.08 & 0.25 & 0.32 & 1.32 & 0.11 & 1.08 & 0.015 & 0.026 & 0.72 & 0.14 & 0.18 & 0.42 + ngc 4365 & mp & 0.34 & 0.27 & 0.017 & 0.021 & 0.24 & 0.41 & 1.12 & 0.43 & 0.64 & 0.27 & 0.40 & 0.54 & 0.29 & 0.39 & 0.009 & 0.007 & 0.17 & 0.10 & 0.41 & 0.31 + ngc 4594 & mp & 0.27 & 0.36 & 0.012 & 0.017 & 0.41 & 1.08 & 1.51 & 0.58 & 0.69 & 0.28 & 0.95 & 0.64 & 0.60 & 1.99 & 0.020 & 0.033 & 0.39 & 0.41 & 0.63 & 0.59 + ngc 3610 & mp & 1.48 & 0.21 & 0.045 & 0.045 & 0.17 & 0.66 & 0.81 & 0.46 & 1.02 & 0.13 & 1.05 & 1.60 & 0.27 & 2.04 & 0.015 & 0.012 & 0.48 & 0.54 & 0.35 & 1.08 + ngc 1052 & mp & 0.95 & 0.67 & 0.069 & 0.061 & 0.45 & 1.16 & 1.51 & 0.47 & 2.16 & 0.37 & 0.67 & 1.78 & 0.57 & 1.77 & 0.038 & 0.026 & 0.99 & 0.91 & 1.10 & 0.32 + ngc 7457 & mp & 0.82 & 0.48 & 0.014 & 0.024 & 0.36 & 0.75 & 0.94 & 0.35 & 1.02 & 0.38 & 0.56 & 0.68 & 0.31 & 1.26 & 0.016 & 0.025 & 0.58 & 0.43 & 0.46 & 0.59 + vcc 1087 & mp & 1.08 & 0.38 & 0.031 & 0.027 & 0.26 & 1.03 & 0.54 & 0.48 & 0.82 & 0.25 & 0.82 & 0.70 & 0.29 & 0.54 & 0.012 & 0.029 & 0.49 & 0.27 & 0.18 & 0.19 + fornax & mp & 0.91 & 0.59 & 0.016 & 0.012 & 0.07 & 0.79 & 1.40 & 0.71 & 0.85 & 0.06 & 0.51 & 0.41 & 0.20 & 0.60 & 0.002 & 0.012 & 0.23 & 0.30 & 0.18 & 0.14 + ngc 1407 & mr & 0.65 & 0.35 & 0.037 & 0.040 & 0.52 & 0.59 & 0.91 & 0.26 & 0.60 & 0.23 & 0.38 & 0.72 & 0.41 & 1.13 & 0.024 & 0.024 & 0.49 & 0.28 & 0.35 & 0.21 + ngc 4365 & mr & 0.54 & 0.40 & 0.032 & 0.036 & 0.16 & 0.44 & 0.71 & 0.41 & 0.56 & 0.19 & 0.44 & 0.66 & 0.26 & 0.65 & 0.009 & 0.012 & 0.20 & 0.25 & 0.21 & 0.18 + ngc 4594 & mr & 2.01 & 0.61 & 0.086 & 0.083 & 0.68 & 0.89 & 1.26 & 0.60 & 1.34 & 0.43 & 0.57 & 2.08 & 0.38 & 1.07 & 0.019 & 0.040 & 0.69 & 0.88 & 0.63 & 0.95 + ngc 3610 & mr & 2.06 & 2.42 & 0.029 & 0.067 & 1.15 & 1.16 & 1.28 & 0.87 & 1.56 & 0.95 & 2.13 & 2.11 & 0.29 & 0.96 & 0.023 & 0.019 & 1.03 & 0.24 & 0.89 & 0.47 + ngc 7457 & mr & 0.60 & 0.71 & 0.010 & 0.020 & 0.12 & 0.50 & 0.22 & 0.24 & 0.64 & 0.05 & 0.26 & 0.54 & 0.14 & 0.31 & 0.005 & 0.011 & 0.25 & 0.66 & 0.67 & 0.25 + ngc 1407 & i m & 0.87 & 0.38 & 0.038 & 0.039 & 0.15 & 0.50 & 0.69 & 0.41 & 0.40 & 0.26 & 0.64 & 0.64 & 0.20 & 0.52 & 0.013 & 0.027 & 0.48 & 0.26 & 0.21 & 0.27 + ngc 4365 & i m & 0.22 & 0.17 & 0.013 & 0.016 & 0.23 & 0.30 & 0.37 & 0.21 & 0.41 & 0.15 & 0.41 & 0.34 & 0.26 & 0.73 & 0.012 & 0.013 & 0.29 & 0.23 & 0.09 & 0.16 +

we study the liquid - solid transition in a collection of interacting particles moving through a dissipative medium under the action of a constant , spatially uniform external force , e.g. a charge - stabilized suspension in a fluidized bed or a flux - point lattice moving through a thin , current - carrying slab of type ii superconductor . the mobility of a given region in these systems is in general a function of the local concentration . we show that the structure factor peak is suppressed in an anisotropic manner as a result of this effect , resulting in a shift of the crystal - liquid phase boundary towards the crystal side . a nonequilibrium phase diagram is presented . epsf , 0_0 ( ) _ 0 1 _

cond-mat9811052

how a liquid freezes to form a solid is a problem of longstanding interest to physicists . freezing at thermal equilibrium can now be said to be quite well understood in terms of a first - principles order - parameter theory @xcite . the effect of steady nonequilibrium driving on the freezing transition has been studied in the context of _ sheared _ colloids @xcite . a uniformly driven state , with no gradients on the average , as would arise in steady sedimentation @xcite or the fluidized bed @xcite is conceptually simpler : unlike in shear flow , an order - parameter description of the mean state of the driven crystal is straightforward . nonetheless , we know of no studies , experimental or theoretical , of the effect of sedimentation on the liquid - solid phase boundary in a suspension settling uniformly under gravity , or of the freezing of moving flux lattices in the _ absence _ of quenched disorder . a study of freezing in the presence of uniform driving is also of interest as a problem in the general class of driven diffusive systems . most of the investigations in this field @xcite are either away from any equilibrium phase transition or in the neighbourhood of a second order phase transition . the effect of uniform driving on a first order phase transition has received very little attention @xcite ; the freezing transition , in particular holds interesting possibilities since the ordering is at non - zero wavenumber . in this paper we consider a collection of interacting particles in a liquid state , i.e. , with strong short - range translational correlations , and ask what happens to the liquid - solid transition in the presence of an imposed steady mean translational motion of the particles with respect to the ambient medium . our model should apply not only to charge - stabilized colloidal suspensions @xcite , but also to the effect of a steady current on the freezing transition of flux - point liquids @xcite in thin slabs of clean type ii superconductors . the main ingredient in our theory is the concentration dependence of the mobility of particles in a suspension , or of the flux - points in a superconductor . in colloidal suspensions , this dependence arises as a result of the hydrodynamic interactions between particles , while in the flux - liquid , it is the electromagnetic interaction that is primarily responsible for the effect . although our approach is fairly general , we use parameter values appropriate to the specific case of sedimenting colloidal suspensions in the rest of the paper . our main results are as follows : as the mean concentration of the suspension is increased , the hydrodynamic interaction causes it to settle more slowly @xcite . we show that this leads to a nonlinearity in the equations of motion whose effect is to reduce the height of the structure factor relative to its equilibrium value . the suppression is strongly anisotropic , largest for wavevectors along the direction of mean drift , and vanishing for wavevectors normal to it . the reduction is most pronounced at the location of the peak , which means the strong short - range correlations in the liquid , which are the precursor of translational order , are suppressed . this inevitably leads to a shift of the phase boundary towards the crystal side ( i.e. , favours the liquid over the crystalline phase ) , by an amount proportional to the square of the pclet number , in a manner reminiscent of a proposed mechanism @xcite for shear - induced melting . this nonequilibrium phase transition now assumes additional importance in the wake of recent experiments on charged fluidized beds @xcite where it should be possible to test our predictions . constraints on the types of systems to which our theory should apply are discussed at the end of section [ sec : conclude ] . the paper is organised as follows . section [ introfreezing ] gives an overview of equilibrium freezing . the equations describing the motion of a sedimenting suspension are derived in section [ freezingeomsection ] . in section [ freezingphasetrans ] we discuss in a general setting the prescription used in this paper to describe a nonequilibrium phase transition . the perturbative calculation leading to the equal time correlation function in the driven system is the content of section [ freezingperturbation ] , and our numerical results for the structure factor are given in section [ freezingstrucfac ] . in section [ freezingfreezing ] we derive our results for the freezing transition , and conclude in section [ sec : conclude ] . the equilibrium phase - transition from a liquid to a crystal is the spontaneous appearance of a static , spatially periodic modulation in the density field of the liquid . an order parameter theory of this transition @xcite . is obtained by writing down the free - energy cost of such a modulation as a functional of the density field . the free energy cost of an inhomogeneous density field @xmath0 relative to the uniform liquid state of density @xmath1 is given by @xcite & = & f / k_b t + & = & c(|- | ) ( ( ) - _ 0 ) ( ( ) - _ 0 ) + & & + , [ ryf ] where @xmath2 is the direct correlation function of the liquid . the second term in ( [ ryf ] ) is an ideal gas entropy , whereas the first comes from correlations in the interacting system . expanding the logarithm and stopping at the second order in the density fluctuation @xmath3 , gives the quadratic part of the free - energy in terms of the fourier modes @xmath4 as : = ^2 in terms of the dimensionless structure factor = . the density correlation function at equilibrium is related to the structure factor by the relation ( - ) = v _ 0 where v is the system volume . all information about temperature and mean density is encoded in the liquid - state direct correlation function @xmath2 . for a given @xmath2 , @xmath5 corresponding to the thermodynamically stable state , within mean - field theory , is the global minimum of @xmath6 $ ] , obtained in principle by solving @xmath7 , and choosing that solution for which @xmath8 is the smallest . at low densities and high temperatures , the global minimum is @xmath9 . for sufficiently dense , cold systems , this uniform liquid state is unstable relative to a spatially periodic @xmath10 , the crystal . terms in @xmath8 of order higher than quadratic in @xmath11 are clearly essential to give such a solution . although the best theories of freezing @xcite go the whole log , we shall sacrifice accuracy in favour of simplicity , and use a quartic polynomial approximation to @xmath8 . since we are looking for periodic density waves , it is convenient to work with @xmath12 , the weights of the bragg peaks at the reciprocal lattice vectors @xmath13 , defined by = _ g _ ( - ) . since the peak of the liquid - state structure factor at its first maximum @xmath14 is much higher than the subsequent peaks , we restrict ourselves to fourier amplitudes of @xmath15 close to this peak . this is equivalent to considering only the first shell of reciprocal lattice vectors for the crystal . only the form of the s(k ) near the peak matters then , and @xmath16 can be replaced by a lorentzian s^-1(g ) = a + ( g^2 - k_0 ^ 2 ) ^2 so that the peak height is s(k_0 ) = a^-1 . the landau free energy per particle is then given by & = & _ + & & - b _ 0 ^ 2 _ , , _ * g*_1 + * g*_2 + * g*_3,0 + & & + c _ 0 ^ 3 _ _ * g*_1 + * g*_2 + * g*_3 + * g*_4 , 0 . + [ flandauintro2 ] if @xmath17 , then free energy costs strongly discourage the fourier amplitudes to stray from the peak of the structure factor , i.e @xmath18 . the free energy can then be expressed as a function of a single scalar order parameter @xmath19 where @xmath20 is the number of reciprocal lattice vectors in the set ( i.e. 6 for a triangular lattice , 12 for bcc etc ) . note that at nonzero wavevector @xmath13 , @xmath21 and @xmath22 are identical . the free energy per particle per @xmath23 is thus a polynomial f = = a r^2 - b r^3 + c r^4 , [ flandauintrooneop ] with @xmath24 { \rho_0}^3 c. \label{landauparameterc}\end{aligned}\ ] ] [ landauparameters ] here @xmath25 and @xmath26 are numbers that depend on the type of lattice being considered : @xmath25 is the number of such triplets or triangles that each rlv belongs to , and @xmath26 is the number of non - planar diamonds that the rlvs can form among themselves @xcite . the value of @xmath27 at which the transition occurs is found by demanding that the free energy be zero at the crystalline minimum . this leads to the value of @xmath27 at the transition , a_c = . the cubic term favours lattice structures in which the rlvs form triangles in sets of three . the candidate lattices , within the one order parameter theory , are the ones whose rlvs correspond to edges of regular polyhedra or polygons with triangular faces , i.e. planar triangular ( whose rlvs form a pair of triangles ) , face centred cubic ( octahedron ) and icosahedral ( icosahedron ) lattices . the theory predicts that all crystals should be triangular lattices in two dimensions and body centred cubic ( which has a face centred cubic reciprocal lattice ) in three dimensions . given the lack of sophistication , the theory is reasonably successful as almost all two dimensional crystals are triangular lattices , and many , if not all elemental solids in three dimensions are indeed bcc , at least close to freezing @xcite . the theory discussed above has several drawbacks : it neglects all structure in s(k ) except that at the peak and assumes that all wavevectors have the same order parameter amplitude . these assumptions can be relaxed and a more complete theory written down , but the landau expansion itself is questionable , in fact inaccurate , since the order parameter jump at freezing is large . it would doubtless be better to generalize the first - principles density - wave approach of ramakrishnan and yussouff @xcite to include the effects of sedimentation . in the present work , however , we have attempted only a nonequilibrium generalizations of the simple and intuitive landau - alexander - mctague theory , as a first step in taking into account the effect of sedimentation on freezing . an extension of our approach to more realistic freezing theories should be fairly straightforward . the frictional force on a spherical particle of radius @xmath28 moving slowly in a fluid with shear viscosity @xmath29 with a uniform speed @xmath30 is given by @xcite = 6 r v [ stokeslaw ] the dimensionless parameter which gives a measure of the relative importance of sedimentation with respect to brownian motion is the pclet number , defined as the ratio of the time taken by the particle to diffuse a characteristic distance @xmath31 to the time taken to settle the same distance under gravity , without diffusing . in terms of the mean settling speed @xmath32 of the particle , the pclet number is given by pe = . the appropriate dimensionless measure ( which we will continue to call the pclet number ) of the relative importance of drift versus diffusion for this problem , will turn out to be proportional to the pclet number defined at the scale of an interparticle spacing , but will incorporate certain collective effects whose nature will become clear below after we construct our equations of motion . although the single particle problem was solved in 1851 @xcite , the sedimentation speed of a suspension was theoretically obtained as recently as 1972 @xcite , and that too in the dilute limit . the speed of a strictly random suspension of volume fraction @xmath15 turns out to be _ less _ than that of a single sphere : v _ = ( 1 - 6.55 + ( ^2 ) ) v _ [ hinderfluid ] thus , at a volume fraction of @xmath33 , the settling speed is reduced to @xmath34 of the stokes value . it has been also observed experimentally @xcite that the mean settling speed of a suspension is less than that of a single particle , and a decreasing function of the concentration of suspended particles . in other words , the mobility of a region is a function of the local concentration . the sedimenting suspension that we speak of is an idealized bottomless tube of infinite height containing a solvent in which there is a steady downward flow of particles ( denser than the solvent ) . an ` infinite ' tube is realized in real life by means of a setup known as a fluidized bed @xcite , a vertical tube through which fluid is pushed up at a steady rate by an externally maintained pressure head , with particles suspended stably on the average by the balance between gravity and the viscous drag of the upflow of fluid . the problem of the steadily sedimenting particle in the comoving frame of reference is identical , in the infinite container limit , or far from the walls of the container , to the fluidized bed problem in the lab frame . for very dilute solutions , the reduced weight of the particle ( i.e. the weight minus the buoyant force ) is balanced by the single particle stokes friction . at higher concentrations , hydrodynamic interactions between the spheres @xcite becomes important , leading to a hindered settling speed , the effect of which we investigate . consider now a collection of interacting particles , on the average in a liquid state with uniform number density @xmath1 , moving through a frictional medium . in the absence of external forces ( such as gravity ) the number density @xmath0 follows the equation : = ^2 + , [ eomeq ] where @xmath35 is the collective diffusion constant , @xmath29 is a gaussian white noise consistent with thermal equilibrium : = 2 k_b t ^2 ( - ) ( t - t ) , [ noiseeq ] and @xmath8 is the free - energy functional defined in [ ryf ] . this equation is simply the equation of continuity = - . with a current with two pieces : a mean @xmath36 and a fluctuating piece @xmath37 , uncorrelated in space and time . the chemical potential @xmath38 is related to the gibbs free energy @xmath8 by the thermodynamic relation @xmath39 which gives equation ( [ eomeq ] ) with a conserving noise @xmath40 . note that the equation of motion [ eomeq ] combined with the noise statistics ( [ noiseeq ] ) will take the system to an equilibrium state with a probability distribution p ~e^- f / k_b t. in the presence of a constant external force , which we take for specificity to be a uniform gravitational field @xmath41 the particles sediment with a steady mean drift velocity @xmath42 @xcite , where @xmath43 is the concentration - dependent mobility , which in the low density limit , is expected to go to the stokes value @xmath44 the current then acquires an extra piece @xmath45 leading to an additional term of the form - m g [ extraterm ] on the right hand side of equation ( [ eomeq ] ) . here @xmath46 is the reduced or archimedean weight of the solute particles , mg = m_0 g - m_g , with @xmath47 being the mass of the particle , and @xmath48 the mass of solvent displaced by submerging it . let us now consider concentration fluctuations in the suspension around the mean density @xmath49 : @xmath50 taylor expanding the mobility about the mean value @xmath51 to second order , we find the equation of motion for the density field to be & + & mg ( + 0 ) + & = & ^2 - + , [ eom0 ] where @xmath52 \nonumber \\ & = & \frac{\muo m g}{\r0 } \ ; { \left [ \frac{{\partial}^2 { \mub}}{\partial { \lr}^2 } \right]}_{\r0}. \label{mutwodef}\end{aligned}\ ] ] the left hand side of ( [ eom0 ] ) is simply the eulerian derivative for a reference frame moving with a velocity _ = g ( + ( ) _ 0 ) ( ) . in the rest of this paper , we will do all our calculations in this reference frame , so that the left hand side of ( [ eom0 ] ) can be replaced by a time derivative . note that this frame is _ not _ the frame in which the mean velocity of the particles is zero . for a colloidal suspension , since @xmath43 decreases on increasing concentration , the frame moves with a velocity less than the mean velocity . thus one would see a net drift along gravity in this frame . however , the quantities of interest to us , i.e. , equal time correlation functions , are invariant under a galilean transformation , so it does nt matter in which frame we do our calculation @xcite . before beginning the calculations , we change variables to the non - dimensionalized fluctuation of the concentration from the mean , @xmath53 , and the dimensionless free energy @xmath54 , in terms of which the equation of motion in the comoving frame becomes @xmath55 [ freom ] where = is the collective diffusion constant , and = - m g _ 0 is a velocity scale set by the stokes velocity modified by the concentration - dependence of the mobility . note that the piece @xmath56 outside the brackets is the mean settling speed . the term within the brackets is a dimensionless number that vanishes in the dilute limit . for sedimenting suspensions , the mobility is a decreasing function of concentration , so @xmath57 since @xmath58 and @xmath59 are a diffusion constant and a velocity respectively , together they give the appropriate dimensionless coupling ( which we will call the pclet number @xmath60 ) for the problem : @xmath61}_{\r0 } , \label{peclet2}\end{aligned}\ ] ] [ peclet ] where @xmath31 is the average interparticle spacing in the suspension . note that we are using @xmath31 rather than the particle size as the length scale to define the pclet number . this is reasonable since we are considering collective rather than single particle effects . in terms of the single particle diffusion constant @xmath62 , which is related to the bare mobility by the einstein relation @xmath63 , the pclet number can be written as pe = ( ) ( ) _ 0 ( ) . [ peclet3 ] the last piece ( in parentheses ) in ( [ peclet3 ] ) is the particle - size pclet number for a single sphere , which is simply the ratio of the gravitational potential energy drop across the size of the particle to the thermal energy @xmath23 . the other three terms are dimensionless numbers that bring in the collective effects . the hydrodynamic interactions appear through the third piece in [ peclet3 ] ( in square brackets ) . it should be noted that this term is indeed nonzero even if the mobility is a linear function of concentration . we pause here to do some power - counting . the change in mobility due to a number density fluctuation depends on the change in _ volume fraction_. the taylor expansion of mobility , for example , is given by @xmath64 $ ] in @xmath65 dimensions , where @xmath66 is a dimensionless number that can be found from experiments . to the lowest order in the polyball size r , therefore , the piece in square brackets in [ peclet3 ] is proportional to @xmath67 . the pclet number thus goes as @xmath68 for particle of mass @xmath69 and radius @xmath28 . as @xmath70 , the hydrodynamic interaction between the spheres goes away , and the pclet number goes to zero as @xmath71 if we keep the mass fixed , and as @xmath72 if we change the radius of the particle keeping the density fixed . the equation of motion ( [ freom ] ) of the density fluctuation is the equilibrium dynamical equation with an additional nonlinearity of the burgers type @xcite . qualitatively , the nonlinear term may be understood in terms of a lattice gas model @xcite , where particles are driven in one direction by a constant field , but a particle is allowed to hop to the next site only if it is empty ( no double occupancy ) . this microscopically models a mobility depending on the local density in regions where the density is large , the probability of hopping being obstructed by the presence of other particles is larger , and hence the mobility smaller . the current at site @xmath73 is then proportional to @xmath74 , where @xmath75 is the discretized density in a bin at @xmath73 . on coarse graining this will give two terms : a term @xmath76 , which can be removed by a galilean transformation ; and another @xmath77 , which persists even in the comoving frame . before moving on to use ( [ freom ] ) to calculate the correlations in the suspension , let us briefly discuss the limitations of our model . first , we have only considered one effect of sedimentation , namely the dependence of concentration on mobility . this is a coarse - grained description of the hydrodynamic interactions , reliable only at long wavelengths . a better theory should consider , for example , gradient - dependence in the mobility , i.e. , the diffusivity at wavenumber @xmath78 should have corrections of @xmath79 for a suspension of particles of size @xmath28 . since we are interested in ordering phenomena at finite wavevectors @xmath80 , our theory should give reliable results for @xmath81 which is satisfied , for example , for strongly charge stabilized colloids where the ordering length scale @xmath82 is of order the screening length , @xmath83 . in addition , we have ignored the dependence of the mobility on more detailed aspects of the local distribution of particles , e.g. , its anisotropy , a point which we comment on at the end of this paper . secondly , we have also ignored the concentration - dependence of the diffusion constant @xmath58 in ( [ eomphi ] ) . strictly speaking , @xmath58 should have the same concentration dependence as @xmath84 , since the two are related by the einstein formula . a correct formulation of the problem with @xmath85 would require a multiplicative noise ( of the form @xmath86 $ ] where @xmath37 is a spatiotemporally white noise with unit variance ) to make the equations of motion consistent with equilibrium . this problem is technically much more difficult @xcite , and we believe it should nt change the results dramatically for the following reason . a concentration - dependent diffusion constant coupled with a consistent noise of the above type is an _ equilibrium _ effect , and an ( effective ) concentration independent @xmath58 with ordinary gaussian noise leads to the same equilibrium distribution . the nonlinear term in ( [ eomphi ] ) is the _ only driving term _ arising from concentration dependence . in any case , ( [ freom ] ) is the simplest driven - diffusive model with a liquid - solid transition and the correct equilibrium limit , and thus merits study in its own right . lastly , we have also ignored the long - ranged effects of the hydrodynamic interaction . our description corresponds to a quasi two - dimensional suspension confined between parallel walls with spacing @xmath87 . the walls cut off the hydrodynamic forces on lateral scales much larger than @xmath87 , making them effectively short range . the _ local _ effects of the hydrodynamic interactions are , however , accounted for by the concentration dependent mobility . to describe the three dimensional problem we have to explicitly include the the hydrodynamic velocity - field of the suspension in the description . although this could have far - reaching effects at small wavenumber , it should be relatively unimportant for the statics of crystallization , whose physics is dominated by wavenumbers near the structure - factor peak . since we are interested in the freezing transition , we set up the equation of motion ( [ freom ] ) for the density field , with a free energy functional that leads to a freezing transition in equilibrium . the simplest of these is the landau - alexander - mctague free energy functional which we have discussed in section [ introfreezing ] . a spatially inhomogeneous density fluctuation @xmath88 ( fourier transform @xmath4 ) has , relative to the uniform fluid , a free energy @xcite & = & + & & - b + & & + c + & & ( * p* ) ( - -*p* ) [ freeenalexmctague ] where @xmath89 is the equilibrium structure factor of the suspension . the density correlation function at equilibrium calculated using the above free energy is , to quadratic order , = _ 0 ( + ) . the equation of motion ( [ freom ] ) becomes in reciprocal space & & = - _ 0 k^2 + & & + i k_z + ( ) , [ eomkwithfeom ] + & & ( , t ) ( , t ) = 2 k^2 _ 0 ( + ) ( t - t ) . [ noisekspace ] [ eomkwithf ] using the form ( [ freeenalexmctague ] ) for the free energy , the equation of motion ( [ eomkwithfeom ] ) becomes & = & - + i k_z + & & + ( ) . [ eomnoneq ] where the contribution of anharmonic terms in the free energy have not been put in . this is because we are iterested in the _ corrections _ to the equilibrium bare correlation funtions due to the driving term . it is easy to see that the additional nonlinear term in equation ( [ freom ] ) or ( [ eomnoneq ] ) can not be expressed in a form so as to be absorbed into an additional term in the free energy . this is done by checking that the functional curl of the term is nonvanishing . we will call this term in ( [ eomnoneq ] ) the sedimentation nonlinearity . to describe phase transitions in this nonequilibrium problem , we follow earlier work on critical phenomena in the presence of shear @xcite and shear - induced melting of colloidal crystals @xcite . we assume that the fokker - planck equation that derives from the above langevin equation ( [ eomnoneq ] ) has a steady solution in the long time limit , which we write as p [ ] ~e^- [ probss ] and expand the functional @xmath90 in a taylor expansion = _ 2 ( 1,2)(1 ) ( 2 ) + _ 3 ( 1,2,3 ) ( 1 ) ( 2 ) ( 3 ) + where @xmath91 etc . are arguments , for example spatial positions @xmath92 etc . the summation convention is implied for repeated arguments . the @xmath93 are the vertex functions . the calculation of moments from this probability functional gives , for the connected parts , the relations @xcite @xmath94 [ corlsandvertex ] and so on , which are the usual relations between the connected correlation functions and vertex parts . the only difference here is that the steady state probability distribution is not the inverse exponential of the equilibrium free energy functional . the vertex functions @xmath95 above have to be therefore calculated from the steady state probability distribution @xmath96 . however , the correlation functions can be calculated from the langevin equation itself by demanding that the time derivative of all moments is zero , and these may be used in ( [ corlsandvertex ] ) to calculate the vertex functions upto any desired order , which gives the steady state @xmath96 to the same order in the taylor expansion . for instance , @xmath97 [ corlsandvertexk ] the second of these relations can be rewritten in terms of the structure factor as = - _ 3 ( , , ) _ 0 ^ 3 s(k_1)s(k_2)s(k_3 ) . [ threeptvertexwithsk ] at thermal equilibrium , i.e. , in the absence of driving , the vertex functions calculated to tree level are simply the inverse correlation function and the higher vertices in the free energy : @xmath98 [ corlsandvertexeq ] we calculate the perturbative correction to the vertices due to the sedimentation nonlinearity , from which we obtain the steady state probability distribution ( [ probss ] ) . this @xmath96 can be treated as a mean - field probability distribution which may be maximized to find the stable phase at any given temperature . in other words , @xmath99 $ ] can be treated as a nonequilibrium free energy . thus the perturbative corrections to the vertices can be interpreted as additional terms generated by sedimentation in an effective free energy . we start with the linearized equation of motion ( [ eomnoneq ] ) for the density field with only the sedimentation nonlinearity , = & - & + & + & i k_z + ( ) , and calculate perturbative correction to the vertex functions . the noise statistics is as given in ( [ noisekspace ] ) which , for momentum labels @xmath100 and @xmath101 reads ( , t ) ( - , t ) = 2 k^2 _ 0 v ( t - t ) [ noisekspacewithvol ] where @xmath102 is the volume of the system . to the lowest ( i.e. first ) order , there is a correction to the three - point vertex , whereas the structure factor gets modified at second order in pclet number . the four - point vertex @xmath103 will also have a correction at second order , but we ignore this correction as this is simply a stabilizing term and small changes in it will not affect the physics of freezing significantly . in this section , we set up a hierarchy of equations for the moments of @xmath15 , which we close by decoupling the four - point function into a product of two two - point functions ( hartree approximation ) . multiplying the equation of motion ( [ eomnoneq ] ) by @xmath104 , symmetrizing in @xmath105 and setting time derivative of the two - point function to zero , we obtain for _ equal time _ a relation between the two - point and the three - point functions : 2 & & = + i k_z & & + & + & 2 n k^2 , [ twopointequaltime ] where we have used the equal - time correlation between the noise and the density fluctuation , ( -,t ) ( , t ) = n k^2 , [ noiseandensityequaltime ] which is proved in detail in the appendix . here @xmath106 is the total number of particles in the system . since the correlation function is symmetric under interchange of @xmath100 and @xmath101 , ( [ twopointequaltime ] ) can be rewritten as & & = + & & + n k^2 , [ twopointequaltimeb ] similarly , multiplying with two fields @xmath15 at different wavevectors but equal times , we obtain a relation involving the third and fourth moments : & & = + & i & k_1z ( - ) + & & [ hier2 ] to obtain this equation we have used the fact that the correlation of two fields @xmath15 and one noise @xmath29 vanishes at equal time ( see appendix ) . if we proceed in this manner we will keep obtaining higher - order correlations in terms of the lower ones . we are , however , interested only in the lowest - order correlations , so we close the hierarchy by replacing the four - point function by a product of two two - point functions . only pairings of @xmath25 with @xmath107 or @xmath108 give nonzero contributions , because of the @xmath109 prefactor , so the dummy index @xmath110 on the right hand side of ( [ hier2 ] ) can only take the values @xmath111 and @xmath112 . the two terms thus obtained are identical , and we arrive at the following relation between the three - point and two - point functions : & & = + & & 2 i _ 0 ( + + ) s_0(k_1 ) s_0(k_2 ) s_0(k_3 ) . + [ threeptfnhartree ] using this in combination with ( [ threeptvertexwithsk ] ) we obtain the correction to the three - point vertex from sedimentation : _ 3 ( , , ) = - ( + + ) . [ correctiontogamma3 ] combining ( [ twopointequaltime ] ) with ( [ threeptfnhartree ] ) , we obtain the correction to the structure factor as an integral over wavevectors . in the limit of infinite volume , the sum over wavenumbers goes from @xmath113 to an upper cutoff determined by the smallest length scale in the problem . our coarse - grained model breaks down on scales of order the particle radius @xmath28 ; the upper cutoff is therefore @xmath114 . non - dimensionalizing the momentum argument in the integral with the @xmath115 outside , i.e. by multiplying all momenta by the average interparticle spacing @xmath31 , we finally obtain s ( ) & = & - 2 ( ) ^2 + & & _ 0^ + & & \ { } , [ skintegral ] where _ all _ the wavevectors in the above equation , including the @xmath78 s multiplying the integral , are non - dimensionalized by multiplying with @xmath31 . note that the upper cutoff and hence the integral goes to @xmath116 in the limit @xmath117 . let us calculate how the structure factor correction behaves in this limit . the integrand , for large arguments , @xmath118 ( note that @xmath119 for large @xmath25 ) , so by power counting , the integral @xmath120 . this apparent divergence is however controlled by the powers of @xmath28 in the pclet number , which was @xmath72 . we are thus left with a correction that @xmath121 , which vanishes as @xmath70 for all @xmath65 . this is consistent with the fact that in this limit there are no hydrodynamic interactions . the results of this section can also be obtained using diagrammatic perturbation theory @xcite . the correlation function calculated to one loop without self consistency and the three point vertex at tree level give results identical to those obtained above . we computed the structure factor of the sedimenting suspension by numerically computing the integral in eq . ( [ skintegral ] ) @xcite . we used a 40-point gaussian quadrature routine @xcite and took the large @xmath78 cutoff of the integral , @xmath122 to be 20 , corresponding to @xmath123 . the integral has reflection symmetry about @xmath124 , so that the structure factor at an angle @xmath66 from the horizontal ( @xmath125 ) is the same as that at an angle @xmath126 from the horizontal ( @xmath127 ) . this follows from the invariance of the integral under @xmath128 . the numerical results are presented in figs . [ fig2ddeltaskvk ] and [ figskvk2d ] . the first shows the fractional change in the structure factor per unit pclet number as a function of wavenumber and angle . the _ reduction _ happens to be a maximum at the peak itself , and is a monotonic function of the angle between the horizontal ( where it is maximum ) and the vertical ( where it goes to zero ) . in fig . [ figskvk2d ] we show the numerically computed structure factor in two dimensions . the equilibrium structure factor has been taken from monte carlo simulations of a two - dimensional system of particles interacting via a screened coulomb potential , just above freezing @xcite . the peak of the equilibrium @xmath129 was at @xmath130 ( which is close to the average interparticle separation @xmath131 in the units we are using ) and the peak value was @xmath132 . we have chosen the pclet number @xmath133 for these plots . this is a rather large number , and our calculation , which goes only upto leading order in perturbation theory is probably not accurate for such large driving . the large value has been chosen to make the deviations from the equilibrium structure easily visible . it is important to note that the split - peak in @xmath129 is a high pclet number effect . since the correction to @xmath129 is peaked at the same place as @xmath129 itself , there can be no shift in the position of the peak of the extremum . at low pclet number , therefore , the structure factor has a global maximum at @xmath134 , which becomes a local minimum at a sufficiently large @xmath60 . at what value of @xmath60 this happens depends on the angle , the required driving going to infinity as the horizontal direction is approached . we have studied the freezing transition only in two dimensions ( section [ freezingfreezing ] ) , but we present structure factor data for three dimensions as well ( figs . [ fig3ddeltaskvk ] and [ fig3dskvk ] ) . in three dimensions the integral in ( [ skintegral ] ) , and also the structure factor correction , have an azimuthal symmetry . in addition there is the reflection symmetry about @xmath124 as in the two - dimensional case . for this case we used a 30-point gaussian quadrature routine @xcite , and checked that the answer does not change significantly between 28-point and 30-point . the structure factor after the sedimentation correction is shown in fig . [ fig3dskvk ] . for the equilibrium structure factor we have used monte carlo results for a @xmath135 liquid @xcite , made continuous by suitable interpolation . the pclet number is once again @xmath136 . the effect of the corrected three - point vertex as well as the structure factor correction have to be considered while describing the liquid - solid transition in the presence of sedimentation . the latter has a relatively simple effect : since @xmath129 is reduced in all directions , there is a tendency towards melting . the correction ( [ correctiontogamma3 ] ) to the three - point vertex , being imaginary , has more interesting implications . combining ( [ correctiontogamma3 ] ) with ( [ corlsandvertexeq2 ] ) we conclude that the nonequilibrium effects shift the cubic vertex , b b - ( ) , where we have scaled all wavevectors by multiplying with the average interparticle separation @xmath31 . with reference to ( [ landauparameters ] ) , we can find the cubic term in the landau expansion for the nonequilibrium system , - b - b - ( 2 p n^-1/2 ) . [ bchange ] introducing a real dimensionless quantity which is first order in pclet number , _ 1 = 2 ( ) , [ beta1 ] we rewrite ( [ bchange ] ) as b b + i ( 2 p n^-1/2 ) _ 1 [ bchange2 ] the correction to the landau parameter @xmath137 is entirely imaginary , and makes the effective @xmath137 complex . the free energy is a real , positive quantity , so a complex coefficient @xmath138 in [ flandauintro2 ] would imply that the complex fourier amplitudes @xmath4 pick up phases . in equilibrium , when @xmath138 is real ( and positive by construction ) , the phases of @xmath139 and @xmath140 add up to @xmath141 so as to make the cubic term as negative as possible . in other words , if we write the amplitude as = || e^i ( ) then ( , , ) & = & ( ) + ( ) + ( ) + & = & 2 for sets of vectors that form triangles , i.e. satisfy @xmath142 . note that the cubic term is always real , whatever the phases of the wavevectors , since for any set \{@xmath143 } there is another with all wavevectors reversed which has exactly the opposite total phase @xmath144 . however the contribution to the free energy from two such terms is @xmath145 , which should be @xmath146 to minimize the free energy . with a complex coefficient as in ( [ bchange ] ) , the free energy is minimized if the phases satisfy ( , , ) + ( ) = 2 . this ` phase shift ' in the order parameters may be visualized in real space as a shift of one set of lattice planes relative to its equilibrium position . note that scattering experiments measure only the squared modulus @xmath147 , and will not see a phase shift . freezing is determined by the _ modulus _ of the modified coefficient of the cubic term , which is b . the prefactor multiplying @xmath148 is a number that depends on the type of lattice . for a triangular lattice in two dimensions , @xmath149 ( six reciprocal lattice vectors in the first shell ) and @xmath150 ( each reciprocal lattice vector belongs to one of the two triangles ) . [ modbcorrection ] clearly , the magnitude of @xmath137 can only increase due to sedimentation . thus , as far as this term is concerned , sedimentation seems to favour ordering . this effect has to compete with the reduction of the structure factor and only a quantitative analysis can decide their combined effect . in the rest of this section we present results for the simplest case , that of a planar triangular lattice . we have described the equilibrium freezing with a one - order parameter landau theory , choosing the landau coefficients to match the structure factor peak height at coexistence for our data , i.e. @xmath151 . our landau parameters were @xmath152 and , at coexistence , @xmath153 , giving a polynomial f = 0.10107 a r^2 - r^3 + 2.5 r^4 where @xmath154 is the reduced order parameter , defined in ( [ opr ] ) . note that the correction @xmath155 [ eqn . ( [ beta1 ] ) ] vanishes if the reciprocal lattice vectors remain on the shell @xmath156 , and the structure factor remains unchanged . this follows from ( [ bchange ] ) and the fact that the sum of @xmath157 has to be zero . thus , to benefit from the free energy reduction that an increase in @xmath137 can cause , the lattice may like to distort , thus sending one of the vectors out of the shell while keeping the sum equal to zero by rotating the other two towards or away from each other . we have considered two such distortions in our analysis , as shown in fig . [ figdistperp ] , which seem to be most likely from the symmetry of the problem . it is also imaginable that the lattice distorts to an asymmetric state with two degenerate minima of the free energy ; we have not made a complete scan of all possible distortions to rule out such possibilities , but looked at only the above two on grounds of plausibility . such distortions apart , the structure factor is itself affected by the driving , so @xmath155 is nonzero even without pushing the reciprocal lattice vectors out of the shell . the undistorted crystal is therefore also a strong candidate . the crucial issue is the relative importance of the two effects : @xmath155 and @xmath129 . though the correction to the three - point vertex is first order in @xmath60 , the correction to @xmath137 for small @xmath60 goes as @xmath158 as can be seen simply by binomially expanding the square - root in ( [ modbcorrection ] ) . for this reason , combined with the fact that the numerical value of @xmath155 is rather small at modest driving rates , the @xmath159 effect is negligible compared to the structure factor reduction . for the landau potential chosen , we have found numerically that the structure factor reduction makes the free energy function go unstable well before @xmath137 can change significantly . the physics , therefore , is dominated entirely by the structure factor . this implies that sedimentation will neccessarily lead to a shift of the liquid - solid phase boundary towards the solid side . in other words , since the correlations are reduced , a system which was at liquid - solid coexistence in the absence of sedimentation will choose the _ liquid _ phase if sedimentation is switched on . the reduction of the structure factor peak height is maximum along the sedimentation direction and falls monotonically as the angle with respect to the horizontal is decreased , going to zero in the horizontal direction . it is therefore expected that one of the wavevectors will point along the @xmath160 direction . this immediately rules out the possibility shown in fig . [ figdistperp](a ) in favour of that shown in fig . [ figdistperp](b ) . even if there is no distortion , this orientation is the best among the undistorted ones . we now ask : is there any reason the lattice may wish to distort due to structure factor effects alone ? consider the distortion sketched in fig . [ figdistperp](b ) the reciprocal lattice vector @xmath161 growing slightly larger , @xmath160 axis so as to keep the sum of the vectors zero . if the pclet number is not too large , @xmath129 continues to have a peak at @xmath162 even in the presence of sedimentation ( section [ freezingstrucfac ] ) , so that there is a free energy cost to pay for the expanding @xmath161 which is moving off the peak . the other two , however , have moved closer to the horizontal , i.e. in a direction such that the driving affects them less , and the peak height at these wavevectors is higher than it would have been had they not rotated at all . there is a competetion between these two effects , and its not _ a priori _ clear which will win , so we computed the free energy of distorted crystals at coexistence to find the most favourable value of the distortion or angle . note that in any case the liquid is the favoured phase , but we can still search among all the crystalline phases and find which of them has the lowest free energy . this , as we demonstrate shortly , gives us some hint as to what to expect at other temperatures . we found that for low pclet numbers the undistorted crystal is favoured , but at around @xmath163 a distorted crystal becomes favourable ( fig . [ figdeltavpe ] ) . there seems to be a first order jump in what we might call the distortion order parameters @xmath164 . this is not surprising , as we may regard the nonequilibrium free energy coming out of our analysis as a function of @xmath164 , whose power series expansion will have odd order terms since @xmath165 is not a symmetry of the problem . the distorted crystal appears at rather high pclet number ( fig . [ figdeltavpe ] ) , certainly well beyond the range of validity of our perturbation theory . the reason for the sudden distortion is probably related to the original peak @xmath162 becoming a minimum and two peaks developing on either side ( fig . [ figskvk2d ] ) . the several plateaus that appear at even higher @xmath60 are even more intriguing . the intriguing results regarding split peaks ( figs . [ figskvk2d],[fig3dskvk ] ) and distortions are however somewhat speculative in the absence of a reliable theory for large driving . we have also compared the configurations shown in figs [ figdistperp](a ) , and [ figdistperp](b ) and found , as expected , that the latter has a lower free energy at all values of the driving . thus one immediate consequence of our theory is that the crystal , when it forms , will form with the orientation shown in fig . [ figdistperp](b ) , i.e. with lattice planes parallel to the sedimentation direction . this agrees with intuition , since there are shocks @xcite travelling along the sedimentation direction . finally , one would like to look at the system below the equilibrium freezing temperature @xmath166 , and find how the melting pclet number depends on @xmath167 . strictly speaking this requires structure factor data in the supercooled liquid , which one should use in ( [ skintegral ] ) to compute the corrected @xmath129 . such data , however , is not readily available , and in particular for the case of the planar triangular lattice that we are considering , it is very hard to sustain a supercooled liquid for too long in a simulation with monodisperse systems . we therefore attempt the following approximate theory . near @xmath166 , the structure factor is that at coexistence plus a perturbative correction . the critical pclet number @xmath168 at which the crystal melts is an object which is itself perturbative in @xmath167 . in calculating @xmath168 , if we neglect the perturbative corrections to @xmath169 , we will be making an error only at a higer order . we can therefore legitimately do a landau theory at a @xmath170 below but close to @xmath166 using the structure factor _ at coexistence_. the @xmath129 corrections will all be calculated using this @xmath171 , the temperature entering only in the landau parameter @xmath27 . the result is shown in fig . [ sedpdg ] . we have shown that a uniform drift , as in steady - state sedimentation , reduces translational correlations in a suspension of interacting brownian particles , as a result of local hydrodynamic interactions that reduce the mobility of a dense region . this reduction is largest in the direction of sedimentation . at sufficiently high pclet number , the first peak of the structure factor splits due to the fact that the reduction is strongest where the initial correlation was largest . for the freezing transition we find that the phase boundary in a charge stabilized colloidal suspension shifts towards the crystal side . the shape of the phase boundary is quadratic in @xmath60 as shown in fig . [ sedpdg ] . let us now make some estimates for pclet numbers in realistic experimental situations to see whether the values of @xmath60 in fig . [ sedpdg ] are accessible . the pclet number for a dense suspension , at the scale of the particle radius , is given by eq . [ peclet3 ] , which can be rewritten as pe = ( ) ( ) _ 0 ( ) , [ pecletinremarks ] provided the hindered settling speed ( or the mobility ) is roughly linear in the volume fraction in the range of interest . taking the specific case of polystyrene spheres of diameter @xmath172 in water @xcite , we note that this is true for volume fraction between @xmath173 and @xmath174 @xcite . the volume fraction used in our calculations ( @xmath123 , from the cutoff used for the integrals , i.e. volume fraction @xmath175 ) falls within this range . > from the data presented in @xcite , the term within the square brackets in ( [ pecletinremarks ] ) is found to be of order @xmath176 . the first term @xmath177 in ( [ pecletinremarks ] ) is the ratio of the bare to the collective diffusion constant , which has a value of about @xmath178 near the freezing transition of a colloidal suspension @xcite . the ratio of the interparticle spacing to the radius of the polyballs , @xmath179 , is @xmath180 . the product of the first three factors in [ pecletinremarks ] is thus @xmath181 . finally , we have to estimate the bare pclet number @xmath182 . this turns out to be @xmath183 for polystyrene @xcite . the pclet number for the interacting suspension is therefore @xmath184 . this is rather large , which means that our lowest - order theory is probably not reliable here . it does however suggest that the effect of steady - state sedimentation on freezing should be large for this system . in general , in choosing a system on which to test our results , two competing needs must be kept in mind . on the one hand , the ordering length scale should be large compared to the particle size , so that the simple concentration - dependence of the mobility ( see the remarks towards the end of section [ freezingeomsection ] ) in our model is valid . on the other hand , the particle size itself should be large enough to give an appreciable peclet number . a comparison with shear induced melting is tempting , since a perturbation theory similar in spirit to the present approach @xcite is known to give a shift in the phase boundary that is quadratic in the shear rate . there are essential differences however : firstly the _ only _ significant effect of shear is to suppress correlations , whereas for a sedimenting suspension , there are nontrivial corrections to the three point vertex . the effect of these latter corrections happens to be negligible on the freezing transition as we have shown , making the phase boundary shift in a manner similar to the shear case . secondly , the sheared colloidal crystal at low shear - rates is a slightly distorted crystal essentially stuck in a statically sheared configuration . at low shear therefore ( where perturbation theory is valid ) , the crystal is essentially in an equilibrium state , and its properties can be derived from an equlibrium picture @xcite . for uniform driving , however , there are nonzero currents even at low pclet numbers , making it a more genuinely nonequilibrium problem . an interesting feature of our theory is that the sign of @xmath185 does not figure in the results . thus even if we have a suspension whose settling speed _ increases _ due to many body effects , our theory will still predict a reduction in the correlations and a consequent shift in the phase boundary towards the crystal side . finally , a few caveats . in @xcite we have shown that a sedimenting colloidal crystalline lattice is likely to be unstable at long wavelength . the theory proposed in this paper is probably nonetheless applicable to the formation of the crystalline phase at least at low pclet numbers , since the physics of freezing is centred primarily around the ordering wavenumber . the coupling of concentration and shear ( tilt ) considered in @xcite may also be important in the liquid state . the detailed consequences will be different because shear stresses have a finite lifetime in the liquid , in contrast to that in a solid where it lasts for ever if the strain is present . the effect of the sedimentation instability is thus very different in a liquid . this is an additional aspect that we have totally ignored in the above treatment , and one that a more sophisticated approach must address . however , further development of the theory should be motivated by experiments , which we eagerly await . consider a set of langevin equations of the form & = & f ( ) + _ , [ fraplang ] + & = & 2 k_b t _ ( t - t ) . [ frapnoise ] [ fraplangandnoise ] where @xmath186 stands for @xmath187 , and @xmath188 can be any arbitrary function , in general nonlinear . we shall calculate the field - noise correlations using novikov s theorem @xcite , which states that for any functional of the noise @xmath189 $ ] , a _ ( t ) = 2 t _ . [ novikovthm ] we first calculate the average of the product of one field with one noise , which by ( [ novikovthm ] ) is _ ( t ) _ ( t ) = . the equation ( [ fraplang ] ) can be formally solved to give _ ( t ) = _ t_0^t dt f ( ( t ) ) + _ t_0^t dt _ ( t ) where the @xmath190 dependence in the first term is present through @xmath191 , which depends on the noise amplitudes at all past times but not those of the future . the derivative vanishes for @xmath192 by causality . to calculate the equal time value of this derivative , let us calculate it first for @xmath193 for infinitesimal positive @xmath194 . for @xmath195 , = _ t_0^t dt ( t ) + 1 . for @xmath196 , all the arguments @xmath197 in the integral are times before @xmath198 , so the functional derivative inside the integral , and hence the integral itself vanishes . thus the forward derivative is simply equal to @xmath199 . the backward derivative is zero , therefore = ( t - t ) . at equal time this is discontinuous , but we recall that the delta function noises are nothing but limits of gaussians , so the theta function is not really a discontinuous function at the origin , but rather rises smoothly from @xmath113 to @xmath199 taking the midpoint value @xmath200 at the origin . thus = , which , combined with ( [ novikovthm ] ) gives _ ( t ) _ ( t ) = k_b t _ . this can be shown with more rigour by starting from a gaussian distribution , calculating the equal time correlation and then taking the width of the gaussian to zero @xcite . in section [ freezingperturbation ] we have used an analogous result for reciprocal space variables . let us now consider the product of two fields and a noise . again , using ( [ novikovthm ] ) , we have _ ( t ) _ ( t ) _ ( t ) = _ ( t ) + _ ( t ) where we have used the chain rule for functional derivatives . now each of the partial derivatives @xmath201 , therefore _ ( t ) _ ( t ) _ ( t ) = _ ( t ) + _ ( t ) . in section [ freezingperturbation ] we do a perturbation theory about the liquid phase , calculating liquid state correlations , so the first moments of @xmath15 are all zero , thus making all averages of the form @xmath202 vanish . for completeness we explicitly check the vanishing of the first moment in the presence of driving . averaging ( [ eomnoneq ] ) over noise , and setting the time derivative of the field average to zero ( steady state ) , we have = i k_z . [ firstmomentaverage ] the second moment on the right is nonzero only if @xmath203 , i.e. @xmath204 , so that the first moment @xmath205 vanishes for all @xmath206 . but since @xmath15 is a deviation variable , its @xmath204 component is zero by construction . we would like to thank mustansir barma , pinaki majumdar , rajaram nityananda and rahul pandit for discussions and sushan konar for the use of interpolation programs . a special word of thanks is due to chinmay das who supplied structure factor data from his monte carlo simulations @xcite . most of the work was done when r.l . was at the physics department , indian institute of science , bangalore . the facilities of the institue are acknowledged , including the the computational resources of the supercomputer education and research centre . was supported by csir ( india ) during the same period . ramakrishnan and m. yussouff , phys . rev . * b 19 * , 2775 , ( 1979 ) ; t.v . ramakrishnan , pramana * 22 * , 365 ( 1984 ) ; y. singh , physics reports * 207 * , 351 ( 1991 ) . b.j . ackerson and n.a . clark , phys . * 46 * , 123 ( 1981 ) ; phys . rev . * a 30 * , 906 ( 1984 ) ; physica * 118 a * , 221 ( 1983 ) ; h. m. lindsay and p.m. chaikin , j. phys . ( paris ) , colloq . * 46 * , c3269 ( 1985 ) ; h.m . lindsay , ph . d. thesis , university of california , los angeles , 1988 ( unpublished ) ; m.j . stevens , m.o . robbins and j.f . belak , phys . * 66 * , 3004 ( 1991 ) ; m. j. stevens and m. o. robbins , phys . rev . * e 48 * , 3778 ( 1993 ) . s. ramaswamy and s.r . renn , phys . 56 * , 945 ( 1986 ) ; b. bagchi and d. thirumalai , phys . rev . * a 37 * , 2530 ( 1988 ) . r. lahiri and s. ramaswamy , phys . lett * 73 * , 1043 ( 1994 ) ; r. lahiri , ph . d thesis , indian institute of science ( 1997 ) , chap . _ , _ colloidal dispersions _ , ( cambridge university press 1989 ) . r. blanc and e. guyon , la recherche * 22 * , 866 ( 1991 ) . m. a. rutgers , ph . d. thesis , princeton university 1995 ( unpublished ) . j. happel and h. brenner , _ low reynolds number hydrodynamics _ ( prentice - hall , englewood cliffs , n.j . , 1965 ) . m. a. rutgers et . al . , phys . rev . * e 51 * , 4674 ( 1995 ) . b. schmittmann and r.k.p . zia , statistical mechanics of driven diffusive systems , in _ phase transitions and critical phenomena _ , edited by c. domb and j. l. lebowitz , vol . 17 ( academic press , 1995 ) . see , however , m. acharyya , a. basu , r. pandit , and s. ramaswamy ( in preparation ) . a.k . sood in solid state physics , edited by h. ehrenreich & d. turnbull ( academic press , 1991 ) vol . g. blatter , m.v . feigelman , v.b . geshkenbein , a.i . larkin and v.m . vinokur , rev . phys . * 66 * , 1125 m.p.a . fisher , d.s . fisher and d.a . huse , phys . rev . * b 43 * , 130 ( 1990 ) ; s. sengupta , c.dasgupta , h.r . krishnamurthy , g.i . menon and t.v . ramakrishnan , phys . lett . * 67 * , 3444 ( 1991 ) ; g.i . menon , ph . d. thesis , indian institute of science , bangalore 1994 ( unpublished ) . * 69 * , 1715 ( 1992 ) . j.m . ham and g.m . homsy , int . j. multiphase flow * 14 * , 533 ( 1988 ) . s. alexander and j. mctague , phys . lett . * 41 * , 702 ( 1978 ) ; p.m. chaikin and t.c . lubensky , _ principles of condensed matter physics _ ( cambridge university press , 1995 ) . p.hansen and i.r . macdonald , _ theory of simple liquids _ ( academic , london , 1986 ) . g. g. stokes , trans camb phil soc . * 9 * , 8 ( 1851 ) . g. k. batchelor , j. fluid . * 52 * , 245 ( 1972 ) . p. mazur and w. van saarloos , physica * 115a * , 21 ( 1982 ) ; w. van saarloos and p. mazur , physica * 20a * , 77 ( 1983 ) . ladd , phys . fluids * a 5 * , 299 ( 1993 ) ; phys . lett * 76 * , 1392 ( 1996 ) . m. a. rutgers , ph . d. thesis , princeton university 1995 ( unpublished ) . kynch , trans . faraday soc . * 4 * , 166 ( 1952 ) . correlation functions calculated at unequal times do have a signature of the reference frame . this should in principle be observable in dynamic light scattering experiments . burgers , adv . appl . math . * 1 * , 171 ( 1948 ) . v. kumaran , j. chem . phys . * 104 * , 3120 ( 1996 ) . s. ramaswamy , unpublished . a. onuki and k. kawasaki , ann . ( ny ) * 121 * , 456 ( 1979 ) . d. j. amit , _ field theory , the renormalization group and critical phenomena _ , ( world scientific 1984 ) . r. lahiri , ph . d thesis , indian institute of science , 1997 , chap . 5 . strictly speaking , the integral should be performed in a region such that both @xmath100 and @xmath207 are within a sphere of radius @xmath208 . we have , for simplicity , simply integrated over all @xmath100 inside the sphere , i.e. our calculation covers a somewhat larger area of integration than the correct one . however , in the limit of large , though finite @xmath209 , the error incurred by this is negligible . w.h . press _ _ , _ numerical recipes _ , ( cambridge university press , 1992 ) . c. das , unpublished . the data has been obtained from 400 particles in a rectangular box . hansen and d. schiff , molecular physics * 25 * , 1281 ( 1973 ) . the results are based on systems with 894 particles in a cubic box with periodic boundary conditions . brewer , mem . sci usa * 2 * , 165 ( 1884 ) . w. van saarloos and d.a . huse , europhys . lett . * 11 * , 107 ( 1990 ) . fig . 1 of @xcite h. lwen , t. palberg and r. simon , phys . lett . * 70 * , 1557 ( 1993 ) ; a.v . indrani and s. ramaswamy , phys . lett . * 73 * , 360 ( 1994 ) . bare peclet number at the scale of particle radius : & = & + & = & in cgs units @xmath210 . for polystyrene , @xmath211 and @xmath212 , giving @xmath213 . r. lahiri and s. ramaswamy , phys . * 79 * , 1150 ( 1997 ) ; cond - mat 9610022 . crowley , j. fluid mech . * 45 * , 151 ( 1971 ) ; phys . fluids * 19 * , 1296 ( 1976 ) . novikov , sov . jetp * 20 * , 1290 ( 1965 ) . ma and g.f . mazenko , phys . b 11 * 4077 ( 1975 ) . c. zinn - justin , _ quantum field theory and critical phenomena _ ( clarendon press , oxford 1990 ) , chapter 3 .

in this work the second of a pair of articles we consider transport through spatially symmetric quantum dots with leads whose widths or positions do not obey the spatial symmetry . we use the semiclassical theory of transport to find the symmetry - induced contributions to weak localization corrections and universal conductance fluctuations for dots with left - right , up - down , inversion and four - fold symmetries . we show that all these contributions are suppressed by asymmetric leads , however they remain finite whenever leads intersect with their images under the symmetry operation . for an up - down symmetric dot , this means that the contributions can be finite even if one of the leads is completely asymmetric . we find that the suppression of the contributions to universal conductance fluctuations is the square of the suppression of contributions to weak localization . finally , we develop a random - matrix theory model which enables us to numerically confirm these results .

0906.0892

this is work the second of a pair of articles on mesoscopic transport through chaotic quantum dots with spatial symmetries ( see ref . @xcite for part i ) . in both works we use recent advances in semiclassical techniques to address the effect of spatial symmetries on weak localization ( wl ) corrections and universal conductance fluctuations ( ucfs ) . the aim of the first article was to identify the microscopic origin of properties that were earlier only known from phenomenological random - matrix theory ( rmt ) @xcite , and furthermore to extend the considerations to situations in which rmt is not easily applicable . in particular , this includes scenarios where symmetries are only partially preserved . to this end , the first article @xcite also considered the combined effects of magnetic fields , a finite ehrenfest time , and dephasing on symmetric systems and also discussed the reduction of symmetry - related interference effects by deformations of the dots . in the present paper , we contrast this ` internal ' symmetry breaking with symmetry breaking which is due to the position or shape of the leads ( for examples of such situations see fig . [ fig : asymmleads ] ) . we ask what happens to the transport if we take a symmetric dot coupled to leads which respect the symmetry , and then start moving one of the leads . in the fully symmetric situation , the magnitude of ucfs is doubled for each independent symmetry , while the weak localization correction can be either increased or reduced ( sometimes remain unaffected ) depending on the spatial symmetry in question @xcite . are these symmetry - induced effects modified when the leads are deformed or displaced ? if so , are they sensitive to displacement on a quantum scale ( of order of a fermi wavelength ) or a classical scale ( of order of a lead width ) ? the present literature does not offer much guidance to answer these questions indeed , the knowledge on transport in spatially symmetric systems with displaced leads is rather limited . reference @xcite reports that the distribution of transmission eigenvalues of a left - right symmetric dot with completely asymmetrically - placed leads differs slightly from the distribution of a completely asymmetric dot . because the difference is small , symmetric systems ( such as stadium billiards ) with displaced leads are indeed often used as representatives of completely asymmetric systems ( see , e.g. , refs.@xcite ) . recent works of one of the authors , on the other hand , identify a huge conductance peak in weakly coupled mirror - symmetric double - dots which still remains large even when the leads are not placed symmetrically @xcite . a simple consideration of weak localization quickly convinces us that it could never be as robust as the above - mentioned huge conductance peak in double dots . in systems without spatial symmetries , weak localization is the counter - part of coherent backscattering particle conservation guarantees that one can not have one without the other . systems with spatial symmetries have addition coherent back- and forward - scattering contributions ( as discussed in the first of this pair of articles @xcite ) . these contributions rely on interference between paths that are related by spatial symmetry . if those paths do not both couple to the leads , they can not generate an interference contribution to conductance . thus , if we displace one lead so much that there is no intersection with its spatially symmetric partner ( @xmath0 in fig . [ fig : asymmleads ] ) then the contributions to coherent forward scattering due to the spatial symmetries must vanish . the precise distance by which one has to move the lead to substantially suppress the symmetry - related contributions depends on the detailed position dependence of the coherent forward- and backscattering peaks . in principle , these coherent interference patterns could oscillate on a scale of a wavelength , and thus one might imagine that a small displacement of that order would suffice . the calculations and numerical computations presented by us here show that this is not the case . instead , the coherent forward- and backscattering peaks have a width of order the lead width , and do not oscillate on the scale of a wavelength . these considerations entail that the displacement of leads in internally symmetric systems offers a unique means to study coherent forward- and backscattering processes . from photonic systems it is known that the shape of the coherent backscattering cone provides valuable information on the multiple scattering in a sample @xcite . based on the results of the present work , transport measurements with gradually displaced leads promise to give similar insight into the dynamics of electronic systems . this work is organized as follows . section [ sec : backgrd ] introduces notation and provides a condensed review of the basic semiclassical concepts elaborated in more detail in the first of this pair of articles @xcite . the following sections describe the consequences of displaced leads for the weak localization correction in systems with left - right symmetry ( sec.[sect : lr ] ) , inversion symmetry ( sec . [ sect : inv ] ) , up - down symmetry ( sec . [ sect : ud ] ) and four - fold symmetry ( sec.[sect:4f ] ) . in sec . [ sect : ucfs ] we study the magnitude of universal conductance fluctuations for all types of symmetry . finally , in section [ sect : rmt ] we generalize the phenomenological rmt model of symmetry breaking ( presented in @xcite ) to the case of displaced leads , and compare the results of numerical computations to the semiclassical predictions . our conclusions are collected in section [ sect : conclusions ] . the appendix contains some further details on the semiclassical calculation of universal conductance fluctuations . to make this article self - contained we here first fix notation and then briefly summarize the main concepts of the theory of semiclassical transport in systems with spatial symmetries , developed in the first of this pair of articles @xcite . we consider chaotic quantum dots of size @xmath1 [ area @xmath2 and circumference @xmath3 which may possess any of the following three types of spatial symmetry ; a left - right mirror - symmetry , an inversion symmetry , and an up - down mirror - symmetry . we also consider four - fold symmetric systems which simultaneously possess all the above symmetries . the quantum dot is perfectly coupled to two leads , labelled left ( l ) and right ( r ) and carrying @xmath4 and @xmath5 modes , where @xmath6 for @xmath7 ( here @xmath8 is the fermi momentum ; we also denote the fermi velocity by @xmath9 ) . the quantum dynamics in the dot is characterized by a number of time scales , given by the time of flight @xmath10 between successive reflections off the boundaries , the dwell time @xmath11 , the dephasing time @xmath12 ( where @xmath13 is the dephasing rate ) , and a time scale @xmath14 on which a magnetic field destroys time - reversal symmetry . here , @xmath15 is a characteristic field strength at which about one flux quantum penetrates the quantum dot . in transport , the effect of a magnetic field is felt at a smaller magnetic field @xmath16 where @xmath17 is a system - specific parameter of order one @xcite . furthermore , the quantum - to - classical crossover is characterized by the open - system ehrenfest time @xmath18 $ ] and the closed - system ehrenfest time @xmath19 $ ] , where @xmath20 is the classical lyapunov exponent and @xmath21 is the fermi wavelength @xcite . in contrast to ref . @xcite , we here consider the possibility that the leads do not respect the symmetry of the dot . as shown in fig . [ fig : asymmleads ] , the displacement from the symmetry - respecting position is characterized by the overlap of leads under the relevant symmetry operation . for left - right mirror symmetry and inversion symmetry , this is the width @xmath22 of the intersection of a lead with the image of the other lead . an up - down symmetry maps each symmetry - respecting lead onto itself . the displacement of lead l ( r ) is then characterized by the width @xmath23 ( @xmath24 ) of the intersection of this lead with its own mirror image . in a four - fold symmetric system , the displacement is characterized by the various widths of intersections with respect to the individual symmetries ( @xmath25 for left - right mirror symmetry , @xmath26 for inversion symmetry , @xmath27 for up - down mirror symmetry of lead l and @xmath28 for up - down mirror symmetry of lead r ) . the semiclassical theory of transport @xcite expresses the transport through a quantum dot in terms of classical paths @xmath29 which connect point @xmath30 lead l to point @xmath31 on lead r. summing over lead modes as in ref . @xcite , the dimensionless conductance ( conductance in units of @xmath32 ) is given by @xmath33 where @xmath34 denotes the classical action of a path , and the amplitude @xmath35 is related to the square - root of the path s stability . for most pairs of @xmath36 and @xmath37 the exponential in eq . ( [ eq : g - doublesum ] ) oscillates wildly as one changes the energy or the dot - shape . thus they make no contribution to the average conductance ( where one averages over energy , dot - shape , or both ) . the contributions that survive averaging are those where the pairs of paths have similar actions @xmath38 for a broad range of energies and dot - shapes . in particular , this is the case for the `` diagonal contributions '' to the above double sum ( with @xmath39 ) , which can be analyzed using the sum rule ( in the spirit of eq . ( b6 ) of ref . @xcite ) @xmath40_{\gamma } \ ! & = & \ ! \ ! \int_{-\frac{\pi}{2}}^{\frac{\pi}{2 } } { \rm d } \theta_{0 } \int_{-\frac{\pi}{2}}^{\frac{\pi}{2 } } { \rm d } \theta\ , p_{\rm f } \cos \theta_0 \ \nonumber\\ & & \qquad \times \tilde{p } ( { \bf y } , { \bf y}_0 ; t ) \left [ \cdots \right]_{{\bf y}_0}. \ \ \label{eq : sum - rule } \end{aligned}\ ] ] here we define @xmath41 as the classical probability for a particle to go from an initial position and momentum angle of @xmath42 on lead l to within @xmath43 of @xmath44 on lead r in a time within @xmath45 of @xmath46 . the average of @xmath47 over an ensemble of dots or over energy results in a smooth function . if the dynamics are mixing on a timescale @xmath48 , one can approximate @xmath49 $ ] , which results in the classical drude conductance @xmath50 quantum corrections to this result originate from correlations of paths @xmath36 and @xmath37 which are not identical , but closely related by additional discrete symmetries in the system . for asymmetric quantum dots the only possible additional symmetry is time - reversal symmetry , which results in the ordinary weak localization correction @xcite and associated coherent - backscattering peak @xcite for systems whose classical dynamics exhibit hyperbolic chaos . the identification of possible pairings is also at the heart of the calculation of the magnitude @xmath51 of universal conductance fluctuations , which in the semiclassical theory naturally takes the form of a quadruple sum over classical paths @xcite . spatial symmetries in such chaotic systems induce further possible pairings both for the average conductance as well as for its variance , which are discussed in detail in the first article in this series @xcite . in the following sections we revisit these results and extend them to the case of displaced leads , which is far richer than the case of symmetry - respecting leads . we first consider a left - right mirror - symmetric system with leads that are ( partially or fully ) displaced from the symmetry - respecting configuration . as shown in fig . [ fig : asymmleads](a ) , the leads are of different widths and centred at different places . the amount of symmetry - breaking is characterized by the ( possibly vanishing ) width @xmath22 of intersection between lead l and the mirror image of lead r. in fig . [ fig : asymmleads - contributions ] we show the path - pairings for all symmetry - induced interference corrections to the average conductance . ( there is a strong resemblance between these contributions and the weak localization correction for systems with leads that contain tunnel barriers ; in particular compare the _ failed coherent forward scattering _ contributions in fig . [ fig : asymmleads - contributions ] of this article with the _ failed coherent backscattering _ contributions in fig . 4 of ref . @xcite . ) none of the contributions listed in fig . [ fig : asymmleads - contributions ] are particularly difficult to calculate using the method presented in the first of this pair of article @xcite . this method involves folding paths under the spatial symmetry to find ways in which one can construct pairings between paths or their images , with pairings switching at `` effective '' encounters ; see fig . [ fig : asymmleads - folds ] . the difficulty is to find all contributions . one crucial check is to verify that the sum of all interference contributions to transmission and reflection gives zero , thereby ensuring particle conservation . the main difference from the equivalent calculation for a system with symmetric leads ( cf . ref . @xcite ) is that here a pair of symmetry - related paths has a shorter _ joint _ survival time than the pairs of identical paths in the diagonal contribution . when the leads are symmetrically placed , the probability of a path staying in the dot ( not hitting a lead ) is strictly identical to the probability of its mirror image staying in the dot . this ceases to be the case when the leads are not symmetric . we deal with this by explicitly considering all situations where a path hits a lead ( in which case it escapes from the system ) or the mirror image of a lead ( in which its mirror image will escapes from the system ) . the probability that either of the processes occurs is @xmath52 per bounce at the boundary of the dot , where @xmath53 is the circumference of the dot . we therefore define a modified dwell time @xmath54 which characterizes the probability @xmath55 $ ] that a path _ and _ its mirror image are _ both _ still in the dot at time @xmath46 . we use this probability in place of @xmath56 $ ] in evaluating all parts of contributions 1 and 2 in fig . [ fig : asymmleads - contributions ] where the paths are the mirror image of each other . the contribution of paths of the type labelled 1 and 2i-2iv in fig . [ fig : asymmleads - contributions ] have an effective encounter close to a lead . these contributions are similar to certain contributions in an asymmetric system with tunnel barriers @xcite , and hence we use a similar method to analyze them here . the behavior of path @xmath37 is completely determined by that of path @xmath36 , so the two paths have the same amplitudes , @xmath57 . the action difference between them is @xmath58 , where @xmath59 is the component of @xmath60 which is perpendicular to the direction of path @xmath36 at @xmath61 @xcite . using the sum rule in eq . ( [ eq : sum - rule ] ) , we see that the contribution 1 in fig . [ fig : asymmleads - contributions ] is given by ( cf.contribution lr : a in ref . @xcite ) @xmath62 . \nonumber \end{aligned}\ ] ] the limits on the integral indicate that we only integrate over the region of the leads which have an overlap with each other under the left - right mirror symmetry ( the regions of width @xmath22 marked in fig . [ fig : asymmleads - contributions ] ) . the survival probability @xmath63 { \delta}r { \delta}\theta /[\pi ( w_{\rm l}+w_{\rm r}-w_\cap){\tau_{\rm d}}']$ ] is that of a path and its mirror image . the probability per unit time for path @xmath36 to hit within @xmath64 of a given point in the region of phase space defined by the union of leads and their mirror images is @xmath65 where @xmath66 . note that it is @xmath67 rather than @xmath68 which gives the decay rate of @xmath69 . we express the @xmath70 integral in terms of the relative coordinates @xmath71 and define @xmath72 and @xmath73 as the time between touching the lead and the perpendicular distance between @xmath36 and @xmath37 becoming of order @xmath74 and @xmath1 , respectively . for times less than @xmath72 , the path segments are almost mirror images of each other , and their joint survival probability is the survival probability of a path and its mirror image . for times longer than this the path - pairs escape independently , but since the pairs are made of a path and its mirror image , the escape rate is @xmath67 not @xmath68 . the @xmath46-integral in eq . ( [ eq : g_lr1 ] ) must have a lower cut - off at @xmath75 , because that is the minimum time for reconvergence . ( for shorter times there is no contribution , because path @xmath36 and @xmath37 must separate to a distance of order the dot size , if they are going to reconverge at the other lead ) . thus we have @xmath76 \over n_{\rm l}+n_{\rm r}-n_\cap } , \label{eq : int - of - p - asymleads}\end{aligned}\ ] ] where @xmath77 are shorthand for @xmath78 . note that the @xmath67 in the denominator of @xmath79 was cancelled when we integrated over all times longer than @xmath80 . for small @xmath81 we find @xmath82 , \qquad \label{eq : t'_l}\end{aligned}\ ] ] and @xmath83 is given by the same formula with @xmath1 replaced by @xmath74 . evaluating the integrals over the relative coordinates @xmath71 as in ref . @xcite , we finally obtain @xmath84^{-1 } \exp[-{\tau_{\rm e}^{\rm c}}/{\tau_{\rm d}}'].\end{aligned}\ ] ] the failed coherent forward - scattering contributions labelled 2i and 2ii in fig . [ fig : asymmleads - contributions ] come from the window of width @xmath85 in the l lead . this causes an enhanced probability of hitting the mirror image of that part of lead l. however , the lead r is not there , so this constructive interference peak gets reflected back into the dot , and has a probability of @xmath86 of going to lead r and a probability of @xmath87 of going back to lead l @xcite . the former is a contribution to transmission ( and hence to the conductance ) while the latter is a contribution to reflection . thus we have @xmath88 \over 2(n_{\rm l}+n_{\rm r}-n_\cap)(n_{\rm l}+n_{\rm r } ) } , \label{eq : g_lr:2i } \\ \langle { \delta}r\rangle_{\rm lr:2ii } & = & { ( n_{\rm l}-n_\cap ) \,n_{\rm l } \ , \exp[-{\tau_{\rm e}^{\rm c}}/{\tau_{\rm d } } ' ] \over 2(n_{\rm l}+n_{\rm r}-n_\cap)(n_{\rm l}+n_{\rm r})}.\end{aligned}\ ] ] by inspection of fig . [ fig : asymmleads - contributions ] it follows that @xmath89 is given by eq . ( [ eq : g_lr:2i ] ) with @xmath5 and @xmath4 interchanged , while @xmath90 . to evaluate the uniform contributions to transmission and reflection , labelled 3i and 3ii in fig . [ fig : asymmleads - contributions ] , we divide the pairs of paths in this contribution into three regions . the first part is when @xmath36 and @xmath37 are the same , and are far from the encounter ( a time @xmath91 or more away from the encounter ) . here the probability for the paths to escape is @xmath92 per unit time . the second region is where @xmath37 and @xmath36 are the mirror image of each other and far from the encounter ( a time @xmath91 or more from the encounter ) . here the probability of one or both paths to escape is @xmath93 per unit time . finally , the third region is close to the encounter ( less than a time @xmath91 away from the encounter ) . here the probability for the paths to escape the first time they pass through this region surrounding the encounter is @xmath94 $ ] . however , the conditional probability to escape the second time the paths pass through this region ( given that they both survived the first time ) is @xmath95 $ ] . it follows that the contribution 3i is given by @xmath96 \big\langle f ( { \bf y}_{0 } , \epsilon ) \big\rangle$ ] , where the action difference @xmath97 is the same as for weak localization in refs . @xcite and @xmath98 since the paths are paired with their mirror image between time @xmath99 and time @xmath100 , the survival rate is @xmath67 during this time , but it is @xmath68 at all other times . evaluating this integral with these survival times gives @xmath101 . \label{eq : f - result - asym}\end{aligned}\ ] ] this has two differences from the result for symmetric leads in ref . @xcite . the exponent contains @xmath67 not @xmath68 , and the prefactor contains @xmath102 not @xmath103 . when integrating over @xmath104 , we obtain a factor of @xmath105^{-1}\exp[-{\tau_{\rm e}^{\rm c}}/{\tau_{\rm d}}']$ ] in place of @xmath106^{-1 } \exp[-{\tau_{\rm e}^{\rm c}}/{\tau_{\rm d}}]$ ] . thus the @xmath67 in the prefactor is cancelled @xcite . evaluating the integrals , we get @xmath107^{-2 } \exp[-{\tau_{\rm e}^{\rm c}}/{\tau_{\rm d } } ' ] , \quad \ \\ \langle { \delta}r\rangle_{\rm lr:3ii } & = & - n_{\rm l}^2 [ n_{\rm l}+n_{\rm r}]^{-2 } \exp[-{\tau_{\rm e}^{\rm c}}/{\tau_{\rm d}}'].\end{aligned}\ ] ] these results are of the same form as the weak localization correction except that the exponent contains @xmath67 in place of @xmath68 . in particular , we recover the familiar factor of @xmath108 even though the joint survival time is reduced when the paths are mirror images of each other . one can next include other suppression effects such as asymmetry in the dot and dephasing , which we discussed for dots with symmetric leads in ref . the only difference caused by asymmetric leads is that now the parts of contributions affected by asymmetries and dephasing ( parts where paths are paired with their mirror image ) decay with a rate @xmath67 instead of @xmath68 . thus we find that all the contributions listed in fig . [ fig : asymmleads - contributions ] are then multiplied by a factor @xmath109 \over 1+(\gamma_{\rm asym}+\gamma_\phi){\tau_{\rm d } } ' } , \label{eq : z'_lr}\end{aligned}\ ] ] where the expression for the decay rates @xmath110 , @xmath13 and timescales @xmath111 , @xmath112 are the same as for a dot with symmetric leads @xcite . as required by particle number conservation , the seven contributions in fig . [ fig : asymmleads - contributions ] sum to zero . in order to obtain the conductance , we sum the four contributions to transmission from the left lead to the right lead ( contributions 1 , 2i , 2iii and 3i ) , and add them to the drude conductance and the weak localization correction . this gives the conductance of a chaotic left - right symmetric quantum dot with many - modes on each lead ( @xmath113 ) , @xmath114 + { \cal o}[n_{\rm l , r}^{-1 } ] , \nonumber\end{aligned}\ ] ] where @xmath115 is given by eq . ( [ eq : z_lr ] ) . the second term in the square brackets is the usual weak localization correction , which is suppressed by magnetic fields and dephasing according to the function @xmath116 \big[1 + ( b / b_c)^2+\gamma_\phi\tau_{\rm d}\big]^{-1}$ ] . for symmetric leads we have @xmath117 ( and hence @xmath118 ) , and this result immediately reduces to the one in ref . @xcite . it is worth considering two special cases . the first case is when the leads are of equal width but not centred at the mirror image of each other , such that @xmath119 . taking @xmath120 as the relative distance ( in units of the lead width @xmath121 ) by which lead l is displaced with respect to the mirror image of lead r , and assuming there is no dephasing , magnetic field , or internal asymmetry , we find @xmath122 \nonumber \\ & & + { \cal o}[n^{-1 } ] . \label{eq : move - lead}\end{aligned}\ ] ] the second special case is when the lead r is narrower but situated entirely within the mirror image of lead l ; we then have @xmath123 . assuming again that there is no dephasing , magnetic field , or internal asymmetry , @xmath124 \nonumber \\ & & + \ { \cal o}[n_{\rm l , r}^{-1 } ] . \label{eq : narrow - rlead}\end{aligned}\ ] ] as one could scan the narrow lead r across the mirror image of the wide lead l , this scenario can be thought of as a probe of the shape of the coherent forward - scattering peak . the fact that our result eq . ( [ eq : narrow - rlead ] ) is independent of the position of lead r tells us that the forward - scattering peak is uniformly distributed over the region defined by the mirror image of lead l. for systems with inversion symmetry the calculation follows much as for a left - right symmetry . the one significant difference is the magnetic - field dependence of the contributions , which was treated in ref . the displacement of the leads simply requires us to replace @xmath68 with @xmath67 in the suppression of contributions by magnetic fields , asymmetries in the dot , and dephasing . the suppression factor therefore takes the form @xmath125 \over 1 + ( b / b'_c)^2 + ( \gamma_{\rm asym}+\gamma_\phi){\tau_{\rm d}}'}. \nonumber \\ \label{eq : z'_inv}\end{aligned}\ ] ] where @xmath126 is given by eq . ( [ eq : bc ] ) with @xmath68 replaced by @xmath67 . as a result , an inversion - symmetric quantum dot with many modes on each lead ( @xmath113 ) has a total average conductance of @xmath127 + { \cal o}[n^{-1}_{\rm l , r}].\end{aligned}\ ] ] with the exception of the magnetic - field dependence of the second term , this formula is the same as eq.([eq : g - lr - final - result ] ) for a left - right symmetric dot . thus the two special cases discussed below eq . ( [ eq : g - lr - final - result ] ) are directly applicable here . for up - down symmetric systems , there are a number of important differences with the case of left - right symmetry discussed in section [ sect : lr ] . firstly , a pair of paths related by the mirror symmetry decays jointly at a rate @xmath128 where @xmath129 is the number of modes in the intersection of lead l with its own mirror image , and @xmath130 is the number of modes in the intersection of lead r with its own mirror image . secondly , the successful and failed forward - scattering contributions for left - right symmetry are converted into successful and failed backscattering contributions for up - down symmetry . in particular , successful backscattering makes no contribute to the conductance . the other contributions to transmission are not very different from those for left - right symmetry , except that one must distinguish @xmath129 from @xmath130 , and one must replace @xmath67 by @xmath131 . summing up the contributions to conductance induced by the spatial symmetry , we find @xmath132 \over ( 2n_{\rm l}+2n_{\rm r}-n_{\rm \cap l}-n_{\rm \cap r } ) ( n_{\rm l}+n_{\rm r})^2 } \nonumber \\ & & \times z'_{\rm ud } ( \gamma_{\rm asym},\gamma_\phi ) , \label{eq : de - g_ud}\end{aligned}\ ] ] where @xmath133 has the same form as @xmath134 given in eq . ( [ eq : z_lr ] ) , but with @xmath67 replaced by @xmath135 . like for left - right mirror - symmetry ( but unlike for inversion symmetry ) this contribution is unaffected by a magnetic field . the average conductance of an up - down mirror - symmetric dot with many modes on each lead is therefore @xmath136 \ z'_{\rm ud } ( \gamma_{\rm asym},\gamma_\phi ) \over 2n_{\rm l}+2n_{\rm r}-n_{\rm \cap l}-n_{\rm \cap r } } \nonumber \\ & & \qquad \qquad \qquad \qquad + \exp[{-{\tau_{\rm e}^{\rm c}}/{\tau_{\rm d } } } ] \ z_{\rm wl } ( b,\gamma_\phi ) \bigg ] \ + \ { \cal o}[n_{\rm l , r}^{-1 } ] . \label{eq : g_ud}\end{aligned}\ ] ] it is worth noting that the spatial symmetry induces a reduction of conductance whenever one lead is close to symmetric , even if the other lead is completely asymmetric ( i.e. when @xmath137 but @xmath138 , or vice versa ) . for example , when both leads have the same width ( @xmath139 ) and the right lead is perfectly on the symmetry axis ( @xmath140 ) , but the left lead is a long way from the symmetry axis ( @xmath141 ) , eq . ( [ eq : g_ud ] ) reduces to @xmath142 + \exp[{-{\tau_{\rm e}^{\rm c}}/{\tau_{\rm d } } } ] \bigg ] \nonumber \\ & & \qquad \qquad + \ { \cal o}[n^{-1}]\end{aligned}\ ] ] assuming no dephasing , magnetic field and no asymmetry in the dot . if the ehrenfest time is much shorter than @xmath68 and @xmath143 , the average conductance of the system with one displaced lead is therefore simply @xmath144 . remarkably , the conductance from the l lead to the r lead is therefore affected by the symmetry of the dot even when the l lead is completely asymmetric . this result is perhaps less counterintuitive when one considers reflection ( rather than transmission ) . if one lead is on the symmetry axis , then reflection back to that lead will be enhanced even if the other lead is a long way from the symmetry axis . since we have particle conservation , there must be an associated reduction in transmission from one lead to the other ( compared to transmission in a completely asymmetric situation ) . a quantum dot with four - fold symmetry simultaneously possesses all three of the spatial symmetries that we discuss in this article . the interference corrections to the conductance of such a system are simply the sum of the corrections due to each of these three symmetries ( i.e. , the presence of the extra symmetries has no effect on the contributions which do not respect those symmetries ) , @xmath145 where @xmath146 is the contribution to the average conductance induced by spatial symmetry @xmath147 . the explicit form of this result is easily extracted from the expressions in the previous sections . instead of writing it out in full , we consider the special case where the two leads have the same width , @xmath148 , the ehrenfest time is negligible and there is no dephasing , magnetic field or asymmetry in the dot . the average conductance then takes the form @xmath149 , \label{eq : delta - g-4fold}\end{aligned}\ ] ] where @xmath150 is the intersection between leads l and r under the left - right symmetry , @xmath151 is the intersection between leads l and r under the inversion symmetry , and @xmath152 ( @xmath153 ) is the intersection of lead l ( r ) with itself under the up - down symmetry . the final term in the square - bracket is the usual weak localization contribution . since the presence of two of the above mentioned symmetries always implies the presence of the third , it is not possible to move the leads such that only one of the @xmath154 parameters changes . without affecting the integrity of the leads there are only two possible modifications for which only two of the parameters change ; starting with perfectly symmetric leads one can ( a ) move both leads upwards by the same amount so that @xmath150 is unchanged , or ( b ) move both leads by the same amount in opposite directions ( one up and one down ) so that @xmath151 is unchanged . in principle , it is also possible to break up a single lead ( say l ) in the middle and move the two parts into opposite directions ( both parts would still be contacted by the same source or drain electrode ) ; this preserves @xmath129 and @xmath155 but affects the other parameters . however , the latter deformation is difficult to realize in practice . now we turn to the magnitude of universal conductance fluctuations ( ucfs ) in symmetric dots with asymmetric leads . their calculation is generally far more complicated than the calculation of the average conductance . this is illustrated by the fact that there is as yet no semiclassical theory of ucfs for leads with tunnel barriers , a problem which has many similarities to the problem we need to solve here . thus we restrict ourselves to the simplest case of quantum dots with negligible ehrenfest time and negligible dephasing , and only consider magnetic fields which are either negligibly small ( @xmath156 ) , or sufficiently strong to break time - reversal symmetry in the asymmetric system ( @xmath157 ) . the magnitude of the ucfs ( with conductances measured in units of @xmath158 ) is given by @xmath159= { \rm var}[t]$ ] , where @xmath160 $ ] and @xmath46 is the block of the scattering matrix @xmath161 associated with transmission from lead l to lead r. for practical calculations it is beneficial to exploit the unitarity of the scattering matrix ( i.e. , current conservation ) , which results in the relations @xmath162 with @xmath163 $ ] and @xmath164 $ ] , where @xmath165 is the block of the scattering matrix associated with reflection back to lead l , and @xmath166 describes reflection back to lead r. as a result we can write the magnitude of the ucfs in any of the following ways , @xmath167 = { \rm var}[r ] = { \rm var}[r ' ] = { \rm covar}[r , r'].\end{aligned}\ ] ] as for conventional ucfs without spatial symmetries @xcite , the semiclassical calculation of @xmath168 $ ] is most straight - forward , thus we base our calculations on this quantity . for the expert reader , appendix [ appendix ] contains an outline of the calculation of @xmath169 $ ] and @xmath170 $ ] , showing that they equal @xmath171 $ ] . all symmetry - induced contributions to @xmath168 $ ] for an up - down symmetric dot are listed in fig . [ fig : ucf - asymleads1 ] . for a left - right or inversion - symmetric dot there are _ additional _ contributions , which are listed in fig . [ fig : ucf - asymleads2 ] . in all cases , when paths 2 and 2 are not paired with each other , they are paired with the _ images _ of paths 1 and 1 under the appropriate symmetry operation . to keep the sketches in figs . [ fig : ucf - asymleads1 ] and [ fig : ucf - asymleads2 ] as clear as possible , we only show these images of paths 1 and 1 ( rather than paths 1 and 1 themselves ) . then the resulting contributions look very much like the usual contributions to ucfs in a system without a spatial symmetry @xcite . in analogy to the situation in asymmetric systems , one would also expect contributions in which paths wind around periodic orbits ( see figs . 1b , c in ref . for example , a symmetric quantum dot will have contributions in which path 1 is the same as path 1 except that it winds around a periodic orbit @xmath172 when path 1 does not ( thus path 1 must come very close to the periodic orbit in phase space ) , while path 2 is the same as path 2 except that it winds around the _ image _ of the orbit @xmath172 . these contributions are proportional to those analyzed for ucfs in asymmetric dots , the only modification being that the joint survival probability of a periodic orbit and its image is again changed to @xmath173 $ ] . drawing on the results of refs . @xcite , it follows that the contributions involving windings around periodic orbits will be negligibly small when the ehrenfest time is small . ( this observation makes the calculation of the ucfs in the present problem significantly simpler than for the case with tunnel barriers , where one can not rule out contributions from periodic orbits which touch the barriers on the leads . ) inspecting the sketches in figs . [ fig : ucf - asymleads1 ] and [ fig : ucf - asymleads2 ] we see that all contributions are doubled when the magnetic field is negligible , because path 2 can either follow the image of path 1 or the time - reverse of path 1. thus we can multiple all terms by @xmath174 , where @xmath175 for a system with negligible magnetic field , @xmath156 and @xmath176 for a system with a finite magnetic field , @xmath177 . in the latter case the presence of mirror - reflection symmetries allows one to define a generalized time - reversal symmetry ; however , this is already accounted for in the construction of all diagrams ( see appendix a of ref . @xcite ) . the general rules for constructing all contributions to the ucfs are the following . each segment where path 2 or 2 is paired with the image of path 1 or 1 gives a factor of @xmath178 , which arises from the survival time @xmath143 given in eq . ( [ eq : tau - updown ] ) . each segment where paths 2 and 2 are paired ( or paths 1 and 1 are paired ) gives a factor of @xmath179 , which comes from the conventional survival time @xmath68 . each segment that touches a lead gives a factor equal to the number of lead modes that the path could couple to ; i.e. , a lead labelled `` @xmath180 '' gives a factor of @xmath181 , while a lead labelled `` r '' simply gives a factor of @xmath5 . an encounter which touches a lead gives the same factor as a simple path - segment that touches a lead , so again if it is labelled `` @xmath180 '' then it gives a factor of @xmath182 ( this rule is proven by applying the same analysis as was used for the successful and failed forward - scattering processes in section [ sect : suc - and failed - forward ] . ) finally , encounters deep in the dot ( i.e. , those which do not touch the leads ) give a factor of @xmath183 ( this rule can be proven by applying the same analysis as was used for the uniform contributions to transmission in section [ sect : uniform - contrib ] ) . with this set of rules we can easily see that contribution ( i ) in fig . [ fig : ucf - asymleads1 ] gives @xmath184 next we see that @xmath185 , and that they are negative because only one of the encounters is deep in the dot ( the other is near a lead ) , resulting in @xmath186 finally @xmath187 gives a positive contribution because it has two encounters deep in the dot , and is given by @xmath188 the total magnitude of the ucfs is given by the ucfs of an asymmetric dot , @xmath189_{\rm asym}$ ] , plus the sum of the terms above , i.e. , @xmath189 = { \rm var}[g]_{\rm asym}+c_{\rm i}+c_{\rm ii}+c_{\rm iii}+c_{\rm iv}$ ] . in the limit of perfectly symmetric leads ( @xmath190 and @xmath191 ) , only @xmath187 survives and the ucfs have double the magnitude as those for an asymmetric dot . in the limit of completely asymmetric leads ( @xmath192 ) , one has @xmath193 , and the ucfs have the same magnitude as those for an asymmetric dot . to express @xmath189 $ ] for arbitrary @xmath4 , @xmath5 , @xmath129 , and @xmath130 , we find it beneficial to introduce the quantities @xmath194 and @xmath195 , where @xmath196 . making use of the fact that @xmath197 , we find @xmath198 & = & { \rm var}[g]_{\rm asym } \\ & + & { 2\over \beta}\ , n_{\rm l}^2 n_{\rm r}^2 \left ( { 1-(1-n_{\rm l})w_{\rm l } - ( 1-n_{\rm r})w_{\rm r}\over 1+n_{\rm l}w_{\rm l } + n_{\rm r}w_{\rm r } } \right)^2 \nonumber\end{aligned}\ ] ] where in this notation @xmath189_{\rm asym } = ( 2/\beta ) n_{\rm l}^2 n_{\rm r}^2 $ ] . in the special case where @xmath199 , displacing the leads suppresses the symmetry - induced contribution to ucfs by a factor @xmath200 ^ 2 $ ] . in terms of the original quantities @xmath4 , @xmath5 , @xmath129 , and @xmath130 , eq . ( [ eq : ucf - updown1 ] ) takes the form @xmath201 & = & { \rm var}[g]_{\rm asym } \\ & + & { 2\over \beta}\ , { n_{\rm l}^2 n_{\rm r}^2 \over ( n_{\rm l}+ n_{\rm r})^4 } \left ( n_{\rm r}n_{\rm \cap l}/n_{\rm l } + n_{\rm l}n_{\rm \cap r}/n_{\rm r } \over 2n_{\rm l}+2n_{\rm r}-n_{\rm \cap l } -n_{\rm \cap r } \right)^2 \nonumber \end{aligned}\ ] ] where @xmath189_{\rm asym } = ( 2/\beta ) n_{\rm l}^2 n_{\rm r}^2 ( n_{\rm l}+ n_{\rm r})^{-4}$ ] . comparison with eq . ( [ eq : g_ud ] ) shows that lead displacement suppresses the symmetry - induced contributions to ucfs by a factor that is the square of the suppression of the symmetry - induced contributions to the average conductance . for a systems with a left - right or an inversion symmetry , we once again find the magnitude of the ucfs by evaluating @xmath171 $ ] . for these symmetries , we must consider the contributions in fig . [ fig : ucf - asymleads2 ] in _ addition _ to those in fig . [ fig : ucf - asymleads1 ] . the origin of the extra contributions in fig . [ fig : ucf - asymleads2 ] is most clearly understood by considering the case of perfectly symmetric leads . then the left - right and inversion symmetries map lead l onto lead r , which means that if path 2 is paired with path 1 then path 2 will hit lead r when path 1 hits lead l ( meaning the image of path 1 hits lead r ) . one can thereby immediately see that the contribution @xmath202 in fig . [ fig : ucf - asymleads2 ] contributes to @xmath203 $ ] ( this was not the case for up - down symmetry , since there path 1 hits the same lead as the image of path 1 ) . for asymmetric leads a similar situation occurs . if path 1 hits the intersection region of width @xmath22 on lead l then its image hits lead r ; thus path 2 will also hit lead r if it is paired with 1 over this segment . the rules to evaluate each contribution are the same as for up - down symmetry , with now necessarily @xmath204 . using these rules , we find that @xmath205 summing these contributions and writing the result with the same denominator as eq . ( [ eq : ucf - updown2 ] ) gives @xmath206 adding this set of contribution to those already calculated in the previous section , we find that the ucfs of a left - right or inversion - symmetric dot with asymmetric leads is given by @xmath167 & = & { \rm var}[g]_{\rm asym } \nonumber \\ & + & { 2\over \beta}\ , { n_{\rm l}^2n_{\rm r}^2 \over ( n_{\rm l}+n_{\rm r})^4 } \left ( { n_\cap \over n_{\rm l}+n_{\rm r}-n_\cap}\right)^2 \label{eq : ucf - leftright}.\end{aligned}\ ] ] by comparing this with eq . ( [ eq : g - lr - final - result ] ) , we find that the suppression of symmetry - induced contributions to ucfs is the square of suppression of the symmetry - induced contributions to the average conductance ( just as we already found for an up - down symmetric system ) . for completeness , we now briefly discuss ucfs in a 4-fold symmetric dot with asymmetric leads . a 4-fold dot has all three of the symmetries discussed above . thus the ucfs in a four - fold symmetric system are given by the sum of all possible symmetry - induced contributions ( just as with symmetric leads @xcite ) . given the results in the preceding sections , the general formula is easily determined . here we give the result for the special case @xmath207 , @xmath167 & = & { 1\over 8\beta}\bigg [ \left({n_{\rm \cap lr}\over 2n - n_{\rm \cap lr}}\right)^2 + \left({n_{\rm \cap inv}\over 2n - n_{\rm \cap inv}}\right)^2 \nonumber \\ & & + \left({n_{\rm \cap ud : l}+n_{\rm \cap ud : r } \over 4n - n_{\rm \cap ud : l}-n_{\rm \cap ud : r } } \right)^2 + 1 \bigg]\end{aligned}\ ] ] where @xmath150 is the intersection between leads l and r under the left - right symmetry , @xmath151 is the intersection between leads l and r under the inversion symmetry , and @xmath152 ( @xmath153 ) is the intersection of lead l ( r ) with itself under the up - down symmetry . the final term in the square - bracket represents the usual ucfs for an asymmetric dot . note that the suppression of each symmetry - induced term goes like the square of the equivalent term in the average conductance , eq . ( [ eq : delta - g-4fold ] ) . in this section we compare the semiclassical predictions derived in the previous sections to numerical results obtained from a phenomenological random - matrix model . this model generalizes the construction discussed in section 9 of part i ( ref.@xcite ) . the general framework is the same as in part i : the conductance is obtained from the landauer formula @xmath208 $ ] , where @xmath46 is the transmission block of a scattering matrix @xmath161 given by @xmath209 here , @xmath210 is an internal unitary evolution operator of dimension @xmath211 while @xmath212 is an @xmath213 dimensional matrix specified below , and @xmath214 . in part i we assumed that the leads respect the geometrical symmetries , which allows to fully desymmetrize the system . one can then introduce a fixed form of the matrix @xmath212 and attribute the effects of symmetries solely to the internal dynamics ( the resulting rmt ensembles for @xmath210 are given in table 2 of ref.@xcite ) . it is clear that this full desymmetrization fails when leads are to be displaced . for up - down symmetry , for instances , desymmetrization identifies two effectively separate systems ( consisting of modes of even and odd parity ) which do not couple to each other . shifting lead modes in this representation has no effect since rmt is invariant under the permutation of matrix indices . a real displacement of leads , however , mixes the states of even and odd parity . the reason for this discrepancy is that leads are defined locally in real space , while parity is a global symmetry which connects remote parts of the system . no spatial sym . & coe(@xmath211 ) & cue(@xmath211 ) + left - right sym . & @xmath215 coe@xmath216(@xmath217 ) @xmath218 & @xmath215 coe(@xmath211 ) @xmath218 + inversion sym . & @xmath219 coe@xmath216(@xmath217 ) @xmath220 & @xmath221 cue@xmath216(@xmath217 ) @xmath220 + up - down sym . & @xmath222 coe@xmath216(@xmath217 ) @xmath223 & @xmath222 coe(@xmath211 ) @xmath223 + four - fold sym . & @xmath224[@xmath215 coe@xmath216(@xmath225 ) @xmath218]@xmath216@xmath227 & @xmath228 [ @xmath215 coe(@xmath217 ) @xmath218]@xmath216@xmath229 + + it is therefore necessary to define both the internal evolution operator @xmath210 as well as the coupling to the leads @xmath212 in a way which resembles modes in a real - space basis . in principle , this can be done , e.g. , based on the sinusoidal transverse mode profiles of a strip resonator . we adopt a similar , but more efficient procedure , whose principle idea is shown in fig.[fig : portmodel ] . the illustration shows an abstract scatterer with @xmath211 ports which serve as possible contacts to the system . for each lead we select @xmath230 ports ( with index @xmath231 for lead l and @xmath232 for lead r ) ; the remaining ports are closed off . the internal evolution operator @xmath210 describes the transport from port to port . the scattering matrix is then given by eq.([eq : dyns ] ) where @xmath233 . a crucial point of the illustration in fig . [ fig : portmodel ] is the numeration of ports , which are grouped into 4 segments that map in specific ways onto each other when symmetry operations are applied . ( i ) left - right symmetry maps segment 1 onto segment 3 and segment 2 onto segment 4 . ( ii ) up - down symmetry maps segment 1 onto segment 2 and segment 3 onto segment 4 . ( iii ) inversion symmetry maps segment 1 onto segment 4 and segment 2 onto segment 3 . ( iv ) four - fold symmetry maps all segments onto each other . in the basis of these ports , the explicit symmetries of the internal evolution operator @xmath210 are specified in table [ tab : f ] . since up - down and left - right symmetry are both manifestations of a reflection symmetry , they are now simply related by a interchanging segments 2 and 3 ( as described by the matrix @xmath53 defined in table [ tab : f ] ) ; this is a consequence of the fact that we do not fully desymmetrize the up - down symmetry ( the left - right symmetric case can never be fully desymmetrized because one has to keep track of the identity of the leads ) . a finite magnetic field breaks these symmetries , but still allows one to define a generalized time - reversal symmetry . similarly , for vanishing magnetic field , inversion symmetry is obtained from reflection symmetry by interchanging segments 3 and 4 ( as described by the matrix @xmath234 defined in the table caption ) . the slightly different systematics in the presence of a magnetic field arises because the orientation of the segments matters ; consequently , for inversion symmetry , time - reversal symmetry is effectively broken but the geometric symmetry itself is still present in the dynamics ( trajectories still occur in symmetry - related pairs ) . a convenient choice of a fully symmetry - respecting arrangement of leads which applies to all internal symmetries is given by @xmath235 where @xmath236 and @xmath237 . the case of a four - fold symmetry in principle allows two symmetry - respecting arrangements ( aligned along each of the two symmetry lines of reflection ) ; these two arrangements are equivalent in rmt and again related by a reshuffling of the 4 segments . the form of @xmath212 for generally placed leads is easily read off fig . [ fig : portmodel ] . figures [ fig : numerics1a ] ( for @xmath238 ) and [ fig : numerics1b ] ( for @xmath177 ) show how the weak localization correction and universal conductance fluctuations are affected when the leads are moved away from the symmetry - respecting positions . the degree of displacement is quantified by a variable @xmath239 ( @xmath240 in the symmetric arrangement , @xmath241 in the asymmetric arrangement ) . the data points are based on an ensemble average over @xmath242 rmt matrices with @xmath243 and @xmath244 , while the curves are the predictions of our semiclassical theory , which can be written as @xmath245\frac{1-\lambda}{1+\lambda } , \label{eq : dglambda}\ ] ] @xmath246\left(\frac{1-\lambda}{1+\lambda}\right)^2 . \label{eq : varglambda}\ ] ] starting from a four - fold symmetry , leads can be displaced in a manner which still preserves left - right , inversion , or up - down symmetry . to preserve up - down symmetry alone , one can imagine splitting one lead in two and moving the two parts in opposite directions ( both parts would remain contacted to the same source or drain electrode ) . the remaining symmetry of the lead arrangement can then be broken by further displacement of the leads . in the figures , the subscript @xmath247 is used to distinguish these situations ( in which the underlying internal symmetry is four - fold ) from the symmetry breaking in systems with only a single internal symmetry . e.g. , the label ` left / right@xmath248 asymmetric@xmath247 ' refers to the displacement of leads out of a left - right symmetric position where the internal symmetry is four - fold , while the label ` left / right @xmath249 asymmetric ' refers to the displacement of leads out of a left - right symmetric position where the internal symmetry is itself only left - right symmetric . according to our theory , the weak localization correction should behave identically in both situations ; this also applies to the ucfs . this statement is validated by the numerical data . indeed , excellent agreement of the numerical data with the semiclassical predictions is observed in all cases . as discussed earlier in this paper , in the up - down symmetric case it is interesting to displace only one lead while the other lead remains on the symmetry line ( the symmetry - preserving positions in the up - down symmetric case are absolute , in contrast to the left - right symmetric case where these positions are relative to each other ) . the effect on the transport is shown in fig.[fig : numerics2 ] , along with the effect of the consecutive displacement of the second lead , and the simultaneous displacement of both leads . according to our theory , the effects of consecutive displacement of the leads are cumulative : the displacement of the first lead is described by eqs . ( [ eq : dglambda ] ) , ( [ eq : varglambda ] ) with @xmath250 ( covering the range [ 0,1/2 ] ) , while the displacement of the second lead completes the transition according to the substitution @xmath251 ( covering the range [ 1/2,1 ] ) . the numerical results are in perfect agreement with this prediction . we conclude with some additional remarks on the rmt model . for leads which respect the symmetries , the construction presented here is equivalent to the model presented in part i ( which then is more efficient ) ; this equivalence also extends to the symmetry breaking in the internal dynamics , which then requires to interpolate between ensembles of table [ tab : f ] . following earlier works , the rmt model can be further utilized to include the effects of dephasing and a finite ehrenfest time . for dephasing , this is achieved by opening additional ports which couple to a voltage probe @xcite or a dephasing stub @xcite . a finite ehrenfest time is obtained when @xmath210 represents a dynamical system , such as the kicked rotator @xcite ( which also possesses discrete symmetries ) . this strategy can also be used to probe the case of dynamics which are not fully chaotic ( which in the kicked rotator is achieved for moderate values of the kicking strength ) . the transport calculations performed here assume that the classical dynamics is uniformly chaotic , and in particular do not apply to system with islands of stability in phase space ( such as the annular billiard studied in refs.@xcite ) , or networks of chaotic dots inter - connected by narrow leads ( such as the double dot in ref . @xcite ) . it would be intriguing to study the shape of the back- and forward - scattering peaks for such systems . rw thanks p. brouwer , p. marconcini and m. macucci for interesting and useful discussions . rw and hs are grateful for the hospitality of the banff international research station , where this work was initiated . in section [ sect : ucfs ] we pointed out that unitarity implies @xmath203= { \rm var}[r_{\rm l}]={\rm var}[r_{\rm r}]$ ] . here we outline a semiclassical calculation of @xmath252 $ ] , which acts as a check of the semiclassical calculation of @xmath203 $ ] in section [ sect : ucfs ] . the rules to calculate each contribution remain the same as for @xmath203 $ ] . however , the contributions that we consider differ by the requirement that all paths start and end on the same lead l. we know that the result must be invariant under the interchange of labels `` l '' and `` r '' , and this invariance is manifestly obvious in the contributions to @xmath203 $ ] . in contrast , this invariance is hidden in the contributions to @xmath252 $ ] that we discuss here . thus the simplest check that one has not missed any contributions is that this invariance is present when one sums the contributions . in the case of an up - down symmetric dot , all contributions in both fig . [ fig : ucf - asymleads1 ] and fig . [ fig : ucf - asymleads2 ] contribute to @xmath252 $ ] once we change all lead labels so that `` r '' @xmath249 `` l '' and `` @xmath253r '' @xmath249 `` @xmath253l '' ( but not vice versa ) . writing contributions to @xmath254 $ ] with a prime(to distinguish them from contributions to @xmath203 $ ] ) we find as in section [ sect : ucfs ] , we find that this sum is most easily evaluated by re - writing the contributions in terms @xmath256 and @xmath257 for @xmath258 . performing a little algebra using @xmath197 , we then recover eq . ( [ eq : ucf - updown1 ] ) , and therefore @xmath254={\rm covar}[r_{\rm l},r_{\rm r}]$ ] . furthermore , expression eq . ( [ eq : ucf - updown1 ] ) is invariant under the interchange of labels `` l '' and `` r '' , which entails @xmath259={\rm var}[r_{\rm l}]$ ] . thus the semiclassical method obeys the relations @xmath252={\rm var}[r_{\rm r } ] = { \rm covar}[r_{\rm l},r_{\rm r}]$ ] , as required by the unitarity of the scattering matrix . this strongly suggests that we have not missed any contributions and gives us confidence in the result ; particularly , it is noteworthy that the individual contributions in @xmath252 $ ] and @xmath260 $ ] combine in very different ways to give the invariance under the interchange of `` l '' and `` r '' . the evaluation of @xmath252 $ ] for a left - right or inversion - symmetric dot is very similar to that for an up - down symmetric dot . however , here , when a path hits the l lead then its image hits the r lead . this means that there are no contributions to @xmath252 $ ] of the form shown in fig . [ fig : ucf - asymleads2 ] , since all paths must go from the l lead to the l lead . thus to get @xmath252 $ ] for a left - right or inversion - symmetric dot , we need to subtract those contributions from the result for @xmath252 $ ] in an up - down symmetric dot . the sum of these contributions to @xmath261 $ ] , written with the same denominator as in eq . ( [ eq : ucf - updown2 ] ) , is @xmath262 this only differs by an overall sign from the sum of contributions in eq . ( [ eq : extra - contributions - for - lr ] ) . subtracting this from the result eq . ( [ eq : ucf - updown2 ] ) , we get @xmath261 $ ] for a left - right or inversion - symmetric dot . the result equals @xmath203 $ ] given by eq . ( [ eq : ucf - leftright ] ) , thus we have @xmath260={\rm var}[r_{\rm l}]={\rm var}[r_{\rm r}]$ ] , as required by the unitarity of the scattering matrix . the fact that the paths in contributions 2i - iv reflect back into the dot slightly changes the action difference between them . it was shown in the context of failed coherent backscattering that this change of the action difference is irrelevant @xcite . this is rather different from the weak localization correction with tunnel barriers @xcite . there the survival probability is only different near the encounter , which leads to an extra prefactor of the ratio of dwell times ( of the form @xmath263 ) . there is no such extra prefactor here .

high resolution x - ray spectroscopy of the warm absorber in the nearby x - ray bright seyfert 1 galaxy , mrk 1040 is presented . the observations were carried out in the 20132014 timeframe using the chandra high energy transmission grating with a total exposure of 200ks . a multitude of absorption lines from ne , mg and si are detected from a wide variety of ionization states . in particular , the detection of inner k - shell absorption lines from ne , mg and si , from charge states ranging from f - like to li - like ions , suggests the presence of a substantial amount of low ionization absorbing gas , illuminated by a steep soft x - ray continuum . the observations reveal at least 3 warm absorbing components ranging in ionization parameter from and with column densities of . the velocity profiles imply that the outflow velocities of the absorbing gas are low and withinkms of the systemic velocity of mrk1040 , which suggests any outflowing gas may have stalled in this agn on large enough scales . the warm absorber is likely located far from the black hole , within 300pc of the nucleus and is spatially coincident with emission from an extended narrow line region as seen in the _ hst _ images . the iron k band spectrum reveals only narrow emission lines , with fe k at 6.4 kev consistent with originating from reflection off compton thick pc - scale reprocessing gas .

1702.00461

material existing within the vicinity of the super - massive black hole in an active galactic nucleus ( agn ) can significantly modify the resultant x - ray spectrum . consequently , detailed studies of the reprocessed primary x - ray emission can reveal valuable information about the kinematics and geometry of the central nucleus . in the soft x - ray band ( below 2kev ) , the dominant reprocessing mechanism is absorption by photoionized ( or `` warm '' ) material along the line of sight . in fact systematic x - ray studies of agn with _ asca _ , _ chandra _ , _ xmm - newton _ and _ suzaku _ have established that at least half of all type 1 seyferts host warm absorbers @xcite . when observed at high spectral resolution , such as with the x - ray grating spectrometers onboard either _ chandra _ or _ xmm - newton _ , the warm absorber is seen to give rise to numerous narrow absorption lines from astrophysically abundant elements such as c , n , o , ne , mg , si , s and fe @xcite . the warm absorption is often seen to be outflowing and can arise from gas covering a wide ( several orders of magnitude ) range of column densities and ionization parameters . indeed the x - ray spectral signatures of the warm absorber range can arise from very low ionization gas from ions with low charge states ; such as from the unresolved transition array ( uta ) of m - shell iron ( @xmath6 fexvii ) at @xmath7 @xcite or from inner ( k - shell ) transitions of l - shell ions from lighter elements , such as o through to si @xcite . on the other hand , the absorbing gas can also be very highly ionized , in some cases arising from h - like and he - like iron , where such highly ionized gas may originate from an accretion disk wind ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? the associated absorption lines are often blueshifted , thus implying outflowing winds , with velocities ranging from a few hundred to a few thousand kms@xmath4 , or even exceeding 10000kms@xmath4 in the case of the so - called ultra fast outflows @xcite . with the exception of the ultra fast outflows , which are often highly ionized and may originate from an accretion disk wind on sub - pc scales , most soft x - ray warm absorbers are thought to be typically located at fairly large distances from the central black hole . this is as a result of their low ionization parameters and outflow velocities ( @xmath8kms@xmath4 ) , their ( relative ) lack of variability , plus in some cases from their inferred low densities ( e.g. , ngc3783 : @xcite ; mrk279 : @xcite ; ngc4051 : @xcite ; mrk290 : @xcite ; and mkn509 : @xcite ) . these soft x - ray warm absorbers can be associated with , for example , a wind originating from the putative parsec scale torus ( @xcite ) or the latter stages of an accretion disc wind which has propagated out to larger radii ( ) , or from gas associated with the agn narrow line regions . although the exact kinematic contribution of the warm absorbers to the total agn luminosity can be difficult to assess , they may still play a key role in shaping the wide scale properties of agn host galaxies . mrk 1040 ( also known as ngc931 ) is a bright , nearby ( @xmath9 , @xcite ) type-1 seyfert galaxy . the likely black hole mass is @xmath10 @xcite , as estimated from its stellar velocity dispersion . in x - rays , mrk1040 was observed by _ asca _ in august 1994 @xcite and the spectral features were interpreted as arising from a warm absorber plus a strong broad , fluorescent fek@xmath5 emission line ( fwhm @xmath11@xmath12kms@xmath4 ; @xmath13ev ) . it is also bright in the hard x - ray bandpass , having been detected in the 70 month swift bat catalogue , with a 14 - 195kev band flux of @xmath14erg@xmath2s@xmath4 @xcite . mrk 1040 was subsequently observed by _ xmm - newton _ in february 2009 when the source flux was high , with a total exposure of @xmath1584ks . the spectrum revealed all the classic seyfert 1 characteristics ; a soft x - ray excess , a deep warm absorber below 2kev and a strong fe k@xmath5 line profile @xcite . the agn is also strongly variable on timescales of @xmath16s and an energy dependent lag was claimed from the _ xmm - newton _ observation @xcite , similar to the soft x - ray ( or negative ) lags that have been claimed in other agn @xcite . yet despite its x - ray brightness , similar to other bright and nearby seyfert 1 galaxies , the x - ray spectrum of mrk1040 is comparatively understudied and no high resolution x - ray spectrum has yet been published on this agn . the host galaxy of mrk1040 is a sbc spiral galaxy and is viewed at high inclination of @xmath17 @xcite . the uv continuum is heavily reddened , which suggested was due to dust in the plane of the host galaxy . based on the width of h@xmath18 , full width at half maximum ( fwhm ) of @xmath19kms@xmath4 , it is classified as a narrow - line seyfert 1 @xcite . if the narrowness of the balmer lines is due to a line - of - sight close to the rotation axis of the black hole / accretion disk @xcite , then our view of the agn in mrk 1040 may be roughly pole - on , which suggests that the agn is highly inclined with respect to the disk of the host galaxy . this paper presents the analysis of a series of _ chandra _ high energy transmission grating ( hetg ) @xcite observations of mrk1040 from october 2013 to march 2014 , with a total exposure of @xmath20ks . indeed the observations provide the first reported high spectral resolution observations of this seyfert 1 galaxy over the x - ray band . subsequent observations with _ xmm - newton _ were also obtained during 2015 , with a total exposure of 175ks in the rgs ; these together with the archival observation from 2009 will be presented in a subsequent paper ( hereafter paper ii ) . one of the primary goals was to study the warm absorber in this agn in unprecedented detail and resolution . here the hetg provides the higher energy coverage of the warm absorber , especially in the 1 kev band and above and covering the ne through fe k - shell lines , with the rgs providing high sensitivity coverage at lower energies down to 0.3kev . the paper is organized as follows . in section 2 , we describe the observations and reduction of the hetg spectra . section 3 is devoted to the spectral fitting of the hetg data , including atomic line detections and identifications as well as the measurements of the kinematics and profiles of the absorption lines . the properties of the emission lines , such as in the iron k - shell band and the he - like triplets , are also presented . section 4 presents photoionization modeling of the x - ray absorption in the hetg , which as will be shown requires multiple absorption components to cover the wide range of ionization states seen in the absorption spectrum . in section 5 , we discuss about the origins and infer some physical properties of the x - ray absorbing media observed in mrk1040 . the properties of the x - ray gas are then compared to the images and kinematics of the nuclear [ oiii ] emitting gas seen from archival _ hubble space telescope _ ( _ hst _ ) observations . as we will show , the x - ray absorber appears to show no evidence of outflow to high precision and the ionized gas may be in part associated with a failed wind located within @xmath21pc of the nucleus . _ chandra _ observed mrk1040 four times between 24 september 2013 to 3 march 2014 ( see table 1 for a summary ) , with three of the four observations occurring within a one week period from 25 february 2014 and 3 march 2014 . spectra were extracted with the ciao package v4.7 @xcite . only the first order dispersed spectra were considered for both the meg ( medium energy grating ) and heg ( high energy grating ) and the @xmath22 orders for each grating were subsequently combined for each sequence . no significant spectral variability was observed between the four sequences , with only modest @xmath23% variations in count rate ( see table 1 ) . therefore the spectra were combined from all four sequences to yield a single 1st order spectrum for each of the meg and heg , yielding respective net source count rates of @xmath24s@xmath4 ( from 0.85.0kev ) and @xmath25s@xmath4 ( from 1.2 - 9.0kev ) respectively for a total exposure time of 197.8ks . the total counts obtained exceeded 90000 and 47000 counts for meg and heg respectively , while the background count rate was negligible . the undispersed zeroth order image shows no evidence for extended x - ray emission in excess of the telescope point spread function . the resultant time - averaged source spectra were subsequently binned at 3 different levels . these correspond to either at half width at half maximum ( hwhm ) of the spectral resolution ( i.e. @xmath26m and @xmath27m bins for meg and heg respectively ) , at the fwhm of the resolution , or at twice the fwhm resolution . note that at 1kev ( or 12.3984 ) , the hwhm binning corresponds to velocity intervals of @xmath28kms@xmath4 and @xmath29kms@xmath4 , per spectral bin , for the meg and heg respectively . the former finer hwhm binning was adopted for the subsequent spectral fitting as well as for obtaining velocity profiles of the lines , while the latter more coarsely binned spectra ( fwhm or twice fwhm ) were used for an initial inspection of the data and for some of the spectral plots . the meg and heg spectra were analyzed over the energy ranges of 0.85.0kev and 1.29.0kev respectively . below 0.8kev , the signal to noise of the meg data was low due to the build up of contamination over time on the acis - s ccds and was excluded from the subsequent fitting . a cross normalization constant between the meg and heg spectra has been included in all the subsequent fits , however in each case it is consistent with 1.0 . note the overall fluxes , obtained against an absorbed power - law , correspond to @xmath30erg@xmath2s@xmath4 and @xmath31erg@xmath2s@xmath4 respectively . the c - statistic @xcite was employed in the subsequent spectral fits to the hetg , as although an overall large number of counts were obtained in the observations , towards the lower energy ( longer wavelength ) end of each grating spectrum the total source counts per bin drops below @xmath32 in some of the spectral bins , which is also true at the centroids of the deepest absorption lines in the spectrum . in the case of @xmath33 minimization , this would lead to the continuum level being somewhat underestimated at soft x - ray energies . all fit parameters are given in the rest frame of the quasar at @xmath34 and are stated in energy units , adopting the conversion factor of 1kev being equivalent to 12.3984 in wavelength . in all of the fits , a galactic absorption of hydrogen column density of @xmath35@xmath2 @xcite was adopted , modeled with the `` tuebingen boulder '' absorption model ( tbabs in xspec ) , see @xcite and references therein for the cross sections and abundances used in this model . note that upon including molecular hydrogen , the galactic column rises to @xmath36@xmath2 , e.g. see @xcite , although the spectra above 0.8kev are not sensitive to the exact value adopted here . values of @xmath37kms@xmath4mpc@xmath4 , and @xmath38 are assumed throughout and errors are quoted at 90% confidence ( @xmath39 ) , for 1 parameter of interest . figure[fluxed ] shows the observed hetg spectrum of mrk1040 , where the spectrum has been unfolded through the instrumental response against a powerlaw model of photon index @xmath40 in order to produce a fluxed spectrum in @xmath41 flux units . the spectrum has been plotted at the fwhm spectral resolution and over the wavelength ( energy ) range from 1.715.2 ( or 0.827.3kev ) and is plotted in wavelength ( in the observed frame ) to show the overall spectral form . the spectrum shows many clear signatures of a warm absorber , with several k - shell absorption lines present from the most abundant elements , where the strongest lines have been identified in the figure panels . aside from the highly ionized absorption lines from he and h - like ions ( e.g. from ne , mg and si , as well as the h - like lines from s and ar ) , evidence for lower ionization absorption is also present . this is in the form of inner k - shell ( @xmath42 ) absorption lines from ions corresponding to charge states from li to f - like ( i.e. where the l - shell is partially occupied ) , see @xcite . these can be seen in the spectrum via absorption lines from charge states varying from:- neiii - viii ( 14.014.8 observed frame ) , mgv - ix ( 9.5 - 10.0 ) and from sivi - xii ( 6.87.2 ) . the most prominent of these lines appear from nev ( c - like ne ) and mgviii ( b - like mg ) . this suggests that the warm absorbing gas covers a wide range of ionization , as has been observed in other seyfert 1 galaxies , e.g. ngc3783 : @xcite , ngc5548 : @xcite , mrk509 : @xcite , ngc7469 : @xcite , ic4329a : @xcite , ngc3516 : , mcg-6 - 30 - 15 : @xcite , mrk279 : @xcite . note that we do not detect absorption due to either the low ionization ions of iron , i.e. in the form of the iron m - shell unresolved transition array ( fei - xvii uta , @xcite ) , or due to oxygen @xcite , as these fall below 0.8kev where the effective area of the hetg drops rapidly . however the study of the absorption from these lower energies features will be presented in paper ii covering the soft x - ray rgs spectrum . in order to measure the absorption line properties , an initial parameterization of the continuum was adopted . the continuum emission was modified by a powerlaw absorbed by the galactic column and a neutral partial covering absorber was also included ( using the xspec model zpcfabs ) , whereby a fraction @xmath43 of the primary x - ray continuum passes through the absorber . this simple model is adopted purely to account for the spectral curvature ( decrease in flux ) towards lower energies due to bound - free absorption , but without accounting for any discrete absorption lines . a full description of the photoionization modeling of the spectrum which accounts for both the absorption lines present and the spectral curvature is given in section 4 and its effect on the overall spectrum is shown later in figure 8 . the photon index of the powerlaw was found to be @xmath44 , while the column of the partial covering absorber was found to be @xmath45@xmath2 with a line of sight covering fraction of @xmath46 , i.e. consistent with a fully covering absorber . while this simple partial covering model is successful in reproducing the spectral curvature , the overall fit is poor , with a fit statistic of @xmath47 ( where @xmath48 is the number of degrees of freedom ) which is rejected at @xmath49% confidence . the residuals to this continuum model in the ne , mg and si bands to the meg spectrum are shown in figure[meg - panels ] , while the equivalent residuals in the mg and si bands to the heg spectrum are shown in figure[heg - panels ] . indeed a wealth of absorption line structure is clearly present in the hetg spectrum ( independently in both the meg and heg gratings ) against the continuum model , especially at energies below 2kev in the ne , mg and si bands . as expected from the initial broad - band spectrum , low ionization gas appears to be present in the form of a multitude of inner k - shell lines of ne , mg and si . we refer to @xcite for a compilation of these inner shell lines and we adopt the known energies ( wavelengths ) of these lines from this paper later in table2 . indeed such lines have been detected in other high signal to noise grating spectra of seyfert 1 agn , such as in ngc3783 @xcite , ngc 4151 @xcite , mrk 509 @xcite , ngc 3516 , ngc4051 @xcite , ngc 5548 @xcite , mr2251 - 178 @xcite and mcg-6 - 30 - 15 @xcite among others . in the ne band of the mrk1040 meg spectrum , residuals due to inner - shell absorption lines due to neiii - vii ( o - like ne through to be - like ne ) are present from 0.85 - 0.90kev ( 13.8-14.6 ) . in addition at higher energies absorption from he and h - like ne are also present , in particular the he - like line series shows absorption from @xmath42 up to @xmath50 ( see figure 2 , panels a and b ) . this indicates the absorber exhibits a wide range of ionization contributing many different charge states to the absorption spectrum . likewise , the spectrum in the mg and si k - shell bands shows similar characteristics . inner - shell lines from mgv - ix ( o - like through to be - like ions ) are apparent from @xmath51kev ( 9.49.8 ) . similarly , inner shell absorption is also detected from sivi - xii ( f - like to li - like ) around 1.741.85kev ( 6.77.1 ) . most of the lines are also independently detected in the heg ( figure[heg - panels ] ) as well as the meg ( figure[meg - panels ] ) spectra . absorption from the resonance ( @xmath42 ) he - like lines of mg and si are also detected ( as well as the h - like absorption for mgxii ) , which again confirms that ions from a wide range of charge states are present in the mrk1040 absorber . we then constructed velocity profiles of the principle absorption lines listed in table2 . we restricted the analysis to those lines which are detected at at least 95% confidence ( corresponding to @xmath52 for 2 interesting parameters ) , noting that most of the lines are detected at least at this level in both gratings ( with the exception of the ne band as noted below ) . a baseline continuum consisting of an absorbed power - law was used , utilizing the neutral partial covering model described in the previous section , as this accounts for the broad - band spectral curvature without imparting any discrete absorption lines upon the spectrum . the lines were then fitted with gaussian profiles in the rest energy frame , allowing their centroid energies , widths and normalisations to vary and accounting for the spectral resolution via the standard instrument responses . to maximize the resolution , hwhm binning was used and we utilized the highest resolution heg data for the mg and si bands ( and above ) , except for the ne band where the meg is adopted due to the low effective area of the heg at 1kev and below . the line profiles were then transposed into velocity space around the known lab frame energy ( or wavelength ) of each line . in the case where a line profile originates from an unresolved doublet ( such as for the h - like lines ) , we took the weighted mean of the respective lab frame energies to calculate the expected line energy . the doublet corresponding to the mgviii @xmath42 lines is resolved by the heg , so in that case we fitted the two lines separately , obtaining the centroid in velocity space separately for each component , i.e. allowing for the different rest - energies of the two lines . the results of the line profile fitting are summarized in table2 , which includes 21 line profiles covering 18 separate ionic species from ne up to ar . the absorption profiles arise mainly from the @xmath42 transitions , except for neix and nex , where the higher order @xmath53 lines are detected ; indeed in these latter cases the he-@xmath18 and ly-@xmath18 profiles provide better constraints on the line widths and velocity centroids , as they are free from contamination from other lines . it is also apparent from inspecting the measured energy centroids in table2 that most of the lines are positioned close to their expected lab frame energies , while generally the absorption lines are unresolved ( or only marginally resolved ) , with typical velocity widths of @xmath54kms@xmath4 ( or even @xmath55kms@xmath4 in the more tightly constrained cases ) . examples of some of the velocity profiles are plotted in figure4 ( for the inner - shell and he - like resonance lines ) and in figure5 ( for the h - like lines ) . in all of the examples , the line centroids are consistent with zero velocity shift within the measurement errors . while it may appear that some of the profiles contain an additional higher velocity component ( either red or blue - shifted ) with respect to the main line centroids , these are due to the presence of nearby lines from adjacent ions as labelled in figures 4 and 5 . for example the strong nev absorption profile shows weak absorption components at @xmath56kms@xmath4 ; these are simply due to neighbouring absorption lines from neiv and nevi either side of nev with no velocity shift . this is also the case for the siviii and sixiii profiles , e.g. the latter he - like profile contains a weak contribution from sixii ( li - like si ) on the red - side of the profile . the velocity widths of these profiles are generally low or unresolved , only nev is significantly broadened ( with @xmath57kms@xmath4 ) . however this may be a result of the line being an unresolved doublet @xcite , or from this deep line becoming saturated in its core ( as the flux reaches close to zero at the center ) . in contrast the he - like profiles of neix ( he-@xmath18 ) and mgxi ( neither of which are doublets , saturated , or contaminated by other lines ) are only marginally resolved , with widths of @xmath58kms@xmath4 and @xmath59kms@xmath4 , respectively . the mgviii profile is clearly resolved in the heg data into a doublet , although the velocity shifts and widths of both components are consistent with zero ( e.g. @xmath55kms@xmath4 ) . a similar story emerges for the h - like ( lyman-@xmath5 ) profiles , as shown in figure 5 for nex , mgxi , sxvi and arxviii ( note sixiv falls just below the detection threshold ) . the nex ly-@xmath5 profile is heavily contaminated by @xmath60 absorption by fexxi and thus the velocity constraints listed in table2 are instead obtained from nex ly-@xmath18 . the highly ionized sxvi and arxviii profiles are both unresolved and their velocity centroids are consistent with zero . only the mgxii profile contains some indication of blueshift . although the main centroid is consistent with zero velocity , a second velocity component emerges at @xmath61kms@xmath4 which appears uncontaminated by other lines ( note that the separation of the mgxii doublet can not account for this as it is too small to be resolved by the heg ) . however upon inspecting the meg data for consistency , no such blueshifted component was confirmed , nor is such a component detected in any of the other high ionization he or h - like lines . even though this component appears formally significant at the @xmath62% level ( @xmath63 ) , as no high velocity component was measured in any of the other profiles we do not consider this detection to be plausible . overall none of the individual profiles , low ionization as well as high ionization ( he and h - like ) , appear to show a significant outflowing component and most of the lines are unresolved within the instrumental resolution . to determine if the absorption profiles as a whole are characterized by a small velocity shift , which is within the statistical error of any individual profile , the mean and dispersion of the velocity shifts of the sample of lines was calculated based on the values reported in table2 . only one value was considered for each ionic species , taking the best determined values in the three cases where multiple measurements exist for a given ion ( e.g. the nex @xmath53 line was favored over the @xmath42 line ) . over 18 ions , the mean velocity shift was found to be @xmath64kms@xmath4 , with a dispersion of @xmath65kms@xmath4 . neither was any trend or deviation found between outflow velocity and rest energy . thus no outflow ( or inflow ) is required to within @xmath66kms@xmath4 of the systemic velocity of mrk1040 , within the absolute wavelength calibration of the hetg gratings . thus unlike for many other agn , the mrk1040 warm absorber does not appear to originate from an outflow , at least to within very tight limits . in the next section we return to model the absorption with photoionization models generated by xstar . in contrast to the absorption lines , there are a paucity of soft x - ray emission lines in the hetg spectrum against the bright x - ray continuum . the emission lines which are detected in the spectrum are listed in table3 . in the soft x - ray band the most significant emission residuals occur around the mgxi triplet , at the expected positions of the forbidden and intercombination transitions , while an absorption line is present at the resonance line energy as is shown in figure6 for the heg data . both emission lines are relatively weak ( equivalent width of @xmath67ev ) and appear unresolved , with an upper limit to their velocity widths of @xmath68kms@xmath4 . the relatively equal contribution of the forbidden and intercombination components appears surprising at first , as this would imply a relatively low ratio of the @xmath69 value between the two lines ( @xmath70 , see @xcite ) and subsequently a high density of @xmath71@xmath72 . however given the large errors on the line normalizations it is not possible to constrain the line ratio and the derived density is also formally consistent with a lower density photoionized plasma , where @xmath73@xmath72 . furthermore a relatively high uv radiation field may also suppress the forbidden line , via photoexcitation @xcite , making the direct measurement of the density from the triplet lines less certain . furthermore the forbidden line may also be absorbed on its blue - wing due to the adjacent mgx resonance absorption line ( see figure 6 ) . note that no other line emission is detected from the other triplets , e.g. as the s / n is too low around neix , while the only other soft x - ray line which is marginally detected arises from sixiv lyman-@xmath5 at 2.0kev , which again is unresolved . the strongest signal present in the iron k - shell band arises from the almost ubiquitous narrow fe k@xmath5 emission line @xcite ; the iron k band spectrum is plotted in figure7 . the line is detected at @xmath74ev and thus is at the expected energy for near neutral iron . the equivalent width of the narrow fe k@xmath5 line is @xmath75ev . the line is unresolved , with a formal upper - limit to its width of @xmath76ev or @xmath77kms@xmath4 in velocity space . this appears consistent with an origin in distant ( e.g. pc scale ) gas , which is often found to be case from fe k@xmath5 line cores seen in other type i seyferts as measured by chandra hetg @xcite . note there is some indication of an excess red - wards of the line core , this is consistent with the presence of a compton shoulder to the line with an energy of @xmath78kev ; however formally the compton shoulder is not detected ( @xmath79 ) and neither is it well constrained . aside from the possibility of a weak compton shoulder , there is no evidence for any additional excess emission on the red - wing of the line , which may be expected to arise from a broad line from the inner accretion disk . the presence ( or lack thereof ) of any relativistic line component in mrk1040 will be discussed in a forthcoming paper . aside from the neutral k@xmath5 emission , there is some evidence ( at @xmath80 ) for the presence of ionized emission exactly coincident with the expected position of the forbidden line from he - like fe ( fexxv ) . although weak ( @xmath81ev ) , the line centroid is accurately measured to @xmath82ev , consistent with the known energy of the forbidden component and the line width appears narrow with @xmath83ev ( or equivalently @xmath84kms@xmath4 ) . thus its origin likely arises from distant photoionized material , as per the soft x - ray line emitting and absorbing gas . note that the residuals at the expected position of the resonance absorption lines from fexxv and fexxvi ( 6.70kev and 6.97kev ) are not formally significant in their own right ( @xmath85 and @xmath86 respectively ) , however in the next section we will place a formal limit on any high ionization absorption component arising from mrk1040 . finally in order to determine whether the observed narrow iron k@xmath5 emission is consistent with an origin in scattering off distant compton thick matter , we instead modeled with spectrum with either a neutral or an ionized compton reflection component in addition to the underlying power - law continuum . an excellent fit ( @xmath87 over the 28kev band ) was found from modeling the line with a neutral reflection component , adopting the pexmon model @xcite within xspec , which returned a reflection fraction of @xmath88 for an assumed inclination angle of @xmath89degrees and solar abundances . note that the photon index of the underlying continuum was @xmath90 . a statistically equivalent fit was also obtained with an ionized reflection model , using the ( angle averaged ) tabulated xillver model spectra of @xcite and @xcite , where the upper - limit on the ionization parameter was found to be @xmath91 , with a consistent photon index of @xmath92 . indeed the spectrum fitted with this model is shown in figure7 . thus the fe k@xmath5 emission appears consistent with an origin from distant , compton thick matter and no additional velocity broadening is required to account for this band . given the substantial presence of partially ionized gas in the x - ray spectrum of mrk1040 , we attempted to model the absorption spectrum with photoionized grids of models using the xstar code v2.2 @xcite . absorption grids were generated in the form of xspec multiplicative tables . the absorption spectra within each grid were computed between 0.120kev with @xmath93 spectral bins . the photoionizing x - ray continuum between 11000rydberg was assumed to be a broken power - law , with a photon index of @xmath94 below a break energy of 0.8kev ( at the low energy end of the hetg bandpass ) and with a photon index of @xmath40 assumed above this break energy . the 11000rydberg photoionizing luminosity is predicted to be @xmath95ergs@xmath4 for mrk1040 using this continuum model . although the power - law break is not required to model the continuum in the hetg data , which is not sensitive to the lowest energies , it is required by the subsequent _ xmm - newton _ spectra ( both for the pn and rgs , paper ii ) to account for the soft x - ray excess in this source . furthermore , as will be discussed in section4.3 , a softer photoionizing continuum is required to model the numerous low ionization inner - shell absorption lines in the spectrum , which otherwise are over ionized by a flatter @xmath40 continuum . with this choice of continuum , we then generated an initial broad grid of models that covered a wide range in ionization and column density:- from @xmath96@xmath2 to @xmath97@xmath2 and @xmath98 ( @xcite ) , where @xmath99 is the @xmath100 rydberg ionizing luminosity , @xmath101 is the electron density and @xmath69 is the distance of the ionizing source from the absorbing clouds . the units of @xmath102 are ergcms@xmath4 . ] in logarithmic steps of @xmath103 and @xmath104 respectively . this broad grid was used to provide an initial characterization of the warm absorber zones ( i.e. to provide a first order estimate of the range of ionization and column needed to account for the absorption ) and to account for any high ionization absorption components . a more finely tuned grid ( covering a narrower range of parameters ) was subsequently generated with the specific purpose of modeling the low ionization absorption in the mrk1040 spectrum , especially the inner - shell lines . the column density of this narrow low ionization grid covered the range from @xmath105@xmath2 in steps of @xmath106@xmath2 , with the ionization range extending from @xmath107 in 20 steps of @xmath108 . a fine spectral resolution of @xmath109 points over an energy range of @xmath110kev was also employed . given the narrow ( or unresolved ) widths of the absorption lines detected in the _ chandra _ hetg spectrum , a turbulence velocity . ] of @xmath111kms@xmath4 was assumed for the absorption models ; grids with much higher turbulences all gave substantially worse fits in the models considered below . solar abundances were adopted for all the abundant elements , using the values of @xcite , except for ni which is set to zero ( the default option within xstar ) . in order to model the absorption spectrum we successively added individual components of absorbing gas , fully covering the line of sight to the source , until the fit statistic was no longer improved at the 99% confidence level ( @xmath112 for 2 parameters ) and no obvious residuals remained after fitting the spectrum . to fit the continuum itself we adopted the broken powerlaw form as described above , with the break energy fixed at 0.8kev and the photon index below the break also fixed at @xmath94 ( as the chandra data are not sensitive to the slope of the soft excess ) . note that the above choice of the continuum is consistent with the later _ xmm - newton _ observation , where the low energy soft excess is required . above the break energy the photon index was allowed to vary , along with the overall normalization of the continuum . after adding the required absorption zones as described below , the photon index was found to be @xmath113 , while after correcting for the absorption , the intrinsic luminosity of mrk1040 was @xmath114ergs@xmath4 and @xmath115ergs@xmath4 over the 210kev and 11000rydberg bands respectively . the emission lines were also added to the model as gaussians , as described in section3.4 ; these arise from the mgxi triplet , from the fe k@xmath5 and fexxv components of the iron k band emission and a weak emission line from sixiv lyman-@xmath5 ( see table 3 ) . three components of absorbing gas are formally required in the _ chandra _ model to account for the soft x - ray absorption , which are listed as zones13 in table4 . the finely tuned grid of absorption models was used for all three soft x - ray zones . a fourth highly ionized absorption zone was also added to the model ( zone 4 , table4 ) , using the coarse grid covering a broader range of column and ionization . however compared to the soft x - ray zones 13 , this high ionization zone is only marginally significant ( @xmath116 ) and mainly contributes by adding the weak absorption lines due to fexxv and fexxvi ( see section 3.4.1 and figure 7 ) . overall with all the zones of absorption applied to the continuum , the fit to the _ chandra _ spectrum is very good , with a final fit statistic of @xmath117 and a corresponding null hypothesis probability of @xmath118 . in comparison the fit with the simple neutral partial covering model yielded a substantially worse fit with @xmath119 ( as this model only accounts for the continuum curvature but not the line absorption ) , while a fit without any absorption zones applied to the continuum ( aside from the galactic column ) is extremely poor with @xmath120 . indeed the effect of the warm absorber upon the overall spectrum is illustrated in figure 8 . here the top panel shows the spectrum fitted with just a power - law continuum ( with @xmath121 ) above 3 kev , where the pronounced spectral curvature due to the warm absorber is apparent towards lower energies . once the warm absorber is applied to the spectral model , then the continuum curvature is clearly accounted for by the absorbing gas ( see figure 8b ) . overall the three soft x - ray warm absorber zones13 that are required to model the _ chandra _ spectrum cover the range in column from @xmath122@xmath2 and ionization from @xmath123 ( see table4 for detailed parameters ) , where the lowest ionization zone 1 ( @xmath124 ) has the highest column . as is expected from the velocity profile analysis ( section 3.3 ) , the outflow velocities of the absorption zones are very low and are typically consistent within @xmath3kms@xmath4 of the systemic velocity of mrk1040 . while the lowest ionization zone1 hints at some mild outflow ( with @xmath125kms@xmath4 ) , this is not supported by either zones 2 or 3 ; e.g. zone 2 is consistent with zero ( @xmath126kms@xmath4 ) and zone 3 suggests very mild inflow ( with @xmath127kms@xmath4 ) . however any slight difference in velocities between the zones are likely to be within the scatter of the measurements . indeed even the highly ionized zone4 ( where @xmath128 ) shows no evidence for outflow , with a formal limit on the outflow velocity of @xmath129kms@xmath4 . figure9 shows the relative contributions of each of the warm absorber components against a power - law continuum . the two lower ionization zones 1 and 2 , with @xmath124 and @xmath130 respectively ( top and middle panels ) , contribute to the lower ionization ions , reproducing the inner - shell lines of ne , mg and si as seen in the spectrum below 2kev . most of the bound - free spectral curvature is also imparted upon the spectrum by these two zones , with the lowest ionization zone1 having the greater opacity towards lower energies . the higher ionization zone3 ( with @xmath131 ) mainly produces the he and h - like absorption lines from ne , mg and si , while the very highly ionized zone4 is essentially transparent at soft x - rays and only makes a weak contribution to some of the h - like lines . the superposition of these zones is then able to model the absorption line structure seen in the _ chandra _ spectrum , where figure10 shows the best - fit model over the 0.8 - 2kev band covering the major absorption lines from ne , mg and si . indeed the xstar model is able to account for the wide range of charge states present , e.g. reproducing the series of absorption lines observed from nev - x , mgvi - xii and sivii - xiii , while simultaneously being able to model for the convex shape of the overall x - ray spectrum . the broad , and relatively flat distribution of @xmath132 with @xmath102 seen in table 4 is commensurate with other slow seyfert outflows ; see @xcite where this was parametrized as @xmath133 . here the value of @xmath134 corresponds to a density profile of @xmath135 , where @xmath136 . formally , the values in table 4 ( for zones 13 ) yield @xmath137 and hence @xmath138 . such a flat distribution can be explained by an mhd outflow with @xmath139 @xcite , but also with a radiation - pressure compressed ( rpc ) cloud @xcite . the mhd wind models predict a well - defined line - of - sight outflow velocity structure of @xmath140 , that for a flat ionization distribution implies @xmath141 that is not quite observed here . the rpc models do not solve for the kinematic structure of the outflow and assume the entire absorber is cruising uniformly at the same velocity ( that does not need to be specified for the ionization distribution to be flat ) . the latter appears to be the case for the absorber of mrk 1040 . clearly , the broad distribution of @xmath132 with @xmath102 is not consistent with a simple radial outflow of @xmath142 , which would naturally produce a narrow distribution of ionization . as discussed above , the shape of the illuminating sed from 11000ryd may effect the properties of the x - ray absorber via the ionization balance . here we adopted a broken powerlaw for the sed , with a steep spectrum below @xmath143kev , motivated by the need to model the lower ionization lines of ne , mg and si in particular . these inner shell ions have relatively low ionization potentials for their l - shell electrons , e.g. 266ev for mgviii and 126ev for nev ; both of the ions produce notable absorption in the spectra as is shown in figures 2 and 3 . the low ionization potentials of these ions suggest that the form of the euv and soft x - ray continuum may have an effect on the warm absorber modeling . in comparison we also tested the effect of a flat @xmath40 continuum on the absorber properties , by replacing the above grids of models with their equivalent versions with the flat sed shape and subsequently minimizing the new spectral fit . in this case the 11000 rydberg ionizing luminosity is a factor of 3 lower than for the broken power - law case ( @xmath144ergs@xmath4 vs. @xmath145ergs@xmath4 ) . the effect of the flat sed on the subsequent warm absorber modeling is quite apparent , which is illustrated in figure11 in the ne and mg k - shell bands . the flat sed model clearly underpredicts the inner shell absorption from mgv - ix and nev - viii , presumably due to the lower luminosity in the euv to soft x - ray range and as an excess of such photons above the x - ray powerlaw is needed to excite these l - shell electrons . on the other hand , both models are able to account for the higher ionization absorption lines from he and h - like ions , as these are more sensitive to the hard powerlaw part of the continuum above 1kev than the soft excess . overall the fit with the flat sed model is significantly worse than the softer broken powerlaw sed model , with @xmath146 vs. @xmath117 respectively ; here the difference in fit statistic arises from the inability of the former model to account for the lower ionization absorption lines . this demonstrates that the low ionization warm absorber is particularly sensitive to the form of the ionizing continuum in the lowest energy band ; this has been found to be the case for the qsos , mr2251 - 178 @xcite and iras13349 + 2438 @xcite . thus potentially the properties of the warm absorber may make it possible to deduce the nature of the soft x - ray continuum and especially the soft excess in agn . while this is unlikely to be the result of the direct thermal emission from the accretion disk @xcite , some authors have suggested that the origin of the soft excess was atomic and may instead arise from an ionized reflector . an ionized reflection spectrum @xcite , when convolved with the relativistic blurring expected in the innermost disk a few gravitational radii from the black hole , can produce an overall smooth continuum capabale of fitting the featureless soft excess below 1kev @xcite . alternatively , comptonization of uv disk photons in a warm , optically thick , accretion disk atmosphere @xcite has also proven to be a plausible form for the euv to soft x - ray excess in many agn @xcite . in mrk1040 , the warm absorber modeling favors a steep ( @xmath147 ) continuum from the euv to soft x - rays . this might imply that the soft excess is part of the intrinsic continuum in this agn and may be produced through disk comptonization , rather than arising from reprocessed or reflected emission , as is also likely to be the case in mr2251 - 178 @xcite . we have presented the first high resolution x - ray spectrum of the nearby , x - ray bright , seyfert 1 galaxy , markarian1040 . the _ chandra _ hetg observations reveal a spectrum showing a wealth of spectral features from the warm absorber , which covers a wide range in ionization ; e.g. from lowly ionized o - like ions up to highly ionized he / h - like like species . the warm absorber parameters for this agn , with column density in the range from @xmath148@xmath2 and @xmath149 , are fairly typical for most nearby seyfert galaxies @xcite , although at the lower range in ionization . perhaps the most surprising discovery is the low ( or zero ) velocity of the gas , consistent within @xmath3kms@xmath4 of the systemic velocity of mrk1040 . thus unlike the majority of seyfert 1 galaxies , whereby the warm absorbing x - ray gas is clearly ascribed to an outflow , the warm absorber in mrk1040 either does not appear to be associated with an outflow , or instead any wind is not able to accelerate out to sufficient distances from the nucleus . indeed the velocities of most x - ray warm absorber components typically fall in the range from 0 up to @xmath150kms@xmath4 @xcite , while the uv velocity components can also cover a similar range of velocity as seen in x - rays . typical outflow velocities of a few hundred kms@xmath4 are also seen for the soft x - ray absorber in many well studied cases ; e.g. ngc 3783 @xcite , mrk 509 @xcite . furthermore , seyfert 1 agn often show multiple velocity components in the x - ray band , as is also commonly observed in the uv . for example in an analysis of a deep observation of mcg6 - 30 - 15 observed with _ chandra _ hetg , @xcite detect a low velocity absorber with @xmath151kms@xmath4 , while a higher velocity absorber emerges with @xmath152kms@xmath4 only in the highest ionization component of the warm absorber . in this case , this was directly seen in some of the h - like lines ( e.g. mgxii and sixiv ) , where the high velocity component was revealed in the blue - wing of the velocity profiles for these lines . while the lower velocity component of the warm absorber in mcg6 - 30 - 15 appears very similar to what is observed here ( with a wealth of inner k - shell and uta absorption present ) , no high velocity component of the high ionization lines is seen in mrk1040 ( e.g. see the velocity profiles in figures 4 and 5 ) . the higher ionization zone3 warm absorber in mrk1040 ( see table 4 ) shows no evidence of outflow ( @xmath153kms@xmath4 ) , while the highest ionization zone 4 ( mainly arising from the weak contribution of he and h - like iron ) has an upper limit of @xmath129kms@xmath4 to the outflow velocity . this is in contrast to the warm absorbers seen in many other seyferts , where higher velocity zones can emerge in the higher ionization gas . for instance in ngc3516 , fast outflowing zones of up to @xmath154kms@xmath4 were present when the agn continuum brightened , which in this case may have coincided with the emergence ( or illumination ) of an inner disk wind component to the absorber . similarly in ngc4051 , both high ( several thousand kms@xmath4 ) and low velocity ( several hundred kms@xmath4 ) warm absorbing zones appear to be present , while in ngc5548 the onset of a high velocity ( bal - like ) component of the uv and x - ray absorber was also recently revealed in during an absorption episode @xcite . thus mrk1040 appears curious in that there is no evidence for the x - ray warm absorber to be outflowing , either through a slow component ( with @xmath155kms@xmath4 ) or a faster higher ionization zone ( with @xmath156kms@xmath4 ) , at least on the basis of the x - ray spectra from the present epoch of observations . both the low velocity and ionization of the absorber in mrk1040 likely places the gas at large distances from the black hole , as we will show below . this is far removed from some of the high velocity and high ionization components of the warm absorber discussed above , which may instead originate from an accretion disk wind on sub - parsec scales ( e.g. * ? ? ? * ) and some of which are powerful enough to drive gas out to kpc scales @xcite . the best physical constraints on the absorbing gas in mrk1040 arise from the zone 3 absorber ( see table 4 ) , as this zone is also associated with the he - like emission from the mg triplet , which makes it possible to calculate the covering fraction of the gas as well as its radial location . photoionized emission spectra , in the form of an additive table of models ( or an atable within xspec ) were also generated with xstar , with the same properties as the zone3 absorber and with the same input continuum and ionization ( @xmath131 ) . from the photoionization modeling , the normalization ( or flux ) , @xmath157 , of an emission component is defined by xstar @xcite in terms of : @xmath158 where @xmath159 is the ionizing luminosity in units of @xmath160ergs@xmath4 over the 11000rydberg band and @xmath161 is the distance to the agn in kpc . here @xmath162 is the covering fraction of the gas with respect to the total solid angle , where @xmath163 and thus for a spherical shell of gas , @xmath164 . for mrk1040 , where @xmath165ergs@xmath4 and @xmath166mpc @xcite , then for a spherical shell the expected xstar normalization is @xmath167 . hence for a given column density of gas , this sets the total luminosity of the soft x - ray photoionized emission ; also see @xcite for a similar calculation . we then applied this xstar emission component to the spectrum of mrk1040 in order to reproduce the he - like emission from mgxi , which is the strongest emission line present in the spectrum and in particular the forbidden line component . adopting the above normalization of the emission component , then a column density of @xmath168@xmath2 is required , in the case where the gas fully covers the x - ray source with a solid angle of @xmath169steradian . alternatively if we fix the column density of the emitter to what was previously measured for the zone3 absorber ( where @xmath170@xmath2 , see table4 ) , then a minimum covering fraction of @xmath171 of @xmath169steradian is required to model the emission . a lower limit on the radial location of the emission can also be obtained from the upper limit to the velocity width of the line emission . taking @xmath172 and with an upper limit for velocity width of the mgxi forbidden emission of @xmath173kms@xmath4 ( see table3 ) , then for an estimated black hole mass of mrk1040 of @xmath174m@xmath175 @xcite , the location of the emitting gas associated with zone3 is found to be @xmath176 cm or @xmath177pc . a typical limit to the radial location of the absorber can be derived from the definition of the ionization parameter and the requirement that the thickness of the absorber does not exceed its distance to the supermassive black hole , i.e. , @xmath178 . then from the definition of the ionization parameter , this yields:- @xmath179 thus for the condition that @xmath180 and for the measured zone3 absorber parameters of @xmath131 , @xmath170@xmath2 and an ionizing luminosity of @xmath165ergs@xmath4 , then @xmath181 cm or @xmath182pc . note that the location of the gas can not be much larger than this to satisfy the above geometric condition , while it can be smaller if @xmath183 or if the gas is clumpy . for comparison , the expected sizescale of the narrow line region ( nlr ) can be estimated by the scaling relation of @xcite:- @xmath184 where here @xmath185 is the bolometric luminosity of the agn in units of @xmath186ergs@xmath4 . if we take the 2 - 10kev luminosity to be a typical 3% of the total bolometric luminosity of mrk1040 , then @xmath187ergs@xmath4 and the estimated distance of the optical nlr would be @xmath188pc , consistent with the above minimum and maximum estimates for the radial distance of the ionized gas ( of @xmath189 ) . if the ionized x - ray gas is extended on distances of up to @xmath190pc , then the density of the gas is low , of the order @xmath191@xmath72 for a column density of @xmath192@xmath2 . this may be expected for large scale gas associated with an extended nlr region . note the distance scale of up to 300pc is consistent with archival hst images @xcite , which show extended [ oiii ] emission within @xmath22 of the nucleus ( see section5.3 ) . we can place upper - limits to the mass outflow rate ( @xmath193 ) , which in the quasi - spherical case is given by @xmath194 , where @xmath162 is the gas covering fraction and @xmath195 is the average particle mass relative to h. from the observations , @xmath170@xmath2 , the covering fraction ( as estimated from the emission ) is @xmath196 , while we adopt a likely upper - limit to the outflow velocity of @xmath197kms@xmath4 . thus for the most likely range in the radial location ( @xmath69 ) of the absorber , i.e. from 3pc to 300pc as above , then the limits on the mass outflow rate are @xmath198yr@xmath4 and @xmath199yr@xmath4 respectively . even for the most conservative upper - limit of @xmath199yr@xmath4 , then the corresponding upper - limit to the kinetic power of the outflow is very low with @xmath200ergs@xmath4 . similar constraints can be placed on the other absorber zones . given the lower ionizations of zones 1 and 2 ( table4 ) , equation 2 would predict @xmath201 values 12 orders of magnitude greater than for the more highly ionized zone3 discussed above ( i.e. on kpc scales ) . however a moderate clumping of the gas would restrict it to spatial scales consistent with the extended [ oiii ] emitting gas , which is discussed below . thus for the lowest ionization zone1 ( with @xmath124 ) , if @xmath202 and for the measured column of @xmath203@xmath2 , then the distance is @xmath204 cm ( @xmath205pc ) and thus the density would be @xmath206@xmath72 . this is typical of the estimated densities for line emitting nlr clouds @xcite . furthermore , gas with this low ionization ( @xmath207 will be more than capable of producing [ oiii ] emission in the optical band . indeed as will be shown from subsequent _ xmm - newton / rgs _ soft x - ray spectroscopy of mrk1040 ( paper ii ) , absorption is seen in the o k - shell band covering the full range of charges states ; i.e. from oi - viii , including the strong detection of k - shell absorption from oiii . the latter absorption is also likely to be associated to the low ionization zone 1 gas . interestingly , from the _ hst _ imaging of the [ oiii ] narrow line emitting gas in mrk1040 @xcite , the nlr appears to be extended on spatial scales of @xmath208 from the nucleus . this is consistent with the above maximum estimate of @xmath209pc for the warm absorber at the distance of mrk1040 . @xcite report that the [ oiii ] emission appears bi - conical in structure , with an opening angle of 120 , with the cone axis approximately perpendicular to the host galaxy disk ; the _ hst _ wide field planetary camera 2 ( wfpc2 ) image of mrk1040 is reproduced in figure12 ( left panel ) from these observations . note that no direct evidence is found for extended x - ray emission from the _ chandra _ image of mrk1040 . however the brightness of the central agn , the lower spatial resolution of _ chandra _ and the moderate photon pile - up present in the zeroth order image precludes a quantitative comparison with the _ hst _ image . two _ hst _ space telescope imaging spectrometer ( stis ) observations of mrk1040 in 2009 were performed with the g430 m grating , sampling two different position angles through the central part of the galaxy . the details of the observations and of the analysis proceedures are reported by @xcite , who perform imaging and kinematics ( primarily in [ oiii ] ) of a sample nearby seyfert galaxies with _ hst _ stis . here we show the results obtained in mrk1040 , in order to compare the location and kinematics of the narrow line region emitting gas with the x - ray warm absorber . these results are summarized in figure12 . the two stis observations give two slit positions through different portions of the galaxy , with slit a along the major axis of the galaxy and slit b approximately perpendicular to this . the positions of the two slits are indicated in the left hand panel of figure12 , overlaid on the non dispersed wfpc2 image of mrk1040 at [ oiii ] @xmath210 . the center and right panels of figure12 show the resulting kinematics of the [ oiii ] line components versus position in arcseconds from the agn center . these plots indicate that there is evidence for some modest outflow located within @xmath211 of the nucleus of mrk1040 , with outflow velocities of typically @xmath212kms@xmath4 . this blue - shifted component of [ oiii ] is visible in both of the slit a and b spectra and occurs well within the inner 100pc of the central agn . slit a , positioned along the major axis , also shows extended , blue - shifted [ oiii ] emission up to 0.5 to the southwest of the nucleus and in the direction of the major axis of the galaxy . this emission may be attributed to rotation , if for instance the disk is rotating in a counterclockwise direction and thus the southwest side would appear blue shifted . thus the likely distance of the ionized x - ray gas is consistent with the extended emission observed from the optical nlr , as seen in the hst observations . the x - ray kinematics are also broadly consistent with what is seen in _ hst _ , where some modest outflow is associated with the nucleus itself of up to @xmath212kms@xmath4 . indeed for the zone 1 absorber , which is most likely associated to the [ oiii ] emission , the outflow velocity is @xmath213kms@xmath4 ( see table 4 ) . the aperture of _ chandra _ hetg is much larger than for the _ hst _ stis and thus given the likely upper limit of @xmath182pc to the location of the x - ray absorber , we can not exclude either an origin either associated with the weak nuclear outflow located within @xmath214pc of the nucleus as seen in [ oiii ] , or with the somewhat more extended gas associated with the rotating galactic disk , with low radial velocities . a tighter constraint on the radial distance of the x - ray absorbing gas would be needed to distinguish between these possibilities ; this will be discussed further in paper ii , where we present evidence for variations in absorber ionization between the different epochs . one interesting scenario is that we are viewing a failed nuclear wind from mrk1040 . the accretion rate of mrk1040 , given its luminosity and likely black hole mass , probably does not exceed 10% of eddington . with such a weak agn , it may be difficult to radiatively accelerate gas at distances beyond @xmath215pc @xcite , although this gas could still be ionized by the agn . indeed the stringent upper limit obtained earlier on the outflow kinetic power ( @xmath200ergs@xmath4 ) implies that this weak wind will leave a minimal kinetic footprint on its host galaxy , while the x - ray outflow velocity is also no larger than the velocity dispersion of the stellar bulge in mrk1040 . the low x - ray velocity may also be in part due to our orientation , if for instance we are viewing mrk1040 pole - on compared to a more equatorial wind . the narrowness of the balmer lines in mrk1040 is more likely due to a pole - on view , rather than a high eddington rate as is usually inferred in more typical nls1s . this paper has presented a 200ks _ chandra _ hetg observation of the nearby seyfert 1 , mrk1040 , which is the first time a high resolution x - ray spectrum of this agn has been discussed . the spectra show the strong signature of a warm absorber , with the lower ionization gas responsible for the strong series of inner k - shell absorption lines from ne , mg and si , originating from ions with low charge states . neither the low or high ionization components of the absorber appear to be outflowing , to typically within @xmath3kms@xmath4 of the systemic velocity of mrk1040 . we also find that the lower ionization ( inner k shell ) absorption lines are best reproduced with a model that requires a steep ( @xmath94 ) photon index below 1kev . this argues for an intrinsic origin for the soft x - ray excess from this agn . the low ionization and low velocity of the absorber likely places the absorbing gas between @xmath216 and @xmath190pc from the black hole and it appears coincident with the optical nlr emitting gas , as seen in [ oiii ] in the _ hst _ images of this agn . this is also broadly similar to the soft x - ray absorbers seen towards other seyfert 1s , which may typically range from a pc scale toroidal wind , to the nlr on scales of several tens to hundreds of parsecs ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? finally we suggest that the lack of an x - ray outflow component in mrk1040 may be due to a failed nuclear wind in this agn , with the gas not able to accelerate out to scales beyond @xmath215pc . we thank the anonymous referee for their positive and constructive comments . j.n . reeves acknowledges chandra grant number go3 - 14123x , as well as nasa grant numbers nnx16ae11 g and nnx15af12 g . t.j . turner acknowledges nasa grant number nnh13ch63c . both j.n . reeves and a. lobban acknowedge support from stfc , via the consolidated grants st / m001040/1 and st / k001000/1 . d. porquet acknowledges financial support from the european union seventh framework program ( fp7/2007 - 2013 ) under grant agreement number 312789 . e. behar is supported by the the european union s horizon 2020 research and innovation programme under the marie sklodowska - curie grant agreement no . 655324 and by the i - core program of the planning and budgeting committee ( grant number 1937/12 ) . e.n also acknowledges funding from the european union s horizon 2020 research and innovation programme under the marie skodowska - curie grant agreement no . tcf was supported by an appointment to the nasa postdoctoral program at the nasa goddard space flight center , administered by universities space research association under contract with nasa . the scientific results reported in this article are based on observations made by the chandra x - ray observatory . this research has made use of software provided by the chandra x - ray center ( cxc ) in the application package ciao . 15075 & 2014 - 02 - 25 04:00:17 & heg & 28.6 & @xmath217 + & & meg & & @xmath218 + 15076 & 2013 - 10 - 19 09:24:53 & heg & 19.7 & @xmath219 + & & meg & & @xmath220 + 16571 & 2014 - 03 - 03 12:26:42 & heg & 89.6 & @xmath221 + & & meg & & @xmath222 + 16584 & 2014 - 02 - 26 01:17:35 & heg & 60.0 & @xmath223 + & & meg & & @xmath224 + total & & heg & 197.8 & @xmath25 + & & meg & 197.8 & @xmath24 + lcccccc neiv ( n - like ) & @xmath225 & @xmath226 & @xmath227 & @xmath228 & @xmath229 & 8.1 + nev ( c - like ) & @xmath230 & @xmath231 & @xmath232 & @xmath233 & @xmath234 & 30.3 + nevi ( b - like ) & @xmath235 & @xmath236 & @xmath237 & @xmath238 & @xmath239 & 6.7 + neix he-@xmath5 & @xmath240 & @xmath241 & @xmath242 & @xmath243 & @xmath244 & 13.0 + neix he-@xmath18 & @xmath245 & 1073.7 & @xmath246 & @xmath247 & @xmath248 & 21.2 + nex ly-@xmath5@xmath249 & @xmath250 & @xmath251 & & & @xmath252 & 11.5 + nex ly-@xmath18 & @xmath253 & @xmath254 & @xmath255 & @xmath256 & @xmath257 & 9.6 + mgv ( o - like ) & @xmath258 & @xmath259 & @xmath260 & @xmath214 & @xmath261 & 6.1 + mgvi ( n - like ) & @xmath262 & @xmath263 & @xmath264 & @xmath265 & @xmath266 & 6.5 + mgvii ( c - like ) & @xmath267 & @xmath268 & @xmath269 & @xmath270 & @xmath266 & 7.1 + mgviii ( b - like)@xmath271 & @xmath272 & @xmath273 & @xmath274 & @xmath275 & @xmath276 & 20.7 + mgviii ( b - like)@xmath271 & @xmath277 & @xmath278 & @xmath279 & @xmath280 & @xmath281 & 7.7 + mgxi he-@xmath5 & @xmath282 & @xmath283 & @xmath284 & @xmath285 & @xmath286 & 19.8 + mgxii ly-@xmath5 & @xmath287 & @xmath288 & @xmath289 & @xmath290 & @xmath291 & 14.6 + sivii ( o - like ) & @xmath292 & @xmath293 & @xmath294 & @xmath295 & @xmath239 & 7.8 + siviii ( n - like ) & @xmath296 & @xmath297 & @xmath298 & @xmath299 & @xmath286 & 12.6 + sixi ( be - like ) & @xmath300 & @xmath301 & @xmath302 & @xmath303 & @xmath266 & 6.4 + sixii ( li - like ) & @xmath304 & @xmath305 & @xmath306 & @xmath307 & @xmath308 & 6.3 + sixiii he-@xmath5 & @xmath309 & @xmath310 & @xmath311 & @xmath307 & @xmath312 & 12.4 + sxvi ly-@xmath5 & @xmath313 & @xmath314 & @xmath315 & @xmath243 & @xmath316 & 22.5 + arxviii ly-@xmath5 & @xmath317 & @xmath318 & @xmath319 & @xmath320 & @xmath321 & 8.5 + [ line - widths ] lccccc mgxi @xmath322 & @xmath323 & @xmath324 & @xmath67 & @xmath325 & 9.7 + mgxi @xmath326 & @xmath327 & @xmath328 & @xmath67 & @xmath325@xmath329 & 9.7 + sixiv ly-@xmath5 & @xmath330 & @xmath331 & @xmath332 & @xmath333 & 8.3 + fe k@xmath5 & @xmath74 & @xmath334 & @xmath75 & @xmath335 & 35.2 + fexxv @xmath322 & @xmath336 & @xmath337 & @xmath338 & @xmath339 & 6.3 + [ emission - lines ] lcccc @xmath132@xmath340 & @xmath341 & @xmath342 & @xmath343 & @xmath344 + @xmath345@xmath346 & @xmath347 & @xmath348 & @xmath349 & @xmath350 + @xmath351@xmath352 & @xmath213 & @xmath353 & @xmath354 & @xmath129 + @xmath355@xmath356 & 946.7 & 130.3 & 46.0 & 10.3 + [ absorbers ]

we develop a generic method to compute the dynamics induced by quenches in completely connected quantum systems . these models are expected to provide a mean - field description at least of the short time dynamics of finite dimensional system . we apply our method to the bose - hubbard model , to a generalized jaynes - cummings model , and to the ising model in a transverse field . we find that the quantum evolution can be mapped onto a classical effective dynamics , which involves only a few intensive observables . for some special parameters of the quench , peculiar dynamical transitions occur . they result from singularities of the classical effective dynamics and are reminiscent of the transition recently found in the fermionic hubbard model . finally , we discuss the generality of our results and possible extensions .

1108.5068

thanks to the fast experimental progress on cold atoms @xcite , where direct control over the parameters of the effective hamiltonian is available , many theoretical questions about out of equilibrium dynamics of quantum systems have been raised and started to be addressed . among them , we cite the problem of thermalization for an isolated system ( see e.g. @xcite ) the existence of long - lived out of equilibrium states ( see e.g. @xcite ) , dynamical transitions out of equilibrium @xcite , etc . in this work we focus on the off - equilibrium dynamics induced by quantum quenches , _ i.e. _ the evolution of an isolated quantum system after a sudden change of a parameter of the hamiltonian . this problem has been intensively studied these last years . it provides an useful idealization of the protocols followed in experiments , where the change of parameters takes instead place at finite rates . the literature on quantum quenches is already quite broad , see for example the reviews @xcite . one dimensional systems have been intensively studied by exact analytical and numerical methods . results for higher dimensional systems are instead scarcer ; the previous methods can not be applied and one has to resort to approximations of some kind . the ones that have been more used are the bogoliubov method @xcite , path integral saddle - point expansions @xcite , the large-@xmath0 limit @xcite and mean - field ( or large dimension ) approximations @xcite . interestingly , mean - field approaches revealed that quantum quenches may lead to out of equilibrium dynamical transitions . this was first found in the analysis of the fermionic hubbard model by time dependent dynamical mean field theory ( t - dmft ) in @xcite . later , a confirmation and explanation of this result was found by using a gutzwiller time - dependent ansatz in @xcite . in @xcite we solved the bose hubbard model on a completely connected lattice and found a very similar dynamical transition at integer fillings . this was also noticed in the hard - core bose - hubbard model in a superlattice potential @xcite . finally , clues to argue that these dynamical transitions are generic even beyond mean - field were recently presented in @xcite by mapping the problem to the one of classical phase transitions in films . in this article we first describe a generic approach to solve the dynamics of quenches in completely connected quantum models . then , we apply our method to the bose - hubbard model , the generalized jaynes - cummings model and the ising model in a transverse field . we find that the dynamical transition , discovered for the hubbard model , is a systematic dynamical effect present in completely connected systems characterized by a quantum phase transition in equilibrium . a shorter version of this work , that focused only on the bose - hubbard model , appeared in @xcite . in this article , we focus on the out of equilibrium dynamics in several completely connected models : the bose hubbard model ( bhm ) , the transverse field ising model ( i m ) and the jaynes - cumming model ( jcm ) . the reason for focusing on systems defined on completely connected graphs is that this allows one to study a well defined model and at the same time to obtain an approximate solution for finite dimensional lattices . indeed , such models are related to several approximations used in the literature . among the most known ones , we cite the limit of infinite dimensions for models defined on a hyper - cubic @xmath1-dimensional lattice and the gutzwiller ansatz , a widespread method to obtain mean field equations . actually , in the case of completely connected models , the gutzwiller ansatz can be shown to be exact by using a hubbard - stratonovich transformation . we shall focus on out of equilibrium dynamics induced by a quantum quench . this procedure is defined as follows . let @xmath2 be the quantum hamiltonian , which depends on a coupling @xmath3 . the system is prepared in its ground state @xmath4 at some coupling @xmath5 . at @xmath6 , the system is driven out of equilibrium in a controlled fashion , tuning the coupling @xmath7 in time according to a predefined procedure . except for quasistatic procedures , often called `` adiabatic '' , the wave function becomes different from the ground state @xmath8 . we call _ sudden quench _ the procedure in which the coupling is suddenly switched to a final value @xmath9 : @xmath10 . as we shall show , the quantum dynamics in a completely connected system can be solved by mapping the unitary evolution onto an effective model undergoing newtonian dynamics . this is a drastic simplification which is possible thanks to the symmetry of the completely connected hamiltonian under any permutation of sites . we restrict our analysis to cases where @xmath11 is the ground state at some coupling @xmath5 . thus , @xmath11 is also symmetric under permutation of sites and since both the initial state and the hamiltonian are symmetric , the generic unitary evolution takes place in the sector of symmetric states only . in this subspace , the states are parametrized using a few local macroscopic observables . the unitary evolution can then be written as a schrdinger equation in the symmetric space , which involve an effective @xmath12 , where @xmath13 is the number of sites of the system . thanks to this property , at the thermodynamic limit , the entire dynamics of the system can be encoded in the one of few macroscopic variables , which undergo an effective classical hamiltonian evolution . we first apply this approach to the bose - hubbard model , our main interest because of its applicability to cold atom experiments . in this case , it is the onsite repulsion @xmath14 that plays the role of the coupling @xmath3 . in a first stage we make the additional assumption that the number of bosons per site is less or equal to @xmath15 . the average superfluid order after the quench is a non - monotonous function of @xmath16 . actually , it decays logarithmically to zero at some special values of @xmath17 and @xmath18 . for these special quenches , the superfluid order relaxes exponentially to zero . we call this peculiar feature a dynamical transition . surprisingly , the microcanonical equilibrium characterized by same energy , towards which the system would relax on large times if it were able to thermalize . ] , has non - zero superfluid order . thus , this transition is a purely _ dynamical _ phenomenon , which has nothing to do with the fact that the superfluid order vanishes in the high temperature phase . we extend our analysis to @xmath19 . in this case the effective dynamics involves @xmath20 degrees of freedom , and the classical trajectories can be either regular or chaotic . apart from this difference , the dynamical transition is found to be qualitatively unchanged and the quantitative differences with @xmath15 are small . finally , we focus on two other systems , for which some out of equilibrium properties have already been studied , a generalized jaynes - cummings model @xcite , and the ising model in a transverse field @xcite . we apply our formalism to both systems and show that also in these cases there is a dynamical transition for sudden quenches . this paper is organized as follows : in section [ models ] , we briefly describe the three considered models . in section [ general ] , we describe how the effective classical dynamics can be derived for an arbitrary hamiltonian . then in section [ special ] , we derive the classical dynamics for the bose - hubbard model with truncation @xmath15 , and analyze quenches and the dynamical transition in section [ quenches ] . the general case @xmath21 is considered in section [ more ] . the effective dynamics and quench properties of the generalized jaynes - cummings model are described in section [ jaynes ] , and in section [ ising ] for the ising model . section [ discussion ] and [ conclusion ] contain discussions and possible extensions of our work . [ gutzwiller ] is devoted to prove that the same effective dynamics is recovered with a gutzwiller ansatz wave function , and in [ wkb_expansion ] the wkb eigenstates of the completely connected model are discussed . the bose - hubbard lattice model has regained a lot of interest recently since it can be realized and studied in experiments on cold atoms @xcite . the hamiltonian of its completely connected version , suited to describe the limit @xmath22 of the lattice model , reads @xmath23 where @xmath24 , @xmath25 are the bosonic creation and annihilation operators , satisfying @xmath26 = \delta_{ij}$ ] , and @xmath27 . the first term is an on - site repulsion between bosons , the second is a tunneling term , of amplitude @xmath28 , and rescaled by the volume @xmath13 ( number of sites ) . this rescaling is needed to obtain a well - defined thermodynamic limit @xmath29 . the limit of infinite dimensions that is equivalent to this model is obtained by taking a coupling @xmath30 for a bhm defined on a @xmath1-dimensional hyper - cubic lattice . this model undergoes a quantum phase transition at _ commensurate _ fillings : @xmath31 with @xmath32 integer . see @xcite for a detailed discussion of the phase diagram . notice that since we consider the dynamical behavior , the number of particles is fixed because it corresponds to a quantity conserved by the dynamics . in consequence we shall discuss the phase diagram in terms of the density and not the chemical potential , as it is instead done usually . let us focus first on cases where the number of particles per site is an integer . to grasp why there is a quantum phase transition , it is helpful to compare the ground state in the two limits @xmath33 ( or @xmath34 ) and @xmath35 . in the first case , the ground state is diagonal in the occupation number @xmath36 . it remains the same for all @xmath37 , and is called a mott insulator . it is incompressible because adding or removing a boson requires an energy of the order of @xmath14 . in the second case , @xmath38 , the ground state is akin to corrections . ] a product of coherent states @xmath39 . for @xmath38 , and also moderate values of @xmath14 , the ground state is superfluid and compressible . the usual order parameter @xmath40 is zero because the density is fixed , hence the appropriate parameter is the off - diagonal long range order measured by @xmath41 . in the completely connected model , all distances @xmath42 between two different sites are equal to @xmath43 . this is the largest distance in the problem . in consequence one can define ] : @xmath44 , the order parameter @xmath45 vanishes and the ground state becomes a mott insulator . this is the mott - superfluid quantum phase transition . for non - integer fillings , the system is instead always superfluid and compressible as it can be understood considering perturbation around the @xmath46 limit . at non zero temperature , there is a second order phase transition from the superfluid phase to a bose gas at a critical temperature @xmath47 . the second model that we consider is the generalized jaynes - cummings model studied in @xcite , which describes @xmath0 distinguishable two - level systems in interaction with a single quantized electromagnetic mode . it is useful in various contexts , and has been suggested as a qualitative description of the bec / bcs crossover in fermionic condensates , of molecular magnetism and of the formation of dimers by pairing of two bosons . recently , new experiments in the setting of cavity quantum electrodynamics @xcite are realizations of the dicke hamiltonian @xcite , which can be mapped onto the jaynes - cummings hamiltonian if the `` rotating wave approximation '' is made . the dynamics of sweeps , starting from an empty bosonic mode , was studied in @xcite using quasiclassical approximations and the truncated wigner approximation . here , we describe the dynamics of sudden quenches from the broken symmetry phase . the model is defined as follows . there is one bosonic @xmath48= 1 $ ] degree of freedom the electromagnetic mode and one su(2 ) spin @xmath49 . the spin degree of freedom keeps track of the number of excited two - level systems , @xmath50 . with @xmath51 , the hamiltonian is @xmath52 it includes a potential energy for bosons ( @xmath53 ) , for the spin ( @xmath54 ) , and a coupling @xmath55 between the two . each time a boson is lost , a two - level system is excited , and reciprocally : the quantity @xmath56 is conserved by the dynamics . it is thus possible to define @xmath57 such that up to a constant term one finds : @xmath58 in the following we take @xmath59 and consider @xmath3 in units of @xmath55 , and the time in units of @xmath60 . we consider the regime @xmath61 ( which can be reduced to @xmath62 ) , for which there is a quantum phase transition . at @xmath63 , the system is in the _ normal _ ground state @xmath64 if @xmath65 and is in the _ super - radiant _ ground state for @xmath66 where @xmath67 . this is the super - radiance quantum phase transition . this transition is the consequence of the competition between the two terms in the hamiltonian : the first one favors removing bosons ( if @xmath68 ) , whereas the second plays the role of a kinetic energy and it can lower the energy if the bosons modes are filled , see section [ jaynes ] . as for the bose - hubbard model , there is a finite temperature phase transition at @xmath69 , above which the super - radiant phase disappears . notice that this phase transition is usually studied in the canonical / microcanonical ensemble , whereas here it is more natural to focus on given value of @xmath70 since this quantity conserved by the dynamics . the ising model in a transverse field is a paradigm of quantum phase transitions @xcite . the hamiltonian of its completely connected version reads @xmath71 where @xmath28 is the ferromagnetic coupling and @xmath72 the transverse field . for simplicity , we set @xmath73 , measure @xmath72 in units of @xmath28 and time in units of @xmath74 . the hamiltonian can be written using a single spin @xmath75 and reads @xmath76 the large @xmath0 limit corresponds to the large spin @xmath49 limit , which is also the classical limit . the ground state is found , minimizing the corresponding classical hamiltonian @xmath77 @xcite . the minimum is at @xmath78 , and for @xmath79 , @xmath80 , which is a _ ferromagnetic _ ground state , in which the spins align toward the @xmath81-axis . for @xmath82 , @xmath83 and the ground state is called _ quantum paramagnet _ , because it is unoriented in the @xmath81-basis @xmath84 , @xmath85 . the quantum phase transition from z - ferromagnet to quantum paramagnet takes place at @xmath86 . starting from the ferromagnetic ground state and increasing the temperature the system undergoes a second - order phase transition at a finite temperature @xmath87 . in the following , we show how the dynamics induced by quantum quenches , in arbitrary completely connected models , can be mapped onto an effective classical hamiltonian dynamics . as already stated , any model defined on a completely connected lattice has a hamiltonian which is symmetric under any permutation of sites . thus , one expects that the ground state of the model is _ also _ site permutation symmetric . we shall study quenches starting from the ground state of the system . since the site permutation symmetry is conserved by the unitary evolution , the dynamics only takes place into the subspace of symmetric wave functions . this is a drastic simplification , because symmetric states can be described using a few variables only . to clarify this point in a concrete but simple case , let us see what these variables are for the bose - hubbard model , with the additional constraint that there are @xmath88 or less bosons per site . first , remark that given a particular fock state @xmath89 , the linear combination of every possible site permuted fock state is a site permutation symmetric state @xmath90 . this new state is completely characterized by the fraction @xmath91 , @xmath92 , @xmath93 of sites with @xmath94 , @xmath43 and @xmath95 bosons respectively , and we call it @xmath96 where @xmath97 is a shorthand for all variables . since the fock states form a basis of generic states , the @xmath98 states form a basis of the symmetric sector . in order to express the schrdinger evolution in this basis , we have to compute the transitions elements @xmath99 . typically , only few transitions @xmath100 are allowed , with @xmath101 a vector of integers . for example in the bose - hubbard model , the matrix elements of the hamiltonian connect states that differs by one boson jump . thus @xmath91 , @xmath92 and @xmath93 can only differ by @xmath102 or @xmath103 . for example , after a jump @xmath104 , @xmath105 , @xmath106 and @xmath107 . this transition is labeled @xmath108 . moreover , the reverse move is a transition characterized by @xmath109 , with same amplitude @xmath110 at dominant order in @xmath13 . thanks to the `` locality '' of transition elements between symmetric states , the schrdinger equation in the symmetric basis @xmath111 takes a simple form . the schrdinger equation projected on @xmath112 reads @xmath113 . denoting the diagonal transition element @xmath114 , we get @xmath115 where we made use of @xmath116 with an implicit summation over the index @xmath117 . strikingly , the schrdinger equation involves an effective @xmath118 , thus the regime of interest @xmath29 corresponds to the _ classical regime_. furthermore , for the ground state , the initial wave function is a narrow wave packet of width @xmath119 ( this general property is discussed in more detail for the bose - hubbard model in section [ classical_evolution_of_packet ] ) . thus , by the heisenberg uncertainty principle , we expect that momentum fluctuations are proportional to @xmath120 . therefore the evolution of the wave - packet can be fully described in the thermodynamic limit by its average ( or center ) @xmath121 and average momentum @xmath122 . both quantities evolve following a classical hamiltonian dynamics obtained from the quantum one by replacing @xmath123 and @xmath124 : @xmath125 @xmath126 = d(x ) - 2 \sum_{m } w_{m}(x ) \cos(m_i p_i)\ ] ] the classical hamiltonian evolution of the variables @xmath127 and @xmath128 yields the evolution of the wave packet after the quantum quench , and give access to all observables as a function of time . a more careful analysis is performed in section [ rigorous ] and fully supports this picture . as a conclusion , the analysis of quench dynamics in connected models is tractable , and takes the form of an effective classical hamiltonian evolution . the initial condition is provided by the ground state values of @xmath129 at the initial coupling @xmath17 . the sudden quantum quench dynamics is then described by the classical dynamics of @xmath130 induced by the effective hamiltonian @xmath131 , with initial conditions @xmath129 . in the following , we study the bose - hubbard model with two or less bosons per site . we derive explicitly the effective classical hamiltonian in section [ bose - hubbard_hamiltonian ] . the nature of the ground state , in particular the peaking of the wave packet when @xmath133 and the quantum phase transition , are discussed in section [ classical_evolution_of_packet ] . a more thorough derivation of the effective hamiltonian dynamics is presented in section [ rigorous ] . finally , we conclude by describing the phase space of effective trajectories in section [ types ] . in this section , the method sketched above is applied to the bose - hubbard model with truncation @xmath134 . this restriction is well - suited to derive analytical expressions , and preserves the main features of the model , such as the mott insulator to superfluid quantum phase transition . the case of a larger number of bosons per site is considered in section [ more ] . the results remain qualitatively the same . the symmetric states are written under the form @xmath135 . since the total number of sites @xmath136 is fixed , and the overall density of bosons @xmath137 is conserved by the dynamics , the symmetric states can be labeled by one variable only . choosing @xmath92 as such a variable , the symmetric normalized wave functions are denoted @xmath138 , where @xmath139 and @xmath140 means the sum over fock states of fixed fraction @xmath92 . in the following for simplicity of notation we drop the subindex one and use @xmath97 instead of @xmath92 . note that contrary to the previous paragraph , @xmath97 is now just a number and not a vector . + to compute the transition rates , we proceed as follows . the repulsion term @xmath141 is diagonal in the @xmath142 basis . the tunneling term @xmath143 allows transitions from @xmath144 to three states , @xmath145 , @xmath146 and @xmath144 . using the previous notations , the only possible transitions correspond to @xmath147 , @xmath148 , @xmath149 representing respectively @xmath150 , @xmath151 and @xmath152 . the first amplitude reads @xmath153 there are @xmath154 factors on the ket side . for each of these factors , there are @xmath155 transition to a state with @xmath156 . therefore using that @xmath157 , and after a partial cancellation of the normalization factors , we obtain : @xmath158^{1/2}\ ] ] in the following , only dominant contributions of order @xmath13 are kept , the sub - leading ones are dropped since they do not matter for dynamics taking place on times not diverging with the system size . rewriting @xmath91 and @xmath93 as functions of @xmath97 , the transition rates are @xmath159^{1/2 } & \equiv - v\ , \hat w_x \\ \displaystyle \langle x |\hat{h}|x \rangle & = & v \ , u(n - x)/2 - v \ , x(2+n-3x)/2 & \equiv v\ , \hat d_x\\ \displaystyle \langle x + 2/v|\hat{h}|x \rangle & = & \langle x -2/v|\hat{h}|x \rangle & \\ \end{array}\ ] ] the schrdinger evolution on @xmath160 in the site permutation symmetrical basis is @xmath113 @xmath161 therefore we have obtained that the hamiltonian in the subspace of symmetric states reads @xmath162 note that the dimension of the hilbert space has been reduced from @xmath163 to @xmath13 in this case . the case @xmath164 is especially convenient because the effective classical motion takes place in one dimension , therefore the equation of motion are integrable and the evolution easy to understand and describe . as in section [ general ] , the schrdinger equation involves an effective @xmath118 , and its classical limit is obtained through @xmath165 and @xmath166 . @xmath167 = d(x ) - 2 w(x ) \cos(2p)\ ] ] the possibility of describing the quantum dynamics after a quench in terms of classical dynamics relies on the fact that the initial wave function is initially of small width ( @xmath168 ) . in order to check this , let s first expand the eigenvalue equation at lowest order in @xmath102 : @xmath169 since the kinetic term has a factor @xmath170 in front , the ground state corresponds to the minimum of @xmath171 and small quantum fluctuations around it . indeed , around the minimum , @xmath172 , a quadratic expansion of @xmath173 and @xmath174 maps this problem onto a quantum harmonic oscillator with @xmath175 and @xmath176 . thus , we find that the ground state wavefunction is centered around the absolute minimum @xmath172 of the potential @xmath177 and has a width @xmath178 , which is of the order of @xmath119 . an appropriate expression for any eigenstate is provided by the wkb expansion is given in section [ wkb_expansion ] . the quantum phase transition taking place as a function of @xmath14 can be recovered within this formalism as we now show . we recall that for the bose - hubbard model , the quantum phase transition and the mott phase are present at commensurate fillings only , so we have to focus on @xmath179 for @xmath15 . the ground state of the quantum hamiltonian corresponds to the global minimum of the effective hamiltonian , given by @xmath180 . these equations lead to @xmath181 and @xmath97 equal to the value @xmath172 , which verifies @xmath182 . it is easy to check that @xmath183 where @xmath184 ( note that @xmath185 ) . in the thermodynamic limit , the mott insulator ground state is @xmath186 . instead the superfluid ground state is @xmath187 , with @xmath188 . therefore , in the superfluid state , a fraction of sites @xmath189 contain two bosons . @xmath190 is proportional to the intensive kinetic energy , and can be computed using the identity @xmath191 . it is the order parameter for the transition : it is positive in the superfluid phase and vanishes in the mott insulating one . in the following we show in detail how the classical hamiltonian dynamics emerges from the quantum evolution of symmetric states in the thermodynamic limit . for simplicity we consider the situation of section [ special ] , where @xmath192 , but the argument is more general . the schrdinger equation reads @xmath193 a ground state wave function is a wave packet characterized by a small width of the order of @xmath119 as shown in section [ classical_evolution_of_packet ] . the width broadens after a long time , which diverges with @xmath13 . on times that do not diverge in the thermodynamic limit , the ground state wave function and its subsequent evolution can be written as @xmath194 with @xmath195 , @xmath55 and @xmath196 being respectively the envelope and the phase of the wave - packet . the envelope @xmath197 has a maximum at @xmath198 , and expanding @xmath197 around @xmath198 shows that indeed this wave function describes a packet of width @xmath119 . after these preliminary considerations , we proceed and evaluate the evolution of the wave function , plugging the expression ( [ psiexp ] ) of @xmath199 into ( [ schrodinger2 ] ) : @xmath200 for the amplitude and the phase , this leads to : @xmath201 the position of the peak @xmath198 corresponds to the maximum of the amplitude , thus it satisfies the implicit equation @xmath202 . to obtain its evolution , we differentiate it with respect to time : @xmath203 the term @xmath204 above can be computed from ( [ amplitude ] ) . at the point where @xmath202 , ( [ first ] ) becomes : @xmath205 this equation involves only one unknown quantity , @xmath206 , which , as we shall show , is akin to a momentum . a self - consistent equation for this quantity can be found taking its derivative with respect to time : @xmath207 in the previous expression , @xmath208 is the space derivative of ( [ phase ] ) . using ( [ evo_r ] ) the two terms in @xmath209 cancel out , and ( [ second ] ) reads : @xmath210 this , together with eq . ( [ evo_r ] ) , provides a set of closed differential equations for @xmath198 and @xmath211 . the last thing we have to show is that the quantity @xmath212 is the average value of the momentum operator @xmath213 , which reads : @xmath214 because of the form ( [ psiexp ] ) of @xmath160 these integrals can be performed by the saddle point method ( in the limit @xmath133 ) . at the saddle point , @xmath215 and @xmath216 has a non zero value , thus @xmath217 . the same is true for @xmath218 . thus , rewriting ( [ evo_r ] ) and ( [ third ] ) , one finds that the phase @xmath219 and the position @xmath198 obey the equations : @xmath220 as a conclusion , the average values @xmath198 and @xmath219 of the position and momentum operator , which describe the global behavior of the wave packet , have the same time evolution as classical canonical variables with hamiltonian @xmath221 = d_x - 2 w_x \cos(2p)$ ] . this property fits into the general picture of the propagation of wave packets in the semi - classical regime ( see for example @xcite for a mathematically oriented review ) and , hence , was expected on general grounds as discussed previously . the generalization to more than one variable is straightforward . the width of the packet can also be evaluated . it is related to the separation in time of two classical trajectories initially separated by the width of the packet . when the effective dynamics is one dimensional , the classical hamiltonian is integrable and periodic orbits separate linearly in time . in this case , since trajectories start from an initial distance @xmath119 , the typical time of separation is @xmath222 . we check numerically that our analysis of the limit of large @xmath13 is correct . in figure [ fig : comparison ] , the exact quantum evolution , obtained by diagonalization of the discrete equation , is compared to the classical evolution for short times . we find an excellent agreement . for @xmath223 , with initial conditions @xmath224 of energy @xmath225 . the dots are obtained by numerical diagonalization of for @xmath226 with a sharp gaussian initial condition . the line is the evolution according to the classical hamiltonian . the two are equivalent on short times . similar results hold for any trajectory and initial condition . ] we now study the effective classical dynamics by focusing on the phase space properties . + since the hamiltonian is one - dimensional , and energy is conserved there are lots of constraints on the motion . in particular all the trajectories are integrable , and can only be periodic in @xmath198 except for separatrix trajectories . + we find that the phase space is divided in two regions by a separatrix . to see this , one can characterize trajectories in terms of their turning points , and considering whether the momentum @xmath227 is bounded or not . there are three types of turning points : the derivative @xmath228 vanishes if either @xmath181 , @xmath229 or @xmath230 . this last case is called `` absorbing '' for a reason that will be explained later . let us compare three trajectories at different energies for @xmath37 , which are representative of all the cases encountered . for this purpose , two representations are shown : the evolution of @xmath198 , @xmath219 in figure [ fig : traj_sample ] , the phase space in figure [ fig : traj ] ( right panel ) . the three types of trajectories are : * \(a ) has two @xmath231 turning points , and thus its momentum @xmath227 is bounded . * \(b ) is a separatrix trajectory in the classical mechanics sense . the left turning point is at @xmath231 , and the right turning point at @xmath232 is `` absorbing '' , the time taken to reach it ( or escape from it ) is infinite . in the equation of the motion , at the point @xmath233 , the effective mass tend to infinity . * \(c ) has one turning point type at @xmath231 and one at @xmath234 . @xmath227 is growing infinitely large with time , which is not pathological because only @xmath227 modulo @xmath235 enters the equations . the diagram in figure [ fig : traj ] ( left panel ) is similar in the spirit to the figures of effective potential shown to explain central motions in textbooks . it allows one to understand the dynamical evolution in a simple way . actually , since the classical energy @xmath236 is conserved , the values of @xmath97 during the motion are restricted by the conditions @xmath237 ( note that @xmath238 for @xmath239 ) . except for the separatrix , all turning points correspond to either @xmath231 at @xmath240 , or @xmath234 at @xmath241 . thus , all trajectories are delimited by the two turning points @xmath242 and @xmath243 , which are at the intersection between @xmath244 and @xmath245 , and @xmath244 and @xmath177 . the evolution of the system consists then in a periodic motion oscillating between @xmath242 and @xmath243 . [ types ] and @xmath246 modulo @xmath235 for the three trajectories @xmath247 , @xmath248 , @xmath249 . the scale is the same for all graphs . ] in the previous discussion we considered a specific form of the phase space and of @xmath250 , @xmath251 which are valid for certain values of @xmath14 only . in the next section a careful investigation of the @xmath14 dependence will be presented . however , before that , we want to stress that the effective potentials @xmath250 , @xmath251 can take three different qualitative form , depending on the coupling @xmath14 . these are shown in figure [ fig:4 ] . the three regimes are separated by two special values of @xmath14 : @xmath252 and @xmath253 , see figure [ fig:4 ] . @xmath253 is the critical coupling of the quantum phase transition , and @xmath254 . the special regime @xmath255 disappears when larger filling numbers @xmath19 are included , so it is a peculiarity of the @xmath15 case and , hence , not very relevant . we now discuss the dynamical evolution following a quantum quench and its dependence on the final and initial value of @xmath14 . at @xmath256 the system is in the ground state at the coupling @xmath17 . at @xmath257 the coupling is switched to @xmath18 and the quench dynamics is computed for @xmath258 . in the following we call a quench `` from superfluid '' when @xmath259 and `` from mott '' when @xmath260 , and `` to superfluid '' or `` to mott '' when @xmath261 and @xmath262 . the results of the following sections are summarized by the dynamical phase diagram shown in [ fig : summary ] ( left panel ) . starting from the mott ground state @xmath263 , the trajectory is stuck at @xmath264 even for large times . in order to check this , let us linearize the equation of motion around @xmath265 . we use that @xmath266 and that @xmath267 can be extracted from @xmath268 to obtain @xmath269 . thus , at dominant order in @xmath270 we obtain the equation for @xmath271 : @xmath272 the trajectory @xmath273 is unstable , and since the wave function has a width @xmath119 , its typical evolution is given by @xmath274 . therefore , in the effective picture , the trajectory is stuck at the mott ground state on times of the order of @xmath275 . this result is a peculiarity , actually a pathology , of mean field models . indeed , in a real finite dimensional system , spatial fluctuations drive the system away from the mott state in a finite time . in a superfluid to mott quench , the initial condition is given by the ground state packet characterized by @xmath276 . the trajectory after the quench is of the type ( c ) , see quench q1 of figure [ fig : quenches ] . the superfluid order parameter @xmath277 ( equation ) oscillates . depending on the value of @xmath17 and @xmath18 , there are three different types of dynamical evolutions in this case , that we call q2 , q3 and q4 in figure [ fig : quenches ] . q2 is of type ( c ) and q4 is of type ( a ) . q3 , reached from a special @xmath279 ( dependent on @xmath18 ) , corresponds to a singular separatrix of type ( b ) and of energy @xmath280 . in this case the packet relaxes exponentially to the _ mott insulator ground state _ @xmath281 , and concomitantly the superfluid order @xmath45 vanishes exponentially in time . this may be seen from the linearization , which becomes @xmath282 for a trajectory relaxing to @xmath265 . approaching the transition oscillations take place on a time scale that diverges as @xmath283 . in consequence we find that a dynamical singularity , or transition , takes place at @xmath284 . the values of @xmath284 depends on @xmath17 and thus defines a dynamical transition line @xmath285 , in the @xmath286 plane . in figure [ fig : summary ] ( right panel ) , as an example of singular behavior , we show the time average @xmath287 as a function of @xmath18 for quenches starting from the non interacting case @xmath288 . + it is interesting to compare to @xmath287 its equilibrium counterpart obtained from the microcanonical average corresponding to the same energy . from this we clearly see that the system is not thermalized . actually , large quenches ( q2 ) have non vanishing oscillations of the superfluid order contrary to the equilibrium value which is instead zero ( because it corresponds to an effective high temperature ) . moreover , we find that the dynamical transition takes place at an energy at which the system , if it were relaxed , would be superfluid . thus , the exponential relaxation of the superfluid order is a purely _ dynamical effect_. in this regime , there is a qualitatively new family of quenches q6 ( comparable to q2 ) . there is also a new special quench q5 to the modified separatrix state at energy @xmath290 , which gives rise to another dynamical transition . however , unlike the other quench cases , these effects are artifacts of the truncation @xmath15 and disappear for @xmath19 . therefore , they are not indicated in figure [ fig : summary ] ( left panel ) , where the outcome of all possible sudden quenches is summarized . the previous analysis focused on two bosons or less per site . this constraint can be relaxed to include up to @xmath291 number of bosons per site . when @xmath291 is sufficiently high compared to the density , @xmath292 , one recovers the behavior of the bose - hubbard model with no constraints on the occupation number per site . for a given maximum number of bosons per site , @xmath291 , any symmetric wave function can be parametrized by the fractions @xmath293 of sites with @xmath117 bosons per site , @xmath294 $ ] . since the @xmath293 are fractions , they verify @xmath295 . moreover the density @xmath296 is fixed , thus there are only @xmath20 free variables left . the wave function is expanded in the symmetric basis like @xmath297 . the transition elements @xmath298 and @xmath299 can be computed as done previously . for instance , there are 3 different types of transitions @xmath300 when @xmath301 and 6 when @xmath302 . performing the classical equivalence for packet states , the resulting hamiltonian can be put in the form : @xmath303 specifically , when @xmath304 , if one chooses @xmath92 and @xmath93 as free variables , the hamiltonian is @xmath305 notice that because @xmath306 , there are constraints on the possible values of @xmath293 , such as @xmath307 . for all @xmath291 , there is a mott insulator to superfluid quantum phase transition at some coupling @xmath253 if the density @xmath32 is an integer ( lower than @xmath291 ) . above the critical coupling @xmath308 the ground state is a mott insulator @xmath309 ( all sites have @xmath32 bosons ) , and below @xmath37 the ground state is superfluid with all @xmath310 . in the previous case @xmath15 , the effective dynamics was one - dimensional , and thus integrable . for @xmath311 , the classical dynamics takes place in two or more dimensions , and the trajectories may be either regular or chaotic . in order to characterize them we study their regularity properties . for chaotic trajectories , neighboring trajectories separate exponentially in time in the phase space @xmath312 , like @xmath313 . the rate of separation @xmath3 is the largest lyapunov exponent @xcite . we computed @xmath3 for @xmath301 at density @xmath179 for several trajectories , using a simplified version of the traditional gram - schmidt orthonormalization of the lyapunov vectors . roughly speaking , if we write the hamiltonian evolution under the form @xmath314 , the deviation satisfies @xmath315 . this can be integrated numerically and normalized at each step to avoid an overflow ( the orthogonalization step is dedicated to find all lyapunov exponents , whereas here we only need the largest one ) . the lyapunov exponents for trajectories with different initial conditions @xmath316 are plotted in figure [ fig : lyap ] in the superfluid phase @xmath317 . trajectories are regular in some regions of the space ( periodic or quasi periodic ) with @xmath318 within the error bar , whereas some other regions are chaotic with @xmath319 . the quench from @xmath280 is exactly at the dynamical transition , the corresponding trajectory is chaotic . for @xmath320 , when the ground state is a mott insulator , all trajectories are regular . we notice that the regularity of a trajectory affects the time of spreading of the packet ( determined by the time of separation of two neighboring trajectories ) . actually , the time of separation is typically polynomial @xmath321 for a regular motion but only @xmath322 for a chaotic one . in quenches from the superfluid phase @xmath323 , like in the previous case @xmath15 , a dynamical transition occurs at the special coupling @xmath324 where the final energy equals the energy @xmath280 of the unstable mott trajectory @xmath309 . direct evidence of this transition is given by the singularity of the time averaged superfluid order @xmath45 as a function of @xmath18 for a given @xmath17 . in figure [ fig5]b , we compare this singularity in @xmath45 for @xmath325 and find that the dependence in @xmath291 of the divergence is very weak beyond @xmath326 . the two curves @xmath302 and @xmath327 are not distinguishable , because they are identical up to @xmath328 . + we show the dynamical phase diagram in figure [ fig : lyap]b for @xmath302 and unit filling factor . we observe that the transition line is near to the transition line for @xmath15 , and that they are asymptotically equal around @xmath253 . because the probability of having more than 4 bosons on the same site is extremely small , of the order of @xmath328 when @xmath329 , we can safely assume that this phase diagram is quantitatively representative of the phase diagram _ without truncation_. even though the existence of the dynamical transition for any @xmath291 is beyond doubt , because the singularity is numerically manifest , more precise results on this transition , even for @xmath301 , are hard to provide . some features of the case @xmath15 persist ; for example , at the dynamical transition , the momentum @xmath330 becomes unbounded , see fig . the fractions @xmath331 are also oscillating , but without definite period . they are either quasi - periodic for low @xmath3 regions , or chaotic . approaching the dynamical transition , the typical time of return to @xmath332 is increasing , possibly logarithmically divergent in @xmath333 as suggested by the numerical divergence in figure [ fig5]b . unfortunately , the analysis of the singularity for @xmath334 turns out to be out of reach . the numerical integration of classical equations of motions is exponentially sensible to numerical errors and thus not reliable . one could imagine that the trajectory at @xmath280 is an unstable manifold , which would support the existence of a singularity for trajectory arbitrarily close to it . however , it is hard to decide whether the trajectory at the dynamical transition goes arbitrarily close to the point @xmath335 ( ground state of the mott insulator ) , at which the trajectory is exponentially slowed down . a possible scenario , in which the surface @xmath280 is ergodic , does not seem to be validated by numerical integration of trajectories . we now consider the generalized jaynes - cummings model and derive its quench dynamics . for simplicity and without loss of generality we focus on @xmath336 . thanks to the conservation of @xmath337 , the spin degrees of freedom can be accounted for in terms of the bosonic ones . the states can be parametrized by the density of bosons @xmath32 only , in the fock basis @xmath338 : @xmath339 the action on @xmath340 of some useful operators can be readily computed at the dominant order in @xmath0 @xmath341 from these equations , the transition elements @xmath342 and @xmath343 are easy to compute , and yield an effective hamiltonian on @xmath32 and @xmath344 its conjugate momentum @xmath345 = d_n - 2w_n \cos(\phi)$ ] @xmath346 = -2 \lambda n + 2 n\sqrt{1-n}\cos(\phi)\ ] ] which is the semi - classical approximation of the original hamiltonian obtained in @xcite , up to a shift in the phase @xmath347 . in @xcite , the authors studied sweeps from the empty state @xmath348 , and the link to the landau - zener problem . here , we focus on sudden quenches and the related dynamical transition . the ground state of this effective system is at @xmath349 and @xmath350 the super - radiance quantum phase transition is the result of a competition between the potential energy term and the kinetic energy term , which can be understood if the quantum hamiltonian is written in the @xmath340 basis : @xmath351 qualitatively , the potential term @xmath352 encourages filling for @xmath353 and discourages it for @xmath354 . the `` kinetic '' contribution @xmath355 can lower the energy . as a matter of fact , the sign of @xmath55 is almost irrelevant . changing the sign of @xmath55 leads to @xmath356 . however , the new hamiltonian can be mapped into the old one by the translation @xmath357 . in consequence , if the ground state of a given hamiltonian is at @xmath181 , @xmath358 $ ] , then the ground state of the hamiltonian with the opposite sign of @xmath55 is at @xmath359 , @xmath360 $ ] , identical but with a staggered sign . as a function of the parameter @xmath3 . the quantum phase transition takes place at @xmath361 . b ) quench diagram for all @xmath362 . the quench trajectory is of bounded momentum @xmath344 ( light gray ) or unbounded @xmath344 ( white ) . on the dashed line , the separatrix states are not singular ( absorbing ) . the thick line is the dynamical transition , where the trajectories are singular . for @xmath363 , in the dark gray region , trajectories are definitely stuck at @xmath364 . otherwise , there is a slow relaxation on times of order @xmath365.,width=529 ] previously , we have shown that the characterization of turning points plays an important role for understanding the dynamical behavior . in the generalized jaynes - cummings model , there are two special turning points , @xmath366 , identified by the equation @xmath367 . like in the bose - hubbard model , only one gives rise to singular trajectories . this is readily seen , for example , computing the time taken to reach the singular point and having zero kinetic energy at this point : @xmath368 using the conservation of the energy @xmath369 , we can eliminate @xmath370^{1/2}$ ] and obtain @xmath371^{-1/2}\ ] ] around @xmath372 , the time is finite @xmath373 , whereas around @xmath374 , the time @xmath375 diverges . in other words , the trajectories touching the turning point @xmath376 have a finite period , whereas the trajectories touching @xmath364 have infinite period . the situation is anologous to the one already studied for the bhm : the absorbing state @xmath377 plays the same role of the mott state , and gives rise to the dynamical transition . + to identify the critical value @xmath378 at which the transition takes place , one can compute the energy after a quench @xmath362 and compare it to the energy of the system at rest in @xmath364 . one can check that a trajectory @xmath379 starting with the same energy , actually energy zero , is indeed driven to the absorbing point @xmath380 . this dynamical transition occurs on the line @xmath381 . a linearization similar to shows that the relaxation is exponential at the transition with a characteristic time @xmath382 . around the transition , the period of a trajectory diverges as @xmath383 . the dynamical phase diagram is shown in figure [ fig : jaynes]b . for @xmath384 , the initial state is in the broken symmetry phase @xmath385 . the situation is very similar to the bhm one : the filling number @xmath379 oscillates in time after the quench , and there is one region with bounded effective momentum @xmath344 , and two regions with unbounded @xmath344 . notice that there are two separatrices between the three regions , and that one is _ not singular _ , and does not give rise to a dynamical transition , whereas the other is . the singular one is the one for which the absorbing state @xmath364 is met , corresponding to the restored symmetry state . for quenches with @xmath386 , the initial state is empty @xmath387 , and either drifts away from zero on large times ( of the order of @xmath365 ) if @xmath388 , and otherwise , the state is definitely stuck at @xmath364 . the bounded or unbounded nature of the final state is also indicated . finally , we remark that eigenstates of the completely connected model can be written within a wkb expansion , this is done for completeness in [ wkb_expansion ] . in our last example , let us consider the dynamics due to quantum quenches in the transverse field ising model . to derive the classical effective dynamics , one can use the fact that the large spin limit is also the classical limit , see e.g. @xcite . to do so , one should make the substitution @xmath77 . the kinetic term ( needed to describe classical dynamics ) in the classical hamiltonian of the large spin limit can be derived using path integral for spins @xcite . here we instead use our generic method based on site permutation symmetry . we denote spin symmetric states of total momentum @xmath49 with the notation @xmath389 . they are such that @xmath390 with @xmath391 $ ] . at dominant order in @xmath0 , @xmath392 . one can proceed as usual to get the effective hamiltonian : @xmath393 = d_s - 2 w_s \cos\phi & = -\displaystyle\frac{1}{2 } s^2 -\gamma\sqrt{\frac{1}{4}-s^2 } \cos \phi\\ \end{array}\ ] ] as expected , the hamiltonian is very reminiscent of the classical hamiltonian for a single rotor once the change of variables @xmath394 is made . indeed , the equations of motion obtained here are the same as is @xcite . in this paper , the authors studied the quench from the paramagnetic phase to the ferromagnetic phase and ac dynamics . proceeding as for the bhm , we analyze the phase space of symmetric trajectories . this is different in the ferromagnetic and paramagnetic phases , see figure [ fig : ising ] . in the latter , obtained for @xmath395 , there is a separatrix . for sudden quenches from the ferromagnetic phase , where @xmath396 , to final values of the transverse field @xmath397 , there is a dynamical transition at @xmath398 . at the transition , the trajectory is a separatrix , and @xmath399 decreases exponentially in time to @xmath400 , which is the quantum paramagnetic ground state . the symmetry is dynamically restored at this point but also for larger quenches : contrary to the two previous examples , for quenches beyond the dynamical transition @xmath401 , the trajectories are symmetric in @xmath402 , and the magnetization is oscillating around zero . averaging over time , the order parameter is zero @xmath403 . for @xmath404 and @xmath405 . the dark square is the quantum paramagnetic ground state , the dark circles are the ferromagnetic ground states . the dark dashed line is the separatrix , quenches on this trajectory are at the dynamical transition.,width=529 ] before concluding , two points need to be addressed further : ( 1 ) the possible link between dynamical transitions and equilibrium quantum phase transitions and ( 2 ) the pro and cons of our approach concerning the physics of finite dimensional systems . * in three examples , we found that dynamical transitions occur in systems characterized by a quantum equilibrium phase transition . therefore , despite the fact that the dynamical transition is not directly related to the equilibrium one , as we stressed before , it is natural to wonder whether instead its _ existence _ is related to it . in support of this , if one studies the completely connected bose - hubbard model in a parameter regime ( non unitary fillings ) where there is no _ equilibrium _ quantum phase transition then one finds that there is no dynamical transition either . actually , we found the very same phenomenon in all models we studied , and this was also noticed for the fermionic hubbard model @xcite . we certainly can not address this issue in general , also because it is still far from clear what `` the dynamical transition '' is beyond mean - field theory . we shall instead endeavor to provide a partial answer at the mean - field level . generically , we observe that in all models we studied , there is a regime in which the symmetric state ( the mott insulator , the paramagnet , etc . ) is not the ground state but it plays the role of an excited metastable state that can trap the system for an infinite time . this gives rise to the dynamical transition . if the effective motion is one dimensional , this state is absorbing and belongs to a separatrix . if the effective motion has two or more degrees of freedom , this state seems to belong to an unstable manifold . thus , it appears that the singularity at the dynamical transition is a result of this singular trajectory , which in turn is a consequence of the existence of a quantum phase transition . a physical argument to support this idea is the following . for a quench from the unbroken symmetry phase to the broken symmetry phase in completely connected model , the dynamical evolution is expected to be very slow . the reason is that the symmetry must break up and the order must develop starting from this point . this , however , is a local phenomenon , for example due to domain growth in finite dimensions , and therefore generically absent in completely connected models , except on times diverging with the system size : indeed , in our examples , we find that the typical time scale to depart from the initial value @xmath406 is of the order of @xmath407 . the point is that this slow trajectory is precisely the _ time reversal _ of the singular trajectory that is responsible for the dynamical transition . to summarize , the existence of a quantum phase transition implies the existence of slow trajectories . they correspond to the escape from the symmetry unbroken state . their time reversed counterparts are the trajectories responsible for the dynamical transition . * thus , the existence of a phase transition indeed appears to be intertwined with the existence of a dynamical transition within mean - field . * * let us now discuss the range of applicability of our results for finite - dimensional systems . first , our analysis can be put on the same footing as the gutzwiller ansatz and functional integral saddle point approximations , since the corresponding dynamical equations of motion coincide ( see [ gutzwiller ] for the former equivalence and also @xcite ) . clearly , our mean - field approximation misses several essential physical effects . relaxation and thermalization , for which spatial and temporal fluctuations must be taken into account , are not present . furthermore , the dynamics of a quench from the unbroken symmetry phase to the broken symmetry phase is not properly described for the reasons mentioned above . inhomogeneities , topological defects , domain growth are out of reach . + on the contrary , our mean field approximation is expected to capture well the evolution of local quantities at least on short times , as long as they are homogeneous across the system . contrary to many other approaches , it is _ not perturbative _ in parameters ( @xmath14 , @xmath28 , @xmath72 ) of the hamiltonian . * unlike pertubative approaches , it can describe the existence of dynamical transitions * , and phenomena related to global observables . the general purpose of this article was twofold . first , we described a method to study the quantum quench dynamics of generic completely connected models , providing a mapping to effective classical dynamics . there are two reasons for considering completely connected models . in some cases , such as in the generalized jaynes - cummings model , and the dicke model , they provide the correct physical description . in others , such as the transverse field ising model or the bose - hubbard model , they lead to an approximative treatment of finite dimensional systems . in the latter case , the range of validity of the approximation is possibly limited to short times only . our second aim was to study and to reveal the existence of out of equilibrium dynamical transitions induced by quantum quenches . in agreement with other works @xcite we showed that within the mean - field approximation dynamical transitions occur in quenches from the broken symmetry phase to other regions of the broken symmetry phase and that this is a quite generic phenomenon for systems that display a quantum phase transition at equilibrium . clearly , a main and pressing question is to understand what this transition is really , beyond mean - field theory . does it become a cross - over in finite dimensions ? if it remains a bona fide transition , what are its critical properties ? in the @xmath408 bhm , a non - monotonic behavior of the propagation velocity as a function of @xmath18 has been found in @xcite by t - dmrg . this could well be a signature of a cross - over related to the dynamical transition found in mean - field . an exact diagonalization study of hard core bosons in @xmath409 suggests the existence of a cross - over or a singularity of the revival time after a quantum quench @xcite . from the analytic point of view going beyond mean - field is a difficult task . an argument for the occurrence of dynamical transitions beyond mean field is given in @xcite , within an imaginary time path integral formalism . clearly , it is crucial to take into account fluctuations not captured by mean - field theory . this was started to be done in @xcite . this work indeed suggests that these fluctuations could alter substantially the mean - field results . promising ways to capture fluctuations are the projector operator formalism @xcite , @xmath410 expansions @xcite ( @xmath81 is the connectivity of the lattice ) . field - theoretic methods are also being developed , as in @xcite for the bose - hubbard model in the grand canonical ensemble ; the use of two particle irreducible actions also seems promising @xcite . we thank c. kollath , m. schir , m. eckstein and j. keeling for interesting discussions on these topics . g. biroli acknowledges partial financial support from anr famous . the previous analysis , based on the symmetry between sites , seems to be related to completely connected models only . it is instead much more robust than it looks , since it is actually equivalent to the time dependent gutzwiller ansatz . the latter can be formulated as a mean field approximation for a finite dimensional model : it consists in providing a trial wave function dependent on some physical parameters that are chosen through a maximization procedure . in its equilibrium version , the estimated ground state is simply found minimizing the energy . the dynamical formulation is a bit more elaborated . for clarity , we show here how to build it for the case @xmath15 only . the following is an adapted version of the formulation developed in @xcite . let us define first the `` full state '' @xmath411 , and let @xmath412 be the projector on site @xmath413 on the state @xmath414 , where @xmath415 . the gutzwiller ansatz wave function is a trial wave function with @xmath416 real , time dependent parameters @xmath417 , @xmath418 : @xmath419 note that for simplicity we dropped the site index . the @xmath420 are the projector average : @xmath421 . as we shall see , they will turn out to coincide with the fraction of sites with @xmath32 bosons . obviously , because this is only a trial wave function , it can not satisfy exactly the time evolution : i.e. @xmath422 . however , one can enforce it in an approximate way . this allows one to obtain the time evolution of @xmath417 and @xmath418 . we impose the following constraints : * the projectors in the heisenberg representation @xmath423 satisfy on average the heisenberg equation of evolution @xmath424 $ ] . this amounts to computing the average of @xmath425 in the schrdinger picture @xmath426 using @xmath427| \psi \rangle\\ \end{array}\ ] ] * the energy is conserved @xmath428 . let us call @xmath429 = \langle \psi | \hat{h } | \psi \rangle$ ] the energy as a function of the parameters . the evolution can also be written , using the explicit form of @xmath430 : @xmath431| \psi \rangle = \displaystyle \frac{\partial}{\partial p_l}\langle \psi | \hat{h } | \psi \rangle \\ & = \displaystyle \frac{\partial \mathcal{h}[x_l , p_l]}{\partial p_l } \\ \end{array}\ ] ] this evolution is very reminiscent of the first hamilton equation of motion . furthermore , the conservation of energy @xmath432}{\partial p_l } \dot{p_l } + \frac{\partial \mathcal{h}[x_l , p_l]}{\partial x_l } \dot{x_l}=0 $ ] leads to the second hamilton equation : @xmath433}{\partial x_l}\ ] ] as a consequence , the gutzwiller ansatz wave function is indeed equivalent to an effective classical hamiltonian evolution of the variables @xmath434 and @xmath435 . the effective hamiltonian @xmath429 = \langle \psi | \hat{h } | \psi \rangle [ x_l , p_l]$ ] must be computed for the model at hand , and is indeed the same as the one discussed in section [ general ] . for instance , for the bose - hubbard model with truncation @xmath15 , one recovers . using a wentzel - kramers - brillouin ( wkb ) approximation @xcite , eigenstates of the schrdinger equation can be found in the limit of large @xmath13 . to do so , we write the eigenstates @xmath436 within the wkb approximation . the derivation is standard , one looks for stationary solutions of the form @xmath437 of the equation , at dominant order in @xmath13 . one finds @xmath438 where @xmath249 is a normalization constant , @xmath244 is the energy , @xmath439 satisfies the implicit equation @xmath440 , and @xmath91 is the left turning point of the classical trajectory of energy @xmath244 . using this expression , we can draw a parallel between the eigenstate @xmath436 and the effective classical trajectories of energy @xmath244 . for an observable @xmath441 , the average reads @xmath442 in the second line , we refer to the effective classical motion @xmath443 , and the average @xmath444 is the average over the effective classical trajectory of @xmath445 over one peridod . in words , we found that in the semi classical regime , the average over a quantum stationary state of energy @xmath244 is given by the average over one period of the classical trajectory of same energy @xmath244 . although this result is not often mentioned , this is a direct consequence of the wkb expression of the wave function .

berry connection is conventionally defined as a static gauge field in the brillouin zone . here we show that for three - dimensional ( 3d ) time - reversal invariant superconductors , a generalized berry gauge field behaves as a dynamical fluctuating field of a chern - simons gauge theory . the gapless nodal lines in the momentum space play the role of wilson loop observables , while their linking and knot invariants modify the gravitational theta angle . this angle induces a topological gravitomagnetoelectric effect where a temperature gradient induces a rotational energy flow . we also show how topological strings may be realized in the 6 dimensional phase space , where the physical space defects play the role of topological d - branes .

1610.05810

topology has played an increasingly important role in condensed matter physics . an early application involved using topological chern - simons theory in quantum hall systems @xcite . more recently , topological ideas applied to brillouin zone ( bz ) have played a significant role in classifying topological phases of matter ( for example , see @xcite and references therein ) . whereas the former application involved a dynamical gauge field in physical spacetime , the latter cases were formulated in the momentum space and involved topological aspects of static gauge fields such as the berry connection . it is thus natural to ask : can dynamical gauge fields naturally occur in a condensed matter system ? the aim of this paper is to argue that this can be done at least in the context of three - dimensional ( 3d ) time reversal invariant ( tri ) superconductors , where ( a slightly modified version of ) berry s gauge field can be viewed as dynamical in the brillouin zone and where the action of the theory is the chern - simons action . in this class of theories , the fluctuations of the berry connection are induced from quantum fluctuations of the superconductor s pairing amplitude . more interestingly , 3d tri superconductors are known to be capable of having gapless nodal lines in the bz @xcite , and we will show that they play exactly the role of wilson loop observables in the chern - simons theory . in nature , a large class of materials belongs to 3d tri superconductors , which includes most of the conventional s - wave superconductors described by the bardeen - cooper - schrieffer theory , and a large number of unconventional superconductors such as many heavy fermion superconductors , cuprates , iron - based superconductors and the 3d topological superconductors . in particular , unconventional superconductors often exhibit a highly anisotropic pairing amplitude such as p - wave , d - wave or their hybridizations with s - wave , and have much richer phenomena in experiments @xcite . previous studies have shown from different perspectives the significance of topology in these tri unconventional superconductors . gapped 3d tri superconductors are shown to have a @xmath0 topological classification @xcite , which fit in the generic k - theory classification framework of gapped topological phases @xcite . topological superconductors are defined as such superconductors with a nonzero topological number ( e.g. the @xmath1he - b phase ) , and are shown to support gapless topological majorana fermions on the surfaces @xcite . on the other hand , the gapless 3d tri superconductors contain gapless nodal lines in the bz that are allowed by the time - reversal symmetry , with known examples such as the heavy fermion superconductor cept@xmath2si and the cuprates @xcite . several recent studies show that topological numbers can also be defined for these nodal - line superconductors in terms of k - theory @xcite , which give rise to various types of topological majorana surface states @xcite . in addition , in analogy to 3d weyl semimetals which can be viewed as intermediate phases between 3d topological insulators and trivial insulators @xcite , nodal - line superconductors can also arise as gapless intermediate phases between conventional tri and topological tri superconductor phases @xcite . these different while related facts strongly indicate the existence of a unified topological field theory that describes both gapped and gapless 3d tri topological superconductors . this also motivates us to consider the dynamical chern - simons theory in the 3d bz for these superconductors . there is already a hint that a topological field theory in bz can be physically relevant . in particular , it has been shown @xcite that for an insulator in odd spatial dimensions , the value of the chern - simons ( cs ) action for berry connection of the filled bands in the bz computes the effective theta angle of the corresponding @xmath3 term ( or a gravitational analog proportional to @xmath4 in the 3d case @xcite ) , where @xmath5 is the electromagnetic field in the physical spacetime . in other words , there is a coupling of the form @xmath6\times \left[\int_{\mathbb r^{2n } } f^n\right]\ .\]]we will be specializing to the case of @xmath7 , i.e. , 3d space in this paper . ( a similar term can be used to compute the gravitational response involving @xmath8 ) . note that the cs action , which is an angular quantity and has a shift ambiguity , has the correct structure to be the coefficient in front of @xmath9 , which needs to be defined only up to shift symmetry . in this context , it is very natural to ask whether the berry connection @xmath10 can fluctuate . in particular , we can imagine having in physical spacetime a pulse where @xmath11 , which leads to an effective level of @xmath12 for the cs theory in the bz . can the berry connection behave as if it were dynamical ? for the answer to be yes , the classical background for @xmath10 must be flat , as is demanded by the cs equations of motion . this is certainly not the case in general . however , as we shall see , it is the case for tri superconductors . with this encouraging result , one may then ask whether there are natural objects in the bz corresponding to wilson loops of the cs theory . indeed , for tri superconductors the symmetries allow gapless nodal lines in the bz , which we will see end up playing the role of wilson lines for the cs theory . for superconductors , the relevant term to compute is the gravitational @xmath13 term . but with the gapless modes , as would be the case if we have nodal lines , the theta angle is ambiguous . it turns out the choices of resolving this ambiguity by introducing infinitesimal time reversal breaking perturbations to get rid of gapless modes are in 1 - 1 correspondence with allowed basic charges for the cs theory ! we thus find that dressing up the nodal lines with this data gives an unambiguous theta angle for the coefficient of @xmath4 term , which is identified with the free energy of cs theory in the presence of wilson loops and leads to physically measurable effects . in particular , as we shall see , when the nodal lines change from linked to unlinked , the theta angle changes in units of @xmath14 . our main discussion is in the context of the abelian chern - simons theory , but we also indicate briefly how the theory gets extended to non - abelian case , and how in particular the @xmath15 case can be potentially realized in experiments . we will also connect aspects of our discussions with topological strings @xcite which is formulated in a 6-dimensional symplectic manifold ( which is typically taken to be a calabi - yau 3-fold ) . in our context , the phase space which is @xmath16 turns out to play the role of this 6d symplectic space . in the topological string setup , cs theory lives on 3d lagrangian defects ( ` d - branes ' ) . if they are oriented along the bz @xmath17 , they give rise to the cs theory we find in the bz . in this context , wilson loop observables arise from a pair of such 3d defects intersecting along a loop @xcite . we will show that line defects in physical space lead to such 3d lagrangian branes and can also give rise to wilson loop observables ( nodal lines ) in the bz . the organization of this paper is as follows . in section ii , we review the hamiltonian of tri superconductors and define the modified berry connection . in section iii , we relate the topological @xmath18 angle to a chern - simons term . section iv incorporates the nodal lines as wilson loops , first in minimal model and then in multi - band system . in section v , we consider non - abelian nodal lines , first in a @xmath15 example and then generalize to the @xmath19 example . in section vi we discuss connections with topological strings and show how line defects in physical space also lead to effective wilson loops in the bz . we present our conclusions in section vii , and in the appendix are some computational details . at the single - particle level , superconductors are described by the bogoliubov - de gennes ( bdg ) hamiltonian which has an inherent charge conjugation symmetry @xmath20 . for a superconductor with @xmath21 electron bands , the bdg hamiltonian can be written in terms of the nambu basis @xmath22 as @xmath23 , where @xmath24 is a @xmath25 matrix , @xmath26 is the momentum , and @xmath27 is the @xmath21-component electron basis of the system . both @xmath28 and @xmath29 are @xmath30 matrices . @xmath28 represents the single - particle hamiltonian of the system before superconductivity arises , while @xmath29 is the pairing amplitude satisfying @xmath31 as required by the fermion statistics . the charge conjugation is defined as @xmath32 , or equivalently @xmath33 , where @xmath34 is a @xmath25 matrix , with @xmath35 denoting the pauli matrices ( for particle - hole basis ) and @xmath36 the @xmath30 identity matrix . the charge conjugation symmetry of the bdg hamiltonian can be seen via the relation @xmath37 . tri superconductors are a large class of superconductors that are quite robust in nature . in the presence of time reversal symmetry @xmath38 , electrons which are time - reversal kramers pairs have the maximal spatial wave - function overlap , thus fall into cooper pairs most easily @xcite . for fermions , the time reversal symmetry @xmath38 is anti - unitary and satisfies @xmath39 . it acts on the fermions as @xcite @xmath40 , where @xmath41 is an @xmath30 matrix satisfying @xmath42 , @xmath43 . the bdg hamiltonian of a tri superconductor can then be shown to satisfy @xmath44 and @xmath45 , namely , @xmath46 is hermitian . these conditions can be more compactly written as @xmath47 , where @xmath48 is the @xmath25 time - reversal transformation matrix of the nambu basis . it is useful to define the chiral transformation @xmath49 , which is a unitary hermitian matrix that anti - commutes with the bdg hamiltonian , @xmath50 . upon diagonalizing @xmath51 to @xmath52 under a new basis @xmath53 , the bdg hamiltonian becomes @xmath54 where @xmath55 . since both @xmath28 and @xmath46 are hermitian , @xmath56 is a general @xmath30 complex matrix . if @xmath56 is non - singular everywhere , the superconductor is in a fully gapped phase . on the other hand , the superconductor becomes gapless ( nodal ) at momentum @xmath26 if @xmath57 . in the absence of additional symmetries other than the time reversal symmetry , the gapless submanifolds in the momentum space are nodal points for two - dimensional ( 2d ) superconductors , and are 1d nodal lines for 3d superconductors @xcite . when the superconductor is centrosymmetric , the nodal lines ( points ) in 3d ( 2d ) become doubly degenerate , as we shall show in sec . [ secu2 ] . in this paper we focus on 3d superconductors with nodal lines , which are widely found in noncentrosymmetric superconductors @xcite and quasi-2d centrosymmetric superconductors such as cuprates and iron - based superconductors @xcite . in general , the matrix @xmath56 can be singular - value - decomposed ( svd ) into @xmath58 , where @xmath59 , @xmath60 are unitary matrices , and @xmath61 is a diagonal matrix with all elements real and nonnegative . the bdg hamiltonian can then be diagonalized as @xmath62 in literature @xcite , it is conventional to define a @xmath63-form u(@xmath21 ) berry connection for the @xmath21 occupied bands with negative energy as @xmath64 , where @xmath65 is the exterior derivative in the momentum space , and @xmath66 is the eigenstate wave function of the @xmath67-th occupied band at momentum @xmath26 . this can be rewritten as @xmath68 where we have used @xmath69 , @xmath70 short for @xmath59 , @xmath60 . however , such a berry connection @xmath71 can not be regarded as a u(@xmath21 ) gauge field for @xmath72 , since the eigenstates @xmath66 are fixed by the hamiltonian and the gauge freedom of @xmath71 is u(@xmath63)@xmath73 instead of u(@xmath21 ) . ( see appendix [ appberry ] ) . in this paper , we define a modified u(@xmath21 ) berry connection as @xmath74 , where @xmath75 is the vacuum state with no electrons , and @xmath67 runs over all the @xmath76 eigenstates . in the tri case here , such a berry connection can be rewritten as follows : @xmath77 physically , @xmath78 is the berry connection felt by an electron when its wave function is projected onto the bdg eigenstates and evolves adiabatically . for superconductors , this definition of @xmath79 is closely related to the conventional berry connection @xmath80 , as we shall see in sec . [ seccsa ] . in particular , in the case @xmath81 , we have @xmath82 , and the two definitions are identical up to a sign . in the generic @xmath21 band case , even though the eigenstates are fixed by the hamiltonian , the u(@xmath21 ) gauge transformation for @xmath79 can be realized via a unitary transformation of the basis @xmath83 ( instead of transformation of eigenstates ) , where @xmath84 is a u(@xmath21 ) matrix . by the definition given above , the berry connection @xmath79 transforms as @xmath85 , which is exactly a u(@xmath21 ) gauge transformation . this is however not the case for @xmath80 ( see appendix [ appberry ] ) . in this sense , the modified berry connection @xmath79 is more naturally defined than the conventional @xmath80 as a u(@xmath21 ) gauge field . accordingly , a slow external potential @xmath86 added to the system will induce an effective hamiltonian perturbation @xmath87 in the momentum space . the topological nature of most weakly interacting condensed matter systems is reflected in the berry connections @xcite . in particular , a 3d gapped superconductor is shown to be described by an effective gravitational topological action @xcite @xmath88 where @xmath89 is the riemann tensor of the @xmath90d spacetime , and @xmath18 is the topological theta angle determined by the superconductor up to multiples of @xmath91 . for spacetimes which are spin manifolds , the topological action is quantized to @xmath92 with @xmath93 being the gravitational instanton number @xcite . in the weak field limit , the einstein gravitation can be reformulated in the gravitoelectromagnetism framework @xcite , and the above action can be rewritten as @xcite @xmath94 where @xmath95 is the effective coupling constant , while @xmath96 and @xmath97 are the gravitoelectric field and the gravitomagnetic field , respectively . when a 3d gapped superconductor is tri , @xmath98 is required and the theta angle @xmath18 must take @xmath99 or @xmath14 . in general , a tri topological superconductor characterized by topological number @xmath100 will have a theta angle @xmath101 @xcite . in particular , a superconductor in the @xmath1he - b phase has @xmath102 . this form of gravitational topological action is in direct analogy with the topological electromagnetic action for 3d topological insulators and axion insulators @xcite . these gravitational fields are closely related to the thermal transport of the superconductor @xcite . the gravitoelectric field @xmath96 can be produced by a gradient of temperature @xmath103 as @xmath104 , while the gravitomagnetic field @xmath97 characterizes the rotational energy flow and is proportional to the effective angular velocity of the matter in the system . as a result , the superconductor s response to gravito - magneto - electric effects is affected by this @xmath18 angle . in the band theories of gapped systems , the topological theta angle is given by the momentum - space chern - simons action of the non - abelian berry connection as follows @xcite : @xmath105=\frac{1}{4\pi}\int\text{tr}\left(a'\wedge da'-i\frac{2}{3}a'\wedge a'\wedge a'\right),\ ] ] where @xmath80 is as defined in eq . ( [ aprime ] ) for superconductors and is defined similarly for occupied bands of insulators , and the integration is in the whole brillouin zone . when the gravitational instanton number @xmath106 , the total action @xmath107 will become a chern - simons action at level @xmath12 however , @xmath108 $ ] can not be regarded as defining a non - abelian chern - simons theory . this is because the berry connection @xmath80 has only u(@xmath63)@xmath73 instead of u(@xmath21 ) gauge freedom , and is generically non - flat for @xmath72 : namely , the non - abelian berry curvature @xmath109 for generic band structures , which is contrary to the classical equation of motion @xmath110 of the chern - simons theory . to obtain @xmath18 for gapped tri superconductors from a well - defined momentum - space chern - simons theory , we consider the modified berry connection @xmath79 , and define @xmath111=-\frac{1}{4\pi}\int\text{tr}\left(a\wedge da - i\frac{2}{3}a\wedge a\wedge a\right).\ ] ] for 3d gapped tri superconductors , it can be shown that @xmath112=-\mathcal{a}_{cs}[a']=n\pi$ ] with @xmath113 an integer , and the two definitions of @xmath18 coincide ( see appendix [ appberry ] ) . formula ( [ csa ] ) has the advantage that @xmath79 is a well - defined gauge field with u(@xmath21 ) gauge freedom as we have shown . more importantly , the equation of motion of chern - simons theory , i.e. , the flatness of @xmath79 , is satisfied by a natural physical condition which is generically true in mean - field theories of tri superconductors : two electrons will form a cooper pair only when they are time - reversal partners and have the same kinetic energy . this condition is equivalent to stating that the two hermitian matrices @xmath28 and @xmath46 commute with each other . as a result , they can be simultaneously diagonalized via a unitary transformation as @xmath114 and @xmath115 . therefore , @xmath59 and @xmath60 in the svd of @xmath56 are related by @xmath116 , where @xmath117 is a diagonal unitary matrix , and @xmath118 is the complex phase of @xmath119 @xmath120 . following eq . ( [ a ] ) , one can readily show the berry connection is in the following form : @xmath121 which is just a gauge transformation of @xmath122 . it is easy to see that @xmath123 as a diagonal matrix of exact @xmath63-forms is flat , namely , @xmath124 . therefore , we reach the conclusion that @xmath79 is also flat , namely , the non - abelian field strength ( modified berry curvature ) @xmath125 at the mean field level . unlike insulators where the berry connection @xmath80 is fixed , superconductors always have quantum fluctuations of the pairing amplitude @xmath29 around the mean - field solution , which lead to fluctuations in both @xmath80 and @xmath79 . the pairing fluctuations @xmath126 generically need not respect the time - reversal symmetry or preserve the commutation of @xmath28 and @xmath46 , therefore leading to non - flat fluctuations in the berry connection @xmath79 . to the lowest order , the fluctuation of @xmath79 is linear to @xmath126 . explicitly , for the simplest @xmath81 case where @xmath79 is a u(1 ) gauge field in 3d , the fluctuation of the phase of @xmath29 breaks the time - reversal symmetry and gives rise to a @xmath127 term in the @xmath128 hamiltonian @xmath129 , which makes the berry connection @xmath79 generically non - flat . in calculating the partition function of the tri superconductors , all the quantum fluctuations in @xmath29 around the mean - field solution should be included in the path integral . correspondingly , any fluctuations in @xmath79 induced by the pairing fluctuation should be taken into account in evaluating the @xmath18 angle . these fluctuation in @xmath79 can then be naturally regarded as gaussian fluctuations in the vicinity of the flat connection which is a classical solution . since the partition function of the gravitational topological action @xmath107 generically contains contributions from topological sectors of all gravitational instanton numbers @xmath93 , the action @xmath92 in each topological sector should be regarded as a physical action . for this reason , we shall view the topological @xmath18 angle as a functional @xmath130 $ ] of the gauge field @xmath79 in the rest of the paper , i.e. , as the free energy of the chern - simons theory . more specifically , we will focus on the case where the gravitational instanton pulse has integer unit @xmath131 , which leads to level 1 chern - simons theory using eq . ( [ grav ] ) , and @xmath18 will be identified with the free energy of this level 1 cs theory . when a tri superconductor has nodal lines , the system becomes gapless and dissipative , and the gravitational topological action @xmath107 does not seem to have a clear physical meaning . on the other hand , nodal lines look like they can behave as wilson line observables of the chern - simons theory . but what picks the charges of these wilson lines ? it turns out that these two problems cancel out , as we will show below : to make @xmath107 well defined , we need to gap the system , which can be done by choosing small time - reversal breaking perturbations . the choices for these perturbation turn out to pick the charges that the wilson lines carry . in other words , we will argue that fixing the charges on the wilson line , which leads to well - defined amplitudes for the cs theory , _ is equivalent _ to choosing small perturbations which gap the system , leading to well defined @xmath18 angle , which will be identified with the free energy of the cs theory in the presence of wilson loop observables . in this section , we will focus on tri superconductors without additional symmetries , such as the noncentrosymmetric superconductors , while tri nodal - line superconductors with inversion symmetry will be discussed in the next section . in a tri superconductor with no other symmetry , the bdg band structure is in general nondegenerate except for some zero - measure submanifolds which generically consists of nodal lines . therefore , each nodal line is associated with a definite pair of bands which are related by the chiral transformation @xmath51 . it is helpful to consider the minimal toy model with @xmath81 before dealing with the general case , though it is not quite realistic as the spin degrees of freedom ( or equivalently , @xmath39 ) requires the total number of bands @xmath21 to be even . the bdg hamiltonian is then a @xmath128 matrix @xmath132 where the matrix @xmath56 from eq . ( [ jk ] ) reduces to a complex number @xmath133 , with @xmath134 being its complex phase . accordingly , in the svd @xmath135 , one can choose @xmath136 and @xmath137 where @xmath138 is an arbitrary real function corresponding to the u(@xmath63 ) gauge freedom , and the berry connection @xmath79 reduces to a u(@xmath63 ) abelian gauge field @xmath139 the nodal lines are characterized by the equation @xmath140 , which are closed loops in the momentum space . therefore , the berry connection @xmath79 becomes ill - defined on each nodal line @xmath141 . as one winds around a nodal line @xmath141 once , the phase @xmath134 will change by @xmath91 as long as @xmath142 is a generic continuous function . besides , the gauge function @xmath138 must be chosen to change by @xmath143 ( @xmath144 ) per winding around the nodal line @xmath141 , so that @xmath59 and @xmath60 are single - valued . the gauge field @xmath79 then satisfies @xmath145 where @xmath146 is a loop that winds around the nodal line @xmath141 once , and @xmath113 is an arbitrary integer . therefore , a nodal line can be viewed as a vortex line in the momentum space that carries a @xmath147 berry flux @xmath148 . such a u(@xmath63 ) berry connection can be derived from the equation of motion of the chern - simons theory by regarding the nodal lines as wilson loops . by modifying the topological theta angle to @xmath149=-\frac{1}{4\pi}\int a\wedge da+\sum_bq_b\oint_{\mathcal{l}_b}a\ ] ] and regarding it as a physical action , one finds the equation of motion @xmath150 where @xmath151 stands for the delta function at the nodal line @xmath141 in the two perpendicular directions @xmath152 and @xmath153 of @xmath141 . the flux formula in eq . ( [ flux ] ) is then satisfied by assigning each wilson loop a half - integral charge @xmath154 . to better understand such a modified theta angle , we can consider a weak time reversal symmetry breaking in the system . more explicitly , we introduce a small tri breaking mass term @xmath155 to the system , so that the hamiltonian becomes @xmath156 and a nodal line will be lifted into a gap of size @xmath157 ( @xmath155 is real ) . this mass term may emerge effectively from magnetic disorders , or could be induced via uniform tuning of the parameters of the system ( e.g. , the phase of @xmath29 ) . if @xmath155 is nonzero on all nodal lines @xcite , the superconductor becomes fully gapped , and the berry connections @xmath79 and @xmath158 become well - defined everywhere . in particular , in the limit @xmath159 , the berry phase along a loop @xmath146 that winds around a nodal line @xmath141 once becomes definite : @xmath160 where @xmath161 denotes the sign of @xmath155 . this is analogous to the berry phase around a dirac point in the 2d graphene , which falls into @xmath162 depending on the sign of the dirac mass added . accordingly , the wilson loop charge becomes @xmath163 and is no longer indefinite , either . in principle , different nodal lines can have different signs of mass , and therefore different wilson loop charges . therefore , the ambiguity of wilson loop charges @xmath154 disappears once the time reversal symmetry of the superconductor is explicitly weakly broken , and are restricted to @xmath164 . let us clarify the meaning of half - integral @xmath165 charge . recall that we are dealing with a @xmath165 chern - simons theory at level @xmath63 . the hilbert space of this theory consists of conformal blocks of the 2d cft associated with a complex free fermion.@xcite this theory also admits an extra field , the spin field , which creates branch cuts for the fermion correlations . using bosonization , we can view the fermion as @xmath166 and the spin field as @xmath167 . the insertion of the @xmath168 fields corresponds to charge @xmath169 states in the @xmath165 chern - simons theory at level 1 . here we have restricted to gravitational instanton number @xmath131 . for a more general @xmath12 , the action is @xmath170 $ ] , which we argue emerges from cs theory at level @xmath12 coupled with wilson loops : for eq . ( [ mono ] ) to still hold , we also need to rescale the charges on the wilson loops so that the charge carried by the wilson loops are now @xmath171 . the gravitational topological action @xmath107 with coefficient @xmath130 $ ] defined in eq . ( [ u1theta ] ) also acquires a physical meaning under such a weak time reversal symmetry breaking . since the mass term @xmath155 gaps out the superconductor , @xmath107 is now well - defined as discussed in sec . [ seccsa ] , and one has to go back to eq . ( [ theta ] ) for the definition of @xmath18 in the absence of time reversal symmetry . on the other hand , making use of the equation of motion in eq . ( [ eom ] ) , one can show @xmath172 and thus the classical value of @xmath130 $ ] defined in eq . ( [ u1theta ] ) is @xmath173=\frac{1}{4\pi}\int a\wedge da=\frac{1}{4\pi}\int a'\wedge da'=\mathcal{a}_{cs}(a')\ , \ ] ] which agrees exactly with the definition of @xmath18 in eq . ( [ theta ] ) , and we have used the fact that @xmath82 for @xmath81 . this verifies the validity of the topological theta angle @xmath130 $ ] defined in the presence of nodal lines . as a topological quantum field theory , the chern - simons theory has an intrinsic connection with the knot invariants as first revealed by witten @xcite , which plays an important role in the description of @xmath174d topological many - body states such as the quantum hall states , and quasiparticle statistics therein @xcite . indeed , we will now argue that the partition function of the theory as measured by the @xmath18-angle , in the presence of nodal lines , behaves exactly as expected for a quantum chern - simons theory with the corresponding wilson loops . in the u(@xmath63 ) case at level @xmath63 , the partition function of cs theory with wilson loops is given by @xcite @xmath175=\pi\sum_{b , c}q_bq_c\phi(\mathcal{l}_b,\mathcal{l}_c)\ , \ ] ] where @xmath176 is the gauss linking number between nodal lines @xmath141 and @xmath177 . in the case of u(@xmath63 ) chern - simons theory , this is also the classical valuation of the action at the critical point ( because the action is quadratic ) . since the wilson loop charges @xmath154 are half integral , changing the linking number between two nodal lines by @xmath63 yields a @xmath178 shift in @xmath130 $ ] . besides , the calculation of eq . ( [ thetalinking ] ) needs a proper regularization of the self - linking number for @xmath179 , which is related to the framing of the wilson loop . change in the @xmath130 $ ] angle . ( b ) this shows how a nodal line @xmath141 can cross itself , after which its self - linking number changes by @xmath180 . ( c ) the generic skein relation for knot polynomials , which gives the rules for the local surgery of a knot configuration.,width=316 ] now we would like to derive this same result using the definition of the @xmath18-angle for our system . we will show that as we change the linking number between nodal lines , the @xmath18-angle changes exactly as expected for @xmath165 cs theory at level 1 . consider the local crossing of two nodal lines @xmath141 and @xmath177 as shown in fig . [ fig1](a ) . @xmath141 and @xmath177 can be either different nodal lines or different parts of the same nodal line . the two configurations in fig . [ fig1](a ) can be adiabatically connected by a parameter @xmath181 $ ] , where @xmath182 and @xmath183 corresponds to the left and right configurations , respectively . for example , this can be realized by assuming the off - diagonal element in eq . ( [ hu1 ] ) to be @xmath184 locally , where @xmath141 and @xmath177 are at @xmath185 and @xmath186 , respectively . accordingly , the field strength in the 4d parameter space @xmath187 is @xmath188 $ ] . besides , the wilson loop integrals @xmath189 and @xmath190 change by @xmath191 and @xmath192 , respectively . the resulting total variation in the expectation value of @xmath130 $ ] is thus @xmath193\big|_{t=-1}^{t=1}=-\int_{-1}^{1}d\mathcal{a}_{cs}[a]+2\pi q_bq_c+2\pi q_bq_c\\ & = -\frac{1}{4\pi}\int_{4d}\text{tr}(f\wedge f)+4\pi q_bq_c=-2\pi q_bq_c+4\pi q_bq_c\\ & = 2\pi q_bq_c=\pm\frac{\pi}{2}\ \mod 2\pi\ , \end{split}\ ] ] where we have used the fact that @xmath194 are @xmath169 . if we define @xmath195=e^{i\theta_{cl}[a]}=\langle e^{i\theta[a]}\rangle$ ] as a knot invariant , the above result yields a simple skein relation @xmath196-e^{i2\pi q_bq_c}w_2[a]=0\ , \ ] ] where @xmath197 $ ] and @xmath198 $ ] correspond to two configurations which are only different locally as shown in fig . [ fig1](a ) . this is a special case of the general skein relation as shown in fig . [ fig1](c ) , with coefficients @xmath199 , @xmath200 and @xmath201 @xcite . it is convenient to assign each nodal line @xmath141 a direction along which the berry flux is @xmath202 , i.e. , @xmath203 , so that skein relation is unambiguous with @xmath204 . in particular , when a nodal line @xmath141 crosses a distinct nodal line @xmath177 the @xmath18-angle changes the contribution to @xmath18-angle changes by @xmath178 exactly as is expected from eq . ( [ thetalinking ] ) . this is easy to explain : both computations can be reduced to finding classical solutions of the cs theory with a given nodal line and evaluating the action at the classical value . however , we also need to discuss the nodal line crossing itself as shown in fig . [ fig1](b ) . here $ ] should also change by @xmath178 due to the skein relation , which indicates the ` self - linking number ' @xmath206 should change by @xmath180 . the self - linking number is also known as the framing of the knot @xcite , which we have not yet discussed in the context of the nodal lines . as we will argue below , due to the doubling of the nodal lines due to time reversal invariance , such self - linking , or framing contributions , cancel pairwise as a nodal line crosses itself . in other words , we can delete the @xmath179 terms in the sum in eq . ( [ thetalinking ] ) as far as its contribution to the total @xmath18-angle is concerned . @xcite in practice , in tri superconductors , nodal lines usually occur in pairs at momenta related by the time - reversal symmetry . therefore , the nodal line crossings are always doubled , and the change in @xmath130 $ ] will be doubled to @xmath99 or @xmath14 , depending on the signs of masses of each nodal line in the weak breaking of time - reversal symmetry . for instance , if one assign @xmath141 and its time - reversal partner the same mass , but assigns @xmath177 and its time - reversal partner opposite masses , the crossing of @xmath141 and @xmath177 ( and their time - reversal ) will induce a @xmath14 shift in @xmath130 $ ] . however , if a nodal line @xmath141 crosses itself once and recovers the original configuration , due to the time - reversal doubling , time - reversal related nodal lines always have opposite self - crossing phases , and the total change in @xmath205 $ ] is always zero . this justifies why we may neglect the contribution of self - linking ( or framing number ) in finding the contribution of linked nodal lines to @xmath18 . now we proceed to consider nodal lines in general @xmath21-band tri superconductors . following the discussion above eq . ( [ flata ] ) , we shall assume @xmath28 and @xmath46 are simultaneously diagonalized by @xmath207 , where @xmath208 is the @xmath67-th normalized eigenvector of @xmath28 ( @xmath209 ) , or the wave function of the @xmath67-th electron band before superconducting . in the absence of additional symmetries , each nodal line @xmath141 is associated with a definite electron band . the modified topological theta angle can then be naturally generalized as @xmath210=-\mathcal{a}_{cs}[a]+\sum_bq_b\oint_{\mathcal{l}_b}u_{\alpha_b}^\dag au_{\alpha_b}\ , \ ] ] where @xmath79 is now a u(@xmath21 ) gauge field , @xmath112 $ ] is the non - abelian chern - simons action , and @xmath211 is the @xmath21-component wave function of the electron band @xmath212 associated with nodal line ( or wilson loop ) @xmath141 . similarly , the wilson loop charges @xmath154 are half - integral . the wilson loops @xmath141 are coupled to the projected u(@xmath63 ) berry connection @xmath213 , and thus breaks the u(@xmath21 ) gauge symmetry . if the 3d momentum space is viewed as a @xmath214d `` spacetime '' , the wilson loops can be interpreted as world lines of particles , while @xmath211 are the u(@xmath21 ) charges carried by the particles . due to such a coupling , the chern - simons theory governed by @xmath130 $ ] is enriched with more structures . the equation of motion then becomes @xmath215 where @xmath216 is now the u(@xmath21 ) berry curvature . when two nodal lines @xmath141 and @xmath177 cross each other as shown in fig . [ fig1](a ) , a calculation similar to eq . ( [ calcrossing ] ) yields the jump in theta angle @xmath217=2\pi q_bq_c|u_{\alpha_b}^\dag u_{\alpha_c}|^2=2\pi q_bq_c\delta_{\alpha_b\alpha_c}\ , \ ] ] provided the eigenvectors @xmath211 and @xmath218 are nonsingular during the crossing . therefore , the crossing produces a change in @xmath205 $ ] only if the two nodal lines are in the same band . in metals which become superconductors at low temperatures , nodal lines always live on the 2d fermi surfaces of the metals . one may therefore wonder what kind of fermi surface can give rise to linked nodal lines . fig . [ fig1](a ) shows how two nodal lines of linking number @xmath63 can be drawn on a @xmath180-torus fermi surface . we note that they are not time - reversal invariant by themselves . therefore , their time - reversal partners should coexist on another @xmath180-torus fermi surface in the bz . to unlink the two nodal lines @xmath219 and @xmath220 , one must imagine a process during which the poloidal radius of a certain part of the torus shrinks to zero and then expands back . the two nodal lines can then cross each other at the zero poloidal radius point . and annihilated at point @xmath221 , whose trajectory is a closed surface @xmath222 . ( c)-(d ) the sign of pairing @xmath223 on a fermi surface of a gapped tri superconductor can be reversed by creating nodal lines and making them sweep around the fermi surface , which leads to a phase transition between topological superconductors if the fermi surface chern number @xmath224 is nonzero.,width=316 ] the theta angle @xmath130 $ ] defined in eq . ( [ untheta ] ) also explains in a different way the fact that @xmath101 for a tri gapped topological superconductor with topological number @xmath225 . this fact comes from the following observation : consider the process in which a nodal line @xmath141 is created at point @xmath226 and then annihilated at point @xmath221 as shown in fig . [ fig2](b ) as a function of adiabatic parameter @xmath181 $ ] , during which the system remains tri . the trajectory of @xmath141 is thus a closed 2d manifold @xmath222 in the 3d momentum space . accordingly , the change in @xmath205 $ ] is @xmath227\big|_{t=-1}^{t=1}=-\int_{-1}^{1}d\mathcal{a}_{cs}[a]+q_b\int_{-1}^1d\oint_{\mathcal{l}_b}a^{(\alpha_b)}\\ & = -\frac{1}{4\pi}\int_{4d}\text{tr}(f\wedge f)+q_b\oint_\mathcal{m}da^{(\alpha_b)}\\ & = 0+q_b\oint_\mathcal{m}f^{(\alpha_b)}=2\pi q_b c_1(\alpha_b)\ , \end{split}\ ] ] where we have used the fact that @xmath228 when there is no crossing of nodal lines , while @xmath229 is the projected u(@xmath63 ) berry curvature of electron band @xmath212 , and @xmath230 is the first chern number of band @xmath212 on the 2d manifold @xmath222 . we note that even if the u(@xmath21 ) berry curvature @xmath231 , the projected u(@xmath63 ) berry curvature @xmath232 and thus @xmath230 can be generically nonzero . therefore , @xmath130 $ ] of a tri gapped superconductor can be changed by creation and annihilation of nodal lines . physically , this can be done by adiabatically changing the pairing amplitude @xmath29 of the bdg hamiltonian . we shall still assume @xmath114 and @xmath115 , so a nodal line in band @xmath67 is given by @xmath233 . we note that @xmath234 defines the fermi surface @xmath235 of electron band @xmath67 , while @xmath236 is a real function of @xmath26 . when the superconductor is gapped , @xmath236 is nonzero everywhere on the fermi surface @xmath235 , and thus has a definite sign @xmath237 there . qi , hughes and zhang @xcite has shown that the topological number @xmath225 of a gapped tri superconductor is given by @xmath238 where @xmath239 is the fermi - surface first chern number of electron band @xmath67 . now consider a fermi surface @xmath235 with @xmath240 and @xmath241 as shown in fig . [ fig2](c ) at the beginning . one can then create two nodal lines @xmath219 and @xmath220 ( which are time - reversal partners ) from the north and south poles of the fermi surface as shown in fig . [ fig2](d ) , and make them annihilate on the equator . since @xmath223 on @xmath235 changes sign across a nodal line , the sign @xmath242 is reversed at the end of the process , and @xmath225 decreases by @xmath63 . on the other hand , eq . ( [ c1 ] ) tells us that the change in the theta angle @xmath18 is @xmath243 , which therefore agrees with the fact @xmath244 for tri topological superconductors . this also verifies that the wilson loop charges @xmath154 have to be half - integral . this also indicates that @xmath130 $ ] for @xmath72 could acquire a geometric contribution solely from the deformation of a nodal line , without involving any nodal - line crossings . this is different from the u(@xmath63 ) chern - simons theory discussed in sec . [ secu1a ] which is purely topological . such an additional phase change is due to the berry curvature of the wave functions @xmath245 . in practice , @xmath39 always requires the total number of bands @xmath21 to be even , so such a geometric contribution is in general present and nonzero . so far , we have discussed tri superconductors with nondegenerate bands , where nodal lines behave as wilson loops coupled to projected u(@xmath63 ) berry connections . to construct wilson loops with a non - abelian coupling to the berry connections , one needs the band structure of the superconductor to have an @xmath21-fold degeneracy @xcite . theoretically , we can discuss all @xmath72 cases , but it turns out that the @xmath246 case is particularly well motivated in certain superconductors : in particular , adding inversion symmetry to the superconductor yields a @xmath180-fold degeneracy and leads to u(@xmath180 ) nodal lines , as we shall show below . in subsection a , we discuss the @xmath15 case and in the subsection b , we turn to the more general @xmath19 case . many tri superconductors are also centrosymmetric , namely , have an inversion symmetry @xmath247 . in this case , the combined symmetry @xmath248 is an anti - unitary symmetry with @xmath249 . since the @xmath248 symmetry keeps the momentum @xmath26 invariant , such a symmetry ensures the @xmath21 electron bands ( @xmath21 is even ) to be doubly degenerate according to the kramers theorem . if one denotes the action of @xmath247 on the electron basis as @xmath250 where @xmath251 , one can show @xmath252 and @xmath253 , namely , the matrix @xmath254 has the same properties as @xmath41 . the inversion symmetry of the superconductor requires @xmath255 , and @xmath256 , where @xmath257 and @xmath258 signs indicate even and odd parity pairings , respectively . note that @xmath31 , so this parity requirement can be rewritten as @xmath259 , i.e. , the matrix @xmath260 is anti - symmetric for even parity , and symmetric for odd parity . when the superconductor has a pairing amplitude @xmath29 of even parity , nodal lines are allowed to arise and are stable against perturbations . this can be seen as follows . assume @xmath261 , where @xmath262 with @xmath263 being a diagonal matrix , while the basis of each @xmath264 ( @xmath128 identity matrix ) consists of two eigenstates related by the @xmath248 transformation . effectively , one can regard the two states as carrying spin up and down , respectively . the matrix @xmath254 can be written as @xmath265 , where @xmath266 denote the pauli matrices in the doublet basis of @xmath248 related eigenstates . we note that @xmath59 is not unique . since @xmath28 commute with the hermitian matrix @xmath46 , the same matrix @xmath59 diagonalizes @xmath46 into @xmath267 blocks . together with the condition that @xmath268 is anti - symmetric ( symmetric ) for even ( odd ) parity , we find @xmath269 for even parity pairing , where @xmath270 are real functions even in @xmath26 , and @xmath271 for odd parity pairing , where @xmath272 are real vector functions odd in @xmath26 . when the parity is odd , the gapless condition for the bdg hamiltonian is @xmath273 , which is overdetermined in 3d and generically has no solution . on the contrary , when the parity is even , the bdg band structure becomes gapless @xmath274 , which defines the nodal lines in the doubly - degenerate electron band @xmath275 . in practice , such nodal lines widely exist in cuprates with @xmath276-wave pairings @xcite and iron - based superconductors with modulated @xmath277-wave pairings @xcite . due to the double degeneracy of the bdg band structure , the nodal lines are coupled to the berry connection as non - abelian u(@xmath180 ) wilson loops . in the @xmath246 case , the berry connection @xmath79 is itself a u(@xmath180 ) gauge field , and the @xmath18-angle is simply given by the conventional u(@xmath180 ) non - abelian chern - simons theory . these wilson loops reside in the fundamental representation @xmath278 of u(@xmath180 ) where the generators are given by @xmath279 and @xmath280 ( @xmath281 ) @xcite , and accordingly the berry connection can be decomposed into @xmath282 . the matrix @xmath56 in this case is simply @xmath283 , where @xmath142 is a complex number function . a nodal line is defined by @xmath140 , while winding around the nodal line increases the phase @xmath134 of @xmath142 by @xmath91 . to compute the classical value of gauge field @xmath79 , one can choose @xmath284 and @xmath285 in the vicinity of the nodal line , in which case the field strength is @xmath286 alternatively , one can choose @xmath287 and @xmath288 with a unit vector @xmath289 , which yields a field strength @xmath290 both choices ensure @xmath59 and @xmath60 to be single - valued . this choice of classical solutions exactly results from the u(@xmath180 ) gauge freedom of the nodal line . similar to the u(@xmath63 ) case , the field strength of a nodal line becomes well defined when the time reversal symmetry or the inversion symmetry of the system is weakly broken . for instance , when the hamiltonian acquires a small mass term and becomes @xmath291 explicitly , there are four degrees of freedom , which are @xmath292 and @xmath293 ( @xmath294 ) , which is the expected number for arbitrary fluctuations of a u(@xmath180 ) gauge field in @xmath295 because @xmath296 , where here @xmath297 and @xmath295 . it is important to note that the field strength @xmath216 will depend on the signs of the eigenvalues of the upper @xmath128 block . if the signs of the mass eigenvalues of the upper @xmath128 block is @xmath298 , this leads to the curvature sitting in the abelian subalgebra of @xmath15 given by @xmath299\ \delta^2(\mathbf{k}_\perp-\mathbf{k}_{b})dk_\perp^{1}\wedge dk_\perp^{2}.\ ] ] this can be interpreted ( when the gravitational instanton number @xmath131 ) as the response of the @xmath15 system to a charged object with charges @xmath300 $ ] in the cartan of @xmath15 . if both signs are the same , the wilson loop will carry only a @xmath301 charge . if the signs are opposite , the two possibilities correspond to the fundamental of @xmath302 together with some overall @xmath165 charge . note that when the inversion symmetry is broken , the u(@xmath180 ) nodal lines naturally breaks down to abelian nodal lines with wilson loop charges @xmath164 , in agreement with our observation in sec . [ secinco ] . the wilson loop observables etc . will work exactly as in the abelian case we already discussed , because the level of the @xmath15 cs theory is 1 . if we change the gravitational instanton number so that the level of the @xmath15 cs theory is changed to @xmath12 , we would need to choose different weights in the @xmath15 weight lattice in order to get the same @xmath216 we computed in eq . ( [ fsigns ] ) . in particular , the cartan weights will now be @xmath303 . apart from the @xmath165 part , this belongs to the @xmath12-fold symmetric representation of the fundamental of @xmath304 , which is known to generate a free current and thus means that its correlations behave as in the abelian case , leading exactly to the same result expected , where @xmath305 gets multiplied by a factor @xmath12 . though uncommon in nature , condensed systems with @xmath113-fold degenerate ( @xmath306 ) band structures are theoretically possible when the symmetries of the systems are high enough . for instance , systems where electrons carry both a spin @xmath307 and a pseudospin @xmath307 may have a @xmath308-fold degenerate band structure when there are no spin - pseudospin couplings . when such a system is tri and develops a superconductivity , nodal lines with a u(@xmath113 ) symmetry may arise . for simplicity , here we only consider the case when the total number of electron bands is @xmath309 , which is the minimal system that realizes u(@xmath113 ) nodal lines . the matrix @xmath56 in eq . ( [ jk ] ) is then of the form @xmath310 with @xmath133 being a complex function . a nodal line is defined by @xmath140 ; it occupies the fundamental representation of u(@xmath21 ) group , of which the generators are @xmath311 and the @xmath312 generators @xmath313 of su(@xmath21 ) group @xcite . the matrices @xmath69 and @xmath70 around the nodal line can be generically chosen as @xmath314 where @xmath315 ( @xmath316 ) , and @xmath317 ( @xmath318 ) generates the element @xmath319 of the center of su(@xmath21 ) . the resulting field strength @xmath216 is @xmath320 which is a classical solution of the u(@xmath21 ) chern - simons theory with u(@xmath21 ) wilson loops . in paricular , just as in the @xmath15 case this will lead to curvature which is proportional to @xmath321 $ ] , which is in the fundamental representation of @xmath19 at level 1 , leading to similar formulae as before . in many cases it is desirable to consider the effect of defects in the physical space of a condensed matter system . in the physical space of 3d superconductors , vortex lines can be naturally created as line defects , while crystallographic line defects ( dislocations and disclinations ) may also be present , which lead to a breaking of the translational symmetry . interestingly , as we shall show below , such line defects in physical space give rise to effective nodal lines in the bz in a semiclassical approximation . for simplicity , let us consider a line defect along @xmath322-direction located at @xmath323 in the physical space . the line defect will generically produce a potential energy @xmath324 for the bdg quasi - particles that is centered at @xmath323 . an quasi - particle with kinetic energy @xmath325 will then have a hamiltonian : @xmath326 where we have used the fact that @xmath327 is still a good quantum number . assume the hamiltonian has a bound state @xmath328 with a bound state energy @xmath329 . such a bound state is in general a wave packet of size @xmath330 in the momentum space with the center located at @xmath331 and @xmath332 , where @xmath333 is the localization length of the bound state ( which we assume is much larger than the lattice constants ) , and the functions @xmath334 and @xmath335 are determined by the line defect hamiltonian @xmath336 . therefore , all the bound states on the physical space line defect live on a `` smeared out '' line @xmath337 in the bz given by @xmath338 , which is a closed loop . now we assume a quasi - particle is bounded on the line defect . we then introduce an additional slow potential @xmath339 which slightly breaks the translational symmetry along the line defect . such a slow potential could result from inhomogeneity of the system or slight distortions of the defect line . consider a local minimum where @xmath340 . one can write down a @xmath63d hamiltonian for the quasi - particle : @xmath341 where @xmath26 is restricted on the loop @xmath337 , the bound state energy @xmath342 , and @xmath343 is the berry connection in the bz defined in sec . ii . for generality , we have also included a new berry gauge field @xmath344 in the lagrangian space @xmath345 @xcite , which may in general arise when the translational symmetry of the system along @xmath346 and @xmath347 directions is broken @xcite . the corresponding lagrangian of the particle in the momentum space is then @xmath348 where @xmath349 is the time derivative of @xmath26 . note that @xmath350 in this lagrangian can be viewed as a periodic potential along the loop @xmath337 . accordingly , the quasi - particle will have a tunnelling amplitude of winding @xmath351 times along the loop , which can be calculated via the wkb approximation as @xmath352 . the path integral of the bounded quasi - particle is thus @xmath353 therefore , one sees the loop @xmath337 plays exactly the role of a wilson loop ( nodal line ) that couples to both @xmath79 and @xmath354 , where the wilson loop charge is now an integer winding number @xmath351 . we note that the slow potential @xmath355 only enters into the tunnelling action @xmath356 , while does not affect the wilson loop integral . in this case , the wilson loop charge @xmath351 and the level of cs action of @xmath79 , i.e. , the gravitational instanton number @xmath12 , are independent of each other . in principle , the gauge field @xmath354 may also have a physical cs action at some level @xmath357 in the lagrangian space @xmath345 . in this case , one will obtain a double cs theory which is more interesting . in general , the number of quasi - particles bounded on the line defect is not limited to one . to take all number of quasi - particles into account , we can second quantize the above action , rewriting the path integral as @xmath358\psi_j}\ , \ ] ] where @xmath359 is the fermion mode associated with a particular closed path @xmath360 on the loop @xmath337 , and @xmath361 is the period in the time direction . we note that the berry connection @xmath79 here should be treated as defined at the specific point @xmath362 of the physical space where the quasi - particles are trapped , i.e. , the local minimum of the line defect . in fact , the strict definition of the berry connection @xmath79 does require specifying a point @xmath363 in the physical space as an origin @xcite , namely , @xmath79 is defined via the bloch wave functions in the unit cell at @xmath363 . in particular , since the bloch wave functions have a shift ambiguity @xmath364 under a translation @xmath365 ( where @xmath366 is a lattice vector ) , the berry connections @xmath79 defined at @xmath363 and @xmath367 are related by @xmath368 . this shift ambiguity is closely related to the 3d quantum hall effect @xcite , but does not affect the calculation of @xmath130 $ ] of tri systems here . similarly , the gauge field @xmath354 and the associated lagrangian space @xmath345 should be viewed as located at the specific position @xmath369 ( local minimum of @xmath355 ) and @xmath338 @xcite . it turns out that the above description is very natural in the context of topological strings , which was introduced in @xcite and gives a realization of chern - simons theory as a string theory . this corresponds to strings propagating on a six dimensional symplectic manifold which is typically taken to be a calabi - yau 3-fold . @xmath21 d - branes wrapping lagrangian 3-cycles lead to @xmath19 chern - simons theory living on it . moreover , intersecting d - branes along a loop leads to insertion of wilson loop observables as in eq . ( [ sum ] ) on the intersection @xcite . to connect this to the present discussion , we note that the phase space @xmath370 is a sympletic manifold @xmath16 . we take two lagrangians to be one wrapped around @xmath371 giving us the cs theory with gauge field @xmath343 , and the other to represent the defect , given by the lagrangian @xmath372 with gauge field @xmath373 living on it . these lagrangians intersect along @xmath327 , leading to a fermion field which couples to the two gauge fields exactly as in eq . ( [ mohem ] ) @xcite . it is remarkable that phase space in 3d can naturally give a realization of topological strings ! though our discussion here is based on a straight line defect in physical space , the result holds for generic smooth line defects . in the semiclassical approximation , one can define a momentum along the defect line which plays the role of @xmath327 here , and all the rest of the derivation will follow . we have seen how a modified berry s connection in the bz of a tri superconductor can behave as a dynamical gauge field in 3d with cs action whose level is given by the gravitational instanton number in the physical spacetime . moreover , we have seen that gapless nodal lines play the role of wilson loop observables for the chern - simons theory . the free energy of this cs theory computes the @xmath18-angle of the topological gravitational response @xmath4 of this system . changing of the topological mutual linking numbers of nodal lines leads to a shift of the @xmath18-angle in units of @xmath14 . whereas the linking of degenerate manifolds is also studied for the 5d weyl semimetals @xcite , it is reasonable to expect linking and knot invariants to play an important role in more generic gapless topological states of matter . the examples we have discussed exhibit mostly an abelian structure . even in the multi - band case where we got @xmath19 cs theory , the computations reduce to @xmath374 abelian cs theory . it would be nice to find ways where the linked nodal amplitudes are genuinely non - abelian . this paper provides an example of how the berry connection in the bz can become dynamical . in addition , in the discussion of physical space defects in sec . [ secdefect ] , we see possibility of realizing topological string theory in a condensed matter set - up . we have seen an intriguing realization of topological strings on the six dimensional phase space where defects play the role of d - branes . it would be very interesting to develop this connection further , for instance , to see how the gauge field @xmath354 in the lagrangian manifold @xmath345 can also be made dynamical and governed by a cs action . it would also be useful to find more examples along these lines in different condensed matter systems in different dimensions , which need not be limited to topological theories . we would like to thank c. liu and c. xu for discussions . we gratefully acknowledge support from the simons center for geometry and physics , stony brook university , where some of the research for this paper was performed during the 2016 simons summer workshop . cv would also like to thank stanford institute for theoretical physics for hospitality during part of this work . the research of bl and scz is supported by the nsf grant dmr-1305677 . the research of cv is supported by the nsf grant phy-1067976 . the research of fv is supported by the nsf grant dmr-1151208 . the berry connection @xmath64 defined in the literature @xcite is usually seen as a natural non - abelian generalization of the u(@xmath63 ) berry connection proposed by berry @xcite , where @xmath66 are the eigenstates of the hamiltonian @xmath375 and @xmath376 runs over the @xmath21 occupied bands . however , when the eigenstates are nondegenerate , the only degrees of gauge freedom in this definition are the phase rotations of the eigenstates @xmath377 , which leads to a gauge transformation @xmath378 with @xmath379 being the kronecker symbol . this just gives @xmath80 a gauge freedom u(@xmath63)@xmath73 . one may wonder what happens if one makes a unitary transformation of the basis used for writing the hamiltonian as @xmath83 , which is also the basis for writing the eigenstates @xmath66 . such a transformation leads to @xmath380 and @xmath381 ( as we will show later ) , and according to eq . ( [ aprime ] ) , leads to a transformation @xmath382\ ] ] which is not a gauge transformation . therefore , the basis transformation does not transform @xmath80 legally . we conclude @xmath80 only has a u(@xmath63)@xmath73 gauge freedom instead of a u(@xmath21 ) gauge freedom , and can not be regarded as a u(@xmath21 ) gauge field . on the other hand , we shall show the modified berry connection @xmath74 is a well - defined u(@xmath21 ) gauge field , which is uniquely defined for superconductors . to see this , we note that under a unitary u(@xmath21 ) basis transformation @xmath83 , one has @xmath383 , therefore @xmath380 and @xmath381 , and @xmath79 given in eq . ( [ a ] ) transforms as @xmath384 which is exactly the gauge transformation of the u(@xmath21 ) gauge field . therefore , we can view the basis transformation @xmath83 as the u(@xmath21 ) gauge transformation of @xmath79 . therefore , for gapped superconductors where @xmath69 and @xmath70 can be well defined everywhere in the brillouin zone , we have @xmath385=-\mathcal{a}[a']$ ] . besides , note that the integral in the 3d brillouin zone @xmath389 is an integer for unitary matrix @xmath69 , we conclude that @xmath385=-\mathcal{a}[a']=n\pi$ ] with @xmath144 , in agreement with the requirement of the time reversal symmetry . in the presence of nodal lines , however , @xmath69 and @xmath70 become ill - defined on the nodal lines , and this conclusion no longer holds .

in order to attain a statistical description of the evolution of cosmic density fluctuations in agreement with results from the numerical simulations , we introduce a probability conditional formalism ( cf ) based on a complete inventory of isolated overdense regions in a density random field . this formalism is a useful tool for describing at the same time the mass function ( mf ) of virialized dark haloes , their mass aggregation histories ( mahs ) and merging rates ( mrs ) . the cf focuses on virialized regions in a self - consistent way rather than in mass elements , and it offers an economical description for a variety of random fields . within the framework of the cf , we confirm that , for a gaussian field , it is not possible to reproduce at the same time the mf , mah , and mr of haloes , both for a constant and moving barrier . then , we develop an inductive method for constraining the cumulative conditional probability from a given halo mf description , and thus , using the cf , we calculate the halo mahs and mrs . by applying this method to the mf measured in numerical simulations by tinker et al . , we find that a reasonable solution , justified by a mass conservation argument , is obtained if a rescaling increment by of the virial mass defined in simulations is introduced , and a ( slight ) deviation from gaussianity is taken into account . thus , both the mah and mr obtained by a monte carlo merger tree agree now with the predictions of numerical simulations . we discuss on the necessity of rescaling the virial mass in simulations when comparing with analytical approaches on the ground of the matter not accounted as part of the halos and the halo mass limit due to numerical resolutions in the simulations . our analysis supports the presence of a diffuse dark matter component that is not taken into account in the measured halo mfs inasmuch as it is not part of the collapsed structures . [ firstpage ] cosmology : dark matter large - scale structure of universe galaxies : haloes galaxies : formation methods : statistical methods : analytical

1304.0765

according to the contemporary cosmological paradigm , cosmic structures emerge from the gravitational growth of primordial dark matter density fluctuations . a central problem in the last four decades has been the connection between this primordial fluctuation field and the abundance and assembly history of the virialized dark matter haloes ; in more detail , their mass function ( mf ) , mass aggregation histories ( mahs ) , and merger rates ( mrs ) . since the seminal work by press & schechter ( 1974 ; hereafter ps ) , many statistically - based analytical formalisms were developed in order to perform such a connection ( see for reviews e.g. , zentner 2007 ; mo , van den bosch & white 2010 ) . in order to achieve an association between the density fluctuation field and the virialized haloes , one needs : a statistical description of the density fluctuation field , an inventory of the overdense regions which will be associated to the virialized objects , and a model for the dynamical evolution of the overdensities , including an operational criterion of collapse ( virialization ) . ps assumed a gaussian density field and used the spherical collapse model ( gunn & gott 1972 ) . a region of the density field is assumed to end as a virialized halo when its linearly evolving overdensity exceeds a critical value @xmath1 . the virialized mass at a fixed scale is assumed to be given by the contributions of the regions of this scale overdense by @xmath1 ( the _ ps ansatz _ ) plus the underdense regions of the same scale contained in larger regions overdense by @xmath1 ; the latter statement enunciates the so - called _ cloud - in - cloud problem_. in order to account for all mass , ps assumed that the second contribution equals to the first one , which justifies this the fudge factor of 2 introduced in their inferred halo mf . based on a more rigorous statistical description , the excursion set ( es ) , also know as extended ps ( eps ) formalism ( peacock & heavens 1990 ; bond et al . 1991 ) , allows to overcome the cloud - in - cloud problem . the es formalism applied to a gaussian field provides a tool to compute , besides the unconditional mf , the conditional mf that can be used for generating halo merger trees ( bond et al . 1991 ; lacey & cole 1993 ) . in this formalism , the result depends critically on the filter used to average the linear density fluctuation field at different mass scales . the mathematical solution is straightforward when a sharp @xmath2-space filter is adopted . for this filter , the ps result including the fudge factor of 2 is recovered . in the calculation of the merger trees , the sharp @xmath2-space filter implies independent steps ( markovian random walks ) along the mass trajectory . note that the es formalism has a conceptual problem for predicting the halo mass in which a particular mass element ends up ( mo et al . 2010 , 7.2.2 ) , problem that may lead to an inaccurate buildup of mahs and mrs . with the advent of large cosmological n - body simulations , the whole non - linear process of dark matter gravitational evolution and collapse into virialized haloes could be followed , though with strong limitations due to mass resolution . do approaches based on the gaussian es formalism allow to describe correctly the results from the simulations ? in particular , _ do results from these approaches agree at the same time with the mf , mahs , and mrs of haloes as measured in simulations ? _ a non - negligible discrepancy between the mfs obtained in the es formalism and the n - body simulations has been early reported . the introduction in the es formalism of the elliptical gravitational collapse instead of the spherical one ( mass - dependent instead of constant @xmath1 , respectively ) helped to overcome this problem ( sheth , mo & tormen 2001 ) . however , in this case the es formalism does not provide an analytic formula for the conditional mf , and the merger trees based on the gaussian es formalism show deviations with respect to simulations in the progenitor mass distributions ( sheth & tormen 2002 ) , the mass contained within all the progenitors ( neistein et al . 2006 ) , and the average main progenitor mahs and mrs ( wechsler et al . 2002 ; van den bosch 2002 ) . on the other hand , direct measures of the conditional mfs in n - body simulations show that they depart from the corresponding functions calculated with the es formalism for a gaussian field ( cole et al . 2008 ; neistein et al . thus , a possible source of the discrepancies lies in the strict assumption of gaussianity in the es formalism . note that the question is not about primordial non - gaussianity ; the n - body simulations use indeed a gaussian density field that is evolved analytically ( e.g. , by means of the zeldovich approximation ) until the quasi - linear regime is reached . the point is that non - negligible deviations from gaussianity at the scales of interest have likely happen already during this regime ( coles & jones 1991 ) . here , we set aside gaussianity and the es formalism and handle the problem of the inventory of overdense regions in arbitrary random fields . we present an approach aimed to state the problem of the inventory of overdense regions in random ( gaussian or non - gaussian ) fields from a very general point of view . we assume that the density fluctuation field is fully characterized by its field conditional probability function . in order to face the problem of associating the halo mass to the overdensity in an alternative way to the es formalism ( see the reference mo et al . 2010 cited above ) we proceed as follows . in our _ conditional formalism _ ( hereafer cf ) , we make use of the concept of _ isolated regions _ introduced by jedamzik ( 1995 ; see also yano , nagashima & gouda 1996 ; nagashima 2001 ) as a natural way to connect overdensities to virialized regions . we enunciate the _ isolated overdense regions inventory _ theorem , through which the given field conditional probability function of the random field is linked to the inventory of isolated overdense regions inside larger isolated regions of lower overdensity , being the former those that eventually collapse into virialized haloes if their overdensities equal to @xmath1 . the cf offers an alternative eps formulation . it gives an easy way to develop a monte carlo algorithm for calculating the mah and mr of the growing virialized haloes , because it supplies naturally the progenitor conditional probability function of finding a specific isolated region ( progenitor ) inside a given isolated region ( descendent ) . when the problem of finding isolated overdense regions inside larger ones is extended to the overall universe , then such conditional probability reduces to an unconditional probability , which gives the halo mf . nevertheless , if one starts from an unconditional probability function , for example on the basis of a halo mf obtained from n - body simulations , and tries to go back to the conditional probability function , in order to build merger trees compatible with the given mf , then multiple possibilities appear . the cf offers useful tools to make a choice . as we will discuss in 4.2 , we decide to leave to the gaussian hypothesis even if , for simplicity , we retain the ( gaussian ) ps ansatz in a generalized version . the cf might be a powerful tool for economically describing _ in a consistent way the simulation results related to the mf , mahs , and mrs of the virialized haloes_. these descriptions can then be easily implemented in semi - analytical and semi - empirical models of galaxy evolution , and can be extended to masses and epochs , where the resolution is a limit for the simulations . on the other hand , the cf may allow to explore economically cosmic structure formation in alternative cosmologies or in cases where the density fluctuation field is intrinsically non - gaussian . our approach could be applied also in other astrophysical problems , where random density fields are introduced , e.g. , the formation and evolution of dense gas structures in the interstellar medium as giant molecular clouds , massive clumps and cores ( hopkins 2012 ) . this paper is organized as follows . section 2 is devoted to present the cf ; in 2.1 , an equation for the isolated overdense regions inventory is derived , and in 2.2 the algorithms to build up the merger trees are introduced . section 3 deals with the particular case of gaussian random fields . the halo mf is derived in 3.1 , while the mah and mr are calculated in a discussion on the moving barrier effects is presented in 3.3 . in view of the discrepancies obtained when using as input the gaussian statistics , in section 4 an heuristic approach , based on the use of the halo mf from simulations as input , is presented . in 4.1 we approach the problem of the difference on the mass estimates in the analytic method and in the simulations , and remark the necessity to introduce a mass rescaling , representative of a diffuse matter component , to connect the two standpoints . the results concerning mahs and mrs are presented in our conclusions are given in section 5 . we adopt a flat @xmath3cdm cosmology with @xmath4 . the es formalism provides a means to derive the halo mf , accounting for the cloud - in - cloud problem , and to build up the halo mahs . however , as mentioned above , the gaussian es formalism seems to have difficulties for predicting at the same time the mf , mah , and mrs in agreement with numerical simulations . also , with the es formalisms it is not easy to deal with non - gaussian random fields ( inoue & nagashima 2002 ) , though some progress has been made recently by introducing a stochastic barrier and non - markovian corrections ( maggiore & riotto 2010 ) , by manipulating the step - size distribution of the random walks ( lam & sheth 2009 ) or by accounting for correlations between steps ( musso & paranjape 2012 ) . on the other hand , being the es formalism focussed on a generic mass element , the virialized halo mass is associated to a smoothing scale , which may lack of a real physical meaning . in order to overcome these shortcomings , we introduce below the conditional formalism ( cf ) inspired by the jedamzik ( 1995 ) approach . this formalism is based on a complete inventory of the isolated overdense regions of a random density field described by a _ cumulative field conditional probability function_. let start with some key definitions . _ isolated regions _ : an isolated region with overdensity @xmath5 is a connected region not included in a larger connected region with overdensity @xmath6 , while at the same time , outside of the totality of isolated regions with overdensity @xmath5 there does not exist any region with overdensity @xmath6 . hereafter we will refer to an isolated region with overdensity @xmath5 by means of the symbol @xmath7 . , one can cut an arbitrary region of such space along a specific border . the consequence will be that some @xmath8 will be internal , others will be external and finally some of them will be on the border of that region . by definition , when no one of the isolated regions @xmath8 will be on the border , such region will be named _ suitable _ with respect to its internal isolated regions overdense by @xmath5 . it is straightforward to see that an isolated region @xmath7 is a _ suitable _ region with respect to all its internal isolated regions more overdense than @xmath5 . another example is given by an isolated region @xmath7 where one ( or more ) internal isolated region(s ) @xmath9 with @xmath10 has ( have ) been removed . the result is a _ suitable _ region @xmath11 with respect to its internal isolated regions @xmath12 , its overdensity is less than @xmath5 , and typically it will not be an isolated region . any region of @xmath11 more overdense than @xmath13 is contained inside an internal isolated region @xmath12 . afterwards when we will refer to a region @xmath11 actually we will refer to a region _ suitable _ with respect to the internal isolated regions overdense by @xmath5 involved by the problem . ] _ virialized regions _ : a virialized region is an isolated region whose overdensity @xmath5 is equal to a critical value @xmath1 . @xmath14 : is the _ field conditional probability _ to find a region of mass @xmath15 with an overdensity between @xmath13 and @xmath16 , contained inside a larger region of mass @xmath17 and overdensity @xmath5 . @xmath18 : is the number of isolated regions , i.e. the _ conditional mf _ , with overdensity @xmath13 and mass between @xmath15 and @xmath19 , contained inside a larger region @xmath11 with overdensity @xmath5 and mass @xmath17 . the _ cumulative field conditional probability _ to find a region of mass @xmath15 with overdensity @xmath20 inside a larger region of mass @xmath17 and overdensity @xmath5 is @xmath21 our cf concerns random fields that can be defined by such conditional probability . following , a theorem focussed on an _ isolated overdense regions inventory _ and aimed to establish a connection between the cumulative field conditional probability , @xmath22 , and the conditional mf , @xmath23 with @xmath24 , is presented . consider a region @xmath11 with overdensity @xmath5 and mass @xmath17 . define an arbitrary mass @xmath25 . a decomposition of the mass range @xmath26 into @xmath27 ordered steps , @xmath28 , identifies a sequence of @xmath29 isolated regions @xmath12 with overdensity @xmath13 and masses between @xmath30 and @xmath31 inside @xmath11 . for any given @xmath32 , the amount of mass in @xmath11 assembled by regions with mass @xmath33 and overdensity between @xmath34 and @xmath35 is @xmath36 this mass is also given by the sum of the regions of mass @xmath33 with overdensity between @xmath34 and @xmath35 located in each one of the isolated regions @xmath12 with masses between @xmath30 and @xmath31 , and overdensity @xmath13 of the above defined sequence : @xmath37 no other contribution exists because outside the totality of the isolated regions @xmath12 in @xmath11 there are not regions with overdensity @xmath38 . by equating the last two contributions , we can write @xmath39 from which we derive the following integral equation : @xmath40 by integration on @xmath34 between @xmath13 and @xmath41 , the previous equation writes in terms of the cumulative field conditional probability : @xmath42 an alternative formulation of eq . ( [ fundeqa ] ) is obtained introducing the _ conditional probability _ @xmath43 of finding an isolated region of mass between @xmath15 and @xmath19 and overdensity @xmath13 inside a region @xmath11 . given a region @xmath11 with overdensity @xmath5 and mass @xmath17 , our interest is now to identify inside it an isolated region with overdensity @xmath13 . the quantity @xmath44 gives the amount of mass of @xmath11 assembled by isolated regions with overdensity @xmath13 and mass between @xmath45 and @xmath46 . therefore , the probability that an arbitrary mass element of @xmath11 is part of an isolated region with overdensity @xmath13 and mass between @xmath45 and @xmath46 is @xmath47 that is @xmath48 the previous statement is equivalent to say that @xmath49 is the probability to find an isolated region with overdensity @xmath13 and mass between @xmath45 and @xmath46 inside @xmath11 . introducing eq . ( [ prob ] ) into eq . ( [ fundeqa ] ) , we obtain @xmath50 which is basically a volterra integral equation . the fundamental aspect of eq . ( [ fundeqa ] ) is that for any given cumulative field conditional probability @xmath51 , it gives the conditional mf , @xmath23 . its main difference with the jedamzik ( 1995 ) result is that eq . ( [ fundeqa ] ) is applied to any region @xmath11 with overdensity @xmath5 and mass @xmath17 and not only to the overall universe . for this reason the cumulative field conditional probability @xmath51 appears now on both sides of the equation . such property highlights the intrinsic meaning of the cf . it is interesting to remark that the previous _ isolated overdense regions inventory _ theorem solves automatically the cloud - in - cloud problem . ( [ fundeqa ] ) represents the generalization of the ps formalism . such formalism may be recovered for the special random fields wherein @xmath52 with @xmath2 constant . in such case eq . ( [ fundeqa ] ) writes @xmath53 the ps ansatz in terms of the conditional probability and including already the fudge factor of 2 , is obtained in the particular case of @xmath54 ( corresponding namely to gaussian fields , see below ) . a random field for which eq . ( [ anscon ] ) is satisfied is very simple to handle in terms of eq . ( [ ans ] ) . note that not only gaussian fields obey eq . ( [ anscon ] ) but also a variety of non - gaussian fields . therefore , in order to find an analytical formalism in agreement with simulations and to avoid the complex numerical integration of eq . ( [ fundeqa ] ) , it is natural for us to look at first among random fields which satisfy eq . ( [ anscon ] ) . this is the strategy followed in section 4 . a merger tree realization needs to identify progenitors of a given descendant . this is done by means of a monte carlo algorithm that uses the conditional probability and mf . the choice of a monte carlo algorithm is not unique and may hide subtle physical and technical questions ( for a review see zhang et . al . 2008 ) . the simplest criterion establishes that the monte carlo algorithm has to reproduce the conditional probability and mf predicted by the theory . however , this criterion is not sufficient , because the overall meaning of the involved probability , i.e. the stocahstic nature of the problem , has to be taken in to account . we will introduce a monte carlo algorithm optimized to build up a merger tree . afterwards we will check our algorithm with tests that involve the conditional probability and mf , as well as the mah and mr in order to verify also the agreement with the stochastic nature of the problem . only in this section , to ease the reading of the formulae , we will indicate the variables of the extracted regions by a lower case letter . integrating @xmath55 between @xmath56 and @xmath57 and identifying the obtained cumulative probability with a random number @xmath58 uniformly distributed in the interval [ @xmath56@xmath59 , we write @xmath60 which represents the basis for a monte carlo approach to build up merger trees . given an isolated region @xmath61 with overdensity @xmath62 and mass @xmath63 , the extraction of the random number @xmath64 by eq . ( [ mont ] ) allows to identify inside it a first isolated region @xmath65 of mass @xmath66 with a given overdensity @xmath13 . taking into account that the _ isolated overdense regions inventory _ theorem has been proved for general regions ( see footnote@xmath67 ) , after the identification of the isolated region @xmath65 , the extraction may continue on the complementary region @xmath68 with mass @xmath69 and volume @xmath70 obtained from @xmath61 removing @xmath65 from its inside . in terms of the density , from the volume relation we write @xmath71 being the region density @xmath72 , where @xmath73 is the background density and @xmath5 the overdensity , we have @xmath74 and finally @xmath75 at this point the extraction of the next isolated region can be carried out on the region @xmath68 with overdensity @xmath76 and mass @xmath77 . to perform such extraction , for simplicity , we will use eq . ( [ mont ] ) with a new random number @xmath78 , applied on the region @xmath68 . by proceeding in this way we implicitly assume that the same statistics of @xmath61 holds for @xmath68 , excluding any sort of correlation . for a discussion of possible correlations , see sheth & lemson ( 1999 ) . according to what has been said above , afterwards we will proceed to test the robustness of such hypothesis . a recurrent algorithm may be particularly useful now . from the @xmath79 complementary region @xmath80 of mass @xmath81 with overdensity @xmath82 calculated by the previous method , using eq . ( [ mont ] ) with a random number @xmath83 we can identify inside it an isolated region @xmath84 with mass @xmath85 and overdensity @xmath13 . at this point the next complementary region @xmath86 with mass @xmath87 will have the overdensity @xmath88 equations ( [ mcon ] ) and ( [ overev ] ) guarantee a mass conservation and make the procedure particularly handy because they do not make explicit reference to the previous i - steps . equations ( [ mont][overev ] ) are particularly useful for constructing the halo merger trees . suppose a simple case when the isolated virialized regions are defined by a critical overdensity @xmath1 function of @xmath89 only . given a _ descendant _ at @xmath89 with overdensity @xmath90 , the _ progenitors _ at @xmath91 have overdensity @xmath92 . the finding of such progenitors ( branches ) is performed by applying the monte carlo extractions of eq . ( [ mont ] ) to the ( complementary ) region described by eqs . ( [ mcon ] ) and ( [ overev ] ) . this procedure identifies a sequence of progenitors at @xmath91 that can be sorted in order of decreasing mass . the largest ( primary ) progenitor identifies the main progenitor by which the _ mass aggregation history , _ mah , is obtained . the overall population of less massive progenitors fulfills the _ merger tree _ and allows to find the _ merger rate _ , mr , along the mah . note that the condition @xmath93 implies @xmath94 . therefore , for the complementary regions , the progenitor extractions can be performed meanwhile @xmath95 . this means that the progenitors are extracted until all matter in the volume is exhausted , which implies that the entire mass contained in the descendant should come from virialized structures . however , the physics of halo mass assembly is more complex . a fraction of the mass can be present as diffuse matter . the strict limit @xmath96 implies lack of diffuse matter . in this paper we explore solutions with @xmath96 , however we will have in mind the physical consequences of such a hypothesis . the above described method has been extensively used in the case of gaussian cdm density fields for generating the initial conditions ( the mahs ) for the virialization process of dark haloes ( avila - reese , firmani & hernandez 1998 ) , and the subsequent formation and evolution of disc galaxies inside them ( firmani & avila - reese 2000,2009 ; avila - reese & firmani 2000 ) . the obtained halo properties agree with those found in cosmological numerical simulations ( avila - reese et al . 1999 ) . it is interesting to remark that the mass distribution of the first progenitor obtained with the monte carlo method can be compared with the conditional probability function , while the mass distribution of all the progenitors can be compared with the conditional mf . this offers a first test of our monte carlo algorithm by comparing its one - step results with the straight analytic formulae from the statistics ( in 3.2 we will show this point in some detail ) . it is important to highlight that the mean halo mah has to be a convergence solution of the monte carlo algorithm for the integration redshift step @xmath97 , otherwise it is physically meaningless . because of the stochastic nature of the problem , this property is dependent on the behavior of the given cumulative field conditional probability . we will focus on _ well behaved cumulative field conditional probabilities , such that guaranty @xmath97 convergence in the mah buildup_. in the case of a gaussian random field , the cumulative field conditional probability to find a region with mass @xmath15 and overdensity @xmath38 contained inside the region of mass @xmath17 with overdensity @xmath5 is given ( see bower 1991 ) by : @xmath98 where @xmath99 and @xmath100 is the mass variance calculated from the density fluctuation power spectrum of the random field . making the limit @xmath101 , eq . ( [ gaussdist ] ) gives @xmath102 which is eq . ( [ anscon ] ) with @xmath54 . ( [ ans ] ) reduces to @xmath103 this is just the ps ansatz in terms of the conditional probability with the fudge factor of 2 included . introducing eq . ( [ gaussdist ] ) into eq . ( [ psans ] ) we obtain @xmath104 it is easy to verify that such conditional probability is normalized , i.e. , its integral in @xmath15 from @xmath56 to @xmath17 ( @xmath105 from @xmath56 to @xmath41 ) is 1 . ( [ psgen ] ) is applied to a region of mass @xmath17 and overdensity @xmath5 . we can extend such region to the entire universe making @xmath106 , @xmath107 , and @xmath108 . hereafter , in such limit case of the entire universe , the superscript @xmath109 will be omitted . a recipe for the collapse criterion may be obtained from the top - hat spherical gravitational collapse , which establishes that an isolated region at redshift @xmath89 is virialized when its overdensity is equal to a critical value @xmath1 ( navarro frenk & white 1997 ) . for a flat cosmology with cosmological constant ( @xmath110 ) : @xmath111 where @xmath112 is the linear growth factor normalized to 1 at @xmath113 . by defining the number density per mass unit @xmath114 definition that is not limited to gaussian fields , and introducing @xmath115 we obtain the halo mf @xmath116 which coincides with the ps mf . in fig . [ psmafun ] , the solid red curves show the gaussian - case ( ps ) halo mf obtained from eq . ( [ ps ] ) at @xmath117 , while black solid curves are the accurate fitting formulae to cosmological n - body simulations provided by tinker et al . ( 2008 ; hereafter tinker08 ) ; the parameters for the virial mass at the overdensity @xmath118 ( bryan & norman 1998 ) corresponding to our cosmology were used . the comparison reveals the well known excess ( deficit ) of intermediate ( high ) mass haloes predicted by the ps formalism . the tinker08 fitting curves are valid for halo masses above @xmath119 , which corresponds to the minimal mass at which haloes are reasonable resolved in the simulations studied by these authors ; this limitation makes uncertain the normalization of the overall mf . just in order to explore possibilities to approximate the ps mf to the simulations mf , @xmath1 has been rescaled by a factor @xmath120 to fit the high mass cut - off of the n - body ( tinker08 ) curves ( for similar earlier attempts see e.g. , carlberg & couchman 1989 ; klypin & rhee 1994 ) , and the normalization has been lowered by a factor @xmath121 to fit the maxima ( red dotted curves ) . note that the ps mf is normalized independently of @xmath1 , and that the dotted curves imply now a not normalized mf . it is evident that in spite of these mass - independent transformations of the ps mf , the comparison with simulation results continue failing . the exercise presented above is congruent with the one carried out in sheth & tormen ( 1999 ) , where they showed that besides these operations , the ps mf should be multiplied by a @xmath122 ( mass)-dependent factor in order to fit the numerical simulations analyzed by them . for a gaussian field and the critical overdensity given by eq . ( [ dlc ] ) , if @xmath123 , then @xmath124 , @xmath125 and @xmath126 . using eqs . ( [ psans ] ) and ( [ gaussdist ] ) , eq . ( [ mont ] ) writes as @xmath127 this operation is equivalent to make a random choice of a number @xmath105 with a normal deviate , zero mean and variance @xmath128 . hence , using eq . ( [ gamma ] ) and making reference to 2.2 , we obtain : @xmath129 where the symbols are the same as in 2.2 , and @xmath130 is the ( rescaled critical ) overdensity of the descendant at redshift @xmath89 , while @xmath131 is the ( rescaled critical ) overdensity of a progenitor at redshift @xmath132 . the @xmath133 sequence defines the complementary regions of mass @xmath134 with overdensity @xmath135 , as well as the progenitors of mass @xmath30 with overdensity @xmath136 . equations ( [ mcon ] ) , ( [ overev ] ) and ( [ meq ] ) , togheter , define the monte carlo algorithm to build the merger tree of a virialized halo with a given mass at a given redshift . in fig . [ psmafr ] , we plot the average of @xmath137 different mahs corresponding to a present - day halo of mass @xmath138 ( red solid line ; in this case , the critical overdensity is not rescaled ) . as in fig . [ psmafun ] , the red dotted line corresponds to the rescaling in the critical overdenstiy and the amplitude - reduction correction of the mf mentioned above . the dashed line represents the fit to the corresponding average mah measured in the millennium simulations ( fakhouri , ma & boylan - kolchin 2010 ) . for present - day haloes of @xmath138 , fig . [ psmefr ] shows the mean mr histories _ per unit of redshift _ with merger ratios greater than @xmath139 ( @xmath140 and @xmath141 are the masses of the primary and secondary progenitors , respectively ) . from bottom to top , @xmath142 and @xmath143 ( green lines ) and @xmath144 , @xmath145 , and @xmath146 ( blue lines ) . the coding of solid and dotted lines is the same as in fig . [ psmafr ] . the average halo mrs as a function of the @xmath147 threshold at @xmath113 compare well with the results reported in fig . 3a by fakhouri et al . the small change of these mean mrs with @xmath89 is also in general agreement with the numerical simulation results reported in fakhouri & ma ( 2008 ) and fakhouri et al . ( 2010 ) . ( red , left abscissa ) and the conditional mass function @xmath148 ( green , right abscissa ) per unitary progenitor mass natural logarithm as a function of the progenitor mass . the dots show the average distribution obtained by the monte carlo algorithm for the first progenitor ( bottom ) and for the entire progenitor collection ( top ) using eqs . ( [ mcon ] ) and ( [ overev ] ) . the descendant mass is @xmath149 , the redshift @xmath113 while the redshift step @xmath150 ( upper panel ) and @xmath151 ( lower panel ) . ] ( green ) per unitary progenitor mass natural logarithm as a function of the progenitor mass . the dots show the average distribution obtained by the monte carlo algorithm for the entire progenitor collection using eqs . ( [ mcon ] ) and ( [ overev ] ) . in the upper panel the descendant masses are @xmath152 and @xmath153 , the redshift @xmath113 and the redshift step @xmath150 , while in the lower panel the descendant mass is @xmath149 , the redshift @xmath154 and the redshift step @xmath150 . ] the main conclusion from the exercise presented here is that for a gaussian density field ( ps ) , the predicted average halo mah and mrs agree reasonably well with the results from numerical simulations . and [ psmefr ] ) . in what follows , unless otherwise stated , we will use the rescaled critical overdensity case . ] however , the predicted halo mf , as it is well known , has an excess at intermediate masses and a deficit at high masses as compared to simulations at the mass range they are able to resolve ( e.g. , tinker08 ) . particularly interesting are the one - step results shown in fig . [ test1 ] for a descendant mass of @xmath149 at @xmath113 and redshift intervals of @xmath150 ( upper panel ) and @xmath151 ( lower panel ) , respectively . the red curve ( left abscissa ) shows the analytic conditional probability @xmath155 per unitary progenitor mass natural logarithm , while the green curve ( right abscissa ) shows the conditional mf @xmath148 per unitary progenitor mass natural logarithm as a function of the progenitor mass . here @xmath5 correspond to @xmath113 while @xmath13 to @xmath156 . the dots show an average distribution obtained with our monte carlo algorithm for the first progenitor ( bottom region ) and for the entire progenitor collection ( top region ) . the total number of extractions , not reported here because unnecessary , is regulated in each case to reduce the scatter at reasonable levels . the marginal differences appearing in the graphics can arise from numerical effects , as well as from some departure from the hypothesis made in 2.2 about the complementary regions . fig . [ test2 ] shows a similar test carried out for the masses @xmath152 and @xmath153 at @xmath113 ( upper panel ) and for @xmath149 but at @xmath154 ( lower panel ) , being in both cases @xmath150 . we conclude that the agreement between the analytic conditional probability and mf and the monte carlo results as well as the comparison of the mahs and mrs with simulations ( see above ) are fully satisfactory , and it indicates that our merger tree algorithm works properly . for our main calculations we will assume a redshift step close to @xmath157 , while for step invariance tests we will assume @xmath158 . so far , we have used the spherical collapse critical overdensity ( eq . [ dlc ] ) , which is function of @xmath89 but not of mass ; this corresponds to a _ fixed barrier _ in the es formalism . the adoption of a mass - dependent critical overdensity , i.e. a _ moving barrier , _ compatible with an ellipsoidal collapse , improves the agreement between the analytic and the numerical results regarding the halo ( unconditional ) mf ( sheth et al . 2001 ; sheth & tormen 2002 ) . we extend now the test to the halo mah and mr . we assume the same gaussian random field as in section 3 described by eqs . ( [ gaussdist])([psgen ] ) and eqs . ( [ cmf])([ps ] ) , and adopt the moving barrier @xmath160 where @xmath161 , @xmath162 is a function that in sheth et al . ( 2001 ) is @xmath163 , and @xmath164 , @xmath165 , and @xmath166 are free parameters . the cf can be used now to explore the effects of a mass - dependent critical overdensity on the mf , mah and mr , under the anstaz that the cumulative field conditional probability is the same as given in eq . ( [ gaussdist ] ) . using eq . ( [ ps ] ) with the parameter values given in sheth et al . ( 2001 ) , @xmath167 , @xmath168 and @xmath169 , we recover almost the same halo mf proposed by sheth & tormen ( 1999 ) as a fit to numerical simulations , and justified in sheth et al . ( 2001 ) as an indication of elliptical collapse ( fig . [ stmf ] ; note that both the obtained mf and the sheth & tormen one are close to the mf given in tinker08 ) . the excellent agreement shows that our ansatz is quite reasonable . within the framework of the es approach , recent work on random walks with correlated steps suggests that substituting @xmath170 into eq . ( [ gamma ] ) could be indeed a good approximation ( c.f . paranjape , lam & sheth 2012 ; musso & sheth 2012 ) . to find the mah and mr , we use eqs . ( [ gaussdist])([psans ] ) with @xmath170 given by eq . ( [ dlec ] ) . for @xmath171 ( @xmath172 ) , @xmath125 only if @xmath173 , while @xmath174 when @xmath175 and @xmath176 for @xmath177 . in these three cases , @xmath178 is @xmath179 , a value between @xmath56 and @xmath179 , and @xmath56 , respectively . ( [ psans ] ) , eq . ( [ mont ] ) now writes @xmath180 this equation does not have a solution when @xmath181 , while it does have a solution when @xmath173 for each value of @xmath182 , and for @xmath175 only if @xmath183 . within the framework of our approach , we conclude that a mass - dependent critical overdensity , as the one inferred from the moving barrier analysis suggested in sheth et . al ( 2001 ) , does not offer a consistent description of the halo mah and mr , though , the limiting `` square - root '' moving barrier ( @xmath175 ) suggested by moreno et al . ( 2008 ) in principle could offer a solution . the results obtained by the monte carlo method using the _ square - root _ moving barrier show that both the average halo mah and mr depart significantly from those found in the simulations . the mah is now too extended towards the past , revealing a low contribution of major mergers in the halo assembly . the major mrs ( @xmath184 ) are indeed rare with respect to the simulation results . we have explored the obtained effects varying the @xmath185 parameters without any interesting result . nonetheless , the most serious difficulty in our formalism when using a mass - dependent critical overdensity is that the convergence of the monte carlo algorithm , when redshift step @xmath97 , does not accomplish anymore . it is not surprising that some rough agreement with the results from simulations is obtained only when @xmath186 , owing to the `` correct '' behavior of @xmath105 as a function of the progenitor mass in this case ( see below ) . indeed , by using the large @xmath186 time step , moreno et al . ( 2008 ) have showed that the moving barrier predictions agree with the simulation conditional probability . simple considerations help us to understand the failure of the mass - dependent critical overdensity in our scheme . figure [ gamplot ] allows to understand the reduction of the major merger rate . for a fixed barrier ( @xmath1 independent of mass ; dashed lines ) , @xmath105 increases monotonically with mass and eq . ( [ psgen ] ) leads to a conditional probability that makes possible the existence of low - mass progenitors . for the `` square - root '' moving barrier ( moreno et al . 2008 ; solid lines ) , @xmath105 shows a decreasing shape on a very extended low - mass range of progenitors ; here the conditional probability eq . ( [ psgen ] ) is physically meaningless and no low mass progenitors are possible . only in a small range immediately below the descendant mass , @xmath105 increases with mass , allowing at this range for a physical conditional probability , such that the progenitors are more massive than in the fixed barrier case . when @xmath187 the `` correct '' monotonic increasing of @xmath105 with the progenitor mass is recovered . furthermore , a necessary condition to obtain a redshift - step convergence is that @xmath125 when @xmath97 with @xmath188 . this tendency is not fulfilled for the moving barrier case . we conclude that , _ when using a mass - dependent critical overdensity ( moving barrier ) in our approach , the halo mf is correctly described but the predicted halo mah and mr fail in reproducing the simulation results . _ an exhaustive analysis of this shortcoming and the reason behind it is deserved for a future work . in view of the difficulty of proposing a random fluctuation field described by an analytic cumulative field conditional probability function and being able to generate at the same time the halo mf , mah , and mr obtained in the numerical simulations , we turn on to an inductive ( semi - empirical ) approach : the probability function will be inferred numerically from the tinker08 halo mf , which was obtained from simulations . we split the problem into two parts . firstly , in 4.1 we extend the unconditional probability corresponding to the tinker08 mf ( defined only above @xmath189 ) down to small masses by an algebraic extrapolation , taking care that such an extrapolation does not imply a sharp change in the functionality ; furthermore , we introduce a mass rescaling in order to take into account any halo mass defect due to some diffuse matter ( see 2.2 ) . secondly , in 4.2 , based on the properties of eq . ( [ fundeqa ] ) , we extend the definition of the unconditional probability to the conditional probability , and using it within the context of the cf , we calculate the halo mah and mrs . tinker08 provided an analytical fit to the halo mfs measured in numerical simulations at different redshifts and above a minimum ( resolution ) mass . we can calculate from these mfs the cumulative unconditional probability functions expressed in terms of the commonly used scaled variable @xmath122 or , what is the same , @xmath190 , where @xmath191 is the same as in eq . ( [ gamma ] ) but applied to the entire universe , i.e. for @xmath192 ( @xmath193 ) and @xmath194 . due to the lower limit in mass in the tinker08 mfs , the cumulative unconditional probability should be constructed by integrating the mf from each mass @xmath195 at a given redshift , corresponding to the argument @xmath191 , to infinity . we define the following function : @xmath196 where @xmath197 , given by eq . ( [ cmf ] ) , is the halo mf , and the mass @xmath198 is an intrinsic function of @xmath191 for a given @xmath89 . as long as the integral in eq . ( [ tpro ] ) is correctly normalized , @xmath199 represents the cumulative probability to find an isolated region with overdensity @xmath90 ( collapsed region ) and mass @xmath200 ( or @xmath191 less than the value corresponding to @xmath195 for the given @xmath89 ) . when we apply eq . ( [ psans ] ) to the gaussian ( ps ) case , we obtain @xmath201 . since in such case @xmath202 is function of @xmath191 only , the dependence of @xmath203 on @xmath89 vanishes ( @xmath204 depends on @xmath89 but through @xmath191 ) ; this is the well known fact that for a gaussian field , the halo multiplicity function is universal when expressed in terms of the scaled variable @xmath122 ( or @xmath191 ) . vs. @xmath191 ( eq . [ tpro ] ) for the tinker08 mf at @xmath117 ( solid black lines ) . in each panel the virial mass in the mf is rescaled by the factor shown by the label inside . the dashed black lines show a cubic extrapolation down to the origin for @xmath113 , fitting the low limit of the cumulative probability up to the second derivative . the ps case is displayed in each panel just for visual reference ( red dot dashed line , see text ) ] by using the tinker08 halo mf in the first equality of eq . ( [ tpro ] ) , the cumulative probability function @xmath199 can be obtained numerically . the result is shown in the top left panel of fig . [ cpf ] down to the limit mass at each epoch ( solid black lines for @xmath117 ) . the corresponding result for the gaussian ( ps ) case , with @xmath1 rescaled by a factor @xmath120 in order to fit the high - mass end of the tinker08 mf , is also plotted ( red dot - dashed curves ) . it is interesting to note that for the tinker08 mf , the dependence of @xmath205 on @xmath89 is negligible , which suggests that @xmath205 as inferred from the simulations deviates only slightly from a universal function . in what follows , we will consider that the function @xmath205 is the same at all redshifts , omitting the explicit argument @xmath89 . in order to make @xmath205 physically meaningful as a cumulative probability , we need to know its entire profile , i.e. down to @xmath206 ( @xmath207 ) . then , we proceed first by smoothly extrapolating the tinker08 @xmath205 applying a cubic polynomial calculated with the conditions to be @xmath56 when @xmath208 , and fitting the low limit of the numerical curve up to the second derivative . the result is shown by the dashed black curve in the left upper panel of fig . the plot highlights the difference between the tinker08 and the ps cases in the @xmath205 shape when @xmath209 . while the latter reaches the origin following a straight line , the first one shows a curved shape . the physical reason of such difference can be due to the lack of normalization in the tinker08 mf ; such a lack of normalization can arise from the presence of diffuse matter in simulations unaccounted in the mf ( see below ) . in the following , we discuss critical issues that appear when a confrontation between analytical approaches ( e.g. , es and cf ) and the n - body simulations is carried out . then , taking into account these issues , we propose an economical strategy that makes compatible for the cf the , the cumulative probability @xmath210 ) inferred from simulations and extrapolated to low masses . in the analytic formalisms , the linear overdense region of mass @xmath17 is linked to a collapsed halo of the same mass ( mass conservation ; see discussion in 2.2 ) , and the progenitor distributions at each redshift take into account _ all the available mass as part of collapsed halos_. the question is whether the simulations account for the mass in the same way . first , in the simulations the range of halo masses is limited ; there is no information below the halo mass resolution limit . if one extrapolates the fitted halo mf to lower masses and integrate the obtained multiplicity function from 0 to infinity , then one finds a deficit . this implies that the fitted mf should change its slope ( be steeper ) at smaller masses , and/or that some fraction of the mass is actually not in the virialized halos . the latter brings to consideration a second issue : in the simulations a non - negligible fraction of mass is indeed _ diffuse _ ( not in halos ) . the presence of diffuse matter is due to several reasons : 1 . a halo is counted only if it contains tens of particles ; groups with less particles are part of what is considered as diffuse matter . several authors have shown that a significant fraction of the gravitationally bounded particles are actually further away from the spherical virial radius ( e.g. , prada et al . 2006 ; cuesta et al . 2008 ; lacerna & padilla 2011 ; anderhalden & diemand 2011 ) . it is not easy to determine the halo radius that contains all the bounded particles , and likely this radius presents a large variation from case to case , depending on environment , epoch , previous assembly history , mass , etc . let us call @xmath211 the average mass fraction of matter gravitationally bounded to halos but not accounted in the conventional spherical virial mass . 3 . due to true dynamical processes , e.g. , when halos collide , some fraction of the particles are ejected from the merged system ( wang et al . 2011 ) ; let @xmath212 be the overall fraction of such ejected mass . + the diffuse mass produced by all of these effects may eventually infall on to the growing halos . therefore , at difference of the analytical formalisms , the mass growth of halos in simulations happens also in the form of diffuse accretion , which is expected to reduce the minor merger rates as compared to the analytical formalisms . in the current simulations , at least 30% of the @xmath113 halo masses came in diffuse accretion ( c.f . genel et al . 2011 ; wang et al . it was suggested also a background of diffuse matter , remnant of the cut - off in the mass power spectrum due to the relativistic free - streaming damping ( e.g. , @xmath213 for the neutralino ) ; at very high redshifts , most of this matter due to the cut - off seems to be diffuse ( angulo & white 2010 ) , but at the redshifts of significant halo mass assembly , the effects mentioned above dominate over this . the diffuse matter in simulations implies that the mass originally linked to a given overdensity is less than that defined inside the spherical virial radius . due to these effects , the conventional virial masses deplete on average by the fraction @xmath214 . now , considering the issues discussed above and that in the cf the diffuse matter is not taken into account , we need to make as compatible as possible the cumulative probability @xmath210 ) inferred from simulations with the one ( correctly normalized from 0 to infinity ) to be used in the cf . in the case of item ( ii ) above , the solution is simple : a mass rescaling allows us to recover the mass inside the collapsed halo proper of the analytical formalism . with this correction the analytic approach is justified . the ejected matter ( item iii ) may be considered a simple extension of this approximation . after several experiments , we find that an economical way is using a strategy based on the following ideas : @xmath164 ) a mass rescaling allows to pass from a model ( tinker08 ) @xmath205 to another @xmath215 , where the complications of the ejected matter and incorrect halo mass definition in simulations are avoided , in such a way that the masses are now roughly compatible with the ones of the analytic case and @xmath215 can be handled with the cf ; @xmath165 ) a good empirical criterion is to choose @xmath215 with a reduced curvature when extrapolated to low masses in agreement with the shape of @xmath216 . finally the choice of the mass rescaling factor is justified by the result , i.e. the mahs and mrs being in agreement with simulations . following this strategy , we explore now the effects of a simple constant mass rescaling of the tinker08 halo mf on @xmath215 . the panels of fig . [ cpf ] show the cases of virial mass rescaling by factors @xmath217 , and @xmath218 , respectively . it is noteworthy that for @xmath219 , the extrapolation of @xmath215 to low @xmath191 values is close to a straight line and , in general , this probability function approaches to @xmath216 . moreover , from the papers mentioned above , it can be said that an average halo mass correction by @xmath220 ( assumed to be mass independent ) is within the uncertainties . we conclude that this exercise , changing the shape of @xmath205 obtained from simulations so that it is closer to the shape of the analytic ( ps ) gaussian probability , implies a mass rescaling of @xmath221 , in rough agreement with measures of gravitationally bounded halo mass and diffuse matter considerations in simulations . we do not attempt to follow a more rigorous exercise , for instance introducing a mass dependence in @xmath162 , because of the large uncertainties involved in the simulation analysis , and because at this point we require only an indicative mass correction that encompasses all the complexity implied in passing from the linear overdensity to a virialized structure . the function @xmath205 given by eq . ( [ tpro ] ) and obtained from the tinker08 halo mf is the cumulative unconditional probability to find an isolated region with overdensity @xmath1 and mass @xmath222 inside the entire universe , being in this case the _ cosmic _ @xmath223 . now , in order to apply the cf , we need to pass from such unconditional probability to a conditional probability . taking into account the generalized ps ansatz given by eq . ( [ ans ] ) and according to the definition of @xmath105 in eq . ( [ gamma ] ) , the simplest random field compatible with the previous reasoning is @xmath224 in this way , even if the statistics is not any more gaussian , eq . ( [ ans ] ) by construction is satisfied and the involved mathematics is simple . from fig . [ cpf ] one can see that our inferred p(@xmath105 ) from the tinker08 mfs is different from the one corresponding to a gaussian field ( see for a similar conclusion cole et al . 2008 ; neistein et al . 2010 ) . the probability @xmath225 to find an isolated region with overdensity @xmath13 and mass between @xmath15 and @xmath226 inside an isolated region with overdensity @xmath5 and mass @xmath17 , according to eqs . ( [ psans ] ) and ( [ nongauss ] ) , is : @xmath227 we can apply now the monte carlo method to find the halo mah and mrs according to the procedure defined in 2.2 . for @xmath137 monte carlo realizations , we calculate the average mahs and mrs by using the cumulative probability @xmath228 inferred from the tinker08 mf ( without any mass rescaling , i.e. @xmath229 ) . the left panels of fig . [ tmah ] show the average mah ( upper panel , solid line ) and mrs ( lower panel , the same five threshold ratios @xmath147 as in fig . [ psmefr ] ) . the corresponding average mah from the millennium simulations ( fakhouri et al . 2010 ) is shown with the dashed line . the predicted mrs are higher at all epochs , but more at lower redshifts , than those measured in the simulations by fakhouri & ma ( 2008 ) and fakhouri et al . ( 2010 ; thick ticks in the @xmath230 axis correspond to the @xmath113 mrs as reported by these authors ) . this produces also a too fast mass growth of haloes as seen in the upper panel . on the ground of these results , we conclude that the cumulative probability inferred from the original tinker08 halo mf leads to halo mahs and mrs in disagreement with those measured in numerical simulations . in the right panels of fig . [ tmah ] we present the same as in the left panels but now for @xmath228 inferred from the tinker08 halo mf with the virial mass rescaled by the factors @xmath231 , @xmath232 , and @xmath233 ( see fig . [ cpf ] ) . the solid curves are for @xmath234 and the left / right ( upper / lower ) curves are for @xmath235 in the top ( bottom ) panel . the predicted average halo mah and mrs are now close to those measured in the simulations ( certainly within the scatter and uncertainties ) for the case @xmath234 . it is remarkable that the value @xmath234 is close to that one which provides a @xmath215 with a shape similar to the shape of @xmath216 according to the discussion in these two findings are completely independent between them . besides , the value @xmath236 seems to be consistent with measures of the diffuse matter component in the simulations _ ( see for references 4.1 ) . note that the mah and mrs measured from the simulations could change if the virial mass is rescaled ; however , we do not expect significant changes because in the case of the mah and mr the physics has to do with _ mass ratios_. for completeness , we have applied the same procedure to the fitting mf given by sheth & tormen ( 1999 ) . we have obtained practically the same results as for the tinker08 mf , which for economy we do not repeat here . summarizing , through a mass rescaling that makes compatible the masses between simulations and the analytic framework , we have derived from the halo tinker08 mf a conditional field probability function given by eq . ( [ nongauss ] ) particularly handy in the sense that it is compatible with the ps ansatz ( eq . [ ans ] ) , in spite of it deviates from the one of a gaussian density fluctuation field . starting from such probability , our cf allows to calculate the halo mahs and mrs . the obtained average mahs and mrs are consistent with numerical simulations when the virial mass in the input tinker08 halo mf is rescaled . for economy , we plotted results regarding the average mah and mrs only for one descendant mass , @xmath237 . the conclusions are similar for other masses . figure [ f - m ] shows the value of @xmath162 required to agree with the fakhouri et al . ( 2010 ) average mahs at the redshift one - half ( solid line ) and one - tenth ( dashed line ) of the mass for @xmath113 descendants of @xmath238 and @xmath239 . as is seen , the mass rescaling factor is roughly the same for masses below @xmath240 , something in between 1.30 and 1.34 , in order to agree with the mahs down to one - tenth of the @xmath113 descendant mass . for larger masses , this factor should be slightly larger . we conclude that , taking into account a mass rescaling of @xmath232 , the cf allows for a reasonable good description of the halo mfs , and the average mahs and mrs of halos @xmath241 obtained in numerical simulations . probably , a better tuning of this description can be attained by varying @xmath162 with mass , however , this task becomes difficult due to by subtle aspects , for example , the way the averaging is carried out when calculating the average mahs and mrs . we deserve such an exploration for future works . the level of normalization on the tinker08 mf influences our result . suppose that @xmath242 is a correction factor on the mf . the model establishes that @xmath243 , that is a renormalization on the mf produces the same effect of a mass rescaling . if @xmath244 , no mass rescaling should be necessary at all . a defect in the mf normalization and a defect in the halo mass are related aspects of a same problem . if the diffuse matter plays the role that we argued in our reasonings above , then the tinker08 mf does not include all the dark matter in the cosmological boxes , and consequently it is not normalizable . tinker08 indeed acknowledge that this is the case . , required to agree with the average mah from fakhouri et al . ( 2010 ) at the redshift one - half ( solid line ) and one - tenth ( dashed line ) obtained for a wide range of present - day halo masses . ] we develop an approach able to connect the overdensities of the density fluctuation field with the abundance and assembly history of the virialized haloes . starting from a very general basis , our _ conditional formalism _ ( cf ) concerns the inventory of the isolated overdense regions given the field conditional probability function of the density fluctuation field . this formalism allows us to calculate at the same time and in a simple way : the halo mass function mf at any @xmath89 , the halo mass aggregation histories mahs , and the halo merger rate histories mrs . after that , our main goal is to identify a strategy that allows us to use the cf for describing optimally the numerical simulation results regarding these three halo statistical and evolutionary features . in the way of attaining this goal , we arrived to some important conclusions : * for the gaussian density field ( the generalized ps ansatz , eq . ( [ ans ] ) , applies ) , the predicted average halo mah and mrs agree roughly with those measured in current cosmological ( @xmath3cdm ) simulations , but the halo mf , as previously found , is overabundant at intermediate masses and deficient at high masses as compared to simulations , e.g. , with the tinker08 mf . * the introduction of @xmath1 depending on mass ( moving barrier ; sheth et al . 2001 ) instead of the constant one , improves the halo mf as compared to simulations , but dramatically spoils the average mah and mrs . nonetheless , the major problem with the mass - dependent @xmath1 is the excessive sensitivity of the mah and mrs to the redshift step used in the monte carlo merger tree construction . * a cumulative field unconditional probability function can be inferred from the halo mf measured in simulations ( tinker08 ) , being this almost the same ( universal ) at different redshifts when expressed in terms of @xmath245 ; its definition has to be complemented by adequately extrapolating it to low masses . this inference is particularly suitable regarding the shape when the halo virial mass is rescaled by @xmath236 . remarkably , this rescaling makes the halo mass compatible with the mass used in the analytical formalisms having in mind that in simulations not all the mass is in halos . on the ground of the cf , the conditional probability function is obtained from the unconditional one by making the former compatible with the ps ansatz , in spite of that the shape of the latter deviates from the one corresponding to a gaussian statistics . such a situation allows us to easily use the cf and our monte carlo algorithm for calculating the halo mahs and mrs . the obtained average mahs and mrs depend critically on the mass rescaling factor . if this factor , assumed constant , is @xmath236 , then the agreement with the average mah and mrs in simulations becomes remarkable , at least for descendant halos up to @xmath240 . it is encouraging that the rescaling of the virial mass in the tinker08 halo mf by a factor @xmath236 allows both for a well behaved overall cumulative field unconditional probability function and for halo mahs and mrs in agreement with simulations . it is important to remark that the value of @xmath162 depends on the resolution limit of numerical simulations . * dark matter in simulations exhibits a rather complex structure . not all the mass is counted in the halos defined up to the conventional virial radius , but a diffuse component is present because of at least three reasons : a ) the mass resolution limit on halo identification , b ) the not counting of gravitationally bounded particles that are further away the virial radius , and c ) the mass ejection generated by mergers . the contribution a ) has to do with the low mass extension of the mf . due to b ) and c ) the mass accounted in the virialized halo is smaller than in the original overdense region , producing this a normalization defect in the mf . the disagreement between the simulation and analytic mfs seems to lie on these last two effects as well as on a departure from the gaussian distribution of the density fluctuations . the conditional probability function inferred from simulations with our approach and its use in the cf allow for an estimate of about @xmath246 for the diffuse mass fraction related to effects b ) and c ) , and show that the conditional probability corresponds to a density fluctuation field ( slightly ) deviated from gaussianity . in agreement with the previous argument , the tinker08 mf does not include all the dark matter present in the cosmological boxes and consequently it is not normalizable . a consistent description of the cosmological simulations results regarding the halo mf , mahs , and mrs through an analytical ( statistical ) formalism has been achieved . the interesting finding is that such a description could not be attained for a gaussian density fluctuation field , as well as without taking into account the complex distribution of dark matter in simulations . in this sense , our results urge for a revision of two questions : 1 . _ the correct accounting in simulations of the mass in virialized halos and of diffuse matter . _ in the analytical formalisms there is an exact equality between the mass of a linear overdensity and that of the collapsed halo used for the counting in the mf , while in the numerical simulations this is not the case because of effects b ) and c ) above mentioned . therefore , when confronting the halo mf calculated in the analytical formalisms with that one measured in simulations , these issues should be taken into account . our approach is a first approximation to this complex problem . further exploration is necessary . 2 . _ the correct introduction of the statistics in analytical formalisms . _ the hierarchical structure formation implies that in the same density field coexist collapsed structures at small scales with regions in the linear regime at large scales . then , even if the primordial statistics is gaussian , at the time the analytical approach is applied the density field at the scales of interest could already have deviated from gaussian initial conditions ( e.g. , coles & jones 1991 ) , as numerical simulations suggest . in any case , the cf and the heuristic approach presented here , allow us to overcome some of the difficulties related to these questions , and attain a consistent description of the halo mf , mahs and mrs measured in cosmological numerical simulations . we thank the anonymous referee for his / her comments and suggestions which helped to improve this paper . c. f. is grateful to dr . flavio firmani from the university of victoria ( bc , ca ) for the encouraging support to carry out this work . v. a. acknowledges conacyt grant 167332-f for partial funding .

a hamiltonian decomposition of is a partition of its edge set into disjoint hamilton cycles . manikandan and paulraja conjectured that if and are hamilton cycle decomposable circulant graphs with at least one of them is nonbipartite , then their tensor product is hamilton cycle decomposable . in this paper , we have proved that , if is a hamilton cycle decomposable circulant graph with certain properties and is a hamilton cycle decomposable multigraph , then their tensor product is hamilton cycle decomposable . in particular , tensor products of certain sparse hamilton cycle decomposable circulant graphs are hamilton cycle decomposable .

1703.03148

all graphs considered here are simple and finite , unless otherwise stated . let @xmath2 denote the cycle ( respectively path ) on @xmath3 vertices . the complete graph on @xmath4 vertices is denoted by @xmath5 and its complement is denoted by @xmath6 for two graphs @xmath0 and @xmath1 their _ tensor product _ , denoted by @xmath7 has vertex set @xmath8 in which two vertices @xmath9 are adjacent whenever @xmath10 the _ wreath product _ of the graphs @xmath0 and @xmath11 denoted by @xmath12 has vertex set @xmath8 in which @xmath13 is an edge whenever @xmath14 is an edge in @xmath15 or @xmath16 and @xmath17 is an edge in @xmath18 similarly , the _ cartesian product _ of the graphs @xmath0 and @xmath11 denoted by @xmath19 has vertex set @xmath8 in which @xmath13 is an edge whenever @xmath16 and @xmath17 is an edge in @xmath11 or @xmath20 and @xmath14 is an edge in @xmath21 the subgraph induced by @xmath22 is denoted by @xmath23 similarly , the subgraph induced by @xmath24 is denoted by @xmath25 for a graph @xmath15 @xmath26 is the graph obtained from @xmath0 by replacing each edge of @xmath0 by @xmath27 parallel edges . for a graph @xmath15 @xmath28 is the symmetric digraph of @xmath21 for two loopless multigraphs @xmath26 and @xmath29 their tensor product , denoted by @xmath30 has the vertex set @xmath8 and its edge set is described as follows : if @xmath31 is an edge of multiplicity @xmath27 in @xmath26 and @xmath32 is an edge of multiplicity @xmath33 in @xmath29 then corresponding to these edges there are edges @xmath34 and @xmath35 each of multiplicity @xmath36 in @xmath37 and @xmath37 is isomorphic to @xmath38 hence @xmath39 if @xmath40 are edge - disjoint subgraphs of @xmath0 and @xmath41 then we write @xmath42 for a graph @xmath15 if @xmath43 can be partitioned into @xmath44@xmath45@xmath46 such that @xmath47 for all @xmath48 then we say that @xmath1 decomposes @xmath15 or _ @xmath1-decomposition _ of @xmath0 exists . clearly , the tensor product is commutative and distributive over edge - disjoint union of graphs , that is , if @xmath49 then @xmath50 let @xmath0 be a finite group and let @xmath51 be a symmetric subset of @xmath0 ( that is , @xmath52 the vertices of the _ cayley graph , _ @xmath53 are the elements of @xmath0 and there is an edge between @xmath54 and @xmath55 if and only if @xmath56 note that @xmath57 is connected if and only if @xmath51 generates the group @xmath21 a _ circulant _ @xmath58 is a graph with vertex set @xmath59 and edge set @xmath60 where @xmath61 and @xmath62 is the set of integers modulo @xmath63 the elements of @xmath64 are called _ jumps . _ clearly , every circulant graph of order @xmath4 is a cayley graph with the underlying group being @xmath65 the problem of finding hamilton cycle decompositions of product graphs is not new . hamilton cycle decompositions of various product graphs , including digraphs , have been studied by many authors ; see , for example , . it has been conjectured @xcite that if both @xmath0 and @xmath1 are hamilton cycle decomposable graphs , then @xmath66 is hamilton cycle decomposable , where @xmath67 denotes the cartesian product of graphs . this conjecture has been verified to be true for a large classes of graphs @xcite . baranyai and szsz @xcite proved that if both @xmath0 and @xmath1 are even regular hamilton cycle decomposable graphs , then @xmath68 is hamilton cycle decomposable . in @xcite , ng has obtained a partial solution to the following conjecture of alspach et al . @xcite : if @xmath69 and @xmath70 are directed hamilton cycle decomposable digraphs , then @xmath71 is directed hamilton cycle decomposable . jha @xcite conjectured the following _ @xcite_[ijpam.23.723 - 729 ] if both @xmath0 and @xmath1 are hamilton cycle decomposable graphs and @xmath72 is connected , then @xmath72 is hamilton cycle decomposable . conjecture [ ijpam.23.723 - 729 ] was disproved , see @xcite . because of this , finding hamilton cycle decompositions of the tensor products of hamilton cycle decomposable graphs is considered to be difficult . though conjecture [ ijpam.23.723 - 729 ] has been disproved , we believe that if the graphs @xmath0 and @xmath1 are suitably chosen , that is , with some suitable conditions imposed on them , then @xmath72 may have hamilton cycle decomposition . in @xcite and @xcite it has been proved that @xmath73 and @xmath74 are hamilton cycle decomposable ; in @xcite it is shown that the tensor product of two regular complete multipartite graphs is hamilton cycle decomposable . hamilton cycle decompositions of the tensor products of complete bipartite graphs and complete multipartite graphs are dealt with in @xcite . also in @xcite , paulraja and sivasankar proved that @xmath75 and @xmath76 are directed hamilton cycle decomposable . it can be observed that @xmath77 are circulant graphs . based on the results of , manikandan and paulraja conjectured the following : _ ( manikandan and paulraja ) @xcite._[con.rsmpp ] if @xmath0 and @xmath1 are hamilton cycle decomposable circulant graphs and at least one of them is non bipartite , then @xmath72 is hamilton cycle decomposable . _ @xcite_[ann.dm.3.21 - 28 ] if both @xmath0 and @xmath1 have hamilton cycle decompositions and at least one of them is of odd order , then @xmath72 admits a hamilton cycle decomposition.@xmath67 one may naturally ask when is @xmath72 hamilton cycle decomposable , if both @xmath0 and @xmath1 are of even order . in this paper it is partially answered . we say that an even regular circulant graph @xmath58 has the _ property @xmath78 _ if ( 1 ) the number of odd and even jumps in @xmath64 are equal ; ( 2 ) odd jumps can be paired with even jumps so that each of the @xmath79 resulting @xmath80-regular graphs is connected . it is known that every @xmath80-regular connected circulant graph is hamilton cycle decomposable ; see @xcite . consequently , every circulant graph with property @xmath81 is hamilton cycle decomposable . here we prove the following main theorem . [ thm.cir.1 ] let @xmath0 be a circulant graph with property @xmath81 and let @xmath1 be any hamilton cycle decomposable multigraph , then @xmath72 is hamilton cycle decomposable . this theorem has many interesting consequences . in particular , we have the following corollary . [ cor.cir.1 ] if @xmath0 and @xmath1 are even regular hamilton cycle decomposable circulant graphs and at least one of them has the property @xmath78 then @xmath72 is hamilton cycle decomposable . one of the consequences of corollary [ cor.cir.1 ] is that if @xmath82 and @xmath83 where @xmath84 and @xmath85 are @xmath86-factors of @xmath87 and @xmath88 respectively , then @xmath72 is hamilton cycle decomposable . in particular , after deleting suitable number of jumps of @xmath87 and @xmath88 the resulting graphs need not be dense but their tensor product is hamilton cycle decomposable . this can not be deduced from the existing results in this direction . first we present the necessary definitions here . the notation that we use here are from @xcite and for the sake of completeness we give them . the cayley graph @xmath89 where @xmath90 is a generating set of the finite abelian group with @xmath91 and @xmath92 is a simple connected graph . we call the edge @xmath93 of @xmath94 an @xmath95-edge ; the subgraph formed by the @xmath95-edges is a disjoint union of cycles , called @xmath95-cycles , each of length @xmath96 the order of the element @xmath97 similarly , we define @xmath98-edges and @xmath98-cycles ; each @xmath98-cycle is of length @xmath99 the order of the element @xmath100 in @xmath101 let us denote the number of @xmath95-cycles by @xmath102 and the number of @xmath98-cycles by @xmath103 since the length of each @xmath95-cycle ( respectively @xmath98-cycle ) is @xmath104 ( respectively @xmath105 ) @xmath106 ( respectively @xmath107 ) . as the graph @xmath94 is connected , the vertices @xmath108 are in the @xmath102 @xmath95-cycles denoted by @xmath109 respectively , and @xmath110 belongs to @xmath111 see @xcite . hence , we have @xmath112 for some @xmath113 with @xmath114 further , every vertex @xmath54 of @xmath94 can be uniquely written as @xmath115 with @xmath116 and @xmath117 using this uniqueness , label the vertices of @xmath94 as @xmath118 with @xmath119 and @xmath120 where the first coordinate indicates the label of the cycle @xmath121 containing the vertex and the second coordinate indicates the position of the vertex on the cycle . _ @xcite_[jctb.46.142 - 153 ] let @xmath123 denote the class of simple graphs on @xmath124 vertices , where @xmath125 and @xmath126 the @xmath124 vertices of the graph can be labeled @xmath118 with @xmath127 taken modulo @xmath102 and @xmath128 taken modulo @xmath129 the edges are _ ( 1 ) _ first kind : @xmath130 and _ ( 2 ) _ second kind : @xmath131 for all @xmath132 and @xmath133 a graph @xmath134 has @xmath102 vertical " disjoint cycles @xmath135 with a natural orientation and @xmath136 horizontal _ parallel matchings _ between the cycles @xmath121 and @xmath137 for @xmath138 and a particular _ parallel matching _ between @xmath139 and @xmath140 ( which depends on the value of @xmath141 a graph @xmath142 consists of two cycles plus two perfect matchings between them . a graph @xmath143 consists of a cycle plus the chords joining @xmath144 to @xmath145 observe that @xmath123 and @xmath146 are isomorphic classes of graphs , that is , an element @xmath147 is isomorphic to an element @xmath148 and vice versa , for example see figure 1 . graphs of figures @xmath149 and @xmath150 are isomorphic . + @xmath151 graph @xmath152 where the edges of jump @xmath153 give @xmath153 disjoint cycles each of length @xmath154 and the edges of jump @xmath155 give @xmath155 disjoint cycles each of length @xmath156 + @xmath157 @xmath0 is an element of @xmath158 with @xmath159 + @xmath160 @xmath0 in @xmath157 is drawn as an element of @xmath161 with @xmath162 _ @xcite_[jctb.46.142 - 153-defn ] a hamilton cycle decomposition of a graph @xmath134 has the property @xmath163 if both hamilton cycles use at least one edge of the matching between @xmath121 and @xmath164 for all @xmath165 where @xmath166 _ @xcite_[jctb.46.142 - 153 - 1 ] the class @xmath123 consists of the @xmath80-regular connected cayley graphs on a finite abelian group with a generating set @xmath167 where @xmath102 is the number of @xmath95-cycles and @xmath168 is the number of @xmath98-cycles . @xmath67 if @xmath0 and @xmath1 are hamilton cycle decomposable graphs with at least one of them having odd order , then @xmath72 is hamilton cycle decomposable , by theorem [ ann.dm.3.21-28 ] ; hence in what follows , * we assume that both @xmath0 and @xmath1 are hamilton cycle decomposable graphs each having an even number of vertices . * if @xmath0 is a connected @xmath80-regular circulant graph of even order with generating set @xmath169 where @xmath95 and @xmath98 are of different parity such that @xmath91 and @xmath92 then by the definition [ jctb.46.142 - 153 ] , @xmath102 and @xmath168 are of different parity . for our convenience , according to the situation we consider the parity of @xmath95 and @xmath100 * in what follows , we assume that @xmath123 denotes the class of cayley graphs on the abelian group @xmath62 * ( that is , circulants ) with generating set @xmath169 and by theorem [ jctb.46.142 - 153 - 1 ] , @xmath123 consists of the class of @xmath80-regular connected circulants of order @xmath4 with generating set @xmath167 where @xmath102 is the number of @xmath95-cycles in a @xmath153-factor and @xmath168 is the number of @xmath98-cycles . in the rest of the paper , * we assume that @xmath134 is a @xmath80-regular circulant graph of even order with generating set @xmath169 where @xmath95 and @xmath98 are of different parity ; consequently , @xmath102 and @xmath168 are of different parity . * [ defn.jha ] let @xmath170 be a hamilton cycle decomposition of a @xmath80-regular graph @xmath0 of even order with at least @xmath154 vertices . if @xmath0 contains a @xmath80-cycle @xmath171 such that the edges @xmath172 and @xmath173 belong to one of the two hamilton cycles and the edges @xmath174 and @xmath175 are on the other hamilton cycle , then the @xmath80-cycle is said to be an _ alternating _ @xmath80-cycle in @xmath15 with respect to @xmath176 and @xmath177 further , if the vertices @xmath95 and @xmath113 _ _ ( _ _ or @xmath98 and @xmath178 _ _ ) _ _ are at an odd distance along each of the two hamilton cycles @xmath179 then the @xmath80-cycle @xmath171 is said to be an odd alternating @xmath80-cycle in @xmath0 with respect to @xmath180 we say that the graph @xmath0 has _ property @xmath181 _ if @xmath0 contains an odd alternating @xmath80-cycle with respect to a hamilton cycle decomposition of it . the _ reduced graph @xmath182 with respect to @xmath183 is defined as follows : delete the vertices of the cycles @xmath184 and @xmath185 of @xmath101 that is , the vertices @xmath186 and @xmath187 and add the edges @xmath188 joining @xmath139 and @xmath111 where @xmath113 is given in the definition of @xmath189 in some cases @xmath190 might be a multigraph and , if it is a simple graph , it is an element of @xmath123 ( @xmath3 and @xmath113 being unchanged ) . successive reductions of the resulting graphs result in a graph @xmath191 for some @xmath192 @xmath191 is also called a reduced graph of @xmath193 the graph @xmath194 is said to be _ lift graph _ of @xmath195 if @xmath94 is obtained as follows : delete the edges @xmath196 joining @xmath197 and @xmath140 of @xmath198 then add two @xmath95-cycles @xmath199 and @xmath200 with vertices @xmath201 and @xmath202 to @xmath203 and add the edges @xmath204 where @xmath113 is as given in the definition of @xmath205 successive liftings of the resulting graphs result in a graph @xmath206 for some @xmath192 @xmath206 is also called a lift graph of @xmath207 * proof . * let @xmath214 since @xmath215 and @xmath216 without loss of generality assume that @xmath217 and @xmath218 consequently , @xmath95 is even and @xmath98 is odd as @xmath219 and @xmath220 clearly , @xmath221 ( otherwise @xmath222 the existence of hamilton cycles @xmath176 and @xmath223 in @xmath94 described below is in @xcite . a hamilton cycle @xmath224 is obtained by deleting the edges @xmath225 and @xmath226 from @xmath140 and @xmath227 respectively , and adding the edges @xmath228 and @xmath229 connecting these two @xmath95-cycles , that is , @xmath230 see figure 2(b ) ; @xmath231 is shown in the solid lines of figure 2(c1 ) . clearly , these two hamilton cycles use both @xmath95 and @xmath98-edges and hence the property @xmath212 holds . a member @xmath94 of @xmath232 + ( b ) . hamilton cycle @xmath176 of @xmath233 + ( c1 ) . the solid lines are edges on @xmath177 other than the two @xmath95-edges @xmath225 and @xmath234 all edges of @xmath223 ( and thus all edges connecting @xmath140 and @xmath235 are @xmath98-edges . + ( c2 ) . another drawing of the hamilton cycle @xmath223 of @xmath236 since @xmath238 there are two @xmath95-cycles @xmath140 and @xmath239 and there are two parallel matchings between @xmath140 and @xmath240 we denote one of the matchings from @xmath140 to @xmath239 as @xmath241 and the other matching from @xmath239 to @xmath140 as @xmath242 let the edges of @xmath241 be @xmath243 and let the edges of @xmath244 be @xmath245 where addition is taken modulo @xmath63 we call the vertex @xmath246 of @xmath140 as the _ corresponding vertex _ of @xmath247 of @xmath239 and vice versa . we prove the existence of the property @xmath237 with respect to the hamilton cycle decomposition @xmath170 of @xmath94 by the parity of @xmath248 first , we shall obtain an even length path along @xmath250 so that this path is a section of a suitable odd length path along @xmath176 having its ends as two opposite vertices " of an odd alternating @xmath80-cycle in @xmath251 the existence of the odd alternating @xmath80-cycle will be proved later . in @xmath252 below , we obtain a required even length path along @xmath176 and in @xmath253 we prove that the path obtained in @xmath252 is a section of a path along @xmath176 joining two opposite vertices " of an odd alternating @xmath80-cycle . * * first we consider @xmath113 to be even and consider the hamilton cycle @xmath176 of @xmath94 obtained above . since @xmath244 is a perfect matching of @xmath94 and @xmath244 matches the vertices in @xmath239 with vertices in @xmath111 @xmath228 is an edge of @xmath254 which is one of the two @xmath244 edges in @xmath250 see figure 3(a ) . clearly , @xmath255 that is , @xmath256 is the @xmath257th vertex of @xmath227 starting from @xmath258 see figure 3(a ) . as @xmath259 is odd , @xmath257 is even and hence the length of the section of @xmath239 from @xmath98 to @xmath256 ( containing the vertex @xmath260 ) is odd . as the length of the section of @xmath239 from @xmath98 to @xmath256 ( containing the vertex @xmath261 ) is the same as the length of the section of @xmath140 from @xmath95 to @xmath262 ( containing the vertex @xmath263 the length of the path from @xmath95 to @xmath264 along @xmath250 denoted by @xmath265 is odd and let this path be @xmath266 now @xmath267 together with the edge @xmath268 is our required even length path @xmath269 see the bold edges of figure 3(a ) . next we suppose that @xmath113 is odd . as in the above paragraph , the vertex @xmath270 in @xmath140 is matched under @xmath244 with the vertex @xmath271 @xmath272 is even as @xmath259 is odd . thus the length of the section of @xmath239 from @xmath98 to @xmath256 ( containing the vertex @xmath261 ) is even and hence the length of the section @xmath273 of @xmath239 from @xmath98 to @xmath274 ( containing the vertex @xmath275 ) is even ( note that @xmath226 is an edge of @xmath239 which we have deleted for the construction of @xmath176 ) . now @xmath273 together with the edges @xmath276 is our required even length path @xmath277 from the vertex @xmath95 to the vertex @xmath260 along @xmath250 see the bold edges of figure 3(b ) . @xmath278 @xmath176 contains exactly two edges of @xmath244 as shown in the figure 3 . hence except the two @xmath80-cycles of @xmath101 namely , @xmath279 and @xmath280 the other @xmath80-cycles , which have two consecutive vertices of @xmath140 and the corresponding two vertices of @xmath239 constitute alternating @xmath80-cycles ( with respect to @xmath176 and @xmath223 of @xmath94 ) . if @xmath113 is even , it is clear that every @xmath80-cycle containing two consecutive vertices of the section @xmath281 of @xmath140 and the two corresponding consecutive vertices of the section @xmath282 of @xmath239 is an alternating @xmath80-cycle of @xmath94 with respect to @xmath176 and @xmath283 see the figure 3 ( note that @xmath284 is ruled out , for otherwise , it would imply @xmath285 for the following reason : @xmath284 implies @xmath286 and since @xmath229 is an edge of @xmath176 joining @xmath239 and @xmath140 implies @xmath287 which is not the case as we consider @xmath113 is even ) . similarly if @xmath113 is odd , every @xmath80-cycle containing two consecutive vertices in the section @xmath288 of @xmath140 and the two corresponding consecutive vertices in the section @xmath289 of @xmath239 is an alternating @xmath80-cycle of @xmath94 with respect to @xmath176 and @xmath290 note if @xmath291 then @xmath292 so @xmath293 but then @xmath294 that is , @xmath295 which is even , a contradiction . next we prove that the alternating @xmath80-cycles described in the above paragraph using the sections of @xmath140 and @xmath239 satisfy the property that the length of the path along @xmath176 between the opposite vertices of the @xmath80-cycle is of odd length and further this odd length path contains @xmath277 ( described above ) . let @xmath296 be an alternating @xmath80-cycle so that @xmath297 and @xmath298 see figure 4 . the vertices @xmath54 and @xmath299 are the corresponding vertices in @xmath140 and @xmath240 if @xmath113 is even , then the @xmath300-section of @xmath250 contained in @xmath111 and the corresponding @xmath301-section of @xmath176 contained in @xmath239 have the same length , see figure 4(a ) . hence the @xmath302-section of @xmath176 ( containing the vertex @xmath303 ) is of odd length as it contains @xmath269 which is of even length , see figure 4(a ) ; that is @xmath304 is odd when @xmath113 is even . similarly , if @xmath113 is odd then , the @xmath305-section of @xmath176 ( containing the vertex @xmath306 ) , contained in @xmath111 and the corresponding @xmath307-section of @xmath250 contained in @xmath239 ( containing the vertex @xmath308 have the same length . thus the @xmath302-section of @xmath176 ( containing the vertex @xmath309 is of odd length as it contains @xmath269 which is of even length , see figure 4(b ) . that is @xmath304 is odd when @xmath113 is odd , see figure 4(b ) . \(a ) the lengths of the sections @xmath300 along @xmath176 ( section of @xmath140 ) and @xmath310 along @xmath176 ( section of @xmath239 ) are same and @xmath311 is an edge + ( b ) the lengths of the sections @xmath305 along @xmath176 ( part of @xmath140 ) and @xmath307 along @xmath176 ( part of @xmath239 ) are same and @xmath312 is an edge . next we shall obtain an appropriate @xmath54 on @xmath140 so that @xmath313 is also odd which gives the property @xmath237 with respect to @xmath176 and @xmath177 as @xmath223 contains exactly two @xmath95-edges , namely , @xmath226 and @xmath225 ( see figure 2(c2 ) ) , @xmath314 is a pair of odd length paths ( since the vertices of these paths alternate between @xmath315 and @xmath316 ) . let @xmath317 and @xmath318 denote these paths , namely , the @xmath319-section and @xmath320-section of @xmath223 , respectively , see figures 2(c1 ) and 2(c2 ) . if @xmath321 ( respectively @xmath322 ) contains a pair of consecutive vertices @xmath54 and @xmath323 of @xmath140 ( note that the edge @xmath93 is in @xmath176 and @xmath321 is contained in @xmath223 ) , then @xmath313 is odd as the vertices of @xmath321 ( respectively @xmath322 ) alternate between @xmath315 and @xmath316 and @xmath312 is an edge in @xmath177 we shall prove the existence of such a pair of consecutive vertices in the sequence of vertices @xmath324 or in @xmath325 see figure 4 . let @xmath113 be even . from the construction of @xmath176 and @xmath283 the vertices @xmath270 and @xmath264 are in @xmath326 ( see figure 2(c1 ) ) . there are even number of vertices in @xmath327 as @xmath277 is of even length and @xmath140 has odd number of vertices ( see figure 3(a ) ) . if there is no pair of consecutive vertices in @xmath327 of the required type in @xmath321 or @xmath328 then the preceding vertex of @xmath329 namely , @xmath330 and the succeeding vertex of @xmath331 namely , @xmath332 are in @xmath328 as the vertices @xmath270 and @xmath264 are in @xmath333 consequently , there must be an odd number of vertices from @xmath334 to @xmath335 in @xmath336 which is not the case ( since in @xmath336 @xmath334 to @xmath337 contains even number of vertices ) . therefore there must exist a pair of consecutive vertices in @xmath327 of the required type and hence @xmath94 satisfies property @xmath237 with respect to @xmath176 and @xmath177 next we assume that @xmath113 is odd . as above , we shall show the existence of the required pair of consecutive vertices in @xmath338 @xmath339 has even number of vertices ( see figure 3(b ) ) . assume that there is no pair of vertices in @xmath339 of the required type in @xmath321 or @xmath322 and hence alternate vertices of @xmath339 are in @xmath321 and @xmath340 then , as the vertices @xmath95 and @xmath303 are in @xmath328 the vertices @xmath306 and @xmath341 must be in @xmath321 and hence there must be an odd number of vertices from @xmath306 to @xmath341 in @xmath342 which is not the case ( as @xmath339 contains even number of vertices from @xmath306 to @xmath341 ) . thus there must exist a pair of consecutive vertices as required and hence @xmath94 satisfies the property @xmath213 since @xmath344 and @xmath345 @xmath113 is always odd , otherwise , the @xmath98-edges will not induce a hamilton cycle . as @xmath217 and @xmath346 @xmath347 as @xmath348 by assumption @xmath98 is odd and hence @xmath349 then @xmath350 for any two vertices @xmath54 and @xmath55 on @xmath227 the path from @xmath54 to @xmath55 along @xmath239 has even length if and only if @xmath351 hence the length of the section of @xmath239 from @xmath98 to @xmath352 ( containing the vertex @xmath353 is odd and the length of the section of @xmath239 from the vertex @xmath274 to @xmath98 ( containing the vertex @xmath354 is even and let this path be @xmath355 ( see figure @xmath356 note that the figure @xmath357 is for the case @xmath358 the figure for the case @xmath344 is similar ) . now the edges of @xmath267 together with the edges @xmath276 induce an even length path , say @xmath359 from the vertex @xmath260 to @xmath97 again , the length of the section of @xmath140 from @xmath95 to @xmath264 ( containing the vertex @xmath306 ) is odd since it is of same length as the section of @xmath239 from @xmath98 to @xmath256 ( containing the vertex @xmath353 . now the edges of the section of @xmath140 from @xmath95 to @xmath264 together with the edge @xmath268 induce an even length path , say @xmath360 as in case 1 , except the two @xmath80-cycles , namely , @xmath279 and @xmath361 in @xmath101 each of the @xmath80-cycles formed by two consecutive vertices of @xmath140 and their corresponding vertices in @xmath239 constitute an alternating @xmath80-cycle with respect to @xmath176 and @xmath177 clearly , one of the two paths along @xmath250 joining a pair of opposite vertices of these alternating @xmath80-cycles contains exactly one of the even length paths @xmath273 or @xmath362 and hence its length along @xmath176 is odd ; hence if @xmath363 is an alternating @xmath80-cycle with @xmath297 and @xmath298 then @xmath304 is odd . we now show that in at least one of these alternating @xmath80-cycles , described above , @xmath313 is odd . let @xmath321 and @xmath322 be as defined in case 1 . suppose the vertices of @xmath140 are alternately in @xmath321 and @xmath328 then @xmath113 is even , but it is not the case . therefore , there exist two consecutive vertices along @xmath140 which are in @xmath321 or @xmath340 thus there exists a pair of opposite vertices of an alternating @xmath80-cycle @xmath364 where @xmath365 and @xmath366 such that @xmath313 is odd and hence @xmath363 is an odd alternating @xmath80-cycle . thus @xmath94 satisfies property @xmath181 with respect to @xmath176 and @xmath177 [ lemma1ofbermond]_@xcite _ let @xmath94 be a graph of @xmath367 if the reduced graph @xmath190 admits a hamiltonian decomposition having the property @xmath212 between @xmath139 and @xmath111 then @xmath94 admits a hamilton cycle decomposition having the property @xmath212 between @xmath185 and @xmath368@xmath67 [ dc.remark1 ] in the construction of two edge disjoint hamilton cycles @xmath176 and @xmath223 of @xmath183 from the two edge disjoint hamilton cycles @xmath369 and @xmath370 respectively , of @xmath371 every @xmath98-edge of @xmath372 connecting the vertices of @xmath373 to @xmath374 is replaced by a path of odd length , see lemma [ lemma1ofbermond ] . hence the parity of length of the path between any pair of vertices in @xmath375 along @xmath376 remains the same as the parity of length of the path between the respective vertices in @xmath377 of @xmath378 where @xmath379 are the @xmath95-cycles of @xmath380@xmath67 [ dc.remark2 ] consider the graph @xmath382 and its reduced graph @xmath383 let @xmath384 be the hamilton cycle decomposition of @xmath190 guaranteed by lemma [ lem2.directcirculant ] . clearly , @xmath369 has two natural orientations one in the clockwise direction and the other in the anticlockwise direction and these two orientations induce orientations for the two @xmath98-edges of @xmath385 namely , @xmath386 and @xmath387 we first fix the clockwise orientation of the hamilton cycle @xmath369 of @xmath203 and let the orientation of the edges @xmath386 and @xmath268 of @xmath369 give two arcs @xmath388 and @xmath389 see figure 5 . with respect to this orientation , we obtain a hamilton cycle decomposition @xmath170 of @xmath101 using the proof of lemma [ lemma1ofbermond ] , with property @xmath163 where the arcs @xmath388 and @xmath390 correpond to the @xmath391 edges @xmath392 and @xmath393 respectively , in the proof of lemma [ lemma1ofbermond ] . the hamilton cycles @xmath176 and @xmath223 shown in figure 5(a ) arise out of the hamilton cycles @xmath369 and @xmath394 with respect to the clockwise orientation of @xmath369 and , the hamilton cycle decomposition @xmath395 shown in figure 5(b ) arises out of the hamilton cycles @xmath369 and @xmath394 with respect to the anticlockwise orientation of @xmath385 where @xmath396 + if we fix the anticlockwise orientation for @xmath385 we obtain another orientation of the edges @xmath386 and @xmath268 of @xmath369 and the resulting arcs are @xmath397 and @xmath398 see figure @xmath399 with respect to this orientation , we obtain a hamilton cycle decomposition @xmath395 of @xmath101 by proof of lemma [ lemma1ofbermond ] , with the property @xmath163 where the arcs @xmath397 and @xmath400 correspond to the @xmath391 edges @xmath392 and @xmath393 respectively , in the proof of lemma [ lemma1ofbermond ] , see figure 5 . irrespective of the orientations of @xmath369 ( that is , clockwise or anticlockwise ) , @xmath401 that is , the two circulant graphs obtained from the union of the hamilton cycles @xmath402 and @xmath403 in two different orientations of the edges of @xmath369 are isomorphic to each other . this isomorphism can be described by mapping @xmath118 to @xmath404 where the even integer @xmath4 is the number of vertices of @xmath203 and the vertices @xmath405 , are the fixed vertices of the isomorphism.@xmath67 [ lem1.directcirculant ] let @xmath183 be a graph with even number of vertices and having the generating set @xmath406 where @xmath407 and @xmath408 are of different parity . if the reduced graph @xmath371 of @xmath101 has the properties @xmath212 and @xmath237 with respect to a hamilton cycle decomposition @xmath384 of @xmath203 then @xmath94 has a hamilton cycle decomposition @xmath170 with the properties @xmath212 and @xmath213 * proof . * throughout the proof we use @xmath406 and @xmath169 as the generating sets for @xmath183 and @xmath371 respectively . without loss of generality assume that throughout the proof @xmath407 is even and @xmath408 is odd . as @xmath94 is of even order and as @xmath407 is even and @xmath408 is odd , @xmath102 is even and @xmath168 is odd , since @xmath409 and @xmath410 where @xmath411 is the number of vertices of @xmath412 let @xmath384 be a hamilton cycle decomposition of @xmath190 and @xmath170 be the corresponding hamilton cycle decomposition of @xmath412 in the construction of @xmath176 and @xmath223 from @xmath369 and @xmath394 ( see @xcite ) , @xmath413 is obtained from @xmath414 by replacing each of the edges of @xmath414 between @xmath373 to @xmath374 by an odd length path , whose origin is in @xmath139 and terminus is in @xmath140 of @xmath94 and the internal vertices of the paths are in @xmath184 and @xmath415 as @xmath384 has the property @xmath163 there exists a hamilton cycle decomposition @xmath170 of @xmath183 satisfying property @xmath212 by lemma [ lemma1ofbermond ] . next we prove that @xmath94 satisfies the property @xmath237 with respect to @xmath180 we prove this by induction on @xmath416 first we explain the idea behind the proof of this lemma . in claim 1 below , we prove the existence of a hamilton cycle decomposition of @xmath382 satisfying @xmath237 from a hamilton cycle decomposition of @xmath417 with property @xmath237 and in claim 2 below , we prove the existence of a hamilton cycle decomposition of @xmath418 satisfying @xmath237 from a hamilton cycle decomposition of @xmath419 with property @xmath213 in claim 3 below , we obtain a hamilton cycle decomposition of @xmath183 with property @xmath237 from a hamilton cycle decomposition of @xmath182 with property @xmath213 first we assume that @xmath384 is the hamilton cycle decomposition of @xmath371 where @xmath420 as described in lemma [ lem2.directcirculant ] . let @xmath417 and let its corresponding graph in @xmath161 be @xmath412 from the definition of @xmath422 the labels of the vertices of the @xmath127th @xmath95-cycle are @xmath423 where @xmath424 is the length of the @xmath127th @xmath95-cycle . we know that @xmath417 has only one @xmath95-cycle and hence its vertices are @xmath425 each vertex @xmath144 of @xmath190 gives rise to three vertices , namely , @xmath426 and @xmath427 in @xmath94 and we call these three vertices of @xmath94 as the corresponding vertices of @xmath144 of @xmath190 and vice versa . @xmath151 labeling of the vertices of the circulant graph @xmath417 on @xmath428 vertices with the group elements , where the generating set is @xmath429 @xmath430 labeling of the vertices of the graph @xmath190 with ordered pairs as described in @xcite . @xmath431 labeling of the vertices of the graph @xmath203 with @xmath432 if the vertex has the label @xmath144 in @xmath430 for our convenience we relabel the vertices @xmath433 of @xmath417 as @xmath434 see figure 6(c ) ; we call it as _ new labelling of @xmath190_(throughout this lemma , the new labelling of @xmath190 is denoted by bold face letters ) ; in fact , in the proof of this lemma each vertex of @xmath190 will have three different labels , namely , @xmath435 and the other one the group element and according to our convenience and circumstances we use one of these labels . but , for the vertices of @xmath382 we use the unique label @xmath118 as in @xcite . in the graph @xmath203 let the vertex @xmath256 be denoted by the label @xmath436 and hence in our new labeling it is denoted by @xmath437 see figure 6 ; for our convenience we write @xmath438 clearly , @xmath439 is the immediate next vertex of @xmath440 in @xmath374 of @xmath190 in the new labeling . corresponding to the vertices @xmath441 and @xmath442 of @xmath203 there are two rows , each having three vertices , in @xmath251 the vertices of these rows are denoted by , @xmath443 and @xmath444 respectively , see figure 7 . in the construction of @xmath382 from @xmath445 the two edges @xmath446 and @xmath447 are in @xmath176 and the two edges @xmath448 and @xmath449 are in @xmath283 where @xmath170 is the hamilton cycle decomposition of @xmath101 corresponding to the hamilton cycle decomposition @xmath384 of @xmath203 as described in @xcite , see figures 7 and 9 . we claim that @xmath94 satisfies the property @xmath237 with respect to the hamilton cycle decomposition @xmath450 where the required @xmath80-cycle of @xmath94 is @xmath451 the hamilton cycle @xmath176 of @xmath94 contains the four vertices @xmath454 and @xmath455 in the clockwise order as shown in the figure 7 ( the order is guaranteed by the corresponding hamilton cycle @xmath369 of @xmath190 ) , where @xmath447 is an edge of @xmath456 as @xmath374 being an even cycle , it can be thought of as a bipartite graph with bipartition @xmath457 and @xmath458 clearly , the vertex @xmath459 the @xmath460th vertex along @xmath461 the group element @xmath462 we do not differentiate the labels @xmath463 and @xmath464 and we denote the vertex @xmath256 by @xmath465 in the three labellings of the vertices of @xmath466 is in @xmath467 because @xmath98 is even implies @xmath256 is even ; as @xmath468 and @xmath95 is odd , @xmath441 is even . further , @xmath469 and @xmath470 implies @xmath471 is even . now consider the hamilton cycle decomposition @xmath384 of @xmath190 as in lemma [ lem2.directcirculant ] . the hamilton cycle @xmath369 of @xmath190 is obtained from @xmath374 by deleting two edges @xmath472 and @xmath473 of @xmath374 and adding two edges @xmath474 and @xmath475 see figure 8 . length of the @xmath476-section along @xmath369 is odd , as @xmath441 ( an even integer ) is the length from @xmath477 to @xmath441 along @xmath478 this implies that the @xmath479-section along @xmath369 is of odd length . next we prove that @xmath480 is odd . from the last paragraph , @xmath481 is even and hence @xmath482 is odd , see figure 8 ; the path from @xmath483 to @xmath441 along @xmath374 is also in @xmath369 and hence @xmath484 is odd ; consequently , as @xmath4 is even,@xmath485 ( where we consider the @xmath479-section of @xmath369 not containing the vertex @xmath441 ) , see figure 8 . next we prove that @xmath452 is odd . we divide the @xmath486-section of @xmath176 ( containing the vertex @xmath487 into three subsections and we show that each one of them is of odd length to conclude @xmath452 is odd ; the subsections are @xmath488 @xmath489 and @xmath490 see figure 7 . from the construction of @xmath176 from @xmath385 it is clear that as @xmath491-section and @xmath492-section of @xmath176 are of length 3 and length 1 , respectively , it is enough to show that @xmath489-subsection is of odd length . but it is easy to see that @xmath489-section of @xmath176 is identical with the @xmath479-section of the hamilton cycle @xmath385 see figure 7 . the @xmath479-section of @xmath369 is already proved to be of odd length , by ( [ directcirculant.equation1 ] ) . thus @xmath493 is odd . next we prove that @xmath494 is odd . as above , we divide the @xmath486-section of @xmath223 ( not containing the vertex @xmath495 ) into four subsections , namely , @xmath488((0,0),(0,1)),@xmath496 and @xmath497-section in the cyclic order are guaranteed by @xmath370 see @xcite and figure 9 . from the construction of @xmath283 @xmath498 is an edge and the @xmath497-section is a path of length 3 , namely , @xmath499 see figure 9 . hence we show that the lengths of the other two sections are of different parity . this is achieved by finding the lengths of the corresponding sections in @xmath394 of @xmath380 the sections in @xmath394 corresponding to the @xmath491-section and @xmath500-section of @xmath223 in @xmath94 are @xmath501-section and @xmath502-section , respectively , in the new labeling of @xmath380 we shall show that @xmath503 is odd and @xmath504 is even . we prove that @xmath503 is odd by observing that the @xmath505-section ( not containing the edge @xmath506 ) of one of the @xmath98-cycles , namely @xmath507 not containing the edge @xmath508 of @xmath190 is a section of @xmath370 and it is of odd length as the @xmath505-section of @xmath394 together with the edge @xmath506 is a @xmath98-cycle , which is of even length in @xmath203 that is , @xmath503 is odd . next we prove that @xmath504 is even . now consider the @xmath502-section of @xmath394 containing only @xmath98-edges . as we move along @xmath98-edges of the @xmath502-section of @xmath370 the alternate labels of the vertices are congruent to @xmath509 or @xmath510 if two nonconsecutive vertices of @xmath394 are both congruent to @xmath511 then their distance along @xmath394 is even . therefore , as @xmath512 implies @xmath513 @xmath514 is even . as observed earlier , the length of the @xmath486-section of @xmath223 is sum of the lengths of the sections @xmath488@xmath515@xmath500 and a path of length @xmath155 from @xmath516 to @xmath455 in @xmath223 of @xmath412 as @xmath503 is odd , @xmath517 is odd by remark [ dc.remark1 ] , the @xmath518-section is an edge , the length of @xmath500-section of @xmath223 is even , as @xmath504 is even and by remark [ dc.remark1 ] and the last section is a @xmath519 thus in @xmath101 @xmath494 is odd . let @xmath384 be the hamilton cycle decomposition of @xmath190 with property @xmath237 obtained in lemma [ lem2.directcirculant ] . as pointed out in remark [ dc.remark2 ] , the hamilton cycle @xmath369 has two natural orientations and with respect to each of these two orientations there is a hamilton cycle decomposition of @xmath520 we show that in at least one of these two orientations of @xmath385 the corresponding hamilton cycle decomposition of @xmath94 has the property @xmath237 ( note that both these two hamilton cycle decompositions satisfy the property @xmath212 as in remark [ dc.remark2 ] ) . consider the new labeling of @xmath190 described in case 1 above . we know that @xmath369 has exactly two @xmath98-edges . in this case we consider the clockwise orientation of @xmath521 consider the @xmath80-cycle @xmath522 in @xmath101 see figure 9 ( the figure for the case @xmath523 also resembles the same as in the case @xmath344 and note that @xmath441 is even ) . to prove @xmath452 is odd , we divide the @xmath486-section of @xmath250 ( containing the vertex @xmath524 ) into three subsections and we show that each of them is of odd length . the subsections are @xmath488 @xmath489 and @xmath490 see figure 7 ( the figure for the case @xmath523 also resembles the same as in the case @xmath344 ) . from the construction of @xmath176 from @xmath385 it is clear that @xmath491-section of @xmath176 is of length 3 , see figure 7 , and @xmath447 is an edge , hence it is enough to show that @xmath489-section ( containing the vertex @xmath525 ) is of odd length . but it is easy to see ( in fact , it is the same proof as in the case @xmath344 ) that this section of @xmath176 is the same as the @xmath479-section of the hamilton cycle @xmath369 of @xmath203 which is of odd length and hence @xmath493 is odd . this completes the proof that @xmath452 is odd . now we prove that @xmath526 is odd . for that we divide the @xmath486-section of @xmath223 ( not containing the vertex @xmath495 ) into four subsections , namely , @xmath527 and @xmath528 see figure 9 . it is clear that the @xmath518-section is an edge and the @xmath497-section is a path of length @xmath529 we shall show that the lengths of the subsections @xmath491 and @xmath500 are of different parity . first we prove that the length of the subsection @xmath491 of @xmath223 is even . clearly , the section corresponding to @xmath491 of @xmath223 is @xmath501-section in @xmath394 of @xmath203 in the new labeling . observe that the @xmath505-section of one of the @xmath98-cycles , not containing the edge @xmath530 of @xmath190 is a section of @xmath370 and it is of even length as the @xmath505-section of @xmath394 together with the edge @xmath506 is a @xmath98-cycle of odd length in @xmath190 ( as @xmath523 and @xmath531 ) , that is , @xmath503 is even . hence by remark [ dc.remark1 ] , @xmath532 is even . next consider the @xmath500-section ( containing the vertex @xmath533 ) of @xmath290 the corresponding @xmath502-section in @xmath394 contains only @xmath98-edges . as @xmath504 is odd , by assumption , and also the section @xmath502 of @xmath394 containing only @xmath98-edges , @xmath534 is odd , by remark [ dc.remark1 ] . this proves @xmath494 is odd . first we prove that @xmath536 is odd . for this , we divide the @xmath537-section of @xmath176 ( containing the vertex @xmath538 ) into three subsections , namely , @xmath539 and @xmath540 clearly , @xmath541-section is the path @xmath542 of length 3 and @xmath543-section is an edge and hence it is enough to prove that @xmath544-section of @xmath176 is of odd length . in @xmath369 of @xmath203 @xmath98 is even implies @xmath256 is even as @xmath4 is even . as @xmath468 and @xmath95 is odd implies @xmath441 is even . it is an easy observation that the vertices with even numbered labels in the new labeling of @xmath369 are at odd distance from the vertex @xmath545 the new label of the vertex @xmath546 since the even numbered vertices are @xmath547 etc . hence @xmath484 is odd as @xmath441 is even . as the @xmath544-section of @xmath176 has the same length as the @xmath548-section of @xmath385 @xmath549 is odd . finally , we show that @xmath550 is odd . to prove this , we divide the @xmath537-section of @xmath223 ( containing the vertex @xmath551 ) into three subsections , namely , @xmath552 and @xmath553 see figure 11 . it is clear that the @xmath518-section is an edge and the @xmath554-section is the path @xmath555 on @xmath556 vertices , see figure 11 . hence it is enough to show that the @xmath500-section is of even length . as @xmath504 is even , by assumption in this subcase , @xmath502-section of @xmath370 containing only @xmath98-edges , is of even length and hence @xmath534 is even , by remark [ dc.remark1 ] . thus @xmath557 is odd . hence the hamilton cycle decomposition @xmath170 of @xmath94 satisfies the property @xmath213 * claim 2 . * for any hamilton cycle decomposition @xmath384 of @xmath419 with the property @xmath181 there is a hamilton cycle decomposition @xmath170 of @xmath418 with the property @xmath213 consider the graph @xmath558 there are two @xmath95-cycles @xmath374 and @xmath559 hence there are two parallel matchings between @xmath374 and @xmath560 we denote the matching from @xmath374 to @xmath561 as @xmath241 and the other matching from @xmath561 to @xmath374 as @xmath242 the edges of @xmath241 are @xmath562 and the edges of @xmath244 are @xmath563 @xmath564 for some @xmath248 clearly , from the construction of @xmath369 and @xmath394 of @xmath203 as in @xcite , every alternating @xmath80-cycle having two consecutive vertices along @xmath374 and two consecutive vertices along @xmath561 uses exactly two @xmath95-edges and two @xmath98-edges and both these two @xmath98-edges are in @xmath241 or in @xmath565 lemma [ lem2.directcirculant ] guarantees the existence of an odd alternating @xmath80-cycle with respect to the hamilton cycle decomposition @xmath384 of @xmath190 and this @xmath80-cycle contains two @xmath98-edges of @xmath566 in the construction of @xmath413 of @xmath94 ( from @xmath372 of @xmath190 ) the @xmath95-edges and @xmath241-edges ( that is , @xmath98-edges that belong to the matching @xmath241 ) of @xmath190 are retained as it is in the transformation of @xmath190 to @xmath412 while obtaining @xmath376 from @xmath378 each of the @xmath98-edges of @xmath244 in @xmath414 are replaced by a path of odd length whose internal vertices are in @xmath567 and @xmath568 hence an odd alternating @xmath80-cycle that exists in @xmath190 becomes an odd alternating @xmath80-cycle with respect to the hamilton cycle decomposition @xmath170 of @xmath412 * claim 3 . * for any hamilton cycle decomposition @xmath384 of @xmath182 with property @xmath181 there is a hamilton cycle decomposition @xmath170 of @xmath183 with property @xmath213 in claims 1 and 2 , we proved that the existence of the property @xmath237 in @xmath569 or @xmath232 implies the existence of the property @xmath237 in @xmath382 or @xmath570 respectively . from the hamilton cycle decomposition and construction of the odd alternating @xmath80-cycle of @xmath94 from @xmath203 we have shown that the odd alternating @xmath80-cycle lies between @xmath140 and @xmath227 that is having two vertices in @xmath140 and two vertices in @xmath240 hence , as in claim 2 , the odd alternating @xmath80-cycle in the hamilton cycle decomposition @xmath384 of @xmath571 is also the odd alternating @xmath80-cycle in the hamilton cycle decomposition @xmath170 of @xmath572 this completes the proof of existence of the property @xmath237 with respect to the hamilton cycle decomposition @xmath170 of @xmath412 [ dc.remark3 ] recall that our graphs of @xmath123 always have even order by assumption . as every member of @xmath123 is isomorphic to a member of @xmath146 and vice - versa , to each member of @xmath210 there corresponds a graph @xmath573 if @xmath102 is odd , then @xmath96 the length of each of the @xmath95-cycles of @xmath574 must be even . let @xmath575 and let @xmath576 be the image of @xmath94 under an isomorphism . clearly , the @xmath95-edges and @xmath98-edges of @xmath94 are the @xmath98-edges and @xmath95-edges of @xmath577 respectively , and vice versa . as @xmath424 is even , each @xmath95-cycle @xmath578 in @xmath94 has two @xmath86-factors , say @xmath579 and @xmath580 in the subgraph induced by @xmath581 that is , for @xmath119 @xmath582 and @xmath583 the edges of the @xmath86-factor @xmath579 of @xmath94 become the @xmath241-edges of @xmath576 and the edges of the @xmath86-factor @xmath580 of @xmath94 become the @xmath244-edges of @xmath584 in the proof of lemma [ lem1.directcirculant ] , we constructed an odd alternating @xmath80-cycle with respect to the hamilton cycle decomposition of @xmath101 so that the two @xmath95-edges use the edges of @xmath585 hence , if @xmath575 has an odd alternating @xmath80-cycle , then the image of the @xmath80-cycle under the isomorphism between the corresponding graphs of @xmath586 and @xmath587 gives the odd alternating @xmath80-cycle in @xmath576 using two @xmath241-edges of @xmath584 [ dc.remark4 ] by saying , for @xmath134 obtain the reduced graph of @xmath94 in @xmath588 we mean the following : the successive reduced graphs of @xmath134 in @xmath589 yields a graph @xmath590 but the class of graphs in @xmath591 are isomorphic to the class of graphs in @xmath592 that is , to each graph of @xmath591 there is an isomorphic copy of it in @xmath593 and vice versa . hence @xmath594 can be considered as a graph @xmath595 then successive reduced graphs of @xmath596 yields a graph @xmath597 as the class of graphs in @xmath598 are isomorphic to the class of graphs in @xmath599 @xmath600 can be considered as a graph @xmath601 we shall call @xmath602 or @xmath603 ( note that @xmath604 ) as a reduced graph of @xmath134 according to the circumstances . similarly , by saying , for @xmath605 obtain its lifted graph @xmath606 where @xmath3 and @xmath607 are not simultaneously zero , we mean the following : for @xmath195 by lifting , we obtain a graph @xmath608 similarly , for @xmath609 by lifting , we obtain @xmath610 successively , we can get the graph @xmath611 but the class of graphs in @xmath612 are isomorphic to the class of graphs in @xmath613 that is , to each graph of @xmath612 there is an isomorphic copy of it in @xmath614 and vice versa . hence @xmath615 can be considered as a graph of @xmath616 from this , by successive liftings , we obtain @xmath617 we call @xmath618 as a lifted graph of @xmath207 as the two classes of graphs @xmath619 and @xmath620 have the same set of graphs , upto isomorphism , we can consider @xmath618 as an element of @xmath621 * proof . * let @xmath214 as observed in lemma [ lem1.directcirculant ] , @xmath102 and @xmath168 are of different parity and hence we assume that @xmath102 is even and @xmath168 is odd . obtain the reduced graph @xmath417 of @xmath101 see remark [ dc.remark4 ] . if @xmath190 is simple , then it has a hamilton cycle decomposition satisfying properties @xmath212 and @xmath181 by lemma [ lem2.directcirculant ] . now we lift " @xmath190 to @xmath622 by lemma [ lem1.directcirculant ] , @xmath623 has the properties @xmath212 and @xmath237 with respect to a hamilton cycle decomposition of it . @xmath623 can be considered as a graph @xmath624 again lift @xmath625 to a graph in @xmath422 which is precisely @xmath101 having the properties @xmath212 and @xmath237 with respect to a hamilton cycle decomposition of it , by lemma [ lem1.directcirculant ] . thus the result is true if @xmath190 is simple . next we assume that @xmath190 is not simple . let @xmath626 be a reduced graph of @xmath134 such that @xmath594 is simple but the reduced graph of @xmath594 in @xmath627 is not simple . the successive reduced graphs of @xmath594 results in @xmath558 because of the above process of reductions and liftings of the classes of graphs , @xmath628 and @xmath629 may have different parity from the parity of @xmath102 and @xmath630 respectively . @xmath190 is not simple only in three cases , see @xcite : as the jumps of @xmath594 are of different parity , the first two cases correspond to the cartesian product of cycles ; in that case @xmath636 and @xmath637 if @xmath638 again we reduce the graph @xmath639 or @xmath640 to a graph in @xmath641 or @xmath642 but @xmath641 and @xmath643 are isomorphic classes of graphs . hence it is enough to consider @xmath641 with @xmath644 then the class of graphs @xmath641 with @xmath636 reduces to the single graph @xmath645 where @xmath67 is the cartesian product of graphs . in figure 12 , a hamilton cycle decomposition is shown , where @xmath646 is an odd alternating @xmath80-cycle . next we consider the last case @xmath648 as @xmath594 is simple , then @xmath649 and @xmath650 as @xmath651 if @xmath652 we first reduce the graph @xmath594 to a graph in @xmath653 in this case there is only one graph in @xmath654 see @xcite , as shown in figure 13 . in the hamilton cycle decomposition shown in figure 13 , @xmath646 is an odd alternating @xmath80-cycle . to complete the proof , first we lift the graph @xmath647 of figure 12 to the graph @xmath655 with properties @xmath212 and @xmath237 with respect to a hamilton cycle decomposition , by lemma [ lem1.directcirculant ] . let @xmath656 be isomorphic to @xmath21 as @xmath0 has the properties @xmath212 and @xmath181 so does @xmath657 now we lift the graph @xmath656 to the graph @xmath214 clearly , @xmath134 has a hamilton cycle decomposition , by lemma [ lem1.directcirculant ] . similarly , we prove it for the graph in figure 13 . _ @xcite_[thm.main.directcirculant ] let @xmath170 be a hamilton cycle decomposition of a @xmath80-regular graph @xmath0 of even order @xmath411 containing an odd alternating four cycle , that is property @xmath181 with respect to @xmath450 then the graph @xmath658 even , admits a hamilton cycle decomposition.@xmath67 let @xmath659 and @xmath660 since @xmath1 is hamilton cycle decomposable , we have @xmath661 where each @xmath662 is a hamilton cycle of @xmath18 if at least one of @xmath0 or @xmath1 has odd order , then @xmath72 is hamilton cycle decomposable , by theorem [ ann.dm.3.21-28 ] , and hence assume that both @xmath0 and @xmath1 are of even order . since @xmath0 has the property @xmath78 @xmath0 can be decomposed into @xmath80-regular connected circulants , that is , @xmath663 where each @xmath664 is a @xmath80-regular circulant graph with jumps of different parity . thus @xmath665 can be decomposed into two hamilton cycles with property @xmath181 by theorem [ thm.directcirculant ] . now @xmath666 but each @xmath667 can be decomposed into hamilton cycles , by theorem [ thm.main.directcirculant ] . * conclusion . * in , existence of hamilton cycle decompositions of the graphs @xmath668 are proved ; the factor graphs in the product graphs are either complete or complete multipartite graphs , which are very dense graphs . however , if we consider one of the factor graphs as circulant , in @xmath7 with same number of odd and even jumps , irrespective of the number of jumps , which can be paired so that the resulting set of edges induces connected @xmath80-regular circulants , then @xmath72 is hamilton cycle decomposable . this proves that with this additional condition on @xmath0 or @xmath11 the factor graphs @xmath0 and @xmath1 need not be dense . for example , hamilton cycle decomposition of @xmath669 where @xmath84 is a @xmath86-factor of @xmath670 follows from our theorem [ thm.cir.1 ] whenever @xmath1 is hamilton cycle decomposable multigraph . also one can conclude that tensor products of certain sparse hamilton cycle decomposable circulant graphs are hamilton cycle decomposable .

we present molecular line observations , made with angular resolutions of , toward the filamentary infrared dark cloud g34.43 + 0.24 using the apex [ , , and transitions ] , nobeyama 45 m [ , , , , and transitions ] , and sest [ and transitions ] telescopes . we find that the spatial distribution of the molecular emission is similar to that of the dust continuum emission observed with 11 resolution ( rathborne et al . 2005 ) showing a filamentary structure and four cores . the cores have local thermodynamic equilibrium masses ranging from and virial masses from , molecular hydrogen densities between and , and column densities ; values characteristics of massive star forming cores . the profile observed toward the most massive core reveals a blue profile indicating that the core is undergoing large scale inward motion with an average infall velocity of 1.3 and a mass infall rate of yr . we report the discovery of a molecular outflow toward the northernmost core thought to be in a very early stage of evolution . we also detect the presence of high velocity gas toward each of the other three cores , giving support to the hypothesis that the excess 4.5 m emission ( `` green fuzzies '' ) detected toward these cores is due to shocked gas . the molecular outflows are massive and energetic , with masses ranging from 25 80 , momentum 2.3 , and kinetic energies 1.1 km s ; indicating that they are driven by luminous , high - mass young stellar objects .

1005.3048

infrared dark clouds ( irdcs ) are cold ( @xmath1625 k ) , massive ( @xmath0@xmath17 @xmath18 ) , and dense ( @xmath19@xmath20 @xmath6 ) molecular clouds with high column densities ( @xmath0@xmath21@xmath22 @xmath8 ) seen as a dark silhouette against the bright mid infrared background emission ( perault et al . 1996 ; egan et al . 1998 ; carey et al . 1998 , 2000 ; hennebelle et al . 2001 ; simon et al . 2006a , 2006b ) . a wealth of recent observations show that irdcs are the sites of high - mass stars and star cluster formation ( rathborne et al . 2006 , 2007 ; pillai et al . 2006 ; jackson et al . 2008 ; chambers et al . dust continuum observations reveal that irdcs harbor compact cold cores with typical masses of @xmath0120 and sizes @xmath160.5 pc ( rathborne et al . the masses and sizes of the irdc cores are similar to those of the massive hot cores ( e.g. , garay & lizano 1999 ) , but irdc cores are much colder . recently , chambers et al . ( 2009 ) investigated a sample of 190 cores found toward 38 irdcs by rathborne et al . ( 2006 ) , and classified them as `` active '' if they are associated with enhanced 4.5 @xmath11 m emission , the so - called green fuzzies , and an embedded 24 @xmath11 m source ; and `` quiescent '' if they contain neither of these indicators . they found 37 active cores and 69 quiescent cores . further , active cores are usually associated with bright ( @xmath191 jy ) methanol and water masers ( pillai et al 2006 ; wang et al . 2006 ; chambers et al . 2009 ) , and have high bolometric luminosities , indicating that they are forming high mass stars ( m @xmath19 8 ) stars . the quiescent cores are the most likely candidates for high mass starless cores . g34.43 + 0.24 is an irdc with a filamentary morphology , extending by @xmath09 from north to south in equatorial projection ( 9.7 pc at the source distance of 3.7 kpc ; fandez et al . 2004 ; simon et al . this irdc is in the list of rathborne et al . ( 2006 ) and chambers et al . it is located roughly 11 north of the ultra - compact ( uc ) h ii complex g34.26 + 0.15 ( molinari et al . 1996 ) . molecular line observations toward this cloud were first reported by miralles et al . ( 1994 ) , who from ammonia observations with 1.5 resolution found an elongated structure in the n - s direction with a total mass of 1000 . near the center of the filament lies the iras point source 18507 + 0121 which has a luminosity of @xmath23 . bronfman et al . ( 1996 ) detected toward iras 18507 + 0121 strong emission with broad line wings , indicating it is associated with a dense massive star forming region . ramesh et al . ( 1997 ) observed the iras source in the , , and molecular lines , with angular resolutions of @xmath016 , modeling the observed line profiles as due to emission from a collapsing hot core that is hidden behind a cold and dense envelope . from observations of 3 mm continuum and h@xmath24co@xmath25 line emissions , with @xmath05 angular resolution , shepherd et al . ( 2004 ) detected two compact molecular cores , separated by 40 , toward the center of the filament . millimeter continuum observations revealed the presence of four dust cores ( fandez et al . 2004 ; garay et al . 2004 ; rathborne et al . 2005 ) , labeled by the latter authors as mm1 , mm2 , mm3 , and mm4 . the mm1 dust core , which corresponds to the northern compact molecular core of shepherd et al . ( 2004 ) , contains a deeply embedded luminous object surrounded by several hundred solar masses of warm gas and dust . based on the weak 6 cm continuum emission and the lack of detection at nir wavelenghts , shepherd et al . ( 2004 ) suggested that the embedded object appears to be a massive b2 protostar in an early stage of evolution . the mm2 dust core , which corresponds to the southern compact molecular core of shepherd et al . ( 2004 ) , is associated with iras 18507 + 0121 ; with a nir cluster of young stars with a central b0.5 star ( shepherd et al . 2004 ) ; with an uc h ii region ( miralles et al . 1994 ; molinari et al . 1998 ) ; with a variable h@xmath26o maser ( miralles et al . 1994 ) ; and with ch@xmath27oh maser emission ( szymczak et al . 2000 ) . the mm3 dust core is located close to the northern edge of the filament , @xmath03.5north of mm2 . garay et al . ( 2004 ) suggested that mm3 corresponds to a massive and dense cold core that will eventually collapse to form a high - mass star . the mm4 core is located @xmath030south of mm2 ( rathborne et al . table [ tbl-1.2 mm ] lists the measured 1.2 mm dust continuum emission properties as measured by rathborne et al . rathborne et al . ( 2005 ) and chambers et al ( 2009 ) reported an excess 4.5 @xmath11 m emission toward mm1 , mm3 , and mm4 cores , which could be produced by either ionized gas and/or shocked gas and acts therefore as a signature of current star formation ( mm2 is classified by the last authors as a `` red '' core , namely with bright 8 @xmath11 m emission ) . they also showed that the four mm cores are associated with 24 @xmath11 m point sources ( emission from warm dust ) , which is an indicator of accretion onto an embedded protostar . water maser emission , a well known tracer of low mass and high mass star formation , has been detected toward all four mm cores ( wang et al . 2006 ; chambers et al . chambers et al . ( 2009 ) detected class i ch@xmath27oh masers , another well know signpost of star formation toward mm1 , mm2 , and mm3 cores . rathborne et al . ( 2006 ) identified nine 1.2 mm continuum cores with masses ranging from 80 to 1300 toward g34.43 + 0.24 . shepherd et al . ( 2007 ) observed the central region of the filament , encompassing the mm1 and mm2 cores , at millimeter wavelengths , in co@xmath28 , @xmath24co@xmath28 , and c@xmath29o@xmath28 at @xmath30 angular resolution and at near infrared wavelengths with the _ spitzer _ space telescope . they discovered five massive outflows , three of them associated with the mm2 core and the remaining two associated with the mm1 core . from the _ spitzer _ data they identified 31 young stellar objects in the filament , with a combined mass of @xmath0127 , and additional 22 sources that they believe could be cluster members based on the presence of strong 24 @xmath11 m emission . cortes et al . ( 2008 ) observed mm1 and mm2 in polarized thermal dust emission at 3 mm and co@xmath28 line emission with bima array . their results suggest that there is a magnetic field othogonal to the axis formed by the mm1 and mm2 cores . recently , rathborne et al . ( 2008 ) carried out high angular submillimeter observations toward the mm1 core using the submillimeter array . they determined , using continuum emission images , that mm1 remains unresolved with a size of @xmath00.03 pc and a mass of 29 . their molecular line spectrum shows the presence of several complex molecules , suggesting that mm1 is a hot core . they also found an extended structure which may be evidence of a rotating envelope surrounding the central high mass star , perpendicular to one of the outflows that shepherd et al . ( 2007 ) detected toward mm1 . in this work we report observations of 11 molecular transitions using the apex 12-m , sest 15-m , and nobeyama 45-m telescopes toward the g34.43 + 0.24 cloud . since different molecular species trace different density and kinematics regimes , the capability of observing multiple spectral lines is particularly advantageous to determine the morphology and kinematics of star forming cores . due to the high abundance of co , the line is an ideal tracer of the low intensity , high velocity wing emission . the line , being more optically thin than , traces , in addition to the ambient gas , the low velocity gas of the molecular outflow . the line is usually associated with the presence of outflows since its abundance is highly enhanced by shocks . methanol transitions are usually used as tracers of dense gas , temperature probes and tracers of shocks , since this molecule is evaporated from grain mantles . the and lines are high density tracers and thus are good probes of the dense envelopes surrounding protostars . the , , , and molecular transitions probe the inner regions of the cores . the last two species are weakly influenced by infall and outflow motions , and are used as reference to estimate the central ambient cloud velocity . the main goals of these multi line observations are to determine the morphology and kinematics of the gas within the molecular cores in the g34.43 + 0.24 filamentary irdc , in particular to investigate the presence of outflowing gas , in order to increase our knowledge of the evolutionary state of the cores and to achieve a better understanding of the overall process of massive star formation along the filament . this is the first survey of several molecular lines mapping portions and the whole cloud . it gives us , for the first time , a comprehensive view of the molecular gas distribution at the scale of parsecs . the observations were made using the 12-m apex telescope , the 15-m sest telescope , both located in chile , and the 45-m telescope of the nobeyama radio observatory ( nro ) located in japan . the molecules , transitions , frequencies and observing parameters are listed in table [ tbl - obsparam ] . with the apex telescope we observed the , , and lines . the observations were carried out during three epochs : 2005 july , 2005 october , and 2006 may - july . the telescope beam size at the frequency of the observed lines ranged from 17 to 19 ( fwhm ) . we used the heterodyne receiver ( apex-2a ) , which provided a channel separation of 122 khz and a total bandwidth of 1 ghz with 8192 spectral channels . in and , we mapped the emission within a 100@xmath31680region which cover a spatial extent similar to that exhibited by the 1.2 mm dust continuum emission map ( see garay et al . the spacing between adjacent points on a regular grid was 20and the position - switching mode was used . the on - source integration time per map position was typically @xmath040 s in both molecules . system temperatures were in the range 110 - 340 k in and 200 - 300 k in , resulting in an rms noise per channel , in antenna temperature , of typically 0.27 k and 0.37 k , respectively . in we mapped the emission within a 80@xmath31100zone , covering the central region of the filament also known as the iras 18507 + 0121 region . system temperatures were typically @xmath0190 k. the on - source integration time per map position was @xmath050 s resulting an rms noise per channel of typically 0.29 k in antenna temperature . in we mapped the emission within a 40@xmath31120region . system temperatures were typically @xmath0200 k. on - source integration time was @xmath080 s resulting an rms noise per channel of typically 0.18 k in antenna temperature . the molecular line data were reduced according to the standard procedure using class , a gildas working group software . with the nro 45 m telescope we observed the , , , , and lines toward the central region of the filament . the observations were undertaken during 1996 february - april . the telescope beam size at the frequencies of the observed lines is 15 - 18 . we used an acousto - optical spectrometer with 37 khz resolution covering a total bandwidth of 40 mhz . the emission was mapped in all lines within a 80@xmath31140region . the spacing between adjacent points on a regular grid was 20 , using the position - switching mode . system temperatures were in the range between 160 and 390 k. on - source integration times per map position ranged from 300 to 600 s resulting in rms noise of typically 0.08 k per channel in antenna temperature in all lines . first order polynomials were fitted in order to estimate and remove the baselines . part of this data were used by ramesh , bronfman & deguchi ( 1997 ) to model the kinematics at the peak position of iras 18507 + 0121 . with the sest telescope we observed the and lines toward the northern part of the filament ( mm3 core ) . the telescope beam size is 24 at the frequency of the line and 51 at the frequency of the line . the observations were carried out in 2003 march . we used the high - resolution acousto - optical spectrometers , which provided a channel separation of 43 khz and a total bandwidth of 43 mhz . we observed , with 30 spacing between adjacent points on a regular grid , 25 positions within a 150@xmath31150 region . the on - source integration times per map position were typically @xmath0130 s for cs and @xmath070 s for c@xmath29o . system temperatures were in the range 230 - 270 k for cs and 400 - 540 k for c@xmath29o , resulting in an rms noise , in antenna temperature , of typically 0.07 k and 0.18 k per channel , respectively . figure [ fig-13co32-spectra ] shows the spectra observed with apex along the whole filamentary structure detected in the 1.2 mm dust continuum observations of garay et al . ( 2004 ; see their figure 4 ) . emission in the line was detected across the whole mapped region . figure [ fig - ambientcentral ] shows contour maps of the velocity integrated ambient gas emission ( velocity range from 52.9 to 61.1 , determined by inspection of the line profiles ) in the , , , , and lines observed toward the central @xmath01.5@xmath313 region of the filament . the maps are overlaid on a grey scale image of the 1.2 mm dust continuum emission observed with simba / sest by garay et al . we find that the morphology of the molecular line emission resembles that of the dust continuum emission , showing two components : a northern component associated with the mm1 core and a southern component associated with the mm2 dust core . there are , however , some differences in the morphology of the emission in the different molecular lines . these are likely due to chemistry and/or optical depth effects and/or differences in the regimes of density and temperature traced by the different lines . in particular , the emission in the line is less extended than in the other transitions and runs nearly parallel to the north - south direction . figure [ ch3oh_mapa_pos_pos ] presents contour maps of the velocity integrated ambient gas emission ( velocity range from 55.8 to 60.6 ) in three methanol lines [ ch@xmath27oh(2@xmath32 ) a+ , ch@xmath27oh(2@xmath33 ) e , ch@xmath27oh(2@xmath34 ) e ] observed toward the central region of g34.43 + 0.24 . the maps are overlaid on a gray scale image of the 1.2 mm dust continuum emission observed with simba / sest . the morphology of the emission in the three methanol lines is all similar . they are , however , different from the morphology exhibited by the dust continuum emission and the rest of the molecular lines . the methanol peaks do not coincide with the dust peaks , on the contrary they seem to be slightly anticorrelated . this could be due to depletion of methanol within the central regions of the dust cores . figure [ fig - ambientmm3 ] shows contour maps of velocity - integrated ambient gas emission ( velocity range from 56.1 to 61.5 ) in the , and lines observed toward the mm3 core , overlaid on a gray scale image of the 1.2 mm dust continuum emission . whereas the morphology of the emission in the line is similar to that of the dust continuum emission , that in the line is somewhat different , showing a peak that is displaced from the peak of the dust emission . table [ tbl - sizes ] gives the observed parameters of the molecular cores . the angular sizes , determined from the contour maps , correspond to the observed semi - major and semi - minor axes . the line center velocity , line width and velocity - integrated brightness temperature were determined from gaussian fits to the composite spectra of the ambient cloud emission within each core . the cores are assumed to be at a distance of 3.7 kpc , which corresponds to the kinematical distance of iras 18507 + 0121 reported by fandez et al . ( 2004 ) and the kinematical distance of the irdc given by simon et al . ( 2006b ) . there are notable differences between the line profiles observed in different molecular species , for different positions . in what follows we describe the spectra observed toward three selected positions ( 0 , 20 ) , ( 20 , -20 ) and ( 40 , 200 ) associated with , respectively , the mm1 , mm2 and mm3 dust peaks . figure [ fig - specpeakmm1 ] shows the spectra observed toward the ( 0 , 20 ) position , which corresponds to the mm1 core . the profile shows , in addition to considerable self - absorption , the presence of both blueshifted and redshifted high velocity emission . the strong wing emission at redshifted velocities is also clearly seen in the , , and spectra . figure [ fig - specpeakmm2 ] shows the spectra observed toward the ( 20,@xmath3520 ) position ( mm2 core ) . the profile shows considerable self - absorption , and the presence of high velocity emission at both blueshifted and redshifted velocities . the line profile exhibits a strong blueshifted feature , with a peak at the velocity of 55.8 , and a flat redshifted shoulder . the line shows strong emission toward blueshifted velocities , with a peak at the velocity of 56.1 , and a skewed profile toward redshifted velocities . a similar profile is seen in the line . the and lines show double - peaked profiles ; the former exhibits a strong blue peak and a weaker red peak whereas the later shows an opposite asymmetry , namely a strong redshifted peak and a weaker blueshifted peak . the and lines exhibit symmetrical profiles , with peak line center velocites , determined from gaussian fits , of 58.1 @xmath36 0.1 and 57.4 @xmath36 0.2 , respectively . ramesh et al . ( 1997 ) modeled the observed , , and profiles toward mm2 as being produced by a collapsing hot core hidden behind a cold intervening screen or envelope . we have modeled here the line profiles of , and using the simple model of spectral - line profiles from contracting clouds of myers et al . we concluded that the mm2 core is undergoing large - scale inward motions with a infall velocity of @xmath01.3 ( see section [ infall ] ) . figure [ fig - specpeakmm3 ] shows the spectra observed toward the ( 40 , 200 ) position , which corresponds to the mm3 core . the spectrum shows , in addition to considerable self - absorption , strong blueshifted wing emission up to a velocity of 42.2 . the spectrum exhibits a strong gaussian line , with a center velocity of 58.5 @xmath36 0.1 , and blueshifted wing emission up to a velocity of 53.6 . these observations strongly indicate the presence of an energy source within the mm3 core . the and spectra exhibit nearly gaussian lines profiles with peak line center velocities of 58.9 @xmath36 0.1 and 58.8 @xmath36 0.1 , respectively . figure [ fig - wingpos - vel_mm1andmm2 ] shows position - velocity diagrams of the , , , and emission , toward the central region of g34.43 + 0.24 , along two cuts in the north - south direction passing through the peak position of the mm1 ( left panel ) and mm2 ( right panel ) cores . strikingly seen in this figure is the presence of high velocity gas associated with the mm1 and mm2 cores , most likely due to molecular outflows emanating from both cores . particularly notable is the detection of emission in the silicon monoxide line , which is a powerful tracer of the presence of strong shocks . the abundance of sio molecules is highly enhanced , with respect to ambient abundances , in molecular outflows as a result of destruction of dust grains by shocks , giving rise to the injection into the gas phase of si atoms and/or si - bearing species . once silicon is injected into the gas phase , chemical models based on ion - molecule reactions predict a large abundance of sio molecules ( turner & dalgarno 1977 ; hartquist et al . 1980 ) . in the cut passing through the peak of mm1 , the most prominent outflow emission is seen redshifted with respect to the ambient gas velocity , extending up to velocities of 80.7 in and 70.2 in and . weak or no outflow emission is detected in the line , indicating low optical depths . in the cut passing through the peak of mm2 , both blueshifted and redshifted high velocity emission are clearly seen emanating from the mm2 core . in the blueshifted and redshifted emissions extend up to velocities of 42.0 and 74.2 , respectively . also seen in this cut is wing emission emanating from the mm1 core as well as from the mm4 core , suggesting that mm4 also contains an outflow . maps of the velocity integrated line wing emission , in the , and lines toward the central 1.5@xmath313 region of g34.43 + 0.24 , including the mm1 , mm2 and mm4 dust cores , are shown in figure [ fig - wingpos - pos_mm1andmm2 ] . in and , the range of velocity integration of the blueshifted emission ( blue contours or solid lines ) is @xmath37 , whereas that of the redshifted emission ( red contours or dashed lines ) is @xmath38 . in the range of velocity integration of the blueshifted and redshifted emissions are , respectively , @xmath39 and @xmath40 . in fig . [ fig - wingpos - pos_mm1andmm2 ] we identify two , possibly three , outflows associated with the dust cores within the region considered . the northernmost outflow ( labeled o - mm1 ) is associated with the mm1 core . it is roughly oriented in the ne - sw direction , with the peaks of the redshifted and blueshifted emission being located , respectively , sw and ne of the 3 mm continuum source detected by shepherd et al . ( 2004 ; marked with a star ) . there is , however , a large spatial overlap between the blueshifted and redshifted emissions . the outflow associated with the mm2 core , ( labeled o - mm2 ) , shows a complex morphology with the blueshifted and redshifted emission exhibiting a considerable spatial overlap . the peak position of the redshifted and blueshifted emissions are located @xmath07 south of the uc hii region ( marked with a triangle ) and are coincident within the errors . this suggest that the symmetry axis of the outflow is close to the line of sight . outflow - mm4 is associated with the mm4 dust core ( marked with a square ) . the blueshifted and redshifted emissions exhibit a large overlap in their spatial distribution , suggesting that the flow is unresolved or that its axis of symmetry is close to the line of sight . figure [ fig - wingpos - vel_mm3 ] shows position - velocity diagrams of the and emission toward the northern region of g34.43 + 0.24 , along a line in the north - south direction passing through the peak position of the mm3 core . clearly seen in the cut is the presence of wing emission emanating from mm3 , with velocities extending up to 42.2 in the blueshifted side and 69.2 in the redshifted side . in the line the wing emission is barely seen . figure [ fig - wingpos - pos_mm3 ] shows maps of the integrated line wing emission , in the and lines , toward the mm3 core . the range of velocity integration of the blueshifted emission ( blue contours ) is @xmath41 for and @xmath42 for ; whereas that of the redshifted emission ( red contours ) is @xmath43 for and @xmath44 for . in the map , the peak positions of the redshifted and blueshifted wing emissions are located , respectively , ne and sw of the peak position of the millimeter dust core . the red and blue wing emissions exhibit however a considerable overlap in their spatial distribution . from the observations of the and lines it is possible to estimate the column density and the mass of molecular gas within the central region ( mm1 and mm2 cores ) as follows . assuming that the line is emitted under local thermodynamic equilibrium ( lte ) conditions and its emission is optically thin , then the column density is given by ( e.g. , bourke et al . 1997 ) @xmath45 where @xmath46 is the optical depth of the line , @xmath47 is the excitation temperature , and @xmath48 is in . the lte mass can be obtained from @xmath49}{3.8\times10^{6}}\right ) \left(\frac{d}{\rm kpc}\right)^2 \ , \frac{(t_{ex}+0.88)\exp(15.81/t_{ex})}{1-\exp(-15.81/t_{ex } ) } \int\!\!\int \tau_{18 } dv d\omega~ , \label{eqn - lte - mass}\ ] ] where @xmath50 is the mean molecular mass per h@xmath26 molecule , @xmath51 is the mass of a hydrogen atom , @xmath52 is the source distance , [ h@xmath26/c@xmath29o ] is the h@xmath26 to c@xmath29o abundance ratio , and @xmath53 is in arcmin@xmath14 . the opacities in the line are obtained from the ratio of the observed brightness temperatures in the and lines as follows ( see bourke et al . 1997 , for derivation ) @xmath54}{[j_{13}(t_{ex})-j_{13}(t_{\rm bg})]}\,\frac{t_{\rm mb}(^{13}{\rm co})}{t_{\rm mb}({\rm c}^{18}{\rm o})}~ , \label{eqn - for - tau}\ ] ] where @xmath55 is the background temperature , @xmath56 is the main beam brightness temperature of the line , @xmath57 is given by ( the subscripts `` 18 '' and `` 13 '' refer to the c@xmath29o and @xmath24co isotopes , respectively ) @xmath58 and , @xmath59 is the optical depth ratio @xmath60\frac{(kt_{ex}/hb_{13 } + 1/3)}{(kt_{ex}/hb_{18 } + 1/3)}\,\frac{\exp ( e_{_{j_{13}}}/kt_{ex})}{\exp ( e_{_{j_{18}}}/kt_{ex})}\,\frac{[1 - \exp ( -h\nu_{_{18}}/kt_{ex})]}{[1 - \exp ( -h\nu_{_{13}}/kt_{ex})]}~ , \label{eqn - razon - taus}\ ] ] where @xmath61 is the rotational constant of the molecule , @xmath62 is the rotational energy at the lower state , and @xmath57 is the rotational quantum number . assuming an [ @xmath24co / c@xmath29o ] abundance ratio of 7.6 , we derive average optical depths in of 0.10 and 0.18 toward the mm1 and mm2 cores , respectively . using an excitation temperature of 30 k and an [ h@xmath26/c@xmath29o ] abundance ratio of @xmath63 ( wilson & rood 1994 ) we derive total masses of 330 and 1460 for the mm1 and mm2 cores , respectively . the derived core parameters are listed in table [ tbl - derivparams1 ] . masses can also be estimated assuming that the clouds are in virial equilibrium . neglecting magnetic fields and external forces , the virial mass , @xmath64 , for a spherical cloud of radius @xmath65 is given by ( maclaren , richardson & wolfendale 1988 ) @xmath66 where @xmath67 is the average line width in , @xmath65 in pc , and @xmath61 is a constant which depends on the density profile of the cloud . assuming that the cores have uniform densities ( @xmath68 ) we derive virial masses of @xmath3 , @xmath2 and @xmath69 for , respectively , the mm1 , mm2 and mm3 cores . these values correspond to the geometric mean of the masses determined from the transitions given in table [ tbl - derivparams2 ] . the molecular hydrogen densities were computed from the mass assuming that the cloud has a spherical morphology and a uniform density . table [ tbl - comparison ] summarizes the core masses determined from different methods and authors ( 2 ) and ( 3 ) give the lte and the virial masses , respectively , determined in this work ; col . ( 4 ) gives the lte mass from shepherd et al . ( 2007 ) ; and , cols . ( 5 ) and ( 6 ) give the masses derived from dust continuum observations by garay et al . ( 2004 ) and rathborne et al . ( 2006 ) , respectively . in our estimate , if the excitation temperature is 25 or 35 k , the lte mass is decreased or increased by @xmath020% , respectively . given the uncertainties associated with each method such as in the abundance , excitation temperature , dust temperature , and dust opacity the agreement is good between the masses determined in this work and the masses determined from dust emission . differences in the mass obtained by different authors using dust emission are mostly due to different dust temperatures and core sizes used in their calculations . there are , however , discrepancies with the masses determined by shepherd et al . ( 2004 ) from h@xmath24co@xmath70 observations , of 4000 and 5000 for mm1 and mm2 , respectively , and a total cloud mass of 50000 . we note that rathborne et al . ( 2005 ) estimated from dust emission observations a total cloud mass of 7500 assuming a dust temperature of 30 k. a possible explanation of the larger mass estimate of shepherd et al . ( 2004 ) is that the actual abundance of hco@xmath25 is probably larger than the adopted one , due to an enhancement produced by the molecular outflows . moreover , shepherd et al . ( 2007 ) determined , from observations of the c@xmath29o@xmath28 line using the ovro array , masses of 75 and 690 for mm1 and mm2 , respectively . these values probably correspond to the inner mass of the cores . we note that masses derived from molecular lines might not be tracing all of the emission , and their emission may be optically thick , specially c@xmath29o@xmath28 . however , the assumption that the is optically thin seems to be adequate and it is supported for the inspection of the line profiles and values obtained for the optical depths . optical depths and column densities toward the peak position of the mm1 and mm2 cores are given in table [ tbl - derivparampeak ] . the and values were obtained in an analogous way to that described above for the and lines , adopting a [ cs / c@xmath71s ] abundance ratio of 22.5 and an [ h@xmath26/cs ] ratio of @xmath72 ( e.g. , van der tak et al . 2000 ) . the line emission at the peak position of the mm2 core shows a bright blueshifted peak and a flat redshifted shoulder . such line shape can be modeled as due to inward motions ( mardones et al . the blue peak is produced by high excitation gas on the far side of the envelope , while the redshifted peak results from low excitation gas in the near side of the envelope . the self absorption due to low excitation , low velocity gas in the near side of the envelope gives rise to the central dip ( evans 1999 ) . an estimate of the infall velocity , @xmath73 , can be obtained using the two layer infall model of myers et al . ( 1996 ) . myers et al . ( 1996 ) made a simple analytic model of radiative transfer in order to inspect if the observed spectra show any evidence of infall motions and to provide an estimate of the characteristic @xmath73 . they assumed two uniform regions of equal temperature and velocity dispersion @xmath74 , whose density and velocity are attenuation - weighted means over the front and rear halves of a centrally condensed , contracting cloud with @xmath75 and @xmath76 . we modeled the profile of the , and lines toward the mm2 core using their prescription and obtained the following model parameters : the peak optical depth , @xmath77 ; the kinetic temperature , @xmath78 ; the infall velocity , @xmath73 ; and the non thermal velocity dispersion , @xmath79 . figure [ fig - myers - model ] shows the observed and the best model line profiles , indicating motions consistent with a `` fast '' infall , @xmath80 . the values derived by the model are shown in the top right corner in the same figure . we note that some of the parameters for each fit are different , for example , the infall velocities are 1.5 , 1.3 , and 1.0 for @xmath24co , c@xmath29o , and cs , respectively . this decreasing order is in accord with the level of asymmetry in the lines . the most asymmetric line , a single peak in the blue and a red shoulder , has the largest values of @xmath73 ( myers et al . 1996 ; fuller et al . 2005 ) . although the velocity dispersion in the cs line is @xmath81 , the other two lines have @xmath82 indicating a fast infall . the derived systemic velocities show some scatter , most likely due to the influence of the molecular outflow , which is more intense toward redshifted velocities . if the contribution of the outflow is removed , the blue peak becomes stronger than the red shoulder , and this will have the effect of increasing the derived infall velocity . ramesh et al . ( 1997 ) modeled the profile of the , , and lines toward the mm2 core using both the large velocity gradient ( lvg ) method and the simple model of myers et al . for the lvg method they adopted a constant temperature of 22 k ( we determined a kinetic temperature of @xmath036 k ) , and found a collapse velocity that varies radially as @xmath83 with @xmath84 cm . at a radius of @xmath85 cm ( equivalent to an angular size of 10 ) we get @xmath86 . given the uncertainties associated with each method , the agreement is good with the values determined by us . the infall speed estimated by ramesh et al . ( 1997 ) using the myers et al . ( 1996 ) analytic model is about 1 order of magnitude smaller than the velocities they determine from the lvg analysis ; however , since the fit of the myers model is admittedly not too good , they argue in favor of the larger value , which agrees well with ours here . the mass infall rate can be estimated assuming @xmath87 , where @xmath88 is the mass within the radius @xmath65 and @xmath73 is the infall speed at radius @xmath65 . using the lte mass derived at the central position of mm2 , of 240 , an average infall velocity of 1.3 , and a radius of @xmath89 cm ( corresponding to a 10angular radius ) ; we obtain a mass infall rate of @xmath9 yr@xmath10 . such accretion rate is high enough to overcome the radiation pressure of the central star and to form a massive star . shepherd et al . ( 2007 ) identified two likely massive stars ( m @xmath19 8 ) with the mm2 core , with one of them , located near the projected center of the uc h ii region , having a spectral type of b0.5 ( shepherd et al . . models of massive envelopes accreting onto this kind of young stars require accretion rates of @xmath90 yr@xmath10 ( osorio , lizano & dalessio 1999 ) , in agreement with our results . to compute physical parameters of the outflowing gas , we followed the standard lte formalism described in bourke et al . ( 1997 ) using the emission in the co(@xmath91 ) line . the main source of error in determining the mass in the outflow arises from the difficulty in separating the contribution to the outflows of the emission in the velocity range of the ambient cloud . we adopt as velocity boundary between the blue and red wing emission and the ambient emission the values of , respectively , @xmath92 and @xmath93 for the mm1 outflow , @xmath94 and @xmath95 for the mm2 outflow , and 56.1 and 61.5 for the mm3 outflow . another source of error is the possibility that the line might not be optically thin , particularly at low flow velocities . fortunately , with the available data , we can asses this effect and make corrections . the mass in the high velocity wings is estimated from the co(3@xmath962 ) observations as follows . assuming that the energy levels of co are populated according to lte and that the outflow emission is optically thin , then the total co column density , @xmath97 , at each observed position is given by ( e.g. , garden et al . 1991 ) @xmath98 } \,\,\frac{1}{[j(t_{ex } ) - j(t_{bg } ) ] } \int t_{b}(v)dv~,\ ] ] where @xmath47 is the excitation temperature of the population in the rotational levels , @xmath99 is the background temperature , @xmath100 is the brightness temperature of the co(3@xmath962 ) emission at velocity @xmath48 , @xmath48 is measured in , and @xmath57 is defined by equation ( [ eqn - j ] ) . the mass is then computed from the derived column densities as , @xmath101~\mu_m~a~ \sum n({\rm co})~,\ ] ] where @xmath50 is the mean molecular mass per h@xmath26 molecule , [ h@xmath26/co ] is the molecular hydrogen to carbon monoxide abundance ratio , @xmath102 is the size of the emitting area in an individual position , and the sum is over all the observed positions inside each lobe . from the observations of the and lines it is possible to estimate the opacities of the flowing gas and therefore to assess whether or not the assumption that the wing emission is optically thin is correct . using expressions ( [ eqn - for - tau ] ) and ( [ eqn - razon - taus ] ) for @xmath103co and @xmath24co , assuming @xmath47 = 30 k , and [ @xmath103co/@xmath24co ] = 50 , we find that the average opacities of the blueshifted gas ( in the velocity range from 49.5 to 53.0 ) and redshifted gas ( in the velocity range from 60.8 to 68.9 ) of the mm1 flow are 3.4 and 2.5 , respectively . the given ranges are those in which emission is detected in both lines . for the mm2 outflow we find that the average optical depth of the blueshifted gas ( in the velocity range from 47.2 to 51.8 ) and redshifted gas ( in the velocity range from 60.9 to 68.7 ) are 5.4 and 4.4 , respectively . for the mm3 outflow we find that the average optical depth of the blueshifted gas ( velocity range from 52.8 to 56.1 ) and redshifted gas ( velocity range from 61.5 to 65.2 ) are 7.1 and 4.2 , respectively . these calculations indicate that the emission from the flowing gas , in the given velocity ranges , is moderately optically thick for the three flows . on the other hand , as the position velocity diagrams suggested , the @xmath24co flow emission is optically thin , with optical depths between 0.05 and 0.13 . to correct the mass determined under the optically thin assumption by the opacity effect we divided the wing emission into two regimes , an optically thin regime , in which no emission is detected , and an optically thick regime . we assumed an [ h@xmath26/co ] abundance ratio of @xmath18 ( frerking , langer & wilson 1982 ) and an excitation temperature of the outflowing gas of 30 k. we derived masses in the optically thin regime of the 0.5 , 1.1 , and 1.6 for the blueshifted lobes of mm1 ( @xmath104 ) , mm2 ( @xmath105 ) , and mm3 ( @xmath106 ) , respectively , and masses of 0.7 , 0.8 , and 0.7 for the redshifted lobes of mm1 ( @xmath107 ) , mm2 ( @xmath108 ) , and mm3 ( @xmath109 ) , respectively . in the optically thick regime the mass determined under the optically thin assumption is multiplied by the factor @xmath110 , where @xmath111 is the average optical depth of wing emission in the optically thick velocity range . we derive masses in the optically thick regime , corrected by opacity , of 2.6 , 13.6 , and 8.5 for the blueshifted lobes of mm1 ( @xmath112 ) , mm2 ( @xmath113 ) , and mm3 ( @xmath114 ) , respectively . similarly , for the redshifted lobes we derived masses of 11.9 , 30.8 , and 7.3 toward mm1 ( @xmath115 ) , mm2 ( @xmath116 ) , and mm3 ( @xmath117 ) , respectively . to estimate the contribution to the flow mass from the low velocity material emitting in the same velocity range as the ambient cloud gas , we followed the prescription of margulis & lada ( 1985 ) . using expression ( a16 ) of bourke et al . ( 1997 ) , @xmath118 where @xmath119 , @xmath120 and @xmath121 , @xmath122 are the brightness temperature and velocity at the blue and red velocity boundaries , respectively . we estimated that the hidden flow masses corrected by opacity in the low velocity range within the blueshifted lobes toward mm1 , mm2 and mm3 are 3.0 , 20.6 and 13.9 , and within the redshifted lobes are 6.2 , 11.9 and 8.9 , respectively ( see table [ tbl - derivparams - outflows ] ) . finally , the total mass of the outflows are : 24.9 for mm1 , 78.8 for mm2 , and 40.9 for mm3 . using the standard lte formalism , it is also possible to estimate the momentum , @xmath123 , and the kinetic energy , @xmath124 , in the flow as follows ( margulis & lada 1985 ; bourke et al . 1997 ) @xmath125 these equations assume no correction for flow inclination , and so are strict lower limits . another method to compute the physical parameters of the outflowing gas is to assume that all the mass is flowing at a velocity characteristic of the entire flow , @xmath126 . the flow parameters are , in this approach , given by @xmath127 where @xmath128 is the total mass of the outflow lobe . the blueshifted and redshifted emissions from the outflows show considerable overlap in their spatial distribution , indicating that the symmetry axis of the flows is likely to be near the line of sight . hence , we adopt as a characteristic flow velocity the average of the maximum observed outflow velocities . the parameters determined under this assumption are likely to correspond to upper limits . the maximum velocities observed toward the blue and red lobes are , respectively , 12.0 and 23.1 for the mm1 outflow , 15.0 and 15.9 for the mm2 outflow , and 17.3 and 11.6 for the mm3 outflow . table [ tbl - derivparams - outflows ] gives the derived parameters of the blueshifted gas and redshifted gas of the flows calculated using both methods . the geometric mean values for the outflow parameters are @xmath129 , @xmath130 , and @xmath131 ; and @xmath132 , @xmath133 , and @xmath134 km@xmath14 s@xmath15 ( subscripts 1 , 2 , and 3 refer to the mm cores ) . if the excitation temperature is increased to 50 k , the flow parameters increase by 7% . the derived values of the outflow parameters are similar to those of massive and energetic molecular outflows driven by high mass young stellar objects ( shepherd & churchwell 1996 ; beuther et al . 2002 ; wu et al . we do not provide estimates of the dynamical timescale , mechanical luminosity , and mass outflow rate of the outflows toward g34.43 + 0.24 because they are likely to be aligned close to the line of sight , and therefore the lengths of the lobes are difficult to estimate . rathborne et al . ( 2005 ) reported an excess 4.5 @xmath11 m emission toward the mm cores . the _ spitzer _ irac 4.5 @xmath11 m band contains the h i br@xmath135 line ( at 4.052 @xmath11 m ) , the co@xmath28 p(8 ) band head ( 4.64.8@xmath11 m ) , and many h@xmath26 vibrational transitions ( the most important being h@xmath136 s(9 ) line at 4.694 @xmath11 m ) , but unlike the other three bands it does not contain any strong polycyclic aromatic hydrocarbon ( pah ) features ( e.g. , smith et al . it has been postulated that the main contributers of the excess 4.5 @xmath11 m emission could be emission from the co@xmath28 p(8 ) band head ( e.g. , marston et al . 2004 ) and/or emission from the h@xmath136 s(9 ) line ( e.g. , noriega crespo et al . 2004 ) . thus , the enhancement of the 4.5 @xmath11 m emission could be a probe of jets and molecular outflows ( e.g. , teixeira et al . 2008 ; smith et al . 2006 ; noriega crespo et al . 2004 ) ; therefore of current star formation . rathborne et al . ( 2005 ) first found evidence for the presence of high velocity gas toward the mm1 and mm3 cores from the broad line widths ( @xmath13710 ) observed in the cs@xmath138 and hcn@xmath138 lines . a detailed study of the central region of g34.43 + 0.24 was carried out by shepherd et al . ( 2007 ) , who reported two molecular outflows emanating from mm1 ( g34.4 mm , in their paper ) , and three outflows originating from the mm2 core . they estimate a total mass , momentum and kinetic energy of 34.8 , 350 , and @xmath139 ergs ( 2460 km@xmath14 s@xmath15 ) , for the outflows in mm1 ; and 111.2 , 830 , and @xmath140 ergs ( 3920 km@xmath14 s@xmath15 ) , for the outflows in mm2 . the masses are @xmath020% greater than our estimate , and the momenta and kinetic energies are within our estimated range . in spite of the different molecular transitions , angular resolutions and assumptions used , shepherd et al . and our results are in agreement , within the observational errors . we report the discovery of one massive molecular outflow associated with the mm3 core discovered by garay et al . they suggested that this massive and dense core , without counterpart at mid infrared ( _ msx _ ) and far infrared ( _ iras _ ) wavelengths implying that it has a low temperature , seemed to be in a quiescent state , and thus appeared to be the youngest of the massive cores in the filament . the presence of 24 @xmath11 m emission , a green fuzzy , water and methanol masers ( rathborne et al . 2005 ; chambers et al . 2009 ) and the strong wing emission in the line , indicating the presence of a powerful molecular outflow , all indicate star formation activity in the northernmost region of the filament . due to its spatial location and mass , yso number 29 in shepherd et al . ( 2007 ) appears to be a good candidate for being the driver of the mm3outflow . it is the most massive stellar object within mm3 , with a mass of @xmath04.86.8 and a luminosity of @xmath0170340 . however , this is an intermediate mass star that may not be able to drive alone the massive outflow . in the same region there are three more stars of intermediate mass ( less massive than the number 29 ) with maximum masses between 45 . interferometric observations are needed to clarify if the mm3outflow corresponds to an overlap of a few individual outflows , such as that seen in the central region of g34.43 + 0.24 , or it is driven by a single star . we carried out a multi line study of cores within the filamentary irdc g34.43 + 0.24 . the emission from the filament was fully mapped in the and lines . in addition , emission from the central region of the filament was observed in eight different molecular transitions [ , , , , , , and ] , and from the north region in the and lines . the results of this study are summarized as follows . the spatial distribution of the molecular emission in all , except methanol , lines is similar to that of the dust continuum emission derived from observations with similar angular resolution , showing a filamentary structure and four cores . the mass of the molecular cores derived assuming lte conditions are 300 and 1460 for the mm1 and mm2 cores , respectively . the masses derived assuming virial equilibrium are 1100 , 1500 and 1400 for the mm1 , mm2 and mm3 cores , respectively . the average molecular hydrogen densities of the cores are @xmath141 @xmath6 in mm1 , @xmath142 @xmath6 in mm2 and @xmath143 @xmath6 in mm3 . we also find that the molecular hydrogen column densities at the peak positions of mm1 and mm2 are , respectively , @xmath144 @xmath8 and @xmath145 @xmath8 . the derived parameters of the mm1 and mm2 cores are typical of massive and dense cores harboring high - mass yso s ( plume et al . 1997 ; garay et al . those of the mm3 core suggest that it harbors young intermediate mass stars , in accord with the result obtained by shepherd et al . ( 2007 ) from the modeling of the sed of the nir sources detected using _ spitzer_. we modeled the molecular emission in the , , and lines toward the mm2 core , concluding that this core is undergoing infalling motions , with an infall velocity of @xmath01.3 and a mass infall rate of @xmath0@xmath146 yr@xmath10 . this value is large enough to allow the formation of massive stars by accretion ( see osorio , lizano & dalessio 1999 ) . we report the discovery of a molecular outflow associated with the massive , dense mm3 core located in the northern region of g34.43 + 0.24 , and possibly of a second one associated with the mm4 core . the molecular outflow within mm3 has a total mass of 41 , a momentum of 270 , and a kinetic energy of 1120 km@xmath14 s@xmath15 ; indicating that it is driven by high intermadiate mass stars . we also detected high velocity molecular gas toward the mm1 and mm2 cores with total masses of 25 and 80 , momenta of 230 and 690 , and kinetic energies of 1290 and 3630 km@xmath14 s@xmath15 , respectively . noriega - crespo , a. , morris , p. , marleau , f. r. , carey , s. , boogert , a. , van dishoeck , e. , evans , ii , n. j. , keene , j. , muzerolle , j. , stapelfeldt , k. , pontoppidan , k. , lowrance , p. , allen , l. , & bourke , t. l. 2004 , , 154 , 352 is from 52.9 to 61.1 . the fwhm beam is shown in the bottom left corner . contour levels are from 20% to 90% ( in steps of 10% ) of the peak integrated intensity of 49.9 k for ( _ a _ ) ; 7.7 k for ( _ b _ ) ; 16.9 k for ( _ c _ ) ; 3.4 k for ( _ d _ ) ; 3.7 k for ( _ e _ ) ; and 14.3 k for ( _ f _ ) . the blue lines indicate the 50% contour level . the gray scale levels are 1 , 2 , 3 , 5 , 7 , 10 , 15 , 20 and 30 @xmath31 0.12 jy beam@xmath10 . the triangles mark the peak positions of the mm1 , mm2 and mm4 dust cores . ] oh emission toward the central region of g34.43 + 0.24 , overlaid on a 1.2 mm dust continuum emission map ( gray scale ) . the fwhm beam is shown in the bottom left corner . contour levels are from 20% to 90% ( in steps of 10% ) of the peak integrated intensity of 5.1 k for ch@xmath27oh(2@xmath32 ) a+ ( _ a _ ) ; 4.1 k for ch@xmath27oh(2@xmath33 ) e ( _ b _ ) ; and 1.1 k for ch@xmath27oh(2@xmath34 ) e ( _ c _ ) . the blue lines indicate the 50% contour level . the gray scale levels are 1 , 2 , 3 , 5 , 7 , 10 , 15 , 20 and 30 @xmath31 0.12 jy beam@xmath10 . the triangles mark the peak positions of the mm1 , mm2 and mm4 dust cores . ] is from 56.1 to 61.5 . contour levels are from 20% to 90% ( in steps of 10% ) of the peak integrated intensity of 12.5 k for ( _ a _ ) and 3.7 k for ( _ b _ ) ; and from 3 to 21 @xmath74 in steps 3 @xmath74 , with @xmath74 = 0.14 k , for ( _ c _ ) . the gray scale levels are 1 , 2 , 3 , 5 , 7 , 10 , 15 , 20 and 30 @xmath31 0.12 jy beam@xmath10 . the blue lines indicate the 50% contour level . the triangle marks the peak position of the mm3 core . the fwhm beam is shown in the bottom left corner . ] 0.27 k for ; 3 , 6 , 9 , 12 , 17 , 22 , 28 , and 34 @xmath31 0.08 k for ; 3 , 4 , 6 , and 8 @xmath31 0.07 k for ; and 1.5 , 3 , 6 , 9 , 12 , 15 , and 19 @xmath31 0.37 k for . _ right _ : cut passing through mm2 or position ( 20 , 0 ) . contour levels are 3 , 8 , 13 , 18 , 28 , 38 , 48 , and 58 @xmath31 0.27 k for ; 3 , 6 , 11 , 16 , 21 , 28 , 35 , and 42 @xmath31 0.08 k for ; 3 , 5 , 7 , and 9 @xmath31 0.07 k for ; and 1.5 , 3 , 6 , 10 , 14 , 19 , 25 , and 31 @xmath31 0.37 k for . ] for ; 3 , 5 , 7 , 9 , 12 , and 16 @xmath74 for ; and 3 , 9 , 15 , 25 , and 40 @xmath74 for . for the redshifted emission , these are 20 , 35 , 50 , 80 , 110 , 150 , 190 , and 230 @xmath74 for ; 3 , 5 , 7 , 9 , 12 , 16 , 24 , and 32 @xmath74 for ; and 3 , 9 , 15 , 25 , 40 , 60 , 85 , and 110 @xmath74 for . the @xmath74 values are : 0.30 k for the co line , and 0.04 k for the sio and cs lines . ] ) . transitions and telescopes are indicated in the top left corner of each panel . the contour levels for are 3 , 6 , 10 , 15 , 22 , 29 and 36 @xmath74 , with @xmath74 = 0.27 k. for are 1.5 , 3 , 6 , 8 and 10 @xmath74 , with @xmath74 = 0.37 k. ] ( @xmath74 = 0.35 k ) ; and for redshifted emission , these are 20 , 30 , 60 , 90 , 120 and 150 @xmath74 ( @xmath74 = 0.25 k ) . _ bottom _ : emission . contour levels for the blueshifted emission are 3 , 4 , 5 , 7 and 9 @xmath74 ( @xmath74 = 0.23 k ) ; and for redshifted emission , these are 3 , 4 , 5 , 6 and 7 @xmath74 ( @xmath74 = 0.19 k ) . ] cccccc & & & + & @xmath135(j2000 ) & @xmath148(j2000 ) & angular & physical & mass + core & & & ( ) & ( pc ) & + mm1 & 18 53 18.0 & 01 25 24 & 16 & 0.19 & 1187 + mm2 & 18 53 18.6 & 01 24 40 & 26 & 0.42 & 1284 + mm3 & 18 53 20.4 & 01 28 23 & 24 & 0.38 & 301 + mm4 & 18 53 19.0 & 01 24 08 & 24 & 0.38 & 253 + lccccccc & frequency & beam ( fwhm ) & spacing & observed & region size & @xmath67 & rms noise ( t@xmath149 ) + observation & ( ghz ) & ( arcsec ) & ( arcsec ) & positions&(arcsec@xmath14 ) & ( km s@xmath10 ) & ( k ) + + & 330.588 & 18.4 & 20 & 145 & @xmath150 clccccc + mm1 & & @xmath156 & 57.38 @xmath36 0.21 & 5.69 @xmath36 0.71 & 6.94 @xmath36 0.80 & @xmath157 + & & @xmath158 & 58.13 @xmath36 0.01 & 5.29 @xmath36 0.02 & 15.10 @xmath36 0.04 & @xmath159 + & & @xmath160 & 58.36 @xmath36 0.02 & 2.75 @xmath36 0.06 & 2.78 @xmath36 0.05 & @xmath161 + mm2 & & @xmath162 & 57.96 @xmath36 0.03 & 5.27 @xmath36 0.09 & 9.41 @xmath36 0.13 & @xmath163 + & & @xmath164 & 57.35 @xmath36 0.01 & 5.19 @xmath36 0.03 & 15.41 @xmath36 0.07 & @xmath165 + & & @xmath166 & 57.88 @xmath36 0.02 & 3.96 @xmath36 0.05 & 3.52 @xmath36 0.04 & @xmath167 + mm3 & & @xmath168 & 58.77 @xmath36 0.02 & 3.01 @xmath36 0.04 & 10.31 @xmath36 0.11 & @xmath169 + & & @xmath170 & 58.73 @xmath36 0.01 & 3.43 @xmath36 0.03 & 3.44 @xmath36 0.02 & @xmath171 + llcccc & & & n & n(h@xmath26 ) & m@xmath172 + core & transition & @xmath173&(@xmath8)&(@xmath8 ) & ( ) + mm1 & & 0.10 & @xmath174 & @xmath175 & 330 + & & 0.75 & @xmath176 & & + mm2 & & 0.18 & @xmath177 & @xmath178 & 1460 + & & 1.39 & @xmath179 & & + clcccc & & radius & @xmath180 & m@xmath181 & n(h@xmath26 ) + core & & ( pc ) & ( ) & ( ) & ( @xmath6 ) + mm1 & & 0.25 & 5.69 & 1700 & @xmath182 + & & 0.35 & 5.29 & 2060 & @xmath183 + & & 0.26 & 2.75 & 410 & @xmath184 + mm2 & & 0.27 & 5.27 & 1580 & @xmath185 + & & 0.35 & 5.19 & 1980 & @xmath186 + & & 0.33 & 3.96 & 1090 & @xmath187 + mm3 & & 0.57 & 3.01 & 1090 & @xmath188 + & & 0.69 & 3.43 & 1710 & @xmath189 + cccccc & lte mass & virial mass & lte mass & + core & ( this work ) & ( this work ) & shepherd et al . ( 2007 ) & garay et al . ( 2004 ) & rathborne et al . ( 2006 ) + ( 1 ) & ( 2 ) & ( 3 ) & ( 4 ) & ( 5 ) & ( 6 ) + mm1 & 330 & 1130 & 75 & 550 & 1187 + mm2 & 1460 & 1510 & 690 & 2700 & 1284 + mm3 & - & 1370 & - & 780 & 301 + clccc & & & n & n(h@xmath26 ) + core & & @xmath173&(@xmath8)&(@xmath8 ) + mm1 & & 0.04 & @xmath190 & @xmath191 + & & 0.30 & @xmath192 & + & & 0.02 & @xmath193 & @xmath178 + & & 0.41 & @xmath194 & + mm2 & & 0.37 & @xmath195 & @xmath196 + & & 2.80 & @xmath197 & + & & 0.03 & @xmath198 & @xmath199 + & & 0.69 & @xmath200 & + + mm1 & blueshifted & 3 & 3 & 6 & @xmath201 & @xmath202 + & redshifted & 13 & 6 & 19 & @xmath203 & @xmath204 + & total & 16 & 9 & 25 & @xmath205 & @xmath206 + mm2 & blueshifted & 15 & 21 & 36 & @xmath207 & @xmath208 + & redshifted & 32 & 12 & 44 & @xmath209 & @xmath210 + & total & 47 & 33 & 80 & @xmath211 & @xmath212 + mm3 & blueshifted & 10 & 14 & 24 & @xmath213 & @xmath214 + & redshifted & 8 & 9 & 17 & @xmath215 & @xmath216 + & total & 18 & 23 & 41 & @xmath217 & @xmath218 +

graphons are analytic objects representing limits of convergent sequences of graphs . lovsz and szegedy conjectured that every finitely forcible graphon , i.e. any graphon determined by finitely many subgraph densities , has a simple structure . in particular , one of their conjectures would imply that every finitely forcible graphon has a weak-regular partition with the number of parts bounded by a polynomial in . we construct a finitely forcible graphon such that the number of parts in any weak-regular partition of is at least exponential in . this bound almost matches the known upper bound for graphs and , in a certain sense , is the best possible for graphons .

1507.00067

the theory of combinatorial limits has recently attracted a significant amount of attention . this line of research was sparked by limits of dense graphs @xcite , which we focus on here , followed by limits of other structures , e.g. permutations @xcite , sparse graphs @xcite and partial orders @xcite . methods related to combinatorial limits have led to substantial results in many areas of mathematics and computer science , particularly in extremal combinatorics . for example , the notion of flag algebras , which is strongly related to combinatorial limits , resulted in progress on many important problems in extremal combinatorics @xcite . theory of combinatorial limits also provided a new perspective on existing concepts in other areas , e.g. property testing algorithms in computer science @xcite . a convergent sequence of dense graphs can be represented by an analytic object called a _ graphon_. let @xmath4 be the density of a graph @xmath5 in a graphon @xmath2 ( a formal definition is given in section [ sec : limits ] ) . a graphon @xmath2 is said to be _ finitely forcible _ if it is determined by finitely many subgraph densities , i.e. there exist graphs @xmath6 and reals @xmath7 such that @xmath2 is the unique graphon with @xmath8 . finitely forcible graphons appear in many different settings , one of which is in extremal combinatorics . it is known that if a graphon is finitely forcible , then it is the unique graphon which minimizes a fixed finite linear combination subgraph densities , i.e. finitely forcible graphons are extremal points of the space of all graphons . the following conjecture ( * ? ? ? * conjecture 7 ) claims that the converse is also true . let @xmath6 be finite graphs and @xmath9 reals . there exists a finitely forcible graphon @xmath2 that minimizes the sum @xmath10 . finitely forcible graphons are also related to quasirandomness in graphs as studied e.g. by chung , graham and wilson @xcite , rdl @xcite and thomason @xcite . in the language of graph limits , results on quasirandom graphs state that every constant graphon is finitely forcible . a generalization of this statement was proven by lovsz and ss @xcite : every step graphon ( i.e. a multipartite graphon with a finite number of parts and uniform edge densities between its parts ) is finitely forcible . in @xcite , lovsz and szegedy carried out a more systematic study of finitely forcible graphons . the examples of finitely forcible graphons that they constructed led to a belief that finitely forcible graphons must have a simple structure . to formalize this , they introduced the ( topological ) space @xmath11 of typical vertices of a graphon @xmath2 and conjectured the following ( * ? ? ? * conjectures 9 and 10 ) . [ conj - compact ] the space of typical vertices of every finitely forcible graphon is compact . [ conj - dimension ] the space of typical vertices of every finitely forcible graphon has a finite dimension . both conjectures were disproved through counterexample constructions in @xcite . conjecture [ conj - dimension ] is a starting point of our work . analogously to weak regularity of graphs , every graphon has a weak @xmath0-regular partition with at most @xmath12 parts . ( see section [ sec : weak ] for the necessary definitions . ) if the space of typical vertices of a graphon is equipped with an appropriate metric , then its minkowski dimension is linked to the number of parts in its weak regular partitions . in particular , if its minkowski dimension is @xmath13 , then the graphon has weak @xmath0-regular partitions with @xmath14 parts . consequently , if conjecture [ conj - dimension ] were true , the number of parts of a weak @xmath0-regular partitions of a finitely forcible graphon would be bounded by a polynomial of @xmath1 . the number of parts in weak @xmath0-regular partitions of a graphon constructed in @xcite as a counterexample to conjecture [ conj - dimension ] is @xmath15 , which is superpolynomial in @xmath1 , but is much smaller than the general upper bound of @xmath12 . we construct a finitely forcible graphon almost matching the upper bound . [ thm - main ] there exist a finitely forcible graphon @xmath2 and positive reals @xmath16 tending to @xmath17 such that every weak @xmath16-regular partition of @xmath2 has at least @xmath18 parts . as pointed out to us by jacob fox , there is no graphon ( finitely forcible or not ) matching the upper bound for infinitely many values of @xmath0 tending to @xmath17 . in light of this , theorem [ thm - main ] is almost the best possible . [ prop - lower - graphon ] there exist no graphon @xmath2 , positive real @xmath19 and positive reals @xmath16 tending to @xmath17 such that every weak @xmath16-regular partition of @xmath2 has at least @xmath20 parts . the proof of this proposition is sketched at the end of section [ sec : weak ] . we will refer to the graphon @xmath2 from theorem [ thm - main ] as the graphon . is the name of a famous ( and fictitious ) brave czech soldier and , more importantly for us , it is the name of the restaurant where we usually ate lunch during our work on this subject while three of us were visiting the university of west bohemia in pilsen . we now introduce notions related to graphons and convergent sequences of graphs . the _ density _ of a graph @xmath5 in @xmath21 , which is denoted by @xmath22 , is the probability that @xmath23 randomly chosen vertices of @xmath21 induce a subgraph isomorphic to @xmath5 , where @xmath23 is the order ( the number of vertices ) of @xmath5 . a sequence of graphs @xmath24 with the number of their vertices tending to infinity is _ convergent _ if the sequence @xmath25 converges for every graph @xmath5 . note that if @xmath26 has @xmath27 edges , then the sequence @xmath24 is convergent for trivial reasons . hence , this notion of graph convergence is of interest for sequences of dense graphs , i.e. graphs with @xmath28 edges . a convergent sequence of dense graphs can be represented by an analytic object called a graphon . a _ graphon _ @xmath2 is a measurable function from @xmath29 ^ 2 $ ] to @xmath29 $ ] that is symmetric , i.e. it holds that @xmath30 for every @xmath31 $ ] . the points in @xmath29 $ ] are often referred to as the _ vertices _ of the graphon @xmath2 . if @xmath2 is a graphon , then a _ @xmath2-random graph _ of order @xmath32 is obtained by sampling @xmath32 random points @xmath33 $ ] uniformly and independently , and joining the @xmath34-th and the @xmath35-th vertex by an edge with probability @xmath36 . the _ density _ of a graph @xmath5 in a graphon @xmath2 is the probability that a @xmath2-random graph of order @xmath23 is isomorphic to @xmath5 . if @xmath24 is a convergent sequence of graphs , then there exists a graphon @xmath2 such that @xmath37 for every graph @xmath5 @xcite . this graphon can be viewed as the limit of the sequence @xmath24 . on the other hand , a sequence of @xmath2-random graphs of increasing orders is convergent with probability one and its limit is the graphon @xmath2 . two graphons @xmath38 and @xmath39 are _ weakly isomorphic _ if there exist measure preserving maps @xmath40 and @xmath41 from @xmath29 $ ] to @xmath29 $ ] such that @xmath42 for almost every pair @xmath43 ^ 2 $ ] . if two graphons @xmath38 and @xmath39 are weakly isomorphic , then @xmath44 for every graph @xmath5 . the converse is also true @xcite : if two graphons @xmath38 and @xmath39 satisfy that @xmath44 for every graph @xmath5 , then @xmath38 and @xmath39 are weakly isomorphic . hence , the limit of a convergent sequence of graphs is unique up to weak isomorphism . we finish with giving a formal definition of a finitely forcible graphon : a graphon @xmath2 is _ finitely forcible _ if there exist graphs @xmath6 such that if a graphon @xmath45 satisfies that @xmath46 for @xmath47 , then @xmath48 for every graph @xmath5 . in this section , we recall some basic concepts related to weak regularity for graphs and graphons and cast the lower bound construction of conlon and fox from @xcite in the language of graphons . since we do not use any other type of regularity partition , we will just say `` regular '' instead of `` weak regular '' in what follows . we start with defining the notion for graphs . if @xmath21 is a graph and @xmath49 and @xmath50 two subsets of its vertices , let @xmath51 be the number of edges @xmath52 such that @xmath53 and @xmath54 . a partition of a vertex set @xmath55 of a graph @xmath21 into subsets @xmath56 is said to be _ @xmath0-regular _ if it holds that @xmath57}\frac{e(v_i , v_j)}{|v_i||v_j|}|v_i\cap a||v_j\cap b|\right|\le\varepsilon |v(g)|^2\ ] ] for every two subsets @xmath49 and @xmath50 of @xmath55 . it is known that for every @xmath0 , there exists @xmath58 ( which depends on @xmath0 only ) such that every graph has an @xmath0-regular partition with at most @xmath59 parts @xcite . this dependence of @xmath59 on @xmath0 is best possible up to a constant factor in the exponent as shown by conlon and fox @xcite . we now define the analogous notion for graphons . let @xmath60 ^ 2\to [ 0,1]$ ] be a graphon . if @xmath49 and @xmath50 are two measurable subsets of @xmath29 $ ] , then the _ density @xmath61 between @xmath49 and @xmath50 _ is defined to be @xmath62 we will omit @xmath2 in the subscript if the graphon @xmath2 is clear from the context . note that it always holds that @xmath63 where @xmath64 is the measure of a set @xmath65 . we would like to mention that the density between @xmath49 and @xmath50 is often defined in a normalized way , i.e. it is defined to be @xmath66 , but this is not the case in this paper . a partition @xmath29 $ ] into measurable non - null sets @xmath67 is said to be _ @xmath0-regular _ if it holds that @xmath68}\frac{d(u_i , u_j)}{|u_i||u_j|}|u_i\cap a||u_j\cap b|\right|\le\varepsilon\ ] ] for every two measurable subsets @xmath49 and @xmath50 of @xmath29 $ ] . the upper bound proof translates directly from graphs to graphons and so we get that for every @xmath0 , there exists @xmath58 such that every graphon has an @xmath0-regular partition with at most @xmath59 parts . likewise , the example of conlon and fox from @xcite can be used to obtain a step - graphon @xmath69 such that every @xmath0-regular partition of @xmath69 has at least @xmath70 parts . however , the construction is probabilistic and the description of @xmath69 is thus not explicit . based on this construction , we will define an explicit graphon @xmath71 which has similar properties as @xmath69 for @xmath72 . in fact , a @xmath71-random graph of order @xmath73 for some @xmath74 close to @xmath17 is the graph constructed by conlon and fox in @xcite . fix an integer @xmath75 . the graphon @xmath71 is a step graphon that consists of @xmath76 parts of equal size . each of the parts is associated with a vector @xmath77 . the part of the graphon @xmath71 corresponding to vectors @xmath78 and @xmath79 is constantly equal to @xmath80 where @xmath81 is equal to @xmath82 if @xmath83 $ ] , it is equal to @xmath17 if @xmath84 and to @xmath85 if @xmath86 . in other words , the operator @xmath87 replaces values smaller than @xmath17 or larger than @xmath85 with @xmath17 and @xmath85 , respectively . observe that @xmath88,[0,1])=1/2 $ ] by symmetry . using the chernoff bound , one can show that the measure of the points @xmath89 with @xmath90 is at least @xmath91 . it would be possible to relate the proof presented in @xcite to arguments on regular partitions of graphons . however , the probabilistic nature of the construction would make this technical and obfuscate some simple ideas . because of this , and to keep the paper self - contained , we decided to present a direct proof following the lines of the reasoning given in @xcite . [ thm - conlon - fox ] if @xmath92 and @xmath93 , then every @xmath0-regular partition of the graphon @xmath71 has at least @xmath94 parts . fix an integer @xmath92 . let @xmath95 be the vertices of @xmath71 in the parts associated with vectors @xmath78 whose @xmath34-th coordinate equals @xmath96 . similarly , @xmath97 are the vertices of @xmath71 in the parts associated with vectors @xmath78 whose @xmath34-th coordinate equals @xmath98 . suppose that @xmath71 has an @xmath0-regular partition @xmath67 with @xmath99 and @xmath100 . we say that a part @xmath101 is _ small _ if @xmath102 . note that the sum of the measures of the small parts is at most @xmath103 . for every @xmath104 $ ] , set @xmath105}|u_t\cap v_i^-|\cdot|u_t\cap v_i^+|\;\mbox{. } \label{eq - st}\ ] ] if @xmath106 , then the number of vectors @xmath107 such that @xmath108 and @xmath109 differ in at most @xmath110 coordinates is at most @xmath111 using the chernoff bound and the fact that @xmath92 . hence , for every @xmath106 , the measure of the vertices in the parts associated with vectors that differ from @xmath108 in at most @xmath110 coordinates is at most @xmath112 . consequently , if @xmath101 is not small , then each vertex of @xmath101 contributes to the sum ( [ eq - st ] ) by at least @xmath113 . we conclude that @xmath114 if @xmath101 is not small . we say that the pair @xmath115\times [ k]$ ] is _ useful _ if @xmath116 . let @xmath117 , @xmath104 $ ] , be the number of indices @xmath118 $ ] such that the pair @xmath119 is useful . since each term in the sum ( [ eq - st ] ) is at most @xmath120 and it is at most @xmath121 if @xmath119 is not useful , it follows that @xmath122 . we conclude that @xmath123 unless @xmath101 is small ( recall that @xmath114 if @xmath101 is not small ) . since the sum of the measures of the parts that are not small is at least @xmath124 , we obtain that @xmath125}m_t|u_t|\ge m/32\;\mbox{. } \label{eq - mt}\ ] ] in particular , there exists @xmath126 $ ] such that the sum of the measure of parts @xmath101 such that the pair @xmath127 is useful is at least @xmath128 . fix such an index @xmath129 for the rest of the proof . let @xmath130 be any measurable subset of @xmath131 such that @xmath132 if @xmath127 is useful , and @xmath133 otherwise . similarly , let @xmath134 be any measurable subset of @xmath135 such that @xmath136 if @xmath127 is useful , and @xmath137 otherwise . such sets @xmath130 and @xmath134 exist because @xmath138 for every @xmath139 such that @xmath127 is useful . note that the sets @xmath130 and @xmath134 have the same measure and the choice of @xmath129 implies this measure is at least @xmath140 . let @xmath141 . since the partition @xmath67 is @xmath0-regular , we get that @xmath142}\frac{d(u_i , u_j)}{|u_i||u_j|}|u_i\cap a^+||u_j\cap b|\right|\le\varepsilon\ ] ] and that @xmath143}\frac{d(u_i , u_j)}{|u_i||u_j|}|u_i\cap a^-||u_j\cap b|\right|\le\varepsilon\mbox{.}\ ] ] since @xmath144 for every @xmath104 $ ] , we infer that @xmath145 . on the other hand , the choices of @xmath130 , @xmath134 and @xmath50 imply that @xmath146 and @xmath147 . in particular , it holds that @xmath148 this contradicts the fact that @xmath67 is an @xmath0-regular partition of @xmath71 with @xmath99 and @xmath100 . at first sight , it might seem natural to consider the limit of the sequence of the graphons @xmath71 , @xmath149 , as a candidate for a graphon with no regular partitions with few parts . it can be shown that the sequence of @xmath71 , @xmath149 , is convergent , however , its limit is ( somewhat surprisingly ) the graphon that maps every @xmath150 ^ 2 $ ] to @xmath124 , which has an @xmath0-regular partition with one part for every @xmath0 . we finish this section with sketching a short argument explaining why proposition [ prop - lower - graphon ] is true . the proof of the weak regularity lemma in @xcite is based on iterative refinements of a partition of a graph(on ) , at each step doubling the number of parts and increasing the `` mean square density '' by at least @xmath151 until the partition becomes weakly @xmath0-regular . suppose that there exists a graphon @xmath2 and @xmath152 as in the proposition . we can assume that @xmath153 for every @xmath154 . we start with a trivial partition into a single part and keep refining it until it becomes @xmath155-regular ; let @xmath156 be the number of steps made . we then continue with refining until it becomes @xmath157-regular and let @xmath158 be the total number of steps made till this point . we continue this procedure and define @xmath159 , @xmath160 , in the analogous way . setting @xmath161 , we conclude that the mean square density after @xmath162 steps is at least @xmath163 however , each @xmath159 must be at least @xmath164 by the assumption of the proposition ( otherwise , @xmath2 would have a weak @xmath16-regular partition with fewer than @xmath20 parts ) . consequently , after @xmath165 steps , the mean square density exceeds one , which is impossible . we now define the graphon @xmath166 . we start with defining a tower function @xmath167 as follows : @xmath168 note that @xmath169 , @xmath170 , @xmath171 , @xmath172 , @xmath173 , etc . the notation that we define next is summarized in table [ tab - brackets ] . for @xmath174 , we define @xmath175_1 $ ] to be the smallest integer @xmath32 such that @xmath176 . in particular , @xmath175_1=1 $ ] iff @xmath177 , @xmath175_1=2 $ ] iff @xmath178 , @xmath175_1=3 $ ] iff @xmath179 , etc . this allows us to view the interval @xmath180 as split into _ segments _ @xmath181 , @xmath182 , @xmath183 , etc . and @xmath175_1 $ ] is the index of the segment containing @xmath82 ( numbered from one ) . we then define @xmath184 to be @xmath185_1}-1)\cdot 2^{[x]_1}$ ] , i.e. @xmath184 is the position of @xmath82 in the @xmath175_1$]-th segment if the @xmath175_1$]-th segment is scaled to the unit interval . .the notation used in the definition of the graphon . [ cols="^ , < " , ] fix a graphon @xmath186 that satisfies all the constraints presented in this section . in particular , @xmath186 satisfies the constraints given by lemma [ lm - partition ] and it is a partitioned graphon with ten parts of the sizes as in the graphon and degrees of vertices in these parts as in table [ tab - degrees ] . these ten parts of @xmath186 will be denoted by @xmath187 , @xmath188 , @xmath189 and @xmath190 in correspondence with the parts of the graphon . we will show that @xmath186 is weakly isomorphic to the graphon . the half - graphon @xmath191 , i.e. the zero - one graphon defined by @xmath192 iff @xmath193 , is finitely forcible @xcite , also see @xcite . by lemma [ lm - subforcing ] , there exists a finite set of decorated constraints such that @xmath186 satisfies these constraints if and only if the subgraphon induced by the part @xmath188 is weakly isomorphic to the half - graphon @xmath191 . we insist that @xmath186 satisfies these constraints . let @xmath194 . we use the symbol @xmath195 to refer to the corresponding element of @xmath196 . by the monotone reordering theorem ( see ( * ? ? ? * proposition a.19 ) for more details ) , there exist measure preserving maps @xmath197 and non - decreasing functions @xmath198 , such that @xmath199 for almost every @xmath200 . since we already know that the subgraphon of @xmath186 induced by @xmath201 is weakly isomorphic to @xmath191 , we must have @xmath202 for almost every pair @xmath203 with @xmath204 , @xmath205 for almost every pair @xmath203 with @xmath206 , and @xmath207 for almost every @xmath208 . set @xmath209 for @xmath200 and @xmath210 . for completeness , let @xmath211 and @xmath212 be any measurable maps from @xmath189 and @xmath190 to @xmath213 and @xmath214 such that for any measurable subset @xmath215 of @xmath216 ( @xmath180 , respectively ) , we have @xmath217 ( @xmath218 , respectively ) . each @xmath219 can be viewed as a map from @xmath195 to the part @xmath65 of the graphon @xmath220 . the maps @xmath221 , @xmath222 , @xmath211 and @xmath212 constitute a map @xmath223 from the vertices of @xmath186 to the vertices of @xmath220 and so to those of @xmath166 . we will argue that the map @xmath223 as a map from the vertices @xmath186 to the vertices of @xmath166 is measure preserving and we will show that @xmath224 for almost every pair @xmath43 ^ 2 $ ] . this will prove that the graphons @xmath186 and @xmath166 are weakly isomorphic . so far , we have established that @xmath225 for almost every pair @xmath226 and the map @xmath223 is measure preserving when restricted to @xmath227 . let us consider the decorated constraints depicted in figure [ fig - coordinate ] . let @xmath228 for @xmath229 and @xmath230 . the first constraint implies that the graphon @xmath186 is zero - one valued almost everywhere on @xmath231 and that @xmath232 or @xmath233 has measure zero for almost every pair @xmath234 and for every @xmath230 . the second constraint in figure [ fig - coordinate ] implies for @xmath235 that the measure of @xmath236 is @xmath237 for almost every @xmath229 . hence , it must hold that @xmath238 for @xmath239 and @xmath202 for almost every @xmath240 with @xmath241 , where @xmath235 . similarly , the third constraint in figure [ fig - coordinate ] implies for @xmath242 that the measure of @xmath236 is @xmath243 for almost every @xmath229 . consequently , it holds that @xmath238 for @xmath239 and @xmath202 for almost every @xmath240 with @xmath244 for @xmath242 . we conclude that @xmath223 is a measure preserving map on the whole domain and @xmath224 for almost every pair @xmath245 . the values of the functions @xmath221 can be understood to be the coordinates of the vertices in @xmath187 , respectively , and the coordinate of a vertex @xmath246 is @xmath247 for @xmath248 . this integral is easily expressible as a decorated density expression since it is just the relative edge density ( degree ) of @xmath82 to @xmath188 . this view allows us to speak about segments and subsegments of the parts @xmath187 . the @xmath32-th segment of @xmath195 , @xmath249 , is formed by those @xmath200 such that @xmath250_1=k$ ] . analogously , the values @xmath250_2 $ ] determine the subsegments . we now force that the parts @xmath187 of @xmath186 are split into segments as in @xmath166 . we also force the structure to recognize the first segment through the clique inside @xmath251 and to have the `` successor '' relation on the segments through the structure inside @xmath252 . all three of these aims will be achieved by the decorated constraints given in figure [ fig - segments ] . the four constraints on the first line in figure [ fig - segments ] force the structure on @xmath253 . the line of arguments in this subsection follows those given in @xcite . the first constraint implies that there exists a set @xmath254 of disjoint measurable subsets of @xmath255 such that for almost every @xmath256 , there exists @xmath257 such that @xmath202 for almost every @xmath258 and @xmath205 for almost every @xmath259 and @xmath260 . hence , @xmath202 for almost every @xmath261 and @xmath205 for almost every @xmath262 . we claim that the second constraint together with the structure on @xmath263 yields that for every set @xmath257 there exists an open interval @xmath264 such that @xmath265 and @xmath266 differ on a set of measure zero . note that such an open interval @xmath267 might be empty . since we use an argument of this kind for the first time in this paper , we give more details . if one of the ( non - null ) sets @xmath265 did not have the property , then a random sampling of three points @xmath268 with @xmath269 would satisfy @xmath270 and @xmath271 with positive probability . for such three points , the probability of sampling the additional two points from @xmath188 is @xmath272 and @xmath273 and the triples of points @xmath274 such that the differences @xmath272 and @xmath273 would be bounded away from zero have positive measure . let @xmath275 be the set of open intervals @xmath264 such that @xmath276 and @xmath265 differ on a set of measure zero for some @xmath257 . since the sets in @xmath254 are disjoint and the sets in @xmath275 are open , the intervals of @xmath275 are also disjoint . the third constraint implies that if @xmath277 for some @xmath278 , then the measure of the interval @xmath267 , assuming it is non - empty , and the measure of the interval @xmath279 are the same . again , we provide a detailed justification since we use an argument of this kind for the first time . almost every choice of the three roots @xmath256 , @xmath280 and @xmath281 ( the order follows that in the figure ) satisfies that @xmath282 ( because of the non - edge between @xmath82 and @xmath283 and the edge between @xmath283 and @xmath284 ) and that there exists @xmath278 such that @xmath285 ( because of the edge between @xmath82 and @xmath284 ) . the left side is then equal to the measure of @xmath276 , which is the measure of @xmath267 . the right side is equal to the measure of those @xmath286 such that @xmath287 and @xmath288 . hence , the right side is equal to @xmath289 . since this holds for almost every triple @xmath82 , @xmath283 and @xmath284 , we conclude that the measure of each non - empty interval @xmath278 is @xmath289 . consequently , each non - empty interval @xmath267 must be of the form @xmath290 for some @xmath291 . since the intervals of @xmath275 are disjoint , there can only be a finite number of intervals to the left of each interval of @xmath275 . this implies that the set @xmath275 is countable . finally , the last constraint on the first line yields that @xmath292 however , this equality can hold only if the intervals contained in @xmath275 are exactly the intervals @xmath293 , @xmath294 . we conclude that @xmath295 and @xmath296 are equal for almost every pair @xmath297 . we now analyze the four constraints on the second line in figure [ fig - segments ] . fix @xmath298 . the first constraint implies that for every @xmath257 , there exists @xmath299 such that @xmath202 for almost every @xmath300 and @xmath205 for almost every @xmath301 . the second constraint yields that the measure of @xmath302 is the same as the measure of @xmath265 . the third constraint implies that there exists an open interval @xmath303 such that @xmath302 and @xmath304 differ on a set of measure zero . finally , the last constraint on the second line yields that each of the intervals @xmath303 is of the form @xmath290 for some @xmath291 . since the length of @xmath303 is the same as the measure of @xmath265 , we conclude that if @xmath305 , then @xmath306 . hence , @xmath295 and @xmath296 are equal for almost every pair @xmath307 , @xmath298 . let us turn our attention to the three constraints on the third line in figure [ fig - segments ] . the first constraint implies that there exists a subset @xmath308 of @xmath309 such that @xmath202 for almost every @xmath310 and @xmath205 for almost every @xmath311 . the second constraint yields that @xmath308 is a subset of @xmath302 for some @xmath257 . finally , the third constraint says that the square of the measure of @xmath308 is @xmath312 , i.e. the measure of @xmath308 is @xmath124 . however , this is only possible if @xmath308 and @xmath313 differ on a set of measure zero . we conclude that @xmath295 and @xmath296 are equal for almost every pair @xmath314 . it remains to analyze the three constraints on the last line in figure [ fig - segments ] . the first constraint yields that for every @xmath294 , there exists @xmath315 such that @xmath316 for almost every @xmath317 and @xmath318 for almost every @xmath319 . the second constraint yields that the measure of @xmath320 is @xmath321 , and the third constraint that @xmath320 is a subset of @xmath322 except for a set of measure zero . hence , @xmath320 and @xmath322 differ on a set of measure zero , and we can conclude that @xmath295 and @xmath296 are equal for almost every pair @xmath323 . in this subsection , we will force a representation of the tower inside @xmath324 . this is achieved using the constraints depicted in figure [ fig - tower ] . before analyzing these constraints , we give an analytic observation based on ( * ? ? * proof of lemma 3.3 ) . [ lm - square ] let @xmath325 be a measurable function . if @xmath326 for almost every @xmath327 , then @xmath328 for almost every @xmath174 . 16.2 cm 16.2 cm the constraint on the first line in figure [ fig - tower ] yields that @xmath329 for almost every @xmath330 and @xmath331 all in the same segment ( this is implied by the presence of the edges between the roots ) . by lemma [ lm - square ] , we get that @xmath332 for almost every @xmath333 and @xmath334 such that @xmath335_1=[g_a(y)]_1 $ ] . since the equality holds for almost every @xmath333 and @xmath334 , it actually holds that @xmath336 for almost every @xmath337 and @xmath338 such that @xmath335_1=[g_a(y)]_1 $ ] . the cauchy - schwartz inequality yields that there exist @xmath339 , @xmath294 , such that @xmath340 for almost every @xmath337 , @xmath338 and @xmath341 with @xmath335_1=[g_a(y)]_1=k$ ] . hence , @xmath342 for almost every @xmath337 and @xmath341 with @xmath335_1=[g_c(z)]_1=k$ ] and @xmath343 for almost every @xmath337 and @xmath341 with @xmath335_1\not=[g_c(z)]_1 $ ] . along the same line , the constraint on the second line implies that there exist @xmath344 , @xmath294 , such that @xmath345 for almost every @xmath346 , @xmath338 and @xmath347 with @xmath348_1=[g_a(y)]_1=k$ ] . consequently , @xmath349 for almost every @xmath350 and @xmath351 with @xmath352_1=[g_b(z)]_1=k$ ] and @xmath343 for almost every @xmath350 and @xmath351 with @xmath352_1\not=[g_b(z)]_1 $ ] . almost every choice of the roots in the first constraint on the third line satisfies that all the roots belong to the same segment and this segment must be the first segment because of the edge between the two roots from @xmath309 . hence , this constraint implies that @xmath353 , i.e. @xmath354 as desired . let us now look at the second constraint . almost every choice of the roots satisfies that if the right root from @xmath355 is in the @xmath32-th segment , then the roots from @xmath255 and @xmath309 are also in the @xmath32-th segment and the left root from @xmath355 is in the @xmath356-th segment . since for every @xmath32 the choice of such roots has positive probability , the constraint implies that the following holds for every @xmath294 : @xmath357 @xmath358 hence , it holds that @xmath359 . we conclude that @xmath295 and @xmath296 are equal for almost every pair @xmath360 . the first three constraints on the fourth line in figure [ fig - tower ] yield that for every @xmath294 either @xmath318 for almost every @xmath346 in the @xmath32-th segment and almost every @xmath361 or there exists @xmath362 such that @xmath316 for almost every @xmath346 in the @xmath32-th segment and almost every @xmath361 in the @xmath362-th segment and @xmath318 for almost every @xmath361 not in the @xmath362-th segment . the last constraint on the fourth line yields that @xmath363 . we now show that the constraint on the fifth line implies that @xmath362 exists and @xmath364 for every @xmath294 . for almost every choice of the roots in the constraint on the fifth line , if the right root from @xmath365 belongs to the @xmath32-th segment , the left root from @xmath365 belongs to the @xmath356-th segment and the left root from @xmath355 belongs to the @xmath366-th segment . we derive that this constraint implies that @xmath367 we conclude that @xmath368 and so @xmath364 . consequently , @xmath295 and @xmath296 are equal for almost every pair @xmath369 . the constraint on the last line in figure [ fig - tower ] yields by considering a choice of the root in the @xmath32-th segment of @xmath365 that @xmath370 we conclude that @xmath371 . hence , @xmath295 and @xmath296 are equal for almost every pair @xmath372 . in analogy to subsection [ sub - segments ] , the first two constraints in figure [ fig - subsegments-1 ] for @xmath373 yield that there exists a set @xmath254 of disjoint open subintervals of @xmath180 such that @xmath202 for almost every @xmath374 for some @xmath257 and @xmath205 for almost all other pairs @xmath375 . the third constraint implies that every set @xmath376 is a subset of @xmath377 except for a set of measure zero for some @xmath294 . hence , each interval @xmath257 is a subinterval of @xmath378 for some @xmath294 . the fourth constraint with @xmath373 yields that the length of each interval @xmath265 is @xmath379 . finally , the first constraint on the second line can hold only if each interval @xmath378 contains @xmath380 such intervals @xmath265 . we conclude that @xmath295 and @xmath296 are equal for almost every pair @xmath375 . for @xmath381 , the first , second and fourth constraints on the first line yield that for each @xmath257 there exists an open interval @xmath267 of the same length as @xmath265 such that @xmath202 for almost every @xmath382 and @xmath205 for almost every @xmath383 . the last constraint on the second line in figure [ fig - subsegments-1 ] gives that the intervals @xmath267 follow in the same order as the intervals @xmath265 . hence , @xmath295 and @xmath296 are equal for almost every pair @xmath384 . the set of constraints in figure [ fig - subsegments-2 ] is analogous to those in figure [ fig - subsegments-1 ] . the main difference is the third constraint , which forces that if an interval @xmath265 is a subinterval of an interval @xmath378 , then @xmath385 . hence , the length of @xmath265 must be @xmath386 . the first constraint on the second line then forces that the interval @xmath378 must contain @xmath387 such intervals @xmath265 and the order of the corresponding pairs of intervals is forced by the last constraint . we can now conclude that @xmath295 and @xmath296 are equal for almost every pair @xmath388 . we now analyze the constraints from figure [ fig - subsegments-3 ] . as in the previous cases , the three constraints on the first line force that each interval @xmath378 , @xmath294 , contain disjoint open intervals @xmath389 and @xmath390 , each of length @xmath386 , such that @xmath202 for almost every @xmath391 and @xmath205 for almost every @xmath392 . fix a choice of the roots in the first constraint on the second line in figure [ fig - subsegments-3 ] ; in almost every choice of the roots , all the three roots belong to the same segment . suppose that they belong to the @xmath32-th segment . the right side is equal to @xmath393 for almost all choices of the roots ( since the structure of @xmath394 has already been forced ) and the left side is equal to @xmath395 multiplied by the number of choices of @xmath396 and @xmath397 such that @xmath396 is contained in the subsegment of the left root and @xmath397 in the subsegment of the right root . since the left side and the right side must be equal for almost all choices of the roots , we conclude that for any pair of subsegments @xmath398 and @xmath399 of the @xmath32-th segment there exists a unique index @xmath34 such that @xmath396 is contained in @xmath398 and @xmath397 in @xmath399 . the last constraint in figure [ fig - subsegments-3 ] enforces that for any fixed subsegment @xmath398 of the @xmath32-th segment and any two subsegments @xmath399 and @xmath400 such that @xmath399 precedes @xmath400 , the pairs @xmath401 and @xmath402 satisfy that the interval @xmath396 precedes the interval @xmath403 . this implies that @xmath295 and @xmath296 are equal for almost every pair @xmath404 . in this section , we force the structure of the graphon inside @xmath405 , @xmath406 , @xmath407 and @xmath408 . this will be achieved using the constraints depicted in figure [ fig - binary ] . the first constraint on the first line in the figure causes that for almost every @xmath409 , there exists @xmath410 such that @xmath202 for almost every @xmath411 with @xmath412 and @xmath295 for almost every @xmath411 with @xmath413 . the second constraint on the line causes that for almost every @xmath409 , it holds that @xmath414_1 } \xi_x = 1 - 2^{-[g_e(x)]_1}-g_e(x)$ ] . it follows that @xmath415_1}-g_e(x)}{2^{-[g_e(x)]_1}}=1-\llbracket g_e(x)\rrbracket_1\ ] ] for almost every @xmath409 . we conclude that @xmath295 and @xmath296 are equal for almost every pair @xmath416 . the first constraint on the second line forces that @xmath186 is almost everywhere @xmath17 or almost everywhere @xmath85 on each product of a subsegment of @xmath417 and a segment of @xmath355 . fix a segment @xmath418 of @xmath417 . the second constraint forces that for almost every @xmath361 the measure of @xmath419 such that @xmath202 is exactly half of the measure of @xmath418 . since the measure of @xmath361 such that @xmath202 is @xmath420_1 $ ] for almost every @xmath409 because of the third constraint on the second line , the choice of the segments and subsegments where @xmath186 is one almost everywhere is unique . so , we get that @xmath295 and @xmath296 are equal for almost every pair @xmath421 . the first constraint on the third line implies that for almost every @xmath422 there exist @xmath423 , @xmath294 , such that @xmath202 for almost every @xmath424 with @xmath425_1=k$ ] and @xmath426 and @xmath205 for almost every other @xmath361 . consider a possible choice of the roots in the second constraint on the third line . for almost every such choice of the roots , if the leftmost root from @xmath255 lies in the @xmath32-th segment , then the middle root from @xmath255 is in the @xmath427-th segment ( because of the already enforced structure of @xmath186 on @xmath428 in particular ) and the rightmost root from @xmath255 is in the @xmath429-th segment where @xmath430 . the left side of the constraint is equal to @xmath431 . the right side of the constraint is equal to @xmath432 where @xmath82 is the root from @xmath433 . this implies that @xmath434 where for almost every @xmath82 from the @xmath32-th segment of @xmath433 and @xmath430 . finally , almost every choice of the roots in the last constraint on the third line satisfies that if the leftmost root from @xmath255 lies in the @xmath32-th segment , then the middle root from @xmath255 is in the @xmath427-th segment and the rightmost root from @xmath255 is in the @xmath429-th segment where @xmath435 . hence , @xmath436 for almost every @xmath82 from the @xmath32-th segment of @xmath433 and @xmath435 . we conclude that @xmath295 and @xmath296 are equal for almost every pair @xmath437 . the first constraint on the fourth line implies that for almost every @xmath361 , if the measure of @xmath82 with @xmath438 from the @xmath32-th segment of @xmath439 is positive , then @xmath202 for almost every @xmath82 from the @xmath32-th segment of @xmath433 . analogously , the second constraint yields that for almost every @xmath361 , if the measure of @xmath82 with @xmath438 from a certain subsegment of @xmath439 is positive , then @xmath202 for almost every @xmath82 from the corresponding subsegment of @xmath433 . consequently , @xmath205 for almost every pair @xmath440 such that @xmath441 . since the last constraint implies that the integral of @xmath186 over @xmath408 is the same as the integral of @xmath166 over @xmath442 , it holds that @xmath295 and @xmath296 are equal for almost every pair @xmath440 . in this subsection , we focus on the pair @xmath433 and @xmath417 of the parts . the first constraint in figure [ fig - linear ] yields that @xmath205 for almost every @xmath443 such that the segments of @xmath444_1\not=[g_e(y)]_1 $ ] , i.e. the segments of @xmath82 and @xmath424 are different . the second constraint implies that for almost every @xmath422 there exists @xmath410 such that @xmath202 for almost every @xmath445 such that @xmath444_1=[g_e(y)]_1 $ ] and @xmath446 and @xmath205 for almost every @xmath445 such that @xmath444_1=[g_e(y)]_1 $ ] and @xmath447 . the third constraint implies that almost every pair of @xmath82 and @xmath283 from the same segment of @xmath433 such that @xmath448 satisfies that @xmath449 . in order to show that @xmath295 and @xmath296 are equal for almost every pair @xmath443 , it is enough to show that @xmath450_1 - 1)^{1/2}\left(\frac{1}{2}-\llbracket g_g(x)\rrbracket_1\right)\label{eq - linear}\ ] ] for almost every @xmath422 . almost every choice of the roots in the last constraint on the first line in figure [ fig - linear ] satisfies that the root from @xmath21 is from the first segment . hence , this constraint implies that almost every @xmath422 with @xmath444_1=1 $ ] satisfies that @xmath451 . since @xmath169 and @xmath452 for such @xmath422 , we obtain that ( [ eq - linear ] ) holds for almost every @xmath82 from the first segment of @xmath433 . we now analyze the constraint on the second and third lines in figure [ fig - linear ] . note that almost every choice of the roots satisfies that if the left root from @xmath49 is in the @xmath32-th segment , then the root from @xmath21 is also in the @xmath32-th segment and the right root from @xmath49 and the right root from @xmath453 are in the @xmath356-th segment . in particular , @xmath454 for almost every choice of the roots . rewriting the densities using the already established structure of the graphon , we obtain that almost every @xmath422 with @xmath444_1=k\ge 2 $ ] satisfies that @xmath455 since it holds that @xmath456 , we can rewrite the right side as follows @xmath457 which can be transformed to @xmath458 this implies that @xmath410 is equal to @xmath459_1 - 1)^{1/2}\left(\frac{1}{2}-\llbracket g_g(x)\rrbracket_1\right)\mbox { or } \ ; \frac{1}{2}-t([g_g(x)]_1 - 1)^{1/2}\left(\frac{1}{2}-\llbracket g_g(x)\rrbracket_1\right)\ ] ] for almost every @xmath422 not contained in the first segment of @xmath433 . recall that almost every pair of @xmath82 and @xmath283 from the same segment of @xmath433 such that @xmath448 , which is equivalent to @xmath460 , satisfies that @xmath449 . this implies that the latter of two options for the values @xmath410 holds for almost no @xmath422 and thus ( [ eq - linear ] ) also holds for almost every @xmath422 not contained in the first segment of @xmath433 . we conclude that @xmath295 and @xmath296 are equal for almost every pair @xmath443 . we now enforce the unique structure inside the pairs @xmath461 and @xmath462 . the first constraint in figure [ fig - dot ] yields that @xmath205 for almost every @xmath463 with @xmath464_1\not=[g_e(y)]_1 $ ] , i.e. the segments of @xmath82 and @xmath424 are different . the second constraint implies that for almost every @xmath465 there exists @xmath410 such that @xmath202 for almost every @xmath445 with @xmath464_1=[g_e(y)]_1 $ ] and @xmath466 and @xmath205 for almost every @xmath445 with @xmath464_1=[g_e(y)]_1 $ ] and @xmath467 . in order to show that @xmath295 and @xmath296 are equal for almost every pair @xmath463 , it is enough to show that @xmath468_2^{\pm1},[g_f(x)]_{3}^{\pm1}\right\rangle}}{4t([g_f(x)]_1 - 1 ) } \label{eq - dot-1}\ ] ] for almost every @xmath465 . let @xmath82 and @xmath283 be two elements of @xmath439 from the same segment , say , the @xmath32-th segment . by the structure on @xmath408 , the measure of @xmath361 such that either @xmath202 and @xmath469 or @xmath205 and @xmath470 is equal to the number of different pairs of coordinates in @xmath471 and @xmath472 multiplied by @xmath473 for almost any pair @xmath82 and @xmath283 . this measure can be rewritten as @xmath474 consider now the constraint on the second line in figure [ fig - dot ] . for almost every choice of the roots , their segments are the same and so they all belong to the first segment of their parts ( because of the edge between the two roots from @xmath309 ) . let @xmath82 be the right root that belongs to @xmath439 and @xmath283 the left one . because of the already enforced structure of the graphon , the subsegment of @xmath283 is @xmath464_{3}$ ] in almost every choice of the roots . using ( [ eq - dot - measure ] ) with @xmath169 and @xmath170 , we derive that it holds for almost every @xmath465 that belongs to the first segment that @xmath475_2^{\pm1},[g_f(x)]_{3}^{\pm1}\right\rangle}}{4}\;\mbox{.}\ ] ] we conclude that the equation ( [ eq - dot-1 ] ) holds for almost every @xmath465 from the first segment . we now consider the constraint on the third and fourth lines in figure [ fig - dot ] . in almost every choice of the roots , the segment of the two roots from @xmath439 is one higher than the segment of the root from @xmath255 adjacent to the root from @xmath355 . also note that for almost every @xmath465 that belongs to the second or higher segment , there is a set of positive measure of possible choices of the roots . let @xmath82 be the right root from @xmath439 , @xmath283 the left one and @xmath32 the common index of their segment . the constraint implies for almost every choice of the roots that @xmath476 this expression readily transforms to @xmath477 since we can choose the roots for almost every @xmath465 that belongs to the second or higher segment with positive probability , almost every such @xmath465 satisfies ( [ eq - dot-1 ] ) . we conclude that @xmath295 and @xmath296 are equal for almost every pair @xmath463 . in the analogy to the first two constraints on the first last in figure [ fig - dot ] , the first two constraints on the last line in the figure yield that for almost every @xmath347 , there exists @xmath478 such that @xmath479 for almost every @xmath480 with @xmath481_1=[g_g(y)]_1 $ ] and @xmath482 and @xmath483 for almost every @xmath480 with @xmath481_1\not=[g_g(y)]_1 $ ] or @xmath484 . we now consider the last two constraints on the last line in figure [ fig - dot ] . fix a choice of roots @xmath465 and @xmath347 . almost every such choice of roots satisfies @xmath464_{2,3}=[g_d(z)]_{2,3}$ ] , which implies that @xmath464_1=[g_d(z)]_1 $ ] . the last but one constraint yields that @xmath485 for almost every @xmath445 , @xmath486 , @xmath487_1=[g_f(x)]_1 $ ] , @xmath488_1=[g_d(z)]_1 $ ] , @xmath466 and @xmath489 . similarly , the last constraint yields that @xmath490 for almost every @xmath445 , @xmath486 , @xmath487_1=[g_f(x)]_1 $ ] , @xmath488_1=[g_d(z)]_1 $ ] , @xmath467 and @xmath491 . the structure of the graphon @xmath186 on @xmath492 implies that @xmath493 for almost every pair of @xmath465 and @xmath347 such that @xmath464_{2,3}=[g_d(z)]_{2,3}$ ] and @xmath464_1=[g_d(z)]_1=k$ ] . the expression ( [ eq - dot-1 ] ) can be rewritten ( for a particular choice of @xmath82 and @xmath494 ) as @xmath495_2^{\pm1},[g_d(z)]_{3}^{\pm1}\right\rangle}}{4t(k-1)}\;\mbox{. } \label{eq - dot-3}\ ] ] we get by substituting ( [ eq - dot-3 ] ) in ( [ eq - dot-2 ] ) that @xmath496_2^{\pm1},[g_d(z)]_{3}^{\pm1}\right\rangle}}{4t(k-1)^{1/2}}\right)}\ ] ] for almost every @xmath347 that belongs to the @xmath32-th segment . we conclude that @xmath497 and @xmath498 are equal for almost every pair @xmath499 . we now focus on the values of @xmath186 on @xmath500 . the first constraint in figure [ fig - main ] implies that @xmath205 for almost every @xmath501 with @xmath502_1\not=[g_e(y)]_1 $ ] . let us consider the second constraint in the figure . for almost every choice of the roots , all the roots belong to the same segment in their respective parts . let @xmath32 be the index of this segment and let @xmath494 and @xmath503 be the choices of the bottom and the top roots that belong to @xmath365 . further , let @xmath398 and @xmath399 be the subsegments of @xmath417 corresponding to the subsegments of @xmath494 and @xmath503 in @xmath365 , respectively . for almost every choice of the roots , the root from @xmath417 belongs to @xmath398 . the second constraint yields that it holds for almost all @xmath504 that @xmath505 along the same lines , the third constraint yields that it holds for almost all @xmath506 that @xmath507 by lemma [ lm - square ] , the identity ( [ eq - main-2 ] ) implies that @xmath508 for almost every @xmath504 . we derive from ( [ eq - main-1 ] ) and ( [ eq - main-3 ] ) using cauchy - schwartz inequality that @xmath509 for almost every @xmath504 and every @xmath510 . since @xmath511 and @xmath512 for almost every such pair @xmath82 and @xmath424 , we conclude that @xmath295 and @xmath296 are equal for almost every pair @xmath513 . fix @xmath210 . the first constraint in figure [ fig - degrees ] implies that @xmath514 for almost every @xmath515 for some @xmath516 . by lemma [ lm - square ] , this also implies that @xmath517 for almost every @xmath518 . hence , there exists a function @xmath519 $ ] such that @xmath520 for almost every @xmath518 and @xmath521 . since the last constraint on the first line yields that @xmath522 for almost every @xmath200 and it holds that @xmath523 for almost every @xmath200 because of the already enforced structure , we conclude that @xmath295 and @xmath296 are equal for almost every pair @xmath524 . the three constraints on the second line in figure [ fig - degrees ] clearly implies that @xmath295 and @xmath296 are equal for almost every pair @xmath525 . again fix @xmath210 . the two constraints on the last line in figure [ fig - degrees ] give using lemma [ lm - square ] that @xmath526 for almost every @xmath200 for some @xmath516 . however , this is only possible if @xmath527 for almost every @xmath200 and almost every @xmath528 . the right side values of the two constraints yield that the values of @xmath516 matches the corresponding values in @xmath166 . so , we can conclude that @xmath295 and @xmath296 are equal for almost every pair @xmath529 . since we have shown that the constraints in figures [ fig - coordinate][fig - degrees ] imply that @xmath295 and @xmath296 are equal for almost every pair @xmath82 and @xmath424 , we have established that the graphon is finitely forcible . this completes the proof of theorem [ thm - main ] . proposition [ prop - lower - graphon ] implies that it is not possible to remove @xmath530 completely from the denominator in the exponent in theorem [ thm - main ] . however , our construction can be modified to replace @xmath531 with a faster growing function of @xmath532 , e.g. with @xmath533 , which would replace the function @xmath530 with a slower growing function of @xmath1 . the authors are grateful to jacob fox for sparking their interest in finitely forcible graphons with weak regular partitions requiring a large number of parts and for pointing out the tightness of their construction as formulated in proposition [ prop - lower - graphon ] . they would also like to thank andrzej grzesik , sune jakobsen and fiona skerman for discussions on the limit of the graphons @xmath534 , and taisa lopes martins for her comments on one of the drafts of the paper . c. borgs , j. chayes , l. lovsz , v.t . ss , b. szegedy and k. vesztergombi : _ graph limits and parameter testing _ , in : proceedings of the 38rd annual acm symposium on the theory of computing ( stoc ) , acm , new york , 2006 , 261270 . a. thomason : _ random graphs , strongly regular graphs and pseudorandom graphs _ , in : c. whitehead ( ed . ) , surveys in combinatorics 1987 , london mathematical society lecture note series , 123 , cambridge univ . press , cambridge ( 1987 ) , 173-195 .

we suggest a variant of the recently proposed experiment for the generation of a new kind of schrdinger - cat states , using two coupled parametric down - converter nonlinear crystalsf . de martini , phys . rev . lett . * 81 * , 2842 ( 1998) . we study the parametric oscillator case and find that an entangled schrdinger - cat type state of two cavities , whose mirrors are placed along the output beams of the nonlinear crystals , can be realized under suitable conditions . 2

quant-ph9902082

schrdinger - cat states @xcite are most important in the domain of fundamental quantum mechanics , since the study of their progressive decoherence @xcite would provide a better understanding of the transition from the _ quantum _ to the _ classical _ world @xcite . however , due to their extreme sensitivity to the decoherence caused by the interaction with the environment , such linear superpositions of macroscopically distinguishable states are difficult to produce and to observe @xcite . in the last few years , a major effort in this field has led to the experimental production and detection of _ mesoscopic _ superpositions of distinct states , both in the context of the single - mode microwave cavities @xcite and of the dynamics of the center of mass motion of a trapped ion @xcite . on the other hand , entanglement has been widely recognized as one of the essential and most puzzling features of quantum mechanics @xcite , in that it allows the existence of _ quantum correlated _ states of two noninteracting subsystems : entangled states play a crucial role in the so called einstein , podolsky and rosen ( epr ) paradox @xcite , and are essential in the rapidly growing field of quantum information , as they allow the feasibility of quantum state teleportation @xcite , quantum cryptography @xcite , and quantum computation @xcite . in two recent papers @xcite , one of us has proposed an original scheme for the generation of a new kind of _ amplified _ schrdinger - cat type states . it is based on the new concept of _ quantum injection _ into an optical parametric amplifier ( opa ) operating in _ entangled _ configuration . as a relevant variant and a natural extension of the above scheme , in the present work we analyze the case of the quantum injection in an optical parametric oscillator ( opo ) in which two optical cavities are added to the opa scheme considered in ref . @xcite : refer to fig . [ fg : scheme ] . since the presence of the cavities leads to a large enhancement of the nonlinear ( nl ) parametric interaction , the number of the photon couples which are expected to be generated , in practical conditions , by the opo scheme is far larger than in the amplifier condition : in addition , the generation of parametrically coupled quasi - coherent fields represents in this context an appealing perspective . the schrdinger - cat state that has been put forward in ref . @xcite and is being analyzed , in a more detailed fashion , in the present paper , is a superposition of two macroscopic states which are distinguished by their polarization . it can be considered as a sort of amplified version of the polarization - entangled states which have been widely used in the last few years for the demonstration of the violation of bell s inequality @xcite , of teleportation @xcite , and for the generation of greenberger - horne - zeilinger ( ghz ) states @xcite . the present paper is organized as follows : in sec . [ spdc ] we briefly describe the process of type - ii parametric down conversion , with an emphasis on the kind of entangled states usually produced in these experiments , and on the state we want to generate . in sec . [ scheme ] we outline the experimental apparatus needed for the realization of our scheme . we devote sec . [ evol ] to the presentation of the dynamical time evolution of the density matrix and of the wigner function in our system , and sec . [ stability ] to the discussion of the stability conditions for our parametric oscillator . in sec . [ choice ] we set the initial conditions for the two coupled nonlinear crystals and the two cavities , whereas the way in which the cat state is produced is discussed in detail in sec . [ genera ] . [ detect ] is devoted to the presentation of the three methods we propose for detecting the schrdinger - cat state : photodetection ( sec . [ photo ] ) , measurement of the second - order quantum coherence ( sec . [ corr ] ) , and wigner function reconstruction ( sec . [ wigner ] ) . we finally summarize and discuss our results in sec . [ conclu ] . the appendix is devoted to the development of the small interaction time approximation . 2 let us first describe the kind of states commonly generated in the experiments aimed at the violation of the bell s inequalities . in these experiments the nl crystal ( typically beta - barium - borate : bbo ) is cut for type ii phase matching where the two down - converted photons are emitted into two cones , one `` ordinary '' polarized ( @xmath2 ) , the other `` extraordinary '' polarized ( @xmath3 ) . when the angle between the pump direction and the nonlinear crystal optical axis is sufficiently large @xcite , the two cones mutually intersect along two lines , lying on opposite sides of the pump beam direction . these ones identify the output modes of the parametric down conversion : @xmath4 ( @xmath5 , @xmath6 ) . therefore the field belonging to the modes @xmath4 can be simultaneously @xmath3- and @xmath2-polarized . in typical conditions , the output state of the emitted photon couple may be expressed by @xcite @xmath7 since we have , for each couple , four degrees of freedom involved , i.e. 2 states of orthogonal linear polarization @xmath3 , @xmath2 for each mode @xmath4 , we can rewrite state ( [ eq : pola ] ) in the more precise form @xmath8 which will be used in the following . the `` schrdinger - cat state '' we want to generate is a sort of amplification of this state , that is , it may be expressed in the form @xmath9 where @xmath10 is a state with a large number of photons in some sense , and the states @xmath11 are to be interpreted here as squeezed vacuum states . this kind of state is different from the traditional schrdinger cat states discussed in the quantum optics literature @xcite , where one has a _ single _ mode of the electromagnetic field in a superposition of two macroscopic states with different phases of the field . the state ( [ eq : catpro ] ) is a _ nonlocal _ superposition in which a macroscopic optical field is `` localized '' simultaneously either in the @xmath3- or in the @xmath2-polarized mode . in other words it is a state more similar to nonlocal field states such as @xmath12 where the field can be simultaneously in one cavity or in another cavity and whose generation is discussed in @xcite . we shall present here an experimental scheme for the generation of a state which is actually a mixed state , but nonetheless , has the same structure of the state of eq . ( [ eq : catpro ] ) , that is , can be represented by the density operator @xmath13 where @xmath14 is a two - mode mixed state with a large number of photons , @xmath15 is a two - mode mixed state with a small number of photons and @xmath16 and @xmath17 are the interference terms . we shall consider an experimental arrangement , fig . [ fg : scheme ] , based on the one proposed in ref . @xcite and similar to that adopted in refs . @xcite to show the realization of inducing coherence , without induced emission . two down - conversion nl crystals , are arranged in such a way that the two corresponding idlers beams are aligned along a common direction @xmath18 . moreover , both idler beams and the signal beam of one nl crystal ( with wave - vector @xmath19 ) are placed within couples of mirrors . this scheme can be thought of to realize the coupling of two nondegenerate opos . the signal beam of the other crystal , emitted along the direction @xmath20 triggers the photodetector @xmath21 . the directions @xmath20 , @xmath18 and @xmath19 are selected to realize for both nl crystals the type - ii phase matching described before . these beams are then associated with six modes , with annihilation operator @xmath22 , @xmath23 , @xmath24 , @xmath25 , @xmath26 and @xmath27 . note that the first two annihilation operators refer to traveling - waves , while the last four refer to cavity modes . the dynamics of the system is determined by the nonlinear parametric interaction at each crystal and by the damping terms associated with losses and dissipation inside the cavities @xcite , as we shall see in the next section . the partial hamiltonian operators describing the unitary dynamics inside the crystals are given by @xcite [ eq : hnl ] @xmath28 where @xmath29 , @xmath30 , @xmath31 is the second - order nonlinear susceptibility of the crystals , and @xmath32 ( @xmath33 ) is the pump intensity in crystals 1 and 2 , respectively , which is assumed to be `` classical '' . due to the explicit presence of dissipation in this problem , one has to write the master equation for the reduced density matrix of the combined system which arises from the hamiltonian terms ( [ eq : hnl1 ] ) and ( [ eq : hnl2 ] ) and from the damping terms @xmath34 for @xmath35 . since the damping constants @xmath36 are essentially connected to the transmittivity of the mirrors , it is quite natural to assume @xmath37 and @xmath38 . upon writing the full master equation for the total density matrix @xmath39 of the ( six - mode ) system , it appears clear that the dynamics of the six modes actually decouples into two independent dynamics for two groups of three modes . in fact , one has @xmath40 where @xmath41 \label{eq : parmeq } \\ & + & \kappa_{2}(2\hat{a}_{2o}\rho_{\rm t}\hat{a}_{2o}^{\dagger } -\hat{a}_{2o}^{\dagger}\hat{a}_{2o}\rho_{\rm t } -\rho_{\rm t}\hat{a}_{2o}^{\dagger}\hat{a}_{2o } ) \nonumber \\ & + & \kappa_{3}(2\hat{a}_{3e}\rho_{\rm t}\hat{a}_{3e}^{\dagger } -\hat{a}_{3e}^{\dagger}\hat{a}_{3e}\rho_{\rm t } -\rho_{\rm t}\hat{a}_{3e}^{\dagger}\hat{a}_{3e})\ ; , \nonumber\end{aligned}\ ] ] and @xmath42 @xmath43 is identical to @xmath44 up to the substitution @xmath45 and @xmath46 . as a consequence , the complete time evolution will be of the form @xmath47 from eq . ( [ eq : timev ] ) it is clear that if the initial condition is factorized , namely , if @xmath48 the state will remain factorized at all times , unless specifically designed conditional measurements @xcite are performed on the system ( for example on the mode @xmath20 ) . due to the decoupling between the @xmath49 and the @xmath50 modes , we can simply restrict ourselves to the investigation of the three - mode problem described by the master equation ( [ eq : meq ] ) with ( [ eq : parmeq ] ) , and we shall drop the subscript @xmath3 and @xmath2 when not needed . the wigner function @xcite @xmath51 , with @xmath52 @xmath53 , resulting from this density matrix @xmath54 will then be a function of six real variables ( or three complex variables ) . its time evolution , upon evaluating the commutator and the damping terms and after some lengthy algebra , is described by the six - dimensional fokker - planck equation @xmath55 where the vector @xmath56 , the matrix @xmath57 , and @xmath58 the solution to eq . ( [ eq:6fp ] ) can be written @xcite as the integral @xmath59 where [ eq : where ] @xmath60\ ; , \nonumber \\ g(t ) & = & \exp\left(-\gamma t\right)\ ; , \label{eq : gt}\end{aligned}\ ] ] and @xmath61 @xmath62 being the transposed of the matrix @xmath63 . the stability properties of the system are intimately connected to the threshold of the overall opo consisting of nl1 and nl2 . below threshold , the system is stable and reaches a stationary state , since all eigenvalues of @xmath64 have positive real parts . on the other hand , above threshold the system is unstable and its energy exponentially increases , because some eigenvalues of @xmath64 have negative real part . this result can be easily checked in the case in which the parametric oscillator associated with nl2 is decoupled from nl1 ( @xmath65 ) : in this case modes @xmath18 and @xmath19 decouple from mode @xmath20 , and we end up with a four - dimensional problem for the modes @xmath18 and @xmath19 , described by a fokker - planck equation of the same type as eq . ( [ eq:6fp ] ) , but with @xmath66 and @xmath67 . in this case the ( doubly degenerate ) eigenvalues of @xmath64 are @xmath68 and the stability condition becomes @xmath69 which coincides with the customary threshold for the parametric oscillator @xcite . however , if we turn on the first parametric amplifier ( @xmath70 ) then the problem turns from 4-dimensional to 6-dimensional , as we have seen : the eigenvalues of @xmath64 change and it is in principle possible to change the threshold , i.e. , the stability condition . as soon as @xmath70 , namely the first parametric amplifier is present , the system becomes unstable , independently on the values of @xmath71 , @xmath72 , and @xmath73 . in fact , the eigenvalue equation for @xmath64 is @xmath74 as a consequence , we have three doubly degenerate eigenvalues ( @xmath75 , @xmath76 , and @xmath77 ) . since @xmath78 , at least one of the @xmath79 has a negative real part . we assume that at the beginning the first crystal is switched off ( the pump strength @xmath80 ) . on the other hand , the second pump is on ( @xmath81 ) and the second parametric oscillator is in its equilibrium state below threshold . we therefore have a factorized initial state @xmath82 where @xmath83 and @xmath84 is the equilibrium state of the oscillator below threshold . this can be easily determined upon considering the limits @xmath85 and results in the following expression @xmath86\;. \nonumber\end{aligned}\ ] ] the equilibrium state is thus a gaussian state in which the modes @xmath18 and @xmath19 are correlated . the initial state @xmath84 is then given by the density matrix corresponding to the wigner function @xmath87 where it is straightforward to realize that the modes 2 and 3 are correlated . moreover , it is not a pure state , because @xmath88^{2 } \nonumber \\ & = & \left(1-\frac{\chi_{2}^{2}}{k^{2}}\right ) < 1 \ ; , \label{eq : trrhosq}\end{aligned}\ ] ] as expected . the reduced density matrices of each mode are identical and coincide with the thermal state @xmath89 with an initial mean number of photons given by @xmath90 which means that when the oscillator is initially sufficiently close to threshold the initial mean number of photons in modes 2 and 3 within the cavities can be large . at time @xmath91 the first pump is turned on ( @xmath92 ) : also the first crystal begins to operate and the two groups of three modes start their joint evolution , according to @xmath93 where the two factorized evolutions are identical because both the operator @xmath94 and the initial condition are identical in the two cases . as a consequence , we end up with two identical six - dimensional problems . the solution of eq . ( [ eq : evolrhot ] ) can be found as in sec . [ evol ] by using the wigner functions @xmath95 where the initial wigner function @xmath96 corresponding to the initial density matrix [ eq . ( [ eq : rho123 ] ) ] is given by [ eq : wwww ] @xmath97\right\ } \label{eq : w123 in } \\ & = & \left(\frac{2}{\pi}\right)^{3}\sqrt{\det c } \exp\left\{-2\langle\vec{z}|c|\vec{z}\rangle\right\}\ ; , \label{eq : wcomp}\end{aligned}\ ] ] with @xmath98 and @xmath99 from eq . ( [ eq : wig123 ] ) one can immediately recognize that since the initial state @xmath100 is gaussian and the propagator @xmath101 is also gaussian , the wigner function @xmath102 of the evolved state must remain gaussian at all times . upon integrating over @xmath103 , eq . ( [ eq : wig123 ] ) can be rewritten as @xmath104 where @xmath105^{-1}\ ; , \label{eq : bt}\ ] ] and @xmath106 and @xmath107 are the six - dimensional matrices defined in eqs . ( [ eq : gamma6 ] ) , ( [ eq : gt ] ) , and ( [ eq : sigma ] ) . this gaussian evolution holds for a short time only . as a matter of fact , one should distinguish between the mode along direction 1 and those along directions 2 and 3 : @xmath108 and @xmath109 denote creation of a photon in the _ stationary - wave _ modes within the cavities , whereas @xmath110 denotes the creation of a photon in the _ traveling - wave _ mode along direction @xmath20 . therefore the interaction @xmath111 exists only for the time period during which this traveling wave mode 1 moves _ within _ the nonlinear crystal . in order to prepare the desired state for the modes 2 and 3 , simultaneously taking full advantage of the degree of freedom represented by the traveling - wave mode 1 , we perform a _ conditional _ @xcite measurement on direction 1 , thereby conditioning the state of the four modes along directions 2 and 3 upon the detection of a photon along direction 1 polarized at @xmath112 with respect to the two output polarizations @xmath3 and @xmath2 , which are orthogonal to each other . in this way we also post - select ( along direction 2 ) the input state of the second crystal . the projection operator associated to such a conditional measurement is therefore given by @xmath113 as a consequence of this measurement ( whose success probability amounts to 0.5 ) the state along direction 1 and directions 2 and 3 factorizes : the state along direction 1 is given by @xmath114 which represents a photon polarized at @xmath112 , whilst the conditional state for directions 2 and 3 is represented by the density matrix @xmath115 which can be rewritten as @xmath116\ ; , \label{eq : rhocon}\end{aligned}\ ] ] where [ eq : rhoconiii ] @xmath117 the state of eq . ( [ eq : rhocon ] ) is of the same form of the desired state , eq . ( [ eq : catpro2 ] ) , and is a linear superposition of distinguishable states , as long as @xmath118 is well distinguishable from @xmath119 . it should also be emphasized at this stage that the density matrix ( [ eq : rhocon ] ) directly corresponds to the wigner function , eq . ( 2 ) of ref . @xcite , obtained in the opa case . the similarity between the opo and the opa configurations is better brought about in the limit of small interaction times ( see the appendix , where it is also shown that in this limit many of our results are very similar to those obtained in the opa case @xcite ) . roughly speaking , one should recover the opa results from the opo ones in the limit @xmath120 , since this condition means absence of cavity mirrors . however , this correspondence does not hold exactly because the initial state in the opo case ( the state present in the cavity at @xmath91 , when the first nonlinear crystal is switched on ) is slightly different . this fact explains the differences between the opa and the opo , which result in a far larger effective number of photons in the latter case . how can we probe the quantum state produced in this parametric - oscillator entangled configuration , and prove that it actually represents a schrdinger - cat state ? in order to do this , one has to independently show that i ) the state is indeed made out of two _ macroscopically distinct components _ , that ii ) these two components exhibit _ quantum interference _ , so that the state can be considered as a true _ linear superposition _ rather than a _ statistical mixture _ , and that iii ) the `` separation '' between the two components _ scales with a macroscopic or mesoscopic parameter _ , usually the number of photons . to achieve this goal , we propose three different and independent methods which can be used either alternatively or simultaneously as we shall explain in detail in the next three subsections . let us employ photon number measurements for the modes along direction 2 , thereby collecting the photon - number distributions @xmath121 and @xmath122 . we therefore consider the reduced density matrix obtained by performing the trace on the state of eq . ( [ eq : rhocon ] ) , that is , @xmath123 { \rm tr}_{23}\left[\rho_{2 - 3}^{(0)}\right]\right . \nonumber \\ & & + \left.{\rm tr}_{3o}\left[\rho_{2e-3o}^{(0)}\right ] { \rm tr}_{23}\left[\rho_{2 - 3}^{(1)}\right]\right\ } \left[p\left(\frac{\pi}{4}\right)\right]^{-1}\ ; , \label{eq : rhored2e}\end{aligned}\ ] ] where @xmath124 is the probability of finding one photon with polarization at @xmath112 , that is , @xmath125 \nonumber \\ & = & { \rm tr}_{2 - 3}\left[\rho_{2 - 3}^{(1)}\right ] { \rm tr}_{2 - 3}\left[\rho_{2 - 3}^{(0)}\right]\ ; , \label{eq : probpi4}\end{aligned}\ ] ] represents the probability of the conditional measurement generating the desired cat state . the interference terms in eq . ( [ eq : rhocon ] ) obviously give no contribution to eq . ( [ eq : rhored2e ] ) , since @xmath126= { \rm tr}_{2 - 3}\left[{}_{1}\langle 1 | \rho_{1 - 2 - 3}(t ) | 0\rangle_{1}\right ] = 0\;. \label{eq : trint}\ ] ] combining eqs . ( [ eq : rhored2e ] ) and ( [ eq : probpi4 ] ) , one obtains for the reduced state @xmath127\ ; , \label{eq:2erhored}\ ] ] with an identical form for the reduced state @xmath128 . the reduced density matrices @xmath129 and @xmath130 can be determined in a similar way . from eq . ( [ eq:2erhored ] ) it is immediate to recognize that the reduced state of the mode @xmath131 is given by the sum of two density matrices , conditioned upon the detection of _ one _ photon and of _ zero _ photons in the mode @xmath132 ( or , more precisely , _ one _ photon in the mode @xmath133 ) , respectively . therefore the two terms of the reduced density matrix can be experimentally obtained by rotating the polarizer in front of the detector d@xmath134 located along the direction @xmath20 : when the polarizer is vertical ( mode @xmath133 ) , we have zero photons in the mode @xmath132 , and only the second term of the sum in the right - hand side ( rhs ) of eq . ( [ eq:2erhored ] ) is realized . on the contrary , if the polarizer is set horizontally ( mode @xmath132 ) , one detects one photon in the mode @xmath132 , projecting the resulting density matrix for the mode @xmath131 onto the second term in the sum ( [ eq:2erhored ] ) . however , both terms are present when the polarizer is set at 45@xmath135 . an experimentalist could then take advantage of this property to test the presence of the two component states : the distinction between the two states in the superposition can be made via photon number measurements , yielding the probability distribution @xmath122 . in fact , one has @xmath136 where @xmath137 [ @xmath138 is the probability distribution obtained when the polarized is set horizontally ( vertically ) . the results an experimentalist would obtain with a simple photodetection in these two situations are shown in fig . [ fg : photod](a ) and ( b ) , together with the probability distribution ( [ eq : pn2e ] ) one would obtain when the polarizer is set at @xmath139 [ fig . [ fg : photod](c ) ] . in this way we have verified the existence of two distinct components in the state ( [ eq:2erhored ] ) . but how can we be sure that these two components form a quantum superposition and not just a classical mixture ? to answer this question , one has to perform a measurement able to distinguish the `` cat state '' @xmath140^{-1 } \nonumber \\ & \otimes & \left[\rho^{(1)}_{2e-3o}\rho^{(0)}_{2o-3e } + \rho^{(0)}_{2e-3o}\rho^{(1)}_{2o-3e } \right . \nonumber \\ & + & \left.\rho^{({\rm int})}_{2e-3o}\rho^{({\rm int})\,\dagger}_{2o-3e } + \rho^{({\rm int})\,\dagger}_{2e-3o}\rho^{({\rm int})}_{2o-3e}\right]\ ; , \label{eq : rhoconcomp}\end{aligned}\ ] ] from the corresponding statistical mixture @xmath141^{-1 } \nonumber \\ & \times & \left[\rho^{(1)}_{2e-3o}\rho^{(0)}_{2o-3e } + \rho^{(0)}_{2e-3o}\rho^{(1)}_{2o-3e } \right]\ ; , \label{eq : rhostat}\end{aligned}\ ] ] which does not exhibit any interference . in order to reach this goal , we perform an interference experiment , involving the modes along direction @xmath18 only , using a detection system similar to the one proposed in ref . @xcite , as schematically described in fig . [ fg : scheme ] . the measured quantity is given by the photocounts at the detector d@xmath142 , as a function of the variable phase @xmath143 . the annihilation operator @xmath144 corresponding to the mode traveling to the detector d@xmath142 can be written in terms of the annihilation operators of the modes @xmath131 and @xmath145 as @xmath146 so that the operator number of photons for the mode @xmath144 will be given by @xmath147 in order to be able to distinguish between the superposition state and the mixture , the expectation value @xmath148 \label{eq : cdccond}\ ] ] has to be different from @xmath149\;. \label{eq : cdcmix}\ ] ] it is then clear that this interference experiment can answer our question whenever the contributions of the off - diagonal terms @xmath150 $ ] and its complex conjugate @xmath151 $ ] are nonzero . let us start by evaluating the contribution of the diagonal terms , namely , eq . ( [ eq : cdcmix ] ) . after explicit integration of the corresponding wigner function , it is easy to prove that the phase - dependent terms [ the third and the fourth term in eq . ( [ eq : nc ] ) ] vanish when one computes the expectation value , eq . ( [ eq : cdcmix ] ) . therefore the diagonal terms yield a phase ( @xmath143)-independent contribution given by [ eq : cdc ] @xmath152 \nonumber \\ & & + \frac{1}{4}\left[\langle n_{2e}\rangle^{(1 ) } + \langle n_{2e}\rangle^{(0)}\right ] \label{eq : cdcexp } \\ & = & \frac{1}{2}\left[\langle n_{2}\rangle^{(1 ) } + \langle n_{2}\rangle^{(0)}\right]\ ; , \label{eq : ncdiagexp}\end{aligned}\ ] ] where @xmath153}{{\rm tr}_{2 - 3}\left[\rho_{2 - 3}^{(i)}\right]}\;,\ ] ] ( @xmath154 , 1 ) is the mean photon number in one of the two diagonal states in eqs . ( [ eq : rhoconcomp ] ) and ( [ eq : rhostat ] ) . in the small interaction - time limit , which is very well justified in the present case ( see appendix ) , @xmath155 , @xmath156 , @xmath157 , we have [ eq:23rho ] @xmath158 as a consequence , the two expectation values in the rhs of eq . ( [ eq : ncdiagexp ] ) can be explicitly evaluated and are given by [ eq : n2 ] @xmath159^{-1}\ ; , \label{eq : n20 } \\ \langle n_{2 } \rangle ^{(1 ) } & = & 2\bar{n}+1\ ; , \label{eq : n21}\end{aligned}\ ] ] where @xmath160 is the initial mean photon number in the cavity . in conclusion , the diagonal contribution to the expectation value in eq . ( [ eq : cdccond ] ) amounts to @xmath161 which is indeed @xmath143-independent as expected . we turn now our attention to the off - diagonal terms in eq . ( [ eq : rhoconcomp ] ) , which are absent in eq . ( [ eq : rhostat ] ) . first we note that the expectation values of the number operators relative to the two polarizations in mode 2 computed on the off - diagonal terms vanish , i.e. , @xmath162 on the other hand , the third and the fourth term in the rhs of eq . ( [ eq : nc ] ) give to the expectation value on the off - diagonal terms the contributions [ eq : phio ] @xmath163 { \rm tr}_{2 - 3}\left[\rho_{2 - 3}^{(0)}\right]\right\}^{-1 } \nonumber \\ & & \times \left(\langle a_{2e}\rangle^{\rm ( int ) } \langle a_{2o}^{\dagger}\rangle^{{\rm ( int)}\ , \dagger}\right . \nonumber \\ & & \,\,\,\,\,\,\,\,\ , + \left.\langle a_{2e}\rangle^{{\rm ( int)}\ , \dagger } \langle a_{2o}^{\dagger}\rangle^{\rm ( int)}\right)\ ; , \label{eq : phiod } \\ e^{-i\phi } \langle a_{2e}^{\dagger } a_{2o } \rangle_{\rm o - d } & = & \left(e^{i\phi}\langle a_{2o}^{\dagger } a_{2e}\rangle_{\rm o - d}\right)^{*}\ ; , \label{eq : phiodst}\end{aligned}\ ] ] where @xmath164\;. \label{eq : a2int}\ ] ] these contributions are generally different from zero , and this observation is sufficient to reach the conclusion that the proposed interference experiment is able to distinguish the cat state from the corresponding mixture . we are able to evaluate these off - diagonal terms in the small interaction - time limit developed in the appendix : at the lowest order in @xmath156 , @xmath157 , and @xmath165 , we have @xmath166 and therefore , using eq . ( [ eq : a2int ] ) , [ eq : inta2 ] @xmath167 on the other hand , [ eq : r23 ] @xmath168 which yield , respectively , [ eq : tr23r23 ] @xmath169 and , finally , @xmath170 in conclusion , considering the off - diagonal contribution , eq . ( [ eq : cdccond ] ) can be rewritten as @xmath171 it is then clear that the photocounts at the detector d@xmath142 exhibit interference fringes as a function of the variable phase @xmath143 , if and only if the state ( [ eq:2erhored ] ) is a true linear superposition and not just a statistical mixture of the two macroscopic components . the visibility of such interference fringes is given by @xmath172 and has therefore the lower bound @xmath173 for @xmath174 . our aim in this subsection is to compute the first- and second - order correlation functions relative to our output modes , in order to make an independent test of the presence of quantum coherence in our system . we keep in mind @xcite that a manifestation of quantum coherence at second order is subpoissonian statistics , i.e. , @xmath175^{2}\ ; , \label{eq : subpo}\ ] ] where @xmath176 and @xmath177 are , respectively , the first- and second - order correlation functions . let us consider the same experimental apparatus we have proposed for the detection of interference ( see fig . [ fg : scheme ] ) . we take now into account both output ports @xmath144 and @xmath178 of the polarizing beam splitter , with annihilation operators [ eq : cd ] @xmath179 and evaluate the correlation functions @xmath180 , @xmath181 , and @xmath182 , where @xmath183 is given by eq . ( [ eq : nc ] ) , and @xmath184 we shall evaluate the functions @xmath185 , @xmath186 , and @xmath187 in the small - time approximation limit ( see appendix ) , in which @xmath188 where @xmath189 is the gaussian state described by the wigner function @xmath190 of eq . ( [ eq : w23bt ] ) , for which the wigner function corresponding to the reduced density matrix of mode @xmath6 alone is given by eq . ( [ eq : redtherm ] ) , that represents a thermal state with a mean number of photons given by @xmath160 of eq . ( [ eq : n20 ] ) . upon evaluating all the required expectation values , we obtain [ eq : corr ] @xmath191 from eqs . ( [ eq : corr ] ) it is clear that the visibility of the fringes in @xmath192 and @xmath193 is given by @xmath194 and monotonically decreases from @xmath195 ( for @xmath196 ) to @xmath197 ( for @xmath198 ) . finally , considering the field at the output port @xmath144 , the first- and second - order correlation functions for mode @xmath6 can be written as [ eq : gg ] @xmath199\ ; , \label{eq : gg2}\end{aligned}\ ] ] respectively . it should be noted that these results map into the corresponding ones obtained in ref . @xcite for the opa case upon a redefinition of the phase angles . by comparing @xmath200^{2}$ ] and @xmath177 it is possible to see that @xmath201^{2}$ ] only at low mean photon number , as it could have been easily expected . the best situation is obtained when @xmath202 , in which case [ eq : ggg ] @xmath203^{2 } & = & \left(1 + 2\bar{n}\right)^{2}\ ; , \label{eq : ggg1 } \\ g^{(2)}(0 ) & = & 2\bar{n}\left(3\bar{n}+2\right)\ ; , \label{eq : ggg2}\end{aligned}\ ] ] and the condition for quantum coherence at second order is reached when @xmath204 . on the other hand , when @xmath205 , @xmath200^{2}=\bar{n}^{2}$ ] , @xmath206 , and therefore @xmath177 is always larger than @xmath200^{2}$ ] . the aim of the present section is to provide a means to represent the essential features of the schrdinger - cat state , eq . ( [ eq : rhocon ] ) , which `` lives '' in a 8-dimensional phase space , in the more customary 2-dimensional phase space , in order to make a comparison with the more conventional cat states @xcite . let us start from eq . ( [ eq : rhocon ] ) which we rewrite here for convenience @xmath116\;. \label{eq : cat}\end{aligned}\ ] ] the wigner function representation of the density matrix ( [ eq : cat ] ) would of course reflect its characteristic schrdinger - cat properties . however , in order to better understand the nature of this state , it would be interesting and desirable to see whether it is possible to find different optical modes in whose terms the state ( and therefore the wigner function ) may be rewritten in a simpler form . our key idea is then to look for linear combinations of mode operators ( which can easily be realized with linear elements : polarizers and beam - splitters ) such as to factorize the state ( [ eq : cat ] ) in smaller subspaces . we first perform a transformation which changes the horizontally and vertically polarized modes into the 45@xmath135-polarized ones , namely , [ eq : modetra ] @xmath207 and the corresponding expressions for mode @xmath20 and for the creation operators . in terms of these new operators , @xmath208 and @xmath209 [ eqs ( [ eq : hnl1 ] ) and ( [ eq : hnl2 ] ) ] can be rewritten as [ eq : hnlp ] @xmath210 we have already assumed [ sec . [ evol ] ] that the cavity decay rates @xmath36 do not depend on the polarization . this in turn means that @xmath211 and @xmath212 , and therefore we have that for the @xmath213-polarized modes we have the same evolution equation as that for the original modes ( except for a minus sign ) . consequently , it is possible to repeat all the same arguments as before [ secs . [ evol ] and [ genera ] ] . in particular , the modes @xmath214 , @xmath215 , and @xmath216 are decoupled from their orthogonal counterparts @xmath217 , @xmath218 , and @xmath219 , and the evolution equation may be rewritten as @xmath220 in eq . ( [ eq : evolrt ] ) the initial condition is given in the same way by @xmath221 where @xmath222 is the equilibrium state below threshold of the parametric oscillator when nl1 is turned off , and the same initial condition holds for the @xmath223-polarized modes . as a consequence , the same gaussian evolution we have found in sec . [ evol ] holds . the only difference is that now the conditional measurement is simply a projection onto the state @xmath224 , i.e. , the one - photon state for the @xmath214 mode , while the @xmath223-polarized modes remain decoupled from the orthogonal ones . the cat state after the conditional detection of the @xmath225 photon for the @xmath226 mode is then written in the following way ' '' '' ' '' '' @xmath227 ' '' '' ' '' '' 2 where @xmath228 and @xmath229 are again given by the expressions ( [ eq : rhocon231 ] ) and ( [ eq : rhocon230 ] ) . it should be noted that , using these new @xmath213-polarized modes , one gets a complete factorization of the @xmath223-polarized modes , which are _ not affected _ by the _ quantum injection _ process induced by the conditional measurement . the @xmath223-polarized modes are not `` interesting '' , in the sense that all the `` cat '' properties of the state ( [ eq : cr01 ] ) are contained in @xmath230 , and therefore we shall neglect them from now on . we are then left with the state @xmath230 , which is an _ entangled _ state of the modes @xmath231 and @xmath232 . as the second step of our procedure aimed at the further simplification of the original 8-dimensional wigner function , we consider the transformation @xmath233 which is suggested by the interaction term in eq . ( [ eq : hnl2p ] ) . in terms of @xmath234 and @xmath235 , eq . ( [ eq : hnl2p ] ) becomes @xmath236 and the two modes @xmath234 and @xmath235 are squeezed by the nonlinear crystal . these modes can be experimentally realized outside the cavity for example with two pbs and a 50%50% bs , as schematically described in fig . [ fg : schemwig ] . the state of these two modes can be represented by the wigner function @xmath237 where [ see eqs . ( [ eq : sol123 ] ) and ( [ eq : bt ] ) ] ' '' '' ' '' '' @xmath238 what is the nature of this state ? in order to answer this question , we are naturally guided by two different approaches : i ) the study of the opa case @xcite and ii ) the use of the small - time limit @xmath239 we have already considered in sec . [ photo ] and worked out in the appendix . in the opa case @xcite the output state at time @xmath240 is given by @xmath241 which can be rewritten in terms of the @xmath213-polarized modes as @xmath242 neglecting the factorized state @xmath243 , and using the @xmath244 modes , we have [ eq : squent ] @xmath245 ' '' '' ' '' '' 2 which is an entangled superposition of the _ squeezed _ one - photon and vacuum states of the modes @xmath234 and @xmath235 . it is quite clear now that if we want to `` isolate '' one mode , say , the @xmath234 mode , we need a second _ conditional measurement _ on the mode @xmath235 , e.g. , a projection onto the state @xmath246 such a conditional measurement could be performed , for example , by sending a two - level atom resonant with the atomic transition through the cavity , and eventually post - selecting its internal state in a corresponding superposition of its ground and excited states . the conditional state , provided the measurement has given a successful result , would then read as @xmath247 we can reach a similar conclusion also by analyzing the opo case using the very well justified small - time approximation ( see appendix ) in the limit @xmath248 , applied to the modes @xmath226 , @xmath231 , and @xmath232 . we have , at the lowest order in @xmath156 , [ eq : r12345 ] @xmath249 where the initial density matrix @xmath84 is the state described by the wigner function ( [ eq : w23bt ] ) . if we now write eq . ( [ eq : w23bt ] ) in terms of the new variables corresponding to the modes @xmath234 and @xmath235 , namely , [ eq : varchange ] @xmath250 we obtain ' '' '' ' '' '' @xmath251 \nonumber \\ & & \times \frac{2}{\pi } \sqrt{1-\frac{\chi_{2}^{2}}{\kappa^{2 } } } \exp\left[-2\left(1+\frac{\chi_{2}}{\kappa}\right ) x_{d_{-}}^{2}\right . % \nonumber \\ % & & \;\;\;\;\;\;\;\;\;\ ; \left . -2\left(1+\frac{\chi_{2}}{\kappa}\right ) y_{d_{-}}^{2}\right]\;. \nonumber \\ \label{eq : wdb}\end{aligned}\ ] ] the initial states for the modes @xmath244 are generalized gaussian states @xcite , of the kind @xmath252 with [ eq : nmgau ] @xmath253 and @xmath254 since the initial state factorizes , we have [ eq : rfac ] @xmath255 which is a mixed state analogous to the pure state ( [ eq : squent ] ) obtained in the opa case . its wigner function can be calculated from eq . ( [ eq : rfac ] ) and is given by @xmath256 e^{-2\left(1-\frac{\chi_{2}}{\kappa}\right ) \left(x_{d_{+}}^{2}+y_{d_{-}}^{2}\right ) -2\left(1+\frac{\chi_{2}}{\kappa}\right ) \left(x_{d_{-}}^{2}+y_{d_{+}}^{2}\right)}\;. \label{eq : wigfour}\end{aligned}\ ] ] ' '' '' ' '' '' 2 two important features should be noted within the form of this wigner function : i ) the interference term ( the last term in the square brackets ) decreases when the number of photons in the initial state increases . this behavior is governed by the factor @xmath257 and by the fact that [ see eqs . ( [ eq : meannumb ] ) and ( [ eq : n2 ] ) ] @xmath198 when @xmath258 . ii ) the wigner function is negative around the origin and its negativity scales to zero as the initial mean photon number @xmath198 . in fact , @xmath259 we have already seen the same scaling behavior of quantum properties with @xmath198 in the calculation of the second - order correlation function @xmath260 : this is one of the desired properties of a schrdinger - cat state , as we have emphasized at the beginning of this section . again , eq . ( [ eq : wigfour ] ) bears a remarkable similarity with the corresponding result obtained in ref . @xcite for the opa configuration , in the limits @xmath120 and of small interaction times . the main advantage of the opo is given by the larger effective number of photons per mode @xmath261 [ see eq . ( [ eq : meannumb ] ) ] with respect to the @xmath262 of the opa @xcite . we have therefore learnt that in order to obtain a one - mode state which embodies all the relevant features of the original four - mode cat state one _ has to perform _ a conditional measurement on the mode @xmath235 . when this is successfully done , the final conditioned state of the mode @xmath234 alone is described by the wigner function @xmath263 where @xmath264 is the wigner function [ see eqs . ( [ eq : w1d+d- ] ) and ( [ eq : w12233 ] ) ] of the state ( [ eq : rfac ] ) , and @xmath265 \nonumber \\ \label{eq : wab}\end{aligned}\ ] ] is the wigner function of the state onto which the conditional measurement projects the mode @xmath235 [ eq . ( [ eq : prosup ] ) ] . according to the small - time limit approximation ( see appendix ) the explicit form of the wigner function ( [ eq : wigd+ ] ) can be derived from eqs . ( [ eq : rfac])([eq : wab ] ) and , after a lengthy calculation , reads as ' '' '' ' '' '' @xmath266 % \label{eq : wd } \\ % \nonumber \\ % & & \times \bigg\{\left(|\alpha|^{2 } + |\beta|^{2}\frac{\chi_{2}^{2}/\kappa^{2}}{4-\chi_{2}^{2}/\kappa^{2}}\right ) % \nonumber \\ % & & \times \left[x_{d_{+}}^{2}\left(2-\frac{\chi_{2}}{\kappa}\right)^{2}\right . \nonumber \\ & & \left.+ y_{d_{+}}^{2}\left(2+\frac{\chi_{2}}{\kappa}\right)^{2 } -1\right ] % \nonumber \\ % & & \;\;\;\ ; + |\beta|^{2 } + 2{\rm re } \left(\alpha^{\star}\beta \left[x_{d_{+}}\left(2-\frac{\chi_{2}}{\kappa}\right)\right.\right . % \nonumber \\ % & & \;\;\;\ ; -iy_{d_{+ } } \left.\left.\left(2+\frac{\chi_{2}}{\kappa}\right)\right]\right)\bigg\}\ ; , \label{eq : wd } % \nonumber\end{aligned}\ ] ] ' '' '' ' '' '' 2 which is in very good agreement with the numerically computed exact one . as desired , the value of @xmath267 at the origin may also be negative ( depending on the parameters @xmath268 and @xmath269 specifying the conditional measurement ) , reflecting the quantum properties of the original 4-dimensional wigner function ( [ eq : wigfour ] ) . explicitly , one has @xmath270 these results are graphically shown in figs . [ fg : wig1][fg : margy ] . in figs . [ fg : wig1 ] and [ fg : wig2 ] we have plotted the wigner function ( [ eq : wd ] ) for two values of @xmath261 . in both figures two different viewpoints have been selected for the tridimensional plots , in order to display most clearly the quantum superposition character of our cat - like state . in particular , one should note that in both cases the wigner function is negative around the origin . however , a comparison between fig . [ fg : wig1 ] and fig . [ fg : wig2 ] shows that even though the two gaussian peaks are better separated for a larger number of photons , the negativity of the wigner function tends to disappear as soon as the initial number of photons increases , as expected . this behavior is further confirmed by the inspection of the corresponding marginal distributions of the wigner function ( [ eq : wd ] ) , shown in figs . [ fg : margx ] and [ fg : margy ] : @xmath271 displays a larger separation between the peaks as the initial mean photon number @xmath272 increases . on the other hand , @xmath273 displays the interference between the two macroscopic components , which tends to be washed out when the number of photons increases . in fact , for @xmath274 , the interference fringes have already disappeared . in this paper we have considered the generation of _ entangled _ schrdinger - cat states in an optical parametric oscillator , as a relevant variant of the original proposal @xcite which instead had considered the amplifier case . in these works , the central point ( both conceptually and experimentally ) is the _ quantum injection _ @xcite of the second nonlinear crystal with the output of the first parametric medium . in the present paper , we have computed the time evolution for the electromagnetic - field and chosen the initial condition needed for the generation of the desired cat state . such a state , however , lives in a eight - dimensional phase - space : therefore we have proposed three methods which are able to prove that it is an actual schrdinger - cat state : direct photodetection , measurement of the correlation functions , and measurement of the wigner function . our calculations show that the state produced in this way has indeed two macroscopic ( mesoscopic ) components which are macroscopically ( mesoscopically ) distinguishable and that they are in a coherent superposition ( and not just in a statistical mixture ) , i.e. they display quantum interference . a comparison with the performance of the corresponding opa scheme @xcite is in order here . first , the opo has a larger conversion efficiency due to the enhancement factor of the parametric interaction , given by the presence of the cavities . this leads to a larger number of photon couples with the same pump power . second , our schrdinger - cat state is confined in the cavities , contrarily to what happens in the opa case , where it is a traveling wave . however , the price one has to pay in order to have these advantages , is given by the unavoidable cavity losses , that tend to destroy the coherence of the state when the number @xmath261 of initial photons tends to infinity . such a phenomenon decoherence @xcite is visualized by the progressive disappearance of the interference fringes and of the negativity of the wigner function when @xmath261 increases . it is then clear that one has to consider a trade - off condition between the enhancement factor ( a large @xmath261 ) and the losses ( a low @xmath275 ) . this may lead to a comparison between the performances of the opo and the opa @xcite : in particular , our opo configuration is preferable when the mean number of initial photons @xmath261 [ see eq . ( [ eq : meannumb ] ) ] is larger than the corresponding parameter ( @xmath262 @xcite ) of the opa . in conclusion , we think that an experiment along the lines outlined in this paper and in @xcite which is realizable using presently available technology is a promising candidate for producing entangled superpositions of macroscopically distinct quantum states . it is a pleasure for us to acknowledge interesting and stimulating discussions with a. ekert and p. grangier . this work has been partially supported by infm ( through the 1997 advanced research project `` cat '' ) , by the european union in the framework of the tmr network `` microlasers and cavity qed '' , and by murst through `` cofinanziamento 1997 '' . the fact that the time @xmath240 during which we have the interaction within the first nonlinear crystal is very short is of fundamental importance , and it allows an immediate description of the experiment . to bring out this most clearly , we develop an approximate treatment , which is however justified by the actual experimental values reported in ref . @xcite . the interaction time @xmath240 , which is the time of flight of the photon within the first nonlinear crystal nl1 , is given by @xmath276 where @xmath277 is the crystal length , @xmath278 its refraction index , and @xmath144 is the speed of light in vacuum . on the other hand , for an average pump power @xmath279 , the coupling strength is of the order of @xmath280 . in order to obtain `` macroscopic '' states , one needs a quite large initial mean number of photons in the parametric oscillator below threshold . this fixes the damping rates @xmath281 to be slightly larger than @xmath71 , since , from eq . ( [ eq : meannumb ] ) , we have @xmath282 therefore , we have @xmath283 , too . since the wavelength of the photon is @xmath284 cm , this amounts to having a _ standard cavity _ , with a quality factor @xmath285 on the other hand , @xmath286 will be of the order of @xmath71 . in summary , we have @xmath287 from eqs . ( [ eq : meq][eq : hnl123 ] ) , ( [ eq : infact ] ) , and ( [ eq : rho123 ] ) , one has , for the time evolution of the combined density matrix , @xmath288 since @xmath289 , @xmath286 , @xmath71 , it is appropriate to expand the exponential @xmath290 in power series up to second order in @xmath291 , @xmath156 , @xmath157 , yielding @xmath292 and @xmath293\ ; , \label{eq : l123}\ ] ] where @xmath294 is that part of the liouvillian which only acts on the modes @xmath18 and @xmath19 , as given by eqs . ( [ eq : parmeq ] ) and ( [ eq : hnl123 ] ) . it is possible in this way to determine the conditional states @xmath295 , @xmath296 , and the interference terms in eqs . ( [ eq : rhocon][eq : rhocon23i ] ) . we compute first @xmath297 the properties of this state are usually characterized by measuring the photon - number distribution of the mode 2 along direction 2 . we have therefore to perform the trace over the mode 3 in eq . ( [ eq:1rho1 ] ) , obtaining @xmath298 we already know that @xmath299 is a thermal state with a mean number of photons given by @xmath160 [ see eqs . ( [ eq : meannumb ] ) , ( [ eq : kchi ] ) , and ( [ eq : n20 ] ) ] , i.e. , @xmath300 [ see eq . ( [ eq : redtherm ] ) ] and consequently @xmath301 which is a sort of _ shifted _ thermal state and is identical to the state obtained in the case of the parametric amplifier @xcite with a mean number of photons given by eq . ( [ eq : meannumb ] ) . on the other hand , we have , at the lowest order in @xmath156 , electronic address : mauro@camcat.unicam.it e. schrdinger , naturwiss . * 23 * , 807 ( 1935 ) . b. yurke and d. stoler , phys . lett . * 57 * , 13 ( 1986 ) . w. schleich , m. pernigo , and fam le kien , phys . a * 44 * , 2172 ( 1991 ) . w. h. zurek , phys . d * 24 * , 1516 ( 1981 ) ; _ ibid . _ * 26 * , 1862 ( 1982 ) ; phys . today * 44*(10 ) , 36 ( 1991 ) , and references therein . m. brune , e. hagley , j. dreyer , x. maitre , a. maali , c. wunderlich , j. m. raimond and s. haroche , phys . rev . lett . * 77 * , 4887 ( 1996 ) . s. habib , k. shizume , and w. h. zurek , phys . lett . * 80 * , 4361 ( 1998 ) . c. monroe , d. m. meekhof , b. e. king , and d. j. wineland , science * 272 * , 1131 ( 1996 ) . e. schrdinger , proc . cambridge philos . * 31 * , 555 ( 1935 ) . a. einstein , b. podolsky , and n. rosen , phys . rev . * 47 * , 777 ( 1935 ) . c. bennett , g. brassard , c. crpeau , r. josza , a. peres , and w. k. wootters , phys . lett . * 70 * , 1895 ( 1993 ) . d. bouwmeester , j .- v . pan , k. mattle , m. eibl , h. weinfurter , and a. zeilinger , nature ( london ) * 390 * , 575 ( 1997 ) ; d. boschi , s. branca , f. de martini , l. hardy , and s. popescu , phys . * 80 * , 1121 ( 1998 ) . a. k. ekert , phys . lett . * 67 * , 661 ( 1991 ) ; a. k. ekert _ et al . _ , _ ibid . _ * 69 * , 2881 ( 1992 ) . divincenzo , science * 270 * , 255 ( 1995 ) . f. de martini , phys . lett . * 81 * , 2842 ( 1998 ) ; phys . a * 250 * , 15 ( 1998 ) . p. g. kwiat , k. mattle , h. weinfurter , a. zeilinger , a. v. sergienko , y. h. shih , phys . lett . * 75 * , 4337 ( 1995 ) . for the generation of an ultra - bright source of polarization - entangled photons also see : p. g. kwiat , e. waks , a. g. white , i. appelbaum , and p. h. eberhard , lanl e - print quant - ph/9810003 . j. s. bell , physics * 1 * , 195 ( 1964 ) . d. m. greenberger , m. a. horne , and a. zeilinger , am . j. phys . * 58 * , 1131 ( 1990 ) ; n. d. mermin , phys . * 65 * , 1838 ( 1990 ) . d. bouwmeester _ et al . _ , to appear ( 1998 ) . m. h. rubin , d. n. klyshko , y. h. shih , a. v. sergienko , phys . a * 50 * , 5122 ( 1994 ) . l. davidovich , m. brune , j. m. raimond and s. haroche , phys . a * 53 * , 1295 ( 1996 ) . x. y. zou , l. j. wang , l. mandel , phys . lett . * 67 * , 318 ( 1991 ) ; l. j. wang , x. y. zou , l. mandel , phys . a * 44 * , 4614 ( 1991 ) ; t. p. grayson , x. y. zou , d. branning , j. r. torgerson , l. mandel , phys . a * 48 * , 4793 ( 1993 ) . a. o. caldeira and a. j. leggett , phys . a * 31 * , 1059 ( 1985 ) ; d. f. walls and g. j. milburn , phys . a * 31 * , 2403 ( 1985 ) . scully and m.s . zubairy , _ quantum optics _ ( cambridge university press , cambridge , 1997 ) . g. j. milburn and d. f. walls , _ quantum optics _ ( springer - verlag , berlin , 1994 ) . b. sherman and g. kurizki , phys . rev . a * 45 * , r7674 ( 1992 ) ; b. sherman , h. moya - cessa , p. l. knight , and g. kurizki , _ ibid . _ * 49 * , 535 ( 1994 ) . k. vogel , v. m. akulin , and w. p. schleich , phys . * 71 * , 1816 ( 1993 ) . m. hillery , r. f. oconnell , m. o. scully , and e. p. wigner , phys . rep . * 106 * , 121 ( 1986 ) . c. w. gardiner , _ quantum noise _ ( springer - verlag , berlin,1991 ) . h. risken , _ the fokker planck equation _ ( springer - verlag , berlin , 1989 ) . g. j. milburn and d. f. walls , _ quantum optics _ ( springer - verlag , berlin , 1991 ) .

scintillation properties of pure csi crystals used in the shower calorimeter being built for precise determination of the decay rate are reported . seventy - four individual crystals , polished and wrapped in teflon foil , were examined in a multiwire drift chamber system specially designed for transmission cosmic muon tomography . critical elements of the apparatus and reconstruction algorithms enabling measurement of spatial detector optical nonuniformities are described . results are compared with a monte carlo simulation of the light response of an ideal detector . the deduced optical nonuniformity contributions to the fwhm energy resolution of the pibeta csi calorimeter for the 69.8mev positrons and the monoenergetic 70.8mev photons were 2.7% and 3.7% , respectively . the upper limit of optical nonuniformity correction to the 69.8mev positron low - energy tail between 5mev and 55mev was.2% , as opposed to the.3% tail contribution for the photon of the equivalent total energy . imposing the 5 mev calorimeter veto cut to suppress the electromagnetic losses , geant - evaluated positron and photon lineshape tail fractions summed over all above - threshold adcs were found to be 2.36.05(stat).20(sys)% and 4.68.07(stat).20(sys)% , respectively . pacs numbers : 87.59.f ; 29.40.mc ; 24.10.lx _ keywords : _ computed tomography ; scintillation detectors ; monte carlo simulations psfig.sty , , , , , , , , , , , ,

hep-ex9905041

the pibeta collaboration has proposed an experimental program @xcite with the aim of making a precise determination of the @xmath0@xmath1@xmath2 decay rate at the paul scherrer institute ( psi ) . the proposed technique is designed to achieve an overall level of accuracy in the range of @xmath70.5@xmath4% , improving thus on the present branching ratio uncertainty of 4@xmath4% @xcite . the ( 1.025@xmath60.034)@xmath8@xmath9 pion beta decay ( @xmath10 ) branching ratio will be remeasured relative to the @xmath11 times more probable @xmath0@xmath1@xmath12 decay rate , that is known with the combined statistical and systematic uncertainty of @xmath70.40@xmath4% @xcite . the standard model description of the @xmath10 decay and its radiative corrections @xcite enables a stringent test of the conserved vector current ( cvc ) hypothesis and the unitarity of the cabibbo - kobayashi - maskawa ( ckm ) quark mixing matrix . the most accurate extraction of the ckm matrix element @xmath13 , based on the superallowed fermi transitions in nuclei , involves the theoretical nuclear overlap corrections that dominate the 0.3@xmath4% total uncertainty @xcite . recent measurements of nine different nuclear transition rates confirm the cvc hypothesis at the level of 4@xmath8@xmath14 , but violate the three - generation ckm unitarity condition by more than twice the estimated error @xcite . complementary neutron decay experiments yield a result that also differs from unitarity by over two standard deviations , but in the opposite sense @xcite . it is of fundamental interest to measure the branching ratio of the @xmath0@xmath1@xmath2 decay at the 0.5@xmath4% level or better to check the consistency with the nuclear and neutron @xmath15 decay values and the unitary prediction of the minimal standard model ( msm ) . moreover , the precisely measured @xmath10 branching ratio could be used to constrain masses and couplings of additional neutral gauge bosons in grand unified theories . if such couplings are directly discovered at future collider experiments , precision low - energy data will be an essential ingredient in extending the msm framework @xcite . the central part of the pibeta detector system is a high - resolution , highly segmented fast shower calorimeter surrounding the active stopping target in a near - spherical geometry . the requirements imposed on the calorimeter design were : * high energy resolution for effective suppression of background processes ; * high segmentation and fast timing response to handle the high event rates ; * acceptable radiation , mechanical and chemical resistance ; * compact geometry to simplify the operation and reduce the cost . these design constraints are met in a spherical detector with individual calorimeter modules made from undoped cesium iodide ( csi ) crystals . the nonactivated alkali iodides have been known for more than thirty years to exhibit a near - ultraviolet emission component when excited by ionizing radiation @xcite . the pure csi material was reintroduced as a fast , rugged , and relatively inexpensive scintillator material by kobayashi et al . @xcite and kubota et at . the scintillation characteristics of pure csi crystals have been reported subsequently by a number of experimental groups . these investigations have covered light readout techniques , scintillation decay times , origins of different emission components , crystal light yields , and energy and timing resolutions @xcite , radiation resistances @xcite , wrapping and tuning methods and the uniformity of light responses @xcite . the design and performance of pure csi electromagnetic calorimeters have been recently described in refs . @xcite ( nms at lampf / bnl ) , @xcite ( ktev at fermilab ) , and @xcite ( phenix at rhic ) . the pibeta calorimeter geometry shown in fig . [ fig : ball ] is obtained by the class ii geodesic triangulation of an icosahedron @xcite . selected geodesic breakdown results in 220 truncated hexagonal , pentagonal and trapezial pyramids covering the total solid angle of 0.77@xmath8@xmath16 sr . additional 20 crystals cover two detector hemispheres open to the beam and act as shower leakage vetoes . the inner radius of the crystal ball is 26@xmath4 cm , and the axial module length is 22@xmath4 cm , corresponding to 12 csi radiation lengths ( @xmath17=1.85@xmath4 cm @xcite ) . there are nine different module shapes : four irregular hexagonal truncated pyramids ( we label them hex a , hex b , hex c , and hex d ) , one regular pentagonal ( pent ) and two irregular half - hexagonal truncated pyramids ( hex d1 and hex d2 ) , and two trapezohedrons which function as vetoes ( vet1 , vet2 ) . the volumes of our csi crystal detector modules vary from 797@xmath4@xmath18 ( hex d1/2 ) to 1718@xmath4@xmath18 ( hex c ) . dimensioned drawings of the hex a and hex d1 crystal shapes are shown as examples in two panels of fig . [ fig : shape ] . all key components of the complete pibeta detector have been prototyped and built and have met the required specifications @xcite . all the detector components were delivered to psi by mid-1998 . the final assembly of the apparatus was followed by the in - beam commissioning and calibration . `` production '' measurements are scheduled for 1999 , and will extend over two more years to complete the first phase of the project , a @xmath70.5@xmath4% measurement of the @xmath0@xmath1@xmath19 decay rate . the quality of the delivered csi calorimeter crystals was controlled in the following program of measurements : 1 . check precisely the crystal physical dimensions against the specifications ; 2 . study the effects of different crystal surface treatments and wrapping configurations on the energy and timing resolution and on the uniformity of light response ; 3 . determine the percentage of the fast scintillation component in the light output ; 4 . measure the contribution of detector photoelectron statistics to energy resolution ; 5 . find the temperature dependence of a detector s adc readings ; 6 . determine axial and transverse optical nonuniformities for each crystal . the ultimate goal of the program was to provide the input parameters for a monte carlo simulation of the calorimeter response and to deduce the effects of csi crystal light yields and optical nonuniformities on the energy lineshapes of detected photons and positrons ranging in energy between 10@xmath4mev and 120@xmath4mev . the spectra generated by cosmic muons have been used to examine the uniformity of light response of two large ( 24@xmath836@xmath4 cm ) cylindrical nai(tl ) calorimeter crystals in the earlier work of dowell et al . their apparatus relied on plastic scintillator hodoscopes with 3.8@xmath83.8@xmath4@xmath20 cross section for the charged particle tracking . the precision of the trajectory reconstruction was therefore limited to @xmath74@xmath4 cm . such positional resolution was deemed inadequate for our application because the pibeta csi crystals have an axial length of 22@xmath4 cm with front ( back ) surface side lengths of less than 4@xmath4 cm ( 7@xmath4 cm ) . most of our measurements were done with the tomography apparatus built around three pairs of multiwire drift chambers ( mwdc ) using cosmic muons as the probe . one dozen pibeta csi detectors were examined in a 350@xmath4mev / c minimum ionizing muon beam of the psi @xmath21e1 area , as well as in the 405@xmath4mev / c penetrating pion and stopping proton beams in the @xmath21m1 channel . selected csi crystals were also scanned with a 662 kev @xmath22cs @xmath23-source . the light yield nonuniformities measured in the beam and with a radioactive - source scan method were compared with more precise cosmic muon tomography data . the experimental apparatus used for a @xmath23-source crystal scans and the subsequent data reduction were simple enough to be easily adopted by the crystal manufacturers , thus accelerating the subsequent quality control cycle . pure csi crystals produced for the pibeta calorimeter came from two different sources . twenty - five crystals were grown and cut to specified shapes in the bicron corporation facility in newbury , ohio . the rest of the scintillators were grown in the institute for single crystals in kharkov ( amcrys ) , ukraine . preliminary examination of fifty - five amcrys csi crystals , including the mechanical and physical tests , was done in the high energy institute of tbilisi university , georgia . manufacturing tolerances of the crystals were specified for the linear dimensions ( @xmath24150 @xmath25m/@xmath2650 @xmath25 m ) and for the angular deviations ( @xmath240.040@xmath27/@xmath260.013@xmath27 ) . geometrical dimensions of the machined crystals were measured upon delivery at psi using the computer - controlled distance - measuring device _ wenzel precision_. the machine was programmed to automatically probe the surfaces of a subject crystal with a predefined shape . each crystal surface plane was scanned with a touch head at six points and the equations of planes were found through these surveyed points . body vertices , measured with an absolute precision of 2 @xmath25 m and reproducible within 20 @xmath25 m , were then compared with the expected theoretical values . those crystals that failed the imposed geometrical tolerances were returned to the manufacturer for reuse as raw crystal growing material . after physical measurements crystal surfaces were polished with a mixture of 0.2 @xmath25 m aluminum oxide powder and etylenglycol . next came a measurement of the light output of the fast csi scintillation component ( f ) that is completely decaying in the first 100 ns , relative to the total csi signal ( t ) , integrated in a 1 @xmath25s adc gate . these measurements were made with unwrapped crystals using air - gap coupled photomultiplier tubes ( pmts ) and a tektronix tds 744 digital oscilloscope . only the crystals with a fast - to - total component ratio ( f / t ) better than 0.7 were accepted . the mean value of fast - to - total ratio for all accepted csi crystals was 0.788 , as shown in fig . [ fig : ft ] . emi photomultiplier tubes 9821qkb @xcite with 75@xmath4 mm diameter cathode were glued to the back faces of hexagonal and pentagonal csi crystals using a 300 @xmath25 m layer of silicone sylgard 184 elastomer ( dow corning rtv silicon rubber plus catalyst ) . smaller half - hexagonal and trapezial detector modules were equipped with two inch emi 9211qka phototubes @xcite . the resulting crystal photomultiplier tube couplings were strong and permanent , but could be broken by application of a substantial tangential force . the pmt quartz window transparency , peaking at @xmath7380 nm @xcite , is approximately matched to the spectral excitation of a pure csi fast scintillation light component with a maximum room temperature emission at @xmath7310 nm @xcite . the pmt high voltage dividers were modified emi - recommended bases designed and built at the university of virginia . these dividers minimized the so - called `` super''-linearity exhibited by many pmts well below the onset of saturation . the maximum pmt nonlinearity measured with a pair of light - emitting diodes was less than 2@xmath4% over the full dynamic range expected in the @xmath10 decay rate measurement @xcite . four different wrapping materials for lateral crystal surfaces were investigated in wrapping and tuning studies with 405@xmath4mev / c pion and proton beams in psi @xmath25m1 area : 1 . one to five layers of a 38@xmath4@xmath25 m ptfe teflon sheet ( cf@xmath28 monomer , @xmath17=16.0@xmath4 cm ) ; 2 . one to five layers of 30@xmath4@xmath25 m mylar sheet ( c@xmath29h@xmath30o@xmath28 , @xmath17=28.7@xmath4 cm ) ; 3 . a 250 @xmath25 m thick tyvek fleece ( polyethylene ch@xmath28 , @xmath17=131.3@xmath4 cm ) ; 4 . a 110@xmath4@xmath25 m thick millipore filter ( polyvinylidene fluoride pvdp , @xmath17=19.0@xmath4 cm ) with 0.22@xmath4@xmath25 m pores ; 5 . a wavelength shifter lacquer treatment of crystal surfaces . in all cases a primary diffuse reflector was protected by an additional 20 @xmath25 m thick aluminized mylar cover . these preliminary tests showed more uniform axial light collection from the crystal front section when a polished front crystal surface was covered with a black paper sheet . a similar tuning method , using black paper strips at the front section of 5@xmath86@xmath836@xmath4@xmath18 csi(tl ) crystals in order to improve the uniformity of light collection , was recently described in ref . @xcite . the maximum fast - to - total ratio was always achieved with unwrapped crystals because all applied wrapping materials absorb more ultraviolet light which dominates the fast component , than visible light . a two - layered teflon cover was found to be superior in delivering @xmath3120@xmath4% more fast scintillation light component than the other reflectors . it was also comparatively better in not degrading the f / t ratio by more than @xmath72@xmath4% . these differences in measured light output were reproducible in the beam tests with @xmath71@xmath4% event statistics and estimated 2@xmath4% systematic uncertainty . they are subsequently confirmed in more controlled cosmic muon tomography measurements . on the basis of these findings the adopted standard wrapping configuration for all studied csi crystals consisted of : ( a ) the lateral surfaces being wrapped with two layers of teflon foil plus one layer of aluminized mylar sheet , and ( b ) the front crystal surface covered with the black paper template . the back crystal surface with the glued phototube was left uncovered . light yield measurements were repeated for a subset of crystals after a six month period to confirm that no appreciable changes occurred as a result of degradation of the surface reflectivities and wrapper material qualities . the measured light output was typically within 5@xmath4% of the originally measured value . the most successful surface treatment of csi crystals , however , involved painting the lateral crystal surfaces with a waveshift lacquer . we recently studied this method on a large sample of csi detectors . the light response and uniformity properties were noticeably improved , resulting in @xmath720@xmath4% better energy resolutions for the 70@xmath4mev monoenergetic @xmath32 and 5082@xmath4mev tagged photons . the possible degradation and change in the crystal surface reflectivities that could be difficult to account for in the multi - year long precision experiment like the pibeta is also expected to be arrested by such a treatment . these measurements will be covered in detail in a forthcoming paper @xcite . a simplified sketch of the tomographic apparatus layout is illustrated in a geant rendering showing a few simulated cosmic muon trajectories on fig . [ fig : drift ] . three identical delay - line readout drift chambers were used to define the cosmic muon tracks intersecting csi crystals . the chambers were built at the los alamos meson physics facility ( lampf ) by the m. sadler group from the abilene christian university . this type of drift chambers has been used reliably for years with several lampf spectrometers @xcite . each chamber consists of two signal and three ground planes with the nominal active area of 60@xmath860@xmath4@xmath20 . the horizontally oriented `` @xmath33 '' and `` @xmath34 '' signal planes are two orthogonal sets of alternating cathode and anode wires evenly spaced at 0.4064@xmath4 cm . the light - tight aluminum box had two drift chamber pairs mounted on its top plate , and one pair fixed below the bottom side . distances between the centers of chamber pairs 13 and 23 were 24@xmath4 cm and 27@xmath4 cm , respectively . the dark box could accommodate up to six csi calorimeter modules at one time . it had feedthrough connectors for twelve signal and twelve high voltage cables as well as six temperature sensor lines and six led calibration signal cables . a pair of 1@xmath4 cm thick plastic scintillators were placed directly below the apparatus and separated from the chamber system by a 5@xmath4 cm layer of lead bricks shielding out the soft cosmic ray component . the monte carlo simulation ( see sec . [ mcd])showed that hard cosmic muons penetrating the frames of the apparatus , csi crystals and shielding material and triggering two scintillators had a smooth energy threshold starting at @xmath7120@xmath4mev . the zenithal angle range of accepted cosmic muon trajectories intersecting all three mwdcs and at least one csi detector was @xmath645@xmath27 . the chambers were operated with anode wires held at positive voltage between 24002600 v. the chamber cathode wires were grounded . the gas mixture was 65@xmath4% argon , 35@xmath4% isobutane plus 0.1@xmath4% isopropyl alcohol . detection efficiencies of individual chambers for penetrating cosmic muons exceeded 90@xmath4% , with the combined six - chamber efficiency routinely surpassing 50@xmath4% . anode wires were attached directly to a fast 2.5 ns / cm delay - line . signals from both ends of the delay line were amplified twenty - fold , discriminated at a threshold of 10 mv and connected to two channels of a time - to - digital converter ( lcr 2229a tdc ) . the time difference in the two tdc readings identified the fired anode wire . cathode lines defining the alternate field @xcite were connected to one `` odd '' ( o ) and one `` even '' ( e ) line , and the signals on these lines were processed by a custom - made electronics unit which added and subtracted the analog pulses @xcite . the electronic sum of cathode pulses ( o@xmath24e ) was discriminated and the resulting delayed signal determined the drift position timing , and was used to stop another tdc channel . the difference of the cathode signals ( o@xmath26e ) was digitized with an analog - to - digital converter ( lcr 2249a adc ) . that information was used to discriminate between the events that produced the ionization tracks left and right of the given anode wire . the 100 ns integration gate timing was defined by the ( o+e ) logic pulse timing . four measurements were therefore required to find an intersection between a cosmic muon track and one drift chamber plane , namely three tdc values and one adc value . the schematic diagram of the electronic logic is shown in fig . [ fig : tmelec1 ] . advantages of the system were good charged particle track resolution ( @xmath70.5@xmath4 mm root - mean - square in the horizontal plane , fig . [ fig : deviation ] ) , stability to fluctuations in the outside environmental parameters ( humidity , temperature and pressure ) and low cost of the associated electronics logic and readout . limited counting rate and restriction to single hit events in the chambers did not represent a drawback in this tomographic application . the apparatus was operated in an air - conditioned room with a controlled humidity level kept below 30@xmath4% . the temperature at six points inside the dark box as well as absolute time were recorded for every triggered event . the temperature range recorded inside the dark box during three years of data taking was ( 22@xmath27@xmath63@xmath35)c . the typical temperature gradient was 0.4@xmath35c / day , leading to average absolute temperature variation of 0.2@xmath35c and average csi adc gain drift of @xmath70.4@xmath4% in a single data taking run ( fixed at 250000 triggers , @xmath76 hours ) . measured light output variations of the csi crystals and the pmt gain drifts caused by daily temperature cycles were compensated for in the replay analysis ( see sec . [ raw ] ) . the computer code used for data acquisition was hix ( heterogeneous information exchange ) , a data acquisition system developed by s. ritt at psi , originally intended for use in small and medium - size nuclear physics experiments @xcite . the `` frontend '' 486 personal computer was connected to a camac crate via a hytec 1331 interface that read adc , tdc , scalar , and temperature sensor units . the c program running under ms - dos accessed data and sent it over the network to a vax 3100 server . a simple communication program on a vaxstation received data from the frontend computer and stored it in a global section buffer . buffered data were passed by the logger application to a user - written analyzer program.analyzed events were subject to predefined cuts , filled in predefined histograms , and stored in a raw data stream directed simultaneously to a hard disk drive and an 8-mm tape system . the experiment was controlled from a pc computer by a microsoft windows control program . a windows - based graphical user interface to the program allowed starting , pausing and stopping data acquisition , as well as online inspection of individual raw data words , calculated data words , scalers , assorted efficiencies , one and two - dimensional histograms , and gate , box and boolean tests . different pc computers on the ethernet could make a connection to the vax server called link and access the same information remotely . during data acquisition only those events for which at least two @xmath33 and two @xmath34 drift chambers had nonzero adc and tdc values were written to an disk - resident ascii data summary file ( dsf ) . one good event contained 55 raw data words specifying the response of the mwdcs and csi detectors , instantaneous temperatures and absolute times . that was a full set of observables for which a cosmic muon trajectory could be unambiguously reconstructed . the raw trigger rate was @xmath713 hz with six csi crystals in the dark box . the average dsf event rate was @xmath74 hz . individual runs were stopped and restarted automatically after 250,000 collected triggers . a total of about 10@xmath36 dsf events per crystal were typically collected in one week of data acquisition . tomography data for one set of six crystals were usually collected over a two week period . two different trigger configurations were used interchangeably during the data collection . drift chamber calibration was done with a trigger requiring a two - scintillator coincidence and good adc and tdc data for at least one drift chamber . in the tomography data acquisition mode , the triple coincidence between two tag scintillators and one csi detector was required , accompanied by at least one good pair of @xmath33 and @xmath34 chamber hits . the rare accidental coincidences involving signals in more than one csi detector were eliminated from the dsf records in the offline analysis . the hit anode wire number @xmath37 was calculated from @xmath38 where @xmath39 and @xmath40 were the tdc values for two ends of the anode delay line , @xmath41 was the number of wires in a chamber ( between 71 and 73 ) , and @xmath42 was the @xmath72.05 ns time delay between the adjacent wires . truncated wire position @xmath43 was defined as the nearest integer multiple @xmath37 of the 0.8128@xmath4 cm wire separation @xmath44 : @xmath45 the anode delay time difference @xmath46 depended nonlinearly on the anode position . linearized anode wire positions for each chamber were approximated by a polynomial expansion @xmath47 where the order @xmath48 of the fit with @xmath49 per degree of freedom @xmath50@xmath511 depended on the particular delay line . for our chambers the orders of the fits were 3 , 4 , or 5 . the coefficients @xmath52 were optimized with the program minuit @xcite that minimized the sum of squared deviations @xmath53 for each chamber in every run . the drift distance @xmath54 was determined from the drift timing @xmath55 using the calibration lookup table @xmath56 @xmath57 where @xmath58 was the drift time offset . the lookup table was calculated assuming that incoming cosmic muons are distributed uniformly over the equidistantly spaced wires . the final hit position @xmath33 was given by expression : @xmath59 where @xmath60 was the chamber @xmath33 coordinate offset , and @xmath61 and @xmath62 were the ( o@xmath26e ) adc signal and associated offset , respectively . the individual chamber phases @xmath63 in the eq . ( 5 ) and absolute horizontal coordinate offsets @xmath60 and @xmath64 between six chambers were adjusted with help of two - dimensional histograms @xmath65 vs @xmath66 and @xmath67 vs @xmath68 @xcite . data summary files were analyzed offline by compressing the dsf event data into paw ntuples @xcite . drift time - to - distance lookup tables and assorted chamber calibration constants , as well as csi detector adc pedestals and adc temperature corrections , were determined for each run separately . the high voltages of the csi detector phototubes were selected to give @xmath78 adc channels / mev scale . the fwhms of the pedestal peaks were typically 2 adc channels . therefore , the pedestal widths corresponded to an energy deposition of 0.25@xmath4mev , 24 @xmath25 m pathlength or 0.5@xmath4% of the most probable energy deposition by a minimum ionizing charged particle . drifts in the pedestal position over a one month period amounted to less than 2 adc channels . the precise horizontal coordinate offsets of six csi crystals inside the dark box were determined in the next stage of analysis . using the preliminary offsets read from the plastic template on which the crystals were laying , between 5 and 10 percent of the reconstructed cosmic muon trajectories were found not to intersect any of the predefined detector volumes . after adjusting the @xmath33 and @xmath34 translation software offset parameters of the csi modules by maximizing the number of tracks intersecting individual crystal volumes with nonzero adcs , the real number of events undergoing scattering in the apparatus or having the improperly reconstructed tracks was shown to be below 1@xmath4% . that percentage was in agreement with the monte carlo simulation of cosmic muons interacting with the experimental apparatus , showing the most probable muon scattering angle of 0.25@xmath27 and the 2.5@xmath27 mean scattering angle . the monte carlo root - mean - square of the scattering angle for the accepted events was 0.66@xmath27 , fig . [ fig : ths ] . that value translates into an average pathlength uncertainty of 1@xmath4 mm the estimated error in finding the correct crystal position inside the dark box was smaller than @xmath70.5@xmath4 mm , fig . [ fig : position22 ] . the energy deposited in csi crystal by cosmic muons along the fixed pathlength is a broadened distribution due to the statistical nature of energy transfers . maximal measured pathlengths were up to 12@xmath4 cm , while the average pathlength was @xmath76@xmath4 cm . a gaussian model for the energy loss distribution of the minimum ionizing particles in csi thicknesses of less than 12@xmath4 cm is not a good approximation . the distribution of energy losses is asymmetric , with a long high energy tail ; the most probable energy loss is smaller than mean energy loss @xcite . light yield temperature coefficients for the individual csi detectors were determined by the `` robust '' estimation @xcite of the adc values per unit pathlength as a linear function of temperature recorded inside the dark box . the least - squares condition assumes normally distributed measurements with constant standard deviations and is therefore not appropriate in this application , as pointed out above . typically , more than @xmath69 cosmic muon events , collected over at least one week and spanning the temperature range of @xmath72@xmath27c , were used as an input experimental data set . the fortran subroutine medfit , documented in the ref . @xcite , was adopted as the fitting procedure by imposing the requirement of minimum absolute deviation between the measured and calculated adc values per unit path that were dependent linearly on the temperature variable . the average light output temperature coefficient for 74 different csi detectors extracted by that method was @xmath70@xmath4%/@xmath27c , both for fast and total scintillation light components . these numbers are in good agreement with the previously reported value of @xmath261.5@xmath4%/@xmath27c @xcite , but the spread of extracted coefficients for different csi detectors was large : @xmath64@xmath4%/@xmath27c ( fig . [ fig : tf ] ) . we point out that the method did not allow us to separate the temperature dependence of the crystal light outputs from the temperature instabilities of the phototubes and high voltage dividers as well as the temperature instabilities of the adc modules themselves . the lecroy catalogue specification @xcite lists the typical temperature coefficient of a lcr 2249a adc unit as zero , and the maximum coefficient up to @xmath63@xmath4% for an adc gate longer than 100 ns and an average adc reading of 75 pc ( about a quarter of a full 256 pc scale ) . the raw adc values were corrected for the temporal temperature variations and written into the revised dsf files used in subsequent analysis . the contribution of the photoelectron statistics to the energy resolution of csi crystals was determined by a photodiode - based system . six individual detectors ( csi crystals with photomultiplier tubes and high voltage dividers ) having identical light emitting diodes ( leds ) coupled to their back sides were placed inside the cosmic muon tomography apparatus . the leds were pulsed at a 10 hz rate using a multichannel diode driver with continuously adjustable output voltage . one split output signal from each channel of the driver generated a 100 ns wide adc gate . the led light in a csi detector produced fast ( @xmath720 ns fwhm ) pmt anode pulses whose integrated values depended on the driving voltage and were equivalent to the fixed cosmic muon energy depositions between 10@xmath4mev and 100@xmath4mev . a total of five different led amplitudes were used in measurements of each csi detector . examples of the led spectra are shown in fig . [ fig : set2 ] . the adc pedestals were recorded simultaneously during the run . the intensity of led light was cross - calibrated against the cosmic muon spectra peaks in csi crystals . these muons were tracked in tomography drift chambers and both their pathlengths and energy depositions in the crystals could be easily calculated . that calculation enabled the establishment of the absolute energy scale in mev . the variance @xmath71 of a photodiode peak depended upon the mean number of photoelectrons @xmath72 on the photocathode created per unit energy deposition in the crystal : @xmath73 where @xmath74 was the led spectrum peak position and @xmath75 s were assorted variances , such as instabilities of the led driving voltage and temporal pedestal variations . five measured points ( @xmath71,@xmath74 ) were fitted with a linear function ( [ eq : led ] ) and the mean number of photoelectrons per mev @xmath76 for each csi detector was determined . the statistical error of the least - squares fits was typically less than one photoelecton / mev . using the previously extracted light output temperature coefficients all @xmath76 values were scaled to the 18@xmath27c point . that value was the designed operating temperature of the pibeta calorimeter . the number of extracted photoelectrons per mev for the fast scintillation component fell into the 20130 range as shown in fig . [ fig : npf ] . the hexagonal and pentagonal crystals equipped with three inch phototubes averaged 73 photoelectrons / mev , while the half - hexagonal and veto detectors with two inch phototubes had a mean of 33 photoelectrons / mev . the 73/33 ratio is very well explained by the ( 3/2)@xmath77 ratio of photosensitive areas for two different photocathode sizes . the measured photoelectron statistics were in agreement with the 100 ns adc gate values of @xmath76=(20260 ) for large ( @xmath710@xmath4 cm ) pure csi crystals equipped with two or three inch pmts reported in the past in refs . @xcite . the quantum efficiency of our bialkali photocathodes for pure csi scintillation light was 23@xmath4% @xcite , the average light collection probability for our detector shapes 23@xmath4% ( sec . [ tks ] ) and the fraction of deposited energy converted into the scintillation light was about 12@xmath4% . therefore , starting from the mean number of 73 photoelectrons / mev , we calculated that 1@xmath4mev energy deposited in the csi crystal produced on average about 10@xmath78 scintillation photons . geometrical layout of the experimental apparatus was defined using the geant detector description and simulation tools @xcite . the active elements of the simulated apparatus were an aluminum box housing six csi modules , six multiwire drift chambers , two scintillator planes and a layer of lead brick shielding , all shown in fig . [ fig : drift ] . generated events were muons with the energy , angular and charge distribution of a hard cosmic ray component at sea level . the zenithal angle @xmath79 distribution of muons at the ground was assumed to be proportional to @xmath80 ; the momentum spectrum between 0.1 and 10@xmath81 gev / c and the energy - dependent ratio of number of positive to number of negative cosmic muons was parameterized from the data given in refs . @xcite . the simulation trigger was defined by requiring a minimum energy deposition of more than 0.2@xmath4mev in each scintillator plane . this threshold was about one - tenth of the minimum - ionizing peak in the triggered scintillator and corresponded to the discriminator level used in the data acquisition electronics . penetrating cosmic muons and generated secondary particles were tracked through the apparatus , and energy depositions , pathlengths in csi crystals and hits in the drift chambers were digitized . all physical processes were turned on in the geant programs with the cutoff energies of 0.2@xmath4mev for charged particles and photons . inspection of simulated energy deposition spectra showed deviations from theoretical vavilov distributions @xcite expected in the planar detector geometries . the differences were caused by multiple scattering in the irregularly shaped crystals , and were particularly prominent for the events with shorter pathlengths close to the crystal edges . the simulated adc histograms revealed that triggering cosmic muons deposit on average 10.33@xmath4mev / cm in a csi detector , with the most probable energy loss 5.92@xmath4mev / cm . cosmic muon spectra in csi detectors were described in a satisfactory way ( @xmath50@xmath511.2 ) by the combination of a gaussian distribution and a falling exponential tail : @xmath82 ^ 2 } + \theta(\epsilon- \epsilon_0)e^{\alpha-\beta\epsilon},\ ] ] where @xmath83 is energy ( in mev ) deposited in one full csi module . parameters @xmath84 , @xmath85 , @xmath86 , @xmath87 , and @xmath15 are extracted from the geant spectra by imposing the least squares constraint to the fit and leaving the pathlength @xmath88 ( cm ) as free parameter : @xmath89 and @xmath90 these monte carlo spectra , broadened with photoelectron statistics , were used to describe our cosmic muon lineshapes produced by the optically uniform csi detectors . propagation of scintillating photons through a uniform detector with ideal reflecting dielectric surfaces and different wrapping materials was studied using the tkoptics simulation program @xcite . the code is a library of fortran subroutines with an x - windows - based user interface written with tck / tk toolkit @xcite . the program can simulate the light output response of an arbitrarily shaped scintillation detector with given bulk and surface optical properties . the polygonal detector shape is defined by the coordinates of its vertices . the detector attributes are reflector types of lateral and front crystal surfaces and wrapper material , crystal surface - wrapper gap distances and characteristics of the photomultiplier tube and a phototube - crystal joint coupling . the program handles normal dielectric , specular , diffuse and partially absorbent reflector surfaces with arbitrary diffuse fraction , roughness , and specular @xmath91 and diffuse reflectivity @xmath92 . predefined bulk properties of a detector are the medium scattering and attenuation length as well as refractive index . a photomultiplier tube is specified by the diameter , position and quantum efficiency of the photocathode . different choices of initial scintillating photon distribution are possible , the default being a uniform starting distribution throughout the scintillating volume . a working volume is a box divided into elementary cubic cells of fixed size . output menu options include the initial and endpoint photon coordinates , and direction vectors and timing distributions of the detected scintillation photons organized in a paw ntuple . every elementary cell is flagged as a bulk or edge crystal cell . center coordinates of the cells and a fraction of photons generated in every cell , and subsequently detected on the pmt sensitive surface , are always recorded . results of high statistics runs with @xmath93 photons generated uniformly through the detector volume were plotted to show the number of photoelectrons as a function of scintillation source position inside the crystal . the relative statistical errors of calculated light collection probabilities for the bulk crystal cells were @xmath942@xmath4% . the bulk attenuation length and the light scattering length of the near - ultraviolet emission component inside the csi crystals were set to 150@xmath4 cm and 200@xmath4 cm , respectively @xcite . the index of refraction for the csi medium increases from 1.82 for the blue - green light to about 2.08 for ultraviolet light @xcite . the simulated light collection probabilities @xmath95 that depend on axial @xmath96 and transverse @xmath33 positions are shown integrated in the vertical coordinate ( @xmath34 ) in figs . [ fig : hexa_9 ] and [ fig : hh1d_9 ] . in the form of `` lego '' plots for one wrapping configuration and two different crystal shapes . the coordinate system is defined with the @xmath96=0@xmath4 cm origin at the front face of the crystal and a photocathode window at the @xmath96=22@xmath4 cm plane . we find that simulated light response of the ideal hexagonal or pentagonal pibeta csi detector with specular lateral surfaces and a diffuse wrapping material can be described with the following parameters : 1 . the average photon collection probability @xmath97 for a three inch photocathode is about 23@xmath4% for ideal @xmath91@xmath981.0 crystal surfaces and a @xmath92@xmath980.9 diffuse wrapper , decreasing by half , to 12@xmath4% for @xmath91@xmath99 . the axial light collection probability variation through the first 10 centimeters is in the @xmath100% range , with a positive slope @xmath101 for higher specular and diffuse reflectivities , namely for @xmath102@xmath310.9 . the axial detected light variation in the back crystal half ( @xmath96@xmath3110@xmath4 cm ) is up to @xmath103% . the transverse light response referenced to the light yield at the crystal axis typically increases towards the crystal surfaces by up to @xmath55@xmath4% for @xmath96@xmath10410@xmath4 cm , but is generally declining away from the central axis by @xmath10530@xmath4% at the @xmath96=18@xmath4 cm plane . the root - mean - square of a three - dimensional light nonuniformity is between 2.5@xmath4% and 3.5@xmath4% for @xmath96@xmath10410@xmath4 cm , compared to a @xmath720@xmath4% root - mean - square for a @xmath96@xmath106@xmath4 cm volume , where a large spread is caused by the inefficient light collection from crystal back corners . for lateral surfaces painted with a @xmath92@xmath510.9 diffuse substance without an air gap we find that the average photon collection probability is lower , @xmath717@xmath4% , and the axial light collection nonuniformity as a function of axial position is always positive and is usually larger than @xmath510@xmath4% in a 22@xmath4 cm long detector , making the root - mean - square of the 3d nonuniformity function @xmath10720@xmath4% . the monte carlo results for the optically uniform half - hexagonal and trapezial csi crystals with the same range of optical properties predict smaller light yields and somewhat higher light collection nonuniformities : 1 . the average photon collection probability with a two inch photocathode is about 12@xmath4% . the typical axial nonuniformity is positive in front , @xmath55@xmath4%/10 cm , and negative in the back crystal half , with a @xmath103%/10@xmath4 cm variation , 3 . the simulated transverse light output from the front crystal half is very similar to one for the full crystal shapes , increasing up to @xmath53@xmath4% away from the detector central axis , but decreasing by almost a third in the back corners of the crystal . the root - mean - square of detected light output varying between 3.0@xmath4% and 4.5@xmath4% , in the front ( @xmath96@xmath10410@xmath4 cm ) and back ( @xmath96@xmath3110@xmath4 cm ) crystal halves , respectively . these light collection probability distributions calculated for optically homogeneous crystals explain the major features of the measured optical nonuniformities presented below . the three - dimensional ( 3d ) spatial distribution of the light output of a scintillation detector can be specified by giving the number of photoelectrons @xmath108 produced by 1@xmath4mev energy deposition at point ( @xmath109 ) ( `` 3d light nonuniformity function '' ) . in the following discussion we limit ourselves to the linear one - dimensional variations of the detected light output separable in the axial and transverse directions : @xmath110 where @xmath111 ( @xmath112 ) is the linear optical nonuniformity coefficient in the @xmath96 ( @xmath33 ) coordinate , and the coordinate system origin is at the center of the detector front face , as explained in sec . [ tks ] . a simple and straightforward parameterization of the detector light response nonuniformity can be made on the basis of scatter plots showing the light output per unit pathlength as a function of longitudinal or transverse cosmic muon - csi crystal intersection coordinates . in this analysis we take only cosmic muon events with almost perpendicular trajectories ( @xmath79@xmath3185@xmath27 ) , so that the measured two - dimensional light outputs are averaged over the pathlengths and over @xmath113@xmath20 vertical cross sections . the cosmic muons deposit the energy along the well - defined ionization track of the length @xmath88 in the crystal . this description will somewhat underestimate the real light collection probabilities ( see sec . [ 137s ] ) due to the integration of a three - dimensional light nonuniformity function along the charged particle track . our work on the fully 3d reconstruction of the scintillator light response will be presented in a forthcoming publication @xcite . [ fig : expected6 ] shows the axial variation of normalized adc values as a function of distance of energy deposition from the front crystal face for the six representative crystals . the axial positions were calculated by averaging two @xmath96 values of the cosmic muon track intersection with the crystal surfaces . panels on figs . [ fig : expected7a ] and [ fig : expected7c ] show the transverse dependence of measured light output per unit pathlength , where the independent variable is the distance from the crystal axis measured in the horizontal plane . the selected scatter plots show the data points for the axial slices at @xmath114=6@xmath1151@xmath4 cm and @xmath114=18@xmath1151@xmath4 cm . we have chosen to parameterize the axial light output nonuniformity with two piecewise linear functions : the light output nonuniformity coefficient @xmath116(@xmath96@xmath10410@xmath4 cm ) for the front half of the crystal , and the light nonuniformity coefficient @xmath117(@xmath96@xmath3110@xmath4 cm ) for the back half of the crystal , both values expressed in % /cm . the transverse light output variation at a fixed axial distance @xmath114 , @xmath118 , is described by the average value of a change in luminosity left and right of the crystal axis . several general features are readily noticeable : the spread of measured points due to energy deposition straggling , the gradual variation in the collected light along and perpendicular to the detector axis , and the decrease in detected light when the photon generation occurs close to lateral detector surfaces and , in particular , near the back corners of the detector volumes . the calculated adc / pathlength data points shown on the panels of figs . [ fig : expected6][fig : expected7c ] have been fitted with linear functions imposing the `` robust '' condition of the minimum absolute deviation between the measured and calculated values , eqs . the normalized light output scatter plots in the axial and transverse coordinates , as well as two - dimensional adc/@xmath88 distributions in the @xmath33-@xmath96 bins and associated linear light output nonuniformity coefficients have been documented for all studied csi crystals and are available for inspection at the pibeta www site @xcite . the average scintillation properties of all the measured csi crystals are listed in tables [ tab1 ] and [ tab2 ] . in summary , for a set of fifty - nine full hexagonal and pentagonal crystals we find : 1 . the average axial light nonuniformities are @xmath116(@xmath119@xmath4cm)=@xmath260.1@xmath4%/cm and @xmath117(@xmath120@xmath4cm)=@xmath261.3@xmath4%/cm , respectively ( see fig . [ fig : a1a2 ] ) . 2 . distribution of nonuniformity coefficients could be described by a gaussian with @xmath121@xmath511.3@xmath4%/cm . 3 . the average transverse light output is flat for @xmath96@xmath10410@xmath4 cm , and the typical light variation is @xmath122% at @xmath96=18@xmath4 cm . 4 . the kharkov - grown crystals on average have twice the axial optical nonuniformity of the bicron - grown crystals . for fifteen half - hexagonal and trapezial crystal shapes we find : 1 . the average axial nonuniformities are @xmath116(@xmath96@xmath10410 cm)=@xmath1050.3@xmath4%/cm and @xmath117(@xmath96@xmath3110@xmath4cm)=@xmath261.0@xmath4%/cm . the transverse light output variation is limited in the front crystal half , but increases up to @xmath10530@xmath4% in the back of the crystal . with the insights gained in the simulation of the light collection probabilities ( sec . [ tks ] ) we conclude that the @xmath62@xmath4%/cm spread of the nonuniformity coefficients corresponds to an equivalent range of 0.81.0 in crystal surface reflectivities , and/or 100250@xmath4 cm range in csi attenuation lengths . a light - tight plywood box was made to house a single csi detector and associated measurement apparatus described below . a 0.662@xmath4mev @xmath22cs gamma source was placed on a moveable support next to the detector and collimated by a 20@xmath810@xmath85@xmath4@xmath18 lead brick with a 6@xmath4 mm collimator hole . the photomultiplier analog signal from the detector was processed with an ortec 454 timing amplifier with a gain of 30 and a 50 ns integration time constant . the amplifier output was digitized with a peak sensing ortec ad811 adc unit . the same signal was discriminated and produced the trigger rate of about 5 khz with a @xmath22cs source present . the background rate , without the source present , was 50100 hz . pedestal runs were taken with a clock trigger and used to properly correct offsets of the adc spectra . temperature variation during one run was less than 0.3@xmath27c , typically causing the overall light output variation of less than 0.5@xmath4% in a adc spectrum that was gated with a 100 ns window . several runs were taken with the source removed to find the shape of the background spectrum . it was confirmed that the background spectrum does not depend on the position of the lead collimator , so the same background lineshape was used in the analysis for all source positions . the collimated @xmath22cs source was placed by remote control in turn at five points along the axis of each crystal at 2 , 6 , 10 , 15 and 20@xmath4 cm from the front face of the crystal . following background subtraction , recorded spectra were fitted with a sum of a gaussian and exponential function , fig . [ fig : cs137 ] . the peak position and the fwhm for each spectrum were extracted with the statistical uncertainty lower than 0.3@xmath4% . the dependence of the peak position on the placement of the source is illustrated in fig . [ fig : cs137f ] and follows the trend of the tomography results . the geant simulations revealed that the 662 kev gamma rays could probe all the volume of csi crystal , but the shower energy deposited at the central axis is only about one - sixth of the energy deposition near the crystal surface . the cosmic muons transfer the energy uniformly along their tracks in csi material : the scintillation volume over which the measured light output is integrated is therefore larger , and inherent averaging leads to smaller extracted optical nonuniformity coefficients . this feature is borne out in the panels of fig . [ fig : cs137 ] where the tomography data for three different crystals yield consistently smaller axial light nonuniformities when compared with the radioactive source measurements . the simple @xmath22cs scans with the described apparatus were used to evaluate the light collection nonuniformities for up to six crystals per day . radioactive source scan measurements similar to ours are described in refs . we have studied the simulated pibeta calorimeter response to the 69.8@xmath4mev @xmath0@xmath1@xmath3 positrons and to the ( 69.8 + 2@xmath123 ) mev photons , where @xmath123 is the positron rest mass . the goal was to find the intrinsic difference between the responses to monoenergetic positrons and photons with identical total incident energies , emphasizing the correct modeling of the low - energy tail below the edge of the michel @xmath25@xmath1@xmath124 spectrum at 52.8@xmath4mev . the 0.46@xmath4% accuracy of the new triumf measurement @xcite of the @xmath0@xmath1@xmath3 branching ratio was limited by event statistics and systematic uncertainties in evaluating the low - energy tail of the positron peak . the @xmath0@xmath1@xmath125 positrons and the forward - peaked bremsstrahlung gammas were detected in a 46-cm - diameter@xmath851-cm - long nai(tl ) crystal `` tina '' . the tail correction of 1.44@xmath60.24@xmath4% for the energy region from 0@xmath4mev to 52.3@xmath4mev was determined by subtracting the measured michel positron spectrum from the detector positron response functions measured with the 2085@xmath4mev @xmath126 monoenergetic beams . an additional tail component of @xmath70.4@xmath4% due to radiative processes was estimated by monte carlo method . the triumf group apparently made no attempt to account for the potential difference in the scintillator light response of positrons and photons and neglected potential nonlinearities of the energy scale . both effects could arise from the light collection nonuniformities of the nai(tl ) detector . in the recent psi @xmath0@xmath1@xmath3 experiment @xcite the quoted 0.29@xmath4% systematic uncertainty of the extracted branching ratio was also dominated by the 1.64@xmath60.09@xmath4% electromagnetic loss fraction as well as by the 0.95@xmath60.19@xmath4% contribution from the photonuclear reactions . the @xmath74@xmath21 calorimeter used in that measurement was made of 132 hexagonally shaped bgo crystals , with the light yield claimed to be homogeneous to within 1.5@xmath4% over the whole 20@xmath4 cm length of every crystal . the lineshapes of positrons and photons in our pibeta calorimeter and their dependence on the optical properties of 240 constituent csi crystals were predicted using the geant detector description and simulation tools . the code defined the geometry and tracking media for the complete pibeta detector @xcite . the inner detector region was occupied by two cylindrical wire chambers and a segmented plastic veto detector . the incident particles , monoenergetic positrons and photons , were generated in the center of the crystal sphere . photonuclear reactions in the active detectors as well as in the surrounding passive material were modeled in a user subroutine @xcite that was added to the default geant version 3.21 . the probabilities of photonuclear reactions were calculated using the published cross sections from the @xmath127 reaction thresholds up to the energy of 120 mev @xcite . the 240 individual csi module shapes were specified , taking into account the mechanical tolerances of the physical csi crystals , with teflon and aluminized mylar wrappings filling the 200 @xmath25 m intermodule gaps . the irregular csi modules had to be constructed from up to six geant generalized trapezoidal wedges . every csi crystal volume was considered a sensitive detector with the associated luminosity @xmath76 expressed as the number of photoelectrons per mev , and two axial and one transverse light collection nonuniformity coefficients , @xmath116 , @xmath117 , and @xmath112 , respectively . a set of seventy - four detector luminosity and nonuniformity values , extracted in the tomography analysis , were initialized in the geant - accessible database . the optical parameters of the remaining 163 modules were drawn randomly from the distributions in figs . [ fig : npf ] and [ fig : a1a2 ] . the assumption was that the crystals produced in the future will be of the same optical quality as the ones that are already delivered . we have always simulated at least 10@xmath36 events for every fixed set of the calorimeter parameters ( @xmath76,@xmath116,@xmath117,@xmath112 ) . the low - threshold trigger was defined by requiring a sum of the adc readings exceed 5@xmath4mev of the light - equivalent energy for one calorimeter supercluster containing 32 crystals . the integrated detector acceptances with a low - level trigger were 85.8@xmath4% for @xmath770@xmath4mev positrons and 85.5@xmath4% for @xmath770@xmath4mev photons , so all extracted quantities had the relative statistical uncertainties of @xmath70.2@xmath4% . the response of the calorimeter was parameterized by the fwhm of the simulated adc spectrum and the tail contribution being between the adjustable low level ( default lt being 5@xmath4mev ) and high level thresholds ( two default values ht@xmath128 and ht@xmath28 being 54@xmath4mev and 55@xmath4mev , respectively ) . for the purposes of comparison and gain normalization we first studied the response of an ideal , homogeneous calorimeter , with realistic csi crystal light outputs . the `` software '' gain factor of every individual csi detector was determined from the fitted peak positions of the simulated adc spectrum sum over the crystal with maximum energy deposition and its nearest neighbors . the peak value positions were found after smoothing the histograms with a multiquadric function , to improve on the limited simulation statistics . the ratios of peak positions in the ideal , homogeneous calorimeter and the corresponding values for the physical , nonuniform calorimeter , were , by definition , the individual detector gain corrections . the values of these ratios were refined in three steps of the iterative procedure . the extracted energy resolution of the nonuniform detector in principle depends on the convergence of the gain - matching process and the resulting uncertainties of the individual detector gains . that gain - matching process was equivalent to the pmt high voltage adjustments in the real experiment that are effected to obtain the best energy resolution with the modular detector . we estimate that our simulation procedure fixes the software detector gains with the accuracy of @xmath71@xmath4% . the set of gain constants depended on the optical properties of our csi crystals as defined in our geant database , but also upon the incident particle chosen for the monte carlo calibration runs and its energy because of differences in shower developments of photons and positrons . the gain normalizations of the calorimeter detectors were calculated by aligning the simulated @xmath0@xmath1@xmath3 positron peak positions . the same procedure for matching the csi detector gains was used in the real data - taking runs . due to the gain renormalization , the influence of transverse light nonuniformities smaller than 10@xmath4% on the adc spectrum peak positions and the adc lineshape could be neglected in comparison with the axial light nonuniformity effects . the simulated adc values were summed over all detectors with the energy deposition above the 1@xmath4mev tdc threshold . we also examined the calculated adc sums for the clusters which contained the crystal with the maximum energy deposition and its nearest neighbors . the full - widths at half maximum for these spectra are labeled in the following figures and tables as fwhm@xmath129 and fwhm@xmath130 , respectively . no event - to - event uncertainties of the adc pedestal values were assumed in the simulation . stability of the pibeta electronics tested under real experimental conditions in 1996 calibration runs and the quality of the algorithms used for the first and second pedestal correction reduces the pedestal peak root - mean - square to @xmath78 channels , equivalent to @xmath70.3@xmath4mev . the geant - calculated resolution of the pibeta calorimeter consisting of 220 csi detector crystals and 20 veto crystals was parameterized by the fractional full width at half maximum : @xmath131 for the positrons with the most probable peak energy @xmath132 mev and @xmath133 for the monoenergetic photons with the detected peak energy @xmath134 mev ( fig . [ fig : fit_fwhm_220 ] ) . the top lines refer to the cases of an optically uniform detector ( @xmath111=0 ) while the bottom equations describe the predicted response of the nonuniform calorimeter ( @xmath111=tomography ) . the average fractional energy resolution fwhm@xmath129 of 5.3@xmath4% ( 6.0@xmath4% ) was achieved for the @xmath770@xmath4mev kinetic energy positrons in the optically uniform ( nonuniform ) calorimeter , as compared to 5.7@xmath4% ( 6.8@xmath4% ) resolution for the equivalent energy photons . that was the resolution found with no cuts applied on the light - equivalent energy generated in the calorimeter vetoes . requiring less than 5@xmath4mev detected in the veto shield decreased the statistics by about 10@xmath4% and suppressed the low - energy tail , but did not improve the energy resolution at the peak position . limiting the adc sums to the nearest - neighbor crystals ( six or seven crystal clusters ) the fractional fwhm@xmath130 energy resolution could be parameterized by : @xmath135 for the positrons with the detected peak energy @xmath132 mev and @xmath136 for the monoenergetic photons with the detected peak energy @xmath134 mev ( fig . [ fig : fit_fwhm_nn ] ) . the average fwhm@xmath130 for the 70@xmath4mev positrons and gammas in nearest - neighbor nonuniform crystal clusters was 8.7@xmath4% ( 6.1@xmath4mev ) and 9.9@xmath4% ( 6.8@xmath4mev ) , respectively . imposing the 5@xmath4mev cut on the detector veto signals improves these resolutions only marginally . these numbers should be compared with the response width of the ideal uniform nearest - neighbor clusters : 8.2@xmath4% ( 5.7@xmath4mev ) and 8.7@xmath4% ( 6.2@xmath4mev ) . the percentage of the events in the tail between the preset low and high energy threshold was tracked in the same geant simulation . the listing of the low - energy tail contributions for 69.8@xmath4mev @xmath126 and 70.8@xmath4mev @xmath23 in table [ tab3 ] shows that in the optically uniform calorimeter they differ by 1.8@xmath4% . the optical nonuniformity does not change the positron tail contribution , but increases the photon tail by @xmath70.2@xmath4% . the imposition of the 5@xmath4mev veto - shield cut decreases the electromagnetic leakage to 2.4@xmath4% for positrons and 4.7@xmath4% for photons , a @xmath126-@xmath23 tail difference of @xmath52.3@xmath4% . table [ tab4 ] shows the low - energy tail components for the simulated nearest - neighbor cluster adc sums : with no applied cuts the contributions were 4.7@xmath4% and 6.7@xmath4% for @xmath126 s and @xmath23 s , respectively . the veto shielding cuts decreased both fractions by about 0.3@xmath4% . if the positron originates from a radiative @xmath0@xmath1@xmath125 decay , its average monte carlo peak position is essentially unchanged . its tail contribution is unaffected if simulated adc sums extend over full calorimeter but are increased by @xmath71.4@xmath4% if only the nearest - neighbor clusters are summed . the radiative decay matrix elements used in the calculation were taken from ref . all quoted numbers have an approximate systematic uncertainty of @xmath70.2@xmath4% . the detector response to positrons and photons in the energy range between 10@xmath4mev and 120@xmath4mev is also nonlinear because the volume distribution of the shower energy deposition depends on the incident particle type and the energy . our geant studies showed that the magnitude and energy dependence of these nonlinearities does not change significantly because of csi crystal light collection nonuniformity . results of the calculations , both for the case of a homogeneous and optically nonhomogeneous calorimeter , are displayed in figs . [ fig : fit_nonlin_220 ] and [ fig : fit_nonlin_nn ] . nonlinearities of the detector response in the covered energy range for positrons and photons are very close in magnitude and shape of energy variation and amounted to @xmath71.5@xmath4% for the adc sums over 220 crystals and up to @xmath72.8@xmath4% if the simulated adc sums were restricted to the over - the - threshold adcs of nearest neighbors . we have measured the optical properties of seventy - four pure cesium iodide crystals that were polished and wrapped in the diffuse teflon reflector . the results are summarized separately for the full - sized and half - sized crystals in tables [ tab1 ] and [ tab2 ] . the deduced light yields parameterized by two axial and one transverse light collection nonuniformity coefficients constitute a minimum set of parameters necessary for a realistic monte carlo simulation of the modular csi calorimeter . the predicted energy resolutions fwhm@xmath129 for @xmath770@xmath4mev positrons and photons in the full pibeta calorimeter with the ideal , optically uniform csi modules with specified luminosities were shown to be close , 3.7@xmath4mev and 4.0@xmath4mev , respectively . the upper limit of the low energy tail contributions in the region between 5@xmath4mev and 55@xmath4mev were calculated to be 6.9@xmath4% and 8.7@xmath4% for the positrons and gammas in an optically homogeneous detector , respectively . after applying the 5@xmath4mev calorimeter veto cut , these tail corrections decrease to 2.2@xmath4% and 4.5@xmath4% . the average deduced axial light nonuniformities of real csi crystals wrapped in a teflon sheet had negative slopes , @xmath1050.18@xmath4%/cm and @xmath1051.6@xmath4%/cm , for the front and back half of the crystal volume , respectively . the corresponding geant simulation of the nonuniform pibeta apparatus for 68.9@xmath4mev @xmath126 and 70.8@xmath4mev @xmath23 s shows that both energy responses will be broadened to the average fwhm@xmath129 of 5.9@xmath4% and 6.9@xmath4% while the associated tail contributions will change only for photon spectra , increasing the low - energy tail by @xmath70.3@xmath4% . the simulated calorimeter adc spectra are shown in figs . [ fig : tails1 ] and [ fig : tails2 ] . the nonlinearity of the measured energy scale caused by the optical nonuniformity is @xmath942.8@xmath4% throughout the relevant @xmath126/@xmath23 energy range . this spread is consistent with the precision of energy calibration required to extract the tail corrections with the systematic uncertainty of @xmath70.2@xmath4% . the predicted adc spectra of the monoenergetic positrons , electrons and tagged photons in the energy range 1070@xmath4mev will be compared with the measured responses of the partial csi calorimeter arrays in a forthcoming publication @xcite . the authors wish to thank micheal sadler of the abilene christian university ( acu ) for lending us the drift chamber tomography apparatus . derek wise , also of acu , has helped with the cosmic muon tomography measurements . roger schnyder of paul scherrer institut has maintained and repaired the faulty electronics modules . their help is gratefully acknowledged . 9 d. poani , d. day , e. frle , r. m. marshall , j. s. mccarthy , r. c. minehart , k. o. h. ziock , m. daum , r. frosch and d. renker , psi r-89.01 experiment proposal ( paul scherrer institute , villigen , 1988 ) . g. czapek , a. federspiel , a. fluckiger , d. frei , b. hahn , c. hug , e. hugentobler , w. krebs , u. moser , d. muster , e. ramseyer , h. scheidiger , p. schlatter , g. stucki , r. abela , d. renker and e. steiner , phys . 70 ( 1993 ) 17 . e. frle , ch . brnnimann , t. flgel , j. e. koglin , b. krause , d. poani , d. renker , s. ritt , p. l. slocum , _ the light response of pure csi calorimeter crystals painted with waveshift lacquer _ , to be submitted to nucl . phys . and meth . e. frle , s. ritt , k. a. assamagan , ch . brnnimann , t. flgel , b. krause , d. poani , d. renker , p. l. slocum , n. soi , i. supek , _ three - dimensional reconstruction of light collection probability function for pure csi scintillators _ , to be submitted to nucl . phys . and meth . e. frle , k. a. assamagan , ch . brnnimann , t. flgel , j. e. koglin , b. krause , d. w. lawrence , r. c. minehart , d. poani , d. renker , b. g. ritchie , s. ritt , p. l. slocum , and i. supek , _ the response of the pibeta csi calorimeter array to 1070 mev positrons , electrons and tagged photons _ , to be submitted to nucl . phys . and meth . .average scintillation properties of hexagonal and pentagonal pibeta csi calorimeter shapes pentas , hex as , hex bs , hex cs and hex ds ( 59 crystals ) . all crystals were polished and wrapped in two layers of a teflon membrane and one layer of aluminized mylar film . the light yields are normalized to the temperature of 18@xmath27c . all other parameters were measured at the average laboratory room temperature of 22@xmath27c . [ cols="<,^,^ " , ]

we study yang baxter deformations of the superstring with non - abelian classical-matrices which satisfy the homogeneous classical yang baxter equation ( cybe ) . by performing a supercoset construction , we can get deformed backgrounds . while this is a new area of research , the current understanding is that abelian classical-matrices give rise to solutions of type iib supergravity , while non - abelian classical-matrices lead to solutions of the generalized supergravity equations . we examine here some examples of non - abelian classical-matrices and derive the associated backgrounds explicitly . all of the resulting backgrounds satisfy the generalized equations . for some of them , we derive `` t - dualized '' backgrounds by adding a linear coordinate dependence to the dilaton and show that these satisfy the usual type iib supergravity equations . remarkably , some of the `` t - dualized '' backgrounds are locally identical to undeformed after an appropriate coordinate transformation , but this seems not to be generally the case . * generalized type iib supergravity equations + and non - abelian classical-matrices * + domenico orlando , susanne reffert , + jun - ichi sakamoto , and kentaroh yoshida _institute for theoretical physics , albert einstein center for fundamental physics , university of bern , sidlerstrasse 5 , ch-3012 bern , switzerland _ +_department of physics , kyoto university , + kitashirakawa oiwake - cho , kyoto 606 - 8502 , japan _

1607.00795

yang baxter deformations provide a systematic way to study integrable deformations of non - linear sigma models in two dimensions . this method was invented by klimcik for principal chiral models based on the modified classical yang baxter equation ( mcybe ) @xcite . it was generalized to the symmetric coset case in @xcite and the homogeneous cybe case in @xcite . it is very interesting to consider some applications of yang baxter deformations in the context of the ads / cft correspondence @xcite . the green - schwarz action of type iib superstring on @xmath0 was constructed by metsaev and tseytlin in terms of the supercoset @xcite @xmath4 this supercoset has a @xmath5-grading , generalizing the @xmath6-grading of the symmetric cosets . hence the @xmath0 superstring exhibits classical integrability in the sense of kinematic integrability @xcite ( for nice reviews , see @xcite ) . the first application of yang baxter deformation to the @xmath0 superstring was carried out by delduc , magro and vicedo @xcite . the classical action was constructed with a classical @xmath1-matrix of drinfeld - jimbo type @xcite and the associated symmetry algebra is a @xmath7-deformation of @xmath8 . this specific example is often called @xmath9-deformation . the metric and ns - ns two - form were subsequently computed by performing a coset construction for the bosonic element @xcite . after that , some works were done towards constructing the full background by directly solving the equations of motion of type iib supergravity @xcite . this strategy was preferred simply because of the technical difficulty of the supercoset construction . in the end , arutyunov , borsato and frolov accomplished doing the supercoset construction and the full background was determined @xcite ( for a good review , see @xcite ) . a remarkable point is that the resulting background does not satisfy the equations of motion of type iib supergravity . however , the situation is not hopeless . the ten - dimensional background is related to the full solution of the usual type iib supergravity via t - dualities up to the linear dependence of the dilaton @xcite . the world - sheet theory is not weyl invariant but still scale invariant , _ i.e. _ , the world - sheet beta function does not vanish but is given by a total derivative . from the viewpoint of the spacetime , this result indicates that the on - shell condition of type iib supergravity is weakened to a generalized set of the equations . in fact , the generalized equations of type iib supergravity were proposed in @xcite and the background obtained in @xcite is a solution of the generalized equations . for a more general argument on the relation between the @xmath10-symmetry of the @xmath9-deformed @xmath0 superstring and the generalized equations , see @xcite . other examples of yang baxter deformations of the @xmath0 superstring are based on the homogeneous cybe @xcite . in comparison to the mcybe case , there are some advantages . one of them is that one can perform partial deformations of @xmath0 , i.e. , only ads@xmath11 or only @xmath12 can be deformed . this class of yang baxter deformations includes well - known examples such as gravity duals of non - commutative gauge theories @xcite , the gamma - deformations of @xmath12 @xcite , and schrdinger spacetimes @xcite . in a series of works @xcite , the associated classical @xmath1-matrices were identified with these backgrounds , flat spacetime and pp - wave , see @xcite . for short summaries , see @xcite . ] . in a remarkable advance , the supercoset construction has been performed in @xcite and now the full background can be obtained for arbitrary classical @xmath1-matrices . for abelian classical @xmath1-matrices , the r - r sector and dilaton have been confirmed for well - known backgrounds and it seems likely that yang baxter deformations work well @xcite . however , for non - abelian classical @xmath1-matrices , the associated backgrounds do not in general satisfy the equations of motion of type iib supergravity @xcite . an interesting question here is whether deformed backgrounds for non - abelian classical @xmath1-matrices satisfy the generalized equations of motion or not . this question can be answered if the deformed string theory is supposed to be the canonical action of the green - schwarz string @xcite . this assumption however has to be confirmed by a separate analysis and it is still an open problem . a confirmation was also given in @xcite based on scaling limits of the @xmath9-deformed ads@xmath13 . it may also be intriguing to study `` t - dualized '' backgrounds related to non - abelian classical @xmath1-matrices and consider their physical interpretation . we use quotation marks for t - dualities involving solutions to the generalized sugra equations , where the usual buscher rules have to be supplemented with a prescription for the behavior of the dilaton @xcite . in this paper , we consider yang baxter deformations of the @xmath0 superstring with the homogeneous cybe and focus on some examples of non - abelian classical @xmath1-matrix . we derive the associated backgrounds explicitly and show that all of the resulting backgrounds satisfy the generalized equations . for some of them , `` t - dualized '' backgrounds are derived by adding a linear coordinate dependence to the dilaton and these satisfy the usual type iib supergravity equations . remarkably , some of the `` t - dualized '' backgrounds are locally t - dual to the undeformed @xmath0 via appropriate coordinate transformations , though it does not seem that this is always the case . at least in cases where the undeformed @xmath0 background can be reproduced , the classical integrability of the `` t - dualized '' background is manifest . this paper is organized as follows . section [ sec : yang - baxter - deformed ] introduces yang baxter deformations of the @xmath0 superstring based on the homogeneous cybe and gives an outline of the supercoset construction . in section [ sec : generalized - type - iib ] , we briefly introduce the generalized type iib supergravity equations of motion . in section [ sec : examples ] , we study six examples of non - abelian classical @xmath1-matrices and the associated full backgrounds . for four of them , `` t - dualized '' backgrounds , which are solutions of the usual type iib supergravity , are found . then , three of the `` t - dualized '' backgrounds are shown to be locally equivalent to the undeformed @xmath0 . section [ sec : conclusions ] is devoted to conclusion and discussion . in this section , we shall give a short summary of yang baxter deformations of the @xmath0 superstring based on the homogeneous cybe and an outline of the supercoset construction . the action of the deformed system is given by @xmath15\ , , \label{ybsm}\end{aligned}\ ] ] where the left - invariant one - form @xmath16 is defined as @xmath17 with the world - sheet index @xmath18 . we are in conformal gauge and the world - sheet metric takes the diagonal form @xmath19 . hence there is no coupling of the dilaton to the world - sheet scalar curvature . the anti - symmetric tensor @xmath20 is normalized as @xmath21 . the constant @xmath22 is the usual t hooft coupling . the deformation is measured by a constant parameter @xmath9 . when @xmath23 , the undeformed @xmath0 action @xcite is reproduced . an important quantity here is the chain of operations @xmath24 defined as @xmath25 where the linear operator @xmath26 is a solution of the homogeneous cybe : @xmath27 this @xmath28-operator is related to the _ skew - symmetric _ classical @xmath1-matrix in the tensorial notation through the following formula : @xmath29 = \sum_i\qty ( a_i \operatorname{str}[b_ix ] - b_i \operatorname{str}[a_ix ] ) \ , . \label{r - r}\end{aligned}\ ] ] here @xmath1 is written as @xmath30 the projection operator @xmath31 is a linear combination of projectors defined as @xmath32 where @xmath33 are projectors to the @xmath5-graded components of @xmath8 . in particular , @xmath34 is a local symmetry of the classical action , @xmath35 . the numerical coefficients in ( [ op - d ] ) are fixed by requiring kappa - symmetry @xcite . the classical action ( [ ybsm ] ) is written in terms of a group element @xmath36 and a coordinate system can be introduced by fixing an explicit parametrization . first of all , the group element @xmath37 is represented by the product of the bosonic and fermionic elements @xmath38 and @xmath39 as follows : @xmath40 for our later convenience , let us parametrize the bosonic element @xmath38 as @xmath41\ , \exp[\qty(\log z ) d ] \ , , \\ g_{\rm b}{}^{\text{s}^5}&= \exp[\frac{i}{2 } ( \phi_1\ , h_1+\phi_2\,h_2+\phi_3\,h_3)]\ , \exp[\,\xi\,{{\rm\bf j}}_{68 } ] \exp[-i\,r\,{{\rm\bf p}}_6]\ , . \end{aligned}\ ] ] here , @xmath42 and @xmath43 are translations and dilatation @xmath43 in the four - dimensional conformal algebra @xmath44 , and @xmath45 are the cartan generators of @xmath46 . for the other generators and the details of the notation , see @xcite . the coordinates @xmath47 and @xmath48 describe the poincar ads@xmath11 , and @xmath49 parametrize the round five - sphere . the resulting metric is given by @xmath50 , where @xmath51 then the fermionic group element @xmath39 is generated by the supercharges @xmath52 as follows : @xmath53 here , @xmath54 are grassmann - odd coordinates and correspond to a couple of 16-component majorana - weyl spinors satisfying the majorana condition ( see @xcite for details ) . the coset construction for the bosonic element @xmath38 is relatively straightforward and the metric and ns - ns two - form of the corresponding deformed geometries are well - studied . however , the supercoset construction including the fermionic element @xmath39 is quite complicated and until very recently , the r - r sector and dilaton have not been studied . eventually , the supercoset construction has been carried out concretely for the @xmath9-deformed @xmath0 superstring @xcite . following this pioneering work , it has been generalized to the homogeneous cybe case @xcite . in the following , we will give an outline of the supercoset construction . in order to understand the geometries corresponding to the integrable deformations , we need to express the deformed action in eq.([ybsm ] ) in terms of the component fields of type iib supergravity . in other words , we need to put it in the canonical form @xmath55 \\ -\frac{\sqrt{\lambda_c}}{2}\ , i\bar{\theta}_i \qty(\gamma^{ab}\delta^{ij}-\epsilon^{ab}\sigma_3^{ij})\ , \tilde{e}_a^m \gamma_m \widetilde{d}^{jk}_{b}\theta_k + \order{\theta^4}\ , , \end{gathered}\ ] ] where the spinorial covariant derivative @xmath56 is defined as @xcite @xmath57 \tilde{e}^m_a \gamma_m\ , . \label{canonical}\end{gathered}\ ] ] here , @xmath58 is the vielbein for the deformed metric , and @xmath59 is a 32-component spinor composed of @xmath60 . although it is quite difficult to rewrite the action at all orders in @xmath61 , the second - order result is enough for our purposes . this expansion simplifies the analysis that remains nevertheless quite involved . for the technical details , see @xcite . one of the main problems in the study of integrable deformations is to determine if the corresponding ten - dimensional field content can be understood as a string theory background . in the case of the @xmath9-deformation , the ten - dimensional background can be obtained via the supercoset construction outlined above and one can see that the fields _ do not satisfy _ the type iib equations of motion @xcite . in fact , as was shown in @xcite , the worldsheet theory is _ not weyl invariant _ , but only scale invariant : the corresponding beta function does not vanish but is a total derivative . this fact weakens the usual on - shell construction of type iib supergravity and has lead to the definition of _ generalized _ type iib equations @xcite . note that as of now , this is a purely on - shell construction for which no ten - dimensional effective action has been found yet . interestingly , in recent work @xcite , it has been shown that the @xmath10-symmetry of the green - schwarz superstring generally leads to the generalized equations of type iib supergravity , while in the older literature @xcite , it was shown that the on - shell condition of the usual type iib supergravity leads to the kappa - invariance of the green - schwarz string action . the generalized type iib supergravity equations @xcite are @xmath62 here @xmath63 . the meaning of these equations is summarized as follows : * the first equation ( [ general1 ] ) is for the metric in the string frame @xmath64 . the matter contribution @xmath65 is given by @xmath66 here @xmath67 are the rescaled r - r field strengths @xmath68 where @xmath69 is the dilaton whose motion is described by ( [ general3 ] ) . * the second equation ( [ general2 ] ) is for the field strength @xmath70 of the ns - ns two - form . * the fourth , fifth and sixth equations ( [ general4 ] ) , ( [ general5 ] ) and ( [ general6 ] ) are for the r - r one - form , three - form and five - form field strengths . the bianchi identities for the r - r field strengths are also generalized as @xmath71 together with the standard type iib fields , the equations ( [ general1])-([general6 ] ) involve the three new vector fields @xmath72 , @xmath73 and @xmath74 . let us consider them in detail . in fact , only two of them are independent as the vector @xmath72 is expressed as @xmath75 @xmath73 and @xmath74 satisfy the following relations : @xmath76 the first equation of ( [ iz ] ) is the killing vector equation . assuming that @xmath77 is chosen such that the lie derivative vanishes , @xmath78 the second equation of ( [ iz ] ) can be solved by @xmath79 thus @xmath74 can be regarded as a generalization of the dilaton gradient @xmath80 . in particular , when @xmath73 vanishes , @xmath81 becomes @xmath80 and the generalized equations ( [ general1])-([general6 ] ) are reduced to the usual type iib supergravity equations . as mentioned at the beginning of this section , the generalized equations ( [ general1])-([general6 ] ) were found in the study of the @xmath9-deformed @xmath0 superstring . however , they also appear for yang baxter deformations with _ non - abelian _ classical @xmath1-matrices satisfying the homogeneous cybe - deformation is an example of yang baxter deformations with a non - abelian classical @xmath1-matrix ( the one of drinfeld jimbo type ) satisfying the mcybe . ] , as we will show in the next section . in this section , we will consider six examples of non - abelian classical @xmath1-matrices satisfying the homogeneous cybe . the supercoset construction is performed explicitly by following @xcite and it is shown that all of the resulting backgrounds satisfy the generalized type iib supergravity equations proposed in @xcite . for four of the examples , we derive `` t - dualized '' backgrounds which are solutions of the usual type iib supergravity . then for three of them , we show that the backgrounds are t - dual to the undeformed @xmath0 after an appropriate coordinate transformation . we will concentrate on deformations of the ads@xmath11 part and basically follow the notation and conventions for the @xmath44 generators in @xcite to represent classical @xmath1-matrices . as a first example , let us consider the following non - abelian classical @xmath1-matrix : @xmath83 here @xmath84 , @xmath85 are the generators of translations in the four - dimensional poincar algebra . the generator @xmath43 represents the dilatation . this is a solution of the homogeneous cybe which was already used to study a yang baxter deformation of four - dimensional minkowski spacetime @xcite . by performing the supercoset construction @xcite , the associated background is found to be @xmath86+\eta^2({{\rm d}}{t}-t z^{-1}{{\rm d}}{z})^2}{z^4+\eta^2(z^2+t^2 ) } + \frac{t^2(-{{\rm d}}{\phi}^2+\cosh^2\phi { { \rm d}}{\theta}^2)}{z^2 } + { { \rm d}}{s_{{\rm s}^5}^2}\ , , \\ b_2 & = \eta\,\frac{t { { \rm d}}{t } \wedge { { \rm d}}{x^1 } + z { { \rm d}}{z}\wedge { { \rm d}}{x^1}}{z^4+\eta^2(t^2+z^2)}\ , , \\ f_3 & = -\frac{4\eta\ , t^2\cosh\phi}{z^4}\left[{{\rm d}}{t } \wedge { { \rm d}}{\theta } \wedge { { \rm d}}{\phi } -\frac{t}{z}{{\rm d}}{\theta}\wedge { { \rm d}}{\phi } \wedge { { \rm d}}{z } \right]\ , , \\ f_5 & = 4\left[\frac{z^4}{z^4+\eta^2(t^2+z^2)}\omega_{{\ensuremath{\text{ads}_{5}}}}+\omega_{{\rm s}^5}\right]\,,\\ \phi & = \frac{1}{2}\log \left[\frac{z^4}{z^4+\eta^2(t^2+z^2)}\right]\ , , \end{aligned } \label{space}\ ] ] where we have rewritten the four - dimensional cartesian coordinates as : @xmath87 note here that the @xmath88 direction has the time - like signature . these fields _ do not _ satisfy the equations of motion of type iib supergravity , but solve the generalized equations of section [ sec : generalized - type - iib ] when supplemented with the following vectors : @xmath89 let us now perform t - dualities for the deformed background ( [ space ] ) . following @xcite , the extra fields are traded for a linear term in the dual dilaton . t - dualising along the @xmath90 and @xmath91 directions , we find : @xmath92 \\ & + { { \rm d}}{r}^2 + \sin^2 r { { \rm d}}{\xi}^2+\cos^2\xi \sin^2 r\ , { { \rm d}}{\phi_1}^2 + \sin^2r\sin^2\xi { { \rm d}}{\phi_2}^2 + \frac { { { \rm d}}{\phi_3}^2}{\cos^2r}\,,\\ \mathcal{f}_5 = { } & \frac{4t^2\cosh\phi } { z^4\cos r } ( { { \rm d}}{t}-\eta t { { \rm d}}{x^1})\wedge ( { { \rm d}}{z}-\eta z { { \rm d}}{x^1 } ) \wedge { { \rm d}}{\theta}\wedge { { \rm d}}{\phi}\wedge { { \rm d}}{\phi_3 } \\ & + 2z\sin^3 r\sin2\xi { { \rm d}}{x^1}\wedge { { \rm d}}{r}\wedge { { \rm d}}{\xi } \wedge { { \rm d}}{\phi_1}\wedge { { \rm d}}{\phi_2 } \ , , \\ \phi = { } & -\eta x^1+\log\left[\frac{z}{\cos r}\right]\ , . \end{aligned}\ ] ] remarkably , this is a solution of the _ usual _ type iib supergravity equations rather than the generalized ones . note , however , that the dilaton has acquired a linear dependence on @xmath90 . this means that @xmath93 is not an isometry and , as it is , the background can not be dualized in the directions @xmath94 to go back to the initial frame . this result is very similar to the fact that the hoare tseytlin solution @xcite is `` t - dual '' to the @xmath9-deformed background . indeed , we have followed the same strategy as in @xcite . the `` t - dualized '' background in eq . is a solution to the standard type iib equations and has a remarkable property : it is _ locally equivalent _ to undeformed @xmath0 . let us first perform the following change of coordinates : @xmath95 note that the new coordinate system does _ not _ cover all of spacetime : the new coordinate @xmath96 has to be restricted to be positive ( negative ) when @xmath97 ( @xmath98 ) . the signature of @xmath9 is fixed when we have chosen the deformation . this change of coordinates achieves the following points : * it diagonalizes the metric ; * it absorbs the @xmath90-dependence of the dilaton into the @xmath99 variable , such that @xmath93 is now a symmetry of the full background . explicitly , we find @xmath100 \\ & + { { \rm d}}{r}^2 + \sin^2 r { { \rm d}}{\xi}^2+\cos^2\xi \sin^2 r\ , { { \rm d}}{\phi_1}^2 + \sin^2r\sin^2\xi { { \rm d}}{\phi_2}^2 + \frac { { { \rm d}}{\phi_3}^2}{\cos^2r}\,,\\ \mathcal{f}_5 = { } & \frac{4\rho^2\cosh\phi } { \tilde{z}^4\cos r } { { \rm d}}{\rho}\wedge { { \rm d}}{\tilde{z } } \wedge { { \rm d}}{\theta}\wedge { { \rm d}}{\phi}\wedge { { \rm d}}{\phi_3 } \\ & + 2\tilde{z}\sin^3 r\sin2\xi { { \rm d}}{\tilde{x}^1}\wedge { { \rm d}}{r}\wedge { { \rm d}}{\xi } \wedge { { \rm d}}{\phi_1}\wedge { { \rm d}}{\phi_2 } \\ \phi = { } & \log\left[\frac{\tilde{z}}{\cos r}\right]\ , . \end{aligned}\ ] ] now we can perform again the two standard t - dualities along @xmath96 and @xmath91 to find , as advertised above , _ undeformed _ @xmath0 . let us stop for a moment to summarize what we have done . we have started with a yang baxter deformation of @xmath101 described by the non - abelian @xmath1-matrix ( eq . ) . using the supercoset construction outlined in section [ sec : yang - baxter - deformed ] , we have found the corresponding ten - dimensional background ( eq . ) that _ does not satisfy _ the usual type iib equations but is a solution to the generalized equations described in section [ sec : generalized - type - iib ] . then we have `` t - dualized '' this background using the rules of @xcite to find a new background ( eq . ) which solves the _ standard _ supergravity equations , but whose dilaton depends linearly on one of the t - dual variables . finally , we have observed that after a change of variables , this last background is nothing else than the t - dual of the standard undeformed @xmath0 . this result points to the fact that the yang baxter deformation with the classical @xmath1-matrix in eq . can be interpreted as an integrable twist , just like in the case of abelian classical @xmath1-matrices ( see for example @xcite ) . it might also suggest a relation between a gravity dual of the omega - background @xcite and a non - abelian classical @xmath1-matrix . our next example is the classical @xmath1-matrix @xmath103 where @xmath104 is the generator of time translations . this is a solution of the homogeneous cybe which was originally utilized to study a yang baxter deformation of 4d minkowski spacetime @xcite . by performing the supercoset construction @xcite , the associated background is found to be @xmath105-\eta^2({{\rm d}}{\rho}-\rho z^{-1}{{\rm d}}{z})^2}{z^4-\eta^2(z^2+\rho^2 ) } + \frac{\rho^2({{\rm d}}{\theta}^2+\sin^2\theta { { \rm d}}{\phi}^2)}{z^2 } + { { \rm d}}{s_{{\rm s}^5}^2}\ , , \\ b_2 = { } & \eta\,\frac{z { { \rm d}}{x^0 } \wedge { { \rm d}}{z } + \rho { { \rm d}}{x^0 } \wedge { { \rm d}}{\rho}}{z^4-\eta^2(z^2+\rho^2)}\ , , \\ f_3 = { } & \frac{4\eta}{z^4}\left[\rho^2\sin\theta { { \rm d}}{\rho}\wedge d \theta\wedge { { \rm d}}{\phi } -\frac{\rho^3\sin\theta}{z}{{\rm d}}{\theta}\,\wedge { { \rm d}}{\phi } \wedge { { \rm d}}{z } \right]\ , , \\ f_5 = { } & 4 \left[\frac{z^4}{z^4-\eta^2(z^2+\rho^2)}\,\omega_{{\ensuremath{\text{ads}_{5 } } } } + \omega_{{\rm s}^5}\right]\,,\\ \phi = { } & \frac{1}{2}\log \left[\frac{z^4}{z^4-\eta^2(z^2+\rho^2)}\right]\ , , \end{aligned}\ ] ] where the cartesian coordinates of the four - dimensional minkowski spacetime are @xmath106 in this case we see that with the following two vectors @xmath107 the above background becomes a solution of the generalized supergravity equations . our next task is to perform `` t - dualities '' for the background ( [ time ] ) in the @xmath108 and @xmath91 directions . we find a solution of the standard type iib supergravity equations with a dilaton that depends linearly on @xmath108 : @xmath109 \\ & + { { \rm d}}{r}^2+\sin^2 r { { \rm d}}{\xi}^2+\cos^2\xi \sin^2 r { { \rm d}}{\phi_1}^2 + \sin^2r\sin^2\xi { { \rm d}}{\phi_2}^2+\frac { { { \rm d}}{\phi_3}^2}{\cos^2r}\,,\\ \mathcal{f}_5 = { } & -\frac{4i\rho^2\sin\theta } { z^4\cos r } ( { { \rm d}}{\rho}+\eta\rho { { \rm d}}{x^0})\wedge({{\rm d}}{z}+\eta z { { \rm d}}{x^0 } ) \wedge { { \rm d}}{\theta}\wedge { { \rm d}}{\phi}\wedge { { \rm d}}{\phi_3 } \\ & + 2i z\sin^3 r\sin2\xi\ , { { \rm d}}{x^0}\wedge { { \rm d}}{r}\wedge { { \rm d}}{\xi } \wedge { { \rm d}}{\phi_1}\wedge { { \rm d}}{\phi_2}\ , , \\ \phi = { } & \eta\ , x^0+\log\left[\frac{z}{\cos r}\right]\ , . \end{aligned}\ ] ] finally , let us show that the `` t - dualized '' background ( [ t-4.2 ] ) is again equivalent to the undeformed @xmath0 . first of all , we perform the following coordinate transformations : @xmath110 note here that the new coordinate @xmath111 is restricted to be positive ( negative ) when @xmath112 ( @xmath113 ) . just like in the previous case , the metric is diagonal and the dilaton does not depend anymore on @xmath111 , such that @xmath114 is an isometry : @xmath115 \\ & + { { \rm d}}{r}^2+\sin^2 r { { \rm d}}{\xi}^2+\cos^2\xi \sin^2 r { { \rm d}}{\phi_1}^2 + \sin^2r\sin^2\xi { { \rm d}}{\phi_2}^2+\frac { { { \rm d}}{\phi_3}^2}{\cos^2r}\,,\\ \mathcal{f}_5 = { } & -\frac{4i\tilde{\rho}^2\sin\theta } { \tilde{z}^4\cos r}\ , { { \rm d}}{\tilde{\rho}}\wedge { { \rm d}}{\tilde{z } } \wedge { { \rm d}}{\theta}\wedge { { \rm d}}{\phi}\wedge { { \rm d}}{\phi_3 } \\ & + 2i \tilde{z}\sin^3 r\sin2\xi\ , { { \rm d}}{\tilde{x}^0}\wedge { { \rm d}}{r}\wedge { { \rm d}}{\xi } \wedge { { \rm d}}{\phi_1}\wedge { { \rm d}}{\phi_2}\ , , \\ \phi = { } & \log\left[\frac{\tilde{z}}{\cos r}\right]\ , . \end{aligned}\ ] ] again , by performing two t - dualities along the @xmath111 and @xmath91 directions , we go back to the undeformed @xmath0 background . let us consider now the classical @xmath1-matrix @xmath117 where @xmath118 is the generator of the lorentz rotation in the plane @xmath119 . performing the supercoset construction @xcite , we obtain the corresponding background : @xmath120({{\rm d}}{x^+})^2 + { { \rm d}}{s_{s^5}^2}\ , , \\ b_2 & = \eta\qty[\frac{\rho { { \rm d}}{x^+ } \wedge { { \rm d}}{\rho}}{z^4}+\frac{1}{z^3}{{\rm d}}{x^+}\wedge { { \rm d}}{z } ] \ , , \\ f_3 & = 4\eta \qty[\frac{\rho^2}{z^5 } { { \rm d}}{x^+ } \wedge { { \rm d}}{\theta } \wedge { { \rm d}}{z } + \frac{\rho}{z^4 } { { \rm d } } { x^+}\wedge { { \rm d}}{\rho } \wedge { { \rm d}}{\theta } ] \ , , \\ f_5 & = 4 \qty(\omega_{{\ensuremath{\text{ads}_{5}}}}+\omega_{{\rm s}^5 } ) \ , , \\ \phi & = \phi_0 \text { ( constant ) , } \end{aligned}\ ] ] where the cartesian coordinates of the four - dimensional minkowski spacetime @xmath121 are @xmath122 this background is a solution of the generalized supergravity equations - deformed ads@xmath13 . ] when supplemented by the vectors @xmath73 and @xmath74 : @xmath123 let us perform four `` t - dualities '' along the @xmath124 and @xmath125 directions , t - dualize in these and finally introduce light - like combinations for the t - dual variables . ] . the resulting background is given by @xmath126\ , , \end{aligned}\ ] ] where the other components are zero . the `` t - dualized '' background in eq . is a solution to the standard type iib equations and is again _ locally _ equivalent to undeformed @xmath0 . let us first change the coordinates as follows : @xmath127 explicitly , we find @xmath128\ , . \end{aligned}\ ] ] now , rewriting the light - like coordinates in terms of the cartesian coordinates as @xmath129 and performing four t - dualities along @xmath111 , @xmath130 , @xmath131 and @xmath125 , we reproduce the _ undeformed _ @xmath0 background . [ [ mixing - of - abelian - and - non - abelian - classical - r - matrices . ] ] mixing of abelian and non - abelian classical @xmath1-matrices . + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + this example admits a generalization , obtained by mixing abelian and non - abelian classical @xmath1-matrices : @xmath132\ , . \label{4.7-r}\ ] ] when @xmath133 , the classical @xmath1-matrix reduces to the one described above ; when @xmath134 , the @xmath1-matrix becomes abelian and the associated background is the hubeny rangamani ross solution of @xcite , as shown in @xcite . in @xcite it was shown that with a supercoset construction , one finds the following ten - dimensional background @xmath135({{\rm d}}{x^+})^2+{{\rm d}}{s}_{\rm s^5}^2\ , , \\ b_2 & = \frac{\eta}{z^4}{{\rm d}}{x^+}\wedge \qty [ a_1 \qty ( \rho { { \rm d}}{\rho } + z { { \rm d}}{z } ) - a_2 \rho^2 { { \rm d}}{\theta } ] \ , , \\ f_3 & = \frac{4\eta\rho}{z^5}{{\rm d}}{x^+}\wedge \qty [ a_1 \qty ( z { { \rm d}}{\rho } - \rho { { \rm d}}{z } ) \wedge { { \rm d}}{\theta } + a_2 { { \rm d}}{\rho}\wedge { { \rm d}}{z } ] \ , , \\ f_5 & = 4 \qty(\omega_{{\ensuremath{\text{ads}_{5 } } } } + \omega_{{\rm s}^5})\ , , \\ \phi & = \phi_0 \text { ( constant ) . } \end{aligned}\ ] ] this background is still a solution of the generalized equations with the two vectors @xmath73 and @xmath74 given by @xmath136 in the special case @xmath134 , the above background reduces to a solution of standard type iib supergravity . let us next take four `` t - dualities '' along the @xmath124 and @xmath125 directions . then we can obtain a solution of the usual type iib supergravity as @xmath137\ , , \end{aligned}\ ] ] where the other components are zero . it is easy to see that this is just a twist of the previous solution ( in eq . ) and in fact there is a change of variables @xmath138 that maps this background to the same local form : @xmath139\ , , \end{aligned}\ ] ] which is a t - dual of the undeformed @xmath140 background . let us now consider the non - abelian classical @xmath1-matrix given by @xmath142 which is another solution of the homogeneous cybe . using the supercoset construction @xcite , the associated background is found to be @xmath143+\frac{{{\rm d}}{\rho}^2+\rho^2{{\rm d}}{\theta}^2}{z^2 } + { { \rm d}}{s_{{\rm s}^5}^2}\ , , \\ b_2 = { } & \eta\,\frac{{{\rm d}}{x^+ } \wedge ( z{{\rm d}}{z}+\rho { { \rm d}}{\rho}-x^+{{\rm d}}{x^-})}{z^4-\eta^2(x^+)^2}\ , , \\ f_3 = { } & 4\eta\,\frac{\rho}{z^4}\left[\frac{\rho}{z}{{\rm d}}{x^+}\wedge { { \rm d}}{\theta}\wedge { { \rm d}}{z } + { { \rm d}}{x^+}\wedge { { \rm d}}{\rho}\wedge { { \rm d}}{\theta } -\frac { x^+}{z}{{\rm d}}{\rho}\wedge { { \rm d}}{\theta } \wedge { { \rm d}}{z}\right]\ , , \\ f_5 = { } & 4\left[\frac{z^4}{z^4-\eta^2(x^+)^2}\,\omega_{{\ensuremath{\text{ads}_{5 } } } } + \omega_{{\rm s}^5}\right]\ , , \\ \phi = { } & \frac{1}{2}\log \left[\frac{z^4}{z^4-\eta^2(x^+)^2}\right]\ , . \end{aligned}\ ] ] here the following new coordinates have been introduced : @xmath144 this background satisfies the generalized sugra equations with the two vectors @xmath73 and @xmath74 given by @xmath145 as of now , we have not found an appropriate t - dual frame in which this background is a solution to the standard type iib equations with a linear dilaton . it would be interesting to understand if this means that we need a different realization of a scale ( non - weyl ) invariant worldsheet theory , different from the one discussed in @xcite . let us next consider the following non - abelian classical @xmath1-matrix : @xmath146\,.\ ] ] this is a solution of the homogeneous cybe which was employed to study a light - like deformation of the four - dimensional poincar algebra @xcite and was also utilized to study a yang baxter deformation of four - dimensional minkowski spacetime @xcite . by performing the supercoset construction @xcite , the associated background is determined to be @xmath147+\frac{{{\rm d}}{\rho}^2+\rho^2{{\rm d}}{\theta}^2}{z^2 } + { { \rm d}}{s_{{\rm s}^5}^2}\ , , \\ b_2 = { } & \eta\,\frac{{{\rm d}}{x^+ } \wedge ( \rho { { \rm d}}{\rho}-x^+{{\rm d}}{x^-})}{z^4-\eta^2(x^+)^2}\ , , \\ f_3 = { } & 4\eta\,\frac{\rho}{z^5}\left(\rho { { \rm d}}{x^+}\wedge { { \rm d}}{\theta}\wedge { { \rm d}}{z } - x^+{{\rm d}}{\rho}\wedge { { \rm d}}{\theta } \wedge { { \rm d}}{z}\right)\ , , \\ f_5 = { } & 4\left[\frac{z^4}{z^4-\eta^2(x^+)^2}\omega_{{\ensuremath{\text{ads}_{5}}}}+\omega_{{\rm s}^5}\right]\,,\\ \phi = { } & \frac{1}{2}\log \left[\frac{z^4}{z^4-\eta^2(x^+)^2}\right]\ , . \end{aligned}\ ] ] here the coordinate system ( [ lccoord ] ) has been used . this background does not satisfy the equations of the usual type iib supergravity , but the generalized equations with the two vectors @xmath73 and @xmath74 given by @xmath148 this example is quite close to the one discussed in the previous subsection . also in this case , a t - duality to a solution of the standard type iib equations is as of now missing . our last example is the classical @xmath1-matrix @xmath149 it was originally studied in @xcite in relation to a scaling limit of the classical @xmath1-matrix of drinfeld - jimbo type . by performing the supercoset construction include @xmath150 . for details , see @xcite . ] , the full background is determined to be . ] @xmath151}{z^4 + 4\eta^2\rho^2 } + \frac{\rho^2{{\rm d}}{\theta}^2}{z^2}+ { { \rm d}}{s}^2_{{\rm s}^5}\ , , \\ b_2 & = \frac{2\eta } { z(z^2 - 4\eta^2 ) } { { \rm d}}{x^0}\wedge { { \rm d}}{z } + \frac{2\eta\ , \rho}{z^4 + 4\eta^2 \rho^2 } { { \rm d}}{x^1}\wedge { { \rm d}}{\rho}\ , , \\ f_{1 } & = \frac{16\eta^2\rho^2}{z^4}{{\rm d}}{\theta}\ , , \\ f_{3 } & = \frac{8\eta \rho^2}{z^3(z^2 - 4\eta^2 ) } { { \rm d}}{x^0}\wedge { { \rm d}}{\theta } \wedge { { \rm d}}{z } + \frac{8\eta \rho}{z^4 + 4\eta^2\rho^2 } { { \rm d}}{x^1}\wedge { { \rm d}}{\rho } \wedge { { \rm d}}{\theta}\ , , \\ f_{5 } & = 4\left[\frac{z^6}{(z^2 - 4\eta^2)(z^4 + 4\eta^2\rho^2)}\,\omega_{{\ensuremath{\text{ads}_{5 } } } } + \omega_{{\rm s}^5}\right ] \ , , \\ \phi & = \frac{1}{2}\log\left[\frac{z^6}{(z^2 - 4\eta^2)(z^4 + 4\eta^2\rho^2)}\right]\ , . \end{aligned}\ ] ] again , this background does not satisfy the usual type iib supergravity equations , but becomes a solution of the generalized equations by taking @xmath73 and @xmath74 as @xmath152{{\rm d}}{z } + \frac{4\eta^2\rho { { \rm d}}{\rho}}{z^4 + 4\eta^2\rho^2}\ , . \label{i-6}\end{aligned}\ ] ] performing `` t - dualities '' for all of the @xmath153 directions , we can obtain a solution of the standard type iib supergravity : @xmath154\ , , \end{aligned } \label{t - hvt}\ ] ] where the other components are zero . just like in the first examples that we have discussed , there is a simple change of coordinates @xmath155 that diagonalizes the metric : @xmath156 + \frac { { \rm e}^{4\eta(x^0-x^1)}{{\rm d}}{\tilde{\rho}}^2+{{\rm d}}{\tilde{z}}^2}{\tilde{z}^2 } + { \rm e}^{-4\eta(x^0-x^1)}\frac{\tilde{z}^2{{\rm d}}{\theta}^2}{\tilde{\rho}^2 } \\ & + { { \rm d}}{r}^2+\sin^2 r { { \rm d}}{\xi}^2 + \frac{{{\rm d}}{\phi_1}^2 } { \sin^2 r\cos^2\xi } + \frac{{{\rm d}}{\phi_2}^2}{\sin^2r\sin^2\xi } + \frac { { { \rm d}}{\phi_3}^2}{\cos^2r}\,,\\ \mathcal{f}_5 = { } & \frac{-4i { \rm e}^{2\eta ( x^0-x^1)}}{\tilde{z}^2\sin^2r\cos r\sin\xi\cos\xi}\ , { { \rm d}}{\tilde{\rho } } \wedge { { \rm d}}{\tilde{z } } \wedge { { \rm d}}{\phi_1 } \wedge { { \rm d}}{\phi_2}\wedge { { \rm d}}{\phi_3 } \\ & + 4i \frac { { \rm e}^{-2\eta\,(3x^0-x^1)}\tilde{z}^3}{\tilde{\rho}}\sin r\ , { { \rm d}}{x^0}\wedge { { \rm d}}{x^1}\wedge { { \rm d}}{\theta } \wedge { { \rm d}}{r } \wedge { { \rm d}}{\xi}\ , , \\ \phi = { } & 2\eta\ , ( x^0 - x^1)+\log\left[\frac{\tilde{z}^3}{\tilde{\rho } \sin^2 r \cos r \sin\xi \cos\xi}\right]\ , . \end{aligned}\ ] ] in this case , however , the linear dependence of the dilaton on the t - dual variables remains . it may be helpful to recall the result for the case of 3d schrdinger spacetime @xcite . an affine symmetry algebra is given by a twisted yangian , which is called an exotic symmetry in @xcite , and can be mapped to the standard yangian by undoing the integrable twist . then , however , the target spacetime is not mapped to the undeformed ads@xmath157 . the resulting geometry is described by a dipole - like coordinate system and hence it is very close to , but not identical to ads@xmath157 . it would be interesting to study also the case without the total derivative term of @xmath158 in ( [ hvt ] ) . then the `` t - dualized '' background is different from the one of eq . ( [ t - hvt ] ) and the resulting background is given by @xmath159\ , . \label{t - hvtno}\end{aligned}\ ] ] this background is a solution of the usual type iib supergravity , and agrees with the one obtained in @xcite after fixing some typos . now it is natural to ask whether ( [ t - hvt ] ) and ( [ t - hvtno ] ) are equivalent or not , and if so , whether this equivalence holds locally or globally . the local equivalence can be shown explicitly by using the coordinate transformations @xmath160\qquad z>2\eta\,,{\nonumber}\\ x^0&\to&-x^0+\frac{1}{4\eta}\log\left[\frac{4\eta^2-z^2}{z^2}\right]\qquad z<2\eta\,.\end{aligned}\ ] ] for the case of the global equivalence , some subtleties arise . the background ( [ t - hvt ] ) is regular while the one ( [ t - hvtno ] ) has a coordinate singularity at @xmath161 and so are the coordinate transformations . moreover , two types of time directions have to be introduced . due to these observations , a more involved analysis is necessary in order to argue the global equivalence . in this paper we have studied the relation between non - abelian classical @xmath1-matrices and the generalized type iib equations . we have discussed several examples of non - abelian classical @xmath1-matrices and derived the corresponding ten - dimensional backgrounds by performing a supercoset construction . our main result is that these backgrounds do not satisfy the standard type iib equations of motion but a set of generalized equations , corresponding to a scale - invariant as opposed to weyl - invariant worldsheet theory . this is consistent with the scaling result of @xcite , and the general result of @xcite under the assumptions that the deformed string action is the canonical green - schwarz string . for some of these examples , we were able to perform a t - duality transformation leading to a new solution of the standard iib equations , where the dilaton depends linearly on some of the t - dual variables . for some of these examples , we have even been able to find a change of variables showing that the linear - dilaton backgrounds are locally t - dual to undeformed @xmath162 . this fact makes the classical integrability of these examples manifest . we have however not been able to find appropriate t - dual frames for all of the examples . at this point , it is unclear whether this is only due to technical issues or whether there is an underlying physical reason for this . it would therefore be of use to explore more general classical @xmath1-matrices , as well as lower - dimensional cases , such as deformations of @xmath163 or @xmath164 @xcite , and the relationship between our construction and the @xmath165 deformation of @xcite . we find that a criterion for distinguishing classes of non - abelian classical @xmath1-matrices is of importance . a possible criterion is whether a given non - abelian classical @xmath1-matrix is constructed as a twist of the drinfeld - jimbo @xmath1-matrix or not ( for details , see sec . 3.1 of @xcite ) . this direction would be useful for the classification of non - abelian classical @xmath1-matrices . gaining a deeper understanding of the underlying algebraic structure giving rise to our findings is essential , as it might throw light on the fundamental mathematical underpinnings of the gravity / cybe correspondence . we are very grateful to gleb arutyunov and hideki kyono for useful discussions . we are indebted to ben hoare and stijn van tongeren for clarifying some points and misinterpretations . would like to acknowledge useful discussions which have taken place during the workshop `` generalized geometry & t - dualities '' at the simons center for geometry and physics , especially valuable comments by a. a. tseytlin and c. klimcik . d.o . and s.r . would like to thank the department of physics of kyoto university for hospitality where this work was initiated . the work of s.r . was supported by the swiss national science foundation ( snf ) under grant number pp00p2_157571/1 . the work of j.s . was supported by the japan society for the promotion of science ( jsps ) . the work of k.y . was supported by the supporting program for interaction - based initiative team studies ( spirits ) from kyoto university and by a jsps grant - in - aid for scientific research ( c ) no.15k05051 . this work was also supported in part by the jsps japan - russia research cooperative program and the jsps japan - hungary research cooperative program . c. klimcik , `` yang baxter sigma models and ds / ads t duality , '' jhep * 0212 * ( 2002 ) 051 [ hep - th/0210095 ] ; `` on integrability of the yang baxter sigma - model , '' j. math . * 50 * ( 2009 ) 043508 [ arxiv:0802.3518 [ hep - th ] ] ; `` integrability of the bi - yang baxter sigma model , '' lett . math . * 104 * ( 2014 ) 1095 [ arxiv:1402.2105 [ math - ph ] ] . i. kawaguchi and k. yoshida , `` hybrid classical integrability in squashed sigma models , '' phys . b * 705 * ( 2011 ) 251 [ arxiv:1107.3662 [ hep - th ] ] ; `` hidden yangian symmetry in sigma model on squashed sphere , '' jhep * 1011 * ( 2010 ) 032 [ arxiv:1008.0776 [ hep - th ] ] . + i. kawaguchi , t. matsumoto and k. yoshida , `` the classical origin of quantum affine algebra in squashed sigma models , '' jhep * 1204 * ( 2012 ) 115 [ arxiv:1201.3058 [ hep - th ] ] ; `` on the classical equivalence of monodromy matrices in squashed sigma model , '' jhep * 1206 * ( 2012 ) 082 [ arxiv:1203.3400 [ hep - th ] ] . i. kawaguchi , d. orlando and k. yoshida , `` yangian symmetry in deformed wznw models on squashed spheres , '' phys . b * 701 * ( 2011 ) 475 [ arxiv:1104.0738 [ hep - th ] ] ; i. kawaguchi and k. yoshida , `` a deformation of quantum affine algebra in squashed wznw models , '' j. math . phys . * 55 * ( 2014 ) 062302 [ arxiv:1311.4696 [ hep - th ] ] . i. kawaguchi and k. yoshida , `` classical integrability of schrdinger sigma models and @xmath7-deformed poincare symmetry , '' jhep * 1111 * ( 2011 ) 094 [ arxiv:1109.0872 [ hep - th ] ] ; `` exotic symmetry and monodromy equivalence in schrdinger sigma models , '' jhep * 1302 * ( 2013 ) 024 [ arxiv:1209.4147 [ hep - th ] ] . + i. kawaguchi , t. matsumoto and k. yoshida , `` schrdinger sigma models and jordanian twists , '' jhep * 1308 * ( 2013 ) 013 [ arxiv:1305.6556 [ hep - th ] ] . f. delduc , m. magro and b. vicedo , `` an integrable deformation of the ads@xmath166s@xmath167 superstring action , '' phys . * 112 * ( 2014 ) 051601 [ arxiv:1309.5850 [ hep - th ] ] ; `` derivation of the action and symmetries of the @xmath7-deformed ads@xmath166s@xmath167 superstring , '' jhep * 1410 * ( 2014 ) 132 [ arxiv:1406.6286 [ hep - th ] ] . v. g. drinfeld , `` hopf algebras and the quantum yang baxter equation , '' sov . * 32 * ( 1985 ) 254;``quantum groups , '' j. sov . * 41 * ( 1988 ) 898 [ zap . semin . * 155 * , 18 ( 1986 ) ] . m. jimbo , `` a @xmath7 difference analog of @xmath169 and the yang baxter equation , '' lett . * 10 * ( 1985 ) 63 . g. arutyunov , s. frolov , b. hoare , r. roiban and a. a. tseytlin , `` scale invariance of the eta - deformed ads@xmath166s@xmath167 superstring , t - duality and modified type ii equations , '' nucl . b * 903 * ( 2016 ) 262 [ arxiv:1511.05795 [ hep - th ] ] . c. p. herzog , m. rangamani and s. f. ross , `` heating up galilean holography , '' jhep * 0811 * ( 2008 ) 080 [ arxiv:0807.1099 [ hep - th ] ] ; j. maldacena , d. martelli and y. tachikawa , `` comments on string theory backgrounds with non - relativistic conformal symmetry , '' jhep * 0810 * ( 2008 ) 072 [ arxiv:0807.1100 [ hep - th ] ] ; a. adams , k. balasubramanian and j. mcgreevy , `` hot spacetimes for cold atoms , '' jhep * 0811 * ( 2008 ) 059 [ arxiv:0807.1111 [ hep - th ] ] . d. israel , c. kounnas , d. orlando and p. m. petropoulos , `` electric / magnetic deformations of s@xmath176 and ads@xmath157 , and geometric cosets , '' fortsch . phys . * 53 * , 73 ( 2005 ) [ hep - th/0405213 ] ; s. detournay , d. orlando , p. m. petropoulos and p. spindel , `` three - dimensional black holes from deformed anti - de sitter , '' jhep * 0507 * , 072 ( 2005 ) [ hep - th/0504231 ] ; d. t. son , `` toward an ads / cold atoms correspondence : a geometric realization of the schrdinger symmetry , '' phys . d * 78 * ( 2008 ) 046003 [ arxiv:0804.3972 [ hep - th ] ] ; k. balasubramanian and j. mcgreevy , `` gravity duals for non - relativistic cfts , '' phys . rev . * 101 * ( 2008 ) 061601 [ arxiv:0804.4053 [ hep - th ] ] . m. crichigno , t. matsumoto and k. yoshida , `` deformations of @xmath178 as yang baxter sigma models , '' jhep * 1412 * ( 2014 ) 085 [ arxiv:1406.2249 [ hep - th ] ] ; `` towards the gravity / cybe correspondence beyond integrability yang baxter deformations of @xmath178 , '' j. phys . * 670 * ( 2016 ) no.1 , 012019 [ arxiv:1510.00835 [ hep - th ] ] . a. borowiec , h. kyono , j. lukierski , j. sakamoto and k. yoshida , `` yang baxter sigma models and lax pairs arising from @xmath10-poincar @xmath1-matrices , '' jhep * 1604 * ( 2016 ) 079 [ arxiv:1510.03083 [ hep - th ] ] ; h. kyono , j. sakamoto and k. yoshida , `` lax pairs for deformed minkowski spacetimes , '' jhep * 1601 * ( 2016 ) 143 [ arxiv:1512.00208 [ hep - th ] ] . t. matsumoto and k. yoshida , `` integrable deformations of the ads@xmath173s@xmath167 superstring and the classical yang _ -towards the gravity / cybe correspondence- _ , '' j. phys . * 563 * ( 2014 ) 1 , 012020 [ arxiv:1410.0575 [ hep - th ] ] ; `` towards the gravity / cybe correspondence current status , '' j. phys . * 670 * ( 2016 ) no.1 , 012033 . a. mikhailov , `` pure spinors in ads and lie algebra cohomology , '' lett . math . * 104 * ( 2014 ) 1201 [ arxiv:1207.2441 [ hep - th ] ] ; o. a. bedoya , l. i. bevilaqua , a. mikhailov and v. o. rivelles , `` notes on beta - deformations of the pure spinor superstring in ads@xmath166s@xmath167 , '' nucl . b * 848 * ( 2011 ) 155 [ arxiv:1005.0049 [ hep - th ] ] . m. t. grisaru , p. howe , l. mezincescu , b. e. nilsson and p. k. townsend , `` @xmath179 superstrings in a supergravity background , '' phys . b * 162 * ( 1985 ) 116 . + e. bergshoeff , e. sezgin and p. k. townsend , `` superstring action in @xmath180 curved superspace , '' phys . b * 169 * ( 1986 ) 191 . s. hellerman , d. orlando and s. reffert , `` the omega deformation from string and m - theory , '' jhep * 1207 * , 061 ( 2012 ) [ arxiv:1204.4192 [ hep - th ] ] . s. hellerman , d. orlando and s. reffert , `` string theory of the omega deformation , '' jhep * 1201 * , 148 ( 2012 ) doi:10.1007/jhep01(2012)148 [ arxiv:1106.0279 [ hep - th ] ] .

universal features in the scalings of shannon - rnyi entropies of many - body groundstates are studied for interacting spin- systems across ( 2 + 1 ) dimensional critical points , using quantum monte carlo simulations on dimerized and plaquettized heisenberg models on the square lattice . considering both full systems and line shaped subsystems , symmetry breaking on the nel ordered side of the transition is characterized by the presence of a logarithmic term in the scaling of shannon - rnyi entropies , which is absent in the disordered gapped phase . such a difference in the scalings allows to capture the quantum critical point using shannon - rnyi entropies for line shaped subsystems of length embedded in tori , as the smaller subsystem entropies are numerically accessible to much higher precision than for the full system . most interestingly , at the quantum phase transition an additive subleading constant emerges in the critical scaling of the line shannon - rnyi entropy . this number appears to be universal for 3d criticality , as confirmed for the finite - temperature transition in the 3d antiferromagnetic spin- heisenberg model . additionally , the phases and phase transition can be detected in several features of the participation spectrum , consisting of the diagonal elements of the reduced density matrix of the line subsystem . in particular the nel ordering transition can be simply understood in the basis by a confinement mechanism of ferromagnetic domain walls .

1402.4813

how do coefficients of a wave - function change at continuous quantum phase transitions ? in a given basis , this question can be addressed by monitoring the behavior of ( inverse ) participation ratios , which have a long history _ e.g. _ in localization physics @xcite . more recently , the study of related quantities such as shannon - rnyi entropies ( which quantify the localization of the wave - function in a given basis ) in _ many - body _ problems has revealed an intriguing aspect : subleading terms in the finite - size scaling of these quantities appear to carry universal information , characteristic of the physics contained in the ground - state wave - function@xcite . for instance , they can characterize the presence of broken continuous or discrete symmetry breaking in the ground - state , as well as information on the universality class of continuous phase transitions . most previous studies@xcite on this topic focused on one - dimensional quantum systems , where both analytical and numerical studies are easiest . in particular , dealing with the exponentially growing size of the hilbert space of many - body problems , while maintaining a large enough total system size to study finite - size dependence , is a hurdle to surmount . recently , we have introduced in ref . convenient numerical methods to study the shannon - rnyi entropies of many - body systems through a quantum monte carlo ( qmc ) sampling of the ground - state wave function . this method allows studies of much larger systems than previously accessible in numerical calculations , which is necessary for the analysis of universal behavior at continuous quantum phase transitions where physical correlation lengths diverge . here , we will study this problem for a non - trivial , yet well - understood quantum phase transition in two - dimensional quantum magnetism : the transition between a nel antiferromagnet and a quantum paramagnet in two @xmath8 quantum spin heisenberg models with varying antiferromagnetic couplings , namely two - dimensional coupled dimers and plaquettes ( see fig . [ fig : lattices ] ) . the variation of the ratio of two exchange couplings @xmath9 allows to couple isolated paramagnetic units ( at @xmath10 ) to form a two - dimensional antiferromagnet ( at @xmath11 ) which spontaneously breaks @xmath2 symmetry at zero temperature . a quantum critical point at @xmath12 in the 3d @xmath1 universality class@xcite separates the quantum disordered and nel ordered phases . the first part of the paper ( sec . [ sec : sr ] ) is devoted to the study of the behavior of subleading terms in the sr entropies of the ground - state of heisenberg magnets . given a density matrix @xmath13 , the sr entropies are defined as : @xmath14 where @xmath15 are states of the computational basis in which sr entropies are calculated . note that the choice of the natural logarithm ( base @xmath16 ) fixes the units of sr entropies to `` nats '' . we will first consider in sec . [ sec : srfull ] the sr entropy of the full system composed of @xmath17 interacting @xmath18 spins on a square lattice , that is choosing @xmath19 in eq . to be the full density matrix @xmath20 of the ground state @xmath21 . sr entropies are generally found to have a leading behavior which is extensive@xcite @xmath22 where the pre - factor @xmath23 ( for spin @xmath0 systems ) depends on details of the model ( @xmath24 in that case ) . in the @xmath7 basis considered throughout this work , we naturally expect @xmath25 to be ` small ' in the nel phase , and ` large ' in the quantum disordered phase . this is easily understood by considering the limit @xmath26 , where @xmath27 . here , latexmath:[$\max(\rho_{ii})=\max_{i } maximal coefficient of the groundstate wave - function expanded in the @xmath7 basis . for antiferromagnetic systems , this is the coefficient of the nel state @xmath29 , which is expected to be much larger in the antiferromagnetically ordered phase than in the disordered phase . in the antiferromagnetic phase , the groundstate spontaneously breaks the continuous @xmath2 symmetry and our previous work @xcite showed that this is reflected in a subleading logarithmic correction : @xmath30 . in the disordered phase , no symmetry is broken and the subleading term is in general a universal constant ( expected to be zero in the paramagnetic phase discussed in this paper ) . for this first part of the paper ( sec . [ sec : sr ] ) , we will consider the case @xmath26 essentially for practical purposes . indeed , @xmath31 is simpler to obtain numerically ( within our qmc simulations ) and more importantly , the leading term prefactor @xmath32 is smaller ( @xmath33 for all finite @xmath34 ) which ensures that we can reach larger system sizes . despite these facts , simulations of the sr entropy of full two dimensional systems are limited to relatively small sizes ( up to @xmath35 ) close to the quantum phase transition , as the prefactor @xmath32 is still quite large in this region ( see also the discussion in appendix [ sec : mc ] ) . to circumvent this , we next consider in sec . [ sec : srline ] the scaling of the sr entropy of a subsystem , composed of a single line of size @xmath3 ( the geometry of the subsystem is defined in fig . [ fig : lattices ] ) embedded in a periodic @xmath4 torus . subsystem sr entropies are defined in analogy to eq . ( [ eq : sr - entropies ] ) , except that we now consider the _ reduced _ density matrix @xmath36 of a subsystem @xmath37 which is obtained by performing a partial trace over the rest of the system @xmath38 : @xmath39 with @xmath40 , _ i.e. _ where the basis state @xmath41 is a tensor product state of subsystems basis states @xmath42 and @xmath43 . the rationale for choosing a line shaped subsystem is two - fold : first , we physically expect that the sr entropy of the line also contains the information about antiferromagnetic ordering ( since for instance the correlation function @xmath44 along the line is defined in terms of diagonal elements of the reduced density matrix ) . second , the sr entropy of the line has a leading term @xmath45 , and therefore takes much smaller values than for the full - system sr entropy for the same value of @xmath3 ( @xmath3 scaling versus @xmath46 scaling ) . this leads to a better accuracy and allows to reach much larger linear sizes @xmath3 in our qmc simulations . our results indicate that the sr entropy of the line also shows a sub - leading @xmath47 term in the nel phase , and a constant term in the disordered phase which turns out to vanish ( see below ) . quite interestingly , the subleading term right at the quantum phase transition is a constant @xmath48 , which appears to be identical ( within error bars ) for the two models studied . this suggest that this constant is characteristic of the 3d @xmath1 universality class to which both quantum phase transitions belong . further simulations of the finite - temperature ordering phase transition of the simple cubic @xmath8 heisenberg model ( also in the same @xmath1 universality class ) support this conjecture for antiferromagnetic interactions . we also expect a universal value @xmath49 for the full system , even though the limited accuracy ( due to the large value of @xmath32 ) of our simulations does not allow to prove this . for the verification of the universality of @xmath50 , we have first performed extensive calculations of the spin stiffness in order to extract the best estimate for the value of the transition temperature , @xmath51 , in agreement with ref . . the sr entropies are global averages of all coefficients of the wave - function and their scaling with the ( sub- ) system size thus capture correctly phases and phase transitions . it is interesting to ask whether _ each individual _ coefficient ( or reduced density matrix diagonal element in the case of subsystems ) also `` sees '' the quantum phase transition when @xmath52 is varied this independently of their scaling with system size , as exemplified with the maximal diagonal entry of the ( reduced ) density matrix , governing @xmath31 ( @xmath53 ) . motivated by this question , we study in the second part of the paper ( sec . [ sec : ps ] ) the behavior of each diagonal element of the reduced density matrix @xmath54 for a line subsystem across the transition . in analogy with the entanglement spectrum @xcite , we define the `` participation spectrum '' as the set of pseudo - energies @xmath55 . the participation spectrum develops into well - defined bands , which can be classified according to the magnetization and the number of ferromagnetic domain - walls separating segments having different nel line configurations ( @xmath56 or @xmath57 . identifying the lowest - lying states in this spectrum allows to understand the quantum phase transition in terms of an effective repulsion between such domain - walls . even though the participation spectra appear to differ at first glance in the quantum disordered phases of the two studied models , we find that this can be understood easily by classifying states according to the number of strong or weak domain walls ( this notion is dictated by the local physics of one of the two models considered ) . a striking outcome of this analysis is that _ all _ individual levels ( even corresponding to assumedly irrelevant states such as the fully polarized state ) harbor signs of the quantum phase transition , as exemplified for instance by an inflection point ( with respect to @xmath52 ) for almost all @xmath58 . we analyze this in detail for the most probable state . another interesting sign of the quantum phase transition is revealed by the study of the finite - size behavior of the width of the lowest - lying bands . we will finally conclude in sec . [ sec : conclusion ] on the implications of our results while the appendices contain specific details of the qmc procedure used ( appendix [ sec : mc ] ) , as well as exact results in the limit of @xmath59 for reference ( appendix [ sec : exact ] ) . let us begin our paper ( sec . [ sec : model ] ) by providing useful details on the models studied as well as on the finite - size scaling analysis . and we refer to them as plaquettes / dimers . the dotted bonds are the weak interplaquette / interdimer couplings @xmath60 . periodic boundary conditions are implicit . ] the two models ( dimerized and plaquettized heisenberg models ) that we study are defined with the same hamiltonian form : @xmath61 with @xmath62 and where the two terms correspond to the summation over stronger bonds for columnar dimers ( plaquettes ) and to the summation over the weaker links between these entities ( see fig . [ fig : lattices ] ) . we only consider @xmath63 here , with @xmath11 yielding the homogeneous heisenberg antiferromagnet on the square lattice . the two models have slightly different critical points at @xmath64@xcite for the columnar dimerized system and @xmath65@xcite for the plaquettized system . for @xmath66 both models display a disordered ground state separated from excited states by a finite energy gap , whereas for @xmath67 antiferromagnetic nel long - range order occurs , with a spontaneous breaking of the @xmath2 symmetry . we considered these two models as they are well - established to harbor the same physical content ( in particular the quantum phase transitions at @xmath12 belong to the same 3d @xmath1 universality class ) , yet with different microscopics : this will allow us to discuss universality of the scaling of sr entropies . we study properties of the groundstates expanded in the @xmath7 basis and note that all results will be identical in any basis obtained by a global @xmath2 transformation , by symmetry of the hamiltonian . we use the index @xmath68 ( respectively @xmath69 ) to denote quantities for the plaquettized ( resp . dimerized ) model . considering the results of ref . , we will perform fits of the sr entropy @xmath31 of the full system to the following forms : @xmath70 and @xmath71 equivalent forms for the line sr entropy @xmath72 are : @xmath73 and @xmath74 note that in general , one also expects @xcite further size corrections @xmath75 and @xmath76 . the second functional forms eqs . and are of course included into the first forms eqs . and , when the fitting parameters @xmath77 or @xmath78 are found to be zero however , given the finite values of @xmath79 and @xmath3 that we can reach and the error bars inherent to qmc , the fits to eqs . and are better controlled ( and errors on estimated parameters smaller ) by forcing @xmath77 to be zero for systems where no log term is present . indeed , putting a log term when not needed can result in an acceptable fit where an artificial @xmath80 compensates wrongly underestimated @xmath32 or @xmath81 . for systems where no log term is present , we must have @xmath82 and @xmath83 ( respectively @xmath84 and @xmath85 ) for large enough sizes , but this scaling regime might be reached earlier by using the second forms eqs . , . let us finally mention the simple argument that if one is looking for universal constants , then only @xmath77 and @xmath78 can be universal ( but not @xmath86 ) in the first forms eqs . and : this is seen by a redefinition of sample size @xmath79 or @xmath3 . with the same reasoning , only @xmath87 can be universal for the second forms eqs . and . for all fits , we used a rigorous bootstrap analysis in order to provide reliable error bars for fit parameters . note , however , that these error bars do not contain systematic effects due to finite system sizes . these effects can nevertheless be estimated by comparison of fits over different system size @xmath79 or @xmath3 ranges ( `` fit windows '' , see ref . for details ) . we also monitored the fit quality @xmath88 ( see ref . ) to ascertain the precision of our fits . throughout this section , we restrict our discussion and analysis to the computationally most accessible sr entropy , when @xmath89 for both the full system ( @xmath31 ) in sec . [ sec : srfull ] and the line subsystem ( @xmath53 ) in sec . [ sec : srline ] . for different values of the plaquette coupling strength @xmath24 . the result for the limit @xmath90 is given by @xmath91 ( see appendix [ sec : exact ] ) . an emerging logarithmic scaling term for @xmath92 can be guessed . lines are guides to the eye . ] figure [ fig : plaq_full_sinf ] shows our qmc result for @xmath93 of the plaquettized model in the range of accessible entropies ( our simulations are limited roughly to @xmath94 as discussed in appendix [ sec : mc ] ) , for different values of the parameter @xmath24 in the range @xmath95 $ ] . in the limit @xmath90 of isolated plaquettes , @xmath96 can be exactly ( _ cf . _ appendix [ sec : exact ] ) shown to be @xmath97 , _ i.e. _ a pure linear scaling with no logarithmic or constant terms . in the uniform heisenberg limit @xmath98 on the other hand , previous results @xcite have shown the existence of a logarithmic scaling correction with the form eq . with @xmath99 . by inspection of the bare sr entropy scaling in fig . [ fig : plaq_full_sinf ] , a nonzero logarithmic scaling term @xmath100 can be presumed for the whole ordered phase @xmath92 ( with curves clearly bending downwards for smaller system sizes ) , while for the disordered phase , the scaling appears linear . in order to quantify this , we have performed fits of the monte carlo data corresponding to the form eq . ( [ eq : logscaling ] ) . we emphasize that the quality of the fits ( in particular the extraction of the logarithmic term ) is reduced when only few system sizes are available , which is specially the case in the disordered regime of the phase diagram ( due to faster growing @xmath31 with system size ) . as the situation is worse for the dimerized model ( we have for instance @xmath101 , _ cf . _ appendix [ sec : exact ] ) , we concentrated our analysis for this section on the plaquettized model . [ fig : plaq_linf ] displays the result of our fits for the prefactor @xmath102 of the logarithmic scaling correction of @xmath103 . the trend with increasing system sizes included in the fit is evident in the ordered phase , as @xmath104 is found to be almost constant with @xmath24 there . in the disordered phase , large finite size effects are observed which are very similar to the oscillations found for the constant term close to the quantum phase transitions of transverse field ising models @xcite . we expect @xmath104 to vanish in the complete quantum disordered phase ( as it is shown analytically for @xmath90 in appendix [ sec : exact ] ) and our data are consistent with this expectation , although the numerical precision is not sufficient for a definite answer . the lack of availability of larger @xmath79 also prevents us to conclude if there is a universal number @xmath77 ( and what is its numerical value ) in the nel phase , even though the plateau shape of the curves tend to indicate that this is possible . the actual universal value @xmath77 ( if any ) may be quite larger than the maximum value here ( found to be @xmath105 for the fit window with the largest @xmath79 ) , as can be seen by the shift of the curves when smaller sizes are removed from the fit . of the logarithmic scaling term of @xmath103 extracted from fits over different size windows for the plaquettized square lattice . @xmath106 is the maximal accessible size for which the entropy is smaller than @xmath107 ( see fig . [ fig : plaq_full_sinf ] ) . we used @xmath108 for @xmath109 and were able to push calculations up to @xmath110 around the critical point and even to @xmath111 for @xmath112 . for @xmath113 , the fits are difficult because of greater errorbars for large entropies and large finite size effects . the behavior is nevertheless consistent with a vanishing @xmath104 in the disordered phase . for @xmath92 a plateau emerges and @xmath104 is found to assume approximately the same nonzero value in the whole ordered phase . the inset shows the subleading constant term @xmath114 obtained from fits excluding a logarithmic scaling term , in the relevant low-@xmath24 phase . ] the inset of fig . [ fig : plaq_linf ] shows our fit results for the same fit windows as in the main panel for @xmath114 as obtained from fits to eq . close to @xmath115 . in this region , large finite size effects are hampering a reliable extraction of the constant but a lower bound for the value @xmath116 at the critical point can be perceived . results from fit windows excluding smaller system sizes seem to indicate that @xmath117 vanishes in the disordered phase . it would be of clear interest to increase the maximum size in the simulation to have a larger fitting range , but this is not possible with the extensive growth of the entropy @xmath31 . to circumvent this problem , we consider in the next section the scaling behavior of a the sr entropy of a subsystem , which grows much more slowly . , top panel ) and dimerization ( @xmath118 , bottom panel ) transitions.,title="fig : " ] , top panel ) and dimerization ( @xmath118 , bottom panel ) transitions.,title="fig : " ] we present in this section our qmc results for the line subsystem sr entropy @xmath119 . its scaling with the length of the line @xmath3 will be shown to also capture the nature of the ordered and paramagnetic phases . @xmath119 is equal to ( minus ) the natural logarithm of the maximum diagonal entry of the line reduced density matrix , which turns out to correspond to the two local nel states @xmath120 and @xmath121 on the line . this is slightly less obvious than the fact that the full nel states are the most probable states on the full lattice , but we checked explicitly that this is the case in all our simulations . by definition of the reduced density matrix , @xmath122 contains now information about all basis states of the full system which fulfill the geometrical condition of forming one of the two nel states on the subsystem . of the subsystem entropy @xmath123 across the plaquettization ( top ) and dimerization ( bottom ) transitions . we show both fits to the forms @xmath124 ( eq . ( [ eq : bscalingline ] ) , valid for @xmath113 , bold in valid regime , pale for @xmath92 ) and @xmath125 ( eq . ( [ eq : logscalingline ] ) , valid for @xmath92 , bold in valid regime , pale for @xmath113 ) . as @xmath126 and @xmath127 in the quantum disordered phase when @xmath113 , the fits forcing @xmath127 are slightly better . in the ordered phase , the fit to eq . does not work because of the existence of the logarithmic scaling term and fit quality factors of @xmath128 ( see _ e.g. _ ref . ) were obtained here . , title="fig : " ] of the subsystem entropy @xmath123 across the plaquettization ( top ) and dimerization ( bottom ) transitions . we show both fits to the forms @xmath124 ( eq . ( [ eq : bscalingline ] ) , valid for @xmath113 , bold in valid regime , pale for @xmath92 ) and @xmath125 ( eq . ( [ eq : logscalingline ] ) , valid for @xmath92 , bold in valid regime , pale for @xmath113 ) . as @xmath126 and @xmath127 in the quantum disordered phase when @xmath113 , the fits forcing @xmath127 are slightly better . in the ordered phase , the fit to eq . does not work because of the existence of the logarithmic scaling term and fit quality factors of @xmath128 ( see _ e.g. _ ref . ) were obtained here . , title="fig : " ] we display our results for the line subsystem sr entropies as a function of the length @xmath3 of the subsystem for both dimerized and plaquettized models in fig . [ fig : all_line_sinf ] . much larger system sizes @xmath17 are accessible now ( when compared to fig . [ fig : plaq_full_sinf ] for the full system ) : this greatly reduces the effect of further finite size corrections beyond eqs . and and makes a reasonable analysis of the scaling of subsystem entropies viable . we now discuss systematically the scaling behavior of the sr entropy @xmath6 across the plaquettization - dimerization transitions , by fitting to the functional forms eq . and , and displaying the estimates of fits parameters . of the line sr entropy @xmath53 across the plaquettization ( top ) and dimerization ( bottom ) transitions , as obtained from fits to eq . . we show fits over different system size windows . the logarithmic term vanishes in the quantum disordered phase , while in the ordered phase it assumes a nonzero , almost constant value , which is similar for both models for a given fitting size window . , title="fig : " ] of the line sr entropy @xmath53 across the plaquettization ( top ) and dimerization ( bottom ) transitions , as obtained from fits to eq . . we show fits over different system size windows . the logarithmic term vanishes in the quantum disordered phase , while in the ordered phase it assumes a nonzero , almost constant value , which is similar for both models for a given fitting size window . , title="fig : " ] of the subsystem entropy @xmath53 across the plaquettization ( top ) and dimerization ( bottom ) transitions , as obtained by a fit to eq . . this form is clearly not valid in the ordered phase ( @xmath92 ) , where a logarithmic scaling term of @xmath129 is found . fit qualities drop to zero for @xmath92 and @xmath130 is therefore shown in pale colors . in the disordered phase @xmath113 , @xmath131 is found to be 0 ( bold ) . the lines cross at the critical point at @xmath132 ( plaquettized model ) and @xmath133 ( dimerized model ) . , title="fig : " ] of the subsystem entropy @xmath53 across the plaquettization ( top ) and dimerization ( bottom ) transitions , as obtained by a fit to eq . . this form is clearly not valid in the ordered phase ( @xmath92 ) , where a logarithmic scaling term of @xmath129 is found . fit qualities drop to zero for @xmath92 and @xmath130 is therefore shown in pale colors . in the disordered phase @xmath113 , @xmath131 is found to be 0 ( bold ) . the lines cross at the critical point at @xmath132 ( plaquettized model ) and @xmath133 ( dimerized model ) . , title="fig : " ] we begin with the linear prefactors @xmath134 , as displayed in fig . [ fig : all_line_ainf ] as a function of @xmath24 for fits over different system size windows . the results for the two functional forms are shown in the same figure , but with a different color coding depending on the regimes : @xmath135 obtained from linear fits [ eq . ] is represented with bold lines for @xmath113 ( in the disordered regime where we find that they represent the correct form ) and pale lines for @xmath92 ( when they are not expected to be valid ) and vice - versa for fits including the logarithmic correction [ eq . ] . for @xmath113 , both results agree very well within error bars , while the linear fit result is slightly more stable and converges faster with system size . this is already a hint that the logarithmic correction @xmath136 presumably vanishes in the disordered phase , which will be verified in the next paragraph . both dimerized and plaquettized models display the same behavior , with @xmath137 taking larger values due to suppressed nel order . one can notice a qualitative change in the extensive contribution to the shannon entropy across the quantum phase transition where @xmath138 changes abruptly . more precisely , its derivative with respect to @xmath24 displays a singularity at the critical point . we discuss in more detail such features in sec . [ sec : detecting ] . the first subleading scaling term is the logarithmic correction @xmath139 as defined in eq . . [ fig : all_line_linf ] represents results of fits obtained from three sets of system size ranges . we find that fits excluding the smallest system sizes generally correspond to higher fit qualities ( quality factor @xmath88 closer to @xmath140 ) while on the other hand , error bars on @xmath139 become larger as the number of data points included in the fit decreases . nevertheless , results are stable with respect to different fit windows : we observe a clear change in the estimated @xmath139 exactly at the transition point for both dimerized and plaquettized models at the respective @xmath115 . deep in the quantum disordered phase , the logarithmic term @xmath141 converges very well towards zero . close to the critical point for @xmath113 , nontrivial finite size effects show up in pronounced oscillations preceding the jump to nonzero @xmath139 in the ordered phase . similar to what is observed in the constant term of the sr entropies of the one - dimensional @xcite and two - dimensional @xcite quantum ising model close to its transition point , the oscillations become narrower and move closer to the critical point with growing system sizes used for the fit . we conclude that @xmath142 in the full disordered phase . in the ordered phase , the behavior is very different and a logarithmic scaling correction emerges with @xmath143 . our results for the fitting window with the larger sizes is @xmath144 and @xmath141 appear identical for both models within the nel phase . however , even though we performed calculations in large systems of up to @xmath145 spins , the asymptotic value of @xmath141 can not be extrapolated from our data . right at the critical point , curves for the estimated @xmath141 for different fit windows cross at a value which is @xmath146 within error bars . in the quantum disordered phase and presumably also at the critical point , the logarithmic term vanishes and therefore the first subleading scaling term is @xmath147 . to best estimate its value , we force @xmath148 by using the functional form eq . in our fit . [ fig : all_line_binf ] shows the result of this analysis , the pale lines correspond to the regime @xmath92 where the fit function does not represent the data correctly ( this is reflected also by strong finite size effects ) . we find exactly the same behavior for both models in the disordered phase with @xmath149 for all @xmath113 . we furthermore find ( see fig . [ fig : all_line_binf ] ) that curves of @xmath150 for different fit windows cross at the critical point , taking a non - trivial value @xmath50 . the absence of finite size effects at the crossing point provides evidence that the logarithmic correction actually vanishes at the critical point . for the plaquettized model , we find @xmath132 ; in the dimerized case we obtain a similar value @xmath133 . this strongly suggests that @xmath50 is _ universal _ at the quantum critical point , and should be identical for all models with a phase transition in the 3d @xmath1 universality class . to test this , we perform large - scale simulations of the _ finite - temperature _ transition in the isotropic 3d @xmath8 heisenberg model on a cubic lattice with antiferromagnetic interactions . this transition belongs to the 3d @xmath1 universality class . we then computed the line sr entropy @xmath123 using the same qmc technique @xcite , but this time at finite temperature , close to the critical point . as a function of temperature @xmath151 for different linear system sizes @xmath3 in the 3d @xmath8 antiferromagnetic heisenberg model on the simple cubic lattice . the inset shows our best estimate for the crossing point @xmath152 of the spin stiffness for sizes @xmath3 and @xmath153 as obtained from a cubic fit to our data including a bootstrap analysis for the error bars ( black points ) . the red line corresponds to @xmath152 calculated from our best fit of our data to a universal function @xmath154 ( _ cf . _ ref . ) , including empirical scaling corrections . we approximated @xmath155 by a third order polynomial . for the critical temperature , we obtain @xmath156 using @xmath157 bootstrap samples ( 1@xmath158 errorbar indicated by the narrow rectangle ) . ] as a preliminary , we want first to extract the best estimate for the critical temperature @xmath159 . we have performed additional simulations , up to @xmath160 sites , studying the crossings of the spin stiffness ( times linear system size ) , a standard method to locate critical points @xcite . these results are reported in fig . [ fig:3dtc ] where we show very precise qmc data for cubic systems , thus allowing to estimate the critical point with a high accuracy to @xmath161 , which agrees with previous estimate @xmath162 from ref . . while measuring the spin stiffness within the sse computation is very standard @xcite and relatively fast , accessing @xmath31 for a single line in the cubic antiferromagnet requires much longer simulation time . we have been able to reach system sizes up to @xmath163 for @xmath164 for which the subleading constant @xmath165 is shown in fig . [ fig:3db ] in the vicinity of @xmath159 . despite the sizable error bars , we can nevertheless observe a clear crossing for various fit windows , drifting towards the actual critical point at @xmath161 where the subleading constant takes a numerical value @xmath166 . this value is in perfect agreement with estimates for the two - dimensional quantum critical points , thus reinforcing the evidence for the universality of @xmath167 for 3d o(3 ) critical points . in the scaling of the sr entropy @xmath168 with system size close to the critical temperature in the 3d heisenberg model . we only performed fits to the form in eq . ( [ eq : bscalingline ] ) , which is strictly only valid in the absence of logarithmic scaling terms ( _ i.e. _ in the disordered phase at @xmath169 ) . at the critical point , @xmath165 converges well with system size and the estimate for the fit window with the largest sizes is given by @xmath170 . the insets show the scaling of @xmath168 as a function of @xmath3 in the ordered phase @xmath171 with a clear sign of the logarithmic scaling correction ( @xmath172 ) and in the paramagnetic phase @xmath173 , where the scaling is purely linear with a vanishing constant @xmath174 . ] away from criticality , we have also checked the scalings of @xmath164 in the low temperature ordered phase at @xmath175 ( left inset of fig . [ fig:3db ] ) and in the high temperature disordered regime at @xmath176 ( right inset of fig . [ fig:3db ] ) . as expected , below @xmath159 , a subleading logarithmic term emerges with @xmath172 , and a purely linear scaling is found above @xmath159 , with a vanishing constant @xmath174 . up to now , we have focused on the finite - size behavior of a single quantity , namely @xmath31 ( @xmath53 ) , which is related to a single diagonal element - the largest - of the ( reduced ) density matrix . let us now inspect the behavior of all the diagonal elements of the reduced density matrix @xmath177 for a subsystem @xmath37 being , as above , a line of @xmath3 spins embedded in a @xmath4 torus . for practical reasons , we again restrict ourselves to the set of bases that are connected to the @xmath7 basis by global @xmath2 transformations , leaving the hamiltonian invariant . inspired by recent insights obtained on the _ entanglement spectrum _ @xcite , we introduce the _ participation spectrum _ obtained from the diagonal of the reduced density matrix @xmath178 in the computational basis @xmath179 @xmath180 using the line shaped subsystem defined in fig . [ fig : lattices ] . from now on , we drop the index ` line ' on the set of pseudo - energies @xmath181 . in order to clarify the tremendous amount of information contained in the participation spectrum , we anticipate ( as detailed below ) that the line participation spectrum will develop well - defined bands that can be classified through specific characteristics of their containing basis states : ( absolute value of ) magnetization @xmath182 and number of ( ferromagnetic ) domain - walls @xmath183 , the later turning out to be the crucial element to classify the spectrum . the @xmath184 operator being diagonal in the computational basis @xmath179 , the magnetization of a basis state @xmath15 is simply defined as @xmath185 . we define the total number of ferromagnetic domain walls in the line as @xmath186 . we assume periodic boundary conditions ( @xmath187 ) along the chain with @xmath3 even ( ensuring @xmath183 to be an even number ) . in other words , @xmath183 is simply the number of bonds along the chain hosting nearest - neighbors spins with the same orientation in basis state @xmath15 . we use the term _ domain walls _ since the most likely states are the two nel states on the chain , as mentioned earlier , which have @xmath188 and @xmath189 ( ferromagnetic orientation of spins on a given bond correspond to disrupting local nel ordering ) . magnetization @xmath182 and number of domain walls @xmath183 run from @xmath146 to respectively @xmath190 and @xmath3 ( for the polarized ferromagnetic state ) for possible basis states on the chain . we expect states with low @xmath183 and low @xmath182 to be more likely , and therefore to have a lower pseudo - energy @xmath181 . and @xmath191 separating the two nel configurations @xmath192 and @xmath193 . ] a typical basis state is illustrated in fig . [ fig : dwspict ] for a line of 30 spins with @xmath194 and @xmath195 . from such a picture one sees that increasing the separation between two domain walls tends to reduce the total staggered magnetization @xmath196 . it is therefore expected that the effective interaction between domain walls will be strongly ( weakly ) attractive for states having long - range ( short - range ) antiferromagnetic correlations . this will be discussed on more quantitative grounds below in sec . [ sec : detecting ] . we finally note that this description is not sufficient for characterizing states of the chain subsystem in the plaquettized lattice , as readily seen in fig . [ fig : lattices ] . while all bonds along the chain are equivalent for the dimerized lattice , this is not the case for the plaquettized lattice with `` strong '' bonds carrying the coupling constant @xmath197 and `` weak '' bonds carrying @xmath60 . we therefore find it useful to define the number of strong @xmath198 domain walls ( @xmath199 ) . we assume that the chain subsystem starts from a strong bond and that @xmath3 is a multiple of @xmath200 , as is the case in fig . [ fig : lattices ] . we expect that ferromagnetic domain - walls on strong bonds will be less likely than on weak bonds and anticipate this notion of strong and weak domain walls to be particularly relevant in the quantum disordered phase . to get a first idea on how the weight of each basis state gets redistributed while passing through the quantum phase transition , it is instructive to consider the density of states @xmath201 corresponding to the participation spectrum of the line subsystem . accross the transition on the plaquettized ( left ) and dimerized ( right ) square lattice . here , @xmath202 . [ fig : dos ] ] figure [ fig : dos ] displays the density of states as a function of @xmath24 across the transition in dimerized ( right ) and plaquettized ( left ) square lattices . while the two density of states naturally develop into the same homogeneous limit of @xmath112 , they appear to differ strongly for dimerized and plaquettized models , specially in the quantum disordered phase ( which is nevertheless physically similar for both models ) . this difference is readily understood by considering the limit of @xmath90 ( see appendix [ sec : exact ] ) where the reduced density matrix for isolated plaquettes and dimers is quite different . indeed , all diagonal elements of the reduced density matrix are equal to @xmath203 for the dimer case , whereas this large degeneracy is lifted by the existence of two different diagonal entries in the reduced density matrix at @xmath90 for a single plaquette . for the plaquettized model , the number @xmath204 of strong domain walls determines the value of the diagonal reduced density matrix element of the line by @xmath205 and therefore labels the different packets of states at @xmath90 . this degeneracy is lifted at @xmath206 and different packets of states tend apart from their initial pseudo - energy @xmath181 . due to the high complexity of the spectrum in the plaquettized model , these packets are mixed in energy at the critical point but become `` fat '' as they cross the critical point . this phenomenon , which will be discussed in detail in sec . [ sec : fullspectrum ] , is even more visible in the participation spectrum of the dimerized model . there , there is no distinction between strong and weak domain walls , and the well - separated bands evolve smoothly with @xmath24 across the phase transition . in the heisenberg limit @xmath112 . basis states with different numbers of domain walls are clearly separated in pseudo - energy @xmath207 , while the bands come closer with system size and have a small overlap . note that the density of states is normalized for each state band . the fine line displays the overall density of states scaled by a factor for visibility . [ fig : dosdw ] ] on the other side of the transition , the density of states in the homogeneous limit @xmath98 shows bands of high density in pseudo - energy , separated by local minima . inspection of the corresponding states confirms our intuition by revealing that the bands can be characterized by the number of domain walls @xmath183 in the basis states . [ fig : dosdw ] illustrates this by displaying the density of states for a fixed number of domain walls @xmath208 @xmath209 where the sum runs only over states with @xmath208 domain walls . we compare the domain wall resolved density of states for two lattice sizes @xmath202 and @xmath210 in fig . [ fig : dosdw ] and it is apparent that the bands seen in fig . [ fig : dos ] in the ordered phase correspond to states with a fixed number of domain walls . this is in contrast with the gapped phase of the plaquettized model , where the number of strong domain walls is the dominant characteristic for the bands ( see fig . [ fig : dos01 ] ) . with growing system size , the number of possible bands grows linearly , as the maximal number of domain walls grows linearly in @xmath3 . also , on the ordered side of the transition the width of the bands grows linearly in @xmath3 ( as will be shown in sec . [ sec : detecting ] ) and the bands come closer together , eventually forming a continuum of states . a similar picture of bands labelled by the number of spin flips has also been proposed for the entanglement spectrum of quantum dimer models on a cylinder @xcite . and @xmath210 ) . the number of domain walls is not a good identifier of the state packets in this case . in the gapped phase , even in the thermodynamic limit there are pronounced gaps between the bands of states which in this case are characterized by the number of strong domain walls @xmath204 ( indicated by bold face ( red ) numbers ) . [ fig : dos01 ] ] we now turn our attention to the fully resolved participation spectrum for the plaquettization transition , where states are distinguished by @xmath211 . the use of magnetization @xmath182 is mainly for clarity reasons : knowing @xmath184 does not help in resolving bands which are characterized by the number of domain walls ( even though some values of magnetization do not accommodate all possible number of domain walls ) . the splitting of the participation spectrum in different sectors of @xmath184 is displayed in fig . [ fig : plaq_spectrum ] for different values of @xmath24 for the plaquettized model , illustrating more clearly how the pseudoenergies @xmath181 vary from being grouped by their number of strong domain walls @xmath212 ( at @xmath90 ) to their total number of domain walls @xmath183 ( at @xmath112 ) . a more detailed look at the development of state packets with fixed @xmath213 is provided in fig [ fig : plaq_spectrum_dw ] where we concentrate on all basis states with exactly @xmath214 domainwalls . ( indicated by the numbers on top of each state column ) across the plaquettization transition for @xmath215 . [ fig : plaq_spectrum ] ] ) . bold face ( red ) numbers indicate the number of strong bonds for the state packet . [ fig : plaq_spectrum_dw ] ] while the previous sections aimed at illustrate how participation spectra evolve though the continuous quantum phase transition , we present now two distinct quantitative features for localizing the quantum phase transition in the ( fully - resolved ) participation spectrum . as visible in figs . [ fig : plaq_spectrum ] and [ fig : plaq_spectrum_dw ] , the behavior of the pseudo - energy @xmath181 of each basis state is peculiar close to the quantum critical point , showing a maximal or minimal steepness of @xmath216 and a zero transition of the second derivative @xmath217 , _ i.e. _ every @xmath216 shows an inflection point at the critical point . while the change of curvature of @xmath216 close to @xmath115 is clearly visible for most basis states in figs . [ fig : plaq_spectrum ] and [ fig : plaq_spectrum_dw ] , a quantitative demonstration that the inflection point lies right at the critical point requires high accuracy of @xmath181 and a fine grid in @xmath24 . we therefore performed this analysis for @xmath218 for best control . for the plaquettized ( left ) and dimerized ( right ) lattices for different lattice sizes . the sign change shows a clear inflection point right at the quantum phase transition , which corresponds to a pronounced minimum of the first derivative of @xmath6 . ] we compute numerically the second derivative of @xmath123 with respect to @xmath24 close to the quantum phase transition and display it in fig . [ fig : plaq_line_sinf_diff2 ] for both the dimerized and plaquettized models , confirming that the quantum phase transition corresponds to a zero in the second derivative . in the limit @xmath219 , the first derivative of @xmath6 diverges right at the critical point and thus precisely marks the quantum phase transition . of state packets as a function of system size @xmath3 for fixed @xmath220 ( top panel ) and @xmath208 ( bottom panel , including states from all @xmath184 sectors ) across the plaquettization ( top ) or dimerization ( bottom ) transition . lower numbers of domain walls @xmath208 correspond to low pseudo - energy part of the participation spectrum . the crossing of @xmath221 for different sizes close to the critical point indicated by a vertical line is most prominent for the subsystem basis state packets with low pseudo - energy ( _ e.g. _ for the packet with @xmath222 domain walls for both panels ) but clearly exists with considerable larger size effects ( such as drift crossing ) in higher parts of the spectrum . [ fig : allpacks],title="fig : " ] of state packets as a function of system size @xmath3 for fixed @xmath220 ( top panel ) and @xmath208 ( bottom panel , including states from all @xmath184 sectors ) across the plaquettization ( top ) or dimerization ( bottom ) transition . lower numbers of domain walls @xmath208 correspond to low pseudo - energy part of the participation spectrum . the crossing of @xmath221 for different sizes close to the critical point indicated by a vertical line is most prominent for the subsystem basis state packets with low pseudo - energy ( _ e.g. _ for the packet with @xmath222 domain walls for both panels ) but clearly exists with considerable larger size effects ( such as drift crossing ) in higher parts of the spectrum . [ fig : allpacks],title="fig : " ] of state packet ( 1 , 2 , 1 ) as a function of inverse system size @xmath223 for different values of @xmath24 in the case of the plaquettization transition . in the limit of large system sizes , @xmath221 vanishes in the quantum disordered phase , while it goes to a constant in the ordered phase . inset : zoom for small values of @xmath24 in the gapped phase . the linear behavior in @xmath223 suggests that @xmath221 vanishes like @xmath223 . [ fig : deltal ] ] another way to quantitatively detect the quantum phase transition is obtained following the previous observation that bands of identical number of domain walls in the participation spectra appear to become fatter " as the system crosses the quantum critical point . we investigate the size of state packets labeled by @xmath224 in the plaquettized case by looking at the difference in pseudo - energy @xmath225 between the basis states with the largest and lowest pseudo - energy within the packet @xmath226 . results for the normalized packet width @xmath227 for different @xmath226 ( see top panel of fig . [ fig : allpacks ] ) display a clear crossing point at the quantum critical point , separating two different regimes . in the quantum disordered phase , the normalized packet width @xmath227 tends to vanish presumably as @xmath223 in the thermodynamic limit , as revealed by the finite - size scaling analysis in the inset of fig . [ fig : deltal ] . on the other hand , the packet width @xmath228 grows as @xmath3 ( with @xmath223 correction ) in the ordered phase as also seen in fig . [ fig : deltal ] . the same behavior is revealed when integrating over all @xmath184 sectors as shown for the case of the dimerized model in the bottom panel of fig . [ fig : allpacks ] , where state packets are defined by a fixed number of domain walls @xmath208 . ( expressed in terms of the chord distance on the ring ) for all states with @xmath222 domain walls in the @xmath194 sector for the plaquettized ( top ) and dimerized ( bottom ) model , for different values of @xmath24 ( equidistant in steps of @xmath229 ) . in order to compare the curves , we subtracted the pseudo - energy of the state with the minimal distance @xmath230 between domain walls ( which is always the lowest ) . , title="fig : " ] ( expressed in terms of the chord distance on the ring ) for all states with @xmath222 domain walls in the @xmath194 sector for the plaquettized ( top ) and dimerized ( bottom ) model , for different values of @xmath24 ( equidistant in steps of @xmath229 ) . in order to compare the curves , we subtracted the pseudo - energy of the state with the minimal distance @xmath230 between domain walls ( which is always the lowest ) . , title="fig : " ] this behavior can be understood following the sketch presented in fig . [ fig : dwspict ] , where one sees that the dynamics of two domain walls is constrained by the relative sizes of two nel patterns , @xmath192 and @xmath193 . since the staggered magnetization of a single basis state depends on the size difference @xmath231 between nel domains of different type , one can predict a qualitative difference for the packet width @xmath228 between ordered and disordered ground states . in the disordered phase @xmath113 where antiferromagnetic correlations are short - ranged , we naturally expect that the pseudo - energies will not be affected ( remaining essentially constant ) when the separation between two domain walls becomes larger than the finite correlation length . on the other hand , for long - range order a confinement mechanism between two domain walls will be necessary to maintain a finite staggered magnetization . there we expect the pseudo - energy spectrum to be controlled by an attractive long - range interaction between domain walls . this picture holds best for a small number of domain walls in the high probability ( low pseudo - energy ) part of the participation spectrum . we illustrate this interpretation by considering the pseudo - energy dependence on the distance between domain walls . [ fig : dwpotential ] displays this pseudo - energy difference for the diluted case of @xmath195 domain - walls with @xmath194 ( in which case the distance has to be even ) , as a function of the chord distance between domain walls for different @xmath24 for both dimerized and plaquettized models . in the disordered phase , the domain walls appear rapidly deconfined with a finite small pseudo - energy difference between states with different domain wall distances and hence a small packet width @xmath228 . in contrast , large domain wall distances are penalized in the magnetically ordered phase by a high pseudo - energy cost , which appears to grow approximatively linearly with distance ( for a large enough distance ) . for states with more than two domain walls , the situation becomes somewhat more complicated . the crossing in @xmath221 still exists but it acquires a drift with system size ( see fig . [ fig : allpacks ] ) . a closer inspection of the involved states in the corresponding packet and their pseudo - energies suggests that multi - domain wall attraction terms play a role in addition to the long - range @xmath222 domain - walls attraction . this is probably the source of the a drift of the crossing as these multi - domain wall terms become eventually less important when the average distance between domain walls becomes large @xmath232 for the high end ( low probability ) of a packet with a fixed number of domain walls . we have analyzed the nel antiferromagnet - paramagnet quantum phase transition in two - dimensional quantum spin systems using the shannon - rnyi entropies for the full system and for subsystems , together with the associated participation spectra , using extensively qmc methods presented in ref . . our study shows that a line shaped one - dimensional subsystem is actually sufficient to capture the quantum phase transition . we confirm that the subleading scaling behavior of shannon - rnyi entropies changes radically at the quantum phase transition , giving rise to a logarithmic scaling term in the nel phase , independent on microscopic details ( such as the choice of dimerization or plaquettization of the lattice in our study ) . similar logarithmic corrections to an _ area law _ have been numerically observed@xcite in studies of the rnyi entanglement entropy of the ground - state of the 2d heisenberg model for _ two - dimensional _ subsystems ( such as half - torus ) , in agreement with theoretical predictions @xcite . for the case of a line - shaped subsystem , we have checked that the rnyi entanglement entropy also exhibits a logarithmic correction to the area law ( indistinguishable from a volume law in this particular case ) . on the disordered side of the phase diagram , only a subleading constant term can be present , which is actually @xmath146 in this phase with no broken symmetry . at the quantum critical point , we find a universal subleading constant ( for the line sr entropy @xmath53 ) with an estimated value of @xmath233 being identical for dimerized and plaquettized models . we have confirmed that a similar constant can be found at the nel - paramagnetic finite - temperature phase transition of the 3d heisenberg @xmath8 antiferromagnet at @xmath161 , strongly suggesting that this is a characteristic of the 3d @xmath1 universality class to which all mentioned transitions belong . we suspect that such a universal subleading constant also exists for the sr entropy of the full system , but the precision of our numerical computations do not allow to prove this . universality could be further checked by considering @xmath1 critical points in other models , either with two - dimensional quantum models such as @xmath8 bilayers @xcite , coupled haldane chains @xcite , staggered - dimer models where anomalously large corrections to scaling are known @xcite , or in 3d classical systems such as the classical heisenberg model on the cubic lattice . it would also be very interesting to extend this study to other universality classes , either of conventional or unconventional type @xcite . motivated by the finding that the one - dimensional subsystem captures the physics of the phase transition , we also performed a phenomenological study of the information contained in the participation spectra of the subsystem . in the participation spectra , states group together in packets of pseudo - energy which can be classified by the number of ferromagnetic domain walls ( number of strong domain walls for the plaquettized lattice ) in the basis states . the development of these packets across the transition is peculiar as packets become `` fat '' in the ordered phase , meaning that their width @xmath228 ( in pseudo - energy ) grows linearly with subsystem size @xmath3 . in the disordered phase , however , @xmath221 tends to zero . this can be phenomenologically explained by an attractive potential between domain walls in the ordered phase , while in the disordered phase , domain walls remain deconfined . in addition , we observe at the critical point an interesting behavior for the pseudo - energies in the participation spectrum : their slope as a function of the control parameter @xmath24 has a pronounced extremum and we show that @xmath6 has an inflection point at @xmath115 . the analysis presented in sec . [ sec : ps ] is fairly simple , but contains the basic ingredients to build a wave - function that describes the studied ( 2 + 1)-dimensional quantum phase transition . let us indeed put in perspective our work with the widely - used variational approach . in the context of two - dimensional heisenberg models , huse and elser @xcite ( see also refs . ) formulated a variational ansatz for the ground - state wave function : @xmath234 where a classical pseudo - energy @xmath181 is associated to the basis state @xmath15 ( also taken as a @xmath7 basis state in ref . ) . here @xmath235 is ( a set of ) variational parameter(s ) used to minimize the total energy of the heisenberg hamiltonian . this ansatz looks similar to our definition of the participation spectrum . the crucial difference is of course that the variational approach _ assumes _ a form for the pseudo - energy ( for instance a power - law ising interaction in ref . ) , while our qmc methods can calculate the _ exact _ ( within statistical accuracy ) value of each @xmath181 . note as well that we did not compute , for essentially practical reasons , the participation spectrum for the full system ( as in the variational ansatz ) but rather on a subsystem . nevertheless , our results give the correct qualitative ingredients to construct variational wave - functions of the form of eq . to describe the nel and paramagnetic phases , as well as the transition in between . this could be useful in particular for frustrated spin systems , where qmc is not available . we would like to thank f. assaad , f. bgue , g. misguich , g. roux , s. pujari and j .- m . stphan for many useful discussions , as well as g. misguich , m. oshikawa and x. plat for collaboration on related topics . our qmc codes are partly based on the alps libraries @xcite . this work was performed using numerical resources from genci ( grants 2013-x2012050225 and 2014-x2013050225 ) and calmip and is supported by the french anr program anr-11-is04 - 005 - 01 . here we provide further details ( in supplement to ref . ) on the qmc procedure to access subsystem entropies and participation probabilities , as well as on dealing with statistical uncertainty . as already presented in ref . , the diagonal elements @xmath236 of the full density matrix are readily available in standard qmc techniques such as the standard stochastic series expansion method ( cf . _ e.g. _ ref . ) . in the same vein , the diagonal elements of the reduced density matrix @xmath237 ( see eq . ) are obtained in the following way . it is clear that the probability of observing the basis state @xmath41 in the sse operator string is directly given by @xmath238 hence , the probability of finding the subsystem basis state @xmath43 in subsystem @xmath37 in the stochastic series expansion operator string is @xmath239 in practice , this means that one has to inspect only the @xmath37-part of each basis state in the simulation , and record a histogram of its frequency . the numerical calculation of large sr entropies corresponds to the observation of rare events with low probabilities . it is easy to obtain an approximation of a reasonable upper limit for the calculation of the entropy @xmath240 given by the observation of the most probable state appearing in the sse markov chain with a probability @xmath241 . for the estimation of the maximally accessible entropy , we make the assumption of statistical independence , as statistical correlations will only reduce the maximally accessible entropy . in the markov chain , we measure the observable @xmath242 , yielding @xmath140 of state @xmath15 corresponds to the state @xmath243 with maximal probability and @xmath146 otherwise . the monte carlo average of this observable is given by @xmath244 where @xmath245 is the number of occurrences of the most probable state in the markov chain @xmath246 of length times the expansion order @xmath247 . ] @xmath248 . the standard error of @xmath249 is given by @xmath250 for sufficiently small @xmath241 and @xmath248 inspected states . now we can estimate the standard error of @xmath31 by the linear approximation @xmath251 . to obtain a given relative error @xmath252 for a given number @xmath248 of inspected states , the maximal entropy is then governed by the equation : @xmath253 in a realistic calculation , we inspect typically @xmath254 states . therefore , for a requirement of a relative error @xmath255 , we obtain a maximally accessible entropy @xmath256 . measuring accuracy on the participation spectrum requires special care . it is indeed impractical to construct an observable for each subsystem basis state @xmath257 and measure @xmath258 as proposed in ref . for each observed state @xmath259 in the sse operator string . for the calculation of the participation spectrum , defined as @xmath260 and inducing a nonlinear transformation of the monte carlo data , we perform several independent monte carlo simulations in parallel , each creating a simple histogram @xmath261 containing the count of all observed subsystem basis states . after applying all model symmetries to the histogram to improve the statistics ( see supplementary material of ref . ) , we perform a bootstrap analysis for each @xmath262 , creating bootstrap samples from the histogram counts of the different independent simulations . this provides an unbiased estimation of the standard errors of the mean of all @xmath262 . we did not display error bars in figures [ fig : plaq_spectrum ] and [ fig : plaq_spectrum_dw ] as they are smaller than the line width but they where used for the estimation of the uncertainty in @xmath221 shown _ e.g. _ in fig . [ fig : deltal ] . we provide here simple exact results for sr entropies and reduced density matrices when @xmath90 . these results are useful in two ways : they provide direct insight for the subleading terms of the sr entropies ( which vanish altogether when @xmath90 ) as well as on the starting point of the participation spectrum . for a dimer shaped subsystem @xmath37 of one plaquette , the reduced density matrix is given ( in the @xmath270 basis ) by @xmath271 in the single dimer case , subsystem @xmath37 is just one site and the reduced density matrix ( in the @xmath272 basis ) is @xmath273 the total reduced density matrix of the line shaped subsystem composed of @xmath190 dimers @xmath37 of single plaquettes ( or @xmath3 sites @xmath37 of single dimers in the dimerized case ) is then obtained from the kronecker product of the reduced density matrices : 37ifxundefined [ 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 link:\doibase 10.1007/bf01325284 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.69.1093 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.84.3690 [ * * , ( ) ] link:\doibase 10.1103/revmodphys.80.1355 [ * * , ( ) ] link:\doibase 10.1103/physrevb.80.184421 [ * * , ( ) ] link:\doibase 10.1103/physrevb.82.125455 [ * * , ( ) ] link:\doibase 10.1103/physrevb.84.195128 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.107.020402 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.112.057203 [ * * , ( ) ] link:\doibase 10.1103/physreve.86.021104 [ * * , ( ) ] http://stacks.iop.org/1742-5468/2012/i=10/a=p10026 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.111.017201 [ * * , ( ) ] link:\doibase 10.1143/jpsj.66.2957 [ * * , ( ) ] link:\doibase 10.1103/physrevb.65.014407 [ * * , ( ) ] link:\doibase 10.1103/physrevb.73.014431 [ * * , ( ) ] link:\doibase 10.1103/physrevb.78.132402 [ * * , ( ) ] link:\doibase 10.1103/physrevb.79.014410 [ * * , ( ) ] \doibase doi:10.1063/1.3518900 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.94.217201 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.101.010504 [ * * , ( ) ] link:\doibase 10.1103/physreva.78.032329 [ * * , ( ) ] http://arxiv.org/abs/1210.3781 [ `` , '' ] ( ) , http://stacks.iop.org/1742-5468/2012/i=02/a=p02003 [ * * , ( ) ] link:\doibase 10.1103/physrevb.84.165134 [ * * , ( ) ] link:\doibase 10.1103/physrevb.86.235116 [ * * , ( ) ] link:\doibase 10.1103/physrevb.85.165121 [ * * , ( ) ] http://arxiv.org/abs/1112.5166 [ ( ) ] link:\doibase 10.1103/physrevlett.72.2777 [ * * , ( ) ] link:\doibase 10.1143/jpsj.58.3131 [ * * , ( ) ] link:\doibase 10.1103/physrevb.83.174416 [ * * , ( ) ] link:\doibase 10.1126/science.1091806 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.60.2531 [ * * , ( ) ] link:\doibase 10.1103/physrevb.40.4904 [ * * , ( ) ] link:\doibase 10.1103/physrevb.40.11437 [ * * , ( ) ] link:\doibase 10.1088/1742 - 5468/2011/05/p05001 [ * * , ( ) ] link:\doibase 10.1103/physrevb.76.214428 [ * * , ( ) ]

we examine various possibilities for realistic 5d higgsless models and construct a full quark sector featuring next - to - minimal flavor violation ( with an exact bulk protecting the first two generations ) satisfying electroweak and flavor constraints . the `` new custodially protected representation '' is used for the third generation to protect the light quarks from flavor violations induced due to the heavy top . a combination of flavor symmetries , and rs - gim for the right - handed quarks suppresses flavor - changing neutral currents below experimental bounds , assuming ckm - type mixing on the uv brane . in addition to the usual higgsless rs signals , this model predicts an exotic charge- quark with mass of about 0.5 tev which should show up at the lhc very quickly , as well as nonzero flavor - changing neutral currents which could be detected in the next generation of flavor experiments . in the course of our analysis , we also find quantitative estimates for the errors of the fermion zero mode approximation , which are significant for higgsless - type models . 1.5 cm * a flavor protection for warped higgsless models * 0.2 cm * csaba cski and david curtin * _ institute for high energy phenomenology + newman laboratory of elementary particle physics + cornell university , ithaca , ny 14853 , usa _ + csaki@cornell.edu , drc39@cornell.edu 0.3truecm

0904.2137

one of the most important unresolved questions in particle physics is how exactly the electroweak symmetry is broken . the standard model higgs mechanism provides ample motivation to come up with alternatives . an interesting new possibility is provided by higgsless models @xcite with a warped extra dimension @xcite . a higgs field localized on the ir brane of an rs background is decoupled by taking its vev to be very large , while the masses of the @xmath2 and @xmath3 bosons remain finite and are set by the size of the extra dimension . unitarity of the gauge boson scattering amplitudes can then be ensured via heavy kk gauge boson exchange . such models would solve the little hierarchy problem of randall - sundrum setups and have very distinctive phenomenological consequences . however , it is not clear whether these higgsless rs models can be made completely viable : a large correction to the s parameter makes it difficult to match electroweak precision data , the cutoff scale has to be adequately raised to ensure unitarization happens at weak coupling , and generically fcnc s are not adequately suppressed . many of these initial difficulties have been at least partially addressed . one can tune the effective s - parameter away by making the fermion left - handed fermion wave functions close to flat @xcite , and choosing the right fermion representations can prevent the large top mass from introducing coupling deviations in the @xmath4-vertex @xcite . the cutoff scale can also be raised by lowering the curvature of the extra dimension @xcite . however , once the fermion wave functions are required to be close to flat , the traditional anarchic rs approach to flavor @xcite ( where fermion wave function overlaps generate fermion mass hierarchies and also give a protection called rs - gim against fcnc s @xcite ) can no longer be applied . a possible resolution to this problem is to introduce a genuine five - dimensional gim mechanism , which uses bulk symmetries to suppress flavor violation @xcite . the trick is to impose global flavor symmetries on the bulk , with a large subgroup left unbroken on the ir brane and flavor mixing forbidden anywhere except the uv brane . one can then construct a model where tree - level fcnc s are genuinely vanishing , with the downside that we are no longer trying to explain the quark mass and mixing hierarchies , merely accommodating them . the aim of this paper is to examine the flavor bounds ( similar to @xcite on higgsless models and to present a viable flavor construction for these theories ( see @xcite for other examples in an rs context ) . we have to circumvent the problems usually associated with higgsless models by ensuring that all fcncs are sufficiently suppressed , all tree - level electroweak precision constraints are satisfied , the cutoff scale is sufficiently high . we show that the simplest versions of such a model can not be realistic : imposing an exact gim mechanism for all three generations either drives up the cutoff scale or prevents the s - parameter from being cancelled . instead , the realistic flavor model we propose will have next - to minimal flavor violation ( nmfv ) @xcite , featuring a custodially protected quark representation for the third generation , and an exact gim mechanism implemented for the first two generations only . this choice of representations allows us to isolate the lighter quarks from the dangerous top mass and prevent a large s - parameter without having to increase the bulk coupling and decrease the cutoff scale . flavor - changing neutral currents are controlled by two main mechanisms : the surviving flavor symmetry between the first two generations forces all the mixing to go through the third generation ( hence nmfv ) , which is vital to reduce @xmath5 and @xmath6 mixing . kinetic mixing terms on the uv confine the right - handed fermions to the uv brane and reduce bulk contributions to the couplings , which are the source of off - diagonal neutral couplings . this results in an rs - gim - like flavor suppression mechanism for the right - handed fermions . we also have some freedom to distribute the required charged - current mixing amongst the up- and down - sectors , which reduces neutral - current mixing in each sector . all of this is necessary to sufficiently suppress flavor violation . we find that experimental fcnc bounds systematically constrain the down - sector mixing angles , forcing them to lie within a volume of angle space that is enclosed by a well - defined surface . _ assuming _ the uv kinetic mixing terms obey a cabibbo - type mixing hierarchy , this volume occupies @xmath7 of available angle space . this paper is structured as follows : in section [ s.background ] we review the 5d gim mechanism and introduce the quark representations we will be using . in section [ s.nmfv ] we outline our nmfv quark model and show compliance with electroweak precision data ( ewpd ) . we also examine in detail the errors introduced by the zero mode approximation , and find that one can have zero s - parameter without flatness , provided there is a lot of kk mixing on the ir brane . the flavor suppression mechanisms of the nmfv model are derived in section [ s.flavor ] and demonstrated with the gluon kk contribution to fcnc s in section [ s.fcncbounds ] . numerical results for the mixing constraints are presented in section [ s.numerical ] , and we conclude with section [ s.conclusion ] . after briefly reviewing the gauge sector , we will discuss the full 5d gim mechanism @xcite and how one could apply it to various simple quark models . this will motivate the construction of our next - to minimal flavor violation model in section [ s.nmfv ] . we work on an ads@xmath8 background and parameterize our space - time using conformal coordinates @xmath9 where the uv and ir branes sit at @xmath10 respectively . our gauge sector takes the standard rs form @xcite with @xmath11 on the ir brane , as outlined in @xcite . this means we have an @xmath12 gauge group @xcite in the bulk ( where @xmath13 ) , which is broken by boundary conditions to @xmath14 on the uv brane and @xmath15 on the ir brane . the custodial @xmath16 symmetry protects the @xmath17 ratio from deviations at tree level , and the gauge boson mass is given by the size of the extra dimension . for future convenience , let us define @xmath18 then to leading order in @xmath19 , @xmath20 on a technical note , we include all brane - localized gauge kinetic terms ( bkts ) that are allowed by symmetries , and include their corrections due to 1-loop running effects on their respective branes @xcite . this means that the effective @xmath21 bkt on the uv brane can be negative at the weak scale . we will also localize some fermions on the uv brane , making the effective @xmath22 bkt always positive ( but the effect is very small ) . unless otherwise mentioned , we set all _ effective _ bkts to zero at the weak scale except for @xmath22 , for which we set the bare term to zero . we also set the effective @xmath23 bkts to zero , but note that they could be made negative at the weak scale . in addition to the bkts , the free input parameters in the gauge sector are @xmath24 and the ratio @xmath25 . the sm @xmath26 masses then determine @xmath27 and @xmath28 , while @xmath29 and @xmath30 at the weak scale set the overall size of the 5d couplings . all of our gauge bosons are canonically normalized , with all electroweak coupling corrections ( including @xmath31 and @xmath32 parameters ) pushed into interaction terms . our theory is valid up to a momentum cutoff , which in ads@xmath8 space is given conservatively by @xmath33 where @xmath34 is the largest 5-dimensional coupling . if we take @xmath35 , then to leading order in @xmath19 the 4d coupling @xmath36 can be expressed as @xmath37 @xcite , which together with eq . ( [ e.wmass ] ) gives @xmath38 if we had a physical higgs on the ir brane we could freely choose our kk scale and make the cutoff large , but in the higgsless model we must choose a low curvature to unitarize @xmath39-scattering before the theory becomes strongly coupled . we set @xmath40 which gives @xmath41 and @xmath42 . we also choose @xmath35 , since gauge matching would decrease the cutoff if we made the couplings different . the @xmath43 mass is therefore fixed and of order @xmath44 . our notation for a 5d dirac fermion will be @xmath45 where @xmath46 and @xmath47 are both lh 2-component weyl spinors , and hence @xmath48 is a rh 2-component spinor . we write the boundary conditions for each dirac spinor as @xmath49 at @xmath50 , where @xmath51 means @xmath52 and @xmath53 means @xmath54 . the normalized left - handed zero - mode profile is given by @xmath55 is the rs flavor function . the right - handed profile @xmath56 is defined identically , with @xmath57 . for three generations of a given quark representation we can impose global flavor symmetries that prevent fcncs at tree - level while generating all the required masses and mixings @xcite . in broad strokes , those symmetries must satisfy the following criteria : we need to have enough freedom to generate 6 different 4d quark masses and reproduce the ckm mixing matrix . to ensure that flavor - violating operators are suppressed by the high scale @xmath58 , we only allow flavor mixing on the uv brane , via right - handed kinetic terms for the up and down sector independently . left - handed mixing is assumed to be forbidden by flavor symmetry , since otherwise the right- and left - handed kinetic terms can not be diagonalized simultaneously and there would be fcncs . if we switch off the charged - current interactions we should be able to use the symmetries of the bulk and ir brane to diagonalize the uv mixing matrices . this makes the neutral currents diagonal in the 4d mass basis and forbids tree - level fcncs . we will now see how this mechanism can be applied to various simple quark models . the simplest potentially realistic quark representation is the left - right symmetric representation , which has been adopted in @xcite : @xmath59 we impose the boundary conditions @xmath60 on @xmath61 and @xmath62 ) on @xmath63 to obtain left - handed and right - handed zero modes ( similarly for @xmath64 ) . on the ir brane , both @xmath65 and @xmath66 are @xmath67 doublets , so we can write a dirac mass mixing term @xcite to lift the zero modes , @xmath68 on the uv brane , the @xmath66 doublet breaks up into two @xmath69 singlets , allowing us to assign brane - localized kinetic terms to @xmath63 and @xmath70 separately and supply the proper mass splitting . if we want to implement the full 5d gim mechanism using this representation , we populate the bulk with three copies of @xmath71 and impose flavor symmetries @xmath72 and @xmath73 . this makes the bulk masses @xmath74 and the ir dirac mass @xmath75 flavor blind . on the uv brane we allow only kinetic mixing of the @xmath66 fields , which take the form @xmath76 for both @xmath63 and @xmath70 , where @xmath77 are two independent hermitian matrices and @xmath78 are flavor indices . to forbid left - handed mixing , we impose the flavor symmetry @xmath79 . if we switch off the charged - current interactions , the @xmath80 symmetries become independent and @xmath81 , similarly in the bulk . these symmetries are broken on the uv brane , but we can use them to diagonalize the @xmath77 mixing matrices and end up with @xmath82 , which prevents tree - level fcncs . another way to see this is as follows : since all other mass and kinetic terms in the action are flavor singlets , we can go to the 4d mass basis by rotating the fermions in flavor space and diagonalizing @xmath83 . this pushes all the physical mixing into the charged - current couplings , giving us exactly the standard model ckm structure . the main problem with this model is the top quark . to make it heavy , we must localize it close to the ir brane and make @xmath75 large . the flavor symmetry then forces _ all _ the quarks to be close to the ir , generating a large negative @xmath31-parameter . the large flavor - blind @xmath75 has two additional dangerous effects : firstly , the @xmath84 mixing causes left - handed quarks to live partially in the @xmath24 representation ( and vice versa ) , which induces even more coupling corrections since it has the wrong quantum numbers . secondly , the @xmath85 mixing causes the light quarks to live partially in kk modes , resulting in dangerously high couplings to gauge kk modes . one could try to address these problems by increasing the @xmath67 ir kinetic term , which corresponds to adding a positive bare @xmath31 parameter on the ir brane , but this is not viable . the matching of gauge couplings would force the coupling in the bulk to increase , lowering our cutoff to @xmath86 and making the theory non - perturbative before the unitarization mechanism of higgsless rs models kicks in . we clearly need some way to protect the other quarks from deviations due to the heavy top mass . focusing on the third generation only for a moment , the problem with the left - right symmetric representation is that the effects of the top mass are felt by the left - handed bottom ( we can localize the right - handed bottom on the uv brane ) . al @xcite realized that deviations to those couplings could be avoided if the @xmath88 is not in the same representation as any field which can mix with @xmath89 , so that the top can have a separate ir dirac mass that is not communicated to the bottom , and the representations that house the left - handed @xmath90 are symmetric under @xmath91 interchange . this ensures that the @xmath19 and @xmath24 couplings are the same , meaning the @xmath89 couples to a linear combination of gauge bosons which is flat near the ir brane . its couplings are therefore protected from deviations due to the @xmath92 breaking . in the notation of @xcite , the simplest representation which ( almost ) satisfies these requirements is : @xmath93 for the @xmath94 quarks we impose the same boundary conditions as for the quarks in the left - right symmetric representation , while we make the exotic @xmath32 , @xmath95 quarks ( with electric charge @xmath96 and @xmath1 respectively ) heavy by imposing mixed boundary conditions @xmath97 for @xmath98 and @xmath99 for @xmath100 . on the ir brane , the @xmath101 bidoublet breaks down into an @xmath67 triplet and a singlet , which can mix with @xmath102 and @xmath88 . @xmath103 hence the allowed dirac mass term on the ir brane is @xmath104,\ ] ] so we can make @xmath105 large to get a heavy top without influencing any field with the quantum numbers of the bottom . also note that the custodial protection granted by this representation is not complete : kk - mixing via the @xmath106 mass term causes the left - handed @xmath90 to live partially in @xmath107 , which is _ not _ part of a @xmath108 symmetric representation . symmetric we would have to extend @xmath102 to a @xmath109 . ] this turns out to be good thing , since the fully protected coupling is a few percent too large in higgsless models ( just the effect of the @xmath31-parameter ) . we can reduce it by localizing the @xmath107 closer to the uv brane , which increases @xmath106 ( for a given 4d bottom mass ) and hence increases kk - mixing . this in turn decreases the coupling , since it makes the lh bottom sensitive to the gauge boson profiles near the ir brane . the unique features of this representation allow us to implement the full 5d gim mechanism in a rather different fashion from the previous case . to protect the up and charm quarks from the heavy top , we make @xmath105 different for each quark generation and forbid all up - sector flavor mixing , including on the uv brane . the down sector symmetries , on the other hand , are chosen very similarly to the left - right symmetric representation : @xmath106 is flavor blind and large enough to generate the _ bottom _ mass , and a kinetic mixing matrix @xmath110 on the uv brane generates quark mixing in the down sector ( and hence all the physical quark mixing ) . this amounts to imposing the flavor symmetry @xmath111 , which gets broken down to @xmath112 on the ir brane , and @xmath113 on the uv brane ( we also set all brane kinetic terms for @xmath114 fields to zero ) . fcncs are prevented in the exact same fashion as for the left - right symmetric model , except we now only have to diagonalize the down sector . nevertheless , a higgsless quark model using _ only _ the custodially protected representation is not viable . @xmath102 must be close to the uv to match @xmath115 to the sm . this fixes @xmath116 while leaving the @xmath101 bulk mass @xmath117 unconstrained due to custodial protection . the problem is a tension between the lh up - type couplings and the rh bottom coupling . @xmath118 can only be matched if @xmath119 uv , since it is not protected and suffers large corrections near the ir brane . @xmath120 has the opposite requirement . if @xmath106 is the minimal value to give @xmath121 , @xmath107 lives entirely in the bulk . it must therefore be close to the uv brane to make it insensitive to the broken symmetry on the ir brane , which is not a problem . however , if @xmath122 uv , then @xmath106 must be very large to generate @xmath121 , which increases kk mixing and makes @xmath120 sensitive to the ir brane again , reducing the coupling below sm . so while matching the lh up - type couplings to the sm requires @xmath123 uv , the rh bottom coupling requires @xmath119 ir . it is not possible to match both simultaneously . one might try to increase @xmath106 and confine the rh bottom to the uv brane , allowing @xmath101 to be closer to the uv , but this also increases @xmath32-mixing with the up - type quarks , which forces @xmath101 _ even closer _ to the uv to get a match . one can not achieve overlap . so while this model can protect the left - handed bottom couplings , the 5d gim mechanism forces _ all _ the up - type quarks to behave like the troublesome top , and their couplings can not be matched to the sm simultaneously with the rh bottom . the complete 5d gim mechanism is too restrictive for higgsless rs model - building . we have to give up a some flavor protection in exchange for agreement with electroweak precision data , while ensuring that fcncs are still under control . this motivates us to combine both representations in a single quark model with _ next - to minimal flavor violation _ , harnessing their complementing strengths while keeping as much flavor symmetry as possible . two copies of @xmath71 with bulk masses @xmath74 make up the first two generations , while the third generation is contained in the custodially protected @xmath124 with bulk masses @xmath125 . this protects the other quarks from the influence of the heavy top while enabling us to match all fermion couplings to experimental data . ( note that the top couplings are poorly constrained . ) the form of the respective ir dirac mass terms are given in equations ( [ e.lrsymirterm ] ) and ( [ e.custodialirterm ] ) . we impose the flavor symmetry @xmath126 in the bulk , which is broken down to @xmath127 on the ir brane . this means that the first two generations have the same ir dirac mass @xmath75 and bulk masses . the third generation has the ir dirac masses @xmath106 for the @xmath67 triplet ( which includes the bottom ) and @xmath105 for the singlet ( which supplies mass to the top ) . to provide flavor mixing and differentiate the quark masses of the first two generations , we must introduce general hermitian @xmath128 kinetic mixing matrices @xmath129 and @xmath110 as in eq . ( [ e.kineticterm ] ) . therefore , the flavor symmetry on the uv brane is @xmath79 ( where the third @xmath65 is contained in @xmath101 ) . we can see immediately that there will be fcncs in this model . the flavor symmetry is explicitly broken by choosing a different quark representation for the third generation . if we switch off the charged currents , we only have @xmath0 symmetries available , which are not enough to diagonalize the kinetic mixing matrices on the uv brane . however , as we will see , this partial symmetry is enough to force all mixing to go ` through the third generation ' and suppress 12-mixings . we can solve the bulk equations with the appropriate bc s to compute the entire kk tower of fermion wave functions . after integrating out the 5@xmath130 dimension , we end up with a 4d action containing the following terms ( using matrix notation in flavor / kk space ) : @xmath131 where @xmath132 is a gauge boson kk index , @xmath133 is a diagonal matrix of the right - handed fermion wave functions evaluated at @xmath134 , and @xmath83 is the uv brane kinetic mixing matrix . to go to 4d mass basis , we must first diagonalize and canonically normalize the kinetic mixing term by rotating the rh spinors with a hermitian matrix @xmath135 . once the kinetic terms are flavor singlets we can diagonalize the mass matrices with the usual biunitary transformation . we will always distinguish quantities in the physical basis with a ` mass ' superscript from quantities in the original flavor basis without superscript . the quark spinors in the mass basis are related to the flavor basis in the following way : @xmath136 applying this transformation to the the mass terms , the left / right - handed neutral couplings ( denoted generically by @xmath137/@xmath138 ) , and the left / right - handed @xmath2 couplings , we get : @xmath139 there is a very useful relation which we will need later . we simply write out @xmath140 ( since @xmath141 is diagonal ) . keeping in mind that the @xmath135 matrices are hermitian and @xmath142 , we find @xmath143 the exotic x - quark with charge @xmath1 is an interesting experimental signature of our model . its mass is roughly half a @xmath144 and it couples to the top via charged - current interactions ( in the flavor basis ) with coupling strength comparable to but generically less than @xmath145 . the coupling in the mass basis is @xmath146 detection could be possible at the lhc with less than @xmath147 pb@xmath148 of integrated luminosity @xcite . it is not hard to see why this model can satisfy electroweak precision constraints . the heavy top mass does not influence the other quarks , and the correct bottom couplings can be achieved by moving @xmath149 and @xmath88 close to the ir brane , while the @xmath150 is close to the uv @xcite . the top couplings will deviate from the sm value , but this is acceptable since it is poorly constrained experimentally . the first- and second - generation couplings can be made to agree with the sm by adjusting @xmath151 , and we have enough freedom to choose ir dirac masses and uv kinetic terms to generate all the different quark masses and mixings . it is worth noting that the @xmath152 bulk masses can take on a range of values , due to the effect of kk - mixing which we will discuss in section [ ss.zeromode ] . we explicitly demonstrated ewpd compliance using two different numerical calculations . in the first , we assumed that there is no flavor mixing and absorbed the diagonal boundary terms into bc s . in the second there was flavor mixing , and we followed the procedure of section [ ss.4dmassbasis ] : using the zero mode approximation in which the boundary terms act as mixing terms between zero modes and kk modes . one of the canonical signatures of higgsless models are light gauge kk - modes with a mass of @xmath153 . this is low enough to warrant closer inspection of current cdf bounds @xcite to make sure our model is not already excluded . the cdf searches for heavy gauge bosons focus on resonant pair production processes of the form ( light quark pair ) @xmath154 ( heavy gauge boson ) @xmath154 ( some fermion pair , e.g. @xmath155 ) . assuming that the coupling to the heavy gauge boson is the same as to the sm counterpart for both the initial and final fermion states , the cdf bounds are @xmath156 . however , those bounds must be adjusted for our model since the coupling of gauge kk modes is very suppressed for the first two quark generations , and somewhat enhanced for the third generation . @xmath157 since the light left - handed quarks are not uv - localized , their couplings depend sensitively on the bulk masses and can be very small . leptons in our model would have similar couplings to the light quarks . it is clear that the coupling suppression increases the @xmath158 bounds from leptonic and @xmath159-channel searches way beyond our kk - scale of 700 gev . due to low @xmath160-detection efficiencies , the @xmath161-channel also does not supply a meaningful @xmath162 bound @xcite . only the constraints on @xmath163 from @xcite require closer inspection . their analysis assumed vector - like couplings to @xmath164 which were parameterized as @xmath165 and @xmath166 . the bounds on @xmath163 depend on @xmath167 and the width @xmath168 of @xmath164 . if we assume that we can use those bounds for our _ chiral _ couplings by simply averaging and setting @xmath169 , we can extract an approximate bound of @xmath170 on the width of our kk - gluon if its mass is @xmath171 . we have not calculated the width of the @xmath164 since it depends on several parameters that are not completely fixed in our model , but @xmath172 is not an atypical value for rs kk - gluons , see for example @xcite . furthermore , we can also decrease @xmath173 by another factor of @xmath174 by taking into account 1-loop rge corrections to the @xmath175 uv brane kinetic term , as outlined in section [ ss.gaugesector ] . this alleviates any concern that our model might be excluded by cdf bounds . however , the relatively light @xmath164 should certainly be detected at the lhc . each @xmath176 hermitian uv kinetic mixing matrix @xmath77 is defined by @xmath177 parameters , @xmath178 real elements and @xmath179 complex phases . for @xmath180 , this gives a total of 12 real parameters and 6 phases . we can always do an @xmath181 flavor rotation , which corresponds to eliminating unphysical parameters : it removes 1 angle and 3 phases . this leaves us with 11 real parameters and 3 phases , which includes the 6 quark masses . hence the parameters in the flavor sector are 6 quark masses , 5 mixing angles and 3 phases , as well as 3 ir dirac masses @xmath182 and @xmath106 . at this point a remark about the _ size _ of the uv kinetic terms is in order . the @xmath83 matrix elements will be very large , generically @xmath183 r$ ] , but this is no cause for concern . after canonically normalizing , the magnitude of the @xmath6 s will merely specify what fraction of the fermions lives on the uv brane ( i.e. is elementary in the ads / cft picture ) , and how much lives in the bulk ( i.e. is composite ) . in our model the right - handed quarks are almost entirely confined to the uv brane , only slightly dipping into the bulk to mix with the left - handed quarks on the ir brane and generate a dirac mass . the fermion zero mode approximation enormously simplifies matching and mixing calculations , and we can use it to gain a great deal of insight into the flavor protection mechanisms of our model . however , kk mixing is much more significant for higgsless models than for standard rs with multi - tev kk masses , so we need to investigate the range of validity of this approximation in detail if we want to trust our calculations . consider a simple toy - model with a single generation of quarks in the left - right symmetric representation eq . ( [ e.qlqrdef ] ) . there is a dirac mass term on the ir brane ( [ e.lrsymirterm ] ) and a uv boundary kinetic term for the right - handed fields ( [ e.kineticterm ] ) . focusing only on the _ left - handed _ fields for the moment , we can incorporate the ir dirac mass term into the @xmath184 boundary conditions of the 5d wave function profiles @xcite , eliminating kk - mixing on the ir brane : @xmath185 similarly for the down sector . we will assume that the errors are small and @xmath186 . if there was no ir mixing , @xmath187 would just be the zero mode @xmath188 ( i.e. @xmath189 from eq . ( [ e.zeromode ] ) with @xmath190 ) , and adding a small amount of mixing should not change the shape of that waveform significantly . the new mode that appears due to mixing is @xmath191 , and its shape is also independent of the size of a small mass . hence it should be the zero mode that is normally projected out when the bc s do not include any mixing , i.e. @xmath192 ( which is different from the usual rh zero mode @xmath193 ) . a simple ansatz to approximately solve the exact bcs is therefore @xmath194 using the fermion normalization condition @xmath195 as well as eq . ( [ e.irbc ] ) , we can solve for the coefficients @xmath196 and @xmath90 . assuming the error is small , one obtains @xmath197 we can now estimate the deviation of a typical coupling to gauge boson @xmath198 compared to the zero mode approximation : @xmath199 \int dz \left ( \frac{r}{z } \right)^4 |g^0_{u_l}|^2 g_5 \psi\ ] ] the correction due to including the @xmath138 is at most of similar order , and in fact much smaller for electroweak couplings since @xmath200 near the ir brane and @xmath192 is extremely ir localized . hence the zero mode approximation _ overestimates _ left - handed couplings by roughly @xmath201 which is a _ relative _ error independent of the gauge charge . by a similar procedure we obtain the error for the right - handed couplings . it is simplest to _ not _ include the uv brane term in the bcs and simply renormalize the bulk wave function . thus we find that the zero mode approximation _ overestimates _ right - handed couplings by @xmath202 which is negligible unless the uv term is very small . both of these error estimates have been confirmed numerically . using eq . ( [ e.wmass ] ) we can express them as @xmath203 to demonstrate how significant those errors can be , we computed the couplings in our toy model numerically for the first quark generation only , incorporating _ both _ uv and ir brane terms into boundary conditions . [ f.1stgensmmatchdiffros ] shows , we find that one can now have _ both _ @xmath65 and @xmath66 localized near the uv brane without any @xmath31-parameter by turning up the value of @xmath75 ! thus one can get around the canonical assumption that fermions in higgsless rs models must be almost flat and near the ir brane . if there is significant kk mixing on the ir brane , the fermions can be uv localized . as we discussed in section [ ss.cdf ] , we have used more accurate calculational methods to confirm that our model can satisfy electroweak precision constraints . the zero mode calculation without fermion kk - modes will only be used to estimate fcnc suppression . we must expect percent - level errors for all diagonal couplings and masses that do not involve the top , and @xmath204 errors for the top mass and off - diagonal couplings in the up sector . going from the zero mode calculation to full kk mixing preserves the general mixing hierarchy , but we would have to restore a full ckm match by adjusting the up - sector mixing angles by order unity . this level of accuracy is sufficient to estimate the tightly constrained down - sector fcncs to within a few percent . our estimate for d - mixing , on the other hand , will only be valid up to a factor of order unity , but this is enough to demonstrate our flavor suppression mechanism . we will now analyze the flavor - protection mechanisms of the nmfv model in the framework of the zero - mode calculation . our first task is to find the correct uv - localized right - handed kinetic mixing matrices @xmath129 and @xmath110 which reproduce the 4d ckm matrix . after obtaining a tree - level match to the standard model we proceed to find the off - diagonal neutral couplings which give rise to dangerous tree - level fcncs . throughout this section we will drop the @xmath205 , @xmath64 subscripts whenever the derivations for the up- and down - sectors are identical . the problem of matching the zero mode calculation to the standard model factorizes into four steps : find bulk masses @xmath206 and @xmath117 which give the correct quark couplings ( noting that some deviation from sm is permissable for the top ) . choose a set of unitary matrices @xmath207 , @xmath208 that match @xmath209 to the experimental value of @xmath210 . there is a lot of freedom to choose mixing matrices here , and we shall address it in section [ s.numerical ] . choose ir dirac masses @xmath211 and @xmath105 which are at least big enough to supply the charm , bottom and top masses , and bigger if we want to confine the rh quarks to the uv brane . find the @xmath77 kinetic matrices which are required to produce the sm masses and mixings . the fourth step works as follows . the total kinetic term is @xmath212 ( see eqn . [ e.matrixnotations ] ) , where @xmath213 is the diagonal matrix of rh quark zero modes evaluated on the uv brane ( note that @xmath214 due to flavor symmetry ) . using eq . ( [ e.getkappa ] ) we can express @xmath215 in terms of quantities that we know : @xmath216 hence we obtain an expression for the uv brane kinetic mixing matrix : @xmath217 f_r^{-1}\ ] ] in the flavor basis , we can always write any left - handed neutral coupling as : @xmath218 where @xmath219 is the wave function of the gauge boson that the fermion couples to , evaluated on the uv brane and multiplied by the appropriate gauge coupling ( in the ads / cft picture , this is the elementary part of the gauge boson ) . all the flavor non - universalities are contained in the diagonal matrix @xmath220 which comes from bulk overlap integrals with the fermions ( and corresponds to the composite gauge boson coupling ) . for a right - handed neutral coupling we also have the contribution from the uv kinetic term , which gives @xmath221 the form of @xmath6 is known from eq . ( [ e.findk ] ) , and we can obtain the right - handed rotation matrix from eq . ( [ e.basischange ] ) @xmath222 the left - handed couplings just rotate by @xmath223 . this is all the information we need to transform the couplings into the physical basis , where the mass matrix is @xmath224 . before we do that , however , it is useful to parameterize the ir dirac masses in terms of roughly how large we want the uv kinetic terms to be , i.e. how strongly we want to confine the rh quarks to the uv brane . in the flavor basis , the mass matrix is @xmath225 where the flavor symmetry forces the first two terms to be the same and @xmath226 @xmath227 corresponds to choosing the minimal ir dirac mass ( and a correspondingly minimal uv kinetic term ) which can generate the @xmath228 4d mass . @xmath229 simply corresponds to increasing the ir dirac mass by that factor , which also increases the uv kinetic term in order to keep the quark mass constant . this localizes the rh quark on the uv brane . now we can apply the basis transformations @xmath223 and @xmath230 to eqns . ( [ e.glflavorbasis ] ) and ( [ e.grflavorbasis ] ) . we obtain expressions for the physical 4d neutral couplings : @xmath231 all the off - diagonal terms come from the flavor non - universal bulk part of the coupling , rotated by the appropriate transformation matrix . let us now find explicit expressions for these off - diagonal neutral coupling elements . our flavor symmetry imposes @xmath232 and @xmath233 . the most general form ( ignoring phases ) that @xmath223 can take is : @xmath234 if we assume completely anarchic uv mixing , the fcnc s will generically be too large . however , if one assumes cabibbo - type hierarchies for both @xmath207 and @xmath208 mixing matrices , @xmath235 then the flavor - changing effects will get an additional cabibbo - suppression . this amounts to assuming that there is some systematic uv physics generating the mixing hierarchies . substituting all this into eq . ( [ e.glgr ] ) and expanding to lowest order in @xmath173 , we obtain simple expressions for the off - diagonal neutral couplings . in the down sector , @xmath236 u_{ld}^{ji } \\ \\ { g_{rd , ij}^\mathrm{mass } } & \approx & \left[\dfrac{g_{rd}^{\mathrm{bulk}3}}{\rho_{b}^2 m_{b}^2 } - \dfrac{g_{rd}^{\mathrm{bulk}1}}{\rho_c^2 m_c^2}\right ] m_d^j m_d^i u_{ld}^{ji } \medskip \end{array } & \phantom{spacer } & \begin{array}{l } \mbox{where } i < j\\ \\ \mbox{and for } ( i , j ) = ( 1,2 ) , \medskip \\ u_{ld}^{ji } \rightarrow u_{ld}^{32 } u_{ld}^{31}. \end{array } \end{array}\ ] ] for the up sector just change subscripts @xmath237 and replace @xmath238 by @xmath239 . the flavor protection of our model is now apparent . firstly , the surviving _ flavor symmetry _ between the first two generations forces all the mixing to go through the third generation ( hence nmfv ) . this is vital to push @xmath5 and @xmath6 mixing below the stringent experimental bounds . secondly , since we are free to increase @xmath240 , there is an _ rs - gim - like flavor suppression mechanism _ for the right - handed fermion couplings . this is due to the kinetic mixing terms , which confine the right - handed quarks to the uv brane and suppress the bulk contributions to the couplings , which are the source of flavor violation . finally , since the charged - current mixing matrix is made up of both the up- and down - sector mixing matrices , we have some freedom to ` divide up the mixing ' between the two sectors and reduce fcncs for each sector accordingly . we are now in a position to estimate the fcncs for our nmfv model and compare them to experimental bounds , as well as to a standard rs setup with a kk scale of @xmath241 and anarchic yukawa couplings , where the only flavor protection is due to rs - gim ( see for example @xcite ) . @xmath242 fcncs are mediated by the 4-fermi operators @xmath243 . for the up sector ( identically for the down sector ) , the relevant terms in the effective lagrangian are given by @xmath244 where greek letters denote flavor indices and @xmath245 are color indices ( if not shown , then color is contracted inside brackets ) . we will compute the fcnc operators by integrating out the massive gauge bosons , then compare them to utfit bounds @xcite . the most relevant constraints come from meson mixing processes , i.e. @xmath5 mixing in the up sector and @xmath6 , @xmath246 and @xmath247 mixing in the down sector : @xmath248 ( similarly for @xmath249 ) . integrating out the massive gauge bosons in our model , we obtain : @xmath250 where `` kk '' indicates that we sum over gauge kk modes , including the sm @xmath3 boson . for each operator , we define a suppression scale @xmath251 by @xmath252 . we want to write down expressions for the suppression scales of the @xmath243 operators due to contributions of the first gluon kk - mode , which dominate if @xmath3-couplings are matched to the sm and ( almost ) flavor universal . the mass of the first kk - gluon is given to @xmath253 accuracy by @xmath254 where we have used eq . ( [ e.wmass ] ) , and @xmath255 is the first root of the bessel function @xmath256 . the bulk part of its coupling to a left - handed zero mode is approximately @xmath257 where @xmath258 is an @xmath204 numerical correction factor @xcite in @xcite is somewhat improved by replacing @xmath259 . gauge matching sets @xmath260 , and we find that numerically , @xmath261 . now put everything together by substituting eq . ( [ e.gluonkkcoupling ] ) into eq . ( [ e.offdiagnc ] ) , and using those couplings in eq . ( [ e.4fermiops ] ) . we obtain the following expressions for the down - sector flavor suppression scales : @xmath262\right|^{-1}\\ \nonumber \lambda_4^{(d)ij } & \approx \dfrac{2 m_w}{u^{ij}_{ld}}\left|m_d^i m_d^j \left[f(c_{q_l } ) - f(c_l)\right]\left[\frac{f(-c_{q_r})}{m_c^2 \rho_c^2 } - \frac{f(-c_{b_r})}{m_b^2 \rho_b^2}\right]\right|^{-1/2}\\ \lambda_5^{(d)ij } & = \sqrt{3 } \lambda_4^{(d)ij}\end{aligned}\ ] ] where we replace @xmath263 . for the up sector just change subscripts @xmath237 and replace @xmath264 by @xmath265 . to see that our model has sufficient flavor protection to satisfy fcnc constraints we plug in some typical numbers and compare them to the rs - gim suppression and the utfit experimental flavor bounds on bsm fcnc contributions @xcite . the results are shown in table [ t.allfcnc ] and demonstrate why we need all our suppression mechanisms . the rs - gim - like mechanism for the right - handed couplings , together with the flavor symmetry , ensures that the @xmath266 and @xmath267 operators are easily below bounds , a great improvement on traditional rs - gim alone . the suppression scales of the @xmath268 operators are set by @xmath269 , which are only cabibbo suppressed , with the direct 12 contribution forbidden by the @xmath0 flavor symmetry . this is another reason why we need the flavor symmetry breaking it would increase 12-mixing by @xmath270 , immediately violating bounds . even with the flavor symmetry , the @xmath268 operators are close to bounds and the greatest source of angle constraints indeed , we can see that most of the mixing will have to be in the up - sector . .we compare lower bounds on the np flavor scale @xmath271 ( all in tev ) for arbitrary np flavor structure from the utfit collaboration @xcite to the effective suppression scale in rs - gim @xcite and our higgsless nmfv model , see eq . ( [ e.supscale ] ) . in this rs - gim model , @xmath272 , @xmath273 and @xmath274 , with a kk scale of @xmath275 . for the higgsless model @xmath41 determines a kk scale @xmath276 . setting @xmath277 gives @xmath278 , and @xmath279 matches the couplings for the first two generations to the sm . a third generation ewpd match is most easily obtained for @xmath280 and @xmath281 . to satisfy the flavor bounds , we need to push more mixing into the up - sector by setting @xmath282 and @xmath283 . [ cols="<,<,<,<,<",options="header " , ] we will calculate fcncs due to tree - level @xmath284-exchange , for one million different down - sector mixings per scan . we can parametrize the ckm mixing matrix as @xmath285 where , for example , @xmath286 would satisfy the pdg constraints on @xmath287 @xcite . naively , we would think that we can obtain the correct ckm matrix by defining our up- and down - sector lh rotations as @xmath288 and letting @xmath289 be an arbitrary unitary rotation matrix which gives the down - sector mixing . this is sufficient for this scan , even though it only gives an order unity estimate of the up - sector mixing angles . in this analysis we shall also ignore phases , since we are after a scan of the _ magnitudes _ of the possible mixing matrices , and for our purposes we define @xmath290 ( otherwise we could never cancel this matrix with a real rotation , introducing an up - mixing bias into our scan ) . to avoid obviously large fcncs , we will make the assumption that the mixing angles of @xmath289 have a _ natural size _ comparable to those of the final @xmath287 mixing matrix . we parametrize @xmath289 with angle - coordinates @xmath291 : @xmath292 note that @xmath293 and @xmath294 put all the mixing into the up- and down - sector respectively , so to avoid a bias in our scan we define the range of the angle coordinates to be @xmath295 . once we determine which points in angle - space satisfy fcncs in the zero mode calculation , we re - check those points using a full kk calculation . as we can see from the similar plots in fig . [ f.scanfit ] , the choice of bulk masses does not have a great effect on the nature of constraints on the down - mixing angles . this is expected since the @xmath268 operators , which are the greatest bottleneck , are only weakly dependent on @xmath296 the dominant contribution comes from the large @xmath297 , see eq . ( [ e.supscale ] ) . in eliminating points which do not satisfy fcnc bounds with full kk mixing , we only loose a few percent of points in each scan . the zero mode calculation is therefore sufficient for estimating the angle constraints . the fcnc bounds impose entirely systematic constraints on the down - sector mixing angles . this can be seen from fig . [ f.scanfitresult44bslices ] , where we take slices at different points on the @xmath90-axis ( i.e. @xmath298 ) and project them onto an @xmath299-plane . a point in angle - space satisfies fcnc bounds if and only if it lies within a well - defined sub - volume , i.e. the constraints are systematic . assuming cabibbo - type mixing , the good points occupy @xmath300 of the total angle space . this is not really ` tuning ' in the usual sense , it merely means that whatever uv - scale mechanism generates the mixings should give a somewhat larger mixing in the up - sector than in the down - sector . we note that while @xmath301 and @xmath302 are correlated , their range is fairly unconstrained , whereas @xmath298 must fall within strict limits to satisfy @xmath303 constraints . roughly speaking , less than half the 23-mixing is allowed to be in the down sector . we can conclude that our higgsless nmfv model should have no trouble satisfying fcnc bounds as long as certain systematic constraints on the down - sector mixing angles are met . [ s.conclusion ] we examined various possibilities for higgsless rs model - building , and constructed a model with next - to - minimal flavor violation satisfying tree - level electroweak precision and meson - mixing constraints , as well as cdf bounds . the theory has a sufficiently high cutoff of @xmath304 to unitarize @xmath39-scattering at lhc energies , the third generation is in the custodial quark representation to protect the bottom couplings , and a combination of flavor symmetries and uv confinement of the right - handed quarks suppress fcncs . using numerical scans , we were able to demonstrate that our model can satisfy flavor bounds as long as the down - sector mixing angles are cabibbo - type and satisfy systematic constraints . we also found quantitative error estimates for the zero mode approximation , which are important for rs model - building with a low kk scale . this model has distinctive experimental signatures , allowing it to be excluded early on at the lhc . apart from the absence of the higgs , the usual higgsless rs signals include @xcite a relatively light @xmath164 with a mass below 1 tev , as well as @xmath43 and @xmath305 which are harder to detect ( see section [ ss.cdf ] ) . more specific to our setup is an exotic @xmath95-quark with charge @xmath1 and a mass of @xmath306 , which could be detected with less than @xmath147 pb@xmath148 of data @xcite . the nmfv model also predicts non - zero correlated flavor - changing neutral currents , which lie relatively close to current experimental bounds and would be detected in the next generation of flavor experiments . we are grateful to andreas weiler , kaustubh agashe and yuval grossman for many useful discussions and comments on the manuscript , and in particular we thank andreas weiler for his help with calculating the rge evolution of the utfit flavor bounds . this research has been supported in part by the nsf grant number nsf - phy-0757868 . was also supported in part by a u.s .- israeli bsf grant . c. csaki , c. grojean , h. murayama , l. pilo and j. terning , phys . d * 69 * , 055006 ( 2004 ) [ arxiv : hep - ph/0305237 ] . c. csaki , c. grojean , l. pilo and j. terning , phys . lett . * 92 * , 101802 ( 2004 ) [ arxiv : hep - ph/0308038 ] . r. s. chivukula , d. a. dicus , h. j. he and s. nandi , phys . b * 562 * , 109 ( 2003 ) [ arxiv : hep - ph/0302263 ] ; r. s. chivukula , e. h. simmons , h. j. he , m. kurachi and m. tanabashi , phys . d * 71 * , 115001 ( 2005 ) [ arxiv : hep - ph/0502162 ] . r. casalbuoni , s. de curtis and d. dominici , phys . d * 70 * ( 2004 ) 055010 hep - ph/0405188 ; r. s. chivukula , e. h. simmons , h. j. he , m. kurachi and m. tanabashi , phys . rev . d * 70 * ( 2004 ) 075008 hep - ph/0406077 ; h. georgi , phys . d * 71 * , 015016 ( 2005 ) hep - ph/0408067 ; r. sekhar chivukula , e. h. simmons , h. j. he , m. kurachi and m. tanabashi , phys . d * 71 * ( 2005 ) 035007 hep - ph/0410154 ; r. s. chivukula , b. coleppa , s. di chiara , e. h. simmons , h. j. he , m. kurachi and m. tanabashi , phys . d * 74 * , 075011 ( 2006 ) [ arxiv : hep - ph/0607124 ] . y. nomura , jhep * 0311 * , 050 ( 2003 ) [ arxiv : hep - ph/0309189 ] ; g. burdman and y. nomura , phys . d * 69 * , 115013 ( 2004 ) [ arxiv : hep - ph/0312247 ] ; r. barbieri , a. pomarol and r. rattazzi , phys . b * 591 * , 141 ( 2004 ) [ arxiv : hep - ph/0310285 ] h. davoudiasl , j. l. hewett , b. lillie and t. g. rizzo , phys . d * 70 * , 015006 ( 2004 ) [ arxiv : hep - ph/0312193 ] ; h. davoudiasl , j. l. hewett , b. lillie and t. g. rizzo , jhep * 0405 * , 015 ( 2004 ) [ arxiv : hep - ph/0403300 ] ; j. l. hewett , b. lillie and t. g. rizzo , jhep * 0410 * , 014 ( 2004 ) [ arxiv : hep - ph/0407059 ] . g. cacciapaglia , c. csaki , c. grojean and j. terning , phys . d * 71 * , 035015 ( 2005 ) [ arxiv : hep - ph/0409126 ] ; r. foadi , s. gopalakrishna and c. schmidt , phys . b * 606 * , 157 ( 2005 ) [ arxiv : hep - ph/0409266 ] ; r. s. chivukula , e. h. simmons , h. j. he , m. kurachi and m. tanabashi , phys . d * 72 * , 015008 ( 2005 ) [ arxiv : hep - ph/0504114 ] ; r. casalbuoni , s. de curtis , d. dolce and d. dominici , phys . d * 71 * , 075015 ( 2005 ) [ arxiv : hep - ph/0502209 ] . y. grossman and m. neubert , phys . b * 474 * , 361 ( 2000 ) [ arxiv : hep - ph/9912408 ] ; t. gherghetta and a. pomarol , nucl . b * 586 * , 141 ( 2000 ) [ arxiv : hep - ph/0003129 ] . g. burdman , phys . d * 66 * , 076003 ( 2002 ) [ arxiv : hep - ph/0205329 ] ; phys . b * 590 * , 86 ( 2004 ) [ arxiv : hep - ph/0310144 ] . s. j. huber , nucl . b * 666 * , 269 ( 2003 ) [ arxiv : hep - ph/0303183 ] ; s. j. huber and q. shafi , phys . b * 498 * , 256 ( 2001 ) [ arxiv : hep - ph/0010195 ] . k. agashe , g. perez and a. soni , phys . d * 71 * , 016002 ( 2005 ) [ arxiv : hep - ph/0408134 ] ; phys . lett . * 93 * , 201804 ( 2004 ) [ arxiv : hep - ph/0406101 ] . k. agashe , m. papucci , g. perez and d. pirjol , arxiv : hep - ph/0509117 ; z. ligeti , m. papucci and g. perez , phys . * 97 * , 101801 ( 2006 ) [ arxiv : hep - ph/0604112 ] . c. csaki , a. falkowski and a. weiler , jhep * 0809 * , 008 ( 2008 ) [ arxiv:0804.1954 [ hep - ph ] ] . s. casagrande , f. goertz , u. haisch , m. neubert and t. pfoh , jhep * 0810 * , 094 ( 2008 ) [ arxiv:0807.4937 [ hep - ph ] ] ; m. blanke , a. j. buras , b. duling , s. gori and a. weiler , jhep * 0903 * , 001 ( 2009 ) [ arxiv:0809.1073 [ hep - ph ] ] ; m. blanke , a. j. buras , b. duling , k. gemmler and s. gori , jhep * 0903 * , 108 ( 2009 ) [ arxiv:0812.3803 [ hep - ph ] ] ; m. e. albrecht , m. blanke , a. j. buras , b. duling and k. gemmler , arxiv:0903.2415 [ hep - ph ] . a. l. fitzpatrick , g. perez and l. randall , arxiv:0710.1869 [ hep - ph ] ; j. santiago , jhep * 0812 * , 046 ( 2008 ) [ arxiv:0806.1230 [ hep - ph ] ] ; c. csaki , a. falkowski and a. weiler , arxiv:0806.3757 [ hep - ph ] ; k. agashe , a. azatov and l. zhu , arxiv:0810.1016 [ hep - ph ] . k. agashe , a. e. blechman and f. petriello , phys . d * 74 * , 053011 ( 2006 ) [ arxiv : hep - ph/0606021 ] ; m. c. chen and h. b. yu , phys . lett . b * 672 * , 253 ( 2009 ) [ arxiv:0804.2503 [ hep - ph ] ] ; g. perez and l. randall , jhep * 0901 * , 077 ( 2009 ) [ arxiv:0805.4652 [ hep - ph ] ] ; c. csaki , c. delaunay , c. grojean and y. grossman , jhep * 0810 * , 055 ( 2008 ) [ arxiv:0806.0356 [ hep - ph ] ] ; k. agashe , t. okui and r. sundrum , phys . rev . lett . * 102 * , 101801 ( 2009 ) [ arxiv:0810.1277 [ hep - ph ] ] . k. agashe , a. delgado , m. j. may and r. sundrum , jhep * 0308 * , 050 ( 2003 ) [ arxiv : hep - ph/0308036 ] . g. cacciapaglia , c. csaki , c. grojean and j. terning , phys . d * 70 * , 075014 ( 2004 ) [ arxiv : hep - ph/0401160 ] . l. randall and m. d. schwartz , jhep * 0111 * , 003 ( 2001 ) [ arxiv : hep - th/0108114 ] ; w. d. goldberger and i. z. rothstein , phys . * 89 * , 131601 ( 2002 ) [ arxiv : hep - th/0204160 ] ; phys . d * 68 * , 125011 ( 2003 ) [ arxiv : hep - th/0208060 ] ; phys . rev . d * 68 * , 125012 ( 2003 ) [ arxiv : hep - ph/0303158 ] ; k. w. choi and i. w. kim , phys . d * 67 * , 045005 ( 2003 ) [ arxiv : hep - th/0208071 ] ; k. agashe , a. delgado and r. sundrum , nucl . b * 643 * , 172 ( 2002 ) [ arxiv : hep - ph/0206099 ] ; annals phys . * 304 * , 145 ( 2003 ) [ arxiv : hep - ph/0212028 ] ; r. contino , p. creminelli and e. trincherini , jhep * 0210 * , 029 ( 2002 ) [ arxiv : hep - th/0208002 ] . t. aaltonen _ et al . _ [ cdf collaboration ] , arxiv:0811.0053 [ hep - ex ] ; t. aaltonen _ et al . _ [ cdf collaboration ] , arxiv:0902.3276 [ hep - ex ] ; a. abulencia _ et al . _ [ cdf collaboration ] , phys . lett . * 95 * , 252001 ( 2005 ) [ arxiv : hep - ex/0507104 ] ; a. abulencia _ et al . _ [ cdf collaboration ] , phys . rev . * 96 * , 211801 ( 2006 ) [ arxiv : hep - ex/0602045 ] ; t. aaltonen _ et al . _ [ cdf collaboration ] , phys . lett . * 99 * , 171802 ( 2007 ) [ arxiv:0707.2524 [ hep - ex ] ] . t. aaltonen _ et al . _ [ cdf collaboration ] , phys . d * 77 * , 051102 ( 2008 ) [ arxiv:0710.5335 [ hep - ex ] ] . cdf collaboration , `` a search for massive gluon decaying to top pair in lepton+jet channel '' [ www-cdf.fnal.gov/physics/exotic/exotic.html ] . b. lillie , l. randall and l. t. wang , jhep * 0709 * , 074 ( 2007 ) [ arxiv : hep - ph/0701166 ] . j. a. bagger , k. t. matchev and r. j. zhang , phys . b * 412 * , 77 ( 1997 ) [ arxiv : hep - ph/9707225 ] ; m. ciuchini , e. franco , v. lubicz , g. martinelli , i. scimemi and l. silvestrini , nucl . b * 523 * ( 1998 ) 501 [ arxiv : hep - ph/9711402 ] ; a. j. buras , s. jager and j. urban , nucl . b * 605 * ( 2001 ) 600 [ arxiv : hep - ph/0102316 ] . a. birkedal , k. matchev and m. perelstein , phys . lett . * 94 * , 191803 ( 2005 ) [ arxiv : hep - ph/0412278 ] ; c. englert , b. jager and d. zeppenfeld , jhep * 0903 * , 060 ( 2009 ) [ arxiv:0812.2564 [ hep - ph ] ] ; a. martin and v. sanz , to appear .

the second - quantization of a scalar field in an open cavity is formulated , from first principles , in terms of the quasinormal modes ( qnms ) , which are the eigensolutions of the evolution equation that decay exponentially in time as energy leaks to the outside . this formulation provides a description involving the cavity degrees of freedom only , with the outside acting as a ( thermal or driven ) source . thermal correlation functions and cavity feynman propagators are thus expressed in terms of the qnms , labeled by a discrete index rather than a continuous momentum . single - resonance domination of the density of states and the spontaneous decay rate is then given a proper foundation . this is a first essential step towards the application of qnms to cavity qed phenomena , to be reported elsewhere .

quant-ph9803014

in this paper we are concerned with quantum fields in _ open _ cavities the obvious example of ultimate interest would be electromagnetic fields in optical cavities @xcite and the associated problem of cavity quantum electrodynamics ( cqed ) . such systems are open because energy leaks to the outside ( e.g. , via output coupling ) , and as dissipative systems can not normally be quantized on their own @xcite ; rather , one must also consider the bath into which energy escapes , so that the total universe is conservative . thus such cavities , say of linear dimension @xmath0 , can be embedded in a universe of dimension @xmath1 . one can then quantize on the modes of the universe , which are labeled by a wavenumber @xmath2 spaced by @xmath3 . the field quanta are then created or destroyed by operators @xmath4 and @xmath5 , and higher - order processes involve integrals @xmath6 . on the other hand , these cavities often have a very small amount of leakage , characterized by a parameter @xmath7 , where the quality factor of the cavity can be as high as @xmath8 or more . if this is the case , the intuition developed from a _ closed _ cavity , also of length @xmath0 , should be relevant . a closed cavity is a conservative system , with normal modes ( nms ) labeled by a _ index @xmath9 , where the wavenumber is @xmath10 , @xmath11 . field quanta in such a closed cavity are created and destroyed by operators @xmath12 and @xmath13 , and higher - order processes involve discrete sums @xmath14 . can quantum fields in an _ open _ cavity be described in a similar way in terms of discrete modes and the corresponding operators ? if this is possible , computations will be simplified and will correspond to physical intuition , with each term @xmath15 associated with a cavity mode " . the connection with the limit of a closed cavity ( @xmath16 ) would also become manifest . quantization of a closed system relies on its nms ; the counterparts in an open system are the quasinormal modes ( qnms ) , which are again factorized solutions @xmath17 where @xmath18 because of the loss of energy . each qnm corresponds to a resonance , with a width @xmath19 . the purpose of this paper , in short , is to develop a formalism whereby field quantization can be implemented in terms of these qnms , and to define and study operators @xmath12 , @xmath20 for these modes . specifically , one wishes to express field correlation functions , feynman propagators and other quantities in terms of the qnms . the dissipative nature of the system is then contained in the qnms themselves . the advantages for cqed would be obvious . the simplest phenomenon to which such a formalism can apply is the well - known enhancement ( or suppression ) of spontaneous decay rates when the emitted radiation falls on ( or between ) resonances @xcite . resonance domination of these processes has been discussed starting with the heuristic argument due to purcell @xcite . he proposed that the fermi golden rule should be generalized : the density of states per unit volume , instead of the usual @xmath21 for vacuum ( where @xmath22 is the velocity of light ) , is to be replaced by @xmath23 for a @xmath24-fold degenerate qnm of width @xmath25 in a cavity of volume @xmath26 . this leads to an enhancement factor of @xmath27 for spontaneous emission on resonance , where @xmath28 is the wavelength of light emitted and @xmath29 is the quality factor of the cavity . the essence of this argument is that each resonance counts as one state , i.e. , in a suitable sense it carries unit weight . while intuitively plausible , this statement is difficult to justify formally since the entire concept of a state , i.e. , an nm , falls apart in an open system . however , this argument , and its many variants and extensions , would find natural expression in a framework that quantizes on the qnms , and we shall in particular show explicitly below that each resonance carries unit weight . in section [ class ] , the qnm expansion of _ classical _ fields outgoing from a cavity is reviewed . the classical results may be organized into two levels . first , under conditions to be specified , the green s function @xmath30 can be expanded in terms of qnms . second , one can try to expand the outgoing classical fields @xmath31 in terms of qnms , and to establish a linear space structure similar to that for conservative systems . in order to do so , it turns out to be necessary to make use of a two - component formalism , dealing with @xmath31 and the conjugate momentum @xmath32 at the same time . the linear space formalism is more elegant , but in its simplest form is limited to only 1 d. the quantum formalism can likewise be approached in two ways . in the first , which we shall call the green s function method ( section [ 1comp ] ) , one focuses on @xmath22-number correlation functions and propagators without explicitly expanding the field operator @xmath31 in terms of qnms . the key idea is that the retarded propagator @xmath33 of the quantum theory is exactly the same as the classical green s function @xmath30 , and the latter has a qnm expansion . once @xmath33 is obtained , it is straightforward to derive a qnm expansion for the correlation function @xmath34 as well . in the second , which we shall call the field expansion method , one tries to establish an expansion of the quantum field @xmath31 , in parallel with the linear space structure established for classical fields this allows us to interpret the expansion coefficients @xmath12 and @xmath13 , roughly speaking , as generalized creation and annihilation operators for the discrete qnms @xmath15 . however , before doing so , it has to be recognized that quantum fields can not be constrained by the outgoing wave condition for the simple reason that zero - point ( and thermal ) fluctuations must contain an incoming component . thus , the first step in developing this method , presented in section [ incoming ] , is to generalize the field expansion to handle _ incoming _ waves as well . with this generalization , one can then subject the fields to canonical quantization in section [ quant ] . this is done by starting with the universe , a closed hermitian system for which the quantization is unambiguously defined . then , in parallel with the usual removal of bath oscillators @xcite , the outside degrees of freedom are eliminated from the equations of motion @xcite . the results will be equations of motion and commutation relations for the _ discrete _ operators @xmath12 and @xmath13 , in which the effects of the outside bath are clearly displayed : the loss of energy of each mode by leakage , and the pumping of each mode by the thermal or quantum fluctuations from the outside . the formalism is then used to evaluate the correlation function @xmath34 in section [ correl ] , and the results are compared with those obtained from the green s function method . interestingly , the results appear to be different those derived from the green s function method contain a single sum @xmath35 over the qnms , while the field expansion method yields a double sum @xmath36 with off - diagonal terms . the two are , however , shown to be equal through an identity on @xmath33 . recalling that the expansion of the classical field is unique only when the second component @xmath32 is considered at the same time , we next show that the expansion of the correlation function @xmath34 is also unique if we consider @xmath32 as well , giving the non - diagonal form . the density of states @xmath37 , which is intimately related to the correlation function , is also expressed in terms of qnms ; in particular it is shown that up to corrections of @xmath38 , each resonance carries unit weight in the density of states . the results on the correlation function are then used , in section [ feyn ] , to evaluate and discuss the feynman propagator @xmath39 , which is the fundamental building block for cqed . again , equivalent diagonal and non - diagonal forms are obtained . particular attention is paid to the equal space propagator @xmath40 , whose imaginary part is related to the life - time of an excited atom placed at @xmath41 . this quantity is discussed in the approximation of domination by a single resonance , providing justification for purcell s heuristic argument @xcite on the enhancement of spontaneous decay rates . the advantage of using the non - diagonal expression is again emphasized . in section [ example ] , a very simple example is studied explicitly , and its correlation function and energy density are expressed in terms of a sum over qnm contributions . some final remarks are then given in section [ remarks ] . we stress that this paper is concerned entirely with _ free _ fields , either as a model of the free electromagnetic field in an optical cavity , or as the zeroth - order building blocks in an interacting theory , e.g. , the propagators as ingredients in higher - order feynman diagrams . the development of the interacting theory and its application to cqed phenomena will be given elsewhere @xcite . a partial account of the present theory has been given in @xcite . in this section , we summarize the qnm expansion for classical fields . in this paper , we deal with scalar fields in 1 d only . for closed , linear systems , eigenfunction expansions , based on the eigenfunctions or nms of their evolution operators , are a tool of vital importance in theoretical physics . however , open systems are not directly amenable to an nm analysis . examples of open systems include optical cavities @xcite , and finite regions of space near astrophysical objects , from which gravitational waves can escape @xcite . in these systems , any initial state decays in time , so stationary nms do not exist . as the simplest example , we shall be concerned with the scalar wave equation in one space dimension , @xmath42 studied in a cavity " @xmath43 , with the nodal boundary condition @xmath44 at one end but with the outgoing one @xmath45 at the other . the latter condition states that , just outside the cavity boundary , the field @xmath46 is an outgoing wave @xmath47 ; the condition is specified at @xmath48 because , as we shall see below , one is often concerned with models in which there is a singularity in @xmath49 at @xmath50 , leading to possible discontinuities in @xmath51 or @xmath52 . the boundary condition ( [ bc - right ] ) turns the cavity into a dissipative system that is leaky but not absorptive . the model ( [ wave - eq ] ) has been widely used as the scalar model of electromagnetism in an optical cavity @xcite . more physically , the 1-d nature is realized in fabry - perot cavities with lengths much smaller than the lateral dimensions , and the scalar field model is rigorously applicable to the transverse electric sector . for the system ( [ wave - eq])([bc - right ] ) , the eigensolutions , labeled by an index @xmath15 , have the form ( [ eigensol ] ) , with the qnms or cavity resonances @xmath53 satisfying @xmath54 \mskip0.5\thinmuskip f_j = 0 \label{def - qnm}\ ] ] and the boundary conditions ( [ bc - left ] ) , ( [ bc - right ] ) translating to @xmath55 it is easily verified that @xmath56 , so that the solution ( [ eigensol ] ) is indeed decaying in time . furthermore , the frequencies @xmath57 , which we suppose to be ordered according to increasing real parts , are spaced by @xmath58 , approximately as for a conservative system of size @xmath0 . with the possible exception of modes with @xmath59 , the qnms always occur in pairs with @xmath60 , and one can choose @xmath61 . the usual formalism concerning eigenfunction expansions relies on the hermiticity of the evolution operator , which only holds in the conservative case , and therefore breaks down for open systems . one possible resolution is to embed the cavity into a universe " @xmath62 with a nodal condition at @xmath63 , and study its nms the modes of the universe . namely , the system ( [ wave - eq])([bc - right ] ) is the restriction to @xmath64 of the problem ( [ wave - eq ] ) on the half line @xmath65 , if one sets @xmath66 and with the extension of the initial conditions to the outside " @xmath67 obeying @xmath68 . however , this has the obvious disadvantage of having to work with a continuum of states ( spaced by @xmath69 ) as opposed to the discrete set of eigenfunctions in the conservative case . besides , the closed system of equations ( [ wave - eq])([bc - right ] ) shows that even in the presence of dissipation the time evolution of the cavity can be studied _ without _ explicit reference to the outside , which is the principal goal of the program of second quantization of the open system . previous work ( see @xcite and references therein ) has established that , in spite of the lack of hermiticity in the conventional sense , an eigenfunction expansion for outgoing waves in classical open wave systems can be formulated in terms of the cavity degrees of freedom only , overcoming the disadvantages of the modes of the universe approach . the sufficient conditions for this qnm expansion are as follows . * the function @xmath49 has at least a step discontinuity at @xmath50 . this demarcates a well - defined cavity region . * the function @xmath49 has no tail outside the cavity , i.e. , @xmath70 . this condition ensures that the outside does not reflect outgoing waves back into the cavity , enabling the complete elimination of the environment from the equations of motion . these conditions are satisfied for optical cavities bounded from extended vacuum by a sharp material interface . the completeness of the qnms can be pursued at two levels . first , one shows that the retarded green s function of the system has the representation @xmath71 for @xmath72 and @xmath73 , where the @xmath53 s are normalized according to ( [ normal ] ) below . thus , the dynamics is contained entirely in the qnms , leading to a simple method of obtaining the retarded propagators and quantum correlation functions , as sketched in section [ 1comp ] . second , realizing that the wave equation ( [ wave - eq ] ) , like any classical hamiltonian problem , requires both position and momentum to be specified as initial data , one introduces function _ pairs _ @xmath74 with the conjugate momentum @xmath75 , so that for eigenfunctions @xmath76 . the space of all function pairs satisfying the boundary conditions ( [ bc - left ] ) and ( [ bc - right ] ) will be denoted as @xmath77 the space of outgoing waves . using these pairs , one can prove that the time evolution generated by ( [ greenfie ] ) can be recast in the form @xcite @xmath78 where the expansion coefficients are given by @xmath79 with @xmath80 and the _ bilinear scalar product _ @xmath81 + \phi(a^+)\mskip0.5\thinmuskip\chi(a^+ ) \right\ } \;. \label{scalarprod}\ ] ] by simply letting @xmath82 in ( [ evolve ] ) one arrives at a _ two - component expansion _ for an arbitrary @xmath83 @xcite . this expansion makes the completeness of the qnms manifest . the normalization used in ( [ greenfie ] ) to ( [ a_n1 ] ) can be concisely expressed as @xmath84 it is seen that ( [ normal ] ) in general is not real , underlining the difference between the product ( [ scalarprod ] ) and a conventional one involving complex conjugation . the fact that ( [ normal ] ) is bilinear also serves to establish a phase convention for the wavefunctions . upon introducing the two - component evolution operator @xmath85 the cavity evolution ( [ wave - eq ] ) can be written as @xmath86 , in striking analogy with quantum mechanics . in this notation , the definition ( [ def - qnm ] ) of @xmath53 takes the form @xmath87 . the operator @xmath88 can be shown to be symmetric with respect to the form ( [ scalarprod ] ) , i.e. , @xmath89 for any @xmath90 . this analog of hermiticity holds even though the system is not conservative . the symmetry of @xmath88 yields the orthogonality relation @xmath91 in an immediate transcription of the usual proof , leading to the uniqueness of the expansion . incidentally , an expansion such as ( [ evolve ] ) but involving the first component alone would not be unique . it is appropriate to contrast the two methods of approaching the classical theory , since they respectively underpin the two methods for dealing with the quantum case . the expansion of the green s function is easy to derive , and is readily generalized to higher dimensions ; however , in itself it does not lead to a _ unique _ expansion of the field , nor to concepts of orthogonality . the two - component approach based on ( [ evolve ] ) is more elegant , exhibits a deeper resemblance to conservative systems , and most importantly leads to a _ unique _ expansion in terms of orthogonal functions . the two - component expansion can in principle be generalized to higher dimensions by treating each angular momentum sector @xmath92 as a 1-d radial problem @xcite , but the degree of complexity increases with @xmath92 . thus each method has its own merits ; both will be pursued below , and the results compared . the quantum mechanics of the system is specified by the hamiltonian @xmath93 \label{def - h}\ ] ] together with the canonical equal - time commutation relation @xmath94=i\delta(x - y ) \;. \label{ccr}\ ] ] time evolution is then generated by means of the heisenberg equation @xmath95 $ ] for an arbitrary operator @xmath96 . however , instead of the equations of motion for the quantum operators , in this section we focus first on the retarded propagator @xmath97\rangle \ ; , \label{def - gr}\ ] ] in which @xmath31 is of course to be regarded as an operator , and @xmath98 denotes the expectation value at a finite temperature @xmath99 ; throughout we take @xmath100 . the central idea is that this propagator defined in terms of the quantum fields can be evaluated without explicitly introducing an expansion for the field operators @xmath31 , by simply noticing that @xmath101 is exactly the same as the green s function @xmath30 of the classical wave equation @xcite , which has the expression ( [ greenfie ] ) in terms of qnms . this relationship between @xmath33 and @xmath30 follows from the commutation relation ( [ ccr ] ) . in terms of @xmath33 , it is straightforward to compute the equilibrium correlation function @xmath102 we shall devote attention to @xmath34 , because the physical quantities of interest in quantum field theory can often be formulated in terms of correlation functions , at either zero or finite temperatures . for example , the casimir force is merely the vacuum expectation value of the electromagnetic stress tensor , which is an equal - time equal - space correlation function of two field operators . the spontaneous decay rate of an atom in an excited state is , in the golden rule approximation , related to the correlation function of two electric field operators . since the correlation function is related to the retarded propagator @xcite , one gets @xmath103 the real - time correlator can be obtained from ( [ f - d ] ) by contour integration , yielding @xmath104\;. \label{f - realt}\end{aligned}\ ] ] the first term in this formula is due to the qnm poles in @xmath105 ; the second term , which has no counterpart in @xmath101 , is caused by the matsubara poles in @xmath105 at frequencies @xmath106 . this very simple derivation has the advantage that it goes through in situations where the two - component formalism may be more complicated . in principle , physical quantities can be expressed in terms of @xmath34 bilinear quantities ( such as the energy density ) as linear combinations of @xmath34 and its derivatives , and other quantities involving products of @xmath34 s . for example , the energy density is @xmath107 \ , f(x , y , t ) \big|_{x = y , t=0}\;. \label{eden1}\ ] ] however , this quantity is divergent . subtracting off the zero - point , we consider @xmath108 \ , f_{\rm s}(x , y , t ) \big|_{x = y , t=0 } \label{edeb2}\end{aligned}\ ] ] in terms of the subtracted correlation function @xmath109 , where @xmath110 . the limit @xmath111 is best taken in ( [ f - d ] ) prior to fourier inversion . we further make use of the expansion for @xmath112 to get @xmath113 + \frac{i}{\beta } \sum_{m=1}^\infty \frac{\mu_m e^{-\mu_m t } } { \mu_m^2 + \omega_j^2 } \nonumber\\ & & -{}\frac{i}{4\pi}e^{i\omega_jt } { \rm e}_1(i\omega_jt ) -{}\frac{i}{4\pi}e^{-i\omega_jt } { \rm e}_1(-i\omega_jt ) \ ; , \label{fr2-realt}\end{aligned}\ ] ] where this and subsequent formulas for @xmath114 are written only for @xmath115 . @xmath116 is the exponential integral function @xmath117 in which the integration contour is defined not to pass through the origin and the negative real axis ; on that semi - axis , the function is defined as the principal value . we further define @xmath118 . alternatively , for greater formal similarity to the conservative case , ( [ fr2-realt ] ) can be rewritten as @xmath119 \;,\label{fs - realt } \\ \tilde{c}_j(t)&= & \frac{e^{-i\omega_j t}}{2(e^{\beta\omega_j}-1 ) } + \frac{i}{\beta } \sum_{m=1}^\infty \frac{\mu_m e^{-\mu_m t } } { \mu_m^2 + \omega_j^2 } \nonumber\\ & & -{}\frac{i}{4\pi}e^{i\omega_jt } { \rm e}_1(i\omega_jt ) -{}\frac{i}{4\pi}e^{-i\omega_jt } { \rm e}_1(-i\omega_jt ) \;.\end{aligned}\ ] ] the prime on the sum in ( [ fs - realt ] ) signifies that terms with @xmath120 are to be taken with weight @xmath121 . the actual evaluation of @xmath122 needs some care in the @xmath123 part of the sums . these details , and the very similar problem for the calculation of the casimir force , will be given elsewhere . the expansion of a classical field sketched in section [ class ] is restricted to outgoing waves , i.e. , to @xmath124 , satisfying ( [ bc - right ] ) . in preparing the ground for the expansion of a _ quantum _ field , it is necessary to remove this restriction , for the simple reason that the zero - point quantum fluctuations will inevitably contain incoming waves as well . moreover , one would wish that the ensuing theory should be applicable to situations where there is an incoming pump field . thus , we study the wave equation ( [ wave - eq ] ) for the system together with the outside bath " , i.e. , on the half line @xmath125 , with @xmath49 satisfying ( [ rho - out ] ) and the boundary condition ( [ bc - left ] ) . the initial conditions are now arbitrary and accordingly the outgoing boundary condition ( [ bc - right ] ) is abandoned , i.e. , the restriction of @xmath126 to the cavity need not lie in @xmath77 . for the outside @xmath67 ( where @xmath127 ) the initial data are decomposed as @xmath128 with @xmath129 satisfying the incoming wave condition @xmath130 , while @xmath131 . for the cavity subsystem this decomposition leads to the boundary condition @xmath132 where the driving force @xmath133 ( see ( [ eqom - a_n ] ) below for its name ) , being determined by the initial data , is supposedly a known function ( at least in a statistical sense ) which characterizes the waves incoming from the outside . inside the cavity , the field is then expanded in terms of qnms by ( [ evolve ] ) with @xmath134 + f_j(a^+)\mskip0.5\thinmuskip\phi(a^+ ) \right\}\;. \label{def - a_n}\end{aligned}\ ] ] that is , we _ retain _ the expansion formula and the inner product definition and notation even though @xmath135 . as a consequence , the sum in ( [ evolve ] ) will in general not converge to @xmath136 at @xmath50 , the point where the boundary condition is imposed . nevertheless , the sum for the first component converges to @xmath31 everywhere , while the sum for the second component converges to @xmath32 everywhere except at @xmath50 @xcite . ( this is most easily appreciated by noticing that upon changing @xmath32 at just one point , the resultant wavefunction can be made to lie in @xmath77 . ) this flaw on a set of measure zero does not lead to problems , however , for the projection formula ( [ def - a_n ] ) renders the coefficients @xmath137 well - defined irrespective of the convergence of the series ( [ evolve ] ) . the equation of motion for @xmath13 , which will survive quantization , will now be derived . by differentiating ( [ def - a_n ] ) with respect to time , and then integrating by parts , one obtains @xmath138 in contrast to the case of pure outgoing waves , there is now an extra term on the right hand side : each qnm is driven by the force " @xmath139 , and at the same time decays because of @xmath140 . the coupling to the force " is determined by the surface value of the qnm wavefunction @xmath141 . another approach to second - quantization proceeds more explicitly by first promoting @xmath31 and @xmath32 to operators @xcite . these fields may be regarded as operators for the entire universe " , which is a conservative system to which canonical quantization can be applied . the same projection formula ( [ def - a_n ] ) as in the classical case now defines the @xmath13 s as hilbert space operators , obeying the equation of motion ( [ eqom - a_n ] ) . the crucial point is that the field commutation relation ( [ ccr ] ) and the projection formula ( [ def - a_n ] ) now lead directly to commutators for these coefficients , viz . , @xmath142&=&\frac{1}{4\omega_j\omega_k } \left[\left<\bbox{f}_{\mskip-0.5\thinmuskip\!j},\bbox{\phi}\right > , \left<\bbox{f}_{\mskip-0.5\thinmuskip\!k},\bbox{\phi}\right>\right]\nonumber\\ & = & -\frac{1}{4\omega_j\omega_k } \biggl\{\int_0^{a^+}\!\!dx\mskip0.5\thinmuskip dy\ , \left(f_j(x)\mskip0.5\thinmuskip\hat{f}_k(y)[\hat{\phi}(x),\phi(y ) ] + \hat{f}_j(x)\mskip0.5\thinmuskip f_k(y ) [ \phi(x),\hat{\phi}(y)]\right)\nonumber\\ & & \hphantom{-\frac{1}{4\omega_j\omega_k}\biggl\ { } + \int_0^{a^+}\!\!dx\,f_j(x)\mskip0.5\thinmuskip f_k(a^+)[\hat{\phi}(x),\phi(a^+ ) ] + \int_0^{a^+}\!\!dy\,f_j(a^+)\,f_k(y)[\phi(a^+),\hat{\phi}(y ) ] \biggr\}\;.\nonumber \\ & & \label{a_n - comm}\end{aligned}\ ] ] in these equations , @xmath31 and @xmath32 are @xmath143-numbers , while @xmath53 , @xmath144 are @xmath22-cumbers . the two surface terms on the last line cancel as long as the delta function at the boundary of the integration interval is interpreted consistently . in the first line , the commutation relation ( [ ccr ] ) gives @xmath145 and cancels one integration . one is then left with @xmath142&=&\frac{\omega_k-\omega_j}{4\omega_j\omega_k } \int_0^{a^+}\!\!dx\,\rho(x)\mskip0.5\thinmuskip f_j(x)\mskip0.5\thinmuskip f_k(x ) \label{comm - res}\\ & = & \frac{i(\omega_j-\omega_k)f_j(a^+)\mskip0.5\thinmuskip f_k(a^+ ) } { 4\omega_j\omega_k(\omega_j+\omega_k)}\ ; , \label{comm - r2}\end{aligned}\ ] ] where the second form follows from the first by means of the orthogonality relation ( [ ortho ] ) , and will be useful later for comparison with results from section [ correl ] . the linear space structure for open systems involves projections based on the generalized inner product ( [ scalarprod ] ) which is bilinear rather than linear in one vector and conjugate linear in the other ; thus the expression in ( [ comm - res ] ) involves an integral over @xmath146 without complex conjugation . however , for the sake of a more transparent comparison with the conservative case , it is useful to re - write these expressions by changing @xmath147 and using @xmath148 , @xmath149 and @xmath150 to give : @xmath151 & = & -\frac{\omega_k+\omega^*_j}{4\omega^*_j \omega_k } \int_0^{a^+}\!\!dx\,\rho(x)\mskip0.5\thinmuskip f^*_j(x)\mskip0.5\thinmuskip f_k(x ) \label{comm - resa } \\ & = & -\frac{i(\omega^*_j+\omega_k)f^*_j(a^+)\mskip0.5\thinmuskip f_k(a^+ ) } { 4\omega^*_j\omega_k(\omega^*_j-\omega_k ) } \;. \label{comm - r2a}\end{aligned}\ ] ] the result in the form ( [ comm - resa ] ) reveals the conservative limit most clearly ; in this limit the integral would simply be @xmath152 . comparison of ( [ comm - resa ] ) and ( [ comm - r2a ] ) shows that @xmath153 in the conservative limit . a more explicit proof is given in appendix [ ident - app ] . the above commutators show that , if we define , for @xmath154 @xmath155 then in the conservative limit these should reduce to the annihilation and creation operators , respectively @xcite . indeed , the qnm expansion ( [ evolve ] ) then takes the form @xmath156 the standard normal - mode field expansion for a closed cavity @xcite . for finite damping , however , the operators @xmath13 have mixed creation and annihilation character . in short , we have established an expansion of the quantum field @xmath31 ( and its conjugate momentum @xmath32 ) in terms of operators @xmath12 and @xmath13 , and then obtained equations of motion and commutation relations for the latter . this , in principle , completes the program of second - quantization , and it remains to use these results to compute correlation functions and propagators , which we proceed to do in the following sections . however , the deviation of the commutators ( [ comm - r2 ] ) and ( [ comm - r2a ] ) from the canonical form prevents the construction of a fock space , as is the case for quantum dissipative systems in general @xcite . the formalism derived in the last section for expanding the quantum field @xmath31 in terms of the operators @xmath12 and @xmath13 will be applied to the calculation of equilibrium correlation functions , yielding discrete representations for the cavity correlator @xmath34 in the presence of dissipation . section [ corr - gen ] investigates the general case , and section [ comp ] compares the results with those obtained from the green s function approach in section [ 1comp ] . section [ dos ] evaluates and discusses the density of states . in equilibrium , the initial conditions for ( [ eqom - a_n ] ) are irrelevant and the dynamics are completely specified by the driving force @xmath133 , i.e. , @xmath157 the nonzero imaginary part of the @xmath57 renders the integral rapidly converging , in contrast to the conservative case . fourier transforming and taking expectation values then lead to @xmath158 since @xmath133 is fully specified by the _ in_coming waves from the _ free _ string @xmath159 , it does not know " about the cavity @xmath64 , so one can use the free infinite - string correlation function to calculate its spectral density from the definition ( [ def - b ] ) as @xmath160}{\omega(1-e^{-\beta\omega})}\right|_{y = x}\nonumber\\ & = & \frac{2\omega}{1-e^{-\beta\omega}}\;. \label{b - corr}\end{aligned}\ ] ] for a simple check , antisymmetrize ( [ a_n - a_m ] ) in @xmath15 and @xmath161 and perform the inverse fourier transform to reproduce ( [ comm - r2 ] ) ( for the expectation value of the commutator ) . incidentally , by assuming other forms for @xmath162 , the theory accommodates various incoming pump fields . given the two - point function ( [ b - corr ] ) for the driving force , it is straightforward to compute the two - point function for the response , namely the field field correlation function inside the cavity . this now merely requires summation , that is , combination of ( [ evolve ] ) , ( [ a_n - a_m ] ) and ( [ b - corr ] ) leads to @xmath163 the above derivation leads to a clear physical interpretation of the pole structure of ( [ ff - corr ] ) in the complex @xmath164-plane : the matsubara poles at @xmath165 ( @xmath166 ) arise from the thermal character of the incoming noise , while the qnm poles correspond to cavity resonances excited by this noise . it will be noticed that we have obtained two different qnm expansions for @xmath167 , namely the double sum in ( [ ff - corr ] ) and the single sum in ( [ f - d ] ) . we next prove their equivalence without invoking the qnm expansion of a quantum field . to do so , we rely on the identity @xcite @xmath168 for @xmath169 . this identity , proved in appendix [ gr - app ] , has no nontrivial counterpart in closed , conservative systems . for an interpretation , notice that @xmath170 vanishes in the conservative limit and hence is a measure of dissipation , which the right hand side states as taking place exclusively at the surface @xmath171 . given this identity , the equivalence of the two expressions for @xmath167 follows simply by canceling the bose prefactors in ( [ f - d ] ) and ( [ ff - corr ] ) and comparing the result with the fourier transform of ( [ greenfie ] ) . although the two forms are equivalent , each has its own attractive properties . the diagonal form ( [ f - d ] ) is simpler , while the non - diagonal form ( [ ff - corr ] ) is manifestly factorizable : @xmath172 @xcite . anticipating a similar structure for feynman propagators , the non - diagonal form permits a quantum in one mode @xmath15 to propagate to another mode @xmath161 , while the diagonal form implies that the mode index is conserved " . the expansion of correlations involving @xmath31 alone is not unique , on account of the doubling of qnms ( @xmath15 and @xmath173 ) compared to nms @xcite . as discussed in section [ class ] , it is more natural to consider @xmath174 , which leads to a unique expansion . thus we define a tensor field - field correlator @xmath175 which can be expressed as @xmath176 where @xmath177 is evaluated to be ( appendix [ app - tens ] ) @xmath178 that is , the non - diagonal expansion ( [ ff - corr ] ) is the unique one which generalizes to the tensor @xmath179 as in ( [ sff - expand ] ) . another important quantity is the density of states , which figures prominently in the heuristic argument of purcell @xcite and others @xcite . the local density of states @xmath180 , given below only for real positive @xmath164 , is related to the correlation function @xmath181 by @xmath182 which allows expression of this important quantity in terms of the qnms . from ( [ f - d ] ) one gets @xmath183 while the non - diagonal expression ( [ ff - corr ] ) gives @xmath184 the second form results from the first by use of the identity @xmath185 which follows from ( [ greenfie ] ) by letting @xmath186 . superficially , the diagonal form is simpler . however , if we take a single resonance approximation , ( [ ldos - d ] ) yields , with one term @xmath15 @xmath187 which is not positive definite . on the other hand , for the non - diagonal form , the appropriate approximation is to take one @xmath15 and @xmath188 in ( [ 2dos - d2 ] ) , leading to @xmath189 } \ ; , \label{2dos - dr}\ ] ] which is manifestly positive and moreover lorentzian . from this expression one finds , to leading order in @xmath190 , that @xmath191 where the @xmath164-integral is over one resonance . this statement is readily derived from ( [ 2dos - dr ] ) by using ( [ ratiolim ] ) and the fact that @xmath192 for a narrow resonance . recall that in the modes of the universe approach @xcite , the unit weight of the resonances emerges simply as a numerical result , and is difficult to understand theoretically . here the same result ( in 1 d ) is justified analytically , and moreover , one can in principle ( a ) estimate the corrections due to other resonances ( note that there is no background " apart from the qnm contributions ) , ( b ) calculate the corrections to higher order in @xmath25 , and ( c ) discuss the local density of states @xmath180 rather than the integrated @xmath193 . incidentally , this discussion shows that of the two equivalent forms ( [ 2dos - d ] ) and ( [ 2dos - d2 ] ) , the latter is the more appropriate , since it leads to a finite integral over @xmath164 in the single - resonance approximation . one can derive another sum rule , @xmath194 for large @xmath164 . this second sum rule @xcite states that the states are merely redistributed without changing their total number . however , this sum rule is not immediately useful when expressed in terms of the qnms , and will not be further discussed here . another important correlation function is the feynman propagator @xmath195 in which t denotes time - ordering . taking the fourier transform of the definition ( [ def - fprop ] ) leads to a direct relation to the correlator ( [ def - f ] ) , @xmath196 we shall limit the discussion below to @xmath197 . substitution of the right hand side of ( [ ff - corr ] ) into ( [ gf - f ] ) yields @xmath198 as @xcite @xmath199 the cavity feynman propagator can also be expressed in diagonal form , by substituting ( [ f - d ] ) into ( [ gf - f ] ) . again taking @xmath197 , this leads to @xmath200 for real @xmath164 . it is stressed that these forms as single sums exist even though @xmath201 in general . the form ( [ gf - d2 ] ) for @xmath198 has been derived from ( [ gf - d ] ) by means of the qnm identity ( [ qnm - ident ] ) . the second form with its divergence @xcite at @xmath202 is less convenient than the first . it has been included to show that caution is needed when speaking about the contribution of one qnm " . in fact the two summands are almost equal if @xmath203 ; such resonances are seen to be exclusively associated to terms with @xmath204 . all of these equivalent expressions ( [ gf - od ] ) , ( [ gf - d ] ) and ( [ gf - d2 ] ) can be written generally as @xmath205 with different forms for @xmath206 . this has an obvious diagrammatic interpretation : the field at @xmath41 ( @xmath207 ) couples to the qnm @xmath15 ( @xmath161 ) with a vertex @xmath208 ( @xmath209 ) , and the qnm propagates from mode @xmath15 to mode @xmath161 with an amplitude @xmath206 . this may be compared with the more familiar case of an infinite conservative system , say @xmath210 it is seen that @xmath211 is replaced by @xmath212 . an important goal of the present second - quantized theory is to study cavity atom interactions @xcite , often referred to as cqed . the objective is to establish a set of qnm feynman rules " , in which each line in a diagram is represented not by a continuous momentum , but by one discrete index ( or a pair of them ) not only for computational convenience , but also because each term can be associated with a cavity resonance . such a discrete representation is especially useful for microscopic cavities , where the resonances are widely spaced in frequency . the above results are crucial for establishing these feynman rules . the possibility of alternate expressions for the propagator may recall a similar situation with gauge theories @xcite , though the reasons are quite different . while the use of the feynman propagators in an interacting theory will be presented elsewhere @xcite , it is neverthess profitable at this point to consider the very simple example of an atom coupled to the field at a fixed point @xmath41 ; in the dipole approximation , the decay rate is related to the equal - space propagator @xmath213 in particular , we shall be interested in the single - resonance approximation ( ra ) for @xmath214 . the obvious choice is to take a single term of the sum in ( [ gf - d ] ) , i.e. , @xmath215 the alternative is to start from ( [ gf - od ] ) , and retain only the @xmath216 terms ( only when @xmath188 does the factor @xmath217 in the denominator of ( [ gf - od ] ) get small close to the conservative limit , which is the only case in which a single resonance can dominate ) to arrive at @xmath218\;. \label{dappr}\ ] ] without loss of generality choosing @xmath154 , the second ( non - resonant ) term in ( [ dappr ] ) is of the same order as those already neglected , and hence for most purposes may be omitted . however , only the sum of the two terms in ( [ dappr ] ) preserves the fundamental relation@xcite @xmath219 for real @xmath164 , where @xmath220 ( @xmath221 ) is the retarded ( advanced ) propagator obtained from @xmath222 by continuation from positive ( negative ) frequencies . as a consequence , it turns out@xcite that keeping both terms and using the ensuing cavity propagator to compute the self - energy of a two - level atom leads to a renormalization of the level splitting that is guaranteed to be real . of course , for @xmath223 the equality ( [ dr - da ] ) is always violated . moreover , @xmath223 does not obey the equally fundamental inequality @xmath224 on the real axis @xcite ( see also ( [ relatedd ] ) below ) , which @xmath225 satisfies term by term . violation of this inequality in general leads to a retarded atom propagator that has poles in the upper half @xmath164-plane @xcite , signifying an unphysical instability . to be sure , in spite of these crucial differences between @xmath225 and @xmath223 their residues at @xmath226 agree in the conservative limit , in which the domination of a single qnm becomes rigourous . for a proof it suffices to note that @xmath227 in this limit ( appendix [ ident - app ] ) . we have discussed the single - resonance approximation to both the density of states @xmath180 and to the equal space propagator @xmath228 . in fact , the arguments are equivalent , which can be appreciated physically from the fact that they both relate to the decay rate , and mathematically from the following identity for real positive @xmath164 : @xmath229 in several places we have remarked that the non - diagonal qnm representation has some nice properties , and is in fact the unique representation if the field @xmath31 and the conjugate momentum @xmath32 are considered together , for example in the tensor correlator ( [ sff - expand ] ) . there are of course many ways to understand why the correlator is non - diagonal ; one of the most direct is via ( [ eqom - a_n ] ) , which shows that all the mode coefficients @xmath13 are driven by the same force @xmath139 , so in general different coefficients will have phase coherence and hence a nonzero correlation . incidentally , this non - diagonal nature is _ not _ a quantum effect , since the property survives at high temperatures , e.g. , @xmath230 in ( [ ff - corr ] ) . however , all the propagators and correlation functions become diagonal in the conservative limit , as they should . in fact , applying ( [ ortho ] ) to ( [ a_n - a_m ] ) in this limit readily yields @xmath231 in agreement with the creation annihilation interpretation of the @xmath13 in this limit given above ( [ nm - exp ] ) . as a result both ( [ ff - corr ] ) for @xmath181 and ( [ gf - od ] ) for @xmath198 become diagonal in the conservative limit as well . a useful check and example of the preceding is given by the dielectric rod " model @xcite : @xmath232 that is , we generalize the condition ( [ rho - out ] ) and allow @xmath233 to be an arbitrary constant @xmath234 . to be sure , this generalization is trivial in principle since a model with parameters @xmath235 can be mapped onto one with parameters @xmath236 by the substitution @xmath237 . yet it is convenient in practice , since we can now deal with two different conservative limits ( see the discussion below ( [ dr - f ] ) ) : @xmath238 ( the nodal " limit ) and @xmath239 ( the antinodal " limit ) , by letting @xmath240 and @xmath241 , respectively , while keeping @xmath242 , @xmath0 , and hence @xmath243 ( see ( [ dr - omega ] ) below ) fixed . the model ( [ rho - dr ] ) can be solved exactly for the qnm frequencies @xcite , which read @xcite @xmath244 both @xmath240 and @xmath241 are indeed seen to be conservative limits ( @xmath245 ) . they correspond to clamped and free ends , respectively , in the interpretation of the wave equation as the transverse vibrations of a string @xcite . on the other hand , for @xmath246 the qnm description breaks down as the dissipation tends to infinity . the qnm wavefunctions are given inside the cavity by @xcite @xmath247 their normalization is still given by ( [ normal ] ) , but our generalization @xmath248 is readily shown to imply a corresponding modification of the surface term in the scalar product definition itself , viz . , @xmath249 + n_0\mskip0.5\thinmuskip\phi(a^+)\mskip0.5\thinmuskip\chi(a^+)\right\}\;. \label{scalarpr2}\ ] ] in the nodal " limit @xmath240 , ( [ dr - omega ] ) and ( [ dr - f ] ) show that @xmath250 , so that the factor @xmath251 in the surface term of the orthogonality relation ( [ ortho ] ) , ( [ scalarpr2 ] ) overcomes the explicit factor @xmath252 , allowing the surface term to be neglected . this nodal limit has a counterpart in the loaded string " model @xmath253 , where @xmath254 can be set to infinity @xcite . on the other hand , in the antinodal limit @xmath241 it is the explicit @xmath252 which allows neglect of the surface term , @xmath255 tending to a constant . hence , the qnm expansion becomes a standard normal - mode expansion also if @xmath256 , which clarifies a detail left open in section [ class ] and ref . @xcite . by means of a partial fraction expansion and the identity ( [ qnm - ident ] ) , the expression ( [ f - d ] ) for @xmath181 can be rewritten as @xmath257\ ; , \label{f - d2}\ ] ] which is analytically more convenient even though the sum over @xmath15 converges more slowly . upon substitution of the dielectric rod qnms ( [ dr - omega ] ) and ( [ dr - f ] ) into ( [ f - d2 ] ) , the sum over @xmath15 can be performed as the _ conventional _ fourier series @xcite @xmath258 implying @xmath259 and some rearrangement yields the correlation function as @xmath260}\;. \label{f - dr}\ ] ] this result also follows from the modes of the universe approach in appendix [ mu - app ] . as discussed in section [ 1comp ] , the subtracted correlation function @xmath114 is directly related to the energy density , and is in fact the squared amplitude of the field strength . figure [ fig1 ] shows @xmath261 versus @xmath41 at @xmath262 for the dielectric rod model with @xmath263 , for different values of @xmath264 ; this shows that the field amplitude is largest near the leaky end of the rod . figure [ fig2 ] shows @xmath261 versus @xmath265 at @xmath266 ( all other parameters the same as before ) . this diagram vividly illustrates the advantage of the qnm approach although the result is in principle obtainable from the modes of the universe method , the clear oscillatory signal is best captured by expressing this quantity in terms of qnms . also , for the feynman propagator the sum ( [ gf - d ] ) can be performed if @xmath267 , yielding @xmath268-in_0\sin[n|\omega|(a - y ) ] } { n\cos(n\omega a)-in_0\sin(n|\omega|a ) } \label{gf - dr}\ ] ] for @xmath269 , while for @xmath270 the propagator is obtained via @xmath271 . notice that the final expressions ( [ f - dr ] ) and ( [ gf - dr ] ) tend to a finite limit if @xmath272 even though the individual terms in ( [ f - d ] ) and ( [ gf - d ] ) do not . in this semi - infinite string limit the very notions of cavity and environment lose their meaning , and indeed the right hand sides of ( [ f - dr ] ) and ( [ gf - dr ] ) are seen to become independent of @xmath0 . to summarize , we have developed the second - quantized version of the field theory using the qnm basis . various physical quantities are then written as sums over qnm contributions either as diagonal sums over a single index @xmath15 , or as non - diagonal sums over a pair of indices @xmath273 . the resonance approximation is studied , leading to a proof of the unit weight of narrow resonances in the density of states , or equivalently the enhancement rate for the decay of excited states as embodied in the behavior of the equal space propagator @xmath228 . as has been mentioned already in section [ intro ] , an important extension of the present work is to include matter in the hamiltonian ( [ def - h ] ) , enabling the application of qnms to quantum optics . this will be the subject of the sequel paper ref . other generalizations include the study of vector fields , and of open systems in three space dimensions . further , a development parallelling the present one could be carried out for the klein gordon equation instead of the wave equation ( [ wave - eq ] ) . since the two evolution equations are directly related by a transformation of the spatial variable @xcite however , this has not been taken up here . instead of generalizing the physical system one can also relax the assumption of global equilibrium made in section [ correl ] . it is recalled here that the formalism of sections [ incoming ] and [ quant ] and in particular the driving force @xmath133 of ( [ def - b ] ) is well - defined for any initial state of the fields ; taking a coherent state for the latter instead of a thermal one enables the study of a pumped cavity . on the theoretical side , it would be interesting to provide a path - integral formulation of qnm quantization . this can supposedly be done on two levels . the first , semi - phenomenological one is to write down an effective action generating dynamics equivalent to ( [ eqom - a_n ] ) . the second , more fundamental one is to start with the action for the whole universe for our model ( [ def - h ] ) , integrate out the degrees of freedom of the outside , and use a qnm basis for the ensuing dynamics of the cavity . for such future developments , this article can hopefully serve as a starting point and reference . in conclusion , we have shown that the qnm expansion is as powerful for open second - quantized systems as it is for their classical counterparts . this work is supported by grant no . 452/95p of the hong kong research grants council . we thank c. k. au , e. s. c. ching , w. m. suen and c. p. sun for discussions . in this appendix we give an alternative proof of the identity ( [ ratiolim ] ) in the conservative limit . for this purpose , generalize to complex classical fields and define the energy density @xmath274 , so that @xmath275 with the current @xmath276 . define the cavity energy @xmath277 , then for a field @xmath278 at @xmath279 one has @xmath280 where @xmath281 . in the conservative limit the first term vanishes , while the integral in the second term tends to unity so that @xmath282 . combination with @xmath283 shows that @xmath284 in this limit , which proves our assertion . in this appendix we derive the green s function identity ( [ g - ident ] ) . to this end , define @xmath285 ( @xmath286 ) as the solution of the homogeneous wave equation ( [ def - qnm ] ) ( upon the substitution @xmath287 ) satisfying the first ( second ) of the boundary conditions ( [ f - bc ] ) @xcite . this allows one to write @xmath288 where one can choose @xmath289 and @xmath290 , and where @xmath291 is the position - independent wronskian of the functions @xmath292 and @xmath293@xcite . then one has @xmath294 the numerator of this last expression is itself another wronskian , and hence can be evaluated at @xmath295 to yield @xmath296 completing the proof of ( [ g - ident ] ) . the coefficients @xmath177 in ( [ sff - expand ] ) are given by the projection @xmath297 in terms of the bilinear form on the product space , which reads @xmath298 substitution of ( [ ftensor ] ) for @xmath299 into ( [ tensor - proj ] ) yields @xmath300 inserting _ any _ qnm expansion @xmath301 such as ( [ f - d ] ) or ( [ ff - corr ] ) and invoking the relations ( [ normal ] ) , ( [ ortho ] ) and ( [ qnm - ident ] ) , the expression ( [ def - ajk ] ) can be evaluated as in ( [ ajk - res ] ) , which is what we set out to show . it is instructive to rederive the correlator @xmath34 using the modes of the universe ( mu ) . the mu expansion of the fields reads @xcite i.e. , it is of the same form as ( [ nm - exp ] ) but in ( [ univ - exp ] ) the sum runs over the mu frequencies @xmath303 ( to leading order in @xmath304 ) , the @xmath305 and @xmath306 being mu annihilation operators and wavefunctions , respectively . insertion of ( [ univ - exp ] ) into ( [ def - f ] ) yields to arrive at the third line , we used @xmath308[0pt ] { { \scriptsize(}\raisebox{-1pt}{+}{\scriptsize)}}}{-}$}}\nu_l)\mskip0.5\thinmuskip u_l^{(\dagger)}$ ] and defined the boson occupation number @xmath309^{-1}$ ] . comparison of ( [ ff - corr ] ) and ( [ f - mu ] ) elucidates why the former factorizes with respect to @xmath41 and @xmath207 : this is seen to be a consequence of the nondegeneracy of the mu spectrum of the semi - infinite string , as opposed to , e.g. , a free string or one with periodic boundary conditions . evaluation of @xmath310 for the dielectric rod " of section [ example ] at once shows that ( [ f - mu ] ) indeed reduces to ( [ f - dr ] ) in this case , providing a further comparison between the various techniques of this paper . r. lang , m. o. scully and w.e . lamb , phys . rev . a * 7 * , 1788 ( 1973 ) ; r. lang and m. o. scully , opt . * 9 * , 331 ( 1973 ) ; h. dekker , opt . comm . * 10 * , 114 ( 1974 ) ; physica c * 83 * , 183 ( 1976 ) ; j. c. penaforte and b. baseia , phys . a * 30 * , 1401 ( 1984 ) ; a. j. campillo , j. d. eversole and h .- b . lin , phys . lett . * 67 * , 437 ( 1991 ) . r. p. feynman and f. l. vernon , ann . * 24 * , 118 ( 1963 ) ; p. ullersma , physica * 32 * , 2 ( 1966 ) ; h. dekker , phys . rep . * 80 * , 1 ( 1981 ) , chapter 5 ; a. o. caldeira and a. j. leggett , ann . phys . ( n.y . ) * 149 * , 374 ( 1983 ) ; h. grabert , u. weiss and p. talkner , z. phys . b. * 55 * , 87 ( 1984 ) ; p. s. riseborough , p. hnggi and u. weiss , phys . rev . a. * 31 * , 471 ( 1985 ) . r. g. hulet , e. s. hilfer and d. kleppner , phys . lett . * 55 * , 2137 ( 1985 ) ; d. j. heinzen _ et al . lett . * 58 * , 1320 ( 1987 ) ; f. de martini and g. r. jacobovitz , phys lett . * 60 * , 1711 ( 1988 ) ; e. yablonovitch and k. m. leung , physica b * 175 * , 81 ( 1991 ) . see ref.@xcite for the relation between the two classes of dissipative systems : one in which the coupling is mediated through a term in the total hamiltonian , and the other in which the coupling is mediated by a boundary condition the continuity of @xmath31 across @xmath50 in the present case . k. c. ho , p. t. leung , a. maassen van den brink and k. young , to appear in the proceedings of the 7th asia - pacific physics conference , august 1997 ; p. t. leung , a. maassen van den brink and k. young , in preparation ( 1997 ) . p. t. leung , s. y. liu and k. young , phys . a * 49 * , 3057 ( 1994 ) ; p. t. leung , s. y. liu , s. s. tong and k. young , phys . a * 49 * , 3068 ( 1994 ) ; p. t. leung , s. y. liu and k. young , phys . a * 44 * , 3982 ( 1994 ) . notice that , while novel for open classical systems as in ( [ evolve ] ) , the concept of two - component expansion where the same @xmath313 are used to expand _ both _ @xmath31 _ and _ @xmath136 as in ( [ nm - exp ] ) below enjoys a wide textbook popularity for closed quantum systems . see , e.g. , ref . @xcite . the fact that the sum in ( [ evolve ] ) converges more slowly if @xmath135@xcite is irrelevant in the present formal development . in any case , for operator fields the large wavenumber components never get small as they do for smooth classical fields , irrespective of the expansion method . for numerical applications , it is best to optimize the final formulas for observable physical quantities for convergence along the lines of section iii of ref . notice , however , that comparison of methods b and c in ref . @xcite shows the surface term in ( [ def - a_n ] ) to improve convergence also for function pairs outside @xmath77 , in particular , to be essential for convergence to the second component . the equations ( [ gf - od ] ) , ( [ gf - d ] ) , ( [ gfd - ra ] ) and ( [ dappr ] ) apply for real @xmath164 , and the usual care @xcite is needed when performing an analytic continuation from either positive or negative frequencies . in ( [ nm - exp ] ) , the normalization ( [ normal ] ) forces the @xmath53 to be real in the conservative limit , in contrast to the standard formalism where their phase is arbitrary . for uniformity we choose the @xmath310 in ( [ univ - exp ] ) to be real as well ; this shows that the result in ( [ f - mu ] ) is well - defined .

the galaxy and mass assembly ( gama ) survey furnishes a deep redshift catalog that , when combined with the wide - field infrared explorer ( _ wise _ ) , allows us to explore for the first time the mid - infrared properties of galaxies over 120deg to . in this paper we detail the procedure for producing the matched gama-_wise _ catalog for the g12 and g15 fields , in particular characterising and measuring resolved sources ; the complete catalogs for all three gama equatorial fields will be made available through the gama public releases . the wealth of multiwavelength photometry and optical spectroscopy allows us to explore empirical relations between optically determined stellar mass ( derived from synthetic stellar population models ) and 3.4 and 4.6 _ wise _ measurements . similarly dust - corrected h-derived star formation rates can be compared to 12 and 22 luminosities to quantify correlations that can be applied to large samples to . to illustrate the applications of these relations , we use the 12 star formation prescription to investigate the behaviour of specific star formation within the gama-_wise _ sample and underscore the ability of _ wise _ to detect star - forming systems at . within galaxy groups ( determined by a sophisticated friends - of - friends scheme ) , results suggest that galaxies with a neighbor withinkpc have , on average , lower specific star formation rates than typical gama galaxies with the same stellar mass .

1401.0837

spanning almost six decades and counting , large area surveys have revolutionised our view of the structure and evolution of the universe , for example , the palomar observatory sky survey ( poss ii , reid et al . 1991 ) , the iiia - j src southern sky survey , the infrared astronomy satellite ( iras ) sky survey atlas ( wheelock et al . 1994 ) , 2df galaxy redshift survey ( 2dfgrs , * ? ? ? * ) , the sloan digital sky survey ( sdss , abazajian et al . 2003 ) , the hi parkes all sky survey ( hipass , meyer et al . 2004 ) , the 6df galaxy survey ( 6dfgs , * ? ? ? * ) , the 2 micron all sky survey ( 2mass , skrutskie et al . 2006 ) and the ukirt infrared deep sky survey ( ukidss , * ? ? ? uncovering the properties of large galaxy populations has proved essential to our understanding of how galaxies develop and transform . increasing sensitivity and scale in the time domain , as with skymapper and the planned lsst ( large synoptic survey telescope ) holds the promise of enormous future scientific returns . in the near future , leading - edge radio telescopes and planned strategic surveys will add a key ingredient to the mix hi and continuum measurements will trace neutral gas reservoirs and activity to unprecedented depth , sky coverage and resolution with wallaby and emu on askap ( australian square kilometer array pathfinder ) , apertif - wnshs , laduma @xcite on meerkat , and the jvla . in order to better understand the key physics at work in star - forming galaxies , and crucially the efficacy of different physical mechanisms , one requires the combination of extended area and sufficient depth in order to capture a cosmologically and evolutionarily significant volume . the gama ( galaxy and mass assembly , * * ; * ? ? ? * ) survey provides exactly this laboratory . at its heart , gama is an optical spectroscopic survey of up to @xmath7 300 000 galaxies @xcite obtained at the anglo - australian telescope situated at siding spring observatory . three equatorial fields ( g09 , g12 and g15 ) covering 180deg@xmath1 sample large - scale structure to a redshift of z@xmath8 , with overall median redshift of z@xmath9 , and two southern fields ( g02 and g23 ) . multiwavelength ancillary data from ultraviolet to far - infrared wavelengths provides comprehensive sed ( spectral energy distribution ) wavelength real estate . at mid - infrared wavelengths it falls to _ wise _ , the wide - field infrared survey explorer , to span the near - infrared transition from stellar to dust emission . launched in december 2009 , _ wise _ surveyed the entire sky in four mid - infrared bands : 3.4 4.6 , 12 and 22 @xcite . in the local universe , the 3.4 ( w1 ) and 4.6 ( w2 ) bands chiefly trace the continuum emission from evolved stars ( see , for example , * ? ? ? the w1 band is the most sensitive to stellar light , typically reaching @xmath10 depths to @xmath11 . the w2 band is additionally sensitive to hot dust ; hence , this makes the 3.4 @xmath12 4.6 color very sensitive to galaxies dominated globally by active galactic nucleus ( agn ) emission ( see e.g. , * ? ? ? * ; * ? ? ? the 12 ( w3 ) band is broad ( see * ? ? ? * for relative system response curves ) and can trace both the 9.7 silicate absorption feature , as well as the 11.3 pah ( polycyclic aromatic hydrocarbon ) and @xmath13neii@xmath14 $ ] emission lines . lastly , the w4 band traces the warm dust continuum at 22 , sensitive to reprocessed radiation from star formation and agn activity . _ wise _ is thus optimally suited to study the diverse emission mechanisms of galaxies . _ wise _ is confusion - noise - limited the structure of the background ( e.g. distant galaxies and scattered light ) increases noise by contributing flux and under these conditions point spread function ( psf ) profile - fitted photometry performs robustly @xcite . for unresolved sources , this is the adopted method of obtaining photometry and accordingly the _ wise _ source catalog is optimised and calibrated for point sources . the angular resolution of _ wise _ frames , used for determining the profile photometry measurement ( keyword _ w@xmath15mpro _ in the _ wise _ all - sky data release ) , is 6.1 , 6.4 , 6.5and 12.0 for w1 , w2 , w3 and w4 , respectively . the _ wise _ public - release data products also includes an image atlas of tiled mosaics ; however , the angular resolution is slightly degraded in the atlas images because they are convolved with a matched filter to optimize point - source detection . the crucial point here is that the _ wise _ catalogs and images are not well - suited to characterization of resolved sources . resolved sources will either be detected and measured as conglomerations of several point sources , or have their flux grossly underestimated by the psf profile . as a consequence , it is left to the community to properly measure sources that are resolved by _ wise _ ( i.e. , nearby galaxies ) . in this work we have produced new , better - optimized images to extract resolved galaxy measurements from the gama fields to complement the point - source measurements obtained from the public - release catalogs . a detailed _ wise _ study of several nearby ( @xmath16 mpc ) galaxies @xcite has highlighted how _ wise _ data can be used in multiwavelength studies of star formation and interstellar medium ( ism ) conditions , as well as tracing global properties such as stellar mass and star formation . the work of @xcite showed the power of combining _ wise _ with a large - area survey ( sdss ) ; they investigated the effect of star formation- and agn - activity on the _ wise _ properties of @xmath17 galaxies , including calibrations using the 12 and 22 luminosities . these empirical calibrations have also been investigated in 22-selected galaxies in sdss galaxies with @xmath18 by @xcite . _ wise _ colours have proved to be an excellent agn selection tool @xcite , and as a diagnostic for determining the accretion modes amongst radio - loud agn @xcite . @xcite combine _ wise _ and sdss to provide a phenomenological characterization for _ wise _ extragalactic sources . in this paper we harness the power of the gama survey , and its value - added data products , cross matching gama galaxies in two completed gama fields ( the g12 and g15 regions ) to their _ wise _ counterparts . the g09 region is not considered in this analysis , but will be included as part of the gama-_wise _ data release . gama counterparts at low redshift will be predominantly resolved , particularly in the w1 and w2 bands . unlike previous studies , we include robust measurements of resolved sources in these fields and distinguish between resolved and unresolved systems to determine the photometry most appropriate for a given source . since gama is highly spectroscopically complete to @xmath19 within the two gama regions , we are able to push _ wise _ to higher redshifts than previously possible for wide - field surveys . we explore empirical relationships for stellar mass and star formation rates , the color distribution of _ wise _ sources in gama and their behaviour in dense environments . the paper is arranged as follows : in 2 we describe the data used in this analysis , 3 discusses how the gama-_wise _ matched catalog is constructed , 4 contains results derived from the combined surveys and 5 illustrates scientific applications of the catalog . conclusions are presented in 6 . the cosmology adopted throughout this paper is @xmath20kms@xmath21 mpc@xmath21 , @xmath22 , @xmath23 and @xmath24 . all magnitudes are in the vega system ( _ wise _ calibration described in jarrett et al . 2011 ) unless indicated explicitly by an ab subscript . our sample is drawn from the g12 and g15 equatorial fields of the gama ii survey @xcite combining high spectroscopic completeness ( @xmath25 97% ) to a limiting magnitude of @xmath26 , with a wealth of ancillary photometric data . details of target selection for the survey are outlined in @xcite and optimal tiling for the survey in @xcite . spectra were obtained primarily with the 2df instrument mounted on the 3.9-m anglo - australian telescope , and additionally from the sloan digital sky survey ( sdss dr7 ; * ? ? ? * ) . reduction and analysis of the spectra is discussed in detail in @xcite and liske et al . ( in prep . ) . photometric data for galaxies within the g12 and g15 volumes ( apmatchedcatv05 ) is drawn from sdss imaging ( @xmath27 ) as outlined in @xcite and vista viking ( @xmath28 ) as detailed in driver et al . ( in prep . ) . photometry is corrected for galactic foreground dust extinction using @xcite . stellar mass estimates ( stellarmassesv15 ) are derived from matched - aperture photometry using synthetic stellar population models from @xcite as detailed in @xcite . we take advantage of the completed observations in the equatorial fields and updated redshifts ( baldry et al . , in prep . ) and stellar masses . throughout our analysis we only use sources with reliable redshifts i.e. @xmath29 @xcite . star formation rates ( sfrs ) can be derived from the h@xmath3 equivalent width , combined with the @xmath30-band absolute magnitude , and are available for the gama phase - i survey ( @xmath31 in g15 and @xmath19 in g12 ) as determined in @xcite . corrections are applied for the underlying balmer stellar absorption , dust obscuration and fibre aperture effects . to maintain consistency we only use sfrs from @xcite derived for galaxies with redshifts matching in both gama i and gama ii ( @xmath32 were excluded because of this criterion ) . the _ wise _ mission provides ` atlas ' images via its data repository , which are co - added and interpolated from the multiple single exposure frames taken during the survey @xcite . these 1.56@xmath33 mosaics have a 1.375 pixel scale , but the resampling and co - addition method is optimised for point source detection ; atlas images have beam sizes of 8.1 , 8.8 , 11.0 and 17.5 for the w1 , w2 , w3 and w4 bands , respectively . to better preserve the native resolution of the single frames we employ a mosaic construction that uses the ` drizzle ' resampling technique . variable - pixel linear reconstruction or drizzling co - addition algorithm using a tophat prf ( point response function ) can be used to improve the spatial resolution compared to the nominal atlas images @xcite . drizzled image mosaics , 1.3@xmath341.3 degrees in size , were created using the software package icore @xcite achieving a resolution of 5.9 , 6.5 , 7.0 and 12.4 in the 3.4 , 4.6 , 12 and 22 , respectively . within icore , background level offset - matching , flagging and outlier rejection and co - addition using overlap area - weighted interpolation ensures optimal background stability . the number of images that are combined is dependent on the depth of coverage and additional orbits , a function of the field s location relative to the ecliptic . for the gama fields the coverage is typically 12 and 24 frames for the g12 and g15 fields , respectively , where g15 benefits from additional scans in that region arising from multiple epoch _ wise _ orbits . galaxies were extracted from the g12 and g15 gama ii catalogs of observed sources ( i.e. tilingcatv41 with redshifts as in specallv21 and @xmath35 ) 69,693 and 64,822 for g12 and g15 , respectively , and cross - matched to the _ wise _ all - sky catalog using the nasa / ipac infrared science archive ( irsa ) and a 3 cone search radius . the number of unique objects in each field were 60,645 gama matches for g12 and 58,199 for g15 , i.e. , a 87.0% and 89.8% match rate for g12 and g15 , respectively , with a higher match rate in g15 due to the increased _ wise _ depth ( from additional scans ) and therefore sensitivity . further details are provided in table [ match ] and a region of g15 shown in figure [ fig1 ] to illustrate the difference between the optical and mid - infrared sky . for this analysis , we rely on several parameters output by the _ wise _ all - sky catalog ; these are listed in table [ param ] with a brief explanation . l l l l l + + band & detections & snr@xmath36 @xmath37 & snr@xmath38 & upper limits@xmath39 + w1 & 100% & 100% & 99.9% & 0% + w2 & 100% & 96% & 77% & 4% + w3 & 99% & 47% & 17% & 52% + w4 & 98% & 19% & 2% & 79% + + + + band & detections & snr@xmath37 & snr@xmath38 & upper limits@xmath39 + w1 & 100% & 100% & 99.9% & 0.01% + w2 & 100% & 98% & 88% & 2% + w3 & 99% & 58% & 29% & 42% + w4 & 98% & 19% & 3% & 78% + + + + + l l + _ designation _ & unique _ wise _ source designation + _ ra _ ( deg ) & right ascension ( j2000 ) + _ dec _ ( deg ) & declination ( j2000 ) + _ sigra _ ( arcsec ) & uncertainty in ra + _ sigdec _ ( arcsec ) & uncertainty in dec + _ w@xmath15mpro _ ( mag ) & instrumental profile - fit photometry magnitude + _ w@xmath15sigmpro _ ( mag ) & instrumental profile - fit photometry uncertainty + _ w@xmath15rchi2 _ & instrumental profile - fit photometry reduced @xmath40 + _ nb _ & number of blend components used in each fit + _ na _ & active deblend flag ( = 1 if actively deblended ) + _ xscprox _ ( arcsec ) & distance between source center and xsc@xmath39 galaxy + _ w@xmath15rsemi _ & ( scaled ) semi - major axis of galaxy from xsc@xmath39 + _ w@xmath15gmag _ & elliptical aperture mag of extracted galaxy + + + + a key feature of a catalog containing _ wise _ photometry is determining which of the galaxies are resolved and then ensuring their fluxes are measured robustly . a broad indication is given by the reduced @xmath40 ( _ w@xmath15rchi2 _ ) of the profile - fit solution , where high values of the _ w@xmath15rchi2 _ parameter indicate that the wpro algorithm measurement is underestimating the flux of the source . sources with _ w@xmath15rchi2 _ @xmath37 are often resolved , with _ @xmath41 usually indicating well - resolved systems . w@xmath15rchi _ is not by itself a robust measure of ` resolvedness ' , particularly where sources have a low signal to noise . the ` reduced ' metric tends to unity as noise begins to dominate the measurement . moreover , the _ w@xmath15mpro _ fitting process can be fooled for sources with relatively bright cores that are just being resolved by _ wise _ this happens with 2mass compact extended sources which have low _ w@xmath15rchi2 _ , but the _ w@xmath15mpro _ and isophotal photometry can be systematically offset . even though 2mass has superior ( 2@xmath34 ) spatial resolution compared to _ wise _ , with its greater sensitivity _ wise _ is able to resolve many nearby , relatively small galaxies . therefore , all 2mass extended source catalog ( xsc , * ? ? ? * ) sources should be tested to determine if they are resolved by _ wise _ ( see [ rfuzzy ] ) . for galaxies not in the 2mass xsc , values of _ w@xmath15rchi2 _ @xmath42 should be used as an initial selection for resolved objects . caution is required , however , since blended objects will also satisfy this criterion and can be a source of false positives , notably when the stellar confusion is significant . sources are measured using custom software adapting tools and algorithms developed for the 2mass xsc @xcite and _ wise _ photometry pipelines @xcite . the process is semi - automated in that photometry measurements are automated , but each result is assessed by visual inspection with intervention where necessary . the first step is to remove point sources by psf subtraction which preserves the structure of the background . if necessary surrounding contaminating sources are masked ( e.g. , bright stars ) . the local background is estimated from pixel value ( trimmed average ) distribution that lies within an elliptical annulus located just outside of an ` active region ' which represents the image area that contains measurable light from the galaxy . the active region is not initially known , but it determined through successive iteration of the characterisation process until convergence is reached . the galaxy is modeled using an ellipsoid built from azimuthally averaging the ( background - subtracted ) surface brightness ; any masked pixels are recovered using a weighted combination of the local background ( to the masked source ) and the galaxy model . the best fit axis ratio and ellipticity are determined using the 3@xmath43 isophote and the galaxy shape is defined by this isophote and is assumed to be fixed at all radii . the primary ` isophotal ' photometry ( w@xmath15iso ) is then measured from the 1@xmath43 ( of the background rms ) elliptical isophote , capturing over 90% of the total light ( see jarrett et al 2013 ) . other measurements include a double srsic fit , to the inner galaxy region ( i.e. , the bulge ) and the outer region ( i.e. , the disk ) , thus allowing estimation of the total integrated flux that extends beyond the 1@xmath43 isophote . since _ wise _ is confusion - noise limited , we track the photometry using a curve - of - growth table . where a large mismatch occurs between the isophotal radius and the radius where the change in flux is less than 2% ( the ` convergence point ' ) , the size of the active region is automatically decreased . this usually occurs where background levels are elevated due to a neighboring bright source which contaminates the background level . the process is iterated to adjust the active region until the source measurements have converged . once all the sources in a field are measured and modeled accordingly , the process is rerun to effectively perform galaxy - galaxy deblending . where a measurement exists for a galaxy when measuring an adjacent source , the galaxy model is subtracted in addition to neighboring point sources . this mitigates contamination from nearby resolved sources . the _ wise _ bands at 3.4 and 4.6 achieve the best spatial resolution , at 6.1 and 6.4 , respectively , and capture extended light arising from evolved stars in the bulge and disk regions . since _ w1rchi2 _ can not discriminate well between faint , resolved and unresolved systems , we use the 2nd order intensity - weighted moment to describe the shape ( major and minor axis ) of the object and measure the intensity - weighted moment of the object on either side of the main axis . if the object is symmetric and resolved on both sides of the major axis ( as with a nominal resolved source ) , then each moment will have a resolved signature ( i.e. , the moment is large compared to a point source ) . if the object is , for example , a star blended with a galaxy , then their moments will be different ( one half will have a moment that is point - like , the other resolved ) . thus , taking the minimum moment value between the two halves , determines whether the source is resolved . fuzzy objects will have a large moment regardless of which major axis half that is measured ; stars or double stars will have a minimum moment that appears to be unresolved . the performance of the rfuzzy parameter is illustrated in [ rfuzz_sec ] of the appendix . isolating the resolved sources using the rfuzzy parameter ( in w1 ) gives 1,390 and 1,368 sources in g12 and g15 , respectively , i.e. @xmath44% of the total _ wise _ matched sources in these fields . it is important to bear in mind that despite the resolution of the w1 and w2 bands , _ wise _ is far more sensitive compared to , for example , 2mass and will detect the extended light profiles from nearby galaxies . the typical w1 1@xmath43 isophotal radius is more than a factor of @xmath252 in scale compared to the equivalent 2mass k@xmath45-band isophotal radius . an illustration of this is shown in figure [ res ] where the 2mass xsc - derived scaled aperture photometry ( _ w@xmath15gmag _ ) are compared to the isophotal photometry of resolved sources in the gama g12 and g15 fields . the 2mass xsc - derived magnitudes underestimate the inferred flux due to the increased sensitivity of _ wise_. note that no star subtraction or deblending has been attempted when measuring the _ w@xmath15gmag_s and contamination may be present ; as a consequence , these measurements should only be used as a last resort ( employing the offset shown in figure [ res ] to obtain the correct galaxy flux ) . the behavior of resolved sources is further explored in [ resphot ] of the appendix . for point sources , the primary photometry are the profile fit measurements ( _ w@xmath15mpro _ ) , and for well - resolved sources the isophotal photometry ( 3.1 ) . when the source is not well resolved , or the s / n ( signal - to - noise ) is low , the following steps are used to choose the best photometry : * _ w@xmath15mpro _ photometry is used when the s / n is low , as measured by the isophotal aperture process for a given band : s / n of 10 , 10 , 15 and 15 for w1 , w2 , w3 and w4 , respectively . * if the rfuzzy parameter is false ( measured in the w1 band ) , classify source as unresolved and use _ w@xmath15mpro _ photometry for bands w1 and w2 . * if _ w3rchi2 _ @xmath46 2 , classify source as unresolved and revert to the _ w3mpro _ photometry . * if _ @xmath46 2 , classify source as unresolved and revert to the _ w4mpro _ photometry . * w1 is the most sensitive _ wise _ band . to obtain an accurate w1@xmath12w2 color for galaxies resolved in w1 and w2 , the following steps are employed . firstly , an accurate w1@xmath12w2 color is determined using a matched aperture derived from the smaller of the two isophotal radii ( which is usually w2 ) . then , from the isophotal magnitude of the w1 band , the corresponding w2 magnitude is determined , thereby reflecting the sensitivity of the w1 band . this is also done for the w4 band using w3 in the same way as w1 , if sources are resolved at these wavelengths , which in gama is rare ( only 16 sources in g12 and g15 ) . we note that the w1 and w3 bands can not be matched in this way since the flux at these wavelengths is produced by different physical processes ( evolved stars vs. pah features ) . aperture corrections are applied to isophotal measurements as detailed in the explanatory supplement to the _ wise _ all - sky data release products ] . in figure [ sens ] we plot the signal - to - noise ratio ( snr ) for the g12 and g15 fields ; due to the additional @xmath252@xmath34 frame coverage the g15 field has higher signal - to - noise detections compared to g12 at a given w1 magnitude . the magnitude sensitivity limit in the w1 band for the g12 field is 16.6 ( 10-@xmath43 ) , 17.3 ( 5-@xmath43 ) and for g15 , 17.0 ( 10-@xmath43 ) and 17.7 ( 5-@xmath43 ) . the combined fields have a 10-sigma magnitude sensitivity limit of 16.6 or 71 @xmath47jy . the redshift distribution of sources in the two fields ( excluding upper limits ) is shown in figure [ cat1 ] . since the g15 field is more sensitive , the peak of the distribution is shifted to a slightly higher redshift compared to g12 . the two fields show very similar redshift distributions for @xmath48 ( figure [ cat1]b ) , although the g12 matches have relatively more high-@xmath49 sources . point sources are both passively and actively deblended by the _ wise _ pipeline and the _ wise _ all - sky catalog provides the _ na _ and _ nb _ blending flags where a value of 0 " indicates no active deblending has been performed , and 1 " if sources have been deblended ; deblending introduces an additional uncertainty on the measurement provided ( nb is the number of components that were deblended ) . however , an additional flag is needed to indicate where _ wise _ can not distinguish between gama sources ( one _ wise _ source for multiple gama sources ) and where contamination is expected to be high since more than one gama source lies within the _ wise _ beam . as such , we use an additional blending flag as determined by the proximity of neighboring gama-_wise _ sources . sources within 5 are viewed as a catastrophic blend due to the size of the _ wise _ beam and multiple gama sources will probably have the same _ wise _ source matched to it and/or have a highly uncertain match . a source within 15 indicates a potential blend or contamination and is also flagged . for the analysis in this study we remove all flagged sources from our sample ( @xmath50 6% of g12 and g15 ) to ensure the cleanest _ wise _ photometry , but may bias the sample against the densest environments . rest - frame colours are determined using the gama redshift by building an sed ( spectral energy distribution ) combining the optical , near - infrared and mid - infrared flux densities and fitting to an empirical template library , consisting of 126 galaxy templates , of local well - studied and morphologically diverse galaxies ( e.g. sings , the _ spitzer _ infrared nearby galaxy survey * ? ? ? * ) from @xcite . the templates are constructed from optical and _ spitzer _ spectroscopy , with matched aperture photometry from _ galex _ @xcite , _ swift _ ( uvot ; * ? ? ? * ) , sdss , 2mass , _ spitzer _ @xcite and _ wise _ , combined with magphys @xcite . the _ wise _ relative spectral response curves are from @xcite and are available as part of the _ wise _ explanatory supplement . the best template fit is determined by minimising the function that describes the difference between the flux density in the measured band , and the model - template flux density . each resulting best - fit is assigned a normalised score derived from the reduced @xmath40 minimisation , with the most consistent fits having values of @xmath51 and relatively poor fits with scores of @xmath52 . in the sections that follow we use only those sources with a score @xmath53 , although the fitting accuracy is minimally important for nearby galaxies ( i.e. with low rest - frame color corrections ) . as shown by @xcite , the 3.4 , 4.5 and 12 bands of _ wise _ can be combined in a mid - infrared color - color diagram ( figure [ col ] ) ; the shortest bands are sensitive to the evolved stellar population and hot dust , therefore their color indicates increased activity ( agn or starburst ) . the 12 band is dominated by the 11.3 pah , as well as the dust continuum , sensitive to star formation . this color - color diagnostic is therefore useful to separate galaxy populations , particularly old stellar population - dominated , star - forming and systems dominated by agn - activity @xcite . in figure [ col]a the observed colors of the g12 and g15 are shown , and the k - corrected version , figure [ col]b , shows the distribution of the rest - frame colors . both fields have similar distributions , biased towards star - forming systems . this is not surprising given the optical selection of the _ gama _ sample . heavily obscured galaxies , notably ultra - luminous infrared galaxies ( ulirgs ) are absent due to the insensitivity ( selective extinction ) of sdss and gama optical catalogs . systems globally dominated by agn - activity also appear sparse within the sample ; this is consistent with the findings of @xcite where spectroscopically - identified agn make up @xmath54 of the gama i emission - line catalog ( see also * ? ? ? * ) . the 3.4 band of _ wise _ is dominated by the light from old stars and can be used as an effective measure of stellar mass @xcite . to explore this further , we calculate the ` in - band ' luminosity for w1 , i.e. the luminosity of the source as measured relative to the sun in the w1 band , and use the stellar masses for gama as determined by @xcite to determine a mass - to - light ( m / l ) ratio . these stellar masses are best constrained for @xmath55 @xcite and we apply this cut to our sample . as shown by @xcite , the stellar mass - to - light has a linear trend with _ wise _ w1@xmath12w2 and w2@xmath12w3 color , reflecting systematic m / l differences between passive and star - forming systems . in order to empirically calibrate a relation we require high signal - to - noise measurements and additionally use only the rest - frame w1@xmath12w2 color since the detection rate is much higher in w2 compared to w3 . we apply a s / n cut of 13.5 in w1 and w2 and remove known agn ( from the gama i spectroscopy measurements of * ? ? ? * ) and also systems with _ wise _ colors consistent with agn activity dominating their global colors ( w1@xmath12w2@xmath56 ) as discussed by @xcite . the first population we investigate are resolved , low-@xmath49 galaxies , shown in figure [ sm1]a . we illustrate the danger of a least - squares minimisation fit compared to a bivariate - gaussian , maximum - likelihood ( or ` best ' fit ) line @xcite . throughout the paper we plot both the least - squares and maximum - likelihood lines to show the large discrepancies that may exist ; we provide relations for the maximum - likelihood ( or best fit ) only . we find very good agreement with the m@xmath57/l@xmath58 relation of @xcite based on their relatively small sample of 17 galaxies and with stellar masses effectively derived from 2@xmath59 @xmath60 band photometry . our relation for resolved low-@xmath49 sources is : @xmath61 with @xmath62 + where m is the absolute magnitude of the source in w1 , m@xmath63 , and @xmath64 reflects the rest - frame color of the source . for comparison to the resolved sample , in figure [ sm1]b we show all sources that meet our s / n selection criterion and redshift cut , which shows the distribution to be shifted to warmer " w1@xmath12w2 colors i.e. signifying systems with increased activity such as star - formation or low - power agn . in addition , the contours suggest that the distribution is made up of two populations , most probably due to passive galaxies having a larger mass - to - light ratio than disk - dominated and dwarf galaxies . we investigate this further in figure [ sm2 ] where we make a color - distinction based on the w3 measurement . by selecting galaxies with w2@xmath12w3@xmath65 ( see figure [ col ] ) we choose systems that are most likely dominated by star formation . these selected types show a clear trend with mass - to - light ( figure [ sm2]a ) , but lie parallel and offset to the relation of @xcite . but now including the passive galaxies , figure [ sm2]b , we see a clear clustering around a fixed mass - to - light ratio of @xmath66 . we conclude that the best - fit for our entire sample is : @xmath67 + and for star - forming ( lower mass - to - light ) systems only : @xmath68 this likely explains the shift observed in the relation for resolved sources and the entire field sample . the resolved sources are relatively nearby and will be a mix of passive and star - forming systems . however , at higher redshifts we add relatively more star - forming galaxies ( higher infrared luminosity ) causing a larger spread in w1@xmath12w2 color . notably galaxies with higher star formation rates will have a larger w1@xmath12w2 color due to more hot dust emission giving a brighter w2 measurement . also , at higher redshifts we will preferentially detect galaxies with higher star formation rates ( see next section ) . it should be noted that agn activity would have a similar effect ( as shown by the upward trend in figure [ col]b ) and some contamination from nuclear activity is inevitable . in figure [ sm3 ] we plot the residuals of the gama stellar masses ( used to calibrate the wise relation in equation 2 ) and the wise - derived values themselves . the mass estimates agree within a factor of 1.2 at @xmath69m@xmath70 ( with a standard deviation of 0.5 ) , within a factor of 0.98 at @xmath71m@xmath70 ( with a standard deviation of 0.4 ) and a factor of 0.96 at @xmath72m@xmath70 ( with a standard deviation of 0.4 ) . for stellar masses @xmath73m@xmath70 the wise - derived masses appear overall lower than the gama stellar masses . this is probably unsurprising given that the sample is dominated by star - forming galaxies ( see figure [ col ] ) and the wise 3.4 band is sensitive to the light from evolved stars in passively evolving galaxies . _ iras _ , _ iso _ and _ spitzer _ revolutionised our understanding of dust emission as a tracer of star formation . in the mid - infrared , the _ spitzer _ mips 24 band measures the warm dust continuum excited by hot , young stars and is therefore sensitive to recent star formation , as well as agn - activity . numerous studies have investigated its stability as a measure of sfr ( see e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) and we are able to transfer this understanding to _ wise _ and its 22 band . we determine the luminosity density ( @xmath74l@xmath75 ) for the 22 band and normalise by the total solar luminosity ( l@xmath76= 3.839@xmath3410@xmath77w ) . the w4 band is the least sensitive _ wise _ band , but we impose a s / n cut of 7 to ensure high quality photometry . sources flagged as agn based on optical spectroscopy diagnostics are removed from the sample . cross - matching these sources with star formation rates available for gama i , using sources with sfr between 0.1 and 100 m@xmath76yr@xmath21 ( see * ? ? ? * ) , yields the best - fit relation shown in figure [ sf2 ] . the optical star formation rates are determined from h@xmath3 equivalent widths applying an extinction - correction based on the balmer decrement ( full details can be found in * ? ? ? * ) and are sampled to @xmath78 ( i.e. beyond which h@xmath3 is redshifted out of the observed spectral range ) . galaxies where the h@xmath3 line is contaminated by atmospheric o@xmath79 ( a band ) absorption , in the redshift range 0.155@xmath80 0.170 are not included in the sample as recommended in @xcite . the points in figure [ sf2 ] are color - coded according to balmer decrement and on average higher infrared luminosities correspond to higher balmer decrements ( i.e. , increased dust obscuration ) . however , without detailed knowledge of the dust geometry this is merely a crude comparison ( see also * ? ? ? * ; * ? ? ? * ) , and as we discuss below , biases due to extinction are probably in play at higher redshifts . balmer decrement has also been shown to be correlated with stellar mass , stellar mass surface density and metallicity ( see for example , * ? ? ? * ) -derived star formation rates as a function of @xmath74l@xmath81 luminosity color - coded by balmer decrement ( h@xmath3/h@xmath82 ) from @xcite.,width=302 ] figure [ sf2 ] also shows the linear fit of @xcite , where the _ wise _ 22 -based sfr relation was calibrated using _ spitzer _ 24 photometry and the relation of @xcite . this fit is somewhat steeper , likely explained by the distinction made by @xcite based on total infrared luminosity of the source and the relatively small number of sources in the @xcite sample ( but which also included local group dwarf galaxies ) . the best fit to the gama sample distribution is : @xmath83 infrared - luminous sources will be detected by _ wise _ 22 out to moderate redshifts ( @xmath84 2 to 3 ) , but most typical galaxies would not . hence there is a strong bias for w4 detections to be luminous galaxies ( lirgs and ulirgs ) , which are highly obscured at optical wavelengths . the _ wise _ 12 band , on the other hand , has greater sensitivity by comparison ( see table [ match ] ) and also probes the ism , sensitive to a larger ( representative ) sampling of galaxies , and thus making it potentially the primary star formation indicator for _ wise_. the dominant feature within this w3 band is the 11.3 pah ( polycyclic aromatic hydrocarbon ) , and to a lesser extent the @xmath13neii@xmath14 $ ] emission line . this pah is large , neutral and excited by ultraviolet radiation from young stars , as well as radiation from older , evolved stars ( see , for example , * ? ? ? @xmath13neii@xmath14 $ ] is associated with emission from hii regions . passive disks generally lack the 7.7 pah tracing current star formation , but still show prominent 11.3 pah features ( see , for example , * ? ? ? as first demonstrated with _ spitzer _ measurements ( see for example * ? ? ? * ; * ? ? ? * ) , pahs can be used to estimate star formation , although there is larger scatter ( and potential biases ) relative to the superior 24 mid - infrared or 70 far - infrared tracers . the _ wise _ 12 band is thus a powerful , yet also more problematic , tracer of recent star formation . we proceed as with w4 , but with a s / n cut of 10 in the w3 ( 12 band ) , similarly color - coding to reflect balmer decrement ( figure [ sf1]a ) . a contour plot of the distribution is shown in figure [ sf1]b and the trend appears very tight for sfr @xmath85 . at low sfr and @xmath74l@xmath75 , however , the distribution appears to flatten and probably accounts for the differences in slope between our relation and the least squares fit of @xcite.the relation of @xcite lies offset below our best fit , most likely as a result of relatively low sfr within their small sample ( sfr @xmath86 ) of nearby galaxies . the best fit relation for the 12 band is : @xmath87 this flattening of the distribution in figure [ sf1 ] at low sfr is most likely a feature of the 11.3 pah tracer ( since we do not observe it at 22 ) and is also observed by @xcite in their study . given that the 12 micron band emission appears too low for the given h@xmath3 sfr , a possible explanation is that the relative abundance of pah molecules to big grain emission is diminished , due to low metalicity in these galaxies . if this is the case , we do not expect to see a similar effect at 22 micron , since this band is dominated by big grains in equilibrium with the strong radiation fields inside star formation regions . indeed , the 22 sample of @xcite does not show such a flattening . an alternative explanation would be that in a low sfr regime , the preponderance of the diffuse medium increases with respect to the star - formation component of the interstellar medium ( galaxies are more quiescent ) . bear in mind that the 24 warm dust emission is powered by the uv radiation fields from the massive stars inside the star forming regions , while the 12 emission is mainly powered by the diffuse interstellar radiation fields @xcite . thus , the pah and small grains emission seen in the 12 band increases with increasing sfr , but this increase is mediated by the propagation of the uv photons in the diffuse ism , while the 24 emission directly traces star formation regions , as does the balmer line corrected h@xmath3 emission from which the sfrs are derived . to investigate this further we show individual points for contiguous redshift slices overplotted on the full sample in figure [ sf3 ] ; this serves as an indication of how the distribution is built up . the flattening at low sfr appears to be a feature of nearby sources alone ( figure [ sf3]a ) supporting the hypothesis that we are observing the pah features and small grain emission of relatively quiescent galaxies , dominated by the diffuse ism . the flattening in the slope is not detected at higher redshifts ( @xmath88 ) as these galaxies have more star formation activity ( i.e. more clumpy ) . in figures [ sf3]b - d we see how as we move to higher redshifts , we sample systems with higher @xmath89 and , on average , higher sfrs . the relatively low number of sources in figure [ sf3]d reflects the atmospheric contamination affecting sources at redshifts of @xmath90 . the higher infrared luminosity in these sources , however , also indicate greater obscuration in the optical bands ; this explains the sharp drop in h@xmath3 luminosities relative to w3 , apparent in figs . 10c and d with sfr(h@xmath3 ) @xmath46 3 m@xmath76yr@xmath21 . corrections using the balmer decrement become ineffective when the extinctions are high , ( av @xmath91 ) , which is why infrared tracers of obscured star formation and uv / optical tracers of unobscured star formation are best combined to estimate the total star formation rate . the gama-_wise _ catalog and empirically - derived relations can be used any number of ways to probe the behaviour of large populations of galaxies . to illustrate , we highlight below how _ wise _ can be used to study the specific star formation of galaxies . l@xmath92 luminosity , for the entire sample @xmath74l@xmath92 as a function of stellar mass , giving the ssfr , color - coded by redshift . lines of constant sfr ( 0.1 , 1 , 10 and 20 m@xmath76yr@xmath21 ) are shown as dotted , solid , dashed and dash - dot lines , respectively.,width=302 ] we can investigate the specific star formation rate ( ssfr ) of the entire sample by using @xmath93-derived star formation rates ( equation 5 ) and the gama stellar masses @xcite . in figure [ ssfr ] we show specific star formation as a function of stellar mass , color - coded by redshift . two clear trends are seen : ( 1 ) lower mass galaxies are actively building their disks while massive galaxies have expended their gas reservoirs rendering mostly passive evolution ( the sfr - m@xmath94 " relation ) and ( 2 ) with increasing redshift , a shift to higher sfr for a given stellar mass as we capture these infrared - luminous systems . the behavior of star formation and specific star formation , as a function of stellar mass and redshift , within the gama sample is explored in detail in @xcite and @xcite . the observed increase in mean ssfr of star - forming galaxies with increasing redshift is well - established ( see for example * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) and we illustrate here that the gama-_wise _ sample is sufficiently large and diverse to explore galaxy evolution between the local universe and @xmath95 0.5 . from the relation of @xcite , the main - sequence " of galaxies with stellar mass of 10@xmath96m@xmath70 at @xmath2 have typical sfrs of @xmath97 m@xmath70yr@xmath21 . converting to luminosity density by way of equation ( 5 ) gives log@xmath98l@xmath70 . as illustrated in figure [ lumw3 ] , _ wise _ can detect these systems that are within the gama-_wise _ sample . a luminosity density of log@xmath98l@xmath70 corresponds to a flux density of 0.24 mjy or 12.8mag ; using the s / n detection statistics of the g15 sources , _ wise _ will detect this source with s / n@xmath97 . the magnitude sensitivity limit in the w3 band for the g12 field is 13.2 ( 10-@xmath43 ) , 14.0 ( 5-@xmath43 ) and for g15 , 13.8 ( 10-@xmath43 ) and 14.5 ( 5-@xmath43 ) . the 12-derived star formation rates , computed from equation ( 5 ) , can also be used to investigate ssfr trends within populations of gama galaxies . for example , we use the updated gama ii galaxy group catalogue ( g@xmath99cv6 ) , as detailed in @xcite , to isolate galaxies that reside within a pair . in g@xmath99c a galaxy pair is defined as two galaxies within 100@xmath100 , in physical separation , and 1000 kms@xmath21 in velocity separation . figure [ pairs ] is a normalised histogram of all galaxies in the sample with stellar mass @xmath101 10@xmath102m@xmath76 and the subset that reside in a pair ( 17,475 galaxies ) and a close pair ( i.e. separation @xmath103 ) . we note that potential blends ( as outlined in [ blend ] ) are removed . the striking feature of figure [ pairs]a is that galaxies that reside within a pair appear to have , on average , lower ssfr than a typical gama galaxy of the same mass , suggesting that instead of star formation being enhanced in these systems , it is suppressed . although initially counter - intuitive , and contrary to studies of local galaxy pairs compared to field control samples ( see , for example , * ? ? ? * ) , it should be noted that @xmath104 of gama galaxies reside within a group , with numerous mechanisms at play . within the context , therefore , of the environment of a typical galaxy this highlights the complexity of interacting and merging systems and is discussed further in a detailed study of merging and interacting galaxies within gama ( robotham et al . , in prep . ) . we note that a lack of sfr enhancement has emerged from studies of galaxy pairs probing higher redshift samples ( see , for example , * ? ? ? * ) , with one suggested explanation that higher gas fractions at higher redshift reduce the efficiency of torque - driven gas infall . further we include in figure [ pairs]b a histogram of galaxies within 20kpc / h of their neighbor , as compared to the distribution from figure [ pairs]a . it appears that this subset shows a broader range of ssfrs compared to the larger sample of galaxies in pairs , and may even be bimodal , consisting of suppressed systems and enhanced systems . adding this kinematic trigger to galaxy evolution is clearly a complex process , but crucial to understand how galaxies evolve in the group environment . finally , we note that blending limitations of the _ wise _ data within pairs that have a smaller separation make this parameter space uncertain , but optical tracers will be exploited here in forthcoming gama papers . we have detailed the procedure for creating a source - matched catalog between galaxies in the gama g12 and g15 fields , and _ wise _ photometry . in particular we have outlined how best to extract photometry for low signal to noise , unresolved and resolved sources in the _ wise _ all - sky catalog . complete gama-_wise _ catalogs for the g09 , g12 and g15 fields will be made available through the gama public releases ( see http://www.gama-survey.org/ ) . using the _ wise _ measurements and matched gama galaxies in the g12 and g15 fields , we have investigated the following : * the _ wise _ color distribution for the sample shows most systems are globally dominated by star formation , with few passive and agn - dominated systems ; this is consistent with the gama sample selection and spectroscopic sensitivity to higher redshift , star - forming galaxies . * empirical relations of stellar mass as a function of 3.4@xmath124.6 color for resolved sources , our entire sample and star formation - dominated galaxies only . we provide relations that can be applied to large samples for @xmath4 . * star formation rate relations can be derived using 22 and 12 luminosities . the 12-derived sfr relation ( equation 5 ) is a complex tracer of the ism and should , however , be used with caution . we show that the distribution of galaxies in the 12 luminosity and h@xmath3 sfr plane as a function of redshift are affected by selection effects and most probably dust geometry . * specific star formation ( using 12-derived sfrs ) for the sample illustrates how the gama-_wise _ catalog detects only the most massive , highest star - forming systems at the highest redshifts . _ wise _ is , however , able to detect star - forming main - sequence systems , of stellar mass @xmath710@xmath96m@xmath70 , to @xmath11 with s / n @xmath105 . * controlling for stellar mass , galaxies with an associated neighbor appear to experience , on average , a shift to lower specific star formation . extracting pairs with relatively small separations ( @xmath46 20@xmath106kpc ) suggests a broader behaviour consistent with populations experiencing either star formation suppression or enhancement . mec and thj acknowledge support from the national research foundation ( south africa ) . mec , mll , mo , aeb and mjib acknowledge support from the australian research council ( fs110200023 , fs100100065 , ft100100280 ) . mlpg acknowledges support from a european research council starting grant ( degas-259586 ) . this publication makes use of data products from the wide - field infrared survey explorer , which is a joint project of the university of california , los angeles , and the jet propulsion laboratory / california institute of technology , funded by the national aeronautics and space administration . gama is a joint european - australasian project based around a spectroscopic campaign using the anglo - australian telescope . the gama input catalog is based on data taken from the sloan digital sky survey and the ukirt infrared deep sky survey . complementary imaging of the gama regions is being obtained by a number of independent survey programs including galex mis , vst kids , vista viking , wise , herschel- atlas , gmrt and askap providing uv to radio coverage . gama is funded by the stfc ( uk ) , the arc ( australia ) , the aao , and the participating institutions . the gama website is http://www.gama - survey.org/. 99 abazajian , k. n. , adelman - mccarthy , j. k. , ageros , m. a. , et al . 2009 , , 182 , 543 abazajian , k. , adelman - mccarthy , j. k. , ageros , m. a. , et al . 2003 , , 126 , 2081 alonso - herrero , a. , rieke , g. h. , rieke , m. j. , et al . 2006 , , 650 , 835 assef , r. j. , stern , d. , kochanek , c. s. , et al . 2013 , , 772 , 26 baldry , i. k. , robotham , a. s. g. , hill , d. t. , et al . 2010 , , 404 , 86 bauer , a. e. , hopkins , a. m. , gunawardhana , m. , et al . 2013 , , 434 , 209 brown , m. j. i. , moustakas , j. , smith , j .- d . t. , et al . 2013 , arxiv:1312.3029 boselli , a. , hughes , t. m. , cortese , l. , gavazzi , g. , & buat , v. 2013 , , 550 , a114 bouch , n. , dekel , a. , genzel , r. , et al . 2010 , , 718 , 1001 bruzual , g. , & charlot , s. 2003 , , 344 , 1000 da cunha , e. , charlot , s. , & elbaz , d. 2008 , , 388 , 1595 calzetti , d. , kennicutt , r. c. , engelbracht , c. w. , et al . 2007 , , 666 , 870 cluver , m. e. , appleton , p. n. , ogle , p. , et al . 2013 , , 765 , 93 colless , m. , dalton , g. , maddox , s. , et al . 2001 , , 328 , 1039 cutri , r. m. , wright , e. l. , conrow , t. , et al . 2012 , explanatory supplement to the _ wise _ all - sky data release products , 1 donoso , e. , yan , l. , tsai , c. , et al . 2012 , , 748 , 80 driver , s. p. , norberg , p. , baldry , i. k. , et al . 2009 , astronomy and geophysics , 50 , 050000 driver , s. p. , hill , d. t. , kelvin , l. s. , et al . 2011 , , 413 , 971 elbaz , d. , daddi , e. , le borgne , d. , et al . 2007 , , 468 , 33 ellison , s. l. , mendel , j. t. , patton , d. r. , & scudder , j. m. 2013 , , 435 , 3627 farrah , d. , bernard - salas , j. , spoon , h. w. w. , et al . 2007 , , 667 , 149 gunawardhana , m. l. p. , hopkins , a. m. , sharp , r. g. , et al . 2013 , , 433 , 2764 gunawardhana , m. l. p. , hopkins , a. m. , sharp , r. g. , et al . 2011 , , 415 , 1647 gurkan , g. , hardcastle , m. j. , & jarvis , m. j. 2013 , arxiv:1308.4843 hill , d. t. , kelvin , l. s. , driver , s. p. , et al . 2011 , , 412 , 765 holwerda , b.w . & blyth , s .- l . 2010 , arxiv:1007.4101 hogg , d. w. , bovy , j. , & lang , d. 2010 , arxiv:1008.4686 hopkins , a. m. , driver , s. p. , brough , s. , et al . 2013 , , 430 , 2047 houck , j. r. , weedman , d. w. , le floch , e. , & hao , l. 2007 , , 671 , 323 jarrett , t. h. , masci , f. , tsai , c. w. , et al . 2013 , , 145 , 6 jarrett , t. h. , masci , f. , tsai , c. w. , et al . 2012 , , 144 , 68 jarrett , t. h. , cohen , m. , masci , f. , et al . 2011 , , 735 , 112 jarrett , t. h. , chester , t. , cutri , r. , et al . 2000 , , 119 , 2498 jones , d. h. , saunders , w. , colless , m. , et al . 2004 , , 355 , 747 kaneda , h. , onaka , t. , sakon , i. , et al . 2008 , , 684 , 270 kennicutt , r. c. , jr . , armus , l. , bendo , g. , et al . 2003 , , 115 , 928 lara - lpez , m. a. , hopkins , a. m. , lpez - snchez , a. r. , et al . 2013 , , 434 , 451 lawrence , a. , warren , s. j. , almaini , o. , et al . 2007 , , 379 , 1599 lee , j. c. , hwang , h. s. , & ko , j. 2013 , , 774 , 62 masci , f. 2013 , astrophysics source code library , 2010 marsh , k. a. , & jarrett , t. h. 2012 , _ pasa _ , 29 , 269 martin , d. c. , fanson , j. , schiminovich , d. , et al . 2005 , , 619 , l1 meidt , s. e. , schinnerer , e. , knapen , j. h. , et al . 2012 , , 744 , 17 meyer , m. j. , zwaan , m. a. , webster , r. l. , et al . 2004 , , 350 , 1195 noeske , k. g. , weiner , b. j. , faber , s. m. , et al . 2007 , , 660 , l43 popescu , c. c. , tuffs , r. j. , dopita , m. a. , et al . 2011 , , 527 , a109 reid , i. n. , brewer , c. , brucato , r. j. , et al . 1991 , , 103 , 661 rieke , g. h. , alonso - herrero , a. , weiner , b. j. , et al . 2009 , , 692 , 556 robotham , a. s. g. , norberg , p. , driver , s. p. , et al . 2011 , , 416 , 2640 robotham , a. , driver , s. p. , norberg , p. , et al . 2010 , _ pasa _ , 27 , 76 rodighiero , g. , cimatti , a. , gruppioni , c. , et al . 2010 , , 518 , l25 roming , p. w. a. , kennedy , t. e. , mason , k. o. , et al . 2005 , , 120 , 95 schlegel , d. j. , finkbeiner , d. p. , & davis , m. 1998 , , 500 , 525 stern , d. , assef , r. j. , benford , d. j. , et al . 2012 , , 753 , 30 taylor , e. n. , hopkins , a. m. , baldry , i. k. , et al . submitted taylor , e. n. , hopkins , a. m. , baldry , i. k. , et al . 2011 , , 418 , 1587 wijesinghe , d. b. , da cunha , e. , hopkins , a. m. , et al . 2011 , , 415 , 1002 wijesinghe , d. b. , hopkins , a. m. , sharp , r. , et al . 2011 , , 410 , 2291 werner , m. w. , roellig , t. l. , low , f. j. , et al . 2004 , , 154 , 1 wheelock , s. l. , gautier , t. n. , chillemi , j. , et al . 1994 , nasa sti / recon technical report n , 95 , 22539 wright , e. l. , eisenhardt , p. r. m. , mainzer , a. k. , et al . 2010 , , 140 , 1868 xu , c. k. , shupe , d. l. , bthermin , m. , et al . 2012 , , 760 , 72 yan , l. , donoso , e. , tsai , c .- w . , et al . 2013 , , 145 , 55 in order to understand the photometry of resolved and partially resolved sources in _ wise _ we use several diagnostics . since _ wise _ is calibrated in the vega system , images and all - sky catalogs , all comparisons are done as such . 1 . the source is rotated to determine the optimal 2nd moment parameters ( rmajor , rminor , position angle of major axis ) 2 . the source is divided along the minor axis into two halves : positive x , and negative x , where the central x position is the nominal position of the source 3 . compute the 2nd order intensity - weighted moment for each half and derive the rmajor for each half : rmajor@xmath107 , rmajor@xmath108 4 . rfuzzy @xmath109 minimum(rmajor@xmath107 , rmajor@xmath108 ) . if rfuzzy @xmath46 rfuzzy@xmath110 , then the source is unresolved . the rfuzzy@xmath110 is determined by measuring sources that are expected to be unresolved . for example , figure [ rfuzz ] shows the rfuzzy values for _ wise _ sources in the gama g12 region with _ w1rchi2_@xmath462 which preferentially selects unresolved sources . the typical value for rfuzzy is @xmath2510.0 for high signal to noise point sources ( e.g. stars ) , but the distribution can be used to identify the limit where resolvedness can be determined . a power series fit to the 2@xmath43 mean of the distribution yields the rfuzzy@xmath110 curve points that lie above this curve are classified as resolved . rfuzzy can be defined in terms of isophotal signal to noise or isophotal magnitude . given the differing exposure coverage ( number of _ wise _ frames that samples a given patch of sky ) for the entire sky , the sensitivity limits vary depending on location on the sky . for this reason , a magnitude - dependent rfuzzy@xmath110 curve will be sensitive to the _ wise _ coverage and would need to be derived for each region . however , a signal to noise function is robust against this and therefore the curve shown in figure [ rfuzz]a is used . as a test of the performance of the rfuzzy parameter , figure [ rfuzz_perf ] plots rfuzzy for a sample from gama g12 expected to be dominated by resolved systems ( based on _ w1rchi2 _ ) . the majority of sources with _ w1rchi2_@xmath1015 lie above the curve defined by the point sources in figure [ rfuzz]a . in figure [ delmag_rfuzz_sn ] we test 2590 sources in g15 chosen using the prescription of 3 , i.e. most of which are expected to be resolved in w1 . this shows that the largest offset between the w1iso and _ w1mpro _ photometry measurements occurs for sources deemed resolved by the rfuzzy parameter . this behavior is consistent with sources whose flux is underestimated by profile fitting i.e. resolved . unresolved sources show the expected scatter around 0 . in figure [ delmag_rfuzz_sn]c we see that few sources are resolved in the w3 band . in order to establish the reliablity of isophotal photometry , we use our test sample ( 2590 potentially resolved sources from the g15 gama field ) . in figure [ delmag_snr ] we compare the difference between w@xmath15iso and _ w@xmath15mpro _ photometry as a function of s / n ratio . at low s / n the isophotal measurement becomes unreliable in some cases the fluxes are inflated , notably for w3 and w4 due to contamination from background sources and _ w@xmath15mpro _ provide the most robust measurement . in figure [ delmag_rchi_sn ] we again use the difference between w@xmath15iso and _ w@xmath15mpro _ photometry to illustrate that the sensitivity of the w1 and w2 bands prevents the _ w1rchi2 _ and _ w2rchi _ values from acting as a reliable discriminator of resolvedness . for w3 and w4 , however , it can be used and the limits derived from the plots are indicated . we explore the relationship between the isophotal photometry and the _ w@xmath15gmag _ photometry for sources in the 2mass xsc . since the _ w@xmath15gmag_s are measured using elliptical apertures with radii scaled to twice the 2mass radii , this provides an indication of the additional sensitivity . we note that _ w@xmath15gmag_s can be contaminated by nearby objects since no attempt is made to remove neighboring sources . in figure [ resplots ] we show the difference between _ w@xmath15gmag _ and w@xmath15iso photometry for resolved galaxies in g12 and g15 . this shows that for fainter sources _ w@xmath15gmag _ is brighter than w@xmath15iso , likely due to contamination . finally , in figure [ radplots ] the _ wise _ isophotal and 2mass isophotal radii of resolved sources which clearly shows that the 1@xmath43 isophotal radii are systematically larger than the 2mass radii by a factor of 2 to 2.5 . the largest offset occurs for the most compact 2mass sources probably due to the increased sensitivity of wise in the w1 and w2 bands resolving more of the galaxy relative to 2mass .

we have observed the prototypical wide - angle tail radio galaxy 3c465 with _ chandra _ and _ xmm - newton_. x - ray emission is detected from the active nucleus and the inner radio jet , as well as a small - scale , cool component of thermal emission , a number of the individual galaxies of the host cluster ( abell 2634 ) , and the hotter thermal emission from the cluster itself . the x - ray detection of the jet allows us to argue that synchrotron emission may be an important mechanism in other well - collimated , fast jets , including those of classical double radio sources . the bases of the radio plumes are not detected in the x - ray , which supports the model in which these plumes are physically different from the twin jets of lower - power radio galaxies . the plumes are in fact spatially coincident with deficits of x - ray emission on large scales , which argues that they contain little thermal material at the cluster temperature , although the minimum pressures throughout the source are lower than the external pressures estimated from the observed thermal emission . our observations confirm both spatially and spectrally that a component of dense , cool gas with a short cooling time is associated with the central galaxy . however , there is no evidence for the kind of discontinuity in external properties that would be required in many models of the jet - plume transition in wats . although the wat jet - plume transition appears likely to be related to the interface between this central cool component and the hotter intra - cluster medium , the mechanism for wat formation remains unclear . we revisit the question of the bending of wat plumes , and show that the plumes can be bent by plausible bulk motions of the intra - cluster medium , or by motion of the host galaxy with respect to the cluster , as long as the plumes are light . galaxies : active x - rays : galaxies galaxies : individual : 3c465 galaxies : jets radiation mechanisms : non - thermal

astro-ph0502575

wide - angle tail radio galaxies ( wats ) present a number of interesting problems for our understanding of the dynamics of extragalactic radio sources and in particular their interaction with the external medium . here we define the class as being sources that are associated with a cluster dominant galaxy , at or near the cluster centre , and that are characterized in the radio by twin , well - collimated jets that abruptly flare into long , often sharply bent plumes or tails . ( throughout this paper we use the term ` jet ' to refer to the inner , well - collimated feature , and ` plume ' to refer to the outer parts of the source . ) the earliest problem associated with wats was identified by burns ( 1981 ) and eilek et al . ( 1984 , hereafter e84 ) : given that wat host galaxies are cluster - centre objects and should be close to the bottom of the cluster potential well ( and so not moving at high speed with respect to the external medium ) what gives rise to the often sharp bends in the plumes ? this problem is in fact not unique to sources that are wats by our definition ; not all cluster - centre sources with bent plumes exhibit the large - scale , narrow , well - collimated jets of a wat . a problem that _ is _ unique to wats ( also identified by e84 ) is how , and why , these jets make the transition into broad plumes where they do : what is it about the external environment that sets the location for the jet - plume transition ? or , since these objects are in general radio - luminous enough to be classical double ( frii ) sources and their jets are very similar to those of friis , we can ask a related question ( e.g. hardcastle 1998 ) : what is it about the external environment that causes these sources to be wats and not friis ? radio , optical and x - ray observations have allowed some progress to be made in answering these questions . large - scale bulk motions in the intra - cluster medium ( icm ) are expected if the host cluster is not relaxed , and the ram pressure from these can bend the plumes : _ einstein _ and _ rosat _ observations have shown that there is indeed x - ray substructure in the host clusters of wats ( e.g. burns et al.1994 ; gmez et al . 1997a ) and this is backed up by optical spectroscopy that shows there is substantial substructure in the galaxy velocity distributions of a few well - studied clusters hosting wats ( pinkney et al . 1993 ; gmez et al . substructure in the host clusters will also give rise to non - radial buoyancy forces that can play a part in bending the plumes ( e.g. sakelliou , merrifield & mchardy 1996 ) . observational input to the questions related to the jet - plume transition has been more limited , but hardcastle ( 1998 ) and hardcastle & sakelliou ( 2004 ) have shown that the jets often terminate in compact structures similar to the hotspots of frii radio galaxies , supporting a picture ( hardcastle 1999 ) in which the current jet - plume transition , where it occurs at a single point , is a result of the interaction between the jet and the edge of the previously existing plume . hardcastle & sakelliou also provided the first direct evidence that the external environment is responsible for the unique structure of wats , by demonstrating an inverse correlation between the jet termination distance and the temperature of the host cluster . however , the physics that governs the location of the base of the plume is still unclear . modellers have concentrated on the idea that it is associated with some localized feature of the external medium , such as propagation across an igm / icm interface or a shock front ( e.g. loken et al . 1995 ; hooda & wiita 1996 ) or interaction with discrete clumps in the external medium ( e.g. higgins , obrien , & dunlop 1999 ) . x - ray observations with the previous generation of observatories had insufficient resolution and sensitivity to test these models . however , _ chandra _ and _ xmm _ have now been used to observe a number of wat host clusters , allowing the observational picture in the x - ray to be updated . jetha et al . ( 2005 ) present observations of two wat host clusters with _ chandra _ , showing that both contain a cool central gas component and that the jet enters the plume close to the transition between this and the hotter gas in the rest of the host cluster . while this has some similarities to the picture in which a jet - plume transition is a result of a propagation across some structure in the igm / icm interface , the detailed _ chandra _ observations show that there are no discontinuities in the physical conditions in the clusters ( as would be expected if the jet - plume transition was a result of propagation across a discrete feature in the external medium , such as a shock or cold front ) and no discrete x - ray emitting features associated with the plume bases . an association between the jet - plume transition and the cold central gas seems likely as a result of this work , but the mechanism for wat formation is still not clear . the observations of jetha et al . were also not sensitive enough to study the x - ray properties of the radio components themselves . in wats the jets , like those in friis , are thought to remain relativistic , well - collimated flows ( ` type ii ' jets ) until they undergo rapid deceleration at a hotspot . in the much commoner twin - jet fri sources , the initially type ii jet flow undergoes significant bulk deceleration on scales of a few kpc , slowing from bulk relativistic speeds to speeds that are trans - sonic or sub - sonic with respect to the external medium ( bicknell 1994 ; laing & bridle 2002 ) and forming a ` type i ' jet that may in some cases make a smooth transition into a plume similar to those of wats . in all cases that have been observed in detail this bulk deceleration in fris is accompanied by high - energy particle acceleration that gives rise to x - ray synchrotron emission ( e.g. hardcastle et al . 2001 , worrall et al . 2001 , hardcastle et al.2002 ) . however , the inner jets of fri sources , while they should still be relativistic , are often also strong sources of x - ray emission ( hardcastle et al . 2001 , 2003 ) . wat jets should be similar to the inner jets of fris , and so we might expect to detect synchrotron x - ray emission from them , particularly on small scales . on the other hand , wat plumes should _ not _ be similar to the brighter outer jets of fris in the deceleration region , since the model outlined by hardcastle & sakelliou ( 2004 ) requires the jet deceleration in the plumes to take place at a single discrete location . investigating these issues requires sensitive x - ray observations of more wat systems . 3c465 , the subject of this paper , is the closest wat in the northern sky ( @xmath0 ) and the best studied at radio frequencies ( riley & branson 1973 ; leahy 1984 ; e84 ; eilek & owen 2002 ; hardcastle & sakelliou 2004 ) , to the extent that it is often taken to be the prototype object of its class . its host galaxy is the d galaxy ngc 7720 , the dominant galaxy of the cluster abell 2634 . a2634 has been well studied in the x - ray , initially with _ einstein _ ( jones et al . 1979 ; e84 ; jones & forman 1984 ) and later with the _ rosat _ pspc ( schindler & prieto 1997 ) and hri ( sakelliou & merrifield 1998 , 1999 ) and with asca ( e.g. fukazawa et al . 1998 ) . the velocity properties of the galaxies in the cluster have also been studied in detail ( pinkney et al . 1993 , scodeggio et al . 1995 ) . from these studies it has been possible to show that a2634 is a non - relaxed cluster , with an x - ray extension around the central cluster galaxy perpendicular to the jet axis ( e.g. schindler & prieto 1997 ) although there is little evidence in the galaxy distributions and velocity data for a recent merger ( scodeggio et al . 1995 ) . on small scales , sakelliou & merrifield ( 1999 ) argued that the compact x - ray core seen by the _ hri was extended on scales of a few arcsec , and that this implied detection of a small - scale igm component . sakelliou & merrifield ( 1998 ) showed that the individual galaxies of the host cluster were of low luminosity in the x - ray , arguing that this implied that hot gas had been stripped from the galaxies in their motion through the igm . because of this wealth of existing observational data , 3c465 is an excellent target for observations with the new generation of x - ray observatories aimed at addressing the outstanding questions described above . in this paper we report on the results of our _ chandra _ and _ xmm - newton _ observations . throughout the paper we use a concordance cosmology with @xmath1 km s@xmath2 mpc@xmath2 , @xmath3 and @xmath4 . at the redshift of 3c465 , 1 arcsec corresponds to 590 pc . spectral indices @xmath5 are the energy indices and are defined in the sense @xmath6 . the photon index @xmath7 is @xmath8 . all fits to the x - ray data include the effects of galactic absorption with a column density of @xmath9 @xmath10 . j2000 co - ordinates are used throughout . 3c465 was observed with the epic cameras on _ xmm _ , using the medium filter , for 11160 s ( mos1 ) 11162 s ( mos2 ) and 8725 s ( pn ) on 2002 jun 22 . the data were filtered using sas to include only the standard flag set , and remove bad columns and rows ; we retained the out of field of view events . high background events were filtered on the count rate above 10 kev using the same criteria as were used for the background dataset we used in subsequent analysis ( read & ponman 2003 ) , i.e. an upper count rate in the 1015 kev energy range with standard flags and pattern=0 of 1 count s@xmath2 ( pn ) and 0.35 count s@xmath2 ( mos ) : this did not significantly affect the mos cameras , but flagged out a substantial fraction of the pn data . the effective exposure times after filtering were 11054 s ( mos1 ) 11105 s ( mos2 ) and 6735 s ( pn ) . after filtering , the data were vignetting - corrected using the sas task _ evigweight _ , and spectra and images were generated using the _ evselect _ task . we used the techniques described by croston et al . ( 2003 ) to interpolate over chip gaps and bad columns to make images in the energy range 0.37.0 kev for presentation . although the source and background datasets had been filtered using the same criteria , the count rate in the out of field of view regions for both mos and pn was significantly higher in our observations than in the background dataset . this implies a higher contribution from particles in our data , which ( if uncorrected ) would bias our extracted spectra . the reason for this discrepancy appears to be that the effective spectrum of the particle background , as measured by _ xmm _ , differs in our data and in the background files ( where it is of course an exposure - weighted average over many observations ) . we generated an approximate correction for the particle background excess in our data by taking the difference between the out of field of view spectra in our data and those in the read & ponman background files . a suitably scaled version of this spectrum ( taking account of the weighting applied to the data ) was then added to the _ evigweight_-corrected background spectra generated for each observation . this had the effect of stabilizing the temperature fitted to large - area regions of the extended x - ray source ( section [ cluster ] ) . while this approach to background correction should provide an accurate correction for the non - x - ray contribution to the observed background , it differs from the double - subtraction method of arnaud et al . ( 2002 ) in that it does not attempt to correct for the difference between the level of the cosmic x - ray background in the background datasets and that in the observation of 3c465 . the reason we do not attempt to do this is that the cluster fills the field of view , so that it is impossible to define a local blank - sky background that does not include cluster emission : any attempt to apply the method of arnaud et al . would tend to over - estimate the cosmic background excess and consequently cause us to under - estimate the true normalization of the models fitted to the data . the systematic error introduced by omitting this double - subtraction correction is small in the case of 3c465 and should predominantly affect low energies , and we have verified that applying the double - subtraction method to the spectra we extract in section [ cluster ] does not significantly affect fitted parameters such as mekal model temperature , though of course it does affect the normalization of the fitted models . we observed 3c465 with _ chandra _ on 2004 aug 31 for 49528 s. the source was positioned near the standard aim point on the back - illuminated s3 ccd . we reprocessed the data in the standard way to apply the latest calibration files ( using ciao 3.1 and caldb 2.28 ) . as the data were taken in vfaint mode , we used ` vfaint cleaning ' to identify and reject some background events . we also removed the 0.5-pixel event position randomization . there were no times of unusually high background , so we were able to use all of the available data . we used events in the energy range 0.55 kev to construct images ( throughout the paper this energy range is used for _ chandra _ imaging ) . the high - resolution radio images used are taken from hardcastle & sakelliou ( 2004 ) , and were made with the nrao very large array ( vla ) at 8.4 ghz . we also make use of a 1.4-ghz vla image taken from the 3crr atlas of leahy , bridle & strom with a resolution of 5.4 arcsec . _ chandra _ spectra were extracted and corresponding response and ancillary response matrices constructed using ciao . _ xmm _ spectral extraction was carried out using sas , and an on - axis ancillary response matrix was generated using the task _ arfgen _ and used for every spectrum . the response matrices used were the ` canned ' responses supplied on the _ xmm _ web site . the use of on - axis ancillary response matrices is made possible by the vignetting correction applied with _ spectral model fitting was carried out in xspec , in the energy range 0.47.0 kev ( _ chandra _ ) and 0.37.0 kev ( _ xmm _ ) . in all cases , except where otherwise stated , the extracted spectra were grouped so that each fitting bin had @xmath11 counts after background subtraction . errors quoted throughout are the @xmath12 value for one interesting parameter , unless otherwise stated . in the _ chandra _ image there is a clear detection of an x - ray core and emission associated with the jet , as well as small - scale extended emission ( fig . [ chandra - image ] ) . on larger scales , thermal emission from the host cluster is clearly visible in the _ chandra _ data and there are detections of several of the individual galaxies of the cluster . no significant excess x - ray emission is seen around the points where the narrow inner jets enter the radio plumes . the _ xmm _ observations ( fig . [ xmm - image ] ) detect the x - ray core and small - scale thermal emission and the thermal emission from the cluster , as well as some of the cluster galaxies and the background cluster known as cl-37 ( scodeggio et al . 1995 ) . in the following sections we discuss each component in turn . 8.5 cm 8.5 cm there is small - scale thermal emission associated with the inner parts of ngc 7720 ( section [ small - sc ] ) , and so measurements of the nuclear x - ray emission need to be carried out with care . we initially used a small region ( a circle of radius 1.5 arcsec centred on the peak of the x - ray emission ) to extract a spectrum from the _ chandra _ dataset . background was taken from a concentric annulus between 1.5 and 2.5 arcsec . this spectrum was very poorly fitted ( @xmath13 for 28 degrees of freedom ) with a single power law model with galactic absorption , and the strong residuals below 1 kev suggested that there was some contribution from thermal emission even in this small region . a model consisting of a thermal ( mekal ) component with 0.5 solar abundance and an unabsorbed power law gave a much better fit ( @xmath14 for 26 degrees of freedom ) , but required an unusually flat photon index ( @xmath15 ) for the power - law component . when the power law was allowed to have some intrinsic absorption , a better fit was obtained ( @xmath16 for 25 d.o.f . ) : in this case the temperature of the thermal component was @xmath17 kev , the power - law photon index was @xmath18 , and the excess absorbing column was @xmath19 @xmath10 . no improvement in the fit was obtained by including further components , such as a second unabsorbed power law . imaging the core in hard and soft bands ( above and below 1 kev ) suggests that it is indeed much more point - like in the hard band than the soft band , supporting the model implied by this fit . the power - law component contributes 260 counts in this model , or about 1/3 of the total in the extraction region , and its unabsorbed 1-kev flux density is @xmath20 njy , taking into account a correction for the small aperture and local background subtraction estimated from psf simulations using the chart and marx tools . errors quoted here are @xmath12 for two interesting parameters to take account of the strong relationship between absorbing column and normalizing flux . to compare the _ chandra _ and _ xmm _ results , we extracted a spectrum for both using a circle of radius 30 arcsec , taking background from an annulus between 30 and 40 arcsec . ( this region includes emission from the companion galaxy , ngc 7720a , which is 14 arcsec to the north , but as discussed in section [ compgs ] this is a weak x - ray source and does not contribute significantly to the fits . ) the results are similar to those found above : the best joint fit to the thermal / absorbed power - law model , with all parameters the same for both _ xmm _ and _ chandra _ datasets , has @xmath21 for 122 d.o.f . and gives a somewhat higher @xmath22 kev , an intrinsic absorbing column for the power - law component @xmath23 @xmath10 , power - law photon index @xmath24 , and unabsorbed 1-kev power - law flux density of @xmath25 njy . the improvement in the fit obtained by allowing the parameters of the power law component in the _ chandra _ and _ xmm _ datasets to vary independently is marginal ( @xmath26 for 119 d.o.f . ) and the best - fitting parameters are similar . a more significant improvement ( @xmath27 for 120 d.o.f . ) is obtained by fitting two mekal models with different temperatures together with the absorbed power law , suggesting that there may be temperature structure in the extraction region : in this model the lower temperature is @xmath28 kev , and so is consistent with the temperature measured in the inner 1.5 arcsec , while the higher temperature is @xmath29 kev . we discuss temperature structure in more detail in section [ small - sc ] . the normalizations for the power - law component in both these fits agree with that determined from the small - aperture _ chandra _ fit described above within the large fitting uncertainties . the _ chandra _ detection of x - ray emission from the inner part of the n jet of 3c465 is the first x - ray detection of a wat jet . x - ray emission appears to come predominantly from the two brightest knots of the radio jet , denoted nj1 by hardcastle & sakelliou ( 2004 ) ( fig . [ chandra - image ] ) . it is hard to assign a statistical significance to the detection of the inner knot ( labelled knot 1 in fig.[chandra - image ] ) , because of the difficulty in measuring a good background so close to the bright , asymmetrical central x - ray emission from the core and central thermal component : our best estimate , based on local matched background circles , is that it contains @xmath30 counts . knot 2 , which corresponds to the brightest discrete feature of the radio jet , contains @xmath31 counts , and there are a few more counts plausibly associated with the jet in the bright jet region beyond it . no x - ray emission from the jet is detected beyond 7.5 arcsec ( 4.4 kpc ) . knot 2 is resolved transversely in the radio , with a deconvolved gaussian fwhm of 0.3 arcsec ; this suggests that it should be essentially unresolved to _ chandra _ , and the observed x - ray emission is consistent with that suggestion . there are too few counts in the region of the jet from knot 2 onwards to extract a spectrum and carry out detailed model fitting , but the two - bin spectrum that can be extracted from the region marked on the figure suggests a steep x - ray power - law photon index ( @xmath32 ) . assuming this photon index and galactic absorption , the x - ray flux density from this region of the jet is @xmath33 njy . the corresponding radio flux density is 6.6 mjy at 8.4 ghz , so that the two - point radio - x - ray spectral index is 0.92 . the unabsorbed jet flux in the _ rosat _ band is @xmath34 ergs @xmath10 s@xmath2 , corresponding to a luminosity of @xmath35 ergs s@xmath2 at the distance of 3c465 ( assuming no beaming ) . the inner region of the jet is clearly different from the outer region ( beyond 7.5 arcsec ) : if the straight region of the outer jet were similar in x - ray properties to the inner parts , we would expect to have detected it at the level of @xmath36 _ chandra _ counts , whereas fewer than 30 ( @xmath37 ) are detected . however , the outer jet shows no clear knots of radio emission until the base of the plume ( see below ) . there is no significant detection of x - ray emission from the radio - bright bases of the plumes in 3c465 ( fig . [ pbase ] ) . the strongest limit on the radio / x - ray ratio from these regions comes from the bright southern plume base . in a circular region around the brightest radio emission from this region we can set an upper limit of 25 _ chandra _ counts , corresponding to @xmath38 njy assuming a power - law model with a photon index of @xmath39 . the radio flux density in the same region is 143 mjy , so that the two - point radio - x - ray spectral index is @xmath40 . this is steeper than is observed for any detected fri jet base ; if the plume base had had a more typical value for fri jets of @xmath41 ( similar to the knots in the inner jet , see above ) , it would have been detected at the level of @xmath42 counts . as pointed out by hardcastle & sakelliou ( 2004 ) , 3c465 does not show any obvious hotspot - like jet termination features , but the best candidate feature , the knot at the end of the northern radio jet ( nj2 in their notation : labelled in fig . [ pbase ] ) is also not significantly detected in the x - ray . to characterize the small - scale thermal emission implied by the images and the spectral fits of section [ core ] , we carried out radial profile fitting to the inner 35 arcsec of the _ chandra _ dataset , taking background from the region between 35 and 45 arcsec and excluding a pie slice at the position angle of the jet and the emission from the nearby companion galaxy ( see section [ compgs ] ) . we use the psf parametrization of worrall et al . the radial profile is best fitted with a combination of a small @xmath43 model and a point - like component ( fig . [ proplot ] ) . the fit is good ( @xmath44 for 40 d.o.f . ) and the best - fitting model parameters are @xmath45 , @xmath46 arcsec ( errors are @xmath12 for 2 interesting parameters ) . in this model the point - like component contributes @xmath47 counts in the 0.55 kev energy range : this compares well with the number estimated from spectral fitting in section [ core ] . these observations confirm the suggestion based on _ rosat _ data ( sakelliou & merrifield 1999 ) that there is a component of extended x - ray emission associated with ngc 7720 . we used spectral fitting in concentric annuli in the inner 30-arcsec region to investigate whether there was evidence for temperature structure in this small - scale thermal component . identical background regions , between 30 and 40 arcsec , were taken for each fit . the results are shown in table [ tgrad ] . either simply fitting spectra to each annulus , or using the xspec deprojection model , we find evidence that there is a temperature gradient in the small - scale thermal material in the centre of a2634 . using the temperatures measured above in combination with the radial profile , we estimate the central density of the small - scale thermal material in 3c465 to be @xmath48 m@xmath49 , while the central pressure is @xmath50 pa . the central cooling time of this material is short ( @xmath51 years ) . these conditions are very similar to those found in the central components of more normal , non - wat fri radio galaxies hosted by groups ( e.g. hardcastle et al . 2002 , 2005 ) . .temperatures in the central region of 3c465 measured with _ chandra _ [ cols= " < , > , > , > , > " , ] the fit to the central bin includes an obscured power - law component as discussed in section [ core ] . abundance is fixed to 0.5 solar ( see the text ) . the total @xmath52 value for the deprojected model was 1671/1847 . it will be seen that the deprojected and independent fits give similar temperatures , around 4.55 kev , for the outer regions of the cluster . these temperatures are in reasonable agrement , within the errors , with the temperatures estimated from _ rosat _ pspc data by schindler & prieto ( 1997 ) , though somewhat higher than the cluster temperature of @xmath53 kev derived from _ asca _ data by fukazawa et al.(1998 ) . it remains possible that the temperatures we measure are contaminated by excess particle background , in spite of the steps we have taken to subtract this off ( section [ xmm - intro ] ) . however , we see no sign of an increase in temperature at large radii , as we might expect in such a scenario ( and as we do in fact see without the particle background corrections ) . instead , the temperatures are largely consistent with a constant value of @xmath54 kev beyond about 1 arcmin from the core . combining the _ xmm _ deprojection with the _ chandra _ deprojection of section [ small - sc ] , we obtain the temperature profile shown in fig . [ tprofile ] . whereas the _ xmm _ data alone might have suggested a significant temperature jump at around the jet termination distance in 3c465 ( @xmath55 arcsec ) the two datasets taken together strongly suggest a smooth temperature gradient as a function of distance , probably levelling out at @xmath56 kev . the deprojection allows us to estimate particle densities in the deprojection annuli , which we can compare with those estimated from the radial profile analysis using the method of birkinshaw & worrall ( 1993 ) . the radial profile values are necessarily calculated assuming a single conversion factor between count density and volume - normalized emission measure , and so can not take account of temperature structure : we carry out the conversion assuming instrumental responses appropriate for temperatures of 4.7 kev for the outer region modelled with _ xmm _ data and 1.0 kev for the _ chandra _ data that describe the inner region . on the other hand , the densities estimated using the deprojection model are emission - weighted densities and can only be at best representative of the density in the bin . the proton densities estimated using the two models are plotted as a function of radius in fig . [ dprofile ] ( we assume for the purposes of the plot that the transition between the two @xmath43 models happens when their densities become equal at around 20 arcsec from the core ) . the plot shows that the agreement between the densities estimated using the two methods is reasonable , except perhaps in the outer bin , which would be expected to be most affected by the assumption of the deprojection algorithm that no cluster emission exists outside the deprojection region . we can therefore have reasonable confidence in pressures estimated using either method . the agreement between the two methods also suggests that the physical conditions we estimate are reliable in spite of the fact that we find lower @xmath43 values and core radii than other workers . pressure as a function of radius , estimated from the deprojection and from the radial profile fitting , is plotted in fig . [ pprofile ] . to convert the density estimated from radial profile fitting to a pressure we have assumed a temperature which is constant at 0.66 kev within 2 arcsec and constant at 4.7 kev beyond 200 arcsec , and has a log - linear profile between those two radii ( i.e. @xmath57 ) : such a temperature profile is a reasonable description of the profile plotted in fig . [ tprofile ] , though others could be devised . it can be seen that the pressure estimates are again reasonably consistent , as expected . also plotted on fig . [ pprofile ] ( dashed lines ) are the minimum pressures in various components of the radio source . these are calculated from flux measurements from the 1.4 ghz and 8.4 ghz maps on the assumption of a synchrotron spectrum with @xmath58 , @xmath59 , a low - energy electron energy index of 2 ( corresponding to @xmath60 ) and a break in the electron spectrum determined by the fluxes in the two maps . we assume no relativistic protons , a filling factor of unity , and a cylindrical geometry for all features . for the bright northern jet , we have estimated the minimum pressure in two ways : firstly assuming no projection or beaming ( as was the case for all the other components ) and secondly assuming an effective beaming speed of @xmath61 and an angle to the line of sight @xmath62 : this latter is a representative beaming value that reproduces the jet sidedness ratio measured by hardcastle & sakelliou ( 2004 ) . these pressures are essentially true minimum pressures : only very implausible assumptions for the electron energy spectra can reduce them significantly , and then only by a factor @xmath63 . fig . [ pprofile ] shows that the inner jet is the only component , on either of these two models , that appears to have a higher minimum internal pressure than the external thermal material over some of its length . all the other regions of the source are underpressured with respect to the external pressure by factors of between about 3 and 30 , as is typically found for the large - scale components of fri sources ( hardcastle & worrall 2000 and references therein ) . this conclusion is different from that of e84 partly because our estimated minimum pressures are slightly lower , but mostly because our external pressures are higher , as a result of a higher measured external temperature ( their value of 1.85 kev is lower than any modern estimate ) and a better - constrained model for the external density . pinkney et al . ( 1993 ) and scodeggio et al . ( 1995 ) both report the existence of a cluster of galaxies 12 arcmin to the nw of ngc 7720 , with a redshift around 0.124 . this cluster is detected in x - rays in the mos1 and mos2 cameras of _ xmm _ ( it lies on the very edge of the pn field of view ) . extracting a spectrum from a 75-arcsec radius circle around the x - ray centroid , with background taken from an adjacent larger circle , we find that the mos1 and mos2 data are well fitted ( @xmath64/d.o.f . = 84/88 ) by a mekal model with @xmath65 kev and abundance @xmath66 solar . the unabsorbed x - ray flux in this region is around @xmath67 ergs @xmath10 s@xmath2 , which is substantially lower than the value predicted by scodeggio et al . from the luminosity - velocity dispersion relationship even after correcting to the cosmology used here , although the extraction region is not large at the distance of cl-37 . however , as the measured temperature would correspond ( e.g. edge & stewart 1991 ) to an expected velocity dispersion in the range 400600 km s@xmath2 , it seems likely that the true velocity dispersion of the cluster is less than the value of @xmath68 km s@xmath2 quoted by scodeggio et al .. given the large uncertainties on the result of scodeggio et al . , their comment that the velocity structure of the cluster is complex , and the large scatter in the @xmath69-@xmath70 relationship , the true velocity dispersion can be estimated to be something closer to @xmath71 km s@xmath2 . it is conventional to believe ( e.g. leahy 1993 ) that the jets of wat sources are physically similar to those in classical double ( frii ) radio galaxies and quasars ; evidence for this comes from their narrow and sometimes one - sided appearance ( as in 3c465 ) , the polarization structure they exhibit and their frequent termination in compact , shock - like features similar to the hotspots in friis ( hardcastle & sakelliou 2004 ) . x - ray emission from frii - like ( ` type ii ' ) jets , particularly in quasars and other sources where the jet is _ a priori _ thought to make a small angle to the line of sight , is often attributed in the literature to inverse - compton scattering of the cosmic microwave background radiation by a jet with a high bulk lorentz factor ( @xmath72 ) , based on the observations that a one - zone synchrotron model can not be fitted to the data and that the inverse - compton process without beaming requires a large departure from the equipartition condition ( e.g. tavecchio et al . 2000 ) . however , the generally accepted picture of x - ray emission from fri - like ( ` type i ' ) jets is that they are due to synchrotron emission related to the strong deceleration of the jet flow expected on other grounds in the inner few kpc of these systems ( e.g. hardcastle et al.2001 , 2002 ) ; in these cases a one - zone synchrotron model is often a good fit to the overall jet spectrum ( although the details of jet structure may in fact be complex on scales that can not be resolved in most sources : hardcastle et al . 2003 ) while a beamed inverse - compton model requires implausible jet parameters . the detection of 3c465 s inner jet in x - rays , at a level similar to what would be predicted using a simple synchrotron model based on observations of type i jets , lends support to the idea that some type ii jets can be synchrotron sources . there is already evidence for this from the comparatively strong detections of the pre - flaring jets , thought to be type ii , in some fri sources , notably 3c66b and cen a ( hardcastle et al . 2001 , 2003 ) as well as x - ray detections of components of the jets of some frii sources ( e.g. 3c219 , comastri et al . 2003 ; 3c452 and 3c321 , hardcastle et al . 2004 ; 3c403 , kraft et al . 2005 ) for which it is highly implausible that a beamed inverse - compton model can be an explanation . 3c465 is particularly interesting in this context for three reasons : ( 1 ) it is generally accepted that wat jets are type ii , ( 2 ) the emission from the jet is not confused by inverse - compton emission from any surrounding lobes , and ( 3 ) there is no confusion between jet knots and hotspots , which can also be synchrotron x - ray sources . 3c465 s detection thus gives us a reason to reiterate a point already made by hardcastle et al . ( 2004 ) and kraft et al . ( 2005 ) : there are clearly x - ray - emitting type ii jets for which the beamed inverse - compton model is _ not _ the explanation . since there is no compelling reason to expect a one - zone synchrotron model to be a good fit in an frii jet in which the loss spatial scale may be two to three orders of magnitude smaller than the region of the jet being considered , observers should be wary of ruling out the possibility that synchrotron emission is responsible for the x - rays from a given type ii jet . on the other hand , the non - detection of the bases of the plumes of 3c465 in x - rays , at a level sufficient to show that their radio to x - ray spectral index is significantly greater than that seen in all fri jets , strongly suggests that these regions are _ not _ analogous to the bases of type i jets . this is as expected in the picture of these objects proposed by hardcastle et al . ( 1998 ) and hardcastle & sakelliou ( 2004 ) , in which the jet decelerates rapidly at a single point close to the base of the plume , rather than there being uniform distributed deceleration over a finite - sized region of the base of the plume . we conclude that the non - detection of the plumes in x - rays in the _ chandra _ observations is evidence that the bases of wat plumes are physically different from the bases of fri type i jets ; wat plumes are probably more similar to the large - scale jets of fris , which typically do not show either strong deceleration or x - ray emission . it is not clear why we do not see x - ray emission associated with the assumed rapid deceleration of the jet at the base of the plume , by analogy with the x - ray emission detected from the hotspots of many low - luminosity frii radio galaxies , thought to be synchrotron in origin ( hardcastle et al . 2004 and references therein ) . features as faint as the brightest knot at the end of the n jet ( nj2 in the notation of hardcastle & sakelliou : see also fig . [ pbase ] ) have been detected in x - ray observations of frii hotspots . however , as we noted above , there is no particularly good candidate feature analogous to a hotspot in the radio images of 3c465 ; the jet flow clearly continues after nj2 . _ chandra _ images of wats that _ do _ exhibit compact bright jet termination structures would be of considerable interest . we begin by noting that the conclusions of jetha et al . ( 2005 ) , based on less sensitive observations , apply to 3c465 as well . there is no evidence in the data either for discrete features associated with the jet termination points or plume bases or for discontinuities in the properties of the external medium at the distances of the jet - plume transition , such as the shock structures in the models of loken et al.(1995 ) ; in particular , our observations show that the apparent temperature discontinuity seen in the inner regions of the sources studied by jetha et al . was probably an unresolved temperature gradient . models that require a discontinuity in the external medium are thus ruled out by observation . the jet termination is on a similar scale to the transition between the central , cool x - ray core and the hotter cluster medium , again as observed by jetha et al . , but as yet we have no clear idea of why this should be the case . as jetha et al.suggest , numerical modelling of realistic ( fast , low - density ) jets and plumes in realistic cluster environments may be the way forward here . it is also interesting to note that the transition occurs close to the point where the jet internal ( minimum ) pressure starts to exceed the external thermal pressure ( section [ cluster ] and fig.[pprofile ] ) . x - ray and radio data of similar quality for a larger number of sources would be required to say whether this is more than a coincidence . our detailed x - ray and radio observations have given a very clear indication that the large - scale plumes of 3c465 behave similarly to the lobes or plumes of normal twin - jet fri sources : the minimum pressures in the lobes are about an order of magnitude below the external thermal pressures . this implies some sort of departure from the minimum - energy condition , either a departure from equipartition between the radiating electrons and magnetic fields , a population of non - radiating relativistic particles , or a contribution to pressure from particles that do not participate in equipartition , such as thermal protons . as usual ( e.g. hardcastle et al . 1998 , croston et al.2003 ) we can not tell which of these is the case . in the inner parts of normal fri jets the jet deceleration is thought to be by entrainment of external material , which has suggested that the entrained material might supply the missing pressure if it could be heated efficiently so as to avoid thermal x - rays from the lobes or plumes ( see e.g. croston et al . 2003 ) . in wat plumes , the jets are not expected to have entrained significant amounts of external material , since they do not show the deceleration - related structures of fri jets , and so this explanation seems less likely ( see also below , section [ bending ] ) . on the other hand , if we assume that equipartition does not hold , it is not clear why such a large departure from equipartition should be required in wat plumes but not in frii lobes , where inverse - compton evidence ( e.g. hardcastle et al . 2002 , belsole et al . 2004 , croston et al . 2004 ) suggests that equipartition is roughly maintained and that the electrons and magnetic fields alone are sufficient to provide pressure balance with the external medium . the departure from equipartition would correspond to around a factor 5 difference between the actual magnetic field strength and the equipartition value ; this is considerably larger than is typically seen in the x - ray - detected lobes of frii radio galaxies ( croston et al . in addition , a departure from equipartition sufficient to make the internal pressure equal to the external pressure in the direction of electron domination would give rise to strong inverse - compton emission from the plumes , at the level of a few thousand _ xmm _ counts , which is not observed the deficits in x - ray count rate seen at the bases of the plumes seem to rule out this model , so that the lobes would have to be magnetically dominated if electrons and magnetic field alone were to supply the required pressure . the third possibility is that the plumes contain an energetically dominant population of relativistic protons : this can not be ruled out directly by the data , but again arguments based on inverse - compton detections suggest that it is not the case in the hotspots ( hardcastle et al . 2004 ) or lobes ( croston et al . 2005 ) of frii sources . we return to the particle content of the plumes in the next section . it is instructive to revisit the arguments of e84 that led to the conclusion that the bending of 3c465 in particular , and of wats in general , was difficult to explain in any model . they assumed that the kinetic energy of the plume must supply the overall observed radio luminosity : that is , the bulk flow is cold and there is _ in situ _ particle acceleration at all points in the plume . this leads to a condition on the density and speed of a non - relativistic plume : @xmath73 ( their eq . 2 ) where @xmath74 is a factor giving the efficiency of conversion between kinetic energy and energy of relativistic electrons , @xmath75 is the plume radius , @xmath76 is its density and @xmath77 is its speed . they assumed _ in situ _ particle acceleration on the basis of estimates of the synchrotron loss timescale at 5 ghz , and so we note to begin with that it may not be necessary to satisfy this condition in its original form . the synchrotron age of material at the end of the plumes is @xmath78 years , based on 1.4 ghz and 330 mhz data taken from the vla archives , if we assume a jaffe & perola ( 1973 ) aged synchrotron spectrum , an injection index of 2.0 and a loss magnetic field strength around the minimum - energy value of 0.4 nt ( taking into account cmb losses ) , and consequently material can be transported out from the jets or plume bases without _ in situ _ reacceleration so long as @xmath79 , or 2500 km s@xmath2 . e84 used the lack of observed depolarization to place a limit ( @xmath80 m@xmath49 ) on the density of thermal material ( effectively , protons ) in the tails . this method of setting a density limit has fallen out of favour in recent years , partly because it does not provide a strong upper limit ( laing 1984 ) and partly because it seems unlikely that the limits it does provide are close to the true values , which may be much lower . it may thus not be appropriate to use a proton density around @xmath81@xmath82 m@xmath49 as a ` typical ' value in the plumes . observationally , there are few other constraints on the internal density available : as discussed above , we know from the x - ray deficits seen in both ordinary fris and in cluster - centre objects ( e.g. brzan et al . 2004 ) that the missing pressure in these sources is _ not _ supplied by thermal protons at the temperature of the external medium , but this only really gives the constraint that the internal density is likely to be much lower than the external one . a true limit comes from considering the minimum possible internal density . if we assume that the wat plumes are in pressure balance and that the internal pressure is supplied by relativistic particles and magnetic field , presumably with some departure from the minimum energy condition , the plume material remains very light compared to the external medium : we require @xmath83 , where @xmath84 , and the effective density of the internal material is @xmath85 , while the external density is @xmath86 , where @xmath87 is the mass per particle in atomic mass units ( about 0.6 ) . the density contrast is thus @xmath88 , or about @xmath89 for the temperature of the 3c465 cluster . the internal density ( in terms of an equivalent proton density ) in this model , which represents a minimum possible internal density if we think that the plumes are not underpressured , would be of the order of 0.04 m@xmath49 , much lower than the ` typical ' values used by e84 . such a light plume is comparatively easy perhaps even too easy to bend by ram pressure ; if the flow speed up the plume , @xmath77 , were comparable to the local external sound speed ( @xmath90 km s@xmath2 ) , bulk motions with speeds of @xmath91 would have a significant effect , while if @xmath77 were comparable to the speed estimated above from spectral ageing arguments , we would require external bulk motions with speeds around @xmath92 . if instead we adopt the luminosity condition ( eq . [ lum ] ) to estimate the flow speed , retaining this low internal density , then we would obtain a bulk speed around @xmath93 [ for a radiative efficiency of 0.1 , which may be generous , in view of the jet luminosity inferred for the weaker source 3c31 by laing & bridle ( 2002 ) ] and , because of the low internal density , bulk motions with speeds around @xmath94 would still be sufficient to bend the plume : as e84 comment , low jet densities are possible , but require high speeds if the luminosity condition is to be satisfied . if , on the other hand , the pressure deficit at minimum energy is supplied by hot , entrained , internal thermal material , as is possible for fri lobes , then we know only that the density contrast is @xmath95 ( since the thermal protons then dominate the density of the flow ) and in that case external bulk motions at the sound speed can bend the plume significantly if @xmath96 , which for moderate heating of the internal protons requires plume speeds that are at most mildly supersonic : for example , if the plume speed is 2500 km s@xmath2 , then the internal protons must be heated to around 20 kev ( which would also give a density for internal protons that was close to the ` limit ' on internal density inferred by e84 ) . bending by subsonic motions requires a correspondingly higher temperature , increasing as @xmath97 , but this is not impossible , as we have little information on the temperature of the internal protons . in the case of a proton - dominated plume the luminosity condition , eq . [ lum ] , requires @xmath98 , and it can be shown that very high internal temperatures ( @xmath99 k ) , correspondingly low densities , and high plume flow speeds ( @xmath100 ) are required for the plume to be bent by sub - sonic bulk motions in the external medium . to summarize , the plumes can be bent by bulk motions in the external medium if they are light , but they will only also meet the luminosity condition of e84 if they are also fast . if we drop the assumption of _ in situ _ acceleration ( and instead assume passive transport of radiating electrons ) then the plumes need only be mildly supersonic , but must still be light . this conclusion revives the possibility that the plumes are in fact bent by the motion of the galaxy with respect to the host cluster . e84 suggest that the systematic motion of ngc 7720 could be of order 100300 km s@xmath2 , or @xmath101@xmath102 . their estimate does not necessarily rule out higher galaxy speeds , because they base it on the radial velocity of the galaxy with respect to the cluster mean ( which does not tell us about motions in the plane of the sky ) and on studies of the radial velocities of cds in general ( which may not be representative of the systems hosting wats ) . the possible detection of an x - ray wake behind 3c465 ( section [ wake ] ) may indicate a higher speed . nevertheless , it is interesting that the speed estimate of e84 is within the range estimated above for plume bending either if the plume s internal pressure is supplied entirely by relativistic particles and the plume is fast , or if a slower plume s internal pressure is supplied by very hot thermal material . is this picture consistent with the fact that the inner jets are not bent ? this question is more difficult to answer because we do not have a reliable estimate for the inner jet speed or density , and we can not assume pressure balance in the jet . if we assume @xmath103 ( from the beaming analysis ) and take the minimum equivalent internal density to be given by @xmath104 , using the minimum pressure plotted in fig . [ pprofile ] , then we can use euler s equation in the form @xmath105 where @xmath106 is the radius of curvature of the jet , @xmath107 is the galaxy speed , and @xmath108 is the width of the jet . ( the non - relativistic approximation is not badly wrong in this case . ) from the radio maps we can see that the radio jet bends , if it bends at all , by less than one jet width over its observed length @xmath109 . taking @xmath110 arcsec ( which includes the effects of projection with @xmath111 ) , @xmath112 arcsec , and an external density comparable to that at the transition between the hot and cool components in the x - ray ( 1500 protons m@xmath49 ) we find that speeds of 120 km s@xmath2 could produce a bend of one jet width . as this is very consistent with the possible values for systematic motion of the galaxy , and as our estimate of @xmath113 is a lower limit , we are satisfied that an explanation for the plume bending in terms of the systematic motion of the galaxy through the cluster is not inconsistent with the observations of essentially straight inner jets , provided that the jets are fast . we emphasise , however , that this model is only viable if the plume is light . if there is substantial entrained thermal material , then the analysis above suggests that sub - sonic motion of the galaxy through the cluster material is too slow to have a significant effect on the plumes . it is interesting in this context that there is a population of wat sources with bent inner jets ( exemplified by 0647 + 693 , hardcastle & sakelliou 2004 ) : if the jets in these sources _ are _ bent by ram pressure due to the host galaxy s motion , then the model we have discussed suggests either that the jets in these sources are weaker ( transport less energy ) or slower , or that the systematic motions of the host galaxies are larger . we will investigate this possibility in more detail elsewhere . [ fri - wat ] these results , combined with those of the previous subsection , mean that there are some interesting possible differences between wats and classical fri sources . on the one hand , we have shown ( section [ pcontent ] ) that the large - scale plumes of 3c465 are very similar to the extended structure of other fri sources in terms of the difference between the minimum internal pressure and the external pressure : in fris one of the best explanations for this observation is that the pressure deficit is supplied by heated , entrained material . on the other hand , as we have seen , bending of the 3c465 plumes by galactic motions puts limits on the amount of entrained material that is present : for example , if we adopt a plume speed of 2500 km s@xmath2 , then significant bending by bulk speeds of @xmath114 km s@xmath2 require thermal densities several orders of magnitude below the external density ( with correspondingly high temperatures if we want this material to supply the missing pressure ) , while satisfying the e84 luminosity condition requires even lower densities . these densities are considerably lower than the densities estimated by odea ( 1985 ) from the bending of narrow - angle tail ( nat ) sources . although it would be interesting to revisit the calculations of odea with modern estimates of the external density , and to compare them to estimates of the proton density required to provide pressure in the lobes , this may be an indication that type ii jets such as those in 3c465 are less efficient at entraining thermal material than the type i jets of nats . our principal results can be summarized as follows : 1 . the detection of x - ray emission from the inner part of the jet of 3c465 adds significantly to the evidence that fast ( at least mildly relativistic ) , efficient , frii - like ( ` type 2 ' ) jets can be sources of x - ray synchrotron radiation . 2 . the non - detection of x - ray emission from the jet termination point and the bases of the plumes rules out a model in which the plume bases resemble the bright bases of fri radio jets , which are often x - ray synchrotron sources : this is consistent with our favoured model for the wat jet - plume transition , in which the deceleration of the fast inner jets takes place rapidly when the jet interacts with the external medium at the edge of the plume , rather than slowly by entrainment as in fri jets . however , it is interesting that there is no detection of synchrotron x - rays from any of the plausible candidates for the jet termination ` hotspot ' in 3c465 , unlike the situation in the ( supposedly physically similar ) hotspots in low - power frii sources . more x - ray observations of wats with bright jet termination hotspots are required . we have made detailed measurements of the temperature and density structure of 3c465 s host cluster , a2634 , and in particular have resolved a central dense , cool thermal component with a relatively short cooling time , which is similar to the components seen in other wats and to the central components of fri sources in groups . there is no evidence for any discontinuity in the properties of the external medium , and so we conclude , following jetha et al . ( 2005 ) , that models of wat formation that require such discontinuities are ruled out . the most likely scenario based on our observations and those of jetha et al . is that the location of the jet - plume transition is related to the emergence of the jet from the small - scale cool component associated with the host galaxy , but the physical mechanism that sets the location of the plume base remains unclear . our observations show that the minimum pressures of the plumes fall below the pressures estimated for the external thermal material : this implies either some departure from the minimum - energy ( equipartition ) condition or an additional contribution to pressure from low - density , hot thermal material . the plumes of wats are thus similar to the large - scale components of more normal fri sources . 5 . revisiting the arguments related to wat plume bending , we find that the plumes can be bent by sub - sonic bulk motions of the icm , or by plausible motion of the host galaxy through the icm , provided that they are extremely light with respect to the external medium ( as would be the case if their internal pressures were dominated by relativistic particles and/or magnetic field ) . we thank dan evans for discussions of the x - ray core and helpful comments on the paper , nazirah jetha for discussions of wat properties , and an anonymous referee for helpful comments . mjh thanks the royal society for a research fellowship . the national radio astronomy observatory is a facility of the national science foundation operated under cooperative agreement by associated universities , inc . this work is based on observations obtained with _ xmm - newton _ , an esa science mission with instruments and contributions directly funded by esa member states and nasa . arnaud , m. , et al . , 2002 , a&a , 390 , 27 belsole , e. , worrall , d.m . , hardcastle , m.j . , birkinshaw , m. , lawrence , c.r . , 2004 , mnras , 352 , 924 bicknell , g.v . , 1994 , apj , 422 , 542 birkinshaw , m. , worrall , d.m . , 1993 , apj , 412 , 568 brzan , l. , rafferty , d.a . , mcnamara , b.r . , wise , m.w . , nulsen , p.e.j . , 2004 , apj , 607 , 800 burns , j.o . , 1981 , mnras , 195 , 523 burns , j.o . , rhee , g. , owen , f.n . , pinkney , j. , 1994 , apj , 423 , 94 comastri , a. , brunetti , g. , dallacasa , d. , bondi , m. , pedani , m. , setti , g. , 2003 , mnras , 340 , l52 croston , j.h . , hardcastle , m.j . , birkinshaw , m. , worrall , d.m . , 2003 , mnras , 346 , 1041 croston , j.h . , birkinshaw , m. , hardcastle , m.j . , worrall , d.m . , 2004 , mnras , 353 , 879 croston , j.h . , hardcastle , m.j . , harris , d.e . , belsole , e. , birkinshaw , m. , worrall , d.m . , 2005 , apj submitted edge , a.c . , stewart , g.c . , 1991 , mnras , 252 , 428 eilek , j.a . , owen , f.n . , 2002 , apj , 567 , 202 eilek , j.a . , burns , j.o . , odea , c.p . , owen , f.n . , 1984 , apj , 278 , 37 [ e84 ] fukazawa , y. , makishima , k. , tamura , t. , ezawa , h. , xu , h. , ikebe , y. , kikuchi , k. , ohashi , t. , 1998 , pasj , 50 , 187 gmez , p.l . , pinkney , j. , burns , j.o . , wang , q. , owen , f.n . , voges , w. , 1997a , apj , 474 , 580 gmez , p.l . , ledlow , m.j . , burns , j.o . , pinkney , j. , hill , j.m . , 1997b , aj , 114 , 1711 hardcastle , m.j . , 1998 , mnras , 298 , 569 hardcastle , m.j . , 1999 , a&a , 349 , 381 hardcastle , m.j . , birkinshaw , m. , worrall , d.m . , 2001 , mnras , 326 , 1499 hardcastle , m.j . , sakelliou , i. , 2004 , mnras , 349 , 560 hardcastle , m.j . , worrall , d.m . , 2000 , mnras , 319 , 562 hardcastle , m.j . , worrall , d.m . , birkinshaw , m. , laing , r.a . , bridle , a.h . , 2002 , mnras , 334 , 182 hardcastle , m.j . , worrall , d.m . , kraft , r.p . , forman , w.r . , jones , c. , murray , s.s . , 2003 , apj , 593 , 169 hardcastle , m.j . , worrall , d.m . , birkinshaw , m. , laing , r.a . , bridle , a.h . , 2005 , mnras in press higgins , s.w . , obrien , t.j . , dunlop , j.s . , 1999 , mnras , 309 , 273 hooda , j.s . , wiita , p.j . , 1996 , apj , 470 , 211 jaffe , w.j . , perola , g.c . , 1973 , a&a , 26 , 423 jetha , n.n . , sakelliou , i. , hardcastle , m.j . , ponman , t.j . , stevens , i.r . , 2005 , mnras , in press jones , c. , forman , w. , 1984 , apj , 276 , 38 jones , c. , mandel , e. , schwarz , j. , forman , w. , murray , s.s . , harnden , f.r . , 1979 , apj , 234 , l21 kraft , r.p . , hardcastle , m.j . , worrall , d.m . , murray , s.s . , 2005 , apj submitted laing , r.a . , 1984 , in bridle a.h . , eilek j.a . , eds , physics of energy transport in radio galaxies , nrao workshop no . 9 , nrao , green bank , west virginia , p. 90 laing , r.a . , bridle , a.h . , 2002 , mnras , 336 , 1161 leahy , j.p . , 1984 , mnras , 208 , 323 leahy , j.p . , 1993 , in rser h .- j . , meisenheimer k. , eds , jets in extragalactic radio sources , springer - verlag , heidelberg , p. 1 loken , c. , roettiger , k. , burns , j.o . , 1995 , apj , 445 , 80 odea , c.p . , 1985 , apj , 295 , 80 pinkney , j. , rhee , g. , burns , j.o . , hill , j.m . , oegerle , w. , batuski , d. , hintzen , p. , 1993 , apj , 416 , 36 read , a.m. , ponman , t.j . , 2003 , a&a , 409 , 395 riley , j.m . , branson , j.b.a . , 1973 , mnras , 164 , 271 sakelliou , i. , merrifield , m.r . , 1998 , mnras , 293 , 489 sakelliou , i. , merrifield , m.r . , 1999 , mnras , 305 , 417 sakelliou , i. , merrifield , m.r . , mchardy , i.m . , 1996 , mnras , 283 , 673 sakelliou , i. , acreman , d.m . , hardcastle , m.j . , merrifield , m.r . , ponman , t.j . , stevens , i.r . , 2005 , mnras submitted schindler , s. , prieto , m.a . , 1997 , a&a , 327 , 37 scodeggio , m. , solanes , j.m . , giovanelli , r. , haynes , m.p . , 1995 , apj , 444 , 41 tavecchio , f. , maraschi , l. , sambruna , r.m . , urry , c.m . , 2000 , apj , 544 , l23 worrall , d.m . , birkinshaw , m. , hardcastle , m.j . , 2001 , mnras , 326 , l7

we describe the different possibilities that a simple and apparently quite harmless classical scalar field theory provides to violate the energy conditions . we demonstrate that a non - minimally coupled scalar field with a positive curvature coupling can easily violate all the standard energy conditions , up to and including the averaged null energy condition ( anec ) . indeed this violation of the anec suggests the possible existence of traversable wormholes supported by non - minimally coupled scalars . to investigate this possibility we derive the classical solutions for gravity plus a general ( arbitrary ) massless non - minimally coupled scalar field , restricting attention to the static and spherically symmetric configurations . among these classical solutions we find an entire branch of traversable wormholes for every . ( this includes and generalizes the case of conformal coupling we considered in phys . lett . * b466 * ( 1999 ) 127134 . ) for these traversable wormholes to exist we demonstrate that the scalar field must reach trans - planckian values somewhere in the geometry . we discuss how this can be accommodated within the current state of the art regarding scalar fields in modern theoretical physics . we emphasize that these scalar field theories , and the traversable wormhole solutions we derive , are compatible with all known experimental constraints from both particle physics and gravity . pacs : 04.60.ds , 04.62.+v , 98.80 hw + keywords : + traversable wormholes , energy conditions , classical solutions , non - minimal coupling . ' '' '' ' '' '' s l

gr-qc0003025

it is often ( mistakenly ) believed that every kind of matter , on scales in which we do not need to consider its quantum features , has an energy density that is everywhere positive . ( in fact , we could think on this property as defining what we understand by classical matter ) . this is more precisely stated by saying that every type of classical matter satisfies the energy conditions of general relativity @xcite . there are several ( pointwise ) energy conditions requiring that various linear combinations of the components of the energy - momentum tensor of matter have positive values . ( or at the very least , non - negative values . ) among them , we will be particularly interested in the null energy condition ( nec ) because it is the weakest one : if the nec is violated all pointwise energy conditions would be violated @xcite . now , if we assume that the energy conditions are satisfied by every kind of classical matter , general relativity leads to many powerful classical theorems . the singularity theorems @xcite , the positive mass theorem @xcite , the superluminal censorship theorem @xcite , the topological censorship theorem @xcite , certain types of no - hair theorem @xcite , and various constraints on black hole surface gravity @xcite , all make use of some type of energy condition . as an illustration of the importance of these results , the conclusion regarding the inevitability of the appearance of singularities @xcite , has been ( for the last thirty years ) one of the central pillars from which many investigations in general relativity start . ( additional powerful technical assumptions are also needed for the singularity theorems to apply ; see @xcite for a critical review ) . in this regard , the alleged impossibility of the existence of traversable wormholes connecting different spatial regions of the universe @xcite , a topic in which we will be particularly interested in this paper , could be seen as a complementary way of phrasing that conclusion ( of course , one should not carry this parallelism too far ) . assuming the positivity of the energy density implies that spacetime geometries containing traversable wormholes are ruled out of the classical realm . specifically , the topological censorship theorem @xcite states that if the averaged null energy ( anec ; the nec averaged over a complete null geodesic ) is satisfied , then there can not be any topological obstruction ( , a wormhole ) inside any asymptotically flat spacetime . in agreement with this result , a purely local analysis by david hochberg and one of the present authors @xcite shows that the violation of nec on or near the throat of a traversable wormhole is a generic property of these objects . for this reason most of the investigations regarding traversable wormholes tend to view these objects as semiclassical in nature , using the expectation value of the quantum operator associated with the energy - momentum tensor as the source of gravity @xcite . however , it is easy to demonstrate that an extremely simple and apparently quite innocuous classical field theory , a scalar field non - minimally coupled to gravity ( that is , with a non - vanishing coupling to the scalar curvature ) , can violate the nec and even the anec @xcite . as we will see , some of the other energy conditions can be violated even by minimally coupled scalar fields . there exist other classical systems that exhibit nec violations , such as brans dicke theory @xcite , higher derivative gravity @xcite or gauss bonnet theory @xcite , but they are all based on modifications of general relativity at high energies . it is the simplicity of the scalar field theory that particularly attracted our attention . in a previous paper @xcite we analyzed a massless scalar field conformally coupled to gravity ( that is , we used the special value @xmath2 ) and found that among the classical solutions for the system there exists an entire branch of traversable wormholes . the case of conformal coupling has many interesting features and , in fact , it seems the most natural behaviour for a scalar field at low energies @xcite . however , in this paper we want to point out that assuming conformal coupling is not the critical issue , and that it is only the fact of non - minimality ( more precisely , positive curvature coupling @xmath0 ) that is important for traversable wormhole solutions to exist . we have found general expressions for the classical solutions of gravity plus a massless non - minimally coupled scalar field . in particular , we have found a sub - class of traversable wormhole solutions for the entire range @xmath3 . in our solutions , apart from the geometry , there is also a scalar field that modifies the effective newton constant . we will not address here the possible effects of this scalar field on the journey of any hypothetical traveller trying to cross the wormhole throat . in the next section we will review the possibilities that scalar fields offer to violate the different energy conditions . ( that scalar fields might potentially cause problems in this regard was first noted at least 25 years ago @xcite , and is an issue that has periodically come in and out of focus since then @xcite . ) then we will obtain and describe the classical solutions for gravity plus a non - minimally coupled massless scalar field , restricting to static and spherically symmetric configurations . among these solutions there are an assortment of naked singularities but we also find an entire branch of traversable wormhole solutions for curvature coupling @xmath0 . we will leave it to section [ s : wormhole ] to analyze these traversable wormhole geometries in detail . finally , in section [ s : discussion ] we will discuss the key features of these solutions and the plausibility with which they might actually be produced in nature , based on the role of scalar fields in modern theoretical physics and the totality of the experimental constraints arising from both particle physics and gravity physics . when a classical scalar field acts as a source of gravity , many of the energy conditions can be violated depending on the form of the scalar potential and the value of the curvature coupling . the einstein equations @xmath4 relate the geometry - dependent einstein tensor @xmath5 to the energy - momentum tensor for the matter field @xmath6 . the symbol @xmath7 represents essentially the inverse of newton s constant , @xmath8 . these equations can be obtained by varying the einstein hilbert action which , for a generically coupled scalar field , reads @xmath9 then , the scalar field energy - momentum tensor has the form @xcite @xmath10_{\mu\nu } = & & \hspace*{-6 mm } \nabla_{\mu } \phi_\xi \ ; \nabla_{\nu } \phi_\xi - { 1 \over 2 } g_{\mu\nu } ( \nabla \phi_\xi)^2 - g_{\mu\nu } \ ; v(\phi_\xi ) \nonumber \\ & & \hspace*{-6 mm } + \xi\left [ g_{\mu\nu } \ ; \phi_\xi^2 - 2 \;\nabla_{\mu } ( \phi_\xi \ ; \nabla_{\nu } \phi_\xi ) + 2 \ ; g_{\mu\nu } \ ; \nabla^{\lambda } ( \phi_\xi \ ; \nabla_{\lambda } \phi_\xi ) \right ] . \label{e : emt}\end{aligned}\ ] ] this energy - momentum tensor has a term that depends algebraically on the einstein tensor . by grouping all the dependence on @xmath5 on the left hand side of einstein equations we can rewrite them , alternatively , by using an effective energy - momentum tensor @xmath11_{\mu\nu } = & & \hspace*{-6 mm } { \kappa \over \kappa- \xi \phi_\xi^2 } \bigg [ \nabla_{\mu } \phi_\xi \ ; \nabla_{\nu } \phi_\xi - { 1 \over 2 } g_{\mu\nu } ( \nabla \phi_\xi)^2 - g_{\mu\nu } \ ; v(\phi_\xi ) \nonumber \\ & & \hspace*{-6 mm } -\xi\left [ 2 \;\nabla_{\mu } ( \phi_\xi \ ; \nabla_{\nu } \phi_\xi ) - 2 \ ; g_{\mu\nu } \ ; \nabla^{\lambda } ( \phi_\xi \ ; \nabla_{\lambda } \phi_\xi ) \right ] \bigg ] . \label{e : emt2}\end{aligned}\ ] ] this is the relevant expression for the analysis of the different energy conditions : since the einstein equations now read @xmath12_{\mu\nu}$ ] , a constraint on this effective stress - energy tensor is translated directly into a constraint on the spacetime curvature , and it is ultimately constraints on the spacetime curvature that lead to singularity theorems and the like . now let @xmath13 be a properly normalized timelike vector , @xmath14 , and , for convenience , let it be locally extended to a geodesic vector field , so that @xmath15 . if @xmath16 denotes a timelike geodesic with tangent vector @xmath17 we have @xmath18 . we can now express the strong energy condition ( sec ) as @xmath19_{\mu\nu } - { 1 \over 2}g_{\mu\nu } \ ; [ t^\eff(\phi_\xi)]\,\right ) v^{\mu } \ ; v^{\nu } \nonumber \\ = & & \hspace*{-6 mm } { 1 \over \kappa- \xi \phi_\xi^2 } \left [ ( \phi_\xi ' ) ^2-v(\phi_\xi)-\xi[(\phi_\xi^2)''-\nabla^\mu ( \phi_\xi\nabla_\mu\phi_\xi ) ] \right]\geq 0 , \label{e : sec}\end{aligned}\ ] ] where @xmath20 is the ricci tensor of the geometry . it is easy to see that even in the minimally coupled case the sec can be violated for positive values of the scalar potential such us a mass term or a positive cosmological constant . every cosmological inflationary process violates the sec @xcite . we could say that violation of the sec is a generic property of scalar fields . in fact , it is so easy to violate sec in many situations that it has almost become to be abandoned @xcite as a reasonable restriction on the properties of matter . with the same definitions , the weak energy condition ( wec ) reads @xmath21_{\mu\nu } \ ; v^{\mu } \ ; v^{\nu } \\ = & & \hspace*{-6 mm } { 1 \over \kappa- \xi \phi_\xi^2 } \left [ ( 1 - 2\xi)(\phi_\xi')^2 + { 1 \over 2}(\nabla \phi_\xi)^2 + v(\phi_\xi ) -\xi[(\phi_\xi^2 ) '' + 2\nabla^\mu(\phi_\xi \nabla_\mu\phi_\xi ) ] \right ] \geq 0 . \nonumber \label{e : wec}\end{aligned}\ ] ] in the minimally coupled case @xmath22 it is generically satisfied . only a large negative potential , for example , a negative cosmological constant , could provide a violation of wec . although recent observations suggest a probable positive value for the effective cosmological constant , a possible negative value can not be ruled out on theoretical grounds . in the non - minimal case there are various terms that can be negative depending on the situation , so the wec can be violated in various ways . finally , let us now analyze the nec . it is the weakest pointwise energy condition , that is , when it is violated the wec and sec are violated too . let @xmath23 be a null vector tangent to the null geodesic @xmath24 , with @xmath25 some affine parameter . in an analogous way as with the previous energy conditions , we arrive at the following expression for the nec @xmath26_{\mu\nu } \ ; k^{\mu } \ ; k^{\nu } = { 1 \over \kappa- \xi\phi_\xi^2 } \left[\phi_\xi'^2 - \xi(\phi_\xi^2)''\right ] \geq 0 . \label{e : nec}\ ] ] this condition is clearly satisfied by minimally coupled scalars . however , for @xmath27 it can be violated in a number of ways : for @xmath28 any local minimum of @xmath29 violates the nec while for @xmath30 and @xmath31 small , [ meaning @xmath32 , any local maximum of @xmath29 violates the nec . finally for @xmath30 and @xmath31 large , [ meaning @xmath33 , roughly corresponding to super planckian values for the scalar field ] , any local minimum of @xmath29 violates the nec . at this point it is worth noticing that , because our analysis is completely classical , it is _ a priori _ conceivable that averaged versions of the energy conditions ( averaged over a geodesic ) could in principle be as easily violated as their pointwise counterparts . in particular , anec violations , critical for traversable wormhole configurations to be able to exist , could in principle be as easy to find as nec violations , and it is to exploring this possibility that we now turn . suppose we take a segment of a null geodesic and consider the anec integral @xmath34_{\mu\nu } \ ; k^\mu \ ; k^\nu \ ; \d\lambda.\ ] ] then @xmath35 integrate by parts @xmath36 now assemble the pieces : @xmath37\over(\kappa-\xi\phi_\xi^2)^2 } \left({\d\phi_\xi \over \d\lambda}\right)^2 d\lambda - \left.\left\ { { 2\xi\kappa\over\kappa-\xi\phi_\xi^2 } \left ( \phi_\xi { \d\phi_\xi\over \d\lambda } \right ) \right\ } \right|_{\lambda_1}^{\lambda_2}.\ ] ] discarding the boundary terms is an issue fraught with subtlety : we start by considering a complete null geodesic and assuming sufficiently smooth asymptotic behaviour . then the boundary terms from asymptotic infinity can be neglected , and the only potential problems with the boundary terms come from the places @xmath38 where @xmath39 . we obtain @xmath40\over(\kappa-\xi\phi_\xi^2)^2 } \left({\d\phi_\xi \over \d\lambda}\right)^2 d\lambda + \sum_i \left.\left\ { { 2\xi\kappa\over\kappa-\xi\phi_\xi^2 } \left ( \phi_\xi { \d\phi_\xi\over \d\lambda } \right ) \right\ } \right|_{\lambda_i^+}^{\lambda_i^-}.\ ] ] 1 . if @xmath41 then there are no places on the geodesic where @xmath42 ^ 2 $ ] , so the boundary terms represent an empty set . but if @xmath41 the _ integrand _ in the above formula is itself guaranteed positive so anec is satisfied . 2 . if @xmath0 , but we have @xmath43 , then again there are no places on the geodesic where @xmath44 ^ 2 $ ] . furthermore the _ integrand _ appearing above is again positive and anec is satisfied . 3 . finally , if @xmath0 , and we have at least some places where @xmath45 , then there are by definition places on the geodesic where @xmath46 ^ 2 $ ] . the boundary terms can no longer be neglected and potentially can contribute typically making negative and infinite contributions to the anec integral . furthermore the _ integrand _ appearing above is no longer guaranteed to be positive . ( if @xmath47 then it is possible to have @xmath48 $ ] and so make the integrand negative . ) in short : there is definitely the _ possibility _ of anec violations under these conditions , and in the exact solutions we investigate below we shall see that for some of these exact solutions the anec is certainly violated . thus we have seen that a rather simple and apparently quite harmless scalar field theory can in many cases violate all the energy conditions . violating all the pointwise energy conditions is particularly simple , and violating the averaged energy conditions , though more difficult , is still generically possible . in the next section we shall exhibit some specific examples of this phenomenon in the form of exact solutions to the coupled einstein scalar - field equations . in the final section we will discuss the extent to which these exact solutions are realistic : we shall discuss the generic role played by scalar fields in modern theoretical physics , and the experimental / observational limits on their existence and behaviour in order to address the physical plausibility of these energy condition violations . as we have just seen , a non - minimally coupled scalar field can violate , in some circumstances , the nec and even the anec . this opens up the possibility of finding some traversable wormholes among the many geometries that can be supported by a classical non - minimally coupled scalar field . that this is in fact the case for a massless conformally coupled scalar field was shown in @xcite . in the present paper , we will see that even for non - conformal coupling ( curvature coupling different from @xmath49 ) , but positive , we can also find traversable wormholes . in this section we will obtain the classical solutions for gravity plus a generic non - minimally coupled scalar field . for simplicity , we will restrict to the spherically symmetric and static configuration and will take the scalar potential @xmath50 equal zero . ( to add a particle mass to the scalar field complicates the equations sufficiently to preclude the possibility of analytic results ) . the first thing that we realize is that for any spatially constant value of the scalar field , @xmath51 , the einstein equations reduce to @xmath52 . by looking also at the scalar field equation @xmath53 , where @xmath54 is the ricci scalar curvature , we can straightforwardly find a variety of trivial solutions for this system . for ( i ) @xmath0 and @xmath55 , or ( ii ) @xmath41 and any value value of @xmath56 , we find that the ordinary vacuum einstein solutions also solve the coupled einstein - scalar equations . as we are here restricted to spherically symmetric and static configurations without cosmological constant , the solutions that show up are the schwarzschild and anti - schwarzschild geometries ( @xmath57 represents a perfectly good solution to the einstein field equations , normally the negative mass schwarzschild geometry is excluded by hand , here it s best to keep it for the time being as an aid in classifying the total solution space . ) more surprising is that for @xmath0 and @xmath58 every ricci - scalar - flat geometry is a solution . ( that is , any geometry satisfying the condition @xmath59 is a solution of the coupled einstein - scalar equations . ) the condition @xmath59 is characteristic of geometries supported by conformally invariant matter . in particular , the geometries that appear in the solutions for the @xmath2 case all satisfy @xmath59 . these geometries were found in froyland @xcite and also in @xcite , and will be re - obtained later on this paper as particular cases . here we want to point out that these geometries are not only associated with the conformal coupling but they appear quite generally , for arbitrary @xmath0 , provided only @xmath60 . to obtain additional ( non - trivial ) solutions for a scalar field non - minimally coupled to gravity we will use a `` solution generating technique '' that relies on knowledge of the solutions for the minimally coupled case @xcite . the classical solutions for a massless scalar field minimally coupled to gravity are very well known . they have been discovered and re - discovered several times in different coordinate systems ( see the articles by fisher @xcite , janis , newman , and winicour @xcite , wyman @xcite , and m. cavaglia and v. de alfaro @xcite ) . they can be expressed @xcite as @xmath61 the same geometry , which has an obvious symmetry under @xmath62 , can exist with a field configuration @xmath63 or the reversed sign configuration @xmath64 . less obvious is that by making a coordinate transformation @xmath65 , one uncovers an additional symmetry under @xmath66 , ( with @xmath67 ) . the key to this symmetry is to realize that @xmath68 in view of these symmetries one can without loss of generality take @xmath69 and @xmath70 $ ] remembering the overall two possible signs for the scalar field . similar symmetries will be encountered for non - minimally coupled scalars . the lagrangian for which these `` minimal '' solutions are extrema is @xmath71 this lagrangian and the lagrangian for a non - minimally coupled massless scalar field , @xmath72 can be related by a conformal transformation of the metric @xmath73 and a redefinition of the scalar field @xmath74 @xcite . rescaling the field @xmath75 we can write the specific transformation as @xmath76 notice that for @xmath77 the absolute value of the non - minimally coupled scalar field can not surpass @xmath78 if we want expression ( [ e : equations ] ) to make sense . for every solution of equation ( [ e : equations ] ) we have a two - parameter family , @xmath79 , of solutions for the non - minimally coupled system . for @xmath80 , the expressions in ( [ e : equations ] ) become the identity transformation , as they must . for @xmath2 we easily find the following solutions : @xmath81 with @xmath82 and @xmath83 two arbitrary real constants . the second set of solutions is unphysical because it gives a negative sign for the metric signature ( opposite to the one we are using ) . however , one can easily demonstrate @xcite that the einstein tensor @xmath5 and the energy - momentum tensor ( [ e : emt ] ) for a massless scalar field are both invariant if we change @xmath84 to @xmath85 leaving the field unchanged . therefore , from the unphysical solutions ( [ e : unphysicalsol ] ) we obtain physical solutions of the form @xmath86 we can also find additional solutions by considering the limiting cases when the constants @xmath82 or @xmath83 tend to ( @xmath87 ) infinity . in this way we find the solutions : @xmath88 all these metrics @xmath89 , described in ( [ e : physicalsol1 ] ) , ( [ e : physicalsol2 ] ) , ( [ e : physicalsol1l ] ) , and ( [ e : physicalsol2l ] ) have a zero scalar curvature , @xmath59 , owing to the conformal coupling features @xcite . therefore , any of these geometries , supplemented with the specific value @xmath90 for the non - minimally coupled scalar field ( with @xmath0 ) are examples of the `` trivial '' solutions for the non - minimally coupled system of which we have spoken at the beginning of this section , ) and ( [ e : physicalsol2l ] ) , are already in this later class . ] and these `` trivial '' solutions no longer seem to be all that trivial . these solutions can not be obtained by means of the conformal transformation procedure described above because the conformal factor has a singular behaviour for those scalar field values . the spacetime geometries of all these solutions were analyzed in a previous paper @xcite although with a slightly different parameterization . here , we will describe them as limiting cases of the solutions that we will find for the interval @xmath77 . however we want to mention that among these conformal solutions we found traversable wormholes and that in the present parameterization they correspond to ( [ e : physicalsol1 ] ) with @xmath91 , ( [ e : physicalsol2 ] ) with @xmath92 , and ( [ e : physicalsol2l ] ) ; all using a value for @xmath93 in ( [ e : msolution ] ) equal to @xmath94 . let us now solve the general equation ( [ e : equations ] ) for an arbitrary @xmath1 . we first rewrite this equation as a sum of two terms @xmath95 . \label{e : requation}\ ] ] for convenience , we will express the solutions of this equation in terms of two functions @xmath96 and @xmath97 as @xmath98 . \label{e : fh}\ ] ] the function @xmath99 will be related with the first term in ( [ e : requation ] ) and the function @xmath100 with the second term . the first term in ( [ e : requation ] ) can easily be integrated by changing to a new variable @xmath102 using this new variable we only have to solve an integral of the form @xmath103 that we express in terms of logarithms . in this way ( once we invert the change of variables ) , we arrive at a closed form for the function f in ( [ e : fh ] ) , for arbitrary @xmath1 . indeed , the above integral gives rise to two possible functions @xmath99 , @xmath104 and @xmath105 , of the form @xmath106 for @xmath0 , and only one , the positive sign @xmath104 , for @xmath107 . this restriction on signs is due to the fact that for @xmath0 the absolute value of the variable @xmath108 can be greater or lower than unity and these two domains have to be analyzed separately . in terms of the scalar field , these two domains are separated by the critical value @xmath109 . at best , the function @xmath104 is only defined for @xmath110 whilst @xmath105 is at best defined for the on region @xmath111 . indeed , in the case @xmath77 , the function @xmath105 is only defined up to @xmath112 . an observation is in order at this point . in the second domain of values for the scalar field the conformal factor @xmath113 is negative and so the geometry obtained is `` unphysical '' in the sense that the metric has reversed signature . as explained before for the conformal case , from this unphysical solution we can obtain a physical solution by changing the sign of the conformal factor to @xmath114 @xcite . another observation concerning the functions @xmath115 is that we have already embedded into them the corresponding integration constants for each solution . the @xmath116 in ( [ e : f ] ) are these integration constants . owing to the logarithmic form in which we have cast the solution these constants are both positive . in summary @xmath41 : + @xmath117 is real and well - defined for all values of @xmath118 ; + @xmath119 is undefined ( complex ) , and un - needed . @xmath80 : + @xmath120 ; + @xmath119 is undefined ( complex ) , and un - needed . @xmath121 : + @xmath117 is real and well - defined for @xmath122 ; + and is undefined outside this range . + @xmath119 is real and well - defined for @xmath123 ; + and is undefined outside this range . @xmath124 : + for conformal coupling there is tremendous simplification @xmath125 @xmath117 is well defined for @xmath126 , whereas @xmath119 is well - defined for @xmath127 . @xmath128 : + @xmath117 is real and well - defined for @xmath129 ; + and is undefined outside this range . + @xmath119 is real and well - defined for @xmath130 ; + and is undefined outside this range . the second term in ( [ e : requation ] ) can be integrated directly yielding different rather complex algebraic expressions for the function @xmath97 depending on the value of @xmath1 1 . @xmath41 : @xmath132 2 . @xmath80 : @xmath133 3 . @xmath77 : @xmath134 4 . @xmath2 : @xmath135 5 . @xmath136 : @xmath137 here , we have not introduced any arbitrary integration constants because , as we mentioned before , we have already included these constants in the @xmath115 functions . at this point we already have ( implicitly ) all the different classical solutions for gravity plus a massless non - minimal scalar field . on one hand , if we substitute the different @xmath96 and @xmath97 in the left hand side of ( [ e : fh ] ) , and the minimally coupled scalar field ( [ e : msolution ] ) in the right hand side , we have a implicit expression for the non - minimally coupled scalar field as a function of the radial coordinate @xmath138^{\pm 1}. \label{e : fieldr}\ ] ] these relations can not be analytically inverted in general . also , depending on whether the function @xmath99 is @xmath104 or @xmath105 we have geometries @xmath139 with different conformal factors : @xmath140 our next step is to analyze the different solutions found . let us begin with some general comments . for convenience , henceforth we will work in isotropic coordinates , @xmath141 in these coordinates the previous implicit expression for the scalar field ( [ e : fieldr ] ) becomes @xmath142^ { { \sin\chi \over \sqrt{3 } } } = [ f(\phi_\xi)h(\phi_\xi)]^{\pm 1}. \label{e : implicitrbar}\ ] ] as in the minimally coupled case , the non - minimal solutions possess a symmetry under @xmath143 going to @xmath144 . also , they exhibit a symmetry under the change of @xmath93 to @xmath145 and a simultaneous flip in the sign of the exponent on the right hand side of ( [ e : implicitrbar ] ) . therefore , we will without loss of generality restrict the analysis to @xmath146 , @xmath147 $ ] and the positive exponent , remembering that for every solution there is a second solution that is geometrically identical but with a reversed sign for the scalar field . after a little algebra we can express the schwarzschild radial coordinate @xmath148 for the different geometries as a function of @xmath149 and , therefore , implicitly as a function of the isotropic radial coordinate @xmath150 , @xmath151 \sqrt{\pm ( 1 - 6\xi\phi_\xi^2)}}. \label{e : src}\ ] ] in the following discussion it is useful to know that , in the case @xmath0 , if we make a perturbative expansion of the functions @xmath152 around @xmath153 ( that is , we take @xmath154 with @xmath155 and small positive quantity ) then the functions @xmath156 as @xmath157 . in the same way , the conformal factor @xmath158 as @xmath159 . in this case , the function @xmath160 is everywhere positive . ( see figure 1 . ) for a certain ( finite ) value of the scalar field ( depending on @xmath161 and @xmath1 ) , @xmath162 . we can easily verify that both @xmath150 and @xmath163 go to infinity at this stage , so they are describing an asymptotic region . indeed , it is an asymptotically flat region because since the scalar field goes to a finite constant in the asymptotic region with @xmath164 the conformal factor tends to a constant , and therefore the geometry behaves as in the minimal solution metric ( [ e : msolution ] ) . decreasing the value of the scalar field we drive ourselves towards the interior of the geometry . as @xmath165 , @xmath166 , and @xmath167 . then , analyzing the asymptotic behavior of @xmath149 in equation ( [ e : src ] ) we can conclude that the schwarzschild radial coordinate shrinks to zero for every @xmath168 . thus , for a non - minimally coupled scalar field with a curvature coupling @xmath41 we find naked singularities , in the same manner as for a minimally coupled scalar field @xcite . for the special case @xmath169 we find solutions with a constant value for the scalar field and a schwarzschild geometry ( of course , in this case the spacetime geometry does not shrink to zero for @xmath170 ) , while for @xmath171 we encounter an anti - schwarzschild geometry . the function @xmath101 in this case becomes a constant . the function @xmath131 is @xmath172 , so in equation ( [ e : fh ] ) we can read that @xmath173 . clearly , we recover the standard minimally coupled solutions with its naked singularities . here we have to analyze separately both signs in equation ( [ e : src ] ) . let us begin with the positive , @xmath104 . as before , for a certain value of the scalar field the function @xmath174 , making both @xmath175 and @xmath176 , thereby describing an asymptotically flat region . ( see figure 2 . ) then , decreasing the value of the field we leave the asymptotic region going towards the interior of the geometry . at the value @xmath177 the @xmath150 coordinate reaches the value @xmath178 with the corresponding zero value for @xmath160 . a perturbative analysis around @xmath177 tells us that for @xmath179 the schwarzschild radial coordinate blows up . for the rest of values the geometry shrinks to a naked singularity except for @xmath169 and @xmath180 in which the schwarzschild radial coordinate goes to a constant value . the solution with @xmath169 is once more the schwarzschild geometry , whilst among the naked singularities we have the @xmath181 case representing the anti - schwarzschild geometry . the solutions with @xmath180 are more bizarre . at @xmath170 the geometry neither shrinks to zero nor blows up to infinity . we can see the reason for this behaviour easily by looking at equation ( [ e : src ] ) . for @xmath180 the exponent of the function @xmath160 becomes unity and so it goes to zero at the same rate as the factor of @xmath182 , with these two terms counteracting each other . moreover , it can be seen that @xmath183 for @xmath184 , that is , we do not find any horizon by going inward from the asymptotic region . in fact , we can extend these geometries to values @xmath185 but we will leave for the next section to describe the traversable wormhole nature of these solutions . the solutions with @xmath179 also deserve some additional attention . they are wormhole - like shaped , that is , they have two asymptotic regions ( at @xmath186 and @xmath170 ) joined by a throat . however , analyzing the @xmath187 component of the ricci tensor ( @xmath188 in an orthonormal coordinate basis ) one can easily realize that it diverges when the scalar field reaches the value @xmath177 , that is , at @xmath170 . therefore , the region @xmath167 is not a proper `` asymptotic '' region . ( see the discussion on this point in @xcite ) . although in these geometries there exist diverging - lens effects ( they have a throat ) , they are not genuine traversable wormholes . the solutions with @xmath105 in equation ( [ e : src ] ) merit a discussion similar to that made for the plus sign . the real function @xmath189 is only defined for absolute values of the scalar field greater than @xmath190 , and of course less than @xmath191 . ( see figure 3 . ) when the scalar field reaches a value slightly lower than @xmath177 the radial coordinate @xmath150 approaches @xmath178 , with @xmath189 going to zero . at this coordinate point we can perform the same analysis as before , concerning the different behaviour as a function of @xmath93 of the schwarzschild radial coordinate @xmath192 . obviously , we find the same results . the case in which we are most interested is that of @xmath180 . in the next section we will see how by extending the geometry beyond @xmath178 we finally get a perfectly well - defined traversable wormhole . however , for this to happen it is necessary that by decreasing the value of the scalar field from the critical value @xmath177 , one must reach an asymptotic region before arriving to the lowest possible value of @xmath193 . that can only be guaranteed if the condition @xmath194 is fulfilled . this condition is obtained by requiring that @xmath189 have a value greater than one for @xmath195 . in the example of figure 3 this is not satisfied . this constraint implies super - planckian values of the scalar field in the wormhole throat , and is a cause for some mild concern we shall return to this point shortly . the coupling constant @xmath2 corresponds to a conformal coupling prescription . we have already found all the classical solutions for this system in @xcite , highlighting the many interesting characteristics possessed by the conformal coupling . here , we present these solutions as a particular case of the curvature coupling . for this particular case the function @xmath100 is unity and the functions @xmath115 become @xmath196 } ) = \phi_\pm \sqrt{\pm { 1+\phi_{[\xi=1/6 ] } \over 1-\phi_{[\xi=1/6 ] } } } .\ ] ] the expression ( [ e : fh ] ) can be inverted . ] yielding the two solutions ( [ e : physicalsol1 ] ) and ( [ e : physicalsol2 ] ) , provided we identify @xmath197 and @xmath198 . the analysis done for the previous @xmath77 case extends directly to @xmath2 . the special geometry corresponding to @xmath180 will be described more fully in the next section in combination with the equivalents for the range @xmath77 . once more it is necessary to take into account the functions @xmath104 and @xmath105 separately . the form of @xmath160 tells us that the scalar field reaches some finite value in an asymptotic region , at the point with @xmath199 . ( see figure 4 . ) then , by going inward from this asymptotically flat region we approach the @xmath200 section , for a value of the scalar field equal to @xmath177 . perturbatively analyzing the form of @xmath160 around @xmath201 we can see that @xmath163 has a different behaviour depending on the value of @xmath93 . one again obtains the same qualitative results as for the previous @xmath77 case . for @xmath202 the geometry blows up in a singular way , for @xmath203 the geometry shrinks to a naked singularity , and in the special @xmath180 case the geometry can be extended to values @xmath204 . once more , this particular case will be described in the next section . if we now consider the function @xmath189 , we can see that it is only defined for @xmath205 . here , for a negative value of the field lower than @xmath177 we have an asymptotically flat region , @xmath176 ( @xmath206 ) , but now this happens independently of the value of @xmath207 . ( see figure 5 . ) increasing the value of the scalar field up to @xmath177 we arrive at the inner @xmath208 region . at this coordinate point we find once more what we now see are the three possible standard behaviours for the geometry depending on @xmath93 . in this section we shall describe in more detail the solutions with @xmath180 and @xmath0 . in all these cases the expression under the square root symbol of the functions @xmath104 or @xmath105 changes its sign when @xmath201 ( @xmath208 ) . it seems that we can not define the scalar field beyond this point because the apparently complex values that @xmath99 would generate . however , we have to realize that for @xmath180 the expression ( [ e : implicitrbar ] ) can be written as @xmath142= [ f(\phi_\xi)h(\phi_\xi)]^2 . \label{e : implicitrbarp3}\ ] ] which we shall see allows us to avoid the problem and permits the extension . by using isotropic coordinates and particularizing to the value @xmath180 one can realize that there is a convenient way in which we can write the metric : @xmath209 \right ] . \label{e : wh}\ ] ] let us see what happens in the @xmath77 case when we choose the function @xmath104 . as we explained in the previous section , the implicit relation ( [ e : implicitrbarp3 ] ) , and the form of the schwarzschild radial coordinate ( [ e : src ] ) , tell us that we have a geometry that is perfectly regular from @xmath176 to @xmath170 , at which point the scalar field acquires the value @xmath177 . the function @xmath210 has a zero for @xmath201 which counteracts the zero in the denominator of the prefactor in equation ( [ e : wh ] ) . therefore , despite naive appearances , this prefactor acquires a finite positive value on @xmath201 . beyond that point , that is for @xmath211 and @xmath212 , we can see that this prefactor continues to be finite and positive [ @xmath210 changes its sign in the same way as @xmath213 even at the lowest possible value for the scalar field @xmath193 . now , if the radial coordinate @xmath150 reaches the value zero before the scalar field reaches its lowest possible value @xmath193 , that is , if the condition @xmath214 is fulfilled , expression ( [ e : wh ] ) tells us that for @xmath215 there is another asymptotically flat region . this spacetime is now a perfectly well - defined traversable wormhole geometry . if the condition ( [ e : condition+ ] ) is _ not _ satisfied , then the scalar field reaches the critical value @xmath193 before the function @xmath210 reaches the value @xmath216 for which the other asymptotic region shows up . in these solutions , we can extend the geometry and the scalar field regularly up to a spherical section at which the scalar field approaches the value @xmath193 . at this value we can see , by differentiating ( [ e : implicitrbar ] ) with respect @xmath150 , that the derivative of the scalar field becomes infinite . this , and the form of the scalar curvature for these systems , @xmath217 tells us that there is a curvature singularity in these solutions . ( this will be a naked singularity hiding behind a wormhole throat , but the lack of a second asymptotic region implies these are not true traversable wormhole solutions . ) we can now realize that by choosing the @xmath105 branch and extending upwards from scalar field values greater than @xmath193 towards @xmath218 , we obtain the same solutions as for @xmath104 , but with interchanged asymptotic regions . the prefactor in equation ( [ e : wh ] ) is @xmath219 in this case . the condition ( [ e : condition- ] ) now plays the same role here as that of ( [ e : condition+ ] ) previously . the traversable wormhole solutions that we found in @xcite for the conformal case can be rediscovered here as the limiting case of those with @xmath77 . the conditions ( [ e : condition- ] ) and ( [ e : condition+ ] ) for the existence of wormholes , become @xmath220 and @xmath221 which yield easily the conditions @xmath222 and @xmath223 for the conformal wormholes . in the case @xmath136 , and using the function @xmath104 , we had solutions with an asymptotic region in which the scalar field acquired a value greater than @xmath177 . from this asymptotic region one can move in the direction in which the scalar field decreases . in this way , first one crosses the non - singular section with @xmath201 . later on , one crosses a spherical section with a minimum diameter ( a throat ) , and from that point on the size of the spherical sections begin to grow to reach another asymptotic region associated with some asymptotic value for the scalar field which is of course lower than @xmath177 . here , one does not have to impose any restriction on @xmath224 in order to obtain a traversable wormhole configuration . by choosing the function @xmath105 we find the same solutions , but coordinatized in a reversed way : the asymptotic region with a greater scalar field value is here that with @xmath215 . in view of the plethora of traversable wormhole solutions we have found , an important observation is in order . in all these solutions the scalar field has to reach absolute values above @xmath225 , where @xmath226 is the planck mass . that is , either the scalar field acquires trans - planckian values or the curvature coupling constant @xmath1 must become disturbingly large . moreover , imagining that we include in the system some additional matter field with an action @xmath227 its contribution to the effective energy - momentum tensor would be @xmath228_{\mu\nu } = { f(\phi_\xi ) \over 1 - 6\xi \;\phi^2_\xi } \ ; t^m_{\mu\nu}.\ ] ] this behaves _ as if _ this matter interacts through an `` effective newton constant '' @xmath229 ( this is not yet quite the physical newton constant even in the standard case @xmath230 ; see section [ s : discussion ] below ) . then , unless @xmath231 is specifically chosen to counteract the @xmath232 factor , this effective newton constant would change its sign form one asymptotic region to the other , producing weird effects on ordinary matter . one can try to `` build '' symmetric wormholes beginning from these asymmetric wormholes @xcite and performing `` thin - shell surgery '' @xcite . in this way , one could potentially restrict the peculiar effects on matter to a thin region around the wormhole throat . we have seen that a non - minimally coupled scalar field can violate all the energy conditions at a classical level . the violation , in principle , of the anec inspired us to look for the possible existence of traversable wormhole geometries supported by these non - minimal scalar fields . we have obtained the classical solutions for gravity plus a massless arbitrarily coupled scalar field in the restricted class of spherically symmetric and static configurations . let separate the different cases 1 . @xmath41 : + we find naked singularity geometries and , in the case of a constant value for the scalar field , we recover the schwarzschild solution . @xmath80 : + in this case we have the usual class of minimally coupled solutions with its naked singularities . again , we recover the schwarzschild solution for a constant field . @xmath77 : + we find assorted naked singularities . in some of them the geometry shrinks to zero , in others the singularity is placed in an asymptotic region , and yet others there is a scalar curvature singularity at a finite size spherical section . + for a constant value of the field we recover the schwarzschild solution , but if this value is exactly @xmath233 the geometry can be arbitrarily chosen from among the solutions for the conformal @xmath49 case . + apart from these solutions we find a two - parameter family of perfectly well - defined traversable wormholes geometries . given an asymptotic value for the scalar field , with absolute value lower than @xmath191 , and a scalar charge , the asymptotic mass for a traversable wormhole is fixed . @xmath2 : + the case of conformal coupling leads to the same type of naked singularities that were seen before , but the scalar curvature singularities are now absent . + the different geometries can appear either in combination with a suitable spatial dependent scalar field , or with a constant field @xmath234 . + here , we have also a two - parameter family of traversable wormholes with the only difference that now the asymptotic value of the scalar field can have arbitrary values . in fact , when one asymptotic region has an infinite value for the scalar field it means that it is not asymptotically flat , degenerating to a cornucopia @xcite . @xmath136 : + once more we find assorted naked singularities of the same type as in the conformal coupling . + for a special constant value of the field , @xmath233 , the geometry can again be arbitrarily selected from among the solutions for the conformal @xmath49 case . for a different constant value we only recover the schwarzschild solution . + also , there are specific traversable wormhole solutions in which an infinite asymptotic value for the scalar field can not be reached . since the potential existence of traversable wormholes is a perhaps somewhat disturbing possibility , we feel it a good idea to see where the potential pitfalls might be mathematically , we have exhibited exact classical traversable wormhole solutions to the einstein equations , and now wish to investigate the extent to which they should physically be trusted . to start with , scalar fields play a somewhat ambiguous role in modern theoretical physics : on the one hand they provide great toy models , and are from a theoretician s perspective almost inevitable components of any reasonable model of empirical reality ; on the other hand the direct experimental / observational evidence is spotty . the only scalar fields for which we have really direct `` hands - on '' experimental evidence are the scalar mesons ( pions @xmath235 ; kaons @xmath236 ; and their `` charm '' , `` truth '' , and `` beauty '' relatives , plus a whole slew of resonances such as the @xmath237 , @xmath238 , @xmath239 , @xmath240 , ) @xcite . not a single one of these particles are fundamental , they are all quark - antiquark bound states , and while the description in terms of scalar fields is useful when these systems are probed at low momenta ( as measured in their rest frame ) we should certainly not continue to use the scalar field description once the system is probed with momenta greater than @xmath241 . in terms of the scalar field itself , this means you should not trust the scalar field description if gradients become large , if @xmath242 similarly you should not trust the scalar field description if the energy density in the scalar field exceeds the critical density for the quark - hadron phase transition . ( note that if the scalar mesons were strict goldstone bosons [ exactly massless ] rather than pseudo goldstone bosons , then they could achieve arbitrarily large values of the field variable with zero energy cost . ) thus scalar mesons are a mixed bag : they definitely exist , and we know quite a bit about their properties , but there are stringent limitations on how far we should trust the scalar field description . the next candidate scalar field that is closest to experimental verification is the higgs particle responsible for electroweak symmetry breaking . while in the standard model the higgs is fundamental , and while almost everyone is firmly convinced that some higgs - like scalar field exits , there is a possibility that the physical higgs ( like the scalar mesons ) might itself be a bound state of some deeper level of elementary particles ( , technicolor and its variants ) . despite the tremendous successes of the standard model of particle physics we do not ( currently ) have direct proof of the existence of a fundamental higgs scalar field accepting for now the existence of a fundamental higgs scalar , what is its curvature coupling ? the parameter @xmath1 is completely unconstrained in the flat - space standard model . if we choose for technical reasons to adopt the `` new improved stress energy tensor '' for the higgs scalar in flat spacetime then one is naturally led to conformal coupling in curved spacetime @xcite . conformal coupling seems to be the `` most natural '' choice for the higgs curvature coupling , and we have seen in this note that both conformal coupling and the entire open half - line @xmath243 surrounding conformal coupling lead to traversable wormhole geometries . unfortunately adding a higgs mass results in analytically intractable equations . a third candidate scalar field of great phenomenological interest is the axion : it is extremely difficult to see how one could make strong interaction physics compatible with the observed lack of strong cp violation , without something like an axion to solve the so - called `` strong cp problem '' . still , the axion has not yet been directly observed experimentally . a fourth candidate scalar field of phenomenological interest specifically within the astrophysics / cosmology community is the so - called `` inflaton '' . this scalar field is used as a mechanism for driving the anomalously fast expansion of the universe during the inflationary era . while observationally it is a relatively secure bet that something like cosmological inflation ( in the sense of anomalously fast cosmological expansion ) actually took place , and while scalar fields of some type are presently viewed as the most reasonable way of driving inflation , we must again admit that direct observational verification of the existence of the inflaton field ( and its variants , such as quintessence ) is far from being accomplished . note that in many forms of inflation trans planckian values of the scalar field are generic and widely accepted ( though not universally accepted ) as part of the inflationary paradigm . a fifth candidate scalar field of phenomenological interest specifically within the general relativity community is the so - called `` brans dicke scalar '' . this is perhaps the simplest extension to einstein gravity that is not ruled out by experiment . ( it is certainly greatly constrained by observation and experiment , and there is no positive experimental data guaranteeing its existence , but it is not ruled out . ) the relativity community views the brans dicke scalar mainly as an excellent testing ground for alternative ideas and as a useful way of parameterizing possible deviations from einstein gravity . ( and experimentally and observationally , einstein gravity still wins . ) in this regard it is important to emphasize that the type of scalar fields we have been discussing in the present paper are completely compatible with current experimental limits on brans dicke scalars , or more generally , generic scalar - tensor theories @xcite . to take the observational limits and utilize them in our formalism you need to use the translations @xmath244 @xmath245 then , the physical newton constant , which takes its meaning within the perturbative ppn expansion around flat minkowski space , can be written as @xmath246 ( here , we have set the function @xmath247 in ( [ effective ] ) equal to one . however it must be pointed out that if @xmath248 then the current theories are more general even than the standard scalar - tensor models ) . in the wormhole solutions that we have found , the scalar field @xmath249 reaches a trans - planckian value in one of the asymptotic regions , making the physical newton constant negative . this means that , contrary to what is commonly done in scalar - tensor theories @xcite , the translated brans - dicke field should not be restricted to only positive values . in order to cover our wormhole solutions , negative values of the brans - dicke field are required . current solar system measurements imply @xmath250 @xcite . more precisely , vlbi techniques allow us to use the deflection of light by the sun to place very strong constraints on the ppn parameter @xmath251 , with @xcite @xmath252 the standard result that for brans dicke theories @xcite @xmath253 now provides the limit on @xmath254 quoted above . for instance , this constraint is very easily satisfied if the scalar field @xmath249 is a small fraction of the naive planck scale ( @xmath255 ) in the solar neighborhood . thus , the scalar theories we investigate in this paper are perfectly compatible with known physics . finally , the membrane - inspired field theories ( low - energy limits of what used to be called string theory ) are literally infested with scalar fields . in membrane theories it is impossible to avoid scalar fields , with the most ubiquitous being the so - called `` dilaton '' . however , the dilaton field is far from unique , in general there is a large class of so - called `` moduli '' fields , which are scalar fields corresponding to the directions in which the background spacetime geometry is particularly `` soft '' and easily deformed . so if membrane theory really is the fundamental theory of quantum gravity , then the existence of fundamental scalar fields is automatic , with the field theory description of these fundamental scalars being valid at least up to the planck scale , and possibly higher . ( for good measure , by making a conformal transformation of the spacetime geometry it is typically possible to put membrane - inspired scalar fields into a framework which closely parallels that of the generalized brans dicke fields . thus there is a potential for much cross - pollination between brans dicke inspired variants of general relativity and membrane - inspired field theories . ) so overall , we have excellent theoretical reasons to expect that scalar field theories are an integral part of reality , but the direct experimental / observational verification of the existence of fundamental fields is still an open question . nevertheless , we think it fair to say that there are excellent reasons for taking scalar fields seriously , and excellent reasons for thinking that the gravitational properties of scalar fields are of interest cosmologically , astrophysically , and for providing fundamental probes of general relativity . the fact that scalar fields then lead to such widespread violations of the energy conditions @xcite , with potentially far - reaching consequences like a `` universal bounce '' ( instead of a big - bang singularity ) @xcite , the traversable wormholes of this paper ( see also @xcite ) , and possibly even weirder physics ( see for example @xcite ) , leads to a rather sobering assessment of the marked limitations of our current understanding . the research of cb was supported by the spanish ministry of education and culture ( mec ) . mv was supported by the us department of energy . mv wishes to thank jacob bekenstein for his interest and comments . 99 s.w . hawking and g.f.r . ellis , _ the large scale structure of space - time _ , ( cambridge university press , england , 1973 ) . m. visser , _ lorentzian wormholes : from einstein to hawking _ , ( american institute of physics , woodbury , 1995 ) . r. schoen and s.t . yau , _ commun . math . * 65 * ( 1979 ) 45 . olum , _ phys . _ * 81 * ( 1998 ) 3567 ; gr - qc/9805003 . m. visser , b. bassett , s. liberati , _ superluminal censorship _ , gr - qc/9810026 . + proceedings of the third meeting on constrained dynamics and quantum gravity , qg99 , villasimius , italy , sept . 1999 . b ( proc . supl . ) 88 ( 2000 ) 267 . m. visser , b. bassett , s. liberati , _ perturbative superluminal censorship and the null energy condition _ , gr - qc/990802 ; + general relativity and relativistic astrophysics , proceedings of the eighth canadian conference , edited by c.p burgess and r.c . meyers , ( aip press , melville , new york , 1999 ) , pages 301305 . j. l. friedman , k. schleich , and d. m. witt , _ phys . * 71 * ( 1993 ) 1486 . mayo and j.d . bekenstein , _ phys . * d54 * ( 1996 ) 5059 . m. visser , _ phys . _ * d46 * ( 1992 ) 2445 ; hep - th/9203057 . senovilla , _ gen . grav . _ * 30 * ( 1998 ) 701 . morris and k.s . thorne , _ am . j. phys . _ * 56 * ( 1988 ) 395 . d. hochberg and m. visser , _ phys . _ * 81 * ( 1998 ) 746 . d. hochberg , a. popov , and s.v . sushkov , _ phys . _ * 78 * ( 1997 ) 2050 . flanagan and r.m . wald , _ phys . rev . _ * d54 * ( 1996 ) 6233 . c. and m. visser , _ phys . b _ * 466 * ( 1999 ) 127 . m. visser and c. , `` energy conditions and their cosmological implications '' , cosmo99 proceedings ; trieste , sept / oct 1999 , world scientific ; gr - qc/0001099 . a. agnese and m. la camera , _ phys . _ * d51 * ( 1995 ) 2011 . nandi , a. islam , and j. evans , _ phys . _ * d55 * ( 1997 ) 2497 . anchordoqui , s. perez bergliaffa , and d.f . phys . rev . _ * d55 * ( 1997 ) 5226 . m. visser and d. hochberg , _ generic wormhole throats _ , + _ in : _ the internal structure of black holes and spacetime singularities , + _ edited by : _ l.m . burko and a. ori , + ( institute of physics press , bristol , 1997 ) . + gr - qc/9710001 . d. hochberg , _ phys . b _ * 251 * ( 1990 ) 349 . b. bhawal and s. kar , _ phys . rev . _ * d46 * ( 1992 ) 2464 . bekenstein , _ phys . rev . _ * d11 * ( 1975 ) 2072 . m. visser , _ science _ , * 276 * ( 1997 ) 8890 . j. froyland , _ phys . rev . _ * d25 * ( 1982 ) 1470 . bekenstein , _ ann . * 82 * ( 1974 ) 535 . duruisseau , _ gen . _ * 15 * ( 1983 ) 285 . fisher , _ zh . exp . fiz . _ * 18 * , 636 ( 1948 ) ; gr - qc/9911008 . janis , e.t . newman , and j. winicour , _ phys . lett . _ * 20 * ( 1968 ) 878 . m. wyman , _ phys . _ * d24 * ( 1981 ) 839 . m. cavaglia and v. de alfaro , _ int . phys . _ * d6 * ( 1997 ) 39 . a. agnese and m. la camera , _ phys . rev . _ * d31 * ( 1985 ) 1280 . m. visser , _ phys . rev . _ * d39 * ( 1989 ) 3182 . particle data group , http://pdg.lbl.gov , _ euro . phys . j. _ * c3 * , 1 ( 1998 ) . c.m . will , _ theory and experiment in gravitational physics _ , revised edition , ( cambridge university press , england , 1993 ) . s. weinberg , _ gravitation and cosmology _ , ( john wiley , new york , 1972 ) . santiago and a.s . silbergleit , _ on the energy momemtum tensor of the scalar field in scalar - tensor theories of gravity _ ; gr - qc/9904003 . eubanks , d.n . matsakis , j.o . martin , b.a . archinal , d.d mccarthy , + s.a . klioner , s. shapiro , and i.i . shapiro , + _ advances in solar system tests of gravity _ , + proceedings of the joint aps / aapt 1997 meeting , 18 - 21 april 1997 , + washington d.c . + ( abstract only ) . d. hochberg , c. molina , and m. visser , _ phys . * d59 * ( 1999 ) 044011 ; gr - qc/9810029 . c. molina and m. visser , _ phys . _ * b455 * ( 1999 ) 90 - 95 ; gr - qc/9810023 .

in a previous paper ( maoz 1999 ) , we reported a hubble space telescope ( hst ) cepheid distance to the galaxy ngc 4258 obtained using the calibrations and methods then standard for the key project on the extragalactic distance scale . here , we reevaluate the cepheid distance using the revised key project procedures described in freedman ( 2001 ) . these revisions alter the zero points and slopes of the cepheid period - luminosity ( p - l ) relations derived at the large magellanic cloud ( lmc ) , the calibration of the hst wfpc2 camera , and the treatment of metallicity differences . we also provide herein full information on the cepheids described in maoz 1999 . using the refined key project techniques and calibrations , we determine the distance modulus of ngc 4258 to be mag ( unique to this determination ) mag ( systematic uncertainties in key project distances ) , corresponding to a metric distance of mpc and 1.2 from the maser distance of mpc . we also test the alternative cepheid p - l relations of feast ( 1999 ) , which yield more discrepant results . additionally , we place weak limits upon the distance to the lmc and upon the effect of metallicity in cepheid distance determinations .

astro-ph0012377

distances to other galaxies obtained by observations of cepheid variables with the hubble space telescope ( hst ) lie at the core of most recent efforts to determine the extragalactic distance scale . the small observed scatter in the relationship between cepheids pulsation periods and luminosities , their large numbers ( which allow many independent measures of the distance to a galaxy ) , and the simplicity of the basic physics underlying their variability all have made them uniquely suitable for this purpose . as a consequence of their integral role in establishing the distance scale , however , any changes in the calibration and application of the cepheid period luminosity ( p l ) relationship will affect many other secondary methods of distance measurement . any improvements in techniques for obtaining cepheid distances are thus extremely valuable ; but this also means that any changes in this vital link should be well scrutinized if they are to be adopted . in a paper describing the final results of the hst key project on the extragalactic distance scale ( henceforward , the key project ) , freedman ( 2001 ) make a number of refinements to the techniques used in earlier papers in the series based upon newly - available information . first and foremost , microlensing experiments ( e.g. , udalski 1999 ) and dedicated efforts ( sebo 2001 ) have greatly enlarged the set of calibrating lmc cepheids beyond what had been studied when the initial key project p l relation was determined ( madore & freedman 1991 ) . the resulting samples have revealed a modest correction to the previously adopted p l slope for @xmath5 observations ; the @xmath6band slope remains unchanged . in the key project methodology , cepheid magnitudes are corrected for extinction using their observed color excess @xmath7 , where @xmath8 is the expected color of an unreddened cepheid of given period based upon the lmc calibrated p l relations . this procedure is sensitive to the reddening , since @xmath9 ) . an error in the p l slope in @xmath5 thus propagates into a larger , period dependent error in the true distance modulus ; in this case , @xmath10 . because in more distant galaxies only brighter , longer - period cepheids are observable , this generally results in a distance - dependent revision to hst cepheid distances which in extreme cases can reach -0.20 mag . further details are given in freedman et al . ( 2001 ) . second , our understanding of charge transfer efficiency and related effects in the wfpc2 ccds has greatly improved in recent years ( e.g. , stetson 1998 , dolphin 2000 ) , motivating revisions in the hill ( 1998 ) photometric zero points used in most earlier key project papers . freedman adopt the calibration of stetson ( 1998 ) , which results in a -0.02 mag mean adjustment in @xmath6 and -0.04 mag in @xmath5 from the hill `` long '' zero points . carrying this through the key project procedures for obtaining reddening - corrected distance moduli , they apply a net correction of -0.07@xmath11 mag to distances obtained using the hill `` long '' zero point ( where the error adopted reflects a conservative estimate of systematic differences among recent calibrations ) . the third change made in the revised key project procedures is the adoption of cepheid distance moduli adjusted for metallicity effects as standard . the typical metallicity of cepheids in the lmc with which the p l relation is calibrated differs by @xmath12 dex from that in many of the fields observed in the course of the key project . metallicity differences may produce measurable differences in the colors and magnitudes of cepheids in those fields from ones found in the lmc ; unfortunately , neither theoretical calculations ( as those of alibert 1999 or musella 1999 ) nor observations ( cf . kennicutt 1998 , sasselov et al . 1997 , kochanek 1997 , or nevalainen & roos 1998 ) have provided a definitive determination of the magnitude , or even sign , of this effect . freedman ( 2001 ) adopt a correction to cepheid magnitudes of @xmath13 mag / dex ( based upon current observational results ) as standard . in previous key project work , distance moduli uncorrected for metallicity effects were primarily used , though results for a correction of -0.24 mag / dex were also given . in their error budget , freedman estimate the potential systematic error in a typical key project cepheid distance measurement due to corrections for differences in metallicity from the lmc to be @xmath14 mag . even with these revisions , important uncertainties remain in the cepheid - calibrated distance scale . foremost among these is the distance to the large magellanic cloud ( lmc ) , using which the key project period luminosity relation has been calibrated . a wide variety of measurements of the distance to the lmc have been performed in the past few years . many of these studies disagree with each other statistically , spanning roughly 0.5 mag in distance modulus . freedman ( 2001 ) adopt @xmath15 mag as the distance modulus of the lmc ; the uncertainty in the lmc distance modulus leads directly to a systematic uncertainty in the cepheid distance scale of @xmath16 mag . after the uncertainties in the lmc distance and metallicity corrections , the next most significant contribution to the key project systematic error budget is the difficulty in determining zero points for wfpc2 photometry , which is estimated to lead to a @xmath17 mag uncertainty in hst cepheid distance moduli ; a variety of other systematic effects could enter at lower levels . given the remaining uncertainties , it is worthwhile to test the revised key project distance scale using a galaxy with a well - known , primary distance and with a metallicity typical of galaxies observed in the course of the key project . to be useful , such a test requires that cepheids be observed with the same instruments , filters , and parameter measurement techniques as used for objects in the key project sample . such a test has already been applied to the original key project distance scale based upon observations of cepheids in ngc 4258 ( maoz 1999 ) . the spiral galaxy ngc 4258 ( sabbc , @xmath18 mag ) presents a unique opportunity for such a test because of the precision with which its distance has been measured in a manner independent of the conventional ladder of astronomical distance scales ( freedman 1998 ) . furthermore , its metallicity and distance are similar to those of typical targets of hst cepheid programs . the distance to ngc 4258 , @xmath19 mpc ( corresponding to distance modulus @xmath20 mag ) , has been determined using its apparently simple , keplerian circumnuclear disk delineated by line - emitting water masers that orbit a supermassive black hole at its center ( miyoshi 1995 , maoz 1995 ) . this disk was discovered by vlbi observations of water maser emission from the central region of the galaxy ( miyoshi 1995 ) . it is about @xmath21 mas in diameter , @xmath22 mas in thickness , rotates at speeds of @xmath23 , and is viewed by us from nearly edge - on . most remarkably , the rotation curve of the maser sources is keplerian to high precision ( @xmath24 ) , which provides very strong evidence for a supermassive black hole at the galaxy center ( maoz 1995 ) . the high angular resolution ( @xmath25mas ) and spectral resolution ( @xmath26 ) of vlba observations ( moran 1995 ) allow a precise definition of the disk structure and kinematics . combining the observed rotation velocities with the measured centripetal acceleration in the disk ( greenhill 1995 ) or with the observed proper motions of the maser sources ( herrnstein 1999 ) permits independent measurements of the physical size of the disk ; comparing these to its observed angular extent yields the distance to the galaxy via simple geometry . these distance measurements are subject to only small uncertainties . the differential systematic error in the maser positions is @xmath27 mas ( moran 1995 ) , translating into a relative distance error smaller than @xmath28 . the distance to the masing disk scales with the disk s inclination as @xmath29 ; since the disk is viewed nearly edge - on , @xmath30 ( herrnstein 1996 ) , an error even as high as @xmath31 would introduce a distance error of only @xmath32 . relativistic corrections due to gravitational redshift and transverse doppler shift have been taken into account and are much smaller . the disk is slightly warped , but the distance determination is not sensitive to the warp model adopted . the warp contributes , though , to the small uncertainty in the disk inclination mentioned above . the total estimated uncertainty in this distance is @xmath33 mpc if the disk is presumed to be circular ( which it appears to be to better than @xmath28 ) . if nonzero eccentricities are allowed , the uncertainty increases to @xmath34 mpc ( we adopt this more conservative value for all further discussions ) . the direct , geometric methods used are believed to have minimal unknown systematic uncertainties . the two routes to a distance ( proper motions and accelerations ) yield results in agreement with each other to @xmath32 . as described in maoz ( 1999 ) , we have therefore obtained and analyzed hst observations of ngc 4258 with the intention of testing , and potentially of better determining , the zero point of the cepheid p - l relation . in that work , we used the then - standard key project methodologies and calibrations , and found a @xmath35 discrepancy between the hst cepheid distance and that obtained from studies of the masing disk . in this paper , we perform a similar test after obtaining a cepheid distance using the revised key project procedures , allowing an evaluation of the new techniques . we also provide herein full information ( locations , light curves , finding charts , etc . ) for the cepheids described in maoz 1999 . we present the observations , the reduction of the data and the search for cepheids in @xmath36 , the resulting derivation of a cepheid distance in @xmath37 , and the implications in @xmath38 . we observed a portion of ngc 4258 using the wide field and planetary camera 2 ( wfpc2 ) instrument and the hubble space telescope on 11 epochs in 1998 . the data were acquired with an optimal power - law spacing between them as described in freedman ( 1994 ) and madore & freedman ( 2001 ) . the @xmath39 and @xmath40 filters were used for a combined total of one orbit at every epoch , with two frames obtained using each filter to limit the effects of cosmic rays ; the exposure time per frame was 500 seconds . a journal of observations is given in table 1 . to simplify analysis , a fixed orientation was maintained for all epochs . the field observed is superimposed on a ground - based image of ngc 4258 in fig . 1 . we have obtained photometry from these data with two commonly - used software packages : daophot / allframe and dophot . for both analyses , the data were first preprocessed via the standard space telescope science institute pipeline ( holtzman 1995 ) . for the allframe photometry each frame was also corrected for vignetting and geometrical effects on the effective pixel area as described in stetson ( 1998 ) . photometry was then performed on each of the data frames using the daophot ii / allframe package ( stetson 1987 ) . allframe fits a predefined point - spread function ( psf ) to all stars on a frame and iteratively determines their magnitudes . the procedures used throughout were similar to those of the key project ( see stetson 1998 ) . the photometric zero points of stetson ( 1998 ) were used . aperture corrections for each frame were derived using a set of bright , isolated stars . in addition to the cepheid analysis described below , we also attempted to determine a tip of the red giant branch ( trgb ) distance ( lee 1993 ) to ngc 4258 using the allframe photometry for wf2 and wf3 . however , our observations were insufficiently deep and contamination by other stellar populations too great to allow any convincing detection of the trgb . the dophot photometry was performed using a variant of the dophot package ( schechter 1993 , saha 1994 ) which was developed especially to deal with the photometry of undersampled images such as those obtained with the hst . in addition to the data processing mentioned earlier , pairs of cosmic ray split images were combined prior to performing the photometric reduction with dophot . the algorithm used for this is designed to reject cosmic ray events ; particular care is taken to ensure that the photometry of real objects is preserved . further discussion of the application of dophot to photometry of hst images can be found in saha ( 1996a ) , ferrarese ( 1996 , 1998 ) , and hill ( 1998 ) . the photometric calibration adopted again follows stetson ( 1998 ) , as per the revised key project procedures of freedman ( 2001 ) . the limited number of bright , uncrowded stars prevented us from deriving reliable spatially - dependent aperture corrections for dophot . we therefore adopted corrections obtained from observations of an uncrowded field in the leo i dwarf galaxy . aperture photometry conducted on the ngc 4258 individual frames , as well as on a deep frame obtained by combining all available epochs , produces results in agreement with the leo i aperture corrections at the 0.03 mag level , which we therefore adopt as a measure of the uncertainty in the correction themselves . searches for cepheid variables using the allframe photometric measurements were conducted using two different algorithms . one of these selects candidate variables via a modified welch - stetson test and performs a nonlinear fit of template cepheid light curves to their photometry to assess their variability and determine their parameters ; see stetson ( 1996 ) and stetson ( 1998 ) for further description . the other linearly fits template light curves defined on a grid in period and phase to all stars with well - determined photometry ; cepheids are then identified by a set of criteria that are effective at eliminating nonvariables . this algorithm is described in more detail in newman et al . both searches independently yielded similar sets of candidate cepheid variables and parameter estimates in good agreement . these algorithms in combination yielded 21 potential cepheid variables . a search for variable stars was also conducted using dophot magnitude measurements for the @xmath6 and @xmath41band frames independently following the procedure described by saha & hoessel ( 1990 ) . we required that a star be detected at at least 8 of the 11 epochs in order to be checked for variability . we also excluded all stars in crowded regions by rejecting candidates that had a companion contributing more than 50% of the total light within a two - pixel radius . a detailed discussion of the search procedure can be found in ferrarese ( 1996 ) . a star meeting the above constraints was flagged as a variable if @xmath42 or @xmath43 where @xmath44 and @xmath45 are as used in saha & hoessel ( 1990 ) . several spurious variables were detected by this procedure as a consequence of non - gaussian sources of error and various anomalies in the images ( e.g. , residual cosmic ray events ) along with the crowding referred to earlier . each variable star candidate was visually inspected by blinking several of the individual frames against each other . the best period for each variable was selected by phasing the data for all periods between 3 and 60 days in incremental steps of 0.1 days . although in most cases the adopted period corresponds to a minimum value of the phase dispersion , in a few cases an obvious improvement of the light curve was obtained for a slightly different period . the dophot analysis identified a total of 28 potential cepheids . all cepheids found in either dataset were on wide field ( wf ) chips 2 or 3 ; we therefore will restrict our discussion to these for the remainder of this paper . on wf2 , the agreement of dophot and allframe results was well within the expected errors in the aperture corrections used ; the mean difference between allframe and dophot magnitudes for 24 bright , isolated stars was @xmath46 ( standard deviation ; standard error of the mean 0.009 ) mag for @xmath6 , and @xmath47 mag for @xmath5 . for wf3 , the mean difference for 30 stars was @xmath48 mag for @xmath6 , and @xmath49 mag for @xmath5 . mean magnitudes for cepheids yielded results consistent with these to within 1@xmath3 , albeit with much larger standard errors due to their fainter magnitudes . we believe that the wf3/@xmath5 results are an aberration closely related to the discrepant distance moduli obtained from allframe magnitudes on the two chips ( q.v . @xmath50 3 ) . this anomoly is plausibly accounted for by the difficulties of determining aperture corrections in the observed fields , which contain few bright , isolated stars . if we presume that this is the case , the wf3/@xmath5 allframe photometry may be corrected by bringing it onto the same system as the wf2 and wf3/@xmath6 allframe magnitudes ; i.e. , adjusting the wf3/@xmath5 allframe magnitudes to be @xmath51 mag fainter than dophot ( the average difference from the other three chip / filter combinations ) , rather than @xmath52 mag without any correction . averaging the allframe - dophot differences from those three cases may be justified by the fact that errors in aperture corrections , etc . are generally highly correlated between @xmath6 and @xmath5 and from one wf chip to another ; and indeed , the allframe - dophot offsets for wf2/@xmath6 and @xmath5 and wf3/@xmath6 are all quite consistent , agreeing to within @xmath53 mag . therefore , in addition to presenting results for unmodified allframe photometry , in the next section we also provide distance measurements obtained from allframe data `` corrected '' by subtracting @xmath54 mag from wf3/@xmath5 magnitudes . we have identified and determined light curves , periods , mean magnitudes , and colors for 15 definitive cepheids in ngc 4258 . all of these stars fulfill four criteria : they are identified as variable by all three search techniques , they fit a cepheid template light curve with reasonable @xmath55 , they visibly vary in blink comparisons in both @xmath39 and @xmath40 images , and they have negligible statistical probability of being misidentified nonvariables . a variety of methods exist for determining the mean magnitudes of cepheids . in this paper , we use the intensity - weighted mean magnitude obtained from a fit to the cepheid light curve using the templates of stetson ( 1996 ) , also known as `` template fit '' mean magnitudes . this is the preferred method for the key project , providing a robust method of determining mean magnitudes analogous to those obtained for more densely sampled datasets ( such as the lmc cepheids used for calibrating the p l relation ) . we note that adopting other standard magnitude averaging methods changes the ngc 4258 distance modulus obtained by no more than 0.04 mag . we use the cepheid periods determined in the dophot analysis for all distance modulus calculations , as these yielded smaller scatter about the p l relation for all datasets and averaging methods than alternatives ( e.g. , periods determined from template fits ) , likely reflecting the fact that those periods were refined by hand when necessary to improve the cepheid light curves , rather than being determined solely by an automated algorithm . we have adopted the dophot photometry for all major conclusions reported here , since as discussed in @xmath50 3 , allframe photometry yielded internally discrepant distances from cepheids on the two wfpc2 chips used ; however , as an additional consistency check we also provide allframe values in much of what follows . we present the locations and characteristics of the cepheids found in table 2 , and complete dophot photometry for those stars in table 3 . the positions of the cepheids are shown in fig . 2 , and detailed finding charts may be found in fig . 3 . dophot light curves for the cepheids found are depicted in fig . 4 . in accordance with the revised key project methodology , we adopt the p - l relation slopes of freedman ( 2001 ) and only fit for differences in the zero point . their lmc calibration yields mean absolute magnitudes for cepheids @xmath56 where @xmath57 and @xmath58 are the intensity - weighted mean absolute johnson @xmath6 and kron - cousins @xmath5 magnitudes of the star and @xmath59 is its period in days . the same relations have been used by freedman ( 2001 ) and macri ( 2001 ) . fitting the observed magnitudes of cepheids in ngc 4258 with such relations yields a measurement of the apparent distance modulus of the galaxy . we have done this fitting with the standard key project processor , which determines overall distance moduli as an unweighted mean of the values for individual stars . from the difference between the absolute magnitudes of lmc cepheids ( for an assumed lmc distance modulus of 18.50 mag ) and the observed magnitudes of ngc 4258 cepheids we then may derive a @xmath6 or @xmath5 distance modulus to ngc 4258 . we present the resulting distance moduli for both allframe and dophot photometry in table 4 , both for the entire sample and the subsets of cepheids on either chip 2 or chip 3 ( which can indicate the presence of catastrophic photometric or other errors ) . the dophot ngc 4258 p - l relations are plotted in figure 5 , and those from allframe in figure 6 . in key project procedures , the colors of the observed cepheids ( compared to those of a calibration set of such stars in the lmc whose reddening has been assumed ) are then used to correct for line - of - sight of extinction assuming a cardelli ( 1989 ) reddening law . this may be done by comparing the distance moduli obtained in the @xmath6 and @xmath5 bands ; the difference of the two measures the average value of @xmath60 . because the key project processor uses only unweighted averages with no rejection , this is equivalent both to applying the extinction correction star - by - star and then averaging and to fitting the p - l relation for a reddening - corrected `` wesenheit '' magnitude , @xmath61 ( madore 1982 , madore & freedman 1991 ) . this method yields @xmath62 mag and an extinction - corrected distance modulus of @xmath63 mag from dophot photometry for all cepheids ; for the allframe photometry , the corresponding numbers are @xmath64 and @xmath65 mag . the true moduli for the subset of cepheids on either wf2 or wf3 are also listed in table 4 . note that in the allframe dataset , cepheids on the two chips yield extinction - corrected distance moduli differing by 0.25 mag , or 1.9@xmath3 ; the dophot moduli differ by only 0.07 mag ( 0.5@xmath3 ) . furthermore , the value for wf2 is quite consistent with that obtained from dophot photometry , both for single chips samples and overall ; this makes the wf3 allframe results particularly suspect . the bulk of this discrepancy is eliminated if we perform the wf3/@xmath5 correction described in @xmath50 2.3 . that adjustment of wf3/@xmath5 magnitudes by @xmath66 mag reduces the overall distance modulus yielded by cepheids on wf3 by @xmath67 mag . such a correction reduces the allframe wf2/wf3 discrepancy to 0.09 mag , a 0.7@xmath3 difference . as seen in table 4 , this correction would leave allframe @xmath6 distance moduli unchanged , alter the wf3 and overall average @xmath5 moduli to 29.77 mag , and reduce the extinction - corrected distance modulus to @xmath68 mag for wf3 , or @xmath69 mag overall ( where the errors include the propagated error from the uncertainty in the allframe - dophot differences , 0.021 mag , added in quadrature ) . these values are in much better agreement with those obtained from dophot photometry ( for which the overall , extinction - corrected distance modulus was @xmath63 mag ) . applying this correction also reduces the scatter in the overall , extinction - corrected allframe distance modulus from 0.35 mag to 0.25 mag . freedman ( 2001 ) find that differences in metallicity have an effect on extinction - corrected cepheid distances of 0.2@xmath70 mag / dex . using the fits to data on ngc 4258 hii regions from zaritsky , kennicutt , and huchra ( 1994 ) , we estimate the metallicity in our hst fields to be 12+log(o / h)=@xmath71 , 0.35 dex higher than that adopted for cepheids in the lmc . this leads to a correction of + 0.07@xmath72 mag to the distance moduli we have derived . the revised key project procedures adopt metallicity - corrected values for the distance modulus as their primary estimate ( in contrast to previous practice ) ; we thus do likewise , and obtain a final true modulus to ngc 4258 of 29.42 mag . because the metallicity of ngc 4258 is quite typical of key project targets , we treat the uncertainty in the metallicity correction as a systematic uncertainty in the distance scale . we present an error budget for our measurement of the distance to ngc 4258 in table 5 . in addition to the random errors determined in the course of fitting p l relations , uncertainties in the aperture corrections used are also random between different cepheid target galaxies . from studies of globular clusters and key project galaxies , we expect these to be approximately 0.05 magnitude or less in both @xmath6 and @xmath5 , and highly correlated between the two bands ; the combined uncertainty due to the aperture corrections in the reddening - corrected distance modulus would then still be 0.05 mag . differences between allframe and dophot photometry of bright stars are much smaller than this in all cases save wf3/@xmath5 . since it appears that the difference between the overall dophot and uncorrected allframe distance primarily reflects a correctable error in the wf3/@xmath5 allframe photometry alone , we estimate that photometric errors in the dophot results which are unique to our study of ngc 4258 may constitute 0.05 mag . adding all potential sources in quadrature , the total random uncertainty in our determination of a cepheid distance to ngc 4258 ( @xmath73 in table 5 ) is 0.09 magnitude . this measurement is also subject to a number of potential sources of systematic error that affect key project distance determinations similarly , as described in table 5 ; their possible contributions have been estimated to total @xmath1 mag ( ferrarese 1999 , freedman 2001 ) . for those potential systematic errors which affect all cepheid distances obtained in the same manner as ours uniformly , we have adopted the uncertainty estimates of the key project ( freedman 2001 ) ; more detailed descriptions may be found therein . we thus obtain a cepheid distance modulus to ngc 4258 of @xmath74 mag ( unique to this determination ) @xmath1 mag ( systematic uncertainties in key project distances ) , corresponding to a metric distance of 7.8 @xmath33 mpc @xmath34 mpc . when treated in the same way , the uncorrected allframe results yield a distance modulus of @xmath75 mag , corresponding to a metric distance of @xmath76 mpc , while the corrected allframe results yield a distance modulus of @xmath77 mag , corresponding to a metric distance of @xmath78 mpc . the distance to ngc 4258 derived from observations of cepheids using the revised key project methodologies is thus not significantly greater than the maser distance of 7.2 @xmath79 mpc ( herrnstein 1999 ) . in assessing the validity of the calibration of the cepheid distance scale using ngc 4258 , we must consider how significant the difference we have found is . first of all , we may examine whether the differences are consistent within the random and systematic error budgets of the two processes ; i.e. , whether previously considered sources of error are sufficient to account for what we have found . the cepheid and maser distances differ by 0.9@xmath3 if we add in quadrature our measurement uncertainty of 0.3 mpc , the key project systematic error estimate of 0.5 mpc , and the maser distance error estimate of 0.5 mpc ; potential systematic errors in either technique do not appear to have been underestimated . in performing a test of the validity of the revised key project distance scale , however , we must consider what sort of a discrepancy we can measure . all hst cepheid distances following the revised techniques will share the systematic errors that affect our results . thus , we should consider only the random errors unique to this measurement and those errors affecting the maser distance in determining whether a recalibration would be an improvement . in that case , we find a 1.2@xmath3 difference , as opposed to a 1.6@xmath3 difference if the results of maoz ( 1999 ) were evaluated with the same error budget . put differently , if we were to presume the maser distance to ngc 4258 is correct and recalibrate the distance scale based upon our cepheid observations of ngc 4258 , all revised key project distances would have to be reduced by 0.19 mag , increasing the resulting measurements of the hubble constant by @xmath80 ; the total systematic error budget for the new calibration would be 0.16 mag ( 8% ) , slightly greater than that resulting from the current , lmc - based methods . modest changes in calibration or methodology would be sufficient to bring the cepheid and maser distances into substantially better agreement ; for instance , the lmc distance need only be reduced by a few hundredths of a magnitude , well within the key project estimate of its uncertainty , to bring the difference below 1@xmath3 . we can similarly use our measurement of the distance to ngc 4258 to test other calibrations of the cepheid p - l relation . in particular , we have applied the milky way - based calibration of feast ( 1999 ) , which corresponds to : following the conventions of equation 1 . note that the slopes adopted in deriving these relations are substantially different from those used by freedman ( 2001 ) . although the cepheids in feast s calibration are not at uniform distance , making their use to calibrate the p - l relation somewhat more complicated , they are independent of any assumptions about the lmc distance and of very similar metallicity to the fields we have studied . using these relations leads to a cepheid distance modulus for ngc 4258 of @xmath82 ( random ) mag , 0.23 mag greater than the maser result ; a metallicity correction of -0.2 mag / dex increases this by only 0.01 mag . this is a total discrepancy of 1.3@xmath3 if we consider all possible errors and assume a systematic uncertainty of 0.05 mag in feast s calibration , 1.2@xmath3 for 0.1 mag , or 1@xmath3 for 0.15 mag uncertainty . if , as above , we compare the discrepancy to what we can measure , we find that the maser and cepheid distances are 1.5@xmath3 apart ; the calibration of feast ( 1999 ) might be improved by rereferencing it to ngc 4258 . some authors ( e.g. gibson 2000 ) have used the results of maoz et al . ( 1999 ) to estimate the distance to the large magellanic cloud , based upon the assumption that the maser distance to ngc 4258 is correct and that any differences between the cepheid distance to that galaxy and the maser distance are due to an error in the distance to the lmc assumed in calibrating p l relations . such a procedure is subject to many sources of uncertainty ; the random uncertainty in the resulting lmc distance will correspond to the sum in quadrature of the random uncertainties in the maser and cepheid distances to ngc 4258 . furthermore , the uncertainty due to potential systematic errors in that procedure includes contributions from all possible systematics in the maser and key project error budgets , save the distance to the lmc itself . if we nevertheless proceed in this manner , we determine an lmc distance modulus of 18.31 @xmath83 0.11 ( random ) @xmath83 0.17 ( systematic ) . similarly , we may use the maser and cepheid distances to ngc 4258 to place limits on the effect of metallicity on cepheid luminosities . in that case , we should consider all random and systematic uncertainties in the maser or cepheid distances save those due to metallicity effects ; they total 0.21 mag . thus , the cepheid distance becomes 1@xmath3 discrepant from the maser distance if there is a total metallicity correction of + 0.09 mag to the uncorrected cepheid distance , and 2@xmath3 discrepant with a correction of + 0.30 mag . as the metallicity of ngc 4258 is @xmath84 dex greater than that of the lmc , we thus can limit the effect of metallicity on cepheid mean magnitudes to -0.26 mag / dex at 1@xmath3 , or -0.9 mag / dex at 2@xmath3 . ngc 4258 has provided the most stringent geometrical test of the revised key project distance scale so far . the 0.26 mag discrepancy between the maser distance and the cepheid distance to ngc 4258 obtained with the original key project methods diminishes to 0.19 mag when the revisions of freedman ( 2001 ) are applied . this provides one piece of evidence that the changes made to the key project distance scale are , in fact , improvements . a stronger test of the hst cepheid distance scale based on the maser distance to ngc 4258 would require a substantially larger sample of cepheids ( which should reduce the uncertainties in determining the @xmath85 p l relations and the reddening ) and better determination of aperture corrections ; these issues can be addressed simultaneously by searching for cepheids with hst in a field that contains more stars and has undergone more recent star formation , preferably with the higher resolution that will be afforded by the advanced camera for surveys . it would be reasonable to expect that observations of a region richer in cepheids might yield as many as 3 times the number of cepheids ( giving a distance modulus uncertainty of 0.04 magnitude ) and aperture corrections accurate to @xmath86 magnitude ; better agreement between allframe and dophot analyses might also occur with improved data . reductions of uncertainties in the maser distance ( e.g. , via improved constraints on the eccentricity of the circumnuclear disk ) , would also be greatly beneficial for its use to calibrate the extragalactic distance scale . successful maser distances to other galaxies , establishing a hubble relation , would more firmly establish this novel and promising technique . the data available at present are sufficient only to test calibrations of the extragalactic distance scale , but that alone is of great value . with improvements in both cepheid and maser analyses , ngc 4258 has great potential for establishing a new primary step in the distance ladder , reducing the potential systematic errors in measurements of the hubble constant to perhaps as little as 5% , and bypassing controversies over the distance to the large magellanic cloud and the effect of metallicity on the colors and magnitudes of cepheids entriely . 999 bartel , n. 1999 , baas , 194.6208 cardelli , j.c . , clayton , g.c . & mathis , j.s . 1989 , ap . j. , 345 , 245 dolphin , a. e. 2000 , pasp , 112 , 1397 eastman , r.g . , schmidt , b.p . , & kirschner , r. 1996 , ap . j. , 466 , 911 feast , m. 1999 , pasp , 111 , 775 ferrarese , l. 1998 , ap.j . , 464 , 568 ferrarese , l.f . 2000 , ap . j. , 529 , 745 freedman , w.l . 1997 , in _ critical dialogues in cosmology _ , ed . n. turok ( world scientific , singapore ) , 92 freedman , w.l . relativistic astrophysics and cosmology _ , eds . v. olinto , d.n . schramm , & j.a . frieman ( world scientific , singapore ) freedman , w.l . 1999 , _ particle physics and the universe _ ( world scientific press , singapore ) freedman , w.l . 2001 , apj , submitted freedman , w.l . 1994 , ap . j. , 427 , 628 gibson , b. k. 1999 , astro - ph/9910547 , preprint gould , a. & uza , o. 1998 , ap . j. , 494 , 118 greenhill , l.j . , henkel , c. , becker , r. , wilson , t.l . , & wouterloot , j.g.a . 1995 , a&a , 304 , 21 hedges , l.v . 1985 , _ scientific methods of meta - analysis _ ( academic press , orlando ) herrnstein , j.r . 1999 , nature , 400 , 539 kennicutt , r.c . jr . , freedman , w.l . , & mould , j.r . 1995 , a.j . , 110 , 1476 kennicutt , r. c. jr . 1998,ap . j. , 498 , 181 krauss , l. m. 1998 , ap . j. , 501 , 461 jacoby , g.h . 1992 , pasp 104 , 599 lee , m.g . , freedman , w.l . , & madore , b.f . 1993 , apj , 417 , l553 macri , l. 2001 , apj , approved madore , b.f . & freedman , w.l . 1991 , pasp , 103 , 933 madore , b. f. 1999 , ap . j. , 515 , 29 maoz , e. 1995 ap.j . , 494 , l181 maoz , e. 1999 , nature , 401 , 351 miyoshi 1995 , nature , 373 , 127 mould , j.r . 2000 , ap.j . , 529 , 786 newman , j.a . 1999 , ap.j . , 523 , 506 panagia , n. 1999 in _ new views of the magellanic clouds _ , eds . chu , n.b . suntzeff , j.e . hesser , & d. bohlender ( a.s.p . , san francisco ) saha , a. , labhardt , l. , schwengeler , h. , macchetto , f.d . , panagia , n. , sandage , a. , & tammann , g.a . 1994 , ap.j . , 425 , 14 sakai , s. 2000 , ap . j. , 529 , 698 schechter , p.l . , mateo , m. , & saha , a. 1993 , pasp , 105 , 1342 stetson , p.b . 1994 pasp , 106 , 250 stetson , p. b. 1996 , pasp , 108 , 851 stetson , p.b . 1998 , ap.j . , 508 , 491 zaritsky , d. , kennicutt , r.c . jr . , & huchra , j.p . 1994 , ap . j. , 420 , 87 lccccc 1 & f555w & 1998 apr 20 & 4901847.600 & 1000 & u4f60101r/2r + & f814w & & 4901847.632 & 1000 & u4f60103r/4r + 2 & f555w & 1998 apr 21 & 4901850.428 & 1000 & u4f60201r/2r + & f814w & & 4901850.460 & 1000 & u4f60203r/4r + 3 & f555w & 1998 apr 23 & 4901854.189 & 1000 & u4f60301r/2r + & f814w & & 4901854.221 & 1000 & u4f60303r/4r + 4 & f555w & 1998 apr 25 & 4901859.086 & 1000 & u4f60401r/2r + & f814w & & 4901858.118 & 1000 & u4f60403r/4r + 5 & f555w & 1998 apr 28 & 4901863.866 & 1000 & u4f60501r/2r + & f814w & & 4901863.898 & 1000 & u4f60503r/4r + 6 & f555w & 1998 may 1 & 4901869.241 & 1000 & u4f60601r/2r + & f814w & & 4901869.273 & 1000 & u4f60603r/4r + 7 & f555w & 1998 may 5 & 4901878.034 & 1000 & u4f60701r/2r + & f814w & & 4901878.066 & 1000 & u4f60703r/4r + 8 & f555w & 1998 may 10 & 4901887.113 & 1000 & u4f60801r/2r + & f814w & & 4901887.145 & 1000 & u4f60803r/4r + 9 & f555w & 1998 may 15 & 4901897.864 & 1000 & u4f60901r/2r + & f814w & & 4901897.896 & 1000 & u4f60903r/4r + 10 & f555w & 1998 may 22 & 4901912.376 & 1000 & u4f61001r/2r + & f814w & & 4901912.408 & 1000 & u4f61003r/4r + 11 & f555w & 1998 may 30 & 4901928.500 & 1000 & u4f61101r/2r + & f814w & & 4901928.532 & 1000 & u4f61103r/4r + lccc|cccc|ccc c1 & 2 & 158.18 & 603.73 & 24.55 @xmath83 0.03 & 23.84 @xmath83 0.03 & 20.5 @xmath83 0.2 & 1.08 & 24.77 & 23.94 & 21.3 + c2 & 2 & 102.72 & 455.97 & 24.70 @xmath83 0.03 & 23.84 @xmath83 0.02 & 17.6 @xmath83 0.2 & 0.99 & 24.77 & 23.84 & 17.2 + c3 & 2 & 364.28 & 317.39 & 24.86 @xmath83 0.03 & 24.07 @xmath83 0.03 & 17.0 @xmath83 0.1 & 1.14 & 24.86 & 24.07 & 17.0 + c4 & 2 & 509.26 & 90.23 & 25.22 @xmath83 0.04 & 24.16 @xmath83 0.03 & 21.4 @xmath83 0.4 & 1.00 & 25.50 & 24.26 & 20.5 + c5 & 2 & 405.30 & 604.14 & 25.15 @xmath83 0.03 & 24.33 @xmath83 0.03 & 16.0 @xmath83 0.3 & 0.68 & 25.27 & 24.40 & 14.0 + c6 & 2 & 431.21 & 554.94 & 25.59 @xmath83 0.04 & 24.35 @xmath83 0.04 & 15.2 @xmath83 0.3 & 0.74 & 25.68 & 24.42 & 13.9 + c7 & 2 & 447.29 & 566.43 & 25.55 @xmath83 0.04 & 24.66 @xmath83 0.05 & 10.4 @xmath83 0.1 & 0.79 & 25.61 & 24.62 & 10.1 + c8 & 3 & 627.67 & 365.75 & 25.58 @xmath83 0.04 & 24.65 @xmath83 0.05 & 11.7 @xmath83 0.2 & 0.60 & 25.71 & 24.80 & 11.2 + c9 & 3 & 276.75 & 618.19 & 25.10 @xmath83 0.03 & 24.24 @xmath83 0.03 & 15.2 @xmath83 0.3 & 0.95 & 25.12 & 24.36 & 15.2 + c10 & 3 & 535.47 & 636.44 & 25.13 @xmath83 0.03 & 24.20 @xmath83 0.04 & 17.3 @xmath83 0.3 & 0.81 & 25.26 & 24.42 & 17.6 + c11 & 3 & 566.89 & 659.25 & 24.61 @xmath83 0.03 & 23.53 @xmath83 0.02 & 31.4 @xmath83 0.5 & 0.82 & 24.59 & 23.64 & 30.7 + c12 & 3 & 597.51 & 717.16 & 24.57 @xmath83 0.02 & 23.75 @xmath83 0.03 & 18.2 @xmath83 0.2 & 0.93 & 24.60 & 23.85 & 17.6 + c13 & 3 & 780.74 & 236.12 & 25.27 @xmath83 0.04 & 24.44 @xmath83 0.04 & 12.3 @xmath83 0.3 & 0.97 & 25.33 & 24.57 & 12.4 + c14 & 3 & 495.77 & 645.99 & 25.26 @xmath83 0.03 & 24.19 @xmath83 0.03 & 13.6 @xmath83 0.2 & 0.74 & 25.37 & 24.32 & 13.7 + c15 & 3 & 599.59 & 740.72 & 25.44 @xmath83 0.04 & 24.59 @xmath83 0.06 & 13.7 @xmath83 1.3 & 0.82 & 25.57 & 24.73 & 13.0 + + 1 & 25.07@xmath830.10 & 25.26@xmath830.11 & 25.12@xmath830.10 & 24.97@xmath830.07 & 25.27@xmath830.09 + 2 & 24.11@xmath830.05 & 24.99@xmath830.10 & 25.27@xmath830.12 & 24.99@xmath830.09 & 24.80@xmath830.08 + 3 & 24.06@xmath830.04 & 24.10@xmath830.08 & 25.41@xmath830.11 & 25.07@xmath830.10 & 24.73@xmath830.06 + 4 & 24.33@xmath830.06 & 24.46@xmath830.08 & 25.30@xmath830.10 & 25.13@xmath830.09 & 25.15@xmath830.08 + 5 & 24.52@xmath830.06 & 24.58@xmath830.06 & 24.31@xmath830.06 & 25.56@xmath830.12 & 25.27@xmath830.11 + 6 & 24.70@xmath830.08 & 24.82@xmath830.08 & 24.47@xmath830.12 & 26.02@xmath830.18 & 25.60@xmath830.12 + 7 & 25.09@xmath830.09 & 25.41@xmath830.11 & 24.72@xmath830.08 & 25.77@xmath830.12 & 25.20@xmath830.11 + 8 & & 24.72@xmath830.09 & 25.40@xmath830.12 & 24.75@xmath830.07 & + 9 & 24.34@xmath830.09 & 24.55@xmath830.06 & 24.25@xmath830.08 & 25.44@xmath830.12 & 25.40@xmath830.10 + 10 & 24.76@xmath830.08 & 25.22@xmath830.10 & 24.91@xmath830.09 & 25.99@xmath830.15 & 25.09@xmath830.08 + 11 & 25.24@xmath830.09 & 24.38@xmath830.06 & 25.20@xmath830.11 & 24.69@xmath830.07 & 25.30@xmath830.10 + + 1 & 24.25@xmath830.13 & 24.15@xmath830.08 & 23.95@xmath830.09 & 23.95@xmath830.08 & 24.34@xmath830.09 + 2 & 23.75@xmath830.07 & 23.95@xmath830.08 & 24.16@xmath830.10 & 24.05@xmath830.09 & 24.34@xmath830.09 + 3 & 23.48@xmath830.08 & 23.63@xmath830.08 & 24.24@xmath830.09 & 23.95@xmath830.08 & 24.06@xmath830.10 + 4 & 23.74@xmath830.09 & 23.50@xmath830.10 & 24.38@xmath830.12 & 24.10@xmath830.09 & 24.32@xmath830.10 + 5 & 23.73@xmath830.08 & 23.70@xmath830.08 & 23.80@xmath830.07 & 24.27@xmath830.08 & 24.39@xmath830.11 + 6 & 23.83@xmath830.09 & 23.87@xmath830.08 & 24.06@xmath830.12 & 24.61@xmath830.13 & 24.49@xmath830.10 + 7 & 23.97@xmath830.09 & 24.07@xmath830.08 & 23.96@xmath830.09 & 24.77@xmath830.12 & 24.40@xmath830.10 + 8 & 24.20@xmath830.09 & 23.82@xmath830.07 & 24.42@xmath830.12 & 23.92@xmath830.07 & 24.05@xmath830.08 + 9 & 23.52@xmath830.06 & 23.79@xmath830.08 & 23.67@xmath830.07 & 24.07@xmath830.08 & 24.75@xmath830.12 + 10 & 23.90@xmath830.07 & 24.14@xmath830.09 & 23.94@xmath830.08 & 24.65@xmath830.11 & 24.23@xmath830.08 + 11 & 24.09@xmath830.08 & 23.48@xmath830.08 & 24.50@xmath830.12 & 23.91@xmath830.06 & 24.39@xmath830.13 + lccccc + 1 & 25.32@xmath830.10 & 25.56@xmath830.14 & 25.57@xmath830.13 & 24.77@xmath830.08 & 24.64@xmath830.06 + 2 & 25.28@xmath830.11 & 25.87@xmath830.15 & 25.51@xmath830.11 & 24.84@xmath830.08 & 24.78@xmath830.08 + 3 & 25.35@xmath830.11 & 25.62@xmath830.11 & 25.95@xmath830.17 & 25.07@xmath830.08 & 25.02@xmath830.08 + 4 & 25.79@xmath830.15 & 25.23@xmath830.09 & 25.77@xmath830.17 & 25.34@xmath830.11 & 25.07@xmath830.09 + 5 & 26.07@xmath830.18 & 25.25@xmath830.10 & 25.24@xmath830.10 & 25.60@xmath830.15 & 25.47@xmath830.15 + 6 & 25.91@xmath830.13 & 25.94@xmath830.17 & 25.37@xmath830.11 & 25.16@xmath830.10 & 25.43@xmath830.12 + 7 & 25.30@xmath830.09 & 25.60@xmath830.11 & 25.98@xmath830.19 & 24.63@xmath830.06 & 25.41@xmath830.13 + 8 & 25.67@xmath830.13 & 25.55@xmath830.12 & 25.33@xmath830.10 & 25.31@xmath830.09 & 24.92@xmath830.12 + 9 & 25.88@xmath830.16 & 25.46@xmath830.10 & 25.78@xmath830.14 & 25.33@xmath830.11 & 25.37@xmath830.10 + 10 & 25.30@xmath830.13 & 25.86@xmath830.17 & 25.37@xmath830.12 & 24.96@xmath830.09 & 25.50@xmath830.12 + 11 & 26.06@xmath830.16 & 25.37@xmath830.12 & 25.85@xmath830.18 & 25.60@xmath830.14 & 25.18@xmath830.14 + + 1 & 24.41@xmath830.11 & 24.46@xmath830.14 & 24.51@xmath830.14 & 24.20@xmath830.11 & 24.05@xmath830.10 + 2 & 24.28@xmath830.12 & 24.84@xmath830.18 & 24.69@xmath830.17 & 23.81@xmath830.09 & 23.90@xmath830.08 + 3 & 24.20@xmath830.13 & 24.89@xmath830.15 & 24.82@xmath830.18 & 24.18@xmath830.10 & 24.04@xmath830.10 + 4 & 24.36@xmath830.13 & 24.53@xmath830.13 & 24.61@xmath830.15 & 24.24@xmath830.12 & 24.07@xmath830.10 + 5 & 24.24@xmath830.12 & 24.55@xmath830.11 & 24.60@xmath830.15 & 24.70@xmath830.18 & 24.30@xmath830.12 + 6 & 24.63@xmath830.15 & 24.82@xmath830.18 & 24.42@xmath830.12 & 24.49@xmath830.13 & 24.55@xmath830.14 + 7 & 24.14@xmath830.10 & 24.55@xmath830.12 & 25.01@xmath830.23 & 24.13@xmath830.11 & 24.46@xmath830.14 + 8 & 24.20@xmath830.11 & 24.78@xmath830.16 & 24.31@xmath830.13 & 24.13@xmath830.09 & 23.94@xmath830.11 + 9 & 24.81@xmath830.15 & 24.46@xmath830.14 & 25.13@xmath830.22 & 24.59@xmath830.13 & 24.26@xmath830.11 + 10 & 24.39@xmath830.12 & 24.73@xmath830.20 & 24.59@xmath830.17 & 24.15@xmath830.10 & 24.61@xmath830.17 + 11 & 24.36@xmath830.11 & 24.78@xmath830.21 & 24.73@xmath830.16 & 24.37@xmath830.13 & 24.29@xmath830.14 + lccccc + 1 & 24.83@xmath830.14 & 24.89@xmath830.11 & 24.82@xmath830.10 & 25.50@xmath830.16 & 25.16@xmath830.11 + 2 & 24.86@xmath830.08 & 24.87@xmath830.08 & 25.10@xmath830.09 & 25.11@xmath830.09 & 25.16@xmath830.10 + 3 & 24.95@xmath830.09 & 24.95@xmath830.08 & 25.32@xmath830.12 & 24.76@xmath830.09 & 25.29@xmath830.11 + 4 & 24.87@xmath830.09 & 24.95@xmath830.08 & 25.80@xmath830.14 & 25.00@xmath830.08 & 25.69@xmath830.17 + 5 & 25.01@xmath830.10 & 24.04@xmath830.06 & 25.52@xmath830.12 & 25.27@xmath830.12 & 26.04@xmath830.24 + 6 & 24.47@xmath830.08 & 24.37@xmath830.09 & 24.80@xmath830.08 & 25.43@xmath830.12 & 25.49@xmath830.14 + 7 & 24.02@xmath830.08 & 24.57@xmath830.08 & 25.35@xmath830.10 & 25.52@xmath830.10 & 25.24@xmath830.11 + 8 & 24.17@xmath830.08 & 24.94@xmath830.09 & 25.77@xmath830.15 & 25.32@xmath830.10 & 25.90@xmath830.17 + 9 & 24.54@xmath830.06 & 23.89@xmath830.05 & 24.97@xmath830.08 & 25.70@xmath830.14 & 25.37@xmath830.13 + 10 & 24.78@xmath830.07 & 24.63@xmath830.08 & 25.88@xmath830.16 & 25.02@xmath830.08 & 25.79@xmath830.15 + 11 & 24.72@xmath830.08 & 24.88@xmath830.11 & 25.33@xmath830.12 & 25.68@xmath830.12 & 25.17@xmath830.11 + + 1 & 23.71@xmath830.09 & 24.03@xmath830.11 & 24.47@xmath830.17 & 24.30@xmath830.12 & 24.25@xmath830.14 + 2 & 23.53@xmath830.09 & 23.96@xmath830.10 & 24.33@xmath830.13 & 24.16@xmath830.10 & 24.54@xmath830.16 + 3 & 23.79@xmath830.07 & 24.10@xmath830.12 & 24.64@xmath830.16 & 23.95@xmath830.10 & 24.35@xmath830.13 + 4 & 23.74@xmath830.07 & 23.57@xmath830.09 & 24.72@xmath830.15 & 24.00@xmath830.12 & 24.75@xmath830.22 + 5 & 23.77@xmath830.07 & 23.36@xmath830.07 & 24.56@xmath830.14 & 24.36@xmath830.14 & 25.03@xmath830.26 + 6 & 23.44@xmath830.06 & 23.58@xmath830.08 & 24.16@xmath830.11 & 24.24@xmath830.12 & 24.98@xmath830.25 + 7 & 23.33@xmath830.09 & 23.77@xmath830.08 & 24.12@xmath830.13 & 24.18@xmath830.11 & 24.52@xmath830.17 + 8 & 23.22@xmath830.05 & 23.97@xmath830.11 & 24.77@xmath830.17 & 24.12@xmath830.11 & + 9 & 23.37@xmath830.06 & 23.39@xmath830.07 & 24.46@xmath830.16 & 24.49@xmath830.14 & 24.41@xmath830.13 + 10 & 23.59@xmath830.07 & 23.94@xmath830.12 & 24.54@xmath830.17 & 24.18@xmath830.10 & + 11 & 23.60@xmath830.06 & 23.86@xmath830.10 & 24.29@xmath830.12 & 24.23@xmath830.13 & 24.43@xmath830.17 + lclll & all & 29.90 @xmath83 0.07 & 29.69 @xmath83 0.05 & * 29.40 @xmath83 0.06 * + & chip 2 & 29.92 @xmath83 0.13 & 29.72 @xmath83 0.085 & 29.43 @xmath83 0.10 + & chip 3 & 29.87 @xmath83 0.10 & 29.67 @xmath83 0.06 & 29.38 @xmath83 0.07 + * allframe * & all & 29.99 @xmath83 0.075 & 29.80 @xmath83 0.05 & * 29.53 @xmath83 0.07 * + & chip 2 & 30.03 @xmath83 0.14 & 29.77 @xmath83 0.09 & 29.40 @xmath83 0.10 + & chip 3 & 29.97 @xmath83 0.08 & 29.84 @xmath83 0.07 & 29.65 @xmath83 0.09 + * corrected allframe * & all & 29.99 @xmath83 0.075 & 29.77 @xmath83 0.05 & * 29.44 @xmath83 0.06 * + & chip 2 & 30.03 @xmath83 0.14 & 29.77 @xmath83 0.09 & 29.40 @xmath83 0.10 + & chip 3 & 29.97 @xmath83 0.08 & 29.77 @xmath83 0.07 & 29.48 @xmath83 0.09 + lll s@xmath87 & systematic errors in lmc p - l calibration & + & _ a . lmc true modulus _ & @xmath830.10 + & _ b . lmc p - l zero point _ & @xmath830.02 + & a and b added in quadrature & @xmath830.10 + + s@xmath88 & systematic errors in wfpc2 zero points & @xmath89 + + s@xmath90 & average metallicity correction & @xmath830.08 + + s@xmath91 & systematic errors unique to ngc 4258 photometry & @xmath830.05 + + s@xmath92 & dependence of ngc 4258 distance modulus on magnitude averaging method & @xmath830.04 + + r@xmath87 & random error in the ngc 4258 extinction - corrected distance modulus & + & _ c . ngc 4258 p - l fit ( v ) _ & @xmath830.07 + & _ d . ngc 4258 p - l fit ( i ) _ & @xmath830.05 + & c and d partially correlated & @xmath830.06 + + r@xmath93 & errors only affecting this determination & * @xmath830.09 * + & ( s@xmath91 , s@xmath92 , and r@xmath87 added in quadrature ) & + s@xmath93 & systematic errors in key project techniques & * @xmath830.15 * + & ( s@xmath87 , s@xmath88 , and s@xmath90 added in quadrature ) & +

let be a real smooth projective 3-fold fibred by rational curves . j. kollr proved that , if is orientable , then a connected component of is essentially either a seifert fibred manifold or a connected sum of lens spaces . our main theorem , answering in the affirmative three questions of kollr , gives sharp estimates on the number and the multiplicities of the seifert fibres and on the number and the torsions of the lens spaces when is a geometrically rational surface . when is seifert fibred over a base orbifold , our result generalizes comessatti s theorem on smooth real rational surfaces : can not be simultaneously orientable and of hyperbolic type . we show as a surprise that , unlike in comessatti s theorem , there are examples where is non orientable , of hyperbolic type , and is minimal . the technique we use is to construct seifert fibrations as projectivized tangent bundles of du val surfaces . _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ msc 2000 : _ 14p25 , 14m20 , 14j26 . _ keywords : _ del pezzo surface , rationally connected algebraic variety , seifert manifold , du val surface _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _

0803.2074

given a smooth real projective variety @xmath5 of dimension @xmath6 , we consider the topology of a connected component @xmath2 of the set @xmath1 of its real points . john nash proved in @xcite that any compact connected differentiable manifold @xmath2 is obtained in this way , and went over to ask whether the same would hold if one assumes @xmath5 to be geometrically rational . however , when @xmath5 is a surface of negative kodaira dimension , one is able , after the work of comessatti @xcite for geometrically rational surfaces , to deduce drastical restrictions for the topology of @xmath2 . namely , if @xmath2 is orientable , then it is diffeomorphic to a sphere or to a torus : in other words , @xmath2 can not be simultaneously oriented and of hyperbolic type . in this note , we make a step towards a complete classification of the topological types for @xmath2 when @xmath5 is a rationally connected 3-fold fibred by rational curves ( this is one of the higher dimensional analogues of comessatti s theorem ) . this study was initiated by jnos kollr , in the third paper @xcite of a ground - breaking series of articles applying the minimal model program to the study of the topology of real algebraic 3-folds . kollr s philosophy is that a very important condition in order to obtain restrictions upon the topological type of @xmath1 is that @xmath5 has terminal singularities and @xmath7 is cartier along @xmath1 . kollr proved in particular that if @xmath5 is a smooth 3-fold fibred by rational curves ( in particular , @xmath5 has negative kodaira dimension ) and such that @xmath1 is orientable , then a connected component @xmath2 of @xmath1 is essentially a seifert fibred 3-manifold or the connected sum of a finite number of lens spaces . note that in @xcite it was shown that conversely all the above manifolds @xmath2 do occur for some smooth 3-fold @xmath5 fibred by rational curves . when @xmath5 belongs to the subclass of rationally connected 3-folds fibred by rational curves , kollr proved some additional restrictions upon @xmath2 and made three further conjectures . in our first note @xcite we proved two of the optimal estimates that kollr conjectured to hold . in the present note we prove the third estimate , which is the most important one since it allows us to conclude in particular that , if @xmath2 is a seifert fibred 3-manifold , then the base orbifold can not be simultaneously oriented and of hyperbolic type . let us now introduce our results in more detail . let @xmath2 be an oriented three dimensional compact connected topological manifold without boundary . take a decomposition @xmath8 with @xmath9 maximal and observe that this decomposition is unique by a theorem of milnor @xcite . we shall focus our attention on the case where @xmath10 is seifert fibred or a connected sum of lens spaces . we consider the integers @xmath11 and @xmath12 , @xmath13 defined as follows : 1 . if @xmath14 is a seifert fibration , @xmath15 denotes the number of multiple fibres of @xmath16 and @xmath17 denote the respective multiplicities ; 2 . if @xmath10 is a connected sum of lens spaces , @xmath15 denotes the number of lens spaces and @xmath18 , @xmath19 , the orders of the respective fundamental groups ( thus we have a decomposition @xmath20 for some @xmath21 relatively prime to @xmath22 ) . observe that when @xmath10 is a connected sum of lens spaces , the number @xmath15 and the numbers @xmath23 , @xmath24 are well defined ( again by milnor s theorem ) . in the case of a seifert fibred manifold @xmath10 , these integers may a priori depend upon the choice of a seifert fibration . three results of our two notes are summarized by the following . [ teo : multiple ] let @xmath25 be a real smooth projective 3-fold fibred by rational curves over a geometrically rational rationally connected and fibred by rational curves . ] surface @xmath3 . suppose that @xmath1 is orientable . then , for each connected component @xmath26 , @xmath27 and @xmath28 . furthermore , if @xmath10 is seifert fibred over @xmath29 , then @xmath30 . this theorem answers , as we already said , some questions posed by kollr , see ( * ? ? ? * remark 1.2 ( 1,2,3 ) ) . in the first note , we proved the estimate @xmath31 and we showed that @xmath30 if @xmath10 is seifert fibred over the torus . thus theorem [ teo : multiple ] follows from ( * ? ? ? * corollary 0.2 , and theorem 0.3 ) and from theorem [ teo : main ] of the present paper using results of @xcite as in @xcite . the present note is mainly devoted to the proof of the inequality @xmath28 , see lemma [ lem : question2 ] . the proof of this inequality goes as follows : let @xmath32 be a real smooth projective 3-fold fibred by rational curves over a geometrically rational surface @xmath3 . using the same arguments as in ( * ? ? ? 3 ) , we reduce the proof of the estimate for the integers @xmath33 to an inequality depending on the indices of certain singular points of a real component @xmath34 of the topological normalization of @xmath35 ( see definition [ def : norm ] ) . in this process , the number @xmath36 can be made to correspond to the number of real singular points on @xmath34 which are of type @xmath37 , and globally separating when @xmath38 is odd ; each number @xmath39 corresponds to the index @xmath40 of the singularity @xmath41 of @xmath34 . the main part of the paper is devoted to the proof of the following . [ teo : main ] let @xmath3 be a projective surface defined over @xmath42 . suppose that @xmath3 is geometrically rational with du val singularities . then a connected component @xmath34 of the topological normalization @xmath43 contains at most 4 singular points @xmath44 of type @xmath41 which are globally separating for @xmath40 odd . furthermore , their indices satisfy @xmath45 let us now give an interpretation of the above results in terms of geometric topology ( see e.g. @xcite for the basic definitions and classical results ) . suppose that @xmath10 admits a seifert fibration with base orbifold @xmath4 . from our main theorem [ teo : multiple ] we infer that , if the underlying manifold @xmath46 is orientable , then the euler characteristic of the compact 2-dimensional orbifold @xmath4 is nonnegative ( see proposition [ proporbi ] ) . thus , by the uniformization theorem for compact 2-dimensional orbifolds , @xmath4 admits a spherical structure or an euclidean structure . in general , a 3-manifold @xmath2 does not possess a geometric structure , but , if it does , then the geometry involved is unique . moreover , it turns out that every seifert fibred manifold admits a geometric structure . the geometry of @xmath2 is modeled on one of the six following models ( see @xcite for a detailed description of each geometry ) : @xmath47 where @xmath48 is the @xmath49-dimensional euclidean space and @xmath50 is the hyperbolic plane . the six above geometries are called the seifert geometries . the appropriate geometry for a seifert fibration is determined by the euler characteristic of the base orbifold and by the euler number of the seifert bundle ( * ? ? ? * table 4.1 ) . let @xmath5 be a real projective 3-fold fibred by rational curves and such that @xmath1 is orientable , let @xmath51 be a connected component and let @xmath10 be the manifold defined as above . suppose moreover that @xmath10 possesses a geometric structure . by theorem ( 1.1 ) , the geometry of @xmath10 is one of the six seifert geometries . conversely , by @xcite , any orientable three dimensional manifold endowed with any seifert geometry is diffeomorphic to a real component of a real projective 3-fold fibred by rational curves . but , when @xmath5 is rationally connected , the following corollary of our main theorem gives further restrictions . [ cor : orbi ] let @xmath5 be a real smooth projective rationally connected 3-fold fibred by rational curves . suppose that @xmath1 is orientable and let @xmath2 be a connected component of @xmath1 . then neither @xmath2 nor @xmath10 can be endowed with a @xmath52 structure or with a @xmath53 structure whose base orbifold @xmath4 is orientable . observe moreover that in @xcite all compact 3-manifolds with @xmath54 or @xmath48 geometry , and some manifolds with @xmath55 geometry , are realized as a real component of a real smooth projective rationally connected 3-fold fibred by rational curves . there remains of course the question about what happens when @xmath2 is seifert fibred over a nonorientable orbifold @xmath4 : is the orbifold still not of hyperbolic type ? in the last section we show that the answer to this question is negative . we produce indeed an example of a smooth 3-fold @xmath5 , fibred by rational curves over a du val del pezzo surface @xmath3 , where @xmath1 is orientable , and contains a connected component which is seifert fibred over a nonorientable base orbifold of hyperbolic type . the striking fact is here that @xmath3 is a real minimal surface : this contrasts with comessatti s theorem : since indeed a real minimal nonsingular geometrically rational surface can not have an component which is of hyperbolic type . [ hyperbolic ] there exists a minimal real du val del pezzo surface @xmath3 of degree @xmath56 having exactly two singular points , of type @xmath57 , and such that the real part @xmath35 has a connected component containing the two singular points and which is homeomorphic to a real projective plane . let @xmath58 be the projectivized tangent bundle of @xmath3 : then @xmath58 has terminal singularities , @xmath59 is contained in the smooth locus of @xmath58 , in particular if @xmath5 is obtained resolving the singular points of @xmath58 , then @xmath60 . moreover @xmath61 is orientable and contains a connected component @xmath2 which is seifert fibred over a non orientable orbifold of hyperbolic type ( the real projective plane with two points of multiplicity 3 ) . briefly , now , the contents of the paper . sections [ sec : recall ] and [ sec : sevenconfig ] are devoted to the reduction of the proof of the main theorem to the assertion of non existence of seven configurations of singular points on a real component of a du val del pezzo surface of degree 1 . two main methods used here are borrowed from @xcite : namely , the generalization of brusotti s theorem to the effect that one can independently take any smoothing of the singularities of a du val del pezzo surface , and also the use of the plane model where the family of hyperplane sections of the quadric cone @xmath62 is represented by the family of parabolae in the plane with a fixed asymptotic direction . these methods combine with a delicate argument , suggested by e. brugalle , excluding the possibility of an intersection of @xmath62 with a cubic surface yielding an irreducible curve @xmath63 with four real cusps ( see [ 4cusps ] ) . section [ sec : config ] introduces the main tools used in the proof ( the topological classification of real smooth del pezzo surfaces of degree 1 , and the choice of the appropriate partial smoothings ) , and ends with the exclusion of two configurations via complicated although elementary topological considerations . section [ sec : euler ] uses a classification of critical points for the projection of @xmath63 and a precise table for the local contributions to the multiplicity of the discriminant and for the local contribution to the euler number in order to exclude two more cases . section [ sec : lasteuler ] proves theorem [ teo : main ] by excluding the three remaining cases by combining all the previous tools with an ad hoc analysis and with two new tools , namely : the use of the comessatti characteristic , relating the total betti number of the real part with the one of the complex part , and the calculation of the contributions of the singularities to the picard and to the various euler numbers . finally , section [ sec:3fold ] is devoted to the proof of lemma [ lem : question2 ] and in section [ orbi ] , after showing that the base orbifold can not be oriented and hyperbolic , we exhibit the example of a projectivized tangent bundle over a du val del pezzo surface for which a component @xmath2 is seifert fibred with base orbifold of hyperbolic type . in the course of this complicated construction we give a quite general method to construct seifert fibrations as projectivized tangent bundles of surfaces with @xmath64- singularities . we want to thank e. brugalle for pointing out the statement of lemma [ 4cusps ] and suggesting the main idea of the proof , and ingrid bauer for helping us to understand the configuration of lines on del pezzo surfaces of degree @xmath56 . using the results and notation of ( * ? ? ? * section 1 ) , we reduce the proof of theorem [ teo : main ] to the proof of a statement about singular del pezzo surfaces of degree 1 with small picard number @xmath65 . recall that a surface singularity which is a rational double point is also called a du val singularity and that a projective surface @xmath3 is called a _ surface if @xmath3 has only du val singularities . a surface singularity is _ of type @xmath37 _ if it is real analytically equivalent to @xmath66 and _ of type @xmath67 _ if it is equivalent to @xmath68 . the type @xmath69 is real analytically isomorphic to @xmath70 ; otherwise , singularities with different names are not isomorphic . we recall some definitions due to kollr ( see ( * ? ? ? * section 1 ) ) . [ def : norm ] let @xmath71 be a simplicial complex with only a finite number of points @xmath72 where @xmath71 is not a manifold . define the _ topological normalization _ @xmath73 as the unique proper continuous map such that @xmath74 is a homeomorphism over the set of points where @xmath71 is a manifold and @xmath75 is in one - to - one correspondence with the connected components of a good punctured neighborhood of @xmath76 in @xmath71 otherwise . observe that if @xmath71 is pure of dimension @xmath77 , then @xmath78 is a topological manifold ( since each point of @xmath78 has a neighbourhood which is a cone over @xmath79 ) . let @xmath3 be a real du val surface , and let @xmath80 be a singular point of type @xmath81 with @xmath38 odd . the topological normalization @xmath82 has two connected components locally near @xmath76 . we will say that @xmath76 is _ globally separating _ if these two local components lie on different connected components of @xmath82 and _ globally nonseparating _ otherwise . let @xmath83 let @xmath3 be a real du val surface , let @xmath84 be the topological normalization , and let @xmath85 be the connected components of @xmath86 . by ( * ? ? ? 9.7 ) , the unordered sequence of numbers @xmath87 is an invariant for extremal birational contractions of du val surfaces . we will now reduce the proof of theorem [ teo : main ] to the proof of the following . [ teo : dps1 ] let @xmath3 be a real du val del pezzo surface of degree 1 with @xmath88 . then @xmath89 , @xmath90 , and moreover for any @xmath91 such that @xmath92 contains @xmath93 where @xmath41 is globally separating for @xmath40 odd , we have : @xmath94 up to section [ sec:3fold ] , the sequel of this paper is devoted to the proof of theorem [ teo : dps1 ] . numerically , the following configurations of @xmath95 singularities are the only ones allowed by the inequality @xmath96 * @xmath97 and the configuration is @xmath98 , * @xmath99 and the configuration is * * @xmath100 , any @xmath38 , or * * @xmath101 , @xmath102 , or * * @xmath103 , * * @xmath104 , * @xmath105 . recall that a du val del pezzo surface @xmath3 is by definition a du val surface ( i.e. , a surface with only rational double points as singularities ) whose anticanonical divisor is ample , see ( * ? ? ? * section 2 ) . the anticanonical model of a del pezzo surface @xmath3 of degree 1 is a ramified double covering @xmath106 of a quadric cone @xmath107 whose branch locus is the union of the vertex of the cone with a curve @xmath63 not passing through the vertex and which is the complete intersection of the cone with a cubic surface . let @xmath3 be a real du val del pezzo surface of degree 1 and let @xmath108 be the singular surface obtained from @xmath3 by blowing up the pull - back by @xmath109 of the vertex of the cone ( which is a smooth point of @xmath3 ) . the surface @xmath108 is a ramified double covering of the hirzebruch surface @xmath110 whose branch curve is the union of the unique section of negative selfintersection , the section at infinity @xmath111 , and the trisection @xmath63 of the ruling @xmath112 , which is disjoint from @xmath111 . the composition @xmath113 is a real elliptic fibration . the different cases that we shall now consider are distinguished by the number of irreducible components of the trisection @xmath63 . notice that if all the singular points are of type @xmath114 , the conclusion of theorem [ teo : dps1 ] follows from ( * ? ? ? * proposition 2.1 ) . if @xmath63 has strictly more than 4 real singular points , all the possible cases are enumerated in ( * ? ? ? * section 2 ) , and an inspection of [ ibid . , figures 1 , 2 , 3 ] shows that for any connected component of the complement @xmath115 , the configuration is @xmath116 or @xmath117 . thus the inequality ( [ eq : main ] ) holds except possibly in the situation where two irreducible components of @xmath63 are tangent to the third one . it turns out that there is only one normal form for this situation , see figure [ fig:2tacnodes ] . indeed , the affine part of @xmath63 is a union of three parabolae and without loss of generality , these three parabolae are given by @xmath118 , @xmath119 and @xmath120 [ ibid . ] . we have @xmath121 , else @xmath63 has a triple point with an infinitely near triple point , contradicting the fact that @xmath3 has only du val singularities . furthermore , in order to get at least three real intersection points , @xmath122 has to be positive . up to reflection @xmath123 , this leads to one possibility . recalling that in this figure two components are connected at infinity if their boundaries have two unbounded arcs belonging to the same pair of parabolae , we see that none of the connected components of @xmath115 contains more than 3 singular points and at most two of them are tacnodes . thus ( [ eq : main ] ) holds also in this case . then @xmath124 where @xmath125 is a bisection of the ruling @xmath126 and @xmath127 is a section . the bisection @xmath125 has arithmetic genus one , hence it has at most one double point @xmath114 or @xmath128 and at most 4 intersection points with the section @xmath127 . if @xmath125 is non singular , we have @xmath129 or @xmath130 or only two singular points . in each case we get an allowed configuration . assume @xmath125 is singular : if @xmath63 has 5 singular points , we are done since either all singular points are of type @xmath114 , see ( * ? ? ? * figures 4 , 5 , 6 ) , or we are in the situation depicted in ( * ? ? ? * figure 7 ) and then the @xmath57 is on a component with only two other singularities , of type @xmath114 . if @xmath63 has 4 singular points , the possibilities are @xmath131 , or @xmath132 . if @xmath63 has 3 singular points , the possibilities are @xmath133 , or @xmath134 , or @xmath135 . thus if @xmath63 has two irreducible components , we get the conclusion of theorem [ teo : dps1 ] unless the configuration of singular points is @xmath136 , or @xmath137 , or @xmath138 . if the trisection is irreducible , then it has at most 4 singular points , since @xmath139 has genus 4 . [ 4cusps ] the real curve @xmath63 can not have 4 real cusps . suppose that @xmath63 is irreducible with 4 real cusps . choose three of them . let @xmath140 be a section of the ruling @xmath126 corresponding to a plane section of @xmath62 passing through these three points ; for an appropriate choice of the plane model of @xmath62 ( see @xcite , beginning of section 2 ) we may assume @xmath140 to be the horizontal @xmath76-axis @xmath141 in the plane since the intersection number @xmath142 , we get that @xmath140 intersects @xmath63 exactly at the three chosen cusps , and transversally . this means that , w.l.o.g . , @xmath63 lies in the upper halfplane : in fact , since @xmath63 is rational and irreducible , then its real part @xmath143 is homeomorphic to @xmath79 , in particular it is connected . observe moreover that none of the cusps is tangent to a fibre , since each cusp gives a contribution at least 3 to the local multiplicity of the discriminant of @xmath63 , and this contribution becomes 4 if the cusp is tangent to the fibre : and the order of the discriminant is 12 . in fact , we get more from this calculation : the projection @xmath126 has no further critical points on @xmath63 . it follows that the projective line with coordinate @xmath76 is divided into 4 open intervals , such that the cardinality of the fibre of @xmath144 varies alternatingly from @xmath49 to @xmath56 . on the intervals where we have 3 counterimages , it makes sense to talk about first , second and third branch ( ordered according increasing value of the @xmath145-coordinate ) , on each interval it makes sense to talk about the highest and the lowest branch . whenever one moves on @xmath146 and goes across a cusp lying on the @xmath76-axis , the highest branch continues to be the highest branch . since three of the cusps lie on the x - axis , we may assume that the fourth cusp is located at @xmath147 , and the three cusps with @xmath148 occur for @xmath149 where @xmath150 . then the highest branch over the interval @xmath151 remains the highest branch on the whole real line by virtue of the previous remark . by compactness of @xmath143 we get a connected component of @xmath143 mapping to @xmath146 homeomorphically , contradicting our previous assertion about the cardinalities of the fibres . thus , if @xmath63 is irreducible , we observe that @xmath63 has arithmetic genus @xmath152 , and nonnegative geometric genus : hence the ` number of double points ' @xmath153 is at most @xmath152 . but each point of type @xmath64 contributes exactly @xmath154 $ ] double points . therefore an elementary calculation shows that we get the conclusion of theorem [ teo : dps1 ] unless the configuration of singular points is one of the following : @xmath155 , @xmath156 , or @xmath157 , or @xmath158 , or @xmath159 . we are going now to exclude the first case by an argument similar to the one of lemma [ 4cusps ] , even if it could also be treated by the same methods used in section [ sec : euler ] . [ 2cusps ] the real curve @xmath63 can not have 2 real @xmath128 singularities and an @xmath160 singularity . we already know that @xmath63 is irreducible and we argue as in lemma [ 4cusps ] , assuming that the three singular points lie on the horizontal @xmath76-axis @xmath161 in the plane and that , since @xmath143 is homeomorphic to @xmath79 , @xmath63 and @xmath140 intersect exactly at the three chosen points , and transversally , hence @xmath143 lies in the upper halfplane . if none of the cusps is tangent to a fibre , since each cusp @xmath162 gives a contribution @xmath163 to the local multiplicity of the discriminant of @xmath63 , and the order of the discriminant is 12 , there is exactly another critical point for the restriction of the projection @xmath126 to @xmath63 , and the same argument as in lemma [ 4cusps ] provides the same contradiction . there remains the case where exactly one cusp is vertical , and there are no further critical points . it follows that the projective line with coordinate @xmath76 is divided into 3 open intervals , and the cardinality of the fibre of @xmath144 must be equal to @xmath56 on the two intervals neighbouring the vertical cusp . at the two other cusps the highest branch remains the highest , and we get the usual contradiction ( since over the third interval we have three branches ) . in any case , regardless of the difference between @xmath164 and @xmath165 , we have reduced the problem to the exclusion of 7 configurations . for any of these configurations , we can suppose that all singular points are of type @xmath164 with @xmath41 globally separating for @xmath40 is odd . indeed , if one of the point is not of this type , the sum @xmath166 restricted to the remaining points if less than or equal to 2 . summarizing , we get seven remaining configurations to be excluded : 1 . [ 2a3+a2 ] @xmath167 ( section [ sec : lasteuler ] ) 2 . [ a3 + 2a2 ] @xmath168 ( section [ sec : euler ] ) 3 . [ a3+a2 + 2a1 ] @xmath169 ( section [ sec : config ] ) 4 . [ a3 + 3a1 ] @xmath170 ( section [ sec : config ] ) 5 . [ 3a2+a1 ] @xmath171 ( section [ sec : euler ] ) 6 . [ 2a2 + 2a1 ] @xmath172 ( section [ sec : lasteuler ] ) 7 . [ a2 + 3a1 ] @xmath173 ( section [ sec : lasteuler ] ) we recall that our problem consists in giving an estimate concerning the configurations of certain singular points lying on a component of the topological normalization of a real du val del pezzo surface @xmath3 . for this purpose , we want to understand as much as possible the topology of @xmath174 , and we do this by taking a global smoothing of @xmath3 , and then using the known topological classification of smooth real del pezzo surfaces of degree 1 . the best strategy is to choose a global smoothing realizing certain local smoothings of the singularities chosen a priori . that this can be done for all choices of the local smoothings holds true by a generalization of the theorem of brusotti which was proven in our preceding paper . 4.3)[teo : generalizedbrusotti ] let @xmath3 be a du val del pezzo surface . one can obtain , by a global small deformation of @xmath3 , all the possible local smoothings of the singular points of @xmath3 . [ sec : dps1 ] let @xmath3 be a real smooth del pezzo surfaces of degree 1 : then the real part @xmath35 is diffeomorphic to one of the surfaces in the following list : * @xmath175 ; * @xmath176 ; * @xmath177 ; * @xmath178 . here @xmath179 denotes the connected sum of @xmath180 copies of the real projective plane , @xmath181 denotes the klein bottle and @xmath182 denotes the disjoint union of @xmath126 copies of the @xmath77-sphere . it is the well - known classification of real smooth del pezzo surfaces , see e.g. @xcite . consider a real singular point of a surface @xmath3 of type @xmath37 , of local equation @xmath183 where @xmath184 vanishes at the origin . then for each case @xmath185 , there exists local smoothings @xmath186 with equation @xmath187 , such that @xmath188 is represented by one of the figures [ fig : a1 ] , [ fig : a2 ] , or [ fig : a3 ] . = 2.5 cm = 2.5 cm = 2.5 cm let @xmath3 be a real du val del pezzo surface of degree 1 . recall that @xmath108 denotes the singular elliptic surface obtained from @xmath3 by blowing up a smooth point . we denote by @xmath189 the topological normalization of the real part and we assume that there is a connected component @xmath190 of @xmath191 whose image by @xmath74 contains at least three singular points of @xmath108 . furthermore , we assume that the singular points are only of type @xmath164 , with @xmath37 globally separating for @xmath38 odd . let @xmath192 , @xmath193 be the other components of @xmath191 such that @xmath194 and @xmath195 intersect ( in some singular point of @xmath196 ) . any singular point @xmath37 of @xmath190 with @xmath38 odd is globally separating , while the ones with @xmath38 even are not , thus in particular the number @xmath197 satisfies @xmath198 . let s denote by @xmath199 the connected component of @xmath191 which meets the section at infinity , i.e. , @xmath200 . in the proof of the main theorem we will often use the distinction between the cases @xmath201 and @xmath202 for some @xmath203 . [ lem : klein ] the component @xmath204 of the topological normalization is a klein bottle unless the elliptic fibration has two white returns ( see table [ tab : euler ] ) . in the latter situation , @xmath191 contains at most another component which is then a sphere . if the fibre of the double covering @xmath205 over a real point @xmath206 contains a real point , we shall say that @xmath206 belongs to the region of positivity , which we denote by @xmath207 . the section @xmath111 is part of the branch locus and is bilateral in @xmath110 . consider @xmath208 which is an oriented @xmath209-bundle over @xmath210 , and indeed homeomorphic to @xmath211 . hence we take corresponding coordinates @xmath212 for the points of @xmath213 . we may assume without loss of generality that @xmath214 for @xmath215 . consider now the function @xmath216 , @xmath217 . therefore , if @xmath218 is a continuous function on @xmath219 , then we have @xmath220 . for further use , we notice that : [ lem : monoton ] let @xmath221 be a connected component of the topological normalization of @xmath196 , and consider @xmath76 as a function on the boundary of @xmath34 : then the number of changes of monotonicity of @xmath76 is even . let @xmath222 be the discriminant of the elliptic fibre over @xmath76 ( i.e. , the discriminant of the degree three polynomial in @xmath145 whose zero set is the trisection ) . in view of table [ tab : euler ] ( go two pages ahead ) , the only root of @xmath223 which can break the continuity of @xmath218 corresponds to a white return ( @xmath224 ) . if the lower part of the white return branch continues and meets as first critical point of @xmath225 a point which contributes one change of monotonicity of @xmath76 ( that is , a black return , or a black node , or a tangent node , or a transversal cusp or tacnode ) , then we can topologically deform to the case where @xmath218 is continuous . the flex and the tangent cusp are clearly irrelevant and if the first met critical point is a white node , we can perform a cut smoothing and pass to the next critical point . the only obstacle is then the case when we meet another white return singularity on the branch curve . in this case , one sees easily that there is another component @xmath226 of @xmath143 disjoint from the component @xmath227 containing the white return branches , hence @xmath228 ( take the 4 cycles respective inverse images of the section at infinity @xmath229 , of @xmath226 , and of two segments , one joining @xmath227 with @xmath229 , the other joining @xmath227 with @xmath226 ) . recall that the topological normalization @xmath86 of the real del pezzo surface @xmath3 can be realised by a global smoothing of @xmath3 , see ( * ? ? ? * lemma 4.4 and theorem 4.3 ) . thus the component of @xmath86 corresponding to @xmath199 has @xmath230 and , by [ sec : dps1 ] , either we have @xmath231 or @xmath232 for some @xmath233 . more generally , by [ sec : dps1 ] , the real part of any global smoothing @xmath234 of @xmath108 , including the case when @xmath235 , is diffeomorphic to one of the surfaces in the following list , * @xmath236 ; * @xmath237 ; * @xmath238 ; * @xmath239 . [ topotype ] let @xmath3 be a real du val del pezzo surface of degree one , and let @xmath108 be the corresponding rational elliptic surface . let @xmath234 be a global smoothing of @xmath108 . then we have the following estimates for the betti numbers @xmath240 * @xmath241 . * @xmath242 . * in any case , @xmath243 . for each node of @xmath195 connecting @xmath195 with some @xmath194 , we choose the cut smoothing if this point is the only singular point on @xmath194 . otherwise , we choose the cylinder smoothing . we do the + sphere smoothing for the cusp . for the tacnode , connecting @xmath195 with some @xmath194 , we choose the cut+sphere smoothing if this point is the only singular point on @xmath194 or if we are in the last two cases in the next list . otherwise , we choose the cylinder smoothing . recalling that @xmath244 , we obtain the following inequalities for the betti numbers of @xmath245 . the different cases are distinguished by the number @xmath197 defined above . * @xmath169 * * @xmath246 ; * * @xmath247 , and @xmath248 ; * * @xmath249 , and @xmath250 . * @xmath170 * * @xmath246 ; * * @xmath247 , and @xmath248 ; * * ( cut+sphere smoothing for the tacnode ) @xmath247 , and @xmath248 . * * ( cut+sphere smoothing for the tacnode ) @xmath249 , and @xmath250 . in each case , these inequalities contradict lemma [ topotype ] . thus cases ( [ a3+a2 + 2a1 ] ) and ( [ a3 + 3a1 ] ) are excluded . recall that @xmath108 is a singular surface obtained from the singular degree 1 del pezzo surface @xmath3 by blowing up a smooth point . it is a ramified double covering of the hirzebruch surface @xmath110 whose branch locus is the union @xmath251 where @xmath63 is a trisection of the ruling @xmath112 . the composition @xmath252 is a real elliptic fibration and @xmath222 denotes the discriminant of the elliptic fibre over @xmath76 ( i.e. , the discriminant of the degree three polynomial in @xmath145 whose zero set is the trisection ) . table [ tab : euler ] gives a local topological description of the fibration over a neighbourhood of a real zero of @xmath253 , in terms of two basic numerical invariants , namely the multiplicity of the zero of @xmath253 , and the euler number of the real part of the singular fibre of the elliptic surface . the table considers only the singular points that we have to deal with , and introduces a name for each case , which will be used in the course of the forthcoming proofs . observe finally that , in drawing as black the region of positivity , we have used the convention introduced in lemma [ lem : klein ] . finally , a point of type @xmath254 is here a smooth point of @xmath63 which is a critical point for the restriction of @xmath126 to @xmath63 . ' '' '' type & fibre type & picture & multiplicity of @xmath253 & euler number @xmath255 + ' '' '' @xmath254 & black return & & @xmath56 & @xmath56 + ' '' '' @xmath254 & white return & & @xmath56 & @xmath256 + ' '' '' @xmath254 & flex & & @xmath77 & @xmath257 + ' '' '' @xmath69 & black node & & @xmath77 & @xmath56 + ' '' '' @xmath69 & white node & & @xmath77 & @xmath258 + ' '' '' @xmath69 & tangent node & & @xmath49 & @xmath257 + ' '' '' @xmath57 & transversal cusp & & @xmath49 & @xmath56 + ' '' '' @xmath57 & tangent cusp & & @xmath152 & @xmath257 + ' '' '' @xmath259 & tacnode & & @xmath152 & @xmath56 + we use now table [ tab : euler ] in order to proceed with our case by case exclusion . * @xmath168 , here @xmath260 , and @xmath261 meet in the tacnode @xmath262 . by doing the + sphere smoothing for each cusp and the cut+sphere smoothing for the tacnode , we get at least 5 connected components , hence a klein bottle and 4 spheres by [ topotype ] , and @xmath263 . we conclude that @xmath264 . + if @xmath265 ( thus @xmath266 and @xmath267 ) , the cusps are transversal . since the total multiplicity of @xmath253 is 12 , the fibration has at most two real @xmath254 singular fibers . but any white return stays on @xmath268 . then on the boundary of @xmath195 the monotonicity of the function @xmath76 has three changes , a contradiction . + if @xmath269 , then the component @xmath270 must contain a black return ( recall that the two cusps belong to @xmath195 ) . the existence of the tacnode on @xmath268 forces a white return the contributions to the multiplicity of @xmath253 are then @xmath271 which implies that the two cusps are transversal . the contributions to the euler characteristic of @xmath196 are then @xmath272 , a contradiction . + thus case ( [ a3 + 2a2 ] ) is excluded . * @xmath171 , here also @xmath260 and the same argument as above shows that @xmath273 and @xmath264 . + if @xmath265 , the three cusps are transversal . by lemma [ lem : monoton ] , if we had a white node , we would have a black return on @xmath195 , thus @xmath274 , a contradiction . if we did not have a white node , then @xmath275 , a contradiction again . + if @xmath269 , then the node on @xmath276 forces a black return . since @xmath277 is a sphere , the contributions to the multiplicity of @xmath253 impose that the fibration has at most one real @xmath254 singular fiber and then that all singular points are of transversal type . whence @xmath278 , a contradiction . + thus case ( [ 3a2+a1 ] ) is excluded . recall that @xmath108 is a singular rational elliptic surface obtained from the del pezzo surface @xmath3 by blowing up a smooth point ; and that a singular point @xmath37 is real analytically equivalent to @xmath279 . let us denote by @xmath280 the minimal resolution of singularities . in this paper , we define , for a real variety @xmath3 , @xmath281 to be the picard number of the surface @xmath3 _ over @xmath42_. it must not be confused with the picard number of the complexification @xmath282 of @xmath3 . we have always @xmath283 but , generally , @xmath284 . [ lem : rho ] suppose that the singularities of @xmath3 ( and then of @xmath108 ) are only of type @xmath37 . then , denoting by @xmath285 the number of singular points which are of type @xmath37 , we have : @xmath286 a local computation shows that @xmath287 and @xmath288 . [ lem : rho - euler ] let @xmath3 be a real du val del pezzo surface of degree 1 . suppose that @xmath88 . suppose moreover that the singularities are only of type @xmath37 , @xmath289 . then we get for the rational elliptic surface @xmath108 : @xmath290 denote by @xmath291 the comessatti characteristic of @xmath292 given by @xmath293 ( see ( * ? ? ? * chap . i ) , and recall that our betti numbers are taken with coefficients @xmath294 ) . the nonsingular rational elliptic surface @xmath292 has total betti number @xmath295 . moreover , for a nonsingular surface @xmath292 with @xmath296 and with @xmath297 simply connected , we have @xmath298 , see @xcite or @xcite . since however @xmath299 we get @xmath300 . by our hypothesis on the picard number of the singular del pezzo surface @xmath3 , we have @xmath301 , thus the formula follows from lemma [ lem : rho ] . in the first case , the branch curve has 2 irreducible components which are rational . indeed one of them is smooth rational and the other has genus 1 and one singular point . furthermore the two irreducible components intersect in a real point , thus the real part of the branch curve is connected . in the last two cases , the branch curve @xmath63 is irreducible , has genus 4 , and has 4 singular points . thus the curve is rational and its real part @xmath143 is connected . it follows that every connected component @xmath34 of the topological normalization has the property that @xmath302 hence @xmath303 is the number of connected components of the normalization . * @xmath167 , @xmath304 and @xmath305 or @xmath306 by lemma [ lem : rho - euler ] . + the total multiplicity of @xmath253 is 12 , thus there is at most one fibre @xmath254 and the contributions to the euler characteristic are @xmath307 , or @xmath308 , or @xmath309 when the cusp is tangent to a fibre . thus @xmath310 would be greater than or equal to 2 , a contradiction . thus case ( [ 2a3+a2 ] ) is excluded . * @xmath172 , @xmath311 and @xmath312 or @xmath257 . + here , the number of components of @xmath191 such that @xmath194 and @xmath195 belong to the same connected component of @xmath196 satisfies @xmath313 . + assume @xmath260 , and do the cylinder smoothing for the nodes , and the + sphere smoothing for the cusps . we obtain @xmath314 , and @xmath248 , a contradiction . + assume @xmath315 , then there are two cases : @xmath316 or @xmath317 . the topological normalization has @xmath318 components , hence @xmath319 . + assume @xmath265 , then any cusp is transversal and yields a @xmath320 contribution to the euler number . for the component @xmath277 , which is distinct from @xmath190 and from @xmath199 , we must have a black node , therefore on it there is also a black return . in order to get @xmath321 , there must be a white return , and then we should have a white node to make @xmath321 . but a white return is necessarily on @xmath268 , and its existence implies the existence of other critical points at @xmath322 , a contradiction . + assume @xmath269 , consider the two components not at @xmath322 , @xmath277 and @xmath323 . on them , a white node implies at least two black returns , while a black or tangent node implies at least one black return . since @xmath253 has degree 12 , there are exactly two black returns ( on each respective @xmath192 ) and two black nodes ( on each respective @xmath192 ) . at @xmath322 , there are as critical points only the singular points , and these are transversal , whence we get 2 transversal cusps , thus @xmath324 , a contradiction . + thus the configuration @xmath172 does not exist . * @xmath173 , @xmath325 and @xmath326 or @xmath56 . + in this case , we have @xmath327 . + if @xmath260 we have two components @xmath190 , @xmath277 and each node on @xmath195 connects with @xmath276 . we do 3 cylinder smoothings and the + sphere smoothing for the cusp . we obtain @xmath328 , and @xmath250 , a contradiction . + if @xmath315 we have three components @xmath190 , @xmath277 , @xmath323 and we let @xmath277 be the component such that there are two nodes on @xmath195 connecting with @xmath276 . we perform 2 cylinder smoothings at these nodes . for the remaining node connecting @xmath195 with @xmath329 , we choose the cut smoothing . we do the + sphere smoothing for the cusp . this gives @xmath330 , and @xmath331 , a contradiction . + assume @xmath332 , and take the normalization . there are 4 components , whence they are @xmath333 with @xmath334 . notice in particular that @xmath326 . + if @xmath269 , assume that we have a black node . for the corresponding @xmath335 component this requires a black return . for a white node , we need at least two black returns . for a tangent node , one needs one black return . however , the number of @xmath254 fibres is @xmath336 , thus there is no white node , and we have exactly 3 black returns . the contribution to the euler number is then @xmath337 . this is a contradiction which excludes the case @xmath269 . + if @xmath265 , the cusp is transversal hence it contributes 1 to the euler number . consider the two nodes not involving the component @xmath199 . necessarily they are black nodes since the other two types of nodes involve the component @xmath199 . these singularities each involve a black return as before . we get a contribution 5 to @xmath326 . hence the remaining node and return must contribute twice a @xmath338 . a @xmath338 contribution is white and involves the component @xmath199 . but because the white return gives for the boundary of @xmath339 some @xmath76 for which the degree is 3 , and others for which it is 1 , there must be another critical point on @xmath339 , a contradiction . + thus case ( [ a2 + 3a1 ] ) is excluded . @xmath340for theorem 1.3 we end this section with the first of all , the reduction from the case of a geometrically rational du val surface to the case of a du val del pezzo surface of degree 1 is done precisely as in @xcite , proposition 2.4 . and the subsequent proof of theorem 0.1 . the same argument given in the proof of cor.0.2 ( @xcite , end of the third section ) shows that it suffices to consider the singular points of type @xmath37 which are globally separating when locally separating . finally , by lemma 1.8 of @xcite , it remains only to check the case where @xmath3 is a real du val del pezzo surface of degree 1 with @xmath88 . then our assertion is exactly reduced to the main assertion of theorem [ teo : dps1 ] . this short section explains how theorem [ teo : main ] implies the following . [ lem : question2 ] let @xmath25 be a real smooth projective 3-fold fibred by rational curves over a geometrically rational surface @xmath3 . suppose that @xmath1 is orientable . then for each connected component @xmath26 , we have @xmath341 let @xmath25 be a real smooth projective 3-fold fibred by rational curves over a geometrically rational surface @xmath3 . suppose that @xmath1 is orientable . let @xmath26 be a connected component . kollr proved ( see also ( * ? ? ? * , and proof of cor . 0.2 ) ) , that there is \1 ) a pair of birational contractions @xmath342 , @xmath343 , where \2 ) @xmath58 is a real projective 3-fold @xmath58 with terminal singularities such that @xmath344 is cartier along @xmath345 , \3 ) @xmath108 is a du val surface \4 ) a rational curve fibration @xmath346 such that @xmath347 is @xmath348-ample and with \5 ) @xmath349 . let @xmath350 be the connected component of the topological normalization @xmath351 such that @xmath350 maps onto @xmath352 . the main property of this construction is that \6 ) @xmath353 . thanks to ( * ? ? ? * theorem 8.1 ) , and ( * ? ? ? * proof of cor . 0.2 , end of section 3 ) , there is a small perturbation @xmath354 of @xmath355 such that @xmath356 is a seifert fibration , and an injection from the set of multiple fibres of @xmath357 to the set of singular points of @xmath108 contained in @xmath358 which are of type @xmath359 and globally separating when locally separating . under this injection , the multiplicity of the seifert fibre equals @xmath360 if the singular point is of type @xmath361 . hence , the desired inequality follows from theorem [ teo : main ] . in this section we derive first some consequences from our main result on the components of the topological normalization of a geometrically rational du val surface . then we construct a real smooth algebraic 3-fold whose real part contains a connected component which is seifert fibred over the real projective plane , with two multiple fibres of multiplicity 3 . * corollary [ cor : orbi ] . * _ let @xmath5 be a real smooth projective rationally connected 3-fold fibred by rational curves . suppose that @xmath1 is orientable and let @xmath2 be a connected component of @xmath1 . then neither @xmath2 nor @xmath10 can be endowed with a @xmath52 structure or with a @xmath53 structure whose base orbifold @xmath4 is orientable . _ \3 ) if @xmath2 or @xmath10 can be endowed with a @xmath52 structure or with a @xmath53 structure , then @xmath2 is seifert fibred and , by the cited theorem of milnor , we have that @xmath10 is seifert fibred by the given rational curve fibration . now , the six geometries for seifert fibrations are distinguished by negativity , nullity or positivity of the euler characteristic @xmath362 of the base orbifold and by the vanishing or non vanishing of the euler number of the seifert bundle ( * ? ? ? * table 4.1 ) . in particular the @xmath52 and the @xmath53 geometry correspond exactly to the ` hyperbolic ' case , where @xmath362 is negative . [ proporbi ] let @xmath2 be as in corollary [ cor : orbi ] . suppose moreover that @xmath2 admits a seifert fibration with base orbifold @xmath4 such that @xmath363 is orientable . then the euler characteristic @xmath362 of the compact 2-dimensional orbifold @xmath4 is nonnegative . by ( * theorem 4.3 and lemma 4.4 ) , the topological normalization @xmath86 can be realized as the real part of a real perturbation @xmath364 of @xmath3 . thanks to comessatti s theorem , an orientable connected component of @xmath365 is a sphere or a torus . in the last case , the seifert fibration @xmath366 has no singular fibre and @xmath4 is a manifold , hence the euler characteristic of @xmath4 is zero . in the latter case , the euler characteristic of @xmath4 is positive . as announced in the introduction , we are going to construct a real smooth 3-fold @xmath5 , fibred by rational curves over a du val del pezzo surface @xmath3 , with the property that @xmath1 is connected and enjoys the following properties : our method of construction is based on a rather general procedure which produces seifert fibrations as projectivized tangent bundles of du val surfaces , so we start with some easy lemmas , the first one being well known . this 3-fold is simply obtained by glueing together the projectivized tangent bundle of the smooth part with the @xmath378 quotient @xmath379 of the projectivized tangent bundle of the affine plane via the action of the @xmath380-th roots of unity induced by the action on @xmath381 yielding the quotient @xmath382 . assume now that @xmath6 is even , so that each nontrivial group element has only two fixed points , namely , for @xmath393 , respectively for @xmath394 . at each point , passing to local coordinates , we see that we have a singularity of type @xmath395 , the quotient @xmath396 by the action where @xmath397 . this singularity is well known to be terminal ( see @xcite ) , and the zariski canonical divisor @xmath398 there is not cartier because the differential form @xmath399 is not invariant , being multiplied by @xmath400 ( only @xmath401 is cartier ) . [ ptb ] let @xmath6 be an even number and define @xmath402 to be the projectivized tangent bundle of a singularity of type @xmath403 , and define analogously @xmath404 for a singularity of type @xmath405 . @xmath406 has terminal isolated singularities and the real part @xmath407 is a pl - manifold of real dimension 3 , while the real part @xmath408 is contained in the smooth locus of @xmath404 . we treat first the @xmath403-case . we consider the real group scheme @xmath411 which acts on the affine plane @xmath412 by @xmath386 , whence its action on @xmath413 , @xmath414 is such that each nontrivial group element has only two fixed points , namely , the point where @xmath393 , respectively the one where @xmath394 . at each point , passing to local coordinates , we see that we have a singularity of type @xmath415 , the quotient @xmath416 by the action where @xmath397 . let us now observe that @xmath415 sits inside a galois sandwich @xmath417 where @xmath418 is the quotient morphism and the composition @xmath419 is given by @xmath420 ( the coordinates of @xmath418 are just a set of invariant monomials including @xmath421 ) . since @xmath422 induces a homeomorphism @xmath423 , our claim is established if we show that in the real part of the sandwich @xmath424 the polynomial map @xmath425 is surjective . take a point @xmath426 : since it maps under @xmath427 to @xmath428 , there exists a real point @xmath429 and elements @xmath430 , for @xmath431 , such that @xmath432 . since however @xmath433 and @xmath434 , we get : @xmath435 , @xmath436 . since @xmath437 is odd , then @xmath438 and @xmath439 : we have thus proven that @xmath440 . the cyclic group stabilizes @xmath452 , and the origin is the only fixed point , while the action on @xmath413 has no real fixed points , hence @xmath409 is a seifert bundle and the multiplicity over the origin is @xmath437 . as a consequence of the previous lemma , given any real du val surface @xmath3 with only @xmath405 singularities with @xmath6 even , the projectivized tangent bundle of @xmath3 , @xmath453 is a 3-fold with terminal singularities , such that _ there exists a minimal real du val del pezzo surface @xmath3 of degree @xmath56 having exactly two singular points , of type @xmath57 , and such that the real part @xmath35 has a connected component containing the two singular points and which is homeomorphic to a real projective plane . _ let @xmath58 be the projectivized tangent bundle of @xmath3 : then @xmath58 has terminal singularities , @xmath59 is contained in the smooth locus of @xmath58 , in particular if @xmath5 is obtained resolving the singular points of @xmath58 , then @xmath60 . moreover @xmath61 is orientable and contains a connected component @xmath2 which is seifert fibred over a non orientable orbifold of hyperbolic type ( the real projective plane with two points of multiplicity 3 ) . we construct the real del pezzo surface @xmath3 as the blow up of the real projective plane in @xmath454 real points . we shall indeed construct a family of such surfaces @xmath455 , having two real @xmath456 singularities , and two real and non isolated @xmath114 singularities . for certain values of the parameters , once we represent the del pezzo surface @xmath455 as the double cover of the quadric cone @xmath62 branched on the vertex of the cone and on a real branch curve @xmath63 , then the two @xmath114 points give rise to two isolated real points of the real part @xmath143 of the branch curve . using then the generalization of brusotti s theorem given in theorem 4.3 of @xcite , we can take a small deformation which leaves unchanged the two real @xmath456 singularities , but deforms the two @xmath114 points replacing the two isolated points of @xmath143 by two small ovals . we obtain now a real del pezzo surface @xmath457 of degree 1 with exactly two real @xmath456 singularities , and our desired real del pezzo surface @xmath3 will be the same complex surface @xmath457 , but with a new real structure @xmath458 , where @xmath459 is the real involution of @xmath457 , and @xmath460 is the * bertini involution * , the covering involution of the bianticanonical morphism , yielding @xmath457 as a double cover of the quadric cone @xmath62 . in terms of this last representation , this simply amounts to exchanging the region of positivity with the region of negativity . for this reason , @xmath3 has now two @xmath57 singularities , and the inside of the two ovals are now regions of positivity ; we conclude that the connected components of @xmath461 consist of two spheres @xmath462 , and of a component homeomorphic to the real projective plane , and containing the two @xmath57 singularities . each pair of lines shall yield a respective @xmath128 configuration on our del pezzo surfaces of degree @xmath56 . we choose in fact eight points in the plane , such that each of the four lines contains four of them , namely the set @xmath466 we see easily that the configuration is symmetric with respect to the real involution @xmath467 such that @xmath468 , and this symmetry is responsible of the fact that there is a conic @xmath226 containing the following symmetrical set of six points : @xmath469 we let @xmath471 be the blow up of the real projective plane in the above eight points . we claim that @xmath471 is a weak del pezzo surface of degree @xmath56 , i.e. , that its anticanonical divisor is nef ( it is big since it clearly satisfies @xmath472 ) . this claim follows right away from the fact that , if we take homogeneous coordinates @xmath473 on @xmath474 , the system of cubics through the eight points is the pencil spanned by the two cubics @xmath475 whose proper transforms meet only ( transversally ) in the point ` at infinity ' @xmath476 . the bianticanonical morphism @xmath477 of @xmath471 is the double covering of the quadric cone @xmath478 given by @xmath479 , where @xmath480 has as set of zeros the union of six lines and passes doubly through the @xmath454 points , but does not vanish on the base point of the anticanonical pencil ( hence , @xmath481 is not a linear combination of @xmath482 ) . the morphism @xmath477 clearly factors through the quotient of @xmath471 by the involution @xmath483 ( @xmath484 ) and we have then a factorization @xmath485 , where @xmath486 , and @xmath487 is the quotient projection . hence it follows that @xmath488 contracts the images under @xmath487 of the @xmath489-curves on @xmath471 , and that the branch curve @xmath490 is the image under @xmath477 of the projective line @xmath491 . the branch curve @xmath63 is irreducible of arithmetic genus @xmath152 , and it has @xmath152 singular points , corresponding to the blow down of the curves @xmath492 : hence we conclude that the only @xmath489-curves on @xmath471 lie on the corresponding fibres of the anticanonical pencil , and a direct inspection shows that there are no other @xmath489-curves on @xmath471 . we want now to find a choice of the parameters such that the curves @xmath493 are real and do not intersect the line @xmath494 , thereby yielding two real isolated double points of the branch curve @xmath143 . using now the plane representation of the quadric cone ( a main tool in @xcite ) given by the rational map @xmath499 , we see that the line @xmath500 maps to @xmath501 , and using the inhomogeneous coordinate @xmath502 on the line @xmath494 , we obtain @xmath503 , @xmath504 . the rational function @xmath76 only ramifies for @xmath513 , and for @xmath514 we get the point @xmath515 . for @xmath516 we must go to the other chart of our segre - hirzebruch surface @xmath110 , and setting @xmath517 , we get the parametrization @xmath518 , showing that the point @xmath519 of @xmath63 is an ordinary cusp with vertical tangent . observe however that , since on our surface @xmath471 @xmath520 and as @xmath521 @xmath522 , the image of @xmath523 ( region of positivity for the double cover ) is the one beneath the curve @xmath143 . moreover , the isolated real double points of the branch curve @xmath143 are not isolated points of the real del pezzo surface @xmath523 , whence they lie in the region of positivity . since we blew up @xmath454 real points in the plane , we have for the euler characteristic ( which we calculate through the @xmath524 betti numbers ) : @xmath525 as already announced , we obtain a du val del pezzo surface @xmath457 with two @xmath456 singularities using a small deformation on @xmath455 which realizes the following local deformations : no local deformation at the @xmath456 singularities , and , at the @xmath114 singularities , which are of the form @xmath527 , for suitable local coordinates @xmath528 on the quadric cone @xmath62 , a small deformation of type @xmath529 , where @xmath530 is a small real number . the existence of this global deformation is guaranteed by ( * ? ? ? * theorem 4.3 ) ; in practice , some simple calculations show that this deformation is obtained by simply moving the @xmath454 points on the four lines @xmath531 but breaking the symmetry @xmath532 . this deformation produces on the one side a real du val del pezzo @xmath457 with @xmath533 on the other side it produces two real ovals in the branch curve @xmath143 , whose respective interiors are now in the region of negativity . we define now @xmath3 as the same complex surface as @xmath457 , but with real structure @xmath534 , @xmath483 being the bertini involution . now the region of positivity and negativity are exchanged , since the old real function @xmath535 is replaced by @xmath536 ; hence , instead of having @xmath537 , we obtain @xmath538 . clearly @xmath35 consists of a real projective plane with two @xmath57 singularities , together with two spheres . the ( @xmath42- ) minimality of @xmath3 ( see e.g. @xcite ) is a consequence of the following _ proof . _ let @xmath540 be the minimal resolution of singularities of @xmath3 , and observe that @xmath541 , since the blow up of a real singular point of type @xmath57 yields a pair of complex conjugate @xmath489-curves . arguing as in lemma [ lem : rho - euler ] we calculate @xmath542 here , by the definition of @xmath291 , @xmath543 the remaining assertions follow from lemma [ ptb ] , and most of them were already mentioned in the previous remark : observe finally that a real projective plane with two points of multiplicity @xmath49 is a orbifold of hyperbolic type , since @xmath546

we present the results of an optical lightcurve survey of 114 jovian trojan asteroids conducted to determine the fraction of contact binaries . sparse - sampling was used to assess the photometric range of the asteroids and those showing the largest ranges were targeted for detailed follow - up observations . this survey led to the discovery of two trojan asteroids , ( 17365 ) and ( 29314 ) displaying large lightcurve ranges ( 1 magnitude ) and long rotation periods ( 2 rotations per day ) consistent with a contact binary nature . the optical lightcurves of both asteroids are well matched by roche binary equilibrium models . using these binary models , we find low densities of 600 kg m and 800 kg m , suggestive of porous interiors . the fraction of contact binaries is estimated to be between 6% and 10% , comparable to the fraction in the kuiper belt . the total binary fraction in the trojan clouds ( including both wide and close pairs ) must be higher .

0706.0233

the existence and importance of binary asteroids in small - body populations has only been realized in the last decade , after the first unambiguous detection of a satellite around main - belt asteroid 243 ida by the galileo spacecraft @xcite . it is now evident that binaries exist in the main - belt asteroids , the near - earth asteroids and in the kuiper belt ( see review by @xcite and references therein ) . apart from spacecraft flybys ( and the rare case of measuring gravitational perturbations of planets by very large asteroids ) , studying the orbital dynamics of binary systems provides the only method available for calculating mass and density . density measurements are important as probes of internal structure , enabling constraints to be placed on the porosity and composition . the jovian trojan asteroids are trapped in a 1:1 mean motion resonance with jupiter . they form two large clouds around the stable ( l4 , l5 ) lagrangian points 60 ahead of and behind the giant planet . it has been estimated that @xmath0 10@xmath3 trojan asteroids with diameters larger than 1-km exist @xcite , comparable in number to the main belt population ( @xmath4 asteroids , @xcite ) , making it clear that they comprise an important reservoir of information . the trojan asteroids of jupiter have yet to be searched systematically for the presence of binaries . despite this fact , two trojan binaries have already been identified : 617 patroclus , a resolved wide binary discovered by merline et al . ( 2001 ) , while 624 hektor has a distinctive lightcurve that indicates it is a close or contact binary ( @xcite , @xcite ) and a widely separated satellite has recently been imaged @xcite . the trojans are intriguing because they show larger photometric ranges when compared with main - belt asteroids @xcite , particularly those with diameters larger than 90-km @xcite . large lightcurve amplitudes suggest elongated shapes or binarity . while it is not clear whether the trojans formed at their current location alongside jupiter or were trapped after forming at larger heliocentric distances @xcite , it is believed that these bodies are primordial . understanding their composition and internal structure is therefore of great interest , making density determination vital . the density of trojan 617 patroclus has been estimated as @xmath5 kg m@xmath2 based on the measured orbital period and size , and on diameter determinations made from infrared data @xcite . this low density contrasts with a comparatively high estimate for 624 hektor , namely @xmath6 = 2480@xmath7 kg m@xmath2 , determined from the lightcurve and a roche binary model @xcite . close or contact binaries are composed of two asteroids in a tight orbit around each other . the trojan contact binary fraction is potentially important in distinguishing between various formation theories . for example , one model of binary formation by dynamical friction predicts that close binaries should be common @xcite while another based on 3-body interactions asserts that they should be rare @xcite . the nature of the trojan binaries can also reveal clues about their formation . it is known that different mechanisms formed binaries in the main belt and the kuiper belt because of the distinct types of binaries found in both populations . it is suspected that gravitational processes predominantly form kuiper belt binaries , the known examples of which have components of comparable mass and large separations @xcite . sub - catastrophic impacts followed by gravitational interaction with the debris formed are the leading way to form tight binary systems with unequal mass components that make up the larger main - belt binary population @xcite . a comparative study of the binaries in the trojan clouds , the main belt and the kuiper belt might illuminate the different roles played by formation conditions in these populations . motivated by the lack of studies about trojan binaries , the aim of this paper is to investigate the fraction of close or contact binary systems among the jovian trojan population . contact binaries are specifically targeted for the ease with which they can be identified using optical lightcurve information . here , we present a technique called sparse sampling , which we used to conduct a lightcurve survey of 114 jovian trojan asteroids . the results of this survey , the discovery of two suspected contact binary asteroids and a discussion of the binary fraction in the jovian trojan population will follow . the maximum photometric range that can be exhibited by a rotationally elongated , strengthless body is 0.9 mag @xcite . ranges larger than 0.9 mag . are strongly suggestive of a contact binary nature , in which mutual gravitational deformation of the components can drive the range up to @xmath0 1.2 magnitudes @xcite . in principle , structurally strong bodies can maintain any shape and show an arbitrarily large photometric range . however , most main - belt asteroids larger than @xmath0 150-m in diameter show little sign of possessing internal strength sufficient to resist gravity and/or rotational deformation @xcite and we expect that the trojan asteroids are similarly structurally weak . in what follows , we assume that objects with photometric range @xmath80.9 mag . are candidate contact binaries . to examine the efficiency of sparse lightcurve sampling , we conducted a series of monte carlo tests . the tests were applied to asteroids with a photometric range of 1.2 magnitudes and double - peaked lightcurve periods uniformly distributed between 6 and 20 hours . the lightcurves were uniformly sampled by n=1,2 ... 10 observations over one night . asteroids for which the sparse - sampling technique detected photometric ranges between 0.9 and 1.2 magnitudes were picked out as successful candidates . monte carlo simulations suggest that between 85% and 92% of asteroids with photometric ranges of 1.2 magnitudes would be identified as contact binary candidates from just five measurements of brightness per night ( see figure 1 ) . ( the efficiency of detecting brightness variations larger than 0.9 magnitudes ranged from @xmath0 71% for asteroids with actual peak - to - peak lightcurve amplitudes of 1.0 magnitudes to @xmath0 81% of asteroids with peak - to - peak amplitudes of 1.1 magnitudes . ) the simulations indicate that the accuracy with which contact binary candidates are identified varies little when sampling between five and eight lightcurve points per asteroid ( see figure 1 ) . the advantage of sparse sampling is clear : estimates of photometric range for a large number of asteroids can be made rapidly , significantly reducing observing time . asteroids exhibiting large photometric ranges in the sparse sampling study are subsequently targeted for detailed follow - up observations with dense coverage in rotational phase space . to further test the sparse sampling technique , we observed 2674 pandarus and 944 hidalgo , two asteroids known to show large photometric variations . from published lightcurves , 2674 pandarus is known to have a photometric range of 0.49 magnitudes @xcite . using the sparse sampling technique , with the same sampling as for all other asteroids in the study ( and without prior knowledge of the rotational phase ) , we measured a lightcurve amplitude of 0.50 @xmath9 0.01 magnitudes for pandarus . hidalgo has shown a maximum photometric variation of 0.60 magnitudes @xcite , whereas sparse sampling measured the brightness range to be 0.58 @xmath9 0.02 magnitudes ( see figures 2 and 3 ) . the agreement results show that the photometric range can be usefully estimated with only five measurements of asteroid brightness . having gained confidence in the technique through simulations and observational tests , we applied sparse sampling to the trojan asteroids . taking five short exposures , while cycling through the asteroids , we were able to obtain limited sampling of 114 asteroid lightcurves in nine good weather nights of observing . we obtained sparsely sampled optical lightcurve data for the jovian trojan asteroids using both the university of hawaii 2.2-m telescope on mauna kea and the lulin one - meter telescope ( lot ) in taiwan . we used a 2048 x 2048 pixel tektronix charge - coupled device ( ccd ) on the 2.2-m telescope . this detector has a 0.219 arcseconds per pixel image scale and a field of view of 7.5 square arcminutes . the ccd on lot ( versarray:1300b ) has 1340 x 1300 pixels with 0.516 arcseconds per pixel scale , and a field of view of 11.5 x 11.2 arcminutes . all images were taken in the r band with exposure times scaled to the brightnesses of the asteroids . on lot , the exposure times ranged from 30 seconds for objects brighter than 15th magnitude , up to 120 seconds for 19th magnitude trojans . at the 2.2-m telescope , the exposure times ranged from 10 seconds for objects brighter than 17th magnitude , to 150 seconds for 20th magnitude asteroids . see table 1 for a description of the observations . lccccr 2005 march 07 & lot 1-m & 2.0 & sparse & full & scattered cirrus + 2005 march 09 & lot 1-m & 2.2 & sparse & full & windy + 2005 march 11 & lot 1-m & 2.0 & sparse & half & cloudy + 2005 march 13 & lot 1-m & 1.7 & sparse & full & clear skies + 2005 april 05 & uh 2.2-m & 0.6 & sparse & full & cirrus + 2005 april 06 & uh 2.2-m & 0.6 - 0.8 & sparse & half & cloudy + 2005 april 07 & uh 2.2-m & 0.6 & sparse & half & photometric + 2005 april 09 & uh 2.2-m & 0.6 - 0.7 & sparse & half & clear + 2005 april 11 & uh 2.2-m & 0.6 & sparse & half & clear + 2005 april 12 & uh 2.2-m & 0.7 & sparse & half & clear + 2005 april 14 & uh 2.2-m & 0.7 & sparse & half & clear + 2005 april 15 & uh 2.2-m & 0.8 & sparse & half & cloudy + 2005 april 17 & uh 2.2-m & 0.8 & dense & half & cloudy + 2005 april 18 & uh 2.2-m & 0.8 - 1.0 & dense & half & moon rising + 2006 february 01 & uh 2.2-m & 1.0 & dense & full & focus problems + 2006 february 02 & uh 2.2-m & 0.6 & dense & full & clear + 2006 february 04 & uh 2.2-m & 1.5 & dense & full & strong winds + 2006 february 24 & uh 2.2-m & 1.0 - 1.2 & dense & full & windy + 2006 april 24 & uh 2.2-m & 0.7 & dense & half & cloudy / clear + 2006 april 29 & uh 2.2-m & 0.8 & dense & half & clear , windy + 2006 april 30 & uh 2.2-m & 0.9 & dense & half & clear , windy + 2006 may 01 & uh 2.2-m & 0.9 - 1.0 & dense & half & windy + [ table : obs_journal ] lcccccccc 884 & uh & 16.37 & 0.13 & -0.09 & -0.07 & 0.08 & -0.05 & 0.22 + 1172 & uh & 15.78 & 0.06 & -0.05 & -0.04 & 0.03 & & 0.11 + 1173 & lot & 16.85 & 0.02 & -0.20 & -0.08 & 0.17 & 0.10 & 0.37 + 1208 & uh & 16.60 & 0.06 & 0.06 & -0.01 & 0.00 & -0.06 & 0.12 + 1583 & uh & 16.87 & -0.02 & -0.07 & 0.00 & 0.04 & 0.04 & 0.11 + 1647 & uh & 18.88 & -0.20 & -0.14 & 0.09 & 0.24 & & 0.44 + 1867 & uh & 15.82 & 0.04 & 0.04 & 0.02 & -0.07 & -0.04 & 0.12 + 1868 & uh & 17.52 & 0.03 & -0.03 & -0.08 & 0.06 & 0.02 & 0.14 + 1869 & uh & 19.51 & -0.18 & 0.01 & 0.03 & 0.07 & 0.08 & 0.26 + 1870 & uh & 17.90 & -0.05 & -0.01 & 0.05 & -0.03 & 0.03 & 0.10 + 1871 & uh & 19.29 & 0.05 & 0.05 & 0.01 & -0.07 & -0.04 & 0.12 + 1872 & lot & 17.99 & 0.09 & -0.03 & -0.01 & 0.01 & -0.06 & 0.15 + 1873 & uh & 17.24 & -0.14 & -0.05 & 0.11 & 0.08 & & 0.25 + 2146 & uh & 17.79 & -0.07 & 0.07 & 0.05 & -0.06 & 0.00 & 0.14 + 2207 & uh & 16.03 & 0.05 & -0.02 & -0.03 & 0.03 & -0.03 & 0.08 + 2241 & uh & 15.95 & 0.11 & -0.15 & 0.01 & 0.03 & & 0.26 + 2260 & uh & 17.47 & 0.03 & 0.12 & -0.09 & -0.03 & -0.03 & 0.22 + 2357 & uh & 15.93 & 0.01 & 0.04 & -0.02 & -0.03 & & 0.07 + 2357 & lot & 15.96 & -0.02 & -0.01 & -0.01 & 0.02 & 0.03 & 0.05 + 2363 & uh & 17.12 & 0.03 & 0.03 & -0.06 & 0.01 & & 0.09 + 2674 & lot & 16.54 & 0.19 & -0.23 & 0.03 & 0.26 & -0.23 & 0.49 + 2893 & uh & 16.62 & 0.14 & -0.03 & -0.11 & 0.00 & & 0.26 + 2895 & uh & 17.24 & -0.01 & -0.04 & 0.08 & -0.02 & & 0.12 + 2895 & lot & 16.73 & -0.02 & 0.05 & 0.01 & -0.01 & -0.04 & 0.09 + 2920 & uh & 16.57 & 0.10 & 0.06 & -0.10 & -0.06 & & 0.20 + 3240 & uh & 18.06 & -0.09 & -0.17 & 0.01 & -0.15 & 0.40 & 0.57 + 3317 & uh & 16.33 & 0.02 & 0.01 & -0.05 & -0.01 & 0.04 & 0.09 + 3451 & uh & 15.91 & -0.10 & 0.14 & 0.04 & -0.02 & -0.06 & 0.25 + 3708 & uh & 17.20 & 0.01 & -0.04 & -0.01 & 0.01 & 0.02 & 0.06 + 3709 & uh & 17.42 & -0.05 & -0.04 & 0.01 & -0.05 & 0.13 & 0.18 + 4068 & uh & 17.41 & -0.07 & -0.06 & 0.04 & 0.04 & 0.04 & 0.11 + 4348 & uh & 17.09 & 0.13 & 0.10 & -0.03 & -0.01 & -0.01 & 0.16 + 4489 & lot & 17.04 & 0.08 & -0.01 & -0.06 & -0.01 & & 0.13 + 4707 & lot & 17.81 & -0.18 & 0.16 & -0.10 & -0.08 & 0.21 & 0.40 + 4708 & lot & 17.35 & -0.20 & 0.13 & -0.04 & 0.11 & & 0.33 + 4709 & uh & 15.92 & -0.05 & 0.05 & 0.09 & -0.05 & -0.06 & 0.15 + 4715 & lot & 17.13 & 0.17 & -0.23 & -0.13 & 0.23 & -0.03 & 0.46 + 4722 & lot & 17.28 & -0.02 & 0.00 & 0.01 & -0.04 & 0.05 & 0.08 + 4754 & lot & 16.95 & 0.02 & 0.00 & 0.01 & -0.01 & -0.01 & 0.03 + 4792 & uh & 17.85 & 0.01 & 0.01 & 0.01 & -0.03 & & 0.05 + 4792 & lot & 17.56 & 0.17 & 0.03 & -0.10 & -0.06 & -0.04 & 0.27 + 4805 & uh & 17.73 & 0.01 & 0.04 & 0.04 & -0.09 & & 0.14 + 4827 & uh & 17.86 & 0.01 & 0.07 & 0.02 & -0.06 & -0.05 & 0.13 + 4828 & uh & 17.63 & 0.13 & 0.11 & -0.06 & -0.19 & & 0.32 + 4828 & lot & 17.47 & 0.06 & 0.00 & -0.11 & 0.06 & & 0.18 + 4832 & lot & 17.55 & 0.01 & 0.00 & 0.01 & 0.00 & -0.02 & 0.03 + 4833 & uh & 17.25 & -0.18 & 0.10 & 0.13 & 0.05 & -0.10 & 0.31 + 4834 & uh & 17.70 & 0.06 & 0.02 & -0.02 & -0.04 & -0.03 & 0.10 + 4867 & lot & 16.97 & 0.02 & -0.01 & -0.02 & -0.03 & 0.04 & 0.07 + 5119 & uh & 17.97 & 0.07 & 0.07 & -0.02 & -0.11 & & 0.18 + 5233 & uh & 18.85 & 0.00 & -0.08 & 0.06 & 0.02 & & 0.15 + 5648 & uh & 17.84 & 0.06 & 0.02 & -0.03 & -0.05 & & 0.11 + 6002 & uh & 18.00 & 0.06 & 0.03 & -0.02 & -0.07 & & 0.13 + 9030 & uh & 18.20 & -0.21 & 0.06 & 0.36 & -0.08 & -0.13 & 0.57 + 9142 & lot & 18.19 & -0.08 & 0.05 & 0.04 & -0.01 & -0.01 & 0.13 + 9431 & lot & 18.19 & 0.07 & -0.01 & -0.12 & -0.06 & 0.13 & 0.25 + 9694 & uh & 17.90 & -0.05 & -0.16 & -0.02 & 0.08 & 0.15 & 0.32 + 11554 & lot & 17.31 & 0.03 & 0.00 & -0.03 & 0.00 & -0.01 & 0.06 + 11668 & uh & 19.33 & -0.05 & -0.02 & 0.14 & -0.03 & -0.08 & 0.22 + 12649 & uh & 19.64 & 0.04 & 0.00 & -0.06 & 0.00 & 0.02 & 0.10 + 13402 & uh & 19.08 & -0.02 & 0.00 & 0.00 & 0.02 & 0.01 & 0.04 + 15527 & lot & 18.50 & 0.05 & 0.29 & -0.13 & -0.20 & & 0.49 + 16667 & uh & 19.02 & -0.11 & 0.06 & 0.05 & 0.01 & 0.00 & 0.17 + 17172 & lot & 17.83 & 0.04 & 0.03 & -0.04 & 0.00 & -0.03 & 0.07 + 17365 & lot & 17.61 & -0.21 & 0.35 & 0.05 & -0.20 & & 0.56 + 17419 & uh & 18.76 & -0.03 & 0.00 & 0.00 & 0.02 & 0.02 & 0.05 + 17442 & uh & 19.39 & 0.11 & 0.00 & 0.06 & -0.04 & -0.13 & 0.24 + 17492 & uh & 17.70 & 0.09 & 0.10 & 0.03 & -0.05 & -0.16 & 0.26 + 18037 & uh & 19.22 & -0.05 & -0.06 & -0.03 & -0.01 & 0.15 & 0.21 + 18054 & uh & 18.22 & -0.06 & 0.02 & -0.01 & -0.01 & 0.05 & 0.11 + 23463 & uh & 19.15 & -0.07 & 0.01 & 0.08 & -0.04 & 0.02 & 0.15 + 23549 & uh & 18.90 & -0.03 & 0.02 & 0.09 & 0.00 & -0.08 & 0.16 + 24018 & uh & 19.19 & 0.09 & 0.02 & -0.18 & -0.11 & 0.17 & 0.35 + 24022 & uh & 19.79 & 0.06 & -0.08 & -0.06 & 0.08 & & 0.16 + 24449 & uh & 19.50 & 0.13 & 0.08 & -0.17 & -0.17 & 0.13 & 0.30 + 24451 & uh & 18.19 & 0.04 & 0.00 & 0.05 & -0.01 & -0.07 & 0.12 + 24452 & uh & 19.06 & -0.03 & 0.03 & -0.03 & 0.01 & 0.01 & 0.06 + 24456 & uh & 19.37 & -0.15 & 0.10 & 0.13 & 0.04 & -0.11 & 0.27 + 24531 & lot & 19.72 & 0.25 & -0.07 & 0.05 & 0.00 & -0.23 & 0.48 + 25344 & uh & 19.22 & 0.13 & 0.01 & -0.13 & -0.11 & 0.09 & 0.26 + 25347 & uh & 19.23 & 0.09 & 0.20 & 0.04 & -0.16 & -0.17 & 0.37 + 29314 & uh & 19.44 & 0.22 & 0.31 & 0.21 & -0.21 & -0.53 & 0.83 + 30498 & uh & 19.59 & 0.00 & -0.07 & -0.12 & 0.10 & 0.09 & 0.22 + 30499 & uh & 19.76 & 0.05 & -0.03 & 0.04 & -0.07 & 0.01 & 0.12 + 30505 & uh & 19.02 & -0.13 & 0.15 & 0.08 & -0.22 & 0.12 & 0.34 + 30506 & uh & 18.78 & -0.19 & -0.18 & -0.02 & 0.19 & 0.20 & 0.39 + 30704 & uh & 18.67 & -0.08 & -0.03 & -0.01 & 0.11 & & 0.19 + 30942 & uh & 18.52 & 0.04 & 0.02 & 0.00 & -0.02 & -0.04 & 0.08 + 31806 & uh & 19.51 & 0.15 & 0.07 & -0.09 & -0.03 & -0.10 & 0.25 + 31814 & uh & 19.81 & -0.11 & 0.11 & 0.23 & -0.09 & -0.16 & 0.39 + 31819 & uh & 18.90 & 0.20 & 0.01 & 0.00 & -0.03 & -0.17 & 0.37 + 31820 & uh & 20.06 & 0.16 & 0.09 & 0.05 & 0.12 & -0.40 & 0.56 + 32482 & lot & 18.68 & 0.13 & -0.14 & 0.13 & 0.03 & -0.15 & 0.27 + 32496 & uh & 18.01 & 0.01 & -0.01 & -0.01 & 0.02 & -0.02 & 0.04 + 32811 & uh & 18.43 & -0.11 & -0.02 & 0.01 & 0.05 & 0.07 & 0.18 + 47962 & uh & 19.59 & 0.04 & -0.05 & -0.02 & 0.00 & 0.03 & 0.09 + 51364 & uh & 18.49 & 0.02 & 0.05 & 0.04 & -0.01 & -0.09 & 0.15 + 53436 & uh & 18.40 & -0.03 & 0.02 & 0.00 & 0.00 & 0.01 & 0.04 + 55060 & lot & 18.85 & 0.27 & -0.09 & -0.22 & 0.03 & & 0.48 + 55419 & lot & 18.68 & 0.01 & -0.22 & -0.05 & 0.20 & 0.06 & 0.42 + 65216 & uh & 19.67 & 0.14 & -0.02 & -0.05 & -0.03 & -0.03 & 0.19 + 67065 & uh & 18.99 & 0.08 & -0.12 & -0.09 & 0.09 & 0.04 & 0.21 + 69437 & uh & 19.54 & -0.06 & 0.01 & 0.01 & 0.02 & 0.01 & 0.08 + 73677 & uh & 19.34 & 0.06 & 0.03 & 0.00 & -0.02 & -0.01 & 0.08 + 85798 & uh & 19.10 & -0.08 & 0.03 & 0.02 & 0.03 & 0.00 & 0.12 + 1999 xj55 & uh & 19.29 & 0.04 & 0.00 & -0.03 & -0.01 & & 0.06 + 2000 tg61 & uh & 19.76 & 0.01 & -0.01 & 0.00 & 0.02 & -0.03 & 0.04 + 2000 sj350 & uh & 20.17 & -0.20 & -0.14 & -0.13 & 0.15 & 0.08 & 0.35 + 2001 qz113 & uh & 19.53 & -0.02 & -0.02 & -0.02 & 0.00 & 0.05 & 0.07 + 2001 xw71 & uh & 20.24 & 0.06 & -0.03 & -0.05 & 0.17 & -0.08 & 0.24 + 2001 qq199 & uh & 20.51 & -0.12 & -0.09 & 0.04 & 0.05 & 0.11 & 0.23 + 2004 bv84 & uh & 20.34 & 0.05 & -0.01 & 0.01 & -0.05 & & 0.10 + 2004 fx147 & uh & 19.67 & 0.06 & -0.16 & -0.13 & 0.02 & 0.20 & 0.36 + 2005 ej133 & uh & 20.15 & -0.11 & 0.01 & 0.00 & 0.08 & 0.01 & 0.18 + lcccccc 884 & 8.53 & 5.66 & 5.34 & 9.9 & 146 & l5 + 1172 & 8.00 & 5.68 & 5.24 & 9.4 & 193 & l5 + 1173 & 9.08 & 6.02 & 5.27 & 6.6 & 150 & l5 + 1208 & 8.86 & 5.69 & 5.17 & 9.0 & 134 & l5 + 1583 & 9.30 & 5.33 & 4.92 & 10.2 & 99 & l4 + 1647 & 11.50 & 5.20 & 4.70 & 10.1 & 37 & l4 + 1867 & 8.18 & 5.34 & 5.12 & 10.7 & 163 & l5 + 1868 & 9.91 & 5.50 & 5.00 & 9.5 & 80 & l4 + 1869 & 12.10 & 5.49 & 4.75 & 7.5 & 34 & l4 + 1870 & 10.29 & 5.42 & 5.03 & 10.1 & 64 & l5 + 1871 & 11.47 & 5.46 & 5.46 & 10.5 & 36 & l5 + 1872 & 10.78 & 5.51 & 4.61 & 4.7 & 84 & l5 + 1873 & 9.71 & 5.11 & 5.02 & 11.3 & 78 & l5 + 2146 & 9.98 & 5.69 & 5.28 & 9.6 & 77 & l4 + 2207 & 8.73 & 5.05 & 4.61 & 10.7 & 127 & l5 + 2241 & 8.34 & 5.17 & 5.17 & 11.1 & 148 & l5 + 2260 & 9.92 & 5.39 & 4.92 & 9.8 & 77 & l4 + 2357 & 8.34 & 5.29 & 4.90 & 10.4 & 149 & l5 + 2357 & 8.79 & 5.29 & 4.51 & 7.1 & 164 & l5 + 2363 & 9.74 & 5.24 & 4.69 & 9.7 & 85 & l5 + 2674 & 9.37 & 5.17 & 4.49 & 8.6 & 111 & l5 + 2893 & 8.75 & 5.56 & 5.50 & 10.4 & 128 & l5 + 2895 & 9.79 & 5.25 & 4.69 & 9.6 & 81 & l5 + 2895 & 9.67 & 5.24 & 4.39 & 6.3 & 118 & l5 + 2920 & 9.23 & 5.25 & 4.64 & 9.3 & 111 & l4 + 3240 & 10.04 & 5.92 & 5.61 & 9.5 & 75 & l5 + 3317 & 8.44 & 5.78 & 5.39 & 9.5 & 157 & l5 + 3451 & 8.38 & 5.44 & 4.90 & 9.4 & 163 & l5 + 3708 & 9.29 & 5.93 & 5.41 & 8.6 & 113 & l5 + 3709 & 9.77 & 5.58 & 5.04 & 9.1 & 87 & l4 + 4068 & 9.97 & 5.33 & 4.78 & 9.5 & 78 & l4 + 4348 & 9.51 & 5.49 & 4.95 & 9.2 & 97 & l5 + 4489 & 9.26 & 5.54 & 5.37 & 10.3 & 104 & l4 + 4707 & 10.60 & 5.53 & 4.62 & 4.4 & 96 & l5 + 4708 & 10.05 & 5.34 & 4.65 & 8.2 & 84 & l5 + 4709 & 8.53 & 5.30 & 4.71 & 9.3 & 153 & l5 + 4715 & 9.85 & 5.30 & 4.62 & 8.4 & 91 & l5 + 4722 & 10.04 & 5.44 & 4.60 & 6.0 & 102 & l5 + 4754 & 10.04 & 5.22 & 4.29 & 4.1 & 129 & l5 + 4792 & 10.00 & 5.69 & 5.25 & 9.5 & 74 & l5 + 4792 & 10.11 & 5.68 & 4.86 & 6.1 & 98 & l5 + 4805 & 10.06 & 5.46 & 5.11 & 10.2 & 71 & l5 + 4827 & 10.51 & 5.08 & 4.70 & 10.9 & 55 & l5 + 4828 & 10.18 & 4.96 & 4.81 & 11.6 & 59 & l5 + 4828 & 10.37 & 4.96 & 4.40 & 10.1 & 63 & l5 + 4832 & 10.00 & 5.94 & 5.03 & 4.2 & 128 & l5 + 4833 & 9.58 & 5.61 & 5.07 & 9.1 & 95 & l4 + 4834 & 9.80 & 5.94 & 5.38 & 8.4 & 91 & l4 + 4867 & 9.86 & 5.20 & 4.43 & 7.4 & 97 & l5 + 5119 & 10.08 & 5.74 & 5.30 & 9.3 & 72 & l5 + 5233 & 11.32 & 5.05 & 4.92 & 11.4 & 35 & l5 + 5648 & 9.76 & 5.88 & 5.62 & 9.7 & 82 & l5 + 6002 & 10.34 & 5.55 & 4.97 & 8.9 & 66 & l5 + 9030 & 11.03 & 5.11 & 4.46 & 9.1 & 49 & l5 + 9142 & 10.41 & 5.84 & 5.27 & 8.4 & 70 & l5 + 9431 & 10.51 & 5.52 & 5.17 & 10.0 & 59 & l4 + 9694 & 10.75 & 5.39 & 4.51 & 5.5 & 78 & l4 + 11554 & 10.12 & 5.32 & 4.53 & 6.9 & 90 & l5 + 11668 & 11.74 & 5.87 & 5.04 & 5.9 & 47 & l4 + 12649 & 11.61 & 5.90 & 5.58 & 9.5 & 36 & l5 + 13402 & 11.20 & 5.72 & 5.35 & 9.6 & 43 & l5 + 15527 & 10.95 & 5.32 & 5.01 & 10.5 & 47 & l4 + 16667 & 10.85 & 6.17 & 5.88 & 9.1 & 54 & l5 + 17172 & 10.59 & 5.45 & 4.61 & 6.0 & 80 & l5 + 17365 & 10.31 & 5.54 & 4.69 & 5.8 & 92 & l5 + 17419 & 11.33 & 5.38 & 4.81 & 9.3 & 43 & l5 + 17442 & 11.62 & 5.43 & 5.35 & 10.6 & 34 & l5 + 17492 & 10.10 & 5.42 & 5.07 & 10.3 & 70 & l5 + 18037 & 11.50 & 5.51 & 5.21 & 10.3 & 37 & l5 + 18054 & 10.85 & 5.19 & 4.74 & 10.3 & 50 & l5 + 23463 & 11.57 & 5.27 & 5.05 & 10.9 & 34 & l5 + 23549 & 11.54 & 5.10 & 4.76 & 11.0 & 35 & l5 + 24018 & 11.65 & 5.44 & 4.95 & 9.7 & 36 & l5 + 24022 & 12.12 & 5.66 & 5.12 & 9.0 & 30 & l5 + 24449 & 11.96 & 5.36 & 4.94 & 10.2 & 30 & l5 + 24451 & 10.33 & 5.89 & 5.39 & 8.8 & 70 & l5 + 24452 & 11.78 & 5.01 & 4.63 & 11.1 & 31 & l5 + 24456 & 11.86 & 5.33 & 4.90 & 10.2 & 31 & l5 + 24531 & 11.79 & 5.76 & 5.57 & 9.9 & 33 & l4 + 25344 & 11.54 & 5.62 & 5.11 & 9.2 & 39 & l5 + 25347 & 11.44 & 5.57 & 5.32 & 10.2 & 38 & l5 + 29314 & 11.84 & 5.46 & 5.02 & 9.9 & 32 & l5 + 30498 & 11.78 & 5.70 & 5.33 & 9.7 & 34 & l5 + 30499 & 12.16 & 5.32 & 5.06 & 10.7 & 26 & l5 + 30505 & 11.60 & 5.32 & 4.76 & 9.5 & 37 & l5 + 30506 & 11.06 & 5.43 & 5.24 & 10.6 & 44 & l5 + 30704 & 11.20 & 5.34 & 4.85 & 9.8 & 43 & l5 + 30942 & 11.20 & 5.17 & 4.64 & 10.0 & 43 & l5 + 31806 & 11.73 & 5.67 & 5.29 & 9.7 & 34 & l5 + 31814 & 12.16 & 5.65 & 5.10 & 8.9 & 30 & l5 + 31819 & 11.65 & 5.14 & 4.57 & 9.8 & 36 & l5 + 31820 & 12.46 & 5.50 & 5.03 & 9.6 & 25 & l5 + 32482 & 11.36 & 5.26 & 4.66 & 9.2 & 42 & l5 + 32496 & 10.30 & 5.63 & 5.17 & 9.5 & 68 & l5 + 32811 & 11.14 & 5.00 & 4.64 & 11.2 & 41 & l5 + 47962 & 12.04 & 5.54 & 4.95 & 8.9 & 32 & l5 + 51364 & 11.42 & 4.95 & 4.34 & 9.9 & 39 & l5 + 53436 & 11.36 & 5.21 & 4.35 & 6.4 & 54 & l4 + 55060 & 11.41 & 5.34 & 4.83 & 9.6 & 40 & l5 + 55419 & 11.12 & 5.51 & 4.98 & 9.1 & 47 & l5 + 65216 & 12.49 & 5.44 & 4.54 & 5.2 & 36 & l4 + 67065 & 12.02 & 5.20 & 4.30 & 5.3 & 44 & l4 + 69437 & 11.89 & 5.55 & 5.10 & 9.7 & 32 & l5 + 73677 & 11.99 & 5.27 & 4.70 & 9.6 & 31 & l5 + 85798 & 11.89 & 5.45 & 4.57 & 5.6 & 45 & l4 + 1999 xj55 & 12.21 & 5.29 & 4.42 & 6.0 & 38 & l4 + 2000 tg61 & 12.23 & 5.47 & 4.92 & 9.2 & 28 & l5 + 2000 sj350 & 12.55 & 5.44 & 5.03 & 10.0 & 23 & l5 + 2001 qz113 & 11.98 & 5.39 & 4.98 & 10.2 & 30 & l5 + 2001 xw71 & 12.71 & 5.51 & 4.89 & 8.7 & 23 & l5 + 2001 qq199 & 12.59 & 6.36 & 5.48 & 4.6 & 37 & l5 + 2004 bv84 & 12.95 & 5.37 & 4.74 & 8.8 & 21 & l5 + 2004 fx147 & 12.61 & 5.25 & 4.39 & 6.0 & 31 & l4 + 2005 ej133 & 12.72 & 5.39 & 4.80 & 9.1 & 23 & l5 + raw data frames were bias subtracted , then flat fielded using a master flat field produced from median filtering dithered images of the sky taken at dusk and dawn . landolt ( 1992 ) standard star fields were imaged and measured to convert the instrumental magnitudes to an absolute magnitude scale . an aperture radius of eight pixels was consistently used throughout the observations for images taken on both telescopes . median sky values were determined using an adjacent annulus around the aperture having an outer radius of 20 pixels . the reason for similar aperture and sky annulus size on both telescopes , despite differing pixel scales was because of the significantly worse seeing conditions at lulin ( see table 1 ) . for the sparse sampling survey , two images were taken in each setting and then averaged to obtain the brightness measurement . the photometric uncertainties are small ( @xmath10 0.02 mag . ) compared to the photometric variability that is the subject of interest and so we have ignored these uncertainties in our presentation of the data . for the densely sampled lightcurves , errors for each observation were calculated using poisson statistics . the instrumental magnitude of the asteroid in each image was subtracted from the brightness of a nearby field star . the field star was chosen to be persistent in all five observations and helped reduce photometric errors by providing a correction for weather variations occuring throughout the night . images in which the asteroid was affected by proximity to a field star were rejected and resulted in some trojans having only four measurements of brightness rather than five . tables 2 and 3 contain results of the sparsely sampled lightcurve survey . in table 2 , the average r band magnitude , @xmath11 is listed , along with the independent measurements of the asteroid s brightness , expressed as deviations from the mean magnitude . the last column shows the maximum deviation measured , which gives a lower limit to the photometric range of each asteroid . table 3 contains the absolute magnitude , @xmath12 , which is defined as the magnitude an object would have if placed at heliocentric ( @xmath13 ) and geocentric @xmath14 distances of 1 au , and at a phase angle of @xmath15 = 0 degrees . the conversion between the apparent magnitude , @xmath16 and absolute magnitude , @xmath12 is @xmath17 where @xmath18 is the phase coefficient for which we used a value of 0.04 magnitudes per degree for the low albedo trojan asteroids @xcite . also listed in table 3 is an estimate of the equivalent circular diameter , @xmath19 which was calculated using @xcite @xmath20.\ ] ] here , _ p _ is the geometric albedo , for which a value of 0.04 was used throughout @xcite and @xmath21 = -27.1 is the apparent red magnitude of the sun @xcite . figures 4 shows the distribution of photometric ranges shown by the trojan asteroids in the sparsely - sampled lightcurve survey . for comparison , figure 5 shows the photometric range distributions of both the trojan and main belt asteroids with diameters between 70-km and 150-km ( main belt asteroid data taken from @xcite ) . figure 5 reveals that a larger fraction of trojan asteroids have photometric ranges larger than main belt asteroids , similar to previous studies by @xcite . a kolmogorov - smirnov statistical test found a 32.1% probability that the two distributions are drawn from the same parent distribution . trojan asteroids ( 17365 ) and ( 29314 ) showed the largest photometric ranges in the sparsely - sampled photometry , with 0.56 @xmath9 0.02 magnitudes and 0.83 @xmath9 0.03 magnitudes , respectively ( see table 2 ) . follow - up observations to obtain densely sampled optical lightcurves for both trojan asteroids were taken using the university of hawaii 2.2-m telescope between 2005 april 9th and 17th . we were unable to complete the observations due to bad weather coupled with the fact the asteroids were quickly setting . we were however , able to confirm the large photometric ranges to motivate further study of these trojan asteroids ( see figures 6 through 9 ) . in our first dense light curve study , in 2005 , asteroid ( 17365 ) had a photometric range of 0.98 @xmath9 0.02 magnitudes , centered at a mean of 10.64 @xmath9 0.01 magnitudes , while asteroid ( 29314 ) had a peak - to - peak lightcurve amplitude of 1.05 @xmath9 0.03 centered on 11.89 @xmath9 0.02 magnitudes . to complete the lightcurve study , we continued optical observations of both candidate contact binary asteroids in 2006 . figures 10 through 13 show the results of the photometric observations . in 2006 , asteroid ( 17365 ) showed a photometric range of 0.81 @xmath9 0.02 magnitudes , centered at a mean absolute magnitude of 10.76 @xmath9 0.01 . asteroid ( 29314 ) shows a peak - to - peak amplitude of 0.86 @xmath9 0.03 magnitudes , with a mean absolute magnitude of 11.80 @xmath9 0.02 . the phase dispersion minimization ( pdm ) method @xcite was used to determine possible rotation periods for each asteroid . figures 14 and 15 show plots of @xmath22 , which characterizes the dispersion in the data phased to a given period ( see @xcite for more information ) . the most likely rotation periods correspond to the smallest values of theta . several periods appeared to minimize theta , but when used to phase the data , the results were not persuasive lightcurves . in fact , only two periods per asteroid produced convincing lightcurve results . for trojan ( 29314 ) , minima consistent with the data occur at periods of 0.3133 @xmath9 0.0003 days ( 7.518 @xmath9 0.007 hr ) , and a double - peaked period of 0.6265 @xmath9 0.0003 days ( 15.035 @xmath9 0.007 hr ) . asteroid ( 17365 ) shows a single - peaked lightcurve period of 0.2640 @xmath9 0.0004 days ( 6.336 @xmath9 0.009 hr ) and double - peaked period of 0.52799 @xmath9 0.0008 days ( 12.672 @xmath9 0.019 hr ) . while both the single - peaked and double - peaked periods produce good fits for trojan asteroid ( 29314 ) , the double - peaked lightcurve is more convincing . the lightcurve of ( 29314 ) shows subtle differences in the shapes of the two minima , which is obvious by the spread in the data when phased to the single - peaked period ( see figure 10 and 11 ) . asteroid ( 17365 ) shows a more obvious double - peaked lightcurve ( see figures 12 and 13 ) with maxima of different shapes . the maxima of ( 17365 ) differ by 0.10 @xmath9 0.01 magnitudes while the minima differ by 0.06 @xmath9 0.01 magnitudes . trojan asteroids ( 17365 ) and ( 29314 ) show strong evidence of being contact binaries . both asteroids reveal photometric ranges greater than 0.9 magnitudes , sufficiently long rotation periods ( @xmath1 2 rotations per day ) and lightcurve profiles ( qualitatively similar to 624 hektor ) containing u - shaped maxima and v - shaped minima . here , we speculate about all possible explanations for the brightness variations in the lightcurve observations of these trojan asteroids , including albedo variations , elongated shapes or binarity @xcite . surface albedo contrasts provide a possible but unconvincing explanation for the large brightness variations of the trojans . amongst solar system objects , only iapetus , a satellite of saturn , shows strong spatial albedo variations which account for its large lightcurve amplitude . however , iapetus synchronous rotation about saturn plays a large role in producing the dichotomous behaviour of the satellite @xcite and this circumstance is not relevant in the context of the trojan asteroids . amongst previously studied asteroids , double - peaked lightcurves are almost always caused by rotational variations in the projected area , and reflect the elongated shapes of the bodies . while albedo contrasts can not be formally ruled out , we feel that they are an unlikely cause of the observed brightness variations . increasing evidence suggests asteroids have little or no internal strength , probably as a result of impacts that disrupt but do not disperse the object @xcite . the trojan asteroids have undergone a collisional history that is either similar to that of the main - belt asteroids @xcite or perhaps even more intense @xcite , making it highly probable that they , too , are gravity dominated rubble piles " , strengthless or nearly so in tension @xcite . studies have found that only the smallest main - belt asteroids , with diameters less than 0.15-km , have sufficient internal strength to overcome gravity @xcite . figure 5 from @xcite shows observations of decreasing maximum spin rate with increasing lightcurve amplitude ( a proxy for elongation ) of near - earth asteroids . this observation indicates a lack of fast rotating elongated bodies , which implies that asteroids larger than @xmath0 0.15-km are structurally weak and lack the tensile strength to withstand high rotation rates without becoming unstable and flying apart . also evident in figure 5 @xcite is the tendency of fast rotators to have spheroidal shapes , an indicator of gravity - dominated bodies which do not possess the internal strength to resist gravity . collectively , the observations point to asteroids being bodies of negligible strength , whose shapes are dominated by rotation and gravity . rotation rates must lie between 4 and 6 rotations per day in order for rotational elongation of a structurally weak body to be maintained . this is the range for which jacobi ellipsoids are possible figures of equilibrium @xcite . if the rotation rate was much higher than 6 rotations per day , the body would fall apart , while at a much lower rotation rate , the body would adopt a spherical figure of equilibrium . in 2005 , both asteroids ( 17365 ) and ( 29314 ) showed photometric variations larger than 0.9 magnitudes , above the threshold for rotational instability in a structurally weak body . additionally , both asteroids have double - peaked lightcurve periods that are too slow to cause sufficient rotational elongation . both observations indicate that rotationally - induced elongation is an insufficient explanation for the brightness variations of these trojan asteroids . we are therefore left with the strong possibility that trojan asteroids ( 29314 ) and ( 17365 ) are contact binaries . figure 16 is a plot of rotation periods and photometric ranges of several well studied kuiper belt objects and main - belt asteroids . it is divided into three main regions : region a spans the photometric ranges that can be explained by albedo variations , elongation or binarity of an asteroid . region b represents the characteristics explained by albedo variations or rotational elongation of an object , while variations in region c can only be explained by binary asteroids . both trojan asteroids lie well within region c , alongside contact binaries 216 kleopatra , 624 hektor and 2001 qg@xmath23 , contributing to their suspected binary nature . the lightcurve of a contact binary is expected to show u - shaped or spread out maxima and v - shaped or notched minima , as shown by the lightcurves of 2001 qg@xmath23 ( see @xcite ) and 624 hektor ( see figure 17 ) . these characteristic lightcurve profiles are unlike the distinctive notched " profile expected for wide , eclipsing binaries which remain flat for the majority of the orbit , and contain sharp dips during the relatively short eclipsing events . the photometric observations of trojan asteroids ( 29314 ) and ( 17365 ) are consistent with lightcurve profiles expected of very close or contact binary systems . 624 hektor was recently discovered to possess a satellite of diameter 15-km using keck laser guide star adaptive optics @xcite , but an independent density estimate derived from the orbital motion of this satellite has not yet been published . additionally , the imaging observations of 624 hektor indicate that its primary component has a double - lobed nature . similarities are obvious between the lightcurves of 624 hektor , ( 17365 ) and ( 29314 ) ( see figures 11 , 13 and 17 ) and consistent with our interpretation that the latter two asteroids are contact binaries . we used equilibrium models of roche binaries to determine how well the photometric observations of ( 17365 ) and ( 29314 ) could be matched by theoretical lightcurves of contact binary systems . a roche binary consists of a pair of homogeneous bodies in hydrostatic equilibrium orbiting each other . a strength of this modeling is the ability to estimate densities for the asteroids without knowing the sizes of the binary components . the exact shapes and rotation rates of the roche binaries were calculated using the mathematical description presented in @xcite ( see also @xcite ) . binary configurations were calculated for secondary to primary mass ratios from @xmath24 to @xmath25 in steps of 0.01 . for each value @xmath26 , equations 1 to 3 of @xcite were solved simultaneously to find possible shapes and orbital frequencies for the primary . the same equations were then solved using mass ratio @xmath27 to calculate the shapes and orbital rates for the secondary . finally , valid binaries are uniquely selected by matching pairs ( q,1/q ) with the same orbital frequency . this procedure is described in detail in @xcite and @xcite . the models were ray - traced using publicly available software pov - ray ( http://www.povray.org ) , but the surface scattering routine of pov - ray was rewritten to allow better control of the scattering function . the scattering law used here was first implemented by @xcite . it linearly combines single ( lommel - seeliger ) and multiple ( lambert ) scattering terms using a parameter @xmath28 @xcite , which varies from 0 to 1 . the resulting reflectance function is @xmath29 where @xmath30 and @xmath31 are the cosines of the incidence and emission angles . when @xmath32 , only single scattering is present , while @xmath33 simulates pure multiple scattering of light off the surface of the binaries . all binary configurations were raytraced for @xmath28 between 0 and 1 in steps of 0.1 . two viewing geometries were modelled , at aspect angles of 75 and 90@xmath34deg ( equator - on ) . the aspect angle lies between the line of sight of the observations and the rotation axis of the body . simulated illumination angles were chosen to match the phase angles at the time the data were taken . in total , nearly 50000 models were computed for comparison with the data . observations of ( 17365 ) and ( 29314 ) were simultaneously fitted for the different viewing orientations in 2005 and 2006 to find the best shape interpretation for the asteroids . we assumed that the objects were viewed equatorially in 2005 , thus producing the larger photometric range in the discovery epoch data . this assumption was encouraged by the fact that an aspect angle of 75 degrees ( rather than 90 degrees ) produced a better fit with the 2006 observations ( see figures 10 and 12 ) . figures 8 , 9 , 11 and 13 show the best - fit models overlaying lightcurve data , with residuals plotted underneath . best fit models were found by minimizing chi - squared . small deviations ( @xmath0 0.1 magnitudes ) from the binary model are evident for both asteroids , but are negligible compared with the total range of the observations , the more important parameter . presumably , the deviations are caused by irregularities on the surface of the asteroids , which were not included in the simple binary model , but without which the asteroids would be considered odd . the ability of the models to simultaneously fit two epochs of photometric observations lends strong support to the idea that we observed contact binary asteroids over two years at different viewing geometries . figures 18 and 19 show the shapes derived from the binary models for ( 17365 ) and ( 29314 ) . orbital periods combined with shape information allowed us to estimate the densities of the asteroids . the components of our model of asteroid ( 29314 ) were found to have a mass ratio of @xmath35 and a density of @xmath36 kg / m@xmath37 , while our best model of asteroid ( 17365 ) has a mass ratio of @xmath38 and density of @xmath39 kg / m@xmath37 . these low densities suggest porous asteroid interiors . if ( 29314 ) and ( 17365 ) have a rock / ice composition similar to the moons of jupiter , ( 29314 ) would have a porosity of @xmath0 64% , while ( 17365 ) would have a smaller porosity of 50% ( see figure 3 from @xcite ) . if ( 17365 ) and ( 29314 ) were composed purely of water ice , their porosities would be 15% and 35% , respectively @xcite . this pure water ice composition is unrealistic , however . it is interesting to note that our low density measurements are consistent with 617 patroclus @xcite . among the trojans , only 624 hektor is known to have a comparable lightcurve amplitude , making ( 29314 ) and ( 17265 ) the 2nd and 3rd known trojans to show such large rotational variations . lightcurve analysis suffers from the notorious non - uniqueness problem , which arises from the ability to reproduce any lightcurve with a complicated pattern of surface markings and shapes . our interpretation is not unique , but is the simplest , most plausible explanation for the behaviour of the trojan asteroids . following the method outlined in @xcite to account for the geometrical circumstances of the observations , we were able to estimate the fraction of contact binary systems among the jovian trojan asteroids . this method uses two very crude approximations . in the first approximation , the binary system is simplified to be an elongated , rectangular object with dimensions a @xmath40 b = c , having a lightcurve amplitude as follows : @xmath41 the range of lightcurve amplitudes used to identify contact binary asteroids is 0.9 to 1.2 magnitudes . for the maximum amplitude of 1.2 magnitudes and viewing angle of @xmath42 = 0@xmath43 , an axis ratio of @xmath44 = 3 is calculated from equation 4 . using this axis ratio and the minimum expected amplitude of 0.9 magnitudes , a viewing angle of 10@xmath43 was determined . therefore , the range of lightcurve amplitudes expected for a contact binary asteroid would only be observed if the earth lies within 10@xmath43 of the equator of the asteroid . the probability that the earth would lie within 10@xmath43 of the equator of a randomly oriented asteroid is p(@xmath45 ) = 0.17 . we found two suspected contact binary asteroids in our sample of 114 trojan asteroids , so the fraction of contact binary jovian trojan asteroids is approximately @xmath46 = 10 % . a second approximation uses an ellipsoid shape to represent the contact binary asteroid , again having dimensions a @xmath40 b = c , and having a lightcurve amplitude expressed by the following : @xmath47\sin^2 \theta + 1 \right\}.\ ] ] using the axis ratio of @xmath44 = 3 , in order to observe photometric ranges between 0.9 and 1.2 magnitudes , the earth must lie within 17@xmath43 of the equator of the ellipsoidal asteroid . the probability of a randomly oriented object having this geometrical orientation relative to the observer is p(@xmath48 ) = 0.29 , implying a contact binary fraction of @xmath49 = 6 % . we conclude that the fraction of contact binary trojan asteroids is @xmath06% to @xmath010% . this is a lower limit to the actual fraction as some of the objects not found in the survey sample to have large amplitudes might in fact have them because the sparse sampling method is not 100% efficient . the existence of likely contact binary 624 hektor separately suggests that the binary fraction is high . binaries with equal - sized components are rare in the main - belt ( the frequency of large main - belt binaries is @xmath0 2% @xcite ) and have yet to be observed in the near - earth asteroid population . however , they are abundant in the observed binary kuiper belt population , where the fraction lies between 10% and 20% @xcite . the results of this study show that there are three jovian trojan asteroids that reside in region c. the observations tend to suggest a closer relationship between the binary populations of the kuiper belt and the trojan clouds . this correlation could signify similar binary formation mechanisms in the two populations . this is an interesting connection considering that in one model of formation , the trojans are actually captured kuiper belt objects @xcite . however , it is clear that the total binary fractions in the kuiper belt and in the trojans needs to be more tightly constrained before conclusions can be made . the contact binaries detected were skewed towards those with components of comparable sizes , which are capable of producing photometric ranges @xmath40 0.9 magnitudes . for mass ratios @xmath50 1 , sparse sampling would more likely miss the eclipsing event and the photometric range would be @xmath51 0.9 magnitudes and would not attract our attention . the method was strongly dependent on geometrical circumstances , and only binaries viewed edge - on or almost equatorially would be detected in our survey . additionally , sparse sampling is only able to put lower limits on the photometric range of an asteroid , making the binary fraction a lower limit estimate . only binaries with sufficiently short orbital periods ( optimally between 6 to 12 hour rotation periods ) would be detected , so wide binaries were not accounted for in this study . therefore , again the measured binary fraction is a strong lower limit to the actual fraction and is suggestive of a significant binary population among the trojan clouds . our project is a pilot study for the much larger scale pan - starrs , which will detect every object with a red magnitude brighter than @xmath52 magnitude . it is estimated that approximately 10@xmath53 jovian trojans exist with red magnitudes lower than 24 , all of which will be detected using pan - starrs @xcite . our results suggest that pan - starrs will reveal between 6000 and 10,000 contact binary systems among the trojan clouds . lccccr ( 17365 ) & 10.76 & 92 & 12.672 & 0.98 & 780 + ( 29314 ) & 11.80 & 32 & 15.035 & 1.05 & 590 + 624 hektor & 7.37 & @xmath54 & 6.921 & 1.10 & 2200 + [ table : obs_journal ] sparsely sampled lightcurve measurements were used to statistically study the photometric variations of 114 jovian trojan asteroids . objects with large photometric ranges were targeted for follow - up in this survey , and are considered as candidate contact binary systems . 1 . the sparse sampling technique successfully confirmed known photometric ranges of both 944 hidalgo ( 0.58 @xmath9 0.02 magnitudes ) and 2674 pandarus ( 0.50 @xmath9 0.01 magnitudes ) . 2 . two of the 114 observed trojans , asteroids ( 17365 ) and ( 29314 ) , were found to show photometric ranges larger than expected for rotationally deformed equilibrium figures , and were targeted for dense follow - up lightcurve observations . the resulting ranges ( 0.98 @xmath9 0.02 mag and 1.05 @xmath9 0.03 mag , respectively ) and long rotation periods ( 12.672 @xmath9 0.019 hr and 15.035 @xmath9 0.007 hr ) of these two trojans are consistent with a contact binary structure for each object . roche binary models give densities of @xmath39 kg / m@xmath37 for asteroid ( 17365 ) and @xmath36 kg / m@xmath37 for asteroid ( 29314 ) , suggestive of porous interiors . if ( 17365 ) and ( 29314 ) are indeed contact binaries , then we estimate from our survey that the binary fraction of the jovian trojans is @xmath06% to 10% or more . the total binary fraction ( including both wide and close pairs ) must be higher . we thank john dvorak , daniel birchall , dave brennan and ian renaud - kim for operating the uh telescope and henry hsieh for assisting with the observations both in taiwan and honolulu . we are grateful for the assistance and expertise of the lulin observatory staff , in particular wen - ping chen , chung - ming ko and hc lin . support for this work by a grant to d.j . from nasa s origins program is greatly appreciated . astakhov , s.a . , lee , e.a . , farrelly , d. 2005 , mnras , 360 , 401 barucci , m.a . , cruikshank , d.p . , mottola , s. , lazzarin , m. 2002 , in asteroids iii , ed . w.f . bottke jr . , a. cellino , p. paolicchi & r. p. binzel ( tucson : univ . arizona press ) , 273 belton , m. , champman , c. , thomas , p. , davies , m. , greenberg , r. , klaasen , k. , byrnes , d. , damario , l. , synnott , s. , merline , w. , petit , j.m . , storrs , a. , zellner , b. 1995 , nature , 374 , 785 binzel , r.p . , & sauter , l.m . 1992 , icarus , 95 , 222 - 238 bowell , e. , hapke , b. , domingue , d. , lumme , k. , peltoniemi , j. , and harris , a.w . 1989 , in asteroids ii , ed . binzel , t. gehrels & m.s . matthews ( tucson : univ . arizona press ) , 524 cellino , a. , pannunzio , r. , zappal , v. , farinella , p. , paolicchi , p . 1985 , a&a , 144 , 355 chandrasekhar , s. 1987 , ellipsoidal figures of equilibrium ( new york : dover ) chapman , c.r . , veverka , j. , thomas , p.c . , klaasen , k. , belton , m.j.s . harch , a. , mcewen , a. , johnson , t.v . , helfenstein , p. , davies , m.e . , merline , w.j . , denk , t. 1995 , nature , 374 , 783 cook , a.f . , franklin , f.a . 1970 , icarus , 13 , 282 cook , a.f . 1971 , in physical studies of minor planets , ed . t. gehrels , 155 cox , a.n . , 2000 , in allen s astrophysical quantities , ( new york : springer - verlag ) , 341 davis , d.r . , durda , d.d . , marzari , f. , campo b.a . , gil - hutton , r. 2002 , in asteroids iii , ed . w.f . bottke jr . , a. cellino , p. paolicchi , & r. p. binzel ( tucson : univ . arizona press ) , 545 dunlap , j. l. , gehrels , t. 1969 , aj , 74 , 796 urech , j. , grav , t. , jedicke , r. , denneau , l. , & kaasalainen , m. 2005 , earth moon and planets , 97 , 179 farinella , p. , paolicchi , p. , tedesco , e.f . , zappal , v. 1981 , icarus , 46 , 114 farinella , p. , zappal , v. 1997 , advances in space research , 19 , 181 fernndez , y.r . , sheppard , s.s . , jewitt , d.c . 2003 , aj , 126 , 1563 funato , y. , makino , j. , hut , p. , kokubo , e. , kinoshita , d. 2004 , nature , 427 , 518 goldreich , p. , lithwick , y. , sari , r. 2002 , nature , 420 , 643 harris , a.w . , lagerkvist , c.i . , zappal , v. , warner , b.d . minor planet lightcurve parameters , jan 15 , 2006 . hartmann , w.k . , cruikshank , d.p . 1978 , icarus , 36 , 353 hartmann , w.k . , binzel , r. p. , tholen , d.j . , cruikshank , d.p . , goguen , j. 1988 , icarus , 73 , 487 holsapple , k. a. 2004 , icarus , 172 , 272 ivezic , z. , et al . 2001 , aj , 122 , 2749 jewitt , d.c . , trujillo , c.a . , & luu , j.x . 2000 , , 120 , 1140 jewitt , d.c . , 2003 , earth , moon , and planets , 92 , 465 jewitt , d. c. , sheppard , s. , porco , c. 2004 , in jupiter : the planet , satellites and magnetosphere . bagenal , f. , dowling , t.e . , mckinnon , w.b . vol . 1 ( cambridge university press ) , 263 kaasalainen , m. , torppa , j. , muinonen , k. 2001 , icarus , 153 , 37 kaasalainen , m. , torppa , j. , piironen , j. 2002 , a&a , 383 , l19 lacerda , p. , jewitt , d. 2007 , aj,133 , 1393 landolt , a.u . 1992 , aj , 104 , 340 leone , g. , paolicchi , p. , farinella , p. , zappali , v. 1984 , a&a , 140 , 265 marchis , f. , et al . 2006a nature , 439 , 565 marchis , f. wong , m. h. , berthier , j. , descamps , p. , hestroffer , d. , vachier , f. , le mignant , d. , keck observatory , w.m . , de pater , i. 2006b , , 8732 , 1 marzari , f. , farinella , p. , davis , d.r . , scholl , h. , campo b.a . 1997 , icarus , 125 , 39 merline , w.j . , et al . 2001 , , 7741 , 2 morbidelli , a. , levison , h. f. , tsiganis , k. , gomes , r. 2005 , nature , 435 , 462 pravec , p. , harris , a.w . , michalowski , t. 2002 , in asteroids iii , ed . w.f . bottke jr . , a. cellino , p. paolicchi , & r. p. binzel ( tucson : univ . arizona press ) , 113 richardson , d.c . , & walsh , k.j . 2006 , annual review of earth and planetary sciences , 34 , 47 russell , h.n . 1916 , aj , 43 , 173 stellingwerf , r.f . 1978 . apj , 224 , 953 sheppard , s. & jewitt , d.c . 2004 , aj , 127 , 3023 takahashi , s. , & ip , w. h. 2004 , , 56 , 1099 weidenschilling , s. j. 1980 , icarus , 44 , 807 weidenschilling , s. j. 1989 , in asteroids ii , ed . r. p. binzel , t. gehrels , m.s . matthews ( tucson : univ . arizona press ) , 643 weidenschilling , s.j . 2002 , icarus , 160 , 212 wisniewski , w.z . , michalowski , t.m . , harris , a.w . , mcmillan , r.s . 1997 , icarus , 126 , 395 yoshida , f. , & nakamura , t. 2005 , , 130 , 2900

rank - three tensor model may be regarded as theory of dynamical fuzzy spaces , because a fuzzy space is defined by a three - index coefficient of the product between functions on it , . in this paper , this previous proposal is applied to dynamical generation of commutative nonassociative fuzzy spaces . it is numerically shown that fuzzy flat torus and fuzzy spheres of various dimensions are classical solutions of the rank - three tensor model . since these solutions are obtained for the same coupling constants of the tensor model , the cosmological constant and the dimensions are not fundamental but can be regarded as dynamical quantities . the symmetry of the model under the general linear transformation can be identified with a fuzzy analog of the general coordinate transformation symmetry in general relativity . this symmetry of the tensor model is broken at the classical solutions . this feature may make the model to be a concrete finite setting for applying the old idea of obtaining gravity as nambu - goldstone fields of the spontaneous breaking of the local translational symmetry . = 17.5pt plus 0.2pt minus 0.1pt # 1([#1 ] ) # 1#2#3#4#5#6 \ { # 1&#2&#3 + # 4&#5&#6 } # 1#2#3#4#5#6 ( # 1&#2&#3 + # 4&#5&#6 )

hep-th0606066

the tensor model is an interesting generalization of the matrix model . while the matrix model gives an analytical method to describe the two - dimensional simplicial quantum gravity @xcite , the tensor model was proposed to describe the simplicial quantum gravity in higher dimensions and dynamical lattice topological field theories @xcite . feynman diagrams of the models are assumed to correspond to dual diagrams of simplicial complexes . in the matrix model , the topological expansion of the two - dimensional manifolds can be realized in the large @xmath1 expansion , and this fact is essentially used in the computation . unfortunately , this kind of expansion has not been realized in the tensor model so far , and it seems hard to use the tensor model as an analytical tool to study quantum gravity . another interpretation of rank - three tensor model was proposed by the present author in @xcite . a fuzzy space can be characterized by a rank - three tensor @xmath2 , which defines the algebraic relation @xmath3 between functions @xmath4 on a fuzzy space . therefore dynamical fuzzy spaces may be described by theory which contains the rank - three tensor as a dynamical variable . a fuzzy space will be obtained as a classical solution to the equation of motion of a tensor model . unlike the original interpretation in the preceding paragraph , it will be easier to deduce physical results from the tensor model under this new interpretation . an important advantage of the tensor model as theory of spacetime is the existence of the symmetry under a fuzzy analog of the general coordinate transformation in general relativity . as discussed in @xcite , the general linear transformation on the rank - three tensor can be naturally identified with it . around the classical solutions , the symmetry will be broken and will be non - linearly realized , which is actually the same situation for the general coordinate transformation on a geometric background . there is an old idea that gravity may be obtained as the nambu - goldstone fields of the spontaneous breaking of the local translational symmetry @xcite . therefore the tensor model may provide a concrete finite model for the old idea . the notion of noncommutative coordinates @xcite-@xcite is a natural application of the quantum procedure to spacetime . in fact , noncommutative field theories are derived as effective field theories on d - branes in string theory with a background @xmath5 field @xcite and of three - dimensional quantum gravity coupled with matter @xcite . on the other hand , nonassociativity is also known to appear in open string theory with a non - constant background @xmath5 field @xcite . it was also argued that the algebra of closed string field theory should be commutative nonassociative @xcite . therefore , though its relation to quantization of spacetime is not clear , it would be fair to say that nonassociativity is also another physically sensible structure of spacetime . moreover , it is generally much easier to obtain commutative nonassociative fuzzy spaces of physical interest @xcite than noncommutative ones which seem to generally require a kind of symplectic structure . in addition , quantum field theory on commutative nonassociative spacetimes seems to be able to respect the principles in physics more faithfully @xcite than noncommutative quantum field theory , which is known to have some unusual properties such as the uv / ir mixing @xcite and the violation of causality @xcite and unitarity @xcite . the main purpose of this paper is to show the existence of some elementary commutative nonassociative fuzzy spaces as classical solutions of the rank - three tensor model . this paper is organized as follows . in the following section , the rank - three tensor model is defined . a constraint is imposed on its dynamical variable to simplify the original model in @xcite . this constraint corresponds to dealing with commutative nonassociative fuzzy spaces . in section [ simpleexample ] , a simple example of a commutative nonassociative space this space is not a solution to the tensor model , but provides a good reference to understand the qualitative features of the numerical solutions . in section [ numericalsolutions ] , the numerical solutions representing fuzzy flat torus and fuzzy spheres of various dimensions are obtained and checked . the final section is devoted to summary and discussions . the original model proposed in @xcite contains a symmetric tensor @xmath6 and a rank - three tensor @xmath7 as real dynamical variables . without any constraints on the variable @xmath7 , there are a lot of possibilities of constructing actions , and the analysis to find solutions to equation of motion becomes quite complicated . to simplify the model , a constraint will be imposed on the variable @xmath7 . to allow fuzzy spaces to become good approximations to ordinary spaces , i assume such a constraint be also satisfied by ordinary spaces . the algebra of functions on an ordinary continuum space is commutative and associative . let @xmath8 denote a complete set of independent functions on such a space . a rank - three tensor defines the product between two functions , f_a * f_b = c_ab^c f_c . the commutativity implies [ commutativity ] c_ab^c = c_ba^c , while the associativity implies [ associativity ] c_ab^d c_dc^e = c_ad^e c_bc^d . let me define a tensor _ ab ^dc_bd^c , which is obviously symmetric for the indices . then let me consider c_abc^d _ dc . by using the associativity , the right - hand side can be rewritten , and c_abc = c_ad^f c_be^d c_cf^e . this shows that the tensor @xmath7 is symmetric under the cyclic permutations of the indices . therefore , from the commutativity , @xmath7 is totally symmetric under the permutations of the indices . if @xmath9 is assumed to be not singular , @xmath6 can be identified with @xmath10 . thus the constraint to be imposed on @xmath7 is the total symmetry for the indices , [ constraintc ] c_abc = c_bca = c_cab = c_bac = c_acb = c_cba . conversely , if one starts with @xmath7 with the constraint , the constraint is not enough to show the associativity , and the algebra generated by @xmath11 will be a commutative nonassociative ( or associative ) algebra in general . under the constraint , the variety of action is drastically simplified , compared with the original proposal @xcite . in the quadratic order of @xmath7 , there is only one possibility , s_0=c_abcc^abc . in the quartic order , there are only two possibilities , s_1&=&c_abcc^dbcc_defc^aef , + s_2&=&c_abcc^ad_e c^b_dfc^cef . the total action is given with three coupling constants @xmath12 by [ totals ] s=- s_0+s_1 -4 s_2 , which is graphically shown in fig.[fig : s012 ] . and a line @xmath6 . ] this action is invariant under the @xmath13 transformation , [ ctrans ] c_abc&=&m_a^a m_b^bm_c^cc_abc , g^ab&= & m^-1_a^am^-1_b^b g^ab , where @xmath14 is the total number of functions on a fuzzy space and @xmath15 . as discussed in @xcite , the transformation can be identified with a fuzzy analog of the general coordinate transformation . the equation of motion for @xmath7 from @xmath16 in is given by [ eom ] -g_0 c_abc+3-g_2 c_ad^ec_be^fc_cf^d=0 , which is graphically represented in fig.[fig : eom ] . as shown in @xcite , the equation of motion for @xmath6 is automatically satisfied due to the @xmath13 symmetry of the action , provided @xmath6 is not singular and the equation of motion for @xmath7 is satisfied . an interesting case occurs when @xmath17 . in this case , if the rank - three tensor satisfies the associativity condition , the second and third terms in the equation of motion cancel with each other . this can be checked by using the graphical representation of the associativity condition in fig.[fig : associativity ] to reconnect the diagrams in fig.[fig : eom ] . therefore , the presence of the first term in guarantees that the algebraic relation obtained from a classical solution is commutative _ nonassociative_. this is important , because a commutative associative space with a positive ( negative ) definite @xmath6 is trivial in the sense discussed in the following paragraph . therefore the case @xmath17 selects automatically non - trivial commutative nonassociative fuzzy spaces as classical solutions . let me show the triviality of a commutative associative fuzzy space which has a positive ( negative ) definite @xmath6 and a finite number of functions , but is not trivial , because there are an infinite number of functions . ] . by using the @xmath13 transformation , a positive ( negative ) definite @xmath6 can be diagonalized to @xmath18 ( @xmath19 ) . then , because of the commutativity @xmath20 , the associativity implies n_a n_b = n_b n_a , where @xmath21 is a symmetric matrix defined by @xmath22 . this mutual commutativity of @xmath21 implies that these matrices can be simultaneously diagonalized by an orthogonal matrix @xmath23 , which keeps @xmath24 invariant , ^b m_c^c ( n_a)_bc=e(a , b ) _ bc , where @xmath25 denotes the diagonal elements . because of the commutativity @xmath20 , this @xmath26 will transform @xmath7 to a totally diagonal form , c_abc= m_a^am_b^b m_c^c c_abc=e(a ) _ ab _ bc . this shows that a commutative associative fuzzy space which has a positive ( negative ) definite @xmath6 and a finite number of functions is just a collection of independent points . conversely , if @xmath7 and @xmath6 have such diagonal forms , the algebra is obviously commutative associative . let me finally comment on the satiability of the solutions . the equation of motion just implies that the solutions are the stationary points of the action , but not that they are the local minima . the analysis of the fluctuations around the classical solutions is required to study the stability , and one can not expect that all the solutions are locally stable . in the euclidean case ( when @xmath6 of a classical solution is positive definite ) , however , there exists an obvious way to define an action in which the solutions to are always stable . this is given by [ sstable ] s_stable = g_aag_bbg_cc . this action will generate various euclidean classical solutions to on equal footing , i.e. as its local minima , @xmath27 . on the other hand , it does not seem obvious whether a solution must be a stable one to become interesting . one of the motivation behind in the present research is to discuss gravity in terms of fuzzy space . it is well known that the conformal mode in the einstein - hilbert action has a reversed kinetic term and therefore the solutions can not be stable in the usual sense . moreover , when a minkowski case is considered , the stationary points will be interesting in general . therefore the preference of the choice of an action will depend on the problems in mind . in this paper , the interest is only in the classical solutions to the equation of motion , and the exact form of the action is not cared . in this section i will give an analytically treatable example of a commutative nonassociative fuzzy space . although this space is not a solution to and the number of functions on it is infinite , it has the properties very similar to the numerical solutions in the following section . therefore this example will provide a good reference to understand the qualitative features of the numerical solutions . the usual space is characterized by the following algebra of product , e^ip^x_e^iq^x_=e^i(p^+q^ ) x _ , where @xmath28 is a plane wave with momentum @xmath29 on a space coordinated by @xmath30 . i consider only euclidean spaces . in @xcite , the following commutative nonassociative deformation is considered , [ deformprod ] e^ip^x_*e^iq^x_=e^-(p^2+q^2+(p+q)^2 ) e^i(p^+q^ ) x _ , where @xmath31 , and @xmath32 is a positive deformation parameter . to describe the deformed space , the index of @xmath7 and @xmath6 of the tensor model is given by the continuous momentum @xmath29 , and their values are taken as [ examplecg ] c_p_1 p_2 p_3&=&^d(p_1+p_2+p_3 ) e^-(p_1 ^ 2+p_2 ^ 2+p_3 ^ 2 ) , + g^p_1p_2&=&^d(p_1+p_2 ) , where @xmath33 is the dimensions of the space , and the vector indices of momenta are suppressed for notational simplicity . the measure of the momentum integration is defined by @xmath34 . in @xcite , the reality of the functions @xmath4 is assumed , which is the basis of the reality of @xmath7 and @xmath6 . however the expression is not in such a real coordinate , because @xmath35 is a complex function . an appropriate real coordinate can be obtained by considering instead the real functions , @xmath36 and @xmath37 . then the indefinite @xmath6 in is transformed to a positive definite one . for the computation in this section , however , the above complex coordinate is much more convenient , and can be used without confusion . on the other hand , the relation to such a real coordinate must be taken into account in the analysis of fluctuations around classical solutions , as done in @xcite . there is a simple useful tensor characterizing a fuzzy space , [ defofk ] k^a_bc^acdc_bcd . in the present case , putting to , one obtains [ ksimple ] k^p_1_p_2= ( ) ^e^-3p_1 ^ 2 ^d(p_1-p_2 ) . it is interesting to note that the factor depending on the momentum in has the form of the heat kernel of the laplacian . as discussed in @xcite , the heat kernel expansion is a powerful method to obtain the low - momentum effective geometry of a fuzzy space . this method requires an operator on a fuzzy space which corresponds to the heat kernel of the laplacian ( or other elliptic differential operators ) . an indispensable feature of the laplacian is the invariance under the general coordinate transformation . as for a fuzzy space , the properties of @xmath38 is invariant under the @xmath13 transformation , which is a fuzzy analog of the general coordinate transformation in general relativity . therefore , the operator @xmath38 will be the primary candidate to be used in applying this heat kernel method to derive the low - momentum effective geometry of a fuzzy space in general . more concretely , the general procedure to obtain the effective geometric quantities of a fuzzy space can be described as follows . suppose @xmath38 is diagonalizable with positive eigenvalues , which is the case studied in this paper . the definition shows that @xmath39 is symmetric and ( semi-)positive definite . therefore , after diagonalizing @xmath6 to @xmath18 by @xmath13 transformation , @xmath39 can be transformed to a diagonal form with positive ( or vanishing ) diagonal elements by the remaining @xmath40 symmetry which keeps @xmath18 . the part with vanishing eigenvalues should be simply discarded in defining @xmath41 . ] . then let me define an operator @xmath42 depending on a parameter @xmath43 by k(t)^a_b= _ i ( e_i)^a ( e_i)_b , [ defofkt ] where @xmath44 and @xmath45 are the eigenvalues and the eigenvectors of @xmath38 , respectively , @xmath46 is the maximum of the eigenvalues , and @xmath47 are the dual vectors of @xmath45 , ( e_i)_a ( e_j)^a=_ij . the operator @xmath41 will be assumed to give a fuzzy analogue of a heat kernel of an invariant operator in low - momentum . then the effective geometric quantities of a fuzzy space such as local curvatures will be obtained by investigating the approximate asymptotic expansion of @xmath48 for appropriate insertion operators @xmath49 , as was discussed in @xcite . the philosophy behind in the above procedure is that an effective geometry should be defined by a probe . the probe here is a scalar field . since @xmath8 span all the functions on a fuzzy space , a real scalar field will be described by a real rank - one tensor @xmath50 . therefore a candidate of an action of a non - self - interacting scalar field is given by s_scalar=^a ( -k_ab+m_0 ^ 2 g_ab ) ^b , [ sscalar ] where @xmath51 is a constant . this action respects the invariance under the @xmath13 symmetry . the dynamics of the scalar field from the action will determine an effective geometry of a fuzzy space . though the form of a scalar field action is constrained by the @xmath13 symmetry , there still remains large ambiguity , and this may harm the existence of a unique effective geometry . in fact , the @xmath39 in should be replaced with its logarithm for a direct link to . moreover @xmath39 in may be replaced with another rank - two symmetric tensor obtained from @xmath7 and @xmath6 . to justify the existence of a unique effective geometry , the low - momentum limit of the various scalar field actions must have a universal character . this is certainly true for the present example . because of , approaches the standard scalar field action in the low - momentum limit @xmath52 . because of the gaussian nature of , it is obvious that another choice of the rank - two tensor composed of @xmath7 and @xmath6 in will also lead to a similar gaussian form like . in addition , it is numerically shown in the following section that the behavior of @xmath38 has a good fit with @xmath53 . although these evidences give only inconclusive support to the procedure and the existence of a unique low - momentum effective geometry , it can safely be insisted that at least some geometric properties can be incorporated in the commutative nonassociative algebra . this is in sharp contrast with the noncommutative geometry . the matrix algebra itself does not contain any information on geometry . various geometries results solely from the choice of laplacian or matter field actions @xcite . now let me discuss another quantity which is not covariant or generally applicable , but is certainly useful to check the numerical results in the following section . to avoid confusion , let me denote the functions on the fuzzy space by @xmath54 . now let me define d(x)d^dp e^-ipx f_p , where @xmath55 is the usual @xmath56-number function and should not be confused with @xmath57 . this @xmath58 is a fuzzy analog of the delta function with distribution on @xmath59 . the product of @xmath58 and @xmath60 can be written in the form , d(x)*d(y)&= & d^dp_1 d^dp_2 e^-ip_1x - ip_2yf_p_1*f_p_2 + & = & d^dp_1 d^dp_2 d^dq e^-ip_1x - ip_2yc_p_1p_2^q f_q + & = & d^dz l(x - z , y - z ) d(z ) , where [ fnl ] l(x , y)=d^d p_1 d^d p_2 d^dq e^-ip_1 x - ip_2 y c_p_1p_2^q . a straightforward computation for the present case shows [ lsimple ] l(x , y)= e^- ( ( x - y)^2+x^2+y^2 ) . as shown in fig.[fig : shapel ] , for general @xmath61 , this two - parameter function has a peak at the origin @xmath62 and is distributed around it . the distribution is a little bit stretched in the direction @xmath63 . the profile becomes sharper for smaller @xmath32 , and in the associative limit @xmath64 , @xmath65 approaches @xmath66 as expected from the product of two delta functions on an ordinary space . in for @xmath67 and @xmath68 is plotted . ] as above , the existence of distribution of @xmath65 around the origin can be regarded as a sign of the fuzziness of a space . the profile will provide a quantitative measure of the fuzziness . as the distribution becomes sharper , the fuzziness becomes smaller , and a fuzzy space approaches more an ordinary continuum space . in the following section , @xmath65 will be computed for the numerical solutions to check how well the numerical solutions approximate ordinary continuum spaces . let me finally discuss the general coordinate transformation . a general coordinate transformation is a transformation of the coordinate @xmath59 to a general invertible function of @xmath59 , [ xpgx ] x=g(x ) . as discussed in @xcite , since the right - hand side of can be generally expanded as a linear combination of the functions @xmath4 , the analog of the general coordinate transformation on a fuzzy space is given by the general linear transformation of @xmath4 . on an ordinary space , the transformation of @xmath35 can be written in such a linear form as e^ipx= e^ipg(x)=d^dq c(p , q)e^iqx , where [ cpq ] c(p , q)=1(2)^dd^dx e^-iqx+ipg(x ) . therefore , the corresponding linear transformation for the present example is given by [ fptrans ] f_p=d^dqc(p , q)f_q . then the product of the transformed functions is given by f_p*f_q&=&d^dpd^dq c(p , p)c(q , q)f_p*f_q + & = & d^dr n(p , q;r ) f_r , where [ npqr ] n(p , q;r)&=&d^dpd^dq c(p , p)c(q , q)c_pq^r & = & d^dy c(p , y)c(q , r - y)e^-(y^2+(r - y)^2+r^2 ) . from and , in the case of the ordinary flat space , @xmath69 , one can easily show that [ npqreqcpqr ] n(p , q;r)=c(p+q , r ) , and therefore , f_p*f_q = f_p+q , which implies that the product is invariant under the general coordinate transformation for @xmath69 . on the contrary , @xmath70 will depend not only on @xmath71 but generally both on @xmath72 and @xmath73 for @xmath61 , and the product can not be invariant under the general coordinate transformation except the obvious rotational and translational symmetry of . an essential distinction of the @xmath61 case is the existence of the dimensionful parameter @xmath32 . this dimensionful parameter characterizes the length scale of the fuzziness of the space . in another word , a kind of geometric structure is incorporated in the rank - three tensor @xmath7 for @xmath61 , while this is not for the ordinary space @xmath69 . therefore , as the metric tensor in general relativity is covariantly transformed , the rank - three tensor @xmath7 should be transformed under the general coordinate transformation , which is nothing but the transformation . this transformation cancels that of the functions , and the product between the functions becomes invariant under the general coordinate transformation . at the classical solutions , the fuzzy analog of the general coordinate transformation symmetry will be broken except some global remaining symmetries of a fuzzy space . in this section i will numerically obtain some solutions to the equation of motion . by imposing the expected global symmetries , i obtain solutions corresponding to fuzzy flat torus and fuzzy spheres of various dimensions . to obtain the non - trivial solutions in the meaning discussed in section [ themodel ] , i only consider the coupling constants @xmath74 . since rescaling of @xmath7 and the overall numerical factor of the action are irrelevant in the following analysis , i can put [ g012 ] & & g_0=1,&&g_1=g_2=2 , or @xmath75 without loss of generality . in the latter case , however , the action has the form , s=12 g^aatr ( n_an_a ) -14 g^aag^bbtr ( [ n_a , n_b][n_a,n_b ] ) with @xmath76 . if @xmath6 is positive definite , this action is a sum of a positive definite term and a semi - positive definite one , and can not be stationary in the direction of rescaling @xmath77 . in fact , only euclidean spaces are considered in the following analysis , and all the @xmath6 s are positive definite in the real coordinates mentioned in the preceding section . therefore the latter case does not have non - vanishing solutions . thus it is enough to consider only the former case . numerical solutions in this section have been obtained by taking @xmath7 of ordinary spaces with an order - one normalization as an initial value , finding a local minimum of v=(equation of motion)^2 , where the sum is over all the equations of motion , and checking @xmath78 at the minimum within computational accuracy . mathematica 5.0 on a desktop personal computer is used . in general relativity , the cosmological constant and the dimensions must be changed to obtain flat spaces and spheres of various dimensions as classical solutions . on the other hand , in this section , fuzzy flat torus and fuzzy spheres of various dimensions will be obtained as the classical solutions of the rank - three tensor model for the same coupling constants . this implies that , in the rank - three tensor model , the cosmological constant and the dimensions are not fundamental , but should be regarded as dynamical quantities . as in the preceding section , the momentum index will be used . the momenta on a torus with unit radius take integer values . they are cut - off to consider a finite number of dynamical variables . thus the index of the model is assumed to be given by p=(n_1,n_2,,n_d ) , |n_i| , where @xmath81 are integers , and @xmath33 is the dimensions of the torus . the cut - off @xmath82 may be different in each direction , but i consider only a common cut - off for simplicity . a @xmath33-dimensional flat torus has the translational @xmath83 symmetry . under this symmetry , each function @xmath84 will be transformed by f_p = e^i n_i ^i f_p , where @xmath85 are the real parameters of the lie group . this symmetry is realized as @xmath86 symmetry in the real coordinate mentioned in the preceding section , and is the remaining symmetry of the breakdown of the @xmath13 symmetry . imposing the @xmath83 symmetry on solutions , the non - vanishing components are restricted to the momentum conserving ones , @xmath87 and @xmath88 . as discussed in @xcite , the @xmath6 can be fixed to a convenient value by virtue of the @xmath13 symmetry , and here let me take [ gpmp ] g^p ,- p=1 . a numerical solution to with and for @xmath89 and @xmath68 is obtained as [ flatsol ] llll c_-5 , 0 , 5= 0.516893 , & c_-5 , 1 , 4= 0.564918 , & c_-5 , 2 , 3= 0.584882 , & c_-4,0 , 4=0.635328 , + c_-4 , 1 , 3=0.665027 , & c_-4 , 2 , 2=0.674038 , & c_-3 , 0,3=0.711723 , & c_-3 , 1 , 2=0.728883 , + c_-2 , 0 , 2=0.759477 , & c_-2 , 1 , 1=0.76713 , & c_-1 , 0 , 1=0.784838 , & c_0 , 0 , 0=0.790784 , where the other momentum conserving components are obtained by the permutation symmetry and another assumption on solutions , @xmath90 , which is a reflection symmetry of @xmath79 . using the solution , @xmath91 defined in is plotted in fig.[fig : kflat ] . the values can be fit well with @xmath92 . is plotted for the fuzzy @xmath79 solution . the solid line is @xmath93 . ] the analog of the function @xmath65 in for discrete momentum is defined by [ fnldiscrete ] l(x , y)=_p_1,p_2,q e^-ip_1 x - ip_2 y c_p_1p_2^q . this is plotted for the solution in the left figure of fig.[fig : lflat ] . is plotted for the fuzzy @xmath79 with @xmath89 in the left figure , and that with @xmath94 in the right figure . the peaks are well concentrated around the origin , similarly to that of the example fuzzy space in fig.[fig : shapel ] . as @xmath82 is increased , the peak becomes sharper , and the ordinary space will be obtained in the limit @xmath95.,title="fig : " ] is plotted for the fuzzy @xmath79 with @xmath89 in the left figure , and that with @xmath94 in the right figure . the peaks are well concentrated around the origin , similarly to that of the example fuzzy space in fig.[fig : shapel ] . as @xmath82 is increased , the peak becomes sharper , and the ordinary space will be obtained in the limit @xmath95.,title="fig : " ] the profile is very similar to fig.[fig : shapel ] of the example fuzzy space . one can check that , as @xmath82 is increased , the profile becomes sharper as shown in the right figure of fig.[fig : lflat ] for @xmath94 , i.e. the ordinary space will be obtained in the limit @xmath95 . the above kind of solution seems to exist also in higher dimensions . a numerical solution for @xmath96 and @xmath97 is obtained as [ flat2dimsol ] lll c_(2 , -2 ) , ( 0 , 0 ) , ( -2 , 2)= 0.26524 , & c_(2 , -2 ) , ( 0 , 1 ) , ( -2 , 1)=0.31018 , & c_(2 , -2 ) , ( 0 , 2 ) , ( -2 , 0)=0.33439 , + c_(2 , -2 ) , ( -1 , 1 ) , ( -1 , 1)=0.31655 , & c_(2 , -1 ) , ( 0 , 0 ) , ( -2 , 1)=0.35363 , & c_(2 , -1 ) , ( 0 , 1 ) , ( -2 , 0)=0.37304 , + c_(2 , -1 ) , ( 0 , 2 ) , ( -2 , -1)=0.36306 , & c_(2 , -1 ) , ( -1 , -1 ) , ( -1 , 2)=0.35707 , & c_(2 , -1 ) , ( -1 , 0 ) , ( -1 , 1)=0.37615 , + c_(2 , 0 ) , ( 0 , 0 ) , ( -2 , 0)=0.38327 , & c_(2 , 0 ) , ( 0 , 1 ) , ( -2 , -1)=0.37304 , & c_(2 , 0 ) , ( -1 , -1 ) , ( -1 , 1)=0.39036 , + c_(2 , 0 ) , ( -1 , 0 ) , ( -1 , 0)=0.39675 , & c_(2 , 1 ) , ( 0 , 1 ) , ( -2 , -2)=0.31018 , & c_(1 , -1 ) , ( 0 , 0 ) , ( -1 , 1)=0.41777 , + c_(1 , -1 ) , ( 0 , 1 ) , ( -1 , 0)=0.42284 , & c_(1 , 0 ) , ( 1 , 1 ) , ( -2 , -1)=0.37615 , & c_(1 , 0 ) , ( 0 , 0 ) , ( -1 , 0)=0.43698 , + c_(0 , 0 ) , ( 0 , 0 ) , ( 0 , 0)=0.45174 , & & where all the other momentum conserving components are obtained by the symmetry and another assumption on solutions , @xmath98 . this assumption will be valid when the torus is a flat 2-torus @xmath80 , and the two @xmath79 s are equivalent and symmetric under a reflection . in fig.[fig : kflat2dim ] and fig.[fig : lflat2dim ] , @xmath99 and @xmath100 are plotted , respectively . the features of the solution again agrees well with those of the example fuzzy space in the preceding section . it has been checked that the shape of the peak becomes sharper as @xmath82 is increased . therefore the ordinary continuum space will be obtained in the @xmath95 limit . is plotted for the fuzzy @xmath80 solution . the vertices of the mesh are the values at the integer points @xmath101 . ] ( left ) and @xmath102 ( right ) are plotted for the fuzzy @xmath80 solution . the qualitative behavior agrees with .,title="fig : " ] ( left ) and @xmath102 ( right ) are plotted for the fuzzy @xmath80 solution . the qualitative behavior agrees with .,title="fig : " ] in the ordinary coordinate description of spacetime , the number of the coordinates can not be a dynamical quantity . in the tensor model , however , there exist flat torus solutions of various dimensions for the same coupling constants . therefore the dimensions can be regarded as a dynamical quantity in the tensor model . in this section , i will show the existence of fuzzy spheres as classical solutions of the tensor model for the same coupling constants . this shows that the cosmological constant is also a dynamical quantity in the tensor model . a complete set of independent functions on ordinary @xmath103 is given by the set of spherical harmonic functions . the harmonic functions can be classified by the index @xmath105 , where @xmath106 is an integer spin of the @xmath107 rotation symmetry of @xmath103 and @xmath108 is a spin component in a direction , @xmath109 . therefore , in this subsection , the index of the tensor model is assumed to be a=(j , m ) , a cut - off . the rank - three tensor of an @xmath107 invariant solution can be assumed to have a form , [ defcjm ] c_(j_1,m_1),(j_2,m_2),(j_3,m_3)=a_j_1,j_2,j_3 ( ccc j_1 & j_2 & j_3 + m_1 & m_2 & m_3 ) , where @xmath110 is real , its indices must satisfy the triangle inequality @xmath111 for non - vanishing components , and @xmath112 is the @xmath113-symbol @xcite . the ordinary spherical harmonics have the product relation such that @xmath110 is non - vanishing only when @xmath114 is an even integer . therefore i also assume a_j_1,j_2,j_3=0 for j_1+j_2+j_3=odd . then the permutation symmetry and the permutation property of the @xmath113-symbol imposes a_j_1,j_2,j_3=a_j_(1),j_(2),j_(3 ) for any permutation @xmath115 . the rank - two tensor is taken as [ defgjm ] g^(j_1,m_1 ) , ( j_2,m_2)=_j_1j_2_m_1+m_2,0 ( -1)^j_1+m_1 , which is invariant under the @xmath107 symmetry . as shown explicitly in @xcite , the expressions and in momentum coordinate can be transformed to real valued components in a real coordinate . by using some properties of @xmath113-symbols @xcite such as _ m_1,m_2 ( ccc j_1 & j_2 & j_3 + m_1 & m_2 & m_3 ) ( ccc j_1 & j_2 & j_3 + m_1 & m_2 & m_3 ) = _ j_3j_3 _ m_3m_3 and _ m_i ( -1)^_i j_i+m_i ( ccc j_1 & j_2 & j_3 + m_1 & -m_2 & m_3 ) ( ccc j_2 & j_3 & j_1 + m_2 & -m_3 & m_1 ) ( ccc j_3 & j_1 & j_2 + m_3 & -m_1 & m_2 ) = \ { ccc j_1 & j_2 & j_3 + j_1 & j_2 & j_3 } , where \ { : : : } is the @xmath116-symbol , the equation of motion can be rewritten as [ eombya ] & & -g_0 a_j_1,j_2,j_3 + & & -g_2 _ j_4,j_5,j_6 \ { ccc j_1 & j_2 & j_3 + j_4 & j_5 & j_6 } a_j_1,j_5,j_6 a_j_2,j_4,j_6 a_j_3,j_4,j_5=0 , where the values of the couplings are given by . a numerical solution to for @xmath89 is obtained as [ sols2 ] llll a_0 , 0 , 0=0.375888 , & a_0 , 1 , 1=0.642392 , & a _ 1 , 1 , 2=0.80089 , & a_0 , 2 , 2=0.893485 , + a_2 , 2 , 2=0.92406 , & a_1 , 2 , 3=0.865991 , & a_0 , 3 , 3=0.887106 , & a_2 , 3 , 3=1.03293 , + a_2 , 2 , 4=1.11042 , & a_1 , 3 , 4=1.08588 , & a_3 , 3 , 4=0.947859 , & a_0 , 4 , 4=1.15568 , + a_2 , 4 , 4=1.03207 , & a_4 , 4 , 4=1.17188 , & a_2 , 3 , 5=1.08615 , & a_1 , 4 , 5=1.0445 , + a_3 , 4 , 5=1.07613 , & a_0 , 5 , 5=1.09768 , & a_2 , 5 , 5=1.09539 , & a_4 , 5 , 5=1.10454 . using the properties of the @xmath113-symbol , the quantity can be rewritten as [ ks2witha ] k^(j , m)_(j,m)&=&_j_1,j_2,m_1,m_2 c^(j , m),(j_1,m_1),(j_2,m_2)c_(j,m),(j_1,m_1),(j_2,m_2 ) & = & ^j_j^m_m 12j+1 _ j_1,j_2 a_j , j_1,j_2 ^ 2 , which basically depends only on @xmath106 due to the @xmath107 symmetry . the @xmath117 for the solution are plotted in fig.[fig : props2 ] . is plotted for the fuzzy @xmath103 solution . the horizontal axis is @xmath106 . the solid line is @xmath118 $ ] ] the values can be fit well by @xmath92 . in the following , let me compare the product relation obtained from the solution and that of the spherical harmonics . one needs to take care of the phase choice of the spherical harmonics . the choice compatible with is such that the spherical harmonics satisfy y_j , m^*()=(-1)^j+m y_j , m ( ) under the complex conjugation . for these spherical harmonics , the rank - three tensor defining the product is given by & & c_(j_1,m_1),(j_2,m_2),(j_3,m_3)= d y_j_1,m_1()y_j_2,m_2()y_j_3,m_3 ( ) & & = ( -1)^(j_1+j_2+j_3)/2 ^12 ( ccc j_1 & j_2 & j_3 + 0&0&0 ) ( ccc j_1 & j_2 & j_3 + m_1&m_2&m_3 ) . therefore the quantity of the ordinary @xmath103 corresponding to @xmath110 is given by [ ats2 ] a_j_1,j_2,j_3 = ( -1)^(j_1+j_2+j_3)/2 ^12 ( ccc j_1 & j_2 & j_3 + 0&0&0 ) . in the left figure of fig.[fig : aats2 ] , the solution @xmath110 in and @xmath119 of the ordinary sphere in are compared . of the fuzzy @xmath103 solution with @xmath89 in ( bold line ) and @xmath120 of the ordinary @xmath103 in ( thin line ) are plotted . the values are ordered in the same order as , i.e. smaller to larger spins . the discrete points are joined by lines for clear view . the solution has a tendency to take smaller values at larger spins . in the right figure , the same for a solution with @xmath94 . comparing the regions below 20 in each figure , it can be observed that a better approximation to the ordinary @xmath103 is obtained for @xmath94 than @xmath89.,title="fig : " ] of the fuzzy @xmath103 solution with @xmath89 in ( bold line ) and @xmath120 of the ordinary @xmath103 in ( thin line ) are plotted . the values are ordered in the same order as , i.e. smaller to larger spins . the discrete points are joined by lines for clear view . the solution has a tendency to take smaller values at larger spins . in the right figure , the same for a solution with @xmath94 . comparing the regions below 20 in each figure , it can be observed that a better approximation to the ordinary @xmath103 is obtained for @xmath94 than @xmath89.,title="fig : " ] a similar diagram is shown for the solution with @xmath94 in the right figure of fig.[fig : aats2 ] . one can see that the solutions approximate well the ordinary product between spherical harmonics . the approximation becomes better for @xmath94 than @xmath89 . therefore , the solution can be identified with a fuzzy @xmath103 , and the ordinary continuum @xmath103 will be obtained in the limit @xmath121 . an analogous quantity for a sphere corresponding to can be defined by l(,)_j_i , m_i c_(j_1,-m_1),(j_2,-m_2),(j_3,-m_3)y_j_1,m_1(_0 ) y_j_2,m_2 ( ) y_j_3,m_3( ) , where @xmath122 denotes a reference point on @xmath103 , and one of the delta functions is located there . in fig.[fig : ls2 ] , @xmath123 for the @xmath89 solution is plotted . the profile is consistent with the fuzzy product between two delta functions at @xmath122 . for the fuzzy @xmath103 solution is plotted . the horizontal axis is the angle between @xmath122 and @xmath124 . the values are well concentrated around the origin , that is consistent with the product between two fuzzy delta functions at @xmath122 . ] let me finally discuss a fuzzy @xmath104 solution . the spherical harmonics on @xmath104 are given by the complete set of representation matrices of @xmath125 , @xmath126 , @xmath127 with @xmath128 @xcite . these wigner @xmath33-functions satisfy @xcite [ intddd ] dg d^j_1_m_1n_1(g ) d^j_2_m_2n_2(g ) d^j_3_m_3n_3(g ) =( ccc j_1 & j_2 & j_3 + m_1 & m_2 & m_3 ) ( ccc j_1 & j_2 & j_3 + n_1 & n_2 & n_3 ) , where the structure of the combination of two @xmath113-symbols comes from the invariance under the two @xmath125 s of @xmath129 . therefore i assume the solution to have the form must be common in the two @xmath113-symbols for a solution to represent an @xmath104 . if they were allowed to be different , a solution would represent @xmath130 instead of @xmath104 . ] , c_(j_1,m_1,n_1),(j_2,m_2,n_2),(j_3,m_3,n_3)=b_j_1,j_2,j_3 ( ccc j_1 & j_2 & j_3 + m_1 & m_2 & m_3 ) ( ccc j_1 & j_2 & j_3 + n_1 & n_2 & n_3 ) , where @xmath131 is assumed to be real and symmetric under any permutation of the indices . the rank - two tensor is taken as g^(j_1,m_1,n_1),(j_2,m_2,n_2)=_j_1,j_2_m_1+m_2,0 _ n_1+n_2,0 ( -1)^m_1-n_1 , which is just obtained by multiplying and its inverse . this choice is consistent with the property of the wigner @xmath33-functions , [ dnormalization ] dg d^j_1_m_1n_1(g ) d^j_2_m_2n_2 ( g ) = up to the factor @xmath132 , which will be considered in due course . because of these double - structures , the equation of motion in @xmath131 can be just obtained by taking the squares of the coefficients in , [ eombyat ] & & -g_0 b_j_1,j_2,j_3 + & & -g_2 _ j_4,j_5,j_6 \ { ccc j_1 & j_2 & j_3 + j_4 & j_5 & j_6 } ^2 b_j_1,j_5,j_6 b_j_2,j_4,j_6 b_j_3,j_4,j_5=0 . the spin will be cut - off by @xmath133 , and a numerical solution for @xmath134 is obtained as [ sols3 ] llll b_0 , 0 , 0 = 0.179229 , & b_0 , 1/2 , 1/2 = 0.352866 , & b_1/2 , 1/2 , 1 = 0.516800 , & b_0 , 1 , 1 = 0.661060 , + b_1 , 1 , 1 = 0.784606 , & b_1/2 , 1 , 3/2 = 0.860387 , & b_0 , 3/2 , 3/2 = 0.899459 , & b_1 , 3/2 , 3/2 = 0.600820 , + b_1 , 1 , 2 = 0.819278 , & b_1/2 , 3/2 , 2 = 1.018466 , & b_3/2 , 3/2 , 2 = 1.153431 , & b_0 , 2 , 2 = 1.252915 , + b_1 , 2 , 2 = 0.889080 , & b_2 , 2 , 2 = 1.078355 , & b_1 , 3/2 , 5/2 = 1.150590 , & b_1/2 , 2 , 5/2 = 1.275880 , + b_3/2 , 2 , 5/2 = 1.393015 , & b_0 , 5/2 , 5/2 = 1.419768 , & b_1 , 5/2 , 5/2 = 1.559359 , & b_2 , 5/2 , 5/2 = 1.683407 , + b_3/2 , 3/2 , 3 = 1.428888 , & b_1 , 2 , 3 = 1.497101 , & b_2 , 2 , 3 = 1.687474 , & b_1/2 , 5/2 , 3 = 1.851444 , + b_3/2 , 5/2 , 3 = 1.738983 , & b_5/2 , 5/2 , 3 = 1.964490 , & b_0 , 3 , 3 = 2.013377 , & b_1 , 3 , 3 = 2.241499 , + b_2 , 3 , 3 = 2.368063 , & b_3 , 3 , 3 = 2.867651 . & & for fuzzy @xmath104 is plotted . the horizontal axis is @xmath106 . the solid line is @xmath135 $ ] ] by taking the square of the coefficient in , the @xmath38 is obtained as [ ks3witha ] k^(j , m , n)_(j,m,n ) = ^j_j^m_m ^n_n 1(2j+1)^2 _ j_1,j_2 b_j , j_1,j_2 ^ 2 , which depends essentially only on @xmath106 . the values for the solution are plotted in fig.[fig : ks3 ] . the values can be fit well with @xmath92 . after including the factor appearing in into the normalization of the spherical harmonics , it can be shown from that the quantity of the ordinary @xmath104 corresponding to @xmath136 is given by [ ords3 ] b_j_1,j_2,j_3= ^12 . in fig.[fig : s3comp ] , this @xmath137 is compared with @xmath131 of the solution . a good agreement is observed , and the solution can be identified with a fuzzy @xmath104 . solution , @xmath131 in , ( bold ) and the values of the ordinary @xmath104 , @xmath138 in , ( thin ) are compared . the values are ordered in the same order as , i.e. smaller to larger spins . the discrete points are joined by lines for clear view . the solution has a tendency to take smaller values at larger spins . ] in this paper , the proposal in @xcite that the rank - three tensor model can be used as theory of dynamical fuzzy spaces is applied to dynamical generation of commutative nonassociative fuzzy spaces . it is numerically shown that fuzzy flat torus and fuzzy spheres of various dimensions are classical solutions of the rank - three tensor model . therefore , in the tensor model , the cosmological constant and the dimensions are not fundamental but can be regarded as dynamically generated quantities . it is also observed that , as the number of the degrees of freedom increases , the fuzziness becomes smaller , and the ordinary spaces will be recovered in the infinite limit . in the tensor model , @xmath13 is a fuzzy analog of the general coordinate transformation symmetry in general relativity . there is an old idea that gravity may be regarded as nambu - goldstone fields of a spontaneous broken local translational invariance @xcite in a similar manner as the abelian gauge theory @xcite . since the @xmath13 symmetry is broken at classical solutions except some global symmetries , the rank - three tensor model is a finite concrete model which may realize the idea . however , the applicability of the idea to the tensor model is not clear at present , since the discussions have been for ordinary continuous spacetime . if it is shown to be applicable , the tensor model can be an interesting candidate for quantum gravity . there was a serious unsatisfactory property in the solution of a _ noncommutative _ two - sphere in the tensor model found in the previous paper @xcite . in the discussions of the interpretation of @xmath13 symmetry as a fuzzy analog of the general coordinate transformation @xcite , it is essential that the set of functions @xmath8 span all the scalar functions on a fuzzy space . however , in the previous _ noncommutative _ solution , the scalar field had a double - index structure @xmath139 , and could not be identified with @xmath8 . on the contrary , the solutions found in this paper are _ commutative nonassociative _ deformations of the algebraic relations of the functions on the ordinary spaces with truncation , and the @xmath13 transformation can be naturally identified with a fuzzy analog of the general coordinate transformation . the interpretation of the rank - three tensor model as a model of fuzzy spaces has some advantages over the original interpretation of tensor models @xcite that feynman diagrams correspond to dual diagrams of simplicial complexes . in the original interpretation , the rank of tensor is related to the dimensions and must be changed to discuss quantum gravity in different dimensions . on the contrary , in the present interpretation , the rank is kept three , but various dimensions are dynamically induced . in addition , in the original interpretation , a kind of topological expansion is required to obtain physical results from tensor models . however , such an expansion has not been found for tensor models . on the other hand , in the present interpretation , the physical meaning is definite . moreover the quantum fluctuations around a classical solution are controllable in principle by taking a semi - classical limit , and several quantities will be computable . there is a worry about the locality of the model . the behavior of @xmath65 of the classical solutions shows that the peak is well localized around the origin , i.e. the classical solutions are local . however , the fluctuations around the classical solutions obviously contain various non - local modes . it seems inevitable that , since there exists no metric at the beginning , a pregeometric system like the tensor model contains modes which are non - local in the sense of the background of classical solutions . therefore the right question would be whether such non - local fluctuations lead to serious contradictions to the principles in physics . there are a number of ways to avoid contradictions like that non - local modes are very heavy or that interactions between local and non - local modes are negligibly small . a systematic analysis of fluctuation modes around classical solutions is required as a future work . the notion of nonassociative space is rather new , compared with noncommutative space . noncommutativity is natural from the viewpoint of quantization , but nonassociativity is rather exceptional in physical systems . however , as explained briefly in the introduction , nonassociativity can be also a physically sensible structure of spacetime . in fact , as shown in fig.[fig : nonassociative ] , the two figures for closed string field theory and a fuzzy space , respectively , have some similarities , and the intuitive argument on closed string field theory in @xcite seems to also hold for a fuzzy space . , since @xmath140 . similar nonassociativity seems to hold for a fuzzy space in the lower figure . ] as discussed in section [ simpleexample ] , an interesting aspect of a nonassociative space is that a kind of metric structure is incorporated in the definition . this would be the most peculiar property distinct from a noncommutative space which seems to be associated with a kind of symplectic structure . the present construction of the tensor model is not unique . although the action is among the simplest , there is no reason to discard the higher order interactions . another ambiguity is the choice of the action of a scalar field on a fuzzy space . as discussed in section [ simpleexample ] , this ambiguity may harm the existence of a unique low - momentum effective geometry defined by the probe of a scalar field . it might turn out to be true that the low - momentum dynamics does not depend on the details of the actions , as the universality in lattice theory . however , although this kind of universality has a chance to hold near some classical solutions , it is hard to believe that the full quantum properties such as transition probabilities between classical solutions do not depend on the details . therefore , even if the tensor model has a chance to provide an interesting model of quantum gravity by itself , it seems very important to find its principle in relation with other systematic approaches of quantum gravity or string theory . the author was supported in part by the grant - in - aid for scientific research no.13135213 , no.16540244 and no.18340061 from the ministry of education , science , sports and culture of japan . * note added in proof : * i was informed that systematic construction of fuzzy spheres in diverse dimensions using quantized nambu brackets , which is distinct from the formulation of this paper , was proposed , and their features were discussed in a series of papers @xcite . j. ambjorn , b. durhuus and t. jonsson , `` three - dimensional simplicial quantum gravity and generalized matrix models , '' mod . lett . a * 6 * , 1133 ( 1991 ) . n. sasakura , `` tensor model for gravity and orientability of manifold , '' mod . lett . a * 6 * , 2613 ( 1991 ) . n. godfrey and m. gross , `` simplicial quantum gravity in more than two - dimensions , '' phys . d * 43 * , 1749 ( 1991 ) . d. v. boulatov , `` a model of three - dimensional lattice gravity , '' mod . lett . a * 7 * , 1629 ( 1992 ) [ arxiv : hep - th/9202074 ] . h. ooguri , `` topological lattice models in four - dimensions , '' mod . phys . lett . a * 7 * , 2799 ( 1992 ) [ arxiv : hep - th/9205090 ] . r. de pietri , l. freidel , k. krasnov and c. rovelli , `` barrett - crane model from a boulatov - ooguri field theory over a homogeneous space , '' nucl . b * 574 * , 785 ( 2000 ) [ arxiv : hep - th/9907154 ] . r. de pietri and c. petronio , `` feynman diagrams of generalized matrix models and the associated manifolds in dimension 4 , '' j. math . phys . * 41 * , 6671 ( 2000 ) [ arxiv : gr - qc/0004045 ] . l. freidel , d. oriti and j. ryan , `` a group field theory for 3d quantum gravity coupled to a scalar field , '' arxiv : gr - qc/0506067 . n. sasakura , `` an invariant approach to dynamical fuzzy spaces with a three - index variable , '' mod . lett . a * 21 * , 1017 ( 2006 ) [ arxiv : hep - th/0506192 ] . a. b. borisov and v. i. ogievetsky , `` theory of dynamical affine and conformal symmetries as gravity theory of the gravitational field , '' theor . * 21 * , 1179 ( 1975 ) [ teor . mat . * 21 * , 329 ( 1974 ) ] . for example , see n. boulanger and i. kirsch , `` a higgs mechanism for gravity . ii : higher spin connections , '' phys . rev . d * 73 * , 124023 ( 2006 ) [ arxiv : hep - th/0602225 ] , for recent developments and further references . s. snyder , `` quantized space - time , '' phys . rev . * 71 * , 38 ( 1947 ) . c. n. yang , `` on quantized space - time , '' phys . rev . * 72 * , 874 ( 1947 ) . a. connes and j. lott , `` particle models and noncommutative geometry ( expanded version ) , '' nucl . . * 18b * , 29 ( 1991 ) . s. doplicher , k. fredenhagen and j. e. roberts , `` the quantum structure of space - time at the planck scale and quantum fields , '' commun . math . phys . * 172 * , 187 ( 1995 ) [ arxiv : hep - th/0303037 ] . a. connes,``noncommutative geometry"(academic press , 1994 ) . s. ramgoolam , `` on spherical harmonics for fuzzy spheres in diverse dimensions , '' nucl . b * 610 * , 461 ( 2001 ) [ arxiv : hep - th/0105006 ] . de medeiros and s. ramgoolam , `` non - associative gauge theory and higher spin interactions , '' jhep * 0503 * , 072 ( 2005 ) [ arxiv : hep - th/0412027 ] . y. sasai and n. sasakura , `` one - loop unitarity of scalar field theories on poincare invariant commutative nonassociative spacetimes , '' arxiv : hep - th/0604194 . t. filk , `` divergencies in a field theory on quantum space , '' phys . b * 376 * , 53 ( 1996 ) . s. minwalla , m. van raamsdonk and n. seiberg , `` noncommutative perturbative dynamics , '' jhep * 0002 * , 020 ( 2000 ) [ arxiv : hep - th/9912072 ] . n. sasakura , `` an invariant approach to dynamical fuzzy spaces with a three - index variable - euclidean models , '' to appear in the proceedings of 4th international symposium on quantum theory and symmetries ( qts-4 ) and 6th international workshop on lie theory and its applications in physics ( lt-6 ) , varna , bulgaria , 15 - 21 aug 2005 [ arxiv : hep - th/0511154 ] . n. sasakura , `` heat kernel coefficients for compact fuzzy spaces , '' jhep * 0412 * , 009 ( 2004 ) [ arxiv : hep - th/0411029 ] . n. sasakura , `` effective local geometric quantities in fuzzy spaces from heat kernel expansions , '' jhep * 0503 * , 015 ( 2005 ) [ arxiv : hep - th/0502129 ] . as a review , see for example , a. p. balachandran , s. kurkcuoglu and s. vaidya , `` lectures on fuzzy and fuzzy susy physics , '' arxiv : hep - th/0511114 . albert messiah , `` quantum mechanics , '' amsterdam , holland : north - holland ( 1961 - 62 ) . d. a. varshalovich , a. n. moskalev and v. k. khersonsky , `` quantum theory of angular momentum , '' singapore , singapore : world scientific ( 1988 ) . for example , see j. s. dowker , `` spherical universe topology and the casimir effect , '' class . . grav . * 21 * , 4247 ( 2004 ) [ arxiv : hep - th/0404093 ] , and the references therein . r. ferrari and l. e. picasso , `` spontaneous breakdown in quantum electrodynamics , '' nucl . b * 31 * ( 1971 ) 316 . r. a. brandt and w. c. ng , `` gauge invariance and mass , '' phys . d * 10 * , 4198 ( 1974 ) . a. b. hammou , m. lagraa and m. m. sheikh - jabbari , `` coherent state induced star - product on r(lambda)**3 and the fuzzy sphere , '' phys . d * 66 * , 025025 ( 2002 ) [ arxiv : hep - th/0110291 ] . + m. m. sheikh - jabbari , tiny graviton matrix theory : dlcq of iib plane - wave string theory , a jhep * 0409 * , 017 ( 2004 ) [ arxiv : hep - th/0406214 ] . + m. m. sheikh - jabbari and m. torabian , `` classification of all 1/2 bps solutions of the tiny graviton matrix theory , '' jhep * 0504 * , 001 ( 2005 ) [ arxiv : hep - th/0501001 ] . + m. m. sheikh - jabbari , `` inherent holography in fuzzy spaces and an n - tropic approach to the cosmological constant problem , '' arxiv : hep - th/0605110 .

deep radio observations at 1.4 ghz for the extended chandra deep field south were performed in june through september of 2007 and presented in a first data release ( miller et al . 2008 ) . the survey was made using six separate pointings of the very large array ( vla ) with over 40 hours of observation per pointing . in the current paper , we improve on the data reduction to produce a second data release ( dr2 ) mosaic image . this dr2 image covers an area of about a third of a square degree and reaches a best rms sensitivity of 6jy and has a typical sensitivity of 7.4jy per 28 by 16 beam . we also present a more comprehensive catalog , including sources down to peak flux densities of five or more times the local rms noise along with information on source sizes and relevant pointing data . we discuss in some detail the consideration of whether sources are resolved under the complication of a radio image created as a mosaic of separate pointings each suffering some degree of bandwidth smearing , and the accurate evaluation of the flux densities of such sources . finally , the radio morphologies and optical / near - ir counterpart identifications ( bonzini et al . 2012 ) are used to identify 17 likely multiple - component sources and arrive at a catalog of 883 radio sources , which is roughly double the number of sources contained in the first data release .

1301.7004

the _ chandra _ deep field south ( cdf - s ) continues to be one of the most important deep fields for multiwavelength investigation of the cosmological evolution of galaxies . originally devoted 1 msec of _ chandra _ integration @xcite , subsequent observations have pushed ever deeper in x - ray flux by increasing the net exposure first to 2 msec @xcite and recently to 4 msec @xcite . this makes the cdf - s the most sensitive x - ray view of the universe , reaching sources with full - band ( 0.5 - 8.0 kev ) fluxes down to @xmath1 erg s@xmath2 @xmath3 . the area covered by relatively deep x - ray observations has also increased over the initial survey , with an additional one msec divided among four _ chandra _ pointings to produce the extended cdf - s ( e - cdf - s ; * ? ? ? * ) . likewise , _ xmm - newton _ has observed the field in increments over the past nine years and has now accumulated over 3 msec of integration in a region that covers much of the e - cdf - s . these data will provide quality x - ray spectra of detected sources ( e.g. , * ? ? ? * ) . in addition to x - ray coverage , _ hubble _ and _ spitzer _ have produced deep near - ultraviolet through infrared data for the field . the deepest images ever obtained with _ hubble _ are located in the cdf - s , with the _ hubble _ ultra deep field using the advanced camera for surveys ( acs ) to obtain deep @xmath4 ( @xmath5 ) , @xmath6 ( @xmath7 ) , @xmath8 ( @xmath9 ) , and @xmath10 ( @xmath11 ) images @xcite and more recently the wide field camera 3 ( wfc3 ) instrument exploring the near - ir @xmath12 ( @xmath13 ) , @xmath14 ( @xmath15 ) , and @xmath16 ( @xmath17 ) with a 192-orbit `` treasury '' program ( pi g. illingworth ) . the great observatories origins deep survey ( goods ; * ? ? ? * ) targets the cdf - s as one of its two fields , using both _ hubble _ and _ spitzer _ images to study galaxy evolution across cosmic time . analogous to the widening of the cdf - s to the e - cdf - s in x - ray surveys , programs have expanded the covered area using _ hubble _ ( galaxy evolution from morphologies and seds , gems ; * ? ? ? * ) and _ spitzer _ ( the _ spitzer _ irac / musyc public legacy survey , simple , at near - ir wavelengths and the far - infrared deep extragalactic legacy survey , fidel , in mid - ir ; * ? ? ? * ; * ? ? ? most recently , the `` cosmic assembly near - ir deep extragalactic legacy survey '' ( candels ) is capitalizing on _ hubble _ s wfc3 camera to provide deep near - ir imaging of several fields including the cdf - s @xcite . in terms of allocated orbits , candels is the largest project in the history of _ hubble _ and its three near - ir filters will identify and characterize galaxies over the approximate redshift range @xmath18 . _ spitzer _ observations continue with warm mission _ activity ( e.g. , the _ spitzer _ extragalactic representative volume survey , or servs ; * ? ? ? * ) . finally , _ herschel _ has pushed the deep ir data into the far - ir portion of the spectrum with the goods-_herschel _ program @xcite and the _ herschel _ multitiered extragalactic survey ( hermes ; * ? ? ? * ) . ground - based observatories have also supplied much ancillary data at optical and near - ir wavelengths . these programs include @xmath19 imaging @xcite with the infrared spectrometer and array camera ( isaac ) of the european southern observatory s very large telescope ( vlt ) and @xmath20 imaging from the garching - bonn deep survey @xcite using the wide field imager ( wfi ) at the 2.2 m la silla telescope . the e - cdf - s is also one of four fields covered by the multiwavelength survey by yale - chile ( musyc ; * ? ? ? * ) , with this program using the cerro tololo inter - american observatory ( ctio ) 4 m telescope and the mosaic ii and ispi instruments to obtain deep coverage of the e - cdf - s region at @xmath21 . the musyc project has also acquired medium - band imaging in 18 filters using the subaru telescope and used these along with the broad - band and _ spitzer _ irac photometry to produce accurate photometric redshifts for tens of thousands of galaxies @xcite . in addition to photometric redshifts , there are a host of spectroscopic programs providing critical spectroscopic redshifts and source classifications . these have often been associated with the goods program and capitalized on the vlt with the fors2 and vimos instruments ( * ? ? ? * ; * ? ? ? * and associated papers ) . many other spectroscopic campaigns have also targeted the field , including specific targeting of radio - selected sources ( e.g. , * ? ? ? * ) and recent efforts to expand the area from the well - sampled goods region ( essentially the cdf - s ) to the full 30 by 30 e - cdf - s ( e.g. , * ? ? ? * using the inamori - magellan areal camera and spectrograph , imacs , on the magellan - baade telescope ) . the field has also received extensive attention at radio wavelengths . @xcite used the australia telescope compact array ( atca ) at 1.4 ghz to observe a large ( 3.7 deg@xmath22 ) area , with the e - cdf - s region reaching down to an rms sensitivity of @xmath23jy per @xmath24 beam . * hereafter k08 ) used the national radio astronomy observatory ( nrao ) very large array ( vla ) to obtain deeper and higher resolution 1.4 ghz data ( @xmath25jy rms noise per @xmath26 beam ) in a field centered on the cdf - s . these latter data were exploited in a sequence of papers on optical counterparts @xcite , radio - x - ray properties @xcite , and source populations and evolution @xcite . in addition to 1.4 ghz , vla observations at 5 ghz are presented in k08 with an rms sensitivity of @xmath25jy rms for a @xmath27 beam and @xcite present an atca 5.5 ghz survey of the e - cdf - s with a @xmath28jy rms per @xmath29 beam . at longer wavelengths , @xcite included giant metre - wave radio telescope data at 610 mhz ( @xmath30jy rms noise for a @xmath31 beam ) in their investigation of evolution in the fir - radio correlation . in june through september 2007 , we observed the e - cdf - s with the national radio astronomy observatory ( nrao ) very large array ( vla ) under program code am889 . the observations were performed at a frequency of 1.4 ghz and consisted of over 250 hours of time ( about 690 ksec time on source ) spread across six separate pointings . the survey description , strategy , and initial image and catalog were presented in @xcite . the rapid turnaround time between data collection and publication was motivated by the strong community interest in the field , which has been justified by the number of studies which have relied upon the data to set or confirm the astrometry for observations made at other wavelengths ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) , identify counterparts to sub - millimeter surveys @xcite , extend the far - infrared / radio correlation out to cosmological redshifts @xcite , and identify and study star - forming galaxies and active galactic nuclei @xcite . naturally , this rapid release of the calibrated radio images necessitated some minor compromises . in regards to the imaging of the data , these amounted to postponing some time- and computational - intensive techniques that produce slight improvements to the depth and cosmetics of the resulting images . similarly , the initial catalog was conservative in only going to a 7@xmath32 point - source detection limit and providing little detail about source morphology . in the current paper , we present the second data release ( dr2 ) associated with this program . it incorporates these finer imaging techniques yielding a typical reduction in the rms noise of about 0.5 @xmath0jy across the full e - cdf - s area , and this improvement plus moving to a 5@xmath32 detection threshold produces a deeper and more comprehensive but still highly - reliable source catalog . we describe the details associated with the improvements to the imaging in section [ sec - image ] . in section [ sec - cat ] we discuss the detection and characterization of sources applied in the construction of a source catalog , and briefly discuss the catalog and future directions in section [ sec - sum ] . to fix our terminology for the ensuing discussion , we provide a brief summary of the observations . in order to cover the full e - cdf - s area at near - uniform sensitivity , we pointed the vla at six separate coordinate locations arranged in a hexagonal grid around the adopted center of the cdf - s , ( j2000 ) @xmath33 @xmath34 . we refer to these as `` pointings '' and often reference them with a numerical designation from 1 to 6 starting due east of the center coordinate and proceeding clockwise ; table [ tbl - ptgs ] includes the numerical designations and coordinate centers . our observations were spread over many days on account of the low declination of the field and typically amounted to five hours of time per calendar date . thus , the full program consists of about fifty of these `` tracks . '' we opted to observe a single pointing on any given calendar date , meaning that each pointing was observed on at least eight separate dates . while this was done largely for efficiency and convenience , it also allows for a deep investigation of possible radio transient populations ( frail et al . , in preparation ) . we refer those interested in the survey strategy and design to ( * ? ? ? * hereafter m08 ) . at the time of the observations the vla was undergoing improvements paving the way toward full `` expanded '' vla @xcite operations , a now complete process that includes the official renaming as the karl g. jansky very large array . in general , these modifications to the array had little effect , and the procedures for calibrating and imaging the data followed the standard prescriptions for deep , multi - channel continuum vla observations at 1.4 ghz using nrao s astronomical image processing system ( aips ) . the most notable exception was the retirement of the original vla control computers on 2007 june 27 , which led to slight errors in how the @xmath35 coordinates were written at various times during the range of dates covering our observations . for the first data release , we relied upon the aips task uvfix to recalculate the @xmath35 data based on the antenna positions which are included among the data tables associated with each observation . several of the observations were also afflicted by a glitch which reversed the channel indexing for brief periods of time , and we fixed this error using a sequence of existing aips procedures . finally , in a few instances the most - recently retrofitted evla antenna had incorrect system temperature values which produced spuriously large weights that we had to manually adjust down to more representative values . each of these problems has now been corrected in the nrao data archive , and in this second pass at data calibration and imaging we essentially started `` from scratch . '' we re - obtained the raw @xmath35 data from the archive , and followed the same general procedure of data editing , calibration , and imaging as was described in m08 . one minor improvement was the inclusion of a very small amount of extra data relative to that previous reduction . as vla antennas are retrofit to become evla antennas , they are placed on a `` master pad '' for testing . there were some times during the course of our observations when such antennas were included in the array and we were able to calibrate such data and include it in our current imaging . although we mainly started from scratch and paralleled the previous reduction procedure , we did have one significant saving : the presence of first data release deep images associated with each pointing . these were used to self - calibrate the @xmath35 data after the initial data edits and amplitude and phase calibration . the basic sequence of initial steps is identical to that presented in m08 . for each observing track , we inspected and edited the calibrator @xmath35 data ( 3c48 and j0340 - 213 ) . these data were then used to establish the bandpass calibration ( using 3c48 ) , flux density scale ( from 3c48 , with j0340 - 213 bootstrapped to 3c48 ) , and phase calibration ( using j0340 - 213 ) for the target data . after application of the calibrations , the @xmath35 data for the target were inspected and edited to remove obvious errors . the resulting @xmath35 data were then self - calibrated on phase using the deep individual pointing images created for the first data release . we proceeded to then image the data , subtract the determined clean components , edit the resulting source - removed @xmath35 file to flag obvious interference and data errors , and then return the clean components . once all the data associated with an individual pointing were thus calibrated and edited , we combined them into single , deep @xmath35 data set associated with that pointing . these combined single - pointing datasets were then imaged and subjected to further self - calibration and editing . first , we applied a strong taper to the @xmath35 data and generated a wide - field , low resolution map . over this image we laid the grid of 127 facets arranged in a `` flys - eye '' pattern that would be used in our imaging , with these smaller facets allowing for correction of image distortions caused by sky curvature . each of these facets was @xmath36 05 pixels , and the flys - eye thus covered about 52 in diameter not quite out to the first null of the vla primary beam at 1.4 ghz . we were thus able to search the wide - field image for additional faint sources outside of our principal imaged area , including a small number within the primary beam and a larger number within the sidelobes . while these same basic steps were performed previously on individual tracks of data , repeating them on the deeper combined datasets warranted an increase in the radial coverage of individual facets ( previously 91 were used ) and revealed additional faint sources outside the flys - eye coverage . we identified about 30 such additional small fields to image for each pointing . finally , we searched within the main 127-facet coverage for particularly bright sources ( those with flux densities uncorrected for primary beam response of a few mjy and higher ) . this identified between seven and 10 additional small facets for imaging , and thus in total each pointing was imaged as the combination of between about 160 and 175 facets . successive rounds of imaging and self - calibration were then performed , with each newer set of images used as the source model for self - calibration in both amplitude and phase . our final steps of self - calibration on the full data for each pointing used the aips task peelr , which `` peels '' off individual bright sources to improve the overall calibration . these sources are returned to the data once all corrections are made . between one and four bright sources per field were subjected to peelr , with these sources typically having an apparent flux density ( uncorrected for primary beam attenuation ) of about 7.5 mjy and greater . the final imaging was done in segments of the full @xmath35 data . this provides a fine correction to several effects , each of which relates to the shape of the primary beam and how it can differ slightly for different portions of the data . these minor differences cause the response to sources to vary over the course of an observing track . first , the vla observed in a pair of frequency bands ( intermediate frequencies , or ifs ) that bracket 1.4 ghz . these separate ifs imply slightly different resolutions . second , the feeds for the right and left circular polarization are not coincident and thus have slightly different pointing centers . third , the power pattern of the vla primary beam is not perfectly circular and over the course of observations as the hour angles of sources change their responses vary . we consequently separated the data into 12 segments , consisting of three ranges in hour angle and separately for each if and polarization . as before , we imaged each segment of the data and performed a round of self - calibration on that segment before producing the final image for that segment . the images corresponding to the 12 segments for each pointing were then combined using variance weighting based on their individual rms sensitivities . relative to the first data release , the final images for the six pointings had net improvements in sensitivity ranging from about @xmath37jy to @xmath38jy . table [ tbl - ptgs ] includes the achieved rms sensitivity per pointing . [ [ sec-34 @xmath3934 ( 0.324 square degrees ) , being @xmath40 05 square pixels . this is the main image associated with this second data release and is shown in figure [ fig - image ] . it will be used in subsequent characterization and compilation of a source catalog . we note that the actual coverage of the individual pointings extends well beyond the boundaries of this mosaic image but experiences the drop - off in sensitivity associated with the power pattern of the vla primary beam . users interested in sources that fall within the coverage of such pointings but outside the final mosaic image are directed to the individual pointing images and a larger mosaic image that incorporates all of the data . the sensitivity and coverage thereof was evaluated via the construction of an rms sensitivity map . to construct this map , we took the final mosaic image and removed all sources with peak flux densities greater than 150 @xmath0jy , chosen as approximately 20@xmath32 point source detections . the rms noise at each pixel in the mosaic map was then determined from this bright - source - removed map , based on the value calculated within a 135-diameter circle centered on that pixel . this `` background mesh '' size is considerably larger than some other radio surveys which suggest that sizes on the order of just ten beam widths across are sufficient ( * ? ? ? * ; * ? ? ? * e.g. , ) . we found that , given the resolution of our survey and the existence of extended sources that approached the size of these smaller background mesh apertures , their usage produced inaccurate small - scale variations in rms sensitivity maps . the larger size that we used was also consistent with what was used in the first data release , although in that release the background mesh was square but of the same equivalent area . the evaluation of the rms noise in constructing the rms sensitivity map was iterative , with points differing by more than three times the calculated rms noise removed prior to re - calculation until convergence was achieved . the final sensitivity map , hereafter called the rms map , is relatively smooth as seen by the overlaid contours of constant rms included in figure [ fig - image ] . figure [ fig - rms ] depicts the area covered at a given sensitivity or better based on this rms map . the most sensitive regions of the mosaic image have an rms noise of 6 @xmath0jy per beam , and half of the full mosaic image has an rms noise of 7.4 @xmath0jy per beam or better . this represents an improvement of about 0.5 @xmath0jy over the first data release mosaic image . the regions at the center of the mosaic where all six individual pointings can contribute are the most sensitive , and these correspond to the cdf - s coverage . the cdf - s 4 msec x - ray data correspond to 54 separate _ chandra _ images collected using acis - i and thus each having a field size of about 169 by 169 . the exposure - weighted center of these data is ( j2000 ) @xmath41 @xmath42 @xcite , and for the range of roll angles and slight shifts in aim point of the separate exposures it is reasonable to assume approximately uniform coverage out to about 7@xmath43 in radius from this central point . within this restricted area of the deepest x - ray coverage ( 0.043 square degrees , or 13% of the mosaic image ) the radio mosaic image has a typical rms noise of 6.3 @xmath0jy and is never worse than 6.7 @xmath0jy . this is depicted with the dotted line in figure [ fig - rms ] . we also define a subset of the full mosaic over which there are enough ancillary multiwavelength data to provide counterpart identification and spectroscopic redshifts or reasonable photometric redshifts ( padovani et al . , in preparation ; this region amounts to 0.282 square degrees or 87% of the mosaic image ) . the rms sensitivity as a function of area for this restricted region is depicted by the dashed line in figure [ fig - rms ] , and is used in correcting the source counts for investigation of contributions by source type . our data release consists of the six @xmath35 data sets , nine images , and one catalog . the @xmath35 data are the final combined and edited collections for each of the six pointings , and allow interested users to perform imaging with their own choices of weighting parameters . the nine images correspond to the final mosaic image and its associated rms map , the final images corresponding to each of the six individual pointings ( each with an rms near 10 @xmath0jy at the field center ) , and a large mosaic map made from these six individual pointing images and representing the full imaging coverage ( i.e. , out past the coverage of the main final mosaic image ) . the catalog is based entirely on the final mosaic map , and will be discussed now in greater detail . the basis for the detection of sources was a signal - to - noise ratio ( @xmath44 ) map , constructed by dividing the final mosaic image by its corresponding rms map . the aips task sad was used to pick out all sources in this @xmath44 map with peaks above above 3 , and thus three times the local noise . we note that the effect of using an rms map generated with a larger background mesh size is to slightly increase the number of sources above the detection threshold by smoothing over small - scale variation produced by the presence of bright sources and the more rapid fall - off in sensitivity at the edges of the mosaic . sad fits gaussians to the sources above the 3@xmath32 threshold and subtracts them from the @xmath44 map to produce a residual image which may then be inspected to find sources that were either missed or poorly fit . these include sources with extended radio morphologies as well as blends of multiple and possibly independent radio sources . the residual image was inspected by eye with the help of overlaid contours to reveal such missed or poorly fit sources ( indicated by large positive or negative values ) and these were manually added to the list for subsequent analysis . we then proceeded to evaluate all sources with peaks above 4@xmath32 identified in the @xmath44 image by either sad or our manual inspection . our initial choice of 3@xmath32 was based on prior knowledge that sometimes single faint sources might be incorrectly split into a pair of fainter sources , and we used the sad list to recover these objects as multiple @xmath45 objects at nearly identical position . in this round of source cataloging , we used the aips task jmfit to fit gaussians to the sources on the final mosaic image in order to evaluate source peak and integrated flux densities . jmfit was provided with a guess for the peak flux density and its position ( based on manual inspection of the image ) , but no corrections for primary beam attenuation or bandwidth smearing were applied . a correction for the former was explicitly applied in construction of the mosaic image , while the latter is complicated by the lack of a single pointing center for the image . the output parameters of jmfit included the source coordinates , major and minor axis size and position angle , and the peak and integrated flux density . we used the rms map to determine the local noise at the coordinates of the fitted source , and accepted those sources with peak flux density greater than five times their local noise value for inclusion in the final catalog . for those sources poorly fit by gaussians in the initial investigation using the @xmath44 map , we directed jmfit to fit multiple components where applicable and performed aperture photometry on the more extended sources using the tvstat task . for some work correlating the radio catalog with other wavelengths , it might seem of value to release the initial automated catalog that extended to down 3@xmath32 however , we deem the danger of misinterpretation of such low - significance sources outweighs the possible utility of such a catalog . in addition to some issues already discussed ( i.e. , false decomposition of single faint sources ) , we note here that if the mosaic image were truly described by gaussian noise there would be over 1,100 noise peaks greater than 3@xmath32 across its large area . this number becomes reasonable at 4@xmath32 , where about 26 noise peaks would be expected . using the traditional @xmath46 detection threshold the number of anticipated false sources has dropped to less than one across the full area of the final mosaic . the true noise distribution is not perfectly gaussian as a result of the presence of real sources and their sidelobes , so these numbers may be considered lower limits to the estimate of the number of false sources at these thresholds . finally , we note that the availability of the actual mosaic image allows users to directly assess possible low - significance radio sources at their positions of interest . source morphology is an important aspect of source catalogs , often providing a coarse way to discriminate between emission mechanisms at appropriate resolutions . radio emission associated with star formation is generally extended across galaxy disks , and thus resolved at arcsecond scales even for sources at high redshifts @xcite . evaluation of the true flux density of a source is also often dependent on source morphology , with compact sources at lower signal - to - noise ratio better described by their peak flux density than by their integrated flux density ( e.g. , see * ? ? ? unfortunately , the careful consideration of source morphology is complicated by the mosaic nature of the final image . every location in the output mosaic is the combination of the contributions of up to six separate pointings , and thus up to six separate corrections each dependent on differing values of angle and radial distance . if these corrections do not alter the apparent morphology of a source but only its apparent flux density , they are easily applied this is the case for primary beam attenuation , and the simplification that the power pattern is radially symmetric is applied and accounted in the creation of the mosaic map . however , other effects such as bandwidth smearing are not so easily handled . a given location in the output mosaic is the combination of up to six separate images each of which suffers some degree of smearing in the radial direction relative to its relevant pointing center , and while corrections for this effect may be made on sources extracted from single - pointing data they are not so easily achieved when multiple pointings are combined . bandwidth smearing does not change the integrated flux density in the limit of a noiseless image , and in m08 we confirmed that for the e - cdf - s radio survey the standard gaussian fitting for source integrated flux densities provided consistent results with aperture photometry . however , the combination of up to six separate bandwidth smearing corrections does affect source morphology and the determination of whether a source is resolved . in light of this complication , we performed multiple tests of source morphology that are encapsulated in the source catalog . the simplest of these is the direct fit to the mosaic image , which produces a major axis , minor axis , and position angle of the gaussian which best fits the source in the mosaic image . in addition to ignoring effects such as bandwidth smearing , these results are also still the convolution of the primary beam and the intrinsic morphology of the source . as such , they are not tremendously useful except in the very general sense of providing a quick handle on which sources are clearly resolved ( i.e. , those with axial sizes well above the 28 by 16 beam size ) . similarly , sources in the mosaic image which were poorly fit were identified in the residual maps produced as the difference between the mosaic image and the best gaussian fits to sources . the presence of large residuals is strongly suggestive of an extended source , and the integrated flux density for these sources was evaluated using manually - sized aperture photometry with the task tvstat . for single pointing data we can incorporate the effect of bandwidth smearing into the source fitting . the distances between each source and the six separate pointing centers was determined , and when such distances were less than the 33% power point of the primary beam these pointings were noted . this provides a listing of the pointings which directly contributed to the mosaic image at the location of each source . we then determined the rms sensitivity of each contributing pointing at the coordinates of each source in order to determine which individual pointing provided the best signal - to - noise for source measurement . this is usually the pointing nearest to the source , but slight differences in achieved depth per pointing did occasionally mean a more distant pointing provided a marginally lower rms sensitivity . we then evaluated the gaussian fit to each source using just this `` best '' single pointing , incorporating the effects of both primary beam attenuation and bandwidth smearing . the parameters of these fits , deconvolved to remove the primary beam geometry , are included in the source catalog . the major and minor axis parameters are provided as @xmath47 ranges , so that sources with fitted axes differing from zero may be identified and thus provide an evaluation that a source is resolved ( i.e. , the minimum size for its major axis is greater than zero ) . this represents the first of our two tests on whether a source is resolved . a second technique to evaluate whether a source is resolved involves comparison of the fitted integrated and peak flux densities . a comparison of the ratio of integrated to peak flux density as a function of signal - to - noise ratio ( defined as the peak flux density divided by the local rms sensitivity ) is an empirical estimate of the errors in source fitting , and can be used to identify resolved sources as those for which the integrated flux density is clearly in excess of the peak flux density after accounting for the fitting errors ( e.g. , * ? ? ? although this method does not provide direct indicators of source sizes and morphologies , it is quite helpful in the determination of the better estimate of a source s true flux density . in figure [ fig - resck ] we plot the ratio of the integrated to peak flux densities as a function of signal - to - noise based on the gaussian fits to sources in the mosaic image . situations where the integrated flux density is less than the peak flux density are clearly the result of errors inherent to source fitting , and we can determine a function which bounds these cases . under the assumption that it is equally likely that such errors can result in fits for which the integrated flux density is greater than the peak flux density , we can mirror this function above the line where the peak flux density equals the integrated flux density and thus define an envelope which contains sources that are consistent with being unresolved . sources for which the ratio of integrated to peak flux density is greater than this envelope are those that are likely to be resolved . we follow the functional form presented in @xcite to fit the lower envelope of the distribution in figure [ fig - resck ] as @xmath48 where @xmath49 and @xmath50 represent the integrated and peak flux densities and @xmath44 is the signal - to - noise ratio , defined as the peak flux density divided by the local rms sensitivity . we determined the constant , 200 , by finding its value such that 95% of all sources with @xmath51 are bound by the function . it is clear from figure [ fig - resck ] that the distribution of the ratio of peak to integrated flux density is not centered on unity . this is the aforementioned effect of combining separate pointings each with their own degrees of bandwidth smearing and the creation of the mosaic image not explicitly accounting for this bandwidth smearing . the resulting source will appear to have a core representing the region for which each pointing contributes to the flux density and a much fainter irregular extended halo consisting of the bandwidth - smeared flux densities of the separate pointings . the net effect is that sources with high signal - to - noise will always have an integrated flux density greater than their peak flux density , although due to the mosaic nature of the map such sources are not unambiguously resolved . we can model the expected size of this effect using reasonable assumptions . the spatial distribution of vla antennas and our weighting of the resulting @xmath35 data imply that our @xmath35 coverage is approximately a circularly - symmetric gaussian , and the individual 3.125 mhz channels after bandpass calibration can be assumed to have nearly square response . under these simplifications , the reduction in measured amplitude of a point source caused by bandwidth smearing relative to its true amplitude is : @xmath52 @xcite , where @xmath53 is the amplitude of the point source in the absence of bandwidth smearing , @xmath54 , and @xmath55 . this latter term is the fractional bandwidth of the observations ( in our case , 3.125 mhz divided by 1.4 ghz ) times the distance from the pointing center expressed in number of beams ( the separation @xmath56 divided by the fwhm of the beam , @xmath57 ) . for simplicity in our estimation of the effect of bandwidth smearing we `` circularize '' the beam and assume it is symmetric with a fwhm of 21 . we can then apply equation [ eqn - bwsmear ] to determine the reduction in peak response associated with each pointing that contributes to the mosaic image at the position of each source , and weight these per - pointing responses in the same manner that was used to generate the mosaic image . this yields the reduction in peak response caused by bandwidth smearing , and since in the absence of noise the net flux density is conserved , the inverse of the left - hand side of equation [ eqn - bwsmear ] is the ratio of the integrated flux density to the peak flux density : the ratio examined in figure [ fig - resck ] . we plot a histogram of the modeled effect of bandwidth smearing in figure [ fig - bwsmear ] , which is based on the positions of actual sources in our mosaic image . the results are consistent with the offset seen in figure [ fig - resck ] , with the mean ratio of @xmath58 to @xmath59 being @xmath60 with a dispersion of @xmath61 . for perspective , the range in values is understood by considering limiting cases . the least possible bandwidth smearing occurs for a source coincident with a single pointing center , for which the mosaic map is dominated by data associated with that pointing but also includes contributions from its two adjacent pointings . in this case , the ratio of @xmath58 to @xmath59 is 1.06 . at the other extreme , a source directly in the center of the mosaic image is equidistant from all six pointings and thus suffers modest bandwidth smearing from each . for such a source , the ratio of @xmath58 to @xmath59 is 1.14 . it can be seen in figure [ fig - bwsmear ] that these values do indeed bracket the majority of sources and the peak in the histogram around @xmath62 indicates that , by design , most sources fall reasonably close to a single pointing center . the small number of sources at @xmath63 correspond to the corners of the mosaic image . returning to figure [ fig - resck ] and the evaluation of whether a source is resolved , instead of simply mirroring equation [ eqn - rescklow ] about @xmath64 we include an offset to account for bandwidth smearing . based on the spread of values for @xmath65 from bandwidth smearing shown in figure [ fig - bwsmear ] and its consistency with that measured for sources that appear unresolved based on the gaussian source fitting to their single best pointing , we set the upper envelope for evaluating whether a source is resolved at @xmath66 sources for which @xmath65 is above this signal - to - noise based threshhold are thus considered to be resolved using this approach . this is indicated by a flag in the source catalog , and additional justification for the 20% offset in the upper envelope is provided in section [ sec - flux ] when we compare our flux densities with those of single - pointing data . the points in figure [ fig - resck ] are coded to reflect whether individual sources appear to be resolved under each of the two techniques . grey asterisks represent all sources for which both fitted axes have minimum values of zero and the ratio of peak to integrated flux density lies below the curve described by equation [ eqn - resckhigh ] ; these points were associated with unresolved sources in each technique . similarly , open circles represent those sources that appear to be resolved when evaluated using both techniques . the filled black symbols represent the less conclusive cases , with filled triangles signifying sources that appear resolved in their best single - pointing data yet do not lie above the fitted relationship of equation [ eqn - resckhigh ] and filled circles signifying sources that appear unresolved in their single - pointing data yet lie above the fitted relationship thus suggesting that they are resolved . powerful radio galaxies are often resolved into multiple components , with the classic example being frii sources typically composed of a pair of bright , extended radio lobes surrounding the progenitor galaxy which sometimes is also coincident with a distinct radio core @xcite . the sensitivity and resolution of the survey can also `` break up '' the contiguous radio emission associated with a single progenitor galaxy into multiple apparently separate components . possible associations are generally made by the experienced radio astronomers who collect , reduce , and analyze the radio data and are thus inherently subjective , although usually accurate . in m08 , we relied upon the listing of identified multiple component sources from k08 and added new possible components that the slightly higher resolution of our survey provided . in the present work , we take advantage of the outstanding available multiwavelength coverage to assess multiple component sources . @xcite use a likelihood ratio technique to identify counterparts for the radio sources associated with this survey , relying on deep optical , near- and mid - infrared data . this is highly effective at assessing possible multiple component sources , as sometimes radio sources thought to be components associated with a nearby but spatially - separate galaxy are shown to have much stronger associations with a unique , fainter coincident galaxy . similarly , the outstanding depth of the optical and ir data make radio sources without a coincident counterpart very rare . this means that close positional coincidence of a radio source with an optical or ir counterpart is almost certainly an association with a single galaxy rather than a component associated to a neighboring galaxy . our radio catalog is thus guided by the counterpart identification . if a single galaxy has multiple associated radio components , we list their combined properties as a single entry in the main catalog and provide a flag indicating that the source consists of multiple components . a subsequent catalog then includes the information on each separate component of the combined source . there are 17 such multiple - component sources , with a total of 49 individual components ascribed to them . we stress that no information is lost by this approach ; every distinct peak in the radio emission of the final image is included in one of the two catalogs , and users interested in the total radio emission associated with a given galaxy will find such information in the first catalog . images of the 17 multiple - component sources are provided in the appendix . of greater practical importance to most users is the question of what flux density should be used for each source ? as previously discussed , this is related to the issue of source morphology even if such detail is not needed by the user . in m08 we confirmed the general consistency of our flux density scale with prior radio observations of this region , particularly the deep single - pointing data of k08 . we also confirmed that the flux densities based on gaussian fitting were consistent with simple aperture photometry , indicating that although considerations such as bandwidth smearing do affect the apparent morphology of sources the standard approach of using gaussian fits to evaluate flux densities was still appropriate . we rely on the same basic approach here to specify which flux density measurement , peak or integrated , should be adopted in order to most accurately reflect the true flux density of a source . this was accomplished by calculating the mean and dispersion of the ratio of flux density from our catalog with that of k08 , using various prescriptions to choose between peak and integrated flux density measurements . in general , these prescriptions paralleled the discussion of source morphology and the two evaluations thereof , one based on whether the fitted deconvolved axial sizes were consistent with zero and one based on the ratio of integrated to peak flux density and the parametrized envelope bounding unresolved sources . these will henceforth be referred to as `` fitted axis '' and `` envelope , '' respectively . for resolved sources we used the integrated flux density measurement while for unresolved sources we used the peak flux density measurement . we also explored the use of a threshold in @xmath44 , using integrated flux densities for sources above the threshold and peak flux densities for those below . combinations of these various prescriptions were also examined and table [ tbl - flxck ] summarizes the findings . the flux densities associated with the fitted axis method are slightly lower than those reported in k08 ( @xmath67 ) , while the envelope method yields slightly higher flux densities ( @xmath68 ) . after performing a single round of 3@xmath32 clipping to remove outliers ( for example , sources with intrinsic variability over the @xmath696 years between the k08 observations and this survey ) , the dispersions associated with the two methods are similar at @xmath70 and @xmath71 , respectively . at this point we note that adding the 20% offset to the upper curve in the envelope method and using this to evaluate whether sources are resolved and hence whether to adopt their integrated flux density measurements has a large effect . simply mirroring the envelope determined for @xmath51 without including the offset produces flux densities that are about 7% higher than those in k08 . this is just a consequence of the convergence to @xmath72 for brighter unresolved sources ( see figures [ fig - resck ] and [ fig - bwsmear ] ) . since the two methods have opposing effects relative to the k08 flux density scale , we investigated logical combinations of `` or '' ( if either method indicated a source was resolved , its integrated flux density was adopted ) and `` and '' ( a source was assumed to be resolved only if both methods agreed that it was ) . neither of these produced an improvement on simply adopting one of the two methods , although the dispersion with the `` and '' combination after clipping outliers was the lowest we found . these findings led us to identify a `` hybrid '' solution . the `` hybrid '' method applies a cut of @xmath73 , with all sources detected at greater than this threshold being represented by their integrated flux density measurement . for sources with lower @xmath44 , the peak flux density measurement is adopted unless both the fitted axis and envelope methods indicate the source is resolved in which case its integrated flux density is used . this prescription yields flux densities nearly identical to those of the k08 survey , @xmath74 and @xmath75 before and after clipping . the dispersion in the ratio of the flux densities after the clipping is also very low . we recommend this hybrid solution for the selection of peak or integrated flux density , and indicate its selection in the final source table . the final source catalog is presented in tables [ tbl - cat ] and [ tbl - multiple ] and encapsulates all of the above discussion . table [ tbl - cat ] is the main catalog of radio sources , wherein the 17 sources thought to consist of multiple components associated with a single host object are listed with a single aggregate integrated flux density . gaussian fits to the individual components associated with these sources are listed in full detail in table [ tbl - multiple ] . the data columns for each table are identical , and are summarized as follows : \(1 ) - ( 6 ) _ source position _ ( j2000 ) . in most cases , the listed position is that of the center of the gaussian that best fits the data . the typical position errors for strong , unresolved sources observed with the vla at 1.4 ghz are 01 , and position errors attributed to the gaussian fitting in the presence of noise are on the order of the beam size divided by twice the signal - to - noise ratio @xcite . thus , the position error in right ascension for a 5@xmath32-point source would be @xmath76 while the position error in declination for the same source would be @xmath77 . for some resolved sources , the radio emission is poorly fit by a gaussian and the source is evaluated using an irregularly - shaped aperture that covers the apparent emission ( aips task tvstat ) . these sources are flagged ( column 29 ) and their positions are usually just the location of the maximum emission within the aperture . \(7 ) _ signal - to - noise ratio_. this is defined as the fitted peak flux density divided by the local rms noise ( i.e. , column 8 divided by column 9 ) . \(8 ) _ peak flux density _ in units of @xmath0jy per beam . \(9 ) _ local rms noise _ in units of @xmath0jy per beam . this is evaluated using the rms map and thus represents the local noise within a 135-diameter circle around the source position . \(10 ) - ( 11 ) _ integrated flux density _ and associated error , in units of @xmath0jy . as with the prior columns , these values represent the gaussian fit to the source unless otherwise noted . for the extended sources poorly fit by a gaussian , the total flux density within the irregularly - shaped aperture is indicated and its error is just the square root of the number of beams covering the aperture times the local rms noise . \(12 ) - ( 14 ) _ source size and position angle from mosaic image _ , in units of arcsec and degrees . these represent the fitted source size including convolution with the 28 by 16 beam , with the position angle measured in degrees east from north ( e.g. , the beam has @xmath78 ) . in addition , these parameters include the differing contributions of bandwidth smearing from each of the pointings contributing to the mosaic image at the position of the source . note that these columns are only present in the online version of the tables . \(15 ) _ best pointing_. the index number of the individual pointing which has the lowest rms noise at the position of the source . the pointings are numbered 1 through 6 starting due east of the center of the mosaic and progressing clockwise ( see table [ tbl - ptgs ] ) . \(16 ) _ rms noise in best pointing _ in units of @xmath0jy per beam . \(17 ) - ( 21 ) _ source size and position angle from best pointing _ , in units of arcsec and degrees . the sizes are presented as @xmath47 range for the major axis ( columns 17 and 18 ) and minor axis ( columns 19 and 20 ) , while the position angle ( column 21 ) is just the nominal value . since these values are based on gaussian fitting to the source in the single pointing with the lowest rms , they have accounted for bandwidth smearing and have been deconvolved to remove the synthesized beam . \(22 ) _ extended flag_. if the source was found to be extended using the `` envelope '' method ( see section [ sec - morph ] ) , this column is set to 1 . sources that were unresolved under this test have a value of 0 in this column . \(23 ) _ flux density choice flag_. the recommended flux density measurement to use for the source ( see section [ sec - flux ] ) . if the value in this column is `` p '' the flux density is better represented by the peak flux density ( column 8) whereas if it is `` i '' the integrated flux density ( column 10 ) is recommended . \(24 ) _ pointings contributing to mosaic image_. this is a listing of the pointings that contribute to the mosaic image at the position of the source . thus , if the source position falls within the @xmath79 power point of the primary beam ( about 20 ) associated with an individual pointing that pointing is noted here . this information allows users to inspect the images associated with the individual pointings to further assess source morphology and characteristics . this column is only available in the online version of the tables . \(25 ) - ( 28 ) _ source information from kellermann et al . ( 2008)_. for sources that were identified in k08 , the associated identification number ( column 25 ) , flux density and error ( columns 26 and 27 ) in units of @xmath0jy , and deconvolved source size ( column 28 , including upper limits ) are provided . these columns are only available in the online version of the tables . \(29 ) _ notes_. flags indicating details such as extended source morphologies and other source fitting details . multiple - component sources are indicated here ( table [ tbl - cat ] ) and cross referenced to their individual fitted components ( table [ tbl - multiple ] ) , with figures provided in the appendix . table [ tbl - cat ] provides a listing of 883 distinct radio sources , 17 of which are single entries for multiple - component sources where the total radio emission appears to be associated with a single host galaxy . the details on the individual separate components for these 17 sources form the basis for the second table , which contains a total of 49 separate components for these sources . to summarize , the final mosaic image consists of 915 separate peaks in radio emission ( 883 in table [ tbl - cat ] minus the 17 multiple component sources , plus the 49 individual components of these sources listed in table [ tbl - multiple ] ) . for comparison , the early data release catalog of m08 consisted of 495 unique components . figure [ fig - snrhist ] reveals the primary reason for this greatly increased number , as the @xmath44 limit for inclusion has decreased from 7 to 5 for this catalog relative to the first data release . in addition , the @xmath44 for any given source will usually be increased in the present catalog on account of the improved imaging , which typically amounts to an rms noise that is 0.5 @xmath0jy lower than that of the first data release at any point in the mosaic image . in addition to the catalogs , this second data release consists of 9 images and 6 calibrated @xmath35 datasets . these all may be accessed online . there are individual images and calibrated @xmath35 files for each of the six pointing centers ( refer to table [ tbl - ptgs ] ) , along with the main mosaic image that uses all six pointings and its associated rms image . the final released image is a larger area mosaic that incorporates all data out to the @xmath79 power point of the vla primary beam . while there is no catalog associated with the additional area provided by this larger mosaic , it is useful for identification and measurement of sources found in other wide - area surveys . the depth of the survey makes the surface density of radio sources competitive with that determined in x - rays . for the full mosaic image ( 0.324 square degree ) and its associated 5@xmath32 catalog presented in table [ tbl - cat ] , there are over 2,700 sources per square degree . this is over 30 times the source density associated to the faint images of the radio sky at twenty centimeters survey ( first ; * ? ? ? because the sensitivity is fairly uniform across the mosaic image this figure increases only slightly towards the image center . for the portion of the mosaic image with good multiwavelength coverage to provide photometric redshifts ( refer to section [ sec- ) the source density approaches 2,800 sources per square degree , and for the region of deep and uniform coverage from the cdf - s 4 msec observations ( i.e. , the 7 radius about the exposure - weighted center point ) it is nearly 3,100 sources per square degree . in the central 5 around the center of the mosaic ( i.e. , the center of the pointing ring ) where the rms sensitivity ranges from 6 @xmath0jy per beam to 6.6 @xmath0jy per beam , the density is over 3,500 sources per square degree . for comparison , the x - ray source density within the inner 3 of the 4 msec cdf - s is roughly 16,700 sources per square degree @xcite but over the full cdf - s coverage it is only about twice the surface density of radio sources although the observing time for the radio observations was about a factor of six less than for the x - ray observations . similarly , simple positional matching of x - ray and radio catalogs provides some illuminating numbers . considering just the region of deep and uniform coverage from the cdf - s 4 msec observations , 16.8% ( 80/475 ) of the @xcite x - ray sources have matches in the radio catalog using a simple positional match of 2 . the detection fraction within this same area only rises slightly as the _ chandra _ integration time declines , being 18.2% ( 50/274 ) for the cdf - s 2 msec catalog of @xcite , 19.0% ( 38/200 ) for the cdf - s 1 msec catalog of @xcite , and 21.8% ( 19/87 ) for the shallower e - cdf - s catalog of @xcite . looking at this from the other direction , the increasing integrations with _ chandra _ are gradually producing larger detection fractions of the radio sources . while only 14.5% of the radio sources within this region are detected in the e - cdf - s observations ( 19/131 ) , 29.0% ( 38/131 ) are detected in the 1 msec data , 38.2% ( 50/131 ) are detected in the 2 msec data , and 61.1% ( 80/131 ) are detected in the full 4 msec data . at this rate of increase , should the cdf - s be expanded to 10 msec ( e.g. , * ? ? ? * ) we can expect x - ray detections for over 90% of the radio sources from within this central area . emphasis on the e - cdf - s continues with additional radio surveys either in process or planned . the rebirth of the vla as the karl g. jansky very large array , with new receivers and electronics as well as a wide - bandwidth , many channel correlator , will greatly advance this objective . the rms sensitivity of a radio image is inversely proportional to the square root of the bandwidth , meaning the new correlator produces a dramatic improvement in the depth that vla images can achieve in reasonable duration programs . consequently , future deep fields will routinely achieve rms sensitivities of @xmath80 @xmath0jy at frequencies near 1.4 ghz ( e.g. , * ? ? ? this will translate to thousands of detected radio sources in the e - cdf - s , and the wide bandwidth will also provide spectral indices for brighter sources where signal - to - noise is high . one of the tradeoffs in survey imaging is the size of the primary beam ( and hence field of view ) , which is inversely proportional to the frequency of observation . this means that doubling the frequency of observation requires four times as many pointings to cover the same area , and the attendant increase in program duration has greatly limited the number of wide - field surveys performed at frequencies above 1.4 ghz . the vast improvement in sensitivity arising from increased bandwidth will counteract this problem , and deep surveys at higher observing frequencies will also be achievable in reasonable observing times . these will provide spectral measurements for a large fraction of the sources . nam gratefully acknowledges support through _ chandra _ award ar8 - 9016x , and a jansky fellowship of the national radio astronomy observatory held during the period when these data were obtained . dummy balestra , i. et al . 2010 , , 512 , 12 beckwith , s.v.w . 2006 , , 132 , 1729 biggs , a.d . 2011 , , 413 , 2314 bonzini , m. et al . 2012 , , in press ( arxiv:1209.4176 ) bourne , n. , dunne , l. , ivison , r.j . , maddox , s.j . , dickinson , m. , & frayer , d.t . 2011 , , 410 , 1155 bridle , a.h . , & schwab , f.r . 1994 , in _ synthesis imaging in radio astronomy _ , eds . perley , f.r . schwab , and a.h . bridle , ( san francisco : asp conference series ) cardamone , c.n . 2010 , , 189 , 270 comastri , a. et al . 2011 , , 526 , l9 condon , j.j . 1997 , , 109 , 166 condon , j.j . , cotton , w.d . , fomalont , e.b . , kellermann , k.i . , miller , n. , perley , r.a . , scott , d. , vernstrom , t. , & wall , j.v . 2012 , , 758 , 23 cooper , m.c . 2012 , , 425 , 2116 coppin , k.e.k . 2009 , , 395 , 1905 damen , m. et al . 2011 , , 727 , 1 danielson , a.l.r . , lehmer , b.d . , alexander , d.m . , brandt , w.n . , luo , b. , miller , n. , xue , y.q . , & stott , j.p . 2012 , , 422 , 494 dye , s. et al . 2009 , , 703 , 285 elbaz , d. et al . 2011 , , 533 , 119 fanaroff , b.l . , & riley , j.m . 1974 , , 167 , 31 fiore , f. et al . 2011 , , 537 , 16 gawiser , e. et al . 2006 , , 642 , l13 giacconi , r. et al . 2002 , , 139 , 369 giavalisco , m. et al . 2004 , , 600 , l93 grogin , n.a . 2011 , , 197 , 35 hildebrandt , h. et al . 2006 , , 452 , 1121 huynh , m.t . , jackson , c.a . , norris , r.p . , & prandoni , i. 2005 , , 130 , 1373 hyunh , m.t . , hopkins , a.m. , lenc , e. , mao , m.y . , middelberg , e. , norris , r.p . , & randall , k.e . 2012 , , 426 , 2342 ivison , r.j . 2010 , , 402 , 245 koekemoer , a.m. et al . 2011 , , 197 , 36 lehmer , b.d . 2005 , , 161 , 21 lehmer , b.d . 2012 , , 752 , 46 luo , b. et al . 2008 , , 179 , 19 kellermann , k.i . , fomalont , e.b . , mainieri , v. , padovani , p. , rosati , p. , shaver , p. , tozzi , p. , & miller , n. 2008 , , 179 , 71 magnelli , b. , elbaz , d. , chary , r.r . , dickinson , m. , le borgne , d. , frayer , d.t . , & willmer , c.n.a . 2011 , , 528 , 35 mainieri , v. et al . 2008 , , 179 , 95 mao , m.y . , huynh , m.t . , norris , r.p . , dickinson , m. , frayer , d. , helou , g. , & monkiewicz , j.a . 2011 , , 731 , 79 mao , m.y . 2012 , , 426 , 3334 mauduit , j .- c . 2012 , , 124 , 714 miller , n.a . , fomalont , e.b . , kellermann , k.i . , mainieri , v. , norman , c. , padovani , p. , rosati , p. , & tozzi , p. 2008 , , 179 , 114 [ m08 ] moncelsi , l. et al . 2011 , , 727 , 83 muxlow , t.w.b . 2005 , , 358 , 1159 norris , r.p . 2006 , , 132 , 2409 oliver , s.j . 2012 , , 424 , 1614 owen , f.n . , & morrison , g.e . 2008 , , 136 , 1889 padovani , p. , mainieri , v. , tozzi , p. , kellermann , k.i . , fomalont , e.b . , miller , n. , rosati , p. , & shaver , p. 2009 , , 694 , 235 padovani , p. , miller , n. , kellermann , k.i . , mainieri , v. , rosati , p. , & tozzi , p. 2011 , , 740 , 20 perley , r.a . , chandler , c.j . , butler , b.j . , & wrobel , j.m . 2011 , , 739 , l1 retzlaff , j. et al . 2010 , , 511 , 50 rix , h .- w . 2004 , , 152 , 163 schinnerer , e. , sargent , m.t . , bondi , m. , smolcic , v. , datta , a. , carilli , c.l . , bertoldi , f. , blain , a. , ciliegi , p. , koekemoer , a. , & scoville , n.z . 2010 , , 188 , 384 scott , k.s . 2010 , , 405 , 2260 tozzi , p. et al . 2009 , , 698 , 740 truch , m.d.p . et al . 2009 , , 707 , 1723 vanzella , e. et al . 2008 , , 478 , 83 vattakunnel , s. et al . 2011 , , 420 , 2190 wei , a. et al . 2009 , , 707 , 1201 white , r.l . , becker , r.h . , helfand , d.j . , & gregg , m.d . 1997 , , 475 , 479 xue , y.q . 2011 , , 195 , 10 yun , m.s . 2011 , , 420 , 957 l l l r ecdfs 1 & 03:33:22.25 & -27:48:30.0 & 10.5 @xmath0jy + ecdfs 2 & 03:32:55.12 & -27:38:03.0 & 9.4 @xmath0jy + ecdfs 3 & 03:32:00.88 & -27:38:03.0 & 9.7 @xmath0jy + ecdfs 4 & 03:31:33.75 & -27:48:30.0 & 9.5 @xmath0jy + ecdfs 5 & 03:32:00.88 & -27:58:57.0 & 10.0 @xmath0jy + ecdfs 6 & 03:32:55.12 & -27:58:57.0 & 9.3 @xmath0jy + l r r r r r fitted axis & 0.985 & 0.248 & 0.978 & 0.218 & 6 + envelope & 1.085 & 0.290 & 1.071 & 0.256 & 6 + envelope+20 & 1.020 & 0.276 & 0.994 & 0.225 & 10 + both - or & 1.041 & 0.278 & 1.013 & 0.227 & 11 + both - and & 0.964 & 0.241 & 0.960 & 0.213 & 5 + snr & 0.938 & 0.421 & 0.938 & 0.421 & 0 + hybrid & 1.005 & 0.258 & 0.997 & 0.220 & 6 + r r r r r r r r r r r r r r r r r r r r r r r r r r r r r 3 & 31 & 10.69 & -28 & 03 & 22.8 & 18.7 & 207.5 & 10.6 & 365.2 & 26.9 & 5 & 14.6 & 1.4 & 2.3 & 0.0 & 0.7 & 169 & 1 & i & + 3 & 31 & 10.81 & -27 & 55 & 52.8 & 13.8 & 123.1 & 8.8 & 156.6 & 18.0 & 4 & 11.8 & 0.0 & 2.0 & 0.0 & 1.4 & 120 & 0 & p & + 3 & 31 & 10.83 & -27 & 55 & 57.8 & 5.3 & 47.3 & 8.9 & 32.7 & 12.1 & 4 & 11.9 & 0.0 & 3.1 & 0.0 & 0.0 & 2 & 0 & p & + 3 & 31 & 10.93 & -27 & 49 & 55.5 & 5.1 & 40.8 & 8.0 & 35.3 & 12.6 & 4 & 10.2 & 0.0 & 2.6 & 0.0 & 2.7 & 168 & 0 & p & + 3 & 31 & 10.95 & -27 & 58 & 10.4 & 12.7 & 121.2 & 9.5 & 136.9 & 17.9 & 4 & 12.9 & 0.0 & 1.3 & 0.0 & 1.5 & 21 & 0 & p & + 3 & 31 & 11.48 & -27 & 52 & 59.0 & 12.8 & 105.9 & 8.0 & 185.8 & 20.5 & 4 & 10.9 & 0.0 & 2.7 & 0.0 & 1.9 & 165 & 1 & p & + 3 & 31 & 11.63 & -27 & 55 & 06.3 & 5.4 & 47.1 & 8.7 & 69.0 & 20.1 & 4 & 11.4 & 0.0 & 4.7 & 0.0 & 1.6 & 49 & 0 & p & + 3 & 31 & 11.69 & -27 & 31 & 44.2 & 211.2 & 2555 . & 12.1 & 3232 . & 65.4 & 3 & 15.2 & 0.2 & 0.5 & 0.0 & 0.0 & 29 & 1 & i & a + 3 & 31 & 11.84 & -27 & 58 & 18.0 & 8.1 & 77.2 & 9.4 & 92.5 & 18.4 & 4 & 12.7 & 0.0 & 2.9 & 0.0 & 1.3 & 162 & 0 & p & + 3 & 31 & 11.89 & -27 & 59 & 52.3 & 6.8 & 65.8 & 9.5 & 90.3 & 20.5 & 5 & 13.7 & 0.0 & 2.7 & 0.0 & 0.6 & 22 & 0 & p & + r r r r r r r r r r r r r r r r r r r r r r r r r r r r r 3 & 31 & 13.99 & -27 & 55 & 19.9 & 144.4 & 1242 . & 8.6 & 7180 . & 57.3 & 4 & 11.3 & 7.1 & 7.2 & 2.0 & 2.1 & 81 & 1 & p & 1 + 3 & 31 & 15.06 & -27 & 55 & 18.9 & 120.9 & 1040 . & 8.6 & 1206 . & 16.5 & 4 & 11.2 & 0.0 & 0.6 & 0.0 & 0.7 & 42 & 1 & p & 1 + 3 & 31 & 17.05 & -27 & 55 & 15 . & 103.6 & 870.0 & 8.4 & 3650 . & 43.0 & 4 & 11.0 & 4.6 & 4.7 & 3.0 & 3.1 & 157 & 1 & p & 1 + 3 & 31 & 14.69 & -28 & 01 & 51.7 & 8.7 & 90.1 & 10.4 & 1367 . & 168.5 & 5 & 13.6 & 16.0 & 20.9 & 1.2 & 2.8 & 77 & 1 & p & 2 + 3 & 31 & 17.35 & -28 & 01 & 47.4 & 26.5 & 270.2 & 10.2 & 358.8 & 21.0 & 5 & 12.9 & 0.0 & 1.6 & 0.0 & 1.0 & 48 & 1 & p & 2 + 3 & 31 & 20.16 & -28 & 01 & 46.2 & 9.9 & 98.5 & 9.9 & 856.7 & 95.5 & 5 & 12.4 & 5.1 & 6.7 & 3.1 & 4.6 & 100 & 1 & p & 2 + 3 & 31 & 24.83 & -27 & 52 & 08.9 & 1043 . & 8032 . & 7.7 & 24670 . & 30.5 & 4 & 10.2 & 4.0 & 4.1 & 1.7 & 1.7 & 64 & 1 & p & 3 + 3 & 31 & 25.01 & -27 & 52 & 07.7 & 960.5 & 7396 . & 7.7 & 15380 . & 22.3 & 4 & 10.2 & 2.1 & 2.2 & 2.0 & 2.0 & 179 & 1 & p & 3 + figures for each of the 17 multiple - component sources described in section [ sec - mult ] and listed in table [ tbl - multiple ] are presented here . unless otherwise noted , each image is 1 by 1 with the greyscale ranging from @xmath81 @xmath0jy beam@xmath2 to 105 @xmath0jy beam@xmath2 . contours are depicted at 5 , 8 , 13 , 21 , ... , and 987 times the local rms noise . information regarding distinct sources may be found in @xcite .

the main goal of the paper is to find the _ absolute maximum _ of the width of the separatrix chaotic layer as function of the frequency of the time - periodic perturbation of a one - dimensional hamiltonian system possessing a separatrix , which is one of the major unsolved problems in the theory of separatrix chaos . for a given small amplitude of the perturbation , the width is shown to possess sharp peaks in the range from logarithmically small to moderate frequencies . these peaks are universal , being the consequence of the involvement of the nonlinear resonance dynamics into the separatrix chaotic motion . developing further the approach introduced in the recent paper by soskin et al . ( _ pre _ * 77 * , 036221 ( 2008 ) ) , we derive leading - order asymptotic expressions for the shape of the low - frequency peaks . the maxima of the peaks , including in particular the _ absolute maximum _ of the width , are proportional to the perturbation amplitude times either a logarithmically large factor or a numerical , still typically large , factor , depending on the type of system . thus , our theory predicts that the maximal width of the chaotic layer may be much larger than that predicted by former theories . the theory is verified in simulations . an application to the facilitation of global chaos onset is discussed .

0908.0459

separatrix chaotic layers ( scls ) play a fundamental role for hamiltonian chaos and may be important in a broad variety of subjects in physics and astronomy @xcite . one of the most important characteristics of the layer is its width in energy or in related quantities . it can be easily found _ numerically _ by means of integration of the hamiltonian equations with a set of initial conditions in the vicinity of the separatrix . but it is important also to be able to find it _ theoretically_. there is a long and rich history of the corresponding studies . the results may be classified as follows . * 1 . heuristic analytic results . * consider a 1d hamiltonian system perturbed by a weak time - periodic perturbation : @xmath0 where @xmath1 possesses a separatrix and , for the sake of notation compactness , all relevant parameters of @xmath2 and @xmath3 , except possibly @xmath4 , are assumed to be @xmath5 . there were a few heuristic criteria set by physicists ( see e.g. @xcite ) which gave qualitatively similar results for the scl width @xmath6 in terms of energy @xmath7 : @xmath8 the quantity @xmath9 is called the _ separatrix split _ @xcite ( see also eq . ( 4 ) below ) : it determines the maximum distance between the perturbed incoming and outgoing separatrices @xcite . it follows from ( 2 ) that the maximum of @xmath6 lies in the frequency range @xmath10 while the maximum itself is @xmath11 : @xmath12 * 2 . mathematical and accurate physical results . * many papers studied the scl by mathematical and accurate physical methods . for the range @xmath13 , there were many works studying the separatrix splitting ( see the review @xcite and references therein ) and the scl width in terms of the normal coordinates ( see the review @xcite and references therein ) . though quantities studied in these works typically differ from those studied by physicists @xcite , they implicitly confirm the main qualitative conclusion from the heuristic formula ( 2 ) in the high frequency range : if @xmath13 the scl width is exponentially small . there were also several works studying the scl in the opposite ( i.e. adiabatic ) limit @xmath14 : see e.g. and references therein . in the context of the scl width , it is most important that @xmath15 for most of the systems . for a particular class of systems , namely for ac - driven spatially periodic systems ( e.g. the ac - driven pendulum ) , the width of the scl part above the separatrix diverges in the adiabatic limit @xcite : the divergence develops for @xmath16 . finally , there is a qualitative estimation of the scl width for the range @xmath10 within the kolmogorov - arnold - moser ( kam ) theory @xcite while the quantitative estimation within the kam theory appears to be very difficult for this frequency range @xcite . it follows from the results in @xcite that the width in this range is of the order of the separatratrix split while the latter is of the order of @xmath17 . thus , from the above results , it could seem to follow that , for all systems except the ac - driven spatially periodic systems , the maximum of the scl width is @xmath18 and occurs in the range @xmath10 , quite in agreement with the heuristic result ( 3 ) . even for the ac - driven spatially periodic systems , this conclusion could seem to apply to the width of the scl part below the separatrix , for the whole frequency range , and to the width of the scl part above the separatrix , for @xmath19 . * 3 . numerical evidences of high peaks in @xmath20 and their rough estimates . * the above conclusion does not agree with several numerical studies carried out during the last decade ( see e.g. @xcite ) which have revealed the existence of sharp peaks in @xmath20 in the frequency range @xmath21 the heights of which greatly exceed @xmath17 ( see also figs . 2 , 3 , 5 , 6 below ) . thus , the peaks represent the general _ dominant feature _ of the function @xmath20 . the peaks were related by the authors of @xcite to the absorption of nonlinear resonances by the scl . for some partial case , rough analytic estimates for the position and magnitude of the peaks were made in @xcite . approach to an accurate description of the peaks . * accurate analytic estimates for the peaks were lacking . it is explicitly stated in the review @xcite that the search for the mechanism of the involvement of resonances into the separatrix chaos and an accurate analytic description of the peaks are being among the most important and challenging tasks in the study of separatrix chaos . the first step towards this accomplishment was done in the recent papers @xcite , where a new approach to the theoretical treatment of the separatrix chaos for the relevant frequency range was developed and applied to the problem of the onset of global chaos between two close separatrices . an application of the approach to the , more common , single - separatrix cases was only discussed in @xcite . the _ present _ paper formulates the basic ideas of the approach in terms more general than @xcite and , on the basis of this approach , develops the _ first ever _ accurate theoretical description of the peaks i.e. of the scl width as a function of frequency in the range of the _ maximum _ of the width , which is the most important range from the physical point of view . in particular , we show that the maximal width of the separatrix chaotic layer may be much larger than it was assumed before . in the latter context , all systems are classified by us into two different types : for systems of type i , the ratio between the maximal width and the perturbation amplitude @xmath17 _ logarithmically diverges _ in the asymptotic limit @xmath22 while , for systems of type ii , it asymptotically approaches a _ constant _ ( still large , typically ) . though the form of our treatment differs from typical forms of mathematical theorems in this subject ( cf . @xcite ) , the results yield the _ exact _ leading - order term in the asymptotic expansion of the width in the parameter of smallness @xmath23 . our theory is in excellent agreement with the results of numerical integration of the equations of motion . ii describes the basic ideas of the approach . iii presents the classification into two types of systems , using rough estimates . iv develops the leading - order asymptotic theory for an archetypal example of type i and compares it with the numerical integration of hamiltonian equations of motion . v develops the leading - order asymptotic theory for two archetypal examples of type ii and compares it with the numerical integration . next - order corrections are estimated in sec . discussion of a few other issues , including in particular an application to the global chaos onset , is presented in sec . conclusions are drawn in sec . the new approach , which is developed in @xcite and here , may be briefly formulated as the matching between the discrete chaotic dynamics of the separatrix map in the immediate vicinity of the separatrix and the continuous regular - like dynamics of the resonance hamiltonian beyond the close vicinity of the separatrix . the present section describes the general features of the approach in more details . the motion near the separatrix may be approximated by the _ separatrix map _ ( sm ) @xcite . it was introduced for the first time in @xcite and its various modifications were used in many studies afterwards , sometimes being called as the _ whisker map_. it was re - derived in @xcite rigorously , as the leading - order approximation of the motion near the separatrix in the asymptotic limit @xmath22 , and an estimate of the errors was carried out too ( see also @xcite and references therein ) . we remind the main ideas which allow one to introduce the sm @xcite . for the sake of simplicity , let us consider a perturbation @xmath3 that does not depend on the momentum : @xmath24 . a system with an energy close to the separatrix value spends most of the time in the vicinity of the saddle(s ) , where the velocity is exponentially small . differentiating @xmath7 with respect to time and allowing for the equations of motion of the system ( 1 ) , we can show that @xmath25 . thus , the perturbation can significantly change the energy only when the velocity is not small i.e. during the relatively short intervals while the system is away from the saddle(s ) : these intervals correspond to _ pulses _ of velocity as a function of time . consequently , it is possible to approximate the continuous hamiltonian dynamics by a discrete dynamics which maps the energy @xmath26 , the perturbation angle @xmath27 and the velocity sign @xmath28 from pulse to pulse . the actual form of the sm may vary , depending on the system under study , but its features , relevant in the present context , are similar for all systems . for the sake of clarity , let us consider the explicit case when the separatrix of @xmath1 possesses a single saddle and two symmetric loops while @xmath29 . then the sm reads @xcite : @xmath30 where @xmath31 is the separatrix energy , @xmath32 is the frequency of oscillation with energy @xmath26 in the unperturbed case ( i.e. for @xmath33 ) , @xmath34 is the instant corresponding to the @xmath35-th turning point in the trajectory @xmath36 , and @xmath37 is an arbitrary value from the range of time intervals which greatly exceed the characteristic duration of the velocity pulse while being much smaller than the interval between the subsequent pulses @xcite . consider the two most general ideas of our approach . if a trajectory of the sm includes a state with @xmath38 and an arbitrary @xmath39 and @xmath40 , then this trajectory is chaotic . * indeed , the angle @xmath39 of such a state is not correlated with the angle of the state at the previous step of the map , due to the divergence of @xmath41 . therefore , the angle at the previous step may deviate from a multiple of @xmath42 by an arbitrary value and , hence , the energy of the state at the previous step may deviate from @xmath31 by an arbitrary value within the interval @xmath43 $ ] . the velocity sign @xmath40 is not correlated with that at the previous step either @xcite . given that a regular trajectory of the sm can not include a step where all three variables of the sm change random - like , we conclude that such a trajectory is chaotic . though the above arguments appear to be obvious , they may not be considered as a mathematically rigorous proof , so that the statement about the chaotic nature of the sm trajectory which includes any state with @xmath38 should be considered as a conjecture supported by the above arguments and by the results of the numerical iteration of the sm . possibly , the mathematically rigorous proof should involve an analysis of the lyapunov exponents for the sm ( cf . @xcite ) but this appears to be a technically difficult problem . we emphasize however that the rigorous proof of the conjecture is not crucial for the validity of the main results of the present paper , namely of the _ leading _ terms in the asymptotic expressions describing the peaks of the scl width as a function of the perturbation frequency . it will be obvious from the next item that in order to derive the leading term it is sufficient to know that the chaotic trajectory does visit areas of the phase space where the energy deviates from the separatrix by values of the order of the separatrix split @xmath44 , which is a widely accepted fact @xcite . as well known @xcite , at the leading - order approximation the frequency of eigenoscillation as function of the energy near the separatrix is proportional to the reciprocal of the logarithmic factor @xmath45 where @xmath46 is the energy of the stable states . given that the argument of the logarithm is large in the relevant range of @xmath26 , the function @xmath47 is nearly constant for a substantial variation of the argument . therefore , * as the sm maps the state @xmath48 onto the state with @xmath49 , the value of @xmath32 for the given @xmath50 is nearly the same for most of the angles @xmath51 ( except in the close vicinity of multiples of @xmath52 ) * , namely @xmath53 moreover , if the deviation of the sm trajectory from the separatrix increases further , @xmath32 remains close to @xmath54 provided the deviation is not too large , namely if @xmath55 . if @xmath56 , then the evolution of the sm ( 4 ) may be regular - like for a long time until the energy returns to the close vicinity of the separatrix , where the trajectory is chaotized . such a behavior is especially pronounced if the perturbation frequency is close to @xmath57 or @xmath58 or to one of their multiples of relatively low order : the resonance between the perturbation and the eigenoscillation gives rise to an accumulation of energy changes for many steps of the sm , which results in a deviation of @xmath26 from @xmath31 that greatly exceeds the separatrix split @xmath59 . consider a state at the boundary of the scl . the deviation of energy of such a state from @xmath31 depends on its position at the boundary . in turn , the maximum deviation is a function of @xmath4 . the latter function possesses the absolute maximum at @xmath4 close to @xmath57 or @xmath58 typically @xcite , for the upper or lower boundary of the scl respectively . this corresponds to the absorption of the , respectively upper and lower , 1st - order nonlinear resonance by the scl . the above intuitive idea has been explicitly confirmed in @xcite : it has been shown in the appendix of @xcite that , in the relevant range of energies , the separatrix map can be reduced to the system of two differential equations which are indentical to the equations of motion of the auxiliary resonance hamiltonian which describes the resonance dynamics in terms of the conventional canonically conjugate slow variables , action @xmath60 and slow angle @xmath61 where @xmath62 is the angle variable @xcite ( see eq . ( 16 ) below ) while @xmath63 is the relevant resonance number i.e. the integer number closest to the ratio @xmath64 . thus , the result of the matching between the discrete chaotic dynamics of the sm and the continuous regular - like dynamics of the resonance hamiltonian is the following @xcite . after the chaotic trajectory of the sm visits any state on the separatrix , the system transits in one step of the sm to a given upper or lower curve in the @xmath65 plane which has been labelled @xcite respectively upper or lower _ generalized separatrix split _ ( gss ) curve @xcite : @xmath66 where @xmath67 is the conventional separatrix split @xcite while @xmath68 is the amplitude of the melnikov - like integral defined in eq . ( 4 ) above ( cf . @xcite ) , and the angle @xmath69 may take any value from one of the two ranges : either @xmath70 $ ] or @xmath71 $ ] @xcite . after that , because of the closeness of @xmath4 to the @xmath63-th harmonic of @xmath32 in the relevant range of @xmath26 @xcite , for a relatively long time the system follows the _ nonlinear resonance _ ( nr ) dynamics ( see eq . ( 16 ) below ) , during which the deviation of the energy from the separatrix value grows , greatly exceeding @xmath67 for most of the trajectory . as time goes on , @xmath69 is moving and , at some point , the deviation in energy from the separatrix value begins to decrease . this decrease lasts until the system hits the gss curve , after which it returns to the separatrix just for one step of the separatrix map . at the separatrix , the slow angle @xmath69 is chaotized , so that a new stage of evolution similar to the one just described occurs , i.e. the nonlinear resonance dynamics starting from the gss curve with a new ( random ) value of @xmath69 . of course , the sm can not describe the variation of the energy during the velocity pulses ( i.e. in between instants relevant to the sm ) : in some cases this variation can be comparable with the change within the sm dynamics . this additional variation will be taken into account below , where relevant ( see sec . v below ) . one might argue that , even for the instants relevant to the sm , the sm describes the original hamiltonian dynamics only approximately @xcite and may therefore miss some fine details of the motion : for example , the above picture does not include small windows of stability on the very separatrix . however these fine details are irrelevant in the present context , in particular the relative portion of the windows of stability on the separatrix apparently vanishes in the asymptotic limit @xmath22 . the boundary of the sm chaotic layer is formed by those parts of the sm chaotic trajectory which deviate from the separatrix more than others . as follows from the structure of the chaotic trajectory described above , the upper / lower boundary of the sm chaotic layer is formed in one of the two following ways @xcite : 1 ) if there exists a _ self - intersecting _ resonance trajectory ( in other words , the resonance separatrix ) the lower / upper part of which ( i.e. the part situated below / above the self - intersection ) touches or intersects the upper / lower gss curve while the upper / lower part does not , then the upper / lower boundary of the layer is formed by the upper / lower part of this self - intersecting trajectory ( figs . 1(a ) and 1(b ) ) ; 2 ) otherwise the boundary is formed by the resonance trajectory _ tangent _ to the gss curve ( fig . it is shown below that , in both cases , the variation of the energy along the resonance trajectory is larger than the separatrix split @xmath67 by a logarithmically large factor @xmath72 . therefore , over the boundary of the sm chaotic layer the largest deviation of the energy from the separatrix value , @xmath73 , may be taken , in the leading - order approximation , to be equal to the largest variation of the energy along the resonance trajectory forming the boundary , while the latter trajectory can be entirely described within the resonance hamiltonian formalism . finally , we mention in this section that , as obvious from the above description of the boundary , @xmath74 possesses a local maximum @xmath75 at @xmath4 for which the resonance separatrix just _ touches _ the corresponding gss curve ( see fig . 1(a ) ) . 0.2 cm 0.2 cm 0.2 cm as obvious from sec . ii above , @xmath76 is equal , in the leading order , to the width @xmath77 of the nonlinear resonance which touches the separatrix . let us make a rough estimate of @xmath77 : it will turn out that it is possible to classify all systems into two different types . with this aim , we expand the perturbation @xmath3 into a fourier series in @xmath78 and a fourier series in @xmath62 : @xmath79 as in the standard theory of a nonlinear resonance @xcite , let us single out the relevant @xmath80 for a given peak , and denote its absolute value by @xmath81 : @xmath82 let us now roughly estimate the width of the resonance , using the pendulum approximation of the resonance dynamics @xcite : @xmath83 of course , this approximation assumes the constancy of @xmath84 in the resonance range of energies , while it is not so in our case : @xmath85 in the vicinity of the separatrix @xcite , so that the relevant derivative @xmath86 strongly varies within the resonance range . however , for our rough estimate we may use the maximal value of @xmath87 , which is equal to @xmath77 approximately . if @xmath4 is of the order of @xmath88 , then eq . ( 10 ) reduces to the following rough asymptotic equation for @xmath77 : @xmath89 the asymptotic solution of eq . ( 11 ) essentially depends on @xmath90 as a function of @xmath77 . in this context , all systems can be divided in the following two types . * type i*. the separatrix of the unperturbed system has _ two or more _ saddles while the relevant fourier coefficient @xmath91 possesses _ different _ values on adjacent saddles . given that , for @xmath92 , the system stays most of time near one of the saddles , the coefficient @xmath93 as a function of @xmath62 is nearly a square wave : it oscillates between the values at the different saddles . the relevant @xmath94 is typically odd and , therefore , @xmath95 approaches a well defined non - zero value . substituting it in eq . ( 11 ) , we conclude that @xmath96 * type ii*. either ( i ) the separatrix of the unperturbed system has a _ single saddle _ , or ( ii ) it has more than one saddle but the perturbation coefficient @xmath97 is _ identical _ for all saddles . then @xmath93 , as a periodic function of @xmath62 , significantly differs from its value at the saddle(s ) only during a small part of the period in @xmath62 : this part is @xmath98 . hence , @xmath99 . substituting this value in eq . ( 11 ) , we conclude that @xmath100 thus , for systems of type i , the maximal width of the sm chaotic layer is proportional to @xmath17 times a logarithmically large factor @xmath101 while , for systems of type ii , it is proportional to @xmath17 times a numerical factor . as shown below , the variation of energy in between the instants relevant to the sm is @xmath18 , which thus it is much less than @xmath77 ( 12 ) for the systems of type i and it is of the same order of @xmath102 ( 13 ) for the systems of type ii . therefore , one may expect that the maximal width of the layer for the original hamiltonian system ( 1 ) , @xmath103 , is at least roughly approximated by that for the sm , @xmath104 , so that the above classification of systems is relevant to @xmath103 too . this is confirmed both by the numerical integration of equations of motion of the original hamiltonian system and by the more accurate theory presented in the next two sections . for the sake of clarity , we consider a concrete example of type i , while the generalization is straightforward . let us consider an archetypal example : the ac - driven pendulum ( sometimes called as a pendulum subject to a dipole time - periodic perturbation ) @xcite : @xmath105 fig . 2 presents the results of computer simulations ( i.e. of a numerical integration of the equations of motion ) for a few values of @xmath17 and several values of @xmath4 . it shows that : 1 ) the function @xmath106 indeed possesses sharp peaks ; their height greatly exceed the estimates by the heuristic @xcite , adiabatic and mathematical moderate - frequency @xcite theories ( see the inset ) ; 2 ) as predicted by the rough estimates in sec . iii , the 1st peak of @xmath107 shifts to smaller values of @xmath4 while its magnitude grows , as @xmath17 decreases . below , we develop the leading - order asymptotic theory and compare it with results of the simulations . 0.2 cm 0.9 cm 0.9 cm 0.2 cm before moving on , we note that the sm ( approximated in the relevant case by the nonlinear resonance dynamics ) considers states of the system only at discrete instants . apart from the variation of energy within the sm dynamics , the variation of energy in the hamiltonian system occurs also in between the instants relevant to the sm . given that @xmath108 , this latter variation may be considered in adiabatic approximation and it is of the order of @xmath17 . as follows from the rough estimates above and from the accurate consideration below , the variation of energy within the sm dynamics for systems of type i is logarithmically larger i.e. larger by the factor @xmath109 . therefore , the variation of energy in between the instants relevant to the sm may be neglected at the leading - order approximation for systems of type i ( the parameter of smallness of the asymptotic theory is @xmath110 ) : @xmath111 . for the sake of notational compactness , we shall omit the subscript `` @xmath112 '' further in this section . for the system ( 14 ) , the separatrix energy is equal to 1 , while the asymptotic ( for @xmath92 ) dependence @xmath32 is @xcite : @xmath113 let us consider the range of energies below @xmath31 ( the range above @xmath31 may be considered analogously ) and assume that @xmath4 is close to one of the odd multiples of @xmath58 . the nonlinear resonance dynamics of the slow variables in the range of the approximately resonant energies may be described as follows @xcite ( cf . also @xcite ) : @xmath114 where @xmath60 and @xmath62 are the canonical variables action and angle respectively @xcite ; @xmath115 is the minimal energy over all @xmath116 , @xmath117 ; @xmath118 is the minimal coordinate of the conservative motion with a given value of energy @xmath26 ; @xmath119 is the number of right turning points in the trajectory @xmath120 $ ] of the conservative motion with energy @xmath26 and given initial state @xmath121 . the resonance hamiltonian @xmath122 is obtained from the original hamiltonian @xmath123 transforming to action - angle variables @xmath124 , with a further multiplication by @xmath63 ; extracting the term @xmath125 ( that corresponds to the transformation @xmath126 ) ; and neglecting all the fast - oscillating terms ( their effect on the dynamics of slow variables is small : see the estimate of the corrections in sec . vi below ) i.e. keeping only the resonance term in the double fourier expansion of the perturbation . let us derive the asymptotic expression for @xmath127 , substituting the asymptotic expression ( 15 ) for @xmath32 into the definition of @xmath127 ( 16 ) and carrying out the integration : @xmath128 as for the asymptotic value @xmath129 , it is easy to see that @xmath130 , as a function of @xmath62 , asymptotically approaches a square wave , oscillating between @xmath131 and @xmath52 , so that , for sufficiently small @xmath132 , @xmath133 the next issue is the analysis of the phase space of the resonant hamiltonian ( 16 ) . substituting @xmath134 ( 16 ) into the equations of motion ( 16 ) , it is easy to see that their stationary points have the following values of the slow angle @xmath135 while the corresponding action is determined by the equation @xmath136 where the sign @xmath137corresponds to @xmath138 ( 19 ) . as usual ( cf . @xcite ) , the term @xmath139 in ( 20 ) may be neglected in the leading - order approximation , and eq . ( 20 ) reduces to the resonance condition @xmath140 the lowest - order solution of which is @xmath141 eqs . ( 19 ) and ( 22 ) together with ( 17 ) explicitly determine the elliptic and hyperbolic points of the hamiltonian ( 16 ) . the hyperbolic point is often called saddle and corresponds to @xmath142 or @xmath143 in ( 19 ) for even or odd @xmath132 respectively . the saddle point generates the resonance separatrix . using the asymptotic relations ( 17 ) and ( 18 ) , we obtain that the resonance hamiltonian ( 16 ) takes the following asymptotic value in the saddle : @xmath144 the second asymptotic equality in ( 23 ) takes into account the relation ( 22 ) . as explained in sec . ii above , @xmath106 possesses a local maximum at @xmath4 for which the resonance separatrix is tangent to the lower gss curve ( fig . 1(a ) ) . for the relevant frequency range @xmath14 , the separatrix split ( which represents the maximum deviation of the energy along the gss curve from @xmath31 ) approaches the following value @xcite , in the asymptotic limit @xmath22 @xmath145 as it is shown further down , the variation of energy along the relevant resonance trajectories is much larger . therefore , in the leading - order approximation , the gss curve may be simply replaced by the separatrix of the unperturbed system i.e. by the horizontal line @xmath38 or , equivalently , @xmath146 . then the tangency occurs at @xmath69 shifted from the saddle by @xmath52 , so that the condition of tangency is written as @xmath147 substituting here @xmath148 ( 23 ) , we finally obtain the following transcendental equation for @xmath149 : @xmath150 fig . 3(b ) demonstrates the excellent agreement between eq . ( 26 ) and the results of simulations for the hamiltonian system in a wide range of @xmath17 . in the asymptotic limit @xmath22 , the lowest - order explicit solution of eq . ( 26 ) is @xmath151 as follows from eq . ( 26 ) , the value of @xmath152 ( 22 ) for @xmath153 is @xmath154 its leading - order expression is : @xmath155 if @xmath156 , then , in the chaotic layer , the largest deviation of energy from the separatrix value corresponds to the minimum energy @xmath157 on the nonlinear resonance separatrix ( fig . 1(a , b ) ) , which occurs at @xmath69 shifted by @xmath52 from the saddle . the condition of equality of @xmath134 at the saddle and at the minimum of the resonance separatrix is written as @xmath158 let us seek its asymptotic solution in the form @xmath159 substituting ( 31 ) and ( 23 ) into eq . ( 30 ) , we obtain for @xmath160 the following transcendental equation : @xmath161 where @xmath149 is given by eq . ( ( 31 ) and ( 32 ) describe the left wing of the @xmath132-th peak of @xmath106 . 3(a ) demonstrates the good agreement between our analytic theory and simulations for the hamiltonian system . as follows from eq . ( 26 ) , eq . ( 32 ) for @xmath153 reduces to the relation @xmath162 , i.e. @xmath163 as follows from eqs . ( 33 ) , ( 31 ) and ( 28 ) , the maximum for a given peak is : @xmath164 fig . 3(c ) shows the excellent agreement of this expression with the results of simulations for the hamiltonian system in a wide range of @xmath17 . the leading - order expression for @xmath165 is : @xmath166 that confirms the rough estimate ( 12 ) . as @xmath4 decreases , @xmath160 increases exponentially sharply , as follows from eq . ( 32 ) . in order to understand how @xmath167 decreases upon decreasing @xmath4 , it is convenient to rewrite eq . ( 31 ) expressing the exponent by means of eq . ( 32 ) : @xmath168 it follows from eqs . ( 32 ) and ( 36 ) that @xmath167 decreases power - like rather than exponentially when @xmath4 is decreased . in particular , @xmath169 for the far part of the wing . as for the right wing of the peak , i.e. for @xmath170 , over the chaotic layer the largest deviation of energy from the separatrix value corresponds to the minimum of the resonance trajectory tangent to the gss curve ( fig . the value of @xmath69 in the minimum coincides with @xmath171 . in the leading - order approximation , the gss curve may be replaced by the horizontal line @xmath146 , so that the tangency occurs at @xmath172 . then the energy at the minimum @xmath157 can be found from the equation @xmath173 let us seek its asymptotic solution in the form @xmath174 substituting ( 38 ) into eq . ( 37 ) , we obtain for @xmath175 the following transcendental equation : @xmath176 where @xmath149 is given by eq . ( 26 ) . ( 38 ) and ( 39 ) describe the right wing of the @xmath132-th peak of @xmath106 . 3(a ) shows the good agreement between our analytic theory and simulations . as follows from eq . ( 26 ) , the solution of eq . ( 39 ) for @xmath177 is @xmath178 , so the right wing starts from the value given by eq . ( 28 ) ( or , approximately , by eq . ( 29 ) ) . expressing the exponent in ( 38 ) from ( 39 ) , we obtain the following equation @xmath179 it follows from eqs . ( 39 ) and ( 40 ) that @xmath180 decreases power - like rather than exponentially for increasing @xmath4 . in particular , @xmath181 in the far part of the wing . the further analysis of the asymptotic shape of the peak is done in sec . vii below . beyond the peaks , the function @xmath106 is logarithmically small in comparison with the maxima of the peaks . the functions @xmath182 and @xmath183 in the ranges beyond the peaks are also logarithmically small . hence , nearly any combination of the functions @xmath183 and @xmath184 which is close to @xmath183 in the vicinity of @xmath149 and to @xmath184 in the vicinity of @xmath185 may be considered as an approximation of the function @xmath106 with a logarithmic accuracy with respect to the maxima of the peaks , @xmath186 and @xmath187 , in the whole range @xmath188 $ ] . one of the easiest combinations is the following : @xmath189 we used this function in fig . 3(a ) , and the analogous combination will be also used in the other cases . in fact , the theory may be generalized in such a way that eq . ( 41 ) would well approximate @xmath106 in the ranges far beyond the peaks with a logarithmic accuracy even with respect to @xmath106 itself rather than to @xmath186 only ( cf . the next section ) . however , we do not do this in the present case , being interested primarily in the leading - order description of the peaks . finally , we demonstrate in fig . 4 that the lowest - order theory describes quite well the layer boundaries even in the poincar section rather than only in energy / action . 0.2 cm 0.2 cm 0.2 cm we shall consider two characteristic examples of type ii , corresponding to the classification given in sec . iii . as an example of the system where the separatrix of the unperturbed system possesses a single saddle , we shall consider the ac - driven duffing oscillator @xcite . as an example of the system where the separatrix possesses more than one saddle while the perturbation takes equal values at the saddles , we shall consider the pendulum with an oscillating suspension point @xcite . the treatment of these cases is similar in many respects to the one presented in sec . so , we present it in less details , emphasizing the differences . consider the following archetypal hamiltonian @xcite : @xmath190 the asymptotic dependence of @xmath32 on @xmath26 for @xmath26 below the separatrix energy @xmath191 is the following @xcite @xmath192 correspondingly , the resonance values of energies ( determined by the condition analogous to ( 21 ) ) are @xmath193 the asymptotic dependence of @xmath127 is @xmath194 the nonlinear resonance dynamics is described by the resonance hamiltonian @xmath134 which is identical to eq . ( 16 ) in form . obviously , the actual dependencies @xmath32 and @xmath127 are given by eq . ( 43 ) and ( 45 ) respectively . the most important difference is in @xmath195 : instead of a non - zero value ( see ( 18 ) ) , it approaches 0 as @xmath92 . namely , it is @xmath196 @xcite : @xmath197 i.e. @xmath198 is much smaller than in systems of type i ( cf . ( 18 ) ) . due to this , the resonance is weaker . at the same time , the separatrix split @xmath67 is also smaller , namely @xmath199 ( cf . @xcite ) rather than @xmath11 as for the systems of type i. that is why the separatrix chaotic layer is still dominated by the resonance dynamics while the matching of the separatrix map and nonlinear resonance dynamics is still valid in the asymptotic limit @xmath22 @xcite . similarly to the previous section , we find the value of @xmath134 in the saddle in the leading - order approximation @xcite : @xmath200 where @xmath152 is given in ( 44 ) . as before , the maximum width of the layer corresponds to @xmath4 , for which the resonance separatrix is tangent to the gss curve ( fig . it can be shown @xcite that the angle of tangency asymptotically approaches @xmath201 while the energy still lies in the resonance range , where @xmath202 . using the expressions for @xmath203 ( cf . ( 16 ) ) , @xmath127 ( 45 ) , @xmath195 ( 46 ) , and taking into account that in the tangency @xmath204 , the value of @xmath134 at the tangency reads , in the leading - order approximation , @xmath205 allowing for eqs . ( 47 ) and ( 48 ) , the condition for the maximum , @xmath206 , reduces to @xmath207 thus , these values @xmath152 are logarithmically smaller than the corresponding values ( 28 ) for systems of type i. the values of @xmath4 corresponding to the maxima of the peaks in @xmath106 are readily obtained from ( 49 ) and ( 44 ) : @xmath208 the derivation of the shape of the peaks for the chaotic layer of the separatrix map in the leading order , i.e. within the nonlinear resonance ( nr ) approximation , is similar to that for type i. so , we present only the results , marking them with the subscript @xmath209 . the left wing of the @xmath132th peak of @xmath210 is described by the function @xmath211 where @xmath160 is the positive solution of the transcendental equation @xmath212 similarly to the type i case , @xmath213 , so that @xmath214 eq . ( 53 ) confirms the rough estimate ( 13 ) . the right wing of the peak is described by the function @xmath215 where @xmath216 is the solution of the transcendental equation @xmath217 similarly to the type i case , @xmath218 . as follows from eqs . ( 49 ) and ( 53 ) , the typical variation of energy within the nonlinear resonance dynamics ( that approximates the separatrix map dynamics ) is @xmath139 . for the hamiltonian system , the variation of energy in between the discrete instants corresponding to the separatrix map @xcite is also @xmath139 . therefore , unlike the case of type i , one needs to take it into account even at the leading - order approximation . let us consider the right well of the duffing potential ( the results for the left well are identical ) , and denote by @xmath219 the instant at which the energy @xmath26 at a given @xmath220-th step of the separatrix map is taken : it corresponds to the beginning of the @xmath220-th pulse of velocity @xcite i.e. the corresponding @xmath221 is close to a left turning point @xmath222 in the trajectory @xmath120 $ ] . let us also take into account that the relevant frequencies are small so that the adiabatic approximation may be used . thus , the change of energy from @xmath219 up to a given instant @xmath78 during the following pulse of velocity ( @xmath223 ) may be calculated as @xmath224 for the motion near the separatrix , the velocity pulse corresponds approximately to @xmath225 ( see the definition of @xmath62 ( 16 ) ) . thus , the corresponding slow angle is @xmath226 . for the left wing of the peak of @xmath106 ( including the maximum of the peak too ) , the boundary of the chaotic layer of the separatrix map is formed by the lower part of the nr separatrix . the minimum energy along this separatrix occurs at @xmath227 . taking this into account , and also that @xmath228 , we conclude that @xmath229 . so , @xmath230 , i.e. it does lower the minimum energy of the layer of the hamiltonian system . the maximum lowering occurs at the right turning point @xmath231 : @xmath232 we conclude that the left wing of the @xmath132-th peak is described by the following formula : @xmath233 where @xmath234 is given by eqs . ( 51)-(52 ) . in particular , the maximum of the peak is : @xmath235 for the right wing of the peak , the minimum energy of the layer of the separatrix map occurs at @xmath69 coinciding with @xmath171 ( fig . 1(c ) ) i.e. equal to 0 . as a result , @xmath236 and , hence , @xmath237 . so , this variation can not lower the minimal energy of the layer for the main part of the wing , i.e. for @xmath238 where @xmath239 is defined by the condition @xmath240 . for @xmath241 , the minimal energy in the layer occurs at @xmath227 , and it is determined exclusively by the variation of energy during the velocity pulse ( the nr contribution is close to zero at such @xmath69 ) . thus , we conclude that there is a bending of the wing at @xmath242 : @xmath243 where @xmath244 is given by eqs . ( 54 ) and ( 55 ) . analogously to the previous case , @xmath107 may be approximated in the whole frequency range by eq . ( 41 ) with @xmath167 and @xmath245 given by eqs . ( 58 ) and ( 60 ) respectively . moreover , unlike the previous case , now the theory accurately describes also the range far beyond the peaks : @xmath246 is dominated in this range by the velocity pulse contribution @xmath247 , which is accurately taken into account both by eq . ( 58 ) and by eq . ( 60 ) . 5 shows a very reasonable agreement between theory and simulations , especially for the 1st peak @xcite . 0.8 cm 0.8 cm 0.2 cm consider the archetypal hamiltonian @xcite @xmath248 though the treatment is similar to the previous case , there are also characteristic differences . one of them is the following : although the resonance hamiltonian is similar to the hamiltonian ( 16 ) , instead of the fourier component of the coordinate , @xmath249 , there should be the fourier component of @xmath250 , @xmath251 , which can be shown to be : @xmath252 the description of the chaotic layer of the separatrix map at the lowest order , i.e. within the nr approximation , is similar to that for the ac - driven duffing oscillator . so , we present only the results , marking them with the subscript @xmath209 . the frequency of the maximum of a given @xmath132-th peak is : @xmath253 this expression well agrees with simulations for the hamiltonian system ( fig . to logarithmic accuracy , eq . ( 63 ) coincides with the formula following from eq . ( 8) of @xcite ( reproduced in @xcite as eq . ( 21 ) ) taken in the asymptotic limit @xmath22 ( or , equivalently , @xmath254 ) . however , the numerical factor in the argument of the logarithm in the asymptotic formula following from the result of @xcite is half our value : this is because the nonlinear resonance is approximated in @xcite by the conventional pendulum model which is not valid near the separatrix ( cf . iii above ) . 0.9 cm 0.9 cm 0.2 cm the left wing of the @xmath132th peak of @xmath210 is described by the function @xmath255 where @xmath160 is the positive solution of the transcendental equation @xmath256 similarly to the previous cases , @xmath257 . hence , @xmath258 eq . ( 66 ) confirms the rough estimate ( 13 ) . the right wing of the peak is described by the function @xmath259 where @xmath216 is the solution of the transcendental equation @xmath260 similarly to the previous cases , @xmath218 . consider now the variation of energy during the velocity pulse . though the final result looks quite similar to the case with a single saddle , its derivation has some characteristic differences , and we present it in detail . unlike the case with a single saddle , the pulse may start close either to the left turning point or to the right turning point , and the sign of the velocity in such pulses is opposite @xcite . as concerns the angle @xmath62 in the pulse , it is close to @xmath261 or @xmath262 respectively . so , let us calculate the change of energy from the beginning of the pulse , @xmath219 , till a given instant @xmath78 within the pulse : @xmath263 here , the third equality assumes adiabaticity while the last equality takes into account that the turning points are close to the maxima of the potential i.e. close to a multiple of @xmath42 ( where the cosine is equal to 1 ) . the quantity @xmath6 ( 69 ) has the maximal absolute value at @xmath264 . so , we shall further consider @xmath265 the last equality takes into account that , as mentioned above , the relevant @xmath266 is either @xmath261 or @xmath262 . for the left wing , the value of @xmath69 at which the chaotic layer of the separatrix map possesses a minimal energy corresponds to the minimum of the resonance separatrix . it is equal to @xmath52 or @xmath267 if the fourier coefficient @xmath268 is positive or negative , i.e. for odd or even @xmath132 , respectively : see eq . thus @xmath269 for any @xmath132 and , therefore , it does lower the minimal energy of the boundary . we conclude that @xmath270 where @xmath234 is given by eqs . ( 64)-(65 ) . in particular , the maximum of the peak is : @xmath271 the expression ( 72 ) confirms the rough estimate ( 13 ) and well agrees with simulations ( fig . 6(c ) ) . at the same time , it differs from the formula which can be obtained from eq . ( 10 ) of @xcite ( using also eqs . ( 1 ) , ( 3 ) , ( 8) , ( 9 ) of @xcite ) in the asymptotic limit @xmath22 : the latter gives for @xmath186 the asymptotic value @xmath272 . though the result @xcite ( referred also in @xcite ) provides for the correct functional dependence on @xmath17 , it is quantitatively incorrect because ( i ) it is based on the pendulum approximation of the nonlinear resonance while this approximation is not valid in the vicinity of the separatrix ( see the discussion of this issue in sec . iii above ) , and ( ii ) it does not take into account the variation of energy during the velocity pulse . the right wing , analogously to the case of the duffing oscillator , possesses a bending at @xmath242 at which @xmath273 , that corresponds to the switching of the relevant @xmath69 by @xmath52 . we conclude that : @xmath274 where @xmath244 is given by eqs . ( 66 ) and ( 67 ) . similarly to the previous case , both the peaks and the frequency ranges far beyond the peaks are well approximated by eq . ( 41 ) with @xmath167 and @xmath245 given by eqs . ( 71 ) and ( 73 ) respectively ( fig . we have explicitly calculated only the leading term @xmath6 in the asymptotic expansion of the chaotic layer width . the explicit calculation of the next - order term @xmath275 is possible but it is rather complicated and cumbersome : see the closely related case with two separatrices @xcite , where most of the next - order contributions are calculated quantitatively @xcite . in the present paper , where the perturbation amplitude @xmath17 in the numerical examples is 4 orders of magnitude smaller than that in @xcite , there is no particular need to calculate the next - order correction @xmath276 quantitatively . let us estimate it just qualitatively , with the main purpose to demonstrate that its ratio to the lowest - order term does vanish in the asymptotic limit @xmath22 . we shall consider separately the contribution @xmath277 stemming from the various corrections _ within _ the resonance approximation ( 16 ) and the contribution @xmath278 stemming from the corrections _ to _ the resonance approximation . the former contribution may be estimated similarly to the case considered in @xcite : it stems from the deviation of the gss curve from the separatrix ( this deviation reaches @xmath67 at certain angles : see eq . ( 7 ) ) , from the difference between the exact resonance condition ( 20 ) and the approximate one ( 21 ) , etc . it can be shown that the absolute value of the ratio between @xmath279 and the leading term is logarithmically small ( cf . @xcite ) : @xmath280 let us turn to the analysis of the contribution @xmath278 , i.e.the contribution stemming from the corrections to the resonance hamiltonian ( 16 ) . it is convenient to consider separately the cases of the left and right wings of the peak . as described in secs . iv and v above , the left wing corresponds in the leading - order approximation to formation of the boundary of the layer by the _ separatrix _ of the resonance hamiltonian ( 16 ) . the resonance approximation ( 16 ) neglects time - periodic terms while the frequencies of oscillation of these terms greatly exceed the frequency of eigenoscillation of the resonance hamiltonian ( 16 ) around its relevant elliptic point i.e. the elliptic point inside the area limited by the resonance separatrix . as is well known @xcite , fast - oscillating terms acting on a system with a separatrix give rise to the onset of an _ exponentially narrow _ chaotic layer in place of the separatrix . in the present context , this means that the correction to the maximal action @xmath281 stemming from fast - oscillating corrections to the resonance hamiltonian , i.e. @xmath278 , is _ exponentially small _ , thus being negligible in comparison with the correction @xmath282 ( see ( 74 ) ) . the right wing , described in secs . iv and v above , corresponds in leading - order approximation to the formation of the boundary of the layer by the resonance trajectory _ tangent _ to the gss curve . for the part of the right wing exponentially close in frequency to the frequency of the maximum , the tangent trajectory is close to the resonance separatrix , so that the correction stemming from fast - oscillating terms is exponentially small , similarly to the case of the left wing . as the frequency further deviates from the frequency of the maximum , the tangent trajectory further deviates from the resonance separatrix and the correction @xmath278 differs from the exponentially small correction estimated above . it may be estimated in the following way . it follows from the second - order approximation of the averaging method @xcite that the fast - oscillating terms lead , in the second - order approximation , to the onset of additional terms @xmath283 and @xmath284 in the dynamic equations for slow variables @xmath281 and @xmath69 respectively , where @xmath285 and @xmath286 are of the order of the power - law - like function of @xmath110 in the relevant range of @xmath281 . the corresponding correction to the width of the chaotic layer in energy may be expressed as @xmath287 where @xmath288 and @xmath289 are instants corresponding to the minimum and maximum deviation of the tangent trajectory from the separatrix of the unperturbed system ( cf . figs . 1(c ) and 4(c ) ) . the interval @xmath290 may be estimated as follows : @xmath291 where @xmath292 is the value of @xmath293 averaged over the tangent trajectory . it follows from ( 16 ) that @xmath294 taking together eqs . ( 75)-(77 ) and allowing for the fact that @xmath295 is of the order of a power - law - like function of @xmath110 , we conclude that @xmath296 where @xmath297 is some power - law - like function . the value @xmath278 is still asymptotically smaller than the absolute value of the correction within the resonance approximation , @xmath298 , which is of the order of @xmath17 or @xmath299 for systems of type i or type ii respectively . thus , we conclude that , both for the left and right wings of the peak , ( i ) the correction @xmath300 is determined by the correction within the resonance approximation @xmath277 , and ( ii ) in the asymptotic limit @xmath22 , the overall next - order correction is negligible in comparison with the leading term : @xmath301 this estimate well agrees with results in figs . 3 - 6 . in this section , we briefly discuss the following issues : 1 ) the _ scaled _ asymptotic shape of the peaks , 2 ) peaks in the range of _ moderate _ frequencies , 3 ) _ steps _ in the amplitude dependence of the layer width , 4 ) an application to the _ global chaos onset_. * let us analyse the scaled asymptotic shape of the peaks . consider first systems of type i. then the peaks are described in the leading - order approximation exclusively within the separatrix map dynamics ( approximated , in turn , by the nr dynamics ) . as follows from eqs . ( 32 ) , ( 34 ) , ( 36 ) , ( 39 ) and ( 40 ) , most of the peak with a given @xmath132 can be written in the universal scaled form : + @xmath302 + where the universal function @xmath303 is strongly asymmetric : + @xmath304 + it is not difficult to show that + @xmath305 + thus , the function @xmath303 is discontinuous at the maximum . to the left of the maximum , the function relatively _ slowly _ approaches the far part of the wing ( which falls down power - like ) while , to the right of the maximum , the function first drops _ jump - wise _ by a factor @xmath306 and then _ sharply _ approaches the far part of the wing ( which falls down power - like ) . + as follows from eqs . ( 80 ) , ( 81 ) , ( 82 ) and ( 27 ) , the peaks are logarithmically narrow , i.e. the ratio of the half - width of the peak , @xmath307 , to @xmath308 is logarithmically small : + @xmath309 + we emphasize that the shape ( 81 ) is not restricted to the example ( 14 ) : it is valid for any system of type i. + for systems of type ii , the contributions from the nr and from the variation of energy during the pulse of velocity , as concerns the @xmath17 dependence , are formally of the same order but , numerically , the latter contribution is typically much smaller than the former one . thus , typically , the function ( 81 ) well approximates the properly scaled shape of the major part of the peak for systems of type ii too . * the quantitative theory presented in the paper relates only to the peaks of _ small _ order @xmath63 i.e. in the range of logarithmically small frequencies . at the same time , the magnitude of the peaks is still significant up to the frequencies of the order of one . this occurs because , for the motion close to the separatrix , the order of magnitude of the fourier coefficients remains the same up to logarithmically large numbers @xmath63 . the shape of the peaks remains the same but their magnitude decreases typically ( but , in some cases , it may even increase in some range of frequencies ) . the quantitative description of this decrease as well as the analysis of more sophisticated cases requires a generalization of our theory , that will be presented elsewhere . * apart from the frequency dependence of the layer width , our theory is also relevant for the amplitude dependence : it describes the jumps @xcite in the dependence of the width on @xmath17 and the transition between the jumps and the linear dependence . the values of @xmath17 at which the jumps occur , @xmath310 , are determined by the same condition which determines @xmath149 in the frequency dependence of the width . the formulas relevant to the left wings of the peaks in the frequency dependence describe the ranges @xmath311 while the formulas relevant to the right wings describe the ranges @xmath312 . * finally we note that , apart from systems with a separatrix , our work may be relevant to _ nonlinear resonances _ in any system . if the system is perturbed by a weak time - periodic perturbation , then nonlinear resonances arise and their dynamics is described by the model of the auxiliary time - periodically perturbed pendulum @xcite . if the original perturbation has a single harmonic , then the effective perturbation of the auxiliary pendulum is necessarily a high - frequency one , and chaotic layers associated with the resonances are exponentially narrow @xcite while our results are irrelevant . but , if either the amplitude or the angle of the original perturbation is slowly modulated , or if there is an additional harmonic of a slightly shifted frequency , then the effective perturbation of the auxiliary pendulum is a low - frequency one @xcite and the layers become much wider @xcite while our theoretical approach becomes relevant . it may allow to find optimal parameters of the perturbation for the facilitation of the onset of global chaos associated with the overlap in energy between different - order nonlinear resonances @xcite : the overlap may be expected to occur at a much smaller amplitude of perturbation in comparison with that one required for the overlap in case of a single - harmonic perturbation . we have further developed the new @xcite approach to the treatment of the separatrix chaos in the range of logarithmically small and moderate frequencies , where the chaos takes the largest possible area in phase space . the approach is based on the matching between the discrete chaotic dynamics of the sm and the continuous regular - like dynamics of the resonance hamiltonian . using this approach and taking also into account the dynamics in between instants corresponding to the sm , we have presented the _ first ever _ accurate asymptotic description of high sharp peaks of the width of the separatrix chaotic layer in energy as function of the frequency of a weak time - periodic perturbation , including in particular the _ absolute maximum _ of the function . our work provides the accurate base to explain former numerical and heuristic results and intuitive assumptions @xcite , corrects the errors of a previous heuristic theory @xcite , discovers new important features , and opens up new horizons for future studies and applications . the observed peaks arise due to the involvement of the nonlinear resonance dynamics into the separatrix chaotic motion . in the context of the magnitude of the peaks , all systems are classified into _ two types _ : the magnitude of the peaks is proportional to the perturbation amplitude @xmath17 times either a _ logarithmically large _ factor @xmath72 ( for systems of type i ) or a _ numerical _ factor ( for systems of type ii ) . type i includes systems for which the separatrix of their unperturbed hamiltonian has more than one saddle while the perturbation is not identical on adjacent saddles ( an example is an ac - driven pendulum ) . all other systems belong to type ii . the latter type includes , in particular , a pendulum with an oscillating suspension point , for which our result differs from the result @xcite since the latter ( i ) is based on the conventional approximation of the nonlinear resonance ( not valid near the separatrix ) , and ( ii ) does not take into account the variation of energy during the velocity pulse ( i.e. in between the instants relevant to the separatrix map ) . our theory is verified by computer simulations . the shape of the peaks is strongly asymmetric . in the asymptotic limit of small amplitudes , the shape of the peaks for type universal_. for type ii , the shape is quite similar , differing only by a typically small contribution stemming from the variation of energy during the velocity pulse . finally , our work suggests a new method for the _ facilitation of global chaos onset _ due to the enhanced overlap of nonlinear resonances . the theoretical approach developed by us may be used to derive the optimal choice of parameters of the perturbation leading to the facilitation . this work was partly supported by the grant within the convention between the institute of semiconductor physics and university of pisa for 2008 and by the royal society grant 2007/r2-ijp . we acknowledge discussions with vassili gelfreich , igor khovanov and oleg yevtushenko . we especially appreciate numerous stimulating discussions with george zaslavsky and his role for the subject of hamiltonian chaos on the whole . we dedicate this paper to the memory of george zaslavsky who recently passed away . soskin , r. mannella , o.m . yevtushenko , in _ chaos , complexity and transport : theory and applications _ , proceedings of the cct-07 , eds . c. chandre , x. leoncini , and g. zaslavsky ( world scientific , singapore , 2008 ) , pp . 119 - 128 . formally , @xmath313 is not defined for @xmath38 but , if to shift @xmath26 from @xmath31 for an infinitesemal value , @xmath313 acquires the value equal to either @xmath314 or @xmath315 , dependently on the sign of the shift . given that @xmath316 is proportional to @xmath317 while the latter is random - like ( as it has been shown above ) , @xmath316 is not correlated with @xmath318 if @xmath319 . for the sm relating to the ac - driven spatially periodic systems , the time during which the sm undergoes a regular - like evolution above the separatrix diverges in the adiabatic limit @xmath14 @xcite , and the width of the part of the sm layer above the separatrix diverges too . however , we do not consider this case here since it is irrelevant to the main subject of the present paper i.e. to to the involvement of the resonance dynamics into the separatrix chaotic motion . namely , that of these two intervals is relevant in which the derivative @xmath320 in the nonlinear resonance equations ( see eq . ( 16 ) below ) is positive or negative , for the case of the upper or lower gss curve respectively . i.e. @xmath26 determined by eq . ( 7 ) for any @xmath69 except from the close vicinity of multiples of @xmath52 . as shown in @xcite , eq . ( 7 ) is irrelevant to the boundary of the chaotic layer in the range of @xmath69 close to multiples of @xmath52 while the boundary in this range of @xmath69 still lies in the resonance range of energies , where @xmath321 . the adiabatic estimate for @xmath106 is equal approximately to @xmath42 . the estimate following from the results of the mathematical work @xcite for @xmath10 is of the same order , so that it is conditionally represented in the inset in fig . 2 by the same line as for the adiabatic estimate ( dashed line ) . the disagrement between the theory and simulations for the magnitude of the 2nd peak is about three times larger than that for the 1st peak , so that the height of the 2nd peak is about 30@xmath323 smaller than that by the asymptotic theory . this occurs because , for energies relevant to the 2nd peak , the deviation from the separatrix is much higher than that for the 1st peak . due to the latter , the fourier coefficient @xmath324 for the relevant @xmath26 is significantly smaller than that by the asymptotic formula ( 42 ) . besides , the velocity pulse contribution @xmath6 also significantly decreases while the separatrix split increases as @xmath4 becomes @xmath5 . the generalization of the theory for @xmath10 will be presented elsewhere . the paper @xcite quantitatively calculates all contributions into the next - order correction , except the contribution stemming from the second - order approximation of the averaging method @xcite , i.e. from the next - order correction to the resonance hamiltonian . this should not be confused with the widening occuring with the separatrix chaotic layer in the _ original _ pendulum if an originally single - harmonic perturbation of a high frequency is completed by one more harmonic of a slightly shifted frequency : see @xcite and references therein .

the closest pair problem ( cpp ) is one of the well studied and fundamental problems in computing . given a set of points in a metric space , the problem is to identify the pair of closest points . another closely related problem is the fixed radius nearest neighbors problem ( frnnp ) . given a set of points and a radius , the problem is , for every input point , to identify all the other input points that are within a distance of from . a naive deterministic algorithm can solve these problems in quadratic time . cpp as well as frnnp play a vital role in computational biology , computational finance , share market analysis , weather prediction , entomology , electro cardiograph , n - body simulations , molecular simulations , etc . as a result , any improvements made in solving cpp and frnnp will have immediate implications for the solution of numerous problems in these domains . we live in an era of big data and processing these data take large amounts of time . speeding up data processing algorithms is thus much more essential now than ever before . in this paper we present algorithms for cpp and frnnp that improve ( in theory and/or practice ) the best - known algorithms reported in the literature for cpp and frnnp . these algorithms also improve the best - known algorithms for related applications including time series motif mining and the two locus problem in genome wide association studies ( gwas ) .

1407.5609

the closest pair problem ( cpp ) has a rich history and has been extensively studied . on an input set of @xmath2 points , the problem is to identify the closest pair of points . a straight forward algorithm for cpp takes quadratic ( in @xmath2 ) time . most of the algorithms proposed in the literature are concerned with the euclidean space . in the rest of this paper , unless otherwise mentioned , we imply the euclidean space . in his seminal paper , rabin proposed a randomized algorithm with an expected run time of @xmath3 @xcite ( where the expectation is in the space of all possible outcomes of coin flips made in the algorithm ) . algorithm used the floor function as a basic operation . in 1979 , fortune and hopcroft presented a deterministic algorithm with a run time of @xmath4 assuming that the floor operation takes @xmath5 time @xcite . both of these algorithms assume a constant - dimensional space ( and the run times have an exponential dependency on the dimension ) . the algorithm of preparata and shamos for points in 2d is deterministic and runs in @xmath6 time @xcite . yao has proven a lower bound of @xmath7 on the algebraic decision tree model ( for any dimension ) @xcite . this lower bound holds under the assumption that the floor function is not allowed . the algorithm of khuller and matias is also randomized and has an expected linear run time utilizing the floor operation @xcite . there are two steps in the algorithm . in the first step , the distance between the closest pair of points is estimated within a factor of 3 . in the second step , the neighborhood of each point @xmath1 is explored to identify those points that are within a distance of @xmath8 from @xmath1 , where @xmath8 is the estimate that step 1 comes up with . using this neighborhood information , the correct pair is identified . one of the major issues with the above algorithms is the fact that their run times are exponentially dependent on the dimension . for example , the expected run time of @xcite s algorithm is @xmath9 on @xmath2 points from a @xmath10-dimensional space . so , even for a moderate value of @xmath10 , the algorithm may be very slow in practice . there are numerous applications for which the dimension is very large . in the following sections , we consider two such application domains . time series motif mining ( tsmm ) is a crucial problem that can be thought of as cpp in a large dimensional space . in one version of the tsmm problem , we are given a sequence @xmath11 of real numbers and an integer @xmath12 . the goal is to identify two subsequences of @xmath11 of length @xmath12 each that are the most similar to each other ( from among all pairs of subsequences of length @xmath12 each ) . these most similar subsequences are referred to as _ time series motifs_. let @xmath13 be a collection of all the @xmath12-mers of @xmath11 . ( an @xmath12-mer is nothing but a contiguous subsequence of @xmath11 of length @xmath12 ) . clearly , the @xmath12-mers in @xmath13 can be thought of as points in @xmath14 . as a result , the tsmm problem is the same as cpp in @xmath14 . any of the above mentioned algorithms can thus be used to solve the tsmm problem . a typical value for @xmath12 of practical interest is several hundreds ( or more ) . for these values of @xmath12 , the above algorithms ( @xcite,@xcite,@xcite,@xcite ) will take an unacceptable amount of time ( because of the exponential dependence on the dimension ) . designing an efficient practical and exact algorithm for the tsmm problem remains an ongoing challenge . mueen , et al . have presented an elegant exact algorithm called mk for tsmm @xcite . mk improves the performance of the brute - force algorithm with a novel application of the triangular inequality . mk is currently the best - performing algorithm in practice for tsmm . a number of probabilistic as well as approximate algorithms are also known for solving this problem ( see e.g. , @xcite ) . for instance , the algorithm of @xcite exploits algorithms proposed for finding @xmath15-motifs from biological data . the idea here is to partition the time series data into frames of certain width . followed by this , the mean value in each frame is computed . this mean is quantized into four intervals and as a result , the original time series data is converted into a string of characters from an alphabet of size 4 . finally , any @xmath15-motif finding algorithm is applied on the transformed string to identify the time series motifs . an application of great interest in bioinformatics is genome wide association study ( gwas ) . a lot of effort has been spent to identify mappings between phenotypical traits and genomic data . due to the advent of next generation high throughput sequencing technologies , nowadays it is possible to study the genomic structure of individuals in detail . single nucleotide polymorphisms ( snps ) are positions in the genome where nucleotides vary among individuals @xcite . snps in the human genome are considered to be responsible for different phenotypical traits . in gwas , two different problems have been focused on . in single locus association study , researchers try to find out the association between phenotypical traits and individual snps . in two locus association study , the goal is to figure out the association between pairs of snps and phenotypical traits . a major task in this study is that of identifying the most correlated pair of snps . two locus associations are also known as gene - gene interactions . such interactions are believed to be major factors responsible for many complex phenotypical traits @xcite . given that the number of snps found in humans is @xmath16 to @xmath17 , a brute force way of scanning through every possible pair of snps to identify the most correlated pair is not feasible in practice . a number of algorithms for the two locus problem can be found in the literature . for instance , genetic algorithms are used in @xcite and @xcite . the algorithms proposed in @xcite and @xcite take @xmath18 time , where @xmath2 is the number of snps and @xmath19 is the number of subjects . an algorithm with an expected run time of @xmath20 , where @xmath21 is a constant , has been presented in @xcite . this algorithm exploits an algorithm known for the light bulb problem @xcite and locality sensitive hashing ( lsh ) @xcite . in this paper we present efficient algorithms for cpp . our algorithms improve the results reported in several papers including @xcite , @xcite , and @xcite . our main contributions can be summarized as follows : 1 . we improve the cpp algorithm of @xcite by introducing a novel idea . specifically , @xcite s algorithm is based on two basic ideas . we contribute a third idea that results in an improvement of the run time for exact tsmm by a factor of around 1.5 . we also show how to extend our algorithm to solve the fixed radius nearest neighbors problem ( frnnp ) . 2 . we present an algorithm for cpp when the domain of interest has strings of characters ( from a finite alphabet ) and the metric is hamming distance . it turns out that mk does not perform well for the case of hamming distance , and to be fair , we note that the authors of mk do not claim it might . a comparison of our algorithm with mk reveals that our algorithm outperforms mk ( by a factor of around 200 ) . our algorithm can also be used in approximate tsmm . specifically , instead of using @xmath15-motif search algorithms , our algorithm can be used in @xcite . in this case , the run time of the algorithm in @xcite will improve significantly , since exact algorithms for solving the @xmath15-motif search problem take time that is exponential in @xmath12 and @xmath10 . our algorithm is a modified version of the light bulb algorithm of @xcite . 3 . the light bulb algorithm of @xcite finds the most correlated pair of bulbs . the light bulb problem can be thought of as cpp in the space of * binary * strings with hamming distance as the metric . we extend this algorithm when the strings are from an arbitrary ( finite ) alphabet . more importantly , we present an algorithm for finding the * least * correlated pair of strings ( from an arbitrary alphabet ) . the algorithm of @xcite also solves this problem utilizing locality sensitive hashing ( lsh ) @xcite . our algorithm does not use lsh . instead , it uses a novel deterministic mapping function that we have come up with . 4 . using the above algorithm for finding the least correlated pair , we present a novel algorithm for solving the two locus gwas problem . an experimental comparison reveals that our algorithm is four times faster than the algorithm of @xcite . we note here that the authors of @xcite use pearson s correlation coefficient to measure the similarity between a pair of snps , whereas we use the complement of the hamming distance as the measure of similarity . the rest of this paper is organized as follows . in section [ sec2 ] we provide some notations . in section [ sec3 ] we present our improved algorithm for cpp ( called mpr ) and compare it with mk . in this section we also provide an analysis of mk and experimentally compare the performances of mk and mpr . section [ sec4 ] deals with the case of character strings and hamming distance . specifically , we show how to modify the light bulb algorithm of @xcite to get an algorithm for finding the most correlated pair of strings from an arbitrary finite alphabet . we compare this algorithm with mk experimentally . in section [ sec5 ] we present an algorithm for finding the least correlated pair of strings . this algorithm is based on a novel mapping function that we have come up with . section [ sec6 ] is devoted to the problem of two locus association in gwas . in particular , we present a novel algorithm for this problem and compare our algorithm with that of @xcite . some concluding remarks are given in section [ sec7 ] . let @xmath22 be a sequence of real numbers ( or characters from a finite alphabet ) . an @xmath12-mer of @xmath23 is nothing but a subsequence of @xmath23 of @xmath12 contiguous elements of @xmath23 . the @xmath12-mers of @xmath23 are @xmath24 , for @xmath25 . if the elements of @xmath23 are real numbers , then the euclidean distance between @xmath26 and @xmath27 , denoted as @xmath28 , is @xmath29 . if the elements of @xmath23 are characters from an alphabet @xmath30 , then the hamming distance between @xmath26 and @xmath27 , denoted as @xmath28 , is @xmath31 where @xmath32 if @xmath33 and @xmath34 if @xmath35 ( for any @xmath36 ) . a sequence of characters can be thought of as a string of characters , since we can obtain a string from the sequence by concatenating the characters . thus we ll use the terms ` a sequence of characters ' and ` a string of characters ' interchangeably . let @xmath37 and @xmath38 be two sequences of characters . also , let the hamming distance between @xmath39 and @xmath40 be @xmath10 . then , by the _ number of matches _ between @xmath39 and @xmath40 we mean @xmath41 . also , the _ correlation _ between @xmath39 and @xmath40 is defined to be @xmath42 . the input for this problem are a sequence @xmath22 and an integer @xmath12 . the goal is to find two @xmath12-mers of @xmath23 that are the closest to each other ( from among all the pairs of @xmath12-mers of @xmath23 ) . a general version of this problem is one where the input consists of @xmath2 points from @xmath14 and we want to identify the two closest points . a straight forward algorithm will compute the distance between every pair of @xmath12-mers and output the pair with the least distance . since we can compute the distance between two @xmath12-mers in @xmath43 time , this simple algorithm for tsmm will run in a total of @xmath44 time . the mk algorithm of @xcite speeds up the brute force method by pruning off a large number of pairs that can not possibly be the closest . there are two main ideas used in mk . the first idea in the algorithm is to speedup the computation of distances . let @xmath45 and @xmath46 be any two @xmath12-mers . to compute the distance between @xmath47 and @xmath48 , the algorithm keeps adding @xmath49 for @xmath50 . when the sum exceeds @xmath51 , this pair is immediately dropped ( without completing the rest of the distance computation ) . this technique is known as _ early abandoning_. the second idea in mk uses the triangular inequality in a novel way . let @xmath47 and @xmath48 be any two @xmath12-mers . at any stage in the algorithm , we have an upper bound @xmath52 on the distance between the closest pair of @xmath12-mers . if @xmath53 can be inferred to be greater than @xmath52 , then we can drop the pair @xmath54 from future consideration ( since this pair can not be the closest ) . ideally , we would like to calculate @xmath53 exactly for every pair of @xmath12-mers @xmath47 and @xmath48 . but this will take too much time . mk circumvents this problem by * estimating * the distance between @xmath47 and @xmath48 via the triangular inequality . in particular , a random reference @xmath12-mer @xmath55 is chosen and the distance between each @xmath12-mer and @xmath55 is computed . the @xmath12-mers are kept in an ascending order of their distances to @xmath55 . from thereon , @xmath56 is used as a lower bound on @xmath53 . if this lower bound is @xmath57 , then @xmath54 is dropped from future consideration . the above algorithm is generalized to employ multiple reference @xmath12-mers . the use of multiple references speeds up the algorithm . in this section we provide an ( informal ) analysis of the mk algorithm to explain why the algorithm has a very good performance . specifically , if we choose multiple random reference points , the algorithm achieves a much better run time than having a single reference point . we explain why this is the case . for ease of understanding consider the 2d euclidean space . the analysis can be extended to points in @xmath14 . for any two @xmath12-mers @xmath47 and @xmath48 , the closer @xmath56 is to @xmath53 , the better will be our estimate and hence the better will be our chance of dropping @xmath54 ( if @xmath54 is not the closest pair ) . it turns out that the quality of the lower bound @xmath56 is decided by two factors : 1 ) the angle @xmath58 and 2 ) @xmath59 . we illustrate this with an example . let @xmath60 and @xmath61 . consider a reference point @xmath62 ( figure [ figure1](a ) ) . note that @xmath63 is at a distance of @xmath64 from @xmath47 . in this case @xmath65 . also , @xmath66 . let @xmath67 be the point we get by keeping the distance between the reference point and @xmath47 the same , but changing this angle to @xmath68 ( figure [ figure1](b ) ) . in this case , @xmath69 improves to @xmath70 . as another example , if the reference point @xmath55 lies on the perpendicular bisector of @xmath47 and @xmath48 , then @xmath71 [ fig:1 ] for any two input points @xmath47 and @xmath48 , if we pick multiple reference points randomly , then we would expect that at least one of these reference points @xmath55 will be such that the angle @xmath58 will be such that @xmath72 will be ` large ' . in contrast , if we have only one reference point , for some pairs of points the corresponding angles may be ` good ' , but for a good percentage of the pairs , the angles may not be ` good ' . ( a reference point @xmath55 is ` good ' if @xmath72 is close to @xmath53 ) . the effect of @xmath59 on @xmath72 can be seen with the same examples . consider the example of figure [ figure1](a ) . assume that we keep the angle the same but increase @xmath73 to @xmath74 and get the reference point @xmath75 ( figure [ figure1](c ) ) . in this case , @xmath76 improves to @xmath77 . also , in the example of figure [ figure1](b ) , say we keep the angle the same but increase @xmath78 to @xmath74 and get the reference point @xmath79 . in this case , @xmath80 improves to @xmath81 . of course , if the reference point @xmath55 lies on the perpendicular between @xmath47 and @xmath48 , then however large @xmath59 could be , @xmath72 will continue to be zero ! however , the probability of this happening is low . for a given angle @xmath82 , we can compute the limit of @xmath72 as @xmath59 tends to @xmath83 . for instance when the angle is @xmath84 ( figure [ figure1](a ) ) , this limit is @xmath85 . our proposed new algorithm indeed exploits the relationship between @xmath59 and @xmath72 . in particular , we pick a collection @xmath13 of random reference points and project each of these points out by multiplying each coordinate value of each point by a factor of @xmath86 . for example , @xmath86 could be @xmath74 . the rest of the algorithm is the same as mk . a pseudocode for our algorithm , called _ motif discovery with projected reference points ( mpr ) _ , is given below . * algorithm * mpr + _ input : _ @xmath22 and an integer @xmath12 , where each @xmath87 is a real number ( for @xmath88 ) . input are also @xmath89 and @xmath86 , where @xmath89 is the number of references and @xmath86 is the projection factor . + _ output : _ the two closest @xmath12-mers of @xmath23 . + * 1 ) * pick @xmath89 random @xmath12-mers of @xmath23 as references ; project these references by multiplying each element in each @xmath12-mer by @xmath86 . let these projected references be @xmath90 . + * 2 ) * compute the distance between every @xmath12-mer of @xmath23 and every projected reference @xmath12-mer . + * 3 ) * sort the @xmath12-mers of @xmath23 with respect to their distances to @xmath63 . let the sorted @xmath12-mers be @xmath91 . + * 4 ) * let @xmath92 ; let @xmath93 ; + * 5 ) for * @xmath94 * to * @xmath95 * do * + @xmath96 * to * @xmath95 * do * + @xmath97 ; + @xmath98 * to * @xmath89 * do * + @xmath99 * then * + @xmath100 ; * exit * ; + @xmath101 * then exit else * + compute @xmath102 ; + @xmath103 * then * + @xmath104 ; @xmath105 ; + * 6 ) output * @xmath106 . * observation : * please note that even though the above algorithm has been presented for solving the tsmm problem , it is straight forward to extend it to solve cpp in @xmath14 . in this section we show why our idea of projecting reference points improves the performance of the algorithm . let the input points be from @xmath107 for some integer @xmath10 . consider any two input points @xmath39 and @xmath40 . let @xmath0 be any reference point . note that any three points are coplanar . consider any hyperplane @xmath108 containing @xmath109 and @xmath0 . if we multiply every coordinate of @xmath0 by the same number , then the resultant point will also lie in @xmath108 . this is because the equation defining @xmath108 will be of the form @xmath110 . thus , in order to see how @xmath111 changes with a scaling of @xmath0 , it suffices to consider the case that these three points are in 2d . without loss of generality let @xmath39 be @xmath112 and @xmath40 be @xmath113 , for some real number @xmath114 . there are two cases to consider for the position of @xmath0 relative to @xmath39 and @xmath40 : 1 ) @xmath0 is to the right of the perpendicular bisector of @xmath39 and @xmath40 ; 2 ) @xmath0 is to the left of the perpendicular bisector between @xmath39 and @xmath40 . these two cases are illustrated in figures [ fig2](a ) and [ fig2](b ) , respectively . note that when @xmath0 lies on the perpendicular bisector of @xmath39 and @xmath40 , @xmath111 will be zero . in case 1 , let @xmath115 be @xmath116 , @xmath117 being a scaling factor . @xmath118 , @xmath119 , and @xmath120 . as a result , @xmath121 @xmath122 . using the fact that @xmath123 ( when @xmath124 ) , @xmath125 . thus , @xmath126 . clearly , when @xmath114 and @xmath82 are the same , the value of @xmath111 increases when @xmath117 increases . in case 2 , let @xmath115 be @xmath116 , for a scaling factor of @xmath117 . clearly , @xmath118 , @xmath127 . thus , @xmath128 . also , @xmath129 . using the approximation mentioned in case 1 , we see that @xmath130 . therefore , @xmath131 . in this case , when @xmath114 and @xmath82 are the same , the value of @xmath132 increases when @xmath117 decreases . but for a given reference point , and two input points @xmath39 and @xmath40 , we do not know which of the two cases will hold . but we can expect that half of the randomly chosen reference points will fall under case 1 and the other half will be expected to fall under case 2 . if we only employ a scaling factor @xmath117 that is greater than one , then , for an expected half of the reference points we expect to see an improvement in the estimate of a lower bound for @xmath133 . this explains why our algorithm performs better than mk . the above analysis can also be used to better understand the mk algorithm . in this problem we are given @xmath2 points @xmath134 in @xmath14 and a radius @xmath0 ( which is a real number ) and the problem is to identify the @xmath0-neighborhood of each input point . if @xmath1 is an input point , its @xmath0-neighborhood is defined to be the set of all input points that are within a distance of @xmath0 from @xmath1 . frnnp has numerous applications . one of the applications of vital importance is that of molecular simulations . we can modify mpr to solve this problem as well . the modified version is given below . let @xmath135 denote the @xmath0-neighborhood of @xmath136 . * algorithm * mpr - frnns + _ input : _ @xmath22 and @xmath0 , where each @xmath87 is a point in @xmath14 ( for @xmath88 ) and @xmath0 is a real number . input are also @xmath89 and @xmath86 , where @xmath89 is the number of references and @xmath86 is the projection factor . + _ output : _ @xmath135 , for @xmath88 . + * 1 ) * pick @xmath89 random points of @xmath23 as references ; project these + references by multiplying each coordinate of each reference + point by @xmath86 . let these projected references be @xmath90 . + * 2 ) * compute the distance between every point of @xmath23 and + every projected reference point . + * 3 ) * sort the points of @xmath23 with respect to their distances + to @xmath63 . let the sorted points be @xmath137 . + * 4 ) for * @xmath94 * to * @xmath2 * do * @xmath138 ; + * 5 ) for * @xmath94 * to * @xmath2 * do * + @xmath96 * to * @xmath2 * do * + @xmath97 ; + @xmath98 * to * @xmath89 * do * + @xmath139 * then * + @xmath100 ; * exit * ; + @xmath101 * then exit else * + compute @xmath102 ; + @xmath140 * then * add @xmath141 to @xmath135 and + add @xmath142 to @xmath143 ; + * 6 ) output * @xmath144 for @xmath88 . a typical algorithm in the literature for cpp has two phases . in the first phase pairs of points that can not possibly be the closest are eliminated . in the second phase distance is computed between every pair of points that survive the first phase . the time spent in the first phase is typically very small and hence is negligible ( compared to the time spent in the second phase ) . also , the time needed to process the pairs in the second phase is linear in the number of surviving pairs . as a result , it suffices to report the number of pairs ( to be processed in the second phase ) as a measure of performance ( see e.g. , @xcite ) . in this paper also we use this measure of performance throughout . we have experimentally compared the performance of mk and mpr on different data sets . the machine we have used has an intel(r ) core(tm ) i7 - 2640 m 2.8 ghz cpu with 8 gb ram running windows 7 ( 64 bit ) . the same machine has been used for all the experiments reported in this paper . as mentioned in @xcite , random walk data set is the most difficult case for time series mining algorithms since the probability of the existence of very close motif pairs is very low . we have also used the same data for our comparison . in particular , we have used 10 different random walk data sets of sizes ranging from 10k to 100k . we have also varied the motif length to see how the performances change . our algorithm performs better than mk for higher motif lengths . both the algorithms have been run 10 times and the averages computed . we do not perform any comparison with the brute force method as that has already been done in @xcite . in table [ table1 ] we show the number of pairs processed ( in the second phase ) in mk and mpr . the size ( i.e. , the length ) of the time series data varies from 10k to 100k , the motif length being 1024 . the following parameter values have been used : @xmath145 and @xmath146 . from this table we see that mk processes around 1.5 times the number of pairs processed by mpr . figure [ figure2 ] presents a graphical plot of the runtime requirements of mk and mpr algorithms . this figure shows that the run time of mk is around 1.5 times the run time of mpr . this improvement is quite significant for the following reason : there are two ideas used in mk , namely , early abandoning and the use of random reference points . as the authors point out in @xcite , the difference between using early abandoning alone and both the ideas is small , especially on random walk data sets . however , mk performs much better than using early abandoning alone on real datasets . table [ table2 ] and figure [ figure3 ] show how the number of pairs reduces with an increase in the number @xmath89 of the reference points for mk and mpr algorithms . from these , we note that as @xmath89 increases , the difference between mk and mpr widens . the bar graph in figure [ figure3 ] pictorially represents this comparison . the blue and red bars represent the number of pairs processed by mk and mpr , respectively . we have run both mk and mpr algorithms on some of the real data sets from http://www.cs.ucr.edu/~mueen / mk/. table [ table3 ] shows the performance of mk and mpr algorithms . on dataset 1 , mpr is around 2 times faster than mk and on dataset 2 , mpr is around 1.5 times faster than mk . in this table , ` - ' indicates that the algorithm did not stop within 40 minutes . in this section we consider the cpp when the space is one of character strings and the metric is hamming distance . an algorithm for this version of cpp has numerous applications including approximate tsmm ( see e.g. , @xcite ) . when the alphabet is @xmath147 , the light bulb algorithm of @xcite can be used to solve this problem . in this section we show how to modify the light bulb algorithm for the case of generic alphabets . before presenting details on the modification , we provide a brief summary of the light bulb problem . the light bulb problem is that of identifying the most correlated pair of bulbs from out of @xmath2 given bulbs @xmath148 . this problem is solved by observing the state of each bulb in @xmath149 discrete time steps ( for some relevant value of @xmath149 ) . the states of bulb @xmath142 in these @xmath149 time steps can be represented as a vector @xmath150 ( for @xmath88 ) . we can think of @xmath151 as a sample from a probability distribution that governs the state of bulb @xmath142 . it can be shown that if @xmath149 is sufficiently large ( e.g. , @xmath152 ) , then the pair of bulbs that is the most correlated in the samples is also the most correlated pair with high probability . thus the light bulb problem can be stated as follows : we are given @xmath2 boolean vectors @xmath153 . the problem is to find the pair of vectors that are the most similar ( i.e. , the hamming distance between them is the smallest ) . from hereon , we will use this formulation of the problem . note that , given two vectors , we can find the hamming distance between them in @xmath154 time . a straight forward algorithm to identify the most correlated pair of bulbs takes @xmath155 time . this algorithm computes the hamming distance between every pair of bulbs . the algorithm of @xcite takes subquadratic time . in particular , the expected run time of this algorithm is @xmath156 assuming that @xmath157 . here , @xmath158 is the correlation between the most correlated pair of bulbs and @xmath159 is the correlation between the second most correlated pair of bulbs . note that if the correlation between two bulbs @xmath142 and @xmath141 is @xmath160 then the expected hamming distance between @xmath151 and @xmath161 is @xmath162 . equivalently , the similarity ( i.e. , the number of matches ) between @xmath151 and @xmath161 is @xmath163 . consider a matrix @xmath164 of size @xmath165 , such that the @xmath142th row of @xmath164 is @xmath151 , for @xmath88 . the algorithm of @xcite iteratively collects pairs of bulbs that are candidates to be the most correlated . once it collects enough pairs , it computes the distance between each pair in this collection and outputs the closest . there are @xmath166 iterations in the algorithm and in each iteration , some candidate pairs are generated and added to the collection @xmath13 . in any iteration , the algorithm picks @xmath167 columns of @xmath164 at random ( for some constant @xmath114 ) . the rows are sorted based on the characters in the randomly chosen columns . as a result of this sorting , the bulbs get partitioned into buckets such that all the bulbs with equal values ( in the @xmath167 random columns ) fall into the same bucket . a pair of bulbs @xmath168 will get added to @xmath13 in any iteration if they fall into the same bucket in this iteration . the authors of @xcite show that after @xmath166 iterations , @xmath13 will have the most correlated pair of bulbs with high probability ( i.e. , with a probability of @xmath169 ) . the above algorithm has been proposed for the case of binary strings . we can modify this algorithm to handle the case of an arbitrary ( finite ) alphabet and get the following theorem . [ theorem0 ] let @xmath164 be a matrix of size @xmath165 . each entry in this matrix is an element from some set @xmath30 of cardinality @xmath170 . we can find the most correlated pair of columns of @xmath164 in an expected @xmath171 time where @xmath158 is the correlation between the most correlated pair of columns , @xmath159 is the correlation between the second most correlated pair of columns , and @xmath172 is the word length of the machine . this expected run time will be @xmath173 if we use a general sorting algorithm . ( here correlation is based on hamming distance . for example , @xmath158 is the largest fraction of rows in which any two columns agree ) . * proof : * the only difference in the algorithm is that instead of sorting binary strings we will have to sort strings from an arbitrary alphabet . without loss of generality , let @xmath174 be the alphabet under concern . in the original algorithm , one has to sort @xmath2 @xmath175-bit integers in every iteration . for a generic alphabet , we have to sort @xmath2 @xmath176-bit integers . if one uses any comparison based sorting algorithm , this sorting takes @xmath6 time . if we use an integer sorting algorithm , this sorting can be done in @xmath177 time where @xmath172 is the word length of the machine . this is the time spent in each iteration of the algorithm . therefore , the total expected run time is @xmath171 . @xmath178 let this modified version of the light bulb algorithm be called mlba . the algorithm of @xcite for approximate tsmm partitions the input time series data @xmath23 based on a window of size @xmath172 ( for an appropriate value of @xmath172 ) , computes the mean of every window , and discretizes the mean into four possible values . as a result , the time series data is transformed into a string @xmath179 of characters form the alphabet @xmath180 . it then uses any @xmath15-motif finding algorithm to find the motifs in @xmath179 . however , all the exact algorithms for finding @xmath15-motifs take time that is exponential on @xmath12 and @xmath10 . note that the last step of finding @xmath15 motifs can be replaced with a problem of finding time series motifs in @xmath179 which is nothing but cpp in the domain of strings of characters , the motif length being @xmath12 . one could employ mk to solve cpp in the domain of character strings . the only difference is that we have to replace euclidean distance with hamming distance . we have implemented this algorithm . it turns out that mk does not perform well for the case of hamming distance . to be fair , the authors of mk have not tested mk for this case . we have compared mk with mlba and the results are shown in table [ table4 ] . as this table reveals , mlba is around 200 times faster than mk . it is also clear that if we employ mlba in place of @xmath181-motif finding algorithms , the performance of the approximate tsmm algorithm given in @xcite will improve significantly . figure [ figure4 ] shows a runtime comparison of mk and mlba for the case of character strings from a finite alphabet . the light bulb algorithm of @xcite identifies the closest pair of strings , from out of @xmath2 given binary strings . an interesting question is if we can use the same algorithm to identify the * furthest * pair of strings . this problem has relevance in many problems including the two locus problem in gwas . the authors of @xcite present an elegant adaptation of the light bulb algorithm to solve this problem when the strings are binary . they also show how to solve this problem for arbitrary alphabets using locality sensitive hashing ( lsh ) @xcite . they map the input strings into binary strings using lsh . lsh closely preserves similarities with a high probability . in this section we show how to avoid lsh . in particular , we present novel deterministic mappings of the input strings to binary strings such that similarities are preserved deterministically . our experimental comparison shows that our algorithm has a significantly better run time than that of @xcite . let @xmath182 stand for the number of matches between two strings ( of equal length ) @xmath47 and @xmath48 . for instance , if @xmath183 and @xmath184 , then @xmath185 ( since they match in positions 2 and 4 ) . let @xmath186 and @xmath187 be two sequences of strings ( each string having the same length ) . we define @xmath188 to be @xmath189 . consider the sequences @xmath190 , for @xmath88 , where each @xmath191 is @xmath192 or @xmath193 ( for @xmath194 ) . note that each @xmath195 is a sequence of strings where each string is of length 1 . let @xmath196 . each @xmath195 can be thought of as a string from the alphabet @xmath197 . in the application of gwas , we can let @xmath195 correspond to the snp @xmath142 , for @xmath88 . specifically , @xmath191 is the value of the @xmath142th snp in subject @xmath141 , for @xmath194 . if we are interested in finding the two most correlated snps , then we can use mlba to identify this pair ( as shown in section [ sec4 ] ) . on the other hand , if our goal is to identify the least correlated pair , then , it is not clear how to do this using mlba . to solve the two locus gwas problem , we have to identify not only the most correlated pair of bulbs but also the least correlated pair . the authors of @xcite present an elegant solution for this problem when each @xmath195 has only zeros and ones . in this case , each @xmath195 can be thought of as a light bulb . the idea is to construct a matrix @xmath198 of size @xmath199 where each column of @xmath198 corresponds to either a bulb or its ` complement ' , specifically , the first @xmath2 columns correspond to the bulbs and the next @xmath2 columns correspond to the complements of the bulbs . in other words , @xmath200=a_j^i$ ] , for @xmath201 and @xmath200=\bar{a}_j^i$ ] for @xmath202 . here , if @xmath47 is any bit , then , @xmath203 denotes its complement . let @xmath204 and @xmath205 . the algorithm of @xcite for finding the least correlated pair works as follows . consider all the pairs of columns @xmath168 such that @xmath206 and @xmath207 . from out of these pairs , identify the pair @xmath208 of columns with the maximum number of matches . if @xmath209 and @xmath210 , then @xmath106 is the least correlated pair of bulbs . finding such a pair @xmath208 can be done using the light bulb algorithm of @xcite . the correctness of this algorithm follows from the fact that if the two bulbs @xmath142 and @xmath141 have the least number of matches , then , column @xmath142 and the complement of column @xmath141 will have the most number of matches . it is not clear how to extend the above idea when the sequences have three ( or more ) possible elements . the authors of @xcite reduce such general cases to the case of zeros and ones using locality sensitive hashing ( lsh ) . the measure of correlation used by @xcite is different from what we use in this paper . we define the correlation between two strings @xmath195 and @xmath211 as @xmath212 . in contrast , @xcite use pearson s correlation coefficient . in this section we present an elegant algorithm for the problem of identifying the least correlated pair of strings without employing lsh . the idea of @xcite is to map input strings into boolean vectors . if @xmath142 and @xmath141 are any two strings , then the sequences @xmath195 and @xmath211 are mapped to boolean vectors @xmath213 and @xmath214 by lsh such that the distance between @xmath195 and @xmath211 will be nearly the same as the distance between @xmath213 and @xmath215 with some probability . the larger the length of @xmath213 is , the better will be the accuracy of lsh in preserving distances . our algorithm also maps each @xmath195 into a boolean vector @xmath213 deterministically such that @xmath216 , for @xmath88 . consider an alphabet @xmath30 with three strings where @xmath217 . clearly , @xmath218 if @xmath219 and @xmath220 if @xmath221 for any @xmath222 . also , @xmath223 if @xmath219 and @xmath224 if @xmath221 . here @xmath225 stands for the string obtained from @xmath48 by complementing each bit . for example , if @xmath226 then @xmath227 . consider the sequences @xmath190 , for @xmath88 , where each @xmath191 is @xmath192 or @xmath193 ( for @xmath194 ) . note that each @xmath195 is a sequence of strings where each string is of length 1 . let @xmath196 . assume now that we encode each @xmath191 as follows ( for @xmath88 and @xmath194 ) : @xmath228 : and @xmath229 . let the encoded version of @xmath195 be denoted as @xmath213 for @xmath88 . note that @xmath230 , for any @xmath88 . it is easy to see that @xmath231 , for any @xmath142 and @xmath141 ( @xmath232 ) . for any @xmath233 , let @xmath234 , for @xmath88 . clearly , @xmath235 for any @xmath232 . clearly , the following statement is true : if , from out of all the pairs of strings , @xmath106 has the largest correlation , i.e. , @xmath236 is the largest , then from out of all the boolean vectors generated , @xmath213 and @xmath215 will have the largest correlation . also , if @xmath236 is the smallest , then , @xmath213 and @xmath237 will have the largest correlation ( from out of the pairs @xmath238 ) . we can now form a matrix @xmath198 of size @xmath239 where the first @xmath2 columns correspond to ( transformed ) strings and the next @xmath2 columns correspond to complements of ( transformed ) strings . let @xmath204 and @xmath205 . consider all the pairs of columns @xmath168 such that @xmath206 and @xmath207 . from out of these pairs , identify the pair @xmath208 of columns with the maximum number of matches . if @xmath209 and @xmath210 , then @xmath106 is the least correlated pair of strings . finding such a pair @xmath208 can be done using the light bulb algorithm of @xcite . [ theorem1 ] given @xmath2 strings , we can find the closest pair of strings in an expected time of @xmath156 , where @xmath158 and @xmath159 are the largest and the second largest correlation values , respectively . also , we can find the least correlated pair of strings in an expected time of @xmath240 , where @xmath241 and @xmath242 are the smallest and the next smallest correlation values , respectively . * proof : * when we transform input strings to binary sequences , the ordering of pairs is preserved in terms of correlations as we have shown before . let @xmath158 be the correlation of the largest correlated pair and @xmath159 be the correlation of the second largest correlated pair . how do these values change in the transformed domain ? if @xmath243 and @xmath244 are the transformed values of these correlations , respectively , it can be seen that @xmath245 and @xmath246 . if @xmath241 and @xmath242 are the correlations of the smallest and the second smallest correlated pairs , respectively , and if @xmath247 and @xmath248 are the transformed values of these , respectively , then we can see that : @xmath249 and @xmath250 . to find the largest correlated pair , we can use mlba ( theorem [ theorem0 ] ) . we use the mapping only to find the least correlated pair . @xmath178 we have thus far considered the case where the alphabet is @xmath197 . we can extend the mapping to a general alphabet and get the following theorem . given @xmath2 strings , we can find the largest correlated pair of strings in an expected time of @xmath156 , where @xmath158 and @xmath159 are the largest and the second largest correlation values , respectively . also , we can find the least correlated pair of strings in an expected time of @xmath251 , where @xmath241 and @xmath242 are the smallest and the next smallest correlation values , respectively , and there are @xmath170 characters in the alphabet . * proof : * consider sequences from the alphabet @xmath252 . in this case we map each element of this alphabet to a binary string of length @xmath170 where there is only one 1 . specifically , we use the following mapping : @xmath253 etc . as before , we do nt need any mapping if our goal is to find the largest correlated pair . the mapping is used only to find the least correlated pair . @xmath178 we can improve the above theorem by employing a random mapping as follows : we will use a binary string of length @xmath170 to encode each symbol in the alphabet . the encoding for each symbol is obtained by ( uniformly ) randomly choosing each bit in the string ( of length @xmath170 ) . let @xmath47 and @xmath48 be any two symbols in the alphabet ( with @xmath221 ) and let @xmath254 and @xmath255 be their encodings , respectively . then , clearly , the expected value of @xmath256 is @xmath257 . also , the expected value of @xmath258 is @xmath257 . if @xmath114 is the correlation between a pair of strings and if @xmath259 is the transformed value , then , it follows that the expected value of @xmath259 is @xmath260 . an application of the chernoff bounds will readily imply that the value of @xmath259 will indeed be very close to this expected value with a probability of @xmath261 . therefore , we get : given @xmath2 strings , we can find the least correlated pair of strings in an expected time of @xmath262 , where @xmath241 and @xmath242 are the smallest and the next smallest correlation values , respectively , and there are @xmath170 characters in the alphabet . @xmath178 the two locus association problem is defined as follows . input is a matrix @xmath164 of size @xmath263 where @xmath264 is the number of patients ( subjects ) each with @xmath2 snps . here @xmath265 is the number of cases and @xmath266 is the number of controls . there are three possible values for each snp , namely , @xmath192 or @xmath193 . the cases are of phenotype @xmath267 and the controls are of phenotype @xmath268 . rows @xmath267 through @xmath265 of @xmath164 correspond to cases . let this submatrix be called @xmath39 . rows @xmath269 through @xmath270 of @xmath164 correspond to controls and let this submatrix be called @xmath40 . each column of @xmath164 corresponds to an snp . the two locus association problem is to identify the pair of snps whose statistical correlation with phenotype is maximally different between cases and controls . as mentioned in @xcite , the goal is to identify the pair : @xmath271 if @xmath272 is any matrix , then , @xmath273 stands for the correlation between the columns @xmath142 and @xmath141 of @xmath272 . the algorithm of @xcite exploits the light bulb algorithm of @xcite and locality sensitive hashing ( lsh ) @xcite . they use lsh to transform matrices @xmath39 and @xmath40 to @xmath274 and @xmath275 , respectively . in particular , each column @xmath276 of @xmath39 is converted to a column @xmath277 of zeros and ones . the size of @xmath276 is @xmath278 and the size of @xmath277 is chosen to be @xmath279 . the matrix @xmath40 is also transformed into @xmath275 in a similar manner using lsh . followed by this , the pair of interest is identified . to be precise , using @xmath274 and @xmath275 , the matrix @xmath198 is formed where @xmath280\ ] ] where @xmath281 is obtained from @xmath275 by complementing every element of @xmath275 . note that @xmath198 is of size @xmath282 . let @xmath283 and @xmath284 . consider all the pairs of columns @xmath106 such that @xmath285 and @xmath286 . from out of these pairs , identify the pair @xmath287 of columns with the maximum number of matches . if @xmath288 and @xmath289 , then @xmath168 is the pair of interest . this pair can be found using the light bulb algorithm of @xcite . we can use our mapping ideas to get the following theorem . [ theorem6.1 ] we can find the pair @xmath106 of snps that maximizes @xmath290 in an expected time of @xmath291 , where @xmath158 and @xmath159 are the smallest and the next smallest values of @xmath290 , respectively , over all possible pairs @xmath106 of snps . * proof : * our algorithm also uses the same method except that instead of using lsh to map @xmath39 and @xmath40 to @xmath274 and @xmath275 , respectively , we employ the deterministic mapping we have proposed in section [ mapping ] . let @xmath142 and @xmath141 be two snps ( i.e. , two columns in @xmath39 and @xmath40 ) . let @xmath292 and @xmath293 . now consider columns @xmath142 and @xmath294 of @xmath198 . what can we say about the correlation of these columns ? from the discussion in section [ mapanalysis ] , we realize that this correlation is @xmath295 . this also proves the correctness of our algorithm . let @xmath158 be the maximum value of @xmath290 over all possible pairs @xmath106 and let @xmath159 be the second largest value . then , the run time follows from theorem [ theorem0 ] . @xmath178 in a similar manner we can also find the pair that maximizes @xmath296 and hence identify the pair that maximizes @xmath297 . the notion of similarity ( between two snps ) used in @xcite is pearson s correlation coefficient . in this paper the similarity we use is based on the hamming distance . specifically , the complement of the hamming distance is the measure of similarity we employ . the authors of @xcite have tested their algorithms on different data sets ( including random data ) . since we do not have access to either these data sets or their algorithms , the only comparison we can do was on the random data . as explained in @xcite , we have also generated snps from binomial distributions . in particular , for each subject , the value of each snp is chosen uniformly randomly to be either 0 or 1 with equal probability . this dataset is called noise data in @xcite ( c.f . table [ table6 ] in @xcite ) . like in @xcite , we have also generated data of sizes @xmath74k , @xmath298k , and @xmath299k . for each size we have generated two different data sets and computed the average number of pairs to be processed . we compare these numbers with the ones reported in @xcite . as can be seen from table [ table5 ] , our algorithm is around 4 times faster than the one in @xcite . note that this a significant improvement since the typical processing times for the two locus problem are quite high . for example , the authors of @xcite report that on some of the data sets ( with no more than @xmath17 snps ) , the brute force algorithm for the two locus problem took several days on 1000 cpus ! thus any improvement in the run time could make a noticeable difference . how does one ensure that the output of an algorithm for the two locus problem is correct ? for small data sizes , one could run the exhaustive brute force algorithm to identify the correct pair and use it to verify correctness . in fact when the number of snps is either @xmath74k or @xmath298k , we first found the correct answer and then used it to measure the run time of our algorithm as follows . we ll run our algorithm one iteration at a time until the correct pair(s ) is ( are ) picked up by our algorithm . at this point we will stop and report the total number of pairs collected . the numbers shown in table [ table5 ] have been obtained in this manner . we could not use this method for @xmath300k , since the brute force algorithm was taking too much time . when @xmath2 is very large , we _ inject _ pairs with known correlations . as an example , consider the problem of finding the largest correlated pair of columns in a @xmath301 matrix @xmath39 . say we generate each column by picking each element to be either 0 or 1 with equal probability . for any two columns , clearly , the expected correlation is @xmath302 . we can perform a probabilistic analysis to get a high probability bound on the largest correlation between any two columns . one could also get this estimate empirically . for example , for @xmath303 , we generated several random data sets and computed the largest correlation in each and calculated an average . the average maximum correlation was @xmath304 . let @xmath1 be this value . to inject a pair with a correlation of @xmath305 where @xmath305 is @xmath306 we generate a column @xmath307 with all ones and another column @xmath308 with @xmath309 ones and @xmath310 zeros . we then replace ( any ) two columns of @xmath39 with @xmath307 and @xmath308 . clearly , the correlation between @xmath307 and @xmath308 is @xmath305 . the expected correlation between @xmath307 and any other column of @xmath39 ( other than @xmath308 ) is @xmath302 . similarly , the expected correlation between @xmath308 any other column of @xmath39 ( other than @xmath307 ) is @xmath302 . thus the pair @xmath168 is likely to be the winner with high probability . we stop our algorithm when this pair is picked by our algorithm . we have picked a value for @xmath305 that is only slightly larger than @xmath1 so as to get an accurate estimate on the run time . we have used a similar technique to inject pairs for the two locus problem as well . in table [ table6 ] we show the results for our algorithm . in these cases we have generated the snps randomly from a binomial distribution as before . we have employed 200 cases and 200 controls . clearly , the expected value of @xmath311 is zero for this data . if we map this data using our deterministic mapping , then the expected correlation between any two snps will be @xmath302 . here again we empirically found that the largest correlation was around 58.3% ( when @xmath2 was @xmath74k ) . therefore , we have injected a pair whose correlation was @xmath312% . this pair is likely to be the winner . our algorithm was run until this pair was picked . at this time , the algorithm was stopped . we also checked if the algorithm picked any pair whose correlation was better than that of the injected one and found none . in all the datasets we tried , we were always able to find a pair ( other than the injected one ) whose correlation was very close to 60% and hence the numbers shown in table [ table6 ] are very close to the case with no injections . * practical considerations : * another important question is for how long we should run the program before we can be sure that the correct pair has been obtained ( with a high confidence ) . please note that we do not know the values of @xmath158 and @xmath159 ( c.f . theorem [ theorem6.1 ] ) . theorem [ theorem6.1 ] suggests that the run time of our algorithm is @xmath313 for some relevant @xmath314 . we can empirically estimate @xmath314 . the idea is to measure the run time of the algorithm for various values of @xmath2 ( as explained above ) , the maximum value of @xmath2 being as much as possible ( for the given computing platform and time constraints ) . then we could use any numerical procedure to estimate @xmath314 . in this paper we have presented novel algorithms for the closest pair problem ( cpp ) . cpp is a ubiquitous problem that has numerous applications in varied domains . we have offered algorithms for the cases of euclidean as well as hamming distances . we have applied our algorithms for two well studied and important problems , namely , time series motif mining and two locus genome wide association study . our algorithms significantly improve the best - known algorithms for these problems . specifically , we improve the results presented in many prior papers including @xcite , @xcite , and @xcite . this research has been supported in part by the nih grant nih - r01-lm-010101 . p. achlioptas , b. scholkopf and k. borgwardt , two - locus association mapping in subquadratic runtime , acm sigkdd international conference on knowledge discovery and data mining ( kdd ) 2011 . aston , d.a . ralph , d.p . lalo , s. manjeshwar , b.a . gramling , d.c . defreese , a.d . west , d.e . branam , l.f . thompson , m.a . craft , d.s . mitchell , c.d . shimasaki , j.j . mulvihill , and e.r . jupe , oligogenic combinations associated with breast cancer risk in women under 53 years of age , _ human genetics _ , 116(3):208 - 221 , feb . p. beaudoin , m. van de panne , p. poulin and s. coros , motion - motif graphs , symposium on computer animation 2008 . m. s. charikar , similarity estimation techniques from rounding algorithms , proc . acm symposium on theory of computing ( stoc ) , 2002 . b. chiu , e. keogh , and s. lonardi , probabilistic discovery of time series motifs , proc . of the 9th international conference on knowledge discovery and data mining ( kdd ) , pp . 493 - 498 , 2003 . cho , d.l . nicolae , l.h . gold , c.t . fields , m.c . labuda , p.m. rohal , m.r . pickles , l. qin , y. fu , j s. mann , b.s . kirschner , e.w . jabs , j. weber , s.b . hanauer , t.m . bayless , and s.r . identification of novel susceptibility loci for inflammatory bowel disease on chromosomes 1p , 3q , and 4q : evidence for epistasis between 1p and ibd1 proceedings of the national academy of sciences of the united states of america , 95(13):7502 - 7507 , june 1998 . t. h. consortium . a second generation human haplotype map of over 3.1 million snps . nature , 449(7164):851 - 61 , oct . 2007 . h. j. cordell . detecting gene - gene interactions that underlie human diseases . nat rev genet , 10(6):392 - 404 , june 2009 . cox , m. frigge , d.l . nicolae , p. concannon , c.l . hanis , g.i . bell , and a. kong , loci on chromosomes 2 ( niddm1 ) and 15 interact to increase susceptibility to diabetes in mexican americans , nature genetics , 21(2):213 - 215 , feb . 1999 . s. fortune and j. s. hopcroft , _ a note on rabin s nearest - neighbor algorithm _ , information processing letters , 8(1 ) , pp . 20 - 23 , 1979 . t. guyet , c. garbay and m. dojat , knowledge construction from time series data using a collaborative exploration system , journal of biomedical informatics 40(6 ) : 672 - 687 ( 2007 ) . s. khuller and y. matias , _ a simple randomized sieve algorithm for the closest - pair problem _ , information and computation , vol 188 ( 1 ) , pp . 3437 , 1995 . j. meng , j.yuan , m. hans and y. wu , mining motifs from human motion , proc . of eurographics , 2008 . d. minnen , c.l . isbell , i. essa , and t. starner , discovering multivariate motifs using subsequence density estimation and greedy mixture learning , proc . 22nd conference on artificial intelligence ( aaai?07 ) , 2007 . a. mueen , e. keogh , q. zhu , s. cash , b. westover , _ `` exact discovery of time series motifs '' _ , siam international conference on data mining ( sdm ) 2009 . s. k. musani , d. shriner , n. liu , r. feng , c. s. coffey , n. yi , h. k. tiwari , and d. b. allison . detection of gene x gene interactions in genome - wide association studies of human population data . human heredity , 63(2):67 - 84 , 2007 . r. nakamichi , y. ukai , and h. kishino , detection of closely linked multiple quantitative trait loci using a genetic algorithm , genetics , 158(1):463 - 475 , may 2001 . r. paturi , s. rajasekaran , and j. reif . the light bulb problem . in proc . workshop on comput . learning theory , pp . 261 - 268 , san mateo , ca , 1989 . morgan kaufmann . f. preparata and m. shamos , _ computational geometry _ , springer verlag , 1986 . m. rabin , _ probabilistic algorithms _ , algorithms and complexity , recent results and new directions , academic press , pp 21 - 39 , 1976 . s. rombo and g. terracina , discovering representative models in large time series databases , proc . 6th international conference on flexible query answering systems , pp . 84 - 97 , 2004 . y. tanaka , k. iwamoto , and k. uehara , discovery of time - series motif from multi - dimensional data based on mdl principle , machine learning , 58(2 - 3):269 - 300 , 2005 . s. tata , declarative querying for biological sequences , ph.d thesis , the university of michigan , 2007 . ( advisor jignesh m. patel ) . y. wang , x. liu , k. robbins , and r. rekaya , antepiseeker : detecting epistatic interactions for case - control studies using a two - stage ant colony optimization algorithm , bmc bioinformatics , 3:117 - 117 , 2010 . j. xu , c.d . langefeld , s.l . zheng , e.m . gillanders , b. chang , s.d . isaacs , a.h . williams , k e. wiley , l. dimitrov , d.a . meyers , p.c . walsh , j.m . trent , and w.b . isaacs , interaction effect of pten and cdkn1b chromosomal regions on prostate cancer linkage , human genetics , 115(3):255 - 262 , aug . lower bounds for algebric computation trees with integer inputs _ , siam j. comput . , 20:4 , pp . 655 - 668 , 1991 . x. zhang , s. huang , f. zou , and w. wang . team : efficient two - locus epistasis tests in human genome - wide association study . bioinformatics ( oxford , england ) , 26(12):i217 - 227 , june 2010 . x. zhang , f. zou , and w. wang , fastanova : an efficient algorithm for genome - wide association study , in proc . 14th acm sigkdd international conference on knowledge discovery and data mining ( kdd ) , pages 821 - 829 , las vegas , nevada , usa , 2008 .

we present an elemental - abundance analysis , in the near - ultraviolet ( nuv ) spectral range , for the extremely metal - poor star , a 9th magnitude subgiant with [ fe / h ] and enhanced carbon , based on data acquired with the space telescope imaging spectrograph on the _ hubble space telescope_. this star is the brightest example of a class of objects that , unlike the great majority of carbon - enhanced metal - poor ( cemp ) stars , does not exhibit over - abundances of heavy neutron - capture elements ( cemp - no ) . in this paper , we validate the abundance determinations for a number of species that were previously studied in the optical region , and obtain strong upper limits for beryllium and boron , as well as for neutron - capture elements from zirconium to platinum , many of which are not accessible from ground - based spectra . the boron upper limit we obtain for , , the first such measurement for a cemp star , is the lowest yet found for very and extremely metal - poor stars . in addition , we obtain even lower upper limits on the abundances of beryllium , , and lead , ( [ pb / fe] ) , than those reported by previous analyses in the optical range . taken together with the previously measured low abundance of lithium , the very low upper limits on be and b suggest that was formed at a very early time , and that it could well be a bona - fide second - generation star . finally , the pb upper limit strengthens the argument for non--process production of the heavy - element abundance patterns in cemp - no stars .

1406.0538

carbon - enhanced metal - poor ( cemp ) stars are a subset of metal - poor ( mp ; [ fe / h ] = @xmath6 , where @xmath7 is the number density of atoms of a given element , and the indices refer to the star ( @xmath8 ) and the sun ( @xmath9).]@xmath101.0 , e.g. , * ? ? ? * ; * ? ? ? * ) and very metal - poor ( vmp ; [ fe / h ] @xmath11 ) stars that exhibit elevated carbon relative to iron , [ c / fe ] , sometimes referred to as carbonicity ( [ c / fe]@xmath121.0 ; * ? ? ? it has recently been recognized that a more appropriate division on [ c / fe ] for the identification of cemp stars is at somewhat lower carbonicity , e.g. , [ c / fe ] @xmath130.7 @xcite . in the past few decades , it has become clear that such stars comprise a significant fraction of all vmp stars ( @xmath14 10 - 20% ; @xcite ) , one that increases strongly with declining metallicity , from 30% for [ fe / h ] @xmath153.0 , to 40% for [ fe / h ] @xmath153.5 , @xmath1675% for [ fe / h ] @xmath154.0 and 100@xmath17 for [ fe / h]@xmath185.0 . this trend has been confirmed with the many thousands of cemp stars identified by @xcite from the sloan digital sky survey ( sdss ; * ? ? ? * ) , and its extensions segue-1 ( sloan extension for galactic exploration and understanding ; * ? ? ? * ) and segue-2 ( c. rockosi et al . , in preparation ) . for most cemp stars there exists a clear correlation between carbon enhancement and the over - abundance of elements produced by @xmath5-process nucleosynthesis , such as ba ( cemp-@xmath5 stars ; see * ? ? ? this behavior is consistent with the hypothesis that these enhancements ( both for carbon and elements produced by the @xmath5-process ) are due to nucleosynthesis processes that took place during the asymptotic giant - branch ( agb ) stage of stellar evolution ( e.g. , * ? ? ? * ; * ? ? ? the resulting abundance pattern can arise from the star itself ( which should rarely be found , but see * ? ? ? * ) or by a now - extinct binary companion that transferred material to the surviving ( observed ) companion @xcite . multi - epoch radial - velocity measurements by @xcite , @xcite , and @xcite demonstrated that the frequency of detected binaries among equivalent population i ba stars and population ii ch stars indicated that essentially all are members of binary systems . @xcite conducted a similar study for members of the more metal - deficient cemp-@xmath5 sub - class of stars , and reached the same conclusion . an intriguing variation on this behavior was initially recognized by @xcite and @xcite . the cemp stars cs 22948 - 027 and cs 29497 - 034 were found not only to be rich in the elements commonly produced by @xmath5-process nucleosynthesis , such as sr , y , ba , and la , but also in eu , an element that , for extremely low - metallicity stars , is more likely produced by @xmath19-process nucleosynthesis . additional studies by many groups have now identified @xmath1450 of these so - called cemp - r / s stars . these cases , once thought to be the rare exceptions , are as commonly represented among cemp stars as the `` @xmath5-only '' variety . the origin of the abundance patterns of the cemp-@xmath20 stars is not yet clear , and many scenarios have been proposed @xcite . for example , an association with a @xmath21ne neutron source in intermediate - mass agb stars has been suggested for progenitors of cemp-@xmath20 stars , rather than the @xmath22c neutron source thought to be active for low - mass agb stars , the likely progenitors of the cemp - s stars ( @xcite , hollek et al . 2014 , submitted ) . mass transfer from a companion that passed through the agb phase of stellar evolution has been suggested previously as an explanation for the cemp-@xmath5 and cemp-@xmath20 classes of stars , based on the high fraction of such stars found in binary systems @xcite . the story has become richer still . @xcite , and others since , have shown that the correlation between carbon enhancement and the over - abundances of @xmath5- or @xmath20-elements no longer persists ( or at least is different in nature ) for the majority of cemp stars with [ fe / h ] @xmath182.7 . these so - called cemp - no stars ( indicating no enhancement of neutron - capture elements ) suggest that a variety of carbon - producing mechanisms , other than that associated with agb stars , may have played a role in the early universe . possible progenitors for this sub - class include massive , rapidly rotating , mega metal - poor ( mmp ; [ fe / h ] @xmath186.0 ) stars , sometimes referred to as `` spinstars '' @xcite , which models suggest have greatly enhanced abundances of cno due to distinctive internal burning and mixing episodes , followed by strong mass loss @xcite . another possible scenario is pollution of the ism by so - called faint supernovae associated with the first generations of stars , which experience extensive mixing and fallback during their explosions @xcite . although more data are desired for cemp - no stars , @xcite report that the fraction of binaries among stars within this sub - class is no higher than expected for random samples of vmp giants . @xcite , @xcite , @xcite , and j. andersen et al . ( in preparation ) reach similar conclusions . thus , contribution of material from an evolved binary companion is apparently not required in order to form cemp - no stars . the recently reported extremely metal - poor damped lyman-@xmath23 system by @xcite ( [ fe / h]@xmath243.0 ) exhibits enhanced carbon ( [ c / fe ] = @xmath251.5 ) and other elemental abundance signatures that @xcite also associate with production by faint supernovae . this observation is suggestive of similar carbon - production and enrichment mechanisms in the early universe both locally and in high - redshift systems . it is presumably no coincidence @xcite that five of the six stars known with [ fe / h ] @xmath26 are confirmed cemp - no stars @xcite . there are two observational keys required to advance our understanding of these ancient stars and how they are related to early galactic chemical evolution . one is to obtain the full set of c , n , and o abundances for as many cemp stars as possible , an activity that is being pursued by a number of groups ( see , e.g. , @xcite ; c. hansen et al . , in preparation ; c. kennedy et al . , in preparation ) . the other is to obtain as complete an inventory as possible of the light and heavy neutron - capture elements for representative examples of the known varieties of cemp stars . although we have partial information from previous ground - based high - resolution spectroscopic observations , there remain many key elements , such as the light species ge and zr , and heavier species such as os and pt , that can only be obtained through near - ultraviolet ( nuv ) spectroscopy ( e.g. , * ? ? ? * ; * ? ? ? the element pb is of particular importance , as it may provide a useful discriminant between a number of possible nucleosynthesis pathways @xcite . since it was installed on board the _ hubble space telescope _ ( _ hst _ ) in 1997 , the space telescope imaging spectrograph ( stis ) has been the only instrument available for the high - resolution nuv spectroscopy required to make these measurements . we have recently completed a new _ hst_/stis observing program to collect high - quality nuv spectroscopy for three cemp stars , including one member of each of the cemp - no , cemp-@xmath5 , and cemp-@xmath20 sub - classes . in this paper , we perform an abundance analysis of high - resolution nuv spectroscopy for the star , the brightest known member of the sub - class of cemp - no stars . we fill in the abundance patterns , as best as possible , for elements beyond the iron peak , including eight species not accessible from ground - based observations . section [ obssec ] describes our observations and reductions , and compares abundances derived from nuv lines with those derived previously from optical lines by @xcite . section [ absec ] describes our abundance analysis in detail . we present a brief discussion and our conclusions in section [ diss ] . new stis @xcite observations of were obtained as part of program go-12554 , using the e230 m echelle grating , centered on 2707 , and the nuv multianode microchannel array detector . there were two observational sequences of four individual exposures , taken on february 28 , 2012 . each sequence had an exposure time of 2 hours and 52 minutes , with a total integration time of 5 hours and 44 minutes . the 006@xmath2702 slit yields a @xmath14 2-pixel resolving power ( r @xmath28 ) @xmath14 30,000 . our setup produced a wavelength coverage from 22803070 in a single exposure . the observations were reduced and calibrated using the standard _ calstis _ pipeline . the s / n of the combined spectrum varies from @xmath1450 pix@xmath29 near 2300 , to @xmath1480 pix@xmath29 near 2700 , to @xmath30100 pix@xmath29 near 3070 . figure [ mg_comp ] shows a portion of the nuv spectra of , in the region of the doublet at 2800 . _ hst_/stis spectra of two metal - poor giants with similar atmospheric parameters are shown for comparison , ( @xmath31= 5100 k , [ fe / h]= @xmath322.53 ; * ? ? ? * ) and ( @xmath31= 5170 k , [ fe / h]= @xmath322.46 ; * ? ? ? * ) . the effective temperatures and surface gravities of and are comparable to , but their metallicities are higher by about 1.5 dex . the lower metallicity of is immediately apparent from inspection of this figure . is moderately enhanced in @xmath19-process material , ( [ eu / fe ] @xmath330.5 ; @xcite ) , and is enhanced in @xmath5-process material ( [ ba / fe ] @xmath331.5 ; @xcite ) . does not exhibit neutron - capture element enhancements . to better illustrate the rather striking differences , figure [ comp ] shows portions of the nuv spectra around the lines of several neutron - capture elements for the same three stars shown in figure [ mg_comp ] . note in particular the absence of absorption by , , , and for , which are well - known abundance markers for the operation of the @xmath5-process @xcite . equivalent widths were obtained by fitting gaussian profiles to the observed atomic lines , using the robospect package @xcite . the line lists were based on the compilation of @xcite , as well as on data retrieved from the vald database @xcite and the national institute of standards and technology atomic spectra database ( nist ; * ? ? ? * ) . abundances for individual and lines , derived from equivalent widths as well as from spectral synthesis , are listed in table [ eqw ] . figure [ iron ] shows a sample of the nuv spectra , with a number of the fe lines used for the synthesis . the abundances of all and lines in our stis spectrum were verified by spectral synthesis . we adopt the model - atmosphere parameters derived by @xcite , @xmath31= 5430 k , log@xmath34= 3.4 ( cgs ) , @xmath35 = 1.3 kms@xmath29 , and [ fe / h ] = @xmath36 . the iron abundances we derive from nuv lines [: log(@xmath37 ) = 3.62 @xmath38 0.02 ; : log(@xmath37 ) = 3.63 @xmath38 0.02 ] differ little from the values derived by @xcite from optical lines [ : log(@xmath37 ) = 3.67 @xmath38 0.01 ; : log(@xmath37 ) = 3.68 @xmath38 0.03 ] . previous work has shown that small differences between optical and nuv and lines may exist @xcite . for example , figure 6 of @xcite reveals a small `` dip '' in the abundances of and lines blueward of the balmer series limit in . @xcite , @xcite , and @xcite investigated several causes of this effect for other metal - poor giants , but the differences are not fully understood at present . to minimize this effect , we reference abundance ratios of other elements derived from nuv transitions to the iron abundance also derived from nuv transitions . as a check on our procedures , we also used the equivalent - width values published in @xcite as input to our machinery . figure [ ito_comp ] shows the differences between the abundances derived by equivalent - width analysis between @xcite and this work . apart from si ( which differs by @xmath320.07 dex relative to ito et al . ) , all of the other differences lie within @xmath39 dex , with a mean difference of @xmath320.01 dex . this test demonstrates that our analysis procedures and machinery can reproduce the @xcite values to excellent precision . chemical abundances or upper limits were obtained from the nuv spectrum of for 26 elements , including measurements for c , o , sc , ti , cr , mn , fe , and ni , and upper limits for be , b , ge , zr , nb , mo , cd , te , ce , nd , eu , gd , yb , lu , hf , os , pt , and pb . our abundance analysis utilizes one - dimensional plane - parallel atlas9 model atmospheres with no overshooting @xcite , computed under the assumption of local thermodynamic equilibrium ( lte ) . we use the 2011 version of the moog synthesis code @xcite for this analysis . to treat isotropic , coherent scattering in this version of moog , the solution of the radiative transfer considers both absorption and scattering components , rather than treating such scattering as pure absorption ( see * ? ? ? * for further details ) . our final abundance ratios , [ x / fe ] , are given with respect to the solar abundances of @xcite . upper limits for elements for which no absorption lines were detected provide additional information for the interpretation of the overall abundance pattern of the stars . based on the s / n ratio in the spectral region of the line , and employing the formula given in @xcite , we derive 3-@xmath40 upper limits for 13 elements . abundances and upper limits for individual lines , derived from both equivalent widths and spectral synthesis , are listed in table [ abtable ] . a summary of the chemical abundances and upper limits for is provided in table [ abfinal ] . the @xmath40 refers to the standard error of the mean . we have also addressed the systematic uncertainties that could affect the model - atmosphere parameters . table [ sys ] shows the effect of changes in each atmospheric parameter on the determined abundances , using spectral lines from which abundances were determined by equivalent - width analysis alone . the adopted variations are 150 k for @xmath31 , 0.5 dex for log@xmath34 , and 0.3 kms@xmath29 for @xmath35 . also shown is the total uncertainty , taken as the quadratic sum of the individual errors . we discuss the determinations of beryllium , boron , carbon , the iron - peak elements , and the neutron - capture elements in more detail in the following subsections . the resonance line at 2348 is not detected in our spectrum of . however , we can make use of this line to place a significantly lower ( by 0.5 dex , a factor of three ) upper limit on the be abundance in ( @xmath412.3 ) than obtained by @xcite from the nuv doublet at 3130 ( @xmath411.8 ) . the blue dotted line in the left panel of figure [ beb_ul ] shows our 3-@xmath40 upper limit for the 2348 line . for comparison , the red dashed line shows the @xcite upper limit . we note , following @xcite , that 3d and nlte effects on the beryllium abundance in metal - poor stars are expected to be small @xcite . the resonance doublet at 2497 is also not detected in our spectrum of . we derive an upper limit of @xmath150.5 from these lines , as shown in the right panel of figure [ beb_ul ] . nlte corrections for measured b abudances appear to be important ( e.g. , @xcite ) , and could possibly perturb the results ( as inferred from the nlte calculations of kiselman & carlsson ) by up to @xmath250.5 dex , for stars of metallicity similar to . for the purpose of comparing our present upper limit on b to previous detections , we prefer to use the lte results . the upper panels of figure [ co ] show the spectral synthesis of the atomic features at 2478.56 and 2967.21 . although the 2967.21 line is clean , the 2478 line is blended with several other species , and in more metal - rich stars these blends prohibit its use as an abundance indicator . the most severe of these blending features is an line at essentially the identical wavelength , 2478.57 . the nist asd database reports uncertainties of @xmath42 18% ( @xmath43 0.09 dex ) on the @xmath44 values of both the lines . thus , to the extent that we know the fe abundance , and are modeling the line formation appropriately , we can use the line as a c abundance indicator in . from the spectral synthesis of these lines , we obtain an average of = 5.78 , which yields a carbonicity of [ c / fe]= @xmath251.23 ( using [ fe / h ] = @xmath323.88 ) . these values agree well with the optical determinations from @xcite (= 5.95 and [ c / fe]= @xmath251.35 , using [ fe / h]=@xmath323.83 ) , but are a few tenths of a dex higher than the near - infrared ( nir ) determinations from @xcite (= 5.69 and [ c / fe]= @xmath250.83 , using [ fe / h ] = @xmath323.68 ) . it must be kept in mind that the optical values are determined from the ch @xmath45-band , and the nir from @xmath46 m lines . to our knowledge , this is the first determination of [ c / fe ] based on atomic lines in the nuv region . the relatively close agreement of this determination with the [ c / fe ] inferred from the molecular ch @xmath45-band is encouraging . previous modeling has suggested that 3d effects on the ch and c@xmath47 features for giants and subgiants at [ fe / h ] @xmath48 can lead to an over - estimate of [ c / fe ] of @xmath49 to @xmath50 dex @xcite , although nlte effects are not expected to be large . another possibility is that uncertainties in the uv opacity determination have the same magnitude as 3d effects . @xcite found that the [ c / fe ] ratio derived from an lte analysis of the [ c i ] forbidden line at 8727 for the cemp star he 1005 - 1429 ( [ fe / h ] @xmath51 ) was on the order of 0.3 - 0.4 dex lower than the value reported by @xcite , based on the molecular c@xmath47 feature at 5170 . future observations of the nuv features for additional ( necessarily bright , and ideally extremely metal - poor ) cemp stars may thus prove illuminating . the lower panel of figure [ co ] shows the spectral region 2965 2972 , where several oh features are available . we were able to obtain adequate fits for 11 lines with the same input abundance , using the line list from @xcite . the value of the o abundance we obtain , [ o / fe ] = @xmath251.54 , is in good agreement with that derived by @xcite ( also using an oh feature as an indicator ) , [ o / fe ] = @xmath251.64 . abundances for sc , ti , cr , and ni were determined with an equivalent - width analysis only . lines are broadened by hyperfine splitting of the @xmath52mn isotope , so we derived those abundances from spectral synthesis . figure [ mnii ] shows the synthesis of four lines in the nuv spectrum of . to make a fair comparison with the @xcite abundances , we recomputed their optical abundances of ti ii , mn ii , and ni i on the same @xmath44 scale we used for the nuv lines . from the @xcite study , we find that the ito et al . ti ii abundance would have decreased by only 0.01 dex . from lines in common with the @xcite study , we find that the ito et al.mn ii abundance would have increased by 0.08 dex . from lines in common with the @xcite study , we find that the ito et al.ni i abundance would have had no change . there would also be no change for the sc ii abundance , since ito et al . used the @xmath44 values reported by @xcite . as a result of this exercise , we find good agreement between the abundance ratios determined by this work and those from @xcite these are [ /fe ] = @xmath250.29 , identical to that obtained from the optical work ( @xmath250.29 ) ; [ /fe ] = @xmath250.36 ( @xmath250.36 in the optical ) ; [ /fe ] = @xmath320.09 ( @xmath320.22 in the optical ) ; [ /fe ] = @xmath320.58 ( @xmath320.79 in the optical ) ; and [ /fe ] = @xmath320.02 ( @xmath250.08 in the optical ) . only upper limits were determined in for neutron - capture elements in the nuv region ; figure [ upperlim ] shows examples of these limits for , , and . figure [ pb_ul ] shows the comparison between the observed and the synthetic spectra around the 2833.05 feature . we determine a 3-@xmath40 ( 2-@xmath40 ) upper limit of @xmath180.23 ( @xmath150.42 ) from this line , assuming the poorly fit absorption features at 2832.9 and 2833.2 are noise , and not absorption lines . our value confirms and strengthens the @xcite upper limit of @xmath180.10 , estimated from the weak optical line at 4057.80 . syntheses of both the 3-@xmath40 and 2-@xmath40 upper limits are shown in figure [ pb_ul ] , as well as for = 0.00 and = @xmath322.00 , for reference . figure [ pb_ul ] reveals that our upper limit on the abundance of pb might indeed be too conservative , depending on the nature of the nearby noise features . the wavelength of the pb i line is known to better than 1 m @xcite , so these supposed noise features are not due to pb i absorption . the feature at 2832.9 , however , is also observed in our spectra of and ( see figure [ comp ] ) , but the nist database does not include any probable lines at this wavelength . the known fe ii line at 2833.09 @xcite does not appear in our spectrum of . the feature at 2833.2 may also appear in and , but there it is weak . from examination of the summed spectra from the first four observations and the second four observations of independently , these unidentified features appear in both . this indicates that they are not random noise spikes . if we do not treat these unfit features as noise , the upper limit on pb becomes tighter by at least several tenths of a dex but it is still limited by our ability to correctly identify the continuum or other contaminants . @xcite have discussed in detail the importance of the low upper limit they derived for be in the optical spectrum of , which we have now lowered by about a factor of three , to @xmath412.3 , based on our nuv _ hst_/stis measurements . our estimate of the upper limit on b ( @xmath53 ) , is also of significance , since these limits are at the lowest level yet determined for very and extremely metal - poor stars . figure [ bbefeh ] shows a comparison between be and b abundances , as a function of [ fe / h ] , for the upper limits determined in this work and data from @xcite , @xcite , and the saga database @xcite . individual references are listed in the caption of figure [ bbefeh ] . how can we account for the low upper limits for be and b , taken together with the fact that the li abundance for ( = 1.0 , reported by * ? ? ? * ) is significantly below the level of the spite plateau ? the low abundance of li in this star , compared to the spite plateau value , could be a result of convective mixing with internal layers in which li is fully depleted . be might also be affected by mixing with material from layers that reach its burning temperature ( @xmath54 k ) , which is slightly higher than that of li ( @xmath55 k ) . compared to these two elements , b has an even higher burning temperature ( @xmath56 ) , and would be expected to be less affected by mixing , although such an interpretation depends on detailed modeling of stellar evolution during the subgiant phase . alternatively , the very low upper limits of be and b are consistent with the view that cemp - no stars such as may have formed in the very early universe , _ prior _ to the establishment of the level of cosmic - ray flux necessary to produce be and b by spallation ( for a more detailed discussion , see the review by @xcite ) . the low abundance of li could then be accounted for by mixing of li - astrated material ( due to burning by first - generation stars ) with primordial li created by big bang nucleosynthesis @xcite . this alternative receives some support from the recent observations of li abundances below the spite plateau for cemp - no stars by @xcite , including two stars with @xmath31 = 6100 k , presumably too warm for conventional li - depletion from convective mixing to have taken place . it is worth recalling that @xcite reports that the cemp - no class _ only _ contains li - depleted stars . unfortunately , be and b abundance estimates are not yet available for the cemp - no stars of @xcite . improved models and , in particular , additional observations of be and b for cemp - no stars , are necessary in order to constrain these ideas further . it should be noted that , as already discussed by @xcite , the progenitor of is unlikely to be a significant source of high - energy cno nuclei that could yield lighter elements by spallation processes . this fact might be a useful constraint on the nature of the progenitor , most likely a faint ( mixing and fallback ) supernova , that produces the high c and o abundances found in . we have discussed above that our measurement of [ c / fe ] , based on the nuv atomic lines at 2478.56 and 2967.21 , provides an important validation of [ c / fe ] estimates for cemp stars based on the ch @xmath45-band in the optical , as well as from lines in the nir . since this is the first determination of [ c / fe ] from nuv spectroscopy , it would clearly be important to carry out similar observations of additonal bright cemp stars , in order to test if this level of agreement holds for stars that are cooler , or more metal - rich , than . the [ ni / fe ] abundance ratio for can also place further constraints on galactic chemical - evolution models . by comparing this quantity with the value measured for the main - sequence turnoff star hd 84937 ( @xmath31= 6300 k , log@xmath34= 4.0 , @xmath35 = 1.5 kms@xmath29 , and [ fe / h ] = @xmath57 ; * ? ? ? * ) , one can see that , even though their metallicities are more than 1.5 dex apart , their [ ni / fe ] ratios exhibit only a 0.05 dex difference . this corroborates the predicted plateau for ni abundances as a function of the metallicity from the theoretical models of @xcite , and can set observational limits on the galactic initial mass function and yields from both core - collapse supernovae and hypernovae . figure [ pattern ] shows a comparison between a set of @xmath5- and @xmath19-process templates and the upper limits on neutron - capture elements determined in this work . also shown are the optical abundances from @xcite . these are used to normalize the models ( to sr in the upper panel and to ba in the lower panel ) . the template for the weak component of the @xmath19-process is the metal - poor giant @xcite . the template for the main component of the @xmath19-process is the metal - poor giant @xcite . the template for the @xmath5-process shows the agb yields from the model presented in @xcite , with m = 1.3 m@xmath58 and [ fe / h ] = @xmath322.5 . this is necessary due to the fact that , in contrast to the `` universal '' @xmath19-process , the @xmath5-process abundances are highly dependent on metallicity . these templates are not intended as firm representations of the expected nucleosynthesis outcomes , since , for example , the yields depend on the physical conditions at the time the nucleosynthesis events took place . stars with metallicities as low as do not exhibit enhancements of @xmath5-process material ( e.g. , @xcite ) , at least none have been identified to date . at somewhat higher metallicities , the @xmath5-process elements in cemp-@xmath5 stars typically show substantial enhancements relative to the solar abundance ratios ( e.g. , @xcite ) , which are attributed to a mass - transfer event from a companion star that passed through the agb phase of evolution . at lower metallicity , the high neutron - to - seed ratios are expected to drive the flow to the most massive stable elements that can be produced by the @xmath5-process , pb and bi ( e.g. , @xcite ) . enhanced pb abundances are considered unmistakable signatures of the operation of @xmath5-process nucleosynthesis in a low - metallicity environment . to the best of our knowledge , pb has not been detected for any stars with [ fe / h ] @xmath153.2 ( see , e.g. , figure 3 of @xcite ) . our results do not change this finding . figure [ bisterzo ] shows the [ pb / fe ] , [ ba / fe ] , and [ pb / ba ] values for , compared with four different model prescriptions from @xcite . even though exhibits a clear lack of pb and ba when compared to the models , the upper limit we place on pb , relative to barium , [ pb / ba ] @xmath592.5 , is on the cusp of excluding @xmath5-process nucleosynthesis in low - mass low - metallicity agb stars with the highest neutron exposures possible . this is not yet definitive evidence against an @xmath5-process origin of the neutron - capture elements in ; however , the lack of significant radial - velocity variations ( * ? ? ? * ; * ? ? ? * spanning over 25 years ) and the sub - solar [ sr / fe ] and [ ba / fe ] ratios are suggestive that enrichment from an unseen companion that passed through the agb phase of evolution appears extremely unlikely , calling for non - agb sources , such as spinstars or faint supernovae explosions that undergo mixing and fallback . some form of @xmath19-process nucleosynthesis may have been the dominant production mechanism for the neutron - capture elements observed in , and other stars at extremely low metallicity ( e.g. , @xcite ) . barium is the heaviest element detected in . the abundance pattern presented in the bottom panel of figure [ pattern ] can not exclude that the neutron - capture abundance pattern in resembles that found in the main or weak components of the @xmath19-process . @xcite also reached this conclusion , based on their analysis of optical spectra of and other members of the cemp - no sub - class of stars . is also the most metal - poor star where the ge abundance can be reasonably evaluated . the upper limit on the ge abundance in we obtain is only about a factor of three higher than the mean [ ge / fe ] ratio found for stars with @xmath323.0 @xmath60 [ fe / h ] @xmath151.6 ( see figure 3 of @xcite ) . this indicates that ge was not manufactured in large quantities by the stars that produced the metals in . the primary nucleosynthesis mechanism of ge assumed for the stars with @xmath323.0 @xmath60 [ fe / h ] @xmath151.6 , some form of explosive or charged - particle nucleosynthesis , is not excluded as a possible source for whatever ge may be present in . the te upper limit we obtain is still about 2 dex higher than what might be required to secure a te detection in , based on the ba abundance in and the @xmath61 ( te / ba ) ratios reported by @xcite . similarly , the cd upper limit is about 1 dex higher than what might be required to secure a cd detection in . if the 3@xmath62 @xmath19-process peak elements in are assumed to follow a scaled - solar @xmath19-process abundance pattern , the pt upper limit is about 0.5 dex higher than what might be required to secure a pt detection in . if the 3@xmath62 @xmath19-process peak elements in are deficient relative to ba and the rare - earth elements , as found in the metal - poor giants and by @xcite , it would be even more challenging to secure a pt detection in . te and pt are expected to be two of the most abundant elements heavier than zr @xcite , yet it is unlikely that either of these elements can be detected in . this highlights the challenge of firmly establishing the nucleosynthetic origins of the elements heavier than the iron group for extremely metal - poor cemp - no stars . v.m.p . acknowledges support from the gemini observatory . t.c.b . and h.s . acknowledge partial support for this work from phy 08 - 22648 ; physics frontier center / joint institute or nuclear astrophysics ( jina ) , awarded by the us national science foundation . is supported by nsf career grant ast-1255160 . was supported by the jsps grants - in - aid for scientific research ( 23224004 ) . generous support for program go-12554 has been provided by nasa through a grant from the space telescope science institute , which is operated by the association of universities for research in astronomy , inc . , under nasa contract nas 5 - 26555 . cccccc + 2283.30 & 0.12 & @xmath322.22 & 21.10 & 3.62 & 3.70 + 2283.66 & 0.11 & @xmath322.22 & & & 3.70 + 2293.85 & 0.09 & @xmath322.37 & 13.10 & 3.43 & 3.60 + 2294.41 & 0.11 & @xmath321.54 & 43.40 & 3.71 & 3.65 + 2296.93 & 0.11 & @xmath322.02 & 29.60 & 3.68 & 3.70 + 2297.79 & 0.05 & @xmath321.10 & 52.80 & 3.59 & 3.60 + 2298.66 & 0.11 & @xmath322.42 & 13.90 & 3.54 & 3.60 + 2299.22 & 0.09 & @xmath321.55 & 38.00 & 3.48 & 3.65 + 2320.36 & 0.05 & @xmath320.99 & 57.80 & 3.66 & 3.65 + 2350.41 & 0.00 & @xmath323.03 & 7.30 & 3.67 & + 2369.46 & 0.11 & @xmath322.19 & 19.10 & 3.49 & 3.50 + 2371.43 & 0.09 & @xmath321.95 & 33.30 & 3.67 & 3.70 + 2374.52 & 0.12 & @xmath322.10 & 22.90 & 3.53 & 3.60 + 2389.97 & 0.09 & @xmath321.57 & 39.70 & 3.50 & 3.60 + 2443.87 & 0.86 & @xmath321.24 & 28.00 & 3.58 & 3.65 + 2445.21 & 0.86 & @xmath322.02 & 5.40 & 3.40 & + 2453.48 & 0.92 & @xmath320.92 & 27.90 & 3.31 & + 2457.60 & 0.86 & @xmath320.32 & 48.00 & 3.32 & + 2462.18 & 0.05 & @xmath321.30 & 49.80 & 3.54 & 3.60 + 2462.65 & 0.00 & @xmath320.32 & 78.40 & 3.46 & 3.60 + 2463.73 & 0.96 & @xmath321.13 & 20.60 & 3.33 & + 2468.88 & 0.86 & @xmath320.62 & 42.50 & 3.41 & 3.55 + 2470.97 & 0.92 & @xmath321.62 & 8.40 & 3.27 & + 2472.89 & 0.05 & @xmath320.08 & 99.50 & 3.65 & 3.60 + 2485.99 & 0.92 & @xmath321.61 & 10.30 & 3.36 & 3.60 + 2486.69 & 0.96 & @xmath320.91 & 28.50 & 3.35 & 3.50 + 2487.07 & 1.01 & @xmath320.75 & 32.90 & 3.38 & 3.55 + 2487.37 & 0.09 & @xmath321.90 & 31.60 & 3.51 & 3.55 + 2491.16 & 0.11 & @xmath250.13 & 104.30 & 3.55 & 3.60 + 2495.87 & 0.86 & @xmath321.76 & 12.70 & 3.56 & 3.65 + 2496.53 & 0.92 & @xmath320.66 & 39.10 & 3.37 & 3.50 + 2501.13 & 0.00 & @xmath320.35 & 81.70 & 3.55 & 3.60 + 2501.69 & 0.86 & @xmath321.51 & 23.50 & 3.69 & 3.70 + 2507.90 & 0.96 & @xmath320.79 & 34.50 & 3.40 & 3.50 + 2508.75 & 0.99 & @xmath321.95 & & & 3.65 + 2517.66 & 0.99 & @xmath320.98 & 32.80 & 3.57 & 3.65 + 2518.10 & 0.09 & @xmath320.26 & 75.10 & 3.37 & 3.60 + 2519.63 & 1.01 & @xmath321.20 & 19.70 & 3.41 & 3.65 + 2522.48 & 0.92 & @xmath321.92 & 12.20 & 3.76 & 3.70 + 2522.85 & 0.00 & @xmath250.26 & & & 3.70 + 2530.69 & 0.09 & @xmath322.37 & & & 3.70 + 2543.92 & 2.45 & @xmath250.70 & 31.40 & 3.32 & 3.55 + 2552.61 & 0.11 & @xmath322.52 & 15.30 & 3.62 & 3.65 + 2556.86 & 0.86 & @xmath322.02 & & & 3.55 + 2560.56 & 1.01 & @xmath322.11 & & & 3.65 + 2569.74 & 0.99 & @xmath322.24 & 3.10 & 3.47 & + 2576.69 & 0.86 & @xmath320.91 & 35.90 & 3.43 & 3.55 + 2584.54 & 0.86 & @xmath320.39 & 51.00 & 3.42 & 3.60 + 2610.75 & 0.09 & @xmath322.96 & 5.10 & 3.45 & 3.70 + 2612.77 & 0.05 & @xmath322.59 & 17.30 & 3.68 & 3.70 + 2618.02 & 0.96 & @xmath320.97 & 38.60 & 3.66 & 3.65 + 2618.71 & 0.00 & @xmath322.43 & 14.20 & 3.35 & 3.50 + 2623.37 & 0.11 & @xmath322.57 & & & 3.65 + 2623.53 & 0.96 & @xmath320.70 & 48.40 & 3.72 & 3.60 + 2632.24 & 0.99 & @xmath321.20 & 25.60 & 3.54 & 3.60 + 2632.59 & 0.09 & @xmath322.33 & 23.80 & 3.67 & 3.65 + 2635.81 & 0.99 & @xmath320.81 & 43.60 & 3.69 & 3.60 + 2636.48 & 0.92 & @xmath322.04 & 6.00 & 3.49 & 3.55 + 2641.03 & 2.45 & @xmath321.25 & 1.70 & 3.72 & + 2644.00 & 1.01 & @xmath320.91 & 29.50 & 3.38 & 3.50 + 2647.56 & 0.05 & @xmath322.42 & 18.50 & 3.55 & 3.60 + 2651.71 & 0.96 & @xmath322.04 & 5.80 & 3.51 & 3.70 + 2656.14 & 2.40 & @xmath320.59 & 3.70 & 3.36 & 3.60 + 2656.79 & 1.49 & @xmath321.77 & 3.90 & 3.61 & 3.65 + 2660.40 & 0.99 & @xmath322.33 & 5.40 & 3.80 & 3.70 + 2662.06 & 0.96 & @xmath321.61 & 9.00 & 3.30 & + 2679.06 & 0.86 & @xmath320.75 & 50.90 & 3.72 & 3.60 + 2680.45 & 0.99 & @xmath321.74 & 7.40 & 3.35 & 3.50 + 2689.21 & 0.92 & @xmath320.89 & 41.70 & 3.61 & 3.60 + 2690.07 & 0.00 & @xmath322.72 & 11.00 & 3.48 & 3.55 + 2699.11 & 0.92 & @xmath321.26 & 24.50 & 3.47 & 3.60 + 2710.54 & 1.61 & @xmath321.33 & 4.30 & 3.34 & 3.60 + 2714.87 & 0.96 & @xmath322.19 & 3.50 & 3.41 & 3.60 + 2723.58 & 0.09 & @xmath320.72 & 73.50 & 3.66 & 3.70 + 2726.05 & 1.01 & @xmath321.21 & 19.60 & 3.37 & 3.55 + 2728.02 & 0.92 & @xmath321.46 & 19.10 & 3.50 & 3.55 + 2735.47 & 0.92 & @xmath320.40 & 54.50 & 3.53 & + 2737.31 & 0.11 & @xmath320.61 & & & 3.70 + 2744.07 & 0.12 & @xmath320.98 & & & 3.65 + 2754.03 & 0.99 & @xmath321.38 & & & 3.60 + 2755.18 & 2.43 & @xmath321.28 & & & 3.65 + 2756.27 & 0.05 & @xmath322.17 & & & 3.65 + 2756.33 & 0.11 & @xmath321.09 & & & 3.65 + 2759.81 & 1.01 & @xmath321.58 & 12.00 & 3.45 & 3.55 + 2772.07 & 0.86 & @xmath321.53 & & & 3.65 + 2772.11 & 0.09 & @xmath321.48 & & & 3.65 + 2813.29 & 0.92 & @xmath320.35 & & & 3.60 + 2823.28 & 0.96 & @xmath320.90 & & & 3.60 + 2827.89 & 0.05 & @xmath322.80 & 8.00 & 3.43 & 3.60 + 2838.12 & 0.99 & @xmath321.11 & & & 3.55 + 2936.90 & 0.00 & @xmath320.79 & & & 3.50 + 2959.99 & 2.69 & @xmath320.07 & 9.90 & 3.55 & 3.55 + 2965.25 & 0.12 & @xmath321.34 & & & 3.70 + 2970.10 & 0.11 & @xmath321.15 & & & 3.65 + 2970.12 & 0.09 & @xmath321.87 & & & 3.65 + 2983.57 & 0.00 & @xmath320.58 & & & 3.65 + 2994.43 & 0.05 & @xmath320.53 & & & 3.70 + 2994.50 & 0.12 & @xmath322.22 & & & 3.70 + 2999.51 & 0.86 & @xmath320.60 & & & 3.70 + 3000.45 & 1.49 & @xmath321.09 & & & 3.50 + 3008.14 & 0.11 & @xmath320.84 & & & 3.60 + 3016.18 & 0.99 & @xmath321.44 & & & 3.60 + 3021.07 & 0.05 & @xmath320.36 & & & 3.60 + 3026.46 & 0.99 & @xmath321.12 & & & 3.60 + 3037.39 & 0.11 & @xmath320.70 & & & 3.60 + 3042.66 & 0.99 & @xmath321.30 & & & 3.60 + 3047.61 & 0.09 & @xmath320.56 & 79.70 & 3.45 & 3.60 + 3059.09 & 0.05 & @xmath320.69 & & & 3.70 + + 2331.31 & 0.23 & @xmath320.68 & 110.40 & 3.58 & 3.70 + 2354.89 & 0.35 & @xmath321.05 & 84.70 & 3.49 & 3.65 + 2359.11 & 0.11 & @xmath320.60 & & & 3.70 + 2359.60 & 2.68 & @xmath320.73 & 21.40 & 3.57 & 3.55 + 2360.00 & 0.23 & @xmath320.52 & & & 3.70 + 2360.29 & 0.30 & @xmath320.51 & & & 3.70 + 2361.73 & 2.69 & @xmath320.79 & 12.30 & 3.31 & 3.50 + 2368.60 & 0.35 & @xmath320.69 & 102.70 & 3.58 & 3.55 + 2370.50 & 0.39 & @xmath321.23 & 76.10 & 3.68 & 3.70 + 2384.39 & 0.39 & @xmath320.96 & 85.30 & 3.60 & 3.70 + 2399.24 & 0.08 & @xmath320.14 & & & 3.65 + 2422.69 & 3.89 & @xmath250.01 & 7.80 & 3.45 & 3.60 + 2429.39 & 2.70 & @xmath320.61 & 26.70 & 3.63 & 3.60 + 2430.08 & 2.83 & @xmath250.23 & 40.50 & 3.37 & + 2432.87 & 4.08 & @xmath250.55 & 12.90 & 3.36 & 3.55 + 2433.50 & 2.68 & @xmath320.86 & 23.80 & 3.76 & 3.70 + 2446.47 & 2.66 & @xmath320.42 & 31.30 & 3.54 & 3.60 + 2447.20 & 3.89 & @xmath320.21 & 8.50 & 3.71 & 3.65 + 2463.28 & 3.15 & @xmath320.19 & 23.20 & 3.54 & 3.60 + 2464.91 & 3.23 & @xmath320.09 & 22.00 & 3.48 & 3.60 + 2468.30 & 2.68 & @xmath321.05 & 15.90 & 3.67 & 3.70 + 2470.41 & 3.23 & @xmath320.48 & 11.20 & 3.47 & 3.50 + 2470.67 & 2.83 & @xmath320.07 & 40.50 & 3.65 & 3.60 + 2476.27 & 2.69 & @xmath321.05 & 12.70 & 3.56 & 3.60 + 2497.82 & 3.23 & @xmath320.03 & 23.10 & 3.44 & 3.60 + 2502.39 & 3.22 & @xmath250.03 & 23.80 & 3.40 & 3.55 + 2506.09 & 3.20 & @xmath320.03 & 20.10 & 3.32 & 3.55 + 2527.10 & 2.66 & @xmath320.45 & 32.10 & 3.56 & 3.70 + 2529.55 & 2.81 & @xmath250.33 & 51.90 & 3.64 & 3.70 + 2533.63 & 2.66 & @xmath250.35 & 48.60 & 3.33 & + 2555.07 & 2.84 & @xmath320.81 & 12.30 & 3.44 & 3.60 + 2555.45 & 2.86 & @xmath320.83 & 11.30 & 3.43 & 3.60 + 2559.77 & 3.23 & @xmath320.72 & 9.30 & 3.59 & 3.60 + 2560.28 & 3.20 & @xmath320.16 & & & 3.60 + 2562.54 & 0.99 & @xmath250.02 & 105.30 & 3.50 & 3.65 + 2563.48 & 1.04 & @xmath320.23 & 94.60 & 3.64 & 3.70 + 2566.62 & 2.81 & @xmath321.07 & 6.60 & 3.34 & 3.55 + 2566.91 & 1.08 & @xmath320.64 & & & 3.70 + 2582.58 & 1.08 & @xmath320.45 & 77.50 & 3.52 & 3.65 + 2590.55 & 2.70 & @xmath321.32 & 7.20 & 3.52 & 3.60 + 2591.54 & 1.04 & @xmath320.46 & 80.40 & 3.56 & 3.65 + 2592.78 & 4.08 & @xmath250.65 & 18.20 & 3.44 & 3.50 + 2608.85 & 2.81 & @xmath321.39 & 7.80 & 3.73 & 3.70 + 2610.63 & 0.05 & @xmath250.92 & & & 3.70 + 2620.17 & 2.84 & @xmath321.15 & & & 3.65 + 2620.70 & 2.83 & @xmath320.55 & & & 3.60 + 2621.67 & 0.12 & @xmath320.94 & 110.30 & 3.63 & 3.70 + 2626.50 & 2.86 & @xmath320.67 & 12.60 & 3.30 & 3.50 + 2637.64 & 3.34 & @xmath320.56 & 6.80 & 3.37 & + 2652.57 & 3.27 & @xmath321.43 & 2.50 & 3.69 & + 2664.66 & 3.39 & @xmath250.31 & 29.90 & 3.43 & 3.55 + 2684.75 & 3.81 & @xmath250.23 & 13.00 & 3.38 & 3.60 + 2721.81 & 3.15 & @xmath321.25 & 4.20 & 3.63 & 3.70 + 2732.45 & 0.23 & @xmath322.96 & 39.70 & 3.68 & 3.70 + 2769.36 & 3.15 & @xmath320.48 & 18.20 & 3.60 & 3.60 + 2917.47 & 1.04 & @xmath322.85 & & & 3.65 + 2965.41 & 3.42 & @xmath322.24 & & & 3.65 + lcccrr & 2348.61 & 0.000 & @xmath250.140 & @xmath182.30 & 1 + & 2496.77 & 0.000 & @xmath320.800 & @xmath180.70 & 1 + & 2497.72 & 0.002 & @xmath320.500 & @xmath180.70 & 1 + & 2478.56 & 2.682 & @xmath321.110 & @xmath255.80 & 1 + & 2967.21 & 0.005 & @xmath326.800 & @xmath255.75 & 1 + & 2552.35 & 0.022 & @xmath250.030 & @xmath320.44 & 2 + & 2524.64 & 0.122 & @xmath321.320 & @xmath251.42 & 3 + & 2525.60 & 0.151 & @xmath320.570 & @xmath251.44 & 3 + & 2531.25 & 0.135 & @xmath320.670 & @xmath251.48 & 3 + & 2534.62 & 0.122 & @xmath320.930 & @xmath251.44 & 3 + & 2841.93 & 0.607 & @xmath320.590 & @xmath251.35 & 3 + & 2888.93 & 0.574 & @xmath321.360 & @xmath251.49 & 3 + & 2891.06 & 0.607 & @xmath321.140 & @xmath251.44 & 3 + & 3058.09 & 1.180 & @xmath320.420 & @xmath251.42 & 3 + & 2740.10 & 1.506 & @xmath321.090 & @xmath251.62 & 4 + & 2751.87 & 1.525 & @xmath320.290 & @xmath251.72 & 4 + & 2757.72 & 1.506 & @xmath320.360 & @xmath251.66 & 4 + & 2762.59 & 1.525 & @xmath250.050 & @xmath251.68 & 4 + & 2766.54 & 1.549 & @xmath250.320 & @xmath251.66 & 4 + & 2576.11 & 0.000 & @xmath250.400 & @xmath251.00 & 5 + & 2605.68 & 0.000 & @xmath250.136 & @xmath251.00 & 5 + & 2933.05 & 1.175 & @xmath320.102 & @xmath251.00 & 5 + & 2949.20 & 1.175 & @xmath250.253 & @xmath250.90 & 5 + & 2289.99 & 0.000 & @xmath250.060 & @xmath252.24 & 6 + & 2293.12 & 0.109 & @xmath320.970 & @xmath252.25 & 6 + & 2312.34 & 0.165 & @xmath250.410 & @xmath252.48 & 6 + & 2325.80 & 0.165 & @xmath250.400 & @xmath252.23 & 6 + & 2346.63 & 0.165 & @xmath320.840 & @xmath252.33 & 6 + & 2356.87 & 0.025 & @xmath321.510 & @xmath252.23 & 6 + & 2360.64 & 0.275 & @xmath321.080 & @xmath252.24 & 6 + & 2376.02 & 0.109 & @xmath321.700 & @xmath252.36 & 6 + & 2386.59 & 0.109 & @xmath321.180 & @xmath252.28 & 6 + & 2419.31 & 0.165 & @xmath321.050 & @xmath252.27 & 6 + & 2821.29 & 0.025 & @xmath321.410 & @xmath252.37 & 6 + & 2943.91 & 0.025 & @xmath321.170 & @xmath252.34 & 6 + & 2992.59 & 0.025 & @xmath321.220 & @xmath252.48 & 6 + & 3003.62 & 0.109 & @xmath320.320 & @xmath252.17 & 6 + & 3012.00 & 0.423 & @xmath250.000 & @xmath252.23 & 6 + & 3031.87 & 0.000 & @xmath321.810 & @xmath252.46 & 6 + & 3037.93 & 0.025 & @xmath320.520 & @xmath252.40 & 6 + & 3050.82 & 0.025 & @xmath320.100 & @xmath252.33 & 6 + & 3054.31 & 0.109 & @xmath320.600 & @xmath252.39 & 6 + & 2278.77 & 1.680 & @xmath250.190 & @xmath252.21 & 7 + & 2297.14 & 1.254 & @xmath320.070 & @xmath252.27 & 7 + & 2297.49 & 1.322 & @xmath320.330 & @xmath252.19 & 7 + & 2356.40 & 1.859 & @xmath320.830 & @xmath252.29 & 7 + & 2387.76 & 1.680 & @xmath321.070 & @xmath252.32 & 7 + & 2394.52 & 1.680 & @xmath250.170 & @xmath252.09 & 7 + & 2416.14 & 1.859 & @xmath250.130 & @xmath252.20 & 7 + & 2437.89 & 1.680 & @xmath320.330 & @xmath252.13 & 7 + & 2651.17 & 0.175 & @xmath250.020 & @xmath180.80 & 8 + & 2691.34 & 0.069 & @xmath320.700 & @xmath180.00 & 8 + & 3039.07 & 0.883 & @xmath250.070 & @xmath180.05 & 8 + & 2567.64 & 0.000 & @xmath320.170 & @xmath181.11 & 9 + & 2699.60 & 0.039 & @xmath320.660 & @xmath180.40 & 10 + & 2700.14 & 0.095 & @xmath320.080 & @xmath181.03 & 9 + & 2732.72 & 0.095 & @xmath320.490 & @xmath180.45 & 9 + & 2758.81 & 0.000 & @xmath320.560 & @xmath180.46 & 9 + & 2915.99 & 0.466 & @xmath320.500 & @xmath180.38 & 9 + & 3054.84 & 1.010 & @xmath250.080 & @xmath180.16 & 10 + & 2950.88 & 0.510 & @xmath250.240 & @xmath180.77 & 11 + & 3028.44 & 0.440 & @xmath320.200 & @xmath180.51 & 11 + & 2871.51 & 1.538 & @xmath250.060 & @xmath180.29 & 12 + & 2288.02 & 0.000 & @xmath250.150 & @xmath181.79 & 13 + & 2385.79 & 0.589 & @xmath320.810 & @xmath180.05 & 14 + & 3063.00 & 0.900 & @xmath250.400 & @xmath180.51 & 15 + & 3014.17 & 0.200 & @xmath320.660 & @xmath180.01 & 16 + & 2906.67 & 0.000 & @xmath320.350 & @xmath181.46 & 16 + & 3010.13 & 0.000 & @xmath250.190 & @xmath181.21 & 17 + & 2891.39 & 0.000 & @xmath321.169 & @xmath181.75 & 16 + & 2615.43 & 0.000 & @xmath320.270 & @xmath182.16 & 18 + & 2641.41 & 1.036 & @xmath250.570 & @xmath181.11 & 19 + & 2820.23 & 0.380 & @xmath320.140 & @xmath181.31 & 19 + & 2929.64 & 0.000 & @xmath320.940 & @xmath181.02 & 19 + & 3058.66 & 0.000 & @xmath320.410 & @xmath180.30 & 20 + & 2282.28 & 0.000 & @xmath320.050 & @xmath180.90 & 20 + & 2659.45 & 0.000 & @xmath320.030 & @xmath181.40 & 21 + & 2929.79 & 0.000 & @xmath320.700 & @xmath180.66 & 21 + & 2833.05 & 0.000 & @xmath320.500 & @xmath180.23 & 22 lcrrrr & 1.38 & @xmath182.30 & @xmath630.20 & & 1 + & 2.70 & @xmath180.70 & @xmath630.48 & & 2 + & 8.43 & 5.78 & @xmath251.23 & 0.20 & 2 + oh & 8.69 & 6.35 & @xmath251.54 & 0.20 & 11 + & 3.15 & @xmath320.44 & @xmath250.29 & 0.10 & 1 + & 4.95 & 1.43 & @xmath250.36 & 0.02 & 8 + & 5.64 & 1.67 & @xmath320.09 & 0.02 & 5 + & 5.43 & 0.97 & @xmath320.58 & 0.03 & 4 + & 7.50 & 3.62 & @xmath323.88 & 0.01 & 98 + & 7.50 & 3.63 & @xmath323.87 & 0.01 & 53 + & 6.22 & 2.32 & @xmath320.02 & 0.02 & 19 + & 6.22 & 2.21 & @xmath320.13 & 0.03 & 8 + & 3.65 & @xmath180.80 & @xmath180.57 & & 3 + & 2.58 & @xmath181.11 & @xmath630.18 & & 7 + & 1.46 & @xmath180.77 & @xmath631.65 & & 2 + & 1.88 & @xmath180.29 & @xmath631.71 & & 1 + & 1.71 & @xmath181.79 & @xmath630.38 & & 1 + & 2.18 & @xmath180.05 & @xmath631.65 & & 1 + & 1.58 & @xmath180.51 & @xmath631.79 & & 1 + & 1.42 & @xmath180.01 & @xmath632.45 & & 1 + & 0.52 & @xmath181.46 & @xmath631.90 & & 1 + & 1.07 & @xmath181.21 & @xmath631.60 & & 1 + & 0.84 & @xmath181.75 & @xmath631.29 & & 1 + & 0.85 & @xmath181.31 & @xmath630.00 & & 1 + & 0.10 & @xmath182.16 & @xmath631.62 & & 1 + & 0.85 & @xmath181.31 & @xmath631.72 & & 3 + & 1.40 & @xmath180.30 & @xmath632.18 & & 1 + & 1.40 & @xmath180.90 & @xmath631.58 & & 1 + & 1.62 & @xmath181.40 & @xmath630.86 & & 2 + & 1.75 & @xmath180.23 & @xmath631.90 & & 1 lcccc & @xmath320.10 & @xmath320.17 & @xmath250.02 & 0.20 + & @xmath320.09 & @xmath320.17 & @xmath250.04 & 0.20 + & @xmath320.07 & @xmath320.16 & @xmath250.11 & 0.21 + & @xmath320.12 & @xmath320.06 & @xmath250.13 & 0.19 + & @xmath320.18 & @xmath250.02 & @xmath250.06 & 0.19 + & @xmath320.05 & @xmath320.15 & @xmath250.05 & 0.17 + & @xmath320.20 & @xmath250.02 & @xmath250.09 & 0.22 + & @xmath320.06 & @xmath320.17 & @xmath250.14 & 0.23

we report on numerically exact determinantal quantum monte carlo simulations of the onset of spin - density wave ( sdw ) order in itinerant electron systems captured by a sign - problem - free two - dimensional lattice model . extensive measurements of the sdw correlations in the vicinity of the phase transition reveal that the critical dynamics of the bosonic order parameter are well described by a dynamical critical exponent , consistent with hertz - millis theory , but are found to follow a finite - temperature dependence that does not fit the predicted behavior of the same theory . the presence of critical sdw fluctuations is found to have a strong impact on the fermionic quasiparticles , giving rise to a dome - shaped superconducting phase near the quantum critical point . in the superconducting state we find a gap function that has an opposite sign between the two bands of the model and is nearly constant along the fermi surface of each band . above the superconducting our numerical simulations reveal a nearly temperature and frequency independent self energy causing a strong suppression of the low - energy quasiparticle spectral weight in the vicinity of the hot spots on the fermi surface . this indicates a clear breakdown of fermi liquid theory around these points .

1609.08620

metallic spin - density wave ( sdw ) transitions are ubiquitous to strongly correlated materials such as the electron - doped cuprates @xcite , the fe - based superconductors @xcite , heavy fermion systems @xcite , and organic superconductors @xcite . in all these materials , unconventional superconductivity is found to emerge near the onset of sdw order , with the maximum of the superconducting t@xmath2 occuring either at or near the underlying sdw quantum phase transition ( qpt ) . in addition , the vicinity of the sdw transition is often characterized by strong deviations from fermi liquid theory - both in thermodynamic and in single electron properties . more broadly , understanding the properties of quantum critical points ( qcps ) in itinerant fermion systems has attracted much interest over the past several decades @xcite . unlike thermal critical phenomena and qcps in insulating systems , here the critical order parameter fluctuations are strongly interacting with low - energy fermionic quasiparticles near the fermi surface . in the traditional approach to this problem due to hertz @xcite , later refined by millis @xcite , the fermions are integrated out from the outset , leading to an effective _ bosonic _ action for the order parameter fluctuations . the dynamics of the order parameter is found to be overdamped due to the coupling to the fermions , that act as a bath . the effective bosonic action is then treated using conventional renormalization group ( rg ) techniques . while physically appealing , this approach has the drawback that integrating out low - energy modes is dangerous , since it generates non - analytic terms in momentum and frequency that are difficult to treat within the rg scheme . an alternative popular approach has been to use a @xmath3 expansion @xcite , where @xmath4 is the number of fermion flavors . however , in the important case of two spatial dimensions , this approach turns out to be uncontrolled , as well @xcite . alternative expansion parameters have been proposed @xcite , but a fully controlled analytical treatment of qcps in itinerant electron systems has remained one of the grand challenges in strongly correlated electron physics . in addition to the bosonic critical fluctuation dynamics , an important open question regards the behavior of the _ fermionic _ quasiparticles in the vicinity of the transition . the exchange of sdw critical fluctuations can mediate a superconducting instability ; however , the same fluctuations also strongly scatter the quasiparticles , causing them to lose their coherence and leading to the formation of a non - fermi liquid metal . it is not clear which of these effects dominates ; i.e. , is there a well - defined non - fermi liquid regime that precedes superconductivity , or does pairing always preempt the formation of a non - fermi liquid @xcite ? in this work , we perform extensive numerically exact determinantal quantum monte carlo ( qmc ) simulations @xcite of a metallic system in the vicinity of an sdw transition . we use the approach of refs . @xcite , that introduced a two - dimensional multi - band lattice model that captures the generic structure of the `` hot spots '' - points on the fermi surface where quasiparticles can scatter off critical sdw fluctuations resonantly . the universal properties of metallic sdw transitions are believed to depend only on the vicinity of the hot spots . at the same time , the model is amenable to qmc simulations without a sign problem @xcite . our goal here is both to understand the generic properties of the transition , and to provide a controlled benchmark to analytic theories . we present detailed information about the bosonic and fermionic correlations and the interplay with unconventional superconductivity in the vicinity of the qcp . previously , the finite - temperature phase diagram of the model has been characterized , and a dome - shaped superconducting phase peaked near the sdw transition was found @xcite . here , we measure the sdw correlations in the vicinity of the transition , above the superconducting @xmath1 . we find that , over a broad range of parameters , the sdw susceptibility is well described by the following form : @xmath5 here @xmath6 is the ordering wavevector [ chosen to be @xmath7 in our model ] , @xmath8 is a matsubara frequency , and @xmath9 is a parameter used for tuning through the sdw qcp , while @xmath10 , @xmath11 , @xmath12 , and @xmath13 are nonuniversal constants . the function @xmath14 extrapolates to @xmath15 as @xmath16 . importantly , eq . ( [ eq : hertz ] ) captures the behavior of both the bosonic sdw correlations and the susceptibility of a fermion bilinear operator with the same symmetry , establishing the consistency of our analysis . interestingly , @xmath17 has precisely the form predicted by hertz and millis ; in particular , the bosonic critical dynamics are characterized by a dynamic critical exponent @xmath0 . the function @xmath14 does not follow the predicted behavior , however . in a window of temperatures above t@xmath2 , we find a power - law dependence @xmath18 with @xmath19 , in contrast to the linear behavior predicted by millis @xcite . the single - fermion properties above @xmath1 are found to depend strongly on the distance from the hot spots . away from the hot spots , a behavior consistent with fermi liquid theory is observed . at the hot spots , a substantial loss of spectral weight is seen upon approaching the qcp . in a temperature window above t@xmath2 , the fermionic self energy is only weakly frequency and temperature dependent , corresponding to a nearly - constant lifetime of quasiparticles at the hot spots . this behavior indicates a strong breakdown of fermi liquid theory at these points . it is not clear , however , whether it represents the asymptotic behavior at the putative underlying sdw qcp , since superconductivity intervenes before the qcp is reached . finally , in order to probe the interplay between magnetic quantum criticality and superconductivity , we measure the superconducting gap , @xmath20 , and the momentum - resolved superconducting susceptibility , @xmath21 , across the phase diagram . no strong feature is found in @xmath20 at the hot spots . rather , @xmath20 and @xmath21 vary smoothly on the fermi surface . while the pairing interaction may be strongly peaked at wavevector @xmath22 , the resulting gap function does not reflect such a strong momentum dependence . this paper is organized as follows . in sec . [ sec : model ] we describe the model and review its phase diagram . [ sec : sdw_corr ] presents a detailed analysis of the sdw susceptibility across the phase diagram . in sec . [ sec : fermion_corr ] we study the single - fermion properties , providing evidence for the breakdown of fermi liquid theory in the vicinity of the hot spots . [ sec : sc ] analyzes the gap structure in the superconducting state . the cumulative results are put into perspective in sec . [ sec : discussion ] . supplementary data sets and some technical details are presented in the appendices . and @xmath23 fermions . hot spots ( indicated by the black dots ) are linked by the vector @xmath24 with points on the other band . the @xmath25 and @xmath23 fermions have stronger dispersion in direction @xmath26 and @xmath27 , respectively . ] our lattice model is composed of two flavors of spin@xmath28 fermions , @xmath25 and @xmath23 , that are coupled to a real bosonic vector field @xmath29 , which represents fluctuations of a commensurate sdw order parameter at wavevector @xmath30 . the two types of fermions exhibit quasi one - dimensional dispersions along momenta @xmath26 and @xmath27 , respectively , which in the absence of intreractions give rise to the fermi surfaces illustrated in fig . [ fig : fermisurfaces ] . it is precisely this two - flavor structure that fundamentally allows us to set up an action completely devoid of the fermion sign problem in qmc simulations @xcite . yet , the fermi surfaces of this model capture the generic structure of the hot spots , which is generally believed to determine the universal physics near the qcp . as in our previous work on this model @xcite , we assume an @xmath31 symmetric sdw order parameter , i.e. @xmath29 is restricted to the @xmath32 plane . in contrast to the case of an @xmath33 order parameter ( studied in refs . @xcite ) , the easy - plane order parameter implies the existence of a _ finite - temperature _ sdw phase transition of berezinskii - kosterlitz - thouless ( bkt ) character , which we can track in our numerics . in addition , the reduced dimensionality of the order parameter brings a welcome computational benefit as it enables a reduction of the dimensions of all single - fermion matrices by half , greatly improving the efficiency of the numerical linear algebra . our lattice model is given by the action @xmath34 with @xmath35 \psi_{j \alpha s } \notag\\ { } & + \lambda \sum_{\substack { i , s , s ' } } [ \vec{s } \cdot \vec{\varphi}_i ] _ { ss ' } \psi^{\dagger}_{ixs } \psi_{iys'}^{\vphantom{\dagger } } + \mathrm{h.c . } , \notag\\ l_{\varphi } = { } & \frac{1}{2 } \sum\limits_{i } \frac{1}{c^2 } \left ( \frac{d\vec{\varphi}_i}{d\tau } \right)^2 + \frac{1}{2 } \sum\limits_{\left\langle i , j \right\rangle } \left ( e^{i \mathbf{q}\cdot \mathbf{r}_i}\vec{\varphi}_i - e^{i \mathbf{q}\cdot \mathbf{r}_j } \vec{\varphi}_j\right)^2 \notag\\ & + \sum\limits_{i } \left [ \frac{r}{2}\vec{\varphi}_i^2 + \frac{u}{4 } ( \vec{\varphi}_i^2)^2 \right ] .\end{aligned}\ ] ] this action is defined on a square lattice with sites labeled by @xmath36 , where @xmath37 are nearest neighbors . the two fermion flavors are indexed by @xmath38 , while @xmath39 index the spin polarizations and @xmath40 are the pauli matrices . imaginary time is denoted by @xmath41 and @xmath42 is the inverse temperature . the fermionic dispersions are implemented by setting different hopping amplitudes along the horizontal and vertical lattice directions . for the @xmath25-fermions they are given by @xmath43 and @xmath44 , respectively , and for the @xmath23-fermions by @xmath45 and @xmath46 . note that our model is fully @xmath47-symmetric with a @xmath48 rotation mapping the @xmath25-band to the @xmath23-band and vice versa . the tuning parameter @xmath9 allows to tune the system to the vicinity of an sdw instability . in an experimental system , @xmath9 could be proportional to a physical tuning parameter such as pressure or doping . we set the chemical potential to @xmath49 , such that the fermi energy , measured relative to the band bottom , is @xmath50 . the quartic coupling is set to @xmath51 . in the following we mostly focus on the case of a bare bosonic velocity of @xmath52 and a yukawa coupling between fermions and bosons of @xmath53 . occasionally , we also consider other values of @xmath54 and @xmath55 . our numerical analysis of model is based on extensive finite - temperature dqmc @xcite simulations . for the general setup and technical details on the implementation of our dqmc simulations we refer to our earlier paper @xcite and its detailed supplemental online material . here we want to single out a few conceptual aspects of our setup , which have allowed to push our simulations down to temperatures of @xmath56 for system sizes up to @xmath57 sites . first , using a replica - exchange scheme @xcite in combination with a global update procedure @xcite thermal equilibration of our simulations is decidedly improved . most of our simulations were performed in the presence of a weak fictitious perpendicular magnetic " field ( designed not to break time - reversal symmetry ) , which serves to greatly speed up convergence to the thermodynamic limit for metallic systems @xcite . since this technique breaks translational invariance of the fermionic green s function , we can not make use of it to study @xmath58-resolved fermionic observables . for this reason we have carried out additional simulations without the magnetic flux , but with twisted boundary conditions , which allows to increase the momentum space resolution . we give details on these procedures in appendix [ sec : dqmcappendix ] to set the stage for our discussion in the following sections , we summarize the finite - temperature phase diagram of model for @xmath59 and three different values of the yukawa coupling @xmath60 , @xmath61 , and @xmath62 in fig . [ fig : phasediagrams ] . besides a paramagnetic regime for sufficiently high temperature , the dominant feature of these phase diagrams is a quasi - long - range ordered sdw phase whose transition temperature @xmath63 is suppressed with increasing tuning parameter @xmath9 . extrapolating the sdw transition down to the zero temperature provides an estimate of the location of the quantum phase transition at @xmath64 ( indicated by the red star ) . at low temperatures , the sdw transition may become weakly first order @xcite . however , in the temperature range considered here , the transition is either continuous or ( possibly ) very weakly first order . while for @xmath65 this sdw phase is the only ordered phase down to temperatures of @xmath66 , there is an additional superconducting phase emerging in the vicinity of the qpt for the two larger values of the yukawa coupling . for @xmath53 we barely observe the tip of this quasi - long - range ordered phase with a maximum critical temperature of @xmath67 , which is our numerical temperature limit . for @xmath68 , we can clearly map out a superconducting dome with the maximum of the critical temperature @xmath69 . at finite temperatures , both the sdw and the sc finite - temperature transitions are expected to be of bkt type . the sdw susceptibility @xmath70 is found to follow a scaling law @xmath71 with a continuously changing exponent @xmath72 . we identify the transition temperature @xmath63 with the point where the exponent takes the universal value @xmath73 . the sc transition temperature can be both determined via a similar @xmath72-fit or by the point where the superfluid density @xmath74 obtains the universal value of @xmath75 . both estimates are found to agree . the error bars in fig . [ fig : phasediagrams ] are mostly due to finite - size effects . for further details on the procedures employed to identify the different phase transitions see ref . @xcite and its accompanying supplementary online material . a common feature to all three phase diagrams is a change of slope of the sdw phase boundary curve at low temperatures ( @xmath76 ) . this apparent bending " is more pronounced at larger yukawa coupling @xmath55 . such behavior is generally expected to occur at the onset of superconductivity @xcite . here , however , the bending does not visibly track the sc transition temperature . the curvature of the @xmath63 line is also reflected in the sdw susceptibility in the paramagnetic region of the phase diagram , see fig . [ fig : chi - inv - color - l1 ] for @xmath65 and fig . [ fig : chi - inv - color - l1.5-l2 ] in appendix [ sec : magn - corr - at ] for @xmath77 . over a broader range of parameters , the maximum sc transition temperature @xmath78 grows monotonically with increasing of either the yukawa coupling or the boson velocity @xmath54 , as illustrated in fig . [ fig : tc_vs_lambda ] . up to an intermediate coupling strength @xmath79 , @xmath78 rapidly grows as @xmath80 , eventually saturating at stronger coupling . qualitatively , these trends are in agreement with results from eliashberg theory @xcite . across the phase diagram for @xmath65 . we show numerical data obtained at @xmath81 at the temperatures that are indicated by the ticks on the inside of the plot . intermediate temperatures are interpolated linearly , while the high resolution in the tuning parameter @xmath9 is achieved by reweighting . contour lines of @xmath82 are marked gray . ] for different values of the yukawa coupling @xmath55 and ( bare ) boson velocity @xmath54 . the hatched gray region indicates temperatures @xmath83 , which are beyond our numerically accessible temperature range . ] we start our discussion of the quantum critical behavior of model with an examination of magnetic fluctuations across its entire finite - temperature phase diagram . we probe the formation of magnetic correlations both through the susceptibility of the bosonic order parameter , @xmath29 , and through a fermionic bilinear of the same symmetry , which we evaluate independently in the same numerical simulations of the action . as we show below , both susceptibilities exhibit the same behavior , supporting the robustness of our conclusions to be presented . we first consider the bosonic susceptibility calculated from the sdw order parameter @xmath29 in action @xmath84 for a given momentum @xmath85 and matsubara frequency @xmath8 . the expectation values are estimated in a dqmc simulation run at finite temperature @xmath86 and for a specific value of the tuning parameter @xmath9 , indicated here as explicit parameters . at low temperature , we use the following form to fit the data , inspired by hertz theory @xcite : @xmath87 where @xmath88 and @xmath12 are non - universal fitting parameters that describe the momentum dependence in the vicinity of the ordering wavevector @xmath89 , landau damping , and the dependence on the tuning parameter @xmath9 , respectively . the fitting parameter @xmath13 indicates the location of the divergence of @xmath90 . due to the appearance of a superconducting phase at low temperatures , @xmath13 may differ from the actual location of the qpt at @xmath64 . however , within our numerical resolution , we find @xmath91 , where @xmath92 is obtained by extrapolating the finite - temperature transition line @xmath93 , as shown in the phase diagrams of fig . [ fig : phasediagrams ] . and the functional form @xmath94 , which has been fitted for small frequencies @xmath95 and momenta @xmath96 at low temperatures @xmath86 and tuning parameters @xmath97 in the magnetically disordered phase , for ( a ) @xmath65 , ( b ) @xmath98 , and ( c ) @xmath99 . data inside the superconducting phase has been excluded from the fit . for temperatures @xmath100 we restrict the fit to finite frequencies @xmath101 . the correspondence of @xmath82 with the fitted form is shown in the form of 2d histograms over all data points , which are normalized over the total area . in each fit we have minimized @xmath102 ^ 2 $ ] , where @xmath103 is the number of degrees of freedom of the fit and @xmath104 is the statistical error of the data . , title="fig : " ] + and the functional form @xmath94 , which has been fitted for small frequencies @xmath95 and momenta @xmath96 at low temperatures @xmath86 and tuning parameters @xmath97 in the magnetically disordered phase , for ( a ) @xmath65 , ( b ) @xmath98 , and ( c ) @xmath99 . data inside the superconducting phase has been excluded from the fit . for temperatures @xmath100 we restrict the fit to finite frequencies @xmath101 . the correspondence of @xmath82 with the fitted form is shown in the form of 2d histograms over all data points , which are normalized over the total area . in each fit we have minimized @xmath102 ^ 2 $ ] , where @xmath103 is the number of degrees of freedom of the fit and @xmath104 is the statistical error of the data . , title="fig : " ] + and the functional form @xmath94 , which has been fitted for small frequencies @xmath95 and momenta @xmath96 at low temperatures @xmath86 and tuning parameters @xmath97 in the magnetically disordered phase , for ( a ) @xmath65 , ( b ) @xmath98 , and ( c ) @xmath99 . data inside the superconducting phase has been excluded from the fit . for temperatures @xmath100 we restrict the fit to finite frequencies @xmath101 . the correspondence of @xmath82 with the fitted form is shown in the form of 2d histograms over all data points , which are normalized over the total area . in each fit we have minimized @xmath102 ^ 2 $ ] , where @xmath103 is the number of degrees of freedom of the fit and @xmath104 is the statistical error of the data . , title="fig : " ] running extensive dqmc simulations for system sizes @xmath105 , we have evaluated @xmath106 across the three principal phase diagrams of fig . [ fig : phasediagrams ] for different values of the yukawa coupling @xmath60 , @xmath61 , @xmath62 and bare bosonic velocity @xmath59 . restricting our analysis to the magnetically disordered side for each coupling and to temperature scales above the superconducting phase , we find that our calculated susceptibilities are in good agreement with the functional form of eq . . the consistency with eq is illustrated in the panels of fig . [ fig : chi_collapses ] , which show data collapses for a range of small momenta @xmath96 , small matsubara frequencies , low temperatures @xmath107 and tuning parameters @xmath108 . finite - size effects are rather small . considering the variation of the yukawa coupling @xmath55 , we find that the fit to the functional form is slightly worse for stronger coupling @xmath55 , which is also indicated by the larger spread of the data points . this decreasing fit quality may be a consequence of the smaller temperature window available above the superconducting @xmath1 , as well as the associated regime of superconducting fluctuations at @xmath109 @xcite , which increases with yukawa coupling ( see also fig . [ fig : tc_vs_lambda ] ) . as a function of the tuning parameter @xmath9 for @xmath53 at @xmath110 . the black line is a linear fit for @xmath111 and @xmath81 . continuous colored lines through data points have been obtained by a reweighting analysis . ] with the data collapse of fig . [ fig : chi_collapses ] asserting the general validity of the functional form ( [ eq : chi_functional ] ) , we now take a closer look at its individual dependence on tuning parameter , frequency and momentum . first , the dependence on the tuning parameter @xmath9 is illustrated for the inverse susceptibility @xmath112 in fig . [ fig : chi_r_l1.5 ] ( for @xmath53 and @xmath113 ) . for tuning parameters @xmath114 we find that the data for different system sizes follows a linear dependence . the moderate deviation from a perfect kink - like behavior at @xmath13 is likely a combination of finite - size and finite - temperature effects ( see also the finite - size trend shown in the inset of fig . [ fig : chi_r_l1.5 ] ) . a very similar picture emerges for the two other coupling parameters @xmath60 and @xmath68 , for which we show analogous plots in fig . [ fig : chi_r_l1_l2 ] of appendix [ sec : magn - corr - at ] . turning to the frequency dependence of @xmath115 next , we find that for a range of values @xmath108 the frequency dependence is linear for small matsubara frequencies @xmath116 with an apparent cusp at @xmath117 , signaling overdamped dynamics of the order parameter field . this holds both for @xmath118 and for small finite momentum differences @xmath119 . see fig . [ fig : chi_freq_multir_l1.5 ] for an illustration at @xmath53 and appendix [ sec : magn - corr - at ] with fig . [ fig : chi_freq_multir_l1_l2 ] for @xmath60 and @xmath68 . at finite matsubara frequencies @xmath116 , finite - size effects are negligibly small , as evident in the data collapse of @xmath82 for different system sizes in the left panel in fig . [ fig : chi_freq_multir_l1.5 ] . to establish the presence of a @xmath120 term in @xmath82 , we fit it at low frequencies to the form @xmath121 . the fits are shown in fig . [ fig : chi_freq_multir_l1.5 ] . the @xmath120 contribution is clearly dominant in this frequency range . inside the superconducting phase , the @xmath120 term is suppressed ( see fig . [ fig : chi_freq_l3_c2 ] in appendix [ sec : magn - corr - at ] ) . this is presumably due to gapping out of the fermions . for @xmath53 at @xmath66 ( a ) shown at @xmath122 for various momenta @xmath123 and ( b ) shown at various values @xmath97 for @xmath124 . the black line is the best fit of a second degree polynomial @xmath125 to the @xmath118 , @xmath81 low - frequency data , yielding a basically straight line.,title="fig : " ] + third , for the same range of @xmath9 the momentum dependence of @xmath115 is consistent with a quadratic form in @xmath119 , which holds both for @xmath117 and small finite frequencies @xmath116 . see fig . [ fig : chi_mom_multir_l1.5 ] for @xmath53 and appendix [ sec : magn - corr - at ] with fig . [ fig : chi_mom_multir_l1_l2 ] for @xmath60 and @xmath68 . note that due to the discretization of the brillouin zone finite - size effects are more pronounced here than for the frequency dependence . as a function of momentum @xmath123 for @xmath98 at @xmath66 ( a ) shown at @xmath122 for various frequencies @xmath116 and ( b ) shown at various values @xmath126 for @xmath127 . the black line is the best fit of @xmath128 to the @xmath127 , @xmath81 small - momentum data . ] an important independent confirmation that the form is generic to the quantum critical regime is to affirm that it also holds for other sdw order parameters that have the same symmetry . we have examined the correlations of a _ fermion _ bilinear order parameter : @xmath129 in the estimation of @xmath130 we make use of spin rotational symmetry around the @xmath131 axis , @xmath132 . here @xmath133 and @xmath134 are inter - flavor fermion spin operators , which are given by @xmath135 indeed , we find that at small frequencies and momenta , the fermion bilinear sdw susceptibility @xmath130 follows the same functional form as the bosonic sdw susceptibility @xmath106 discussed above . the momenta and frequency dependences of the fermionic bilinear susceptibility at @xmath98 are shown in fig . [ fig : chi_freq_mom_f_l1.5 ] , with the respective dependences of the bosonic susceptibility appearing in figs . [ fig : chi_freq_multir_l1.5 ] and [ fig : chi_mom_multir_l1.5 ] . additional data for the fermionic sdw susceptibility at @xmath60 and @xmath68 is given in fig . [ fig : chi_freq_mom_f_l1_l2 ] of appendix [ sec : magn - corr - at ] . in summary , the dependence of both the bosonic and fermionic sdw susceptibilities on the tuning parameter , frequency , and momentum stand in good agreement with the form . for @xmath98 at @xmath66 and @xmath136 . ( left hand side ) frequency dependence for various momenta @xmath123 . the black line is a fit of the second degree polynomial @xmath125 to the @xmath118 , @xmath81 low - frequency data , yielding a basically straight line . ( right hand side ) momentum dependence for various frequencies @xmath116 . the black line is a fit of @xmath128 to the @xmath127 , @xmath81 small - momentum data . ] and inverse fermionic sdw susceptibility @xmath137 as a function of temperature for @xmath53 at ( a ) @xmath136 and at ( b ) @xmath138 . solid lines indicate fits of @xmath139 to the @xmath81 data at intermediate temperatures . dashed lines are linear fits to the high - temperature data . in each figure the inset shows the same data as the main plot over a more extended temperature range . , title="fig : " ] + and inverse fermionic sdw susceptibility @xmath137 as a function of temperature for @xmath53 at ( a ) @xmath136 and at ( b ) @xmath138 . solid lines indicate fits of @xmath139 to the @xmath81 data at intermediate temperatures . dashed lines are linear fits to the high - temperature data . in each figure the inset shows the same data as the main plot over a more extended temperature range . , title="fig : " ] we now turn to the temperature dependence of the numerically computed bosonic and fermionic sdw susceptibilities @xmath82 and @xmath140 . our numerical data for the temperature dependence of @xmath82 and @xmath140 is shown in fig . [ fig : chi_temp_l1.5 ] for fixed yukawa coupling @xmath98 and two different values of the tuning parameter on the paramagnetic side of the qcp , i.e. for @xmath97 . this data is complemented with similar results for @xmath65 and @xmath68 in fig . [ fig : chi_temp_l1_l2 ] of appendix [ sec : magn - corr - at ] . evidently , the data shows different scaling regimes with increasing temperature . at sufficiently high temperatures , @xmath141 the susceptibilities @xmath142 and @xmath143 are approximately linearly dependent on temperature , as shown in the insets of fig . [ fig : chi_temp_l1.5 ] . in an intermediate temperature regime , however , we observe a crossover to a different functional temperature dependence as shown in the main panels of fig . [ fig : chi_temp_l1.5 ] . in this intermediate temperature window @xmath144 our numerical data is found to reasonably fit functions of the _ quadratic _ form @xmath145 with @xmath146 . unlike the leading dependences on the tuning parameter , frequency and momentum discussed in the previous section , this power - law dependence is not as robust . note that the crossover temperature between the high-@xmath86 linear and intermediate-@xmath86 quadratic behaviors does not depend strongly on the tuning parameter @xmath9 . notably , even for @xmath122 this intermediate regime does not disappear . at still lower temperatures @xmath147 our data might indicate a second crossover to yet different behavior . with the tuning parameter @xmath9 tuned close to its critical value @xmath13 both @xmath106 and @xmath130 are found to be non - monotonic for the smallest temperatures and largest system sizes accessed in this study . the apparent upturn , whose precise location is hard to determine due to finite - size effects ( which are strongest for @xmath148 ) and the enhanced statistical uncertainty at low temperatures , is most likely a precursor effect of superconductivity @xcite , which for @xmath98 sets in just at the lowest temperature we have accessed in this work , @xmath149 . for larger yukawa coupling @xmath68 , where @xmath1 is higher , this non - monotonic behavior is indeed found to be more pronounced as shown in fig . [ fig : chi_temp_l1_l2 ] of appendix [ sec : magn - corr - at ] . note that over the range of temperatures displayed in fig . [ fig : chi - inv - color - l1 ] the leading temperature dependence of @xmath82 is quadratic . this is reflected in the contour lines of @xmath82 in the @xmath150 plane , which have a form @xmath151 , approaching infinite slope at low temperatures . since the data does not allow us to identify a simple functional form for the temperature dependence of @xmath82 , we have opted against taking into account any temperature dependence in the fits for the data collapses shown in fig . [ fig : chi_collapses ] . instead we have constrained the included data to @xmath152 where the overall temperature dependence is rather weak . we now turn to examine the fermionic spectral properties in the metallic state above the superconducting @xmath1 . as our dqmc simulations are performed in imaginary time , there is an inherent difficulty in probing real - time dynamics . to partially circumvent this issue , we use the relation @xcite @xmath153 which connects the readily available imaginary - time ordered green s function @xmath154 , where @xmath155 , with the spectral function @xmath156 of interest . here and in the following , we focus on a single flavor of fermions @xmath23 , suppressing band and spin indices . close inspection of reveals that the behavior of the green s function @xmath157 at long times , i.e for imaginary times close to @xmath158 , provides information about the spectral function integrated over a frequency window of width @xmath86 . in fig . [ fig : fs ] we present the evolution of the fermi surface across the phase diagram . for orientation , the fermi surfaces of the noninteracting system are shown in panel ( a ) . in panels ( b - d ) we show @xmath159 across a quadrant of the brillouin zone for a low temperature @xmath160 . near the magnetic qcp ( panel c ) , there is a clear loss of spectral weight in the immediate vicinity of the hot spots , as compared to the magnetically disordered phase ( panel d ) . upon entering the magnetic phase ( panel b ) , a gap opens around the hot spots . in this section we focus on the parameter set @xmath53 and @xmath59 . the dqmc simulations are carried out with different sets of twisted boundary conditions ( see appendix [ sec : dqmcappendix ] for details ) , providing a four - fold enhancement in @xmath58-space resolution . a fermi liquid is usually characterized by the quasiparticle weight @xmath161 and the fermi velocity @xmath162 . we note that these quantities are only strictly defined at zero temperature . given that the zero - temperature ground state of our model is probably always superconducting , our strategy is to consider finite - temperature proxies for @xmath161 and @xmath162 , and study their behavior over an intermediate temperature range @xmath163 . such proxies , @xmath164 and @xmath165 , can be extracted by considering the imaginary time dependence of @xmath157 near @xmath166 and fitting it to the fermi liquid form @xcite @xmath167 where @xmath168 . in a complementary approach we consider the matsubara frequency dependence of the green s function @xmath169 . in a fermi liquid at low temperatures we have @xcite @xmath170^{-1 } \label{eq : g_fl}\ ] ] up to higher order terms in temperature , frequency or the distance from the fermi surface . it is then natural to define the finite - temperature quantities @xmath171 and @xmath172 where @xmath173 is the first matsubara frequency at temperature @xmath86 . in the zero temperature limit , @xmath174 , and similarly for @xmath162 . we therefore use the finite - temperature observables and as alternative proxies for the quasiparticle spectral weight and fermi velocity , respectively . in a quadrant of the brillouin zone . the dashed line in panel ( a ) corresponds to the noninteracting fermi surface of the @xmath25 fermions , shifted by @xmath22 . ( c - d ) the quasiparticle weights @xmath175 and @xmath176 along the fermi surface . the location of the hot spot is indicated by the red marker . here we show data obtained for @xmath177 . ] figure [ fig : z_vs_k ] shows the momentum dependence of @xmath178 for temperature @xmath179 . with @xmath9 tuned close to the location of the qcp at @xmath92 , @xmath178 is suppressed in the vicinity of the hot spots , as shown for one quadrant of the brillouin zone in fig . [ fig : z_vs_k](a ) and along the fermi surface in fig . [ fig : z_vs_k](c ) . this stands in sharp contrast to the featureless behavior of @xmath178 in the magnetically disordered phase , as shown in figs . [ fig : z_vs_k](b , d ) . we find qualitative agreement between the two proxies @xmath178 and @xmath180 throughout , as illustrated in panels ( c ) and ( d ) of fig . [ fig : z_vs_k ] . here , we numerically identify and track the fermi surface as the maxima of @xmath181 at fixed @xmath26 . for different momenta @xmath26 along the fermi surface ( a ) in the vicinity of the qcp at @xmath92 and ( b ) in the disordered side . here we show data obtained for @xmath177 . ] along the fermi surface . ( c)-(d ) the velocity renormalization @xmath182 . the location of the hot spot is indicated by the red marker . here we show data obtained for @xmath177 at @xmath183 . ] the temperature dependence of @xmath184 is shown in fig . [ fig : z_vs_t ] . for momenta away from the hot spots , we find @xmath185 to be nearly flat in temperature and to approach a constant as @xmath16 . a different picture emerges at the hot spots , i.e. for @xmath186 . here @xmath187 remains flat only in the magnetically disordered phase @xmath188 , whereas the quasiparticle weight @xmath187 decreases substantially at the critical coupling @xmath189 as the temperature is lowered towards the qcp , see fig . [ fig : z_vs_t](a ) . while our numerical data does not allow for a simple extrapolation towards @xmath190 , the results are not inconsistent with a vanishing of @xmath191 , indicating a breakdown of fermi - liquid behavior at this point . for different temperatures and two momenta along the fermi surface , ( a ) in the vicinity of the qcp at @xmath92 and ( b ) on the disordered side . here we show data obtained for @xmath81 . the data for @xmath192 is indicated by full circles , the momentum away from the hotspot is indicated by empty squares . ] figure [ fig : vf_vs_k ] shows the velocity @xmath193 along the fermi surface . the qualitative behavior of the velocity does not differ substantially between @xmath194 [ fig . [ fig : vf_vs_k](a ) ] and @xmath195 [ fig . [ fig : vf_vs_k](b ) ] . the insets show the ratio @xmath196 , where @xmath197 is the fermi velocity of the noninteracting system . a small feature might be visible in the vicinity of the hot spot , but there is certainly no evidence of a substantial suppression of @xmath198 close to @xmath92 . having found a substantial suppression of the quasiparticle weight tuned close to the qcp , we now directly examine the frequency dependence of the self energy , @xmath199 , defined via @xmath200 . the imaginary part of the self energy is shown in fig . [ fig : self_energy ] . with @xmath194 tuned close to the qcp , see fig . [ fig : self_energy](a ) , the self energy close to the hot spots is found to be nearly frequency independent , consistent with a _ constant _ , yet small , scattering rate @xmath201 which is only weakly dependent on temperature . away from the hot spot the self energy is linear with frequency , with a substantially smaller intercept . moving away from the critical point [ fig.[fig : self_energy](b ) ] , the self energy at all momenta decreases rapidly as the frequency is lowered . after concentrating our discussion on the manifestation of quantum critical behavior in the normal state , we now consider the effect of the qcp on the superconducting state that emerges in its vicinity . we begin by considering the fermionic green s function for temperatures @xmath202 . in this regime , the single - particle excitation energy @xmath203 can be extracted , as demonstrated in appendix [ sec : delta ] , from the decay of the single - particle green s function at intermediate times @xmath204 ( where @xmath205 is a microscopic scale ) . the so - extracted excitation energy @xmath203 is plotted in fig . [ fig : ek_from_gtau ] , which shows that , across the brillouin zone , @xmath203 has a broad minimum in the vicinity of the noninteracting fermi surface . from these momentum - resolved energy bands we extract the superconducting gap @xmath206 as the minimum of @xmath203 with respect to @xmath27 . as seen in fig . [ fig : ek_from_gtau ] ( b ) , the superconducting gap @xmath206 varies smoothly across momentum space , without any significant features at the hot spots . in this section we choose parameters @xmath207 and @xmath208 , as in ref . the maximal @xmath1 for this value of @xmath55 is high enough to allow us to explore properties of the superconducting state significantly below @xmath1 . of the @xmath23 fermions , as extracted from the imaginary - time evolution of the green s function @xmath209 across the brillouin zone , cf . [ fig : ek_from_gtau_fits ] of the appendix . ( b ) single - particle gap @xmath206 . for both panels data is for parameters @xmath210 , @xmath211 , @xmath212 and @xmath213 @xcite and a system size of @xmath214 . several twisted boundary conditions were combined for a four - fold enhancement of the resolution in @xmath58-space , see appendix [ sec : dqmcappendix ] . ] with the band @xmath215 shown in panel ( a ) and the band @xmath216 shown in ( b ) . data is calculated for parameters @xmath210 , @xmath211 , @xmath212 and @xmath217 and system size @xmath81 . ] at higher temperatures , close to @xmath1 , additional information can be obtained by considering the momentum - resolved superconducting susceptibility @xmath218 where @xmath219 is the singlet superconducting pair amplitude on the band @xmath38 . here we focus on the intraband , spin - singlet channel since it is the leading instability @xcite . figure [ fig : pairing_p ] shows the optimal pair amplitude @xmath220 , corresponding to the maximal eigenvalue of the matrix @xmath221 at a temperature slightly above @xmath1 . the pair amplitude of the band @xmath215 , shown in fig . [ fig : pairing_p ] ( a ) , is of the opposite sign to the amplitude on the band @xmath216 , shown in fig . [ fig : pairing_p ] ( b ) . in fact , the two amplitudes are related precisely by a @xmath48 rotation , highlighting the @xmath222-wave symmetry of the superconducting order parameter . the optimal pair amplitude is found to be maximal around the ( noninteracting ) fermi surface . the variation of @xmath220 along the fermi surface is weak , again showing no strong features at the hot spots . in this work , we have explored the properties of a metal on the verge of an sdw transition . we focused on the critical regime upon approaching the transition , characterized by a rapid growth of the sdw correlations , but still above the superconducting transition temperature . our main conclusion is that , in this regime , the sdw correlations are remarkably well described by a form similar to that predicted by hertz - millis theory , eq . ( [ eq : hertz ] ) ( although the temperature dependence of the sdw susceptibility deviates from the expected form ) . this holds both for the correlations of the bosonic sdw order parameter field , and for an sdw order parameter defined in terms of a fermion bilinear . in the same regime , we find evidence for strong scattering of quasiparticles near the hotspots , leading to a breakdown of fermi liquid theory at these points on the fermi surface . the scattering rate at the hotspots ( extracted from the fermion self energy ) is only weakly temperature and frequency dependent , down to @xmath223 , where we suspect that superconducting fluctuations begin to play a role ; it is out of this unusual metallic state that the superconducting phase emerges . in addition , we have studied the structure of the superconducting gap near the sdw transition . unlike the single - fermion green s function in the normal state , it does not have a sharp feature at the hot spots ; rather , it is found to vary smoothly across the fermi surface . experimentally , a broad maximum of the superconducting gap near the hot spots was observed in a certain electron doped cuprate @xcite . eliashberg theory predicts a peak of the gap function at the hot spots at weak coupling @xcite and it remains to be seen whether such behavior appears in our model at weaker coupling . it is interesting to discuss our results in the context of the existing theories of metallic sdw transitions . first , the fact that hertz - millis theory successfully describes many features of our data is non - trivial , in view of the fact that it has no formal justification , even in the large @xmath4 limit @xcite . however , as we saw , an extension of the hertz - millis analysis to finite temperature predicts that at criticality , @xmath224 , in apparent disagreement with our data . this may be due to the limited temperature window we can access without hitting the superconducting t@xmath2 , or to effects beyond the one - loop approximation . an important conclusion of our study is that the sdw critical point is always masked by a superconducting phase that we have studied , where the maximal t@xmath2 is smaller than the lowest temperature in our study , we see enhancement of diamagnetic fluctuations , indicating that we are not too far above t@xmath2 . ] . as a result , it seems likely that the critical metallic regime is never parametrically broad , and one can not sharply define scaling exponents within the metallic phase xy universality class , since the fermi surface is gapped . ] . as mentioned above , the sdw correlations follow a hertz - millis form and hence it is tempting to associate with them critical exponents , i.e. a mean - field value @xmath225 for the correlation length exponent , and a dynamical critical exponent @xmath0 . the fermionic quasiparticles at the hot spots , however , do not exhibit this scaling behavior . in particular , the expected scaling law @xmath226 for the fermion self - energy at the hot spots is not seen within our accessible temperature range . one can imagine trying to access the `` bare '' metallic quantum critical point by suppressing the superconducting transition . presumably , this can be done by adding to the model a term that breaks time reversal and inversion symmetries ( such a term would lift the degeneracy of fermionic states with opposite momenta , and hence remove the cooper instability ) . breaking time reversal symmetry , however , gives rise to a sign problem , so it is not clear whether the critical behavior can be accessed within the qmc technique . alternatively , one could try to understand the metallic regime above t@xmath2 in our model as a crossover regime of an underlying `` nearby '' metallic critical point , where some correlators already exhibit their asymptotic behavior ( such as the sdw order parameter correlations ) , while others do not ( e.g. , the single - fermion green s function ) . interestingly , a simple , non - self consistent one - loop calculation of the fermionic self energy in our model does show a broad range of temperature and frequency where the self energy at the hot spot is nearly constant , before eventually settling into the expected @xmath227 behavior ( see appendix [ sec : perturbation_theory ] ) . this calls for a more detailed comparison between our numerically exact dqmc results and a detailed self - consistent one - loop analysis . preliminary results show that this approximation is surprisingly successful in capturing at least some of the physics of our model @xcite . to what extent such a crossover behavior , characterized by a nearly - constant fermionic lifetime at the hotspots , is ubiquitous across different models , as well as in real materials , remains to be seen . it is interesting to note , however , that a similar behavior has been observed in a study of a _ nematic _ transition in a metal @xcite . it would be interesting to systematically look for such behavior in angle - resolved photoemission spectroscopy in the electron - doped cuprates , where anomalously large broadening of the quasiparticle peaks is seen near the hot spots @xcite . another important aspect of the metallic state in the critical regime , which we have not addressed in this work , is the electrical conductivity . the optical conductivity may be strongly affected by the presence of an sdw critical point , even without quenched disorder @xcite . extracting the conductivity from quantum monte carlo simulations requires an analytic continuation , and is therefore intrinsically more difficult ( and involves more uncertainties ) than calculating thermodynamic and imaginary - time quantities . nevertheless , we have obtained preliminary results showing strong effects of the critical fluctuations on the low - frequency optical conductivity @xcite . a full analysis of the conductivity is deferred to future work . we thank a. chubukov and c. varma for useful discussions on this work . e. b. and y. s. also thank r. fernandes , s.a . kivelson , s. lederer and x. wang for numerous discussions and collaborations on related topics . thanks the hospitality of the aspen center for physics , where part of this work was done . the numerical simulations were performed on the cheops cluster at rrzk cologne , the juropa / jureca clusters at the forschungszentrum jlich , and the atlas cluster at the weizmann institute . y.s . and e.b . were supported by the israel science foundation under grant no . 1291/12 , by the us - israel bsf under grant no . 2014209 , and by a marie curie reintegration grant . e. b. was supported by an alon fellowship . m. g. thanks the bonn - cologne graduate school of physics and astronomy ( bcgs ) for support . in this appendix we elaborate on a number of specific technical aspects of our numerical implementation of the determinantal quantum monte carlo ( dqmc ) approach . we refer readers looking for a more comprehensive discussion of the general dqmc setup to our previous paper @xcite and in particular its supplementary online material . we study the lattice model described by the action ( [ eq : action ] ) at finite temperature in the grand canonical ensemble . after discretizing imaginary time and integrating out the fermionic degrees of freedom , the partition function reads @xmath228 where @xmath229 is the equal - time green s function matrix for a fixed configuration of the bosonic order parameter field @xmath29 . we use an imaginary time step of @xmath230 in all calculations . the dqmc method samples configurations of @xmath29 according to their weight @xmath231 . for efficient monte carlo sampling , is it highly advantageous to consider models in which the determinant in is guaranteed to be positive , thereby avoiding the notorious fermion sign problem . for the two - band model that has been proposed in ref . @xcite this is ensured by an antiunitary symmetry of the action @xcite , written in first quantization as @xmath232 here @xmath233 is the complex conjugation operator and @xmath234 ( @xmath235 ) are pauli matrices acting in spin ( flavor ) space , respectively . in this manuscript we have modified the model of ref . @xcite in two regards . first , as in our previous paper @xcite , we consider an easy - plane sdw order parameter , rather than an @xmath33 symmetric order parameter used in @xcite . second , we couple the system to a fictitious , spin and band dependent orbital `` magnetic '' field , whose flux through the system is given by @xmath236 . here , we place the two - dimensional lattice ( @xmath237 directions ) on a torus . the @xmath238 and @xmath239 components of the flux twist the boundary conditions along the @xmath239 and @xmath238 directions , respectively , while the @xmath131 component acts as an orbital magnetic field with a uniform flux per plaquette through the torus . as in the definition of the action , @xmath38 is a fermion flavor index and @xmath240 is a spin index . the flux @xmath241 is restricted to be an integer multiple of the flux quantum @xmath242 . in order to preserve the symmetry , fermions of different spin and flavor are coupled to this fictitious flux as @xmath243 with @xmath244 . note that the inter - band sign change is not strictly necessary to avoid the fermion sign problem . importantly , in the thermodynamic limit @xmath245 the fictitious field vanishes . in our previous paper @xcite and for some of the results in the present one , we have chosen @xmath246 . although such a flux is useful in reducing spurious finite - size effects at low temperatures @xcite , it also breaks lattice translational symmetry , hampering the analysis of non - local , fermionic correlation functions such as the momentum - resolved green s function . in order to measure such quantities , for the present paper , we have run additional simulations without applying a perpendicular flux @xmath247 , but instead with _ in - plane _ fields , such that @xmath248 , where @xmath249 . this procedure is equivalent to having twisted boundary conditions , such that the allowed momenta for the @xmath250 fermions are @xmath251 where @xmath252 , @xmath253 are integers , thereby enhancing the momentum - space resolution of fermionic observables fourfold . * ( a ) * @xmath98 + for ( a ) @xmath98 and ( b ) @xmath99.,title="fig : " ] + * ( b ) * @xmath99 + for ( a ) @xmath98 and ( b ) @xmath99.,title="fig : " ] in this appendix we present additional data for the bosonic and fermionic sdw susceptibilities @xmath254 and @xmath255 . we begin with fig . [ fig : chi - inv - color - l1.5-l2 ] , which shows @xmath142 across the phase diagrams for @xmath98 and @xmath99 . it illustrates how the bending of the sdw finite - temperature transition line is also visible in @xmath82 for @xmath97 , similarly to the behavior for @xmath65 in fig . [ fig : chi - inv - color - l1 ] , where , however , no superconducting phase has been observed within our temperature resolution . next , to complement the discussion in sec . [ sec : sdw_corr ] , where we have focused on the yukawa coupling @xmath53 and a bosonic velocity of @xmath52 , we here show detailed data for values @xmath60 and @xmath68 at the same velocity @xmath54 . in fig . [ fig : chi_r_l1_l2 ] we show @xmath256 as a function of @xmath9 for constant @xmath113 , where the same linear dependence as for @xmath65 ( fig . [ fig : chi_r_l1.5 ] ) is apparent . as we show in fig . [ fig : chi_freq_multir_l1_l2 ] both for @xmath122 and a range of @xmath97 the leading frequency dependence of @xmath82 is clearly linear , similarly to @xmath98 ( fig . [ fig : chi_freq_multir_l1.5 ] ) , whereas the leading momentum dependence shown in fig . [ fig : chi_mom_multir_l1_l2 ] is quadratic in @xmath119 , which is again comparable to @xmath98 ( fig . [ fig : chi_mom_multir_l1.5 ] ) . note that for @xmath99 we show data at a higher temperature @xmath179 rather than at @xmath66 , because otherwise the system would be in the superconducting phase , where the frequency dependence is significantly altered . for small frequencies and momenta the fermionic susceptibility @xmath140 in fig . [ fig : chi_freq_mom_f_l1_l2 ] behaves similarly to the bosonic @xmath82 ( see fig . [ fig : chi_freq_mom_f_l1.5 ] for @xmath98 ) . the temperature dependence of both @xmath82 and @xmath140 is demonstrated in fig . [ fig : chi_temp_l1_l2 ] . as in the case of @xmath53 ( fig . [ fig : chi_temp_l1.5 ] ) , we observe a linear regime at high temperatures , shown in the insets , and a crossover region , where we can fit a quadratic law . for @xmath65 this second region extends down to lower temperatures than for @xmath53 , where @xmath1 is higher . at @xmath68 the data for @xmath257 is from within the superconducting phase . moreover , at @xmath258 the system is partially inside the magnetic quasi - long range order phase ( cf . the phase diagram in fig . [ fig : phasediagrams ] ) . finally , to illustrate the influence of superconductivity on the frequency dependence of the sdw susceptibility , we show data from deep within the superconducting phase in fig . [ fig : chi_freq_l3_c2 ] . here we have chosen a data set with different values of the yukawa coupling @xmath210 and the bosonic velocity @xmath211 ( as in ref . @xcite ) since @xmath259 is about twice as high for these parameters as for @xmath68 , @xmath59 . in contrast to the data at @xmath260 shown in figs . [ fig : chi_freq_multir_l1.5 ] and [ fig : chi_freq_multir_l1_l2 ] , the low - frequency behavior is clearly no longer purely linear indicative of a suppression of landau damping in the superconducting phase . + for @xmath210 , @xmath211 at @xmath261 and @xmath262 , close to where @xmath1 is highest for this set of parameters , shown for various momenta @xmath123 . the black line is the best fit of a second degree polynomial @xmath125 to the @xmath118 , @xmath81 low - frequency data . ] in this appendix , we consider the fermionic self energy in a one - loop approximation . to this order , the self energy is given by @xmath263 where @xmath264 is the non - interacting green s function and @xmath265 is a bosonic matsubara frequency . deferring more systematic calculations for future work , here we do not attempt a full , self - consistent solution of the coupled eliashberg equations @xcite for the sdw correlations and the fermionic green s function . instead , we use the non - interacting green s function and @xmath106 taken from a lattice , discretized imaginary time version of eq . @xmath266\\ & + \frac{2a_{\omega}}{\delta\tau}\left|\sin\left(\frac{\delta\tau\omega_{m}}{2}\right)\right| , \end{split } \label{eq : chi_perturbative}\ ] ] where the parameters @xmath267 are taken from a fit to the dqmc data , see section [ sec : boson - sdw - susc ] . strictly speaking , this procedure is not justified . however , since within a self - consistent eliashberg theory , @xmath106 has the form ( [ eq : chi_functional],[eq : chi_perturbative ] ) , we expect our simplified approximation to capture the general behavior of the self - consistent theory . , ( b ) further away from the qcp . an sdw correlator of the form was used , with the parameters taken from the fit to the dqmc data for @xmath98 , @xmath59 , with @xmath268 , and @xmath269 , as shown in fig . [ fig : chi_collapses](b ) . the data for @xmath192 is indicated by full circles , the momentum away from the hotspot is indicated by empty squares . ] in fig . [ fig : perturbative_self_energy ] we show the imaginary part of the self energy . the results bear some similarities to the dqmc data , shown in fig . [ fig : self_energy ] . whereas at moderate @xmath270 or away from the hotspots the self energy is rapidly diminished as the frequency @xmath116 is lowered , the behavior at the hotspots as @xmath9 approaches @xmath92 is different . there , as a function of temperature , a change of slope occurs in the frequency dependence of the self energy , where at intermediate temperatures @xmath271 the self energy is nearly frequency independent . only at lower temperatures , @xmath272 starts resembling the expected @xmath273 form . in comparing with the dqmc results in fig.[fig : self_energy ] , we note the similar magnitude of the self energy . however , the dqmc results show a far weaker temperature dependence at the hotspots for @xmath9 close to @xmath92 . in this appendix we provide a detailed description of the procedure by which we extract the single - particle excitation energy @xmath203 in the superconducting state , which we discuss in sec . [ sec : sc ] of the main text . the single - particle green s function @xmath209 is found to exhibit a characteristic imaginary - time evolution as shown in fig . [ fig : ek_from_gtau_fits ] . at intermediate times , @xmath204 , where @xmath274 is some microscopic time scale , the single - particle green s function decays exponentially as @xmath275 and similarly , for times @xmath276 , @xmath277 at long times , @xmath278 the green s function is substantially suppressed and statistical errors dominate the signal . we extract the decay constants @xmath279 and @xmath280 from appropriate exponential fits and define the single - particle excitation energy as their minimum @xmath281 . for ( a ) several momenta along the noninteracting fermi surface and ( b ) several momenta along a cut perpendicular to the fermi surface . here , @xmath214 , @xmath66 , @xmath210 , and @xmath211 . shaded regions indicate the statistical uncertainty . the solid lines are exponential fits . ] for a qualitative understanding of these results , we consider the behavior of the green s function in a fermi liquid and in a bardeen - cooper - schrieffer ( bcs ) superconductor . the fact that @xmath209 exhibits exponential behavior can be interpreted as arising from a peak in the spectral function @xmath282 , occuring at a non - zero frequency , as can be seen from . in a fermi liquid , the spectral function at a given momentum has a single , sharp peak at the energy of the quasiparticle . it then follows that @xmath209 has the form of a single exponential , with hole - like quasiparticles obeying and particle - like quasiparticles obeying . indeed , in our simulations in the normal state we find monotonic behavior of @xmath283 ( not shown ) . it is illuminating to contrast this behavior with the bcs state , where a superposition of hole - like and particle - like excitations is allowed . in this case , the spectral function consists of two delta - function peaks at @xmath284 , such that the green s function takes the form @xmath285 here @xmath286 with the quasiparticle dispersion @xmath287 and the gap @xmath20 , and @xmath288 , @xmath289 are particle and hole amplitudes , respectively . the resulting green s function is non - monotonic , showing a minimum at a finite imaginary time . while our numerical data below @xmath1 shares some similarities with the bcs form , it differs in two notable ways . first , clear exponential behavior is not seen at short times @xmath290 . second , the particle and hole excitation energies differ , i.e. @xmath291 , with the difference more pronounced away from the fermi surface , as illustrated in fig . [ fig : ek_from_gtau_fits](b ) .

a directed graph ( digraph ) is formed by vertices and arcs ( directed edges ) from one vertex to another . a feedback vertex set ( fvs ) is a set of vertices that contains at least one vertex of every directed cycle in this digraph . the directed feedback vertex set problem aims at constructing a fvs of minimum cardinality . this is a fundamental cycle - constrained hard combinatorial optimization problem with wide practical applications . in this paper we construct a spin glass model for the directed fvs problem by converting the global cycle constraints into local arc constraints , and study this model through the replica - symmetric ( rs ) mean field theory of statistical physics . we then implement a belief propagation - guided decimation ( bpd ) algorithm for single digraph instances . the bpd algorithm slightly outperforms the simulated annealing algorithm on large random graph instances . the predictions of the rs mean field theory are noticeably lower than the bpd results , possibly due to its neglect of cycle - caused long range correlations . janary 17 , 2016 ( first version ) ; april 04 , 2016 ( revised version )

1604.00873

directed graphs ( digraphs ) are widely used to describe interactions in technological , biological , and social complex systems @xcite . a digraph is composed of vertices and arcs ( directed edges ) which link between vertices . for instance , the gene regulation system of a biological cell can be modeled as a digraph of genes , in which an arc points from gene @xmath0 to gene @xmath1 if @xmath0 regulates the expression of @xmath1 @xcite . real - world complex systems are full of feedback interactions and adaptation mechanisms , and the digraphs of these systems usually contain an abundant number of directed cycles , which make the system s dynamical properties difficult to predict and to externally control @xcite . not all vertices and arcs are equally important in feedback interactions . some vertices and arcs may participate in a much greater number of directed cycles than other vertices and arcs . a minimum feedback vertex set ( fvs ) contains a smallest number of vertices whose deletion from the digraph destroys all the directed cycles . similarly , a minimum feedback arc set ( fas ) is an arc set of smallest cardinality such that every directed cycle of the digraph has at least one arc in this set . in terms of collective effect , the vertices in a fvs and the arcs in a fas may be most significant to the dynamical complexity of a complex networked system @xcite , and these vertices and arcs may also be optimal targets of distributed network attack processes @xcite . the directed feedback problems , namely constructing a minimum fvs and a minimum fas for a generic digraph , are combinatorial optimization problems in the nondeterministic polynomial - hard ( np - hard ) class and therefore are intrinsically very difficult @xcite . some progresses have been achieved by computer scientists and applied mathematicians in understanding this classic hard problem since the 1990s @xcite . approximate algorithms with provable bounds have been designed ( see , for example @xcite ) and efficient heuristic algorithms such as greedy local search @xcite and simulated annealing @xcite have been implemented and tested on benchmark small problem instances . yet unlike the fvs problem on an undirected graph @xcite , it is still very difficult to give tight bounds on the minimal cardinality of the directed fvs and fas problems . in this paper we study the directed fvs problem using statistical physics methods . our method is also applicable to the fas problem since it is essentially equivalent to the fvs problem @xcite . we construct a spin glass model for the directed fvs problem and derive the replica - symmetric ( rs ) mean field theory for this model . the mean field theory enables us to estimate the minimum directed fvs cardinality for random digraphs . based on this mean field theory we implement a belief propagation - guided decimation ( bpd ) algorithm and apply this message - passing algorithm to large random digraph instances . the bpd algorithm slightly outperforms the simulated annealing algorithm @xcite in terms of the fvs cardinality and the arc density of the remaining directed acyclic graph ( dag ) , while the computing times of the two algorithms are comparable to each other . our work will be helpful for future investigations on the dynamical complexity of various real - world networked systems and for further studies of targeted attacks on directed networks . in this paper we also point out a major difficulty of the rs mean field theory in treating directed cycle - caused long range correlations . directed and undirected cycles cause long - range correlations and frustrations in dynamical systems and spin glass models . they are very important network structural properties , but theoretical efforts directly treating them as constraints are quite lacking . directed cycles are global properties of a digraph ( each of which may involve a lot of vertices and arcs ) . compared with problems with local constraints ( such as the @xmath2-satisfiability problem @xcite ) , optimization problems constrained by directed cycles are much more difficult to tackle theoretically . the present paper is a continuation of our earlier report @xcite which treated the undirected fvs problem successfully . we realized that the directness of the cycle constraints makes the directed fvs problem much harder than its undirected counterpart . it is clear that further efforts are needed to overcome the gap between the bpd results and the rs predictions in fig . [ fig : rdfvs ] . the next section defines the directed fvs and fas problems and explains their essential equivalence . we then introduce a fvs spin glass model in sec . [ sec : model ] and study it by rs mean field theory in sec . [ sec : rs ] . in sec . [ sec : bpd ] we describe the bpd message - passing algorithm and compare its performance with the simulated annealing algorithm on random digraphs . we conclude this paper in sec . [ sec : conclusion ] . the two appendices contain some additional technical discussions . consider a simple digraph @xmath3 . this digraph contains @xmath4 vertices whose indices ( say @xmath5 ) range from integer values @xmath6 to @xmath4 . there are @xmath7 arcs in this digraph , each of which points from one vertex ( say @xmath8 ) to another different vertex ( say @xmath9 ) and is denoted as @xmath10 $ ] . there is no self - arc pointing from a vertex to itself , and there is at most one arc from any vertex @xmath8 to any another different vertex @xmath9 . the arc density @xmath11 of a digraph is just the ratio between arc number @xmath7 and vertex number @xmath4 , that is @xmath12 . each vertex @xmath8 of the digraph has a positive weight @xmath13 whose meaning is context - dependent . for instance , in the problem of complex - system control @xcite , the weight @xmath13 may be the cost of monitoring the vertex @xmath8 . the vertex weights are fixed parameters of the digraph and can not be modified . in the actual numerical computations of this paper , each vertex weight @xmath13 is set to be @xmath14 just for simplicity . if there is an arc @xmath10 $ ] from a vertex @xmath8 to another vertex @xmath9 but the reverse arc @xmath15 $ ] is absent , then @xmath9 is said to be a _ child _ of @xmath8 and @xmath8 is said to be a _ parent _ of @xmath9 . if there are two arcs @xmath10 $ ] and @xmath15 $ ] of reverse direction between vertices @xmath8 and @xmath9 , then vertex @xmath8 is said to be a _ brother _ of vertex @xmath9 and @xmath9 a brother of @xmath8 . the in - degree and out - degree of vertex @xmath9 are defined as the total number of arcs pointing to @xmath9 and pointing from @xmath9 , respectively . for example , vertex @xmath16 in fig . [ fig : digraph ] is a parent of vertex @xmath6 and is a child of vertex @xmath17 , while vertices @xmath17 and @xmath6 are brother of each other . a directed path in a digraph @xmath3 is a sequence of arcs which connect a sequence of vertices , for example a directed path @xmath18 , \ [ j_1 , j_2 ] , \ \ldots , \ [ j_{n-1 } , j_{n } ] , \ [ j_n , j]$ ] from a start vertex @xmath8 to an end vertex @xmath9 . if the start vertex and the end vertex of a directed path are identical , then this path is referred to as a directed cycle . some examples of directed paths and cycles can be found in the small graph of fig . [ fig : digraph ] . a feedback vertex set of digraph @xmath3 is a subset @xmath19 of the @xmath4 vertices such that if all the vertices of this set and the attached arcs are removed from @xmath3 the remaining digraph will have no directed cycle . in other words , the set @xmath19 contains at least one vertex of every directed cycle . as an example , the set @xmath20 is a fvs for the small digraph of fig . [ fig : digraph ] . similarly , a feedback arc set is a subset of the @xmath7 arcs such that it contains at least one arc of every directed cycle of digraph @xmath3 . for the example of fig . [ fig : digraph ] one can easily verify that the arc set @xmath21 , [ 13 , 14]\}$ ] is a minimum fas . the weight @xmath22 of a feedback vertex set @xmath19 is just the total weight of its ingredient vertices : @xmath23 . the fundamental goal of the feedback vertex set problem is to construct a fvs for a given digraph with total weight as small as possible . this np - hard problem has been studied by mathematicians and computer scientists , see for example references @xcite , but it appears that this important problem has not yet been treated by statistical physics methods . here we work on the directed fvs problem using the mean field spin glass theory and the belief propagation message - passing algorithm . the feedback arc set problem for a digraph @xmath3 ( left ) can be converted into a feedback vertex set problem for a new digraph @xmath24 ( right ) . each arc @xmath25 $ ] of @xmath3 is mapped to a vertex ( denoted by a square ) in @xmath24 , and an arc is created in @xmath24 from a vertex @xmath10 $ ] to another vertex @xmath26 $ ] if and only if @xmath27 . the left digraph becomes acyclic after arc @xmath28 $ ] is deleted , and correspondingly the right digraph becomes acyclic after vertex @xmath28 $ ] and its attached arcs are deleted . , scaledwidth=60.0% ] the spin glass model of this work and the corresponding bpd algorithm is also applicable to the fas problem . this is because the fas problem is equivalent to the fvs problem on a modified digraph @xcite ( fig . [ fig : fastofvs ] ) . to construct a fas for a given digraph @xmath3 , we can first map each arc @xmath10 $ ] of @xmath3 to a vertex ( also denoted as @xmath25 $ ] ) of a new digraph @xmath24 , and then we setup an arc in @xmath24 from a vertex @xmath25 $ ] to another vertex @xmath26 $ ] if and only if @xmath27 . it is easy to verify that there is a one - to - one correspondence between a directed cycle of digraph @xmath3 and a directed cycle in the digraph @xmath24 , and a fvs of digraph @xmath24 is just a fas of digraph @xmath3 ( fig . [ fig : fastofvs ] ) . in the present paper we focus on the directed fvs problem . treatment of the fas problem on random and real - world networks and its application to the network destruction problem will be reported in a separate paper @xcite . we now describe in detail a spin glass model for the directed fvs problem , which was briefly introduced in the last section of @xcite by the present author . similar model systems have also been discussed by other authors , see , e.g. , @xcite . first , a height state @xmath29 is assigned to each vertex @xmath8 of a digraph @xmath3 . this height state can take integer values in the interval @xmath30 , with @xmath31 being the allowed maximal height in the model . the value of @xmath31 is adjustable in the theoretical calculations , but it should not exceed the total number @xmath4 of vertices . setting @xmath31 to a larger value will result in better theoretical results . on the other hand , the memory space and the computing time both increase linearly with @xmath31 . in this paper we mainly work with @xmath32 but we will also discuss the effect of @xmath31 on the theoretical results . a generic configuration of the system is denoted as @xmath33 with @xmath34 . if the height of a vertex @xmath8 is @xmath35 , this vertex is referred to as being unoccupied or empty , otherwise @xmath8 is referred to as being occupied . to ensure that there are no directed cycles within the set of occupied vertices , we impose the following constraint on each arc @xmath10 $ ] of digraph @xmath3 : if both vertex @xmath8 and vertex @xmath9 are occupied ( @xmath36 and @xmath37 ) , then the height of vertex @xmath8 must be less than that of vertex @xmath9 ( @xmath38 ) . a configuration @xmath33 is referred to as a legal configuration if it satisfies all the @xmath7 arc constraints . if @xmath33 is a legal configuration of digraph @xmath3 , then the vertex heights along any directed path of occupied vertices form a strictly increasing sequence . therefore there must be no directed cycles within the set of occupied vertices of configuration @xmath33 and the set of unoccupied vertices of this configuration must form a fvs . in other words , each legal configuration @xmath33 corresponds to exactly one fvs . on the other hand , since two or more legal configurations may have the identical set of unoccupied vertices , a single fvs often corresponds to more than one legal configuration . in this sense the above introduced arc constraints are not very restrictive . we can achieve a one - to - one mapping between a legal configuration and a fvs by requiring that if the height of a vertex is positive it must exceed the maximal height of its parent and brother vertices by @xmath6 ( see [ sec : modelb ] ) . but the spin glass model with such a set of more restrictive arc constraints is more complicate for numerical treatment , and it does not lead to better fvs solutions for single digraph instances . in the present paper we stick to the weaker arc constraints . a comparative study on this weaker model and the stronger model of [ sec : modelb ] will be carried out in another technical paper . for the convenience of theoretical development , if there is an arc @xmath10 $ ] between a vertex @xmath8 and another vertex @xmath9 but the reverse arc @xmath15 $ ] is absent , we denote this situation by the notation @xmath39 ; if both the forward arc @xmath10 $ ] and the reverse arc @xmath15 $ ] exist between vertices @xmath8 and @xmath9 , we denote this situation by the notation @xmath40 . in later discussions , @xmath39 and @xmath41 are referred to as _ links _ of digraph @xmath3 . a link factor is defined for each link @xmath39 as @xmath42 where @xmath43 is the kronecker symbol such that @xmath44 if @xmath45 and @xmath46 if @xmath47 , and @xmath48 is the heaviside step function such that @xmath49 for integers @xmath50 and @xmath51 for @xmath52 . the link factor @xmath53 if either @xmath54 ( vertex @xmath9 is unoccupied ) or @xmath55 ( vertex @xmath9 is occupied and its height is higher than that of vertex @xmath8 ) , otherwise @xmath56 . a different link factor is defined for each link @xmath41 as @xmath57 this link factor @xmath58 if either @xmath59 ( vertex @xmath8 is unoccupied ) or @xmath54 ( vertex @xmath9 is unoccupied ) or both , otherwise its value is zero . we now introduce a partition function @xmath60 for the digraph @xmath3 as @xcite @xmath61 \prod\limits_{(i\rightarrow j)\in g } c_{(i\rightarrow j)}(h_i , h_j ) \prod\limits_{(k \leftrightarrow l ) \in g } c_{(k\leftrightarrow l)}(h_k , h_l ) \ ; . \label{eq : zxd1}\ ] ] because of the product term of link factors in the above expression , only legal configurations have positive contributions to the partition function . the parameter @xmath62 of ( [ eq : zxd1 ] ) is a re - weighting parameter ( the inverse temperature ) which favors configurations of larger total weight of occupied vertices . as @xmath62 becomes sufficiently large , the partition function @xmath60 will be contributed predominantly by those height configurations @xmath33 which correspond to minimum feedback vertex sets . given a height configuration @xmath33 , the total number of occupied vertices , @xmath63 , and the total number of occupied arcs , @xmath64 , are computed respectively through @xmath65 \in g } \bigl(1- \delta_{h_i}^0 \bigr ) \bigl(1-\delta_{h_j}^0 \bigr ) \ ; .\ ] ] the free entropy @xmath66 of the spin glass system is defined as @xmath67 for a digraph @xmath3 containing a large number @xmath4 of vertices , we expect the free entropy @xmath66 to be an extensive thermodynamic quantity , namely @xmath68 with @xmath69 being the free entropy density . the free entropy density @xmath70 depends on the the re - weighting parameter @xmath62 but not on the vertex number @xmath4 ( for @xmath4 sufficiently large ) . an interesting connection which we wish to point out is that the directed fvs problem can be viewed as an irreversible opinion propagation problem . consider a simple contagion process in which an initially occupied vertex will decay to be unoccupied if all its parent and brother vertices become unoccupied @xcite . for this process to reach all the vertices of the digraph we need to fix some vertices to be unoccupied as the initial condition . this set of externally fixed vertices must form a fvs . in this opinion - dynamics context the height @xmath29 of a vertex @xmath8 should be understood as the discrete time at which vertex @xmath8 jumps from being occupied to being unoccupied ( see @xcite for related theoretical and computational efforts on irreversible and reversible opinion spreading processes ) . we now study the model ( [ eq : zxd1 ] ) by the rs mean field spin glass theory . this theory is based on the bethe - peierls approximation @xcite , and it can also be derived through the method of partition function expansion @xcite . a quantity of central importance is the probability that a vertex belongs to a fvs ( i.e. , being unoccupied ) . to calculate this marginal probability , let us denote by @xmath71 the probability of vertex @xmath9 taking the height state @xmath72 . if there is a link between two vertices @xmath8 and @xmath9 , we denote by @xmath73 the marginal probability that vertex @xmath9 would take the height state @xmath72 if the effect of vertex @xmath8 to @xmath9 was not considered . consider a randomly chosen vertex @xmath9 of the digraph @xmath3 . in general this vertex is connected to a set of parent , child , and brother vertices ( fig . [ fig : bpapprox ] ) . let us denote by @xmath74 , @xmath75 , and @xmath76 the set of parent , child , and brother vertices of vertex @xmath9 , respectively . in mathematical terms , @xmath77 , @xmath78 , and @xmath79 . the set of neighboring vertices of vertex @xmath9 , denoted as @xmath80 , is the union of these three disjoint sets , namely @xmath81 . to compute the marginal probability @xmath82 , let us first remove vertex @xmath9 and all the attached arcs from digraph @xmath3 . the remaining digraph is referred to as a cavity graph and is denoted as @xmath83 ( fig . [ fig : bpapprox ] ) . as a first approximation we assume that in the cavity digraph @xmath83 all the vertices in the set @xmath80 are mutually independent and therefore the joint distribution of these vertices height states can be written as a product of the marginal height distributions of single vertices ( the bethe - peierls approximation @xcite ) . under this approximation , when vertex @xmath9 is added to the cavity digraph @xmath83 to form the whole digraph @xmath3 , its marginal height distribution @xmath71 can be expressed as @xmath84 where @xmath85 is a shorthand notation for the expression @xmath86 \prod\limits_{k\in c(j ) } \bigl [ q_{k \rightarrow j}^0 + \sum\limits_{h^{\prime\prime } \geq h+1}^{d } q_{k\rightarrow j}^{h^{\prime\prime } } \bigr ] \ ; , \ ] ] and the normalization constant @xmath87 is expressed as @xmath88 \prod\limits_{k\in c(j ) } \bigl [ q_{k \rightarrow j}^0 + \sum\limits_{h^{\prime\prime } \geq h+1}^{d } q_{k\rightarrow j}^{h^{\prime\prime } } \bigr ] \ ; .\ ] ] the neighborhood of a vertex @xmath9 . the whole square region represents the digraph @xmath3 , and the shaded region represents the cavity digraph @xmath83 after vertex @xmath9 and all its attached arcs are removed from @xmath3 . the neighboring vertices of @xmath9 can be divided into three disjoint sets , the parent set @xmath74 , the child set @xmath75 , and the brother set @xmath76 . for instance , vertex @xmath89 , @xmath90 , and @xmath91 . each double - arrowed link in this figure denotes two oppositely directed arcs.,scaledwidth=30.0% ] the free entropy contribution of vertex @xmath9 is defined as @xmath92 \ ; , \ ] ] where @xmath93 denotes the partition function of the cavity digraph @xmath83 . under the bethe - peierls approximation , the explicit expression of @xmath94 is @xmath95 \prod\limits_{k\in c(j ) } \bigl [ q_{k\rightarrow j}^0 + \sum\limits_{h^{\prime\prime } \geq h+1}^{d } q_{k\rightarrow j}^{h^{\prime\prime } } \bigr ] \biggr ) \ ; . \label{eq : phiv}\ ] ] consider a link @xmath39 which indicates a single arc @xmath10 $ ] between vertices @xmath8 and @xmath9 . the free entropy contribution of this link is defined as @xmath96 } \bigr ) } \bigr ] \ ; , \ ] ] where @xmath97}\bigr)$ ] is the partition function of the cavity graph @xmath98}$ ] obtained by removing the single arc @xmath10 $ ] from @xmath3 . its explicit expression under the bethe - peierls approximation is @xmath99 \biggr ) \ ; . \label{eq : phiarc}\ ] ] similarly , the free entropy contribution @xmath100 of a link @xmath101 ( due to the existence of arcs @xmath10 $ ] and @xmath15 $ ] of opposite direction between vertices @xmath8 and @xmath9 ) is expressed as @xmath102 under the bethe - peierls approximation , the whole free entropy @xmath66 of digraph @xmath3 can be evaluated through the following simple expression : @xmath103 the first term of the above expression is the sum of free entropy contributions of all the vertices . since each link @xmath39 contributes to the free entropies of vertices @xmath8 and @xmath9 , the free entropy contribution of this link should be subtracted once from the total free entropy . this explains the second term of ( [ eq : phi ] ) . similarly the free entropy contribution of each link @xmath104 should be subtracted once from the total free entropy , leading to the third term of ( [ eq : phi ] ) . the free entropy density is computed from ( [ eq : phi ] ) as @xmath105 . the entropy density @xmath106 of the system is then calculated through @xmath107 \ ; , \ ] ] where @xmath108 is the relative total weight of the occupied vertices : @xmath109 to actually compute the free entropy density @xmath69 and other thermodynamic quantities , we need to compute the two cavity probability distributions @xmath110 and @xmath111 associated with every link @xmath112 and also the two distributions @xmath111 and @xmath110 associated with every link @xmath113 . under the same bethe - peierls approximation we can write down the following self - consistent equations for two connected vertices @xmath9 and @xmath114 : @xmath115 where @xmath116 and @xmath117 are computed through the following two expressions , respectively : @xmath118 \prod\limits_{k\in c(j)\backslash j^\prime } \bigl [ q_{k \rightarrow j}^0 + \sum\limits_{h^{\prime\prime } \geq h+1}^{d } q_{k\rightarrow j}^{h^{\prime\prime } } \bigr ] \ ; , \\ \hspace*{-2.0 cm } a_{j\rightarrow j^\prime}(h ) = e^{x w_j } \prod\limits_{l\in b(j)\backslash j^\prime } q_{l\rightarrow j}^0 \prod\limits_{i \in p(j)\backslash j^\prime } \bigl [ \sum\limits_{h^\prime=0}^{h-1 } q_{i\rightarrow j}^{h^\prime } \bigr ] \prod\limits_{k\in c(j)\backslash j^\prime } \bigl [ q_{k\rightarrow j}^0 + \sum\limits_{h^{\prime\prime } \geq h+1}^{d } q_{k\rightarrow j}^{h^{\prime\prime } } \bigr ] \ ; .\end{aligned}\ ] ] notice that eq . ( [ eq : qjjpmarginal ] ) holds for all vertices @xmath119 , and it differs from eq . ( [ eq : qjmarginal ] ) only by discarding vertex @xmath114 from the set @xmath80 . equation ( [ eq : qjjpmarginal ] ) is referred to as the belief propagation ( bp ) equation . this equation together with the free entropy expression ( [ eq : phi ] ) forms the rs mean field theory of model ( [ eq : zxd1 ] ) . the bp equation can be solved by iterations . in each time step of this iteration we consider all the vertices consecutively and in a random order , and for each vertex @xmath9 under consideration we update all its output cavity probabilities @xmath120 using ( [ eq : qjjpmarginal ] ) . the difference between the updated cavity probability @xmath121 at time @xmath122 and the old cavity probability @xmath123 is measured by @xmath124 if the maximal value of all the differences @xmath125 is less than certain small threshold @xmath126 then we regard the bp iteration as converging to a fixed point . in this work we set the convergence criterion as @xmath127 . compared with the bp iteration process of the undirected fvs problem @xcite , the bp iteration for the directed fvs problem is more time- and computer memory - consuming , since each height state can have @xmath128 possible values . after the bp equation converges on all the arcs we can then compute the free entropy density @xmath70 and the entropy density @xmath106 . in addition , the mean fraction @xmath129 of vertices in the fvs is equal to the mean fraction of unoccupied vertices : @xmath130 1.0 cm as a simple illustration of the rs mean field theory , let us apply it to a digraph composed of a single directed cycle of @xmath4 vertices and @xmath131 arcs ( in the example of fig . [ fig : cycle : a ] , @xmath132 ) . the bp equation ( [ eq : qjjpmarginal ] ) always converges for this system , and at the bp fixed point all the parent - to - child cavity distributions @xmath133 for arcs @xmath10 $ ] are identical to the same probability function @xmath134 which satisfies @xmath135 the unique solution of this equation is @xmath136 where the constant @xmath0 is determined by @xmath137 . similarly all the child - to - parent cavity distributions @xmath138 on the arcs @xmath10 $ ] are identical to the same probability function @xmath139 with @xmath140 and @xmath141 ( @xmath142 ) . the mean field theory then predicts that the fvs relative size @xmath129 , the free entropy density @xmath70 , and the entropy density @xmath106 are @xmath143 which are all independent of the cycle length @xmath4 . as @xmath144 we have @xmath145 and @xmath146 , see fig . [ fig : cycle : b ] . the minimum fvs size @xmath147 predicted by the rs theory is @xmath148 , which is higher than the true minimum size @xmath6 if @xmath149 and is lower than the true value if @xmath31 is longer than the cycle length @xmath4 . this exactly solvable example clearly demonstrates that the rs mean field theory is only an approximate theory . a major shortcoming of the rs theory is that it neglects all the cycle - caused long range correlations and all the cycle corrections to the free entropy @xmath66 @xcite , even though the constraints in the directed fvs and fas problems are induced by directed cycles . in [ sec : appendix ] we consider another simple random graph example to discuss more about the relationship between predicted minimum fvs relative size ( @xmath147 ) and the parameter @xmath31 . the numerical results shown in fig . [ fig : rrra20 ] demonstrates that the predicted value of @xmath147 decreases considerably with @xmath31 for an infinitely large random digraph . we now perform bp iterations on two types of random digraphs , namely erds - rnyi ( er ) digraphs and regular random ( rr ) digraphs @xcite . we generate a random digraph @xmath3 of @xmath4 vertices and @xmath150 arcs in two steps . first we generate an undirected random simple graph @xmath151 of @xmath4 vertices and @xmath7 undirected edges . then we turn each undirected edge of @xmath151 into a directed edge ( an arc ) by assigning it a direction uniformly at random from the two possible directions . the resulting digraph is then the digraph @xmath3 . an undirected er graph @xmath151 is generated from an empty graph of @xmath4 vertices . two different vertices @xmath8 and @xmath9 are randomly drawn from the whole set of @xmath4 vertices , if these two vertices are not yet connected we draw an edge between them . this edge addition process continues until @xmath7 edges have been created . an undirected rr graph has the property that each vertex is attached by exactly @xmath152 edges , therefore to generate such a graph we first assign @xmath153 @xmath154half - edges to each vertex , and then repeatedly glue two randomly chosen half - edges into a full edge if this full edge is neither a self - connection nor a multi - edge between two vertices . the circular and diamond points of figure [ fig : entropyrspop10 ] are the bp iteration results on a single random er digraph with @xmath155 vertices and arc density @xmath156 . as the re - weighting parameter @xmath62 increases from zero , the relative size @xmath129 of fvs and the entropy density @xmath106 decrease continuously . however we find that the bp iteration process no longer converges if @xmath62 exceeds @xmath157 . the non - convergence of bp iteration indicates the bethe - peierls approximation is no longer a good approximation at @xmath158 for this system ( see also sec . [ sec : cycle ] ) . on the other hand , when bp converges on a single digraph instance , the computed entropy density @xmath106 and fvs relative size @xmath129 are in good agreement with ensemble - averaged results ( see next subsection ) . the rs mean field theory can also be used to calculate ensemble - averaged properties of random digraphs at the thermodynamic limit of @xmath159 . from eq . ( [ eq : phi ] ) we know that to compute the free entropy density @xmath69 we need only to compute the mean value of vertex contribution @xmath94 over all vertices @xmath8 and the mean value of link contribution @xmath160 over all links @xmath39 ( the sum of link contributions @xmath100 can be neglected since there is only a vanishing fraction of links @xmath41 in random er or rr digraphs ) . these mean values and the fvs relative size @xmath129 are easy to compute by population dynamics simulations @xcite . corresponding to eq . ( [ eq : rhoexp ] ) for the fvs relative size @xmath129 of a single digraph , the ensemble - averaged value of @xmath129 is obtained by the following probabilistic equation ( assuming all the vertex weights @xmath161 ) @xmath162 \prod\limits_{k}^{d_{c } } \int \mathcal{d } q_{k\rightarrow j } q^{(c)}[q_{k\rightarrow j } ] \nonumber \\ \quad \quad \quad \quad \times \frac{1}{1 + e^{x } \sum\limits_{h=1}^{d } \prod\limits_{i \in p(j ) } \bigl [ \sum\limits_{h^\prime=0}^{h-1 } q_{i\rightarrow j}^{h^\prime } \bigr ] \prod\limits_{k\in c(j ) } \bigl [ q_{k \rightarrow j}^0 + \sum\limits_{h^{\prime\prime } \geq h+1}^{d } q_{k\rightarrow j}^{h^{\prime\prime } } \bigr ] } \ ; . \label{eq : rhopd}\end{aligned}\ ] ] in this equation , @xmath163 is the probability that a randomly chosen vertex has @xmath164 parent vertices and @xmath165 child vertices . for er digraphs @xmath163 is the product of two poisson distributions with mean value @xmath11 , @xmath166 , while for rr digraphs @xmath163 is the binomial distribution @xmath167 . the probability functional @xmath168 $ ] of eq . ( [ eq : rhopd ] ) is the probability that the parent - to - child message on an arc @xmath25 $ ] is the height distribution function @xmath169 ; similarly , @xmath170 $ ] is the probability that the child - to - parent message on an arc @xmath171 $ ] is the height distribution function @xmath172 . the ensemble - averaged expressions for the vertex free - entropy contribution @xmath173 [ eq . ( [ eq : phiv ] ) ] is very similar to eq . ( [ eq : rhopd ] ) , while the ensemble - averaged expression for the arc free - entropy contribution @xmath174 is computed through @xmath175 \int \mathcal{d } q_{j\rightarrow i } q^{(c)}[q_{j\rightarrow i } ] \ln \biggl ( q_{j\rightarrow i}^0 + \sum\limits_{h_j=1}^{d } q_{j\rightarrow i}^{h_j } \bigl[\sum\limits_{h_i=0}^{h_j-1 } q_{i\rightarrow j}^{h_i } \bigr ] \biggr ) \ ; .\ ] ] in the population dynamics simulations the probability functional @xmath176 $ ] is represented by a large array of @xmath177 cavity probability distributions @xmath178 of the parent - to - child type ( arcs @xmath179 $ ] ) . similarly @xmath180 $ ] is also represented by a large array of @xmath177 cavity probability distributions @xmath178 of the child - to - parent type ( arcs @xmath181 $ ] ) . to update an element of the array @xmath182 $ ] , we first generate an in - degree @xmath183 and an out - degree @xmath184 for the parent vertex @xmath9 according to the cavity degree distribution @xmath185 and we then select @xmath183 parent - to - child probability distributions @xmath169 from the array @xmath176 $ ] and @xmath186 child - to - parent probability distributions @xmath172 from the array @xmath180 $ ] , finally we compute a new probability distribution @xmath178 according to eq . ( [ eq : qjjpmarginal ] ) and replace an old element of the array @xmath176 $ ] with this new element . the array @xmath187 $ ] is updated according to the same procedure , but the degree distribution of the child vertex @xmath9 is changed to be @xmath188 as an example , fig . [ fig : entropyrspop10 ] shows the ensemble - averaged mean field results for er digraphs with arc density @xmath189 . in our calculations the weight of each vertex @xmath8 is set to be @xmath14 for simplicity . the fvs relative size @xmath129 and the entropy density @xmath106 both decrease with the re - weighting parameter @xmath62 , while the entropy density @xmath106 as a function of @xmath129 appears to be concave and its value becomes negative as @xmath129 is lower than certain threshold value @xmath190 which depends slightly on the parameter @xmath31 ( the maximal height ) , e.g. , @xmath191 at @xmath192 and @xmath193 at @xmath32 . 0.2 cm at a given value of @xmath31 we take the value of @xmath147 obtained by the rs population dynamics at entropy density @xmath194 as the predicted minimum relative size of fvs . the relationship between @xmath147 and the arc density @xmath11 obtained at @xmath32 is shown in fig . [ fig : rdfvs ] for er and rr random digraphs . in view of the discussion made in sec . [ sec : cycle ] the predicted @xmath147 should only be considered as an educated guess on the true value of fvs relative size . through bp iteration we can obtain an estimate about the height probability distribution @xmath71 of each vertex @xmath9 , see eq . ( [ eq : qjmarginal ] ) . this estimate might not be very accurate if the underlying bethe - peierls approximation is not a good approximation ( especially when the bp iteration fails to converge ) , however it still contains very useful information for constructing a near - optimal feedback vertex set . we now describe a belief propagation - guided decimation ( bpd ) algorithm for solving the directed fvs problem ( and also the fas problem ) on single digraph instances . initially all the vertices of the digraph @xmath3 are declared as active , and the candidate fvs is initialized as empty , @xmath195 . the bp iteration then runs on this digraph for @xmath196 time steps and the height probability distribution @xmath71 of each vertex @xmath9 is computed by eq . ( [ eq : qjmarginal ] ) . we set @xmath197 in this paper . the bpd algorithm then repeatedly performing the following fixing - and - updating procedure : in the fixing stage , the active vertices @xmath9 are ranked in descending order of their empty probability @xmath198 , and the vertices at the top @xmath199 percent of this ranked sequence are all added to the set @xmath19 and declared as inactive . we set @xmath200 in this paper . the digraph induced by the remaining active vertices is then simplified by repeatedly turning an active vertex into inactive if this vertex has no active parent and brother vertices , or no active child and brother vertices . then in the updating state , we run bp iteration on the simplified digraph of active vertices for @xmath201 time steps ( @xmath202 in this paper ) . then the height probability distributions of all the active vertices are computed by eq . ( [ eq : qjmarginal ] ) again . all the vertices in the original digraph @xmath3 will be turned into inactive by repeating the fixing - and - updating process . the resulting set @xmath19 forms a fvs of @xmath3 . we then further polish this set by examining , in a random order , each vertex of @xmath19 and deleting it from @xmath19 if the reduced set is still a fvs . the performance of the bpd algorithm also depends on the maximal height @xmath31 and the re - weighting parameter @xmath62 . for a digraph with @xmath4 vertices and @xmath7 arcs , the required memory space for running the bpd algorithm is of order @xmath203 while the time complexity is of order @xmath204 . the source code of the bpd algorithm is accessible at the author s webpage ( power.itp.ac.cn/@xmath205zhouhj/codes.html ) . figure [ fig : bpd10 ] shows the mean value of the relative sizes of constructed feedback vertex sets and the mean value of arc densities of the complementary directed acyclic digraphs , obtained by averaging the results of a single run of the bpd algorithm on @xmath206 random er digraphs of arc density @xmath207 and vertex number @xmath208 . when the maximum height @xmath31 is fixed , the performance of bpd improves considerably as @xmath62 increases from @xmath209 to @xmath210 and then saturates as @xmath62 goes beyond @xmath211 . if the re - weighting parameter @xmath62 is fixed but the maximal height changes from @xmath192 to @xmath32 , there is considerable improvement in the performance of bpd . however if @xmath31 is further increased to @xmath212 the additional improvement in performance is much weaker . based on the empirical observations of fig . [ fig : bpd10 ] we set @xmath213 and @xmath32 when applying bpd to all the other random digraph instances mentioned below , even if these digraphs have different arc density @xmath11 . we apply the bpd algorithm to a set of er digraphs with @xmath208 vertices and mean arc density @xmath11 ranging from @xmath6 to @xmath214 . at each value of @xmath11 the fvs mean relative size , denoted as @xmath215 , is computed by averaging over the results of a single run of the bpd algorithm on @xmath206 independent random digraph instances . when @xmath216 we find the value of @xmath215 is very close to the minimum fvs relative size @xmath217 as predicted by the rs mean field theory at @xmath32 , see fig . [ fig : erfvs ] . however , the difference between @xmath215 and @xmath217 becomes noticeable at @xmath218 and the positive gap @xmath219 increases continuously as @xmath11 further increases . we have tested on several digraphs of larger size @xmath155 and found that the bpd results @xmath215 are not sensitive to @xmath4 . at @xmath32 , the discrepancy between the rs mean field results and the bpd algorithmic results therefore is unlikely to be caused by finite - size effects . instead we tend to believe that this gap indicates that the rs mean field prediction @xmath217 is lower than the true minimum fvs size . we expect that the minimum fvs size obtained by the first - step replica - symmetric breaking ( 1rsb ) mean field theory @xcite will be closer to the empirical bpd results , but we have not yet carried out such an investigation . we notice that an elegant 1rsb study has been carried out in the context of the minimal contagious set problem @xcite , which is an inspiration for our future 1rsb work . we also apply the bpd algorithm to a set of rr digraphs with @xmath208 vertices and arc density @xmath11 ranging from @xmath6 to @xmath214 , and compare the results with rs mean field predictions at @xmath32 , see fig . [ fig : rrfvs ] . when @xmath220 the mean fvs relative size @xmath215 obtained by bpd is very close to the minimum fvs relative size @xmath217 of mean field theory , but the difference between @xmath215 and @xmath217 becomes noticeable at @xmath221 and is more and more pronounced as @xmath11 further increases . recently an efficient heuristic local algorithm was proposed in @xcite , which repeatedly refine the height configuration @xmath33 of a digraph through simulated annealing under the constraints ( [ eq : linkc1 ] ) and ( [ eq : linkc2 ] ) . when tested on some small digraph instances with @xmath222 vertices , the sa algorithm outperforms the greedy adaptive searching process ( grasp ) @xcite , one of the most successful local algorithms which repeatedly reduces the number of directed cycles by deleting vertices with highest value of in- and out - degree product @xmath223 . at each elementary step of the sa algorithm a trial is proposed to update the fvs : ( a ) first a randomly chosen vertex @xmath9 of the fvs ( height @xmath54 ) is chosen , and ( b ) its height @xmath72 is either set to a maximal value while satisfying all the out - going arcs @xmath224 $ ] or set to a minimal value while satisfying all the in - coming arcs @xmath10 $ ] , and then ( c ) the parent vertices @xmath8 of all the unsatisfied arcs @xmath25 $ ] or the child vertices @xmath225 of all the unsatisfied arcs @xmath224 $ ] are put to the fvs ( @xmath59 or @xmath226 ) . such a trial is accepted for sure if the resulting fvs size does not increase , otherwise it is accepted with a low probability @xcite . here we compare the performances of bpd and sa on large digraph instances of @xmath208 vertices , see fig . [ fig : rdfvs ] and fig . [ fig : rdag ] . first , we notice that the mean relative sizes of fvs constructed by bpd and by sa are very close to each other both for er and for rr digraphs . ( we have also tested the bpd algorithm on the same set of small digraph instances used in @xcite and found that bpd performs equally good as sa . ) if we assume that the fvs solutions reached by the sa algorithm are close to optimal solutions , the results of fig . [ fig : rdfvs ] then indicate that ( 1 ) the bpd algorithm is able to construct close - to - minimum feedback vertex sets , and that ( 2 ) the rs population dynamics predictions underestimate the minimum fvs size of finite random digraph instances . although the sizes of fvs solutions constructed by bpd and by sa are almost equal , we find that the bpd algorithm is more likely to select a vertex of low connectivity into the fvs than the sa algorithm . as a result the dag obtained by the bpd algorithm has considerably higher arc density ( fig . [ fig : rdag ] ) . in other words , if we choose the bpd algorithm instead of the sa algorithm , we can make the digraph free of directed cycles by deleting much fewer arcs . in this latter sense the bpd algorithm outperforms the sa algorithm ( besides the advantage that bpd is much faster than sa ) . of course both the bpd algorithm and the sa algorithm can be modified by adding another re - weighting parameter to avoid deleting highly connected vertices . such an extension might be necessary for some practical applications . in this paper we introduced a spin glass model ( [ eq : zxd1 ] ) for the directed feedback vertex set problem , approximately solved this model by the replica - symmetric mean field theory of statistical physics , and implemented a belief propagation - guided decimation algorithm to construct nearly optimal feedback vertex sets for single digraph instances . the theory and algorithm of this paper is also applicable to the equivalent feedback arc set problem . the bpd algorithm slightly outperforms the simulated annealing algorithm ( fig . [ fig : rdfvs ] and fig . [ fig : rdag ] ) , it is therefore an efficient algorithm to approach a minimum feedback vertex set for single difficult digraph instances . the rs mean field theory appears to compute a lower - bound for the fvs minimum relative size @xmath129 ( fig . [ fig : rdfvs ] ) . this theory assumes that the height states of all the neighbors of a vertex @xmath9 are independent of each other after vertex @xmath9 is deleted ( fig . [ fig : bpapprox ] ) . this approximation essentially ignores all the correlations propagated along the directed cycles ( e.g. , in fig . [ fig : cycle : a ] the heights @xmath29 and @xmath227 of vertices @xmath8 and @xmath225 must be strongly correlated at low temperatures even if vertex @xmath9 is removed ) . conceptually speaking , the rs mean field theory of sec . [ sec : rs ] can not be a satisfactory approach to tackle cycle constraints . much theoretical efforts ( including 1rsb mean field computations ) are needed to achieve a deeper understanding on the directed fvs problem . an especially interesting but challenging question would be to design a spin glass model with truly local interactions for the cycle constraints ( like the edge - constrained model for the undirected fvs problem @xcite ) , without using height as the vertex state . the author thanks yang - yu liu , shao - meng qin , chuang wang and jin - hua zhao for helpful discussions , and the school of physics of northeastern normal university for hospitality during his two visits in august 2013 and august 2014 . this work was supported by the national basic research program of china ( grant number 2013cb932804 ) , by the national natural science foundation of china ( grant numbers 11121403 and 11225526 ) , and by the knowledge innovation program of chinese academy of sciences ( no . kjcx2-ew - j02 ) . in this appendix we describe a more restrictive spin glass model for the directed feedback vertex set problem . the model in the main text can be regarded as a relaxed version of this model . for simplicity we assume that there is no bi - directional edges in the digraph @xmath3 so that the set of brother vertices of each vertex @xmath9 is empty : @xmath228 . the partition function of this new model is defined as @xmath229 \ ; , \ ] ] where @xmath230 returns the maximal height value among all the parent vertices of @xmath9 . a height configuration @xmath231 with @xmath232 zero elements will contribute a term @xmath233 to this partition function if , for every vertex @xmath9 , the height state @xmath72 is either zero ( vertex @xmath9 being unoccupied ) or @xmath72 is exceeding the maximal height of its parent vertices by one . such a height configuration is called a legal configuration , and all other height configurations are illegal and have no contribution to @xmath234 . there is a one - to - one correspondence between a legal configuration @xmath33 of model ( [ eq : modelb ] ) and a fvs of the digraph @xmath3 . first , it is obvious that the set formed by all the zero - height vertices of a legal configuration @xmath33 must be a fvs . second , given any feedback vertex set @xmath19 of the digraph we can construct a unique legal height configuration by recursion : first assign all vertices in @xmath19 the height value @xmath209 and put them into an initially empty vertex set @xmath235 ; then assign the height value @xmath6 to all the vertices whose parent vertices are completely contained in set @xmath235 and then add these newly assigned vertices to set @xmath235 ; then repeat this process and assign the remaining vertices the height values @xmath157 , @xmath236 , @xmath237 , until all the vertices have been exhausted . the resulting height configuration must be a legal configuration . the spin glass model ( [ eq : modelb ] ) is more restrictive than the model ( [ eq : zxd1 ] ) . this is due to the fact that in the new model each vertex @xmath9 causes a many - body constraint among @xmath9 and all its parent vertices . for convenience of discussion let us denote by @xmath238 the constraint caused by vertex @xmath9 . to solve this more difficult model by the replica - symmetric mean field theory , we denote by @xmath239 the probability that vertex @xmath225 will be in height state @xmath227 if it is only constrained by constraint @xmath240 . similarly , for each parent vertex @xmath9 of vertex @xmath225 , we denote by @xmath241 the probability that @xmath9 will be in height state @xmath72 if it is only constrained by the constraint @xmath240 . we can write down the following set of belief - propagation equations for these two height distributions : @xmath242 \biggr\ } \ ; , \label{eq : bpmany1 } \\ & \hspace*{-2.2 cm } p_{\boxed{\scriptstyle{k}}\rightarrow j}^{h_j } = \frac{1}{z_{\boxed{\scriptstyle{k}}\rightarrow j } } \biggl\ { e^{-x } q_{k\rightarrow \boxed{\scriptstyle{k}}}^0 + q_{k\rightarrow \boxed{\scriptstyle{k}}}^{h_j+1 } \prod\limits_{i\in p(k)\backslash j } q_{i\rightarrow \boxed{\scriptstyle{k}}}^{h_j } + \nonumber \\ & \hspace*{2.0 cm } \sum\limits_{h_k\geq h_j+2 } q_{k\rightarrow \boxed{\scriptstyle{k}}}^{h_k } \bigl [ \prod\limits_{i\in p(k)\backslash j } q_{i\rightarrow \boxed{\scriptstyle{k}}}^{h_k-1 } -\prod\limits_{i\in p(k)\backslash j } q_{i\rightarrow \boxed{\scriptstyle{k}}}^{h_k-2 } \bigr ] \biggr\ } \ ; . \label{eq : bpmany2 } \end{aligned}\ ] ] in the above equations @xmath243 and @xmath244 are two probability normalization constants ; the quantity @xmath245 denotes the probability that vertex @xmath225 will be in height state @xmath227 if it is _ not _ constrained by constraint @xmath240 ; the quantity @xmath246 is a partial sum defined as @xmath247 where @xmath248 denotes the probability that the parent vertex @xmath9 of vertex @xmath225 will be in height state @xmath72 if the constraint @xmath240 is absent . the self - consistent bp equations for @xmath245 and @xmath248 are much simplier : @xmath249 where again @xmath250 and @xmath251 are two probability normalization constants . the marginal probability @xmath252 that vertex @xmath9 will be in height state @xmath72 in the digraph @xmath3 is then evaluated at a fixed - point of the bp iteration as @xmath253 with @xmath87 being a normalization constant . the mean fraction of unoccupied vertices ( i.e. , the fvs relative size ) @xmath129 is computed as @xmath254 the bp equations ( [ eq : bpmany1])([eq : bpmany5 ] ) are rather slow to iterate . our preliminary numerical results indicated that the feedback vertex set solutions offered by this more complicated bp scheme are not better than the solutions obtained by the relaxed bp scheme of the main text . in this work we therefore give up further exploration of the model ( [ eq : modelb ] ) of many - body interactions . to be complete , here we also list the explicit expression for the total free entropy @xmath66 of the new model : @xmath255 \,\ ] ] where @xmath256 is the number of child vertices of vertex @xmath225 ( the out - degree ) . in this expression , @xmath257 is the free entropy contribution of constraint @xmath240 and all its involved vertices , and @xmath258 is the free entropy contribution of vertex @xmath225 . their respective expressions are @xmath259 \nonumber \\ & \hspace{2.0 cm } + \prod\limits_{l\in c(k ) } p_{\boxed{\scriptstyle{l}}\rightarrow k}^1 \prod\limits_{j\in p(k)}\bigl [ p_{\boxed{\scriptstyle{j}}\rightarrow j}^{0 } \prod\limits_{m\in c(j)\backslash k } p_{\boxed{\scriptstyle{m}}\rightarrow j}^{0 } \bigr ] \nonumber \\ & \hspace{2.0 cm } + \sum\limits_{h_k\geq 2 } \prod\limits_{l\in c(k ) } p_{\boxed{\scriptstyle{k}}\rightarrow k}^{h_k } \biggl ( \prod\limits_{j\in p(k)}\bigl [ \sum\limits_{h_j=0}^{h_k-1 } p_{\boxed{\scriptstyle{j}}\rightarrow j}^{h_j } \prod\limits_{m\in c(j)\backslash k } p_{\boxed{\scriptstyle{m}}\rightarrow j}^{h_j } \bigr ] \nonumber \\ & \hspace{4.0 cm } - \prod\limits_{j\in p(k)}\bigl [ \sum\limits_{h_j=0}^{h_k-2 } p_{\boxed{\scriptstyle{j}}\rightarrow j}^{h_j } \prod\limits_{m\in c(j)\backslash k } p_{\boxed{\scriptstyle{m}}\rightarrow j}^{h_j } \bigr ] \biggr ) \biggr\ } \ ; , \\ & \hspace*{-1.0 cm } \phi_{k } = -\frac{1}{x } \ln \biggl\ { \sum\limits_{h_k } p_{\boxed{\scriptstyle{k}}\rightarrow k}^{h_k } \prod\limits_{l\in c(k ) } p_{\boxed{\scriptstyle{l}}\rightarrow k}^{h_k } \biggr\ } \ ; . \end{aligned}\ ] ] in sec . [ sec : cycle ] we studied a single directed cycle and found that the rs mean field predicted that the relative size @xmath147 of minimum fvs approaches zero as the maximal height parameter @xmath260 . in this appendix we study another exactly solvable digraph and show that for this system the value of @xmath147 may converge to a strictly positive value at @xmath261 . the digraph we now consider is a completely random digraph with the constraint that each vertex has exactly @xmath11 parent vertices and @xmath11 child vertices ( i.e. , the in - degree and out - degree of every vertex is an integer @xmath11 ) . the weight of every arc in this digraph is set to be unity . for such a random digraph the bp equation ( [ eq : qjjpmarginal ] ) has a fixed point with all the parent - to - child cavity distributions @xmath262 for arcs @xmath10 $ ] being identical to the same probability function @xmath134 and all the child - to - parent cavity distributions @xmath263 being identical to the same probability function @xmath139 . furthermore we find that @xmath264 and @xmath265 for @xmath266 . following the general bp equation ( [ eq : qjjpmarginal ] ) , we obtain the self - consistent condition for the function @xmath134 as @xmath267^{\alpha } \bigl [ \sum\limits_{h^{\prime \prime}=0}^{d-\tilde{h } } q_{p c}^{h^{\prime \prime } } \bigr]^{\alpha -1 } } \ ; , \label{eq : qpc0rrr } \\ q_{p c}^h & = & \frac{e^x \bigl [ \sum\limits_{h^\prime=0}^{h-1 } q_{p c}^{h^\prime } \bigr]^{\alpha } \bigl [ \sum\limits_{h^{\prime \prime}=0}^{d - h } q_{p c}^{h^{\prime \prime } } \bigr]^{\alpha -1 } } { 1 + e^x \sum\limits_{\tilde{h}=1}^{d } \bigl [ \sum\limits_{h^\prime=0}^{\tilde{h}-1 } q_{p c}^{h^\prime } \bigr]^{\alpha } \bigl [ \sum\limits_{h^{\prime \prime}=0}^{d-\tilde{h } } q_{p c}^{h^{\prime \prime } } \bigr]^{\alpha -1 } } \quad \quad \quad ( h=1 , 2 , \ldots , d ) \ ; . \label{eq : qpchrrr } \end{aligned}\ ] ] these two equations have only a unique solution for any fixed values of the parameters @xmath62 and @xmath31 . we determine the value of @xmath268 through the following numerical procedure : ( 1 ) set @xmath269 to a value @xmath270 $ ] ; ( 2 ) determine the value @xmath271 by solving @xmath272 , and then determine @xmath273 according to @xmath274 ; ( 3 ) then determine the value of @xmath275 by solving @xmath276 and then determine @xmath277 according to @xmath278 ; ( 4 ) then continue to determine all the remaining probability items @xmath268 in the same sequential maner as steps ( 2 ) and ( 3 ) ; ( 5 ) then compute a new value of @xmath269 according to eq . ( [ eq : qpc0rrr ] ) ; ( 6 ) if @xmath269 is different from the input value @xmath279 we then change @xmath279 appropriately and repeat steps ( 1)(5 ) until convergence is reached . at the fixed point of eqs . ( [ eq : qpc0rrr ] ) and ( [ eq : qpchrrr ] ) we then compute the relative size @xmath129 of fvs is @xmath280^{\alpha } \bigl [ \sum\limits_{h^{\prime \prime}=0}^{d-\tilde{h } } q_{p c}^{h^{\prime \prime } } \bigr]^{\alpha -1 } } \ ; .\ ] ] the free entropy density @xmath70 is @xmath281^{\alpha } \bigl [ \sum\limits_{h^{\prime \prime}=0}^{d-\tilde{h } } q_{p c}^{h^{\prime \prime } } \bigr]^{\alpha -1 } \biggr ] \nonumber \\ & & - \frac{(\alpha - 1)}{x } \ln \biggl [ 1 + e^x \sum\limits_{\tilde{h}=1}^{d } \bigl [ \sum\limits_{h^\prime=0}^{\tilde{h}-1 } q_{p c}^{h^\prime } \bigr]^{\alpha } \bigl [ \sum\limits_{h^{\prime \prime}=0}^{d-\tilde{h } } q_{p c}^{h^{\prime \prime } } \bigr]^{\alpha } \biggr ] \ ; . \end{aligned}\ ] ] the entropy density is then simply computed as @xmath282 . at a given value of maximal height @xmath31 we can obtain the function @xmath283 by performing numerical calculations at different values of @xmath62 . the minimum relative size @xmath147 at this value of @xmath31 is then determined by solving @xmath284 . the plus symbols of fig . [ fig : rrra20 ] are the predicted values of @xmath147 up to @xmath285 for the particular case of arc density @xmath286 . qualitatively the same mean field results are obtained for other values of the arc density @xmath11 . the predicted minimum value @xmath147 of fvs relative size for the special regular random digraph in which every vertex is connected to @xmath214 parent vertices and @xmath214 child vertices . the value of @xmath147 depends on the maximal height @xmath31 used in the rs mean field theory . the three fitting curves are @xmath287 ( long - dashed blue line ) , @xmath288 ( dashed blue line ) , and @xmath289 ( dotted red line ) . , scaledwidth=55.0% ] figure [ fig : rrra20 ] demonstrates that @xmath147 decreases slowly with maximal height @xmath31 . we can fit the behaviour of @xmath147 as @xmath290 the fitting parameter @xmath291 is then the predicted minimum fvs relative size for an infinite random digraph . the fitted value of @xmath291 is sensitive to the range of @xmath31 values included in the fitting . for example , if we only use the data points of @xmath292 , we obtain @xmath293 , see the green long - dashed line of fig . [ fig : rrra20 ] ; if we use the data points of @xmath294 we obtain a value @xmath295 , see the blue dashed line of fig . [ fig : rrra20 ] . it s not easy for us to compute @xmath147 for @xmath296 , but we anticipate that if more @xmath147 data obtained at very large @xmath31 values are included in the fitting , the fitted @xmath291 will further decrease . on the other hand , we observe that if we fix @xmath297 the fitting can not capture the asymptotic behaviour of @xmath147 at @xmath298 ( see dotted red curve of fig . [ fig : rrra20 ] ) . therefore we believe @xmath291 should be strictly positive . 10 a. mochizuki , b. fiedler , g. kurosawa , and d. saito . dynamics and control at feedback vertex sets . ii : a faithful monitor to determine the diversity of molecular activities in regulatory networks . , 335:130146 , 2013 . p. festa , p. m. pardalos , and m. g. c. resende . feedback set problems . in d du and p. m. pardalos , editors , _ handbook of combinatorial optimization _ , pages 209258 . springer , berlin , germany , 1999 . c .- l . chang and y .- d . lyuu . triggering cascades on strongly connected directed graphs . in _ proceedings of the fifth international symposium on parallel architectures , algorithms and programming ( paap ) _ , pages 9599 . ieee , 2012 .

the chromosphere of the quiet sun is a very important stellar atmospheric region whose thermal and magnetic structure we need to decipher for unlocking new discoveries in solar and stellar physics . to this end , we need to identify and exploit observables sensitive to weak magnetic fields ( g ) and to the presence of cool and hot gas in the bulk of the solar chromosphere . here we report on an investigation of the hanle effect in two semi - empirical models of the quiet solar atmosphere with different chromospheric thermal structures . our study reveals that the linear polarization profiles produced by scattering in the ca ii ir triplet have thermal and magnetic sensitivities potentially of great diagnostic value . the linear polarization in the 8498 line shows a strong sensitivity to inclined magnetic fields with strengths between 0.001 and 10 g , while the emergent linear polarization in the 8542 and 8662 lines is mainly sensitive to magnetic fields with strengths between 0.001 and 0.1 g. the reason for this is that the scattering polarization of the 8542 and 8662 lines , unlike the 8498 line , is controlled mainly by the hanle effect in their ( metastable ) lower levels . therefore , in regions with magnetic strengths sensibly larger than 1 g , their stokes and profiles are sensitive only to the orientation of the magnetic field vector . we also find that for given magnetic field configurations the sign of the and profiles of the 8542 and 8662 lines is the same in both atmospheric models , while the sign of the linear polarization profile of the 8498 line turns out to be very sensitive to the thermal structure of the lower chromosphere . we suggest that spectropolarimetric observations providing information on the relative scattering polarization amplitudes of the ca ii ir triplet will be very useful to improve our empirical understanding of the thermal and magnetic structure of the quiet chromosphere .

1007.0727

the chromosphere of the quiet sun is one of the most complex stellar atmospheric regions ( e.g. , harvey 2006 , 2009 ; judge 2006 , 2009 ) . lying between the thin photosphere and the extended @xmath5 k corona , it is here where the magnetic field becomes the globally dominating factor ruling the outer solar atmosphere . measuring the magnetic field vector in the solar chromospheric plasma is , however , notoriously difficult , especially outside sunspots and related active regions . quantitative information can be obtained through spectropolarimetry , but the measurement and physical interpretation of the weak polarization signatures that the hanle and zeeman effects produce in the few chromospheric lines that can be observed from ground - based telescopes is not an easy task ( see trujillo bueno 2010 , for a recent review ) . it is thus important to develop novel methods of measuring " the chromospheric magnetic field , ideally based on the action of the hanle and zeeman effects in spectral lines whose line - center intensity images reveal the fine - scale thermodynamic structuring of the solar chromosphere . here we show that the scattering polarization observed in the ir triplet of ca ii contains valuable information on the thermal and magnetic structure of the quiet " chromosphere , from the shock dominated region of the lower chromosphere " to the upper chromosphere " below the interface region to the @xmath5 k solar corona . the only way to obtain direct empirical information on the intensity and topology of the magnetic fields of the solar chromosphere is via the measurement and interpretation of polarization signals in chromospheric spectral lines . in regions with high concentrations of magnetic flux , such as in sunspots and plages , the polarization patterns are dominated by the zeeman effect . hence , diagnostic techniques based on this effect are quite useful ( e.g. , socas - navarro , trujillo bueno & ruiz cobo 2000 ; socas - navarro 2005 ) . however , in the quiet sun , which covers most of the solar disk at any given time during the solar magnetic activity cycle , the zeeman splitting between the @xmath6 ( @xmath7 ) and @xmath8 ( @xmath9 ) components ( @xmath10 being the magnetic quantum number ) is only a small fraction of the width of the spectral lines formed in the chromosphere . as a result , the transverse zeeman effect makes a very small or insignificant contribution to the emergent linear polarization amplitudes . on the other hand , the circular polarization of the zeeman effect as a diagnostic tool is of limited practical interest for the exploration of the magnetism of the upper chromosphere because the response function of the emergent stokes @xmath11 profiles of strong lines like h@xmath12 and ca ii 8542 is significant only in the photosphere and/or lower chromosphere ( socas - navarro & uitenbroek 2004 ; uitenbroek 2006 ) . there is yet another mechanism producing linear polarization in the spectral lines that can be exploited to diagnose the solar chromosphere . atoms align due to the optical pumping caused by the incident anisotropic radiation field ; i.e. , the individual magnetic @xmath10 sublevels of energy levels with total angular momentum @xmath13 are unevenly populated ( in such a way that the populations of substates with different values of @xmath14 are unequal ) and coherences between them may appear . this , in turn , gives rise to a sizable amount of linear polarization in chromospheric spectral lines , either by selective emission or absorption of polarization components in the spectral line . a remarkable example is provided by the ir triplet of ca ii . its linear polarization pattern was considered `` enigmatic '' for a while ( stenflo , keller & gandorfer 2000 ) because it was impossible to achieve even a qualitative fit to the observed relative polarization amplitudes considering that scattering line polarization results only from selective emission processes caused by the population imbalances of the line s upper level . it was later realized that selective absorption of polarization components caused by the presence of populations imbalances in the line s lower level is a very efficient mechanism for producing linear polarization in the spectral lines of a stellar atmosphere ( trujillo bueno & landi deglinnocenti 1997 ; trujillo bueno 1999 ; trujillo bueno et al . in fact , manso sainz & trujillo bueno ( 2003 ) demonstrated quantitatively that the physical origin of the linear polarization pattern observed in the ca ii lines at 8542 and 8662 is indeed zero - field " dichroism , i.e. , selective absorption of polarization components caused by the presence of atomic alignment in the metastable lower levels @xmath15 and @xmath16 ( see fig . 1 ) , while that observed in the 8498 line is controled by both dichroism and by the selective emission of polarization components that result from the population imbalances between the sublevels of its upper level @xmath17 . the alignment of the atomic @xmath18-levels and the ensuing emergent linear polarization is modified by the presence of a magnetic field through the hanle effect , even if the magnetic field turns out to have a complex , tangled topology , with mixed polarities at subresolution scales ( e.g. , landi deglinnocenti & landolfi 2004 ) . in the absence of collisional depolarization the magnetic field strength @xmath19 ( in g ) that is sufficient to produce a significant change is @xmath20 where @xmath21 and @xmath22 are , respectively , the lifetime ( in seconds ) and the land factor of the @xmath18 level under consideration . interestingly enough , since the lifetimes of the metastable lower levels of the ca ii ir triplet are about two orders of magnitude larger than the upper level lifetimes and levels with @xmath23 or @xmath24 can not carry atomic alignment , the linear polarization of the ca ii line at 8662 ( whose upper level has @xmath25 ) is expected to be sensitive to mg fields only , while the scattering polarization of the 8498 and 8542 lines ( which share the same upper level whose @xmath26 ) can in principle be sensitive also to magnetic fields in the gauss range . in the quiet solar chromosphere depolarization of the ca ii atomic levels by elastic collisions with neutral hydrogen atoms is insignificant for the upper levels of the h and k lines , while such collisions have a rather small impact on the atomic polarization of the metastable lower levels ( manso sainz & trujillo bueno 2003 ; derouich , trujillo bueno & manso sainz 2007 ) . actually , the mere fact that the ca ii 8662 line shows a clear @xmath3 polarization signal ( see stenflo et al . 2000 ) implies that its lower level ( whose @xmath27 ) is significantly polarized in the quiet solar chromosphere ( manso sainz & trujillo bueno 2003 ) . the aim of this paper is to investigate the thermal and magnetic sensitivity of the scattering polarization @xmath3 and @xmath4 profiles of the ca ii ir triplet . after formulating in 2 the multilevel problem of scattering polarization and the hanle effect , 4 discusses the unmagnetized case in the hot " and cool " chromospheric models described in 3 . the impact of the hanle effect on the linear polarization profiles of the ca ii ir triplet is discussed in 5 , showing for each solar atmospheric model detailed results for the cases of disk center ( 5.1 ) and close to the limb observations ( 5.2 ) . the ensuing information is summarized in 5.3 in the form of hanle effect diagrams for each of the spectral lines of the ir triplet . finally , 6 summarizes our main conclusions with an outlook to future research . the relevant radiative transfer problem here implies calculating , at each spatial point of the ( generally magnetized ) astrophysical plasma model under consideration , the values of the diagonal and non - diagonal elements of the atomic density matrix corresponding to each atomic level of total angular momentum @xmath18 , which quantify its overall population , the population imbalances between its magnetic sublevels , and the quantum coherences between each pair of them . the values of such density - matrix elements have to be consistent with the intensity , polarization and symmetry properties of the radiation field generated within the medium . this requires solving jointly the rt equations for the stokes parameters and the statistical equilibrium equations for the density - matrix elements . once such a self - consistent solution is obtained it is straightforward to compute the emergent stokes profiles via the formal solution of the stokes - vector transfer equation for any desired line of sight ( los ) . the aim of this section is to formulate such equations with emphasis on the case of the ir triplet of ca ii . the excitation state of the atomic system is described through the density matrix ( fano 1957 ) . we consider the populations ( @xmath28 ) , and coherences ( @xmath29 ) among the @xmath30 magnetic sublevels @xmath10 of each given atomic level with total angular momentum @xmath18 . alternatively , we use the multipolar components of the atomic density matrix ( omont 1977 ) : @xmath31 where @xmath32 , @xmath33 , and the expression in brackets is the wigner 3j - symbol ( e.g. , brink & satchler 1968 ) . the @xmath34 elements with @xmath35 are real numbers measuring population imbalances between sublevels , while @xmath36 is the total population of the level . the @xmath37 elements are combinations of the complex non - diagonal components @xmath29 . hermiticity of the density matrix ( @xmath38 ) implies @xmath39^*$ ] ( @xmath40 meaning complex conjugate ) . in the absence of net circular polarization in the incident radiation @xmath41 , i.e. , atoms _ align _ but they do not _ orientate_. hence , odd-@xmath42 @xmath43 elements vanish . therefore , the 5-level atomic model of ca ii considered here ( fig . 1 ) requires @xmath44 @xmath43 elements to fully characterize the excitation state of the atomic system . the transfer of spectral line polarization is described by the following equations for the stokes parameters [ rte ] @xmath45 \\ \frac{\rm d}{{\rm d}s}q&=\epsilon_q-\eta_qi-\eta_iq-\rho_vu+\rho_uv , \label{rte2 } \displaybreak[0 ] \\ \frac{\rm d}{{\rm d}s}u&=\epsilon_u-\eta_ui+\rho_vq-\eta_iu-\rho_qv , \label{rte3 } \displaybreak[0 ] \\ \frac{\rm d}{{\rm d}s}v&=\epsilon_v-\eta_vi-\rho_uq+\rho_qu-\eta_iv.\label{rte4}\end{aligned}\ ] ] in the hanle effect regime the magnetic splitting of @xmath46 and @xmath6 components of the line transitions are negligible and the spectral line emissivity thus depends exclusively on the excitation state of the upper level of the transition ( i.e. , on the @xmath47 values ) . in particular , linearly polarized emission is due to the presence of alignment and coherences between atomic sublevels of the upper level of the transition . moreover , since levels are not oriented , @xmath48 . more explicitly ( landi deglinnocenti & landolfi 2004 ; see also manso sainz & trujillo bueno 1999 ) : [ epsilons ] @xmath49 - \sin \chi { \rm im}[{{{\rho}}^2_1 } ] ) \\ + \frac{\sqrt{3}}{2 } ( 1-\mu^2 ) ( \cos 2\chi \ , { \rm re}[{{{\rho}}^2_2}]-\sin 2\chi \ , { \rm im}[{{{\rho}}^2_2 } ] ) \big{\}},\end{gathered}\ ] ] @xmath50 - \sin \chi { \rm im}[{{\rho}}^2_1 ] ) \\ - \frac{\sqrt{3}}{2 } ( 1+\mu^2 ) ( \cos 2\chi \ , { \rm re}[{{\rho}}^2_2]-\sin 2\chi \ , { \rm im}[{{\rho}}^2_2 ] ) \big{\}},\end{gathered}\ ] ] @xmath51+\cos \chi { \rm im}[{{\rho}}^2_1 ] ) + \mu ( \sin 2\chi \ , { \rm re}[{{\rho}}^2_2 ] + \cos 2\chi \ , { \rm im}[{{\rho}}^2_2 ] ) \big{\}},\ ] ] where @xmath52 ( with @xmath53 the total number of atoms per unit volume , @xmath54 the einstein coefficient for spontaneous emission and @xmath55 the voigt profile , with @xmath56 and @xmath57 the transition s frequency ) , @xmath58 and @xmath59 are the inclination and azimuth of the ray , and @xmath60 is a numerical coefficient that depends only on @xmath61 and @xmath62 in particular , @xmath63 , @xmath64 , and @xmath65 ( see table 10.1 in landi deglinnocenti & landolfi 2004 ) . the absorption coefficient @xmath66 , and the dichroism coefficients @xmath67 and @xmath68 are given by analogous expressions but with @xmath69 instead of @xmath70 , @xmath71 instead of @xmath60 ( @xmath72 , @xmath73 ) , and with the @xmath74 values of the _ lower _ level of the line transition instead of those of the upper level ( we neglect stimulated emissions ) . likewise , since we are assuming that the atomic levels are not oriented , @xmath75 ( when neglecting the wavelength shifts between the @xmath46 components caused by the zeeman effect ) . finally , the anomalous dispersion terms ( @xmath76 ) have the same functional dependence on the @xmath74 values of the _ lower _ level , but with an antisymmetric dispersive dependence on frequency , i.e. , @xmath77 , with @xmath78 the dispersion profile . additionally , we consider an unpolarized background continuum providing an extra absorption coefficient @xmath79 and emissivity @xmath80 ( where @xmath81 is the planck function ) , a very good approximation toward the red part of the solar spectrum ( e.g. , gandorfer 2000 ) . therefore , in eqs . ( 3 ) , @xmath82 , @xmath83 , while all other emission , dichroism , and anomalous dispersion coefficients have only line contributions ( @xmath84 , @xmath85 , @xmath86 , @xmath87 , @xmath88 , @xmath89 ) . note that @xmath90 and @xmath91 depend on both the population imbalances ( @xmath92 ) and on the coherences ( @xmath93 , with @xmath94 ) , while @xmath95 and @xmath96 depend _ only _ on the quantum coherences is the _ perpendicular _ to the plane formed by the propagation direction and the quantization z - axis of total angular momentum ] . the @xmath97 elements produce the dominant contribution to the stokes @xmath98 profile , while the @xmath93 elements ( the _ alignment _ components ) contribute to the _ linear _ polarization signals , which we quantify by the stokes parameters @xmath1 and @xmath2 . although @xmath99 ( because , as mentioned above , @xmath100 ) according to eq . ( 3d ) it is possible to generate stokes @xmath11 in the spectral line due to the @xmath101 ( second - order ) terms . this contribution is however negligible compared to the longitudinal zeeman effect signal produced by the splitting of the @xmath46-components of the emission profile which we do not consider here . since the degree of anisotropy of the radiation field in solar - like atmospheres is weak , the population imbalances and coherences of each @xmath18-level are only a small fraction of its overall population . therefore , second - order terms can be safely neglected in the stokes - vector transfer equation and the relevant equations for calculating @xmath3 and @xmath4 are @xmath102 -\eta_i q , \\ \frac{\rm d}{{\rm d}s}u & \approx [ \epsilon_u-\eta_ui ] -\eta_i u. \end{aligned}\ ] ] nevertheless , all the calculations of this paper were carried out by solving numerically the full stokes vector transfer equations ( 3 ) through the application of the formal solver mentioned in appendix a. the complication with respect to the unmagnetized case considered by manso sainz & trujillo bueno ( 2003 ) is that now the presence of a magnetic field of given strength @xmath19 , inclination @xmath103 and azimuth @xmath104 at each grid point of the stellar atmosphere model under consideration implies that we have to consider also the quantum coherences ( @xmath93 , with @xmath105 ) between each pair of magnetic substates pertaining to each @xmath18 level , in addition to their individual populations ( @xmath92 ) . for this reason , not only stokes @xmath1 but also stokes @xmath2 is now non - zero , in general . the expressions above for the components of the emission vector and propagation matrix ( eqs . 4 and paragraphs following them ) simplify considerably for several cases of practical interest . for a one - dimensional atmosphere , either unmagnetized , or permeated by a _ microturbulent _ and _ isotropically distributed _ magnetic field of strength @xmath19 , or in the presence of a deterministic magnetic field with a fixed inclination but with a random azimuth below the line s photon mean - free - path , or in the presence of a deterministic magnetic field with a fixed orientation but with a strength in the saturation regime of the hanle effect , the quantum coherences ( i.e. the real and imaginary parts of the @xmath106 and @xmath107 components ) vanish in a reference system whose quantization axis must be the parallel to the magnetic field vector for the deterministic field case and the the normal to the stellar surface for the three other cases : @xmath108},\ ] ] @xmath109 @xmath110 where the @xmath34 ( @xmath111 ) values are those of the _ upper _ level of the line transition under consideration , and where the orientation of the ray is specified by the azimuthal angle @xmath59 and by @xmath112 ( with @xmath113 the angle between the direction of the radiation beam and the quantization axis , which must be the magnetic field direction itself for the deterministic field case and the normal to the surface of the stellar atmospheric model for the two other cases ) . likewise , the only non - zero elements of the propagation matrix are @xmath114 and @xmath91 which are given by identical expressions ( i.e. by @xmath115 and @xmath116 ) , but with @xmath117 instead of @xmath70 , @xmath71 instead of @xmath60 and with the @xmath34 values of the _ lower _ level of the line transition instead of those of the upper level ( for the case in which stimulated emissions are neglected ) . under such assumptions , and taking into account the low polarization level in solar - like atmospheres ( i.e. , that @xmath118 and @xmath119 ) , an approximate formula can be derived to estimate the emergent fractional linear polarization amplitude at the center of a strong spectral line ( trujillo bueno 2003 ) : @xmath120,\ ] ] where @xmath121 must be evaluated at height in the model atmosphere where the line - center optical distance along the line of sight is unity . its first term is due to _ selective emission _ of polarization components ( caused by the population imbalances of the upper level ) , while its second term accounts for `` zero - field '' dichroism . note that since @xmath63 and @xmath72 the scattering polarization of @xmath122 is exclusively due to `` zero - field '' dichroism . general equations for the multivel atom model which neglects quantum interferences between the sublevels pertaining to different @xmath18-levels are derived in landi deglinnocenti & landolfi ( 2004 ) . for the particular case of the 5-level atom of fig . 1 the following rate equations for the density matrix elements follow : @xmath123 \rho^0_0(1 ) + a_{41}\rho^0_0(4 ) + \sqrt{2 } a_{51}\rho^0_0(5 ) + \sum_{i\neq 1 } c_{i1 } \sqrt{\frac{2j_i+1}{2 } } \rho^0_0(i ) , \end{split } \displaybreak[0 ] \\ \begin{split}\label{see02 } \frac{\rm d}{{\rm d}t } \rho^0_0(2 ) = & -\biggl[\sum_{u=4}^5 b_{2u}j^0_0(2\rightarrow u ) + \sum_{i\neq 2 } c_{2i } \biggr ] \rho^0_0(2 ) - \sum_{q'=-2}^2 \biggl ( \frac{1}{\sqrt{2 } } b_{24}j^2_{q'}(2\rightarrow 4 ) -\frac{2\sqrt{2}}{5 } b_{25}j^2_{q'}(2\rightarrow 5 ) \biggr ) \rho^2_{q'}(2 ) \\ & + \frac{1}{\sqrt{2 } } a_{42}\rho^0_0(4 ) + a_{52}\rho^0_0(5 ) + \sum_{i\neq 2 } c_{i2 } \frac{\sqrt{2j_i+1}}{2 } \rho^0_0(i ) , \end{split } \displaybreak[0 ] \\ \begin{split}\label{see03 } \frac{\rm d}{{\rm d}t } \rho^2_q(2 ) = & -\biggl [ 2\pi{\rm i}\nu_{\rm l}g_2 q + \sum_{u=4}^5 b_{2u}j^0_0(2\rightarrow u ) + \sum_{i\neq 2 } c_{2i } -d_2^{(2)}\biggr ] \rho^2_q(2 ) \\ & - \biggl ( \frac{1}{\sqrt{2 } } b_{24}j^2_{-q}(2\rightarrow 4 ) -\frac{2\sqrt{2}}{5 } b_{25}j^2_{-q}(2\rightarrow 5 ) \biggr)(-1)^q \rho^0_0(2 ) \\ & + \frac{1}{5 } a_{52}\rho^2_q(5 ) + \sum_{i=3 , 5 } c_{i2}^{(2 ) } \frac{\sqrt{2j_i+1}}{2 } \rho^2_q(i ) , \end{split } \displaybreak[0 ] \\ \begin{split}\label{see04 } \frac{\rm d}{{\rm d}t } \rho^0_0(3 ) = & -\biggl[b_{35}j^0_0(3\rightarrow 5 ) + \sum_{i\neq 3 } c_{3i } \biggr ] \rho^0_0(3 ) - \sum_{q=-2}^2 \frac{\sqrt{7}}{5 } b_{35}j^2_q(3\rightarrow 5 ) \rho^2_q(3 ) \\ & + \sqrt{\frac{2}{3 } } a_{53 } \rho^2_q(5 ) + \sum_{i\neq 3 } c_{i3 } \sqrt{\frac{2j_i+1}{6 } } \rho^0_0(i ) , \end{split } \displaybreak[0 ] \\ \begin{split}\label{see05 } \frac{\rm d}{{\rm d}t } \rho^2_q(3 ) = & -\biggl [ 2\pi{\rm i}\nu_{\rm l}g_3 q + b_{35}j^0_0(3\rightarrow 5 ) + \sum_{i\neq 3 } c_{3i } -d_3^{(2)}\biggr ] \rho^2_q(3 ) - b_{35}(-1)^q j^2_{-q}(3\rightarrow 5)\frac{\sqrt{7}}{5 } \rho^0_0(3 ) \\ & - \sum_{q , q=-2}^2 b_{35}j^2_{q}(3\rightarrow 5 ) \biggl [ -\sqrt{\frac{5}{7 } } \biggl ( \begin{matrix } 2 & 2 & 2 \\ q & -q ' & q \end{matrix } \biggr ) ( -1)^{q ' } \rho^2_{q'}(3 ) + \frac{9}{2}\sqrt{\frac{3}{35 } } \biggl ( \begin{matrix } 2 & 4 & 2 \\ q & -q ' & q \end{matrix } \biggr ) ( -1)^{q ' } \rho^4_{q'}(3 ) \biggr ] \\ & + \frac{2}{5}\sqrt{\frac{7}{3 } } a_{53 } \rho^2_q(5 ) + \sum_{i=2 , 5 } c_{i3}^{(2 ) } \sqrt{\frac{2j_i+1}{6 } } \rho^2_0(i ) , \end{split } \displaybreak[0 ] \\ \begin{split}\label{see06 } \frac{\rm d}{{\rm d}t } \rho^4_q(3 ) = & -\biggl [ 2\pi{\rm i}\nu_{\rm l}g_3 q + b_{35}j^0_0(3\rightarrow 5 ) + \sum_{i\neq 3 } c_{3i } -d_3^{(4)}\biggr ] \rho^4_q(3 ) \\ & - \sum_{q , q'=-2}^2 b_{35 } j^2_q(3\rightarrow 5 ) \biggl [ \frac{9}{2}\sqrt{\frac{3}{35 } } \biggl ( \begin{matrix } 4 & 2 & 2 \\ q & -q ' & q \end{matrix } \biggr)(-1)^{q ' } \rho^2_{q'}(3 ) + 3\sqrt{\frac{11}{70 } } \biggl ( \begin{matrix } 4 & 4 & 2 \\ q & -q ' & q \end{matrix } \biggr)(-1)^{q ' } \rho^4_{q'}(3 ) \biggr ] . \end{split } \displaybreak[0 ] \\ \begin{split}\label{see07 } \frac{\rm d}{{\rm d}t } \rho^0_0(4 ) = & -\biggl [ \sum_{\ell=1}^2 a_{4\ell } + \sum_{i\neq 4 } c_{4i } \biggr ] \rho^0_0(4 ) + \sum_{\ell=1}^2 b_{\ell 4 } j^0_0(\ell \rightarrow 4 ) \sqrt{\frac{2j_\ell+1}{2 } } \rho^0_0(\ell ) \\ & + \sum_{q=-2}^2 b_{24 } j^2_q(\ell \rightarrow 4 ) \rho^2_q(2 ) + \sum_{i\neq 4 } c_{i4 } \sqrt{\frac{2j_i+1}{2 } } \rho^0_0(i ) , \end{split } \displaybreak[0 ] \\ \begin{split}\label{see08 } \frac{\rm d}{{\rm d}t } \rho^0_0(5 ) = & -\biggl [ \sum_{\ell=1}^3 a_{5\ell } + \sum_{i\neq 5 } c_{5i } \biggr ] \rho^0_0(5 ) + \sum_{\ell=1}^3 b_{\ell 5 } j^0_0(\ell \rightarrow 5 ) \frac{\sqrt{2j_\ell+1}}{2 } \rho^0_0(\ell ) \\ & + \sum_{q=-2}^2 \biggl ( -\frac{2\sqrt{2}}{5 } b_{25 } j^2_q(2\rightarrow 5 ) \rho^2_q(2 ) + \frac{\sqrt{42}}{10 } b_{35 } j^2_q(3\rightarrow 5 ) \rho^2_q(3 ) \biggr ) + \sum_{i\neq 5 } c_{i5 } \frac{\sqrt{2j_i+1}}{2 } \rho^0_0(i ) , \end{split } \displaybreak[0 ] \\ \begin{split}\label{see09 } \frac{\rm d}{{\rm d}t } \rho^2_q(5 ) = & -\biggl [ 2\pi{\rm i}\nu_{\rm l}g_5 q + \sum_{\ell=1}^3 a_{5\ell } + \sum_{i\neq 5 } c_{5i } -d_5^{(2)}\biggr ] \rho^2_q(5 ) + \frac{1}{5}b_{25}j^0_0(2\rightarrow 5)\rho^2_q(2 ) + \frac{\sqrt{21}}{5}b_{35}j^0_0(3\rightarrow 5)\rho^2_q(3 ) \\ & + -\frac{2\sqrt{2}}{5 } b_{25 } j^2_{-q}(2\rightarrow 5 ) \rho^0_0(2 ) + \sqrt{\frac{3}{100 } } b_{35 } j^2_{-q}(3\rightarrow 5 ) \rho^0_0(3 ) \\ & + \sum_{q'=-2}^2 \biggl ( 2\sqrt{\frac{7}{5 } } b_{25 } j^2_{q'}(2\rightarrow 5 ) \rho^2_{q+q'}(2 ) -\sqrt{\frac{3}{5 } } b_{35 } j^2_{q'}(3\rightarrow 5 ) \rho^2_{q+q'}(3 ) \biggr ) \biggl ( \begin{matrix } 2 & 2 & 2 \\ -q & q - q ' & -q ' \end{matrix } \biggr ) \\ & + \sum_{i=2 , 3 } c_{i5}^{(2 ) } \frac{\sqrt{2j_i+1}}{2 } \rho^2_0(i ) . \end{split}\end{aligned}\ ] ] in the previous equations @xmath124 and @xmath125 are the einstein coefficients for spontaneous emission and absorption , respectively ( see table 1 ) . @xmath126 ( @xmath127 ; @xmath128 ) are the radiation field tensors introduced by landi deglinnocenti ( 1984 ) . they are integrals over frequency and solid angle of the stokes parameters ( for their explicit expressions appropiate for computational work see manso sainz & trujillo bueno 1999 ) . in particular , @xmath129 is the mean intensity integrated over the absorption profile , while @xmath130\,.\ ] ] the contribution of other @xmath131 components is very small here , where only the magnetic field breaks the symmetry of the radiation field . thus , @xmath132 throughout the whole outer atmosphere in the five transitions relevant for our problem . @xmath133 and @xmath134 are the excitation and desexcitation collisional rates , respectively ( see table 1 ) , between levels @xmath135 and @xmath136 , while @xmath137 and @xmath138 are collisional transfer rates for alignment between polarizable levels 2 , 3 , and 5 . derouich et al . ( 2007 ) have computed values for these transitions in ca ii . alternatively one can follow the approach suggested by landi deglinnocenti & landolfi ( 2004 ) , and consider @xmath139 where @xmath140 is just a numerical factor depending on the quantum numbers of the transition . @xmath141 is the depolarization rate of the @xmath42-th multipole of level @xmath142 due to elastic collisions with neutral hydrogen . theoretical values for ca ii are given by derouich et al . ( 2007 ) , who confirmed that such @xmath141 terms of the statistical equilibrium equations have only a very small impact on the atomic polarization of the metastable lower levels ( cf . , manso sainz & trujillo bueno 2003 ) . the first term on the r.h.s . ( [ see03 ] ) , ( [ see05])-([see06 ] ) , and ( [ see09 ] ) represents the relaxation of coherences in the presence of a magnetic field with larmor frequency @xmath143 ( @xmath19 expressed in gauss and @xmath144 in hz ) , and @xmath145 the land factor of the @xmath18-level under consideration . equations ( [ see01])-([see09 ] ) are expressed in a reference system with the quantization axis along the magnetic field . the radiative and collisional terms remain formally invariant in any other reference system although clearly , their actual numerical values may change due to the different values of @xmath131 in the new reference system . in an arbitrary reference system where the magnetic field has an inclination @xmath103 with respect to the quantization axis and an azimuth @xmath104 , the magnetic relaxation term reads : @xmath146 , \end{gathered}\ ] ] where the term multiplying @xmath147 ( @xmath148 ) is not considered when @xmath149 ( @xmath150 ) . the statistical equilibrium equations ( [ see01])-([see09 ] ) have been derived under several hypotheses . collisions are considered under the impact approximation . the interaction time of the collision is negligible and hence , collisional and radiative terms can be added independently . the colliders are assumed to be isotropically distributed ( they are maxwellian ) , and they can not generate atomic polarization , though they can transfer it between levels . radiative rates have been derived within the framework of the quantum theory of spectral line polarization described in landi deglinnocenti and landolfi ( 2004 ) , assuming that the radiation field is _ spectrally flat _ ( i.e. , lacking spectral structure ) , in spectral ranges of the order of the energy separation between levels with coherences . this is an excellent approximation for the case under consideration since we are considering relatively weak fields ( less than @xmath151100 g ) and we neglect coherences between different @xmath18 levels . neglecting @xmath18-level interferences is a good approximation for modeling the ca ii ir triplet and the core of the h- and k - lines , but it can not account for the general @xmath3 pattern of the uv doublet ( see stenflo 1980 ) , which is beyond the scope of this work . finally , we neglect doppler correlations due to the thermal movement of the atoms . all these hypotheses together account for what is called complete frequency redistribution ( crd ) in the laboraroty frame , since correlations between the incident and scattered radiation fields are completly neglected . we calculate the 29 @xmath74 unknowns at each grid point of the chosen stellar atmosphere model assuming statistical equilibrium @xmath152 the resulting system of equations is linearly dependent and we substitute one of the equations , say , the one for the ground - level population , by the trace equation of the density matrix , which establishes the conservation of the number of particles @xmath153 we have developed a general multilevel radiative transfer computer program for the numerical solution of the non - lte problem of the second kind , which we have applied in this investigation to obtain the self - consistent solution of the previous equations . see appendix a for details . cccccccc @xmath154 ( ) & @xmath136 & @xmath135 & @xmath124 ( s@xmath155 ) & + & & & & 3000 k & 5000 k & 7000 k & 9000 k + + k & 5 & 1 & @xmath156 & 3.60@xmath157 & 2.92@xmath157 & 2.59@xmath157 & 2.39@xmath157 + h & 4 & 1 & @xmath156 & 3.60@xmath157 & 2.92@xmath157 & 2.59@xmath157 & 2.39@xmath157 + 8498 & 5 & 2 & @xmath158 & 3.68@xmath157 & 2.93@xmath157 & 2.51@xmath157 & 2.23@xmath157 + 8542 & 5 & 3 & @xmath159 & 1.59@xmath160 & 1.29@xmath160 & 1.12@xmath160 & 1.01@xmath160 + 8662 & 4 & 2 & @xmath161 & 1.61@xmath160 & 1.31@xmath160 & 1.15@xmath160 & 1.04@xmath160 + + & 5 & 4 & & 3.94@xmath157 & 3.05@xmath157 & 2.58@xmath157 & 2.27@xmath162 + & 4 & 3 & & 2.69@xmath157 & 2.08@xmath157 & 1.76@xmath157 & 1.55@xmath157 + & 3 & 1 & & 2.20@xmath157 & 1.71@xmath157 & 1.44@xmath157 & 1.27@xmath157 + & 3 & 2 & & 5.51@xmath157 & 4.27@xmath157 & 3.61@xmath157 & 3.19@xmath157 + & 2 & 1 & & 2.20@xmath157 & 1.71@xmath157 & 1.44@xmath157 & 1.27@xmath157 + high spatial and temporal resolution intensity images taken at the line center of strong absorption lines , like h@xmath12 and ca ii 8542 , show that the chromosphere of the quiet sun is a highly inhomogeneous and dynamic medium ( e.g. , review by rutten 2007 ) . the spatial and dynamic complexity is such that it is obvious that the two one - dimensional ( 1d ) atmospheric models we have chosen for this investigation ( i.e. , the hot " fal - c model of fontenla et al . 1993 and the cool " m - co model of avrett 1995 ) should only be considered as rough representations of the stratification of the kinetic temperature and density at two different phases of inter - network oscillations ( see fig . 2 ) . nevertheless , as we shall see below , our solution of the multilevel hanle - effect problem for the ir triplet of ca ii in such semi - empirical models of the solar atmosphere allows us to reach the main goal of this paper , namely to demonstrate the diagnostic potential of the linear polarization signals produced by atomic polarization in the levels of the ca ii ir triplet . we consider a 5-level atomic model of ca ii including the ground level , the metastable @xmath163 levels and the @xmath164 upper levels of the @xmath165 and @xmath42 lines ( see fig . the excitation state in such a model is given by 29 statistical tensor elements @xmath74 ( @xmath166 , 2 , 4 ; @xmath167 ) , including the total populations of the five levels . the radiative transition probabilities are those of edln & risberg ( 1956 ) compiled at the nist database . ( note that these transition probabilities for the infrared triplet are slightly larger than those used by manso sainz & trujillo bueno ( 2003a ) for the unmagnetized case ; correspondingly , the @xmath3 values for the zero - field case are slightly larger here than in our previous work . ) inelastic collisional rates and line broadening parameters have been taken from shine & linsky ( 1974 ; see table 1 ) . elastic depolarizing collisions are treated according to lamb & ter haar ( 1971 ) , which gives results similar to those obtained using the collisional rates given by derouich et al . the collisional pumping of @xmath168 statistical tensors are treated following landi deglinnocenti & landolfi ( 2004 ) . we have calculated the number density of ca ii ions at each atmospheric height of the model under consideration by solving the standard non - lte radiative transfer problem for a realistic atomic model that includes also the bound - free transitions from all the bound levels of ca ii to the ground level of ca iii . with the resulting number densities of ca ii ions we then solved the non - lte problem of the second kind for the atomic model of fig . 1 ( see the appendix ) , which is sufficiently realistic for investigating the polarization signatures produced by atomic level polarization and the hanle effect in the lines of the ca ii ir triplet . we point out that while @xmath169 ( see eq . 1 ) is of the order of a few gauss for the upper level of the 8498 and 8542 lines , it lies in the milligauss range for the lower ( metastable ) levels of the ca ii ir triplet . for symmetry reasons , in a one - dimensional model atmosphere without magnetic fields the only non - zero stokes parameters are @xmath170 and @xmath171 ( choosing the polarization directions as in eqs . ( [ epsilons ] ) ) . figure 3 shows the wavelength variation of the emergent fractional linear polarization ( @xmath3 ) in the ca ii ir triplet , calculated for a line of sight with @xmath172 in the hot " fal - c model ( middle panels ) and in the cool " m - co model ( bottom panels ) . while @xmath3 in the 8542 and 8662 lines are positive in both models , the @xmath3 profile of the 8498 line is positive in the hot " model but negative in the cool " model . moreover , the 8498 @xmath3 profiles calculated in such atmospheric models have even different shapes , and their amplitudes are one order of magnitude smaller than those of the 8542 and 8662 lines . the ( positive ) @xmath3 amplitudes of the 8542 and 8662 lines turn out to be significantly larger in the cool " model . this dependence of the amplitude of the @xmath3 profiles on the chosen model atmosphere is not at all surprising , because the atomic polarization that anisotropic pumping processes induce in the atomic energy levels sensitively depends on the temperature structure of the model . in particular , the steeper the temperature gradient , the larger the anisotropy factor and hence , the induced atomic level polarization ( e.g. , see figure 4 in trujillo bueno 2001 ) . the upper panels of fig . 4 show the spatial variation of the degree of anisotropy " in each of the considered atmospheric models , indicating for each los characterized by @xmath173 the atmospheric height where the line - center optical depth is unity along the los . note that at the height corresponding to a los with @xmath172 the anisotropy is larger in the m - co model . the decrease of the anisotropy factor at very low optical depths is due to the rapid widening of the @xmath174 profile with which the radiation field is averaged ( e.g. , see eq . ( 18 ) for the definition of the @xmath175 radiation field tensor component ) . this is a consequence of the temperature rising in the upper chromosphere - transition region . the two lower panels of fig . 4 show the behavior of the fractional atomic alignment of the levels of fig . 1 having @xmath13 ( i.e. , the two metastable levels , which are the lower levels of the ca ii ir triplet , and the upper level of the k - line ) . the fact that the sign of the @xmath3 profile of @xmath176 ( i.e. , the weakest line of the ca ii ir triplet ) is very sensitive to the thermal structure of the lower chromosphere can be easily understood by combining the information provided in the lower panels of fig . 4 and the approximate eddington - barbier formula ( 8) . 3 suggests that the scattering polarization of the ca ii ir triplet can be used as a sensitive thermometer of the quiet " regions of the solar chromosphere , as already noted by manso sainz & trujillo bueno ( 2001 ) . in this respect , a particularly useful quantity is the center - limb variation of the @xmath3 line - center amplitudes ( see fig . 5 ) . as seen in this figure , the most noteworthy feature is that in the hot " fal - c model the ( positive ) @xmath3 amplitude of the 8498 line is larger at @xmath177 than at @xmath178 and that it becomes negative for line of sights with @xmath173 values significantly smaller than 0.1 . the aim of this section is to show how the hanle effect modifies the emergent linear polarization of the ir triplet of ca ii . to this end , we consider the following two geometries . in the absence of magnetic fields and horizontal atmospheric inhomogeneities the polarization of the atomic levels do not produce any linear polarization in the spectral line radiation observed at disc center ( @xmath179 ; see fig . the same is true in the presence of a vertical magnetic field , because in a one - dimensional stellar atmospheric model the vertical direction coincides with the symmetry axis of the anisotropic radiation field that induces atomic level polarization . however , in the presence of an inclined magnetic field the symmetry of the problem is broken and , as a result , the ensuing atomic level polarization can generate linear polarization even for a line of sight with @xmath179 . in this case , the linear polarization is created by the hanle effect of the inclined field ( see a demonstrative observational example in trujillo bueno et al . 2002 ) . figure 6 shows the fractional linear polarization of the ir triplet of ca ii generated in forward scattering geometry by the hanle effect of a horizontal magnetic field in the fal - c ( top panels ) and m - co ( bottom panels ) models . in the ( hot " ) fal - c model , the linear polarization amplitudes are at the level of @xmath180 when the magnetic strength of the horizontal field is similar or larger than only 0.1 gauss , with @xmath181 ( i.e. , parallel to the horizontal magnetic field ) for the 8498 line and with @xmath182 ( i.e. , perpendicular to the horizontal magnetic field ) for the 8542 and 8662 lines . in the ( cool " ) m - co model @xmath181 in the 8542 and 8662 lines , with amplitudes again at the @xmath180 level for @xmath183 gauss , while the shape of @xmath3 in the 8498 line is very peculiar ( i.e. , similar to the typical observational signature of the transverse zeeman effect ) and with its maximum amplitude one order of magnitude smaller . figure 7 shows the results of our rt calculations for the case of a magnetic field inclined by only @xmath184 . note that now , @xmath185 . at present , detection of these weak linear polarization signals is challenging , though feasible . but they will become straigtforward with future large - aperture solar telescopes such as the advanced technology solar telescope ( keil , rimmele & wagner 2009 ) or the european solar telescope ( collados 2008 ) . in forward scattering geometry , symmetry imposes that linear polarization vanishes if the magnetic field is inclined but has a random azimuth distribution , unlike in the fixed inclination and azimuth case just considered . however , for observations away from the solar disk center ( @xmath186 ) , scattering may produce linear polarization in the emergent spectral line radiation even for magnetic fields with a uniformly distributed azimuth within the spatio - temporal resolution element of the observations . this is illustrated in fig . 8 , which shows the linear polarization amplitudes of the ca ii ir triplet in the fal - c model ( top panels ) and in the m - co model ( bottom panels ) vary with the magnetic strength ( @xmath19 ) and the inclination ( @xmath103 ) of the assumed random - azimuth field for a close - to - the - limb observation ( @xmath172 ) . figure 8 shows how sensitive the @xmath3 line - center amplitude of the 8498 line is to the model s thermal structure , and its very peculiar behavior in the sub - gauss range . the 8542 and 8662 lines are mostly sensitive to magnetic fields between 0.001 and 0.1 gauss , although the 8542 line is also weakly sensitive to fields between 1 and 10 gauss . the 8542 line enters into the saturation regime of the upper - level hanle effect for fields stronger than @xmath15110 gauss , while just 0.1 gauss saturate the 8662 line . on the contrary , the 8498 line is highly sensitive in the whole regime 0.00110 gauss . as seen in fig . 3 , in the absence of magnetic fields the linear polarization amplitude of the 8498 line for a los with @xmath172 is one order of magnitude smaller than those corresponding to the 8542 and 8662 lines . however , the three ca ii lines may show similar @xmath3 amplitudes if there is a magnetic field of the order of only 0.1 gauss in the atmospheric region where the linear polarization of the 8498 line originates . figure 9 shows the emergent @xmath3 amplitudes averaging the @xmath3 profiles of both atmospheric models with a weight of 0.55 for the fal - c profiles and 0.45 for the m - co profiles . these weights are similar to those used by avrett ( 1995 ) to simultaneously reproduce the intensity profiles observed in the 4.6 @xmath187 lines of co and in the h & k lines of ca ii , and to those used by holzreuter et al . ( 2006 ) to fit the @xmath3 profile of the ca ii k - line observed in a quiet regions very close to the solar limb . these type of fits obtained by mixing a cool and a hot component should not be considered as a realistic representation of the chromospheric conditions , but as further indication of the complexity of the ( time - dependent ) three - dimensional thermal structure of the solar chromosphere . nevertheless , this two - component model shows that the @xmath3 observations of the ir triplet of ca ii reported by stenflo et al . ( 2000 ) are compatible with very weak magnetic fields in the region of formation " of the 8498 line ( i.e. , in the lower chromosphere ) and with fields stronger than 10 gauss ( and inclined by about @xmath184 ) in the regions of formation " of the 8542 and 8662 lines ( i.e. , in the upper chromosphere ) . stokes @xmath2 is zero if the magnetic field has a random - azimuth distribution , and it is non - zero if the atmosphere is permeated by a magnetic field vector with a well - defined inclination and azimuth . this is illustrated in the upper panels of fig . 10 , which corresponds to the fal - c model . in the absence of magnetic fields , only @xmath3 is non - zero ( see the dashed lines ) , but @xmath4 becomes significant even at fields strengths as low as 0.005 gauss . note that for this case ( b=0.005 gauss ) , characterized by a horizontal magnetic field pointing towards the observer ( triangles ) , perpendicular to the los ( squares ) and away from the los ( diamonds ) , the stokes @xmath3 profiles of the ca ii ir triplet lines are always positive , while @xmath4 changes its sign . the dotted lines of fig . 10 show the case of a 100 gauss horizontal field pointing towards the observer ( i.e. , a strength for which the three ir lines of ca ii are in the saturation regime of the upper - level hanle effect ) . the bottom panels of fig . 10 show the results of similar calculations for the cool m - co model . in this model the @xmath3 profile of the 8498 line changes its sign and shape when the magnetic field azimuth changes from @xmath188 ( or @xmath189 ) to @xmath190 a hanle - effect diagram shows the line - center amplitudes of @xmath3 and @xmath4 varying with the strength and the azimuth of the magnetic field corresponding to a given line - of - sight and inclination of the magnetic field vector . figure 11 shows restricted hanle - effect diagrams corresponding to the configurations considered in fig . 10 ( close - to - the - limb @xmath172 observation ) . in particular , the thin solid lines show how linear polarization varies with the magnetic strength , for a horizontal magnetic field pointing towards the observer . we recall that 50 gauss are needed to completely depolarize the 8498 line , while just 10 gauss are sufficient for the 8542 line and only 0.1 gauss for the 8662 line ( see fig . 8) . thick solid lines show , for the case of a 0.005 gauss horizontal field , how the @xmath3 and @xmath4 line center amplitudes vary as the orientation of the magnetic field vector changes from pointing towads the observer ( triangles ) , to being perpendicular to the line of sight ( squares ) to pointing away from the observer ( diamonds ) . comparing the upper and lower panels we see the differences due to the different thermal structures of the fal - c and m - co models . figures 12 and 13 show full hanle - effect diagrams obtained from the emergent stokes @xmath3 and @xmath4 profiles calculated for a close - to - the - limb ( @xmath172 ) line of sight in the hot " and cool " atmospheric models , respectively . each panel corresponds to a fixed inclination of the magnetic field vector . each of the four thick lines in any given panel correspond to a fixed strength of the magnetic field with azimuth ranging from @xmath188 to @xmath189 ( solid line ) , and from @xmath189 to @xmath191 ( dashed line ) . thin lines correspond to magnetic fields ranging between 0 g and 1000 g for a fixed @xmath104 . they show the pattern of depolarization ( @xmath3 decreases ) and rotation of the direction of linear polarization ( @xmath4 appears ) characteristic of the hanle effect in a close to the limb scattering geometry . finally , figs . 14 and 15 show hanle - effect diagrams for a disc center observation ( @xmath179 ) . note that in this forward scattering geometry the hanle effect of a magnetic field inclined with respect to the solar local vertical creates linear polarization . we have investigated theoretically the sensitivity of the scattering polarization @xmath3 and @xmath4 profiles of the ca ii ir triplet to the presence of magnetic fields through the hanle effect . to this end , we have applied a multilevel radiative transfer code ( see appendix a ) for treating the transfer of spectral line polarization due to scattering in weakly magnetized stellar atmospheres . we find that the ca ii 8542 and 8662 lines are mostly sensitive to fields between 0.001 and 0.1 g , the former being also weakly sensitive to fields between 1 and 10 g. otherwise the linear polarization signals are dependent only on the magnetic field geometry , but not to its intensity . the 8498 line is highly sensitive in the whole regime 0.00110 g , but its scattering polarization amplitude is significantly lower in the absence of magnetic fields . probably , the most interesting line of the ca ii ir triplet because of its diagnostic potential is the strongest one : @xmath192 . its linear polarization is sensitive to the orientation of the magnetic field in the chromosphere of the quiet sun , though not too much to its strength unless @xmath193 g there . although its scattering polarization amplitude depends on the thermal structure of the solar chromosphere , the emergent @xmath3 and @xmath4 profiles do not change sign between the ( hot " ) fal - c model and the ( cool " ) m - co model , unlike @xmath194 , which is very sensitive to the physical conditions of the lower chromosphere . the thermal sensitivity of the @xmath195 line ( which is sensitive only to milligauss field strengths ) is similar to that of the 8542 line . therefore , spatial fluctuations in the sign of the emergent @xmath4 profiles of the ca ii 8542 and 8662 lines may be ascribed more safely to spatial variations in the azimuth of the chromospheric magnetic field than is the case for @xmath194 . more generally , the emergent @xmath3 and @xmath4 profiles are produced by the joint action of atomic level polarization , and the hanle and transverse zeeman effects . in this paper we have considered only the action of atomic level polarization and the hanle effect . these effects dominate the emergent linear polarization profiles for inclined magnetic fields with strengths weaker than @xmath196 , where the @xmath196 value depends on the scattering geometry . in the forward scattering geometry of a disk center observation ( @xmath179 ) the linear polarization of the ca ii ir triplet is dominated by the hanle effect if the magnetic field is weaker than about 10 g. in fact , while the contribution of the transverse zeeman effect is negligible for @xmath197 g the hanle effect creates fractional linear polarization signals of the order of @xmath180 already for horizontal magnetic fields as weak as 0.1 g ( see figs . 6 and 7 ) . for magnetic strengths @xmath198 g the contribution of the transverse zeeman effect to the linear polarization observed at the solar disk center should not be neglected . detection of @xmath3 and @xmath4 disk center signals caused either by the hanle effect alone ( if @xmath199 g ) or by the joint action of the hanle and transverse zeeman effects ( if @xmath198 g ) would require very high polarimetric sensitivity together with a spatial and temporal resolution sufficient to at least resolve partially the magnetic field azimuth . given the weakness of the theoretical scattering polarization signals ( @xmath200 at @xmath179 ; @xmath201 at @xmath172 ) , the most favourable geometry for observing them is in quiet regions close to the limb . in this geometry the emergent linear polarization of the ca ii ir triplet is non - zero even if the magnetic field has a random azimuth within the spatio - temporal resolution element of the observation ( see figs . 8 and 9 ) . moreover , for magnetic strengths sensibly weaker than 100 g the contribution of the transverse zeeman effect to the emergent linear polarization is expected to be smaller than that caused by atomic level polarization . for stronger fields ( e.g. , @xmath202 g ) the wings of the emergent @xmath3 and @xmath4 profiles may show sizable features produced by the transverse zeeman effect , but the line - center amplitudes would still be dominated by atomic level polarization . finally , we point out that information on the relative amplitudes and signs of the scattering polarization profiles of the ca ii ir triplet would be very useful to help constrain the thermal and magnetic structure of the quiet solar chromosphere . in principle , to determine the structure of the magnetic field via stokes inversion techniques is possible , although it would be easier if one is able to infer first the atmospheric thermal structure from the observed intensity spectrum . the @xmath3 and @xmath4 signals we have studied here should also be exploited to investigate the reliability of three - dimensional models of the solar chromosphere resulting from holistic magnetohydrodynamic ( mhd ) simulations of the solar atmosphere . to this end , it suffices with confronting spectropolarimetric observations of the ca ii ir triplet with synthetic stokes profiles obtained by solving the non - lte problem of the 2nd kind in 3d snaphot models resulting from such mhd simulations . modeling the spectral line polarization produced by atomic level polarization and its modification by the hanle effect requires calculating , for multilevel systems , the excitation and ionization state of chemical species of given abundance that is consistent with both the intensity and polarization of the radiation field generated within the ( generally magnetized ) plasma under consideration . this so - called non - lte problem of the 2nd kind ( cf . , landi deglinnocenti & landolfi 2004 ) is a very involved non - local and non - linear radiative transfer problem which requires solving the statistical equilibrium equations for the elements of the atomic density matrix and the stokes - vector transfer equation for each of the allowed transitions in the multilevel model . once such a self - consistent excitation state is known throughout the medium , it is then straightforward to solve the transfer equation for any desired line of sight in order to obtain the emergent stokes profiles to be compared with spectropolarimetric observations . to this end , we developed a general computer program , which we will describe in greater detail in a future publication along with the efficient and accurate radiative transfer methods on which it is based ( for a first advance see the workshop contribution by manso sainz & trujillo bueno 2003 ; see also manso sainz 2002 ) . in this appendix we present only a brief summary of the code with emphasis on its application to solve the problem of scattering polarization and the hanle effect in the ir triplet of ionized calcium . 1 . the equations are formulated within the spherical tensors representation of the density matrix and radiation field tensor . there are @xmath203 unknowns @xmath74 with @xmath42 even , for each level with integer angular momentum @xmath18 , and @xmath204 unknowns for each level with semi - integer angular momentum @xmath18 . 2 . the radiative transfer equations for the stokes parameters are integrated along short characteristics using the quasi - parabolic delo method ( delopar ; trujillo bueno 2003 ) . the iterative corrections for the unknowns ( the statistical tensors @xmath74 ) are calculated applying a suitable generalization to the multilevel atom case of the fast iterative methods described in trujillo bueno & manso sainz ( 1999 ) . this implies writing down the statistical equilibrium equations taking explicitly into account the contribution of the diagonal components of the @xmath205 operator and linearizing according to eqs . ( 53 ) and ( 54 ) of trujillo bueno ( 2003 ) . the iterative scheme can speeded up further through polynomial ng - acceleration ( e.g. , auer 1987 ) . the number density of the ion under consideration is computed in non - lte . at each point in the atmosphere we solve the statistical equilibrium equations for the multipolar components of the atomic density matrix plus the conservation of particles equation to calculate the excitation state of the atomic system . we calculate the radiation field at each point in the model atmosphere for each of the allowed radiative transitions in the model atom by formal integration of the radiative transfer equations for the stokes parameters applying the delopar method . then , radiation field tensors @xmath131 are calculated and used to work out the new excitation state of the atomic system and so on , iteratively . figure 16 shows the convergence behavior of the above - mentioned iterative method for the numerical solution of the ( ca ii ) non - lte problem of the second kind .

we consider a random sparse graph with bounded average degree , in which a subset of vertices has higher connectivity than the background . in particular , the average degree inside this subset of vertices is larger than outside ( but still bounded ) . given a realization of such graph , we aim at identifying the hidden subset of vertices . this can be regarded as a model for the problem of finding a tightly knitted community in a social network , or a cluster in a relational dataset . in this paper we present two sets of contributions : we use the cavity method from spin glass theory to derive an exact phase diagram for the reconstruction problem . in particular , as the difference in edge probability increases , the problem undergoes two phase transitions , a static phase transition and a dynamic one . we establish rigorous bounds on the dynamic phase transition and prove that , above a certain threshold , a local algorithm ( belief propagation ) correctly identify most of the hidden set . below the same threshold _ no local algorithm _ can achieve this goal . however , in this regime the subset can be identified by exhaustive search . for small hidden sets and large average degree , the phase transition for local algorithms takes an intriguingly simple form . local algorithms succeed with high probability for and fail for ( with , the average degrees inside and outside the community ) . we argue that spectral algorithms are also ineffective in the latter regime . it is an open problem whether any polynomial time algorithms might succeed for .

1502.05680

the problem of finding a highly connected subset of vertices in a large graph arises in a number of applications across science and engineering . within social network analysis , a highly connected subset of nodes is interpreted as a community @xcite . many approaches to data clustering and dimensionality reduction construct a ` similarity graph ' over the data points . a highly connected subgraph corresponds to a cluster of similar data points @xcite . a closely related problem arises in the analysis of matrix data , e.g. in microarray data analysis . in this context , researchers are often interested in a submatrix whose entries have an average value larger ( or lower ) than the rest @xcite . such an anomalous submatrix is interpreted as evidence of association between gene expression levels and phenotypes ( e.g. medical conditions ) . if we consider the graph adjacency matrix , a highly connected subset of vertices corresponds indeed to a principal submatrix with average value larger than the background . the special case of finding a completely connected subset of vertices ( a clique ) in a graph has been intensely studied within theoretical computer science . assuming p@xmath6np , the largest clique in a graph can not be found in polynomial time . even a very rough approximation to its size is hard to find @xcite . in particular , it is hard to detect the presence of a clique of size @xmath7 in a graph with @xmath8 vertices . such hardness results motivated the study of random instances . in particular , the so - called ` planted clique ' or ` hidden clique problem ' @xcite requires to find a clique of size @xmath9 that is added ( planted ) in a random graph with edge density @xmath10 . more precisely , for a subset of vertices @xmath11 $ ] , all edges @xmath12 , with @xmath13 are present . all other edges are present independently with probability @xmath10 . such a clique can be found reliably by exhaustive search as soon as @xmath14 @xcite . however , despite many efforts , no algorithm is known that achieves this goal for @xmath15 @xcite . in other words , the problem of finding cliques of size @xmath16 is solvable , but possibly hard . proving that indeed it is computationally hard to find cliques in this regime is an outstanding problem in theoretical computer science . for general polynomial algorithms , it is known since @xcite that a clique of size @xmath17 can be found in time @xmath18 for any @xmath19 fixed . hence , if we allow any time complexity polynomial in @xmath8 , then the question is whether the planted clique can be found for @xmath20 . a more stringent computational constraint requires that the clique is found in nearly - linear time , i.e. in time of order @xmath21 . note that the number of bits required to encode an instance of the problem is of order @xmath22 , so @xmath23 is a logarithmic multiple of the time required to read an instance . dekel , gurel - gurevittch and peres @xcite developed a linear - time algorithm ( i.e. with complexity @xmath24 ) that finds the hidden clique with high probability , provided @xmath25 . in @xcite it was proved that , if @xmath26 , then there exists a message passing algorithm that finds with high probability the clique with @xmath27 operations . the same paper provided evidence that a certain class of ` local ' algorithms fails at the same threshold . among other motivations , the present paper generalizes and supports the existence of a fundamental threshold for local algorithms at least in the sparse graph setting . in the present paper , we consider the problem of finding a highly connected subset of vertices in a _ sparse graph _ , i.e. in a graph with _ bounded average degree_. in this case , the hidden set size must scale linearly with @xmath8 to obtain a non - trivial behavior . somewhat surprisingly , we find that the phase transition ` at @xmath28 ' leaves a trace also in the sparse regime . more precisely , we consider a random graph generated as follows . we select a subset of vertices @xmath29 of size @xmath30 , uniformly at random given its size . we connect any two vertices in the set independently with probability @xmath31 . any other edge is added independently with probability @xmath32 , @xmath33 . the problem distribution is therefore parametrized by @xmath34 and we will be therefore interested in the limit @xmath35 , with @xmath36 fixed . a more intuitive parametrization is obtained by replacing @xmath37 with the average degrees for vertices @xmath38 , and @xmath39 denoted , respectively , by @xmath4 , @xmath5 our main rigorous result is a sharp phase transition in the following double asymptotics : * first @xmath35 . this corresponds to considering large graphs . * then @xmath40 and @xmath41 . this corresponds to focusing on small hidden sets , but still linear in @xmath8 . the requirement @xmath41 is a necessary consequence of @xmath40 : it can be shown that otherwise the hidden set can not possibly be detected . our main rigorous result ( theorem [ thm : main ] ) establishes that , in the above double asymptotics , a phase transition takes place for local algorithms at @xmath42 namely , we consider the problem of testing whether a vertex @xmath43 is in @xmath29 or not . we say that such a test is reliable if , in the above limit , the fraction of incorrectly estimated vertices vanishes in expectation . then : * for @xmath44 , a local algorithm can estimate reliably @xmath29 in time of the order of the number of edges . this is achieved for instance , by the belief propagation algorithm . * for @xmath45 , no local algorithm can reliably reconstruct @xmath29 . analogously to the classical hidden clique problem , there is a large gap between what can be achieved by local algorithms , and optimal estimation with unbounded computational resources . proposition [ propo : exhaustive ] estabilishes that exhaustive search will find @xmath29 in exponential time , as soon as @xmath46 for some positive @xmath47 ( in the same double limit ) . note that , in both cases , a small fraction of the vertices in @xmath29 remains undetected because of the graph sparsity , for any @xmath48 . in particular , the number of vertices of degree @xmath49 is linear in @xmath8 , and such nodes can not be identified . let us finally mention the degree of a vertex @xmath43 is a poisson with mean @xmath4 if @xmath38 and mean @xmath5 if @xmath39 . hence , the degree standard deviation ( outside @xmath29 ) is @xmath50 . therefore , the ratio @xmath51 is the difference in mean degree divided by the standard deviation , and has the natural interpretation of a ` signal - to - noise ratio . ' while our rigorous analysis focuses on the limit @xmath40 and @xmath41 ( after @xmath35 ) , we will use the cavity method from spin glass theory to investigate the model behavior for _ arbitrary _ @xmath52 ( in the @xmath35 limit ) or , equivalently , arbitrary @xmath36 . we will use two approaches to obtain concrete predictions from the cavity method : * for general @xmath37 ( bounded degree ) , we derive the cavity predictions for local quantities , as well as for the free energy density . we show that this indeed coincide ( up to a shift ) with the mutual information per variable between the hidden set @xmath29 and the observed graph @xmath53 . + we use the ` population dynamics ' ( or ` sampled density evolution ' ) algorithm @xcite to solve numerically the cavity equations . * we then consider the limit of large @xmath37 ( large degree ) for arbitrary @xmath54 . in order to obtain a non - trivial limit , the signal - to - noise ratio @xmath55 $ ] is kept fixed in this limit , together with @xmath56 $ ] . + the cavity equations simplify in this limit ( the cavity field distributions become gaussian ) , and we can derive an exact phase diagram , without recourse to intensive numerical methods , cf . figure [ fig : phasediaglargedeg ] . this two approaches are complementary in that the large - degree asymptotics yields closed - form expressions . the qualitative features of the resulting phase diagram should remain unchanged at moderately small values of @xmath37 . our population dynamics analysis confirms this . as already mentioned , one of the motivations for the present work was to better understand the computational phase transition discovered in @xcite for the classical hidden clique problem . for background edge density @xmath10 , this takes place when the size of the hidden clique is @xmath57 . this phase transition can indeed be formally recovered as a dense limit of the results presented in this paper . more precisely , the phase transition for hidden cliques @xcite is captured by eq . ( [ eq : basiccondition ] ) , once we rewrite the latter in terms @xmath58 , the variance of the degrees of nodes @xmath39 . in the sparse regime , the degree is approximately poisson distributed , and hence @xmath59 . therefore eq . ( [ eq : basiccondition ] ) can be rewritten as @xmath60 for the classical ( dense ) hidden clique problem , we have @xmath61 , @xmath62 and @xmath63 , and hence we recover the condition @xmath64 . from a different perspective , the present work offers a statistical mechanics interpretation of the phase transitions in the hidden clique problem . namely , the latter can be formally recovered as the @xmath40 limit of the phase diagram in figure [ fig : phasediaglargedeg ] below . in particular , the computational phase transition at @xmath65 corresponds to a dynamical phase transition ( a spinodal point ) in the underlying statistical mechanics model . the rest of the paper is organized as follows . in the next section we define formally our model and some related notations . section [ sec : cavity ] derives the phase diagram using the cavity method . in particular , we show that the model undergoes two phase transitions as the signal - to - noise ratio increases ( for @xmath66 small enough ) : a static phase transition and a dynamic one , the two phase transitions are well separated . section [ sec : rigorous ] presents rigorous bounds on the behavior of local algorithms and exhaustive search , that match the above phase transitions for small @xmath66 . this section is self - contained and the interested reader can move directly to it , after the model definition ( some useful , but elementary results are presented in section [ sec : treeinterpretation ] ) . proofs are deferred to the appendix . finally , section [ sec : related ] positions our results in the context of recent literature . several research communities have been working on closely related problems : statistical physics , theoretical computer science , machine learning , statistics , information theory . we tried to write a paper that could be accessible to researchers with different backgrounds , both in terms of tools and of language . we apologize for any redundancy that might have followed from this approach . we use @xmath67= \{1,\dots,\ell\}$ ] to denote the set of first @xmath68 integers , and @xmath69 to denote the size ( cardinality ) of set @xmath70 . ) for a set @xmath71 , we write @xmath72 to indicate that @xmath12 runs over all unordered pairs of distinct elements in @xmath71 . for instance , for a symmetric function @xmath73 , we have @xmath74 } f(i , j ) \equiv \prod_{1\le i < j\le n } f(i , j)\ , . \ ] ] if instead @xmath75 is a set of edges over the vertex set @xmath71 ( unordered pairs with elements in @xmath71 ) we write @xmath76 to denote elements of @xmath75 . we use @xmath77 to denote the gaussian distribution with mean @xmath78 and variance @xmath79 . other classical probability distributions are denoted in a way that should be self - explanatory ( bernoulli@xmath80 , poisson@xmath81 , and so on ) . we consider a random graph @xmath82 with vertex set @xmath83 \equiv \{1,\dots , n\}$ ] and random edges generated as follows . a set @xmath84 is chosen at random . introducing the indicator variables @xmath85 we let @xmath86 independently with @xmath87 in particular @xmath88 is a binomial random variable , and is tightly concentrated around its mean @xmath89 . edges are independent given @xmath29 , with the following probability for @xmath90 distinct : @xmath91 we let @xmath92 denote the vector identifying @xmath29 . by using bayes theorem , the conditional distribution of @xmath93 given @xmath53 is easily written @xmath94}\big(\frac{\kappa}{1-\kappa}\big)^{x_i } \prod_{(i , j)\subseteq [ n ] } \big(\frac{1-a / n}{1-b / n}\big)^{x_ix_j}\prod_{(i , j)\in e } \rho_n^{x_ix_j}\ , , \label{eq : exactposterior } \ ] ] where @xmath95 . we next replace the last probability distribution with one that is equivalent as @xmath35 , and slightly more convenient for the cavity calculations of the next section . ( these simplifications will not be used to prove the rigorous bounds in section [ sec : rigorous ] . ) we first note that , as @xmath35 , we have @xmath96 with @xmath97 next , letting @xmath98 , we can rewrite the second product in eq . ( [ eq : exactposterior ] ) as @xmath74 } & \big(\frac{1-a / n}{1-b / n}\big)^{x_ix_j } = \big(\frac{1-a / n}{1-b / n}\big)^{\binom{|x|}{2 } } = c\ , \big(\frac{1-a / n}{1-b / n}\big)^{(2\kappa n-1)(|x|-\kappa n)/2 } \cdot \big(\frac{1-a / n}{1-b / n}\big)^{\frac{1}{2}(|x|-\kappa n)^2}\label{eq : simplificationglobal } \ ] ] with @xmath99 a constant independent of @xmath93 . notice that @xmath100 is tightly concentrated around @xmath101 . in particular @xmath102 with high probability , and therefore the last term in eq . ( [ eq : simplificationglobal ] ) is of order @xmath103 . we will neglect it , thus obtaining @xmath74 } & \big(\frac{1-a / n}{1-b / n}\big)^{x_ix_j } \approx c'\ , \big(\frac{1-a / n}{1-b / n}\big)^{\kappa n(|x|-\kappa n ) } \approx c '' \ , e^{-\kappa(a - b)|x|}\ , . \ ] ] the error incurred by neglecting the last term in eq . ( [ eq : simplificationglobal ] ) can be corrected by considering the following approximate conditional distribution of @xmath93 given the graph @xmath53 @xmath104 where @xmath105 is the indicator function on condition @xmath70 and @xmath106 note that we multiplied @xmath107 by the indicator function @xmath108 . this can be interpreted as replacing the i.i.d . bernoulli distribution ( [ eq : kappadef ] ) with the uniform distribution over @xmath29 with @xmath109 , which is immaterial as long as local properties of @xmath110 are considered . in the following , we shall compare different reconstruction methods . any such method corresponds to a function @xmath111 of vertex @xmath43 and graph @xmath53 , with the interpretation @xmath112 we characterize such a test through its rescaled success probability @xmath113 note that a trivial test ( assigning @xmath111 at random independently of @xmath53 ) achieves @xmath114 , while a perfect test has @xmath115 . we shall often omit the arguments @xmath116 , @xmath117 from @xmath118 in the following . we note in passing that the optimal estimator with respect to the metric ( [ eq : psucct ] ) is the maximum - likelihood estimator @xmath119 namely , for any other estimator @xmath116 , we have @xmath120 ( see , for instance , the textbook @xcite for a proof of this fact ) . the resulting success probability coincides with the total variation distance between the conditional distribution of @xmath53 given the two hypotheses @xmath38 and @xmath121 . recall that , given two probability measures @xmath122 on the same finite space @xmath123 , their total variation distance is defined as @xmath124 . then we have @xmath125 in this section we use the cavity method to derive an exact phase diagram of the model . it is convenient to introduce the following signal - to - noise - ratio parameter : @xmath126 using the fact that the degree outside @xmath29 is poisson with mean @xmath127 , and inside is poisson with mean @xmath128 , we also have @xmath129 we will therefore think in terms of the three independent parameters : @xmath54 ( the relative size of @xmath88 ) ; @xmath130 ( the average degree in the background ) ; @xmath131 ( the signal - to - noise ratio ) . we generically find two solutions of the cavity recursion , that possibly coincide depending on the parameters values . this correspond to two distinct phases of the statistical mechanics models , and also have a useful algorithmic interpretation , which will be spelled out in detail in section [ sec : treeinterpretation ] . initializing the recursion with the ` exact solution ' of the reconstruction problem ( ` plus ' initialization ) , we converges to a ferromagnetic fixed point . this provides an upper bound on the performance of any reconstruction algorithm . initializing the recursion with a completely oblivious initialization ( ` free ' initialization ) , we converge to a paramagnetic fixed point . this also corresponds to the performance of the best possible local algorithm ( see next section for a formal definition ) . a very similar qualitative picture is found in other inference problems on random graphs , one early example being the analysis of sparse graph codes @xcite . an important simplification is that we do not expect replica - symmetry breaking in these models @xcite . depending on the model parameters , we encounter two types of behaviors as @xmath131 increases . * for large @xmath54 or small @xmath130 , the two fixed points mentioned above coincide for all @xmath131 and no phase transition takes place . * for small @xmath54 and large @xmath130 , two phase transitions take place : a static phase transition at @xmath132 and a dynamic phase transition at a larger value @xmath133 . in addition , a spinodal point occurs at @xmath134 . + for @xmath135 the two fixed point above coincide , and yield bad reconstruction . for @xmath136 they coincide and yield good reconstruction . in the intermediate phase @xmath137 , the two fixed points do not coincide . the relevant fixed point for bayes - optimal reconstruction corresponds to the one of smaller free energy , and the transition between the two takes place at @xmath138 . the reader might consult fig . [ fig : phasediaglargedeg ] for an illustration . also , a very similar phase diagram was obtained in the related problem of sparse principal component analysis in @xcite . fixing @xmath43 , let @xmath139 ( respectively @xmath140 ) be the law of @xmath53 subject to @xmath29 containing ( respectively not containing ) vertex @xmath43 . consider the random variable @xmath141 the likelihood ratio test ( maximizing @xmath142 ) amounts to choosing . this is achieved by setting @xmath143 . ] @xmath144 as @xmath35 , the distribution of @xmath145 under @xmath139 converges to the law of a certain random variable @xmath146 , and the distribution of @xmath145 under @xmath140 converges instead to @xmath147 . the cavity method allows to write fixed point equations for these limit distributions . we omit details of the derivation since they are straightforward given the model ( [ eq : approxmodel ] ) , and since it is sufficient here to consider the replica - symmetric version of the method . general derivations can be found in ( * ? ? ? * chapter 14 ) . a closely related calculation is carried out in @xcite , which studies a more general random graph model , the so - called stochastic block model . the distribution of @xmath146 , @xmath147 are fixed point of the following recursion ( the symbol @xmath148 means that the distributions of quantities on the two sides are equal ) @xmath149 here @xmath150 are independent copies of @xmath151 . further @xmath152 , @xmath153 , @xmath154 , @xmath155 , are independent poisson random variables , independent of the @xmath156 . finally , @xmath157 and the function @xmath158 is given by @xmath159 ( recall that @xmath160 , cf . ( [ eq : rhodef ] ) . ) the cavity method predicts that the asymptotic distribution of @xmath145 ( conditional to @xmath38 or @xmath121 ) is a fixed point of eqs . ( [ eq : generalcavity1 ] ) , ( [ eq : generalcavity2 ] ) . in order to find the fixed points , we iterate these distributional equations with two types of initial conditions ( that correspond , respectively , to the poor reconstruction and good reconstruction phases ) @xmath161 we refer to section [ sec : treeinterpretation ] for the interpretation and monotonicity properties of these conditions : in particular it can be proved that @xmath151 converge in distribution if initialized in this manner . we implemented eqs . ( [ eq : generalcavity1 ] ) , ( [ eq : generalcavity2 ] ) numerically using the ` population dynamics ' method in the sampled density evolution . this was done using the method of @xcite . ] of @xcite ( also known as ` sampled density evolution ' @xcite ) . ( left ) , @xmath162 ( right ) and @xmath163 ( corresponding to average degree outside the set @xmath29 , @xmath164 ) . red curves correspond @xmath165 ( i.e. free boundary / initial conditions ) , and provide to the optimal performance of local algorithms . blue curves yield @xmath166 ( i.e. plus boundary / initial conditions ) and yield an upper bound on the performance of any algorithm . the continuous black line at @xmath167 coincides with the phase transition of bayes - optimal estimation . these curves were computed by averaging over @xmath168 runs of the population dynamics algorithm with @xmath169 samples and @xmath170 iterations.,title="fig : " ] ( left ) , @xmath162 ( right ) and @xmath163 ( corresponding to average degree outside the set @xmath29 , @xmath164 ) . red curves correspond @xmath165 ( i.e. free boundary / initial conditions ) , and provide to the optimal performance of local algorithms . blue curves yield @xmath166 ( i.e. plus boundary / initial conditions ) and yield an upper bound on the performance of any algorithm . the continuous black line at @xmath167 coincides with the phase transition of bayes - optimal estimation . these curves were computed by averaging over @xmath168 runs of the population dynamics algorithm with @xmath169 samples and @xmath170 iterations.,title="fig : " ] ( -385,23)@xmath171 ( -365,23)@xmath138 ( -345,23)@xmath172 in figure [ fig : sparsepsucc ] , we plot the predicted behavior of @xmath142 for @xmath163 and two different values of the clique size : @xmath173 . the success probability is predicted to be ( for @xmath35 ) @xmath174 we denote by @xmath175 and @xmath166 the predictions obtained with the two initializations above . as anticipated two behaviors can be observed . for @xmath54 sufficiently large , the curves @xmath165 and @xmath166 coincide for all @xmath131 . when this happens , this is also the success probability of the optimal likelihood ratio test @xmath176 , and the latter can be effectively approximated using a local algorithm ( e.g. belief propagation ) , see section [ sec : treeinterpretation ] . for @xmath54 small the two curves remain distinct in an intermediate interval of values : @xmath177 . in this regime , the asymptotic behavior of the bayes - optimal test is captured by the fixed point that yields the lowest free energy . it is convenient to define the rescaled free energy density as follows ( assuming that the limit exists ) @xmath178 the reason for this choice of the additive constants is that the resulting free energy is also equal to the asymptotic mutual information between the hidden set @xmath29 and the observed graph @xmath53 @xmath179 this quantity has therefore an immediate interpretation and several useful properties . the replica symmetric cavity method ( equivalently , bethe - peierls approximation ) predicts @xmath180 where the supremum is over all probability distributions @xmath181 over the real line satisfying the following symmetry property ( see section [ sec : treeinterpretation ] for further clarification on this property ) : @xmath182 the functional @xmath183 is defined as follows @xmath184 here expectation is taken with respect to the following independent random variables : * @xmath185 that are i.i.d . random variables with distribution @xmath186 ; * @xmath187 that are i.i.d . random variables with distribution @xmath188 ; * @xmath189 with joint distribution @xmath190 , @xmath191 , @xmath192 , where @xmath193 . * @xmath194 with the following mixture distribution . with probability @xmath54 : @xmath195 , @xmath196 . with probability @xmath197 : @xmath195 , @xmath198 . . the horizontal line corresponds to the maximal mutual information @xmath199 . , the vertical line at @xmath167 corresponds to the phase transition of the bayes - optimal estimator . this curve was computed by averaging over @xmath168 runs of the population dynamics algorithm with @xmath169 samples and @xmath170 iterations.,title="fig : " ] ( -145,23)@xmath138 let @xmath200 and @xmath201 the distributions of the fixed points obtained with plus and free initial conditions . in figure [ fig : sparsefree ] we plot the minimum of the corresponding bethe free energies @xmath202 and @xmath203 for @xmath163 , @xmath204 ( as obtained by the population dynamics algorithm ) . this is the cavity prediction for the free energy density @xmath205 . the value of @xmath131 for which @xmath206 corresponds to the phase transition point @xmath138 between paramagnetic and ferromagnetic phases from the reconstruction point of view , this is the phase transition for bayes - optimal estimation : @xmath207 notice from figure [ fig : sparsefree ] that as expected @xmath208 is monotone increasing in the signal - to - noise ratio @xmath131 , with @xmath209 as @xmath210 , and @xmath211 as @xmath212 ( here @xmath213 is the entropy of a bernoulli random variable with mean @xmath54 ) . also , the curve fig . [ fig : sparsefree ] presents some ` wiggles ' at large @xmath54 that are due to the limited numerical accuracy of the population dynamics algorithm . in the previous section we solved numerically the distributional equations ( [ eq : generalcavity1 ] ) , ( [ eq : generalcavity2 ] ) . this approach is somewhat laborious and its accuracy is limited . asymptotic expansions provide complementary analytical insights into the solution of these equations . with @xmath131 and @xmath54 fixed . here @xmath214 . left frame : success probability for free boundary condition ( equivalently , local algorithms , red curve ) , and plus boundary condition ( equivalently , general upper bound , blue curve ) . center frame : free energy ( equivalently , mutual information per vertex ) with same boundary conditions . right frame : zoom of the free energy curves.,title="fig : " ] with @xmath131 and @xmath54 fixed . here left frame : success probability for free boundary condition ( equivalently , local algorithms , red curve ) , and plus boundary condition ( equivalently , general upper bound , blue curve ) . center frame : free energy ( equivalently , mutual information per vertex ) with same boundary conditions . right frame : zoom of the free energy curves.,title="fig : " ] with @xmath131 and @xmath54 fixed . here left frame : success probability for free boundary condition ( equivalently , local algorithms , red curve ) , and plus boundary condition ( equivalently , general upper bound , blue curve ) . center frame : free energy ( equivalently , mutual information per vertex ) with same boundary conditions . right frame : zoom of the free energy curves.,title="fig : " ] ( -450,17)@xmath171 ( -423,17)@xmath138 ( -380,17)@xmath172 ( -292,17)@xmath171 ( -270,17)@xmath138 ( -233,17)@xmath172 ( -120,17)@xmath171 ( -80,17)@xmath138 here we consider @xmath215 with @xmath54 fixed , and @xmath216 converging to a limit . in particular , the signal - to - noise ratio @xmath131 is also a constant . let us emphasize once more that these limits are taken _ @xmath35 and hence the graph is still sparse . with @xmath131 and @xmath54 fixed . here we plot the ( shifted ) free energy function @xmath217 for @xmath214 and @xmath218 . comparing with figure [ fig : imu ] we see that @xmath219 , @xmath220 , @xmath221 . ] in this limit , the fixed points of eqs . ( [ eq : generalcavity1 ] ) , ( [ eq : generalcavity2 ] ) take the form @xmath222 further @xmath78 satisfies the fixed point equation @xmath223 where the function @xmath224 is defined by @xmath225 with expectation being taken with respect to @xmath226 . in other words , the distributional equations ( [ eq : generalcavity1 ] ) , ( [ eq : generalcavity2 ] ) reduced to a single nonlinear equation for the scalar @xmath78 . large @xmath78 correspond to accurate recovery . more formally , we expect the distributional solutions of eqs . ( [ eq : generalcavity1 ] ) , ( [ eq : generalcavity2 ] ) to converge to solutions of eqs . ( [ normalapprox1 ] ) to ( [ eq : cavitylargedegreelast ] ) . we do not provide a ` physicists ' derivation of this statement since this follows heuristically does not prove rigorously that the fixed points of eqs . ( [ eq : generalcavity1 ] ) , ( [ eq : generalcavity2 ] ) converge to fixed points of eqs . ( [ normalapprox1 ] ) to ( [ eq : cavitylargedegreelast ] ) . a complete proof would require controlling the convergence rate to fixed points . however in heuristic statistical physics derivation this is typically not done . also , the proof of lemma [ lemma : gaussianapprox ] follows the same strategy that would be employed in a heuristic derivation . ] from lemma [ lemma : gaussianapprox ] . the latter establishes that , iterating the cavity equations eqs . ( [ eq : generalcavity1 ] ) , ( [ eq : generalcavity2 ] ) any fixed number of times @xmath227 is equivalent ( in the large - degree limit ) to iterating eqs . ( [ normalapprox1 ] ) to ( [ eq : cavitylargedegreelast ] ) . , with @xmath228 ( relative size of the hidden set ) and @xmath229 ( signal - to - noise ratio ) fixed . the three curves are , from top to bottom @xmath230 , @xmath231 and @xmath232 . ] the free energy ( [ eq : freeenergy ] ) becomes a function of @xmath78 ( we still denote it by @xmath183 with a slight abuse of notation ) : @xmath233 where expectation is with respect to independent random variables @xmath234 and @xmath226 . its local minima are solutions of eq . ( [ eq : cavitylargedegree ] ) . equation ( [ eq : cavitylargedegree ] ) can be easily solved numerically , yielding the phase diagram in figure [ fig : phasediaglargedeg ] . as before , we obtain phase transitions @xmath235 as long as @xmath54 is below a critical point @xmath236 . the critical point location is @xmath237 the free energy @xmath238 has two local minima @xmath239 for @xmath236 , @xmath240 , and one local minimum otherwise . the local minimum @xmath241 is the global minimum for @xmath242 , while @xmath243 is the global minimum for @xmath244 . we refer to figures [ fig : imu ] and [ fig : psimu ] for illustration . of particular interest is the case of small hidden subsets , i.e. the limit @xmath40 ( note that @xmath88 is still linear in @xmath8 ) . for small @xmath54 we have @xmath245 . hence the solutions ( [ eq : cavitylargedegree ] ) that stay bounded converges to the solution of @xmath246 this equation has two solutions for @xmath247 and no solution for @xmath248 . this implies that @xmath249 which is the result announced in eq . ( [ eq : basiccondition ] ) . it is also easy to see that @xmath250 as @xmath40 . the distributional equations ( [ eq : generalcavity1 ] ) and ( [ eq : generalcavity2 ] ) define a sequence of probability distributions indexed by @xmath251 . more precisely , for every @xmath227 the recursion defines the probability distributions @xmath252 ( the distribution of @xmath253 ) and @xmath254 ( the distribution of @xmath255 ) . when specialized to the free / plus initial conditions ( cf . ( [ eq : freebc ] ) , ( [ eq : freebc ] ) ) , these probability distributions have a simple and useful interpretation that we will now explain . define @xmath256 to be the ball of radius @xmath227 centered at @xmath257 , in graph @xmath53 . namely , this is the subset of vertices of @xmath53 whose distance from @xmath43 is at most @xmath227 . by a slight abuse of notation , this will also denote the subgraph induced in @xmath53 by those vertices . the following remarks are straightforward . * free boundary condition . * consider the optimal test @xmath258 among those that only use local information . in other words , @xmath258 is the optimal test that is a function of @xmath256 . this is again a likelihood ratio test . concretely , we can define the log - likelihood ratio @xmath259 then the optimal test takes the form @xmath260)$ ] ( if we are interested in maximizing @xmath261 ) or @xmath262 ( if we are interested in minimizing the expected number of incorrectly assigned vertices ) . fixing the depth parameter @xmath227 , the distribution of @xmath263 converges ( as @xmath264 ) to @xmath265 for @xmath38 , and to @xmath266 for @xmath121 . mathematically , for any fixed @xmath43 @xmath267 in particular , for any fixed @xmath227 , the success probability @xmath268\big ) + \p_1^{(t),\fr}\big(\xi\ge \log[\kappa/(1-\kappa)]\big)-1\ , \label{eq : psuccfree } \ ] ] is the maximum asymptotic success probability achieved by any test that is @xmath227-local ( in the sense of being a function of depth-@xmath227 neighborhoods ) . it follows immediately from the definition that @xmath269 is monotone increasing in @xmath227 . its @xmath270 limit @xmath165 is the maximum success probability achieved by any local algorithm . this quantity was computed through population dynamics in the previous section , see figure [ fig : sparsepsucc ] . * plus boundary condition . * let @xmath271 be the complement of @xmath256 , i.e. the set of vertices of @xmath53 that have distance at least @xmath227 from @xmath43 . then @xmath272 has the interpretation of being the log - likelihood ratio , when information is revealed about the labels of vertices in @xmath273 . namely , if we define @xmath274 then we have @xmath275 in particular , @xmath166 is an upper bound on the performance of any estimator . in the previous section we computed this quantity numerically through population dynamics . let us finally comment on the relation ( [ eq : symmetry ] ) between @xmath186 and @xmath188 . this is an elementary consequence of bayes formula : consequences of this relation have been useful in statistical physics under the name of ` nishimori property ' @xcite . it is also known in coding theory as ` symmetry condition ' @xcite . consider the general setting of two random variables @xmath276 , with @xmath277 , @xmath278 , and let @xmath279 $ ] . then for any interval @xmath70 ( with non - zero probability ) , applying bayes formula , @xmath280 which is the claimed property . in the previous section we relied on the non - rigorous cavity method from spin glass theory to derive the phase diagram . most notably we used numerical methods , and formal large - degree asymptotics to study the distributional equations ( [ eq : generalcavity1 ] ) , ( [ eq : generalcavity2 ] ) . here we will establish rigorously some key implications of the phase diagram , namely : * by exhaustive search over all subsets of @xmath9 vertices in @xmath53 , we can estimate @xmath29 accurately for any @xmath281 and @xmath54 small . * local algorithms succeed in reconstructing accurately @xmath29 if @xmath248 , and fail for @xmath247 ( assuming large degrees and @xmath54 small ) . given a set of vertices @xmath282 $ ] , we let @xmath283 denote the number of edges with both endpoints in @xmath284 . exhaustive search maximizes this quantity among all the sets that have the ` right size . ' namely , it outputs @xmath285}\big\{\ , e(r)\ , : |r| = \lfloor\kappa n \rfloor\big\}\ , . \ ] ] ( if multiple maximizers exist , one of them is selected arbitrarily . ) we can also define a test function @xmath258 by letting @xmath286 for @xmath287 and @xmath288 otherwise . note that , for @xmath289 growing with @xmath117 , this algorithm is non - polynomial and hence can not be used in practice . it provides however a useful benchmark .. we have the following result showing that exhaustive search reconstructs @xmath29 accurately , for any constant @xmath131 and @xmath54 small . we refer to section [ sec : proofexhaustive ] for a proof . [ propo : exhaustive ] let @xmath290 be the asymptotic success probability of exhaustive search and assume @xmath291 . then @xmath292 in particular , we have the following large degree asymptotics as @xmath293 with @xmath294 fixed @xmath295 and @xmath296 as @xmath40 for any @xmath281 fixed . we next give a formal definition of @xmath227-local algorithms . let @xmath297 is the space of unlabeled rooted graphs , i.e. the space of graphs with one distinguished vertex ( see for instance @xcite for more details ) . formally , an estimator @xmath258 for the hidden set problem is a function @xmath298 . since the pair @xmath299 is indeed a graph with one distinguished vertex ( and the vertices labels clearly do not matter ) , we can view @xmath116 as a function on @xmath297 : @xmath300 the following definition formalizes the discussion in section [ sec : treeinterpretation ] ( where the definition of @xmath256 is also given ) . the key fact about this definition is that @xmath227 ( the ` locality radius ' ) is kept fixed , while the graph size can be arbitrarily large . given a non - negative integer @xmath227 , we say that a test @xmath116 is _ @xmath227-local _ if there exists a function @xmath301 such that , for all @xmath302 , @xmath303 we say that a test is local , if it is @xmath227-local for some fixed @xmath227 . we denote by @xmath304 and @xmath305 the sets of @xmath227-local and local tests . the next lemma is a well - known fact that we nevertheless state explicitly to formalize some of the remarks of section [ sec : treeinterpretation ] . recall that @xmath261 denotes the success probability of test @xmath116 , as per eq . ( [ eq : psucct ] ) , and let @xmath269 be defined as in eq . ( [ eq : psuccfree ] ) , with @xmath265 , @xmath266 , the laws of random variables @xmath306 , @xmath307 . we have @xmath308 in particular @xmath309 further , the maximal local success probability @xmath269 can be achieved using belief propagation with respect to the graphical model ( [ eq : approxmodel ] ) in @xmath310 time . we will therefore valuate the fundamental limits of local algorithms by analyzing the quantity @xmath165 . the following theorem establishes a phase transition for this quantity at @xmath311 . [ thm : main ] consider the hidden set problem with parameters @xmath36 , and let @xmath312 . then : * if @xmath247 , then all local algorithms have success probability uniformly bounded away from one . in particular , letting @xmath313 to be the smallest positive solution of @xmath314 , we have @xmath315 * if @xmath248 , then local algorithms can have success probability arbitrarily close to one . in particular , considering the large degree asymptotics @xmath215 with @xmath316 fixed @xmath317 we have @xmath318 as a useful technical tool in proving part @xmath319 of this theorem , we establish a normal approximation result in the spirit of eqs . ( [ normalapprox1 ] ) , ( [ normalapprox2 ] ) . in order to state this result , we recall the definition of wasserstein distance of order @xmath320 , @xmath321 between two probability measures @xmath78 , @xmath322 on @xmath323 , with finite second moment @xmath324 , @xmath325 . namely , denoting by @xmath326 the family of couplings if it is a probability distribution on @xmath327 such that @xmath328 and @xmath329 for all @xmath70 . ] of @xmath78 and @xmath322 , we have @xmath330 given a sequence of probability measures @xmath331 with finite second moment , we write @xmath332 if @xmath333 . [ lemma : gaussianapprox ] for @xmath334 , let @xmath335 be the random variables defined by the distributional recursion ( [ eq : generalcavity1 ] ) , ( [ eq : generalcavity2 ] ) , with initial condition ( [ eq : freebc ] ) , and denote by @xmath265 , @xmath266 the corresponding laws . further let @xmath336 be defined recursively by letting @xmath337 and @xmath338 then , considering the limit @xmath215 with @xmath54 fixed and @xmath339 , we have @xmath340 the proof of this lemma is presented in section [ sec : proofgaussian ] . as mentioned in the introduction , the problem of identifying a highly connected subgraph in an otherwise random graph has been studied across multiple communities . within statistical theory , arias - castro and verzelen @xcite established necessary and sufficient conditions for distinguishing a purely random graph , from one with a hidden community . with the scaling adopted in our paper , this ` hypothesis testing ' problem requires to distinguish between the following two hypotheses : @xmath341 note that this problem is trivial in the present regime and can be solved for instance by counting the number of edges in @xmath53 . the sparse graph regime studied in the present paper was also recently considered in a series of papers that analyzes community detection problems using ideas from statistical physics @xcite . the focus of these papers is on a setting whereby the graph @xmath53 contains @xmath342 non - overlapping communities , each of equal size @xmath343 . using our notation , vertices within the same community are connected with probability @xmath31 and vertices belonging to different communities are connected with probability @xmath32 . interestingly , the results of @xcite point at a similar phenomenon as the one studied here for @xmath344 . namely , for a range of parameters the community structure can be identified by exhaustive search , but low complexity algorithms appear to fail . let us mention that the very same phase transition structure arises in other inference problem , for instance in decoding sparse graph error correcting codes , or solving planted constraint satisfaction problems @xcite . a unified formalism for all of these problems is adopted in @xcite . all of these problems present a regime of model parameters whereby a large gap separates the optimal estimation accuracy , from the optimal accuracy achieved by known polynomial time algorithms . establishing that such a gap can not be closed under standard complexity - theoretic assumptions is an outstanding challenge . ( see @xcite for partial evidence in this direction albeit in a different regime . ) one can nevertheless gain useful insight by studying classes of algorithms with increasing sophistication . local algorithms : : are a natural starting point for sparse graph problems . the problem of finding a large independent set in a sparse random graph is closely related to the one studied here . indeed an independent set can be viewed as a subset of vertices that is ` less - connected ' than the background ( indeed is a subset of vertices such that the induced subgraph has no edge ) . + the largest independent set in a uniformly random regular graph with @xmath8 vertices of degree @xmath345 has typical size @xmath346 where , for large bounded degree @xmath345 , @xmath347 . hatami , lovsz and szegedy @xcite conjectured that local algorithms can find independent sets of almost maximum size up to sublinear terms in @xmath8 . gamarnik and sudan @xcite recently disproved this conjectured and demonstrated a constant multiplicative gap for local algorithms . roughly speaking , for large degrees no local algorithm can produce an independent set of size larger than @xmath348 of the optimum . this factor of @xmath348 was later improved by rahman and virag @xcite to @xmath349 . this gap is analogous to the gap in estimation error established in the present paper . we refer to @xcite for a broader review of this line of work . + as mentioned before , belief propagation ( when run for an arbitrary _ fixed _ number of iterations ) is a special type of local algorithm . further it is basically optimal ( among local algorithms ) for bayes estimation on locally tree like graphs . the gap between belief propagation decoding and optimal decoding is well studied in the context of coding @xcite . spectral algorithms . : : let @xmath350 be the adjacency matrix of the graph @xmath351 ( for simplicity we set @xmath352 for @xmath38 , and @xmath353 for @xmath121 ) . we then have @xmath354 this suggests that the principal eigenvector of @xmath355 should be localized on the set @xmath29 . indeed this approach succeeds in the dense case ( degree of order @xmath117 ) , allowing to reconstruct @xmath29 with high probability @xcite . + in the sparse graph setting considered here , the approach fails because the operator norm @xmath356 is unbounded as @xmath35 . concretely , the sparse graph @xmath351 has large eigenvalues of order @xmath357 localized on the vertices of largest degree . this point was already discussed in several related problems @xcite . several techniques have been proposed to address this problem , the crudest one being to remove high - degree vertices . + we do not expect spectral techniques to overcome the limitations of local algorithms in the present problem , even in their advanced forms that take into account degree heterogeneity . evidence for this claim is provided by studying the dense graph case , in which degree heterogeneity does not pose problems . in that case spectral techniques are known to fail for @xmath358 @xcite , and hence are strictly inferior to ( local ) message passing algorithms that succeed in the present paper correspond to @xmath359 in ( * ? ? ? * ; * ? ? ? ] for any @xmath248 . semidefinite relaxations . : : convex relaxations provide a natural class of polynomial time algorithms that are more powerful than spectral approaches . feige and krauthgamer @xcite studied the lovsz - schrijver hierarchy of semidefinite programming ( sdp ) relaxations for the hidden clique problem . in that setting , each round of the hierarchy yields a constant factor improvement in clique size , at the price of increasing complexity . it would be interesting to extend their analysis to the sparse regime . it is unclear whether sdp hierarchies are more powerful than simple local algorithms in this case . let us finally mention that the probability measure ( [ eq : approxmodel ] ) can be interpreted as the boltzmann distribution for a system of @xmath30 particles on the graph @xmath53 , with fugacity @xmath360 , and interacting attractively ( for @xmath361 ) . statistical mechanics analogies were previously exploited in @xcite . ( see also @xcite for the general community detection problem . ) i am grateful to yash deshpande for carefully reading this manuscript and providing valuable feedback . this work was partially supported by the nsf grants ccf-1319979 and dms-1106627 , and the grant afosr fa9550 - 13 - 1 - 0036 . for the sake of simplicity , we shall assume a slightly modified model whereby the hidden set @xmath29 is uniformly random with size @xmath362 , with @xmath363 . recall that , under the independent model ( [ eq : kappadef ] ) @xmath364 and hence is tightly concentrated around its mean @xmath289 . hence , the result the independent model follows by a simple conditioning argument . let @xmath365 . by exchangeability of the graph vertices , we have @xmath366 where the last inequality follows since , without loss of generality , @xmath367 . setting @xmath368 , we will prove that for any @xmath369 there exists @xmath370 such that @xmath371 the claim the follows by using the inequality ( [ eq : proofpropo - simple ] ) together with the fact that @xmath372 . for two sets @xmath70 , @xmath373 $ ] , we let @xmath374 the number of edges @xmath76 such that @xmath375 , but @xmath376 . in order to prove eq . ( [ eq : basicproofpropo ] ) note that , for @xmath377 @xmath378 to see this notice that , by definition , if @xmath379 then @xmath380 . this mean that there must exists at least one set @xmath381 $ ] satisfying the following conditions : * @xmath382 . * @xmath383 . * @xmath384 . indeed @xmath385 is such a set . this immediately implies eq . ( [ eq : tobeexplained ] ) by noticing that @xmath386 and @xmath387 . by a union bound ( setting @xmath388 ) : @xmath389 in the last inequality we used union bound and the fact that edges contributing to @xmath390 and @xmath391 are independent . using chernoff bound on the tail of binomial random variables ( with @xmath392 the kullback - leibler divergence between two bernoulli random variables ) , we get @xmath393\cap \naturals } \prob\big(\binom(m;a / n)\le j\big ) \,\,\prob\big(\binom(m;b / n)\ge j\big)\\ & \le ( m+1)\binom{k}{\ell}\binom{n - k}{k-\ell}\ , \exp\big\{-m\ , \min_{j\in [ bm / n , am / n ] } \big[d(j / m|| a / n)+d(j / m|| b / n)\big]\big\ } , . \label{eq : boundwkl } \ ] ] here , the first inequality follows because both probabilities are increasing for @xmath394 and decreasing for @xmath395 . we further note that , @xmath396 and therefore , for @xmath397 $ ] , @xmath398 this implies that , for @xmath399 , we have @xmath400 } \big[d(x||p_1)+d(x||p_2)\big]&\ge \frac{1}{2}\big(\frac{1}{p_2}+1\big)\ , \min_{x\in [ p_1,p_2 ] } \big[(x - p_1)^2+(x - p_2)^2\big]\\ & \ge \frac{1}{4}\big(\frac{1}{p_2}+1\big)\ , ( p_1-p_2)^2\ , . \ ] ] we substitute the last inequality in eq . ( [ eq : boundwkl ] ) , together with the bounds @xmath401 $ ] @xmath402 we let @xmath403 whence @xmath404 we therefore get @xmath405 for @xmath406 , the argument in parenthesis is smaller than @xmath407 and therefore @xmath408 summing over @xmath409 , we get @xmath410 which implies the claim ( [ eq : basicproofpropo ] ) , after eventually adjusting @xmath411 , since @xmath412 . recall that convergence in @xmath415 distance is equivalent to weak convergence , plus convergence of the first two moments ( * ? ? ? * theorem 6.9 ) . we will prove by the following by induction over @xmath227 : these claims obviously hold for @xmath423 . next assuming that they hold up to iteration @xmath227 , we need to prove them for iteration @xmath424 . for the sake of brevity , we will only present this calculation for @xmath425 , since the derivation for @xmath426 is completely analogous . let us start by considering eq . ( [ eq : limitexpxi0 ] ) . first notice that the absolute value of right - hand side of eq . ( [ eq : generalcavity1 ] ) is upped bounded by @xmath427 and hence @xmath428 follows from the induction hypothesis i@xmath429 and the fact that @xmath430 are poisson . next to prove eq . ( [ eq : limitexpxi0 ] ) , we take expectation of eq . ( [ eq : generalcavity1 ] ) , and let , for simplicity , @xmath431 : @xmath432 where the last equality follows from bounded convergence , since , for all @xmath433 , @xmath434 . note that the laws of @xmath253 and @xmath255 satisfy the symmetry property ( [ eq : symmetry ] ) . hence , for any measurable function @xmath435 such that the expectations below make sense , we have @xmath436 in particular applying this identity to @xmath437 and @xmath438 ^ 2 $ ] , we get @xmath439 substituting in eq . ( [ eq : exiexpansion ] ) , and expressing @xmath440 in terms of @xmath441 we get @xmath442 where @xmath443 denotes a quantity vanishing as @xmath444 . the last equality follows from induction hypothesis iii@xmath429 and the fact that @xmath445 is bounded continuous , with @xmath226 . this yields the desired claim ( [ eq : limitexpxi0 ] ) after comparing with eq . ( [ eq : mutrecursion ] ) . consider next eq . ( [ eq : limitvarxi0 ] ) . the upper bound on the right - hand side of eq . ( [ eq : generalcavity1 ] ) given by eq . ( [ eq : upperboundrhs ] ) immediately imply that @xmath446 . in order to estabilish eq . ( [ eq : limitvarxi0 ] ) , we recall an elementary formula for the variance of a poisson sum . if @xmath447 is a poisson random variable and @xmath448 are i.i.d . with finite second moment , then @xmath449 applying this to eq . ( [ eq : generalcavity1 ] ) , and expanding for large @xmath130 thanks to the bounded convergence theorem , we get @xmath450 ^ 2\right\}+\kappa b \ , \e\left\{\big[\log\left(1+(\rho-1)\frac{e^{\xi_1^{(t ) } } } { 1+e^{\xi_1^{(t)}}}\right)\big]^2\right\}\\ & = ( 1-\kappa ) \frac{(a - b)^2}{b } \e\left\{\left(\frac{e^{\xi_0^{(t)}}}{1+e^{\xi_0^{(t)}}}\right)^2\right\ } + \kappa\ , \frac{(a - b)^2}{b}\e\left\{\left(\frac{e^{\xi_1^{(t)}}}{1+e^{\xi_1^{(t)}}}\right)^2\right\ } + o(b^{-1/2})\\ & = \kappa \frac{(a - b)^2}{b}\e\left(\frac{e^{\xi_1^{(t)}}}{1+e^{\xi_1^{(t)}}}\right ) + o(b^{-1/2})\ , , \ ] ] where the last equality follows by applying again eq . ( [ eq : secondidentity ] ) . by using the induction hypothesis iii@xmath319 and the fact that @xmath451 is bounded lipschitz , @xmath452 which is eq . ( [ eq : mutrecursion ] ) . we finally consider eq . ( [ eq : limitweakp0 ] ) . by subtracting the mean , we can rewrite eq . ( [ eq : generalcavity1 ] ) as @xmath453 where @xmath454 , @xmath455 . note that @xmath456 , @xmath457 have zero mean and , by the calculation above , they have variance @xmath458 . denoting the right hand side by @xmath459 : @xmath460 because ( for instance ) @xmath461 is a sum of order @xmath462 independent random variables with zero mean and variance of order @xmath463 . note that @xmath464 + \kappa b\e [ f(\xi^{(t)}_{1,1})^2 ] \big\ } = \mu^{(t+1)}\ , , \ ] ] where the last equality follows by the calculation above . hence , by applying the central limit theorem to each of the four terms in eq . ( [ eq : sb ] ) and noting that they are independent , we conclude that @xmath465 converges in distribution to @xmath466 . define the event @xmath468 , and write @xmath414 for @xmath469 . from eq . ( [ eq : psuccfree ] ) we have @xmath470 using eq . ( [ eq : symmetry ] ) , and the fact that @xmath471 , we get @xmath472 call @xmath473 . by the initialization ( [ eq : freebc ] ) , @xmath474 . taking exponential moments of eq . ( [ eq : generalcavity1 ] ) , we get @xmath475 + \kappa b \,\e\left[\left(\frac{1+\rho\ , e^{\xi_1^{(t ) } } } { 1 + e^{\xi_1^{(t ) } } } \right)^2\right ] \right\}\ , . \ ] ] note that by eq . ( [ eq : symmetry ] ) , for any measurable function @xmath435 such that the expectations below make sense , we have @xmath476 applying this to @xmath477 , we get @xmath478 \right\}\ , . \ ] ] now we claim that , for @xmath479 , we have @xmath480 this can be checked , for instance , by multiplying both sides by @xmath481 and simplifying . using @xmath482 and @xmath483 , we get @xmath484 let @xmath485 be the solution of the above recursion with equality , i.e. @xmath486 and @xmath487 it is a straightforward exercise to see that @xmath485 is monotone increasing in @xmath227 and @xmath131 . further , for @xmath488 , @xmath489 the smallest positive solution of @xmath490 , and @xmath491 . hence @xmath492 which , together with eq . ( [ eq : basictvbound ] ) finishes the proof . note that by monotonicity @xmath493 , and hence it is sufficient to lower bound the limit of the latter quantity . by lemma [ lemma : gaussianapprox ] , we have @xmath494 where @xmath495 is the gaussian distribution , and @xmath336 is defined recursively by eq . ( [ eq : mutrecursion ] ) with @xmath337 . hence for all @xmath334 @xmath496 it is therefore sufficient to prove that @xmath497 now by monotone convergence , we have @xmath498 further @xmath499 increases monotonically towards its limit as @xmath40 . furthermore , @xmath500 is increasing in @xmath78 for any fixed @xmath501 . by induction over @xmath227 we prove that @xmath502 ( the limit being monotone from below ) , where @xmath503 and for all @xmath334 @xmath504 in order to prove this claim , note that the base case of the induction is trivial and ( writing explicitly the dependence on @xmath54 @xmath505 on the other hand for a fixed @xmath506 @xmath507 the claim follows since @xmath508 can be taken arbitrarily small . noga alon , michael krivelevich , and benny sudakov , _ finding a large hidden clique in a random graph _ , proceedings of the ninth annual acm - siam symposium on discrete algorithms , society for industrial and applied mathematics , 1998 , pp . 594598 . emmanuel abbe and andrea montanari , _ conditional random fields , planted constraint satisfaction and entropy concentration _ , approximation , randomization , and combinatorial optimization . algorithms and techniques , springer , 2013 , pp . 332346 . dimitris achlioptas and federico ricci - tersenghi , _ on the solution - space geometry of random constraint satisfaction problems _ , proceedings of the thirty - eighth annual acm symposium on theory of computing , acm , 2006 , pp . 130139 . aurelien decelle , florent krzakala , cristopher moore , and lenka zdeborov , _ asymptotic analysis of the stochastic block model for modular networks and its algorithmic applications _ , physical review e * 84 * ( 2011 ) , no . 6 , 066106 . alexandre gaudillire , benedetto scoppola , elisabetta scoppola , and massimiliano viale , _ phase transitions for the cavity approach to the clique problem on random graphs _ , journal of statistical physics * 145 * ( 2011 ) , no . 5 , 11271155 . dandan hu , peter ronhovde , and zohar nussinov , _ phase transitions in random potts systems and the community detection problem : spin - glass type and dynamic perspectives _ , philosophical magazine * 92 * ( 2012 ) , no . 4 , 406445 . subhash khot , _ improved inapproximability results for maxclique , chromatic number and approximate graph coloring _ , foundations of computer science , 2001 . proceedings . 42nd ieee symposium on , ieee , 2001 , pp . 600609 . florent krzakala , cristopher moore , elchanan mossel , joe neeman , allan sly , lenka zdeborov , and pan zhang , _ spectral redemption in clustering sparse networks _ , proceedings of the national academy of sciences * 110 * ( 2013 ) , no . 52 , 2093520940 . andrea montanari , daniel reichman , and ofer zeitouni , _ on the limitation of spectral methods : from the gaussian hidden clique problem to rank one perturbations of gaussian tensors _ , arxiv:1411.6149 ( 2014 ) .

we study the thermal stability of the periodic ( p ) and quasi - periodic ( q ) precessional modes of the nanoparticle magnetic moment induced by a rotating magnetic field . an analytical method for determining the lifetime of the p mode in the case of high anisotropy barrier and small amplitudes of the rotating field is developed within the fokker - planck formalism . in general case , the thermal stability of both p and q modes is investigated by numerical simulation of the stochastic landau - lifshitz equation . we show analytically and numerically that the lifetime is a nonmonotonic function of the rotating field frequency which , depending on the direction of field rotation , has either a pronounced maximum or a deep minimum near the larmor frequency .

1108.3559

magnetic nanoparticles are of great interest because of their nanoscale physical properties and many current and potential applications . these applications range from high - density storage media@xcite and spintronic devices@xcite to biomedical applications like drug delivery , cell separation , cancer treatment and many others ( for a review , see refs . ) . since the physical properties of nanoparticles play a decisive role in all these applications , their study is of fundamental importance . in particular , for high - density storage media , e.g. , bit - patterned media@xcite where each nanoparticle is a carrier of information , the thermal stability of a given direction ( or magnitude ) of the nanoparticle magnetic moment is one of the most important problems . the reason is that under thermal fluctuations the magnetic moment can be randomly switched to a new state leading to the loss of information . in the case of ferromagnetic nanoparticles , the fluctuation dynamics of the nanoparticle magnetic moment can be described by the stochastic landau - lifshitz equation . if the noise term in this equation is approximated by the gaussian white noise , then the probability density of the magnetic moment satisfies the fokker - planck equation whose properties are well known.@xcite this approach , introduced by brown@xcite almost five decades ago , has become an important tool in the study of stochastic magnetic dynamics ( see ref . and references therein ) . to characterize the thermal stability in the case of uniaxial nanoparticles , it is often enough to determine the lifetimes of the nanoparticle magnetic moment in the up " and down " states . in the above approach , the lifetime in a given state is usually interpreted as the relaxation time . however , from a theoretical point of view , the lifetime is reasonable to associate with the mean first - passage time ( mfpt ) , i.e. , average time that a random process dwells in a prescribed state . an additional advantageous feature of this definition of the lifetime is that the mfpt method is mathematically well developed.@xcite this approach was first applied to study the magnetic relaxation in systems of noninteracting@xcite and dipolar interacting@xcite nanoparticles subjected to a constant magnetic field . in general , the lifetime depends on both intrinsic properties of nanoparticles and external magnetic fields . the case when the external fields contain a rotating magnetic field applied perpendicular to the easy axes of nanoparticles has a particular interest . on the one hand , this is because the rotating field plays a key role in the microwave - assisted switching@xcite and , on the other hand , because the corresponding dynamical equations ( without accounting the thermal fluctuations ) can often be solved analytically.@xcite specifically , it has been shown@xcite ( see also ref . ) that the rotating field can induce two types of the stable precessional modes of the magnetic moment , namely , the periodic ( p ) and quasi - periodic ( q ) modes . under certain conditions,@xcite one mode can exist in the up state of the magnetic moment and the other in the down state . the thermal fluctuations make the random transitions between these modes possible , and the problem of the lifetime of the p and q modes appears . some aspects of this problem have already been considered previously in the context of magnetic relaxation and induced magnetization.@xcite but the dependence of the lifetime on the parameters of the rotating field has not been studied systematically . at the same time , the effect of strong dependence of the lifetime on the rotating field frequency , which is expected to exist in the vicinity of the larmor frequency , could be important for applications . therefore , in this paper we present a detailed analytical and numerical analysis of the above mentioned problem . the paper is organized as follows . in sec . [ life ] , we describe the model and define the lifetime of both the p and q modes . here we also derive the boundary conditions and transformation properties of the lifetime . in sec . [ anres ] , we develop an analytical method for calculating the frequency dependence of the lifetime of the p mode in the case of high anisotropy barrier . our numerical results obtained by the simulation of the deterministic and stochastic landau - lifshitz equations are presented in sec . [ numres ] . specifically , the features of the p and q modes at zero temperature are studied in sec . [ prec ] and the effects of thermal fluctuations are considered in sec . [ simulat ] . finally , in sec . [ con ] we summarize our findings . to study the influence of the rotating magnetic field on the thermal stability of the nanoparticle magnetization , we use a minimal model with coherent spin dynamics . within this model , which is applicable to particles whose exchange energy is comparatively large , the magnetic state of each particle is completely characterized by the magnetic moment @xmath0 of a fixed magnitude @xmath1 . due to its interaction with a heat bath , @xmath2 is a vector random process which can be described by the stochastic landau - lifshitz equation@xcite @xmath3 where @xmath4 is the gyromagnetic ratio , @xmath5 is the damping parameter , the cross denotes the vector product , and @xmath6 and @xmath7 are the effective magnetic fields . the first field is given by @xmath8 , where @xmath9 is the magnetic energy of the particle which , in the case under consideration , contains only the uniaxial anisotropy energy @xmath10 and the zeeman energy @xmath11 , i.e. , @xmath12 . here , the @xmath13-axis of a cartesian coordinate system @xmath14 with unit vectors @xmath15 , @xmath16 and @xmath17 is chosen to be parallel to the easy axis of magnetization , @xmath18 is the anisotropy field , @xmath19 , the dot denotes the scalar product , and @xmath20 is the rotating magnetic field . we assume that @xmath20 is applied perpendicular to the @xmath13-axis , so that @xmath21 where @xmath22 is the field amplitude , @xmath23 is the angular rotation frequency , and @xmath24 ( for clockwise rotation ) or @xmath25 ( for counterclockwise rotation ) . it is this field that induces the precessional modes of @xmath2 . the effective field @xmath26 accounts for the thermal fluctuations . it is assumed that the cartesian components @xmath27 @xmath28 of @xmath26 are independent gaussian white noises with zero means and correlation functions @xmath29 . here , the angular brackets denote averaging over all realizations of @xmath26 , @xmath30 is the noise intensity , @xmath31 is the boltzmann constant , @xmath32 is the absolute temperature , and @xmath33 is the dirac @xmath34 function . in accordance with this definition of @xmath26 , the random process @xmath2 is markovian and can be described within the fokker - planck formalism . because eq . ( [ l - l ] ) preserves the length of the magnetic moment @xmath2 , it is convenient to write the fokker - planck equation that corresponds to eq . ( [ l - l ] ) in spherical coordinates . introducing the polar and azimuthal angles @xmath35 and @xmath36 of @xmath2 , the forward and backward fokker - planck equations for the conditional probability density @xmath37 ( @xmath38 ) can be written as@xcite @xmath39 & \displaystyle -\frac{2a}{\lambda } \frac{\partial } { \partial\psi } [ v(\theta,\psi ) - \rho \tilde{\omega}]p = \frac{2a}{\lambda } \frac{\partial p}{\partial \tilde{t } } & \label{f - pfw}\end{aligned}\ ] ] and @xmath40 & \displaystyle + \frac{2a}{\lambda } [ v(\theta',\psi ' ) - \rho \tilde{\omega}]\frac{\partial p}{\partial\psi ' } = -\frac{2a}{\lambda } \frac{\partial p}{\partial \tilde{t } ' } , & \label{f - pbw}\end{aligned}\ ] ] respectively . here , @xmath41 is the dimensionless time , @xmath42 is the larmor frequency , @xmath43 is the dimensionless frequency of the rotating field , and @xmath44 is a dimensionless parameter that characterizes the anisotropy barrier height in the units of the thermal energy @xmath45 . finally , the variables @xmath46 and @xmath47 are associated with @xmath48 and @xmath49 , respectively , and the functions @xmath50 and @xmath51 are expressed through the dimensionless magnetic energy @xmath52 ( @xmath53 ) as follows : @xmath54 v(\theta,\psi ) \!&=&\ ! \frac{1}{\sin^{2 } \theta } \bigg ( \sin \theta \frac{\partial}{\partial \theta } - \lambda \frac{\partial}{\partial \psi } \bigg ) \tilde{w } \nonumber\\ \!&=&\ ! \cos\theta - \tilde{h}\cot\theta\cos\psi - \lambda\tilde{h}\frac{\sin\psi}{\sin\theta}. \label{def uv}\end{aligned}\ ] ] it is assumed that the probability density @xmath55 satisfies the initial condition @xmath56 . moreover , if the absorbing boundary conditions are not imposed , then @xmath55 is properly normalized : @xmath57 . the system of two stochastic differential equations @xmath58 & \displaystyle \frac{d}{d\tilde{t}}\psi(\tilde{t } ) = v\big(\theta(\tilde{t } ) , \psi(\tilde{t})\big ) - \rho \tilde{\omega } + \sqrt{\frac{\lambda } { a}}\,\frac{1}{\sin \theta(\tilde{t } ) } \ , \eta_{2}(\tilde{t}),&\qquad \label{l - l2}\end{aligned}\ ] ] where @xmath59 @xmath60 denote independent gaussian white noises with zero means and correlation functions @xmath61 , leads to the same fokker - planck equation ( [ f - pfw]).@xcite this means that the above system is stochastically equivalent to the stochastic landau - lifshitz equation ( [ l - l ] ) . in what follows , we will use eq . ( [ l - l2 ] ) to numerically study the thermal stability of the precessional modes of the magnetic moment @xmath2 . at zero noise intensity , the rotating magnetic field @xmath20 can induce stable precessional modes of @xmath2 of two types.@xcite in the first , p mode , the precession angle @xmath62 is a constant and @xmath2 in the laboratory frame is a periodic function of time . in the second , q mode , the precession angle varies periodically and @xmath2 becomes a quasi - periodic function of time ( because the periods of @xmath62 and @xmath20 are in general not commensurable ) . some properties of these modes related to the steady - state will be considered in sec . [ prec ] . here , we use only the fact that , depending on the parameters @xmath63 and @xmath64 , one or two precessional modes may exist in steady state . in the latter case , one mode occurs in the up state ( @xmath65 ) and the other in the down state ( @xmath66 ) of @xmath2 . we assume that for a given set of the above parameters the magnetic moment @xmath2 is in the state @xmath67 if @xmath68 tends to @xmath69 as @xmath70 _ slowly _ decreases to zero . it should be noted that the last condition is important because a sharp decrease of @xmath70 can switch @xmath2 to another state . in steady - state , the reference modes are stable and transitions between them are impossible . however , these transitions can occur under thermal fluctuations . in this case the precessional modes become metastable and the magnetic moment remains in a given state @xmath67 for some ( dimensionless ) random time @xmath71 . the average value of this time , i.e. , the lifetime @xmath72 of the metastable state , can be determined using the mfpt method.@xcite the basis of this method is the backward fokker - planck equation ( [ f - pbw ] ) , which should be written for a given state @xmath67 . to this end , we add the index @xmath67 to all angle variables , replace the conditional probability density @xmath55 by @xmath73 , @xmath74 and @xmath75 . here , the angle @xmath76 ( @xmath77 ) should be chosen so that the time average of the precession angle @xmath78 [ we recall that @xmath79 for the p modes does not depend on @xmath80 is relatively close to @xmath76 and to @xmath81 for @xmath66 and + 1 , respectively . to meet these requirements , in our numerical simulations we assume that @xmath82 . in accordance with the mfpt approach , we consider a circular cone with the cone angle @xmath83 as the absorbing boundary for the magnetic moment in the @xmath67 state , i.e. , @xmath84 . in this case , taking into account that @xmath85 ( @xmath86 ) , the lifetime can be defined as @xmath87 , where @xmath88 is the probability that the magnetic moment stays in the state @xmath67 up to a given value of the time difference @xmath89 . finally , using the relations @xmath90 and @xmath91 is governed by the partial differential equation @xmath92 & \displaystyle + \frac{2 a}{\lambda } [ v(\theta ' _ { \sigma } , \psi'_{\sigma } ) - \rho \tilde{\omega } ] \frac{\partial \mathcal { t}_{\sigma } } { \partial\psi ' _ { \sigma } } = -\frac{2a}{\lambda}. & \label{mfpt1}\end{aligned}\ ] ] at the absorbing boundary , the solution of this equation must satisfy the condition @xmath93 one more important property of the lifetime is that it is a finite function of @xmath94 and @xmath95 . in order to prove this statement , let us first approximate the stochastic dynamics of the magnetic moment by a random walk on the sphere characterized by a dimensionless discrete time @xmath96 , where @xmath97 and @xmath98 is the time step . then , denoting @xmath99 the probability that the magnetic moment stays in the state @xmath67 after the @xmath100th step , we can write @xmath101 . if the maximal element of the set @xmath102 equals @xmath103 then @xmath104 and , as a consequence , @xmath105 . finally , taking into account that the condition @xmath106 implies that @xmath107 , we obtain the desired result : @xmath108 . it should also be noted that since the maximum angular distance to the absorbing boundary occurs at @xmath109 , i.e. , @xmath110 , the condition of finiteness of the lifetime can be written in the form @xmath111 the above result shows the importance of knowing the solution of eq . ( [ mfpt1 ] ) in a small vicinity of the point @xmath112 . assuming that @xmath113 ( @xmath114 ) , this equation at @xmath115 reduces to @xmath116 & \displaystyle + \lambda \sin \psi'_{\sigma } ) \frac{\partial \mathcal { t}_{\sigma } } { \partial \psi'_{\sigma } } = -\frac{2a}{\lambda}. & \label{aseq_mfpt1}\end{aligned}\ ] ] its solution can be represented as @xmath117 where @xmath118 and @xmath119 are constant parameters , and the functions @xmath120 satisfy the ordinary differential equations @xmath121 and @xmath122 with @xmath123 and @xmath124 being the kronecker delta . according to eq . ( [ sol_aseq ] ) , the finiteness condition ( [ fint ] ) takes the form @xmath125 , which in turn is equivalent to @xmath126 . we note that at @xmath125 the derivative @xmath127 in the point @xmath109 , in general , does not vanish : @xmath128 . however , if @xmath129 then @xmath130 ( this is so because in this case @xmath131 does not depend on @xmath132 ) , and the finiteness condition ( [ fint ] ) reduces to the reflecting boundary condition @xmath133 it should also be noted that at @xmath134 the same reflecting boundary condition holds for the average lifetime @xmath135 , since @xmath136 . now , using the general equation ( [ mfpt1 ] ) , we can establish the transformation properties of its solution @xmath137 ( for clarity , the dependence of @xmath131 on @xmath138 is shown explicitly ) , which is assumed to obey the conditions ( [ abst ] ) and ( [ fint ] ) . toward this end , let us introduce the change of variables @xmath139 taking into account that @xmath140 and @xmath141 , one can make sure that eq . ( [ mfpt1 ] ) in the new variables @xmath142 and @xmath143 becomes @xmath144 \displaystyle \times \frac{\partial \mathcal { h } _ { \sigma } } { \partial\theta'_{-\sigma } } + \frac{2 a } { \lambda } [ v(\theta'_{-\sigma } , \psi'_{-\sigma } ) + \rho \tilde{\omega } ] \frac{\partial \mathcal { h}_{\sigma } } { \partial \psi'_{-\sigma } } = -\frac{2a}{\lambda } , \end{array } \label{mfpt2}\ ] ] where @xmath145 . in accordance with the transforms ( [ change ] ) , the conditions ( [ abst ] ) and ( [ fint ] ) for @xmath146 take the form @xmath147 and @xmath148 , respectively . therefore , comparing eqs . ( [ mfpt1 ] ) and ( [ mfpt2 ] ) and the corresponding absorbing and finiteness conditions , one can conclude that @xmath149 is equal to @xmath150 , i.e. , @xmath151 for the average lifetime @xmath152 this transformation property reads : @xmath153 . in general , the arguments @xmath154 and @xmath155 of @xmath131 are arbitrary and can be properly chosen to best suit the problem . in this paper we are interested in the lifetime of the precessional modes _ reaching _ the steady state . therefore , the angles @xmath156 and @xmath155 should be associated with the solution of deterministic ( when @xmath157 ) landau - lifshitz equations ( [ l - l2 ] ) at some time @xmath158 , i.e. , @xmath159 and @xmath160 . to be sure that the magnetic moment is near the steady state , we assume that @xmath161 , where @xmath162 is the time of increasing the rotating field amplitude from @xmath163 to a given value @xmath70 , and @xmath164 is the relaxation time . it should be emphasized that the time @xmath165 must be chosen large enough to prevent the dynamical switching from the state @xmath67 to the state @xmath166 . in the p mode , the angles @xmath167 and @xmath168 approach the limiting precession angle @xmath169 and the limiting difference of phases @xmath170 , respectively . as a consequence , the lifetime of this mode @xmath171 does not depend on the precise choice of @xmath172 . in contrast , in the q mode the precession angle @xmath173 is a periodic function of time @xmath174 with a period @xmath175 and the difference of phases is given by @xmath176 , where @xmath177 and @xmath178 is also a periodic function with the same period @xmath179 ( see sec . [ prec ] ) . thus the lifetime of the q mode , in general , depends on @xmath180 : @xmath181 however , at @xmath182 this dependence is very weak and can be safely neglected ( see sec . [ simulat ] ) . in the case of p mode the limiting angles @xmath183 and @xmath184 depend also on the parameters @xmath70 , @xmath185 , @xmath64 and @xmath138 . however , for brevity , we keep only the parameter @xmath138 , i.e. , @xmath186 $ ] and @xmath187 $ ] , which together with the state parameter @xmath67 describes the transformation properties of these angles . to find them , we use eq . ( [ def uv ] ) for representing the stationary landau - lifshitz equations @xmath188 , \psi_{\sigma}[\rho])=0 $ ] and @xmath189 , \psi_{\sigma } [ \rho ] ) = \rho \tilde { \omega}$ ] as @xmath190 - \rho \kappa \tan\theta_{\sigma}[\rho ] ) , & \nonumber\\[3pt ] & \displaystyle \sin \psi_{\sigma}[\rho ] = - \frac{\rho \lambda \kappa}{\tilde{h } } \sin \theta_{\sigma}[\rho ] , & \label{st1}\end{aligned}\ ] ] where @xmath191 . from these equations , it is straightforward to obtain the desired result : @xmath192 = \pi - \theta_{-\sigma}[-\rho ] , \qquad \psi_{\sigma}[\rho ] = 2\pi - \psi_{-\sigma}[-\rho ] . \label{sym_sol}\ ] ] because the transformation properties ( [ sym_sol ] ) are similar to those in eq . ( [ change ] ) , from eq . ( [ transf1 ] ) one gets the transformation property of the lifetime of the p mode @xmath193,\psi _ { \sigma}[\rho];\rho ) = \mathcal { t}_{-\sigma } ( \pi - \theta_{\sigma}[\rho ] , 2\pi - \psi_{\sigma}[\rho];-\rho ) . \label{transf2}\ ] ] it shows that the lifetimes characterized by the pairs @xmath194 and @xmath195 are the same . the analytical determination of the lifetimes of the precessional modes implies the solution of eq . ( [ mfpt1 ] ) with the absorbing and finiteness conditions ( [ abst ] ) and ( [ fint ] ) . since the lifetime @xmath196 is a periodic function of @xmath155 with the period @xmath197 , it can be expressed as the fourier series @xmath198 to guarantee the reality of @xmath196 , we assume that the coefficients @xmath199 of the series satisfy the condition @xmath200 ( the asterisk denotes complex conjugation ) . substituting this series into eq . ( [ mfpt1 ] ) and introducing the differential operators @xmath201 & & \ ! + \ , i\frac{2an}{\lambda } ( \cos \theta'_{\sigma } - \rho \tilde{\omega } ) , \nonumber\\[4pt ] \hat{n}_{n } \!&=&\ ! ( \lambda \cos \theta'_{\sigma } - i ) \frac{d}{d\theta'_{\sigma } } + \frac{\lambda n } { \sin \theta'_{\sigma } } - i n \cot \theta'_{\sigma } , \label{l , n}\end{aligned}\ ] ] one obtains for @xmath202 an infinite set of coupled equations @xmath203 like @xmath204 , the coefficients @xmath205 must satisfy both the absorbing boundary condition @xmath206 and the finiteness condition @xmath207 . in the three - mode approximation , when @xmath208 for all @xmath209 , an infinite set of equations ( [ tn ] ) reduces to a set of three equations for @xmath210 , @xmath211 and @xmath212 . since @xmath204 is real , it is convenient to consider instead of @xmath213 and @xmath214 the real and imaginary parts of @xmath213 , i.e. , @xmath215 and @xmath216 . then , in this approximation , the lifetime ( [ solt ] ) is given by @xmath217 and , denoting the real and imaginary parts of an operator @xmath218 as @xmath219 and @xmath220 , respectively , from eqs . ( [ tn ] ) and ( [ l , n ] ) one obtains a set of three coupled equations for @xmath210 , @xmath221 and @xmath222 : @xmath223 & \displaystyle \hat{l}_{1}^{+ } \mathcal{t}_{\sigma 1}^{+ } - \hat{l}_{1}^{- } \mathcal{t}_{\sigma 1}^{- } = - \frac{a \tilde{h}}{\lambda } \hat{n}_{0}^{+ } \mathcal{t}_{\sigma 0 } , & \label{eqt+ } \\[4pt ] & \displaystyle \hat{l}_{1}^{+ } \mathcal{t}_{\sigma 1}^{- } + \hat{l}_{1}^{- } \mathcal{t}_{\sigma 1}^{+ } = \frac{a \tilde{h}}{\lambda } \hat{n}_{0}^{- } \mathcal{t}_{\sigma 0}. & \label{eqt-}\end{aligned}\ ] ] using eq . ( [ eqt0 ] ) , the function @xmath224 can be expressed through @xmath225 and @xmath226 . indeed , considering @xmath227 as a given function of @xmath156 , one can write a formal solution of eq . ( [ eqt0]).@xcite from this solution , by satisfying the absorbing and reflecting boundary conditions @xmath228 and @xmath229 ( since @xmath230 does not depend on @xmath94 , the last condition is equivalent to the finiteness condition @xmath231 , see sec . [ def ] ) , we obtain @xmath232 e^{ay^{2}}. \label{t0}\ ] ] substituting this result into eqs . ( [ eqt+ ] ) and ( [ eqt- ] ) , one can readily get the coupled integro - differential equations for @xmath221 and @xmath222 which , however , are too complicated to be solved in the general case . moreover , these equations are not closed because , in accordance with eqs . ( [ deff ] ) , ( [ eqt+ ] ) and ( [ eqt- ] ) , the function @xmath233 depends on @xmath230 . fortunately , an important case of high anisotropy barrier can be studied in detail . in the case of high anisotropy barrier ( or low temperatures ) , when the condition @xmath234 holds , the main contribution to the first integral in eq . ( [ t0 ] ) comes either from a small vicinity of the point @xmath235 ( if @xmath94 is not too close to @xmath236 and @xmath237 ) or from the vicinity of the point @xmath238 ( if @xmath239 ) . it should be noted that the latter situation can be realized only in the case of q mode with @xmath240 or @xmath241 . here , we restrict our theoretical analysis by considering small amplitudes of the rotating field that do not exceed the threshold amplitude of the q mode . such rotating field can induce only the p mode and , as a consequence , the former situation is always realized ( see also sec . [ prec ] ) . therefore , taking into account that in this case the integrals @xmath242 and @xmath243 at @xmath244 can be approximated by @xmath245 and @xmath246 , respectively , eq . ( [ t0 ] ) reduces to @xmath247 in order to evaluate the integral in eq . ( [ t02 ] ) , we note that small vicinities of the limits of the integration determine the asymptotic behavior of this integral at @xmath248 . more precisely , the lower limit is responsible for the high - frequency behavior of this integral and the upper one for its behavior in the resonant case . hereafter , we call the rotating field resonant if @xmath249 and the direction of its rotation coincides with the direction of the natural precession of the magnetic moment , i.e. , if @xmath250 . to study these cases in a unified way , it is convenient to write the integral in eq . ( [ t02 ] ) as a sum of two terms , @xmath251 and @xmath252 , which correspond to the resonant and high frequencies , respectively . thus , taking into account that @xmath253 and @xmath254 , eq . ( [ t02 ] ) yields @xmath255 . \label{ast0}\ ] ] it is important to stress at this point that eq . ( [ ast0 ] ) is not an exact asymptotic formula for @xmath230 because the terms @xmath256 and @xmath257 correspond to different frequencies . to calculate @xmath256 , we need to solve eqs . ( [ eqt+ ] ) and ( [ eqt- ] ) in the vicinity of the point @xmath259 . assuming that @xmath260 ( @xmath261 ) and @xmath262 , these equations can be rewritten as @xmath263 & \displaystyle \hat{l}_{1}^{+}\mathcal{t}_{\sigma 1}^{- } + \frac{2a}{\lambda } ( \sigma - \rho \tilde{\omega } ) \mathcal{t}_{\sigma 1}^{+ } = -\sigma \frac{a\tilde{h}}{\lambda}\frac{d\mathcal{t}_{\sigma 0 } } { d\eta_{\sigma } } , & \label{eqst+-a}\end{aligned}\ ] ] where @xmath264 since the approximate formula ( [ ast0 ] ) does not depend on @xmath265 , it can not be used to determine the derivative @xmath266 . therefore we use an exact result : @xmath267 e^{ay^{2 } } , \label{dert0}\ ] ] following from eq . ( [ t0 ] ) , which at @xmath268 gives @xmath269\ ! \eta_{\sigma}. \label{t0a}\ ] ] it is not difficult to verify that an exact solution of eq . ( [ eqst+-a ] ) that vanishes as @xmath270 has the form @xmath271 } \frac{d\mathcal{t}_{\sigma 0 } } { d\eta_{\sigma } } , & \nonumber \\[4pt ] & \displaystyle \mathcal{t}_{\sigma 1}^{- } = \sigma \frac{\lambda ( 2-\sigma\rho\tilde{\omega } ) \tilde{h } } { 2[(1-\sigma\rho \tilde{\omega})^{2 } + \lambda^{2 } ] } \frac{d\mathcal{t}_{\sigma 0 } } { d\eta_{\sigma } } , & \label{solt+-}\end{aligned}\ ] ] where @xmath266 is given by eq . ( [ t0a ] ) . according to this equation , the solution ( [ solt+- ] ) contains an unknown parameter @xmath272 , which can be determined from the fitting condition @xmath273 [ see eq . ( [ deff ] ) ] . taking into account that @xmath274 and @xmath275 , this condition reduces to @xmath276 finally , substituting eq . ( [ t0a ] ) into eq . ( [ fitt ] ) and imposing the commonly used condition @xmath277 , one gets @xmath278 to find @xmath280 , we assume that @xmath281 ( @xmath282 ) and , as before , @xmath283 . since in this case @xmath284 , @xmath285 and @xmath286 , eqs . ( [ eqt+ ] ) and ( [ eqt- ] ) take the form @xmath287 & \displaystyle \hat{l}_{1}^{+}\mathcal{t}_{\sigma 1}^{- } - \frac{2a}{\lambda}(\xi_{\sigma } + \rho\tilde{\omega } ) \mathcal{t}_{\sigma 1}^{+ } = -\frac{a\tilde{h } } { \lambda}\frac{d\mathcal{t}_{\sigma 0}}{d\xi_{\sigma } } , & \label{eqst+-b}\end{aligned}\ ] ] where @xmath288 and , as it follows from eq . ( [ dert0 ] ) , the derivative @xmath289 at @xmath290 is given by @xmath291 . \label{t0b}\ ] ] in general , the solution of eq . ( [ eqst+-b ] ) can be represented in the form of the taylor series : @xmath292 . however , since the main quantity of our interest is @xmath293 ( here , @xmath294 and @xmath295 ) , we can restrict ourselves to the linear approximation , i.e. , @xmath296 . assuming that @xmath297 , after straightforward calculations one obtains @xmath298 and @xmath299 therefore , the fitting condition @xmath300 can be reduced to the following one : @xmath301 which together with eq . ( [ t0b ] ) at @xmath277 yields @xmath302 \tilde{h}^{2 } } { 2[g(\tilde { \omega } ) + \sigma \rho a\tilde{h}^{2 } ] } , \label{f2b}\ ] ] where @xmath303 and @xmath304 is defined in eq . ( [ f1 ] ) . now we are in a position to determine the lifetime ( [ pmod ] ) of the p mode . using eq . ( [ appr ] ) and the fact that @xmath210 does not depend on @xmath94 and @xmath95 , the desired lifetime can be expressed as @xmath305 - \mathcal{t}_{\sigma 1}^{- } \sin \psi_{\sigma}[\rho]\big ) \big|_{\theta_{\sigma } ' = \theta_{\sigma}[\rho]}. \label{lt1}\ ] ] if the field amplitude @xmath70 is small enough , then @xmath306 $ ] in eq . ( [ st1 ] ) can be replaced by @xmath67 . one can easily check that in this case @xmath307 = -\rho \frac{\lambda \tilde { \omega } } { 1+\lambda^{2}- \sigma \rho \tilde{\omega } } , \label{psi}\ ] ] and the precession angle @xmath308 $ ] is given by @xmath192 = \frac{\pi}{2}(1 - \sigma ) + \sigma \tilde{h } \sqrt { \frac{1+\lambda^{2 } } { ( 1-\sigma \rho \tilde{\omega})^{2}+\lambda^{2}}}. \label{theta}\ ] ] the last result shows that the functions @xmath309 in eq . ( [ lt1 ] ) must be taken from eq . ( [ solt+- ] ) with @xmath310 + \pi(1- \sigma)/2 $ ] . but , according to eq . ( [ ast0 ] ) , the term @xmath311 contains an additional factor @xmath312 , and so the second term in eq . ( [ lt1 ] ) which describes the dependence of @xmath131 on @xmath313 $ ] can be safely neglected at @xmath182 . thus , using eqs . ( [ ast0 ] ) , ( [ f1 ] ) , and ( [ f2b ] ) , for the lifetime of the p mode we obtain @xmath314 where @xmath315 and @xmath316 if @xmath250 then eq . ( [ lt2 ] ) correctly describes the frequency dependence of the lifetime in the vicinity of the point @xmath317 and at @xmath318 . since @xmath319 , @xmath277 and @xmath320 , in the former case we obtain @xmath321 . to put it differently , the rotating magnetic field whose direction of rotation coincides with the direction of the natural precession decreases the lifetime in a resonant manner , i.e. , a resonant suppression of the thermal stability of the p mode occurs . taking into account that @xmath322 and @xmath323 , in the latter case eq . ( [ lt2 ] ) yields @xmath324 . in contrast , at @xmath325 eq . ( [ lt2 ] ) describes only the high - frequency behavior of the lifetime : @xmath326 . combining the last two results , we obtain the expression @xmath327 ( @xmath328 ) , which shows that while at @xmath329 the rotating field suppresses the lifetime of the p mode , the rotating field with @xmath325 enhances it . for the evaluation of the lifetime at @xmath330 it is convenient to use the one - dimensional approximation , which consists in replacing @xmath47 by @xmath163 in the magnetic energy ( [ defw ] ) . in this case @xmath331 , the lifetime @xmath131 does not depend on @xmath155 , and the partial differential equation ( [ mfpt1 ] ) reduces to the ordinary one : @xmath332 its exact solution satisfying the conditions ( [ abst ] ) and ( [ reflt ] ) [ we note that for this equation the finiteness condition ( [ fint ] ) is equivalent to the reflecting boundary condition ( [ reflt ] ) ] can be written in the form @xmath333 where @xmath334 ( @xmath335 ) is the symmetric function with @xmath336 , @xmath337 , and @xmath338 . as before , we are interested in the behavior of @xmath72 at @xmath182 . in this case a small vicinity of the point @xmath235 gives the main contribution to the first integral in eq . ( [ tzer1 ] ) . this contribution depends not only on @xmath339 but also on the parameter @xmath340 . in particular , if @xmath262 then , putting @xmath235 everywhere except for @xmath341 , representing @xmath341 in the vicinity of the point @xmath235 as @xmath342 and extending the limits of integration to infinity , eq . ( [ tzer1 ] ) yields @xmath343 then , taking into account that @xmath344 \!&=&\ ! \frac{e^{a } } { 2a}\,(1 + \sqrt{\pi a}\,\tilde{h } ) \label{relc1}\end{aligned}\ ] ] ( @xmath244 , @xmath262 ) and neglecting terms proportional to @xmath345 , from eqs . ( [ tzer2 ] ) and ( [ relc1 ] ) one finds @xmath346 comparing this result with eq . ( [ thigh3 ] ) , we conclude that @xmath347 , and so at @xmath348 the frequency dependence of the lifetime has a maximum exceeding the limiting value @xmath349 . it should be noted that the lifetime @xmath350 strongly decreases with increasing @xmath70 . for example , if @xmath351 then , instead of eq . ( [ tzer2 ] ) , we obtain @xmath352 to evaluate the integral in eq . ( [ tzer4 ] ) , it is convenient to divide the interval of integration @xmath353 into two parts , @xmath354 and @xmath355 , and apply the laplace method.@xcite by this way , the corresponding integrals can easily be evaluated yielding @xmath356 \!&=&\ ! \frac{\tilde{h}}{2 } \sqrt{\frac{\pi } { a}}\ , e^{a(1 + \tilde{h } ^{2 } ) } \mathrm{erf } \bigg ( \frac{\sqrt{a}}{\tilde{h}}\bigg ) \label{relc2}\end{aligned}\ ] ] and @xmath357 \!&=&\ ! \frac{\tilde{h}}{2 } \sqrt{\frac{\pi } { a}}\ , e^{a(1 + \tilde{h } ^{2 } ) } \mathrm{erf } \bigg ( \frac{\sqrt{a}\tilde{h}}{2}\bigg ) , \qquad \label{relc3}\end{aligned}\ ] ] where @xmath358 is the error function . finally , using eqs . ( [ tzer4])([relc3 ] ) and the approximate formula @xmath359 ( @xmath360 ) , we find the following expression for the lifetime at @xmath361 : @xmath362 , \label{tzer5}\ ] ] which is valid if @xmath244 and @xmath70 is of the order of @xmath363 . it is not difficult to see that for these conditions the strong inequality @xmath364 holds . taking into account also eq . ( [ thigh3 ] ) , we can conclude that @xmath72 as a function of @xmath185 at @xmath365 has a local maximum ( see sec . [ simulat ] ) . the analytical results suggest that the numerical analysis of the lifetimes of the precessional modes should start with the study of these modes without thermal fluctuations . more precisely , it is necessary ( i ) to determine the conditions when for a given rotating field one precessional mode exists in the up state of the magnetic moment and one in the down state , and ( ii ) to study the steady - state properties of these modes . to solve these problems , we use the landau - lifshitz equation written in the form of eq . ( [ l - l2 ] ) with @xmath157 . in general , the finite state of the magnetic moment ( i.e. , state at @xmath366 ) depends not only on the parameters @xmath70 , @xmath185 and @xmath138 characterizing the rotating field , but also on how this field is switched on . in particular , a sharp switching of the rotating field induces dynamical effects that may result in a change of the initial state @xmath67.@xcite although these effects can be important for applications , they are out of our scope here . therefore , to minimize the role of dynamical effects , we assume that the switch - on of the rotating field is slow enough . in order to determine the character of the precessional modes for a given rotating field , the following numerical scheme is used . first , the field amplitude is discretized as @xmath367 , where @xmath368 and @xmath369 is the amplitude increment . then , putting @xmath370 and using the initial conditions @xmath371 and @xmath372 , the fourth order runge - kutta method with a time step @xmath373 is applied to solve eq . ( [ l - l2 ] ) at @xmath374 on the time interval @xmath375 $ ] ( @xmath376 ) . it is assumed that at @xmath377 the field amplitude jumps from @xmath369 to @xmath378 , i.e. , @xmath100 becomes equal to two , and eq . ( [ l - l2 ] ) is solved again on the interval @xmath379 $ ] . but now the initial conditions are the solutions obtained for @xmath380 at the previous stage : @xmath381 . continuing this procedure , eq . ( [ l - l2 ] ) can be solved for an arbitrary @xmath100 . it is worth to note that since @xmath382 the solutions of this equation at @xmath383 are expected to be quite close to the steady - state solutions @xmath384 and @xmath385 . using the above procedure with @xmath386 , @xmath387 and @xmath388 [ this value of @xmath64 , which belongs to the interval @xmath389 of typical values of the damping parameter in the case of co samples,@xcite is used in all our numerical calculations ] , we determined the character of the precessional modes for a wide range of parameters characterizing the rotating field . it is established that if @xmath325 then only the p mode is realized for all @xmath70 and @xmath185 . in contrast , the precessional modes at @xmath250 exhibit a much more complex behavior . the results related to the character of these modes are summarized in the diagram shown in fig . we note that the difference between two p modes , which exist in the regions @xmath390 and @xmath391 , is that the precession angle @xmath183 as a function of @xmath70 is discontinuous at the boundary between them . it should also be emphasized that the transitions between the modes with @xmath329 , which occur under changing the field amplitude @xmath70 , are reversible . for clearness of presentation , this fact is illustrated by the horizontal bidirectional arrows . in contrast , the transitions to the p mode with @xmath325 are irreversible ( they are depicted by the horizontal unidirectional arrows ) . ( color online ) diagram of the precessional modes for @xmath250 . the regions in the @xmath70-@xmath185 plane where different p modes exist at @xmath250 are denoted as @xmath390 ( white ) and @xmath391 ( light - green ) . the q mode is realized in the white shaded region . in the region denoted as @xmath392 ( blue ) the stable precessional modes with @xmath250 do not exist . here , only the p mode with @xmath325 is realized . the vertical dotted lines ( a ) , ( b ) , ( c ) and ( d ) correspond to @xmath393 and @xmath394 , respectively . ] moreover , using eq . ( [ st1 ] ) and the stability criterion for the p mode,@xcite we independently confirmed the correctness of this diagram by calculating the lines that separate the regions with @xmath250 . since the rotating field is switched on during the time interval of duration @xmath395 , the above procedure is time consuming . however , in view of the importance of the diagram of the precessional modes for the problem of lifetimes , its use for the precise determination of this diagram is quite acceptable . at the same time , the application of this method to the study of the steady - state properties of a given mode , whose character is already known from the diagram , is clearly redundant . therefore , to reduce the computational time , next we use a modified numerical procedure with @xmath396 and @xmath397 leading to @xmath398 . figure [ fig2 ] shows the frequency dependence of the precession angle @xmath399 for different precessional modes . ( color online ) frequency dependence of the precession angle @xmath399 for different modes that exist at @xmath400 . the frequencies @xmath401 , @xmath402 and @xmath403 are the coordinates of the points in which the vertical dotted line ( d ) , see fig . [ fig1 ] , crosses the boundaries of the diagram ( @xmath404 , @xmath405 , and @xmath406 depend on @xmath70 ) . the green line ( with squares ) and brown line ( with triangles ) show the frequency dependence of @xmath407 and @xmath408 , respectively , in the case of q mode . ] the dashed ( black ) and solid ( blue ) lines correspond to the p modes with @xmath365 and @xmath409 , respectively . in accordance with fig . [ fig1 ] , the q mode occurs at @xmath410 ( we recall that this mode can exist only if @xmath409 ) . for this mode , the time dependence of the precession angle @xmath411 and the time dependence of the difference of phases @xmath412 are illustrated in figs . [ fig3 ] and [ fig4 ] . ( color online ) time dependence of the precession angle @xmath413 in the case of q mode . the parameters of the rotating field are as follows : @xmath414 , @xmath415 , and @xmath400 . ] ( color online ) time dependence of the difference of phases @xmath416 in the case of q mode . insert : time dependence of the function @xmath417 . the parameters of the rotating field are the same as in fig . [ fig3 ] and @xmath418 . ] we note that for a given set of parameters @xmath419 , i.e. , the precession angle @xmath173 in the case of q mode can cross the anisotropy barrier . moreover , since @xmath420 and @xmath421 have the same period @xmath422 , the parameter @xmath423 and the period @xmath422 ( its frequency dependence illustrates figs . [ fig5 ] ) are connected by the condition @xmath424 , where @xmath425 is a nonnegative integer ( color online ) frequency dependence of the period @xmath422 of the precession angle @xmath411 in the case of q mode . the rotating field parameters are taken as @xmath426 and @xmath400 . ] that depends on @xmath70 and @xmath427 . in particular , if @xmath400 then @xmath428 at @xmath429 , where @xmath430 is the solution of the equation @xmath431 at @xmath432 . it should also be stressed that in the last case the function @xmath433 shows a cosine - like behavior , in contrast to that shown in fig . [ fig4 ] . as it was mentioned earlier , the thermal fluctuations can cause the transitions between different precessional modes induced by a given rotating field . according to the diagram in fig . [ fig1 ] , these modes , one in the up state of the magnetic moment and the other in the down state , exist only if the amplitude and frequency of the rotating field belong to the region @xmath434 , @xmath435 or q. in this case , the lifetime of a given mode can be calculated from the numerical solution of the stochastic equations ( [ l - l2 ] ) . our testing calculations showed that the solution of these equations by the euler method gives practically the same lifetime obtained by the fourth order runge - kutta method . but in the first case the calculation time is almost four times less . therefore , because the procedure of determining the frequency dependence of the lifetime is extremely time - consuming , we used the euler method . the time step @xmath436 is chosen to be @xmath437 and the initial conditions are given by @xmath438 and @xmath439 . to prevent the appearance of singularities in eq . ( [ l - l2 ] ) at @xmath440 , we assume that the point @xmath441 acts on the process @xmath442 as a reflecting screen . since our interest here is the lifetimes of the precessional modes _ reaching _ the steady state , the thermal fluctuations are switched on at @xmath443 ( see sec . [ def ] ) with @xmath444 ( see fig . [ fig6 ] ) . in the case of p modes we chose @xmath445 , while for the q mode @xmath446 $ ] . in the latter case , the thermal fluctuations can be switched on at a certain instant of time , e.g. , when the precession angle @xmath447 reaches maximum or minimum . finally , in all our numerical simulations the parameter @xmath44 is chosen to be ten . ( color online ) schematic time dependence of the polar angle @xmath448 in the regions @xmath390 and @xmath391 shown in fig . the change of the magnetic moment state @xmath67 from @xmath25 to @xmath449 occurs at @xmath450 , when @xmath451 reaches the angle @xmath452 ( the horizontal dashed line ) for the first time . for a given trajectory @xmath448 , the lifetime of the p mode in the state @xmath453 is equal to @xmath454 . running @xmath455 trajectories , the mean lifetime can be evaluated as @xmath456 . ] in fig . [ fig7 ] , we show the frequency dependencies of the lifetime @xmath457 for the rotating field amplitudes indicated in fig . each point of these curves is determined by running @xmath458 trajectories of the polar angle @xmath448 ( see the caption to fig . [ fig6 ] ) . ( color online ) frequency dependencies of the lifetime of the precessional modes induced by the rotating field with @xmath414 in the up state ( @xmath459 ) of the magnetic moment . ] if @xmath460 then the dependence of @xmath457 on @xmath185 is continuous and exhibits a resonant minimum at @xmath461 . the plot of the resonant frequency @xmath462 versus the field amplitude @xmath70 is shown in fig . in contrast , if @xmath463 then @xmath457 is discontinuous : at @xmath464 the function @xmath457 does not exist . this result is a consequence of the fact that in the region @xmath392 ( see fig . [ fig1 ] ) there are no stable precessional modes with @xmath465 . ( color online ) dependence of the resonant frequency of the lifetime @xmath457 on the rotating field amplitude . ] one more important feature of the lifetime @xmath457 is that it is practically not sensitive to changing the character of the precessional modes . in particular , for @xmath466 the frequency dependence of @xmath457 ( see the red line with circles in fig . [ fig7 ] ) is continuous at the point @xmath467 ( i.e. , point separating the regions @xmath390 and @xmath468 at @xmath466 ) , while the precession angle @xmath469 is discontinuous . this insensibility of the lifetime to changing the precessional modes with changing the field frequency @xmath427 is especially surprising when the q mode appears . for example , even at @xmath400 , when the precession angle can cross the anisotropy barrier ( see fig . [ fig3 ] ) , the character of the frequency dependence of @xmath457 ( see the green line with stars in fig . [ fig7 ] ) is changed at @xmath470 so small that it is not visible on this scale . moreover , the lifetime of the q mode does almost not depend on @xmath471 $ ] . this result is counterintuitive because different @xmath472 may correspond to very different values of the precession angle @xmath473 . therefore , in order to get more insight into the problem , we calculated the probability density function @xmath474 of the lifetime of the q mode for two special values of @xmath475 which correspond to @xmath476 and @xmath477 , respectively . as seen from fig . [ fig9 ] , these density functions are somewhat different from each other only if @xmath478 . in this region the probability density @xmath479 depends on @xmath472 and exhibits local minima and maxima which come from a complex behavior of the q mode . in contrast , at @xmath480 the memory about the chosen value of @xmath472 and periodicity of @xmath411 and @xmath481 is lost . as a consequence , in this region the difference between density functions vanishes and the local minima and maxima disappear . such behavior of @xmath474 on @xmath482 confirms that the lifetime @xmath483 practically does not depend on @xmath472 . ( color online ) probability density functions of the lifetime of the q mode for two values of @xmath475 . the blue solid and red dashed lines correspond to such @xmath475 that @xmath484 and @xmath485 , respectively . insert : the same density functions in a larger time scale . the parameters of the rotating field are as follows : @xmath414 , @xmath486 , and @xmath400 . ] finally , the influence of the direction of field rotation on the frequency dependence of the lifetime of the precessional modes is illustrated in fig . [ fig10 ] . in accordance with our analytical results , the rotating field with @xmath414 and @xmath487 influences the lifetime @xmath457 in a different way . specifically , @xmath457 as a function of @xmath427 at @xmath414 displays a deep minimum , while at @xmath487 it shows a pronounced maximum . ( color online ) frequency dependencies of the lifetime of the precessional modes in the up state ( @xmath459 ) of the magnetic moment . it is assumed that @xmath400 for both clockwise ( @xmath487 ) and counterclockwise ( @xmath414 ) rotation of the magnetic field . ] this difference in the behavior of the lifetime results from that the rotating fields with different @xmath138 act on the magnetic moment in a given state @xmath67 quite differently . from a physical point of view , the reason is that the magnetic moment has a definite direction of the natural precession . we note also that the numerical data @xmath488 and @xmath489 are in a good agreement with the analytical results @xmath490 and @xmath491 obtained from the asymptotic formulas ( [ thigh3 ] ) and ( [ tzer5 ] ) , respectively . some difference between them can be caused by that the asymptotic formulas , which were obtained at @xmath248 , are applied to @xmath492 . we have studied in detail the thermal stability of the precessional modes of the nanoparticle magnetic moment induced by the rotating magnetic field whose plane of rotation is perpendicular to the easy axis of the nanoparticle . if the direction of field rotation and the direction of the natural precession of the magnetic moment are opposite , i.e. , if the condition @xmath325 holds , then only periodic ( p ) stable mode is induced by this field . in contrast , if the above mentioned directions coincide , i.e. , if @xmath250 , then the magnetic moment exhibits a much more complicated behavior . the numerical solution of the deterministic landau - lifshitz equation in the long - time limit has shown that , depending on the rotating field amplitude and frequency , in this case the magnetic moment can be in one of two p modes , in the quasi - periodic ( q ) mode , or even be unstable . these results obtained in the absence of thermal fluctuations have been collected in the diagram shown in fig . 1 . if the amplitude and frequency of the rotating field are chosen so that a stable precessional mode exists in both up ( @xmath459 ) and down ( @xmath493 ) states of the magnetic moment , then the thermal fluctuations can cause transitions between these modes . one of the most important parameters characterizing these transitions is the lifetime of a given mode . since it can be naturally associated with the mean first - passage time for the magnetic moment , we have used the fokker - planck formalism to define this quantity and calculate its properties . in particular , we have determined the boundary conditions and transformation properties of the lifetime and have developed an analytical method for finding its frequency dependence in the case of high anisotropy barrier and small amplitudes of the rotating field . using this method , it has been shown that the rotating field ( a ) slightly decreases ( if @xmath250 ) or increases ( if @xmath494 ) the lifetime of the p mode at large frequencies and ( b ) strongly decreases it ( if @xmath250 ) in the vicinity of the larmor frequency . we have also established that at zero frequency the lifetime is always less than in the limit of large frequencies . these analytical findings for the lifetime of the p mode have been confirmed by our numerical simulations of the stochastic landau - lifshitz equation . moreover , the numerical simulations of this equation for not too small amplitudes of the rotating field permitted us to solve the problem of the lifetime of the q mode . since in this case the precession angle is a periodic function of time which can cross the anisotropy barrier , the solution of this problem is of particular interest . it has turned out that , although the precessional angle depends on time , the lifetime of the q mode practically does not depend on this time , i.e. , on the precession angle . we have also verified this result by calculating the lifetime from the first - passage time distributions that correspond to different values of the precession angle . we are grateful to the institute of applied physics nas of ukraine for granting us access to their computing facilities . i. uti , j. fabian , and s. das sarma , rev . mod . phys . * 76 * , 323 ( 2004 ) . s. a. wolf , d. d. awschalom , r. a. buhrman , j. m. daughton , s. von molnr , m. l. roukes , a. y. chtchelkanova , and d. m. treger , sciense * 294 * , 1488 ( 2001 ) . s. laurent , d. forge , m. port , a. roch , c. robic , l. vander elst , and r. n. muller , chem . rev . * 108 * , 2064 ( 2008 ) . m. ferrari , nat . cancer * 5 * , 161 ( 2005 ) . q. a. pankhurst , j. connolly , s. k. jones , and j. dobson , j. phys . d : appl . * 36 * , r167 ( 2003 ) . p. hnggi , p. talkner , and m. borkovec , rev . phys . * 62 * , 251 ( 1990 ) . c. w. gardiner , _ handbook of stochastic methods _ , 2nd ed . ( springer - verlag , berlin , 1990 ) . s. redner , _ a guide to first - passage processes _ ( cambridge university press , cambridge , 2001 ) . c. thirion , w. wernsdorfer , and d. mailly , nat . * 2 * , 524 ( 2003 ) . z. z. sun and x. r. wang , phys . b * 73 * , 092416 ( 2006 ) . z. z. sun and x. r. wang , phys . b * 74 * , 132401 ( 2006 ) . j. podbielski , d. heitmann , and d. grundler , phys . * 99 * , 207202 ( 2007 ) . g. woltersdorf and c. h. back , phys . lett . * 99 * , 227207 ( 2007 ) . g. bertotti , c. serpico , and i. d. mayergoyz , phys . lett . * 86 * , 724 ( 2001 ) . g. bertotti , i. d. mayergoyz , and c. serpico , physica b * 343 * , 325 ( 2004 ) . s. i. denisov , t. v. lyutyy , p. hnggi , and k. n. trohidou , phys . b * 74 * , 104406 ( 2006 ) . g. bertotti , i. d. mayergoyz , c. serpico , m. daquino , and r. bonin , j. appl . * 105 * , 07b712 ( 2009 ) . l. chotorlishvili , p. schwab , and j. berakdar , j. phys . : condens . matter * 22 * , 036002 ( 2010 ) . s. i. denisov , t. v. lyutyy , p. hnggi , phys . lett . * 97 * , 227202 ( 2006 ) . s. i. denisov , k. sakmann , p. talkner , and p. hnggi , europhys . lett . * 76 * , 1001 ( 2006 ) . s. i. denisov , k. sakmann , p. talkner , and p. hnggi , phys . b * 75 * , 184432 ( 2007 ) .

we report new properties of the 11 and 12.7 m emission complexes of polycyclic aromatic hydrocarbons ( pahs ) by applying a gaussian - based decomposition technique . using high - resolution _ spitzer _ space telescope data , we study in detail the spectral and spatial characteristics of the 11 and 12.7 m emission bands in maps of reflection nebulae ngc 7023 and ngc 2023 ( north and south ) and the star - forming region m17 . profile variations are observed in both the 11 and 12.7 m emission bands . we identify a neutral contribution to the traditional 11.0 m pah band and a cationic contribution to the traditional 11.2 m band , the latter of which affects the pah class of the 11.2 m emission in our sample . the peak variations of the 12.7 m complex are explained by the competition between two underlying blended components . the spatial distributions of these components link them to cations and neutrals . we conclude that the 12.7 m emission originates in both neutral and cationic pahs , lending support to the use of 12.7/11.2 intensity ratio as a charge proxy .

1604.08589

prominent infrared ( ir ) emission bands between 3 and 20 m are observed in many astronomical environments . these spectral features are attributed to the vibrational fluorescence of polycyclic aromatic hydrocarbons ( pahs ) , which are electronically excited by ultraviolet ( uv ) photons . pah molecules are composed of a hexagonal honeycomb carbon lattice , typically containing 50 - 100 carbon atoms , with a dusting of hydrogen atoms about the periphery . compact structures are generally the most stable , but a variety of pah shapes and sizes are expected to exist in space @xcite . the strongest pah emission bands are observed at 3.3 , 6.2 , 7.7 , 8.6 , 11.2 , and 12.7 m. a variety of weaker bands are also seen in the observational spectra ( at , e.g. , 11.0 , 12.0 and 13.5 , 14.2 , 16.4 m ) . the pah bands can be associated with the following vibrations : c - h stretching ( 3.3 m ) ; c - c stretching ( 6.2 m ) ; c - c stretching and c - h in - plane bending ( 7.7 , 8.6 m ) ; and c - h out - of - plane bending ( hereafter ch@xmath0pah bands in the 10 - 15 m region ) . it is the number of adjacent c - h groups that determines the wavelength of the emission in the 10 - 15 m region , i.e.solo , duo , trio and quartet c - h groups . the relative emission intensities in these bands are known to be highly variable between sources and within individual resolved objects ( e.g. @xcite ) . the charge state of the pah population has been identified in laboratory studies as the most important parameter in driving variations in the relative emission intensities , sometimes reaching one order of magnitude between charge states @xcite . likewise , the profiles are known to vary in shape and peak position , which have been linked to object type ( e.g. @xcite ) . the variability of the pah profiles is thought to represent differences in pah sub - populations , possibly in , e.g. , size or structure ( e.g. @xcite ; see @xcite for a detailed overview ) . decomposing the pah emission bands with a mixture of functions ( e.g. gaussians , lorentzians , drude profiles ) is one way to investigate the origins of the observed spectral variability @xcite . a recent result by peeters et al . ( 2015 ) showed that the 7.7 and 8.6 m emission bands can be decomposed into four gaussian components , revealing that at least two pah sub - populations contributed to this emission . motivated by this result , we apply here a similar approach to the 11 m emission complex ( i.e. , both the 11.0 and 11.2 m bands ) and the 12.7 m emission complex . since the variations of the 11 and 12.7 m complexes are relatively minor when compared to the 7.7 and 8.6 m emission bands , it is critical to examine high spectral resolution observations . in addition , if band substructure indeed traces pah sub - populations , the astronomical data considered must span a sufficiently wide swath of physical conditions , such that any intrinsic differences are reflected in the observational band profiles . we present here new decompositions of the 11 and 12.7 m emission complexes in high - resolution _ spitzer_/irs maps of rne and a star - forming region in order to understand the pah sub - populations that produce the blended emission bands . we organize the paper as follows : the targets , observations , and continuum determination are presented in section [ sec : obs ] . the spectral variability in the spectra prior to any further analysis are examined in section [ sec : specprof ] . we introduce new methods for decomposing the 11 and 12.7 m pah emission bands in section [ sec : methods ] . results are presented in section [ sec : results ] and we discuss the implications of these results in section [ sec : discussion ] . we present a brief summary in section [ sec : conclusion ] . we chose targets with _ spitzer_/irs - sh maps that exhibit strong emission in the 11 and 12.7 m pah complexes . the chosen sources were the rne ngc 7023 , ngc 2023 ( two pointings north and south ) , and the star - forming region m17 . spectroscopic maps were obtained using the infrared spectrograph ( irs ; @xcite ) on - board the _ spitzer _ space telescope @xcite . these data , spanning 10 - 20 m at resolving power r@xmath1600 , were obtained from the nasa / ipac _ spitzer _ heritage archive . the spectral maps included in this work are summarized in table [ table : target_properties ] . astrometry for ngc 7023 and ngc 2023 south is presented in fig . [ fig : astrometry ] , and for ngc 2023 north and m17 in fig . [ fig : astrometry2 ] . [ cols="<,^,^,^,^,^,^,^,^ " , ] @xmath2 field of view of the sh maps . @xmath3 the aorkey uniquely identifies _ space telescope observations . @xmath4 the major exciting star is thought to be cen 1 , a double o4-type star @xcite . [ table : target_properties ] the ngc 2023 north and south sh maps were previously presented by @xcite and @xcite . these data were processed by the _ spitzer _ science center ( pipeline version s18.7 ) . further processing was accomplished with the ` cubism ` tool @xcite , including coaddition and bad pixel cleaning . for the purpose of spectral extraction , a @xmath5-pixel aperture was stepped across each map . this ensured that the extraction apertures matched the point - spread function , removing non - independent pixels . further details of the reduction process can be found in @xcite . a similar approach was applied in the reduction of the sh maps of ngc 7023 and m17 . the map of ngc 7023 has been previously analyzed by @xcite and @xcite . _ spitzer_/irs observations of m17-sw have been previously examined by @xcite and @xcite . extinction is small in ngc 2023 , as are the variations in extinction , as concluded by @xcite and @xcite , with a@xmath6 values on the order of 0.1 . we thus do not correct for it in this source . the extinction in ngc 7023 and m17 was investigated by @xcite . these authors computed the extinction with the iterative spoon method @xcite using _ spitzer_/irs - sl data . regarding ngc 7023 , @xcite found significant extinction ( a@xmath7 ) in the lower left corner of the map , as also reported by @xcite . the small amount of extinction in the rest of the map , combined with the fact that the gradients of the extinction curve in the 11.0 - 11.6 and 12.3 - 13 m regions are small , leads to a change in profile shapes of the 11.2 and 12.7 m features at approximately the 4% level towards ngc 7023 . in contrast , the extinction towards m17 reaches a maximum a@xmath6 value of 1.48 in our field , with typical values near unity . therefore , we dereddened our spectra using the chiar & tielens ( 2006 ) extinction law . further discussion and analysis on this topic is presented in section [ subsec:127_over_112 ] , including its effect on the 12.7/11.2 band strength ratio . a local spline continuum was chosen to isolate the pah emission features . a series of anchor points were chosen between 10 - 20 m ( fig . [ fig : spectrum ] ) to define the spline . this type of continuum determination has been performed several times in the literature and is chosen here for the purposes of comparison ( e.g. , @xcite ) . the pah band profiles and fluxes derived from the spline method are not very sensitive to the precise position of the anchor points , depending on the apparent smoothness of the underlying continuum ( i.e. spectra with , e.g. , undulating continua will lead to less precise pah band flux measurements ) . we estimate the influence of the continuum choice on our 11.2 and 12.7 m band fluxes to be at the 5% level . we focus here specifically on the pah emission bands at 11.0 , 11.2 and 12.7 m. the former two bands we will generally refer to as the 11 m " emission complex , and the latter as the 12.7 m emission complex . the 12.0 m pah band is excluded as it does not exhibit spectral blending with the 12.7 m emission . the variability of the 11 m and 12.7 m emission complexes in ngc 7023 is examined in fig . [ fig : raw ] . a representative set of pixels are chosen from the map such that each position is a different distance from the illuminating source . the relative strengths of the 11.0 and 11.2 m emission bands show significant variations across the map ( upper panel ) . as stellar distance decreases , four effects are simultaneously visible : the 11.0 m band strength relative to the 11.2 m emission increases monotonically ; the peak position of the 11.2 m band moves to shorter wavelengths ( from class ( a ) to class a(b ) ; see also @xcite ) ; a small , narrow feature at 11.20 m appears ; and the red wing of the 11 m complex either shifts towards shorter wavelengths or decreases in intensity ( also reported by @xcite for orion and ngc 7023 , respectively ) . for these same positions , we inspect the behaviour of the 12.7 m complex ( fig . [ fig : raw ] , lower panel ) . as the stellar distance decreases , there is a clear transition of the 12.7 m peak position towards shorter wavelengths from approximately 12.77 to 12.71 m. the red wing also decreases in intensity ( or shifts to shorter wavelengths ) during this transition . in addition , there is a difference in the blue wing of the 12.7 m complex between the map positions : in the range 12.5 - 12.7 m , positions closer to the star display greater emission intensities than those further away . it does not appear that the entire emission complex is shifting , as the offsets of the red wing , peak position and blue wing are not consistent . likewise , the emission blueward of 12.5 m are identical in each position . the other sources exhibit only minor variations . m17 displays a similar peak transition in the 11.2 m band , while ngc 2023 north and south show little peak variation in the 11 m complex . at 12.7 m , ngc 2023 south and m17 have small variations in the peak profile and red wing , but not as clearly as for ngc 7023 . the north map of ngc 2023 has no discernible 12.7 m variations . the observed sub - structure of the emission complexes in ngc 7023 suggests that subtle and possibly important clues about pah sub - populations may be accessible . we now discuss traditional methods and a newly proposed method for analyzing the 11 and 12.7 m pah bands . the 11 m emission complex consists of two components : a strong , broad feature near 11.2 m and a weaker band near 11.0 m. they are generally blended together near 11.1 m. there are multiple ways in which to measure the fluxes in these two emission bands . the simplest method is to integrate the 11 m complex shortward of 11.1 m ( representing the 11.0 m band ) and longward of 11.1 m ( representing the 11.2 m band ) . a more complex method involves using gaussian components to help disentangle the spectral blend ( see , e.g. , @xcite , peeters et al . 2015 , apj , submitted ) . in this case , one simultaneously fits a gaussian to the 11.0 m band and a gaussian to the blue wing of the 11.2 m complex . the 11.0 m flux is then the flux contained within the fitted gaussian at 11.0 m. the 11.2 m band flux is measured by subtracting the 11.0 m gaussian from the original spectrum and integrating the remainder . a good single gaussian fit to the 11.2 m band is not possible due to the asymmetric red wing ( @xcite ) . a third method to separate the 11.0 and 11.2 m emission was introduced by @xcite . the authors used a five - component gaussian decomposition in which one component was responsible for the 11.0 m emission . the sum of the remaining components then represented the 11.2 m emission . a fourth possibility is to fit the 11 m emission with two drude profiles , an approach used by the pahfit tool @xcite . for the 12.7 m band , one measurement method ( besides direct integration ) was presented by @xcite . their approach uses the average 12.7 m emission profile from @xcite as a template . this template is scaled to the data in the range 12.4 - 12.7 m , where only pah emission is expected . there is frequently an adjacent [ ne ii ] 12.81 m atomic line and h@xmath8 emission at 12.3 m , which prevents scaling beyond this spectral window . after fitting the pah template and subtracting its profile , only atomic and/or molecular emission lines remain . these are fit with gaussian functions in accordance with the instrumental spectral resolution of the data . afterwards , the atomic and molecular lines are subtracted from the original 12.7 m emission complex , leaving only the 12.7 m pah emission . as with the 11 m emission , another possible 12.7 m decomposition is to fit the band with two drude profiles ( the pahfit approach ) . to account for the diversity of the profile variations ( see section [ sec : specprof ] ) , we have introduced a new decomposition method which we will explore in detail . we attempted to fit the 11 and 12.7 m complexes with a variety of gaussians and lorentzians . we used the non - linear least squares fitting tool mpfit @xcite , which is an implementation of the levenberg - marquardt algorithm @xcite . we compute the reduced-@xmath9 of the fit for each pixel in our spectral cubes . a histogram of all pixels in each cube is prepared , which is used to evaluate the overall fit . the best reduced-@xmath9 values resulted from fitting the 11 m complex with five gaussians and the 12.7 m complex with four . to determine their nominal parameters we applied an iteration method . initially , the components were permitted to move , as long as they did not overlap . their widths were also allowed to vary . after allowing several free runs " we observed that some of the parameters were converging on particular values ( those areas of the spectrum with less complex structure ) . we fixed these converging parameters and recomputed our fits . the process was iterated in this manner until all parameters had converged . the final fits are thus obtained with fixed peak positions and full - width at half - maximum ( fwhm ) values while only the peak intensity of each component was allowed to vary . it is important to emphasize that _ the decompositions we adopt are arbitrary_. we do not suggest that they reflect a true decomposition " or any a priori knowledge . rather , this method is applied to understand what ( if anything ) may be learned about the pah bands through simple fitting . [ [ the-11-mcomplex ] ] the 11 m complex + + + + + + + + + + + + + + + + the best fits resulted from a five - component gaussian fit . the gaussians were fixed to the following positions : 11.021 m ( fwhm : 0.066 m ) , 11.000 m ( 0.021 m ) , 11.199 m ( 0.021 m ) , 11.320 m ( 0.118 m ) and 11.245 m ( 0.055 m ) . an example of the decomposition we adopted is shown in fig . [ fig:112_3pos ] . we see that the 11.0 m emission is determined by two components , and the 11.2 m emission by three components . the two components of the 11.0 m emission can be qualitatively interpreted as an underlying broad component ( centered near 11.0 m ) and a narrower symmetric feature placed upon it . the 11.2 m profile displays a strong gaussian feature centered near 11.25 m with an accompanying red wing ( out to 11.6 m ) . the slight asymmetry of the 11.2 m peak , namely on the short - wavelength edge ( near 11.20 m ) , is coincident with the fifth component . [ [ the-12.7-mcomplex ] ] the 12.7 m complex + + + + + + + + + + + + + + + + + + we applied the same methodology to the 12.7 m emission complex and found that the best statistics resulted from a four - component decomposition . a sample decomposition is presented in fig . [ fig:127_3pos ] . one component of the fit is linked to the broad blue wing , upon which a weaker , narrower component is placed . the other two components compete to fit the peak position of the band and the spectral profile of its long - wavelength wing . the adopted central wavelengths of our components are as follows : 12.55 m ( fwhm : 0.160 m ) , 12.54 m ( 0.035 m ) , 12.72 m ( 0.090 m ) and 12.78 m ( 0.080 m ) . we present our decompositions of the 11 and 12.7 m emission complex in figs . [ fig:112_3pos ] and [ fig:127_3pos ] , respectively , for a choice of three locations in ngc 7023 . these positions are chosen along a radial vector emanating outward from the central star . the positions are 34 , 43 and 53 distant from the star , respectively . considering the 11 m emission first , we see that at the closest location the 11-g1 component dominates the 11.0 m emission ( fig . [ fig:112_3pos ] , upper panel ) . here , the 11-g2 component contributes weakly . at further distances , however , the flux of the 11-g1 component decreases significantly , while the flux of the 11-g2 component is relatively unchanged . at the furthest distance , the peak flux density of the 11-g2 component slightly exceeds that of the 11-g1 component . the 11.2 m emission profile varies across the three map positions . near the star , the 11.2 m peak is asymmetric and centered near 11.20 m. far from the star , we see that the peak of the 11.2 m emission has shifted to @xmath111.25 m and it is now more symmetric . note that the fit has a slight mismatch on the peak emission at the closer positions . the 11-g3 component decreases in intensity with distance , similar to that of the 11-g1 emission . the 11-g3 component appears to be at least partially responsible for the asymmetry of the peak 11.2 m emission at positions near the star . the other two components , 11-g4 ( which traces the red - shaded wing ) and 11-g5 ( which contains the bulk of the flux at the traditional position of 11.2 m ) increase in flux density with increasing distance . in fig . [ fig:127_3pos ] we see that at the position nearest the star the peak of the 12.7 m complex lies blueward of 12.75 m ( which we use as a reference ) . at this location , the flux of 12-g3 is clearly greater than that of 12-g4 . we observe strong asymmetry of the 12.7 m peak at this position . note the fit can not completely reproduce the peak shape observed . at the position furthest from the central star ( lower panel ) , however , the peak is now located redward of 12.75 m. the flux of the 12-g4 component now exceeds that of the 12-g3 component considerably , indicating that it is the relative strengths of the 12-g3 and 12-g4 components that determines the peak position of the 12.7 m emission . at an intermediate distance from the central star in ngc 7023 ( middle panel ) , we see that 12-g3 and 12-g4 are of roughly equal strength , and that the 12.7 m emission has a peak that is an intermediate between the extremes of the upper and lower panels . the broad component in this decomposition , 12-g1 , is generally unchanged across the three map positions . the 12-g2 component however decreases in relative intensity as one moves to further distances from the central star . + + + + starting with ngc 7023 , we present spectral maps of the five components of the 11 m decomposition in fig . [ fig : map_7023 ] . we have included maps of the traditional 11.0 and 11.2 m decomposition , defined here as direct integration , for comparison ( see section [ sec : tradition ] ) . we observe that the distribution of the 11-g1 component is extremely similar to that of the 11.0 m band . recall that the 11-g1 component is the broad underlying feature centered near 11.0 m. the 11-g2 component , which is the smaller feature perched on top of the 11-g1 plateau , shows a spatial distribution that is intermediate between that of the 11.0 and 11.2 m emission . this is also true for the 11-g3 component , which appears to be intermediate in morphology between the 11.0 and 11.2 m bands . the 11-g4 and 11-g5 components are both coincident with the 11.2 m emission . similar results are found in ngc 2023 south ( fig . [ fig : map_2023s ] ) . the map of the 11-g1 component s flux is an extremely good match to that of the traditional 11.0 m component . the 11-g2 and 11-g3 components are also intermediate in morphology between that of the 11.0 and 11.2 m bands . however , it is clear that in ngc 2023 south the 11-g3 component is closer in spatial distribution to that of the 11.0 m emission and the 11-g2 component is closer in spatial distribution to that of the 11.2 m emission . the 11-g4 and 11-g5 components again match the 11.2 m emission , though the 11-g4 component is less extended in comparison . the spectral maps of ngc 2023 north are presented in fig . [ fig : map_2023n ] . we observe that the 11.0 m band and the 11-g1 components are similar , each peaking along a vertical slice in the map . additionally , the 11-g3 , 11-g4 and 11-g5 components exhibit similar morphologies , each peaking along a common horizontal strip . the 11-g2 component appears to share the peak positions of these two groups , forming an almost l " shape from the horizontal and vertical peak pah zones . in m17 ( fig . [ fig : map_m17 ] ) , there is generally very little variation between the traditional 11.0 and 11.2 m emission maps . they appear to peak in the same position in the map . however , there is one apparent difference : there is 11.2 m emission in the upper corner , forming a small ridge along the boundary of the field of view . we will hereafter refer to this feature as the m17 `` spur . '' using this as a distinguishing characteristic , we observe that the 11.2 m band , the 11-g4 component and the 11-g5 component all have emission in this region . conversely , the 11.0 m band and the 11-g1 , 11-g2 and 11-g3 components all have very little emission in comparison . we also observe that the 11-g2 component is more extended than the 11-g1 and 11-g3 emission , and is closer in morphology to that of the 11-g4 and 11-g5 bands . this suggests that the 11-g1 and 11-g3 bands are more closely related to the 11.0 m emission than the 11-g2 component . in brief summary , the emission of the 11-g1 component closely spatially matches the 11.0 m band . the maps of the 11-g2 and 11-g3 components are a mixture of the 11.0 and 11.2 m maps , while the 11-g4 and 11-g5 distributions are well - matched to that of the traditional 11.2 m band . in fig . [ fig : map_7023 ] we present maps of the four components of the 12.7 m decomposition . we again include maps of the 11.0 and 11.2 m emission bands as measured with the traditional decomposition ( see section [ sec : tradition ] ) . as shown in fig . [ fig:127_3pos ] , it is the 12-g3 and 12-g4 components whose relative strengths determine the position of the 12.7 m peak . the 12-g3 component of the 12.7 m decomposition has a very similar spatial distribution to that of the 11.0 m emission , peaking in generally the same location . the morphologies of the 12-g4 component and the 11.2 m emission are likewise very similar . there is generally very little overlap between the 12-g3 and 12-g4 components in these maps . the 12-g2 component generally peaks where the 11.0 m emission does , and therefore also the 12-g3 component , but it is clearly more extended than either of these features . the 12-g1 component is the most extended , with emission at both the 11.0 and 11.2 m peaks . this is consistent with the findings in fig . [ fig:127_3pos ] , in which the 12-g1 component showed little variation in flux density for three chosen positions in the ngc 7023 map and the 12-g2 component increased in strength when approaching the star . we perform a similar analysis on the map of ngc 2023 south ( fig . [ fig : map_2023s ] ) . we also observe strong similarities between the spatial distributions 11.0 m and 12-g3 fluxes , and the 11.2 m and 12-g4 fluxes . the 12-g1 and 12-g2 components both appear to involve a mixture of the spatial emission of the 11.0 and 11.2 m bands , which is generally consistent with what was observed in ngc 7023 . the spatial distributions of the emission bands in ngc 2023 north ( fig . [ fig : map_2023n ] in the appendix ) are difficult to interpret . very broadly , the peak of the 11.2 m emission is coincident with the peak of the 12-g4 component s emission . the 11.0 m emission peaks along a vertical line in the map ; the 12-g3 component also has strong emission in this region , but there appears to also be significant emission in the same location as that of the 11.2 m band ( spanning a horizontal zone ) . the 12-g1 and 12-g2 components are broad and generally overlap the locations in which the 11.0 and 11.2 m emission originates . the 12-g1 component has the most extended emission in this map . spectral maps of the emission in m17 are presented in the appendix in fig . [ fig : map_m17 ] . m17 is also difficult to disentangle , but there is roughly a commonality between the 11.2 m emission , 12-g4 , and 12-g1 components . likewise , a grouping of the 11.0 m emission , 12-g3 and 12-g2 components is observed , as found in ngc 7023 . summarizing , the 12-g3 emission spatially matches that of the 11.0 m band , while the 12-g4 emission component is a close spatial match to the 11.2 m emission . the spatial distributions of the 12-g1 and 12-g2 emission maps are clearly a mixture of the 11.0 and 11.2 m emission maps . the profile variability of the 11 and 12.7 m pah complexes can be explained within the framework of our adopted decompositions . the asymmetry of the 11.2 m peak is determined by the strength of the 11-g3 component , and the asymmetry of the 12.7 m peak is determined by the competition between the 12-g3 and 12-g4 components . the spectral maps reveal that some components are spatially coincident while others display distinct spatial morphologies . to quantify the morphological similarities of the spectral maps we introduce the structural similarity algorithm of @xcite . this is an image processing method to evaluate the similarities between images based on local luminance , contrast and structure . the method produces a structural similarity index ( ssim ) to quantify how alike two images are . the ssim value ranges from -1 to 1 , where values approaching 1 represent very similar images ( only identical images have an ssim index of unity ) . the ssim index is computed by first comparing sub - regions , or windows , of the two images . the windows are used to compare each portion of the corresponding images , before producing a single number to encapsulate the similarity between the images as a whole . the ssim index between two images @xmath10 and @xmath11 of common size @xmath12 is defined as follows : @xmath13 where @xmath14 and @xmath15 are the mean and variance of each window , respectively . the constants @xmath16 and @xmath17 are variables for preventing instability when the denominator would otherwise approach zero . they are defined as @xmath18 and @xmath19 , in which @xmath20 is the dynamic range of the image , and @xmath21 and @xmath22 are canonically 0.01 and 0.03 , respectively @xcite . we use the structural_similarity subpackage of the scikit - image python package @xcite to compute the ssim values . the default window size of 7 pixels is used , as are the canonical @xmath21 and @xmath22 values . we present the ssim values comparing the maps from our decomposition in table [ tables : ssim7023 ] . ssim indices exceeding 0.90 are presented in bold face . due to the nans in the m17 map and the ssim requirement for rectangular windows we could not analyze the m17 map in this manner . we find the same results as were determined from visual inspection : the morphologies of the 11-g1 and 12-g3 emission components are well - matched with that of the 11.0 m emission ; the 11-g4 , 11-g5 and 12-g4 emission components are spatially well - matched with the 11.2 m emission ; and the 11-g2 , 11-g3 , 12-g1 and 12-g2 emission components are spatially a mixture of the two . similar conclusions about the 12.7 m components are reached when examining correlation plots of band flux ratios ( fig . [ fig : corr ] ) . l|*13c + 11.0 & & & & & & & & & & & + 11.2 & 0.30 & & & & & & & & & & + 11-g1 & & 0.30 & & & & & & & & & + 11-g2 & 0.75 & 0.73 & 0.74 & & & & & & & & + 11-g3 & 0.69 & 0.78 & 0.69 & & & & & & & & + 11-g4 & 0.24 & & 0.24 & 0.66 & 0.71 & & & & & & + 11-g5 & 0.32 & & 0.32 & 0.75 & 0.79 & & & & & & + 12-g1 & 0.63 & 0.68 & 0.64 & 0.74 & 0.71 & 0.67 & 0.69 & & & & + 12-g2 & 0.75 & 0.71 & 0.75 & & & 0.64 & 0.74 & 0.74 & & & + 12-g3 & 0.88 & 0.60 & 0.88 & & & 0.52 & 0.63 & 0.75 & & & + 12-g4 & 0.22 & & 0.22 & 0.65 & 0.69 & & & 0.60 & 0.64 & 0.51 & + & 11.0 & 11.2 & 11-g1 & 11-g2 & 11-g3 & 11-g4 & 11-g5 & 12-g1 & 12-g2 & 12-g3 & 12-g4 + + + + 11.0 & & & & & & & & & & & + 11.2 & 0.52 & & & & & & & & & & + 11-g1 & & 0.50 & & & & & & & & & + 11-g2 & 0.66 & 0.86 & 0.60 & & & & & & & & + 11-g3 & 0.81 & 0.60 & 0.80 & 0.70 & & & & & & & + 11-g4 & 0.47 & & 0.44 & 0.81 & 0.53 & & & & & & + 11-g5 & 0.55 & & 0.53 & 0.87 & 0.63 & & & & & & + 12-g1 & 0.66 & 0.87 & 0.63 & 0.83 & 0.70 & 0.83 & 0.87 & & & & + 12-g2 & 0.68 & 0.61 & 0.65 & 0.72 & 0.77 & 0.56 & 0.63 & 0.55 & & & + 12-g3 & 0.84 & 0.53 & 0.83 & 0.70 & & 0.46 & 0.57 & 0.61 & 0.82 & & + 12-g4 & 0.55 & & 0.52 & 0.86 & 0.60 & & & 0.87 & 0.64 & 0.53 & + & 11.0 & 11.2 & 11-g1 & 11-g2 & 11-g3 & 11-g4 & 11-g5 & 12-g1 & 12-g2 & 12-g3 & 12-g4 + + + + 11.0 & & & & & & & & & & & + 11.2 & 0.61 & & & & & & & & & & + 11-g1 & & 0.63 & & & & & & & & & + 11-g2 & 0.76 & 0.89 & 0.75 & & & & & & & & + 11-g3 & 0.66 & & 0.68 & & & & & & & & + 11-g4 & 0.56 & & 0.57 & 0.85 & & & & & & & + 11-g5 & 0.68 & & 0.70 & & & & & & & & + 12-g1 & 0.64 & 0.76 & 0.64 & 0.78 & 0.77 & 0.75 & 0.76 & & & & + 12-g2 & 0.64 & 0.64 & 0.65 & 0.69 & 0.75 & 0.60 & 0.68 & 0.43 & & & + 12-g3 & 0.87 & 0.76 & 0.88 & 0.88 & 0.83 & 0.70 & 0.82 & 0.64 & 0.78 & & + 12-g4 & 0.77 & 0.76 & 0.76 & 0.83 & 0.79 & 0.71 & 0.83 & 0.74 & 0.70 & 0.80 & + & 11.0 & 11.2 & 11-g1 & 11-g2 & 11-g3 & 11-g4 & 11-g5 & 12-g1 & 12-g2 & 12-g3 & 12-g4 + since early on , astronomical observations of extended sources have revealed that the major pah bands exhibit spatially different behaviour . in particular , the 8.6 and 11.0 m bands peak closer to the exciting star than the 3.3 and 11.2 m bands , which peak further away ( e.g. @xcite ) . the behaviour has been attributed to the pah charge state , as laboratory experiments have shown that the 8.6 and 11.0 m emission are dominated by cations , and the 11.2 m emission by neutral pahs @xcite . a comparison of astronomical spectra to laboratory and theoretically calculated spectra by @xcite reinforced the assignment of the 11.0 and 11.2 m bands to solo ch@xmath0 bending in cations and neutrals , respectively . more recently , computed spectra of ( very ) large pahs by @xcite and @xcite showed that the solo ch@xmath0 emission from pah neutrals becomes blueshifted upon ionization , supporting the same assignment . using blind signal separation , @xcite also identified the 11.0 m band as cationic and the 11.2 m band as neutral . recently , however , there have been suggestions of further complexity in the assignments of the 11.0 and 11.2 m bands , which we address here . for one , @xcite showed that _ neutral _ acenes produce emission near 11.0 m. at present it is not clear how significant their contribution will be to the astronomical 11.0 m emission band , as they constitute a small set of the pah family . the nasa ames pah ir spectroscopic database @xcite , hereafter referred to as pahdb , has few included acenes at this time . using pahdb , @xcite showed that _ nitrogen - substituted pahs , or panhs , were required to fit the 11.0 m astronomical emission in ngc 7023 . another possibility is that [ sipah]@xmath23 complexes may contribute to the 11.0 m band , as shown by quantum chemical calculations by @xcite . the resulting ir emission intensities of such complexes are expected to be similar to those of pure pah cations . to confirm this assignment further laboratory and theoretical work are required @xcite . regarding the 11.2 m emission , it has been proposed that its red wing ( out to 11.4 - 11.6 m ) is due to the emission from very small grains ( vsgs ; @xcite ) . if vsg abundances are sufficiently high , they may influence the peak position of the 11.2 m complex . apart from neutral acenes , all assignments of the 11.0 m band point towards a cationic carrier . similarly , apart from vsgs , all assignments of the 11.2 m band point towards neutral pahs . we adopt such charge assignments , leading to the following conclusions : 1 . the 11 m complex : 1 . 11-g1 dominated by cations 2 . 11-g2 mixed charge 3 . 11-g3 mixed charge 4 . 11-g4 dominated by neutrals 5 . 11-g5 dominated by neutrals 2 . the 12.7 m complex : 1 . 12-g1 mixed charge 2 . 12-g2 mixed charge 3 . 12-g3 dominated by cations 4 . 12-g4 dominated by neutrals the mixed behaviour of the 11-g2 component indicates that the traditional 11.0 m band is not purely cationic . the 11-g1 component however dominates the 11.0 m flux , meaning that the emission is still primarily cationic . we estimate the contributions of these two components to the traditional 11.0 m emission in fig . [ fig : stats ] . the 11-g1 component in general carries 60 - 80% of the flux : @xmath24% ( ngc 7023 ) , @xmath25% ( ngc 2023 south ) , @xmath26% ( ngc 2023 north ) and @xmath27% ( m17 ) . the 11-g2 component generally is the complement to these values : @xmath28% ( ngc 7023 ) , @xmath29% ( ngc 2023 south ) , @xmath30% ( ngc 2023 north ) and @xmath31% ( m17 ) . in some instances there is a minor contribution ( less than 5% ) from either the 11-g3 or 11-g4 component . we use the structural similarity indices ( table [ tables : ssim7023 ] ) to quantify the mixed behaviour of the 11-g2 component . to do so , we must make the assumption here that the traditional 11.2 m component traces neutral pahs and the 11-g1 component traces pure cations . we can reach the same conclusions if we instead use the traditional 11.0 m band as a tracer of pure cations , though with greater uncertainty ( since the 11-g2 component is a minor contributor to the 11.0 m emission ) . the similarity indices show that in ngc 7023 , the 11-g2 component is equally similar to the 11.2 m emission ( ssim = 0.73 ) as it is to the 11-g1 emission ( ssim = 0.74 ) . in ngc 2023 south and north , however , the 11-g2 component is more similar to the 11.2 m band than the 11-g1 emission ( ssim = 0.86 versus 0.60 in the southern map and ssim = 0.89 versus 0.75 in the northern map ) . this implies that at minimum 50% of the flux of the 11-g2 component is from neutral pahs . in turn , this then implies that _ the traditional 11.0 m band is roughly @xmath32% neutral and @xmath33% cationic , with uncertainties approaching ten percentage points . _ recently , @xcite reported that the solo ch@xmath0 mode of neutral acenes falls near 11.03 m and argues for a possible contribution from neutral acenes to the 11.0 m band . the results reported here , that there is a small neutral contribution to the 11.0 m band in our sample , provide the first observational evidence in support of a partial origin in a neutral carrier . @xcite applied blind signal separation to ngc 7023 and identified three basis vectors : pah@xmath34 , pah@xmath23 and vsgs . one interesting result of their analysis is that the pah@xmath34 signal , which is associated with neutral pahs , showed a local emission peak at 11.0 m. the authors suggest this is an artifact of the applied method , and may be compensated for by a local minimum in their vsg signal . however , since we have identified a component of the traditional 11.0 m band that is linked to neutral pahs , we suggest that it may be due to the non - cationic component that we have found . in the traditional assignment of the 11.0 and 11.2 m bands , i.e. originating from solo ch@xmath0 modes in cations and neutrals , respectively , the 11.0/11.2 ratio traces the pah charge fraction @xcite . to examine this , one must determine the intrinsic 11.0/11.2 flux ratio of a pah after a single photon excitation event . the authors chose circumcoronene as being representative of a typical pah , which then relates the observed 11.0/11.2 flux ratio to an implied neutral - to - cation fraction . the neutral fraction was shown to decrease in orion from 80% to 65% when moving away from the star , before returning to 80% at further distances . the unexpected diminution of the neutral fraction with increasing distance may reflect dehydrogenation , which would affect the measured 11.0/11.2 m ratio , or it may indicate that circumcoronene is not a reasonable proxy for the total pah population @xcite . our results show that neutral pahs contribute to the 11.0 m band , at approximately the 8 - 16% level . this may be an additional contributing factor to the unusual 11.0/11.2 behaviour observed by @xcite . mean spectra of species in pahdb were presented by @xcite , binned by size and charge state ( their fig . there does not appear to be neutral emission at 11.0 m. this may reflect biases or limitations of the database ( see @xcite ) . the neutral emission we deduce to exist at 11.0 m is a small fraction of the cationic emission at 11.0 m , and thus it may be hidden ( if present in the pahdb spectra ) by averaging over all species . the 11.2 m band displays two asymmetries : a prominent red wing in the range 11.4 - 11.6 m and a narrow peak near 11.20 m that appears only when sufficiently close to the illuminating source ( c.f . [ fig : raw ] ) . we exclude the 11.0 m band from consideration here as it is understood to be a separate band . in our decomposition , the 11-g3 , 11-g4 and 11-g5 components all emit significantly at 11.20 m ( fig . [ fig:112_3pos ] ) . however , the 11-g3 component is much narrower than the others and it is located at exactly 11.20 m , meaning that it might provide some clues about the origin of the peak asymmetry . with the caveats in mind , the results show that the spatial distribution of the 11-g3 component is a mixture of the traditional 11.0 and 11.2 m emission . the 11-g3 component has almost no spectroscopic overlap with the traditional 11.0 m emission , and the 11-g3 component is strongly blended at 11.2 m with the neutral - carrying 11-g4 and 11-g5 bands , which dominate the 11.2 m flux . this implies that _ we identify a cationic contribution to the 11.2 m emission . _ furthermore , the strength of the 11-g1 and 11-g3 components both increase when approaching the star ( as shown in fig . [ fig:112_3pos ] ) . this is complemented by a transition of the 11.2 m peak position from class a(b ) ( near 11.24 m ) to class a ( near 11.20 m ) as reported by @xcite ( see @xcite for classification details ) . since the 11-g3 component is coincident with the nominal class a position , it suggests that it is the relative intensity of the 11-g3 component to that of 11.2 m emission peak that determines the pah class . knowing also that the 11-g3 component has a cationic contribution , this implies that the pah classification of the 11.2 m band is partially moderated by the relative fraction of emission from pah cations to neutrals . this effect is likely in addition to that reported by @xcite , who studied neutral pahs and identified that the class variations , from a to a(b ) , result from changes in the distribution of pah masses . during the class transition from a(b ) to a , the flux in the red wing decreases significantly in ngc 7023 ( c.f . [ fig : raw ] ) , and less so in ngc 2023 south . the spectral asymmetry due to the variable red wing is well known in the literature @xcite . many pah bands display a red wing due to anharmonicity , but the magnitude of variations seen in the observations can not be reconciled with the expected degree of asymmetry from this effect @xcite . possible explanations have been presented in the literature but its origin has not yet been established . suggestions include emission from vsgs @xcite , pah clusters @xcite , pah anions @xcite , low - mass pahs @xcite and superhydrogenated pahs @xcite . @xcite found a spatial separation between neutral pahs and very small grains in ngc 7023 based on blind signal separation . in our decomposition , the flux of the red wing is carried by the 11-g4 component , while the symmetric " 11.2 m emission is carried primarily by the 11-g5 component . as introduced in section [ sec : results_maps ] , the spectral maps of the 11-g4 and 11-g5 components are similar but have subtle differences . we observe in ngc 7023 and 2023 south that the 11-g4 ( red wing ) emission is less extended than the 11-g5 emission , despite peaking in the same map position . the discrepancy with the results of @xcite is likely a consequence of the large width of our 11-g4 component , which has significant contributions at 11.1 m and greater . we conclude that the entire 11 m complex traces the following ( not necessarily unique ) populations : at 11.0 m , pah cations and a small fraction of neutrals ; and at 11.2 m , pah neutrals and a small fraction of pah cations . with our decomposition we are unable to deduce the carrier of the red wing . the 12.7 m band is quite asymmetric , exhibiting a blue - shaded wing @xcite . our decomposition accounts for the blue wing primarily through the flux of the 12-g1 component , which is broad and centered at 12.55 m. in ngc 7023 , the 12-g1 component is very extended , easily encompassing the regions containing strong 11.0 and 11.2 m emission . this suggests that the broad component , as defined here , has no charge preference . in ngc 2023 south , it is also quite extended , though to a lesser extent than in ngc 7023 . the structural similarity indices show that the map of the 12-g1 component is equal parts 11.0 and 11.2 m emission in ngc 7023 . the 12-g1 component is slightly more similar to the 11.2 m emission than the 11.0 m emission in ngc 2023 south . similar results to ngc 2023 south are found in the northern map , suggesting only a weak dependence on charge . this likely originates in the broadness of the component and reflects that the proposed decomposition does not completely disentangle all pah sub - populations . @xcite used density functional theory ( dft ) to compute the absorption spectra of large symmetric pahs and large irregular pahs . near 12.7 m , these authors found that the mean emission from pah cations is blueshifted from the neutral position . in our decomposition , we find the same relationship : the 12-g3 component at 12.72 m is cationic , while the 12-g4 component at 12.78 m is neutral . hence , the intrinsic spectra and observations are consistent with each other . @xcite decomposed the pah emission in ngc 7023 with pahdb . in fitting the 12.7 m emission the authors find that , in the dense region , neutrals are responsible for the majority of the total intensity . in the diffuse region(s ) , the cations instead carry most of the intensity . we observe that the ratio of the cationic 12-g3 emission to the neutral 12-g4 emission is higher in the diffuse region than the dense region , consistent with their result . @xcite also found that the dense region was characterized by small pahs ( those with fewer than 50 carbon atoms ) and the diffuse region was dominated by large pahs ( those with more than 50 carbon atoms ) . within this framework , the 12-g3 component then originates in large pahs preferentially . the 12-g4 component corresponds to emission that was comparable between small and large pahs in the pahdb fit @xcite , suggesting an equal mixture of sizes . it has been noted in the literature that some pah correlation plots involving the 12.7 m band display a bifurcation @xcite . as our results show that the profile of the 12.7 m band can be strongly dependent on ionization , perhaps the bifurcated correlations simply trace the two different charge states of the 12.7 m emission . the 12.7/11.2 ratio has long been understood as probing hydrogen adjacency , as the 11.2 m band is associated with solo ch@xmath0 bending modes and the 12.7 m band with duo and trio ch@xmath0 bending modes @xcite . the astronomical range of 12.7/11.2 intensity ratios has been shown to be consistent with computed spectra of irregular pahs , which have solos , duos , trios and quartets @xcite . pahs with only solos and duos , such as those from the coronene family , lead to 12.7/11.2 ratios that lie at the low end of the astronomical range @xcite . in addition , this ratio can be further enhanced due to the coupling between the ch@xmath0 modes and the c - c ring deformation mode in ( elongated ) armchair pahs contributing to the 12.7 m emission @xcite . through the use of pahdb , @xcite showed that the spatial variation of the 12.7/11.2 ratio in ngc 7023 seems to primarily trace ionization rather than edge structure . based on their database fits , the authors observed that the ionization fraction increases by 200% across ngc 7023 , whereas the hydrogen adjacency only drops by 25% . since we have presented a way of generally disentangling ions from neutrals in the 11 and 12.7 m complexes , we can probe the dependence of the 11.2/12.7 ratio on charge and molecular edge structure .. for the mixed charge bands ( e.g. 11-g2 , 11-g3 ) , we use the spatial maps and structural similarity indices to discern if one charge state appears to be dominant . the 11-g2 and 12-g2 components are found to be on average equally similar to the cationic- and neutral - dominated bands . thus , we assume half of their flux contributes to the total cationic emission , and half to the neutral emission . for the other mixed charge bands ( 11-g3 , 12-g1 ) , we tested a series of different charge fractions ( 25% , 50% , or 75% cationic ) . the fluxes of these components relative to the other fluxes in the calculation are sufficiently low that they do not affect our conclusions . since the 11 m emission is thought to trace solo ch@xmath0 structures , and the 12.7 m emission duo and trio ch@xmath0 structures , we finally measure the solo / duo+trio ratio for cations , and separately for neutrals . ] we found that the ionization fraction ( defined as the flux from cations divided by the flux from both cations and neutrals ) spans a factor of of 3.1 @xmath35 1.4 across the ngc 7023 map . although the uncertainty is large , the derived ionization fraction is very similar to the 200% increase ( i.e. a factor of 3 ) determined by @xcite . to study hydrogen adjacency , we measured the 12.7/11.2 ratio separately for cations and neutrals . for cations , hydrogen adjacency varies across the map by approximately 30% , while for neutrals it varies by about 10% . these results are generally consistent with the 25% value presented by @xcite . we also applied this analysis to ngc 2023 south and found that the ionization fraction spans a range of 1.9 @xmath35 0.8 ( or a 100% increase across the map ) . the hydrogen adjacency varied by approximately 25% for cations , but only 7 or 8% for neutrals . this analysis shows that the 12.7/11.2 ratio depends largely on the ionization fraction and to a lesser extent on molecular edge structure . however , when charge state is taken into account ( by only considering neutral pahs or only pah ions ) , we can trace molecular edge structure of the pah population . the spatial morphology of the 12.7/11.2 emission was compared against the spatial distribution of the 6.2/(@xmath36 m ) fraction and the pahdb - derived fractional emission in pah cations @xcite . the latter two both display smooth gradations across the pdr , whereas the 12.7/11.2 map shows pockets of enhanced ratios . the authors interpreted this as a consequence of the mixed - charge behaviour of the 12.7 m band . we re - examine this relationship in fig . [ fig : map_7023_127 ] . we now include two additional maps : the 12-g3/11.2 and 12-g4/11.2 ratios . we find a stark increase in the contrast between the diffuse and dense regions when examining 12-g3/11.2 instead of 12.7/11.2 . this originates in the fact that , to first order , the neutral dependence is removed . the 12-g4/11.2 ratio shows little emission in the diffuse region and relatively little variation across the region of peak 11.2 m emission . one maximum is observed in the dense region . clearly the 12-g3/11.2 and 12-g4/11.2 ratios are tracing different pah populations . taking these effects into account , it is apparent that the 12.7/11.2 ratio probes both ionization and molecule structure , which both depend on the local physical conditions ( c.f . @xcite ) . it s worth noting that the 12.7/11.2 ratio can be greatly affected by extinction . to illustrate this , we quantify this effect by comparing the 12.7/11.2 ratio before and after correcting for extinction in m17 ( fig . [ fig : extinction ] ) and two regions from @xcite . using the interstellar extinction curves of @xcite , we corrected for extinction . in m17 , the 12.7/11.2 ratio decreases from a range of 0.37 - 0.72 ( before correction ) to 0.35 - 0.52 ( after correction ) . some pixels exhibit a greater than 30% decrease in the 12.7/11.2 ratio . we also examined two regions from the sample of @xcite , using the extinction measurements of @xcite : iras 15384 - 5348 ( a@xmath37 ) and iras 18317 - 0757 ( a@xmath38 ) . the latter source exhibits the largest 12.7/11.2 ratio in the study of @xcite before extinction correction ( 1.49 ) , but this value is near unity after correction ( 1.03 ) . extinction correction thus significantly reduces the large range in 12.7/11.2 ratios observed in regions . with regards to the 12.7/11.2 ratio in ngc 7023 , only three pixels in the map meeting the signal - to - noise criterion are affected by extinction to any substantial degree ( approximately 10 - 15% flux difference in the 12.7 m band ) . these pixels are on the lower edge of the map and do not affect the map or our conclusions . we have examined the spatial and spectral behavior of the 11 and 12.7 m pah emission complexes in high - resolution _ spitzer_/irs maps of ngc 7023 , ngc 2023 south and north and m17 . we have introduced a five - component gaussian decomposition of the 11 m emission and a four - component decomposition of the 12.7 m emission . at 11 m , two components are centered near 11.0 m , one of which is relatively broad and the other relatively narrow . a narrow , weak gaussian component is located near the peak of the 11.2 m complex , slightly offset towards the blue . a strong band carries most of the peak 11.2 m emission , and a final broad emission component is responsible for the red wing of the 11.2 m band . the 12.7 m decomposition consists of one broad feature on the blue wing , two components near 12.75 m , and a narrow feature that appears on the blue wing in some positions . we investigated the spatial distributions of these components in our spectral maps and quantified their similarities the structural similarity index , in addition to flux correlation plots . we have arrived at the following conclusions : 1 . the traditional 11.0 m emission band has a contribution from neutral pahs . we identify a broad cationic feature at 11.00 m and a weaker , narrower mixed - charge state band at 11.00 m. in total , 78 - 88% of the traditional 11.0 m flux is carried by cations , and 8 - 16% of the flux is carried by neutral pahs . this may be supporting evidence for the contribution of neutral acenes to the 11.0 m band @xcite . the traditional broad 11.2 m emission band has a small cationic contribution at 11.20 m. the relative strength of this cationic feature to the broad 11.2 m emission determines the overall peak position of the 11.2 m complex . this implies that the pah classification of the 11.2 m emission is partially determined by the fraction of pah cations to neutrals , in addition to the varying distribution of pah masses reported by @xcite . the variable peak position of the 12.7 m complex can be explained by the relative strengths of two competing gaussian components at 12.72 and 12.78 m. the features are spectroscopically blended , yet they distinctly trace cations and neutrals , respectively , in the spectral maps . 4 . the component responsible for the bulk of the blue - shaded wing of the 12.7 m band appears to be only weakly dependent on charge . this may indicate that the proposed decomposition does not completely disentangle all pah sub - populations . the observed contribution of both cation and neutral pahs to the 12.7 m band supports the use of the 12.7/11.2 intensity ratio as a charge proxy @xcite . however , after accounting for pah charge , structural variations are still probed by the 12.7/11.2 ratio . these results illustrate the power of spectral maps for understanding the complicated spectral profiles of pahs , wherein blended spectral components can be understood as independent spatial components . the next step is to apply this technique to other high - resolution spectral maps as well as integrated spectra of individual objects . this will help improve the quantification of the non - cationic emission at 11.0 m and understand the behaviour of the broad blue 12.7 m wing , which appears to be insensitive to charge based on the applied decomposition . this work and the interpretation of the pah emission bands will strongly benefit from the heightened spectral resolution and sensitivity of the forthcoming james webb space telescope mission . the authors acknowledge support from an nserc discovery grant , nserc acceleration grant and era . mjs acknowledges support from a qeii - gsst scholarship . the irs was a collaborative venture between cornell university and ball aerospace corporation funded by nasa through the jet propulsion laboratory and ames research center @xcite . this research has made use of nasa s astrophysics data system bibliographic services , and the simbad database , operated at cds , strasbourg , france . this work has also made use of the matplotlib python plotting library @xcite and the seaborn python visualization library . the maps of ngc 2023 north and m17 are presented in figs . [ fig : map_2023n ] and [ fig : map_m17 ] , respectively . these are discussed in the main text in sections [ sec:11decomp ] and [ sec:12decomp ] . here we give a brief summary . in general , ngc 2023 north displays a much more complex morphology than ngc 2023 south or ngc 7023 . the peak pah emission appears to be oriented either vertically in this map ( e.g. 11.0 m , 11-g1 emission ) or horizontally ( e.g. 11.2 m , 11-g3 , 11-g4 , 11-g5 , 12-g4 emission ) , with significant overlap . however , the similarities between bands found in ngc 2023 north are generally consistent with those found for ngc 2023 south and ngc 7023 ( see main text ) . likewise , m17 is a complicated environment for analysis . in fig . [ fig : map_m17 ] , we observe that the peak 11.0 and 11.2 m emission regions are nearly coincident . however , the key distinguishing feature seems to be the m17 spur " emission in the upper right corner of the map , which is stronger in the 11.2 m band . based only on this distinguishing feature , key results from studying ngc 7023 and ngc 2023 south are also present here . the most robust trends that the 11.0 m and 12-g3 emission are well - matched , as are the 11.2 m and 12-g4 emission are both observed in m17 when examining the spur , despite its complexity . we examine ratios of band fluxes to probe for the presence of correlations between emission features ( fig . [ fig : corr ] ) . we compare first the 12-g3 and 12-g4 band fluxes to that of the 11.0 m emission ( all normalized to the 11.2 m band ) . we also plot the 12-g3 and 12-g4 band fluxes against that of the 11.2 m band ( then normalized to the 11.0 m emission ) . we identify a strong correlation between the 12-g4 and 11.2 m bands in ngc 7023 and ngc 2023 south and a mild correlation in ngc 2023 north ( with associated weighted pearson correlation coefficients of 0.98 , 0.97 and 0.58 , respectively ) . another strong correlation is identified between the 12-g3 and 11.0 m bands in these same sources , with coefficients 0.96 , 0.89 and 0.87 , respectively . the cross terms " of these correlations , i.e. the 12-g3 band versus the 11.0 m band , and the 12-g3 versus the 11.2 m emission , reveal some residual structure and/or mild correlations . there is clearly some cross - contamination " , but the strongest relationships are nevertheless between the 12-g3 and 11.0 m bands , and between the 12-g4 and the 11.2 m bands . similar correlation plots for the 12-g1 and 12-g2 components are presented in the lower panels of fig . [ fig : corr ] . in general , both the 12-g1 and 12-g2 components correlate with the 11.0 and 11.2 m bands , implying a mixed charge origin . m17 is an unusual case , as the only distinction between the traditional 11.0 and 11.2 m emission in the spectral maps was the m17 spur . the correlation plots involving m17 are also outlying cases , as the 11.0/11.2 ratio is constant .

in this paper we address the inconsistency of the ekf - based slam algorithm that stems from non - observability of the origin and orientation of the global reference frame . we prove on the non - linear two - dimensional problem with point landmarks observed that this type of inconsistency is remedied using the invariant ekf , a recently introduced variant of the ekf meant to account for the symmetries of the state space . extensive monte - carlo runs illustrate the theoretical results .

1510.06263

the problem of simultaneous localization and mapping ( slam ) has a rich history over the past two decades , which is too broad to cover here , see e.g. @xcite . the extended kalman filter ( ekf ) based slam ( the ekf - slam ) has played an important historical role , and is still used , notably for its ability to close loops thanks to the maintenance of correlations between remote landmarks . the fact that the ekf - slam is inconsistent ( that is , it returns a covariance matrix that is too optimistic , see e.g. , @xcite , leading to inaccurate estimates ) was early noticed @xcite and has since been explained in various papers @xcite . in the present paper we consider the inconsistency issues that stem from the fact that , as only relative measurements are available , the origin and orientation of the earth - fixed frame can never be correctly estimated , but the ekf - slam tends to think " it can estimate them as its output covariance matrix reflects an information gain in those directions of the state space . this lack of observability , and the poor ability of the ekf to handle it , is notably regarded as the root cause of inconsistency in @xcite ( see also references therein ) . in the present paper we advocate the use of the invariant ( i)-ekf to prevent covariance reduction in directions of the state space where no information is available . the invariant extended kalman filter ( iekf ) is a novel methodology introduced in @xcite that consists in slightly modifying the ekf equations to have them respect the geometrical structure of the problem . reserved to systems defined on lie groups , it has been mainly driven by applications to localization and guidance , where it appears as a slight modification of the multiplicative ekf ( mekf ) , widely known and used in the world of aeronautics . it has been proved to possess theoretical local convergence properties the ekf lacks in @xcite , to be an improvement over the ekf in practice ( see e.g. , @xcite and more recently @xcite where the ekf is outperformed ) , and has been successfully implemented in industrial applications to navigation ( see the patent @xcite ) . in the present paper , we slightly generalize the iekf framework , to make it capable to handle very general observations ( such as range and bearing or bearing only observations ) , and we show how the derived iekf - slam , a simple variant of the ekf - slam , allows remedying the inconsistency of ekf - slam stemming from the non - observability of the orientation and origin of the global frame . the issue of ekf - slam inconsistency has been the object of many papers , see @xcite to cite a few , where empirical evidence ( through monte - carlo simulations ) and theoretical explanations in various particular situations have been accumulated . in particular , the insights of @xcite have been that the orientation uncertainty is a key feature in the inconsistency . the article @xcite , in line with @xcite , also underlines the importance of the linearization process , as linearizing about the true trajectory solves the inconsistency issues , but is impossible to implement in practice as the true state is unknown . it derives a relationship that should hold between various jacobians appearing in the ekf equations when they are evaluated at the current state estimate to ensure consistency . a little later , the works of g.p . huang , a.i . mourikis , and s. i. roumeliotis @xcite have provided a sound theoretical analysis of the ekf - slam inconsistency as caused by the ekf inability to correctly reflect the three unobservable degrees of freedom ( as an overall rotation and translation of the global reference frame leave all the measurements unchanged ) . indeed , the filter tends to erroneously acquire information along the directions spanned by those unobservable transformations . to remedy this problem , the above mentioned authors have proposed various solutions , the most advanced being the observability constrained ( oc)-ekf . the idea is to pick a linearization point that is such that the unobservable subspace seen " by the ekf system model is of appropriate dimension , while minimizing the expected errors of the linearization points . our approach , that relies on the iekf , provides an interesting alternative to the oc - ekf , based on a quite different route . indeed , the rationale is to apply the ekf methodology , but using alternative estimation errors to the standard linear difference between the estimate and the true state . any non - linear error that reflects a discrepancy between the true state and the estimate , necessarily defines a local frame around any point , and the idea underlying the iekf amounts to write the kalman jacobians and covariances in this frame . we notice and prove here that an alternative nonlinear error defines a local frame where the unobservable subspace is _ everywhere _ spanned by the same vectors . using this local frame at the current estimate to express kalman s covariance matrix will be shown to ensure the unobservable subspace seen " by the ekf system model is _ automatically _ of appropriate dimension . we thus obtain an ekf variant which automatically comes with consistency properties . moreover , we relate unobservability to the inverse of the covariance matrix ( called information matrix ) rather than on the covariance matrix itself , and we derive guarantees of information decrease over unobservable directions . contrarily to the oc - ekf , and as in the standard ekf , we use here the latest , and thus best , state estimate as the linearization point to compute the filter jacobians . in a nutshell , whereas the key fact for the analysis of @xcite is that the choice of the linearization point affects the observability properties of the linearized state error system of the ekf , the key fact for our analysis is that the choice of the error variable has similar consequences . theoretical results and simulations underline the relevance of the proposed approach . robot - centric formulations such as @xcite , and later @xcite are promising attempts to tackle unobservability , but they unfortunately lack convenience as the position of all the landmarks must be revised during the propagation step , so that the landmarks estimated position becomes in turn sensitive to the motion sensor s noise . they do not provably solve the observability issues considered in the present paper , and it can be noted the oc - ekf has demonstrated better experimental performance than the robocentric mapping filter , in @xcite . in particular , the very recent papers @xcite propose to write the equations of the slam in the robot s frame under a constant velocity assumption . using an output injection technique , those equations become linear , allowing to prove global asymptotic convergence of any linear observer for the corresponding deterministic linear model . this is fundamentally a deterministic approach and property , and as the matrices appearing in the obtained linear model are functions of the observations , the behavior of the filter is not easy to anticipate in a noisy context : the observation noise thus corrupts the very propagation step of the filter . some recent papers also propose to improve consistency through local map joining , see @xcite and references therein . although appealing , this approach is rather oriented towards large - scale maps , and requires the existence of local submaps . but when using submap joining algorithm , inconsistency in even one of the submaps , leads to an inconsistent global map " @xcite . this approach may thus prove complementary , if the iekf slam proposed in the present paper is used to build consistent submaps . note that , the iekf slam can also be readily combined with other measurements such as the gps , whereas the submap approach is tailored for pure slam . from a methodology viewpoint , it is worth noting our approach does not bring to bear estimation errors written in a robot frame , as @xcite . although based on symmetries as well , the estimation errors we use are slightly more complicated . finally , nonlinear optimization techniques have become popular for slam recently , see e.g. , @xcite as one of the first papers . links between our approach , and those novel methods are discussed in the paper s conclusion . the paper is organized as follows . in section [ sec:1 ] , the standard ekf equations and ekf - slam algorithm are reviewed . in section [ sec:2 ] we recall the problem that neither the origin nor the orientation of the global frame are observable , but the ekf - slam systematically tends to think " it observes them , which leads to inconsistency . in section [ sec:22 ] we introduce the iekf - slam algorithm . in section [ sec:3 ] we show how the linearized model of the iekf always correctly captures the considered unobservable directions . in section [ sect::tools ] we derive a property of the covariance matrix output by the filter that can be interpreted in terms of fisher information . in section [ sec:4 ] simulations support the theoretical results and illustrate the benefits of the proposed algorithm . finally , the iekf theory of @xcite is briefly recapped in the appendix , and the iekf slam shown to be an application of this theory indeed . the equations of the iekf slam in 3d are then also derived applying the general theory . consider a general dynamical system in discrete time with state @xmath0 associated to a sequence of observations @xmath1 . the equations are as follows : @xmath2 @xmath3 where @xmath4 is the function encoding the evolution of the system , @xmath5 is the process noise , @xmath6 an input , @xmath7 the observation function and @xmath8 the measurement noise . the ekf propagates the estimate @xmath9 obtained after the observation @xmath10 , through the deterministic part of : @xmath11 the update of @xmath12 using the new observation @xmath13 is based on the first - order approximation of the non - linear system , around the estimate @xmath14 , with respect to the estimation errors @xmath15 defined as : @xmath16 using the jacobians @xmath17 , @xmath18 , and @xmath19 , the combination of equations , and yields the following first - order expansion of the error system @xmath20 where the second order terms , that is , terms of order @xmath21 have been removed according to the standard way the ekf handles non - additive noises in the model ( see e.g. , @xcite p. 386 ) . using the linear kalman equations with @xmath22 the gain @xmath23 is computed , and letting @xmath24 , an estimate @xmath25 of the error @xmath26 accounting for the observation @xmath13 is computed , along with its covariance matrix @xmath27 . the state is updated accordingly : @xmath28 the detailed equations are recalled in algorithm [ algo::ekf ] . the assumption underlying the ekf is that through first - order approximations _ of the state error _ evolution , the linear kalman equations allow computing a gaussian approximation of the error @xmath29 after each measurement , yielding an approximation of the sought density @xmath30 . however , the linearizations involved induce inevitable approximations that may lead the filter to inconsistencies and sometimes even divergence . define @xmath31 and @xmath32 through and . define @xmath33 as @xmath34 and @xmath35 as @xmath36 . * propagation * @xmath37 @xmath38 * update * @xmath39 @xmath40 , @xmath41 @xmath42p_{n|n-1 } $ ] for simplicity s sake let us focus on the standard steered " bicycle ( or unicycle ) model @xcite . the state is defined as : @xmath43 where @xmath44 denotes the heading , @xmath45 the 2d position of the robot / vehicle , @xmath46 the position of unknown landmark @xmath47 ( landmarks or synonymously features , constitute the map ) . the equations of the model are : @xmath48 where @xmath49 denotes the odometry - based estimate of the heading variation of the vehicle , @xmath50 the odometry - based indication of relative shift , @xmath51 and @xmath52 their associated noises , and @xmath53 is the matrix encoding a rotation of angle @xmath54 : @xmath55note that a forward euler discretization of the continuous time well - known unicycle equations leads to @xmath50 having its second entry null . more sophisticated integration methods or models including side slip may yet lead to non - zero values of both entries of @xmath56 so we opt for a more general model with @xmath50 . the covariance matrix of the noises will be denoted by @xmath57 with @xmath58 . a general landmark observation in the robot s frame reads : @xmath59+v_n^1 \\ \vdots \\ \tilde h \left [ r(\theta_n)^t \left ( p^k - x_n \right ) \right]+v_n^k \end{pmatrix}\ ] ] where @xmath60 ( or @xmath61 for monocular visual slam ) is the observation of the features at time step @xmath62 , and @xmath8 the observation noise , and @xmath63 is any function . only a subset of the features is actually observed at time @xmath62 . however , to simplify the exposure of the filters equations , we systematically assume in the sequel that all features are observed . we let the output noise covariance matrix be @xmath64 note that , the observation model encompasses the usual range and bearing observations used in the slam problem by letting @xmath65 . if we choose instead the one dimensional observation @xmath66 we recover the 2d monocular slam measurement . note also we do not provide any specific form for the noise in the output : this is because the properties we are about to prove are related to the observability and thus only depend on the deterministic part of the system , so they are in fact totally insensitive to the way the noise enters the system . we merely apply here the methodology of ekf to the slam problem described in section [ sect::slam_problem ] . the first - order expansions , applied to equations , yield : @xmath67 with @xmath68 , @xmath69 and @xmath70 denotes the jacobian of @xmath63 computed at @xmath71 \in { { \mathbb r}}^2 $ ] . the obtained ekf - slam algorithm is recaped in algorithm [ algo::ekf_slam_linear ] . define @xmath31 and @xmath32 as in . define @xmath33 , @xmath35 as in and . * propagation * for all @xmath72 @xmath38 * update * @xmath73 \\ \vdots \\ \tilde h \left [ r(\hat \theta_{n|n-1})^t \left ( \hat p^k_{n|n-1}- \hat x_{n|n-1 } \right ) \right ] \end{pmatrix}$ ] @xmath40 , @xmath41 @xmath42p_{n|n-1 } $ ] in this section we come back to the general framework , . the standard issue of observability @xcite is fundamentally a deterministic notion so the noise is systematically turned off . [ def::non_obs ] we say a transformation @xmath74 of the system - is unobservable if for any initial conditions @xmath75 and @xmath76 the induced solutions of the dynamics with noise turned off , i.e. , @xmath77 yield the same output at each time step @xmath78 , that is : @xmath79 it concretely means that ( with all noises turned off ) if the transformation is applied to the initial state then none of the observations @xmath13 are going to be affected . as a consequence , there is no way to know this transformation has been applied . in line with @xcite we will focus here on the observability properties of the linearized system . to that end we define the notion of non - observable ( or unobservable ) shift which is an infinitesimal counterpart to definition [ def::non_obs ] , and is strongly related to the infinitesimal observability @xcite : [ def::non_obs_first_order ] let @xmath80 denote a solution of with noise turned off . a vector @xmath81 is said to be an unobservable shift of - around @xmath82 if : @xmath83 where @xmath32 is the linearization of @xmath7 at @xmath84 and where @xmath85 is the solution at @xmath62 of the linearized system @xmath86 initialized at @xmath87 , with @xmath88 denoting the jacobian matrix of @xmath89 computed at @xmath90 . in other words ( see e.g. @xcite ) , for all @xmath91 , @xmath85 lies in the kernel of the observability matrix between steps @xmath92 and @xmath62 associated to the linearized error - state system model , i.e. , @xmath93=0 $ ] . the interpretation is as follows : consider another initial state shifted from @xmath82 to @xmath94 . saying that @xmath87 is unobservable means no difference on the sequence of observations up to the first order could be detected between both trajectories . formally , this condition reads : @xmath95 , i.e. , @xmath96 . an estimation method conveying its own estimation uncertainty as the ekf , albeit based on linearizations , should be able to detect such directions and to reflect that accurate estimates along such directions are beyond reach . in the present paper we consider unobservability corresponding to the impossibility to observe the position and orientation of the global frame @xcite . the corresponding shifts have already been derived in the literature . [ prop::first_order_rotations_prelim]@xcite let @xmath97 be an estimate of the state . only one feature is considered , the generalization of the proposition to several features is trivial . the first - order perturbation of the estimate corresponding to an infinitesimal rotation of angle @xmath98 of the global frame consists of the shift @xmath99 with @xmath100 . in the same way , the first - order perturbation of the estimate corresponding to an infinitesimal translation of the global frame of vector @xmath101 consists of the shift @xmath102 . when rotating the global frame the heading becomes : @xmath103 the position of the robot becomes : @xmath104 the position of the feature becomes : @xmath105 stacking these results we obtain the first - order variation of the full state vector ( regarding the rotation only , the effect of infinitesimal translation being trivial to derive ) : @xmath106 [ prop::first_order_rotations]@xcite the shifts of proposition [ prop::first_order_rotations ] that correspond to infinitesimal rotations , are unobservable shifts of - in the sense of definition [ def::non_obs_first_order ] . the intuitive explanation is clear @xcite : `` if the robot and landmark positions are shifted equally along those vectors , it will not be possible to distinguish the shifted position from the original one through the measurements . '' this section recalls using the notations of the present paper , a result of @xcite . it shows the infinitesimal rotations defined in proposition [ prop::first_order_rotations ] are not , in general , unobservable shifts of the system linearized about the trajectory estimated by the ekf . indeed , applying definition [ def::non_obs_first_order ] to in the case of a single feature ( the generalization being straightforward ) with @xmath107 and @xmath108 yields the condition for an infinitesimal rotation of the initial state to be unobservable for the linearized system . this condition writes @xmath109 and boils down to have for any @xmath110 ( see @xcite ) : @xmath111=0\ ] ] where @xmath112 is the jacobian of @xmath63 computed at @xmath113 . for example , if @xmath63 is invertible the condition boils down to @xmath114=0.\label{never_verified}\end{aligned}\ ] ] we see the quantities involved are the updates of the state . as they depend on the noise , there is a null probability for the condition to be respected , and it is always violated in practice . but the point of the present paper is to show that the problem is related to the ( arbitrary in a non - linear context ) choice to represent the estimation error as the linear difference @xmath115 , not to an inconsistency issue inherent to ekf - like methods applied to slam . by devising an ekf - slam based on another estimation error variable , which in some sense amounts to change coordinates , the false observability problem can be corrected . the qualitative reason why this is sufficient is related to the basic cause of false observability : a given fixed shift may or may not be observable depending on the linearization point @xmath116 , as proved by proposition [ prop::first_order_rotations ] . it turns out that the latter property is not inherently related to the slam problem : it is in fact a mere consequence of the errors definition . defining those errors otherwise can dramatically modify the condition . this is the object of the remainder of this article . building upon the theory of the invariant ( i)ekf on matrix lie groups , as described and studied in @xcite , we introduce in this section a novel iekf for slam . in appendix [ primer]-[gen : iekf ] the general theory of the iekf is recalled and slightly extended to account for the very general form of output , and the algorithm derived herein is shown to be a direct application of the theory . to spare the reader a study of the lie group based theory , we attempt to explain in simple terms the iekf methodology on the particular slam example throughout the present section . consider the model equations with state @xmath84 given by . exactly as the ekf , the iekf propagates the estimated state obtained after the observation @xmath10 of through the deterministic part of i.e. , @xmath117 , @xmath118 , @xmath119 for all @xmath72 . to update the predicted state @xmath120 using the observation @xmath13 we use a first order taylor expansion of the error system . but , _ instead of considering the usual state error @xmath121 _ , we rather use the ( linearized ) estimation error defined as follows @xmath122and @xmath123 is analogously defined . for close - by @xmath124 , this represents an error variable in the usual sense indeed , as @xmath125 if and only if @xmath126 . as in the standard ekf methodology , let us see how this _ alternative _ estimation error is propagated through a first - order approximation of the error system . using the propagation equations of the filter , and , we find @xmath127 where terms of order @xmath128 , @xmath129 , and @xmath130 have been neglected as in the standard the ekf handles non - additive noises @xcite . to derive we have used the equalities @xmath131 , @xmath132 : @xmath133 note that , the odometer outputs @xmath134 have miraculously vanished . this is in fact a characteristics - and a key feature - of the iekf approach . let us now compute the first - order approximation of the observation error , using the alternative state error . define @xmath32 as the matrix , depending on @xmath120 only , such that for all @xmath135 defined by , the innovation term @xmath136 \\ \vdots \\ \tilde h \left [ r(\theta_n)^t \left ( p^k - x_n \right ) \right ] \end{pmatrix}-\begin{pmatrix } \tilde h \left [ r(\hat \theta_{n|n-1 } ) ^t \left(\hat p^1_{n|n-1}-\hat x_{n|n-1 } \right ) \right ] \\ \vdots \\ \tilde h \left [ r(\hat \theta_{n|n-1 } ) ^t \left ( \hat p^k_{n|n-1}-\hat x_{n|n-1 } \right ) \right ] \end{pmatrix}\]]is equal to @xmath137 . using that @xmath138 $ ] , and @xmath139 from , we see that @xmath32 is defined as in below . thus the linearized ( first - order ) system model with respect to alternative error writes @xmath140 with @xmath141 , and @xmath142 where @xmath143 is the jacobian of @xmath63 computed at @xmath144 . as in the standard ekf methodology , the matrices @xmath145 allow to compute the kalman gain @xmath23 and covariance @xmath146 . letting @xmath147 be the standardly defined innovation ( see algorithm [ algo::iekf_slam ] just after update " ) , @xmath148 is an estimate of the linearized error @xmath149 accounting for the observation @xmath13 , and @xmath27 is supposed to encode the dispersion @xmath150 . the final step of the standard ekf methodology is to update the estimated state @xmath120 thanks to the estimated linearized error @xmath148 . there is a small catch , though : @xmath151 being not anymore defined as a mere difference @xmath121 , simply adding @xmath123 to @xmath152 would not be appropriate . the most natural counterpart to in our setting , would be to choose for @xmath152 the values of @xmath153 making the right member of equal to the just computed @xmath123 . however , the iekf theory recalled in appendix [ gen : iekf ] , suggests an update that amounts to the latter to the first order , but whose non - linear structure ensures better properties @xcite . thus , the state is updated as follows @xmath154 , with @xmath155 defined by @xmath156where @xmath157 . algorithm [ algo::iekf_slam ] recaps the various steps of the iekf slam . define @xmath31 and @xmath32 as in . define @xmath33 , @xmath35 as in and . * propagation * for all @xmath72 @xmath38 * update * @xmath73 \\ \vdots \\ \tilde h \left [ r(\hat \theta_{n|n-1})^t \left ( \hat p^k_{n|n-1}- \hat x_{n|n-1 } \right ) \right ] \end{pmatrix}$ ] @xmath40 , @xmath41 @xmath42p_{n|n-1 } $ ] in this section we show the infinitesimal rotations and translations of the global frame are unobservable shifts in the sense of definition [ def::non_obs_first_order ] regardless of the linearization points used to compute the matrices @xmath158 and @xmath32 of eq . , a feature in sharp contrast with the usual restricting condition on the linearization points . in other words we show that infinitesimal rotations and translations of the global frame are always unobservable shifts of the system model _ linearized _ with respect to error regardless of the linearization point , a feature in sharp contrast with previous results ( see section [ sect::eks_slam8inconsistency ] and references therein ) . we can consider only one feature ( @xmath159 ) without loss of generality . the expression of the linearized system model has become much simpler , as the linearized error has the remarkable property to remain constant during the propagation step in the absence of noise , since @xmath160 in - . first , let us derive the impact of first - order variations stemming from rotations and translations of the global frame on the error as defined by , that is , an error of the following form @xmath161 [ prop::first_order_rotations_non_linear ] let @xmath162 be an estimate of the state . the first - order perturbation of the _ linearized _ estimation error defined by around 0 , corresponding to an _ infinitesimal _ rotation of angle @xmath98 of the global frame , reads @xmath163 in the same way , an _ infinitesimal _ translation of the global frame with vector @xmath101 implies a first - order perturbation of the error system of the form @xmath164 according to proposition [ prop::first_order_rotations_prelim ] , an infinitesimal rotation by an angle @xmath165 of the true state corresponds to the transformation @xmath166 . @xmath167 and @xmath168 . regarding @xmath151 of eq it corresponds to the variation @xmath169this direction of the state space is seen " by the _ linearized _ error system as the vector @xmath170 . similarly , a translation of vector @xmath171 of the global frame yields the transformation @xmath172 . the effect on the linearized error @xmath151 of is obviously the perturbation @xmath173 neglecting terms of order @xmath174 . we can now prove the first major result of the present article : the infinitesimal transformations stemming from rotations and translations of the gobal frame are unobservable shifts for the iekf linearized model . [ slam : thm : obs ] consider the slam problem defined by equations and , and the iekf - slam algorithm [ algo::iekf_slam ] . let @xmath87 denote a linear combination of infinitesimal rotations and translations @xmath175 of the whole system defined as follows @xmath176 then @xmath87 is an unobservable shift of the linearized system model - of the iekf slam in the sense of definition [ def::non_obs_first_order ] , and this whatever the sequence of true states and estimates @xmath177 : the very structure of the iekf is consistent with the considered unobservability . note that definition [ def::non_obs_first_order ] involves a propagated perturbation @xmath85 , but as here @xmath158 is @xmath178 : we have @xmath179 . thus , the only point to check is : @xmath180 i.e. , @xmath181 . this is straightforward replacing @xmath87 with alternatively @xmath182 and @xmath183 . we obtained the consistency property we were pursuing : the linearized model correctly captures the unobservability of global rotations and translations . as a byproduct , the unobservable seen by the filter is automatically of appropriate dimension . the standard ekf is tuned to reduce the state estimation error @xmath184 defined through the original state variables @xmath124 of the problem . albeit perfectly suited to the linear case , the latter state error has in fact absolutely no fundamental reason to rule the linearization process in a non - linear setting . the basic difference when analyzing the ekf and the iekf is that * in the standard ekf , there is a trivial correspondence between a small variation of the true state and a small variation of the estimation error . but the global rotations of the frame make the error vary in a non - trivial way as recalled in section [ sec:2 ] . * in the iekf approach , the effect of a small rotation of the state on the variation of the estimation error becomes trivial as ensured by proposition [ prop::first_order_rotations_non_linear ] . but the error is non - trivially related to the state , as its definition explicitly depends on the linearization point @xmath116 . many consistency issues of the ekf stem from the fact that the updated covariance matrix @xmath27 is computed before the update , namely at the predicted state @xmath120 , and thus does not account for the updated state s value @xmath152 , albeit supposed to reflect the covariance of the updated error . this is why the oc - ekf typically seeks to avoid linearizing at the latest , albeit best , state estimate , in order to find a close - by state such that the covariance matrix resulting from linearization preserves the observability subspace dimension . the iekf approach is wholly different : the updated covariance @xmath27 is computed at the latest estimate @xmath120 , which is akin to the standard ekf methodology . but it is then indirectly adapted to the updated state , since it is _ interpreted _ as the covariance of the error @xmath123 . and contrarily to the standard case , the definition of this error depends on @xmath152 . more intuitively , we can say the confidence ellipsoids encoded in @xmath27 are attached to a basis that undergoes a transformation when moved over from @xmath120 to @xmath152 , this transformation being tied to the unobservable directions . this prevents spurious reduction of the covariance over unobservable shifts , which are not identical at @xmath120 and @xmath152 . finally , note the alternative error is all but artificial : it naturally stems from the lie group structure of the problem . this is logical as the considered unobservability actually pertains to an _ invariance _ of the model - , that is the slam problem , to global translations and rotations . thus it comes as no surprise the _ invariant _ approach , that brings to bear invariant state errors that encode the very symmetries of the problem , prove fruitful ( see the appendix for more details ) . our approach can be related to the previous work @xcite . indeed , according to the latter article , failing to capture the right dimension of the observability subspace in the linearized model leads to `` spurious information gain along directions of the state space where no information is actually available '' and results in `` unjustified reduction of the covariance estimates , a primary cause of filter inconsistency '' . theorem [ slam : thm : obs ] proves that infinitesimal rotations and translations of the global frame , which are unobservable in the slam problem , are always `` seen '' by the iekf linearized model as unobservable directions indeed , so this filter does not suffer from `` false observability '' issues . this is our major theoretical result . that said , the results of the latter section concern the system with noise turned off , and pertain to an automatic control approach to the notion of observability as in @xcite . the present section is rather concerned with the estimation theoretic consequences of theorem [ slam : thm : obs ] . we prove indeed , that the iekf s output covariance matrix correctly reflects an absence of `` information gain '' along the unobservable directions , as mentioned above , but where the information is now to be understood in the sense of fisher information . as a by - product , this allows relating our results to a slightly different approach to slam consistency , that rather focuses on the fisher information matrix than on the observability matrix , see in particular @xcite . the exposure of the present section is based on the seminal article @xcite . see also @xcite for related ideas applied to slam . consider the system with output . define the collection of state vectors and observations up to time @xmath62 : @xmath185the joint probability distribution of the @xmath186 vector @xmath187 and of the @xmath188 vector @xmath189 is @xmath190the bayesian fisher information matrix ( bifm ) is defined as the following @xmath191 matrix based upon the dyad of the gradient of the log - likelihood : @xmath192[\nabla_{\tilde x_n } \log p(\tilde y_n,\tilde x_n]^t)\]]and note that , for the slam problem it boils down to the matrix of @xcite . this matrix is of interest to us as it yields a lower bound on the accuracy achievable by any estimator used to attack the filtering problem - . indeed let @xmath193 be defined as the _ inverse _ of the @xmath194 right - lower block of @xmath195^{-1}$ ] . this matrix provides a lower bound on the mean square error of estimating @xmath84 from past and present measurements @xmath189 and prior @xmath196 . indeed , for any unbiased estimator @xmath197 : @xmath198[t(\tilde y_n)]^t)\succeq j_n^{-1}\ ] ] where @xmath199 means @xmath200 is positive semi - definite . @xmath201 is called the bayesian or posterior cramr - rao lower bound for the filtering problem @xcite . most interestingly , in the case where @xmath4 and @xmath7 are linear , the prior distribution is gaussian , and the noises are additive and gaussian , we have @xmath202where @xmath27 is the covariance matrix output by the kalman filter . thus , in the linear gaussian case , @xmath203 reflects the statistical information available at time @xmath62 on the state @xmath84 . by extension in the slam literature @xmath203 is often simply referred to as the information matrix , in non - linear contexts also , e.g. when using extended information filters @xcite . in the last section , we have recalled that in the linear gaussian case , the inverse of the covariance matrix output by the kalman filter is the fisher information available to the filter ( this is also stated in @xcite p. 304 ) . in the light of those results , it is natural to expect from any ekf variant , that the inverse of the output covariance matrix @xmath203 reflect an absence of information gain along unobservable directions indeed . if the filter fails to do so , the output covariance matrix will be too optimistic , that is , inconsistent , and wrong covariances yield wrong gains @xcite . the following theorem shows the linearized system model of the iekf allows ensuring the desired property of the covariance matrix . it is our second major result . [ slam : big : thm ] consider the slam problem defined by equations - and the iekf - slam algorithm [ algo::iekf_slam ] . let @xmath87 denote a linear combination of infinitesimal rotations and translations @xmath175 of the whole system , as defined in theorem [ slam : thm : obs ] . @xmath87 is thus an unobservable shift . if the matrix @xmath27 output by the iekf remains invertible , we have at all times : @xmath204 as @xmath205 in , the unobservable shifts remain fixed i.e. @xmath206 . at the propagation step we have : @xmath207 as @xmath208 is positive semi - definite . and at the update step ( see the kalman information filter form in @xcite ) we have : @xmath209 \delta x_0 = \delta x_0^t p_{n|n-1}^{-1 } \delta x_0\ ] ] as @xmath210 as shown in the proof of theorem [ slam : thm : obs ] . thus @xmath211 is non - increasing over time @xmath62 . note that , the proof evidences that if @xmath27 is not invertible , the results of the theorem still hold , writing the iekf in information form . our result essentially means the _ linearized model _ of the iekf has a structure which guarantees that the covariance matrix at all times reflects an absence of spurious " ( bayesian fisher ) information gain over directions that correspond to the unobservable rotations and translations of the global frame . in this section we verify in simulation the claimed properties on the one hand , and on the other we illustrate the striking consistency improvement achieved by the iekf slam . to that end , we propose to consider a similar numerical experiment as in the sound work @xcite dedicated to the inconsistency of ekf and the benefits of the oc - ekf . the iekf is compared here to the standard ekf , the ukf , the oc - ekf , and the ideal ekf , which is the - impossible to implement - variant of the ekf where the state is linearized about the _ true _ trajectory . . we see the indicator stays around 1 for iekf slam and oc - ekf slam over the whole time interval , as expected from a consistent estimation method . the ideal " ekf , where the system is linearized on the true value of the state , yields similar results . to the opposite , we see the ekf is inconsistent , and the ukf also . ] uncertainty envelope computed by the filter . filters whith theoretical properties regarding non - observable directions ( iekf , oc - ekf and ideal ekf ) remedy this problem . ] . bottom plot is a zoom of the first time steps . the information over an infinitesimal perturbation corresponding to a rotation of the whole system is decreasing for the iekf slam , which is a consistent behavior as this perturbation is unobservable . ideal ekf and ocekf give similar results , but ekf and ukf do not . the plot also confirms ekf and ukf slam tend to acquire spurious information over this unobservable direction.,title="fig : " ] . bottom plot is a zoom of the first time steps . the information over an infinitesimal perturbation corresponding to a rotation of the whole system is decreasing for the iekf slam , which is a consistent behavior as this perturbation is unobservable . ideal ekf and ocekf give similar results , but ekf and ukf do not . the plot also confirms ekf and ukf slam tend to acquire spurious information over this unobservable direction.,title="fig : " ] the simulation setting we chose is ( deliberately ) similar to the one used in @xcite ( section 6.2 ) . the vehicle ( or robot ) drives a 15m - diameter loop ten times in the 2d plane , finding on its path 20 unknown features as displayed on figure [ fig::map ] . the velocity and angular velocity are constant ( 1 m / s and 9 deg / s respectively ) . the relative position of the features in the reference frame of the vehicle is observed once every second ( where @xmath63 of is the identity ) . the standard deviation @xmath212 of the velocity measurement on each wheel of the vehicle is @xmath213 of the velocity . this yields the standard deviation @xmath214 of the resulting linear velocity and @xmath215 of the rotational velocity : @xmath216 and @xmath217 , where @xmath218 is the distance between the drive wheels ( see @xcite ) . the features are visible if they lie within a sensing range of 5 m , in which case they are observed with an isotropic noise of standard deviation 10 cm . the initial uncertainty over position and heading is zero - which will prove a condition not sufficient to prevent failure of the ekf . each time a landmark is seen for the first time , its position is initialized in the earth frame using the current estimated pose of the robot , the associated uncertainty is set to a very high value compared to the size of the map , then a kalman update is performed to correlate the position of the new feature withe the other variables . each second , all visible landmarks ( i.e. those in a range of 5 m ) are processed simultaneously in a stacked observation vector . five algorithms are compared : 1 . the classical ekf , described in algorithm [ algo::ekf_slam_linear ] . 2 . the proposed iekf slam algorithm described in algorithm [ algo::iekf_slam ] . the ideal ekf as defined in @xcite , i.e. , a classical ekf where the riccati equation is computed at the true trajectory of the system instead of the estimated trajectory . although not usable in practice , the latter is a good reference to compare with , as it is supposed to be an ekf with consistent behavior . the oc - ekf described in @xcite , which is so far the the only method that guarantees the non - observable subspace has appropriate dimension . 5 . the unscented kalman filter ( ukf ) , known to better deal with the non - linearities than the ekf . before going further , the next subsection introduces the nees indicator used in the simulations to measure the consistency of these methods . classical criteria used to evaluate the performance of an estimation method , like root mean squared ( rms ) error do not inform about consistency as they do not take into account the uncertainty returned by the filter . this point is addressed by the normalized estimation error squared ( nees ) , which computes the average squared value of the error , normalized by the covariance matrix of the ekf . for a sample @xmath219 of error values having dimension @xmath220 , each of them with a covariance matrix @xmath221 of size @xmath222 , the nees is defined by : @xmath223 if each @xmath224 is a zero mean gaussian with covariance matrix @xmath221 , then for large @xmath225 we have nees @xmath226 . the case nees @xmath227 reveals an inconsistency issue : the actual uncertainty is higher than the computed uncertainty . this situation typically occurs when the filter is optimistic as it believes to have gained information over a non - observable direction . the nees indicator will be used , along with the usual rms , to illustrate our solution to slam inconsistency in the sequel . figure [ fig::nees ] displays the nees indicator of the vehicle pose estimate ( heading and position ) over time , computed for 50 monte - carlo runs of the experiment described in section [ sect::exp_setting ] . as expected , the profile of the nees for classical ekf , ideal ekf and oc - ekf is the same as in the previous paper @xcite which inspired this experimental section . note that we used here a normalized version of the nees , making its swing value equal to 1 . we see also that the result is similar for oc - ekf slam and ideal ekf slam : the nees varies between 1 and 1.7 , in contrast to the ekf slam and ukf slam which exhibit large inconsistencies over the robot pose we see here that the iekf remedies inconsistency , with a nees value that remains close to 1 . note that , it performs here even better than oc - ekf and ideal ekf ( whose results are very close to each other ) , in terms of consistency . the basic difference between iekf and these filters lies in using or not the current estimate as a linearization point . uncertainty directions being very dependent from the estimate , what figure [ fig::nees ] suggests is that they may not be correctly captured if computed on a different point . the other aspect of the evaluation of an ekf - like method is performance : regardless of the relevance of the covariance matrix returned by the filter ( i.e. consistency ) , pure performance can be evaluated through rms of the heading and position error , whose values over time are displayed in figure [ fig::heading_rms ] . they confirm an expectable result : solving consistency issues improves the accuracy of the estimate as a byproduct , as wrong covariances yield wrong gains @xcite . selecting a _ single run _ , we can also illustrate the inconsistency issue in terms of covariance and information . figure [ fig::ekf_enveloppe ] displays the heading error for ekf , ukf , iekf , oc - ekf and ideal ekf slam , and the @xmath228 envelope returned by each filter . this illustrates both the false observability issue and the resulting inconsistency of ekf and ukf : the heading uncertainty is reduced over time while the estimation error goes outside the @xmath228 envelope . to the opposite , the behavior of iekf , oc - ekf and ideal ekf is sound . figure [ fig::ekf_enveloppe2 ] shows the map and the landmarks @xmath228 uncertainty ellipsoids : similarly , both ekf and ukf fail capturing the true landmarks positions within the @xmath228 ellipsoids whereas the three over filters succeed to do so . finally , figure [ fig::info_rot ] displays the evolution of the information over a shift corresponding to infinitesimal rotations as defined in theorem [ slam : big : thm ] , that is the evolution over time of the quantity @xmath229 . the theorem is successfully illustrated : the latter quantity is always decreasing for the iekf , ideal ekf , oc - ekf but not for the ekf and the ( slightly better to this respect ) ukf . this work evidences that the ekf algorithm for slam is not inherently inconsistent - at least regarding inconsistency related to unobservable transformations of the global frame - but the choice of the right coordinates for the linearization process is pivotal . we showed that applying the recent theory of the iekf - an ekf ( slight ) variant - leads to provable properties regarding observability and consistency . extensive monte - carlo simulations have illustrated the consistency of the new method and the striking improvement over ekf , ukf , oc - ekf , and more remarkably over the ideal ekf also , which is the - impossible to implement - variant of the ekf where the system is linearized about the true trajectory . note that , the iekf approach may prove relevant beyond slam to some other problems in robotics as well , such as autonomous navigation ( see @xcite ) , and in combination with controllers , notably for motion planning purposes , see @xcite . in @xcite the iekf has proved to possess _ global _ asymptotic convergence properties on a simple localization problem of a wheeled robot , which is a strong property . the iekf has also been patented for navigation with inertial sensors @xcite . nowadays nonlinear optimization based slam algorithms are becoming popular as compared with ekf slam , see e.g. @xcite for one of the first papers on the subject . we yet anticipate a simple ekf slam with consistency properties will prove useful to the research community , the ekf slam having been abandoned in part due to its inconsistency . the general ekf has proved useful in numerous industrial applications , especially in the field of guidance and navigation . it has the benefits of being 1-recursive , avoiding to store the whole trajectory and 2-suited to on - line real - time applications . moreover the aerospace and defense industry has developed a corpus of experience for its industrial implementation and validation . and the iekf is a variant that , being in every respect similar to ekf , retains all its advantages , but which possesses additional guaranteed properties . note also that , all the improvements of the ekf for slam such as e.g. , the slam of @xcite and sparse extended information filters @xcite , can virtually be turned into their invariant counterpart . the high dimensional optimization formulation of the slam problem being prone to local minima , having an accurate initial value ( i.e. a small initial estimation error ) is very critical @xcite . the iekf slam algorithm proposed in the present paper may thus be advantageously used to initialize those methods in challenging situations . besides , we anticipate our approach based on symmetries could help improve ( at least first order ) optimization techniques for slam . to understand why , assume by simplicity the sensors to be noise free . then , moving a candidate trajectory along unobservable directions will not change the cost function , and an efficient optimization algorithm should account for this . and when a gradient descent algorithm is used , only a first - order expansion of the cost function is considered . our lie group approach will allow defining steepest descent directions in a alternative geometric way , that will `` stick '' to the unobservable directions , and the corresponding update will move along the ( lie group ) state space in a non - linear yet relevant way . this issue is left for future work , but a thorough understanding of the interest of the invariant approach for the ekf , is a first step in this direction . the authors would like to thank cyril joly for his advice . in this section we provide more details on the iekf theory on matrix lie groups , and show how the underlying lie group structure of the slam problem has been used indeed to build the iekf slam algorithm [ algo::iekf_slam ] . we also provide the iekf equations for 3d slam . for more information on the iekf see @xcite and references therein . a matrix lie group @xmath230 is a subset of square invertible @xmath231 matrices @xmath232 verifying the following properties : @xmath233 where @xmath234 is the identity matrix of @xmath235 . if @xmath236 is a curve over @xmath230 with @xmath237 , then its derivative at @xmath238 necessarily lies in a subset @xmath239 of @xmath240 . @xmath239 is a vector space and it is called the lie algebra of @xmath230 . it has same dimension @xmath241 as @xmath230 . thanks to a linear invertible map denoted by @xmath242 , one can advantageously identify @xmath239 to @xmath243 . besides , the vector space @xmath239 can be mapped to the matrix lie group @xmath230 through the classical matrix exponential @xmath244 . thus , @xmath245 can be mapped to @xmath230 through the lie exponential map defined by @xmath246 for @xmath247 . this map is invertible for small @xmath151 , and we have @xmath248 . the well - known baker - campbell - hausdorff ( bch ) formula gives a series expansion for the product @xmath249 . in particular it ensures @xmath250 , where @xmath251 is of the order @xmath252 . for any @xmath253 , the adjoint matrix @xmath254 is defined by @xmath255 for all @xmath256 . we now give explicit formulas for two groups of particular interest for the slam problem . this famous group in robotics can be defined using homogeneous matrices , i.e. , @xmath258 . @xmath259 , then @xmath260 , where @xmath261 and @xmath262 . we have @xmath263 . the lie exponential writes @xmath264 where @xmath157 . we have @xmath265 . we now introduce a simple extension of @xmath257 , inspiring from preliminary remarks in @xcite . for @xmath267 and @xmath268 , consider the map @xmath269 defined by @xmath270 \vdots & \\[-0.5ex ] 0_{1,2 } & \end{array } \right)\ ] ] and let @xmath271 be defined by @xmath272and denote it by @xmath273 . note that , we recover @xmath257 for @xmath159 , i.e. , @xmath274 . letting @xmath261 and @xmath275 yields @xmath276 \vdots & \\[-0.5ex ] 0_{1,2 } & \end{array } \right)$ ] and @xmath277 . it turns out , by extension of the @xmath257 results , that there exists a closed form for the lie exponential @xmath278 that writes @xmath279 with @xmath157 . the @xmath280 is also easily derived by extension of @xmath257 , but to save space , we only display it once : @xmath281 is defined as the matrix @xmath282 of eq . this section is a summary of the iekf methodology of @xcite . let @xmath230 be a matrix lie group . consider a general dynamical system @xmath283 on the group , associated to a sequence of observations @xmath1 , with equations as follows : @xmath284 @xmath285 where @xmath286 is an input matrix which encodes the displacement according to the evolution model , @xmath287 is a vector encoding the model noise , @xmath288 is the observation function and @xmath289 the measurement noise . the iekf propagates an estimate obtained after the previous observation @xmath10 through the deterministic part of : @xmath290 to update @xmath291 using the new observation @xmath13 , one has to consider an estimation error that is _ well - defined _ on the group . in this paper we will use the following right - invariant errors @xmath292 which are equal to @xmath234 when @xmath293 . the terminology stems from the fact they are invariant to right multiplications , that is , transformations of the form @xmath294 with @xmath253 . note that , one could alternatively consider left - invariant errors but it turns out to be less fruitful for slam . the iekf update is based upon a first - order expansion of the non - linear system associated to the errors around @xmath234 . first , compute the full error s evolution @xmath295note that the term @xmath296 has disappeared ! this is a key property for the successes of the invariant filtering approach @xcite . to linearize this equation we define @xmath297 around @xmath234 through @xmath298as in the standard non - additive noise ekf methodology @xcite all terms of order @xmath299 , are assumed small and are neglected . using the bch formula , and neglecting the latter terms , we get @xmath300 using the local invertibility of @xmath301 around @xmath302 , we get the following linearized error evolution in @xmath243 : @xmath303 where @xmath304 and @xmath305 . to linearize the output error , we now slightly adapt the iekf theory @xcite to account for the general form of output . note that , @xmath306 . as @xmath149 is assumed small , and as @xmath307 , a first - order taylor expansion in @xmath247 arbitrary , allows definit @xmath32 as follows @xmath308 as in the standard theory , the kalman gain matrix @xmath23 allows computing an estimate of the linearized error after the observation @xmath13 through @xmath309 , where @xmath310 . recall the state estimation errors defined by - are of the form @xmath311 , that is , @xmath312 . thus an estimate of @xmath313 after observation @xmath13 which is consistent with - , is obtained through the following lie group counterpart of the linear update @xmath314 the equations of the filter are detailed in algorithm [ algo::iekf ] . choose initial @xmath315 and @xmath316 [ algo::iekf ] define @xmath32 as in and let @xmath317 and @xmath318 . define @xmath33 as @xmath34 and @xmath35 as @xmath36 . * propagation * @xmath319 @xmath320 * update * @xmath321 @xmath322 , @xmath41 @xmath42p_{n|n-1 } $ ] the lie group that underlies the slam problem , is @xmath273 introduced in appendix [ sect::tuto_se23 ] . let us apply the general theory of the iekf to this group . to define the lie group counterpart @xmath313 of the state @xmath84 defined by , we let @xmath323 . the model equations write @xmath324 with @xmath325 . at the propagation step , the iekf propagates the estimate through the corresponding deterministic equations @xmath326with @xmath327 . a simple matrix multiplication shows that @xmath328 , where @xmath329 \\ p^1_n-[r \left ( \theta_n -{\hat \theta}_{n|n-1}\right ) \hat p^1_{n|n-1 } ] \\ \vdots \\ p^k_n- [ r \left ( \theta_n -{\hat \theta}_{n|n-1}\right ) \hat p^k_{n|n-1 } ] \end{pmatrix}\]]and where @xmath330 is defined analogously . the linearized error is defined by , that is , @xmath331 . as terms of order @xmath332 are to be neglected in the linearized equations , it suffices to compute a first - order approximation of @xmath333 . first note that @xmath149 defined by is a linear approximation to @xmath334 , that is , @xmath335 . this readily implies @xmath336recalling @xmath337 we see is a first approximation of @xmath333 as defined in indeed . the way the linearized error propagates has already been computed and consists of , which is the same as the first equation of . it exactly matches what can be expected from the general theory , that is , equation , recalling that @xmath282 of is the map @xmath338 of the group @xmath273 indeed . extending the group @xmath266 to the 3d case , and applying the general iekf theory of section [ gen : iekf ] , we derive in the present section an iekf for 3d slam . due to space limitations an as it is not the primary object of the present paper we pursue extreme brevity of exposure . see also @xcite . note that , although the 3d slam equations make use of rotation matrices , they are in fact totally intrinsic : when using quaternions ( recommended ) or euler angles ( not recommended ) they write the same as the group @xmath342 we introduce does in fact not depend on a specific representation of rotations . the equations of the robot in 3d and in continuous time write : @xmath343 where @xmath344 is a rotation matrix that represents the robot s orientation at time @xmath345 , @xmath346 denotes the angular velocity of the robot measured by a gyrometer or by odometry ( in combination with a unicycle model for a terrestrial vehicle ) , @xmath347 the velocity in the robot s frame , and @xmath348 is the position of landmark @xmath47 , and where @xmath349 for @xmath350 denotes the skew symmetric matrix of @xmath351 such that for any @xmath352 we have @xmath353 . finally @xmath354 and @xmath355 denote ( resp . ) the noise on angular and linear velocities . although the theory of iekf could very well be applied directly to this continuous time dynamics as in @xcite , we apply it here to a discretized model , to be consistent with the rest of the article . although exact discretization of the noisy model on the group is beyond reach @xcite , letting @xmath356 be the time step , the following first - order integration scheme is widely used : @xmath357\omega_n , \quad x_n = x_{n-1}+r_{n-1 } ( v_n+ w_n^v ) , \\ p_n^j & = p_{n-1}^j,\quad 1\leq j\leq k \end{aligned}\ ] ] where the increments @xmath358 are obtained solving the noise - free initial conditions during the @xmath62-th time step with initial condition @xmath359 , and where the following discrete noise @xmath360 is obtained by integration of the corresponding white noises . note that , this scheme is accurate to first - order terms in @xmath356 . a general landmark observation in the car s frame reads : @xmath361+v_n^1 \\ \vdots \\ \tilde h \left [ r_n^t \left ( p^k - x_n \right ) \right]+v_n^k \end{pmatrix}\ ] ] where @xmath362 ( or @xmath363 for monocular visual slam ) is the observation of the features at time step @xmath62 , and @xmath8 the observation noise . we let the output noise covariance matrix be @xmath364 ( not to be confused with the rotation @xmath35 ) . the lie group that underlies the problem is the group @xmath365 that we introduce as follows . for @xmath366 and @xmath367 let @xmath368 \vdots & \\[-0.5ex ] 0_{1,3 } & \end{array } \right)\ ] ] and let @xmath369 be defined as @xmath370and denote it by @xmath365 . we then have @xmath371 \vdots & \\[-0.5ex ] 0_{1,3 } & \end{array } \right)$ ] and @xmath372 . for @xmath373 , by extension of the @xmath374 results , we have the closed form : @xmath375 where @xmath376 . as easily seen by analogy with @xmath374@xmath377 \vdots & \\[-0.5ex ] ( p^k)_\times r & \end{array } \right)\end{aligned}\ ] ] let the state be @xmath378 , and let @xmath379 be @xmath380 , and let @xmath116 and @xmath381 be their estimated counterparts . it is easily seen that up to terms that will disappear in the linearization process anyway , the model for the state is mapped through @xmath382 defined at , to a model of the form . using the matrix logarithm , define @xmath384 as the solution of @xmath385 = r\hat r^{t}$ ] . neglecting terms of order @xmath386 , we have @xmath387 . a first order identification as in appendix [ lnz : sec ] thus yields @xmath388 as a vector that satisfies the definition @xmath389 up to terms of order @xmath390 ) . as in the standard ekf theory , the iekf propagates an estimate obtained after the previous observation @xmath10 through the deterministic part of , or equivalently in matrix form . thus , the propagation equation is given by where @xmath391 and @xmath392 as a direct application of the theory . let @xmath391 , @xmath32 as in , @xmath396 using define @xmath33 by . @xmath397 is the observation noise cov . propagation * for all @xmath72 @xmath38 * update * @xmath398 \\ \vdots \\ \tilde h \left [ r_{n|n-1}^t \left ( \hat p^k_{n|n-1}- \hat x_{n|n-1 } \right ) \right ] \end{pmatrix}$ ] @xmath399 , @xmath41 @xmath42p_{n|n-1 } $ ] t. bailey , j. nieto , j. guivant , m. stevens , and e. nebot . consistency of the ekf - slam algorithm . in _ intelligent robots and systems , 2006 ieee / rsj international conference on _ , pages 35623568 . ieee , 2006 . s. bonnabel , p. martin , and e. salaun . invariant extended kalman filter : theory and application to a velocity - aided attitude estimation problem . in _ ieee conference on decision and control _ , pp . 1297 - 1304 , 2009 . g. p. huang , a. mourikis , st . an observability - constrained sliding window filter for slam . in _ intelligent robots and systems ( iros ) , 2011 ieee / rsj international conference on _ , pages 6572 . ieee , 2011 . g. p. huang , a. i. mourikis , and s. i roumeliotis . analysis and improvement of the consistency of extended kalman filter based slam . in _ robotics and automation , 2008 . icra 2008 . ieee international conference on _ , pages 473479 . ieee , 2008 . s. j. julier and j. k. uhlmann . a counter example to the theory of simultaneous localization and map building . in _ robotics and automation , 2001 . proceedings 2001 icra . ieee international conference on _ , volume 4 , pages 42384243 . ieee , 2001 . k. w. lee , w. s. wijesoma , and j. i. guzman . on the observability and observability analysis of slam . in _ intelligent robots and systems ( iros ) , 2006 ieee / rsj international conference on _ , pages 35693574 . ieee , 2006 . a. martinelli , n. tomatis , and r. siegwart . some results on slam and the closing the loop problem . in _ intelligent robots and systems ( iros ) , 2012 ieee / rsj international conference on _ , pages 334339 . ieee , 2005 . s. thrun , y. liu , d. koller , a. y. ng,2 . ghahramani and h. durrant - whyte . simultaneous localization and mapping with sparse extended information filters . in _ the international journal of robotics research _ , 23(7 - 8 ) : 693 - 716 , 2004 . l. zhao , s. huang , and g. dissanayake . linear slam : a linear solution to the feature - based and pose graph slam based on submap joining . in _ intelligent robots and systems ( iros ) , 2013 ieee / rsj international conference on . _ ieee , 2013 .

obscured or narrow - line active galaxies offer an unobstructed view of the quasar environment in the presence of a luminous and vigorously accreting black hole . we exploit the large new sample of optically selected luminous narrow - line active galaxies from the sloan digital sky survey at redshifts , in conjunction with follow - up observations with the low dispersion survey spectrograph ( ldss3 ) at magellan , to study the distributions of black hole mass and host galaxy properties in these extreme objects . we find a narrow range in black hole mass (log/ ) and eddington ratio (log ) for the sample as a whole , surprisingly similar to comparable broad - line systems . in contrast , we infer a wide range in star formation properties and host morphologies for the sample , from disk - dominated to elliptical galaxies . nearly one - quarter have highly disturbed morphologies indicative of ongoing mergers . unlike the black holes , which are apparently experiencing significant growth , the galaxies appear to have formed the bulk of their stars at a previous epoch . on the other hand , it is clear from the lack of correlation between gaseous and stellar velocity dispersions in these systems that the host galaxy interstellar medium is far from being in virial equilibrium with the stars . while our findings cast strong doubt on the reliability of substituting gas for stellar dispersions in high luminosity active galaxies , they do provide direct evidence that luminous accreting black holes influence their surroundings on a galaxy - wide scale .

0907.1086

over recent years the cosmological status of accreting supermassive black holes ( bhs ) has been elevated from an extreme rarity to a ubiquitous phase in the life cycle of all massive galaxies ( e.g. , * ? ? ? * ) . there appears to be direct evidence for a connection between bh and galaxy growth in the tight scaling relations between bh mass and bulge properties ( e.g. , the @xmath7 relation ; * ? ? ? nonetheless , all of the classic conundrums regarding bh remain . there is not yet a reliable measurement of the length of typical bh growth episodes @xcite , or the typical mode of bh growth ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? furthermore , while it has been argued both on observational ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) and theoretical @xcite grounds that star formation and accretion activity are temporally coincident , quantitative stellar population constraints for large samples of accreting bhs at high luminosities are notoriously difficult to obtain ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? . the accretion power from a luminous active galactic nucleus ( agn ) may easily outshine the surrounding host galaxy by a factor of 100 , which makes studies of agn host galaxy stellar populations and morphologies a considerable challenge . occasionally nature is particularly kind and puts obscuring material between us and the accretion disk . in these cases , we know that an accreting bh is present due to the ratios of strong emission lines that imply a very hard ionizing spectrum ( e.g. , * ? ? ? * ) , and bhs selected this way are known as narrow - line or type 2 agns . these objects are an ideal tool with which to study agn host galaxies , since the galaxy to agn contrast ratio is maximized . much of what we know about the stellar populations of agn host galaxies is derived from narrow - line objects ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? however , until recently there were not large samples of homogeneously selected , _ luminous _ narrow - line agns . the situation has been remedied by the large spectroscopic database provided by the sloan digital sky survey ( sdss ; * ? ? ? @xcite used the strong and ubiquitous [ ] @xmath8 line emission as a proxy for bolometric luminosity in order to select a sample of narrow - line agns as luminous as any known broad - line active galaxies with @xmath9 . more recently @xcite presented a sample of nearly 1000 narrow - line agns with @xmath10 and bolometric luminosities as high as @xmath11 erg s@xmath12 or more . we have initiated a program to obtain deeper follow - up long - slit spectroscopy of the reyes et al . sample at @xmath13 . our primary goal is to study the evolutionary state of the host galaxies of vigorously accreting bhs . both stellar velocity dispersions ( @xmath14 ) and stellar population age can be measured from the stellar continua . from @xmath14 we derive not only a mass - scale for the galaxy , but also an estimate of the bh mass and in turn eddington fractions for the sample , assuming that the @xmath7 relation holds ( which may not be the case for active galaxies at these redshifts ; e.g. , * ? ? * ; * ? ? ? * ) . by comparing the structural properties of the hosts to those of inactive galaxies at a similar mass , combined with our measurements of stellar ages , we can hope to determine whether the host galaxies are evolving coevally with their bhs . of course , as we discuss in 7 , it is not yet clear whether broad- and narrow - line agns are truly identical populations seen at different orientations ( as suggested by unification ; * ? ? ? * ) , or whether they represent an evolutionary sequence ( as suggested by the differences in star formation properties , among other things , e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) . throughout we assume the following cosmological parameters to calculate distances : @xmath15 km s@xmath12 mpc@xmath12 , @xmath16 , and @xmath17 . our goal was to select nearby targets , for which we could hope to resolve the emission - line structure , that are nevertheless at the upper end of the luminosity function . we thus limited our attention to targets with @xmath18 , such that [ ] is accessible in the observing windows ( see below ) and with @xmath19}}}}\geq 10^{42}$ ] erg s@xmath12 . while arbitrary , our luminosity cut is chosen to match the classic definition of a quasar . altogether , we obtained spectroscopic observations for 15 targets . our sample spans , by construction , a much narrower range in redshift ( and thus luminosity ) than the parent sample . all of the data discussed in this paper were collected over two observing runs using the low dispersion survey spectrograph ( ldss3 ) at the clay - magellan telescope on las campanas . ldss3 is a wide - field ( 8.3 diameter ) imaging spectrograph , which was upgraded from the original ldss2 @xcite to have increased sensitivity in the red . for each target we obtained 25 min acquisition images in @xmath20 and @xmath21 , and sometimes @xmath22 . we then spent at least 1 hr per target at the primary slit position , as well as additional time as noted in table 1 , using a 1@xmath234slit . slit position angles were chosen by manual inspection of the sdss color images , with an eye to investigate color gradients or low - surface brightness material ( table 1 ) . the seeing ranged from 06 to 15 over the two runs , but was typically 1 ; with a plate scale of 0189 the spectra are well sampled in the spatial direction . two spectroscopic settings were required in order to ensure that [ ] was observed over the full redshift range of our sample . the primary setting was the reddest position of the vph - blue grism , which covers @xmath24 . for the highest redshift targets , we needed a redder setting , and used the bluest setting of the vph - red grism , which covers @xmath25 . our velocity resolution was @xmath26 km s@xmath12 for both settings . each night , in addition to the science targets , we observed two to four spectro - photometric standards covering a range of airmass . finally , over the course of the runs we also observed a large library of velocity template stars , consisting mostly of g m giants . the data reduction followed standard procedures , with the exception of pattern - noise removal as discussed below . we first performed cosmic - ray 0.2 in removal on each individual frame using the spectroscopic version of lacosmic @xcite . we performed bias - subtraction , flat - fielding , wavelength calibration , pattern noise removal ( when necessary ; see below ) and rectification using the carnegie observatories reduction package cosmos . the rectification works extremely well ; in the rectified images the trace centroid typically migrates by less than two spatial pixels across the full spectrum , with the maximum drift being five pixels . cosmos optionally performs sky subtraction prior to rectification @xcite , but we have found that for the purposes of stellar velocity dispersion measurements we achieve significantly better sky subtraction using _ apall _ in iraf on the rectified , but not sky - subtracted , frames . we have used several apertures for spatial extractions , but for most of this paper we focus on a 225 spatial extraction , which is well - matched to the point - spread function ( psf ) for the majority of our observations . generally we did not use optimal extraction @xcite in order to avoid emission - line clipping . once the spectra were extracted , we performed flux calibration , minor shifts to the wavelength calibration and telluric absorption corrections using idl routines , following methods described by @xcite . these routines provide considerable flexibility and control in the selection of bandpasses for spectrophotometric calibration used to perform a spline fit to the continuum . small corrections are made , if necessary , to the wavelength solutions using a cross - correlation with sky lines , and a heliocentric correction is applied . our spectrophotometric standards are typically white dwarfs , and thus the smooth continuum is ideally suited to telluric absorption removal as well ( e.g. , * ? ? ? * ; * ? ? ? below we show that our flux calibration agrees reasonably well with the sdss by comparing flux measurements of the [ ] line . the final spectra have a considerably higher signal - to - noise ratio ( s / n ) than the sdss spectra . for @xmath27 , the median s / n at @xmath28 is 150 per pixel , which is typically 4 times higher than the sdss , while for @xmath29 , the median s / n over the same region is 11 per pixel and 8 times higher than the sdss spectra . -0 mm ldss3 , particularly during the march 2007 observation period , had intermittent but substantial pattern noise . pattern noise appears as periodic striping in the data , with a typical scale of a few pixels and an amplitude ( in these data ) of 50100 counts . the amplitude and phase vary in time and so standard calibrations do not remove it but there is a cosmos task that fourier transforms the data and removes the high - frequency noise . our concern is that the frequency of the pattern noise is comparable in scale to that of the expected stellar velocity dispersions , potentially leading to biases in our final measurements . in order to test this possibility , we have performed three independent pattern - noise removal runs for the first night , which has particularly bad pattern noise . we have measured the dispersions independently for each of these spectra . the rms in the measured @xmath14 values for the four objects range from 3 to 20 km s@xmath12 , and , most crucially , we see no systematic differences in the measured dispersion for different versions of the pattern - noise removal . we conservatively add an additional 20 km s@xmath12 uncertainty in quadrature to the measured dispersions that suffer from pattern noise . one of the primary uncertainties in interpreting observations of narrow - line quasars comes from their unknown bh masses . for the median @xmath30}}$]@xmath31 erg s@xmath12 of the magellan sample , we estimate that the bolometric luminosity is @xmath32 erg s@xmath12 , which suggests bh masses of @xmath33 @xmath3 if they emit at or below their eddington limits [ where we take the eddington luminosity to be @xmath34 ( @xmath2/@xmath3 ) erg s@xmath12 ] . however , given the scatter between [ ] and bolometric luminosity , the range of bolometric luminosities in the sample , and their ( presumed ) range of intrinsic eddington ratios , it would be nice to have an independent estimate of the bh masses . at the same time , it would be very useful to have a mass scale for the host galaxies . although it is indirect , if we assume that the @xmath7 relation holds , then we can use measurements of bulge velocity dispersions both to provide an estimate of the bh masses in the sample and as a proxy for the mass of the host galaxy ( e.g. , * ? ? ? we describe the @xmath14 measurements and uncertainties ; the bh masses are presented in 5 . there are a variety of commonly used techniques for measuring stellar velocity dispersions in galaxies . for computational expediency , most of them rely on fourier techniques , including cross - correlation @xcite , the fourier quotient method @xcite , and the fourier correlation quotient @xcite . it is also possible to directly compare broadened templates with data in pixel space ( e.g. , * ? ? ? direct - pixel fitting methods , while relatively computationally expensive , provide many benefits over fourier techniques ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) , including the ability both to fit over restricted regions and to mask complicating emission or absorption features . as described in detail in @xcite , even faint emission lines can lead to substantial biases in derived @xmath14 measurements if not properly accounted for . our galaxy models are built from a linear combination of stellar spectra . the program object is de - redshifted using the sdss redshift , but inevitably small velocity shifts are present between the templates -0.2 in and the target . thus , each template is shifted to zero velocity , diluted by a constant or power - law component @xmath36 , convolved with a gaussian @xmath37 , and multiplied by a polynomial @xmath38 : @xmath39 + c(\lambda ) \}\ ] ] the nonlinear levenberg - marquardt algorithm , as implemented by mpfit in idl @xcite , is used to minimize @xmath40 in pixel space between the galaxy and the model . our code is based on that presented in @xcite , except that the model spectrum is actually a linear combination of stars ( typically four in this work ) with their relative amplitudes as additional free parameters . this version of the code was developed to model the high - quality nuclear spectra from the palomar spectroscopic survey of nearby galaxies @xcite and is presented and extensively tested in @xcite . unless otherwise noted we have used the following four stars as our template set : hd 26574 ( f2 iii ) , hd 107950 ( g5 iii ) , hd 18322 ( k1 iii ) , and hd 131507 ( k4 iii ) . example fits are shown in figure 1 . while the primary purpose of the multi - template fitting is to achieve reliable fits , we also extract crude stellar population information in 7 . in order to use the full spectral range of the observations , we can not rely only on our own template stars , whose wavelength coverage does not go blueward of @xmath414300 . rather , we utilize a large library of high - s / n , high - spectral resolution ( @xmath42 km s@xmath12 ) stellar templates from @xcite . it is best to use template stars observed with an identical instrumental setup as the targets , so that the measured broadening can be ascribed solely to internal kinematics in the galaxy . however , the valdes templates not only allow us to include spectral features such as the g - band @xmath43 to the highest redshifts , but also provide a wide range of spectral types to help minimize the impact of template mismatch . we thus use a simple bootstrapping procedure to remove the impact of differing instrumental resolutions on the measured dispersions . first of all , we measure the intrinsic broadening of the -0 mm ldss3 template stars using the valdes library . while we did not observe identical stars , we can match the stars in spectral type nearly exactly . we expect the relative broadening to be @xmath44 km s@xmath12 . we measure a mean broadening over 10 stars of 60@xmath457 km s@xmath12 . dividing the sample into the two runs demonstrates the stability of the spectrograph ; we find 61@xmath458 km s@xmath12and 59@xmath457 km s@xmath12 , respectively . here and throughout , when we quote errors on mean quantities , we are quoting the standard deviation of the distribution rather than the error on the mean . we then confirm that our conversion works properly by measuring the velocity dispersions of our targets with the valdes templates and comparing the values to those measured from our internal templates , again being careful to match the spectral types of the template stars and the fitting regions . the resulting dispersions are shown in figure 2 , where the reported valdes dispersions have been reduced by @xmath46 km s@xmath12 in quadrature . it appears that our conversion is quite robust . from now on we report results based entirely on the valdes stars with the above correction , conservatively adding 7 km s@xmath12 in quadrature to our error budget . the resulting @xmath14 measurements are shown in table 2 . in the interest of uniformity , in all cases we use the spectral region 41005400 , which is accessible for every observation . we follow @xcite and mask out the @xmath47 @xmath48 triplet . while the @xmath47 triplet is one of the highest equivalent width ( ew ) features in this spectral region , the well - known correlation between the [ mg / fe ] ratio and @xmath14(e.g . * ; * ? ? ? * ) leads to a systematic overestimate of @xmath14 based on @xmath47 @xcite . -0 mm in order to increase our sample size , we have also considered all of the targets in the @xcite sample . since most of the spectra do not have the required s / n for reliable velocity dispersion measurement , we have limited our attention to objects with a median s / n of @xmath4915 per pixel between 5080 and 5400 , and an ew in the ca k line greater than 3 . there are 547 objects with @xmath50 } > 10^{42}$ ] erg s@xmath12 ; of those , 111 make our cut . the sdss velocity dispersions are measured in an identical fashion as the fits described above , using the same mix of valdes template stars . in this case , we assume an instrumental resolution of 71 km s@xmath12 for the sdss spectra ( see discussion in * ? ? ? the sdss - derived @xmath30}}$ ] values , our measured velocity dispersions , and [ ] line widths for this additional sample are all presented in table 3 . two objects are unresolved ( @xmath14@xmath51 km s@xmath12 ) . there are only two objects from the magellan sample for which we can measure reliable dispersions from the sdss spectra , and the resulting dispersions agree with ours within the quoted uncertainties . our quoted uncertainties are a quadrature sum of the formal errors from the fit and the 7 km s@xmath12 uncertainty incurred in converting to the valdes system . our typical uncertainties thus derived range from @xmath52 for the lower - redshift targets to @xmath53 for the most distant and luminous targets . as the s / n in the spectrum decreases , our ability to recover the true dispersion decreases . also , an increasing contribution to the continuum from sources apart from old stars lowers their ew and thus the s / n . to a small degree the relative incurred error also depends on the true dispersion . -0.1 in we run a suite of simulations designed to quantify each of these effects by adding gaussian random noise and a constant continuum of various levels to one of the template stars that has been broadened to various widths . we create ten mock spectra for each s / n value , and then measure @xmath14 for each mock spectrum . this simple input - output experiment shows that our error bars are reasonable . we expect @xmath54 errors for the targets with @xmath55 and @xmath56 continuum dilution ( fig . systematic uncertainties may be introduced in the gas and stellar dispersions due to our use of fixed angular rather than physical apertures . in the case of the stars , we are typically extracting the spectra at @xmath57 . since the velocity dispersion profiles of bulge - dominated galaxies are known to be flat ( e.g. , * ? ? ? * ) , in general aperture bias should be small compared to the measurement uncertainties . the only potential exceptions are the disk - dominated galaxies ( j1106 , j1222 , and j1253 ) for which disk contamination may be significant . in the case of the gas , line widths can change dramatically as a function of aperture ( typically getting narrower at large radius ) . unfortunately , the dependence of gas dispersion on radius varies substantially from object to object ( see , e.g. , * ? ? ? * ; * ? ? ? * ) and there is no physically motivated aperture to choose . in future work we will explore the full variation in gas velocity dispersion and luminosity as a function of radius . a convenient additional product of our velocity dispersion fits is a continuum - subtracted emission - line spectrum for each target . since we are interested in the [ ] line when it falls in the bandpass , we extend the model out to the bluest available pixels although we do not use those wavelengths in the fitting . examples of the continuum - subtracted spectra are shown in figure 1 , and we note that all spectra have been extracted with a 225 aperture . we fit the h@xmath58 , [ ] , and [ ] lines with multi - component fits . briefly , the [ ] @xmath59 lines are modeled with identical sets of gaussian components , where the centroid shifts are fixed to the laboratory value . the h@xmath58 line is fit with a combination of two gaussians ( but see 7 for an alternate fit ) . our prescription follows that described in detail in greene & ho ( 2005a ) , except that unlike that work in general we require more than two components to adequately model the lines . furthermore , we do not place any restrictions on the relative widths or strengths of the lines . the [ ] @xmath60 lines are modeled with identical sets gaussian components ( usually two ) , and the line ratios in the doublet are fixed to one . this procedure is necessary to decouple true velocity dispersion from the small velocity separation of the two components of the doublet . the line widths and strengths are shown in table 4 . the [ ] line is known to have an asymmetric profile , often with a prominent blue wing ( e.g. , * ? ? ? @xcite find that using a two - component fit and removing the blue wing results in a line width that better tracks @xmath14 . however , we find that in many cases a two - component model is not an adequate description of these lines . as shown by many authors ( e.g. , * ? ? * ; * ? ? ? * ; * ? ? ? * ) , the fwhm of the line is relatively free of bias from the velocity structure at the base of the line . therefore , we simply take @xmath61=fwhm/2.35 , as measured from our best - fit emission - line model . although in the bulk of this paper we focus on the extractions with a 225 aperture , we have made a variety of other extractions , with widths of 378 , 567 , and 756 . the last has an effective area comparable to the sdss fiber ( 7 arcsec@xmath62 ) . in figure 4 we compare the @xmath30}}$ ] measurements for our objects extracted ( typically ) within a 756 aperture with the sdss measurements . we find overall reasonable agreement , with @xmath63 , sdss } - \log l_{\rm [ o { \tiny iii } ] , our}\rangle = 0.01 \pm 0.2 $ ] . one object is a significant outlier ; our measurement of j2212 is @xmath64 times fainter than that derived from the sdss spectrum . this problem persists regardless of the flux calibrator used . it is curious , since we do not see a similar problem for any other target observed on this night . since the discrepancy is seen in both the continuum level and the @xmath30}}$ ] value , and since j2212 is a highly disturbed and spatially extended system ( 4 ) , we suspect that we were just unlucky in our slit placement and missed some luminous region of the galaxy . none of the conclusions presented here are impacted if we use the sdss luminosity rather than ours . if we remove j2212 , then the scatter between the two sets of measurements drops to 0.15 dex . we note that in our own observations @xmath30}}$ ] can vary by as much as @xmath65 between different aperture positions , and so we are satisfied with this level of agreement . we have also reduced the acquisition images ( table 1 ) using standard bias - subtraction and flat - fielding routines within iraf . we do not attempt to remove the pattern noise in the images . each frame is tied to the sdss photometric system using three to six field stars . we use the iraf package _ phot _ to perform aperture photometry on these relatively uncrowded fields , and then calculate a weighted average zeropoint based on the reported photometric errors in the sdss photometry . the formal uncertainties in the final zeropoints are @xmath66 mag . the ground - based imaging reaches a typical surface brightness limit of @xmath67 mag arcsec@xmath68 , and it is worth using the images to derive bulge - to - disk decompositions for the sample . our goal is to isolate the bulge - like component of the host galaxy in order to compare with the @xcite relation of inactive bulge - dominated galaxies . the faber - jackson relation can provide a valuable diagnostic as to whether the galaxy has reached the relaxed state of local elliptical galaxies . we use the two - dimensional surface brightness fitting program galfit @xcite to model each galaxy . our band of choice is sloan @xmath21 , since we have it for the majority of the targets . however , for various reasons we are forced in a few cases to model either the @xmath20- or @xmath22-band image . -0 mm in these cases we convert to @xmath21 band using the measured color from the sdss petrosian magnitudes . throughout , magnitudes are quoted in the ab system . in comparison with broad - line agns the modeling is relatively straightforward in this case since there is no nuclear point source . as a rule , however , parameter coupling is a major complication ( e.g. , * ? ? ? * ) , and thus our operating principle is to model each galaxy with the minimum number of components . we are particularly interested in the bulge components of these galaxies , and we only introduce a disk component if visible in the images . in distinguishing between bulges and bars for disk - dominated galaxies we use the reduced @xmath40 to determine the preferred model , but we generally can not constrain more than two galaxy components robustly . for the purposes of this paper we have not placed rigorous limits on the presence of low surface - brightness components ( cf . * ? ? ? our basic model is the @xcite function : @xmath69,\ ] ] where @xmath70 is the effective ( half - light ) radius , @xmath71 is the intensity at @xmath70 , @xmath72 is the srsic index , and @xmath73 is chosen such that @xmath74 we adopt the analytic approximation for @xmath73 from @xcite , as adapted from @xcite : @xmath75 the srsic model reduces to an exponential profile for @xmath76 and a @xcite profile for @xmath77 . bars may be modeled as ellipsoids with very low axial ratios . we follow ( * ? ? ? * ; see also @xcite @xcite ) and model the intensity distribution in the bar as a gaussian . in total for a given srsic component , the model parameters include the two - dimensional centroid , the total magnitude , the srsic index , the effective radius , the position angle , and the ellipticity ( all of which are constants with radius for a given component in this version of galfit ) . the sky is modeled as a constant pedestal offset determined using multiple boxes placed in empty regions of the image , and levenberg - marquardt minimization of @xmath40 in pixel space is used for parameter estimation . -0 mm even without a nuclear point source , proper modeling of the observed light distribution requires convolution with a psf . we follow @xcite and use _ daophot _ as implemented within iraf to build a psf model , using no fewer than five isolated field stars on the same chip as the target galaxy . the psf thus constructed is both noise - free and directly based on the psf from the current image . two - dimensional modeling provides a few advantages worth noting . for one thing , we gain the freedom to assign different position angles and ( if necessary ) centroids to different components ( e.g. , * ? ? ? * ) . another major benefit of galfit for these galaxies , which often have nearby companions , is that we can simultaneously model the surrounding galaxies rather than mask them , which provides a considerably better model of the background . all fits are shown in figure 5 . uncertainties in the measured parameters are driven primarily by the psf model and the sky level , as well as inadequacies in the assumed model for the surface brightness profile . in order to estimate the former , we adopt a single star from the field as an additional psf model . we further run a pair of models in which the sky level is increased and decreased by @xmath78 , which is the typical dispersion in sky values across the field . the final error bars represent the dispersion in each parameter from these four alternate runs . our methodology is described in more detail in @xcite . the hardest uncertainties to estimate arise when our assumed model is incorrect . indeed , in j1356 and j2212 ( see table 2 and fig . 5 ) , the isophotes are so disturbed due to ongoing merging activity that we do not achieve a sensible fit from a simple combination of one or two srsic components . in these two cases , due to the presence of tidal features and spatially overlapping nuclei , our models are not particularly meaningful . instead , we adopt a nonparametric @xcite magnitude for clearly disturbed systems , as noted in table 2 . furthermore , although the direct light from the active nucleus is extincted by many orders of magnitude , some of it escapes along unobscured directions , scatters off of the interstellar matter in the host galaxy and reaches the observer . both our photometric modeling ( see below ) and spectroscopic fitting ( 7 ) may be affected by this component . finally , we do not have usable magellan images for j1124 or j1142 , and so we adopt sdss magnitudes in table 2 . as a sanity check , we compare our final galaxy luminosities to model magnitudes from sdss , as well as to magnitudes from the _ hst _ survey of @xcite for overlapping galaxies . the total magnitudes we measure agree very well with the sdss model magnitudes , with an average difference ( ours@xmath79sdss ) of @xmath80 mag , excluding j0157 because the photometry is exceedingly uncertain due to the proximity of a bright star . we are thus confident that our photometry is reasonable , but we adopt an additional 0.15 mag uncertainty in the bulge luminosities to account for the systematic differences between the two sets of measurements . there are two objects in common between this paper and the _ hst_sample , j1106 and j1413 . again , the total magnitudes agree reasonably well , within 0.07 mag in the case of j1106 and 0.4 mag in the case of j1413 . ( color differences are derived from the ldss3 spectra , and so strictly speaking only apply in the inner regions of the galaxy , and we account only for galactic extinction . ) while we derive similar b / t ratios for j1106 from the two images , the derived bulge effective radii are different ; we find @xmath81 kpc , while it is found to be 3.8 kpc in the nearest _ hst _ band ( f550 m ) . the explanation for this difference can be seen in the radial profiles derived from the ldss3 and _ hst _ images respectively ( fig . 5 ) . a scattered - light component is detected in _ images as a biconical structure with a maximum extent of 09 and appears as a central brightness peak in the one dimensional brightness profile . this component is not resolved with the ldss3 images but may bias our derived bulge size to low values . at the same time , the strong dust lanes may mimic the transition between bulge and disk in the _ hst_analysis , causing the _ hst _ bulge measurement to be too high . this one example emphasizes that we are potentially biased in our bulge sizes and luminosities due to scattered light . we do not detect the scattered light signal strongly in the spectrum , presumably because of internal reddening in the host galaxy . in addition to reliable galaxy luminosities , we also have derived bulge - to - total ratios ( b / t ; table 2 ) and general morphological information for this sample . there are four disk galaxies ( j1106 , j1222 , j1253 , and j2126 ) while the rest consist of a single bulge - like component , or in the case of j1356 and j2212 , train - wrecks . based on sdss imaging , j1124 appears to be extended , but not obviously disky , while j1142 is compact . it is interesting to note , furthermore , that a large fraction of the sample have companions ( eight out of 15 , most confirmed with the long - slit spectroscopy ) . four objects are significantly tidally disturbed , and thus obviously currently interacting ( j0841 , j1356 , j1222 , and j2212 ) . there is no correlation between agn luminosity or stellar velocity dispersion and galaxy morphology . even if we include the _ hst _ observations of @xcite , the median @xmath30}}$ ] for objects with and without a disk component are identical to within 0.1 dex . on the other hand , the sample is quite small . @xcite find a hint that @xmath30}}$ ] and morphology are correlated , in the sense that b / t decreases with agn luminosity , but for objects extending to somewhat lower luminosities . our ultimate goal is to compare the structural properties of the narrow - line quasars to those of inactive galaxies , which involves translating our observations to match the exact observing procedures and evolutionary state of the local samples . since our galaxies span a redshift range of @xmath83 , we must assume a model of the intrinsic spectral shape in order to calculate an effective rest - frame luminosity . we use the publicly available program _ k - correct _ based on the sdss photometry @xcite . it would be preferable to use the bulge component alone to derive the @xmath82-corrections , but given our heterogeneous coverage in different filters for different subsets of objects we use the sdss photometry in the interest of uniformity . additionally , we must estimate the contamination from emission lines . these objects can have high ews in [ ] , meaning that the line emission contributes significantly to the broad - band luminosity . from @xmath84 , [ ] contaminates the @xmath21-band magnitude , while at higher redshift it falls in the @xmath22 band . we calculate the total ( extended ) [ ] luminosity from our widest slit extraction , and convert to an ab magnitude using _ synphot _ in iraf . the corrections thus derived , as well as the @xmath82-corrections for the corrected colors , are shown in table 2 . along the same lines , scattered light from the nucleus may contribute a sizable fraction of the luminosity in the blue . we have tried to disentangle contributions from both star formation and scattered light to the total luminosity ( 7 ) . @xmath82-correct _ with the blue continuum removed leads to only @xmath85 mag of change in the final @xmath82-corrections . finally , of course , stellar populations fade as they age , and the magnitude of that fading is dependent on the detailed star formation history of the galaxies in question . we will explore this last issue in 7 . -0 mm it is hard to interpret observations of narrow - line agns in the absence of direct measurements of bh mass . the targets may be @xmath86 @xmath3 bhs radiating close to their eddington luminosity or they may be more massive systems in a low accretion state . their importance to the accretion history of the universe , the interpretation of the observed properties of the host galaxies , and comparisons with broad - line objects all depend critically on the assumed mass scale . the stellar velocity dispersions and inferred range of bh masses thus constitute one of the main results of this paper . using the @xmath7 relation of @xcite , we convert @xmath14 into bh mass and in figure 6 we plot @xmath2 versus @xmath30}}$ ] for the sample . lines of constant eddington ratio are highlighted , where the bolometric correction , @xmath87 } + 3.5 $ ] is taken from @xcite . this correction is derived in two steps . first , the conversion between @xmath30}}$ ] and @xmath88 is calibrated using sdss broad - line agns @xcite , and then the relation between @xmath88 and bolometric luminosity is derived based on two recent compilations of spectral energy distributions ( seds ) , @xcite and @xcite . in the latter case , we have extrapolated the uv - optical slope downward beyond the near - infrared minimum in order to estimate the full contribution of the big blue bump , but exclude the reradiated mid - infrared emission . the estimated scatter is 0.4 dex , accounting for the 0.2 dex systematic difference between the two chosen bolometric corrections , the 0.36 dex scatter between @xmath30}}$ ] and continuum luminosity in broad - line objects , and the 0.05 dex scatter in the single - band bolometric correction from marconi et al . @xcite . as seen in figure 6 , our narrow - line quasars appear to span @xmath89 dex in bh mass , luminosity , and eddington ratio . nominally , the eddington ratios cluster around the value log @xmath5@xmath90 with a dispersion of 0.7 dex ( see the best - fit gaussian fit to the distribution in figure 6_b _ ) , but it is important to bear in mind how the eddington ratios are obtained . as was mentioned above , there is an 0.4 dex uncertainty in the conversion from @xmath30}}$ ] to @xmath91 . furthermore , there is a @xmath92 dex uncertainty associated with each @xmath2 measurement which arises from the uncertainty in the velocity dispersion . therefore , the observations are consistent with a population of sources radiating close to their eddington luminosities , with the distribution of eddington ratios broadened only by observational uncertainty . now let us compare our observations with the masses and eddington ratios found for broad - line agns at similar redshifts . @xcite find a strikingly narrow range in eddington ratio for luminous broad - line agns at @xmath93 . @xcite and @xcite report similar results , although the latter paper finds evidence for an increased spread in @xmath5 at lower luminosity . it appears that our sample is consistent with a similar behavior : rather than spanning a wide range in accretion rate , the objects are clustered close to their eddington limits . at face value , not only do the narrow- and broad - line objects share comparable space - densities at the same luminosity @xcite , but their distributions of mass and eddington ratio are also surprisingly similar . probably our application of a luminosity threshold is partially responsible for the observed narrow distribution in bh mass . the eddington limit effectively determines the lowest - mass bhs to be included in the sample , while the space density of high - mass bhs declines exponentially ( e.g. , * ? ? ? nevertheless , while a variety of correlated errors or selection effects may contribute to the apparent narrow distributions observed in each population , the point we emphasize here is that the selections and biases are quite different between the broad and narrow - line objects , and thus the similarity in derived distributions is surprising . there are several steps involved in calculating @xmath2 and @xmath5 , each of which has the potential to introduce systematic biases into the measurements . first , there is still substantial disagreement about the calculation of @xmath2 in broad - line quasars on the basis of continuum luminosity and broad line width ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? second , the bolometric luminosity correction for narrow - line quasars uses a conversion from @xmath30}}$ ] to @xmath94 derived for broad - line quasars , and this procedure may systematically bias the bolometric luminosities towards low values @xcite . furthermore , one may question our assumption that the @xmath7relation holds for narrow - line quasars . for one thing , they are active ( i.e. , currently accreting ) bhs and thus may be offset from the local relation @xcite . in addition , there may or may not be significant redshift evolution of the local @xmath7relation that we use to calculate @xmath2 , and both the sign and the magnitude of this change are at present unknown @xcite . each of these effects may potentially bias the calculation of eddington ratios by as much as 0.5 dex . the similarity of eddington ratios of broad- and narrow - line quasars , as well as the narrowness of their distribution , is an interesting and suggestive result , but we caution that some systematic effects may be affecting the comparison . it has long been known that the [ ] line width traces the bulge velocity dispersion in low - luminosity agns ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? recently , with the discovery of the @xmath7 relation @xcite , interest in the gas velocity dispersion ( @xmath61 ) has increased , due to the possibility of using it as a proxy for @xmath14 in distant or luminous active galaxies for which it is prohibitive to measure @xmath14 directly ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) . as a result , it becomes all the more crucial to understand the physical origin of the observed correlation , and to characterize whether it depends on properties of the agn , such as bolometric luminosity , eddington ratio or radio luminosity . this work increases the luminosity range over which such comparisons have been done by nearly two orders of magnitude , providing a crucial check of the assumption that @xmath14@xmath95@xmath61 . by the same token , given that @xmath61 and @xmath14 have now been measured for large samples of weakly active galaxies over a wide range of hubble types @xcite , it is possible to use the variations in this ratio as a _ probe _ of the impact of the agn on the surrounding interstellar medium ( ism ) . as shown in figure 7 , there appears to be no correlation between either @xmath96}$ ] or @xmath97}$ ] and @xmath14 at these high luminosities ( the spearman correlation coefficient is @xmath98 with a probability @xmath99 that no correlation is present ) . we note that while we span a narrow range in luminosity , we actually span a wide range in @xmath14 , suggesting that the lack of correlation is a generic property of luminous obscured agns , and may well be a generic property of all agns at such high luminosities . we try to salvage the correlation by relying on lines with lower ionization potential , since many studies have found systematic trends between line shape and the critical density or ionization potential of the lines ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? generically speaking the high - ionization and/or high - critical density lines originate from closer to the bh and are thus typically broader , making low - ionization lines a better choice as proxies to @xmath14(e.g . * ; * ? ? ? instead , we find that the [ ] and [ ] gas widths are strongly correlated ( @xmath100 , @xmath101 ; @xmath97}$]/@xmath96}$]@xmath102 ) , but neither correlates well with @xmath14 . in search of some physical insight into the processes driving the gas dispersion , we explore correlations between various agn properties and @xmath61/@xmath14 . for convenience , we define the logarithmic difference @xmath103@xmath104log @xmath61@xmath79log @xmath14 . within the sample as a whole , @xmath1@xmath103@xmath105 . this is to be compared with @xmath1@xmath103@xmath106 for low - luminosity agns in early - type galaxies @xcite and @xmath1@xmath103@xmath107 for local narrow - line agns from sdss @xcite . historically , many papers have found a correlation between radio power and [ ] line width ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? we see no evidence for increasing @xmath103 or @xmath61 with increasing radio power over a range of @xmath108/w hz@xmath12@xmath109 at 1.4 ghz . the nelson & whittle sample does extend an order of magnitude higher in radio luminosity , but it is clear that jets alone can not explain the sources with very large gas dispersions ( see also * ? ? ? * ; * ? ? ? according to @xcite , the deviation of @xmath61 from @xmath14 shows a clear dependence on agn luminosity and eddington ratio ( see also * ? ? ? * ) , in the sense that more luminous agns have a larger @xmath61/@xmath14 ratio on average . we see no statistically significant correlation between @xmath30}}$ ] and @xmath103(@xmath110 , @xmath111 ) . granted , the dynamic range in luminosity is very narrow for this sample alone , but , if anything , there is a weak negative trend of @xmath103 with luminosity . indeed , the lower - luminosity data from @xcite shown in figure 7_cd _ clearly demonstrate that there is not a well - defined correlation between @xmath103 and @xmath30}}$ ] over a much larger luminosity range . while these results are discouraging in terms of the use of @xmath96}$]to approximate @xmath14 , they do tell us in no uncertain terms that the ism in the hosts of these luminous quasars is very disturbed . as mentioned above , gas line widths in the centers of inactive and weakly active galaxies are well - described by virial or slightly sub - virial motions due to turbulence @xcite . in low - luminosity active galaxies , there is some evidence that agn activity is , on average , increasing the gas line widths @xcite , but overall the gas still appears to be dominated by virial motion , at least in a luminosity - weighted sense . in stark contrast , at the highest agn luminosities , we see no correlation of gas and stellar velocity dispersions , suggesting that the ism is completely disturbed . despite decades of research , the prevalence of temporally coincident star formation and accretion remains one of the outstanding problems in agn physics ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? generically speaking , it seems difficult to inhibit star formation when there is a large enough gas supply to fuel a luminous agn . on the other hand , robust measurements of star formation are difficult to obtain , particularly with a luminous broad - line agn spilling light in all directions and at all wavelengths . our hope is to gain new insights from the study of obscured but intrinsically quite luminous accreting bhs . of course , there are always complications . in our case , although the primary emission from the agn is obscured , nevertheless accretion luminosity manages to contaminate many standard star formation indicators . the most reliable instantaneous ( i.e. @xmath112 yr ) rate diagnostic , h@xmath113 luminosity ( e.g. , * ? ? ? * ) , is excited by accretion power rather than young stars in our case . likewise , blue / uv light is prone to contamination from scattered agn light , whose magnitude can be very significant at these high luminosities ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) . in what follows we show that all of our galaxies are bluer than their inactive counterparts , and we use a variety of indirect techniques to disentangle the young starlight from scattered agn light . we begin by modeling the sdss colors using stellar population synthesis techniques . as input , we use the sdss @xmath114 magnitudes , with the [ ] and @xmath82-corrections applied , as described in 4.3 . we do not apply any correction for potential scattered light , as our goal is to establish whether or not there is a blue component apparent in the integrated colors of the galaxies . the stellar population synthesis code is described in detail in @xcite . in short , the code utilizes isochrones from padova @xcite , the basel3.1 stellar libraries with thermally pulsating asymptotic branch stars added @xcite , and ( in this case ) a @xcite initial mass function . dust attenuation optical depths are allowed to vary separately for old and young stars , following the model of @xcite . -0 mm the fraction of total stellar mass formed in the last 300 myr for each galaxy is shown in table 4 , with a median value of @xmath115 at a median stellar mass of @xmath116 @xmath3 . already we see that while the bhs are experiencing a major growth episode , the galaxies are not . the integrated colors show that blue light is prevalent , but spectroscopy is required to determine its origin . for the purpose of comparing with @xcite , we measure the d@xmath117(4000 ) and h@xmath118 indices . this two - dimensional diagnostic is predominantly sensitive to age , and relatively insensitive to variations in metallicity or dust . we follow their conventions , wherein d@xmath117(4000 ) is defined as the ratio of fluxes in bands at 38503950 and 40004100 @xcite , and h@xmath118 is defined in @xcite . as a function of age , galaxies follow a well - defined sequence in d@xmath117(4000)-h@xmath118 space , and the locus of sdss galaxies in this space is shown in figure 8 . unfortunately , because our agn - powered emission lines have such high ews , we can not measure a meaningful h@xmath118 index directly from the spectra . we can measure d@xmath117(4000)after correcting for [ ne iii]@xmath119 , and those values are shown in the top histogram in figure 8 . [ we have verified that we can reproduce the values in kauffmann et al . measuring the 30 most luminous objects in their dr2 catalog , we find @xmath1[d@xmath117(4000)@xmath120-d@xmath117(4000)@xmath121/d@xmath117(4000)@xmath122 . ] the median of our sample , uncorrected for any blue component , is @xmath1d@xmath117(4000)@xmath123 . for comparison , the thirty most luminous targets in the kauffmann et al . dr2 sample , with a median @xmath30}}$ ] of @xmath124 erg s@xmath12 ( @xmath92 dex less luminous than our median luminosity ) , have @xmath1d@xmath117(4000)@xmath125 . our first task is to measure the luminosity of the blue continuum . the ldss3 spectra do not always include the 4000 break , and so we use the sdss spectra for this measurement . we wish to model the various spectral components , but to mitigate degeneracy we restrict the model to nothing more than a k star , an a star , and a blue power law whose slope is fixed to @xmath126 . in principle the slope of the blue component will depend on the physical process responsible for it ; given the modeling degeneracies and the uncertainties in internal reddening and flux calibration , we do not attempt to measure the true spectral shape of the blue light . likewise @xmath14 is fixed to the best - fit value determined in 3 . while the fits are not unique , we believe each component is reasonably well constrained : a significant a or f star contribution is fixed by higher - order balmer absorption lines , while the ew of the old stellar component provides a direct probe of the dilution by blue light . by repeating the fitting with a power - law slope of @xmath127 , we confirm that the derived luminosities are insensitive to the assumed power - law slope . more reassuringly , fits performed without an a star component yield blue luminosities that agree within 0.05 dex of the nominal result . -0 mm these fits provide two useful measurements : one is a blue luminosity , and the other is a ( model - dependent ) measurement of the d@xmath117(4000 ) and h@xmath118indices for the underlying stellar population , as shown by red crosses in figure 8 . the stellar population gets older as one moves to the lower right of this diagram . note that three objects have d@xmath117(4000)indices that are larger than values seen in real galaxies . presumably the discrepancy arises because in these cases the stellar continuum is modeled by a single k star , whereas in real stellar populations there will be some contribution from m stars as well as younger populations . the d@xmath117(4000 ) and h@xmath118 indices derived this way represent a lower limit on the young population , while below we will attempt to evaluate what fraction of the blue light may originate in young stars . if the blue light is predominantly scattered agn light , we can predict the corresponding reflected broad h@xmath58 component , using the well - known tight correlation between optical / uv continuum luminosity and balmer emission ( e.g. , * ? ? ? * ) ; we use the recent calibration of @xcite . the expected line fluxes are low ( @xmath128 erg s@xmath12 cm@xmath68 ) , but we can at least determine whether or not our spectra are consistent with a broad component at this level . in the most luminous narrow - line quasars , @xcite robustly detect a broad h@xmath58component in a significant fraction of their sample and determine the scattered light contribution to the blue continuum ( see also * ? ? ? * ; * ? ? ? * ; * ? ? ? , because the lines are much weaker , we rather take the indirect approach of comparing the expected scattered emission from the continuum fits with a direct fit to the h@xmath58 line . if the two h@xmath58 fluxes are consistent , then we conclude that the blue light may be ascribed exclusively to scattered light . since the narrow - line shapes are quite complex and highly non - gaussian , care must be taken to ensure that structure in the narrow lines does not masquerade as a bona fide scattered broad component . we thus repeat the narrow - line fits as described above , except that in this case we fix the narrow components of h@xmath58 to that of [ ] , and allow the fit to include an additional broad h@xmath58 component . we constrain @xmath130fwhm@xmath131 km s@xmath12(liu et al . find h@xmath58@xmath132 km s@xmath12 ) , since otherwise the broad line will often attempt to fit low - level wiggles in the continuum . -0 mm in all cases the [ ] and h@xmath58 line shapes match each other extremely well ( fig . 9 ) , and in eight cases we detect a broad component ( table 4 ; fig . 9 _ bottom _ ) . we convert the rms residuals into an upper limit on broad h@xmath58 assuming a fwhm of 4000 km s@xmath12 . we take as the upper limit on the scattered broad component either the detected broad h@xmath58 luminosity or the upper limit thereof , whichever is larger , and then compare with the luminosity we expect based on the detected blue continuum . if the two agree within a factor of 2 , then we believe scattered light can fully account for the blue light , whereas when the continuum level is higher , we ascribe the excess to current star formation ( the five cases are noted with * in table 4 ) . when detected , the blue light accounts for anywhere from 5% to 60% of the total continuum flux at 4000 . after the blue light is removed , our galaxies occupy a sensible region of the h@xmath118-d@xmath117(4000 ) plane relative to the inactive galaxy population . the median d@xmath117(4000 ) rises from 1.2 to 1.6 . even excluding those galaxies with d@xmath117(4000 ) larger than what is seen in sdss galaxies yields an average value of 1.5 . in other words , while there is clearly an age spread , the corrected index values suggest that the galaxy light is dominated by old stellar populations in the majority of cases . in five galaxies ( j0841 , j1222 , j1253 , j1356 , and j2212 ) , we find a blue continuum in excess of what can be explained by scattered light , which we interpret as a star - forming component . it is interesting to note that these systems tend to have higher star formation rates inferred from their broad - band colors as well . even more intriguing is the observation that all of the highly disturbed systems have an excess ( star - forming ) blue component . on the other hand , we see no correlation between the d@xmath117(4000)values and host morphology . in summary , these host galaxies are forming @xmath133 of their stellar mass at the present time . about half of the objects are consistent with having only old stellar populations , while the others show evidence for ongoing ( in the form of excess blue light ) or recent [ low values of d@xmath117(4000 ) ] star formation at a low level . while we can not place absolute limits on the current star formation rate to any great precision , we argue here , and based on the structural parameters investigated below , that the galaxies are effectively fully formed . in figure 10 we show the @xmath21-band faber - jackson relation for the galaxy bulge components , with the most luminous targets circled in blue and the disk- or bulge - dominated galaxies differentiated as triangles or circles , respectively . obviously there is a large amount of scatter in this diagram , much of it measurement - based . it appears that nearly half of the targets are offset toward more luminous bulges at a given @xmath14 , but we caution that there are substantial systematic biases hidden in the bulge luminosities , as discussed below . we do not see any correlations between location in the faber - jackson diagram and agn luminosity , young stars ( as traced by the blue light ) , or intermediate - age stars [ d@xmath117(4000 ) ] . partly , the sample is just too small , spanning too narrow a range in luminosity , to display any such trends . at the same time , a number of potential uncertainties may impact the measured luminosities . there is firstly the contamination from scattered blue light , which boosts the observed @xmath21-band magnitudes by as much as 0.5 mag . also , the targets span a range of redshifts ( @xmath0 ) , but we know that over time a collection of stars will become less luminous as the most massive stars evolve . unfortunately , as shown by , for example , @xcite , the precise magnitude of this fading is highly uncertain ( by @xmath134 magnitude ) , due to our inadequate knowledge of the slope of the initial mass function around the turnoff mass . taking a value of 0.9@xmath135 mag in @xmath21 for the purposes of illustration ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) , and removing the detected blue light yields the magenta points , substantially lowering the discrepancy with the local faber - jackson relation . we are reluctant to provide any additional interpretation , given the small sample and substantial uncertainties in the measurements . our basic conclusion , combining the results of the stellar population modeling , stellar population indices , and structural properties , is that for the most part the host galaxies had their stellar content in place prior to this epoch of substantial bh growth . the general result , that we see a range of star - forming properties with no clear correlation between host galaxy morphology and star formation properties , is in broad agreement with previous results for lower - luminosity narrow - line agns ( e.g. , * ? ? * ; * ? ? ? * ; * ? ? ? most interestingly , the situation appears rather different for objects at the highest luminosities . @xcite see clear evidence for wolf - rayet features in @xmath136 of their targets ( we see none in this sample ) , which constitute a clear signature of a recent burst of star formation ( e.g. , * ? ? ? furthermore , @xcite find very high star formation rates from mid - infrared spectroscopy , for a different sample of luminous narrow - line agns , rates that are substantially higher than either those seen in inactive galaxies of similar mass or those in typical broad - line agn samples of similar luminosity . now , @xcite find compelling evidence that the post - starburst fraction increases with increasing [ ] luminosity . we may be seeing a hint that the galaxy growth rate scales with agn luminosity even up to starbursts at the highest agn luminosities ( see also * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? images show that hosts of luminous narrow - line quasars from the sdss - selected sample are bulge - dominated @xcite , whereas those of somewhat less luminous ( by about 0.5 dex ) ir - selected narrow - line quasars are dusty , disky systems @xcite . this paper presents observations of objects with luminosities very similar to those in our hst sample , with similar results : about a third of the objects are found in strongly disturbed interacting systems , while the rest are found in bulge - dominated galaxies . we previously suggested that the difference in morphology may reflect the difference in the quasar fueling mechanisms as a function of quasar luminosity @xcite , but of course this interpretation is only tentative since the comparison samples were selected using very different methods . a detailed study of hosts of optically selected narrow - line quasars at lower luminosity is needed to understand the origin of the morphological difference . we have chosen to explore the host galaxy properties of obscured active galaxies in particular because of the increased contrast between galaxy and nucleus . on the other hand , it is our burden to determine whether or not our conclusions can be generalized to all accreting bhs at the epoch being probed . rather surprisingly , the distributions of bh mass and eddington ratio for our sample of narrow - line agns are quite similar to those seen in broad - line samples at similar luminosity and redshift ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? this fact , combined with the observation of @xcite that the space densities of the broad- and narrow - line objects are comparable at fixed luminosity , suggest that the two may originate from similar populations . there is another aspect to the story , as illuminated by the differences in star - forming properties between the narrow- and broad - line objects . a variety of recent papers have devised clever techniques for disentangling photoionization by stellar and nonstellar sources using the line strengths of high- and low - ionization lines , where the former are certainly dominated by the agn but the latter can be significantly contaminated by star formation ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? these studies all come to the same conclusion ; star formation is more energetically significant in narrow - line than in broad - line agns , particularly at the luminous end ( see also * ? ? ? * ) . the excess star formation in narrow - line objects has been interpreted as evidence for an evolutionary scenario , in which obscured accretion and ongoing star formation coexist , and then accretion terminates further star formation , yielding an unobscured object ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) . in this picture some event ( usually a merger ) induces both the star formation and nuclear activity , and the broad - line object signifies the final stage in the accretion cycle . perhaps , on the other hand , there is not necessarily a causal connection between the star formation event and the accretion event , but when both occur simultaneously the odds are higher that the accretion event will be obscured by star formation in the host galaxy @xcite . our data alone clearly can not distinguish between these two possibilities . however , we do see a range of star formation properties across even this relatively small sample . naively , if there were truly a causal connection between star formation and nuclear activity , one would expect that the majority of our sample would show a clear sign of ongoing star formation , even if only in the form of post - starburst signatures . we look forward to analyzing the full @xcite sample to more definitively quantify the star formation properties as a function of agn luminosity @xcite . we have seen that luminous obscured active galaxies with @xmath137 and @xmath138 erg s@xmath12 are composed of bhs with masses @xmath139 @xmath3 radiating at a high fraction of their eddington limits . we are seeing unambiguous evidence that bh accretion impacts the host on a galaxy - wide scale ; the host galaxy ism is clearly substantially disturbed relative to less active systems . as a result , we can not recommend the use of gas line widths as a substitute for stellar velocity dispersions at such high luminosities , at least not in obscured sources . there is a clear blue component in the galaxies , apparent both in the overall colors and in nuclear spectroscopy , and in most cases the blue light is consistent with being scattered emission from the obscured nucleus . nevertheless , nearly half of the sample shows evidence for recent or ongoing star formation , while the rest are consistent with predominantly old stellar populations . although star formation activity may rise with vigorous agn activity , there is a wide dispersion at all luminosities , suggesting that we are far from having a consistent picture of the coevolution of host galaxies and bh growth .

we present an-parity conserving model of sneutrino dark matter within a higgs - philic extension of the minimal supersymmetric standard model . in this theory , the parameter and light dirac neutrino masses are generated naturally upon the breaking of the gauge symmetry . one of the right - handed senutrinos is the lsp . the leptonic and hadronic decays of another sneutrino , taken to be the next - to - lightest superpartner , allow for a natural fit to the recent results reported by the pamela experiment . we perform a detailed calculation of the dark matter relic density in this scenario , and show that the model is consistent with the atic and fermi - lat experiments .

0906.3540

the question of the nature of dark matter ( dm ) , which forms nearly one - fourth of the total mass in the universe , is a pivotal question for cosmology , astrophysics , and particle physics . other than its relic density @xcite , little is known about the structure of dm . recently , there has been great excitement that new information about dm may be revealed by the results from the pamela satellite @xcite claiming an increase in positron fraction in cosmic rays with energies above @xmath3 . others , including the ppb - bets @xcite and hess @xcite experiments , also claim enhancements in electron / positron flux for energies above @xmath4 . more recently , the fermi collaboration released a measurement of the @xmath5 flux in the @xmath6 to @xmath7 range @xcite . all of these experiments are consistent with a new primary source contributing to local electron / positron fluxes in the @xmath8 range @xcite . these anomalies in cosmic ray fluxes may well originate from astrophysical sources such as pulsars or nearby supernova remnants @xcite . however , they might also result from the presence of dm particles , providing a golden opportunity to learn more about the properties of dm . this intriguing possibility has led to a number of proposals that interpret the data as arising from dark matter annihilation @xcite ( which necessitates an @xmath9 boost factor ) or decays @xcite ( which must proceed much slower than the present value of the hubble parameter ) . moreover , pamela , while reporting an excess in the electron / positron flux , claims no excess in the proton / anti - proton flux @xcite . thus , one intriguing idea is that the dm carries lepton number , _ e.g. _ , it is composed of right - handed scalar neutrinos ( sneutrinos ) in minimal @xcite and extended @xcite supersymmetric models . this framework might explain the preference for leptonic annihilation / decay products @xcite . in the simplest extensions of the minimal supersymmetric model ( mssm ) that include right - handed neutrinos , the right - handed neutrinos are gauge singlets . hence the strength of their yukawa interactions with the standard model ( sm ) fields governs whether they thermalize together with the rest of the mssm matter @xcite . on the other hand , if the lightest sneutrino @xmath10 is the lightest supersymmetric particle ( lsp ) and therefore is a potentially viable dm candidate , it can be produced with the right relic density from the decays of heavier superpartners @xcite . if the heavier right - handed scalar neutrino @xmath11 is the next - to - lightest superpartner ( nlsp ) , its lifetime turns out to be longer than the age of the universe @xcite . therefore , decays of @xmath11 should still be active , resulting in observable effects like the excess energetic electrons / positrons in cosmic rays . in the mssm with right - handed neutrinos , these features have been analyzed with respect to the observed cosmic ray anomalies @xcite . however , such right - handed sneutrino dark matter scenarios generically suffer from two aesthetic problems : _ ( i ) _ the origin of the suppressed dirac neutrino yukawa couplings ( though such small couplings are allowed on the basis of technical naturalness @xcite ) , and _ ( ii ) _ the origin of the supersymmetric @xmath2 parameter @xcite . in this paper , we present a @xmath1-extended mssm model in which these aesthetic problems are remedied , and the desirable features of sneutrino dm , including the ability to account for the pamela positron excess , can be correctly produced . it is well known that the mssm suffers from a naturalness problem due to the presence of the superpotential bilinear operator @xmath12 @xcite . though the mass parameter @xmath2 enters from the superpotential , it must be of the order of the mass terms associated with the supersymmetry breaking sector . this puzzle can be remedied by extending the matter and gauge structure of the mssm , _ e.g. _ within unified and/or string models @xcite . to this end , theories with an extra @xmath13 broken at the electroweak- to- tev scale by sm singlets are known to be able to generate an appropriately sized @xmath2 parameter ( see _ e.g. _ @xcite ) . the @xmath13 symmetry can also play a crucial role in generating neutrino masses . the right - handed neutrino sector and the @xmath2 parameter can be correlated for both majorana @xcite and dirac masses @xcite . we assume here that lepton number is an accidental symmetry that is conserved at the perturbative level . hence , the neutrinos are dirac fermions , requiring yukawa couplings of @xmath14 . these couplings are technically natural , but an explanation for such a strong suppression is clearly desirable . one way this can occur is if the @xmath13 invariance suppresses leading order contributions to dirac neutrino masses and allows higher - dimensional operators @xcite . in this work , we assume that the @xmath1 charges satisfy @xmath15 to forbid the bare @xmath2 term , and @xmath16 to forbid a bare neutrino yukawa coupling . after including an sm - singlet chiral superfield @xmath17 , the relevant part of the superpotential takes the form @xcite @xmath18 in which the @xmath13 invariance requires @xmath19 and @xmath20 . in the above , @xmath21 is a large mass scale , and @xmath22 and @xmath23 are the yukawa couplings responsible for generating the @xmath2 parameter and neutrino masses . upon the breaking of the electroweak and @xmath13 gauge symmetries , the effective low energy parameters include the @xmath2 parameter @xmath24 and dirac neutrino masses @xmath25 the effective neutrino yukawa coupling @xmath26 leads to neutrino masses in good agreement with experiment : @xmath27 for @xmath28 , @xmath29 , and @xmath30 hence , in this framework the @xmath2 problem is resolved and appropriately suppressed dirac neutrino masses are generated upon @xmath13 breaking . parameter values were chosen to obtain neutrino masses which are in the right range . in the next sections , we give more details of the spectrum constrained by the experimental data . to account for the anomalous cosmic ray fluxes , there arise several additional constraints on the model . first , we assume the right - handed neutrino superfields are _ total gauge singlets _ to avoid unacceptably fast decays that would otherwise occur via gaugino mediated processes . second , the @xmath13 charges must be assigned such that the successful generation of the @xmath2 parameter and neutrino masses are maintained . the option that we pursue in this work is to have a higgs - philic " @xmath13 , similar in spirit to @xcite . enforcing these constraints leads to the charge assignments displayed in table [ charges ] . note that of the mssm fields , only the right - handed up quarks @xmath31 and the higgs fields @xmath32 and @xmath17 have nonvanishing @xmath13 charges . the charge assignment in table [ charges ] is anomalous as it stands . however , all gauge and gravitational anomalies can be cancelled either by invoking family - dependent @xmath13 charges as in @xcite ( though in this case one needs to worry about constraints from flavor violation ) , or by augmenting the matter content of the theory by sets of vector - like quark and lepton fields , as well as additional sm singlets that are charged under the @xmath1 gauge symmetry . the details will be addressed somewhere else in our future studies . with this charge assignment , scattering processes that involve @xmath32 , @xmath17 and @xmath31 are influenced by the @xmath13 gauge boson ( @xmath33 ) and the @xmath13 gaugino ( @xmath34 ) @xcite . at hadron colliders , the right - handed up - type quarks ( @xmath35 and @xmath36 ) can undergo drell - yan annihilation through @xmath33 to produce higgs fields . the decays of the up - type squarks also exhibit novel branchings due to @xmath34 exchange @xcite . ccccc + @xmath37 & @xmath38 & @xmath39 & @xmath40 & @xmath13 + + + @xmath41 & @xmath42 & @xmath43 & @xmath44 & @xmath45 + + @xmath31 & @xmath46 & @xmath47 & @xmath48 & @xmath49 + + @xmath50 & @xmath46 & @xmath47 & @xmath51 & @xmath45 + + @xmath52 & @xmath47 & @xmath43 & @xmath53 & @xmath45 + + @xmath54 & @xmath47 & @xmath47 & @xmath45 & @xmath45 + + @xmath55 & @xmath47 & @xmath47 & @xmath47 & @xmath45 + + @xmath32 & @xmath47 & @xmath43 & @xmath56 & @xmath57 + + @xmath58 & @xmath47 & @xmath43 & @xmath53 & @xmath45 + + @xmath17 & @xmath47 & @xmath47 & @xmath45 & @xmath49 + + we now explore the possibility that the excess positron flux observed in the cosmic ray data is due to the presence of sneutrino dm . to do so , we will assume that the soft supersymmetry breaking sector of the theory is such that the lightest right - handed sneutrino @xmath10 is the lsp and the next - to - lightest right - handed sneutrino @xmath11 is the nlsp . more explicitly , @xmath59 where @xmath60 denotes the remaining superpartner masses , including the heaviest right - handed sneutrino , @xmath61 . the third heavy particle , the nnlsp , is chosen to be the lightest neutralino , @xmath62 and it requires special attention since it decays late enough to have some non - thermal contribution to the dm relic density . the hadronic big bang nucleosynthesis ( bbn ) constraints will also be considered for the late decaying particles in the model . we discuss these issues further in the next section . for the spectrum considered above , the dm has two components , the stable lsp and the practically stable nlsp which lives longer than the age of the universe . together , their relic density should reproduce the wmap @xcite result . the right - handed sneutrino population forms during the cosmic evolution from the decays of the mssm spectrum plus exotics , which , with a few exceptions , should already have reached thermal equilibrium . a given sparticle decays into right - handed sneutrinos with a rate @xmath63 . therefore , the conversion rate into right - handed sneutrinos is quadratic in the effective neutrino yukawa coupling @xmath64 , and the growth of the lsp number density is fast enough to produce the observed dm relic density @xcite . we expect the third sneutrino to also decay to the lsp at freeze - out and thus contribute to the relic density . however , in our chosen parameter space , this sneutrino is taken very heavy ( quantified below ) , much heavier than other sparticles . it would likely mix more with the left - handed sneutrino and decay very quickly compared to the nlsp . our detailed analysis shows that its life - time ( @xmath65 sec ) is larger than its freeze - out time ( @xmath66 sec ) but its non - thermal contribution to the relic density is negligible since the branching ratio for decaying into lsp or nlsp is suppressed ( less than @xmath67 ) . the right - handed sneutrinos exhibit a non - thermal distribution . to see this , recall that the right - handed sneutrinos mix with the left - handed sneutrinos and therefore interact with the gauge sector . for right - handed sneutrino annihilation , only a four - point interaction , a left - handed sneutrino , or a higgsino exchange can contribute . the annihilation is out of equilibrium as long as the interaction rate is smaller than the expansion rate @xmath68 . the rate of the four - point interaction @xmath69 , since @xmath70 . the left - handed sneutrino and higgsino exchanges also have rates @xmath71 , and hence the right - handed sneutrinos do not thermalize before the electroweak phase transition . this is maintained after the phase transition if the mixing with the active ( left - handed ) sneutrino is small @xcite . for a detailed calculation of the relic density , the model must be considered in detail and all the mixings , masses and branching ratios must be determined so that we would calculate the relic density by counting all possible channels . moreover , we need to evaluate the would - be relic density of the late decaying nnlsp which is computed with through conventional methods . we implement the model fully into the calchep @xcite package program with the help of lanhep @xcite program . once the calchep model files are provided , we use the micromegas @xcite software for calculating the nnlsp relic density . in the implementation , we consider a normal hierarchy for the neutrinos and chose the mixing parameters to be @xmath72 , @xmath73 , and @xmath74 . the parametrization of the mixing matrix is identical to the one in the quark sector and the additional cp - violating phases are taken zero . as the positron excess is proportional to the neutrino masses , we keep these non - zero , consistent with the following constraints : @xmath75 and @xmath76 @xcite . we take @xmath21 , @xmath2 , @xmath77 , and @xmath78 as free parameters and express @xmath79 , @xmath80 and @xmath64 in terms of them . in this model the sneutrinos also mix , and the mixing matrix can in general be expressed as @xmath81 m_{\widetilde \nu_{r l}^{ij}}^{2 } & m_{\widetilde \nu_{r r}^j}^{2 } \end{array } \right ) \left ( \begin{array}{c } \widetilde \nu_l^i\\ [ 1.ex ] \widetilde \nu_r^j \end{array } \right),\end{aligned}\ ] ] where @xmath82 are the flavor indices and the matrix elements are given by @xmath83.\end{aligned}\ ] ] here @xmath84 and @xmath85 are the soft mass terms and @xmath86 are the trilinear couplings ( assumed diagonal ) . @xmath87 and @xmath2 are given in ( [ mneut ] ) and ( [ mueff ] ) , respectively . note that due to the specific @xmath1 charge assignments in table [ charges ] , some of the terms in ( [ mix_comp ] ) are zero . the sneutrino mass eigenstates @xmath88 are given by @xmath89 -\sin\theta^i_{\widetilde\nu_l-\widetilde\nu_r } & \cos\theta^i_{\widetilde\nu_l-\widetilde\nu_r } \end{array } \right ) \begin{pmatrix } \widetilde \nu_l^i\\ \widetilde \nu_r^i \end{pmatrix } \quad \label{higgs458}\end{aligned}\ ] ] with the left - right sneutrino mixing angles @xmath90 : @xmath91 such that @xmath92 . in the rest of the paper we will refer to the right - handed sneutrinos as @xmath93 for simplicity , but in the numerical analysis all the mixings are implemented . note that even though such a mixing is small numerically , it must be retained . the alternative is to introduce mass insertions to obtain most of the decay channels discussed below and calchep can not handle such insertions . cccccc + @xmath94 & @xmath95 & @xmath96 & @xmath97 & @xmath98 & @xmath99 + + + 100 & 2000 & 1200 & 2000 & 2000 & 5000 + + + @xmath100 & @xmath101 & @xmath102 & @xmath103 & @xmath104 & @xmath105 + + + 1392 & 1519 & 1561 & 1650 & 2537 & 2538 + + + @xmath106 & @xmath107 & @xmath108 & @xmath109 & @xmath110 & @xmath111 + + + 1513 & 1648 & 87 & 2556 & 5410 & 2536 + + + @xmath112 & @xmath113 & @xmath114 & @xmath115 & @xmath116 & @xmath117 + + + 1999 & 1999 & 1999 & 1999 & 1999 & 1999 + + cccccc + @xmath118 & @xmath119 & @xmath120 & @xmath121 & @xmath122 & @xmath123 + + + 0 & 3.95 & @xmath124 & 3.95 & 3.95 & @xmath125 + + + @xmath126 & @xmath127 & @xmath128 & @xmath129 & @xmath130 & @xmath131 + + + @xmath132 & 0.30 & 1.0 & 1.5 & 6.2 & 11.6 + + + @xmath133 & @xmath134 & @xmath135 & @xmath136 & @xmath137 & @xmath138 + + + 0.13 & 1.9 & 0.0041 & 24.6 & 27176.5 & 63.7 + + + @xmath139 & @xmath140 & @xmath141 & @xmath142 & @xmath143 & @xmath144 + + + 3.9 & @xmath145 & 3.9 & @xmath146 & 2.2 & 1.8 + + for the numerical analysis , we use the following input values : @xmath147 where @xmath148 and @xmath149 are the soft gaugino mass terms for @xmath150 , @xmath39 and @xmath151 , respectively . for the chosen point in the parameter space , the masses entering the calculations are given in table [ spectrum ] . clearly , the lsp is @xmath152 and the nlsp is @xmath153 , denoted as @xmath154 and @xmath155 in the rest of the text . the third right - handed sneutrino @xmath156 becomes @xmath157 . the nnlsp is @xmath158 as previously chosen . the total widths of these particles are also given in table [ widths ] . we first address the hadronic bbn constraints on several particles in the table , as it is known that late decaying particles may spoil the predictions of bbn and thus need special attention . from table [ widths ] , there are only two particles , namely @xmath159 and @xmath157 , with lifetimes @xmath160 sec and @xmath65 sec , respectively , which survive and decay after the freeze - out of the nnlsp . in our scenario , the relevant quantity @xcite that needs to be evaluated and compared with the observational constraints on light elements like @xmath161 , @xmath162 , @xmath163 etc is @xmath164 where @xmath165 are the energies carried by the hadrons and @xmath166 is the yield variable , defined as @xmath167 with @xmath168 the number density of @xmath159 and @xmath169 the total entropy density . this quantity is calculated with the help of micromegas . we also generated all the relevant three - body decays eventually leading to hadronic final states . these are @xmath170 and @xmath171 . then @xmath172 , and @xmath173 decay hadronically . we found that @xmath174 , @xmath175 , and @xmath176 . when all of the quantities are entered numerically , the variable @xmath177 is around @xmath178 gev . in the range of the lifetimes of particles , the constraint coming from the overproduction of @xmath179he is relevant . the observational constraint requires @xmath180 @xcite , which is satisfied for the @xmath159 case . we repeated the analysis for the heaviest right - handed sneutrino ( @xmath157 ) and found @xmath177 even smaller than @xmath178 gev . after insuring that the hadronic bbn constraints are satisfied , we proceed with the relic density calculation . as mentioned earlier , there are two particles , @xmath154 and @xmath155 as the components of the dm . including the contributions after freeze - out , the total relic density of the right - handed sneutrinos is given @xmath181 where `` ce '' ( `` fo '' ) refers the contributions from decays at chemical equilibrium ( after freeze - out ) . the contribution to the relic density from the heaviest right - handed sneutrino is small and neglected here . thus , @xmath182 has contributions only from the nnlsp @xmath159 . it is defined as @xmath183 where @xmath82=1,2,3 ( we also include the conjugated states ) . of course , the sum of the branching ratios has to be unity , _ i.e. _ , @xmath184 . numerically , we found @xmath185 and @xmath186 in the parameter set ( [ inputs ] ) . @xmath187 is the would - be relic density of nnlsp for the case where it is stable . we have calculated @xmath187 by using the micromegas @xcite through its calchep interface . numerically , we found @xmath188 with the usual dimensionless parameter @xmath189 . here @xmath190 and @xmath191 are the hubble parameter and the freeze - out temperature . for example , the two most significant contributions to @xmath192 from the ( co-)annihilation channels are @xmath193 @xmath194 and @xmath195 @xmath196 . ccccc + @xmath197 & @xmath198 & @xmath199 & @xmath200 & @xmath201 + + + 0.000112 & 0.0444 & 0.000929 & 0.0665 & @xmath202 + + for the calculation of @xmath203 and @xmath204 , the following decay channels are included ( conjugated states are not listed ) : @xmath205 we performed a complete parameter scan , and found these are the only decay modes which are numerically significant . the individual contributions to the relic density are summarized in table [ relics ] . in the table , we split the freeze - out contribution @xmath206 into two parts , @xmath207 and @xmath208 , which show the contributions of the nnlsp decaying to both the lsp and the nlsp , @xmath209 , and @xmath210 , respectively . as can be seen from table [ relics ] , the dominant contributions are coming from @xmath211 @xmath212 and @xmath208 @xmath213 and only @xmath214 comes from decays to the lsp , as expected . from ( [ sneutrino_angle ] ) , the mixing angle is inversely proportional to @xmath215 . thus , as the mass difference between left and right handed sneutrinos becomes smaller , their mixing becomes larger . from the fact that @xmath216 and @xmath217 as well as @xmath218 , the nlsp @xmath155 mixes largely with left - handed sneutrino fields and becomes most likely to be produced through the decays in ( [ channels ] ) . the total relic density is 0.1036 , which lies in the @xmath219 of the wmap value ( with those from the sloan digital sky survey ) @xcite : @xmath220 having shown that the parameter set of ( [ inputs ] ) can reproduce the required relic density , we next investigate the positron flux and discuss the pamela , atic and fermi - lat data for the same parameter set . unlike the decays of heavy superpartners , the decay of the nlsp into the lsp proceeds much more slowly as seen from table [ widths ] ( see also @xcite ) . the reason is as follows . as gauge singlets , the sneutrinos do not couple to gauginos but only to higgsinos , via @xmath221 { \bf y_{\nu } } \tilde \nu_{r},\nonumber\\\end{aligned}\ ] ] in which the yukawa interactions of the higgsinos are @xmath222 these interaction terms show that the exchanges of @xmath223 and @xmath224 induce @xmath225 and the @xmath226 exchange gives rise to the decay @xmath227 the rate of this dileptonic decay is given by @xmath228 after summing over all three lepton generations . integrating over the dalitz density gives @xmath229 \nonumber\\ & - & 2 y^2 \log y - \big[2 \left(y - x\right ) \left(x -1\right)\nonumber\\ & \times & \left(y + x + x y - 3 x^2 \right ) \log \left(\frac{x - y}{x-1}\right)\big ] , \ ] ] which appears in ( [ rate ] ) with the indicated arguments . the two chargino states @xmath230 ( @xmath231 ) with masses @xmath232 , also appear in the decay rate via @xmath233 , in which @xmath234 where @xmath235 , @xmath236 , @xmath237 , and @xmath238 is the @xmath239 gaugino mass . various analyses @xcite of pamela and other satellite data suggest that , rather model independently , the dm candidate must have a lifetime of @xmath240 . this is consistent with the dileptonic decay rate ( [ rate ] ) in that for the parameter choice in ( [ inputs ] ) , one finds @xmath241 leading to a @xmath240 lifetime . this arises from the small @xmath242 ratio that sets the neutrino mass scale . one notices that the value of the rate involves the sum of all three generations of charged leptons ( @xmath243 ) as well as all the conjugated modes . note that there is no contribution to the positron production from the third right - handed sneutrino since it only lives @xmath244 sec and thus decays much faster . in contrast to the case of the mssm with right - handed neutrinos @xcite , the @xmath2 term and the neutrino yukawa couplings are induced dynamically in our model . the two models also differ in terms of the predicted correlation between the neutrino masses ( [ mneut ] ) and the dileptonic decay rate ( [ rate ] ) . in the mssm , these quantities are directly correlated via the yukawa couplings @xmath245 . in our model , the rate involves both @xmath246 and @xmath247 , and , in the heavy higgsino regime where @xmath248 , it scales as @xmath249 , not as @xmath250 . hence , in the large @xmath2 limit there is only a mild dependence on @xmath251 , and the see - saw scale " that suppresses the neutrino masses is then independent of the scale that governs the nlsp decays . however , in this regime the dm lifetime is longer than @xmath252 , and hence the heavy higgsino regime is not particularly preferred given present data . we now crystallize the qualitative approach presented above with a more detailed calculation of the positron flux originating from the decay ( [ leptonic ] ) with the decay width ( [ rate ] ) . we follow the procedure in @xcite , to which we refer the reader for further details . the source term of the diffusion equation is @xmath253 where @xmath254 is the number density of the nlsp , @xmath255 is the lifetime ( which will be set as a free parameter to be fit ) , and @xmath256 is the energy distribution of @xmath257 from the decay @xmath258 . before proceeding to evaluate @xmath256 , we first determine @xmath255 . since both the lsp and the nlsp are components of the dm , other particles decay to either of them . then , the ratio of the number densities of the lsp and nlsp should be equal to the ratios of the branching rates into either state . explicitly , @xmath259 where @xmath260 , and @xmath261 represent all possible particles of the model and the right hand side is known once we fix the free parameters . for example , numerical evaluation yields , @xmath262 and @xmath263 for the particular parameter set considered here . there are two unknown in the left hand side of ( [ ndensity_1 ] ) . the second relation needed arises from the mass density of the dm , @xmath264 , as @xmath265 where @xmath266 is further parametrized by adopting the navarro - frank - white density profile @xcite . for the diffusion parameters , we use the three sets from models m1 , m2 and med @xcite . we return to the evaluation of the @xmath256 term . the definition and normalization of this term depends on whether the positron / electron pair is produced directly from the decay @xmath267 ( that is , @xmath268 ) or through some cascade decays . our numerical analysis shows that in addition to the direct @xmath269 production case , the @xmath270 has also significant decay width . in this case the muon @xmath271 will decay further as @xmath272 where the appropriate combinations of @xmath273 and @xmath274 should be understood based on the charge of @xmath2 or @xmath275 . in principle , the @xmath276 lepton can also be produced directly which then requires two cascade decays to get the positron / electron pair . so , we consider direct productions of positron / electron , or indirect production only through the muon cascade . the details of how to treat the cascade productions of positron / electron from the muon cascade can be found for example in @xcite . if the electron / positron pair is produced directly , then the @xmath256 is given @xmath277}_{e^\pm}}{de_{e^\pm}}=\frac{1}{\gamma_{e^+e^-}}\int de_{e^\mp } \frac{d^2\gamma}{de_{e^+}de_{e^-}}. \label{direct}\end{aligned}\ ] ] where @xmath278 is the decay width for the direct electron / positron production case . if positron / electron is produced from the decay @xmath279 followed by the decay of the muon @xmath280 , we write the @xmath281 term for the positron only for notational simplicity ( the electron case is very similar ) : @xmath282}_{e^+}}{de_{e^+}}=\frac{1}{n_{e^\pm\mu^\mp } } \left ( \frac{d\gamma^{e^+\mu^-}}{de_{e^+ } } + \int de_{\mu^+ } \frac{d^2\gamma^{e^-\mu^+}}{de_{e^+}de_{\mu^+ } } \right ) \label{cascade}\end{aligned}\ ] ] where the second term is represented generically after carrying out the integral over the electron energy @xmath283 . in this term , the muon decays further to a positron with energy @xmath284 . the combined contribution is included in the second term . the first term corresponds to the case where the leptons in the final state are @xmath285 so that there is no further decay considered for the positron case . the normalization factor @xmath286 is given by @xmath287 the solution of the diffusion equation for the flux @xmath288 is given in terms of the green s function in @xcite as @xmath289 where either ( [ direct ] ) or ( [ cascade ] ) should be used for the @xmath290 term depending on the relevant decay mode . here @xmath291 is an effective mass term originating from the solution of ( [ ndensity_1 ] ) and ( [ ndensity_2 ] ) . we further approximate the green s function as in @xcite to maximize computational efficiency . the flux @xmath288 obtained as a solution of the diffusion equation is then used to define the ratio for the positron flux @xmath292 : @xmath293 where @xmath294 is the sum of @xmath288 and the background flux in our galaxy and we define @xmath294 @xmath295 where @xmath296 represents the uncertanities in the @xmath297 background . @xmath298 is used in the numerical study . the following electron and positron backgrounds @xmath299 in our galaxy are approximated as @xcite in units of @xmath300 @xmath301 the quantity @xmath292 in ( [ ratio - positron ] ) can be used to compare with the pamela data . similarly the total flux @xmath302 scaled by @xmath303 is the relevant observable for the atic and the fermi - lat experiments . as a function of positron energy in the m1 propagation model . the direct positron channel ( @xmath304 ) and the indirect one through muon decay ( @xmath305 ) are shown separately . the background and the pamela data are also shown . the fitted lifetime of nlsp is also indicated.,width=345,height=230 ] -0.2 in as a function of positron energy in the m2 propagation model . the direct positron channel ( @xmath304 ) and the indirect one through muon decay ( @xmath305 ) are shown separately . the background and the pamela data are also shown . the fitted lifetime of nlsp is also indicated.,width=345,height=230 ] -0.2 in as a function of positron energy in the med propagation model . the direct positron channel ( @xmath304 ) and the indirect one through muon decay ( @xmath305 ) are shown separately . the background and the pamela data are also shown . the fitted lifetime of nlsp is also indicated.,width=345,height=230 ] -0.2 in our results for the positron flux are shown in figs . ( [ flux1 ] ) , ( [ flux2 ] ) , and ( [ flux3 ] ) as the ratio @xmath306 as a function of the positron energy @xmath284 for each propagation model m1 , m2 , and med , respectively . the direct and indirect production of the positron as well as the expected background are shown separately . the figures show that the models m1 and med fit much better than the m2 model . the fitted curves significantly deviate from the pamela data ( shown in the figures ) for energies less than about @xmath307 but are consistent with the experimental data for @xmath308 . the fitted values of the lifetimes are @xmath309 , which are in the expected range . the lifetimes for the m1 and med models are close to each other , and the value is only slightly smaller for the m2 model . note also that the constraint on the lifetime of the decaying dm from the cosmic microwave background ( cmb ) is analysed in @xcite , and our results obey the lower bounds . the positron emitted through the @xmath305 channel deviates only from the direct channel part of the high energy tail . similar graphs are shown in figs . ( [ atfer1 ] ) , ( [ atfer2 ] ) , and ( [ atfer3 ] ) for the total flux scaled by the positron energy @xmath303 for the propagation models m1 , m2 , and med , respectively . again the direct and indirect production of the positron as well as the expected background are shown separately . the atic and the fermi - lat data are also included . as seen from the figures , our signal is a reasonably good fit to the atic and fermi data . our signal explains the fermi - lat data very well up to @xmath310 gev for the @xmath311 case and gets even better for the @xmath305 case , where the signal is consistent up to @xmath312 gev . in the atic case , the signal explains the low energy as well as the peak regions better than the intermediate enery interval . once again , considering the @xmath305 mode makes the situation better in the end - region . in either case ( atic or fermi - lat ) , the signal would fit better especially in the high energy region for another set of parameter values . for example , a bigger gap between the lsp and nlsp mass would allow more energetic @xmath313 which would shift the dying tail of the signal to the right . unlike the pamela case , where m1 and med scenarios are favored , the m2 model works better for atic , while the fermi - lat data slightly favors the m1 and med models . the fitted lifetimes for the atic data are very close for the propagation models and are also quite consistent with the ones for the pamela case ( especially for the m1 and med models ) . on the other hand , the lifetimes for the fermi - lat data are the same for the three models considered ( the fitted curves are less sensitive to the types of the model used as long as the lifetimes are in the @xmath314 sec range ) . for the m1 and med models , the fermi - lat data requires lifetimes about twice bigger than than the ones for the atic and pamela cases . the ratio is bigger for the m2 propagation model . as a function of the positron energy in the m1 propagation model . both the direct positron channel ( @xmath304 ) and the indirect one through muon decay ( @xmath305 ) are shown separately . the background and the atic and fermi - lat data are also shown . the fitted lifetimes of nlsp for each data are also indicated.,width=403,height=230 ] -0.2 in as a function of the positron energy in the m2 propagation model . both the direct positron channel ( @xmath304 ) and the indirect one through muon decay ( @xmath305 ) are shown separately . the background and the atic and fermi - lat data are also shown . the fitted lifetimes of nlsp for each data are also indicated.,width=403,height=230 ] -0.2 in as a function of the positron energy in the m2 propagation model . both the direct positron channel ( @xmath304 ) and the indirect one through muon decay ( @xmath305 ) are shown separately . the background and the atic and fermi - lat data are also shown . the fitted lifetimes of nlsp for each data are also indicated.,width=403,height=230 ] -0.2 in from the results , we can conclude that it is possible to satisfactorily explain the pamela+atic data simultaneously ( except the data in the 100 - 200 gev range for the atic case ) with a consistent lifetime for the dm . it is also possible to explain the fermi - lat data , which requires a slightly larger lifetime for the dm and works well for energies up to 500 gev . having discussed the electron / positron excess as well as the total flux , we now consider briefly the issue that no significant excess was observed in the anti - proton flux @xcite . unlike the case of the leptonic decays , the hadronic decays are generated only through the @xmath315-term interactions in the superpotential . the relevant interaction terms in the lagrangian are given by @xmath316 in which @xmath317 ( @xmath318 ) are the neutral cp - even higgs bosons and @xmath319 is the fraction of @xmath317 in @xmath320 @xcite . higgs bosons @xmath317 produced in the @xmath321 decay @xcite further decay into pairs of quarks ( including top quark pairs as suggested by the mass spectrum in ( [ inputs ] ) and in table [ spectrum ] , gauge bosons , and higgs bosons . the decay rate to quarks is @xmath322 the integrated dalitz density has the functional form @xmath323\end{aligned}\ ] ] with @xmath324 , @xmath325 , and @xmath326 . the light quarks are produced directly or indirectly through heavy quark , gauge boson , and higgs boson decays , and hadronize to produce protons and anti - protons . an inspection of ( [ rateq ] ) and ( [ rate ] ) shows that the anti - proton signal is suppressed compared to the positron signal by a kinematic factor of order @xmath327 , which is @xmath328 for @xmath329 quarks and @xmath330 for @xmath331 quarks , as needed for consistency with pamela @xcite . however , the higgs bosons that are produced in nlsp to lsp decays fragment efficiently into gauge bosons and top quarks , which in turn give off light quarks and anti - quarks with no apparent yukawa suppression . moreover , the leptons produced with the rate given in ( [ rate ] ) can give rise to hadronic final states at the loop level . these indirect contributions are expected to yield anti - protons with an efficiency around @xmath332 @xcite . the needed suppression of the anti - proton flux with respect to the positron flux can stem from various effects , such as the interaction strengths in ( [ lagran ] ) . one way to suppress @xmath333 with respect to @xmath334 is to have @xcite @xmath335 which imposes an overall suppression on the flavor - changing entries of the neutrino yukawa matrices with respect to the flavor - conserving entries . a justification or realization of such a structure would in principle require a detailed knowledge of the flavor structure of ( [ mneut ] ) . however , it is worth noting that there is no direct correspondence in general between the right - handed neutrino and the active ( left - handed ) neutrino sectors , which can lend credence for the needed suppression . we have presented a simple extension of the mssm with an additional @xmath1 gauge symmetry that couples predominantly to higgs fields . while other sneutrino dark matter models have been proposed to explain the cosmic ray observations , the model presented here does so while also resolving the naturalness problems of the mssm . in this model , an electroweak scale @xmath2 term and appropriately suppressed dirac neutrino masses are generated upon @xmath1 breaking . the right - handed sneutrinos @xmath336 and @xmath11 are the lightest and next - to - lightest superpartners , allowing for a natural explanation of pamela and other experiments like atic and fermi - lat in the context of sneutrino dark matter . a complete study of the dm relic density in this model includes contributions from the lsp and the nlsp , which is essentially stable with respect to the lifetime of the universe . we also analysed the contributions coming from late decaying particles , _ i.e. _ , particles that decay after freeze - out . for the parameter set in ( [ inputs ] ) , we obtain @xmath337 for the total relic density of right - handed sneutrino , consistent with the current wmap value . we then discussed the pamela , atic and fermi - lat data , considering @xmath338 as the source decay for the observed positron excess . the possibility of producing a final positron through muon is also discussed separately . the suppression of the anti - proton flux compared to the positron flux suggests the model - building constraint that the flavor - changing entries of the neutrino yukawa coupling matrix are suppressed with respect to the flavor - conserving entries . the fitted lifetimes for the dm , of the order of @xmath339 sec , fit the pamela data well for positron energies greater than @xmath340 gev , as well as the atic and fermi - lat cases . the fit is better for the m1 and med models described in the text . it can be concluded from the values of the fitted lifetimes that it maybe possible to explain the pamela+atic and fermi - lat data simultaneously in the corresponding energy ranges , though the fermi - lat data requires a slightly larger lifetime values for the dm . in principle , it is foreseeable that a better scan of the model parameters as well as implementing an improved fitting procedure could offer an even closer simultaneous explanation of these three sets of experimental data . we thank the referee for helpful comments and feedback , and for pointing out several crucial issues . it also thanks heather logan for useful discussions . the work of dad is supported by the turkish academy of sciences via the gebip grant and by the turkish atomic energy authority via the cern - cms research grant . lle is supported by the doe grant de - fg-02 - 95er40896 . the work of mf and it is supported in part by the nserc of canada under the grant no . the work of ls is supported in part by a tubitak post - doctoral fellowship . 999 e. komatsu _ et al . _ [ wmap collaboration ] , astrophys . j. suppl . * 180 * , 330 ( 2009 ) [ arxiv:0803.0547 [ astro - ph ] ] . o. adriani _ et al . _ [ pamela collaboration ] , nature * 458 * , 607 ( 2009 ) [ arxiv:0810.4995 [ astro - ph ] ] . s. torii _ et al . _ [ ppb - bets collaboration ] , arxiv:0809.0760 [ astro - ph ] . f. aharonian _ et al . _ [ h.e.s.s . collaboration ] , phys . * 101 * , 261104 ( 2008 ) [ arxiv:0811.3894 [ astro - ph ] ] . a. a. abdo _ et al . _ [ the fermi lat collaboration ] , phys . rev . lett . * 102 * , 181101 ( 2009 ) [ arxiv:0905.0025 [ astro-ph.he ] ] . s. chang and l. goodenough , arxiv:0908.2429 [ astro-ph.he ] ; p. scott , j. conrad , j. edsjo , l. bergstrom , c. farnier and y. akrami , arxiv:0909.3300 [ astro-ph.co ] ; k. y. choi and c. e. yaguna , arxiv:0906.0736 [ hep - ph ] . t. kobayashi , y. komori , k. yoshida and j. nishimura , astrophys . j. * 601 * , 340 ( 2004 ) [ arxiv : astro - ph/0308470 ] ; d. hooper , p. blasi and p. d. serpico , jcap * 0901 * , 025 ( 2009 ) [ arxiv:0810.1527 [ astro - ph ] ] ; h. yuksel , m. d. kistler and t. stanev , arxiv:0810.2784 [ astro - ph ] ; s. profumo , arxiv:0812.4457 [ astro - ph ] ; k. ioka , arxiv:0812.4851 [ astro - ph ] ; n. j. shaviv , e. nakar and t. piran , arxiv:0902.0376 [ astro-ph.he ] ; d. malyshev , i. cholis and j. gelfand , arxiv:0903.1310 [ astro-ph.he ] ; n. kawanaka , k. ioka and m. m. nojiri , arxiv:0903.3782 [ astro-ph.he ] ; y. fujita , k. kohri , r. yamazaki and k. ioka , arxiv:0903.5298 [ astro-ph.he ] . v. barger , w. y. keung , d. marfatia and g. shaughnessy , phys . lett . b * 672 * , 141 ( 2009 ) [ arxiv:0809.0162 [ hep - ph ] ] ; m. cirelli , m. kadastik , m. raidal and a. strumia , nucl . b * 813 * , 1 ( 2009 ) [ arxiv:0809.2409 [ hep - ph ] ] ; l. bergstrom , t. bringmann and j. edsjo , phys . rev . d * 78 * , 103520 ( 2008 ) [ arxiv:0808.3725 [ astro - ph ] ] ; g. olivares , f. atrio - barandela and d. pavon , phys . d * 77 * , 103520 ( 2008 ) [ arxiv:0801.4517 [ astro - ph ] ] ; m. cirelli and a. strumia , arxiv:0808.3867 [ astro - ph ] ; j. h. huh , j. e. kim and b. kyae , arxiv:0809.2601 [ hep - ph ] ; i. cholis , d. p. finkbeiner , l. goodenough and n. weiner , arxiv:0810.5344 [ astro - ph ] ; n. arkani - hamed , d. p. finkbeiner , t. r. slatyer and n. weiner , phys . rev . d * 79 * , 015014 ( 2009 ) [ arxiv:0810.0713 [ hep - ph ] ] ; j. hisano , m. kawasaki , k. kohri and k. nakayama , phys . d * 79 * , 063514 ( 2009 ) [ arxiv:0810.1892 [ hep - ph ] ] ; m. fairbairn and j. zupan , arxiv:0810.4147 [ hep - ph ] ; a. e. nelson and c. spitzer , arxiv:0810.5167 [ hep - ph ] ; d. feldman , z. liu and p. nath , phys . d * 79 * , 063509 ( 2009 ) [ arxiv:0810.5762 [ hep - ph ] ] ; y. bai and z. han , arxiv:0811.0387 [ hep - ph ] ; j. zhang , x. j. bi , j. liu , s. m. liu , p. f. yin , q. yuan and s. h. zhu , arxiv:0812.0522 [ astro - ph ] . w. shepherd , t. m. p. tait and g. zaharijas , arxiv:0901.2125 [ hep - ph ] ; s. c. park and j. shu , arxiv:0901.0720 [ hep - ph ] ; i. z. rothstein , t. schwetz and j. zupan , arxiv:0903.3116 [ astro-ph.he ] ; d. s. m. alves , s. r. behbahani , p. schuster and j. g. wacker , arxiv:0903.3945 [ hep - ph ] ; v. barger , y. gao , w. y. keung , d. marfatia and g. shaughnessy , arxiv:0904.2001 [ hep - ph ] ; j. d. march - russell and s. m. west , phys . b * 676 * , 133 ( 2009 ) ; j. march - russell , s. m. west , d. cumberbatch and d. hooper , jhep * 0807 * , 058 ( 2008 ) . c. r. chen , f. takahashi and t. t. yanagida , phys . b * 671 * , 71 ( 2009 ) [ arxiv:0809.0792 [ hep - ph ] ] ; p. f. yin , q. yuan , j. liu , j. zhang , x. j. bi and s. h. zhu , phys . d * 79 * , 023512 ( 2009 ) [ arxiv:0811.0176 [ hep - ph ] ] ; k. hamaguchi , e. nakamura , s. shirai and t. t. yanagida , arxiv:0811.0737 [ hep - ph ] ; a. ibarra and d. tran , jcap * 0902 * , 021 ( 2009 ) [ arxiv:0811.1555 [ hep - ph ] ] ; c. r. chen , m. m. nojiri , f. takahashi and t. t. yanagida , arxiv:0811.3357 [ astro - ph ] ; s. de lope amigo , w. y. cheung , z. huang and s. p. ng , arxiv:0812.4016 [ hep - ph ] ; c. r. chen , k. hamaguchi , m. m. nojiri , f. takahashi and s. torii , arxiv:0812.4200 [ astro - ph ] ; f. takahashi and e. komatsu , arxiv:0901.1915 [ astro - ph ] ; c. h. chen , c. q. geng and d. v. zhuridov , arxiv:0901.2681 [ hep - ph ] ; k. cheung , p. y. tseng and t. c. yuan , arxiv:0902.4035 [ hep - ph ] ; k. ishiwata , s. matsumoto and t. moroi , arxiv:0903.0242 [ hep - ph ] . o. adriani _ et al . _ , phys . lett . * 102 * , 051101 ( 2009 ) [ arxiv:0810.4994 [ astro - ph ] ] . t. asaka , k. ishiwata and t. moroi , phys . d * 73 * , 051301 ( 2006 ) [ arxiv : hep - ph/0512118 ] ; phys . d * 75 * , 065001 ( 2007 ) [ arxiv : hep - ph/0612211 ] . f. deppisch and a. pilaftsis , jhep * 0810 * , 080 ( 2008 ) [ arxiv:0808.0490 [ hep - ph ] ] . j. mcdonald , jcap * 0701 * , 001 ( 2007 ) [ arxiv : hep - ph/0609126 ] . h. s. lee , k. t. matchev and s. nasri , phys . d * 76 * , 041302 ( 2007 ) [ arxiv : hep - ph/0702223 ] . d. g. cerdeno , c. munoz and o. seto , phys . d * 79 * , 023510 ( 2009 ) [ arxiv:0807.3029 [ hep - ph ] ] . d. g. cerdeno and o. seto , arxiv:0903.4677 [ hep - ph ] . c. r. chen and f. takahashi , jcap * 0902 * , 004 ( 2009 ) [ arxiv:0810.4110 [ hep - ph ] ] . r. allahverdi , b. dutta , k. richardson - mcdaniel and y. santoso , phys . d * 79 * , 075005 ( 2009 ) [ arxiv:0812.2196 [ hep - ph ] ] ; arxiv:0902.3463 [ hep - ph ] ; b. dutta , l. leblond and k. sinha , arxiv:0904.3773 [ hep - ph ] . e. ma , arxiv:0904.4450 [ hep - ph ] . i. gogoladze , n. okada and q. shafi , arxiv:0904.2201 [ hep - ph ] . x. j. bi , x. g. he and q. yuan , arxiv:0903.0122 [ hep - ph ] ; b. kyae , arxiv:0902.0071 [ hep - ph ] . h. s. goh , l. j. hall and p. kumar , arxiv:0902.0814 [ hep - ph ] . m. pospelov and m. trott , jhep * 0904 * , 044 ( 2009 ) [ arxiv:0812.0432 [ hep - ph ] ] . j. e. kim and h. p. nilles , phys . b * 138 * , 150 ( 1984 ) . s. m. barr , phys . rev . lett . * 55 * ( 1985 ) 2778 ; f. del aguila , g. a. blair , m. daniel and g. g. ross , nucl . b * 272 * ( 1986 ) 413 ; j. l. hewett and t. g. rizzo , phys . rept . * 183 * , 193 ( 1989 ) ; m. cvetic and p. langacker , phys . d * 54 * , 3570 ( 1996 ) [ arxiv : hep - ph/9511378 ] ; g. cleaver , m. cvetic , j. r. espinosa , l. l. everett and p. langacker , phys . d * 57 * , 2701 ( 1998 ) [ arxiv : hep - ph/9705391 ] ; nucl . b * 525 * , 3 ( 1998 ) [ arxiv : hep - th/9711178 ] ; d. m. ghilencea , l. e. ibanez , n. irges and f. quevedo , jhep * 0208 * ( 2002 ) 016 [ arxiv : hep - ph/0205083 ] ; s. f. king , s. moretti and r. nevzorov , phys . rev . d * 73 * ( 2006 ) 035009 [ arxiv : hep - ph/0510419 ] . m. cvetic , d. a. demir , j. r. espinosa , l. l. everett and p. langacker , phys . d * 56 * , 2861 ( 1997 ) [ erratum - ibid . d * 58 * , 119905 ( 1998 ) ] [ arxiv : hep - ph/9703317 ] . d. a. demir and y. farzan , jhep * 0603 * , 010 ( 2006 ) [ arxiv : hep - ph/0601096 ] . d. a. demir , l. l. everett and p. langacker , phys . rev . lett . * 100 * , 091804 ( 2008 ) [ arxiv:0712.1341 [ hep - ph ] ] . d. a. demir , g. l. kane and t. t. wang , phys . d * 72 * , 015012 ( 2005 ) [ arxiv : hep - ph/0503290 ] . d. hooper , j. march - russell and s. m. west , phys . b * 605 * ( 2005 ) 228 ; c. arina and n. fornengo , jhep * 0711 * ( 2007 ) 029 . see the url : http://theory.sinp.msu.ru/ pukhov / calchep.html ; a. pukhov , arxiv : hep - ph/0412191 . a. semenov , arxiv:0805.0555 [ hep - ph ] ; a. semenov , comput . commun . * 115 * , 124 ( 1998 ) . g. belanger , f. boudjema , a. pukhov and a. semenov , comput . commun . * 180 * , 747 ( 2009 ) [ arxiv:0803.2360 [ hep - ph ] ] ; g. belanger , f. boudjema , a. pukhov and a. semenov , comput . commun . * 176 * , 367 ( 2007 ) [ arxiv : hep - ph/0607059 ] . e. aliu _ et al . _ [ k2k collaboration ] , phys . rev . lett . * 94 * , 081802 ( 2005 ) [ arxiv : hep - ex/0411038 ] ; t. araki _ et al . _ [ kamland collaboration ] , phys . lett . * 94 * , 081801 ( 2005 ) [ arxiv : hep - ex/0406035 ] . j. l. feng , s. su and f. takayama , phys . d * 70 * , 075019 ( 2004 ) [ arxiv : hep - ph/0404231 ] . d. n. spergel _ et al . _ [ wmap collaboration ] , astrophys . j. suppl . * 170 * , 377 ( 2007 ) [ arxiv : astro - ph/0603449 ] . a. ibarra and d. tran , jcap * 0807 * , 002 ( 2008 ) [ arxiv:0804.4596 [ astro - ph ] ] . e. a. baltz and j. edsjo , phys . d * 59 * , 023511 ( 1998 ) [ arxiv : astro - ph/9808243 ] . l. zhang , x. chen , m. kamionkowski , z. g. si and z. zheng , phys . d * 76 * , 061301 ( 2007 ) [ arxiv:0704.2444 [ astro - ph ] ] . d. a. demir and l. l. everett , phys . d * 69 * , 015008 ( 2004 ) [ arxiv : hep - ph/0306240 ] .

a kinetic treatment is developed for collisionless magnetized plasmas occurring in high - temperature , low - density astrophysical accretion disks , such as are thought to be present in some radiatively - inefficient accretion flows onto black holes . quasi - stationary configurations are investigated , within the framework of a vlasov - maxwell description . the plasma is taken to be axisymmetric and subject to the action of slowly time - varying gravitational and electromagnetic fields . the magnetic field is assumed to be characterized by a family of locally nested but open magnetic surfaces . the slow collisionless dynamics of these plasmas is investigated , yielding a reduced gyrokinetic vlasov equation for the kinetic distribution function . for doing this , an asymptotic quasi - stationary solution is first determined , represented by a generalized bi - maxwellian distribution expressed in terms of the relevant adiabatic invariants . the existence of the solution is shown to depend on having suitable kinetic constraints and conditions leading to particle trapping phenomena . with this solution one can treat temperature anisotropy , toroidal and poloidal flow velocities and finite larmor - radius effects . an asymptotic expansion for the distribution function permits analytic evaluation of all of the relevant fluid fields . basic theoretical features of the solution and their astrophysical implications are discussed . as an application , the possibility of describing the dynamics of slowly time - varying accretion flows and the self - generation of magnetic field by means of a kinetic dynamo effect is discussed . both effects are shown to be related to intrinsically - kinetic physical mechanisms .

1102.0179

this paper is part of an investigation concerning the theoretical formulation of kinetic theory for collisionless astrophysical plasmas in accretion disks ( ads ) around compact objects , and its application to the study of their equilibrium properties and dynamical evolution . note that what is meant here by the word equilibrium is in general a stationary - flow solution , which can also include a stationary radial accretion velocity . in contrast with the majority of previous treatments , which are based on fluid approaches within the context of hydrodynamics ( hd ) or magnetohydrodynamics ( mhd ) @xcite , here we adopt a kinetic approach . this provides a phase - space treatment allowing us to formulate a consistent description of plasma dynamics . kinetic theory is essential for studying both stationary configurations and dynamical evolution of plasmas when kinetic effects are relevant , such as ones associated with conservation of particle adiabatic invariants , temperature anisotropy , finite larmor - radius ( flr ) effects ( as pointed out in ref.@xcite ) and kinetic trapping phenomena . these properties are relevant for magnetized plasmas and in particular for those arising in ads @xcite whenever the plasma is regarded as collisionless or weakly collisional cremasch2008 - 2,cremasch2009 . in the context of astrophysical ads , there are several examples of collisionless plasmas of this kind , with both strong and weak magnetic fields . one is the case of radiatively inefficient accretion flows ( riafs ) @xcite , in geometrically thick disks around black holes consisting of two - temperature plasma , with the ion temperature being much higher than the electron one , and the timescale of the coulomb collision frequency being much longer than the inflow time . other interesting applications occur in ads around neutron stars and white dwarfs : in the inner regions of such disks , where the magnetic field of the central object becomes dominant , ions and electrons can be collisionally decoupled and sustain different temperatures . this happens , in particular , if the radiative cooling time - scale of the electrons is much shorter than the time - scale for electron - ion collisions . in this way , electrons and ions are thermally decoupled : the two charged species acquire unequal temperatures and the accretion flow becomes a two - temperature flow saxton2005,saxton2007 . in our earlier paper ( ref.@xcite , hereafter referred to as paper i ) , we presented preliminary results in this direction , concerning formulation of kinetic theory for investigating stationary solutions for collisionless ad plasmas , focusing on configurations with locally - closed magnetic flux surfaces . the present paper is intended as a continuation of the previous one , with the aim of generalizing the previous solution to arbitrary magnetic field configurations , which are no longer restricted to localized spatial domains in the disk . we refer to fig.1 below and the discussion in section ii for an explicit comparison of the two configurations . more specifically , in paper i the treatment concerned collisionless magnetized plasmas characterized by locally closed and nested magnetic surfaces . for such configurations , it was shown that suitable kinetic distribution functions ( kdfs ) are permitted , describing both kinetic and gyrokinetic ( gk ) equilibria ( see definition in paper i ) , which are represented by generalized maxwellian and bi - maxwellian kdfs . a main feature was the inclusion in the kinetic treatment of both temperature anisotropy and flr effects . in particular , in paper i and in ref.@xcite , it was proved that these equilibria can sustain a stationary kinetic dynamo . as a basic consequence , it was found that both toroidal and poloidal equilibrium magnetic fields can be self generated for quasi - neutral plasmas , _ without ongoing instabilities and/or turbulence phenomena_. in particular , in closed nested field configurations [ and hence in the local absence of net accretion flow ] , the toroidal field component was found to be produced by diamagnetic effects driven by the species temperature anisotropies . as a further development , in paper i and in ref.@xcite , it was pointed out that the kinetic treatment allows one to construct exact fluid equilibria ( identically satisfying the corresponding fluid equations ) . using a perturbative expansion , a well - defined set of kinetic closure conditions was determined analytically for the relevant stationary moment equations . despite more than forty years of observations and theoretical investigations , there is a lot remaining to be understood about the physical processes governing the structure and evolution of ads . they are observed in a wide range of astrophysical contexts @xcite and consist of plasma orbiting a central object with the velocities of the inward accretion flow usually being much smaller than the rotational velocities . in order for the accretion to happen , there needs to be a net outward transport of angular momentum and there are several conceivable mechanisms for producing this ( see for example @xcite ) . the most obvious one is fluid viscosity , but this would need to be an anomalousviscosity , driven by some type of turbulence , rather than a standard viscosity connected with coulomb collisions ( spitzer viscosity ) which would be much too small to explain the observed accretion rates under the conditions actually found in accretion disks ( see lomi09 for a review of turbulence mechanisms ) . however , other collisionless physical mechanisms are possible in principle , such as kinetic instabilities , radiation effects and magnetic reconnection . the aim of the present paper is to help in preparing the way for a discussion of these . we focus on ad plasmas immersed in slowly time - varying magnetic fields , characterized locally by open nested magnetic surfaces . for these systems , no kinetic treatment has been available up to now . the origin of their magnetic fields varies depending on the type of the central object : in the case of black holes , the fields are only ones self - generated by currents in the plasma itself via dynamo effects , while with neutron stars and white dwarfs there can also be a magnetic field intrinsic to the central object @xcite . the interplay between magnetic fields and accretion plasmas can affect the overall velocity profile of the disk , as well as giving rise to species - dependent velocities and rotational frequencies naso , cremasch2008 - 1 , cremasch2008 - 2 . moreover , the magnetic field can be a source of anisotropies in the kdf and allow particular symmetries which influence both the single particle and collective plasma behavior . the transport of angular momentum , the accretion flow and the possible generation of jets @xcite are all strongly dependent on the magnetic field structure and so magnetic fields play an important role for ad physics . the purpose of this paper is to formulate a comprehensive kinetic treatment for collisionless axisymmetric ad plasmas including both accretion flows and collisionless dynamo effects . we include general relative orderings between the magnitudes of the external and self - generated magnetic fields and allow the magnetic field be non - uniform and slowly time - varying while possessing locally nested open magnetic surfaces . extending the investigation developed in paper i , we do this by constructing particular quasi - stationary solutions of the vlasov - maxwell equations , characterized by generalized bi - maxwellian phase - space distributions ( see also paper i ) , which are referred to here as quasi - stationary asymptotic kdfs ( qsa - kdfs ) . as discussed below , the functional form of these solutions is physically motivated . we will show that this makes possible the explicit inclusion of both _ temperature anisotropies _ and _ parallel velocity perturbations _ in the qsa - kdfs ( see the definition below in section 4 ) . this is done , first , by developing an ad hoc formulation for gk theory in the presence of a gravitational field , making it possible to directly construct the relevant particle guiding - center adiabatic invariants . the qsa - kdfs are then expressed in terms of these . remarkably , this allows also the consistent treatment of trapping phenomena due to spatial variations both of the magnetic field and of the total effective potential ( gravitational em trapping ) . second , the qsa - kdfs are constructed by imposing appropriate _ kinetic constraints _ ( see section 4 ) , requiring that suitable _ structure functions _ ( see below ) which enter the definition of the qsa - kdfs , depend only on the azimuthal canonical momentum and total particle energy . by invoking suitable perturbative expansions , it follows that the relevant moments and moment equations can be evaluated analytically . the solution thus obtained can be used for investigating the quasi - stationary dynamics of magnetized ad plasmas , including description of quasi - stationary accretion flows and allowing for the generation of finite poloidal and toroidal magnetic fields . in particular , the kinetic theory predicts the possibility of pure matter inflows as well as the independent coexistence of both inflows and outflows . the paper is organized as follows . in section 2 we summarize the basic assumptions and definitions of the theory . in section 3 we formulate the gk theory for magnetized accretion disk plasmas , deriving the relevant integrals of motion and guiding - center adiabatic invariants and discussing the particle trapping phenomenon . section 4 deals with the construction of a generalized asymptotic stationary kdf , with the inclusion of parallel velocity perturbations and the adoption of suitable kinetic constraints . in section 5 we give an analytic expansion for the kdf and discuss its main features . section 6 deals with the relationship between kinetic theory and the corresponding fluid treatment , which concerns the validity of moment equations and the analytic calculation of fluid fields . section 7 is dedicated to discussing the temporal evolution of the gk equilibria and derivation of the dynamical equation for the gk kdf . in section 8 we investigate the implication of the kinetic solution for the ampere equation and the existence of the kinetic dynamo effect . then , in section 9 we discuss the treatment of quasi - stationary accretion flow within the present formulation , showing that solutions with net radial accretion are admitted consistently with the constraints imposed by the maxwell equations . finally , section 10 contains a summary of the main results with closing remarks . ignoring possible weakly - dissipative effects ( coulomb collisions and turbulence ) , we shall assume that the kdf and the em fields associated with the plasma obey the system of vlasov - maxwell equations , with maxwell s equations being considered in the quasi - static approximation . for definiteness , we shall consider here a plasma consisting of at least two species of charged particles : one species of ions ( @xmath0 ) and one of electrons ( @xmath1 ) . following the treatment presented in paper i , we shall take the ad plasma to be : a ) _ non - relativistic _ , in the sense that it has non relativistic species flow velocities , that the gravitational field can be treated within the classical newtonian theory , and that the non - relativistic vlasov kinetic equation is used as the dynamical equation for the kdf ; b ) _ collisionless _ , so that the mean free path of the plasma particles is much longer than the largest characteristic scale length of the plasma ; c ) _ axisymmetric _ , so that the relevant dynamical variables characterizing the plasma ( e.g. , the fluid fields ) are independent of the azimuthal angle @xmath2 when referred to a set of cylindrical coordinates @xmath3 ; d ) acted on by both gravitational and em fields . the kinetic formulation is intrinsically asymptotic . this means that the theory ( in particular the gk theory formulated in the next section ) is characterized by a suitable species - dependent dimensionless physical parameter @xmath4 , where @xmath5 denotes the species index . here @xmath6 is the species average larmor radius , with @xmath7 denoting the species thermal velocity perpendicular to the magnetic field and @xmath8 being the species larmor frequency . moreover , @xmath9 is the characteristic length - scale of the spatial inhomogeneities of the em field , defined as @xmath10 , where @xmath11 and @xmath12 are the characteristic magnitudes of the gradients of the absolute values of the magnetic field @xmath13 and the electric field @xmath14 , defined as @xmath15 and @xmath16 , where the vector @xmath17 denotes @xmath18 . then , in analogy with paper i , we define a unique parameter @xmath19 . for temperatures and magnetic fields typical of ad plasmas , we have @xmath20 . in the following we will focus on solutions for the equilibrium magnetic field @xmath21 which admit , at least locally , a family of nested and _ open _ axisymmetric toroidal magnetic surfaces @xmath22 , where @xmath23 denotes the poloidal magnetic flux of @xmath21 . see fig . 1 for a schematic comparison between the configuration of locally closed magnetic surfaces considered in paper i and the case of open magnetic surfaces analyzed in the present study . a set of magnetic coordinates ( @xmath24 ) can be defined locally , where @xmath25 is a curvilinear angle - like coordinate on the magnetic surfaces @xmath26 each relevant physical quantity @xmath27 can then be conveniently expressed either in terms of the cylindrical coordinates or as a function of the magnetic coordinates , i.e. @xmath28 where the @xmath29 dependence has been suppressed due to the axisymmetry . we require the em field to be slowly varying in time , i.e. , to be of the form @xmath30 , \label{b0}\]]with @xmath31 being a suitable integer . this time dependence is connected with either external sources or boundary conditions for the kdf . in particular , we shall assume that the magnetic field is of the form @xmath32where @xmath33 and @xmath34 denote the self - generated magnetic field produced by the ad plasma and a finite external magnetic field produced by the central object ( in the case of neutron stars or white dwarfs ) . for greater generality , we shall not prescribe any relative orderings between the various components of the total magnetic field , which are taken to be of the form@xmath35 in particular , here @xmath36 and @xmath37 are the toroidal and poloidal components of the self - field , while the external magnetic field @xmath34 has to be purely poloidal , as a consequence of the axisymmetry , and is defined in terms of the vacuum potential @xmath38 . as a consequence , the magnetic field can also be written in the equivalent form @xmath39where the function @xmath40 is defined as @xmath41 with @xmath42 and @xmath43 defining a set of local magnetic coordinates ( as implied by the equation @xmath44 which is identically satisfied ) . also , it is assumed that the charged particles of the plasma are subject to the action of _ effective _ _ em potentials _ @xmath45 where @xmath46 is the vector potential corresponding to the magnetic field of eq.([b field ] ) , while @xmath47 is given by@xmath48with @xmath49 @xmath50 and @xmath51 denoting the _ effective _ electrostatic potential and the electrostatic and generalized gravitational potentials ( the latter , in principle , being produced both by the central object and the accretion disk ) . the effective electric field @xmath52 can then be defined as@xmath53 in this section we recall the gk theory appropriate for the description of ad plasmas . its formulation is in fact a prerequisite for the construction of the kinetic quasi - stationary equilibria to be developed later . the appropriate generalization of gk theory allowing for the presence of strong gravitational fields should in principle be based on a covariant formulation [ see @xcite ] . however , for non - relativistic plasmas within a gravitational field , the appropriate formulation can also be directly recovered via a suitable reformulation of the standard non - relativistic theory holding for magnetically confined plasmas catto1978,littlejohn1979,littlejohn1981,littlejohn1983,dubin1983,hahm1988,bernstein1985,balescu , meiss1990 . in this case , the appropriate particle lagrangian function can be represented in terms of the effective em potentials @xmath54 with @xmath31 , where @xmath55 is defined in eq.([efp ] ) . in terms of the hybrid variables @xmath56**@xmath57**@xmath58 ( with @xmath17 and @xmath59 denoting respectively the particle position and velocity vectors ) , this is expressed as@xmath60where @xmath61 $ ] and@xmath62denotes the corresponding hamiltonian function in hybrid variables . the gk treatment for the lagrangian ( [ lagrangian ] ) involves the construction - in terms of a perturbative expansion determined by means of a power series in @xmath63 - of a diffeomorphism of the form @xmath64referred to as the _ gk transformation_. note that , in the following , we shall use a prime @xmath65 to denote a dynamical variable defined at the _ guiding - center position _ @xmath66 ( or @xmath67 in axisymmetry ) . here , by definition , the transformed variables @xmath68 * * * * ( _ gk state _ ) are constructed so that their time derivatives to the relevant order in @xmath63 have at least one ignorable coordinate ( a suitably - defined gyrophase @xmath69 ) . as an illustration , we show the formulation of the perturbative theory to leading - order in @xmath70 in this case the gk transformation becomes simply@xmath71where @xmath72 , with @xmath69 denoting the gyrophase angle . in the following , the gk transformation will be performed on all phase - space variables @xmath73 , _ except _ for the azimuthal angle @xmath74 which is left unchanged @xcite and is therefore to be considered as one of the gk variables . here @xmath75 with @xmath76**@xmath57**@xmath77 while @xmath78 and @xmath79 are respectively the guiding - center larmor frequency and the _ effective drift velocity _ produced by @xmath80 namely@xmath81the rest of the notation is standard , with @xmath82 and @xmath83 denoting respectively the parallel and perpendicular ( guiding - center ) velocities , both defined relative to the frame locally moving with velocity @xmath84 . it follows that , when expressed in terms of the gk variables @xmath68 , the gk lagrangian and hamiltonian functions , @xmath85 and @xmath86 , can be evaluated with the desired order of accuracy . in particular , to leading - order , i.e. neglecting corrections of @xmath87 with @xmath88 , @xmath89 and @xmath90 , where @xmath91 and @xmath92 recover the customary expressions@xmath93@xmath94with the magnetic moment @xmath95 to leading order , while the gyrophase - independent _ modified em potentials _ are@xmath97 in the same approximation . it is important to stress that the gk theory can be performed in principle to arbitrary order in @xmath63 catto1978,littlejohn1979,littlejohn1981,littlejohn1983,dubin1983,hahm1988,bernstein1985,balescu , meiss1990 , thus permitting the explicit determination of @xmath98 and the modified em potentials as well as the relevant guiding - center canonical momenta . the exact integrals of motion and the relevant adiabatic invariants corresponding respectively to eqs.([lagrangian ] ) and ( [ lagr girocinetica ] ) can be immediately recovered . by definition , an adiabatic invariant @xmath99 of order @xmath100 with respect to @xmath63 is conserved only in an asymptotic sense , i.e. , in the sense that @xmath101 , where @xmath102 is a suitable integer . first we notice that , under the assumptions of axisymmetry and of eq.([b0 ] ) , the only first integral of motion is the canonical momentum @xmath103 conjugate to the ignorable azimuthal angle @xmath29:@xmath104since the azimuthal angle @xmath29 is ignorable also for the gk lagrangian @xmath85 , it follows that the quantity @xmath105 is an adiabatic invariant of the prescribed order , according to the accuracy of the gk transformation used to evaluate @xmath106 . we shall refer to @xmath107 as the _ guiding - center canonical momentum_. in particular , correct to @xmath108 , with @xmath31 , one obtains@xmath109which is an adiabatic invariant of @xmath110 , with @xmath111 . furthermore , the total particle energy @xmath112with @xmath88 , and the gk hamiltonian @xmath113 are also adiabatic invariants of order @xmath100 . finally , in gk theory , by construction , the momentum @xmath114 conjugate to the gyrophase , as well as the related magnetic moment @xmath98 defined as@xmath115 are adiabatic invariants . as shown by kruskal ( 1962 @xcite ) it is always possible to determine @xmath85 so that @xmath116 is an adiabatic invariant of arbitrary order in @xmath117 . in particular , the leading - order approximation is @xmath118 . note that the allowance of slow time variations for @xmath119 is an elementary consequence of assumption ( [ b0 ] ) , which allows us to describe realistic configurations of ad plasmas which slowly evolve in time . gk theory permits explicit treatment of particle trapping corresponding to the existence of forbidden regions for the motion of charged particles arising from conservation of energy and magnetic moment . conservation of the guiding - center hamiltonian ( [ ham giricinetica ] ) and the magnetic moment @xmath120 ( leading - order approximation ) give rise to some implications . combining the two identities to express the parallel velocity @xmath121 , and using the definition ( [ v^drift ] ) , we find@xmath122 } . \label{trap}\]]therefore @xmath82 is a local function of the guiding - center position vector @xmath67 and , thanks to axisymmetry , of the corresponding flux coordinates ( @xmath123 ) . since the argument of the square root must be non - negative , this means that @xmath121 is only defined in the subset of the configuration space spanned by ( @xmath123 ) where this property holds . it follows that if the argument becomes null for given @xmath124 and @xmath125 , the parallel velocity must change sign so that the particle undergoes a spatial reflection . the points of the configuration space where this occurs are the so - called _ mirror points_. the existence of these points may generate various kinetic phenomena in ad plasmas . in particular , for open magnetic surfaces , particles can in principle experience zero , one or two reflections corresponding respectively to _ passing particles ( pps ) , bouncing particles ( bps ) _ and _ trapped particles ( tps)_. in the present case , since the right hand side of eq.([trap ] ) depends on the magnitude of the magnetic field ( @xmath126 ) , the effective potential energy ( @xmath127 ) and the centrifugal potential ( @xmath128 ) , we shall refer to the tps case as _ gravitational em trapping_. in section vii we shall investigate some consequences of trapping phenomena for the dynamics of ads . in this section we show that the equilibrium generalized bi - maxwellian solution for the kdf obtained in paper i can be extended to qsa - kdfs describing axisymmetric ad plasmas with the following features : \1 ) the kdf is also axisymmetric ; \2 ) each species in the collisionless plasma is considered to be associated with a suitable set of _ sub - species _ ( referring to the different populations mentioned above ) , each one having a different kdf ; \3 ) temperature anisotropy : for all of the species , it is assumed that different parallel and perpendicular temperatures are allowed ( with respect to the local direction of the magnetic field ) ; \4 ) accretion flow velocity : a non - vanishing species dependent poloidal flow velocity is prescribed ; \5 ) open , locally nested magnetic flux surfaces : the magnetic field is taken to allow quasi - stationary solutions with magnetic flux lines belonging to open and locally nested magnetic surfaces ; \6 ) kinetic constraints : suitable functional dependencies are imposed so that the kdf is an adiabatic invariant ; \7 ) analytic form : the solution is required to be asymptotically close to a local bi - maxwellian in order to permit comparisons with previous literature dealing with maxwellian or a bi - maxwellian kdfs ( see for example @xcite ) . requirement 2 ) is suggested by observations of collisionless plasmas . for example , in the solar wind plasma both ion and electron species are described by superpositions of shifted bi - maxwellian distributions . requirements 1 ) - 7 ) clearly imply that the solution can not generally be a maxwellian . however , in analogy with paper i , it is possible to show that they can be fulfilled by a suitable modified bi - maxwellian expressed solely in terms of first integrals of motion and adiabatic invariants cremasch2008 - 2,cremasch2009,kocha10 . it follows that this is necessarily a qsa - kdf . a set of fluid equations can then readily be determined using this solution , expressed in terms of four moments of the kdf [ corresponding to the species number density , flow velocity and the parallel and perpendicular temperatures ] . these equations which , by construction , satisfy a kinetic closure condition , are also useful for comparing with previous fluid treatments . for consistency with the notation of paper i , we again use the symbol @xmath129 to denote physical quantities which refer to the treatment of anisotropic temperatures , unless otherwise specified , but in the present work , for greater generality , the symbol @xmath130 is used to denote variables which depend on both the canonical momentum @xmath131 and the total particle energy @xmath119 . in line with all of the previous requirements , it is possible to show that a particular solution for the qsa - kdf is given by:@xmath132which we refer to as the _ generalized bi - maxwellian kdf _ _ with parallel velocity perturbations_. here @xmath133 is defined in the phase - space @xmath134 , where @xmath135 and @xmath136 are both identified with suitable subsets of the euclidean space @xmath137 . the notation is as follows:@xmath138with @xmath119 and @xmath131 given by eqs.([total_energy ] ) and ( p_fi ) respectively , while @xmath139 . by construction , @xmath140 has the dimensions of an angular momentum , while @xmath141 has those of a frequency . in contrast with the solution obtained in paper i , @xmath142 is not necessarily associated here with a purely azimuthal leading - order velocity . in general @xmath143 can , in fact , be represented as@xmath144here @xmath145 has the same meaning as the analogous quantity used in paper i , with @xmath146 being related to the azimuthal rotational frequency . in eq.(kstar ) @xmath147 is a frequency associated with the leading - order guiding - center canonical momentum @xmath107 defined in eq.([p_fi - hat ] ) , which is an adiabatic invariant depending on @xmath82 and , by definition , is independent of the gyrophase angle . as we shall show at the end of this section , this feature can be used to require that the qsa - kdf carries a non - vanishing parallel flow velocity . this can be related to a net accretion flow arising in the ad plasma . finally , by substituting eq.([kstar ] ) into eq.([sol1 ] ) we reach the equivalent representation for the qsa - kdf:@xmath148 in order for the solution ( [ sol2 ] ) [ or equivalently ( [ sol1 ] ) ] to be a function of the integrals of motion and of the adiabatic invariants , the functions @xmath149 , which we will refer to as _ structure functions _ , must be adiabatic invariants by themselves . to further generalize the solution of paper i , we shall here retain a functional dependence on both the total particle energy and the canonical momentum , thus imposing the functional dependencies@xmath150which will be referred to in the following as _ kinetic constraints_. the kinetic constraints ( [ kincon ] ) provide the most general solution for @xmath133 . it can be shown that the physical motivation behind imposing these dependencies lies essentially in the fact that the asymptotic condition of small inverse aspect ratio ( adopted previously in paper i ) is no longer valid . in the present context , the kinetic solution is no longer restricted to localized spatial domains in the disk but applies to the general configuration of open magnetic surfaces . this in turn implies that the structure functions are generally not simply flux - functions on the magnetic surfaces . in previous treatments ( paper i and ref.@xcite ) , the structure functions were identified with @xmath151 and @xmath152 for non - isotropic and isotropic generalized maxwellian kdfs respectively . some basic properties of @xmath133 are : property 1 : @xmath133 is itself an adiabatic invariant , and is therefore an asymptotic solution of the stationary vlasov equation , i.e. , a qsa - kdf ; property 2 : @xmath133 is only defined in the subset of phase - space where the adiabatic invariants @xmath107 , @xmath153 and @xmath98 are defined . it follows that @xmath133 is suitable for describing both circulating and trapped particles ( see the related discussion in section 7 ) ; property 3 : all of the velocity - moment equations obtained from the vlasov equation ( and in particular the continuity and linear momentum fluid equations ) are identically satisfied in an asymptotic sense , i.e. , neglecting corrections of @xmath154 ; property 4 : its velocity moments , to be identified with the fluid fields , are unique once @xmath133 is prescribed in terms of the structure functions ; property 5 : it generalizes the solution earlier presented in paper i : a ) by using both @xmath107 and @xmath98 as adiabatic invariants and b ) because of the new kinetic constraints . it follows immediately that the solution ( [ sol2 ] ) does indeed carry finite parallel velocity perturbations . invoking the definitions ( p_fi - hat ) and ( [ kstar ] ) , eq.([sol2 ] ) can be re - written as@xmath155 } { \left ( 2\pi /m_{s}\right ) ^{3/2}\left ( t_{\parallel \ast s}\right ) ^{1/2}}\right . \label{sol3 } \\ & & \times \exp \left\ { -\frac{m_{s}\left ( \mathbf{v}-\mathbf{v}_{\ast s}-u_{\parallel \ast s}^{\prime } \mathbf{b}^{\prime } \right ) ^{2}}{% 2t_{\parallel \ast s}}-m_{s}^{\prime } \widehat{\alpha _ { \ast s}}\right\ } , \notag\end{aligned}\]]where @xmath156 and@xmath157here the function @xmath158 is defined as @xmath159with @xmath160 . note that @xmath161 is non - zero only if the toroidal magnetic field is non - vanishing . this quantity is independent of @xmath162 and is clearly associated with a parallel flow velocity ( i.e. , having both poloidal and toroidal components ) , referred to here as a _ parallel velocity perturbation_. this perturbation enters the solution via the adiabatic invariant @xmath107 and therefore its inclusion is consistent with the requirement that kdf is an adiabatic invariant . finally we note that the same kinetic constraints ( [ kincon ] ) also apply to the solution ( [ sol3 ] ) . however , the functions @xmath163 , $ ] @xmath164 @xmath165 and @xmath166 can not be directly regarded as _ fluid fields _ , since they still depend on the single particle velocity via the canonical momentum @xmath131 and the particle energy @xmath119 . based on properties 1 - 5 , in this section we determine an approximate analytical expression for @xmath133 obtained by means of suitable asymptotic expansions . these are carried out in terms of the following two dimensionless parameters : \1 ) @xmath167 : which is related to the canonical momentum @xmath168 . this is defined as ( cf paper i ) @xmath169 , where @xmath170 and @xmath171 denotes the species particle angular momentum . we refer to the ad plasma as being _ strongly magnetized _ if @xmath172 ; \2 ) @xmath173 : which is related to the total particle energy @xmath119 . this is defined as @xmath174 , i.e. , it is the ratio between the kinetic energy and potential energy of the particle . for bound orbits @xmath175 , and so @xmath176 . in the following , we treat @xmath167 and @xmath173 as infinitesimals of the same order , with @xmath177 and then @xmath167 and @xmath173 can be used for performing a taylor expansion of the implicit dependencies contained in the structure functions by setting @xmath178 and @xmath179 to leading order . this implies that the linear asymptotic expansion for the structure functions , obtained neglecting corrections of @xmath180 as well as of @xmath181 and @xmath182 , with @xmath183 , is@xmath184 _ { \substack { \psi _ { \ast s}=\psi \\ e_{s}={z_{s}e}\phi _ { s}^{eff}}}+ \notag \\ & & + \left ( e_{s}-{z_{s}e}\phi _ { s}^{eff}\right ) \left [ \frac{\partial \lambda _ { \ast s}}{\partial e_{s}}\right ] _ { \substack { \psi _ { \ast s}=\psi \\ e_{s}=% { z_{s}e}\phi _ { s}^{eff } } } , \label{espan0}\end{aligned}\]]where @xmath185to perform the corresponding expansion for @xmath133 , we leave unchanged the dependence in terms of the guiding - center canonical momentum @xmath186 , while retaining the leading - order approximation for the magnetic moment only in the linear perturbation terms of eq.(espan0 ) . then , it is straightforward to prove that for strongly magnetized and bound plasmas , the following relation holds to leading - order : @xmath187 , \label{solo}\]]where @xmath188 and @xmath189 represent the so - called _ diamagnetic parts _ of @xmath133 ( see the definition below ) . the definitions are then as follows : _ first _ , the leading - order distribution @xmath190 is expressed as@xmath191which we will here call the _ bi - maxwellian kdf with parallel velocity perturbations_. here @xmath192 is related to the temperature anisotropy , the number density is defined as@xmath193\]]and@xmath194with @xmath195 denoting the _ pseudo - density_. the function @xmath196 is defined as @xmath197note that @xmath198 and @xmath199 define , respectively , the leading - order azimuthal flow velocity and the leading - order parallel velocity perturbation of the fluid . then , the following kinetic _ _ _ _ constraints are implied from ( [ kincon ] ) , to leading - order , for the structure functions:@xmath200 _ second _ , the diamagnetic parts @xmath188 and @xmath189 of @xmath201 , due respectively to the expansions of the canonical momentum and the total energy , are given by@xmath202 + \frac{% m_{s}r}{t_{\parallel s}}y_{2}\right\ } \left ( \mathbf{v\cdot } \widehat{e}% _ { \varphi } \right ) , \label{hd1 } \\ h_{ds}^{2 } & = & \frac{m_{s}}{2z_{s}e}\left\ { y_{4}-\frac{z_{s}e}{t_{\parallel s}}y_{5}+\frac{p_{\varphi s}^{\prime } \xi _ { s}}{t_{\parallel s}}% c_{5s}\right\ } v^{2}. \label{hd2}\end{aligned}\]]here @xmath203 @xmath204 , is defined as @xmath205 , \\ y_{2 } & \equiv & \omega _ { s}\left [ 1+\psi a_{3s}\right ] , \\ y_{3 } & \equiv & \left [ \frac{p_{\varphi s}^{\prime } \xi _ { s}}{t_{\parallel s}}% a_{5s}-a_{2s}\frac{p_{\varphi s}^{\prime } \xi _ { s}}{t_{\parallel s}}\right ] , \\ y_{4 } & \equiv & \left [ c_{1s}+c_{2s}\left ( \frac{h_{s}}{t_{\parallel s}}-% \frac{1}{2}\right ) -\mu _ { s}^{\prime } \widehat{c_{4s}}\right ] , \\ y_{5 } & \equiv & \left [ 1+\frac{\omega _ { s}\psi } { c}c_{3s}\right ] , \end{aligned}\]]where @xmath206 and the following definitions have been introduced : @xmath207 @xmath208 @xmath209 @xmath210 @xmath211 and @xmath212 @xmath213 @xmath214 , @xmath215 @xmath216 . we should make a number of comments here : \1 ) the functional forms of the leading - order number density , the parallel and azimuthal flow velocities and the temperatures carried by the bi - maxwellian kdf , are naturally determined in terms of @xmath217 and @xmath218 . the effective potential @xmath55 is generally a function of the form @xmath219 with @xmath220 , since generally neither the gravitational potential nor the electrostatic potential are expected to be flux functions in the present case . hence , in magnetic coordinates , it follows that the structure functions are of the form @xmath221 ; \2 ) the coefficients @xmath222 and @xmath223 , @xmath204 , can be identified with effective _ thermodynamic forces _ : @xmath224 carries the contribution of the parallel velocity perturbation , while the @xmath223 , @xmath204 , are due to the energy dependence contained in the structure functions ; \3 ) we stress that the energy dependence contained in the kinetic constraints is non trivial and can not be included simply by redefining the structure functions ( e.g. , by transforming the magnetic coordinates ) . in fact , besides modifying the leading order structure functions ( see point 1 above ) , it gives rise to the new diamagnetic contribution @xmath189 . eq.([solo ] ) is therefore a generalization of the analogous solution obtained in paper i , which also appears in standard tokamak transport theory @xcite , where the relevant structure functions were considered solely as flux functions . including the effect of the parallel velocity perturbations gives rise to contributions to @xmath189 which are even with respect to @xmath121 ; \4 ) in the analytical expansion , we have assumed that the scale - length @xmath225 is of the same order in @xmath167 as the characteristic scale - lengths associated with the structure functions ; \5 ) we have performed the analysis distinguishing between the different plasma species . since this is an asymptotic estimation , the analytical expansion can be different for ions and electrons , particularly for the terms appearing in the diamagnetic part , depending on the relative magnitudes of the parameters @xmath167 and @xmath173 . on the other hand , because of the double expansion and the energy dependence , the asymptotic solution for the two species can hold also in different spatial domains ; \6 ) the kdf @xmath226 also satisfies property 2 : namely , it is only defined in the subset of phase - space where the parallel velocity @xmath227 is a real function . it is therefore suitable for properly describing particle trapping ; \7 ) finally , we stress that the qsa - kdf ( [ sol2 ] ) obtained here , reduces asymptotically to the expression reported in previous paper ( see eq.(10 ) in paper i ) when the following conditions are satisfied : a ) parallel velocity perturbations are ignored , namely the structure function @xmath147 is set to zero ; b ) closed nested magnetic surfaces are considered ; c ) large aspect ratio ordering , @xmath228 , is invoked ( see the definition in paper i ) . in this case , the effective potential is solely a flux - function to leading order , while the diamagnetic contribution @xmath189 can be shown to be of higher order than @xmath188 . in this section we discuss the connection between the kinetic treatment presented here and the corresponding fluid approach , obtained by describing the plasma in terms of a suitable set of fluid fields . the latter can in principle be specified as required by experimental observations and identified with the relevant physical observables . important practical aspects of the present theory concern the explicit evaluation of the fluid fields associated with the qsa - kdf , and the conditions for validity of the relevant moment equations . for definiteness , let us require that : 1 . the kdf , the em fields @xmath229 and the corresponding em potentials @xmath230 are all exactly axisymmetric and , moreover , stationary in an asymptotic sense , i.e. neglecting corrections of @xmath231 ; 2 . the kdf is identified with _ _ _ _ the qsa - kdf @xmath232 which , by assumption , is required to be an adiabatic invariant of @xmath231 . by construction @xmath233 is a solution of the _ _ asymptotic vlasov equation__@xmath234this equation holds by definition up to infinitesimals of @xmath235 , where @xmath100 is an arbitrary positive integer ; 3 . the magnetic field is taken to be of the form ( [ b field ] ) . as a basic consequence of these assumptions , the stationary fluid equations following from the vlasov equation are necessarily all identically satisfied in an asymptotic sense , i.e. , again neglecting corrections of @xmath235 . in fact if @xmath236 is an arbitrary weight function , identified for example with @xmath237 then the generic moment of eq.([asvla ] ) is:@xmath238where @xmath136 denotes the appropriate velocity space of integration . using the chain rule , this can be written as@xmath239on the other hand , eq.([dd ] ) can also be represented as@xmath240 -\widehat{f_{\ast s}}\frac{d}{dt}z\right\ } = 0+o\left ( \varepsilon _ { m}^{n+1}\right ) , \]]which recovers the usual form of the velocity - moment equations in terms of suitable ( and _ uniquely defined _ ) fluid fields . for @xmath241 one obtains , in particular , that the species continuity and linear momentum fluid equations are satisfied identically up to infinitesimals of @xmath242 @xmath243 @xmath244 similarly , the law of conservation of the species total canonical momentum can be recovered by setting @xmath245 , namely@xmath246 = 0+o\left ( \varepsilon _ { m}^{n+1}\right ) .\]]in the stationary case this implies the _ _ species angular momentum conservation law__@xmath247 + \frac{z_{s}e}{c}\nabla \psi \cdot n_{s}^{tot}\mathbf{v}_{s}^{tot}=0 \label{angcons}\]]for the species angular momentum@xmath248here the notation is standard . in particular the following velocity moments of the qsa - kdf can be introduced : \a ) _ _ species number density__@xmath249 \b ) _ _ species flow velocity__@xmath250 \c ) _ _ species tensor pressure__@xmath251 \d ) _ _ species canonical toroidal momentum__@xmath252 it is worth remarking here that _ the velocity moments are unique once the qsa - kdf _ @xmath133 [ see eq.([sol1 ] ) ] _ is prescribed in terms of the structure functions _ @xmath253 on the other hand , as a result of eqs.([asvla ] ) and ( [ dd ] ) , it follows that the stationary fluid moments calculated in terms of the qsa - kdf @xmath254 are identically solutions of the corresponding stationary fluid moment equations . in particular , imposing the _ quasi - neutrality condition _ in the sense@xmath255with @xmath256 , the total fluid canonical toroidal momentum and the fluid angular momentum necessarily coincide , namely@xmath257let us now illustrate explicitly how it is possible to carry out such a calculation within the present theory . the evaluation of the previous fluid fields can be made by using the asymptotic analytical solution of the qsa - kdf @xmath133 derived in the previous section and given by eq.([solo ] ) . for example , adopting this expansion in the limit of strongly magnetized plasmas , from eq.([fff1 ] ) the species number density becomes@xmath258 \right\ } , \]]in which the diamagnetic corrections to the bi - maxwellian kdf @xmath259 are polynomial functions of the particle velocity . analogous expressions can also be obtained in a straightforward way for the remaining fluid moments . as pointed out in paper i and subsequently in ref.@xcite , the expansion procedure for @xmath133 can in principle be performed to higher order , allowing for the analytical computation of the corresponding quasi - stationary fluid fields and the determination of the relevant kinetic closure conditions for the stationary moment equations . in the present context we stress that the theory allows the treatment of multiple - species plasmas including , in particular , particle trapping phenomena . this is taken into account by proper definition of the velocity sub - space @xmath260 in which the integrations are performed . in fact , charged particles in both open and closed configurations can have mirror points ( tps and bps ) or be pps , which are free to stream through the boundaries of the domain . these populations give different contributions to the relevant fluid fields and therefore require separate statistical treatments . the explicit calculation of fluid fields requires also a preliminary _ inverse transformation _ representing all quantities in terms of the actual particle positions ( the flr expansion , see eq.(trasformazionui guirocinetiche ) ) . this introduces further correction terms of order @xmath261 , @xmath31 , into the final analytical expressions . in contrast with the conclusion reached in paper i , here we expect these flr corrections to be non - negligible due to the requirement @xmath262 holding for open - field configurations . in this section we investigate the temporal evolution of the axisymmetric qsa - kdf , consistent with the assumptions of section 2 and the results of sections 3 and 4 . two different issues must be addressed : giving an estimate of the maximum time interval over which the qsa - kdf can be regarded as an asymptotic stationary solution ; and determining the solution of the vlasov equation for time intervals longer than the equilibrium one . for our explicit determination of the time evolution of the qsa - kdf , we make the following assumptions : \1 ) that the plasma can be treated as a continuous medium in the kinetic description . this requires that the species kinetic equation holds on time and spatial scales which are much longer than the corresponding langmuir characteristic times and debye lengths ; \2 ) that we are considering timescales much shorter than the species characteristic collisional time @xmath263 , so that it is appropriate to use the vlasov equation ; \3 ) that the species kdf and the em fields vary slowly in time and space with respect to the corresponding larmor times and radii , so that the gk description is valid ; \4 ) that the em and gravitational fields vary slowly in time , in accordance with eq.([b0 ] ) , so that the total energy @xmath119 is an adiabatic invariant . in particular , we require:@xmath264which implies that @xmath265 , with @xmath266 . consistently with the properties of solution ( [ sol2 ] ) , we take @xmath267 as a specific case . note that from here on , @xmath268 will denote the species characteristic time associated with the larmor rotation ( the _ larmor rotation time _ ) . since @xmath269 , it follows that@xmath270 ; \ ] ] \5 ) that the magnetic moment @xmath98 and the guiding - center canonical momentum @xmath107 can be taken as adiabatic invariants of @xmath271 with @xmath272 . the ordering @xmath273 holds for the leading - order expression for @xmath186 adopted here as follows from eq.([p_fi - hat ] ) and the fact that , by definition , higher - order correction terms , @xmath274 , to @xmath107 are independent of the gyrophase angle @xmath275 . in fact , denoting by @xmath276 the second - order gk lagrangian , @xmath277 can be estimated as @xmath278 $ ] where , by construction , @xmath279 and @xmath280 are both gyrophase independent . note that the assumption made here requires the construction of a higher - order gk theory in order to correctly determine @xmath98 to the required order in the larmor - radius expansion . the time evolution of the qsa - kdf is in principle determined by two different mechanisms : the explicit time variation of the em and gravitational fields , and the time variation of the guiding - center adiabatic invariants . however , the choice of the orderings in 4 ) and 5 ) above , allows the time dependence produced only by the em and gravitational fields to be singled out . when assumptions 1 ) - 5 ) above hold , it follows that @xmath281 , with @xmath282 being determined by eq.([dten ] ) . then , ignoring higher - order corrections@xmath283where@xmath284and so the solution @xmath133 can be regarded as an exact kinetic equilibrium for all times @xmath285 such that@xmath286where @xmath287 . within the scope of the above assumptions , we now determine the dynamical evolution equation which describes the slow time - evolution of the qsa - kdf @xmath254 , for time intervals such that @xmath288 is within @xmath289 in analogy with ref.@xcite , we denote by@xmath290the exact solution of the collisionless vlasov equation , for which @xmath291 . here @xmath292 is referred to as the _ reduced kdf_. following the discussion in ref.@xcite , regarding the evaluation of @xmath293 : it is straightforward to prove that @xmath292 is gyrophase independent , to lowest order , in the sense that @xmath294 . therefore , identifying the gk variables with the set @xmath295 , we shall assume that @xmath292 is axisymmetric and of the form @xmath296 . the gyro - averaged dynamical equation for @xmath292 can then be obtained to next order by introducing the gyro - average operator @xmath297 defined as@xmath298with the operation being performed while all of the other gk variables are held fixed @xcite . it follows that , to leading - order , @xmath299 , where the time variation of the guiding - center magnetic coordinate @xmath300 is given by @xmath301 \cdot \nabla ^{\prime } \vartheta ^{\prime } $ ] to leading - order , with the equation of motion for @xmath302 following from the gyrokinetic lagrangian [ e.g. from the leading - order eq.([lagr girocinetica ] ) ] . then , consistently with these assumptions and ignoring higher - order corrections , it is found that the gk reduced kdf @xmath292 obeys the _ _ reduced gk - vlasov equation__@xmath303it follows that , to leading - order@xmath304denoting @xmath305 , @xmath306 is then defined as @xmath307 . the remaining gyrophase average in the last equation can be performed in a straightforward way using eqs.([dten ] ) and ( [ s ] ) . eq.([timeevol2 ] ) clearly also holds in the time interval ( [ timeint1 ] ) , and so it determines the slow time - evolution for all times @xmath308 . for consistency , the non - stationary maxwell equations must also be solved with the same accuracy . eq.([timeevol2 ] ) must be supplemented by appropriate boundary conditions : for open magnetic surfaces with boundaries prescribed on a given magnetic surface @xmath309 , at @xmath310 and @xmath311 , with @xmath312 and @xmath313 , @xmath314 representing the internal and external boundaries , these are defined respectively either by prescribing @xmath315 or @xmath316 ( see fig.2 for a schematic view of the configuration geometry and the meaning of the notation ) . both @xmath317 and @xmath318 are necessarily of the form ( [ pos1 ] ) but their moments remain arbitrary in principle . as indicated below , this is essential for making comparisons with experimental observations . the results obtained here have important consequences for the kinetic description of slow time - evolution of collisionless ad plasmas . fluid fields and moment equations can be explicitly determined in terms of eq.(timeevol2 ) by invoking the perturbative expansion outlined in section 5 and the relations given in section 6 . in this section we apply the kinetic solution for the qsa - kdf to discuss the properties of the ampere equation and the implications for the self - generation of magnetic field by the quasi - stationary ad collisionless plasma . we refer here to this phenomenon as a _ quasi - stationary kinetic dynamo effect_. generalizing the treatment presented in paper i , the ampere equation for the self magnetic field becomes:@xmath319where @xmath33 has been defined in eq.([bself ] ) and here we have distinguished between the contributions arising from pps , bps and tps , denoting the corresponding total current densities as @xmath320 and @xmath321 . as described in sections 5 and 6 , these fluid fields can be calculated in closed analytic form to the required order , by using the asymptotic expansion of the qsa - kdf . this gives:@xmath322 \right\}\end{aligned}\]]for @xmath323 and where @xmath324 denotes the appropriate velocity space domain of integration for trapped , bouncing and passing particles respectively . for convenience of notation , in the following we shall denote as @xmath325 the total current density entering eq.([ampere-2 ] ) . it is possible to prove that in the case of open magnetic surfaces the total current density @xmath326 in general has non - vanishing components along all of the three directions identified by the set of magnetic coordinates @xmath43 . hence , @xmath326 can be represented as@xmath327let us now proceed with the study of the ampere equation . the toroidal component of eq.([ampere-2 ] ) gives , as usual , the generalized grad - shafranov equation for the poloidal flux function @xmath328:@xmath329where the elliptic operator @xmath330 is defined as @xmath331 . the remaining terms of eq.([ampere-2 ] ) along the directions @xmath332 and @xmath333 give two equations for the toroidal component of the magnetic field @xmath334 . these are respectively@xmath335yielding the constraint@xmath336which is a solubility condition for the structure functions . in this regard we notice that as a consequence of the kinetic constraints the function @xmath337 in the previous equations is of the form @xmath338 , i.e. , in contrast to paper i it is no longer a flux - function . therefore , the solubility condition ( [ solubility ] ) can always be satisfied . eqs.([gsapp])-([solubility ] ) therefore provide consistent solutions for both poloidal and toroidal self magnetic fields in a collisionless ad plasma . it is remarkable that in principle all of the populations of charged particles ( pps , bps and tps ) can contribute to the generation of the toroidal magnetic field . more precisely , the following mechanisms can be involved : # 1 ) flr and diamagnetic effects , driven by temperature anisotropy , of the same kind as those described in paper i ; # 2 ) parallel velocity perturbations @xmath165 , which generate a poloidal flow velocity , giving a related contribution to the electric current density through @xmath339 and @xmath340 ; # 3 ) flr effects driven by the remaining thermodynamic forces ( see section 5 ) . these contributions are produced by the diamagnetic kdf and arise because of the asymptotic ordering introduced here ; # 4 ) gyrophase - dependent contributions driven by the same thermodynamic forces . these are originated by the inverse gk transformation of the guiding - center quantities in the qsa - kdf . as discussed above , contributions # 2 and # 4 were negligible under the circumstances discussed in paper i. therefore they should be considered as characteristic features of open - field configurations . we refer to the mechanism of self - generation of both poloidal and toroidal magnetic fields as a _ quasi - stationary kinetic dynamo effect_. in contrast to customary mhd treatments , this type of dynamo effect occurs _ in the absence of possible instabilities or turbulence phenomena_. in particular , in the case of tps , the self generation of toroidal field could take place even _ without any net accretion _ in the domain of interest , in presence of open magnetic field lines . this phenomenon is analogous to that treated in paper i for closed - field configurations . in particular , the toroidal field is associated with the existence of torques which cause redistribution of angular momentum , producing radial inflows and outflows of disk material . as a consequence , various scenarios can be envisaged in which stationary radial flows and kinetic dynamos are present in ad plasmas , both affected by processes of type # 1-#4 . let us now consider specifically the application of the kinetic solution developed here to the investigation of the accretion process in ad plasmas . the inward accretion flow in ads is usually `` slow '' in comparison with the characteristic larmor time @xmath341 . for example , ad plasmas with @xmath342 have hydrogen - ion larmor rotation times in the range @xmath343 which is shorter than the dynamical timescale at most relevant radii . for typical plasma densities and temperatures in the range @xmath344 and @xmath345 , the ( spitzer ) ion collision time ( below which the plasma can be considered collisionless ) is in the range @xmath346 ( the upper value corresponding to high temperature and low density ) . independent of the physical origin of the accretion process , we can therefore expect that the present theory correctly describes phenomena occurring on all time - scales in the range @xmath347 . we next determine the local poloidal and radial flow velocities for the various particle sub - species . by definition , these are given by@xmath348 s}}\left [ 1+g_{s}^{\prime } \right ] , \right . \\ & & \left . v_{rs}\equiv \mathbf{v}_{s}\cdot \mathbf{e}_{r}=\sum% \limits_{sub - species}\frac{1}{n_{s}^{tot}}j_{rs}^{l},\right . \\ & & \left . j_{rs}^{l}\equiv \int_{\gamma _ { u}^{l}}d^{3}v\left [ \mathbf{v}\cdot \mathbf{e}_{r}\right ] \widehat{f_{\ast s}}\left [ 1+g_{s}^{\prime } \right ] , \right.\end{aligned}\]]where @xmath349 and @xmath350 and the summations are performed over the particle sub - species for @xmath323 . we stress that the velocity - space integrals indicated above must contain the contributions from pps , bps and tps and so @xmath351 , where @xmath352 @xmath353 and @xmath354 are the corresponding mass currents . as an example , let us consider the leading - order contributions obtained ignoring flr corrections . explicit calculation gives@xmath355where @xmath356 and the functional dependence @xmath357 is prescribed by the kinetic constraints ( [ kinkin ] ) . we stress that the precise form of @xmath358 still has to be chosen to satisfy the solubility constraints imposed by ampere s law ( see the discussion in previous section ) and so the radial mass current density @xmath359 is generally a function of the form @xmath360 . we are interested in situations where there is a _ net radial accretion flow _ i.e. where the average radial mass current @xmath361 ( with @xmath362 and @xmath363 being suitably prescribed ) is negative . there are local contributions to @xmath364 from tps , bps and pps , but the overall accretion flow is mainly associated with pps . let us show that such a solution exists . we seek particular vlasov - maxwell equilibria which are globally quasi - neutral , in the sense of eq.([qneu ] ) . these equilibria are uniquely defined once @xmath365 , @xmath366 , @xmath367 and the structure functions are prescribed . the latter , by definition , are arbitrary smooth real functions of the specified variables as required by the kinetic constraints . notice that , if the quasi - neutrality condition is valid , an analytical solution for the es potential can be obtained as shown in paper i. furthermore , we consider an example in which by assumption 1 . @xmath368 is non - vanishing and dominant with respect to flr effects , so that eqs.([vp ] ) and ( [ vr ] ) apply . 2 . particular solutions have a definite parity property with respect to the spatial reflection @xmath369 as a specific case , the poloidal flux @xmath217 is assumed here to be antisymmetric , i.e. , @xmath370 , while both the toroidal and poloidal magnetic fields are symmetric . as a consequence , the toroidal current density must be antisymmetric . this can be realized only if a vertical electric field is present ( i.e. , one in the @xmath371 direction ) , consistent with the quasi - neutrality condition . if these assumptions are valid , ampere s law demands that , to leading order , i.e. , neglecting diamagnetic flr effects,@xmath372namely @xmath373 at this order of approximation . therefore , in this case a solution consistent with the requirement of net radial accretion flow and vlasov - maxwell equilibrium is obtained imposing that the species number density @xmath374 is even in @xmath217 , while the species structure function @xmath375 is odd with respect to the same variable . a solution of this type is consistent with the angular momentum conservation law ( [ angcons ] ) ; in order to obtain the solution , suitable kinetic boundary conditions must be prescribed ( see the discussion following eq.([dopo ] ) ) . this proves that stationary accretion solutions exist and are admitted by the present kinetic theory for the incoming qsa - kdf , namely for @xmath376 in the subset @xmath377 ( see fig.2 ) . the same conclusion is in principle applicable for outflows , by appropriate prescription of the outgoing qsa - kdf @xmath378 in the subset @xmath379 . in fact , the angular momentum conservation law ( [ angcons ] ) allows both inward and outward radial fluid velocities for each species , namely having @xmath380 or @xmath381 respectively . indeed , for a collisionless plasma the species tensor pressure is generally non isotropic ( see the related discussions in paper i and ref.@xcite ) such that eq.([angcons ] ) is identically satisfied . unlike the customary view based on ideal mhd , for which a self - consistent treatment of inflow and outflow solutions is usually difficult , within the present theory both inflows and outflows can occur independently and are described consistently by their respective qsa - kdfs . in particular , eq.([angcons ] ) shows that radial flows arise due both to the parallel velocities @xmath368 and to the kinetic effects carried by the flr diamagnetic corrections . as a result , species radial flow velocities appear necessarily in combination with non - isotropic tensor pressures and a non - vanishing toroidal magnetic field . in conclusion , the theory predicts the possibility of having purely inflowing matter in quasi - stationary ad plasmas , or of having co - existing inflows and outflows . in this paper , a consistent theoretical investigation of the slow kinetic dynamics of collisionless non - relativistic and axisymmetric ad plasmas has been presented . the formulation is based on a kinetic approach developed within the framework of the vlasov - maxwell description . we have considered here plasmas immersed in quasi - stationary magnetic fields characterized by open nested magnetic surfaces . this can be appropriate for radiatively inefficient accretion flows onto black holes , some of which are believed to be associated with a plasma of collisionless ions and electrons having different temperatures , and there can be other related applications to the inner regions of accretion flows onto magnetized neutron stars and white dwarfs . the discussion presented here provides a background for future investigations of instabilities and turbulence occurring in these plasmas . we have shown that a new type of asymptotic kinetic equilibria exists , which can be described by qsa - kdfs expressed in terms of generalized bi - maxwellian distributions . these solutions permit the consistent treatment of a number of physical properties characteristic of collisionless plasmas . the existence of these equilibrium solutions has been shown to be warranted by imposing suitable kinetic constraints for the structure functions entering the definition of the qsa - kdfs . in terms of these solutions , the slow dynamics of collisionless ad plasmas has been described by means of a suitable reduced gk - vlasov equation . in addition , the theory permits the consistent treatment of gravitational em particle trapping phenomena , allowing one to distinguish between different populations of charged particles . we have shown that the kinetic approach is suitable for the description of quasi - stationary ad plasmas subject to accretion flows and kinetic dynamo effects responsible for the self - generation of both poloidal and toroidal magnetic fields . four intrinsically - kinetic physical mechanisms have been included in the treatment of this , related to temperature anisotropy , parallel velocity perturbations and flr - diamagnetic effects . the novelty of the present approach , with respect to traditional fluid treatments , lies in the possibility of explicitly constructing asymptotic solutions for the fluid equations : the calculation of all of the relevant fluid fields involved ( e.g. the plasma charge and mass current densities and the radial flow velocity ) can be performed in a straightforward way using a species - dependent asymptotic expansion of the qsa - kdf . we believe that this study makes a relevant contribution for the description of two - temperature collisionless ad plasmas and the improvement of our understanding of their physical properties . the kinetic treatment developed here can also provide a convenient starting point for making a kinetic stability analysis of these plasmas . this work has been partly developed in the framework of miur ( italian ministry of university and research ) prin research programs and the consortium for magnetofluid dynamics , trieste , italy . c. cremaschini , j.c . miller and m. tessarotto , _ theory of quasi - stationary kinetic dynamos in magnetized accretion disks _ , proceedings iau symposium no . 274 , 2010 , _ advances in plasma astrophysics _ a. bonanno , e. de gouveia dal pino and a. kosovichev , cambrideg university press , in press . c. cremaschini , j.c . miller and m. tessarotto , _ kinetic closure conditions for quasi - stationary collisionless axisymmetric magnetoplasmas _ , proceedings iau symposium no . 274 , 2010 , _ advances in plasma astrophysics _ , ed . a. bonanno , e. de gouveia dal pino and a. kosovichev , cambrideg university press , in press . m. tessarotto , c. cremaschini , p. nicolini and a. beklemishev , proc . 25th rgd ( international symposium on rarefied gas dynamics , st . petersburg , russia , july 21 - 28 , 2006 ) , ed . ivanov and a.k . rebrov ( novosibirsk publ . house of the siberian branch of the russian academy of sciences ) , p.1001 ( 2007 ) , isbn / issn : 978 - 5 - 7692 - 0924 - 6 .

during the accretion phase of a core - collapse supernova ( sn ) , the deleptonization flux has recently been found to develop a global dipole pattern ( lesa lepton emission self - sustained asymmetry ) . the number flux is much larger than in one direction , whereas they are approximately equal , or even , in the opposite direction . we use a linearized stability analysis in a simplified sn model to study the impact of the flux asymmetry on self - induced neutrino flavor conversion . while a small lepton - number flux facilitates self - induced flavor conversion , `` multi - angle matter suppression '' is more effective . overall we find that for large matter densities which are relevant below the shock wave , self - induced flavor conversion remains suppressed in the lesa context and thus irrelevant for neutrino - driven explosion dynamics .

1412.0670

the neutrino and antineutrino flux spectra emitted by a core - collapse supernova ( sn ) significantly depend on flavor . therefore , flavor conversion can strongly modify what neutrinos do after decoupling , notably in driving the explosion , determining nucleosynthesis yields in the neutrino - driven wind , and the expected signal in large - scale detectors @xcite . even though neutrino mixing angles are large , in a dense medium the eigenstates of propagation and those of interaction are very nearly the same @xcite . therefore , significant flavor conversion would only occur by the msw effect @xcite at a large distance from the collapsed core . this situation can fundamentally change when taking neutrino - neutrino refraction into account @xcite . it can lead to self - induced flavor conversion @xcite even when the matter density is large and therefore the effective mixing angle is small @xcite . the propagation eigenmodes of the collective neutrino ensemble include run - away solutions in flavor space @xcite , leading to this phenomenon . one question is if self - induced flavor conversion would occur in regions below the stalled shock wave during the sn accretion phase . flavor conversion could then modify neutrino energy deposition and impact the explosion dynamics in the framework of the neutrino - driven mechanism of sn explosion . while self - induced flavor conversion can occur at much higher density than msw conversion , it is still suppressed by the `` multi - angle matter effect '' @xcite , although the exact conditions where conversions would occur require a linearized stability analysis or a numerical solution of the neutrino - flavor equations of motion . dedicated studies , using one - dimensional ( 1d ) sn models found that the `` onset radius '' of self - induced flavor conversion would always lie beyond the shock front where matter densities are much smaller @xcite . subsequent studies using other sn models found similar results @xcite . since that time , our theoretical understanding of both self - induced flavor conversion and of flavor - dependent sn neutrino emission have both evolved . we are here especially concerned with developments having to do with spontaneous breaking of symmetries that were previously taken for granted . axial symmetry of the neutrino radiation field around a given radial direction had been assumed in studies of collective flavor oscillations . however , a new class of run - away solutions breaks this symmetry spontaneously , allowing self - induced flavor conversion for both neutrino mass orderings @xcite . moreover , spatial uniformity can also be broken spontaneously , leading to small - scale instabilities @xcite . however , very recently it was shown that in the sn context , after including multi - angle matter effects , the uniform mode provides the most sensitive test for instability @xcite . therefore , in our present study we do not worry about small - scale instabilities when addressing the impact of an unusually small lepton - number flux . the other new finding is that sn neutrino emission in 3d models can develop a global dipole pattern termed lesa for lepton emission self - sustained asymmetry @xcite . while the overall neutrino luminosity remains nearly spherically symmetric , the deleptonization flux ( @xmath0 minus @xmath4 ) develops a strong dipole pattern within the first @xmath5150 ms after collapse irrespective of other hydrodynamical instabilities , notably even in the presence of the standing accretion shock instability ( sasi ) . more recent 3d models of the garching group @xcite as well as the oak ridge group @xcite show explosions . while all garching models exhibit the lesa effect very clearly and suggest that it persists at least until some 100 ms after the onset of the explosion @xcite , the 3d explosion model of oak ridge shows a less stationary dipole with an amplitude of just about 10% of the monopole over most of the pre - explosion evolution and a more lesa - like episode with a dipole - to - monopole ratio of @xmath525% and stable direction only in a time interval between 360 and 410 ms after bounce @xcite . the reason for these differences between the garching and oak ridge models is not clear yet , but they may be connected to different structural or transport conditions in the newly formed neutron star . the lesa phenomenon , when it is fully developed , implies that during the accretion phase , the deleptonization flux emerges primarily in one hemisphere . in some directions of the opposite hemisphere , it can be very small and sometimes even slightly negative . one may wonder if an unusually small lepton - number flux would facilitate self - induced flavor conversion , perhaps even below the shock front , and thus would require a fundamental re - examination of neutrino heating in directions of small lepton - number flux . for isotropic neutrinos or sn neutrinos in the single - angle approximation , self - induced flavor conversion is unavoidable in a @xmath0-@xmath4 symmetric ensemble @xcite . on the other hand , multi - angle effects caused by matter or even by neutrinos themselves can suppress the instability . the main purpose of our work is a parametric study of the impact of the @xmath0-@xmath4 flux asymmetry on self - induced flavor conversion . we use a linearized stability analysis , where we neglect several issues that have been discussed in the recent literature . we ignore spin - flip effects caused by neutrino magnetic dipole moments @xcite or simply by refraction in inhomogeneous or anisotropic media @xcite , and we ignore the role of anomalous neutrino - antineutrino pair correlations @xcite . realistically , all of these are probably small effects . it could be more questionable that we also ignore the `` halo flux '' produced by residual scattering beyond the neutrino sphere and especially its `` backward going '' component @xcite . moreover , we do not worry about the question if our assumed quasi - stationary neutrino source indeed produces a quasi - stationary solution . a true assessment of these latter assumptions requires a deeper conceptual development irrespective of the lesa context . we begin in sec . [ sec : flv - evol ] with setting up the linear stability analysis and clarifying our conventions and normalizations of crucial input parameters . in sec . [ sec : analysis ] we perform the stability analysis for a simplified setup and we discuss the role of the @xmath0-@xmath4 asymmetry . we conclude in sec . [ sec : conclusions ] with a summary and discussion of our findings . we set up our stability analysis along the lines of ref . @xcite , slightly adapting some conventions . the flavor - dependent sn neutrino fluxes depend , at a given radius @xmath6 , on energy @xmath7 . the neutrino direction of propagation in the transverse direction is described by the zenith angle @xmath8 relative to the radial direction and by the azimuth angle @xmath9 . we assume that the fluxes do not depend on coordinates in the transverse direction . the impact of small - scale variations in the transverse direction was recently studied in ref . @xcite with the result that , for a stability analysis , the largest - scale modes are most relevant . we use a very basic `` neutrino bulb model '' to describe the initial fluxes , i.e. , we assume that neutrinos are produced at a spherical surface with radius @xmath10 and emitted isotropically ( blackbody like ) into the outside half - space . it is convenient to describe the zenith direction with a fixed label @xmath11 , where @xmath12 is the zenith angle at radius @xmath10 . our blackbody like emission model corresponds to a uniform distribution on the interval @xmath13 . we stress that our reference radius @xmath10 is not a physical quantity but a quantity of mathematical convenience . it need not coincide with the neutrino sphere @xmath14 where typical neutrinos decouple and which is the physical emission region . neutrinos emerging at @xmath14 tend to show a forward - peaked zenith - angle distribution . the approximate width of this distribution provides us with the approximate radius @xmath10 of a hypothetical blackbody surface that would produce a flux with similar angular divergence . our main point of convenience is to use a @xmath15 distribution on the unit interval and we define @xmath10 accordingly . notice that we do not integrate the flavor evolution equations , but the philosophy of the stability analysis is to assume that nothing happens until our chosen test radius @xmath6 , i.e. , neutrinos have remained in flavor eigenstates which coincide with propagation eigenstates in a dense medium . therefore , the exact path taken before reaching the radius @xmath6 is irrelevant . we only need to know the flavor - dependent number fluxes and their angular divergence at radius @xmath6 . we combine the flavor - dependent fluxes , assumed to be stationary , into a matrix @xmath16 which depends on the variables denoted by subscripts . matrices in flavor space are written as capital sans - serif letters . the diagonal entries are the @xmath17 equivalent ordinary neutrino fluxes @xmath18 for species notice that @xmath18 is differential with regard to energy @xmath7 , i.e. , it is a flux spectrum , and it is also differential with regard to the zenith - angle variable @xmath15 and the azimuth angle @xmath9 . the off - diagonal elements of @xmath20 contain flavor coherence information . we use negative energies @xmath7 and negative flux values @xmath18 to denote antineutrinos . this `` flavor isospin convention '' considerably simplifies the analysis because we do not need to distinguish explicitly between neutrinos and antineutrinos in the equations . however , this notation can also be confusing . @xmath21 is a positive number for a positive @xmath7 , but a negative number for negative @xmath7 and then denotes the antineutrino flux in the sense that @xmath22 if we take @xmath23 to be the usual positive - valued @xmath4 flux . assuming that the solution does not depend on global direction , the radial evolution is given by @xcite @xmath24\ , , \label{eq : eom1}\ ] ] where we have suppressed the index @xmath6 on all quantities . the hamiltonian matrix governing the evolution is @xcite @xmath25 where @xmath26 is the neutrino mass - squared matrix , causing vacuum oscillations . the factor @xmath27 arises from projecting the propagation path onto the radial direction . this factor causes the multi - angle matter effect : neutrinos traveling in different directions accrue different phases along the radial direction . the matter term is given by the matrix @xmath28 of net charged - lepton densities which is diagonal in the weak - interaction basis . the third term represents neutrino - neutrino refraction which is given by the phase - space integral @xmath29 , where @xmath30 is understood as @xmath31 at radius @xmath6 . the radial velocity of a given mode is @xmath32 @xcite . the transverse velocity @xmath33 depends on the azimuth angle @xmath9 . one finds @xmath34 and @xmath35 @xcite . we restrict our stability analysis to a two - flavor system and express the flux matrices in the form @xmath36 where @xmath37 is real , @xmath38 complex , and @xmath39 . the fluxes with superscript @xmath10 are the ones at radius @xmath10 where @xmath40 and @xmath41 . since the overall neutrino flux is conserved , the trace term is conserved and drops out of the commutator equation . because the @xmath42 are @xmath17 equivalent fluxes , they do not decrease with distance , only the flavor content evolves . the flavor evolution is encoded in the normalized matrix @xmath43 . next we change the energy variable @xmath7 to the frequency variable @xmath44 which is more convenient in the flavor oscillation context . negative @xmath45 values denote antineutrino modes of the neutrino field . we imagine that self - induced flavor conversion is driven by the atmospheric mass difference . our sign convention is such that a positive @xmath46 refers to inverted mass ordering . actually we will always assume @xmath46 to be a positive parameter and will explain later on how to implement normal mass ordering in our equations . as far as flavor conversion is concerned , what enters is the two - flavor flux difference @xmath47 , not the individual flavor fluxes . therefore , it proves useful to introduce a dimensionless neutrino spectrum in the form @xmath48 where @xmath49 and all fluxes depend on @xmath45 , @xmath15 , and @xmath9 , and @xmath50 we recall that antineutrino fluxes in the flavor - isospin convention are negative and that typically the @xmath0 flux is larger than the @xmath51 flux , and similar for antineutrinos . therefore , in the sn context , @xmath49 is typically positive for positive @xmath45 ( neutrinos ) and negative for negative @xmath45 ( antineutrinos ) , although the opposite sign can appear for some range of @xmath45 , corresponding to a `` crossed over '' number flux between @xmath0 and @xmath51 . our normalization differs from our previous studies where we often used in the denominator , i.e. , the total number flux of @xmath4 minus that of @xmath52 . however , in the maximal lepton - flux direction of our 3d models , this quantity becomes very small or even vanishes and thus can not be used to normalize other fluxes . moreover , our new definition is more symmetric and more intuitive . with the spectrum thus normalized , we define the asymmetry parameter as @xmath53 if we denote with @xmath54 the total positive number flux for neutrinos or antineutrinos @xmath19 , the asymmetry parameter corresponds to @xmath55 in the sn context we have approximately @xmath56 , further simplifying this expression . the normalization in our previous papers was @xmath57 in this notation . the main point of our study is to investigate how flavor stability depends on the asymmetry parameter @xmath58 . in the lesa context , we have a strong lepton - number flux dipole , whereas the @xmath0 plus @xmath4 and the @xmath51 plus @xmath52 fluxes remain nearly @xmath17 symmetric , i.e. , the denominator in our definition of @xmath58 is nearly independent of direction . we finally linearize the equations of motion in two ways . we assume @xmath59 as we want to investigate exponentially growing modes in @xmath38 , implying @xmath40 to linear order . furthermore , we use the large - distance approximation , @xmath60 , leading to @xcite , @xmath61 s \nonumber \\ & - & \mu \int d \gamma^\prime [ u + u^\prime - 2\sqrt{u u^\prime } \cos ( \varphi-\varphi^\prime)]g^\prime s^\prime \label{eq : lin } \,,\end{aligned}\ ] ] where @xmath38 depends on @xmath45 , @xmath15 and @xmath9 , whereas @xmath62 and @xmath63 depend on @xmath64 , @xmath65 and @xmath66 . we then seek eigenvalues @xmath67 for solutions of the form @xmath68 , where a positive imaginary part @xmath69 reflects unstable solutions . the effective multi - angle strength of the neutrino - neutrino and neutrino - matter interaction are described by the parameters [ eq : mu - lambda - def ] @xmath70 where @xmath71 is the net electron density ( @xmath72 minus @xmath73 ) . on the other hand , @xmath74 is the @xmath0 density at radius @xmath6 and @xmath75 is the @xmath4 number density . therefore , in the definition of @xmath76 , besides subtracting the @xmath51 and @xmath52 densities , we use the sum @xmath77 to describe the neutrino matter effect . our definition of @xmath76 differs somewhat from the previous literature , corresponding to our modified definition of @xmath49 . what enters in the equations is the combination @xmath78 , which is independent of chosen normalizations of either quantity . in the lesa context , the total number fluxes @xmath79 and @xmath80 are almost @xmath17 symmetric . therefore , our new normalization of @xmath76 implies that it is nearly @xmath17 symmetric in our models . notice also that @xmath76 explicitly shows the usual @xmath81 variation with distance . therefore , we will write @xmath82 in our explicit analysis we will use @xmath83 and @xmath84 km , whereas the physical neutrino sphere is approximately at @xmath85 km . either @xmath76 or @xmath6 can be used as a measure of distance from the sn core . we assume axially symmetric neutrino emission , implying @xmath86 , but not necessarily an axially symmetric solution . in the lesa context , the local neutrino radiation field at some distance @xmath6 and some specific direction is not axially symmetric . however , if we assume lesa to represent an exact dipole , the system shows global axial symmetry relative to the dipole direction . our schematic neutrino distributions always refer to the direction of minimal or maximal lepton - number emission , i.e. , to the positive or negative lesa dipole directions . in these extreme lepton - asymmetry directions the fluxes can be assumed to be locally axially symmetric and in this sense axial symmetry pertains to the emitted fluxes . we do not think that our overall conclusions are strongly affected by this simplification . as derived in ref . @xcite , in the axially symmetric case one finds the eigenvalue equations @xmath87 where @xmath88 these eigenvalue equations are to be solved to find the eigenfrequencies @xmath89 and the range of parameters where these have an imaginary part , signifying instabilities with regard to self - induced flavor conversion . notice that the first block in eq . ( [ eq : stabilityconidtion ] ) provides solutions for the axially symmetric case , the so - called bimodal instability ( for inverted mass ordering ) and the multi - zenith - angle ( mza ) instability ( for normal mass ordering ) . the second block provides the instability where axial symmetry is spontaneously broken , the multi - azimuth - angle ( maa ) instability which appears only for normal mass ordering . our equations are written for inverted neutrino mass ordering ( io ) , whereas normal ordering ( no ) corresponds to a negative @xmath90 . however , we prefer to keep @xmath46 a positive parameter so that negative @xmath91 continues to denote antineutrinos . we can account for the mass ordering by introducing a parameter @xmath92 in front of the vacuum oscillation term @xmath93 in the hamiltonian matrix eq . ( [ eq : hamiltonian ] ) . this parameter carries through and finally appears in the first line of eq . ( [ eq : linearizedeom ] ) in front of the term @xmath45 . we can then , for no , multiply this equation with @xmath94 , restoring the original @xmath45 term , changing the sign of @xmath76 and @xmath95 , and the l.h.s . of the equation . for flavor oscillations , it is irrelevant if neutrinos oscillate `` clock - wise '' or `` counter - clockwise '' in flavor space , i.e. , the sign change on the l.h.s . is not important to find the eigenvalues . in other words , normal mass ordering is covered by @xmath96 therefore , we should solve the eigenvalue equation for @xmath97 . the first quadrant , where both parameters are positive , corresponds to io , the third quadrant , where both parameters are negative , to no . we are mainly concerned with the role of the asymmetry parameter @xmath58 . of course , there are many ways one can modify the spectrum @xmath98 to achieve a different asymmetry @xmath58 . however , as a simple case we could imagine to switch the role of @xmath0 and @xmath4 , leaving all else unchanged . this modification corresponds to @xmath99 which implies @xmath100 . in other words , we consider a new spectrum @xmath101 . following the propagation of signs in the equations , one finds that this new system is equivalent to the original one with the sign change @xmath102 , whereas @xmath76 remains unchanged . more precisely , if instead of switching the sign of @xmath58 we switch the sign of @xmath95 , the eigenvalue will also change sign , i.e. , @xmath103 . whenever @xmath89 has an imaginary part , there exists also the complex conjugate solution , i.e. , an exponentially growing and an exponentially damped solution . therefore , it is irrelevant if we find @xmath89 or @xmath104 because the growth rate is the same in both cases . finally , then , for a given spectrum @xmath98 and studying the stability condition in the full parameter range @xmath97 , all four quadrants have a physical interpretation according to @xmath105 \lambda>0&\fbox{\mystrut no , $ \epsilon<0$}&\fbox{\mystrut io , $ \epsilon>0$}\\[1ex ] \lambda<0&\fbox{\mystrut no , $ \epsilon>0$}&\fbox{\mystrut io , $ \epsilon<0$}\\ \end{matrix}\ ] ] where the sign change in @xmath58 is understood in the spirit of eq . ( [ eq : eps - switch ] ) . as a next step we perform the stability analysis for a simplified spectrum . we are not concerned with an exact numerical result for a specific sn model , but rather we wish to understand the qualitative impact of a modified @xmath0-@xmath4 asymmetry when all else is kept fixed . therefore , we construct the simplest possible toy model which allows us to develop an understanding of this question . for the zenith - angle distribution we assume black - body like emission from the neutrino sphere , i.e. , we assume a uniform distribution of the @xmath15-variable on the interval @xmath106 , together with a suitable reference radius @xmath10 as explained earlier . for the relationship between test radius @xmath6 and effective neutrino interaction energy @xmath76 we use the relationship shown in and around eq . ( [ eq : mur ] ) . our choices are motivated by the 3d models that have led to the discovery of the lesa phenomenon @xcite . specifically , we use the model with @xmath107 progenitor mass as a benchmark for our study . during the accretion phase , this model shows large - scale convective overturn , but it does not develop the standing accretion shock instability ( sasi ) , in contrast to the models with larger progenitor masses . after about 150 ms post bounce ( p.b . ) , this and the other models develop a large - scale anisotropy of lepton - number emission . , @xmath4 and @xmath51 . the @xmath51 flux ( representing any of @xmath108 , @xmath109 , @xmath110 and @xmath111 ) is similar in both directions . the @xmath4 flux develops a large asymmetry at around 150 ms p.b . in the direction of minimal lepton number flux , the @xmath4 flux is as large as the @xmath0 flux , in the opposite direction it is as small as the @xmath51 flux . _ middle : _ asymmetry parameter @xmath58 of the lepton - number flux as defined in eq . ( [ eq : eps - definition ] ) . _ bottom : _ effective neutrino - neutrino interaction strength @xmath112 at the reference radius @xmath84 km , see eq . ( [ eq : mur]).,scaledwidth=40.0% ] we illustrate this behavior in fig . [ fig - epsilon ] where we show in the top panel the flavor - dependent neutrino number fluxes in two opposite directions , roughly corresponding to the directions of maximal and minimal lepton - number flux , respectively . the @xmath51 fluxes in the two opposite directions are similar . however , in the minimum direction , the @xmath4 flux is similar to the @xmath0 flux , corresponding to a small lepton asymmetry . in the maximum direction , the @xmath4 and @xmath51 fluxes are similar and much smaller than the @xmath0 flux . from these number fluxes we can derive the asymmetry parameter @xmath58 which we show as a function of time for our two extreme directions in the middle panel of fig . [ fig - epsilon ] . when lesa is fully developed , the @xmath0 and @xmath4 fluxes are approximately equal in the direction of minimum lepton - number flux so that @xmath113 . in the maximum lepton number flux direction , the @xmath114 and @xmath115 fluxes are almost exactly equal . in our formulation , this situation corresponds to @xmath116 . in this extreme case , self - induced flavor conversion is not possible because @xmath117 can not be swapped with @xmath118 if the latter is zero . therefore , a stability analysis for @xmath116 is moot , but we will consider a case with @xmath119 . as an intermediate benchmark case we consider a situation where @xmath118 is half of @xmath117 , corresponding to @xmath120 . another crucial parameter is the effective neutrino - neutrino interaction strength defined in eq . ( [ eq : mu ] ) . we show @xmath112 as a function of time for our model in the bottom panel of fig . [ fig - epsilon ] . a typical value is a few @xmath121 . we will specifically use the value @xmath83 which is typical for the period after 150 ms when the lesa dipole is fully developed . notice that the choice of @xmath112 is not relevant for the stability analysis per se , but only to establish a relationship between physical radius @xmath6 and corresponding @xmath76-value . from past experience we know that the exact neutrino energy distribution tends to be relatively insignificant . we have explicitly checked this point by using top - hat shaped , maxwell - boltzmann , and more general gamma distributions . the locus of the instability region in the @xmath76@xmath95plane is primarily determined by the average oscillation frequency @xmath122 . because we study the differential effect caused by modified @xmath58 values it is most transparent to use the simplest possible model which exhibits these effects . therefore , we describe all @xmath0 and @xmath4 by a single energy , i.e. , the system is described by a vacuum oscillation frequency @xmath123 and one for @xmath124 . moreover , we can always go to a rotating coordinate frame in flavor space such that effectively @xmath125 . therefore , our simplified model is described by a spectrum @xmath126 which does not depend on @xmath15 . here , @xmath127 where the first term is for neutrinos , the second one for antineutrinos . the abstract stability analysis does not depend on the numerical choice of @xmath128 in the sense that all other quantities of dimension `` frequency , '' i.e. , @xmath95 , @xmath76 and @xmath69 are expressed in units of @xmath128 . the chosen value only matters when translating the stability region in the parameter space of @xmath76 and @xmath95 to physical sn parameters . specifically we use @xmath129 mev as our single energy , corresponding to @xmath130 . other choices would slightly shift the sn density profile relative to the instability region in our final plots figs . [ fig : stab - io ] and [ fig : stab - no ] . the main practical reason for using a single - energy neutrino spectrum is that it allows a simple analytic integration of the integrals eq . ( [ eq : in - definition ] ) . avoiding a numerical integration considerably accelerates the numerical search for the eigenvalues . the imaginary part of @xmath89 , the growth rate @xmath69 , is always of order @xmath128 , whereas in our range of interest , @xmath76 and @xmath95 are up to @xmath131 in these units and the real part of @xmath89 can also take on such large values . therefore , finding the eigenvalues can be numerically challenging . we provide the explicit eigenvalue equations for our simplified spectrum in appendix [ sec : explicit ] . for general values of @xmath76 , @xmath95 and @xmath58 , the eigenvalue @xmath89 and its imaginary part must be found numerically . the structure of typical unstable solutions is best explained using a simple example with @xmath132 , corresponding to twice as many @xmath0 than @xmath4 after subtracting the @xmath51 and @xmath52 fluxes . we first consider the first block in eq . ( [ eq : stabilityconidtion ] ) which derives for those solutions which preserve axial symmetry around a given radial direction . in fig . [ fig : contour1 ] we show the growth rate @xmath69 for unstable solutions in the parameter space consisting of the neutrino - neutrino interaction strength @xmath76 and the multi - angle matter effect @xmath95 as defined in eq . ( [ eq : mu - lambda - def ] ) . the upper right quadrant ( @xmath133 ) of the right panel corresponds to io and shows the traditional bimodal instability region . for a given neutrino density @xmath76 , the system is stable if the matter density is either too small or too large . the same effect can be described as a shift of the unstable @xmath76 range to larger @xmath76 values for increasing @xmath95 . this effect has been termed `` multi - angle matter suppression '' of the instability , although the instability is not suppressed , it is shifted to larger neutrino densities . however , we can say that it is suppressed relative to a specific locus in the @xmath76-@xmath95 plane which corresponds to the sn density profile . in the lower left quadrant ( @xmath134 ) of the right panel , corresponding with our conventions to no , there is no instability . physically , of course , @xmath76 and @xmath95 are positive changing their sign allows us to show the no case in the same plot as explained earlier . when we include axial symmetry breaking a new set of solutions arises , deriving from the second block in eq . ( [ eq : stabilityconidtion ] ) , and we find the solutions shown in the left panel . this is the multi - azimuth angle ( maa ) instability discussed in the previous literature . overall , then , in io we find the traditional bimodal instability , in no the more recent maa instability . the upper left and lower right quadrants ( @xmath76 and @xmath95 have opposite signs ) are unphysical in the sense that these quadrants require an excess @xmath4 flux . in the sn context , the collapsed material is lepton rich , implying that both the electron and neutrino densities are dominated by particles , not antiparticles , and thus that @xmath95 and @xmath76 have the same sign . however , in the lesa context it can happen that in the direction of minimal lepton - number flux the @xmath4 flux actually dominates . in this sense , a small negative @xmath58 is not entirely hypothetical , although a large negative value such as @xmath135 would be unrealistic . therefore , the upper left and lower right quadrants in fig . [ fig : contour1 ] are primarily shown for mathematical completeness . in the lower right quadrant , corresponding to io with @xmath135 , the instability range is rather insensitive to @xmath95 . in the upper left quadrant , corresponding to no with @xmath135 , we find a total of three instabilities . from a mathematical perspective , @xmath136 and @xmath76 is a more natural pair of parameters . notice that in the eigenvalue equations , the matter effect enters in the form @xmath136 , where @xmath137 represents the matter effect caused by neutrinos on each other . in the @xmath76-@xmath138 parameter space , the two mza instabilities in the right panel of fig . [ fig : contour1 ] appear in different quadrants ( cf . 9 of ref . in other words , the line @xmath139 , equivalent to @xmath140 , cuts between the two mza instabilities . as a second case we show in fig . [ fig : contour2 ] a similar contour plot for the fully symmetric case @xmath141 . as expected , the plot is now symmetric under the exchange @xmath102 . for @xmath141 and in the absence of matter ( on the line @xmath142 ) , the growth rate becomes arbitrarily large for @xmath143 . in this symmetric case without matter , there is no `` sleeping top regime , '' i.e. , the system is unstable for any @xmath76 above a very small threshold . notice that the eigenvalue equation can be solved analytically for @xmath144 and any value of @xmath58 as shown in appendix [ sec : explicit ] . we are here not particularly interested in the details of the growth rates of the various instabilities , but rather , if neutrinos streaming from a sn core along a given density profile will encounter any instability . in this sense it is enough to find the region in the @xmath76-@xmath95 parameter space where the system is unstable , a region which we call the `` footprint of the instability . '' this footprint is not quite uniquely defined . for example , in fig . [ fig : contour1 ] the region to the right of the bimodal instability is perfectly stable in the sense that @xmath145 , whereas in the region between mza and bimodal instability , @xmath69 is never exactly zero but only becomes extremely small . therefore , as a specific criteria we adopt somewhat arbitrarily @xmath146 as a stability condition . the natural dimension for @xmath69 is the vacuum oscillation frequency which in our case is of order @xmath147 . therefore , our instability criteria corresponds to one @xmath148-folding of growth on a radial distance of around 100 km . choosing a smaller @xmath69 criteria would imply a larger radial distance for one @xmath148-folding of growth . notice also that the exact locus of the limiting @xmath69 contour depends somewhat on the exactly chosen neutrino energy distribution which here was taken to be monochromatic . also notice that the region nominally found unstable in this sense may not be unstable enough to lead to any appreciable growth in realistic situations . our somewhat arbitrary stability criteria is probably conservative in this sense . in any case , we here only study the differential effect of modifying the @xmath58 value so that our exact choice of stability criteria is not crucial for our discussion . in fig . [ fig : footsix ] we show the instability footprints of figs . [ fig : contour1 ] and [ fig : contour2 ] , essentially providing the same information as the contour plots , but now on a logarithmic scale encompassing the range of parameters relevant in the sn context . for inverted mass ordering and @xmath149 , the upper right panel ( bimodal instability ) applies . for large @xmath76 and large @xmath95 , the red instability range ( @xmath120 ) is a narrow band . as shown in appendix [ sec : explicit ] , the asymptotic behavior of the instability strip is @xmath150 except for a logarithmic correction . the curves delimiting the asymptotic footprint are provided explicitly appendix [ sec : explicit ] . for @xmath141 , the asymptotic instability region is below the large-@xmath58 case , but in the approximate range @xmath151 , reducing @xmath58 creates an instability region above the large-@xmath58 footprint . for the most part , however , reducing @xmath58 extends the unstable region to the lower right of the original footprint . we will see that this region is not of interest in the sn context . for normal mass ordering , the panels with @xmath134 apply , i.e. , the lower mza and maa panels . the mza footprint is covered by the maa one , i.e. , in practice the maa instability is the only one relevant in the sn context . we see in the lower left panel that the maa footprints are qualitatively similar to bimodal ones . here reducing @xmath58 has the clear effect of shifting the instability region to an area between the red footprint and the horizontal axis , which makes it `` less dangerous '' in the sn context . the findings of the previous section become more explicit if we focus on the physical panels of fig . [ fig : footsix ] , i.e. , the upper right and the lower left , and show the same information where @xmath76 and @xmath95 are both plotted as positive variables and moreover , where we transform the @xmath76 parameter to radial distance in a sn model . in figs . [ fig : stab - io ] and [ fig : stab - no ] we show the bimodal and maa instability regions and also show a representative sn density profile inspired by our numerical lesa model . besides the footprints already shown in fig . [ fig : footsix ] , we now also include the case @xmath119 near to the largest possible value of @xmath152 . as @xmath153 , the footprint becomes an ever more narrow sliver of parameters . near - maximal @xmath58 values are relevant in the maximum - lepton number flux direction . , now for the maa instability region , corresponding to the lower left panel of fig . [ fig : footsix ] . [ fig : stab - no ] ] , now for the maa instability region , corresponding to the lower left panel of fig . [ fig : footsix ] . [ fig : stab - no ] ] in agreement with the previous literature we find that the most `` dangerous '' region with regard to a possible instability is at the shock wave where the density drops by a large factor , corresponding to a much reduced multi - angle matter effect . coincidentally , in our shown example of a sn density profile , it never intersects the instability region , but of course small changes in chosen parameters or shifts in the density profile will lead to an unstable region just outside of the shock wave . for a similar result see , for example , fig . 2 of ref . @xcite . we also confirm previous findings that the maa instability region tends to be nearer to the sn density profile . of course , which of the two instabilities applies depends on the neutrino mass ordering chosen by nature . in the maa case , reducing @xmath58 has the effect of shifting the unstable region away from the sn density profile . in the bimodal case , the effect is not monotonic , but typically the @xmath141 case tends to be `` most dangerous '' in the shock - wave region and comparable to the large-@xmath58 case of the maa case . motivated by the lesa effect , we have studied a simplified example of the impact of an unusually small lepton - number flux on the interplay between self - induced neutrino flavor conversion and its multi - angle matter suppression . the instability regions are strongly enlarged into regions of small matter density and large neutrino density as perhaps expected . however , the stability plots in figs . [ fig : stab - io ] and [ fig : stab - no ] show that for the most part this is a region which does not correspond to a realistic sn density profile . lowering the lepton - number flux by itself does not have the effect of leading to self - induced flavor conversion in regions below the shock front . the main point of our study was to explore the direction of the effect of lowering the lepton - number flux . according to our simplified models even strong modifications of the asymmetry parameter do not have a dramatic impact on the stability question along realistic sn density profiles . for io , where the bimodal instability applies , a small @xmath58 value can have the effect of creating an unstable region beyond the shock - wave radius . for no , where the maa instability applies , the largest @xmath58 values provide the `` most dangerous '' instability regions . overall it is primarily the multi - angle matter effect which controls the instability region for densities relevant along the sn profile . for practical sn simulations with regard to the explosion mechanism , it appears that unusually small @xmath58 values do not require a reconsideration of neutrino flavor evolution for distances below the shock front . of course , flavor conversion will be important beyond of the shock front , certainly at large distances by the msw effect , and perhaps in some cases directly outside of the shock front by self - induced flavor conversion . such effects would be important for the interpretation of the neutrino signal from the next nearby sn , but would not directly affect the explosion dynamics . our stability analysis was somewhat schematic with the purpose of exploring the general direction of what an unusually small lepton - number flux caused by the lesa effect would do . in principle , of course , in any given sn simulation one could perform an _ a posteriori _ stability analysis , based on the numerical neutrino radiation field and density profile , to verify that it was justified to ignore flavor conversion . in 3d simulations , the local neutrino radiation field will not be axially symmetric , not even in the lesa dipole directions . therefore , the stability analysis has to be based on neutrino fluxes with a non - trivial azimuth - angle distribution . this is not conceptually difficult , but such cases have not yet been explored in practice and await systematic study . in principle , one could eventually develop a numerical tool that could flag possible flavor instabilities as the numerical simulation progresses . of course , should cases be found where flavor conversion below the shock front occurs after all , a completely new challenge has to be faced in sn modeling . we stress , however , that not finding flavor - conversion instabilities in such an approach does not necessarily prove that none exist . as explained in the introduction , we have ignored a number of new issues that have emerged in the recent literature on neutrino - neutrino refractive effects . these topics should be sorted out before worrying further about self - induced flavor conversion in practical sn simulations . in particular , the multi - angle matter effect which provides the stabilizing ingredient in the present context relies on the assumption that the neutrino radiation field and its flavor properties vary only along the radial direction . this is an imposed symmetry assumption which can hide unstable solutions that might exist otherwise . likewise , the assumption of a purely stationary solution , allowing us to treat the problem in the form of an ordinary ( rather than partial ) differential equation has never been strictly justified . the core - collapse sn explosion mechanism is one of the few physical phenomena where neutrinos play a dominant dynamical role and where the flavor dependence of the fluxes matters . yet flavor conversion , in spite of large mixing angles , does not seem to figure at all for sn dynamics due to matter suppression of flavor oscillations , i.e. , because in dense matter , propagation and interaction eigenstates are almost the same . however , a final verdict on the role of active - active flavor conversion for sn dynamics requires more theoretical work to fully appreciate the role of possible flavor instabilities in the interacting neutrino field . this work was partly supported by the deutsche forschungsgemeinschaft under grant no . exc-153 ( cluster of excellence `` origin and structure of the universe '' ) and by the european union under grant no . pitn - ga-2011 - 289442 ( fp7 initial training network `` invisibles '' ) . s.c . acknowledges support from the european union through a marie curie fellowship , grant no.piif-ga-2011-299861 . . has been supported by the australian research council through a discovery early career researcher award ( grant no . de150101145 ) . in order to solve the eigenvalue equation we need the integrals defined in eq . ( [ eq : in - definition ] ) . in our simplified model , they take on the form @xmath154 where @xmath136 and @xmath155 and @xmath156 . we find explicitly as a next step we adopt the monochromatic spectrum eq . ( [ eq : monospectrum ] ) and express all frequencies in units of @xmath128 , i.e. , @xmath76 is understood to mean @xmath158 and so forth . with @xmath136 we find explicitly @xmath159\,,\\[1ex ] \kern-1em i_1&=&\frac{\mu}{\bar\lambda^2}\left[\epsilon\bar\lambda+ \left(1+\frac{\epsilon}{2}\right)(\omega-1)\log\left(1-\frac{\bar\lambda}{\omega-1}\right ) -\left(1-\frac{\epsilon}{2}\right)(\omega+1)\log\left(1-\frac{\bar\lambda}{\omega+1}\right)\right]\,,\\[1ex ] \kern-1em i_2&=&\frac{\mu}{\bar\lambda^3}\left[\bar\lambda\,\left(\frac{\epsilon\bar\lambda}{2}+\epsilon\omega-2\right)+ \left(1+\frac{\epsilon}{2}\right)(\omega-1)^2\log\left(1-\frac{\bar\lambda}{\omega-1}\right ) -\left(1-\frac{\epsilon}{2}\right)(\omega+1)^2\log\left(1-\frac{\bar\lambda}{\omega+1}\right)\right]\,.\end{aligned}\ ] ] @xmath162 \kern-3em \omega_{\rm mza}&=&+\frac{2\sqrt3 - 3}{12}\,\epsilon\mu \nonumber\\ & \pm&\sqrt{1+\frac{2\sqrt3 - 3}{3}\,\mu+\frac{7 - 4\sqrt3}{48}\left(\epsilon\mu\right)^2 } \,,\\[1ex ] \kern-3em \omega_{\rm maa}&=&\frac{\epsilon\mu}{4}\pm \sqrt{1+\mu+\left(\frac{\epsilon\mu}{4}\right)^2 } \,.\end{aligned}\ ] ] for these solutions to have a nonvanishing imaginary part , the bimodal solution requires @xmath163 , whereas the two others require @xmath164 . for all three cases , one finds that the maximum growth rate to be @xmath165 which of course only applies on the locus @xmath144 , i.e. , on the line @xmath140 . in other regions of the @xmath76-@xmath95 parameter space , @xmath166 could be larger . notice that in our convention , @xmath167 so that @xmath166 is indeed real and positive . for small @xmath58 , we have latexmath:[$\kappa_{\rm max}\approx the situation changes in the absence of any @xmath0-@xmath4 asymmetry where @xmath141 . in this case the imaginary parts of all three solutions grow without limit for @xmath169 . for @xmath170 , the edge of the plotting region in fig . [ fig : contour2 ] , and for @xmath142 we find @xmath171 , @xmath172 , and @xmath173 , in agreement with the contours in fig . [ fig : contour2 ] . in the opposite limit of very large effective matter density @xmath174 we may also gain considerable insight by analytic techniques . to this end we observe that unstable solutions , for large @xmath95 values , also require large @xmath76 values . therefore , to expand in powers of @xmath175 , we express @xmath176 , where @xmath177 is a dimensionless parameter . of course , at this stage this is only an ansatz because @xmath76 could also scale with another power of @xmath138 . considering first @xmath178 we assume that the real part of @xmath89 does not become large , i.e. , that @xmath179 can be considered small relative to @xmath138 or @xmath76 . under these conditions , to lowest order in @xmath175 the first block of the eigenvalue equation ( [ eq : stabilityconidtion ] ) becomes @xmath180 where @xmath181 the second block , yielding the maa instability , provides to lowest order @xmath182 therefore , to lowest order in @xmath175 , any solution , stable or unstable , requires @xmath183 . therefore , with @xmath136 we find @xmath184 this results corresponds to the very thin footprint in the upper left panel of fig . [ fig : footsix ] . next we turn to @xmath185 . we notice without further discussion that in this case the real part of @xmath89 becomes of order @xmath138 so that we write @xmath186 . performing now the expansion , we find for both blocks of eq . ( [ eq : stabilityconidtion ] ) a result of the form of eq . ( [ eq : logequation ] ) , but with modified expressions equation ( [ eq : logequation ] ) can be solved explicitly for our benchmark example @xmath120 where it is essentially a quadratic equation . in this case one finds @xmath188 this solution has a nonvanishing imaginary part for @xmath189 . however , for @xmath190 the imaginary part ( the growth rate ) is exponentially suppressed . the formal criteria @xmath191 leads to the requirement @xmath192 . therefore , the system is unstable for @xmath193 , where to make these results more explicit for the various instabilities , we first solve eq . ( [ eq : am1 ] ) , corresponding to @xmath195 . we find @xmath196 together with @xmath197 and with the explicit values for @xmath198 and @xmath199 we can find the limiting contours in the @xmath76-@xmath95-plane as a parametric plot depending on the variable @xmath138 . l. wolfenstein , `` neutrino oscillations in matter , '' phys . d * 17 * , 2369 ( 1978 ) . s. p. mikheev and a. yu . smirnov , `` resonance enhancement of oscillations in matter and solar neutrino spectroscopy , '' sov . j. nucl . phys . * 42 * , 913 ( 1985 ) [ yad . fiz . * 42 * , 1441 ( 1985 ) ] ; `` neutrino oscillations in a variable density medium and neutrino bursts due to the gravitational collapse of stars , '' sov . jetp * 64 * , 4 ( 1986 ) [ zh . eksp . * 91 * , 7 ( 1986 ) ] . a. s. dighe and a. yu . smirnov , `` identifying the neutrino mass spectrum from the neutrino burst from a supernova , '' phys . d * 62 * , 033007 ( 2000 ) [ hep - ph/9907423 ] . j. pantaleone , `` neutrino oscillations at high densities , '' phys . b * 287 * , 128 ( 1992 ) . v. a. kosteleck and s. samuel , `` neutrino oscillations in the early universe with an inverted neutrino mass hierarchy , '' phys . b * 318 * , 127 ( 1993 ) . s. samuel , `` bimodal coherence in dense selfinteracting neutrino gases , '' phys . d * 53 * , 5382 ( 1996 ) [ hep - ph/ 9604341 ] . h. duan , g. m. fuller and y .- z . qian , `` collective neutrino flavor transformation in supernovae , '' phys . d * 74 * , 123004 ( 2006 ) [ astro - ph/0511275 ] . h. duan , g. m. fuller , j. carlson and y .- z . qian , `` simulation of coherent non - linear neutrino flavor transformation in the supernova environment : correlated neutrino trajectories , '' phys . d * 74 * , 105014 ( 2006 ) [ astro - ph/0606616 ] . h. duan , g. m. fuller and y .- z . qian , `` collective neutrino oscillations , '' annu . sci . * 60 * , 569 ( 2010 ) [ arxiv:1001.2799 ] . a. banerjee , a. dighe and g. raffelt , `` linearized flavor - stability analysis of dense neutrino streams , '' phys . d * 84 * , 053013 ( 2011 ) [ arxiv:1107.2308 ] . a. esteban - pretel , a. mirizzi , s. pastor , r. toms , g. g. raffelt , p. d. serpico and g. sigl , `` role of dense matter in collective supernova neutrino transformations , '' phys . d * 78 * , 085012 ( 2008 ) [ arxiv:0807.0659 ] . s. chakraborty , t. fischer , a. mirizzi , n. saviano and r. toms , `` no collective neutrino flavor conversions during the supernova accretion phase , '' phys . lett . * 107 * , 151101 ( 2011 ) [ arxiv:1104.4031 ] . s. chakraborty , t. fischer , a. mirizzi , n. saviano and r. toms , `` analysis of matter suppression in collective neutrino oscillations during the supernova accretion phase , '' phys . rev . d * 84 * , 025002 ( 2011 ) [ arxiv:1105.1130 ] . b. dasgupta , e. p. oconnor and c. d. ott , `` the role of collective neutrino flavor oscillations in core - collapse supernova shock revival , '' phys . rev . d * 85 * , 065008 ( 2012 ) [ arxiv:1106.1167 ] . s. sarikas , g. g. raffelt , l. hdepohl and h .- janka , `` suppression of self - induced flavor conversion in the supernova accretion phase , '' phys . lett . * 108 * , 061101 ( 2012 ) [ arxiv:1109.3601 ] . g. raffelt , s. sarikas and d. de sousa seixas , `` axial symmetry breaking in self - induced flavor conversion of supernova neutrino fluxes , '' phys . lett . * 111 * , 091101 ( 2013 ) [ arxiv:1305.7140 ] . g. raffelt and d. de sousa seixas , `` neutrino flavor pendulum in both mass hierarchies , '' phys . d * 88 * , 045031 ( 2013 ) [ arxiv:1307.7625 ] . a. mirizzi , `` multi - azimuthal - angle effects in self - induced supernova neutrino flavor conversions without axial symmetry , '' phys . rev . d * 88 * , 073004 ( 2013 ) [ arxiv:1308.1402 ] . s. chakraborty and a. mirizzi , `` multi - azimuthal - angle instability for different supernova neutrino fluxes , '' phys . d * 90 * , 033004 ( 2014 ) [ arxiv:1308.5255 ] . r. s. hansen and s. hannestad , `` chaotic flavor evolution in an interacting neutrino gas , '' phys . d * 90 * , 025009 ( 2014 ) [ arxiv:1404.3833 ] . s. chakraborty , a. mirizzi , n. saviano and d. de sousa seixas , `` suppression of the multi - azimuthal - angle instability in dense neutrino gas during supernova accretion phase , '' phys . d * 89 * , 093001 ( 2014 ) [ arxiv:1402.1767 ] . g. mangano , a. mirizzi and n. saviano , `` damping the neutrino flavor pendulum by breaking homogeneity , '' phys . d * 89 * , 073017 ( 2014 ) [ arxiv:1403.1892 ] . h. duan and s. shalgar , `` flavor instabilities in the neutrino line model , '' phys . b * 747 * , 139 ( 2015 ) [ arxiv:1412.7097 ] . a. mirizzi , g. mangano and n. saviano , `` self - induced flavor instabilities of a dense neutrino stream in a two - dimensional model , '' phys . d * 92 * , 021702 ( 2015 ) [ arxiv:1503.03485 ] . a. mirizzi , `` breaking the symmetries of the bulb model in two - dimensional self - induced supernova neutrino flavor conversions , '' arxiv:1506.06805 . s. chakraborty , r. s. hansen , i. izaguirre and g. raffelt , `` self - induced flavor conversion of supernova neutrinos on small scales , '' arxiv:1507.07569 . i. tamborra , f. hanke , h .- janka , b. mller , g. g. raffelt and a. marek , `` self - sustained asymmetry of lepton - number emission : a new phenomenon during the supernova shock - accretion phase in three dimensions , '' astrophys . j. * 792 * , 96 ( 2014 ) [ arxiv:1402.5418 ] . i. tamborra , g. raffelt , f. hanke , h .- t . janka and b. mller , `` neutrino emission characteristics and detection opportunities based on three - dimensional supernova simulations , '' phys . d * 90 * , 045032 ( 2014 ) [ arxiv:1406.0006 ] . t. melson , h .- janka and a. marek , `` neutrino - driven supernova of a low - mass iron - core progenitor boosted by three - dimensional turbulent convection , '' astrophys . j. * 801 * , l24 ( 2015 ) [ arxiv:1501.01961 ] . t. melson , h .- janka , r. bollig , f. hanke , a. marek and b. mller , `` neutrino - driven explosion of a 20 solar - mass star in three dimensions enabled by strange - quark contributions to neutrino - nucleon scattering , '' astrophys . j. 808 , l42 ( 2015 ) [ arxiv:1504.07631 ] . e. j. lentz _ et al . _ , `` three - dimensional core - collapse supernova simulated using a @xmath202 progenitor , '' astrophys . j. * 807 * , l31 ( 2015 ) [ arxiv:1505.05110 ] . janka , `` three - dimensional core - collapse simulations by the garching group : the route to explosions , '' talk at f.o.e . fifty - one erg , raleigh , nc , 15 june 15 , + http://www.physics.ncsu.edu/foe2015 + /presentations / janka_raleigh-2015.pdf e. mller , h .- janka and a. wongwathanarat , `` parametrized 3d models of neutrino - driven supernova explosions : neutrino emission asymmetries and gravitational - wave signals , '' astron . astrophys . * 537 * , 63 ( 2012 ) [ arxiv:1106.6301 ] . s. hannestad , g. g. raffelt , g. sigl and y. y. y. wong , `` self - induced conversion in dense neutrino gases : pendulum in flavour space , '' phys . d * 74 * , 105010 ( 2006 ) ; erratum _ ibid . _ * 76 * , 029901 ( 2007 ) [ astro - ph/0608695 ] . h. duan , g. m. fuller , j. carlson and y .- z . qian , `` analysis of collective neutrino flavor transformation in supernovae , '' phys . d * 75 * , 125005 ( 2007 ) [ astro - ph/0703776 ] . g. g. raffelt and g. sigl , `` self - induced decoherence in dense neutrino gases , '' phys . d * 75 * , 083002 ( 2007 ) [ hep - ph/0701182 ] . a. esteban - pretel , s. pastor , r. toms , g. g. raffelt and g. sigl , `` decoherence in supernova neutrino transformations suppressed by eleptonization , '' phys . d * 76 * , 125018 ( 2007 ) [ arxiv:0706.2498 ] . c. giunti and a. studenikin , `` neutrino electromagnetic interactions : a window to new physics , '' rev . * 87 * ( 2015 ) 531 [ arxiv:1403.6344 ] . a. de gouvea and s. shalgar , `` effect of transition magnetic moments on collective supernova neutrino oscillations , '' jcap * 1210 * , 027 ( 2012 ) [ arxiv:1207.0516 ] . a. de gouvea and s. shalgar , `` transition magnetic moments and collective neutrino oscillations : three - flavor effects and detectability , '' jcap * 1304 * , 018 ( 2013 ) [ arxiv:1301.5637 ] . a. i. studenikin , `` neutrinos in electromagnetic fields and moving media , '' phys . atom . * 67 * ( 2004 ) 993 [ yad . fiz . * 67 * ( 2004 ) 1014 ] . a. vlasenko , g. m. fuller and v. cirigliano , `` neutrino quantum kinetics , '' phys . d * 89 * , 105004 ( 2014 ) [ arxiv:1309.2628 ] . v. cirigliano , g. m. fuller and a. vlasenko , `` a new spin on neutrino quantum kinetics , '' phys . b * 747 * , 27 ( 2015 ) [ arxiv:1406.5558 ] . a. vlasenko , g. m. fuller and v. cirigliano , `` prospects for neutrino - antineutrino transformation in astrophysical environments , '' arxiv:1406.6724 . c. volpe , d. vnnen and c. espinoza , `` extended evolution equations for neutrino propagation in astrophysical and cosmological environments , '' phys . d * 87 * , 113010 ( 2013 ) [ arxiv:1302.2374 ] . d. vnnen and c. volpe , `` linearizing neutrino evolution equations including neutrino - antineutrino pairing correlations , '' phys . d * 88 * , 065003 ( 2013 ) [ arxiv:1306.6372 ] . j. serreau and c. volpe , `` neutrino - antineutrino correlations in dense anisotropic media , '' phys . d * 90 * , 125040 ( 2014 ) [ arxiv:1409.3591 ] . a. kartavtsev , g. raffelt and h. vogel , `` neutrino propagation in media : flavor , helicity , and pair correlations , '' phys . d * 91 * , 125020 ( 2015 ) [ arxiv:1504.03230 ] . j. f. cherry , j. carlson , a. friedland , g. m. fuller and a. vlasenko , `` neutrino scattering and flavor transformation in supernovae , '' phys . lett . * 108 * , 261104 ( 2012 ) [ arxiv : 1203.1607 ] . j. f. cherry , j. carlson , a. friedland , g. m. fuller and a. vlasenko , `` halo modification of a supernova neutronization neutrino burst , '' phys . d * 87 * , 085037 ( 2013 ) [ arxiv:1302.1159 ] . s. sarikas , i. tamborra , g. raffelt , l. hdepohl and h .- janka , `` supernova neutrino halo and the suppression of self - induced flavor conversion , '' phys . d * 85 * , 113007 ( 2012 ) [ arxiv:1204.0971 ] . a. mirizzi and p. d. serpico , `` flavor stability analysis of dense supernova neutrinos with flavor - dependent angular distributions , '' phys . d * 86 * , 085010 ( 2012 ) [ arxiv:1208.0157 ] . a. marek and h .- t . janka , `` delayed neutrino - driven supernova explosions aided by the standing accretion - shock instability , '' astrophys . j. * 694 * , 664 ( 2009 ) [ arxiv:0708.3372 ] . b. mller , h .- janka and a. marek , `` a new multi - dimensional general relativistic neutrino hydrodynamics code of core - collapse supernovae iii . gravitational wave signals from supernova explosion models , '' astrophys . j. * 766 * , 43 ( 2013 ) [ arxiv:1210.6984 ] . janka , f. hanke , l. hdepohl , a. marek , b. mller and m. obergaulinger , `` core - collapse supernovae : reflections and directions , '' ptep * 2012 * , 01a309 ( 2012 ) [ arxiv:1211.1378 ] .

the high - cadence , comprehensive view of the solar corona by sdo / aia shows many events that are widely separated in space while occurring close together in time . in some cases , sets of coronal events are evidently causally related , while in many other instances indirect evidence can be found . we present case studies to highlight a variety of coupling processes involved in coronal events . we find that physical linkages between events do occur , but concur with earlier studies that these couplings appear to be crucial to understanding the initiation of major eruptive or explosive phenomena relatively infrequently . we note that the post - eruption reconfiguration time scale of the large - scale corona , estimated from the euv afterglow , is on average longer than the mean time between cmes , so that many cmes originate from a corona that is still adjusting from a previous event . we argue that the coronal field is intrinsically global : current systems build up over days to months , the relaxation after eruptions continues over many hours , and evolving connections easily span much of a hemisphere . this needs to be reflected in our modeling of the connections from the solar surface into the heliosphere to properly model the solar wind , its perturbations , and the generation and propagation of solar energetic particles . however , the large - scale field can not be constructed reliably by currently available observational resources . we assess the potential of high - quality observations from beyond earth s perspective and advanced global modeling to understand the couplings between coronal events in the context of cmes and solar energetic particle events .

1305.0801

the atmospheric imaging assembly ( aia , * ? ? ? * ) on the solar dynamics observatory(sdo , * ? ? ? * ) provides continuous full - disk observations of the solar chromosphere and corona in seven extreme - ultraviolet ( euv ) channels , spanning a temperature range from some 20,000k to in excess of 20mk @xcite . the 12-second cadence of the image stream with 4096 by 4096 pixel images at 0.6arcsec / pixel provides unprecedented views of the various phenomena that occur within the evolving solar outer atmosphere . with the full - disk coverage and the multitude of events going on on the sun at any given time , it is not surprising that many events are seen to occur synchronously or near - synchronously . in many cases , this ( near-)synchronicity is a matter of chance . during the 976-day time interval from the start of the sdo prime mission ( 2010/05/01 ) through the last of the searched dates for this study ( 2012/12/31 ) the noaa / goes logs contained 2881 flares of class c1 or above , or on average 3.0 flares per day . the study by @xcite counts typically some 4 coronal mass ejections for an average day . for a causal link via an alfvnic signal , we can take a typical time delay @xmath0 for a signal to travel a distance @xmath1 around the solar circumference with a characteristic coronal alfvn speed @xmath2 . for events separated by , e.g. , 90@xmath3 , @xmath4 and with @xmath5km / s , we find that @xmath6s , or about 1/40-th of a day . with 3 flares and 2 cmes a day on the earth - facing hemisphere , each of which can last for up to multiple hours , it is not surprise that many events in the sdo / aia data are seen to overlap in time ( quantitative estimates are provided in section [ sec : discussion ] ) , overlap or proximity in time is not an adequate distinguishing criterion for inferring causal links between events on the sun . in order to assess the importance of causal linkages in the triggering of near - synchronous events , we first must identify and classify the types of pathways that may connect them . to that end , we have reviewed many near - synchronous events in sdo / aia observations , and here present a selection of those to illustrate the causal linkages . with the available present - day observations , we can see that many of the events observed with sdo / aia either reveal direct magnetic connections between near - synchronously flaring or erupting regions , while others are highly suggestive of it . the idea of causal linkage between flaring in different regions goes back to @xcite . @xcite , following up on his initial report , noted that `` the formal statistical results as well as the visual impression conveyed from inspection of the [ caiik ] photographs suggests that some form of coupling may exist between widely separated spot groups '' , but with the data at hand concluded that `` the question must still be regarded as open . '' since then , multiple studies were published on the possible linkages not only between flares but also between flares and filament activations and even between sets of filament activations ; @xcite , who list many of these early studies , differentiate the linkage between relatively distant regions from another form of linkage , namely that of multiple flares occurring sequentially in the same region , which includes the subset of homologous flares ( with a much - larger scale equivalent in repeating pseudo - streamer blowouts , e.g. , * ? ? ? * ) . among the possible scenarios for linkage between solar impulsive events that were proposed already early on were fast - moving energetic particles and also shock waves traveling at up to 2000km / s . evidence for the existence and possible role of these was found at radio and microwave wavelengths ( e.g. , * ? ? ? * ; * ? ? ? thermal energy input conducted via direct field connections was also proposed as a possible causal agent in sympathetic activity ( e.g. , * ? ? ? larger data bases , some supported by space - based observations of coronal connections , led to statistical studies in the context of connection patterns in the coronal magnetic field . @xcite and , for example , following similar studies and conclusions by others cited in their work , used those statistical methods on samples of events to conclude that sympathetic flaring was at most a weak phenomenon , although apparently significant for regions in close proximity of each other ( which those studies found to be closer than @xmath7 and @xmath8 , respectively ) . suggestions of causal linkage between events are not limited to the flare - flare or flare - filament events , but also include couplings of filament eruptions . for example , @xcite discuss the possibility of two quiet - sun filament eruptions occurring in the wake of an active - region eruption , with the coupling agent being the field deformation by an eruption ( often with a signature `` coronal dimming '' ) that is instrumental in causing other field configurations that are either connected to it or that lie contained within it to lose their stability . another such pair of related quiet - sun filament eruptions , along with other flaring activity , was discussed by @xcite . a magnetic configuration of three adjacent flux ropes was subsequently modeled by @xcite , whose detailed mhd work illustrated how stretched field by one eruption can destabilize one or more other flux - rope configurations nested within a common overlying field . such linkages , particularly those involving fairly high field , are often not directly observable but can be inferred from models , as was done in the studies mentioned in this paragraph , and in the work by , for example , @xcite and @xcite . despite the many studies cited above and referenced within those , the phenomenon of sympathetic activity in the solar corona remains elusive . yet , in the context of forecasting solar activity , the influence of adjacent or distant regions on the loss of stability of a given region needs to be understood . this is particularly important for the development and propagation of cmes and the resulting particle events and geomagnetic storms : given the frequency of eruptive events on the sun , many cmes are composite events , but understanding their makeup from different events with either physical connections or chance coincidences is important both to the interpretation and the forecast of any heliospheric event under study . the complete coverage of the earth - facing hemisphere of the solar surface and corona by sdo , supported by far - side observations from the stereo spacecraft @xcite , complemented by the stereo and soho / lasco @xcite coronagraphs and full - sphere coronal field modeling ( such as work by * ? * ; * ? ? ? * that we use in the present study ) enable a comprehensive empirical assessment of the linkages that exist in the solar corona that may play a role in sympathetic activity . in view of the mixture of positive and negative findings in the literature of the significance of causal linkages between explosive or eruptive events on the sun , the aim of this study is not to quantify the frequency of sympathetic couplings , but to assess the evidence for , and to discuss examples of , any of the proposed causal pathways by which couplings may occur . in selecting the events discussed here , we reviewed much of the sdo / aia data for the period from 2010/05/01 through 2012/12/31 . based on the review of those observations , we discuss evidence for four fundamentally distinct plausible causal pathways : ( evolving ) direct magnetic connections , waves or propagating disturbances , distortions of and reconnection with overlying field by the eruption of one or more flux ropes elsewhere in the corona , and evolving indirect connections . one aim of this study is to present evidence for each of these pathways in at least one well - observed case . our other aim is to illustrate why it is proving difficult to assess the prevalence of sympathetic couplings in the causes of space weather and to point towards future opportunities to alleviate the difficulties in the study of this now nearly 80-year old problem . section [ sec : data ] describes the various data sets used in this study and the selection criteria for initial review of candidate data sets . section [ sec : casestudies ] reviews the most illuminating cases that support causal linkage of solar events or clearly illustrate a particular difficulty in establishing whether events are synchronous by chance or by interaction . in reviewing each of the events , we typically look at events over a full 24-h period and sometimes up to several days . after reviewing section [ sec : data ] the reader may choose to jump to the final two sections , where we discuss our findings in a summarizing section [ sec : conclusions ] and in the concluding section [ sec : discussion ] , before reviewing the detailed case studies in section [ sec : casestudies ] . to aid in the review of the supporting images and movies , we created an on - line table http://www.lmsal.com/forecast/styd.html and include direct links throughout the manuscript shown by a superscript `` s '' followed by a number . the primary data source for this study is the complete archive of sdo / aia observations . we selected candidate events using several different approaches . first , we created 30-minute summed images , downgraded to 64 by 64 pixels , in the 193 channel , using the on - line 2-minute cadence synoptic data set at 1024 by 1024 pixels from 2010/05/01 through 2012/12/31 ( http://jsoc.stanford.edu/data/aia/synoptic/ ) . we chose the 193 channel because it reveals both flaring activity and the lower - energetic phenomena of eruptions from quiet - sun regions . we then selected all events that exceeded a flare - like intensity and remapped their coordinates to an evolving synoptic map set , letting the signal fade over a 2-week period to allow patterns to stand out clearly . all pronounced clusters of events were subsequently reviewed in the daily summary movies of aia observations , in the process eliminating instrumental artifacts related to spacecraft rolls and off - points , data gaps , and calibration mode data . as a next pre - selection criterion , we reviewed all m- and x - class flares in the aia archive during the same period , and once again reviewed all daily summary movies for those dates . we also collected candidate events during daily reviews of aia data as annotated into the heliophysics events registry @xcite . from the sample of some five dozen candidate events , we selected the ten cases that were most compelling by visual inspection in supporting causal connections . for selected events , we also made and reviewed running - ratio movies in which the 211 , 193 , and 171 channels were combined . the frames in these image sequences were created by first computing logarithmic differences for time - averaged images ( for 264s averages of fixed - exposure frames taken at 24s cadence and 2minute offsets between successive frames to be differenced for these movies ) , and then combining these in sets of three into the rgb color planes of a movie ( clipping the scales at relative brightening or dimming to range from 0.1 to 10 ) . these `` tri - ratio''image sets readily reveal changes in intensity , be they in bright active - region settings or in quiet - sun or off - disk signals . the summary sdo / aia data for the selected events can be reviewed via the on - line table@xmath9 at http://www.lmsal.com/forecast/styd.html . for further review of the possible physical links between the selected data , we used potential - field source - surface ( pfss ) models using the magnetogram assimilation code developed by @xcite , updated to assimilate sdo / hmi data ( see @xcite for the instrument description ) , using the solarsoft pfss tool ( see http://www.lmsal.com / forecast / surfflux - model - v2/. links to visualizations of the field and an interactive viewing tool are included in the on - line table@xmath9 . in addition , we also compare the aia observations with a magnetofrictional ( mf ) model developed by @xcite , for which snapshots are included in the table . the mf model incorporates dynamics of the coronal field through the introduction of a dimensionless number that describes the balance between relaxation and diffusion and is the product of a friction coefficient and a diffusion coefficient @xcite . a second dimensionless parameter in the mf model is the ratio of the surface and coronal diffusion coefficients . these two dimensionless parameters were originally calibrated by @xcite ( a ) such that flux ropes form above the internal polarity - inversion line of each active - region bipole with roughly one turn of twist and roughly once per 27 days , and ( b ) such that the field lines at the top of the computational box remained radial . the models shown in this study use the same parametrization as in the original model . we note that the mf model run requires selection of the magnitude and sign of the helicity for each emerging bipole ; as these quantities are as a rule not known for active regions , these quantities are selected from a latitude - dependent distribution as described by @xcite . consequently , the mf model should be viewed as one possible incarnation of a dynamic corona , to be contrasted with the static pfss model , but subject to major uncertainties related to the choice of the above mentioned parameters . other comparison material included the soho / lasco data viewed using the jhelioviewer package ( http://www.jhelioviewer.org/ ) and the stereo browse data ( with links in our on - line table , and directly accessible at http://stereo-ssc.nascom.nasa.gov/browse ) . in reviewing and selecting the events discussed in section [ sec : casestudies ] we paid particular attention to the following possible mechanisms for sympathetic coupling : 1 ) ( evolving ) direct magnetic connection , 2 ) wave / front perturbation reaches distant region , and 3 ) distortion of overlying field by an eruption . we always allowed for mere synchronicity , and include discussion of such events in our case studies in the next section . while reviewing events , we generally tracked the duration of the glow of post - eruption arcades in the 171/193/211 - channel composite images ; these durations can be found in the case studies denoted as @xmath10 . this section describes the events on a selection of 11 dates in detail . these dates are listed in table [ tab : couplings ] , which also summarizes our conclusions regarding likely coupling mechanisms ( identifying the times and the regions involved in each such set of events ) . on first reading , it may be easier first proceed to section [ sec : conclusions ] in which we summarize the variety of processes before reviewing the following case studies in detail . an on - line table@xmath9 provides access to the observational and modeling materials used the case studies . the coronal events on 2010/08/01 have been extensively described by @xcite . following that study , the events in the solar corona and their impacts throughout the heliosphere have been studied in a series of papers that include the work by @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite - who describe the coronal field topology in detail - , and @xcite . @xcite pointed out that the emergence of several new bipolar regions on the far side involved field configurations that directly connected all of the main sites of activity that day , involving a large - scale quasi - separatrix , connecting over several null points above the sites of filament destabilizations , and reaching into a flaring active region and a far - side cme . we refer to that paper for a detailed description of the events . @xcite argued that the magnetic topology was such that all of the major coronal events were connected by a quasi - separatrix involving several coronal null points . in an mhd modeling study of a simplified configuration , @xcite demonstrated that the evolution of the common - envelope field over several filaments can cause flux ropes to erupt in the wake of a primary event . the evolution of the corona , visible in the aia tri - ratio movies @xmath11 as colored events , offers additional support for this process . in particular , the eruption of a large filament in the western hemisphere appears to induce large - scale reconnection that then allows a second filament eruption ( starting around 06:40 ) . the discoloration in the tri - ratio images , interpreted as a signature of coronal reconnection induced by the eruption , starts during the first , largest filament eruption . it is most evident from about 2010/08/01 7:45ut through the end of that day ( @xmath12h ) . [ fig : lcurve20100801 ] shows the aia light curves for 2010/08/01 , revealing ( a ) that the duration of the events differs considerably between the 94 , 131 , and 335 light curves , where there is one dominant peak with different decay time scales . moreover , the 171 , 193 , and 211 light curves have a more complex structure that does not lend themselves to the interpretation as a single event . hence , here ( as in other cases below ) , we can not rely on the disk - integrated light curves , but instead perform a visual inspection of the intensity and tri - ratio image sets to estimate the fading time scales in the emission patterns associated with the events discussed . the second main filament , lying to the south of and largely parallel to the first , erupts starting about 16:30ut with a noticeable rise speed and developing into a rapid eruption by 19:30ut . the delay of 9 to 12h between the initially observed thermal changes following the first filament eruption and the ultimate eruption of the second filament may appear rather long , but the thermal signatures of reconnection high in the corona in the wake of the preceding eruption of the largest filament after about 7ut continues until past the end of the day , so is clearly an ongoing process over many hours . for these events , there is model support for causal linkages , both through the evolution of a series of pfss models that illustrate the possible evolution of the field subject to flux emergence on the far hemisphere of the sun as well as in a simplified mhd model of parallel flux ropes sharing a joint overlying field configuration . there are direct field connections between the destabilizing configurations , seen in the aia images as in the field models . but despite all this indirect evidence of causal couplings , direct observational evidence is tenuous , leaving the ultimate interpretation rather subjective as to whether the overall field configuration is responsible for each event occurring with direct connections to the main topological feature , and as to whether the time of occurrence of the destabilizations is in fact affected by the evolution of the surrounding field . the period around february 14 and 15 , 2011 , shows considerable activity on the sun , with multiple flares and eruptions in the main region , ar11158 , as well as other activity seen on disk and towards the northeast limb . the main center of activity is a substantial active region in the southern hemisphere , just past central meridian , ar11158 . the region s x - class flare and associated cme are described in detail by , e.g. , @xcite and @xcite . here , we focus on february 14 and 15 , when at least 20 c - class events occurred , one m2.2 flare , and the first x - class flare of sunspot cycle 23 ( see the goes light curve in fig . [ fig : goes20110214 ] ) . we start our description with a high c - class flare around 2011/02/14 12:50ut . this event shows an eruption ( @xmath13 m within the active - region core , with coronal signatures elsewhere last at least for 80 m ) from the trailing region of ar11158 towards the trailing part of ar11159 to its north , associated with a coronal propagating front that is most noticeable to the north of ar11158 . this eruption occurs as a coronal reconfiguration is already in progress towards the north - east limb . the latter , although extending off the disk into the high corona , is not associated with substantial high - field activity in the soho / lasco - c2 field of view . there is no obvious field connection to that region in either the pfss or mf models on that day@xmath14 , but there is one in the model fields for a time five days later when more of the near - limb and over - the - limb field has been assimilated ; we return to its discussion below . around 2011/02/14 17:30ut , the same following part of ar11158 is involved in an m2.2 flare . this event is associated with a pronounced coronal propagating front , that is visible most prominently in all but the southerly directions , appearing to reach to the edges of the streamer belt towards the northeast , and out to the edges of three large topological domes converging on ar11159 ( fig . [ fig : topo20110215 ] ) ( see also the on - line topology movie@xmath15 ) . associated with this flare is a dark 304 arc or surge from the trailing part of ar11158 moving towards the east and south , occurring around 18ut . at the same time , a small filament destabilizes , once again on this day , towards the northwest around @xmath16 ( in arsceconds from disk center ) , in the same location as the field reconfiguration noted for the events following 12:50ut , and within 20minutes of the propagating front reaching a decayed active region just to the southwest of it near @xmath17 . then , around 18:30ut , there is another eruption from the same part of ar11158 , again associated with activity over the filament configuration to the ne . this happens once more just after 19ut ( associated with another dark 304 feature moving westward ) in events that are then followed by a high c - class flare around 19:30ut ( @xmath18 m within the region and about 2h in its immediate vicinity ) . in the trailing end of that event , another dark 304 feature is seen , forming threads that reach a long way towards ar11161 . the x2.2 flare occurs at 2011 - 02 - 15 starting at 01:45ut ( @xmath19 m within the region and about 90 m outside it ) . this flare is associated with a pronounced coronal propagating front and cme , studied in detail by @xcite . the filament region to the northeast near @xmath16 again shows limited synchronous activity . this event set appears coupled through the propagating disturbance , which may be either a pure wave or an expansion front that couples to the filament in the northeast through higher field breached by the initiating cme . in the decay phase of this main flare ( brightest within the central regions of ar11158 ) another eruption from the trailing segment or the active region occurs around 04:30ut ( @xmath20 m in the active region core , some 90 m in its far surroundings ) . this time , the 304 channel shows a moving , dark connection from ar11158 to ar11161 towards the northeast limb . this field deformation is a precursor to a mid - c class flare from that region around 05ut , that shows a similar , subsequent dark structure in 304 continuing its motion towards the east and south , now with synchronous brightening and deformations in ar11161 , around 05:40ut being highly suggestive of a direct connection between the region , but falling short of showing convincing evidence , even in high - contrast running - difference image sequences ( not shown ) . by 07ut , this pair of regions again shows synchronous mild flaring and surging , now involving also a ribbon - like brightening , at least in 304 , in the quiet - sun small filament towards the northeast limb near @xmath16 where repeated activity occurred as described above . later in the event sequence , ar11161 and the filament configuration around @xmath16 show pronounced synchronous activity once 2011/02/15 12:20ut , with some activity also at @xmath21 in the compact bipole north of ar11159 . in between the several near - synchronous events pointed out above , each of the regions involved exhibits activity ( both flares and eruptions ) in the absence of synchronous activity in the other regions , including , for example , the ultimate eruption of the filament towards the northeast limb after about 23:20ut . the general geometry of the corona on the earth - facing side of the sun@xmath22 appears to be described quite well by the pfss configuration shown in fig . [ fig : pfss20110215 ] ( showing the configuration one day later when ar11161 that emerged on the far side had been assimilated into the model ) . the magnetofrictional model for that day@xmath23 shows a connectivity similar to that of the pfss model , suggesting that there are no pronounced flux - rope configurations for 2011/02/14 . we return to this below , however , arguing that this may be because that formation is delayed in the mf model relative to reality . the model field and its topological summary in figs . [ fig : pfss20110215 ] and [ fig : topo20110215 ] show ( a ) a null point floating above about @xmath24 , and ( b ) a helmet structure that arches over ar11158 connecting to the surface northeastward of ar11161 . the repeated eruptions in the field connecting ar11158 and ar11159 arch towards the coronal null , and must somehow connect through it if it is to explain the synchronous activity in ar11161 . any reconnection occurring through the coronal null requires corresponding changes on both sides , thus providing a pathway to explain some of the sympathetic activity in ar11158 and the field of and near ar11161 in terms of coupled fields , and thus in terms of sympathy . eruptions that become cmes have to breach the field under the helmet configuration , which involves field reaching down to the ar11169 and its surroundings ( as can be seen in the left panel of fig . [ fig : topo20110215 ] ) . in view of the repeated near - synchronous activity and suggestions of connecting signals ( surges or waves ) , one might conclude that the dynamics of the evolving field forced by the frequent activity in ar11158 is likely not adequately captured by the approximations of a static ( pfss ) or slowly evolving ( mf ) field . we also note that the details of the high configuration of the field for these days is quite sensitive to changes in the solar surface field . the pfss field based on the revised assimilation procedure ( `` version 2 '' , see http://www.lmsal.com/forecast/surfflux-model-v2/ ) shown in this study and that computed for version 1 in @xcite show different configurations for the helmet streamer than those in fig . [ fig : topo20110215 ] . on the other hand , most of the solar surface for this date appears to lie under the warped , undulating helmet configuration , so invoking the disruption of that configuration by nascent cmes in causal linkages would not be readily falsifiable in this case . moreover , the many non - synchronous events in the main activity centers demonstrates that care needs to be taken in interpreting the synchronous ones as consequences of sympathetic coupling even if pathways in the magnetic field are identified . in the end , we propose that the events on 2011/02/14 - 15 are , in fact , causally connected . this is based on a view of the mf configuration for 2011/02/16 , when more near - limb field has been assimilated into the models . this configuration in shown in fig . [ fig : mf20110216 ] shown from the earth s perspective for 2011/02/14 . this field model clearly shows a rope connecting ar11158 to ar11161 , as it links through ar11159 , and it shows another high rope - like configuration ( shown in red ) from ar11161 to above the location from where the northernmost quiet - sun filament lifted off . this model connection is supported by the repeated activation of the filament towards the northeast limb closely following activity in ar11158 . the southern base of that rope connects to a region of major flaring in the trailing part of ar11158 , where another rope goes southeastward into the direction of the major surge around 2011/02/15 04ut . here , what appear to be ropes connect all of the regions of activity involved , and the footpoints of all ropes are directly adjacent from one connecting segment to the next , if not in fact interlinked . in conclusion , we find evidence of synchronous activity in directly connected regions where the connecting field is affected by flares or eruptions in at least one of these connected regions . the synchronicity of events in more distant regions , both in quiet - sun field and to active - region field , may be by chance , but the mf model for the configuration a few days after the above events ( when more of the field involved is on disk and assimilated into the models ) is highly suggestive of direct connections between all regions involved . the final week of september of 2011 showed considerable activity : along with a multitude of lesser events , there were 8 m flares on the 24th , 6 on the 25th , and 2 more on the 26th . here , we focus on the activity on 2011/09/25 and 26 when the main regions were well onto the disk and the overall magnetic patterns readily observable and models subject to less uncertainty . the 25th begins in the decay phase of an eruptive m1.0 flare from ar11303 in the southwest ( see fig . [ fig : pfss20110925 ] ) , associated with a cme as seen in soho / lasco . this is followed , around 01:15ut , by near - synchronous eruptions from the northwest limb and from the trailing side of ar11302 . these events occur on opposite sides of the helmet streamer and are unlikely to be physically connected ( cf . [ fig : topo20110925 ] ) . the first large flare , class m.4 . , starts in the goes data at 02:27ut in ar11302 ( @xmath20 m ) , in the northern hemisphere and east of central meridian . at about 02:35ut , the flare reaches its peak in the aia channels ( with a very pronounced diffraction pattern revealing a compact kernel ) . at that time ar11303 at the southwest limb is also brightening , reaching its peak brightness ( also with a pronounced diffraction pattern ) just minutes later . the diffraction patterns in these two events , indicative of compact flare kernels , brighten within about 70s of each other ; for a signal traveling between these two regions over the solar surface , this would require velocities in excess of 20,000km / s . the synchronicity is thus either associated with energetic particles , occurs simply by chance , or there is some other evolution that causes both of these events to go off simultaneously . we find no unambiguous evidence for such an intermediary event in either the observations or in the model fields , but do point out that there was a preceding cme from the direction of ar11303 visible in soho / lasco c2 from 00:36ut onward and some 10min . earlier in stereo / a cor1 ; that cme propagates in fairly narrow cone towards the southwest , with no obvious signatures over the northeast limb above ar11302 . the goes light curves ( fig . [ fig : goes20110925 ] ) show signatures of a double flare , being dominated by the signal from ar11302 . the flare from ar11303 is associated with a mild coronal propagating front traveling over some tens of degrees , and an eruption into the overlying high corona . the flare / eruption from ar11303 develops into a cme over the southwest limb . we note that for this pair of distant events , the field models in figs . [ fig : mf20110925 ] and [ fig : topo20110925 ] ( see also the on - line topology movie@xmath25 ) do not reveal connections between these regions ( with the pfss model putting ars11302 and11303 under disjoint domains of connectivity ) , nor do running - ratio movies reveal any wavelike perturbation connecting these events that stands out above the noise . at 2011/09/25 04:31ut an eruptive m7.4 flare initiates from ar11302 , with a pronounced coronal propagating front moving in southerly directions ( @xmath26h for the coronal region towards the southwest from the active region , still evolving when an earth transit starts at 06ut ) that develops into a cme from the sun in a southern to southeasterly direction . apart from a synchronous brightening of the core of ar11303 and compact , weak brightenings in quiet sun in the region leading ar11301 , there is no other obvious synchronous activity elsewhere . but ar11303 does brighten once more , starting around 05:20ut , around the time that the perturbation front from the eruption of ar11302 approaches it . there is a cme visible in soho / lasco c2 from 8ut onward , most strongly from the direction of ar11303 ; no aia observations of its origin are available owing to an earth transit from 06:05ut to 07:20ut . the m3.1 flare starting at 2011/09/25 08:46ut in ar11302 ( @xmath20 m ) is compact once more with no clear sign of any high coronal field being breached or distorted . shortly after 09ut an eruption begins from ar11303 , clearly opening field starting around 09:35ut , coincident with activity at the top of a polar - crown prominence in the southwest this eruption is associated with an m1.5 flare starting in the goes data at 09:25ut . around 09:30ut there is a compact brightening in the core of ar11302 . again , there is synchronicity between ars11302 and 11303 without model or observational support for a causal connection . at other times , however , such as the flare in ar11302 around 10:10ut , no synchronicity between events in these regions is observed . at 15:26ut a pronounced southward eruption ( and surge / spray ) is associated with an m3.7 from ar11302 ( @xmath18 m in the region , and until past the next flare 90 m later in its surroundings ) , associated with a wide - angle cme from the southeast to the north . this is followed by a compact brightening into an m2.2 flare starting at 16:51ut in the same region . around 2011/09/25 18:45ut , an eruption starts over the northeast limb over regions trailing ar11302 ( with a rather subjective estimate of @xmath27h ) . once that eruption accelerates around 19ut , a high loop configuration over the western equatorial limb , connecting regions near ar11303 and ar11295 , also begins to rise . around 21ut , high loops connecting ars11301 , 11302 , and 11304 expand ( @xmath28h ) , as the western - equatorial loop system continues to expand , followed by repeated mid - c level flaring in ar11303 with an eruption around 2011/09/26 02:20ut . field models do not suggest how any causal linkage between these events might occur , except for the direct linkage of ars11301 and 11302 . then , there is another compact brightening in ar11302 associated with an m4.0 flare starting at 2011/09/26 05:06ut ( @xmath29h with reconfiguration of the surroundings , but observations are cut short by an earth transit ) , and finally a moderately disruptive m2.6 starting at 2011/09/26 14:37ut ( @xmath30 m ) . both events are limited to activity within ar11302 . the overall topology of the coronal field , as summarized in fig . [ fig : topo20110925 ] , does not support connections between the ( near-)synchronous events in ars11302 , 11303 , and 11295 . these regions appear isolated by separatrices : ars11302 and 11295 lie on either side of the tilted ( yellow ) separatrix surface ( the `` helmet '' ) seen in the pfss model , while ar11303 lies under that helmet . hence , the eruption of one region is not connected to either of the others through topological structures . on the other hand , the reconfiguration of the loops seen between ars11302 and 11301 are directly mappable to the magnetic field , most clearly seen in the mf model in fig . [ fig : mf20110925 ] . these are not , however , associated with major flaring . we conclude that in this case , despite several instances of synchronicity , there is no compelling observational evidence for causal connections between the main events on 2011/09/25 and 26 except , perhaps , for the coupling through a field disturbance ( propagating front ) around 05:20ut on 2011/09/25 . around 07:30ut , a large quiet - sun filament configuration erupts in the southeastern quadrant ; the post - eruption arcade continues to gradually evolve throughout the remainder of the day . around 11:30ut , an eruption initiates behind the northwestern limb , with effects in the visible part of the corona fading out around 12:30ut . neither of these events appear to be linked , either in the aia observations or in the field models , with the major flaring that follows . at about 12:25ut an m1.1 flare begins ( flagged as 13:04 in goes data ) , initiated in ar11341 , associated with a large filament eruption from neighboring ar11342 that comes to full development around 13:00ut , resulting in a full - blown cme ( @xmath31h ) . the rising eruption distorts and disrupts overlying field , with signatures seen ( in , e.g. , the tri - ratio movie@xmath32 ) , for example , into the inter - regional separatrix in the trailing polarity of ar11339 to the west of the erupting region that are visible until past 14ut , with the post - eruption arcade itself glowing until at least 20ut . after the thermal signatures of the previous eruption over the surrounding quiet sun have faded away , by about 16:30ut , there are coronal deformations to the north of ar11339 just south of the polar coronal hole boundary . these appear to be caused by an eruption ( @xmath33h , poorly determined because of the slow evolution in the end ) , of which we see faint ribbon - like signatures northwest of ar11339 starting around 17:45ut , as well as high off - limb field deformation towards the north from there , with tilting field reaching from there to about 45 degrees clockwise from the north . soho / lasco and stereo / secchi show faint blowouts associated with this event . when the 94 channel is suggestive of high - temperature loops forming after that eruption , there is simultaneous formation of such hot loops within an evolving environment in the region adjacent to ar11342 towards the northwest , both lasting until at least 21ut . there are high coronal flux ropes connecting these features between 1.3 and 1.7@xmath34 in the magnetofrictional model ( fig . [ fig : mf20111109_ropes ] , in particular at 50@xmath3 north over ar11342 and surroundings ) . fig . [ fig : mf20111109_tip57 ] shows a tilted perspective of the magnetofrictional field model , revealing that the erupting ar11341 and the eruption north of ar11339 have connections that end adjacent to each other ( close to the central meridian at the center of the disk in this tilted perspective ) showing that there are likely indirect magnetic interactions between the evolution of these structures , but not demonstrating their involvement in sympathy . shortly after 2011/11/22 04ut , a c7 flare goes off in a compact emerging - flux region slightly to the south of the line connecting ars11355 and11356 ( see fig . [ fig : pfss20111122 ] ) . just as that event ends , a small activation and eruption ( and cme observed in both soho / lasco c2 and stereo / secchi - a cor1 ) occurs to the north of it ( @xmath35h ) , in the leading edge of ar11356 , expanding somewhat to the northeast and southwest through about 07ut . there are synchronous brightenings visible at the trailing end of the polar - crown configuration north of ar11353 , which itself shows a high coronal reconfiguration starting at about 07:25ut , when there are simultaneous brightenings south of ar11355 and in ar11357 far to its west . these events are associated with a cme first seen in soho / lasco c2 after 06:24ut . all of these regions lie under a large dome of connectivity in the pfss model , mostly near its periphery , and thus suggest involvement of field that connects through a coronal domain near a high coronal null ( see fig . [ fig : topo20111122 ] ) . around 10:30ut , the filament configuration connecting ar11353 in the northern hemisphere and ar11354 in the southern hemisphere activates ( @xmath36h ) , synchronous with brightenings and loop shifts from there to the northern polar crown , including the corona to the north of , and apparently high over , ar11353 . around 13:50ut , there are simultaneous brightenings reaching from ar11356 to near that same position in the northern polar crown configuration . here , [ fig : topo20111122 ] suggests the connections to involve two adjacent domains connected at a coronal null westward of ar11353 . at around 17:10ut there is another , more gradual mid - c class brightening in the emerging flux region southeast of ar11355 ( @xmath37 m ) , not obviously associated with any other substantial activity . then , just past 18ut , a filament - prominence polar - crown configuration to the northwest accelerates its rise , transitioning into a cme ( @xmath27h ) , with first signatures in soho / lasco c2 after 20:48ut . there is some activity near ar11356 , and the filament between ars11353 and11354 continues to evolve , but neither is obviously connected to the erupting filament - prominence to the northwest . around 2011/11/30 12ut a prominence over the east limb begins to rise . the prominence trails ar11362 which shows some brightening at that time , but without obvious goes signatures above the background variability ( fig . [ fig : goes20111130 ] ) . around about 19ut , the prominence rise accelerates and transitions into an eruption ( @xmath38h ) . on the other side of the sun , a filament in the southern reaches of ar11355 begins to rise around 19:50ut , erupting after 20:25ut , in association with a mid - c class flare . the latter eruption occurs even as the off - limb corona to the east is still deforming in association with the prominence eruption . around 21:40ut , the same filament configuration in ar11355 erupts once more ( @xmath20h ) , into a c - class event of very comparable peak strength , but more gradual in both onset and decay . neither the pfss model ( fig . [ fig : pfss20111130 ] ) nor the mf field@xmath39 show obvious connections between these two eruptions . the soho / lasco images show signatures of the two eruptions starting on the east limb around 19ut and from ar11355 in the west around 20:25ut entering the c2 field essentially simultaneously just after 21ut . the second eruption from ar11355 is a much brighter cme that becomes visible between 22:00ut and 22:36ut . in c3 images , the set of three eruptions looks like a single complex event , extending from east to west , with some brightness variations at all hour angles , giving this set of events the appearance of a single irregular halo cme . stereo a / b cor2 data reveal that the southern component of this apparent halo event as seen from earth perspective is , in fact , a slower moving cme that erupted from the far - side southern hemisphere ( in an eruption starting around 18ut ) even before the above eruptions started . we find no compelling evidence for a physical linkage between these three eruptions , although all three appear to occur from underneath the helmet streamer . these events are an excellent demonstration that what appears as a single cme in coronagraph images , in fact is a composite event from two main eruptions , overlapping with even other events . there is no obvious evidence for causal couplings between these two eruptions within the low solar corona , however . this case shows the eruption , starting around 5ut ( @xmath38h ) , of a quiet - sun filament ( confirmed by h@xmath40 observations by meudon ) with its enveloping rope configuration ( clearly seen in the mf field model@xmath41 ) . during the course of the eruptions it is revealed that the quiet - sun filament in fact connects to a trailing negative - polarity region some 60@xmath3 away . thus , it is a demonstration of a direct field connection some 60@xmath3 trailing the initial eruption site that is evident only from field modeling , but that is not recognizable as such in a single coronal or h@xmath40 image prior to the filament eruption . the destabilization , flare , and eruption of a region at the limb region ( reaching no higher than b9 ) was a direct consequence of the eruption of a filament starting almost a solar radius away . prior to that eruption , a high rope was seen to move out to about 1.2r@xmath42 over the north - polar region , but without erupting into the soho / lasco cor2 field of view ; its relationship relative to the eruption and weak limb flare remains unknown . the long quiet - sun filament configuration in the northeast quadrant destabilizes around 04ut . by 05:45ut , coronal brightenings are suggestive of eruption - driven reconnection ( with a mild goes increase at that time , but rising from about b7 to b8 ) . in the subsequent 20 minutes , ribbon - like features and dimming regions develop , most prominently on the trailing side , that are commonly associated with eruptions of flux ropes . by 06:15ut , there are brightenings some 10 degrees eastward of the trailing end of the filament , revealing more extended field involvement . the images from aia s 304 channel most clearly reveal a direct coupling with a distant trailing region just at the east limb ( fig . [ fig : taffy304 ] ) . the erupting filament is seen to have , or to develop , an extension that propagates towards the region at the limb . by 07ut , this filament extension is seen to distort the field in the region at the limb , pulling a segment of the field out of its initial configuration , stretching it high into the corona . around 08ut , the distorted configuration of the limb region itself erupts ( goes 0.5 - 4.0 rising to about b9 starting at 08:05ut ) , with the last obvious signatures of that eruption ending around 08:45ut as the eastern end of the erupting filament exits the aia field of view . we note that around 08:20ut , a small filament leading the main erupting filament by about 40 degrees initiates an eruption ( @xmath20h ) that fully develops around 08:45ut . stereo - a cor1 ( then leading earth by 107@xmath3 ) shows a cme first developing at about 06:45ut , followed by a second at 09:10ut . soho / lasco c2 images confirm that there are two distinct eruptions , so the timing of the eruption of the leading smaller filament around the end of the activity at the east limb appears not to be indicative of a single composite eruption . the bottom end of a u - shaped configuration detaches from the stereo - a cor1 occulting disk at about 03:45ut , continuing a very gradual rise through at least 07:45ut when it fades from the running difference images at about 1.2r@xmath42 above the solar surface . there is no obviously identifiable counterpart of this in soho / lasco data , likely because the structure faded , if not stalled , before reaching the c2 field of view . it is possible that this reconfiguration is part of the overall set of events , either as another consequence or as an element of the subsequent eruption . prior to the primary eruption of the quiet - sun filament , stereo - a / b cor1 images show a gradual , faint eruption of a rope - like configuration already in progress at very high northerly latitudes shortly after midnight . the pfss field shows two connected arcades : a long arcade over the erupting filament , and a side channel connecting to the region at the east limb . the mf field reveals a rope configuration that distributes its field between a bipolar area just south of the trailing end of the initially erupting filament ( that in later magnetograms is shown to have a small active region injected into it while on the far side ) and a negative - polarity region at the limb , that corresponds to the erupting limb region . the mf configuration for 2011/12/11 shows the low - lying flux rope ( at and below 1.1 r@xmath42 ) extending over 130 degrees , with a higher rope configuration ( around @xmath43r@xmath42 ) over a 70-degree span above the main filament eruption site ; this high configuration may be what the stereo a / b cor1 showed as a gradually rising rope , although then manifested at apparently higher latitudes . six days later , the high rope has mostly disappeared from the model , having evolved away while the low - lying rope has lost its leading segment around where the eruption occurred , both presumably in part because of newly emerged field on the far side , and possibly because of a loss of equilibrium in the overall configuration . in the eruption evolving around 8ut , the 304 image sequences ( see the on - line sdo / stereo 304 movie@xmath9 and , in particular , the running difference image set@xmath44 described in the caption to fig . [ fig : taffy304 ] ) are highly suggestive of a field connection that the field models suggested possible , but did not themselves directly contain . with these events occurring at the east limb , however , no model could have captured this well given the unavailability of up - to - date magnetic information for the far hemisphere of the sun as seen from earth . [ fig : lcurve20111225 ] early on 2011/12/25 , two quiet - sun filaments erupt from near central meridian ( leading ars11384 and11383 by about 30 degrees , cf . , [ fig : pfss20111225 ] ) in the northern hemisphere . the first , northernmost filament is in early eruption at 2011/12/24 23:45ut , with ribbons forming around 2011/12/25 00:15ut ( @xmath45h ) . the erupting filament rapidly rises , exiting the aia field of view by 01ut , leaving behind a post - eruption arcade that relaxes at least until after 06ut . the southernmost signal that we see in aia running - ratio movies@xmath46 extend southward of the next event : the second filament , just to the south of it , responds gradually starting from about 03ut , accelerating from about 05:30ut onward , with reconnection signatures starting around 07ut . around 07:50ut , another eruption begins , this time from a region just over the northwest limb . it forces its way through the coronal field starting from 08:25ut onward ( @xmath28h ) . starting at 08:45ut ar11385 flares ( @xmath47c7 ) and erupts , associated with a coronal propagating front , and once more ( @xmath47c9 ) around 11:15ut . in the goes light curves , all of the above events form a single curve with two c - spikes on it ( fig . [ fig : goes20111225 ] ) . at 18:11ut , an m4.0 starts from ar11385 , associated with a very pronounced , far - reaching coronal propagating front ( @xmath48h inside the region , and approximately 3 in its surroundings ) . as that front reaches the trailing ar11386 , just after 18:30ut , a relatively minor eruption occurs within it . connections between these regions were suggested earlier in the day when , in association with the 08:45ut flare in ar11385 , brightenings in ar11386 were observed , and again in events of different magnitude at 17:05ut and 18:15ut , interspersed with others in which no such synchronicity was observed . the pfss ( fig . [ fig : pfss20111225 ] ) and mf@xmath49 models do not reveal field connections or topological relationships that might connect the various eruptions of the on - disk filaments with the over - the - limb configuration , with ar11385 , or of ar11385 with ar11386 . there are several high ropes ( fragments ) in the magnetofrictional model for this date ( fig . [ fig : mf20111225_ropes ] ) : one between 1.25 and 1.6@xmath34 between ars11385 and 11386 , overlying a much lower one below 1.1@xmath34 , and above 1.6@xmath34 overlying the two erupting filaments . we note that whereas the goes light curve in fig . [ fig : goes20111225 ] suggests that the eruptions form a long - duration event in which energy conversion occurs over some 24h before the corona relaxes to its pre - events emission levels . aia s light curves in that figure show a similar behavior for the shortest wavelength channel at 94 , which has a strong contribution from fexviii lines . the channels responsive to lower coronal temperatures , specifically 171 , 193 , and 211 , show a strong peak developing from about 15ut onward , which has only a very weak counterpart in the goes light curves . these signals are dominated by post - eruption emission from an event that started behind the northwestern limb , of which the post - eruption loops are showing up over the limb in the aia observations ( @xmath50h ) . we propose that the eruptions of the on - disk filaments may be coupled as in the mechanism explored by @xcite , and that the propagating front emanating from the m4.0 flare in ar11385 was instrumental in upsetting the balance of the field in ar11386 or at least in accelerating the destabilization . we find no evidence for direct field couplings in the equilibrium field models , however , between the on - disk filament eruptions and the over - the - limb eruption , or between those and the subsequent mid - c class flaring in ar11386 . around 2012/02/09 07ut changes begin to occur in a quiet - sun region around @xmath51arcsec from disk center . these changes suggest an eruption , with intensity signatures coupling in a compact , unnumbered , emerging bipolar region in the southern hemisphere around @xmath52 . no obviously related additional activity occurs in that emerging region as a result of this , although some activity is ongoing in association with its emergence throughout the day . the thermal evolution of the erupted region in the north continues until at least 12ut ( @xmath20h ) . starting around 17ut , brightness and thermal changes are seen in that erupted region around @xmath51 ( @xmath38h ) , which continues as a filament / prominence due east at the limb begins to erupt , accelerating into a cme around 19ut ( visible in soho / lasco c2 after 20:48ut ) , together with a quiet - sun filament reconfiguration to the southeast of , and connected with , the bipolar region at @xmath52 ( @xmath31h ) . as this joint eruption continues , the regions in the vicinity of @xmath53 to @xmath54 exhibit a pronounced change , starting with an apparent cooling ( dominant 171 channel ) from 20ut onward , then recovering by the end of the day . the goes light curve reaches c level only twice , and then briefly , during these events ( fig . [ fig : goes20120209 ] ) . the pfss model ( fig . [ fig : pfss20120209 ] ) does not show obvious connections between these regions of activity , and neither does the mf@xmath55 model field . soho / lasco c2 data suggest , however , that the prominence eruption from the northeast limb region is coupled to regions to a clock angle of about 08h , i.e. , to the general direction of the emerging bipolar region in the southern hemisphere and the nearby quiet - sun filament , even deforming the streamer slightly to the south of that by its expansion . the absence of clear stretched - field signatures , other than the expanding cme ( shock ) front , in stereo / a cor1 , however , suggests that the structures seen from soho / lasco c2 may not , in fact , map onto the disk as seen from earth . there is no obvious direct evidence for coupled substantial activity for 2012/02/09 , although the pronounced cooling seen in the quiet - sun filament region around @xmath51 during the rise and eruption of the prominence over the northeast limb is suggestive of the existence of long - range couplings in the field that are manifested during the eruption possibly through high - arching field next to open field that is rooted in a patch around @xmath56 . we note that with these events occurring on the eastern hemisphere , and major parts of that on the east limb , the model field configuration may not reflect all the relevant parts as the field underlying these events is far from fully assimilated into the field models . another interpretation is the analogy with the model discussed by @xcite , in which reconnection during the eruption process forms long - range connections that can lead to dimming and adiabatic cooling of plasma on loops connecting far away . we have shown evidence that directly - connected , adjacent coronal volumes ( in any combination of active and quiet regions ) can be involved in sympathetic events when the connecting field is affected by a flare or eruption ( examples discussed in this study include cases a , b , c , g ) . similarly , adjacent filaments , nested within a larger overall configuration , can be associated in coupled eruptions , as explored in the mhd model by @xcite , for which cases a and h contain good illustrations . for case i , there is only weak evidence of long - range couplings in the pfss and mf field models between an erupting east - limb filament and a dimming , cooling region in the quiet sun on disk . this coupling appears to be through high - arching field under a helmet adjacent to an open - field patch ( reminiscent of the model by lugaz et al , 2012 ) . although the field configuration for such a volume of the corona near the eastern limb is subject to major uncertainties , the unusual evolution of intensity profiles and thermal signatures over an on - disk region of quiet - sun as an east - limb filament eruption proceeds suggests that such connections do , in fact , exist . we suggest that these eastern - hemispheric events illustrate that magnetic connections in this case are not correctly approximated by the pfss and mf field models simply because much of the surface field involved is not , or only poorly , visible by existing magnetographs . another illustration of the problems associated with the poor magnetograph coverage is found in case b : a magnetofrictional field model for 2011/02/16 reveals a set of connecting field structures that ties together all of the major activity for the preceding days of activity , but these connections had not fully developed in the model by then . given the patterns of coronal activity on the 14th and 15th , it is possible that connections that already exist in the real corona were yet to develop in the model . another likely example of the consequences of not having all of the information of the surface field near the edge of the visible disk and beyond is that of case g , where the aia observations show a direct connection between two distant regions linked by a coupled eruption and flare . a fourth example of this is that of case a described in detail by @xcite who explicitly show the substantial coronal evolution between a pfss model that includes regions that emerged on the far solar hemisphere and one that does not . in other cases , field models are helpful in revealing connections where nestings of field , some including ropes , could not evolve without upsetting adjacent fields ( of which case c is a good example ) . in other cases , these field models reveal that evolution in one region affects the evolution in a distant region via connections that sit side by side in a third location ( e.g. , cases b and d ) and that thus may involve a coronal null . detailed mhd modeling is needed to reveal to what degree such multi - hop connections , mutually influencing each other as they evolve , communicate evolutionary signals leading to sympathetic activity . among the cases discussed , we have identified a few in which coronal propagating expansion fronts or shock waves ( or both ) reach ( the vicinity of ) other regions ( quiet or active ) that then destabilize and erupt ( such as in cases c and h , and perhaps b ) . understanding the potential couplings in such cases also calls for detailed evolutionary mhd models . in other event sets , such as some in case h and for case f , there may be no evidence for connections of any kind between two distant eruptions . in those cases distinct eruptions from different sites overlap in time so that the resulting inner - heliospheric perturbation develops as a single composite event . knowing of such composite events is important to understanding the possible impacts on geospace and to the forecasting of such impacts . in contrast to this , case e shows how much of the front - side corona can be evolving , but without introducing noticeable confusion about the source regions and triggering events of inner - heliospheric perturbations . our study leads us to two primary conclusions : ( 1 ) long - range couplings by several distinct physical mechanisms affect the evolution of the large - scale corona , and ( 2 ) the structure of the large - scale coronal field is the product of insertion , buildup , ejection , and relaxation of electrical current systems that are inadequately represented by present - day modeling and impossible to reconstruct from present - day observational data . these properties of the dynamic , 3-dimensional , dynamic corona need to be incorporated into assimilative models that couple the solar surface to the heliosphere to accurately forecast the solar wind , its variable properties , and the pathways of solar energetic particles . present - day instrumentation supported by relatively elementary modeling for the global coronal field has revealed at least three different ways eruptive and explosive events in the solar outer atmosphere influence the destabilization of other regions : evolving direct magnetic coupling , distortion of the enveloping field by a large eruption developing into a cme , and the effects of an expansion front or coronal wave associated with a major flare or eruption . observations suggest that indirect coupling through an intermediate region adjacent to a mutual separatrix surface could be another pathway for couplings to occur . whereas in some cases the evidence for coupling between events is direct and uncontroversial , in many others it depends on the applicability of the model fields used in the interpretation . this presents a major problem for several reasons . first , the overall coronal field is built up through flux emergence , shear , and eruptions that require a full - sphere model and likely at least a full rotation of model time prior to the event to be studied ( see the model and discussion by * ? ? ? . it may require more than a year of solar time for the high - latitude fields that involve , among others , the polar crown filament configurations which are long - term assemblages of years of active - region decay products ( e.g. , * ? ? ? this is not only computationally demanding , but also requires that we increase our observational coverage of the solar surface in at least line - of - sight , perhaps even vector - magnetograms , from only the earth - facing side as available at present to a substantially larger coverage of the solar surface . in order to succeed in obtaining an acceptable model representation of the real - world coronal field , we have to realize that field disturbed by a cme requires a good fraction of a day to relax . a lower limit to that time scale is set by the euv afterglow of post - eruption arcades ( extending beyond , for example , the time interval during which supra - arcade downflows might be visible , e.g. * ? ? ? * ) . for 21 eruptions from outside active - region core domains studied here , the average coronal relaxation time estimated from 171 , 193 , and 211 signals is @xmath57h . the time scale for the euv afterglow following quiet - sun eruptions of @xmath57h is much longer than the corresponding signatures of flares and eruptions within active regions . there @xmath58h ( estimated from 8 events , with one strong outlier ) . we can compare that ratio of @xmath59 with a ratio of typical timescales involved . for length scales @xmath60 and alvn speeds @xmath61 we can derive a time scale ratio of @xmath62 ) . for typical length scales of 250 mm and 50 mm , field strengths of 10 g and 100 g , and densities of @xmath63@xmath64 and @xmath65@xmath64 for quiet - sun coronal regions and active - region flaring interiors , we find @xmath66 . that ratio is compatible with the observed ratio of time scales , consistent with the interpretation that the euv eruption afterglow in quiet sun is a signature of the post - eruption reconnection like the relaxation that occurs in active regions after eruptions . with 4 cmes going off on an average day , this long reconnection time scale means that the high coronal field is most likely often evolving from the disturbing effects of previous eruptions at the time of any event that we elect to study . how much of the coronal field is involved in such reconfiguration ? one might look at the angular extent of cmes to estimate this . @xcite show a power - law distribution of cme opening angles with an average power - law index of @xmath67 that holds for opening angles of 10@xmath3 up to above 120@xmath3 . the average opening angle is about 50@xmath3 , corresponding to about 1/10th of the full sky as seen from the sun . one might thus infer that roughly 1/10th of the corona by volume is involved in an average eruption . however , even a cme with a small opening angle must break all field that closes over its site of origin . with a typical base cross section of order 100@xmath3 for the helmet streamer belt in the pfss approximation , and assuming a comparable width along and across the helmet direction , then up to @xmath68rd or the solar surface may have a fraction of its field forced to reconnect during a cme . the latter estimate is consistent with the extent over which perturbations are seen to travel in the tri - ratio movies such as shown in the on - line table@xmath9 . if we take @xmath69 to be characteristic of the reconnection time scale for the high coronal field , and for an average of 4 cmes / day that each reach over about 1/3rd of the solar surface , about one in @xmath70 of all cmes affect field that is still relaxing from preceding cmes . the time scale for ongoing reconnection is likely larger than @xmath69 ; if we use a reconnection time scale of @xmath71 then close to one in @xmath72 cmes occur while the field that it encompasses is still relaxing . in view of this , one should question the validity of any global coronal model at the start of a selected event that does not include in its computation at least several preceding eruptions . the events of , and model for , case a are an excellent illustration of this : even as the model shows how one filament can be destabilized in the wake of another s eruption , and although this certainly appears to be the case on that date , the second filament does not take off until half a day after the nearby afterglow of the reconnection arcades associated with the first eruption faded away ; only the continuing afterglow in a more distant location ( at the edge of the northern coronal hole ) bore witness to the ongoing deformation of the field enveloping the filaments and flux - rope configurations . the double filament eruptions in case h are similar in that respect , with the eruptions in that case separated by approximately 7h . this work has illustrated that we should anticipate that in many instances , connections in the solar corona exist that present - day coronal field models do not reveal . the fundamental problem that we face in understanding the connections in the solar corona even in the absence of eruptive events is that of the dependence of its instantaneous state on what happened before , i.e. , its magnetohydrodynamic hysteresis . in addition to the effects on time scale from flux emergence in days to field evolution over months , there are those that occur on time scales of hours to a day associated with cmes . these are particularly important to understanding the evolution , and at least in some cases the triggering , of cmes as well as the generation and escape pathways for solar energetic particles into the heliosphere ( see also , for example , the study by * ? ? ? * and references therein on the role of precursor activity in seeding the coronal environment with particles that can be effectively accelerated to high energies in subsequent events ) . as we are faced with inadequate observational coverage and consequent model abilities , it remains difficult to establish how often the physical coupling of one event with another impacts coronal evolution . the importance for flare / eruption forecasting and thereby space - weather forecasting therefore remains unquantified . the evidence presented in this study motivates development of advanced capabilities and further study of the available data sets . over the years , mhd and pfss field models ( see , e.g. , * for a comparison between these ) have demonstrated that the largest - scale coronal configuration is represented fairly well : coronal hole boundaries and even their dynamics are approximated fairly well by open field regions in global pfss models @xcite , solar wind models based on series of pfss models are reasonable approximations of the quiescent solar wind particularly in view of our limited observational coverage of the solar surface magnetic field @xcite , coronal streamers modeled for eclipses approximate some of the observed ones well , and even a fly - through of comet lovejoy shows reasonable agreement with an mhd model @xcite . on the other hand , each of these tests of our understanding of the large - scale coronal structure and its evolution exhibits substantial mismatches ( see , e.g. , * ? ? ? * ; * ? ? ? * and references therein for discussions of these aspects ) : modeled open - field regions do differ substantially from observed coronal - hole outlines , solar wind predictions are far from perfect in both speed and field polarity , the model streamer structures do not always reflect the directions of observed streamers differ while some observed helmet structures are missed altogether , the tail dynamics of comet lovejoy suggests mismatches in field patterns of at least several degrees , and even the evolution of the sun s open flux over time remains unexplained to within a range of a factor of at least two . all of these mismatches are likely caused by a mixture of the incomplete knowledge of the surface magnetic field , the evolving patterns of the coronal electromagnetic field that continually evolves following flux emergence and decay as well as eruptive events , and of assumptions and approximations made in the modeling . numerical experiments can help us understand the relative weighting of the impacts of observational and modeling limitations , but ultimately sustained observations from perspectives well away from the sun - earth line , of the high corona , and of the details of current buildup and disappearance are essential in making the next leap forward in understanding the corona and the space weather that it drives . we thank nariaki nitta , bernhard fleck , and miho janvier for discussions of early versions of the manuscript . this work was supported by nasa contract nng04ea00c in support of the sdo atmospheric imaging assembly .

we perform a statistical study of permanent changes in longitudinal fields associated with solar flares by tracking magnetic features . the yafta feature tracking algorithm is applied to gong++ one - minute magnetograms for 77 x - class and m - class flares to analyze the evolution and interaction of the magnetic features and to estimate the amount of canceled magnetic flux . we find that significantly more magnetic flux decreases than increases occurred during the flares , consistent with a model of collapsing loop structure for flares . correlations between both total ( unsigned ) and net ( signed ) flux changes and the goes peak x - ray flux are dominated by x - class flares at limb locations . the flux changes were accompanied in most cases by significant cancellation , most of which occurred during the flares . we find that the field strength and complexity near the polarity inversion line are approximately equally important in the flux cancellation processes that accompany the flares . we do not find a correlation between the flux cancellation events and the stepwise changes in the magnetic flux in the region .

1210.4200

it is now known that photospheric magnetic fields change significantly , abruptly and permanently as a result of major x- and m - class flares ( _ e.g. _ ; ; , @xcite ; ; ; ; ; ; ) . detection and study of photospheric magnetic field changes during solar flares can aid in a better understanding of the mechanisms of energy buildup and release in flares . despite these observations , most models of flares assume that these events are coronal and that no photospheric magnetic field changes occur over the flare duration @xcite . ideally , coronal field models should be based on boundary conditions observed in the lower atmosphere . however , it is not clear how to relate the restructuring of approximately force - free or current - free coronal field with the changes in the forced photosphere . a few possible mechanisms were offered in past works that could cause the sudden and permanent photospheric field changes observed . the most frequently mentioned processes are the emergence of new magnetic flux and changes in the direction of the field @xcite . supporting magnetic flux emergence as a possible mechanism , observed emergence of new flux in sunspot umbra immediately after an x - class flare . suggested the alfvn wave scenario as an explanation for the observed rapid variations in the line - of - sight component of the photospheric magnetic field in the flare impulsive phase : large - scale alfvn wave pulses may transport both energy and changes in the magnetic field rapidly from the flare site in the corona to the lower atmosphere . estimated back reaction of the photospheric magnetic field during a solar flare and proposed a model of magnetic implosion and collapsing loop structure towards the neutral line . , , , and found strong evidence that this prediction is true . analyzed pixel - by - pixel the field changes accompanying 77 x- and m - class flares using global oscillation network group ( gong ) full - disk magnetograms . in this work , we use the same flares from the gong flare list to study flare - related photospheric field changes by decomposing the field into individual magnetic features . as have noted , the change in magnetic flux may be a more important physical quantity than the largest local field change at one particular location . by only picking up representative pixels with the fastest and largest field changes , free of artifacts @xcite , one may not capture a complete picture of the changes in the magnetic field in the region . however , the value of magnetic flux changes may capture the effects of flares on the entire active region . the calculations of magnetic flux made by past authors have either summed the flux of changing pixels over the entire active region @xcite or confined attention to simple bipolar active regions with highly sheared neutral lines ( _ e.g. _ ; ) . in the present work the active regions are divided into magnetic features ( substructures ) whose flux changes are calculated individually . this allows us to identify which parts of a region change significantly and also to investigate the interactions among different features in an active region . one of the processes involving magnetic features associated with flares is flux cancellation . flux cancellation plays an important role in some theories of solar eruptions ( _ e.g. _ ; ) . it is found to be essential for the formation of prominences @xcite , and may possibly trigger filament eruptions and coronal mass ejections @xcite . the mechanism for flux cancellation is suggested to be a necessary condition of flare initiation as a part of slow reconnection processes in the lower atmosphere @xcite . the first evidence of the possibility of magnetic reconnection in flux cancellation was presented by , who analyzed vector magnetic field changes and demonstrated how features previously not directly connected by transverse fields partially reconnected during flux cancellation . strong observational evidence of magnetic reconnection in the lower atmosphere leading to instability and flare / coronal mass ejection ( cme ) events was presented by , , and . reconnection can occur at different levels of the solar atmosphere ; above , below , or at the photospheric level @xcite . observed that magnetic flux disappeared in the chromosphere earlier than in the photosphere at the cancellation sites , which suggests submergence of the flux implying that reconnection occurs above the photosphere . confirms this finding . however , reported flux cancellation , following a c - class flare , with the presence of upflows in the canceling feature , which suggests the reconnection below or at the level of the photosphere . in the case of the formation of prominences , cancellation is explained by the emergence of u - shaped loops @xcite . most of these studies suggest flux cancellation as a pre - flare condition , with some exceptions ( _ e.g. _ ) . photospheric magnetic shear increases have been observed to occur during and after flares @xcite , and before and during flares @xcite . in this work , we estimate flux cancellation in the magnetic features at the polarity inversion line ( pil ) associated with a major flare , at the time before , during , and after the flare . in this study full - disk photospheric magnetograms from the gong sites are used . gong full - disk magnetograms are obtained with a 1-minute cadence , 2.5-arcsec pixel size , and a noise level of 3 gauss ( g ) per pixel . we have analyzed 77 sets of gong magnetograms for 38 x- and 39 m - class flares . we confined our attention to the most energetic flares with the best data coverage . we eliminated all flares of geostationary operational environmental satellite ( goes ) class weaker than m5.0 and all with an apparent central meridian distance ( longitude distance from central meridian ) greater than @xmath0 . we further limited our attention to those flares for which gong magnetograms are available over approximately 1 hour before and after the flare . the flares studied are identified in tables 1 ( m - class ) and 2 ( x - class ) of . studied , pixel by pixel , the field changes accompanying the flares . in this work we study flare - related photospheric field changes analyzing collections of pixels ( magnetic features ) . the preliminary data reduction procedure is as in . _ i _ ) the magnetograms are remapped to azimuthal equidistant projection , @xmath1 field of view , local to a flare . _ ii _ ) the maps are corrected for differential rotation and registered , as a reference , to a ten - minute average of magnetograms prior to the flare . we use the yet another feature tracking algorithm ( yafta ) feature - tracking code @xcite to decompose the field into individual magnetic features and to study their evolution and interplay during the flare . we have used yafta s `` downhill '' algorithm to group together more than four pixels lying on the same `` hill '' in the absolute field strength above 100 g. the 100 g threshold identifies active region fields in gong magnetograms . small features often have short lifetimes . the smallest feature that showed abrupt and permanent flux change in this work consisted of eight pixels ( which is @xmath2 0.5 heliographic degrees in diameter ) , and the biggest one was 650 pixels ( @xmath2 3 degrees in diameter ) . the program accomplishes matching of features between two consecutive magnetograms . most of the features fragment , merge , appear , and disappear between the magnetograms , which makes the matching process complex . as a result , features change their labels . only features that do not appear , disappear , fragment , or merge between the magnetograms `` propagate '' themselves ; they keep their label through all steps . to track distinct magnetic features , for each flaring active region we identify the features on a reference image composed of ten magnetograms immediately preceding the flare . if some of these ( two or more ) feature - hills appear as a single feature - hill in most of the frames during the whole analyzed time interval and each of the individual features does not show significant temporal variations , they are merged into one feature on the reference image . an example of an active region with 18 magnetic features found is shown in figure 1(a ) . we locate each of these features on each of our images . we analyze 4 h of data covering each flare . we calculate the total flux and the flux - weighted average @xmath3- and @xmath4-coordinates of each feature . -map of stepwise field changes found by during this flare . the locations of abrupt flux decreases ( blue circles ) and increases ( red circles ) of magnetic features found in this work are overplotted . ] most of the previous feature - tracking works have applied a radial field correction to the magnetograms before processing , dividing the longitudinal fields by the cosine of the heliocentric angle ( _ e.g. _ ) . for most photospheric fields this is generally a valid assumption @xcite . however , for the photospheric fields in a flaring active region , particularly those fields associated with strong , twisted magnetic structures directly involved in the eruption , this assumption may not be applicable . reported evidence of this in the result that the average magnitude of photospheric field changes tends to increase as a function of distance from the disk center , indicating that the fields changing the most tend to be nearly horizontal . the radial field correction is necessary when features inside of an active region are tracked for a long period of time to ensure a constant threshold during the whole tracking interval ( _ e.g. _ , see ) , while we track features only up to 4 h. we analyze the magnitude and sign of flux changes in each magnetic feature during a flare . the magnitude and sign of flux changes are defined by fitting the time evolution of the total flux in each magnetic feature from before the onset to the end of a flare with a straight line , assuming that the flux evolution around the flare is linear and ignoring long - term variations and trends . figure 2 demonstrates an example of the strongest field change found in our analysis . the spread in the points is mainly caused by seeing changes in the earth s atmosphere with some contribution from ongoing solar field changes . we selected magnetic features showing stepwise flux changes during flares . each of the flux changes satisfies the following criterion . the value of the flux change should exceed 1.3 @xmath5 of the scatter in the measured background flux . the background flux is defined as a median value of the pre - flare flux in the feature . the minimal and maximal values of the background fluxes we found are @xmath6 and @xmath7 mx ( maxwell ) . the 1 @xmath5 level is between @xmath8 and @xmath9 mx . the maximal and median values of all the stepwise flux changes are 7.3 @xmath5 and 3.0 @xmath5 , respectively . it has been noted that the duration of the flux changes is on the scale of 1040 min ( _ e.g. _ ; ; ) . we select flux changes which are completed within 30 min or less . the statistics for the flux increases and decreases for the 77 flares are presented in table 1 . in total 114 stepwise flux changes which satisfied the above criteria were selected . we find that the field decreases are significantly more numerous than the increases ( 70 versus 44 ; see the first two columns under the ` all ' heading in table 1 ) . the bias towards the field decreases is noticeable for both disk - center and limb locations with a slightly larger ratio for the limb regions ( the first two columns under ` disk - center ' and ` limb ' headings in table 1 ) . we designate `` near disk - center '' for the locations @xmath10 and `` near limb '' for the locations @xmath11 , where @xmath12 is the solar disk radius in the image plane . median longitudinal field changes are larger in the regions close to the limb ( compare the last two columns under the ` disk - center ' and ` limb ' headings in table 1 ) , indicating that their field vector is nearly horizontal in general . limb changes are more than twice as numerous as disk - center changes . x - class flares show more numerous ( 74 versus 40 ) and strong photospheric flux changes in comparison to m - class flares ( adding the numbers of flux increases and decreases under the ` x - class ' and ` m - class ' headings in table 1 ) . for x - class flares , the flux decreases are more than twice as numerous as the increases in both disk - center and limb regions , while for m - class flares the ratio of decreases to increases is close to 1:1 for disk - center regions and has an opposite bias ( 1:2 ) towards flux increases in regions close to the limb . @ l @ * 5c @ field & & minimum & maximum & median + changes & increases & decreases & change & change & change + all & 44 & 70 & 0.02 & 21.04 & 0.61 + x - class & 21 & 53 & 0.02 & 21.04 & 0.81 + m - class & 23 & 17 & 0.10 & 4.10 & 0.48 + disk - center & 15 & 22 & 0.10 & 1.84 & 0.35 + limb & 29 & 48 & 0.02 & 21.04 & 0.82 + disk - center x - class & 4 & 11 & 0.11 & 1.84 & 0.42 + disk - center m - class & 11 & 11 & 0.10 & 1.60 & 0.27 + limb x - class & 17 & 42 & 0.02 & 21.04 & 0.82 + limb m - class & 12 & 6 & 0.15 & 4.10 & 0.84 + these statistics for magnetic features agree with the basic result reported by that decreasing longitudinal fields tend to outnumber increasing longitudinal fields , except that here magnetic features of near - limb m - class flares have twice as many longitudinal flux increases as decreases . mx in panel ( a ) separate features whose flux change was anti - correlated with their background flux from the big features whose flux change did not conform to this general pattern . ] the overall bias towards flux decreases in comparison with flux increases can be seen in the scatter plot of the change in the signed total magnetic flux against the background signed total magnetic flux ( figure 3(a ) ) . the flux changes are calculated for individual magnetic features . in figure 3 , we designate a magnetic flux change `` signed '' if magnetic polarity is taken into account and `` unsigned '' if it is ignored . each feature consists of pixels of like polarity ; thus the flux is total flux in all the features . the strongest flux change was found during an x1.2 flare on 26 october 2003 with its onset at 05:57ut , which is much larger than all other flux changes found in this work . on the other hand , the change in the unsigned total magnetic flux against the background unsigned total magnetic flux ( figure 3(b ) ) demonstrates that this case of maximum flux change is consistent with the common trend in flux changes , showing that this value may not be an outlier . however , this is a log - log plot . this data point has the same order of magnitude as the rest , but otherwise is an outlier . in any case , large flux changes tend to be associated with x - class flares . figure 3(c ) shows the change in the unsigned total magnetic flux against the radial position of the flare on the solar disk . we can see that the large flux changes associated with x - class flares occur closer to the limb . in total , a larger range of flux changes is found in a flare near the limb than near the disk center . these conclusions are consistent with the statistics of flux changes from @xcite ( @xcite ; see sections 5 and 7 of their paper ) . figure 3(d ) shows the change in the unsigned total magnetic flux as a function of the goes peak x - ray flux . we see a weak overall increasing trend of the flux change with respect to the flare strength . we note that the size of magnetic features changes in time , causing an uncertainty in the flux estimates . some fields crossing the 100 g threshold introduce this error into the flux estimate . however , such errors do not change the basic patterns of increasing and decreasing fluxes . @ l * 4c @ & flux change vs. & flux change vs. + & background flux & peak x - ray flux + all & -0.42 ( @xmath13 ) & 0.38 ( @xmath14 ) + x - class & -0.50 ( @xmath15 ) & 0.35 ( @xmath16 ) + m - class & 0.22 ( 0.19 ) & 0.07 ( 0.67 ) + disk - center & -0.64 ( @xmath17 ) & 0.13 ( 0.46 ) + limb & -0.40 ( @xmath18 ) & 0.36 ( @xmath19 ) + disk - center x - class & -0.74 ( @xmath20 ) & 0.14 ( 0.64 ) + disk - center m - class & 0.10 ( 0.68 ) & 0.02 ( 0.93 ) + limb x - class & -0.49 ( @xmath21 ) & 0.33 ( @xmath22 ) + limb m - class & 0.23 ( 0.37 ) & -0.01 ( 0.96 ) + we then compute pearson s correlation coefficients between the magnetic flux change and the background magnetic flux . we find that the flux change in the features whose absolute value of the background flux is @xmath23 mx is anti - correlated with their background flux , while the larger flux features do not conform to this general pattern ( see figure 3(a ) ) . this bimodal behavior seems to be analogous to the dichotomy between weak ( @xmath24 g ) and strong ( @xmath25 g ) field changes found in . the correlation coefficients computed for magnetic features with absolute values of the background flux @xmath23 mx are shown in table 2 . the @xmath26-value , the probability that the observed correlation occurs by chance , is also included in this table . we find a modest anti - correlation between the flux change and the background flux for both disk - center and limb locations . the correlation is stronger for near disk - center locations and is dominated by x - class flares . m - class flares show no correlation between the flux change and the background flux ; in this case we find about equal numbers of flux decreases and increases ( see table 1 ) . pearson s correlation coefficients between the magnetic flux change and the goes peak x - ray flux show a weak but statistically significant overall correlation dominated by x - class flares at limb locations ( see table 2 ) . in table 2 , the goes peak x - ray flux and the position on the disk are approximately equally influential in the correlations . the magnetic flux changes estimated here represent flux changes at certain locations ( in individual magnetic features ) within active regions . to obtain a fuller picture of the flare effect on active regions , summed pixel - wise field changes over the entire active region . we calculate the flux change in a flare by summing abrupt flux changes in the magnetic features during the flare in the active region . some magnetic features in our analysis did not show stepwise flux changes . the summation of the flux changes over many magnetic features may cause a mixture of some pixels that do not change their flux in an abrupt , stepwise way and some pixels with abrupt , stepwise changes in the flux of opposite signs . this can cause even those magnetic features with many pixels changing stepwise to result in an overall flux profile that does not show a clear stepwise change . we restrict our attention to those magnetic features whose overall flux shows a clear , stepwise , permanent change . we expect these changes to dominate the flux evolution of an active region during a flare . an example of the @xmath27-map ( a map of stepwise field changes defined by ) is given in figure 1(b ) for the x6.5 flare on 6 december 2006 . the locations of magnetic features with abrupt flux changes found in this work are overplotted on the same figure . feature 3 in figure 1(a ) , which demonstrated one of the strongest negative stepwise flux changes in this work ( see figure 2 ) , is fully populated by pixels with strong stepwise changes of the same sign during the flare . feature 1 included both field decreases and increases . although most of its pixels did not exhibit clear , stepwise field changes because of spectral line changes during the flare , the summed flux change showed a clear stepwise flux increase . feature 5 also included pixels with field decreases , and showed a flux decrease twice weaker than that of feature 3 . feature 2 partially consisted of pixels with the strongest negative field changes , but its flux change did not exceed 1.3 @xmath5 of the scatter in the background flux , which is the criterion we adopted at the beginning of this section . the other features had smaller stepwise field decreases , and are not specified in figure 1(a ) . they also showed stepwise flux changes twice smaller than those of feature 3 . for similar reasons , some of the 77 flares did not include magnetic features whose flux changed in a clear stepwise manner even though each of these flares did include at least one pixel that showed a stepwise flux change @xcite . also , the imposed threshold for the flux changes has restricted our analysis to strong field regions , which may have left some weaker flux changes in magnetic features beyond our analysis . thus , among 77 flares analyzed , only in 45 ( 25 x- and 20 m - class ) flares we found at least one magnetic feature with strong stepwise flux change . the location of the active region which did not involve stepwise flux changes during some flares does not seem to be correlated with the distance of the active region from the disk center . some flares in the same active region and at a similar separation from the disk center involved magnetic features with stepwise flux changes and others did not . the detection of the stepwise flux changes did not vary between gong stations . we note that all cases of the absence of the stepwise flux changes are associated with flare class x1.8 and less , except for one x3.8 flare on 17 january 2005 that occurred near the disk - center location . on the other hand , among 45 flares showing stepwise flux changes , only 12 were above flare class x1.8 . @ l @ * 5c @ field & & minimum & maximum & median + changes & increases & decreases & change & change & change + all & 18 & 27 & 0.002 & 22.17 & 1.17 + x - class & 8 & 17 & 0.002 & 22.17 & 1.63 + m - class & 10 & 10 & 0.12 & 2.52 & 0.80 + disk - center & 5 & 13 & 0.002 & 2.65 & 0.63 + limb & 13 & 14 & 0.29 & 22.17 & 1.75 + disk - center x - class & 2 & 5 & 0.002 & 2.65 & 1.32 + disk - center m - class & 3 & 8 & 0.12 & 2.07 & 0.55 + limb x - class & 6 & 12 & 0.29 & 22.17 & 2.89 + limb m - class & 7 & 2 & 0.30 & 2.52 & 1.36 + @ l @ * 5c @ field & & minimum & maximum & median + changes & increases & decreases & change & change & change + all & 19 & 26 & 0.04 & 17.50 & 1.16 + x - class & 6 & 19 & 0.29 & 17.50 & 1.75 + m - class & 13 & 7 & 0.04 & 2.16 & 0.63 + disk - center & 6 & 12 & 0.07 & 2.65 & 0.84 + limb & 13 & 14 & 0.04 & 17.50 & 1.75 + disk - center x - class & 1 & 6 & 0.37 & 2.65 & 1.32 + disk - center m - class & 5 & 6 & 0.07 & 2.16 & 0.55 + limb x - class & 5 & 13 & 0.29 & 17.50 & 2.73 + limb m - class & 8 & 1 & 0.04 & 2.14 & 1.36 + the statistics of net ( signed ) and total ( unsigned ) flux changes are shown in tables 3 and 4 , respectively . we find generally more decreases than increases in both net and total flux changes ( tables 3 and 4 under the ` all ' heading ) . the bias between the flux decreases and increases is significant in the case of disk - center flares for both net and total flux changes ; the ratios of the flux decreases to increases are 2.6:1 and 2:1 for net and total flux changes , respectively ( tables 3 and 4 under ` disk - center ' heading ) . limb flares show approximately equal numbers of increases and decreases for both net and total flux changes ( tables 3 and 4 under ` limb ' heading ) . calculated the longitudinal field component associated with tilt and azimuth angles of the magnetic vector at different locations on the solar disk . they found that if the field vector becomes more horizontal during a flare , then the longitudinal field can either increase or decrease , both near the disk center or near the limb . however , decreases would likely outnumber increases at all parts of the disk investigated ( _ i.e. _ , within @xmath0 from the disk center ) , and this bias is expected to be greater near the disk center than near the limb . we find these patterns in our data . for the x - class flares , the flux decreases are at least twice as numerous as the increases at both the disk center and limb locations and for both net and total flux changes ( see tables 3 and 4 under ` x - class ' , ` disk - center x - class ' , and ` limb x - class ' headings ) . reported a similar trend for the total flux changes in x - class flares ; for the net flux changes they did not find any significant trend . we find more than twice as many flux decreases as increases for the net flux changes in m - class flares at the disk - center locations , which is consistent with the same statistics from , while the trend is opposite for the limb flares ( table 3 under ` m - class ' , ` disk - center m - class ' , and ` limb m - class ' headings ) . for the total flux changes in m - class flares , the increases are nearly twice as numerous as the decreases in general ( table 4 under ` m - class ' heading ) . a significant bias towards total flux increases in near - limb m - class flares is observed , while the same category from does not show this trend . similar to the feature - wise statistics , the median flux changes are larger in the regions close to the limb for both total and net flux changes and larger in the case of x - class flares in comparison with m - class flares . @ l * 4c @ & + & net flux & total flux + & change & change + all & -0.28 ( 0.09 ) & 0.55 ( @xmath28 ) + x - class & -0.39 ( 0.08 ) & 0.55 ( @xmath29 ) + m - class & 0.32 ( 0.20 ) & 0.47 ( @xmath30 ) + disk - center & -0.56 ( @xmath31 ) & 0.68 ( @xmath20 ) + limb & -0.26 ( 0.23 ) & 0.56 ( @xmath32 ) + disk - center x - class & -0.63 ( 0.19 ) & 0.68 ( 0.10 ) + disk - center m - class & -0.42 ( 0.24 ) & 0.87 ( @xmath33 ) + limb x - class & -0.41 ( 0.13 ) & 0.52 ( @xmath34 ) + limb m - class & 0.55 ( 0.17 ) & -0.33 ( 0.44 ) + @ l * 4c @ & + & net flux & total flux + & change & change + all & 0.49 ( @xmath35 ) & 0.56 ( @xmath36 ) + x - class & 0.39 ( 0.07 ) & 0.47 ( @xmath37 ) + m - class & -0.11 ( 0.65 ) & 0.21 ( 0.40 ) + disk - center & 0.34 ( 0.17 ) & 0.36 ( 0.14 ) + limb & 0.48 ( @xmath38 ) & 0.56 ( @xmath39 ) + disk - center x - class & 0.20 ( 0.72 ) & 0.20 ( 0.68 ) + disk - center m - class & -0.15 ( 0.65 ) & 0.09 ( 0.80 ) + limb x - class & 0.40 ( 0.11 ) & 0.50 ( @xmath40 ) + limb m - class & 0.003 ( 0.99 ) & 0.81 ( @xmath41 ) + figure 4 shows scatter plots of the changes in the net and total magnetic flux against the background net and total magnetic flux , against the radial position of the region on the solar disk , and against the goes peak x - ray flux . the correlations between the magnetic flux changes and the background magnetic flux and the goes peak x - ray flux are summarized in tables 5 and 6 . the only statistically significant modest anti - correlation between the net flux change and the net background flux is found for the disk - center flares ( see table 5 under ` disk - center ' heading ) , which agrees with the results in table 3 . some weak anti - correlation is also noted for the x - class flares ( table 5 under ` x - class ' heading , first column ) , which is consistent with figure 4(a ) and the statistics in table 3 . the distribution of points in figure 4 is not much different from the one shown in figure 3 . figure 4(b ) shows more points that do not conform to the general trend for the total flux changes in m - class flares , while the flux changes in x - class flares maintain this pattern quite well , except for one point corresponding to one of the limb flares . figures 4(c ) and 4(d ) clearly show that stronger flux changes occur preferentially closer to the limb and that a larger range of the flux changes corresponds to the limb locations . figures 4(e ) and 4(f ) show an overall increasing trend of the net and total flux changes as a function of the peak x - ray flux . we do find moderately significant correlation with the peak x - ray flux for both net and total flux changes , which are dominated by the x - class flares near the limb locations . this tendency is clearer for the total flux changes ; the net flux changes show statistically significant correlation only for limb flares ( compare table 6 under ` all ' vs. ` limb ' and ` x - class ' vs. ` limb x - class ' ) . we also identify episodes of flux cancellation , when closely spaced , opposite - polarity features simultaneously lose flux @xcite . we estimate the flux cancellation as described in . the magnetic flux of features derived from high - cadence data suffers short - term fluctuations ( noise ) with large amplitudes in this study , as the gong data are also influenced by the earth s atmosphere . the fluctuations significantly affect the measurements of flux cancellation , leading to exaggerated values of the flux losses in some cases by summing up noisy data instead of real flux evolution . we smooth the total flux in each feature with a 30-min running mean . we then mostly see a long - term evolutionary trend of the flux that we use in further investigations of flux cancellation . first , we define instances when the boundaries of closely spaced , opposite - polarity features come into contact , using the technique described in . for our analysis , the closest distance between the features is limited by the pixel size of the gong data to about 1.8 mm on the sun at disk center . further , for a pair of neighboring features to be partially canceled in the next magnetogram , the centers of magnetic features are required to approach and the flux of both features to decrease . this approach is subject to many sources of uncertainties , related to the applied threshold ( some flux may be missed by the threshold ) , fluctuations in magnetic flux due to measurement noise and true evolution , as well as possible interaction among several neighboring like- or opposite - polarity features . to account for these uncertainties , flux cancellation is computed in three ways as in : * estimated minimum value of flux loss from the pair of canceling features , summed over all cancellation events ; * estimated average value of flux loss from the pair of features exhibiting flux cancellation over at least two consecutive magnetograms ( multi - step cancellation ) , summed over all multi - step cancellation events ; * estimated minimum value of flux loss from the pair of features exhibiting flux cancellation over at least two consecutive magnetograms , summed over all multi - step cancellation events . @ l * 7c @ date & goes & radial position & number of & number of & all events & multi - step & multi - step + & class & ( in @xmath42 ) & events & multi - step events & @xmath43 & @xmath44 & @xmath43 + 22 jun 2001 & m6.2 & 0.74 & - & - & - & - & - + 23 jun 2001 & m5.6 & 0.43 & 29 & 2 & 5.5 & 0.2 & 0.2 + 05 sep 2001 & m6.0 & 0.49 & 9 & - & 0.7 & - & - + 09 sep 2001 & m9.5 & 0.70 & 14 & - & 1.1 & - & - + 16 sep 2001 & m5.6 & 0.90 & 2 & - & 0.4 & - & - + 22 oct 2001 & m6.7 & 0.52 & 35 & 8 & 2.2 & 1.3 & 0.6 + 23 oct 2001 & m6.5 & 0.43 & 39 & 13 & 5.0 & 5.7 & 2.5 + 07 nov 2001 & m5.7 & 0.74 & 96 & 21 & 18.2 & 7.1 & 4.6 + 08 nov 2001 & m9.1 & 0.48 & 5 & - & 0.4 & - & - + 28 nov 2001 & m6.9 & 0.28 & 53 & 13 & 7.7 & 3.1 & 2.0 + 29 nov 2001 & m5.5 & 0.19 & 133 & 29 & 11.2 & 4.2 & 2.5 + 26 dec 2001 & m7.1 & 0.83 & 26 & 4 & 12.9 & 4.8 & 3.6 + 09 jan 2002 & m9.5 & 0.29 & 60 & 9 & 10.8 & 3.7 & 2.4 + 14 mar 2002 & m5.7 & 0.41 & 28 & 9 & 3.6 & 2.7 & 1.5 + 11 jul 2002 & m5.8 & 0.80 & - & - & - & - & - + 17 jul 2002 & m8.5 & 0.33 & 26 & 7 & 1.1 & 1.7 & 0.5 + 26 jul 2002 & m8.7 & 0.49 & 173 & 38 & 21.7 & 9.5 & 6.2 + 16 aug 2002 & m5.2 & 0.39 & 62 & 11 & 11.7 & 4.4 & 3.0 + 05 oct 2002 & m5.9 & 0.52 & 10 & 2 & 0.3 & 0.2 & 0.1 + 18 nov 2002 & m7.4 & 0.80 & - & - & - & - & - + 20 dec 2002 & m6.8 & 0.65 & 183 & 52 & 12.2 & 7.1 & 3.9 + 26 oct 2003 & m7.6 & 0.63 & 95 & 12 & 44.8 & 13.3 & 6.1 + 27 oct 2003 & m5.0 & 0.54 & 110 & 28 & 11.0 & 10.8 & 3.3 + 20 nov 2003 & m5.8 & 0.33 & 45 & 9 & 3.5 & 1.5 & 0.5 + 20 nov 2003 & m9.6 & 0.18 & 88 & 28 & 7.9 & 6.2 & 2.6 + 17 jan 2004 & m5.0 & 0.37 & - & - & - & - & - + 20 jan 2004 & m6.1 & 0.32 & 7 & - & 0.1 & - & - + 13 jul 2004 & m6.2 & 0.83 & 10 & - & 0.8 & - & - + 13 jul 2004 & m6.7 & 0.72 & 14 & 2 & 2.5 & 0.5 & 0.5 + 20 jul 2004 & m8.6 & 0.58 & - & - & - & - & - + 22 jul 2004 & m9.1 & 0.35 & 31 & 2 & 1.3 & 0.04 & 0.01 + 25 jul 2004 & m7.1 & 0.39 & 21 & 6 & 1.5 & 1.1 & 0.4 + 14 aug 2004 & m7.4 & 0.56 & 43 & 10 & 8.5 & 2.5 & 1.3 + 30 oct 2004 & m5.9 & 0.39 & 37 & 10 & 3.1 & 4.9 & 0.6 + 15 jan 2005 & m8.4 & 0.33 & 62 & 12 & 16.9 & 5.7 & 2.0 + 15 jan 2005 & m8.6 & 0.31 & 63 & 12 & 20.8 & 10.7 & 5.3 + 06 dec 2006 & m6.0 & 0.90 & - & - & - & - & - + 04 jun 2007 & m8.9 & 0.75 & 34 & 7 & 2.4 & 1.5 & 0.5 + @ l * 7c @ date & goes & radial position & number of & number of & all events & multi - step & multi - step + & class & ( in @xmath42 ) & events & multi - step events & @xmath43 & @xmath44 & @xmath43 + 02 apr 2001 & x20.0 & 0.93 & 54 & 17 & 63.5 & 23.9 & 17.0 + 23 jun 2001 & x1.2 & 0.41 & 42 & 12 & 5.9 & 2.3 & 1.3 + 25 aug 2001 & x5.3 & 0.66 & 58 & 10 & 2.7 & 0.9 & 0.4 + 19 oct 2001 & x1.6 & 0.50 & 27 & 7 & 1.8 & 0.8 & 0.5 + 22 oct 2001 & x1.2 & 0.47 & 26 & 9 & 5.8 & 4.8 & 2.4 + 11 dec 2001 & x2.8 & 0.65 & 8 & 2 & 0.8 & 0.9 & 0.3 + 20 may 2002 & x2.1 & 0.91 & 2 & - & 0.2 & - & - + 21 aug 2002 & x1.0 & 0.72 & 19 & 6 & 5.9 & 3.8 & 3.0 + 17 mar 2003 & x1.5 & 0.59 & 70 & 26 & 28.6 & 17.6 & 12.5 + 18 mar 2003 & x1.5 & 0.71 & 113 & 31 & 24.2 & 10.7 & 6.6 + 27 may 2003 & x1.3 & 0.36 & 72 & 11 & 11.2 & 3.8 & 2.4 + 28 may 2003 & x3.6 & 0.38 & 84 & 17 & 12.1 & 6.1 & 3.0 + 10 jun 2003 & x1.3 & 0.66 & 96 & 14 & 11.7 & 3.7 & 1.7 + 11 jun 2003 & x1.6 & 0.85 & 77 & 11 & 14.5 & 3.2 & 1.6 + 19 oct 2003 & x1.1 & 0.81 & 14 & - & 4.5 & - & - + 26 oct 2003 & x1.2a & 0.73 & 84 & 17 & 25.1 & 16.4 & 6.0 + 26 oct 2003 & x1.2b & 0.62 & 127 & 16 & 33.3 & 15.2 & 4.0 + 29 oct 2003 & x10.0 & 0.34 & 153 & 37 & 40.4 & 25.6 & 14.4 + 02 nov 2003 & x8.3 & 0.85 & - & - & - & - & - + 26 feb 2004 & x1.1 & 0.43 & 19 & 2 & 2.7 & 0.8 & 0.4 + 15 jul 2004 & x1.6 & 0.72 & 44 & 6 & 12.4 & 2.8 & 1.5 + 15 jul 2004 & x1.8 & 0.79 & 2 & - & 0.2 & - & - + 16 jul 2004 & x1.1 & 0.56 & 59 & 10 & 14.0 & 2.4 & 1.4 + 16 jul 2004 & x1.3 & 0.63 & 51 & 15 & 10.6 & 6.4 & 3.2 + 16 jul 2004 & x3.6 & 0.56 & 71 & 16 & 18.7 & 6.0 & 3.7 + 13 aug 2004 & x1.0 & 0.49 & 63 & 16 & 4.9 & 1.9 & 1.4 + 30 oct 2004 & x1.2 & 0.40 & 29 & 4 & 2.2 & 0.4 & 0.1 + 01 jan 2005 & x1.7 & 0.56 & - & - & - & - & - + 15 jan 2005 & x1.2 & 0.32 & 194 & 35 & 16.0 & 7.5 & 2.7 + 15 jan 2005 & x2.6 & 0.30 & 297 & 95 & 28.3 & 19.2 & 7.8 + 17 jan 2005 & x3.8 & 0.46 & 203 & 54 & 73.2 & 31.2 & 17.4 + 20 jan 2005 & x7.1 & 0.88 & 7 & - & 7.6 & - & - + 30 jul 2005 & x1.3 & 0.86 & - & - & - & - & - + 10 sep 2005 & x1.1 & 0.75 & 39 & 3 & 9.8 & 3.2 & 0.6 + 10 sep 2005 & x2.1 & 0.73 & 43 & 8 & 12.5 & 4.7 & 1.9 + 13 sep 2005 & x1.5 & 0.32 & 120 & 15 & 9.5 & 2.2 & 1.1 + 15 sep 2005 & x1.1 & 0.40 & 90 & 17 & 16.3 & 5.6 & 3.7 + 06 dec 2006 & x6.5 & 0.89 & 60 & 10 & 44.1 & 14.4 & 8.0 + 14 dec 2006 & x1.5 & 0.75 & 48 & 8 & 7.5 & 4.5 & 1.8 + the total number of cancellation events and the estimates of the flux losses during the whole 4-h interval around each flare of m- and x - classes are listed in tables 7 and 8 , respectively . ordering of all the estimates obtained with the three methods for each flare is consistent with that of . the third technique is the most restrictive , and thus should be the most reliable , showing a real physical pattern . in the active regions over 60 min before ( blue ) , 60 min after ( red ) , and during ( green ) a flare in multi - step events in which flux cancellation was observed over at least two consecutive magnetograms . the duration of flares varies from 8 to 57 min . panels ( a ) and ( b ) show the results for m- and x - class flares , respectively . in each panel , the upper graph shows the sum of minimum flux loss for each pair of magnetic features in each event ; the lower graph shows the average flux loss rate [ in mx min@xmath45 ( the amount of lost flux divided by the observed time interval ) . flux losses are plotted in ascending order of flare class . the inset plots show the total flux loss over the entire time interval around each flare ( one bar per flare ) . ] we also compute the amount of canceled magnetic flux separately for three time intervals : before the flare onset , during the flare , and after the end of the flare . the flare start , peak , and end times are taken from goes data . by this division of our data sets by three time intervals , we end up with different durations of these intervals for each flare . for a fair comparison , we restrict the time intervals before and after the flare to 60 min each . if a flare data set is shorter in at least one of the two time intervals , we exclude the flare from comparison , thus obtaining a final inclusion of 29 x- and 29 m - class flares . the duration of the flare is different for all flares and varies from 8 to 57 min . note that the ` during - flare ' interval is always shorter than 60 min . flux losses for 29 x- and 29 m - class flares obtained by the third , most restrictive , technique for estimating flux cancellation are shown in figure 5 . we compute mean and median values of flux losses in flux cancellation events before , during , and after the flare for both the amount of flux loss and the average flux loss rate . the results are shown in table 9 . the amount of flux canceled was greater over the time during the flare than before and after the flare for x - class flares . the ratios of the flux losses before , during , and after the flare to the amount of flux canceled before the flare are 1:3:0.7 ( median values ) and 1:4:0.5 ( mean values ) . some of the strongest x - class flares showed significantly greater flux loss during the flare time . the flux loss before the flare was greater than that after the flare by a smaller factor . for the m - class flares no significant trend was found . @ l r @ * 3c @ & before flare & during flare & after flare + median flux loss & flux loss [ @xmath46 mx ] + & m - class flares & 5.2 & 1.8 & 6.6 + & x - class flares & 6.1 & 7.1 & 4.4 + & flux loss rate [ @xmath46 mx min@xmath45 + & m - class flares & 0.09 & 0.1 & 0.1 + & x - class flares & 0.1 & 0.3 & 0.07 + mean flux loss & flux loss [ @xmath46 mx ] + & m - class flares & 5.5 & 6.5 & 9.4 + & x - class flares & 12.9 & 22.3 & 6.3 + & flux loss rate [ @xmath46 mx min@xmath45 + & m - class flares & 0.09 & 0.2 & 0.16 + & x - class flares & 0.2 & 0.8 & 0.1 + the flux losses in figure 5 are plotted in ascending order of flare class . we note an overall increasing trend of the flux losses as a function of flare power , which is more obvious for x - class flares than for m - class flares . this is clear from the inset plots of figure 5 , where each bar represents flux loss over the entire 4-h interval around each flare ( from the last columns of tables 7 and 8) and is plotted in the ascending order of flare class . can the correlation between the flux loss and the total flux in the canceling features or in the active region account for this trend ? figure 6 shows , as a function of the goes peak x - ray flux , the amount of magnetic flux loss , total magnetic flux in the active region , the number of canceling magnetic features , and the total magnetic flux in the canceling features , respectively . the pearson correlation coefficients for all the quantities are summarized in table 10 . the increasing trend as a function of the peak x - ray flux is dominated by x - class flares , as can be seen in figure 6 for the flux loss and the total flux in the active region but is less obvious for the number of canceling features and the total flux in canceling features . we find that all four quantities show weak - to - moderate correlation with peak x - ray flux at disk - center locations , and the correlation is dominated by x - class flares . the exception is the number of canceling features , which shows only weak correlation with the peak x - ray flux for disk - center flares and no statistically significant correlation for the disk - center x - class flares ( table 10 under ` disk center ' and ` disk - center x - class ' headings ) . the correlation is highest for the flux loss , which also shows moderate correlation for the limb flares dominated by the x - class flares ( the ` flux loss ' column of table 10 ) . the m - class flares revealed no correlation with the peak x - ray flux in any of the categories . previous studies found that the total magnetic flux in the active region and the magnetic flux at the vicinity of the polarity inversion line ( pil ) correlate with flare activity @xcite . the latter was found to best correlate with the peak x - ray flux @xcite . we find about the same level of low - to - moderate correlation , dominated by x - class flares at disk - center locations , between the peak x - ray flux and both the magnetic flux in the canceling features and the total flux in the active region . here , only some of the pil features contribute to the multi - step flux cancellation events ; thus the flux in the canceling features and the flux at the pil are not exactly the same quantities . the relation between the flux in pil features and the flare strength will be discussed in the next section . we have explored how the amount of magnetic flux loss , the total magnetic flux in the active region , the number of canceling features , and the total flux in the canceling features correlate with each other . the results are shown in table 11 . we find that both the number of canceling features and the total flux in the canceling features show moderate positive correlation with the flux loss in all categories . the exceptions are limb flares , particularly limb x - class flares , which did not show any correlation with the flux loss . however , some moderate correlation is found for the limb m - class flares ( see the last two columns in table 11 under ` limb m - class ' heading ) . the correlation for the flux in the canceling features is dominated by the disk - center x - class flares . in contrast , the total flux in the active region shows weak - to - moderate correlation only with the flux loss for the disk - center flares and with contribution only of the x - class flares . we also note that the flux loss is approximately equally correlated with the number of canceling features and the total flux in these features ( see the last two columns in table 11 ) , indicating that the field strength and complexity are approximately equally important . @ l * 4c @ & flux loss & total flux in & number of & total flux in + & & active region & canceling features & canceling features + all & 0.66 ( @xmath47 ) & 0.07 ( 0.57 ) & 0.08 ( 0.56 ) & 0.09 ( 0.49 ) + x - class & 0.66 ( @xmath48 ) & 0.16 ( 0.32 ) & 0.01 ( 0.95 ) & 0.10 ( 0.60 ) + m - class & 0.12 ( 0.58 ) & 0.18 ( 0.29 ) & -0.02 ( 0.93 ) & -0.02 ( 0.93 ) + disk - center & 0.74 ( @xmath49 ) & 0.35 ( @xmath50 ) & 0.47 ( @xmath51 ) & 0.56 ( @xmath52 ) + limb & 0.68 ( @xmath53 ) & -0.01 ( 0.94 ) & -0.19 ( 0.33 ) & -0.11 ( 0.56 ) + disk - center x - class & 0.75 ( @xmath54 ) & 0.61 ( @xmath34 ) & 0.39 ( 0.20 ) & 0.64 ( @xmath55 ) + disk - center m - class & 0.28 ( 0.30 ) & 0.27 ( 0.25 ) & 0.24 ( 0.37 ) & 0.22 ( 0.43 ) + limb x - class & 0.70 ( @xmath56 ) & 0.05 ( 0.81 ) & -0.22 ( 0.38 ) & -0.12 ( 0.63 ) + limb m - class & -0.14 ( 0.71 ) & 0.02 ( 0.92 ) & -0.44 ( 0.21 ) & -0.32 ( 0.38 ) + @ l * 4c @ flux loss & total flux in & number of & total flux in + & active region & canceling features & canceling features + all & 0.21 ( 0.11 ) & 0.47 ( @xmath57 ) & 0.50 ( @xmath33 ) + x - class & 0.29 ( 0.10 ) & 0.40 ( @xmath58 ) & 0.51 ( @xmath54 ) + m - class & 0.31 ( 0.12 ) & 0.70 ( @xmath59 ) & 0.60 ( @xmath60 ) + disk - center & 0.40 ( @xmath61 ) & 0.70 ( @xmath62 ) & 0.78 ( @xmath63 ) + limb & 0.004 ( 0.98 ) & 0.14 ( 0.47 ) & 0.21 ( 0.27 ) + disk - center x - class & 0.63 ( @xmath64 ) & 0.66 ( @xmath65 ) & 0.86 ( @xmath66 ) + disk - center m - class & 0.25 ( 0.36 ) & 0.69 ( @xmath54 ) & 0.58 ( @xmath55 ) + limb x - class & -0.10 ( 0.69 ) & 0.02 ( 0.93 ) & 0.15 ( 0.56 ) + limb m - class & 0.41 ( 0.25 ) & 0.69 ( @xmath31 ) & 0.61 ( 0.06 ) + thus , the level of correlation between the flux loss and the total flux in the canceling features is approximately as high as the correlation between the flux loss and the flare strength . therefore , it can account for the correlation between the flux loss and the goes peak x - ray flux we observe in general , and with the x - class flares at disk - center locations in particular . however , we find no correlation between the flux loss and the total flux in the canceling features ( or with the number of canceling features ) for the limb x - class flares ; thus it does not explain the correlation between the flux loss and the peak x - ray flux in this case . the correlation appears to work better for regions at disk - center locations than far from the disk center , suggesting that it is the radial flux loss that correlates best with the peak x - ray flux . we need vector magnetic field data to know the radial flux ( not the nearly horizontal field seen in the line - of - sight magnetograms near the limb ) to confirm this conjecture . measured the total magnetic flux ( @xmath12 ) near the pil and proposed this quantity as a measure for major flare potential rather than the total flux in an active region . in this work , the total flux in the pil features can be considered as an equivalent to the parameter @xmath12 . figure 7 shows the relationship between the total magnetic flux in the pil features against the total flux in the active region and the goes peak x - ray flux . the total magnetic flux in the pil features correlates with the total flux in the active region with the correlation coefficient of @xmath2 0.60 , which is in agreement with . figure 7(b ) indicates that large flares are associated with strong pil for the flares at disk - center locations , consistent with the finding of . such strong - gradient pils are generally associated with flux emergence and flux cancellation driven by photospheric flows . the net flux change in the pil features during its time evolution from before to after the flare shows that the pil flux decreases before the flare outnumber the increases in x - class flares , while m - class flares show an approximately equal number of flares with pil flux increases and decreases . this could hint that there is more canceling flux than emerging flux on pils before the flare . however , after the flare we find approximately equal numbers of pil flux decreases and increases in both x- and m - class flares . figure 8 shows , as a function of the radial position of the region on the solar disk , the amount of magnetic flux loss , the total flux in the active region , the number of canceling features , and the total flux in the canceling features . this figure suggests that the magnetic flux loss , the total flux in the active region , and total flux in the canceling features are not correlated with the radial position of the region on the disk . the magnetic fields related to these categories have no apparent directional preferences , whereas figures 4(c ) and 4(d ) show correlation between the flux changes and the radial position , consistent with nearly horizontal fields . on the other hand , figure 8(c ) does show correlation , indicating that the number of canceling features per active region is affected by foreshortening , decreasing towards the limb . vector magnetic field data are needed to clarify whether the spatial resolution in the observations is the reason for this tendency . the pearson correlation coefficients , summarized in table 12 , show weak anti - correlation between the total magnetic flux in the active region and its radial position . the number of canceling features indicates slightly higher , weak - to - moderate anti - correlation with the radial position of the region for the x - class flares only . the flux loss and the total flux in the canceling features show no statistically significant correlation with the radial position of the region . @ l * 4c @ & flux loss & total flux in & number of & total flux in + & & active region & canceling features & canceling features + all & 0.18 ( 0.19 ) & -0.23 ( @xmath67 ) & -0.18 ( 0.19 ) & -0.08 ( 0.53 ) + x - class & 0.10 ( 0.61 ) & -0.24 ( 0.14 ) & -0.43 ( @xmath50 ) & -0.18 ( 0.33 ) + m - class & 0.17 ( 0.40 ) & -0.21 ( 0.22 ) & 0.05 ( 0.80 ) & -0.02 ( 0.93 ) + cancellation events were not observed in all 77 flares . the absence of flux cancellation events generally corresponds to the active regions with flares far from the disk center ; here the longitudinal component of the magnetic field is believed to be reduced in comparison to the disk - center locations because the photospheric field is approximately radial . however , there are exceptional cases where the disk - center flares do not involve flux cancellation or the near - limb flares show significant flux loss . the m5.0 flare on 17 january 2004 that occurred at @xmath2 0.37 solar radii in noaa 10540 did not involve flux cancellation events . the active region was diffuse with the total flux in all magnetic features lower than the median value of the total flux of all active regions analyzed . the same active region hosted another flare in our analysis , m6.1 on 20 january 2004 , which also occurred at a near disk - center location ( @xmath2 0.32 solar radii ) and showed a very small amount of flux loss with no multi - step flux cancellations . the active region was located at the coordinates of noaa 10486 three rotations earlier , which was associated with halloween s flare . after two following rotations ( noaa 10508 and 10523 ) , the active region almost completely decayed . in another rotation , new flux emerged on the front side of the region and evolved into noaa 10540 , hosting a few c - class and m - class flares . three other flares ( an m9.1 flare on 8 november 2001 , an m9.1 flare on 22 july 2004 , and an m6.0 flare on 5 september 2001 ) that took place at near disk - center locations ( @xmath2 0.48 , 0.35 , and 0.49 solar radii ) did not involve multi - step flux cancellations . in the first flare , the active region had a diffuse structure and a short pil . in the last two flares , the region was concentrated and had large total magnetic flux , but the total flux in the pil features was lower than the median value of that for all the active regions analyzed . all active regions that hosted x - class flares in the near disk - center locations involved flux cancellation events . the absence of flux cancellation events around the x - class flares is generally seen in active regions with flares far from the disk center . however , we note the opposite cases in which some active regions with flares close to the limb locations involved significant flux cancellations during m- and x - class flares ( see tables 7 and 8 for details ) . they are the x20.0 flare on 2 april 2001 ( noaa 9393 ) , m5.7 on 7 november 2001 ( noaa 9690 ) , m7.1 on 26 december 2001 ( noaa 9742 ) , m6.8 on 20 december 2002 ( noaa 10226 ) , m7.6 on 26 october 2003 ( noaa 10484 ) , x1.5 on 17 march 2003 ( noaa 10314 ) , and x6.5 on 6 december 2006 ( noaa 10930 ) . all of the active regions had concentrated , essentially balanced flux of both polarities and long and tight ( strong ) pils . do the flux cancellation events correlate with stepwise flux changes during the flares ? the estimated amount of flux cancellation during some flares indicates a very large amount of flux loss ( see figure 5 ) . we select the set of canceling features that contributed to the flux loss over the entire 4-h interval around a flare and consider also the subset of these features whose flux loss occurred only over the duration of the flare . among these two sets of canceling features , we have counted the number of features which involve stepwise flux changes . we have compared 45 flares where at least one magnetic feature with stepwise flux change was found . table 13 shows that only a small portion of magnetic features involving flux cancellation showed stepwise flux changes ; thus we find no relationship between these two types of events . the lack of relationship between the stepwise field changes and the cancellation suggests that the photospheric cancellation process is independent of the abrupt flare - related field changes . @ l @ * 4c @ number & & + of features & during flare & total & during flare & total + stepwise flux change & 0 & 2 & 4 & 13 + no stepwise flux change & 19 & 47 & 37 & 79 + we have analyzed magnetic flux changes and studied flux cancellations related to 77 major flares . the results of this work can be summarized as follows . we found that significantly more magnetic flux decreases occur during flares than increases . this result is consistent with the model of magnetic implosion and collapsing loop structure proposed by . the change in the longitudinal magnetic field is larger in a region close to the limb , indicating that the involved field component is nearly horizontal . the x - class flares show larger flux changes and stronger bias towards flux decreases in comparison with m - class flares . both net and total flux changes show an increasing trend with the flare strength . we find that a linear correlation between the flux changes and the goes peak x - ray flux is dominated by x - class flares at limb locations , in agreement with the study of . we identified episodes of flux cancellation and estimated the amount of flux canceled during three intervals around each flare : before the flare onset , during the flare , and after the flare end . we found that the amount of canceled flux is generally greater over the time during the flare than before and after the flare . this tendency is mostly contributed from x - class flares . some of the strongest x - class flares showed significantly greater flux loss during the flare time . m - class flares did not show a significant bias in the amount of flux canceled during either time frame . the flux loss is approximately equally correlated with the number of canceling features and the total flux in these features . this leads us to conclude that the field strength and complexity are approximately equally important in the flux cancellation processes during flaring activity . the increasing trend of the flux losses as a function of the flare power is mostly explained by this correlation for the disk - center x - class flares , but not for the limb x - class flares . a less significant correlation between the flux losses and the total flux in the active region is also found for the flares at disk - center locations , dominated by the x - class flares . the flux loss is correlated with the peak x - ray flux for both disk - center and limb x - class flares . thus , the correlation between the flare strength and both the flux losses and the flux changes is dominated by x - class flares . we find that strong magnetic fields at pils are associated with the occurrence of large flares , in agreement with . the correlation with the flare strength is higher for the flux loss than for the total flux in canceling features . we did find contraction of the field towards the neutral line in the form of flux cancellation events , but we found no correlation between the flux cancellation and the abrupt , stepwise photospheric field changes associated with coronal field implosion during flares . we speculate that the stepwise changes and flux cancellation could be unrelated in general , as the stepwise flux changes may be caused by coronal magnetic restructuring while the flux cancellation may be caused by photospheric dynamics . for future study , vector magnetic field data are needed to better interpret the longitudinal field changes . high - cadence , high - sensitivity full vector data from the synoptic optical long - term investigations of the sun ( solis ) and the helioseismic and magnetic imager ( hmi ) will be very useful . also a combined analysis of chromospheric and coronal data is needed to clarify the following : are the canceling features part of the emerging u - loops or submerging @xmath68-loops ? are the cancellations associated with reconnection ? does the reconnection occur above or below the photosphere ? we thank brian t. welsch for providing the yafta feature - tracking code , fruitful discussions on this work , and useful comments . this work utilizes data obtained by the global oscillation network group ( gong ) program , managed by the national solar observatory , which is operated by aura , inc . under a cooperative agreement with the national science foundation . the data were acquired by instruments operated by the big bear solar observatory , high altitude observatory , learmonth solar observatory , udaipur solar observatory , instituto de astrofsica de canarias , and cerro tololo inter - american observatory . hudson , h. s. , fisher , g. h. , welsch , b. t. : 2008 , in : howe , r. , komm , r.w . , balasubramaniam , k.s . , petrie , g.j.d . ( eds . ) , _ subsurface and atmospheric influences on solar activity , asp conf . ser . _ * 383 * , 221 .

turbulence modeling is a critical component in numerical simulations of industrial flows based on reynolds - averaged navier - stokes ( rans ) equations . however , after decades of efforts in the turbulence modeling community , universally applicable rans models with predictive capabilities are still lacking . large discrepancies in the rans - modeled reynolds stresses are the main source that limits the predictive accuracy of rans models . identifying these discrepancies is of significance to possibly improve the rans modeling . in this work , we propose a data - driven , physics - informed machine learning approach for reconstructing discrepancies in rans modeled reynolds stresses . the discrepancies are formulated as functions of the mean flow features . by using a modern machine learning technique based on random forests , the discrepancy functions are trained by existing dns databases and then used to predict reynolds stress discrepancies in different flows where data are not available . the proposed method is evaluated by two classes of flows : ( 1 ) fully developed turbulent flows in a square duct at various reynolds numbers and ( 2 ) flows with massive separations . in separated flows , two training flow scenarios of increasing difficulties are considered : ( 1 ) the flow in the same periodic hills geometry yet at a lower reynolds number , and ( 2 ) the flow in a different hill geometry with a similar recirculation zone . excellent predictive performances were observed in both scenarios , demonstrating the merits of the proposed method . model - form uncertainty , turbulence modeling , reynolds - averaged navier stokes equations , boussinesq assumption , data - driven modeling , machine learning

1606.07987

computational fluid dynamics ( cfd ) simulations have been widely used in aerospace , mechanical , and chemical industries to support engineering design , analysis , and optimization . two decades ago when large eddy simulations ( les ) started gaining popularity with the increasing availability of computational resources , it was widely expected that les would gradually displace and eventually replace reynolds - averaged navier - stokes ( rans ) equations in industrial computational fluid dynamics ( cfd ) work - flows for decades to come . in the past two decades , however , while les - based methods ( including resolved les , wall - modeled les , and hybrid les / rans methods ) did gain widespread applications , and the earlier hope certainly did not diminish , the predicted time when these methods would replace rans has been significantly delayed . this observation is particularly relevant in light of the recent discussions on the ending of the `` moore s law era '' with transistor sizes approaching their theoretical lower limit @xcite . rans solvers , particularly those based on standard eddy viscosity models ( e.g. , @xmath0@xmath1 @xcite , @xmath0@xmath2 @xcite , s a @xcite , and k@xmath2sst @xcite , are still and will remain the dominant tool for industrial cfd in the near future . this is likely to be true even in mission critical applications such as aircraft design . interestingly , even the advanced rans models such as reynolds stress transport models @xcite and explicit algebraic reynolds stress models @xcite have not seen much development in the past few decades . these advanced models are computationally more expensive and less robust compared to the standard eddy viscosity rans models . as such , it is still practically important to further develop the standard rans models for industrial cfd applications . however , improving the predictive capabilities of these models is critical yet technically challenging . while traditional development of turbulence models has focused on incorporating more physics to improve predictive capabilities , an alternative approach is to utilize data . in the past few years , a number of data - driven approaches have been proposed . researchers have investigated the use of both offline data ( i.e. , existing dns data for flows different from that to be predicted @xcite ) and online data ( streamed monitoring data from the flow to be predicted @xcite ) . dow and wang @xcite used direct numerical simulation ( dns ) data from a plane channel flow to infer the full - field discrepancy in the turbulent viscosity @xmath3 modeled by the @xmath0@xmath2 model . to predict flows in channels with wavy boundaries , they modeled the ( log-)discrepancies of @xmath3 in the new flows as gaussian random fields , with the discrepancy field inferred above as mean . duraisamy and co - workers @xcite introduced a full - field multiplicative discrepancy term @xmath4 into the production term of the transport equations of turbulent quantities ( e.g. , @xmath5 in the sa model and @xmath2 in the @xmath0@xmath2 models ) . they used dns data to calibrate and infer uncertainties in the @xmath4 term . it is expected that the inferred discrepancy field can provide valuable insights to the development of turbulence model and can be used to improve rans predictions in similar flows . xiao et al . @xcite used sparse velocity measurements ( online data ) to infer the full - field discrepancies @xmath6 in the rans - predicted reynolds stress tensors , or more precisely the physical projections thereof ( turbulent kinetic energy , anisotropy , and orientations ) . throughout this paper it is understood that @xmath7 indicates the physical projections and not the individual components of the reynolds stress tensor . good performance was demonstrated on several canonical flows including flow past periodic hills , flow in a square duct @xcite , and flow past a wing body junction @xcite . all three approaches @xcite discussed above can be considered starting points toward the same destination : the capability of predictive turbulence modeling by using standard rans models in conjunction with offline data . to this end , the respective discrepancies terms ( @xmath8 , @xmath4 , and @xmath6 ) are expected to be extrapolated to similar yet different flows . these contributions are all relatively recent and much of the research is still on - going . duraisamy et al . @xcite performed _ a priori _ studies to show the potential universality of their discrepancy term @xmath4 among a class of similar flows , but their performances in _ a posteriori _ tests , i.e. , using the calibrated discrepancy in one flow to predict another flow , have yet to be demonstrated . dow and wang @xcite extrapolated the logarithmic discrepancies @xmath8 calibrated in the plane channel flow to flows in channel with slightly wavy walls , where velocity predictions were made . similarly , further pursuing the approach of xiao et al . @xcite , wu et al . @xcite showed that the reynolds stress discrepancy calibrated with sparse velocity data can be extrapolated to flows at reynolds number more than an order of magnitude higher than that in the calibration case . the extrapolated discrepancy has lead to markedly improved predictions of velocities and other quantities of interest ( qois ) , showing the potential of the approach in enabling data - driven predictive turbulence modeling . however , an intrinsic limitation in the approach of wu et al . @xcite is that they inferred the functions @xmath9 , or simply denoted as @xmath10 , in the space of _ physical coordinates _ therefore , strictly speaking they only demonstrated that the discrepancy @xmath6 can be extrapolated to flows in the same geometry at the same location . consequently , their attempts of extrapolation to the flow in a different geometry ( e.g. , from a square duct to a rectangular duct ) encountered less success . the approach of dow and wang @xcite would share the same limitation since they built gaussian random fields indexed by the physical coordinates @xmath11 . a natural extension that overcomes the key limitation in the calibration prediction approach of wu et al . @xcite is to build such functions in a space of well - chosen features @xmath12 instead of physical coordinates @xmath11 . despite its limitations , a key factor in the success of the original approach is that the reynolds stress discrepancies are formulated on its projections such as the anisotropy parameters ( @xmath13 and @xmath14 ) and orientation ( @xmath15 ) of the reynolds stresses and not directly on the individual components . these projections are normalized quantities @xcite . we shall retain this merits in the current approach and thus use data to construct functions @xmath16 instead of @xmath17 . this extension would allow the calibrated discrepancies to be extrapolated to a much wide range of flows . in other words , the discrepancies of the rans - predicted reynolds stresses can be quantitatively explained by the mean flow physics . hence , these discrepancies are likely to be universal quantities that can be extrapolated from one flow to another , at least among different flows sharing the same characteristics ( e.g. , separation ) . as such , discrepancies in reynolds stress projections are suitable targets to build functions for . with the function targets identified , two challenges remain : ( 1 ) to identify a set of mean flow features based on which the discrepancies functions @xmath16 can be constructed and ( 2 ) to choose a suitable method for constructing such functions . duraisamy and co - workers @xcite identified several features and used neural network to construct functions for the multiplicative discrepancy term . ling and templeton @xcite provided a richer and much more complete set of features in their pioneering work , and they evaluated several machine learning algorithms to predict point - based binary confidence indicators of rans models @xcite . ling et al . @xcite further used machine learning techniques to predict the reynolds stress anisotropy in jet - in - cross flows . based on the success demonstrated by ling and co - workers @xcite , we will use machine learning to construct the functions @xmath16 in the current work . specifically , we will examine a class of supervised machine learning techniques , where the objective of the learning is to build a statistical model from data and to make predictions on a response based on one or more inputs @xcite . this is in contrast to unsupervised learning , where no response is used in the training or prediction , and the objective is to understand the relationship and structure of the input data . unsupervised learning will be explored as an alternative approach in future works . the objective of this contribution is to present an approach to predict reynolds stress modeling discrepancies in new flows by utilizing data from flows with similar characteristics as the prediction flow . this is achieved by training regression functions of reynolds stress discrepancies with the dns database from the training flows . in light of the consensus in the turbulence modeling community that the reynolds stresses are the main source of model - form uncertainty in rans simulations @xcite , the current work aims to improve the rans modeled reynolds stresses . in multi - physics applications the qois might well be the reynolds stresses and/or quantities that directly depend thereon . in these applications the current work is significant by itself in that it would enable the use of standard rans models in conjunction with an offline database to provide accurate reynolds stress predictions . moreover , the improvement of reynolds stresses enabled by the proposed method is an important step towards a data - driven turbulence modeling framework . however , the reynolds stresses corrected by the constructed discrepancy function from dns databases can not necessarily guarantee to obtain improved mean flow fields . there are a number of challenges associated with propagating the improvement of reynold stresses through rans equation to the mean velocity field , which will be addressed in future works . the rest of this paper is organized as follows . section 2 introduces the components of the predictive framework , including the choice of regression inputs and responses as well as the machine learning technique used to build the regression function . section 3 shows the numerical results to demonstrate the merits of the proposed method . further interpretation of the feature importance and its implications to turbulence model development are discussed in section 4 . finally , section 5 concludes the paper . the overarching goal of the current and companion works is a physics - informed machine learning ( piml ) framework for predictive turbulence modeling . here , `` physics - informed '' is to emphasize the attempt of accounting for the physical domain knowledge in every stage of machine learning . the problem targeted by the piml framework can be formulated as follows : given high - fidelity data ( e.g. , reynolds stresses from dns or resolved les ) from a set @xmath18 of @xmath19 training flows , the framework shall allow for using standard rans turbulence models to predict a new flow @xmath20 for which data are not available . the flows @xmath21 for which high - fidelity simulation data are available are referred to as _ training flows _ , and the flow @xmath20 to be predicted is referred to as _ test flow_. the lack of data in test flows is typical in industrial cfd simulations performed to support design and optimization . furthermore , we assume that the training flows and the test flow have similar complexities and are dominated by the same characteristics such as separation or shock boundary layer interaction . this scenario is common in the engineering design process , where the test flow is closely related to the training flows . ultimately , the envisioned machine learning framework will be used in scenarios where the training flows consist of a wide range of elementary and complex flows with various characteristics and the test flow has a subset or all of them . however , the latter scenario is much more challenging and is outside the scope of the current study . considering the proposed method is a completely new paradigm , we decide to take small steps by starting from the closely related flows and to achieve the overarching goal gradually . in the proposed approach we utilize training data to construct functions of the discrepancies ( compared to the dns data ) in the rans - predicted reynolds stresses and use these functions to predict reynolds stresses in new flows . the procedure is summarized as follows : 1 . perform baseline rans simulations on both the training flows and the test flow . 2 . compute the feature vector field @xmath22 , e.g. , pressure gradient and streamline curvature , based on the rans - predicted mean flow fields for all flows . 3 . compute the discrepancies field @xmath23 in the rans modeled reynolds stresses for the training flows based on the high - fidelity data . 4 . construct regression functions @xmath24 for the discrepancies based on the training data prepared in step 3 . compute the reynolds stress discrepancies for the test flow by querying the regression functions . the reynolds stresses can subsequently be obtained by correcting the baseline rans predictions with the evaluated discrepancies . in machine learning terminology the discrepancies @xmath6 here are referred to as _ responses _ or _ targets _ , the feature vector @xmath12 as _ input _ , and the mappings @xmath25 as _ regression functions_. a regression function @xmath26 maps the input feature vector @xmath12 to the response @xmath27 , and the term `` function '' shall be interpreted in a broad sense here . that is , depending on the regression technique used , it can be either deterministic ( e.g. , for linear regression ) or random ( e.g. , gaussian process ) @xcite , and it may not even have an explicit form . in the case of random forests regression used in this work @xcite , the mapping does not have an explicit expression but is determined based on a number of decision trees . in the procedure described above , after the baseline rans simulations in step 1 , the input feature fields are computed in step 2 , the training data are prepared in step 3 , and the regression functions are constructed in step 4 . finally , the regression functions are evaluated to make predictions in step 5 . it is worth noting that in each stage domain knowledge is incorporated , e.g. , physical reasoning for identification of input features and consideration of realizability constraints of reynolds stress in learning - prediction process . each component is discussed in detail below . the choice of features and responses are presented in sections [ sec : feature ] and [ sec : discrepancy ] , respectively , and the machine learning algorithm chosen to build the regression function is introduced in section [ sec : regression ] . as has been pointed out in section [ sec : intro ] , mean flow features are better suited as input of the regression function than physical coordinates as they allow the constructed functions to predict flows in different geometries . ling and templeton @xcite proposed a rich set of twelve features based on clear physical reasoning . the set of features used in the present study mostly follow their work , except that we excluded the feature `` vortex stretching '' ( input 8 in table ii of ref . @xcite ) . this feature is present only in three - dimensional flows , but the test cases presented here are two - dimensional flow . we excluded two additional features related to linear and nonlinear eddy viscosities ( features 6 and 12 in ref . these quantities were specifically chosen for evaluating qualitative confidence indicators of rans predictions and , in our opinion , are not suitable input for regression functions of reynolds discrepancies . finally , experiences in the turbulence modeling communities suggest that mean streamline curvature has important influences on the predictive performance of rans models @xcite . therefore , curvature is included as an additional feature . the complete list of the mean flow features chosen as regression inputs in this work is summarized in table [ tab : feature ] . in choosing the mean flow features as regression inputs , we have observed a few principles in general . first , the input and thus the obtained regression functions should be galilean - invariant . quantities that satisfy this requirement include all scalars and the invariants ( e.g. , norms ) of vectors and tensors . an interesting example is the pressure gradient along streamline ( see feature @xmath28 in table [ tab : feature ] ) . while neither velocity @xmath29 nor pressure gradient @xmath30 ( both being vectors ) is galilean - invariant by itself and thus is not a suitable input , their inner product @xmath31 is second , since the truth of the mean flow fields in the test flows are not available , an input should solely utilize information of the mean flow field produced by the rans simulations @xcite . therefore , all of the adopted features are based on rans - predicted pressure @xmath32 , velocity @xmath33 , turbulent kinetic energy @xmath0 , and distance @xmath34 to the nearest wall . finally , to facilitate implementation and avoid ambiguity , only local quantities ( i.e. , cell- or point - based quantities in cfd solvers ) of the flow field are used in the formulation of features , with the distance @xmath34 to nearest wall being a notable exception . this principle is similar to that in choosing variables for developing turbulence models @xcite . the interpretation of most feature variables are evident from the brief descriptions given in the table , but a few need further discussions . first , feature @xmath35 ( q - criterion ) is based on the positive second invariant @xmath36 of the mean velocity gradient @xmath37 , which represents excess rotation rate relative to strain rate @xcite . for incompressible flows , it can be computed as @xmath38 , where @xmath39 and @xmath40 are rotation rate and strain rate tensors , respectively ; @xmath41 and @xmath42 , with superscript @xmath43 indicating tensor transpose and tr indicating trace . the q - criterion is widely used in cfd simulations as a post - processing tool to identify vortex structures for the visualization of flow structures @xcite . second , the wall distance based reynolds number @xmath44 in @xmath45 is an indicator to distinguish boundary layers from shear flows . this is an important feature because rans models behave very differently in the two types of flows . this quantity is frequently used in wall functions for turbulence models . third , feature @xmath46 defines the deviation from orthogonality between the velocity and its gradient @xcite , which indicates the deviation of the flow from parallel shear flows ( e.g. , plane channel flows ) . since most rans models are calibrated to yield good performance on parallel shear flows , this deviation is usually correlated well with large discrepancies in rans predictions . however , since it only accounts for misalignment angle and not the velocity magnitude , in regions with near - zero velocities this quantity becomes the angle formed by two zero - length vectors and is thus mostly noise . finally , we remark that most of the features in table [ tab : feature ] including the q - criterion and wall - distance based reynolds number are familiar to cfd practitioners . most of the features presented in table [ tab : feature ] are formulated as ratios of two quantities of the same dimension , either explicitly ( @xmath35 , @xmath47 , @xmath48 , @xmath49 , @xmath50 ) or implicitly ( @xmath51 and @xmath45 ) . hence , they are non - dimensional by construction . features @xmath28 and @xmath46 involve inner product of vectors or tensors , and thus they are normalized by the magnitude of the constituent vectors or tensors . finally , feature @xmath52 ( streamline curvature ) is normalized by @xmath53 , where @xmath54 is the characteristics length scale of the mean flow , chosen to be the hill height @xmath55 ( see fig . [ fig : flowgeo ] ) in the numerical examples . we followed the procedure of ling and templeton @xcite to normalize the features . except for feature @xmath45 , each element @xmath56 in the input vector @xmath12 is normalized as : @xmath57 where the summation on repeated indices is not implied , @xmath58 are raw values of the features , and @xmath59 are the corresponding normalization factors . this normalization scheme limits the numerical range of the inputs within @xmath60 $ ] and thus facilitates the regression . the normalization is not needed for feature @xmath45 ( wall distance based reynolds number ) since it is already in a non - dimensional form and in a limited range @xmath61 $ ] . it can be seen that the choices of features and normalization factors heavily rely on physical understanding of the problem ( turbulence modeling ) . that is , in the present data - driven modeling framework , the data are utilized only _ after _ physical reasoning from the modeler has been applied . this task can be a burden in certain applications . it is worth noting that the recent work of ling et al . @xcite aimed to relieve the modeler from such burdens by using a basis of invariants of tensors relevant in the specific application ( e.g. , strain rate @xmath40 in turbulence modeling ) . their work has the potential to systematically construct the input features based on raw physical variables and thus makes data - driven modeling even `` smarter '' . p2.0 cm | p5.0 cm p3.0 cm p4.0 cm feature ( @xmath56 ) & description & raw feature ( @xmath58 ) & normalization factor ( @xmath62 ) + @xmath35 & ratio of excess rotation rate to strain rate ( q - criterion ) & @xmath63 & @xmath64 + @xmath51 & turbulence intensity & @xmath0 & @xmath65 + @xmath45 & wall - distance based reynolds number & @xmath66 & not applicable@xmath67 + @xmath28 & pressure gradient along streamline & @xmath68 & @xmath69 + @xmath47 & ratio of turbulent time scale to mean strain time scale & @xmath70 & @xmath71 + @xmath48 & ratio of pressure normal stresses to shear stresses & @xmath72 & @xmath73 + @xmath46 & non - orthogonality between velocity and its gradient @xcite & @xmath74 & @xmath75 + @xmath49 & ratio of convection to production of tke & @xmath76 & @xmath77 + @xmath50 & ratio of total to normal reynolds stresses & @xmath78 & @xmath0 + @xmath52 & streamline curvature & @xmath79 where @xmath80 , @xmath81 & @xmath82 + in section [ sec : motivation ] the reynolds stress discrepancies @xmath83 have been identified as the responses for the regression functions . the response quantities should also be based on galilean invariant quantities due to the same consideration as in the feature choice . as such , individual components of the reynolds stresses or the discrepancies based thereon are not suitable , but those based on their eigenvalues or invariants are preferred . in turbulence modeling , the lumley triangle has been widely used for the analysis of turbulence states related to realizability @xcite . it is formulated based on the second and third invariants ( @xmath84 and @xmath85 ) of the anisotropy tensor . recently , banerjee et al . @xcite proposed an improved formulation in which the eigenvalues of the anisotropy tensor are mapped to barycentric coordinates as opposed to the variants @xmath84 and @xmath85 as in the lumley triangle . an important advantage of their formulation is that the mapping to barycentric coordinates is linear , which is in contrast to the nonlinear mapping to invariants @xmath84 and @xmath85 . therefore , barycentric coordinate provides a non - distorted visual representation of anisotropy and is easier to impose realizability constraints . the formulation of discrepancy starts with the eigen - decomposition of the reynolds stress anisotropy tensor @xmath86 : @xmath87 where @xmath0 is the turbulent kinetic energy , which indicates the magnitude of @xmath88 ; @xmath89 is the second order identity tensor ; @xmath90 $ ] and @xmath91 $ ] with @xmath92 are the orthonormal eigenvectors and eigenvalues of @xmath86 , respectively , indicating its shape and orientation . in the barycentric triangle , the eigenvalues @xmath93 , @xmath94 , and @xmath95 are mapped to the barycentric coordinates @xmath96 as follows : [ eq : lambda2c ] @xmath97 with @xmath98 . as shown in fig . [ fig : bary ] , the barycentric coordinates of a point indicate the portion of areas of three sub - triangles formed by the point and with edges of barycentric triangle . for example , a point located on the top vertex corresponds to @xmath99 while a point located on the bottom edge has @xmath100 = 0 . similar to the lumley triangle , all realizable turbulences are enclosed in the barycentric triangle ( or on its edges ) and have positive barycentric coordinates @xmath101 , @xmath102 , and @xmath100 . the barycentric triangle has been used by emory et al . @xcite as a mechanism to impose realizability of reynolds stresses in estimating uncertainties in rans simulations . placing the triangle in a cartesian coordinate system @xmath103 , the location of any point within the triangle is a convex combination of those of the three vertices , i.e. , @xmath104 where @xmath105 , @xmath106 , and @xmath107 denote coordinates of the three vertices of the triangle . an advantage of representing the anisotropy of reynolds stress in the barycentric coordinates is that it has a clear physical interpretation , i.e. , the dimensionality of the turbulence state @xcite . typically , the standard - rans - predicted reynolds stress at a near wall location is located close to the isotropic , three - component state ( vertex 3c - i ) in the barycentric triangle , while the truth is near the two - component limiting states ( bottom edge ) . moreover , the spatial variations from the near - wall region to the shear flow region are indicated as arrows in fig . [ fig : bary ] . it is clear that the trend of spatial variation predicted by a standard rans model is opposite to that of the truth . the three mutually orthogonal , unit - length eigenvectors @xmath108 , @xmath109 , and @xmath110 indicate the orientation of the anisotropy tensor . they can be considered a rigid body and thus its orientation has three degrees of freedom , although they have nine elements in total . we use the euler angle with the @xmath111@xmath112@xmath113 convention to parameterize the orientation following the convention in rigid body dynamics @xcite . that is , if a local coordinate system @xmath114@xmath115@xmath111 spanned by the three eigenvectors of @xmath116 was initially aligned with the global coordinate system ( @xmath117@xmath118@xmath119 ) , the current configuration could be obtained by the following three consecutive intrinsic rotations about the axes of the local coordinate system : ( 1 ) a rotation about the @xmath111 axis by angle @xmath120 , ( 2 ) a rotation about the @xmath114 axis by @xmath121 , and ( 3 ) another rotation about its @xmath111 axis by @xmath122 . the local coordinate axes usually change orientations after each rotation . in summary , the reynolds stress tensor is projected to six physically interpretable , galilean invariant quantities representing the magnitude ( @xmath0 ) , shape ( @xmath13 , @xmath14 ) , and orientation ( @xmath120 , @xmath121 , @xmath122 ) . they are collectively denoted as @xmath7 . the truths of these quantities can be written as baseline rans predictions corrected by the corresponding discrepancy terms , i.e. , [ eq : delta ] @xmath123 the discrepancies ( @xmath124 , @xmath125 , @xmath126 , @xmath127 , @xmath128 , @xmath129 , denoted as @xmath130 with @xmath131 ) in the six projections of the reynolds stress tensor are responses of the regression functions . we utilize data consisting of pairs of @xmath132 from training flow(s ) to construct the functions @xmath133 . it is assumed that the discrepancies in six quantities @xmath6 are independent , and thus separate functions are built for each of them . this simplification is along the same lines as that made in previous works @xcite . with the input ( mean flow features @xmath12 ) and responses ( reynolds stress discrepancies @xmath134 ) identified above , a method is needed to construct regression functions from training data and to make predictions based on these functions . supervised machine learning consists of a wide variety of such methods including k - nearest neighbors @xcite , linear regression and its variants ( e.g. , lasso ) @xcite , gaussian processes @xcite , tree - based methods ( decision trees , random forests , bagging ) @xcite , neural networks @xcite , and support vector machine @xcite , among others . a major consideration in choosing the regression method is the high dimensionality of the feature space , which is typically ten or higher in our application . the curse of dimensionality makes such methods as k - nearest neighbors , linear regression , and gaussian processes not suitable . secondary considerations , which we believe are also important for turbulence modeling applications , are the capability to provide predictions with quantified uncertainties as well as physical insights ( e.g. , on the importance of each of the features and their interactions ) . after evaluating a number of existing machine learning techniques in light of these criteria , we identified _ random forests _ @xcite as the optimal approach for our purposes , which is an ensemble learning technique based on decision trees . in simple decision tree learning , a tree - like model is built to predict the response variable by learning simple if - then - else decision rules from the training data . decision trees have the advantage of being easy to interpret ( e.g. , via visualization ) and implement . they are also computationally cheap . however , they tend to overfit the data and lack robustness . that is , a small change in the training data can result in large changes in the built model and its predictions . random forests learning is an ensemble learning technique proposed by ho @xcite and briemann @xcite that overcomes these shortcomings of simple decision trees . since these techniques are generally not familiar to readers in the fluid dynamics community , here we use an illustrative example in the context of turbulence modeling to explain the algorithm . a simple decision tree model is illustrated in fig . [ fig : tree - schematic ] . for clarity we consider an input with only two features : pressure gradient @xmath135 ( normalized and projected to the streamline tangential ) and wall distance based reynolds number @xmath136 , as defined in table [ tab : feature ] . it can be also interpreted as wall distance in viscous unit . the response is the discrepancy @xmath126 of the vertical coordinate in the barycentric triangle of the rans - predicted reynolds stress ( see fig . [ fig : bary ] ) . during the training process , the feature space is _ successively _ divided into a number of boxes ( leafs ) based on the training data ( shown as points in the @xmath135@xmath137 plane in fig . [ fig : tree - schematic]a ) . in the simplest decision tree model used for regression , the feature space is stratified with the objective of minimizing the total in - leaf variances of the responses at each step , a strategy that is referred to as greedy algorithm . after the stratification , a constant prediction model is built on each leaf . when predicting the response @xmath126 for a given feature vector @xmath12 , the constructed tree model in fig . [ fig : tree - schematic]b is traversed to identify the leaf where @xmath12 is located , and the mean response on the leaf is taken as the prediction @xmath138 . the tree model has a clear physical interpretation in the context of turbulence modeling . for example , it is well known that a standard isotropic eddy viscosity model has the largest discrepancy when predicting anisotropy in the viscous sublayer ( @xmath139 ) . this is because the truth is located on the bottom , corresponding to a combination of one- or two - component turbulence , while a typical isotropic eddy viscosity model would predict an isotropic state located on the top vertex ( see fig . [ fig : bary ] ) . in contrast , far away from the wall within the outer layer ( @xmath140 ) , the rans - predicted anisotropy is rather satisfactory . therefore , the first two branches divide the space to three regions based on the feature @xmath137 : outer layer ( region 1 ) , viscous sublayer ( region 2 ) , and buffer layer ( regions 3 and 4 ) . in the buffer layer the pressure gradient plays a more important role than in the outer and viscous layers . larger pressure gradients correspond to larger discrepancies , which can be explained by the fact that favorable pressure gradients ( negative @xmath135 values ) tend to thicken the viscous sublayer @xcite , which leads to larger discrepancies in @xmath14 . therefore , a further division splits the buffer layer states to two regions in the feature space , i.e. , those with strong ( region 3 ) and mild ( region 4 ) pressure gradients . a simple regression tree model described above tends to overfit for high dimensional input space , i.e. , yielding models that explain the training data very well but predict poorly for unseen data . in general the decision trees do not have the same level of predictive accuracy as other modern regression methods . however , by aggregating a large number of trees ( ideally with minimum correlation ) , the predictive performance can be significantly improved and , the overfitting can be largely avoided . in random forests an ensemble of trees is built with bootstrap samples ( i.e. , sampling with replacement ) drawn from the training data @xcite . moreover , when building each tree , it utilizes only a subset of @xmath141 randomly chosen features among the @xmath142 features , which reduces the correlation among the trees in the ensemble and thus decreases the bias of the ensemble prediction . random forest regression is a modern machine learning method with predictive performance comparable to other state - of - the - art techniques @xcite . in decision tree models the maximum depth of trees must be limited ( e.g. , by pruning the branches far from the root ) to ensure a sufficient number of training point ( e.g. , 5 ) on each node . in contrast , in random forests , one can build each tree to its maximum depth by successive splitting the nodes until only one training data point remains on each leaf . while each individual tree built in this manner may suffer from overfitting and has large prediction variances , the use of ensemble largely avoids both problems . moreover , random forest regression is simple to use with only two free parameters , i.e. , the number @xmath143 of trees in the ensemble and the number @xmath144 of selected features . in this work we used an ensemble of @xmath145 trees and the a subset of features ( i.e. , @xmath146 ) to build each tree . as a standard practice in statistical modeling , we performed cross - validations to optimize these parameters and performed sensitivity analysis to ensure that the predictions are not sensitive to the parameter choices . almost all industrial flows involve some characteristics ( e.g. , strong pressure gradient , streamline curvature , and separation ) that break the equilibrium assumption of rans model . therefore , we have these challenges in mind when developing the data - driven approach . in this study , we focus on the cases where training and test flows have similar characteristics . specifically , we evaluate the proposed method on two classes of flows : ( 1 ) fully developed turbulent flows in a square duct at various reynolds numbers and ( 2 ) flows with massive separations . the flow in a square duct at reynolds number @xmath147 and the flow in a channel with periodic hills at reynolds number @xmath148 are chosen as the prediction ( test ) flows for the respective flow classes . the square duct flow has an in - plane secondary flow pattern induced by the normal stress imbalance , while the periodic - hill flow features a recirculation bubble , non - parallel shear layer and mean flow curvature . all these characteristics are known to pose challenges for rans based turbulence models , and thus large model - form discrepancies exist in the rans - modeled reynolds stresses . in the two test flows , the relative importance of reynolds stress projections to the mean flow prediction are different . the reynolds stress anisotropy plays an important role in obtaining the accurate secondary mean motion in the duct flow @xcite . in contrast , the anisotropy is less important to predict the mean flow in the periodic - hill case , where the turbulent shear stress component is more essential to obtain an accurate mean velocity field @xcite . therefore , we use these two types of flows to highlight the improvements in the different reynolds stress components that are important for the predictions of qois in the respective flow classes . in both cases , all rans simulations are performed in an open - source cfd platform , openfoam , using a built - in incompressible flow solver ` simplefoam ` @xcite . mesh convergence studies have been performed . the fully developed turbulent flow in a square duct is a challenging case for rans - based turbulence models , since the secondary mean motion can not be captured by linear eddy viscosity models ( e.g. , @xmath0@xmath1 , @xmath0@xmath2 ) , and even the reynolds stress transport models ( rstm ) can not predict it well @xcite . in this test , we aim to improve the rans - modeled reynolds stresses of the duct flow at reynold number @xmath147 by using the proposed piml approach . the training data are obtained from dns simulations @xcite of the duct flows in the same geometry but at lower reynolds numbers @xmath149 and @xmath150 . the dns data of the prediction flow ( @xmath147 ) are reserved for comparison and are not used for training . coordinate represents the streamwise direction . secondary flows induced by reynolds stress imbalance exist in the @xmath115@xmath111 plane . panel ( b ) shows that the computational domain covers a quarter of the cross - section of the physical domain . this is due to the symmetry of the mean flow in both @xmath115 and @xmath111 directions as shown in panel ( c).,scaledwidth=80.0% ] the geometry of this flow case is shown in fig . [ fig : domain_duct ] . the reynolds number is based on the edge length d of the square duct and the bulk velocity @xmath151 . all lengths presented below are normalized by @xmath152 . the baseline rans simulations are performed for all training and test flows . the purpose is twofold : to obtain the mean flow feature fields @xmath22 as inputs and to obtain the discrepancies of reynolds stress by comparing with the dns data . to enable the comparison , the high - fidelity data are interpolated onto the mesh of rans simulation . the launder - gibson rstm @xcite is adopted to perform the baseline simulations , since all the linear eddy viscosity models are not able to capture the mean flow features of the secondary motions . the @xmath153 of the first cell center is kept less than 1 and thus no wall model is applied . as indicated in fig . [ fig : domain_duct ] , only one quadrant of the physical domain is simulated owing to the symmetry of the mean flow with respect to the centerlines along @xmath115- and @xmath111-axes . no - slip boundary condition is applied on the walls , and symmetry boundary condition is applied on the symmetry planes . we first investigate the prediction performance on the reynolds stress anisotropy tensor , since its accuracy is important to capture the secondary flow . figure [ fig : bayre ] shows piml - corrected anisotropy in barycentric triangle compared with baseline and dns results . the comparisons are performed on two representative lines at @xmath154 and @xmath155 on the in - plane cross section ( fig . [ fig : domain_duct]b ) . the two lines are indicated in the insets on the upper left corner of each panel . the arrows denote the order of sample points plotted in the triangle , which is from the bottom wall to the outer layer . the general trends of spatial variations of the dns reynolds stress anisotropies are similar on both lines . that is , from the wall to the outer layer , the reynolds stress starts from the two - component limiting states ( bottom edge of the triangle ) toward three - component anisotropic states ( middle area of the triangle ) . this trend is captured by the baseline rstm to some extent , especially in the regions away from the wall . however , significant discrepancies still can be observed in the near - wall region . very close to the wall , the dns reynolds stress is nearly the two - component limiting state . this is because the velocity fluctuations in the wall - normal direction are suppressed by the blocking of the bottom wall . moreover , before approaching three - component anisotropic states , the dns - predicted anisotropy first moves toward the one - component state ( 1c ) as away from the wall . in contrast , the rans - predicted anisotropy near the wall is closer to the two - component isotropic state ( 2c - i ) , and it approaches toward the three - component anisotropic state directly . therefore , in the near - wall region there are large discrepancies between the rans predicted reynolds stress anisotropy and the dns result , particularly in the horizontal coordinate @xmath13 . by correcting the baseline rstm results with the trained discrepancy function , the predicted anisotropy of reynolds stress is significantly improved . for both lines , the predicted anisotropy ( circles ) agrees well with the dns results ( squares ) . especially on the line @xmath156 , the piml - predicted anisotropy is almost identical to the dns data . significant improvement of the piml - predicted anisotropy can be seen from the barycentric maps shown in fig . [ fig : bayre ] . similar improvements have also been demonstrated in the other physical projections ( tke and orientations ) of the piml - corrected reynolds stresses . therefore , it is expected that the reynolds stress tensor components should be also improved over the rstm baseline . and ( b ) @xmath157 of corrected reynolds stress with the discrepancy model . the profiles are shown at four streamwise locations @xmath158 and @xmath159 . corresponding dns and baseline ( rstm ) results are also plotted for comparison.,title="fig:",scaledwidth=40.0% ] + in the six tensor components , two normal stress components @xmath160 and @xmath157 are among the most important ones to the mean velocity field since the normal stress imbalance ( @xmath161 ) is the driving force of the secondary flow @xcite . figures [ fig : tau_duct]a and [ fig : tau_duct]b show the profiles of normal components @xmath160 and @xmath157 of the reynolds stress reconstructed from the piml - corrected physical projections . corresponding baseline ( rstm ) and dns results are also plotted for comparison . both @xmath160 and @xmath157 are overestimated by the rstm over the entire domain . the discrepancy of the rstm - predicted @xmath160 is large near the wall and decreases when moving away from the wall . in contrast , @xmath157 is significantly overestimated far from the wall but the discrepancy decreases toward the wall . as a results , the rstm - predicted normal stress imbalance is pronouncedly inaccurate , which leads to unreliable secondary mean flow motion . as expected , the piml predictions nearly overlap with the dns results for both @xmath160 and @xmath157 and show considerable improvements over the rstm baseline . in fact , the improvements are observed in all the tensor components , which are omitted here for brevity . the results shown above demonstrate excellent performance of the proposed piml framework by using rstm as the baseline . the turbulent flow in a channel with periodic hills is another challenging case for rans models due to the massive flow separations in leeward of the hill . here , we examine two training scenarios with increasing difficulty levels . in the first scenario the training flows have the same geometry as the test ( prediction ) flow but are different in reynolds numbers . in the second scenario the training flows differ from the prediction case not only in reynolds numbers but also in geometry . p2.5 cm | p8.0 cm | p4.5 cm training flow scenario & training flow & symbol & high fidelity data + & periodic hills , re = 1400 ( ph1400 ) & dns by breuer et al . @xcite + & periodic hills , re = 5600 ( ph5600 ) & dns by breuer et al . @xcite + & wavy channel , re = 360 ( wc360 ) & dns by maa et al . @xcite + & curved backward facing step , re = 13200 ( cs13200 ) & les by bentaleb et al . @xcite + four training flows with dns / les data to build random forest regressors are summarized in table [ tab : database ] . in the first scenario two flows ph1400 and ph5600 are used for training , both of which are flows over periodic hills ( same in geometry ) at @xmath162 and @xmath163 ( different in reynolds numbers ) , respectively . for the second scenario , the training data are obtained from two different flows : one in a channel with a wavy bottom wall at @xmath164 and one over a curved backward facing step at @xmath165 , indicated as flows wc360 and cs13200 , respectively . + + a schematic of the flow geometry and rans - predicted velocity contour for each case are presented in fig . [ fig : flowgeo ] . the dimensions of each case are normalized with the respective hill heights @xmath55 . although the geometries of the training flows are different , all three flows share a similar characteristic as the test flow , i.e. , separation on the leeward side of the hill or step . however , the separation bubbles are different in sizes and shapes . the flow over periodic hills has a stronger separation compared to the other two due to the steeper slope of the hill . relatively mild separation can be observed in the flow over the wavy channel . for all cases , both high - fidelity data and rans - predicted results are available . the high - fidelity data are obtained from dns or resolved les simulations , which have been reported in literature ( see references in table . [ tab : database ] ) . the dns data for flows ph1400 and ph5600 are only available on vertical lines at eight streamwise locations @xmath166 . on the other hand , full - field high - fidelity results are available for flows wc360 and cs13200 , but only the lower part of the channel is adequately resolved . since the separated flow is of interest in this study , only the data in the separation region ( i.e. , region below @xmath167 ) are included . in this test , the performance of the proposed piml framework is evaluated on standard rans models . specifically , the baseline rans predictions are obtained by using the two - equation launder - sharma @xmath0@xmath1 model @xcite . the reason for choosing standard turbulence models here is because of two considerations . first , the standard rans models are the dominant tools for industrial cfd applications , while other sophisticated rans models have been rarely used . therefore , it is more significant to improve the widely used standard rans models . second , we understand that improvement of the reynolds stresses starting from a standard rans model is challenging . nonetheless , this challenging scenario also can better explore the capability of machine learning approach . the functional forms of discrepancies in the six physical projections of reynolds stress are learned from the training flows as mentioned in section [ sec : result : separation : setup ] and are used to correct the rans - predicted reynolds stress field of the test flow ( ph10595 ) . however , since the baseline rans model used in this case is the standard eddy viscosity model , the reynolds stress anisotropy can not be accurately predicted . therefore , the baseline rans - predicted anisotropy is unphysical and is significantly different from the dns result ( see fig . [ fig : bary ] ) . nonetheless , after the correction by using the discrepancy function learned from the training flows , the anisotropy of the test flow shows an excellent agreement with the dns results @xcite . the improvements are observed in the both training scenarios i and ii , demonstrating that the discrepancy function even in the standard rans - predicted anisotropy does exist and can be learned from the closely related flows based on the mean flow features @xmath12 . as mentioned above , in the periodic - hill flow , the correctness of reynolds stress anisotropy is of little consequence to the prediction of the mean velocity , and the correct shear stress component and magnitude of the reynolds stress are most important to obtain an accurate mean flow field . therefore , the anisotropy prediction results are omitted here , and only the turbulence kinetic energy ( tke ) and shear stress component of the piml - corrected reynolds stress are presented and discussed in detail . . corresponding baseline and dns results are also plotted for comparison . the hill profile is vertically exaggerated by a factor of 1.3.,scaledwidth=80.0% ] the comparison of the tke profiles of the baseline , dns , and piml - predicted results in the training scenario i are shown in fig . [ fig : kre ] . the tke predicted by the baseline rans model has notable discrepancies compared to the dns result , particularly in the region with non - parallel free shear flow ( @xmath168 to @xmath169 ) . the poor performance of rans model in such region is typical in these flows @xcite . the rans model underestimates the turbulence intensity along the free shear at @xmath170 , especially near the leeward side of the hill ( @xmath171 to @xmath172 ) . in the upper channel ( @xmath173 to @xmath174 ) , the dns tke is slightly smaller than the baseline rans prediction . the profiles of tke corrected by the piml - predicted discrepancy @xmath124 are significantly improved . the peaks along the streamwise free shear in the dns profiles are well captured in the corrected results with the random forest prediction . it can be seen that the predicted tke profiles ( solid lines ) nearly overlap with the dns results ( dashed lines ) . this clearly indicates that the tke discrepancies can be learned from the data of the training flows . . corresponding baseline and dns results are also plotted for comparison . the hill profile is vertically exaggerated by a factor of 1.3.,scaledwidth=80.0% ] it is also of interest to investigate the tensor components @xmath175 of reynolds stress , which are more relevant for predicting velocities and other qois of the flow fields . for the plane shear flows , the turbulent shear stress @xmath176 is important to predict the velocity field . figure [ fig : taure ] compares the turbulence shear component @xmath176 of predicted reynolds stress with the dns . as expected , significant improvements are observed compared to the baseline results , which underestimate the peak of @xmath176 on the leeward hill side but overestimate it on the windward hill side . as shown in fig . [ fig : taure ] , the profiles of predicted @xmath176 agree well with the dns results . the results above demonstrate that the discrepancy function of reynolds stress in its physical projections ( i.e. , magnitude , shape , and orientation ) trained from the flows at re = 2800 and 5600 can be used to predict the reynolds stress field of the flow at re = 10595 . significant improvements are observed in the predicted reynolds stress compared to the baseline rans results . although in this scenario the training and test flows are quite similar ( with the same geometry ) , and the success of extrapolation has been demonstrated in physical space by wu et al . @xcite , it should not be taken for granted that the accurate prediction is also guaranteed in feature space . since the regressions are performed in the ten - dimensional feature space and there is no direct reference to the physical coordinate , the success is not trivially expected _ a priori_. we investigate a more challenging scenario where the training flows have different geometries from the prediction case . this scenario is also more realistic in the context of using rans simulation to support engineering design and analysis . specifically , the data are more likely to be available for a few flows with specific reynolds numbers and geometries , but predictions are needed for the similar flows yet at different reynolds numbers and with modified geometries . . corresponding baseline and dns results are also plotted for comparison . the hill profile is vertically exaggerated by a factor of 2.4 . , scaledwidth=90.0% ] the comparison of the tke profiles on eight lines is shown in fig . [ fig : kgeo ] . note that only the domain below @xmath167 is investigated due to the lack of reliable high - fidelity training data in the upper channel region . this inadequacy of data quality can be exacerbated when the reynolds stress is decomposed to its physical projections . moreover , the flow separation is the phenomenon of concern in this study , and thus we only focus on the recirculation region . in fig . [ fig : kgeo ] , the random forest predicted tke ( solid lines ) is significantly improved over the baseline results ( dotted lines ) and better agrees with the dns profiles ( dash - dotted lines ) . the agreement is particularly good in the region from the center of recirculation bubble ( @xmath177 ) to the beginning of flow contraction ( @xmath178 ) . nonetheless , the piml - predicted tke does not show any improvement and even deteriorates compared to the baseline results near the windward side of the hill ( @xmath179 ) , where the flow starts to be contracted . as shown in fig . [ fig : kgeo ] , the predicted tke is markedly overestimated at @xmath180 . this is because the flow features in the contraction region ( @xmath179 ) are not supported in the training set , since the contracted flow does not exist in the training flow cs13200 ( fig . [ fig : flowgeo]c ) and is much weaker in the training flow wc360 ( fig . [ fig : flowgeo]b ) due to the mild slope of wavy bottom in this geometry . . corresponding baseline and dns results are also plotted for comparison . the hill profile is vertically exaggerated by a factor of 2.4 . , scaledwidth=90.0% ] finally , we compare the predicted turbulent shear stress @xmath176 with the dns profiles in fig . [ fig : taugeo ] . similar to the results of tke , the piml - predicted turbulent shear stress @xmath176 shows notable improvements in the recirculation region . however , deterioration occurs in the flow contraction region . at @xmath181 and @xmath182 , the magnitudes of turbulent shear stresses are overestimated with the correction based on the predicted discrepancies . this is consistent with the results observed in physical projections of reynolds stress . such a small region with abnormal reynolds stress corrections ( artificial peaks or bumps ) can introduce large errors to the velocity predictions . in general , the physical projections ( i.e. , magnitude , shape , and orientation ) of reynolds stress corrected by random forest predicted discrepancies are still significantly improved with the training flows in different geometries ( wc360 and cs13200 ) . the reynolds stress is markedly improved in the separated flow region , but not in the contracted flow region . this is because the features in training flows can not well support the predicted flow , and thus more extrapolations are expected . although the improvement is less significant compared to that in the scenario , the random forest predictions in this more realistic scenario are still satisfactory , demonstrating the merits of the proposed piml framework . in addition to the predictive capability of the regression model , it is also important to interpret the functional relation between the mean flow features and the discrepancies of the rans modeled reynolds stresses . for example , it is useful to find the most important features to the reynolds stress discrepancy in each of its physical projection ( i.e. , magnitude @xmath0 , shape @xmath183 , and orientation @xmath15 ) , and how each of these features impacts the regression response . identification of such correlation or causal relationship enables modelers to improve the rans turbulence models . the random forest regressor used in the proposed piml framework can also shed light on this issue by calculating the feature importance , which is a measure to evaluate the relative importance of a feature variable for predicting response variables @xcite . the bar plots of the importances of feature vector @xmath12 with respect to the discrepancies @xmath126 and @xmath184 are shown in figs . [ fig : import]a and [ fig : import]b , respectively . for discrepancy @xmath126 in the anisotropy , feature @xmath45 ( i.e. , wall - distance based reynolds number @xmath137 ) is the most important one . as discussed in section [ sec : feature ] , @xmath137 is the wall distance normalized by the approximate viscous unit . therefore , the result of feature importance is consistent with the piml prediction , which have shown that the discrepancy @xmath126 is notably dependent on the distance away from the wall @xcite . figure [ fig : import]b shows that the most important feature for predicting discrepancy @xmath184 of turbulence kinetic energy is feature @xmath51 , turbulence intensity . it is demonstrated that random forest used in the proposed framework can interpret the relationship between the features and the response to a certain extent , although the feature importance has its limitation due to bias introduced under certain conditions @xcite . in the machine learning community , improving interpretability of random forest is an active research topic , e.g. , several improvements of the importance measure have been proposed @xcite . moreover , in addition to calculating the feature importance , it is also helpful to examine the interactions among features , which have important implications for the interpretation of the regression models . as the base learner in random forest is a decision tree , which can capture the feature interactions , it is possible to further investigate the interacting relationship among mean flow variables . breiman et al . have studied feature interaction in random forest method @xcite , but more research is still ongoing . a better understanding of the physics behind the regression model for reynolds stress discrepancies has a profound implication to rans turbulence modeling . therefore , to explore the correlation or causal relationship between the mean flow features and the discrepancies of rans modeled reynolds stress is an important and promising extension of the proposed framework . the objective of the proposed framework is to improve the baseline rans - predicted reynolds stresses of a flow where high - fidelity ( e.g. , dns , les , experimental ) data are not available . the main novelty lies in using machine learning techniques to find the functional forms of reynolds stress discrepancies with respect to mean flow features by learning from the existing offline database of the closely related flows . numerical simulation results have demonstrated the feasibility and merits of the framework . moreover , the excellent performance of the piml predicted reynolds stress in not only the anisotropy but also in the tke and turbulent shear stress shows the fact that reynolds stress discrepancies can be extrapolated even to complex flows sharing similar characteristics . this finding is noteworthy by itself . the improvement of the rans - predicted reynolds stress is considered a viable and promising path toward obtaining better predictions of velocities and other quantities of interest . however , due to a few limitations of the current framework , the improvement of the propagated velocities from the corrected reynolds stress field can not be guaranteed . a small region with abnormal reynolds stress corrections ( e.g. , non - smoothness or artificial peaks ) can introduce large errors to the velocity predictions . for example , the small wave - number variations in reynolds stresses are visible in fig . [ fig : taugeo ] . these fluctuations , despite being small in amplitude , can lead to abnormal behaviors in the divergence term and thus in the predicted velocities . these abnormal predictions of reynolds stress corrections can be caused by several factors . first , the features in certain regions of the prediction flow may not be well supported in the training flows , e.g. the contraction region of periodic - hill flow mentioned in sec . [ sec : result : separation ] . second , the random forest regression used here only provides pointwise estimations but can not consider the spatial information of the reynolds stress field . therefore , the smoothness of the prediction can not be guaranteed . finally , although the input feature space is constructed based on physical reasoning , it is possible that the input features are not rich enough , and thus the randomness in the ensemble of the trained decision trees is significant . in this work , we proposed a physics - informed machine learning approach to reconstruct reynolds stresses modeling discrepancies by utilizing dns databases of training flows sharing similar characteristics as the flow to be predicted . for this purpose , we formulated discrepancy of reynolds stresses ( or more precisely its magnitude and the shape and orientation of the anisotropy ) as target functions of mean flow features and used modern machine learning techniques based on random forest regression to learn the functions . the obtained functions are then used to predict reynolds stress discrepancies in new flows . to evaluate the performance of the proposed approach , the method is tested by two classes of flows : ( 1 ) fully developed turbulent flows in a square duct at various reynolds numbers and ( 2 ) flows with massive separations . in the separated flows , two training flow scenarios of increasing difficulties are considered : in the less challenging scenario , data from two flows in the same periodic hill geometry at lower reynolds numbers ( @xmath185 and 5600 ) are used for training . in a more challenging scenario , the training data come from separated flows in different geometries ( wavy channel and curved backward facing step ) . in all test cases the corrected reynolds stresses are significantly improved compared to the baseline rans predictions , demonstrating the merits of the proposed approach . in the scenario , where the training flows and the prediction flow have different geometries , the improvement is not as drastic as in the the scenario with the same geometry . this is expected since the prediction involves more extrapolations in the feature space for this more challenging scenario . in other words , compared to the first scenario where the training and prediction flows have identical geometry , the prediction flow is less `` similar '' to the training flows in this scenario . the extent to which the training and prediction flows are `` similar '' to each other can be assessed _ a priori _ based on their respective rans predicted mean flow field , and methods for such assessment are presented in companion publications @xcite . as the inaccuracy in modeled reynolds stresses is the dominant source of model - form uncertainty in rans simulations , the proposed method for improving rans - predicted reynolds stresses is an important step towards the goal of enabling predictive capabilities of rans models . moreover , the random forests regression technique adopted in this work can provide physical insights regarding the relative importance of mean flow features that contributed to the discrepancies in the rans predicted reynolds stresses . this information can be used to assist future model development in that developers can devise models that are aware of and correctly respond to these flow features . however , a number of challenges need to be tackled before the improved reynolds stresses can be used to predict more accurate quantities of interests that are needed in engineering design ( e.g. , draft and lift coefficients ) . this topic will be investigated in future research . we thank dr . julia ling of sandia national laboratories and dr . eric g. paterson of virginia tech for helpful discussions during this work . we also thank the anonymous reviewers for their comments , which helped improving the quality and clarity of the manuscript . e. dow , q. wang , quantification of structural uncertainties in the @xmath0@xmath2 turbulence model , in : 52nd aiaa / asme / asce / ahs / asc structures , structural dynamics and materials conference , aiaa , denver , colorado , 2011 , aiaa paper , 2011 - 1762 . h. xiao , j .- l . wu , j .- x . wang , r. sun , c. roy , quantifying and reducing model - form uncertainties in reynolds - averaged navier stokes simulations : a data - driven , physics - informed bayesian approach , journal of computational physics 324 ( 2016 ) 115136 . c. gorle , m. emory , g. iaccarino , rans modeling of turbulent mixing for a jet in supersonic cross flow : model evaluation and uncertainty quantification , in : ichmt digital library online , begel house inc . , 2012 . f. billard , development of a robust elliptic - blending turbulence model for near - wall , separated and buoyant flows . , thesis , school of mechanical aerospace and civil engineering , the university of manchester . , manchester , uk ( 2011 ) . h. xiao , j .- l . wu , j .- x . wang , e. g. paterson , physics - informed machine learning for predictive turbulence modeling : progress and perspectives , in proceedings of the 2017 aiaa scitech , in press ( 2017 ) . a. dobra , j. gehrke , bias correction in classification tree construction , in : proceedings of the eighteenth international conference on machine learning , morgan kaufmann publishers inc . , 2001 , pp .

we explore the limits of photometric reductions of crowded stellar fields observed with the wide field and planetary camera 2 on board the _ hubble space telescope_. two photometric procedures , based on the dophot and daophot / allframe programs are tested , and the effects of crowding , complex sky background and cosmic - ray ( cr ) contamination are discussed using an extensive set of artificial star simulations . as a specific application of the results presented in this paper , we assess the magnitude of photometric biases on programs aimed at finding cepheids and determining distances . we find that while the photometry in individual images can be biased too bright by up to 0.2 mag in the most crowded fields due to confusion noise , the effects on distance measurements based on cepheid variables are insignificant , less than 0.02 mag ( 1% in distance ) even in the most problematic cases . this result , which is at odds with claims recently surfaced in the literature , is due to the strict criteria applied in the selection of the variable stars , and the photometric cross checks made possible by the availability of multiple exposures in different filters which characterizes cepheid observations . .2 in _ to appear in the february 2000 issue of the pasp _

astro-ph9911193

during its six years of operation , the wide field and planetary camera 2 ( wfpc2 ) on board the _ hubble space telescope _ ( _ hst _ ) has produced several thousand hours worth of observations of crowded stellar fields , from galactic globular clusters to distant galaxies . the photometric reduction of such fields is challenging . much effort has been spent in calibrating the photometric zero points and assessing the magnitude of non - linearity effects ( holtzmann 1995 , whitmore & hayer 1997 , hill 1998 , stetson 1998 , saha 2000 ) , but not enough attention has been devoted to understanding whether and to what extent biases are introduced by the procedures used for the photometric analysis . recently , mochejska ( 1999 ) and stanek & udalski ( 1999 ) addressed the effects of confusion noise on distance measurements based on cepheid variable stars , and concluded that for galaxies at 20 mpc observed with _ , distances are underestimated by up to 15% due to blending . these studies are based on a simple extrapolation to large distances of ground - based cepheid observations in m31 and lmc fields . unfortunately , they make two invalid assumptions : that the photometric reduction procedures are incapable of recognizing and correcting for the increasing level of crowding at larger distances , and that the stellar background in the m31 and lmc fields are representative of those of the more distant galaxies . a more direct assessment of the effects of crowding on the dophot ( schechter 1993 ) photometry of one particular galaxy , ngc 4639 , which was targeted for cepheids by saha ( 1997 ) using _ hst _ , is presented by saha ( 2000 ) . for this galaxy , which is at a distance of @xmath0 mpc , the authors find a 5% bias in distance due to crowding , much smaller than inferred by mochejska and stanek & udalski , but still significant . in this paper we test both dophot and daophot / allframe ( allframe from now on for simplicity , stetson 1994 ) photometric procedures using an extensive set of artificial star simulations of fields spanning a wide range of crowding and complexity of the underlying background . the analysis presented in this paper has specific applications to the results of the ` _ hst _ key project on the extragalactic distance scale ' ( hereafter key project , kennicutt , freedman , & mould 1995 ) , and the ` type ia supernovae calibration project ' ( e.g. , sandage et al . 1992 ) , but is also relevant to any dophot and allframe photometric reduction of crowded stellar fields observed with the wfpc2 . the construction of the artificial star frames is discussed in 2 , and the procedure used in the photometric analysis in 3 . the results are presented in 4 , 5 and 6 . section 4 describes control tests done on uncrowded , constant background artificial star fields , while 6 deals with the effects of confusion noise for both dophot and allframe . dophot has also been subjected to two additional test : section 5 dissects the effects of crs on the dophot photometry of crowded fields , and a simple test of the accuracy of the dophot photometric calibration is given in 7 . these tests were done for dophot and not allframe without implying that we expect the former to be more affected ; similar tests could also be carried out for allframe . finally , 8 presents a summary of results . of the fields targeted by the key project we selected two at opposite extremes in terms of complexity of the background , confusion from crowding and incidence of cr hits : the nearby galaxy ngc 2541 ( d@xmath112 mpc ) and ngc 1365 in the fornax cluster ( d@xmath119 mpc ) . most of the galaxies observed by the key project fall in between these two extremes in terms of sky brightness and crowding , both across the field and specifically in the regions where cepheids are found , with very few galaxies approaching the level of complexity presented by the ngc 1365 field . montages of the four wfpc2 ccds for each field ( from which crs have been removed ) are shown in figures 1 and 2 . for each galaxy , four out of the twelve existing epochs of observations were selected , each with pairs of back to back exposures ( referred to as a ` cr - split pairs ' ) in both the f555w and f814w filters . exposure times and other relevant information are given in table 1 . the artificial stars used to test both dophot and allframe are created using exclusively allframe parameters , in particular : 1 ) the average psfs constructed for the allframe analysis are assumed to be faithful reproductions of the true system psfs ; and 2 ) the allframe photometric calibration is used to transform the instrumental psf magnitudes to f555w and f814w magnitudes in the holtzmann ( 1995 ) ground based system ( see 3 ) . in the allframe procedure followed by the kp , the psf is constructed from wfpc2 images of uncrowded stellar fields , and is then used for the allframe photometric analysis of _ all _ of the observed galaxies . however , the wfpc2 psf changes with time due , for example , to focus changes and telescope jitter : these will lead to systematic variations in the profile - fitting magnitudes in frames that have true psfs very different from the mean psfs . in the allframe reduction stream , magnitude corrections are determined by measuring a number of hand - selected local standards in each field through a series of synthetic apertures up to one - half arcsecond in radius . these aperture magnitudes , projected to a fixed aperture radius of one - half arcsecond through a growth - curve analysis ( stetson 1990 ) are presumed to be insensitive to focus and tracking errors ; the mean difference between these aperture magnitudes and the profile - fitting magnitudes for a given frame are applied to the latter as an ` aperture correction ' . dophot takes the opposite approach : for each photometered frame , the psf is determined within the frame using suitable bright and isolated stars throughout the field , so while different frames will generally have different psfs , the aperture corrections are maintained relatively constant . adding stars described by the allframe psfs to an existing galaxy frame has the inevitable consequence that the artificial stars added to the images will by definition have the fiducial psfs and will not be subject to the same aperture corrections as the actual stars in those same images . therefore , the artificial star experiments described here are not effective in testing the validity of the aperture corrections or photometric zero points for either reduction procedures : by definition , the artificial star magnitudes recovered by allframe will agree _ in the mean _ with the input magnitudes , while the errors in both the dophot and allframe aperture corrections will limit the significance of any difference between the input and dophot recovered magnitudes . adopting an allframe psf for the artificial stars does not , however , affect our ability to study crowding - induced photometric scatter and scale errors . two thousand artificial stars with given right ascension , declination and @xmath2 and @xmath3 magnitudes were generated using the allframe psf and the luminosity functions shown by the solid black lines in figures 3 and 4 . figures 1 and 2 show images of the artificial star frames . poisson noise was added to the artificial stars , however no read noise or poisson noise was added to the sky since these noise sources are already present in the real galaxy frames . no artificial stars were placed close to the edges of the chips , which are vignetted by the pyramid mirror . the artificial stars were then added to each of the ngc 2541 and ngc 1365 frames listed in table 1 . in what follows we will refer to the frames containing only the artificial stars on a zero background as the ` artificial frames ' , to the original galaxy frames as the ` real frames ' and to the original galaxy frames to which the artificial stars have been added as the ` real+artificial frames ' . we will speak of an ` exposure ' in reference to a single image of a cr - split pair ( for example u2s72901 t , see table 1 ) , and of an ` epoch ' in reference to the image obtained by combining two cr - split exposures ( for example u2s72901 t and u2s72902 t , see 3.1 ) . ` input magnitudes ' are the ones assumed in constructing the artificial star frames , by definition they do not have an associated error . ` recovered magnitudes ' are the ones measured by running allframe or dophot on the frames ; they will differ from the input magnitudes due to the modeled photon statistics , the underlying read noise and sky noise in the real frames , and the effects of crowding . the real+artificial frames and the artificial frames were photometered by nas using allframe ( stetson 1994 ) and by lf using a variant of dophot especially formulated to deal with the undersampling of wfpc2 data ( schechter , mateo , & saha 1993 ; saha 1994 ) . neither nas nor lf had access to the artificial star list ( created by pbs ) until after the photometric analysis was complete . we followed the methodology of the key project ( e.g. , hill 1998 , ferrarese 1996 ) , except that only a subset of eight of the 24 key project exposures of each field ( four of 12 epochs ) , were used to derive the master star list . as usual , all images are multiplied by four and converted to integer before the photometric analysis is performed . therefore , the effective gain is 1.75 e@xmath4 , as opposed to 7 e@xmath4 of real wfpc2 data . each pair of back to back cr - split exposures was combined to obtain what we will be referring to as an ` epoch frame' and crs were removed prior to the dophot run . further details of the dophot procedure can be found in ferrarese ( 1998 ) and saha ( 1994 ) . table 2 lists the psf parameters ( fwhm along the major and minor axes and tilt angle ) reported by dophot for the real frames , the artificial star frames , and the real+artificial frames . the artificial stars luminosity function is more heavily weighted towards bright magnitudes , so while the psf fitted by dophot to the real frames is slightly different from the artificial star psf , the psf fitted to the real+artificial star frames is very close . the dophot aperture corrections needed to transform the fitted magnitudes to the 05 magnitudes defined by holtzmann ( 1995 ) were calculated independently for each photometered frame , following ferrarese ( 1998 ) . some of the dophot aperture corrections are listed in table 3 for the artificial star frames , and for the first exposure and first epoch ( the two exposures of the first cr - split pair combined , see table 1 ) of the real+artificial star frames . the number of stars used in deriving the aperture corrections is shown in parentheses in each case . because of jitter and focus changes , aperture corrections for other exposures / epochs differ from those of the first epoch and from each other by up to 0.05 mag . for comparison , the aperture corrections derived for an uncrowded field in leo i are also shown . notice that because the dophot aperture corrections are derived from a mixture of real and artificial stars , which ( for dophot ) are not described by an identical psf , the dophot aperture corrections might not be appropriate for either real or artificial stars , and any zero point comparison quoted in this paper is of very limited significance . finally , the aperture corrected magnitudes were transformed to the f555w and f814w magnitudes on the holtzmann ground based system following hill ( 1998 ) ; the final transformation to @xmath2 and @xmath3 magnitudes , which requires the addition of a color term ( the same for dophot and allframe , and up to 0.02 mag for f555w and 0.04 mag for f814w ) , is not necessary for our purpose . as customary , allframe was run on all eight exposures simultaneously . notice that for a typical galaxy observed by the kp , allframe would be run simultaneously on a larger sample of @xmath5 frames , therefore in the case discussed here allframe has a less than typical ability to recognize blended images . the allframe fitted psf is , of course , identical to the artificial psf by construction . the aperture corrections for the first and second exposures are as given in table 4 , notice that these are more recent than the ones used by ferrarese ( 1998 ) and silbermann ( 1999 ) for the original reduction of the galaxy frames . as for the dophot procedure , the aperture corrected magnitudes are transformed to f555w and f814w magnitudes following hill ( 1998 ) . a control experiment was done by running dophot and allframe on the uncrowded , zero background , cr - free artificial star frames constructed for ngc 1365 ( the ngc 2541 artificial frames will not be discussed as they do not differ from the ngc 1365 frames in any relevant way ) . the luminosity function of the recovered stars is shown in figure 3 by the thick red and green lines for allframe and dophot . only the pc and wf2 chips are shown as the results for the other two chips are comparable . notice that at faint magnitudes dophot is more complete than allframe in both chips , _ a posteriori _ we attribute this result to a more sensitive detection threshold ( the number of sigma above the background needed to trigger a detection ) adopted in the dophot run . figures 5 and 6 show a comparison of input magnitudes and the magnitudes recovered by allframe and dophot for the pc and wf2 ( the results for the other chips are similar ) . a weighted mean and standard deviation @xmath6 ( where the weights are given by the inverse of the dophot reported errors ) of the difference @xmath7 between input and recovered magnitudes is calculated by excluding all points deviating by more than 2@xmath6 from the mean , and iterating the process until convergence . this typically excludes 25% of the stars in f555w and 35% in f814w . stars fainter than 28 mag are also excluded . the results are listed in table 5 , where we also report the slope @xmath8/input mag ] and standard deviation derived from a least - squares fit to all data points . a non - zero @xmath9 would indicate the presence of non - linearities in the photometry : for example , if the sky were systematically overestimated , fainter stars would be affected more than bright ones , and @xmath9 would be negative . as expected , both dophot and allframe perform well on these simple frames . no non - linearity effects are seen . notice that the perfect agreement in magnitude between input and recovered magnitudes for allframe is artificial and results from having adopted the allframe photometric calibration for the artificial stars . in view of the fact that allframe aperture corrections for ngc 1365 are based only on a handful of stars ( table 4 ) , and the differences between the allframe and dophot psfs , especially for the pc ( table 2 ) , the agreement between input and recovered magnitudes for dophot is also good . the solid flaring curves in figures 5 and 6 represent the typical error reported by dophot and allframe for stars at any given magnitude . the percentage of stars with difference between input and measured magnitude larger than three times the reported @xmath6 is about 0.6% for dophot and 0.7% for allframe , which confirms the reliability of the error estimates as given by both photometric procedures . allframe and dophot take very different approaches when dealing with cr hits . because allframe runs simultaneously on all available exposures , information on the position and magnitudes of each object is carried from frame to frame and crs are easily flagged . dophot identifies crs as differing significantly from a stellar psf . objects in each frame are identified with the help of a master star list created from a deep , cr - free image obtained by combining all exposures . dophot parameters can be set so that objects identified in a single exposure but not present in the master star list will be classified as crs unless their psf is virtually identical to a stellar psf . this flags the majority of cr hits , but dophot can still err when crs hit the centers of stars included the master star list . for this reason , dophot is more reliable when applied to frames from which crs have been removed . this is an easy task for the key project galaxies since , with very few exceptions , cr - split pairs are available for each epoch . the cr - split exposures are combined before processing with dophot and crs are identified when the difference in counts between corresponding pixels ( after accounting for a global difference in the sky level ) is larger than four times a local sigma calculated from the combined effects of poisson statistics and local noise . this procedure poses some questions . because of the severe undersampling of the wfpc2 , particular care must be taken to assure that the peaks of bright stars are not erroneously identified as cr hits while combining the cr - split pairs . in addition , unidentified cr events could lead to an overestimate or an underestimate of the stellar magnitudes , depending on whether the hits fall on top of a stellar psf or in the nearby region used to measure the sky respectively . we have conducted the following test to assess the effects of cr events on the dophot photometry . for the real+artificial star frames of both ngc 2541 and ngc 1365 we ran dophot independently on the first and second exposure of the first epoch for each galaxy , which are heavily affected by crs , and on the cr - cleaned first epoch frame , produced by combining the first and second exposure . in all cases the master star list derived from all exposures combined was used . figure 7 shows a comparison of sky values measured in the first exposure and the first epoch for the f555w filter and wf2 chip , which for both galaxies shows the largest background excursions . the total counts ( dn ) collected for the sky are given , with the number of counts for the single exposure scaled to match the exposure time of the combined frame . two conclusions can be drawn . first , the mean sky difference at low background levels is not zero . inspection of the frames shows that this corresponds to a difference in the level of scattered light between the two consecutive frames due to the change in orbital position of the spacecraft . more importantly , there is a non - linear effect . a least - squares fit to all available data points gives @xmath10 for ngc 2541 , and @xmath11 for ngc 1365 , where @xmath12 is the difference in sky values between the first exposure and the first epoch of the cr - split pair ( scaled to the same exposure time ) , and @xmath13 is the sky value for the first epoch . the sense of the difference is that the sky is measured brighter in a single , cr - affected exposure , the more so the higher the background level . notice that the agreement in the slopes measured for the ngc 2541 and ngc 1365 frames ( which have exposure times differing by over a factor two ) implies that the effect depends on total number of counts in the sky rather than on sky surface brightness . responsible for this scale error is the presence , in the single exposures , of cr hits which are not flagged and therefore are folded in when calculating the background . because hits are more easily missed on brighter backgrounds , this produces a scale error in the sky measurements . in the epoch frame , the cr hits are correctly identified and removed , and the sky is measured accurately . we therefore recommend that cr - cleaned frames be used for dophot reductions whenever possible , as was done for the photometric reduction of the galaxies observed as part of the kp and the type ia supernovae calibration project . how does this affect the stellar magnitudes ? figure 8 shows the magnitude difference @xmath7 between the first exposure and first epoch of the wf2 cr - split pair : these are the stars whose background is plotted in figure 7 . given the slope of the scale error measured in the sky , and assuming that the stellar magnitudes are not affected by the presence of crs , two stars of 25th and 27th magnitude projecting onto a sky background of 20 mag arcsec@xmath14 ( corresponding to 200 and 480 dn in the bottom and top panel of figure 7 respectively ) would be measured too faint by 0.01 and 0.1 mag respectively . this would produce a scale error when comparing magnitudes recovered from frames with and without crs . no obvious scale error is however observed in figure 8 : a least square fit to the data points produces slopes @xmath8/input mag ] smaller than 0.01 mag / mag . several causes contribute to this result . first of all , faint stars are more easily lost on brighter backgrounds . figure 9 shows the recovery rate for stars of given magnitude as a function of backgrounds brightness . therefore , most of the stars that would show the largest deviations , and contribute more to produce a scale error ( i.e. faint stars on bright backgrounds ) are not even detected . second , just as the presence of crs leads dophot to overestimate the background brightness , it is reasonable to expect that the stars magnitudes will be somewhat overestimated as well , with the consequence that the two effects cancel , at least partially . support to this point comes from the fact that when the difference @xmath7 is plotted as a function of sky brightness , no obvious correlation is observed . figure 8 shows an increased scatter in the sense that stars are measured preferentially brighter in the single exposure than in the cr - free epoch frame . visual inspection of the images confirms that these ( the red crosses in figure 8) are unfortunate stellar images that happen to be hit head on by a cr . the cr is not identified in the single exposure , and the star is measured too bright . unfortunately , these stars can not be discriminated based on their reported error @xmath6 : for all of them @xmath15 and their reported photometric errors are perfectly normal for their measured magnitude . therefore , the bias would be undetected if no combined , cr - free images were available . is the problem solved in the case of the key project galaxies , for which dophot is run on cr - cleaned images ? not entirely . figure 10 shows the same red stars as in figure 8 but this time the values plotted are the magnitude difference between the first exposure and the first epoch ( in red as in figure 8) , between the second exposure and the first epoch ( in black ) , and between the second exposure and the deep frame created by combining all available exposures ( in blue ) . it can be readily seen that the blue points have a mean positive @xmath7 , while the black points are well distributed around a zero @xmath7 . we remind the reader that the stars plotted here were selected by the circumstance of having been affected by a cr in the first exposure ; in very few cases will the star also have been affected by a cr in the second exposure . the fact that the black points show systematically positive magnitude residuals ( by 0.10@xmath160.16 mag ) argues that the procedure used in combining the first and second exposures to produce the epoch frame must leave some of the cr hit behind , so that the star is still measured brighter in the combined frame than in the second exposure . however , when all eight available frames are combined to create the deep image , the process of eliminating cosmic rays is much more effective , and the bias disappears ( blue points ; the weighted mean in this case is 0.01@xmath160.17 ) . in conclusion , the simulations show that when dophot is applied following the kp prescriptions , about 3% of the stars at any given epoch are measured too bright by an average of 0.1 mag . for the kp galaxies , which are observed for a total of 12 epochs , this means that about 1/3 of the cepheids will have one out of twelve measurements biased high by 0.1 mag , which has no impact on the period - luminosity ( pl ) relation . we now turn our attention to the ngc 1365 and ngc 2541 frames to which the artificial stars have been added . from now on , dophot is always applied to frames which are cr - cleaned ( 3 ) , while allframe uses the original , cr - split exposures . the luminosity function of the recovered artificial stars is shown in figures 3 and 4 by the thin yellow and blue lines for the pc and wf2 chips and the allframe and dophot reductions . figures 11 and 12 show the comparison between input and recovered magnitudes for both allframe and dophot . the results do not vary significantly from chip to chip for the same filter , and we only plot them for the wf2 chip of both ngc 1365 and ngc 2541 . the same parameters listed in table 5 for the artificial frames , i.e. , the weighted mean difference between input and recovered magnitudes , and the slope of a least - squares fit line through the data , can be found in table 6 for the real+artificial frames . again , no significant scale errors are found , and the agreement in the zero points is within the uncertainties in the dophot and allframe aperture corrections ( see 3 ) . figures 11 and 12 , however , deserve some closer inspection . the distributions show larger scatter for ngc 1365 than for ngc 2541 , as expected due to the more crowded nature of the former . unlike the control case described in 4 , however , the number of points which deviate by more than three times the reported error does not follow a gaussian distribution . in ngc 1365 , 16% and 24% of the f555w and f814w dophot stars brighter than 28 mag deviate by more than @xmath17 , in ngc 2541 the percentages are 14% and 21% . the same result , but not quite as extreme has been discussed by saha ( 2000 ) , who further point out that this translates into a bias of up to 0.1 mag in the cepheid distance moduli derived using dophot in crowded fields . the plots based on allframe photometry have larger scatter , however , the photometric errors reported by allframe are larger than measured by dophot , with the consequence that a smaller fraction of the allframe stars , 8% and 5% in ngc 1365 and ngc 2541 respectively , deviate by more than @xmath17 . for all stars deviating by more than @xmath17 ( the circles in figures 11 and 12 ) we plot in figure 13 the difference between the dophot magnitudes measured in the first epoch and in all subsequent epochs . because the main features of this plot repeat for all chips and both filters , we only show the case of the wf2 f814w reduction in ngc 1365 . some of the points ( shown by the circles ) are highly correlated . the magnitudes for these stars are mis - measured in the first epoch because of some transient event , such as the presence of a nearby hot pixel , or an undetected cr , but are measured correctly in all other epochs . while there is really no way to identify these stars in single epoch programs , in multi - epoch observations ( such as cepheid finding programs ) the corrupted epoch would easily be flagged . the points shown by the crosses , however , show no obvious correlation : they scatter around a zero mean ( within the uncertainties in the aperture corrections ) and show no scale errors . these stars are measured systematically too bright in all of the epochs . the reason is confusion due to unresolved stellar blends or rapidly changing background level with position on the chip . because these patterns repeat identically from epoch to epoch , they produce a bias which is impossible to detect unless artificial star experiments , of the type described in this paper , are performed in each individual case . following saha ( 2000 ) we plot in figures 14 to 15 the correlations between the f555w and f814w residuals for the first epoch of both galaxies , and for both allframe and dophot . the results shown for the pc and wf2 are representative of the other chips . in each figure crosses are for stars that deviate by more than @xmath17 in both passbands , filled and open circles for stars that deviate by more than 3@xmath6 in either f555w or f814w , and dots for well behaved stars in both bands . in all cases , we notice that there is no correlation for stars which are not affected by confusion noise ( dots and circles ) . in figure 13 , these stars would be plotted as circles , i.e. , they happen to be measured incorrectly in one particular epoch and one particular filter due to some reason that does not repeat for all other epochs ( for example the vicinity of a crs hit , or defecting pixels ) , and would therefore not produce a bias in the derivation of cepheid distance moduli . however , there is a strong correlation for stars for which the difference between measured and expected magnitude is statistically larger _ in both passbands _ than expected given their reported photometric errors ( crosses ) . as pointed out above , the magnitude of these stars is overestimated in each epoch and both filters , and will introduce a bias in the derived cepheid pl relations . a quantitative discussion of the photometric biases arising from the effects described above when the photometry is performed on single frames is discussed in 6.1 . section 6.2 applies to the specific case of the cepheid observations carried out by the kp . the amount of bias from confusion noise in single - epoch observations can be quantified from plots such as the ones shown in figure 11 and 12 : the bias is simply represented by how much , in the mean , the recovered magnitudes differ from the input magnitudes . this is estimated by : @xmath18 defining a robust weighted mean @xmath19 of the difference between input and recovered magnitudes in the real+artificial frames as described in 4 . only bright , isolated , and therefore well measured stars contribute to @xmath20 , which therefore represents the ` true ' value of the mean for any sample of stars in the absence of biases ; @xmath18 computing a straight mean @xmath21 of all available data points recovered in the real+artificial frames ; @xmath18 calculating the difference @xmath22 . this represents the amount of bias for single epoch observations , and is listed in table 7 . we note that while for single epoch one band photometry of crowded fields allframe is more affected by confusion noise than dophot ( table 7 ) , allframe was specifically designed to take optimum advantage of the information content of multi - filter , multi - epoch observations , such as searches for variable stars . while allframe shows larger scatter than dophot in figures 11 and 12 , the deviant points have large photometric errors in allframe , which is not the case for dophot . while it would be difficult to discriminate against these stars in individual exposures ( apart from rejecting them on the basis of their large errorbars ) , the problem is lessened in most practical cases : photometric observations often involve multiple filters , for example , and for cepheids observations carried out using the kp procedure , the large photometric errors attached to the allframe magnitudes would prevent one from selecting most of the deviant stars as cepheid variables . for multi - epoch programs such as observations of cepheid variables , the bias discussed in 6.1 is somewhat lessened . this is because some of the deviant stars will be flagged once the photometry obtained for consecutive exposure or in different filters is compared . let us consider the case of the galaxies observed by the kp , whose goal is to discover cepheids and measure their period and mean magnitudes . observation in two filters and multiple epochs are available . first of all , we pointed out already that stars which deviate in _ only one _ of the filters by more than three times their reported error ( figures 14 and 15 ) are corrupted by a transient phenomenon ( for example a cosmic ray ) affecting that particular epoch , but none of the others . when comparing photometry from different epochs , the corrupted measurement would easily be flagged ( this would correspond , for example , to an obviously deviant point in an otherwise well phased light curve ) . we can therefore remove the corrupted measurements from our sample . second , fits to the cepheid pl relation are not weighted , however cepheids which deviate from the ridge - line of the pl relation by a significant amount ( generally three times the intrinsic 1@xmath6 width of the relation ) are rejected . the kp adopted @xmath23 mag in @xmath2 and @xmath24 mag in @xmath3 for the pl relation . when calculating the bias introduced by confusion noise in the cepheid sample we do the following ( the artificial star magnitudes in this case can be related to the cepheids mean magnitudes ) : @xmath18 include in the sample only measurements that deviate from the straight mean of all data points by more _ or _ less than three times their reported error in _ both _ f555w or f814w . this addresses the first point mentioned above . farther reduce the sample by excluding all stars which deviate from the mean by more than 0.81 mag in f555w and more than 0.54 mag in f814w . this addresses the second point above and flags only a few of the most extreme points . @xmath18 for this sample , calculate @xmath20 as done in the case of single - epoch observations . effectively , because @xmath20 is a robust weighted mean , the prior exclusion of the data satisfying the conditions above has no effects . @xmath18 for the same sample , calculate the mean difference @xmath21 between input and recovered magnitudes . given the distances to ngc 2541 and ngc 1365 , and the fact that the maximum period ( and therefore maximum mean magnitude ) of the cepheids is constrained by the length of the observing window ( 60 days ) , the range in @xmath2 magnitudes spanned by the cepheids is @xmath25 mag and @xmath26 mag in ngc 2541 and ngc 1365 respectively ( ferrarese 1998 , silbermann 1999 ) . only these magnitude ranges are therefore considered in calculating @xmath27 . notice that the exclusion of the brighter , least biased , part of the distribution in figures 11 and 12 , and the inclusion of the faint tail up to @xmath28 mag ( when in fact few cepheids ever get fainter than 27.5 mag ) leads to an increase in @xmath27 , compared to the case in which the entire wavelength range were to be used . @xmath18 finally , calculate the difference @xmath29 . this is the bias due to confusion noise expected in the _ apparent _ cepheid distance moduli , and is listed in table 8 . the bias in the _ final _ , dereddened distance modulus is given by @xmath30 when the magnitudes in both @xmath2 and @xmath3 passbands are mis - measured by approximately the same amount ( i.e. the correlations of residuals in figures 14 and 15 has a unit slope ) ( saha 2000 ) . from table 8 , the first conclusion is that even for the most crowded fields , i.e. wf2 of ngc 1365 , the bias is not larger than 0.07 - 0.08 mag , in agreement with the findings by saha ( 2000 ) in the very crowded field of ngc 4639 ( the authors estimate the bias to be less than 0.08 mag ) . in the specific example of ngc 1365 , for which silbermann ( 1999 ) derived a distance modulus of @xmath31 0.20 ( random ) @xmath16 0.18 ( systematic ) based on allframe photometry , the bias estimated from a mean of all four chips in table 8 would be @xmath32 mag , and can be easily neglected . in the less crowded fields of more nearby galaxies , such as ngc 2541 , the bias is further reduced . there is an additional factor which , while not easily quantifiable , will farther decrease the already insignificant amount of bias in the cepheid distance modulus . contamination by an underlying companion would artificially decrease the amplitude of the cepheid light curve , and either lead to a reduction of confidence in the light curve , or preclude detection as a variable star altogether . we have conducted experiments by adding increasing amount of contamination to real cepheid light curves observed by the key project team . while we could not discriminate against cepheids for which the contamination amounts to 30% or less of the cepheid mean flux , in most cases the cepheids were recognized as affected when the contamination level was increased beyond 60% . one last points needs to be discussed . the bias calculated above can be considered representative of the bias affecting the cepheid sample , and therefore the cepheid distance modulus , only if the artificial stars and the cepheids are affected by the same level of crowding . figure 16 shows the mean and rms contamination affecting the sample of cepheids ( solid circles and smaller errorbars ) and artificial stars ( open circles and larger errorbars ) as a function of f555w magnitudes . to calculate the quantity on the y - axis , we counted the number of real stars detected within a 5 pixel radius of each cepheid and artificial star . we then calculated the ratio between the total flux contributed by these nearby companions to the flux of the central star . we refer to this quantity , multiplied by 100 , as the ` percent contamination ' : its mean and standard deviation , binned in 0.5 magnitude intervals , are plotted as a function of the f555w magnitude of the central star in figure 16 , for both ngc 2541 and ngc 1365 . in making the plot , we used the real sample of cepheids from which a distance was derived ( from ferrarese 1998 and silbermann 1999 ) . so , for example , the total flux of the stars within a 5 pixel radius of the ngc 2541 cepheids with f555w magnitudes between 24.5 and 25.0 is equal , in the mean , to @xmath33% the flux of the cepheid . in the same magnitude range , @xmath34% for the artificial stars . it is easy to see from the figure that , if anything , the artificial stars are _ more _ contaminated than the cepheids . therefore the bias listed in table 8 can be considered as a hard upper limit to the amount of bias affecting the cepheid sample , and the derived cepheid distances . the results of the tests presented in this work are at odds with the finding of mochejska ( 1999 ) and stanek & udalski ( 1999 ) , who speculate from the analysis of cepheids in lmc and m31 fields that the cepheid distances published by the kp and the type ia supernovae calibration team are underestimated by up to 15% ( 0.3 mag ) due to neglecting the effects of blending . the mochejska and stanek & udalski analysis , while correctly assuming that unresolved blending will artificially increase a star s magnitude , do not consider that for the same reason the underlying sky will be brightened . the interplay of the two effects can not be estimated unless the photometry is actually _ performed on _ ( as opposed to extrapolated to ) the distant galaxy fields , using some kind of control test which in our case is provided by the artificial stars . in addition , mochejska and stanek & udalski assume that the stellar background in the lmc and m31 fields is representative of the more distant fields observed with hst , when in fact it is significantly brighter . the tests discussed so far say very little about the accuracy of the dophot or allframe zero points , since the artificial star magnitudes are based on the allframe photometric calibration . the dophot zero point was tested for the most crowded of the fields considered , the wf2 chip of ngc 1365 , using as ` artificial stars ' a real field observed in the dwarf local group galaxy sextans a. the sextans a images used were obtained as part of program go-5915 on december 1 , 1995 , for a total of 1800 seconds of integration in both f555w and f814w filters . based on 4 , we expect the magnitudes reported by dophot when run on this uncrowded field to be very accurate . a comparison between these magnitudes , and the magnitudes recovered when the sextans a field is added to the f555w wf2 ngc 1365 frame is shown in figure 17 . a weighted fit to all stars gives perfect agreement in the absolute zero points , @xmath35 , and no scale error . we have discussed the performance of the dophot and allframe photometric procedures when applied to crowded fields observed with the wfpc2 on board _ hst_. we have focused on searching for biases that can affect the determination of the cepheid distance moduli derived by the _ hst _ key project on the extragalactic distance scale and the type ia supernovae calibration project , but our results are relevant for any photometric study requiring high precision measurements . the following conclusions and guidelines are the results of this paper . @xmath18 when doing any photometric analysis , tests should be performed to asses the impact of cr events on the photometry . in the case of dophot , it is highly recommended to use frames from which crs have been removed . when crs are present , dophot has a tendency to overestimate the sky brightness by amounts that can be significant when total counts per pixel in the sky exceed a few hundred dn . the problem is completely solved only when several exposures are available and the photometry is obtained from an image created by combining all exposures . in the specific case of most of the galaxies observed as part of the _ hst _ key project , where only two exposures are available for each epoch of observation , the stars affected are different in each epoch and are not present in a large enough number to introduce a bias in the derived cepheid distance modulus . the effects of crs on the allframe photometry remain to be tested , however given the good agreement between the allframe and dophot photometry obtained for all kp galaxies ( better than 0.05 mag , or well within the uncertainties in the aperture corrections ) , it is unlikely that crs introduce a significant bias , if any . @xmath18 both dophot and allframe perform extremely well , both in recovering magnitudes and in estimating their errors , in uncrowded fields with limited variation in the background brightness across the chip . @xmath18 in crowded fields , such as the ones observed by the _ hst _ key project team , a bias is introduced by the presence of confusion noise due to crowding and rapidly varying background levels across the chip . this is in the sense that a fraction ( @xmath1 5 - 10% for allframe and up to 25% for dophot ) of the stars are measured consistently too bright . the effect on the photometry of single epoch observations can be significant , 0.05 mag even for not too crowded fields , such as the wf4 chip of ngc 2541 , and up to 0.2 mag for the most crowded and distant fields observed by the kp , such as ngc 1365 . the bias is somewhat worse when allframe , rather than dophot photometry is used . when the photometry is applied to derive cepheid distance moduli from multi - epoch observations , however , the bias is significantly reduced because strict criteria are imposed in selecting the variable stars . in the case of allframe , the bias in the final distance modulus is negligible , leading to underestimate the distance by 1% ( 0.02 mag ) at most for the most crowded and distant fields observed by the _ hst _ key project . the effects is slightly larger , @xmath1 2% mag , when dophot photometry is used , but is in no case as large as extrapolated by mochejska ( 1999 ) and stanek & udalski ( 1999 ) from ground based studies of m31 and lmc fields . we wish to thank the anonymous referee for the useful suggestions that helped improve the quality of this manuscript . lf acknowledges support by nasa through hubble fellowship grant hf-01081.01 - 96a and through long term space astrophysics program nra-98 - 03-ltsa-03 . we have concluded that the key project distance scale is compressed by photometric confusion problems by no more than 0.025 mag . it is interesting to consider whether there is any way of checking that @xmath36 . one possibility is to examine residuals from the tully fisher relation and to seek correlations with distance . if distances for remote galaxies are underestimated due to their higher probability of image blending ( stanek & udalski 1999 ) , those galaxies would lie low in the tully fisher diagram of sakai ( 2000 ) , as also pointed out by gibson ( 2000 ) . figure 18 shows h - band residuals from sakai s equation ( 10 ) plotted against distance . such correlation as there is is in the opposite sense . in fact , calculating the slope in figure 18 and its uncertainty , we can rule out the @xmath37 correlation one might expect due to blending at the 1.85@xmath6 level . this supports our conclusion that the key project distance scale is not compressed by photometric confusion problems by more than 0.025 mag , a conclusion also reached by gibson llrrr 1st v epoch , 1st exposure & u2s72901 t & 1100 & f555w & 30/10/951st v epoch , 2nd exposure & u2s72902 t & 1100 & f555w & 30/10/951st i epoch , 1st exposure & u2s72903 t & 1300 & f814w & 30/10/951st i epoch , 2nd exposure & u2s72904 t & 1300 & f814w & 30/10/952nd v epoch , 1st exposure & u2s73001 t & 1100 & f555w & 5/11/952nd v epoch , 2nd exposure & u2s73002 t & 1100 & f555w & 5/11/952nd i epoch , 1st exposure & u2s73003 t & 1300 & f814w & 5/11/952nd i epoch , 2nd exposure & u2s73004 t & 1300 & f814w & 5/11/953rd v epoch , 1st exposure & u2s73402 t & 900 & f555w & 20/11/953rd v epoch , 2nd exposure & u2s73403 t & 900 & f555w & 20/11/953rd i epoch , 1st exposure & u2s73405 t & 1100 & f814w & 20/11/953rd i epoch , 2nd exposure & u2s73406 t & 1100 & f814w & 20/11/954th v epoch , 1st exposure & u2s73901 t & 1100 & f555w & 8/12/954th v epoch , 2nd exposure & u2s73902 t & 1100 & f555w & 8/12/954th i epoch , 1st exposure & u2s73903 t & 1300 & f814w & 8/12/954th i epoch , 2nd exposure & u2s73904 t & 1300 & f814w & 8/12/951st v epoch , 1st exposure & u2s71201 t & 2400 & f555w & 19/09/951st v epoch , 2nd exposure & u2s71202 t & 2700 & f555w & 19/09/951st i epoch , 1st exposure & u2s71203 t & 2700 & f814w & 19/09/951st i epoch , 2nd exposure & u2s71204 t & 2700 & f814w & 19/09/952nd v epoch , 1st exposure & u2s70201 t & 2400 & f555w & 6/08/952nd v epoch , 2nd exposure & u2s70202 t & 2700 & f555w & 6/08/952nd i epoch , 1st exposure & u2s70203 t & 2700 & f814w & 6/08/952nd i epoch , 2nd exposure & u2s70204 t & 2700 & f814w & 6/08/953rd v epoch , 1st exposure & u2s70301 t & 2400 & f555w & 14/08/953rd v epoch , 2nd exposure & u2s70302 t & 2700 & f555w & 14/08/953rd i epoch , 1st exposure & u2s70303 t & 2700 & f814w & 14/08/953rd i epoch , 2nd exposure & u2s70304 t & 2700 & f814w & 14/08/954th v epoch , 1st exposure & u2s70701 t & 2400 & f555w & 29/08/954th v epoch , 2nd exposure & u2s70703 t & 2300 & f555w & 29/08/954th i epoch , 1st exposure & u2s70705 t & 2300 & f814w & 29/08/954th i epoch , 2nd exposure & u2s70706 t & 2700 & f814w & 29/08/95 lrrrrrrrrrrrr real frame & 1.382 & 1.318 & 3062 & 1.275 & 1.262 & @xmath382280 & 1.272 & 1.226 & 2116 & 1.260 & 1.140 & @xmath386844 artificial frame & 1.271 & 1.109 & 3771 & 1.174 & 1.130 & @xmath382081 & 1.328 & 1.224 & 3723 & 1.299 & 1.159 & @xmath386772 real+art . , 1st exp . & 1.257 & 1.138 & 3297 & 1.168 & 1.133 & @xmath384276 & & & & & & real+art . , 1st epoch & 1.274 & 1.163 & 3365 & 1.159 & 1.114 & @xmath383915 & 1.263 & 1.165 & 3170 & 1.251 & 1.120 & @xmath386835 real frame & 3.924 & 3.819 & 3914 & 1.291 & 1.226 & 8099 & 1.299 & 1.223 & 196 & 1.313 & 1.213 & @xmath388003 artificial frame & 3.987 & 3.837 & 4342 & 1.160 & 1.143 & 7578 & 1.246 & 1.161 & 3203 & 1.240 & 1.153 & @xmath387204 real+art . , 1st exp . & 3.810 & 3.734 & 4262 & 1.206 & 1.175 & @xmath387443 & & & & & & real+art . , 1st epoch & 3.856 & 3.761 & 4832 & 1.196 & 1.166 & 8876 & 1.217 & 1.154 & 3438 & 1.223 & 1.144 & @xmath387632 real frame & 1.198 & 1.118 & 3735 & 1.195 & 1.112 & @xmath383604 & 1.199 & 1.122 & 1737 & 1.226 & 1.074 & @xmath387042 artificial frame & 1.247 & 1.111 & 3905 & 1.171 & 1.123 & @xmath383070 & 1.329 & 1.221 & 3199 & 1.285 & 1.151 & @xmath386445 real+art . , & 1.268 & 1.127 & 3442 & 1.149 & 1.113 & @xmath382023 & & & & & & real+art . , 1st epoch & 1.273 & 1.144 & 3521 & 1.145 & 1.113 & @xmath382646 & 1.255 & 1.164 & 3092 & 1.201 & 1.092 & @xmath386309 real frame & 3.764 & 3.651 & 4269 & 1.189 & 1.151 & @xmath384764 & 1.185 & 1.148 & 1328 & 1.206 & 1.107 & @xmath387235 artificial frame & 4.010 & 3.848 & 4499 & 1.158 & 1.148 & @xmath384909 & 1.227 & 1.169 & 2721 & 1.225 & 1.156 & @xmath386785 real+art . , 1st exp . & 3.870 & 3.733 & 4185 & 1.152 & 1.131 & 667 & & & & & & real+art . , 1st epoch & 3.898 & 3.753 & 4158 & 1.163 & 1.156 & @xmath388561 & 1.191 & 1.154 & 3668 & 1.218 & 1.135 & @xmath387648 lllll leoi values & @xmath380.880 & @xmath380.703 & @xmath380.614 & @xmath380.6821 n2541 , real+art . 1st epoch & @xmath380.869@xmath160.003 ( 262 ) & @xmath380.714@xmath160.005 ( 175 ) & @xmath380.618@xmath160.004 ( 220 ) & @xmath380.712@xmath160.004 ( 210)n2541 , real+art . 1st exposure & @xmath380.880@xmath160.005 ( 151 ) & @xmath380.713@xmath160.007 ( 111 ) & & n2541 , artificial frame & @xmath380.872@xmath160.002 ( 524 ) & @xmath380.683@xmath160.001 ( 737 ) & @xmath380.581@xmath160.001 ( 558 ) & @xmath380.682@xmath160.002 ( 535)n1365 , real+art . 1st epoch & @xmath380.891@xmath160.005 ( 138 ) & @xmath380.693@xmath160.008 ( 80 ) & @xmath380.617@xmath160.006 ( 156 ) & @xmath380.708@xmath160.006 ( 97)n1365 , real+art . 1st exposure & @xmath380.892@xmath160.011 ( 34 ) & @xmath380.730@xmath160.02 ( 27 ) & & n1365 , artificial frame & @xmath380.869@xmath160.002 ( 610 ) & @xmath380.677@xmath160.001 ( 726 ) & @xmath380.579@xmath160.001 ( 730 ) & @xmath380.685@xmath160.001 ( 662 ) + leoi values & 1.066 & @xmath380.740 & @xmath380.763 & @xmath380.7641 n2541 , real+art . 1st epoch & 1.081@xmath160.003 ( 423 ) & @xmath380.734@xmath160.004 ( 219 ) & @xmath380.755@xmath160.003 ( 271 ) & @xmath380.773@xmath160.003 ( 263)n2541 , real+art . 1st exposure & 1.079@xmath160.005 ( 136 ) & @xmath380.757@xmath160.008 ( 64 ) & & n2541 , artificial frame & 1.107@xmath160.001 ( 558 ) & @xmath380.739@xmath160.001 ( 802 ) & @xmath380.739@xmath160.001 ( 724 ) & @xmath380.763@xmath160.001 ( 712)n1365 , real+art . 1st epoch & 1.059@xmath160.005 ( 210 ) & @xmath380.766@xmath160.008 ( 96 ) & @xmath380.754@xmath160.004 ( 219 ) & @xmath380.780@xmath160.006 ( 140)n1365 , real+art . 1st exposure & 1.041@xmath160.013 ( 30 ) & @xmath380.814@xmath160.028 ( 18 ) & & n1365 , artificial frame & 1.105@xmath160.001 ( 697 ) & @xmath380.740@xmath160.001 ( 784 ) & @xmath380.735@xmath160.001 ( 861 ) & @xmath380.763@xmath160.001 ( 763 ) lrrrr + n2541 1st exp . & 0.009@xmath160.012 ( 21 ) & 0.007@xmath160.0085 ( 24 ) & 0.040@xmath160.011 ( 21 ) & 0.023@xmath160.020 ( 13 ) n2541 2nd exp . & @xmath380.023@xmath160.011 ( 23 ) & @xmath380.012@xmath160.0090 ( 23 ) & 0.021@xmath160.013 ( 19 ) & @xmath380.009@xmath160.023 ( 10 ) n1365 1st exp . & @xmath380.186@xmath160.043 ( 13 ) & @xmath380.046@xmath160.025 ( 16 ) & @xmath380.011@xmath160.035 ( 8) & @xmath380.082@xmath160.023 ( 18)n1365 2nd exp . & @xmath380.166@xmath160.031 ( 10 ) & @xmath380.036@xmath160.026 ( 16 ) & @xmath380.001@xmath160.036 ( 9 ) & @xmath380.082@xmath160.022 ( 18 ) + n2541 1st exp . & 0.013@xmath160.012 ( 18 ) & 0.047@xmath160.0084 ( 23 ) & 0.090@xmath160.010 ( 22 ) & 0.069@xmath160.014 ( 17 ) n2541 2nd exp . & 0.032@xmath160.012 ( 16 ) & 0.022@xmath160.0083 ( 22 ) & 0.037@xmath160.010 ( 20 ) & 0.038@xmath160.015 ( 15 ) n1365 1st exp . & @xmath380.063@xmath160.024 ( 9 ) & @xmath380.031@xmath160.024 ( 13 ) & 0.023@xmath160.035 ( 10 ) & 0.058@xmath160.025 ( 12 ) n1365 2nd exp . & @xmath380.342@xmath160.111 ( 9 ) & @xmath380.039@xmath160.028 ( 14 ) & @xmath380.008@xmath160.031 ( 10 ) & 0.047@xmath160.021 ( 13 ) lcccc pc , f555w & @xmath380.20 @xmath16 0.04 & @xmath380.0010 @xmath16 0.0010 & 0.00 @xmath16 0.03 & 0.0013 @xmath16 0.0006 wf2 , f555w & @xmath380.04 @xmath16 0.03 & @xmath380.0020 @xmath16 0.0008 & 0.01 @xmath16 0.03 & 0.0006 @xmath16 0.0006 pc , f814w & @xmath380.08 @xmath16 0.04 & @xmath380.0018 @xmath16 0.0009 & 0.00 @xmath16 0.03 & 0.0004 @xmath16 0.0006 wf2 , f814w & @xmath380.08 @xmath16 0.03 & @xmath380.0009 @xmath16 0.0007 & 0.00 @xmath16 0.03 & 0.0008 @xmath16 0.0006 lrrrr pc , f555w & 0.12 @xmath16 0.05 & @xmath380.0009 @xmath16 0.0014 & 0.01 @xmath16 0.05 & 0.0007 @xmath16 0.0016wf2 , f555w & 0.02 @xmath16 0.05 & 0.0019 @xmath16 0.0015 & 0.03 @xmath16 0.10 & @xmath380.0022 @xmath16 0.0031wf3 , f555w & 0.02 @xmath16 0.04 & 0.0025 @xmath16 0.0011 & 0.00 @xmath16 0.04 & 0.0012 @xmath16 0.0014 wf4 , f555w & @xmath380.05 @xmath16 0.05 & 0.0014 @xmath16 0.0013 & 0.01 @xmath16 0.05 & 0.0026 @xmath16 0.0018pc , f814w & @xmath380.02 @xmath16 0.06 & 0.0013 @xmath16 0.0018 & 0.01 @xmath16 0.04 & 0.0023 @xmath16 0.0016 wf2 , f814w & 0.07 @xmath16 0.06 & 0.0015 @xmath16 0.0019 & 0.03 @xmath16 0.07 & 0.0009 @xmath16 0.0025wf3 , f814w & 0.09 @xmath16 0.04 & 0.0006 @xmath16 0.0011 & 0.00 @xmath16 0.04 & @xmath380.0010 @xmath16 0.0013 wf4 , f814w & 0.10 @xmath16 0.04 & @xmath380.0015 @xmath16 0.0013 & 0.00 @xmath16 0.04 & 0.0014 @xmath16 0.0015 pc , f555w & 0.02 @xmath16 0.05 & 0.0005 @xmath16 0.0013 & 0.09 @xmath16 0.05 & @xmath380.0009 @xmath16 0.0013 wf2 , f555w & 0.04 @xmath16 0.04 & 0.0000 @xmath16 0.0012 & 0.01 @xmath16 0.07 & @xmath380.0053 @xmath16 0.0021 wf3 , f555w & 0.07 @xmath16 0.04 & 0.0017 @xmath16 0.0011 & 0.01 @xmath16 0.05 & @xmath380.0007 @xmath16 0.0014 wf4 , f555w & 0.02 @xmath16 0.05 & @xmath380.0004 @xmath16 0.0012 & 0.00 @xmath16 0.04 & @xmath380.0034 @xmath16 0.0013 pc , f814w & 0.00 @xmath16 0.03 & @xmath380.0002 @xmath16 0.0009 & 0.00 @xmath16 0.03 & @xmath380.0007 @xmath16 0.0010 wf2 , f814w & 0.01 @xmath16 0.04 & 0.0023 @xmath16 0.0012 & 0.01 @xmath16 0.05 & @xmath380.0022 @xmath16 0.0015 wf3 , f814w & 0.08 @xmath16 0.04 & 0.0019 @xmath16 0.0010 & 0.00 @xmath16 0.03 & @xmath380.0011 @xmath16 0.0011 wf4 , f814w & 0.02 @xmath16 0.03 & 0.0009 @xmath16 0.0009 & 0.00 @xmath16 0.04 & @xmath380.0002 @xmath16 0.0010 lcccc pc & 0.069 & 0.066 & 0.124 & 0.134wf2 & 0.138 & 0.179 & 0.207 & 0.191wf3 & 0.074 & 0.072 & 0.119 & 0.124wf4 & 0.082 & 0.107 & 0.158 & 0.166pc & 0.040 & 0.041 & 0.034 & 0.080wf2 & 0.088 & 0.118 & 0.121 & 0.145wf3 & 0.039 & 0.043 & 0.051 & 0.098wf4 & 0.041 & 0.037 & 0.049 & 0.068 lcccc pc & 0.040 & 0.037 & 0.026 & 0.021 wf2 & 0.086 & 0.075 & 0.036 & 0.056 wf3 & 0.033 & 0.041 & 0.021 & 0.018 wf4 & 0.055 & 0.056 & 0.026 & 0.031 pc & 0.027 & 0.027 & 0.000 & 0.017 wf2 & 0.063 & 0.067 & 0.037 & 0.028 wf3 & 0.023 & 0.011 & 0.009 & 0.021 wf4 & 0.031 & 0.018 & 0.022 & 0.013

we solve the gross - pitaevskii ( gp ) and bogoliubov equations to investigate the metastability of superfluidity in a bose - einstein condensate in the presence of a uniformly moving defect potential in a two - dimensional torus . we calculate the total energy and momentum as functions of the driving velocity of the moving defect and find metastable states with negative effective - mass near the critical velocity . we also find that the first excited energy ( energy gap ) in the finite - sized torus closes at the critical velocity , that it obeys one - fourth power - law scaling , and that the dynamical fluctuation of the density ( amplitude of the order parameter ) is strongly enhanced near the critical velocity . we confirm the validity of our results near the critical velocity by calculating the quantum depletion . we find an unconventional swallowtail structure ( multiple - swallowtail structure ) through calculations of the unstable stationary solutions of the gp equation .

1407.7915

the breakdown of superfluidity is a long - standing but still central issue regarding quantum fluids @xcite . it has been observed in experiments of cold atomic gases trapped in simply connected geometry @xcite and multiply connected geometry @xcite . the latter experiments have exhibited various properties of superfluidity , including critical velocity , vortex nucleation , decay of persistent current , phase slip , and hysteresis . the breakdown of superflow stability in cold atoms has been studied theoretically @xcite using the gross - pitaevskii ( gp ) equation @xcite . for example , frisch _ et al_. @xcite showed that , in the presence of a defect potential , a vortex pair is nucleated when the velocity is above a critical value . nucleation of solitons was studied in one - dimensional superfluids @xcite , and the results aid in our understanding of the relationship between the nucleation of a topological defect and the breakdown of superfluidity . the breakdown of superfluidity can be understood , within the mean - field theory , through an energy diagram of stable or unstable states as functions of a control parameter ( angular velocity of a container or driving velocity of an optical lattice , for example ) . superfluidity breaks down at the control parameter when the metastable superflow state meets an unstable state in the energy diagram . from condensate wave function of the unstable state , furthermore , we see that dynamics of topological defects causes decay of superflow . as a typical energy diagram of superfluids , swallowtail structure @xcite has been investigated for one - dimensional optical lattices and ring - shaped systems with narrow widths . those theories seem to explain experimental results . more recently , experimental results @xcite on vortex nucleations and breakdown of superfluids have been discussed on the basis of swallowtail structure and corresponding energy landscape . in two- and three - dimensional superfluids , however , the whole structure of energy diagram of stable or unstable branches has not yet been known . our aim is to find the whole structure of energy diagram in two - dimensional systems and gain physical insight into breakdown of superfluids related to vortex nucleation . for this purpose , we solve the gp and bogoliubov equations in two - dimensional torus with a uniformly moving defect potential . in our previous work @xcite , we showed the properties of excitations and fluctuations near the critical velocity . in this full paper , we present the whole structure of energy diagram and related results ; material not reported in ref . @xcite includes the existence or absence of a ghost vortex pair , quantum depletion near the critical velocity , and energy diagram , which we call _ multiple - swallowtail structure_. this paper is organized as follows : in sec . [ sec : model ] , we introduce our model . in sec . [ sec : results ] , we present the stable stationary solutions of the gp and bogoliubov equations @xcite . in sec . [ subsec : energy - diagram_2d ] , the velocity dependence of the total energy and the total momentum are presented , and we compare our results with those for optical lattice systems . in sec . [ subsec : density_and_phase_profile ] , we show the density and the phase profiles below and above the critical velocity and discuss the appearance of a ghost vortex pair . in secs . [ subsec : excitation_spectra_2d ] and [ subsec : fluctuation_2d ] , we demonstrate the properties of the excitation and the fluctuation . in sec . [ subsec : quantum depletion ] , we show the results for quantum depletion and discuss the validity of the gp and the bogoliubov approximation in this system . in sec . [ sec : results_unstable ] , we present unstable stationary solutions of the gp equation . we show that a multiple - swallowtail structure appears in this system . in sec . [ sec : discussion ] , we discuss the bifurcation structure of the system , the relation between the fluctuations and the energy landscape near the critical velocity , and the possible effects of the multiple - swallowtail structure on the decay of supercurrent . finally , we summarize our results in sec . [ sec : summary ] . the numerical methods used in the present work are summarized in appendixes [ app : methods_of_numerical_calculations ] and [ app : pacm ] . we consider a system in which @xmath0 bosons of mass @xmath1 are confined in a two - dimensional torus @xmath2 . in the mean - field approximation , the physical properties of the system can be described by a complex order parameter ( condensate wave function ) @xmath3 , where @xmath4 and @xmath5 denote the coordinate and time in the laboratory frame , respectively . the subscript @xmath6 denotes the variables in the laboratory frame . the condensate wave function obeys the gp equation @xcite : @xmath7 where @xmath8 represents a moving defect potential with a constant velocity @xmath9 and @xmath10 is the strength of the interaction . we use the gaussian potential : @xmath11},\label{eq : model_defect_potential}\end{aligned}\ ] ] where @xmath12 and @xmath13 are the strength and the width of the potential , respectively . throughout this paper , the velocity of the potential is in the direction of positive @xmath14 ( @xmath15 , where @xmath16 is a unit vector in the direction of @xmath14 . ) periodic boundary conditions are imposed on @xmath3 , because there is a requirement that the condensate wave function should be single valued : @xmath17 where @xmath18 is a unit vector in the direction of positive @xmath19 . from this boundary condition , we can define the winding number : @xmath20 where @xmath21 is the phase of the condensate wave function . the gp equation in the laboratory frame ( [ eq : model : gp_equation_in_lab_frame ] ) depends explicitly on time . we remove @xmath22 dependence by performing a coordinate transformation from the laboratory frame to a moving frame @xcite , as follows : @xmath23 using eqs . ( [ eq : model_change_of_the_coordinate])@xmath24([eq : model_transformation_time_derivative ] ) , the gp equation in the moving frame is given by @xmath25 as a result of transformation ( [ eq : model_local_gauge_transoformation ] ) , the periodic boundary condition becomes twisted @xcite : @xmath26 the stationary solution of the gp equation ( [ eq : model_gp_equation_in_mov_frame ] ) is given by @xmath27 , where @xmath28 is the chemical potential . substituting this relation into eq . ( [ eq : model_gp_equation_in_mov_frame ] ) , we obtain the time - independent gp equation : @xmath29 the chemical potential @xmath28 is determined by the condition @xmath30 the bogoliubov equation @xcite can be derived by linearizing the gp equation around the stationary solution @xmath31 . substituting @xmath32\label{eq : model_linearizing_gp_equation}\end{aligned}\ ] ] into the time - dependent gp equation ( [ eq : model_gp_equation_in_mov_frame ] ) , and neglecting the higher - order terms of @xmath33 and @xmath34 , we obtain the bogoliubov equation , @xmath35 ^ 2 \\ g[\psi^{\ast}(\bm{r})]^2 & -\mathcal{l } \end{bmatrix } \begin{bmatrix } u_i(\bm{r } ) \\ v_i(\bm{r } ) \end{bmatrix } = \epsilon_i \begin{bmatrix } u_i(\bm{r } ) \\ v_i(\bm{r } ) \end{bmatrix } , \label{eq : model_bogoliubov_equation}\\ \mathcal{l}\equiv -\frac{\hbar^2}{2m}\nabla^2+u(\bm{r})-\mu+2g|\psi(\bm{r})|^2.\label{eq : model_definition_of_l}\end{aligned}\ ] ] here @xmath33 and @xmath34 are the wave functions of the @xmath36-th excited state with an excitation energy @xmath37 . the boundary conditions for @xmath33 and @xmath34 are determined by the condition in which eq . ( [ eq : model_linearizing_gp_equation ] ) satisfies the twisted periodic boundary conditions ( [ eq : model_twisted_periodic_boundary_condtion_x ] ) and ( [ eq : model_twisted_periodic_boundary_condtion_y ] ) : @xmath38 the wave functions for the excited states satisfy the following orthonormal conditions : @xmath39&=&\delta_{i j},\label{eq : model_orthonormal_condition_for_u_and_v_1}\\ \int d\bm{r}\left[u_i(\bm{r})v_j(\bm{r})-v_i(\bm{r})u_j(\bm{r})\right]&=&0.\label{eq : model_orthonormal_condition_for_u_and_v_2}\end{aligned}\ ] ] throughout this paper , length , energy , and time are normalized by the healing length @xmath40 , @xmath41 , and @xmath42 , respectively , where @xmath43 ( @xmath44 is the area of the system ) is the mean particle density . the velocity is normalized by the sound velocity @xmath45 or @xmath46 . numerically solving the gp and bogoliubov equations yields the condensate wave function , excitation spectra , and wave functions for the excited states . the methods we used for the numerical calculations are summarized in appendixes [ app : methods_of_numerical_calculations ] and [ app : pacm ] . , @xmath47 ) . the red , green , blue , purple , and orange lines correspond to branches with a winding number @xmath48 , and 4 , respectively . open ( solid ) symbols represent energy branches with a nonpositive ( positive ) winding number.,width=302 ] first , we show the energy diagram , which represents the total energy in the moving frame as a function of driving velocity of moving defect . the energy diagram yields the superfluid fraction ( the nonclassical rotational inertia ) @xcite , the critical velocity , and the metastability and hysteresis of superflow states @xcite . the total energy in the moving frame is defined by @xmath49.\label{eq : results_total_energy_in_moving_frame}\end{aligned}\ ] ] figure [ fig : energy - diagram_32_5_25 ] shows the results for the stable branches . the total energy is periodic with respect to the driving velocity @xmath50 ; it stems from the periodicity of the boundary condition ( [ eq : model_twisted_periodic_boundary_condtion_x ] ) . the lowest energy state under a given @xmath50 is the ground state , and the other states are metastable . we confirm the metastability by calculating the excitation spectra of the bogoliubov equation around each stationary state of the gp equation ( see sec . [ subsec : excitation_spectra_2d ] for details ) . the energy branches shown in fig . [ fig : energy - diagram_32_5_25 ] are almost parabolic , except in the vicinity of the termination points , which correspond to the critical velocity @xmath51 . each branch can be specified by the winding number @xmath52 , which we defined by ( [ eq : model_definition_of_winding_number ] ) . for example , the red branch denoted by open circles ( @xmath53 ) , which continuously connects the ground state at @xmath54 , has @xmath55 . the green branch , denoted by open triangles ( @xmath56 ) , has @xmath57 . the winding number can serve as an _ adiabatic invariant _ under an adiabatic change of @xmath50 @xcite . suppose that the system is in the ground state ( @xmath55 ) , and @xmath50 increases adiabatically from @xmath58 to @xmath51 . we then expect that the system will evolve along the red branch , and the winding number will remain unchanged . ring trap experiments @xcite used the ground state of the noncirculating state as the initial condition in order to see the dynamics under a change of @xmath50 . for the gp equation , the adiabatic condition is determined by the bogoliubov spectrum @xcite ; in the present case , the dynamics is regarded as adiabatic when there is little change in @xmath50 within the time interval @xmath59 ( where @xmath60 denotes the lowest excitation energy of the bogoliubov spectrum ) . as we will show in sec . [ subsec : excitation_spectra_2d ] , @xmath60 vanishes at @xmath61 . when @xmath50 approaches @xmath51 , the adiabaticity condition is violated at a certain @xmath50 , and a transition from @xmath55 to a circulating state @xmath62 occurs , which corresponds to a phase slip @xcite . the flow properties of each branch can be seen in the velocity dependence of the total momentum shown in figs . [ fig : momentum_velocity_32_5_25](a ) and [ fig : momentum_velocity_32_5_25](b ) , where the total momentum of the moving frame and laboratory frame are , respectively , given by @xmath63,\label{eq : results_definition_of_total_momentum}\\ p_{\rm lab}&=p - n m v.\label{eq : results_definition_of_total_momentum_lab_frame}\end{aligned}\ ] ] the @xmath19-component of the total momentum is zero by symmetry , and we thus consider only the @xmath14-component . for the red branch , the total momentum of the moving frame has a linear dependence for small @xmath50 ; see fig . [ fig : momentum_velocity_32_5_25](a ) . the superfluid fraction can be calculated through the following relations@xcite : @xmath64 where we used the relation @xmath65 . the calculated value of the superfluid fraction is @xmath66 for @xmath67 @xcite . if we consider a uniform system , the superfluid fraction becomes unity because the total momentum is given by @xmath68 . the deviation from unity for our system is due to the presence of the external potential . from eq . ( [ eq : results_definition_of_superfluid_fraction ] ) and ref . @xcite , we can show that the effective mass @xmath69 at @xmath54 is related to @xmath70 . in the presence of the external potential , usually @xmath71 holds and hence @xmath72 . figure [ fig : momentum_velocity_zoom_32_5_25 ] shows a blow - up of the region near the critical velocity for the @xmath55 branch . we note that the effective mass [ @xmath73 becomes negative near the critical velocity ; this implies that the mass flow of the condensate in the moving frame decreases while the velocity of the moving defect becomes larger . negative effective - mass states have been found in the gp equation for a bec in an optical lattice near the critical velocity ( see fig . 7 in ref . however , negative effective - mass states in an optical lattice are subject to dynamical instability ( di ) @xcite , while they are metastable in our case . this difference comes from that the di in the optical lattice systems is due to the formation of the long - period structures such as period - doubling solutions @xcite or bright gap solitons @xcite . these structures are prohibited in a torus and thus the negative - effective mass states maintain metastability . we note that the qualitatively same behavior for negative effective - mass states is found for other values of the parameters , as follows : @xmath74 , @xmath75 , and @xmath76 . we thus believe that the results in this section are generic for a superfluid in a torus near the critical velocity in the presence of a moving defect . component of the total momentum per particle for the stable branches [ @xmath77 ) ] in ( a ) the moving frame and ( b ) the laboratory frame . the red , green , blue , purple , and orange lines correspond to branches with winding numbers @xmath48 , and 4 , respectively . open ( solid ) symbols represent nonpositive ( positive ) winding numbers . the dashed black lines in ( a ) and ( b ) indicate @xmath78 and @xmath79 , respectively , which represent the velocity dependence of normal fluids.,width=302 ] branch in fig . [ fig : momentum_velocity_32_5_25](a ) near the critical velocity.,width=302 ] we next consider the spatial profiles of the density and the phase of the condensate wave function for a strong potential @xmath80 and a weak potential @xmath81 . figures [ fig : density_and_phase_profile_48_10_25_0.43034](a ) and [ fig : density_and_phase_profile_48_10_25_0.43034](b ) show the spatial profiles of the density and phase , respectively , of the condensate wave function for @xmath82 and @xmath83 . we find a vortex pair in the low - density region ; this is called a ghost vortex pair ( gvp ) @xcite . this is a ( meta)stable stationary solution , because no di occurs in the solution ( see sec . [ subsec : excitation_spectra_2d ] ) . the gvp is regarded as being pinned to the defect potential , and thus it could be depinned above the critical velocity . in fact , we calculated the real - time dynamics above the critical velocity and found that vortex nucleation occurred as shown in figs . [ fig : density_48_10_25_04_08_t20_t50](a ) and [ fig : density_48_10_25_04_08_t20_t50](b ) . we generated the real - time dynamics by the crank - nicholson scheme . and white and black circles represent the position of the gvp.,width=264 ] and @xmath84 , at ( a ) @xmath85 and ( b ) @xmath86 . the initial condition is the stationary solution for @xmath87 , which contains the gvp.,width=264 ] ghost vortices were first reported in refs . @xcite , where vortex invasions of rotating condensates were numerically studied . a gvp was found in a numerical study of condensates in the presence of an oscillating defect @xcite . our result yields an example of a gvp accompanying a defect moving with a constant subcritical velocity . a gvp does not appear in the presence of a weak potential ; this is shown in figs . [ fig : density_and_phase_profile_48_1_25_0.4608505](a ) and [ fig : density_and_phase_profile_48_1_25_0.4608505](b ) , where we present the density and phase profiles for @xmath88 ) and @xmath89 . typically , investigations are for a velocity in the range of @xmath90 . the dynamics above the critical velocity in the weak - potential case are shown in figs . [ fig : density_48_1_25_04_08_t20_t50](a ) and [ fig : density_48_1_25_04_08_t20_t50](b ) . these figures clearly show that vortices nucleate even when the initial state contains no gvps . and @xmath91.,width=264 ] and @xmath92 , at ( a ) @xmath85 , and ( b ) @xmath86 . the initial condition is the stationary solution for @xmath93.,width=264 ] ; and ( b ) @xmath94 . up to the seventh excitation energy is shown.,width=302 ] and @xmath80 . the solid black line shows the energy gap for @xmath95 . ( b ) fitting results for ( a ) . the solid purple line represents @xmath96^{1/4}$].,width=302 ] and @xmath97 . ( b ) fitting results for ( a ) . the solid purple line represents @xmath96^{1/4}$].,width=302 ] we next present the results for excitations . figure [ fig : excitation_all_32_and_48 ] shows energy spectra of the bogoliubov equation as a function of @xmath50 . in systems of finite size , the bogoliubov spectra are discretized . the excitation energy is always positive , and thus the solutions shown in the figure are stable or metastable . we now focus on the first excited energy , which we call an energy gap ( denoted by @xmath98 ) . figures [ fig : gap_scaling_32_10_d](a ) and [ fig : gap_scaling_32_48_64_10_25](a ) show the energy gap as a function of the velocity . we first notice a linear decrease in the region in which the velocity is small . this reflects the energy gap in uniform systems , and it is given by @xmath99 $ ] . in fact , the solid black line in fig . [ fig : gap_scaling_32_10_d](a ) , which represents @xmath100 , almost overlaps the numerical data for @xmath101 when the velocity is small , except near @xmath54 . the deviation between @xmath100 and the numerical data near @xmath54 is due to the splitting of levels , since the first excited state in a uniform system is fourfold degenerate . we also notice a sharp decrease in the energy gap near the critical velocity . to characterize this behavior , we used the function @xmath102^{c}$ ] , where @xmath103 , and @xmath104 are parameters to fit four sets of data points near the critical velocity . the detailed data are shown in table 1 of ref . figures [ fig : gap_scaling_32_10_d](b ) and [ fig : gap_scaling_32_48_64_10_25](b ) show the results . we determined that the scaling for the energy gap @xmath102^{1/4}$ ] near the critical velocity . so far we have shown only the excitation spectra . we will show the wave functions for excited states in this subsection . using the wave functions of the excited states , we can obtain the properties of the fluctuations . substituting eq . ( [ eq : model_linearizing_gp_equation ] ) into @xmath105 and @xmath106 , and neglecting the higher - order terms of @xmath33 and @xmath34 , we obtain @xmath107,\label{eq : results_density_fluctuation}\\ \frac{\psi(\bm{r } , t)}{|\psi(\bm{r } , t)|}&=e^{-i\mu t/\hbar}e^{i\varphi(\bm{r})}\nonumber \\ & \quad \times \left\{1+\frac{i}{|\psi(\bm{r})|^2}{\rm im}\left[\delta p_i(\bm{r})e^{-i\epsilon_it/\hbar}\right]\right\},\label{eq : results_phase_fluctuation}\end{aligned}\ ] ] where the local density fluctuation @xmath108 and the local phase fluctuation @xmath109 for mode @xmath36 are defined by @xcite @xmath110 in fig . [ fig : fluctuation_density_and_phase_48_5_25 ] , we show the spatial profiles of the density and phase fluctuations for the first excited state . we can see that both the density and phase fluctuations are enhanced when the velocity of the moving defect becomes larger . the enhancement for the density fluctuation is a few orders of magnitude greater than that for the phase fluctuation . we plot the energy dependence of the density fluctuations in fig . [ fig : fluctuation_energy_48_5_25 ] , where the spectral intensity shifts to lower energy and is enhanced when the velocity approaches the critical value . similar behavior was observed in one - dimensional systems , and was related to soliton nucleation @xcite . we briefly remark on the effect of the existence or absence of a gvp on the fluctuations . for the weak potential case , there is no gvp near the critical velocity , as described in sec . [ subsec : density_and_phase_profile ] . in this case , the spatial patterns of the density and phase fluctuations are slightly different ( data not depicted here ) . however , enhancement of the density fluctuation also occurs in the absence of a gvp . we will present the results for quantum depletion ( qd ) , which represents the number of atoms not in the condensate . the expression for the qd is @xmath111 the condition for the gp and bogoliubov approximations to be valid is @xmath112 . therefore , we can check the self - consistency of these approximations by examining the qd . in previous works , qd in nonuniform systems was calculated in the following cases : these studies used perturbative approaches @xcite and many - body calculations for the ground state @xcite . our results presented below are the first examples of the qd near the critical velocity in the presence of a moving defect . and @xmath81 , and ( b)@xmath113 and @xmath97 . the solid purple line represents the qd in the absence of the defect potential . ( c ) normalized velocity dependence of the qd . in all cases , we set @xmath114.,width=302 ] we show the velocity dependence of the qd in fig . [ fig : depletion_32_and_48 ] @xcite . if the system is uniform ( @xmath115 ) and does not exhibit spontaneous translational symmetry breaking , the qd does not depend on the velocity because the velocity dependence of @xmath34 is present only in the plane wave component ( @xmath116 ) . our results show that the qd depends on the velocity , due to the presence of the defect potential . when the velocity is small , the qd is almost the same as that for uniform systems . it is consistent with that for the energy gap when the velocity is small . near the critical velocity , we find that the qd increases steeply and attribute it to enhancement of the low - energy density of states and density fluctuations . within the range of @xmath50 used in our calculations , we do not find power - law scaling of the qd , in contrast to that found for the energy gap ; see fig . [ fig : depletion_32_and_48 ] ( c ) , and note that the curves are not straight lines in the log - log plot near the critical velocity . there may be a narrow scaling region for the qd in this system . although the qd increases near the critical velocity , the value of @xmath117 is still much smaller than unity for @xmath118 and @xmath119 ; see fig . [ fig : depletion_32_and_48 ] ( c ) . here , @xmath120 is the ratio between the healing length and the mean particle distance . a small @xmath120 corresponds to a weakly interacting case . this shows that the gp and bogoliubov approximations are valid even near the critical velocity , for sufficiently small @xmath120 . we can easily obtain the qd for other values of @xmath120 , because the @xmath120 dependence of @xmath34 is given by @xmath121 . figure [ fig : dep_different_delta_48_1_25 ] shows the @xmath120 dependence of the qd . these results show that the qd is much smaller than unity near the critical velocity for @xmath122 , and for @xmath123 , the bogoliubov approximation breaks down near the critical velocity . , for four different values of @xmath120.,width=302 ] in this section , we will present the results for unstable branches , which we calculated by the pseudo - arclength continuation method @xcite . the details are summarized in appendix [ app : pacm ] . figure [ fig : energy_diagram_sw_32_5_25 ] shows the energy diagram containing the stable and the unstable branches for @xmath124 . our results do not exhibit the conventional swallowtail structure but a multiple structure of the unstable branches , in contrast to one - dimensional lattices @xcite and ring systems @xcite . we call it the multiple - swallowtail structure . the branch ( a ) in the inset of fig . [ fig : energy_diagram_sw_32_5_25 ] continuously connects the stable branch with @xmath55 at the critical velocity and contains one vortex pair as shown in fig . [ fig : density_and_phase_swallow_tail_32_5_25 ] ( a ) . the upper ( lower ) vortex has negative ( positive ) vorticity . this solution is similar to the unstable solution reported in ref . @xcite . at the left termination point , the branch ( a ) folds back and connects with the branch ( b ) in the energy diagram . we have two pairs of vortices in the branch ( b ) . at @xmath125 on the top of the unstable branch ( d ) , the self - induced phase slip @xcite occurs due to formation of the dark soliton [ see figs . [ fig : density_and_phase_swallow_tail_32_5_25](d ) and [ fig : density_and_phase_swallow_tail_32_5_25](h ) ] . consequently , the winding number changes from @xmath55 to @xmath57 @xcite . after the self - induced phase slip , finally , the unstable branch ( g ) connects with the stable branch for @xmath57 . in fig . [ fig : energy_diagram_sw_48_5_25 ] , we show the energy diagram for a larger system with @xmath126 . there are more unstable branches with more pairs of vortices compared to the case of @xmath124 . we consider that the number of the branches of the multiple swallowtail structure is sensitive to the width of the superflow path ( spatial extension in the @xmath19-direction ) . we expect that the number of the unstable branches reduces to unity in systems with a narrow path of superflow comparable to the healing length and the multiple swallowtail becomes conventional swallowtail . we show the excitation spectra for an unstable branch ( a ) and the stable branch near the critical velocity in fig . [ fig : excitation_sw ] . as expected , the di occurs in the unstable branch ( a ) . we confirm that the di occurs in other unstable branches ( b ) to ( g ) in fig . [ fig : energy_diagram_sw_32_5_25 ] ( data not depicted here ) . . the red dotted ( solid ) line shows the stable ( unstable ) solution for @xmath55 . the blue dotted ( solid ) line shows the stable ( unstable ) solution for @xmath57 . the inset shows magnified view around @xmath125.,width=302 ] . the red dotted ( solid ) line shows the stable ( unstable ) solution for @xmath55 . the blue dotted ( solid ) line shows the stable ( unstable ) solution for @xmath57 . the inset shows magnified view around @xmath125.,width=302 ] . the solid red curve represents the excitation energy of the stable branch . the green circle and the blue square represent the real and imaginary part of the excitation energy for branch ( a ) in fig . [ fig : energy_diagram_sw_32_5_25 ] , respectively.,width=302 ] in this section , we will discuss the bifurcation structure of the system , the relation between the fluctuation and the energy landscape , and effects of the multiple - swallowtail structure on the superfluidity based on the results presented in secs . [ sec : results ] and [ sec : results_unstable ] . from the results for the stable and unstable stationary solutions and their excitation spectra , we can discuss the bifurcation structures of the system . as shown in figs . [ fig : energy_diagram_sw_32_5_25 ] and [ fig : energy_diagram_sw_48_5_25 ] , the stable and the unstable branches merge at the critical velocity . it implies that a saddle - node ( sn ) type bifurcation , _ i.e. _ , either conventional sn or hamiltonian sn , occurs in this system . the energy diagram and the scaling law shown in sec . [ subsec : excitation_spectra_2d ] are consistent with the hamiltonian saddle node ( hsn ) bifurcation @xcite . the normal form of the hsn bifurcation is given by @xmath127 where @xmath128 is the amplitude of the critical mode , @xmath22 is the time , @xmath129 denotes a bifurcation parameter , and @xmath130 is a constant related to the system parameters . the linear stability analysis around the stationary solution of eq . ( [ eq : results_normal_form_hsn ] ) ( @xmath131 ) shows that the frequency is proportional to @xmath132 . this is the same behavior of our systems . physically , the normal form ( [ eq : results_normal_form_hsn ] ) shows that the breakdown of the metastable state is caused by the disappearance of the energy barrier , as described in fig . [ fig : schematic_picture ] . we note that a similar mechanism for the destabilization of the metastable state has been found in attractive becs in harmonic traps @xcite . the attractive bec collapses when the number of atoms in the harmonic trap exceeds a critical value . near the collapse point , the monopole mode , which is a low - lying excitation in the attractive bec , obeys a one - fourth power law , as found by variational calculations @xcite and numerical calculations @xcite . in ref . @xcite , the hsn bifurcation was reported to occur in this system . we anticipate that the origin of the scaling law is the same as that for our system . however , there is some evidence for the conventional sn bifurcation near the critical velocity of a superfluid in the presence of an obstacle @xcite . the time scale characteristic to the conventional sn bifurcation is proportional to the square root of the distance from the bifurcation points . in future work , we will seek to determine under what conditions the bifurcation near the critical velocity of a superfluid is a conventional sn or a hamiltonian sn . another area of future work is to derive the normal form given by eq . ( [ eq : results_normal_form_hsn ] ) , starting from the gp and bogoliubov equations . we can also discuss the relation between the fluctuations and the energy landscape . the dynamics near the critical velocity are often discussed in terms of the energy landscape , which is schematically depicted in fig . [ fig : schematic_picture ] . near the critical velocity , the metastable and unstable states approach each other in the configuration space , and thus the landscape around the local minimum becomes flat along a particular direction in the coordinate space . this results in the gap closing in the excitation spectrum , as shown in sec . iii c. the meaning of the abscissa in the energy landscape ( _ i.e. _ coordinate of configuration space ) is not generally clear except for a few cases where it was associated with collective coordinates @xcite . in the present case , we can identify the abscissa with the amplitude of the bogoliubov eigenstate that has the lowest excitation energy . on the basis of the results in sec . [ subsec : fluctuation_2d ] , we see that the motion of @xmath133 from the metastable state to the unstable state is accompanied by a density fluctuation near the defect potential . the decay of supercurrent is frequently understood on the basis of the energy landscape as shown in fig . [ fig : schematic_picture ] . recall that the phase slip rate for one - dimensional ring systems has been calculated in @xcite , where the conventional swallowtail structure plays a crucial role . our results imply that the conventional energy landscape , _ i.e. , _ the two local minima are separated by one local maximum , is valid only in the vicinity of the critical velocity ( see figs . [ fig : energy_diagram_sw_32_5_25 ] and [ fig : energy_diagram_sw_48_5_25 ] ) . away from the critical velocity , the energy landscape can not be written in the form shown in fig . [ fig : schematic_picture ] and we expect that the multiple - swallowtail structure affects the phase slip rate . in summary , we investigated the metastability , excitations , fluctuations , and swallowtail structures of the bec with a uniformly moving defect in a two - dimensional torus system . we first calculated the total energy and the total momentum as functions of the driving velocity of the moving defect . a negative effective - mass region appears near the critical velocity , as is the case for optical lattice systems . in contrast to optical lattice systems , the negative effective - mass states are metastable . this difference comes from that the di in the optical lattice systems is due to the formation of the long - period structures , such as the period - doubling states and the bright gap solitons , which are prohibited in the torus systems . we also found gvps in stationary states in the presence of a strong defect . using the results of the gp equation , we solved the bogoliubov equation and obtained the excitation spectra . we determined that near the critical velocity , the scaling of the energy gap followed a one - fourth power law . this implies an algebraic divergence of the characteristic time scale toward the critical velocity and a violation of the adiabaticity condition at the critical velocity . from wave functions of the excited states @xmath33 and @xmath34 , we obtained the fluctuation properties and showed that the density ( amplitude of the order parameter ) fluctuations are enhanced near the critical velocity . we also calculated the qd and found that it increased near the critical velocity . we confirmed the validity of the gp and the bogoliubov approximations on the basis of these results . we found unconventional swallowtail structures ( multiple - swallowtail structure ) by the direct calculations of the unstable solutions . we discussed that the number of unstable branches depends on the width of the superflow path and expect that the multiple - swallowtail structure reduces to the conventional one in the narrow superflow path limit . we discussed the bifurcation structure of the system . the results for the one - fourth power - law scaling near the critical velocity and the unstable branches imply that the hsn bifurcation occurs in the system , which describes the disappearance of the energy barrier that protects a metastable state . we pointed out that the same scaling law appears in the attractive bec in a harmonic trap near the collapse point . we also discussed the effects of the multiple - swallowtail structure on the calculations for the phase slip rate . in future work , we will attempt to determine why the hamiltonian saddle - node bifurcation appears at the critical velocity and to derive the normal form from the gp and bogoliubov equations . the energy landscape away from the critical velocity remains open . full knowledge of the energy landscape of the multiple - swallowtail structures will serve as the understanding of the breakdown of the superfluidity due to the vortex nucleations . a possible further extension of the present work is to study the effects of quantum fluctuations on the metastability of superfluidity . in particular , these effects are crucial for cases that are not weakly interacting and that are near the critical velocity . in fact , a nonzero drag force acting on a defect below the critical velocity in a one - dimensional system has been reported in ref . this phenomenon is due to quantum fluctuations . it is important to understand the effects of quantum fluctuations on vortex nucleation . another extension of the present work is to study multicomponent systems , such as spinor becs @xcite , and to reveal the effects of the internal degrees of freedom on the metastability of the superfluidity . we thank s. watabe and i. danshita for fruitful discussions and d. yamamoto for useful comments . we also thank an anonymous referee for his or her indication of the unstable solutions , which trigger the findings of the multiple - swallowtail structures . m. k. acknowledges the support of a grant - in - aid for jsps fellows ( grant no . this work is supported by jpsj kakenhi ( grant no . 24543061 ) . in this appendix , we explain the methods used for the numerical calculations . similar methods were used in our previous work @xcite . in order to obtain the solutions of the time - independent gp equation ( [ eq : model_time - independent_gp_equation_in_mov_frame ] ) , we used imaginary time propagation . the imaginary - time gp equation is given by @xmath134 where @xmath135 is the time - dependent chemical potential , whose time dependence is determined by the total number of particles ( [ eq : model_total_particle_number_condtion ] ) . the time - independent solution of eq . ( [ eq : appendix_imaginary_time_gp ] ) coincides with the original gp equation ( [ eq : model_time - independent_gp_equation_in_mov_frame ] ) . from the twisted periodic boundary conditions ( [ eq : model_twisted_periodic_boundary_condtion_x ] ) and ( [ eq : model_twisted_periodic_boundary_condtion_y ] ) , we can expand the condensate wave function as a series of plane waves , as follows : @xmath136 where @xmath137 is an expansion coefficient , @xmath138 , and @xmath139 and @xmath140 . substituting eq . ( [ eq : appendix_expansion_of_the_condensate_wave_function ] ) into eq . ( [ eq : appendix_imaginary_time_gp ] ) and using the orthonormal condition @xmath141 , we obtain the imaginary time gp equation for @xmath137 : @xmath142c_{\bm{q}+\bm{g}}(t)\nonumber \\ + \frac{1}{s}\sum_{\bm{g}'}\bar{u}(\bm{g}-\bm{g}')c_{\bm{q}+\bm{g}'}(t)\nonumber\\ + g n_0\sum_{\bm{g},\delta\bm{g}}c^{\ast}_{\bm{q}+\bm{g}'+\delta\bm{g}}(t)c_{\bm{q}+\bm{g}'}(t)c_{\bm{q}+\bm{g}+\delta\bm{g}}(t),\label{eq : appendix_imaginary_time_gp_for_coefficients}\end{aligned}\ ] ] where @xmath143 is the fourier transformation of the external potential : @xmath144 the total particle number condition for @xmath137 is given by @xmath145 from the boundary conditions ( [ eq : model_boundary_condtion_for_u_x ] ) , ( [ eq : model_boundary_condtion_for_v_x ] ) , ( [ eq : model_boundary_condtion_for_u_y ] ) , and ( [ eq : model_boundary_condtion_for_v_y ] ) , the wave functions @xmath33 and @xmath34 can be expanded as a series of plane waves , as follows : @xmath146 where @xmath147 and @xmath148 are the expansion coefficients of mode @xmath36 . the normalization condition for the wave functions of the excited states becomes @xmath149=1.\label{eq : appendix_normalization_condition_u_and_v}\end{aligned}\ ] ] substituting eqs . ( [ eq : appendix_expansion_of_the_condensate_wave_function ] ) , ( [ eq : appendix_expansion_for_u ] ) , and ( [ eq : appendix_expansion_for_v ] ) into eq . ( [ eq : model_bogoliubov_equation ] ) , we obtain the bogoliubov equation for the expansion coefficients : @xmath150 where we have introduced the following variables in order to simplify the notation : @xmath151 numerically diagonalizing eqs . ( [ eq : appendix_bogoliubov_equation_for_a ] ) and ( [ eq : appendix_bogoliubov_equation_for_b ] ) , we obtain the excitation spectra and the wave functions of the excited states . we introduced the cutoff @xmath152 to calculate the summation of @xmath153 . we used the cutoff wave number ( the number of bases ) @xmath154 , and @xmath155 for @xmath156 and @xmath157 , and @xmath158 for @xmath113 . we checked that the cutoff dependence of the present results is negligibly small , other than for the calculation of the quantum depletion ( see ref . in this appendix , we explain the pseudo - arclength continuation method ( pacm ) @xcite . we used this method to obtain the unstable solutions . this method was applied to the spin-1 gp equation in ref . @xcite . we consider nonlinear algebraic equations , @xmath159 where @xmath160 and @xmath161 is a parameter . one way to solve this equation is the newton method : @xmath162 where @xmath163 and @xmath164 are the initial and next values of @xmath165 , respectively . solving eq . ( [ eq : appendix_newton_method_applying_to_g ] ) until convergence , we can obtain a solution for eq . ( [ eq : appendix_equation_we_want_to_solve ] ) . however , this method fails if the saddle - node bifurcation point exists . to avoid this difficulty , we use the pcam method . in the pcam method , instead of solving eq . ( [ eq : appendix_equation_we_want_to_solve ] ) for fixed @xmath166 , we regard @xmath166 as a variable and solve the following equations : @xmath167 where @xmath163 is a solution of eq . ( [ eq : appendix_equation_we_want_to_solve ] ) for @xmath168 , @xmath169 and @xmath170 are the normalized tangent vector in the @xmath171 space at the point @xmath172 , and @xmath173 is the arclength ( see fig . [ fig : schematic_picture_pcam ] ) . equation @xmath174=0 represents the plane that is perpendicular to the tangent vector @xmath175 and is separated by the distance @xmath173 from the point @xmath172 . the tangent vector @xmath175 is determined by the following way : let @xmath176 be a parameter that assigns the position of the orbit in the space @xmath171 . the tangent vector can be obtained by solving the following equation : @xmath177 where @xmath178 is the value of @xmath176 at the point @xmath179 . equation ( [ eq : appendix_derivative_of_g_with_respective_to_s ] ) reduces to @xmath180 the normalization condition for the tangent vector is given by @xmath181 from the normalization condition , the overall sign of the tangent vector is not determined . the overall sign can be determined so that the following condition is satisfied : @xmath182 where @xmath183 is the tangent vector of the previous step . * prepare @xmath172 . * set the appropriate value of @xmath173 . * calculate the tangent vector @xmath175 from eqs . ( [ eq : appendix_determine_tangent_vector ] ) , ( [ eq : appendix_normalization_condition ] ) , and ( [ eq : appendix_overall_factor ] ) . * calculate @xmath184 ( euler predictor ) . this is the initial condition of the newton method ( see fig . [ fig : schematic_picture_pcam ] ) . * solve eq . ( [ eq : appendix_equation_for_pcam ] ) by the newton method : @xmath185 * iterate ( v ) until convergence . 99 l. d. landau , j. phys . ( ussr ) * 5 * , 71 ( 1941 ) . r. p. feynman , prog . * 1 * , 17 ( 1995 ) . a. j. leggett , phys . fenn . * 8 * , 125 ( 1973 ) . p. nozires and d. pines , _ the theory of quantum liquids _ vol . ii ( addison - wesley , reading , 1990 ) . a. j. leggett , rev . phys . * 73 * , 307 ( 2001 ) . a. j. leggett , _ quantum liquids _ ( oxford university press , oxford , 2006 ) . c. raman , m. khl , r. onofrio , d. s. durfee , c. e. kuklewicz , z. hadzibabic , and w. ketterle , phys . lett . * 83 * , 2502 ( 1999 ) . r. onofrio , c. raman , j. m. vogels , j. r. abo - shaeer , a. p. chikkatur , and w. ketterle , phys . lett . * 85 * , 2228 ( 2000 ) . s. inouye , s. gupta , t. rosenband , a. p. chikkatur , a. grlitz , t. l. gustavson , a. e. leanhardt , d. e. pritchard , and w. ketterle , phys . . lett . * 87 * , 080402 ( 2001 ) . p. engels and c. atherton , phys . * 99 * , 160405 ( 2007 ) . t. w. neely , e. c. samson , a. s. bradley , m. j. davis , and b. p. anderson , phys . lett . * 104 * , 160401 ( 2010 ) . r. desbuquois , l. chomaz , t. yefsah , j. lonard , j. beugnon , c. weitenberg , and j. dalibard , nature phys . * 8 * , 645 ( 2012 ) . c. ryu , m. f. andersen , p. clad , v. natarajan , k. helmerson , and w. d. phillips , phys . lett . * 99 * , 260401 ( 2007 ) . a. ramanathan , k. c. wright , s. r. muniz , m. zelan , w. t. hill , c. j. lobb , k. helmerson , w. d. phillips , and g. k. campbell , phys . rev . lett . * 106 * , 130401 ( 2011 ) . s. moulder , s. beattie , r. p. smith , n. tammuz , and z. hadzibabic , phys . a * 86 * , 013629 ( 2012 ) . s. beattie , s. moulder , r. j. fletcher , and z. hadzibabic , phys . lett . * 110 * , 025301 ( 2013 ) . t. w. neely , a. s. bradley , e. c. samson , s. j. rooney , e. m. wright , k. j. h. law , r. carretero - gonzlez , p. g. kevrekidis , m. j. davis , and b. p. anderson , phys . 111 * , 235301 ( 2013 ) . c. ryu , p. w. blackburn , a. a. blinova , and m. g. boshier , phys . 111 * , 205301 ( 2013 ) . k. c. wright , r. b. blakestad , c. j. lobb , w. d. phillips , and g. k. campbell , phys . lett . * 110 * , 025302 ( 2013 ) . k. c. wright , r. b. blakestad , c. j. lobb , w. d. phillips , and g. k. campbell , phys . a * 88 * , 063633 ( 2013 ) . c. ryu , k. c. henderson , and m. g. boshier , new j. phys . * 16 * , 013046 ( 2014 ) . s. eckel , j. g. lee , f. jendrzejewski , n. murray , c. w. clark , c. j. lobb , w. d. phillips , m. edwards , and g. k. campbell , nature ( london ) * 506 * , 200 ( 2014 ) . f. jendrzejewski , s. eckel , n. murray , c. lanier , m. edwards , c. j. lobb , and g. k. campbell , phys . 113 * , 045305 ( 2014 ) . s. eckel , f. jendrzejewski , a. kumar , c. j. lobb , and g. k. campbell , phys . x * 4 * , 031052 ( 2014 ) . l. corman , l. chomaz , t. bienaim , r. desbuquois , c. weitenberg , s. nascimbne , j. dalibard , and j. beugnon , phys . lett . * 113 * , 135302 ( 2014 ) . t. frisch , y. pomeau , and s. rica , phys . lett . * 69 * , 1644 ( 1992 ) . y. pomeau and s. rica , comptes rendus acad . sc . ( paris ) * t. 316 * srie ii , 1523 ( 1993 ) . v. hakim , phys . e * 55 * , 2835 ( 1997 ) . b. jackson , j. f. mccann , and c. s. adams , phys . lett . * 80 * , 3903 ( 1998 ) . c. josserand , y. pomeau , and s. rica , physica d * 134 * , 111 ( 1999 ) . t. winiecki , j. f. mccann , and c. s. adams , phys . rev . lett . * 82 * , 5186 ( 1999 ) . b. jackson , j. f. mccann , and c. s. adams , phys . a * 61 * , 051603 ( 2000 ) . c. huepe and m - e . brachet , physica d * 140 * , 126 ( 2000 ) . j. brand and w. p. reinhardt , j. phys . b : at . mol . . phys . * 34 * , l113 ( 2001 ) . c. t. pham and m. brachet , physica d * 163 * , 127 ( 2002 ) . n. pavloff , phys . a * 66 * , 013610 ( 2002 ) . a. aftalion , q. du , and y. pomeau , phys . lett . * 91 * , 090407 ( 2003 ) . g. a. el , a. gammal , and a. m. kamchatnov , phys . * 97 * , 180405 ( 2006 ) . f. piazza , l. a. collins , and a. smerzi , phys . a * 80 * , 021601(r ) ( 2009 ) . k. sasaki , n. suzuki , and h. saito , phys . lett . * 104 * , 150404 ( 2010 ) . t. aioi , t. kadokura , t. kishimoto , and h. saito , phys . x * 1 * , 021003 ( 2011 ) . s. j. woo and y. w. son , phys . a * 86 * , 011604(r ) ( 2012 ) . r. dubessy , t. liennard , p. pedri , and h. perrin , phys . rev . a * 86 * , 011602 ( 2012 ) . f. piazza , l. a. collins , and a. smerzi , j. phys . b : at . mol . phys . * 46 * , 095302 ( 2013 ) . e. p. gross , nuovo cimento * 20 * , 454 ( 1961 ) . l. p. pitaevskii , sov . jetp * 13 * , 451 ( 1961 ) . d. diakonov , l. m. jensen , c. j. pethick , and h. smith , phys . a * 66 * , 013604 ( 2002 ) . e. j. mueller , phys . a * 66 * , 063603 ( 2002 ) . m. machholm , c. j. pethick , and h. smith , phys . a * 67 * , 053613 ( 2003 ) . b. wu and q. niu , new . * 5 * , 104 ( 2003 ) . c. menotti , a. smerzi , and a. trombettoni , new . * 5 * , 112 ( 2003 ) . e. taylor and e. zaremba , phys . a * 68 * , 053611 ( 2003 ) . m. machholm , a. nicolin , c. j. pethick , and h. smith , phys . rev . a * 69 * , 043604 ( 2004 ) . b. t. seaman , l. d. carr , and m. j. holland , phys . a * 71 * , 033622 ( 2005 ) . o. morsch and m. oberthaler , rev . phys . * 78 * , 179 ( 2006 ) . i. danshita and s. tsuchiya , phys . a * 75 * , 033612 ( 2007 ) . r. kanamoto , l. d. carr , and m. ueda , phys . a * 79 * , 063616 ( 2009 ) . o. fialko , m .- c . delattre , j. brand , and a. r. kolovsky , phys . lett . * 108 * , 250402 ( 2012 ) . s. baharian and g. baym , phys . a * 87 * , 013619 ( 2013 ) . m. kunimi and y. kato , j. low temp . phys . * 175 * , 201 ( 2014 ) . n. n. bogoliubov , j. phys . ( ussr ) * 11 * , 23 ( 1947 ) . a. l. fetter , ann . phys . * 70 * , 67 ( 1972 ) . e. h. lieb , r. seiringer , and j. yngvason , phys . b * 66 * , 134529 ( 2002 ) . j. liu , b. wu , and q. niu , phys . lett . * 90 * , 170404 ( 2003 ) . p. w. anderson , rev . 38 * , 298 ( 1966 ) . j. s. langer and m. e. fisher , phys . * 19 * , 560 ( 1967 ) . g. baym , in _ mathematical methods in solid state and superfluid theory _ , edited by r. c. clark and g. h. derrick ( oliver and boyd , edinburgh , 1969 ) . to obtain this value , we use a linear fit to more than @xmath186 data points of the total momentum for @xmath187 . m. krmer , c. menotti , l. pitaevskii , and s. stringari , eur . j. d * 27 * , 247 ( 2003 ) . v. v. konotop and m. salerno , phys . a * 65 * , 021602(r ) ( 2002 ) . b. eiermann , th . anker , m. albiez , m. taglieber , p. treutlein , k. -p . marzlin , and m. k. oberthaler , phys . lett . * 92 * , 230401 ( 2004 ) . m. tsubota , k. kasamatsu , and m. ueda , phys . a * 65 * , 023603 ( 2002 ) . k. kasamatsu , m. tsubota , and m. ueda , phys . a * 67 * , 033610 ( 2003 ) . k. fujimoto and m. tsubota , phys . a * 83 * , 053609 ( 2011 ) . c . wu and a. griffin , phys . a * 54 * , 4204 ( 1996 ) . a. l. fetter and d. rokhsar , phys . a * 57 * , 1191 ( 1998 ) . y. kato and s. watabe , phys . lett . * 105 * , 035302 ( 2010 ) . s. watabe and y. kato , phys . a * 88 * , 063612 ( 2013 ) . k. huang and h .- f . meng , phys . lett . * 69 * , 644 ( 1992 ) . m. kobayashi and m. tsubota , phys . b * 66 * , 174516 ( 2002 ) . c. a. mller and c. gaul , new . j. phys . * 14 * , 075025 ( 2012 ) . g. e. astrakharchik and k. v. krutitsky , phys . rev . a * 84 * , 031604(r ) ( 2011 ) . g. e. astrakharchik , k. v. krutitsky , and p. navez , phys . a * 87 * , 061601(r ) ( 2013 ) . we use extrapolation to evaluate the quantum depletion . the method is as follows : in uniform two - dimensional systems , the cutoff dependence of the quantum depletion is given by @xmath188 , where @xmath189 is the cutoff wave number . therefore , we assume the same cutoff dependence of the quantum depletion even in the presence of the defect potential . by adjusting @xmath190 and @xmath191 , we can fit the function @xmath192 to three data points , and in this way , we can calculate the quantum depletion . the error bars for the fitting are not shown in figs . [ fig : depletion_32_and_48 ] and [ fig : dep_different_delta_48_1_25 ] because these are sufficiently small in the range of our graph . h. b. keller , _ lectures on numerical methods in bifurcation problems _ ( springer - verlag , berlin / heidelberg / new york / tokyo , 1987 ) . r. kanamoto , l. d. carr , and m. ueda , phys . lett . * 100 * , 060401 ( 2008 ) . m. ueda and a. j. leggett , phys . lett . * 80 * , 1576 ( 1998 ) . c. huepe , l. s. tuckerman , s. mtens , and m. e. brachet , phys . a * 68 * , 023609 ( 2003 ) . d. e. mccumber and b. i. halperin , phys . b * 1 * , 1054 ( 1970 ) . a. polkovnikov , e. altman , e. demler , b. halperin , and m. d. lukin , phys . a * 71 * , 063613 ( 2005 ) . a. g. sykes , m. j. davis , and d. c. roberts , phys . lett . * 103 * , 085302 ( 2009 ) . y. kawaguchi and m. ueda , phys . rep . * 520 * , 253 ( 2012 ) . d. m. stamper - kurn and m. ueda , rev . phys . * 85 * , 1191 ( 2013 ) . m. kunimi and y. kato , phys . b * 86 * , 060510(r ) ( 2012 ) . j .- chen , i .- chern , and w. wang , j. comp . phys . * 230 * , 2222 ( 2011 ) . chen , i .- chern , and w. wang , j. sci . doi : 10.1007/s10915 - 014 - 9924-z .

we investigate a model for driven exclusion processes where internal states are assigned to the particles . the latter account for diverse situations , ranging from spin states in spintronics to parallel lanes in intracellular or vehicular traffic . introducing a coupling between the internal states by allowing particles to switch from one to another induces an intriguing polarization phenomenon . in a mesoscopic scaling , a rich stationary regime for the density profiles is discovered , with localized domain walls in the density profile of one of the internal states being feasible . we derive the shape of the density profiles as well as resulting phase diagrams analytically by a mean - field approximation and a continuum limit . continuous as well as discontinuous lines of phase transition emerge , their intersections induce multicritical behavior .

cond-mat0702022

non - equilibrium critical phenomena arise in a broad variety of systems , including non - equilibrium growth models @xcite , percolation - like processes @xcite , kinetic ising models @xcite , diffusion limited chemical reactions @xcite , and driven diffusive systems @xcite . the latter provide models for transport processes ranging from biological systems , like the motion of ribosomes along a _ m_-rna chain @xcite or processive motors walking along cytoskeletal filaments @xcite , to vehicular traffic @xcite . in this work , we focus on the steady - state properties of such one - dimensional transport models , for which the totally asymmetric simple exclusion process ( tasep ) has emerged as a paradigm ( for reviews see e.g. @xcite ) . there , particles move unidirectionally from left to right on a one - dimensional lattice , interacting through on - site exclusion . the entrance / exit rates at the open left / right boundary control the system s behavior ; tuning them , one encounters different non - equilibrium phases for the particle densities @xcite . intense theoretical research has been devoted to the classification of such non - equilibrium phenomena . for example , within the context of reaction - diffusion systems , there is strong evidence that phase transitions from an active to an absorbing state can be characterized in terms of only a few universality classes , the most important being the one of directed percolation ( dp ) @xcite . to search for novel critical behavior , fruitful results have been obtained by coupling two reaction - diffusion systems @xcite , each undergoing the active to absorbing phase transition . due to the coupling , the system exhibits a multicritical point with unusual critical behavior . we want to stress that already in equilibrium physics seminal insights have been gained by coupling identical systems . for instance , spin - ladders incorporate several heisenberg spin chains @xcite . there , quantum effects lead to a sensitive dependence on the chain number : for even ones a finite energy gap between the ground state and the lowest excitation emerges whereas gapless excitations dominate the low - temperature behavior if the number of spin chains is odd . in this work , we generalize the totally asymmetric exclusion process ( tasep ) in a way that particles possess two internal states ; we have recently published a short account of this work in ref . @xcite . allowing particles to occasionally switch from one internal state to the other induces a coupling between the latter ; indeed , the model may alternatively be regarded as two coupled taseps . when independent , each of them separately undergoes boundary - induced phase transitions @xcite . the coupling is expected to induce novel phenomena , which are the subject of the present work . ( @xmath0 ) enter with rates @xmath1 ( @xmath2 ) , move unidirectionally to the right within the lattice , may flip at rate @xmath3 and leave the system at rates @xmath4 ( @xmath5 ) , always respecting pauli s exclusion principle . [ cartoon_spin ] ] exclusion is introduced by allowing multiple occupancy of lattice sites only if particles are in different internal states . viewing the latter as spin-@xmath6 states , i.e spin - up ( @xmath7 ) and spin - down ( @xmath0 ) , this directly translates into pauli s exclusion principle ; see fig . [ cartoon_spin ] . indeed , the exclusion process presented in this work may serve as a model for semiclassical transport in mesoscopic quantum systems @xcite , like hopping transport in chains of quantum dots in the presence of an applied field @xcite . our model incorporates the quantum nature of the particles through pauli s exclusion principle , though phase coherence is ignored . a surprising analogy to a simple spintronics scheme , the datta - das spin field - effect transistor @xcite , holds . there , electrons move unidirectionally through a ferromagnetic metal or a semiconductor . the polarization of the electrons is controllable by a source for spin injection , a drain for spin extraction as well as a gate in the form of a tunable magnetic field that controls the strength of spin precession . in our model , this is mimicked by considering the spin - flip rate as a control parameter . the model is potentially relevant within biological contexts , as well . in intracellular traffic @xcite , molecular motors walking on parallel filaments may detach from one lane and attach on another , resulting in an effective switching between the lanes . in our model , identifying the two internal states with different lanes , one recovers a transport model on two lanes with simple site exclusion . in the same way , the system presented in this work serves as a highly simplified cartoon model of multi - lane highway traffic taking lane switching into account @xcite . and the lower one as lane @xmath8 . they possess individual entering rates , @xmath9 resp . @xmath10 as well as exiting rates , @xmath11 resp . @xmath12 . [ cartoon_two_lane ] ] significant insight into multi - lane traffic has been achieved ( see ref . @xcite and references therein ) . in particular , novel phases have been discovered in the case of indirect coupling , i.e. the velocity of the particles depends on the configuration on the neighboring lane @xcite . recently , models have been presented that allow particles to switch between lanes , and the transport properties have in part been rationalized in terms of an effective single lane tasep @xcite . there , the case of strong coupling has been investigated : the timescale of lane switching events is the same as of forward hopping . in our model , we explicitly want to ensure a competition between the boundary processes and the switching between the internal states . we therefore employ a mesoscopic scaling , i.e. we consider the case where the switching events are rare as compared to forward hopping . this is the situation encountered in intracellular traffic @xcite where motors nearly exclusively remain on one lane and switch only very rarely . in the context of spin transport , it corresponds to the case where forward hopping occurs much faster than spin precession ( weak external magnetic field ) . the outline of the present paper is the following . in sec . [ model ] we introduce the model in the context of spin transport as well as two - lane traffic . symmetries and currents are discussed , which play a key role in the following analysis . section [ mean - cont ] describes in detail the mean - field approximation and the differential equations for the densities obtained therefrom through a continuum limit . the mesoscopic scaling is motivated and introduced , the details of the analytic solution for the spatial density profiles being condensed in [ app_first_order ] . we obtain the generic form of the density profiles in sec . [ gen_dens ] , and compare our analytic results to stochastic simulations . we find that they agree excellently , suggesting the exactness of our analytic approach in the limit of large systems . as our main result , we encounter the polarization phenomenon , where the density profiles in the stationary non - equilibrium state exhibit localized ` shocks ' . namely , the density of one spin state changes abruptly from low to high density . the origin of this phenomenon is rationalized in terms of singularities in coupled differential equations we partition the full parameter space into three distinct regions , and observe a delocalization transition . the methods to calculate the phase boundaries analytically are developed simultaneously . section [ stoch_sim ] presents details on the stochastic simulations which we have carried out to corroborate our analytic approach . the central result of this work is then addressed in sec . [ phase_diag ] , where two - dimensional analytic phase diagrams are investigated . our analytic approach identifies the phases where the polarization phenomenon occurs , as well as the continuous and discontinuous transitions that separate the phases . the nature of the transitions is explained by the injection / extraction limited current which is conserved along the track . as a second remarkable feature of the model , we uncover multi - critical points , i.e. points where two lines of phase boundaries intersect or the nature of a phase transition changes from discontinuous to a continuous one . although multi - critical point are well - known in equilibrium statistical mechanics , a fundamental description for such a behavior for systems driven far from equilibrium still constitutes a major challenge . a brief summary and outlook concludes this work . in this section , we describe our model in terms of spin transport as well as two - lane traffic . though we will preferentially use the language of spins in the subsequent sections , the two - lane interpretation is of no lesser interest , and straightforwardly obtained . furthermore , we introduce two symmetries which are manifest on the level of the dynamical rules . we consider hopping transport on a one - dimensional lattice , composed of @xmath13 sites , with open boundaries , see fig . [ cartoon_spin ] . particles possess internal states , which we restrict to two different kinds ; adopting a spin notation , they are referred to as spin - up ( @xmath7 ) and spin - down ( @xmath0 ) . they enter at the left boundary at rates @xmath14 resp . @xmath2 , and move unidirectionally from left to the right through the lattice . the timescale is fixed by putting the rate for these hopping events to unity . within the bulk , particles may also flip their spin state , from spin - up to spin - down and back , at rate @xmath3 . finally , having reached the right boundary , particles may exit the system at rates @xmath15 resp . @xmath5 , depending on their spin state . we allow all of these processes only under the constraint of pauli s exclusion principle , meaning that every lattice site may at most be occupied by one particle of a given state . spin - up and spin - down thus may simultaneously occupy the same site , however two particles with identical spin polarization can not share a lattice site . in summary , our dynamical rules are the following : * at site @xmath16 ( left boundary ) , particles with spin - up ( spin - down ) may enter at rate @xmath14 ( @xmath2 ) * at site @xmath17 ( right boundary ) , particles with spin - up ( spin - down ) leave the lattice at rate @xmath15 ( @xmath5 ) * particles may hop at unit rate from site @xmath18 to the neighboring site @xmath19 for @xmath20 , i.e. within bulk * within bulk , particles can flip their spin state with rate @xmath3 , i.e. spin - up turns into spin - down and vice versa always respecting pauli s exclusion principle . processes ( i ) to ( iii ) constitute the totally asymmetric simple exclusion processes ( tasep ) for the two different states separately , while rule ( iv ) induces a coupling between them . indeed , when the spin - flip rate @xmath3 vanishes , we recover the trivial situation of two independent taseps , while we will show that a proper treatment of @xmath3 through a mesoscopic scaling induces nontrivial effects . having introduced our model in the language of semi - classical spin transport , where pauli s exclusion principle is respected while phase coherence completely ignored , we now want to show that it also describes transport with site exclusion on two parallel lanes . as schematically drawn in fig . [ cartoon_two_lane ] , we consider two parallel lanes , each consisting of @xmath13 sites , labeled as upper lane ( i ) and lower lane ( ii ) . they are identified with the internal states of the particles considered before : a particle with spin - up ( spin - down ) now corresponds to a particle on lane i ( lane ii ) . the processes ( i ) and ( ii ) describe entering of particles at lane i ( ii ) at rate @xmath21 ( @xmath22 ) and exiting of lane i ( ii ) at rate @xmath23 ( @xmath24 ) . due to ( iii ) , particles hop unidirectionally to the right on each individual lane ; at rate @xmath3 , they may switch from lane i to ii and back . pauli s exclusion principle translates into simple site exclusion : all the above processes are allowed under the constraint of admitting at most one particle per site . again , we clearly observe that it is process ( iv ) that couples two taseps , namely the ones on each individual lane , to each other . already on the level of the dynamical rules ( i)-(iv ) presented above , two symmetries are manifest that will prove helpful in the analysis of the system s behavior . we refer to the absence of particles with certain state as holes with the opposite respective state . considering their motion , we observe that the dynamics of the holes is governed by the identical rules ( i ) to ( iv ) , with `` left '' and `` right '' interchanged , i.e. with a discrete transformation of sites @xmath25 as well as rates @xmath26 . the system thus exhibits a _ particle - hole symmetry_. even more intuitively , the two states behave qualitatively identical . indeed , the system remains invariant upon changing spin - up to spin - down states and vice versa with a simultaneous interchange of @xmath27 and @xmath28 , constituting a _ spin symmetry _ ( in terms of the two - lane interpretation , it translates into a _ lane symmetry _ ) . + when analyzing the system s behavior in the five - dimensional phase space , constituted of the entrance and exit rates @xmath29 and @xmath3 , these symmetries allow to connect different regions in phase space , and along the way to simplify the discussion . in this section , we shall make use of the dynamical rules introduced above to set up a quantitative description for the densities and currents in the system . within a mean - field approximation , their time evolution is expressed through one - point functions only , namely the average occupations of a lattice site . such mean - field approximations have been successfully applied to a variety of driven diffusive systems , see e.g. ref . we focus on the properties of the non - equilibrium steady state , which results from boundary processes ( entering and exiting events ) as well as bulk ones ( hopping and spin - flip events ) . both types of processes compete if their time - scales are comparable ; we ensure this condition by introducing a _ mesoscopic scaling _ for the spin flip rate @xmath3 . our focus is on the limit of large system sizes @xmath13 , which is expected to single out distinct phases . to solve the resulting equations for the densities and currents , a continuum limit is then justified , and it suffices to consider the leading order in the small parameter , viz . the ratio of the lattice constant to system size . such a mesoscopic scaling has been already successfully used in @xcite in the context of tasep coupled to langmuir dynamics . + let @xmath30 resp . @xmath31 be the fluctuating occupation number of site @xmath19 for spin - up resp . spin - down state , i.e. @xmath32 if this site is occupied at time @xmath33 by a particle with the specified spin state and @xmath34 otherwise . performing ensemble averages , the expected occupation , denoted by @xmath35 and @xmath36 , is obtained . within a mean - field approximation , higher order correlations between the occupation numbers are neglected , i.e. we impose the factorization approximation @xmath37 equations of motion for the densities can by obtained via _ balance equations : _ the time - change of the density at a certain site is related to appropriate currents . the spatially varying _ spin current _ @xmath38 quantifies the rate at which particles of spin state @xmath7 at site @xmath18 hop to the neighboring site @xmath19 . within the mean - field approximation , eq . ( [ decoupling ] ) , the current is expressed in terms of densities as @xmath39 \ , , \qquad i \in \{2,\ldots , l\ } \ , , \label{curr_a}\ ] ] and similarly for the current @xmath40 . the sum yields the total _ particle current _ @xmath41 . due to the spin - flip process ( iv ) , there also exists a _ leakage current _ @xmath42 from spin - up state to spin - down state . within mean - field @xmath43\ , , \ ] ] and similarly for the leakage current @xmath44 from spin - down to spin - up state . now , for @xmath45 we can use balance equations to obtain the time evolution of the densities , @xmath46 this constitutes an exact relation . together with the mean field approximation for the currents , eqs . ( [ curr_a ] , [ curr_updown ] ) , one obtains a set of closed equations for the local densities @xmath47-\rho_{i}^\uparrow(t)[1-\rho_{i+1}^\uparrow(t)]+\omega\rho_i^\downarrow(t ) -\omega\rho_i^\uparrow(t ) \ , . \label{occ_num}\end{aligned}\ ] ] at the boundaries of the track , the corresponding expressions involve also the entrance and exit events , which are again treated in the spirit of a mean - field approach @xmath48 -\rho_{1}^\uparrow(t)[1-\rho_{2}^\uparrow(t ) ] + \omega\rho_1^\downarrow(t)-\omega\rho_1^\uparrow(t)\ , , \label{bound_1 } \\ \frac{\textnormal{d}}{\textnormal{d } t } \rho_l^\uparrow(t)=~&\rho_{l-1}^\uparrow(t)[1-\rho_l^\uparrow(t ) ] -\beta^\uparrow\rho_{l}^\uparrow(t)+\omega\rho_l^\downarrow(t)-\omega\rho_l^\uparrow(t ) \label{bound_n}\ , .\end{aligned}\ ] ] due to the spin symmetry , i.e. interchanging @xmath7 and @xmath0 , an analogous set of equations hods for the time evolution of the density of particles with spin - down state . + in the stationary state , the densities @xmath49 do not depend on time @xmath33 , such that the time derivatives in eqs . ( [ occ_num])-([bound_n ] ) vanish . therefrom , we immediately derive the spatial conservation of the particle current : indeed , summing eq . ( [ occ_curr ] ) with the corresponding equation for the density of spin - down states yields @xmath50 such that the particle current does not depend on the spatial position @xmath19 . note that this does not apply to the individual spin currents , they _ do _ have a spatial dependence arising from the leakage currents . + in a qualitative discussion , let us now anticipate the effects that arise from the non - conserved individual spin currents as well as from the conserved particle current . the latter has its analogy in tasep , where the particle current is spatially conserved as well . it leads to two distinct regions in the parameter space : one where the current is determined by the left boundary , and the other where it is controlled by the right one . both regions are connected by the discrete particle - hole symmetry . thus , in general , discontinuous phase transitions arise when crossing the border from one region to the other . in our model , we will find similar behavior : the particle current is either determined by the left or by the right boundary . again , both regions are connected by the discrete particle - hole symmetry , such that we expect discontinuous phase transitions at the border between both . except for a small , particular region in the parameter space , this behavior is captured quantitatively by the mean - field approach and the subsequent analysis , which is further corroborated by stochastic simulations . the phenomena linked to the particular region will be presented elsewhere @xcite . + on the other hand , the non - conserved spin currents may be compared to the current in tasep coupled to langmuir kinetics ; see refs . @xcite . due to attachment and detachment processes , the in - lane current is only weakly conserved , allowing for a novel phenomena , namely phase separation into a low - density and a high - density region separated by a localized domain wall . the transitions to this phase are continuous considering the domain wall position @xmath51 as the order parameter . in our model , an analogous but even more intriguing phase will appear as well , with continuous transitions being possible . phases and corresponding phase transitions are expected to emerge in the limit of large system size , @xmath52 , which therefore constitutes the focus of this work . we expect interesting phase behavior to arise from the coupling of spin - up and spin - down states via spin - flip events , in addition to the entrance and exit processes . clearly , if spin - flips occur on a fast time - scale , comparable to the hopping events , the spin degree of freedom is relaxed , such that the system s behavior is effectively the one of a tasep . previous work on related two - lane models @xcite focused on the physics in that situation . in this work , we want to highlight the dynamical regime where coupling through spin - flips is present , however not sufficiently strong to relax the system s internal degree of freedom . in other words , we consider physical situations where spin - flips occur on the same time - scale as the entrance / exit processes . defining the _ gross _ spin - flip rate @xmath53 yields a measure of how often a particle flips its spin state while traversing the system . to ensure competition between spin - flips with boundary processes , a _ mesoscopic scaling _ of the rate @xmath3 is employed by keeping @xmath54 fixed , of the same order as the entrance / exit rates , when the number of lattice sites becomes large @xmath52 . the total length of the lattice will be fixed to unity and one may define consistently the lattice constant @xmath55 . in the limit of large systems @xmath56 , a _ continuum limit _ is anticipated . we introduce continuous functions @xmath57 resp . @xmath58 through @xmath59 resp . @xmath60 at the discrete points @xmath61 . expanding these to first order in the lattice constant , @xmath62 the difference equations ( [ occ_num])-([bound_n ] ) turn into differential equations . observing that @xmath63 is already of order @xmath64 , we find that the zeroth order of eq . ( [ occ_num ] ) vanishes , and the first order in @xmath64 yields @xmath65\partial_x\rho^\uparrow(x ) + \omega\rho^\downarrow(x)-\omega\rho^\uparrow(x)=0 \label{cont_a } \ , .\ ] ] similarly , the same manipulations for @xmath66 yield @xmath67\partial_x\rho^\downarrow(x ) + \omega\rho^\uparrow(x)-\omega\rho^\downarrow(x)=0 \ , . \label{cont_b}\ ] ] the expansion of eqs . ( [ bound_1 ] ) and ( [ bound_n ] ) in powers of @xmath64 , yields in zeroth order @xmath68 which impose _ boundary conditions_. since two boundary conditions are enough to specify a solution of the coupled first order differential equations , the system is apparently over - determined . of course , the full analytic solution , i.e. where all orders in @xmath64 are incorporated , will be only _ piecewise _ given by the first - order approximation , eqs . ( [ cont_a])-([bound_cond ] ) . between these branches , the solution will depend on higher orders of @xmath64 , therefore , these intermediate regions scale with order @xmath64 and higher . they vanish in the limit of large systems , @xmath56 , yielding _ domain walls _ or _ boundary layers_. + let us explain the latter terms . at the position of a domain wall , situated in bulk , the density changes its value discontinuously , from one of a low - density region to one of a high - density . boundary layers are pinned to the boundaries of the system . there as well , the density changes discontinuously : from a value that is given by the corresponding boundary condition to that of a low- or high - density region which is imposed by the opposite boundary . in the following , we reflect important properties of the system , symmetries and currents , on the level of the first - order approximation , eqs . ( [ cont_a])-([bound_cond ] ) . the explicit solution of the latter can be found in [ app_first_order ] . the particle - hole symmetry , already inferred from the dynamical rules , now takes the form @xmath69 interchanging @xmath7 and @xmath0 in the densities as well as the in- and outgoing rates yields the spin symmetry , @xmath70 the individual spin currents as well as the particle current have been anticipated to provide further understanding of the system s behavior . in the continuum limit the zeroth order of the spin currents is found to be @xmath71 $ ] , such that eqs . ( [ cont_a ] ) , ( [ cont_b ] ) may be written in the form @xmath72~ , \quad \partial_x j^\downarrow=\omega[\rho^\uparrow-\rho^\downarrow]\ , . \label{cont_curr}\ ] ] the terms on the right - hand side , arising from the spin - flip process ( iv ) , are seen to violate the spatial conservation of the spin currents . however , due to the mesoscopic scaling of the spin flip rate @xmath3 , the leakage currents between the spin states are only weak , see eq . ( [ curr_updown ] ) , locally tending to zero when @xmath73 , such that the spin currents vary _ continuously _ in space . this finding imposes a condition for the transition from one branch of first - order solution to another , as described above : such a transition is only allowed when the corresponding spin currents are continuous at the transition point , thus singling out distinct positions for a possible transition . + finally , summing the two equations in eq . ( [ cont_curr ] ) yields the spatial conservation of the particle current : @xmath74 . the parameter space of our model , spanned by the five rates @xmath29 , and @xmath54 , is of high dimensionality . however , in this section , we show that it can be decomposed into only three basic distinct regions : the maximal - current region ( mc ) as well as the injection - limited ( in ) and the extraction - limited one ( ex ) . while trivial phase behavior occurs in the mc region , our focus is on the in and ex region ( connected by particle - hole symmetry ) , where a striking polarization phenomenon occurs . the generic phase behavior in these regions is derived , exhibiting this effect . effectively acts as @xmath75 . the analytic predictions correspond to the solid lines , the results from stochastic simulations for @xmath76 are indicated by the wiggly line . with increasing spatial position , the densities approach a common value @xmath77 . the parameters used are @xmath78.[eff_rates ] ] the entrance and exit rates as well as the carrying capacity of the bulk impose restrictions on the particle current . for example , the capacity of the bulk limits the individual spin currents @xmath79 to a maximal values of @xmath80 . the latter occurs at a density of @xmath6 , as seen from the previous result @xmath81 $ ] . to illustrate the influence of the injection and extraction rates , we first consider an `` open '' right boundary i.e. @xmath82 . particles then leave the system unhindered , such that only the entrance rates may limit the particle current . provided one of these rates , say @xmath14 , exceeds the value @xmath6 , the current of the corresponding state ( @xmath7 ) is limited by the capacity of the bulk to a value of @xmath80 in the vicinity of the left boundary . a boundary layer thus forms in the density profile of spin - up state at the left boundary , connecting the value of the injection rate @xmath14 to the value @xmath6 . up to this boundary layer , the density profile @xmath57 is identical to the one where @xmath14 takes a value of @xmath6 , c.f . [ eff_rates ] . similar reasoning holds for the extraction rates @xmath83 . they as well behave effectively as @xmath6 when exceeding this value . to treat these findings properly , we introduce the _ effective rates _ @xmath84\ , , \\ \beta_{\textnormal{\scriptsize eff}}^{\uparrow(\downarrow)}&=&\textnormal { min}\big[\beta^{\uparrow(\downarrow)},\frac{1}{2}\big ] \ , . \end{aligned}\ ] ] the system s bulk behavior will only depend on them , and , in particular , remain unaffected when a rate is varied at values exceeding @xmath6 . equipped with these results , in the case of an `` open '' right boundary , the spin currents in the vicinity of the left boundary are given by @xmath85 resp . @xmath86 , resulting in a particle current @xmath87 imposed by the injection rates : @xmath88 . the analogous relations , with the injection and extraction rates interchanged , hold for the case of an `` open '' left boundary , @xmath89 . the particle current is then controlled by the right boundary : @xmath90 . in general , depending on which imposes the stronger restriction , either the left or the right boundary limits the particle current : @xmath91 . indeed , @xmath92 holds except for an anomalous situation , where the current is lower than this value . , discussed in sec . [ phase_diag ] ] depending on which of both cases applies , two complementary regions in phase space are distinguished : @xmath93 is termed _ injection - limited region _ ( in ) , while @xmath94 defines the _ extraction - limited region _ ( ex ) . since they are connected by discrete particle - hole symmetry , we expect _ discontinuous phase transitions _ across the border between both , to be referred as _ in - ex boundary_. right at the in - ex boundary , the system exhibits coexistence of low- and high - density phases , separated by domain walls . interestingly , this phase coexistence emerges on _ both _ lanes ( states ) , which may be seen as follows . recall that a domain wall concatenates a region of low and another of high density however , while the densities exhibit a discontinuity , the spin currents must be continuous . in other words , the spin currents , and therefore the particle currents , imposed by the left and right boundary must match each other . this yields the condition @xmath95 , which is nothing but the relation describing the in - ex boundary . actually , what we have shown with this argument is that domain walls on both lanes ( states ) are _ at most _ feasible at the in - ex boundary . however , it turns out that there , they do indeed form , and are delocalized . we refer to our forthcoming publication @xcite for a detailed discussion of this phenomenon . away from the in - ex boundary , it follows that at most on one lane ( state ) a domain wall may appear . when both entrance rates @xmath96 as well as both exit rates @xmath97 exceed the value @xmath6 , the particle current is limited by neither boundary , but only through the carrying capacity of the bulk , restricting it to twice the maximal value @xmath80 of the individual spin currents : @xmath98 . the latter situation therefore constitutes the _ maximal current region _ ( mc ) . as we have seen in the previous section , particularly simple density profiles emerge in the mc region . there , up to boundary layers , the density profiles remain constant at a value @xmath6 for each spin state . another special region in parameter space is the in - ex boundary , characterized by the simultaneous presence of domain walls in both spin states , as we discuss elsewhere @xcite . ) red ( @xmath76 ) lines . they piecewise obey the first order approximations ( black ) , grey lines indicate continuations of the latter into regions where the densities are no longer given by them . the parameters are @xmath99 , and @xmath100.[el ] ] away from these regions , the generic situation for the density profiles is illustrated in fig . [ el ] . here , we have considered parameters belonging to the in region ; the behavior in the ex region follows from particle - hole symmetry . a domain wall emerges for one spin state and a boundary layer for the other one . for specificity , we consider a domain wall for the spin - up state , the other situation is obtained from spin symmetry . the density profiles @xmath101 close to the left boundary are given by the solution of the first - order differential equations ( [ cont_a]),([cont_b ] ) , obeying the left boundary conditions @xmath102 and @xmath103 . for the density profiles @xmath104 in the vicinity of the right boundary , we use the fact that the current in bulk is determined by the injection rates , @xmath105 ( which defines the in region ) . therefore , the densities satisfy right boundary conditions which are given by @xmath106 ; and @xmath107 is found from the conservation of the particle current : @xmath108 \ , .\end{aligned}\ ] ] at some point @xmath51 in bulk , the left and right solutions have to be concatenated by a domain wall for spin - up . to determine the position @xmath51 of this domain wall , we use the _ continuity _ of the spin currents ; see fig . [ el](b ) . was seen to cause a only slowly varying spatial dependence ; in the continuum limit , the spin currents are continuous . ] this continuity condition singles out a distinct spatial position for the domain wall : denote by @xmath109 the value of the density to the left of @xmath51 , and @xmath110 the value to the right . from @xmath111 together with @xmath112 , we arrive at the condition @xmath113 for the domain wall position . from the conservation of the particle current @xmath114 , it follows that the density @xmath66 is continuous at the position @xmath51 . when considering the internal states as actual spins , the appearance of a domain wall in the density profile of one of the spin states results in a _ spontaneous polarization phenomenon_. indeed , while both the density of spin - up and spin - down remain at comparable low values in the vicinity of the left boundary , this situation changes upon crossing the point @xmath51 . there , the density of spin - up jumps to a high value , while the density of spin - down remains at a low value , resulting in a polarization in this region . comparing the generic phase behavior to the one of tasep , we observe that the in region can be seen as the analogue to the low - density region there : within both , a low - density phase accompanied by a boundary layer at the right boundary arises . following these lines , the ex region has its analogue in the high - density region , while the mc region is straightforwardly generalized from the one of tasep . furthermore , the delocalization transition across the in - ex boundary is similar to the appearance of a delocalized domain wall at the coexistence line in tasep . in the generic situation of fig . [ el ] , the density of spin - down is in a homogeneous low - density ( ld ) state , while for spin - up , a low - density and a high - density region coexist . we refer to the latter as the @xmath115 phase , as the phase separation arises within the in region , to be contrasted from a @xmath116 phase which may arise within the ex region . clearly , the @xmath115 phase is only present if the position @xmath51 of the domain wall lies within bulk . tuning the system s parameter , it may leave the system through the left or right boundary , resulting in a homogeneous phase . indeed , @xmath117 marks the transition between the @xmath115 phase and the pure ld state , while at @xmath118 the density changes from the @xmath115 to a homogeneous high - density ( hd ) state . regarding the domain wall position @xmath51 as an order parameter , these transitions are continuous . implicit analytic expressions for these phase boundaries , derived in the following , are obtained from the first - order approximation , eqs . ( [ cont_a ] ) and ( [ cont_b ] ) . spin symmetry yields the analogous situation with a domain wall appearing in the density profile of spin - down , while particle - hole symmetry maps it to the ex region , where a pure hd phase arises for one of the spins . discontinuous transitions accompanied by delocalized domain walls appear at the submanifold of the in - ex boundary ( see @xcite for a detailed discussion ) . the phase boundaries may be computed from the condition @xmath118 and @xmath117 in the situation of fig . [ el ] . consider first the case of @xmath118 . there , the density profiles are fully given by the first - order approximation @xmath119 satisfying the boundary conditions at the right . the condition ( [ cond_xw ] ) translates to @xmath120 which yields an additional constraint on the system s parameters . this defines the hyper - surface in the in region where @xmath118 occurs , and thus the phase boundary between the @xmath115 and the pure hd phase . + similarly , if @xmath117 , the densities follow the left solution @xmath121 , determined by the left boundary conditions , within the whole system . from eq . ( [ cond_xw ] ) we obtain @xmath122 again , the latter is a constraint on the parameters and defines the hyper - surface in the in region where @xmath117 is found , being the phase boundary between the @xmath115 and the homogeneous ld phase . the conditions ( [ xw=0 ] ) , ( [ xw=1 ] ) yield implicit equations for the phase boundaries . the phase diagram is thus determined up to solving algebraic equations , which may be achieved numerically . further insight concerning the phase boundaries is possible and may be obtained analytically , which we discuss next . first , we note that in the case of equal injection rates , @xmath123 , the density profiles in the vicinity of the left boundary are constant . if in addition @xmath124 , we observe from eq . ( [ xw=1 ] ) that a domain wall at @xmath117 emerges . therefore , this set of parameters always lies on the phase boundary @xmath117 , independent of the value of @xmath54 . second , we investigate the phase boundary determined by @xmath118 . comparing with fig . [ el ] , we observe that the first - order approximation @xmath125 for the density of spin - up may reach the value @xmath75 at a point which is denoted by @xmath126 : @xmath127 . this point corresponds to a branching point of the first - order solution . increasing @xmath54 , the value of @xmath126 increases as well . the domain wall in the density of spin - up can only emerge at a value @xmath128 . at most , @xmath129 , in which case a domain wall with infinitesimal small height arises . for the phase boundary specified by @xmath118 , this implies that it only exists as long as @xmath130 . the case @xmath131 corresponds to a domain wall of infinitesimal height , which is only feasible if @xmath132 . now , for given rates @xmath133 , the condition @xmath134 yields a critical rate @xmath135 , depending on the rates @xmath136 . the situation @xmath118 can only emerge for rates @xmath137 . varying the rates @xmath138 and @xmath15 , the critical rate @xmath135 changes as well . in [ app_first_order ] , we show that its largest value occurs at @xmath139 . they yield the rate @xmath140 , which is calculated to be @xmath141 the critical @xmath135 are lying in the interval between @xmath142 and @xmath143 : @xmath144 $ ] , and all values in this interval in fact occur . the rate @xmath143 defines a _ scale _ in the spin - flip rate @xmath54 : for @xmath145 , the phase boundary determined by @xmath118 exists , while disappearing for @xmath146 . third , we study the form of the phase boundaries for large @xmath54 , meaning @xmath147 . in this case , the phase boundary specified by @xmath118 is no longer present . furthermore , it turns out that in this situation , the densities close to the left boundary quickly approximate a common value @xmath77 . the latter is found from conservation of the particle current : @xmath148 . we now consider the implications for the phase boundary determined by @xmath117 . with @xmath149 , ( [ xw=1 ] ) turns into @xmath150 , yielding @xmath151 this condition specifies the phase boundary @xmath117 , asymptotically for large @xmath54 . it constitutes a simple quadratic equation in the in- and outgoing rates , independent of @xmath5 , and contains the set @xmath152 . to confirm our analytic findings from the previous section , we have performed stochastic simulations . the dynamical rules ( i)-(iv ) described in subsec . [ dynamics ] were implemented using random sequential updating . in our simulations , we have performed averages over typically @xmath153 time steps , with @xmath154 steps of updating between successive ones . finite size scaling singles out the analytic solution in the limit of large system sizes , as exemplified in figs . [ eff_rates ] and [ el ] . for all simulations , we have checked that the analytic predictions are recovered upon approaching the mesoscopic limit . we attribute the apparent exactness of our analytic approach in part to the exact current density relation in the steady state of the tasep @xcite . the additional coupling of the two taseps in our model is only weak : the local exchange between the two states vanishes in the limit of large system sizes . correlations between them are washed out , and mean - field is recovered . the observed exactness of the analytic density profiles within the mesoscopic limit implies that our analytic approach yields exact phase diagrams as well . the latter are the subject of the subsequent section . in this section , we discuss the phase behavior on two - dimensional cuts in the whole five - dimensional parameter space . already the simplified situation of equal injection rates , @xmath123 , yields interesting behavior . there as well as in the general case , we investigate the role of the spin - flip rate @xmath54 by discussing the situation of small and large values of @xmath54 . for simplicity , we start our discussion of the phase diagram with equal injection rates , @xmath123 . then , the spin polarization phenomenon , depicted in fig . [ el ] , becomes even more striking . starting from equal densities at the left boundary , and hence zero polarization , spin polarization suddenly switches on at the domain wall position @xmath51 . the particular location of @xmath51 is not triggered by a cue on the track , but tuned through the model parameters . the phase transitions from ld to the @xmath115 arising in the in region take a remarkably simple form . their location is found from @xmath117 , and is determined by eq . ( [ xw=1 ] ) ( if phase coexistence arises for spin - up ) . since @xmath155 for @xmath156 , eq . ( [ xw=1 ] ) turns into @xmath157 . the latter transition line intersects the in - ex boundary , given by @xmath95 , at @xmath158 , i.e. at the point where all entrance and exit rates coincide . at this _ multicritical point _ @xmath159 , a continuous line intersects a discontinuous one . the same transition in the density of spin - down state is , from similar arguments , located at @xmath160 , and also coincides with the in - ex boundary in @xmath159 . neither the multicritical point @xmath159 nor these phase boundaries depend on the magnitude of the gross spin flip rate @xmath54 . therefore , qualitatively tuning the system s state is possible only upon changing the injection or extraction rates . the other phase transitions within the in region , namely from the hd to the @xmath115 phase , are more involved . the analytic solution ( [ x_rhoa ] ) , ( [ x_rhob ] ) has to be considered together with the condition ( [ xw=0 ] ) for the transition . however , at the end of subsec . [ phases_bound ] , we have found that these transitions ( determined by @xmath118 ) disappear for sufficiently large @xmath146 . and large @xmath54 . the phases of the densities of spin - up ( spin - down ) state are shown in ( a ) resp . ( b ) for a value @xmath161 . at a multicritical point @xmath159 , continuous lines ( thin ) intersect with a discontinuous line ( bold ) , the in - ex boundary . if @xmath162 , the maximal current phase appears for spin - up , see ( c ) , as well for spin - down , drawn in ( d ) . in the first situation , the switching rate is @xmath163 , while @xmath164 in the second.[phasediag_eqal_largeomega ] ] in the situation of large @xmath146 , phase transitions arising from @xmath118 in the in region or from the analogue in the ex region do not emerge , as discussed at the end of subsec . [ phases_bound ] . we have drawn resulting phase diagrams in fig . [ phasediag_eqal_largeomega ] , showing the phase of spin - up ( spin - down ) in the left ( right ) panels , depending on @xmath165 and @xmath15 . along the in - ex boundary , being the same line ( shown as bold ) in the left and right panels , a delocalization transitions occur . at the multicritical point @xmath159 , it is intersected by continuous lines emerging within the in resp . the ex region . when @xmath166 , a maximal current(mc ) phase emerges in the upper right quadrant , see fig . [ phasediag_eqal_largeomega ] ( c)-(d ) . to illustrate the system s phase behavior , let us consider what happens along a horizontal line in the phase diagrams ( a ) and ( b ) , at a value @xmath167 . at such a line , for small values of @xmath165 , both spin states are in low - density ( ld ) phases . upon crossing a certain value of @xmath165 , a domain wall enters at @xmath117 in the spin - down density profile . then , spin - down exhibits phase coexistence ( @xmath115 ) , while spin - up remains in a ld phase . further increasing @xmath165 , the bold line is reached , where delocalized domain walls arise in both spin states . for larger values of @xmath165 , a localized domain wall emerges for spin - up ( implying a @xmath116 phase ) , and a pure hd phase for spin - down . if @xmath165 is further increased , the domain wall in the spin - up density profile leaves the system through the left boundary ( at @xmath118 ) , and pure hd phases remain for both spin states . while we have found the transitions within the in region by simple expressions in the previous subsection , the ones emerging in the ex region are more complex and involve the full analytic solution ( [ x_rhoa ] ) , ( [ x_rhob ] ) . their most notable feature is that the width of the corresponding coexistence phase decreases with increasing spin - flip rate @xmath54 , until it finally vanishes in the limit @xmath168 . this may be seen by considering the analogue of eq . ( [ xw1_largeomega ] ) in the ex region , which describes the phase boundary as it is asymptotically approached when @xmath169 : @xmath170 it coincides with the in - ex boundary . ( a ) , ( b ) , but with @xmath54 decreased to a small value , @xmath171 . additional phase transitions emerge in the in as well as the ex region , accompanied by multicritical points @xmath172 and @xmath173 . caused by them , phase transitions do no longer appear across some parts of the in - ex boundary , which is there shown as dashed line . [ phasediag_eqal_smallomega ] ] as discussed at the end of subsec . [ phases_bound ] , when @xmath174 , the appearance of additional phase transitions becomes possible . for example , within the in region , the situation @xmath118 may emerge . it describes the transition from the hd to the @xmath115 phase ; the analogue occurs in the ex region . in fig . [ phasediag_eqal_smallomega ] , we show resulting phase diagrams for the spin - up ( left panel ) and spin - down ( right panel ) , resp .. the additional transition lines intersect the in - ex boundary ( bold ) at additional multicritical points @xmath172 and @xmath173 . also , they partly substitute the in - ex boundary as a phase boundary : across some parts of the latter , phase transitions do not arise . this behavior reflects the _ decoupling _ of the two states for decreasing spin - flip rate @xmath54 . indeed , for @xmath175 , the states become more and more decoupled , such that the in - ex boundary , involving the combined entrance and exit rates of both states , loses its significance . although the shapes of most of the transition lines appearing in the phase diagrams shown in fig . [ phasediag_eqal_smallomega ] are quite involved , they also exhibit simple behavior . _ pairwise _ , namely one line from a transition in spin - up and another from a related transition in spin - down states , they intersect the in - ex boundary in the same multicritical point . this intriguing phenomenon may be understood by considering the multicritical points : e.g. , at @xmath159 , the transition line from the ld to the @xmath115 phase in the density profile of spin - up intersects the in - ex boundary , which implies that there we have a domain wall in the density profile of spin - up at the position @xmath117 . however , being on the in - ex boundary , the condition @xmath95 implies that in this situation a domain wall forms as well in the density of spin - down states , also located at @xmath117 . consequently , @xmath159 also marks the point where the transition line specified by @xmath117 for spin - down states intersects the in - ex boundary . due to the special situation of equal entrance rates , one more pair of lines intersects in this point . similarly , at @xmath172 , the transition line from the hd to @xmath115 phase in the density profile of spin - up intersects the in - ex boundary , such that a domain wall forms in the density of spin - up at @xmath118 . again , as @xmath95 holds on the in - ex boundary , this implies the formation of a domain wall in the density of spin - down at @xmath117 , corresponding to the transition from the ld to the @xmath116 phase for spin - down in the ex region . having focused on the physically particularly enlightening case of equal entering rates in the previous subsection , we now turn to the general case . to illustrate our findings , we show phase diagrams depending on the injection and extraction rates for spin - up states , @xmath14 and @xmath15 . similar behavior as for equal entrance rates is observed . the multicritical point @xmath159 now splits up into two distinct points @xmath176 and @xmath177 . . lines of continuous transitions ( thin ) within the in resp . ex region intersect the delocalization transition line ( bold ) in multicritical points @xmath176 resp . @xmath177 . both of these points appear in ( a ) , ( b ) ( @xmath178 ) while only @xmath177 is present in ( c ) , ( d ) ( @xmath179 ) and @xmath176 alone in ( e ) , ( f ) ( @xmath180 ) , yielding different topologies . [ phasediag_largeomega1 ] ] and @xmath181 [ upper right quadrant in ( a ) , ( b ) ] , delocalized domain walls form in the density profiles of both spin states . if instead @xmath182 , the maximal current phase emerges , see ( c ) , ( d).[phasediag_largeomega2 ] ] again , large @xmath54 prohibit the emergence of the phase transition from the hd to the @xmath115 phase in the in region as well as from the ld to the @xmath116 phase within the ex region , see end of subsec . [ phases_bound ] . in this paragraph , we consider phase diagrams which are approached asymptotically when @xmath169 . convergence is fast in @xmath54 , and the asymptotic phase boundaries yield an excellent approximation already for @xmath183 . the transition from ld to the @xmath115 phase in the in region asymptotically takes the form of eq . ( [ xw1_largeomega ] ) , and the one from hd to the @xmath116 phase in the ex region is obtained by particle - hole symmetry . all phase boundaries , including the in - ex boundary , are thus given by simple quadratic expressions . phase diagrams with different topologies that can emerge are exhibited in fig . [ phasediag_largeomega1 ] and [ phasediag_largeomega2 ] . as in the previous subsection , we show the phases of spin - up ( spin - down ) states on the left ( right ) panels . the phase boundaries between the ld and the @xmath115 phase in the in region for spin - up as well as spin - down both intersect the in - ex boundary in a multicritical point @xmath176 , being located at @xmath184 . similarly , the lines of continuous transitions within the ex region both coincide with the in - ex boundary in a multicritical point @xmath177 , which is situated at @xmath123 . note that the phase transitions emerging in the density profile of spin - down within the in region do not depend on @xmath15 , thus being horizontal lines in the phase diagrams . within the ex region they are independent of @xmath14 , yielding vertical lines . for @xmath185 , fig . [ phasediag_largeomega1 ] shows different topologies of phase diagrams , which only depend on which of the multicritical points @xmath176 , @xmath177 is present . if both appear , see fig . [ phasediag_largeomega1 ] ( a ) , ( b ) , the @xmath115 and the @xmath116 phase for spin - up are adjacent to each other , separated by the in - ex boundary . although in both phases localized domain walls emerge , their position changes discontinuously upon crossing the delocalization transition . e.g. , starting within the @xmath115 phase , the domain wall delocalizes when approaching the in - ex boundary , and , having crossed it , relocalizes again , but at a different position . when @xmath186 , a subtlety emerges , see fig . [ phasediag_largeomega2 ] ( a ) , ( b ) . if both @xmath187 and @xmath188 , i.e. in the upper right quadrant of the phase diagrams , these rates effectively act as @xmath6 , and the condition @xmath95 for the in - ex boundary is fulfilled _ in this whole region . _ therefore , delocalized domain walls form on both lanes within this region , as is confirmed by our stochastic simulations @xcite . the maximal current phase ( mc ) emerges when all rates exceed or equal the value @xmath6 , corresponding to the upper left quadrant of the phase diagrams in fig . [ phasediag_largeomega2 ] ( c ) , ( d ) . . the nodal point @xmath189 remains unchanged when @xmath54 is varied . the appearance of the multicritical point @xmath172 is accompanied by the non - occurrence of phase transitions across parts of the in - ex boundary , then shown as dashed line . the multicritical point @xmath177 emerges in ( a ) , ( b ) , but not in the situation of ( c ) ( , ( d ) . parameters are @xmath190 in ( a ) , ( b ) and @xmath191 in ( c ) , ( d ) . [ phasediag_smallomega ] ] when @xmath174 , the transitions from ld to @xmath115 within the in region as well as the analogue in the ex region are possible . as in the case of equal entering rates , the corresponding transition lines pairwise intersect the in - ex boundary in multicritical points @xmath172 and @xmath173 . as all transitions between phases of the spin - down density within the in region are independent of @xmath15 , the corresponding lines are simply horizontal ; and within the ex region , their independence of @xmath14 implies that they yield vertical lines . the phase diagram for the density of spin - down is thus easily found from the in - ex boundary given by @xmath95 together with the locations of the multicritical points @xmath176 , @xmath177 , @xmath172 and @xmath173 . the latter follow from the intersection of phase transition lines for the density of spin - up , involving the whole analytic solution ( [ x_rhoa ] ) , ( [ x_rhob ] ) , with the in - ex boundary . in fig . [ phasediag_smallomega ] two interesting topologies that may arise are exemplified . induced by the presence of the multicritical point @xmath172 , phase transitions do not occur across all the in - ex boundary , which is then only shown as dashed line . in fig . [ phasediag_smallomega ] ( a ) , ( b ) , the points @xmath177 and @xmath172 are present . the @xmath115 phase for spin - up intervenes the ld and the hd phase ; the @xmath116 phase for spin - up is also present , though very tiny . in the phase diagram of spin - down , the @xmath115 phase intervenes the ld and the hd phase accompanied by continuous as well as discontinuous transitions . again , the presence of the multicritical points induces the topology ; e.g. in fig . [ phasediag_smallomega ] ( c ) , ( d ) , only @xmath172 appears . for the discussion of the possible topologies , we encounter the restriction that @xmath176 and @xmath172 can not occur together , as well as @xmath177 and @xmath173 exclude each other ( otherwise , the lines determined by @xmath118 and @xmath117 would cross ) . we now discuss the influence of the spin - flip rate @xmath54 on the continuous transition lines for spin - up . in subsec . [ phases_bound ] the manifold defined by @xmath192 was found to be a sub - manifold of the phase boundary specified by @xmath117 in the in region . independent of @xmath54 , the point @xmath192 , denoted by @xmath189 , thus lies on the boundary between the ld and the @xmath115 phase ( determined by @xmath117 ) . for large @xmath54 , this boundary approaches the one given by eq . ( [ xw1_largeomega ] ) . + regarding the transition from the hd to the @xmath115 within the in region ( determined by @xmath118 ) , subsec . [ phases_bound ] revealed that for increasing @xmath54 it leaves the in region at a critical transfer rate @xmath193 . in the limit @xmath194 , the densities @xmath57 and @xmath58 approach constant values , and both the curve @xmath117 as @xmath118 for spin - up in the in region approach the line @xmath195 for @xmath196 ; the phase in the upper right quadrant in the phase diagram converges to the mc phase , such that in this limit , the case of two uncoupled taseps is recovered . we have presented a detailed study of an exclusion process with internal states recently introduced in @xcite . the totally asymmetric exclusion process ( tasep ) has been generalized by assigning two internal states to the particles . pauli s exclusion principle allows double occupation only for particles in different internal states . occasional switches from one internal state to the other induce a coupling between the transport processes of the separate states . such a dynamics encompasses diverse situations , ranging from vehicular traffic on multiple lanes to molecular motors walking on intracellular tracks and future spintronics devices . we have elaborated on the properties of the emerging non - equilibrium steady state focusing on density and current profiles . in a mesoscopic scaling of the switching rate between the internal states , nontrivial phenomena emerge . a localized domain wall in the density profile of one of the internal states induces a spontaneous polarization effect when viewing the internal states as spins . we provide an explanation based on the weakly conserved currents of the individual states and the current - density relations . a quantitative analytic description within a mean - field approximation and a continuum limit has been developed and solutions for the density and current profiles have been presented . a comparison with stochastic simulations revealed that our analytic approach becomes exact in the limit of large system sizes . we have attributed this remarkable finding to the exact current - density relation in the tasep , supplemented by the locally weak coupling of the two taseps appearing in our model : @xmath197 in the limit of large system sizes . local correlations between the two internal states are thus obliterated , as particles hop forward on a much faster timescale than they switch their internal state . furthermore , the parameter regions that allow for the formation of a localized domain wall have been considered . analytic phase diagrams for various scenarios , in particular the case of equal entrance rates , have been derived . the phase diagrams have been found to exhibit a rich structure , with continuous as well as discontinuous non - equilibrium phase transitions . the discontinuous one originates in the conserved particle current , which is either limited by injection or extraction of particles . at the discontinuous transition between both regimes , delocalized domain walls emerge in the density profiles of _ both _ internal states . multicritical points appear at the intersections of different transition lines organizing the topology of the phase diagrams . two classes of multicritical points are identified , one of them arises only for sufficiently small gross spin - flip rate @xmath174 . the value @xmath143 , calculated analytically , provides a natural scale for the rate @xmath54 . it would be of interest to see which of the described phenomena qualitatively remain when generalizing the model to include more than two internal states . indeed , within the context of molecular motors walking on microtubuli @xcite , between 12 and 14 parallel lanes are relevant . also , the internal states might differ in the sense of different switching rates from one to another @xcite and the built - in asymmetry may result in different phases . in the context of intracellular transport it appears worthwhile to investigate the consequences of a coupling to a bulk reservoir , c.f . @xcite ; in particular , to study the interplay of domain wall formation induced by attachment and detachment processes as well as rare switching events . the authors are grateful for helpful discussions with felix von oppen , ulrich schollwck , paolo pierobon and mauro mobilia . financial support of the german excellence initiative via the program `` nanosystems initiative munich ( nim ) '' is gratefully acknowledged . in this appendix , we give details on the derivation of the analytic solution of the mean - field approximation in the continuum limit to first order in @xmath64 , i.e. the system of differential equations ( [ cont_a ] ) , ( [ cont_b ] ) . summing them we find @xmath198 ^ 2 + \partial_x[2\rho^\downarrow(x_i)-1]^2=0\ , , \label{start}\end{aligned}\ ] ] such that @xmath199 ^ 2 + [ 2\rho^\downarrow(x_i)-1]^2=j\ , , \ ] ] constitutes a first integral . remember that @xmath200+\rho^\downarrow(x_i)[1-\rho^\downarrow(x_i)]$ ] , such that @xmath114 is given by the total current : @xmath201 this equation suggests the following parameterization : @xmath202 the derivative reads @xmath203 which leads to the differential equation @xmath204 this may be solved by a separation of variables : @xmath205 to perform the integral , the substitution @xmath206 is useful . we obtain the inverse function @xmath207 : @xmath208 here we defined the function @xmath209 by @xmath210 and @xmath211 is an constant of integration . to obtain the inverse functions @xmath212 and @xmath213 , we have to express @xmath214 by @xmath215 resp . @xmath66 . recognize that @xmath214 can be positive or negative . we therefore define @xmath216 and analogously @xmath217 with @xmath7 and @xmath0 interchanged . now @xmath218 the inverse functions @xmath212 and @xmath213 thus read : @xmath219 the constants of integration @xmath211 and @xmath220 are determined by matching the boundary conditions . the inverse functions of eqs . ( [ x_rhoa ] ) , ( [ x_rhob ] ) constitute the solution to eqs . ( [ cont_a ] ) and ( [ cont_b ] ) , within the first - order approximation to the mean - field equations for the densities in the continuum limit . + next , we derive the result on @xmath143 given at the end of subsec . [ phases_bound ] . therefore , consider fig . we are interested in the point @xmath126 , and thus in the right branch of the spin - up density profile . as the spin - down density is in the ld phase , i.e. it is smaller than @xmath75 , we have @xmath221 in the above solution for @xmath215 . thus , @xmath222 in eq . ( [ x_rhoa ] ) . at the branching point of the analytic solution , i.e. the point @xmath126 , we have the density @xmath75 , such that there @xmath223 , implying @xmath224 . now , if this branching point lies on the right boundary , @xmath134 , as it does for the critical @xmath225 , this yields @xmath226 in eq . ( [ x_rhoa ] ) . on the other hand , the right branch satisfies the boundary condition on the right : @xmath227 . upon substitution into eq . ( [ x_rhoa ] ) , we obtain @xmath228 which for given @xmath229 is an equation for @xmath193 . in particular , @xmath193 is monotonically increasing in @xmath230 . investigating @xmath209 , it turns out that @xmath209 is in turn increasing in @xmath231 . since @xmath231 is bounded from above by @xmath232 , the maximal value for @xmath209 is provided by @xmath233 . next , we note that @xmath193 is an increasing function of @xmath234 . with the constraint that @xmath132 , which is necessary for @xmath134 , the largest @xmath234 arises for @xmath235 , i.e. @xmath236 . combining both results , the maximal value for the critical rates @xmath193 occurs for @xmath235 and @xmath232 . both conditions together yield @xmath237 finally , we note that @xmath235 and @xmath232 implies @xmath238 , such that @xmath143 arises if @xmath235 and @xmath238 . < 11 > natexlab#1#1bibnamefont # 1#1bibfnamefont # 1#1citenamefont # 1#1url # 1`#1`urlprefix[2]#2 [ 2][]#2

a `` friends - of - friends '' percolation algorithm has been used to extract a catalogue of density enhancements ( groups ) from the six slices of the las campanas redshift survey ( lcrs ) . the full catalogue contains 1495 groups and includes 35% of the lcrs galaxy sample . a clean sample of 394 groups has been derived by culling groups from the full sample which either are too close to a slice edge , have a crossing time greater than a hubble time , have a corrected velocity dispersion of zero , or contain a 55-arcsec `` orphan '' ( a galaxy with a mock redshift which was excluded from the original lcrs redshift catalogue due to its proximity to another galaxy i.e. , within 55 arcsec ) . median properties derived from the clean sample include : line - of - sight velocity dispersion km s , crossing time , harmonic radius mpc , pairwise separation mpc , virial mass , total group-band luminosity , and-band mass - to - light ratio ; the median number of observed members in a group is 3 .

astro-ph0006153

loose groups of galaxies are important but little - understood entities . they are intermediate in scale between galaxies and rich clusters , and thus their dynamics are important in the study of the distribution of dark matter on scales greater than haloes of galaxies but smaller than the typical sizes of large clusters [ see , for example , the review by oemler ( 1988 ) ] . their environment is also intermediate between that of isolated galaxies and that of the cores of rich clusters , and therefore the study of groups may provide clues to the processes that create the observed dependency of galaxy morphology on environment ( postman & geller 1984 ; oemler 1992 ; allington - smith et al . 1993 ; whitmore , gilmore , & jones 1993 ; zabludoff et al . 1996 ; hashimoto et al . 1998 ) . only in the past 15 years , however , with the advent of extensive galaxy redshift surveys , have suitably uncontaminated , objective group catalogues been available for study . galaxies and rich clusters of galaxies are generally easy to identify . they are high - contrast objects compared with their immediate surroundings . unfortunately , loose groups of galaxies , which are neither particularly dense nor exceptionally populous , are much more difficult to distinguish from their surroundings . early group catalogues were based upon the identification of galaxy concentrations on the sky , first primarily by visual inspection of photographic plates ( e.g. , holmberg 1969 , de vaucouleurs 1975 ) , and , later , via objective group - finding algorithms ( e.g. , turner & gott 1976 ) . since these group catalogues relied especially on just the two dimensions of spatial information available on the plane of the sky , they were greatly subject to contamination from projection effects . projection effects are largely mitigated ( although never fully eliminated ) in group catalogues derived from galaxy redshift surveys . in the early-1980 s , huchra & geller ( 1982 ; hg82 ) , pioneers in the extraction of groups from redshift surveys by means of objective , `` friends - of - friends '' percolation algorithms , compiled a group catalogue from a shallow ( @xmath10 ) whole - sky redshift catalogue containing 1312 galaxies . they later derived a group catalogue from the original ( @xmath11 ) cfa survey ( hereafter , cfa1 ) ( geller & huchra 1983 ; gh83 ) . the cfa1 has in fact proved to be a popular testing ground for group - finding algorithms ; additional group catalogues drawn from the cfa1 include those by nolthenius & white ( 1987 ; nw87 ) , nolthenius ( 1993 ; n93 ) , and moore , frenk , & white ( 1993 ; mfw93 ) . groups have also been identified in a @xmath12 slice from the cfa extension to @xmath13 [ henceforth , cfa2 ; ramella , geller , & huchra 1989 ( rgh89 ) ] and in the diameter - limited southern sky redshift survey [ ssrs ; maia , da costa , & latham 1989 ( mdcl89 ) ] . more recently , group catalogues have been extracted from the full northern cfa2 by ramella , pisani , & geller ( 1997 ; rpg97 ) , from the pisces - perseus redshift survey ( pps ) by trasarti - battistoni ( 1998 ; tb98 ) , and from the eso slice project ( esp ) galaxy redshift survey by ramella et al . ( 1999 ; rzz99 ) . in this paper , we will present a group catalogue based upon the las campanas redshift survey ( lcrs ; shectman et al . 1996 ) . due to the large volume this survey samples , the lcrs includes numerous `` great wall''-like structures within its borders and is therefore one of the first redshift surveys which can claim to enclose a reasonably fair sample of the nearby universe . with the exception of the esp , the redshift surveys from which the aforementioned group catalogues have been derived have all tended to be dominated by a very few large structures . therefore , a group catalogue based upon the lcrs should be found to contain groups in a wider range of environments than the groups identified from these shallower surveys . a census of group properties based upon lcrs groups would thus be more complete , and therefore more useful for studies of both galaxy dynamics and environmental dependences . in fact , this characteristic is so important that earlier variations of the present catalogue have already been used in studies of the environmental influence on galaxy morphology ( hashimoto & oemler 1998 ) , on the presence of `` e+a '' galaxies ( zabludoff et al . 1996 ) , and on the general rate of star formation within galaxies ( hashimoto et al . 1998 , allam et al.1999 ) . we divide the remainder of this paper as follows : we describe the lcrs galaxy sample in 2 , discuss the modified `` friends - of - friends '' algorithm used to extract the lcrs group catalogue in 3 , present the catalogue itself in 4 , and compare it with various group catalogues and with abell clusters [ abell 1958 ; abell , corwin , & olowin 1989 ( aco ) ] in 5 and 6 ; in 7 , we summarize and conclude . the lcrs is an optically selected galaxy redshift survey which extends to a redshift of 0.2 and which is composed of a total of 6 alternating @xmath14 slices , 3 each in the north and south galactic caps . completed in 1996 , the lcrs contains 26,418 galaxy redshifts , of which 23,697 lie within the official geometric and photometric limits of the survey . accurate @xmath7-band photometry and sky positions for program objects were extracted from ccd drift scans obtained on the las campanas swope 1-m telescope ; spectroscopy was performed at the las campanas du pont 2.5-m telescope , originally via a 50-fiber multi - object spectrograph ( mos ) , and later via a 112-fiber mos . for observing efficiency , all the fibers were used , but each mos field was observed only once . hence , the lcrs is a collection of 50-fiber fields ( with nominal apparent magnitude limits of @xmath15 ) and 112-fiber fields ( with nominal apparent magnitude limits of @xmath16 ) ; see figure 1 . thus , selection criteria vary from field to field , but these selection criteria are carefully documented and therefore easily taken into account . observing each field only once , however , creates an additional selection effect : the protective tubing of the individual fibers prevents the spectroscopic observation of both members of galaxy pairs within 55 arcsec of each other . hence , groups and clusters can be undersampled , potentially causing physical groups to be split by a `` friends - of - friends '' percolation algorithm and resulting in the mis - estimate of general group properties . we will return to this problem in the next section . in constructing the group catalogue , we have considered only those lcrs galaxies within the official geometric and photometric borders of the survey ; we have furthermore limited this sample to galaxies having redshifts in the range @xmath17 and luminosities in the range @xmath18 to avoid group - member incompleteness at the extremal distances of the sample , only groups within @xmath19 were admitted into the final group catalogue . [ n.b . : unless otherwise noted , all redshifts @xmath20 in this text are corrected for motion relative to the dipole moment of the cosmic microwave background ( cmb ; lineweaver et al . 1996 ) . the lcrs group catalogue was extracted by means of an adaptive `` friends - of - friends '' percolation algorithm based upon that of hg82 and modified for use with comoving distances and field - to - field sampling variations . we outline the procedure as follows : first , a seed galaxy ( `` galaxy @xmath21 '' ) is selected which has not yet been classified as either a group member or an isolated galaxy . every other non - classified galaxy @xmath22 in the survey sample is then tested to see if it lies within a projected separation @xmath23 and a velocity difference @xmath24 of the seed galaxy ( note that both @xmath23 and @xmath24 are functions of both the field @xmath25 and of the mean distance to galaxy pair @xmath26 ) : @xmath27 where @xmath28 and @xmath29 the distances @xmath30 are comoving , @xmath31\ ] ] ( @xmath32 and @xmath33 km s@xmath2 mpc@xmath2 ) . the variable @xmath34 is the angular separation between the two galaxies . if no companions are found within @xmath23 and @xmath35 of the seed galaxy , it is assigned `` isolated '' status and another seed galaxy is sought . if companions are found , they are added along with the seed galaxy to a list of group members forming a new group . in turn , the surroundings of each of these companions are combed for the next level of `` friends . '' this loop is repeated until no further companions are located , and the process is begun again by pursuing another seed galaxy . the group catalogue is complete only once every galaxy in the redshift sample has been classified as either `` isolated '' or `` grouped . '' only those groups containing three or more members are included in the final catalogue . the linking parameters , @xmath23 and @xmath24 , are specified in a manner which compensates for both the radial selection function and the field - to - field sampling variations characteristic of the lcrs . for each pair of galaxies , @xmath36 where @xmath37 and @xmath38 are @xmath23 and @xmath24 , respectively , for a given fiducial field at given fiducial redshift , and where @xmath39 is a linking scale which takes into account variations in galaxy sampling rate . it is defined by @xmath40^{-1/3 } , \ ] ] where @xmath41 is the number density of galaxies one would expect to observe at a comoving distance @xmath26 in field @xmath25 for a randomly homogeneous distribution of galaxies having the same selection function and sampling fraction as the lcrs redshift catalogue ; @xmath42 is @xmath41 for a given fiducial field at a given fiducial redshift ( fig . 2 ) . both @xmath42 and @xmath41 are computed by numerically integrating the field s selection function ( see appendix a ) . to elaborate , the fiducial field is an idealized field with a given set of characteristics . the fiducial field is itself never generated . it merely serves as the basis for the normalization of the linking scale @xmath39 in equation 9 . for simplicity , we have chosen our fiducial field to have 100% sampling , flux limits of @xmath43 , and a schechter ( 1976 ) luminosity function with the same parameter values as the lcrs 112-fiber sample : @xmath44 ( lin et al . since it is roughly the median redshift of the survey , we have chosen the fiducial redshift @xmath45 to be 30,000 km s@xmath2 . moving on , we note that field - to - field sampling variations are of particular concern when linking occurs across a field border . this concern is especially important in the case of a group situated on the border between a 50-fiber field and a 112-fiber field , where a factor of 2 discontinuity can occur in the expected surface density of galaxies on the sky . ( fortunately , only @xmath46 of the groups in the final catalogue 28 out of 1495 straddle a 50/112 border . ) therefore , when calculating the linking scale @xmath39 for two galaxies in two different fields , @xmath41 is taken to be its average from the two fields , @xmath47 but what of the artificial splitting of groups due to the lcrs s 55 arcsec fiber separation limit ? to avoid this problem , each of the @xmath48 galaxies originally excluded from lcrs redshift catalogue due to the fiber separation limit has been re - introduced into the sample by assigning it a redshift equal to that of its nearest neighbor convolved with a gaussian of width @xmath49 km s@xmath2 ( roughly the median line - of - sight velocity dispersion of a cleaned lcrs group sample which excludes the 55-arcsec `` orphans '' ) . the re - included galaxies subscribe to all the same photometric limits and spatial borders that are imposed upon the original galaxy sample ( 2 ) . in closing , we note that , for a given choice of @xmath37 and @xmath38 , this algorithm leads to a unique group catalogue independent of the choice of the original seed galaxy . due to the aforementioned field - to - field variations in sampling , however , it is more intuitive to characterize a group catalogue extracted from the lcrs not with @xmath37 for a certain fiducial field , but with @xmath50 the corresponding number density enhancement of the surface contour which delimits each group ; the value @xmath51 characterizes the groups more generally , as it is valid no matter the field in which a given group resides . so far , it has been shown how the superstructure of the group - finding algorithm has been set into place . the choice of values for @xmath52 ( @xmath37 ) and @xmath38 , however , has yet to be presented and explained . we shall follow a course very similar to that of hg82 in our justification of the two values ultimately adopted . take figure 3 as a guide to the choice of @xmath51 ( @xmath37 ) and @xmath38 . their selection should satisfy a few basic criteria . first , the density enhancement contour sought should be high enough to limit the number of interloper galaxies contained within a group , but not so high that only the cores of rich clusters are found . on the other hand , for a thin - wedge geometry like that of an lcrs slice , the groups selected should not be too loose ; otherwise , edge effects become excessive . furthermore , the value of @xmath38 , as with the value of @xmath51 ( @xmath37 ) , should minimize the number of interlopers , but without biasing group line - of - sight velocity dispersions toward artificially low values . as a definite upper limit , @xmath38 should not exceed the radius ( in km s@xmath2 ) of a typical void observed at @xmath45 . the range of acceptable values for @xmath51 ( @xmath37 ) and @xmath38 are enclosed by the solid border in figure 3 . by means of the following semi - quantitative arguments , it will be concluded that the most reasonable values for the two linking parameters are @xmath53 which are denoted in figure 3 by an asterisk . first , consider the number of interlopers per galaxy , @xmath54 , within a specified @xmath23 and @xmath24 of some galaxy . for a given galaxy , @xmath54 can be roughly estimated by the equation ( hg82 ) @xmath55 \sigma_{\rm gal } , \ ] ] where @xmath56 is the surface number density of galaxies in the redshift catalogue , @xmath57 is the total number of galaxies in the sample , and @xmath58 is the number of galaxies within @xmath24 of the galaxy s velocity @xmath59 . ( this measure is actually an underestimate of @xmath54 , since it neglects the correlation of galaxy positions on the sky , but , if we stringently limit the number of interlopers per galaxy , it is adequate for our purposes . ) the lower curve in figure 3 denotes the locus @xmath60 for the the fiducial field at the fiducial redshift @xmath61 30,000 km s@xmath2 . below and to the right of this curve , where @xmath62 , the number of interlopers per galaxy is considered excessive , and thus this curve constitutes one of our boundaries . figure 4 presents @xmath54 for the redshift of each galaxy in the lcrs sample for the values of @xmath51 ( @xmath37 ) and @xmath38 listed in equation 13 . the bumps and dips in figure 4 are due to inhomogeneities within the distribution of galaxy velocities ( the wall - like structures ) in the lcrs ( shectman et al . note that the median number of interlopers per galaxy @xmath63 for our eventual choice of @xmath51 ( @xmath37 ) and @xmath38 . there is only a slight large - scale trend evident in figure 4 , the fact of which argues that the number of spurious groups in the final catalogue should not be a strong function of redshift . to preclude only finding the dense central regions of rich clusters , an arbitrary upper limit to the density enhancement contour cut is set at @xmath64 , equivalent to assigning a value of 0.528 @xmath65 mpc to @xmath37 . in the quest for a catalogue of loose groups , a lower density contrast cutoff is preferred . upon testing , it was discovered that , below a contrast of @xmath66 , edge effects become a problem . at these cutoffs , over half of the groups must be excluded from the clean sample used in the study of group properties ; the group radii encroach upon the slice s borders . thus , the density contrast @xmath66 was chosen for the group catalogue , since it provides a reasonable compromise between the pursuit of loose groups and the desire for a large clean sample . next , the value of @xmath38 should be chosen such that group velocity dispersions are not overconstrained . to avoid seriously underestimating group velocity dispersions , we must set @xmath38 to a value which accommodates the maximum _ likely _ physical velocity dispersion , @xmath67 . mathematically , the maximum _ possible _ velocity dispersion of a group , @xmath68 , is that obtained when the group is maximally spread out in redshift i.e. , when each group member is just within the linking velocity @xmath24 of its nearest neighbor : @xmath69^{\case{1}{2 } } .\ ] ] for example , the maximum velocity dispersion possible for a group containing @xmath70 galaxies would be @xmath71 at the very least , we want this theoretical maximum _ possible _ velocity dispersion to encompass i.e. , to be greater than the maximum _ likely _ physical velocity dispersion . due to the lcrs galaxy selection function , a group containing @xmath70 galaxies in a fiducial field at the fiducial velocity ( @xmath7230,000 km s@xmath2 ) will typically be of abell richness class @xmath73 ( abell 1958 ) . if we make the conservative assumption that the maximum likely @xmath74 for a physical @xmath75 `` group '' is no more than about @xmath76 km s@xmath2 ( zabludoff et al.1993 ; rpg97 ) , we can set a lower limit for @xmath38 by means of the relation @xmath77 which yields @xmath78 400 km s@xmath2 . this value provides the leftmost boundary to the region of acceptable @xmath79 in figure 3 . at the other extreme , the velocity linking parameter should not be so large that galaxies are linked across a void diameter . thus , we give @xmath38 a maximum limit equal to the typical radius of observed voids near @xmath80 30,000 km s@xmath2 , or @xmath81 2000 km s@xmath2 . this maximum @xmath38 provides the rightmost border in figure 3 . to fine - tune @xmath38 , the group algorithm was run , using @xmath66 , for four different values : @xmath82 500 km s@xmath2 , @xmath82 1000 km s@xmath2 , @xmath82 1500 km s@xmath2 , and @xmath82 2000 km s@xmath2 ( fig . many of the group velocity dispersions are unaffected by the change in @xmath38 , but a significant number of groups , by adding progressively more outlying galaxies , see a dramatic increase in @xmath74 as @xmath38 is increased from 500 km s@xmath2 to 2000 km s@xmath2 . in the end , @xmath82 500 km s@xmath2 was chosen , due to its lower probability of incorporating spurious groups into the catalogue , and due to its ability to generate a catalogue with a relatively low redshift - dependence in @xmath74 . this latter property especially makes for a more homogeneous group sample . the full catalogue contains 1495 groups and includes 35% of the lcrs galaxy sample ( figs . 6 , 7 , & 8) . the complete list of lcrs groups and their individual properties is compiled in table 1 , which , due to its size , is confined to the electronic version of this text . members of the clean sample of 394 groups those which have barycenters more than two pairwise separations [ column ( 12 ) ] from a slice edge , crossing times [ column ( 16 ) ] less than a hubble time , corrected line - of - sight velocity dispersions [ column ( 9 ) ] greater than zero , and no 55-arcsec `` orphan '' galaxies as group members are marked by an asterisk in column ( 1 ) . group properties are calculated according to the prescriptions of rgh89 , but modified for use with comoving distances and field - to - field sampling variations . before we enter a description of the tabulated group properties , we must note two caveats . first , even the clean sample is unlikely to be perfectly clean . as noted , one of the rejection criteria used is the removal of groups which are closer than two pairwise separations from a slice edge . in doing so , we have followed the lead of rgh89 , who used this same rejection criterion to define their clean group sample . clearly , having no `` edge - proximity '' rejection criterion would include many groups which overflow the survey boundaries . the measured properties of these groups would be biased from their true values due to their truncated membership . since the pairwise separation , @xmath83 , is a measure of the group radius , excluding groups which are closer than one @xmath83 to a slice edge should go far in counteracting this effect . one must be careful , though , since a truncated membership will both offset the position of the measured group barycenter away from the slice edge and bias the measure of @xmath83 itself toward small values . thus , a group which in fact extends over a slice border could still be accidentally included in the clean sample . therefore , choosing an edge - proximity rejection criterion of @xmath84 although still not perfect is much safer . one could even make more stringent demands , requiring groups in the clean sample to be at least three or four times their measured @xmath83 from a slice edge . here , however , one must worry about biasing the clean sample unnecessarily towards only the most compact systems . in table 2 , we see the effects of steadily increasing the threshold of edge - proximity rejection on the resulting clean samples : although the sample line - of - sight velocity dispersion , @xmath74 , is not strongly affected , the sample pairwise separation and harmonic radius [ @xmath85 ; see column ( 14 ) of table 1 ] both drop precipitously for these larger values of the rejection criterion . therefore , as a compromise , we have chosen a value of @xmath84 . our second caveat is that the clean sample is not necessarily the best sample to use for all purposes . since the clean sample excludes groups containing 55-arcsec `` orphans '' , on average this sample discriminates against groups with dense cores . on some occasions ( as in 6 ) , a superset of the clean sample one which does not exclude groups merely because they harbor an `` orphan '' is the preferred catalogue . the clean sample , as defined , however , is the most conservative catalogue the sample with the fewest extrinsic assumptions attached to it . therefore , in matters of discussing intrinsic mean group properties , we shall use this clean group sample . with these caveats in mind , the columns in table 1 are as follows : column ( 1 ) : a running group identification number , @xmath86 , for a given lcrs slice . the slice declination and @xmath86 form the basis of the iau - registered naming convention for lcrs loose groups , which is of the form @xmath87 where lclg stands for `` las campanas loose group , '' @xmath88 is the ( zero - padded ) declination for the lcrs slice wherein the group resides , and @xmath89 is the ( zero - padded ) @xmath90 . for more information , see the online dictionary of nomenclature of celestial objects ( http://cdsweb.u-strasbg.fr/cgi-bin/dic ) . an asterisk appended to @xmath90 in table 1 indicates that the group is a member of the clean sample . columns ( 2 4 ) : a weighted measure of the b1950.0 right ascension ( in hh mm ss.ss format ) of the group s barycenter , @xmath91 where @xmath92^{-1}\ ] ] this weighting factor which is proportional to the inverse of the selection function helps to counteract a bias resulting from a group straddling two fields with different galaxy sampling characteristics ; a discussion on how @xmath93 is estimated can be found in appendix a. @xmath94 , the number of observed group members , is listed in column ( 11 ) . column ( 5 - 7 ) : a weighted measure of the b1950.0 declination ( in sdd mm ss.s format ) of the group s barycenter , @xmath95 where @xmath96 is as defined in equation 19 . @xmath94 is listed in column ( 11 ) . column ( 8) : the group s redshift , @xmath97 , with respect to the local comoving frame . it is taken as the ( unweighted ) mean of the members redshifts . ( see figs . 9 & 10 . ) column ( 9 ) : the group line - of - sight velocity dispersion , in km s@xmath2 , corrected for relativistic effects ( harrision 1974 ) , @xmath98 ( fig . the random errors in the lcrs galaxy redshift measurments are also removed , in quadrature , assuming @xmath99 km s@xmath2 ( shectman et al . 1996 ) : @xmath100 @xmath94 is listed in column ( 11 ) . column ( 10 ) : a formal estimate of the standard error in @xmath74 , in km s@xmath2 . if a normal distribution is assumed for the line - of - sight velocities , the standard error for @xmath101 can be expressed as @xmath102 ( deming 1950 ) . by means of propagation of errors ( bevington 1969 ) , this expression yields @xmath103 which is employed for the estimates listed in column ( 10 ) . @xmath94 is listed in column ( 11 ) . column ( 11 ) : @xmath94 , the observed number of lcrs galaxies ( including any 55-arcsec `` orphans '' ) comprising the group ( fig . included in @xmath94 are only those lcrs galaxies which the lie within the official geometric and photometric borders of the survey and which subscribe to the redshift and absolute magntitude limits set forth in equations 1 and 2 . as is typical for `` friends - of - friends '' group catalogues , the distribution of @xmath94 for the lcrs group catalogue is heavily skewed toward small values : the median @xmath94 is 3 . column ( 12 ) : the mean pairwise separation , @xmath104 , \ ] ] where @xmath105 and where @xmath106 is the comoving distance to the group , @xmath107 is the angular separation between group members @xmath21 and @xmath22 , and @xmath96 and @xmath108 are the respective weights for @xmath21 and @xmath22 ( equation 19 ) . @xmath83 has dimensions @xmath65 mpc . see figure 13 . column ( 13 ) : an estimate of the rms error in @xmath83 , @xmath109 where @xmath110 where @xmath111 is the number of distinct galaxy pairs in the group , and where @xmath106 , @xmath107 , @xmath96 , and @xmath108 are as described for @xmath83 ( equations 25 and 26 ) . equation 27 was derived from equation 25 , assuming that @xmath112@xmath107@xmath113@xmath107@xmath114 for the typical group angular sizes encountered in this group catalogue , and that the contribution to @xmath115 by the rms error in @xmath106 is insignificant compared to the contribution by @xmath116@xmath107@xmath114 . equation 28 was derived from equation 26 via a straightforward ( albeit tedious ) application of propagation of errors . the units for @xmath115 are @xmath65 mpc . column ( 14 ) : the harmonic radius , @xmath117 , \ ] ] where @xmath118 @xmath106 , @xmath107 , and @xmath96 and @xmath108 are as described for @xmath83 ( equations 25 and 26 ) . @xmath85 has dimensions @xmath65 mpc . see figure 14 . column ( 15 ) : an estimate of the rms error in @xmath85 , @xmath119 where @xmath120 @xmath111 is the number of distinct galaxy pairs in the group ; @xmath106 , @xmath107 , @xmath96 , and @xmath108 are as described for @xmath83 ( equations 25 and 26 ) . the derivation of equations 31 and 32 closely mimics that of equations 27 and 28 . the units for @xmath121 are @xmath65 mpc . column ( 16 ) : the crossing time for the group , @xmath122 in units of the hubble time ( @xmath123 ) ; see figure 15 . this measure is heavily influenced by the relative values of the linking parameters @xmath124 ( which determines @xmath85 ) and @xmath125 ( which determines @xmath74 ) . following gott & turner ( 1977 ) , it can be estimated that the time for a uniform sphere to undergo complete virialization is @xmath126 hence , all groups with @xmath127 should have had enough time to virialize completely within the age of the universe [ but see diaferio et al . ( 1993 ) ] . thus , it can be deduced that roughly half of the lcrs clean sample could have undergone complete virialization . column ( 17 ) : an estimate of the rms error in @xmath128 , @xmath129 in units of the hubble time ( @xmath123 ) ; @xmath74 , @xmath130 , @xmath85 , and @xmath121 come from columns ( 9 ) , ( 10 ) , ( 14 ) , and ( 15 ) , respectively . equation 35 was derived by standard propagation of errors analysis of equation 33 . column ( 18 ) : the group s virial mass , @xmath131 where @xmath132 is the gravitational constant ( fig . this estimate assumes that the groups are virialized and that the galaxies trace the mass distribution within the group [ see , for example , binney & tremaine ( 1987 ) , chapter 10 , section 2.3 ] . @xmath133 is in units of @xmath134 . column ( 19 ) : an estimate of the rms error in @xmath133 , @xmath135 where @xmath74 , @xmath130 , @xmath85 , and @xmath121 come from columns ( 9 ) , ( 10 ) , ( 14 ) , and ( 15 ) , respectively . equation 37 was derived by propagation of errors analysis of equation 36 . note that the large random errors inherent to @xmath74 and to @xmath85 propagate into estimates for @xmath133 . the units for @xmath136 are @xmath134 . column ( 20 ) : the total group luminosity in the lcrs @xmath7-band , @xmath137 , corrected for selection effects to account for galaxies not observed by the lcrs ( fig . 17 ) ; @xmath137 is in units of solar luminosity ( @xmath138 ) , in which the @xmath7-band absolute magnitude of the sun is taken to be @xmath139 ( pinsonnealt 1992 ) . the mathematical apparatus behind the correction factor can be found in appendix b. column ( 21 ) : an estimate of the rms error in @xmath137 , @xmath140 , obtained by summing the rms errors of the individual components of @xmath137 in quadrature . details can be found in appendix b. the units for @xmath140 are @xmath138 ( lcrs @xmath7-band ) . column ( 22 ) : the ratio , @xmath141 , by which the sum of the luminosities of the observed @xmath94 galaxies must be multiplied in order to obtain an estimate of the group s total ( lcrs @xmath7-band ) luminosity , @xmath137 , @xmath142 over the clean sample , the median @xmath141 is @xmath143 for the 50-fiber groups and @xmath144 for the 112-fiber groups . column ( 23 ) : the group mass - to - light ratio in the lcrs @xmath7-band , @xmath145 , in units of @xmath146 ( fig . 18 ) . for comparison with the mass - to - light ratios for groups from other redshift catalogues ( in particular , those based upon the de vaucouleurs @xmath147-band ) , it is convenient to note that @xmath148 the uncertainties in @xmath133 and in @xmath137 tend to give large errors for the mass - to - light ratios of individual groups . column ( 24 ) : an estimate of the rms error in @xmath145 , @xmath149 where @xmath133 , @xmath136 , @xmath137 , and @xmath140 come from columns ( 18 ) , ( 19 ) , ( 20 ) , and ( 21 ) , respectively . equation 40 was obtained by means of propagation of errors analysis of equation 39 . the units for @xmath150 are @xmath146 . column ( 25 ) : an estimate of the group s abell counts , @xmath151 ( fig . abell ( 1958 ) defined his counts to be the number of galaxies , corrected for background contamination , in the magnitude interval @xmath152 to @xmath153 , that lie within a 1.5 @xmath65 mpc projected separation of a cluster s center ; the magnitude @xmath152 is the magnitude of the third brightest cluster member . due to the sampling characteristics and the relatively small range of apparent magnitudes within the lcrs , we could not use abell s definition directly . we had instead to derive each group s abell counts via a schechter function ; the details of the method can be found in appendix c. when compared with actual abell clusters within the lcrs volume ( 6 ) , we find the following relation between the lcrs group counts estimate , @xmath151 , and the revised abell cluster values given by aco : @xmath154 column ( 26 ) : an estimate of the error in the abell counts @xmath151 , based upon poisson statistics of the observed number of lcrs member galaxies within a projected distance of 1.5 @xmath65 mpc of the group center ( @xmath155 ) , @xmath156 column ( 27 ) : the type of group one within the borders of a 50-fiber field ( type 1 ) , one within the borders of a 112-fiber field ( type 2 ) , or one straddling the border of a 50- and a 112-fiber field ( type 3 ) . column ( 28 ) : a column to reference any applicable notes for the given group . these table notes are as follows : a : : : the group s crossing time @xmath128 is greater than a hubble time ( @xmath123 ) . b : : : the group s barycenter is closer than @xmath157 to a slice edge . c : : : the group contains at least one 55 arcsec `` orphan '' with a mock redshift . d : : : the group s line - of - sight velocity dispersion , corrected for galaxy velocity errors , is less than or equal to 0 km s@xmath2 . table 3 lists the members of each group . due to its great size ( 10761 lines ) , it is provided only in electronic format . each group is introduced by a header composed of its group number designation @xmath90 , its redshift @xmath97 , and its line - of - sight velocity dispersion @xmath74 [ columns ( 1 ) , ( 8) , and ( 9 ) of table 1 , respectively ] . the columns for table 2 are as follows : column ( 1 ) : a group member identification number . columns ( 2 - 4 ) : the group member s right ascension in 1950.0 coordinates . columns ( 5 - 7 ) : the group member s declination in 1950.0 coordinates . column ( 8) : the group member s lcrs @xmath7-band isophotal magnitude . column ( 9 ) : the group member s lcrs @xmath7-band central magnitude . column ( 10 ) : the group member s heliocentric velocity , @xmath158 , in km s@xmath2 . column ( 11 ) : the group member s spectrum type ( e @xmath159 contains strong emission lines ; c = a continuum with absorption lines ; b @xmath159 contains both absorption lines and moderate emission lines ; m @xmath159 mock velocity i.e. , the galaxy was one of the 55-arcsec `` orphans '' excluded from the spectroscopic survey due to its proximity to another galaxy ) . the individual group member properties are derived directly from a working copy of the published lcrs galaxy catalogue ; further details can be found in shectman et al . table 4 contains information regarding the general properties of groups both in the lcrs catalogue and from other catalogues . the values tabulated for the eight non - lcrs group catalogues have been taken from the original papers , and , where necessary , these values were converted to be consistent with the property definitions detailed in 4 . the median properties for the lcrs catalogue were derived from the clean sample of 394 groups ; the fraction of grouped galaxies , from the full sample of 1495 groups . we have broken the lcrs groups catalogue into three sub - catalogues for the purpose of inter - comparison ; the sub - catalogues include the lcrs groups from 50-fiber fields , the lcrs groups from 112-fiber fields , and the lcrs groups which straddle a 50/112 border . furthermore , we can compare the present lcrs group catalogue with a precursor based upon the lcrs @xmath160 slice ( tucker 1994 ; henceforth , t94 ) . the properties of these lcrs group samples are tabulated under the lcrs main heading in table 4 . note that there is substantial variation in group properties among the three sub - catalogues . these variations are also apparent in figures 11 19 . first , we can discount the oddities apparent in the 50/112 group properties due to poor number statistics ( there are only 5 of these groups in the clean sample ) and due to the difficulties of linking group members across a 50/112 border . in any case , if not for its aberrantly large @xmath74 s which in turn affect estimates of @xmath133 and @xmath145 the 50/112 sample properties would closely match those of the 50-fiber sample . unfortunately , the discrepancies between the 50- and the 112-fiber samples group properties are somewhat harder to dismiss . great effort was put forth into accounting for field - to - field sampling variations in the group - finding algorithm , including using appropriate luminosity functions and surface brightness cutoff functions for each of the northern 50-fiber , the southern 50-fiber , the northern 112-fiber , and the southern 112-fiber field types ( lin et al . apparently , some small residual selection bias between the 50- and 112-fields remains . fortunately , we can still follow one of two options : ( 1 ) we can choose to ignore the 50-fiber groups altogether ( they make up only @xmath161 of the total number of groups ) , or ( 2 ) we can note that both the 50-fiber and the 112-fiber group properties fall within the general range of typical group properties from other surveys ( 5.2 ) and thus consider the combined lcrs group catalogue as being representative of groups as a whole . unless otherwise noted , we will take the latter course in the following sections . finally , consider the properties from an earlier version of the lcrs group catalogue ( t94 ) . this group catalogue is composed of only 50-fiber data from the lcrs @xmath160 slice , so it is not surprising that many of its general properties have values which approximate those of the present 50-fiber sample . one major difference is the percentage of galaxies in groups . the relatively small value for the t94 sample may be attributed at least in part to the fact that the earlier lcrs group - finding algorithm ignored 55-arcsec `` orphans , '' thus disrupting many groups into doublets or isolated galaxies which were excluded from the resulting t94 catalogue . it is instructive to compare the lcrs group catalogue with those derived from other galaxy redshift surveys . in this section we shall look at nine other group catalogues extracted from five different surveys and relate their properties to those of lcrs groups . the nine group catalogues considered are those by gh83 , nw87 , n93 , mfw93 , mdcl89 , rgh89 , rpg97 , tb98 , and rzz99 . the five surveys are the cfa1 ( huchra et al . 1983 ) , the ssrs ( da costa et al . 1988,1991 ) , the cfa2 ( de lapparent , geller , & huchra 1988 ; huchra et al . 1990 ; huchra , geller , & corwin 1995 ) , the pps ( giovannelli & haynes 1993 ; wegner , haynes , & giovanelli 1993 ) , and the esp ( vettolani et al . 1997 ) . table 5 summarizes the general properties of the different survey samples used to generate these group catalogues ; for comparison , the characteristics of the lcrs sample used in this paper are also included ( note that the lcrs sample contains substantially more galaxies and encloses the largest volume of any of the survey samples listed ) . the interested reader is urged to consult the original papers for details . now , as one might suspect from the small - but - statistically - significant differences between subsamples of the lcrs group catalogue ( 5.1 ; figs . 11 19 ) , it is unlikely that another group catalogue extracted from a different survey with different galaxy selection criteria using another variation of a `` friends - of - friends '' group - finding algorithm would be from the same statistical parent population of groups as the lcrs group catalogue . to verify this , we have used the kolmogorov - smirnov ( ks ) test to compare the physical properties of the lcrs groups with those of the five group catalogues available in machine - readable form ( gh83 , mdcl89 , rpg97 , tb98 , and rzz99 ) . the results of our analysis are shown in figures 21 27 and summarized in table 6 . notice that most of the formal ks probabilities are so low that we needed to list them in logarithmic form in order to make table 6 readable . thus , these group catalogues are not extracted from the same parent population as the lcrs groups : their properties , although roughly similar ( table 4 ) , do differ significantly . there is a hidden benefit in these very low ks probabilities . the other four group catalogues are not available in machine - readable form ( and two of those were not even published in paper form ) . since these four have roughly the same number statistics and variations in physical properties as the other five , we can with some safety assume that the physical properties of the groups from these four catalogues also differ significantly from those of the lcrs group sample . we can therefore exploit the one statistic that is published for the properties of all nine of these group catalogues the median as a reasonable measure of comparison . now , in comparison with these other group catalogues , it is clear from table 4 that the median properties of the full lcrs group catalogue ( those under the `` all '' heading ) are fairly typical . this is comforting , since it shows that all these group catalogues are looking at roughly the same sort of systems . but can more be said ? due to differing survey characteristics and differing group - extracting parameters , it is notoriously difficult to make cross - catalogue comparisons . nonetheless , rough comparisons can be made . of the properties that table 4 lists , two are primarily @xmath162-dependent ( @xmath116@xmath85@xmath163 , @xmath116@xmath83@xmath163 ) , one is primarily @xmath24-dependant ( @xmath116@xmath164@xmath163 ) , and several are dependent in more - or - less complex ways on both @xmath23 and @xmath24 ( # of groups in catalogue , # of groups in clean sample , % of galaxies in groups , @xmath116@xmath165@xmath166 , @xmath116@xmath133@xmath163 , @xmath116@xmath137@xmath163 , @xmath116@xmath145@xmath163 ) . let us designate the properties which depend primarily only on @xmath23 _ or _ @xmath24 as _ simply derived _ properties , and those that depend more complexly on both @xmath23 _ and _ @xmath24 as _ complexly derived _ properties . of the simply derived properties , those depending mainly on @xmath23 can be considered density defining quantities ; the one that mainly depends on @xmath24 ( @xmath116@xmath164@xmath163 ) defines the mean gravitational energy content of the systems . many of the complexly derived properties are straightforward functions of combinations of the simply derived properties ( see 4 ) . let us first consider the simply derived properties . what can we say about the relative mean densities of the groups from the different catalogues ? consider @xmath116@xmath85@xmath163 . for the catalogues which on average contain the denser groups , @xmath116@xmath85@xmath163 should be relatively small . under this criterion , we find that the tb98 and rpg97 catalogues contain the densest systems on average ; the lcrs , gh83 , mdcl89 , rgh89 , and n93 catalogues , those of average density ; and the nw87 catalogue , the least dense . @xmath116@xmath83@xmath163 is another indicator of relative group density , and , albeit with some changes in rank , roughly the same trend is seen . thus , we can conclude that the tb98 and rpg97 groups are the densest on average ; the lcrs , gh83 , mdcl89 , rgh89 , and n93 groups are average systems ; and the nw87 groups are the least dense . next , what can we infer about the mean gravitational energy content of the various catalogues i.e , which are the `` hottest '' systems , the systems with the highest @xmath116@xmath164@xmath163 ? from table 4 , we see that the gh83 and rgh89 catalogues have high @xmath116@xmath164@xmath163 , the lcrs , mdcl89 , rpg97 , tb98 , and rzz99 catalogues have intermediate values , and the nw87 , n93 , and mfw93 catalogues are the `` coolest '' of the systems listed ( this is a reflection in part on the functional forms of the linking lengths used in nw87 , n93 , and mfw93 , which strongly limit velocity outliers ) . the values for the complexly derived quantities are by their very natures less certain . we will consider three specifically @xmath116@xmath133@xmath163 , @xmath116@xmath165@xmath163 , and @xmath116@xmath145@xmath163 . consider @xmath116@xmath133@xmath163 : the gh83 and the rgh89 systems are the most massive , followed in order of decreasing mass by the lcrs , the rpg97 , the mdcl89 , the nw87 , and the tb98 systems . it may seem strange that the catalogue containing the densest systems ( tb98 ) is also the catalogue with the least massive systems . in fact , this is a consequence of the relatively small linking lengths @xmath23 that tb98 used to extract his groups from the pps : his groups do not extend out to the same radii as those in the `` looser '' catalogues , so they enclose less mass . on the other hand , his groups are the most virialized systems in table 4 ( based upon their short crossing times ) , so his catalogue estimate for @xmath116@xmath133@xmath114 is probably the most accurate . ( the lcrs groups catalogue have a somewhat longer median crossing time , but one can still expect a large fraction of them to be virialized see equation 34 . ) we can also ask which group catalogues give evidence for the most ( or least ) dark matter . the lcrs group catalogue has a @xmath116@xmath145@xmath166 of @xmath167 in the lcrs @xmath7-band ; if we convert this to the de vaucouleurs @xmath147-band ( equation 39 ) used by the other group catalogues with measured mass - to - light ratios , we find @xmath116@xmath145@xmath168 . this is fairly average . of the other group catalogues with measured values of @xmath116@xmath145@xmath163 , the n93 and rgh89 catalogues show lower mass - to - light ratios , and the rpg97 and nw87 show higher . in conclusion , we can state the lcrs group catalogue is quite average in its simply derived properties its groups are of average density and gravitional energy content . furthermore , it is not particularly distinct in its complexly derived properties the groups being moderately massive with relatively long crossing times and average mass - to - light ratios . it is , however , a very large catalogue of groups in a wide variety of environments . the lcrs values for the properties tabulated here are among the least - biased presently available . the original abell catalogue ( abell 1958 ) is comprised of 2712 rich , highly dense clusters in the northern sky . revised and expanded to include southern clusters , an updated abell catalogue ( aco ) now contains 4073 rich clusters over a significant fraction of the whole sky . each cluster contains within a projected radius of @xmath169 ( @xmath170 mpc ) from its center at least 30 member galaxies in the magnitude range @xmath152 to @xmath171 , where @xmath152 is the apparent magnitude of the third brightest cluster member . the abell / aco catalogue nominally encompasses a redshift range of @xmath172 ; the redshifts are estimated empirically based upon the apparent magnitude of the tenth brightest member , @xmath173 , and tend to be accurate to within a factor of 2 . each cluster is classified according to distance class @xmath174 , estimated from @xmath173 , and according to richness class @xmath7 , based on the number of members meeting the above criteria for projected radial distance and apparent brightness . furthermore , a supplementary catalogue of southern clusters too poor or too distant to be included in the main catalogue contains an additional 1174 clusters ( abell s001 s1174 ) . how many abell clusters do we expect to identify within the lcrs group catalogue ? there are 206 abell clusters ( including those from the supplementary catalogue ) within the sky - projected confines of the six lcrs slices . if we confine ourselves to a superset of the clean sample one which includes groups with 55 arcsec `` orphans '' meeting the other criteria for inclusion into the clean sample there are 735 lcrs loose groups which could be matched . the abell catalogue is purported to be complete to a redshift of 0.20 , but the lcrs group catalogue only goes to a redshift of 0.15 ; so we expect @xmath175 of abell clusters to be lost due to distance . furthermore , if we assume a typical pairwise separation of @xmath176 mpc , the effective volume of the lcrs slices for the superset sample is again reduced by about half . therefore , we expect about @xmath177 matches between abell clusters and lcrs groups . we find 54 matches , which are described in table 7 . the columns are as follows : column ( 1 ) : the lcrs group s running identification number for the given slice ( @xmath90 ) . column ( 2 ) : the lcrs group s @xmath178 in equinox 1950.0 coordinates ( @xmath179 ) . column ( 3 ) : the lcrs group s @xmath180 in equinox 1950.0 coordinates ( @xmath181 ) . column ( 4 ) : the lcrs group s redshift ( @xmath182 ) . column ( 5 ) : the observed number of galaxies from the lcrs official spectroscopic sample that lie within the group ( @xmath183 ) . column ( 6 ) : an estimate of the abell richness for that group ( @xmath151 ) . column ( 7 ) : the name of the abell cluster match to this group . column ( 8) : the abell cluster s @xmath178 in equinox 1950.0 coordinates ( @xmath184 ) . column ( 9 ) : the abell cluster s @xmath180 in equinox 1950.0 coordinates ( @xmath185 ) . column ( 10 ) : the abell cluster s redshift ( @xmath186 ) , if known . column ( 11 ) : the abell cluster s richness class ( @xmath7 ) . column ( 12 ) : the abell cluster s distance class ( @xmath174 ) . column ( 13 ) : the abell cluster s abell counts , as measured by aco ( @xmath187 ) . column ( 14 ) : the angular separation in arcminutes between the measured group center and the abell cluster center as reported by aco . a group - cluster pair was considered a `` match '' if this angular separation was less than 12 arcmin ( @xmath188 mpc at the median redshift of the lcrs ) . of these matches , we see that some are quite good ( @xmath189 , separation @xmath190 6 arcmin , convergent cluster distance / group redshift , distance class @xmath116 6 ) , such as the match between lclg-42 010 and abell 2758 , and that others are not so good ( separation @xmath191 6 arcmin , widely divergent cluster distance / group redshift ) . of the latter , some of the lcrs groups are likely in the foreground ( lclg-39 172 , lclg-39 202 ) or in the background ( lclg-03 126 , lclg-39 256 , lclg-42 052 , lclg-42 234 ) . furthermore , some abell clusters have been split into two or more groups by the `` friends - of - friends '' algorithm ( abell 1200 , s418 , 2969 , s281 , s253 , s286 ) . all in all , however , the matchups are not too bad , especially for those groups with @xmath192 observed members and those clusters of abell distance class @xmath193 . if we consider just the non - split @xmath194 matches , we find by least - squares fit that @xmath195 ( eq . [ 41 ] ; see fig . such a poor correspondence between @xmath196 and @xmath187 may be disheartening at first at least until one realizes that independent measures of @xmath187 among abell , corwin , and olowin _ themselves _ often had random and systematic offsets of up to 50 galaxy counts ( see aco , figs . 6 & 7 ) . furthermore , aco stress that the measured counts for individual clusters are nearly meaningless . therefore , we also stress that @xmath151 ( or the corresponding @xmath187 from eq . [ 31 ] ) for individual lcrs groups will likely be very noisy ; it is better to consider mean or median values of these estimates for sets of lcrs groups . for instance , the median @xmath151 for lcrs groups is 10.8 ( fig . this value for @xmath151 indicates a median @xmath197 , which implies that lcrs groups , on average , can be thought of as very poor clusters . ( the negative values for the median counts is based in part to the different means of background subtraction used here and in aco . ) clearly , the lcrs groups do define an environment intermediate between that of isolated galaxies and that of rich clusters . we have presented in this paper a catalogue of loose groups within the lcrs . these groups were extracted from the lcrs galaxy catalogue by means of a standard huchra - geller `` friends - of - friends '' percolation algorithm , modified for comoving distances and for the field - to - field sampling variations characteristic of this redshift survey . internal comparisons of characteristics within the lcrs group catalogue indicated some minor differences between groups extracted from the 50-fiber fields and those extracted from the 112-fiber fields . we attributed these differences to some small but still - hidden residual selection bias between galaxies in the 50-fiber and in the 112-fiber fields . since groups in the 50-fiber sample comprise only @xmath161 of the lcrs group catalogue , and since the properties of both the 50-fiber and the 112-fiber groups fall within the general regime of other group catalogues , we found this discrepancy to be of only minor importance . external comparison of the lcrs group catalogue with nine other group catalogues , all based upon other redshift surveys , showed that the general properties of lcrs groups are quite typical of current group catalogues . nonetheless , it , along with the esp group catalogue ( rzz99 ) , is the only group catalogue based upon a redshift survey covering a reasonably fair sample of the local universe . therefore , the properties of the lcrs group catalogue , containing groups from a wide range of environments , should be among the least biased to date . matchups of the lcrs groups with abell clusters indicated , not surprisingly , that , on average , these groups are much poorer than abell - class clusters , and therefore that lcrs groups do indeed inhabit a range of parameter space intermediate to that of individual galaxies and to that of rich clusters . we therefore conclude that this catalogue will be useful for a variety of studies requiring an unbiased census of loose groups , including the measurement of the luminosity function of group members versus that of field galaxies , the investigation of various morphology - environment relations , and the study of the clustering of groups . we thank the anonymous referee for the many useful comments . dlt would also like to thank marcio maia , ben moore , richard nolthenius , massimo ramella , and roberto trasarti - battistoni for valuable discussions by e - mail or in person concerning their group catalogues . the las campanas redshift survey has been supported by nsf grants ast 87 - 17207 , ast 89 - 21326 , and ast 92 - 20460 . this research has made use of the nasa / ipac extragalactic database ( ned ) , which is operated by the jet propulsion laboratory , caltech , under contract with the national aeronautics and space administration . this work was supported in part by the us department of energy under contract no . de - ac02 - 76ch03000 . we estimate @xmath198 via @xmath199 equation a1 is just the schechter ( 1976 ) luminosity function multiplied by a corrective factor @xmath200 and integrated over the interval @xmath201 . @xmath202 and @xmath203 are the extremal luminosities observable at a comoving distance @xmath174 under the flux and luminosity limits imposed on field @xmath25 , @xmath200 is the field - to - field sampling fraction for field @xmath25 , and @xmath204 , @xmath205 , and @xmath178 are the standard schechter function parameters . the value for @xmath42 is calculated in the same manner , @xmath206 where the values for @xmath200 , @xmath202 , @xmath207 , @xmath204 , @xmath205 , and @xmath178 are those for the fiducial field . as in lin et al . ( 1996 ) and tucker et al . ( 1997 ) , we make use of two different lcrs luminosity functions . the first version is the standard lcrs luminosity function ; it best describes the data from the 112-fiber fields and the northern galactic cap 50-fiber fields . since these data compose @xmath208 90% of the full lcrs sample , it is this version of the luminosity function which we use for our fiducial field . the schechter parameters for this luminosity function are as follows : @xmath44 ( @xmath209 is the absolute magnitude equivalent of @xmath205 ) . the measured luminosity function for the 50-fiber southern galactic cap data differs significantly from that of the other lcrs data ( lin et al . the reason behind this difference has never been fully resolved , but it is thought to be the effect of subtle selection problems in the very early lcrs data . we use the following values to describe the luminosity function of these southern 50-fiber data : @xmath210 [ note : both equations a1 and a2 are actually simplified forms of the integral found in the group - finding code . the integral in the code also includes effects due to apparent magnitude and surface brightness incompleteness and due to central surface brightness selection ; furthermore , in the code , this integral is convolved with a gaussian flux error of @xmath211 mag . for a detailed discussion of these additional selection effects , see 3.2 of lin et al . to correct for selection effects , @xmath137 is calculated by means of the following equation : @xmath212 where @xmath213 where @xmath94 is the number of _ observed _ members in the group , @xmath214 is the luminosity of group member @xmath21 , @xmath215 is the _ total _ expected luminosity density of _ all _ galaxies in the local universe , and @xmath216 is the expected luminosity density for only those galaxies that would lie within the photometric boundaries of lcrs field @xmath217 at a comoving distance @xmath218 . to estimate @xmath219 , we integrate the luminosity - weighted schechter function over all luminosities ( @xmath220 ) : @xmath221 where @xmath204 , @xmath178 , and @xmath205 are the standard schechter parameters , and where @xmath222 is the complete gamma function from mathematics ( abramowitz & stegun 1970 ) . we estimate @xmath223 via @xmath224 similar to the cases of equations a1 and a2 , equation b4 is just the luminosity - weighted schechter function multiplied by a corrective factor @xmath225 and integrated over the interval @xmath226 . @xmath202 and @xmath207 are the extremal luminosities observable at redshift @xmath227 under the given flux and luminosity limits imposed on field @xmath217 ; @xmath225 is the field - to - field sampling fraction for field @xmath217 . [ note : equation b4 is actually a simplified form of the integral found in the group - finding code . the integral in the code also includes effects due to apparent magnitude and surface brightness incompleteness and due to central surface brightness selection ; furthermore , in the code , this integral is convolved with a gaussian flux error of @xmath228 mag . for a detailed discussion of these additional selection effects , see 3.2 of lin et al . 1996 . ] the rms error in @xmath137 is estimated by summing the individual contributions from the rms errors in @xmath229 in quadrature , yielding @xmath230 where @xmath231 and @xmath232 we wish to make a quantitative , unbiased estimate of the richnesses of the las campanas loose groups . a useful and historically motivated method is to calculate their abell counts , @xmath233 . @xmath233 was defined by abell ( 1958 ) to be the number of galaxies , corrected for background contamination , in the magnitude interval @xmath152 to @xmath153 that lie within a @xmath234 ( @xmath235 mpc ) projected separation of a cluster s center ; the magnitude @xmath152 is the magnitude of the third brightest cluster member . due to the sampling characteristics and the relatively small range of apparent magnitudes within the lcrs , we can not use abell s definition directly . we must take a more circuitous path , via the use of the schechter function for groups and clusters of galaxies . for simplicity , consider a group which lies entirely within a single field . in this case , @xmath236 , the observed number of galaxies within a @xmath237 mpc projected separation of a group s barycenter , should fit the relation @xmath238 which is just the integral of the schechter function over the luminosity range @xmath201 . as in appendix b , @xmath202 and @xmath207 are the extremal luminosities observable at redshift @xmath227 under the given flux and luminosity limits imposed on field @xmath217 , and @xmath225 is the field - to - field sampling fraction for field @xmath217 . here , @xmath239 is a normalization factor for galaxy number counts within a projected radius of @xmath237 mpc of a given group s barycenter ; it is a counterpart to @xmath204 , which is used for the field galaxy luminosity function . the value for @xmath239 is itself a measure of richness ; the richer the group or cluster , the higher the value of @xmath239 ( assuming constant @xmath178 and @xmath205 ) . we will use @xmath239 to make an estimate of group abell richness . observationally , we can calculate a value for @xmath236 in equation c1 by subtracting off the estimated number of interlopers ( @xmath54 ; equation 14 ) from the total number of observed galaxies in a group within @xmath237 mpc of the group s barycenter ( @xmath240 ) , @xmath241\ ] ] then , placing this result into equation c1 , @xmath242 since the lcrs is a @xmath208 75%-sampled redshift catalogue with both bright and faint apparent magnitude cutoffs , it is quite possible that the third brightest group member may not be in the lcrs spectroscopic sample . therefore , an estimate for the absolute magnitude of the third brightest group member must be computed . we can do this by integrating the group s luminosity function , @xmath243 where @xmath244 is the incomplete gamma function ( abramowitz & stegun 1970 ) , and solving the equation @xmath245 numerically for the luminosity of the third brightest cluster member , @xmath246 . finally , taking our estimate of @xmath239 from equation c3 and our estimate of @xmath246 from equation c5 , we can calculate the group s abell richness , which we will call @xmath196 , @xmath247 where @xmath248 is the luminosity associated with the apparent magnitude @xmath153 at the redshift of the group in question . a further complication exists in that the linking radius employed in the `` friends - of - friends '' percolation algorithm is only @xmath208 1 @xmath65 mpc at the fiducial redshift @xmath249 ; for some tightly configured groups containing only a few observed members , the full abell radius of 1.5 @xmath65 mpc ( projected ) may not be completely searched , resulting in underestimates of the abell counts @xmath233 . since most lcrs groups are not very tightly configured ( consider the mean pairwise separations , @xmath83 ) , this is not likely to be a significant effect for the catalogue as a whole . in a similar case , apm clusters , which number counts only within a projected radius of 0.75 @xmath65 mpc from the cluster center , have been shown to underestimate their abell counts typically by only @xmath208 20% or less ( bahcall & west 1992 ) . the lcrs percolation algorithm , with its relatively large linking parameter @xmath23 ( typically @xmath250 mpc ) , should perform much better than do the apm counts . when compared with actual abell clusters within the lcrs volume , we find the following relation between the lcrs group counts estimate , @xmath151 , and the revised abell cluster values given by aco , @xmath187 : @xmath195 ( eq . [ 41 ] ) . we discuss this relation in more detail in 6 .

work on electroweak precision calculations and event generators for electroweak physics studies at current and future colliders is summarized .

hep-ph0501246

apart from the still missing higgs boson , the standard model ( sm ) has been impressively confirmed by successful collider experiments at the particle accelerators lep , slc , and tevatron during the last decade with high precision , at the level of radiative corrections . future colliders like the upcoming lhc or an @xmath0 international linear collider ( ilc ) , offer an even greater physics potential , and in turn represent a great challenge for theory to provide even more precise calculations . accurate predictions are necessary not only to increase the level of precision of sm tests , but also to study the indirect effects of possible new particles . with increasing energies many new processes with heavy particle production will be accessible . such heavy particles are unstable , so that their production leads to many - particle final states , with e.g. four or six fermions . predictions for such reactions should be based on full transition matrix elements , improved by radiative corrections as far as possible , and call for proper event generators . joint work done within the network is reviewed here ; more details can be found also in previous studies for future colliders @xcite . the precise measurement of the muon lifetime , or equivalently of the fermi constant @xmath3 , sets an important constraint on the sm parameters , = ( 1+r ) , with @xmath4 , where the quantity @xmath5 comprises the radiative corrections to muon decay ( apart from the photonic corrections in the fermi model ) . solving this relation for the w - boson mass @xmath6 yields a precise prediction for @xmath6 that can be compared with the directly measured value . recently the full electroweak two - loop calculation has been completed , with the contributions from closed fermion loops @xcite , from bosonic loops involving virtual higgs - bosons @xcite , and the complete bosonic corrections @xcite . the two - loop fermionic contributions influence the @xmath6 prediction at the level of @xmath7 , the two - loop bosonic corrections by @xmath8 . the predictions at the two - loop level have been further improved by universal higher - order contributions to the @xmath9-parameter . the terms of the order @xmath10 and @xmath11 have been obtained in @xcite and were found to change @xmath6 at the level of @xmath12 and @xmath13 , respectively . the leading three - loop bosonic contribution to the @xmath9-parameter in the large higgs mass limit @xcite , yielding the power @xmath14 , is opposite in sign to the leading two - loop correction @xmath15 . the two terms cancel each other for a higgs boson mass around 480 gev . this interesting new result stabilizes the perturbative expansion and makes a strongly interacting higgs sector very unlikely in view of the electroweak precision data . the prediction for @xmath6 , including the above - mentioned two - loop and leading three - loop effects , carries , besides the parametric uncertainty , an intrinsic theoretical uncertainty , which is estimated to be of @xmath16 @xcite , which has to be compared with the aimed precision of @xmath17 in the @xmath6 determination at a future ilc @xcite . in order to describe the z - boson resonance at lep1 within satisfactory precision it was possible to parametrize the cross section near the resonance in such a way @xcite that a born - like form with generalized `` effective '' couplings is convoluted with qed structure functions modeling initial - state radiation ( isr ) . from these effective z - boson fermion couplings so - called `` pseudo - observables '' were derived , such as various asymmetries , the hadronic z - peak cross section , partial z - decay widths , etc . the precisely calculated pseudo - observables are implemented in the programs zfitter and topaz0 @xcite . a critical overview on high - precision physics at the z pole , in particular focusing on the theoretical uncertainties , can be found in @xcite . the status of precision pseudo - observables in the mssm is summarized in @xcite . following the formal tree - level like parametrization of the couplings , an `` effective weak mixing angle '' , @xmath18 , was derived for each fermion species . among these parameters the leptonic variable @xmath19 plays a particularly important role , since it is measured with the high accuracy of @xmath20 and is very sensitive to the higgs - boson mass . quite recently , a substantial subclass of the electroweak two - loop contributions , the fermionic contributions to @xmath19 , were calculated @xcite ; they reduce the intrinsic theoretical uncertainty to @xmath21 . whether the pseudo - observable approach will also be sufficient for z - boson physics at the high - luminosity gigaz option remains to be investigated carefully . in any case , substantial theoretical progress will be needed to match the aimed gigaz precision on the theoretical side , e.g. for the expected experimental accuracy in @xmath19 of about @xmath22 . a full control of observables at the two - loop level , improved by leading higher - order effects , seems indispensable . an important entry for the precision observables with a large parametric uncertainty is the photonic vacuum polarization at the z scale . the hadronic part is determined via a dispersion relation from the cross section for @xmath23 with experimental data as input in the low - energy regime . possible scans with the radiative - return method require a careful theoretical treatment to reach the required precision @xcite . an independent precise determination of the electroweak mixing angle in terms of the @xmath24 ratio has been done in deep - inelastic neutrino scattering off an isoscalar target by the nutev collaboration @xcite , yielding a deviation of about 3 standard deviations from the value derived from the global analysis of the other precision observables . a new calculation of the electroweak one - loop corrections was performed @xcite to investigate the stability and size of the quantum effects , showing that the theoretical error of the analysis in @xcite was obviously underestimated . the new theoretical result is now being used for a re - analysis of the experimental data ( see also @xcite for another recent recalculation of the electroweak radiative corrections ) . although there are no complete next - to - next - to - leading ( nnlo ) predictions for @xmath25 scattering reactions and @xmath26 decays ( with one truly massive leg ) available yet , significant progress was reached in this direction in recent years . complete virtual two - loop amplitudes for ( massless ) bhabha scattering @xcite and light - by - light scattering @xcite have been worked out . also first steps have been made towards massive bhabha scattering in ref . @xcite the coefficient of the @xmath27 fixed - order contribution to elastic large - angle bhabha scattering is derived by adapting the classification of infrared divergences that was recently developed within dimensional regularization , and applying it to the regularization scheme with a massive photon and electron . the subset of factorizable corrections , resulting from one - loop subrenormalization , is considered in @xcite . this requires the evaluation of one - loop diagrams in arbitrary dimension @xmath28 . the @xmath29-expansion covering the orders @xmath30 and @xmath29 , in particular for the box graphs needed in bhabha scattering , was performed in @xcite , based on the work of @xcite . for the genuine two - loop qed corrections to bhabha scattering , the master integrals for the box with the fermionic loop were calculated @xcite , and the cross section with the corrections at two loop resulting from these diagrams @xcite . a complete set of master integrals for massive two - loop bhabha scattering has been derived in the meantime @xcite , and for form factors with massless @xcite and massive propagators @xcite . moreover , two - loop qcd corrections for the electroweak forward - backward asymmetries were obtained @xcite . also two - loop master integrals and form factors from virtual light fermions were derived and applied to higgs boson production and decays @xcite . electroweak corrections far above the electroweak scale , e.g. in the tev range , are dominated by soft and collinear gauge - boson exchange , leading to corrections of the form @xmath31 with @xmath32 . the leading terms ( @xmath33 ) are called sudakov logarithms . at the one - loop ( @xmath34 ) and two - loop ( @xmath35 ) level the leading and subleading corrections to a @xmath25 process at @xmath36 typically amount to @xcite _ ^ & ~ & -^2 ( ) -26% , + _ ^ & ~ & + ( ) 16% , + _ ^ & ~ & + ^4 ( ) 3.5% , + _ ^ & ~ & -^3 ( ) -4.2% , revealing that these corrections become significant in the high - energy phase of a future ilc . in contrast to qed and qcd , where the sudakov logarithms cancel in the sum of virtual and real corrections , these terms need not compensate in the electroweak sm for two reasons . the weak charges of quarks , leptons , and electroweak gauge bosons are open , not confined , i.e. there is ( in contrast to qcd ) no need to average or to sum over gauge multiplets in the initial or final states of processes . even for final states that are inclusive with respect to the weak charges sudakov logarithms do not completely cancel owing to the definite weak charges in the initial state @xcite . moreover , the large w- and z - boson masses make an experimental discrimination of real w- or z - boson production possible , in contrast to unobservable soft - photon or gluon emission . in recent years several calculations of these high - energy logarithms have been carried out in the sudakov regime , where all kinematical invariants @xmath37 of different particle momenta @xmath38 , @xmath39 are much larger than all particle masses . a complete analysis of all leading and subleading logarithms at the one - loop level can be found in @xcite . diagrammatic calculations of the leading two - loop sudakov logarithms have been carried out in @xcite . diagrammatic results on the so - called `` angular - dependent '' subleading logarithms have been presented in @xcite . all these explicit results are compatible with proposed resummations @xcite that are based on a symmetric su(2)@xmath40u(1 ) theory at high energies matched with qed at the electroweak scale . in this ansatz , improved matrix elements @xmath41 result from lowest - order matrix elements @xmath42 upon dressing them with ( operator - valued ) exponentials , ~_0 ( _ ) ( _ ) . explicit expressions for the electroweak and electromagnetic corrections @xmath43 and @xmath44 , which do not commute with each other , can , for instance , be found in @xcite . for @xmath25 neutral - current processes of four massless fermions , also subsubleading logarithmic corrections have been derived and resummed @xcite using an infrared evolution equation that follows the pattern of qcd . applications to vector - boson pair production in proton proton collisions can be found in @xcite . in supersymmetric models the form of radiative corrections at high energies has also been worked out for a broad class of processes @xcite . based on one - loop results their exponentiation has been proposed . photon radiation off initial - state electrons and positrons leads to large radiative corrections of the form @xmath45 . these logarithmic corrections are universal and governed by the dglap evolution equations . the solution of these equations for the electron - photon system yields the structure functions , generically denoted by @xmath46 below , which can be used via convolution to improve hard scattering cross sections @xmath47 by photon emission effects , ( p_e^+,p_e^- ) & = & _ 0 ^ 1 x_+(x_+ ) _ 0 ^ 1 x_-(x_- ) ( x_+p_e^+,x_-p_e^- ) . while the soft - photon part of the structure functions can be resummed , resulting in an exponential form , the contributions of hard photons have to be calculated order by order in perturbation theory . in @xcite the structure functions are summarized up to @xmath48 . describes a calculation of the ( non - singlet ) contributions up to @xmath49 and of the small-@xmath50 terms @xmath51^n$ ] to all orders ( for previous calculations see papers cited in ) . [ [ section ] ] four - fermion final states in @xmath52 collisions , which involve electroweak boson pair production , are of special interest since they allow the mechanism of spontaneous symmetry breaking and the non - abelian structure of the standard model to be directly tested by the experiments . moreover they provide a very important background to most searches for new physics . lep2 has provided in this respect an ideal testing ground for the sm . the w profile has been measured with great accuracy , new bounds on anomalous trilinear gauge - boson couplings have been set , and single w , single z , zz , and z@xmath53 cross sections have been determined for the first time . these studies will be continued with much higher statistics and energy at a future @xmath54 linear collider . in this context , the montecarlo four fermion generator wphact 2.0 has been completed @xcite , adapted to experimental requests and used for simulation of the lep2 data @xcite . wphact 2.0 computes all standard model processes with four fermions in the final state at @xmath52 colliders , it makes use of complete , fully massive helicity amplitude calculations and includes the imaginary fermion loop gauge restoring scheme is discussed in terms of an effective lagrangian approach . ] . thanks to these features and new phase space mappings , wphact has been extended to all regions of phace space , including kinematical configurations dominated by small momentum transfer and small invariant masses like single w , single z , z@xmath55 and @xmath56 processes . special attention has been devoted to qed effects , which have a large numerical impact , with new options for the description of initial state radiation ( isr ) and of the scale dependence of the electromagnetic coupling . moreover , there is the possibility of including ckm mixing , and to account for resonances in @xmath57 channels . the theoretical treatment and the presently gained level in accuracy in the description of w - pair - mediated @xmath58 production were triggered by lep2 , as it is reviewed in . the @xmath59 bosons are treated as resonances in the full @xmath58 processes , @xmath60 . radiative corrections are split into universal and non - universal corrections . the former comprise leading - logarithmic corrections from isr , higher - order corrections included by using appropriate effective couplings , and the coulomb singularity . these corrections can be combined with the full lowest - order matrix elements easily . the remaining corrections are called non - universal , since they depend on the process under investigation . for lep2 accuracy , it was sufficient to include these corrections by the leading term of an expansion about the two @xmath59 poles , defining the so - called double - pole approximation ( dpa ) . different versions of such a dpa have been used in the literature @xcite . although several monte carlo programs exist that include universal corrections , only two event generators , yfsww @xcite and racoonww @xcite , include non - universal corrections , as well as the option of anomalous gauge couplings @xcite . in the dpa approach , the w - pair cross section can be predicted within a relative accuracy of @xmath61 in the energy range between @xmath62 ( @xmath63 ) and @xmath64 , which was sufficient for the lep2 accuracy of @xmath65 for energies @xmath66 . at threshold ( @xmath67 ) , the present state - of - the - art prediction results from an improved born approximation based on leading universal corrections only , because the dpa is not reliable there , and thus possesses an intrinsic uncertainty of about @xmath68 , which demonstrates the necessary theoretical improvements for the threshold region . at energies beyond @xmath64 , effects beyond @xmath69 , such as the above - mentioned sudakov logarithms at higher orders , become important and have to be included in predictions at per - cent accuracy . at lep2 , the w - boson mass has been determined by the reconstruction of w bosons from their decay products with a final accuracy of about @xmath70 . in @xcite the theoretical uncertainty is estimated to be of the order of @xmath71 . theoretical improvements are , thus , desirable . the above discussion illustrates the necessity of a full one - loop calculation for the @xmath72 process and of further improvements by leading higher - order corrections . the single - w production process @xmath73 plays a particularly important role among the @xmath58 production processes at high scattering energies . the process is predominantly initiated by @xmath74 collision , where the photon is radiated off the electron ( or positron ) by the weizscker williams mechanism , i.e. with a very small off - shellness @xmath75 . consequently the cross section rises logarithmically with the scattering energy and is of the same size as the w - pair production cross section around @xmath76 ; for higher energies single - w dominates over w - pair production . theoretically the dominance of photon exchange at low @xmath75 poses several complications . technically , @xmath77 means that the electrons ( or positrons ) are produced in the forward direction and that the electron mass has to be taken into account in order to describe the cross section there . moreover , the mere application of @xmath78-dependent leading - logarithmic structure functions does not describe the leading photon - radiation effects properly , since isr and final - state radiation ( fsr ) show sizeable interferences for forward scattering . thus , the improvement of lowest - order calculations by leading radiation effects is more complicated than for @xmath78-channel - like processes . finally , the running of the electromagnetic coupling @xmath79 has to be evaluated in the region of small momentum transfer ( @xmath80 ) where the fit for the hadronic part of the photon vacuum polarisation @xcite should be used . the monte carlo generator koralw @xcite has been updated to include the isr - fsr interference effects as well as the proper running of @xmath79 . therefore , this program now has reached a level of accuracy similar to the other state - of - the - art programs for single - w production : grc4f @xcite , nextcalibur @xcite , swap @xcite , wphact @xcite , and wto @xcite . it should be kept in mind that none of these calculations includes non - universal electroweak corrections , leading to a theoretical uncertainty of about @xmath81 in cross - section predictions . although the final solution for a high - energy linear collider certainly requires a full @xmath69 calculation of the @xmath58-production process , a first step of improvement could be done by a careful expansion about the propagator poles of the photon and w boson . the electroweak @xmath69 corrections to the process @xmath82 , which are known @xcite , represent a basic building block in this calculation . -1 one - loop integrals become more and more cumbersome if the number @xmath83 of external legs in diagrams increases . for @xmath84 , however , not all external momenta are linearly independent because of the four - dimensionality of space - time . as known for a long time @xcite , this fact opens the possibility to relate integrals with @xmath84 to integrals with @xmath85 . in recent years , various techniques for actual evaluations of one - loop integrals with @xmath86 have been worked out @xcite ( see also references therein for older methods and results ) . the major complication in the treatment of @xmath87 processes at one loop concerns the numerical evaluation of tensor 5-point integrals ; in particular , the occurrence of inverse gram determinants in the usual passarino veltman reduction to scalar integrals leads to numerical instabilities at the phase - space boundary . a possible solution to this problem was worked out in @xcite where the known direct reduction @xcite of scalar 5-point to 4-point integrals was generalized to tensor integrals , thereby avoiding the occurrence of leading gram determinants completely . more work on one - loop @xmath83-point integrals can be found in @xcite . in the evaluation of real corrections , such as bremsstrahlung , a proper and numerically stable separation of infrared ( soft and collinear ) divergences represents one of the main problems . in the phase - space slicing approach ( see @xcite and references therein ) the singular regions are excluded from the `` regular '' phase - space integration by small cuts on energies , angles , or invariant masses . using factorization properties , the integration over the singular regions can be done in the limit of infinitesimally small cut parameters . the necessary fine - tuning of cut parameters is avoided in so - called subtraction methods ( see @xcite and references therein ) , where a specially tuned auxiliary function is subtracted from the singular integrand in such a way that the resulting integral is regular . the auxiliary function has to be chosen simple enough , so that the singular regions can be integrated over analytically . in @xcite the so - called `` dipole subtraction approach '' has been worked out for massless qcd . and subsequently extended for photon emission off massive fermions @xcite and for qcd with massive quarks @xcite . applications were performed for complete one - loop calculations of electroweak radiative corrections for specific @xmath87 processes of special interest for a future ilc , @xmath88 @xcite and @xmath89 @xcite . in @xcite the technique @xcite for treating tensor 5-point integrals was employed . while @xcite make use of the slicing approach for treating soft - photon emission , the results of @xcite have been obtained by dipole subtraction and checked by phase - space slicing for soft and collinear bremsstrahlung . analytic results for the one - loop corrections to @xmath88 are provided in @xcite . the yukawa coupling of the top quark could be measured at a future ilc with high energy and luminosity at the level of @xmath81 @xcite by analyzing the process @xmath90 . a thorough prediction for this process , thus , has to control qcd and electroweak corrections . results on the electroweak @xmath91 corrections of show agreement within @xmath92 . the results of the previous calculation @xcite roughly agree with the ones of at intermediate values of @xmath93 and @xmath94 , but are at variance at high energies ( tev range ) and close to threshold ( large @xmath94 ) . the large variety of different final states for multi - particle production renders multi - purpose monte carlo event generators rather important , i.e. generators that deliver an event generator for a user - specified ( as much as possible ) general final state based on full lowest - order amplitudes . as results , these tools yield lowest - order predictions for observables , or more generally monte carlo samples of events , that are improved by universal radiative corrections , such as initial - state radiation at the leading - logarithmic level or beamstrahlung effects . most of the multi - purpose generators are also interfaced to parton - shower and hadronization programs . the generality renders these programs , however , rather complex devices and , at present , they are far from representing tools for high - precision physics , because non - universal radiative corrections are not taken into account in predictions . the following multi - purpose generators for multi - parton production , including program packages for the matrix - element evaluation , are available : * amegic @xcite : helicity amplitudes are automatically generated by the program for the sm , the mssm , and some new - physics models . various interfaces ( isr , pdfs , beam spectra , isajet , etc . ) are supported . the phase - space generation was successfully tested for up to six particles in the final state . * grace @xcite : the amplitudes are delivered by a built - in package , which can also handle susy processes . the phase - space integration is done by bases @xcite . tree - level calculations have been performed for up to ( selected ) six - fermion final states . the extension of the system to include one - loop corrections is the grace - loop @xcite program . * madevent @xcite + madgraph @xcite : the madgraph algorithm can generate tree - level matrix elements for any sm process ( fully supporting particle masses ) , but a practical limitation is 9,999 diagrams . in addition , madgraph creates madevent , an event generator for the requested process . * phegas @xcite + helac @xcite : the helac program delivers amplitudes for all sm processes ( including all masses ) . the phase - space integration done by phegas has been tested for selected final states with up to seven particles . recent applications concern channels with six - fermion final states @xcite . * whizard @xcite + comphep @xcite / madgraph @xcite / omega @xcite : matrix elements are generated by an automatic interface to ( older versions of ) comphep , madgraph , and ( the up - to - date version of ) omega . phase - space generation has been tested for most @xmath95 and some @xmath96 processes ; unweighted events are supported , and a large variety of interfaces ( isr , beamstrahlung , pythia , pdfs , etc . ) exists . the inclusion of mssm amplitudes ( omega ) and improved phase - space generation ( @xmath95 ) are work in progress . * alpgen @xcite is a specific code for computing the perturbative part of observables in high energy hadron hadron collisions , which require a convolution of the perturbative quantities with structure or fragmentation functions that account for non - perturbative effects . tuned comparisons of different generators , both at parton and detector level , are extremely important , but become more and more laborious owing to the large variety of multi - particle final states . some progress to a facilitation and automization of comparisons are made by mc - tester project @xcite and java interfaces @xcite . particular progress was reached in recent years in the description of six - fermion production processes . apart from the multi - purpose generators listed in the previous section , also dedicated monte carlo programs and generators have been developed for this class of processes : * sixfap @xcite : matrix elements are provided for all @xmath98 final states ( with finite fermion masses ) , including all electroweak diagrams . the generalization to qcd diagrams and the extension of the phase - space integration for all final states is in progress . * eett6f @xcite : only processes relevant for @xmath99 production are supported ( a new version includes @xmath100 in the final state and qcd diagrams ) ; finite fermion masses are possible . * lusifer @xcite : all @xmath98 final states are possible , including qcd diagrams with up to four quarks ; representative results for all these final states have been presented . external fermions are massless . an unweighting algorithm and an interface to pythia are available . * phase @xcite : all standard model process with six fermions in the final state at the lhc . it employs the full set of tree - level feynman diagrams , taking into account a finite mass for the b quark . an interface to hadronization is provided . a comparison of results for some processes @xmath101 relevant for @xmath99 production for massless external fermions reveals good agreement between the various programs @xcite , where minor differences are presumably due to the different treatments of the bottom - quark yukawa coupling , which is neglected in some cases . a tuned comparison of results obtained with lusifer and whizard for a large survey of @xmath98 final states has been presented in . once the necessary techniques and theoretical subtleties of a perturbative calculation are settled , to carry out the actual calculation is an algorithmic matter . thus , an automization of such calculations is highly desirable , in order to facilitate related calculations . various program packages have been presented in the literature for automatized tree - level , one - loop , and multi - loop calculations . a comprehensive overview can , e.g. , be found in @xcite ; in the following we have to restrict ourselves to a selection of topics , where emphasis is put on electroweak aspects . the generation of feynman graphs and amplitudes is a combinatorial problem that can be attacked with computer algebra . the program packages feynarts @xcite ( which has been extended in @xcite for the mssm ) and diana @xcite ( based on qgraf @xcite ) are specifically designed for this task ; also the grace - loop @xcite system automatically generates feynman diagrams and loop amplitudes . moreover , the task of calculating virtual one - loop and the corresponding real - emission corrections to @xmath102 scattering reactions is by now well understood . such calculations are widely automated in the packages formcalc combined with looptools @xcite , and grace - loop @xcite . an illustrating example was provided for the differential cross section for @xmath103 in lowest order as well as including electroweak @xmath69 corrections . a program fa+fc @xcite was obtained from the output of the feynarts and formcalc packages and makes use of the looptools library for the numerical evaluation . another program , topfit @xcite , was developed from an algebraic reduction of feynman graphs ( delivered from diana ) within form ; for the numerics looptools is partially employed . a completely independent approach has been made by the sanc project @xcite . the sanc program contains another independent calculation of the @xmath69 corrections to @xmath103 , the results of which are also included in @xcite . more details on comparisons , including also other fermion flavours , can be found in @xcite . the agreement between the numerical results at the level of 10 digits reflects the enormous progress achieved in recent years in the automatization of one - loop calculations . the one - loop calculation for the process @xmath104 including hard bremsstrahlung was originally performed in @xcite without full automatization ; it was repeated in later course ( apart from the hard bremsstrahlung part ) as an example for automatization in @xcite . the extension of diana towards full automatization in terms of the package is a new development @xcite ; automatization of the full calculation is performed including renormalization and the creation and running of a fortran code . applications to the calculation of the processes @xmath105 for various final fermions : @xmath106 and also @xmath107 are performed . for further work in automatization see @xcite . most of the various techniques of performing loop calculations share the common feature that the integration over the loop momenta is performed analytically . this procedure leads to complications at one loop if five or more external legs are involved , since both speed and stability of programs become more and more jeopardized . at the two - loop level , already the evaluation of self - energy and vertex corrections can lead to extremely complicated higher transcendental functions that are hard to evaluate numerically . an idea to avoid these complications is provided by a more or less purely numerical evaluation of loop corrections . there are two main difficulties in this approach . firstly , the appearing ultraviolet and infrared divergences have to be treated and canceled carefully . secondly , even finite loop integrals require a singularity handling of the integrand near particle poles , where feynman s @xmath108 prescription is used as regularization . in @xcite-@xcite a method for a purely numerical evaluation of loop integrals is proposed . each integral is parametrized with feynman parameters and subsequently rewritten with partial integrations . the final expression consists of a quite simple part containing the singular terms and another more complicated looking part that can be integrated numerically . the actual application of the method to a physical process is still work in progress . there are five papers in a series devoted to the numerical evaluation of multi - loop , multi - leg feynman diagrams . in @xcite the general strategy is outlined and in @xcite a complete list of results is given for two - loop functions with two external legs , including their infrared divergent on - shell derivatives . results for one - loop multi - leg diagrams are shown in @xcite and additional material can be found in @xcite . two - loop three - point functions for infrared convergent configurations are considered in @xcite , where numerical results can be found . presents a detailed investigation of the algorithms , based on the bernstein - tkachov ( bt ) theorem @xcite , which form the basis for a fast and reliable numerical integration of one - loop multi - leg ( up to six in this paper ) diagrams . the rationale for this work is represented by the need of encompassing a number of problems that one encounters in assembling a calculation of some complicated process , e.g. full one - loop corrections to @xmath109fermions . furthermore , in any attempt to compute physical observables at the two - loop level , we will have to include the one - loop part , and it is rather obvious that the two pieces should be treated on equal footing . all algorithms that aim to compute feynman diagrams numerically are based on some manipulation of the original integrands that brings the final answer into something smooth . this has the consequence of bringing the original ( landau ) singularity of the diagram into some overall denominator and , usually , the method overestimates the singular behavior around some threshold . in these regions an alternative derivation is needed . instead of using the method of asymptotic expansions , a novel algorithm is introduced based on a mellin - barnes decomposition of the singular integrand , followed by a sector decomposition that allows one to write the laurent expansion around threshold . particular care has been devoted to analyze those situations where a sub - leading singularity may occur , and to properly account for those cases where the algorithm can not be applied because the corresponding bt factor is zero although the singular point in parametric space does not belong to the integration domain . finally , a description of infrared divergent one - loop virtual configurations is given in the framework of dimensional regularization : here both the residue of the infrared pole and the infrared finite remainder are cast into a form that can be safely computed numerically . the collection of formulas that cover all corners of phase space have been translated into a set of form codes and the output has been used to create a fortran code whose technical description will be given elsewhere . addresses the problem of deriving a judicious and efficient way to deal with tensor feynman integrals , namely those integrals that occur in any field theory with spin and non trivial structures for the numerators of feynman propagators . this paper forms a basis for a realistic calculation of physical observables at the two - loop level . the complexity of handling two - loop tensor integrals is reflected in the following simple consideration : the complete treatment of one - loop tensor integrals was confined to the appendices of @xcite , while the reduction of general two - loop self - energies to standard scalar integrals already required a considerable fraction of @xcite ; the inclusion of two - loop vertices requires the whole content of this paper . the past experience in the field has shown the encompassed convenience of gathering in one single place the complete collection of results needed for a broad spectrum of applications . in recent years , the most popular and quite successful tool in dealing with multi - loop feynman diagrams in qed / qcd ( or in selected problems in different models , characterized by a very small number of scales ) , has been the integration - by - parts identities method @xcite . however , reduction to a set of master integrals is poorly known in the enlarged scenario of multi - scale electroweak physics . in @xcite another new method is presented in which almost all the work can be performed numerically : one - loop tensor integrals are numerically reduced to the standard set of one - loop scalar functions and any amplitude is calculated simply contracting the numerically computed tensor integrals with the external tensors . to this aim , a recursion relation is introduced that links high - rank tensor integrals to lower - rank ones . singular kinematical configurations give a smoother behaviour than in other approaches because , at each level of iteration , only inverse square roots of gram determinants appear . in refs . @xcite ( see also @xcite ) it was proved the importance of electroweak one - loop corrections to hadronic observables , such as @xmath110 , ` prompt photon + jet ' and ` @xmath2 + jet ' production at tevatron and lhc and jet and @xmath110 production in linear colliders , which can compete in size with qcd corrections . their inclusion in experimental analyses is thus important , especially in view of searches for new physics . in case of ` @xmath2 + jet ' production they can rise to @xmath111 at large transverse momentum at the lhc , while being typically half the size in case of ` photon + jet ' production . as these two channels are the contributors to the drell - yan process , and since the latter is envisaged to be used as one of the means to measure the lhc luminosity , it is clear that neglecting them in experimental analyses would spoil the luminosity measurements . ref . @xcite emphasised the importance of nlo electroweak effects in three - jet production in @xmath112 scattering at the @xmath2-pole ( slc , lep and gigaz ) , showing typical corrections of @xmath113 ( e.g. in jet - rates and thrust ) , comparable to the slc and lep experimental accuracy and certainly larger than the one expected at gigaz . they also introduce sizable parity - violating effects into the fully differential structure of three - jet events in presence of beam polarisation , which are of relevance as background to new physics searches in slc and gigaz data . the complete set of electroweak @xmath114 corrections to the drell yan - like production of @xmath2 and @xmath1 bosons have been studied in @xcite . these corrections are phenomenologically relevant both at the tevatron and the lhc . it is shown that the pole expansion yields a good description of resonance observables , but it is not sufficient for the high - energy tail of transverse - momentum distributions , relevant for new - physics searches . horace and winhac are monte carlo event generators @xcite , developed for single w production and decay , which in their current versions include higher - order qed corrections in leptonic w decays , a crucial entry for precision determination of the w mass and width at hadron colliders . production of vector - boson pairs is an important probe for potential non - standard anomalous gauge couplings . in order to identify possible deviations from the sm predictions , an accurate knowledge of the electroweak higher - order contributions is mandatory as well , in particular for large transverse momenta . a complete electroweak one - loop calculation was performed for @xmath115 production @xcite ; for other processes like @xmath116 the large logarithms in the sudakov regime were derived @xcite . a further aspect that should be recalled is that weak corrections naturally introduce parity - violating effects in observables , detectable through asymmetries in the cross - section . these effects are further enhanced if polarisation of the incoming beams is exploited , such as at rhic - spin @xcite and will be used to measure polarised structure functions . during the recent years there has been continuous progress in the development of new techniques and in making precise predictions for electroweak physics at future colliders . however , to be prepared for the lhc and a future @xmath0 linear collider with high energy and luminosity , an enormous amount of work is still ahead of us . not only technical challenges , also field - theoretical issues such as renormalization , the treatment of unstable particles , etc . , demand a higher level of understanding . both loop calculations as well as the descriptions of multi - particle production processes with the help of monte carlo techniques require and will profit from further improving computing devices . it is certainly out of question that the list of challenges and interesting issues could be continued at will . electroweak physics at future colliders will be a highly exciting issue . j. a. aguilar - saavedra _ et al . _ , tesla technical design report part iii : _ physics at an @xmath117 linear collider _ [ hep - ph/0106315 ] . t. abe _ et al . _ [ american linear collider working group collaboration ] , in _ proc . of the aps / dpf / dpb summer study on the future of particle physics ( snowmass 2001 ) _ ed . r. davidson and c. quigg , slac - r-570 , _ resource book for snowmass 2001 _ [ hep - ex/0106055 , hep - ex/0106056 , hep - ex/0106057 , hep - ex/0106058 ] . k. abe _ et al . _ [ acfa linear collider working group collaboration ] , acfa linear collider working group report , [ hep - ph/0109166 ] . u. baur _ et al . _ , in _ proc . of the aps / dpf / dpb summer study on the future of particle physics ( snowmass 2001 ) _ ed . n. graf , hep - ph/0111314 and hep - ph/0202001 . a. freitas , w. hollik , w. walter and g. weiglein , phys . b * 495 * ( 2000 ) 338 , e : ibid . b * 570 * ( 2003 ) 260 [ hep - ph/0007091 ] and nucl . phys . b * 632 * ( 2002 ) 189 , e : ibid . b * 666 * ( 2003 ) 305 [ hep - ph/0202131 ] . m. awramik and m. czakon , phys . b * 568 * , 48 ( 2003 ) [ hep - ph/0305248 ] . m. awramik and m. czakon , phys . * 89 * ( 2002 ) 241801 [ hep - ph/0208113 ] . berends _ et al . _ , in _ z physics at lep 1 _ , eds . g. altarelli , r. kleiss and c. verzegnassi ( cern 89 - 08 , geneva , 1989 ) , vol . 1 , p. 89 ; + w. beenakker , f. a. berends and s. c. van der marck , z. phys . c * 46 * ( 1990 ) 687 ; + a. borrelli , m. consoli , l. maiani and r. sisto , nucl . b * 333 * ( 1990 ) 357 . d. y. bardin and g. passarino , hep - ph/9803425 ; + g. montagna , o. nicrosini , f. piccinini and g. passarino , comput . commun . * 117 * ( 1999 ) 278 [ hep - ph/9804211 ] ; + g. montagna , f. piccinini , o. nicrosini , g. passarino and r. pittau , comput . commun . * 76 * ( 1993 ) 328 ; + g. montagna , f. piccinini , o. nicrosini , g. passarino and r. pittau , nucl . b * 401 * ( 1993 ) 3 ; + d. y. bardin , m. grnewald and g. passarino , hep - ph/9902452 ; + d. y. bardin _ et al . _ , comput . commun . * 133 * ( 2001 ) 229 [ hep - ph/9908433 ] . contributions of w. hollik , w. van neerven , and g. passarino to the proceedings of the zeuthen workshop _ electroweak precision data and the higgs mass _ , zeuthen , 2003 , eds . s. dittmaier and k. mnig , desy - proc-2003 - 01 and mpp-2003 - 50 . g. p. zeller _ et al . _ [ nutev collaboration ] , phys . * 88 * ( 2002 ) 091802 [ erratum - ibid . * 90 * ( 2003 ) 239902 ] [ hep - ex/0110059 ] . k. p. o. diener , s. dittmaier and w. hollik , phys . d * 69 * ( 2004 ) 073005 [ hep - ph/0310364 ] . z. bern , l. j. dixon and a. ghinculov , phys . d * 63 * ( 2001 ) 053007 [ hep - ph/0010075 ] . z. bern _ et al . _ , jhep * 0111 * ( 2001 ) 031 [ hep - ph/0109079 ] . glover , j.b . tausk and j.j . van der bij , phys . b * 516 * ( 2001 ) 33 [ hep - ph/0106052 ] . j. fleischer , t. riemann , o.v . tarasov and a. werthenbach , nucl . . suppl . * 116 * ( 2003 ) 43 [ lc - th-2003 - 082 , hep - ph/0211167 ] . j. fleischer , t. riemann and o.v . tarasov , acta phys . b * 34 * ( 20003 ) 5345 . j. fleischer , f. jegerlehner and o.v . tarasov , nucl . b * 672 * ( 2003 ) 303 , [ hep - ph/0307113 ] . j. fleischer , o.v . tarasov and v.o . tarasov , phys . b * 584 * ( 2004 ) 294 , [ hep - ph/0401090 ] . birthwright , e.w.n . glover and p. marquard , jhep * 0409 * , 042 ( 2004 ) [ hep - ph/0407343 ] . u. aglietti and r. bonciani , nucl . phys . b * 668 * , 3 ( 2003 ) [ hep - ph/0304028 ] . u. aglietti and r. bonciani , hep - ph/0401193 . r. bonciani , p. mastrolia and e. remiddi , nucl . b * 690 * , 138 ( 2004 ) [ hep - ph/0311145 ] . u. aglietti , r. bonciani , g. degrassi and a. vicini , hep - ph/0407162 . u. aglietti , r. bonciani , g. degrassi and a. vicini , phys . b * 595 * , 432 ( 2004 ) hep - ph/0404071 ] . a. denner , m. melles and s. pozzorini , nucl . b * 662 * ( 2003 ) 299 [ hep - ph/0301241 ] . m. ciafaloni , p. ciafaloni and d. comelli , phys . b * 501 * ( 2001 ) 216 [ hep - ph/0007096 ] . a. denner and s. pozzorini , eur . j. c * 18 * ( 2001 ) 461 [ hep - ph/0010201 ] and eur . j. c * 21 * ( 2001 ) 63 [ hep - ph/0104127 ] . w. beenakker and a. werthenbach , phys . b * 489 * ( 2000 ) 148 [ hep - ph/0005316 ] and nucl . b * 630 * ( 2002 ) 3 [ hep - ph/0112030 ] ; + m. melles , phys . b * 495 * ( 2000 ) 81 [ hep - ph/0006077 ] ; + m. hori , h. kawamura and j. kodaira , phys . b * 491 * ( 2000 ) 275 [ hep - ph/0007329 ] . v. s. fadin , l. n. lipatov , a. d. martin and m. melles , phys . d * 61 * ( 2000 ) 094002 [ hep - ph/9910338 ] ; + j. h. khn , a. a. penin and v. a. smirnov , eur . j. c * 17 * ( 2000 ) 97 [ hep - ph/9912503 ] ; + m. melles , phys . * 375 * ( 2003 ) 219 [ hep - ph/0104232 ] and eur . j. c * 24 * ( 2002 ) 193 [ hep - ph/0108221 ] . j. h. khn , s. moch , a. a. penin and v. a. smirnov , nucl . b * 616 * ( 2001 ) 286 [ erratum - ibid . b * 648 * ( 2003 ) 455 ] [ hep - ph/0106298 ] . e. accomando , a. denner and s. pozzorini , phys . d * 65 * ( 2002 ) 073003 [ hep - ph/0110114 ] . e. accomando , a. denner and a. kaiser , hep - ph/0409247 . m. beccaria _ et al . _ , phys . d * 64 * ( 2001 ) 053016 [ hep - ph/0104245 ] ; phys . d * 65 * ( 2002 ) 093007 [ hep - ph/0112273 ] ; lc - th-2002 - 005 [ hep - ph/0203254 ] ; hep - ph/0212167 ; lc - th-2002 - 017 [ hep - ph/0212247 ] ; hep - ph/0304110 and nucl . b * 663 * ( 2003 ) 394 [ hep - ph/0304175 ] . w. beenakker _ et al . _ , in _ physics at lep2 _ , eds . g. altarelli , t. sjstrand and f. zwirner ( cern 96 - 01 , geneva , 1996 ) , vol . 1 , p. 79 [ hep - ph/9602351 ] . j. blmlein and h. kawamura , acta phys . b * 33 * ( 2002 ) 3719 [ hep - ph/0207259 ] . e. accomando , a. ballestrero and e. maina , 150 2003 166 . a. ballestrero , r. chierici , f. cossutti , e. migliore , 152 2003 175 . w. beenakker , a. p. chapovsky , a. kanaki , c. g. papadopoulos and r. pittau , nucl . b * 667 * , 359 ( 2003 ) hep - ph/0303105 ] . m. w. grnewald _ et al . _ , in _ reports of the working groups on precision calculations for lep2 physics _ , eds . s. jadach , g. passarino and r. pittau ( cern 2000 - 009 , geneva , 2000 ) , p. 1 [ hep - ph/0005309 ] . w. beenakker , f. a. berends and a. p. chapovsky , nucl . b * 548 * , 3 ( 1999 ) [ hep - ph/9811481 ] ; + y. kurihara , m. kuroda and d. schildknecht , nucl . b * 565 * ( 2000 ) 49 [ hep - ph/9908486 ] and phys . lett . b * 509 * ( 2001 ) 87 [ hep - ph/0104201 ] . s. jadach _ et al . _ , phys . d * 61 * ( 2000 ) 113010 [ hep - ph/9907436 ] ; phys . d * 65 * ( 2002 ) 093010 [ hep - ph/0007012 ] ; comput . commun . * 140 * ( 2001 ) 432 [ hep - ph/0103163 ] and comput . phys . commun . * 140 * ( 2001 ) 475 [ hep - ph/0104049 ] . a. denner , s. dittmaier , m. roth and d. wackeroth , phys . b * 475 * ( 2000 ) 127 [ hep - ph/9912261 ] ; lc - th-2000 - 014 [ hep - ph/9912447 ] ; nucl . b * 587 * ( 2000 ) 67 [ hep - ph/0006307 ] ; hep - ph/0101257 and hep - ph/0209330 . a. denner , s. dittmaier , m. roth and d. wackeroth , nucl . b * 560 * ( 1999 ) 33 [ hep - ph/9904472 ] and eur . j. c * 20 * ( 2001 ) 201 [ hep - ph/0104057 ] . a. denner , s. dittmaier , m. roth and d. wackeroth , hep - ph/0110402 . s. jadach _ et al . _ , phys . b * 523 * ( 2001 ) 117 [ hep - ph/0109072 ] . f. jegerlehner , lc - th-2001 - 035 [ hep - ph/0105283 ] . s. jadach _ et al . _ , comput . commun . * 119 * ( 1999 ) 272 [ hep - ph/9906277 ] and eur . j. c * 27 * ( 2003 ) 19 [ hep - ph/0209268 ] . j. fujimoto _ et al . _ , hep - ph/9603394 and comput . commun . * 100 * ( 1997 ) 128 [ hep - ph/9605312 ] . f. a. berends , c. g. papadopoulos and r. pittau , hep - ph/0002249 and comput . commun . * 136 * ( 2001 ) 148 [ hep - ph/0011031 ] . g. montagna _ el al . _ , eur . j. c * 20 * ( 2001 ) 217 [ hep - ph/0005121 ] . e. accomando and a. ballestrero , comput . commun . * 99 * ( 1997 ) 270 [ hep - ph/9607317 ] ; + e. accomando , a. ballestrero and e. maina , phys . b * 479 * ( 2000 ) 209 [ hep - ph/9911489 ] and comput . phys . commun . * 150 * ( 2003 ) 166 [ hep - ph/0204052 ] . g. passarino , comput . commun . * 97 * ( 1996 ) 261 [ hep - ph/9602302 ] ; hep - ph/9810416 and nucl . b * 578 * ( 2000 ) 3 [ hep - ph/0001212 ] . a. denner and s. dittmaier , nucl . b * 398 * ( 1993 ) 239 ; + m. bhm and s. dittmaier , nucl . b * 409 * ( 1993 ) 3 . d. b. melrose , _ nuovo cimento _ * xl a * ( 1965 ) 181 . j. fleischer , f. jegerlehner and o. v. tarasov , nucl . b * 566 * ( 2000 ) 423 [ hep - ph/9907327 ] ; + t. binoth , j. p. guillet and g. heinrich , nucl . b * 572 * ( 2000 ) 361 [ hep - ph/9911342 ] ; + f. tramontano , phys . d * 67 * ( 2003 ) 114005 [ hep - ph/0211390 ] ; + t. binoth , g. heinrich and n. kauer , nucl . b * 654 * ( 2003 ) 277 [ hep - ph/0210023 ] . a. denner and s. dittmaier , nucl . b * 658 * ( 2003 ) 175 [ hep - ph/0212259 ] . @xcite w. t. giele and e. w. n. glover , jhep * 0404 * , 029 ( 2004 ) hep - ph/0402152 ] . + s. dittmaier , nucl . b * 675 * , 447 ( 2003 ) [ hep - ph/0308246 ] . s. dittmaier , nucl . b * 565 * ( 2000 ) 69 [ hep - ph/9904440 ] . l. phaf and s. weinzierl , jhep * 0104 * , 006 ( 2001 ) [ hep - ph/0102207 ] ; + s. catani , s. dittmaier , m. h. seymour and z. trcsnyi , nucl . b * 627 * ( 2002 ) 189 [ hep - ph/0201036 ] . g. blanger _ et al . _ , b * 559 * ( 2003 ) 252 [ hep - ph/0212261 ] . a. denner , s. dittmaier , m. roth and m. m. weber , phys . b * 560 * ( 2003 ) 196 [ lc - th-2003 - 008 , hep - ph/0301189 ] and nucl . b * 660 * ( 2003 ) 289 [ lc - th-2003 - 007 , hep - ph/0302198 ] . y. you _ et al . _ , b * 571 * , 85 ( 2003 ) hep - ph/0306036 ] . f. krauss , r. kuhn and g. soff , jhep * 0202 * ( 2002 ) 044 [ hep - ph/0109036 ] ; + a. schalicke , f. krauss , r. kuhn and g. soff , jhep * 0212 * ( 2002 ) 013 [ hep - ph/0203259 ] . t. ishikawa _ et al . _ , [ minami - tateya collaboration ] , kek-92 - 19 ; + h. tanaka _ et al . _ , [ minami - tateya collaboration ] , nucl . instrum . meth . a * 389 * ( 1997 ) 295 ; + f. yuasa _ et al . _ , prog . * 138 * ( 2000 ) 18 [ hep - ph/0007053 ] . s. kawabata , comput . commun . * 88 * ( 1995 ) 309 . j. fujimoto _ el al . _ , acta phys . b * 28 * ( 1997 ) 945 . f. maltoni and t. stelzer , jhep * 0302 * ( 2003 ) 027 [ hep - ph/0208156 ] . t. stelzer and w. f. long , comput . * 81 * ( 1994 ) 357 [ hep - ph/9401258 ] ; + h. murayama , i. watanabe and k. hagiwara , kek-91 - 11 . c. g. papadopoulos , comput . commun . * 137 * ( 2001 ) 247 [ hep - ph/0007335 ] . a. kanaki and c. g. papadopoulos , comput . commun . * 132 * ( 2000 ) 306 [ hep - ph/0002082 ] . t. gleisberg , f. krauss , c. g. papadopoulos , a. schaelicke and s. schumann , eur . j. c * 34 * , 173 ( 2004 ) [ hep - ph/0311273 ] . w. kilian , lc - tool-2001 - 039 . a. pukhov _ et al . _ , hep - ph/9908288 . m. moretti , t. ohl and j. reuter , lc - tool-2001 - 040 , hep - ph/0102195 . mangano , m. moretti , f. piccinini , r. pittau and a.d . polosa , jhep * 0307 * , 001 ( 2003 ) [ hep - ph/0206293 ] ; + m.l . mangano , m. moretti and r. pittau , nucl . b * 632 * ( 2002 ) 343 ; + f. caravaglios , m.l . mangano , m. moretti and r. pittau , nucl . b * 539 * , 215 ( 1999 ) ; + f. caravaglios and m. moretti , phys . b * 358 * ( 1995 ) 332 . p. golonka , t. pierzchala and z. was , comput . commun . * 157 * , 39 ( 2004 ) [ hep - ph/0210252 ] . m. t. ronan , lc - tool-2003 - 015 [ physics/0306019 ] . g. montagna , m. moretti , o. nicrosini and f. piccinini , eur . j. c * 2 * ( 1998 ) 483 [ hep - ph/9705333 ] ; + f. gangemi _ et al . _ , eur . j. c * 9 * ( 1999 ) 31 [ hep - ph/9811437 ] and nucl . b * 559 * ( 1999 ) 3 [ hep - ph/9905271 ] ; + f. gangemi , hep - ph/0002142 . k. koodziej , eur . j. c * 23 * ( 2002 ) 471 [ hep - ph/0110063 ] and comput . phys . commun . * 151 * ( 2003 ) 339 [ hep - ph/0210199 ] ; + a. biernacik and k. koodziej , nucl . * 116 * ( 2003 ) 33 [ hep - ph/0210405 ] . + a. biernacik , k. kolodziej , a. lorca and t. riemann , acta phys . b * 34 * ( 2003 ) 5487 [ hep - ph/0311097 ] . r. harlander and m. steinhauser , prog . * 43 * ( 1999 ) 167 [ hep - ph/9812357 ] . j. kblbeck , m. bhm and a. denner , comput . commun . * 60 * ( 1990 ) 165 ; + t. hahn , comput . commun . * 140 * ( 2001 ) 418 [ hep - ph/0012260 ] . t. hahn and m. perez - victoria , comput . commun . * 118 * ( 1999 ) 153 [ hep - ph/9807565 ] ; + t. hahn , nucl . * 89 * ( 2000 ) 231 [ hep - ph/0005029 ] . j. fleischer , t. hahn , w. hollik , t. riemann , c. schappacher and a. werthenbach , lc - th-2002 - 002 [ hep - ph/0202109 ] . j. fleischer , a. leike , t. riemann and a. werthenbach , eur . j. c * 31 * ( 2003 ) 37 [ hep - ph/0302259 ] g. passarino and s. uccirati , nucl . b * 629 * ( 2002 ) 97 [ hep - ph/0112004 ] . a. ferroglia , m. passera , g. passarino and s. uccirati , nucl . b * 650 * ( 2003 ) 162 [ hep - ph/0209219 ] . a. ferroglia , g. passarino , s. uccirati and m. passera , nucl . instrum . a * 502 * ( 2003 ) 391 . a. ferroglia , m. passera , g. passarino and s. uccirati , nucl . b * 680 * ( 2004 ) 199 [ hep - ph/0311186 ] . f. v. tkachov , nucl . instrum . a * 389 * ( 1997 ) 309 [ hep - ph/9609429 ] ; + l. n. bernstein , functional analysis and its applications , * 6 * ( 1972 ) 66 . g. t hooft and m. veltman , nucl . b * 44 * ( 1972 ) 189 ; + f. v. tkachov , new methods for evaluation of multi - loop feynman diagrams , phd thesis , inr , moscow , march 1984 ; + f. v. tkachov , phys . b * 100 * ( 1981 ) 65 ; + k. g. chetyrkin and f. v. tkachov , nucl . b * 192 * ( 1981 ) 159 . e. maina , s. moretti , m. r. nolten and d. a. ross , hep - ph/0401093 . e. maina , s. moretti and d. a. ross , hep - ph/0403050 . e. maina , s. moretti , m. r. nolten and d. a. ross , hep - ph/0407150 . e. maina , s. moretti , m. r. nolten and d. a. ross , phys . b * 570 * ( 2003 ) 205 [ hep - ph/0307021 ] . e. maina , s. moretti and d. a. ross , jhep * 0304 * ( 2003 ) 056 [ hep - ph/0210015 ] and hep - ph/0403269 . o. brein , u. baur , w. hollik , c. schappacher , d. wackeroth , phys . d * 65 * ( 2002 ) 033007 [ hep - ph/0108274 ] . u. baur and d. wackeroth , hep - ph/0405191 . c. m. carloni calame , s. jadach , g. montagna , o. nicrosini and w. placzek , acta phys . b * 35 * , 1643 ( 2004 ) [ hep - ph/0402235 ] . w. hollik and c. meier , phys . b * 590 * ( 2004 ) 69 [ hep - ph/0402281 ] .

we present a theoretical investigation of the electronic and optical properties of v- and t - shaped quantum wires . valence band mixing as well as realistic sample geometries are fully included through an accurate and efficient approach that is described here in detail . we investigate the resulting valence band structure , which shows some significant peculiarities , such as an anomalously large spin splitting in the lowest heavy hole subband of t - shaped wires . for both classes of wires we obtain good agreement between calculated optical absorption and recent experimental spectra , and we demonstrate that the analysis of optical anisotropy can be used as an effective tool to extract information on valence states , usually very difficult to obtain otherwise .

cond-mat9612251

in recent years , one - dimensional ( 1d ) semiconductor nanostructures have received increasing attention . the potential technological application of quantum wires ( qwrs ) , e.g. in laser devices , has fueled a search for new fabrication techniques and improved sample quality . @xcite in this area , recent investigations have focused on two classes of structures , the so - called v - shaped @xcite and t - shaped @xcite qwrs ( v - qwrs and t - qwrs ) . due to high control on growth conditions and strong confinement of the electron and hole wavefunctions on the scale of few nanometers , v - qwrs and t - qwrs share desirable optical properties for device applications , such as large exciton binding energy and small linewidth . v - qwrs are obtained from a gaas substrate grown along the [ 001 ] crystallographic direction , patterned with [ @xmath010]-oriented v - shaped grooves obtained by chemical etching . the active region consists of a gaas layer cladded between two al@xmath1ga@xmath2as regions @xcite or gaas / alas superlattices @xcite ( sls ) overgrown on the patterned substrate . the confining potential ( see fig . [ fig : geom ] ) has a crescent shape profile . t - qwrs are obtained by first growing a gaas / al@xmath1ga@xmath2as sl on a ( 001 ) substrate . after cleavage , a gaas quantum well ( qw ) is grown over the exposed ( 110 ) surface , resulting in a t - shaped active region . @xcite in both cases , the electron and hole wavefunctions are confined in the [ 001 ] and [ 110 ] crystallographic directions , while the qwr free axis is parallel to the [ @xmath010 ] direction . the optical spectroscopy of qwrs is more complex than for qws of similar lateral dimension , since in qwrs linewidths can be comparable to intersubband splittings . on the other hand , a remarkable peculiarity of qwrs with respect to qws is that the optical activity is strongly anisotropic when light is linearly polarized , with the electric field directed parallel or perpendicular to the wire axis . this has long been recognized to be a band structure effect due to the quasi-1d character of electronic state , combined with heavy- and light - hole ( hh and lh ) mixing . @xcite the anisotropic absorption is therefore used as a simple tool to reveal the 1d character of electronic states in nanostructured materials . in principle , the optical anisotropy can be exploited to single out detailed information on the electronic states , since , as we will show , it is very sensitive to specific details of the band structure . in practice , this approach has been sofar limited by the lack of realistic calculations for complex geometries , as the present v - qwrs and t - qwrs . indeed , common theoretical methods , even within semi - empirical schemes as the tight - binding or the envelope function approach , require a large scale computational effort . in order to keep calculations tractable , up to now the optical properties of qwrs have been investigated theoretically only for rather idealized structures @xcite , yielding results that can not be directly compared with experimental spectra . calculations have been performed for realistic qwr geometries @xcite , but they have sofar neglected hh - lh mixing . recently , we have demonstrated , by a combined theoretical - experimental study of v - qwrs , @xcite that accurate band structure calculations for realistic structures provide quantitative predictions of photoluminescence excitation ( ple ) spectra , and that detailed information on the valence band states can be singled out of the ple anisotropy , despite the dominant role of the light conduction electrons in the optical spectra . such calculations were based on a recently devised method which provides the band structure for qwrs of arbitrary geometry at a relatively small computational cost . the accuracy and the short computer times make such calculations a practical characterization tool in conjunction with experimental results , as well as a predictive tool for new devices . in this paper , we present a theoretical investigation of the electronic and optical properties of v - qwrs and t - qwrs . we focus on the relationship between optical anisotropy and band structure , and we show how the analysis of optical anisotropy permits a detailed spectroscopy of valence states , even when the large linewidth of the spectra does not allow an identification of the valence - to - conduction subband transitions . for t - qwrs we also predict a huge spin - splitting of the lowest valence subband , originating from the interaction between the lowest hh levels of the intersecting qws . for both classes of wires , calculations are performed by the numerical method introduced in ref . here described in detail , thereby demonstrating its accuracy and flexibility . the main approximation that is still present in our approach is the neglect of excitonic effects . indeed , recent calculations which fully include coulomb interaction in realistic qwr profiles @xcite ( but do not include hh - lh mixing ) demonstrate that electron - hole interaction , besides giving rise to bound excitonic states below the band edge , also modifies the excitonic continuum above the band edge . however , our previous investigation in v - qwrs @xcite has shown that quantitative agreement is obtained between the observed ple anisotropy , which probe the excitonic continuum , and the anisotropy in absorption spectra calculated neglecting excitonic effects . we interpret this result as an indication that the electron - hole coulomb interaction , by mixing isotropically the ( optically anisotropic ) electron and hole states , does not change the average anisotropy as obtained by single particle band structure calculations . this is of course compatible with the possibility that the relative intensity of the absorption peaks for a given polarization may be strongly affected . @xcite the above arguments suggest that the approximation of neglecting electron - hole coulomb coupling is a reasonable one for our purpose of studying optical anisotropies . of course , its accuracy for the present v- and t - qwrs must be established a posteriori by comparison with experiments , as we will do later on in this paper . the theoretical background and the numerical method used in our calculations are outlined in sec . [ sec : metodo ] . sections [ sec : v ] and [ sec : t ] report the results of our calculations , focusing on the band structure and optical anisotropy , for v - qwrs and t - qwrs , respectively . in this section we describe the theoretical framework of our calculations . as we are interested in optical transitions near the direct band gap of gaas , and because qwr electronic states are extended over several nanometers , we work within the envelope function approximation . for the wires of interest in this paper , we choose the following cartesian reference frame ( see also fig . [ fig : geom ] ) : @xmath3 along the [ 110 ] direction , @xmath4 along the [ 001 ] direction , and @xmath5 along the [ @xmath610 ] direction . therefore , for both classes of wires , the qwr section extends in the @xmath7 plane , while its free axis is parallel to the @xmath5-direction . due to translational invariance , along this direction it is possible to define a 1d wavevector @xmath8 . in the @xmath3 and @xmath4 directions we assume a supercell periodicity , i.e. we describe arrays of qwrs . the size of the supercell can be taken large enough to describe effectively isolated wires when needed . electron and hole states will be described separately by different effective mass equations . for conduction electrons , we assume a single band approximation , which implies a parabolic energy dispersion in the free direction ; the wavefunctions of the electron subbands are @xmath9 where @xmath10 is the atomic @xmath11 state with spin @xmath12 , @xmath13 is the subband index , and @xmath14 is the @xmath13-th solution of the envelope function equation . since we deal with a two - dimensional confinement potential @xmath15 , we can factorize @xmath16 , where @xmath17 is the @xmath13-th solution of @xmath18 with the electron effective mass hamiltonian @xmath19;\ ] ] here @xmath20 is the electron effective mass , and @xmath21 , @xmath22 . of course , in this one - band description , the electron subbands @xmath23 depend quadratically on @xmath8 : @xmath24 where the @xmath25 s are the confinement energies ( i.e. , the subband edges ) . the subbands @xmath26 are spin - degenerate , and the @xmath17 s are @xmath8-independent . conduction band states do not directly contribute to optical anisotropy , being mainly of isotropic @xmath11-type character , and in this case the above one - band description is sufficient . conversely , it is essential to give accurate account of @xmath27-type valence states by a multi - band description , in order to investigate optical anisotropy . due to mixing between hh and lh subbands , both eigenvalues and envelope functions depend non - trivially on the wavevector @xmath8 and the hole subband index @xmath28 . using the compact notation @xmath29 , we write the hole wavefunctions @xmath30 where @xmath31 are the four atomic states with total angular momentum @xmath32 and projection @xmath33 of @xmath34 along a quantization axis . again , each envelope function @xmath35 can be factorized as @xmath36 , and the four - component vector @xmath37 is the @xmath28-th solution at point @xmath8 of the multi - band effective mass equation @xcite @xmath38 where @xmath39 is the luttinger hamiltonian . @xcite with the above choice of coordinate axes and the quantization axis of @xmath40 along the [ 110 ] direction , following ref . the luttinger hamiltonian reads @xmath41 where @xmath42 , \\ \hat{q } & = & -2\sqrt{3}\left(\gamma_3 \hat{k}_y - i\gamma_2 k_z\right ) \hat{k}_x .\end{aligned}\ ] ] this hamiltonian provides the ( positive definite ) hole subbands , referred to the bulk valence band edge , as a function of the in - wire wavevector @xmath8 , including hh - lh mixing . the hole hamiltonian for structures grown along crystallographic directions different from the present ones can be obtained along the lines of ref . . in the above electron and hole effective mass hamiltonians we neglect the material dependence of the electron effective mass @xmath20 and the luttinger parameters @xmath43 , @xmath44 , and @xmath45 , and we always use the bulk gaas values listed in table [ tab : parameters ] . although our approach can be extended to account for the material dependence , this would be a small effect in our calculations . one possible approach to solve the multi - band equation ( [ eq : multiband ] ) is to split the problem in two parts : the first step is to solve the two schrdinger - like equations arising from the diagonal terms of ( [ eq : h_l ] ) , @xmath46\phi^1_i(x , y ) = \epsilon^1_i\phi^1_i(x , y ) , \\ \left[\hat{p}_2+v(x , y)\right]\phi^2_i(x , y ) = \epsilon^2_i\phi^2_i(x , y ) . \end{aligned}\ ] ] the second step is to use the set @xmath47 as a basis to expand the components @xmath48 of the vector @xmath49 , and the set @xmath50 as a basis for the components @xmath51 . in this representation , the diagonal matrix elements of the @xmath52 are given by the two sets of scalar numbers @xmath53 , @xmath54 , and matrix elements need only to be calculated for off - diagonal terms , using the functions @xmath47 , @xmath50 . this approach has the following drawback : since the effective masses appearing in @xmath55 , @xmath56 ( the so - called hh and lh effective masses ) are very different , the two sets of eigenvalues @xmath53 , @xmath54 span different energy ranges . the set @xmath54 , being the spectrum of a light particle , will have larger gaps , and the ground state will be higher in energy than for the set @xmath53 . since we are mainly interested in the low - lying hole subbands near @xmath57 , this representation , although exact in principle , is poorly convergent with respect to the number of states @xmath47 , @xmath50 included in the basis , and it is not practical in numerical calculations . to solve this problem , we propose a scheme in which we expand the components of @xmath58 using the solutions of two schrdinger - like equations with two fictitious , arbitrary masses @xmath59 , @xmath60 , which then we tune in order to improve the convergence : @xmath61 \phi^+_\nu ( x , y ) & = & \epsilon^+_\nu\phi^+_\nu ( x , y ) , \label{eq : fict - f}\\ \left[\frac{\hbar^2}{2m^-}\left(\hat{k}_x^2+\hat{k}_y^2\right)+v(x , y)\right ] \phi^-_\mu ( x , y ) & = & \epsilon^-_\mu\phi^-_\mu ( x , y ) \label{eq : fict - l}\end{aligned}\ ] ] ( here and in the following we use the index @xmath62 for @xmath63 states , and @xmath64 for @xmath65 states ) . in this way , we diagonalize exactly only _ part _ of the kinetic energy terms @xmath55 and @xmath56 of @xmath39 , but the potential @xmath15 is exactly diagonalized . of course , in this representation @xmath66 , @xmath67 are _ not _ the diagonal elements of the matrix representing the hamiltonian , and @xmath55 , @xmath56 give rise to additional off - diagonal terms . however , the time spent to calculate the additional terms is more than compensated by the improved convergence which can be achieved by properly choosing @xmath59 and @xmath60 . in the end , we shall find it convenient to choose @xmath68 , and equal to the heavy hole effective mass along the [ 001 ] direction . to implement this idea , we add and subtract a term @xmath69 to @xmath55 , and a term @xmath70 to @xmath56 . then we obtain : @xmath71 where @xmath72 then we solve the equations ( [ eq : fict - f])-([eq : fict - l ] ) by a plane - wave expansion , as outlined in ref . . typically , we fix two energy cutoffs , @xmath73 and @xmath74 , and we find the @xmath75 and @xmath76 eigenstates which fall below the cutoffs . using the eigenfunctions @xmath77 , @xmath78 , we form the following basis set : @xmath79 with @xmath80 and @xmath81 , and we expand @xmath49 in this basis : @xmath82 the explicit matrix elements of @xmath39 in this basis are given in appendix [ sec : app - a ] . the total dimension of the hamiltonian matrix in this representation is @xmath83 . all we need to compute , in order to evaluate the matrix elements , are integrals of the kind @xmath84^ * & \hat{k}_\beta & \phi^{\gamma^\prime}_j(x , y ) \,dx\,dy , \\ \int_\omega \left[\phi^{\gamma}_i(x , y)\right]^ * & \hat{k}_\beta & \hat{k}_{\beta^\prime } \phi^{\gamma^\prime}_j(x , y ) \,dx\,dy,\end{aligned}\ ] ] evaluated over the supercell volume @xmath85 , where @xmath86 and @xmath87 ; these are easily obtained given the plane - wave expansions of the @xmath88 s and @xmath89 s . the choice of @xmath59 , @xmath60 , which is important in order to obtain an efficient convergence , is discussed in detail in appendix [ sec : app - b ] ; here we only anticipate that in all our calculations we use @xmath90 . once we have calculated the electron and hole subbands by the above method , we are in the position to evaluate the absorption spectrum @xmath91 in the dipole approximation , summing the dipole matrix element , with the appropriate polarization of light @xmath92 , over all electron and hole states : @xmath93 where the optical matrix elements @xmath94 are given in appendix [ sec : app - c ] , and @xmath95 is the bulk energy gap of gaas ; tipically , a set of sixty @xmath8 points have been included in the summation . the absorption spectra shown in this paper have been obtained by superimposing a gaussian broadening @xmath96 to @xmath91 , in order to simulate the inhomogenous broadening due to structural imperfections of the samples . the broadened spectrum is obtained as @xmath97 as a prototype of v - qwrs , we first consider a sample described in refs . . this consists of an active gaas layer embedded in a ( alas)@xmath98/(gaas)@xmath99 sl , overgrown by molecular beam epitaxy on the exposed surface of the etched substrate . as in refs . , we use the v - shaped potential profile obtained by digitalizing a tem micrograph of the sample . we also adopt the same supercell geometry as in refs . . the complicated structure of the sl which provides the quasi-1d confinement is modelled by a homogeneous barrier with effective conduction and valence band offsets , @xmath100 and @xmath101 , respectively . based on previous investigations @xcite for the same sample , we take @xmath102 and @xmath103 . it should be noted that the effort of including exactly the confining sl in the calculations would not necessarily result in improved accuracy , as the envelope function approximation itself looses its validity for such short - period sls . in order to investigate the role of the confinement in the optical properties , we shall consider two sample profiles , which differ in the value of the confinement length @xmath104 at the bottom of the v - shaped region ( see fig . [ fig : geom ] ) : profile a ( @xmath105 ) and profile b ( @xmath106 ) . these sl - embedded qwrs will be labelled a / sl and b / sl . a key issue which make nm - scale qwrs interesting for electro - optical applications are large confinement energies which can be obtained with large band - offsets in addition to geometric confinement . in view of this fact , we shall compare the samples described above with samples having the same profiles and barriers constituted by pure alas , that we will label a / alas and b / alas . the parameters of the four samples are summarized in table [ tab : samples ] . a qualitative interpretation of the band structure of a v - qwr can be obtained by adding to a qw of width @xmath104 an additional lateral confinement due to the crescent shape of the profile . in the lowest approximation , the latter can be thought of as a parabolic potential @xcite which splits each subband of the parent qw into a new set of subbands . since the additional confinement is less effective than the confinement due to the original qw , the new sets of subbands have smaller gaps with respect to the subband splittings of the parent qw . this simplified picture will serve as a guideline for the discussion of numerical results obtained for the actual samples . the calculated energies at @xmath107 of the lowest conduction and valence states are listed for reference in table [ tab : v - edges ] for the four samples . in the following we shall focus on hole subbands , which are shown in fig . [ fig : v - bande ] ( right panel ) for sample a / sl . first , we note that at @xmath107 each subband is doubly degenerate , while at finite @xmath8 the subbands are spin - split , due to the lack of inversion symmetry of the confining potential . @xcite splittings are in the range of few mev . secondly , a strongly non - parabolic energy dispersion is evident . this fact , familiar from qws , is due to mixing of states with hh ( @xmath108 ) character and lh ( @xmath109 ) character . the hh / lh character of hole states influences the optical properties of the sample , since different atomic orbital components have different oscillator strenghts ; of particular interest from this point of view , are the @xmath107 states which , due to the large density of states ( dos ) stemming from their quasi-1d character , mainly contribute to the absorption intensity . contrary to the case of qws , hole subbands can not be strictly classified as hh and lh states even at @xmath107 . to define the hh / lh character in the present samples , we note that the direction of strongest confinement is the [ 001 ] direction , as demonstrated by charge density maps of the lowest lying states which we will show later in fig . [ fig : v - chd ] . it is therefore meaningful to calculate the hh / lh character along this direction , because this would be the quantization axis of @xmath34 in an equivalent [ 001]-grown qw of width @xmath104 . to do this , we calculate the rotated vector @xmath110 , where @xmath111 @xmath112 is obtained by diagonalizing the matrix @xmath113 written in the representation in which @xmath114 is diagonal , with eigenvalues @xmath115 . then we define the hh- and lh - projected charge densities @xmath116 finally , the hh and lh character is obtained by integrating the above charge densities over all space . ( in the above equations , the real - space representation is chosen for clarity . the corresponding expressions in fourier space , which we use in the numerical implementation , are very easy to obtain and are not explicitated here . ) in the left panel of fig . [ fig : v - bande ] we show the calculated lh character of the hole subbands at @xmath117 . this is best analized in connection with fig . [ fig : v - chd ] , where we show the total and projected charge densities of the hole states at @xmath107 for the same a / sl sample . in the ground state , the lh component is rather small ( @xmath118 ) , but it increases rapidly for the excited subbands . correspondingly , the ground state is well localized [ fig . [ fig : v - chd](a ) ] , while the lowest excited states [ figs . [ fig : v - chd](b ) , [ fig : v - chd](c ) ] have wavefunctions which increasingly extend along the v - qwr sidewalls . the regular increase of the lh component is interrupted by the @xmath119 levels ( 13th doublet : recall that each point at @xmath107 is doubly degenerate ) at @xmath120 which is mainly of lh character ( @xmath121 ) . this can be interpreted as the reminescent of the lh state of a qw with confinement length @xmath104 . the wavefunction of this level [ fig . [ fig : v - chd](d ) ] , in fact , is again well localized , analogously to the ground state , and in contrast to the wavefunctions of nearby states . we shall comment later on the fingerprints of this strongly lh - like state in the optical spectra , and , particularly , in the optical anisotropy . note also that in fig . [ fig : v - bande ] there are other `` jumps '' in the lh character at higher energies ; these correspond to energies where ladders of levels of different simmetries , like an additional nodal plane parallel to the [ 110 ] direction , begin . in fig . [ fig : sl - vs - bk ] we compare the lh character vs subband energy at @xmath107 for samples with different barriers ( sl and pure alas ) . for both profiles a and b , the strong confinement due to alas barriers induces not only the expected blue - shift of the subbands , but also reduces the lh character of the lowest lh - like state , as compared to sl barriers : in sample a / sl , the @xmath119 doublet is 56% lh , while its counterpart in sample a / alas , the @xmath122 doublet , is only 42% lh . in sample b / sl , the @xmath123 doublet is 75% lh , while its counterpart in sample b / alas , the @xmath124 doublet , is only 52% lh . in fig . [ fig : v - abs ] we show the calculated absorption intensity for the four samples of table [ tab : samples ] and for light linearly polarized parallel to the wire axis , @xmath125 , and perpendicular to it along the [ 110 ] direction , @xmath126 . a gaussian broadening of @xmath127 has been included . @xcite for all samples , we also report the relative optical anisotropy ( thick lines in fig . [ fig : v - abs ] ) , defined as @xmath128 . the optical absorption spectra are obtained integrating over the whole band structure ( i.e. , integrating over @xmath8 and summing over electron and hole subbands and spin ) ; however , for illustration we show the strongest optical transitions due to @xmath107 states , and for one spin orientation of the conduction electrons , for light polarization perpendicular to the qwr axis ( histograms in fig . [ fig : v - abs ] ) . to identify the electron and hole states involved in each transition , we use the label @xmath129 , where @xmath13 and @xmath28 are , respectively , the indices of the conduction and valence states . we first focus on sample a / sl , for which experimental data are available . @xcite the calculated anisotropy in the low energy part ( @xmath130 ) is @xmath131 , with a deep minimum at @xmath132 , where the anisotropy is almost suppressed . both the average anisotropy and the position of the minimum are in _ quantitative _ agreement with experimental data . @xcite the agreement worsens at higher energies ( @xmath133 ) , where the calculated anisotropy drops rapidly and finally changes sign , while experimental data @xcite show an increase . we believe that this discrepancy is due to our `` effective '' description of the barriers which affects particularly the higher - lying hole states . by comparing the calculated anisotropy of the four samples , it appears that a more or less pronounced dip over a range of @xmath134 ( i.e. , the linewidth of the broadened spectra ) is always present in the low energy range of the spectrum , superimposed onto a background of an otherwise large and positive anisotropy . additional structure , particularly for sample a / alas , is also present in the high energy range . note also that the maximum anisotropy is in the range @xmath135 , and does not change dramatically in the different samples . by studying the optical matrix elements , it can be shown that the dips in the anisotropy are due to states with a large lh character , @xcite e.g. , the @xmath136 subband in sample a / sl . since this is a localized state and has a large spatial overlap with the first conduction subband , it contributes to the low energy part of the spectrum . this is true in general : as can be seen from fig . [ fig : v - abs ] , for all samples the anisotropy dip corresponds to a @xmath107 transition ( highlighted as black bars in the histograms ) between the electron ground state ( @xmath137 ) and an excited hole state which is the @xmath136 , the @xmath138 , the @xmath139 , and the @xmath140 level for a / sl , a / alas , b / sl , and b / alas , respectively . for sample a / sl , comparison with fig . [ fig : v - bande ] shows that the involved @xmath136 hole state has a strong lh character and its wavefunction ( fig . [ fig : v - chd](d ) ) is strongly localized . the large lh component makes that the intensity for the two polarizations is reversed with respect to the strongly hh - like ground state , causing the dip in the anisotropy at @xmath141 . a similar correspondence between the dip and a localized lh - like state applies also to the other samples . an immediate consequence of the above result is that , since both the ground hh state and the lh - like state couple with the lowest electron subband , _ the difference in energy between the onset of the continuum and the position of the dip in the anisotropy is a direct measure of the energy splitting between the ground hh and the first lh state , independently of the electron confinement_. note that such information can not be extracted from the absorption spectra alone , as the large broadening prevents the identification of any single transition apart from the fundamental one . the hh / lh splitting obtained in the above manner can be used to extract band structure parameters as , for example , the effective hole confinement @xmath101 . in the measured ple spectra for sample a / sl , the anisotropy dip lies @xmath142 above the onset of the continuum ; in fig . [ fig : e - hh - lh ] we report the hh / lh splitting calculated for several values of @xmath101 ( full dots ) for this sample . we also show , for comparison , the hh / lh splitting calculated by a simple square well model for a qw of width @xmath104 ( empty dots ) , using the bulk hh and lh effective mass along [ 001 ] . it can be seen that the splitting is quite sensitive to @xmath101 , and that the experimental value is compatible with @xmath143 . this observation provides a good criterion for choosing the confinement energy of holes , which would otherwise be rather arbitrary . using this procedure , @xmath101 was finally taken equal to @xmath144 in ref . . note that a variation of the valence band offset of @xmath145 shifts the hh / lh splitting of @xmath146 , well within the experimental accuracy for the determination of the hh / lh splitting . @xcite note also that , for a rough estimation of @xmath101 , calculations using a simple square well model may be sufficient . the comparison with the absorption spectra demonstrates that @xmath107 transitions alone give a poor estimate of the integrated spectrum . this is due to the dos contribution of the hole subbands which are strongly non parabolic , and have a large dos also for @xmath8-points away from @xmath107 ( see fig [ fig : v - bande ] ) . finally , note the relaxation of selection rules for @xmath107 transitions shown in fig . [ fig : v - abs ] . indeed , if the envelope - function parity - conserving selection rule would be obeyed , only transitions of the type @xmath147 or @xmath148 ( depending on electron spin orientation ) would be allowed . we investigate a t - qwr with the geometry sketched in fig . [ fig : geom ] . in our calculation , the sl grown along the [ 001 ] direction is constituted by 5.3 nm wide qws ( qw1 ) , separated from each other by 50 nm wide alas barriers , while the qw grown along the [ 110 ] direction ( qw2 ) is 4.8 nm wide . these parameters correspond to a sample for which polarization - dependent ple spectra are available . @xcite note that the t - qwrs which form at the intersection between qw1 and qw2 are not uncoupled in the [ 001 ] direction , due to the sl structure of qw1s , a fact which we fully take into account in the supercell representation used in our calculations . on the other hand , the t - qwrs are isolated along the [ 110 ] direction ; in our calculations , this is simulated by truncating qw1 with a alas barrier 50 nm on the left hand side . the nature of the 1d confinement in t - qwrs is rather different with respect to the v - qwrs case . there , we used the picture of a lateral confinement added on a qw structure which , therefore , localizes all states in the @xmath7 plane . conversely , t - qwr states can be better interpreted as the result of the coupling between 2d states of the qws , qw1 and qw2 . beside perturbing the 2d states , this coupling also induces localized states or resonances ; these can be also seen as due to a 1d defect in the otherwise translationally invariant ( in the qw plane ) 2d states of a qw , due to the coupling with the other qw . this picture has non trivial consequences , particularly for hole states , which we analyze in the following . first , we focus on the lowest lying states at @xmath107 . electrons and holes are very different from this point of view : contrary to the electron effective mass which is isotropic in gaas , the hh effective mass is strongly anisotropic between the [ 001 ] direction ( @xmath149 ) and the [ 110 ] direction ( @xmath150 ) . the effect of this difference is shown in fig . [ fig : t - chd-1 ] , where we report the total charge densities for the lowest electron and hole states at @xmath107 . for the present structure , being qw1 and qw2 of comparable width , the electron ground state [ fig . [ fig : t - chd-1](a ) ] is a quasi-1d state extending both in qw1 and qw2 ; the lowest excited subbands [ figs . [ fig : t - chd-1](b),(c),(d ) ] belong to the continuum of 2d states of qw1 , since qw1 is wider ( actually , discrete levels are obtained here , due to the supercell method ) ; the localized quasi-1d electron lies 15 mev below the qw1 continuum . on the contrary , for holes the larger mass along [ 110 ] more than compensate for the smaller width of qw2 , and the hole ground state [ fig . [ fig : t - chd-1](a ) ] , although localized in the center of the t , is much more a qw2-like state , only weakly perturbed by coupling to qw1 ; accordingly , the lowest excited hole subbands [ fig . [ fig : t - chd-1](b),(c ) ] are basically 2d states belonging to qw2 , up to the fourth level [ fig . [ fig : t - chd-1](d ) ] , which is a qw1 state . the quasi-1d ground state is practically degenerate with the qw2 continuum ; note however , that , in contrast to conduction electrons , there is a second hole level [ fig . [ fig : t - chd-1](b ) ] with a significant component along qw1 . the full band structure of holes is shown in the right - hand panel of fig . [ fig : t - bande ] . in the rather complex dispersion of the subbands , we can distinguish a peculiar feature of t - qwr , namely a huge spin - splitting of the lowest doublet at finite @xmath8 which , at @xmath151 , is @xmath152 for the present structure . before discussing the origin of such a large spin - splitting , it is useful to examine the actual localization of the calculated states in the qwr . to this aim , we need to discriminate the states peaked around the center of the t from states that are typical of one of the parent qws and are left essentially unchanged by the interaction with the other one . this is particularly useful because we expect that most of the states of fig . [ fig : t - bande ] simply arise from folding of qw states induced by the supercell periodicity and to the truncation of the t along the [ 110 ] direction . therefore , in fig . [ fig : t - bande ] we identify by full circles those states which , by direct inspection of the wavefunction , show a strong localization in the center of the t. indeed , it appears that only a limited subset has predominantly a localized character , and can therefore be assigned to quasi-1d qwr - like states or resonances . these include the localized states arising from the lowest two doublets , already discussed above , as well as the resonant states falling around 40 mev . the remaining states are similar to 2d qw - like states , with charge density localized mostly in qw1 or qw2 . to clarify the origin of the large spin - splitting , it is useful to compare the band structure of the t - qwr with the hole subbands of the parent isolated qw1 and qw2 . in fig . [ fig : t - bande ] we show with open squares and open circles the lowest hh subband of qw1 and qw2 ; the lh subbands , for these thicknesses , lie high in energy ( @xmath153 in qw1 and @xmath154 in qw2 ) . @xcite a small gap of @xmath155 separates the two levels at @xmath107 ; as already noted above , the lowest state is qw2 , due to the larger hh mass . owing to different hh - lh mixing in each qw separately , the hh subbands of the two parent qws have different energy dispersion as a function of @xmath8 and cross at some finite @xmath8 . large spin splittings are in general an effect of hh - lh mixing in asymmetric structures . @xcite in the present case , however , the energy difference between the hh and lh levels in each isolated qw is too large to explain the huge splitting that we have found in the lowest energy subbands of the qwr . naively , one would rather expect small gaps ( induced only indirectly by coupling to the far - lying lh states ) to open at the crossing of the qw1 and qw2 subbands . note , however , that the hh states of the parent qws are eigenstates of @xmath40 with @xmath108 , but _ with the quantization axis along different directions_. taking , for example , the @xmath40 quantization axis along [ 110 ] , a hh state of qw2 is @xmath156 ( for one of the degenerate spin orientations ) . a hh state of qw1 _ written in the same basis _ , is @xmath157 in words , a hh state of qw1 has a strong lh component from the point of view of qw2 . therefore , the lowest hh subbands arising from qw1 and qw2 are strongly coupled , which results in a strong avoided - crossing behaviour and a very large spin - splitting . this interpretation is supported by our calculation of the lh character , which is shown in the left panel of fig . [ fig : t - bande ] . in view of the fact that the hole ground state is rather qw2-like , we have computed the lh character with the quantization axis along [ 110 ] , i.e. , it is obtained , as in sec . [ sec : v ] , by integrating the projected charge densities [ see eqs . ( [ eq : rho - hh])-([eq : rho - lh ] ) ] , but using the @xmath49 instead of the rotated @xmath158 . as expected , the lowest subbands are nearly pure hh , being localized in qw2 ( see fig . [ fig : t - chd-1 ] ) , while the @xmath159 subbands are more than 60% lh . this should not be interpreted as a mixing with lh states of qw2 ; rather , it is a manifestation of the fact that this state is localized in qw1 and , therefore , does not have a well defined orbital character along [ 110 ] . we conclude our analysis of the lowest spin - split doublet by noting that , as @xmath8 increases , only the lowest level remains a well localized , quasi-1d state , while its spin companion gradually merges into the qw1 quasi - continuum ( at large wavevectors , qw1 becomes the ground state ) . for illustration , we show in fig . [ fig : t - chd-2 ] the charge density of the two lowest levels at @xmath160 ( close to the wavevector where the qw2 and qw1 dispersions cross ) . the lowest state ( labelled @xmath161 ) , which falls far in energy from the qw bands , is strongly localized at the intersection of the qws . the higher state ( labelled @xmath161 ) is peaked at the center of the t but it also extends quite far into qw1 . in fig . [ fig : t - abs ] we show the calculated absorption intensity for the t - qwr and for light linearly polarized parallel to the wire axis , @xmath125 , and perpendicular to it along the [ 001 ] direction , @xmath126 . a gaussian broadening of @xmath162 has been included . we can identify four main structures , a peak around 1.65 ev , another peak around 1.68 ev ( with a minor shoulder on the low - energy side ) , a large composite structure beginning above 1.7 ev , and another large structure at 1.74 ev . of these , only the lowest peak involves essentially pure qwr - like localized states [ fig . [ fig : t - chd-1](a ) ] . the second structure involves transitions from the next electron state , extending in qw1 [ see fig . [ fig : t - chd-1](b ) ] , to the first hole states with significant spatial overlap [ see fig . [ fig : t - chd-1](d ) ] . the structure above 1.7 mev involves predominantly qw2 states , with a contributions from higher qwr resonances which produce the high energy shoulder in @xmath125 . finally , the large peak at 1.74 ev is due to many subbands , with a significant contribution from lh states belonging to both qws . the lowest three structures are polarization dependent with @xmath163 . the anisotropy is maximum for the second structure , for which @xmath126 is very small , consistently with the expectation for a ( 001 ) qw . the higher structure , on the other hand , is nearly polarization independent , due to the lh contributions . these results can be compared with the experimental ple spectra of ref . . there , three main structures occur at 1.64 , around 1.671.68 , and above 1.7 mev . the agreement with the calculated spectra is again surprisingly good , taking into account that these neglect excitonic effects . the experimental assignment of the first structure to a qwr - like state and the successive structures to qw1-like states , based on the comparison with reference qws , is fully consistent with our picture . as concerns intensities , the agreement is also reasonably good if one considers that the weight of the qwr peak is sensitive to the relative volume occupied by the qws , which enters the calculation through the choice of the supercell . on the other hand , the reasons of an enhanced intensity of qw1-like features in the ple experimental data are discussed in ref . . finally we compare our results with the observed values of the anisotropy @xcite . from ple , akiyama estimate @xmath164 for the qwr peak , and @xmath165 for the next structure assigned to qw1 . the corresponding theoretical values from fig . [ fig : t - abs ] are approximately @xmath166 and @xmath167 ( the coefficients are slightly larger , 0.57 and 0.17 respectively , if the broadening is reduced to @xmath168 ) . the very different anisotropy of the two structures is therefore in qualitative agreement with experiments , although further investigation would be required to understand the origin of the difference . we have presented an accurate and efficient approach that allows us to calculate the electronic and optical properties of quantum wires , taking into account valence band mixing effects together with realistic profiles of the confining potentials . we have studied specifically v- and t - shaped quantum wires , where the shape of the confinement region differs considerably from the model geometries assumed in most of the previous investigations . the two classes of wires differ significantly in the structure of their energy spectra : while the crescent shape of v - qwrs induces a series of localized quasi-1d levels , only the lowest states of t - qwrs are clearly localized in the wire because of the subsequent onset of the continua of the parent quantum wells . the consequences on the optical spectra have been discussed in detail . in particular , we have focused on the optical anisotropy , and demonstrated that the analysis of anisotropy spectra can be used as an effective tool to extract information on valence states , usually very difficult to obtain otherwise . comparison with very recent ple spectra for both classes of wires shows good agreement , in spite of our neglecting excitonic effects . as we discussed in the introduction , this agreement might be due to the symmetry properties of the coulomb interaction and , therefore , might be a rather genereal feature . however , while the approximation of neglecting excitonic effects is very convenient from the computational point of view , its accuracy for a given class of materials should be established a posteriori from comparison to experiments , as we have positively tested in this paper for v - qwrs and t - qwrs . as a final remark , we stress that all our calculations have been performed by a numerical method which prooved computationally very convenient . furthermore , our method lends itself to include calculations of coulomb correlation effects on the linear and non linear optical properties of these wires which are currently implemented only for non - interacting valence bands,@xcite as well as to include external magnetic fields @xcite to interpret magneto - luminescence experiments . this work was supported in part by the ec commission through the hcm network `` ultrafast '' . the fictitious masses @xmath59 , @xmath60 entering equations ( [ eq : fict - f])-([eq : fict - l ] ) can be chosen arbitrarily . in this appendix we will show that a judicious choice can lead to a significant improvement in the basis set convergence and lowering of the computational cost of the calculation . note that the convergence with respect to the number of functions @xmath171 , @xmath172 included in the basis set , which we investigate below , is a separate problem from the convergence in the plane - wave expansion of the @xmath171 , @xmath172 themselves which , for a given structure , must therefore be checked once and for all _ before _ the diagonalization of the luttinger hamiltonian is started . in fig . [ fig : conv ] we show the behaviour of the lowest hole eigenvalues as a function of the number of basis functions @xmath173 . the calculations have been performed for the v - qwr a / sl , which is described in detail in section [ sec : v ] . the eigenvalues reported in fig . [ fig : conv ] are highlighted by black triangles in the full band structure of the same v - qwr shown in fig . [ fig : v - bande ] : they are the doubly degenerate lowest eigenvalue at @xmath174 [ panel ( a ) ] , and the two spin - split lowest eigenvalues at @xmath175 [ panels ( b ) and ( c ) ] . in each panel of fig . [ fig : conv ] we show two sets of calculations , both obtained with @xmath176 , but with different choices of @xmath60 . the empty dots are obtained with @xmath177 . with this choice , @xmath59 and @xmath60 are the hh and lh effective masses along the [ 001 ] crystallographic direction ; since this is the direction of strongest confinement for these v - qwrs ( see next section ) , these are the `` physical '' masses in the sense that , for example , they would determine the hh and lh levels of a qw with comparable confinement length grown in this direction . it can be noted , however , that convergence is achieved only within @xmath178 with @xmath179 as high as 280 , which corresponds to @xmath180 and @xmath181 ; these , in turn , should be compared with the low barrier height of this sample , which is @xmath144 . therefore , one needs to reach energies high in the continuum to achieve convergence . the couples of empty dots at @xmath182 are obtained with two different choices of the pair @xmath183 : @xmath184 for the upper points , and @xmath185 for the lower points . we conclude that , with fixed @xmath186 , the convergence improves with increasing @xmath76 , suggesting that the @xmath65 states are responsible for the slow convergence . in fact , lowering the @xmath60 mass improves the convergence . in fig . [ fig : conv ] , the black dots correspond to the extremal choice @xmath187 which , as we anticipated in sec . [ sec : metodo ] , is our final choice , and has been used in all calculations presented in this paper . with respect to the previous case , the convergence is much faster : the well - converged values at @xmath188 ( i.e. , @xmath189 ) are obtained with @xmath190 ; indeed , we find that the convergence nearly saturates when @xmath191 , i.e. , just above the valence band offset for this qwr . note that , in addition to the improved convergence , the choice @xmath187 implies that only one of the two equations ( [ eq : fict - f])-([eq : fict - l ] ) need to be solved . as a final remark , we note that the convergence is slower for the eigenvalues at @xmath175 than at @xmath107 , due to the strong hh - lh mixing for large wavevectors . an accurate convergence at these wavevectors , as it can be achieved by our method , is e.g. necessary to calculate in - wire effective mass at the fermi wavevector . the slow convergence of the in - plane effective mass at the fermi edge is a well - known problem in qws . once we have calculated the functions @xmath192 , @xmath193 , we compute the following integrals : @xmath194^ * \phi^-_\mu(x , y ) \,dx\,dy , \\ w_\beta(\nu,\mu ) & = & \int_\omega \left[\phi^+_\nu(x , y)\right]^ * \hat{k}_\beta \phi^-_\mu(x , y ) \,dx\,dy , \\ w_{\beta\beta^\prime}(\nu,\mu ) & = & \int_\omega \left[\phi^+_\nu(x , y)\right]^ * \hat{k}_\beta \hat{k}_{\beta^\prime } \phi^-_\mu(x , y ) \,dx\,dy , \\ v^+_{\beta}(\nu,\nu^\prime ) & = & \int_\omega \left[\phi^+_\nu(x , y)\right]^ * \hat{k}^2_\beta \phi^+_{\nu^\prime}(x , y ) \,dx\,dy , \\ v^-_{\beta}(\mu,\mu^\prime ) & = & \int_\omega \left[\phi^-_\mu(x , y)\right]^ * \hat{k}^2_\beta \phi^-_{\mu^\prime}(x , y ) \,dx\,dy , \end{aligned}\ ] ] where @xmath195 . in our implementation , the functions @xmath77 , @xmath78 are expanded in plane waves . although this is not necessary , it makes very easy to compute the above integrals , where the operators @xmath196 are just substituted by scalar numbers @xmath197 . with the above definitions , and using the short notations @xmath198 , @xmath199 @xmath200 , @xmath201 , and @xmath202 , the only non - zero matrix elements of @xmath39 in the basis set ( [ eq : basis - f])-([eq : basis - l ] ) are @xmath203 \delta_{\nu\nu^\prime } + p^+_x v^+_x + p^+_y v^+_y \langle\left.+,\nu,\downarrow\right|{\bf h}_l\left|+,\nu^\prime,\downarrow\right.\rangle & = & \langle\left.+,\nu,\uparrow\right|{\bf h}_l\left|+,\nu^\prime,\uparrow\right.\rangle \\ \langle\left.-,\mu,\uparrow\right|{\bf h}_l\left|-,\mu^\prime,\uparrow\right.\rangle & = & \left[\epsilon^-_\mu+\left(\gamma_1+\frac{\gamma_2 - 3\gamma_3}{2}\right)k_z^2\right ] \delta_{\mu\mu^\prime } + p^-_x v^-_x + p^-_y v^-_y \\ \langle\left.-,\mu,\downarrow\right|{\bf h}_l\left|-,\mu^\prime,\downarrow\right.\rangle & = & \langle\left.-,\mu,\uparrow\right|{\bf h}_l\left|-,\mu^\prime,\uparrow\right.\rangle \\ \langle\left.+,\nu,\uparrow\right|{\bf h}_l\left|-,\mu,\uparrow\right.\rangle & = & -\frac{\sqrt{3}}{2}\left[2\left(\gamma_2 w_{yy}-2i\gamma_3 k_z w_y\right ) -\left(\gamma_2+\gamma_3\right)k_z^2 s -\left(\gamma_2-\gamma_3\right ) w_{xx } \right ] \\ \langle\left.+,\nu,\downarrow\right|{\bf h}_l\left|-,\mu,\downarrow\right.\rangle & = & -\frac{\sqrt{3}}{2}\left[2\left(\gamma_2 w_{yy}+2i\gamma_3 k_z w_y\right ) -\left(\gamma_2+\gamma_3\right)k_z^2 s -\left(\gamma_2-\gamma_3\right ) w_{xx } \right ] \\ \langle\left.+,\nu,\downarrow\right|{\bf h}_l\left|-,\mu,\uparrow\right.\rangle & = & 2\sqrt{3}\left(\gamma_3 w_{xy}+i\gamma_2 k_z w_x\right ) \\ \langle\left.+,\nu,\uparrow\right|{\bf h}_l\left|-,\mu,\downarrow\right.\rangle & = & -2\sqrt{3}\left(\gamma_3 w_{xy}-i\gamma_2 k_z w_x\right)\end{aligned}\ ] ] we define the electron - hole overlap integrals @xmath204^ * \phi^+_\nu(x , y)\,dx\,dy , \\ t^-(n,\mu ) & = & \int_\omega \left[\psi^e_n(x , y)\right]^ * \phi^-_\mu(x , y)\,dx\,dy . \end{aligned}\ ] ] @xmath207^*\psi^h_{j_m}(x , y)\,dx\,dy \nonumber\\ & = & \frac{p}{\sqrt{2}}\left[\sum_\nu c^+_\alpha(\nu,\uparrow)t^+(n,\nu)- \frac{1}{\sqrt{3}}\sum_\mu c^-_\alpha(\mu,\uparrow)t^-(n,\mu)\right]\end{aligned}\ ] ] @xmath208^*\psi^h_{j_m}(x , y)\,dx\,dy \nonumber\\ & = & \frac{p}{\sqrt{2}}\left[-\sum_\nu c^+_\alpha(\nu,\downarrow)t^+(n,\nu)+ \frac{1}{\sqrt{3}}\sum_\mu c^-_\alpha(\mu,\downarrow)t^-(n,\mu)\right]\end{aligned}\ ] ] @xmath209^*\psi^h_{j_m}(x , y)\,dx\,dy \nonumber\\ & = & \frac{ip}{\sqrt{2}}\left[\sum_\nu c^+_\alpha(\nu,\uparrow)t^+(n,\nu)+ \frac{1}{\sqrt{3}}\sum_\mu c^-_\alpha(\mu,\uparrow)t^-(n,\mu)\right]\end{aligned}\ ] ] @xmath210^*\psi^h_{j_m}(x , y)\,dx\,dy \nonumber\\ & = & \frac{ip}{\sqrt{2}}\left[\sum_\nu c^+_\alpha(\nu,\downarrow)t^+(n,\nu)+ \frac{1}{\sqrt{3}}\sum_\mu c^-_\alpha(\mu,\downarrow)t^-(n,\mu)\right]\end{aligned}\ ] ] for a review see : r. cingolani and r. rinaldi , rivista nuovo cimento * 16 * , 1 ( 1993 ) . e. kapon , d.m . hwang , and r. bhat , phys . * 63 * , 430 ( 1989 ) . m. gailhanou _ et al . _ , lett . * 62 * , 1623 ( 1993 ) . s. tiwari _ et al . _ , phys . lett . * 64 * , 3536 ( 1994 ) . r. rinaldi , _ et al . _ , phys . rev . b * 50 * , 11795 ( 1994 ) . m. grundmann , _ et al . _ , j. nonlin . optical phys . and materials * 4 * , 99 ( 1995 ) . l.n . 56 * , 1697 ( 1990 ) . d. gershoni _ et al . * 65 * , 1631 ( 1990 ) . et al . _ , lett . * 58 * , 726 ( 1991 ) . a.r . goi _ et al . _ , 61 * , 1956 ( 1992 ) . w. wegscheider _ et al . _ , j. cryst . growth * 150 * , 285 ( 1995 ) . h. akiyama , t. someya , and h. sakaki , phys . b * 53 * , r4229 ( 1996 ) . h. akiyama , t. someya , and h. sakaki , phys . b * 53 * , r10520 ( 1996 ) . t. someya , h. akiyama , and h. sakaki , phys . lett * 76 * , 2965 ( 1996 ) . sercel and k.j . vahala , app . . lett . * 57 * , 545 ( 1990 ) . d.s . citrin and y .- c . chang , phys . b * 40 * , 5507 ( 1989 ) . u. bockelmann and g. bastard , europhys . lett . * 15 * , 215 ( 1991 ) ; phys . b * 45 * , 1688 ( 1992 ) . d.s . citrin and y .- c . chang , phys . b * 43 * , 11703 ( 1992 ) . h. ando , s. nojima , and h. kanbe , j. appl . phys.*74 * , 6383 ( 1993 ) . g. goldoni and a. fasolino , phys . b * 52 * , 14118 ( 1995 ) ; physica b * 211 * , 444 ( 1995 ) . a.a . yamaguchi and a. usui , j. appl . phys . * 78 * , 1361 ( 1995 ) . f. rossi and e. molinari , phys . lett . * 76 * , 3642 ( 1996 ) ; phys . b * 53 * , 16462 ( 1996 ) ; f. rossi and e. molinari , in `` the physics of semiconductors '' , edited by m. scheffler and r. zimmermann ( world scientific , singapore 1996 ) , p. 1161 . g. goldoni , f. rossi , e. molinari , a. fasolino , r. rinaldi , and r. cingolani , appl . lett . in press . note that the electron subband index @xmath13 does not include spin , and two spin orientations correspond to each @xmath13 , while for holes the index @xmath28 includes the spin character . j. m. luttinger , phys . rev . * 102 * , 1030 ( 1956 ) . xia , phys . b * 43 * , 9856 ( 1991 ) . for holes , the word `` spin '' is used in a generalized way , and does not correspond to a well defined spin orientation . broido and l.j . sham , phys . b * 31 * , 888 ( 1985 ) ; l.c . andreani , a. pasquarello , and f. bassani , phys . b * 36 * , 5887 ( 1987 ) ; g. goldoni and a. fasolino , phys . lett . * 69 * , 2567 ( 1992 ) ; * 69 * , 2567 ( 1992 ) . the qw subbands are calculated by a finite element solution of the luttinger hamiltonian in the relevant crystallographic directions . see j. los , a. fasolino , and a. catellani , phys . b * 53 * , 4630 ( 1996 ) . note that the qwr ground state is actually sligtly above the qw2 continuum , contrary to the expectation for a localized level . actually , this is a fictitious result due to the different accuracy between the plane wave expasion of qwr states and the real - space representation used to calculate the qw states . in reality , the qwr ground state , being localized , lies slightly below the qw continuum . however , for this structure the binding energy is very small . g. bastard , _ wave mechanics applied to semiconductor heterostructures _ ( les ditions des physique , les ulis , paris , france 1988 ) . u. bockelmann and g. bastard , phys . b * 45 * , 1700 ( 1992 ) . t. ando , j. phys . japan * 54 * , 1528 ( 1985 ) . n & a / sl & a / alas & b / sl & b / alas + 1 & 43.3 & 68.3 & 55.5 & 98.8 + 2 & 57.3 & 91.9 & 63.2 & 117.0 + 3 & 65.3 & 112.5 & 67.0 & 124.0 + 4 & 69.8 & 126.1 & 72.4 & 132.2 + + m & a / sl & a / alas & b / sl & b / alas + 1,2 & 10.7 & 14.4 & 13.1 & 18.7 + 3,4 & 13.1 & 18.1 & 13.8 & 19.5 + 5,6 & 14.4 & 20.0 & 15.6 & 21.7 + 7,8 & 15.8 & 21.8 & 17.1 & 23.7

the shels ( smithsonian hectospec lensing survey ) is a complete redshift survey covering two well - separated fields ( f1 and f2 ) of the deep lens survey . both fields are more than 94% complete to a galactic extinction corrected r = 20.2 . here we describe the redshift survey of the f1 field centered at r.a. = 00.3 and decl. = 12 ; like f2 , the f1 field covers deg . the redshift survey of the f1 field includes 9426 new galaxy redshifts measured with hectospec on the mmt ( published here ) . as a guide to future uses of the combined survey we compare the mass metallicity relation and the distributions of d as a function of stellar mass and redshift for the two fields . the mass - metallicity relations differ by an insignificant 1.6 . for galaxies in the stellar mass range 10 to 10m , the increase in the star - forming fraction with redshift is remarkably similar in the two fields . the seemingly surprising 31 - 38% difference in the overall galaxy counts in f1 and f2 is probably consistent with the expected cosmic variance given the subtleties of the relative systematics in the two surveys . we also review the deep lens survey cluster detections in the two fields : poorer photometric data for f1 precluded secure detection of the single massive cluster at that we find in shels . taken together the two fields include 16,055 redshifts for galaxies with r and 20,754 redshifts for galaxies with r 20.6 . these dense surveys in two well - separated fields provide a basis for future investigations of galaxy properties and large - scale structure . key words : cosmology : observations galaxies : abundances galaxies : distances and redshifts galaxies : evolution large - scale structure of universe surveys

1603.06885

redshift surveys and weak lensing maps are two powerful and independent tools for tracing the matter distribution in the universe . approaches to combining these two powerful tools are developing rapidly as dense redshift surveys access the intermediate redshift universe and as weak lensing maps become increasingly extensive ( geller et al . 2005 ; kurtz et al . 2012 ; shan et al . 2012 ; van waerbeke et al . 2013 ; chang et al . 2015 ) . the shels ( smithsonian hectospec lensing survey ) project began as a platform for exploring the combination of dense , complete foreground redshift surveys with lensing maps ( geller et al . 2005 ; geller et al . 2010 ; utsumi et al . shels consists of dense redshift surveys of two of the five fields of the deep lens survey ( wittman et al . 2006 ; dls ) , f1 centered at r.a.@xmath1 = 00@xmath253@xmath325.3@xmath4 , decl.@xmath1 = 12@xmath533@xmath655@xmath7 and f2 centered at r.a.@xmath1 = 09@xmath219@xmath332.4@xmath4 and decl.@xmath1 = + 30@xmath1800@xmath1900@xmath7 . each of the dls fields covers 4 square degrees . the shels redshift surveys of the two fields are more than 94% complete to an extinction corrected r@xmath0 = 20.2 . these two fields currently represent the most densely sampled surveys to this magnitude limit . the dense , complete sampling makes the surveys useful for a wide range of astrophysical applications . the hectospec , a 300-fiber spectrograph with a 1@xmath5 field of view mounted on the mmt , enables surveys like shels . kochanek et al . ( 2012 ) also carried out a large redshift survey , ages , with hectospec . their survey focuses on agn evolution . the ages survey covers a 7.7@xmath5 contiguous region with a complex sampling strategy in several photometric bands . to a extinction corrected limit of r@xmath0 = 20.2 ( we use the subscript @xmath0 throughout to denote an extinction corrected magnitude ) , the number density of ages redshifts is @xmath81350 galaxies deg@xmath20 in contrast with the shels mean averaged over the two fields of 1961 galaxies deg@xmath20 . shels and ages are thus complementary in both their geometry and selection . here we describe the shels survey of the f1 dls field . geller et al . ( 2014 ) describe the survey of the f2 field . the f1 field contains a total of 9426 new galaxy redshifts and is 94% complete to an extinction corrected r@xmath0 = 20.2 . the f2 survey is somewhat deeper ; it is 95% complete to an observed r = 20.6 and 97% complete to r@xmath0 = 20.2 . taken at face value , the two fields seem remarkably different ; the raw counts to r@xmath0 = 20.2 differ by @xmath831% . the f2 field contains a prominent complex of rich clusters associated with abell 781 ( abell 1958 ) and easily detected with weak lensing ; f1 contains no clusters detected with weak lensing in the entire volume covered by the redshift survey ( ascaso et a. 2014 ) . we compare these observed differences with the expected cosmic variance for these fields . we also provide a guide to the properties of the survey and to some of its potential uses . we have already used both the f2 and f1 shels data for a variety of applications . for essentially all of these applications , the straightforward , complete magnitude limited survey is a critical underpinning . in concert with the original intent of the surveys , we have used the f2 survey as a basis for testing weak lensing maps against a foreground redshift survey ( geller et al . 2005 ; geller et al . 2010 ; utsumi et al . 2014 ; see also viola et al . we have also compared x - ray , spectroscopic and lensing selection demonstrating that the most massive systems are detected robustly with all three techniques ( starikova et al . 2014 ) . the extensive shels spectroscopy enables exploration of a variety of benchmarks for tracking galaxy evolution including the evolution of the h@xmath21 luminosity function ( westra et al . 2010 ) , the faint end of the composite galaxy luminosity function ( geller et al . 2012 ) , the nature of star - forming galaxies detected with wise ( hwang et al . 2012 ) , the impact of close pairs on star formation ( freedman woods et al . 2010 ) , the determination of central velocity dispersions for individual galaxies ( fabricant et al . 2013 ) , and the evolution of the mass - metallicity relation ( zahid et al . 2013 ; 2014 ) . here we provide guides to the quality and potential uses of the shels surveys by revisiting the stellar mass metallicity relation ; we compare the determinations for the f1 and f2 fields separately and in combination . similarly , we compare the d@xmath104000 distributions as a function of stellar mass and redshift for the two fields . we describe the f1 data in section 2 with attention to the differences between the photometric bases for the f1 and f2 surveys . we compare the completeness of the shallower f1 survey with f2 and lay the foundation for comparing the surveys to the same limiting extinction corrected apparent magnitude . in section 3 we compare various aspects of the f1 and f2 survey regions including the mass metallicity relation ( section 3.1 ) , the d@xmath104000 distributions as a function of redshift and stellar mass ( section 3.2 ) , galaxy counts ( section 3.3 ) , and cluster observations ( section 3.4 ) . in sections 3.3 and 3.4 we consider whether the differences in galaxy and cluster counts are consistent with the expected cosmic variance for fields of this size . we conclude in section 4 . we adopt h@xmath0 = 70 km s@xmath22 mpc@xmath22 , @xmath23 and @xmath24 = 0.3 throughout . all quoted errors in measured quantities are 1@xmath11 . the shels redshift survey covers two 4 deg@xmath9 fields originally selected as part of the deep lens survey ( dls ; wittman et al 2006 ) . like most fields chosen for deep photometric surveys , the selection of the deep lens survey fields avoids bright stars , nearby bright galaxies , and nearby rich clusters with redshift @xmath25 . this selection biases the fields toward low density at @xmath25 . geller et al . ( 2014 ) describe the redshift survey of the f2 field of the dls . in this field , the redshift survey is 95% complete to r = 20.6 ( observed total magnitude uncorrected for galactic extinction ) . the complete portion of the survey includes 12,705 redshifts . it is interesting to note that the f2 field contains an impressive complex of rich clusters at z @xmath8 0.3 and another at z @xmath8 0.5 ( geller et al . 2010 ; ascaso et al . 2014 ; utsumi et al . 2014 ; starikova et al . 2014 ) . the original intent of shels was to complete the redshift survey of the f1 field to r = 20.6 . although we measured many redshifts for galaxies to this limit ( and we report them here ) , observing conditions only allowed a survey 93% complete to an extinction corrected r@xmath0 = 20.2 . furthermore the dls reported no significant weak lensing peaks in f1 ( see ascaso et al . 2014 ) thus diminishing the incentive for further spectroscopy ( but see the discussion of clusters in section 3.4 ) . for the f1 field @xmath26 = 0.16 in contrast with @xmath26 = 0.05 for f2 . in the discussion below we compare the properties of the two fields at the limiting r@xmath0 = 20.2 . to this limit , the f2 field contains 9489 galaxies ( 9216 of these have redshifts ) ; remarkably , the f1 field contains only 7261 galaxies ( 6839 of these have redshifts ) . together the two fields currently constitute the largest complete redshift surveys to this limit . for comparison vvds - wide is deeper ( i@xmath27 22.5 ) than shels but sparsely sampled ( at the 20 - 25% level ) ; the target density is 4800 deg@xmath20 and it covers a similar area ( garilli et al . the gama survey is somewhat shallower ( r @xmath2819.8 ) and has a target density about a factor of 2 smaller than shels , but it covers a much wider area , and is more complete ( liske et al . 2015 ) . geller et al . ( 2014 ) describe the construction of the f2 catalog directly from the dls r - band photometry ( wittman et al . for this field we masked out regions around bright stars ( equation ( 1 ) in geller et al . 2014 ) , thus reducing the effective survey area to 3.98 square degrees . in the f1 field we began by selecting targets from the sdss dr9 ( ahn et al . 2012 ) in a region somewhat larger than the dls f1 footprint . the main galaxy candidates are extended sources based on the sdss point / extended source flag . we do not used the point sources in the analysis ; we use only the extended sources . we then use the available dls data to construct an r - limited galaxy catalog . for each sdss galaxy candidate with r@xmath29 , we searched for a dls counterpart within a 2@xmath7 tolerance . if there is a unique dls counterpart , we adopt the dls r - band magnitude . if there is no dls counterpart ( generally as a result of a nearby saturated star , a bleed trail , or proximity to the dls survey edges ) , we transform the magnitudes to the same system we use in f2 ( geller at al . we base the transformation on a fit to all of the galaxies we identify in both the dls f2 field and the sdss . we fit the dls f2 r - band magnitude ( r ) as a function of sdss @xmath30 and @xmath31 color to derive the transformation @xmath32 figure [ magcomp ] shows the residuals between the dls r - band magnitude and @xmath33 as a function of the sdss @xmath34 . the total number of objects with r@xmath16 that require this conversion is 710 or 9.8% of the sample to this limit . most of these objects are near the boundaris of the f1 redshift survey region and thus just outside the dls f1 footprint . there is also a small offset between the dls f2 magnitudes provided in their on - line database and the magnitudes reported in geller et al ( 2014 ) . we note that the f2 magnitudes reported in geller et al . ( 2014 ) are total magnitudes calibrated to the vega system and extrapolated from the isophotal magnitude within the 28.7 mag arcsec@xmath20 isophote . the zero point offset between these magnitudes and those provided on the dls website is 0.038 magnitudes . we add this small offset to magnitudes for galaxies with f1 photometry in order to put the two fields on the same system for a direct comparison . for f1 we do not mask regions around bright stars ; the effective area covered by the survey is 4.2 square degrees . there may be some undersampling of the galaxy distribution around bright stars , but based on the area excised in f2 , a much deeper survey than the sdss , we expect the effect to be substantially less than 5% . in the f2 field approximately 3% of the galaxy candidates in our original catalog turned out to be stars . in the f1 field , a comparable 4% of the galaxy candidates selected from the sdss turned out to be stars . the smaller fraction of stars in f2 is qualitatively consistent with better seeing for the dls imaging . the dls photometry is much deeper than the sdss . furthermore the seeing in f2 was 0.90@xmath7 with a @xmath35% variation among subfields ; for f1 the average seeing was 0.98@xmath7 varying from 0.94@xmath7 to 1.01@xmath7 . for the sdss , the median seeing ( for dr7 ) was 1.43@xmath7 . thus the dls photometry may reveal low surface brightness objects and/or objects that are unresolved in the sdss . remarkably , the number of f2 galaxies with r@xmath16 missed in the sdss photometry is only 37 , corresponding to 0.4% of total number of f2 galaxies with r@xmath3620.2 . these objects are indeed either low surface brightness galaxies or they have very close neighbor galaxies unresolved by the sdss . we may be missing a comparable fraction of galaxies in f1 because we constructed the catalog directly from the sdss ; we conclude that the impact of differences in catalog construction on the numbers of objects is negligible at this limiting magnitude . we note that failure to include low surface brightness objects is a potentially important limitation for investigating some issues including the faint end of the luminosity function ( blanton et al . 2005 ; geller et al . 2012 ) . we used the 300-fiber hectospec instrument ( fabricant et al . 1998 , 2005 ) on the mmt to acquire spectroscopy for galaxy candidates typically brighter than r = 20.6 . we observed the f1 field in queue mode during dark runs in four periods : october 24 - 28 , 2005 ; october 17 - november 22 , 2006 ; october 10 - december 10 , 2012 ; september 26 - november 28 , 2014 . in the 2012 and 2014 observations we filled unused fibers with wise sources ( wright et al . 2010 ) . to obtain a highly complete survey over the entire f1 field we applied the roll et al . ( 1998 ) observation planning software and we revisited each hectospec positioning more than 7 times on average . we use the same oberving protocols and reduction procedures for the f1 and f2 fields . figure [ fegsp ] shows typical high and low signal - to - noise spectra acquired in our 0.75 to 2 hour integration time . we show the spectra for a window in the rest frame . the wavelength range covered by hectospec in the observer s frame is 3,700 9,100 with a resolution of @xmath85 . the hectospec fibers have a 1.5@xmath7 diameter . we reduced the 2005 , 2006 , and 2012 data with the mink et al . ( 2007 ) hectospec pipeline and derived redshifts with rvsao ( kurtz & mink 1998 ; see also fabricant et al.2005 ) . we reduced the 2014 observations with hsred v2.0 developed by the sao telescope data center . this pipeline is a revision of the idl pipeline originally written by richard cool ( see http://www.mmto.org/book/export/html/55 ) . there is no difference between the redshifts derived from the two pipelines . in the analysis of the f2 data , we used repeat hectospec observations of 1651 pairs of spectra of 1651 unique objects to compute internal errors in the redshift . we used the generally brighter ovelapping sdss galaxies to estimate an external error . for emission line objects the internal error ( normalized by ( 1 + z ) ) is 48 km s@xmath22 and for absorption line objects it is 24 km s@xmath22 . the typical external error , which may be underestimated from comparisons with brighter sdss objects , is 37 km s@xmath22 regardless of spectral type . there is a small offset between the sdss and mmt redshifts , @xmath37 = 3.4 @xmath38 3.9 km s@xmath22 . these errors also apply for f1 ; the instrument , the procedures , and the reduction are essentially identical . as a result , in part , of variable conditions , the quality of spectra that yield acceptable redshifts varies significantly . the pipeline provides a standard measure of the quality of the spectrum , @xmath39 , a measure of the width of the cross - correlation peak originally defined by tonry & davis ( 1979 ) . based on the value of @xmath39 we compute the redshift error as in kurtz & mink ( 1998 ) . figure [ bfzxcr ] shows @xmath39 as a function of the extinction corrected r@xmath0 apparent magnitude for the f1 field . kurtz & mink ( 1988 ) show ( their figure 8) that @xmath39 measures the signal - to - noise of the spectrum . for surveys like f1 that span a significant redshift range @xmath39 is a better indicator of the quality of the redshift than signal - to - noise at a fixed wavelength . this figure is in the same format as figure 2 in geller et al . ( 2014 ) for f2 to facilitate direct comparison . here we limit the figure to the highly complete r@xmath0 @xmath40 sample in f1 . although figure [ bfzxcr ] suggests that we could select reliable redshifts based merely on the value of @xmath39 , we inspect each spectrum visually after the pipeline processing . we then classify the spectra with three flags : q for high quality , ? for dubious cases , and x for completely unacceptable data . we report only q redshifts here just as we did for f2 . visual inspection identifies ( 1 ) spectra corrupted by badly subtracted night sky , ( 2 ) cases where there are two objects at different redshift in a single fiber , and ( 3 ) a small number of quasars . otherwise the visual classification is essentially a reflection of @xmath39 . in the following sections we limit discussion to the redshift survey for r@xmath0@xmath41 , the faintest limiting magnitude where the completeness in f1 significantly exceeds 90% . in table [ tab20.2 ] and table [ tabfaint ] we list all of the redshifts we measured in the f1 field . in total , the tables include 9861 redshifts ; 9639 are new measurements with hectospec , 185 are from the sdss dr12 ( alam et al . 2015 ) , and 37 are from ned . table [ tab20.2 ] is the magnitude limited sample we analyze and table [ tabfaint ] contains redshifts we measured for fainter objects . the tables include the shels i d , the sdss objid , the total extinction corrected r@xmath0 magnitude from the dls along with its error , the redshift and its formal error derived from the @xmath39 value . the table includes a flag if the r - magnitude is converted from an sdss r - band magnitude or if the redshift comes from the sdss or ned . we also indicate whether the source is classified as a point source in the sdss based on the probpsf parameter in the sdss database . these sdss point sources were observed as wise qso candidates to fill unused fibers . we do not use any of these point sources in the analysis below and include them solely to publish the redshifts . as for the f2 field , tables [ tab20.2 ] and [ tabfaint ] also include three derived quantities : d@xmath104000 , the stellar mass and its error , and the metallicity ( for emission - line galaxies in the redshift range 0.2@xmath42 ( as we discuss in section 3.1 below ) . section 2.3 of geller at al . ( 2014 ) describes the derivation of these quantities and we do not repeat the discussion here . our procedures for f1 are identical to those for f2 . the computation of d@xmath104000 is based on the procedures described in fabricant et al . ( 2008 ) following the definition of balogh et al . we derive stellar masses according to the procedures described by zahid et al . ( 2013 ) based on the le phare code written by arnouts & ilbert ( arnouts et al . 1999 ; ilbert et al . the derivation of metallicities follows zahid et al . ( 2013 , 2014 ) based on the r23 line ratio calibration by kobulnicky & kewley ( 2004 ) . briefly , to construct the sample of star - forming galaxies we require a s / n@xmath433 in the line flux measurements of [ oii ] @xmath443727 , 3729 , h@xmath45 , h@xmath21 , and [ nii ] @xmath446584 . we remove agn from the sample of star - forming galaxies based on the bpt method ( baldwin et al . 1981 , bpt ) as updated by kewley et al . ( 2006 ) . again as in the f2 field we do not report unphysical values of d@xmath104000 ( d@xmath104000 @xmath46 or d@xmath104000@xmath473 . these values result from poor spectra that are merely adequate to yield a redshift . in section [ distdn4000 ] we compare the distributions of d@xmath104000 as a function of stellar mass and redshift for the f1 and f2 fields . the fraction of objects with unmeasurable d@xmath104000 in both fields ( as detailed in section [ distdn4000 ] ) is so small that the effect on this comparison is negligible . to make the definition of the sample with r@xmath48 clear , we list the galaxy candidates without redshifts to this limit in table [ tabnoz ] . among these candidates , visual inspection of the dls images suggests that @xmath49% of these remaining objects could be stars . in table [ tabstar ] , for completeness , we list the objects that we identify spectroscopically as stars . for both the f1 and f2 fields , table [ numbers ] lists the number of photometric objects and the number of spectroscopically confirmed galaxies for several magnitude limited samples . we focus on comparing two samples with high spectroscopic completeness to the extinction corrected r@xmath0 = 20.2 . in fact , the completeness listed in table [ numbers ] is a lower limit because 3 - 4% of the objects without a measured redshift are stars . we also list the number of redshifts ( published here ) for galaxies fainter than our initial nominal limit r = 20.6 . at face value there are several interesting aspects of the numbers in table [ numbers ] . first , at every magnitude limit , the raw number of galaxies per unit area is larger for f2 than for f1 . as the extinction corrected sample shows , this difference is not a result of the greater extinction for f1 . we show in section [ variance ] that this difference is actually consistent with the expected cosmic variance . figure [ bfcomplete ] provides a more detailed picture of the completeness of the f1 field . the figure is in the same format as figure 4 of geller et al . ( 2014 ) ; corrected for extinction , the magnitude limits in the corresponding panels are essentially the same . the 422 objects without a redshift in f1 are clearly not uniformly distributed over the field . in the map for the survey limited to r@xmath0 = 20.2 ( left panel ) most of the pixels that are @xmath50% complete lie along the edges , but there are a few within the central 8 @xmath51 8 pixel region . the top panel shows the steep drop in average completeness as a function of limiting extinction corrected magnitude . the right - hand panel shows the highly variable , relatively poor completeness of the f1 field in the interval @xmath52r@xmath0 @xmath53 . this panel substantiates our decision to limit discussion of this region to the brighter magnitude limit . the color - magnitude diagram ( figure [ fcmr ] ) for the objects without a redshift shows that at this bright limit , there is essentially no color dependence among the missing objects ( middle panel ) . as expected , the number of objects without spectroscopy increases significantly for the faintest objects in the survey reaching nearly 25% in the faintest 0.1 magnitude bin ( top and bottom panels ) . figure 5 of geller et al . ( 2014 ) shows the analogous plot for f2 where only 273 objects with r@xmath0@xmath40 lack a redshift ; the salient features ( the lack of color dependence and the rise in incompleteness with apparent magnitude ) are similar . the cone diagram of f1 projected in r.a . ( figure [ brafcone ] ) reveals the characteristic cosmic web structure ( geller & huchra 1989 ) . the color coding as a function of d@xmath104000 is the same as in the cone diagram for f2 in figure 6 of geller et al . ( 2014 ) . in both fields , the tendency toward galaxies with younger populations in lower density regions is evident in the bluer color of the points . in f1 there are no prominent concentrations of rich clusters ; in f2 , the a781 complex dominates the survey at z@xmath54 0.29 - 0.30 . a finger corresponding to a cluster at z@xmath55 is visible in the cone diagram . a movie with broader binning in d@xmath104000 shows the 3d structure in f1 and highlights the @xmath15 cluster by zooming in on it . section [ variance ] contains a more extensive discussion of clusters in the two fields . although the construction of the surveys of f1 and f2 is not identical , the surveys are sufficiently similar to provide an interesting comparison of two widely separated similarly observed regions of the universe . our goal is to investigate a few of the differences and similarities between the fields to highlight the quality of the data and to provide a benchmark for further scientific applications . these applications include but are not limited to planning , calibration , and analysis of color - selected surveys ( see e.g. damjanov et al . 2015 ; geller & hwang 2015 ) . two straightforward figures provide an introduction to the salient differences between the two fields . first , figure [ bnzf1f2 ] shows the normalized redshift histograms for the two regions . the difference is striking . the f2 field ( blue histogram ) has a broad , prominent peak centered near @xmath56 . the most prominent peak in the f1 field ( red ) is near @xmath57 . a k s test rejects the hypothesis that the distributions of the two samples are extracted from the same parent population at a confidence level of 99.9% . figure [ maps ] shows the distribution of galaxies with @xmath58 on the sky for both f1 and f2 . the galaxy isodensity contours show that f1 lacks any regions that reach the highest projected number density in f2 . for f2 the map shows the four high confidence clusters ( diamonds ) detected by three methods : shels spectroscopy , x - ray and weak lensing . all of these systems lie in the highest density regions of the map . in the f1 field we show one candidate cluster that is indicated only by shels and galaxy counts . we discuss the clusters further in section [ clusters ] . the combination of spectral properties with stellar masses of galaxies provides a powerful basis for understanding the nature of galaxies and their evolution ( e.g. brinchmann & ellis 2000 ; kauffmann et al . 2003 ; bell et al . brinchman & ellis ( 2000 ) emphasize the power of exploring galaxy evolution with cosmic time by combining a more transitory spectral signature with the generally slowly varying stellar mass derived from multi - band imaging . here we combine stellar masses derived from @xmath59 imaging with the results of strong - line metallicity diagnostics to explore the robustness of the mass - metallicity ( mz ) relation for survey regions like the dls fields . in section [ distdn4000 ] we examine the distribution of d@xmath104000 as a function of stellar mass and redshift as another measure of variations among fields similar to those probed by the dls . as in previous metallicity analyses of these fields ( zahid et al . 2013 ; geller et al . 2014 ; zahid et al . 2014 ) , we compute metallicities for star - forming galaxies in the redshift range 0.2@xmath42 . these limits are set by the 1.5@xmath7 fiber size and by the bandpass of hectospec . the fiber aperture is too small to include enough of the galaxy light ( @xmath60% ) for z@xmath17 0.2 ( see kewley et al 2005 ) , and h@xmath21 shifts out of the bandpass for redshifts @xmath61 0.38 making the exclusion of agn difficult . as for the f2 field ( geller et al . 2014 ) , we include metallicities for individual objects in table [ tab20.2 ] . figure [ fmzrel ] shows the mz relations for the two fields . the curves are fits of the form @xmath62^\gamma\bigr)\bigr]\ ] ] zahid et al . ( 2014 ) discuss the physical interpretation of this form for the mz relation . we simply note that @xmath63 is the saturation metallicity . for stellar masses @xmath64 , the metallicity approaches the saturation limit , @xmath63 . the slope , @xmath65 characterizes the mz relation for m@xmath66m@xmath0 . table [ tblmz ] lists the parameters of the mz relation fits in figure [ fmzrel ] . the lower panel of figure [ fmzrel ] shows the 95% confidence error ellipses for three samples : f1 ( red ) , f2 ( blue ) , and the f1 plus f2 sample from zahid et al . ( 2014 ) ( black ) . we quote values of the fit parameters along with the formal 1@xmath11 errors in the table . we also include the combined result from zahid et al . ( 2014 ) in table [ tblmz ] for reference . the results for f1 and f2 are two independent measurements of the mz relation in the redshift range 0.2@xmath42 . they provide some assessment of the potential impact of cosmic variance on the determination of the mz relation for samples covering a volume of @xmath8 10@xmath67 mpc@xmath68 in each field . the difference between the mz relations for the two fields is small . all of the 95% confidence ellipses for the mz relation parameters overlap ( figure [ fmzrel ] , lower panel ) . the parameter most sensitive to evolutionary effects is the transition stellar mass , m@xmath0 ( zahid et al . thus it is perhaps not surprising that the z@xmath0 vs log(m@xmath0/m@xmath14 ) error ellipses have the smallest fractional overlap , but the offset is only at the @xmath691.6@xmath11 level . peng & maiolino ( 2014 ) use the sdss to show that the mz relation has some environmental dependence : satellite ( generally less massive ) galaxies have greater metallicity in denser environments . this trend is generally consistent with the offset in the z@xmath0 vs log(m@xmath0/m@xmath14 ) error ellipses and the offset in the corresponding mz relations of figure [ fmzrel ] . the lower panel of figure [ fmzrel ] also shows confidence ellipses for the f1 plus f2 sample which effectively averages over the differing overall densities of the f1 and f2 fields . the error ellipses for the combined sample are located , mainly and not surprisingly , where the ellipses for f1 and f2 overlap . the corresponding parameters for the f1 plus f2 sample listed in table [ tblmz ] are the best estimate of the mz relation for the entire shels survey covering the redshift range 0.2 @xmath42 . the small difference between the independent estimates based on f1 and f2 suggests that the result for the combined sample is a robust representation of the properties of star - forming galaxies in this redshift range . the spectral indicator d@xmath104000 has a rich history as a measure of galaxy properties and their evolution ( mignoli et al . 2005 ; bundy et al . 2006 ; roseboom et al . 2006 ; noeske et al . 2007 ; vergani et al . 2008 ; freedman woods et al . 2010 ; moresco et al . 2010 ; moustakas et al . 2013 ; moresco et al . 2013 ; geller et al . 2014 ) . here we examine the distributions of d@xmath104000 as a function of both stellar mass and redshift for the f1 and f2 fields . as for the mz relation , we explore the impact of the different overall galaxy density in the two fields on these distributions . we seek to link any differences with differences in features of the large - scale galaxy distribution in the two fields . figure [ fmassz ] ( left ) shows stellar mass as a function of redshift for f1 . the points representing galaxies in the survey are color - coded by d@xmath104000 . the segregation of d@xmath104000 with stellar mass is obvious ; galaxies with larger stellar mass have larger d@xmath104000 suggesting older ages and perhaps higher metallicities . there is also the known evolutionary trend that large values of d@xmath104000 occur for galaxies with lower stellar mass at lower redshift . the right - hand panel of figure [ fmassz ] shows the k - corrected absolute magnitude ( shifted to @xmath70 ) for the f1 survey . again the points are color - coded with the value of d@xmath104000 . the contrast between the left and right panels demonstrates the known advantage of using stellar mass rather than absolute magnitude to characterize galaxy populations . the advantages of d@xmath104000 include the strength of the feature , its essential redshift independence ( in contrast with colors ) , and its insensitivity to reddening . the simplest interpretation of the d@xmath104000 indicator as a population age indicator is complicated somewhat by metallicity dependence ( e.g. balogh et al . 1999 ; kauffmann et al 2003 ) . based on repeat hectospec observations , fabricant et al . ( 2008 ) show that the error in the d@xmath104000 is 0.045 times the value of the index . comparisons with measurements derived from the larger sdss fiber apertures show that there is no apparent bias in the hectospec 1.5@xmath7-fiber d@xmath104000 values relative to the 3@xmath7-fiber sdss values . although there may be subtle aperture effects as a function of redshift and stellar mass , we assume the values we derive for the galaxies in f1 and f2 are representative . our goal here is comparison of the distributions of the indicator in the two fields where we segregate objects by both redshift and stellar mass ; thus aperture effects should be irrelevant . as we noted in section [ spectroscopy ] , the spectra of some objects do not provide a measure of d@xmath104000 . over the range in redshift and stellar mass explored in figure [ subdn4000 ] , there are only 3 objects in the f1 field ( 0.1% of the total survey ) without a measure of d@xmath104000 . in f2 there are 198 objects ( 2.7% of the sample ) without a measure of d@xmath104000 . only the bin 0.5 @xmath71 0.6 , 11@xmath72 11.5 is significantly affected by this incompleteness : 11% of the galaxies in this bin lack a secure d@xmath104000 . for all other bins , the small fractions of missing objects do not affect the comparison of d@xmath104000 distributions . figure [ subdn4000 ] shows the d@xmath104000 distributions as a function of redshift and stellar mass for f1 ( red ) and f2 ( blue ) . the histograms are not normalized to emphasize the impact of large - scale structure on the occupancy of these bins . in each panel we list the number of galaxies in each survey and the fraction , @xmath73 , with d@xmath104000@xmath74 . the fiducial value 1.5 effectively separates star - forming from quiescent galaxies ( freedman woods et al . 2010 ) . remarkably , the values of @xmath73 are consistent for the two samples in nearly every redshift - mass bin in figure [ subdn4000 ] even though the numbers of galaxies in the bin may sometimes differ significantly . well - populated bins with the most significant differences in raw counts occur in the ranges 0.2 @xmath75 0.3 , 10@xmath76 11.5 and 0.1@xmath77 , 9@xmath76 9.5 . the presence of massive clusters of galaxies ( see section [ clusters ] ) accounts for the difference at 0.2@xmath78 ; in f2 the a781 complex of two massive clusters contributes substantially in this range ; there is no comparable system in f1 . in the range 0.3@xmath79 , the a781 complex still contributes in f2 and f1 contains a massive system with mean redshift 0.35 ( see section 3.4 ) . at 0.2@xmath78 , the most significant differences occur in the stellar mass range 10@xmath72 11.5 as a result of the stellar mass dependence of galaxy clustering ( e.g. bielby et al . 2014 ) ; galaxies of greater stellar mass preferentially inhabit denser regions . in the low redshift bin , 0.1 @xmath77 , the greater absolute abundance of galaxies with 9@xmath72 9.5 in f2 reflects the presence of a z@xmath80 structure where the survey is deep enough to include these low stellar mass galaxies . it is worth emphasizing that the dls field selection avoids regions with clusters at @xmath81 0.1 , but the presence / absence of structures near this limit affects the galaxy count particularly at low stellar mass . the largest differences in the relative abundances of galaxies in the bins of figure [ subdn4000 ] primarily reflect the differences in the populations of clusters near the peak sensitivity of the redshift survey . comparison of the d@xmath104000 distributions for the f1 and f2 fields shows that , averaged over large redshift bins that encompass both dense structures and low density regions , the fractions of quiescent ( large d@xmath104000 ) and star - forming ( small d@xmath104000 ) galaxies are surprisingly similar in the two fields . the lower panels of figure [ massivecone ] show the behavior of the fractions of star - forming and quiescent galaxies in f1 ( left ) and f2(right ) as a function of redshift for galaxies with stellar masses in the range 10@xmath12 to 10@xmath13 m@xmath14 . the upper panels show the corresponding cone diagrams . comparison of the cone diagrams with the fractions in the lower panels shows the expected dominance of star - forming galaxies in low density regions along with the enhancement of quiescent galaxies in dense regions of the survey . the plots also show clearly that the general trend of the star - forming fraction with redshift is essentially the same for the two fields . this result is similar to the early conclusion of bundy et al . ( 2006 ) who show that the fractions of quiescent and star - forming galaxies in samples drawn from the deep2 data ( davis et al . 2003 ; newman et al . 2013 ) are relatively insensitive to selection effects in their survey . bundy et al . ( 2006 ) conclude , as figure [ subdn4000 ] emphasizes for f1 and f2 , that the quiescent fraction rises with decreasing redshift in every mass bin and the star - forming fraction correspondingly declines . furthermore , the fraction of star - forming galaxies increases as the stellar mass decreases at fixed redshift . the d@xmath104000 distributions for f1 and f2 are also consistent with analyses of the zcosmos samples ( scoville et al . 2007 ; lilly et al . 2009 ) by moresco et al . based on 1000 early - type galaxies , they show that in the redshift range 0.45@xmath82 , d@xmath104000 decreases with redshift in a given stellar mass range . the shels survey offers a larger , complete sample of quiescent galaxies overlapping this redshift range and as a platform for extending a more detailed analysis to lower redshift . although we do not pursue the detailed relationship between galaxy properties and environment here , the shels data are very well - suited to such investigations . the redshift survey is dense and complete and the redshift errors are small ( @xmath8350 km s@xmath22 ) . thus velocity dispersions even in the thin structures typical of the cosmic web can be measured robustly . with well - controlled galaxy selection , the combination of the f2 and f1 fields covers a range of environments ranging from a dense complex of massive clusters in f2 to the many obvious low density regions . the f1 and f2 fields are currently unique in the completeness of the redshift surveys to the limiting apparent magnitude . it is thus interesting to compare the galaxy and cluster counts in the two fields as a measure of the potential impact of cosmic variance . it is evident from sections [ mzrelation ] and [ distdn4000 ] that the impact is negligible on scaling relations like the mz relation and on population fractions ( as opposed to absolute abundances ) as a function of redshift.the potential impact of the variance is obviously largest on quantities dependent on an absolute normalization as a function of redshift . table [ numbers ] lists the count of galaxies in the two fields for various samples . the cleanest samples for comparing f1 with f2 are those limited to r@xmath0 = 20.2 . there are 7261 and 9489 objects in f1 and f2 , respectively to this limit . most of these objects have a redshift , but among those without a redshift , a small fraction are stars . using the measured fraction of stars among the objects with spectroscopy in the two fields ( 3% were stars in f2 and 4% were stars in f1 ) we can reasonably convert the number of photometric objects to the number of probable galaxies brighter than the limit : 7244(f1 ) and 9481(f2 ) . in the f1 field we selected objects from the sdss and thus we did not remove regions around stars that were saturated in the dls photometry as we did for f2 . thus the effective area covered by f1 is larger . there may be some diminution in the counts in f1 as a result of this difference in procedure , but based on the observations , this difference should be @xmath84% , essentially the ratio of the effective areas in table [ numbers ] . thus the ratio of counts r@xmath85 in the two fields lies in the range @xmath86 . for the lower limit we simply take the ratio of the observed counts ( we assume that undercounting in f1 roughly compensates for the larger areal coverage ) ; for the upper limit we normalize the observed counts by the relative areas covered by the two surveys thus assuming that any undercounting of galaxies near bright stars in f1 is negligible . we can also compare the counts of massive galaxies seen throughout the range @xmath87 , a volume of @xmath88 mpc@xmath68 , in each field . based on the data in table [ tab20.2 ] and table 3 of geller et al . ( 2014 ) , the numbers of galaxies with stellar masses in the range 10.5@xmath76 11.5 in the two fields are 2848 ( f1 ) and 3315 ( f2 ) . correcting for the incompleteness and for the small difference in the relative volumes of the surveys , the ratio of these populations is 1.19@xmath380.03 . this ratio assumes that the counts of these massive objects are unaffected by bright stars in f1 ( in other words we simply correct the relative counts by the small difference in the relative volumes of the two surveys ( 4.2 ( f1)/3.98 ( f2 ) ; table [ numbers ] ) and by the small difference in incompleteness ( 0.94 ( f1)/0.97 ( f2 ) ; table [ numbers ] ) ) . the count ratio is at best marginally consistent with the range we obtain from the full magnitude limited count . the relative counts of massive galaxies are probably a more robust reflection of any difference between the two fields than the raw counts . because the observing strategies are not identical , there are a number of subtle systematics that probably have a larger effect on the total count . these include failure to include low surface galaxies in f1 and failure to eliminate regions around bright stars . the magnitude transformation ( figure [ magcomp ] ) may also introduce systematics . we have checked the impact of the magnitude conversion by examining the central 3.57 deg@xmath9 of the f1 field where only 2.2% of the objects require a magnitude transformation ( as opposed to the 9.8% for the full field ) . the average number density of galaxies in this central field is only @xmath8 1% less than in the full field we analyze in detail . finally f2 contains a set of massive clusters at z@xmath8 0.53 ( note the peak in the redshift histogram of figure [ bnzf1f2 ] ) that also enhance the total count relative to f1 . in contrast , the count of massive galaxies concentrates on high surface brightness objects over the redshift range where both fields are well - sampled . we next compare the count ratios with the expected impact of cosmic variance on the relative counts in f1 and f2 . driver & robotham ( 2010 ) use sdss dr7 as a basis for developing a formula for computing the expected cosmic variance for a survey covering volumes @xmath89 10@xmath90 mpc@xmath68 ( for h@xmath0 = 70 km / s / mpc ) with arbitrary shape and depth ( their equation ( 4 ) ) . they apply their approach to surveys including the vvds ( le fevre et al . 2005 ; de la torre et al . 2007 ) that covers similar volumes in the redshift range of shels . the vvds is sparse to the shels limiting apparent magnitude . thus , for comparing the counts in f1 and f2 , the analytic approach of driver & robotham ( 2014 ) provides a good guide . technically their approximation applies to galaxies with absolute luminosities m@xmath91@xmath381 ( the numbers of galaxies in this magnitude range are similar to the numbers in the mass range we explore ) . they also show ( their table 1 ) that their estimates agree well with independent calculations for galaxies of stellar mass @xmath92 = 10.75 ( moster et al . 2010 ) , analogous to our massive galaxy sample . the formula given by driver & robotham ( 2010 ) provides the expected cosmic variance as a function of the median redshift transverse length and the median radial depth of the survey . the variance they compute for a 4 square degree field covering 0 @xmath93 with a 1:1 aspect ratio is 10% ( their table 2 ) . from their equation ( 4 ) we derive an identical result for a 4 square degree region with the median redshift , 0.29 , that we measure for f2 limited to z = 0.5 . for a field like our 4.2 square degree field ( f1 ) with its slightly lower median redshift , 0.27 , for the sample limited to z = 0.5 , the predicted cosmic variance is also 10% . driver & robotham ( 2010 ) show that the distribution of galaxy counts in cells lies between a gaussian and a log normal distribution . they use the gaussian approximation in their exploration of variance issues and we follow the same route . if we draw two independent samples randomly from a gaussian with dispersion @xmath11 , the mean of the absolute difference is ( 1.13@xmath38 0.85)@xmath11 . the result is based on 100,000 draws . the mean of the absolute difference that we obtain numerically agrees with the analytic prediction , ( 2/@xmath94)@xmath11 . for a 10% cosmic variance ( @xmath11 ) , the observed fractional difference between between f2 and f1 differs from the simulated mean ( 1.13@xmath11 = 0.113 ) lies in the range @xmath95 times the expected error in the mean ( 0.85@xmath11 = 0.085 ) . the lower limit corresponds to the count of bright galaxies and the upper limit derives from the raw count ratios . thus the fractional difference in the abundance of massive galaxies ( and the fractional difference in counts to the magnitude limit ) is probably consistent with the expected impact of cosmic variance to within our ability to control for the differing systematics in the two surveys . the original goal of the shels survey of the dls f1 and f2 fields was comparison of the matter distribution traced by galaxies in a redshift survey with the matter distribution inferred from a weak lensing map . in the f2 field we cross - correlated the galaxy map with the lensing map to show that indeed the lensing map images the projected matter density traced by galaxies in the foreground redshift survey ( geller et al . later we examined the correspondence of weak lensing peaks with clusters of galaxies identified from the redshift survey ( geller et al . 2010 ; utsumi et al . 2014 ) and from an x - ray survey ( starikova et al . 2012 ) . these studies underscored some of the difficulties in constructing cluster catalogs . in particular they showed that a threshold signal - to - noise of 4.5 for detection of a weak lensing peak yields an essentially false positive free catalog . here we compare weak lensing cluster detections in f2 and f1 for the redshift range of maximum sensitivity , 0.15 @xmath93 the f2 field is distinctive because nearly all of the spectroscopic and weak lensing cluster candidates have been observed in the x - ray with either _ chandra _ or _ xmm - newton_. approximately 75% of the clusters identified from optical spectroscopy are extended x - ray sources ; 60% of the weak lensing peaks with a signal - to - noise of 3.5 correspond to extended x - ray sources . in the redshift range 0.15 @xmath93 where clusters are well - sampled in the redshift survey and where the weak lensing map has maximum sensitivity , there are 4 clusters identified cleanly and independently by all three methods : weak lensing , x - ray , and shels spectroscopy ( these clusters are x0918 + 2953 , x0920 + 3028 , x0920 + 3030 , and x0921 + 3027 in table 1 of starikova et al ( 2014 ) ) . their inferred masses ( m@xmath96 ) range from 7@xmath97 to 5@xmath98 m@xmath14 ; the corresponding rest frame line - of - sight velocity dispersions range from 657@xmath3893 km s@xmath22 to 997@xmath38120 km s@xmath22 . figure [ maps ] shows the positions of the four clusters in f2 . they coincide with the regions on the sky most densely populated by galaxies brighter than our apparent magnitude limit . in f1 , the observations are much less extensive . there are no x - ray observations of extended sources in this field . ascaso et al . ( 2014 ) identified no weak lensing peaks exceeding the detection threshold . however , the dls data for f1 are of poorer quality than those for f2 ; the typical seeing was 0.98@xmath7 as opposed to 0.90@xmath7 for f2 . in the redshift survey , we identify a single potentially massive cluster in the redshift range @xmath99 . this cluster appears as a finger centered at z = 0.35 in the cone diagram of figure [ brafcone ] . its position corresponds to the largest peak in the galaxy number density map of ascaso et al . ( 2014 : figure a1 ) . in the ascaso et al . ( 2014 ) map , this feature is the only case where a low significance weak lensing peak overlies the peak in the galaxy surface number density . the cluster bcg is located at @xmath21 = 13.628@xmath5 , @xmath100 = 12.552@xmath5 and z = 0.35 . the shels survey contains 37 cluster members with a rest frame line - of - sight velocity dispersion of 834@xmath38126 km s@xmath22 ( figure [ rvdiag ] ) within a projected radius of 1.5 mpc . the cluster is isolated in redshift space and we make no attempt here to refine the membership . obviously , tighter limits on the rest - frame velocity would result in a smaller effective line - of - sight velocity dispersion . this cluster is coincident on the sky with a redmapper cluster ( rykoff et al . 2014 ) , rmj005430.7 + 123305.9 ; however , the redmapper photometric redshift is 0.37 , inconsistent with our data . a cluster like rmj005430.7 + 123305.9 would probably be detected as a weak lensing peak in f2 at well above the @xmath101 level ( see geller et al . 2010 ; figure 12 ) . in f1 , however , the poorer image quality implies that a detection would be marginal . the worse and more variable seeing along with the greater galactic extinction toward f1 degrade the f1 lensing map relative to the analogous map for f2 . furthermore scattered light from stars outside the field created diffuse sprays of light in the poorly baffled camera during the f1 observations . these effects reduce the effective number of resolved sources by a factor of 2.1 per unit area relative to f2 . this difference substantially reduces the detection limits for weak lensing . more precisely , in f2 clusters with rest frame line - of - sight velocity dispersions in the range 603 - 697 km s@xmath22 would be weak lensing detections at a signal - to - noise of 3 - 4 ( geller et al . 2010 ) ; in f1 the corresponding range computed as in geller et al . ( 2010 ) is 753 - 870 km s@xmath22 . this reduction in sensitivity probably accounts for the absence of a weak lensing detection by ascaso et al . ( 2014 ) . we conclude that although the difference between f1 and f2 seems remarkable at first glance , the relative galaxy counts are consistent with cosmic variance estimates . the difference in the number of clusters detected in the weak lensing maps is driven primarily by shot noise along with significant differences in the observing conditions that affect the construction of a weak lensing map for f1 . the shels project covers two widely separated 4 square degree fields of the deep lens survey . these fields include 16,055 redshifts and they are both @xmath61 94% complete to an extinction corrected r@xmath0 = 20.2 . other surveys covering comparable areas to the same apparent magnitude limit are sparser and/or color - selected . the straightforward selection in apparent magnitude makes shels a useful benchmark for evaluating selection effects based on other approaches . the median redshift of both shels fields is z @xmath8 0.3 . the redshift histograms of the two regions nonetheless differ significantly . this difference is largely driven by differences in the details of the large - scale structure in each region , particularly the presence of several massive clusters of galaxies in f2 . we have previously used the shels data to determine the stellar mass - metallicity relation for the redshift interval 0.2@xmath42 . comparison of the relation determined for the f1 and f2 fields separately shows that the fiducial mass , m@xmath0 characterizing the relation changes insignificantly ( at the 1.6 @xmath11 level ) from one field to the other . the relation for the two fields is a remarkably robust estimate for this redshift range . we provide the metallicities and stellar masses as we did for f2 in geller et al . ( 2014 ) . as we did for the f2 field ( geller et al . 2014 ) we use the distribution of the spectral indicator d@xmath104000 as a proxy to discriminate between the star - forming and quiescent populations as a function of redshift and stellar mass . the behavior of the d@xmath104000 distributions for the two fields is remarkably similar with salient differences driven either by low redshift structures sampled to low stellar mass or to the presence of massive clusters of galaxies . for galaxy stellar masses in the range 10@xmath12 to 10@xmath13m@xmath14 , the star - forming population fraction as a function of redshift are remarkably similar for the two fields in spite of the difference in the overall mean galaxy density . these broadly binned results are a guide to the use of the data for more detailed analysis of the properties of galaxies as a function of their environment . as a result of the small redshift errors ( @xmath102 km s@xmath22 ) , the shels data are particularly well - suited to this task . in contrast with the stable population fractions , the raw counts of galaxies and massive clusters in the two fields seem , at first glance , to differ significantly . the procedures in observing the two fields were not identical thus complicating the comparison . however , the 31 - 38% difference in the counts to the limiting magnitude ( and the smaller 19% in the count of massive galaxies ) is probably consistent with the expected cosmic variance to within our ability to control for the relative systematics in the two surveys . comparing the abundance of massive clusters in the two fields is complicated by the much poorer data available for f1 . in the f2 field there are extensive x - ray observations and the weak lensing data is of higher quality than for f1 . in fact , the shels survey uncovers a cluster in f1 at z = 0.35 that probably should have been detected in f1 if the lensing data were of the same quality as for the f2 field . the difference in the cluster count is dominated by shot noise , but the apparent difference is accentuated by the poorer f1 lensing data . this comparison underscores the need for well - controlled calibration of surveys and underscores the subtle issues that enter the comparison of data for fields observed under different conditions and with different observational approaches . the shels survey covers 8 square degrees in two widely separated fields and includes 20,754 redshifts for galaxies with r@xmath17 20.6 along with 4457 redshifts for fainter objects . the complete surveys of the f1 and f2 fields provide a resource for many investigations of galaxy properties and their environments . the completeness of the surveys to the apparent magnitude limit provides benchmarks for color - selected surveys and for the development of new strategies based on combinations of imaging and spectroscopy . we appreciate a thoughtful , careful review by the referee that substantially improved this paper . we thank changbom park of kias for generous support that enabled completion of this project . we thank kairy herrera of brown university for examining images of faint galaxies with superimposed stars to refine the construction of table 4 . we thank warren brown for remeasuring two redshifts for two of the objects examined by kairy herrera . scott kenyon provided advice on numerical methods along with comments on the manuscript . perry berlind and mike calkins masterfully operated the hectospec . susan tokarz and sean moran assisted with the data reduction . the smithsonian institution supported the research of margaret geller , ho seong hwang , michael kurtz , daniel fabricant and jabran zahid . kias supported the research of ho seong hwang . ccccccccrc 12.324533 + 12.480500 & 1237678919673053653 & @xmath103 & @xmath104 & 1 & 2 & 0 & 1.76 & @xmath105 & ... + 12.324566 + 12.653818 & 1237678859550589594 & @xmath106 & @xmath107 & 1 & 2 & 0 & 1.23 & @xmath108 & 9.05 + 12.324614 + 12.798466 & 1237678859550589182 & @xmath109 & @xmath110 & 1 & 2 & 0 & ... & @xmath111 & ... + 12.325052 + 11.897449 & 1237678858476913241 & @xmath112 & @xmath113 & 1 & 2 & 0 & 1.82 & @xmath114 & ... + 12.325632 + 12.798472 & 1237678859550589183 & @xmath115 & @xmath116 & 1 & 2 & 0 & 1.28 & @xmath117 & ... + 12.327368 + 12.549498 & 1237678919673053413 & @xmath118 & @xmath119 & 1 & 2 & 0 & 1.96 & @xmath120 & ... + 12.327427 + 12.550200 & 1237678919673053412 & @xmath121 & @xmath122 & 1 & 2 & 0 & 1.78 & @xmath123 & ... + 12.327430 + 11.594275 & 1237678918599311681 & @xmath124 & @xmath125 & 1 & 2 & 0 & 1.01 & @xmath126 & 8.77 + 12.327575 + 12.491597 & 1237678919673053661 & @xmath127 & @xmath128 & 1 & 2 & 0 & 1.40 & @xmath129 & 9.14 + 12.327611 + 12.244888 & 1237678859013783788 & @xmath130 & @xmath131 & 1 & 2 & 0 & 1.22 & @xmath132 & ... + ccccccccrc 12.389126 + 13.290399 & 1237678920746795438 & @xmath133 & @xmath134 & 1 & 1 & 0 & 1.34 & @xmath135 & ... + 12.389223 + 13.548410 & 1237678860624396698 & @xmath136 & @xmath137 & 1 & 2 & 0 & 1.71 & @xmath138 & ... + 12.389453 + 13.234157 & 1237678860087525736 & @xmath139 & @xmath140 & 1 & 1 & 0 & 1.92 & @xmath141 & ... + 12.390178 + 12.722615 & 1237678859550654860 & @xmath142 & @xmath143 & 1 & 1 & 0 & 1.16 & @xmath144 & ... + 12.391906 + 12.125328 & 1237678919136182371 & @xmath145 & @xmath146 & 1 & 1 & 0 & 1.36 & @xmath147 & ... + 12.392926 + 13.139955 & 1237678860087525754 & @xmath148 & @xmath149 & 1 & 1 & 0 & 1.43 & @xmath150 & ... + 12.396500 + 12.704612 & 1237678859550654877 & @xmath151 & @xmath152 & 1 & 1 & 0 & 1.15 & @xmath153 & 8.77 + 12.396858 + 13.245761 & 1237678920746795610 & @xmath154 & @xmath155 & 1 & 1 & 0 & 1.34 & @xmath156 & ... + 12.397086 + 12.221524 & 1237678859013783938 & @xmath157 & @xmath158 & 1 & 1 & 0 & 1.79 & @xmath159 & ... + 12.397125 + 12.852595 & 1237678920209924690 & @xmath160 & @xmath161 & 1 & 1 & 0 & 1.31 & @xmath162 & ... + ccc 12.315080 + 12.318782 & 1237678859013718622 & @xmath163 + 12.339393 + 12.349747 & 1237678859013783808 & @xmath164 + 12.352705 + 13.175315 & 1237678860087525659 & @xmath165 + 12.358372 + 11.543474 & 1237678857940042136 & @xmath166 + 12.359576 + 12.377343 & 1237678859013783569 & @xmath167 + 12.360730 + 12.853304 & 1237678920209924618 & @xmath168 + 12.363524 + 13.473994 & 1237678860624396650 & @xmath169 + 12.363848 + 12.351914 & ... & @xmath170 + 12.365371 + 12.855583 & 1237678920209924173 & @xmath171 + 12.367078 + 12.860638 & 1237678920209924839 & @xmath172 + cccccc 12.311022 + 13.024869 & 1237678920209924333 & @xmath173 & @xmath174 & 2 & 2 + 12.322512 + 12.238044 & 1237678859013783624 & @xmath175 & @xmath176 & 1 & 2 + 12.326695 + 12.168094 & 1237678919136182450 & @xmath177 & @xmath178 & 1 & 2 + 12.331113 + 13.070127 & 1237678860087460426 & @xmath179 & @xmath180 & 1 & 2 + 12.331615 + 12.833143 & 1237678920209924451 & @xmath181 & @xmath182 & 2 & 2 + 12.334228 + 12.140061 & 1237678919136182455 & @xmath183 & @xmath184 & 2 & 2 + 12.339199 + 11.762720 & 1237678918599311568 & @xmath185 & @xmath186 & 1 & 1 + 12.343303 + 12.813752 & 1237678859550654681 & @xmath187 & @xmath188 & 1 & 1 + 12.350332 + 11.883782 & 1237678858476912807 & @xmath189 & @xmath190 & 1 & 1 + 12.355689 + 12.097490 & 1237678919136182468 & @xmath191 & @xmath192 & 1 & 1 + lccr e(b@xmath193v ) & 0.06 ( a@xmath194=0.16 ) & 0.02 ( a@xmath194=0.05 ) + survey area ( deg@xmath9 ) & 4.20 & 3.98 ( excludes masked area ) + n@xmath195 & 7261 & 9489 + n@xmath196 & 6839 & 9216 + completeness@xmath197 & 94.2% & 97.1% + z@xmath198 & 0.282 & 0.294 + n@xmath199 & 6626 & 9946 + n@xmath200 & 6345 & 9643 + completeness@xmath201 & 95.8% & 97.0% + n@xmath202 & 9207 & 13408 + n@xmath203 & 8049 & 12705 + completeness@xmath204 & 87.4% & 94.8% + n@xmath205 & 1515 & 2942 + n@xmath206 & 9564 & 16319 + lccl n@xmath207&1446&2131&3577 + log(m@xmath0/m@xmath14 ) & 9.56@xmath380.03&9.50@xmath380.02&9.52@xmath380.02 + z@xmath0 & 9.10@xmath380.01 & 9.10@xmath380.01&9.10@xmath380.004 + @xmath65&0.49@xmath380.03&0.52@xmath380.03&0.52@xmath380.02 + . , between the dls r - band f2 magnitude ( r ) and r@xmath208 ( equation 1 ) as a function of the sdss r - band petrosian magnitude , r@xmath209 . the typical magnitude errors are 0.01 ( dls r ) and 0.11 ( sdss r@xmath209 ) . [ magcomp],width=672 ] -value ( tonry & davis 1979 ) , a redshift quality indicator , as a function of redshift ( upper panel ) . the center panel shows apparent r@xmath0 magnitude as a function of redshift . we display only 50% 0f the data for clarity . the lower panel shows a redshift histogram in bins of @xmath210 . [ bfzxcr],width=672 ] magnitude . the color bar shows the completeness fractions for the spatial completeness displays in the lower two panels . the lower left panel shows the completeness in 12@xmath5112 arcminute bins for galaxies with r@xmath16 . the yellow points indicate galaxies in the photometric sample without a measured redshift . the right hand plot shows the completeness in the interval 20.2@xmath211r@xmath0@xmath212 in the same format . note that for the r@xmath16 survey the most significant incompleteness occurs at the corners and edges of the field . [ bfcomplete],width=672 ] magnitude ( upper panel ) . color - magnitude diagram for the 422 objects without a redshift ( middle panel ) with r@xmath16 . contours indicate the relative density of objects with a redshift ; the absence of a slope as a function of r@xmath0 suggests that there is little obvious color bias in these objects . the bottom panel shows the number of objects with redshifts in the shels survey ( open histogram ) and the number of unobserved galaxy candidates ( red hashed histogram ) as a function of the r@xmath0 magnitude . [ fcmr],width=624 ] 20.2 shels f1 survey projected on the r.a.@xmath1 direction . the color coding indicates the value of d@xmath104000 . in the low density regions , galaxies with d@xmath104000 @xmath89 1.5 predominate as expected . the online journal includes a video display of the data . the color - coding of the video is in broader bins : d@xmath104000 @xmath213 ( blue ) , 1.3 @xmath17 d@xmath104000 @xmath214 ( green ) , and d@xmath104000 @xmath215 1.7 ( red ) . [ brafcone],width=624 ]

absorption features from the lyman and werner bands of interstellar molecular hydrogen were recorded by the interstellar medium absorption profile spectrograph ( imaps ) at in the spectra of ori a and ori . the objective was to find and study more examples of an unusual phenomenon found for one of the velocity components of h in the spectrum of ori by jenkins & peimbert ( 1997 ) . specifically , they detected a gradual shift in velocity and broadening for features arising from progressively higher rotational excitations . this effect appears to be absent in the spectra of both and ori , which are only a few degrees away in the sky from ori . the absence of atomic material at a large negative velocity in the spectra of and ori ( and its presence in ori ) supports a proposal by jenkins & peimbert that the line of sight to intercepts a bow shock facing away from us , perhaps created by the collision of wind - like material with some foreground obstruction . one edge of the molecular cloud complex lynds 1630 is situated close to ori in the sky , but we present some evidence that seems to indicate that the cloud is more distant , in which case it could not serve as the obstruction . however it is possible that the outermost extension of a high - speed jet from a star forming within the cloud can explain the high - velocity material and the shock front created by it . for both stars , the h absorption features are separated into two velocity components . total h column densities toward and ori are and , respectively . when these values are compared to the column densities of h i , the fractions of h atoms bound in molecular form= 7\times 10^{-6}$ ] for and for . the rotation temperatures of the molecules with toward ori indicate that the gas is in the general vicinity of the stars that emit uv fluxes capable of rotationally pumping the molecules . for the strongest component of h toward ori , the pumping rate is lower and consistent with a general uv flux level in the plane of the galaxy .

astro-ph0003132

the absorption bands of the lyman and werner systems arising from the ground vibrational state of h@xmath3 provide a rich array of electronic transitions from different rotational levels . it is often true that these are the most conspicuous and plentiful features in the spectra of early - type stars below about 1100 ( morton 1975 ; jenkins et al . 1989 ; snow , allen , & polidan 1990 ) . in the interstellar medium , the rotational excitation of molecular hydrogen is driven by a number of processes , such as collisional excitation and de - excitation , optical pumping by starlight , and the introduction of newly formed molecules . for this reason , the relative rotational populations of h@xmath3 have been used to gain insights on the characteristic molecule formation ( and destruction ) rates , temperatures , and densities in the h@xmath3-bearing clouds ( spitzer & cochran 1973 ; spitzer et al . 1973 ; spitzer , cochran , & hirshfeld 1974 ; spitzer & zweibel 1974 ; jura 1975a , b ; morton & dinerstein 1976 ; shull 1977 , 1979 ; shull & york 1977 ) . the uv data provide critical insights on the character of h@xmath3 that are useful to other kinds of observations and our current understanding of how various environmental factors influence the rates of molecule formation and destruction ( shull & beckwith 1982 ) . in some cases , initial studies with the _ copernicus _ satellite revealed changes in the widths of the profiles for absorption features arising from levels with different values of the rotational quantum number @xmath5 ( spitzer & cochran 1973 ; spitzer , cochran , & hirshfeld 1974 ) . more exacting analyses of high quality recordings of the h@xmath3 profiles in several stars by spitzer & morton ( 1976 ) indicated that there were shifts in rotational temperatures for components with small changes in radial velocity , as one might expect for the gas behind shock fronts created by stellar winds . more recently , a significant improvement in the ability to study this phenomenon has come from the much better wavelength resolving power of the interstellar medium absorption profile spectrograph ( imaps ) ( jenkins et al . 1988 ) , which has flown on both sounding rockets and orbital missions ( jenkins 1993 , 1995 ; jenkins et al . 1996 ) . using data acquired by imaps on the orfeus - spas i mission launched on the sts-51 space shuttle flight in 1993 , jenkins & peimbert ( 1997 ) discovered that one of the velocity components of h@xmath3 in front of @xmath4 ori a showed small , progressive velocity offsets and profile broadenings as the rotation states increased from @xmath11 to 5 . they made the conjecture that this behavior could arise from h@xmath3 forming in the collapsing column of partially ionized , cooling gas behind a nearly stationary bow shock facing away from us . this shock front was thought to be created by the abrupt deceleration of wind - like material from the star ( or other stars nearby ) as it collided with a dense cloud , perhaps similar to ones seen elsewhere in orion that have comet - like tails facing away from prominent stars in the association ( bally et al . 1991 ; cernicharo et al . this interpretation was supported by spectroscopic evidence for low - ionization material at @xmath12 ( interpreted as the high speed material feeding the shock on its upstream side ) , and more highly ionized atoms at lower velocities ( immediate post - shock gas at @xmath13 ) . these velocities , relative to the observed pileup of cold material at about @xmath14 , implied a low compression ratio of 2.6 . jenkins & peimbert concluded that this outcome could be produced by an embedded magnetic field in the pre - shock material that is oriented nearly parallel to the flow . the motivation of the present investigation of h@xmath3 in front of the other two bright stars in the belt of the orion constellation , @xmath1 ori a and @xmath2 ori , was to gather independent evidence on the phenomenon studied by jenkins & peimbert ( 1997 ) for @xmath4 ori a. since there are indications that the three stars are within a large cavity containing x - ray emitting gas ( burrows et al . 1993 ) and enclosed by an expanding shell of cooler material ( reynolds & ogden 1979 ) powered by either stellar winds ( castor , mccray , & weaver 1975 ; weaver et al . 1977 ) or by one or more explosive events ( cowie , songaila , & york 1979 ) , it was not unreasonable to expect that we could witness additional manifestations of the unusual behavior of h@xmath3 . tantalizing indications that a phenomenon similar to that found by jenkins & peimbert was occurring on the line of sight to @xmath1 ori a were discussed by spitzer & morton ( 1976 ) . the two additional orion belt targets were observed by imaps on its second flight conducted in late 1996 aboard the orfeus - spas ii mission on sts-80 ( hurwitz et al . we present here the results of our investigation of the absorption features arising from h@xmath3 . the paper is organized as follows . the section on observations ( [ obs ] ) starts with a description of some fundamental properties of the two target stars ( [ tgts ] ) , @xmath1 and @xmath2 ori , along with @xmath4 ori a which is featured as a contrasting case in later discussions . the imaps instrument and some properties of the data that are relevant to our investigation are described in the next subsection ( [ inst ] ) , and this is followed by a brief description of the exposures that were taken ( [ exposures ] ) . the section on analysis ( [ analysis ] ) describes some general steps that were taken in the data reduction ( [ general_steps ] ) , outlines some special procedures to register the velocities of features from h@xmath3 in different @xmath5 levels ( [ vel_regs ] ) , shows the results of an investigation to check on the relative @xmath15 values of the weakest lines out of @xmath16 , 2 and 3 ( [ validation ] ) , and ends with a description of how we constructed profiles of column density vs. velocity for each @xmath5 level . the fundamental qualitative properties of the h@xmath3 features in the spectra of @xmath1 and @xmath2 ori are discussed in [ qualitative ] . the distributions of h@xmath3 over different @xmath5 levels are presented in [ distribution ] , using the concept of a rotation temperature over specific @xmath5 levels as a convenient way to describe the relative populations . we use these rotational temperatures to synthesize a few approximate conclusions about the strengths of the uv pumping fields . in the section that describes our interpretation of the results ( [ interpret ] ) , we note that the unusual shifts and broadening of the h@xmath3 profiles for different @xmath5 found by jenkins & peimbert ( 1997 ) seem to be correlated with the presence ( for @xmath4 ori a ) or absence ( for @xmath1 and @xmath2 ori ) of high velocity gas ( [ hi_vel ] ) or the possible proximity of @xmath4 ori to a dense cloud with a high rate of star formation inside ( [ structures ] ) . finally , in [ remarks ] we present a few remarks that pull together some of the concepts presented earlier and discuss their possible meanings . the two stars studied in this paper , @xmath1 ori a and @xmath2 ori , have a separation in the sky of 144 and are members of the b subgroup defined by blaauw ( 1964 ) within the orion ob1 association . later in this paper , we will be making comparisons with data recorded on an earlier imaps flight for @xmath4 ori a and reported by jenkins & peimbert ( 1997 ) . this star is nearly on the same line as that joining @xmath1 and @xmath2 , but it is 140 degrees from @xmath2 ( see fig . [ dust ] ) . for the benefit of readers who wish to compare the h@xmath3 results with other observations of interstellar material , the equatorial and [ gaustad ] because nearly all of the maps of interstellar structures in the current literature use this coordinate system . ] and galactic coordinates of the three stars are given in table [ coord ] . other important properties of these stars are also given in the table . all radial velocities reported in this paper are heliocentric ; the table lists the numerical values that must be added to the heliocentric velocities to obtain those in the local standard of rest . l c c c @xmath17&@xmath18&@xmath19&@xmath20@xmath21&@xmath22017567&@xmath2211269&@xmath22156333@xmath23&@xmath24&@xmath25&@xmath26@xmath27&@xmath22020044&@xmath22113562&@xmath22158030@xmath28&20386&20521&20645@xmath29&@xmath30&@xmath31&@xmath32@xmath33&@xmath34&@xmath35&@xmath36spectral type&b0@xmath37iii + o9@xmath37v&b0@xmath37iae&o9.5@xmath37ibe@xmath38 ( @xmath39)&144&91&140@xmath40&2.23&1.70&2.05@xmath41&0.09&0.05&0.05@xmath42$]&20.19&20.48 & 20.39 the orion ob1 b subgroup is at a distance modulus of approximately 7.8 ( 360 pc ) ( brown , de geus , & de zeeuw 1994 ) to 8.0 ( 400 pc ) ( warren & hesser 1978 ) . the parallaxes measured by hipparcos , @xmath43 for @xmath1 ori a and @xmath44 for @xmath2 ori ( perryman et al . 1997 ) , are consistent with this range of distances . both targets are bright , hot , and have a large projected rotation velocity ( see table [ coord ] ) , making them excellent uv light sources for viewing narrow interstellar absorption lines . since the instrument has no entrance slit , it is important to know if there are companion stars which could introduce contamination signals to the main spectra . there are two companions to @xmath1 ori a with wide separations from the visual primary . one is a type b2v with a separation of 52 ( hd36485 ) , but its visual magnitude is fainter by 4.6 , so its effect on the main spectrum is negligible . the other at 33 ( bd@xmath2200 983b ) has @xmath45 . two bright components of @xmath1 ori a were identified by heintz ( 1980 ) and later confirmed by the speckle interferometry observations ( mcalister & hendry 1982 ) . recent determinations of their separation yield a value 0267 ( lindegren et al . 1997 ; mason et al . 1998 ) , which is much smaller than the equivalent of our wavelength resolution projected onto the sky ( about 10 ) . finally , the visual primary is a partially eclipsing , spectroscopic binary ( koch & hrivnak 1981 ) with a period of 5.73 days ( harvey et al . our other target , @xmath2 ori , is a spectroscopic binary with also a visual companion that is 3 away and fainter by 8.7 magnitudes ( hoffleit & jaschek 1982 ) . detailed descriptions of the imaps instrument and how it performs have been presented by jenkins et al . ( 1988 , 1996 ) . briefly , imaps is an objective - grating echelle spectrograph whose optical configuration is very simple , consisting of a mechanical collimator ( to reject the off - axis light from stars other than the target ) , an echelle grating with a 63 blaze angle , a parabolic cross - disperser grating , and , at the focal point , an electron - bombarded , windowless , intensified ccd image sensor . the nominal wavelength coverage of @xmath46 is obtained in 4 sequential exposures that span the free spectral range of the echelle format at different angles of grating tilt . spectra taken at different angles were not of uniform quality because they had different deviations off the optimum blaze angle of the echelle grating . this effect , combined with much lower reflection efficiencies of the gratings below about 1000 , resulted in large changes in signal quality over different regions of the spectral format . nevertheless , we were able to get good selections for lines out of various @xmath5 levels ( tables [ h2linelist_dori ] and [ h2linelist_eori ] ) , even after defining a quality threshold for acceptance at a reasonably high level . the instrument is capable of attaining a wavelength resolution of order @xmath47 , but on this mission an out - of - focus condition resulted in @xmath48 and some overlap in energy between adjacent orders . the effective area of imaps for wavelengths greater than about 1000 ( where most of the h@xmath3 features that we used are located ) is about @xmath49 . there was a significant difference between the observations made on the orfeus - spas i and orfeus - spas ii flights , i.e. , those reported by jenkins & peimbert ( 1997 ) for @xmath4 ori a and those given in this paper , respectively . on the earlier flight , there was inadequate control of the humidity of the atmosphere inside the payload prior to launch . as a result , small water droplets formed on the hygroscopic ( and easily damaged ) k@xmath37br photocathode during the decompression and adiabatic cooling of the air inside the instrument that occurred during the initial phases of the shuttle launch . as shown by jenkins et al . ( 1996 ) , all of the data recorded on the 1993 flight exhibited a blotchy pattern of efficiency variations caused by the damage to the photocathode . within each blotch , the quantum efficiency was about 50% of the peak sensitivity in the unaffected regions . for the 1996 flight of imaps , more stringent precautions were taken to control the instrument s internal atmosphere prior to launch . as a result , the damage was prevented and photocathode response was very high , and uniformly so over the entire format . the photocathode damage on the first flight had both good and bad consequences . on the bad side , considerable effort during the reduction of the data was needed to make the required corrections for the variations . in some places , these corrections may not have been perfect , so unpredictable , systematic variations in intensity could arise . also , the overall efficiency of the detector was somewhat less than it would have been otherwise . on the good side , the damage to the photocathode had a beneficial effect of providing an accurate indication of the detector s geometrical mapping function . by correlating the apparent positions of the blotches in the data with a post - flight photograph of the damaged photocathode , accurate corrections for the distortion of the recorded spectral image could be implemented during data processing , with the end result of producing accurate wavelength determinations for all of the absorption features . for the data from the second flight reported on here , we had excellent looking raw data frames but no information about the geometric distortions ( aside from the bending of the orders information which is of no use in understanding perturbations in the wavelength scale ) . it was clear that we could not use the mapping function from the previous flight for the @xmath1 and @xmath2 ori data because there was a change in the stray magnetic fields from other systems ( which affect the electron trajectories in the detector ) from the first flight to the second . thus , in short , for the @xmath4 ori data extraordinary measures were needed to correct for photometric errors , while wavelength derivations were relatively easy . for the @xmath1 and @xmath2 ori data the reverse was true : obtaining relative photometric information of good quality was straightforward , but special efforts were needed ( described in [ vel_regs ] ) to obtain accurate radial velocities . the exposures for @xmath1 ori a with a total integration time of 6064 s were spread over 3 sessions : 1996 november 22 01:3902:15 ut , november 29 12:1312:54 ut , and december 3 21:3322:12 ut . the last session had all exposures taken at a single grating tilt to enhance the quality of the spectra at the very short wavelengths of the deuterium l@xmath1 and l@xmath2 transitions , used ultimately for the study of atomic d / h ( jenkins et al . 1999 ) . an additional set of exposures was taken when the spacecraft was in a minimum drag configuration [ to avoid a possible collision with another satellite ( hurwitz et al . 1998 ) ] , but the unusual aspect of the sun created an unexpected change in thermal conditions that shifted the cross - disperser coverage to much longer wavelengths , ones that were inappropriate for studying the h@xmath3 transitions . for @xmath2 ori , a contiguous set of exposures was taken during a single orbit over the time interval november 29 04:3305:20 ut , for a total exposure time of 2452 s. sample portions of the spectra covering two h@xmath3 absorption features are shown in fig . [ h2sample ] . to derive an approximate wavelength scale , dispersion constants for the various exposures were derived from the observed positions of h@xmath3 features . however , we did not assume that lines out of different @xmath5 levels or with very different strengths necessarily had the same apparent radial velocity . this analysis could be carried out because the h@xmath3 lines are so plentiful ( we used approximately twice as many lines as those selected for analysis and listed in tables [ h2linelist_dori ] and [ h2linelist_eori ] ) . we were able to create outcomes that were internally consistent to within about @xmath50 for @xmath1 ori a and @xmath51 for @xmath2 ori , as judged by the rms dispersion of apparent residual wavelength offsets for lines coming from any given @xmath5 level . the lower rms dispersion of @xmath52 obtained by jenkins & peimbert ( 1997 ) reflected the better control they had over their wavelength scales , as discussed in [ inst ] . an important objective of our investigation was to detect any shifts in velocity across different @xmath5 levels . for this purpose , we devised a more precise tuning of the wavelengths for just the h@xmath3 profiles , so that we could register them with better accuracy . this procedure is described in [ vel_regs ] below . before extracting the spectra from the echelle orders , a distortion correction procedure was invoked to remap the images so that the orders were approximately straight and horizontal . even though this was done , we found it necessary for the extraction program to determine the exact small residual vertical shifts ( in the cross - dispersion direction ) for each order as a function of horizontal position ( along the echelle dispersion direction ) . the distribution of energy across each order in the vertical direction varied with position , owing to small changes in aberrations in the image . thus , in addition to characterizing the changes in the centers of the orders , we also measured their widths and how they varied with position . after the orders had been characterized , we performed an optimum extraction of spectral information using the formulae given by jenkins et al . this method not only gives the best signal - to - noise ratio , but it also compensates for the small spillover of energy between adjacent orders . this is needed to prevent the appearance of spurious , ghost - like features that would ordinarily arise if a simple `` dumb - slit '' extraction were performed . for deriving the scattered light background levels , we obtained guidance from saturated atomic lines with flat bottoms . we also verified that our equivalent widths agreed reasonably well with those measured by spitzer , cochran & hirshfeld ( 1974 ) . in some cases , we had to make adjustments of up to 20% in the background levels to make some lines agree with others . to reduce the errors in radial velocities caused by image distortions , we found pairs of h@xmath3 features that were near each other on the echelle spectrum format , but which came from different @xmath5 levels . lines on a single order that were separated by about 0.1 ( or less ) or lines on adjacent orders ( or nearly so ) with a very small horizontal separation were deemed to be suitable for this purpose . the identities of these combinations are given in the endnotes of tables [ h2linelist_dori ] and [ h2linelist_eori ] . we then measured the differences in radial velocity and used them to define the relative shifts of the absorptions from different @xmath5 levels with respect to each other . the shifts averaged out to nearly @xmath14 for the features in both @xmath1 and @xmath2 ori , with the largest single deviation being about @xmath53 . in using the line pairs to define the velocity offsets , our working assumption was that image distortions over very small scales were absent , and thus the offsets were identical for both lines . to determine an absolute reference of the velocity scale , we compared the positions of the 5@xmath220 r(0 ) and 5@xmath220 r(1 ) lines to three prominent telluric features of o i out of excited fine - structure levels . these lines are rarely seen to arise from interstellar clouds , and thus they serve as excellent velocity calibration features in our spectra . jenkins & peimbert ( 1997 ) encountered some difficulty in obtaining consistent results for different combinations of lines of h@xmath3 arising from the weakest members of the lyman bands out of @xmath11 , 1 and 2 . it was possible that these inconsistencies could have arisen from either uncorrectable saturation effects , systematic errors in the background levels , errors in the adopted @xmath15-values , or some combination of the three . the amount of h@xmath3 in the direction of @xmath1 ori a is not very high . for this reason , the weak h@xmath3 transitions should have very small corrections for saturation , which leads to an opportunity to check on the reasonableness in the relative @xmath15-values . this is an important check to make , since these lines provide the only means for us to measure the strongest parts of the features out of @xmath11 , 1 and 2 for @xmath2 ori . figure [ weaklins ] shows the curves of growth for weakest lines out of @xmath16 and 2 in our spectrum of @xmath1 ori a. unfortunately , the weakest lines out of @xmath11 were too weak to give meaningful information . to estimate the amount of possible saturation in the h@xmath3 lines , we computed their expected behavior for a velocity profile that was identical to the superposition of components derived for na i in front @xmath1 ori a by welty , hobbs & kulkarni ( 1994 ) , based on their much higher resolution recordings of the d1 line . ( there is no reason to expect the h@xmath3 to exactly mimic na i , but this is the best example that we can draw upon . we do not want to use our own h@xmath3 data because there may be very narrow features that are not evident at our resolution . ) the reasonably good correspondence between our measurements and the curves in the figure indicates that the adopted relative @xmath15-values from the theoretical calculations of abgrall & roueff ( 1989 ) are reasonably accurate . after correcting individual line velocity offsets ( [ vel_regs ] ) , we derived the apparent differential column densities per unit velocity for each transition given by the expression @xmath54 where @xmath15 is the transition s @xmath15-value and @xmath55 is expressed in . the apparent absorption optical depth @xmath56 is defined in terms of the continuum level @xmath57 and measured intensity @xmath58 by the relation @xmath59 for each @xmath5 level , we consolidated the @xmath60 s for different lines into a single composite profile with the technique devised by jenkins & peimbert ( 1997 ) . not all of the lines that were visible in the spectrum were included in the composite . some were rejected because they had interference from other lines , were too close to the edge of an image , or fell below a threshold that we defined for the minimum acceptable signal - to - noise ratio . the lines that were chosen ( except where noted ) are listed in tables [ h2linelist_dori ] and [ h2linelist_eori ] . for @xmath2 ori , even the weakest lines out of @xmath11 , 1 and 2 were saturated . in the central portions of these lines , we implemented the correction technique developed by jenkins ( 1996 ) to compensate for the error between the apparent optical depth @xmath56 and a smoothed representation of the proper optical depth @xmath61 . the pairs of lines that were used in each case are identified in table [ h2linelist_eori ] . the other lines for these three @xmath5 levels were used to define the shoulders of the @xmath60 distributions . the usefulness of the correction technique is strongly driven by how well one knows the zero intensity level . if it is poorly known , there can be an ambiguity in the interpretation of lines that do not have @xmath56 scaling in proportion to @xmath62 . a poor scaling can arise either if the background levels are wrong or if the lines have unresolved , saturated substructures . unfortunately , there are no strong atomic lines in the vicinity of the weakest h@xmath3 lines . thus , for these cases we must acknowledge the existence of background uncertainties and their possible influence on the accuracy of the corrections for saturated absorptions for @xmath11 , 1 or 2 in @xmath2 ori . the increased errors are reflected in column 5 of table [ ns_epsilon ] . lines out of @xmath10 for @xmath2 ori and all of the h@xmath3 lines in the spectrum of @xmath1 ori a either showed good consistency in their @xmath60 profiles from strong to weak lines , or they were so weak that the distortions caused by saturated sub - structures should not be a problem . also , the background levels for these lines are probably more reliable than for those that we had to use for @xmath11 , 1 and 2 for @xmath2 ori because there were nearby , saturated atomic lines . r r r r 1@xmath220 r(0)&1092.195&0.802&894@xmath220 r(0)&1049.367&1.383&19 , 365@xmath220 r(0)&1036.545&1.447&46 , 787@xmath220 r(0)&1012.810&1.483&26 , 3610@xmath220 r(0)&981.437&1.314&17w 0@xmath220 r(0)&1008.552&1.647&22 0@xmath220 r(1)&1108.632&0.086&351@xmath220 p(1)&1094.052&0.340&341@xmath220 r(1)&1092.732&0.618&322@xmath220 p(1)&1078.927&0.624&19 , 252@xmath220 r(1)&1077.700&0.919&18 , 353@xmath220 p(1)&1064.606&0.805&223@xmath220 r(1)&1063.460&1.106&15 , 164@xmath220 p(1)&1051.033&0.902&25 , 51 , 714@xmath220 r(1)&1049.960&1.225&24 , 345@xmath220 p(1)&1038.157&0.956&23 , 395@xmath220 r(1)&1037.149&1.271&42 , 767@xmath220 p(1)&1014.325&0.960&14 , 217@xmath220 r(1)&1013.434&1.307&26 , 36 , 708@xmath220 p(1)&1003.294&0.931&21 , 238@xmath220 r(1)&1002.449&1.256&1810@xmath220 p(1)&982.834&0.825&14 , 21 , 3610@xmath220 r(1)&982.072&1.138&17w 0@xmath220 q(1)&1009.770&1.384&16w 0@xmath220 r(1)&1008.498&1.326&21w 1@xmath220 q(1)&986.796&1.551&14 , 29w 2@xmath220 q(1)&966.093&1.529&14 , 16 , 23 1@xmath220 p(2)&1096.438&0.420&282@xmath220 p(2)&1081.266&0.706&53 , 952@xmath220 r(2)&1079.226&0.866&323@xmath220 r(2)&1064.994&1.069&25 , 314@xmath220 r(2)&1051.498&1.168&30 , 40 , 805@xmath220 p(2)&1040.366&1.017&28 , 415@xmath220 r(2)&1038.689&1.221&447@xmath220 p(2)&1016.458&1.007&318@xmath220 r(2)&1003.982&1.232&3512@xmath220 p(2)&966.273&0.798&31w 0@xmath220 p(2)&1012.169&0.746&29w 0@xmath220 r(2)&1009.024&1.208&26 , 29 , 58w 2@xmath220 r(2)&965.791&1.490&25 2@xmath220 r(3)&1081.712&0.840&52 , 713@xmath220 p(3)&1070.141&0.910&23 , 333@xmath220 r(3)&1067.479&1.028&284@xmath220 p(3)&1056.472&1.006&35 , 42 , 694@xmath220 r(3)&1053.976&1.137&275@xmath220 p(3)&1043.502&1.060&295@xmath220 r(3)&1041.157&1.222&31 , 1037@xmath220 r(3)&1017.422&1.263&30 , 368@xmath220 p(3)&1008.383&1.004&219@xmath220 r(3)&995.970&1.229&24 , 5811@xmath220 p(3)&978.217&0.817&32w 0@xmath220 q(3)&1012.680&1.386&24 , 40w 1@xmath220 p(3)&991.378&1.075&39 r r r r 0@xmath220 r(0)&1108.127&0.275&341@xmath220 r(0)&1092.195&0.802&544@xmath220 r(0)&1049.367&1.383&325@xmath220 r(0)&1036.545&1.447&467@xmath220 r(0)&1012.810&1.483&34 0@xmath220 p(1)&1110.063&@xmath220.191&110@xmath220 r(1)&1108.632&0.086&412@xmath220 p(1)&1078.927&0.624&292@xmath220 r(1)&1077.700&0.919&18.13@xmath220 r(1)&1063.460&1.106&303@xmath220 p(1)&1064.606&0.805&464@xmath220 p(1)&1051.033&0.902&474@xmath220 r(1)&1049.960&1.225&395@xmath220 r(1)&1037.149&1.271&557@xmath220 r(1)&1013.434&1.307&458@xmath220 p(1)&1003.294&0.931&26w 2@xmath220 q(1)&966.093&1.529&17 0@xmath220 r(2)&1110.120&0.018&111@xmath220 p(2)&1096.438&0.420&432@xmath220 p(2)&1081.266&0.706&532@xmath220 r(2)&1079.226&0.866&343@xmath220 r(2)&1064.993&1.069&474@xmath220 p(2)&1053.284&0.982&224@xmath220 r(2)&1051.498&1.168&485@xmath220 p(2)&1040.366&1.017&445@xmath220 r(2)&1038.689&1.221&207@xmath220 p(2)&1016.458&1.007&268@xmath220 r(2)&1003.982&1.232&379@xmath220 p(2)&994.871&0.937&15w 0@xmath220 p(2)&1012.169&0.746&30w 0@xmath220 r(2)&1009.024&1.208&33 w 0@xmath220 q(2)&1010.938&1.385&18 1@xmath220 r(3)&1096.725&0.531&45 , 452@xmath220 r(3)&1081.712&0.840&524@xmath220 p(3)&1056.472&1.006&483@xmath220 r(3)&1067.479&1.028&20.54@xmath220 r(3)&1053.976&1.137&345@xmath220 p(3)&1043.502&1.060&22.15@xmath220 r(3)&1041.157&1.222&426@xmath220 p(3)&1031.192&1.055&447@xmath220 r(3)&1017.422&1.263&278@xmath220 r(3)&1006.411&1.207&9.59@xmath220 r(3)&995.970&1.229&34w 0@xmath220 q(3)&1012.680&1.386&28 3@xmath220 p(4)&1074.313&0.923&443@xmath220 r(4)&1070.900&1.012&434@xmath220 p(4)&1060.581&1.017&335@xmath220 r(4)&1044.542&1.195&326@xmath220 r(4)&1032.349&0.814&229@xmath220 r(4)&999.268&1.204&24 4@xmath220 r(5)&1061.697&1.140&3010@xmath220 p(5)&996.181&1.112&35w 0@xmath220 q(5)&1017.831&1.388&36 figure [ naplts ] shows logarithmic depictions of the apparent column densities ( from eq . [ n_a ] ) for all observable @xmath5 levels for the two stars . it is clear from the shapes of the profiles in fig . [ naplts ] that our results agree qualitatively with the conclusions derived by spitzer et al . ( 1974 ) from _ copernicus _ data , refined by a more meticulous study by spitzer & morton ( 1976 ) in the case of @xmath1 ori , that the general structures seen toward both @xmath1 and @xmath2 show some broadening with increasing @xmath5 . for instance , the main peak in the spectrum of @xmath1 ori a centered at @xmath63 is accompanied by a shoulder at more negative radial velocities that becomes more prominent with increasing @xmath5 . the shape of the peak itself is invariant with @xmath5 however [ measured widths ( fwhm ) of linear representations of the profiles shown in fig . [ naplts ] are 7 , 6 , 7 and @xmath64 for @xmath65 , respectively , with the higher result for @xmath66 probably being attributable to the stronger contribution from the shoulder ] . the broadening with increasing @xmath5 is less clear for @xmath2 ori . the most conspicuous effect here is the decrease in the apparent depth of the gap between the two peaks when going from @xmath67 to @xmath68 . for the h@xmath3 lines in the spectrum of @xmath2 ori recorded by _ copernicus _ , the two components are not resolved , but appear as a single peak with a slight asymmetry ( shull 1979 ) . the three orion belt stars seem to have in common a main peak centered at a velocity of approximately @xmath69 . it is strongest toward @xmath2 , of intermediate strength for @xmath4 [ compare our fig . [ naplts ] with figs . @xmath70 of jenkins & peimbert ( 1997 ) ] , and weakest in the spectrum of @xmath1 . maps of 21-cm emission over a velocity range + 18 to @xmath71 ( + 1 to @xmath72 lsr ) show a pronounced north - south ridge of atomic hydrogen to the east of @xmath4 ori a ( chromey , elmegreen , & elmegreen 1989 ; green 1991 ) . if the main h@xmath3 peak were closely coupled with this atomic hydrogen , we would have expected @xmath4 ori to show the largest column density , @xmath2 an intermediate value , and @xmath1 the smallest . instead , @xmath2 actually shows the most , with @xmath4 being second and @xmath1 third . at @xmath73 @xmath2 shows a peak that is almost as strong as its main peak at @xmath74 , while for @xmath1 this peak is weaker and spread over a large range of velocities . h@xmath3 at this velocity shows up very weakly in the spectrum of @xmath4 [ identified as component 2 by jenkins & peimbert ( 1997 ) ] . this behavior seems to agree with the 21-cm results : an isolated cloud of hydrogen , about 1 degree in diameter , is seen at the approximate location of @xmath2 in the maps of chromey , et al . ( 1989 ) and green & padman ( 1993 ) at @xmath75 to @xmath76 ( i.e. , consistent with our heliocentric velocity of @xmath73 ) . finally , the existence of a component at about @xmath77 seems to be unique to @xmath4 , but we could not identify in the 21-cm maps any distinctive atomic hydrogen cloud that was linked to it . while the h@xmath3 profiles toward @xmath1 and @xmath2 ori show some changes in shape as the @xmath5 values progress to higher values , there seems to be no manifestation of the distinctive broadening and shifts toward more negative velocities exhibited by the most negative velocity component [ component 1 of jenkins & peimbert ( 1997 ) ] seen in the direction of @xmath4 ori a. our original aim in this investigation was to study possible new examples of this phenomenon , but none were found . this conclusion is interesting in view of fact that these stars are separated by only about 10 pc in projection on the sky . as long as we are not being misled by the existence of a considerably larger separation longitudinally , our initial expectation was that the stellar winds from these stars probably create a single interstellar bubble that is common to all of them ( castor , mccray , & weaver 1975 ; weaver et al . other factors being equal , it was natural to expect that the interactions of these winds with the ambient medium should be similar in all three cases . the logarithms of the column densities @xmath78 over different velocity intervals and their errors are listed in tables [ ns_delta ] and [ ns_epsilon ] , along with results reported by spitzer et al . ( 1974 ) . for all @xmath5 levels in both stars , the two investigations agree within their combined errors . the atomic hydrogen column densities toward these stars are given in table [ coord ] . the fractions of h atoms bound into molecular form @xmath79 $ ] are @xmath80 for @xmath1 and @xmath9 for @xmath2 . c c c c c c c c 0&12.84&13.60&13.67&0.05&&13.54&@xmath811&13.60&14.44&14.50&0.05&&14.41&@xmath812&13.38&13.81&13.95&0.05&&13.87&@xmath813&13.58&13.73&13.96&0.05&&13.97&@xmath814&&&&&&@xmath82&5&&&&&&@xmath83&total&14.04&14.64&14.74&0.05&&14.66&@xmath81 c c c c c c c c 0&15.38&15.65 & 15.84&0.20&&16.08&@xmath841&15.53&15.75 & 15.96&0.20&&16.36&@xmath852&14.90&15.27 & 15.42&0.20&&15.34&@xmath843&14.22&14.42&14.64&0.05&&14.51&@xmath814&13.59&13.52&13.86&0.15&&13.82&@xmath815&13.29&13.10&13.50&@xmath86&&@xmath87&total&15.83&16.09&16.28&0.20&&16.57&@xmath85 figure [ h2temps ] shows a plot of @xmath88 for different components of the h@xmath3 absorption observed toward @xmath1 and @xmath2 ori , where @xmath89 is the column density in level @xmath5 , @xmath90 is the degeneracy of this level , and @xmath91 is its energy above the @xmath11 level . one can use the boltzmann equation to define an effective rotational temperature @xmath92 between two levels @xmath5 and @xmath93 @xmath94}\ ] ] which leads to the fact that @xmath92 is proportional to the negative inverse of the slope of a line joining the points for the respective levels in fig . [ h2temps ] . for @xmath2 ori , both components show nearly identical relative populations , but they seem to exhibit two values for @xmath92 . one applies to the interval @xmath65 and another is appropriate for @xmath95 . we have determined each by finding least - squares , linear fits to the points in fig . [ h2temps ] , and the outcomes are stated near these trend lines . the abrupt change in slope usually signifies a transition from temperatures that are coupled to collisions at low @xmath5 to ones that are dominated by optical pumping at higher @xmath5 . another effect that may contribute to the relationship at low @xmath5 is a diminution of the pumping rate at the lowest @xmath5 levels caused by the pumping lines becoming optically deep . the radiative decay rates at large @xmath5 are so fast that collisions have a practically negligible effect . for this reason , to a reasonable approximation we can use the ratios @xmath96 and @xmath97 as indicators of the density of pumping radiation ( when the radiation density increases , a large proportion of molecules in @xmath66 are pumped to the upper electronic level before they have a chance to decay to @xmath16 see v(a ) of spitzer & morton ( 1976 ) for a more complete discussion and representative examples ) . when we compare our ratios for @xmath2 ori with models presented in table 5 of spitzer & morton ( 1976 ) , we conclude that the pumping rate @xmath98 for component 1 and @xmath99 for component 2 . for comparison , jura ( 1974 ) calculated that @xmath100 in the general vicinity of the sun . the elevation in @xmath101 for h@xmath3 in front of @xmath2 ori seems consistent with what one would expect for gas that is in the general vicinity of the association but not especially close to one of the bright , early - type stars . we have less complete information on the distribution with @xmath5 for @xmath1 ori a because the overall column density of h@xmath3 is lower . since our values for @xmath89 are nearly the same as those of spitzer & morton ( 1976 ) , we adopt their interpretation for the conditions that lead to the observed relative populations . spitzer & morton found that for their component 3 of h@xmath3 for @xmath1 ori a , which approximately corresponds to our main peak , @xmath102 for @xmath11 and 1 , but @xmath100 for @xmath103 . this seems to indicate that the gas in the main component is rather distant from the orion stars . the lower pumping rate in the lowest @xmath5 levels probably arises from the diminution of flux from distant stars by molecular complexes unrelated to the one seen toward @xmath1 ori a , since in our spectrum only the strongest lines , such as those the werner 1@xmath220 q(1 ) and 2@xmath220 q(1 ) transitions , have central optical depths appreciably greater than about 2 . the suggested remoteness for this component of h@xmath3 from the orion association is an unexpected conclusion when one considers the fact that its velocity is very similar to one of the main components in front of @xmath2 ori ( and also @xmath4 ori a ) . perhaps the molecules in front of @xmath1 ori a are heavily shielded from the pumping radiation by other parts of the same gas complex in the vicinity of the orion association . for the shoulder component toward @xmath1 ori a , the upper limit for @xmath104 found by combining our results with those of spitzer et al . ( 1974 ) indicates that @xmath105 . in the limit of a low density , the high rotation temperature @xmath106k for @xmath65 would be consistent with a considerably larger value of @xmath101 , so perhaps the local kinetic temperature is about 850k and the density is large enough to have collisions couple @xmath92 to this temperature . we turn now to the question of what we can learn from the absence of the broadening and velocity shifts in the spectra of @xmath1 and @xmath2 ori that jenkins & peimbert ( 1997 ) found in the most negative velocity h@xmath3 feature for @xmath4 ori a. in particular , are there other differences between the two stars featured here and @xmath4 ori a that are noteworthy and might give us further insights on the phenomenon ? does this information support or refute the interpretation that it arises from h@xmath3 forming in a cooling zone behind a bow shock , one that could be created by the collision of a gas flow coming toward us and a foreground obstruction , as suggested by jenkins & peimbert ( 1997 ) ? in addressing the question just posed , we examined two types of evidence . first , we searched for differences revealed by other spectral features for these stars ( [ hi_vel ] ) . second , we looked for relationships of the stars with structural features of the interstellar medium detected by other means ( [ structures ] ) . in both cases , we noted certain properties that set @xmath4 ori a apart from @xmath1 and @xmath2 ori . figures [ ciispec ] and [ niispec ] show imaps recordings of the strong resonance lines of the singly ionized forms of c and n , respectively , for the three orion belt stars . the most striking difference that sets @xmath4 apart from @xmath1 and @xmath2 is the absorption from material that is moving at a radial velocity that ranges between @xmath107 and @xmath108 . in the spectrum of @xmath4 ori a , similar components at these velocities are seen in c iii , n iii , si iii and s iii ( cowie , songaila , & york 1979 ) and al iii ( medium - resolution ghrs spectrum in the _ hst _ archive ) , but are not visible in the strong transitions of o i and n i. the interpretation of jenkins & peimbert ( 1997 ) regards this as the upstream material that feeds the shock front that is responsible for the h@xmath3 component at @xmath77 which shows the unusual behavior with increasing @xmath5 . the correlation of the presence or absence of the high velocity gas with the phenomenon noted by jenkins & peimbert seems to favor their shock interpretation over completely unrelated alternatives . the interstellar material in the general vicinity of the orion ob1 association contains a number of well - studied gas complexes and prominent emission regions [ see goudis ( 1982 ) for a detailed summary ] . the lines of sight to @xmath2 and @xmath1 seem unremarkable , but that toward @xmath4 ori a points in a direction that is very close to the densest concentration of material within the orion b molecular cloud complex . this cloud has a large visual extinction and appears as entry number 1630 in the catalog of dark nebulae compiled by lynds ( 1962 ) . it has a significant internal density , and it contains a large number of very young , but highly obscured stars ( strom , strom , & vrba 1976 ) . the striking difference between the direction toward @xmath4 and those toward the other two stars is shown in a map of thermal emission from dust ( schlegel , finkbeiner , & davis 1998 ) recorded by the iras satellite , depicted in fig . [ dust ] , and likewise in maps of co emission ( kutner et al . 1977 ; maddalena et al . 1986 ) . maps of 21-cm emission in this region of orion show no indication of this strong contrast ( chromey , elmegreen , & elmegreen 1989 ; green 1991 ; green & padman 1993 ) , probably because the emission from this cold , highly condensed cloud is very badly saturated . could lynds 1630 and its bright edge ic 434 ( see fig . [ gaustad ] ) on the side toward @xmath4 and @xmath109 ori ( pottasch 1956 ) be the obstruction proposed by jenkins & peimbert ( 1997 ) that is needed to create a stationary bow shock in the stellar wind material ? the plausibility of this idea very much depends on whether or not lynds 1630 is in front of or behind @xmath4 ori . at first glance , one might suppose that the cloud must be in front , since the main body of the cloud appears much darker in h@xmath110 emission than the edge toward @xmath4 ori . also , the existence of the horsehead nebula as a dark silhouette against the h@xmath110 emission is an additional clue , although there is a reasonable chance that the background excitation at this particular location could arise from radiation from @xmath109 ori ( which might be more distant ) rather than @xmath4 ori . the bright , foreground h ii region ngc 2024 [ also known as the orion b radio source ( barnes et al . 1989 ; van der werf et al . 1993 ) ] is powered by uv radiation from one or more highly obscured star(s ) embedded within it , not by @xmath4 ori ( grasdelen 1974 ) . on closer inspection , however , we conclude that the indications offered by the structure of the most easily evident h@xmath110 emission could be misleading . if we disregard the very bright patch of h@xmath110 emission associated with the edges of the lynds 1630 cloud and consider the fainter , more broadly distributed emission in the general region , a different picture emerges . nearly all of the volume inside barnard s loop ( pickering 1890 ; barnard 1894 ) is filled with a faint glow created by the ionizing radiation from the orion association ( sivan 1974 ) . [ gaustad ] shows a contour map of h@xmath110 emission in the vicinity of @xmath4 and @xmath2 ori , well inside the perimeter of barnard s loop . this map was constructed from the survey of gaustad et al . ( 1999 ) , and it clearly shows that the brightness level at the locations where there is significant obscuration from the cloud ( but somewhat removed from ngc 2024 ) is comparable to the flux from directions approximately midway between @xmath4 and @xmath2 ori , well beyond the cloud s edge to the northwest . [ an earlier study of h@xmath110 emission in this vicinity by vidal ( 1980 ) also shows this result , but not as clearly . ] on the one hand , this finding suggests that the lynds 1630 cloud is behind the belt stars , if they were responsible for ionizing hydrogen and producing h@xmath110 emission in front of the cloud or right on the cloud s front edge . on the other hand , we could argue that one or more stars other than @xmath4 ori , such as @xmath109 ori ( spectral type o9.5@xmath37v ) , are responsible for most of the emission and are closer to us than @xmath4 ori . unfortunately , spectroscopic or hipparcos parallaxes are not accurate enough to resolve this issue . if , by contrast , we had found that the h@xmath110 emission were darker in front of lynds 1630 than in directions close to @xmath2 ori , we might have had a more positive indication that the cloud was well in front of the belt stars and appears as a silhouette . other evidence that disfavors lynds 1630 being in the foreground , but again not conclusively , arises from the distribution of dust shown in fig . the amount of infrared 100@xmath111 emission from the cloud s dust in the exact direction of @xmath4 ori leads to an expected b@xmath22v color excess that seems to be far greater than the star s actual @xmath41 = 0.05 . a hires processing of the 100@xmath111 image ( see inset of fig . [ dust ] ) shows that we are not being misled by the smoothing in the representation generated by schlegel , et al . ( 1998 ) combined with the star s proximity to a sharp edge of the cloud . nevertheless , the gas could be clumpy on scales even smaller than the hires picture , and the line of sight to @xmath4 ori could miss the clumps . the indications that lynds 1630 may be behind @xmath4 ori cast doubt on the interpretation that it is a conspicuous barrier for the stellar wind from the star . nevertheless , we still should not overlook the possibility that its presence in the same general direction of the star may still be significant . the cloud contains very dense concentrations of molecular material , as revealed by high - resolution maps of molecular emission by lada , bally & stark ( 1991 ) ( cs ) , bally , langer & liu ( 1991 ) ( @xmath112 ) , and stark & bally ( 1982 ) ( @xmath113 ) . it is often true that clouds containing compact molecular regions are sites of star formation , and the orion b cloud is no exception ( bence & padman 1996 ; chernin 1996 ; launhardt et al . 1996 ; zinnecker , mccaughrean , & rayner 1996 ; sandell et al . one product of star formation is the creation of high speed , bipolar flows of gas ( bally 1982 ; genzel & downes 1982 ; knigl 1996 ) with velocities that have about the same magnitude as the component in front of @xmath4 ori a shown in [ hi_vel ] ( eislffel 1996 ; hirth , mundt , & solf 1996 ) . if such flows penetrate beyond the edge of the dense cloud and into a low - density medium , they could extend over large distances ( reipurth , bally , & devine 1997 ) , perhaps even farther than the remote , multiple shocks ( bence & padman 1996 ) that can be detected in emission . still another possibility is that the high speed gas arises from the breakout of material from an embedded , recently formed h ii region ( tenorio - tagle 1979 ; sandford , whitaker , & klein 1982 ; subrahmanyan et al . 1997 ) . in [ hi_vel ] we pointed out that @xmath1 and @xmath2 ori do not show gas at high negative velocities , while @xmath4 ori does . we can investigate this phenomenon further by examining the outcome of a survey of uv absorption features in b - type stars carried out by shore ( 1982 ) and sonneborn , et al . ( 1988 ) . in fig . [ dust ] we indicate the positions of these stars and with different symbols show whether they do ( * ) or do not ( + ) exhibit features from singly - charged ions at velocities equal to or more negative than @xmath114 . it seems clear that stars in the vicinity of lynds 1630 usually show the high - velocity features , while those that are somewhat removed in a lateral direction , such as the targets in the general vicinity of @xmath1 and @xmath2 ori , do not . this outcome suggests the possibility that the dense molecular cloud may be the source of high velocity gas . while this may be so , we should not overlook the fact that high velocity features are seen elsewhere in the orion region ( cohn & york 1977 ; cowie , songaila , & york 1979 ) , with some of them at locations outside the coverage of fig . [ dust ] . with this in mind , we could equally well propose that a more general , coherent expansion of gas from a larger scale event is taking place . in this picture , the appearances of high velocity features in fig . [ dust ] could arise from just random fragments of enhanced density in the expanding material . in studying the spectra of @xmath1 and @xmath2 ori , we had high expectations of finding duplications of the unusual findings of jenkins & peimbert ( 1997 ) for one of the h@xmath3 components toward @xmath4 ori a. they reported that the profile became broader and shifted toward more negative radial velocities as the rotational excitation increased from @xmath11 to 5 . our optimism about seeing more examples of this phenomenon was based on the premise that , in jenkins & peimbert s interpretation of a bow shock as its source , the stellar wind that may have created this shock front was probably a composite phenomenon common to all three stars in the orion belt . the apparent lack of this effect for @xmath1 and @xmath2 ori deprives us of the chance to see additional manifestations of it , but opens the way for us to examine circumstantial evidence that could give us more insight on the validity of the shock interpretation . aside from not showing the strange behavior of h@xmath3 , @xmath1 and @xmath2 ori differ from @xmath4 ori in two other respects that could have a bearing on the shock interpretation . first , @xmath4 shows absorption by singly and doubly ionized species at a velocity of about @xmath115 , while the others do not . this high velocity material was a good candidate for the upstream gas that could have created the bow shock . second , in projection @xmath4 ori is very near the edge of the dense cloud lynds 1630 , an accumulation of gas that exhibits a bright edge of ionized material ( ic 434 ) on the side towards @xmath4 ori and holds within it many compact regions of star formation . the sight lines to @xmath1 and @xmath2 ori are , by contrast , unremarkable . the real source of the gas at @xmath115 that is seen in the spectrum of @xmath4 ori is uncertain . it may be some byproduct of the winds from the stars , but it is not easy to understand why similar components are not seen toward @xmath1 and @xmath2 ori . an interesting alternative is the possibility that the @xmath116 component has nothing to do with stellar winds from the belt stars . the complex of very dense gas near the @xmath4 ori line of sight is known for its plentiful star forming regions . such regions typically discharge gas at high velocities , starting with flows that are highly collimated . perhaps one such flow could escape from its gas envelope and be responsible for the high velocity component that shows up for @xmath4 ori in figs . [ ciispec ] and [ niispec ] . a well - directed flow could provide a reasonable way to explain the absence in the spectra of @xmath1 and @xmath2 ori of any feature analogous to the @xmath115 component exhibited by @xmath4 ori a. new insights on the problem discussed in this paper may arise from observations by the far ultraviolet spectroscopic explorer ( fuse ) , which has the capability of examining a very much broader sample of cases in its lifetime . however its lower wavelength resolution will make it harder to study the detailed structures in the h@xmath3 profiles , and productive research in this area may require the existence of even more extreme cases of profile variations with @xmath5 than those shown toward @xmath4 ori a. the observations reported in this paper were recorded by imaps during its second orbital flight , which was carried out on the orfeus - spas ii mission in 1996 . this effort was a joint undertaking of the us and german space agencies , nasa and dara . the successful execution of our observations was the product of efforts over many years by engineering teams at princeton university observatory , ball aerospace systems group ( the industrial subcontractor for the imaps instrument ) and daimler - benz aerospace ( the german firm that built the astro - spas spacecraft and conducted mission operations ) . contributions to the success of imaps also came from the generous efforts by members of the optics branch of the nasa goddard space flight center ( grating coatings and testing ) and from o. h. w. siegmund and s. r. jelinsky at the berkeley space sciences laboratory ( deposition of the photocathode material ) . we are grateful b. t. draine for his comments on an early draft of the paper and to j. gaustad for supplying the digital representation of the h@xmath110 image that was the basis of our fig . [ gaustad ] . this research was supported by nasa grant nag5@xmath22616 to princeton university . bally , j. , langer , w. d. , wilson , r. w. , stark , a. a. , & pound , m. w. 1991 , in fragmentation of molecular clouds and star formation , ed . e. falgarone , f. boulanger , & g. duvert ( dordrecht : kluwer ) , p. 11 gaustad , j. e. , rosing , w. , mccullough , p. , & van buren , d. 1999 , in amateur - professional partnerships in astronomical education and research , ed . j. percy ( san francisco : astr . pacific ) , p. in press jenkins , e. b. , joseph , c. l. , long , d. , zucchino , p. m. , carruthers , g. r. , bottema , m. , & delamere , w. a. 1988 , in ultraviolet technology ii , ed . r. e. huffman ( bellingham : the international society for optical engineering ) , p. 213 lindegren , l. , mignard , f. , sderhjelm , s. , badiali , m. , bernstein , h. h. , lampens , p. , pannunzio , r. , arenou , f. , bernacca , p. l. , falin , j. l. , froeschl , m. , kovalevsky , j. , martin , c. , perryman , m. a. c. , & wielen , r. 1997 , , 323 , l53 perryman , m. a. c. , lindegren , l. , kovalevsky , j. , hog , e. , bastian , u. , bernacca , p. l. , crze , m. , donati , f. , grenon , m. , grewing , m. , van leeuwen , f. , van der marel , h. , mignard , f. , murray , c. a. , le poole , r. s. , schrijver , h. , turon , c. , arenou , f. , froeschl , m. , & petersen , c. s. 1997 , , 323 , l49 sandell , g. , avery , l. w. , baas , f. , coulson , i. , dent , w. r. f. , friberg , p. , gear , w. p. k. , greaves , j. , holland , w. , jenness , t. , jewell , p. , lightfoot , j. , matthews , h. e. , moriarty - schieven , g. , prestage , r. , robson , e. i. , stevens , j. , tilanus , r. p. j. , & watt , g. d. 1999 , , 519 , 236

we present the results of a multi - scale analysis of tec fluctuations using a roughly five - hour observation of the bright radio source virgo a with the very large array ( vla ) at 74 mhz in its b configuration . our analysis combines data sensitive to fine - scale structure ( km and tecu in amplitude ) along the line of sight to virgo a as well as larger structures ( hundreds of km ) observed using several ( ) moderately bright sources in the field of view . the observations span a time period from midnight to dawn local time during 1 march 2001 . several groups of magnetic eastward directed ( med ) , wavelike disturbances were identified and determined to be located within the plasmasphere ( ) . we have also detected evidence of non - wavelike structures associated with these disturbances which are propagating roughly toward magnetic north . these likely represent a non - uniform density flow from the plasmasphere toward the nighttime ionosphere . ae and indices and gps tec data indicate that during the observations , there were low levels of geomagnetic activity accompanied by somewhat localized depletions in ionospheric density . thus , the observed plasmaspheric disturbance may be part of a flow triggered by these ionospheric depletions , likely associated with forcing from the lower atmosphere which is typically more prominent during quiet geomagnetic conditions . in addition , we have also observed several roughly westward directed and southeast directed waves located within the ionosphere . they are coincident in time with the plasmaspheric disturbances and may be related to the deposition of material onto the nighttime ionosphere .

1205.1426

because of the negative impact on satellite - based communication , satellite operation , and gps precision , disturbances within both the ionosphere and plasmasphere as well as interactions between the two have generated much interest and investigations . these disturbances run the gamut from geomagnetic storms , substorms , streamers , ionospheric scintillations , storm enhanced densities , and traveling ionospheric disturbances ( tids ) , to name a few . consequently , detailed explorations of these phenomena have been conducted using a wide variety of instruments including satellites , magnetometers , ionosondes , radar arrays , and gps and other satellite beacon receivers . within this field of study , radio - frequency interferometers are a relatively underused resource . designed to operate as interferometric telescopes , such arrays are affected by the ionosphere and plasmasphere in the same way as ground - base satellite receivers . consequently , much effort has been devoted to compensating for these effects , especially within the vhf regime where the impact is much greater . these corrections contain information about density fluctuations within the ionosphere and/or plasmasphere , being sensitive to the gradient of the total electron content ( tec ) rather than the tec itself ( see , e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) . when operating within the vhf regime and observing a bright source , interferometers can measure differences in tec , or @xmath5 between pairs of antennas to a precision of nearly @xmath6 tecu ( 1 tecu@xmath7 m@xmath8 ) and tec gradients to a precision of @xmath9 tecu km@xmath10 ( see , e.g. , * ? ? ? because of its relatively stable electronics , large collecting area , and available vhf system ( bands at 74 and 330 mhz ) , the very large array ( vla ; located at @xmath11 n , @xmath12 w ) has been used on a somewhat limited basis to study the ionosphere and plasmasphere . seminal investigations by @xcite and @xcite used 330 mhz observations with the vla to identify and explore several wave - like phenomena identified within the data . follow - up work done by @xcite showed that a new class of waves found by @xcite were located within the plasmasphere . they were shown to essentially be density fluctuations formed perpendicular to magnetic field lines at mcilwain @xmath13-parameters between about 1.8 and 4 which co - rotated with the plasmasphere with a moderate amount of ( mostly ) westward convection . in addition , a statistical description of the typical , predominantly turbulent behavior of the ionosphere over the vla using a large 74 mhz survey @xcite was presented by @xcite . recently , @xcite detailed new spectral analysis techniques that enabled the simultaneous identification and characterization of wave - like disturbances within vla vhf data from scales of a few kilometers up to more than 100 km . here , we detail a new interferometer - based approach that builds upon this previous work of @xcite , @xcite , and @xcite by using a @xmath14 hour 74 mhz observation of a very bright vhf source as well as several moderately bright sources spread over a @xmath15 area on the sky ( or , about 80 km at an altitude of 300 km ) . this new technique allows for a more thorough evaluation of the environment of fluctuations on medium to large scales than can be obtained with observations of a single source . this multi - source method also has better precision than would be possible with a similar gps - based approach . when combined with simultaneous data for a single bright source in the field of view , we demonstrate that a unique characterization of both plasmaspheric and ionospheric wave properties from scales of a few to hundreds of kilometers can be attained . in 2 , we describe the observations and the processing / analysis of the data with more detailed descriptions included as appendices . in 3 , we detail the properties of plasmaspheric disturbances within the data . in 4 , we similarly describe ionospheric disturbances contained within the data . we include additional contemporaneous ae index and gps data while interpreting the results in 5 and summarize in 6 . the data - set we have chosen to analyze ( vla program number ad441 ) is a roughly five hour long observation of the bright source virgo a , or `` vir a '' ( also commonly known as 3c274 ) with the vla at 74 mhz . vir a is among the brightest cosmic vhf sources with a flux density at 74 mhz of about 1,700 jy ( 1 jy = @xmath16 w m@xmath8 hz@xmath10 ) . the observations were taken in the post - midnight hours local time of 1 march 2001 starting at 06:08 ut using the vla `` 4-band '' system @xcite . while observations with this system are generically referred to as 74 mhz observations , the central frequency was more precisely 73.8 mhz with a 1.5 mhz wide bandpass . in this case , dual polarization was also used . the @xmath4 index over this time period was 12 and the solar activity was fairly average ( @xmath17 sfu ) . the vla was in its b configuration with the antennas spanning an area about 11 km in diameter and laid out according to the diagram displayed in fig . [ layout ] . the calibration , processing , and spectral analysis of the observations are described in detail in appendices a b . briefly , we used techniques developed by @xcite to convert the observations of vir a to measurements of the tec gradient time series for each vla antenna . this conversion relies on a technique known as `` self - calibration '' which is discussed in detail in @xcite . hereafter , use of the term self - calibration in reference to any results presented implies that they were based on these measurements . two fourier - based approachs were used to convert the tec gradient time series ( temporal sampling of 6.59 seconds ) into mean spatial frequency spectra to identify disturbances that may be present within the data . the first technique is described in detail by @xcite which uses the phases of the temporal fourier - transforms to determine a spatial frequency , @xmath18 , and direction for each temporal mode . because it is based solely on phase information , we refer to this technique as `` phase - based . '' within the phase - based technique , the determined direction of each mode is that of the line along which the phases of the fourier transforms of the two tec gradient components ( i.e. , north - south and east - west ) agree . this gives this approach the ability to sense the direction of motion for non - wavelike disturbances ( see appendix b for a more detailed discussion ) . the second method uses a straightforward discrete fourier transform ( dft ) of the antenna - based temporal fourier transforms to map the spectral power as a function of north - south and east - west spatial frequency . this `` dft - based '' method also employs an algorithm designed to mitigate the effects of the rather poorly behaved impulse response of the y - shaped vla within this procedure ( again , see appendix b for more discussion ) . because of the low spectral resolution of the relatively compact b - configuration vla ( 11-km across ) , this technique still only allows one to reliably estimate a single spatial frequency and direction for each temporal mode ( see appendix b ) . since it is a more standard spectral approach , the dft - based technique is mostly sensitive to wavelike disturbances for which it measures wavefront orientations and not necessarily directions of motion . this makes the two techniques complementary when both wavelike and non - wavelike disturbances are present . we have applied both techniques to the vir a self - calibration data to produce two - dimensional spectral maps averaged within one - half - hour windows as functions of spatial frequency , @xmath18 , and azimuth angle measured clockwise from north . to aid with the identification of small - scale , small amplitude disturbances , we have normalized these spectra by dividing each @xmath18 bin by the total power over all azimuth angles within that bin . for convenience , we will refer to these specially normalized spectra as `` azimuth spectra . '' the azimuth spectra derived from the vir a data are displayed in fig . [ paimg ] . the results for the phase - based technique are shown as greyscale images and the dft - based results are plotted as color - coded contours . the dft - based azimuth spectra show several prominent features directed roughly toward magnetic east ( azimuth of @xmath19 for the vla ) between local times of 01:30 and 03:00 . their orientation implies that these disturbances are aligned along magnetic field lines . they also have relatively large speeds , mostly spanning the range 200400 m s@xmath10 . these properties imply that they are likely located within the plasmasphere and are the same magnetic eastward directed ( med ) disturbances discovered by @xcite with the vla . these were described in more detail by @xcite as relatively long plasmaspheric irregularities stretched along field lines . after 03:00 , westward and southwest directed waves begin to dominate the dft - based azimuth spectra . the phase - based azimuth spectra also show detections of the same phenomena , but with notably different characteristics . specifically , the med waves are broken into several groups and the estimated azimuths for each group seem to `` drift '' along linear structures toward the north with decreasing spatial frequency . this may indicate that the med waves are associated with a significant amount of non - wavelike structure which is moving in a different direction , possibly along the field lines toward magnetic north . we will discuss these results and the properties of the med waves in more detail in 3 . in contrast , the phase - based and dft - based spectra agree rather well for the westward and southwestward directed phenomena seen after 03:00 local time , indicating they are likely dominated by wavelike structures . in addition to vir a , several other moderately bright sources were also visible with the @xmath20-wide 74 mhz vla field of view . in particular , we identified 29 sources from the vla low frequency sky survey @xcite which could be used to measure the tec gradient as a function of time along the line of sight to each source , using a technique called `` field - based calibration '' ( see * ? ? ? * and appendix a ) . we will use the term field - based calibration to refer to any results that were based on this analysis . since the sources used within this technique were not nearly as bright as vir a ( between 2.5 and 10 jy versus 1,700 jy for vir a ) , each of them could not be used to perform self - calibration . this is because the relative contributions of other sources to the measured visibilities is significant for fainter sources . for a source as bright as vir a , these contributions are negligible . however , for a moderately bright source , one can measure the effect of the ionosphere along the line of sight by making images of the source on relatively short time scales and measuring how its observed position changes with time ( see appendix a and * ? ? ? * for more details ) . these position shifts yield measurements of the mean tec gradient over the array toward a particular source , implying that these gradients have been smoothed with a roughly 11-km wide circular kernel ( i.e. , the diameter of the vla b configuration ) . in order to have a high enough signal - to - noise ratio within each image to detect the source , each image used one minute of data , implying that the data were effectively smoothed in time as well . this spatial and temporal smoothing reduced the sensitivity of these measurements to small - scale , small amplitude fluctuations . specifically , the measured spectral power drops to one half the actual power at a spatial frequency of 0.04 km@xmath10 ( or , a wavelength of 25 km ) and/or at a temporal frequency of 26.5 hr@xmath10 ( or , a period of 2.26 minutes ) . the spectral response drops to zero at a wavelength of 11 km and/or a period of 1 minute . when projected onto ionospheric heights ( here , we used 300 km ; see appendix a ) , the pierce - points for these sources spanned nearly 100 km . this is illustrated in fig . [ calib ] where we have plotted the projection of the vla antennas onto a height of 300 km centered on each of the 29 sources . with such a distribution of pierce - points , one can perform a straightforward spectral analysis by applying a fourier transform in three - dimensions , one temporal and two spatial . while such power spectrum `` cubes '' will be less sensitive to small - scale structures , they can be used to separate different fourier modes with better fidelity than either the phase - based or dft - based techniques applies to the self - calibration data . this is because both of these methods rely on the conversion of temporal modes into spatial ones to enhance spectral resolution , and if there are several waves with the same temporal frequency but different spatial scales , only the strongest of the waves will be detected . thus , these two methods are quite complementary . we have developed a procedure for using the 29 moderately bright sources to produce power spectrum cubes which is detailed in appendix b. we used this method to yield spectral cubes within the same half - hour intervals as were used for the azimuth spectra displayed in fig . [ paimg ] for the self - calibration data . to display the wealth of information contained within these cubes , we have binned them by temporal frequency into four `` channels , '' with frequency ranges of 06 , 613 , 1321 , and 2130 hr@xmath10 . we have plotted the mean power spectra within each half - hour interval for each channel in fig . [ chmap1][chmap4 ] as greyscale contours . for each of these `` channel maps , '' we have plotted the shape of the impulse response in the lower left corner as an ellipse that represents the full width at half maximum ( fwhm ) of the impulse response function ( irf ) . we have also displayed maps of spectral power derived from the dft - based technique applied to the self - calibration data as false - color images with the channel maps . each channel map shows varying levels of significant ( i.e. , @xmath21 the size of the irf ) structure that , in some cases , overlap with features seen within the self - calibration - based maps . these structures will be discussed further in 4 . as demonstrated in 2.2.2 , our analysis of the vir a - based self - calibration data has revealed a group of med waves similar to those detected by @xcite and established by @xcite to be plasmaspheric irregularities . we show the properties of these med waves in more detail in fig . [ med ] . for each of the three half - hour intervals between 01:30 and 03:00 local time , we have plotted in the left panels the spectral power versus spatial frequency projected along magnetic east , @xmath22 , for temporal modes with azimuths between @xmath23 and @xmath24 and speeds @xmath25 km hr@xmath10 ( 556 m s@xmath10 ) . the latter criterion was used to exclude modes which were essentially noise . most of the med waves have wavelengths between 25 and 100 km with a strong concentration near roughly 30 km . this is very consistent with the range in wavelengths measured for similar plasmaspheric structures by @xcite . for the first two half - hour intervals , there also appears to be a distinct group of relatively large waves ( wavelength @xmath26 km ) which have spectral power nearly an order of magnitude stronger than the bulk of the med waves . for the same temporal modes , we have also plotted spectral power as a function of speed projected along magnetic east , @xmath27 , in the middle panels of fig . as noted before , most of the speeds are relatively high , ranging from about 150500 m s@xmath10 . this is entirely consistent with the trace speeds observed for med waves detected previously with the vla @xcite . assuming that the vast majority of the measured speed of any temporal mode is simply the projection of the co - rotation speed of the plasmasphere along the vector normal to the wavefronts ( i.e. , magnetic east ) , we can estimate the heights and l - shells of these detected med waves . the heights and l - shells corresponding to observed speeds of 100 , 300 , and 500 m s@xmath10 are labeled in the middle panels of fig . they show that most of the med waves are consistent with plasmaspheric heights between 2000 and 7000 km with a prominent peak at about 3300 km for the middle time interval . for the geomagnetic latitudes of the vla line of sight at these heights and times , this peak height corresponds to l@xmath28 . the median l - shell for the vla - detected med waves detailed in @xcite was 2.2 , quite consistent with this estimate . however , we note that the separate group of large / strong waves seen within the first two time intervals have speeds @xmath29 m s@xmath10 . these are therefore likely ionospheric disturbances whose exact heights are less certain due to the influence of other factors on their motion such as the wind within the @xmath30-region . as stated in 2.2.2 , our data are entirely consistent with the model of @xcite and @xcite for med waves that describes them as plasmaspheric irregularities stretched out along field lines . to estimate the zonal angular sizes of these disturbances as they map to lower altitudes , we have used the estimated heights and spatial frequencies to compute angular spatial frequencies , @xmath31 , in the magnetic eastward direction . the power as a function of @xmath31 is plotted in the right panels of fig . [ med ] . from these plots , one can see that most of the structures occupy scales between about @xmath32 and @xmath33 with noticeable peaks in the second and third half - hour intervals corresponding to scales of about @xmath34 and @xmath35 . as noted in 2.2.2 , the results from the phase - based spectral analysis technique show evidence for non - wavelike structure associated with the med waves . this is illustrated by the approximately linear features seen near magnetic east in fig . [ paimg ] . we have postulated that this may indicate movement of non - wavelike structures northward of magnetic east . since these structures are likely associated with the plasmaspheric med waves , the most logical direction of motion for such structures would be toward magnetic north along the field lines . to demonstrate that this provides a good description of the phase - based results , we have constructed a `` toy '' model of these irregularities . this model consisted of @xmath36 gaussian tec fluctuations with randomly generated sizes with a uniform distribution ranging from 15 to 40 km ( here , `` size '' is @xmath37 , where @xmath38 is the standard deviation for the gaussian ) . the amplitudes of the gaussians were drawn randomly from a rayleigh distribution with a mode of unity ( arbitrary units ) . the amplitudes of these gaussians were then modified by a cosine density perturbation with a wavelength of @xmath39 km and an azimuth of @xmath40 by multiplying by a factor of @xmath41 $ ] where @xmath42 is the position vector of each gaussian and @xmath43 is a unit vector perpendicular to the cosine wavefronts . the initial field of fluctuations is displayed in the left panel of fig . [ toy ] with the positions of the vla antennas plotted as points in the center for reference . within the model , both the gaussian fluctuations and the cosine plane wave were given co - rotation speeds of 100 m s@xmath10 due east . the gaussian fluctuations were also given an additional flow velocity parallel to the cosine wavefronts of 200 m s@xmath10 . we then simulated vla measurements by advancing the field of fluctuations using these constant velocities with 10-second time steps over a one - hour time period . we ran the simulation twice , once with the flow velocity directed toward magnetic north , and once toward magnetic south . as with the actual self - calibration data , we fit a second order two - dimensional polynomial to the simulated @xmath5 measurements ( see appendix a ) and applied both the phase - based and dft - based spectral analysis methods to the results . as with the real vla data , we used a sliding half - hour wide boxcar , thus yielding dynamic spectra over the central half - hour of the simulated observations . the results for the phase - based and dft - based analyses for both the northward and southward flows are shown in the right panels of fig . [ toy ] as images of the spectral amplitude as a function of spatial frequency and azimuth . as one would expect , the dft - based approach successfully recovers the properties of the cosine wave with little evidence of the gaussian fluctuations , albeit with a somewhat overestimated size for the southward flow ( see the bottom right panel of fig . [ toy ] ) . conversely , the phase - based spectra show ample evidence of the influence of the gaussian disturbances . for both the northward and southward flows , the measured azimuths skew toward the direction of motion for the gaussians toward lower frequencies where they tend to dominate the spectral power . they form roughly linear structures very similar to those evident in fig . [ paimg ] , even reproducing the `` overshoot '' to more southeastward azimuths ( i.e. , @xmath44 ) for higher spatial frequencies seen within some of the features in fig . [ paimg ] . the exact shapes of these spectral features depend on several factors such as the gaussian fluctuations distributions of sizes and amplitudes and their flow speed as well as the number , sizes , and strengths of the field - aligned cosine waves . it is unclear to what degree the structures detected within our phase - based spectra constrain the properties of the flow(s ) and associated med waves . however , it is clear that our toy model has demonstrated that a substantial flow of non - uniform density material along field lines toward magnetic north provides a adequate explanation of the structures observed within the phase - based azimuth spectra . given that the vla lines of sight throughout the observations probed latitudes above the geomagnetic equator , this implies flow of material down flux tubes . this is consistent with the fact that the observations were conducted at night when some refilling of the ionosphere from the plasmasphere is expected ( see , e.g. , * ? ? ? in addition to the significant presence of plasmapheric activity within the data , there is also evidence of ionospheric disturbances , chiefly among the field - based calibration data . since the field - based data are based on observations of several sources with non - parallel lines of sight , it is much more effective at probing ionospheric rather than plasmaspheric structures . for instance , at a height of 300 km , the mean separation among the lines of sight to the 29 sources used is about 12 km . this implies an approximate nyquist sampling limit of a minimum of 24-km - sized structures . indeed , the channel maps shown in fig . [ chmap1][chmap4 ] show little if any structure beyond spatial frequencies of @xmath45 km@xmath10 . if we increase the assumed height to 3000 km , the approximate height of the plasmaspheric med waves described in the previous section , the minimum probable scale likewise increases to 240 km . this is much too large to detect any of the plasmaspheric med waves which have wavelengths @xmath46 km . the channel maps in fig . [ chmap1][chmap4 ] give a glimpse of the large amount of information about ionospheric disturbances contained within the field - based spectral cubes . at the beginning of the observing run , relatively low - frequency , northeastward - directed waves are prominent with weaker , higher - frequency waves directed toward the south / southeast . these all have wavelengths between 50 and 100 km . starting at a local time of 01:47 , the ionospheric med waves described in 3 appear over a wide range of temporal frequencies , as high as 20 hr@xmath10 . unlike the self - calibration data , they are seen within the field - based data through the remainder of the observing run . westward / southwestward directed waves are seen to overlap with those detected with the self - calibration data starting at a local time of about 03:30 , most prominently for frequencies @xmath47 hr@xmath10 . however , unlike the self - calibration data , roughly westward ( some northwestward , some southwestward ) waves are apparent at all times starting at the third time bin centered at a local time of 01:47 . they span nearly the entire range of temporal frequency and have wavelengths of between 50 and 100 km , similar to the ionospheric med waves . the relatively low @xmath4 indices for the time of our observations and the days preceding it ( @xmath48 ) indicate that the plasmaspheric disturbances we observed are not related in any way to geomagnetic storm activity . they instead appear to be the same plasmaspheric irregularities elongated along magnetic field lines discovered by @xcite and described in more detail by @xcite and @xcite . however , unlike the methods employed by these authors which rely on fitting a single plane wave to the vla @xmath5 measurements for a given time interval , our spectral analysis techniques can break the data up into many wavelike components . these techniques reveal that at a given time , there can be many plasmaspheric med waves occupying a wide range of l - shells and sizes as well as ionospheric med waves . in addition , our specialized phase - based spectral analysis method has revealed the existence of non - uniform density flows of material down the flux tubes occupied by these irregularities toward the nighttime ionosphere . in order for such a flow to take place , there needs to be some depletion within the ionosphere . of course , during the night , recombination naturally causes such a depletion . however , it is well established that during times of low geomagnetic activity , the density within the ionosphere can fluctuate substantially , between about 25% and 40% on scales of a few hours to several days @xcite . it may be that during our vla observations , larger than average nighttime depletions occurred as a result of quiet - time variability caused by forcing from the lower atmosphere via planetary waves , gravity waves , and/or tides @xcite . in the upper panel of fig . [ tecrat ] we establish that indeed , the level of geomagnetic activity was rather low during our observations . we have done this by plotting the ae index measure within 5-minute intervals on the date of the observation , 1 march 2001 , and the mean within @xmath49 days , both as functions of vla local time . while evidence of substorm activity can be seen near sunset and midmorning , the ae index during the time of our observations was substantially lower that the typical value for that time of year ( i.e. , late winter / early spring 2001 ) . to show that this low level of activity was accompanied by ionospheric depletions , we obtained gps - based vertical tec maps for north america from the mit haystack observatory madrigal database for 1 march 2001 and for 15 days before and after this date . we used these data to construct maps of vertical tec at @xmath50 resolution for geomagnetic longitudes between @xmath51 and @xmath52 and geomagnetic latitudes from @xmath53 to @xmath54 at 5-minute intervals . we then computed a 30-day mean map at each 5-minute interval and divided it into the corresponding map for 1 march 2001 to remove typical intraday variations . we then computed the mean of the log of this ratio at each latitude and longitude during the time of our observations and have displayed the result in the lower panel of fig . [ tecrat ] . from this map , one can see that there were several persistent , localized depletions relative to what was typical for the late winter / early spring of 2001 . there were also several locations of relatively enhanced tec , but most of the area in the map shown seems to be below average . this is qualitatively consistent with the idea that these variations are the result of forcing from the lower atmosphere which can have a non - uniform effect on the observed tec ( see , e.g. , * ? ? ? * ) . for comparison , we have also plotted contours of spectral power as functions of estimated geomagnetic longitude and invariant latitude , @xmath55 , for the plasmaspheric med waves described in 3 and fig . this shows that the l - shells occupied by these irregularities map well to relatively depleted regions of the ionosphere . it may then be that the irregularities themselves are tied to substructure within these depleted regions . such structure may cause some flux tubes to be more depleted than others as they refill the ionosphere , causing wavelike features as the ion density fluctuates as a function of geomagnetic longitude . if there are in fact significant structures on scales of a few tenths of a degree in geomagnetic longitude ( see 3 and the right panels of fig . [ med ] ) , this could contribute to the formation of the observed plasmaspheric med waves . the gps - based data does not provide the level of spatial resolution required to search for such structures . our own vla data has the ability to detect features on these scales , but if they exist , they are unfortunately overwhelmed by the plasmaspheric med waves and ionospheric disturbances . future studies based on satellite beacon data may be able to better test this hypothesis . there are two classes of ionospheric disturbances coincident in time with the plasmaspheric disturbances , field - aligned waves directed toward magnetic east and roughly westward directed waves . both classes of waves have wavelengths between about 50 and 100 km , and therefore can be categorized as medium scale traveling ionospheric disturbances ( mstids ) . they occupy a relatively large range in temporal frequencies , including higher frequencies than have been previously observed . with amplitudes @xmath56 tecu , they are also weaker than typical mstids ( see , e.g. , * ? ? ? * ; * ? ? ? this is likely a reflection of the limitations in sensitivity and time resolution of the instruments used to characterize mstids in past studies ( typically gps receivers ) . the fact that they occur at the same time as the observed plasmaspheric disturbances may indicate that the generation of these waves may be related to the deposition of material from the plasmasphere onto the ionosphere . the med , field - aligned waves may be generated in the topside ionosphere where other factors such as winds will have less of an effect and the disturbances are more likely to remain aligned with the magnetic field . the westward directed waves may arise lower within the @xmath30-region where the wind has more of an influence . according to the publicly available gpi data - driven runs of the tiegcm code ( * currently available at http://www.hao.ucar.edu/modeling/tgcm/ ) , the zonal wind above about 350 km is generally westward near the location of the vla during the time of day and year of our observations . using the unique capabilities of the vla vhf system , we have developed a multi - scale approach to the spectral analysis of tec gradients . probing a wide range of spatial scales and amplitudes , we have identified both plasmaspheric and ionospheric waves present during a @xmath14 hour observation during the night of 1 march 2001 . contemporaneous data suggest that geomagnetic activity was relatively low during this time , implying that the plasmaspheric disturbances were not part of storm - related activity . instead , combining the vla data with contemporaneous ae index and gps data , we have concluded that the field aligned waves in the plasmasphere appear to be associated with the flow of non - uniform density material along their wavefronts toward relatively depleted regions of the ionosphere . the low level of geomagnetic activity and the spatial distribution of ionosphere depletions hint that the depletions may be linked to density fluctuations caused by forcing from the lower atmosphere via planetary waves , gravity waves , and/or tides . we are currently conducting a more detailed investigation of this possibility using seven years of gps - based tec maps , ae index data , and 74 mhz vla data from the vla low - frequency sky survey @xcite . we have also simultaneously detected relatively fast ( roughly 150350 m s@xmath10 ) and weak ( amplitudes @xmath56 tecu ) mstids within the ionosphere . they were coincidence in time with the plasmaspheric disturbances and may have been generated by the deposition of material from the plasmasphere onto the nighttime ionosphere . they were directed rough toward the west and toward magnetic east . this may be a reflection of the altitudes at which these waves were generated . the field - aligned waves may have been generated relatively high within the ionosphere whereas the westward moving waves may have formed lower in the middle @xmath30-region where the wind has a larger effect . the data were obtained from the national radio astronomy observatory ( nrao ) data archive and initially processed within the astronomical image processing system ( aips ; * ? ? ? * http://www.aips.nrao.edu ) . within aips , the complex bandpass response of each antenna was computed and the data were corrected for this . an initial calibration was performed using a model of the distribution of the intensity of vir a on the sky presented by @xcite ( currently publicly available at http://lwa.nrl.navy.mil/tutorial/ ) . this model was used to compute model values of the correlations of signals between pairs of antennas , or `` visibilities . '' this model visibility data - set was then divided into the observed visibilities and the antenna - based gains were solved for within each scan ( i.e. , a contiguous block of observing time ) using standard aips tasks . this calibration technique is commonly referred to as `` self - calibration '' ; its application and robustness are detailed in @xcite . in this instance , self - calibration was used over relatively long time intervals to correct for amplitude and phase variations in the responses of the individual antennas which generally vary on relatively long time scales . this also has the effect of removing any slowly varying ionospheric contribution to the observed phases , but this has little impact on our analysis since we are focussing on variations on much shorter time scales ( see below ) . following the initial calibration , a second round of self - calibration was performed , this time at much shorter time intervals of 6.59 seconds , the shortest possible for this data - set given the time - averaging used to measure the visibilities . the motivation for performing two separate rounds of self - calibration will become apparent in the following section where we will detail a separate calibration performed over the entire 74 mhz field of view . this second round of self - calibration was used to compute the difference in tec between each antenna and the reference antenna ( see * ? ? ? * and fig . [ layout ] ) according to @xcite . unlike the data - set used by @xcite , we found no obvious jumps in the self - calibration determined antenna phases which would negatively impact the process of unwrapping the phases needed to compute the values of @xmath5 as functions of time . instead , we were able to simply unwrap the phases as functions of time within each scan and polarization . the phases were then de - trended by subtracting smoothed versions of the phase time series using a 1/2 hour wide , sliding boxcar filter . near the edges of each scan , a linear fit to the phases as functions of time were also subtracted to mitigate edge effects within the smoothing process . the boxcar was chosen to have the narrowest width that would not substantially dampen the longest period waves expected within the spectral analysis to be perform ( see 4 ) . we expect to see waves with periods up to about one hour ( e.g. , * ? ? ? * ; * ? ? ? the subtraction of a version of the data smoothed with a 1/2 hour wide boxcar will reduce the amplitudes of such waves by a factor of about 2.75 which is acceptable given that these waves also tend to have relatively large amplitudes . after this , the two polarizations were averaged and divided by a factor of 114 ( see , e.g. , equation ( 3 ) of * ? ? ? * ) to put them in units of tecu . the rms deviations between the two polarizations within bins of 10 time steps were used to estimate the @xmath57 uncertainties in the @xmath5 values . [ dtec_n ] shows examples of @xmath5 time series for antennas in the northern arm of the vla ( see fig . [ layout ] ) . the typical error in the @xmath5 values is about @xmath58 tecu . this is very similar to what was achieved by @xcite using a source that is about ten times brighter than vir a with similar time averaging ( 6.67 seconds ) . this implies that the accuracy attained is largely influenced by the time averaging used in the observations and that there are real ionospheric fluctuations on the order of @xmath6 tecu on time scales of about 6.5 seconds . to obtain some measure of ionospheric fluctuations over the entire @xmath15 field of view of the 74 mhz vla , we have also performed a procedure referred to as `` field - based '' calibration . the principles behind this method and its common application are detailed by @xcite and @xcite . in short , the first order effect of the ionosphere on an image of a cosmic source made from interferometric data is to shift its position . the observed position shifts of a particular source are proportional to the components of the tec gradient along the line of sight to the source averaged over the span of the array and the time interval used to make the image . at 74 mhz , images of sources with flux densities of more than about 2.5 jy can typically be made with 12 minutes of vla data to the quality needed to accurately measure the positions of the sources . field - based calibration uses a predetermined catalog of sources to choose a set of bright `` calibrators '' and makes images of each calibrator within 12 minute intervals , commonly referred to as `` snapshots '' . the position offset from its expected location is then measured for each source and a zernike polynomial is fit to all of the position offset data for a particular time interval . the zernike polynomials are then used to correct the entire field of view for ionospheric effects . for the observation of the field surrounding vir a , an additional step needed to be performed before running field - based calibration . since vir a is such a bright source , the dominant source of noise within an image made of any source in its vicinity is secondary peaks in the impulse response function ( irf ) , or `` sidelobes '' from vir a. it was therefore necessary to subtract the contribution of vir a from the observed visibilities before field - based calibration was implemented . this was done by using the second round of self - calibration to add the effects of the ionosphere to the model visibility data - set computed with the model of the sky brightness distribution of vir a. this altered model data - set was then subtracted from the visibility data . since the self - calibration process was done in two steps , this subtracted data was still corrected for instrumental effects while the short time - scale ionospheric variations remained . following the subtraction of vir a , field - based calibration was performed using the software package obit ( http://www.cv.nrao.edu/@xmath59bcotton/obit.html ) using one minute snapshots of all calibrators within the field of view brighter than 2.5 jy at 74 mhz , 48 sources in all ( excluding vir a ) . the input source catalog was that produced by @xcite from the vla low frequency sky survey ( vlss ) data . this run of field - based calibration was somewhat compromised by the presence of radio frequency interference ( rfi ) which is usually seen within 74 mhz vla data , especially on shorter baselines . however , a relatively new technique to subtract the effects of stationary sources of rfi from interferometric data has been developed by @xcite and implemented within obit . this technique requires that relatively bright sources be subtracted from the data before it is used , otherwise it may subtract potions of the intensity of such sources at times when their visibility fringe - rates are similar to those of the rfi . thus , we used the first run of field - based calibration to subtract out all sources detected within the field of view ( this is automatically done within obit ) and ran the rfi subtraction algorithm on the resulting residual data . we then ran an additional rfi flagging routine within obit which flags data with unusually large amplitudes , especially those with large stokes v signals as few if any cosmic vhf sources have non - zero stokes v intensities . following this , the previously subtracted sources were added back to the data , and field - based calibration was performed a second time . on average , the rfi subtraction and flagging reduced the noise in the calibrator snapshot images by 20% . following the completion of field - based calibration , we used the measured calibrator position shifts to compute the tec gradient as a function of local time for each source . we limited ourselves to only those sources that were detected in their snapshot images more than half of the time so that the the resulting time series would be reasonably complete which will be important for subsequent spectral analysis ( see appendix b ) . there were 29 calibrators that met this criterion . the relative positions of their associated ionospheric pierce - points are plotted in fig . [ calib ] for a height of 300 km . since we do not know a priori what height most of the observed fluctuations will occur at and the disturbances will likely span a range of heights , the choice of altitude is somewhat arbitrary . we have chosen 300 km as an approximate location within the lower f region where medium scale disturbances are likely to be found . in fig . [ calib ] , the positions of the vla antennas are also plotted around each calibrator position at a height of 300 km to represent the amount of smoothing that has been effectively applied by using the position offsets to measure the local tec gradients . after the position shifts were converted tec gradients assuming a height of 300 km , they were de - trended in a similar manner to the self - calibration - derived @xmath5 values for consistency . in particular , for each component of the gradient and at each time interval , a line was fit to the data for all time intervals within @xmath601/4 hour . any data with absolute deviations from the fitted line more than three times the median absolute deviation ( mad ) were then identified and the fit was redone without them . if the absolute difference between the value at the current time interval and the final linear fit was @xmath61mad , the fitted value was subtracted from it . otherwise , this time interval was flagged as being possibly spurious . as an example of the results of this processing , we have plotted the de - trended north - south components of the tec gradients as functions of time for all 29 calibrators in fig . [ grad_ns ] with the source peak intensities printed in each panel . note that these are the observed peak intensities which are lower than the sources cataloged intensities because of the response of the individual antennas which decreases significantly with angular separation from the center of the field of view . flagged time intervals are highlighted in red . these data include error - bars computed using the errors in the position offsets . these were computed at each time interval according to @xcite , where the error in the source position is given by @xmath62 where @xmath38 is the noise in the image , @xmath63 is the angular diameter of the source , and @xmath64 is the peak intensity on the image . here , we have assumed that the sources are unresolved and consequently have the same angular diameter as the irf , in this case , 76 arcseconds . the dominant source of noise within each snapshot is sidelobes from other sources within the filed of view . to estimate the typical value of @xmath38 for the snapshots , we made an image of the full filed of view using all 270 minutes of data with no deconvolution applied ( i.e. , without removing source sidelobes ) . the rms noise for this image was 0.042 jy beam@xmath10 , implying @xmath65 jy beam@xmath10 for each one - minute snapshot . given the intensities of the calibrators used ( see fig . [ grad_ns ] ) , this further implies a typical tec gradient uncertainty of about @xmath6 tecu km@xmath10 . we have performed separate spectral analyses on both the self - calibration and field - based calibration data . as we will detail below , the methods used are quite complementary and we include some comparison of the two in the main body of this article . the spectral analysis of the de - trended @xmath5 data derived from self - calibration using vir a ( see fig . [ dtec_n ] for examples ) was performed using two complementary techniques . the first follows that developed by @xcite and the second uses a straightforward fourier - style analysis . both techniques start with a second order , two - dimensional polynomial fit over the array to obtain approximate values for the tec gradient at each antenna s associated pierce - point . the locations of these pierce - points were computed assuming a nominal height of 300 km . however , since the lines of sight from the antennas to the sources are essentially parallel , the effect of the assumed height on these computed positions is marginal unless the observed source is at a relatively low elevation . during the vla observations , vir a was no fewer than @xmath66 above the horizon , indicating that the projected antenna separations changed very little throughout the observations ( see * ? ? ? * ) and had only a weak dependence on assumed height . for instance , the relative pierce - point positions among the antennas for a height of 300 km differ by at most a few percent from those computed for a height of 3000 km . the time series for the coefficients of the polynomials fit using these pierce - point positions were then fourier transformed ( here , within a 1/2 hour sliding window ) to obtain a complex spectrum of the gradient for each antenna . the two spectral analysis techniques both use the temporal fourier transforms of the antenna - based tec gradient time series . the first technique from @xcite uses the phase information from the fourier transform at each time step to estimate the weighted mean direction and magnitude of the wavenumber vector , @xmath67 , for each temporal mode . these results were combined with the temporal frequencies to also estimate pattern speeds . this technique uses both components ( i.e. , north - south and east - west ) components of the tec gradient . the technique assumes that the line along which the phases of the fourier transforms of the two components agrees is the direction for that temporal mode . this is done by first fitting a plane to the fourier transform phase of each tec gradient component after unwrapping the phases along each vla arm . then , the parameters of the two planar fits are used to determine the line defined by the intersection of the two planes . it was shown by @xcite that this phase - based approach can mitigate the influence of wavefront distortions on the analysis of relatively large waves ( i.e. , those substantially larger than the array ) . however , it also has the potential to sense the direction of motion for non - wavelike patterns . imagine , for instance , a distortion of arbitrary shape passing over the array . if one chooses a series of points along its direction of motion , the time series will be identical for each point except that they will be out of phase with one another . if the velocity is constant , then the phase difference between two of these points will just be proportional to the separation between them . this holds true for the tec times series as well as for the time series of the two components of the gradient . therefore , the `` phase - based '' technique of @xcite is capable of estimating the direction of motion , rather than orientation , when non - wavelike disturbances are encountered . the second technique uses a two - dimensional , spatial discrete fourier transform ( dft ) of each of the tec gradient components measured at each antenna$ ] ] . since the array is `` y''-shaped , one can imagine that the irf for this dft is rather poorly behaved . we show in fig . [ dft ] that this is in fact the case . the left panel of fig . [ dft ] displays the irf for the first half - hour of vla data . as one might expect , there are substantial sidelobes in the regions between vla arms extending to relatively large spatial frequencies . to mitigate the effect of these sidelobes , we have implemented a version of the clean algorithm @xcite . this algorithm is often used with images produced using radio interferometers to remove the effects of sidelobes associated with bright sources . briefly , it is an iterative algorithm that identifies a peak in the image plane at each iteration and subtracts a scaled version of the irf centered on that peak where the scale factor is usually the intensity multiplied by a predetermined gain value . the iterations are continued until a maximum iteration limit is reached or some convergence criteria are met . often , this is done interactively using boxes drawn by eye to limit the locations of the model delta functions , or `` clean components '' to specific regions . after the cleaning is finished , the clean components are convolved with a more well - behaved irf , usually a gaussian whose parameters are determined from the main peak of the irf . following this , the re - convolved components are added back to the image . our implementation is only slightly different since the spatial spectra we are `` cleaning '' are complex and not real valued functions like the images made with interferometers . consequently , at each iteration , we used the spectral power to determine the location of the next clean component to be subtracted , and the clean components all had complex amplitudes . for each clean component , a scaled version of the irf was used that was computed using the same grid spacing as the spectra but with @xmath68 pixels where @xmath69 is the number used for the actual spectra ( here , @xmath70 ; see fig . [ dft ] ) . this was done so that the irf could be easily shifted to be centered directly on a particular spectral pixel . after trying several convergence criteria , maximum iteration limits , and gain values , we found that satisfactory results were obtained by simply running 20 clean iterations with a gain of 0.5 and no limits on the locations of the clean components ( i.e. , no `` boxing '' ) . this cleaning was run separately on each temporal frequency and tec gradient component within each time bin . after restoring the re - convolved clean components to their respective spectra , the power spectra from both components of the tec gradient were combined into a single gradient power spectrum for each temporal frequency and time bin . [ dft ] shows examples of the results of the clean algorithm applied to the first half - hour of self - calibration data . in the right panels , we show the spectral power for the strongest temporal frequency , 6.375 hr@xmath10 , and for a randomly chosen frequency , 75.374 hr@xmath10 , both before and after cleaning . from these , on can see that nearly all the structure seen within the raw spectral maps was from the irf . it also shows that the overarching assumption of @xcite that each temporal mode is well approximated by a single wavenumber vector is essentially correct . because of this , and so we can more directly compare the two methods , we used the cleaned spectral power to computed the peak power and a single weighted ( with the spectral power ) mean value of the wavenumber vector for each temporal mode for each time step . because it is a more conventional fourier - based analysis , this `` dft - based '' method is most sensitive to wavelike structures and can only measure the wavefront orientation , which dow not necessarily give the direction of motion . the importance of this distinction is made apparent in the discussion in 3 . because both techniques estimate a single wavenumber vector for each temporal mode , they essentially use both the apparent motion of vir a and the motion of the observed disturbances to convert the properties of temporal oscillations observed with a compact array into spatial information . this allows a much higher spatial frequency resolution to be achieved than could normally be obtained with such a small array . for instance , the temporal spectral resolution , @xmath71 , is related to its spatial counterpart , @xmath72 , simply via the pattern speed , i.e. , @xmath73 . for the 1/2 hour wide window we have used here , @xmath74 . therefore , for a disturbance traveling 100 m s@xmath10 , @xmath75 which is the equivalent of what can be obtained with an array with a diameter of nearly 200 km . the results from both techniques are discussed in more detail in 2 . while the self - calibration - based spectral analysis supplies results with remarkable sensitivity and spatial frequency resolution , it suffers from two significant drawbacks . first , even though it uses a relatively long temporal baseline to achieve the equivalent of a long spatial baseline , the measurements are still confined to a narrow strip , in this case , about 11 km wide . this makes it difficult to accurately estimate the direction and speed of relatively large disturbances in the presence of significant wavefront distortions . second , wave properties are derived from the data for each temporal frequency . this implies that if there are several wave patterns with different sizes but similar periods , there is no way within this method to separate them from one another and all that can be estimated is a combined speed and direction for all of them . for isotropic phenomena such as turbulence , this may be acceptable , but for characterizing wave properties , it can be problematic , especially at temporal frequencies where there is no clearly dominant wave pattern . what is needed to supplement this analysis is another array that provides tec gradient measurements over similar time intervals but spread out over a much larger area . the position offset data used in the field - based calibration method detailed in appendix a provides just such an array . as fig . [ calib ] demonstrates , the distribution of calibrator sources on the sky provides and array of tec gradient measurements that spans an area with a diameter of about 100 km or more ( at a height of 300 km ) with a quasi - random distribution of ionospheric pierce - points . with this data , it is therefore possible to perform a relatively straightforward fourier analysis on the data without utilizing any `` tricks '' to retrieve spatial information about the observed tec fluctuations . one may simply perform a fourier transform in three dimensions , one temporal and two spatial , over a given time interval . this will yield spectral `` cubes '' so that different spatial modes with similar temporal frequencies may be separated from one another . we note that the field - based data has effectively been smoothed in time ( with a one minute window ) and space ( with an 11 km diameter circular window ) so that higher frequency modes are damped out . however , the self - calibration data provide ample information about such modes , making the two analyses quite complementary . to begin the analysis of the field - based data , we first put the data into 8 temporal bins . to minimize sidelobes in the temporal components of the fourier transforms , we have weighted the tec gradient time series within each bin by a unit amplitude gaussian with a full width at half maximum ( fwhm ) of 1/2 hour . we therefore constructed the temporal bins to be one hour wide , but separated by 1/2 hour so that all time intervals would have gaussian weights @xmath76 in at least one time bin . for each source , there were also time intervals where either the source s position could not be measured , or the data was otherwise flagged as spurious ( see appendix a ) . such gaps in temporal coverage can also lead to increased sidelobes in frequency space . we therefore reduced the gaussian weight of any time interval adjacent to such a gap by a factor of two to introduce some amount of tapering near these gaps to mitigate their effects on the impulse response . using this weighting scheme , we performed a discrete fourier transform ( dft ) of each component of the tec gradient time series for each source within each of the 8 time bins at 60 frequencies up to the nyquist limit of 30 hr@xmath10 . in each instance , we also computed the amplitude of the impulse response at @xmath77 and normalized each complex temporal spectrum by this value . following the generation of temporal spectra for the calibrators , the mean north - south and east - west position of each calibrator s pierce - point relative to that of the center of the filed of view at a height of 300 km was computed for each time bin . these were then used with a dft to compute a complex spectrum on a grid of north - south and east - west spatial frequencies for each time bin and temporal frequency and for each component of the tec gradient . again , these spectra were normalized by the amplitude of the irf at a frequency of zero . given the relatively sparse sampling provided by the distribution of calibrators , each resulting spatial spectrum was significantly affected by sidelobes . we sought to mitigate these effects using an implementation of the clean algorithm as was done with the dft - based method that was applied to the self - calibration data ( see b1 and fig . [ dft ] ) . as before , 20 iterations were used per temporal frequency and gradient component with a gain of 0.5 . in fig . [ clean ] , we show the mean gradient power over all temporal frequencies at each time bin both before and after cleaning . for reference , we have also included images of the irf power spectra . one can see that the cleaning process has substantially improved the fidelity of the spectral images . unlike the self - calibration data , the multiple lines of sight used in this analysis are far from parallel . this means that each of them `` views '' any disturbance from a different angle causing the observed amplitude of such a disturbance to vary across the vla field of view . for instance , if there is an optimum viewing angle for a particular disturbance , the detected amplitude for a given line of sight is @xmath78 , where @xmath79 is the angle between the actual line of sight and the optimum one ( i.e. , the dot product between the unit vectors for the actual and optimum lines of sight ) . the effect of this on our analysis is that this position dependent weighting of the measured gradients alters the irf from its expected shape . to illustrate the magnitude of this effect , we show in the panels of fig . [ orien ] the observed irf for values of @xmath79 ranging from @xmath80@xmath81 . for this demonstration , we assumed the optimum view angle was toward the north ( i.e. , its unit vector had no east - west component ) . as expected , the most notable change is for the largest @xmath82 . the lower panels of fig . [ orien ] show this better where they display the difference between the observed irf ( scaled to a maximum of unity ) and the expected irf . the maximum difference for @xmath82 is nearly 10% , but ranges from @xmath83% to a few percent for the other values of @xmath79 . therefore , when we clean the field - base spectral cubes , we are basically using the wrong irf , leading to sidelobes that are not properly subtracted and even adding additional but weak sidelobes . since there are typically several detected features within each field - based spectral cube which likely have a distribution of optimum viewing angles , the net effect of this on our analysis is to increase the noise of the spectral maps . however , as fig . [ orien ] indicates , this is likely a relatively small effect which is typically on the order of 1% and only as high as 10% in extreme cases . basic research in astronomy at the naval research laboratory is supported by 6.1 base funding . the national radio astronomy observatory is a facility of the national science foundation operated under cooperative agreement by associated universities , inc . the authors are grateful to mit haystack observatory for gps tec data obtained from their madrigal data center . cornwell , t. , r. braun , and d. s. briggs ( 1999 ) , deconvolution , in _ synthesis imaging in radio astronomy ii , asp conference ser . 180 , edited by g. b. taylor , c. l. carilli , and r. a. perley , pp . 151170 , asp , san francisco , calif . cornwell , t. and e. b. fomalont ( 1999 ) , self - calibration , in _ synthesis imaging in radio astronomy ii , asp conference ser . 180 , edited by g. b. taylor , c. l. carilli , and r. a. perley , pp . 187199 , asp , san francisco , calif . helmboldt , j. f. , t. j. w. lazio , h. intema , and k. dymond ( 2012 ) , a new technique for spectral analysis of ionospheric tec fluctuations observed with the very large array vhf system : from qp echoes to mstids _ rad . _ , rs0l02 jacobson , a. r. , r. c. carlos , r. s. massey , and g. wu ( 1995 ) , observations of traveling ionospheric disturbances with a satellite - beacon radio interferometer : seasonal and local time behavior , _ j. geophys . res . , 100 _ , 16531665 jacobson , a. r. , g. hoogeveen , r. c. carlos , g. wu , b. g. fejer , and m. c. kelley ( 1996 ) , observations of inner plasmasphere irregularities with a satellite beacon radio - interferometer array , _ j. geophys . res . , 101 _ , 665682 kassim , n. e. t. j. w. lazio , w. c. erickson , r. a. perley , w. d. cotton , e. w. greisen , a. s. cohen , b. hicks , h. r. schmitt , and d. katz ( 2007 ) , the 74 mhz system on the very large array , _ astrop . j. supp . , 172 _ , 686719 larsen , m. f. , d. l. hysell , q. h. zhou , s. m. smith , j. friedman , and r. l. bishop ( 2007 ) , imaging coherent scatter radar , incoherent scatter radar , and optical observations of quasiperiodic structures associated with sporadic @xmath84 layers , _ j. of geophys . res . , 112 _ , a06321 mikhailov , a. v. , a. k. depueva , and t. y. leschinskaya ( 2004 ) , morphology of quiet time f2-layer disturbances : high and lower latitudes , _ int . j. geomag . aeronom . , 114 _ , gi1004 , doi:10.1029/2003gi000058 park , c. g. , and p. m. banks ( 1974 ) , influence of thermal plasma flow on the mid - latitude nighttime @xmath86 layer : effects of electric fields and neutral winds inside the plasmasphere , _ j. geophys . , 79 _ , 46614668 tsugawa , t. y. , y. otsuka , a. j. coster , and a. saito ( 2007 ) , medium - scale traveling ionospheric disturbances detected with dense and wide tec maps over north america , _ geophys . lett . , 34 _ , l22101 wang , w. , t. l. killeen , a. g. burns , and r. g. roble ( 1999 ) , a high - resolution , three - dimensional , time dependent , nested grid model of the coupled thermosphere - ionosphere , _ phys , 61 _ , 385397 corresponds to north , @xmath23 to east , @xmath87 to west , and @xmath88 to south . within each spatial frequency bin , the values have been normalized by the total over all azimuth angle bins to make high - frequency features more apparent . the local time at the center of each time bin is given in each panel . ] plotted as greyscale contours . the shape of the impulse response function ( irf ) for each map is plotted in the lower left of each panel as an ellipse representing the full width at half maximum ( fhwm ) of the irf . the underlying image within each panel is a map of the log of the total power from the self - calibration data using the dft - based spectral analysis technique ( see 2 , appendix b , and fig . [ paimg ] ) . ] ) , the properties of the field - aligned plasmaspheric disturbances . the properties of each temporal mode from the dft - based analysis with azimuths between @xmath23 and @xmath24 and speeds @xmath89 km hr@xmath10 are plotted as separate points with magenta contours indicating the density of points within each panel . the left panel shows the spectral power as a function of the spatial frequency vector projected along magnetic east , @xmath22 . the middle panel plots the spectral power versus the velocity projected along magnetic east , @xmath27 , with corresponding estimates of the height , @xmath90 , and l - shell computed assuming all of the motion is from co - rotation . the right panels combine the results from the other two columns of plots to show the spectral power as a function of angular spectral frequency , @xmath31 ( i.e. , the disturbance scale in geomagnetic longitude ) . ] gaussian fluctuations modified by a 30-km - wavelength plane wave aligned along the magnetic field . the positions of the vla antennas are plotted as magenta points . results for both the phase - based and dft - based analysis techniques are shown in the right panels for flows directed toward magnetic north ( left panels ) and toward magnetic south ( right panels ) . in all four panels , the spectral amplitude is displayed as a function of spatial frequency ( abscissa ) and azimuth ( ordinate ) . white dotted lines in all four panels indicate the actual azimuth and wavelength of the field - aligned plane wave . ] days of the observations ( grey curve ) as functions of vla local time . lower : for the time range during the vla observations ( @xmath91ut@xmath92 ) , the mean of the log of the ratio of the vertical tec on 1 march 2001 to the 30-day average ( centered on 1 march 2001 ) within @xmath93 bins in geomagnetic latitude and longitude from the mit madrigal database . white contours represent the spectral power of field - aligned plasmaspheric disturbances as a function of geomagnetic longitude and estimated invariant latitude , @xmath55 . ] relative to the reference antenna ( see fig . [ layout ] and appendix a ) as functions of local time . the baseline lengths and median @xmath94 uncertainties estimated using both polarizations are printed for reference within the panels . ]

solar prominences are an important tool for studying the structure and evolution of the coronal magnetic field . here we consider so - called `` hedgerow '' prominences , which consist of thin vertical threads . we explore the possibility that such prominences are supported by tangled magnetic fields . a variety of different approaches are used . first , the dynamics of plasma within a tangled field is considered . we find that the contorted shape of the flux tubes significantly reduces the flow velocity compared to the supersonic free fall that would occur in a straight vertical tube . second , linear force - free models of tangled fields are developed , and the elastic response of such fields to gravitational forces is considered . we demonstrate that the prominence plasma can be supported by the magnetic pressure of a tangled field that pervades not only the observed dense threads but also their local surroundings . tangled fields with field strengths of about 10 g are able to support prominence threads with observed hydrogen density of the order of . finally , we suggest that the observed vertical threads are the result of rayleigh - taylor instability . simulations of the density distribution within a prominence thread indicate that the peak density is much larger than the average density . we conclude that tangled fields provide a viable mechanism for magnetic support of hedgerow prominences .

1001.2757

solar prominences ( a.k.a . filaments ) are relatively cool structures embedded in the million - degree corona at heights well above the chromosphere . above the solar limb , prominences appear as bright structures against the dark background , but when viewed as filaments on the solar disk they can be brighter or darker than their surroundings , depending on the bandpass used to observe them . magnetic fields are thought to play an important role in supporting the prominence plasma against gravity , and in insulating it from the surrounding hot corona . most quiescent prominences exhibit intricate filamentary structures that evolve continually due to plasma flows and heating / cooling processes . in some cases the threads appear to be mostly horizontal , while in other cases they are clearly radially oriented ( `` hedgerow '' prominences ) . figure [ fig : proms ] shows several examples of prominences observed in h@xmath2 at the big bear solar observatory ( bbso ) and the dutch open telescope ( dot ) . the examples in figs . [ fig : proms]a and [ fig : proms]b show mainly vertical threads , while the prominence in fig . [ fig : proms]c shows horizontal threads . off - limb observations in he ii 304 indicate that there are higher altitude parts that are optically thin in h@xmath2 and therefore not visible on the disk ( or at least have not been clearly identified in disk observations ) . figure [ fig : proms]d shows a prominence above the solar limb as observed in he ii 304 with the secchi / euvi instrument on the stereo spacecraft . the upper parts of the prominence consist of vertical threads with an intricate fine - scale structure . movie sequences of quiescent and erupting prominences can be found at the stereo website . prominence plasma is highly dynamic , exhibiting horizontal and vertical motions of order 1070 km s@xmath3 @xcite . recent high - resolution observations of filaments on the solar disk indicate that they consist of a collection of very thin threads with widths of about 200 km , at the limit of resolution of present - day telescopes @xcite . individual threads have lifetimes of only a few minutes , but the filament as a whole can live for many days . it seems likely that these thin threads are aligned with the local magnetic field . high - resolution images of prominences above the limb have been obtained with the solar optical telescope ( sot ) onboard hinode . for example , @xcite observed horizontal threads in a prominence near an active region , and studied the oscillatory motions of these threads . @xcite observed a hedgerow prominence consisting of many thin vertical threads , and they used multi - wavelength observations to estimate the amount of absorption and `` emissivity blocking '' in the prominence and surrounding cavity . @xcite observed another hedgerow prominence and found that the prominence sheet is structured by both bright quasi - vertical threads and dark inclusions . the bright structures are downflow streams with velocity of about 10 km s@xmath3 , and the dark inclusions are highly dynamic upflows with velocity of about 20 km s@xmath3 . the downflow velocities are much less than the free - fall speed , indicating that the plasma is somehow being supported against gravity . @xcite proposed that the dark plumes contain relatively hot plasma that is driven upward by buoyancy . @xcite observed a persistent flow of h@xmath2 emitting plasma into a prominence from one side , leading to the formation of vertical threads . they suggested that the vertical threads are stacks of plasma supported against gravity by the sagging of initially horizontal magnetic field lines . direct measurements of the prominence magnetic field can be obtained using spectro - polarimetry . a comprehensive effort to measure prominence magnetic fields was conducted in the 1970 s and early 1980 s using the facilities at pic du midi ( france ) and sacramento peak observatory ( usa ) . this work showed that ( 1 ) the magnetic field in quiescent prominences has a strength of 315 g ; ( 2 ) the field is mostly _ horizontal _ and makes an angle of about @xmath4 with respect to the long axis of the prominence ; ( 3 ) the field strength increases slightly with height , indicating the presence of dipped field lines ; ( 4 ) most prominences have _ inverse _ polarity , i.e. , the component of magnetic field perpendicular to the prominence axis has a direction opposite to that of the potential field . these earlier data likely included a variety of quiescent prominences , some with predominantly horizontal threads , others with more vertical threads . in more recent work , @xcite reported full - stokes observations of a limb prominence in he i 5876 ( he @xmath5 ) , and derived magnetic field strengths of 3045 g , somewhat larger than the values reported in earlier studies . @xcite published the first vector - field map of a prominence with a spatial resolution of a few arcseconds . they found that the average magnetic field in this prominence is mostly horizontal with a strength of about 20 g and with the magnetic vector pointing @xmath6 to @xmath7 off the prominence axis , consistent with the earlier studies . however , the map also shows clearly organized patches where the magnetic field is significantly stronger than average , up to 80 g . it is unclear how these patches are related to the fine threads seen at higher spatial resolution . recently , @xcite observed he i 10830 in a polar crown prominence above the limb , and found evidence for fields of about 30 g that are oriented only @xmath8 from the vertical direction . how is the plasma in hedgerow prominences supported against gravity ? many authors have suggested that quiescent prominences are embedded in large - scale _ flux ropes _ that lie horizontally above the polarity inversion line . the prominence plasma is thought to be located near the _ dips _ of the helical field lines . the magnetic field near the dips may be deformed by the weight of the prominence plasma . others have suggested that the magnetic field in hedgerow prominences is vertical along the observed threads , and that the plasma is supported by mhd waves . however , relatively high frequencies and wave amplitudes are required , and it is unclear why such waves would not lead to strong heating of the prominence plasma . @xcite and @xcite showed that the prominence plasma can be supported by the pressure of a coronal plasma lower down along an inclined field line ; however , this mechanism only works for nearly horizontal field lines . furthermore , hedgerow prominences are located in coronal cavities where the plasma pressure is very low . therefore , it seems unlikely that the vertical threads seen in hedgerow prominences can be supported by coronal plasma pressure on nearly vertical field lines . in this paper we propose that hedgerow prominences are embedded in magnetic fields with a complex `` tangled '' structure . such tangled fields have many dips in the field lines where the weight of the prominence plasma can be counteracted by upward magnetic forces . our purpose is to demonstrate that such tangled fields provide a viable mechanism for prominence support in hedgerow prominences . @xcite recently invoked tangled fields in the interpretation of spectropolarimetric observations of an active region filament . while such filaments are quite different from the hedgerow prominences considered here , this work shows that tangled fields have important effects on the measurement of prominence magnetic fields . such effects will not be considered in this paper . the paper is organized as follows . in section 2 we propose that hedgerow prominences are supported by tangled magnetic fields , and we discuss how such fields may be formed . in section 3 we present a simple model for the dynamics of plasma along the tangled field lines , and we show that weak shock waves naturally occur in such plasmas . in section 4 we develop a magnetostatic model of tangled fields based on the _ linear force - free field _ approximation , and in section 5 we study the response of such fields to gravitational forces . in section 6 we simulate the distribution of plasma in a cylindrical prominence thread . in section 7 we discuss the formation of vertical threads by rayleigh - taylor instability . the results of the investigation are summarized and discussed in section 8 . the spectro - polarimetric observations of prominences described in section [ sect : intro ] are consistent with the idea that quiescent prominences are embedded in coronal flux ropes that lie horizontally above the polarity inversion line ( pil ) . figure [ fig : cartoon2]a shows a vertical cross - section through such a flux rope . the magnetic field also has a component into the image plane , so the field lines are helices , and the plasma is assumed to be located at the dips of these helical windings . a dip is defined as a point where the magnetic field is locally horizontal and curved upward . as indicated in the figure , the magnetic field may be deformed by the weight of the prominence plasma , creating v - shaped dips . the magnetic field near such dips is well described by the kippenhahn - schlter model , and several authors have developed local magnetostatic models of the fine structures observed in quiescent prominences ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? however , recent observations of `` dark plumes '' and rotational motions @xcite within prominences remind us again that prominences have complex internal motions , and it is not clear how such motions can be explained in terms of a single large flux rope . perhaps the magnetic structure of hedgerow prominences is more complicated than that predicted by the flux rope model ( figure [ fig : cartoon2]a ) . in this paper we propose an alternative model , which is illustrated in figure [ fig : cartoon2]b . following @xcite , we suggest that hedgerow prominences are formed in current sheets that overlie certain sections of the pil on the quiet sun . unlike those previous authors we suggest that the current sheet extends only to limited height ( @xmath9 mm ) , and may extend only a limited distance along the pil . furthermore , we propose that _ tangled _ magnetic fields are present within these current sheets . a tangled field is defined as a magnetic structure in which the field lines are woven into an intricate fabric , and individual field lines follow nearly random paths . we suggest that the field is tangled on a spatial scale of 0.11 mm , comparable to the pressure scale height @xmath10 of the prominence plasma ( @xmath11 mm ) . the prominence plasma is assumed to be located at the many dips of the tangled field lines . the tangled field is confined horizontally by the vertical fields on either side of the sheet , and vertically by the weight of the prominence plasma . a key feature of a tangled field is that the plasma and field are in magnetostatic equilibrium , i.e. , the lorentz force is balanced by the gas pressure gradients and gravity . therefore , a tangled field is quite different from `` turbulent '' magnetic fields in which large - amplitude alfvn waves are present ( e.g. , the solar wind ) . in a tangled field the magnetic perturbations do not propagate along the field lines . in this paper we examine the basic properties of tangled fields , and we investigate their ability to support the prominence plasma . we suggest that the tangled field may be formed as a result of magnetic reconnection , not the twisting or stressing of field lines . quiescent prominences are located above polarity inversion `` lines '' that are often more like wide bands of mixed polarity separating regions with dominantly positive and negative polarity . in these mixed - polarity zones , magnetic flux elements move about randomly and opposite polarity elements may cancel each other . new magnetic bipoles frequently emerge from below the photosphere . these processes causes the `` recycling '' of the photospheric flux about once every 2 to 20 hours , and the coronal flux is recycled even faster @xcite . it is likely that the interactions between these flux elements produce a complex , non - potential magnetic field in the low corona . within this environment magnetic reconnection is likely to occur frequently at many different sites in the corona above the inversion zone . each reconnection event may produce a bundle of twisted field , and the twisted fields from different events may collect into larger conglomerates to form a tangled field . the tangled field may rise to larger heights ( as a result of its natural buoyancy ) , and may collect into a thick sheet that is sandwiched between smoother fields , as illustrated in figure [ fig : cartoon2]b . the observed prominence consists of plasma that is trapped within this sheet . new tangled field is continually injected into the sheet from below , producing vertical motions within the sheet . we suggest that the `` dark plumes '' observed by @xcite may be a manifestation of such vertical motions of the tangled field . the spatial distribution of plasma within the prominence is determined in part by the dynamics of plasma along the tangled field lines . figure [ fig : flow_cartoon ] shows the contorted ( but generally downward ) path of an individual field line in the tangled field . note that there are several `` dips '' where the field line is horizontal and curved _ upward _ , and `` peaks '' where the field is horizontal and curved _ downward_. tracing the field line upward from a dip , one always reaches a peak where the field line again turns downward . therefore , the question arises whether the plasma collected in the dips would remain in these dips or be siphoned out of the dips via the peaks of the field lines . to answer this question , we consider a simple model for the motion of the prominence plasma along the magnetic field . for simplicity we assume that the flow takes place in a thin tube surrounding the selected field line ( i.e. , the divergence of neighboring field lines is neglected ) , and the cross - sectional area of this tube is taken to be constant . we assume a steady flow is established along the tube . let @xmath12 and @xmath13 be the plasma velocity and density as functions of position @xmath14 along the tube , then conservation of mass requires @xmath15 = constant . the equation of motion of the plasma is @xmath16 where @xmath17 is the plasma pressure , @xmath18 is the height above the photosphere , and @xmath19 is the acceleration of gravity . the equation of state is written in the form @xmath20 , where @xmath21 and @xmath22 are constants ( we use @xmath23 to describe non - adiabatic processes ) . eliminating @xmath17 and @xmath13 from equation ( [ eq : flow1 ] ) , we obtain the following equation for the parallel flow velocity : @xmath24 where @xmath25 is the sound speed ( @xmath26 ) . the above equation has a critical point where the flow velocity equals the sound speed ( @xmath27 ) . therefore , a transition from subsonic to supersonic flow can occur only at points where the rhs of this equation vanishes , @xmath28 . these sonic points are located at the peaks of the field lines where matter can be siphoned out of one dip and deposited into another dip at lower height . the resulting flow pattern is indicated in figure [ fig : flow_cartoon ] . as the supersonic flow approaches the next dip , it must slow down to subsonic speeds , which can only happen in a shock . therefore , the tube has a series of subsonic and supersonic flows separated by shocks and sonic points . the role of these shocks is to dissipate the gravitational energy that is released by the falling matter . the position and strength of the shocks can be computed if the height @xmath18 of the flow tube is known . neighboring peaks are generally not at the same height . therefore , each section between neighboring peaks is approximated as a large - amplitude sinusoidal perturbation superposed on a generally downward path : @xmath29 where @xmath14 is the position along the flow tube , @xmath30 is the distance between neighboring peaks ( as measured along the flow tube ) , @xmath31 is the amplitude of the perturbation in height , @xmath32 is a phase angle , and @xmath33 is the background slope . the phase angle is chosen such that the peaks in the flow tube ( where @xmath28 ) are located at @xmath34 and @xmath35 , then the slope is given by @xmath36 the sonic points will then be located at @xmath34 and @xmath35 . figure [ fig : flow_shock]a shows the height @xmath18 for @xmath37 and @xmath38 rad , so that @xmath39 . let @xmath40 , @xmath41 and @xmath42 be the pressure , density and sound speed at the sonic points , then @xmath43 , and the sound speed can be written as @xmath44^{(\gamma-1)/2 } = c_0 [ v(s)/c_0 ] ^{-(\gamma-1)/2 } .\ ] ] inserting this expression into equation ( [ eq : flow2 ] ) , we obtain @xmath45 where @xmath46 , and @xmath47 is the pressure scale height at the sonic points . equation ( [ eq : flow3 ] ) can be integrated as follows : @xmath48 + \frac{1}{\gamma-1 } \left [ u^{1-\gamma}(s ) - 1 \right ] = - \frac{z(s)-z(s_0 ) } { \gamma h_{p,0 } } , \label{eq : flow4}\ ] ] where @xmath49 is the position of a sonic point . equation ( [ eq : flow4 ] ) can be solved for @xmath50 by newton - raphson iteration . when @xmath51 , there is an analytic solution for @xmath50 in terms of the lambert @xmath52 function ( see * ? ? * ) ; in the present paper we assume @xmath53 . the supersonic solution @xmath54 is computed with @xmath55 , and the subsonic solution @xmath56 is computed with @xmath57 . the dashed curves in figure [ fig : flow_shock]b show @xmath54 and @xmath56 , and figure [ fig : flow_shock]c shows the corresponding sound speeds @xmath58 and @xmath59 . here we assumed a scale height @xmath60 , equal to the amplitude of the field - line distortions . the mach number of the flow is given by @xmath61^{(\gamma+1)/2 } .\ ] ] the shock is located at the point @xmath62 where the mach number @xmath63 before the shock and the mach number @xmath64 after the shock satisfy the following relationship : @xmath65 which follows from the rankine - hugionot conditions for parallel shocks @xcite . therefore , the actual flow velocity @xmath12 between the two sonic points is given by the full curve in figure [ fig : flow_shock]b , and the sound speed @xmath25 is given by the full curve in figure [ fig : flow_shock]c . the plasma density @xmath13 along the flow tube is determined by mass conservation ( @xmath15 = constant ) , and is plotted in figure [ fig : flow_shock]d . note that there is a strong peak in the density at the dip in the field line , @xmath66 . the dashed curve in figure [ fig : flow_shock]d shows the density profile that would exist if the plasma were in hydrostatic equilibrium ( he ) : @xmath67 , \label{eq : rho}\ ] ] where @xmath68 and @xmath10 are the density and pressure scale height at the dip , @xmath69 . the deviations from hydrostatic equilibrium are significant only in those regions where the flow velocity is comparable to the sound speed . we define an average flow velocity @xmath70 by @xmath71 for the case shown in figure [ fig : flow_shock ] we find @xmath72 , so the average flow speed is less than the sound speed . therefore , the contorted shape of the flow tube significantly reduces the flow velocity compared to the supersonic free fall that would occur in a straight vertical tube . the cooler parts of the prominence are thought to have a temperature @xmath73 k. assuming a hydrogen ionization fraction of 10% , a helium abundance of 10 % and @xmath53 , the sound speed @xmath74 km s@xmath3 , and we predict an average flow velocity @xmath75 km s@xmath3 . the vertical component of this velocity is @xmath76 km s@xmath3 , less than the observed vertical velocities in prominence threads ( 510 km s@xmath3 ) . note that the predicted velocity is relative to the _ pattern _ of the tangled field , therefore , if the tangled field expands in the vertical direction it will push the prominence plasma upward . we speculate that the observed upward motions in hedgerow prominences ( e.g. , * ? ? ? * ) are due to such large - scale changes in the tangled field . in this section simple models for tangled fields are developed . a volume @xmath77 in the corona is considered , and the plasma inside this volume is assumed to be in magnetostatic equilibrium , @xmath78 , where @xmath79 is the plasma pressure , @xmath80 is the density , @xmath81 is the acceleration of gravity , and @xmath82 is the lorentz force . all quantities are functions of position @xmath83 within the volume . the lorentz force is given by @xmath84 where @xmath85 is the electric current density and @xmath86 is the magnetic field . using tensor notation , equation ( [ eq : f ] ) can also be written as @xmath87 , where @xmath88 is the magnetic stress tensor , a special case of maxwell s stress tensor @xcite : @xmath89 the first term describes magnetic pressure , and the second term describes magnetic tension . in a tangled field both pressure and tension forces are important . if gravity and plasma pressure gradients are neglected , then @xmath90 , so the magnetic field @xmath91 must satisfy the force - free condition : @xmath92 where @xmath93 may in general be a function of position . in the special case that @xmath2 is constant throughout the volume , equation ( [ eq : fff ] ) becomes a linear equation for @xmath94 , and the solutions are called linear force free fields ( lfff ) . @xcite has shown that in a closed magnetic system with a prescribed magnetic helicity ( @xmath95 , where @xmath96 is the vector potential ) the lowest - energy state is a lfff . therefore , in this paper only lfffs are considered , and @xmath2 is treated as a free parameter . we find that lfffs can be tangled . the typical length scale of the tangled field is given by the inverse of the @xmath2 parameter , @xmath97 . in the following we first consider the case that @xmath98 is small compared to the domain size @xmath99 in _ any _ direction , and then consider the boundary effects . section [ sect : cylinder ] describes tangled fields in a cylindrical domain . in the absense of boundary conditions , the solution of equation ( [ eq : fff ] ) can be written as a superposition of planar modes : @xmath100 , \label{eq : sum}\ ] ] where @xmath101 is the number of modes , @xmath102 is the wave vector ( @xmath103 ) , @xmath104 is the mode amplitude , @xmath105 is a phase angle , and @xmath106 $ ] are unit vectors that are mutually orthogonal and form a right - handed basis system : @xmath107 here @xmath108 and @xmath109 are the direction angles of the wave vector relative to the cartesian reference frame . figure [ fig : volume ] shows an example of a field with @xmath110 modes , an isotropic distribution of direction angles @xmath111 , and randomly selected phase angles @xmath105 . the starting points of the field lines are randomly selected from the central part of the box , and the field lines are traced until they reach the box walls . note that individual field lines follow random paths , and that different field lines are tangled together . now consider an ensemble @xmath112 of fields with different ( randomly distributed ) phase angles @xmath105 , but with a fixed number of modes @xmath101 , and with fixed mode amplitudes @xmath104 and direction angles @xmath113 . the phase angles are assumed to be uniformly distributed in the range @xmath114 $ ] , and angles from different modes @xmath115 and @xmath116 are assumed to be uncorrelated . then the ensemble average of the magnetic field vanishes , @xmath117 where @xmath118 denote the average over phase angles @xmath105 . also , the average of the tensor @xmath119 is given by @xmath120 @xmath121 where @xmath122 is the unit tensor ( @xmath123 ) . expression ( [ eq : tens1 ] ) is independent of position @xmath83 , so the magnetic field is statistically _ homogeneous _ , but is not necessarily isotropic . we study the statistical properties of field lines in models with different values of the mode number @xmath101 . for each @xmath101 we construct a series of models ( @xmath124 ) with different phase angles , but with constant values of the mode amplitudes @xmath104 and direction angles @xmath125 . the mode amplitudes are taken to be the same for all modes ( @xmath126 ) . for each realization @xmath127 of the phase angles @xmath105 we trace out the field line that starts at the origin ( @xmath128 ) , and we measure the square of the radial distance @xmath129 as function of position @xmath14 along the field line : @xmath130 we then average this quantity over @xmath131 realizations of the phase angles to obtain the mean square distance @xmath132 . for @xmath133 both mutually orthogonal directions @xmath125 and randomly chosen directions are considered . in both cases we find that the field lines follow long helical paths , and @xmath134 increases quadratically with @xmath14 . therefore , for @xmath133 the field lines do not behave randomly . for @xmath135 only randomly chosen directions are considered . for @xmath136 some of the field lines are long helices , while others have more random paths , and for @xmath137 all field lines seem random , however , in both cases @xmath132 is not well fit by a power law . true random walk behavior of the field lines , as indicated by a _ linear _ dependence of @xmath138 on distance @xmath14 , is found only when the number of modes is increased to @xmath139 . in the limit of large @xmath101 , @xmath140 . we now consider a larger ensemble @xmath141 in which not only the phase angles @xmath105 but also the mode amplitudes @xmath104 and direction angles @xmath113 are allowed to vary . from now on @xmath142 will denote an average over this larger ensemble . the direction angles are assumed to have an isotropic distribution , i.e. , the angle @xmath109 is uniformly distributed in the range @xmath143 $ ] , and @xmath144 is uniform in the range @xmath145 $ ] , so that @xmath146 . furthermore , the mode amplitudes @xmath104 are assumed to be uncorrelated with the direction angles , and @xmath147 , where @xmath148 is a constant . then @xmath149 , so the mean magnetic field vanishes . further averaging of equation ( [ eq : tens1 ] ) shows that @xmath150 is an isotropic tensor : @xmath151 it follows that @xmath152 , so @xmath148 equals the r.m.s . value of the total field strength . the ensemble average of the magnetic stress tensor , equation ( [ eq : t ] ) , is given by @xmath153 which is also isotropic . note that the diagonal components of @xmath154 are negative , so the effects of magnetic pressure dominate over the effects of magnetic tension . therefore , the isotropic tangled field has a positive magnetic pressure , @xmath155 . the average energy density is @xmath156 . the relationship @xmath157 is similar to that for a relativistic gas ( e.g. , * ? ? ? the tangled field must be confined within a certain volume ( e.g. , a current sheet , see figure [ fig : cartoon2]b ) , and the confinement must be effective for a period much longer than the alfven travel time across the volume . what are the conditions for such confinement ? to answer this question we must consider the boundary region between a tangled field and a smooth field . the tangled field is assumed to be characterized by a high value of @xmath158 , and the smooth field presumably has a much lower value of @xmath159 . to last a long time , the magnetic field near the boundary must be approximately in equilibrium ( non - linear force - free field ) . the force - free condition ( [ eq : fff ] ) implies that @xmath2 is constant along field lines , so there can not be many field lines that pass from the smooth region to the tangled region . therefore , one important condition for the survival of the tangled field is that the two regions are nearly disconnected from each other magnetically . another requirement is that the two regions are approximately in pressure balance . to show that these conditions can be satisfied , we now consider a simple model for the boundary region . the interface between the tangled and smooth fields is approximated by a plane surface , here taken to be the plane @xmath160 in cartesian coordinates . the above - mentioned condition on the lack of connectivity between the smooth and the tangled fields requires @xmath161 at @xmath160 . the tangled field in @xmath162 is assumed to be a lfff with a specified value of @xmath2 . the solution of the lfff equation is again written as a superposition of planar modes . however , in the present case the modes are grouped into pairs with closely related wave vectors @xmath163 and @xmath164 , and with the same amplitude @xmath104 and phase @xmath105 : @xmath165 . \label{eq : sum2}\end{aligned}\ ] ] here @xmath101 is the total number of modes , @xmath166 and @xmath167 are defined in equations ( [ eq : e1 ] ) and ( [ eq : e2 ] ) , @xmath168 is the modified wave vector , and the unit vectors @xmath169 and @xmath170 are defined by @xmath171 note that @xmath170 differs from @xmath172 only in the sign of the @xmath173-component , whereas @xmath169 has the sign of the @xmath174- and @xmath175-components reversed . therefore , the unit vectors @xmath176 $ ] again form a right - handed basis system . the magnetic field at the boundary ( @xmath177 is given by @xmath178 @xmath179 , \label{eq : sum3}\ ] ] which satisfies @xmath180 therefore , it is possible to construct a tangled field that is disconnected from its surroundings . we now consider the statistical average of the tensor @xmath181 at @xmath160 . averaging over phase angles @xmath105 , we obtain @xmath182 @xmath183 and further averaging over mode amplitudes and direction angles yields @xmath184 here we assume an isotropic distribution of direction angles , and we use @xmath185 , where @xmath148 is the r.m.s . field strength in the interior of the tangled field ( see section [ sect : volume ] ) . note that at the boundary @xmath186 , while in the interior @xmath152 , so the r.m.s . field strength at the boundary is reduced by a factor @xmath187 compared to that in the interior . the magnetic pressure at @xmath160 is given by @xmath188 where @xmath189 is the average pressure in the interior of the tangled field [ see equation ( [ eq : stress ] ) ] . equation ( [ eq : pext ] ) shows that it is possible to maintain pressure balance between the tangled field and its surroundings . here an infinitely long cylinder with radius @xmath190 is considered . we adopt a cylindrical coordinate system @xmath191 , and we assume that the radial component of magnetic field vanishes at the cylinder wall , @xmath192 . in the appendix we analyze the eigenmodes of the lfff equation in the domain @xmath193 subject to the above boundary condition . we find that this eigenvalue problem has a discrete set of modes , and the number of modes depends on the dimensionless parameter @xmath194 . figure [ fig : tube ] shows the resulting magnetic fields for @xmath195 , 4.5 and 6.0 . in the first case only the axisymmetric ( lundquist ) mode is present , so the field lines are helical . assuming the cylinder axis is vertical , there are no dips in the field lines . if cool plasma were to be injected into such a structure , it would spiral down along the field lines and quickly reach supersonic speeds . in contrast , for @xmath196 and @xmath197 there are multiple modes of the lfff , and the random superposition of these modes creates a tangled field with many dips where prominence plasma can be supported . the field - line dips ( i.e. , sites where @xmath198 and @xmath199 ) are indicated by _ magenta _ dots in the middle and right panels of figure [ fig : tube ] . we will return to this model in section [ sect : thread ] . the above models for a tangled field are purely force - free and do not have any magnetic forces to support the prominence plasma against gravity . to include such effects , we now consider the `` elastic '' properties of the tangled field , i.e. , its response to external forces . specifically , the weight of the prominence causes the tangled field to be compressed in the vertical direction , resulting in a radially outward force on the plasma . also , shearing motions may occur within the tangled field as dense plasma moves downward and less dense `` plumes '' move upward ( e.g. , * this results in shear deformation of the tangled field and associated magnetic stresses that counteract the plasma flows . in the following both of these effects are considered in some detail . we first consider the effects of gravity on a layer of tangled magnetic field . the magnetostatic equation ( @xmath200 ) can not be solved analytically for a tangled field , so we make the following approximation : @xmath201 = \left [ { \bf b } ( { \bf r } ) - \frac{1}{\alpha h_b } \hat{\bf z } \times { \bf b } ( { \bf r } ) \right ] e^{-z / h_b } . \label{eq : compress}\ ] ] here @xmath91 is the lfff given by equation ( [ eq : sum ] ) , and @xmath202 is the magnetic scale height of the modified field ( we assume @xmath203 ) . note that @xmath204 as required . this modified field @xmath205 is no longer force - free , but has the following statistical properties : @xmath206 e^{-2z / h_b } , \label{eq : tens2b } \end{aligned}\ ] ] where @xmath207 . therefore , the magnitude of the modified field drops off exponentially with height @xmath175 . let @xmath208 be the magnetic stress tensor of the modified field . taking its statistical average , we find for the nonzero components of the stress tensor : @xmath209 note that for @xmath210 the stress tensor is nearly isotropic . the net force on the plasma is given by @xmath211 , and the average force follows from equations ( [ eq : mxx ] ) and ( [ eq : mzz ] ) : @xmath212 note that the average force acts in the positive @xmath175 direction , i.e. , the magnetic force counteracts the force of gravity . in effect , the plasma is being supported by the magnetic pressure of the tangled field . the tangled field acts like a hot gas that has a significant pressure but no mass . the _ average _ density of the plasma that can be supported by the tangled field is given by @xmath213 where @xmath19 is the acceleration of gravity . the horizontal components of lorentz force , @xmath214 and @xmath215 , do not vanish for any particular realization of the tangled field , and can not be written as the gradients of a scalar pressure @xmath79 . the reason is that expression ( [ eq : compress ] ) is not an exact solution of the magnetostatic equilibrium equation . however , equation ( [ eq : fxyp ] ) shows that the horizontal forces vanish when averaged over the fluctuations of the isotropic tangled field . therefore , expression ( [ eq : compress ] ) is thought to give a good approximation for the effects of gravity on the tangled field . we now apply the above model to the vertical threads observed in hedgerow prominences . to explain the observed heights of such prominences , we require that the magnetic scale height @xmath202 is at least 100 mm . the size @xmath98 of the magnetic tangles is assumed to be in the range 0.11 mm , so @xmath216 . for @xmath217 g and @xmath218 mm , we find @xmath219 @xmath220 , which corresponds to an average ( total ) hydrogen density @xmath221 @xmath222 . this is only about 0.05 times the density @xmath223 @xmath220 or @xmath224 @xmath1 typically observed in hedgerow prominences . this comparison shows that the pressure of the tangled field _ inside _ a prominence thread is not sufficient to support the weight of the prominence plasma . to support the plasma with tangled fields , we need to take into account the magnetic coupling between the vertical thread and its surroudings . such coupling is neglected in the above plane - parallel model . we now assume that the tangled field pervades not only the observed vertical threads but also their local surroundings . the density in the surroundings is less than that in the threads , so the force of gravity is also much lower . this difference in gravitational forces leads to vertical motions ( downflows in the dense threads , upflows in the tenuous surroundings ) that create magnetic stresses in the tangled field . the magnetic coupling between the prominence and its surroundings causes the weight of the dense prominence to be distributed over a wider area . in effect , the prominence plasma is being supported by the radial gradient of the magnetic pressure of the tangled field over this larger area . in the following we estimate the magnetic stresses and vertical displacements resulting from these forces . the tangled field is modeled either as a vertical slab with half - width @xmath190 , or as a vertical cylinder with radius @xmath190 . the prominence is located at the center of this slab or cylinder , and has a half - width or radius @xmath225 . then the average density in the tangled field region is given by @xmath226 where @xmath41 is the density of the prominence , and we neglect the mass of the surroundings . the exponent @xmath227 for the slab model or @xmath228 for the cylindrical model . as discussed in section [ sect : compress ] , observations of hedgerow prominences indicate @xmath229 @xmath220 , and to explain the observed heights of such prominences with @xmath217 g , we require @xmath230 . according to equation ( [ eq : rhoa2 ] ) , this implies @xmath231 for the slab model , or @xmath232 for the cylindrical model . the observed threads have widths down to about 500 km , which corresponds to @xmath233 km . therefore , the magnetic coupling by the tangled field must extend to a surrounding distance of at least 5 mm for the slab model , or 1.1 mm for the cylindrical model . more generally , equations ( [ eq : rhoa1 ] ) and ( [ eq : rhoa2 ] ) yield the following expression for the magnetic scale height of tangled field : @xmath234 therefore , the maximum height of the prominence depends strongly on the magnetic field strength @xmath148 . according to the present model , magnetic stress builds up in the tangled field as a result of the difference in gravitational forces between the thread and its surroundings . can the field support such shear stress ? to answer this question we examine the effect of vertical displacements on the tangled field . for simplicity we neglect the mean vertical force given by equation ( [ eq : fzp ] ) , and we focus on _ relative _ displacements . let @xmath235 be the new position of a fluid parcel originally located at position @xmath236 . in the limit of a perfectly conducting plasma , the deformed field @xmath237 at the new position @xmath235 is given by @xmath238 where @xmath91 is the original field , and @xmath239 is the jacobian of the transformation . we assume @xmath240 which yields @xmath241 where @xmath242 is the vertical displacement . the original field @xmath91 is assumed to be a realization of the isotropic tangled field given by equation ( [ eq : sum ] ) . using equation ( [ eq : tens2 ] ) , we obtain @xmath243 , \label{eq : tens3 } \end{aligned}\ ] ] where @xmath244 this yields the following expressions for the off - diagonal components of the stress tensor : @xmath245 the lorentz force is given by @xmath246 , and since @xmath247 is independent of @xmath175 , the average vertical force is given by @xmath248 where @xmath249 is the density perturbation ( @xmath250 ) . using this equation , we can determine the vertical displacement @xmath242 for a given density variation @xmath251 . in the following subsections we solve the above equation for the slab and cylinder models . we first consider a slab with infinite extent in the @xmath252 , @xmath253 and @xmath254 directions . the coordinate @xmath173 perpendicular to the slab is in the range @xmath255 , where @xmath190 is the half - width of the sheet in which the tangled field is embedded ( see figure [ fig : cartoon2]b ) . then the density perturbation is given by @xmath256 & \mbox{for } ~ |x| < r_0 , \\ - \rho_0 r_0/r & \mbox{otherwise } . \label{eq : drho1 } \end{array } \right.\ ] ] inserting this expression into equation ( [ eq : fzpp ] ) and solving for the vertical displacement , we obtain @xmath257 ( x^2 - r_0 ^ 2 ) & \mbox{for } ~ |x| < r_0 , \\ c ( r_0/r ) [ ( r - r_0)^2 - ( r-|x|)^2 ] & \mbox{otherwise } , \end{array } \right.\ ] ] where @xmath258 . here we applied no - stress boundary conditions ( @xmath259 ) at @xmath260 . note that @xmath261 and its derivative are continuous at the edges of the prominence ( @xmath262 ) . the relative displacement across the tangled field is given by @xmath263 figure [ fig : shear]a shows the function @xmath261 for @xmath264 mm , @xmath265 mm , @xmath217 g and @xmath266 @xmath220 , so that @xmath267 mm . note that a relatively small deformation of the tangled field ( @xmath268 ) is sufficient to redistribute the gravitational forces over the full width of the tangled field . however , @xmath269 is larger than the pressure scale height of the prominence plasma ( @xmath11 mm ) . therefore , the deformation of the tangled field in the neighborhood of the prominence may have a significant effect on the distribution of the prominence plasma . this issue will be further discussed in section [ sect : thread ] . for comparison of the tangled field slab model with the flux rope model ( figure [ fig : cartoon2 ] ) , we define the average _ sag angle _ @xmath270 of the prominence relative to its surroundings : @xmath271 where @xmath148 is the r.m.s . field strength of the tangled field , and we assumed @xmath272 . this expression is similar to that derived for the kippenhahn - schlter model : @xmath273 where @xmath274 is the magnetic field through the mid - plane of the prominence . equation ( [ eq : ks ] ) describes the angle of the field lines in the flux rope model shown in figure [ fig : cartoon2]a . therefore , the flux - rope and tangled field models are similar in their ability to explain the magnetic support of the prominence plasma , provided the half - width @xmath190 of the tangled field region is similar to the radius @xmath190 of the flux rope . we now consider a cylindrical model for a prominence thread with @xmath275 the distance from the ( vertical ) thread axis . the density perturbation is given by @xmath276 & \mbox{for } ~ r < r_0 , \\ - \rho_0 ( r_0/r)^2 & \mbox{for } ~ r_0 < r < r . \label{eq : drho2 } \end{array } \right.\ ] ] the vertical displacement is obtained by solving equation ( [ eq : fzpp ] ) , which yields @xmath277 & \mbox{for } ~ r < r_0 , \\ h_0 [ 2 r^2 \ln ( r / r_0 ) -r^2 + r_0 ^ 2 ] /(r^2-r_0 ^ 2 ) & \mbox{for } ~ r_0 < r < r , \end{array } \right . \label{eq : hr}\ ] ] where @xmath278 and we applied no - stress boundary conditions at @xmath279 . then the total displacement across the tangled field is @xmath280 figure [ fig : shear]b shows the vertical displacement @xmath281 for @xmath282 mm , @xmath283 mm , @xmath284 g and @xmath285 @xmath220 , which yields @xmath286 mm and @xmath287 mm . in this case @xmath269 is significantly larger than the pressure scale height ( @xmath11 mm ) , mainly because of the lower field strength compared to the case shown in figure [ fig : shear]a . in section [ sect : thread ] we consider the effect of such deformation on the field - line dips , and on the spatial distribution of the prominence plasma . the above analyses only provide an rough estimate for the density of prominence plasma that can be supported by the tangled field . the actual density distribution @xmath288 is likely to be much more complex for several reasons . first , plasma will tend to collect at the dips of the field lines , so the density will vary on the spatial scale @xmath98 of the tangled field and on the scale of @xmath10 ; this effect will be considered in more detail in section [ sect : thread ] . second , the density will vary with time because there are flows along the field lines ( see section [ sect : flows ] ) and these flows are likely to be non - steady . also , the magnetic structure is not fixed and will continually evolve as dipped field lines are distorted by the weight of the prominence plasma . to predict the actual density will require numerical simulations of the interaction of tangled fields with prominence plasma , which is beyond the scope of the present paper . according to the present theory , hedgerow prominences are supported by the pressure of a tangled magnetic field , which acts like a tenuous gas and is naturally buoyant . it is well known that a tenuous medium supporting a dense medium is subject to rayleigh - taylor ( rt ) instability . therefore , we suggest that the observed vertical threads may be a consequence of rt instability acting on the tangled field and the plasma contained within it . as cool plasma collects in certain regions of the tangled field , the weight of the plasma deforms the surrounding field , which causes even more plasma to flow into these regions . a detailed analysis of the formation of prominence threads by rt instability is complicated by the fact that we presently do not understand how a tangled field responds to shear deformation . in section [ sect : shear ] we estimated the relative vertical displacement @xmath269 of the prominence plasma assuming no reconnection occurs during the deformation of the magnetic field by gravity forces [ see equations ( [ eq : dh1 ] ) and ( [ eq : dh2 ] ) ] . in this case the tangled field behaves as an `` elastic '' medium with magnetic forces proportional to the displacement . however , it is not clear that this approximation is valid . high - resolution observations of prominences indicate that the dense threads move downward relative to their more tenuous surroundings with velocities of the order of 10 - 30 @xmath289 . if the threads and their surroundings are indeed coupled via tangled fields , these relative motions imply that the field is continually being stretched in the vertical direction . therefore , the shear stress continually increases with time , unless there is internal reconnection that causes the shear stress to be reduced . we speculate that tangled fields have a tendency to relax to the lfff via internal reconnection . a similar relaxation processes occurs in the reversed field pinch and other laboratory plasma physics devices . therefore , the long - term evolution of prominence threads likely involves small - scale reconnection within the tangled field . the tangled field may behave more like a `` plastic '' medium that is irreversibly deformed when subjected to shear stress . such plasticity makes it possible to understand how the dense threads can move downward relative their the surroundings at a small but constant speed . these flows significantly deform the tangled field , but the field is nevertheless able to support the plasma against gravity . a detailed analysis of reconnection in tangled fields and its effect on the prominence plasma is beyond the scope of the present paper . the observed vertical structures likely reflect the non - linear development of the rt instability in hedgerow prominences . to establish a vertical column of mass resembling a prominence thread will likely require vertical motions over a significant height range ( tens of mm ) . starting from a homogeneous density distribution , it may take several hours for the threads to form by rt instability . we now construct a model for the density distribution in a fully formed ( vertical ) prominence thread supported by a tangled field . it is assumed that the rt instability has produced a vertical thread that is clearly separated from the rest of the prominence plasma . therefore , only a single thread and its local surroundings are considered , and the tangled field is assumed to be contained in a vertical cylinder with radius @xmath282 mm . as discussed in section [ sect : flows ] , there will in general be mass flows along the tangled field lines , but for the purpose of the present model we neglect such flows and we assume that the plasma is in hydrostatic equilibrium along the field lines . to construct the density model , we first compute a particular realization of the lfff with @xmath290 ( see appendix for details ) . to account for the weight of the prominence plasma , this field is further deformed as described by equation ( [ eq : hr ] ) . the deformation parameters are @xmath291 mm and @xmath292 mm , which yields @xmath293 mm , somewhat larger than the values used in figure [ fig : shear]b . as shown in section [ sect : compress ] , the weight of the prominence plasma causes the strength of the tangled field to decrease with height [ see equation ( [ eq : compress ] ) ] , but for simplicity this gradient is neglected here . the pressure scale height is assumed to be constant , @xmath294 mm , which corresponds to a temperature of about 8000 k , typical for h@xmath2 emitting plasma in prominences . we introduce cartesian coordinates @xmath295 with the @xmath175 axis along the cylinder axis ; the @xmath173 and @xmath174 coordinates are in the range @xmath296 $ ] , and the height @xmath175 is in the range @xmath297 $ ] . the density @xmath298 in this volume is computed on a grid with @xmath299 grid points , using the following method . we randomly select a large number of points within the cylinder and trace out the field lines that pass through these points . for each field line we plot the height @xmath18 as function of position @xmath14 along the field line , and we find the peaks and dips in the field line . for each dip we find the two neighboring peaks ( @xmath300 and @xmath301 ) and we determine the depth @xmath302 of the valley between these peaks . we then iteratively remove shallow dips with @xmath303 by concatenating neighboring sections . figure [ fig : dips ] shows an example for one particular field line ; the remaining dips with @xmath304 are indicated by squares . we then compute the density @xmath13 in each interval , assuming hydrostatic equilibrium along the field line [ see equation ( [ eq : rho ] ) ] . since we are interested only in _ relative _ densities , we set @xmath305 , the same for all dips on all field lines . finally , we distribute the density @xmath13 onto the 3d grid by finding the grid points that lie closest to the path of the field line . this process is repeated for 8000 field lines to obtain the density @xmath298 throughout the 3d grid . figure [ fig : mass_plot ] shows the resulting density distribution . the three panels show different projections obtained by integration in the @xmath173 , @xmath174 and @xmath175 directions , respectively [ for example , figure [ fig : mass_plot]a shows @xmath306 . note that the plasma is concentrated in the central part of the cylinder ; this is due to the deformation of the magnetic field described by the vertical displacement @xmath281 , which changes the distribution of the field - line dips compared to that in the lfff . the plasma is highly inhomogenous ( @xmath307 ) , and is distributed in sheets corresponding to surfaces of field - line dips . in some regions there are multiple sheets along the line of sight ( los ) . this is consistent with observations of the h i lyman lines in solar prominences , which indicate multiple threads along the los . the above model is quasi - static and does not take into account the expected mass flows along the field lines ( section [ sect : flows ] ) , nor the dynamical processes by which the threads are formed ( section [ sect : rt ] ) . constructing a more realistic 3d dynamical model of a prominence thread supported by tangled fields is beyond the scope of the present paper . however , our main conclusion that the density distribution within the thread is highly inhomogeneous is likely to be valid also in the dynamical case . we propose that hedgerow prominences are supported by magnetic fields that are `` tangled '' on a spatial scale of 1 mm or less . a key feature of the model is that the plasma is approximately in magnetostatic balance , therefore , the model is different from earlier models in which the plasma is supported by mhd - wave pressure . in the present case the perturbations of the magnetic field do not propagate along the field lines . the tangled field is located within a large - scale current sheet standing vertically above the pil ( figure [ fig : cartoon2]b ) , and is not magnetically connected to the photosphere on either side of the pil . such tangled fields may be formed by flux emergence followed by magnetic reconnection in the low corona . in this paper we use a variety of methods to explore the interactions of prominence plasma with tangled fields . in section [ sect : flows ] a simple model for the downward flow of plasma along the distorted field lines was developed . it was shown that such flows naturally develop standing shocks where the gravitational energy of the plasma is converted into heat ; this may be important for understanding the heating of prominence plasmas . the average flow velocity is less than the sound speed , indicating that the tangled field is able to support the prominence plasma against gravity . in section [ sect : lfff ] linear force - free models of tangled fields were constructed . tangled fields can be described as a superposition of planar modes . we studied the statistical properties of such fields , and found random - walk behavior of the field lines when the number of modes is sufficiently large ( @xmath308 ) . to produce tangling of the field lines on a scale of 1 mm or less , as required for our model of hedgerow prominences , we need @xmath309 @xmath310 , much larger than the values typically found in measurements of photospheric vector fields . therefore , according to the present model , hedgerow prominences are embedded in magnetic fields with high magnetic helicity density . we also considered the conditions for confinement of a tangled field by the surrounding smooth fields , and showed that despite the high helicity density tangled fields can be in pressure balance with their surroundings . in section [ sect : forces ] the `` elastic '' properties of a tangled field were considered , i.e. , their linear response to gravitational forces assuming ideal mhd . we found that the weight of the prominence plasma can be supported by the nearly isotropic magnetic pressure of the tangled field . the tangled field pervades not only the observed vertical threads , but also their local surroundings . the magnetic coupling between the threads and their surroundings is quite strong : vertical displacements of only 0.52 mm are sufficient to counteract the shear stress resulting from the different gravitational forces . in effect , the weight of a dense thread is distributed over an area that is larger than the cross - sectional area of the thread . as discussed in section [ sect : shear ] , the observed densities in prominence threads ( @xmath311 @xmath1 ) can be supported by tangled fields with field strengths in the range 315 g. in section [ sect : rt ] we proposed that the observed vertical structures in hedgerow prominences are a consequence of rayleigh - taylor ( rt ) instability acting on the tangled field and the plasma contained within it . the tangled field acts like a hot , tenuous gas and is naturally buoyant ; its support of the dense prominence plasma is likely to be unstable to flows that separate the gas into dense and less - dense columns . the observed vertical structures likely reflect the non - linear development of the rt instability . a detailed analysis of this instability is complicated by the fact that it requires internal reconnection to occur within the tangled field , and it is unclear how rapidly such reconnection can proceed . therefore , it is difficult to predict the vertical velocities in prominence threads . clearly , more advanced numerical simulations of the interaction of tangled fields with prominence plasma are needed . finally , in section [ sect : thread ] we simulated the density distribution in a prominence thread , using a cylindrical model for the tangled field and its deformation by gravitational forces . the results indicate that the threads have an intricate fine - scale structure . multiple structures are superposed along the los , consistent with observations of the h i lyman lines @xcite . while our model does not take into account any dynamical processes , the main conclusion that the density distribution is highly inhomogeneous is likely to be valid also in the dynamical case . the images shown in figures [ fig : proms]a and [ fig : proms]b were obtained at the big bear solar observatory , which is operated by the new jersey institute of technology . the image shown in figure [ fig : proms]c was obtained at the dutch open telescope , which is operated by utrecht university in the observatorio del roque de los muchachos of the instituto de astrofsica de canarias , la palma , spain . according to the model presented in this paper , the vertical threads in hedgerow prominences are supported by tangled magnetic fields that pervade the dense threads and their local surroundings . in this section we construct a cylindrical model of this tangled field . we use a cylindrical coordinate system @xmath191 with @xmath275 the distance from the cylinder axis ( @xmath193 where @xmath190 is the cylinder radius ) , @xmath312 the azimuthal angle , and @xmath175 the height along the axis . we assume that the radial component of the field vanishes at the cylinder wall , @xmath192 , so the field lines are confined to the interior of the cylinder . it follows that the axial magnetic flux must be constant along the cylinder : @xmath313 in cylindrical coordinates the force - free condition reads : @xmath314 we again take @xmath2 to be constant ( lfff ) . the general solution of the above equations with the boundary condition @xmath192 can be written as a superposition of discrete eigenmodes enumerated by an index @xmath115 : @xmath315 where @xmath316 is the mode amplitude ; the functions @xmath317 , @xmath318 and @xmath319 describe the radial dependence of each mode ; @xmath320 is the axial wavenumber ; @xmath321 is the azimuthal wavenumber ( non - negative integer ) ; and @xmath322 is the phase . here @xmath227 refers to a fully symmetric mode with @xmath323 ( also known as the lundquist mode ) , while @xmath324 refers to modes that have either a @xmath312 or @xmath175 dependence . the latter modes only exist under certain conditions ( see below ) . inserting expressions ( [ eq : br ] ) , ( [ eq : bp ] ) and ( [ eq : bz ] ) into equations ( [ eq : f1 ] ) , ( [ eq : f2 ] ) and ( [ eq : f3 ] ) , we find that the radial dependencies can be expressed in terms of bessel functions : @xmath325 where @xmath326 is the radial wavenumber , @xmath327 is the bessel function of order @xmath321 , and @xmath328 is its derivative . the boundary condition at the cylindrical wall requires @xmath329 for all modes . introducing a parameter @xmath108 in the range @xmath330 $ ] such that @xmath331 and @xmath332 , we obtain the following equation for @xmath108 : @xmath333 where @xmath194 . the roots of this equation can be found numerically for any given values of @xmath334 and @xmath127 . depending on the value of @xmath334 , the equation may have one or more solutions : 1 . there always exists at least one solution , the axisymmetric mode ( @xmath335 ) with @xmath336 . in this case @xmath337 and @xmath338 , so this mode is invariant with respect to translation along the @xmath175 axis ( lundquist mode ) . 2 . there may be additional axisymmetric modes that are not invariant to translation . in the case @xmath339 and @xmath340 , equation ( [ eq : bc ] ) yields @xmath341 , where @xmath342 is a root of @xmath343 ( @xmath344 , @xmath345 , @xmath346 , etc . ) . solutions exist only when @xmath347 ; for @xmath334 in the range @xmath348 $ ] there exists one such solution with @xmath349 ; for @xmath334 in the range @xmath350 $ ] there exist two solution , etc . since @xmath351 is symmetric with respect to @xmath352 , each solution @xmath108 also has a second solution @xmath353 , but the magnetic structure of these solutions is the same , so we do not count it as a separate mode . 3 . for higher values of @xmath334 there exist non - axisymmetric solutions ( @xmath354 ) . the first modes with @xmath355 occur at @xmath356 ; these `` kink '' modes are apparently more easily excited than the axisymmetric modes with @xmath357 . the first modes with @xmath358 occur at @xmath359 . for a given value of @xmath334 , we systematically find all roots of equation ( [ eq : bc ] ) , starting with @xmath360 and then increasing @xmath127 until no more roots are found . figure [ fig : modes ] shows @xmath108 as function of @xmath334 . note that the number of modes @xmath101 increases with @xmath334 . we now compute the axial flux and magnetic energy density . inserting equations ( [ eq : bz ] ) and ( [ eq : fz ] ) into equation ( [ eq : phi1 ] ) , we find that only the lundquist mode contributes to the axial magnetic flux : @xmath361 where we used @xmath362 for axisymmetric modes . the mean magnetic energy density is defined by @xmath363 where @xmath142 denotes an average over @xmath312 and @xmath175 . let @xmath364 and @xmath365 denote two _ different _ modes , then averages of cross products such as @xmath366 @xmath367 vanish . therefore , the magnetic energy can be written as a sum over individual modes , @xmath368 , where @xmath369 r ~dr , \label{eq : e1 } \\ e_n & = & \frac{\tilde{b}_n^2}{8 \pi r^2 } \int_0^r \left [ f_{r , n}^2 ( r ) + f_{\phi , n}^2 ( r ) + f_{z , n}^2 ( r ) \right ] r ~dr ~~~~~ \mbox{for } ~ n > 1 . \label{eq : en}\end{aligned}\ ] ] for simplicity we consider _ equipartition _ tangled fields in which the various modes @xmath115 of the lfff have equal magnetic energy . this implies @xmath370 , the same for all modes ( including the lundquist mode ) , which provides a relationship between the mode amplitudes @xmath316 . the phase angles @xmath322 of the non - axisymmetric modes are assigned random values in the range @xmath143 $ ] . this results in a tangled magnetic field @xmath91 described by equations ( [ eq : br ] ) , ( [ eq : bp ] ) and ( [ eq : bz ] ) . bommier , v. , & leroy , j .- 1998 , in _ new perspectives on solar prominences _ , iau colloq . 167 , eds . d. webb , d. rust , b. schmieder . series , vol . 150 ( astron . soc . of the pacific , san francisco ) , p. 434 leroy , j. l. 1989 , in dynamics and structure of quiescent solar prominences ; proc . of workshop , palma de mallorca , spain , nov . 1987 , ed . e. r. priest ( dordrecht , kluwer academic publishers ) , p. 77 - 113 . lin , y. , martin , s. f. , engvold , o. 2008 , in _ subsurface and atmospheric influences of solar activity _ , ed . r. howe , r.w . komm , k.s . balasubramaniam , g.j.d . series , vol . 383 ( astron . soc . of the pacific , san francisco ) , p. 235 lpez ariste , a. , aulanier , g. 2007 , in _ the physics of chromospheric plasmas _ , ed . by p. heinzel , i. dorotovic , r.j . series , vol . 368 ( astron . soc . of the pacific , san francisco ) , p. 291 malherbe , j .- 1989 , in _ dynamics and structure of quiescent solar prominences _ , proc . of workshop , palma de mallorca , spain , nov . 1987 , ed . e.r . priest ( dordrecht , kluwer academic publishers ) , p. 115 paletou , f. 2008 , in _ sf2a-2008 : proc . of the annual meeting of the french society of astronomy and astrophysics _ , ed . c. charbonnel , f. combes , r. samadi , p. 559 . available online at http://proc.sf2a.asso.fr

we develop numerical homotopy algorithms for solving systems of polynomial equations arising from the classical schubert calculus . these homotopies are optimal in that generically no paths diverge . for problems defined by hypersurface schubert conditions we give two algorithms based on extrinsic deformations of the grassmannian : one is derived from a grbner basis for the plcker ideal of the grassmannian and the other from a sagbi basis for its projective coordinate ring . the more general case of special schubert conditions is solved by delicate intrinsic deformations , called pieri homotopies , which first arose in the study of enumerative geometry over the real numbers . computational results are presented and applications to control theory are discussed .

alg-geom9706004

suppose we are given linear subspaces @xmath0 of @xmath1 with @xmath2 and @xmath3 . our problem is to find all @xmath4-dimensional linear subspaces of @xmath1 which meet each @xmath5 nontrivially . when the given linear subspaces are in general position , the condition @xmath6 guarantees that there is a finite number @xmath7 of such @xmath4-planes . the classical schubert calculus @xcite gives the following recipe for computing the number @xmath8 . let @xmath9 be indeterminates with @xmath10 . for each integer sequence @xmath11 we define the following polynomial : @xmath12 here @xmath13 and @xmath14 if @xmath15 or @xmath16 . let @xmath17 be the ideal in @xmath18 $ ] generated by those @xmath19 with @xmath20 and @xmath21 . the quotient ring @xmath22/i$ ] is the cohomology ring of the grassmannian of @xmath4-planes in @xmath1 . it is artinian with one - dimensional socle in degree @xmath23 . in the socle we have the relation @xmath24 thus we can compute the number @xmath8 by normal form reduction modulo any grbner basis for @xmath17 . more efficient methods for computing in the ring @xmath25 are implemented in the maple package sf @xcite . in the important special case @xmath26 there is an explicit formula for @xmath8 : @xmath27 the integer on the right hand side is the degree of the grassmannian in its plcker embedding . this formula is due to @xcite ; see also ( * ? ? ? * xiv.7.8 ) and section 2.3 below . the objective of this paper is to present semi - numerical algorithms for computing all @xmath8 solution planes from the input data @xmath0 . this amounts to solving certain systems of polynomial equations . our algorithms are based on the paradigm of _ numerical homotopy methods _ @xcite . homotopy methods have been developed for the following classes of polynomial systems : 1 . complete intersections in affine or projective spaces @xcite , 2 . complete intersections in products of projective spaces @xcite , 3 . complete intersections in toric varieties @xcite . in these cases the number of paths to be traced is optimal and equal to the standard combinatorial bounds : 1 . the bzout number (= the product of the degrees of the equations ) 2 . the generalized bzout number for multihomogeneous systems 3 . the bkk bound @xcite (= mixed volume of the newton polytopes ) none of these known homotopy methods is applicable to our problem , as the following simple example shows : take @xmath28 , @xmath29 , and @xmath30 , that is , we seek the @xmath31-planes in @xmath32 which meet six general @xmath33-planes nontrivially . by formula ( 1.3 ) there are @xmath34 solutions . formulating this in plcker coordinates gives @xmath35 homogeneous equations in ten variables , the five quadrics in display ( [ i23 ] ) below and six linear equations ( [ sixlin ] ) . a formulation in local coordinates ( [ sagbieqs ] ) has 6 quadratic equations in 6 unknowns , giving a bzout bound of 64 . these 6 equations all have the same newton polytope , which has normalized volume 17 , giving a bkk bound of 17 . in section 2 we give two homotopy algorithms which solve our problem in the special case @xmath36 , when the number of solutions equals ( 1.3 ) . the first algorithm is derived from a grbner basis for the plcker ideal of a grassmannian and the second from a sagbi basis for its projective coordinate ring . ( see @xcite or ( * ? ? ? 11 ) for an introduction to sagbi bases ) . both the _ grbner homotopy _ and _ sagbi homotopy _ are techniques for finding linear sections of grassmannians in their plcker embedding . in section 3 we address the general case of our problem , that is , we describe a numerical method for solving the polynomial equations defined by _ special schubert conditions_. this is accomplished by applying a sequence of delicate intrinsic deformations , called _ pieri homotopies _ , which were introduced in @xcite . pieri homotopies first arose in the study of enumerative geometry over the real numbers @xcite . for the experts we remark that it is an open problem to find _ littlewood - richardson homotopies _ , which would be relevant for solving polynomial equations defined by general schubert conditions . a main challenge in designing homotopies for the schubert calculus is that one is not dealing with complete intersections : there are generally more equations than variables . in section 4 we discuss some of the numerical issues arising from this challenge , and how we propose to resolve them . in section 5 we discuss applications of these algorithms to control theory and real enumerative geometry . finally , in section 6 we present computational results . in closing the introduction let us emphasize that all homotopies described in this paper are optimal in the sense the number of paths to be traced equals the number @xmath8 . this means that for generic input data @xmath0 no paths diverge . the set of @xmath4-planes in @xmath1 , @xmath37 , is called the _ grassmannian of @xmath4-planes in @xmath1_. this complex manifold of dimension @xmath23 is naturally a subvariety of the complex projective space @xmath38 . to see this , represent a @xmath4-plane in @xmath1 as the column space of an @xmath39-matrix @xmath40 . the _ plcker coordinates _ of that @xmath4-plane are the maximal minors of @xmath41 , indexed by the set @xmath42}{p}$ ] of sequences @xmath43 : @xmath44\ \longrightarrow\ \det \left[\begin{array}{ccc } x_{\alpha_1\,1}&\cdots & x_{\alpha_1\,p}\\ \vdots&\ddots&\vdots\\ x_{\alpha_p\,1}&\cdots & x_{\alpha_p\,p}\end{array}\right ] .\ ] ] this section deals with the `` @xmath45 '' case of the problem stated in the introduction . given @xmath23 general @xmath46-planes @xmath47 , we wish to find all @xmath4-planes @xmath41 which meet @xmath47 nontrivially . this geometric condition translates into linear equations in the plcker coordinates : represent @xmath41 as an @xmath48-matrix as above , represent @xmath5 as an @xmath49-matrix , and form the @xmath50-matrix @xmath51 $ ] . then @xmath52 \,\,\,= \,\,\ , 0 .\ ] ] laplace expansion with respect to the first @xmath4 columns gives @xmath53 \quad = \quad \sum_{\alpha\in\binom{[m+p]}{p } } c_{\alpha}^{i } \cdot [ \alpha],\ ] ] where @xmath54 is the correctly signed maximal minor of @xmath5 complementary to @xmath55 . hence our problem is to solve @xmath23 linear equations ( [ lineqs ] ) on the grassmannian . the number of solutions is the degree of the grassmannian in its plcker embedding , which is given in ( 1.3 ) . the grassmannian is represented either _ implicitly _ , as the zero set of polynomials in the plcker coordinates , or _ parametrically _ , as the image of the polynomial map ( [ minors ] ) . these two representations lead to two different numerical homotopies . the implicit representation gives the grbner homotopy in section [ grobnerhomotopy ] while the parametric representation gives the sagbi homotopy in section [ sagbihomotopy ] . the first is conceptually simpler but the second is more efficient . in both methods the number of paths to be traced equals the optimal number in ( [ grassdeg ] ) . we describe the two approaches for the case @xmath56 . the grassmannian of 2-planes in @xmath32 has dimension 6 and is embedded into @xmath57 . its degree ( [ grassdeg ] ) is five . the grbner homotopy works directly in the ten plcker coordinates : @xmath58,\ [ 13],\ [ 14],\ [ 15],\ [ 23],\ [ 24],\ [ 25],\ [ 34],\ [ 35],\ [ 45].\ ] ] the ideal @xmath59 of the grassmannian in the plcker embedding is generated by five quadrics : @xmath60[23 ] } \ -\ [ 13][24]\ + \ [ 12][34 ] , \\ \underline{[15][23]}\ -\ [ 13][25]\ + \ [ 12][35 ] , \\ \underline{[15][24]}\ -\ [ 14][25]\ + \ [ 12][45 ] , \\ \underline{[15][34]}\ -\ [ 14][35]\ + \ [ 13][45 ] , \\ \underline{[25][34]}\ -\ [ 24][35]\ + \ [ 23][45 ] . \end{array}\ ] ] this set is the reduced grbner basis for @xmath59 with respect to any term order which selects the underlined terms as leading terms ( see proposition 2.1 below ) . our problem is to compute all @xmath31-planes which meet six sufficiently general @xmath33-planes @xmath61 nontrivially . this amounts to solving ( [ i23 ] ) together with six linear equations @xmath62 \ , + \ , c_{13}^{i } \cdot [ 13 ] \ , + \ , c_{14}^{i } \cdot [ 14 ] \ , + \ , c_{15}^{i } \cdot [ 15 ] \ , + \ , c_{23}^{i } \cdot [ 23 ] \phantom{\ , = \ , 0 } \\ \ , + \ , c_{24}^{i } \cdot [ 24 ] \ , + \ , c_{25}^{i } \cdot [ 25 ] \ , + \ , c_{34}^{i } \cdot [ 34 ] \ , + \ , c_{35}^{i } \cdot [ 35 ] \ , + \ , c_{45}^{i } \cdot [ 45 ] \,\,\ , = \,\,\ , 0 , \end{array}\ ] ] for @xmath63 . this is an overdetermined system of @xmath35 equations in @xmath64 homogeneous variables . to solve it we introduce a parameter @xmath65 into ( [ i23 ] ) as follows : @xmath66[23]\ -\ t\;\,\cdot [ 13][24]\ + \ t^2 \cdot [ 12][34 ] \quad = \quad 0 , \\ { [ 15][23]\ -\ t^2 \cdot [ 13][25]\ + \ t^4 \cdot [ 12][35]}\quad = \quad 0 , \\ { [ 15][24]\ -\ t\;\,\cdot [ 14][25]\ + \ t^5 \cdot [ 12][45]}\quad = \quad 0 , \\ { [ 15][34]\ -\ t^2 \cdot [ 14][35]\ + \ t^4 \cdot [ 13][45]}\quad = \quad 0 , \\ { [ 25][34]\ -\ t\;\,\cdot [ 24][35]\ t^2 \cdot [ 23][45 ] \quad = \quad 0 . } \end{array } \eqno ( \ref{i23}')\ ] ] we call ( [ i23]@xmath67 ) the _ grbner homotopy _ because this is an instance of the flat deformation which exists for any grbner basis ; see ( * ? ? ? * theorem 15.17 ) . the flatness of this deformation ensures that , for almost every complex number @xmath65 , the combined system @xmath68 has five roots . for @xmath69 the equations ( [ i23]@xmath67 ) are square - free monomials . we decompose their ideal : @xmath70[23],\ , [ 15][23],\ , [ 15][24],\ , [ 15][34],\ , [ 25][34 ] \,\ , \rangle \hspace{2.2 in } \\\hspace{1.3 in } = \quad \ , \langle \ , [ 23 ] , [ 24 ] , [ 34 ] \,\rangle \,\ , \cap \,\ , \langle \ , [ 15 ] , [ 23 ] , [ 34 ] \,\rangle \,\ , \cap \,\ , \langle \ , [ 15 ] , [ 23 ] , [ 25 ] \,\rangle \\ \hspace{1.8 in } \,\ , \cap \,\ , \,\ , \langle \ , [ 14 ] , [ 15 ] , [ 34 ] \,\rangle \,\ , \ , \cap \,\,\ , \langle \ , [ 14 ] , [ 15 ] , [ 25 ] \,\rangle . \end{array}\ ] ] the five distinct solutions for @xmath71 are computed by setting each listed triple of variables to zero and then solving the six linear equations ( [ sixlin ] ) in the remaining seven variables . thereafter we trace the five solutions from @xmath71 to @xmath72 by numerical path continuation . at @xmath72 we get the five solutions to our original problem . we next describe the sagbi homotopy . for this we choose the local coordinates @xmath73\ ] ] on the grassmannian . substituting @xmath41 into ( [ sixlin ] ) we get six polynomials in six unknowns : @xmath74 to solve these six equations we introduce a parameter @xmath65 as follows : @xmath75 the system ( [ sagbieqs]@xmath67 ) has five complex roots for almost all @xmath76 . for @xmath71 we get a _ generic unmixed sparse system _ ( in the sense of @xcite ) with support @xmath77 we identify @xmath78 with a set of ten points in @xmath79 . their convex hull _ conv_@xmath80 is a @xmath81-dimensional polytope with normalized volume five . we can therefore solve @xmath82 for @xmath71 using the homotopy method in @xcite or @xcite , provided the input data @xmath83 are sufficiently generic . tracing the five roots from @xmath71 to @xmath72 by numerical path continuation , we obtain the five solutions to our original problem . we next describe a quadratic grbner basis for the defining ideal of @xmath37 . let @xmath84 be the polynomial ring over @xmath85 in the variables @xmath86\,$ ] where @xmath87 } { p}$ ] . we define a partial order on these variables as follows : @xmath86 \leq [ \beta]\,$ ] if and only if @xmath88 for @xmath89 . this partially ordered set is called _ young s poset_. figure [ youngs_poset ] shows young s poset for @xmath90 . @xmath91 fix any linear ordering on the variables in @xmath84 which refines the ordering in young s poset , and let @xmath92 denote the induced degree reverse lexicographic term order on @xmath84 . let @xmath93 be the ideal of polynomials in @xmath84 which vanish on the grassmannian , that is , @xmath93 is the ideal of algebraic relations among the maximal minors of a generic @xmath48-matrix @xmath41 . the grbner homotopy is based on the following well - known result . [ prop : inimp ] the initial ideal @xmath94 is generated by all quadratic monomials @xmath95[\,\beta\ , ] \,$ ] where @xmath96 and @xmath97 for some @xmath98 . in other words , @xmath94 is generated by products of incomparable pairs in young s poset . let @xmath99 denote the set of all maximal chains in young s poset . for example , @xmath100,[13],[14],[15],[25],[35],[45]\ } \ , , \,\ , \\ & & \hspace{6.6pt}\{[12],[13],[14],[24],[34],[35],[45]\}\,,\,\ , \{[12],[13],[23],[24],[34],[35],[45]\}\,,\\ & & \hspace{5.2pt}\left . \{[12],[13],[23],[24],[25],[35],[45]\}\,\right\}\end{aligned}\ ] ] a standard result in combinatorics @xcite states that the cardinality of @xmath99 equals the number ( [ grassdeg ] ) . from proposition [ prop : inimp ] we read off the following prime decomposition which generalizes ( [ primedec ] ) : @xmath101\,\ , : \,\ , [ \alpha ] \not\in c \,\rangle .\ ] ] for a proof of proposition [ prop : inimp ] see @xcite or ( * ? ? ? * theorem ( 4.3 ) ) or @xcite . in these references one finds an explicit minimal grbner basis for @xmath93 , which is classically called the set of _ straightening syzygies_. in the special case @xmath29 the straightening syzygies coincide with the reduced grbner basis : if @xmath29 then the reduced grbner basis of @xmath93 consists of the three - term plcker relations @xmath102[k j ] } - [ i k][j l ] + [ i j][k l ] \,$ ] where @xmath103 . for @xmath104 the straightening syzygies and the reduced grbner basis do not coincide , and they are complicated to describe . for our purposes the following coarse description suffices . let @xmath105 be the set of all quadratic monomials in @xmath84 which do not lie in @xmath106 . the reduced grbner basis consists of elements of the form @xmath107[\beta ] \quad - \quad \sum_{[\gamma][\delta ] \in std } e^{\alpha,\beta}_{\gamma,\delta } \cdot [ \gamma][\delta]\ ] ] where @xmath108[\beta]$ ] runs over all generators of @xmath106 . the constants @xmath109 are integers which can be computed by substituting ( [ minors ] ) into ( [ redgb ] ) and solving linear equations . the term order @xmath92 can be realized for the ideal @xmath93 by the following choices of weights . we define the _ weight _ of the variable @xmath108 = [ \alpha_1 \alpha_2 \cdots \alpha_p]$ ] to be @xmath110 if we replace each variable @xmath86\,$ ] in ( [ redgb ] ) by @xmath86 \cdot t^{v_\alpha}\,$ ] and clear @xmath65-denominators afterwards , then we get the _ grbner homotopy _ : @xmath111[\beta ] \quad - \quad \sum_{[\gamma][\delta ] \in std } e^{\alpha,\beta}_{\gamma,\delta } \cdot [ \gamma][\delta ] \cdot t^{v_\gamma + v_\delta - v_\alpha - v_\beta } \eqno ( \ref{redgb}')\ ] ] it can be checked that all exponents @xmath112 appearing here are positive integers . the special case @xmath56 is presented in ( [ i23]@xmath67 ) . in the grbner homotopy algorithm , we first solve systems of @xmath23 linear equations , one for each chain @xmath113 . these systems consist of the @xmath23 equations ( [ lineqs ] ) , one for each @xmath5 , and the @xmath114 equations @xmath111\qquad\mbox{for}\qquad [ \alpha]\not\in c,\ ] ] suggested by the prime decomposition ( [ genprimedec ] ) . once this is accomplished , we trace each of these @xmath8 solutions from @xmath71 to @xmath72 in the grbner homotopy ( [ i23]@xmath67 ) . clearly , the weights @xmath115 of ( [ weights ] ) are not best possible for any specific value of @xmath46 and @xmath4 . smaller weights can be found using linear programming , as explained e.g. in the proof of ( * ? ? ? * proposition 1.11 ) . another method would be to adapt the `` dynamic '' approach in @xcite to our situation . this is possible since the grbner basis in ( [ redgb ] ) is reverse lexicographic : first deform the lowest variable to zero , then deform the second lowest variable to zero , then the third lowest variable , and so on . let @xmath116 be an @xmath48-matrix of indeterminates . we identify the coordinate ring of the grassmannian with the @xmath85-algebra generated by the @xmath117-minors of @xmath41 . call this algebra @xmath118 and write @xmath108 ( x_{ij } ) $ ] for the minor indexed by @xmath55 . reinterpreting classical results in @xcite , it was shown in ( * ? ? ? * theorem 3.2.9 ) that these generators form a _ sagbi basis _ with respect to the lexicographic term order induced from @xmath119 . this means that the _ initial algebra _ @xmath120\,$ ] is generated by the main diagonal terms of the @xmath117-minors , @xmath121(x_{ij } ) \bigr ) \quad = \quad x_{\alpha_1,1 } \,x_{\alpha_2,2 } \,x_{\alpha_3,3 } \cdots \cdots x_{\alpha_m , m } .\ ] ] the resulting flat deformation can be realized by replacing @xmath122 with @xmath123 for @xmath124 in the matrix @xmath41 . if we expand @xmath86 ( x_{ij } t^{(i-1)(p - j ) } ) \,$ ] as a polynomial in @xmath65 , then the lowest term equals @xmath125 times the main diagonal monomial ( [ inr ] ) , where @xmath126 in what follows we multiply that polynomial by @xmath127 . for any @xmath76 consider the algebra @xmath128 ( x_{ij } t^{(i-1)(p - j ) } ) \,\,:\,\ \alpha \in { \textstyle \binom{[m+p]}{p } } \,\right].\ ] ] then @xmath129 is the coordinate ring of the grassmannian , and @xmath130 is the algebra generated by the monomials ( [ inr ] ) . this is a flat deformation of @xmath85-algebras ; see @xcite and @xcite . the _ sagbi homotopy _ is the following system of @xmath23 equations : @xmath131}{p } } c^{i}_\alpha \cdot t^{-w_\alpha } \cdot [ \alpha ] ( x_{ij } t^{(i-1)(p - j ) } ) \quad = \quad 0 \quad \qquad ( i=1,\ldots , mp).\ ] ] we reduce the number of variables to @xmath23 by introducing local coordinates as follows : @xmath132 for @xmath89 and @xmath133 for @xmath134 or @xmath135 . our original problem is to solve the system ( [ sagbih ] ) for @xmath72 . the flatness of the family of algebras @xmath136 guarantees that the system ( [ sagbih ] ) has the same finite number of complex solutions ( counting multiplicities ) for almost every @xmath76 . for @xmath71 we get a system of linear equations in @xmath137 : @xmath138}{p } } c^{i}_\alpha \cdot x_{\alpha_1,1 } x_{\alpha_2,2 } x_{\alpha_3,3 } \cdots x_{\alpha_p , p } \quad \qquad ( i=1,\ldots , mp).\ ] ] in order to solve these equations we apply the symbolic - numeric algorithm in @xcite , while taking advantage of the following combinatorial structures described in ( * ? ? ? * remark 11.11 ) . the common newton polytope of the equations ( [ r0eqs ] ) equals the _ order polytope _ of the product of an @xmath46-chain and a @xmath4-chain ( sturmfels , remark 11.11 ) . we have the following combinatorial result . the following five numbers coincide : * the right hand side of ( [ grassdeg ] ) , * the number of linear extensions of the product of a @xmath46-chain and a @xmath4-chain , * the number of maximal chains in young s poset , * the normalized volume of the order polytope , and * the number of roots in @xmath139 of a generic system ( [ r0eqs ] ) . the equality of ( 4 ) and ( 5 ) is a special case of kouchnirenko s theorem @xcite , as all the equations have the same newton polytope . the order polytope has a distinguished unimodular regular triangulation with simplices indexed by the chains in young s poset . this regular triangulation is induced by the system of weights given in ( [ weights ] ) . we may use these weights to define a numerical homotopy for finding all isolated solutions of ( [ r0eqs ] ) . once this is accomplished , we trace these roots from @xmath71 to @xmath72 in the homotopy ( [ sagbih ] ) . we first describe a purely combinatorial method for computing the number @xmath8 of solution planes . instead of the algebraic relation ( 1.2 ) we shall make use of young s poset which was introduced in section 2.2 . a cover @xmath108\lessdot [ \beta]$ ] in young s poset determines a unique index @xmath140 for which @xmath141 a chain @xmath142\lessdot [ \alpha^1]\lessdot \cdots\lessdot[\alpha^l]\,$ ] in young s poset is _ increasing at @xmath143 _ if either @xmath144 , or else @xmath145 and @xmath146 . for instance , @xmath147\lessdot [ 13]\lessdot [ 14]\lessdot [ 24]$ ] is increasing at 1 and 2 , but decreasing at 3 . given positive integers @xmath148 , the _ pieri tree _ @xmath149 consists of all chains of length @xmath150 in young s poset which start at the bottom element @xmath151 $ ] and which increase everywhere , except possibly at @xmath152 . here we include all initial segments of such chains and we order the chains by inclusion . label a node in the pieri tree by the endpoint of the chain which that node represents . then the sequence of labels from the root to that node is the chain which that node represents . for example , here is @xmath153 when @xmath154 , @xmath29 : @xmath155 to compute the number @xmath8 , partition the integer sequence @xmath156 into three parts @xmath148 , @xmath157 , and @xmath158 . then @xmath8 is the number of pairs @xmath159 where @xmath118 is a leaf of @xmath149 , @xmath84 is a leaf of @xmath160 , and the endpoints @xmath108 $ ] of @xmath118 and @xmath161 $ ] of @xmath84 satisfy _ pieri s condition _ : @xmath162 call this set of pairs @xmath163 . for instance , @xmath164 for the sequence @xmath165 with @xmath166 : of the 9 pairs @xmath159 of leaves of @xmath153 , only the following 6 satisfy ( [ eq : pieri ] ) : ( here we represent a leaf by its label . ) @xmath167,[34]),\ ( [ 34],[25]),\ ( [ 25],[25]),\ ( [ 25],[16]),\ ( [ 16],[25 ] ) , \ ( [ 16],[16 ] ) \right\}.\ ] ] this combinatorial rule for the number @xmath8 gives the same answer as the algebraic rule ( 1.2 ) because the pieri tree and pieri s condition ( [ eq : pieri ] ) encode the structure of the cohomology ring @xmath25 with respect to its schur basis @xmath168 ; see @xcite or @xcite . specifically , @xmath169 where @xmath170 is the number of leaves in @xmath149 with label @xmath171\quad : = \quad [ \lambda_p+1,\lambda_{p-1}+2,\ldots,\lambda_1+p].\ ] ] the numbers @xmath170 are called _ kostka numbers_. in @xmath25 we calculate @xmath172 we evaluate this expression with pieri s formula ( proposition [ prop : triple ] below ) : if @xmath173 , then @xmath174 is either @xmath175 or 0 depending upon whether or not @xmath176 the condition ( [ eq : pieri]@xmath67 ) is equivalent to ( [ eq : pieri ] ) under the transformation @xmath177 $ ] . these methods correctly enumerate the @xmath4-planes which meet each @xmath0 nontrivially because , under the isomorphism between @xmath25 and the cohomology ring of @xmath37 , the indeterminate @xmath178 corresponds to the cohomology class poincar dual to @xmath179 , the set of @xmath4-planes which meet @xmath5 nontrivially . moreover , @xmath175 represents the class dual to a point . the pieri tree models certain intrinsic deformations ( described in 3.3 and [ explicit_pieri ] ) of the grassmannian which establish this isomorphism , and which we shall use for computing the @xmath4-planes which meet each @xmath0 nontrivially . for vectors @xmath180 in @xmath1 , let @xmath181 be their linear span . fix the columns @xmath182 of the identity matrix as a standard basis for @xmath1 . for @xmath183}{p}$ ] , define @xmath184}{p}$ ] by @xmath185 . a sequence @xmath183}{p}$ ] determines a _ schubert variety _ @xmath186 this variety has complex codimension @xmath187 . similarly define @xmath188 for a linear subspace @xmath189 of @xmath1 of dimension @xmath190 , define the _ special schubert variety _ @xmath191 this has codimension @xmath158 . if @xmath192 , then @xmath193}$ ] . the special schubert variety @xmath194 is cut out by the system of @xmath195 polynomial equations : @xmath196 \ , \mbox{are zero,}\ ] ] where @xmath197 is represented by a @xmath39-matrix . the laplace expansion of these equations in terms of the plcker coordinates of @xmath41 define @xmath194 as a subscheme of @xmath37 . these equations are redundant : select @xmath198 rows of @xmath199 $ ] such that the corresponding maximal minor of @xmath189 is invertible . consider the set of maximal minors of @xmath199 $ ] which cover all the rows selected . this gives @xmath200 polynomial equations which generate the same ideal as all @xmath195 minors of @xmath199 $ ] . for a purely set - theoretic ( but not scheme - theoretic ) representation of @xmath194 a further substantial reduction in the number of equations is possible using the results of @xcite . an intersection @xmath201 of subvarieties is _ generically transverse _ if every component of @xmath201 has an open subset along which @xmath202 and @xmath203 meet transversally . in this case the following identity in the cohomology ring holds : @xmath204\quad = \quad [ y]\cdot[z],\ ] ] where @xmath205 $ ] denotes the cycle class of a subvariety @xmath206 . by kleiman s transversality theorem @xcite , subvarieties of @xmath207 in general position meet generically transversally . transversality and generic transversality coincide when @xmath201 is finite . [ prop : triple ] let @xmath208}{p}$ ] with @xmath209 and let @xmath189 be a linear subspace of @xmath1 with dimension @xmath190 none of whose plcker coordinates vanish . then the intersection @xmath210 either is transverse consisting of a single @xmath4-plane or is empty , depending upon whether or not ( [ eq : pieri ] ) holds . proof and algorithm : the intersection @xmath211 is nonempty if and only if @xmath212 for @xmath213 . these are the weak inequalities in ( [ eq : pieri ] ) . we shall assume that they hold in what follows . the @xmath4-planes in @xmath211 are represented by @xmath39-matrices @xmath214 such that @xmath215 consider the nonzero coordinate subspaces @xmath216 , set @xmath217 , and note that @xmath218 . from ( [ eq : alpha ] ) we see that @xmath219 implies @xmath220 and hence @xmath221 . therefore the triple intersection ( [ triple_int ] ) is nonempty only if the following equivalent conditions hold : @xmath222 @xmath223 in this case we determine@xmath224 by computing vectors @xmath225 such that @xmath226 . ( this computation is the `` algorithm '' part in this proof . ) the desired @xmath4-plane @xmath41 satisfies @xmath227 , and , in view of ( [ eq : alpha ] ) , this implies @xmath228 . transversality of ( [ triple_int ] ) is verified in local coordinates for @xmath229 by considering @xmath230 independent linear forms which vanish on @xmath189 . for @xmath231}{m}$ ] define @xmath232 by @xmath233 for @xmath234 . then @xmath235 represents the cycle class of @xmath236 ( equivalently , of @xmath237 ) . if @xmath238 , then @xmath239 is the cycle class of @xmath194 . . then proposition [ prop : triple ] implies the following identity in @xmath240 : @xmath241 this identity implies ( via poincar duality ) that @xmath242 the sum over all @xmath243 with @xmath244 for which @xmath245 satisfy pieri s condition ( [ eq : pieri ] ) . call this set @xmath246 , which is also the set of endpoints of increasing chains of length @xmath158 in young s poset that begin at @xmath55 . this last form has geometric content . in @xcite , explicit deformations were given that transform the irreducible intersection @xmath247 into the cycle @xmath248 . moreover , the branching of the components of the cycles in these deformations reflects the branching among these increasing chains above @xmath55 . this process may be iterated to transform an intersection of several special schubert varieties into a sum of triple intersections of the form ( [ triple_int ] ) , indexed by pairs @xmath249 . from this sum , we obtain a set of start solutions indexed by _ sols_. also , every intermediate cycle in these deformations consists of the same number ( counting multiplicities ) of @xmath4-planes . the pieri homotopy begins with one of the start solutions and uses numerical path continuation to trace the sequence of curves defined by these deformations which connect that start solution to a solution of the original problem . given linear subspaces @xmath0 in general position with @xmath250 and @xmath251 , first partition @xmath0 into three lists : @xmath252 where @xmath253 , @xmath254 , and @xmath238 . construct the pieri trees @xmath149 and @xmath160 , and form the set _ sols_. change coordinates so that @xmath255 and @xmath256 . set @xmath257 . given a chain @xmath118 in the pieri tree and a positive integer @xmath258 , let @xmath259 be the @xmath258th element in that chain , or , if @xmath258 exceeds the length of @xmath118 , then let @xmath259 be the endpoint of @xmath118 . for each @xmath260 and @xmath258 from @xmath261 to @xmath262 we shall construct ( in definition [ def : zrkt ] below ) one - parameter families @xmath263 and @xmath264 of pure - dimensional subvarieties of @xmath207 with the following properties : 1 . @xmath265 and @xmath266 . 2 . for @xmath71 or @xmath267 and each @xmath258 , @xmath268 is transverse and 0-dimensional . 3 . @xmath269 and @xmath270 . 4 . @xmath271 is a component of @xmath272 . likewise , @xmath273 is a component of @xmath274 . @xmath275 and @xmath276 . property 4 is a consequence of proposition [ prop : pieri ] , the others follow from the assumption of genericity and the definition ( definition [ def : zrkt ] ) of the families @xmath263 and @xmath264 . by 2 , the family @xmath277 over @xmath85 whose fibre at general @xmath65 ( including @xmath71 and @xmath72 ) is @xmath278 consists of a finite number of curves . in fact , for general @xmath65 ( including @xmath71 and @xmath72 ) , @xmath277 has the following general form ( see definition [ def : zrkt ] for the precise form ) : @xmath279 where @xmath280 are linear subspaces with @xmath281 , which depend upon @xmath282 , the subspaces @xmath283 , @xmath284 , and at most two of the @xmath285 depend upon @xmath65 . also , @xmath55 and @xmath286 depend upon @xmath287 , and @xmath258 with the typical case being @xmath288 and @xmath289 . the numerical homotopy defined by the curves @xmath277 may be expressed in a parameterization @xmath290 of an open subset of @xmath291 : @xmath292 where @xmath293 . the equations for @xmath277 are then @xmath294\ = \ 0 \qquad i=1,\ldots , s.\ ] ] the curves of @xmath277 define the sequences of homotopies in the pieri homotopy algorithm as follows : for @xmath295 , let @xmath296 be the ( unique by proposition [ prop : triple ] ) @xmath4-plane in @xmath297 . by 3 and 4 , @xmath298 and hence lies on a unique curve in @xmath299 . use numerical path continuation to trace this curve from @xmath71 to @xmath72 to obtain @xmath300 , which is a point of @xmath301 , by 4 . then @xmath300 lies on a unique curve in @xmath302 , which we trace to find @xmath303 . continuing this process , after tracing @xmath261 curves , we obtain @xmath304 , which is a solution to the original system , by 5 . we show shall prove in theorem [ allsols ] that @xmath305 consists of all the solutions to the original system . the cycle @xmath263 will depend upon the choice of a general upper triangular @xmath306-matrix @xmath307 with 1 s on its anti - diagonal , @xmath308 the @xmath258th link in the chain @xmath118 , and the data @xmath283 . the key ingredient of this definition of @xmath263 is the construction of a one - parameter family of linear subspaces @xmath309 in definition [ def : lambda ] , which depends upon @xmath307 . the matrix @xmath307 is fixed throughout the algorithm , its purpose is that @xmath310 equals the span of the first @xmath311 columns of @xmath307 , and these columns are in general position with @xmath182 . the subtle linear degeneracies of @xmath309 as @xmath312 are at the heart of this homotopy algorithm , as well as the explicit proof of pieri s formula ( * ? ? ? * theorem 3.6 ) , which we state below ( proposition [ prop : pieri ] ) . [ def : zrkt ] 1 . if @xmath313 , then set @xmath314 . 2 . otherwise , define @xmath315 by @xmath316 , and set @xmath317 , @xmath318 , and @xmath319 . 1 . if @xmath320 and @xmath321 , then @xmath322 , and we set @xmath323 2 . otherwise , let @xmath309 be the 1-parameter family of linear subspaces given by definition [ def : lambda ] , where we let @xmath324 and @xmath325 . + if @xmath326 , then @xmath327 , @xmath328 , and we set @xmath329 if @xmath320 , let @xmath311 be maximal such that @xmath330 . then @xmath331 as @xmath332 is case 2(a ) . set @xmath333 we define @xmath264 similarly , but with the matrix @xmath307 replaced by a lower triangular matrix with 1 s on its diagonal , and @xmath334 replaced by @xmath335 . [ def : lambda ] let @xmath307 be an upper triangular matrix with 1 s on the anti - diagonal and @xmath336 be a general @xmath337-plane , represented as a @xmath338-matrix with columns @xmath339 . construct a @xmath39-matrix @xmath340 as follows : reverse the last @xmath341 columns of @xmath307 , then remove the columns indexed by @xmath342 . for each @xmath343 , define a one - parameter family of @xmath338-matrices @xmath309 for @xmath344 as follows : 1 . if @xmath326 , then the @xmath345th column of @xmath346 is @xmath347 . 2 . for @xmath348 , the @xmath345th column of @xmath309 is @xmath349 we give an example illustrating these definitions and the pieri homotopy algorithm . let @xmath350 , and @xmath189 be general 4-planes in @xmath351 . we give a sequence of homotopies @xmath277 for @xmath352 for finding one of the six @xmath31-planes which meet each of the five given 4-planes nontrivially . here , @xmath166 and @xmath353 so that @xmath354 . construct the set _ sols _ as in ( [ sols ] ) . let @xmath249 be the following two sequences : @xmath355\lessdot[13]\lessdot[14]\lessdot[24]\lessdot[25],\qquad s\ : = \ [ 12]\lessdot[13]\lessdot[14]\lessdot[15]\lessdot[16].\ ] ] let @xmath356 be the columns of a @xmath357-identity matrix , a basis for @xmath351 . suppose that @xmath358 and @xmath359 and represent @xmath360 and @xmath361 as @xmath362-matrices . then @xmath363}\cap\omega_l\cap\omega'_{[14]}\cap \omega_{l'}\cap\omega_n$ ] is the set of 2-planes which meet all five linear subspaces nontrivially . we first find the plane @xmath364}\cap\omega'_{[16]}\cap\omega_n$ ] , using the algorithm in the proof of proposition [ prop : triple ] . suppose that @xmath189 has the form @xmath365,\ ] ] where @xmath17 is the @xmath366 identity matrix , and @xmath367 is a @xmath368-matrix . in this case , @xmath369 and @xmath370 , hence @xmath371 . thus the intersection @xmath372 is generated by the third column of @xmath189 , and so @xmath373 is represented by the matrix : @xmath374.\ ] ] following definition [ def : lambda ] , we have : @xmath375 @xmath376 is defined similarly . we describe the families @xmath277 for @xmath352 in local coordinates for @xmath363}\cap \omega'_{[14]}$ ] determined by the sequence @xmath377 $ ] : @xmath378 \ .\ ] ] the family @xmath379 is the constant family @xmath380 . assuming that @xmath381 , which is the @xmath382$]th plcker coordinate of @xmath189 , is non - zero , then @xmath373 may be expressed in these local coordinates : @xmath383 \ .\ ] ] when @xmath384 , first consider the definition of @xmath385 . here we are in case 2(b ) with @xmath386 $ ] and @xmath320 , so that @xmath387 $ ] . since @xmath388 , we have @xmath389}\cap \omega_{\lambda_1(t)}$ ] . for the definition of @xmath390 , we are in case 2(a ) , so that @xmath391}$ ] . hence @xmath392}\cap\omega_{\lambda_1(t ) } \cap\omega'_{[16]}\cap \omega_n.\ ] ] this has 3 linear equations @xmath393 , which describe @xmath394}\cap\omega'_{[16]}$ ] , and 7 non - trivial equations , the vanishing of the maximal minors of @xmath395 $ ] and @xmath396 $ ] , which describe @xmath397 . for @xmath398 , @xmath326 and we are in case 2(b ) for both @xmath399 and @xmath400 so that @xmath401}\cap\omega_{\lambda_0(t ) } \cap\omega'_{[14]}\cap\omega_{\lambda'_0(t)}\cap \omega_n.\ ] ] this has 21 non - trivial equations , the vanishing of the maximal minors of @xmath402 $ ] , @xmath403 $ ] , and @xmath199 $ ] . we describe the pieri deformations linking the families @xmath263 for @xmath258 from @xmath404 to @xmath405 , which establishes property 4 of @xmath263 in [ pieri_alg ] . we also show that the set @xmath305 consists of all the solutions to the original system . consider the dynamic part of @xmath263 , namely whichever of @xmath406 appeared in the definition of @xmath263 . we call this cycle @xmath407 , where @xmath324 and @xmath408 , and @xmath409 . for @xmath410 and @xmath411 write @xmath412 if @xmath413 covers @xmath243 and @xmath414 . this partitions @xmath415 into sets @xmath416 for @xmath417 . * theorem 3.6)[prop : pieri ] let @xmath418 , and @xmath407 be as above . then 1 . for all @xmath65 , @xmath419 is generically transverse . @xmath407 is free of multiplicities for all @xmath65 and irreducible for @xmath420 . 3 . if @xmath421 , then @xmath422 . 4 . if @xmath423 , then @xmath424 . by 3 , the cycle class of @xmath425 is independent of @xmath143 and @xmath65 , and it equals the cycle class of @xmath426 . by 4 , we see that the cycle classes of @xmath427 and @xmath428 coincide , furnishing another proof of pieri s formula . property 4 of @xmath263 follows from assertion 3 . [ allsols ] when @xmath0 are generic , the pieri homotopy algorithm finds all @xmath4-planes which meet each @xmath0 nontrivially . that is , @xmath429 * proof . * note that for any @xmath430 , the families @xmath431 , and @xmath277 depend only upon the initial segments @xmath432 and @xmath433 of @xmath118 and @xmath84 . by construction , the original system @xmath434 coincides with @xmath435 , for any @xmath249 . we inductively construct chains @xmath436 and @xmath437 , and @xmath4-planes @xmath438 for @xmath439 such that @xmath440 and @xmath441 lie on the same curve of @xmath277 . then @xmath442 is the start solution @xmath296 , which shows that @xmath443 . first set @xmath444 to be the unique chain from @xmath445 $ ] to @xmath446 $ ] , and similarly for @xmath447 . then @xmath448 and hence lies on a unique curve in @xmath449 . let @xmath450 be the point on that curve with @xmath71 . by proposition [ prop : pieri ] ( 4 ) , @xmath451 let @xmath452 be the index @xmath243 such that @xmath453 . define @xmath454 similarly . then @xmath455 . in general , suppose that we have constructed @xmath432 , @xmath433 , and @xmath456 . then @xmath438 lies on a unique curve in @xmath277 . let @xmath457 be the point on that curve at @xmath71 . let @xmath315 be minimal subject to @xmath458 and set @xmath408 and @xmath409 . if @xmath459 , then by proposition [ prop : pieri ] ( 3 ) , there is a unique index @xmath460 such that @xmath461 . if @xmath462 then by proposition [ prop : pieri ] ( 4 ) , there is a unique index @xmath460 such that @xmath463 . set @xmath464 and likewise define @xmath465 . continuing in this fashion , we construct the chains @xmath118 and @xmath84 , and @xmath466 for @xmath467 . we show that @xmath118 increases everywhere , except possibly at @xmath468 , and hence @xmath469 . similar arguments show that @xmath470 , which will complete the proof . suppose @xmath471 . let @xmath315 be minimal subject to @xmath458 and let @xmath408 , @xmath409 , and @xmath472 . then by proposition [ prop : pieri ] ( 3 ) and ( 4 ) , @xmath412 . the condition @xmath473 in the definition of @xmath412 ensures that @xmath118 increases at @xmath474 . numerical homotopy continuation is a method for finding the isolated solutions of a system @xmath475 where @xmath476 are polynomials in the variables @xmath477 . first , a _ homotopy _ @xmath478 is found with the following properties : 1 . 2 . the isolated solutions of @xmath480 are known . the system @xmath481 defines finitely many ( rational ) curves @xmath482 , and each isolated solution of ( [ system ] ) is connected to an isolated solution @xmath483 of @xmath480 by one of these curves . given such a homotopy , numerical path continuation is used to trace these curves from solutions of @xmath480 to solutions of the original system ( [ system ] ) . when there are fewer solutions to @xmath484 than to @xmath480 , some curves will diverge or become singular as @xmath485 , and it is expensive to trace such a curve . when @xmath486 , the system ( [ system ] ) is _ square _ and the homotopy @xmath487 where @xmath488 with @xmath489 and @xmath490 , gives @xmath491 curves . this is the bzout bound for a generic dense system @xmath307 . in practice , @xmath484 may have fewer than @xmath491 solutions and we desire a homotopy with no divergent curves . methods for such deficient systems which reduce the number of divergent curves are developed in @xcite . when the polynomials @xmath492 have special forms @xcite , then such homotopies ( [ convex ] ) are constructed where @xmath493 shares this special form . when the polynomials @xmath492 are sparse , polyhedral methods @xcite give a homotopy . the sagbi homotopy algorithm ( 2.3 ) is in the same spirit . we exploit a special feature of the coordinate ring of the grassmannian to obtain a homotopy between the system ( 2.2 ) we wish to solve and one ( 2.12 ) whose solution may be obtained using polyhedral methods . moreover , there are generically no divergent curves to be followed . the overdetermined situation of @xmath494 is more delicate . one difficulty is finding a homotopy @xmath478 for an overdetermined system as generic perturbations of @xmath307 have no solutions . in @xcite , this difficulty is avoided as follows : the system @xmath476 is replaced by @xmath189 random linear combinations of the @xmath492 yielding a square system whose isolated solutions include all isolated solutions of @xmath484 , but typically many more . they then find all isolated solutions of this random square subsystem . for the grbner and pieri homotopy algorithms , we gave ( in 2.2 and 3.23 ) homotopies @xmath495 and solutions @xmath483 at @xmath71 as above . for these , there are generically no divergent curves . to efficiently follow the curves @xmath482 , we select a square subsystem @xmath496 of @xmath478 ( @xmath497 is an @xmath498-matrix ) . if the jacobian of @xmath499 at each @xmath483 has the same rank ( @xmath189 ) as does the jacobian of @xmath478 , then the curves @xmath482 remain components of the algebraic set defined by the equations @xmath500 moreover , other components of this set meet the curves @xmath482 in at most finitely many points @xmath65 in @xmath501 . thus , we may use the square subsystem @xmath496 to trace the curves @xmath482 along some path from 0 to 1 in the complex plane . we remark that in practice , @xmath497 may be chosen at random . the algorithms of sections 2 and 3 are useful for studying both the pole assignment problem in systems theory @xcite and real enumerative geometry @xcite . we describe the connection to the control of linear systems following @xcite . suppose we have a system ( for example , a mechanical linkage ) with inputs @xmath502 and outputs @xmath503 for which there are internal states @xmath504 such that the evolution of the system is governed by the first order linear differential equation @xmath505 if the input is controlled by constant output feedback , @xmath506 , then we obtain @xmath507 the natural frequencies of the controlled system are the roots @xmath508 of @xmath509 the pole assignment problem asks , given a system ( [ linsystem ] ) and a polynomial @xmath510 of degree @xmath367 , which feedback laws @xmath307 satisfy ( [ charpoly ] ) ? a standard transformation ( _ cf . _ @xcite ) transforms the input data @xmath511 into matrices @xmath512 , @xmath513 of polynomials with @xmath514 and @xmath515 such that @xmath516.\ ] ] here @xmath17 is the @xmath517-identity matrix and the feedback law @xmath307 is an @xmath518-matrix . if we let @xmath519 \qquad { \rm and}\qquad k(s)\ : = \ \left[\begin{array}{c } d(s)\\n(s)\end{array}\right],\ ] ] then @xmath307 gives local coordinates on @xmath37 and ( [ schubert_form ] ) is equivalent to @xmath520 these conditions are independent for generic @xmath511 and distinct @xmath521 , hence @xmath522 is necessary for there to be any feedback laws @xmath307 . the critical case of @xmath523 is an instance of the situation in 2 . in @xcite homotopy continuation was used to solve a specific feedback problem when @xmath56 . from this result , they deduced that the pole assignment problem is not in general solvable by radicals . despite this success , only few non - trivial examples have been computed in the control theory literature @xcite . an important question is whether a given system may be controlled by _ real _ output feedback @xcite . that is , if all roots of @xmath510 are real , are there real feedback laws @xmath307 satisfying ( [ schubert_form ] ) ? real enumerative geometry @xcite asks a similar question : are there real linear subspaces @xmath0 in general position with @xmath250 and @xmath6 such that _ all _ @xmath4-planes meeting each @xmath5 nontrivially are real ? when either @xmath46 or @xmath4 is 2 @xcite , @xmath524 @xcite , or when @xmath525 and the @xmath526 @xcite , the answer is yes . in fact , the pieri homotopies arose from these investigations . b. shapiro and m. shapiro give a precise conjecture relating both applications . suppose @xmath527,\ ] ] where @xmath528 is a parameterization of a rational normal ( non - degenerate ) curve in @xmath529 of degree @xmath530 . one such choice is @xmath531.\ ] ] geometrically , @xmath532 is the @xmath533-plane which osculates the curve @xmath528 at @xmath534 . such osculating @xmath46-planes have been used to prove non - degeneracy results in control theory . [ conj : ss ] let @xmath508 be distinct real numbers and suppose @xmath535 osculates @xmath413 at @xmath521 and @xmath6 . then each of the finitely many @xmath4-planes @xmath536 satisfying @xmath537 for @xmath538 is defined over the reals . our algorithms have been tested successfully in matlab , finding all 14 solutions in the case @xmath539 for both the sabgi and grbner homotopy algorithm , and all 15 solutions when @xmath540 and @xmath541 for the pieri homotopy algorithm . at present , the sagbi and grbner homotopy algorithms have been fully implemented . some timings from trial runs of these algorithms on a sparc 20 are displayed in table [ sagbi_times ] . the input for these were @xmath23 random complex @xmath49-matrices . we provide a comparison to methods based upon grbner bases . table [ sagbi_times ] also gives the time on the sparc 20 for the system singular @xcite to compute a degree reverse lexicographic grbner basis for the polynomial systems : @xmath542\ = \ 0 , \quad\mbox{for}\quad i=3,\ldots , k_{mp}.\ ] ] here @xmath41 is expressed in local coordinates for @xmath543 } \,\cap \,\omega'_{[13 ] } \,$ ] and the @xmath544 are @xmath49-matrices with random integral entries between @xmath545 and @xmath546 . a degree reverse lexicographic grbner basis is the input for some alternative numerical polynomial systems solvers ( e.g. eigenvalue methods @xcite ) . we note that the grbner basis calculation did not terminate within one week in the case @xmath540 . , _ an elimination algorithm for the computation of the zeros of a system of multivariate polynomial equations _ , in numerical mathematics , singapore , 1988 , r. agarwal , y. chow , and s. wilson , eds . 86 of internat . birkhuser , 1988 , pp . 1130 . , _ control theory , inverse spectral problems , and real algebraic geometry _ , in diffferential geometric methods in control theory , r. w. brockett , r. s. millman , and h. j. sussmann , eds . , birkhuser , boston , 1982 , pp . 192208 . , _ pole assignment by output feedback _ , in three decades of mathematical systems theory , h. nijmeijer and j. m. schumacher , eds . 135 of lecture notes in control and inform . , springer - verlag , berlin , 1989 , pp . , _ global properties of the root - locus map _ , in feedback control of linear and non - linear systems , d. hinrichsen and a. isidori , eds . , vol . 39 of lecture notes in control and inform . , springer - verlag , berlin , 1982 .

background : : : the scissors mode is a rotational mode of isovector character in deformed nuclei . together with the isoscalar rotation , it is the prominent collective mode at low excitation energies . purpose : : : we use a simple field theory model to investigate the role of the nucleon spin for the magnetic sum rules associated with the low - lying scissors mode . special emphasis is put on the coupling of the spin part of the m1 operator to the scissors mode . methods : : : we apply the mean field approximation and random phase approximation ( rpa ) to a model hamiltonian based on a simple quadrupole - quadrupole interaction . the effects of the spin - orbit interaction are included in the mean field hamiltonian . ward - takahashi relations are used to derive the m1 sum rules . results : : : the presence of the spin - orbit interaction leads to interference terms in the inverse energy weighted sum rule . it is shown that the low - lying scissors mode , which is generated by the isovector combination of proton and neutron total angular momenta , gives the main contribution to the spectral sum for this case . conclusions : : : the basic concept of the scissors mode as an isovector vibrational rotation remains valid in the presence of the nucleon spin . the inverse energy weighted sum rule , however , receives non - trivial modifications because of interference terms .

1312.6925

the scissors mode , which is a rotational mode of isovector character in deformed nuclei@xcite , is usually visualized by an out of phase orbital motion of protons against neutrons . this intuitive picture lead to the prediction of the scissors mode both in terms of nucleonic@xcite and bosonic@xcite degrees of freedom , and has formed the basis of many investigations @xcite . in some respect , the scissors mode can be thought of as an extension of the familiar isoscalar collective rotation@xcite to the isovector rotational motion . indeed , in a recent work @xcite we have used a simple field theory model for the collective orbital motion to show how an effective hamiltonian of the 2-rotor type emerges naturally from the cooperation of the isoscalar ( goldstone ) and isovector ( scissors ) degrees of freedom . in usual nuclear hamiltonians , the orbital angular momentum alone is not conserved . a familiar example is the spin - orbit potential , which explicitly breaks the separate spin and orbital symmetries on the level of the mean field hamiltonian . the explicit breaking of the spin symmetry on the mean field level has important consequences for the properties of nuclear rotational states , and we mention two of them here in more detail : first , if we use the mechanism of spontaneous symmetry breaking to generate a deformed mean field , it is the _ total _ angular momentum symmetry which will be restored by the collective rotation of the whole system . that is , the isoscalar goldstone modes are generated essentially by the total angular momentum @xmath0 , and their degrees of freedom will lead to an effective rotational hamiltonian of the schematic form @xmath1 , where @xmath2 is the total moment of inertia . in analogy to the situation found for the pure orbital case@xcite , we can then anticipate the existence of a mode at low excitation energy , which is generated by the isovector combination of @xmath3 and @xmath4 , and which cooperates with the goldstone mode so as to give a total effective rotational hamiltonian of the schematic 2-rotor form @xmath5 . the structure of the goldstone and scissors modes can therefore be expected to come out similar to the case of the pure orbital motion , but with the orbital angular momenta replaced by the total angular momenta . second , the explicit breaking of the spin symmetry on the mean field level leads to new types of correlation functions , because the spin operators ( @xmath6 and @xmath7 ) can generate particle - hole states of finite excitation energy . these new types of non - interacting correlators , which are visualized by the bubble graphs shown by fig . 2 of the next section with one of the external operators being a spin operator , will obviously modify the sum rules , because the latter ones can be represented in terms of correlation functions@xcite . in particular , we will see that the presence of these correlators implies that the spin part of the m1 operator can couple to the low energy scissors modes . this observation is of particular interest , because it is known that there exists a discrepancy between the theoretical and experimental values of the inverse energy weighted m1 sum rule , for which the low energy scissors mode is expected to give the dominant contribution@xcite . the purpose of the present paper is to consider a `` minimal model '' to get analytic insight into the role of the spin for the low energy rotational states in deformed nuclei . by `` minimal '' we mean that the spin dependence of the nuclear hamiltonian is taken into account only via the one - body spin - orbit potential , and not via the two - body interactions . besides simplicity , the reason to restrict ourselves to this `` minimal model '' is to keep the discussions distinct from the collective spin modes at higher energies , which are known to be strongly modified by interactions of the spin - spin type@xcite . in order to concentrate on the important role of the spin - orbit potential , we also will not include the effects of pairing correlations in this paper . the role of spin dependent interactions and pairing correlations for the m1 sum rules will be discussed in a separate work@xcite . in sect . ii we will briefly explain the model and the approximations ( mean field approximation and random phase approximation ( rpa ) ) . in sect . iii we will explain the constraints which follow from angular momentum conservation , and in sect . iv we will discuss the m1 sum rules , with a special emphasis on the role of the spin in the inverse energy weighted ( iew ) sum rule . v gives a summary of our results . more formal discussions and derivations are presented in four appendices . our model is based on the simple quadrupole - quadrupole ( @xmath8 ) interaction , which we used in our earlier work@xcite to describe the pure orbital collective motion . because the mean field approximation and rpa for this model are discussed in detail in that paper , we only summarize the main points here . the one - body part of the hamiltonian contains a spherical mean field @xmath9 , which we will not need to specify , and a spin - orbit interaction of the simplest possible form from ( see eq.([h0 ] ) below ) . the two - body interaction , which we need to describe the collective rotational states , is of the form @xmath10 where @xmath11 denotes protons ( @xmath12 ) and neutrons ( @xmath13 ) , and for the coupling constants we assume @xmath14 and @xmath15 . the quadrupole operator is defined by @xmath16 where @xmath17 . because the spin - orbit potential breaks the spin symmetry explicitly from the outset , the hamiltonian is only invariant under rotations generated by the total angular momentum @xmath18 . the assumption that the @xmath19 component of the quadrupole operator has a finite ground - state expectation value @xmath20 leads to spontaneous breaking of the rotational symmetry down to rotations generated by @xmath21 . the first term in ( [ mfa ] ) expresses a deformed mean field for protons and neutrons , and the dots in the second term denote normal ordering . by requiring self consistency for the strength of the deformed mean fields , one finally arrives at the following form of the hamiltonian @xcite : @xmath22 where the one - body parts are given by @xmath23 and @xmath24 is a constant given by @xmath25 the parameters @xmath26 in ( [ h0 ] ) represent the strength of the spin - orbit potential . the deformation parameters @xmath27 obey the self consistency relations @xmath28 where the parameters @xmath29 , which describe the explicit breaking of the rotational symmetry via a term @xmath30 in the hamiltonian , are introduced only for technical reasons and will be set to zero in all final results . the excitation energies and wave functions of the collective states of the system are determined by the poles and the residues of particle - hole t - matrix , which is a solution of the bethe - salpeter ( bs ) equation , or equivalently the rpa equation . because we are assuming axial symmetry , we can characterize the states by the quantum number @xmath31 , which is the projection of the total angular momentum on the symmetry ( @xmath32 ) axis . since we are interested in the isoscalar goldstone and the isovector scissors rotational states , we consider the bs equation in the @xmath33 channel , which is graphically represented by fig . 1 . ) and the positions @xmath34 , @xmath35 are fixed , and an integral is taken over the other variables . ] its solution is of the following form : @xmath36 where the reduced t - matrix satisfies the equation @xmath37 here the quantities @xmath38 , the coupling constants @xmath39 and the non - interacting polarizations ( bubble graphs ) with two external quadrupole operators @xmath40 are considered as @xmath41 matrices in charge space . having in mind later applications , we define the proton and neutron bubble graphs for arbitrary external tensor operators ( @xmath33 components @xmath42 and @xmath43 ) as follows ( see fig.2 ) : ) . ] @xmath44 \label{piprop } \\ & = - 2 \sum_{(\alpha i ) \in \tau } \langle \alpha | w'^1 | i \rangle^ * \langle \alpha | w^1 | i \rangle \frac{\omega}{\omega^2 - \omega_{\alpha i}^2 + i \delta } \ , . \label{pi}\end{aligned}\ ] ] here @xmath45 denotes the feynman propagator in the mixed representation , @xmath46 are the non - interacting particle - hole energies for protons ( case @xmath47 ) or neutrons ( case @xmath48 ) , and the symbol @xmath49 means @xmath50 if the operators @xmath51 and @xmath52 have the same time reversal ( @xmath53 ) symmetry ( case @xmath54 ) , and @xmath55 if they have the opposite @xmath53-symmetry ( @xmath56).-symmetry of an operator @xmath57 is defined as in eq.(a.22 ) of @xcite by @xmath58 , where the time - reversed single particle states @xmath59 have the opposite values of @xmath60 but the same energies as @xmath61 for the case of axial symmetry . the sign is @xmath62 for a @xmath53-even operator ( like the quadrupole operator ) , and @xmath63 for a @xmath53-odd operator ( like the angular momentum operators ) . ] from ( [ t1 ] ) , the poles of the t - matrix ( @xmath64 ) are determined by the equation @xmath65 and the reduced vertex functions for the collective states ( @xmath66 ) ) are defined by the residues of the reduced t - matrix : @xmath67 from this definition and eq.([t1 ] ) one obtains the homogeneous bs equation and the normalization condition for the reduced vertex functions : @xmath68 the prime in ( [ norm ] ) indicates differentiation w.r.t . @xmath69 , i.e. , @xmath70 the full vertex functions @xmath71 for the @xmath33 collective states with excitation energy @xmath72 are defined similarly to ( [ pole ] ) by the residues of the full t - matrix ( [ t ] ) , and are given in the present model by @xmath73 the relation of this bs formalism to the usual rpa is explained in appendix a , where it is shown that ( [ norm ] ) is equivalent to the familiar normalization condition for the forward and backward rpa amplitudes . by definition , a tensor operator @xmath74 of rank @xmath75 with spherical components @xmath76 satisfies the following commutations relations with the total angular momentum operators@xcite , @xmath77 for all vector operators @xmath78 , including the case of the angular momentum operators ( @xmath79 ) . therefore the sign of the @xmath80 component of the angular momentum operators is opposite to ref.@xcite or @xcite . ] @xmath81 , @xmath82 : @xmath83 & = \mp \frac{1}{\sqrt{2 } } \sqrt{k(k+1 ) - q ( q \pm 1 ) } \ , w^{q \pm 1 } \label{comm1 } \\\ \left [ j^0 , w^q \right ] & = q \ , w^q \label{comm2 } \,.\end{aligned}\ ] ] the commutator of @xmath84 ( eq.([h0 ] ) ) with total angular momentum operators then becomes @xmath85 = \mp \sqrt{3 } \beta_{\tau } q_{\tau}^{\pm 1}\ , , \,\,\,\,\,\,\left [ h_{0 \tau } , j_{\tau}^{0 } \right ] = 0 \ , , \label{comm0}\end{aligned}\ ] ] and if we consider matrix elements of these identities between non - interacting particle - hole states , we obtain the useful relations @xmath86 here @xmath87 , and the above relations hold for non - degenerate particle - hole states ( @xmath88 ) . let us note here two applications of the relations ( [ rel ] ) : first , consider the bubble graph ( [ pi ] ) for the case where @xmath89 is the quadrupole operator which is @xmath53-even , and another operator which is @xmath53-odd , for example one of the angular momentum operators @xmath90 or @xmath91 . ( [ rel ] ) for @xmath33 then leads to the relation @xmath92 second , the derivative of the @xmath8 bubble graph ( [ pip ] ) at @xmath93 can be expressed in terms of the @xmath94 bubble graph as follows : @xmath95 the quantities @xmath96 are actually the moments of inertia of protons or neutrons , according to the familiar inglis formula@xcite , expressed in terms of the total angular momentum operators : @xmath97 here @xmath98 is the total moment of inertia . next let us discuss the ward - takahashi identities@xcite for the full @xmath33 correlation functions , which follow from angular momentum conservation . we consider the time derivative of the 2-point function @xmath99 with one arbitrary tensor operator @xmath100 of rank @xmath75 and the total angular momentum operator @xmath101 . using the heisenberg equation of motion @xmath102 = - i \sqrt{3 } \sum_{\tau } \varepsilon_{\tau } q_{\tau}^{1 } , \label{heis}\end{aligned}\ ] ] and the equal time commutator @xmath103 $ ] from ( [ comm1 ] ) , and performing a fourier transformation @xmath104 , we obtain the ward - takahashi identity for the exact correlators and the correlators in the chain approximation ( rpa ) , and @xmath105 for the non - interacting ones . although not indicated explicitly , all correlators refer to the case @xmath33 . to get the last term in eq.([wt1 ] ) we assumed the relation @xmath106 , which is satisfied for all operators of the main text . ( only for the operator @xmath78 defined by ( [ v ] ) the sign is opposite . ) ] @xmath107 here we defined the _ exact _ 2-point functions for arbitrary tensor operators @xmath52 and @xmath51 by @xmath108 to visualize these correlators in our chain approximation ( rpa ) , we represent them graphically in fig.3 , which translates into the following expressions in terms of the reduced particle - hole t - matrix of eq.([t1 ] ) : @xmath109 for later reference , we also note that the spectral representation of the exact correlators is given in analogy to eq.([pi ] ) for the non - interacting ones by @xmath110 where @xmath111 are the exact excitation energies of the @xmath33 eigenstates @xmath112 of the hamiltonian @xmath113 , and @xmath49 is defined as in eq.([pi ] ) by @xmath114 ( if @xmath115 ) and @xmath116 ( if @xmath117 ) . returning to the ward - takahashi relation ( [ wt1 ] ) , there are two important limits , namely the `` goldstone limit '' ( @xmath118 for finite @xmath119 ) , and the `` exact symmetry limit '' ( @xmath120 for finite @xmath121 ) . in the goldstone limit we have @xmath122 while in the exact symmetry limit we have @xmath123 each of these limits has important implications . consider first the goldstone limit ( [ w1 ] ) for the case where @xmath51 is @xmath53-odd . we know from the spectral representation eq.([spec1 ] ) that the l.h.s . of ( [ w1 ] ) then vanishes , which implies that the ground state expectation value of the @xmath19 component of any @xmath53-odd tensor operator is zero : @xmath124 if @xmath125 . using this result in the exact symmetry limit ( [ w2 ] ) , we obtain the constraint @xmath126 later , in connection with the m1 sum rules , we will verify eq.([ang ] ) explicitly for the case where @xmath51 is an angular momentum operator and @xmath93 . the relation ( [ ang ] ) holds if the tensor operator @xmath51 is @xmath53-odd . if @xmath51 is @xmath53-even and its @xmath19 component has a non - vanishing ground state expectation value , the rotational symmetry is spontaneously broken , and @xmath127 is nonzero . the spectral representation ( [ spec1 ] ) shows that only the goldstone mode ( @xmath128 ) contributes here , because the total angular momentum @xmath129 can not give rise to finite energy excitations because of @xmath130=0 $ ] . then it follows from the spectral representation ( [ spec1 ] ) that for this case @xmath131 , where @xmath31 is a constant , in agreement with ( [ w2 ] ) . the goldstone limit ( [ w1 ] ) for the case @xmath132 can be used to confirm the existence of 2 independent goldstone modes , corresponding to the 2 broken generators @xmath133 and @xmath134 , and to determine the vertex functions of those modes . the arguments are very similar to those given in ref.@xcite , and we briefly summarize the main points here . first , to show the existence of the goldstone modes , we note that the dyson equation ( [ rpag1 ] ) for the case @xmath135 takes the matrix form @xmath136 . multiplying then ( [ w1 ] ) for @xmath132 from left by @xmath137 we obtain @xmath138 where we used the self consistency relation ( [ gap ] ) to eliminate @xmath139 . ( in this notation , @xmath140 and @xmath141 are considered as vectors in charge space . ) we then obtain the identity @xmath142 by which the self consistency relation ( [ gap ] ) can be rewritten as @xmath143 in the limit of exact rotational symmetry ( @xmath144 ) , this equation leads to the condition @xmath145 for a nontrivial solution . comparing this with the pole equation ( [ det ] ) , we see that in the limit of exact rotational symmetry the self consistency relation guarantees the existence of a goldstone pole ( @xmath146 ) in the @xmath33 channel . a similar argument holds for the case @xmath147 , which completes the proof for the existence of 2 goldstone modes . the reduced t - matrix ( [ pole ] ) therefore behaves in the exact symmetry limit ( @xmath148 ) as follows : @xmath149 second , to determine the vertex functions of the goldstone modes , we note that a comparison of the homogeneous bs equation ( [ hombs ] ) with ( [ gap1 ] ) in the limit of exact symmetry ( @xmath148 ) allows to specify the charge dependence of the goldstone vertex functions as @xmath150 where the charge independent constant @xmath151 is determined from the normalization condition ( [ norm ] ) as @xmath152 where we used the relations ( [ pip1 ] ) - ( [ ingtot ] ) . the form of the reduced and full @xmath33 goldstone vertex functions ( see eq.([fullv ] ) ) can then be summarized as @xmath153 the vertex function for the @xmath147 goldstone mode is obtained from ( [ fgv ] ) by replacing @xmath154 . next we turn to the exact symmetry limit ( [ w2 ] ) for the case @xmath132 . since for the case of spontaneously broken symmetry the r.h.s . of this relation is a non - zero constant , we obtain in the limit @xmath155 , where @xmath156 is one of the nonzero solutions of the eigenvalue equation , @xmath157 inserting here the rpa form ( [ rpag2 ] ) and using the pole behavior of the reduced t - matrix ( [ pole ] ) , we obtain @xmath158 the physical meaning of this relation is that the total angular momentum operator @xmath159 can not cause transitions from the ground state to a state with finite excitation energy . this follows easily from ( [ fo1 ] ) by noting that the transition matrix element of the @xmath33 component of any tensor operator @xmath160 from the ground state to an excited state is given by ( see fig.4 ) @xmath161 for the case @xmath162 we obtain the physical interpretation of the relation ( [ fo1 ] ) as explained above . ) . ] in appendix a , the creation and annihilation operators for the goldstone modes and the resulting isoscalar part of the effective rotational hamiltonian are presented .. ] in particular , it is shown that the total angular momentum operator determines the form of the creation and annihilation operators for the zero modes and the effective rotational hamiltonian . in appendix b , a simple analytic approximation , which takes into account only the isoscalar zero modes and the isovector low energy modes , is used to present a similar discussion for the isovector scissors mode . in particular , it is shown that the isovector combination of @xmath163 and @xmath164 determines the form of the rotational part of the creation and annihilation operators and the effective rotational hamiltonian . by adding the contributions from the goldstone and scissors degrees of freedom , we derive an effective hamiltonian of the 2-rotor form . in this section we wish to discuss the inverse energy weighted ( iew ) and energy weighted ( ew ) magnetic sum rules as they emerge in the simple framework discussed in the previous sections . our main interest is to see how the presence of the spin part of the m1 operator changes the sum rules , and to gain some intuitive understanding of the results . we express the @xmath33 component of the m1 operator in the form @xmath165 where the values of the free proton and neutron @xmath166-factors are @xmath167 , @xmath168 , @xmath169 , @xmath170 . in the second equality of ( [ m1 ] ) we used @xmath171 , and defined @xmath172 the free nucleon values of @xmath173 are therefore @xmath174 and @xmath175 . we remark that in principle the @xmath166-factors appearing in ( [ m1 ] ) should be regarded as effective ones@xcite , which include short range processes ( tensor correlations , meson exchange currents , etc ) , which are not taken into account by the rpa - type correlations . to derive the sum rules , we follow ref.@xcite and consider the following 2-point function with external m1 operators : @xmath176 where @xmath177 , and we used the notations of eq.([spec1 ] ) . the iew and ew sum rules can then be expressed as follows@xcite : @xmath178 the quantity @xmath179 of ( [ specm ] ) is related to the correlators @xmath180 of ( [ corr ] ) by @xmath181 and therefore the calculation of m1 sum rules reduces to the calculation of the correlators @xmath182 in the limits indicated by eq.([iew ] ) and ( [ ew ] ) , where @xmath183 or @xmath184 denotes @xmath91 or @xmath185 . in our simple rpa framework , those correlators are given by ( [ rpag2 ] ) , i.e. , @xmath186 for the iew sum rule , we need the correlator ( [ rr ] ) for @xmath93 . because of the relation ( [ rel1 ] ) , the bubble graphs in the second term on the r.h.s . of ( [ rr ] ) bring in a factor @xmath69 , and therefore only the singular term in the reduced t - matrix of eq.([sing ] ) contributes . using the relation ( [ rgv ] ) , we obtain the very simple result @xmath187 by taking @xmath184 or @xmath183 equal to @xmath91 in this expression , and using the definition of the moments of inertia as given by eq.([ingpn ] ) and ( [ ingtot ] ) , it is then evident that @xmath188 this is a special case of the condition ( [ ang ] ) for @xmath93 , which follows from angular momentum conservation . the relation ( [ conf ] ) implies that only the isovector combination of the @xmath189 and @xmath164 contributes . it is therefore useful to split the orbital @xmath166-factors into isoscalar and isovector parts according to @xmath190 where @xmath191 and @xmath192 . the factors @xmath193 and @xmath194 are the isoscalar and isovector orbital @xmath166-factors , with the free nucleon values given by @xmath195 . because of ( [ conf ] ) , we can effectively replace @xmath196 in the correlator ( [ pimm ] ) for @xmath93 . on the other hand , the terms involving the spin operators give both isoscalar and isovector contributions , and the result does not simplify by splitting @xmath173 into isoscalar and isovector pieces like in ( [ iso ] ) . we therefore express the final result by using the factors @xmath197 in the following way ( see appendix d ) : @xmath198 the term @xmath199 in ( [ scsg ] ) is the familiar scissors iew sum rule @xcite . in order to express the contributions from the spin part of the m1 operator , we defined the spin and mixed moments of inertia ( @xmath200 and @xmath201 ) as follows : @xmath202 and @xmath203 in ( [ scsg ] ) means the real part is actually real . formally this is seen by noting that the expansion coefficients of deformed states in terms of spherical states ( see eq.([expand ] ) ) are real , and matrix elements of the spherical components of angular momentum operators between the spherical states are also real . ] of @xmath204 . in appendix b we show that the contribution ( [ scsg ] ) , which arises from the terms @xmath94 , @xmath205 and @xmath206 in the correlator ( [ pimm ] ) , can be reproduced in a simple analytic approximation which takes into account only the low energy scissors mode , in addition to the goldstone mode . this makes it plausible that the contribution ( [ scsg ] ) arises mainly from the low energy scissors mode in the spectral sum ( [ iew ] ) . the term ( [ ss ] ) , on the other hand , can not be reproduced in this way . in fact , because this term arises from the @xmath207 term in the m1 correlator ( [ pimm ] ) , it will be modified by spin dependent interactions , and can not be reliably discussed within our `` minimal model '' . nevertheless , it is worth while to point out that the @xmath8-type interaction , which leads to the last term of ( [ ss ] ) in the rpa , gives a negative contribution to the spin - spin part of the iew sum rule , that is , the spin part of the sum rule is smaller than the naive ( non - interacting ) part , which is given by the first two terms in ( [ ss ] ) ) is of course positive , because it corresponds to the sum rule ( [ iew ] ) with @xmath208 replaced by @xmath209 . ] . if the spin - orbit interaction is assumed to be zero , all spin correction terms in ( [ scsg ] ) , and also all terms in ( [ ss ] ) , vanish . this is easily seen because in this case the relation @xmath210=0 $ ] is satisfied , which implies @xmath211 for @xmath212 . that is , if @xmath213 commutes with the mean field hamiltonian , it can not cause particle - hole excitations with finite excitation energy . the spectral representation ( [ pi ] ) then indicates that @xmath214 , where @xmath215 or @xmath185 . ) do not contribute from the outset , because the correlator ( [ pi ] ) involves an explicit factor @xmath216 for external operators with the same @xmath53-symmetry . if one needs the limit @xmath217 , this factor should be taken to zero first . ( for the same reason , the goldstone modes do not contribute to the correlator ( [ pimm ] ) . ) we also note that the argument based on the commutation relation given above can be applied also to the interacting case , where @xmath218 is the conserved quantity . the resulting identity was already expressed by eq.([ang ] ) . ] therefore , if the spin - orbit interaction is set to zero , the iew sum rule reduces to the pure orbital result@xcite , as expected . we also remark that , for the case of spherical symmetry , all terms except the pure spin terms ( the first 2 terms in ( [ ss ] ) ) vanish , because in this case the relation @xmath219=0 $ ] is satisfied , and a similar argument as given above implies that all bubble graphs which involve at least one operator @xmath220 vanish . in particular , the mixed moments of inerita , which are responsible for the spin correction terms in in ( [ scsg ] ) , separately vanish in the limit of spherical symmetry . from the above discussions it is clear that the correction terms in ( [ scsg ] ) , which mainly arise from the coupling of the external spin operator to the scissors mode , are non - zero only in the case of finite spin - orbit interaction and finite deformation . this is also clear from the basic formulae for the matrix elements of the orbital and spin angular momentum operators between deformed single particle states , which are collected in appendix c. the sign of the mixed polarization @xmath221 is , however , a delicate matter , because it depends on the quantum numbers of the valence nucleon states , the degree of deformation , and also further model assumptions on the mean field hamiltonian . if the sign of @xmath222 turns out to be negative , the spin correction terms in ( [ scsg ] ) will decrease the value of @xmath223 , which would help to explain the discrepancy between the theoretical and experimental values . in order to see the sign and the size of the correction terms , a numerical analysis in a nilsson - type approach@xcite is necessary , which , however , goes beyond the purpose of the analytical approach pursued in this paper . the evaluation of the ew sum rule ( [ ew ] ) is again based on the expressions ( [ pimm ] ) and ( [ rr ] ) for the correlators . because now we have to consider the limit @xmath224 , the @xmath53-matrix @xmath225 in the rpa correlators ( [ rr ] ) reduces to the 4-fermi coupling constant @xmath226 , as is clear from the expression ( [ t1 ] ) . by using the expression ( [ pi ] ) for the bubble graphs , we obtain @xmath227 where @xmath183 ( or @xmath184 ) denotes either @xmath91 or @xmath185 . for the case @xmath228 , one can use the relation ( [ rel ] ) to replace the operators @xmath229 by the quadrupole operators @xmath230 , and make use of the self consistency relation ( [ gap ] ) to obtain the result derived in ref.@xcite . a similar calculation is possible also for the case where @xmath183 ( or @xmath184 ) is the spin operator @xmath185 . the derivation is explained in appendix d , which in particular shows that there are no crossing terms between @xmath91 and @xmath185 in the ew sum rule , i.e. , @xmath231 the final result becomes @xmath232 here @xmath26 is the strength of the spin - orbit interaction , see eq.([h0 ] ) . the second term in ( [ ewfin ] ) , which comes from the spin part of the m1 operator , is positive because @xmath233 . ( a formal argument for this is given in appendix d. ) however , as we noted already in connection with the iew sum rule , spin dependent interactions will give additional contributions to the @xmath234 correlator , which will modify the result ( [ ewfin ] ) . unlike the case of the iew sum rule , there is no spin effect in the ew sum rule which reflects the excitation of the low energy scissors mode . in this paper we used a simple field theory model for nucleons to get analytic insight into the role of the spin for low energy nuclear rotational states . our starting point was a nuclear hamiltonian which explicitly breaks the separate spin and orbital symmetries due to the presence of the spin - orbit potential from the outset , and we used the mechanism of spontaneous symmetry breaking to generate a deformed mean field . the residual interaction , for which we used a simple @xmath8 interaction , was taken into account in the framework of the rpa . our main focus was on the inverse energy weighted ( iew ) magnetic sum rule , for which we found that the spin part of the m1 operator gives nonzero contributions , even in the absence of spin dependent residual interactions . we argued that the spin - orbit potential gives rise to new non - interacting polarizations @xmath235 , where one of the operators @xmath184 , @xmath183 is a spin operator . in particular , the polarizations @xmath236 modify that part of the iew sum rule which arises mainly from the excitation of the low energy scissors mode . in order to assess the effect of those spin corrections quantitatively , however , a numerical approach is necessary . by using a simple analytic approximation , we also confirmed that the structures of the isoscalar goldstone and isovector scissors modes , and of the resulting effective rotational hamiltonian , are similar to the case of the pure orbital description , if one replaces the orbital angular momenta by the total angular momenta . finally we wish to address the question whether it is possible to construct models where the pure orbital nature of the low energy scissors mode is maintained . this is indeed possible if one starts from a hamiltonian which conserves the orbital and spin symmetries separately , and generates the spin - orbit interaction by the mechanism of spontaneous symmetry breaking . the order parameter which describes this breaking is the ground state expectation value @xmath237 , and the three associated goldstone bosons are represented by @xmath238 . if combined with the pattern of dynamical symmetry breaking used in the present paper , the associated 5 isoscalar goldstone degrees of freedom lead to separate orbital and spin rotational bands , and for each there exists an isovector counterpart at low excitation energy . condensate ) has been used for the description of superfluid helium 3 in the b phase@xcite . the spin rotational bands mentioned above might be connected to the so called magnetic rotational bands observed in weakly deformed nuclei@xcite . ] we plan to present such an approach in a separate paper@xcite . acknowledgements one of the authors ( w.b . ) wishes to express his thanks to prof . k. yazaki for very helpful discussions . we first recall that the @xmath53-symmetry relation for the particle - hole matrix elements of any operator @xmath239 is given by ( see footnote 1 of sect . ii ) @xmath240 where the time - reversed single particle states @xmath241 have the opposite values of @xmath242 but the same energies as @xmath243 for the case of axial symmetry , and @xmath244 . for the case of the @xmath53-even quadrupole operator we have @xmath245 . in order to establish the connection of the bs formalism used in the main text to the more conventional rpa formalism@xcite , we return to the expression ( [ trans ] ) for the transition matrix element of a tensor operator @xmath160 and insert the spectral representation of the correlator @xmath246 in the form @xmath247 \ , . \label{corrf}\end{aligned}\ ] ] by using the @xmath53-symmetry relation ( [ ts ] ) for the second term in ( [ corrf ] ) to combine it with the first one , it is easy to see that this relation is the same as ( [ pi ] ) of the main text . using ( [ corrf ] ) , the transition matrix element ( [ trans ] ) then takes the form @xmath248\ , . \label{tr}\end{gathered}\ ] ] we used the notation @xmath249 for the particle - hole matrix elements of an operator @xmath239 , and defined the @xmath33 components of the rpa amplitudes by @xmath250 in the conventional rpa formalism , the states @xmath251 are then expressed by @xmath252 where @xmath253 with @xmath254\ , . \label{cr}\end{aligned}\ ] ] here we introduced the creation and annihilation operators for a particle - hole pair by @xmath255 the rpa consists in @xmath256 \stackrel{{\rm rpa}}{\equiv } \delta_{\alpha \beta } \delta_{i j } \ , , \label{comm}\end{aligned}\ ] ] while the commutators of two creation ( or two annihilation ) operators vanish . if we also expand the external operator @xmath160 in terms of the creation and annihilation operators as @xmath257 \ , \label{expandw}\end{aligned}\ ] ] and require the rpa ( [ comm ] ) , it is easy to check that @xmath258 |0 \rangle \label{tr1}\end{aligned}\ ] ] gives the same result as ( [ tr ] ) . by using the forms ( [ yrpa ] ) and ( [ zrpa ] ) and the @xmath53-symmetry relation ( [ ts ] ) , it is also easy to confirm the familiar normalization relation for the rpa amplitudes : @xmath259 where the derivatives of the @xmath8 bubble graphs were given in ( [ pip ] ) , and in the last step of ( [ conf1 ] ) we used the condition ( [ norm ] ) of the main text . in the rest of this appendix , we concentrate on the case of the @xmath260 goldstone modes ( @xmath261 ) , where the limit @xmath262 needs special care . we first wish to confirm that our method gives creation and annihilation operators for those modes , which are independent of each other and satisfy the correct boson commutation relations . by using the relation ( [ rel ] ) in the expressions ( [ yrpa ] ) and ( [ zrpa ] ) of the rpa amplitudes , and the form ( [ rgv ] ) of the normalization factors , we obtain the following expressions up to the order @xmath263 : @xmath264 the important points to note for the subsequent calculations are : ( i ) the squares of those amplitudes contain , besides the divergent terms @xmath265 , also finite terms of order @xmath266 . ( ii ) in the subsequent expressions , the divergent terms cancel between the forward and backward amplitudes , on account of the @xmath53-symmetry relation ( [ ts ] ) for the @xmath53-odd operators @xmath267 . by noting these two points , it is then easy to confirm that ( [ y0 ] ) and ( [ z0 ] ) and their @xmath147 counterparts , which are obtained by @xmath268 , satisfy the following relations : + ( 1 ) normalization of rpa amplitudes : @xmath269 where in the last step we used the form ( [ ingtot ] ) of the total moment of inertia . the same argument can be applied also to the @xmath147 mode . + ( 2 ) orthogonality of @xmath33 and @xmath147 modes : + the rpa amplitudes for the @xmath147 goldstone mode are obtained by replacing @xmath268 in the expressions ( [ y0 ] ) and ( [ z0 ] ) . indicating the @xmath31-values explicitly , we obtain @xmath270 \\ = \frac{-2}{i^j } \sum_{\tau } \sum_{(\alpha i)\in \tau } \frac{1}{\omega_{\alpha i } } \langle \alpha|j_{\tau}^{-1}|i \rangle^ * \langle \alpha|j_{\tau}^{1}|i \rangle = 0 \ , , \label{c2}\end{gathered}\ ] ] where in the last step we used the fact that for axial symmetry @xmath271 of the single particle states is a good quantum number . this implies that the state @xmath272 can not have the values @xmath273 and @xmath274 at the same time . + ( 3 ) boson commutation relations for the goldstone modes : + the creation and annihilation operators for the @xmath33 goldstone mode are obtained by inserting the forms ( [ y0 ] ) and ( [ z0 ] ) into ( [ cr ] ) and the h.c . of ( [ cr ] ) . this gives @xmath275 \nonumber \\ \label{o0d } \\ \lefteqn{{\cal o}(\omega_0 , k=1 ) = \frac{-1}{\sqrt{2 \omega_0 i^j } } \sum_{\tau } \sum_{(\alpha i ) \in \tau } } \nonumber \\ & & \hspace{-1 cm } \times \left [ \left(1 + \frac{\omega_0}{\omega_{\alpha i } } \right ) \langle \alpha|j_{\tau}^1 \langle i|j_{\tau}^1 |\alpha \rangle^ * b^{\dagger}_{\alpha i } \right]\ , . \nonumber \\ \label{o0}\end{aligned}\ ] ] the operators for the @xmath147 mode are obtained by @xmath268 in the above expressions . by using the @xmath53-symmetry relation ( [ ts ] ) for the operators @xmath229 , it is then easy to confirm that @xmath276 } \nonumber \\ & & \hspace{-1 cm } = \frac{2}{i^j } \sum_{\tau } \sum_{(\alpha i)\in \tau } \frac{1}{\omega_{\alpha i } } \label{commg}\end{aligned}\ ] ] where in the last step we used the form ( [ ingtot ] ) of the total moment of inertia . similar arguments hold also for the case @xmath147 . the conservation of @xmath277 of the single particle states can be used to show that the commutators between the operators for the @xmath33 and @xmath147 modes vanish , similar to the argument given below eq.([c2 ] ) . the commutation relation ( [ commg ] ) can be used to confirm the correct normalization of the goldstone boson states @xmath278 , namely @xmath279 . + finally in this appendix , we note that the goldstone modes give the following contribution to the effective rotational hamiltonian of the system : @xmath280 where we used the forms ( [ o0d ] ) and ( [ o0 ] ) . we see that , because of the overall factor @xmath281 , only the singular terms in ( [ o0d ] ) and ( [ o0 ] ) contribute to the effective rotational hamiltonian , although it would be meaningless to take the limit @xmath262 directly in those expressions . in this appendix we discuss a simple analytic approximation for the isovector scissors mode . one of our intentions is to show that this approximation can reproduce the contribution ( [ scsg ] ) to the iew m1 sum rule , which suggests that also in the `` exact '' rpa approach of the main text this term comes mainly from the scissors mode contribution to the transition matrix element depicted in fig . analytic approximations are obtained by assuming simple pole forms for the bubble graphs ( [ pi])@xcite . here we discuss the simplest approximation of a one - pole form , where the energies of the particle - hole states are replaced by an average excitation energy : @xmath282 . in this approximation , which is similar to the `` closure approximation '' ( effective energy denominators ) used in different contexts@xcite , the pole equation ( [ det ] ) gives only two solutions , which correspond to the isoscalar goldstone mode and the low energy isovector scissors mode . to obtain the excitation energy and the vertex functions for the scissors mode in the schematic model , we need the form of @xmath283 , see ( [ pi ] ) . if we use the identity ( [ rel ] ) and make the one - pole approximation as explained above , it can be expressed in terms of the moments of inertia ( [ ingpn ] ) : @xmath284 inserting this form into the pole equation ( [ det ] ) we obtain @xmath285 on account of the self consistency relation ( [ gap1 ] ) for exact symmetry ( @xmath144 ) , the goldstone pole @xmath286 is a solution of ( [ pole0 ] ) . the other solution is then obtained as @xmath287 where in the second step we used again the self consistency relation ( [ gap1 ] ) for exact symmetry ( @xmath144 ) . by using the above relations , one can obtain the reduced vertex functions from ( [ hombs ] ) and ( [ norm ] ) . the result for the ratio follows from ( [ hombs ] ) as @xmath288 and the individual factors are then obtained from ( [ norm ] ) as @xmath289 the opposite sign for protons and neutrons indicates the isovector character of this mode . the contribution of the scissors mode @xmath290 to the iew sum rule ( [ iew ] ) is given by @xmath291 where we used the second form given in ( [ m1 ] ) for the m1 operator . to evaluate this , we need the form of the transition matrix element for an angular momentum operator @xmath292 or @xmath185 , which is obtained from ( [ trans ] ) as @xmath293 if we use the results ( [ normp ] ) and ( [ normn ] ) for the normalization factors , and the following form of the polarization @xmath294 in the one - pole approximation @xmath295 we obtain for the transition matrix elements @xmath296 the ordinary moments of inertia @xmath297 and the mixed ones @xmath204 have been defined in the main text as ( see eq.([ingpn ] ) and ( [ ingmpn ] ) ) @xmath298 it is then clear from ( [ tp ] ) and ( [ tn ] ) that + @xmath299 , which is a consequence of angular momentum conservation . ( see eq.([fo1 ] ) and the discussions below that equation . ) therefore the isoscalar part of the first term in the m1 operator ( [ m1 ] ) does not contribute to the transition matrix element , and we can replace @xmath196 in ( [ iew0 ] ) , see eq.([iso ] ) . the transition matrix element of the isovector @xmath91 - part is then obtained from ( [ tp ] ) and ( [ tn ] ) as @xmath300 we insert this expression , as well as ( [ tp ] ) , ( [ tn ] ) for the spin operator ( @xmath301 ) , where also the isoscalar contribution is nonzero , into ( [ iew0 ] ) , and obtain finally @xmath302 if we compare this result to the full rpa result of our model , which was given by ( [ scsg ] ) and ( [ ss ] ) , we see that the terms ( [ scsg ] ) and ( [ scsg0 ] ) , which originate from the products @xmath303 , @xmath304 and @xmath305 in the sum rule , are identical , while the terms ( [ ss ] ) and ( [ ss0 ] ) , which come from the product @xmath306 , are different . this observation suggests that the contribution ( [ scsg ] ) of the full rpa sum rule comes mainly from the excitation of the scissors mode , while the spin contribution ( [ ss ] ) receives also contributions from other modes . using the results of sect . ( b.1 ) for the excitation energy and the reduced vertex functions , it is straight forward to determine the rpa amplitudes and the creation operators for the scissors mode from ( [ yrpa ] ) , ( [ zrpa ] ) and ( [ cr ] ) , and also the contribution to the effective hamiltonian of the system . the calculation is the same as in ref.@xcite , with the only difference that we now have the total angular momentum operators instead of the orbital ones . we therefore simply quote the results here . the creation operator for the @xmath33 scissors mode is obtained from an expression similar to eq.([cr ] ) , with the rpa amplitudes ( [ yrpa ] ) , ( [ zrpa ] ) determined from the scissors vertex functions ( [ normp ] ) and ( [ normn ] ) . it has the form @xmath307 , where @xmath308 here the isovector moment of inertia is defined as @xmath309 the form ( [ oj ] ) of the creation operator clearly indicates the scissors character of the mode . it is important , however , to note that it involves the total angular momenta @xmath310 of protons and neutrons , instead of the orbital ones . the annihilation operator for the @xmath33 scissors mode , @xmath311 , is obtained by replacing @xmath268 and @xmath312 in ( [ oj ] ) and ( [ oq ] ) . the creation and annihilation operators for the @xmath147 modes can then be expressed in terms of the operators for @xmath33 by @xmath313 , and @xmath314 . the contribution of the scissors mode to the effective hamiltonian of the system is given in analogy to the relation ( [ hr1 ] ) by @xmath315 inserting here the forms of the operators , we see that the effective hamiltonian has a rotational part and a vibrational part , which are given by @xmath316 + \frac{i^j_p}{i^j_n } \left[(j_n^x)^2 + ( j_n^y)^2 \right ] \right . \no \\ & \left . - 2 \left(j_p^x j_n^x + j_p^y j_n^y \right ) \right\ } , \label{hrot1 } \\ h_q(\omega_1 ) & = - \frac{\chi_{pn } \left(i^{j}\right)^2}{4 \beta_p \beta_n i_p^j i_n^j } \left ( \frac{2 i^j_n}{i^j } \beta_n q_p^1 - \frac{2 i^j_p}{i^j } \beta_p q_n^1 \right)^{\dagger } \no \\ & \times \left ( \frac{2 i^j_n}{i^j } \beta_n q_p^1 - \frac{2 i^j_p}{i^j } \beta_p q_n^1 \right ) . \label{hq}\end{aligned}\ ] ] the term ( [ hq ] ) represents the restoring force which acts against the proton - neutron oscillations . if one adds the contribution of the goldstone mode ( [ hr2 ] ) to that of the scissors mode ( [ hrot1 ] ) , one obtains the kinetic part of the 2-rotor model : @xmath317 this is analogous to the result derived in ref.@xcite , but now the 2-rotor hamiltonian is expressed in terms of the total proton and neutron angular momentum operators instead of the orbital ones . for the numerical evaluation of the non - interacting correlators , one may use the nilsson model@xcite for single particle states in a deformed potential with axial symmetry . assigning the `` asymptotic quantum numbers''@xcite @xmath318 to a single particle state @xmath319 , one can expand it in a spherical basis with quantum numbers @xmath320 as follows : @xmath321\ , , \nonumber \\ \label{expand}\end{aligned}\ ] ] where @xmath322 corresponds to spin up or down . the coefficients @xmath323 must be obtained by diagonalizing the mean field hamiltonian in the spherical basis , and an example for the results is given in table 5 - 2b of ref.@xcite . in the phase convention used in this paper and in ref.@xcite , the coefficients @xmath323 are real and obey the orthonormalization relation @xmath324 the matrix elements of @xmath325 and @xmath326 are given in the spherical basis by the well known expressions of elementary quantum mechanics , and are also real . we then obtain the following results : @xmath327 @xmath328 in eqs . ( [ l1 ] ) and ( [ s1 ] ) , @xmath329 is defined as @xmath330 , and we made use of the angular momentum conservation @xmath331 . the above expressions can be used to calculate the @xmath93 polarizations needed for the iew sum rule , see ( [ ingmpn ] ) and ( [ splitb ] ) . the particle states ( @xmath332 ) and hole states ( @xmath333 ) , which contribute to the correlators , depend on the deformation , the number of protons and neutrons , and model assumptions for the single particle hamiltonian , like the strength of the @xmath334 term or the spin - orbit interaction . therefore the correlators must be calculated in the nilsson model for each nucleus separately . in this appendix we first briefly explain the derivation of the iew sum rule @xmath223 given in ( [ scsg ] ) and ( [ ss ] ) , before turning to the derivation of the ew sum rule @xmath335 given in ( [ ewfin ] ) . we consider the @xmath93 limit of the correlator @xmath336 of ( [ pimm ] ) , using the form ( [ simple ] ) . as explained in the main text below eq.([iso ] ) , we can replace the orbital @xmath166-factors by their isovector parts , @xmath196 . this gives @xmath337 using the form ( [ simple ] ) and the definitions @xmath338 , @xmath339 , @xmath340 , we easily obtain for the individual terms in ( [ ins ] ) : @xmath341 @xmath342 where @xmath343 means the real part of @xmath201 , and @xmath344 using these forms in ( [ ins ] ) , we arrive at ( [ scsg ] ) and ( [ ss ] ) of the main text . we first briefly recapitulate the evaluation of ( [ ewfree ] ) and ( [ ewint ] ) for the case @xmath228 . we make use of the identity ( [ rel ] ) to express the result in terms of the @xmath8 bubble graph @xmath345 , and the self consistency relation ( [ gap ] ) for @xmath144 to express the final result in terms of the quadrupole field @xmath346 : @xmath347 because of the self consistency relation ( [ gap ] ) for exact symmetry , it is clear that this expression vanishes if we sum over @xmath11 or @xmath348 . therefore we can replace the orbital @xmath166-factors in the first term of ( [ pimm ] ) by their isovector parts , i.e. , @xmath196 , and @xmath349 . the result can again be simplified by using the relation ( [ gap ] ) , and becomes finally @xmath350 next we will show that the second and third terms in the correlator ( [ pimm ] ) vanish in the limit which is needed for the ew sum rule of ( [ ew ] ) . for this purpose , we need the counterparts of some of the identities in the main text for the spin operator . the identities analogous to ( [ comm0 ] ) and ( [ rel ] ) are @xmath351 & = & \sqrt{2 } \gamma_{\tau } v_{\tau}^{\pm 1 } \label{spin1 } \\ \langle \alpha| s_{\tau}^{\pm 1 } |i \rangle & = & \frac{\sqrt{2}\gamma_{\tau}}{\omega_{\alpha i } } \langle \alpha |v_{\tau}^{\pm 1 } |i \rangle \ , . \label{spin2}\end{aligned}\ ] ] here @xmath26 is the strength parameter of the spin - orbit interaction ( see eq.([h0 ] ) ) , and we defined the operator @xmath352 as the tensor product of order @xmath266 of the orbital and spin angular momentum operators . in the notation of first quantization , @xmath353_{(1)}^k = \frac{i}{\sqrt{2 } } \left ( \vec{l } \times \vec{s } \right)^k\ , , \label{v}\end{aligned}\ ] ] where the product in the last expression denotes the usual vector product . then , without making use of any symmetry constraints , the following identity for the correlator @xmath354 with one arbitrary operator ( @xmath33 component @xmath43 ) and the spin operator @xmath355 can be derived ( compare to the relation ( [ wt1 ] ) in the main text ) : @xmath356 \rangle\ , . \label{wts}\end{aligned}\ ] ] by setting @xmath93 in this relation , we obtain the identity ( compare to ( [ w1 ] ) ) @xmath357 \rangle\ , . \label{gs}\end{aligned}\ ] ] in particular , for the case @xmath132 this identity implies that @xmath358 , because the quadrupole operator obviously commutes with the spin operator . by using the rpa equation ( [ rpag1 ] ) for @xmath359 and @xmath360 , we see that also the non - interacting correlator vanishes for @xmath93 , i.e. , @xmath361 returning now to the evaluation of ( [ ewfree ] ) and ( [ ewint ] ) of the main text for the case @xmath215 and @xmath301 , we obtain from the identities ( [ rel ] ) and ( [ spin2 ] ) @xmath362 where we used ( [ vanish ] ) in the last step . we therefore conclude that there are no @xmath304 crossing contributions , corresponding to the second and third terms in ( [ pimm ] ) , to the ew sum rule ( [ ew ] ) . turning finally to the @xmath234 contribution to the ew sum rule , we obtain for the case @xmath363 in ( [ ewfree ] ) and ( [ ewint ] ) @xmath364 where we used the result ( [ vanish ] ) in the last step . we therefore see that in our present schematic model , which does not include spin - spin interactions , the @xmath234 correlator in the limit @xmath224 is simply the non - interacting one . in order to express @xmath365 in terms of a ground state expectation value , we use the identity ( [ gs ] ) for the case @xmath366 . the commutator on the r.h.s . can be decomposed into tensors of rank @xmath367 , @xmath266 and @xmath368 . because only the scalar term can have a ground state expectation value , the identity ( [ gs ] ) gives @xmath369 from the rpa equation ( [ rpag1 ] ) for @xmath370 and the result ( [ vanish ] ) , we see that @xmath371 , and we obtain from ( [ ssew ] ) and ( [ pivv ] ) @xmath372 using this result in the @xmath234 term of the ew sum rule , which arises from the last term in ( [ pimm ] ) , we obtain the result given in eq.([ewfin ] ) of the main text . it is clear from the identity ( [ pivv ] ) that @xmath373 , because the correlator @xmath374 is positive due to the spectral representation ( [ pi ] ) . it is also easy to see that this term is nothing but the contribution from the mean field hamiltonian ( @xmath375 ) to the familiar double commutator , i.e. @xmath376 , m^{-1 } \right ] | 0 \rangle = \frac{2}{3 } \sum_{\tau } h_{\tau}^2 \gamma_{\tau } \langle { \bold l}_{\tau } \cdot { \bold s}_{\tau } \rangle \ , . \label{ewh0}\end{aligned}\ ] ] in the diagrammatic approach used in the main text of this paper , this is reflected by the fact that the second term in ( [ ewfin ] ) comes from the non - interacting correlator @xmath377 , see in particular eq.([ssew ] ) . we finally add a remark on eq.([vanish ] ) : because the operators @xmath378 and @xmath78 have the same @xmath53-symmetry ( both @xmath53-even ) , the identity @xmath358 does not follow from the spectral representation ( [ spec1 ] ) . also , this identity does not rely on rotational symmetry , i.e. , it holds also for finite symmetry breaking parameters @xmath119 . therefore , if we use @xmath379 on the l.h.s . of the identity ( [ w1 ] ) of the main text for finite @xmath119 , we can confirm that @xmath380 . d. bohle , a. richter , w. steffen , a.e.l . dieperink , n. lo iudice , f. palumbo , and o. scholten , phys . lett . * b 137 * ( 1984 ) 27 . n. lo iudice , rivista nuovo cimento * 9 * ( 2000 ) 1 . n. lo iudice , and f. palumbo , phys . * 41 * ( 1978 ) 1532 . f. iachello , nucl . phys . * a 358 * ( 1981 ) 89c . k. heyde , p. von neumann - cosel , and a. richter , rev . * 82 * ( 2010 ) 2365 . n. lo iudice , nucl . phys . * a 605 * ( 1996 ) 61 . d. zawischa , j. phys . g * 24 * ( 1998 ) 683 . a. bohr , and b.r . mottelson , _ nuclear structure _ , vol . ii ( world scientific , 1998 ) . w. bentz , a. arima , j. enders , a. richter , j. wambach , phys . * 84 * ( 2011 ) 014327 . h. ui and g. takeda , prog . theor . phys . * 70 * ( 1983 ) 176 ; + k. fujikawa and h. ui , prog . theor . * 75 * ( 1986 ) 997 . t. papenbrock , nucl . phys . * a 852 * ( 2011 ) 36 ; + t. papenbrock and h. a. weidenmller , arxiv:1307.1181 [ nucl - th ] . e. lipparini , and s. stringari , phys . lett . * b 130 * ( 1983 ) 139 . e. lipparini , and s. stringari , phys . rept . * 175 * ( 1989 ) 103 . n. lo iudice , phys . * c 57 * ( 1998 ) 1246 . j. enders , p. von neumann - cosel , c. rangacharyulu , and a. richter , phys . rev . * c 71 * ( 2005 ) 014306 . d. zawischa , m. macfarlane , and j. speth , phys . rev . * c 42 * ( 1990 ) 1461 . w. bentz , a. arima , a. richter , and j. wambach , _ analytic approach to nuclear rotational states : the role of spin - a model including pairing and spin dependent interactions- _ ( in preparation ) . rowe , nuclear collective motion , methuen and co. , 1970 . a. messiah , quantum mechanics ( dover publications , 1999 ) , appendix c. d.r . inglis , phys . rev . * 96 * ( 1954 ) 1059 . y. takahashi , nuovo cim . * 6 * ( 1957 ) 371 . migdal , _ theory of finite fermi systems and applications to atomic nuclei _ , wiley , new york , 1967 . nilsson , mat . * 29 * , no.16 ( 1955 ) 1 . y. nambu , physica * 15 d * ( 1985 ) 147 . s. frauendorf , rev . * 82 * ( 2010 ) 2365 ; + l.f . zhao , s.q . zhang , p. ring , j. meng , phys . rev . * c 85 * ( 2012 ) 024318 . w. bentz , _ analytic approach to nuclear rotational states : the role of spin - a model with dynamical breaking of spin symmetry - _ ( in preparation ) . p. ring , and p. schuck , the nuclear many - body problem , springer , 1980 . see for example : k. shimizu , z. physik * 278 * ( 1976 ) 201 .

we report results from a search for gravitational waves produced by perturbed intermediate mass black holes ( imbh ) in data collected by ligo and virgo between 2005 and 2010 . the search was sensitive to astrophysical sources that produced damped sinusoid gravitational wave signals , also known as ringdowns , with frequency and decay timescale characteristic of those produced in mergers of imbh pairs . no significant gravitational wave candidate was detected . we report upper limits on the astrophysical coalescence rates of imbhs with total binary mass and component mass ratios of either 1:1 or 4:1 . for systems with total mass , we report a 90%-confidence upper limit on the rate of binary imbh mergers with non - spinning and equal mass components ofmpcyr . we also report a rate upper limit for ringdown waveforms from perturbed imbhs , radiating 1% of their mass as gravitational waves in the fundamental , , oscillation mode , that is nearly three orders of magnitude more stringent than previous results .

1403.5306

intermediate mass black hole ( imbh ) binary systems represent a potential strong source of gravitational radiation accessible to ground - based interferometric detectors such as the laser interferometer gravitational - wave observatory ( ligo ) @xcite and virgo @xcite . although yet to be discovered , binary systems with total masses in the range @xmath8 could form in dense star clusters such as globular clusters @xcite . the coalescence of a compact binary system generates a gravitational wave signal consisting of a low frequency inspiral phase when the compact objects are in orbit around each other , a merger phase marking the coalescence of the objects and the peak gravitational wave emission , and a high frequency ringdown phase after the objects have formed a single perturbed black hole @xcite . for low mass systems , most of the signal - to - noise ratio comes from the inspiral phase of the coalescence . several searches for gravitational waves from the inspiral of low mass compact objects have been performed by ligo and virgo @xcite . however , since the merger frequency is inversely proportional to the mass of the system , it is shifted to lower frequencies for higher mass binaries . searches for gravitational waves from the inspiral , merger and ringdown of binary black holes with total masses @xmath9 have also been performed in ligo - virgo data @xcite . for an imbh binary , typically only the merger and ringdown parts of the signal fall above the low frequency cutoff of @xmath10hz for the initial ligo and virgo detectors . thus it is sufficient to conduct a search solely for these particular phases of the gravitational wave signal @xcite . a binary black hole merger is expected to result in a single perturbed black hole , and black hole perturbation theory and numerical simulations provide us with a well - understood ringdown signal model , a superposition of quasinormal modes that decay exponentially with time @xcite . indeed , any perturbed black hole , not just that produced by a compact merger ( e.g. , a black hole formed as the result of the core collapse of a very massive star @xcite ) , will emit ringdown gravitational waves described by its quasinormal modes . since the gravitational waveform of perturbed black holes has a well - defined model , the method of matched filtering is used to search for ringdown signals . the first such search was carried out on data from the fourth ligo science run ( s4 ) which took place between february 22 and march 24 , 2005 @xcite . additionally , two burst searches with less - constrained waveform models looked for gravitational waves from mergers of imbhs in data collected by ligo and virgo between 2005 and 2010 @xcite . no events were observed in these searches . in this paper , we present the results of a matched filter ringdown search of data from ligo s fifth and sixth science runs and virgo s science runs 2 and 3 . we compare the resulting rate upper limits to the previous searches for gravitational waves from imbhs . sections [ sec : source ] and [ sec : waveform ] describe the expected ringdown sources and waveform . section [ sec : data ] provides a brief description of the detectors and their sensitivities during the data collection epochs . section [ sec : search ] describes the search , and results are presented in section [ sec : results ] . upper limits are presented in section [ sec : ul ] and discussed in section [ sec : summary ] . observed black holes of known masses fall into two broad mass ranges . stellar mass black holes have masses @xmath11 @xmath12m@xmath13 @xcite although theoretical modeling of stellar evolution and population synthesis raises the possibility that significantly heavier stellar black holes could exist @xcite . supermassive black holes have masses @xmath14 @xmath15m@xmath13 @xcite and are thought to be cosmological in origin , possibly formed through galactic mergers leading to their growth through coalescences and accretion @xcite . the large gap between the mass ranges of stellar and supermassive black holes is predicted to be populated by an elusive class of objects known as intermediate mass black holes ( imbhs ) @xcite . observational evidence from ultra- or hyper - luminous x - ray sources and star cluster dynamics suggest a population of imbhs with masses in the range @xmath16m@xmath13 to @xmath17m@xmath13 @xcite . ultra - luminous x - ray sources with angle - averaged fluxes many times that of a stellar mass black hole accreting at the eddington limit @xmath18erg@xmath19s@xmath20 may be explained by black holes with masses larger than any known stellar mass black hole . the brightest known hyper - luminous x - ray source and the strongest imbh candidate is the point - like x - ray source hlx-1 . its maximum x - ray luminosity of @xmath21erg@xmath19s@xmath6 requires a black hole mass @xmath14 a few @xmath22m@xmath13 @xcite . other hyper - luminous x - ray sources include m82 x-1 @xcite , cartwheel n10 @xcite , and cxo j122518.6 @xcite . furthermore , the excess of dark mass at the centers of globular clusters could be explained by @xmath23m@xmath13 imbhs formed from repeated mergers between other compact objects and/or stars @xcite . however , both hyper - luminous x - ray sources and central globular cluster masses can be explained via phenomena that do not include imbhs @xcite . still , most observational evidence for globular cluster imbhs using radio emissions can place upper bounds of @xmath24m@xmath13 @xcite , and do not rule out lower mass systems that are above the expected maximum mass of a normal stellar mass black hole @xcite . thus , the existence of imbhs currently remains speculative . numerical simulations suggest that imbh binaries could form in collisional runaway scenarios in young dense star clusters . initially , in young star clusters , imbhs could form via the runaway collapse of very massive stars @xcite . after separate formation , two imbhs could settle to the core of the cluster through dynamical friction and form a common binary via dynamical interactions . the binary would tighten due to three - body encounters , finally merging quickly via gravitational radiation @xcite . from @xcite , we know that the astrophysical rate of imbh binary coalescence in globular clusters ( gc ) should be no higher than @xmath25gc@xmath6gyr@xmath6 assuming that all globular clusters are sufficiently massive and have a sufficient binary fraction to form this type of binary once in their lifetime of @xmath26gyr @xcite . also , globular clusters have a space density of roughly @xmath27gc@xmath19mpc@xmath5 @xcite . this allows us to convert the astrophysical upper limit to @xmath28mpc@xmath5yr@xmath6 . if we assume that only 10% of globular clusters meet these requirements , the rate would still be as high as one tenth this value @xcite . numerical simulations also suggest the possibility of forming intermediate mass ratio inspirals ( imris ) ( e.g. , a coalescence of an imbh with a compact stellar mass companion ) in these same dense star clusters . this occurs through a combination of gravitational wave emission , binary exchange processes , and secular evolution of hierarchical triple systems @xcite . ringdown searches in the advanced detector era could be important for detecting imris , particularly if the inspiraling companion is a black hole with @xmath29 @xmath30m@xmath13 or if the system is a compact object coalescing with a slowly - spinning imbh with @xmath29 @xmath31m@xmath13 @xcite . a black hole can be perturbed in a variety of ways , e.g. , by interaction with a companion , by accretion or infall of matter , or in its formation through asymmetric gravitational collapse . a perturbed kerr black hole will emit gravitational waves , relaxing to a stable configuration through radiation generated by a superposition of quasinormal modes of oscillation @xcite . the emitted gravitational waves are exponentially decaying sinusoid signals characterized by a complex angular frequency @xmath32 from which we can derive both the real frequency @xmath33 and the quality factor @xmath34 : @xmath35 where @xmath36 and @xmath37 are the spheroidal harmonic indices and @xmath38 denotes the overtones of each mode . overtones with @xmath39 are generally negligible in amplitude compared with the fundamental @xmath40 mode . numerical simulations have demonstrated that the @xmath41 fundamental mode dominates the gravitational wave emission , particularly in the case of an equal mass compact object merger @xcite . the ringdown search uses single - mode waveform templates . however , other modes can contribute significantly to the gravitational wave signal , particularly in cases where the binary s mass ratio @xmath42 where @xmath43 and @xmath44 . reference @xcite reports that single - mode templates can result in a loss @xmath45 in detected events over a significant mass range and also result in large errors in the estimated values of parameters ( especially the quality factor ) . a multimode ringdown search would perform better both in efficiency and parameter estimation @xcite . nevertheless , we show that the single - mode ringdown search will still provide good sensitivity to comparable mass binary systems ( see average sensitive distances given in section [ sec : eobul ] ) . the response of an interferometric detector to a gravitational wave is @xmath46 where @xmath47 and @xmath48 are the antenna pattern functions that depend on the direction to the source as described by a polar angle @xmath49 , an azimuthal angle @xmath50 , and a polarization angle @xmath51 . the plus and cross polarizations @xmath52 and @xmath53 of a single - mode @xmath54 ringdown waveform take the approximate form @xmath55 , \end{split}\ ] ] @xmath56 , \end{split}\ ] ] for @xmath57 where @xmath58 and @xmath59 are the oscillation frequency and the quality factor of the @xmath54 mode , @xmath60 is the distance to the source , @xmath61 is the initial phase of the mode , and @xmath62 is the inclination angle . the oscillation amplitude of the @xmath54 mode , @xmath63 , is given approximately by ( see appendix [ sec : appendix1 ] ) @xmath64 where @xmath65 is the gravitational constant , @xmath66 is the black hole mass , @xmath67 is the speed of light , @xmath68 , known as the ringdown efficiency , is the fraction of the black hole s mass radiated , @xmath69 where @xmath70 is the black hole s spin angular momentum , @xmath71 and @xmath72 $ ] [ cf . ( [ eq : f0 ] ) , ( [ eq : q ] ) , and ( [ eq : rdamp ] ) ] . the total ringdown efficiency of a black hole binary with non - spinning components is known to scale with the square of the symmetric mass ratio , @xmath73 , as @xmath74 @xcite . thus , for @xmath75 , @xmath76 and , for @xmath77 , @xmath78 . gravitational waves from extreme mass ratio systems will not be detectable unless the system is sufficiently close ( see section [ sec : eobul ] ) . a black hole binary with spinning components will radiate more energy if the spins are aligned with the orbital angular momentum and less if the spins are anti - aligned @xcite . the black hole mass @xmath66 and dimensionless spin parameter @xmath79 can be determined numerically using fitting formulae to kerr quasinormal mode frequency and quality factor parameters tabulated in table viii of @xcite . for the @xmath54 mode , the fits are of the form : @xmath80\ , , \label{eq : f0}\\ q & = & 0.7000 + 1.4187\left(1-\hat{a}\right)^{-0.4990}\ , . \label{eq : q}\end{aligned}\ ] ] these fitting functions allow us to relate a measurement of the frequency and quality factor from a match filter ringdown template to the mass and angular momentum of the final perturbed black hole . we can approximate the ringdown gravitational wave strain by @xmath81\ , , \ ] ] for @xmath57 where @xmath82 and @xmath83 is the effective distance to the source and @xmath84 is the effective initial phase depending on the initial phase @xmath61 as well as on the signal polarization [ see eq . ( 1.7 ) and ( 1.9 ) in @xcite ] . note that both @xmath84 and time of arrival at the detector @xmath85 are set to zero for simplicity in the template waveform given in section [ sec : algorithm ] . the data analyzed spans multiple science runs for both the ligo and virgo detectors . we report results both for data collected between november 2005 and september 2007 and between july 2009 and october 2010 . the first time period covers ligo s fifth science run ( s5 ) . the ligo site in hanford , washington hosted two collocated interferometers : a @xmath86 km detector h1 and a @xmath87 km detector h2 . the ligo site in livingston , la hosted one @xmath86 km detector l1 . additionally , the virgo @xmath27 km detector in cascina , italy operated from may 2007 to september 2007 during its first science run ( vsr1 ) which overlapped with the last few months of ligo s s5 run . however , this search did not analyze vsr1 data . thus , for the first time period , which we designate period 1 , we report results for the three - fold coincident search of the h1h2l1 detector network . we also report results for two - detector combinations of this network including h1l1 and h2l1 . we chose to exclude h1h2 coincident events since accurately measuring the significance of gravitational wave candidates is complicated by this network s correlated detector noise . the second time period covers ligo s sixth science run ( s6 ) during which only the h1 and l1 ligo detectors were operating . the virgo detector conducted two science runs during this period : vsr2 which ran from july 2009 to january 2010 , and vsr3 which ran from august 2010 to october 2010 . for this second time period , which we designate period 2 , we report results for the coincident search of the h1l1v1 detector network . we also report results for all two - detector combinations within this network . [ fig : rdhorizon ] ligo s s5 run marked the final data collection of the _ initial _ ligo detector configuration during which design sensitivity was achieved @xcite . figure [ fig : sensitivity ] ( _ left _ ) demonstrates the h1 , h2 , and l1 detectors sensitivities to ringdown signals from spinning black holes with @xmath88 and @xmath89 for typical period 1 performance . this figure shows the horizon distance @xmath90 divided by the square root of the ringdown efficiency @xmath68 , scaled to a canonical value @xmath89 , as a function of the final black hole mass . the horizon distance is the distance at which a given source with optimal location and orientation would produce a snr of 8 in a given detector ; some details of its derivation for ringdowns are given in appendix [ sec : appendix2 ] . dips in the ringdown horizon distance correspond directly to features of the detectors noise spectral density curves . for instance , the strong dip in sensitivity at @xmath91m@xmath13 is due to @xmath92hz electric power noise . the s6 run , during the phase of the _ enhanced _ ligo detector configuration , followed a series of upgrades to the initial detectors to improve sensitivity . these enhancements included a higher power laser and a new dc readout system @xcite . similarly , the virgo detector saw several improvements between its vsr1 and vsr2 runs including a more powerful laser , a thermal compensation system , and improved scattered light mitigation . before virgo s vsr3 run in early 2010 , monolithic suspensions with fused - silica fibers were installed @xcite . figure [ fig : sensitivity ] ( _ right _ ) demonstrates the h1 , l1 , and v1 detectors sensitivities to ringdown signals from spinning black holes with @xmath88 and @xmath89 for typical period 2 performance . gravitational - wave strain data from each of the detectors are known to be both non - gaussian and non - stationary . non - gaussianity is often manifested as noise transients , or glitches , in the strain data . efforts are made to diagnose and remove glitches and stretches of elevated noise from the data set using environmental and instrumental monitors @xcite . in this search , as in previous searches of ligo - virgo data , we apply three levels of data quality vetoes @xcite ( see appendix a of @xcite for more details ) . data remaining after the first and second veto levels have been applied are searched for possible detection candidates ( see section [ sec : results ] ) . data remaining after all three veto levels have been applied are searched for detection candidates and are also used in constraining the imbh merger rate ( see section [ sec : ul ] ) . table [ tab : livetime ] gives the total analyzed time after all three veto levels are applied and after the removal of the playground " data set used for pipeline tuning as described in section [ sec : sim ] . the total analysis time for both period 1 and period 2 was @xmath93years . .length of each network s total analyzed time after the third level of vetoes has been applied and the playground data set has been removed . [ cols="<,^,^ " , ] the ringdown search algorithm , first introduced in @xcite , is based on the optimal method for finding modeled signals buried in gaussian noise , the matched filter @xcite . the data from multiple gravitational wave detectors are match filtered with single - mode ringdown templates to test for the presence or absence of signals in the data . the output is a signal - to - noise ratio ( snr ) time series @xcite from which local maxima above a pre - determined snr threshold , called triggers , are retained for further analysis . since the noise in the detector data is non - stationary and non - gaussian , matched filtering alone is not enough to establish that a trigger is a gravitational wave signal . since detector noise can often mimic the signal for which we are searching , additional tests are employed including detector coincidence and snr consistency . we use a search pipeline similar to the ihope pipeline described in @xcite . here we summarize the main steps of the ringdown search pipeline . the data conditioning and segmentation is discussed in detail in @xcite . each segment of data is filtered using a bank of ringdown templates characterized by frequency @xmath94 and quality factor @xmath95 . following @xcite , the template used in this search is @xmath96 [ cf . ( [ eq : waveform ] ) ] , with a length of 10 e - folding times , @xmath97.hz ) and low dimensionless spin parameter ( @xmath98 ) where significant power is lost by using a cosine template @xcite . however , allowing an arbitrary phase would increase the noise level of the search . furthermore , any ringdown signal would follow a preceding waveform and there is some arbitrariness in the division of one from the other . ] the template bank is tiled in ( @xmath94 , @xmath95)-space according to the analytic approximate metric computed assuming white detector noise as described in @xcite so that no point in the parameter space has an overlap of less than 97% with the nearest template . the template parameters cover a frequency band between @xmath99hz and @xmath87khz and quality factor in the physical range between 2 and 20 . this corresponds roughly to masses in the range @xmath100 to @xmath101 , and spins in the range 0 to 0.99 . a fixed bank of 616 templates was used for all detectors . triggers with an snr statistic above a predetermined threshold @xmath102 are retained for further analysis . for both period 1 and period 2 , we set @xmath103 . for the least sensitive detector in each analysis period , we set lower thresholds : @xmath104 and @xmath105 . once triggers are found in a single detector , we apply a coincidence test , analogous to the one introduced in @xcite , to check for multi - detector parameter and arrival time consistency . in order to include information about time coincidence d@xmath106 and template coincidence for d@xmath94 and d@xmath95 in a single coincidence test , we construct a 3d - metric @xcite to calculate the distances d@xmath107 between two triggers in ( @xmath108 , @xmath95 , @xmath106)-space . the quantity @xmath109 is a measure of normalized signal mismatch . to account for the finite travel time between non - collocated detectors , we minimize d@xmath107 for each detector pair over a range of allowed time differences . only pairs of triggers for which d@xmath110 are kept as coincident candidates . during times when three detectors are operating , triple coincident events are constructed from sets of three triggers if each trigger in the set passes the coincidence test with every other one . we also consider h1l1 coincidences in a h1h2l1 network . we also apply second and third level vetoes to segments of poor data quality as described in @xcite . additionally , for period 1 , we apply a number of amplitude consistency tests that exploit the coalignment of h1 and h2 @xcite . these tests allow us to apply cuts to reduce the background of false alarms . finally , the pipeline ranks the coincidences and determines significance . for this purpose , a detection statistic is designed to separate signal - like coincidences from noise - like coincidences . given the large number of parameters that describe multi - detector coincidences , we employ a multivariate analysis using cuts on multiple parameters to help in classifying coincidences as signals or false alarms : i.e. , a multivariate statistical classifier . the parameters provided to the classifier to aid in characterizing the multi - detector coincidences included single - detector snrs and differences in time and template parameters between detectors , recovered effective distances , composite snr statistics , the 3d - metric distance between triggers and the metric coefficients as well as data quality information from the hierarchical veto method described in @xcite . additional details of these parameters will be described in a future paper . to perform the multivariate analysis , we use a machine learning algorithm known as a random forest of bagged decision trees @xcite . similar techniques have been implemented for detecting gravitational - wave bursts @xcite and cosmic strings @xcite . the training of the classifier uses two sets of data : a collection of coincidences associated with simulated signals and a collection of accidental coincidences that act as a proxy for the background . the simulated signal set is generated by adding software - generated gravitational waveforms to the data and running a separate search . the simulated waveforms , described in more detail in section [ sec : sim ] , included both full coalescence imbh merger signals and lone ringdown signals . the set of accidental coincidences is generated using the method of time - shifted data that takes advantage of the fact that a real signal will produce triggers in each detector that are coincident in time . the data streams of detectors are shifted in time with respect to one another by intervals longer than the light travel time between sites plus timing uncertainties , then a search for coincidences is performed . these time - shifted coincidences are then almost certainly due to noise . for period 1 , the l1 data stream was shifted by multiples of 5 seconds relative to h1 and h2 for a total of 100 time - shifted analyses ; the h1 and h2 data streams were not time - shifted relative to one another . for period 2 , the l1 data stream was shifted by multiples of 5 seconds and the v1 data stream was shifted by multiples of 10 seconds relative to h1 for a total of 100 time - shifted analyses . the classifier assigns a likelihood ranking statistic @xmath111 to each coincidence . a high likelihood implies the coincidence is signal - like ; a low likelihood implies the coincidence is noise - like . for each candidate , we need to be able to assign a significance to its likelihood ranking . this is done by mapping a false alarm rate ( far ) to a candidate s rank in order to assess its significance . we count the number of false coincidences in the time - shifted searches , record their likelihood values , and determine the analysis time @xmath112 of all the time - shift searches for a particular experiment time ( e.g. , h1l1 coincidences in a h1l1v1 network , h1l1v1 coincidences in a h1l1v1 network , etc . ) . we perform this calculation separately for each type of coincidence in each of the different experiment times . then , for each candidate in each experiment , we determine the far at its likelihood value @xmath113 with the expression : @xmath114 where @xmath115 is the measured number of coincidences with @xmath116 in the @xmath117 shifted analysis . we performed a total of 100 time - shifted analyses . finally , we can rank candidates by their fars across all types of experiment times into a combined ranking , known as combined far , for a single experiment time as described in detail in @xcite . the combined far is the final detection statistic that allows us to combine the candidate rankings from the various experiment types into a single list of candidates ordered from most significant to least significant . the analysis was tuned using the set of false alarm coincidences obtained from time - shifted searches , a set of simulated signals ( injections " ) added to the detectors data streams in a separate stage of data analysis , and a small chunk of the actual search data , approximately 10% , designated playground " , that was later excluded from the analysis to preserve blindness . the goal of tuning the analysis is to maximize the sensitivity of the search while minimizing the false alarm rate . for this , we injected a set of ringdown - only waveforms with @xmath89 into the data set . the waveforms were determined by eq . ( [ eq : hoft ] ) , ( [ eq : hplus ] ) , and ( [ eq : hcross ] ) with sky location and source orientation sampled from an isotropic distribution . several sets of ringdown waveforms were injected with a uniform distribution in @xmath94 and @xmath95 to cover the parameter range of the ringdown template bank . also , in order to cover the broad mass and spin range accessible to the ringdown search when signals have @xmath89 , several sets of ringdown waveforms were injected with a uniform distribution in @xmath66 and @xmath79 : @xmath118 and @xmath119 . additionally , we also injected a set of full coalescence waveforms with isotropically - distributed sky location and source orientation parameters into the data . these full coalescence waveforms included the recently - implemented non - spinning eobnrv2 family @xcite and the spinning phenomb family @xcite . the eobnrv2 injections were distributed uniformly in total mass @xmath2 and in mass ratio @xmath120 . the phenomb injections were given the same mass distribution and a uniform dimensionless spin parameter @xmath121 where @xmath122 for the spin angular momentum @xmath70 and the mass @xmath123 of the two binary components . for a discussion of the injection sets used in computing rate upper limits , see section [ sec : ul ] . the search yielded no significant gravitational wave candidates , as all events were consistent , within 1 sigma , with the background from accidental coincidences . figure [ fig : ifar ] shows the cumulative distributions of coincident events found as a function of inverse combined false alarm rate after all vetoes up to the third level are applied . these plots combine results from both triple and double coincident searches over the total analysis time of period 1 and period 2 . the most significant event was found in triple coincidence during period 1 in h1 , h2 , and l1 . after the first and second level vetoes were applied , it was found with a combined @xmath124yr@xmath6 and , after the third level vetoes were additionally applied , with a combined @xmath125yr@xmath6 . thus we expect an accidental coincidence to be found by the search with this significance @xmath126 once per two years of analysis . since the total analysis time was @xmath93years , the event is consistent , within 1 sigma , with the accidental coincidence rate . in both h1 and h2 , a trigger was found barely above threshold with matched filter snrs of 5.5 and 4.4 , respectively . however , the candidate was found as a very loud trigger in l1 with a matched filter snr of 48.9 . performing a coherent bayesian parameter estimation follow - up @xcite on these triggers , we found that a coherent analysis favored a solution for the binary s sky location and orientation that yield a very strong signal in l1 , but virtually no response in h1 and h2 detectors . while it is theoretically possible that very particular location and orientation parameters could produce such a signal , an excursion from stationary , gaussian noise ( a glitch ) in l1 is more likely . through 5@xmath127 region of the expected background from accidental coincidences . no search candidates stand out from the background.,scaledwidth=50.0% ] in this section , we compute the 90%-confidence upper limits on imbh coalescence rates and imbh black hole ringdown rates . the former will allow us to make an astrophysical statement as well as to compare the sensitivity of the ringdown search to various other searches that have made statements in this mass regime , including @xcite . we used a procedure similar to that discussed in @xcite for the upper limit calculation based on the loudest event statistic @xcite . in order to capture the variability of the detector noise and sensitivity , we analyzed the data in periods of @xmath128 to @xmath129 months . in each of these analysis times , we estimate the volume to which the ringdown search is sensitive by injecting many simulated signals into the data and performing an analysis to recover them . in section [ sec : eobul ] , we describe the distribution of eobnrv2 waveforms used to model the source population of imbh binaries . our sensitivity to these signals depends on total mass , mass ratio , source distance , and sky location as well as other parameters such as component spins . we explore the changing sensitivity of the ringdown search to these binaries over a range of total masses for both equal mass and 4:1 mass ratio systems . other distance and orientation parameters are randomly sampled . due to the significant variation of the search sensitivity over the large mass and mass ratio parameter space that we explore in section [ sec : eobul ] , we have chosen to include only systems with non - spinning components in this study . in section [ sec : rdul ] , we describe the distribution of ringdown waveforms used to model the population of perturbed black holes first explored in @xcite . for each of these injection sets , we compute the sensitive volume for a given mass range and mass ratio by integrating the efficiency of the search over distance : @xmath130 where the efficiency @xmath131 is calculated as the number of injections found with a lower combined far than the most significant coincident event in each analysis time for the search divided by the total number of injections made at a given distance . as described in @xcite , we estimate the likelihood parameter @xmath132 of the loudest event being a signal versus being caused by an accidental coincidence for each type of coincident network time and each mass and mass ratio bin . for each analysis time ( excluding playground time ) , effective volume from eq . ( [ eq : veff ] ) , and estimated @xmath132 , we marginalize over statistical uncertainties given in section [ sec : uncertainty ] and construct a marginalized likelihood as a function of the astrophysical rate in units of mergers per mpc@xmath133 per year for our eobnrv2 injection sets and in units of ringdowns per mpc@xmath133 per year for our ringdown injections . in order to obtain a combined posterior probability distribution for the rate over all the analysis times , we multiply a prior on the rate by the product of the marginalized likelihood functions to obtain a posterior probability and integrate to 90% to obtain the 90%-confidence upper limit on the rates . for our combined period 1 result , we assumed a uniform prior on the rate . however , for the main period 2 result , we were able to use the period 1 posteriors over coalescence or ringdown rate as priors for the upper limit calculation . we must account for several sources of random and systematic error when computing rate upper limits . uncertainties on the sensitive volume as well as incomplete knowledge of waveforms and source populations form the largest contributors . as described in earlier search papers @xcite , we marginalize over random uncertainty ( i.e. calibration and statistical monte carlo uncertainties ) for each analysis time . the 90%-confidence upper limits based on the marginalized posterior distributions are the main results of this search . the calibration of the data is a source of both random and systematic error . reference @xcite reports uncertainties on the magnitude of the response function for each detector in period 1 . we find an overall distance uncertainty of 8% . thus , the random uncertainty on the visible volume for period 1 is approximately 8% cubed , or 24% . for period 2 , references @xcite and @xcite report uncertainty on @xmath134 for ligo and virgo detectors additionally , an uncertainty on the scaling of @xmath134 was reported in @xcite and should be treated as a systematic error similar to the systematic waveform uncertainties discussed below that could over- or under - bias the amplitude of a signal . however , the uncertainty on the scaling of h(t ) also has an associated random error that we fold into the random uncertainty calculation for period 2 . we find an overall distance uncertainty of 14% corresponding to a 42% uncertainty on the visible volume for period 2 . see @xcite for a detailed explanation of how the uncertainties were propagated . in addition to the systematic error associated with the overall scaling of @xmath134 that could lead to amplitude bias as mentioned above , there is a larger source of systematic error due to differences between the injected model waveforms and the true waveform . for eobnrv2 waveforms below @xmath135m@xmath13 , comparisons with numerical models indicate that uncertainties in these waveforms result in @xmath136% systematic uncertainty in the snr , corresponding to a @xmath137% uncertainty in sensitive volume . for higher masses , the systematic uncertainty in the snr could be as high as 25% . due to our incomplete knowledge of the true waveform and its changing uncertainty over the mass range we have explored , no systematic errors associated with imperfect waveform modeling were applied to the rate upper limits reported in this paper . systematic errors were also not applied to previous searches @xcite using full coalescence waveforms up to @xmath138m@xmath13 and thus we can compare the upper limits directly with those results . a previous weakly modeled burst search @xcite used waveform errors of @xmath13915% . thus , in order to compare with these results , the upper limits reported here should be rescaled as described below . regarding ringdown waveforms , due to our lack of knowledge about the population of black holes producing the waveforms and the waveforms themselves , we again assign no systematic error to rate upper limits computed with ringdown waveforms . in general , we can rescale our rate upper limits by any systematic uncertainty by applying the scaling factor @xmath140 where @xmath127 is the systematic uncertainty . thus , we can apply a conservative systematic uncertainty of 15% by rescaling our rate upper limit upward by a factor of 1.63 . the statistical error originating from the finite number of monte carlo injections that we have performed is the final source of error for which we must account . these errors on the efficiency at a given distance are found to range between 1.7% and 6.2% and were marginalized over using the method described in @xcite . in order to evaluate the sensitivity of the ringdown search to waveforms from binary imbh coalescing systems with non - spinning components , we used a set of injections from the eobnrv2 waveform family described in section [ sec : sim ] . due to the variation in ringdown search sensitivity over different mass ratios , we chose to compute imbh coalescence rate upper limits separately for @xmath75 and @xmath77 . the injection sets were distributed uniformly over a total binary mass range from @xmath2 and upper limits were computed in mass bins of width @xmath99m@xmath13 . the final black hole spins of these injections can be determined from the mass ratios and zero initial component spins @xcite . for @xmath75 , we find @xmath141 , and for @xmath77 , we find @xmath142 . the average sensitive distances of the ringdown search to imbh binaries described by eobnrv2 signal waveforms for both @xmath75 and @xmath77 are shown in fig . [ fig : s5s6sensitivedist ] for period 1 and period 2 . the most sensitive mass bin in both cases is @xmath143 corresponding roughly to @xmath144 near the peak sensitivity of the ligo detectors . for @xmath75 , the average sensitive distance of the @xmath3 mass bin was @xmath145mpc . for @xmath77 , the average sensitive distance for this mass bin decreases by more than a factor of two to @xmath146mpc . as discussed in section [ sec : waveform ] , the reduced ringdown efficiency for @xmath77 binary systems leads to lower amplitude waveforms and hence , to lower average sensitive distances . additionally , the lower final black hole spin for @xmath77 binary systems acts to decrease the average sensitive distance relative to @xmath75 binary systems for which the final spin is larger . the sensitive distance of higher mass bins drops off significantly due to the steeply rising seismic noise in the detector at low frequencies . this affect is accentuated for @xmath77 systems relative to @xmath75 systems at a fixed mass because a smaller final spin leads to a lower frequency ringdown . the sensitive distance of mass bin @xmath147 is over an order of magnitude less than the sensitive distance of our most sensitive mass bins for both @xmath75 and @xmath77 cases . ( _ yellow _ ) , @xmath77 ( _ green _ ) ] and period 2 [ @xmath75 ( _ cyan _ ) , @xmath77 ( _ blue _ ) ] . these distances are equivalent to appropriate averages over each of the detector networks shown for period 1 and period 2 in table [ tab : livetime ] , weighted by the percentage of time analyzed for each network . thus , while in general the h1l1v1 and h1l1 networks during period 2 were more sensitive than the h1h2l1 and h1l1 networks during period 1 , the consistently smaller average sensitive distances for period 2 reflect the large duty cycle of its least sensitive detector networks compared to period 1.,scaledwidth=50.0% ] figure [ fig : s5s6ul ] shows the 90%-confidence upper limits on non - spinning imbh coalescence rates for a number of mass bins . we find an upper limit of 0.069@xmath148mpc@xmath5 yr@xmath6 on the coalescence rate of equal mass imbh binaries with non - spinning components and total masses @xmath3 . from the discussion of astrophysical rates of imbh mergers in section [ sec : source ] , we see that this rate upper limit is still several orders of magnitude away from constraining the astrophysical rate from gcs . yr@xmath6 as a function of total binary masses , evaluated using eobnrv2 waveforms with @xmath75 ( _ slate grey _ ) and @xmath77 ( _ grey _ ) . in both cases , upper limits computed using period 2 with period 1 as a prior are shown in a darker shade . overlaid in a lighter shade are upper limits computed using only period 1 data with a uniform prior on rate.,scaledwidth=50.0% ] previous searches for weakly - modeled burst signals found no plausible events @xcite . the most recent search reports a rate upper limit for non - spinning imbh coalescences of @xmath149mpc@xmath5yr@xmath6 at the 90%-confidence level for the mass bin centered on @xmath150m@xmath13 @xcite . a direct comparison of our @xmath75 upper limits shown in fig . [ fig : s5s6ul ] to this burst search result should be made with care due to the following differences between the two analyses : statistical approaches leading to different search thresholds , treatment of uncertainties , analyzed detector networks , and mass and distance binnings . additionally , while the ringdown search employed the bayesian formulation @xcite for calculating the rate upper limit , the burst search used a frequentist method . nevertheless , although the impact of the reported differences is hard to quantify , the upper limits determined by the two analyses can be considered consistent with each other . a more robust comparison of the sensitivity of the burst searches and an earlier version of the ringdown search without a multivariate classifier will be presented in a future paper @xcite . additionally , we can make a comparison with the upper limits reported from the matched filter search for gravitational waves from the inspiral , merger , and ringdown of non - spinning binary black holes with total masses @xmath151 @xcite . this search considered similar uncertainties and similar analyzed networks to those used by the ringdown search so a result comparison is fairly straight - forward . from table i of @xcite , we find that for systems with @xmath75 , the rate upper limits for masses @xmath152 to @xmath153 vary in the range @xmath154mpc@xmath155 to @xmath156mpc@xmath155 . from fig . [ fig : s5s6ul ] , we find a rate upper limit for mass bin @xmath157 of @xmath158mpc@xmath155 , a value consistent with the bbh rate upper limit range for these masses and mass ratio . note that we can rescale our rate upper limits by a 15% systematic uncertainty by applying the scaling factor of 1.63 as described in section [ sec : uncertainty ] . from fig . [ fig : s5s6ul ] , we find a rescaled rate upper limit of @xmath159mpc@xmath5yr@xmath6 for mass bin @xmath143 and @xmath160mpc@xmath5yr@xmath6 for mass bin @xmath161 . in order to compare with @xcite , we determined a 90%-confidence upper limit of @xmath162mpc@xmath5yr@xmath6 on rates of pure ringdowns from perturbed black holes with uniformly distributed masses @xmath163 , uniformly distributed spins @xmath164 , and a fixed ringdown efficiency of @xmath89 . we expect ringdown signals from imbh mergers to emit near this efficiency in the @xmath165 fundamental mode if the mass ratio is near unity . however , for other sources of perturbed black holes , such as a hypermassive star collapse directly to a perturbed imbh , we expect @xmath166 . thus , the rate upper limit reported in this section will not be applicable to such sources . reference @xcite placed a 90% confidence upper limit on the rate of ringdowns from black holes with frequencies distributed uniformly in log@xmath167 in the range @xmath168 and uniformly in quality factor @xmath169 of @xmath170mpc@xmath5yr@xmath6 . thus , a rough comparison indicates an improvement of nearly three orders of magnitude . a significant portion of this improvement results from a huge increase in the analysis time . due to the high false alarm rate in double coincident analysis time , an upper limit was set in @xcite using only triple coincident time , a total of @xmath171years . we analyzed both triple and double coincident time in both period 1 and period 2 , a total of @xmath93years . such an increase in analysis time results in a factor of @xmath126 32 improvement in the upper limit . additionally , a significant improvement in detector sensitivity due to detector upgrades between science runs contributed to a better upper limit . furthermore , since only triple coincident time was analyzed in @xcite , the sensitivity was limited by the least sensitive detector , h2 , which was shown to have a horizon distance of @xmath126 @xmath172mpc at @xmath173m@xmath13 as shown in fig . 2 in @xcite . however , since we analyzed both triple and double coincident triggers , the limiting detector was typically the l1 detector . we can compare the h2 horizon distance in fig . 2 in @xcite to the l1 horizon distance in fig . [ fig : sensitivity ] at @xmath173m@xmath13 to see that the horizon distance of the limiting detector improved by a factor of @xmath126 3 for @xmath88 . since the upper limit scales with volume , a factor of @xmath126 3 in distance results in a factor of @xmath126 27 in the upper limit . however , we expect this factor of improvement to decrease for the lower masses on which the ringdown upper limit was set . thus , from the improvements both in analysis time and detector sensitivity , we find already roughly three orders of magnitude improvement . however , several caveats would apply to a direct comparison : different injection distributions in @xmath174-space , the improvements from pipeline enhancements such as the implementation of a machine - learning algorithm , differences in the fitting functions for final black hole mass and spin defined in eq . ( [ eq : f0 ] ) and ( [ eq : q ] ) , differences in the method used in the volume integral in eq . ( [ eq : veff ] ) , and differences in marginalization over errors . a careful study of the improvement due to the use of a machine - learning algorithm will be presented in a future paper . this paper presents the results of the search for ringdown gravitational waves in data collected by ligo and virgo between 2005 and 2010 . no significant gravitational wave candidate was identified . we place upper limits on the merger rates of non - spinning imbh binaries as well as on the rates of ringdowns from perturbed black holes . we conducted a detailed study of the pipeline s sensitivity to full coalescence imbh merger signals using non - spinning eobnrv2 waveforms . for simplicity , we focused our studies on only two mass ratios : @xmath75 and @xmath77 . the average sensitive distances in our most sensitive total mass bin , @xmath3 , indicate that the ringdown search is sensitive to an equal mass system at twice the distance of a 4:1 mass ratio system . the most efficiently detected mass bin gives an upper limit on the rate of non - spinning , equal mass imbh mergers with total masses @xmath3 of @xmath4mpc@xmath5yr@xmath6 . this does not account for any uncertainty in the waveform , which could be as high as 10% for the mass bin . our upper limits for ringdown waveforms from perturbed imbhs with masses @xmath175 and spins @xmath164 show an improvement of nearly three orders of magnitude over the previous result reported @xcite , which we can attribute to improved detector sensitivity , increased livetime , and pipeline enhancements . while our rate upper limits are still two to three orders of magnitude away from constraining the astrophysical imbh merger rate from globular clusters , we note that we will soon approach this optimistic rate with the improved sensitivity of advanced ligo and virgo detectors expected to begin operation in 2015 . with the improved low frequency performance of the advanced detectors , we will have sensitivity to gravitational waves from perturbed intermediate mass black holes with masses up to @xmath126 1000 to @xmath176 . at peak sensitivity , the advanced ligo ringdown horizon distance for black holes with @xmath89 will approach cosmological distances . * acknowledgments * the authors gratefully acknowledge the support of the united states national science foundation for the construction and operation of the ligo laboratory , the science and technology facilities council of the united kingdom , the max - planck - society , and the state of niedersachsen / germany for support of the construction and operation of the geo600 detector , and the italian istituto nazionale di fisica nucleare and the french centre national de la recherche scientifique for the construction and operation of the virgo detector . the authors also gratefully acknowledge the support of the research by these agencies and by the australian research council , the international science linkages program of the commonwealth of australia , the council of scientific and industrial research of india , the istituto nazionale di fisica nucleare of italy , the spanish ministerio de educacin y ciencia , the conselleria deconomia hisenda i innovaci of the govern de les illes balears , the foundation for fundamental research on matter supported by the netherlands organisation for scientific research , the polish ministry of science and higher education , the focus programme of foundation for polish science , the royal society , the scottish funding council , the scottish universities physics alliance , the national aeronautics and space administration , the carnegie trust , the leverhulme trust , the david and lucile packard foundation , the research corporation , and the alfred p. sloan foundation . the amount of energy d@xmath177 carried by gravitational radiation crossing an area d@xmath178 orthogonal to its propagation direction in a time d@xmath106 is given by the energy flux equation , @xmath179 where @xmath52 and @xmath53 are given by the generalized forms of eq . ( [ eq : hplus ] ) and ( [ eq : hcross ] ) for an arbitrary location on a 2-sphere with @xmath180 and time of arrival @xmath85 set to zero . taking the time derivative and squaring the plus and cross polarizations , we find @xmath181 , \end{split}\ ] ] @xmath182 . \end{split}\ ] ] integrating this flux over a sphere with area element d@xmath183 , we find that the trigonometric functions simplify greatly , leaving only the exponential time dependence over which to integrate @xmath184 finally , we note that the energy radiated as gravitational waves during the ringdown phase is @xmath185 where @xmath68 is the ringdown efficiency discussed in sec . [ sec : waveform ] . thus , the amplitude can be found by solving eq . ( [ eq : denergy ] ) for @xmath186 , @xmath187 the ringdown horizon distance , similar to the inspiral horizon distance , is a useful measure of the sensitivity of the detectors to ringdown gravitational waves from a particular type of black hole . it is equal to the distance at which an optimally oriented and located imbh merger would produce an snr of 8 in the detector . the horizon distance is derived from the representative strain noise power spectral density of a detector and the @xmath188 , or root sum squared of the strain , for a signal with optimal orientation at @xmath189mpc . the definition of @xmath188 comes from the need to measure the amplitude of a gravitational wave without reference to a particular detector . in general , it is @xmath190 where @xmath52 and @xmath53 are given in eq . ( [ eq : hplus ] ) and ( [ eq : hcross ] ) for the single - mode @xmath54 ringdown waveform . here , under the assumption of optimal orientation , we set @xmath191 . we find that the @xmath188 takes the form @xmath192 where @xmath186 is derived in eq . ( [ eq : rdamp ] ) . if @xmath193 represents the fourier transform of the expected signal , then the average snr this signal would attain in a detector with spectral density @xmath194 is given by @xmath195 typically , the horizon distance is found by setting @xmath196 and solving for the distance @xmath60 which parameterizes the waveform @xmath197 . we can use the fact that the single - mode ringdown signal is quasi - monochromatic and @xmath194 assumes approximately one value for each @xmath94 so it can be treated as a constant : @xmath198 using parseval s theorem , we can write eq . ( [ eq : rho2 ] ) as @xmath199 also , since optimally oriented and located sources imply maximization over all the angles @xmath49 , @xmath50 , and @xmath51 in @xmath47 and @xmath48 , then @xmath200 and @xmath201 . this then gives us the result that @xmath202 ( which is defined for @xmath203 ) so eq . ( [ eq : rho3 ] ) becomes @xmath204 where @xmath205 is eq . ( [ eq : hrsssq ] ) evaluated at a distance of @xmath189mpc . then , we simply solve eq . ( [ eq : rho4 ] ) for the horizon distance , @xmath206 we then set @xmath196 to define the ringdown horizon distance used in fig . [ fig : sensitivity ] . 107ifxundefined [ 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://stacks.iop.org/0034-4885/72/i=7/a=076901 [ * * , ( ) ] http://stacks.iop.org/1748-0221/7/i=03/a=p03012 [ * * , ( ) ] link:\doibase 10.1142/s0218271804004426 [ * * , ( ) ] link:\doibase 10.1086/425910 [ * * , ( ) ] link:\doibase 10.1086/507106 [ * * , ( ) ] @noop * * , ( ) @noop * * , ( ) link:\doibase 10.1103/physrevd.79.122001 [ * * , ( ) ] link:\doibase 10.1103/physrevd.80.047101 [ * * , ( ) ] link:\doibase 10.1103/physrevd.85.082002 [ * * , ( ) ] link:\doibase 10.1103/physrevd.83.122005 [ * * , ( ) ] link:\doibase 10.1103/physrevd.87.022002 [ * * , ( ) ] link:\doibase 10.1103/physrevd.60.022001 [ * * , ( ) ] http://stacks.iop.org/0264-9381/22/i=18/a=s27 [ * * , ( ) ] link:\doibase 10.1103/physrevd.71.103005 [ * * , ( ) ] link:\doibase 10.1086/152444 [ * * , ( ) ] @noop * * , ( ) link:\doibase 10.1103/physrevd.2.2141 [ * * , ( ) ] link:\doibase 10.1086/180849 [ * * , ( ) ] link:\doibase 10.1103/physrevd.5.2439 [ * * , ( ) ] @noop * * , ( ) link:\doibase 10.1103/physrevd.30.295 [ * * , ( ) ] @noop * * ( ) link:\doibase 10.1088/0004 - 637x/706/2/1184 [ * * , ( ) ] link:\doibase 10.1007/s10509 - 009 - 0134 - 3 [ * * , ( ) ] link:\doibase 10.1088/0004 - 637x/752/1/43 [ * * , ( ) ] link:\doibase 10.1103/physrevd.80.062001 [ * * , ( ) ] link:\doibase 10.1103/physrevd.85.102004 [ * * , ( ) ] @noop ( ) , @noop * * , ( ) link:\doibase 10.1086/523755 [ * * , ( ) ] link:\doibase 10.1086/588096 [ * * , ( ) ] link:\doibase 10.1088/0004 - 637x/714/2/1217 [ * * , ( ) ] link:\doibase 10.1088/0004 - 637x/759/1/52 [ * * , ( ) ] , link:\doibase 10.1086/378122 [ * * , ( ) ] link:\doibase 10.1038/nature11592 [ * * , ( ) ] link:\doibase 10.1007/s00159 - 010 - 0029-x [ * * , ( ) ] link:\doibase 10.1103/physrevlett.111.151101 [ * * , ( ) ] link:\doibase 10.1086/174548 [ * * , ( ) ] http://stacks.iop.org/1538-4357/551/i=1/a=l27 [ * * , ( ) ] link:\doibase 10.1086/338118 [ * * , ( ) ] link:\doibase 10.1046/j.1365 - 8711.2002.05112.x [ * * , ( ) ] link:\doibase 10.1038/nature08083 [ * * , ( ) ] http://stacks.iop.org/2041-8205/747/i=1/a=l13 [ * * , ( ) ] in @noop _ _ , , vol . , ( ) p. @noop * * , ( ) link:\doibase 10.1111/j.1365 - 2966.2010.16943.x [ * * , ( ) ] http://stacks.iop.org/1538-3881/124/i=6/a=3255 [ * * , ( ) ] link:\doibase 10.1086/344584 [ * * , ( ) ] link:\doibase 10.1086/497023 [ * * , ( ) ] http://stacks.iop.org/1538-4357/589/i=1/a=l25 [ * * , ( ) ] link:\doibase 10.1111/j.1365 - 2966.2005.08634.x [ * * , ( ) ] link:\doibase 10.1051/0004 - 6361/201118022 [ * * , ( ) ] link:\doibase 10.1088/0004 - 637x/743/1/11 [ * * , ( ) ] link:\doibase 10.1051/0004 - 6361/201218928 [ * * , ( ) ] link:\doibase 10.1088/0004 - 637x/745/2/175 [ * * , ( ) ] link:\doibase 10.1088/2041 - 8205/750/2/l27 [ * * , ( ) ] link:\doibase 10.1088/0004 - 6256/142/4/113 [ * * , ( ) ] @noop * * , ( ) http://stacks.iop.org/1538-4357/528/i=1/a=l17 [ * * , ( ) ] link:\doibase 10.1086/381968 [ * * , ( ) ] link:\doibase 10.1016/s1384 - 1076(96)00003 - 6 [ * * , ( ) ] http://stacks.iop.org/0004-637x/599/i=2/a=1129 [ * * , ( ) ] link:\doibase 10.1088/0264 - 9381/27/17/173001 [ * * , ( ) ] link:\doibase 10.1086/588246 [ * * , ( ) ] @noop * * , ( ) link:\doibase 10.1086/339472 [ * * , ( ) ] link:\doibase 10.1086/424809 [ * * , ( ) ] link:\doibase 10.1103/physrevd.76.061504 [ * * , ( ) ] link:\doibase 10.1103/physrevd.75.124018 [ * * , ( ) ] link:\doibase 10.1103/physrevd.76.104044 [ * * , ( ) ] link:\doibase 10.1103/physrevd.76.064034 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.109.141102 [ * * , ( ) ] link:\doibase 10.1103/physrevd.85.024018 [ * * , ( ) ] link:\doibase 10.1103/physrevd.74.041501 [ * * , ( ) ] link:\doibase 10.1103/physrevd.73.064030 [ * * , ( ) ] link:\doibase 10.1103/physrevd.88.122002 [ * * , ( ) ] https://dcc.ligo.org/ligo-t060156/public [ ] ( , ) http://stacks.iop.org/0264-9381/28/i=2/a=025005 [ * * , ( ) ] http://stacks.iop.org/0264-9381/25/i=18/a=184004 [ * * , ( ) ] http://stacks.iop.org/0264-9381/27/i=19/a=194010 [ * * , ( ) ] http://stacks.iop.org/1742-6596/243/i=1/a=012005 [ * * , ( ) ] @noop ph.d . thesis , ( ) http://stacks.iop.org/0264-9381/27/i=16/a=165023 [ * * , ( ) ] http://books.google.com/books?id=rnayaaaamaaj[__ ] , international series of monographs on electronics and instrumentation ( , ) link:\doibase 10.1103/physrevd.87.024033 [ * * , ( ) ] @noop ph.d . thesis , ( ) link:\doibase 10.1103/physrevd.68.102003 [ * * , ( ) ] link:\doibase 10.1143/ptp.111.781 [ * * , ( ) ] , link:\doibase 10.1103/physrevd.78.062002 [ * * , ( ) ] link:\doibase 10.1088/0264 - 9381/28/23/235005 [ * * , ( ) ] @noop * * , ( ) @noop ( ) link:\doibase 10.1103/physrevd.88.062006 [ * * , ( ) ] @noop * * , ( ) @noop ph.d . thesis , ( ) link:\doibase 10.1103/physrevd.84.124052 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.106.241101 [ * * , ( ) ] link:\doibase 10.1103/physrevd.88.062001 [ * * , ( ) ] http://stacks.iop.org/0264-9381/26/i=17/a=175009 [ * * , ( ) ] http://stacks.iop.org/0264-9381/30/i=7/a=079502 [ * * , ( ) ] \doibase http://dx.doi.org/10.1016/j.nima.2010.07.089 [ * * , ( ) ] https://dcc.ligo.org/docdb/0033/t1100071/009/errorbudget.pdf [ ] ( , ) @noop ( ) , https://dcc.ligo.org/ligo-g1300649 [ ] ( , ) link:\doibase 10.1088/0004 - 637x/704/1/l40 [ * * , ( ) ] https://dcc.ligo.org/p1100198/public [ `` , '' ] ( )

in search of the counterpart to the brightest unidentified gamma - ray source 3eg j2020 + 4017 ( 2cg078 + 2 ) we report on new x - ray and radio observations of the-cygni field with the _ chandra _ x - ray observatory and with the green bank telescope ( gbt ) . we also report on reanalysis of archival rosat data . with _ chandra _ it became possible for the first time to measure the position of the putative gamma - ray counterpart rx j2020.2 + 4026 with sub - arcsec accuracy and to deduce its x - ray spectral characteristics . these observations demonstrate that rx j2020.2 + 4026 is associated with a k field star and therefore is unlikely to be the counterpart of the bright gamma - ray source 2cg078 + 2 in the snr g78.2 + 2.1 as had been previously suggested . the _ chandra _ observation detected 37 additional x - ray sources which were correlated with catalogs of optical and infrared data . subsequent gbt radio observations covered the complete 99% egret likelihood contour of 3eg j2020 + 4017 with a sensitivity limit of which is lower than most of the recent deep radio search limits . if there is a pulsar operating in 3eg j2020 + 4017 , this sensitivity limit suggests that the pulsar either does not produce significant amounts of radio emission or that its geometry is such that the radio beam does not intersect with the line of sight . finally , reanalysis of archival rosat data leads to a flux upper limit of for a putative point - like x - ray source located within the 68% confidence contour of 3eg j2020 + 4017 . adopting the snr age of 5400 yrs and assuming a spin - down to x - ray energy conversion factor of this upper limit constrains the parameters of a putative neutron star as a counterpart for 3eg j2020 + 4017 to be , and g.

astro-ph0405146

the error boxes of unidentified gamma - ray sources are usually large , and thus the task of finding appropriate candidate counterparts at other wavelengths has not been easy . about 20 bright point - like gamma - ray sources were found near the galactic plane using cos - b ( swanenburg et al . 1981 ) some of which may be concentrations of molecular hydrogen ( mayer - hasselwander & simpson 1990 ) . another few , such as the crab and vela pulsars , were identified based on their periodic emission ( thompson et al . the nature of the other objects remained unknown . the much more sensitive egret telescope on the _ compton gamma - ray observatory _ ( _ cgro _ ) was expected to contribute decisively to the identification of the cos - b sources . and indeed , the higher count rates and tighter source locations provided by egret confirmed the existence of most of the cos - b sources and led to the identification of several other sources . most prominent were the gamma - ray pulsars geminga and psr b170644 which could be identified on the basis of detections at x - ray and radio wavelengths ( see e.g. kanbach 2002 and becker & pavlov 2001 for a review and references ) . at high galactic latitudes about 90 new high - energy sources could be correlated with blazars . the final egret catalog of gamma - ray sources lists 271 objects ( hartman et al . 1999 ) of which about 170 are unidentified . their distribution suggests that most of them are galactic . the origin and nature of this population of extremely energetic objects is clearly of interest . seven of the galactic gamma - ray sources are rotation - powered pulsars , identified through the periodic modulation of their gamma - ray fluxes . these seven are also persistent , point - like sources at gamma - ray energies . in the 100 mev 1 gev energy range , these sources have hard , power - law - like spectra with high - energy cut - offs at a few gev . although rotation - powered pulsars are best known as radio sources , this is not true for all geminga , for example , shows at best marginal evidence of pulsed radio emission ( kuzmin & losovkii 1997 ) . geminga is thus taken as the prototype of a ` radio - quiet ' gamma - ray pulsar of which many more should exist in the galaxy . although geminga s gamma - ray luminosity is rather low ( its small distance of about 160 pc makes it a bright source ) the property of radio faintness could be indicative of pulsar emission where the beamed radiation at different wavelengths is emitted into widely different directions . such a model may be applicable to young , high luminosity pulsars as well ( yadigaroglu & romani 1995 ) . there are other models to explain geminga s radio faintness though ( e.g. gil et al . 1998 ) . a review of the spectra of unidentified low galactic latitude egret sources ( bertsch et al . 2000 ; merck et al . 1996 ) shows that about 10 objects exhibit the very hard power - law type spectra with a cut - off at several gev as seen also in the identified pulsars . these objects would be prime targets for identification efforts at other wavelengths . relatively deep radio searches ( at 770 mhz ) at the positions of several of these sources have not found radio counterparts ( nice & sayer 1997 ) . population studies of the unidentified gamma - ray sources close to the galactic plane indicate that their luminosities are also quite compatible with the luminosities of the younger identified pulsars ( kanbach et al . suggestions , other than pulsars , for the nature of these gamma - ray sources have also been widely discussed . energetic objects , like massive young stars or ob associations and snrs have been correlated with the 3eg catalog ( e.g. romero et al . 2000 ) and certainly indicate a close relationship with the gamma - ray sources . multi - wavelength observations focusing on promising candidate sources have been quite successful in recent years . observations in x - rays have been useful : e.g. in the cases of 3eg j2006@xmath112321 = pmn2005@xmath112310 ( wallace et al . 2002 ) and 3eg j2016 + 3657 = b2013 + 370 ( halpern et al . new pulsar / isolated neutron star identifications were reported , e.g. 3eg j2227 + 6122 ( halpern et al . 2001b ) by discovery of the characteristic pulsar period of rx / ax j2229.0 + 6114 . x - ray observations were used to relate the high galactic latitude source 3eg j1835 + 5918 to an isolated neutron star , rx j1836.2 + 5925 ( reimer et al . 2001 ; mirabal & halpern 2001 ; halpern et al . 2002 ) . with the wealth of incoming discoveries from the parkes multi - beam pulsar survey , promising associations between newly discovered radio pulsars and egret sources have also been discussed . these associations include the two young pulsars psr j14206048 and psr j18370604 ( damico et al . 2001 ) in the vicinity of 3eg j14206038 and 3eg j18370606 , respectively ; psr j10165857 near the snr g284.31.8 is a plausible counterpart for 3eg j10135915 ( camilo et al . 2001 ) . in a recent survey of 56 unidentified egret sources roberts et al . ( 2004 ) found a radio pulsar located inside the 95% likelihood map in six of the investigated gamma - ray sources . the discovery of psr j2021 + 3651 in the error box of gev 2020 + 3658 using the 305 m arecibo radio telescope is another positive example ( roberts et al . 2002 ; 2004 ; hessels et al . however , torres , butt & camilo ( 2001 ) , and more recently kramer et al . ( 2003 ) , who have summarized the observational status of the radio pulsars and egret - detected gamma - ray sources concluded that , in many cases , further multi - frequency investigations are required in order to conclusively translate a proposed association into a final source identification . 3eg j2020 + 4017 is among the brightest persistent sources in the egret sky . originally listed as a cos - b source ( 2cg@xmath12 ) it is still unidentified . its gamma - ray flux is consistent with constant flux ( hartman et al . 1999 ) , and the spectrum is hard and best described by a power - law with photon - index of @xmath13 . merck et al . ( 1996 ) found evidence for a spectral break at @xmath14 gev which has been confirmed in recent studies by bertsch et al . ( 2000 ) and reimer & bertsch ( 2001 ) . examining all archival egret data and using photons @xmath15 gev , brazier et al . ( 1996 ) found a best position at @xmath16 , @xmath17 with a @xmath18 95%-confidence error box . this position was consistent with the 2eg catalog position and placed the egret source within the @xmath0-cygni supernova remnant g78.2 + 2.1 . the remnant g78.2 + 2.1 consists of a @xmath19-diameter , circular radio shell with two bright , broad opposing arcs on its rim ( higgs , landecker & roger 1977 ; wendker , higgs & landecker 1991 ) . g78.2 + 2.1 has a kinematic distance of @xmath20 kpc @xmath21 ( landecker , roger & higgs 1980 ; green 1989 ) and is estimated to have an age of 5400 yr ( sturner & dermer 1995 ) . a very bright star , @xmath0-cygni ( @xmath22 , spectral type f8iab ) lies on the eastern edge and lends its name to the remnant . a small hii region , located close to the star , forms the so - called @xmath0-cygni nebula . brazier et al . ( 1996 ) analyzed rosat pspc data viewing the @xmath0-cygni region . six pspc observations were targeted at celestial positions within 40 arcmin of the egret source . these rosat observations are combined and shown in figure [ f : rosat ] . the point source rx j2020.2 + 4026 is located within the 95% confidence contour of the 2eg position of 2cg078 + 2 ( brazier et al . 1996 ) , and was suggested by these authors to be the x - ray counterpart to the gamma - ray source . brazier et al . ( 1996 ) and carraminana et al . ( 2000 ) provided a possible optical counterpart for rx j2020.2 + 4026 . optical follow - up observations revealed a 14.5 magnitude k0v star nearby and within the @xmath23 rosat error circle . the x - ray to optical flux ratio of this star was found to be marginally consistent with that found for late - type stars ( stocke et al . 1991 ; fleming et al . 1995 ) , so that an association of rx j2020.2 + 4026 with the gamma - ray source could not be excluded ( brazier et al . 1996 ) . with the 3eg catalog ( hartman et al . 1999 ) , an improved position of 2cg078 + 2 became available : @xmath24 , @xmath25 , i.e. shifted in right ascension and declination by a few arc - minutes with respect to the 2eg position used by brazier et al . ( 1996 ) . with this improved position the proposed counterpart rx j2020.2 + 4026 is no longer located within the 95% contour of 3eg j2020 + 4017 . the 99% likelihood contour , however , still includes rx j2020.2 + 4026 ( figure [ f : rosat ] ) . in this paper we report on follow - up studies of rx j2020.2 + 4026 with _ chandra _ and the green bank radio telescope . the _ chandra _ observations were taken with the aim to determine the position and spectrum of rx j2020.2 + 4026 with high precision and to explore the possible connection with 3eg j2020 + 4017 . gbt observations at 820 mhz were made in order to search the egret error box of 3eg j2020 + 4017 for a young radio pulsar . our 30 ksec _ chandra _ observation ( obsid 3856 ) was taken on 2003 january 26 using three advanced ccd imaging spectrometer ( acis ) ccds ( s2,3,4 ) in the faint , timed - exposure mode with a frame time of 3.141 s. standard _ chandra _ x - ray center ( cxc ) processing ( v.6.8.0 ) has applied aspect corrections and compensated for spacecraft dither . level 2 event lists were used in our analyses . events in pulse invariant channels corresponding to 0.5 to 8.0 kev were selected for the purpose of finding sources on s2 and s4 ( after destreaking ) . the energy range 0.25 to 8.0 kev was utilized for source finding with s3 . due to uncertainties in the low energy response , data in the range 0.5 to 8.0 kev were used for spectral analysis . there were no instances of increased background . the center position used for the _ chandra _ pointing was that of the rosat source rxj2020.2 + 4026 , at @xmath26 and @xmath27 . the acis image overlaid with the 3eg likelihood contour lines of 3eg j2020 + 4017 is shown in figure [ f : chandra ] . the positions of rx j2020.2 + 4026 and 37 other x - ray sources detected by _ chandra _ are also indicated . rx j2020.2 + 4026 still appears to be the brightest x - ray source in the field . we used the same source finding techniques as described in swartz et al . ( 2003 ) with the circular - gaussian approximation to the point spread function , and a minimum signal - to - noise ratio ( s / n ) of 2.6 , expected to result in much less than 1 accidental detection in the field . the corresponding background - subtracted point source detection limit is @xmath2810 counts . nineteen sources were found on the s2 chip , 16 on s3 , and three on the s4 ccd . table [ gamcyg_source_table ] lists the 38 x - ray sources . to simplify the discussion and to show the ccd in which a source was detected ( the s2,4 and s3 chips have different sensitivities ) sources are denoted as sxyy with x=2,3,4 indicating the ccd and yy=01,02,03 indicating the ordering in ra . table [ gamcyg_source_table ] gives the source positions , the associated uncertainty in these positions , and the signal - to - noise ratio . the positional uncertainty listed in column 7 is given by @xmath29 where @xmath30 is the size of the circular gaussian that approximately matches the point - spread function ( psf ) at the source location , @xmath31 is the vignetting - corrected number of source counts , and @xmath32 represents the systematic error . the factor 1.51 sets the radius to that which encloses 68% of the circular gaussian . the table also lists potential counterparts in either the united states naval observatory catalog ( usno - b1.0 ; monet et al . 2003 ) , the two micron all sky survey ( 2mass ) , or the various rosat catalogs . with the exception of the target source , rosat sources were listed if their position was within @xmath33 of the _ chandra _ position . uncertainties in the plate scale imply a systematic uncertainty of @xmath34 , and , given the differences in the usno and 2mass and _ cxo _ positions , we have used @xmath35 as a conservative estimate for @xmath32 . the non - x - ray candidate counterparts were selected by searching a circular region centered on the x - ray position and whose radius was the 99%-confidence radius ( 3.03/1.51 times the positional uncertainty listed in column 7 ) continuing the assumption that the point - spread function is described by a circular gaussian . we do recognize , of course , that the assumption is not accurate far off - axis , however , this is partially compensated in that there is a good deal of conservatism built into the definition of the positional uncertainty and by the fact that there is very little impact on the position centroid . as a further check on our estimate of the systematic error in the correlation of the x - ray and optical positions a position - error weighted least squares fit was performed . right ascension , declination , and roll angle of the pointing position were taken as free parameters of this fit . the uncertainties in the optical positions were taken from the usno - b1.0 catalog . the uncertainty in the x - ray position was that discussed previously , including the systematic uncertainty @xmath36 . there are 26 optical candidates selecting only one of the two possible counterparts to s401 . the fit was excellent ( @xmath37 of 55 for 52 degrees of freedom ) , independent of which candidate was associated with s401 . the fitted quantities for right ascension , declination , and roll angle were @xmath38 , and @xmath39 , respectively . considering that we have ignored any possible systematic errors in the non - x - ray positions , such as those due to proper motion , we feel that applying this offset is unjustified . the browse feature was used to search for cataloged objects at or near the x - ray positions listed in table [ gamcyg_source_table ] . all available browse catalogs were selected to be interrogated and the 99%-confidence region around the x - ray sources were searched for possible counterparts . there are 3387 usno - b1.0 sources in a @xmath40 radius centered on the x - ray pointing direction . to the extent that these are uniformly distributed there are @xmath41 and this density was used to calculate the expected average number of accidental coincidences listed in column 4 of table [ t : counterparts ] . the probability of getting one or more matches by chance is given by the poisson probability @xmath42 which for small values is approximately @xmath43 . these probabilities are always below 7% and most ( 20 of 26 ) below 2% . the separation between the x - ray source and the optical source is listed in the 8th column of table [ gamcyg_source_table ] . the position of the candidate optical counterpart is also listed in table [ t : counterparts ] . there are two optical candidate counterparts for s401 . the position of the source s312 is found to match that of the optical k0v - star which has a usno - b1.0 position of @xmath44 and @xmath45 . there are 5061 2mass sources in the @xmath40 radius circle centered on the pointing position and the inferred density was used to calculate the probability of an accidental coincidence that is listed in column 7 of table [ t : counterparts ] . the probabilities are always below 12% and about half of them are below 2% . other pertinent information concerning the potential infrared counterparts is listed in tables [ gamcyg_source_table ] and [ t : counterparts ] . in table [ t : counterparts ] we have also listed , where relevant , the separation between the optical and the infrared candidate counterparts . in all cases but s209 , these separations are sub - arcsecond , implying , apart from the exception , that the optical and the infrared sources are the same . table [ t:2mass_colors ] shows the magnitudes and colors of the 2mass counterparts . figure [ f:2mass_colors ] shows the colors of all of the 2mass sources in a 12-arcmin radius circle centered on the pointing direction . with two possible exceptions , the inferred counterparts of the x - ray sources appear to be distributed as the field sources . those that do not appear to be reddened have colors of moderately late - type stars . this is not surprising since the @xmath0-cygni field is close to the galactic plane where approximately 90% of 2mass sources are stars . although most galactic - plane 2mass objects are normal stars , the majority of objects identified with x - ray sources need not be stars . for example , the x - ray emission may arise from a compact companion . the distribution of the colors of the x - ray selected subset , however , seems to reproduce the distribution of the field objects . the two most reddened sources are s208 and s309 which have j@xmath11k@xmath46 and thus may possibly be background sources ( agns ) absorbed by the galactic column . the position of the source s312 is found to match that of the optical k0v - star which has a 2mass position of @xmath44 and @xmath47 . point - source counts and spectra were extracted from within the radii listed in column 4 of table [ gamcyg_source_table ] . the background estimation was determined from creating data sets for each ccd after removing the events from each source region out to a radius 10 times the extraction radius listed in table [ gamcyg_source_table ] . only a few of the 38 detected sources have sufficient counts to warrant an attempt at an individual spectral analysis . in descending order of the number of detected counts , these are sources s312 ( 253 cts ) , s206 ( 213 cts ) , s219 ( 201 cts ) , s305 ( 176 cts ) , s214 ( 101 cts ) , and s204 ( 85 cts ) . all spectral analyses used ciao 3.0.2 to extract the pulse invariant ( pi ) files and cxc caldb 2.25 calibration files ( gain maps , quantum efficiency uniformity and effective area ) to generate the appropriate effective area and response functions . the spectral data were corrected for the effects of charge transfer inefficiency produced by proton damage to the front - illuminated ccds early in the mission . finally , we accounted for the impacts of molecular contamination on the acis filters with the number of days since launch set at 1275 . for the absorbing column we used tbabs in xspec ( v.11.2 ) with the default cross sections but with the abundances set to wilms , allen & mccray ( 2000 ) . the data were binned with no less than 10 counts per spectral bin . all given errors on spectral parameters are extremes on the two interesting parameters at the 68% confidence range . as noted previously , spectral analysis were restricted to the energy range @xmath48 kev because of the large uncertainties in the acis spectral response at low energies . three spectral models were applied . the first model , an absorbed power law , resulted in a statistically excellent fit ( @xmath37 of 16.7 for 15 degrees of freedom ) but with physically unreasonable parameters such as a very steep power law spectral index of 8.2 . the second model was the thermal , emission - line xspec model mekal . with @xmath49 ( solar metallicity ) , we obtained an acceptable fit ( @xmath50 for 15 degrees of freedom ) with @xmath51 , and @xmath52 kev . the 1@xmath30 confidence ranges for the column absorption and temperature are @xmath53 and @xmath54 kev , respectively . the third model , a blackbody with @xmath55 and @xmath56 kev , provides an alternative statistically acceptable representation of the x - ray spectrum ( table [ t : spectra ] ) . the spectrum and residuals of that fit are shown in figure [ f : spectrum_s312_bbody ] . from the hi in the galaxy ( dickey & lockman 1990 ) we compute the column absorption through the galaxy in the direction to @xmath0-cygni to be @xmath57 . this is comparable with the column absorption of sources in @xmath0-cygni ( maeda et al . 1999 ; uchiyama et al . 2002 ) and significantly higher than what is found from spectral fits of s312 , suggesting that this source is a foreground object . the low - column mekal spectrum together with the positional identification would appear to establish s312 ( rx j2020.2 + 4026 ) as the x - ray counterpart to the k0v - star , invalidating its association with the unidentified egret source 3eg j2020 + 4017 . this conclusion is bolstered by considering the upper limit to the x - ray luminosity of @xmath58 which we obtained from the x - ray flux of @xmath59 and the distance upper limit of @xmath60 pc . the latter was derived from the star s spectral type k0 ( brazier et al . 1996 ) and distance modulus . we note that the luminosity is slightly high suggesting the star is rotating `` rapidly '' which is mildly inconsistent with the optical spectra . as with s312 ( rx j2020.2 + 4026 ) we binned the data to assure at least 10 counts per spectral bin and fit the data for the other relatively bright sources to the absorbed power law , mekal and blackbody models . the results are summarized in table [ t : spectra ] . [ [ s206-s206 ] ] s206 [ s206 ] + + + + + + + + + + + about 213 source counts were detected from this source . the data clearly favor the power law spectrum with a photon - index in the 68% confidence range @xmath61 . [ [ s219-s219 ] ] s219 [ s219 ] + + + + + + + + + + + this source is detected with 201 source counts near to the edge of the s2 ccd and was seen by rosat ( 1rxh j202111 + 402807 ) . in contrast to s206 , a more complicated spectrum is called for . we tried to fit the data with a two - temperature mekal model ( not unusual for stars ) which did provide a good fit ( @xmath62 on 14 degrees of freedom ) . the two best - fit temperatures were 0.2 and 24 kev with the highly uncertain higher temperature component providing only 3% of the total flux . the best - fit absorbing column was @xmath63 . [ [ s305-s305 ] ] s305 [ s305 ] + + + + + + + + + + + similar to s312 , all three spectral models provided statistically adequate fits to the spectral data . in this case , however , it is not valid to argue that the power law index is unphysical . [ [ s214-s204-s214 ] ] s214 & s204 [ s214 ] + + + + + + + + + + + + + + + + + + none of the three model spectral fits are acceptable ( as with s219 ) , however , with only a total of seven and five bins of spectral data respectively , it would not be surprising to be able to fit these data with a more complicated model . both sources have both usno - b1.0 and 2mass counterparts . for the remaining 32 sources , all with fewer than 69 detected source counts , no spectral fitting was attempted . these sources were , however , included in the x - ray color - color diagrams presented in the following section . we show two x - ray color - color " diagrams in figures [ f : s3bi_colors ] and [ f : s2fi_colors ] . because of differences in the spectral responses , the data were separated between back- and front - illuminated ccds , the back - illuminated ccds being somewhat more sensitive to low energy x - rays . clearly , since most of these sources were detected with a small number of total counts , the uncertainties are large and it is difficult to draw any firm conclusions . thus the figures are primarily included for completeness . we note that the hardest sources ( s203 , s302 , s313 , s316 ) , those that occupy the upper - right portions of the diagrams , are among those with no usno and/or 2mass counterparts and these may be background sources ( agn ) absorbed by the galactic column . the very soft and unabsorbed sources ( s315 , s219 , s403 , s207 , s210 , s217 ) are likely to be associated with foreground stars and all have optical and infrared candidate counterparts . the infrared colors ( table [ t:2mass_colors ] ) of all of these objects , except for the candidate counterpart to s219 , are those expected for evolved main sequence ( primarily k ) stars . the counterpart to s219 would have to be a giant or supergiant . the paucity of detected counts for the x - ray sources limit the ability to draw many conclusions from the time series . in one case however , s206 , the source was quiescent for most of the observation and then suddenly flared as shown in figure [ f : s206flare ] . there is also , less compelling , evidence that s219 flared . the flaring nature of s206 and a 2mass counterpart are all consistent with coronal emission from a star . we have re - analyzed the archival rosat pspc data used by brazier et al . source detection algorithms ( box car as well as maximum likelihood ) found many sources , most of which appear to be associated with the diffuse emission of the remnant g78.2 + 2.1 as neither of these methods is ideally suited for searching for point sources embedded in extended sources ( i.e. patchy background ) . the x - ray source rx j2020.2 + 4026 seen by brazier et al . ( 1996 ) was the only point source detected . no other point source was found within the 3eg j2020 + 4017 error box . using these data we can thus set a @xmath64 count rate upper limit of @xmath65 pspc counts / s for a putative x - ray point source located within the region defined by the 68% confidence contour of 3eg j2020 + 4017 . assuming a power law spectrum with a photon index of two and a column absorption of @xmath66 ( dicky & lockman 1990 ) , the count rate upper limit corresponds to an energy flux upper limit of @xmath67 and @xmath68 , respectively . for a distance of 1.5 kpc this yields @xmath69 for the upper limit to the isotropic x - ray luminosity . in addition to the rosat pspc observations , there are four hri data sets in the rosat archive which partly cover the 3eg j2020 + 4017 error box . these data were taken between 1994 and 1997 with varying exposure times from 10 ksec to 36.5 ksec , respectively , and were not considered in the analysis of brazier et al . table [ hri ] list the observational details . a maximum - likelihood source - detection algorithm with a threshold of @xmath70 found six point sources in the data set 400899h , two in 202534h , one in 500339h and none in 202033h . the source properties are given in table [ hri_sources ] and the positions are shown in figure [ f : rosat ] . the region observed in the hri observation 400899h mostly overlaps our _ chandra _ observations . of the six sources detected in this hri observation , two new x - ray sources not seen by _ chandra _ are detected ; rx j202137.6 + 402959 and rx j202057.8 + 402829 . thus , these sources appear to be variable . the other four hri sources detected are s214 , s219 , s204 and s312 ( the last is the putative x - ray counterpart of 3eg j2020 + 4017 proposed by brazier et al . 1996 ) . among the other three newly detected sources , only rx j202150.5 + 401837 is located within the 95% likelihood region for the position of 3eg j2020 + 4017 . the other two sources are located far outside the 99% contour and can not account for an x - ray counterpart of 3eg j2020 + 4017 . a deep search for radio pulsations from 2cg078 + 2 was carried out using the gbt on 2003 december 27 . the observations were made at a center frequency of 820 mhz using an identical data acquisition and analysis scheme as that described by camilo et al . ( 2002 ) in their detection of 65-ms radio pulsations from the pulsar in snr 3c 58 . given the uncertain position of the putative pulsar , the observing time was divided into four separate pointings of the 15@xmath71 ( fwhm ) beam : one pointing was approximately centered on rx j2020.3 + 4026 , and the remainder were arranged so as to cover much of the egret error region likely to contain the @xmath0-ray source with 99% probability ( figure [ radio_pointings ] ) . due to time constraints , dwell times of 4 hr for three of the pointings , and 3 hr for the fourth , were used . the pointings are summarized in table [ radio_pointings_tabel ] . data were acquired with the berkeley caltech pulsar machine ( bcpm ) , an analog / digital filter bank ( backer et al . 1997 ) that divides the frequency band into 96 contiguous channels and samples the incoming voltages of the two orthogonal circular polarizations received by the telescope every 72 @xmath72s . for these observations , the channel bandwidth was set to 0.5 mhz so that a total band of 48 mhz was recorded . after summing the polarizations , the resulting total - power time samples were written to disk with 4-bit precision for subsequent off - line processing . known pulsars were successfully detected before and after the 2cg078 + 2 observations . the data analysis was carried out at jodrell bank using standard fourier - based pulsar search software routines ( for full details , see lorimer et al . 2000 ) which are freely availabledrl / seek ] . in order to reduce the volume of data , 12 ( 10 ) adjacent time samples were added together prior to dedispersion of the 4-hr ( 3-hr ) observations . the resulting decimated time series had effective sampling times of 864 and 720 @xmath72s . this choice of decimation was convenient for the periodicity search , which uses a base - two fourier transform algorithm . for a @xmath73-point fourier transform , the effective integration times were 3.9 hr and 3.2 hr . the latter data were zero padded . each observation was analyzed separately . the data were first dedispersed at 301 trial dispersion measure ( dm ) values between 0 and 300 @xmath74 pc . the expected dm from the ne2001 electron density model ( cordes & lazio 2002 ) for @xmath75 and @xmath76 is @xmath77 @xmath74 pc , assuming a distance of 1.5 kpc ( landecker , roger & higgs 1980 ) . the maximum dm in this direction is 350 @xmath74 pc ( cordes & lazio 2002 ) . the resulting time series were then fourier transformed and the amplitude spectra searched for significant features . to increase sensitivity to narrow duty - cycle pulses , individual spectra summing the first 2 , 4 , 8 and 16 harmonics were also searched . the resulting list of candidate signals above a s / n threshold of 6 were then folded in the time domain for visual inspection . no convincing pulsar - like signals were found . based on the known system parameters ( camilo et al . 2002 ) we estimate the sensitivity of our observations to be @xmath78 mjy at periods @xmath79 ms and dms consistent with the 1.5 kpc distance . with a typical duty cycle @xmath80 , this yields @xmath81jy . for a distance of 1.5 kpc the 820-mhz luminosity limit @xmath82 is @xmath83 . for an assumed radio spectral index of @xmath84 ( see e.g. lorimer et al . 1995 ) the corresponding luminosity at 1400 mhz is @xmath85 which is lower than most of the recent deep radio search limits ( see e.g. camilo 2003 ) . if there is a pulsar operating in 3eg j2020 + 4017 , our sensitivity limit suggests that the pulsar either does not produce significant amounts of radio emission or that its geometry is such that the radio beam does not intersect with the line of sight . we have searched a portion of the @xmath0-cygni field for possible x - ray counterparts to the intriguing gamma - ray source 3eg j2020 + 4017 ( 2cg078 + 2 ) using _ chandra _ and rosat . we have shown that a previous candidate , rx j2020.2 + 4026 , is almost certainly not the gamma - ray source but identified with a normal star . this conclusion is based on the refined position of the x - ray source , its spectrum and coincidence with both optical and infrared sources and the inferred x - ray luminosity . further , we have found a total of 38 x - ray sources in the _ chandra _ s2- , s3- and s4-fields which covers only part of the much larger error box containing the location of the egret source . a re - analysis of archival rosat hri data revealed three more x - ray sources within the egret error box which are not detected in the _ chandra _ observations . two of these sources are surely variable whereas the third source was found in a region not covered by the _ chandra _ observation . we found that some of the _ chandra _ sources have counterparts that may be main - sequence stars based on their identification with optical objects and 2mass sources of normal colors . of course the x - ray emission may not be due to the main sequence star , but can arise from an accreting compact companion . none of the x - ray sources appear to be radio pulsars , down to a limiting sensitivity of @xmath86 for an assumed pulse duty cycle of 4% . this limit also applies to the entire region associated with the 99%-confidence position contours of 3eg j2020 + 4017 . determining an upper limit for a putative x - ray point source located within the 68% confidence contour of 3eg j2020 + 4017 using archival rosat pspc data we found a @xmath64 luminosity upper limit of @xmath87 which is four times smaller than the rosat pspc - deduced luminosity observed from the vela pulsar ( @xmath88 , e.g. table 3 in becker & aschenbach 2002 ) but about four times higher than the total rosat observed x - ray luminosity from geminga ( @xmath89 ) . we therefore consider it as a valid option that the counterpart of 3eg j2020 + 4017 is a neutron star with an x - ray luminosity similar to that observed from vela - like to middle - aged pulsars . an object with such luminosity would not have been detected in the x - ray data from rosat which cover that region of the sky . adopting the snr age of 5400 yrs and assuming a spin - down to x - ray energy conversion factor of @xmath3 ( becker & trmper 1997 ) we are able to constrain the spin - parameters of such a putative neutron star to be @xmath90 , @xmath91 and @xmath92 g , which are consistent with the properties of known vela- to middle - aged pulsars ( e.g. gonzalez & safi - harb 2003 , becker & pavlov 2001 ) , given the uncertainty of this approach . the ratio of the @xmath0-ray to soft x - ray flux deduced from our upper limit , @xmath93 , is consistent with this conclusion . in order to obtain a full census of the x - ray population in the error box of 3eg j2020 + 4017 further observations with _ chandra _ are required . as the discovery of geminga has taught us , deep follow - up optical observations of new x - ray sources can also lead to the identification of the nature of a high - energy source . finally , the measurements we expect from the _ glast _ mission ( launch 2007 ) should provide a much improved signal - to - noise ratio and a source location better than @xmath94 for this gamma - ray source . this will open the possibility to directly search for pulsar periodicities in the gamma - ray data . in case no pulsar is found , the restricted number of _ chandra _ sources in the _ glast _ error box will then be prime candidates for even deeper searches for counterparts . those of us at the marshall space flight center acknowledge support from the _ chandra _ project . mcw acknowledges with gratitude conversations with marshall joy and roc cutri that clarified some of the mysteries of the infrared portion of the spectrum . za was supported by nasa grant nra-99 - 01-ltsa-070 . drl is a university research fellow funded by the royal society . fc is supported in part by nsf grant ast-02 - 05853 . this publication makes 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 the national aeronautics and space administration and the national science foundation . in addition , this research has made use of data obtained from the high energy astrophysics science archive research center ( heasarc ) , provided by nasa s goddard space flight center . backer , d.c . , dexter , m.r . , zepka , a. , ng , d. , werthimer , d.j . , ray , p.s . , foster , r.s . , 1997 , pasp , 109 , 61 becker , w. , aschenbach , b. , 2002 , in proceedings of the we - heraeus seminar on neutron stars , pulsars and supernova remnants , eds . w.becker , h.lesch & j / trmper , mpe - report 278 , 64 , ( available from astro - ph/0208466 ) camilo , f. , stairs , i.h . , lorimer , d.r . , backer , d.c . , ransom , s.m . , klein , b. , wielebinski , r. , kramer , m. , mclaughlin , m.a . , arzoumanian , z. , mller , p. , 2002 , apj , 571 , l41 camilo , f. , 2003 , in _ radio pulsars _ , eds m.bailes , d.j.nice , and s.e.thorsett , astronomical society of the pacific , san francisco carraminana , a. , chavushyan , v. , zharikov , s. , et al . , 2000 , in _ proc . 5th compton symposium _ , aip - cp 510 , p49 gil , j.a . , khechinashvili , d.g . , melikidze , g.i . , 1998 , mnras , 298 , 1207 gonzalez , m. , safi - harb , s. , 2003 , apj , 591 , 143 halpern , j.p . , gotthelf , e.v . , mirabal , n. , camilo , f. , 2002 , apj , 573 , l41 halpern , j.p . , eracleous , m. , mukherjee , r. , et al . 2001a , apj 551 , 1016 hessels , j.w.t . , roberts , m.s.e . , ransom , s.m . , kaspi , v.m . , romani , r.w . , ng , c.y . , freier , p.c.c . , gaensler , b.m . , 2004 , astro - ph/0403632 higgs , l.a . , landecker , t.l . , and roger , r.s . , 1977 , aj , 82 , 718 lorimer , d.r . , kramer , m. , mller , p. , wex , n. , jessner , a. , lange , c. , wielebinski , r. , 2000 , a&a , 358 , 169 lorimer , d.r . , yates , j.a . , lyne , a.g . , gould , d.m . , 1995 , mnras , 273 , 411l maeda , y. , koyama , k. , yokogawa , j. , skinner , s. , 1999 , apj 510 , 967 mayer - hasselwander , h.a . , simpson , g. , 1990 , in _ the egret science symposium _ , eds c.fichtel , s.huntre , p.sreekumar , f.stecker , nasa , vp-3071 , p153 & & & & & & & & & + & & & ( @xmath95 ) & & & ( @xmath95 ) & ( @xmath95 ) & ( @xmath95 ) & ( @xmath95 ) + s401 & 304.86758 & 40.436279 & 13.6 & 49 & 5.7 & 1.25 & 0.61&1.55 & & + s402 & 304.88129 & 40.373795 & 13.8 & 25 & 3.9 & 1.72 & & 1.65 & + s403 & 304.92621 & 40.362309 & 10.1 & 69 & 6.7 & 0.87 & 1.05 & 1.13 & + s301 & 304.97339 & 40.455490 & 4.54 & 13 & 3.0 & 0.88 & 0.70 & 0.63 & + s302 & 304.97769 & 40.468494 & 4.61 & 36 & 5.2 & 0.65 & & & + s303 & 304.99615 & 40.398434 & 3.75 & 14 & 3.1 & 0.76 & 0.85 & 1.29 & + s304 & 305.00381 & 40.365112 & 5.03 & 14 & 3.0 & 0.94 & 1.15 & 1.41 & + s305 & 305.00769 & 40.434021 & 2.63 & 176 & 11.2 & 0.47 & & 0.56 & + s306 & 305.01837 & 40.457535 & 2.47 & 23 & 4.3 & 0.55 & & & + s307 & 305.02713 & 40.412449 & 2.19 & 25 & 4.5 & 0.52 & 0.37 & 0.39 & + s308 & 305.02737 & 40.374905 & 3.60 & 28 & 4.8 & 0.61 & 0.25 & 0.36 & + s309 & 305.05789 & 40.394272 & 2.04 & 15 & 3.3 & 0.55 & & 0.17 & + s310 & 305.06354 & 40.486912 & 2.46 & 12 & 3.1 & 0.63 & 0.95 & 1.05 & + s311 & 305.07104 & 40.446281 & 1.21 & 44 & 5.8 & 0.47 & 0.30 & 0.25 & + s312 & 305.07147 & 40.437424 & 1.14 & 253 & 13.7 & 0.46 & 0.27 & 0.30 & 5.9@xmath96 + s313 & 305.08884 & 40.448517 & 1.21 & 20 & 3.7 & 0.48 & & & + s314 & 305.09262 & 40.490631 & 2.57 & 22 & 4.0 & 0.56 & & 0.13 & + s315 & 305.10596 & 40.484375 & 2.37 & 12 & 3.0 & 0.61 & 0.65 & 0.41 & + s316 & 305.11560 & 40.440331 & 1.40 & 11 & 3.1 & 0.52 & & & + s201 & 305.14136 & 40.473434 & 2.68 & 10 & 2.9 & 0.68 & 0.34 & 0.62 & + s202 & 305.15140 & 40.494373 & 3.97 & 17 & 3.5 & 0.73 & 0.80 & 0.64 & + s203 & 305.15570 & 40.495060 & 4.18 & 11 & 2.8 & 0.88 & & & + s204 & 305.17081 & 40.451115 & 3.26 & 85 & 7.9 & 0.50 & 0.22 & 0.35 & 1.2@xmath97 + s205 & 305.17349 & 40.380219 & 4.65 & 29 & 4.7 & 0.69 & 0.73 & & + s206 & 305.18216 & 40.450073 & 3.82 & 213 & 12.4 & 0.48 & & 0.31 & + s207 & 305.18600 & 40.430450 & 3.93 & 10 & 2.9 & 0.87 & 0.78 & 1.03 & + s208 & 305.21335 & 40.403984 & 6.05 & 16 & 3.4 & 1.02 & & 0.84 & + s209 & 305.21414 & 40.509041 & 8.20 & 28 & 4.2 & 1.04 & 1.73 & 1.66 & + s210 & 305.21881 & 40.473904 & 6.70 & 24 & 3.8 & 0.95 & 0.55 & 0.55 & + s211 & 305.21906 & 40.408849 & 6.35 & 66 & 7.0 & 0.65 & 0.43 & 0.62 & + s212 & 305.22256 & 40.507294 & 8.70 & 47 & 5.9 & 0.89 & 1.54 & 1.61 & + s213 & 305.23041 & 40.474827 & 7.60 & 33 & 4.2 & 0.95 & 0.91 & 0.63 & + s214 & 305.24078 & 40.474091 & 8.44 & 101 & 8.2 & 0.68 & 0.60 & 0.40 & 2.9@xmath98 + s215 & 305.24927 & 40.378075 & 9.95 & 17 & 3.1 & 1.52 & 0.91 & 0.95 & + s216 & 305.25110 & 40.485874 & 9.80 & 33 & 4.3 & 1.12 & 1.63 & 1.10 & + s217 & 305.27808 & 40.430069 & 11.2 & 24 & 3.8 & 1.47 & 0.97 & 0.97 & + s218 & 305.29422 & 40.513210 & 15.8 & 35 & 4.0 & 1.66 & & 0.57 & + s219 & 305.29767 & 40.468163 & 13.9 & 201 & 11.5 & 0.74 & 0.38 & 0.70 & 2.4@xmath99 + & & & & & & & + & & & & & & & ( @xmath95 ) + & & & & & & & + s401 & 304.867803 & 40.436275 & 0.041 & & & & + s401 & 304.868070 & 40.436495 & 0.041 & & & & + s402 & & & & 304.881105 & 40.373360 & 0.117 & + s403 & 304.926575 & 40.362225 & 0.020 & 304.926589 & 40.362183 & 0.030 & 0.16 + s301 & 304.973164 & 40.455400 & 0.021 & 304.973295 & 40.455330 & 0.031 & 0.44 + s303 & 304.996045 & 40.398212 & 0.015 & 304.995913 & 40.398125 & 0.023 & 0.48 + s304 & 305.003723 & 40.364800 & 0.023 & 305.003740 & 40.364723 & 0.035 & 0.28 + s305 & & & & 305.007841 & 40.434124 & 0.009 & + s307 & 305.027000 & 40.412425 & 0.007 & 305.027043 & 40.412365 & 0.011 & 0.25 + s308 & 305.027320 & 40.374964 & 0.010 & 305.027281 & 40.374832 & 0.015 & 0.49 + s309 & & & & 305.057841 & 40.394245 & 0.012 & + s310 & 305.063206 & 40.486845 & 0.010 & 305.063218 & 40.486755 & 0.015 & 0.33 + s311 & 305.070998 & 40.446359 & 0.006 & 305.071003 & 40.446217 & 0.009 & 0.51 + s312 & 305.071387 & 40.437464 & 0.005 & 305.071389 & 40.437366 & 0.008 & 0.35 + s314 & & & & 305.092591 & 40.490601 & 0.012 & + s315 & 305.105898 & 40.484550 & 0.010 & 305.105912 & 40.484482 & 0.015 & 0.25 + s201 & 305.141242 & 40.473467 & 0.012 & 305.141188 & 40.473324 & 0.018 & 0.54 + s202 & 305.151131 & 40.494459 & 0.014 & 305.151193 & 40.494289 & 0.021 & 0.64 + s204 & 305.170775 & 40.451170 & 0.007 & 305.170747 & 40.451031 & 0.010 & 0.51 + s205 & 305.173353 & 40.380392 & 0.013 & & & & + s206 & & & & 305.182074 & 40.450016 & 0.009 & + s207 & 305.186253 & 40.430348 & 0.020 & 305.186255 & 40.430241 & 0.030 & 0.39 + s208 & & & & 305.213189 & 40.403786 & 0.041 & + s209 & 305.214453 & 40.508625 & 0.028 & 305.213771 & 40.508675 & 0.042 & 1.88 + s210 & 305.218609 & 40.473920 & 0.024 & 305.218624 & 40.473850 & 0.035 & 0.26 + s211 & 305.219192 & 40.408425 & 0.011 & 305.219198 & 40.408352 & 0.017 & 0.26 + s212 & 305.223092 & 40.507159 & 0.021 & 305.223083 & 40.507088 & 0.031 & 0.26 + s213 & 305.230259 & 40.475053 & 0.024 & 305.230230 & 40.474934 & 0.036 & 0.44 + s214 & 305.240995 & 40.474125 & 0.012 & 305.240870 & 40.474003 & 0.018 & 0.56 + s215 & 305.249534 & 40.378228 & 0.061 & 305.249506 & 40.378269 & 0.090 & 0.17 + s216 & 305.250842 & 40.486821 & 0.033 & 305.250918 & 40.486145 & 0.049 & 0.53 + s217 & 305.278420 & 40.429992 & 0.057 & 305.278386 & 40.429932 & 0.085 & 0.24 + s218 & & & & 305.294173 & 40.513363 & 0.109 & + s219 & 305.297639 & 40.468059 & 0.015 & 305.297601 & 40.467976 & 0.022 & 0.32 + ccccccccccccccc source & j & @xmath30 & h & @xmath30 & k@xmath100 & @xmath30 & j@xmath11h & @xmath30 & h@xmath11k@xmath100 & @xmath30 & j@xmath11k@xmath100 & @xmath30 + + s201 & 14.821 & 0.038 & 14.136 & 0.053 & 13.706 & 0.061 & 0.685 & 0.065 & 0.430 & 0.081 & 1.115 & 0.072 + s202 & 14.458 & 0.038 & 13.702 & 0.037 & 13.485 & 0.05 & 0.756 & 0.053 & 0.217 & 0.062 & 0.973 & 0.063 + s204 & 13.243 & 0.029 & 12.566 & 0.032 & 12.359 & 0.036 & 0.677 & 0.043 & 0.207 & 0.048 & 0.884 & 0.046 + s206 & 14.516 & 0.051 & 13.921 & 0.051 & 13.796 & 0.069 & 0.595 & 0.072 & 0.125 & 0.086 & 0.72 & 0.086 + s207 & 15.235 & 0.052 & 14.678 & 0.070 & 14.58 & 0.12 & 0.557 & 0.087 & 0.098 & 0.139 & 0.655 & 0.131 + s208 & 17.597@xmath101 & & 15.246 & 0.133 & 14.614 & 0.12 & @xmath102 & & 0.632 & 0.179 & @xmath103 & + s209 & 14.281 & 0.052 & 13.232 & 0.068 & 12.725 & 0.063 & 1.049 & 0.086 & 0.507 & 0.093 & 1.556 & 0.082 + s210 & 14.648 & 0.038 & 13.685 & 0.039 & 13.249 & 0.043 & 0.963 & 0.054 & 0.436 & 0.058 & 1.399 & 0.057 + s211 & 13.503 & 0.026 & 12.695 & 0.021 & 12.338 & 0.03 & 0.808 & 0.033 & 0.357 & 0.037 & 1.165 & 0.040 + s212 & 11.468 & 0.022 & 11.075 & 0.020 & 10.9 & 0.023 & 0.393 & 0.030 & 0.175 & 0.030 & 0.568 & 0.032 + s213 & 14.99 & 0.045 & 14.151 & 0.052 & 13.802 & 0.063 & 0.839 & 0.069 & 0.349 & 0.082 & 1.188 & 0.077 + s214 & 12.674@xmath101 & & 11.851 & 0.035 & 11.606 & 0.039 & @xmath104 & & 0.245 & 0.052 & @xmath105 & + s215 & 14.361 & 0.036 & 13.29 & 0.036 & 11.96@xmath101 & & 1.071 & 0.051 & @xmath106 & & @xmath107 & + s216 & 13.736 & 0.028 & 12.929 & 0.031 & 12.615 & 0.034 & 0.807 & 0.042 & 0.314 & 0.046 & 1.121 & 0.044 + s217 & 11.733 & 0.021 & 11.497 & 0.018 & 11.463 & 0.018 & 0.236 & 0.028 & 0.034 & 0.025 & 0.27 & 0.028 + s218 & 14.9@xmath101 & & 15.591 & 0.136 & 13.743@xmath101 & & @xmath108 & & @xmath109 & & n / a & + s219 & 11.927 & 0.021 & 11.252 & 0.018 & 11.155 & 0.017 & 0.675 & 0.028 & 0.097 & 0.025 & 0.772 & 0.027 + s301 & 14.915 & 0.039 & 13.742 & 0.037 & 13.337 & 0.042 & 1.173 & 0.054 & 0.405 & 0.056 & 1.578 & 0.057 + s303 & 15.776 & 0.069 & 14.678 & 0.064 & 14.283 & 0.091 & 1.098 & 0.094 & 0.395 & 0.111 & 1.493 & 0.114 + s304 & 15.764 & 0.068 & 14.914 & 0.076 & 14.479 & 0.103 & 1.750 & 0.102 & -0.465 & 0.128 & 1.285 & 0.123 + s305 & 11.185 & 0.023 & 10.783 & 0.018 & 10.673 & 0.016 & 0.402 & 0.029 & 0.11 & 0.024 & 0.512 & 0.028 + s307 & 14.559 & 0.033 & 13.225 & 0.026 & 12.636 & 0.026 & 1.334 & 0.042 & 0.589 & 0.037 & 1.923 & 0.042 + s308 & 14.29 & 0.039 & 13.375 & 0.063 & 13.007 & 0.036 & 0.915 & 0.074 & 0.368 & 0.073 & 1.283 & 0.053 + s309 & 17.863@xmath101 & & 15.472 & 0.118 & 14.683 & 0.125 & @xmath102 & 0.118 & 0.789 & 0.172 & @xmath110 & + s310 & 15.37 & 0.052 & 14.331 & 0.049 & 14.127 & 0.076 & 1.039 & 0.071 & 0.204 & 0.090 & 1.243 & 0.092 + s311 & 15.181 & 0.054 & 14.246 & 0.054 & 13.957 & 0.067 & 0.935 & 0.076 & 0.289 & 0.086 & 1.224 & 0.086 + s312 & 12.373 & 0.022 & 11.83 & 0.021 & 11.686 & 0.017 & 0.543 & 0.030 & 0.144 & 0.027 & 0.687 & 0.028 + s314 & 16.492 & 0.135 & 15.4 & 0.107 & 14.906 & 0.158 & 1.092 & 0.172 & 0.494 & 0.191 & 1.586 & 0.208 + s315 & 15.365 & 0.052 & 14.738 & 0.062 & 14.38 & 0.101 & 0.627 & 0.081 & 0.358 & 0.119 & 0.985 & 0.114 + s402 & 14.456 & 0.038 & 13.702 & 0.035 & 13.477 & 0.045 & 0.754 & 0.052 & 0.225 & 0.057 & 0.979 & 0.059 + s403 & 13.262 & 0.023 & 12.749 & 0.024 & 12.634 & 0.028 & 0.513 & 0.033 & 0.115 & 0.037 & 0.628 & 0.036 + cccccc source & model@xmath101 & @xmath37 & @xmath111 & @xmath112 & @xmath113 or kt ( kev ) + & & & & @xmath114 & + + s312 & pl & 16.6 & 15 & 1.07 ( 0.80@xmath1151.45 ) & 8.2 ( 6.8@xmath11510.0 ) + s312 & mekal & 22.3 & 15 & 0.00 ( 0.00@xmath1150.05 ) & 0.77 ( 0.68@xmath1150.83 ) + s312 & bb & 14.6 & 15 & 0.37 ( 0.22@xmath1150.61 ) & 0.14 ( 0.12@xmath1150.15 ) + s206 & pl & 18.7 & 16 & 0.01 ( 0.00@xmath1150.15 ) & 1.92 ( 1.67@xmath1152.36 ) + s206 & mekal & 43.4 & 16 & 1.2 & 1.0 + s206 & bb & 32.8 & 16 & 0.0 & 0.5 + s219 & pl & 29.8 & 16 & 0.5 & 5.4 + s219 & mekal & 49.2 & 16 & 0.0 & 1.0 + s219 & bb & 33.9 & 16 & 0.0 & 0.2 + s305 & pl & 18.2 & 12 & 0.48 ( 0.16@xmath1150.82 ) & 2.98 ( 2.51@xmath116 3.89 ) + s305 & mekal & 17.8 & 12 & 1.28 ( 1.07@xmath1151.51 ) & 0.70 ( 0.59@xmath117 0.82 ) + s305 & bb & 21.7 & 12 & 0.0 ( 0.0@xmath1180.20 ) & 0.39 ( 0.32@xmath117 0.44 ) + s214 & pl & 12.3 & 6 & 0.2 & 3.2 + s214 & mekal & 9.7 & 6 & 0.9 & 0.9 + s214 & bb & 14.6 & 6 & 0.0 & 0.3 + s204 & pl & 7.75 & 4 & 0.0 & 2.6 + s204 & mekal & 11.6 & 4 & 1.0 & 1.0 + s204 & bb & 12.0 & 4 & 0.0 & 0.4 + . ccccc seq . & start date & exposure & r.a . ( j2000 ) & dec . ( j2000 ) + number & ( ymd ) & ( sec ) & ( hms ) & ( dms ) + + 202033h & 1994 11 15 & 10536 & 20 20 28.08 & 41 21 36 + 202534h & 1997 06 06 & 10370 & 20 22 14.04 & 40 15 36 + 400899h & 1996 11 02 & 36552 & 20 21 04.08 & 40 26 24 + 500339h & 1994 11 14 & 18761 & 20 19 48.00 & 40 03 00 + cccccc data & source & r.a . ( j2000 ) & dec . ( j2000 ) & hri rate & _ chandra _ + & & & & cts / s @xmath119 & + + 400899h & rx j202137.6 + 402959 & 305.407035 & 40.499979 & @xmath120 & + 400899h & rx j202057.8 + 402829 & 305.240996 & 40.474830 & @xmath121 & s214 + 400899h & rx j202111.3 + 402806 & 305.297479 & 40.468571 & @xmath122 & s219 + 400899h & rx j202040.8 + 402704 & 305.170322 & 40.451285 & @xmath123 & s204 + 400899h & rx j202130.5 + 402649 & 305.377281 & 40.447124 & @xmath124 & + 400899h & rx j202016.8 + 402614 & 305.070074 & 40.437480 & @xmath125 & s312 + 202534h & rx j202240.0 + 401900 & 305.666926 & 40.316753 & @xmath126 & + 202534h & rx j202150.5 + 401837 & 305.460420 & 40.310447 & @xmath127 & + 500339h & rx j201950.8 + 395752 & 304.961886 & 39.964551 & @xmath128 & +

this paper is intended to provide an introduction to the theory of substitution tilings . for our purposes , tiling substitution rules are divided into two broad classes : geometric and combinatorial . geometric substitution tilings include self - similar tilings such as the well - known penrose tilings ; for this class there is a substantial body of research in the literature . combinatorial substitutions are just beginning to be examined , and some of what we present here is new . we give numerous examples , mention selected major results , discuss connections between the two classes of substitutions , include current research perspectives and questions , and provide an extensive bibliography . although the author attempts to fairly represent the as a whole , the paper is not an exhaustive survey , and she apologizes for any important omissions .

0705.1142

a _ tiling substitution rule _ is a rule that can be used to construct infinite tilings of @xmath0 using a finite number of tile types . the rule tells us how to substitute " each tile type by a finite configuration of tiles in a way that can be repeated , growing ever larger pieces of tiling at each stage . in the limit , an infinite tiling of @xmath0 is obtained . in this paper we take the perspective that there are two major classes of tiling substitution rules : those based on a linear expansion map and those relying instead upon a sort of concatenation " of tiles . the first class , which we call _ geometric tiling substitutions _ , includes _ self - similar tilings _ , of which there are several well - known examples including the penrose tilings . in this class a tile is substituted by a configuration of tiles that is a linear expansion of itself , and this geometric rigidity has permitted quite a bit of research to be done . we will note some of the fundamental results , directing the reader to appropriate references for more detail . the second class , which we call _ combinatorial tiling substitutions _ , is sufficiently new that it lacks even an agreed - upon definition . in this class the substitution rule replaces a tile by some configuration of tiles that may not bear any geometric resemblance to the original . the difficulty with such a rule comes when one wishes to iterate it : we need to be sure that the substitution can be applied repeatedly so that all the tiles fit together without gaps or overlaps . the examples we provide are much less well - known ( in some cases new ) and are ripe for further study . the two classes are related in a subtle and interesting way that is not yet well understood . the study of aperiodic tilings in general , and substitution tilings specifically , comes from the confluence of several discoveries and lines of research . interest in the subject from a philosophical viewpoint came to the forefront when wang @xcite asked about the decidability of the tiling problem " : whether a given set of prototiles can form an infinite tiling of the plane . he tied this answer to the existence of aperiodic prototile sets " : finite sets of tiles that can tile the plane , but only nonperiodically . he saw that the problem is undecidable if an aperiodic prototile set exists . berger @xcite was the first to find an aperiodic prototile set and was followed by many others , including penrose . it turned out that one way prove a prototile set is aperiodic involves showing that every tiling formed by the prototile set is self - similar . independently , work was proceeding on one - dimensional symbolic substitution systems , a combination of dynamical systems and theoretical computer science . symbolic dynamical systems had become of interest due to their utility in coding more complex dynamical systems , and great progress was being made in our understanding of these systems . queffelec @xcite summarized what was known about the ergodic and spectral theory of substitution systems , while a more recent survey of the state of the art appears in @xcite . substitution tilings can be seen as a natural extension of this branch of dynamical systems ; insight and proof techniques can often be borrowed for use in the tiling situation . we will use symbolic substitutions motivate our study in the next section . from the world of physics , a major breakthrough was made in 1984 by schechtman et . @xcite with the discovery of a metal alloy that , by rights , should have crystalline since its x - ray spectrum was diffractive . however , the diffraction pattern had five - fold rotational symmetry , which is not allowed for ideal crystals ! this type of matter has been termed quasicrystalline " , and self - similar tilings like the penrose tiling , having the right combination of aperiodicity and long - range order , were immediately recognized as valid mathematical models . dynamical systems entered the picture , and it was realized that the spectrum of a tiling dynamical system is closely related to the diffraction spectrum of the solid it models @xcite . thus we find several points of departure for the study of substitution tilings and their dynamical systems . let @xmath1 be a finite set called an _ alphabet _ , whose elements are called _ letters_. then @xmath2 , the set of all finite _ words _ with elements from @xmath1 , forms a semigroup under concatenation . a _ symbolic substitution _ is any map @xmath3 . a symbolic substitution can be applied to words in @xmath2 by concatenating the substitutions of the individual letters . a block of the form @xmath4 will be called a _ level-@xmath5 block of type @xmath6_. [ const1d ] let @xmath7 and let @xmath8 and @xmath9 . beginning with the letter @xmath6 we get @xmath10 where we ve added spaces to emphasize the breaks between substituted blocks . notice that the block lengths triple when substituted . [ nonconst1d ] again let @xmath11 ; this time let @xmath12 and @xmath13 . if we begin with @xmath6 we get : @xmath14 note that in this example block lengths are 1 , 2 , 3 , 5 , 8 , 13 , ... , and the reader can verify that they will continue growing as fibonacci numbers . these examples illustrate the major distinction we make between substitutions . in the first example , the length of a substituted letter is always 3 and thus the size of any level-@xmath5 block must be @xmath15 ; this is a _ substitution of constant length_. in the second example the size of a substituted letter depends on the letter itself , and the size of a level-@xmath5 block is computed recursively ; this is a _ substitution of non - constant length_. this is the essence of the distinction between geometric and combinatorial tiling substitutions . it is interesting to consider infinite sequences of the form @xmath16 in @xmath17 . such a sequence is said to be _ admitted _ by the substitution if every finite block of letters is contained in some level-@xmath5 block . in the theory of dynamical systems , the space of all sequences admitted by the substitution is studied using the shift action @xmath18 ( basically , moving the decimal point one unit to the right ) . an interested reader should see @xcite to find out more . the most straightforward generalization to tilings of @xmath19 ( or @xmath0 ) is to use unit square tiles labeled ( colored ) by the alphabet @xmath1 . these tilings can be considered as sequences in @xmath20 , and substitutions can take letters to square or rectangular blocks of letters . we only need to ensure that all of the blocks fit " to form a sequence without gaps or overlaps . the constant length case is to expand each colored tile by some integer @xmath21 and then subdivide into @xmath22 ( or @xmath23 ) colored unit squares . a simple method for the non - constant length case is to take the direct product of one - dimensional substitutions of non - constant length . [ const2d ] let @xmath24 , where we represent @xmath25 as a white unit square tile and @xmath26 as a blue unit square tile . suppose the length expansion is 3 and that the tiles are substituted by a three - by - three array of tiles , colored as in figure [ square1 ] . starting with the blue level-0 tile , level-0 , level-1 , level-2 and level-3 tiles are shown in figure [ square2 ] . one sees in this example the tiling version of the rule creating the sierpinski carpet . [ nonconst2d ] this time , let the alphabet be @xmath27 ; for simplicity of notation we put @xmath28 . the direct product of the fibonacci substitution of example [ nonconst1d ] with itself is shown in figure [ dp1 ] . using only colors without the numbers we show the level-0 through level-4 blocks of type @xmath25 in figure [ dp2 ] . the characteristic plaid " appearance of the direct product is evident . some literature on @xmath29-dimensional symbolic substitutions exists . in the non - constant length case , direct product substitutions , with a generalization allowing randomness in the choice of substitution from level to level , are studied in @xcite . an extension of this idea , allowing substitutions with restrictions forcing the substitutions to fit " , are studied in @xcite . in the constant - length case , a partial survey and spectral analysis of this class from the dynamical systems viewpoint appears in @xcite . for those wishing to experiment with various substitutions of both constant and non - constant length , the author maintains a matlab freeware computer program that allows the user to generate these tilings of @xmath20 and manipulate them in several ways @xcite . let us introduce some terminology that will be useful throughout the paper . tile _ is a set @xmath30 that is the closure of its interior . we will always assume that tiles are bounded ; in the literature it is frequently assumed that tiles are connected or even homeomorphic to topological balls . in fact it is often required that the tiles be polygonal , but in substitution tiling theory tiles with fractal boundary occur naturally . when it is desirable to distinguish between congruent tiles they can be _ labeled _ ( also called _ marked _ or _ colored _ ) . two tiles are considered _ equivalent _ if they differ by a rigid motion and carry the same label . prototile set _ is a finite set @xmath31 of inequivalent tiles . given a prototile set @xmath31 , a _ tiling _ of @xmath0 is a set @xmath32 of tiles , each equivalent to a tile from @xmath31 , such that 1 . @xmath32 covers @xmath0 : @xmath33 , and 2 . @xmath32 packs @xmath0 : distinct tiles have non - intersecting interiors . a _ @xmath32-patch _ is a finite union of tiles with nonintersecting interiors covering a connected set ; two patches are equivalent if there is a rigid motion between them that matches up equivalent tiles . a tiling is said to be of _ finite local complexity ( flc ) _ ( also known as having a _ finite number of local patterns _ ) if there are only finitely many two - tile @xmath32-patches up to equivalence . a tiling is called _ repetitive _ ( also called _ almost periodic _ or the _ local isomorphism property _ ) if for any @xmath32-patch @xmath34 there is an @xmath35 such that in every ball of radius @xmath36 there is a patch equivalent to @xmath34 . in dynamical systems theory the most work has been done on repetitive tilings with finite local complexity . given a tiling substitution , it is possible to construct infinite tilings and tiling spaces from that substitution in a few different ways . ( this is also true for symbolic substitutions ) . our description will be necessarily imprecise as different substitutions can require different definitions of some of the terms ; we give the main ideas here and refer the reader to sources such as @xcite , and @xcite to get more details . one way to get an infinite tiling is to begin with some initial block or tile and substitute ad infinitum . in many cases a limiting sequence or tiling @xmath37 will exist . sometimes it will cover only a half - line , quarter - plane , or some other unbounded region of space , and sometimes it will cover the entire line or plane . a less constructive method is to define a tiling @xmath32 as _ admitted _ by the substitution if every finite configuration of tiles in @xmath32 is equivalent to a configuration found inside a level-@xmath5 tile , for some @xmath5 . the _ tiling space _ associated to a substitution is the set of all tilings admitted by that substitution . another way to obtain this space is to take the closure ( in a suitable metric ) of all rigid motions of a limiting tiling @xmath37 . in either case , a point in the tiling space @xmath38 is an infinite tiling , and any nontrivial rigid motion of that tiling is considered a different point in the tiling space . substitutions of constant length have a natural generalization to tilings in higher dimensions , which we introduce in section [ geom.section ] . these generalizations , which include the well - studied self - similar tilings , rely upon the use of linear expansion maps and are therefore rigidly geometric . we present examples in varying degrees of generality and include a selection of the major results in the field . extending substitutions of non - constant length to higher dimensions seems to be more difficult , and is the topic of section [ comb.section ] . to even define what this class contains has been problematic and there is not yet a consensus on the subject . for lack of existing terminology we have decided to call this type of substitution combinatorial as tiles are combined to create the substitutions without any geometric restriction save that they can be iterated without gaps or overlaps , and because in certain cases it is possible to define them in terms of their graph - theoretic structure . in many cases one can transform combinatorial tiling substitutions into geometric ones through a limit process . in section [ connections.section ] , we will discuss how to do this and what the effects are to the extent that they are known . we conclude the paper by discussing several of the different ways substitution tilings can be studied , and what sorts of questions are of interest . although the idea had been around for several years , self - similar tilings of the plane were given a formal definition and introduced to the wider public by thurston in a series of four ams colloquium lectures , with lecture notes appearing thereafter @xcite . throughout the literature one finds varying degrees of generality and some commonly used restrictions . we make an effort to give precise definitions here , adding remarks which point out some of the differences in usage and in terminology . for the moment we assume that the only rigid motions allowed for equivalence of tiles are translations ; this follows @xcite and @xcite . we give the definitions as they appear in @xcite , which includes that of @xcite as a special case . let @xmath39 be a linear transformation , diagonalizable over @xmath40 , that is expanding in the sense that all of its eigenvalues are greater than one in modulus . a tiling @xmath32 is called _ @xmath41-subdividing _ if 1 . for each tile @xmath42 , @xmath43 is a union of @xmath32-tiles , and 2 . @xmath44 and @xmath45 are equivalent tiles if and only if @xmath43 and @xmath46 form equivalent patches of tiles in @xmath32 . a tiling @xmath32 will be called _ self - affine with expansion map @xmath41 _ if it is @xmath41-subdividing , repetitive , and has finite local complexity . if @xmath41 is a similarity the tiling will be called _ self - similar_. for self - similar tilings of @xmath47 or @xmath48 there is an _ expansion constant _ @xmath49 for which @xmath50 . the rule taking @xmath51 to the union of tiles in @xmath43 is called an _ inflate - and - subdivide rule _ because it inflates using the expanding map @xmath41 and then decomposes the image into the union of tiles on the original scale . if @xmath32 is @xmath41-subdividing , then it will be invariant under this rule , therefore we show the inflate - and - subdivide rule rather than the tiling itself . the rule given in figure [ square1 ] is an inflate - and - subdivide rule with @xmath52 . however , the rule given in figure [ dp1 ] is not an inflate - and - subdivide rule . [ chair ] the l - triomino " or chair " substitution uses four prototiles , each being an l formed by three unit squares . we have chosen to color the prototiles since they are inequivalent up to translation . the expansion map is @xmath53 and in figure [ chair1 ] we show the substitution of the four prototiles . this geometric substitution can be iterated simply by repeated application of @xmath41 followed by the appropriate subdivision . parallel to the symbolic case , we call a tile that has been inflated and subdivided @xmath5 times a _ level-@xmath5 tile_. in figure [ chair2 ] we show level-@xmath5 tiles for @xmath54 and @xmath55 . one of the earliest results was a characterization of the expansion constant @xmath56 of a self - similar tiling of @xmath40 . ( thurston @xcite , kenyon @xcite ) a complex number @xmath49 is the expansion constant for some self - similar tiling if and only if @xmath49 is an algebraic integer which is strictly larger than all its galois conjugates other than its complex conjugate . the forward direction was proved by thurston and the reverse direction by kenyon . in @xcite , kenyon extends the result to self - affine tilings of @xmath0 in terms of eigenvalues of the expansion map . in the study of substitutions , from one - dimensional symbolic substitutions to very general tiling substitutions , the _ substitution matrix _ is an indispensable tool . ( this matrix has also been called the transition " , composition " , subdivision " , or even abelianization " matrix ) . suppose that the prototile set ( or alphabet ) has @xmath57 elements labeled by @xmath58 . the _ substitution matrix _ @xmath59 is the @xmath60 matrix with entries given by @xmath61 for example , the substitution in example [ const2d ] has substitution matrix @xmath62 when we label @xmath63 and @xmath64 . if an initial configuration of tiles has @xmath5 white tiles and @xmath57 blue tiles , then @xmath65^t$ ] is the number of white and blue tiles after one application of the substitution . since the substitution matrix is always an integer matrix with nonnegative entries , perron - frobenius theory is relevant ( see for example @xcite ) . the results we need require @xmath59 to be _ irreducible _ : for every @xmath66 there exists an @xmath5 such that @xmath67 . among other things , the perron - frobenius theorem states that if @xmath59 is irreducible , then the largest eigenvalue will be a positive real number that is larger in modulus than any of the other eigenvalues of the matrix . this eigenvalue is unique , has multiplicity one , and is called the _ perron eigenvalue _ of the matrix . primitivity , a special case of irreducibility , is particularly important . a matrix @xmath59 is _ primitive _ if there is an @xmath68 such that @xmath69 has strictly positive entries . primitivity of @xmath59 means if one substitutes any tile ( or letter ) a fixed number of times , one will see all of the other tiles ( or letters ) . this is a relatively strong property , and one that is almost always assumed in the literature . all of the substitutions in the paper are primitive except examples [ const2d ] and [ chacon1 ] . the following theorem , mentioned in @xcite , is proved in @xcite . ( praggastis @xcite ) an flc @xmath41-subdividing tiling is repetitive if and only if its substitution matrix is primitive . solomyak s papers @xcite and @xcite give several other key results for the dynamical systems of self - similar or self - affine tilings . the following is stated as a corollary to the perron - frobenius theorem . ( solomyak @xcite ) if @xmath59 is primitive , the perron eigenvalue of @xmath59 is the volume expansion @xmath70 . the perron left eigenvector gives the relative volumes of the prototiles . an algebraic integer is a _ pisot number _ if all of its algebraic conjugates ( except its complex conjugate ) are smaller than one in modulus . whether or not the expansion constant @xmath49 is a pisot number is especially important from a dynamical point of view . in dimensions one and two , solomyak @xcite has shown that a self - similar tiling dynamical system is not weakly mixing if and only if its expansion constant is a complex pisot number . we will see this number - theoretic property having other effects in sections [ comb.section ] and [ connections.section ] . so far , the results in this section have depended only on the substitution matrix and expansion constant , and not the geometry of the substitution . the final theorem in this section uses the notion of matching rules " , which are fundamentally geometric . roughly speaking , a set of matching rules determine which patches are allowed in a tiling . a simple yet classic example is the penrose tiling with marked rhombs , which we will encounter in example [ penrose.rhombs ] . the markings give matching rules that enforce " the substitution in the sense that any tiling of the plane constructed following the matching rules must be admitted by the penrose substitution . goodman - strauss was able to generalize this result to most geometric tiling substitutions : ( goodman - strauss @xcite ) every ( geometric ) substitution tiling of @xmath0 , @xmath71 can be enforced with finite matching rules , subject to a mild condition : we require that tiles admit a set of hereditary edges " such that the substitution tiling is sibling edge - to - edge " . we leave a discussion of the particulars to @xcite and note only that the sibling edge - to - edge " condition is mild enough to encompass most of the known examples . in the previous sections the only rigid motions allowed for tile equivalence were translations . however , there are natural tiling substitutions that require relaxing this to allow rotations . [ pinwheel ] the pinwheel " substitution rule acts on right triangles of side lengths 1 , 2 , and @xmath72 , inflating them by a factor of @xmath72 and subdividing into five triangles as shown in figure [ pinwheel1 ] . radin introduces pinwheel tilings in @xcite , attributing them to unpublished work of john h. conway . radin proves that the tiles appear in an infinite number of orientations that are uniformly distributed mod @xmath73 , and calls the tiling space statistically round " because it is invariant under rotations of infinite order . he also establishes matching rules for the pinwheel substitution , giving us the first amazing example where local matching rules produce infinite rotations ! [ pinwheelgenex ] sadun @xcite comes up with an interesting twist on the previous example : the generalized pinwheel " tilings . instead of requiring that the tiles be isometric to members of some finite prototile set , he requires only that they be conformally equivalent . the subdivision rules are quite straightforward but some choices arise for the inflation portion and we do not attempt to explain those here . at the first level , one takes a right triangle with side lengths @xmath74 and @xmath75 and subdivides it into 5 similar triangles as in figure [ pinwheelgen ] . the subdivision at the next stage takes place only on the largest of the triangles . one can continue subdividing indefinitely ; for the appropriate inflation at any stage and for precision of the results mentioned below we refer the reader to @xcite . all salient properties of a tiling admitted by the substitution depend on the angle @xmath76 . for the original pinwheel tiling @xmath77 , and that is one of the angles for which the tiling has finite local complexity , but the tiles appear in an infinite number of orientations . for other values of @xmath76 the tiles will only appear in a finite number of sizes . there is only one value , @xmath78 , for which the tiles appear in both a finite number of sizes and a finite number of orientations . figure [ pinwheel4 ] shows a few subdivisions of this case . . ] a close cousin of the self - similar tiling is the pseudo - self - similar tiling , which is generated by a variant of the inflate - and - subdivide rule . tiles are still inflated and then replaced by tiles from the original scale , but these may stick out of or not completely cover the inflated tile , so the substitution rule is imperfect " . we will show two well - known substitutions in this section , both of which can be converted into proper inflate - and - subdivide rules in different ways . [ penrose.rhombs]the penrose inflation using marked rhombs is shown in figure [ penroseinfl ] . when the reader attempts to inflate and subdivide a second time , she will notice that the subdivisions of adjacent tiles overlap . this is not a contradiction , however , because the overlapping tiles are equal and will therefore be considered the same tile . penrose tilings appear in many equivalent forms , with alternative tile shapes such as triangles or the famous kites and darts " . the scientific american article by gardner @xcite introduced the penrose tilings and many of their interesting properties to the general public . in chapter 10 of @xcite , the penrose tiles are studied as kites and darts and in other forms . it is possible to slice the penrose rhombs in half , creating triangles ( known as robinson triangles ) on which a proper inflate - and - subdivide rule can be defined ; we show this in figure [ penrose3 ] . chapter 6 of @xcite gives a detailed analysis of penrose tilings and includes the robinson triangles . penrose tilings have a number of interesting properties , most of which can be found in other tilings but were first observed in the penrose tilings . we mention a few of the highlights and leave the details to the references . there are ( at least ! ) three ways , other than substitution , to generate penrose tilings . one , which we have already discussed , is via matching rules : if an infinite tiling is constructed from penrose tiles with the requirement that adjacent tiles have matching arrows ( both in number and direction ) , this tiling will be a penrose tiling . amazingly , this local activity of matching the arrows forces the global property of being generated by a substitution ! two other methods for generating penrose tilings are the multigrid method and the projection method . since both methods rely on lattices , they can be used to prove that the nonperiodicity of the penrose tilings is a tightly controlled form of disorder . the _ multigrid method _ was discovered by debruijn @xcite . in this method one superimposes five grids of lines to create a pentagrid " ; every pentagrid is dual to a penrose tiling . this method is used by e. a. robinson , jr . @xcite to understand the penrose tilings as a dynamical system . a nice description of the _ projection method _ appears in @xcite , p. 195 - 196 . in this method a copy of @xmath0 is embedded in @xmath79 , and some lattice @xmath80 is chosen . points from @xmath81 are projected onto the copy of @xmath0 to form a tiling of @xmath0 . projection tilings have been studied extensively and are only sometimes obtainable by methods of substitution . a characterization of the intersection is given in a certain case in @xcite . point sets generated by generalized projection methods are called model sets " and are of great interest in mathematics and physics , and their spectrum is the subject of intense study ( see e.g. @xcite ) . if the reader were to experiment with a large set of penrose rhombs , he would quickly discover that it is difficult to tile a large region without disobeying the matching rules . indeed , almost all finite configurations of penrose rhombs , no matter how large , can not be extended to infinite tilings @xcite ! thus the fact that an infinite penrose tiling exists at all is a major result , and it can be proved using the presence of the inflation rule . [ binaryex ] the binary tilings " ( see @xcite ) are generated by the substitution rule shown in figure [ binary1 ] using the unmarked penrose tiles . the volumes expand linearly by a factor of @xmath82 . this substitution is interesting from the dynamical viewpoint as it produces a weakly mixing tiling dynamical system . the fact that it is weakly mixing means that it has a level of disorder not present in the penrose tilings , despite the fact that the relative frequencies of thick to thin rhombs is the same in both tilings ! weak mixing is evident in the diffraction spectrum , which is pictured in @xcite and is analyzed in a few papers including @xcite . one can redraw " the tiles of the binary tiling in such a way that the substitution rule becomes a perfect inflate - and - subdivide rule that gives rise to a self - similar tiling of the plane . the method is to take the limit as @xmath5 goes to infinity of the support of the level-@xmath5 tiles divided by the @xmath5th power of the expansion constant ( see figure [ binaryfractal ] , courtesy of e. arthur robinson , jr . ) . a general version of this process has been shown to work for all pseudo - self - similar tilings of the plane @xcite . this result is extended to tilings of @xmath0 in @xcite , provided the tiles appear in a finite number of orientations . in section [ connections.section ] we will use a similar process to transform combinatorial substitutions into geometric ones , with some unexpected results . the self - similar tilings and their close relatives in the previous section come from substitution rules that have one thing in common : a single similarity ( or expanding linear map , in the self - affine case ) governs the inflation of all of the tiles . now we consider substitutions for which a tile and its replacement may be geometrically unrelated , or for which there are several linear maps governing the tile inflations . there is no unified definition for this class of substitutions . attempts to define a tiling substitution based on the dual graph of the tiling have been made @xcite , and we call this method constructive " . however , there are perfectly reasonable tiling substitutions for which combinatorial information is insufficient to define the substitution rules . substitutions of this type we call non - constructive " . leaving a formal presentation of the definitions to the references , we simply present ideas and some examples . in this type of substitution , level-@xmath5 tiles can be constructed using only the information about adjacencies between tiles , making it possible to iterate the substitution using only local information . a tiling of @xmath19 makes a drawing of a planar graph : there is a vertex wherever three or more tiles meet , an edge wherever exactly two tiles meet , and the tiles themselves are the facets . the elements of this graph can be labeled according to the tile and adjacency types they represent ; we may choose a labeling scheme that provides as much or as little information about the surrounding tiles as we wish . any planar tiling has a _ dual graph _ : vertices of the dual graph are tiles of the tiling , there is an edge between two vertices if the corresponding tiles are adjacent , and facets correspong to vertices in the tiling . the elements of a dual graph inherit the labels of their dual counterparts . part of a tiling and its dual graph are shown in figure [ dual.graph ] , using the conventions that numbers represent both the label and the vertex of the graph , and that the edge and facet labels are suppressed . the dual graph is a natural object on which to define a substitution , by analogy with one - dimensional symbolic substitutions . the labeled vertices are the letters of an alphabet that corresponds to our prototile set . any rule that replaces a vertex with a finite graph might correspond to a substitution rule for a prototile , provided the geometry of the tiles allows it . a difficulty is that two vertices ( tiles ) may be adjacent in many different ways , so to make the dual graph see " this , we keep track of the adjacency types in our edge and facet labels . constructive combinatorial substitutions specify exactly how the substituted graphs of two adjacent elements should be attached . the next example , taken from @xcite , will provide some intuition . [ fib.dpv ] we obtain a _ direct product variation _ ( dpv ) substitution by rearranging some of the tiles in the fibonacci substitution of figure [ dp1 ] to break up the direct product structure . here we have carefully rearranged the tiles in the substitution of the type 1 tile so that the substitution can be iterated without inconsistency . we show the result in figure [ graph.sub ] , along with the induced graph substitution . to obtain the level-@xmath5 block , one simply concatenates the level-@xmath83 blocks in the obvious " manner as shown in figure [ dpv2 ] , matching sides that have the same length in the order prescribed by figure [ graph.sub ] . since the side lengths of the level-@xmath83 blocks are fibonacci numbers , the fit is guaranteed at each stage . note that the plaid " appearance of the direct product ( figure [ dp2 ] ) has disappeared . but what if you want to substitute a pair of adjacent tiles within a level-@xmath5 block ? you can not do it consistently without knowing the larger context of the adjacency , that is , the tiles that surround them in the tiling . for example , consider two horizontally adjacent tiles of type 1 . that adjacency appears twice in the level-3 tile of type 1 and we have circled them on the left side of figure [ adj.subs ] ; what happens under substitution is shown on the right of the same figure . for this graph substitution , the problem can be handled by relabeling the edges and facets of the graph in terms of the immediate configuration of tiles the edge or facet is contained in . the basic idea of a constructive combinatorial substitution for tilings as it appears in @xcite and in a similar form in @xcite is this . given a labeled vertex set @xmath84 representing the prototile types , a map from @xmath84 to the set of nonempty labeled graphs on @xmath84 is the basis for the substitution rule . the edges and facets of these graphs are labeled to give information about the types of adjacencies they represent in a tiling . the substitution rule also specifies how to substitute the labeled edges and facets so that we know how to connect the vertex substitutions contained in certain labeled graphs . ( the need to specify graph substitutions on facets and not just edges is illustrated in an example in @xcite ) . defining a tiling substitution rule this way is quite tricky since most labeled graphs do not represent the dual graph of a tiling . this interplay between combinatorics and geometry is where the technicalities come in to the formal definitions in the literature . [ rauzy.cst ] the tiling substitution of figure [ rauzy1 ] , introduced in @xcite , is based on a variation of the one - dimensional rauzy substitution " @xmath85 , @xmath86 . figure [ rauzy1 ] is obviously not enough information to iterate the substitution , so we specify how to substitute the important adjacencies " in figure [ rauzy2 ] . this is enough @xcite : there are no ambiguities when substituting other adjacencies , and facet substitutions do not include any new information . we show a few iterates of the tile of type 1 in figure [ rauzy3 ] , starting with the level-2 tile of type 1 . the fact that this substitution rule can be extended to an infinite tiling of the plane is proved using noncombinatorial methods in @xcite ; a combinatorial proof of existence would be welcomed . when trying to make up new examples of combinatorial tiling substitutions it is easy to create examples that fail to be constructive . the problem arises in the substitution of adjacencies : it may happen that no finite label set can be chosen to describe all adjacencies sufficiently to know how to substitute them . there is evidence to suggest that this sort of example can arise when the constant which best approximates the linear growth of blocks is not a pisot number . the author is not aware of any formal definition containing this group and so proposes the following definition , which works directly with the tiling and does not involve dual graphs . [ non.constructive.cst ] a _ ( non - constructive ) tiling substitution _ on a finite prototile set @xmath31 is a set of nonempty , connected patches @xmath87 satisfying the following : 1 . for each prototile @xmath88 and tile @xmath89 , and for each integer @xmath90 , there are rigid motions @xmath91 such that @xmath92 , where 2 . for any @xmath93 in @xmath94 , the patches @xmath95 and @xmath96 intersect at most along their boundaries . we say a tiling @xmath32 is _ admitted by the substitution @xmath97 _ if every patch in @xmath32 appears as a subpatch of some element of @xmath97 . this very general definition is satisfied by every substitution appearing in this paper except the penrose substitution on rhombi and those generalized pinwheel tilings that do not allow a finite number of tile sizes . the rauzy substitution of example [ rauzy.cst ] has a particularly efficient representation by this definition . the patches @xmath94 are given to the right of the arrows in figure [ rauzy1 ] , with all lower right corners at the origin . now , @xmath98 and @xmath99 , so @xmath100 and @xmath101 are the identity map . we find @xmath102 , so @xmath103 is the identity map and all we have left to figure out is the formula for @xmath104 . it turns out that @xmath104 is translation by a vector @xmath105 that can be computed recursively . let @xmath106 and @xmath107 ; for @xmath108 we have that @xmath109 . the fibonacci dpv of example [ fib.dpv ] also has a relatively simple formuation in terms of definition [ non.constructive.cst ] . the side lengths of level-@xmath5 tiles are given recursively , and the placement of the level-@xmath83 tiles to create level-@xmath5 tiles depends only on these side lengths . thus the translations @xmath110 are computable recursively as well . the next example is also a dpv , but it can not be defined in terms of dual graphs and is non - constructive . we encourage the reader to think about how to write up definition [ non.constructive.cst ] in this case . [ np.dpv ] consider a dpv arising from a one - dimensional substitution @xmath111 . from the direct product of this substitution with itself , we choose only to rearrange the substitution of the type-1 tile as in figure [ non - pisot1 ] . the substitution matrix of this one - dimensional substitution has perron eigenvalue @xmath112 , which is not a pisot number : its algebraic conjugate @xmath113 is larger than one in modulus . using analysis similar to what we will see in section [ connections1d ] , we find that constants times @xmath114 are the best approximation to the lengths of level-@xmath5 blocks . the reader can jump ahead to the left side of figure [ non - pisot3 ] to see the effect of @xmath115 being non - pisot on the substitution . there we find an adjacent pair of tiles ( circled in red ) which , under substitution , go to disjoint level-1 tiles ( circled in green ) . it is possible to show that for any large @xmath35 there is a level-@xmath5 tile containing adjacent tiles @xmath116 and @xmath117 such that the distance between the substitutions of @xmath116 and @xmath117 in the level-@xmath118 tile is greater than @xmath36 . the method of proof relies on the fact that @xmath115 is non - pisot and the result in @xcite . to illustrate that non - constructive tiling substitutions can have non - square tiles , to show an interesting connection to fault lines " ( see section [ curiousconnections ] ) , and for its entertainment value , we include an unpublished substitution discovered by the author in 2002 . [ nptriangles ] this substitution uses eight prototiles , shown to the left of the arrows in figure [ non - pisottriangles1 ] . the perron eigenvalue of the substitution matrix is the same non - pisot number as in the previous example . again an argument can be made to show that there are always adjacencies that pull apart under the substitution , no matter how well we try to label them . the left side of figure [ non - pisottriangles2 ] shows four iterations of the square formed by two red triangles . in example [ binaryex ] , we saw an improper substitution that could be fixed " by redrawing the tile boundaries with the replace - and - rescale method . in this method , the new tile boundary is the limit of the boundaries of the level-@xmath5 tiles , rescaled by the @xmath5th power of the inflation factor . in example [ binaryex ] , the result was a set of fractal prototiles satisfying a proper inflate - and - subdivide rule . it turns out that this process can be used to create prototile sets for self - similar tilings in the case of some combinatorial substitutions as well . we begin with the one - dimensional case , which is well - understood . given a symbolic substitution of non - constant length , it is easy to create an inflate - and - subdivide rule on labeled intervals ( tiles ) that has the same combinatorics as the symbolic substitution . the expansion constant is the perron eigenvalue of the substitution matrix and the tile lengths are given by the perron eigenvector . we illustrate with the fibonacci substitution in a way that looks ahead to the higher - dimensional case . [ nonconst1dsst ] we introduce some geometry by thinking of two unit length prototiles represented by different colors in figure [ fib.subs.unit ] . the reader can check that the level-@xmath5 tiles will have lengths given by the entries of @xmath119 , where @xmath59 is the substitution matrix @xmath120 $ ] . this matrix has eigenvalues given by @xmath121 : the golden mean @xmath122 and its conjugate @xmath123 . let @xmath124 and @xmath125 denote the associated left eigenvectors . there are constants @xmath126 and @xmath127 so that @xmath128 , which gives us the vector of level-@xmath5 tile lengths : @xmath129 the lengths of the intervals for our self - similar tiling are the entries of @xmath130 . the length of the type-@xmath6 tile is @xmath131 and the length of the type-@xmath132 tile is @xmath126 . these lengths form an eigenvector for @xmath59 , so there exists an inflate - and - subdivide rule , which we have shown in figure [ fib.subs ] . * notes : * ( 1 ) this process works on any substitution on @xmath57 letters provided that the vector @xmath133 lies in the span of the left eigenvectors of the substitution matrix of the substitution . it works trivially on constant length substitutions since @xmath134 , the vector of unit tile lengths , already forms a perron eigenvector for the substitution matrix . \(2 ) in the fibonacci example , since @xmath122 is a pisot number ( its conjugate @xmath135 is smaller than one in modulus ) , the higher the inflation the less important the second term of equation [ fib.length.eqn ] becomes . thus the lengths of the level-@xmath5 tiles of the inflate - and - subdivide rule are asymptotically close to the lengths of the non - constant length substitution , and therefore approximately integers ! the situation is dramatically different , of course , if any of the secondary eigenvalues are strictly greater than one in modulus ( the strongly non - pisot case ) . the reader should not be too surprised to discover that this process will work for direct product substitutions , such as example [ nonconst2d ] , and their variations , such as examples [ fib.dpv ] and [ np.dpv ] . the level-@xmath5 blocks are rectangular and have side lengths given by the lengths of the one - dimensional substitution . rescaling by the expansion factor gives us rectangular tiles whose side lengths are determined by the perron eigenvector as before . thus , if the original one - dimensional substitutions have a proper inflate - and - subdivide rule , so will any dpvs associated with them . still , it is instructive to consider the two - dimensional replace - and - rescale method as it applies in this simple case . [ fib.sst ] consider the fibonacci dpv substitution of example [ fib.dpv ] . there are four tile types and the substitution matrix is @xmath136.$ ] if there is a proper inflate - and - subdivide rule corresponding to our substitution , it must have the same substitution matrix . this tells us that the volume expansion of the rule must be the perron eigenvalue , which in this case is @xmath137 , the square of the golden mean . ( the other eigenvalues are 1 , 1 , and @xmath138 . ) the level-@xmath5 tiles of the dpv substitution are supported on rectangles with side lengths given by either the @xmath5th or the @xmath83st fibonacci numbers . we rescale the volumes by @xmath139 to obtain prototiles for our self - similar tiling . the right " way to see this process is to consider a linear map @xmath140 that expands with the perron eigenvalue of the substitution matrix @xmath59 . ( how to find this map in general is quite unclear . ) in this example @xmath41 is given by the matrix @xmath141 $ ] . denoting the support of the level-@xmath5 tile of type @xmath142 as @xmath143 , we can find the support of the prototile @xmath144 for the inflate - and - subdivide rule that corresponds to @xmath142 by setting @xmath145 in figure [ dpv3 ] we compare level-@xmath146 tiles from the dpv ( left ) and the self - similar tiling ( right ) . .,title="fig:",width=192 ] .,title="fig:",width=192 ] [ rauzy.sst ] the self - similar tiling associated with the rauzy two - dimensional substitution of example [ rauzy.cst ] has as its volume expansion the largest root of the polynomial @xmath147 . the three tile types obtained by the replace - and - rescale method are shown in figure [ rauzy4 ] , compared with a large iteration of the substituton . the replace - and - rescale method can produce intriguing results , especially if the substitution is not constructive or not primitive . we look at the former case in examples [ np.sst ] and [ trinpsst ] and discover that the associated geometric substitution tilings may lose local finiteness . in example [ chacon ] we consider the latter case to see how a lack of primitivity can impact the geometric substitution ; in this case an attempt to fix " the situation yields new tiling substitutions that fail to have the expected relationship to one another . [ np.sst ] applying the replace - and - rescale method to the substitution in example [ np.dpv ] produces the inflate - and - subdivide rule of figure [ non - pisot2 ] . . ] it is proved in @xcite that any tiling admitted by this inflate - and - subdivide rule does not have finite local complexity since there are two tiles that meet in infinitely many different ways . ( examination of the tiling on the right of figure [ non - pisot3 ] may convince the reader that this is plausible ) . this lack of local finiteness means that the dynamical results found in @xcite , most notably that the system should be weakly mixing , can not be directly applied . , with a loss of flc on the right.,title="fig:",height=288 ] , with a loss of flc on the right.,title="fig:",height=288 ] the loss of finite local complexity happens along arbitrarily long line segments composed of tile edges ( proved for this example in @xcite but was found as a necessary condition in general in @xcite ) . as you travel along such a segment , a discrepancy in the number of short tile edges from one side of the line to the other appears ; on longer segments this discrepancy increases as more and more short edges pile up on one side than the other . because the tile edge lengths are not rationally related , this means that we must keep seeing new adjacencies as the discrepancy grows . in the limit one will see infinite _ fault lines _ along which tiles may slide across one another with arbitrary offsets . the growth of these discrepancies is made possible by the fact that the expansion constant s algebraic conjugate is greater than one in modulus ( i.e. it is strongly non - pisot ) . this is also responsible for the fact that original dpv has adjacencies that are ripped apart when substituted , as shown in the left of figure [ non - pisot3 ] . [ trinpsst ] as in the previous example , the substitution of example [ nptriangles ] gives rise to a self - similar tiling that does not have finite local complexity . in the previous example , fault lines could occur both horizontally and vertically . in this example , fault lines can occur horizontally , vertically , and diagonally , as one can see from the right side of figure [ non - pisottriangles2 ] . the author has not seen examples allowing fault lines in more than three distinct directions . [ chacon ] a famous one - dimensional dynamical system is given by the chacon cut - and - stack construction , which provided the first example of a weakly but not strongly mixing system ( see @xcite , p. 133 for a synopsis of the results in the one - dimensional case ) . the cut - and - stack system can be recoded by the symbolic substitution @xmath148 , and figure [ chaconsq ] shows a dpv substitution based on this . another tiling version of the construction , shown in figure [ chacon1 ] , is analyzed from a dynamical systems viewpoint in @xcite and put in the context of combinatorially substitutive tilings in @xcite . the four prototiles used in those works are not square , but are a rescaling of the supports of the level-1 tiles in figure [ chaconsq ] . this substitution is not primitive since no matter how many times we substitute the three small tiles , they will never contain the large square . because of this we can not obtain a meaningful self - similar tiling directly using the replace - and - rescale method : the replacements of all but the large square will have volumes that go to zero under rescaling , leaving us with only the first tile and a trivial substitution . there is a way to recode the system into a primitive one , producing the prototiles shown to the right of the arrows in figure [ chacon3 ] . by referring to figure [ chacon2 ] , the reader may be convinced that knowing the surroundings of a particular tile is enough to decide unambiguously with which new prototile to replace it . the nonprimitive chacon substitution turns into the primitive one of figure [ chacon4 ] . because there is a locally defined map taking tilings admitted by the nonprimitive substitution to tilings admitted by the primitive substitution , and vice versa , the tiling spaces are considered _ mutually locally derivable_. this means that the dynamical systems are equivalent in the sense of topological conjugacy " @xcite , thus important dynamical features are preserved . one such dynamical feature , proved in @xcite , is that the dynamical system under the action of @xmath20 is weakly mixing . the larger eigenvalue of the substitution matrix of the chacon primitive substitution is 9 , and it is not difficult to see that the length expansion is governed by powers of 3 . the replace - and - rescale method produces a prototile set of five congruent squares ; the inflate - and - subdivide rule is shown in figure [ chacon5 ] . what makes this example curious is that the dynamical systems of the combinatorially substitutive tiling and its associated self - similar tiling are distinctly different . one can use results from @xcite to show that under the @xmath19 action , the self - similar tiling dynamical system is not weakly mixing . an embedded @xmath20 action would therefore also fail to be weakly mixing . this stands in contrast to the weakly mixing @xmath20 action proved in @xcite when the substitution is only combinatorial . one can see that the systems are misaligned " by considering figure [ chacon6 ] and comparing the location of the red circle in each substitution , which for @xmath149 and @xmath26 represents the central level-@xmath5 tile within its level-@xmath118 tile . one can check that as @xmath5 grows without bound so does the distance between the red dots . that is , if any corners of the level-@xmath5 tiles are lined up , the red dots will move further and further away from one another ! substitution tilings are being studied from topological , dynamical , physical , combinatorial , and other perspectives , often in conjunction with one another . in this section we will briefly outline areas of current interest and possible questions for future study . tilings make good models for the atomic structure of crystals and quasicrystals , and perhaps the most exciting work on them is being done at the intersection of physics and topology . methods for investigating certain tiling spaces via @xmath150-algebras have been developed and are nicely summarized in @xcite . the types of tilings that are most easily evaluated this way are self - similar tilings and tilings generated by the projection method . ( some tilings , such as the penrose and the octagonal " tilings , fall into both categories ) . the k - theory of these @xmath150-algebras are of interest to both mathematicians and physicists . the possible energy levels of electrons in a material modeled by a tiling determine gaps in the spectrum of the associated schrdinger operator . the k - theory gives a natural labeling of the spectral gaps , thus providing theoretically relevant physical information ( see @xcite for a detailed discussion of this branch of study ) . it is believed that there may be additional physical interpretations for k - theory and other topological invariants of tiling spaces . there is more promising topological work being done as well . for instance , it has been shown in varying degrees of generality ( see @xcite and references therein ) that flc tiling spaces are inverse limits . successful efforts to compute the homology and cohomology of tiling spaces , and to connect these results to k - theory , have been plentiful . a nice summary of the current state of the art , along with the discovery of torsion and its ramifications , appears in @xcite with a primary emphasis on canonical " projection tilings . an informal discussion of the connections between some physical and mathematical problems appears in @xcite , with a focus on recent progress in the cohomology of tiling spaces . included is a summary of the work in @xcite involving cohomological analysis of the deformations of tiling spaces . an important question is the extent to which the homology and cohomology of tiling spaces has physical interpretations . almost all of the existing literature on the topology of tiling spaces makes the assumption of local finiteness . this is , after all , an appropriate restriction , given that the model of atomic structure requires tiles ( atoms ) to fit together in only a finite number of ways . however , examples exist of geometric tiling substitutions that result in non - flc tilings , for example danzer s triangular tilings @xcite and the tiling from @xcite , which is easily generalized using the methods of section [ connections.section ] . in @xcite , kenyon was the first to consider the conditions under which a tiling of @xmath19 with finitely many tile types can lose local finiteness : the tile boundaries must contain circles or arbitrarily long line segments , thus substitution tilings without finite local complexity have fault lines along which tiles can slide past one another . in @xcite , the cohomology of a highly restricted class of non - flc substitution tilings was successfully computed , and it was shown that each fault line leaves a sort of signature in the cohomology in dimension 3 , even though the tilings are two - dimensional . it is a topic of current interest to understand the topology of tiling spaces without finite local complexity . at the intersection of mathematical physics and dynamical systems is the connection between the diffraction spectra of quasicrystalline solids and the dynamical spectra of the tilings that model them . the fact that these spectra are related at all is first established in @xcite , while the mathematical description of the diffraction spectral measure is given sound theoretical footing in @xcite . much of the work to date has centered around discrete point sets called delone sets , which can be thought of as locations of molecules and which can be converted into tilings in a few different ways . ever since schectman et . @xcite discovered quasicrystals in a laboratory experiment , people have been trying to figure out which delone sets are diffractive " in that their spectra exhibit sharp bright spots . mathematically it is interesting to ask when the spectra consists only of sharp bright spots , i.e. when it has pure point spectrum " . more precisely , one defines a _ spectral measure _ which can be broken into pure point , singular continuous , and absolutely continuous pieces with respect to lebesgue measure . great progress has been made for model sets " ( obtained by a generalized projection method ) , and for delone sets generated by substitutions ; a current synopsis of the state of the art appears in @xcite . it is now known that for certain locally finite delone sets the notions of pure point dynamical and pure point diffraction spectra coincide @xcite . this was generalized in @xcite in a measure - theoretic setting which allows for a lack of local finiteness . the question of whether certain substitution systems consist of model sets can be investigated by looking for modular coincidences " ; @xcite has an algorithm and many examples , which build upon the work in @xcite . questions remain regarding the connections when there is any continuous portion of the spectral measures . the dynamical spectra of specific geometric tiling substitutions have been studied ( @xcite and others ) but are not completely understood . related to the study of tilings and model sets is a question in dynamical systems theory . for one - dimensional symbolic substitutions , it is sometimes possible to find a geometric realization " of the substitution . a formal definition appears in @xcite , p. 140 , but the idea is that a geometric realization is a geometric dynamical system , ( such as an irrational rotation of the circle ) , which encodes the system via partition elements . for example , the fibonacci substitution sequences can be seen to code , in an almost one - to - one fashion , addition by @xmath151 on a one - dimensional torus , where @xmath122 is the golden mean ( see @xcite , p. 199 for details ) . orbits in this geometric realization look like " one - dimensional tilings . several more examples are given in @xcite , p. 231 . if a tiling dynamical system arises from a model set , then it can be seen as a geometric realization . for example , it is shown in @xcite that the penrose rhombic tiling dynamical system is an almost one - to - one extension of an irrational rotation on a 4-dimensional torus . in general , we do not know when substitution tilings have geometric realizations . this paper provides the full extent , to the author s knowledge , of known classes of combinatorial tiling substitutions . all of the examples in section [ comb.section ] are obtained by various means from one - dimensional symbolic substitutions . what other mechanisms exist for generating combinatorial substitutions ? is there a method for obtaining non - geometric substitutions from geometric ones ? it seems clear that there should be a multitude of other examples waiting to be discovered , and finding them is of paramount importance . combinatorial tiling substitutions have hardly begun to be studied from the dynamical systems viewpoint . in analogy with the self - similar case and many of its generalizations , we would like to investigate basic ergodic - theoretic properties such as repetitivity , unique ergodicity , and recognizability . this program was carried out in @xcite on a restricted class of two - dimensional symbolic substitutions of non - constant length for which standard " techniques could be applied . unfortunately , these techniques do not necessarily work in the non - geometric case . the crucial missing piece is that the substitution rule can not always be seen as an action from the tiling space to itself : the substitution can be applied only to level-@xmath5 tiles , not to entire tilings . many combinatorial tiling substitutions do not extend to maps of the tiling space in a canonical way , and it is unclear whether ( or when ) any of them do . new methods will need to be devised to tackle even the most basic questions in the dynamical systems and ergodic theory of combinatorial substitution tilings . a closely related concept , essential to many standard arguments , is whether the substitution map can be locally undone " so that one can detect the level-1 tile in a given region without requiring infinite information about the tiling . this is called _ recognizablility _ in the sequence case and the _ unique composition property _ in the self - similar tilings case . when the substitution acts as a continuous map on the tiling space , unique composition is equivalent to the substitution map being invertible . in the event of non - periodicity , recognizability and unique composition were proved in @xcite and @xcite , respectively . although the substitution map may not make sense on tiling spaces in the non - geometric case , the notion of unique composition still does , and a natural conjecture is that combinatorial substitutions possess it whenever they are non - periodic . in one dimension , there is great interest in the theory of combinatorics on words " ( see part i of @xcite for an extensive exposition ) . in this theory , one considers finite blocks of letters and investigates how often they appear , and in what combinations , within sequences . substitution sequences are particularly fertile for this type of study . the _ complexity _ of a sequence is a function @xmath152 telling how many words of length @xmath5 exist in the sequence ; this can be used to compute the topological entropy @xcite , p. 4 . any non - periodic sequence with minimal complexity is called _ sturmian _ , the classic example being the sequence given by the fibonacci substitution . one can read about the numerous consequences of being sturmian in chapter 6 of @xcite . the notion of complexity can be generalized to higher dimensions and some results exist in this direction ( see @xcite and references therein ) . combinatorial substitutions such as the rauzy substitution of example [ rauzy.cst ] are a natural place to look for examples of low - complexity sequences . some problems that have been at least partially resolved for geometric substitutions are still open for combinatorial ones . for instance it is completely unclear whether there should exist matching rules which force tiles to fit together as prescribed by combinatorial substitutions . would it be possible to use the matching rules for their associated self - similar tilings , which we know exist by @xcite , to find them ? another question is , since the connection to the atomic structure of solids is an important motivation for the study of tiling spaces , can we identify the diffraction or dynamical spectrum of combinatorial substitution ? it is known that the dynamical spectrum of the chacon substitution is trivial since it is weakly mixing @xcite , and following @xcite it is reasonable to conjecture that dpv substitutions without pisot expansions might also be weakly mixing . for combinatorial substitutions , is the spectrum largely dependent on the perron eigenvalue of the substitution matrix , as it is in the geometric case @xcite ? or is the situation like the one - dimensional symbolic case , which is also highly sensitive to the combinatorics of the substitutions ? the first open question is , when does a combinatoral tiling substitution give rise to a reasonable geometric one ? we have already seen that the non - primitive chacon substitution of example [ chacon ] does not . there must be substitutions for which the limit in the replace - and - rescale method does not exist , or produces topologically unpleasant tiles . in fact , it is unclear exactly how the replace - and - rescale method ought to properly be applied : determining the appropriate linear expansion map is problematic for at least two reasons . first , the combinatorial substitution might be encoding an inflate - and - subdivide rule that does not inflate as a similarity . this means that knowing the volume expansion would not tell us the appropriate length expansions . second , if the linear map is a similarity , there may be some rotation inherent in the combinatorial substitution that would need to be expressed in the linear map , as in the rauzy substitution of example [ rauzy.cst ] . in the best circumstances we could hope to find conditions under which the expansion can be found , the limiting tiles are topologically nice " , and a proper inflate - and - subdivide rule exists . it is interesting to consider the relationship between the dual graphs of a combinatorial substitution and its associated geometric substitution ( if it exists ) . it is clear from figure [ dpv3 ] that the dual graphs must always have the same labeled vertices , but the edge and facet sets do not seem to bear a consistent relationship to one another . in the case of example [ fib.sst ] the edge set of the dpv s dual graph is contained in that of its self - similar tiling , but this is not true in general . since the unlabeled dual graphs are not isomorphic , there is no homeomorphism of the plane taking one to another ( see @xcite , p. 169 ) . can an understanding of the combinatorial properties of one tiling still give us insight into the other ? in one - dimensional symbolic dynamics , the curtis - lyndon - hedlund theorem ( see @xcite ) states that homeomorphisms between symbolic dynamical systems are equivalent to local maps called sliding block codes " . a sliding block code transforms one sequence into another element by element , deciding what to put in the new sequence by looking in a finite window in the old one . similarly , one tiling can be transformed locally into another ; if the process is invertible the tilings are mutually locally derivable . it has come to light that there is no curtis - lyndon - hedlund theorem for tiling dynamical systems @xcite . using the basic method of @xcite , one can show that example [ fib.dpv ] and the associated self - similar tiling of example [ fib.sst ] have topologically conjugate dynamical systems without the possibility of mutual local derivability . we conjecture that in the pisot case , combinatorial substitutions have topologically conjugate dynamical systems with their geometric counterparts . in general one would not expect the conjugacy to be through mutual local derivability . our final question takes note of the fact that the dynamical relationship between substitution sequences and self - similar tilings of the line is especially subtle . on a sequence space there is a @xmath153-action ; passing to a tiling by choosing tile lengths provides a natural action by @xmath47 called a suspension " . surprisingly , the continuous action of the tiling space is probably better understood than the discrete action ! for instance , the presence or absence , and nature of , eigenvalues of the tiling dynamical system can be understood in terms of the expansion constant along with certain geometric information @xcite . this situation is far more complicated in the symbolic case and the interested reader should see @xcite , section 7.3 for a synopsis . also , topological conjugacies between different suspensions have been thoroughly considered in @xcite , where it is seen that the eigenvalues of the substitution matrix play a critical role . we can consider tilings such as those in figure [ non - pisot3 ] as being suspensions of the same sequence in @xmath20 , and ask similar questions about their spectra and topological properties . more generally , we can consider tilings such as those in figure [ non - pisottriangles2 ] as being suspensions of the same labeled graph . this perspective yields an interesting set of problems at the intersection of dynamics and combinatorics . j. bellissard , d. hermmann , and m. zarrouati , hull of aperiodic solids and gap labeling theorems , in _ directions in mathematical quasicrystals _ ( m. baake and r. v. moody eds . ) crm monograph series , american mathematical society , providence , 2000 . j. kellendonk and i. putnam , tilings , @xmath150-algebras , and k - theory , in _ directions in mathematical quasicrystals _ ( m. baake and r. v. moody , eds . ) , crm monograph series , american mathematical society , providencee , 2000 . b. praggastis , markov partitions for hyperbolic toral automorphisms , ph . d. thesis , university of washington , 1994 . n. priebe , towards a characterization of self - similar tilings in terms of derived orono tessellations , _ geom . * 79 * ( 2000 ) , no . 3 , 239 - 265 .

gj1214b is a warm sub - neptune transiting in front of a nearby m dwarf star . recent observations indicate the presence of high and thick clouds or haze whose presence requires strong atmospheric mixing . in order to understand the transport and distribution of such clouds / haze , we study the atmospheric circulation and the vertical mixing of gj1214b with a 3d general circulation model for cloud - free hydrogen - dominated atmospheres ( metallicity of 1 , 10 and 100 times the solar value ) and for a water - dominated atmosphere . we analyze the effect of the atmospheric metallicity on the thermal structure and zonal winds . we also analyze the zonal mean meridional circulation and show that it corresponds to an anti - hadley circulation in most of the atmosphere with upwelling at mid - latitude and downwelling at the equator in average . this circulation must be present on a large range of synchronously rotating exoplanets with strong impact on cloud formation and distribution . using simple tracers , we show that vertical winds on gj1214b can be strong enough to loft micrometric particles and that the anti - hadley circulation leads to a minimum of tracers at the equator . we find that the strength of the vertical mixing increases with metallicity . we derive 1d equivalent eddy diffusion coefficients and find simple parametrizations from m/s for solar metallicity to m/s for the 100solar metallicity . these values should favor an efficient formation of photochemical haze in the upper atmosphere of gj1214b .

1509.06814

gj1214b is a warm sub - neptune orbiting a nearby m dwarf and discovered by the mearth survey @xcite . it is one of the rare low - mass planets whose atmosphere is characterizable in transit spectroscopy by current telescopes . gj1214b is therefore considered as a privileged archetype of this new category of planets also called mini - neptunes . it has a mass of 6.55@xmath40.98 @xmath5 and a radius of 2.68@xmath40.13 @xmath6 , giving it a density of around 1.88 g @xmath7 @xcite . this low density necessarily implies the presence of a thick atmosphere . interior models suggest that the planet possesses either a dense iron / rock core surrounded by a thick hydrogen / helium - rich atmosphere or a water - rich core surrounded by a thick steam atmosphere @xcite . @xcite found that a water - rich planet depleted in hydrogen would require a water - to - rock ratio larger than 6:1 . they considered such a large ratio as unlikely and favored instead an intermediate case with a hydrogen / helium / water - rich atmosphere . gj1214b has a tight orbit ( i.e. semi - major axis of 0.014 au ) and is very likely tidally locked , with a permanent dayside and nightside . several observations in transit spectroscopy were performed to probe gj1214b s atmosphere and to break the degeneracy of possible compositions . first observations by @xcite revealed a lack of spectral features between 0.78 and 1 @xmath8 m , ruling out the hypothesis of a cloud - free solar composition . other observations @xcite confirmed this flat spectrum between 0.6 and 4.5 @xmath8 m while one group found a significant difference in the transit depth measured in j - band ( @xmath91.25 @xmath8 m ) and in ks - band ( @xmath92.15 @xmath8 m ) @xcite . finally , precise measurements with the wide field camera 3 ( wfc3 ) instrument on the hubble space telescope ( hst ) revealed a very flat spectrum between 1.15 and 1.65 @xmath8 m @xcite . that implied the presence of high cloud / haze diffusing or absorbing the stellar radiation at low pressure . the presence of clouds is required even for a high mean - molecular - mass atmosphere with a composition dominated by water , methane , carbon monoxide , nitrogen or carbon dioxide . the cloud - top pressure in transit spectroscopy should be less than 10@xmath10 mbar for a solar - like composition and 10@xmath11 mbar for a water - dominated atmosphere @xcite . for the conditions on gj1214b , these clouds / haze could be either condensate clouds of potassium chloride ( kcl ) and zinc sulfide ( zns ) or photochemical haze produced by the photolysis of methane in the upper atmosphere @xcite . kcl and zns clouds are supposed to form between 0.1 and 1 bar @xcite . their presence at low pressure would thus require a strong atmospheric circulation lofting cloud particles over several scale heights . the persistence of photochemical haze at low pressure would also require a sufficiently strong vertical mixing , counteracting particle sedimentation and bringing methane to altitudes where photolysis occurs . therefore , the formation of high and thick cloud / haze on gj1214b is strongly linked to the atmospheric circulation . the latter has been investigated in previous studies using general circulation models ( gcms ) for cloud - free atmospheres @xcite . @xcite used gray opacities and showed that the superrotating zonal winds and the thermal phase curves would vary greatly with the atmospheric metallicity . @xcite analyzed the atmospheric dynamics and thermal phase curves with a more accurate gcm using non - gray opacities for different atmospheric compositions , but they did not analyze the vertical winds and the vertical mixing . in this paper , we analyze the circulation and the vertical mixing of gj1214b for a cloud - free atmosphere with different compositions using the generic lmdz gcm , a very versatile 3d model developed to simulate any kind of planetary atmosphere . in the next section , we describe the model and the parameters used for gj1214b . in section 3 , we analyze the thermal structure and atmospheric circulation ( zonal , meridional and vertical winds ) . in section 4 , we analyze the vertical mixing of tracers and discuss its implications for cloud / haze formation in section 5 . the 3d modeling of realistic clouds with our gcm will be described in a dedicated paper , analyzing their formation , dynamics and impacts on spectra . we simulated the atmosphere of gj1214b using the generic lmdz gcm . this model has been specifically developed for exoplanet and paleoclimate studies . it has been used for studying the early atmospheres of the earth , mars and titan @xcite and for studying the habitable zone for terrestrial exoplanets @xcite . the model is derived from the lmdz earth gcm @xcite , which solves the primitive hydrostatic equations of meteorology using a finite difference dynamical core on an arakawa c grid . this dynamical core has been used in gcms dedicated to the present earth @xcite , mars @xcite , venus @xcite , and titan @xcite . in this paper , simulations were performed with a horizontal resolution of 64@xmath348 ( corresponding to resolutions of 3.75@xmath12 latitude by 5.625@xmath12 longitude ) . we also did a few tests for gj1214b with a 128@xmath396 resolution but we did not notice significant differences . for the vertical discretization , the model uses pressure coordinates . in this work , we used 45 layers equally spaced in log pressure , with the first level at 80 bars and the top level at 3 pa . the radiative scheme is based on the correlated - k model , with k - coefficients calculated from high resolution absorption spectra , computed by kspectrum , a line - by - line model using the hitran 2012 database @xcite . kspectrum is a tool developped by vincent eymet and available online at http://www.meso-star.com/en_products.html . for water opacities , we used the high - temperature database hitemp 2010 @xcite . for methane , the hitran database does not include absorption lines between 9000 and 11000 cm@xmath11 and between 11500 and 12000 cm@xmath11 . we added the missing lines using exomol @xcite . at a given pressure and temperatures , correlated - k coefficients in the gcm are interpolated from a matrix of coefficients stored in a 12 @xmath3 9 temperature and log - pressure grid : t = 100 , 200 , .... , 1000 , 1200 , 1500 , 2000 k , p = 10@xmath11 , 10@xmath14 , 10@xmath15 , .... , 10@xmath16 pa . to facilitate comparison with the work of @xcite , we computed k - coefficients over the same 11 spectral bands . the h@xmath17-h@xmath17 and h@xmath17-he cia from the hitran database were included . rayleigh scattering by h@xmath17 , h@xmath17o and he was included , based on the method described in @xcite and using the @xcite scheme to compute the radiative transfer . the model also includes a scheme for adiabatic temperature adjustment triggered in the regions where the atmosphere is convectively unstable . however , this condition almost never occurs in our simulations of gj1214b . we ran simulations with four different atmospheric compositions : * h@xmath17-rich atmosphere at 1@xmath3solar metallicity * h@xmath17-rich atmosphere at 10@xmath3solar metallicity * h@xmath17-rich atmosphere at 100@xmath3solar metallicity * pure h@xmath17o atmosphere for the h@xmath17-rich atmospheric composition , we used the solar nebula atomic abundances from @xcite . the 10@xmath3solar metallicity is similar to saturn s metallicity , and the 100@xmath3solar is quite similar to uranus and neptune s metallicity @xcite . population synthesis models generally predict high metallicities between 100 - 400@xmath3solar for low - mass planets but lower metallicities can also arise @xcite . for the h@xmath17-rich atmospheres , we computed the compositions at thermochemical equilibrium , assuming only h@xmath17 , he , h@xmath17o , ch@xmath18 , n@xmath17 , nh@xmath19 , co and co@xmath17 . we computed the mixing ratios analytically following the method described in the appendix of @xcite , using molecular gibbs energies from @xcite . we computed abundances from the reactions : @xmath20 @xmath21 @xmath22 the computation requires solving a cubic equation . figure [ figure_1 ] shows the vertical molecular abundances computed for 1 , 10 and 100 times the solar composition cases using temperature profiles from the 1d model ( see figure [ figure_2 ] ) . the opacities for a h@xmath17-rich atmosphere are dominated by molecular absorption ( mainly h@xmath17o ) . however , at high pressure and for temperatures higher than around 1000 k , the pressure broadening of atomic alkalin ( na and k ) spectral lines has a strong impact on the opacity in the visible . for a solar composition , atomic potassium should be the dominant k - bearing gas for temperatures higher than around 980k at 1 bar @xcite . we did not include na opacities in our model , as na absorbs at wavelengths that are shorter than the bulk of the radiation emitted by a later - type m dwarf like gj1214 . also , at the high metallicities considered here , na is likely to condense as na@xmath17s clouds in the deeper atmosphere @xcite , removing it as an atmospheric opacity source . for k , we simply added a continuum between 1.18@xmath23 - 1.43@xmath23cm@xmath11 , assuming pressure broadening and a reference cross section for atomic k of 10@xmath24cm@xmath1/atom at 1 bar and 1000 k for solar metallicity . this value corresponds to the absorption from @xcite at the edges of the spectral band we considered . model parameters are given in tables 1 and 2 . for the stellar spectrum , we used a blackbody at 3026 k. we also tested with spectra of ad leo and gliese 581 but these more realistic spectra did not produce significantly different results . we assumed a stellar flux of 23600 w / m@xmath1 at the top of the atmosphere for a null zenith angle . we also assumed an internal luminosity of 0.73 w / m@xmath1 corresponding to an intrinsic effective temperature ( i.e. the blackbody temperature corresponding to the internal luminosity ) of 60 k , as suggested by @xcite . as initial states for the 3d simulations , we used the 1d temperature profiles computed with the 1d version of the model ( see figure [ figure_2 ] ) and with no wind . the 1d model uses the same vertical discretization and physics ( e.g. radiative transfer , convective adjustment ) as the gcm , assuming a stellar zenith angle of 60@xmath26 and a redistribution of energy over the entire planet . for the 1d calculation , we ran the model for 1000 gj1214b days ( i.e. 1600 days ) and increased the radiative heating / cooling by a factor proportional to the pressure when the pressure was higher than 0.1 bar . this enables faster convergence and is equivalent to running the model for around 10@xmath27 days for the deepest level at 80 bars . this technique resulted in simulations very close to equilibrium after these 1600 days of integration , with relative differences between total emitted radiation and total absorbed radiation lower than 0.1@xmath28 and with no observable variation in the temperature profile for longer simulations . the 3d simulations were also run for 1600 days but we did not use the technique to accelerate the convergence since the atmosphere is not in radiative equilibrium in 3d . the differences between total emitted radiation and total absorbed radiation for the final 3d states were lower than 1@xmath28 . simulations were performed with a dynamical timestep of 60 seconds and a physical / radiative timestep of 300 seconds . the gcm outputs are either instantaneous values or daily averaged values . all the results we present here use daily averaged values . however , the 3d simulations show small time variability with , for instance , standard deviations for temperature generally lower than 10 k. the left panel in figure [ figure_2 ] shows the temperature profiles obtained from the 1d model to initialize the 3d runs . dashed and dotted lines correspond to the saturation vapor pressure curves for kcl and zns for the solar and 100@xmath3solar metallicity from @xcite . the temperature profile is only adiabatic below the isothermal region ( i.e. below 40 bars for the 100@xmath3solar composition and below 100 bars for the other cases ) where it is controlled by the internal heat flux . for these 1d temperature profiles , kcl and zns should condense at around 0.4 bar for the solar metallicity and at around 60 mbar for the 100@xmath3solar metallicity . these altitudes are similar to those obtained by previous 1d models @xcite and are fairly strongly dependent on the metallicity . our 3d simulations globally give the same condensation altitudes , however there are latitudinal variations because of the equator - pole temperature gradient . for instance , for the 100@xmath3solar metallicity ( the case with the strongest variations ) , kcl condensation would start at around 40 mbar at the equator and at around 200 mbar at the poles . in the 3d simulations , there are strong day - night temperature variations that become more pronounced with increasing atmospheric metallicity ( see right panel in figure [ figure_2 ] ) . at 0.1 mbar the day - night temperature difference reaches 500 k for the water atmosphere but is lower than 100 k for the solar metallicity . the atmospheric opacity increases with metallicity while the specific heat capacity decreases . both effects lead to larger heating / cooling rates above 0.1 bar for higher atmospheric metallicity ( see figure [ figure_3 ] left ) . the altitude where stellar energy is deposited increases with metallicity ( see figure [ figure_3 ] right ) . in our model the pure water and the 100@xmath3solar atmospheres have similar net stellar flux . the lower water vapor absorption in the 100@xmath3solar atmosphere compared to the pure water atmosphere is compensated by absorption by other gases , in particular methane . with our model , we obtained net stellar and thermal fluxes similar to those in @xcite for our four different cases . however , the day - night temperature difference for the solar metallicity is a factor of two weaker than in their model . this may be primarily due to the stronger equatorial jet we obtained for this case ( see next paragraph ) . concerning the zonal winds , all simulations lead to the development of an equatorial superrotating jet as expected for strongly irradiated tidally locked exoplanets @xcite . figure [ figure_4 ] shows the zonally averaged zonal wind for our four atmospheric compositions . for the 1@xmath3solar metallicity , there are also two high - latitudes jets . these jets are strongly reduced for the 10@xmath3solar metallicity and vanish for the 100@xmath3solar metallicity . their development is controlled by latitudinal temperature gradients through the thermal wind equation @xcite . with higher metallicity , stellar energy is absorbed higher in the atmosphere leading to weaker latitudinal temperature gradients in the deep atmosphere and so to less significant and shallower jets . in contrast , the depth of the equatorial jet tends to increase with the metallicity leading to stronger dayside - nightside temperature gradients . this behavior ( increase of the equatorial jet and decrease of the high - latitude jets with metallicity ) was also obtained in the simulations of @xcite . compared to the results from @xcite , our gcm produces shallower high - latitude jets for the solar and 10@xmath3solar metallicity cases and no jet for the 100@xmath3solar metallicity . in their model , the high - latitude jets reach a maximum speed between 10@xmath11 and 10@xmath10 bar , while maximum speed occurs at the top of our model ( at 0.03 mbar ) for the solar and 10@xmath3solar metallicity case . therefore , high - latitude jets are stronger at low pressures in our model but weaker at high pressures . for instance , for the solar metallicity , the high - latitude jet speed is around 1100 m / s and 500 m / s at respectively 10@xmath29 and 10@xmath11 bar in our model whereas it is around 900 m / s and 600 m / s at the same pressures in @xcite . concerning the equatorial jet , our model produces a stronger equatorial jet with a wind speed of 1500 m / s at 10@xmath29 bar compared to 900 m / s in @xcite . however , the equatorial jets are very similar between both models for higher metallicity . these differences between both models for h - rich atmospheres may be due to some variations in atmospheric opacities . otherwise , it may be due to the different dynamical cores , based on a standard longitude - latitude grid with a small dissipation in the generic lmdz and on a cubed - sphere grid with no dissipation in @xcite . however , our results globally remain similar to theirs . concerning the pure water atmopshere , our model produces an equatorial jet weaker than for h - rich atmospheres , with a maximum wind speed of around 1 km / s compared to 2 km / s for the others cases . for the water atmosphere , there also are two weak high - latitude jets with a maximum speed of around 500 m / s at 70@xmath12n / s and at 1 mbar . @xcite obtained the same structure and speeds for the zonal wind in the pure water atmosphere . for a tidally locked , synchronously - rotating exoplanet , the heat redistribution from dayside to nightside is mostly done by the superrotating jets . there is also a meridional circulation redistributing heat from low to high latitudes . this is illustrated in figure [ figure_5 ] , which shows - for the 100@xmath3solar metallicity - equatorial temperature and vertical wind as a function of pressure and longitude , as well as maps of temperature and winds at 1.1 mbar ( altitude close to the infrared photosphere ) . the dayside is characterized by upwelling centered near the sub - solar point with poleward winds transporting heat to high latitudes . the dayside poleward circulation crosses poles and is equatorward on the nightside , allowing a stronger heat redistribution . this circulation is shaped by standing rossby and kelvin planetary waves , responsible for the superrotation by pumping eastward momentum from high latitudes to the equator @xcite . winds and temperatures are characterized by a chevron - shape pattern , centered at the equator and pointing to the east @xcite . two strong downdrafts are present at the equator close to the terminators . they are associated with an adiabatic heating explaining the hot spot west of the anti - stellar point in fig . [ figure_5 ] . they are also associated with a shock - like feature with a strong updraft at the substellar and anti - stellar points , also observed in 3d simulations of hot jupiters @xcite . figure [ figure_6 ] shows the mass streamfunction for the 100@xmath3solar metallicity case for the dayside , the nightside and globally ( i.e. dayside+nightside ) . for the dayside , we integrated longitudes between -39@xmath12 and + 90@xmath12 e only , to avoid the very strong downdraft located west of the substellar point . for the nightside , we integrated the mass streamfunction over all the longitudes not included in the dayside integration . the dayside is characterized by a large hadley circulation for pressures below 0.1 bar with two cells extending from the equator to the poles . on the nightside , the mean meridional circulation reverses . this reversal is due to equatorward winds on the nightside and to the two very strong downdrafts located at the equator and close to terminators . these downdrafts are typically twice as strong as the equatorial updrafts ( see figure [ figure_5 ] ) . with an integration between -90@xmath12 and + 90@xmath12 e for the dayside , so including the downdraft located west of the substellar point , we obtained similar circulations but weaker at the equator . when looking at the global mass streamfunction , it appears that the anti - hadley circulation is stronger in the deep atmosphere ( i.e. between 0.1 and 0.01 bar ) than the dayside hadley circulation , but weaker in the upper atmosphere . in reality , the meridional circulation corresponds rather to two large cells crossing the poles and going from the dayside to the nightside but the division into a dayside hadley circulation and a nightside anti - hadley circulation allows an evaluation of which one dominates . these results are for the 100@xmath3solar atmospheric composition , and yet the other atmospheric compositions present the same circulations for pressure lower than around 0.1 bar . figure [ figure_7 ] shows a few contours of the mass streamfunctions averaged zonally over all longitudes and superposed on the zonal mean vertical winds . the mass streamfunctions follow variations of the zonal mean vertical winds . for the 1@xmath3 and 10@xmath3solar metallicity at pressures lower than 0.1 bar , the zonally averaged circulation is characterized by an upwelling at mid - latitudes ( between around 20@xmath12 - 60@xmath12 ) and a downwelling at the equator . the zonal mean circulation is thus dominated by the nightside anti - hadley circulation . the circulations for the 100@xmath3solar metallicity and the pure water atmospheres are more complex , with the anti - hadley circulation dominating at pressures greater than 10 mbar and the hadley circulation above . as detailed before , the dayside hadley circulation becomes stronger than the nightside anti - hadley circulation in the upper atmosphere , explaining this transition . therefore the zonal mean anti - hadley circulation is present in every case but it is located deeper in the atmosphere for higher metallicity . the top panel in figure [ figure_7b ] shows the zonal mean difference between the outgoing longwave radiation ( olr ) and the absorbed stellar radiation ( asr ) . positive values correspond to regions with surplus energy ( compared with the radiative equilibrium ) transported by the atmospheric circulation from regions with a deficit of energy ( negative values ) . globally , the meridional circulation transports heat from low latitudes to high latitudes . however , the anti - hadley circulation limits this poleward transport at low latitudes . the equatorial downdrafts produce an adiabatic heating ( see for instance the equatorial temperature at 150@xmath12e in figure [ figure_5 ] ) . the difference between olr and asr exhibits a local maximum at the equator due to this adiabatic heating . for the 100@xmath3solar metallicity , the meridional circulation contributes to heating at the equator in average . the bottom panel in figure [ figure_7b ] shows the vertically and zonally - integrated heat flux and its decomposition into components of the heat flux transported by mean meridional circulation , stationary waves and transient perturbations . we computed the heat flux from the meridional wind @xmath30 and the dry static energy @xmath31 , and we decomposed the heat flux as : @xmath32=[\overline{e}][\overline{v}]+[\overline{e^*v^*}]+\overline{[e'][v']}\ ] ] where square brackets are time averages and asterisks are deviations from time averages , overbars are zonal averages and primes are deviations from zonal averages . the terms on the right represent , respectively , the transport by the mean meridional circulation , the transport by the stationary waves and the transport by the transient ( time - dependent ) perturbations . for the 100@xmath3solar metallicity case , the total heat transport is strongly impacted by the sationary waves . at a given pressure level , the dry static energy is higher ( both @xmath33 and @xmath34 are higher ) on the dayside with net poleward winds than on the nightside with net equatorward winds . therefore , the stationary waves associated with the day - night contrast produce a net poleward heat transport , counterbalancing the equatorward heat transport associated with the mean anti - hadley circulation below 10 mbar . finally , the anti - hadley circulation we described in this section should be present on a large range of warm tidally locked exoplanets and has a fundamental impact on vertical mixing and cloud transport . we performed an analysis of the strength of the vertical mixing in gj1214b s atmosphere using the generic lmdz gcm with simple tracers . the goal was to understand how the general circulation , and in particular the zonally averaged circulation described in the previous section , controls the transport and distribution of potential cloud or haze particles . from this 3d analysis , equivalent 1d eddy diffusion coefficients were obtained , which may be of use for 1d chemical models . we used a similar method to the analysis of vertical mixing in the hot jupiter hd 209458b by @xcite . we added particles with a fixed diameter and with a constant abundance @xmath35 below 1 bar , with this region corresponding to the source / sink of particles . we let the circulation transport these particles and allowed them to sediment at a terminal velocity described in the next section . contrary to @xcite , we do not assume that particles evaporate on the dayside , which would reduce the downward sedimentation tracer flux . retaining the dayside particles is based on our model dayside temperatures , which are not high enough to significantly evaporate kcl or zns clouds even for the 100@xmath3solar metallicity case . we ran simulations for the different atmospheric compositions with particle radii from 0.1 to 10 microns and with a particle density corresponding to solid kcl ( 2000 kg / m@xmath36 ) , a possible composition for gj1214b s clouds @xcite . we considered micrometric particles because they correspond to the estimated size for possible high cloud / haze particles that would produce a flat transit spectrum in visible and near - infrared @xcite . for the initial state , we assumed a uniform particle abundance everywhere in the atmosphere and equal to @xmath35 ( the constant abundance below 1 bar ) . we used @xmath35=10@xmath37 kg / kg in all simulations . this value is arbitrary and the next figures always show the relative abundance @xmath38 . we computed the sedimentation of particles assuming that they fall at the terminal velocity given by @xcite : @xmath39 where @xmath40 is the particle radius , @xmath41 is the gravitational acceleration of gj1214b , @xmath42 is the particle density ( we chose @xmath42=2000 kg / m@xmath36 corresponding to solid kcl ) , @xmath43 is the atmosphere density , @xmath44 is the viscosity of the atmospheric gas and @xmath45 is the cunnigham slip factor , which describes non - continuum effects . an experimental expression of the cunnigham slip factor is @xcite : @xmath46 where @xmath47 is the knudsen number , equal to the ratio of the mean free path to the size of radius of the particle : @xmath48 the mean free path @xmath49 is given by : @xmath50 with @xmath51 the boltzmann constant , @xmath33 and @xmath52 the temperature and pressure of the gas and @xmath53 the molecular diameter . for the hydrogen - rich atmospheres , we considered that the dynamical viscosity was equal to that of pure h@xmath17 . following @xcite and @xcite , we used the formula : @xmath54 where m is the molecular mass , @xmath53=2.827@xmath55 is the molecular diameter and @xmath56=59.7k@xmath57 k is the lennard - jones potential for h@xmath17 . the ratio on the right in the formula corresponds to the deviation from the hard - sphere model . h@xmath17o is a polar molecule that deviates too strongly from the hard - sphere model to be described by formula ( 8) @xcite . for the water atmosphere , we used a parameterization of the dynamical viscosity of h@xmath17o for dilute conditions and given by @xcite : @xmath58 with @xmath59= 10@xmath60 pa / s , @xmath61= 647.27 k , @xmath62= 0.0181583 , @xmath63= 0.0177624 , @xmath64= 0.0105287 and @xmath65= -0.0036744 . this parameterization is valid for temperatures ranging from around 300 k to 1100 k. the dilute approximation is valid up to 10 bars for gj1214b conditions ( see @xcite for an expression with correction terms to the dilute approximation ) , and is thus compatible with the pressure range we considered for vertical mixing . for the calculation of @xmath47 for the water - rich atmosphere , we used a molecular diameter @xmath66 depending on temperature and given by @xcite : @xmath67 the top panel in figure [ figure_8 ] shows the terminal velocity for particle radii from 0.1 to 10 microns , using @xmath42=2000 kg m@xmath29 ( similar to the density of kcl ) and for the 100@xmath3solar and the pure water atmospheric composition cases ( temperature profiles from the 1d model ) . for both compositions , two regimes are present . at high pressure , the knudsen number is smaller than 1 . the cunningham slip factor and thus the terminal velocity are almost constant . at low pressure , the knudsen number is much larger than 1 . the cunningham slip factor and thus the terminal velocity are inversely proportional to the pressure . for a given particle radius , the terminal velocity and the corresponding sedimentation rate are always lower in the h@xmath17o atmosphere than in the h - rich atmosphere . in the upper atmosphere , the terminal velocities are lower by around a factor 3 . this is due to both the larger molecular diameter and the larger viscosity of h@xmath17o compared to h@xmath17 . the sedimentation timescale is defined as the ratio of the scale height to the sedimentation velocity @xmath68 . the bottom panel in figure [ figure_8 ] shows the sedimentation timescales for the 100@xmath3solar and the pure water atmosphere . the vertical mixing simulation is valid when the duration of the simulation is longer than the sedimentation timescale . since our simulations were run for around 1600 days , the vertical mixing simulations are valid for pressures lower than 0.1 bar for a particle radius of 1 micron , and for pressure lower than 1 bar for particle radii of 3 - 10 micron . for evaluating the vertical mixing of the tracers , it is useful to perform simulations with different particle radii . large particles provide the most robust results at higher atmospheric pressures , but can be completely removed from the upper atmosphere , and so smaller particles give more accurate estimates for vertical mixing at these levels . figure [ figure_9 ] shows the ratio ( in log scale ) of the sedimentation timescale to the advection timescale ( @xmath69 with @xmath70 gj1214b s radius and @xmath71 the mean zonal wind ) for the 100@xmath3solar case . this ratio is very similar for the different atmospheric compositions . when the sedimentation timescale is longer than the advection timescale ( i.e. particles fall by less than a scale height during a full rotation around the planet ) , the particle transport is mostly driven by the zonal mean circulation and there are small longitudinal variations in the particle distribution . according to figure [ figure_9 ] , the particle distribution is mostly controlled by the zonal mean circulation and there should be almost no variations in longitude for pressures higher than 1 mbar . moreover , longitudinal variations should be enhanced at high latitude . figure [ figure_10 ] shows the global mean particle abundance as a function of pressure for the different atmospheric compositions for a particle radius of 1 micron and for the 100@xmath3solar metallicity case for particle radii from 0.1 to 10 micron . the particle abundance in the upper atmosphere increases with atmospheric metallicity . the abundance is reduced by 50@xmath28 at around 5 mbar for the solar and 10@xmath3solar metallicity cases , and at 0.2 - 0.4 mbar for the other cases . the strength of the atmospheric circulation above 0.1 bar increases with metallicity . for the solar metallicity case , the circulation can not efficiently transport micrometer particles to the upper atmosphere ( 0.1 - 0.01 mbar ) . for the 100@xmath3solar metallicity , the vertical mixing is weak between 1 bar and 0.1 bar and very strong between 0.1 bar and 1 mbar . particles with a radius of 0.1 micron are well mixed above the particle source layer while particles with a radius of 10 microns can not go higher than 0.5 bar . figure [ figure_11 ] shows the zonal mean tracer abundance for a particle radius of 1 micron . the black lines correspond to the 50@xmath28 abundance contours . tracers reach the highest altitudes at mid - latitude ( at around 40@xmath12n / s ) . except at low pressure for high metallicity ( i.e. 100@xmath3solar and the pure water atmosphere ) , there is a minimum of particle abundance at the equator . as explained before , the particle distribution is mostly controlled by the zonal mean circulation with in particular the anti - hadley circulation ( see figure [ figure_7 ] ) . on average , particles are thus lofted at mid - latitudes and sediment at the equator . this behavior on a synchronously rotating planet is opposite to that for a non - synchronously rotating planet like the earth , for which there is an upward transport of water vapor and chemical species at the itcz ( inter - tropical convergence zone ) , and thus a maximum of clouds at the equator . figure [ figure_12 ] shows maps of particle abundance at 12 , 1.1 and 0.1 mbar . as discussed in the section 4.1 , there is almost no longitudinal variation at pressures higher than 1 mbar . the zonal wind spatially homogenizes the particle distribution . at pressures lower than 1 mbar , strong longitudinal variations appear . these variations are controlled by local upward / downward particle fluxes and also by meridional winds , producing more complex distributions . in particular , in the upper atmosphere , the meridional circulation tends to transport particles poleward . this is clear in figure [ figure_12 ] for h@xmath17o and the 100@xmath3solar metallicity case at 0.1 - 1 mbar . the poleward transport is mostly due to the stationary waves , similar to the heat transport in figure [ figure_7b ] . indeed , at low pressures , the dayside - nightside contrast is strong and stationnary waves dominate the transport . the particle abundance is higher in the dayside , where there is upwelling , than in the nightside , where there is downwelling . the poleward wind in the dayside and the equatorward wind in the nightside , therefore produce a net poleward particles flux . this meridional transport maintains a high particle abundance at high latitudes in the upper atmosphere , and the maximum of particle abundance migrates poleward with altitude ( see figure [ figure_11 ] ) . most studies concerning the photochemistry or the cloud formation on exoplanets are based on 1d models that take into account the vertical mixing by using a single eddy diffusion coefficient @xmath72 to represent the global 3d mixing . in these studies , the values used for the eddy diffusion coefficient are quite arbitrary , exploring a range of values and generally without altitude dependence . however , our 3d model results can be used to provide eddy diffusion coefficients for use by 1d models . at steady state , the downward particle flux by sedimentation is compensated by the upward vertical mixing . thus the particle abundance @xmath73 ( in kg / kg or mol / mol ) in 1d is given by : @xmath74 with the outputs of the gcm , an equivalent 1d @xmath72 for the global 3d mixing can therefore be defined as @xcite : @xmath75 where the brackets correspond to a horizontal and time average over the entire planet . defining the diffusive time scale @xmath76 and the sedimentation time scale @xmath77 ( both depending on altitude ) , we can also write equation ( 11 ) in pressure coordinate as @xcite : @xmath78 if we assume 1 ) a simple form for the dependence of @xmath72 on pressure , @xmath72=@xmath79 with @xmath80 0 and @xmath801 ( in our simulation @xmath81 @xmath82 0.4 ) , 2 ) an isothermal atmosphere and 3 ) @xmath45=@xmath83 for the terminal velocity in equation [ eq_beta ] ( corresponding to a regime where the sedimentation speed is proportional to the pressure ) , then the particle abundance @xmath73 at a pressure @xmath52 can be expressed in function of @xmath35 at a pressure @xmath84 as ( see also @xcite ) : @xmath85 this relation can match quite well the mean tracer abundance in figure [ figure_10 ] , for instance for the 100@xmath3solar metallicity using @xmath86 m@xmath1/s , @xmath84=1 bar , @xmath81=0.4 and @xmath35=1 . figure [ figure_13 ] shows @xmath72 obtained from the 3d simulations using equation ( 12 ) . the left panel corresponds to the 100@xmath3solar metallicity case . @xmath72 was computed for particle radii from 0.1 to 3 microns ( red lines ) and is similar for these different radii . for the dependence on pressure , a good parametrization is @xmath87 m@xmath1/s . we also compared @xmath72 obtained from the gcm to previous estimates using 1 ) the simple formula @xmath88 where @xmath89 is the root mean square of the vertical wind and @xmath90 is the mixing length assumed to be equal to the scale height @xmath91 @xcite , and 2 ) the following formula for thermal vertical mixing from @xcite and used in the 1d cloud model of @xcite : @xmath92 where @xmath70 is the perfect gas constant , @xmath8 is the mean molecular weight , @xmath93 is the atmospheric density , @xmath94 is the specific heat and @xmath95 is the convective heat flux . the latter is computed from our gcm by subtracting the mean net thermal flux by the mean net stellar flux ( figure [ figure_3 ] ) . the factor 1/3 on the left is empirical , derived from observations of giant planets in our own solar system . this formula was also used by @xcite for clouds on gj1214b . the equivalent 1d @xmath72 from our model is generally one order of magnitude lower than these two estimations . the mixing length has to be lower than the scale height , as predicted by @xcite . a much better estimate is obtained for both formulae using a mixing length 10 times lower than the scale height . since 1d simulations are generally at radiative equilibrium ( @xmath95=0 ) , @xcite assumed @xmath96 for the convective heat flux . while such an estimation can be quite good for the convective region of exoplanets or brown dwarfs , it strongly overestimates the convective heat flux for irradiated planets . according to our gcm , @xmath95 varies between 1 to 100 w / m@xmath1 . the use of @xmath96 overestimates @xmath72 from formula ( 15 ) by one order of magnitude . therefore , the vertical mixing is likely overestimated by one order of magnitude in @xcite . this will tend to overestimate the particle abundance and size at high altitude in their model . consequently lofting larger particles to higher altitudes to explain gj1214b s flat spectrum may be more challenging than their results suggest . such simple parametrizations of @xmath72 should therefore be used with caution since they give a very rough estimate of the mixing . however , they can give a good trend of the evolution of @xmath72 with pressure . in particular , the variations with pressure of @xmath72 computed with our gcm follow quite well those from formula ( 15 ) . even if the atmosphere is never convective in our gcm , we can expect that the global circulation , driven by the horizontal temperature gradient , globally acts as convection for the vertical mixing . the equivalent convective flux @xmath95 in our simulations slowly decreases from the altitude where most of the stellar flux is absorbed . for the 100@xmath3solar case , it is more or less constant from 0.1 bar to 0.1 mbar . since @xmath97 , equation ( 15 ) gives @xmath98 . @xcite found @xmath99 , while in our model the best fit is with -0.4 as the exponent . in fact , an exponent of -0.4 or -1/3 can also provide a good fit to their values . the vertical mixing in planetary atmospheres is mostly done by molecular diffusion above the homopause and by waves and convection below . on earth , the vertical mixing is primarily done by convection ( dry and moist ) in the troposphere and by gravity waves in the stratosphere . the amplitude of the latter grows as @xmath100 until they break in the mesosphere @xcite . in warm exoplanets , the mixing should mostly be done by both largescale upwelling / downwelling ( acting as convection ) and planetary waves , which are well simulated by gcms and generally grow as @xmath100 . we can therefore expect for such atmospheres an eddy diffusion coefficient with an exponent between -1/3 and -1/2 for the pressure dependence , justifying the intermediate exponent used in our fit . the right panel in figure [ figure_13 ] shows the equivalent 1d @xmath72 for the different atmospheric compositions . it decreases with metallicity in the deeper atmosphere ( i.e. at around 1 bar ) and increases with metallicity above 0.1 bar . this is explained by the difference in altitude where most of the stellar energy is deposited ( i.e. below 100 mbar for the solar metallicity case and at around 10 mbar for the 100@xmath3solar metallicity case ) . the equivalent 1d eddy diffusion coefficients can be parametrized as @xmath72=@xmath101 ( p in bar ) with @xmath102=@xmath103 , @xmath104 , @xmath105 , @xmath106 m@xmath1/s for the 1 , 10 , 100@xmath3solar metallicity and pure water case respectively . the parameters in our fits were chosen to primarily match @xmath72 and the mean tracer abundance ( fig . [ figure_10 ] ) for pressures lower than 10 mbar . at higher pressures , there are quite strong deviations from this simple power - law dependence , in particular for the pure water case . for the h - dominated atmosphere , we find that @xmath72 increases with metallicity . a higher metallicity leads to stronger dayside / nightside temperature contrasts and stronger vertical motions producing a stronger vertical mixing . @xmath72 for the pure water case is lower than for h - rich atmospheres except between 0.1 - 0.01 bar . however , the pure water case always leads to a higher particle abundance because of its smaller scale height and the weaker sedimentation velocity . the atmospheric circulation and the vertical mixing we described in the last section have strong implications for the formation of high clouds / haze on gj1214b and other tidally locked exoplanets . the circulation patterns we obtained , and in particular the anti - hadley cell , are very general and can be applied to a wide range of warm synchronously rotating exoplanets , from mini - neptunes to jupiter - mass exoplanets . the temperature profile in the atmosphere of gj1214b allows the condensation of potassium chloride ( kcl ) and zinc sulfide ( zns ) @xcite . according to our simulations for a h - rich atmosphere , the condensation should occur between 40 mbar and 0.4 bar ( see section 3.1 . ) . if the atmospheric circulation can transport cloud particles to the upper atmosphere ( i.e. 0.1 mbar and less ) , they should remain condensed ( in solid phase ) and could produce the observed flat transit spectrum @xcite . micrometric - size particles ( or bigger ) are required to produce a flat spectrum between 1.1 and 1.7 micron . transporting such particles from around 0.1 bar to 0.01 mbar requires a strong circulation that can be accurately quantified only with gcms . our 3d simulations with simple tracers can represent quite well clouds that form deeper in the atmosphere and do not evaporate at higher altitudes on the dayside . they reveal that micrometer particles can be lofted to the upper atmosphere quite efficiently . for instance , the abundance is reduced by around a factor 5 at 0.1 mbar for the 100@xmath3solar metallicity with a particle radius of 1 @xmath8 m . the vertical mixing tends to increase with metallicity . therefore , metallicity impacts the formation of clouds not only by fixing the amount of condensable species but also by controlling the vertical mixing . a high metallicity should therefore favor the formation of high clouds on gj1214b . assuming the solar nebula atomic abundances from @xcite , the abundance of kcl vapor in the deep atmosphere would be around 8.3@xmath310@xmath60kg / kg for the solar metallicity and around 4.3@xmath310@xmath107kg / kg for the 100@xmath3solar metallicity . using the mean profiles of tracer relative abundances for particle radii of 1 @xmath8 m ( fig . [ figure_10 ] ) , we find that kcl clouds would be optically thick in transit at visible and near - infrared wavelength at pressures higher than 3 mbar for the solar metallicity and at pressures higher than 0.1 mbar for the 100@xmath3solar metallicity . therefore the latter case is more likely to produce the observed flat transit spectrum than the former case . clouds could also modify the circulation and the vertical mixing by absorbing and scattering stellar radiation . if clouds are strongly scattering , we can expect a high planetary albedo and in particular a cooling of the atmosphere below the cloud deck ( i.e. below 0.1 - 1 bar ) impacting the atmospheric circulation . the impact of this cloud - climate feedback on the possibility of high clouds on gj1214b is best understood using a gcm that can take into account the radiative effect of clouds . finally , our gcm predicts a minimum of tracers at the equator for pressures higher than 0.1 mbar for all cases considered here , and we expect the cloud distribution to have the same behavior . a minimum of cloud at the equator could have observable impacts on transmission spectra and also on reflection or emission spectra of the planet . another possible explanation for the flat transit spectrum of gj1214b is the formation of photochemical haze , possibly produced by titan - like methane photolysis . such a process can be more efficient than cloud condensation at producing high aerosols because particles are produced directly in the upper atmosphere . however , sedimentation may be too strong to maintain a particle density large enough for blocking stellar light . even on titan , which is very hazy and has a low gravity ( with correspondingly small sedimentation velocities ) , the haze is not opaque enough to block all methane absorption bands in near - infrared transit spectra @xcite . therefore gj1214b s flat spectrum can only be produced by photochemical haze if there is a large methane flux counter - balancing the haze sedimentation flux . based on the atmospheric circulation simulations , we predict that methane would be supplied at mid - latitudes ( see figure [ figure_7 ] ) . once produced , haze would be removed mostly at the poles and at the equator . this would impact the hazes latitudinal distribution . haze particles should fall until the temperature is high enough to pyrolyze them , mostly into methane . the huygens probe aerosol collector and pyrolyser ( acp ) experiment showed that titan s hydrocarbon haze was completely pyrolized at temperatures close to 900k @xcite . in our simulations of gj1214b , that temperature corresponds to pressure of around 0.1 bar for high metallicity and 1 bar for solar metallicity ( see fig . [ figure_2 ] ) . therefore , there could be a complete methane cycle for pressures lower than these values , that could recycle haze into methane . @xcite and @xcite explored the photochemistry in gj1214b with a 1d model using eddy diffusion coefficient values from 10@xmath108 to 10@xmath109 cm@xmath1/s and predicted that more methane and more haze would be present at high altitude with a high @xmath72 ( i.e. 10@xmath109 cm@xmath1/s ) . with our model , we predict eddy diffusion coefficient values of around 3@xmath310@xmath16 cm@xmath1/s at 1 bar to 8@xmath310@xmath109 cm@xmath1/s at 10@xmath60 bar for the 100@xmath3solar metallicity case . the photolysis of methane , which occurs mostly above 10@xmath110 bar , is thus better simulated with @xmath72=10@xmath109 cm@xmath1/s in @xcite and @xcite . moreover the radiative effect of haze , which can absorb stellar radiation , could increase the vertical mixing in the upper atmosphere and support the lofting of aerosol particles , increasing the lifetime of the haze , and this feedback is best studied with a gcm . we conclude that the formation of high haze on gj1214b is likely and it is also possible that both high condensate clouds and photochemical haze are formed . in this study , we analyzed the atmospheric dynamics of gj1214b for a cloud - free atmosphere with different metallicities as a first application of the generic lmdz gcm to a non - rocky exoplanet . we obtained results for radiative transfer , temperatures and winds very similar to those from the other state - of - the - art gcm @xcite , validating our model for this kind of planet . in addition to this intercomparison , we showed that the zonal mean circulation of gj1214 is characterized by the presence of an anti - hadley circulation on the nightside , which leads to zonally - averaged subsidence at the equator . this particular regime likely occurs on a wide range of synchronously rotating exoplanets , from warm mini - neptunes to hot jupiters . using simple tracers , we analyzed the vertical mixing in the atmosphere of gj1214b . the tracers showed that atmospheric transport is primarily driven by the zonal mean circulation , leading to an upwelling at mid - latitudes and a downwelling at the equator . in particular , a minimum of tracer abundance appears at the equator , strengthened in the upper atmosphere by a poleward meridional transport . this should be a fundamental feature of the cloud / haze distribution on many synchronously rotating warm exoplanets . we also found that the vertical mixing increases with metallicity for h - rich atmospheres . for a solar metallicity , the circulation can not loft micrometer particles into the upper atmosphere . therefore , if the upper atmosphere of gj1214b is opaque because of condensate clouds , the atmospheric metallicity must be higher than solar and likely more than 10@xmath3solar . from these simulations , we derived equivalent 1d eddy coefficients . we found that : @xmath0 m@xmath1/s is a good fit for 1@xmath3solar metallicity and @xmath2 m@xmath1/s for 100@xmath3solar metallicity . we compared these values to classical formulae for @xmath72 used for instance in @xcite , showing how these simple formulae can overestimate the mixing by one order of magnitude . our parametrizations of @xmath72 can be used in 1d cloud or photochemical models @xcite for which they are expected to lead to an efficient formation of photochemical haze . finally , we showed that the atmospheric circulation of gj1214b could be strong enough to loft micrometric particles of kcl or zns from 1 bar to 0.1 - 0.01 mbar . however , the radiative effects of clouds , by absorbing or diffusing stellar radiation , could modify the strength of the circulation . the next step will be to simulate the atmosphere of gj1214b with realistic kcl and zns clouds , taking into account the latent heat release and the radiative effects . for this next study , the generic lmdz gcm benefits from all the developments done for previous studies of cloudy rocky planets . such 3d simulations would give strong indications of the possibility of high condensate clouds producing the observed flat transit spectrum . such simulations could also help identify the best observations for probing cloudy atmospheres , with , for example , emission / reflection spectra and phase curves . b.c . acknowledges support from an appointment to the nasa postdoctoral program at nai virtual planetary laboratory , administered by oak ridge affiliated universities . vsm is supported by the nasa astrobiology institute s virtual planetary laboratory lead team , under the national aeronautics and space administration solicitation nnh12zda002c and cooperative agreement number nna13aa93a . this work was facilitated though the use of advanced computational , storage , and networking infrastructure provided by the hyak supercomputer system at the university of washington . solar ( black ) metallicity . right panel : maximum dayside - nightside temperature difference as a function of pressure . the differences are computed at each latitude and pressure level . the figure shows their maximum value at a given pressure.,title="fig:",width=264 ] solar ( black ) metallicity . right panel : maximum dayside - nightside temperature difference as a function of pressure . the differences are computed at each latitude and pressure level . the figure shows their maximum value at a given pressure.,title="fig:",width=264 ] ( right ) as a function of pressure . solid lines correspond to the heating rate and the stellar net flux . dashed lines correspond to the cooling rate and the thermal net flux . , title="fig:",width=264 ] ( right ) as a function of pressure . solid lines correspond to the heating rate and the stellar net flux . dashed lines correspond to the cooling rate and the thermal net flux . , title="fig:",width=264 ] solar metallicity . the top panels show the temperature ( left ) and vertical velocity ( right ) at the equator versus longitude and pressure . the bottom panels show the map of temperature ( left ) and vertical velocity ( right ) at 1.1 mbar . black vectors correspond to the directions of horizontal winds.,title="fig:",width=264 ] solar metallicity . the top panels show the temperature ( left ) and vertical velocity ( right ) at the equator versus longitude and pressure . the bottom panels show the map of temperature ( left ) and vertical velocity ( right ) at 1.1 mbar . black vectors correspond to the directions of horizontal winds.,title="fig:",width=264 ] solar metallicity . the top panels show the temperature ( left ) and vertical velocity ( right ) at the equator versus longitude and pressure . the bottom panels show the map of temperature ( left ) and vertical velocity ( right ) at 1.1 mbar . black vectors correspond to the directions of horizontal winds.,title="fig:",width=264 ] solar metallicity . the top panels show the temperature ( left ) and vertical velocity ( right ) at the equator versus longitude and pressure . the bottom panels show the map of temperature ( left ) and vertical velocity ( right ) at 1.1 mbar . black vectors correspond to the directions of horizontal winds.,title="fig:",width=264 ] solar metallicity case for the dayside ( top ) , the nightside ( middle ) and globally ( bottom ) . positive ( negative ) values correspond to anti - clockwise ( clockwise ) circulation . dashed ( solid ) lines are contours for the value of @xmath410@xmath111 kg / m / s . the dayside streamfunction was calculated by integrating longitudes between -39@xmath12 and + 90@xmath12e . the nightside streamfunction was calculated by integrating the other longitudes ( -180 @xmath12 to -39@xmath12e and + 90@xmath12 to + 180@xmath12 e ) . the global mass streamfunction was calculated by integrating all longitudes and is equal to the sum of the dayside and nightside streamfunctions.,title="fig:",width=264 ] + solar metallicity case for the dayside ( top ) , the nightside ( middle ) and globally ( bottom ) . positive ( negative ) values correspond to anti - clockwise ( clockwise ) circulation . dashed ( solid ) lines are contours for the value of @xmath410@xmath111 kg / m / s . the dayside streamfunction was calculated by integrating longitudes between -39@xmath12 and + 90@xmath12e . the nightside streamfunction was calculated by integrating the other longitudes ( -180 @xmath12 to -39@xmath12e and + 90@xmath12 to + 180@xmath12 e ) . the global mass streamfunction was calculated by integrating all longitudes and is equal to the sum of the dayside and nightside streamfunctions.,title="fig:",width=264 ] + solar metallicity case for the dayside ( top ) , the nightside ( middle ) and globally ( bottom ) . positive ( negative ) values correspond to anti - clockwise ( clockwise ) circulation . dashed ( solid ) lines are contours for the value of @xmath410@xmath111 kg / m / s . the dayside streamfunction was calculated by integrating longitudes between -39@xmath12 and + 90@xmath12e . the nightside streamfunction was calculated by integrating the other longitudes ( -180 @xmath12 to -39@xmath12e and + 90@xmath12 to + 180@xmath12 e ) . the global mass streamfunction was calculated by integrating all longitudes and is equal to the sum of the dayside and nightside streamfunctions.,title="fig:",width=264 ] solar metallicity case into heat transport by the mean meridional circulation ( black dashed line ) , by stationary waves ( black dashed - dotted line ) and by transient perturbations ( black dotted line ) . positive ( negative ) values correspond to northward ( southward ) heat fluxes.,title="fig:",width=377 ] solar metallicity case into heat transport by the mean meridional circulation ( black dashed line ) , by stationary waves ( black dashed - dotted line ) and by transient perturbations ( black dotted line ) . positive ( negative ) values correspond to northward ( southward ) heat fluxes.,title="fig:",width=377 ] solar metallicity with particle radii from 0.1 to 10 microns . for both figures , the relative abundance is fixed to 1 below 1 bar . the dashed lines correspond to the analytical profiles using a parametrized @xmath72 ( see section 4.4 ) . , title="fig:",width=264 ] solar metallicity with particle radii from 0.1 to 10 microns . for both figures , the relative abundance is fixed to 1 below 1 bar . the dashed lines correspond to the analytical profiles using a parametrized @xmath72 ( see section 4.4 ) . , title="fig:",width=264 ] , 10@xmath3 , 100@xmath3solar and pure h@xmath17o from top to down ) and for pressure of 12 mbar ( left ) 1.1 mbar ( middle ) and 0.1 mbar ( right).,title="fig:",width=188 ] , 10@xmath3 , 100@xmath3solar and pure h@xmath17o from top to down ) and for pressure of 12 mbar ( left ) 1.1 mbar ( middle ) and 0.1 mbar ( right).,title="fig:",width=188 ] , 10@xmath3 , 100@xmath3solar and pure h@xmath17o from top to down ) and for pressure of 12 mbar ( left ) 1.1 mbar ( middle ) and 0.1 mbar ( right).,title="fig:",width=188 ] , 10@xmath3 , 100@xmath3solar and pure h@xmath17o from top to down ) and for pressure of 12 mbar ( left ) 1.1 mbar ( middle ) and 0.1 mbar ( right).,title="fig:",width=188 ] , 10@xmath3 , 100@xmath3solar and pure h@xmath17o from top to down ) and for pressure of 12 mbar ( left ) 1.1 mbar ( middle ) and 0.1 mbar ( right).,title="fig:",width=188 ] , 10@xmath3 , 100@xmath3solar and pure h@xmath17o from top to down ) and for pressure of 12 mbar ( left ) 1.1 mbar ( middle ) and 0.1 mbar ( right).,title="fig:",width=188 ] , 10@xmath3 , 100@xmath3solar and pure h@xmath17o from top to down ) and for pressure of 12 mbar ( left ) 1.1 mbar ( middle ) and 0.1 mbar ( right).,title="fig:",width=188 ] , 10@xmath3 , 100@xmath3solar and pure h@xmath17o from top to down ) and for pressure of 12 mbar ( left ) 1.1 mbar ( middle ) and 0.1 mbar ( right).,title="fig:",width=188 ] , 10@xmath3 , 100@xmath3solar and pure h@xmath17o from top to down ) and for pressure of 12 mbar ( left ) 1.1 mbar ( middle ) and 0.1 mbar ( right).,title="fig:",width=188 ] , 10@xmath3 , 100@xmath3solar and pure h@xmath17o from top to down ) and for pressure of 12 mbar ( left ) 1.1 mbar ( middle ) and 0.1 mbar ( right).,title="fig:",width=188 ] , 10@xmath3 , 100@xmath3solar and pure h@xmath17o from top to down ) and for pressure of 12 mbar ( left ) 1.1 mbar ( middle ) and 0.1 mbar ( right).,title="fig:",width=188 ] , 10@xmath3 , 100@xmath3solar and pure h@xmath17o from top to down ) and for pressure of 12 mbar ( left ) 1.1 mbar ( middle ) and 0.1 mbar ( right).,title="fig:",width=188 ] for the 100@xmath3solar metallicity case . red lines are from the gcm for particle radii of 0.1 , 0.3 , 1 , 3 microns . the blue line is the simple fit ( @xmath112 @xmath3p@xmath113@xmath114 m@xmath1/s ) . solid and dashed black lines are estimations from simple formula @xcite . the right panel shows k@xmath115 derived from the gcm for the different atmospheric compositions with particle radius of 1 micron . the dashed lines are the simple fits ( @xmath72=@xmath116 with @xmath102=@xmath103 , @xmath104 , @xmath105 , @xmath106 m@xmath1/s for the 1 , 10 , 100@xmath3solar metallicity and pure water case respectively ) . , title="fig:",width=302 ] for the 100@xmath3solar metallicity case . red lines are from the gcm for particle radii of 0.1 , 0.3 , 1 , 3 microns . the blue line is the simple fit ( @xmath112 @xmath3p@xmath113@xmath114 m@xmath1/s ) . solid and dashed black lines are estimations from simple formula @xcite . the right panel shows k@xmath115 derived from the gcm for the different atmospheric compositions with particle radius of 1 micron . the dashed lines are the simple fits ( @xmath72=@xmath116 with @xmath102=@xmath103 , @xmath104 , @xmath105 , @xmath106 m@xmath1/s for the 1 , 10 , 100@xmath3solar metallicity and pure water case respectively ) . ,

we unveil in concrete terms the general machinery of the syzygy - based algorithms for the implicitization of rational surfaces in terms of the monomials in the polynomials defining the parametrization , following and expanding our joint article with m. dohm . these algebraic techniques , based on the theory of approximation complexes due to j. herzog , a , simis and w. vasconcelos , were introduced for the implicitization problem by j .- p . jouanolou , l. bus , and m. chardin . their work was inspired by the practical method of moving curves , proposed by t. sederberg and f. chen , translated into the language of syzygies by d. cox . our aim is to express the theoretical results and resulting algorithms into very concrete terms , avoiding the use of the advanced homological commutative algebra tools which are needed for their proofs .

1502.00890

let @xmath0 be a field . we can assume @xmath1 ( or any computable field ) when dealing with implementations . all the varieties , rings and vector spaces we will consider are understood to be taken over @xmath0 . consider a rational parametrization @xmath2 of a ( hyper)surface @xmath3 , where @xmath4 $ ] is a non - constant polynomial and @xmath5 $ ] . ( as usual , the dashed arrow means that @xmath6 is defined on a dense open set of @xmath7 . ) an important problem in computer aided geometric design is to switch from parametric to implicit representations of rational surfaces @xcite , that is the _ parametrization _ @xmath6 is assumed to be known and one seeks for the _ implicit equation _ @xmath8 ( which is defined only up to multiplicative constant ) . in fact , we will assume that @xmath6 is given and our aim will not be to get the implicit equation @xmath8 of @xmath9 written in terms of its monomials , but a _ matrix representation _ of the surface . [ def : matrep ] a matrix representation @xmath10 of @xmath9 is a matrix with entries in @xmath11 $ ] , generically of full rank , which verifies the following condition : for any point @xmath12 , the rank of @xmath13 drops if and only if @xmath14 lies on @xmath9 . the use of matrix representations goes back to manocha and canny @xcite , and to chionh and goldman @xcite . having the matrix @xmath10 is sufficiently good for many purposes and it is cheaper to compute . the well - developed theory and tools of linear algebra can be applied to solve geometric problems . we can certainly use the ( numerical ) rank dropping condition in definition [ def : matrep ] to check membership in @xmath9 , and , moreover , the whole structure of minors of @xmath10 is related to the singularities of the parametrization @xcite and gives a way to invert it when the fiber has a single point @xcite . matrix representations are also useful for solving intersection problems as is shown in @xcite . much of the computational difficulty in these problems lies on computing ranks for polynomial matrices ( cf . @xcite as well as section 5 in the nice and interesting paper @xcite ) . the motivation for this paper is to present in the simplest possible terms procedures for the implicitization of rational surfaces via matrix representations , based on the syzygies @xmath15 of the input polynomials , that is , 4-tuples of polynomials in the @xmath16 variables verifying the linear relation @xmath17 . the theoretical justification is not naive and requires a good command of techniques of ( homological ) commutative algebra . however , the algorithms do not require a heavy background and are easy to explain . we will show that they perform very well , and moreover , they work even better in the presence of base points . call @xmath18 the coordinates in the target of @xmath6 . our question is an instance of elimination of variables , where we want to find the algebraic relations among the variables @xmath18 under the assumption that @xmath19 , for some @xmath16 in the domain of @xmath6 . the eliminant polynomial by excellence is the determinant @xmath20 , a polynomial with integer coordinates on the coefficients of a square matrix @xmath21 , which vanishes on those coefficients for which there exists a nonzero solution @xmath22 to the equations @xmath23 . elimination of variables is done in the literature through different incarnations of the following general strategy : 1 . reduce the problem to a linear algebra problem . 2 . hide the variables one wants to eliminate in the ( typically monomial ) bases . 3 . use determinants . this strategy is also the core in our syzygy - based algorithms . the following short account of the approach of the use of syzygies in our context is reconstructed from david cox s lecture at the conference pasi on commutative algebra and its connections to geometry honoring wolmer vasconcelos , held in brazil in 2009 ( * ? ? ? * mini course 1 ) . the use of syzygies for the implicitization of ( conic ) surfaces goes back to steiner in 1832 @xcite . in 1887 , meyer describes in @xcite syzygies of three polynomials and makes a general conjecture proved by hilbert in 1890 @xcite . surface implicitization by eliminating parameters was studied by salmon in 1862 @xcite and dixon in 1908 using resultants @xcite . in 1995 , sederberg and chen reintroduced the use of syzygies , by a method termed as _ moving curves and surfaces _ @xcite . cox realized they were using syzygies @xcite , and produced several papers with other coauthors ( bus , chen , dandrea , goldman , sederberg , zhang @xcite ) . in 2002 , jouanolou and bus @xcite abstracted and generalized on a sound basis the method of sederberg - chen via approximation complexes , a tool in homological commutative algebra that had been developed by herzog , simis and vasconcelos @xcite . bus , chardin , jouanolou and simis produced further advances in the homogeneous case @xcite . goldman et al . studied the cases of planar and space curves @xcite . a generalization of the linear syzygy method when the support of the input polynomials is a square ( that is , bihomogeneous of degree @xmath24 ) was proposed by bus and dohm @xcite , and for any polygon by botbol , dickenstein and dohm @xcite , and botbol @xcite . this method is particularly adapted when the polynomials defining the parametrization are _ sparse _ , which is often the case . this will be our point of view in this article . so , we want to solve the following problem . [ pb : main ] given a rational parametrization @xmath6 as in , find a matrix representation @xmath10 of the surface @xmath9 by means of syzygies and the monomial structure of @xmath25 . the main general algorithmic answer to this problem is given in algorithm [ algo : mainalgo ] ( see theorem [ th : mainalgo ] ) . our assumption that the dimension of @xmath9 is @xmath26 is equivalent to the fact , when we extend the map to the algebraic closure @xmath27 of @xmath0 , that for almost all @xmath28 in the image of @xmath6 , the number of preimages by @xmath6 is finite . this number is called the degree of @xmath6 and noted @xmath29 . the matrix representations @xmath10 of @xmath9 provided by algorithms [ algo : mainalgo ] and [ algo : hirzebruch ] moreover satisfy that the greatest common divisor of all minors of @xmath10 of maximal size equals @xmath30 . we present in section [ sec : naive ] the first naive linear algebra algorithm to compute the implicit equation @xmath8 , which requires to solve a huge linear algebra system . moreover , this naive method `` forgets '' the parametrization and thus in general it is not useful in computer aided geometric design . in section [ sec : syzygies ] we recall previous results on the implicitization of curves and surfaces using syzygies and present our general methods of implicitization via linear syzygies , which requires to solve considerably smaller systems . we highlight in section [ ss : curves ] the main elimination step , which was termed _ instant elimination _ in @xcite ( see also the references therein ) . in section [ ssec : bihom ] we present in theorem [ th : hirzebruch ] a refinement of theorem [ th : mainalgo ] for bihomogeneous parametrizations , in the same spirit . technicalities are avoided in our presentation in these sections , and in particular in the statement of our main results theorems [ th : mainalgo ] and [ th : hirzebruch ] . detailed hypotheses and proofs are deferred to section [ sec : tgtools ] , where we introduce the necessary background on toric geometry . we collect in appendix [ sec : catools ] a general overview of the rationale of the tools and results from homological commutative algebra required for the proofs . a reader only interested in the application of our results , can skip these two sections . section [ sec : examples ] illustrates the practicality and advantages of our approach . for our computations , we use implementations in macaulay 2 , which need different type of homogenizations to use current routines ( via a toric embedding or a multihomogenization via an abstract toric cox ring ) @xcite . for the best performance of our algorithms , it would be important to design ad - hoc structured linear algebra strategies to compute syzygies in the sparse case . the convex hull in @xmath31 of the exponents of the monomials occurring in a non zero ( laurent ) polynomial @xmath32 in @xmath33 variables is called the newton polytope @xmath34 of @xmath32 . when @xmath32 is a polynomial in @xmath35 of degree ( at most ) @xmath36 , its newton polygon @xmath37 is contained in the triangle @xmath38 with vertices @xmath39 . the euclidean area @xmath40 of this triangle is @xmath41 and its lattice area @xmath42 is equal to @xmath43 , which is always an integer . we have the following classical result ( c.f . for instance @xcite ) : for generic polynomials @xmath44 of degree @xmath36 , the degree of the implicit equation @xmath8 is @xmath45 and its newton polytope is the tetrahedron with vertices @xmath46 . in the sparse case , the following generalization holds @xcite . for generic polynomials @xmath25 with the same newton polygon @xmath47 , the degree of @xmath8 is the lattice area @xmath48 and its newton polytope is the tetrahedron with vertices @xmath49 , @xmath50 . a first naive algorithm would then be the following . assume the newton polytope @xmath37 of @xmath8 is known ( as in the previous theorems ) and number @xmath51 the integer points ( also called lattice points ) in @xmath37 . consider indeterminates @xmath52 and write @xmath53 . substitute @xmath54 and equate to zero the coefficient of each power of @xmath35 that occurs ( clearing the denominator ) . this sets a system @xmath55 of linear equations in @xmath56 , with solution space of dimension @xmath57 . any nonzero solution @xmath56 of @xmath55 will give a choice of implicit equation @xmath8 . this solves the problem , but , which is the size of this linear system @xmath55 ? the number of lattice points in @xmath38 equals @xmath58 . in the the sparse case , the number of lattice points of a given lattice polygon @xmath47 can be computed via a theorem of ehrhart valid for any @xmath33 @xcite , which amounts to pick s formula in the case @xmath59 . given a positive integer @xmath60 , we denote by @xmath61 the minkowski sum of @xmath47 with itself @xmath60 times , i.e. @xmath62 . the number of lattice points in @xmath61 equals @xmath63 where @xmath64 denotes the number of lattice points in the boundary of @xmath47 . in particular , @xmath65 . the proof of the following result is straightforward : [ lem : size ] in case @xmath25 are generic polynomials of degree @xmath36 in @xmath66 , the number of unknowns in the linear system @xmath55 in the coefficients of the implicit equation @xmath8 is @xmath58 ( approximately @xmath67 ) and the number of equations is @xmath68 ( approximately @xmath69 ) . for any lattice polygon @xmath47 and generic polynomials @xmath70 with newton polytope @xmath47 , the linear system @xmath55 has @xmath71 ( approximately @xmath72 ) variables and @xmath73 equations ( approximately @xmath74 ) . we will see in remark [ rem : sizemcl ] of section [ sec : syzygies ] that the size of the involved linear systems in the syzygy based methods is drastically smaller . our main result is theorem [ th : mainalgo ] , which has a wide applicability . we distill and state it in naive terms , which do not call upon the more sophisticate tools recalled in section [ sec : tgtools ] and appendix [ sec : catools ] required for its proof this is why we postpone the detail of hypotheses [ hyp1 ] and [ hyp1xp ] until section [ sec : tgtools ] . our approach is an inhomogeneous translation of the basic general algorithm for the sparse case in @xcite , which were inspired by the methods @xcite for classical homogeneous polynomials . before moving to the implicitization of rational surfaces , we recall the practical approach of moving lines proposed by sederberg and chen @xcite for the implicitization of planar curves . a planar rational curve @xmath75 over a field @xmath0 is given as the image of a map @xmath76 with @xmath77 $ ] polynomials of degree @xmath36 in @xmath16 . we can assume without loss of generality that @xmath78 . remark that a linear syzygy can be represented as a linear form @xmath79 in the new variables @xmath80 with @xmath81 $ ] such that @xmath82 with this incarnation , a linear syzygy was termed a _ moving line_. for any @xmath83 , consider the finite - dimensional @xmath0-vector space @xmath84 of linear syzygies satisfying @xmath85 , and call @xmath86 its dimension . pick a @xmath0-basis @xmath87 of @xmath84 . consider the monomial basis @xmath88 of polynomials in @xmath16 of degree at most @xmath89 and write for each syzygy @xmath90 : @xmath91 let @xmath92 be the @xmath93 matrix of coefficients of the @xmath94 s with respect to the basis @xmath88 : @xmath95 observe that the variable @xmath16 has disappeared . this is the * main elimination step * ! it is known that for @xmath96 , the matrix @xmath92 is a square matrix with @xmath97 , where @xmath8 is an implicit equation of @xmath75 . in case @xmath98 , then @xmath92 is a non - square matrix with more columns than rows , but still the greatest common divisor of its minors of maximal size equals @xmath30 . in both cases , for @xmath96 , a point @xmath99 lies on @xmath75 iff the rank of @xmath100 drops . in other words , one can always represent the curve as a square matrix of linear syzygies , which gives a _ matrix representation _ of the implicit equation . in principle , one could now actually calculate the implicit equation , but the matrix @xmath92 is easier to get and well suited for numerical methods @xcite . as we remarked in the surface case , testing whether a point @xmath14 lies on the curve only requires computing the rank of @xmath92 evaluated in @xmath14 . also , the singularities of @xmath101 can be read off from @xmath92 @xcite . in the absence of common zeros of @xmath102 , it is possible to find the implicit equation via a resultant computation . note that for a parametrization with polynomials of degree @xmath36 , the sylvester resultant matrix uses a matrix of size @xmath103 , while the syzygy method uses @xmath26 matrices of size @xmath36 , as the bzout resultant . assume we are given a rational parametrization of a surface @xmath9 as in . we aim at finding a matrix representation for @xmath9 . note that we can in principle assume that @xmath104 are laurent polynomials admitting negative exponents , but after multiplying them by a common monomial , we get a new rational parametrization of @xmath9 defined by polynomials @xmath105 $ ] . we will then assume , without loss of generality , that we are have a lattice polygon @xmath47 which lies in the first orthant of @xmath106 and contains the newton polytopes of @xmath25 . we saw that in the curve case , it is always possible to find a square matrix representation . in the surface case , however , linear syzygies provide in general rectangular matrix representations and the implicit equation ( raised to the degree of the map @xmath6 ) equals the great common divisor of the maximal minors ( or the determinant of a complex ) . a recent paper by bus @xcite presents a very interesting square matrix representation out of a matrix representation @xmath10 when we work over the real numbers , by considering the square matrix @xmath107 . this approach is natural because of the properties of the rank of a real matrix with respect to its singular value decomposition . the determinant of @xmath107 gives an implicit equation for @xmath9 ( in general , it gives @xmath8 with multiplicity ) , which is moreover a sum of squares . as bus observes , for complex matrices it would be enough to replace the transpose @xmath108 by the conjugate transpose . the use of quadratic relations ( i.e. linear syzygies among the products @xmath109 of any of two of the polynomials @xmath70 defining the parametrization ) was proposed to construct square matrices @xcite . khetan and dandrea generalized in 2006 @xcite the method of moving quadrics to the toric case . the choice of the quadratic syzygies is in general not canonical and the cost of computing syzygies is increased . note that syzygies in @xmath104 and the implicit equation @xmath8 have a common shape . indeed , linear syzygies @xmath110 of degree @xmath89 correspond to polynomials @xmath111 such that @xmath112 , with @xmath113 in the @xmath16 variables equal to @xmath89 , and @xmath113 in the @xmath114 variables equal to @xmath57 . also , quadratic syzygies of degree @xmath115 correspond to polynomials @xmath116 such that @xmath117 , with @xmath113 in the @xmath16 variables equal to @xmath115 , and @xmath113 in the @xmath114 variables equal to @xmath26 . the implicit equation ( of degree @xmath118 ) is a polynomial @xmath119 such that @xmath120 . thus , @xmath113 in the @xmath16 variables equals @xmath121 , and @xmath113 in @xmath114 variables equals @xmath118 . so to go from linear syzygies to the implicit equation , in some sense one has to play the game of lowering the degree in the @xmath16 variables to @xmath121 ( which increases the degree in the @xmath114 variables up to @xmath118 ) . we now present our main general algorithm to construct matrix representations of parametrized surfaces . clearly , given any lattice polygon @xmath122 , @xmath123 is again a lattice polygon . moreover , in dimension two , any lattice polygon is normal , which means that @xmath124 , which is implicitly used in the algorithm . [ algo : mainalgo ] the following algorithm produces a matrix of polynomials in @xmath125 out of the input polynomials @xmath25 in variables @xmath126 : * * input : * a lattice polytope @xmath47 and polynomials @xmath127 with no common factor and newton polytopes @xmath128 contained in @xmath47 . * * step 1 : * consider syzygies @xmath15 with @xmath129 . let @xmath130 , @xmath131 , be a @xmath0-basis of such syzygies . * * step 2 : * represent the syzygies as linear forms @xmath132 . write @xmath133 and switch : @xmath134 * * output : * the matrix @xmath10 of linear forms @xmath135 . we illustrate the steps in algorithm [ algo : mainalgo ] in the following example . [ example : illustalg ] let @xmath47 be the lattice polygon with vertices @xmath136 , @xmath137 , @xmath138 and @xmath139 , with lattice points @xmath140 , as in the figure . we consider the following four polynomials with support in @xmath47 , where we denote @xmath141 , and given @xmath142 we write @xmath143 : @xmath144 , @xmath145 , @xmath146 , @xmath147 . [ fig : p ] to compute the syzygies in step 1 , we consider the morphism @xmath148 , where @xmath149 are polynomials with support in @xmath150 . let @xmath151 be the matrix of this map in the monomials bases . since @xmath150 has @xmath152 lattice points and @xmath153 has support in @xmath154 , which has @xmath155 lattice points , then @xmath151 is a _ sylvester _ matrix of size @xmath156 . it can be easily checked that @xmath151 is full ranked ( same as for generic polynomials ) . thus , the kernel of @xmath151 has dimension @xmath157 , which is the number of linearly independent syzygies . to construct the matrix @xmath10 , assume that we choose as our first syzygy the following @xmath158-tuple of polynomials @xmath159 with @xmath160 : @xmath161 , @xmath162 , @xmath163 , and @xmath164 . we consider @xmath165 and we write @xmath166 @xmath167 which gives the first column of the @xmath168-matrix @xmath10 ( computed with macaulay2 computer - algebra software @xcite ) .... 12 26 matrix ( qq[t , t , t , t ] ) < --- ( qq[t , t , t , t ] ) 0 1 2 3 0 1 2 3 .... the columns of this matrix @xmath10 are given by a choice of a basis of syzygies with support in @xmath169 . the corresponding linear forms @xmath170 are known as the _ moving planes _ defining the surface parametrized by @xmath171 . the associated rational map @xmath6 has @xmath172 . it can be checked that the common factor of any maximal minor of @xmath10 is the degree @xmath173 implicit equation of the closed image of @xmath6 : @xmath174 as asserted by theorem [ th : mainalgo ] below . note that we have to write the lattice points in @xmath150 as a sum of two points in @xmath47 , but in general there is not a unique way of doing this . in our example , for instance , @xmath175 , so a choice was made . in fact , it is possible to make a coherent choice in general with the use of weight vectors , but any choice will work since in the quotient ring @xmath21 defined in below , it holds that @xmath176 and @xmath177 are identified . we now state our main result . the proof will be given in section [ sec : tgtools ] . [ th : mainalgo ] given @xmath127 with no common factor , with newton polytopes contained in @xmath47 and satisfying hypotheses [ hyp1 ] below , algorithm [ algo : mainalgo ] computes a presentation matrix of the implicit equation of the rational map @xmath6 . that is , the rank of the matrix @xmath10 drops precisely when evaluated at the points in the closure of the image of @xmath6 . moreover , the implicit equation @xmath8 can be computed as @xmath178 the main ingredient for the validity of algorithm [ algo : mainalgo ] to give a matrix representation is the choice @xmath179 of the support of the linear syzygies . again , the `` instant '' elimination is done in step 2 , where the @xmath16 variables give the monomial basis which is used to compute the matrix @xmath10 ( and thus they disappear from the output ! ) . in fact , algorithm [ th : mainalgo ] can be run without checking hypotheses [ hyp1 ] . we point out in remark [ rem : whatifnot ] the possible outcomes . the general algorithm can be refined using theorem 11 in @xcite . [ th : refinement ] assume @xmath6 satisfies the hypotheses of theorem [ th : mainalgo ] . if the lattice polygon @xmath47 can be written as a multiple @xmath180 of another another lattice polygon @xmath181 _ without interior lattice points _ , then we can consider in * step 1 * of algorithm [ algo : mainalgo ] syzygies @xmath182 with smaller support @xmath183 contained in @xmath184 ( which is strictly contained in @xmath169 ) , and the * output * will still be a matrix representation for @xmath6 . moreover , in case @xmath181 is the unit simplex , it is enough to consider syzygies with support inside @xmath185 . we then have the following comparison between the general syzygy method and the naive linear algebra method described in section [ sec : naive ] . [ rem : sizemcl ] assume that @xmath47 is the triangle of size @xmath36 . then , as it is enough to consider syzygies of degree @xmath186 , they can be found by solving a linear system on @xmath187 variables with @xmath188 equations . that is , both sizes , as well as the vector space dimension of the space of syzygies in this degree , are quadratic in @xmath36 . the matrix @xmath10 has then a number of rows quadratic in @xmath36 . the number of its columns equals @xmath189 , again quadratic in @xmath36 . comparing with the sizes in lemma [ lem : size ] , which are of degree @xmath190 in @xmath36 , we observe that the syzygy method is a great improvement on the naive linear algebra method ! the same improvement occurs for any lattice polygon @xmath47 . using , we see that syzygies with support in @xmath150 can be obtained by solving a system with approximately @xmath191 equations in @xmath192 variables and both row and column sizes of the matrix representation @xmath10 are of the order of @xmath193 and not of its cube , as in lemma [ lem : size ] . as we have mentioned , the main motivation for the implicitization problem comes from computer aided geometric design and geometric modeling . in this area , bihomogeneous surfaces ( corresponding to rectangular support @xmath47 ) are known as tensor product surfaces , and they play a central role , in particular the bzier surfaces . quoting ulrich dietz @xcite : _ in current cad systems tensor product surface representations with their rectangular structure are a de facto standard _ " . these surfaces ( called _ nurbs _ ) are given by pieces of parametrized surfaces cut by curves . so , it is necessary to control the location of the parameter , which can be achieved by computing the kernel of the matrix representation we give , as explained in @xcite . due to the nature of the base locus of the parametrization , many of the current geometric modeling systems do not satisfy the hypotheses to be detailed in [ hyp1 ] needed for theorem [ th : mainalgo ] to hold , if considered as homogeneous polynomials ( with @xmath47 an equilateral triangle ) . but if we use a rectangle @xmath47 as the input in algorithm [ algo : mainalgo ] , it is possible to get a full - ranked matrix representation by theorem [ th : mainalgo ] . in this bihomogeneous case , the detailed study of regularity in @xcite allows to get the following improvement in the support of the proposed linear syzygies in step 1 of algorithm [ algo : mainalgo ] : it is enough that the support of these syzygies is contained in a polygon obtained by only enlarging the rectangle support @xmath47 of the input polynomials ( approximately ) to its double in _ only _ the horizontal _ or _ the vertical direction , instead of considering syzygies with support in ( the lattice points of ) @xmath150 . a more general result can be obtained for bigraded toric surfaces , and in particular for lattice polygons defining a hirzebruch surface , that it , for hirzebruch quadrilaterals @xmath194 with vertices @xmath195 and @xmath196 , for any @xmath197 . we state this extension in theorem [ th : hirzebruch ] below . [ algo : hirzebruch ] take as * input * a hirzebruch lattice polygon @xmath198 and bivariate polynomials @xmath199 with newton polygons contained in @xmath194 , and which satisfy the hypotheses [ hyp1xp ] . run algorithm [ algo : mainalgo ] with the following modification : in * step 1 * consider a basis of syzygies with support in the smaller lattice quadrilaterals @xmath200 ( or @xmath201 instead ) . the * output * is the corresponding matrix @xmath10 of linear forms . in most cases , it is convenient to consider syzygies with support in @xmath200 rather than in @xmath201 since the first one has less lattice points . in general , a hirzebruch lattice polygon @xmath202 has the shape in the diagram on the right . [ fig : h ] [ rem : hirzbigrad ] note that for @xmath203 , @xmath204 is the standard lattice rectangle with vertices in @xmath205 , and thus algorithm [ algo : hirzebruch ] works in particular in a standard bihomogeneous setting . [ th : hirzebruch ] given @xmath127 with no common factor , with newton polytopes contained in a @xmath194 and satisfying hypotheses [ hyp1xp ] below , algorithm [ algo : hirzebruch ] computes a presentation matrix of the implicit equation of the rational map @xmath6 . that is , the rank of the output matrix @xmath10 drops precisely when evaluated at the points in the closure of the image of @xmath6 . moreover , the implicit equation @xmath8 can be computed as @xmath206 the proof of theorem [ th : hirzebruch ] will be also given in section [ sec : tgtools ] . in this section we will recall a minimum of theoretical tools from toric geometry in order to state the hypotheses needed for theorems [ th : mainalgo ] and [ th : hirzebruch ] and to give their proofs . the main homological commutative algebra tools that are the core of the proofs are recalled in appendix [ sec : catools ] . we refer to @xcite and ( * ? ? ? * ch.5&6 ) for the general notions , and to @xcite , @xcite for applications to the implicitization problem . any reader only interested in the application of algorithm [ algo : mainalgo ] or its bihomogeneous ( toric ) refinement given in algorithm [ algo : hirzebruch ] can skip this section . as usual , we denote by @xmath207 the multiplicative group of units of @xmath0 . the first observation is that we can equivalently consider our parametrization as a map @xmath208 or @xmath209 with image inside @xmath173-dimensional projective space , and domain a dense open set @xmath210 in affine space @xmath7 or the torus @xmath211 over @xmath0 , given by @xmath212 for any @xmath213 , and we have the commutative diagram @xmath214^{f } \ar@{-->}[rd]_{\tilde{f } } & \kk^3 \ar@{^(->}[d]_\iota\\ & \pp^3.}\ ] ] in fact , if @xmath8 is the implicit equation of the ( closure of the ) image of @xmath6 , the ( closure of the ) image of @xmath215 is the closure of @xmath9 under the standard embedding @xmath216 and its equation is the homogenization of @xmath8 . similarly , we can consider our rational parametrization from any algebraic variety which contains the domain of @xmath6 as a _ dense _ subset . we will choose embedded or abstract compact toric varieties to get a degree or multidegree notion that will allow us to get homological arguments to `` bound '' the support of the syzygies in theorem [ th : mainalgo ] and in [ th : hirzebruch ] . let @xmath215 be a rational map as in . the base points of the parametrization are the common zeros of @xmath25 , that is , the points where the map is not defined . we assume that @xmath215 is a generically finite map onto its image and hence it parametrizes an irreducible surface @xmath217 . we also assume without loss of generality that @xmath218 , which means that there are only finitely many base points . let @xmath122 be a lattice polygon with @xmath219 lattice points , which contains the newton polygons @xmath220 . write @xmath221 . the polygon @xmath47 determines a projective toric surface @xmath222 as the closed image of the embedding @xmath223 where @xmath224 . for example , the unit triangle with vertices @xmath137 , @xmath225 and @xmath136 ( or any lattice translate of it ) corresponds to @xmath226 , and any lattice rectangle gives a segre - veronese projective embedding of @xmath227 , which are special cases of toric embeddings . [ ex : p1p1 ] assume @xmath47 is the unit square , with @xmath228 integer points : @xmath229 a polynomial @xmath70 with newton polytope contained in @xmath47 looks like @xmath230 we take @xmath158 new variables @xmath231 as the homogeneous coordinates in @xmath232 . the toric variety @xmath233 is the projective variety in @xmath232 cut out by the relation @xmath234 . this binomial equation comes from the primitive affine relation @xmath235 , which implies the multiplicative relation @xmath236 between the monomials with these exponents . the coordinate ring of @xmath233 is the quotient ring @xmath237/\langle x_0 x_3 - x_1 x_2\rangle.$ ] in general , we will call @xmath238 the homogeneous coordinate variables in @xmath239 . write @xmath221 . we set one variable @xmath240 for each integer point @xmath241 in @xmath47 and we record multiplicatively ( by binomial equations ) the affine relations among these points . these binomials generate the toric ideal @xmath242 , which defines the variety @xmath243 . to each @xmath244 we associate the homogeneous linear form @xmath245 for instance , in example [ ex : p1p1 ] , the polynomial @xmath70 gets translated to @xmath246 and over @xmath233 , we have the relation @xmath247 . the rational map @xmath215 factorizes through @xmath233 in the following way @xmath248^{\tilde{f } } \ar@{^(->}[d]^{\rho } & \pp^3 \\ \tc_p \ar@{-->}[ur]_g } \ ] ] where @xmath249 is given by four homogeneous linear polynomials @xmath250 in @xmath219 variables . thus , we have a new homogeneous parametrization @xmath251 of the closed image of @xmath215 from @xmath233 . the polynomials @xmath252 generate an ideal @xmath253 in the coordinate ring @xmath254 / j_p\ ] ] of @xmath233 . this ideal @xmath253 defines the structure of the base locus in @xmath233 . the embedding @xmath255 provides a @xmath256-grading in the coordinate ring @xmath21 of @xmath233 , which is used to study the map @xmath251 with the tools recalled in appendix [ sec : catools ] . given a lattice polygon @xmath47 , one can also associate to it an abstract compact toric variety @xmath257 that naturally contains the torus @xmath211 as a dense open set ( via the map we call @xmath258 below ) , adjoining a torus invariant divisor to each edge of @xmath47 . we refer the reader to @xcite for the theory and details . the map @xmath215 also defines a rational map @xmath259 that makes the following diagram commutative : @xmath260^{\overline{f } } & \\ ( \kk^*)^2 \ar@{^(->}[u]_{j}\ar@{^(->}[d]_{\rho}\ar@{-->}[r]_{\tilde{f } } & \pp^3 \\ \tc_p \ar@{-->}[ur]_{g } & } \ ] ] [ ex : p1p1b ] assume @xmath47 is the unit square , with @xmath261 edges : the segments @xmath262 $ ] , @xmath263 $ ] , @xmath264 $ ] , and @xmath265 $ ] . the respective inner normal vectors @xmath266 , @xmath267 satisfy the linear relations @xmath268 , @xmath269 , which give rise to two homogeneities . we introduce four associated variables @xmath270 . a polynomial @xmath70 with newton polytope @xmath47 as in defines a bihomogeneous polynomial ( in @xmath271 and @xmath272 ) : @xmath273 these polynomials @xmath274 define the map @xmath275 . the main motivation for this change of perspective comes again from the commutative algebra results needed for the proof of theorems [ th : mainalgo ] and [ algo : hirzebruch ] . the cox ring of @xmath257 is endowed with a more natural multigrading , which is finer than the grading obtained via the embedded projective variety @xmath233 . also , this point of view has an impact in the computations , as the number of variables to eliminate is smaller ( one for each edge of @xmath47 , instead of one for each lattice point in @xmath47 ) . in our small example [ ex : p1p1b ] , there are four edges and four lattice points , but the number of edges can remain constant while the number of lattice points goes to infinity . in this subsection we detail the precise hypotheses that ensure the validity of theorem [ th : mainalgo ] and we prove it , based on results in @xcite . we first need to recall a few standard definitions from commutative algebra . given ( nonzero ) homogeneous polynomials @xmath276 , defining a rational map @xmath277 as in , a point @xmath278 is a base point of @xmath251 if it is a common zero set of @xmath279 , that is , if @xmath14 is a zero of the ideal @xmath280 in @xmath233 . let @xmath278 be a base point of @xmath251 . the local ring of @xmath14 is the ring @xmath281 , with the natural operations induced from @xmath21 ( in turn , naturally induced from the polynomial ring ) . let @xmath282 be the ideal generated by ( the classes of ) @xmath283 in @xmath284 . we say that @xmath14 is a local complete intersection base point if @xmath282 can be generated by only @xmath26 elements . we say that @xmath14 is an almost complete intersection base point if @xmath282 can be generated with @xmath173 elements . we have similar definitions for the map @xmath285 . for a given lattice polygon @xmath47 , here are the hypotheses we need in terms of @xmath251 : [ hyp1 ] there are only finitely many base points of @xmath251 on @xmath233 which are local complete intersections . we can not easily find hypotheses on @xmath6 equivalent to hypotheses [ hyp1 ] . given a lattice polygon @xmath47 , an edge @xmath286 of @xmath47 , and a polynomial @xmath70 with @xmath128 contained in @xmath47 as in , the restriction @xmath287 of @xmath70 to @xmath286 is defined as the sub - sum of the monomials with exponents @xmath241 in @xmath286 . we have the following partial translation . [ prop : inequiv ] let @xmath288 and @xmath251 be as in . then 1 . [ one ] there are only finitely many base points of @xmath251 on @xmath233 if and only if there are only finitely many isolated base points of @xmath6 in the torus and for each edge @xmath286 of @xmath47 , at least one of the restrictions @xmath287 is nonzero . [ two ] if @xmath251 has finitely many isolated base points on @xmath233 which are local complete intersections , then the base points of @xmath6 in the torus are local complete intersections . the map @xmath289 defines an isomorphism between @xmath211 and its image ( which is an open dense subset of @xmath233 ) , sending a base point @xmath290 of @xmath215 ( that is , a point where @xmath291 ) to a base point @xmath292 of @xmath251 , and reciprocally , any base point @xmath14 of @xmath251 in @xmath293 is the image of a base point @xmath290 of @xmath6 . moreover , we have an isomorphism between the ideal generated by @xmath25 at @xmath290 and @xmath282 . any base point of @xmath251 outside the image of @xmath289 can not be seen in the torus . but these points are either the fixed torus points corresponding to the finitely many vertices of @xmath47 , or they lie in the torus of the toric divisor @xmath294 in @xmath233 associated to an edge @xmath286 of @xmath47 . as @xmath294 has dimension @xmath57 , there are finitely many solutions as long as at least one of the @xmath287 is nonzero . the following example shows that the converse of item in proposition [ prop : inequiv ] does not hold . [ ex : nonlaci ] consider the parametrization with @xmath190 monomials : @xmath295 . then , @xmath6 has no base points in the torus . but if we consider their standard homogenizations to degree @xmath296 polynomials ( that is , we take @xmath47 equal to @xmath296 times the standard unit simplex in the plane ) , the corresponding homogeneous polynomials @xmath297 have one base point `` at infinity '' which is not even an almost locally complete intersection . we now give the proof of theorem [ th : mainalgo ] . given @xmath127 with no common factor and newton polytopes contained in @xmath47 , the corresponding polynomials @xmath252 associated to @xmath70 are homogeneous of degree @xmath298 and satisfy hypotheses [ hyp1 ] . from ( * ? ? ? 14 ) one has that for @xmath298 , the matrix of coefficients of a @xmath299-basis of the module of syzygies of @xmath251 in any degree @xmath300 with respect to a @xmath299-basis of the graded piece @xmath301 of @xmath21 , is a matrix representation for the closure of the image of @xmath6 , which equals the closure of the image of @xmath251 . in particular , we can take @xmath302 . in step 1 of algorithm [ algo : mainalgo ] , the syzygies @xmath130 for @xmath131 with @xmath303 for all @xmath304 , provide a @xmath0-basis of the module of syzygies of @xmath251 in degree @xmath26 , since classes of monomials of degree @xmath26 in @xmath21 correspond to monomials in the @xmath16 variables with exponents in @xmath150 . equality follows from theorem 13 in @xcite . in principle , given a rational map @xmath215 , we could take any lattice polygon @xmath47 containing the union @xmath305 of newton polytopes of @xmath25 . note that the hypothesis that @xmath6 is generically finite implies that @xmath305 is two - dimensional . taking @xmath47 strictly containing @xmath305 will increase the number of exponents and will in general produce bad behaviour of @xmath251 at the fixed points in @xmath233 corresponding to the vertices of @xmath47 which do not lie in @xmath305 . [ rem : whatifnot ] we can check algorithmically if @xmath25 have finitely many solutions over @xmath306 and if for any edge @xmath286 at least one of the restrictions @xmath287 is nonzero . so , by proposition [ prop : inequiv ] , we can check whether @xmath251 has finitely many base points in @xmath233 . assume the dimension of the base locus of @xmath251 is zero . as we remarked in example [ ex : nonlaci ] , even if we could check the local behavior of the base points of @xmath6 in the torus , this would not imply the satisfiability of hypotheses [ hyp1 ] . but what if we do nt check this and run algorithm [ algo : mainalgo ] ? nothing bad ! we then check whether the output matrix @xmath10 in has full rank . * if the rank of @xmath10 is not maximal , then there is at least one base point @xmath14 of @xmath251 which is not an almost local complete intersection . in this case , we can not get the implicit equation , but we get a certificate of the bad behavior of the base locus ( without computing it ) . * if the rank of @xmath10 is maximal , it may happen that the its rank drops when evaluated at points outside @xmath9 due to the existence of an almost complete intersection but non complete intersection base point . in this case , the greatest common divisor of the maximal minors of @xmath10 would have irreducible factors other than the implicit equation @xmath8 . in fact , the existence of other irreducible factors is equivalent to the fact that there exists a base point which is an almost local complete intersection but not a local complete intersection . in this subsection we detail the precise hypotheses that ensure the validity of theorem [ th : hirzebruch ] and we prove it , based on results in @xcite . given @xmath47 and @xmath6 , here are the hypotheses we need in terms of the map @xmath259 in : [ hyp1xp ] there are only finitely many base points of @xmath259 on @xmath257 which are local complete intersections . again , we can not easily find hypotheses on @xmath6 equivalent to hypotheses [ hyp1 ] , since good algebraic behaviour of the base points in the torus does not imply the same behaviour for the possible base points of @xmath259 at the invariant divisors in @xmath257 associated with the edges of @xmath47 . [ prop : inequivxp ] let @xmath307 and @xmath259 as in [ eq : diagram2 ] . then 1 . [ oneb ] there are only finitely many isolated base points of @xmath259 on @xmath257 if and only if there are only finitely many isolated base points of @xmath6 in the torus and for each edge of @xmath47 , at least one of the restrictions of the @xmath70 is nonzero . [ twob ] if @xmath259 has finitely many base points on @xmath257 which are local complete intersections , then the base points of @xmath6 in the torus are local complete intersections . we next give the proof of theorem [ th : hirzebruch ] . by hypothesis , there are only finitely many isolated base points of @xmath259 on the toric variety @xmath257 associated with @xmath308 , which are local complete intersections . there are four primitive inner normal vectors of @xmath47 : @xmath309 , which satisfy the linear relations @xmath310 . so any multidegree @xmath89 can be described by a `` bidegree '' @xmath311 given by the degrees with respect to the first normals and which fixes ( up to translation ) the associated polytope @xmath312 with the same normals as @xmath47 . thus , by ( * ? ? ? 5.5 ) the matrix of coefficients of a @xmath299-basis of the module of syzygies of @xmath259 in any bidegree @xmath313 with @xmath314 and @xmath315 ( or @xmath316 ) . in ( * ? ? ? 5.5 ) , or the analysis of the bidegree in the standard bigraded case in ( * ? ? ? 7.1 ) ] with respect to a @xmath299-basis of the bigraded piece @xmath317 of the cox ring of @xmath257 , is a matrix representation for the closure of the image @xmath9 of @xmath6 ( which equals the closure of the image of @xmath259 ) . taking @xmath318 one has that in step 1 a basis of syzygies @xmath130 , for @xmath319 with all @xmath320 with support in the quadrilateral @xmath200 , provides a @xmath0-basis of the module of syzygies of @xmath259 in bidegree @xmath321 . hence , the matrix @xmath10 of coefficients of such syzygies obtained in step 2 gives a representation matrix for @xmath9 . equality also follows from theorem 5.5 in @xcite . this section consists of four examples which highlight the usefulness of our approach . example [ ex : toricnotp2 ] is taken from a case studied in ( * ? ? ? @xmath322 ) of a sparse parametrization where projective implicitization does not work due to the nature of the base locus of the map , but algorithm [ algo : mainalgo ] is applicable with a right choice of polygon @xmath47 read from the monomials of the input polynomials . in example [ ex : highdeg ] we show how the method in algorithm [ algo : mainalgo ] works for a parametrization given by fewnomials of high degree , where classical resultant tools fail due to the computational complexity . in example [ ex : bptorus ] , classical resultant tools fail because of the existence of a base point in the torus . finally , in example [ ex : withoutemb ] we compare the methods in algorithms [ algo : mainalgo ] and [ algo : hirzebruch ] . consider the parametrization with @xmath190 monomials given in ( * ? ? ? @xmath322 ) : @xmath295 . the matrix representation can be computed using the package _ matrixreptoric.m2 _ @xcite in the computer algebra software _ macaulay2 _ @xcite . one first defines the map @xmath6 given by polynomials in the ring @xmath323 $ ] ( note that for easiness of typing , we call the variables @xmath324 instead of @xmath66 ) : .... s = qq[s , t ] ; f = { s*t^6 + 2 , s*t^5 - 3*s*t^3 , s*t^4 + 5*s^2*t^6 , 2+s^2*t^6 } ; .... consider @xmath47 the lattice triangle with vertices @xmath325 and @xmath326 . one can compute @xmath47 by the command : .... p = polynomialstopolytope l .... the lattice - points of @xmath47 can be computed using the auxiliary macaulay2 package ` polyhedra ` as : .... latticepoints p .... by taking syzygies with support in @xmath150 , on gets a matrix representation of size @xmath327 . the greatest common divisor of the @xmath328-minors of this matrix is the homogeneous implicit equation of the surface : @xmath329 we can set @xmath330 to get the affine equation . the map @xmath251 is computed with the following command : .... g = tetoricrationalmap f ; .... the matrix representation and the implicit equation are computed as follows : .... m = representationmatrix ( tetoricrationalmap f,2 ) ; impliciteq ( l,2 ) .... notice that the @xmath26 as second parameter in the computation of @xmath331 is precisely the @xmath26 in the support @xmath150 of the syzygies . for a deeper understanding of the choice of the parameter @xmath26 , see appendix [ sec : catools ] . in the language of section [ sec : tgtools ] , the coordinate ring associated to @xmath233 is @xmath332/j_p$ ] , where @xmath333 . the parametrization @xmath251 over @xmath233 is given by @xmath334 . this matrix can be computed as the right - most map of the @xmath335 strand of a graded complex as explained in appendix [ sec : catools ] . the method fails over @xmath226 ( i.e @xmath336 the triangle with vertices @xmath337 ) due to the nature of the base locus . one can see this just by computing a matrix representation and verifying that it is not full - ranked . this example contributes to show how the method works fine for high degree fewnomials involved in the parametrization . consider the polynomials @xmath338 . let @xmath339 be the parametrization that maps @xmath340 . the implicit equation of the closure of the image of @xmath6 could be computed by eliminating the variables @xmath324 as follows ( using general elimination procedures based on grbner bases in macaulay2 ) : .... r = qq[s , t , x , y , z , w ] f0 = 1 + s*t + s^37 ; f1 = s^7 + s^47 ; f2 = s^37 + s^59 ; f3 = s^61 l1 = x*f1 - y*f0 ; l2 = x*f2 - z*f0 ; l3 = x*f3 - w*f0 eliminate ( { s , t } , ideal(l1,l2,l3 ) ) .... in a 2014 standard desktop computer this routine does not end after one hour of computation . we also tried the well implemented _ eliminate _ command in singular @xcite , but with the same lack of answer after a couple of hours of computation . by homogenizing with an auxiliary variable @xmath341 we could try eliminate @xmath342 using macaulay resultant methods , but we easily figure out that the homogenized forms @xmath343 vanish identically over the point @xmath344 . this implies in particular that the macaulay resultant @xmath345 is identically zero . finally one can compute the implicit equation ( and matrix representation ) by implementing the implicitization techniques described in this article . a not very efficient ( but efficient enough ) routine in @xcite gives the toric map @xmath251 in less than @xmath26 minutes and the desired matrix representation @xmath10 in less than one more minute . this example shows a case where classical resultants can not be applied to compute the implicit equation , but the techniques in the paper can . anyway , we recall that the aim of the matrix representations is to provide a better and more complete tool for representing a surface , and hence , the point presented with this example is just one extra advantage of the method . consider the following parametrization @xmath346 given by four polynomials that define a parametrization @xmath339 that maps @xmath340 . the implicit equation can not be computed by eliminating the variables @xmath324 with classical resultants , because the point @xmath139 is in the base locus . however , this fact does not imply any problem in algorithm [ algo : mainalgo ] . with algorithm [ algo : hirzebruch ] with a rectangular @xmath347 , it takes 0.058 seconds in a standard 2014 desktop computer to get a matrix representation @xmath10 of the closure of the image of @xmath6 . the size of @xmath10 is @xmath348 and the gcd of its maximal minors has degree @xmath349 . while algorithm [ algo : mainalgo ] holds with great generality , when dealing with polynomials with rectangular support ( which can be interpreted as bihomogeneous polynomials ) , algorithm [ algo : hirzebruch ] provides a smaller matrix representation . consider the following four polynomials @xmath350 : @xmath351 , @xmath352 , @xmath353 , and @xmath354 . the newton polytope @xmath355 is the rectangle @xmath356 . if we apply algorithm [ algo : mainalgo ] ( as we illustrated in example [ example : illustalg ] ) , we obtain a matrix representation @xmath10 of size @xmath357 . the associated toric variety @xmath257 can be identified with the @xmath358 segre - veronese embedding of @xmath359 in @xmath360 ( see @xcite ) . by means of algorithm [ algo : hirzebruch ] we get a matrix representation @xmath10 from a basis of linear syzygies of bidegree @xmath361 . this matrix representation can be computed using the algorithm developed in @xcite and implemented in m2 , as the matrix @xmath92 for bidegree @xmath362 , and one obtains a square @xmath363-matrix . its determinant equals the implicit equation @xmath8 : .... 8 7 6 2 5 3 4 4 f = 63569053x - 159051916x x + 175350068x x - 82733240x x + 2363584x x + ... 0 0 1 0 1 0 1 0 1 .... notice that the matrix @xmath364 is considerably smaller than the @xmath357-matrix @xmath10 because instead of considering syzygies with support in the rectangle @xmath365 , the syzygies are taken with support in the smaller rectangle @xmath366 . this appendix is devoted to highlight the tools of homological commutative algebra and algebraic geometry that are needed to justify the validity of algorithms [ algo : mainalgo ] and [ algo : hirzebruch ] , and to explain the choice of the support of the syzygies in step 2 which define the matrix representation of the parametrized surface . given @xmath47 , the toric embedding @xmath367 in section [ ssec : toricemb ] provides a @xmath256-grading in the coordinate ring @xmath368/j_p$ ] in of @xmath233 , which can be used to study the map @xmath251 in diagram [ diagram2 t ] and its associated rees and symmetric algebras , denoted by @xmath369 and @xmath370 respectively . notice also that the graded ring @xmath21 coincides with the affine semigroup ring of the lattice polytope @xmath47 , which is _ cohen - macaulay _ and normal because @xmath47 has dimension @xmath26 . the grading in @xmath21 plays a key role in the elimination process . the matrix representation @xmath10 of section [ ssec : general ] depends on a choice of degree @xmath89 , as was shown in section [ ss : curves ] for the case of curves . the reason why @xmath89 needs to be considered is rather technical , and a complete explanation involves sheaf cohomology . from a more naive point of view , the implicit equation of the surface @xmath371 is written in the variables @xmath372 but depends on the algebraic relations among the polynomials @xmath252 , which lie in @xmath21 . fixing a degree @xmath89 in @xmath21 can be thought as _ eliminating the variables of @xmath21 , by hiding them in the monomial basis of the graded piece @xmath301_. in turn , recall that the variables @xmath373 in @xmath21 code monomials in the original @xmath16 variables , with exponents in the lattice points in @xmath47 . more geometrically , consider the graph variety @xmath374 of @xmath251 where both group of variables @xmath375 and @xmath114 are involved . the elimination process can be understood geometrically as projecting @xmath374 via @xmath376 in the following diagram @xmath377^{\pi_2 } \ar[d]^{\pi_1 } & \\ \tc_p \ar@{-->}[r]_g & \pp^3 \supset \overline{\im(g ) } = \overline{\im(f)}.}\ ] ] in the correspondence between subvarieties of @xmath378 and bigraded algebras , the inclusion of the graph @xmath379 corresponds to the surjection @xmath380 \twoheadrightarrow { \mathrm{rees}_a ( i)}$ ] , the rees algebra of the ideal @xmath253 generated by @xmath297 over the coordinate ring @xmath21 . the projection @xmath381 corresponds to eliminating the variables @xmath240 of @xmath369 . we denote by @xmath382 the defining ideal of @xmath383 . how to eliminate the @xmath375 variables from @xmath369 algebraically ? a standard procedure is to find a free graded presentation @xmath384 and a degree @xmath89 ( in the @xmath375 variables ) such that the fitting ideal @xmath385 generated by the maximal minors of @xmath92 ( in the graded strand @xmath386 ) computes @xmath382 . it happens that no universal way to compute such a free presentation is available , so the idea is to `` approximate '' @xmath369 by the ( hopefully ) similar algebra @xmath370 that admits such a universal resolution . these resolutions of the symmetric algebras are known as approximation complexes , they were introduced in @xcite and their application on elimination theory was done in @xcite . the last map of the approximation complex is the following in our case : @xmath387{\stackrel}{m'}{{\longrightarrow}}a[t_0,t_1,t_2,t_3]\to { \mathrm{sym}_a ( i)}\to 0,\ ] ] where @xmath388 is the first module of syzygies of @xmath250 and @xmath389^t$ ] , that is , @xmath390 the cokernel of @xmath391 is @xmath392/ j$ ] , where @xmath393\mbox { and } \sum a_ig_i=0\}.\ ] ] we can recognize the origin of the linear forms @xmath94 in step 2 of our algorithms ! but there is a remaining question : which is the relation between @xmath369 and @xmath370 ? which variety does @xmath370 define ? can we use @xmath394 to compute @xmath395 for some @xmath89 ? the answer is that in case there are finitely many base points and for each base point @xmath14 , the local @xmath282 is a local complete intersection , then @xmath369 and @xmath370 define the same scheme in @xmath396 ( thus , in @xmath397 ) . as @xmath369 is @xmath398-torsion free , both algebras coincide module the @xmath398-torsion of @xmath370 , @xmath399 , where @xmath398 is the maximal ideal generated by @xmath400 . thus if @xmath253 is a local complete intersection and @xmath89 is such that @xmath401 , then @xmath402 . this happens for @xmath403 by theorem 11 in @xcite ( in fact , that result also proves theorem [ th : refinement ] as remarked before ) . in particular , in this case , @xmath394 computes @xmath395 for any @xmath404 . in fact , the condition of @xmath253 being a local complete intersection can be relaxed to the condition of being locally an almost complete intersection . ( i.e. @xmath282 can be generated by @xmath173 elements , for any @xmath14 in the finite set @xmath405 ) . in this case , @xmath406 . since there is always a surjective map @xmath407 then @xmath408 , where @xmath210 has the same dimension . in particular , @xmath409 . for @xmath404 , @xmath370 is @xmath398-torsion free , and @xmath394 computes @xmath410 for any @xmath404 . so , the gcd @xmath411 of the maximal minors of @xmath412 contains has the homogenization of the implicit equation @xmath8 as a factor . in the bigraded case of hirzebruch surfaces , in particular in the standard bigraded case , the basic ideas are similar , but new technical details have to be managed in order to determine the bidegrees for which the torsion of the symmetric algebra vanishes . we refer the reader to @xcite for the details and proofs . we are grateful to the organizers of the european marie curie itn saga network for the invitation to deliver a mini - course at the fall school shapes , geometry , and algebra held on october 2010 , at kolympari , greece , which was the starting point of this article . we thank laurent bus for interesting conversations . dhavide a. aruliah , robert m. corless , laureano gonzalez - vega , and azar shakoori . geometric applications of the bezout matrix in the lagrange basis . in _ proceedings of the 2007 international workshop on symbolic - numeric computation _ , 5564 . acm , 2007 . laurent bus . elimination theory in codimension one and applications . inria research report 5918 , page 47 . notes of lectures given at the cimpa - unesco - iran school in zanjan , iran , july 9 - 22 2005 , 2006 . david a. cox . equations of parametric curves and surfaces via syzygies . in _ symbolic computation : solving equations in algebra , geometry , and engineering ( south hadley , ma , 2000 ) _ , volume 286 of _ contemp . math _ soc , providence , ri , 2001 . gema m diaz - toca , mario fioravanti , laureano gonzalez - vega , and azar shakoori . using implicit equations of parametric curves and surfaces without computing them : polynomial algebra by values . , 30(1):116139 , 2013 . jrgen herzog , aron simis , and wolmer v. vasconcelos . koszul homology and blowing - up rings . in _ commutative algebra ( trento , 1981 ) _ , volume 84 of _ lecture notes in pure and appl . dekker , new york , 1983 . dinesh manocha and john canny . a new approach for surface intersection . in _ proceedings of the first acm symposium on solid modeling foundations and cad / cam applications _ , sma 91 , 209219 , new york , ny , usa , 1991 . acm . luu ba thang , laurent bus , and bernard mourrain . curve / surface intersection problem by means of matrix representations . in _ proceedings of the 2009 conference on symbolic numeric computation _ , 7178 . acm , 2009 . jianmin zheng , thomas w. sederberg , eng - wee chionh , and david a cox . implicitizing rational surfaces with base points using the method of moving surfaces . in _ topics in algebraic geometry and geometric modeling _ , volume 334 of _ contemp . _ , 151168 . soc . , providence , ri , 2003 .

this paper presents _ xmm - newton _ and _ chandra _ x - ray spectroscopy of ten flat - spectrum radio quasars ( fsrq ) which are candidates to have an x - ray spectrum dominated by jet synchrotron emission . in all these fsrq , which are less strongly relativistically beamed than blazars , a considerable contribution from a power - law component similar to that present in radio - quiet quasars is required to explain the observed x - ray fluxes and x - ray spectral slopes . and as in radio - quiet quasars , their relatively high optical / uv fluxes can be accounted for by a significant contribution from thermal accretion disk emission . the lack of success in finding radio quasars with synchrotron x - rays is attributed to the adopted selection criteria , which were based on the multiwavelength flux ratios of bl lacertae ( bl lac ) objects . a refined selection technique , which additionally involves radio imaging , is proposed to search for these important candidates for the gamma ray large area space telescope ( glast ) . on the other hand , the discovered fsrq with their strong accretion disk signatures are expected to be important probes for studies of the poorly known accretion disk - jet connection . = cmr7 0= = 0 @=@

0712.1689

blazars offer us the unique possibility to study the spectral energy distributions ( seds ) of relativistic jets . being radio galaxies with their jets oriented at relatively small angles with respect to our line of sight they profit from the relativistic beaming effect . this enhances dramatically ( by factors up to @xmath0 ) the intrinsic jet ( core ) flux and thus makes it the dominant continuum emission at all frequencies [ see @xcite for a review ] . as first discussed by @xcite two emission processes dominate the multiwavelength seds of relativistic jets and , therefore , of blazars , namely , synchrotron and inverse compton . these give rise to two prominent emission peaks in a logarithmic @xmath1 plot , which were observationally demonstrated by @xcite and @xcite . the synchrotron emission peak ( @xmath2 ) in blazars is observed within a wide range of frequencies ( from infrared to soft - x - rays ) in sources with weak emission lines , the bl lacertae objects ( bl lacs ; e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? however , strong - lined blazars , the flat - spectrum radio quasars ( fsrq ) , had until recently synchrotron emission peaks at relatively low energies and thus x - rays dominated by inverse compton emission ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? based on this finding some authors proposed the so - called `` blazar sequence '' @xcite . this scenario advocates that the frequency of the synchrotron emission peak is governed by particle compton cooling by an external radiation field . such a field , produced by , e.g. , the accretion disk or broad emission line region ( blr ) , both more luminous in radio quasars than in bl lacs , is expected to interact with the particles in the jet via the inverse compton process causing them to lose energy . therefore , within this scenario radio quasars with @xmath3 hz and thus x - rays dominated by synchrotron emission are not expected to exist . two new surveys have recently shown that mere selection effects had prevented us so far from finding `` x - ray loud '' radio quasars , allowing for the possibility that radio quasars with high @xmath4 existed . about 10% of the fsrq discovered in the deep x - ray radio blazar survey ( dxrbs ; * ? ? ? * ; * ? ? ? * ) and @xmath5 of the ones identified in the _ rosat _ all - sky survey ( rass ) - green bank ( rgb ; * ? ? ? * ; * ? ? ? * ) have multiwavelength flux ratios similar to those of bl lacs with synchrotron x - rays @xcite . in addition , subsequent investigations of the _ einstein _ medium sensitivity survey ( emss ) and the slew survey revealed that these radio quasars had gone undetected in previous surveys @xcite . the definite proof of the synchrotron nature of the x - ray emission in these newly discovered blazars , however , requires x - ray spectroscopy . @xcite presented _ beppo_sax observations of four `` x - ray loud '' fsrq and found that the synchrotron emission of one source , rgb j1629@xmath64008 , peaked at uv frequencies ( @xmath7 hz ) and dominated the x - rays . in this paper we present _ xmm - newton _ and _ chandra _ observations of a further ten of these sources . the observed sample was selected as discussed in section 2 and the data acquired and analyzed as described in section 3 . the multiwavelength seds of the sample are presented in section 4 and the results discussed in section 5 . section 6 gives a brief summary . throughout this paper cosmological parameters @xmath8 km s@xmath9 mpc@xmath9 , @xmath10 , and @xmath11 have been assumed . energy spectral indices have been defined as @xmath12 and photon spectral indices as @xmath13 , where @xmath14 . we have selected for x - ray spectroscopy fsrq from the dxrbs and rgb survey which had multiwavelength flux ratios similar to those of bl lacs with x - rays dominated by synchrotron emission . to this selection criterion we have added different constraints in the case of the _ xmm - newton _ and the _ chandra _ observed samples ( see below ) . as first suggested by @xcite bl lacs that emit synchrotron radiation which peaks at low ( ir / optical ) frequencies , the so - called low energy - peaked bl lacs ( lbl ) , can be distinguished from those that emit synchrotron radiation which peaks at higher ( uv / soft - x - ray ) frequencies , the so - called high energy - peaked bl lacs ( hbl ) , by their distinct ratios between their radio , optical and x - ray fluxes . these ratios are usually studied by plotting the objects in an ( @xmath15 ) plane , where @xmath16 and @xmath17 are the usual rest - frame effective spectral indices defined between 5 ghz , 5000 , and 1 kev . in this plane constant @xmath18 values are represented by straight lines of slope @xmath19 ( see fig . [ aroaox ] ) . a sequence of synchrotron emission peak frequencies in bl lacs produces a specific trail in the ( @xmath20 ) plane @xcite . while the synchrotron emission peak is at low enough frequencies for the x - rays to be dominated by inverse compton radiation ( as in lbl ) , @xmath18 is roughly constant , whereas @xmath16 decreases and @xmath21 increases as the peak moves to higher energies ( see also @xcite ) . when the synchrotron radiation starts to dominate the x - ray band ( as in hbl ) , @xmath18 starts to decrease , with lbl and hbl having values of @xmath22 and @xmath23 , respectively @xcite . now @xmath16 is roughly constant ( and low ) and @xmath21 decreases . based on these considerations @xcite have defined a region in the ( @xmath20 ) plane expected to be populated by high - energy peaked blazars , both bl lacs and fsrq ( see their fig . this so - called `` hbl box '' represents the 2 @xmath24 region around the mean @xmath25 , and @xmath18 values of all hbl in the multifrequency active galactic nuclei ( agn ) catalog of @xcite . we have selected for x - ray spectroscopy with _ xmm - newton _ and _ chandra _ fsrq that populate the `` hbl box '' ( fig . [ aroaox ] ) . for observations with _ xmm - newton _ we have chosen fsrq from dxrbs only and have additionally required that the synchrotron emission peak frequency , estimated from a crude multiwavelength sed including only radio , optical and x - ray data points @xcite , was @xmath26 hz . eight sources satisfied these criteria and were proposed . we were granted observing time in cycle 3 for 4/8 sources ( pi : padovani ) . for observations with _ chandra _ we have selected sources from the sample of dxrbs and rgb fsrq imaged with the vla by @xcite and made an additional constraint that extended radio structure was present . in this case the goal was to study the extended x - ray morphology in addition to the core spectrum . six sources satisfied these criteria and were proposed . we were granted observing time in cycle 6 for all six sources ( pi : landt ) . table [ general ] summarizes the general properties of our sample . for all ten x - ray observed sources we have deep radio observations obtained with the vla at 1.4 ghz in a configuration and for 8/10 sources also in c configuration . these data have been published recently @xcite and we reproduce here for convenience also some of these radio measurements . on - board _ xmm - newton _ we used the european photon imaging camera ( epic ) . in table [ xmm ] we give the journal of observations for the three ccd cameras , mos 1 , mos 2 and pn . all three cameras were operated in small window mode and the thin filter was used . the source wga j1026@xmath66746 was observed twice . however , the first observation taken on 23 march 2004 was strongly affected by high radiation and , therefore , not useful for our purposes . on - board _ chandra _ we used the advanced ccd imaging spectrometer ( acis ) with the back - illuminated s3 chip at the aim - point . all sources were observed with a 1/8 subarray , except for wga j2347@xmath60852 which was observed with a 1/4 subarray . the observations were performed in faint timed mode . in order to avoid that the readout streak falls on the extended jet structures of our sources we additionally imposed roll angle constraints . in table [ chandra ] we give the journal of observations . the epic data were processed using the xmm science analysis software ( sas ; version 6.1.0 ) . the initial data files were reprocessed with the emproc and epproc scripts with default settings , using the latest known calibration files ( as of august 2005 ) . we used x - ray events with patterns @xmath27 and energies in the range @xmath28 kev for the two mos cameras . for the pn , we selected x - ray events with patterns @xmath29 ( single and double pixel events ) and energies in the range @xmath30 kev . the non - x - ray background was relatively high only in the case of wga j0110.5@xmath311647 . for this source we excluded from the analysis of the pn in the energy range @xmath32 kev time intervals with count rates @xmath33 counts s@xmath9 . source spectra were extracted from circular regions of @xmath34 radius for the two mos and @xmath35 radius for the pn . this corresponds to an encircled energy of @xmath36 . background spectra were taken from a similar circular region , offset from the source position . none of the source light curves showed variability over the duration of the observations after the removal of the background . the acis initial event files were reprocessed using the ciao software package ( version 3.3 ) applying the latest known calibration files ( as of february 2006 ) . we filtered for bad grades and used only x - ray events with energies in the range @xmath37 kev . high non - x - ray background periods were not present in our data . source spectra were extracted from circular regions of @xmath38 radius , corresponding to an encircled energy of @xmath39 at 1.5 kev . background spectra were taken from a similar circular region , offset from the source position . none of the source light curves showed variability over the duration of the observations after the removal of the background . the background subtracted epic and acis spectra were fit using xspec ( version 12.2 ) , with the individual response and ancillary matrices produced with sas and ciao , respectively , from the source spectra . spectra were binned to a minimum of 20 counts per bin in order to apply the @xmath40 minimization technique . we initially fit the individual x - ray spectra from the three epic detectors in order to check for consistency between the datasets . from a simple power - law fit we found that the photon indices agreed within @xmath41 and that the relative normalizations were consistent to within 10% . therefore , for all sources we proceeded to fit the three datasets simultaneously , allowing for the individual normalizations to vary . we fit the data both with single and broken power - law models with photo - electric absorption using wisconsin cross - sections from @xcite . the hydrogen column densities @xmath42 were fixed to the galactic value @xcite and in the case of the single power - law fits also allowed to vary in order to check for internal absorption and/or indications of a `` soft excess '' . the results for the single power - law fits to the epic and acis spectra are reported in tables [ epic1 ] and [ acis1 ] , respectively . the results for the broken power - law fits to the epic and acis spectra are reported in tables [ epic2 ] and [ acis2 ] , respectively . in the following we discuss for each object the results of the x - ray spectral analysis . in figs . [ epicfit ] and [ acisfit ] we show for the _ xmm - newton _ and _ chandra _ observed sources , respectively , the best - fit for the model which we consider the most appropriate . this is a single power - law with galactic absorption for 7/10 sources and a broken power - law with galactic absorption for 3/10 sources , namely , wga j0447@xmath310322 , rgb j0112@xmath63818 , and rgb j2229@xmath63057 . _ wga j0110@xmath311647 . @xmath31 _ the epic data are well fit by a single power - law with galactic absorption and a spectral index of @xmath43 . the data suggest that a broken power - law is a better fit ( @xmath44-test @xmath45 ) than a single power - law . but since the break is at relatively high energies ( @xmath46 kev ) , the hard power - law can not be constrained by the present data . the soft spectral index is in this case similar to the one obtained for the single power - law fit . additionally , there is a hint in the data of a soft excess ; the fit is marginally improved ( @xmath44-test @xmath47 ) assuming an absorption lower than galactic , but in this case the @xmath42 is not well constrained . _ wga j0304@xmath60002 . @xmath31 _ the epic data are well fit by a single power - law with galactic absorption and a spectral index of @xmath48 . there is a hint in the data of a soft excess ; the fit is marginally improved ( @xmath44-test @xmath49 ) assuming an absorption lower than galactic . the fit is also marginally improved ( @xmath44-test @xmath47 ) if we assume a broken power - law instead of a single power - law . but in this case both the soft and hard spectral indices are similar to the one obtained for the single power - law . _ wga j0447@xmath310322 . @xmath31 _ the epic data are well fit by a broken power - law with galactic absorption . this fit represents a significant improvement relative to a fit by a single power - law ( @xmath44-test @xmath45 ) . the resulting spectral indices of the soft and hard power - laws are @xmath50 and @xmath51 , respectively , with an observed break at energy @xmath52 kev . _ wga j1026@xmath66746 . @xmath31 _ the epic data are well fit by a single power - law with galactic absorption and a spectral index @xmath53 . _ rgb j0112@xmath63818 . @xmath31 _ the acis data are well fit by a broken power - law with galactic absorption . this fit represents a significant improvement relative to a fit by a single power - law ( @xmath44-test @xmath45 ) . the resulting spectral indices of the soft and hard power - laws are @xmath54 and @xmath51 , respectively , with an observed break at energy @xmath55 kev . _ rgb j0254@xmath63931 . @xmath31 _ the acis data are well fit by a single power - law with galactic absorption and a spectral index @xmath53 . _ rgb j2229@xmath63057 . @xmath31 _ the acis data are well fit by a broken power - law with galactic absorption . this fit represents a significant improvement relative to a fit by a single power - law ( @xmath44-test @xmath45 ) . the resulting spectral indices of the soft and hard power - laws are @xmath50 and @xmath51 , respectively , with an observed break at energy @xmath56 kev . _ rgb j2256@xmath62618 . @xmath31 _ the acis data are well fit by a single power - law with galactic absorption and a spectral index @xmath57 . _ rgb j2318@xmath63048 . @xmath31 _ the acis data are well fit by a single power - law with galactic absorption and a spectral index of @xmath57 . there is a hint in the data of a soft excess ; the fit is marginally improved ( @xmath44-test @xmath58 ) assuming an absorption lower than galactic , but in this case the @xmath42 is not well constrained . the fit is also marginally improved ( @xmath44-test @xmath47 ) if we assume a broken power - law instead of a single power - law . but in this case both the soft and hard spectral indices are similar to the one obtained for the single power - law . _ wga j2347@xmath60852 . @xmath31 _ the acis data are well fit by a single power - law with galactic absorption and a spectral index of @xmath48 . there is a hint in the data of a soft excess ; the fit is marginally improved ( @xmath44-test @xmath59 ) assuming an absorption lower than galactic . the fit is also marginally improved ( @xmath44-test @xmath47 ) if we assume a broken power - law instead of a single power - law . but in this case the break is at relatively high energies ( @xmath60 kev ) and the resulting soft spectral index is similar to the one obtained for the single power - law fit . in order to constrain the origin of the x - ray emission in our sources , and in particular to understand if it is produced by the synchrotron jet component , we have related it to the multiwavelength sed . additionally , we have estimated the flux contribution expected from the two thermal emission components present in radio quasars , namely , the accretion disk and the host galaxy . we have used the following uniform multiwavelength data , which , with the exception of the magnitudes from the optical monitor ( om ) on - board _ xmm - newton _ , are non - simultaneous with our x - ray observations ( but , as noted below , in part simultaneous with each other ) : 1 . radio core fluxes at 1.4 ghz obtained with the vla a array ( from * ? ? ? * reproduced in table [ general ] ) ; 2 . near - ir @xmath61 , @xmath62 and @xmath63 magnitudes ( simultaneous to each other ) from the two micron all sky survey ( 2mass ; * ? ? ? * ) point source and extended source catalogs ( see table [ mag ] , where the extended fluxes are listed in parenthesis ) ; 3 . optical red and blue magnitudes ( simultaneous to each other ) from the automatic plate measuring catalogs ( apm ; * * ) ( see table [ mag ] ) ; 4 . optical magnitudes ( simultaneous to each other ) from the sloan digital sky survey ( sdss ; * ? ? ? * ) data release 6 for wga j0304@xmath60002 , our only source included in this survey ( @xmath64 mag , @xmath65 mag , @xmath66 mag , @xmath67 mag , and @xmath68 mag ) ; 5 . far- and near - uv magnitudes ( simultaneous to each other ) from the galaxy evolution explorer ( galex ; * ? ? ? * ) data release 3 ( see table [ mag ] ) ; 6 . optical and uv magnitudes ( simultaneous to each other ) from the om for the four sources observed with _ xmm - newton _ ( see table [ om ] ) ; 7 . dereddened optical spectra from @xcite and @xcite for the dxrbs sources and from @xcite for the rgb sources ; and 8 . unabsorbed _ rosat _ x - ray flux densities at 1 kev ( see table [ general ] ) . the multiwavelength seds of our sources are shown in figs . [ xmmsed ] and [ chandrased ] for our _ xmm - newton _ and _ chandra _ observed sources , respectively . we have plotted the seds in the observer s frame as log @xmath69 vs. log @xmath70 . in such a representation the peak indicates directly at which frequency most of the energy is emitted . all magnitudes have been corrected for galactic extinction prior to their inclusion in figs . [ xmmsed ] and [ chandrased ] . for this purpose we used the analytical expression for the interstellar extinction curve of @xcite and a parameter of @xmath71 to transform the @xmath72 values to @xmath73 values . we note that the optical spectra were obtained through a relatively small ( @xmath74 ) aperture , and , therefore , in particular for nearby objects , are not expected to account for all the flux from extended emission components , such as , e.g. , the host galaxy ( see section [ host ] ) . in addition to the relativistically beamed , non - thermal jet emission , two thermal emission components are expected to be present in radio quasars : the accretion disk and the host galaxy . we have estimated their contribution to the multiwavelength sed as follows . we have calculated accretion disk spectra assuming a steady geometrically thin , optically thick accretion disk . in this case the emitted flux is independent of viscosity and each element of the disk face radiates roughly as a blackbody with a characteristic temperature , which depends only on the mass of the black hole , @xmath75 , the accretion rate , @xmath76 , and the radius of the innermost stable orbit ( e.g. , * ? ? ? * ; * ? ? ? we have adopted the schwarzschild geometry ( non - rotating black hole ) and for this the innermost stable orbit is at @xmath77 , where @xmath78 is the gravitational radius defined as @xmath79 with @xmath80 the gravitational constant and @xmath81 the speed of light . furthermore , we have assumed that the disk is viewed face - on . the accretion disk spectrum is fully constrained by the two quantities , accretion rate and mass of the black hole , which we have estimated as follows . we have calculated the accretion rate from the luminosity of the broad emission lines using the two relations : ( 1 ) @xmath82 , where @xmath83 and @xmath84 are the total ionizing power of the disk and the efficiency for converting matter to energy , respectively , with @xmath856% in the case of a schwarzschild black hole ; and ( 2 ) @xmath86 , where @xmath87 and @xmath88 are the blr covering factor and luminosity , respectively . we have calculated blr luminosities from the fluxes of the observed broad emission lines as described in @xcite . the optical spectra of our sources cover mostly the broad emission lines @xmath89 and h@xmath90 , and in two cases also h@xmath91 . the blr covering factor is not well known , but is derived to be in the range of @xmath92% ( * ? ? ? * and references therein ) and in general a canonical value of @xmath93% is assumed @xcite . assuming that the blr is gravitationally bound , the mass of the black hole can be estimated based on the virial theorem from the width of a broad emission line and the ionizing luminosity . in this case the ionizing power is used as a surrogate for the blr radius . following @xcite we have calculated black hole masses for 6/10 sources using the full width at half maximum ( fwhm ) of h@xmath90 and the ionizing power @xmath83 , the latter calculated from the blr luminosity as above . for 4/10 sources , for which the optical spectrum does not cover the wavelength of h@xmath90 , we have used the fwhm as a substitute for that of h@xmath90 . as @xcite have shown , the widths of h@xmath90 and are almost identical , indicating that , as expected from their ( similar ) ionization potentials , these two broad emission lines are produced at similar radii . in table [ acctable ] we list for our sources the observed blr luminosities and line widths along with the accretion rates and black hole masses calculated assuming @xmath94% . we note that in two sources , namely , wga j0304@xmath60002 and wga j2347@xmath60852 , h@xmath90 has a clear narrow component , which we have subtracted before measuring the width . in figs . [ xmmsed ] and [ chandrased ] we show accretion disk spectra ( black , dotted curves ) for three sets of accretion rates and black holes masses corresponding to @xmath95 and 30% ( from top to bottom ) . the host galaxies of radio quasars are bright ellipticals that exhibit a relatively narrow range in luminosity ( standard deviation @xmath96 mag ; e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? therefore , we have estimated the contribution from this emission component using the elliptical galaxy template of @xcite , which extends from near - ir to uv frequencies , and assuming a fixed absolute luminosity in the ( rest - frame ) @xmath97-band of @xmath98 mag . this is the average value obtained by @xcite for the host galaxy luminosities of a sample of 41 radio galaxies imaged with the _ hubble space telescope _ in the ( observer s frame ) _ i_-band . the green , solid curves in figs . [ xmmsed ] and [ chandrased ] show our results . we now address the main question of this paper : have we found among this subsample of `` x - ray loud '' radio quasars as expected strong - lined blazars with @xmath3 hz and thus x - rays dominated by synchrotron emission ? in a representation of the sed as a logarithmic @xmath99 plot ( as in figs . [ xmmsed ] and [ chandrased ] ) the two main jet components , synchrotron and inverse compton radiation , form two humps peaking at lower and higher energies , respectively . the synchrotron emission in our sources presumably peaks at uv / soft - x - ray frequencies . therefore , if the x - ray emission was synchrotron in origin , it would be necessarily the part of the synchrotron spectrum after the emission peak and thus curved downward . in the simplest approximation this means that in this case a single power - law with slope @xmath100 is expected . in our sample , 7/10 sources are well fit by a single power - law , however , all with values @xmath101 . therefore , synchrotron x - ray emission is not observed in the large majority of our sources . three sources , namely , wga j0447@xmath310322 , rgb j0112@xmath63818 , and rgb j2229@xmath63057 , are well fit by a broken power - law . the best - fits give in all three cases soft and hard spectral indices @xmath102 and @xmath103 , respectively . this suggests that , as in `` intermediate '' bl lacs , we observe the spectral transition between the synchrotron and inverse compton jet components , which in these radio quasars occurs at relatively low x - ray energies of @xmath104 kev . in the seven sources well fit by a single power - law synchrotron emission does not dominate at x - ray frequencies . therefore , the x - ray emission is either produced by the inverse compton jet component as in `` classical '' fsrq , indicating a relatively low @xmath2 , or by an emission component unrelated to the jet , indicating that the jet is generally weak . the first scenario is excluded by the @xmath105 values of our sources , which are much lower than the typical value of @xmath106 of `` classical '' fsrq @xcite . the latter scenario seems likely . as table [ general ] shows , these fsrq are either lobe - dominated ( i.e. , have @xmath107 , where @xmath108 and @xmath109 are the radio core and extended luminosity , respectively ; 3/7 objects ) or are only slightly core - dominated ( i.e. , have @xmath110 ; 4/7 objects ) , which suggests that their jets are weakly beamed . in this case a @xmath111 hz can not be excluded , but then these sources were not selected as high - energy peaked blazar candidates based on their jet sed . out of the three sources well fit by a broken power - law , only wga j0447@xmath310322 has an sed sampled simultaneously at optical / uv and x - ray frequencies ( by _ xmm - newton _ ) and an extrapolation of its soft x - ray power - law to lower energies predicts fluxes a factor of @xmath112 below the observed om magnitudes . this discrepancy suggests that a single emission component can not account for the fluxes at both frequencies . therefore , either the soft - x - ray flux is synchrotron and the optical / uv fluxes are not , in which case the accretion disk could produce the optical / uv emission ( see section [ selection ] ) , or the optical / uv magnitudes sample the synchrotron jet component ( which has @xmath111 hz ) , but the observed soft x - ray spectral slope is too flat ( we refer here and in the following to `` flat '' and `` steep '' in @xmath113 ) to be synchrotron emission alone . a similar argument for an additional emission component can also be made for rgb j2229@xmath63057 , but less stringently so , since its optical / uv magnitudes and x - ray spectrum are not simultaneous . nevertheless , it is likely that rgb j2229@xmath63057 and wga j0447@xmath310322 are similar cases , since their best - fit x - ray broken power - laws have similarly flat spectral slopes and low break energies . for rgb j0112@xmath63818 , an extrapolation of the soft x - ray power - law to lower energies predicts fluxes similar to the observed optical magnitudes . however , also for this source the optical and x - ray observations are not simultaneous and , moreover , its soft x - ray power - law is not as well constrained as that of wga j0447@xmath310322 and rgb j2229@xmath63057 . the fact that an emission component in addition to synchrotron is required at either optical / uv or soft x - ray frequencies ( or both ) for wga j0447@xmath310322 and rgb j2229@xmath63057 means that these sources are not strong - lined analogs to hbl . on the other hand , if instead the analogy to `` intermediate '' bl lacs holds for the three sources well fit by a broken power - law , as suggested by their x - ray spectrum , we expect their synchrotron emission peak to be at relatively low frequencies ( typically @xmath114 hz ; e.g. , * ? ? ? . given that in none of our sources we detected an x - ray spectrum dominated by jet synchrotron emission the important question arises : why do the selected radio quasars have multiwavelength flux ratios typical of hbl ? our method selected sources based on their high x - ray to radio flux ratios ( low @xmath18 ) as well as high optical to radio flux ratios ( low @xmath16 ) . therefore , if it is not a jet synchrotron spectrum peaking at uv / soft - x - ray frequencies and dominating the x - rays that increases both the optical and x - ray fluxes relative to that at radio frequencies ( as in hbl ) , a different emission component must cause a similar effect . since our sources are radio quasars with strong broad emission lines , the most likely component increasing the optical flux relative to that at radio frequencies is the accretion disk . however , low - redshift sources have usually low - luminosity agn and their optical flux could be instead enhanced by the host galaxy emission . our estimate of the fluxes of these two thermal emission components ( section [ thermal ] ) suggests that the accretion disk can contribute significantly to the optical fluxes of the majority of our sources ( 8/10 objects ; see figs . [ xmmsed ] and [ chandrased ] ) . the host galaxy , on the other hand , appears to dominate the optical magnitudes of only the two most nearby objects , namely , rgb j2256@xmath62618 ( @xmath115 ) and rgb j2318@xmath63048 ( @xmath116 ) . the estimated accretion disk spectrum approximate well the optical spectra of 5/8 , 2/8 and 1/8 objects if a blr covering factor of @xmath117 , 10 and 30% , respectively , is assumed . in 5/8 sources we observe moderate variability ( by factors of @xmath118 ) between the optical spectra and the magnitudes . this variability behavior is compatible with that of radio - quiet quasars ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) and , as in these , it could be due to a change in accretion rate ( * ? ? ? * and references therein ) . in this respect we note that in particular the simultaneous om magnitudes ( and in the case of wga j0304@xmath60002 also the simultaneous sdss magnitudes ) mimic the slope of the estimated accretion disk spectrum . our interpretation that the host galaxy dominates the optical fluxes of the two most nearby objects is supported by the fact that they appear extended on digitized images from the second palomar sky survey ( poss2 ) . additionally , their extended near - ir magnitudes lie well above those from the 2mass point source catalog and comply with the estimated flux and spectral shape of the host galaxy . similarly , their optical spectra mimic the expected spectral shape of the host galaxy , but the small aperture used in spectroscopy underestimates its flux . generally , accretion disk emission is expected to be unimportant in fsrq . their jets are assumed to be strongly beamed and to dominate the emission even at uv frequencies , where the accretion disk spectrum peaks @xcite . however , ( * ? ? ? * see their fig . 11 ) have recently shown that a flat radio spectrum does not always ensure that the radio source is highly core - dominated and thus strongly beamed . this means that in some ( few ) fsrq the accretion disk emission could dominate over the ( weakly beamed ) jet spectrum . as already discussed in section [ uvpeak ] , our sample seems to fall in this category . in this respect we note that the _ chandra _ sources were chosen also because of their extended radio emission . nevertheless , among the _ xmm - newton _ sources , which were chosen based on their estimated high @xmath2 , 2/4 sources are lobe - dominated and one source has log @xmath119 . given that thermal accretion disk emission is pronounced in our sources one can assume that their x - ray fluxes are increased relative to that at radio frequencies by the same component that renders also radio - quiet quasars strong x - ray emitters . in support of the notion that both the optical / uv and x - ray fluxes in our fsrq have a similar ( non - jet ) origin as in radio - quiet quasars are also their @xmath17 values ( see fig . [ aroaox ] ) , which are not inconsistent with those of radio - quiet quasars ( e.g. , * ? ? ? * ; * ? ? ? the x - ray emission of radio - quiet quasars is usually dominated by a power - law with a flat spectral index [ e.g. , @xmath120 @xcite ; @xmath121 @xcite ] , which is generally interpreted as the comptonized hot corona of the accretion disk . the seven sources well fit by a single power - law have an average x - ray spectral index of @xmath122 , similar ( @xmath123 level ) to the averages found for radio - quiet quasars . but more than half of these sources ( 4/7 objects ) show also a hint of a broken power - law in their x - ray spectra ( thin , red solid lines in figs . [ xmmsed ] and [ chandrased ] ; see also section [ discuss ] ) . in particular , in wga j0110@xmath311647 , wga j2347@xmath60852 and rgb j2318@xmath63048 a different ( steep ) component seems to emerge at higher x - ray energies ( @xmath124 , 5 and 3 kev , respectively ) , and in wga j0304@xmath60002 a broken power - law similar in break energy to that observed in wga j0447@xmath310322 and rgb j2229@xmath63057 but with much flatter both soft and hard x - ray spectral indices is evident at a lower significance level . especially the case of wga j0304@xmath60002 suggests that in the selected fsrq the flat x - ray power - law component present in radio - quiet quasars adds onto the jet sed , leading to an increase in total x - ray flux and to a flattening of both soft and hard x - ray spectral slopes . then in sources where this component is particularly strong relative to the jet , such as , e.g. , wga j0110@xmath311647 , the jet inverse compton component can emerge only at higher energies . to a first approximation we expect in this scenario that the stronger the relativistic beaming , the stronger the jet contribution and , therefore , the less flattened the x - ray spectral slopes . and indeed , two of the three sources well fit by a broken power - law are also those with the highest radio core - dominance values in the sample ( log @xmath125 for wga j0447@xmath310322 and log @xmath126 for rgb j0112@xmath63818 ) . the additional flat x - ray power - law will dominate the jet sed in particular around the spectral transition point between the synchrotron and inverse compton components . this is exemplified by our simulations in fig . [ corona ] for which we have used the average blazar sed of @xcite and to which we have added a component with a spectral index of @xmath127 , assuming flux ratios at the jet sed transition point of 1 , 5 , 10 , 20 , 30 , 40 , 50 , 100 , and 150 . therefore , an increased x - ray flux relative to that at radio frequencies is expected to be more pronounced in fsrq whose spectral break is at or around the selection x - ray energy . in this respect we note that 4/10 sources in our sample and 2/4 sources observed by @xcite have a detected jet sed transition point around @xmath128 kev ( the x - ray energy used to calculate the @xmath18 and @xmath21 values ) . @xcite and this work presented x - ray spectroscopy for a total of 14 `` x - ray loud '' radio quasars and only one source , namely , rgb j1629@xmath64008 , had x - rays dominated by synchrotron radiation . we have shown that this modest success in finding radio quasars with synchrotron x - rays is due to the adopted selection criteria , which efficiently select bl lacs with synchrotron x - rays , but do not work well for radio quasars . nevertheless , the important question remains : does a large population of radio quasars with synchrotron x - rays exist , and , if yes , how can they be efficiently found ? an efficient selection of radio quasars candidate to have synchrotron x - rays could be based on the diagram presented by ( * ? ? ? * see their fig . 12 ) , which plots log @xmath129 versus log @xmath97 , where @xmath108 and @xmath130 are the radio core and total x - ray luminosity , respectively . in this plot high - energy peaked fsrq are expected to separate from the rest of the population , since they should be those strongly core - dominated radio quasars ( i.e. , with log @xmath131 ) that have the lowest @xmath129 ratios . for bl lacs values of log @xmath132 are known to be indicative of synchrotron x - rays @xcite . however , based on the results in the previous section lower values must be chosen for radio quasars . e.g. , whereas rgb j1629@xmath64008 , which is strongly core - dominated ( log @xmath133 ) , has a ratio of log @xmath134 @xcite , the two strongly core - dominated sources in our sample , namely , wga j0447@xmath310322 and rgb j0112@xmath63818 , have ratios of log @xmath135 and @xmath136 @xcite , respectively , which are apparently not low enough to be indicative of synchrotron x - rays . but it remains to be shown if radio quasars with synchrotron x - rays exist in large numbers and both a positive and a negative result will reveal highly significant agn physics . a negative result could tell us that the so - called `` blazar sequence '' @xcite holds to some level @xcite , which would mean that the particles in the relativistic jets of agn strongly interact with the ambient photon field , such as the one produced by , e.g. , the accretion disk - blr complex , already in the most inner regions where they are produced . a positive result , on the other hand , could mean that the basic properties of agn jets , such as , e.g. , their powers , magnetic field strengths or velocities , are not determined by the presence or absence of emission line regions in the nuclear regions . finding radio quasars with synchrotron x - rays in large numbers will also provide present and up - coming @xmath137-ray missions , such as agile and the gamma ray large area space telescope ( glast ) , with plenty of targets . a synchrotron spectrum peaking at uv / soft - x - ray energies produces an inverse compton spectrum with peak at @xmath137-ray frequencies . but such jets have been detected so far only in a ( small ) part of the bl lac population , the hbl . since at the high fluxes required by the sensitivity of these missions radio quasars are by far more abundant than bl lacs ( e.g. , * ? ? ? * ) , high - energy peaked fsrq instead could become the prime targets . a considerable fraction of fsrq discovered in the dxrbs and rgb survey have multiwavelength flux ratios similar to those of bl lacs with synchrotron x - rays . however , the definite proof of the synchrotron nature of their x - ray emission requires x - ray spectroscopy . @xcite observed four `` x - ray loud '' radio quasars and found one object ( rgb j1629@xmath64008 ) to have x - rays dominated by synchrotron radiation . in this paper we have presented x - ray spectroscopy of a further ten of these sources . the main results are : \1 . the x - ray spectrum of 7/10 sources is well fit by a single power - law with spectral index @xmath138 , indicating that synchrotron x - ray emission is not the dominant component . the remaining three sources are well fit by a broken power - law with soft and hard spectral indices @xmath139 and @xmath140 , respectively , which , as in `` intermediate '' bl lacs , suggests that we observe the spectral transition between the synchrotron and inverse compton jet components . the lack of success in finding radio quasars with x - ray spectra dominated by jet synchrotron emission can be attributed to the employed selection method . this was developed for bl lacs and requires the sources to have high both optical / uv and x - ray fluxes relative to that at radio frequencies ( low @xmath18 and @xmath16 values ) . in the case of radio quasars these criteria yield predominantly those ( few ) sources that have a thermal and non - thermal accretion disk component strong enough to dominate over the ( weakly beamed ) jet emission . the majority of the observed fsrq are either lobe - dominated ( i.e. , have log @xmath141 ; 4/10 objects ) or are only slightly core - dominated ( i.e. , have @xmath110 ; 4/10 objects ) . the discovered radio quasars with their relatively low @xmath142 values and strong accretion disk signatures represent a population intermediate between `` classical '' fsrq and radio - quiet quasars . this makes them unique probes for studying the poorly known accretion disk - jet connection . \4 . a refined selection technique based on the work of @xcite is proposed to be used to search for high - energy peaked fsrq , which , since their inverse compton emission is expected to peak at @xmath137-ray frequencies , could be prime targets for glast and agile . the recipe is : ( i ) choose highly core - dominated radio quasars , and among these ( ii ) choose radio quasars with low radio core to x - ray luminosity ratios of log @xmath143 . h.l . acknowledges financial support from the deutsche akademie der naturforscher leopoldina ( grant no . bmbf - lpd 9901/8 - 99 ) and from nasa ( grants no . nng04gn17 g and go5 - 6102x ) . thanks the european southern observatory and st . john s college , oxford , where part of this work was conducted , for their hospitality . we thank sally laurent - muehleisen for providing the spectra of sources from the rgb survey in electronic format . giommi , p. , capalbi , m. , fiocchi , m. , memola , e. , perri , m. , piranomonte , s. , rebecchi , s. , & massaro , e. 2002 , in blazar astrophysics with bepposax and other observatories , ed . e. m. p. giommi & g. palumbo ( asi / esa ) , 63 perlman , e. s. , padovani , p. , landt , h. , stocke , j. t. , costamante , l. , rector , t. , giommi , p. , & schachter , j. f. 2001 , in blazar demographics and physics , ed . p. padovani & c. m. urry ( a.s.p . , san francisco ) , 200 lccccrrrcccc rgb j0112@xmath63818 & 01 12 18.049 & @xmath638 18 56.90 & 0.333 & 0.096 & @xmath1440.09 & 32.5 & @xmath1440.65 & 0.38 & 1.38 & 0.72 & 5.64 + rgb j0254@xmath63931 & 02 54 42.629 & @xmath639 31 34.75 & 0.291 & 0.344 & @xmath310.31 & 160.2 & @xmath1440.24 & 0.44 & 1.39 & 0.76 & 9.54 + rgb j2229@xmath63057 & 22 29 34.151 & @xmath630 57 12.10 & 0.322 & 0.203 & @xmath1440.15 & 44.2 & @xmath310.52 & 0.40 & 1.42 & 0.74 & 6.79 + rgb j2256@xmath62618 & 22 56 39.163 & @xmath626 18 43.55 & 0.121 & 0.208 & @xmath310.02 & 21.4 & @xmath1440.19 & 0.43 & 1.19 & 0.69 & 5.20 + rgb j2318@xmath63048 & 23 18 36.905 & @xmath630 48 37.00 & 0.103 & 0.185 & @xmath1440.12 & 17.5 & @xmath1440.12 & 0.29 & 1.35 & 0.65 & 6.34 + wga j0110@xmath311647 & 01 10 35.516 & @xmath3116 48 27.80 & 0.781 & 0.176 & @xmath1440.35 & 72.0 & @xmath1440.22 & 0.36 & 1.35 & 0.69 & 1.62 + wga j0304@xmath60002 & 03 04 58.973 & @xmath600 02 35.85 & 0.563 & 0.121 & @xmath1440.40 & 8.4 & @xmath311.22 & 0.50 & 1.21 & 0.74 & 7.00 + wga j0447@xmath310322 & 04 47 54.727 & @xmath3103 22 42.20 & 0.774 & 0.348 & @xmath1440.47 & 68.0 & @xmath1452.63 & 0.32 & 1.31 & 0.66 & 4.01 + wga j1026@xmath66746 & 10 26 33.850 & @xmath667 46 12.10 & 1.181 & 0.069 & @xmath1440.49 & 84.7 & @xmath310.50 & 0.51 & 1.25 & 0.76 & 2.18 + wga j2347@xmath60852 & 23 47 38.144 & @xmath608 52 46.35 & 0.292 & 0.176 & @xmath1440.58 & 8.3 & @xmath311.30 & 0.38 & 1.34 & 0.71 & 6.02 + cccccccccccc wga j0110@xmath311647 & 7506 & 3106 & 812 & 7508 & 3787 & 1004 & 7242 & 4233 & 5785 & 0203160401 & 12/17/2003 + wga j0304@xmath60002 & 15157 & 14638 & 939 & 15162 & 14642 & 928 & 14964 & 10432 & 2733 & 0203160201 & 07/19/2004 + wga j0447@xmath310322 & 7657 & 7439 & 1131 & 7662 & 7443 & 1162 & 7464 & 5238 & 3129 & 0203160101 & 03/06/2004 + wga j1026@xmath66746 & 9957 & 9571 & 433 & 9962 & 9576 & 412 & 9764 & 6779 & 1420 & 0203160601 & 04/19/2004 + lcccccc rgb j0112@xmath63818 & 16183 & 13815 & 1897 & 172 & 5641 & 10/07/2005 + rgb j0254@xmath63931 & 10708 & 8613 & 1584 & 153 & 5638 & 10/26/2005 + rgb j2229@xmath63057 & 10628 & 9235 & 3884 & 91 & 5639 & 05/04/2005 + rgb j2256@xmath62618 & 9854 & 8936 & 2042 & 91 & 5642 & 05/04/2005 + rgb j2318@xmath63048 & 11275 & 9219 & 1168 & 316 & 5643 & 01/25/2005 + wga j2347@xmath60852 & 16675 & 14440 & 4936 & 128 & 5640 & 08/20/2005 + lllccll wga j0110@xmath311647 & 1.62 fixed & 1.96@xmath1460.04 & 1.44e@xmath3112 & 1.19e@xmath3112 & 1.08/306 & + & 0.62@xmath147 & 1.88@xmath148 & 1.48e@xmath3112 & 1.28e@xmath3112 & 1.07/305 & 0.011 + wga j0304@xmath60002 & 7.00 fixed & 1.93@xmath1460.05 & 2.98e@xmath3113 & 3.53e@xmath3113 & 0.92/195 & + & 4.90@xmath149 & 1.83@xmath150 & 2.96e@xmath3113 & 3.75e@xmath3113 & 0.90/194 & 0.019 + wga j0447@xmath310322 & 4.01 fixed & 2.11@xmath1460.04 & 7.31e@xmath3113 & 5.63e@xmath3113 & 1.11/219 & + & 1.79@xmath151 & 1.97@xmath148 & 7.60e@xmath3113 & 6.34e@xmath3113 & 1.04/218 & 7.10e@xmath3105 + wga j1026@xmath66746 & 2.18 fixed & 1.76@xmath1460.08 & 1.99e@xmath3113 & 2.39e@xmath3113 & 1.06/97 & + & 0.60@xmath152 & 1.68@xmath153 & 2.08e@xmath3113 & 2.55e@xmath3113 & 1.05/96 & 0.171 + lllccll rgb j0112@xmath63818 & 5.64 fixed & 1.89@xmath1460.08 & 4.49e@xmath3113 & 5.44e@xmath3113 & 1.07/69 & + & 4.36e@xmath3108@xmath154 & 1.67@xmath150 & 4.77e@xmath3113 & 6.25e@xmath3113 & 0.88/68 & 2.32e@xmath3104 + rgb j0254@xmath63931 & 9.54 fixed & 1.80@xmath1460.08 & 5.74e@xmath3113 & 8.99e@xmath3113 & 0.83/62 & + & 6.85@xmath155 & 1.71@xmath156 & 5.80e@xmath3113 & 9.47e@xmath3113 & 0.81/61 & 0.163 + rgb j2229@xmath63057 & 6.79 fixed & 2.04@xmath1460.05 & 1.39e@xmath3112 & 1.40e@xmath3112 & 1.31/128 & + & 4.84e@xmath3107@xmath157 & 1.76@xmath158 & 1.48e@xmath3112 & 1.71e@xmath3112 & 0.98/127 & 8.69e@xmath3110 + rgb j2256@xmath62618 & 5.20 fixed & 1.66@xmath1460.06 & 7.12e@xmath3113 & 1.18e@xmath3112 & 1.07/77 & + & 4.85@xmath159 & 1.65@xmath160 & 7.13e@xmath3113 & 1.19e@xmath3112 & 1.08/76 & 0.789 + rgb j2318@xmath63048 & 6.34 fixed & 1.67@xmath1460.09 & 3.93e@xmath3113 & 6.70e@xmath3113 & 0.78/46 & + & 1.91@xmath161 & 1.52@xmath162 & 4.04e@xmath3113 & 7.38e@xmath3113 & 0.72/45 & 0.039 + wga j2347@xmath60852 & 6.02 fixed & 1.86@xmath1460.04 & 1.10e@xmath3112 & 1.40e@xmath3112 & 1.34/147 & + & 3.83@xmath163 & 1.78@xmath1460.07 & 1.12e@xmath3112 & 1.49e@xmath3112 & 1.30/146 & 0.031 + llllcccll wga j0110@xmath311647 & 1.62 fixed & 1.97@xmath1460.04 & @xmath313.00(unc . ) & 6.91@xmath164 & 1.43e@xmath3112 & 1.68e@xmath3112 & 1.03/304 & 1.29e@xmath3104 + wga j0304@xmath60002 & 7.00 fixed & 2.07@xmath165 & 1.81@xmath166 & 1.26@xmath167 & 2.92e@xmath3113 & 3.78e@xmath3113 & 0.89/193 & 0.012 + wga j0447@xmath310322 & 4.01 fixed & 2.28@xmath168 & 1.74@xmath169 & 1.62@xmath170 & 7.26e@xmath3113 & 7.27e@xmath3113 & 0.94/217 & 9.71e@xmath3109 + wga j1026@xmath66746 & 2.18 fixed & 1.96@xmath171 & 1.66@xmath172 & 0.99@xmath173 & 2.02e@xmath3113 & 2.61e@xmath3113 & 1.05/95 & 0.246 + llllcccll rgb j0112@xmath63818 & 5.64 fixed & 2.57@xmath174 & 1.74@xmath1460.09 & 0.86@xmath175 & 4.72e@xmath3113 & 6.18e@xmath3113 & 0.82/67 & 6.89e@xmath3105 + rgb j0254@xmath63931 & 9.54 fixed & 1.97@xmath176 & 1.72@xmath177 & 1.17(unc . ) & 5.76e@xmath3113 & 9.50e@xmath3113 & 0.81/60 & 0.223 + rgb j2229@xmath63057 & 6.79 fixed & 2.31@xmath178 & 1.70@xmath179 & 1.50@xmath180 & 1.40e@xmath3112 & 1.78e@xmath3112 & 1.02/126 & 3.79e@xmath3108 + rgb j2256@xmath62618 & 5.20 fixed & 1.69@xmath181 & 1.59@xmath182 & 2.34(unc . ) & 7.10e@xmath3113 & 1.22e@xmath3112 & 1.09/75 & 0.775 + rgb j2318@xmath63048 & 6.34 fixed & 1.78@xmath183 & 1.00@xmath184 & 3.01@xmath185 & 3.91e@xmath3113 & 8.72e@xmath3113 & 0.65/44 & 0.007 + wga j2347@xmath60852 & 6.02 fixed & 1.90@xmath1460.05 & 0.60@xmath186 & 4.64@xmath187 & 1.10e@xmath3112 & 1.82e@xmath3112 & 1.28/145 & 0.013 + lccc|cc|cc rgb j0112@xmath63818 & 15.5 & 14.6 & 13.9 & 17.3 & 16.5 & & + rgb j0254@xmath63931 & 15.9 ( 16.1 ) & 15.3 ( 15.4 ) & 14.4 ( 13.9 ) & 16.4 & 15.4 & & + rgb j2229@xmath63057 & 15.5 & 14.7 & 14.3 & 16.4 & 15.8 & & + rgb j2256@xmath62618 & 15.9 ( 15.0 ) & 15.1 ( 13.8 ) & 14.6 ( 13.8 ) & 18.3 & 16.7 & 21.7 & 21.0 + rgb j2318@xmath63048 & 15.2 ( 14.3 ) & 14.4 ( 13.9 ) & 13.9 ( 13.3 ) & 17.3 & & 20.2 & 19.5 + wga j0110@xmath311647 & 15.0 & 14.7 & 14.0 & 15.9 & 16.2 & 17.2 & 16.5 + wga j0304@xmath60002 & 16.5 & 15.9 & 15.2 & & 18.4 & 19.7 & 19.1 + wga j0447@xmath310322 & 14.7 & 14.5 & 13.8 & 16.3 & 16.0 & & + wga j2347@xmath60852 & 15.6 ( 15.2 ) & 14.8 ( 14.2 ) & 13.8 ( 13.5 ) & 17.1 & 16.2 & 18.0 & 17.7 + lcccccc wga j0110@xmath311647 & 16.42@xmath1460.01 & 16.642@xmath1460.007 & 15.587@xmath1460.006 & 15.241@xmath1460.008 & & 12/17/2003 + wga j0304@xmath60002 & 18.76@xmath1460.02 & 19.04@xmath1460.07 & 18.07@xmath1460.05 & 17.75@xmath1460.05 & & 07/19/2004 + wga j0447@xmath310322 & 16.33@xmath1460.03 & 16.66@xmath1460.01 & 15.76@xmath1460.01 & & & 03/06/2004 + wga j1026@xmath66746 & 18.41@xmath1460.06 & & 17.49@xmath1460.02 & 17.07@xmath1460.02 & 17.23@xmath1460.05 & 03/23/2004 + & & & & 17.09@xmath1460.03 & & 04/19/2004 + lcccccc rgb j0112@xmath63818 & 2.7e@xmath644 & 0.8 & 3144 & & 3e@xmath608 & 0.08 + rgb j0254@xmath63931 & 1.4e@xmath644 & 0.4 & 3244 & h@xmath90 & 2e@xmath608 & 0.05 + rgb j2229@xmath63057 & 2.6e@xmath644 & 0.8 & 5467 & h@xmath90 & 8e@xmath608 & 0.03 + rgb j2256@xmath62618 & 6.9e@xmath642 & 0.02 & 7888 & h@xmath90 & 3e@xmath608 & 0.002 + rgb j2318@xmath63048 & 5.9e@xmath642 & 0.02 & 6645 & h@xmath90 & 2e@xmath608 & 0.003 + wga j0110@xmath311647 & 1.5e@xmath645 & 5 & 3853 & & 1e@xmath609 & 0.1 + wga j0304@xmath60002 & 2.7e@xmath644 & 0.8 & 7404 & h@xmath90 & 2e@xmath609 & 0.01 + wga j0447@xmath310322 & 6.5e@xmath645 & 19 & 4595 & & 3e@xmath609 & 0.2 + wga j1026@xmath66746 & 1.1e@xmath646 & 32 & 6646 & & 8e@xmath609 & 0.1 + wga j2347@xmath60852 & 8.1e@xmath643 & 0.2 & 7926 & h@xmath90 & 9e@xmath608 & 0.005 +

* abstract * + the horton - strahler ordering method , originating in hydrology , formulates the hierarchical structure of branching patterns using a quantity called the bifurcation ratio . the main result of this paper is the central limit theorem for bifurcation ratio of general branch order . this is a generalized form of the central limit theorem for the lowest bifurcation ratio , which was previously proved . some useful relations are also derived in the proofs of the main theorems .

1701.03213

branching objects are found very widely @xcite , ranging from natural patterns like river networks , plants , and dendritic crystals , to conceptual expressions like binary search trees in computer science @xcite and phylogenetic trees in taxonomy @xcite . the topological structure of a branching pattern is modeled by a binary tree if a segment bifurcates ( does not trifurcate or more ) at every branching point . let @xmath0 denote the set of the different binary trees having @xmath1 leaves . the number of leaves is called the _ magnitude _ in research of branching patterns . as known well @xcite , the number of the different binary trees of magnitude @xmath1 is given by @xmath2 which iscalled the @xmath3st catalan number . in fig . [ fig2 ] , @xmath0 for @xmath4 , and 4 are schematically shown . introducing the uniform probability measure @xmath5 on @xmath0 ( so that each binary tree is assigned equal probability @xmath6 ) , we obtain the probability space @xmath7 referred to as the _ random model _ @xcite . the formation of real - world branching patterns more or less involves stochastic effects , and the random model is a kind of mathematical simplification of such random factors . in hydrology , methods for measuring the hierarchical structure of a river network have been proposed by horton @xcite , strahler @xcite , shreve @xcite , tokunaga @xcite , and other researchers . their methods define how to assign an integer number ( called the _ order _ ) to each stream . among all , strahler s method is currently the most popular because of its simple computation rule . strahler s method is a refinement of horton s method , so it is sometimes called the _ horton - strahler ordering method_. the horton - strahler method recursively defines the order of each node by the following rules . ( i ) the leaf nodes are defined to have order one . ( ii ) a node whose children have different order @xmath8 and @xmath9 ( @xmath10 ) has order @xmath11 . ( iii ) a node whose two children have the same order @xmath12 has order @xmath13 . we define a _ branch _ of order @xmath12 as a maximal connected path made by nodes of equal order @xmath12 . ( a branch here is called a _ stream _ in the analysis of river networks . ) an example of strahler s ordering is shown in fig . [ fig1 ] . for a binary tree @xmath14 , we let @xmath15 denote the number of branches of order @xmath12 in @xmath16 . by the definition of the order , @xmath17 and @xmath18 ( @xmath19 ) . note that @xmath20 if @xmath21 , because a node of order 2 is produced by the merge of two leaves . for the binary tree @xmath22 in fig . [ fig1 ] , @xmath23 , @xmath24 , @xmath25 , and @xmath26 for @xmath27 . @xmath28 is a random variable on @xmath7 , and its stochastic property is of main interest in this study . for any function @xmath29 , @xmath30 is a real - valued random variable on @xmath0 . according to ref . @xcite , the recursive relation between the averages of the @xmath12th and @xmath31st variables @xmath32 } = \frac{n!(n-1)!(n-2)!}{(2n-2)!}\sum_{m=1}^{{\lfloorn/2\rfloor } } \frac{2^{n-2m}}{(n-2m)!m!(m-1)!}{e\left[f(s_{r-1,m})\right ] } \label{eq:1 - 1}\ ] ] holds , where @xmath33}$ ] denotes the average on the random model . the coefficient @xmath34 represents the probability @xmath35 . in particular , putting @xmath36 in eq . , we have @xmath37}=\frac{n!(n-1)!(n-2)!}{(2n-2)!}\sum_{m=1}^{{\lfloorn/2\rfloor}}\frac{2^{n-2m}}{(n-2m)!m!(m-1)!}f(m ) . \label{eq2}\ ] ] mathematical properties of @xmath38 have been investigated thoroughly . for instance , the average and variance are respectively given by @xcite @xmath39 } = \frac{n(n-1)}{2(2n-3)},\quad { \operatorname{var}\left[s_{2,n}\right ] } = \frac{n(n-1)(n-2)(n-3)}{2(2n-3)^2(2n-5)}. \label{eq : werner}\ ] ] moreover , from eq . , the moment generating function @xmath40 of @xmath38 is given by @xmath41 } = \frac{n!(n-1)!(n-2)!}{(2n-2)!}\sum_{m=1}^{{\lfloorn/2\rfloor}}\frac{2^{n-2m}}{(n-2m)!m!(m-1)!}e^{mt},\ ] ] and this summation can be expressed using the gauss hypergeometric function @xmath42 @xcite : @xmath43 the ratio @xmath44 is called the _ bifurcation ratio _ of order @xmath12 or simply the @xmath12th bifurcation ratio . hydrologists have empirically confirmed that the bifurcation ratios of an actual river network become almost constant for different orders , and this relation is referred to as _ horton s law of stream numbers_. by definition , the bifurcation ratio is always smaller than or equal to @xmath45 . when @xmath46 , we reasonably define @xmath47 . the random variable @xmath48 is also called the @xmath12the bifurcation ratio . the lowest bifurcation ratio @xmath49 is relatively easy to deal with , because it is similar to @xmath38 . the central limit theorem for @xmath50 has been shown by wang and waymire @xcite : [ thm1 ] on the random model , @xmath51 where `` @xmath52 '' denotes convergence in distribution , and @xmath53 is the normal distribution with mean @xmath54 and variance @xmath55 . it is a simple and natural idea that we extend theorem [ thm1 ] to general order @xmath12 . compared with @xmath38 , however , higher - order branches @xmath28 for @xmath56 and the bifurcation ratio of order @xmath19 is difficult to handle and less studied . in this paper , we generalize theorem [ thm1 ] in two ways ( theorems [ thm2 ] and [ thm3 ] in [ sec2 ] ) , and further generalize them ( theorem [ thm4 ] in [ sec6 ] ) . in [ sec3][sec5 ] , we give proofs of lemmas , which are necessary for the main theorems . in these proofs , eq . and its variant @xmath57 } = \frac{n!(n-1)!(n-2)!}{(2n-2)!}\sum_{m=1}^{{\lfloorn/2\rfloor } } \frac{2^{n-2m}}{(n-2m)!m!(m-1)!}{e\left[f\left(\frac{s_{r , m}}{s_{r-1,m}}\right)\right]}. \label{eq:1 - 3}\ ] ] are very useful . the following two theorems are the main results of the present paper . [ thm2 ] for any order @xmath58 , the @xmath12th bifurcation ratio @xmath48 satisfies @xmath59 [ thm3 ] for any order @xmath58 , the number @xmath60 of @xmath13st branches satisfies @xmath61 these two theorems are generalization of theorem [ thm1 ] to general order @xmath12 ; they are reduced to theorem [ thm1 ] by setting @xmath62 . theorem [ thm2 ] states the property of the bifurcation ratio @xmath48 , and theorem [ thm3 ] states the property of the number of branches @xmath60 . the limit variance @xmath63 in theorem [ thm2 ] becomes large as @xmath12 increases , whereas the limit variance in theorem [ thm3 ] becomes small as @xmath12 increases . from theorem [ thm2 ] , the following property , which can be regarded as horton s law of stream numbers , is easily derived . [ cor1 ] for any order @xmath64 , the @xmath12th bifurcation ratio @xmath48 converges in probability to the common value @xmath65 : @xmath66 where `` @xmath67 '' denotes convergence in probability . let us introduce the asymptotic equality , since this study mainly focuses on the asymptotic behavior ( the limit @xmath68 ) of @xmath28 . the average value @xmath69}$ ] is asymptotically equivalent to @xmath70 if @xmath71}}{g_r(n)}=1,\ ] ] and this is denoted by @xmath32}\sim g_r(n).\ ] ] for example , from eq . , @xmath39}=\frac{n(n-1)}{2(2n-3)}\sim\frac{n}{4}.\ ] ] theorems [ thm2 ] and [ thm3 ] are easily proved by using the following lemmas [ lem1 ] and [ lem2 ] , respectively . [ lem1 ] for @xmath72 and @xmath73 , @xmath74 } \sim \frac{(2s-1)!!}{4^{2s}}\left(\frac{n}{4^{r-1}}\right)^{-s},\quad { e\left[\left(\frac{s_{r+1,n}}{s_{r , n}}-\frac{1}{4}\right)^{2s+1}\right ] } \sim \frac{(2s+1)!!}{2\cdot 4^{2s+1}}\left(\frac{n}{4^{r-1}}\right)^{-s-1}\ ] ] [ lem2 ] for @xmath72 and @xmath73 , @xmath75 } & \sim \frac{(2s-1)!!}{4^{2sr}}\left(\frac{4^r-1}{3}\right)^s n^s,\\ { e\left[\left(s_{r+1,n}-\frac{n}{4^r}\right)^{2s+1}\right ] } & \sim \frac{(2s+1)!!}{2\cdot 4^{(2s+1)r}}\left(\frac{4^r-1}{3}\right)^s\frac{1}{5}\left(\frac{4^{r+1}-1}{3}+\frac{4^{r-1}-1}{3}4(2s+1)\right)n^s.\end{aligned}\ ] ] the odd - power result has a more complicated form than the even - power one . we let @xmath76 denote the characteristic function of the left - hand side of eq . , where the subscript @xmath12 and superscript @xmath13 respectively correspond to @xmath28 in the denominator and @xmath60 in the numerator in eq . . by definition , @xmath76 is calculated as @xmath77}\\ & = { e\left[\sum_{k=0}^\infty\frac{(iz\sqrt{n})^k}{k!}\left(\frac{s_{r+1,n}}{s_{r , n}}-\frac{1}{4}\right)^k\right]}\\ & = \sum_{k=0}^\infty\frac{(iz\sqrt{n})^k}{k!}{e\left[\left(\frac{s_{r+1,n}}{s_{r , n}}-\frac{1}{4}\right)^k\right]}\\ & = \sum_{s=0}^\infty\frac{(iz\sqrt{n})^{2s}}{(2s)!}{e\left[\left(\frac{s_{r+1,n}}{s_{r , n}}-\frac{1}{4}\right)^{2s}\right ] } + \sum_{s=0}^\infty\frac{(iz\sqrt{n})^{2s+1}}{(2s+1)!}{e\left[\left(\frac{s_{r+1,n}}{s_{r , n}}-\frac{1}{4}\right)^{2s+1}\right]},\end{aligned}\ ] ] where @xmath78 . at the last equality , we have split the sum into even @xmath79 ( @xmath80 ) and odd @xmath79 ( @xmath81 ) . by lemma [ lem1 ] , the terms of the first sum ( even @xmath79 ) are @xmath82 , whereas the terms of the second sum ( odd @xmath79 ) are @xmath83 . hence , the second sum can be neglected in the limit @xmath68 , so that @xmath84}\\ & \sim\sum_{s=0}^\infty\frac{(-z^2n)^s}{(2s)!}\frac{(2s-1)!!}{4^{2s}}\frac{n^{-s}}{4^{-(r-1)s}}\\ & = \sum_{s=0}^\infty\left(-\frac{4^{r-3}z^2}{2}\right)^s\frac{1}{s!}\\ & = \exp\left(-\frac{4^{r-3}z^2}{2}\right).\end{aligned}\ ] ] recall that the characteristic function of @xmath53 is @xmath85 . since @xmath86 converges pointwise to the characteristic function of @xmath87 , convergence in distribution in theorem [ thm2 ] is proved . ( for the properties of a characteristic function , see feller @xcite for example . ) keep in mind that the neglect of the odd - power terms is a crucial point also in the other central limit theorems in this paper . as with the above proof of theorem [ thm2 ] , the characteristic function @xmath88 of the left - hand side of eq . is @xmath89}\\ & = \sum_{k=0}^\infty\frac{(iz\sqrt{n})^k}{k!}{e\left[\left(\frac{s_{r+1,n}}{n}-\frac{1}{4^r}\right)^k\right]}\\ & = \sum_{s=0}^\infty\frac{(iz\sqrt{n})^{2s}}{(2s)!n^{2s}}{e\left[\left(s_{r+1,n}-\frac{n}{4^r}\right)^{2s}\right ] } + \sum_{s=0}^\infty\frac{(iz\sqrt{n})^{2s+1}}{(2s+1)!n^{2s+1}}{e\left[\left(s_{r+1,n}-\frac{n}{4^r}\right)^{2s+1}\right]}.\\\end{aligned}\ ] ] by lemma [ lem2 ] , the second sum is neglected and the dominant terms are calculated to @xmath90}\\ & \sim\sum_{s=0}^\infty\frac{(-z^2n)^s}{(2s)!n^{2s}}\frac{(2s-1)!!}{4^{2rs}}\left(\frac{4^r-1}{3}\right)^s n^s\\ & = \sum_{s=0}^\infty\left(-\frac{z^2}{2}\frac{4^r-1}{3\cdot16^r}\right)^s\frac{1}{s!}\\ & = \exp\left(-\frac{z^2}{2}\frac{4^r-1}{3\cdot16^r}\right).\end{aligned}\ ] ] therefore , the converges in distribution to @xmath91 is proved . we give the proofs of lemmas [ lem1 ] and [ lem2 ] in the following three sections . we show lemmas [ lem1 ] and [ lem2 ] by induction on @xmath12 . in this section , the case of @xmath62 in lemmas [ lem1 ] and [ lem2 ] is proved ( cor . [ cor3 ] ) . [ prop1 ] for a two - variable polynomial @xmath92 of finite degree , @xmath93}= \frac{n}{2}{e\left[p(s_{2,n } , n)\right]}-\frac{n(n-2)}{2(2n-3)}{e\left[p(s_{2,n-1 } , n)\right]}. \label{eq : prop1}\ ] ] here , @xmath94}$ ] and @xmath95}$ ] are taken over @xmath0 , whereas @xmath96}$ ] are over @xmath97 . because of the linearity of @xmath33}$ ] , it is sufficient to check the case @xmath98 . the average is expressed using the moment generating function @xmath40 in eq . . @xmath99}&=n^l{e\left[s_{2,n}^{k+1}\right ] } = n^l\left.\frac{d^{k+1}}{dt^{k+1}}m_{2,n}(t)\right|_{t=0}\\ & = n^l \frac{2^{n-2}n!(n-1)!}{(2n-2)!}\left.\frac{d^{k+1}}{dt^{k+1}}e^t{f\left(\frac{2-n}{2 } , \frac{3-n}{2 } , 2 ; e^t\right)}\right|_{t=0}\\ & = n^l\frac{2^{n-2}n!(n-1)!}{(2n-2)!}\left.\frac{d^k}{dt^k}e^t\left[{f\left(\frac{2-n}{2 } , \frac{3-n}{2 } , 2 ; e^t\right)}+\frac{d}{dt}{f\left(\frac{2-n}{2 } , \frac{3-n}{2 } , 2 ; e^t\right)}\right]\right|_{t=0}.\end{aligned}\ ] ] using the derivative of the gauss hypergeometric function @xcite @xmath100\ ] ] and the symmetry @xmath101 , we have @xmath102 , \label{eq : derivative}\ ] ] thereby @xmath99 } & = n^l \frac{2^{n-2}n!(n-1)!}{(2n-2)!}\frac{d^k}{dt^k}e^t\left[\frac{n}{2}{f\left(\frac{2-n}{2 } , \frac{3-n}{2 } , 2 ; e^t\right)}-\frac{n-2}{2}{f\left(\frac{3-n}{2 } , \frac{4-n}{2 } , 2 ; e^t\right)}\right]\\ & = n^l \left(\frac{n}{2}{e\left[s_{2,n}^k\right]}-\frac{2n(n-1)}{(2n-2)(2n-3)}\frac{n-2}{2}{e\left[s_{2,n-1}^k\right]}\right)\\ & = \frac{n}{2}{e\left[s_{2,n}^k n^l\right]}-\frac{n(n-2)}{2(2n-3)}{e\left[s_{2,n-1}^k n^l\right]}.\end{aligned}\ ] ] here we comment on the utility of prop . [ prop1 ] . we can calculate the moments @xmath103}$ ] recursively using eq . : @xmath104}&={e\left[s_{2,n}\cdot1\right]}=\frac{n}{2}{e\left[1\right]}-\frac{n(n-2)}{2(2n-3)}{e\left[1\right]}=\frac{n(n-1)}{2(2n-3)},\\ { e\left[s_{2,n}^2\right]}&={e\left[s_{2,n}s_{2,n}\right]}=\frac{n}{2}{e\left[s_{2,n}\right]}-\frac{n(n-1)}{2(2n-3)}{e\left[s_{2,n-1}\right]}=\frac{n(n-1)(n^2-n-4)}{4(2n-3)(2n-5)},\\ { e\left[s_{2,n}^3\right]}&={e\left[s_{2,n}s_{2,n}^2\right]}=\frac{n}{2}{e\left[s_{2,n}^2\right]}-\frac{n(n-1)}{2(2n-3)}{e\left[s_{2,n-1}^2\right]}=\frac{n(n-1)(n^4 - 2n^3 - 15n^2 + 32n+8)}{8(2n-3)(2n-5)(2n-7)},\end{aligned}\ ] ] and so on . the first and second moments were individually calculated @xcite ( see eq . ) . note , however , that prop . [ prop1 ] provides a systematic calculation method of the @xmath79th moment of @xmath38 in a bottom - up way . furthermore , we easily obtain @xmath103}\sim(n/4)^k$ ] for @xmath105 using eq . . [ cor2 ] subtracting @xmath106}/4 $ ] from eq . , we have @xmath107}= \frac{n}{4}{e\left[p(s_{2,n } , n)\right]}-\frac{n(n-2)}{2(2n-3)}{e\left[p(s_{2,n-1 } , n)\right]}.\ ] ] [ lem3 ] for @xmath105 , @xmath108}\sim\frac{k!}{2^{\lfloor(k+1)/2\rfloor}\lfloor k/2\rfloor!4^k}n^{\lfloor k/2\rfloor}.\ ] ] that is to say , the asymptotic form of @xmath109}$ ] depends on whether @xmath79 is even or odd : @xmath110}\sim\frac{(2s-1)!!}{4^{2s}}n^s,\quad { e\left[\left(s_{2,n}-\frac{n}{4}\right)^{2s+1}\right]}\sim\frac{(2s+1)!!}{2\cdot4^{2s+1}}n^s , \label{eq : lem3}\ ] ] for @xmath73 . in @xmath111}$ ] ( @xmath112 ) , the leading @xmath113 terms are canceled because of @xmath114}\sim n/4 $ ] . similarly , in general @xmath79 , @xmath115 terms are successively canceled , so that the resultant leading order of @xmath109}$ ] becomes @xmath116 . this effect makes the estimation of @xmath109}$ ] difficult . the proof is by induction on @xmath79 . the statement is true for @xmath117 , and 2 , because @xmath118}&=1,\\{e\left[\left(s_{2,n}-\frac{n}{4}\right)^1\right]}&=\frac{n(n-1)}{2(2n-3)}-\frac{n}{4}=\frac{n}{4(2n-3)}\sim\frac{1}{8},\\{e\left[\left(s_{2,n}-\frac{n}{4}\right)^2\right]}&={e\left[s_{2,n}^2\right]}-\frac{n}{2}{e\left[s_{2,n}\right]}+\frac{n^2}{16}=\frac{n(4n^2 - 17n+16)}{16(2n-3)(2n-5)}\sim\frac{n}{16}.\ ] ] assume that it is true up to @xmath119 , and we show it is true for @xmath120 . it follows from cor . [ cor2 ] that @xmath121}&={e\left[\left(s_{2,n}-\frac{n}{4}\right)\left(s_{2,n}-\frac{n}{4}\right)^k\right]}\\ & = \frac{n}{4}{e\left[\left(s_{2,n}-\frac{n}{4}\right)^k\right]}-\frac{n(n-2)}{2(2n-3)}{e\left[\left(s_{2,n-1}-\frac{n}{4}\right)^k\right]}\\ & = \frac{n}{4}{e\left[\left(s_{2,n}-\frac{n}{4}\right)^k\right]}-\frac{n(n-2)}{2(2n-3)}{e\left[\left(s_{2,n-1}-\frac{n-1}{4}-\frac{1}{4}\right)^k\right]}\\ & = \frac{n}{4}{e\left[\left(s_{2,n}-\frac{n}{4}\right)^k\right]}-\frac{n(n-2)}{2(2n-3)}\sum_{l=0}^k\binom{k}{l}\left(-\frac{1}{4}\right)^l{e\left[\left(s_{2,n-1}-\frac{n-1}{4}\right)^{k - l}\right]}.\end{aligned}\ ] ] in the following , we separately investigate odd and even @xmath79 . case 1 : @xmath79 is odd ( @xmath81 ) . the dominant terms are @xmath122 and 1 in the summation , and the others can be neglected . thus , @xmath123}\\ & \sim\frac{n}{4}{e\left[\left(s_{2,n}-\frac{n}{4}\right)^{2s+1}\right ] } -\frac{n(n-2)}{2(2n-3)}\left({e\left[\left(s_{2,n-1}-\frac{n-1}{4}\right)^{2s+1}\right]}-\frac{2s+1}{4}{e\left[\left(s_{2,n-1}-\frac{n-1}{4}\right)^{2s}\right]}\right)\\ & \sim\frac{n}{4}\frac{(2s+1)!!}{4^{2s+1}}-\frac{n}{4}\left(\frac{(2s+1)!!}{4^{2s+1}}(n-1)^s-\frac{2s+1}{4}\frac{(2s-1)!!}{4^{2s}}(n-1)^s\right)\\ & \sim\frac{(2s+1)!!}{4^{2s+2}}n^{s+1}.\end{aligned}\ ] ] case 2 : @xmath79 is even ( @xmath80 ) . picking out the terms up to @xmath124 carefully , we obtain @xmath125}\nonumber\\ & = \frac{n}{4}{e\left[\left(s_{2,n}-\frac{n}{4}\right)^{2s}\right]}-\frac{n(n-2)}{2(2n-3)}\left({e\left[\left(s_{2,n-1}-\frac{n-1}{4}\right)^{2s}\right]}-\frac{2s}{4}{e\left[\left(s_{2,n-1}-\frac{n-1}{4}\right)^{2s-1}\right]}\right.\nonumber\\ & \qquad\left.+\frac{2s(2s-1)}{16\cdot2}{e\left[\left(s_{2,n-1}-\frac{n-1}{4}\right)^{2s-2}\right]}+o(n^{s-1})\right).\label{eq : lem3case2}\end{aligned}\ ] ] we introduce the coefficient @xmath126 as @xmath110}=\frac{(2s-1)!!}{4^{2s}}n^s + a_s n^{s-1 } + o(n^{s-1}),\ ] ] and expand each term on the right - hand side of eq . up to @xmath124 using the induction hypothesis : @xmath127 } & = \frac{(2s-1)!!}{4^{2s+1}}n^{s+1}+\frac{a_s}{4}n^s+o(n^{s}),\\ \frac{n(n-2)}{2(2n-3)}{e\left[\left(s_{2,n-1}-\frac{n-1}{4}\right)^{2s}\right ] } & = \frac{(2s-1)!!}{4^{2s+1}}n^{s+1}-\frac{(2s+1)!!}{2\cdot4^{2s+1}}n^s+\frac{a_s}{4}n^s+o(n^s),\\ \frac{n(n-2)}{2(2n-3)}\frac{2s}{4}{e\left[\left(s_{2,n-1}-\frac{n-1}{4}\right)^{2s-1}\right ] } & = \frac{(2s-1)!!}{4^{2s+1}}sn^s+o(n^s),\\ \frac{n(n-2)}{2(2n-3)}\frac{2s(2s-1)}{16\cdot2}{e\left[\left(s_{2,n-1}-\frac{n-1}{4}\right)^{2s-2}\right ] } & = \frac{(2s-1)!!}{4^{2s+1}}sn^s+o(n^s).\end{aligned}\ ] ] therefore , @xmath128}=\frac{(2s+1)!!}{2\cdot4^{2s+1}}n^s+o(n^s).\ ] ] note that @xmath129 terms and those including @xmath126 are all cancelled . therefore , the statement is true for any @xmath79 . [ cor3 ] multiplying eq . by @xmath130 , we have @xmath131}\sim\frac{(2s-1)!!}{4^{2s}}n^{-s},\quad { e\left[\left(\frac{s_{2,n}}{n}-\frac{1}{4}\right)^{2s+1}\right]}\sim\frac{(2s+1)!!}{2\cdot4^{2s+1}}n^{-s-1}.\ ] ] this result corresponds to the case @xmath62 in lemmas [ lem1 ] and [ lem2 ] . using this corollary , we can provide another proof of theorem [ thm1 ] as follows . the characteristic function @xmath132 for the lowest bifurcation ratio in theorem [ thm1 ] is calculated as @xmath133}\\ & = { e\left[\sum_{k=0}^\infty \frac{(iz\sqrt{n})^k}{k!}\left(\frac{s_{2,n}}{n}-\frac{1}{4}\right)^k\right]}\\ & = \sum_{k=0}^\infty \frac{(iz\sqrt{n})^k}{k!}{e\left[\left(\frac{s_{2,n}}{n}-\frac{1}{4}\right)^k\right]}\\ & = \sum_{s=0}^\infty \frac{(iz\sqrt{n})^{2s}}{(2s)!}{e\left[\left(\frac{s_{2,n}}{n}-\frac{1}{4}\right)^{2s}\right]}+\sum_{s=0}^\infty \frac{(iz\sqrt{n})^{2s+1}}{(2s+1)!}{e\left[\left(\frac{s_{2,n}}{n}-\frac{1}{4}\right)^{2s+1}\right]}\\ & \sim\sum_{s=0}^\infty \left(\frac{-z^2}{32}\right)^s\frac{1}{s!}\\ & = \exp\left(-\frac{z^2}{32}\right),\end{aligned}\ ] ] so the convergence of @xmath134 to @xmath135 is proved . we first derive the asymptotic form of @xmath136}$ ] , which is needed in the proof of lemma [ lem1 ] . [ prop2 ] for @xmath105 , we have @xmath137}\sim \left(\frac{n}{4}\right)^{-k}.\ ] ] since @xmath20 for any @xmath14 , @xmath138 surely takes a finite value and @xmath136}$ ] is not divergent . let us introduce the operator @xmath139 defined by @xmath140 where @xmath141 is integrable on any interval @xmath142 . note that @xmath139 is the inverse of @xmath143 . owing to the property @xmath144 the average of @xmath138 is expressed by @xmath145}&=\frac{n!(n-1)!(n-2)!}{(2n-2)!}\sum_{m=1}^{{\lfloorn/2\rfloor}}\frac{2^{n-2m}}{(n-2m)!m!(m-1)!}m^{-k}\\ & = \frac{n!(n-1)!(n-2)!}{(2n-2)!}\sum_{m=1}^{{\lfloorn/2\rfloor}}\frac{2^{n-2m}}{(n-2m)!m!(m-1)!}\left.\left(\frac{d}{dt}\right)^{-k}e^{mt}\right|_{t=0}\\ & = \frac{2^{n-2}n!(n-1)!}{(2n-2)!}\left.\left(\frac{d}{dt}\right)^{-k}e^t{f\left(\frac{2-n}{2 } , \frac{3-n}{2 } , 2 ; e^t\right)}\right|_{t=0}.\end{aligned}\ ] ] let us derive a relation between @xmath136}$ ] and @xmath146}$ ] as in prop . [ prop1 ] . @xmath147}&=\frac{2^{n-2}n!(n-1)!}{(2n-2)!}\left.\left(\frac{d}{dt}\right)^{-k}\left(\frac{d}{dt}\right)^{-1}e^t{f\left(\frac{2-n}{2 } , \frac{3-n}{2 } , 2 ; e^t\right)}\right|_{t=0}\\ & = \frac{2^{n-2}n!(n-1)!}{(2n-2)!}\left.\left(\frac{d}{dt}\right)^{-k}\int_{-\infty}^t e^s{f\left(\frac{2-n}{2 } , \frac{3-n}{2 } , 2 ; e^s\right)}ds\right|_{t=0}\\ & = \frac{2^{n-2}n!(n-1)!}{(2n-2)!}\left.\left(\frac{d}{dt}\right)^{-k}\left\{\left[e^s{f\left(\frac{2-n}{2 } , \frac{3-n}{2 } , 2 ; e^s\right)}\right]_{s=-\infty}^t -\int_{-\infty}^t e^s\frac{d}{ds}{f\left(\frac{2-n}{2 } , \frac{3-n}{2 } , 2 ; e^s\right)}ds\right\}\right|_{t=0}\\ & = \frac{2^{n-2}n!(n-1)!}{(2n-2)!}\left.\left(\frac{d}{dt}\right)^{-k}\left\{e^t{f\left(\frac{2-n}{2 } , \frac{3-n}{2 } , 2 ; e^t\right)}-\left(\frac{d}{dt}\right)^{-1 } e^t\frac{d}{dt}{f\left(\frac{2-n}{2 } , \frac{3-n}{2 } , 2 ; e^t\right)}\right\}\right|_{t=0}\\ & = { e\left[s_{2,n}^{-k}\right]}-\frac{2^{n-2}n!(n-1)!}{(2n-2)!}\left.\left(\frac{d}{dt}\right)^{-(k+1)}e^t\frac{d}{dt}{f\left(\frac{2-n}{2 } , \frac{3-n}{2 } , 2 ; e^t\right)}\right|_{t=0}.\end{aligned}\ ] ] by using the derivative of the hypergeometric function in eq . , @xmath148}={e\left[s_{2,n}^{-k}\right]}-\frac{n-2}{2}{e\left[s_{2,n}^{-(k+1)}\right]}+\frac{n(n-2)}{2(2n-3)}{e\left[s_{2,n-1}^{-(k+1)}\right]},\ ] ] whose asymptotic form is @xmath149}={e\left[s_{2,n}^{-k}\right]}+\frac{n(n-2)}{2(2n-3)}{e\left[s_{2,n-1}^{-(k+1)}\right ] } \sim{e\left[s_{2,n}^{-k}\right]}+\frac{n}{4}{e\left[s_{2,n}^{-(k+1)}\right]},\ ] ] or @xmath148}\sim\frac{4}{n}{e\left[s_{2,n}^{-k}\right]}.\ ] ] considering @xmath150}=1 $ ] , we obtain @xmath137}\sim\left(\frac{n}{4}\right)^{-k}.\ ] ] by induction on @xmath12 . for @xmath62 , the statement is equivalent to cor . [ cor3 ] . assume that it is true up to @xmath31 , and we show it is true for @xmath12 . using eq . , @xmath151 } & = \frac{n!(n-1)!(n-2)!}{(2n-2)!}\sum_{m=1}^{{\lfloorn/2\rfloor}}\frac{2^{n-2m}}{(n-2m)!m!(m-1)!}{e\left[\left(\frac{s_{r , m}}{s_{r-1,m}}-\frac{1}{4}\right)^k\right]}.\end{aligned}\ ] ] case 1 : @xmath79 is even ( @xmath80 ) . @xmath152 } & = \frac{n!(n-1)!(n-2)!}{(2n-2)!}\sum_{m=1}^{{\lfloorn/2\rfloor}}\frac{2^{n-2m}}{(n-2m)!m!(m-1)!}{e\left[\left(\frac{s_{r , m}}{s_{r-1,m}}-\frac{1}{4}\right)^{2s}\right]}\\ & \sim\frac{n!(n-1)!(n-2)!}{(2n-2)!}\sum_{m=1}^{{\lfloorn/2\rfloor}}\frac{2^{n-2m}}{(n-2m)!m!(m-1)!}\frac{(2s-1)!!}{4^{2s}}\left(\frac{m}{4^{r-2}}\right)^{-s}\\ & = \frac{(2s-1)!!}{4^{2s}4^{-s(r-2)}}\frac{n!(n-1)!(n-2)!}{(2n-2)!}\sum_{m=1}^{{\lfloorn/2\rfloor}}\frac{2^{n-2m}}{(n-2m)!m!(m-1)!}m^{-s}.\end{aligned}\ ] ] by using eq . again and prop . [ prop2 ] , @xmath74 } \sim\frac{(2s-1)!!}{4^{2s}4^{-s(r-2)}}{e\left[s_{2,n}^{-s}\right ] } \sim\frac{(2s-1)!!}{4^{2s}4^{-s(r-2)}}\left(\frac{n}{4}\right)^{-s } = \frac{(2s-1)!!}{4^{2s}}\left(\frac{n}{4^{r-1}}\right)^{-s}.\ ] ] case 2 : @xmath79 is odd ( @xmath81 ) . using eq . and prop . [ prop2 ] as above , @xmath153 } & = \frac{n!(n-1)!(n-2)!}{(2n-2)!}\sum_{m=1}^{{\lfloorn/2\rfloor}}\frac{2^{n-2m}}{(n-2m)!m!(m-1)!}{e\left[\left(\frac{s_{r , m}}{s_{r-1,m}}-\frac{1}{4}\right)^{2s+1}\right]}\\ & \sim\frac{n!(n-1)!(n-2)!}{(2n-2)!}\sum_{m=1}^{{\lfloorn/2\rfloor}}\frac{2^{n-2m}}{(n-2m)!m!(m-1)!}\frac{(2s+1)!!}{2\cdot4^{2s+1}}\left(\frac{m}{4^{r-2}}\right)^{-s-1}\\ & \sim\frac{(2s+1)!!}{2\cdot4^{2s+1}4^{(r-1)(-s-1)}}{e\left[s_{2,n}^{-s-1}\right]}\\ & = \frac{(2s+1)!!}{2\cdot4^{2s+1}}\left(\frac{n}{4^r}\right)^{-s-1}.\end{aligned}\ ] ] in the proof of lemma [ lem2 ] , we use the following relations . [ lem4 ] for @xmath154 , @xmath155}\sim\left(\frac{n}{4}\right)^l\frac{(2s-1)!!}{4^{2s}}n^s , \quad { e\left[s_{2,n}^l\left(s_{2,n}-\frac{n}{4}\right)^{2s+1}\right]}\sim\left(\frac{n}{4}\right)^l\frac{(2s+1)!!}{2\cdot4^{2s+1}}(2l+1)n^s.\ ] ] the complicated form of the odd - power result in lemma [ lem2 ] is actually due to the factor `` @xmath156 '' . by induction on @xmath157 . for @xmath122 , the statement is equivalent to lemma [ lem3 ] . assume that it is true up to @xmath158 , and we show for @xmath159 . using prop . [ prop1 ] , @xmath160 } & = \frac{n}{2}{e\left[s_{2,n}^l\left(s_{2,n}-\frac{n}{4}\right)^k\right]}-\frac{n(n-2)}{2(2n-3)}{e\left[s_{2,n-1}^l\left(s_{2,n-1}-\frac{n}{4}\right)^k\right]}\\ & = \frac{n}{2}{e\left[s_{2,n}^l\left(s_{2,n}-\frac{n}{4}\right)^k\right]}-\frac{n(n-2)}{2(2n-3)}{e\left[s_{2,n-1}^l\left(s_{2,n-1}-\frac{n-1}{4}-\frac{1}{4}\right)^k\right]}\\ & = \frac{n}{2}{e\left[s_{2,n}^l\left(s_{2,n}-\frac{n}{4}\right)^k\right]}-\frac{n(n-2)}{2(2n-3)}\sum_{p=0}^k\binom{k}{p}\left(-\frac{1}{4}\right)^p{e\left[s_{2,n-1}^l\left(s_{2,n-1}-\frac{n-1}{4}\right)^{k - p}\right]}.\end{aligned}\ ] ] case 1 : @xmath79 is even ( @xmath80 ) . in the summation , only @xmath161 is dominant , so that @xmath162 } \sim\frac{n}{2}{e\left[s_{2,n}^l\left(s_{2,n}-\frac{n}{4}\right)^{2s}\right]}-\frac{n(n-2)}{2(2n-3)}{e\left[s_{2,n-1}^l\left(s_{2,n-1}-\frac{n-1}{4}\right)^{2s}\right]}.\ ] ] by using the induction hypothesis , @xmath162 } \sim\frac{n}{2}\left(\frac{n}{4}\right)^l\frac{(2s-1)!!}{4^{2s}}n^s-\frac{n}{4}\left(\frac{n-1}{4}\right)^l\frac{(2s-1)!!}{4^{2s}}(n-1)^s \sim\left(\frac{n}{4}\right)^{l+1}\frac{(2s-1)!!}{4^{2s}}n^s.\ ] ] case 2 : @xmath79 is odd ( @xmath81 ) . note that @xmath161 and 1 in the summation are the dominant terms . @xmath163}\\ & \sim\frac{n}{2}{e\left[s_{2,n}^l\left(s_{2,n}-\frac{n}{4}\right)^{2s+1}\right]}\\ & \quad-\frac{n(n-2)}{2(2n-3)}\left({e\left[s_{2,n-1}^l\left(s_{2,n-1}-\frac{n-1}{4}\right)^{2s+1}\right]}-\frac{(2s+1)}{4}{e\left[s_{2,n-1}^l\left(s_{2,n-1}-\frac{n-1}{4}\right)^{2s}\right]}\right)\\ & \sim\frac{n}{2}\left(\frac{n}{4}\right)^l\frac{(2s+1)!!}{2\cdot4^{2s+1}}(2l+1)n^s\\ & \quad-\frac{n(n-2)}{2(2n-3)}\left[\left(\frac{n-1}{4}\right)^l\frac{(2s+1)!!}{2\cdot4^{2s+1}}(2l+1)(n-1)^s-\frac{(2s+1)}{4}\left(\frac{n-1}{4}\right)^l\frac{(2s-1)!!}{4^{2s}}(n-1)^s\right]\\ & \sim\left(\frac{n}{4}\right)^{l+1}\frac{(2s+1)!!}{2\cdot4^{2s+1}}(2l+3)n^s.\end{aligned}\ ] ] by induction on @xmath12 . for @xmath62 , the statement is equivalent to lemma [ lem3 ] . assume that it is true up to @xmath31 , and we show it is true for @xmath12 . using eq . to calculate @xmath164 } & = \frac{n!(n-1)!(n-2)!}{(2n-2)!}\sum_{m=1}^{{\lfloorn/2\rfloor}}\frac{2^{n-2m}}{(n-2m)!m!(m-1)!}{e\left[\left(s_{r , m}-\frac{n}{4^r}\right)^k\right]}\\ & = \frac{n!(n-1)!(n-2)!}{(2n-2)!}\sum_{m=1}^{{\lfloorn/2\rfloor}}\frac{2^{n-2m}}{(n-2m)!m!(m-1)!}{e\left[\left(s_{r , m}-\frac{m}{4^{r-1}}+\frac{m}{4^{r-1}}-\frac{n}{4^r}\right)^k\right]}\\ & = \frac{n!(n-1)!(n-2)!}{(2n-2)!}\sum_{m=1}^{{\lfloorn/2\rfloor}}\frac{2^{n-2m}}{(n-2m)!m!(m-1 ) ! } \sum_{l=0}^{k+1}\binom{k}{l}\left(\frac{m}{4^{r-1}}-\frac{n}{4^r}\right)^{k - l}{e\left[\left(s_{r.m}-\frac{m}{4^{r-1}}\right)^l\right]}.\end{aligned}\ ] ] we split the summation over @xmath157 according to the parity of @xmath157 , and use the induction hypothesis . the estimation of the summation is complex compared with the other proofs above . case 1 : @xmath79 is even ( @xmath80 ) . @xmath75 } & = \frac{n!(n-1)!(n-2)!}{(2n-2)!}\sum_{m=1}^{{\lfloorn/2\rfloor}}\frac{2^{n-2m}}{(n-2m)!m!(m-1)!}\\ & \quad\times \left[\sum_{l=0}^{s}\binom{2s}{2l}\left(\frac{m}{4^{r-1}}-\frac{n}{4^r}\right)^{2s-2l}{e\left[\left(s_{r.m}-\frac{m}{4^{r-1}}\right)^{2l}\right]}\right.\\ & \qquad \left.+\sum_{l=0}^{s-1}\binom{2s}{2l+1}\left(\frac{m}{4^{r-1}}-\frac{n}{4^r}\right)^{2s-2l-1}{e\left[\left(s_{r , m}-\frac{m}{4^{r-1}}\right)^{2l+1}\right]}\right]\\ & \sim\frac{n!(n-1)!(n-2)!}{(2n-2)!}\sum_{m=1}^{{\lfloorn/2\rfloor}}\frac{2^{n-2m}}{(n-2m)!m!(m-1)!}\\ & \quad\times \left[\sum_{l=0}^{s}\binom{2s}{2l}\left(\frac{1}{4^{r-1}}\right)^{2s-2l}\left(m-\frac{n}{4}\right)^{2s-2l}\frac{(2l-1)!!}{4^{2l(r-1)}}\left(\frac{4^{r-1}-1}{3}\right)^l m^l\right.\\ & \qquad \left.+\sum_{l=0}^{s-1}\binom{2s}{2l+1}\left(\frac{1}{4^{r-1}}\right)^{2s-2l-1}\left(m-\frac{n}{4}\right)^{2s-2l-1}\frac{(2l+1)!!}{2\cdot4^{(2l+1)(r-1)}}\left(\frac{4^{r-1}-1}{3}\right)^{l+1}m^l\right]\\ & = \sum_{l=0}^s\binom{2s}{2l}\left(\frac{1}{4^{r-1}}\right)^{2s-2l}\frac{(2l-1)!!}{4^{2l(r-1)}}\left(\frac{4^{r-1}-1}{3}\right)^l{e\left[\left(s_{2,n}-\frac{n}{4}\right)^{2s-2l}s_{2,n}^l\right]}\\ & \quad+\sum_{l=0}^{s-1}\binom{2s}{2l+1}\left(\frac{1}{4^{r-1}}\right)^{2s-2l-1}\frac{(2l+1)!!}{2\cdot4^{(2l+1)(r-1)}}\left(\frac{4^{r-1}-1}{3}\right)^{l+1}{e\left[\left(s_{2,n}-\frac{n}{4}\right)^{2s-2l-1}s_{2,n}^l\right]}. \displaybreak[1]\end{aligned}\ ] ] from lemma [ lem4 ] , the first summation is @xmath124 , while the second summation is @xmath165 . thus , we can neglect the second sum , so that @xmath75 } & \sim \sum_{l=0}^s\binom{2s}{2l}\left(\frac{1}{4^{r-1}}\right)^{2s-2l}\frac{(2l-1)!!}{4^{2l(r-1)}}\left(\frac{4^{r-1}-1}{3}\right)^l \frac{1}{4^l}\frac{(2s-2l-1)!!}{4^{2s-2l}}n^s\\ & = \frac{(2s)!n^s}{s!2^s 4^{2sr}}\sum_{l=0}^s \binom{s}{l } \left(\frac{4^r-4}{3}\right)^l\\ & = \frac{(2s-1)!!n^s}{4^{2sr}}\left(1+\frac{4^r-4}{3}\right)^s\\ & = \frac{(2s-1)!!n^s}{4^{2sr}}\left(\frac{4^r-1}{3}\right)^s.\end{aligned}\ ] ] case 2 : @xmath79 is odd ( @xmath81 ) . @xmath166 } & = \frac{n!(n-1)!(n-2)!}{(2n-2)!}\sum_{m=1}^{{\lfloorn/2\rfloor}}\frac{2^{n-2m}}{(n-2m)!m!(m-1)!}\\ & \quad\times \left[\sum_{l=0}^{s}\binom{2s+1}{2l}\left(\frac{m}{4^{r-1}}-\frac{n}{4^r}\right)^{2s+1 - 2l}{e\left[\left(s_{r.m}-\frac{m}{4^{r-1}}\right)^{2l}\right]}\right.\\ & \qquad \left.+\sum_{l=0}^{s}\binom{2s+1}{2l+1}\left(\frac{m}{4^{r-1}}-\frac{n}{4^r}\right)^{2s-2l}{e\left[\left(s_{r , m}-\frac{m}{4^{r-1}}\right)^{2l+1}\right]}\right]\displaybreak[1]\\ & \sim\frac{n!(n-1)!(n-2)!}{(2n-2)!}\sum_{m=1}^{{\lfloorn/2\rfloor}}\frac{2^{n-2m}}{(n-2m)!m!(m-1)!}\\ & \quad\times \left[\sum_{l=0}^{s}\binom{2s+1}{2l}\left(\frac{1}{4^{r-1}}\right)^{2s+1 - 2l}\left(m-\frac{n}{4}\right)^{2s+1 - 2l}\frac{(2l-1)!!}{4^{2l(r-1)}}\left(\frac{4^{r-1}-1}{3}\right)^l m^l\right.\\ & \qquad \left.+\sum_{l=0}^{s}\binom{2s+1}{2l+1}\left(\frac{1}{4^{r-1}}\right)^{2s-2l}\left(m-\frac{n}{4}\right)^{2s-2l}\frac{(2l+1)!!}{2\cdot4^{(2l+1)(r-1)}}\left(\frac{4^{r-1}-1}{3}\right)^l\right.\\ & \qquad\left.\times\frac{1}{5}\left(\frac{4^r-1}{3}+\frac{4^{r-2}-1}{3}4(2l+1)\right)m^l\right]\displaybreak[1]\\ & = \sum_{l=0}^s\binom{2s+1}{2l}\left(\frac{1}{4^{r-1}}\right)^{2s+1 - 2l}\frac{(2l-1)!!}{4^{2l(r-1)}}\left(\frac{4^{r-1}-1}{3}\right)^l{e\left[\left(s_{2,n}-\frac{n}{4}\right)^{2s+1 - 2l}s_{2,n}^l\right]}\\ & \quad+\sum_{l=0}^s\binom{2s+1}{2l+1}\left(\frac{1}{4^{r-1}}\right)^{2s-2l}\frac{(2l+1)!!}{2\cdot4^{(2l+1)(r-1)}}\left(\frac{4^{r-1}-1}{3}\right)^l\\ & \qquad\times\frac{1}{5}\left(\frac{4^r-1}{3}+\frac{4^{r-2}-1}{3}4(2l+1)\right){e\left[\left(s_{2,n}-\frac{n}{4}\right)^{2s-2l}s_{2,n}^l\right]}.\end{aligned}\ ] ] both two sums are @xmath167 , so we need to consider them . by using lemma [ lem4 ] , @xmath166 } & \sim\sum_{l=0}^s\binom{2s+1}{2l}\left(\frac{1}{4^{r-1}}\right)^{2s+1 - 2l}\frac{(2l-1)!!}{4^{2l(r-1)}}\left(\frac{4^{r-1}-1}{3}\right)^l\left(\frac{n}{4}\right)^l\frac{(2s-2l+1)!!}{2\cdot4^{2s-2l+1}}(2l+1)n^{s - l}\\ & \quad+\sum_{l=0}^s\binom{2s+1}{2l+1}\left(\frac{1}{4^{r-1}}\right)^{2s-2l}\frac{(2l+1)!!}{2\cdot4^{(2l+1)(r-1)}}\left(\frac{4^{r-1}-1}{3}\right)^l\\ & \qquad\times\frac{1}{5}\left(\frac{4^r-1}{3}+\frac{4^{r-2}-1}{3}4(2l+1)\right)\left(\frac{n}{4}\right)^l\frac{(2s-2l-1)!!}{4^{2s-2l}}n^{s - l}\\ & = \frac{(2s+1)!!n^s}{2\cdot4^{r(2s+1)}}\left[\sum_{l=0}^s\binom{s}{l}\left(\frac{4^r-4}{3}\right)^l(2l+1)+\frac{4}{5}\sum_{l=0}^s\binom{s}{l}\left(\frac{4^r-4}{3}\right)^l\left(\frac{4^r-1}{3}+\frac{4^{r-2}-1}{3}4(2l+1)\right)\right]\\ & = \frac{(2s+1)!!n^s}{2\cdot4^{r(2s+1)}}\frac{4^r-1}{15}\sum_{l=0}^s\binom{s}{l}\left(\frac{4^r-4}{3}\right)^l(2l+5)\\ & = \frac{(2s+1)!!n^s}{2\cdot4^{r(2s+1)}}\left(\frac{4^r-1}{3}\right)^s\frac{1}{5}\left(\frac{4^{r+1}-1}{3}+\frac{4^{r-1}-1}{3}4(2s+1)\right),\end{aligned}\ ] ] where we have used @xmath168 theorems [ thm2 ] and [ thm3 ] are further generalized to the central limit theorem as follows . [ lem5 ] for @xmath169 and @xmath73 , @xmath173}\sim 4^{s(q-1)}\frac{(2s-1)!!}{4^{2sr}}\left(\frac{4^r-1}{3}\right)^sn^{-s},\\ & { e\left[\left(\frac{s_{q+r , n}}{s_{q , n}}-\frac{1}{4^r}\right)^{2s+1}\right]}\sim 4^{s(q-1)}\frac{(2s+1)!!}{4^{(2s+1)r}}\left(\frac{4^r-1}{3}\right)^s\frac{1}{5}\left(\frac{4^{r+1}-1}{3}+\frac{4^{r-1}-1}{3}4(2s+1)\right ) n^{-s}.\end{aligned}\ ] ] assume that the statement is true for @xmath178 , and we show that it is true for @xmath179 . using eq . and prop . [ prop2 ] , @xmath180 } & = \frac{n!(n-1)!(n-2)!}{(2n-2)!}\sum_{m=1}^{{\lfloorn/2\rfloor}}\frac{2^{n-2m}}{(n-2m)!m!(m-1)!}{e\left[\left(\frac{s_{q+r , n}}{s_{q , n}}-\frac{1}{4^r}\right)^{2s}\right]}\\ & \sim\frac{n!(n-1)!(n-2)!}{(2n-2)!}\sum_{m=1}^{{\lfloorn/2\rfloor}}\frac{2^{n-2m}}{(n-2m)!m!(m-1)!}4^{s(q-1)}\frac{(2s-1)!!}{4^{2sr}}\left(\frac{4^r-1}{3}\right)^s m^{-s}\\ & = 4^{s(q-1)}\frac{(2s-1)!!}{4^{2sr}}\left(\frac{4^r-1}{3}\right)^s{e\left[s_{2,n}^{-s}\right]}\\ & \sim 4^{sq}\frac{(2s-1)!!}{4^{2sr}}\left(\frac{4^r-1}{3}\right)^s n^{-s}.\end{aligned}\ ] ] similarly , @xmath181 } \sim 4^{sq}\frac{(2s+1)!!}{4^{(2s+1)r}}\left(\frac{4^r-1}{3}\right)^s\frac{1}{5}\left(\frac{4^{r+1}-1}{3}+\frac{4^{r-1}-1}{3}4(2s+1)\right ) n^{-s}.\end{aligned}\ ] ] 10 p. ball , branches ( oxford university press , oxford , 2011 ) . d. e. knuth , the art of computer programming , vol . 3 ( addison wesley , reading , 1973 ) . j. d. archibald , aristotle s ladder , darwin s tree : the evolution of visual metaphors for biological order ( columbia university press , new york , 2014 ) . r. e. horton , geol . am . bull . 56 , 275 ( 1945 ) . a. n. strahler , trans . un . 38 , 913 ( 1957 ) . r. l. shreve , j. geol . 75 , 178 ( 1967 ) . e. tokunaga , geogr . tokyo metrop . 13 , 1 ( 1987 ) . r. p. stanley , enumerative combinatorics , vol . 2 ( cambridge university press , cambridge , 1999 ) . k. yamamoto , j. stat . 139 , 62 ( 2010 ) . c. werner , canadian geographer 16 , 50 ( 1972 ) . k. yamamoto , phys . e 78 , 021114 ( 2008 ) . s. x. wang and e. c. waymire , siam j. discr . math . 4 , 575 ( 1991 ) . w. feller , an introduction to probability theory and its applications , vol . 2 ( wiley , new york , 1968 ) . m. abramowitz and i. a. stegun , handbook of mathematical functions ( dover , new york , 1972 ) .

we introduce a new class of noncommutative spectral triples on kellendonk s-algebra associated with a nonperiodic substitution tiling . these spectral triples are constructed from fractal trees on tilings , which define a geodesic distance between any two tiles in the tiling . since fractals typically have infinite euclidean length , the geodesic distance is defined using perron - frobenius theory , and is self - similar with scaling factor given by the perron - frobenius eigenvalue . we show that each spectral triple is-summable , and respects the hierarchy of the substitution system . to elucidate our results , we construct a fractal tree on the penrose tiling , and explicitly show how it gives rise to a collection of spectral triples .

1504.05296

a tiling of the plane is a covering of @xmath2 by a collection of compact subsets , called _ tiles _ , for which two distinct tiles can only meet along their boundaries . the building blocks of a tiling are the _ prototiles _ : a finite set of tiles with the property that every other tile is a translation of some prototile . a tiling is said to be _ nonperiodic _ if it lacks any translational periodicity . one method of producing tilings is via a substitution rule ; a rule that expands each tile , and breaks it into smaller pieces , each of which is an isometric copy of an original tile . a nonperiodic substitution rule gives rise to a dynamical system , called the _ continuous hull _ , that consists of all tilings whose local patterns appear in some finite substitution of a prototile . the continuous hull becomes a dynamical system where the homeomorphism is induced by translation . in order to associate a particularly tractable @xmath0-algebra to a nonperiodic tiling , kellendonk @xcite places punctures in each tile , which he then uses to define a discrete subset of the continuous hull , which we refer to as the _ discrete hull_. in this paper , we define spectral triples on kellendonk s @xmath0-algebra @xmath3 associated to a tiling . the fundamental new ingredients for these spectral triples , are the recently developed fractal dual substitution tilings @xcite . suppose @xmath4 is a nonperiodic substitution tiling with finite prototile set @xmath5 . for each prototile @xmath6 , a fractal dual tiling defines a fractal tree , in fact infinitely many , on a self - similar tiling @xmath7 ; a tiling constructed from the substitution rule on @xmath8 . each of our fractal trees defines a unique fractal path between the punctures of any two tiles in @xmath7 . moreover , each fractal tree on @xmath9 respects the hierarchy of the substitution rule . given a fractal tree on @xmath9 , we apply perron - frobenius theory to the substitution matrix associated to the edges of the fractal dual tiling , to define a length function on each fractal edge in the fractal tree . this extends to a self - similar length function on the entire fractal tree , with scaling factor given by the perron - frobenius eigenvalue @xmath10 . if @xmath11 is the scaling factor for the original tiling , the scaling factor @xmath10 of the fractal tree , is related to the hausdorff dimension @xmath12 of the fractal dual tiling by the formula @xmath13 . the fractal tree is then used to define a length function between any two tiles of @xmath7 using perron - frobenius theory . let @xmath14 denote the fractal length between the punctures of two tiles @xmath15 and @xmath16 in @xmath7 . to each substitution tiling with a fractal dual tiling , we construct spectral triples on kellendonk s @xmath0-algebra @xmath3 , which we now outline . for each @xmath6 , let @xmath17 , with canonical basis @xmath18 , and define an unbounded multiplication operator @xmath19 . we show that @xmath20 is a @xmath1-summable ( positive ) spectral triple . let @xmath21 . for each function @xmath22 , we define an unbounded multiplication operator @xmath23 . then , @xmath24 is also a @xmath1-summable spectral triple . this defines a collection of spectral triples on kellendonk s algebra @xmath3 that each respect the hierarchy of the substitution rule . using operator algebras as the basic framework , alain connes developed noncommutative geometry @xcite , and has shown its significance to many fields of mathematics . in particular , one of the overarching themes of noncommutative geometry is to describe a consistent mathematical model for quantum physics . dynamical systems are particularly well suited to the tools of noncommutative geometry , and provide dynamical invariants in a noncommutative framework . of particular importance to connes program are spectral triples , which typically define a noncommutative riemannian metric on a @xmath0-algebra . a spectral triple @xmath25 consists of a @xmath0-algebra @xmath26 faithfully represented on a separable hilbert space @xmath27 , and a self - adjoint unbounded operator @xmath28 on @xmath27 with compact resolvent , whose commutators with a dense @xmath29-subalgebra of @xmath26 are bounded . the noncommutative topology of tilings has a long history . alain connes initiated the study of substitution tilings in a noncommutative framework by giving a detailed description of a @xmath0-algebra associated with the penrose tiling in his seminal book @xcite . in 1982 , dan shechtman discovered quasicrystals @xcite , a type of material that is neither crystalline nor amorphous . the mathematical theory explaining shechtman s discovery had already been developed in the context of purely mathematical research ; nonperiodic tilings provide an excellent model for quasicrystals . in an attempt to understand the physics of quasicrystals , bellissard defined a crossed product @xmath0-algebra by a family of schrdinger operators . years later , kellendonk defined a discrete version of the continuous hull and constructed a groupoid @xmath0-algebra associated with a tiling @xcite . soon afterwards , anderson and putnam @xcite showed that the continuous hull @xmath30 of a tiling is a smale space , and used this observation to describe the @xmath31-theory of the crossed product @xmath32 . more recently , kellendonk s construction was generalised to tilings with infinite rotational symmetry in @xcite , and the rotationally equivariant @xmath31-theory of these algebras was completely worked out in @xcite . only recently has there been a breakthrough in the noncommutative geometry of tilings . the primary interest in spectral triples on tilings is that the continuous hull of a nonperiodic tiling is not only a topological object , it also has rich geometric structure . the groundbreaking spectral triple for tilings appeared in john pearson s 2008 thesis @xcite , and the subsequent joint paper with bellissard @xcite . these spectral triples were defined on the commutative @xmath0-algebra associated with the hull of a tiling . a few years later , the second author constructed spectral triples on the unstable @xmath0-algebra of a smale space @xcite , which is strongly morita equivalent to kellendonk s algebra . however , in the special case of tiling algebras , this spectral triple essentially measured the euclidean distance between two tilings in the groupoid used to define the @xmath0-algebra , and ignored the substitution system . since bellissard and pearson s seminal result there have been a number of papers on spectral triples of tilings , see for example @xcite . the survey article @xcite explains these constructions and their relationship to one another . we thank the anonymous reviewer for suggesting revisions that greatly improved the exposition . the tilings in this paper are built from _ prototiles _ , a finite collection @xmath5 of labelled compact subsets of @xmath2 that each contain the origin , and are equal to the closure of their interior . we denote the label of a prototile @xmath6 by @xmath33 , the support of @xmath8 by @xmath34 , and the boundary of @xmath35 by @xmath36 . in general , given a subset @xmath37 , we write @xmath38 for the boundary of @xmath39 . the labels allow us to have two distinct prototiles with the same support , and we often denote the labels by colours . a _ tile _ is defined to be any translation of a prototile . so , for any @xmath6 and @xmath40 , the labelled subset @xmath41 is a tile with label @xmath42 , and support @xmath43 . [ defn : tiling ] let @xmath5 be a set of prototiles . tiling of the plane _ is a countable collection @xmath44 of tiles , each of which is a translate of some prototile @xmath45 , such that 1 . @xmath46 ; and 2 . @xmath47 whenever @xmath48 . a tiling @xmath4 is said to be _ edge - to - edge _ if whenever two tiles intersect , they meet full edge to full edge , or at a common vertex ( see ( * ? ? ? * section 3.2 ) for the definition of an edge in the case that the tiling does not have polygonal prototiles ) . if tiles in an edge - to - edge tiling @xmath4 only intersect along at most one edge , then @xmath4 is said to be _ singly edge - to - edge_. a _ patch _ @xmath49 is a finite collection of tiles in @xmath4 such that the interior of the support of @xmath50 is connected . for @xmath40 and @xmath51 , let @xmath52 denote the ball of radius @xmath53 centred at @xmath54 . given a tiling @xmath4 , we require the following collections of tiles . for @xmath40 and @xmath55 , let @xmath56 then , @xmath57 is the collection of tiles @xmath58 whose support is completely contained in @xmath52 . suppose @xmath4 is a tiling , and @xmath40 . the _ translate _ of @xmath4 by @xmath54 is @xmath59 . the _ orbit _ of @xmath4 is the set @xmath60 . if there exists @xmath61 such that @xmath62 , we say that @xmath4 is _ periodic_. if @xmath63 implies that @xmath64 , then we say that @xmath4 is _ nonperiodic_. [ tiling metric ] let @xmath4 be a tiling . for @xmath65 the _ tiling metric _ is given by @xmath66 two tilings @xmath67 are close in the tiling metric if @xmath4 and @xmath68 have the same pattern of tiles on a large ball centred about the origin , up to a small translation . the _ continuous hull _ of a tiling @xmath44 , denoted by @xmath69 , is the completion of @xmath70 in the tiling metric . if every tiling in @xmath30 is nonperiodic , then we say that @xmath4 is _ strongly nonperiodic_. it is straightforward to check that the limit of a cauchy sequence of tilings in @xmath71 is a tiling with the same prototile set as @xmath4 , see ( * ? ? * section 2.1 ) for details . in this paper , the tilings we investigate are those whose continuous hull is a compact topological space . in @xcite , radin and wolff give a characterisation of such tilings . a tiling @xmath4 has _ finite local complexity _ if there are only a finite number of two tile patches in @xmath4 , up to translation . in this paper , we are concerned with tilings built from finite prototile sets , whose tiles meet singly edge - to - edge . thus , the tilings we consider here have finite local complexity . [ thm : compact ] a tiling @xmath4 has finite local complexity if and only if @xmath30 is compact . let @xmath5 be a set of prototiles . a _ substitution rule _ on @xmath5 is a map @xmath72 on @xmath73 , with a scaling factor @xmath74 , such that for each @xmath6 , @xmath75 is a patch satisfying the following properties : 1 . @xmath76 ; and 2 . @xmath77 for all @xmath78 with @xmath48 . suppose @xmath72 is a substitution rule on a set of prototiles @xmath5 with scaling factor @xmath79 . let @xmath6 , @xmath40 , and consider the tile @xmath80 . we extend @xmath72 to tiles by @xmath81 , and hence , to tilings by @xmath82 . a similar formula holds for a patch @xmath50 . in particular , for each @xmath6 , we may apply @xmath72 to the patch @xmath75 . we denote @xmath83 by @xmath84 , and inductively define @xmath85 for each @xmath86 . a _ supertile of order @xmath87 _ is a translate of the patch @xmath88 for some @xmath6 . [ eg : penrose ] the version of the penrose tiling we consider , consists of the four prototiles depicted in figure [ fig : penrose ] , and all rotations by multiples of @xmath89 . this gives a total of forty prototiles . the substitution rule @xmath72 applied to each of the prototiles is also illustrated in figure [ fig : penrose ] , and extends to all forty tiles by rotation . the scaling factor for this substitution is the golden ratio , @xmath90 . a patch of the penrose tiling appears in figure [ fig : penrose_patch ] . [ notation : substitution ] let @xmath91 be a substitution rule on a set of prototiles @xmath73 . for each @xmath6 and @xmath92 , let @xmath93 denote the number of tiles in the patch @xmath88 . we write @xmath94 . we note that @xmath95 is well defined , since prototile sets are finite . moreover , for each @xmath6 and @xmath92 , it is clear that @xmath96 . [ defn : submatrix ] let @xmath72 be a substitution rule on a set of prototiles @xmath5 . for each pair of prototiles @xmath97 , define the following set : @xmath98 the matrix @xmath99 is called the _ digit matrix for @xmath72_. for each pair of prototiles @xmath97 , let @xmath100 , so that @xmath101 is the number of translates of @xmath102 in the patch @xmath75 . the matrix @xmath103 is called the _ substitution matrix for @xmath72_. let @xmath72 be a substitution rule on a set of prototiles @xmath5 , and let @xmath104 be the substitution matrix for @xmath72 . if @xmath105 is irreducible , in the sense that there is an @xmath106 sufficiently large such that @xmath107 for all pairs @xmath108 , we say that @xmath91 is a _ primitive substitution rule_. following sadun ( * ? ? * theorem 1.4 ) , we obtain a tiling from a primitive substitution rule @xmath72 on @xmath5 in the following way . fix a prototile @xmath6 . since @xmath72 is primitive , there exists @xmath92 sufficiently large such that there exists a translate of @xmath8 , say @xmath15 , in the patch @xmath88 whose support is contained in the interior of the support of @xmath109 . thus , there is a fixed point in @xmath15 with respect to the map from @xmath110 to @xmath111 . if we place the origin of @xmath2 at this fixed point , we obtain the sequence @xmath112 the union @xmath113 is a tiling of @xmath2 . we may modify the prototile set and the substitution rule so that @xmath114 for each @xmath6 . this implies that the fixed points under the substitution rule are at the origin . we make this assumption for the remainder of this paper . [ tilinggeneratedbyp ] let @xmath72 be a primitive substitution rule on a set of prototiles @xmath5 . for each @xmath6 , the _ tiling generated by @xmath8 _ is the tiling @xmath115 . for each @xmath6 , the tiling generated by @xmath8 is a self - similar tiling in the sense of ( * ? ? ? * definition 2.5 ) , since @xmath116 . in the following sections , we will often require the map @xmath117 , which applies the substitution rule to prototiles , and scales the resulting patch by @xmath118 . for each @xmath97 , and each non - empty subset @xmath119 , we define the following notation : @xmath120 note that if we identify each prototile with its support , then for each @xmath45 , we have @xmath121 from the digit matrix , we build a map @xmath122 . consider the set of non - empty compact subsets of @xmath2 , denoted by @xmath123 . the domain of @xmath122 is given by @xmath124 , and for @xmath125 , we write @xmath126 for the @xmath127 coordinate in the disjoint union . then , for any @xmath125 and @xmath6 , we define @xmath128 and set @xmath129 . we note that @xmath130 for each @xmath131 so that @xmath132 . the map @xmath122 is called the _ contraction map for @xmath72 _ , and can be found in ( * ? ? ? * definition 4.2 ) . let @xmath133 denote the disjoint union of the supports of each prototile @xmath45 . then , @xmath134 is an element of @xmath135 , and it is straightforward to check that @xmath136 . in the sequel , we will often abuse notation , and identify the prototile set @xmath73 with the corresponding supports @xmath137 . then , it is implicitly understood that these compact subsets of @xmath135 carry their label . for example , @xmath138 is the collection of labelled compact subsets @xmath139 in @xmath140 . moreover , @xmath141 can be iterated , and for each @xmath86 , we call each scaled down tile @xmath142 an _ @xmath87-subtile _ , and each @xmath143 a _ subtile_. in this subsection , we look at a particular subset of the continuous hull of a tiling , called the _ discrete hull_. let @xmath4 be a tiling with prototile set @xmath5 . following kellendonk ( * ? ? ? * section 2.1 ) , for each @xmath6 , we choose a point in the interior of @xmath8 . this point is called the _ puncture _ of @xmath8 , denoted @xmath144 . for any vector @xmath145 , we choose the puncture of the tile @xmath146 to be @xmath147 . in ( * ? ? ? * section 2.1 ) , it is shown that each tile @xmath15 of every tiling in the continuous hull @xmath30 , can be assigned a puncture that is consistent in the sense that if @xmath146 , then @xmath148 . [ defn : discretehull ] the _ discrete hull _ of a tiling @xmath44 , is given by @xmath149 the tilings in the discrete hull are those that contain a tile whose puncture is at the origin of @xmath2 . for each @xmath150 , let @xmath151 denote the unique tile @xmath152 whose puncture is the origin . in the relative topology inherited from @xmath30 , and assuming finite local complexity , the discrete hull @xmath153 is a cantor set . in particular , @xmath153 is a compact metric space which has a basis of both open and closed sets . indeed , given a patch @xmath50 and a tile @xmath15 in @xmath50 , the set @xmath154 is both open and closed in @xmath153 , and the set of all such sets forms a basis for the topology on @xmath153 . let @xmath155 . we say that @xmath68 and @xmath156 are _ translation equivalent _ if there exists @xmath152 such that @xmath157 . we denote by @xmath158 $ ] the translation equivalence class of @xmath68 . let @xmath72 be a primitive substitution rule on a set of prototiles @xmath5 . for the remainder of this paper , we assume that @xmath159 for each @xmath6 , so that each puncture is a fixed point under the substitution rule . the assumptions we have made on the substitution rule @xmath72 , and the set of prototiles @xmath5 , force @xmath160 for all @xmath161 . furthermore , for each @xmath6 , @xmath159 implies @xmath162 . in @xcite , the authors construct fractal dual substitution tilings from a given substitution tiling . at the core of their construction , is the notion of a _ recurrent pair_. in this section , we give a basic construction using dual trees , rather than quasi - dual trees , of the fractal dual substitution tilings described in @xcite . we note that the constructions in this paper work in full generality of quasi - dual trees , and are stated as such . in each definition , we refer the reader to the more general definition for quasi - dual trees found in @xcite . we begin with the definition of a consistent dual tree . let @xmath4 be a substitution tiling with finite local complexity , and prototile set @xmath5 . a _ dual tree _ @xmath163 in a prototile @xmath6 , consists of a vertex @xmath164 in the interior of @xmath8 , and a collection of non - overlapping edges connecting @xmath164 to a boundary vertex in the interior of each edge of @xmath8 . considering @xmath73 as an element of @xmath39 , we say that @xmath165 is a _ consistent dual tree _ in @xmath73 , if @xmath163 is a dual tree in @xmath8 for each @xmath45 , and if two translated prototile edges meet in the tiling @xmath4 , then the associated boundary vertices of @xmath166 meet in @xmath4 as well . let @xmath44 be a tiling with finite local complexity , and prototile set @xmath73 . suppose that @xmath167 is a consistent dual tree in @xmath73 . for each @xmath45 , we denote the set of edges in @xmath168 by @xmath169 , the set of vertices in @xmath168 by @xmath170 , and @xmath171 for those vertices which lie on @xmath36 . moreover , for each @xmath86 , we write @xmath172 for the images of edges in each @xmath173 under @xmath174 . let @xmath4 be a substitution tiling with finite local complexity , substitution rule @xmath91 , and prototile set @xmath73 . a pair of consistent dual trees @xmath175 on @xmath5 , is called a _ recurrent pair _ if @xmath176 for some @xmath177 . under certain conditions , developed in detail in ( * section 7.1 ) , the attractor of a recurrent pair @xmath175 is a consistent dual tree @xmath26 inscribed in the prototile set @xmath5 . since the graph @xmath26 is the attractor of the graph iterated function system defined by equation ( 7.2 ) in ( * ? ? ? * section 7.1 ) , we call @xmath26 a _ fractal graph_. moreover , @xmath26 is invariant under this graph iterated function system . it is well known that embedding a consistent dual tree into each prototile of a tiling yields a new tiling , often called the combinatorial dual of @xmath4 . in this case , if we embed the dual trees in @xmath26 into each tile of @xmath4 , we obtain a new tiling that is also a substitution tiling , since @xmath26 is built from the substitution of @xmath4 . [ thm6.6-fww ] let @xmath4 be a substitution tiling with finite local complexity , whose tiles meet singly edge - to - edge . then @xmath4 has an infinite number of distinct fractal quasi - dual substitution tilings . if the prototiles of @xmath4 are all convex , then @xmath4 has an infinite number of distinct fractal dual substitution tilings . [ penrose_recurrentpair ] recall the penrose tiling constructed in example [ eg : penrose ] . the trees giving rise to a recurrent pair @xmath175 are depicted in figure [ eg : penrosefractal ] on two prototiles , and extend uniquely to all forty prototiles by rotation and reflection . the unique fractal graph @xmath26 defined by the recurrent pair @xmath175 is depicted on the right hand side of figure [ eg : penrosefractal ] , and also extends to all forty prototiles . \(a ) at ( 0,4 ) [ label = below:@xmath166 ] ; ( b ) at ( 3.5,4 ) [ label = below:@xmath179 edge[<- ] ( a ) ; ( c ) at ( 7,4 ) [ label = below:@xmath180 edge[<- ] ( b ) ; ( d ) at ( 10.5,4 ) [ label = below:@xmath26 ] edge[<-,dashed ] ( c ) ; \(e ) at ( 1,1.5 ) [ label = below:@xmath166 ] ; ( f ) at ( 4,1.5 ) [ label = below:@xmath179 edge[<- ] ( e ) ; ( g ) at ( 7,1.5 ) [ label = below:@xmath180 edge[<- ] ( f ) ; ( h ) at ( 10,1.5 ) [ label = below:@xmath26 ] edge[<-,dashed ] ( g ) ; placing the graph @xmath26 in each tile of a penrose tiling @xmath4 , defines a new tiling whose tiles have fractal borders . figure [ eg : penrosefractaltiling ] shows a patch of the fractal dual tiling of the penrose tiling overlaid on a patch of the original penrose tiling . the substitution on the fractal realisation is inherited from the original penrose tiling . see ( * ? ? ? * appendix a ) for several additional examples . now that we have defined recurrent pairs and the associated fractal dual , we are now able to define fractal trees . these are defined as decorations on the tiling @xmath7 for each @xmath6 . in short , a fractal tree on a tiling is a network of fractal edges , such that the punctures of two distinct tiles are uniquely connected by a fractal path . we use fractal trees to define spectral triples . let @xmath175 be a recurrent pair of quasi - dual trees , such that for each @xmath6 , the puncture @xmath182 is an interior vertex in @xmath168 . the algorithm in the proof of ( * ? ? ? * theorem 6.6 ) implies it is always possible to set the corresponding interior vertex in @xmath183 to be @xmath144 as well . this forces an interior vertex of the fractal graph @xmath184 to be @xmath144 . we make this additional assumption for the remainder of this paper . [ fractaltree ] let @xmath72 be a primitive substitution rule on a set of prototiles @xmath5 , with a recurrent pair @xmath175 of quasi - dual trees . then , for each @xmath6 , there exists a geometric graph @xmath185 in @xmath7 such that @xmath185 is a connected tree , and @xmath186 for each tile @xmath187 . let @xmath175 be a recurrent pair of quasi - dual trees on @xmath5 , and let @xmath26 be the associated fractal graph . we first define a subgraph @xmath188 of @xmath26 that does not meet the boundary of any prototile . for each @xmath6 , view @xmath189 as a graph consisting of edges in @xmath190 , and define @xmath191 then @xmath192 is a subgraph of @xmath26 that does not meet the boundary of any prototile . we now claim that we can construct a graph @xmath193 in @xmath5 such that for each @xmath45 , 1 . @xmath194 is a connected tree ; [ connect ] 2 . @xmath195 for all @xmath196 ; and [ puncture ] 3 . @xmath197 with @xmath198 a subgraph of @xmath194 . [ interior ] before constructing the graph @xmath199 , we comment on the significance of conditions . condition ensures that any pair of vertices in @xmath194 are connected by a unique path for each @xmath6 . condition says that the graph @xmath194 passes through the puncture of every subtile in @xmath138 for each @xmath6 . this means that every subtile is connected to the graph @xmath199 via its puncture . condition ensures that the graph @xmath194 is wholly contained within the interior of the support of @xmath8 for each @xmath6 . this means that the graph contains no edges which have a boundary vertex as one of its endpoints . moreover , condition also says that we are retaining the basic structure of @xmath200 . we now construct the graph @xmath201 inductively . first , set @xmath202 . for @xmath203 , if @xmath204 does not satisfy conditions , then there exists @xmath45 , and a subtile @xmath196 such that @xmath205 . we construct @xmath206 as follows . for each @xmath207 such that @xmath208 , we define @xmath209 . now , since @xmath190 connects the punctures of the subtiles in @xmath210 , there is a path @xmath211 in @xmath190 connecting @xmath212 to @xmath213 which intersects @xmath213 at a single vertex . define @xmath214 to be the graph @xmath213 along with the edges comprising the path @xmath211 . then @xmath212 is connected to the graph @xmath214 as a degree one vertex . since there are only a finite number of prototiles , and hence subtiles , there exists @xmath215 sufficiently large such that @xmath216 for all subtitles @xmath217 . then it is routine to check that the graph @xmath218 satisfies , proving the claim . for @xmath219 , we inductively define @xmath220 by the formula , @xmath221 we claim that for all @xmath92 , @xmath220 is a graph in @xmath73 such that for each @xmath45 , a. @xmath222 is a connected tree ; [ ind_connect ] b. @xmath223 for all @xmath224 ; and [ ind_puncture ] c. @xmath225 . [ ind_interior ] since @xmath226 is built from the attractor @xmath200 , the formula for @xmath226 , given in equation , gives a well - defined subgraph of @xmath227 . for the remainder of the claim , we proceed by induction on @xmath87 . by conditions , @xmath199 satisfies . assume @xmath228 satisfies . we prove @xmath220 does as well . a. by our inductive hypothesis , @xmath229 is a connected tree in each subtile @xmath143 . moreover , there exists a unique path between any two subtiles in @xmath230 via the connected tree @xmath231 . thus , the union @xmath232 , which is a subgraph of @xmath227 , is a connected tree in @xmath233 . we remark that the invariance of @xmath26 under the map @xmath122 is essential for this step to work , which is why we require fractal trees rather than trees . b. by our inductive hypothesis , @xmath234 for all @xmath235 . thus , @xmath236 for all @xmath224 . since @xmath237 , we have @xmath223 for all @xmath224 . c. by our inductive hypothesis , @xmath238 and hence , @xmath229 is contained in the union of the interior of the subtiles in @xmath138 . that is , @xmath239 . thus by induction , there exists a geometric graph @xmath222 satisfying conditions for each @xmath92 . we now define a fractal graph @xmath240 on @xmath241 for each @xmath92 in the following way . for each @xmath6 and @xmath92 , define @xmath242 . by construction , @xmath243 is a geometric graph in @xmath88 containing the puncture of each tile in @xmath88 . moreover , @xmath244 for each @xmath92 . by construction of @xmath7 , we have @xmath245 for all @xmath86 , so that @xmath246 is a geometric graph on @xmath7 connecting the punctures in @xmath7 by a unique fractal path . let @xmath91 be a primitive substitution rule on a set of prototiles @xmath73 , with a recurrent pair @xmath247 or quasi - dual trees . for each @xmath6 , we will refer to a geometric graph @xmath185 , as in the statement of theorem [ fractaltree ] , as a _ fractal tree _ on @xmath7 . for fixed @xmath6 , suppose that @xmath7 is endowed with a fractal tree @xmath185 . we define a distance between two tiles in @xmath7 via @xmath185 . since fractal edges typically have infinite euclidean length , the length of a fractal path is defined using perron - frobenius theory . let @xmath175 be the recurrent pair which we used to construct the fractal tree @xmath185 , as in theorem [ fractaltree ] , and let @xmath26 be the corresponding fractal graph . since @xmath185 consists of edges in @xmath26 , we assign a length to each edge in @xmath26 and then add the appropriate edge lengths to obtain a length between punctures of tiles in @xmath7 . since @xmath175 is a recurrent pair , @xmath166 and @xmath248 are homeomorphic geometric graphs ( planar graphs embedded in @xmath2 ) . let @xmath249 denote the homeomorphism . for each edge @xmath250 , the edge @xmath251 is composed of edges of the form @xmath252 where @xmath253 , and @xmath40 . we may view @xmath254 as a substitution rule on the set of edges . let @xmath73 be a set of prototiles , and @xmath247 a recurrent pair in @xmath73 . denote by @xmath255 the homeomorphism between @xmath173 and @xmath256 . for each @xmath97 , @xmath257 and @xmath258 , define @xmath259 the matrix @xmath260 is called the _ edge digit matrix_. further , let @xmath261 . that is , @xmath262 is the number of scaled copies of @xmath263 in @xmath264 . the matrix @xmath265 is the _ edge substitution matrix_. when @xmath266 is primitive , the perron - frobenius theorem defines a unique real eigenvalue @xmath267 , with associated eigenvector @xmath268 , whose entries are all positive real numbers . we note that @xmath266 may not be primitive for all choices of recurrent pairs , however , it is routine to see that a recurrent pair @xmath175 with primitive substitution matrix @xmath266 can always be chosen for all tilings satisfying the hypothesis of ( * ? ? ? * theorem 6.6 ) ( see also theorem [ thm6.6-fww ] ) . suppose that @xmath247 is a recurrent pair in a set of prototiles @xmath73 , and let @xmath200 be the associated fractal graph . let @xmath268 be the perron - frobenius eigenvector for the primitive edge substitution matrix @xmath266 . for each @xmath269 , we write @xmath270 for the _ length _ of @xmath271 . [ defn : fractal_distance ] let @xmath73 be a set of prototiles , and fix @xmath45 . suppose @xmath272 , and @xmath185 is a fractal tree on @xmath7 . the _ fractal distance _ between @xmath15 and @xmath16 , denoted @xmath14 , is the sum of the fractal edges making up the unique fractal path in @xmath185 between the punctures @xmath273 and @xmath274 . [ eg : fractaltree ] in this example , we illustrate the construction of a fractal tree for the recurrent pair defined on the penrose tiling in example [ penrose_recurrentpair ] . we then use the perron - frobenius theorem to assign lengths to the fractal edges depicted in figure [ fig : fractal_edges ] , which extend to the edges in all forty prototiles by rotation . putting these two steps together defines a fractal distance between the punctures of tiles @xmath15 and @xmath16 in @xmath7 , for all @xmath6 . to elucidate the algorithm used in the proof of theorem [ fractaltree ] , we restrict our attention to prototile @xmath275 . starting with the recurrent pair in example [ penrose_recurrentpair ] , the first step is to construct the graph @xmath276 by removing the edges of @xmath277 in @xmath278 that intersect the boundary of @xmath279 . from here , we need to add in extra edges from @xmath280 to form a graph @xmath281 in @xmath279 which passes through the puncture of each subtile in @xmath230 . note that in this example , we do not need to add in any extra edges since all punctures are already contained in @xmath282 . the graph @xmath283 appears in figure [ fig : fractal_tree1 ] . the second step in the algorithm , illustrated in figure [ fig : fractal_tree2 ] , is to consider the graph @xmath284 , which will define a graph in each prototile that intersects each puncture in @xmath285 . notice , however , that for each @xmath207 , @xmath286 contains no edges whose endpoints are boundary vertices , and hence , not all pairs of punctures of the @xmath287-subtiles in @xmath288 are connected via the graph @xmath289 . we define @xmath290 , so that each pair of connected graphs in @xmath201 are joined by additional fractal edges . we note that it is essential that we begin with a fractal graph , so that @xmath199 and @xmath284 intersect along complete edges . in figure [ fig : fractal_tree3 ] , we illustrate the graph @xmath291 . inductively , we obtain graphs @xmath220 in @xmath73 for all @xmath92 . for each @xmath92 , consider @xmath292 which connects the punctures of each tile in each patch @xmath88 for each @xmath6 . finally , the infinite graph @xmath293 on tiling @xmath294 , is defined to be the infinite union @xmath295 . thus we have a fractal tree on the tiling @xmath294 that jointly extends each of the fractal trees @xmath296 , and the patches @xmath297 . we now assign lengths to the edges @xmath298@xmath299 in figure [ fig : fractal_edges ] . the matrix @xmath266 for the recurrent pair @xmath175 defined in example [ penrose_recurrentpair ] , is a primitive @xmath300 matrix . however , since the edges in figure [ fig : fractal_edges ] extend to the @xmath301-fold rotations of the @xmath302 prototiles , we need only consider a @xmath303 matrix . the perron - frobenius eigenvalue is @xmath304 with associated eigenvector @xmath305^t.\ ] ] for each @xmath306 , entry @xmath307 is the length @xmath308 of edge @xmath309 . to obtain the fractal distance @xmath310 between two tiles @xmath272 , we sum the lengths of the edges between the punctures @xmath273 and @xmath274 thus , the reader can verify that the fractal distances between the middle pink tile in figure [ fig : fractal_tree3 ] , and each of its three adjacent tiles is @xmath311 , @xmath312 , and @xmath313 in the northwest , northeast , and south directions , respectively . for the remainder of this section , fix a prototile set @xmath73 with a substitution rule @xmath91 . furthermore , for each @xmath45 , let @xmath314 be a fractal tree in @xmath9 . write @xmath315 for the perron - frobenius eigenvalue of the associated edge substitution matrix . we explore some properties associated with the fractal tree , and make some new definitions which will aid us in the definition of a spectral triple . [ root_of_prototile ] let @xmath45 , @xmath316 , and @xmath86 . the _ root _ of the patch @xmath317 is the tile @xmath318 . in particular , taking @xmath319 , @xmath8 is the root of the patch @xmath88 for each @xmath45 , and @xmath92 . [ fscale ] fix @xmath45 , and suppose that @xmath320 are tiles in @xmath9 such that @xmath53 is the root of @xmath321 , and @xmath322 is the root of @xmath323 . then , @xmath324 choose @xmath325 sufficiently large such that @xmath326 as @xmath327-subtiles . by construction , there is a unique fractal path between @xmath328 and @xmath329 via the tree @xmath330 , as in the proof of theorem [ fractaltree ] . by equation , we have @xmath331 so that @xmath332 . thus , the fractal path between @xmath333 and @xmath334 in @xmath335 , is identical to the fractal path between @xmath336 and @xmath337 in @xmath338 . scaling appropriately , the fractal path between @xmath53 and @xmath322 is the image under the edge digit matrix of the fractal path between @xmath15 and @xmath16 . since the length of fractal edges were assigned by the eigenvector corresponding to the edge substitution matrix with perron - frobenius eigenvalue @xmath315 , we conclude that @xmath339 . [ radii ] for each @xmath6 , and @xmath92 , the _ @xmath87th major branch radius of @xmath8 _ is the number @xmath340 a tile @xmath111 satisfying @xmath341 is called an _ @xmath87th major leaf of @xmath8_. the _ major branch radius _ is @xmath342 . similarly , for each @xmath6 and @xmath343 such that @xmath344 the _ @xmath345th minor branch radius of @xmath8 over @xmath87 _ is the number @xmath346 if @xmath347 , we have @xmath348 , and we write @xmath349 for the _ @xmath345th minor branch radius of @xmath8_. a tile @xmath350 satisfying @xmath351 is called an _ @xmath345th minor leaf of @xmath8 over @xmath87_. in particular , a tile @xmath352 satisfying @xmath353 is called an _ @xmath345th minor leaf of @xmath8_. the _ minor branch radius _ is given by @xmath354 . [ longest ] for each @xmath6 , and @xmath92 , we have @xmath355 . we proceed by induction . by defintion , we have @xmath356 , and hence , @xmath357 for all @xmath6 . assume for some @xmath203 , we have @xmath358 for all @xmath45 . fix @xmath6 , and suppose @xmath15 is a first major leaf of @xmath8 . let @xmath53 be the root of @xmath359 . since @xmath8 is the root of @xmath360 , @xmath361 by lemma [ fscale ] . thus , if we view the patch @xmath362 as a collection of supertiles of order @xmath363 , the fractal distance from @xmath8 to the root of one of the supertiles of order @xmath363 is at most @xmath364 . let @xmath16 be a @xmath365th major leaf of @xmath8 . then @xmath366 for some @xmath367 . that is , @xmath16 must be contained in one of the supertiles of order @xmath363 . let @xmath322 be the root of @xmath368 . then , by our inductive hypothesis , @xmath369 [ shortcorona ] for each @xmath6 , and @xmath343 with @xmath344 , we have @xmath370 . let @xmath6 , and @xmath371 . suppose that @xmath372 . then @xmath373 suppose that @xmath374 . let @xmath15 be a first minor leaf of @xmath8 . let @xmath53 be the root of @xmath375 , and note that @xmath376 . moreover , @xmath377 satisfies @xmath378 . let @xmath16 be an @xmath345th minor leaf of @xmath8 over @xmath87 , then the fractal path between @xmath8 and @xmath16 must pass through the root of @xmath379 for some @xmath367 . thus , @xmath380 . given a set of prototiles @xmath73 with a substitution rule @xmath91 , and fractal trees on @xmath9 for each @xmath45 , we now have a means of estimating the various major and minor branch radii . we will use these estimates later in the proofs of our results in section [ sec : spectraltriples ] , where we define spectral triples . in this section , we investigate the @xmath0-algebra associated to a nonperiodic substitution tiling given by kellendonk in @xcite , on which we define a spectral triple . let @xmath72 be a primitive substitution rule on a set of prototiles @xmath5 , and @xmath153 be the associated discrete hull . recall the translational equivalence relation on the discrete hull , given in section [ sec : discretehull ] , which may be written as : @xmath381 we endow @xmath382 with the metric given by , @xmath383 where @xmath384 , @xmath58 , and @xmath385 . in this topology , @xmath382 is an tale equivalence relation . note that the topology here is not the same as the topology inherited from the product topology on @xmath386 . in particular , in the latter topology , @xmath382 is not tale . following the construction of groupoid @xmath0-algebras given by renault in @xcite , kellendonk constructed a @xmath0-algebra in @xcite , which we denote by @xmath3 . we now outline this construction . consider the complex vector space @xmath387 ; the continuous functions of compact support on @xmath382 . endow @xmath387 with multiplication and involution given by , @xmath388 } f(t , t '' ) g(t'',t ' ) , \ \text{and } \\ f^*(t , t ' ) : = \overline{f(t',t)},\end{gathered}\ ] ] respectively , where @xmath389 , and @xmath390 . under these operations , @xmath387 is a @xmath391-algebra . for each @xmath392 , we view @xmath44 as a countable collection of tiles , and consider the _ induced representation from the unit space _ , @xmath393 , given in ( * ? ? ? * section 3 ) . explicitly , @xmath394 where @xmath395 , @xmath396 , and @xmath58 . it is well known that each induced representation @xmath397 is non - degenerate ( * ? ? ? * proposition 1.7 ) . completing @xmath387 in the _ reduced @xmath0-algebra norm _ , @xmath398 where @xmath399 , we obtain kellondonk s @xmath0-algebra @xmath3 . given kellendonk s @xmath0-algebra @xmath3 , we now describe a dense spanning set , see ( * ? ? ? * section 2.2 ) . let @xmath50 be a patch , and let @xmath400 be tiles in @xmath50 . define a function @xmath401 by @xmath402 then , @xmath403 is a partial isometry in @xmath387 . the linear span of all such functions @xmath403 forms a dense spanning set for @xmath3 . we note some of the relations satisfied by these functions . let @xmath404 and @xmath405 be patches , @xmath406 be tiles in @xmath404 , and @xmath407 be tiles in @xmath405 . it is easy to check that @xmath408 furthermore , if there exists a patch @xmath50 such that @xmath409 and @xmath410 , then @xmath411 we note that @xmath3 is a unital @xmath0-algebra , with unit @xmath412 to close this section , we define another representation of @xmath413 that we require in the next section . for each @xmath6 , consider the map @xmath414 defined by @xmath415 where @xmath399 , @xmath416 and @xmath417 . we note that @xmath418 is similar to an induced representation of the unit space . in fact , the only difference here is that we are removing the tile @xmath8 from the tiling @xmath7 . the reason we have removed this particular tile will become apparent in the next section . that @xmath418 is a non - degenerate representation on @xmath387 is similar to the proof for the induced representation found in ( * ? ? ? * proposition 1.7 ) . since the substitution on @xmath419 is assumed to be primitive , every finite patch appearing in any tiling in @xmath419 also appears in @xmath420 . using this fact , it is routine to show that @xmath3 is isometrically isomorphic to the closure of @xmath387 is the operator norm of @xmath421 . thus , @xmath418 extends to a faithful representation of @xmath3 . in this section , we define a spectral triple on kellendonk s @xmath0-algebra @xmath3 , introduced in the previous section . we begin with the definition of a spectral triple on a @xmath0-algebra . [ spectral_triple ] let @xmath26 be a unital @xmath0-algebra . spectral triple over @xmath26 _ , @xmath25 , consists of a separable hilbert space @xmath27 , a faithful representation @xmath422 , and an unbounded , self - adjoint operator @xmath28 on @xmath27 , satisfying the following properties : [ nposspectrip ] let @xmath73 be a set of prototiles with nonperiodic substitution rule @xmath91 . for each @xmath6 , let @xmath7 be the self - similar tiling from definition [ tilinggeneratedbyp ] . let @xmath175 be a recurrent pair on @xmath5 satisfying the hypotheses of theorem [ fractaltree ] , so that there is a fractal tree @xmath185 on @xmath7 , for all @xmath6 . 1 . for each @xmath6 , there is a @xmath1-summable spectral triple @xmath20 where @xmath427 , @xmath428 is defined in equation , and @xmath429 is an unbounded self - adjoint operator on @xmath430 defined on the canonical basis @xmath18 by @xmath431 . [ spec_trip_p ] 2 . for each @xmath432 there is a @xmath1-summable spectral triple @xmath24 where @xmath433 [ spec_trip_p ] the remainder of the paper will be dedicated to proving theorem [ nposspectrip ] . to this end , we fix a set of prototiles @xmath73 , a nonperiodic substitution rule @xmath91 on @xmath73 with scaling factor @xmath434 , and a fractal tree @xmath314 in @xmath9 for each @xmath45 , with @xmath315 corresponding to the perron - frobenius eigenvalue of the edge substitution matrix . we note that the spectral triples constructed in of theorem [ nposspectrip ] , are bounded below , and therefore , @xmath429 may be taken to be a positive operator ( by suitably scaling the perron - frobenius eigenvector so that its minimal entry is greater than @xmath435 ) . this is the primary reason for the more sophisticated spectral triple appearing in of theorem [ nposspectrip ] . we first concentrate on part of theorem [ nposspectrip ] . fix @xmath436 , and consider the @xmath0-algebra @xmath3 , the hilbert space @xmath437 , and the faithful representation @xmath438 of @xmath413 on @xmath439 as defined in equation , which we saw in the previous section . for @xmath272 , recall the fractal distance @xmath440 defined in definition [ defn : fractal_distance ] . we define @xmath429 on the canonical basis @xmath441 of @xmath430 by @xmath442 where @xmath417 , and we extend @xmath443 by linearity to @xmath444 . since @xmath445 , we have @xmath446 for all @xmath417 , and hence , @xmath447 is well - defined . since @xmath448 for each @xmath417 , @xmath429 is a densely defined operator on @xmath430 . furthermore , @xmath429 is an unbounded operator . to see this , fix a tile @xmath417 , and @xmath92 . choose @xmath215 such that @xmath449 , and let @xmath16 be the root of @xmath450 . since @xmath8 is the root of @xmath451 , lemma [ fscale ] implies @xmath452 then , considering the operator norm of @xmath443 , we obtain @xmath453 so that @xmath429 is unbounded . finally , we show that @xmath429 is self - adjoint . that @xmath429 is symmetric follows immediately from the definition of @xmath429 on @xmath430 . to see that @xmath429 is self - adjoint , consider @xmath454 since @xmath429 is symmetric , we have @xmath455 . we show that @xmath456 . let @xmath457 . then the map @xmath458 from @xmath459 to @xmath460 is bounded . by the riesz representation theorem , there exists @xmath461 such that @xmath462 for all @xmath463 . for each @xmath464 , we have @xmath465 , and hence , @xmath466 furthermore , @xmath467 , so that @xmath468 . thus , we obtain , @xmath469 so that @xmath470 . thus , @xmath456 , and hence , @xmath429 is self - adjoint . given our unbounded and self - adjoint operator @xmath429 on @xmath430 , we show that @xmath20 is a @xmath471-summable spectral triple . to this end , we show that conditions and in definition [ spectral_triple ] are satisfied . proposition [ condition_3a ] implies condition in definition [ spectral_triple ] , since the linear span of functions @xmath403 is dense in @xmath3 . we require a sequence of lemmas before arriving at the proof of proposition [ condition_3a ] . given a patch @xmath50 in a tiling @xmath472 , we partition @xmath7 in such a way that we confine where @xmath50 can appear in @xmath7 . the following lemma will help us to do this , and the corollary which follows describes what sort of control we have gained . first , we need the following definition . by corollary [ patchinsupertile ] , there exists @xmath215 sufficiently large such that @xmath495 , or @xmath496 for some @xmath479 . suppose that @xmath495 . by definition [ radii ] , we have @xmath499 otherwise , @xmath496 for some @xmath479 . by definition of the @xmath500th minor branch radius of @xmath8 over @xmath363 in definition [ radii ] , the fact that @xmath501 , and lemmas [ longest ] and [ shortcorona ] , we have @xmath502 thus , combining equations and , we obtain @xmath503 let @xmath50 be a patch in a tiling @xmath472 , @xmath473 , and @xmath504 . making the necessary modifications to the equation immediately preceding ( * ? ? ? * proposition 3.3 ) , we have the following formula for the representation of @xmath403 in @xmath505 : @xmath506 it is important to note that if @xmath507 , then @xmath508 . given equation , we are now in a position to prove proposition [ condition_3a ] . let @xmath50 be a patch in a tiling @xmath472 , @xmath473 and @xmath504 . if @xmath509 , let @xmath510 denote the tile @xmath511 . we begin by calculating @xmath512 \delta_{t '' } & = d_p \pi_p(e(p , t , t ' ) ) \delta_{t''}-\pi_p(e(p , t , t ' ) ) d_p \delta_{t '' } \\ & = d_p \delta_s-\ln(d_{\mathfrak{f}_p}(t'',p ) ) \pi_p(e(p , t , t ' ) ) \delta_{t '' } \\ & = \ln(d_{\mathfrak{f}_p}(s , p ) ) \delta_s-\ln(d_{\mathfrak{f}_p}(t'',p ) ) \delta_s \\ & = \ln \left [ \frac{d_{\mathfrak{f}_p}(s , p)}{d_{\mathfrak{f}_p}(t'',p ) } \right ] \delta_s.\end{aligned}\ ] ] by equation , if @xmath513 then @xmath514 \delta_{t '' } \equiv 0 $ ] . suppose then , that @xmath515 . thus , to prove that @xmath514 \in \bb(h_p)$ ] , we show that @xmath516 \right| \leq m,\ ] ] for some fixed @xmath517 . using lemma [ allofwork ] , fix @xmath215 such that equation holds . we claim @xmath518,\ ] ] does the job . suppose that @xmath519 . then , by lemma [ allofwork ] , we have @xmath520 \leq m.\ ] ] otherwise , @xmath521 and again by lemma [ allofwork ] , we have @xmath522 \leq m.\ ] ] thus , from equations and , we have @xmath523 \right| = \left| \ln \left [ \frac{d_{\mathfrak{f}_p}(s , p)}{d_{\mathfrak{f}_p}(t'',p ) } \right ] \right| \implies 0 \leq \left| \ln \left [ \frac{d_{\mathfrak{f}_p}(s , p)}{d_{\mathfrak{f}_p}(t'',p ) } \right ] \right| \leq m,\ ] ] as claimed . thus , for any patch @xmath524 in any tiling @xmath525 , and @xmath526 , we have @xmath474 \in \bb(h_p ) . \qedhere\ ] ] by proposition [ condition_3a ] , we now know that @xmath20 satisfies condition of definition [ spectral_triple ] . for condition of definition [ spectral_triple ] , we need to show that the operator @xmath527 . we show that @xmath528 is trace class for all @xmath426 , and invoke the following result found in ( * ? ? ? * section 7 ) . if we are able to show that @xmath531 is trace class for all @xmath532 , then lemma [ thetatocompact ] will imply that @xmath533 is a @xmath471-summable spectral triple , and hence , we will be able to prove part of theorem [ nposspectrip ] . before proceeding , we need the following technical lemma . fix @xmath426 . for any @xmath417 , we have @xmath543 from this , we calculate @xmath544 since @xmath545 and @xmath426 , we have @xmath546 . fix @xmath215 such that @xmath547 . by lemma [ leftfield ] , there exists @xmath534 such that @xmath548 for all @xmath536 . since @xmath549 for all @xmath417 , we have @xmath550 from here , let us restrict our attention to the sum in equation , which we rewrite as follows . for each @xmath417 , we have @xmath551 , or @xmath552 for some @xmath371 . to simplify notation , define @xmath553 for each @xmath371 . thus , @xmath554 the first sum on the right hand side of equation is finite since @xmath555 is a patch . thus , we only concern ourselves with the second sum on the right hand side of equation . for each @xmath371 , lemma [ shortcorona ] , with @xmath556 , implies that @xmath557 for any @xmath558 . thus , @xmath559 for each @xmath371 , the sum over @xmath558 no longer depends on @xmath15 . thus , we multiply @xmath560 by the number of tiles in @xmath561 , which is certainly smaller than @xmath562 . thus , we have @xmath563 by our choice of @xmath325 , we have @xmath564 , so we have a convergent geometric series . tracing back to equation , we have @xmath565 . thus , @xmath528 is trace class , as required . the operator @xmath429 is an unbounded and self - adjoint operator on @xmath430 , which satisfies condition of definition [ spectral_triple ] , by proposition [ condition_3a ] . moreover , since the operator @xmath528 is trace class for all @xmath426 , by proposition [ thetasum ] , lemma [ thetatocompact ] implies that @xmath566 , which is condition of definition [ spectral_triple ] . thus , @xmath567 is a @xmath1-summable spectral triple . we now turn our attention to part of theorem [ nposspectrip ] . by part of theorem [ nposspectrip ] , we have a collection of spectral triples @xmath567 for each @xmath6 . given @xmath568 , define @xmath569 the map @xmath570 is a non - degenerate faithful representation since @xmath418 is a non - degenerate faithful representation for each @xmath6 . thus , to show that @xmath571 is a spectral triple , we check @xmath572 is an unbounded , self - adjoint operator satisfying conditions and of definition [ spectral_triple ] . for each @xmath45 , we have @xmath576 \in b ( h_p ) $ ] . thus , @xmath577 & = d _ { \sigma } \pi(e(p , t , t ' ) ) - \pi(e(p , t , t ' ) ) d _ { \sigma } \\ & = \bigoplus_{p \in \pp}\ \sigma(p ) d_p \pi_p(e(p , t , t ' ) ) - \sigma(p ) \pi_p(e(p , t , t ' ) ) d_p \\ & = \bigoplus_{p \in \pp}\ \sigma(p ) [ d_p,\pi_p(e(p , t , t ' ) ) ] \in \bb(h ) . \qedhere\end{aligned}\ ] ] proposition [ thetasum ] implies that @xmath565 for each @xmath426 and @xmath6 . fix @xmath426 . then , by the holomorphic functional calculus we obtain @xmath579 since @xmath426 was arbitrary , the operator @xmath578 is trace class for all @xmath426 . the operator @xmath572 is an unbounded and self - adjoint operator on @xmath27 , which satisfies condition of definition [ spectral_triple ] , by lemma [ cond_3a_d ] . moreover , since the operator @xmath578 is trace class for all @xmath426 , by lemma [ trace_class ] , lemma [ thetatocompact ] implies that @xmath580 , which is condition of definition [ spectral_triple ] . thus , @xmath581 is a @xmath1-summable spectral triple . 1 . suppose @xmath24 is a spectral triple on @xmath3 associated with a collection of fractal trees @xmath582 , as in theorem [ nposspectrip ] . under what conditions will @xmath24 be a _ quantum metric space _ in the sense of rieffel @xcite ? if @xmath24 is a quantum metric space , then when does this property extend to the crossed product @xmath583 ? we note that @xmath583 is a ruelle algebra , as defined by putnam in @xcite . the paper of bellissard , marcolli , and reihani @xcite is likely to factor into the resolution of this question . 2 . suppose @xmath4 is tiling satisfying the hypotheses of ( * ? ? ? * theorem 6.6 ) ( see also theorem [ thm6.6-fww ] ) . we conjecture that there exists a collection of fractal trees @xmath582 such that the @xmath31-homology group @xmath584 is generated by @xmath585 { : } \ , \sigma:\pp \to \{-1,1\}\}$ ] , where each such @xmath586 defines a spectral triple @xmath24 as in theorem [ nposspectrip ] . j. bellissard , _ gap labelling theorems for schrdinger operators _ , in from number theory to physics ( m. waldschmidt , et al , eds . ) , proc . les houches meeting ( march 1989 ) , springer- verlag , berlin , 1992 .

the fundamental task of a digital receiver is to decide the transmitted symbols in the best possible way , i.e. , with respect to an appropriately defined performance metric . examples of usual performance metrics are the probability of error and the mean square error ( mse ) of a symbol estimator . in a coherent receiver , the symbol decisions are made based on the use of a channel estimate . this paper focuses on examining the optimality of usual estimators such as the minimum variance unbiased ( mvu ) and the minimum mean square error ( mmse ) estimators for these metrics and on proposing better estimators whenever it is necessary . for illustration purposes , this study is performed on a toy channel model , namely a single input single output ( siso ) flat fading channel with additive white gaussian noise ( awgn ) . in this way , this paper highlights the design dependencies of channel estimators on target performance metrics . minimum mean square error ( mmse ) , minimum variance unbiased ( mvu ) , probability of error , single input single output ( siso ) .

1303.4289

estimation and detection are two main concerns in the course of designing a communication system @xcite . the main goal is to design optimal demodulators at the receiver side providing the detector with the necessary sufficient statistics for its decision on the transmitted symbol at a specific observation interval . furthermore , the optimization of the decision device is also a target , i.e. , its design based on such statistical tests which rely on sufficient statistics and minimize the probability of error . a different setup of optimal designs related to radar and sonar systems is to detect the presence of either a deterministic or random signal in noise with least probability of error or false alarm @xcite . although the two aforementioned setups have conceptual differences , they are usually treated in the same fashion . first , an optimal demodulator is necessary to deliver the sufficient statistics to the decision device . then , the decision device , that optimally uses these sufficient statistics , has to be derived . the optimal design of the decision device is formulated in any case as a hypotheses testing problem . moreover , the optimization of the transmitter is another related problem . in this case , the problem turns to be the design of optimal transmission sets , such that the end performance metric , i.e , the probability of error is minimized . depending on the degree of knowledge about the transmission channel at the receiver side , the detector can be coherent , semi - coherent or noncoherent @xcite . the more information about the transmission channel is available , the better the receiver s performance will be . this justifies the fact that the receivers usually have a built - in channel estimator . in the communication and signal processing literature , the usual channel estimators are the minimum variance unbiased ( mvu ) and the minimum mean square error ( mmse ) estimators @xcite . the combination of these channel estimators with the optimal decision devices is usually considered to address the problem of determining the optimal receiver . current physical layer ( phy ) standards that have attracted a lot of attention both from the mobile industry and the research community are the wireless interoperability for microwave access ( wimax ) , the long term evolution ( lte ) and the digital video broadcasting ( dvb ) either in its terrestrial ( dvb - t ) or its handheld ( dvb - h ) versions @xcite . these standards are orthogonal frequency division multiple access ( ofdma ) based and they can satisfy the need for shorter communication links to provide truly broadband connectivity services . in these systems , either mvu / least squares ( ls ) or mmse channel estimators are used , usually employing some sort of estimate interpolation through the frame if the goal is to track a time - varying channel @xcite . in this paper , we re - examine the validity of the common belief that the mvu and mmse channel estimators are the best choices to be combined with the optimal detectors , delivering an overall optimal receiver , when finite - sample training is used to estimate the channel . to this end , ideas originating from the system identification field are employed . recent results in optimal experiment design indicate that it is better to design the optimal training for the estimation of a certain set of unknown parameters with respect to optimizing the end performance metric rather than the mean square error of the parameter estimator itself @xcite . we will slightly modify this idea and we will examine if the aforementioned channel estimators are the best choices , when the selection of the channel estimator is made with respect to an appropriately defined end performance metric . for illustration purposes , this study is performed on a toy channel model , namely a single input single output ( siso ) flat fading channel with additive white gaussian noise ( awgn ) . the initial focus is on two different mse criteria . these mse criteria serve to demonstrate the dependence of the optimal channel estimators on the end performance metrics . their choice is based on the simplicity of the analysis that they allow . then , using the obtained results , we will examine the case of the error probability as the performance metric of interest . we show that for several performance metrics examined in this paper , the mvu and mmse channel estimators are suboptimal , while we propose ways to obtain better channel estimators . finally , we numerically compare the performances of the derived channel estimators with those of the mvu and mmse channel estimators for all performance metrics in this paper . these comparisons verify that the optimality of the usual channel estimators with respect to common end performance metrics is questionable . this paper is organized as follows : section [ sec : probst ] defines the problem of designing the channel estimator with respect to the end performance metric . section [ sec : prelim ] presents some results and comments that will be useful in the rest of the paper , while it introduces approximations of the performance metrics that the rest of the analysis will be based on . the optimality of the mvu and mmse channel estimators with respect to the minimization of the symbol estimate mse is examined in section [ sec : dmse ] and subsections therein , while uniformly better channel estimators are also proposed . the same analysis as in section [ sec : dmse ] is pursued in section [ sec : emse ] for a differently defined symbol estimate mse and in section [ sec : minpe ] for a rough approximation ( variation ) of the error probability performance metric . section [ sec : sims ] illustrates the validity of the derived results . finally , section [ sec : concl ] concludes the paper . the received signal model for a siso system , when the channel is considered to be narrowband block fading , is given as follows : @xmath0 where @xmath1 is the observed signal at the receiver side at time instant @xmath2 , @xmath3 is the complex channel impulse response coefficient , @xmath4 is the transmitted symbol at the same time instant taken from an m - ary constellation @xmath5 and @xmath6 is complex , circularly symmetric , gaussian noise with zero mean and variance @xmath7 . given an equiprobable distribution on the constellation symbols , we further assume that @xmath8=0 $ ] and @xmath9=\sigma_x^2 $ ] , while our modulation method is memoryless . in addition , @xmath6 and @xmath4 are independent random sequences , while @xmath6 is a white random sequence . assume that a maximum energy @xmath10 and a training length of @xmath11 time slots are available at the transmitter for training . we can collect the received samples corresponding to training in one vector : @xmath12 where @xmath13^{t}$ ] is the vector of @xmath11 received samples corresponding to training , @xmath14^{t}$ ] is the vector of @xmath11 training symbols and @xmath15^{t}$ ] is the vector of @xmath11 noise samples . considering the class of linear channel estimators , the channel is estimated as follows : @xmath16 where @xmath17 is a @xmath18 channel estimating filter . with the assumptions in ( [ eq : sm ] ) , if the constellation symbols are equiprobable and the channel is perfectly known , the ml detector is optimal @xcite . this is with respect to minimizing the probability that a different symbol from the one transmitted is decided given the transmitted symbol . the ml decision rule is given by the following expression : @xmath19(h)=\arg\min_{\hat{x}(n)\in \mathcal{x}}|y(n)-h\hat{x}(n)|^2 . \label{eq : mldec}\ ] ] here , @xmath20 $ ] denotes the decision of the detector , when the transmitted symbol is @xmath4 . in essence , the ml detector minimizes the probability of error , when the transmitted symbols are equiprobable . when the receiver has a channel estimate @xmath21 , @xmath3 is replaced by @xmath21 in the last expression . a different kind of performance metric is the mse of a _ linear _ symbol estimator . in this paper , we will call the symbol estimator an _ equalizer_. the equalizer uses the channel knowledge and delivers a soft decision of the transmitted symbol , i.e. , a symbol estimate . we will call _ clairvoyant _ the equalizer that has perfect channel knowledge . denoting this equalizer by @xmath22 , we can find its mathematical expression as follows : @xmath23 , \label{eq : mmsechoice}\ ] ] where the expectation is taken over the statistics of @xmath4 and @xmath6 . if we set the derivative of the last expression with respect to @xmath24 to zero and we solve for @xmath24 , then the optimal clairvoyant equalizer is given by the expression @xmath25 we will call this the mmse clairvoyant equalizer . we observe that as the snr increases , i.e. , @xmath26 , @xmath27 . we will call @xmath28 the _ zero forcing _ ( zf ) clairvoyant equalizer . using the above definitions and assuming that the receiver has only an estimate of the channel , the system performance metric is the symbol estimate mse : @xmath29.\label{eq : msedirect}\ ] ] the mse given by ( [ eq : msedirect ] ) can be defined in two different ways : if we assume that the channel is an unknown but otherwise deterministic quantity , then the expectation in ( [ eq : msedirect ] ) does not consider @xmath3 . this leads to an mse expression dependent on the unknown channel @xmath3 . in this case , only the channel estimators that treat the channel as an unknown deterministic variable are meaningful . if we assume that the unknown channel is a random variable , then we can average the mse expression over @xmath3 . in this case , both the estimators that treat the channel as an unknown deterministic variable or as a random variable are meaningful . the former represents the case where the system designer chooses to ignore the knowledge of the channel statistics in the selection of the channel estimator for some reason . in the following , we focus on the zf equalizer , which becomes optimal as the snr increases this choice is made to preserve the simplicity of this paper and to highlight the derived results . the previous mse definition implies the definition of yet another mse that is meaningful in the context of communication systems . given an equalizer , we can define the excess of the symbol estimate based on an equalizer that only knows a channel estimate over the equalizer with perfect channel knowledge , thus leading to @xmath30.\label{eq : mseexcess}\ ] ] in the sequel , this metric will be called _ excess _ mse . our goal will be to determine the optimal channel estimators for fixed training sequences so that each performance metric based on a given equalizer is minimized . to this end , the following section presents some useful ideas . consider the mvu estimator . since it is an unbiased estimator , it satisfies @xmath31 . this condition implies that @xmath32=h$ ] . for our problem assumptions , the mvu estimator can be found by solving the following optimization problem : @xmath33 forming the lagrangian for this problem and zeroing its gradient with respect to @xmath34 , we get : @xmath35 for the sake of completeness , this estimator coincides with the ml and ls channel estimators under our assumptions . if we assume that the prior distribution of @xmath3 is known , then instead of the mvu one could use the mmse channel estimator . with our assumptions and the extra assumption that @xmath36=0 $ ] , one can obtain @xcite @xmath37{\mbox{\boldmath $ x$}_{\rm tr}^{}}}{e[|h|^2]\|{\mbox{\boldmath $ x$}_{\rm tr}^{}}\|^2+\sigma_w^2}.\label{eq : mmse}\ ] ] the @xmath38 of the zf equalizer using a deterministic channel ( `` dc '' ) assumption is @xmath39\sigma_x^2+\sigma_w^2e\left[\frac{1}{|\hat{h}|^2}\right],\label{eq : mseddczf}\ ] ] the corresponding for random channel ( `` rc '' ) is : @xmath40\right]\sigma_x^2+\sigma_w^2e_h\left[e\left[\frac{1}{|\hat{h}|^2}\right]\right],\label{eq : msedrczf}\ ] ] while for the @xmath41 we accordingly have : @xmath42\left(\sigma_x^2+\frac{\sigma_w^2}{|h|^2}\right)\label{eq : emsezf}\ ] ] ( c.f . ( [ eq : mseexcess ] ) ) . the @xmath43 is obtained by averaging the last expression over @xmath3 . depending on the probability distributions of @xmath44 and @xmath45 , the above mse expressions may fail to exist . the mses will be finite if the probability distribution function ( pdf ) of @xmath44 is of order @xmath46 as @xmath47 . a similar condition should hold for the pdf of @xmath45 in the case of @xmath43 . in the opposite case , we end up with an _ infinite moment _ problem . in order to obtain well - behaved channel estimators that will be used in conjunction with the actual performance metrics , some sort of regularization is needed . some ideas for appropriate regularization techniques to use may be obtained by modifying robust estimators ( against heavy - tailed distributions ) , e.g. , by trimming a standard estimator , if it gives a value very close to zero @xcite . an example of such a trimmed estimator is given as follows : @xmath48 where @xmath17 can be any estimator and @xmath49 a regularization parameter . _ remark : _ clearly , the reader may observe that the definition of the trimmed @xmath21 preserves the continuity at @xmath50 . additionally , the event @xmath51 has zero probability since the distribution of @xmath52 is continuous . therefore , in this case @xmath21 can be arbitrarily defined , e.g. , @xmath53 . we focus now on the @xmath54 . assume a fixed @xmath49 . in the appendix , we show that , for a sufficiently small @xmath49 and a sufficiently high snr during training , minimizing @xmath55 is equivalent to minimizing the following approximation @xmath56_0=\frac{e\left[|\hat{h}-h|^2\right]}{e\left[|\hat{h}|^2\right]}\sigma_x^2+\sigma_w^2\frac{1}{e\left[|\hat{h}|^2\right]}.\label{eq : mseddczf0}\ ] ] following similar steps and using some minor additional technicalities , we can work with @xmath57_0=\frac{e_h\left[e\left[|\hat{h}-h|^2\right]\right]}{e_h\left[e\left[|\hat{h}|^2\right]\right]}\sigma_x^2+\sigma_w^2\frac{1}{e_h\left[e\left[|\hat{h}|^2\right]\right]},\label{eq : msedrczf0}\ ] ] instead of @xmath58 . moreover , @xmath59_0 $ ] and @xmath60_0 $ ] can be defined accordingly . we will call the last approximations _ zeroth order _ symbol estimate mses and excess mses , respectively . the following analysis and results will be based on the zeroth order metrics and they will reveal the dependency of the channel estimator s selection on the considered ( any ) end performance metric . _ remarks : _ 1 . a useful , alternative way to consider the zeroth order mses is to view them as affine versions of normalized channel mses , where the actual true channel is @xmath21 and the estimator is @xmath3 . 2 . in the definition of ( [ eq : mseddczf0 ] ) , one can observe that after approximating the mean value of the ratio by the ratio of the mean values the infinite moment problem is eliminated . in the following , all zeroth order metrics will be defined based on the _ non - trimmed _ @xmath21 to ease the derivations . this treatment is approximately valid when @xmath49 is sufficiently small as it is actually shown in eq . ( [ eq : msexdcapp ] ) of the appendix . we now examine the zeroth order symbol estimate mse in the case of the zf equalizer . the optimality of the mvu and mmse channel estimators will be investigated . additionally , the training sequence is assumed fixed . the channel is considered either deterministic or random , depending on the available knowledge of a priori channel statistics and the will of the system designer to ignore or to exploit this knowledge . the expectation operators in eq . ( [ eq : mseddczf0 ] ) are with respect to @xmath61 and @xmath6 . we have : @xmath62_0=\sigma_x^2\frac{\left[|h|^2\left|{\mbox{\scriptsize $ { \mbox{\boldmath $ f$}}$}}^{h}{\mbox{\scriptsize $ { \mbox{\boldmath $ x$}_{\rm tr}^{}}$}}-1\right|^2+\sigma_w^2\left\|{\mbox{\scriptsize $ { \mbox{\boldmath $ f$}}$}}\right\|^2\right]+\frac{\sigma_w^2}{\sigma_x^2}}{|h|^2\left|{\mbox{\scriptsize $ { \mbox{\boldmath $ f$}}$}}^{h}{\mbox{\scriptsize $ { \mbox{\boldmath $ x$}_{\rm tr}^{}}$}}\right|^2+\sigma_w^2\left\|{\mbox{\scriptsize $ { \mbox{\boldmath $ f$}}$}}\right\|^2}.\end{aligned}\ ] ] the numerator of the gradient of the above expression with respect to @xmath34 discarding the outer @xmath63 is given by the following expression is not affected by these operations . ] : @xmath64\left[|h|^2\left(\varphi-1\right)^{*}{\mbox{\boldmath $ x$}_{\rm tr}^{}}+\sigma_w^2{\mbox{\boldmath $ f$}}\right]\nonumber\\ & & -\left[|h|^2\varphi^{*}{\mbox{\boldmath $ x$}_{\rm tr}^{}}+\sigma_w^2{\mbox{\boldmath $ f$}}\right]\left[\frac{\sigma_w^2}{\sigma_x^2}+|h|^2\left|\varphi-1\right|^2+\sigma_w^2\|{\mbox{\boldmath $ f$}}\|^2\right],\nonumber\\ \label{eq : ddczfnom}\end{aligned}\ ] ] where @xmath65 . setting @xmath66 , we obtain : @xmath67\neq { \mbox{\boldmath $ 0$}}. \label{eq : nommvue}\ ] ] note that no choice of @xmath68 will zero this expression for any @xmath69 . therefore , the mvu is not an optimal channel estimator in this case . we can state this result more formally : [ thrm:1 ] the mvu estimator is _ not _ an optimal channel estimator for the task of minimizing @xmath70_0 $ ] , when the channel is considered deterministic but otherwise an unknown quantity . the question that arises in this case is how to find the optimal channel estimator in this setup or generally how to determine a uniformly better channel estimator for minimizing @xmath71_0 $ ] . equating ( [ eq : ddczfnom ] ) to @xmath72 and taking the inner product of both sides with @xmath17 , we obtain the following necessary condition that every optimal channel estimating filter @xmath17 must satisfy given the training sequence @xmath68 : latexmath:[\[{\mbox{\boldmath $ f$}_{}^{h}}{\mbox{\boldmath $ x$}_{\rm tr}^{}}=\left(1+\frac{\sigma_w^2}{\sigma_x^2 @xmath17 that satisfies this condition is @xmath74 . clearly , ( [ eq : optfzf ] ) is sufficient for ( [ eq : ddczfnom ] ) to become zero however , ( [ eq : optfzf ] ) has another problem , namely that the optimal solution depends on the unknown channel @xmath3 . in order to deal with the dependence of the optimal estimator on the unknown channel , we will resort to a stochastic approach . we will assume a _ noninformative _ prior distribution for the unknown channel . if the real and imaginary parts of the channel are considered bounded in the intervals @xmath75 and @xmath76 then the receiver can treat them as independent random variables uniformly distributed on @xmath77 and @xmath78 , respectively . the @xmath79_0 $ ] is now replaced by @xmath80_0\right\}$ ] , where @xmath81 $ ] denotes the expectation with respect to the joint ( uniform ) distribution of the real and imaginary parts of @xmath3 . applying again the zeroth order approximation and following the above analysis . ] we can easily show that the eqs . ( [ eq : nczf ] ) , ( [ eq : optfzf ] ) and ( [ eq : zf - mvue - opt ] ) give again the necessary condition and optimal estimators in this case with the substitution of @xmath82 by @xmath83 $ ] . in this case , the actual prior statistics of the channel are known . the zeroth order symbol estimate mse is given by @xmath84_0= \sigma_x^2\frac{\left[e[|h|^2]\left|{\mbox{\scriptsize $ { \mbox{\boldmath $ f$}}$}}^{h}{\mbox{\scriptsize $ { \mbox{\boldmath $ x$}_{\rm tr}^{}}$}}-1\right|^2+\sigma_w^2\left\|{\mbox{\scriptsize $ { \mbox{\boldmath $ f$}}$}}\right\|^2\right]+\frac{\sigma_w^2}{\sigma_x^2}}{e[|h|^2]\left|{\mbox{\scriptsize $ { \mbox{\boldmath $ f$}}$}}^{h}{\mbox{\scriptsize $ { \mbox{\boldmath $ x$}_{\rm tr}^{}}$}}\right|^2+\sigma_w^2\left\|{\mbox{\scriptsize $ { \mbox{\boldmath $ f$}}$}}\right\|^2}.\ ] ] differentiating this expression with respect to @xmath34 , we get the numerator of the gradient which is given by ( [ eq : ddczfnom ] ) , but with @xmath82 replaced by @xmath85 $ ] . it can be easily shown that this numerator is different from zero if @xmath66 or @xmath86 . we therefore have a formal statement of this result : the mvu and mmse estimators are _ not _ optimal channel estimators for the task of minimizing @xmath87_0 $ ] , when the prior channel distribution is known.[thrm:2 ] the optimal channel estimator @xmath88 satisfies ( [ eq : nczf ] ) , ( [ eq : optfzf ] ) and ( [ eq : zf - mvue - opt ] ) , but with @xmath82 replaced by @xmath85 $ ] . _ remarks : _ 1 . considering ( [ eq : optfzf ] ) and the corresponding expression for the random channel case , we observe that the design of the estimator with respect to the end performance metric introduces a bias to the mvu estimator in the form of scaling , leading to a smaller value of the end performance metric than the one that we would obtain by using the mvu estimator . this bias introduction mechanism has similarities with the introduction of bias in estimators to reduce their mse ( the mse here is the average square distance of the parameter estimator from the true value of the parameter ) @xcite . nevertheless , the reader may observe the conceptual differences in the motivation and goals behind the end performance metric estimator designs presented in this paper and the ideas in @xcite . the claimed optimality of the derived estimators in this section but also in this paper is with respect to the zeroth order performance metrics . these estimators turn out to be uniformly better than the mvu and mmse estimators also when comparing against the _ true _ end performance metrics as we demonstrate in the simulation section . an alternative way to express eq . ( [ eq : optfzf ] ) is @xmath89 where @xmath90 turns to a tuning parameter controlling the introduction of bias in the mvu estimator . we numerically demonstrate this very interesting aspect of the derived estimators in figs . [ fig : zfmse_biased ] and [ fig : zfexmse_biased ] . since the channel estimator is selected in order to optimize the final performance metric of the communication system , one may consider the problem of selecting optimally the training vector @xmath68 under a training energy constraint @xmath91 to serve the same purpose . to optimize the training vector , one should first fix the channel estimator . this is a `` complementary '' problem with respect to the approach that we have followed so far . suppose that we use either the mvu or the mmse channel estimators . one can observe that for @xmath92 the problem of selecting optimally the training vector is meaningless . therefore , we will end up using an inferior channel estimator ( i.e. , the mvu or the mmse ) than the one given by ( [ eq : zf - mvue - opt ] ) and its random channel counterpart . in the case that @xmath93 , fixing for example @xmath94 one can observe that again the problem of selecting optimally the training vector is meaningless . consider for example the case of @xmath95_0 $ ] . we then have : @xmath57_0=\frac{\frac{\sigma_x^2\sigma_w^2}{\|{\mbox{\boldmath $ x$}_{tr}^{}}\|^2}+\sigma_w^2}{e[|h|^2]+\frac{\sigma_w^2}{\|{\mbox{\boldmath $ x$}_{tr}^{}}\|^2}},\ ] ] which only depends on @xmath96 . furthermore , setting @xmath97 , it follows that @xmath98_0/d\theta<0 $ ] at sufficiently high snr , i.e. , @xmath99_0 $ ] is minimized when @xmath100 , which is intuitively appealing . therefore , any @xmath101 with energy equal to @xmath10 is an equally good training vector for the mvu estimator . thus , for the same @xmath101 , the estimator @xmath102 will be better than the mvu . similar conclusions can be reached for the mmse estimator , as well . we now examine the zeroth order excess mse in the case of the zf equalizer . in this case , we have : @xmath103_0= \frac{|h|^2\left|{\mbox{\scriptsize $ { \mbox{\boldmath $ f$}}$}}^{h}{\mbox{\scriptsize $ { \mbox{\boldmath $ x$}_{\rm tr}^{}}$}}-1\right|^2+\sigma_w^2\left\|{\mbox{\scriptsize $ { \mbox{\boldmath $ f$}}$}}\right\|^2}{|h|^2\left|{\mbox{\scriptsize $ { \mbox{\boldmath $ f$}}$}}^{h}{\mbox{\scriptsize $ { \mbox{\boldmath $ x$}_{\rm tr}^{}}$}}\right|^2+\sigma_w^2\left\|{\mbox{\scriptsize $ { \mbox{\boldmath $ f$}}$}}\right\|^2}\left(\sigma_x^2+\frac{\sigma_w^2}{|h|^2}\right)\ ] ] the numerator of the gradient of the above expression with respect to @xmath34 is given by the following expression : @xmath64\left[|h|^2\left(\varphi-1\right)^{*}{\mbox{\boldmath $ x$}_{\rm tr}^{}}+\sigma_w^2{\mbox{\boldmath $ f$}}\right]\nonumber\\ & & -\left[|h|^2\varphi^{*}{\mbox{\boldmath $ x$}_{\rm tr}^{}}+\sigma_w^2{\mbox{\boldmath $ f$}}\right]\left[|h|^2\left|\varphi-1\right|^2+\sigma_w^2\|{\mbox{\boldmath $ f$}}\|^2\right]\nonumber\\ \label{eq : ezfnom}\end{aligned}\ ] ] setting @xmath66 , one can easily check that the above expression becomes zero . therefore : the mvu _ is _ an optimal channel estimator for the task of minimizing @xmath104_0 $ ] , when the channel is considered a deterministic but otherwise unknown quantity . _ remark : _ note that even if @xmath105_0 $ ] depends on the unknown channel @xmath3 , the optimal channel estimator does not in this case . in this case , the prior statistics of the channel are known . the zeroth order excess mse is given by : @xmath106_0&=&\frac{\left|\varphi-1\right|^2(e[|h|^4]\sigma_x^2+e[|h|^2]\sigma_w^2)}{e[|h|^4]|\varphi|^2+\sigma_w^2\left\|{\mbox{\scriptsize $ { \mbox{\boldmath $ f$}}$}}\right\|^2 e[|h|^2]}\nonumber\\ & & + \frac{\sigma_w^2\left\|{\mbox{\scriptsize $ { \mbox{\boldmath $ f$}}$}}\right\|^2(e[|h|^2]\sigma_x^2+\sigma_w^2)}{e[|h|^4]|\varphi|^2+\sigma_w^2\left\|{\mbox{\scriptsize $ { \mbox{\boldmath $ f$}}$}}\right\|^2 e[|h|^2]}\end{aligned}\ ] ] differentiating this expression w.r.t . @xmath34 and setting @xmath66 we zero the gradient . therefore : the mvu _ is _ an optimal channel estimator for the task of minimizing @xmath107_0 $ ] , when the channel is considered random . via tedious calculations , we can show that the mmse channel estimator does not zero the gradient . _ remark : _ this result is _ counterintuitive _ : it says that when one has knowledge of the channel statistics but uses a zf equalizer , one should ignore these statistics in choosing a channel estimator for minimizing the zeroth order excess mse . it is straightforward to see that the decision rule given by ( [ eq : mldec ] ) is equivalent to : @xmath19(h)=\arg\min_{\hat{x}(n)\in \mathcal{x}}\left|\frac{y(n)}{h}-\hat{x}(n)\right|^2 \label{eq : mldec1}\ ] ] with a given channel estimate , @xmath3 is replaced by @xmath21 in the last expression . in the case of a perfectly known channel , the division @xmath108 results in an awgn channel with information bearing signal power @xmath63 and noise variance @xmath109 . if only an estimate of the channel , @xmath110 , is available , then the division results in @xmath111 . here , @xmath112 is the channel estimation error , which is gaussian distributed according to our assumptions with @xmath113=h\left({\mbox{\boldmath $ f$}_{}^{h}}{\mbox{\boldmath $ x$}_{\rm tr}^{}}-1\right)$ ] and variance @xmath114 . also , @xmath115=|h|^2\left|{\mbox{\boldmath $ f$}_{}^{h}}{\mbox{\boldmath $ x$}_{\rm tr}^{}}-1\right|^2+\sigma_w^2\|f\|^2 $ ] . for the case of most common constellations and an awgn channel , the error probability is given by @xcite : @xmath116 where @xmath117 and @xmath118 are positive constants depending on the geometry of the constellation . with a channel estimation error , the useful signal power is again @xmath63 . the noise variable is now @xmath119 and therefore @xmath120=0 $ ] . for the power of the noise component , we have : @xmath121=e\left[\left|\frac{w(n)}{\hat{h}}\right|^2\right]+e\left[\left|\frac{\epsilon}{\hat{h}}\right|^2\right]\sigma_x^2\label{eq : noisepbpsk}\ ] ] here , we face again the infinite moment problem . using again similar arguments as in the appendix for approximating @xmath122 $ ] by @xmath123/e[y]$ ] when @xmath124 , we define the corresponding zeroth order version of @xmath125 $ ] : @xmath126\right]_0=\frac{\sigma_w^2}{e[|\hat{h}|^2]}=\frac{\sigma_w^2}{\left[|h|^2\left|{\mbox{\scriptsize $ { \mbox{\boldmath $ f$}}$}}^{h}{\mbox{\scriptsize $ { \mbox{\boldmath $ x$}_{\rm tr}^{}}$}}\right|^2+\sigma_w^2\left\|{\mbox{\scriptsize $ { \mbox{\boldmath $ f$}}$}}\right\|^2\right ] } \\ \left[e\left[\left|\frac{\epsilon}{\hat{h}}\right|^2\right]\right]_0=\frac{e[|\epsilon|^2]}{e[|\hat{h}|^2]}=\frac{|h|^2\left|{\mbox{\scriptsize $ { \mbox{\boldmath $ f$}}$}}^{h}{\mbox{\scriptsize $ { \mbox{\boldmath $ x$}_{\rm tr}^{}}$}}-1\right|^2+\sigma_w^2\left\|{\mbox{\scriptsize $ { \mbox{\boldmath $ f$}}$}}\right\|^2}{|h|^2\left|{\mbox{\scriptsize $ { \mbox{\boldmath $ f$}}$}}^{h}{\mbox{\scriptsize $ { \mbox{\boldmath $ x$}_{\rm tr}^{}}$}}\right|^2+\sigma_w^2\left\|{\mbox{\scriptsize $ { \mbox{\boldmath $ f$}}$}}\right\|^2}\\ \end{array}\right.\label{eq : usefulratios}\end{aligned}\ ] ] a variation of the error probability performance metric for any of the commonly used linear modulation schemes , named _ zeroth order _ error probability , will be given by the following expression : @xmath127_0= & aq\left(b\sqrt{\frac{\left[|h|^2\left|{\mbox{\scriptsize $ { \mbox{\boldmath $ f$}}$}}^{h}{\mbox{\scriptsize $ { \mbox{\boldmath $ x$}_{\rm tr}^{}}$}}\right|^2+\sigma_w^2\left\|{\mbox{\scriptsize $ { \mbox{\boldmath $ f$}}$}}\right\|^2\right]}{\sigma_w^2/(\sigma_x^2)+\left[|h|^2\left|{\mbox{\scriptsize $ { \mbox{\boldmath $ f$}}$}}^{h}{\mbox{\scriptsize $ { \mbox{\boldmath $ x$}_{\rm tr}^{}}$}}-1\right|^2+\sigma_w^2\left\|{\mbox{\scriptsize $ { \mbox{\boldmath $ f$}}$}}\right\|^2\right]}}\right)\nonumber\\ = & aq\left(b\sqrt{\frac{\sigma_x^2}{\left[{\rm mse}_{x}^{dc}({\rm zf})\right]_0}}\right ) \label{eq : pebpsk1}\end{aligned}\ ] ] where we have used the _ zeroth order _ snr approximation given by : @xmath128_0=\frac{\sigma_x^2}{\left[{\rm mse}_{x}^{dc}({\rm zf})\right]_0}.\ ] ] clearly , ( [ eq : pebpsk1 ] ) is an artificial performance metric that appears in this paper for the sake of our arguments . it is used as a variation of the error probability to help us extract useful conclusions . since @xmath129 is a strictly decreasing function , the zeroth order probability of error for a given channel @xmath3 is minimized when @xmath79_0 $ ] is minimized . therefore , the results of subsection [ subsec : dmsedczf ] apply : the mvu estimator is _ not _ an optimal channel estimator for the task of minimizing @xmath130_0 $ ] given the true channel , when using any of the well - known digital modulations in a flat - fading awgn channel.[thrm:6 ] suppose now that we average @xmath130_0 $ ] with respect to _ any _ given channel distribution . we can then make the following statement : the mvu estimator is _ not _ an optimal channel estimator for the task of minimizing the average @xmath130_0 $ ] , when using any of the well - known digital modulations in a flat - fading awgn channel.[thrm:7 ] assume that the pdf of the fading coefficient magnitude is @xmath131 and @xmath132,\alpha,\beta\geq 0 $ ] , @xmath133 possibly equal to @xmath134 . the average @xmath130_0 $ ] is given by the expression : @xmath135}_0= \int_{\alpha}^{\beta } aq\left(b\sqrt{\left[{\rm snr}\right]_0}\right)p(|h|)d|h|\ ] ] assuming that the differentiation and integral operators can be interchanged , we can set the gradient of the above expression with respect to @xmath136 to zero to get the equation : @xmath137}_0= \int_{\alpha}^{\beta } a \nabla_{{\mbox{\scriptsize $ { \mbox{\boldmath $ f$}}$}}^{h}}q\left(b\sqrt{\left[{\rm snr}\right]_0}\right)p(|h|)d|h|={\mbox{\boldmath $ 0$}_{}^{h}}\\ & & { \rm or } \\ & & \int_{\alpha}^{\beta } a\nabla_{x}q\left(x\right)\left|_{x = b\sqrt{\left[{\rm snr}\right]_0}}\right.\frac{b\nabla_{{\mbox{\scriptsize $ { \mbox{\boldmath $ f$}}$}}^{h}}{\left[{\rm snr}\right]_0}}{2b\sqrt{\left[{\rm snr}\right]_0}}p(|h|)d|h|\\ & & = { \mbox{\boldmath $ 0$}_{}^{h}}\end{aligned}\ ] ] where in the second equation we have used the chain rule of differentiation . @xmath129 is strictly decreasing in @xmath138 , thus @xmath139_0}}\right.<0\ ] ] for any value of @xmath3 . also @xmath140 for every value of @xmath3 since it is a distribution function . additionally , @xmath141_0}\geq 0 $ ] for every value of @xmath3 . finally , the numerator of @xmath142_0}$ ] for the mvu estimator is given by ( [ eq : nommvue ] ) multiplied by @xmath143 and by a positive scalar . therefore , it is either positive or negative with respect to @xmath3 in a componentwise fashion depending on the sign of the corresponding element in @xmath144 may be zero but not all of them simultaneously . ] . these arguments verify that @xmath145}_0\neq { \mbox{\boldmath $ 0$}}$ ] . this concludes the proof . if we assume that the prior distribution of @xmath3 is known , then instead of the mvu , one could use the mmse channel estimator . plugging @xmath146 into the negative of ( [ eq : ddczfnom ] ) , one can obtain that @xmath147_0}\right|_{{\mbox{\scriptsize $ { \mbox{\boldmath $ f$}}$}}={\mbox{\scriptsize $ { \mbox{\boldmath $ f$}_{\rm mmse}^{}}$}}}\neq { \mbox{\boldmath $ 0$}}\label{eq : snrnommmse}$ ] . since in the case of the mmse estimator , the assumption is that we always know the prior channel fading distribution , we can make the following statement : the mmse estimator is _ not _ optimal for the task of minimizing @xmath148}_0 $ ] , when using any of the well - known digital modulations in a flat - fading awgn channel . [ thrm:8 ] the result follows along the same lines as in proposition [ thrm:7 ] . the problems of determining the optimal channel estimator for the task of minimizing @xmath130_0 $ ] for a given channel @xmath3 and @xmath148}_0 $ ] was already solved in subsections [ subsec : dmsedczf-1 ] and [ subsec : dmserczf ] , respectively . in the case of @xmath148}_0 $ ] , we can only assess their optimality analytically , using the following argument : we use the upper bound @xmath149 , which becomes tight as @xmath138 increases @xcite . in our case , @xmath150_0}$ ] , and we have already assumed high snr , therefore high @xmath151_0 $ ] , to justify the use of the zf equalizer . using this bound and the relationship between @xmath151_0 $ ] and @xmath152_0 $ ] , we get : @xmath153_0})&<\frac{\sqrt{\left[{\rm mse}_x^{dc}({\rm zf})\right]_0}}{b\sigma_x}\frac{1}{\sqrt{2\pi}}e^{-\frac{b^2\sigma_x^2}{2\left[{\rm mse}_x^{dc}({\rm zf})\right]_0}}\nonumber\\&<\frac{\sqrt{\left[{\rm mse}_x^{dc}({\rm zf})\right]_0}}{b\sigma_x}\frac{1}{\sqrt{2\pi}}\end{aligned}\ ] ] where the last inequality holds for large snr and therefore small @xmath154_0 $ ] . the right hand side function is concave with respect to @xmath154_0>0 $ ] , therefore averaging over any channel distribution , we get : @xmath155_0}\right)\right]&<&\frac{\sqrt{e_{h}\left[\left[{\rm mse}_x^{dc}({\rm zf})\right]_0\right]}}{b\sigma_x\sqrt{2\pi}}\end{aligned}\ ] ] we can use one more time the zeroth order approximation to approximate @xmath156_0\right]$ ] . the right hand side is minimized when this last zeroth order approximation is minimized . thus , the estimators derived in subsections [ subsec : dmsedczf-1 ] and [ subsec : dmserczf ] are optimal for the task of minimizing @xmath148}_0 $ ] , in the sense that they minimize an upper bound to @xmath157_0}\right)\right]$ ] . _ remark _ : although , we have shown that the mvu and mmse estimators are not optimal for the task of minimizing the zeroth order probability of error , we will see in the simulation section that their _ actual _ probability of error performance is almost identical with that of the optimal estimators for the zeroth order probability of error . this is due to two facts : first , the zeroth order probability of error is a variation of the actual probability of error and second , in practice the difference in the channel estimates must be large enough to give rise to a notable difference in the probability of error . nevertheless , we conjecture that such a difference may be more clear in the case of multiple input multiple output ( mimo ) systems if tight approximations of the error probability functions are used to derive the corresponding channel estimators . [ cols="^ " , ] in this section we present numerical results to verify our analysis . in all figures , @xmath158 and qpsk modulation is assumed . the snr during training highlights how good the channel estimate is . the parameter @xmath49 has been empirically selected to be @xmath159 . all schemes in figs . [ fig : zfmse]-[fig : zfexmse_biased ] use ( [ eq : wellbehest ] ) for the same @xmath49 . in figs . [ fig : zfmse]-[fig : zfpe ] , @xmath160 $ ] is chosen to be @xmath161=3 $ ] , i.e. , the real and imaginary parts of @xmath3 are assumed i.i.d . following a uniform distribution in @xmath162 $ ] . in figs . [ fig : zfmse_biased ] and [ fig : zfexmse_biased ] , @xmath160 $ ] equals @xmath163 and @xmath164 , respectively . in fig . [ fig : zfmse_zeroth ] , @xmath165_0 $ ] is presented for @xmath166 and snr during training equal to @xmath167 db . the derived optimal estimators in this paper are better than the mvu and mmse estimators . additionally , the mvu estimator appears to be better than the mmse estimator for this performance metric . this is a new observation contradicting what one would expect and verifying the motivation of this paper . [ fig : zfexmse_zeroth ] presents the corresponding results for @xmath168_0 $ ] . the mvu is the best estimator as proved . this is another example contradicting what one would expect and verifying the motivation of this paper . furthermore , fig . [ fig : zfpe_zeroth ] shows the performance of all schemes in the case of an approximation to the error probability equal to @xmath169_0}\right)$ ] . here , we have assumed that the constants @xmath118 are equal to @xmath170 , since their specific values are irrelevant to the purpose of this simulation plot . the derived estimators in this paper are better than the mvu and mmse estimators as proved in the previous section . the difference of the curves is present in the low snr regime . we now examine the performance of the derived estimators in this paper for the true performance metrics . all the estimators are implemented based on @xmath171 to combat the infinite moment problems . in fig . [ fig : zfmse ] , @xmath172 is presented for @xmath166 and snr during training equal to @xmath167 db . the derived optimal estimators in this paper are better than the mvu and mmse estimators . we can see that the zeroth order approximations in this case are satisfactory even for a low snr during training , in the sense that the corresponding optimal estimators outperform the mvu and mmse estimators for the true performance metric . additionally , the mvu estimator appears to be better than the mmse estimator for this performance metric . this is yet a new snapshot contradicting what one would naturally expect . [ fig : zfexmse ] presents the corresponding results for @xmath173 . the mvu is better than the mmse estimator , coinciding with the analysis based on the zeroth order approximation . note however that the other two estimators appear to be better than the mvu . to obtain a well - behaved @xmath174 in this case , regularization of the same form as in @xmath171 is applied to @xmath3 to avoid values around zero . in this sense , [ fig : zfexmse ] serves more as a proof that the application - oriented estimator selection is valid and less as an actual scenario present in the real world . furthermore , fig . [ fig : zfpe ] shows the performance of all schemes in the case of the error probability performance metric . monte carlo simulations have been used to compute the actual error probability . all schemes coincide because the differences in the channel estimates are not so large to appear in the error probability . nevertheless , these differences may clearly appear in a mimo scenario if tight approximations of the error probability function are used to derive the corresponding channel estimators . finally , figs . [ fig : zfmse_biased ] and [ fig : zfexmse_biased ] demonstrate the validity of the remark @xmath175 in the end of subsection [ subsec : dmsedczf ] . these plots correspond to figs . [ fig : zfmse ] and [ fig : zfexmse ] but with @xmath160=1/2 $ ] and @xmath164 , respectively . they verify that the zeroth order metrics used in this paper are good approximations in terms of indicating the structure of uniformly better estimators than the mvu and mmse . nevertheless , the zeroth order metrics can not really determine the best possible bias with respect to the mvu estimator that the estimators in this paper must have in order to yield the best possible performance against the _ true _ performance metrics . the bias terms are only optimal with respect to the zeroth order metrics . in this paper , application - oriented channel estimator selection has been compared with common channel estimators such as the mvu and mmse estimators . we have shown that the application - oriented selection is the right way to choose estimators in practice . we have verified this observation based on three different performance metrics of interest , namely , the symbol estimate mse , the excess symbol estimate mse and the error probability . this section proposes a simplification of the @xmath176 metric for the estimator given in ( [ eq : wellbehest ] ) with a fixed @xmath49 . due to the gaussianity of @xmath177 , @xmath178 for any @xmath179 ( infinite moment problem ) . using ( [ eq : wellbehest ] ) , the corresponding mean square error becomes : @xmath180_{\rm reg}={\rm pr}\left\{|{\mbox{\boldmath $ f$}_{}^{h}}{\mbox{\boldmath $ y$}_{\rm tr}^{}}|>\lambda\right\}\cdot\nonumber\\ & & e\left[\sigma_x^2\left|1-\frac{h}{{\mbox{\boldmath $ f$}_{}^{h}}{\mbox{\boldmath $ y$}_{\rm tr}^{}}}\right|^2+\frac{\sigma_w^2}{|{\mbox{\boldmath $ f$}_{}^{h}}{\mbox{\boldmath $ y$}_{\rm tr}^{}}|^2 } ; |{\mbox{\boldmath $ f$}_{}^{h}}{\mbox{\boldmath $ y$}_{\rm tr}^{}}|>\lambda\right]\nonumber\\ & & + { \rm pr}\left\{|{\mbox{\boldmath $ f$}_{}^{h}}{\mbox{\boldmath $ y$}_{\rm tr}^{}}|\leq \lambda\right\ } \cdot\nonumber\\ & & e\left[\frac{\sigma_x^2}{\lambda^2}\left|\lambda \frac{{\mbox{\boldmath $ f$}_{}^{h}}{\mbox{\boldmath $ y$}_{\rm tr}^{}}}{|{\mbox{\boldmath $ f$}_{}^{h}}{\mbox{\boldmath $ y$}_{\rm tr}^{}}|}-h\right|^2+\frac{\sigma_w^2}{\lambda^2 } ; where @xmath181 denotes conditioning and `` reg '' signifies the use of the regularized channel estimator in ( [ eq : wellbehest ] ) . to simplify this expression , we observe that @xmath182 , since by the mean value theorem this probability is equal to the area of the region @xmath183 , which is of order @xmath184 , multiplied by some value of the probability density function of @xmath185 in that region , which is of order @xmath186 . in addition , @xmath187=\frac{\sigma_x^2}{\lambda^2}|h|^2\nonumber\\ & & + \frac{\sigma_w^2}{\lambda^2}-2\frac{\sigma_x^2}{\lambda}\re\left\ { h^{*}\frac{{\mbox{\boldmath $ f$}_{}^{h}}{\mbox{\boldmath $ y$}_{\rm tr}^{}}}{|{\mbox{\boldmath $ f$}_{}^{h}}{\mbox{\boldmath $ y$}_{\rm tr}^{}}|}\right\}\nonumber\end{aligned}\ ] ] if in addition the snr during training is sufficiently high and the probability mass of @xmath185 is concentrated around @xmath45 , then it can be shown that @xmath188\approx \nonumber\\ & & \frac{\sigma_x^2e[|{\mbox{\boldmath $ f$}_{}^{h}}{\mbox{\boldmath $ y$}_{\rm tr}^{}}-h|^2 ; |{\mbox{\boldmath $ f$}_{}^{h}}{\mbox{\boldmath $ y$}_{\rm tr}^{}}|>\lambda]+\sigma_w^2}{e[|{\mbox{\boldmath $ f$}_{}^{h}}{\mbox{\boldmath $ y$}_{\rm tr}^{}}|^2 ; |{\mbox{\boldmath $ f$}_{}^{h}}{\mbox{\boldmath $ y$}_{\rm tr}^{}}|>\lambda]}.\end{aligned}\ ] ] the same holds even if @xmath52 is a biased estimator of @xmath3 at high training snr and @xmath185 tends to concentrate around a value @xmath189 bounded away from @xmath45 ( and of course from @xmath167 ) . to show the last claim , we set @xmath190 and @xmath191 . since @xmath192 , it also holds that @xmath193>\lambda^2 $ ] . furthermore , it can be seen that @xmath194-\frac{e[x]}{e[y]}\right|\leq \frac{1}{\lambda^4}e\left[\left|xe[y]-ye[x]\right|\right].\label{eq : stochineq}\end{aligned}\ ] ] at high training snr , @xmath195 $ ] and @xmath196 $ ] in the mean square sense and therefore it can be easily shown that the right hand side of ( [ eq : stochineq ] ) converges to @xmath167 . to see this , notice that the cauchy - schwarz inequality yields @xmath197-ye[x]\right|\right]\leq \frac{1}{\lambda^4}\left(e\left[\left|xe[y]-ye[x]\right|^2\right]\right)^{1/2}\nonumber\\ & = \frac{1}{\lambda^4}\left(e^2[y]e[x^2]+e[y^2]e^2[x]-2e[xy]e[x]e[y]\right)^{1/2}.\end{aligned}\ ] ] since @xmath195 $ ] and @xmath196 $ ] in the mean square sense , @xmath198\rightarrow e^2[x]$ ] , @xmath199\rightarrow e^2[y]$ ] and @xmath200\rightarrow e[x]e[y]$ ] . for the last case , notice that @xmath201-e[x]e[y]\right|&=\left|e\left[(x - e[x])(y - e[y])\right]\right|\nonumber\\&\leq e\left[\left|x - e[x]\right|\left|y - e[y]\right|\right]\nonumber\\ & \leq \sqrt{e\left[\left|x - e[x]\right|^2\right]e\left[\left|y - e[y]\right|^2\right]},\end{aligned}\ ] ] where the last inequality follows again from the cauchy - schwarz inequality . by the mean square convergence of @xmath202 to @xmath123 $ ] and @xmath203 to @xmath204 $ ] the right hand side of ( [ eq : cauchyineq2 ] ) tends to @xmath167 . therefore , the right hand side of ( [ eq : cauchyineq1 ] ) tends to @xmath167 . furthermore , under the high snr assumption the conditional expectations can be approximated by their unconditional ones , since for a sufficiently small @xmath49 their difference is due to an event of probability @xmath184 . therefore , @xmath188\approx\nonumber\\ & & \left\{\frac{\sigma_x^2e[|{\mbox{\boldmath $ f$}_{}^{h}}{\mbox{\boldmath $ y$}_{\rm tr}^{}}-h|^2]+\sigma_w^2}{e[|{\mbox{\boldmath $ f$}_{}^{h}}{\mbox{\boldmath $ y$}_{\rm tr}^{}}|^2]}\right\}+o(\lambda^2).\end{aligned}\ ] ] combining all the above results yields @xmath205+\sigma_w^2}{e[|{\mbox{\boldmath $ f$}_{}^{h}}{\mbox{\boldmath $ y$}_{\rm tr}^{}}|^2]}\right\}+o(1).\ ] ] the @xmath186 term is not negligible but for sufficiently small @xmath49 its dependence on @xmath17 is insignificant . hence , for a sufficiently small @xmath49 and a sufficiently high snr during training , minimizing @xmath55 is equivalent to minimizing the following approximation @xmath56_0=\frac{e\left[|\hat{h}-h|^2\right]}{e\left[|\hat{h}|^2\right]}\sigma_x^2+\sigma_w^2\frac{1}{e\left[|\hat{h}|^2\right]}.\label{eq : mseddczf0_app}\ ] ] 99 f. f. abari , f. k. sharifabad , o. edfors , `` low complexity channel estimation for lte in fast fading environments for implementation on multi - standard platforms '' , _ proc . vtc fall 2010 _ , ottawa , canada , sept . 69 , 2010 . j. g. andrews , a. ghosh , and r. muhamed , _ fundamentals of wimax : understanding broadband wireless networking , _ prentice - hall , 2007 . x. bombois , g. scorletti , m. gevers , p. m. j. van den hof , r. hildebrand , least costly identification experiment for control , " _ automatica _ , vol . 42 , no . 10 , pp . 16511662 , 2006 . chen , w - h . he , h - s . chen , y. lee , `` mode detection , synchronization , and channel estimation for dvb - t ofdm receiver '' , _ proc . globecom 2003 _ , san francisco , usa , december 15 , 2003 . j. m. cioffi , _ course readers for ee379a , ee379b , ee379c , ee479 _ , available at : http://www.stanford.edu / group / cioffi/. y. c. eldar , _ rethinking biased estimation : improving maximum likelihood and the cramer - rao bound _ , foundations and trends in signal processing , vol . 1 4 , pp . 305449 , 2008 . j. fan , g. ye li , q. yin , b. peng , x. zhu , `` joint user pairing and resource allocation for lte uplink transmission '' , _ ieee trans . on wireless communications _ , vol . 11 , no . 8 , pp . 28382847 , aug . o. n. gharehshiran , a. attar , v. krishnamurthy , `` collaborative sub - channel allocation in cognitive lte femto - cells : a cooperative game - theoretic approach '' , _ ieee trans . on communications _ , accepted for publication , doi : 10.1109/tcomm.2012.100312.110480 . m. k. hati , t. k. bhattacharyya , `` digital video broadcast services to handheld devices and a simplified dvb - h receiver subsystem '' , _ proc . ncc 2012 _ , indian institute of technology kharagpur , india , february 35 , 2012 . h. hjalmarsson , system identification of complex and structured systems , " _ plenary address european control conference / european journal of control _ , vol . 15 , no . 4 , pp . 275310 , 2009 . s. hu , j. xie , f. yang , `` improved pilot - aided channel estimation in lte uplink '' , _ proc . iccp 2011 _ , chengdu , china , october 2123 , 2011 . j. huber , _ robust statistics _ , john wiley @xmath206 sons , 2005 . a. goldsmith , _ wireless communications _ , cambridge university press , 2005 . h. jansson , h. hjalmarsson , input design via lmis admitting frequency - wise model specifications in confidence regions , " _ ieee trans . 50 , no . 10 , pp . 15341549 , 2005 . d. katselis , c. r. rojas , h. hjalmarsson , m.bengtsson , application - oriented finite sample experiment design : a semidefinite relaxation approach , " _ proc . sysid-2012 , _ brussels , belgium , july 2012 . s. m. kay , _ fundamentals of statistical signal processing , volume i : estimation theory _ , prentice hall , 1993 . s. m. kay , _ fundamentals of statistical signal processing , volume ii : detection theory _ , prentice hall , 1998 . d. lee , m. choi , s. choi , `` channel estimation and interference cancellation of feedback interference for docr in dvb - t system '' , _ ieee trans . on broadcasting _ , 1 , pp . 8797 , march 2012 . j. c. lee , d. s. han , s. park , `` channel estimation based on path separation for dvb - t in long delay situations '' , _ ieee trans . on consumer electronics _ , vol . 2 , pp . 316321 , may 2009 . y. liu , s. sezginer , `` iterative compensated mmse channel estimation in lte systems '' , _ proc . icc 2012 _ , ottawa , canada , june 1015 , 2012 . d. lpez - prez , x. chu and j. zhang , `` dynamic downlink frequency and power allocation in ofdma cellular networks '' , _ ieee trans . on communications _ , vol . 29042914 , oct . j. g. proakis , _ digital communications _ , 3rd edition , mcgraw - hill , 1995 . d. schafhuber , g. matz , f. hlawatsch , p. loubaton , `` mmse estimation of time - varying channels for dvb - t systems with strong co - channel interference '' , _ proc . eusipco 2002 _ , toulouse , france , sept . p. siebert , `` dvb : developing global television standards for today and tomorrow '' , _ proc . itu wt 2011 _ , geneva , switzerland , 2427 october , 2011 . i. l. j. da silva , a. l. f. de almeida , f. r. p. cavalcanti , r. baldemair , s. falahati , `` improved data - aided channel estimation in lte pucch using a tensor modeling approach '' , _ proc . icc 2010 _ , cape town , south africa , may 2327 , 2010 . i. siomina , d. yuan , `` analysis of cell load coupling for lte network planning and optimization '' , _ ieee trans . on wireless communications _ , 11 , no . 6 , pp . 22872297 , june 2012 . l. yang , g. ren , b. yang , z. qiu , `` fast time - varying channel estimation technique for lte uplink in hst environment '' , _ ieee trans . on vehicular technology _ , 9 , pp . 40094019 , nov . 2012 .

given an iterated function system ( ifs ) of contractive similitudes , the theory of gromov hyperbolic graph on the ifs has been established recently . in the paper , we introduce a notion of simple augmented tree which is a gromov hyperbolic graph . by generalizing a combinatorial device of rearrangeable matrix , we show that there exists a near - isometry between the simple augmented tree and the symbolic space of the ifs , so that their hyperbolic boundaries are lipschitz equivalent . we then apply this to consider the lipschitz equivalence of self - similar sets with or without assuming the open set condition . moreover , we also provide a criterion for a self - similar set to be a cantor - type set which completely answers an open question raised in . our study extends the previous works .

1612.00209

a self - similar set @xmath0 is defined as the limit set of an iterated function system ( ifs ) of contractive similitudes , and the iteration is addressed by a symbolic space of finite words which forms a tree . the topological boundary of the tree is a cantor - type set . however the tree does not capture all the geometric and analytic properties of @xmath0 . by incorporating more information of @xmath0 onto the tree , kaimanovich @xcite first explored the concept of `` augmented tree '' on the sierpinski gasket such that it is a hyperbolic graph in the sense of gromov ( @xcite@xcite ) and its hyperbolic boundary is homeomorphic to the gasket . lau and wang ( @xcite@xcite ) extended this idea on more general self - similar sets satisfying the open set condition ( osc ) or the weak separation condition . recently they @xcite completed the previous studies by removing some superfluous conditions , and obtained that for any ifs , the augmented tree is always a hyperbolic graph . moreover , the hyperbolic boundary is hlder equivalent to the @xmath0 . the setup of augmented trees connects fractal geometry , graph theory and markov chains , it has been frequently used to study the random walks on self - similar sets and the induced dirichlet forms ( @xcite@xcite@xcite@xcite@xcite ) . on the other hand , the author and his coworkers made a first attempt to apply augmented trees to the study on lipschitz equivalence of self - similar sets . in a series of papers ( @xcite@xcite@xcite ) , we investigated in detail the lipschitz equivalence of totally disconnected self - similar sets with equal contraction ratio and their hyperbolic boundaries . recall that two metric spaces @xmath1 and @xmath2 are said to be _ lipschitz equivalent _ , write @xmath3 , if there exists a bi - lipschitz map @xmath4 , i.e. , @xmath5 is a bijection and there exists a constant @xmath6 such that @xmath7 the lipschitz equivalence of cantor sets was first considered in @xcite and @xcite . for its extension on self - similar sets , it has been undergoing rapid development recently ( @xcite@xcite@xcite@xcite@xcite@xcite@xcite@xcite@xcite@xcite@xcite@xcite ) . however , most of the studies are based on the nice geometric structure of self - similar sets such as cantor sets or totally disconnected self - similar sets with osc . there are few results on the non - totally disconnected self - similar sets ( @xcite@xcite ) or self - similar sets without osc . in this paper , we unify our previous approaches on augmented trees and consider the lipschitz equivalence of more general self - similar sets which allow non - equal contraction ratios and substantial overlaps . let @xmath8 be an ifs on @xmath9 where @xmath10 has a contraction ratio @xmath11 , let @xmath0 be the self - similar set of the ifs satisfying @xmath12 . let @xmath13 and @xmath14 be the symbolic space ( by convention , @xmath15 ) . for @xmath16 , we denote by @xmath17 , and @xmath18 . let @xmath19 . we also define a new symbolic space @xmath20 in the special case that all @xmath21 are equal , @xmath22 . in general , @xmath23 is a proper subset of @xmath24 . if @xmath25 , we denote the length by @xmath26 , and say that @xmath27 lies in level @xmath28 . the @xmath23 has a natural tree structure according to the standard concatenation of finite words , we denote the edge set by @xmath29 ( @xmath30 for vertical ) . we also define a horizontal edge for a pair @xmath31 in @xmath32 if @xmath33 and @xmath34 where @xmath35 is a fixed constant , and denote this set of edges by @xmath36 ( @xmath37 for horizontal ) . let @xmath38 , then the graph @xmath39 is an augmented tree in the sense of kaimanovich ( @xcite@xcite ) . lau and wang @xcite already showed that the augmented tree @xmath39 is a gromov hyperbolic graph and its hyperbolic boundary @xmath40 under a visual metric is hlder equivalent to the @xmath0 . based on this , our approach to the lipschitz equivalence of self - similar sets is to lift the consideration to the augmented tree @xmath41 . we define a _ horizontal component _ of @xmath23 to be the maximal connected horizontal subgraph @xmath42 in some level with respect to @xmath36 . let @xmath43 be the set of all horizontal components of @xmath23 . for @xmath44 , we use @xmath45 to denote the union of @xmath42 and its all descendants , with the subgraph structure inherited from @xmath39 . we say that @xmath46 are equivalent if @xmath45 and @xmath47 are graph isomorphic . we call @xmath39 _ simple _ if there are only finitely many equivalence classes in @xmath43 . similarly , we can define a simple tree for @xmath48 if there are only finitely many equivalence classes of vertices . obviously , if @xmath39 is simple then @xmath49 is simple . we use @xmath50 ( resp . @xmath51 ) to denote the incidence matrix of @xmath39 ( resp . @xmath49 ) , which encodes graph relation of the equivalence classes . let a vector @xmath52 represent the finite classes in @xmath53 by means of the classes of vertices . ( see definition [ def of simple tree ] and remark [ b - u - remark ] . ) the following is our first main result [ th - main ] suppose an augmented tree @xmath54 is simple , and suppose the incidence matrix @xmath50 is @xmath55-rearrangeable . then there exists a near - isometry between the augmented tree @xmath54 and the tree @xmath56 , so that @xmath57 . the concept of rearrangeable matrix was initiated by the author ( @xcite@xcite@xcite ) in studying the lipschitz equivalence . a lot of nonnegative integer matrices including primitive matrices are rearrangeable . it is a combinatorial device to arrange the vertices and edges of the augmented tree properly to construct the near - isometry , which is stronger than the rough isometry in literature . in the paper , we extend the original definition of rearrangeable matrix so as to solve more general situations ( please see definition [ def rearrangeable ] ) . for an ifs with substantial overlaps ( i.e. , without the osc ) , it may happen that @xmath58 for @xmath59 . in this case , the augmented tree @xmath39 may be no longer simple . but we can modify the augmented tree by identifying @xmath60 for @xmath61 and @xmath58 , and let @xmath62 denote the quotient space with the induced graph . moreover , we can further reduce the graph @xmath63 into a tree @xmath64 ( see section 2 ) . then following the same proof of theorem [ th - main ] , we have [ th - quotient ] suppose the quotient space @xmath65 of an augmented tree is simple , and suppose the incidence matrix @xmath50 is @xmath55-rearrangeable . then there exists a near - isometry between the quotient space @xmath65 and the reduced tree @xmath66 , so that @xmath67 . an ifs is called _ strongly separated _ @xcite if @xmath68 for @xmath69 , in this case the @xmath0 is called _ dust - like _ ( @xcite@xcite ) . by reducing the lipschitz equivalence on the augmented trees in theorems [ th - main ] and [ th - quotient ] to the self - similar sets , we have our second main result [ th - main2 ] under the assumption of theorem [ th - main ] , the associated self - similar set @xmath0 is lipschitz equivalent to a dusk - like self - similar set with the same contraction ratios as for @xmath0 . under the assumption of theorem [ th - quotient ] , the associated self - similar set @xmath0 is lipschitz equivalent to a cantor - type set ( may be not a dust - like self - similar set ) . the first part of theorem [ th - main2 ] is applied for the self - similar sets satisfying the osc ; while the second part is applied for the self - similar sets satisfying the weak separation condition ( wsc ) @xcite . the wsc contains many important overlapping cases , it has been studied extensively in connection with the multifractal structure of self - similar measures ( see @xcite@xcite@xcite and the references therein ) . finally , we provide a criterion for the hyperbolic boundary ( or the associated self - similar set ) to be totally disconnected ( i.e. , a cantor - type set ) from the perspective of augmented trees which also answers an open question raised in @xcite . [ th - main3 ] let @xmath54 be an augmented tree of bounded degree . then the hyperbolic boundary @xmath40 ( or the self - similar set @xmath0 ) is totally disconnected if and only if the sizes of horizontal components are uniformly bounded . the theorem is still valid if we replace @xmath23 by its quotient space @xmath70 . the paper is organized as follows : in section 2 , we recall basic results on hyperbolic graphs and introduce the notion of simple augmented tree ; moreover , we state theorem [ th - main ] and prove theorem [ th - quotient ] there . in section 3 , we generalize the concept of rearrangeable matrix and prove theorem [ th - main ] . in section 4 , we prove theorem [ th - main2 ] and give two examples to illustrate our results . theorem [ th - main3 ] will be showed in section 5 . we include some remarks and open questions in section 6 . in this paper , an infinite connected graph @xmath41 will be considered , where @xmath23 is a set of vertices which is countably infinite , @xmath71 is a set of edges which is a symmetric subset of @xmath72 . by a _ path _ in @xmath41 from @xmath27 to @xmath73 , we mean a finite sequence @xmath74 such that @xmath75 . we call @xmath23 a _ tree _ if the path between any two distinct vertices is unique . we equip a graph @xmath23 with an integer - valued metric @xmath76 , which is the length of a geodesic , denoted by @xmath77 , from @xmath27 to @xmath73 . let @xmath78 be a reference point in @xmath23 and call it the _ root _ of the graph . we use @xmath79 to denote @xmath80 and say that @xmath27 belongs to the @xmath28-th level of the graph if @xmath81 . recall that a graph @xmath23 is called _ hyperbolic _ ( with respect to @xmath82 ) if there is @xmath83 such that @xmath84 where @xmath85 is the _ gromov product _ ( @xcite@xcite ) . we choose @xmath86 , and define for @xmath87 and a root @xmath82 @xmath88 this is not necessarily a metric unless @xmath23 is a tree . however @xmath89 is equivalent to a metric as long as @xmath90 is small enough ( i.e. , @xmath91 @xcite ) . hence we regard @xmath89 as a _ visual metric _ for convenience . the _ hyperbolic boundary _ of @xmath23 is defined by @xmath92 , where @xmath93 is the @xmath89-completion of @xmath23 . it is often useful to identify @xmath94 with the _ geodesic rays _ in @xmath23 that converge to @xmath95 , i.e. , an infinite path @xmath96 $ ] such that @xmath97 and any finite segment of the path is a geodesic . it is known that two geodesic rays @xmath96 , \pi[y_1,y_2,\dots]$ ] are equivalent as cauchy sequence if and only if @xmath98 for all but finitely many @xmath28 , where @xmath99 is independent of the rays @xcite . both gromov product and the visual metric @xmath89 can be extended onto the completion @xmath93 @xcite , and the hyperbolic boundary @xmath40 is a compact subset of @xmath93 . we mention that a tree is always a hyperbolic graph with @xmath100 and its hyperbolic boundary is a cantor - type set . throughout the paper , we let @xmath8 be an ifs of contractive similitudes on @xmath9 with contraction ratios @xmath101 . then there exists a unique non - empty compact subset @xmath0 of @xmath9 @xcite such that @xmath102 we call such @xmath0 a _ self - similar set _ , and call @xmath0 _ dust - like _ if it satisfies @xmath103 for @xmath104 . a dust - like self - similar set must be totally disconnected , but the converse is not true . let @xmath13 and let @xmath14 be the symbolic space representing the ifs ( by convention , @xmath15 ) . for @xmath105 , we denote by @xmath17 , @xmath106 and @xmath18 . let @xmath107 , we define a new symbolic space @xmath20 in general @xmath23 is a proper subset of @xmath24 . if @xmath108 , we denote the length by @xmath26 , and say that @xmath27 is in level @xmath28 ( here @xmath79 is not the real length of the word @xmath27 in @xmath24 ) . note that for each @xmath109 there exists a unique @xmath110 and @xmath111 such that @xmath112 in the standard concatenation . this defines a natural tree structure on @xmath23 : @xmath113 are said to be joined by a vertical edge denoted by @xmath31 ( or @xmath114 ) . let @xmath29 be the set of all vertical edges , then @xmath48 is a tree with root @xmath115 . for @xmath116 , let @xmath117 denote @xmath118 such that @xmath119 and @xmath120 , we call such @xmath27 an ancestor of @xmath73 . define inductively @xmath121 to be the @xmath28-th generation ancestor of @xmath73 . given a tree @xmath48 , we can add more ( horizontal ) edges according to @xcite : @xmath122 trivially , @xmath36 is a symmetric subset of @xmath32 and satisfies : @xmath123 let @xmath124 , then the graph @xmath125 is called an _ augmented tree _ in the sense of kaimanovich @xcite . the above horizontal edge set @xmath36 is modified from the previous works ( @xcite@xcite@xcite ) where the last condition was @xmath126 ( i.e. , @xmath127 ) . the advantage of this new definition of @xmath36 is that it allows us to eliminate some superfluous conditions added before . we say that a geodesic @xmath77 is a _ horizontal geodesic _ if it consists of horizontal edges only ; it is called a _ canonical geodesic _ if there exist @xmath128 such that @xmath129 with @xmath130 a horizontal path and @xmath131 vertical paths . moreover , the horizontal part of the canonical geodesic is required to be on the highest level ( i.e. , for any geodesic @xmath132 , @xmath133 ) . then the gromov product of an augmented tree can be expressed by latexmath:[\[\label{eq.cano.geo.identity } level and the length of the horizontal part of the canonical geodesic @xmath77 . [ th2.3 ] an augmented tree @xmath54 is hyperbolic if and only if there exists a constant @xmath135 such that any horizontal geodesic is bounded by @xmath136 . by using the above criterion for hyperbolicity , lau and wang @xcite showed that the augmented tree @xmath39 is always hyperbolic . on an augmented tree , for @xmath137 , we use @xmath45 to denote the union of @xmath42 and its all descendants , that is , @xmath138 where @xmath139 is the initial part of @xmath27 with length @xmath28 . if @xmath42 is connected , then @xmath45 , equipped with the edge set @xmath71 restricted on @xmath45 , is a subgraph of @xmath23 . moreover , if @xmath125 is hyperbolic , then @xmath45 is also hyperbolic . we say that @xmath42 is an _ @xmath140-horizontal component _ if @xmath137 is a maximal connected subset with respect to @xmath141 . we let @xmath142 denote the family of all @xmath140-horizontal components , and let @xmath143 . for @xmath144 , we say that @xmath42 and @xmath145 are _ equivalent _ , denote by @xmath146 , if there exists a graph isomorphism @xmath147 , i.e. , the map @xmath148 and the inverse map @xmath149 preserve the vertical and horizontal edges of @xmath150 and @xmath151 . we denote by @xmath152 $ ] the equivalence class containing @xmath42 . [ @xcite ] [ def of simple tree ] we call an augmented tree @xmath39 _ simple _ if the equivalence classes in @xmath153 is finite , i.e. , @xmath154 is finite . let @xmath155,\dots,[t_m]$ ] be the equivalence classes , and let @xmath156 denote the cardinality of the horizontal components of offspring of @xmath157 $ ] that belong to @xmath158 $ ] . we call @xmath159_{m\times m}$ ] the incidence matrix of @xmath41 . [ b - u - remark ] similarly , for the tree @xmath49 , we say that two vertices @xmath113 of @xmath23 are _ equivalent _ , denote by @xmath160 , if the two subtrees @xmath161 and @xmath162 ( defined as @xmath45 above ) are graph isomorphic . we call the tree @xmath49 _ simple _ if @xmath23 has only finitely many non - equivalent vertices , say @xmath163,\dots , [ \bt_n]$ ] ( for convenience we always assume the root @xmath164 $ ] ) . in this case , we denote the incidence matrix by @xmath165_{n\times n}$ ] , where @xmath166 means the cardinality of offspring of a vertex @xmath167 $ ] that belong to @xmath168 $ ] . obviously , if an augmented tree @xmath39 is simple then the associated tree @xmath49 is also simple . therefore any horizontal component @xmath42 can be represented by the equivalence classes of vertices @xmath163,\dots , [ \bt_n]$ ] . more precisely , for @xmath169 $ ] where @xmath170 , suppose @xmath42 consists of @xmath171 vertices belonging to @xmath168 , j=1,\dots , n$ ] , then we represent it by using a vector @xmath172 $ ] ( relative to @xmath163,\dots , [ \bt_n]$ ] ) , and let @xmath173 $ ] . then @xmath52 is a representation of the classes @xmath155,\dots,[t_m]$ ] with respect to the classes @xmath163,\dots , [ \bt_n]$ ] . [ th.simple tree ] let @xmath49 and @xmath174 be two simple trees defined as above . if they have the same incidence matrix @xmath51 , then they are graph isomorphic . let @xmath175 where @xmath176 are their roots . let @xmath165_{n\times n}$ ] be the incidence matrix associated to the equivalence classes of vertices @xmath163,\dots,[\bt_n]$ ] of @xmath23 ( or the equivalence classes of vertices @xmath177,\dots,[\bt_n']$ ] of @xmath178 respectively ) . write @xmath179_{n\times n}$ ] for @xmath180 . since @xmath164 $ ] , by the definition of incidence matrix , @xmath181 consists of @xmath182 vertices belonging to @xmath168 $ ] where @xmath183 . similarly , @xmath184 consists of @xmath182 vertices belonging to @xmath185 $ ] where @xmath183 as @xmath186 $ ] . hence we can define a bijective map @xmath187 such that if @xmath188 , then @xmath189 ; and if @xmath190 $ ] then @xmath191 $ ] . that @xmath5 is the desired graph isomorphism . let @xmath192 be two connected graphs . we say @xmath193 is a near - isometry if it is a bijection and there exists @xmath194 such that for any @xmath87 , @xmath195 let @xmath23 , @xmath178 be two hyperbolic graphs that are equipped with the visual metrics with the same parameter @xmath90 ( as in ( [ eq2.1 ] ) ) . suppose there exists a near - isometry @xmath196 , then @xmath197 . now we state our main theorem as below . [ th - main - part1 ] suppose an augmented tree @xmath54 is simple , and suppose the incidence matrix @xmath50 is @xmath55-rearrangeable ( where @xmath51 and @xmath52 are defined as in remark [ b - u - remark ] ) . then there exists a near - isometry between the augmented tree @xmath54 and the tree @xmath56 , so that @xmath57 . the concept of _ rearrangeable matrix _ is a generalization of the one in @xcite , which is a crucial device to construct the near - isometry . since the definition is a little complicated , we will introduce it in detail and prove the theorem in the next section . recall that a graph @xmath41 is of _ bounded degree _ if @xmath198 where @xmath199 is the total number of edges connecting @xmath27 . lau and wang @xcite proved that _ the augmented tree @xmath54 is of bounded degree if and only if the ifs satisfies the osc . _ if the augmented tree @xmath41 is simple then it is of bounded degree , hence the ifs satisfies the osc . however , for an overlapping ifs that does not satisfy the osc , it may happen that @xmath58 for @xmath200 . then the augmented tree may be no longer simple . in this situation , we need to modify the augmented tree by identifying @xmath113 for @xmath61 and @xmath58 , and define a quotient space @xmath70 of @xmath23 by the equivalence relation . let @xmath201 be the sets of vertical edges and horizontal edges inherited from the original ones and let @xmath62 denote the quotient space with the induced graph . moreover , it was also proved in @xcite that _ the graph @xmath65 is of bounded degree if and only if the ifs satisfies the weak separation condition ( wsc)_. the concept of the wsc was first introduced in @xcite to study multifractal measures . it is well - known that the wsc is strictly weaker than the osc and the finite type condition @xcite . we refer to a survey paper @xcite for the detailed discussion on various separation conditions in fractal geometry . under the quotient space @xmath62 , we can still define the equivalence classes of horizontal components and vertices and the concept of `` simple '' as above . but it should be more careful that we can not reduce the quotient space @xmath62 into a tree by only removing the horizontal edges , that is , @xmath63 will not be a tree . in that situation , one vertex of @xmath70 may have multi parents , that destroys the tree structure . however , we can further reduce the graph @xmath63 into a real tree @xmath64 in the following way : for a vertex @xmath202 , let @xmath203 be all the parents of @xmath73 such that @xmath204 and @xmath205 where @xmath206 . suppose @xmath207 in the lexicographical order for the finite words , then we keep only the edge @xmath208 and remove all other vertical edges . we denote by @xmath209 the set of remained edges . obviously @xmath64 indeed is a tree . moreover , if the quotient space @xmath62 is simple then the reduced tree @xmath64 is also simple . [ example - quotienttree ] let @xmath210 be an ifs . as @xmath211 , the ifs satisfies the wsc , but not the osc ( see figure [ quotienttree ] ( a ) ) . hence the augmented tree @xmath39 is not simple . but we can consider the quotient space @xmath62 by identifying vertices , then @xmath212 , as an equivalence class , is a vertex in @xmath70 . let @xmath213 and @xmath214 be the horizontal components of the quotient space @xmath62 . we shall show that there are only three equivalence classes : @xmath155 , [ t_2 ] , [ t_3]$ ] by lemma 5.1 of @xcite , hence @xmath62 is simple , and the incidence matrix is @xmath215.\ ] ] indeed , we can find out all the equivalence classes in the first three iterations as follows : in the first iteration , @xmath216 ; in the second iteration , @xmath217 and @xmath218 are all equivalent to @xmath219 ; in the third iteration , @xmath220 , @xmath221 are equivalent to @xmath219 , @xmath222 , @xmath223 are equivalent to @xmath224 . on the other hand , there are two different paths @xmath225 $ ] and @xmath226 $ ] joining @xmath82 and @xmath27 in @xmath70 , hence @xmath63 is not a tree ( see figure [ quotienttree ] ( c ) ) . however , we can reduce @xmath63 into a tree @xmath64 by using the method in the previous paragraph , for example removing the vertical edge @xmath227 as @xmath228 ( see figure [ quotienttree ] ( d ) ) . it is not hard to check that the equivalence classes of vertices of @xmath64 are @xmath229 , [ \{2\}]$ ] . therefore @xmath64 is simple and the incidence matrix is @xmath230.\ ] ] [ h ] + [ th - quotient - tree ] suppose the quotient space @xmath231 is simple , and suppose the incidence matrix @xmath50 is @xmath55-rearrangeable ( where @xmath51 and @xmath52 are defined as in remark [ b - u - remark ] ) . then there exists a near - isometry between @xmath65 and the reduced tree @xmath64 , so that @xmath232 . let @xmath233 and @xmath234 . trivially , the identity map between @xmath235 is a near - isometry . the same proof of theorem [ th - main ] can be applied to build a near - isometry from @xmath236 to @xmath237 . by composing the two near - isometries , we obtain the desired one . in order to define a desired map from an augmented tree to the original tree , we need a combinatorial device to rearrange the vertices and edges of the graphs in a proper way . in a series of papers ( @xcite@xcite@xcite ) , the author developed a concept of rearrangeable matrix to handle this case . in the present paper , we generalize the definition of rearrangeable matrix so that it helps to solve more general problems . for a vector @xmath173\in { \mathbb r}^{nm}$ ] with blocks @xmath172\in { \mathbb r}^n$ ] , we denote by @xmath238 $ ] . we also let @xmath239 $ ] for convenience such that the involved matrix product is well - defined . [ def rearrangeable ] let @xmath51 be an @xmath240 nonnegative integer matrix , let @xmath173 $ ] where @xmath172\in { \mathbb z}^n_+$ ] . an @xmath241 matrix @xmath50 is said to be @xmath55-rearrangeable if for each row vector @xmath242 of @xmath50 , there exists a nonnegative matrix @xmath243_{p\times m}$ ] ( rearranging matrix ) where @xmath244 such that @xmath245:=\left[\begin{array}{c } { \mathbf c}_1 \\ \vdots \\ { \mathbf c}_p \end{array } \right]\ ] ] and @xmath246 , where @xmath247 is the @xmath248-th row vector of @xmath51 . note that when the matrix @xmath51 degenerates to an integer , the above definition coincides exactly with the original one in @xcite ( or @xcite ) . [ exa1 ] let @xmath249 $ ] where @xmath250 , \bu_2=[2,1],\bu_3=[3,1]$ ] . let @xmath251,\quad b=\left[\begin{array}{cc } 3 & 1\\ 2 & 1 \end{array } \right].\ ] ] then @xmath50 is @xmath55-rearrangeable ; @xmath252 is @xmath253-rearrangeable . in fact , let @xmath242 be the @xmath254-th row vector of @xmath50 and @xmath255 the @xmath254-th row vector of @xmath51 . for @xmath256 , there exists a @xmath257 matrix @xmath258 $ ] such that @xmath259 and @xmath260\left[\begin{array}{c } \bu_1 \\ \bu_2 \\ \bu_3 \end{array } \right]=\bu_1+\bu_2=\bb_1\ ] ] which is the first row vector of @xmath51 . for @xmath261 , there exists a @xmath262 matrix @xmath263 $ ] such that @xmath264 and @xmath265\left[\begin{array}{c } \bu_1 \\ \bu_2 \\ \bu_3 \end{array } \right]=\left[\begin{array}{c } \bu_1+\bu_2 \\ \\ \bu_3 \end{array } \right]=\left[\begin{array}{c } \bb_1 \\ \bb_2 \\ \bb_1 \end{array } \right].\ ] ] for @xmath266 , there exists a @xmath267 matrix @xmath268 $ ] such that @xmath269 and @xmath270\left[\begin{array}{c } \bu_1 \\ \bu_2 \\ \bu_3 \end{array } \right]=\left[\begin{array}{c } \bu_1+\bu_2 \\ \bu_2 \\ \bu_3 \\ \bu_3 \end{array } \right]=\left[\begin{array}{c } \bb_1 \\ \bb_2 \\ \bb_1 \\ \bb_1 \end{array } \right].\ ] ] for @xmath271 , b^2=\left[\begin{array}{cc } 11 & 4\\ 8 & 3 \end{array } \right]$ ] , we can use the same argument as above to show that @xmath252 is @xmath253-rearrangeable . the corresponding matrices @xmath272 for each row vector of @xmath252 are listed by @xmath273 , \quad \left[\begin{array}{ccc } 3 & 4 & 0\\ 1 & 2 & 1 \\ 0 & 1 & 3 \end{array } \right ] , \quad \left[\begin{array}{ccc } 3 & 4 & 0\\ 0 & 1 & 3 \\ 0 & 1 & 3 \\ 2 & 3 & 0 \end{array } \right].\ ] ] [ exa - addition ] let @xmath274 $ ] where @xmath250 , \bu_2=[2,0 ] , \bu_3=[0,2 ] , \bu_4=[3,0 ] , \bu_5=[1,2]$ ] . let @xmath275,\quad b=\left[\begin{array}{cc } 3 & 2\\ 1 & 2 \end{array } \right].\ ] ] then @xmath50 is @xmath55-rearrangeable . analogous to example [ exa1 ] , we can check the definition by letting the corresponding rearranging matrices @xmath272 for each row vector of @xmath50 as follows : @xmath276 , \left[\begin{array}{ccccc } 1 & 1 & 1 & 0 & 0\\ 0 & 0 & 1 & 1 & 0 \end{array } \right ] , \left[\begin{array}{ccccc } 1 & 0 & 1 & 0 & 0\\ 0 & 0 & 0 & 0 & 1 \end{array } \right],\end{aligned}\ ] ] @xmath277 , \left[\begin{array}{ccccc } 1 & 1 & 1 & 0 & 0\\ 0 & 0 & 0 & 0 & 1 \\ 0 & 0 & 0 & 0 & 1 \end{array } \right].\end{aligned}\ ] ] [ th - identity ] let @xmath50 be @xmath55-rearrangeable as in definition [ def rearrangeable ] , then we have \(i ) @xmath278 ; \(ii ) @xmath279 is @xmath280-rearrangeable for any @xmath281 . \(i ) following the notation of definition [ def rearrangeable ] , for each row vector @xmath242 of @xmath50 , then @xmath282 we prove ( ii ) by induction . suppose @xmath279 is @xmath280-rearrangeable for some @xmath180 . for @xmath283 , we let @xmath284 be the @xmath254-th row vector of @xmath279 and @xmath285 the @xmath248-th row vector of @xmath286 . then there exists a @xmath287 matrix @xmath272 where @xmath288 such that @xmath289 and @xmath290^t,\ ] ] where the transpose means transposing the row of matrices into a column of matrices ( without transposing the @xmath291 itself ) . since @xmath292^t\quad\text{and}\quad b^{k+1}=b^k b=\big[{\boldsymbol \beta}_1 b , \dots , { \boldsymbol \beta}_n b\big]^t.\ ] ] it follows that @xmath293 . letting @xmath294 , then by ( i ) , we have @xmath295^t.\ ] ] therefore , @xmath296 is @xmath297-rearrangeable by definition [ def rearrangeable ] . * proof of theorem [ th - main ] : * by the assumption , let @xmath298\}_{i=1}^m$ ] be the equivalence classes of horizontal components in @xmath41 , and let @xmath299\}_{j=1}^n$ ] be the equivalence classes of vertices in @xmath49 . without loss of generality , we can assume that @xmath300 for otherwise , by the definition of incidence matrix @xmath51 of @xmath48 , each row sum and each column sum of @xmath51 must be positive . hence every row sum of @xmath286 will go to infinity as @xmath301 does . denote by @xmath302 the entries of @xmath286 , and let @xmath301 be sufficiently large such that @xmath303 by proposition [ th - identity ] , @xmath279 is @xmath280-rearrangeable . the ifs of the @xmath301-th iteration of @xmath304 has symbolic space @xmath305 and the augmented tree has incidence matrix @xmath279 ; moreover , the two hyperbolic boundaries @xmath306 and @xmath40 are identical . hence we can consider @xmath279 instead if is not satisfied . let @xmath307 . we define @xmath187 to be a one - to - one map from @xmath140 ( in @xmath23 ) to @xmath140 ( in @xmath178 ) inductively as follows : let @xmath308 suppose @xmath5 is defined on @xmath140 such that \(1 ) for every horizontal component @xmath42 , @xmath309 have the same parent ( using ) , i.e. , @xmath310 \(2 ) @xmath311 , i.e. , @xmath27 and @xmath312 lie in the same class @xmath168 $ ] for some @xmath248 . in order to define the map @xmath5 on @xmath313 , we note that @xmath42 in @xmath140 gives rise to horizontal components in @xmath313 , which are accounted by the incidence matrix @xmath50 . we write the descendants of @xmath42 as @xmath314 where @xmath315 are horizontal components consisting of @xmath313-descendants of @xmath42 . if @xmath42 belongs to some equivalence class @xmath316 $ ] ( corresponding to @xmath172 $ ] ) , then @xmath42 consists of @xmath171 vertices belonging to the class @xmath168 $ ] where @xmath317 . the total number of vertices of @xmath42 is @xmath318 . by the definition of incidence matrix @xmath50 , the total number of the @xmath313-descendants of @xmath42 is @xmath319 since @xmath50 is @xmath55-rearrangeable , there exists a nonnegative matrix @xmath243_{p\times m}$ ] such that @xmath245:=\left[\begin{array}{c } { \mathbf c}_1 \\ \vdots \\ { \mathbf c}_p \end{array } \right].\ ] ] according to the matrix @xmath272 , we decompose @xmath242 into @xmath320 groups as follows : for each @xmath321 and @xmath322 , we choose @xmath323 of those @xmath315 that belong to @xmath158 $ ] , and denote by @xmath324 the set of all the chosen @xmath301 with @xmath322 . then the index set @xmath325 can be written as a disjoint union @xmath326 hence @xmath314 can be rearranged as @xmath320 groups @xmath324 , and each @xmath324 corresponds to @xmath327 . namely , @xmath328 with @xmath329 , the @xmath330 denote two classes of vertices.,width=529 ] suppose the component @xmath331 , we have defined @xmath5 on @xmath140 and @xmath332 such that @xmath333 are equivalent . as @xmath334 $ ] , by the previous argument , among @xmath335 ( or @xmath336 ) , there are @xmath171 vertices belonging to the class @xmath168 $ ] , @xmath317 . since @xmath337 , for @xmath338 , we have @xmath339 which is equal to the total number of descendants of a vertex lying in @xmath168 $ ] ( due to the incidence matrix @xmath51 ) . we can define a one - to - one map @xmath5 on @xmath340 by assigning @xmath341 to the descendants of @xmath342 which belongs to @xmath168 $ ] . moreover , @xmath338 also means that the number of vertices belonging to @xmath343 $ ] in @xmath341 is the same as the number of descendants of the @xmath342 which belong to @xmath343 $ ] for @xmath344 . hence we can require the @xmath27 and its image @xmath312 are equivalent ( see figure [ fig.re ] ) . therefore , @xmath5 is well - defined on the descendants of @xmath42 and satisfies conditions ( 1)(2 ) . by applying the same construction of @xmath5 on the descendants of every horizontal component in @xmath140 . it follows that @xmath5 is well - defined on @xmath313 as desired . inductively , @xmath5 can be defined from @xmath23 to @xmath178 bijectively . the proof that @xmath5 is a near - isometry is the same as in @xcite . let @xmath87 , @xmath77 be the canonical geodesic connecting them , which can be written as @xmath345\ ] ] where @xmath346 $ ] is the horizontal part and @xmath347 , [ y^{-m},\dots , y^{-1 } , y]$ ] are vertical parts . clearly , @xmath346 $ ] must be included in a horizontal component of @xmath23 , we denote it by @xmath42 . by , it follows that @xmath348 where @xmath349 and @xmath350 are levels and lengths of @xmath351 , respectively . then we have @xmath352 where @xmath136 is a hyperbolic constant as in theorem [ th2.3 ] . if @xmath42 is a singleton , then @xmath353 ; otherwise , the elements of @xmath354 share the same parent . hence @xmath355 , and @xmath356 that completes the proof of the theorem . lipschitz equivalence of self - similar sets is an interesting problem , which has been undergoing considerable development recently . it is well - known that the hausdorff dimension is invariant under a bi - lipschitz map . however , it is impractical to find the hausdorff dimension of a fractal by using the relation of lipschitz equivalence to other ones . usually , hausdorff dimension is a prerequisite for lipschitz equivalence in the studies . we know that , under the osc , the hausdorff dimension of a self - similar set can be calculated easily ( @xcite@xcite ) . in the absence of the osc , it is much harder to compute the hausdorff dimension . nevertheless , the following formula is very useful for a large class of self - similar sets with overlaps . [ thm - finitetype ] let @xmath0 be the self - similar set on @xmath357 defined by ( [ set equ ] ) . if the finite type condition holds , then the hausdorff dimension of @xmath0 is given by @xmath358 where @xmath359 is the spectral radius of a matrix associated with the finite type condition , @xmath360 is the minimum contraction ratio . the following result was recently proved in @xcite , which establishes a very important relationship between the hyperbolic boundary of an augmented tree and the self - similar set . [ th - holder - equ ] with the same notation as in section 2 , the augmented tree @xmath54 is hyperbolic , and the hyperbolic boundary @xmath40 of @xmath54 is hlder equivalent to the self - similar set @xmath0 , i.e. , there exists a natural bijection @xmath361 and a constant @xmath6 such that @xmath362 with @xmath363 . by using the above theorem , we can obtain the lipschitz equivalence of self - similar sets . [ th - lip - sets ] let @xmath0 and @xmath364 be self - similar sets that are generated by two ifs s as in ( [ set equ ] ) with the same ratios @xmath365 . if the corresponding augmented trees @xmath366 are both simple and their incidence matrices are @xmath367-rearrangeable and @xmath368-rearrangeable respectively , where @xmath51 is the common incidence matrix of their tree structures . then @xmath0 and @xmath364 are lipschitz equivalent , and are also lipschitz equivalent to a dust - like self - similar set with the same ratios @xmath365 . if we replace the above @xmath366 by their quotient spaces @xmath369 , then @xmath0 and @xmath364 are still lipschitz equivalent , and are lipschitz equivalent to a cantor - type set ( may be not a dust - like self - similar set ) . it follows from theorem [ th - main ] and proposition [ th.simple tree ] that @xmath370 let @xmath371 be a bi - lipschitz map . by theorem [ th - holder - equ ] , there exist two bijections @xmath372 and @xmath373 satisfying with constants @xmath374 , respectively . now we define @xmath375 as @xmath376 then @xmath377 let @xmath378 , then @xmath379 . moreover , we have @xmath380 by using another inequality of . therefore @xmath381 is a bi - lipschitz map and @xmath382 . if we regard @xmath49 as the augmented tree of an ifs that is strongly separated , then apply the above conclusion to obtain that @xmath0 or @xmath364 is lipschitz equivalent to the dust - like self - similar set . the second part follows from theorem [ th - quotient - tree ] and proposition [ th.simple tree ] by the same argument . since @xmath383 is a tree , its hyperbolic boundary is a cantor - type set . however , due to the identification of vertices , it seems hard to regard @xmath383 as an augmented tree of an ifs that is strongly separated . finally we illustrate our main results by two concrete examples . the first one with the osc was considered in @xcite ( or @xcite@xcite ) by using algebraic properties of contraction ratios , we will give a different proof from our perspective ; the second one without the osc is completely new . the higher dimensional case can also be handled but need more complicated discussions , we omit it here . let @xmath384 and @xmath385 where @xmath386 . let @xmath0 ( resp . @xmath364 ) be the self - similar set generated by the ifs @xmath387 ( resp . @xmath388 ) ( see figure [ fig - dust - touch ] ) . then @xmath382 . obviously both the two ifs s satisfy the osc , and @xmath364 is dust - like while @xmath0 is not as @xmath389 ( see figure [ fig - dust - touch ] ) . let @xmath39 be the augmented tree of @xmath0 defined as in section 2 . then @xmath49 can be regarded as the augmented tree of @xmath364 . by a few calculations , it can be seen that there are only five equivalence classes of the horizontal components in @xmath39 : @xmath155=[\{\vartheta\}]$ ] ; @xmath390=[\{2 , 3\}]$ ] ; @xmath391=[\{12 , 13\}]$ ] ; @xmath392=[\{22 , 23 , 311\}]$ ] ; @xmath393=[\{122 , 123 , 131\}]$ ] . hence @xmath41 is simple and incidence matrix is @xmath275.\ ] ] for the tree @xmath49 , the equivalence classes of vertices are @xmath163=[\{\vartheta\}]$ ] , @xmath394=[\{13\}]$ ] and the incidence matrix is @xmath395.\ ] ] [ h ] let @xmath274 $ ] be the representation of @xmath155,\dots , [ t_5]$ ] with respect to @xmath163 , [ \bt_2]$ ] , where @xmath250 , \bu_2=[2,0 ] , \bu_3=[0,2 ] , \bu_4=[3,0 ] , \bu_5=[1,2]$ ] . it follows from example [ exa - addition ] that the matrix @xmath50 is @xmath367-rearrangeable . therefore @xmath382 by theorem [ th - lip - sets ] let @xmath396 be an ifs where @xmath397 $ ] , and let @xmath398 be the self - similar set . then @xmath399 if and only if @xmath400 $ ] . by the assumption , @xmath401 $ ] . it is easy to see that @xmath211 ( see figure [ fig - exa ] ) , hence the osc does not hold for all @xmath402 . however , if @xmath403 $ ] , the ifs satisfies the finite type condition @xcite ( hence satisfies the wsc @xcite ) . by lemma [ thm - finitetype ] , all such self - similar sets @xmath398 have the common hausdorff dimension @xmath404 where @xmath405 is the spectral radius of the following matrix associated with the finite type condition @xmath406.\ ] ] [ h ] if @xmath407 , then all self - similar sets @xmath398 are lipschitz equivalent to each other as their augmented trees have the same quotient space ( see figure [ fig - exa ] ( a ) ) . hence it suffices to show @xmath408 where @xmath409 . by example [ example - quotienttree ] , the quotient space @xmath62 of @xmath410 is simple . similarly it can be seen that the quotient spaces of @xmath411 and @xmath412 are also simple . we write the incidence matrices of @xmath413 as follows @xmath414 ; \quad a_2=\left[\begin{array}{rr } 2 & 1\\ 3 & 2 \end{array}\right ] ; \quad a_3=\left[\begin{array}{rrrr } 1 & 1 & 0 & 1 \\ 1 & 1 & 1 & 0 \\ 2 & 2 & 1 & 2 \\ 1 & 1 & 1 & 2 \end{array}\right].\end{aligned}\ ] ] moreover , the corresponding reduced trees @xmath64 share the same incidence matrix @xmath415 $ ] , thus they are graph isomorphic by proposition [ th.simple tree ] . let @xmath249 $ ] be the representation for the quotient space of @xmath410 where @xmath250 , \bu_2=[2,1],\bu_3=[3,1]$ ] . then @xmath416 is @xmath367-rearrangeable by example [ exa1 ] . by checking the definition , similarly we have @xmath417 is @xmath367-rearrangeable where @xmath418 $ ] is a representation for the quotient space of @xmath411 where @xmath250 , \bu_2=[1,1]$ ] ; and @xmath419 is @xmath367-rearrangeable where @xmath420 $ ] is a representation for the quotient space of @xmath412 where @xmath421 , \bu_2=[1,1 ] , \bu_3=[2,1 ] , \bu_4=[3,1]$ ] . from theorem [ th - quotient - tree ] , it concludes that @xmath232 . hence the corresponding self - similar sets are lipschitz equivalent by theorem [ th - lip - sets ] . if @xmath422 $ ] , then the overlaps of the ifs become more complicated but it is not hard to see that @xmath423 @xcite . therefore , the @xmath424 can not be lipschitz equivalent to @xmath412 . in this section , we provide a criterion for the hyperbolic boundary @xmath40 to be totally disconnected , which answers a question posed in @xcite . [ lem - tree1 ] let @xmath125 be an augmented tree defined in section 2 , and let @xmath77 be a geodesic with @xmath425 and be of the trapezoidal form ( not necessarily canonical ) . \(i ) if @xmath426 , then @xmath427 is a geodesic ; \(ii ) if @xmath428 , then @xmath429 can be joined by a path from @xmath430 of length less than @xmath136 which is the hyperbolic constant as in theorem [ th2.3 ] . \(i ) is obvious ; ( ii ) the assumption implies that @xmath77 is a horizontal geodesic . hence @xmath431 by theorems [ th - holder - equ ] and [ th2.3 ] . by taking the path formed by the parents of the vertices in @xmath77 , we see that @xmath429 are joined by a path in @xmath430 with length less than @xmath136 . let @xmath125 be an augmented tree of bounded degree . let @xmath432 be a sequence of horizontal components such that @xmath433 . then there exist two sequences of points @xmath434 with @xmath435 and @xmath436 . without loss of generality , we assume @xmath437 . suppose the statement is false , then there exists @xmath438 such that @xmath439 this implies all geodesics connecting @xmath440 have length less than @xmath441 . there exists @xmath442 ( independent of @xmath28 ) such that all such geodesics are contained in @xmath443 ( they are trapezoidal , but not necessary canonical ) . consider @xmath444 , the set of @xmath301-th generation ancestors of @xmath445 . it is clear that @xmath444 is connected by the definition of horizontal edge set @xmath36 . note that if @xmath446 and @xmath447 ( case ( ii ) in lemma [ lem - tree1 ] ) . then @xmath448 are joined by a path in @xmath430 of length less than @xmath136 , and inductively , @xmath449 are joined by a path in @xmath450 of length less than @xmath136 . if @xmath446 but @xmath451 ( case ( i ) in lemma [ lem - tree1 ] ) , then there exists @xmath452 such that @xmath453 . in this case we can derive as the above that @xmath449 again can be joined by a path in @xmath450 of length less than @xmath136 . from the above discussion , we see that for any two vertices in @xmath454 can be joined by a path in @xmath450 of length less than @xmath136 . as @xmath455 has bounded degree , say @xmath456 . then @xmath457 on the other hand , @xmath56 has bounded degree , say @xmath458 . then @xmath459 this leads to contradiction as @xmath433 . under the same hypothesis of the above lemma , the set of limit points of @xmath432 is not a singleton . [ lem - tree3 ] let @xmath125 be an augmented tree , let @xmath460 ( @xmath28-th level ) be a non - self - intersecting path . if @xmath461 , then there exists @xmath462 such that @xmath463 . let @xmath464 , then either @xmath465 or @xmath466 . that is , @xmath467 , as a function of @xmath468 , increases either by at most @xmath469 or decreases by at most @xmath469 at each step . then the lemma follows . [ th - bdy is total.dis ] let @xmath54 be an augmented tree of bounded degree . then the hyperbolic boundary @xmath40 ( or the self - similar set @xmath0 ) is totally disconnected if and only if the sizes of horizontal components are uniformly bounded . the theorem remains valid if we replace @xmath23 by its quotient space @xmath70 . we prove the theorem by contrapositive . the sufficiency is essentially the same as in @xcite by using topological arguments , we sketch the main idea here . if @xmath40 is not totally disconnected , so is the self - similar set @xmath0 by theorem [ th - holder - equ ] . then there is a connected component @xmath470 contains more that one point . note that for any @xmath471 , @xmath472 where @xmath473 . let @xmath474 . if @xmath475 , then it is a relatively open set in @xmath272 , and as @xmath272 is connected , @xmath476 ( @xmath477 means the relative boundary of @xmath478 in @xmath272 ) . let @xmath479 be in the intersection , there exists @xmath480 such that @xmath481 and @xmath482 . inductively , if @xmath483 does not cover @xmath272 , then we can repeat the same procedure to find @xmath484 such that @xmath485 since @xmath472 , this process must end at some step , say @xmath486 , and in this case @xmath487 . since the diameter @xmath488 as @xmath489 , @xmath486 must tend to infinity , which contracts the uniform boundedness of the horizontal components on the levels @xmath140 . for the necessity , let @xmath432 be a sequence of horizontal components satisfying @xmath490 . without loss of generality , we assume @xmath437 . let @xmath42 be the set of limit points of @xmath432 . then @xmath491 and is a compact set . since @xmath40 is totally disconnected , there exists non - empty compact subsets @xmath492 such that @xmath493 let @xmath494 such that @xmath495 and @xmath496 . hence for @xmath28 large enough , @xmath497 observe that @xmath498 ( here @xmath499 ) . for otherwise , there exists @xmath99 and @xmath500 such that @xmath501 this implies that there exist @xmath502 such that @xmath503 . hence the limit points of @xmath504 and @xmath505 are identical in the hyperbolic boundary @xmath40 , a contradiction . let @xmath506 , by @xmath507 and @xmath445 is connected and @xmath508 , we conclude that @xmath509 ( similar argument as the above ) . then by lemma [ lem - tree3 ] , there exists @xmath510 such that @xmath511 let @xmath512 be a limit point of @xmath513 , then @xmath514 , a contradiction . this completes the proof of the first part . the similar argument can also be used to show the second part without making many changes . in @xcite , we observed that if the incidence matrix @xmath50 of an augmented tree is primitive , then @xmath279 is rearrangeable for some @xmath281 , by using this , we obtained a stronger but simpler result ( theorem 1.1 of @xcite ) . however , we do not know if the observation is still true for the generalized rearrangeable matrix of this paper . so we ask later in @xcite , for the equicontractive ifs with the osc , by some modification , we removed the primitive ( or rearrangeable ) assumption on the incidence matrix . so for the the ifs with non - equal contraction ratios and overlaps , we wonder on the other hand , a more general augmented tree was introduced by lau and wang @xcite , which is defined on a tree ( independent of the ifs ) with an associated set - valued map . they proved that such augmented tree is hyperbolic and any compact set in @xmath357 can be hlder equivalent to the hyperbolic boundary of certain augmented tree . hence more flexible iterated schemes can also be fit into the framework . as for the application to lipschitz equivalence , by using the method of this paper , it is still interesting to carry out a similar study on moran sets or more general fractals . * acknowledgements : * the author gratefully acknowledges the support of k. c. wong education foundation and daad . he also would like to thank professor ka - sing lau for his support and valuable discussions , especially on section 5 . m. r. bridson and a. haefliger , _ metric spaces of non - positive curvature _ , grundlehren der mathematischen wissenschaften [ fundamental principles of mathematical sciences ] 319 ( springer , berlin , 1999 ) xxii+643pp . q.r . deng , k.s . lau and s.m . ngai , _ separation conditions for iterated function systems with overlaps _ , fractal geometry and dynamical systems in pure and applied mathematics . i. fractals in pure mathematics , 1 - 20 , contemp . math . , 600 , amer . soc . , providence , ri , ( 2013 ) .

from the general assumption that gravity originates from the coupling and thermal equilibrium between matter and vacuum , after a derivation of newton s law of gravitation and an interpretation of the attractive gravity force between two classical objects , we consider the macroscopic quantum gravity effect for particles whose wave packets are delocalized at macroscopic scale . we predict an abnormal repulsive gravity effect in this work . for a sphere full of superfluid helium , it is shown that with a gravimeter placed in this sphere , the sensitivities of the gravity acceleration below could be used to test the abnormal quantum gravity effect , which satisfies the present experimental technique of atom interferometer , free - fall absolute gravimeters and superconducting gravimeters . we further propose a self - consistent field equation including the quantum effect of gravity . as an application of this field equation , we give a simple interpretation of the accelerating universe due to dark energy . based on the idea that the dark energy originates from the quantum gravity effect of vacuum excitations due to the coupling between matter and vacuum , without any fitting parameter , the ratio between dark energy density and matter density ( including dark matter ) is calculated as , which agrees quantitatively with the result obtained from various astronomical observations .

1101.4890

although newton s gravity law and einstein s general relativity have given marvelous understanding about gravity , it is still the most mysterious problem in the whole field of science @xcite . it is widely believed that the unification of gravity and quantum mechanics and the unification of gravity and other three fundamental forces are impossible in the foreseeable future . one of the main reason is that the abnormal effects observed by experiments on earth are highly scarce . to observe possible abnormal effects , most recently , adler , mueller and perl @xcite proposed a terrestrial search for dark contents of the vacuum using atom interferometry , somewhat similarly to the michelson - morley experiment . recently , verlinde s work @xcite has renewed enthusiasm @xcite of the possibility that gravity is an entropy force , rather than a fundamental force @xcite . of course , if gravity is not a fundamental force , we should reconsider the unification of quantum mechanics and gravity . inspired by the thermodynamic origin of gravity for classical objects , we will try to consider several fundamental problems : ( i ) why the gravity between two classical objects is attractive ? ( ii ) does there exist new and observable quantum gravity effect ? ( iii ) whether there is new clue to solve the mechanism of accelerating universe ? in this paper , the quantum effect of gravity is studied based on the general principle that gravity originates from the coupling and thermal equilibrium between matter and vacuum background . for classical particles , this general principle gives the newton s gravity law . for particles described by quantum wave packets , we predict an abnormal quantum effect of gravity . based on this abnormal quantum gravity effect , we consider the possible origin of the dark energy from the coupling and thermal equilibrium between matter and vacuum background . quite surprising , the ratio of the dark energy in our simple calculations is @xmath2 , which agrees quantitatively with the result 7/3 obtained from various astronomical observations @xcite . our works also show that , with a sphere full of superfluid helium , there is a feasible experimental scheme to test our idea with a gravimeter placed in the sphere . the sensitivities of @xmath0 below @xmath1 could be used to test our idea , which satisfy the present experimental technique of atom interferometer @xcite , free - fall absolute gravimeters @xcite and superconducting gravimeters @xcite . the paper is organized as follows . first , in sec . ii , we consider the change of entropy for a particle with a displacement in space . for a particle with an acceleration , we give a derivation of the vacuum temperature due to the coupling between matter and vacuum . based on the consideration of local thermal equilibrium , we give the acceleration of a particle in the presence of a finite vacuum temperature field distribution . in sec . iii , we give a derivation of newton s law of gravitation . in particular , we explain the physical mechanism of the attractive gravity between two classical objects . in sec . iv , we give a brief discussion of the physical mechanism of the equivalence principle based on the thermodynamic origin of gravity . in sec . v , we consider an abnormal quantum gravity effect for a particle described by the wave function in quantum mechanics . in sec . vi , we consider an experimental scheme to test this abnormal quantum gravity effect with a superfluid helium sphere . the application of this abnormal quantum gravity effect to test many - world interpretation and de broglie - bohm theory is discussed . we also discuss the application of this abnormal quantum gravity effect to condensed matter physics . in sec . vii , we give the field equation including the quantum gravity effect of vacuum excitations . this gives a possible interpretation of the repulsive gravity effect of dark energy . in sec . viii , we calculate the dark energy density from the general principle in this work . in sec . ix , the general field equation including the classical and quantum gravity effect of matter and radiation is given . we give relevant summary and discussions in the last section . compared with electromagnetism , weak interaction and strong interaction , the gravitation has several different features . ( i ) the gravitation is universal . ( ii ) the gravitation `` charge '' is in a sense the energy - momentum tensor . hence , the gravitation `` charge '' is not quantized . ( iii ) the coupling between the energy - momentum tensor and spacetime leads to the gravity force for another particle . ( iv ) the gravity law closely resembles the laws of thermodynamics and hydrodynamics @xcite . these features strongly suggest that the gravitation deserves studies with completely different idea , compared with other forces . it is well known that the force in classical and quantum gases can be understood in a natural way with statistical mechanics . following the intensive pioneering works to eliminate the gravitation from a fundamental force , we will study the thermodynamic origin of gravitation , and in particular the quantum gravity effect when both quantum mechanics and thermodynamics are considered . although the theoretical predication about the abnormal quantum gravity effect addresses a lot of subtle problems , our starting points are the following two formulas about the change of entropy for a displacement of a particle , and the vacuum temperature because of the vacuum excitations induced by an accelerating object . \(1 ) the formula about the change of entropy @xmath4 after a displacement @xmath5 for a particle with mass @xmath6 . @xmath7 in the original work by verlinde @xcite , this postulation motivated by bekenstein s work @xcite about black holes and entropy , plays a key role in deriving the newton s law of gravitation . the above formula means that after a displacement @xmath5 for a particle with mass @xmath6 , there is an entropy increase of @xmath4 for the whole system . to understand this formula , we emphasize three aspects about this formula . ( i ) the whole system about the entropy includes the vacuum background . ( ii ) there is a strong coupling between the particle and vacuum background . this can be understood after little thought . without a strong coupling , it is nonsense to define the location and time for a particle existing in the spacetime ( @xmath8 the vacuum background ) . ( iii ) as a medium for various matters , the zero - point ( or ground - state ) energy density of the vacuum background is extremely large with the standard quantum field theory . the vacuum zero - point energy density is of the order of @xmath9 if the energy cutoff is the planck energy . this also gives the reason why there is a strong coupling between matter and vacuum . in a sense , the motion of a particle in the vacuum background is a little similar to a speedboat moving in the sea . the speedboat left behind a navigation path in the sea . after a navigation of distance @xmath5 , the speedboat stopped . after waiting sufficient time , we can not know the navigation path in the sea . if the location resolution in the navigation path in the sea is @xmath10 , about @xmath11 bits of information are lost . in this situation , there is an entropy increase of @xmath12 . for a matter ( similar to the speedboat ) moving in the vacuum background ( similar to the sea ) , it is similar to understand the relation @xmath12 with @xmath10 being the location resolution in the sight of the vacuum background . here we consider a method to calculate the location resolution @xmath10 . from special relativity , the energy of the particle is @xmath13 . together with quantum mechanics , the eigenfrequency is @xmath14 . it is understandable that various gauge fields in the vacuum have the propagation velocity of light velocity @xmath15 . hence , the relative velocity between matter and the gauge field in the vacuum is @xmath15 . in this situation , we get the coherence length @xmath10 of the particle in the sight of the vacuum background . with the standard quantum mechanics method , we have @xmath16 . this coherence length can also be regarded as the location resolution in the sight of the vacuum background . more detailed discussions about this coherence length and entropy are given in the appendix . \(2 ) vacuum temperature induced by a uniformly accelerating object in the vacuum background . considering a particle with acceleration @xmath17 , we have @xmath18 here @xmath19 . in this paper , all bold symbols represent vectors . from eq . ( [ entropy ] ) , we have @xmath20 in addition , from @xmath21 , we have@xmath22 using the fundamental thermodynamic relation @xmath23 for the whole system including the vacuum background , we have@xmath24 in this process , there is no entropy increase for the particle itself . the entropy increase comes from the vacuum . hence , @xmath25 refers to the vacuum temperature at the location of the particle . because temperature is a concept of statistical average , rigorously speaking , the right - hand side of the above expression should be written as the form of statistical average , @xmath8@xmath26 if the fluctuations of the acceleration can be omitted , we have@xmath27 this shows that the acceleration of a particle will induce vacuum excitations , and thus lead to finite vacuum temperature . although the above formula is the same as the unruh temperature @xcite , the physical meaning is in a sense different from that of the unruh effect . in the present work , the temperature @xmath25 denotes the vacuum temperature due to the vacuum excitations . because the unruh effect itself addresses a lot of subtle problems , in this paper , we will not discuss the detailed difference between @xmath25 and unruh temperature . we will apply the above expression with our understanding of @xmath25 in this work . for the particle b , the acceleration induces a temperature field distribution shown by the blue dashed line , with the peak value given by eq . ( [ eq : u2 ] ) . the local thermal equilibrium requests that this peak value is equal to the temperature of the vacuum temperature field ( red line ) at location b. this gives the physical mechanism why the particle b will accelerate in the presence of a finite vacuum temperature field.,scaledwidth=80.0% ] the meaning of the above equation is further shown in fig . 1 . for a particle at location b with acceleration @xmath28 ( shown by the dashed red arrow ) , the coupling between the particle and the vacuum establishes a temperature field distribution shown by the dashed blue line with peak value @xmath29 . in a sense , the strong coupling between the particle and the vacuum leads to a `` dressed '' state which includes the local vacuum excitations and the particle itself . if the particle has no size , the width of the local vacuum excitations is of the order of the planck length , with the same consideration of the derivation of the newton s law of gravitation in the following section . in fig . 1 , there is a temperature field distribution shown by the red line . at location b , there is a particle denoted by red sphere . to establish local thermal equilibrium , the red sphere will accelerate so that the peak temperature of the dressed state is equal to the temperature of the vacuum temperature field at the location b , @xmath8 @xmath30 . in this situation , we have@xmath31 however , the acceleration is a vector , while the temperature is a scalar . therefore , the above formula is not clear about the direction of the acceleration . we further consider the free energy of the whole system which is defined as @xmath32 . here @xmath33 is the overall energy of the system , which is a conserved quantity . because the system evolution has the tendency to decrease the free energy with the most effective way , we get the following formula to determine the magnitude and direction of the acceleration in the presence of a vacuum temperature field.@xmath34 here @xmath35 denotes a three - dimensional spatial vector . in fig . 1 , for the particle at location b , the direction of the acceleration is determined based on the consideration of the free energy . we will show in due course that the above equation will interpret why the gravity force is attractive between two spatially separated objects . note that eq . ( [ unruhvector ] ) is invalid for uniformly distributed vacuum temperature field . for uniformly distributed vacuum temperature field such as location 1 , from the consideration of the free energy , the direction of the acceleration of a particle is completely random . this means that the direction of the acceleration is highly fluctuating . although @xmath36 for this case , @xmath37 if the uniformly distributed vacuum temperature is nonzero . by using eq . ( [ eq : statemperature ] ) , we have @xmath38 . from @xmath39 and @xmath23 , we have @xmath40 . using eqs . ( [ x])-([t ] ) , we have@xmath41 from the above equation , we get the following inertia law ( newton s second law ) @xmath42 of course , from @xmath39 we can directly get this inertia law . the above derivation of the inertia law shows that it is self - consistent to consider the origin of classic force from statistical mechanics and the coupling between matter and vacuum background . if @xmath43 , we have @xmath44 . this is newton s first law . the coupling between the particle and vacuum background , and the consistency with newton s first law suggest that the vacuum background is in a sense a superfluid . in this superfluid , the propagator velocity of various gauge fields is the light velocity @xmath15 . if @xmath15 is regarded as the sound velocity of the vacuum background , the critical velocity to break the superfluidity is @xmath15 . as discussed in previous section , the coupling between particle and vacuum background leads to an entropy increase for a displacement . the physical mechanism of this entropy increase is that there is a strong coupling between matter and vacuum . in this section , we give a derivation of newton s law of gravitation physically originating from these vacuum excitations . ) for the vacuum fluctuations . the red sphere shows a particle which leads to a vacuum temperature field distribution , through the coupling with the vacuum.,scaledwidth=80.0% ] when there are vacuum excitations due to the coupling between matter and space , we assume @xmath45 as the correlation length of the vacuum excitations . at least for the situations considered in the present work , the vacuum excitation energy density is much smaller than the vacuum zero - point energy density . we will show in due course that @xmath45 is of the order of the planck length @xmath46 with @xmath47 being the gravitational constant , which is the length scale at which the structure of spacetime becomes dominated by quantum gravity effect . in fig . 2 , we consider the three - dimensional space with @xmath45 showing the structure of the space due to quantum gravity effect . although the microscopic mechanism of @xmath45 is not completely clear , it is not unreasonable to assume the existence of @xmath45 in the structure of the space . these lowest space structures are a little similar to atoms in a solid . we assume the energy of a particle shown by red sphere in fig . 2 is @xmath48 . we further assume the freedom of the lowest space structure is @xmath49 . from the local thermal equilibrium , at the location of the particle , the temperature of the vacuum is determined by@xmath50 here @xmath51 denotes a dimensionless coupling strength between matter and space . from the ordinary statistical mechanics , @xmath52 should be of the order of @xmath53 . we have@xmath54 when @xmath55 , @xmath43 . hence , one expects the temperature field distribution shown in fig . 3 . for another particle with mass @xmath6 in this temperature field distribution , from eq . ( [ unruhvector ] ) , the acceleration field distribution is then@xmath56 here the radial unit vector @xmath57 . to get the above expression , the spherical symmetry of the system for @xmath58 is also used . this explains the attractive gravity force between two classical objects . it is worthy to point out that both in newton s gravitation law and einstein s general relativity , this attractive gravity force is imposed from the observation results , rather than microscopic mechanism . one of the merit of thermodynamics lies in that even we do not know the exact collision property such as scattering length between atoms , the macroscopic force such as pressure can be derived . when the thermodynamic origin of gravity is adopted , there is a request that one gets the correct result of the direction of gravity force . .,scaledwidth=80.0% ] from the continuous property ( gauss s flux theorem ) of the force @xmath59 , we have@xmath60 note that the above expression holds for @xmath61 , so that the spherical symmetry approximation can be used . combined with eq . ( [ eq : tvr0 ] ) , we have @xmath62 . although we do not know the exact value of @xmath63 , it is understandable that @xmath63 is of the order of @xmath53 . in this situation , we have@xmath64 to get the above expression , we have used @xmath65 . using eq . ( [ eq : ar ] ) , we have@xmath66 assuming@xmath67 we get the standard result of newton s gravitation law@xmath68 we see that the correlation length of the vacuum excitation is of the order of the planck length . the temperature field distribution for @xmath58 becomes@xmath69 we consider above the situation of a particle whose size is of the order of the planck length . if the size of the particle is larger than the planck length , by using the gauss s flux theorem , for @xmath70 being much larger than the size of the particle , the above result still holds . it is worthy to further consider the meaning of the vacuum excitations due to the coupling between matter and space . the above derivations give the temperature field distribution . assuming the vacuum zero - point energy density is @xmath71 , we consider the situation that a particle with mass @xmath72 suddenly emerges at the location @xmath73 . this will lead to the establishment of the vacuum temperature field distribution ( [ eq : tr ] ) . however , one should note that in the establishment process of the temperature field distribution , the whole energy should be conserved . hence , assuming that @xmath74 is the vacuum energy density in the presence of @xmath72 , we have@xmath75 @xmath76 this physical picture is further shown in fig . 4 . , the blue line shows the fluctuating vacuum energy distribution @xmath74 around the zero - point vacuum energy density @xmath71 . we will show in due course that when the whole universe is considered , @xmath74 will exhibit fluctuation behavior around @xmath77 with @xmath78 being the dark energy density.,scaledwidth=80.0% ] for an assembly of classical fundamental particles ( here classical means that the quantum wave packet effect is negligible ) shown in fig . 5 , we assume the temperature field distribution due to the @xmath49th particle is @xmath79 . because there is no quantum interference effect between different classical particles , the force ( a measurement result in quantum mechanics ) on the object with mass @xmath6 is @xmath80 the above expression is based on the assumption of the linear superposition of gravity force @xmath81 . in the newton s law of gravitation for an assembly of classical particles , this implicit assumption is also used . we stress here that , rigorously speaking , the summation in the above expression is about all fundamental particles . one may still consider the following possibility of calculating the acceleration field@xmath82 obviously , the above expression contradicts with the newton s law of gravitation . considering the astrodynamics addressing three bodies , this method to calculate the gravity force should be ruled out . the reason of the error in the above calculations lies in the method to get @xmath25 . even for the ordinary situation , the above method to get @xmath25 is also not correct . we consider the presence of @xmath83 thermal sources at different location @xmath84 . the temperature increase for an observer due to these @xmath83 thermal sources is @xmath85 . if only one thermal source exists , we assume that the temperature increase for the observer is @xmath86 . it is obvious that when @xmath83 thermal sources coexist , the temperature increase for the observer is not @xmath87 . it is well known that the equivalence principle is based on the assumption that the gravitational mass @xmath88 is equal to the inertial mass @xmath89 . it is still a mystery why @xmath90 because it seems that the gravitational mass and inertial mass are different physical concept , if the gravity force is regarded as a fundamental force , similarly to other fundamental forces . for example , for electromagnetism , the inertial mass is completely different from the electric charge leading to electromagnetism . previous studies clearly show that if the thermodynamic origin of gravity force is adopted , the gravitational mass and the inertial mass are the same mass in the mass - energy relation in special relativity . in all our derivations , we do not need especially introduce the gravitational mass at all . in this situation , the gravitational mass and inertial mass are in fact the same physical concept . hence , it is not surprising that @xmath90 . as discussed previously , the velocity of light is in a sense the sound velocity of the vacuum . if the vacuum is regarded as a superfluid , there is a breakdown of the superfluidity if the velocity of an object is larger than @xmath15 . this gives a possible physical mechanism why @xmath15 is the speed limit for all objects . hence , the validity of special relativity is in a sense determined by the vacuum background where there is a finite and extremely large zero - point energy density . in other words , it s the vacuum background which leads to the theory of special relativity . this is the reason why there is an internal relation between @xmath15 and vacuum zero - point energy density . in a sufficiently small region of a freely falling system , the temperature due to the acceleration of this system is equal to the vacuum temperature field . in this small region , the object will not experience the finite vacuum temperature effect . this suggests the validity of the strong principle of equivalence . hence , when the quantum effect is included in the field equation of gravitation , we will adopt the strong principle of equivalence , which is also adopted by einstein in deriving his field equation for gravitation . from @xmath12 and @xmath91 , we see that there is another definition of the mass . for different fundamental particles , there are different correlation lengths in the coupling between particles and vacuum . the mass is then @xmath92 . considering previous studies that newton s first law , second law and newton s gravitation law can be derived from @xmath12 , this gives the definition of mass with the view of information . for the particle with inertial mass @xmath6 , we consider the possible change of @xmath6 due to a large mass near us , such as the milky way galaxy . at the location of this particle , assuming the presence of the milky way galaxy leads to a change of the correlation length @xmath10 , we have @xmath93 . assuming at the location of the particle , the vacuum temperature field due to the milky way galaxy is @xmath25 , because @xmath10 comes from the coupling between the vacuum and particle , to the first order , @xmath94 would be of the order of @xmath95 with @xmath96 being the planck energy . in this situation , we have @xmath97 . we see that , to observe the possible change of the inertial mass , extremely large @xmath25 would be needed . in most situations we can imagine , we can not distinguish the gravity effect between einstein s general relativity and mach s principle . it is not clear , whether at the singular point of black hole or during the planck epoch et al . , there would be obvious effect predicted by mach s principle . in the thermodynamic origin of gravity , for an object with mass @xmath72 , it establishes a temperature field of @xmath98 . using the formula @xmath99 about the relation between acceleration and temperature , we get the newton s gravity law @xmath100 between two classical objects . in particular , the attractive gravity force between two classical objects is explained . in this section , we will consider the quantum gravity effect by including quantum mechanics in the thermodynamic origin of gravity . to give a general study on the quantum gravity effect , we consider the following wave function for a fundamental particle ( such as electron ) with mass @xmath101,@xmath102 the average density distribution @xmath103 is shown by the black quantum sphere in fig . 6(a ) . for @xmath104 , similarly to the consideration of a classical particle , it is easy to get @xmath105 at @xmath73 , from the consideration of the spherical symmetry , @xmath106 . hence , we have @xmath107 . it is clear that the temperature field distribution is that shown in fig . 7 . for @xmath108 , using again the spherical symmetry , we have@xmath109 ) .,scaledwidth=80.0% ] for the quantum wave packet shown in fig . 6(a ) , from the relation between acceleration field and temperature field distribution given by eq . ( [ unruhvector ] ) , we have@xmath110 this means a remarkable predication that in the interior of the quantum sphere , the gravity force is repulsive ! this abnormal effect is further shown in fig . it is clear that this abnormal quantum gravity effect physically originates from the quantum wave packet effect for the particle in the black sphere . at @xmath111 , the average acceleration field is zero . however , because of the finite vacuum temperature , the direction of the acceleration is highly fluctuating , @xmath8 @xmath112 and @xmath113 . it s this highly fluctuating acceleration field leads to different value of @xmath114@xmath115 in the interior and exterior of the quantum sphere . if there are @xmath83 particles in the same quantum state given by eq . ( [ wavefunction ] ) , we have@xmath116 for this abnormal quantum gravity effect , one should be very careful . without experimental verification , it is only a theoretical predication . although we will show that it is possible that this abnormal quantum gravity effect just gives an interpretation to the accelerating universe due to dark energy , it is necessary to consider whether this theoretical predication is self - consistent . to further understand the abnormal quantum gravity effect , we consider a classical sphere with the following density distribution@xmath117 this density distribution for a classical sphere is shown in fig . at first sight , one may get the same acceleration field distribution , compared with the situation of the quantum sphere . this is not correct . for a classical sphere , assuming there are @xmath83 particles , it is very clear that the wave packets of all particles are highly localized . hence , for a particle at location @xmath118 , the temperature field distribution due to this particle is @xmath119 . from @xmath120 , we get the result given by newton s law of gravitation , shown in fig . it is easy to show that the quantum states of the quantum sphere and classical sphere are completely different . in the quantum sphere , the many - body wave function is@xmath121 here @xmath122 is given by eq . ( [ wavefunction ] ) . for the classical sphere , the many - body wave function is@xmath123.\ ] ] here @xmath124 , @xmath125 , @xmath126 are highly localized wave functions . @xmath127 denotes all the permutations for the particles in different single - particle state . @xmath15 is a normalization factor . we consider here the situation of bosons . it is similar for fermi system . it s this essential difference of the many - body wave function that leads to different gravity effect . assuming there are @xmath83 fundamental particles whose wave functions are @xmath128 , we give here the formulas to calculate the acceleration field due to these @xmath83 particles . previous studies directly lead to the following two formulas . and@xmath130 here @xmath131 is the mass of the @xmath132th fundamental particle . the integral in the right - hand side of eq . ( [ temfield ] ) is due to the quantum wave packet of the @xmath132th fundamental particle , while the norm of the vector after calculating this integral is due to the fact that @xmath133 is a scalar field larger than zero and @xmath133 is an observable quantity based on eq . ( [ acc ] ) . it is easy to show that if all these @xmath83 particles are highly localized classical particles , we get the newton s gravity law @xmath134 , with @xmath135 being the location of the @xmath132th particle . we assume that a fundamental particle with mass @xmath72 is uniformly distributed in the sphere with radius @xmath136 . for the vacuum temperature field due to this particle , the maximum temperature is@xmath137 from the mass - energy relation , even all the energy is transferred into temperature with only one freedom , we get a limit temperature@xmath138 obvious , there is a request of@xmath139 this means that @xmath140 this result shows that any object having strong coupling with the vacuum background can not be distributed within a sphere of radius @xmath141 , irrelevant to its mass . this provides a physical mechanism why in our calculations of @xmath25 , the volume of the smallest space unit is of the order of @xmath142 . now we turn to consider a feasible experimental scheme to test further the quantum gravity effect with superfluid @xmath143he , shown in fig . 8 . for brevity s sake , we consider a sphere full of superfluid @xmath143he . there is a hole in this sphere . in this situation , from eqs . ( [ acc ] ) and ( [ temfield ] ) , the gravity acceleration in the sphere due to superfluid @xmath143he can be approximated as @xmath144 here the liquid helium density is @xmath145 kg / m@xmath146 . from this , the anomalous acceleration is @xmath147 . the gradient of this anomalous acceleration is @xmath148 . even only the condensate component of superfluid @xmath143he is considered , for a superfluid @xmath143he sphere whose radius is @xmath149 , the maximum anomalous acceleration is about @xmath150 . quite interesting , this value is well within the present experimental technique of atom interferometer @xcite to measure the gravity acceleration . nevertheless , this is a very weak observable effect . thus , it is unlikely to find an evidence to verify or falsify this anomalous acceleration without future experiments . apart from atom interferometer , the measurement of gravity acceleration with bloch oscillation @xcite for cold atoms in optical lattices , superfluid helium interferometry @xcite , free - fall absolute gravimeters @xcite and superconducting gravimeters @xcite provide other methods to test the abnormal quantum gravity effect . in particular , the standard deviation of free - fall absolute gravimeters in the present technique is about @xmath150 @xcite , while the superconducting gravimeters have achieved the sensitivities of one thousandth of one billionth ( @xmath151 ) of the earth surface gravity . because this abnormal quantum gravity effect could be tested by contrast experiment with superfluid helium and normal helium , and the prediction that this abnormal quantum gravity effect is location - dependent , the sensitivities of a gravimeter could be used to test the abnormal quantum gravity effect . this makes it very promising to test the abnormal quantum gravity effect in future experiments . because highly localized helium atoms will not lead to quantum effect of gravity , the measurement of acceleration will give us new chance to measure the fraction of highly localized helium atoms , which is still an important and challenging topic in condensed matter physics . in our theory , the quantum effect of gravity does not rely on superfluid behavior . it depends on whether the wave packet of a particle is localized . it is well known that whether there is wave packet localization is a central topic in condensed matter physics and material physics , such as the long - range order problem at a phase transition . considering the remarkable advances in various gravimeters , it is promising that the quantum gravity effect would have potential applications in our understanding of condensed matter physics and material physics , _ etc_. a possible risk in the decisive test of the abnormal quantum gravity effect with superfluid helium sphere lies in our understanding of the superfluid behavior of liquid helium . in the ordinary understanding of superfluid helium , the superfluid fraction can achieve almost @xmath152 while the condensate fraction is about @xmath153 @xcite . because of the strong interaction between helium atoms , the liquid helium is a very complex strongly correlated system . considering the fact that there are a lot of open questions in strongly correlated systems , we can not absolutely exclude the possibility that the wave packets of all helium atoms are localized , although the whole system still exhibits superfluid behavior . this significantly differs from the bose - einstein condensate in dilute gases , where the wave function of the atoms in the condensate is certainly delocalized in the whole condensate . the abnormal quantum gravity effect gives a possible experimental scheme to test many - world interpretation @xcite . in many - world interpretation , there is no true wave packet collapse process . for a particle described by a wave packet , the measurement result of the particle at a location does not mean that the wave packet of this particle at other locations disappears . it suggests that the wave function of the whole universe evolves into a series of orthogonal many - body wave functions due to the interaction between the measurement apparatus and the particle . the observed result of the particle at a location corresponds to one of the orthogonal many - body wave functions . considering again the superfluid helium sphere , if we increase the temperature so that it becomes normal liquid , based on the many - world interpretation , the wave packets of helium atoms ( at least the fraction of the helium atoms initially in the condensate ) are still delocalized in the whole sphere . in this situation , if the many - world interpretation is correct , it is possible that one may also get the abnormal quantum gravity effect . this would mean a gravity effect dependent on the history of a system . at least , it seems that all previous experiments or astronomical observations do not overrule this possibility . the present work clearly shows that it s time to consider more seriously the new view of gravity , in particularly by future experiments . particles with the same wave function @xmath154 , based on the de broglie - bohm theory.,scaledwidth=80.0% ] in the above studies of the quantum gravity effect , we adopt in a sense the ordinary understanding of quantum mechanics . it is worthy to consider the quantum gravity effect based on other understanding of quantum mechanics . here we give a brief discussion with the de brogliebohm theory . we consider @xmath83 particles with the same wave function @xmath155 shown in fig . 9 . in this pilot - wave theory , the particle is still a highly localized particle with a well - defined trajectory guided by the wave function . we can not predict the exact location of a particle because the initial position of the particle is not controllable by the experimenter , and the motion of the particle guided by the wave function is determined by the initial position of the particle . if we assume further that the energy is mainly carried by the particle , rather than the wave , for the situation shown in fig . 9 , the gravity effect due to these @xmath83 particles would be the same as that of @xmath83 classical particles whose density is @xmath156 . hence , this experimental scheme also gives us chance to test the de broglie - bohm theory . if the abnormal quantum gravity effect is verified , the de broglie - bohm theory would be excluded in a sense . as shown in sec . iv , the gravitational mass and inertial mass are the same mass in einstein s mass - energy relation , when gravity force is regarded as a thermodynamic effect . this explains in a natural way the equivalence principle . for the existence of vacuum excitations at location @xmath157 , a local reference system with acceleration given by eq . ( [ acc ] ) will not experience these vacuum excitations . therefore , in this reference system , the physical law of special relativity would hold . this strongly suggests that einstein s general relativity should be included to consider the quantum gravity effect . because different locations can have different accelerations , different locations have different reference systems satisfying the physical law of special relativity . to construct the connection between the reference systems at different locations , riemannian geometry is the most convenient mathematical tool . therefore , although we try to argue here that thermodynamics and the coupling between matter and vacuum are the essential mechanisms of gravity force , riemannian geometry is still needed to construct a systematic theory , because of the same reason in constructing einstein s general relativity . to clearly introduce the new field equation including the quantum effect of gravity , we first give a brief introduction of einstein s field equation for classical object ( see ref . @xcite ) . in weak field approximation , we have @xmath158 with @xmath159 being the gravitational potential . from @xmath160 and @xmath161 , we have @xmath162 if @xmath163 is due to classical particles , from the attractive gravity force between classical objects originating from the consideration of the free energy , the negative sign in the right - hand side of the above equation is included . we stress again that , in einstein s derivation of his field equation , the negative sign in the right - hand side of the above equation is due to the observed phenomena that the gravity force between two classical objects is attractive , rather than a property derived from fundamental principle . the development of the above equation to relativistic case gives@xmath164 with general considerations of riemannian geometry and general covariance , we get the following einstein s field equation@xmath165 note that we have adopted the unit with @xmath166 and the following minkowski space - time@xmath167 the presence of matter and radiation in the universe will establish various temperature field distributions . these temperature field distributions lead to the force for the matter and radiation in the universe . as shown previously , the physical mechanism of these temperature field distributions is due to the vacuum excitations . in higher - order calculations , the gravity effect of the vacuum excitations themselves should also be considered . although this would be a complex nonlinear coupling process , general conditions ( such as the principle of general covariance ) could be imposed to attack this challenging problem . in a reasonable field equation including the gravitation effect of vacuum excitations themselves , one should introduce an extra energy - momentum tensor @xmath168 , apart from @xmath169 for ordinary matter and radiation . because the vacuum excitations are virtual processes from the view of quantum field theory , these vacuum excitations are different from the ordinary matter . because of this , these vacuum excitations can not be described by the form of the energy - momentum tensor for the ordinary matter or radiation . in a sense , these vacuum excitations are integrated into the space time . fortunately , in principle , one can add an extra term in eq . ( [ efe ] ) without the violation of riemannian geometry and principle of general covariance . the sole form of the field equation is @xcite@xmath170 here@xmath171 this general consideration leads to a request that the energy density in @xmath168 is uniform . assuming @xmath71 is the ground - state energy density of the vacuum , @xmath172 is the energy density of the vacuum in the presence of matter and radiation , we have @xmath173 . here @xmath174 is determined by eq . ( [ unruh ] ) . the role of this spatially - dependent @xmath175 and @xmath176 has been included in einstein field equation ( [ efe ] ) . the energy density of the vacuum excitations is @xmath177 . there are two equivalent ways to understand the average @xmath178 considered here : the average about the time interval @xmath179 which is much larger than the response time of the vacuum , and the average about the space scale @xmath180 which is much larger than the planck length . we see that although @xmath175 is spatially dependent , in principle , @xmath181 could be spatially independent . of course , this is only a statement that there would be no contradiction between spatially dependent @xmath175 and spatially independent @xmath181 . the spatial independence of @xmath181 is due to the general principle of covariance for spacetime itself . the role of @xmath181 is included through @xmath168 in eq . ( [ newefe ] ) . from @xmath182 , we have@xmath183 for a flat universe , assuming the scale factor as @xmath53 in the fiedmann - robertson - walker ( frw ) metric at the present time of our universe , we have@xmath184 considering the fact that various fields in the vacuum have the propagation velocity of @xmath15 , it is not unreasonable to assume that various fields in the vacuum background have delocalized wave packets in the whole observable universe . another reason is that when the big - bang model of the universe is adopted , various fields just have sufficient time to propagate in the whole observable universe . this could lead to delocalized wave packets for the vacuum excitation fields at least within the observable universe . the third reason is that when the coupling between matter and vacuum is considered , the vacuum has the characteristic of superfluidity . these analyses suggest a repulsive gravity effect by nonzero and positive @xmath78 . this requests that one should take the negative sign in the above expression of @xmath168 . we finally get@xmath185 here @xmath186 . we see that vacuum excitations as a whole have positive energy and abnormal negative pressure . in our theory , it is clearly shown that the abnormal negative pressure physically originates from the quantum characteristic of vacuum excitations . in this section , we will try to calculate @xmath78 based on the whole thermal equilibrium in the coupling between matter and vacuum . if the gravity effect is regarded as the thermodynamics and the coupling between matters and vacuum , it is a natural request that the whole vacuum excitations ( dark energy ) should be determined based on the thermodynamic origin of gravity . now we turn to consider the whole average effect of these vacuum excitations for large scale of the universe . we will try to calculate the whole average vacuum excitations , @xmath8 the dark energy density . in calculating the dark energy density , we will use the assumption of isotropy and homogeneity on the large scale of the universe . another assumption is the spatially independent @xmath78 discussed in previous section . using spherical polar coordinates for three - dimensional space , we have@xmath187\ ] ] here @xmath188 . @xmath189 , @xmath190 , @xmath154 are time - independent co - moving coordinates . @xmath191 is the scale factor . @xmath192 , @xmath193 , @xmath194 for spherical , hyperspherical and euclidean cases . in the present work , we consider the case of @xmath195 verified by various astronomical observations . if the evolution of the universe is considered , the radial coordinate @xmath196 of a source that is observed now with redshift @xmath197 is @xcite @xmath198 here @xmath199 , @xmath200 , @xmath201 . @xmath202 , @xmath203 and @xmath204 are the present values of the dark energy density , average cold matter ( @xmath205 dust ) energy density and hot matter ( @xmath205 radiation ) energy density . @xmath206 and @xmath207 are the present values of the scale factor and hubble constant . because the gravity field ( @xmath208 the vacuum excitations ) has the propagation velocity of @xmath15 , for the observer at @xmath209 , the above expression of @xmath189 about @xmath197 has the merit that , @xmath196 is the radial coordinate of a source whose gravity field is observed now with redshift @xmath197 . this gravity field at earlier time is that the observer could experience at the present time . this is the same reason why the above expression is very useful in calculating the luminosity of distant stars . various astronomical observations have given precision measurement of @xmath210 and @xmath211 . assuming @xmath212 , we have@xmath213 from this equation , we can also get @xmath214 . note that the radial coordinate @xmath215 for co - moving sources is time - independent . we will use this sort of radial coordinate to calculate the dark energy density . for an observer at @xmath209 of the present time , the overall energy of cold matter and radiation the observer experiences is then@xmath216 here @xmath217 for cold matter is given by@xmath218 @xmath219 is the cold matter density in the co - moving radial coordinate @xmath215 , without considering the expansion of the universe . when this radial coordinate is adopted , the cold matter density in this coordinate system should not depend on the scale factor . hence , when the present value of the scale factor @xmath220 is takes as @xmath53 , @xmath221 . the existence of the average relative velocity @xmath222 between the observer and the matter at @xmath189 makes the cold matter energy density for the observer becomes @xmath223 . this is the physical origin of the factor @xmath224 in the above expression . because @xmath225 denotes the redshift , we have @xmath226 . the factor @xmath227 in the right - hand side of the above equation originates from two physical effects . when the thermodynamic origin of the gravity is considered , there are various sounds in the temperature field of the vacuum due to the presence of matter . in this situation , for the observer , the rate of arrival of the individual sounds in the gravitation field ( temperature field ) is reduced by the redshift factor @xmath228 . on the other hand , the energy of the individual sounds experienced by the observer is also reduced by the redshift factor @xmath228 . hence , the effective energy density is reduced by the factor @xmath227 in the above equation . there is another way to understand the factor @xmath227 in the above equation . for the matter with the radial coordinate @xmath196 , we consider the gravity field emitted at appropriate time @xmath229 . when this gravity field emitted at time @xmath229 just arrives at the observer , the effective distance between the matter that we consider and the observer becomes @xmath230 because of the expansion of the universe . for the observer , this is equivalent to the case that the energy the observer experiences for the matter at @xmath196 is reduced by the multiplication of the factor @xmath227 . as for @xmath231 , because the radiation field propagates at a speed equal to @xmath15 , we have@xmath232 it is understandable that the vacuum excitations leading to the gravity effect have also the propagation velocity of @xmath15 . in this situation , it is similar to get @xmath233 for the vacuum excitations ( dark energy ) , which is given by@xmath234 assuming that there is a thermal equilibrium between matter and vacuum on the large scale of the universe , we have@xmath235 the above equation is also due to the reason that the smallest space unit ( @xmath236 ) is the same for cold matter , hot matter and vacuum excitation , so that the overall freedom ( @xmath237 with @xmath238 being the volume of the universe ) is the same for cold matter , hot matter and vacuum excitation . it is straightforward to get@xmath239 combined with eq . ( [ eq : radial2 ] ) to get @xmath225 , one can get the ratio @xmath240 . the numerical result is @xmath241 , agrees quantitatively with the result @xmath3 in astronomical observations @xcite . in the co - moving coordinate , the dark energy density is then@xmath242 because the co - moving radial coordinate @xmath196 is time - independent , @xmath203 and @xmath204 can be also regarded as the cold matter density and radiation energy density in the co - moving coordinate . for an observer at other times , in the co - moving coordinate , @xmath203 and @xmath204 are also the cold matter density and radiation energy density for this observer . it is easy to show that the dark energy density is still given by the above equation . together with this time - independent dark energy density in the co - moving coordinate , we see that @xmath243 is a universal value once the following conditions are satisfied . \(i ) there is a big - bang origin of the universe . \(ii ) the universe and its evolution are isotropic and homogeneous . \(iii ) in the co - moving coordinate , the radiation energy density is much smaller than the cold matter energy density . \(iv ) the evolution of the universe is a quasi - equilibrium process . this condition suggests that for cosmic inflation process , the calculation of @xmath240 deserves further studies . one may consider the problem that why we do not introduce the similar concept of luminosity distance @xmath244 , so that the temperature field due to matter and radiation at the location of the observer becomes@xmath245 with the similar idea , the temperature field at the location of the observer due to vacuum excitations is@xmath246 from @xmath247 , we get@xmath248 this method to calculate the dark energy density is wrong . in this method , at the location of the observer , the sum of the vacuum temperature due to the matter and radiation at different locations is adopted . as shown already previously , this sort of sum of the vacuum temperature due to the matter and radiation is nonsense . with the above equation , the numerical result of @xmath240 is about @xmath249 . as expected , it does not agree with the astronomical observations . in eq . ( [ eq : dark ] ) , the dark energy density is given in the time - independent co - moving coordinate . if the evolution of the scale factor @xmath191 is considered , @xmath250 , @xmath251 and @xmath252 in the ordinary coordinate ( where the proper distance @xmath253 is adopted ) need further studies . as shown previously , if the energy - momentum tensor given by eq . ( [ eq : darkenergytensor ] ) is adopted for the dark energy , the whole energy - momentum tensor can be written as@xmath254 here @xmath255 and @xmath256 are the energy - momentum tensors for cold matter and radiation . in the evolution of the universe , when the friedmann - robertson - walker metric is used , the energy conservation law @xmath257 leads to@xmath258 from this we have @xmath259 . if @xmath260 for dark energy as discussed in previous section , this requests that the dark energy density is constant in the evolution of the universe . it is similar to get the well - known results of @xmath261 and @xmath262 . for coexisting cold matter , radiation and dark energy , when there is no interchange of energy between different components , these evolutions for different components always hold . it is consistent with the result of the dark energy density in previous calculations . the above energy conservation law clearly shows a further condition in previous calculations of the dark energy density , @xmath8 the time - independent dark energy density relies on the condition that there is no energy interchange ( or material conversion ) between cold matter and hot matter including radiation . if this energy interchange happens , it is possible that the dark energy density becomes time - dependent . this deserves further theoretic studies for the early non - equilibrium universe , to find observable effect . most quantum field theories predict a huge value for the quantum vacuum . it is generally agreed that this huge value should be decreased @xmath263 times to satisfy the observation result . because there are no true wave functions for these quantum zero - point states , based on our theory , even there are huge vacuum energy due to quantum zero - point states , the gravity effect should be multiplied by zero ! another reason is that , the temperature of the vacuum including only quantum zero - point states is zero . the finite temperature characteristic of the vacuum is due to the excitations from the vacuum , which influence the motion of the matters . based on our theory , the cosmological constant problem is not a true problem at all . in this work , the vacuum energy calculated by us is due to the coupling and thermal equilibrium between matter and vacuum background . the coupling and thermal equilibrium lead to various true excitations from the vacuum . what we calculated in this work is in fact aims at these excitations which can be described by sophisticated and delocalized wave functions . in fig . 10 , we give a summary of the role of different forms of vacuum energy . . because of the whole thermal equilibrium between matter and vacuum , and the request of general covariance on the gravity effect of the vacuum excitations , @xmath74 has a fluctuation around @xmath77 . the fluctuations in @xmath74 lead to the gravity effect in the general relativity for the ordinary matter and radiation , while @xmath78 ( dark energy ) leads to repulsive gravity effect which gives a physical mechanism of the accelerating universe.,scaledwidth=80.0% ] for the sake of completeness , now we consider the evolution of our universe in a brief way . @xmath264 for flat universe is@xmath265 we assume the present value of the scale factor @xmath220 as @xmath53 . if the matter ( including dark matter ) and radiation are regarded as classical when the evolution of the whole universe is studied , @xmath266 in eq . ( [ newefe ] ) takes the ordinary form . from eqs . ( [ newefe ] ) and ( [ frw ] ) , we have @xcite @xmath267,\ ] ] and@xmath268.\ ] ] the quantum gravity effect of vacuum excitations has been included in einstein s field equation , which explains in a simple way the remarkable astronomical observations about dark energy . the basic clues of the combination of quantum gravity effect of matter and general relativity are : \(1 ) the direction of the acceleration field due to gravity force should be determined by the tendency of decreasing the free energy . \(2 ) the field equation should satisfy the principle of general covariance . we first consider further the abnormal quantum gravity effect based on eqs . ( [ acc ] ) and ( [ temfield ] ) . for a particle with mass @xmath6 and wave function @xmath154 , eqs . ( [ acc ] ) and ( [ temfield ] ) have another equivalent form as follows.@xmath269 @xmath270 here @xmath271 is the acceleration field without considering quantum effect of gravity . @xmath272 is the acceleration field when quantum effect of gravity is considered . for a classical particle , @xmath273 . note that @xmath274 . from eq . ( [ aclassical ] ) , we have @xmath275 . in this situation , we get a simple relation between @xmath271 and @xmath272 , which is given by@xmath276 here @xmath277 . the sign @xmath278 at a location @xmath35 is determined by the rule that the direction of the acceleration @xmath279 should point to the increasing of @xmath280 in the neighboring region including @xmath35 . in classical case , this rule physically originating from the property of the free energy has explained why the gravity force between two classical objects is attractive . obviously , it is natural to generalize this rule to quantum wave packet , because classical mechanics has been replaced by quantum mechanics , when fundamental physical law is addressed . in reality , the wave function of a lot of particles may be very complex . for brevity s sake , we consider a system of identical bosons which can be directly generalized to more complex case . the many - body wave function is assumed as@xmath281 the single - particle density matrix is given by@xmath282 this single - particle density matrix can be diagonalized , i.e. written in the form @xcite@xmath283 it is natural to consider the field equation of quantum gravity with this diagonalized single - particle density matrix . to calculate the gravity field of the whole system when quantum effect is sufficiently considered , we first calculate the following field equation for @xmath284 bosons described by the wave function @xmath124 @xmath285 here @xmath286 $ ] can be calculated with the standard method . the calculation based on the above equation does not consider the sign problem . from the following formula@xmath287 we can get the acceleration field @xmath288 . with the rule to solve the sign problem , we get a new field equation@xmath289 here , @xmath290 is determined by the sign rule , and the acceleration field @xmath288 obtained from eqs . ( [ eq1f ] ) and ( [ gamma ] ) . it is clear that the above new field equation still satisfies general covariance . one should note that if at @xmath291 , @xmath292 based on the sign rule . in calculating the above new field equation at @xmath293 , @xmath294 in the whole space time is still the result based on the standard method . more specifically , eq . ( [ newfe ] ) is not equivalent to the following field equation@xmath295 it is very clear that this ` wrong ' field equation is not self - consistent . it is similar to consider @xmath296 bosons described by the wave function @xmath297 . it is straightforward and natural to get the following field equation.@xmath298 here @xmath299 is the ordinary energy - momentum tensor for @xmath300 identical particles described by the wave function @xmath301 . @xmath302 is determined by the sign rule for this mode . for the sake of completeness , @xmath168 is also included in the above equation . in summary , einstein s field equation is extended to include the quantum effect of gravity , from the general assumption that gravity force originates from the coupling and thermal equilibrium between matter and vacuum background . with this new field equation and physical mechanism of gravity force , without any fitting parameter , the accelerating universe of remarkable astronomical observations is quantitatively explained . we stress that the pioneering works on the connection between thermodynamics and gravity force , and the astronomical observations of accelerating universe play key role in the construction of the present theory . in fact , the present work is the result of the direct stimulation of these two works . the present paper is a significant extension of our previous works @xcite , with a lot of physical concepts being clarified and new method added to calculate the dark energy density by including general relativity . the present theory has several significant features . \(1 ) it is a simple extension of einstein s field equation to include quantum effect of gravity . \(2 ) different from newton s gravity law and einstein s general relativity , the present theory gives an interpretation why the gravity force between two classical objects is attractive . at least , the weak equivalence principle also becomes a derivable property . \(3 ) with the same reason to explain the attractive gravity force between two classical objects , the present theory predicts new physical effect , i. e. abnormal quantum gravity effect . \(4 ) the present theory quantitatively agrees with the astronomical observations of accelerating universe without any fitting parameter . \(5 ) if the present theory is verified , it has wide applications , even in condensed matter physics and experimental test of the many - world interpretation and de broglie - bohm theory , etc . \(6 ) last but maybe most important , our theory can be falsified by experiment on earth by measuring gravity acceleration in the interior of a superfluid helium sphere . because of the above features , we believe the present theory should be taken seriously . in particular , these features mean great opportunity to relevant experiments on earth and astronomical observations . one of the main results of the present work is the possible emergence of the quantum effect of gravity at the macroscopic scale . nevertheless , it is an interesting issue to consider whether this macroscopic and abnormal quantum gravity effect could be derived from the microscopic mechanism about the quantum gravity , such as superstring with a positive cosmological constant @xcite , loop quantum gravity and twistor theory @xcite , _ etc_. recently , the relation between gravity and thermodynamics has been studied in the frame of loop quantum gravity @xcite , which provides possible clue to this problem . although the microscopic interaction mechanism of quantum gravity at planck scale is an unsolved problem , the thermodynamics of the macroscopic quantum gravity effect of this work is still meaningful . when statistical mechanics was initiated in 1870 by boltzmann , the concept of atom is still an unrecognized hypothesis , not to mention the collision mechanism between atoms due to electromagnetic interaction . however , this does not influence the power of statistical mechanics in the description of gas dynamics . on the contrary , the theoretical and experimental advances of statistical mechanics greatly promoted further understanding of atoms . once the thermodynamic origin of gravity is verified by future experiments and astronomical observations , the abnormal quantum gravity effect would also greatly promote the understanding of microscopic quantum gravity at planck scale . in this work , the repulsive gravity effect for superfluid helium sphere and dark energy physically originate from the same mechanism that the wave packet of energy is delocalized in the space scale we study . if the de broglie - bohm theory is correct , we would not observe the abnormal gravity effect for superfluid helium . however , for dark energy originating from various vacuum excitations , it is still possible that the energy is carried by various wave packets , rather than particles , because there is no stable particles in the vacuum excitations . hence , even the abnormal gravity effect for superfluid helium is excluded in future experiments , it is still possible that the theory of the present work could be applied to accelerating universe . one may consider the problem that whether there is abnormal quantum gravity effect in the interior of a neutron star . if the temperature ( not the vacuum temperature ) of the neutron star is zero and the wave packets of all neutrons are delocalized in the whole interior of the neutron star , we would expect an abnormal quantum gravity effect . however , the temperature of the neutron state is in fact extremely high . the typical temperature of a neutron star is about @xmath303 kelvins . in addition , the neutron star is not an ideal fermi system without the consideration of the interaction between neutrons . the neutron star comprises very complex structures , such as the out core consisting of neutron - proton fermi liquid and the inner core consisting of possible quark gluon plasma . in quantum statistical mechanics , the approximate model of the neutron star as an ideal fermi system holds because of the validity of the local density approximation . the validity of this ideal fermi system does not mean that the wave packets of all neutrons are delocalized in the whole interior of the neutron star . considering the extremely high temperature and complex structure of the neutron star , it is more reasonable to assume that the delocalized scale of the wave packet of the neutrons is much smaller than the size of the neutron star ( the typical size of a neutron star is about @xmath304 km ) . in this situation , we still expect the normal gravity effect in the interior of the neutron star . we thank the discussions with prof . biao wu , and his great encouragements . we also thank prof . w. vincent liu s great encouragements . this work was supported by national key basic research and development program of china under grant no . 2011cb921503 and nsfc 10875165 . to get eq . ( [ entropy ] ) about the entropy increase of a particle having a displacement , there is another request of the loss of information about the trajectory of the particle . in fig . 11(a ) , we show the motion of a particle along the dashed line . because of the location resolution from the view of the vacuum background , the dashed line is partitioned by the box with side length @xmath10 . at time @xmath305 , the wave packet of the particle is shown by the dashed line . at a later time @xmath306 ( @xmath307 ) , the wave packet of the particle is shown by the solid line . at time @xmath306 , the information of the location of the particle at time @xmath305 recorded by the vacuum background will be lost because of two reasons : ( 1 ) at time @xmath306 , although the location of the particle with spatial resolution @xmath10 is recorded by the vacuum background , the velocity information is highly uncertain . from @xmath308 , the velocity uncertainty of the particle is @xmath309 . in this situation , at time @xmath306 , after the position of the particle is recorded by the vacuum background , the history of the particle is lost from the view of the vacuum background . ( 2 ) at time @xmath305 , although the location with spatial resolution @xmath10 is recorded by the vacuum background , it will be lost rapidly after a displacement at time @xmath306 . at time @xmath305 , the location record is due to the coupling between the particle and the vacuum background . in other words , due to the coupling between the particle and the vacuum background , at the location of the particle at time @xmath305 , the vacuum has larger fluctuations . it s these spatial - dependent vacuum fluctuations record the location of the particle with resolution @xmath10 . because the vacuum background is highly fluctuating , at time @xmath306 , the location information at time @xmath305 will be lost because of the re - establishment of the thermal equilibrium in the vacuum background . \(i ) in the above derivation , we assume a strong coupling between the particle and the vacuum background , so that the location of the particle can be recorded by the vacuum background . when quantum field theory is used , the vacuum background is in fact full of various gauge fields , virtual matter - antimatter pairs . this implies a strong coupling between the particle and the vacuum background . \(ii ) one should distinguish two sorts of coherence lengths : the coherence length @xmath10 and the ordinary thermal de broglie wave length . in the non - relativistic approximation , the thermal de broglie wave length reflects the spatial coherence length which is @xmath311 . we stress that @xmath10 and @xmath310 have different physical origins . @xmath10 originates from the coupling between the particle and the vacuum background . because of the extremely large vacuum zero - point energy and strong coupling between matter and vacuum , the special relativity and quantum mechanics are used to calculate @xmath10 . in contrast to @xmath10 , @xmath310 originates from the coupling or interaction between the particle we study and other particles or environment . to clearly elaborate this , we take a hydrogen atom as an example , which consists of an electron and a proton . for a system consisting of a large number of hydrogen atoms , @xmath312 with @xmath313 being the ordinary temperature in thermal equilibrium . @xmath310 physically originates from the interatomic interaction or the coupling with the environment . for bose - einstein condensation of atomic hydrogen @xcite , @xmath310 can arrive at @xmath314 . in fig . 11(c ) , we show @xmath10 and @xmath310 for an atom . \(iii ) although there is a velocity uncertainty @xmath309 for a particle due to the strong coupling with the vacuum , we stress that this does not mean that the particle will have highly random motion in the vacuum without other forces . the reason lies in that the coupling between a particle and the vacuum will lead a `` dressed '' state including the local vacuum excitations . as a whole , the velocity uncertainty could be much smaller than @xmath15 . this is a little similar to the electron in a hydrogen atom . in the view of proton in the hydrogen atom , the velocity uncertainty of the electron is about @xmath315 with @xmath316 being the electron mass and @xmath317 the size of the hydrogen atom . however , for the hydrogen atom as a whole , the velocity uncertainty is determined by the wave packet of the hydrogen atom , rather than the electron . in a sense , it is possible that there is a physical mechanism of spatially resolved quantum non - demolition detection @xcite about the particle by the vacuum with spatial resolution @xmath10 . see , e.g. , r. g. cai , l. m. cao and n. ohta , phys . d * 81 * , 061501(r ) ( 2010 ) ; t. padmanabhan , phys . rev . d * 81 * , 124040 ( 2010 ) ; f. w. shu and y. g. gong , arxiv:1001.3237 ( 2010 ) ; m. li and y. wang , phys . b * 687 * , 243 ( 2010 ) ; t. w. wang , phys . d * 81 * , 104045 ( 2010 ) ; p. nicolini , phys . d * 82 * , 044030 ( 2010 ) ; y. f. cai and e. n. saridakis , phys . b * 697 * , 280 ( 2011 ) ; r. banerjee and b. r. majhi , phys . d * 81 * , 124006 ( 2010 ) . t. m. niebauer , g. s. sasagawa , j. e. faller and f. klopping , metrologia * 32 * , 159 ( 1995 ) ; g. dagostino , s. desogus , a. germak , c. origlia , d. quagliotti , g. berrino , g. corrado , v. derrico and g. ricciardi , ann . geophys . * 51 * , 39 ( 2008 ) ; s. svitlov , p. maslyk , c. rothleitner , h. hu and l. j. wang , metrologia * 47 * , 677 ( 2010 ) . n. poli , f. y. wang , m. g. tarallo , a. alberti , m. prevedelli and g. m. tino , phys . lett . * 106 * , 038501 ( 2011 ) ; p. clade _ et al_. , europhys . lett . * 71 * , 730 ( 2005 ) ; g. ferrari , n. poli , f. sorrentino and g. m. tino , phys . lett . * 97 * , 060402 ( 2006 ) ; f. sorrentino _ et al_. , phys . a * 79 * , 013409 ( 2009 ) .

observations of two hco ( and ) lines and continuum emission at 1.3 mm towards sgr b2(n ) and sgr b2(m ) have been carried out with the sma . the mosaic maps of sgr b2(n ) and sgr b2(m ) in both continuum and lines show a complex distribution of dust and molecular gas in both clumps and filaments surrounding the compact star formation cores . we have observed a decelerating outflow originated from the sgr b2(m ) core , showing that both the red - shifted and blue - shifted outflow components have a common terminal velocity . this terminal velocity is 58 km s . it provides an excellent method in determination of the systematic velocity of the molecular cloud . the sma observations have also shown that a large fraction of absorption against the two continuum cores is red - shifted with respect to the systematic velocities of sgr b2(n ) and sgr b2(m ) , respectively , suggesting that the majority of the dense molecular gas is flowing into the two major cores where massive stars have been formed . we have solved the radiative transfer in a multi - level system with lvg approximation . the observed hco line intensities and their ratios can be adequately fitted with this model for the most of the gas components . however , the line intensities between the higher energy level transition hco ( ) and the lower energy level transition hco ( ) is reversed in the red - shifted outflow region of sgr b2(m ) , suggesting the presence of inversion in population between the ground levels in the two k ladders ( k= 0 and 2 ) . the possibility of weak maser processes for the hco emission in sgr b2(m ) is discussed .

0801.0348

the giant molecular cloud sgr b2 , located close to the galactic center ( @xmath644 arcmin from sgr a * ) , is a well - known massive star - forming region in our galaxy . sgr b2 consists of an extended envelope , a hot ring and a few compact cores ( e.g. goicoechea , rodriguze - fernandez & cernicharo 2004 ) . the radio continuum and recombination line observations of the compact hii regions suggest that sgr b2(n ) and sgr b2 ( m ) are the two most active star forming cores in this region ( gaume & claussen 1990 ; gaume et al . 1995 , mehringer et al . 1993 ; de pree et al . 1995 , 1996 , 1998 ) . masers , outflows and possible rotation of the two dense cores have been revealed from observations of various molecular lines at centimeter and millimeter wavelengths ( reid et al . 1988 ; gaume & claussen 1990 ; martin - pintado et al . 1990 ; mehringer , goss & palmer 1994 ; lis et al . 1993 ; kuan & snyder 1996 ; liu & snyder 1999 ) . in addition , previous observations have shown evidence for the two hot cores to be at different evolutionary stages and to have different molecular abundances ( e.g. vogel et al . 1987 ; lis et al . 1993 ; miao et al . 1995 ; kuan , mehringer & snyder 1996 ; liu & snyder 1999 ) . h@xmath0co pervades the interstellar medium and it has a simple chemical reaction path which has been proven to be a useful probe of physical conditions ( e.g. mangum & wootten 1993 ) . the h@xmath7co ( 1@xmath8 ) transition at 6 cm was observed in absorption against discrete continuum sources towards sgr b2 complex with an angular resolution of @xmath9 , showing nearly the same radial velocity pattern as that of the radio recombination lines ( martin - pintado et al . 1990 ; mehringer , palmer & goss 1995 ) . these authors suggested that the h@xmath7co ( 1@xmath8 ) transition probably arises from the surrounding gas with a relatively low mean h@xmath7 density of @xmath10 @xmath11 ( martin - pintado et al . 1990 ; mehringer , palmer & goss 1995 ) . the millimeter h@xmath7co lines are an excellent tracer of h@xmath7 density @xmath12 @xmath11 ( e.g. , mangum & wootten 1993 ) . in addition , h@xmath7co is a planar asymmetric top molecule with very little asymmetry . the symmetry of the spin function of the molecule leads to two transition classes : ortho - h@xmath7co levels if the spin wavefunction is symmetric and para - h@xmath7co levels if antisymmetric . since para - h@xmath7co is 1 - 3 times less abundant than ortho - h@xmath7co , observations of para - h@xmath7co have less opacity effect ( kahane et al 1984 ; mangum & wootten 1993 ) . hence , para - h@xmath7co appears to be a better probe to determine the physical conditions of the massive star formation regions . the millimeter / submillimeter transitions of h@xmath7co gas require relatively high excitation temperature and high h@xmath7 density compared to those in the centimeter wavebands . if the brightness temperature of the continuum emission is higher than the excitation temperature , the absorption against the continuum cores can be observed in millimeter and sub - millimeter wavebands with the high angular resolution of an interferometric array ( such as the submillimeter array , hereafter sma ) . taking advantage of the large bandwidth coverage of the sma , we have observed multiple h@xmath7co lines towards sgr b2 at 1.3 mm within a bandwidth of 2 ghz . thus , with the same telescope system and calibration procedure , the uncertainties due to absolute flux density calibration among the different line transitions can be mitigated by measuring the line - intensity ratios , which are needed to determine physical conditions , such as kinetic temperature and h@xmath7 number density , of the gas . in addition , the sma is not sensitive to extended larger scale emission ( @xmath6 50 ) . thus , the sma observations are sensitive to the clumps of high density gas rather than the extended diffuse components . in this paper , we present the results from the sma observations of sgr b2 at the h@xmath7co lines and continuum at 1.3 mm . the paper is organized as follows : 2 discusses the observations and data reduction . in 3 we present the data analysis and results . in 4 , we present the kinematics in sgr b2(m ) by a model incorporating a spherically symmetric inflow along with a decelerating outflow . in 5 we model the physical properties of the h@xmath7co gas in sgr b2 using the large velocity gradient ( lvg ) approach . in 6 , we discuss the important results derived from our observations and analysis . we summarize the results in 7 . we adopt a distance of 8 kpc to sgr b2 . observations towards sgr b2(n ) and sgr b2(m ) were carried out at 218 ( lower sideband ) and 228 ghz ( upper sideband ) with seven antennas of the sma in the compact - north array on august 1 , 2005 for 8 hours . the projected baselines ranged from 5 to 50 k@xmath13 . the weather was good during the observations with @xmath14@xmath60.09 at 225 ghz . the typical system temperature was 130 k. sgr b2(n ) and sgr b2(m ) were observed in separate fields with 15 min observing time on each source interleaving 5 min on the phase reference source sgr a * ( @xmath15 mas , @xmath63 jy at 230 ghz , @xmath645 arcmin from sgr b2 ) . our two target fields were centered at ra(j2000)=17@xmath1647@xmath1719@xmath18882 , dec(j2000)=@xmath1922@xmath2018@xmath2137 and ra(j2000)=17@xmath1647@xmath1720@xmath22156 , dec(j2000)=@xmath1923@xmath2003@xmath2156 , for sgr b2(n ) and sgr b2(m ) , respectively . in addition , callisto ( 4.1 jy ) and the qso 3c454.3 ( 32 jy ) were also observed for the flux - density and bandpass calibrations . the flux density was estimated from callisto with the assumption that its brightness temperature was 120 k and its angular size was 1@xmath2114 . the three transitions h@xmath7co ( 3@xmath23 ) , ( 3@xmath24 ) and ( 3@xmath25 ) were positioned in the lower 2 ghz sideband ( lsb ) . h@xmath7co ( 3@xmath24 ) emission appears to be blended with a transition of ch@xmath26oh and will be not used in the analysis in this paper . the two unblended transitions , h@xmath7co , ( 3@xmath23 ) and ( 3@xmath25 ) , have upper level energies of 21.0 and 67.8 k and rest frequencies of 218.2222 and 218.7601 ghz , respectively . the spectral resolution of 0.8125 mhz corresponds to a velocity resolution of 1.1 km s@xmath4 . the data reduction was carried out in _ miriad_. we recomputed the doppler velocity for each of the target sources since the on - line doppler tracking was only made on sgr b2(n ) . system temperature corrections were applied . antenna - based bandpass ripples were corrected by applying the linear interpolation of the bandpass solutions determined from callisto and 3c454.3 . there are spectral - window - based offsets in amplitude and phase on some baselines due to the correlator errors , which were also corrected . the residual errors due to the bandpass shape of the individual spectral windows were reduced to a level below 1% of the continuum level . in the antenna - based gain corrections , we chose the visibilities of sgr a * in the uv range between 20 kilo wavelengths and longer in order to eliminate the contamination from the extended dust and hii emission . the gains determined from the point source ( emission from sgr a * ) were applied to the sgr b2 data . the molecular lines were identified in the rest frame by use of the jpl catalog , as done by others ( sutton et al . 1985 and nummelin et al . 1998 ) . the h@xmath7co transitions and a few other molecular lines have been unambiguously identified ( see spectrum in fig . the continuum was subtracted with a linear fitting to the spectrum of line - free channels in each baseline . for sgr b2(m ) , the line - free channels can be easily selected from the spectrum in the ( u , v ) domain ( fig . however , for sgr b2(n ) , the molecular spectral lines are crowded over each 2 ghz band and it is difficult to choose the line - free channels . we developed a procedure to select the line - free channels for the continuum subtraction . first , after reducing of the spectral resolution to 1 km s@xmath4 , we made each of the channel maps ( 2600 channels over the 2 ghz band ) including both the line and continuum . from the channel maps , we selected the channels without extended emission ( @xmath274@xmath28 ) . the continuum level was determined from these apparently line - free channels . we selected 16 and 37 line - free spectral windows for the fields of sgr b2(m ) and sgr b2(n ) , respectively . then , using the task uvlin in _ miriad _ , we determined the continuum level by fitting to the line - free channels . the uvlin gives two output ( u , v ) data sets , one for the continuum and the other for continuum - free spectral line . self - calibration was performed to the continuum data using the compact sources of the continuum cores for a few iterations in order to eliminate residual errors . the gain solutions determined from the continuum data were applied to the line data , from which we constructed the spectral data cube . the preliminary images of the continuum and lines were made using natural weighting . the clean algorithm was applied to remove the effects of the side - lobes . the synthesized beam sizes of the continuum and line images were approximately 5@xmath29 ( pa=12.5@xmath30 ) . the mosaic maps of the two fields , sgr b2(n ) and sgr b2(m ) , were made using a simple linear mosaic algorithm . the primary beam attenuation was also corrected in the final line cubes and continuum image . the statistical 1@xmath28 rms noise of the continuum was 0.1 jy beam@xmath4 . the 1@xmath28 rms noise levels of the line images were 0.18 and 0.16 jy beam@xmath4 per channel for the h@xmath7co ( 3@xmath23 ) and ( 3@xmath25 ) lines , respectively . the higher noise in the h@xmath7co ( 3@xmath23 ) line images was due to its strong line intensity and the limit of the dynamic range in the clean process . 2 shows the mosaic continuum map of sgr b2(n ) and sgr b2(m ) at 1.3 mm . the brightest components of sgr b2(n ) ( i@xmath31=29.2@xmath32.1 jy beam@xmath4 ) and sgr b2(m ) ( i@xmath31=20.2@xmath31.3 jy beam@xmath4 ) , are associated with the massive star forming cores k1 - 3 and f1 - 4 ( gaume & claussen 1990 ) , respectively . in addition to the emission from these cores , a few nearby continuum clumps were detected , including the components k4 ( i@xmath32=1.40@xmath30.12 jy beam@xmath4 ) , ne ( i@xmath31=1.20@xmath30.14 jy beam@xmath4 ) , nw ( i@xmath32=1.00@xmath30.11 jy beam@xmath4 ) , me ( i@xmath32=0.58@xmath30.15 jy beam@xmath4 ) and mw ( i@xmath32=1.70@xmath30.18 jy beam@xmath4 ) . gaussian fitting to the individual continuum components was carried out . the individual emission clumps near the compact cores or the k1 - 3 and f1 - 4 clusters were modeled as simple gaussian components . the two compact cores , k1 - 3 and f1 - 4 , appeared to be too complicated to be fitted with single gaussian components . a model consisting of a gaussian and two unresolved compact components was used to fit to the data of the core ( k1 - 3 ) . the peaks of the gaussian component and one of the point components are consistent with k3 and k2 positions , respectively . another point component agrees ( within @xmath6 1@xmath33 ) with the 3.5 mm continuum source that is located south of k3 ( see fig.1(b ) of liu & snyder 1999 ) . sgr b2(m ) core ( f1 - 4 ) is fitted well with a gaussian component and a point component ( close to f3 within @xmath6 0@xmath215 ) . the peak positions , deconvolved angular sizes , peak intensities and total flux densities of the continuum components are summarized in table 1 . the components k4 , mw and z10.24 have been detected at radio and millimeter wavelengths ( lis et al . 1993 ; kuan & snyder 1994 ; gaume et al . 1995 ; liu & snyder 1999 ) . z10.24 located in the middle between sgr b2(n ) and sgr b2(m ) ( see fig . 2 ) shows a unique filamentary structure at 1.3 mm continuum . the designation of z10.24 follows that used by gaume et al . ( 1995 ) and de pree et al . ( 1996 ) who detected the h66@xmath34 line towards it . z10.24 was marginally detected at 1.3 mm by lis et al . ( 1993 ) with higher angular resolution ( 4@xmath35 ) and poorer sensitivity . the sma observations show an elongated structure in z10.24 with a peak intensity of 2.02 @xmath3 0.16 jy beam@xmath4 ( 12 @xmath28 ) . the core of z10.24 is unresolved in the sub - arcsec resolution images at both 1.3 cm and 3.5 mm ( gaume et al . 1995 ; liu & snyder 1999 ) . the 1.3 mm continuum image and the detected vibrational hc@xmath26n emission ( de vicente et al . 2000 ) suggest that the 1.3 mm continuum of z10.24 is dominated by the dust emission and z10.24 is likely to be a younger massive star formation region . there have been no detections of the components nw , ne and me in the previous observations at longer wavelengths . these three components are possibly the dust emission from sub - cores at a relatively early stage of star formation . the detections need to be verified with further observations at shorter wavelengths and at higher angular resolutions . the continuum - free channel maps in both h@xmath7co ( 3@xmath23 ) and h@xmath7co ( 3@xmath25 ) lines were constructed in the velocity range from 8 to 151 km s@xmath4 at intervals of 1 km s@xmath4 . the channel maps of the h@xmath7co transitions are complicated , containing several kinematical features in either emission or absorption . those emission and absorption components are separated well in our higher spectral resolution maps but some of them ( in the continuum core regions ) are overlapped . in the moment analysis , the negative intensity value from the absorption and the positive value from the emission may cancel each other in the overlapping regions and thus the resultant moment maps might not reflect the true gas distribution . hence , the emission and absorption need to be handled separately . fig . 3 is the integrated line emission images constructed from the channel maps for the two transitions , h@xmath7co ( @xmath2 ) and ( @xmath1 ) , respectively . the moment 0 images were made with a 4 @xmath28 cutoff in each channel maps . the less significant emission and the absorption are excluded in the moment analysis . most of emission is distributed around the two cores of sgr b2(n ) and sgr b2(m ) . clearly , the distribution of the h@xmath7co emission is not spherically symmetric with respect to each of these cores . in the sgr b2(n ) region , in addition to the gas concentration at the core , gas clumps located north - east and south - west of the core are observed in both h@xmath7co transitions . we note that in sgr b2(n ) , the morphology of h@xmath7co ( @xmath2 ) emission , the higher transition gas , is similar to that of the emission from the lower transition gas h@xmath7co ( @xmath1 ) . hereafter we refer h@xmath7co ( @xmath2 ) as the higher transition and h@xmath7co ( @xmath1 ) as the lower transition . in the sgr b2(m ) region , a strong emission component elongated in northwest - southeast direction is observed in both h@xmath7co transitions . an arch structure ( m1 30@xmath33 long and 10@xmath33 wide ) , 8@xmath33 northwest of the sgr b2(m ) core appears in both h@xmath7co transitions . the major difference in the emission distribution from the two transitions occurs in the outflow ( lis et al . 1993 ) region located south - east of the sgr b2(m ) core . a significant emission `` tongue '' ( m5 ) ( 15@xmath36 ) was detected from the higher h@xmath7co transition gas while no significant detection was made of the lower h@xmath7co transition gas . based on a line survey at 340 ghz from single dish observations , sutton et al . ( 1991 ) obtained mean systematic velocities 65 and 61 km s@xmath4 of sgr b2(n ) and sgr b2(m ) . their observations showed significant velocity variations among the different species . the differences are mostly caused by the chemical differences of the molecules . the different species sample different physical environments . the high angular resolution observations of h66@xmath34 ( de pree et al . 1995 ; 1996 ) showed that the mean systematic velocities are 69.8 and 65.3 km s@xmath4 for the sgr b2(m)-f and sgr b2(n)-k clusters , respectively . the relatively higher mean velocity of 69.8 km s@xmath4 in sgr b2(m ) is likely caused by the high velocity motion of the ionized gas of the uchii regions with respect to the centroid of the system . in section 4 , we will show that the systematic velocity determined from the terminal velocity of the outflow in sgr b2(m ) is 58 km s@xmath4 ( see also appendix a ) . in the rest of the paper , we adopt systematic velocities of 58 km s@xmath4 and 65 km s@xmath4 for sgr b2(m ) and sgr b2(n ) , respectively . the strong continuum cores are excellent probes of absorption by the cold gas that resides in front of them . absorption is observed towards both sgr b2(n ) and sgr b2(m ) continuum cores ( see the top - left and bottom - left panels of fig . multiple gaussian line components were fitted to the spectra for each of the two transitions in both cores . the parameters of these fits are listed in table 2 . the systematic velocities are marked with the vertical lines in fig . the majority of the absorbing gas in sgr b2(n ) and sgr b2(m ) is red - shifted with respect to the systematic velocities 58 and 65 km s@xmath4 , respectively . the red - shifted absorption gas provides evidence for the existence of gas accreting onto the two cores . if the absorbing gas covers the continuum source completely and the emission from the gas is insignificant , the line intensity is @xmath37 . the optical depth ( @xmath38 ) can be derived from the formula @xmath39 where @xmath40 is the observed line intensity and the @xmath41 is the observed continuum intensity . the errors in @xmath42 based on the fractional errors @xmath43 of each line channel and @xmath44 if the line is optically thin ( @xmath45 ) are given by @xmath46 in the optically thick case , the channels in the line center are saturated and the line - to - continuum ratio only gives a lower limit to the optical depth of the line . for example if @xmath47 spectral channels are saturated by the absorption gas component at a velocity , we have @xmath48 in 3 @xmath49 . substituting this formula into the equation ( 1 ) , the lower limit of the optical depth is @xmath50 with the rayleigh - jeans approximation , the value of 1 jy beam@xmath4 in our sma observations is equivalent to 1.5 k. the observed peak continuum intensities of the sgr b2(n ) and sgr b2(m ) cores are 29.2 and 20.2 jy beam@xmath4 , which correspond to brightness temperature of 44 and 30 k in our observations , respectively . the solution of radiation transfer function in terms of the observed brightness temperature of the line ( @xmath51 ) is given by @xmath52 where @xmath53 is the excitation temperature of the molecular line and @xmath54 is the observed brightness temperature of the continuum emission with the true brightness temperature of @xmath55 ; @xmath56 is the optical depth of the molecular cloud ; for given solid angles of the molecular cloud ( @xmath57 ) , continuum source ( @xmath58 ) and the telescope beam ( @xmath59 ) , @xmath60 and @xmath61 are the beam filling factors of the line and continuum , respectively , if both the source and the telescope beam are in gaussian shape ; @xmath62 denotes the fraction of the continuum source covered by the molecular cloud . if the molecular cloud is in front of the larger continuum source ( @xmath63 ) , the observed brightness temperature of the line ( @xmath64 ) becomes : @xmath65 if the molecular clouds cover the continuum cores with the same beam filling factor ( @xmath66 ) , the upper limits of the excitation temperatures h@xmath7co absorbing gas ( @xmath67 ) are imposed by the observed continuum brightness temperatures of the continuum cores , i.e. 44/@xmath68 and 31/@xmath68 k , for the sgr b2(n ) and sgr b2(m ) cores . for a given filling factor @xmath69 , the excitation temperatures of h@xmath7co would be less than 150 and 103 k for sgr b2(n ) and sgr b2(m ) , respectively . fig . 4 shows the spectra of the h@xmath7co lines ( averaged over one beam ) made from the image cube for the rest of components . each of spectral panels includes both the lower ( 3@xmath23 , the green profile ) and the higher ( 3@xmath25 , the red profile ) transition lines . gaussian fits to the h@xmath7co spectra were carried out for both absorption and emission components . the emission and absorption probably come from the different regions along the line of sight . the angular resolution of our observations appears to be inadequate to distinguish the discrete components in the cores . however , our high spectral resolution is adequate to separate the emission from the absorption in the gaussian fits . the parameters derived from the gaussian fits to the two h@xmath7co transitions are summarized in table 2 including the central line velocity ( v@xmath70 ) , the full width of half maximum ( @xmath71v ) and peak intensity ( i@xmath32 ) . the m1 , m4 and m5 are located along the major axis of the bipolar outflow originated from the f - cluster ( lis et al . m1 is on the blue - shifted side of the outflow . both the high- and low - transition spectra can be fitted with two gaussian components at 52 km s@xmath4 in emission and 64 km s@xmath4 in absorption . the weak red - shifted absorption with respect to the systematic velocity suggests that a relatively cold gas component in front of the continuum source are moving towards it . the strong blue - shifted emission with respect to the systematic velocity is the highly - excited gas emission in the outflow ( in front of the continuum source ) likely mixed with the infall gas emission ( behind the continuum source ) . sgr b2(m)-m4 is located close to the hii regions , _ i.e. _ , the f1 - 4 cluster . the higher transition spectrum can be fitted with two gaussian components in emission at 66 and 70 km s@xmath4 , both of which are red - shifted with respect to the systematic velocity 58 km s@xmath4 . the lower transition spectrum can be fitted with five gaussian components at velocity 51 , 65 , 70 , 75 km s@xmath4 in absorption and 100 km s@xmath4 in emission . the morphology of this region is a complex . a possible model to interpret the spectral characteristics of m4 is considered here . if we assume a non - lte condition for the gas and that both higher and lower transition gas comes from the same gas clump located in front of the continuum core , the excitation temperature of the higher transition gas is larger than the brightness temperature of the continuum , while the excitation temperature of the lower transition gas is less than the brightness temperature of the continuum . the red - shifted absorption suggests that the gas flows towards the continuum core . the nature of absorption and emission at m4 suggests that the process of excitation of the infall molecular cloud is complicated . the radiative excitation by the strong fir radiation field might play an important role in the region near the core since the collision alone can not produce the observed line ratio or the inverse population between the higher transition in k@xmath5=2 and the lower transition in k@xmath5=0 based on the lvg fitting ( see appendix b ) . the inversion in population between the two lowest k ladders is observed best in the red - shifted outflow component . at m5 , the spectrum of the higher transition gas shows a very significant line emission ( 7 - 8 @xmath28 ) , fitted to a gaussian at 66 km s@xmath4 with a line width of 16 km s@xmath4 , while less significant emission ( 1 - 2 @xmath28 ) of the lower transition gas is shown at the same position . the highly reversed line ratio @xmath72 suggests that local thermodynamic equilibrium ( lte ) is not valid at this location and a weak maser process is active in the outflow region . the gas components m2 and m3 and mw show significant line emission ( @xmath73 @xmath28 ) from the lower transition gas while the higher transition emission is relatively weak . the line ratio @xmath74 in those isolated component varies in the range between 1.5 to 5.2 . maser at 6 cm ( mehringer , goss & palmer 1994 ) was detected in the z10.24 region . the sma spectrum of the lower transition at z10.24 shows that a significant amount of gas is in absorption while the emission is present but red - shifted with respect to the absorption feature , a typical p - cygni profile suggesting an outflow in this region . the spectrum of the higher h@xmath7co transition can be fitted with two emission gaussian components at 50 and 84 km s@xmath4 with an absorbing gaussian at 75 km s@xmath4 , which is consistent with the h66@xmath34 transition ( de pree , et al . 1996 ) . excluding the possibility of the expanding shell model , de pree et al . argued that an ionized outflow is likely centered at z10.24 . our h@xmath7co observations appear to favor their argument of a bipolar outflow from the uchii region . towards the ne continuum source , both the higher and lower h@xmath7co transitions show that the majority of the gas in absorption is red - shifted with respect to the mean systematic velocity 65 km s@xmath4 . towards nw , a broad ( @xmath75 km s@xmath4 ) absorption ( @xmath76 jy beam@xmath4 ) from the lower transition line is detected at 71 km s@xmath4 . the spectrum of the higher transition shows no significant lines in either emission or absorption . 5 and 6 show the images of intensity weighted velocity ( or moment 1 ) of the h@xmath7co emission gas from both transitions . the moment 1 maps were constructed with a cutoff of 8 @xmath28 from each of the channel images in the velocity range of 8 - 151 km s@xmath4 . in the sgr b2(m ) region , the kinematical structure observed from the lower transition ( see fig . 5 ) consists of the highly red - shifted components 5@xmath33 south - east of the compact core and a northeast - southwest arch structure . the component m4 appears to be a fast moving compact component ( v@xmath77 km s@xmath4 ) ejected from the core . the morphology of the northeast - southwest arch in sgr b2(m ) from the velocity field in the higher transition gas ( see fig . 6 ) appears to be consistent with that observed in the lower transitions . a velocity gradient is present south - east of m4 ( shown in fig . 6 ) , which appears to indicate a decelerating outflow . in the case of sgr b2(m ) , the higher h@xmath7co transition appears to trace outflow well , which is consistent with the interpretation of the larger scale mass outflow based on the lower angular - resolution observations of nh@xmath26 and so ( vogel , genzel & palmer 1987 ) . the higher angular - resolution observations of nh@xmath26 ( guame & claussen 1990 ) showed that the red - shifted emission is located south of the f3 hii region , and the blue - shifted absorption is located north of the red - shifted emission , showing a velocity gradient in north - south direction . based on their higher angular - resolution observations of oh maser and nh@xmath26 , guame & claussen ( 1990 ) suggested that the north - south velocity gradient can be explained by a rotating disk or a torus of material with an extent of 2@xmath215 surrounding the sgr b2(m)-f cluster . although the angular resolution in our observations is not adequate to verify the kinematical model proposed by guame & claussen ( 1990 ) , our observed arch - liked morphology of the h@xmath0co gas shown in blue - shifted emission from northeast to southwest in the larger scale also indicates that the gas is undergoing a complicated infall process interacting with the outflow while the gas is spiraling onto the core rather than being in simple free fall . in the sgr b2(n ) region , the lower transition map ( see fig . 5 ) shows the red - shifted velocity located north of k2 and blue - shifted velocity located south of k2 with a north - south velocity gradient across k2 . a similar velocity gradient has been observed in the hc@xmath26n line ( lis et al . these authors argued that the north - south velocity gradient in sgr b2(n ) traces rotation . the kinematics of the lower h@xmath7co transition in emission gas of sgr b2(n ) is likely dominated by the gas rotating around the core . in addition to the north - south velocity gradient from the higher transition h@xmath7co ( @xmath2 ) map as observed in the lower transition gas , an east - west velocity gradient is also observed in sgr b2(n ) ( see fig . the outflow in sgr b2(n ) was observed in east - west direction ( lis et al . 1993 ) . the higher angular - resolution observations of nh@xmath26 ( gaume & claussen 1990 ) appeared to show a southeast - northwest velocity gradient across k2 . the southeast - northwest velocity gradient in the larger scale from our observations appears to be consistent with the kinematical structure observed in nh@xmath26 . based on the higher angular - resolution observations of nh@xmath26 , ruling out a simple outflow / rotation model , gaume & claussen ( 1990 ) suggested that several kinematic components in outflow , infall and rotation might be involved in the sgr b2(n ) core . the southeast - northwest velocity gradient observed from the h@xmath7co emission gas in sgr b2(n ) appears to be caused by a combination of rotation , infall and outflow . the angular resolution of our observations is not adequate to discern the details of these motions . in order to better understand the on - going astrophysical processes in the star formation cores , we modeled the observed kinematics and the ratio of the line intensities . in comparison to sgr b2(n ) , the kinematics observed from sgr b2(m ) appears to be relatively simple and characterized by infall and outflow . 7 shows the single field map of h@xmath0co ( @xmath2 ) centered at the f - cluster . the major axis of the red - shifted outflow ( pa=158@xmath30 ) can be drawn by connecting the continuum core and the tips of the outflows ( m1 and m5 ) . along the major axis of outflow ( pa=158@xmath30 ) , we have made a position - velocity diagram ( pv ) from the higher h@xmath7co transition line cube ( see fig . 8) . the solid contours show the emission and dashed contours indicate the absorption . this diagram shows two distinct decelerating outflow components clearly . the red - shifted component shows that the emission near the core ( position at 0 in the vertical axis ) starts with a high velocity of 106 km s@xmath4 and the velocity declines to a terminal velocity at 58 km s@xmath4 as the gas goes to the outer region ( 20away from the core ) . if we shifted velocity to the terminal velocity 58 km s@xmath4 or in the system rest frame , the decelerating red - shifted outflow is mainly located in the first quadrant with the absolute velocity decreasing from @xmath648 to 0 km s@xmath4 . the third quadrant shows the decelerating blue - shifted outflow component with deceleration from @xmath650 to 0 km s@xmath4 . appendix a describes the numerical calculations that have been carried out to model the kinematical characteristics of a decelerating outflow combined with spherically infall . with a simple assumption of mass conservation and power law distributions in both outflow velocity and molecular density , the observed kinematics of the outflow components can be well fitted to the decelerating outflow model with an initial velocity of v@xmath78=83 km s@xmath4 at r@xmath78=1 ( 0.04 pc ) from the core center with the outflow opening angle of @xmath7930@xmath80 and inclination of @xmath8145@xmath80 for the red - shifted component ( the thick red curve in fig . the blue - shifted component can be fitted with v@xmath78=33 km s@xmath4 at r@xmath78=1 ( 0.04 pc ) , @xmath7930@xmath80 and @xmath8145@xmath80 ( the thick blue curve ) . the dashed curves range initial velocity v@xmath78 at r@xmath78=1@xmath33 given the same opening angle and inclination angle . for the red - shifted outflow , v@xmath78 is from 250 km s@xmath4 to 17 km s@xmath4 . the blue - shifted velocity ranges from 83 km s@xmath4 to 4 km s@xmath4 . our observations and analysis have shown the presence of a decelerating outflow from the core in sgr b2(m ) . such a decelerating outflow can be caused by the entrainment of gas through the interaction between a fast wind flow from the central stars and an ambient cloud core , where the incorporation of additional mass into the outflow decelerates the wind ( lizano et al . 1988 ) . as the wind moves farther from the cores , the outflow velocity terminates at the systematic velocity 58 km s@xmath4 of the molecular cloud . we note that both the red - shifted and blue - shifted outflows terminate at a common velocity , which is close to the mean systematic velocity 61 km s@xmath4 of sgr b2(m ) ( sutton et al . in fact , the common terminal velocity 58@xmath32 km s@xmath4 of blue - shifted and red - shifted outflow appears to provide an unambiguous method , independent of the chemical processes in molecular clouds , for accurately measuring the systematic velocity of sgr b2(m ) . the uncertainty in the terminal velocity is mainly due to the velocity resolution in the pv diagram and the uncertainty in the model fitting process . finally , the absorption located in the second quadrant is shown to be red - shifted with respect to the systematic velocity 58@xmath32 km s@xmath4 . thus , with the accurately determined systematic velocity we can be certain that this absorption feature arises from gas flowing towards the continuum source . based on the line ratio of the two transitions from the spectra , most of the gas in the sgr b2 region does not satisfy the lte condition . under the assumption that the molecular cloud has spherical geometry , the non - lte calculation requiring the collisional excitation rates for h@xmath0co ( green 1991 ) has been carried out for sgr b2 using the large velocity gradient ( lvg ) approximation ( e.g. evans , davis & plambeck 1979 ; scoville & solomon 1974 ; lucy 1971 ) . the radiative transfer calculations and model fitting are discussed in details in appendix b. the fitting results for the line components for models with various sets of parameters are given in table 3 . we summarize the excitation properties for the various @xmath82 components in the following : \1 . towards the absorption components of the cores ( k1 - 3 , f1 - 4 ) , the kinetic temperature depends on the beam filling factor of the continuum emission , ranging from a few tens k for a large continuum beam @xmath83 ( 0.3 ) to a few hundreds k for a small @xmath83 ( 0.05 ) . high angular resolution observations are necessary to precisely determine the range of t@xmath84 . the typical values of @xmath85 are 10@xmath86 @xmath11 for the core components . the derived column densities @xmath87 are in the range of 10@xmath88 @xmath89 km@xmath4 s. \2 . for the discrete @xmath90 components , no lvg solutions could be fitted to the observations with either model a or model b in which @xmath83 is large and the brightness temperature of the background radiation is too small to produce significant absorption lines as observed . for excitation temperature of a few tens k and higher , the intensity of absorption ( see equation(5 ) ) suggests that the beam filling factor must be small , _ i.e. _ @xmath910.1 or smaller , for a region with a continuum peak intensity of @xmath62 jy beam@xmath4 in our observations . the kinetic temperatures derived from the lvg fitting are all below 100 k for both emission and absorption components except for m3 emission component in which t@xmath92 k. in general the components outside the cores are cooler . the typical values of @xmath85 are 10@xmath93 @xmath11 for the discrete components . the derived column density @xmath94 is in the range of 10@xmath95 @xmath89 km@xmath4 s. \3 . for the components in the red - shifted outflow of sgr b2(m ) , the line ratio @xmath96 is significantly smaller than unity , suggesting that the population of the ground levels in the two k ladders ( @xmath97 and 2 ) is inverted . hence , a weak maser process between the two different k ladders must occur in this region . based on our radiative transfer code with the collisional excitation rates for kinetic temperature t@xmath84@xmath98 300 k ( green 1991 ) , we can not rule out the possibility that the h@xmath0co ( @xmath2 ) line in sgr b2(m ) outflow is excited in the high temperature c - shocks . on the other hand , the strong fir radiation field near the sgr b2(m ) core might also play a role in the population inversion of the molecule . the assessment of molecular cloud mass from molecular lines can be affected by the excitation , opacity , abundance variations and gas dynamics of the molecular lines . the optically thin submillmeter dust continuum emission has been proven to be a good tracer of molecular cloud mass ( pierce - price et al . 2000 , gordon 1995 ) . if we take an average grain radius of 0.1 @xmath99 m and grain density of 3 g @xmath11 and a gas to dust ratio of 100 ( hildebrand 1983 , lis , carlstrom & keene 1991 ) , the dusty cloud mass and column density are given by the formulae ( lis , carlstrom & keene 1991 ) @xmath100 @xmath101 where @xmath102 is the mean dust temperature ( k ) , @xmath103 is grain emissivity at frequency @xmath104 , @xmath105 is the flux density corrected for free - free emission , @xmath106 is the solid angle subtended by the source . assuming @xmath103 at 1.3 mm is 2@xmath10710@xmath108 and the dust temperature is 150 k for sgr b2 ( carlstrom & vogel 1989 ; lis et al . 1993 ; kuan , mehringer & snyder 1996 ) , we derived the masses , h@xmath7 column densities and number densities . because the continuum at 1.3 mm contains free - free emission , we estimate the physical parameters using the flux densities of the continuum corrected for free - free emission . assuming 3.6 cm continuum emission with fwhm beam of 3@xmath109 ( mehringer et al . 1993 ) of sgr b2(n ) ( k1 - 3 ) and sgr b2(m ) ( f1 - 4 ) come from optically thin free - free emission ( @xmath110 ) , we estimate the free - free contribution of 4.7 and 8.4 jy ( @xmath6 9% and 24% of the total flux densities at 1.3 mm ) at 1.3 mm towards the continuum cores k1 - 3 and f1 - 4 in our observations . our estimates are consistent with the determinations of lis et al . ( 1993 ) ( 6% and 33% of the total flux densities at 1.3 mm ) and martin - pintado et al . ( 1990 ) ( @xmath6 9% and 28% of the total flux densities at 1.3 mm ) for k1 - 3 and f1 - 4 . the continuum flux densities at 1.3 mm corrected for free - free emission are 47.4 and 27.2 jy for k1 - 3 and f1 - 4 , respectively . from the flux densities at 1.3 cm ( gaume et al . 1995 ) , the estimated free - free emission contributions at 1.3 mm are 0.02 jy , 0.04 jy and 1.26 jy for k4 , z10.24 and mw , respectively . the corresponding continuum flux densities corrected for free - free emission are 3.5 , 9.1 and 11.0 jy , respectively . the derived h@xmath7 masses , column densities and number densities are summarized in table 4 . the estimated h@xmath7 masses of the sgr b2(n ) core ( k1 - 3 ) and the sgr b2(m ) core ( f1 - 4 ) are larger than those given by lis et al . ( 1993 ) , while the h@xmath7 number densities are less than their results . this result is caused by the relatively larger size and higher flux densities of the continuum in our observations . the h@xmath7co ( @xmath1 ) spectra show absorption against both sgr b2(n ) and sgr b2(m ) compact cores and multiple absorbing peaks . the absorptions are dominated by red - shifted gas , suggesting that the lower transition h@xmath7co ( @xmath1 ) traces the cold gas in front of the continuum cores falling into the two compact cores . previous observations showed multiple massive young stars in sgr b2(n ) and sgr b2(m ) ( e.g. , gaume et al . 1995 , de pree , goss & gaume 1998 ) . there are multiple absorbing peaks with different optical depths from our h@xmath7co spectra in the sgr b2(n ) and sgr b2(m ) cores , which appears to indicate that the gas is falling into the massive stars or massive star forming cores embedded at different depths in the molecular clouds ( mehringer , palmer & goss 1995 ) . however , the angular resolution of our observations is inadequate for us to determine whether there are multiple regions present or whether the overall gravitation potential dominates the infalling gas . if infalling gas is in simple free - fall , the infalling velocities can be estimated by @xmath111 where @xmath112 is the infall radius , @xmath113 is the sum of the gas and star masses included in the @xmath112 and @xmath114 is the gravitation constant . the h@xmath7 masses derived from the continuum are 1.4@xmath10710@xmath115 and 7.9@xmath10710@xmath116 m@xmath117 for the cores of sgr b2(n ) and sgr b2(m ) , respectively . the vla observations of radio continuum at 1.3 cm ( gaume et al . 1995 ) showed that there are three uchii regions ( k1 , k2 and k3 ) in the core of sgr b2(n ) and four uchii regions ( f1 , f2 , f3 and f4 ) in the core of sgr b2(m ) . by use of the relationships between stellar spectral type and stellar mass ( vacca , garmany & shull 1996 ) , a total stellar mass of 68 m@xmath117 was inferred for the massive stars in the core of sgr b2(n ) . the higher resolution observations ( 0@xmath2105 ) at 7 mm ( de pree , goss & gaume 1998 ) resolved out f1 , f2 , f3 and f4 into twenty - one uchii regions , and a stellar mass of 443 m@xmath117 was inferred corresponding to the massive stars in the core of sgr b2(m ) . the mass of the massive stars in the sgr b2(m ) core is six times larger than that in the sgr b2(n ) core . taking the major axis sizes of 0.24 and 0.29 pc as the infall radii of the two cores ( k1 - 3 and f1 - 4 ) , we inferred the infall velocities of 21 and 15 km s@xmath4 for sgr b2(n ) and sgr b2(m ) , respectively . hence , based on our sma observations , we have shown that high - density molecular gas is continuously feeding onto the active star formation cores in both sgr b2(m ) and sgr b2(n ) . the continuum emission and h@xmath7co spectral lines at 1.3 mm were observed with the sma . we detected the continuum emission from the compact cores of sgr b2(n ) and sgr b2(m ) . outside of the two compact cores , a few clumps , k4 , ne , nw , me , mw and the filament of z10.24 were also observed . except for the three newly observed objects , ne , nw and me , all of the other compact cores , clumps and the filament have radio counterparts at centimeter wavelengths . the spectra in the two h@xmath7co lines towards the continuum and outside of the continuum showed the gas components in either emission or absorption . from the integrated line intensity and intensity weighted velocity maps in line emission , we identified possible outflow motion in sgr b2(n ) and sgr b2(m ) cores . the infalling gas was detected by the red - shifted absorbing gas against the continuum in sgr b2(n ) and sgr b2(m ) compact cores and the nearby dust clumps and filament z10.24 . the absorbing gas in sgr b2(n ) and sgr b2(m ) cores is red - shifted with respect to the systematic velocities , suggesting that the gas in front of the continuum cores is flowing onto the continuum cores . in comparison to sgr b2(n ) , the kinematics in sgr b2(m ) are relatively simple . we interpreted the observed outflow and infall in sgr b2(m ) by a model incorporating a spherically symmetric inflow with a decelerating outflow . a decelerating outflow from sgr b2(m ) was evident in our observations . we showed that both the red - shifted and blue - shifted outflow components share a common terminal velocity . from the outflow terminal velocity , we determined the systematic velocity of 58@xmath32 km s@xmath4 for sgr b2(m ) . with the systematic velocity well determined in a manner independent of chemistry in the molecular cloud , we are confident that the majority of the absorbing gas at least in sgr b2(m ) is red - shifted and flowing towards the active star forming core . using the observed two h@xmath7co lines , we have derived the excitation conditions of sgr b2 region in lvg approximation . for the absorbing components in the sgr b2(n ) and sgr b2(m ) cores , the kinetic temperature , the h@xmath0 density and column density are in the range of several tens to hundreds k , 10@xmath86 @xmath11 , 10@xmath88 @xmath89 km@xmath4 s , respectively . the components outside of the cores are relatively cooler , and the h@xmath0 density and column density are one order of magnitude less than those in the sgr b2(n ) and sgr b2(m ) cores . in the sgr b2(m ) outflow region , the intensity of the h@xmath7co ( 3@xmath25 ) is larger than that of the h@xmath7co ( 3@xmath23 ) , suggesting the inversion of the distribution of the particle number between the two ground levels of k@xmath5 = 0 and 2 . the h@xmath0co ( @xmath2 ) might be excited by c - shocks in the outflow . the radiative excitation by the strong fir radiation field in sgr b2(m ) may play an important role in the population inversion . we thank the sma staff for making the observations possible . wang acknowledges support from national natural science foundation of china ( no . 10328306 ) . in this appendix , we discuss a model incorporating a spherically symmetric inflow ( shu 1977 ) with a decelerating outflow ( cabrit & bertout 1990 , raga et al . we calculated position - velocity ( pv ) diagrams based on this infall - outflow model . these pv diagrams can be compared with the observed pv diagrams from sgr b2(m ) to better understand the physical processes in the sgr b2(m ) molecular cloud . for an ideal isothermal flow , under the assumption of spherically symmetric collapse , the volume density @xmath118 and velocity profiles @xmath119 as functions of radius @xmath120 within the collapsing inner envelope can be described by the power - laws ( shu 1977 ) , @xmath121 where @xmath122 is outer boundary of the collapsing cloud in spherical geometry , @xmath123 is the density at @xmath122 , @xmath124 is the mass of the cloud interior to @xmath120 , @xmath114 is the gravitational constant . because the actual core has a finite size , the infall is terminated at an inner radius of @xmath125 . the negative sign in the velocity indicates that the direction of the inflow is opposite that of the unit vector of radius in the spherical geometry . the geometry of a cloud with spherical inflow and decelerating outflow is shown in the cartesian coordinate system ( @xmath126 , @xmath127 , @xmath128 ) in fig . 9 . the sky plane is the x - y plane . the column density ( @xmath129 ) at a sky position ( x , y ) can be calculated by integrating the volume density @xmath118 . the column density in the sky plane is @xmath130 where @xmath131 , @xmath132 . in order to get an analytical form for the column density , we express @xmath133 in a taylor expansion . neglecting the contribution from higher order terms since the interesting region corresponds to @xmath134 ( or @xmath135 ) in the practical case , the final solution of equation ( a2 ) is @xmath136 at ( x , y)=(0,0 ) , the inflow terminates at the inner boundary @xmath125 . assuming the emission from the far - side gas is blocked by the central compact object , the effective column density along the line - of - sight is @xmath137.\ ] ] from equation ( a1 ) , excluding the mass of the central compact object , the mass @xmath124 of the spherically symmetric cloud within @xmath120 can be derived by @xmath138.\ ] ] the mass interior to @xmath122 is @xmath139 if @xmath140 . therefore the equation ( a5 ) can be re - written as @xmath141 the radial velocity of infall gas is the projection of the velocity vector * v * on @xmath128 axis . the radial velocity averaged along the line - of - sight is zero at a position away from the central position ( x , y)=(0,0 ) due to the cancellation of the velocity in a spherically - symmetric inflow . at the position ( x , y)=(0,0 ) , ignoring the contribution of the emission from the region behind the central object , the velocity at ( x , y)=(0,0 ) in front of the central object can be expressed by @xmath142 we note that the sign convention in the spherically symmetric description of infalling gas in front of the central object is opposite to the convention of the radial velocity with respect to the central object in the local standard rest ( lsr ) frame . in addition , in the derivation above , the systematic velocity is assumed to be zero . molecular outflows are observed in star formation regions . various models have been also proposed to interpret the observed outflow morphology ( e.g. , raga et al . 1993 ; cabrit & berout , 1990 ) . in this appendix , a pv diagram for a decelerating flow is calculated . the outflow velocity field and density are assumed to have power - law distributions along the major axis of the flow ( cf . cabrit & bertout 1990 ) @xmath143 where @xmath123 and @xmath144 are the volume density and velocity at the inner radius @xmath145 , respectively ; @xmath146 , the distance from the center , because the flow is in the x - z panel , i.e. , @xmath147 . the geometry of the blue - shifted outflow is illustrated in fig . 9 by the cone in the @xmath148 plane with an opening angle of 2@xmath149 . the inclination angle of @xmath150 is the angle between the z - axis and the axis of the blue - shifted outflow . the angle between the north and x - axis is the negative position angle of the blue - shifted outflow if we take the convention of position angle in aips . the radius of an outflow cross - section is equal to @xmath151 . given conservation of mass flux across a cross - section and @xmath152 , the relationship of the power - law indices @xmath153 and @xmath34 can be derived @xmath154 the size of a cut across the major - axis of the outflow along the line - of - sight ( @xmath128-axis ) can be expressed approximately by , @xmath155 if we take @xmath156 and @xmath157 ( cabrit & bertout 1990 ) , then the column density along the line - of - sight can be calculated from @xmath158 the result of the integration is @xmath159 the column density depends on the radius of the inner cross section ( @xmath145 ) , volume density @xmath123 at @xmath145 , the opening angle ( @xmath149 ) and inclination angle ( @xmath150 ) . in practice , the @xmath150 and @xmath149 are in the ranges of @xmath160 and @xmath152 , respectively . a blue - shifted outflow corresponds to @xmath161 and @xmath162 to a red - shifted outflow . 10 plots @xmath163 as function of @xmath149 and @xmath150 . @xmath163 is a weak function of @xmath149 and @xmath150 . given @xmath164 and @xmath165 , @xmath166 is close to unity . the mean radial velocity of the outflow can be expressed as @xmath167 the radial velocity of the outflow is inversely proportional to @xmath126 along the major axis of the outflow and is a weak function ( @xmath168 ) of @xmath149 and @xmath150 . as noted in the calculation of the infall velocity , the sign of the velocity used in the equation ( a13 ) is opposite to the convention of the radial velocity with respect to the central source . the velocity of the central source with respect to the lsr , or systematic velocity , is not included here . 11 shows the plot of @xmath168 as a function of @xmath149 and @xmath150 . in the range of @xmath169 , @xmath168 has a negative value corresponding to a red - shifted outflow , while @xmath170 is for a blue - shifted outflow . given @xmath164 and @xmath165 , @xmath171 is about 0.6 . in addition , for given @xmath150 , @xmath145 , and @xmath172 , the observed radial velocity decreases as @xmath149 increases . assuming @xmath173 km s@xmath4 at @xmath174 ( 0.04 pc ) , @xmath164 and @xmath165 , we can calculate a pv diagram of an outflow ( the thick solid contour in fig . 12 shows a pv diagram for the parameter @xmath175 in a range between 0.5 to 5.1 km s@xmath4 pc assuming that the blue and red - shifted outflows are symmetric around the central source , whose velocity is 58 km s@xmath4 with respect to the lsr . in this diagram , we also includes an absorption component of a spherical inflow with a mean infall velocity of 8 km s@xmath4 with respect to the central source . the pv - diagram ( fig . 12 ) shows the typical configuration of a spherically inflow and decelerating outflows in an active star formation region . in order to model the excitation conditions and the physical properties of the molecular cloud components observed in sgr b2 , we solve for the radiation transfer in a multi - level system with the large velocity gradient ( lvg ) approximation . in this model , a molecular cloud is assumed to be spherically uniform at kinetic temperature ( @xmath176 ) , column density per unit velocity interval ( @xmath94 ) and h@xmath0 density ( @xmath177 ) with a volume filling factor @xmath178 of the line emitting gas . the escape probability , @xmath179 , is used to account for photon trapping . the volume filling factor @xmath178 is incorporated into the opacity calculation via the equation where @xmath181 is the optical depth at the line center , a@xmath182 is the einstein coefficient for spontaneous emission , @xmath183 is the abundance of para - h@xmath0co , @xmath184 is the velocity gradient , @xmath185 and @xmath186 are the fractional population and statistical weight of level i , respectively . the column density per unit velocity interval at the line center in the lvg model is given by given @xmath85 , @xmath176 and @xmath87 , the local lvg model gives the radiative temperature or the intensity of the line requiring as input the collisional excitation rates . we adopted the collision rates calculated by green ( 1991 ) . green used the interaction potential between h@xmath0co and he . since he has twice mass of h@xmath0 , collisional excitation due to h@xmath0 should be 2.2 times more effective than collisional excitation by he ( green 1991 ; mangum & wootten 1993 ) . the collisional excitation rates from green were multiplied by this factor in our calculations . the accuracy in the green s calculation of the total collisional excitation rates is @xmath620% . incorporating the green s collisional excitation rates and the volume filling factor discussed above , we used the radiative transfer code in _ miriad _ to solve for the radiative intensities from the para - h@xmath0co for the two transitions @xmath1 and @xmath2 . in the line emission case , the line intensity ratio of the two transitions is a good indicator of the kinetic temperature in molecular clouds ( _ e.g. _ mangum & wootten 1993 ) . this ratio becomes strongly dependent upon the volume density ( @xmath85 ) when the kinetic temperature is greater than the upper state energy of the highest excitation transition ( @xmath188 ) . given a column density ( @xmath94 ) , both the intensity of the lower transition i(@xmath1 ) and the intensity ratio ( i(@xmath1)/i(@xmath2 ) ) as a function of @xmath189 and @xmath85 can be calculated . we calculate five models with different sets of volume filling factor of the para - h@xmath0co gas ( @xmath190 ) , beam filling factor of the line emission or absorption ( @xmath68 ) and beam filling factor of the continuum emission ( @xmath83 ) . in model a , b , c and d , the line filling factors @xmath191 and @xmath192 are assumed while @xmath193 and @xmath194 are assumed in model e. the continuum filling factors of @xmath195 and 0.05 are assumed in model a , b , c , d and e respectively . fig.13 shows the lvg results calculated with model c for the component k1 - 3 at 72 km s@xmath4 . the absorption line intensities of the two transitions @xmath1 and @xmath2 are plotted as function of the kinetic temperature and h@xmath0 density . the line intensity ratio i(@xmath1)/i(@xmath2 ) is also shown ( bottom panel ) . in the high density and cold region ( the right - bottom corner ) , both transitions becomes optically thick and the excitation temperature is much smaller than the brightness temperature of the background continuum radiation and the line ratio becomes unity . when the gas becomes hot and less dense ( left - top corner ) , the line ratio becomes large . fitting the observed results of the line intensity and the intensity ratio to the lvg model , we can determine @xmath176 and @xmath85 . we find that the absorption case is different from the case of emission , and the absorption line intensity of the lower energy transition , i(@xmath1 ) , is a good indicator of the kinetic temperature ( @xmath176 ) of a cloud component while the line intensity ratio i(@xmath1)/i(@xmath2 ) places a strong constraint on the h@xmath0 volume density ( @xmath85 ) ( see the top panel of fig . 14 shows examples of the lvg fitting with model c ( @xmath196 , @xmath192 , and @xmath197 ) to the observed results from three typical regions . top panel in fig . 14 is the result showing the lvg model curves fitting to the observed i(@xmath1 ) and i(@xmath1)/i(@xmath2 ) for the velocity component at 72 km s@xmath4 in the region k1 - 3 of sgr b2(n ) . the dashed lines are absorption in the units of jy beam@xmath4 . the thick dashed line and two thin dashed lines in light blue are the observed values of i(@xmath1 ) and 1 @xmath28 uncertainty . the solid lines are the intensity ratio i(@xmath1)/i(@xmath2 ) . the thick line and two thin lines in red correspond to the observed ratio and 1 @xmath28 uncertainty . the hatched region marks the solution ranges of 2.9@xmath1987.9@xmath198 @xmath11 in @xmath85 and @xmath199 k for given @xmath200 @xmath89 km@xmath4 s. we searched for lvg solutions in a wide range of @xmath201 @xmath89 km@xmath4 s in log(@xmath87 ) . the bottom - left panel in fig . 14 shows the velocity component at 68 km s@xmath4 in the f1 - 4 region of sgr b2(m ) . given @xmath202 @xmath89 km@xmath4 s , two possible solutions were found for this region , one with higher temperature ( t@xmath203 150 k ) and lower density @xmath204 @xmath11 and other with lower temperature ( t@xmath203 110 k ) and higher density @xmath205 @xmath11 . the bottom - right panel in fig . 14 shows a lvg model to the observed results from the isolated emission region ( m3 ) at 53 km s@xmath4 . given @xmath206 @xmath89 km@xmath4 s , t@xmath207 k and @xmath208 @xmath11 were found . the large uncertainties in t@xmath84 and @xmath85 are due to relatively weaker line emission . k1 - 3@xmath209&17 47 19.89 & @xmath30.1&28 22 17.6 & @xmath30.3&6@xmath210 ) & 29.2@xmath32&52.1@xmath34 + k4 & 17 47 20.02 & @xmath30.1 & 28 22 04.7 & @xmath30.1&5@xmath211 ) & 1.4@xmath30.1&3.5@xmath30.3 + ne&17 47 20.56 & @xmath30.5 & 28 22 14.4 & @xmath30.3 & 5@xmath212 ) & 1.2@xmath30.1&2.5@xmath30.3 + nw&17 47 19.27 & @xmath30.3 & 28 22 11.8 & @xmath30.3&5@xmath213 ) & 1.0@xmath30.1&2.3@xmath30.2 + z10.24 & 17 47 20.04 & @xmath30.3&28 22 41.2 & @xmath30.4 & 13@xmath214 ) & 2.0@xmath30.2&9.1@xmath30.8 + f1 - 4@xmath215&17 47 20.17 & @xmath30.2&28 23 04.9 & @xmath30.3 & 7@xmath216 ) & 20.2@xmath31.3&35.6@xmath33 + me&17 47 21.65 & @xmath30.6&28 22 57.0 & @xmath30.7&4@xmath217)&0.58@xmath30.15&1.0@xmath30.3 + mw&17 47 19.56 & @xmath30.8&28 23 04.3 & @xmath30.6&12@xmath218 ) & 1.7@xmath30.2&12.3@xmath31.3 + k1 - 3@xmath209 & 13.7@xmath31&21.2@xmath32&18.7@xmath32 + k4 & 1.0@xmath30.1&1.5@xmath30.1&1.3@xmath30.1 + ne&0.73@xmath30.06&3.2@xmath30.3 & 4.8@xmath30.4 + nw&0.66@xmath30.06&1.2@xmath30.1&1.1@xmath30.1 + z10.24 & 2.6 @xmath30.2&1.7@xmath30.1&0.9@xmath30.1 + f1 - 4@xmath215 & 7.8@xmath30.8&9.0@xmath31.3&6.6@xmath30.7 + me&0.28@xmath30.03&1.0@xmath30.1&2.4@xmath30.3 + mw&3.2@xmath30.4&1.1@xmath30.1&0.46@xmath30.06 +

we present deep , near - infrared images of the circumbinary disk surrounding the pre main - sequence binary star , a , obtained with nicmos aboard the hubble space telescope . the spatially resolved proto - planetary disk scatters.5% of the stellar flux , with a near - to - far side flux ratio of.4 , independent of wavelength , and colors that are comparable to the central source ( , ) ; all of these properties are significantly different from the earlier ground - based observations . new monte carlo scattering simulations of the disk emphasize that the general properties of the disk , such as disk flux , near side to far side flux ratio and integrated colors , can be approximately reproduced using ism - like dust grains , without the presence of either circumstellar disks or large dust grains , as had previously been suggested . a single parameter phase function is fitted to the observed azimuthal variation in disk flux , providing a lower limit on the median grain size of . our analysis , in comparison to previous simulations , shows that the major limitation to the study of grain growth in t tauri disk systems through scattered light lies in the uncertain ism dust grain properties . without explicit determination of the scattering properties it is not possible to differentiate between geometric , scattering and evolutionary effects . finally , we use the 9 year baseline of astrometric measurements of the binary to solve the complete orbit , assuming that the binary is coplanar with the circumbinary ring . we find that the estimated 1 range on disk inner edge to semi - major axis ratio , , is larger than that estimated by previous sph simulations of binary - disk interactions .

astro-ph0204465

gg tau a is one of the few t tauri systems in which a disk has been spatially resolved at both millimeter and near - infrared wavelengths and is therefore an ideal system for detailed studies of proto - planetary disk geometry and composition . as the second brightest millimeter object in the survey of beckwith et al . ( 1990 ) , gg tau was an early target for millimeter wave interferometry studies ( simon & guilloteau 1992 ; kawabe et al . 1993 ; dutrey , guilloteau & simon 1994 ; guilloteau , dutrey & simon 1999 , hereafter g99 ) . these observations showed the emission to be from a massive ( @xmath6 ) disk surrounding gg tau a , the closest pair of stars in the gg tau quadruple stellar system , with the bulk of the emission arising from a distinct ring structure extending from 180 to 260 au ( assuming a distance of 139 pc ; bertout , robichon & arenou , 1999 ) . the circumbinary disk was subsequently detected via near - infrared scattered light in multi - wavelength , ground - based adaptive optics ( ao ) images ( roddier et al . 1996 , hereafter known as r96 ) and in space - based hubble space telescope ( hst ) 1 polarimetric images ( silber et al . 2000 ) and wfpc2 optical images ( krist , stapelfeldt & watson 2002 ) . analysis of the ground - based near - infrared images resulted in disk colors that appear to be redder than the central stars ( r96 ) . such a red color excess could either indicate that the circumbinary disk has a substantial population of large dust grains , or that circumstellar disks , which are coplanar with the circumbinary disk , redden the stellar light before it scatters off the circumbinary disk . comparing the variation of disk magnitude with color around the disk , roddier et al . find a trend for fainter regions of the disk to be redder , leading them to suggest that circumstellar disks are the cause of the observed red color excess . the existence of circumstellar disks around the individual components in gg tau a is supported by observations of near - infrared excesses and strong hydrogen emission lines in each component ( white et al . no information , however , is currently available regarding their orientation and in fact many disks in binary systems show evidence of non - coplanarity ( e.g. , stapelfeldt et al . 1998 ; monin , mnard & duchne , 1998 ; jensen et al . 2000 ; wood et al . 2001 ) . to test the circumstellar disk hypothesis , wood , crosas & ghez ( 1999 ) ran monte carlo scattering simulations of the gg tau circumbinary ring with and without circumstellar disks being present . using the kim , martin & hendry ( 1994 ) interstellar medium ( ism ) grain properties and the disk geometry derived by dutrey et al . ( 1994 ) , they find that the circumstellar disks are required to match the disk to star flux ratios and near side to far side flux ratios observed in the ground - based ao images . while these simulations roughly reproduce the estimated quantities due to the large observational uncertainties , they predict blue rather than red disk color excesses . in order to make further progress in understanding the role of circumstellar disks and the possibility of large particles , much higher signal to noise measurements of the circumbinary disk are required at multiple wavelengths . in this paper we present new high angular resolution , multi - wavelength observations of the gg tau system taken with nicmos aboard hst . these deep , space - based observations offer a more stable point spread function ( psf ) than that obtainable with ground - based ao systems , allowing a more accurate determination of the disk properties . the observations are outlined in [ obs ] and the data reduction and psf subtraction method used are explained in [ reduc ] . newly derived stellar and circumbinary disk properties are presented in [ res ] and are discussed in sections [ discus ] and [ implications ] in light of additional monte carlo simulations . gg tau a ( @xmath7 , @xmath8 ) was observed using the near - infrared camera , nicmos , aboard hst on 1997 october 10 . deep images were obtained with the f110w ( @xmath9 , @xmath10 ) , f160w ( @xmath11 , @xmath12 ) , and f205w ( @xmath13 , @xmath14 ) filters0.3 shorter than the standard k band filter and is @xmath00.2 wider . ] . in the f110w and f160w filters , gg tau was imaged 9 times on nic1 , the nicmos camera with the smallest field of view , 11@xmath15 11@xmath16 , and pixel scale of 0.@xmath16043 per pixel ( thompson 1995 ) . each image was offset by 2.@xmath162 in a spiral dither pattern in order to reduce residual flat field uncertainties and the dither size was set to a non - integral number in order to reduce intra - pixel sensitivity variations , while also keeping the circumbinary disk in the field of view . each individual dither image was integrated for 128 seconds in the multiaccum mode . the f205w filter data were taken using nic2 , which has a 19.@xmath162 @xmath17 19.@xmath162 field of view and a pixel scale of 0.@xmath16075 per pixel . due to the significant thermal background at this wavelength a dither - chop observing pattern was employed , whereby each image is composed of one 128 second integration on gg tau followed by a 32@xmath16 chop and a 128 second integration on a blank piece of sky . nine images were obtained in this manner . because the aim of our research was to detect faint , diffuse emission close to bright sources , a good characterization of the point spread function was required . to this end , images of a calibrator star , ntts 042417 + 1744 , were obtained in the same manner as gg tau in each of the above mentioned filters . this weak - lined t tauri star has a number of properties which were used to select it as a calibrator : no infrared excess indicative of circumstellar material ( walter et al . 1988 ) , known to be single from previous speckle interferometry observations ( ghez et al . 1993 ) , only half a magnitude brighter than gg tau aa in the j band ( 1.25 ) , and a spectral type of k1 , which is similar to the stellar components of the binary star ( k7 and m0.5 for gg tau aa and ab respectively ; white et al . initial data reduction was carried out by stsci , providing a quick analysis of the data quality . the images were recalibrated with the latest calibration reference files using the stsci s pipeline routine calnica ( bushouse , skinner & mackenty , 1996 ) , which performs standard image reduction tasks : removal of bias and dark current levels , interpolation over known bad pixels , flatfielding , and removal of all cosmic ray hits . once the data were calibrated , each individual image was sub - pixelated by a factor of 2 . in order to obtain estimates of the properties of the central binary ( [ orbit ] ) and the circumbinary disk ( [ cbdisk ] ) , a model of the central binary star was generated through psf fitting and then subtracted from the individual images . both the empirical psf ( observations of ntts 042417 + 1744 ) and tinytim model psfs ( krist , 1993 ) were used . the psf - fitting was carried out for each of the nine gg tau images by minimizing the chi - squared between the psf model and the data in the regions dominated by the binary star ( the central area of @xmath18 ) and , in the case of the empirical psf , the diffraction spikes ( an additional @xmath19 comprising 4 rectangular regions covering the diffraction spikes beyond 2distance from the central binary ) . for the empirical psf , all nine psf estimates were evaluated and the one that produced the best fit , i.e. , the smallest residuals , was selected for each gg tau image . although the tinytim routine allowed us to match the colors of the two gg tau a stellar components as well as the focus position of the observations ( see 3.2 ) , the empirical psf models consistently produced smaller residuals than those created using tinytim , and thus the final values for the central binary and the circumbinary disk were derived using the empirical psf model subtracted images . the difference in the quality of the fits between the empirical psf and the tinytim psf can be understood in terms of ( a ) small scale psf variations due to the telescopes `` breathing '' , the change in the focus of the telescope on timescales of an orbit due to a varying thermal load , and ( b ) the presence of scattered light off the telescope optics ( krist and hook , 1999 ) . neither of these effects are incorporated into the tinytim psf models used , and both are measured with the empirical psf . the nine model subtracted images were combined together to form an average subtracted image at each wavelength , which are shown in figure [ mosaic ] . azimuthal averages of the model subtracted images are plotted in the first row of figure [ profile ] . a map of the measurement uncertainties was created by taking the standard deviation of the mean from all nine psf subtracted images of gg tau . azimuthal averages of these statistical uncertainties are shown in the second row of figure 2 ; while these values provide a good estimate of the uncertainties associated with the measurement and subtraction process , they do not take into account the presence of any systematic effects . two systematic effects that alter the structure of the psf have been identified . the first , which affects all three band passes , arises because the observed psf , ntts 042417 + 1744 , is slightly bluer than either component of gg tau . the second is an inadvertent focus offset between the object and psf in the f205w data . these effects create a systematic bias and an additional source of uncertainty , both of which were modeled and accounted for . in order to investigate the consequences of the small color differences between the object and psf , we generated polychromatic models of gg tau aa , ab and ntts 042417 + 1744 using tinytim , matching the spectral shape of these objects using previous photometry from ghez et al . ( 1997 ) and walter et al . the psf fitting and subtraction process was run on the tinytim model images in order to produce a model of the bias term ( i.e. , the subtraction residuals caused by this color mismatch ) ; azimuthal averages of this color bias are plotted in the first row of figure [ profile ] . to obtain an estimate of the uncertainty associated with this measurement of the color bias , the method was repeated using a psf that varied in spectral type by the assumed uncertainty in the calibrators spectral type ( one subclass ) . the difference in residuals between using a psf of spectral type k0 and k1 is our estimate of the statistical uncertainty on the color bias term ( plotted in the second row of figure [ profile ] ) ; the uncertainty on the color bias is significantly smaller than the measurement uncertainties . while this method was sufficient for the f110w and f160w data , the f205w data required an additional systematic effect to be accounted for . observations of gg tau on nic2 were run in parallel mode with polarization observations ( which are not presented here ) on nic1 . due to the unexpected thermal problems , each nicmos camera had a different optimal focus position and therefore parallel observations were taken at non - optimal focus positions . since these observations were scheduled prior to the recognition of this problem , the polarization data were arbitrarily regarded as the prime observation . therefore both the polarization and f205w observations were made with the nic1 focus setting , resulting in slightly out of focus f205w observations . unfortunately , the f205w data on the psf were taken in a separate orbit with no parallel observations and therefore at the optimal nic2 focus setting . this mismatch in focus between the object and calibrator in the f205w data set was quantified using the same tinytim modeling used to determine the color bias in the f110w and f160w data , but incorporating both the color difference and mean focus offset . as shown in the third column of figure [ profile ] , the focus bias is the source of a considerably large systematic error . to estimate the systematic uncertainty associated with the measurement of this bias , a polychromatic tinytim model of gg tau was created with the focus set at the _ maximum _ focus position of the data set , and then subtracted using a k0 type psf set at the _ mean _ focus position of the psf data set . as focus and color mismatches can cause similar problems , this model set - up maximizes the difference between the psf for the object and calibrator data sets . in contrast with the f110w and f160w data sets , the uncertainty associated with the bias term in the f205w data dominates the measurement uncertainties . all of the results presented in [ res ] are obtained from analysis of images that have had estimates of the systematic bias subtracted out and uncertainties that have been combined in quadrature , as described in equations [ eq : final ] and [ eq : errors ] . @xmath20 @xmath21 @xmath22 is the median of the nine psf subtracted images and @xmath23 is the systematic bias from the mismatch in color and focus position . the two error terms in equation [ eq : errors ] are the statistical uncertainties of the image measurement and bias modeling respectively . the final row in figure [ profile ] displays the azimuthally average of the bias corrected signal - to - noise ( e.g. , @xmath24 ) . analysis of the f205w dataset is limited to the measurement of the binary star properties ( [ orbit ] ) and integrated disk photometry ( [ colors ] ) , since it is significantly affected by the focus bias . although this bias and its associated uncertainty have been modeled , the models were made using tinytim psfs that have been shown to deviate somewhat from the observed psfs . in this section we present the observed properties of the stellar system and the circumbinary disk . the binary orbital motion has been followed for the past 9 years and the newly estimated disk mass and inclination ( g99 ) allow us to derive estimates of all the orbital parameters for the binary system using only minimal assumptions ( [ orbit ] ) . for the circumbinary disk ( [ cbdisk ] ) , we focus on the following observational quantities that have , in the past , been considered useful tools in comparing observations and disk models : in [ 1-d ] , the disk morphology ( wood et al . 1999 ) , in [ colors ] , the disk to star flux ratio ( wood et al . 1999 ) and color excesses ( r96 ) , and in [ azvar ] , the spatially resolved color - magnitude relationships ( r96 ) and near side to far side flux ratios ( close et al . 1998 ; wood et al . 1999 ) . these properties are then compared to earlier measurements . deviations from previous estimates of these quantities are most likely attributable to a combination of low snr and over - deconvolution of the disk in the more challenging ao observations . table [ astrom ] lists the values for the binary separation , position angle ( p.a . ) and magnitude of each component obtained from the psf fitting of the central binary star . the uncertainties are the standard deviation of the estimates obtained from the nine individual measurements , plus additional contributions from the uncertainties of the spacecraft orientation ( 003 ; holtzmann et al . , 1995a ) and the pixel scale ( 0.5% ; nicmos instrument handbook ) . the 5% uncertainty in the absolute photometry calibration , however , is not included as our analysis rests on the relative photometry of the disk compared to the central stars . the reported individual stellar flux density uncertainties are dominated by psf variations . wide aperture photometry was carried out with the assumption that the disk is a small contamination on the stellar light ( @xmath0 1% , see [ colors ] ) ; this both checks the psf fitting measurements and provides a more precise value of the total stellar flux which is used in [ colors ] . using an aperture 26 in radius , we obtain a total binary flux density that is consistent to within 1@xmath4 with the values found through psf fitting ( see table [ astrom ] ) . over the 9 year period 1990 to 1998 , the binary system was repeatedly observed ( leinert et al . 1993 ; ghez et al . 1995 , 1997 ; roddier et al . 1996 ; silber et al . 2000 ; white & ghez 2001 ; woitas , khler & leinert 2001 ; krist et al . 2002 ) . both the new hst photometric and astrometric measurements agree well with the other reported measurements . furthermore , the astrometric collection of measurements has grown significantly such that estimates of the relative velocities in the plane of the sky can be greatly improved . combining the hst measurements with those published previously ) , speckle interferometry ( @xmath25 ) and adaptive optics ( @xmath26 ) astrometric results , respectively . furthermore , the astrometry from leinert et al . ( 1993 ) , obtained from three separate 1d measurements is not included in our fit . ] and assuming constant linear motion , we find a velocity of @xmath27 mas yr@xmath28 at a position angle of @xmath29 , with an average relative position of @xmath30 at the position angle of @xmath31 . this velocity has a factor of 5 smaller uncertainties than previous estimates ( ghez et al . 1995 ; woitas et al . 2001 ) , mostly resulting from including more measurements . with the greater precision in both the velocity reported here and the total mass of the system , derived from the rotation of the circumbinary disk ( g99 ) , it is now possible to solve for the whole orbit , assuming only that the orbit is coplanar with the circumbinary ring , and that the ring is intrinsically circular . appendix [ getorb ] provides the details of how the orbital parameters , @xmath32 , @xmath33 , @xmath34 , @xmath35 , @xmath36 and their uncertainties are derived from the observables , @xmath37 , @xmath38 , @xmath39 , @xmath40 , @xmath41 , @xmath42 . we find that the stars are in an elliptic orbit with an eccentricity @xmath43 and a semi - major axis @xmath44au . the corresponding orbital period is @xmath45yr . we emphasize that the uncertainties quoted here include the uncertainty on the distance to the taurus molecular cloud , which must also be taken into account when considering the system mass ( g99 ) . indeed , together with the uncertainty in the measured velocity , this is the dominant source of error in the quoted uncertainties . although the orbital parameters derived here still have large uncertainties , this analysis shows the applicability of this method . additional astrometric data over the next few years should help to significantly decrease these uncertainties and ascertain the exact orbit of the binary . we note that our orbital solution differs significantly from that obtained in the past , in the sense that we find that the stars are close to apoastron while r96 concluded that the stars must be close periastron . this contradiction is the result of our estimated velocity being much smaller than the earlier estimate , and smaller than the velocity one would observed if the orbit was circular . a similar conclusion , based on velocity comparison alone , was recently reached by krist et al . our derived orbit is shown in figure [ orb ] , where we have also plotted yearly weighted averages of the data points used in the fit . in the framework of investigating interactions between the ring and the inner binary , it is interesting to compare the inner radius of the circumbinary disk to the semi - major axis of the system . the ratio of these two quantities lies in a 1@xmath46 range @xmath5 ; the two stars have cleared a wide gap around them . artymowicz & lubow ( 1994 ) have used analytical and numerical approaches to study the formation of such a gap through gravitational resonances in geometrically thin disks around circular and eccentric binary systems . the orbital properties of gg tau a provide us with the first direct test of these models in pre main - sequence visual binary systems , although uncertainties prevent us from reaching definitive conclusions . artymowicz & lubow found that this gap clearing phenomenon is primarily driven by the orbital eccentricity , with a smaller dependence on the mass ratio , ( @xmath47 in the case of gg tau a ; white et al . the results of artymowicz & lubow can be summarized as follows : while the @xmath48 ratio is on the order of 1.7 for circular orbits , it can grow up to @xmath49 for highly eccentric binaries ( @xmath50 ) . the derived orbital parameters appear somewhat problematic , with a @xmath48 ratio that seems to be larger than those expected for a moderately eccentric system . this may mean that the dynamical evolution of the gg tau circumbinary ring is more complex than previously thought . azimuthal averages of the psf subtracted images ( see figure [ profile ] ) show that the circumbinary disk is easily detected at radii greater than @xmath01 at all three wavelengths . the circumbinary disk is detected with an average snr @xmath51 5 per pixel at radii of 09 to @xmath019 ( f110w and f160w ) and 11 to 19 ( f205w data ) with an average peak surface brightness of 15.3 , 14.1 and 14 mag / arcsec@xmath52 in f110w , f160w and f205w respectively . at smaller radii the 1@xmath4 subtraction noise level rises dramatically , from @xmath014 mag / arcsec@xmath52 at 05 to @xmath011 mag / arcsec@xmath52 at 02 , with each filter having a slightly different noise level ( see figure [ profile ] ) . this prevents a meaningful investigation of gap material that has been posited in varying forms in earlier observations ( r96 ; silber et al . 2000 ; krist et al . 2002 ) . a model of the two - dimensional near - infrared apparent disk geometry is constructed from the f110w and f160w data by analyzing the intensity profile of the disk as a function of azimuth . note that what we are observing here is the geometry of the optically thick scattered light distribution , not the true disk geometry . for every 10 degree segment of the disk , an average radial profile is produced . the position of the radial profile peak value is found using two methods . the first calculates a weighted centroid over a 05 region centered on the maximum value of the radial profile , which works well on sections of the disk that have a distinct peak . the south side of the disk , however , has a much flatter radial profile , and the weighted centroid method is not as stable for these segments . the second method fits the radial profile with a 4 degree polynomial . the ` center ' of the disk is then calculated by finding the midpoint between the two radii at which the disk value is equal to half the peak value . for the north side of the disk where the radial profile is almost gaussian in shape , this method finds the same disk peak . for the flatter southern profiles it provides a more robust method of finding the disk center . the final peak position values are taken from the results of the polynomial fit . this method of finding the disk peak location also provides an estimate of the width of the radial profile , which is taken to be the distance between the two half - peak radii . uncertainties for these values are estimated by taking the largest of either the difference between the values found through the two methods or the standard deviation of the values found from applying the same analysis to 3 subsets of the data . the overall apparent shape of the disk is explored by fitting an ellipse to the positions of the radial profile peak values . we find an ellipse with a semi - major axis of 142 @xmath53 006 ( 200 au ) orientated at a position angle , @xmath54 of 21@xmath55 @xmath53 9@xmath55 and an eccentricity of 0.64 @xmath53 0.02 . although only g99 provided uncertainties for their analysis , our measurements of the ring geometry are consistent with that found in both the previous near - infrared ( r96 ) and millimeter observations ( dutrey et al . 1994 ; g99 ) . if you assume that the disk is intrinsically circular and geometrically thin , the observed eccentricity corresponds to a disk inclination of 40@xmath56 . while this is consistent with inclination measurements made from the optically thin millimeter images of the circumbinary ring ( @xmath57 , @xmath58 ; g99 ) , the disk is known to be geometrically thick , which should bias the observed inclination towards slightly larger values . g99 also note that the geometrical thickness of the inclined ring will cause the northern inner edge of the scattered light ring to appear closer to the center of mass of the system than the inner edge observed in the optically thin millimeter images . this offset can be used to calculate the total height of the disk above the midplane , i.e. , where the nir light becomes optically thin . this height is where the disk is physically truncated in the monte carlo simulations ( see [ discus ] ) . in the nicmos images , the northern inner edge ( the peak of the radial profile ) occurs @xmath59 away from the center of mass . from the disk model fits to the millimeter images ( g99 ) , the inner edge of the millimeter disk occurs at a projected distance of @xmath60 . the offset , @xmath61 , corresponds to a total disk height of @xmath62au ( using the relation from g99 ) . this is smaller than the 025 offset derived by g99 from the deconvolved adaptive optics images of r96 . in addition , the disk height will also translate into an offset between the center of the ring and the center of mass of the system . the disk fitting routine finds that the center of the ring is offset from the center of mass of the binary by 021 @xmath53 003 along a position angle of @xmath63 degrees , consistent with the offset measured from the inner edge . the width of the disk has a clear azimuthal dependence , with the north side being narrower than the south side , as noted by both silber et al . ( 2000 ) and krist et al . figure [ width ] shows the measured width , which ranges from 03 to @xmath01 , as a function of position angle . we fit these measurements with a modified henyey - greenstein scattering phase function ( described in @xmath64[azvar ] ) , which is physically meaningless in this context but provides an analytical description of the data . this analysis shows that the thinnest portion of the disk occurs at a position angle of 356@xmath55 , independent of wavelength ( see table [ azfit ] ) . because the function we used for the fit is unphysical and provides only a moderately good fit , we do not estimate uncertainties . at the lowest contour levels we observe a deviation away from the elliptical fit along the southern edge of the disk ( see figure [ kink ] ) . this deviation consists of a sharp elbow , or kink , in the south - east edge of the disk and a straightening of the isophotes along the southern edge which makes the disk outer edge appear boxy . this effect is present in all 3 filters , and coincides with the kink detected by silber et al . ( 2000 ) and noted by krist et al . ( 2002 ) . in order to assign flux densities to the disk , we define an annular aperture with an eccentricity of 0.64 centered at the position of the center of the ellipse with an inner semi - major axis at 1 and outer semi - major axis at 22 , based on the results given in [ 1-d ] . the area affected by the diffraction spikes is particularly noisy and is excluded from the aperture by masking off these regions , reducing the effective aperture area from 9.2 to 5.82 square arcseconds . the spike mask comprises two diagonal stripes , 08 in width , centered on the position of the primary . summing the counts over this area and correcting the flux for the masked off regions ( by multiplying the area of the masked regions with the mean disk flux per pixel ) , provides the estimates of the total disk magnitudes given in table [ astrom ] . since the pixel scale is smaller than the diffraction limit , the estimated uncertainties are based on maps that have been averaged over regions corresponding to the diffraction limited beam sizes ( 01 , 016 and 02 for f110w , f160w and f205w respectively ) . these measurements of the integrated disk intensity have an order of magnitude higher signal to noise ratios than earlier ground - based measurements . the overall disk magnitudes are compared to those derived for the binary system ( [ orbit ] ) to obtain disk to star flux ratios and color excesses ( see table [ nearfar ] ) . the nicmos images presented here lead to a disk to star flux ratio of @xmath01.5% , which is 2.5 times larger than that derived by r96 . likewise , the color of the circumbinary disk is comparable to that of the stars : @xmath65 + @xmath66 where @xmath67 is the color excess of the disk , or @xmath68 . this is in contrast to the large red excess suggested earlier . the spatially resolved color properties of the circumbinary disk are investigated by calculating the disk magnitudes within circular apertures which are 019 in diameter ( roughly the size of the 2 diffraction limit ) placed around the disk . for this comparative analysis , the f110w and f160w bias - subtracted images have been convolved to the resolution of the f205w image . only areas of the disk used to calculate the integrated disk photometry are included in this analysis . figure [ diskmag ] displays the resulting color - magnitude plot , which shows no significant trend in color with respect to magnitude . unlike the results from r96 , the nicmos measurements provide no evidence for extinction within the disk . as originally mentioned by r96 , the disk exhibits azimuthal variations in intensity that are presumed to arise from angular variations in the scattering efficiency , with the brightest side of the ring corresponding to forward scattering from the edge of the disk that is nearest to our line of sight . here we present a detailed quantitative analysis of these azimuthal variations , using normalized peak flux densities derived from the 10@xmath55 azimuthal averages discussed in [ 1-d ] . to estimate the position angle of the brightest region , the data are fit , using a chi - square minimization technique , with a modified , is added . ] form of the henyey - greenstein scattering phase function ( henyey & greenstein , 1941 ) @xmath69[1+g'^{2 } - 2g'cos(pa - pa_{0})]^{-3/2}\ ] ] where @xmath70 is the value of position angle that maximizes the function . while the use of this function is not physically meaningful , as the phase function depends on scattering angle , not position angle on the sky , it does provide a convenient method of characterization . the best - fit functions are displayed with the data in the left hand column of figure [ phase ] while the values of @xmath70 and the fitted peak - to - peak near side to far side flux ratios are listed in table [ azfit ] . the peak near side to far side flux ratio is @xmath04 with no significant wavelength dependence . the position angle of the brightest portion of the disk ( @xmath71 ) is roughly consistent with that measured by krist et al . ( 2002 ) in i band images ( 25@xmath55 ) and both the position angle of the semi - minor axis ( @xmath72 , g99 ) and the position angle of the thinnest region ( @xmath73 , derived in [ 1-d ] ) . this provides strong support for the hypothesis that the disk is geometrically thick . the azimuthal variations in intensity arise from variations in dust grain scattering efficiencies as a function of scattering angle , rather than the observed position angle around the disk . assuming a single scattering scenario in a geometrically thin disk , close et al . ( 1998 ) derive scattering properties of the dust grain population by matching the intensity extrema to scattering angles of @xmath74 and @xmath75 , which , with an assumed inclination , @xmath40 , of 35@xmath55 , results in scattering angles of 55@xmath55 and 125@xmath55 . the nicmos images , however , contain much more information than just the intensity extrema , as the circumbinary disk samples a large range of scattering angles as a function of position angle around the disk . we therefore take the additional step of fitting the entire azimuthal variation of flux with a scattering phase function . using the known disk geometry , the observed variation with position angle can be translated , or de - projected , into a variation with scattering angle , @xmath76 , which is defined as the angle of deflection away from the forward direction of the incoming light . @xmath76 is geometrically related to the position angle around the disk , @xmath77 , the position angle which maximizes the azimuthal variation , @xmath70 , and the disk inclination , @xmath40 , by : @xmath78 where @xmath79 if @xmath80 + @xmath81 if @xmath82 a derivation of this relation can be found in appendix [ getscatang ] . @xmath83 is the opening angle of the disk as seen from the stars ; including this factor converts the relation from a geometrically thin disk to a thick disk case . it is assumed here that , at all position angles on the sky , the intensity is dominated by the portion of the ring corresponding to the smallest scattering angle , which is true for forward - scattering grains . using the values for the inner radius of the disk and a disk height of 35au , we calculate @xmath83 to be 11@xmath55 . in this case , with an inclination of 37@xmath55 , the scattering angle ranges from @xmath040@xmath55 at the closest edge to @xmath0120@xmath55 at the edge furthest from us . using the above relationship to translate the observed position angle into scattering angles , the intensity variations are then fitted with a normalized henyey - greenstein scattering phase function , in order to estimate the value of the dust asymmetry parameter , @xmath84 . this parameter is the averaged cosine of the scattering angle which ranges in value from -1 to 1 . given that we are using the one parameter henyey - greenstein scattering phase function we can only investigate forward throwing dust grains with @xmath85 . in this case , if the light is scattered isotropically , then the amount of forward scattered light equals that being back - scattered and @xmath84=0 . as the light becomes more strongly forward - scattered , the value of @xmath84 increases . the results of the fit for the dust asymmetry are summarized in table [ azfit ] . we find that omitting the disk opening angle will slightly overestimate @xmath84 . it should be recalled that these values are obtained under the assumption of single scattering ( see [ multiplescat ] for further discussion ) . since the system may be modulated by geometric factors , such as a possible shadowing by circumstellar disks , wood et al . ( 1999 ) found it more convenient to work with an integrated quantity to describe the near side to far side flux ratio . integrating the flux over the northern and southern halves of the disk , as defined by the position angle of the disk , we find the ratio of scattered light to be @xmath01.4 , with no significant wavelength dependence ( see table [ nearfar ] ) . this value is roughly a factor of two smaller than that seen in the ground - based ao images ( wood et al . in this section , we use the various quantities derived in [ res ] to revisit the question of whether the existence of large dust grains or opaque circumstellar disks are required to explain the observed color excesses . monte carlo scattering simulations are presented , qualitatively reproducing the nicmos data ( [ mcmodel ] ) and emphasizing the key role played by grain properties , especially the dust asymmetry ( [ multiplescat ] ) . the scattering simulations show that neither circumstellar disks or large grains are absolutely needed to reproduce the data and that our fitted scattering phase function can be used to derive a lower limit on the median size of the scatterers . since the nicmos measurements of the gg tau circumbinary disk provide different results from both those previously reported and modeled , we ran new monte carlo simulations to investigate whether the observed disk colors can be reproduced using a standard ism grain model . we began , like wood et al . ( 1999 ) , with the simplest case : modeling the scattered light distribution from a thick torus of ism - like dust , without the additional complications of either circumstellar disks or grain growth ( their model 1 ) . we used a monte carlo code that was readily available to us ; the same model which successfully reproduced the nicmos polarization results ( silber et al . the monte carlo code is based on a program from mnard ( 1990 ) that has been modified such that each photon interacts with a randomly sampled dust grain from the input grain size distribution . details of the modifications can be found in duchne ( 1999 ) . the disk is modeled as a thick torus , using a total disk height of 38 au at the inner edge of the ring , close to that estimated from the observed inner edge offset ( see [ 1-d ] ) . this torus has a total disk mass of 0.13 @xmath86 and a scale height of 21au at the inner edge of the ring ( 180au ) that varies as @xmath87 ( g99 ) . outcoming photons are sorted by inclination : all results presented here are for the range of inclinations closest to the actual inclination of the system , @xmath88 . the simulations follow 1 million photons per filter . the dust grain properties are taken from the mathis & whiffen ( 1989 , hereafter known as mw ) dust model . while the dust parameters ( e.g. , albedo and dust asymmetry ) in the mw dust model differ significantly from the kim , martin & hendry ( 1994 ; hereafter known as kmh ) model between 0.03 and 0.9 ( mw model a for an @xmath89 = 3.1 ) . in contrast , the kmh dust grains are composed of separate silicate and graphite particle populations , with a size distribution of @xmath90 between 0.005 and 1 . ] used by wood et al . ( 1999 ) , both models successfully reproduce the ism extinction curve . the simulated disk properties can be found in table [ nearfar ] . the resulting monte carlo simulations ( see model 1 in table [ nearfar ] ) roughly recreate the observed disk to star flux ratios and near side to far side flux ratios . they also produce a considerably red disk color excess , @xmath91 , @xmath92 . this is distinctly larger than the color excess observed with nicmos , and very different from the blue scattered light disk seen in the wood et al . ( 1999 ) models . the two main differences between the wood et al . simulations and those presented here are the dust grain properties used and a slightly different disk geometry . both the overall disk height and the degree of flaring differ ( wood et al . use a scale height that varies as @xmath93 from the earlier millimeter work of dutrey et al . 1994 ) . to fully compare our numerical results with those of wood et al . , we ran monte carlo simulations where the only difference is the dust grain properties . the results from these simulations ( see model 2 in table [ nearfar ] ) show that the red disk color excesses remain and therefore are predominantly caused by the dust grain properties . the apparently contradictory simulations presented here and in wood et al . can be reconciled by investigating the effect the different dust grain properties have on the scattering process . in optically thick situations , photons are scattered multiple times . the resulting scattered light surface brightness is dependent on two key dust parameters , the albedo of the dust , @xmath36 , and the azimuthal scattering dependence given by the dust asymmetry parameter , @xmath84 . in terms of disk geometry , the inclination of the system is also an important factor . at any one scattering angle , the output scattered surface brightness is dependent on both the albedo of the dust and the number of scattering events @xmath94 , such that @xmath95 ( e.g. , witt & oshel , 1977 ; witt 1985 ) . the greater the number of scatterings , which is a function of the dust asymmetry parameter , the greater reduction in the resulting scattered surface brightness . the simulations presented here and those in wood et al . ( 1999 ) are based on the same monte carlo scheme and have the same disk geometry . the main difference lies in grain properties used , and this alone is enough to explain the disk color reversal between the two sets of models . table [ dustprops ] compares the dust properties used in the two simulations . the dust asymmetry of the median scatterer in our simulations is @xmath02 times larger than that used by wood et al . , while the albedos are essentially the same . how does this affect the observed scattered light distribution ? consider first the far side ( southern edge ) of the ring where back - scattering is the dominant effect . for low @xmath84 values , such as those used in the wood et al . simulations , the scattering is not far from being isotropic , which means that the photons leaving the disk toward the line of sight will in general have gone through a small number of scatterings ( @xmath0 1 ) . thus the resulting surface brightness distribution will not be strongly affected by multiple scattering effects and the scattered light colors will be similar to the wavelength dependence of the dust albedo , resulting in wood et al.s case in blue disk color excesses . for more strongly forward scattering dust particles ( large @xmath84 ) , back - scattering is rarer and in this region of the disk most of the scattering events will cause photons to move further into the disk , away from our line of sight . the photons that do escape from the surface of the disk toward our line of sight will have undergone a much larger number of scatterings and the resulting scattered light surface brightness will be significantly changed by the wavelength dependence of the dust asymmetry parameter . for the mw dust properties , not only is @xmath84 larger than those in the kmh model , they also have a stronger wavelength dependence . the scattered light surface brightness will be more reduced at 1 than it is at 2 producing significantly red disk colors . the same story applies to the nearest ( northern ) edge of the disk , where we are predominantly seeing forward scattering off the upper limb of the ring . this type of scattering is enabled by both low and high @xmath84 values ( near - isotropic and strongly forward scattering ) ; while the number of scattering events remains small in both simulations there will still be more scattering events for the photons in our simulation than those in wood et al.s due to the inclination of the system favoring low @xmath84 values slightly more . this , combined with the corresponding slightly longer pathlength in our simulation , enabling more extinction , is enough of an effect to compensate for the intrinsically blue color of the scattering , as seen in our simulations . taking the disk as a whole , this explains why our simulations predict a red disk color , while those of wood et al . , being dominated by the flux from the far side of the disk , result in overall blue disk colors . the same @xmath96 dependency is responsible for the reduced disk to star flux ratios and increased near to far flux ratios observed in our simulations , as photons undergo more scattering events in the far side of the disk than in wood et al.s ( see table [ nearfar ] ) . given the range of colors that can be simulated using the two grain models , it could be expected that a grain population with properties intermediate of those in the kmh and mw models would result in neutral near - infrared colors , as observed . an additional effect of multiple scattering observed in our simulations is that the disk inclination and optical thickness combine to modify the azimuthal intensity variation from that expected from single scattering . given that the observed surface brightness is dominated by the least scattered photons ( @xmath97 dependency ) , for grains that are predominantly forward - scattering , photons received from the back side of the ring are biased towards the back scattering part of the phase function . statistically , most of the photons will be scattered towards the middle of the ring and will either not reach the observer or be of an extremely low intensity . scattering from the front edge of the ring , however , will remain mostly unbiased . this biasing effect will cause the derived @xmath84 value measured in [ azvar ] to underestimate the true dust asymmetry , and hence our value is only a lower limit . the same reasoning applies to the close et al.s analysis , as a consequence of considering only single scattering . as an example of this effect , in our simulations , the size of the median scattering grain is 0.56 and these have an asymmetry parameter @xmath98 at 1 . however , the measured @xmath84 from the simulations ( obtained using the same method as in [ azvar ] ) is only @xmath99 . the observed lower limit on g , measured in [ azvar ] provides a lower limit on the median size of the scatterers as the asymmetry value is a sensitive function of the size parameter , @xmath100 ( @xmath101 ) , for @xmath84 less than @xmath00.8 . from the mw grain properties used , we find that the measured asymmetry of @xmath102 at 1 corresponds to a grain size , @xmath103 0.23 . the gg tau disk is clearly detected ( @xmath104 ) in all three nicmos filters between radii of 10 to 22 , scattering approximately 1.5% of the stellar light . no discernible wavelength dependence of scattered light is observed , although azimuthal variations in both width and intensity are seen and modeled . we confirm the observation of a kink in the disk by silber et al . ( 2000 ) . observed as a sharp elbow in the disk on the south - east edge , followed by a flattening of the isophotes along the southern edge of a disk , this feature is seen in all three filters . the projected distance of the northern edge from the center of mass , related to the total height of the ring , provides the first hint that the ring may not be as thick as previously estimated , with a total height of 35au at the inner edge of the disk ( 180au ) . the presence of such a well - observed circumbinary ring around this system , combined with the 9 year baseline of binary observations has allowed us to constrain the orbital parameters of the binary . we find a slightly elliptical orbit @xmath105 , with a semi - major axis @xmath44au and a corresponding orbital period of @xmath45yr . the 1@xmath4 range of estimated orbital semi - major axis is such that the ratio of semi - major axis to inner disk radius is larger than expected from previous sph simulations ( artymowicz & lubow , 1994 ) . this suggests that the dynamical evolution of the circumbinary ring may be more complex than previously thought . further astrometric observations of this close binary are needed to constrain the binary orbit further . the monte carlo simulations summarized in [ mcmodel ] show the wide range of colors one can ` predict ' from the gg tau disk , from blue to red , using different dust size distributions and grain properties ( @xmath36 , @xmath84 ) . specifically , our monte carlo simulations show that the observed neutral disk colors , ratio of disk to star light , and near side to far side flux ratio can probably be reproduced with some combination of an ism - like grain size distribution and dust properties and does not necessarily require either the presence of large dust grains or prior extinction by circumstellar disks . however , this does not rule out that either is present . because the disk is optically thick in the near - infrared , studies of scattered light are only sensitive to those grains that are located near the surface of the ring . any conclusion about grain growth only applies to those regions and does not provide any information the size of grains deep in the ring . for instance , it has been shown that grains larger than @xmath106 tend to segregate in the disk midplane because of their larger mass ( e.g. , suttner & yorke 2001 ) and possibly as the result of the enhanced grain - grain collision rate in higher density regions . the presence of such large grains in the midplane can not be determined in our near - infrared observations . circumstellar disks are known to surround both components of gg tau a , however , the possible effects they have on the integrated circumbinary disk colors have been shown to be subtle and somewhat surprising . while krist et al . ( 2002 ) suggest that the red disk colors observed in the optical could be caused by an @xmath107 1.2 mag of extinction , possibly in the form of circumstellar disks , wood et al . ( 1999 ) find that the inclusion of small circumstellar disks can actually cause the integrated disk colors to become _ bluer _ rather than redder . analysis of the disk colors at both optical and near - infrared wavelengths is planned for future work . while the color of scattered light in optically thick disks can not currently be used to constrain grain properties , the azimuthal intensity variations can be used to provide a lower limit on the median grain size , independent of grain model . this lower limit ( @xmath108 at 1 ) corresponds to a median grain size of @xmath3 . given that the ism dust models produce a wide range of median grain sizes ranging from @xmath00.16 ( mathis , rumpl & nordsieck 1977 ; kmh ) to 0.56 ( mw ) , this lower limit does not suggest that grain growth has occured in the surface layers of the disk in this young ( @xmath01 myr old ; white et al . 1999 ) system . there is still considerable uncertainty regarding the ism dust properties ( e.g. , witt 2000 ) , a point which is well illustrated here by the comparison of monte carlo scattering results using two of the more well known dust grain models . both the kmh and mw grain models , like most of the ism dust models to date , have been constrained using the ism extinction curve , which they reproduce equally well . the results presented here suggest that using the extinction curve _ alone _ to constrain the dust properties introduces a significant level of uncertainty in the scattering properties of the grains . as an example , kmh and mw grain models predict v band @xmath84 s of @xmath109 and @xmath110 respectively . observations of reflection nebulae in the same bandpass infer @xmath111 , significantly different from either of these models ( witt , oliveri & schild , 1990 ) . additionally , models of t tauri disks seen in scattered light with wfpc2 find that the observations are best fit by @xmath112 ( burrows et al . 1996 ; stapelfeldt et al . 1998 ; krist et al . 2002 ) , although these values may be affected by grain growth . combining additional sources of information , such as observations of dust scattering ( e.g. , witt , et al . 1990 ) , polarization ( e.g. , zubko & laor 2000 ) , and dust thermal emission ( e.g. , li & draine 2001 ) will provide additional constraints . without a more detailed knowledge of the ism dust grain properties it is unlikely that we will be able to unambiguously determine whether grain growth is occuring in t tauri disks through scattered light imaging . although dust properties in the disk can not currently be constrained from either colors or dust asymmetry alone , this work provides an outline for future analysis and modeling . observationally , further resolved intensity and polarization maps at other wavelengths are key in sampling a large enough grain size parameter ( @xmath100 ) range to constrain the dust grain properties . for instance , at 3 to 5 the dust asymmetry is expected to be much more isotropic , allowing significant color variations ( e.g. , @xmath113 ) between the front and back sides , unless significant grain growth has already occurred . such observations should be carried out in parallel with a full exploration of the parameter space ( grain size distribution , grain properties , ring geometry ) through monte carlo scattering simulations . the authors thank mike jura , alycia weinberger , franois mnard , kenny woods , lisa prato , and angelle tanner for enlightening discussions . we also thank the anonymous referee for their constructive comments . support for this work was provided by nasa through grant number go-06735.01 - 95a from the space telescope institute , the nasa astrobiology institute and the packard foundation . this research was based on observations made with the nasa / esa hubble space telescope , obtained at the space telescope science institute , which is operated by the association of universities for research in astronomy , inc , under nasa contract nas5 - 26555 . this research has made use of the simbad database , operated at cds , strasbourg , france . in this appendix , we present the method used to derive the orbital parameters of the binary from a single measurement of its separation and relative velocity with some additional knowledge available from previous studies . the orbit of a binary system can be described with a set of 9 independent parameters : mass ratio ( @xmath114 ) , eccentricity ( @xmath32 ) , total mass ( @xmath39 ) , orbital period ( @xmath34 ) , time of periastron ( @xmath115 ) , inclination of the orbital plane ( @xmath40 ) , position angle on the sky of the ascending node of the orbit ( @xmath42 ) , angle in the orbital plane between periastron and the ascending node ( @xmath36 ) , and distance to the system ( @xmath41 ) . note also that @xmath34 and @xmath39 are linked to the semi - major axis of the orbit , @xmath33 , through kepler s third law . when considering the _ relative _ motion of one component around the other one , @xmath114 is no longer a relevant parameter . therefore , one must measure 8 independent quantities to solve the orbit unambiguously . in the case of ggtau , the millimeter interferometric measurement of the circumbinary ring by g99 yields precise estimates of three independent parameters : the inclination of the ring , @xmath116 , the position angle of the apparent semi - minor axis of the ring , @xmath117 , and the total system mass @xmath118 . if the ring is intrinsically circular then @xmath119 ( depending on the motion of the binary ) . furthermore , if we assume that the orbital plane corresponds to the ring mid - plane , we have @xmath120 . these two assumptions ( circular ring and coplanar orbit ) are the only assumptions we make in the following analysis . additionally , the hipparcos determination of the distance to the taurus star - forming region , @xmath121pc , by bertout et al . ( 1999 ) , allows us to convert the apparent separation and velocity of the binary into absolute values . prior to our work , the problem could thus be simplified to a problem with only four unknowns . the orbit of the system has been followed over the last ten years or so and , since all measurements so far are consistent with a uniform linear motion , we combined them all to obtain the average binary separation and linear velocity projected on the sky . these quantities can be represented for instance by the respective amplitude of the binary separation and of its velocity ( @xmath122 and @xmath123 , measured in arcsec and arcsec.yr@xmath28 ) and by the angle between these two vectors in the plane of the sky ( @xmath124 ) . in the following derivation , a useful quantity is the angle @xmath76 between the apparent semi - minor axis of the ring and the binary separation , which represents the polar coordinate of the current measurement with respect to the disk itself . this angle can be expressed as @xmath125 . our data imply @xmath126 , @xmath127 , @xmath128 and @xmath129 ( see [ orbit ] ) . the orientation of the relative velocity implies that @xmath130 . we have thus obtained the four last measurements needed to solve the whole orbit . the two - body problem is a classical mechanics problem , the solution of which can be found in many textbooks . we are interested here in the case where only relative measurements of the 3d binary separation ( @xmath131 ) and velocity ( @xmath132 ) are available from the de - projection of the observed separation and motion in the orbital plane . the eccentricity and semi - major axis can then be derived from measurements of the angular momentum , @xmath133 , and total energy , @xmath134 , _ per unit mass_. we have the following relations : @xmath135 and @xmath136 . these relations can be translated into the observables presented above as follows : @xmath137 -2\frac{gm\cos i}{\rho d } \right\}\ ] ] and @xmath138 \right\}^{-1}\ ] ] one can immediately derive the orbital period : @xmath139 . uncertainties for all of these quantities were obtained by allowing all seven observables to vary within 1@xmath46 of their nominal value and retaining the extreme values for each orbital parameter . special care must be taken when considering the uncertainties induced by the distance estimate , as the total mass of the system is also affected in a _ systematic _ manner by the distance . the distance and relative velocity of the binary estimates combine for the largest part of the total uncertainty for the orbital elements . for completeness , we derive the remaining two orbital parameters that have not been addressed so far , @xmath36 and @xmath115 . the former is the angle , in the orbital plane , between periastron and the ascending node . this can be derived from the angular distance between the current location of the companion and periastron . the latter angle can be determined using the general equation of the elliptical orbit , @xmath140 where @xmath141 and @xmath142 are the polar coordinates of a point running along the orbit and @xmath143 is the position angle of periastron . using the current de - projected separation of the system , 43.0au , we find that the binary is about @xmath144 away from periastron . combining this information with the angle between the current binary location and the ascending node ( @xmath145 ) , we conclude that @xmath146 . to estimate how much time it takes for the companion to go from one point to another along the orbit , we use kepler s second law and numerically integrate the area @xmath147 defined by the section of an ellipse between these two points . we can easily calculate the time derivative of the area , which is related to the angular momentum per unit mass by @xmath148 . applying this to the average position of the companion and periastron , we find that the next periastron passage will occur about 61yrs from now or , equivalently , that @xmath149jd . both @xmath36 and @xmath115 depend heavily on the previously derived quantities , therefore , we emphasize that the uncertainties on these parameters are rather large . in this section we derive a relation between position angle around a disk and the angle a photon is scattered through . at each position angle around the disk , we assume that the observed photons are scattered off the closest disk edge which corresponds to the smallest available scattering angle ( see [ azvar ] ) . consider a circular ring that is inclined to the line of sight at an angle @xmath40 . this ring is projected on the sky as an ellipse ( see figure [ model]a ) , with normalized semi - major axis , @xmath150 and aspect ratio @xmath151 . the position angle on the sky of a random point around the disk , @xmath77 , is defined as @xmath152 . in figure [ model]a we have rotated the image so that the semi - minor axis of the disk is now vertical . the amount of rotation , @xmath70 , is the @xmath77 of the semi - minor axis of the disk , defined in the usual manner . the scattering angle is defined as the angle between the pre- and post- scattering directions , @xmath153 and @xmath154 respectively , as shown in figure [ model]b . this angle , @xmath155 , is @xmath156 , where @xmath83 is the angle subtended by the disk height . if the disk can be assumed to be flat , we have @xmath157 . we first consider this case before generalizing our results to @xmath158 . the coordinates of the unit vector @xmath153 , @xmath159 are along the photon s initial path to a random point on the ring , @xmath160 , in the @xmath161 frame are : artymowicz , p. , & lubow , s.h . , 1994 , , 421 , 651 beckwith , s. v. w. , sargent , a. i. , chini , r. s. & guesten , r. , 1990 , , 99 , 924 bertout , c. , robichon , n. , & arenou , f. , 1999 , , 352 , 574 bohren & huffman , 1983 , scattering of light from small particles bushouse , h. , skinner , c. , & mackenty , j. , 1996 instrument science report isr nicmos-009 burrows , a. , stapelfeldt , k. , watson , a. , krist , j.e . , ballester , g.e . 1996 , , 473 , 437 close , l.m . , dutrey , a.m. , roddier , f. , guilloteau , s. , roddier , c.,duvert , g. , northcott , m. , mnard , f. , graves , j.e . & potter , d. 1998 , , 499 , 883 duchne , g. phd , universit de grenoble , 1999 dutrey , a. , guilloteau , s. , and simon , m. 1994 , , 286 , 149 ghez , a.m. , neugebauer , g. and matthews , k. 1993 , , 106 , 2000 ghez , a.m. , weinberger , a.j . , neugebauer , g. , matthews , k. , mccarthy , d.w . , jr . , 1995 , , 110 , 753 ghez , a.m. , white , r. and simon , m. 1997 , , 490 , 353 guilloteau , s. , dutrey , a. , and simon , m. 1999 , , 348 , 570 * ( g99 ) * henyey , l.g . , & greenstein , j.l . , 1941 , , 93 , 70 holtzmann , j. a. et al . 1995 , , 107 , 156 jensen , e.l.n . , donar , a. x. , & mathieu , r. d. , 2000 , iau symp . 200 , the formation of binary stars , 85 kawabe , r. , ishiguro , m. , omodaka , t. kitamura , y. & miyama , s. m. 1993 , , 404 , l63 kim , s. , martin , p.g . , & hendry , p.d . , 1994 , , 422 , 164 * ( kmh ) * krist , j. 1993 , a.s.p . conference series , astronomical data analysis software and systems ii , vol . 2 , 536 krist , j. & hook , r.n . , 1999 , the 1997 hst calibration workshop with a new generation of instruments , 192 krist , j.e . , stapelfeldt , k.r . & watson , a.m. , 2002 , , in press leinert , c. , haas , m. , mundt , r. , richichi , a. , & zinnecker , h. , 1991 , , 250 , 407 leinert , c. , zinnecker , h. , weitzel , n. , christou , j. , ridgway , s.t . , jameson , r. , haas , m. & lenzen , r. , 1993 , , 278 , 129 li , a. , & draine , b.t . , 2001 , , 554 , 778 mathis , j.s . , rumpl , w. , & nordsieck , k.h . , 1977 , , 217 , 425 mathis , j.s . , & whiffen , g. , 1989 , , 341 , 808 * ( mw ) * mnard , f. phd , universit de montreal , 1990 monin , j .- l . , mnard , f. , & duchne , g. , 1998 , , 339 , 113 roddier , c. , roddier , f. , northcott , m.j . , graves , j.e . , and jim , k. 1996 , , 463 , 326 * ( r96 ) * silber , j. , gledhill , t. , duchne , g. , mnard , f. , 2000 , , 536 , l89 simon , m. , & guilloteau , s. , 1992 , , 397 , l47 stapelfeldt , k.r . , krist , j.e . , mnard , f. , bouvier , j. , padgett , d.l . , & burrows , c.j . , 1998 , , 502 , l65 suttner , g. & yorke , h.w . , 2001 , apj , 551 , 461 thompson , r. , 1995 , , 2475 , 367 walter , f. , brown , a. , mathieu , r.d . , myers , p.c . , and vrba , f.j . 1988 , , 96 , 297 white , r.j . , ghez , a.m. , reid , n. , schultz , g. 1999 , , 520 , 811 white , r.j . , & ghez , a.m. , 2001 , , 556 , 265 witt , a.n . , & oshel , e.r . , 1977 , , 35 , 31 witt , a.n . , 1985 , , 294 , 216 witt , a. , oliveri , m.v . , & schild , r.e . , 1990 , , 99 , 888 witt , a. , 2000 , iau symp . 197 , astrochemistry : from molecular clouds to planetary systems , 317 woitas , j. , khler , r. , & leinert , ch . , 2001 , , 369 , 249 wood , k. , crosas , m. & ghez , a. , 1999 , , 516 , 335 wood , k. , smith , d. , whitney , b. , stassun , k , kenyon , s.j . , wolff , m.j . , & bjorkman , k.s . , 2001 , , 561 , 299 zubko , v.g . , & laor , a. , 2000 , , 128 , 245 cllcccc f110w & 0.248 @xmath53 0.002 & 353.9 @xmath53 0.4 & 9.37 @xmath53 0.05 & 10.27 @xmath53 0.07 & 8.98 @xmath53 0.02 & 13.59 @xmath53 0.01 + f160w & 0.251 @xmath53 0.002 & 353.8 @xmath53 0.3 & 8.39 @xmath53 0.01 & 9.08 @xmath53 0.04 & 7.91 @xmath53 0.02 & 12.42 @xmath53 0.01 + f205w & 0.257 @xmath53 0.003 & 353.6 @xmath53 0.8 & 7.94 @xmath53 0.01 & 8.61 @xmath53 0.01 & 7.46 @xmath53 0.02 & 12.01 @xmath53 0.05 + lll|llll @xmath174 & 0.014 @xmath53 0.0003 & 0.0043 @xmath53 0.0012 & 0.007 & 0.012 & 0.030 & 0.008 + @xmath175 & 0.10 @xmath53 0.03 & 0.65 @xmath53 0.32 & 0.50 & 0.47 & -0.07 & -0.31 + @xmath176 & -0.04 @xmath53 0.06 & -0.34 @xmath53 0.45 & 0.28 & 0.20 & -0.21 & -0.20 + @xmath177 & 1.41 @xmath53 0.03 & 3.6 @xmath53 1 & 1.90 & 2.35 & 0.67 & 4.11 + @xmath178 & 1.39 @xmath53 0.02 & 2.6 @xmath53 0.25 & 1.82 & 2.03 & 0.67 & 4.07 + @xmath179 & & 1.5 @xmath53 0.5 & 1.70 & 1.741 & 0.67 & 3.65 +

we present a multi - wavelength photometric study of,000 resolved stars in the nearby spiral galaxy m83 ( ngc5236 ,mpc ) based on wide field camera 3 observations using four filters : f336w , f438w , f555w , and f814w . we select 50 regions ( an average size of 260pc by 280pc ) in the spiral arm and inter - arm areas of m83 , and determine the age distribution of the luminous stellar populations in each region . this is accomplished by correcting for extinction towards each individual star by comparing its colors with predictions from stellar isochrones . we compare the resulting luminosity weighted mean ages of the luminous stars in the 50 regions with those determined from several independent methods , including the number ratio of red - to - blue supergiants , morphological appearance of the regions , surface brightness fluctuations , and the ages of clusters in the regions . we find reasonably good agreement between these methods . we also find that young stars are much more likely to be found in concentrated aggregates along spiral arms , while older stars are more dispersed . these results are consistent with the scenario that star formation is associated with the spiral arms , and stars form primarily in star clusters and then disperse on short timescales to form the field population . the locations of wolf - rayet stars are found to correlate with the positions of many of the youngest regions , providing additional support for our ability to accurately estimate ages . we address the effects of spatial resolution on the measured colors , magnitudes , and age estimates . while individual stars can occasionally show measurable differences in the colors and magnitudes , the age estimates for entire regions are only slightly affected .

1204.6045

understanding the properties of stars and the history of star formation in galaxies remains one of the most fundamental subjects in astrophysics . the _ hubble space telescope _ ( @xmath7 ) provides an important tool for this endeavor , since it enables the detailed study of stars and star clusters , not only in the milky way and its nearest neighbors , but in galaxies well beyond the local group . the wide field camera 3 ( wfc3 ) provides a particularly valuable new panchromatic imaging capability , with spectral coverage from the near - uv to the near - ir . this capability is especially useful for studying the stellar populations in nearby galaxies where individual stars are resolved . a good example of a multi - wavelength survey of nearby galaxies which employs resolved stars and star clusters is the acs nearby galaxy treasury ( angst ) program @xcite . this study includes @xmath065 galaxies out to @xmath03.5 mpc , and provides uniform multi - color ( @xmath8 ) catalogs of tens of thousands of individual stars in each galaxy . although the angst program provides an excellent survey for a wide range of studies , it does not provide observations in the @xmath9-band since the observations were obtained before wfc3 was installed on @xmath7 . the @xmath9-band is particularly useful for age - dating populations of young stars , which is the focus of the current paper . phat ( panchromatic hubble andromeda treasury ) is a related study that does take advantage of the new @xmath9-band capability on wfc3 . while nowhere as extensive as the multiple cycle treasury program phat survey , the current study of m83 is complementary in the sense that it provides similar observations of the resolved stellar component of a nearby spiral galaxy , and uses quite different analysis techniques , as will be discussed in sections 2 and 3 . future observations of m83 ( id:12513 , pi : william blair ) will expand the wfc3 dataset available for m83 from 2 to 7 fields . m83 ( ngc5236 ) , also known as the `` southern pinwheel '' galaxy , is a slightly barred spiral galaxy with a starbursting nucleus located at a distance of @xmath3mpc , i.e. , @xmath10 @xcite . the h@xmath11 emission can be used to pinpoint regions of recently formed massive stars along the spiral arms , while red supergiants can be found throughout the galaxy . the color - magnitude diagram ( cmd ) of resolved stars is a powerful diagnostic tool for understanding the stellar evolution and history of star formation of galaxies in detail . by comparing stellar evolution models to observed cmds , we are able to determine the ages of the stellar populations in galaxies . however , in active star - forming regions , the stars are often partially obscured by dust . applying a single extinction correction value for the whole galaxy often results in over- or under - estimates of the ages of individual stars . if the cmd is the only tool used to determine the ages , the spatial variations of dust extinction in galaxies and their effects on the determined ages are not readily apparent . the additional information available from the color - color diagram can remedy this problem . by using techniques developed in this paper , we can constrain the variations of dust extinction across m83 , and make corrections for individual stars . we also focus on spatial variations in the stellar properties throughout the galaxy . the @xmath7 wfc3 observations of m83 were performed in august 2009 as part of the wfc3 science oversight committee ( soc ) early release science ( ers ) program ( id:11360 , pi : robert oconnell ) . the central region ( 3.2kpc@xmath123.2kpc ) of m83 was observed in august , 2009 . a second adjacent field to the nnw was observed in march 2010 , and will be included in future publications . details of the wfc3 ers data calibration and processing are given by @xcite and @xcite . in this paper we present @xmath13 photometry of resolved stars in m83 and the resulting color - color and color - magnitude diagrams . we use @xmath14 vs. @xmath15 color - color diagrams to constrain the variation of dust extinction along the line - of - sight of each individual star , and correct @xmath16 vs. @xmath15 cmds for the extinction of individual stars to determine their ages . the high sensitivity and the superb resolving power in the wfc3 f336w - band plays a key role in allowing us to develop our extinction correction techniques , and demonstrate the performance of the newly installed wfc3 . this results in improved stellar age estimates , and a better understanding of the recent star - formation history of m83 . our investigations are focused on the followings : 1 . do we see spatial variations of stellar ages in m83 ? if so , can we use these variations to learn more about the evolution of the galaxy and what triggers star formation ? 2 . how well do various age estimates correlate ( i.e. , resolved stars , integrated light , clusters , number ratio of red - to - blue stars , h@xmath11 morphology , stellar surface brightness fluctuations , presence of wolf - rayet stars ) ? this paper is organized as follows . in 2 , we describe the observations , photometric analysis , and the extinction correction method . cmds and color - color diagrams of the 50 selected regions are then used to provide stellar age estimates , as described in 3 . comparisons with other age estimates ( i.e. , integrated light and star clusters ) are also made . in 4 , we compare the stellar ages to a variety of other parameters that correlate with age , including red - to - blue star ratios , @xmath17 morphology , and surface brightness fluctuations . section 5 includes a discussion of how these comparisons can be used as diagnostics , with special attention paid to the question of what spatial variations can tell us about the star formation history in m83 , and on the question of how star clusters might dissolve and populate the field . a summary of our primary results is provided in 6 . in appendix a , we investigate the locations of sources of wolf - rayet stars in our m83 field and discuss the correlation with the positions of young regions . in appendix b , we describe how the spatial resolution affects our measured colors and magnitudes , and our age estimates of stars in our m83 field . the data were obtained in seven broad- and seven narrow - band filters in the wfc3 uvis and ir channels . details are described in @xcite and @xcite . in the current paper we use four broad - band images . the filters and exposure times are f336w ( 1890 sec ) , f438w ( 1880 sec ) , f555w ( 1203 sec ) , and f814w ( 1203 sec ) . although we do not make transformations to the johnson - kron - cousins @xmath13 system , we will refer to these filters as @xmath9 , @xmath18 , @xmath19 , and @xmath20 for convenience . three exposures were taken in each filter at different dithered positions to remove cosmic rays , the intrachip gap , and to partially compensate for the undersampled point - spread function ( psf ) of the wfc3/uvis channel . the raw data were processed using the multidrizzle software @xcite with an effective pixel size of 0@xmath210396 , which corresponds to 0.885 pc per pixel at the distance of 4.61mpc @xcite . the resulting multidrizzled images are combined , cosmic - ray removed , and distortion - corrected . a color composite of the f336w , f555w , f814w ( broad bands ) and f502n , f657n ( narrow bands ) images is shown in figure [ fig : m83_reg50 ] . it covers the nuclear region of m83 , part of its eastern spiral arm , and the inter - arm region . stars in the cores of compact star clusters are not resolved at this resolution , but the brighter stars in the outskirts of the star clusters and in the field are generally resolved into individual stars . details of the effects of spatial resolution are discussed in appendix b. since we are interested in spatial variations of stellar ages in m83 , we selected 50 regions in the spiral arm and inter - arm areas . boxes outlined in blue ( very young age@xmath2210myr , see 3 ) , yellow ( young 10@xmath22age@xmath2220myr ) , and red ( intermediate - aged age@xmath2320myr ) in figure [ fig : m83_reg50 ] show the 50 regions selected for detailed study in this paper . photometric analysis of the wfc3 m83 data was performed on the @xmath9 , @xmath18 , @xmath19 , and @xmath20 images using the dophot package @xcite with modifications made by a. saha . additional routines to derive parameters for the kurtosis of the analytic psfs used by dophot , and for post - processing dophot output to obtain calibrated aperture - corrected photometry , were performed using customized idl code written by saha . a more detailed description of these procedures can be found in @xcite . the kurtosis terms were derived by fitting the functional form of the dophot analytic psf to 20 relatively bright and isolated stars . the remaining shape parameters ( @xmath24 , @xmath25 , and @xmath26 ) are dynamically optimized within dophot . the difference between the shape parameters for an individual object and those for a typical star was calculated and used to classify the object as a star , a galaxy , or a double - star . the shape parameters were determined separately for each image . the numbers of objects classified as stars are approximately 20,000 in @xmath9and @xmath18 , 17,000 in @xmath19 , and 15,000 in @xmath20band images , respectively . the calculated full width at half maximum ( fwhms ) are @xmath9=0@xmath21096 ( 2.42 pixel ) , @xmath18=0@xmath21093 ( 2.34 pixel ) , @xmath19=0@xmath21091 ( 2.29 pixel ) , and @xmath20=0@xmath21113 ( 2.85 pixel ) . all magnitudes are in the wfc3-uvis vegamag magnitude system , calculated using equation ( 4 ) in @xcite and the latest zeropoint magnitudes : @xmath9=23.46 , @xmath18=24.98 , @xmath19=25.81 , and @xmath20=24.67 mag , provided by stsci at the @xmath7/wfc3 website . a multi - wavelength @xmath13 catalog was then constructed by matching the individual objects in each catalog between these filters . interestingly , we found that the extreme image crowding in these images , combined with the excellent panchromatic sensitivity and unprecedented spatial resolution provided by the newly installed wfc3 , introduced a new challenge during this matching step , with large numbers of `` pseudo '' matches occurring if too - large a match radius was used . we discuss the effects of spatial resolution in appendix b. figure [ fig : pseudo ] demonstrates this problem by showing the @xmath27 and color composite image cut - outs near regions # 26 and # 27 . in the middle panel of figure [ fig : pseudo ] , stars circled in red are objects with @xmath28 colors that are very red ( @xmath29 mag ) , but have @xmath30 values that are fairly blue ( @xmath31 mag ) . these can be seen as the spray of points in the upper right parts of the top right panels of figures [ fig : allstar_no ] and [ fig : allstar_cor ] , which will be discussed below . as shown by the postage stamp images in figure [ fig : pseudo ] , about half of these are likely to be blue stars with very high values of reddening ( i.e. , the bottom panel and the small postage stamps on the right show that they are in dusty regions ) , while the other half are very close superpositions of at least one red and one blue star , which we will call `` pseudo '' matches ( i.e. , the panels on the left show a strong blue and red gradient across many of the objects ) . we found that using a very small matching radius of 0.5 pixels was needed to minimize `` pseudo '' matches in our catalog . earlier attempts using a matching radius of 3 pixels resulted in @xmath010 % pseudo matches . matching also requires precise geometric distortion corrections in all filters . however , even when these conditions are met , there is a finite chance that two different stars will fall within the same aperture . this is most easily seen by blinking the @xmath9- and @xmath20-band images . while the single stars stay in the same position , the pseudo match stars move slightly , showing they are two different stars with different colors . the final stellar catalog contains @xmath015,000 objects . we find that our procedure results in only a few percent of pseudo matches , as will be discussed in 2.4 ( i.e. , the upper right panel of figure [ fig : allstar_cor ] ) . to measure the photometric completeness , we performed artificial star tests with the same detection and photometry procedure as applied to the actual stars . we inserted 1000 artificial stars with gaussian psfs and fwhms appropriate for these stars at random positions into all four images . the magnitude of the inserted stars varied from 22 to 28 mag in steps of 0.25 mag . the 50% completeness levels at 5 @xmath32 detection thresholds for a typical region are reached at recovery magnitudes of approximately 24.3 , 25.1 , 25.3 , and 24.8 mag for the @xmath9 , @xmath18 , @xmath19 , and @xmath20 images , respectively . these limits can vary by about a magnitude , depending on the brightness of the background in a given region . the exception is region # 48 , which includes the nucleus . the completeness thresholds are two or more magnitudes brighter in this region due to the very high background . this region has therefore been excluded from the rest of the discussion , but is included in table [ tab:50reg ] for reference . the top panels of figure [ fig : allstar_no ] show the @xmath20 vs. ( @xmath33 ) cmd ( _ left _ ) and the @xmath30 vs. @xmath28 color - color diagram ( _ right _ ) of resolved stars extracted from the cross - matched catalog . the cmd shows the presence of young main - sequence ( ms ) stars , transition or he - burning `` blue loop '' stars , red giant stars ( hydrogen shell - burning and hydrogen@xmath34helium shell - burning stars ) , and a relatively small number of low - mass red - giant branch ( rgb ) stars . this is because the tip of the low - mass rgb stars is at @xmath20= 24.7 mag @xcite , which is roughly the same as our completeness threshold . the ages of these stars range from @xmath35 1 to @xmath35 100 myr . the padova isochrones @xcite ranging in age from log @xmath36 ( age / yr ) = 6.55 ( 3.5myr ) to log @xmath36 ( age / yr ) = 8.0 ( 100 myr ) are included in figure [ fig : allstar_no ] for reference . we note that the younger isochrones ( e.g. , 1myr ) fall nearly on top of the 3.5 myr isochrone , and hence are not included in the diagram . the arrows in both panels show @xmath37 reddening vectors for m83 using the @xmath38 extinction curve of @xcite . corrections have been made for foreground galactic extinction @xcite using @xmath39 , @xmath40 , @xmath41 , and @xmath42 mag , respectively . in figure [ fig : allstar_no ] , we only plot stars with photometric errors less than 0.25 mag in the filters used for each diagram . the numbers of objects plotted in the cmd and the color - color diagram are about 12,000 and 8,500 stars , respectively . since approximately 30% of the stars are not detected in one or more of the @xmath9 , @xmath18 , or @xmath20 filters , due to either reddening or intrinsically red or blue colors , fewer stars are plotted in the color - color diagram . we adopt a value of 1.5 times solar metallicity ( z=0.03 ) , the highest metallicity isochrones available from the padova database , for m83 based on @xcite . a test using solar metallicity ( z=0.019 ) isochrones showed that the ages for a typical region would be @xmath35 2 myr older if the lower metallicity is adopted . this agrees with the study by @xcite , who found that the simulated cmds of young massive clusters in m83 with solar and super - solar metallicity isochrones would not look much different . theoretical isochrones calculated for the wfc3 filters and z=0.03 ( 1.5 z@xmath43 ) from the padova database @xcite are overlaid on both panels in figure [ fig : allstar_no ] . the dashed line in cyan shows the 50% completeness level in @xmath20 and @xmath28 , using the completeness threshold numbers from the artificial star test in 2.2 . the bottom panels of figures [ fig : allstar_no ] and [ fig : allstar_cor ] show the histograms of ages and masses for the same stars plotted in the cmd and color - color diagram . the ages and masses of stars were estimated by finding the closest match between the @xmath20 and @xmath28 values for each star , using a fine mesh of the stellar isochrones plotted in the cmd and color - color diagram in figures [ fig : allstar_no ] and [ fig : allstar_cor ] . more details about the stellar age - dating from the cmds and the stellar isochrones are given in 3.1 . since m83 has an intricate structure of dust lanes associated with active star forming regions in the spiral arms and thin layers of dust in the inter - arm area in figure [ fig : m83_reg50 ] , we can not use a single value of internal extinction and apply it to correct for the extinction of all stars in a given region . in the color - color diagram in figure [ fig : allstar_no ] , we notice that while the bluest stars match the model isochrones quite well after we correct for galactic foreground extinction , the majority of stars are found redward of the models , implying a typical extinction of @xmath44 mag ( for the densest part of the data `` swarm '' ) . as described below , we can determine the reddening of these stars _ if _ we assume that they are intrinsically blue stars that belong on the padova models ( i.e. , with @xmath45 mag ) . a visual inspection supports this interpretation . stars with observed values of @xmath46 mag are found in areas with no obvious dust features surrounding them while stars with @xmath47 mag are near dust filaments . our basic approach to correct the colors ( and luminosities ) of individual stars for the effects of extinction is to backtrack the position along the reddening vector for each data point , until we hit the stellar isochrones in the color - color diagram . what makes this method work is the fact that all the isochrones are nearly on top of each other in the range @xmath48 mag . the way we apply this in practice is to match the observed and predicted values of the reddening free @xmath49 parameter defined by : @xmath50 @xcite . using observed values of @xmath13 , we compute the observed @xmath51 values by using the standard slope of @xmath52 @xcite , which corresponds to @xmath53 . the predicted @xmath49 values are calculated using the 1.5 z@xmath54 padova isochrones . a complication is that we can not simply use the matched @xmath49 values to determine corrected values of @xmath30 and @xmath28 , since they would , by definition , fall precisely on the isochrones ( i.e. , we would be using circular reasoning ) . instead , we solve for the extinction in @xmath55 , which is partly independent and uses the longest wavelength baseline . we then calculate the extinction values in @xmath13 and the color excess values in @xmath30 and @xmath28 . the averages of the corrected internal extinction and color excess values by our star - by - star correction method are @xmath56 0.696 , @xmath57 0.559 , @xmath58 0.441 , @xmath59 0.256 mag , and @xmath60 0.137 and @xmath61 0.185 mag . this is similar to the average internal extinction ( @xmath62 mag ) estimated from the cluster sed fitting by @xcite . however , our results differ from the average extinction ( @xmath63mag ) of 45 clusters in the nucleus ( @xmath64 in diameter ) of m83 , determined from @xmath65 ratios by @xcite . this is consistent with the finding of larger extinction in the nuclear region by @xcite . in the color - color diagram ( top right panel ) of figure [ fig : allstar_no ] , we note that while most of the data points are consistent with the standard reddening vector , the stars located in the green triangle appear to follow a flatter reddening vector . whether this is actually due to a different reddening law in m83 ( e.g. , as has been suggested for heavily extincted regions such as # 4 , # 12 , and # 20 ) , or is some sort of artifact ( e.g. due to photometric uncertainties , unresolved star clusters , or the `` pseudo '' matching problem discussed above ) is difficult to determine from our present dataset . however , for our specific needs the precise answer to this question is not critical . our approach will be to correct the data points in the triangle back to a position near the top of the isochrones , as shown by the intersection of the red dotted line and the isochrones in figure [ fig : allstar_cor ] this is equivalent to using a range in @xmath66 between 3.1 and 5.7 ( the value represented by the red dotted line ) . the upper panels of figure [ fig : allstar_cor ] show the cmd and color - color diagram corrected for the internal extinction in m83 . the arrows in both panels are the reddening vectors . we do not apply an extinction correction for stars with @xmath67 mag , or above an extrapolation of the flatter extinction vector from the bluest possible isochrones ( see the red dotted lines in the upper right panel of figure [ fig : allstar_cor ] ) , since we believe that the colors of many of these sources are incorrect ( i.e. , roughly half are likely to be pseudo matches , as discussed in 2.2 and shown in figure [ fig : pseudo ] ) . we note that only about 4% of the stars fall in this part of the diagram . these stars are removed from the subsequent analysis . if we include these stars , our age estimates typically increase by about 1myr . also , stars below the blue dotted line in figure [ fig : allstar_cor ] are not corrected for extinction , since these are main sequence turnoff stars ( i.e. , blue loop or transition stars in general ) , and hence , do not have unique @xmath13 colors for the stars along a given reddening line , unlike the main sequence stars with @xmath48 mag . based on the extinction corrected cmd in figure [ fig : allstar_cor ] , we note that : ( 1 ) the corrected data now shows a relatively narrow distribution of points along the left side of the cmd in agreement with ages of @xmath03 myr , and ( 2 ) while the data now aligns with the model isochrones much better , there is still a fair amount of scatter ( especially for @xmath20 @xmath68 24 ) , primarily due to observational uncertainties from the fainter stars . the bottom panels of figure [ fig : allstar_cor ] show the distribution of ages and masses of stars after we apply our extinction correction technique . based on the automatic method described in 3.1 , we note that the luminosity weighted mean ages determined from the corrected colors and luminosities are significantly younger ( @xmath014 myr ; figure [ fig : allstar_cor ] ) than the mean ages determined from the uncorrected colors and luminosities ( @xmath027 myr ; figure [ fig : allstar_no ] ) , as we expected . in this section we discuss several different methods of age - dating stellar populations in m83 , using both individual stars and star clusters , and then intercompare the results . we do not necessarily expect field star and cluster ages to agree , since the dissolution of clusters may bias the observed cluster population towards younger ages than the surrounding field stars . as shown in figure [ fig : m83_reg50 ] , we selected 50 regions that cover spiral arm and inter - arm areas in order to study the recent star - formation history of m83 . figures [ fig : reg12_7 ] and [ fig : reg29_2 ] show cutouts of regions # 12 , # 7 , # 29 , and # 2 . these figures include color - color diagrams of the stars in these regions , as well as cmds that are uncorrected ( upper ) and corrected ( lower ) for extinction . these four regions highlight the range in age of the dominant stellar population , from very young to intermediate ages : i.e. , region # 12 with very strong h@xmath11 emission superposed on the cluster stars ; region # 7 with a large bubble of h@xmath11 emission surrounding the stars and clusters ; region # 29 with no h@xmath11 emission , but large numbers of bright red and blue stars ; region # 2 with no h@xmath11 emission and fainter stars . a cursory glance at the cmds of these four regions ( figures [ fig : reg12_7 ] and [ fig : reg29_2 ] ) show clear differences . the primary differences are , as predicted by the isochrones , that younger regions contain : ( 1 ) bluer main sequence stars ; ( 2 ) larger numbers of upper main sequence stars ; and ( 3 ) larger ratios of blue to red stars . our regions do not , however , contain only stars of a single age . even these four regions , which were chosen to include a single dominant stellar population , contain a mix of young and old stars . in this paper , we are primarily interested in the ages of the bright young stars in these regions , which dominate cmds in luminosity - limited samples . in practice , this focus on the younger population is carried out by imposing a magnitude cutoff of @xmath69 mag ( i.e. , @xmath70 mag which corresponds to the age cutoff of @xmath060myr in the cmd ) for the stars used to estimate ages ( i.e. , the magenta dashed line in figure [ fig : allstar_cor ] ) . this luminosity limit allows us to stay above the completeness limit for all but the very reddest stars ( see figure [ fig : allstar_no ] ) , while also focusing on evolved stars and those on the upper main sequence , which are most sensitive to the age of a stellar population . our primary goal is to obtain reliable `` relative '' estimates for the 50 regions ( rather than absolute ages ) , and our magnitude limit is sufficiently deep to accomplish this goal . we compare the extinction corrected colors and magnitudes of stars with the padova stellar isochrones in two ways to estimate the age of the dominant stellar population in each region . first , two of the authors ( hk and bcw ) independently estimated ages based on a visual comparison of the cmds and isochrones . we focused on features such as the color of the main sequence stars , the number of stars in the upper main sequence , and the number ratio of blue to red stars . the left panel of figure [ fig : cmdages ] shows that the independent , visually determined age estimates are in good agreement with a 7.9 @xmath32 ( slope / uncertainty of the best linear fit ) correlation and a slope within 1 @xmath32 of the unity value . the average ages of these two independent ( manual ) estimates are listed in column @xmath71@xmath72 in table [ tab:50reg ] . next , we estimated the ages automatically , finding the closest match in both age and mass for each star by comparing the extinction corrected @xmath20 magnitude and @xmath28 color with a fine mesh of stellar isochrones generated from the padova models ( log@xmath36 ( age / yr ) ranging from 6.05 to 8.35 in steps of 0.05 ) . we then calculate a luminosity - weighted age from the distribution of individual stellar ages in each region . the result of the automatic age estimate is listed in column @xmath71@xmath73 in table [ tab:50reg ] . to assess the uncertainties in our age and mass determinations for each star , we ran a test for four different cases by adding and subtracting the photometric errors in the @xmath28 color and @xmath20-band magnitude . this moves each object on the cmd to the right ( @xmath74 ) , left ( @xmath75 ) , down ( @xmath76 ) , and up ( @xmath77 ) . we then estimated the ages for these four cases in the same way as we did to get @xmath71@xmath73 . the results of this test are summarized in column @xmath78@xmath71@xmath73 in table [ tab:50reg ] . as shown in the @xmath78@xmath71@xmath73 column in table [ tab:50reg ] , the variation in the age estimates are typically only a few myr ( i.e. , @xmath2210% ) . figure [ fig : age_mass ] shows the age and mass histograms of individual stars in the four regions shown in figure [ fig : reg12_7 ] and [ fig : reg29_2 ] . luminosity - weighted mean values of the stellar ages are shown in the upper right of the left panels . the age sequence of these regions is as expected , from the youngest in region # 12 to the oldest in region # 2 . one unexpected result was that while region # 2 was selected to be representative of an older field population , it also contains a fair number of bright young stars . this suggests that young individual stars may form in the field in low numbers . another possibility is that many of these are `` runaway '' stars , i.e. , young stars that have been dynamically ejected from their birth - sites within young star clusters due to interactions with other stars . @xcite also suggested this as a possible explanation for isolated hii regions ionized by a single , massive star , i.e. , `` single - star '' hii ( sshii ) regions ( see figure 3 from whitmore 2011 ) . either explanation would be interesting , and should be investigated in a future study . figure [ fig : reg9 ] shows a similar `` field '' region # 9 with primarily old red giant stars ( barely discernible ) , but which also includes a small population of isolated young blue stars and four compact clusters ranging in age from intermediate to old . the right panel of figure [ fig : cmdages ] compares the @xmath71@xmath72 and @xmath71@xmath73 estimates for the different regions . a unity line is included for comparison , and shows that there is good agreement ( 7.6 @xmath32 correlation ) between the two methods which varies from the unity line by only 1 @xmath32 . we use the objective luminosity - weighted age estimates ( @xmath71@xmath73 ) as our preferred method for estimating ages of the resolved stellar populations in the remainder of this paper . @xcite recently estimated the ages of two star clusters in m83 by comparing photometry of individual stars in the outskirts of the clusters with isochrones . these clusters are within regions # 3 and # 11 here . @xcite used a more sophisticated method , comparing observed and predicted magnitude distributions for red and blue stars using hess diagrams . they did not however , correct the magnitudes or colors of their stars for the effects of spatial variation of extinction , but since these are intermediate age clusters with minimal internal extinction , this is not a large effect in these particular cases . they find log @xmath36 of 7.44 ( 27.5 myr ) and 7.45 ( 28.2 myr ) for the clusters in regions # 3 and # 11 . we find similar ages of 21.9 myr ( region # 3 ) and 26.7 myr ( region # 11 ) for the dominant luminous stellar populations , despite the fact that our photometry includes a significantly larger area around each cluster . hence we find good agreement between our age estimates and those of @xcite . each method has its own strengths and weaknesses . our approach has the advantage of correcting the photometry of individual stars for the effects of extinction to improve our age estimates , but the disadvantage that this requires imaging in more than three optical bands with a crucial need for the @xmath9-band observation . the @xcite method does not correct for the effects of extinction , but allows for a better determination of the star formation history over a wider age range ( i.e. , out to @xmath01gyr ) . much of our recent work has involved age - dating star clusters using the integrated light ( @xmath13 and h@xmath11 emission ) from the cluster and model spectral energy distribution ( sed ) fitting . to compare the age of star clusters in each region to the age determined from individual stars in each region , we adopt the ages and luminosities of compact clusters identified in our previous work @xcite . details of the cluster age - dating method can be found in that paper . we note that the regions sampled in this study are large ( i.e. , several hundredpc@xmath79 ) and may include stars and star clusters spanning a wide range of ages . this requires making the comparison between the ages of field stars and star clusters with caution . only the largest star forming complexes within spiral galaxies have such dimensions ( e.g. , the giant hii region ngc604 in m33 ; freedman et al . 2001 ) . more typical regions are much smaller , and hence would not dominate the entire field . this will tend to weaken our correlations , especially when comparing resolved stellar ages with cluster ages . using the ages of star clusters determined from the sed fitting , we compute the luminosity - weighted mean age of each region , as we did for stars in the same region . the ages of 50 regions are listed in column @xmath80 in table [ tab:50reg ] . the upper panel of figure [ fig : sedages ] shows a comparison between the luminosity - weighted mean age ( @xmath80 ) of the compact clusters in a given region and our stellar age ( @xmath71@xmath73 ) estimates based on the cmds of all the resolved stars in the region ( as discussed in 3.1 ) . the correlation between these two ages is fair ( 5.4 @xmath32 correlation ) . we note that while the midpoints for the two methods are similar , the slope is steeper , as discussed in more details in the next section . we obtain an independent estimate of the age of each region by measuring the colors of the entire region , and perform a simple sed fit , in the same way as we did for star clusters ( i.e. , * ? ? ? the lower panel of figure [ fig : sedages ] shows a comparison between our age estimates for the resolved stellar populations using cmds ( @xmath71@xmath73 ) and from the integrated light ( @xmath81 ) in each region . while there is a fair correlation ( 4.6 @xmath32 ) between the cmd and integrated photometric age estimates , there is also a fair amount of scatter and an apparent gap in the region ages in the range 6.8 @xmath68 log ( age / yr ) @xmath68 7.2 . this gap is similar to the well known artifact for cluster age estimates which is due to the looping of predicted cluster colors in this age range ( see chandar et al . we also note that the slope of the relation is steeper than the unity vector , with integrated light ages ranging to much lower ages than the cmd ages . this is similar to the comparison with @xmath80 , as discussed above . this is probably caused by a variety of effects including : 1 ) the integrated age estimates take into account h@xmath11 emission ( which is very sensitive to massive young stars ) while the cmd estimates do not ; 2 ) the 1 myr isochrones are essentially on top of the 3 myr isochrones , making it rare that the minimum age ever gets selected by the software that does the matching with the isochrones ; and 3 ) the adoption of a @xmath82 mag cutoff removes all stars with ages @xmath23100 myr from the @xmath71@xmath73 estimates , while the light from older stars is still included in the integrated light used for the @xmath81 estimate ( hence , the @xmath71@xmath73 estimate can never be @xmath2350 myr , while the @xmath81 can be older ) . this difference in slope in figure [ fig : sedages ] highlights the fact that each age - dating method has its own idiosyncrasies . by comparing a number of different methods , we begin to understand these artifacts better , and learn how large the true systematic uncertainties can be . we are now ready to compare our age estimates for the resolved stellar components in 50 regions to other observables that correlate with age . these include : ( 1 ) number ratio of red - to - blue stas ( e.g. , * ? ? ? * ) ; ( 2 ) morphological categories @xcite ; and ( 3 ) stellar surface brightness fluctuations @xcite . one property that is expected to correlate with age is the number ratio of red - to - blue stars . at very young age , all stars are blue . as the population ages , the number of red giant stars ( h - shell burning and h@xmath34he - shell burning stars ) increases . this effect can be seen by noting that nearly all of the stars in region # 12 ( figure [ fig : reg12_7 ] ) are blue , while there are large numbers of red stars in regions # 29 and # 2 ( figure [ fig : reg29_2 ] ) . the number ratio of red - to - blue stars ( column red - to - blue ratio in table [ tab:50reg ] ) is calculated by using a criteria that @xmath28 be redder than 0.8 mag for red stars . the top panel of figure [ fig : rb_ha_rms ] shows a fairly good correlation ( 7.2 @xmath32 ) between @xmath71@xmath73 and the red - to - blue ratio . the middle panel of figure [ fig : rb_ha_rms ] shows the correlation between our automatic cmd age estimates and the morphological classification for a given region , as defined in @xcite . briefly , regions with h@xmath11 emission superposed on top of the stellar component are type 3 ( emerging star clusters ) , regions with small h@xmath11 bubbles are category 4a ( very young ) , regions with large h@xmath11 bubbles are category 4b ( young ) , and regions with no h@xmath11 bubbles are category 5 ( intermediate age ) . based on this criteria , the regions displayed in figures [ fig : reg12_7 ] and [ fig : reg29_2 ] are classified as categories 4a ( region # 12 ) , 4b ( region # 7 ) , 5a ( region # 29 ) , and 5a/5b ( region # 2 ) , respectively . each region was classified independently by two authors ( hk and bcw ) . the mean of the two determinations is used in what follows , and is listed in column `` @xmath17 morphology '' of table [ tab:50reg ] . the middle panel of figure [ fig : rb_ha_rms ] shows that stellar cmd age estimates ( @xmath71@xmath73 ) and morphological categories ( h@xmath11 morphology ) show a fair correlation ( 5.8 @xmath32 ) , but not as good as the strong correlations ( 9 @xmath32 for log @xmath83 and 5 @xmath32 for log @xmath84 ) found with ages for compact clusters in @xcite . this is probably due to the fact that most of the regions are not dominated by a single age stellar population , making the morphological classification for an entire region problematic ( i.e. , there is strong h@xmath11 emission in part of the region but none in other parts ) . in @xcite , we developed a method to age - date star clusters based on an observed relation between pixel - to - pixel flux variations ( rms ) within star clusters and their ages . this method relies on the fact that the young clusters with bright stars have higher pixel - to - pixel flux variations , while the old clusters with smoother appearance have small variations . this is because the brightest stars in a 100 myr population have @xmath85 mag , which is below our 50% photometric completeness limit ( @xmath86mag ) and well below our magnitude cutoff of @xmath87mag used in our stellar age - dating method described in 3.1 . the bottom panel of figure [ fig : rb_ha_rms ] shows the correlation between the resolved stellar ages ( @xmath71@xmath73 ) and the surface brightness fluctuations ( pixel - to - pixel rms ) measured in the 50 selected regions . we find little or no correlation ( 1.1 @xmath32 ) with a relatively large amount of scatter , especially for the younger clusters . part of this scatter may be caused by the two - valued nature of the effect as discussed in @xcite , with a maximum value of the surface brightness fluctuations around 10 myr , and lower values for both smaller and larger ages . another reason for the scatter is the fact that many of the regions have mixed populations , rather than single age populations . we first examine how the ages for the 50 regions are distributed in figure [ fig : m83_reg50 ] . the boxes in this figure are color - coded as follows : blue ( very young with ages less than about 10 myr ) , yellow ( young with ages between about 10 and 20 myr ) , and red ( intermediate aged with ages greater than about 20 myr ) . as expected , we find that the very young regions are primarily associated with the spiral arms , the young ryoungs tend to be `` downstream '' from the spiral structure ( i.e. , on the clockwise side away from the dust lanes ) , while the intermediate aged regions are found on both sides of the spiral arms in the interarm regions . a similar , more detailed analysis is being done by chandar and chien using the cluster ages in m51 ( 2012 , in preparation ) and m83 ( 2012 , in preparation ) . they compare the age gradients they find with models developed by @xcite , assuming a variety of different triggering mechanisms ( i.e. , spiral arms , bars , stochasticity , and tidal disturbances ) . we can perform a similar experiment here , since we have age estimates for each individual star . the three panels in figure [ fig : pellerin ] show the distribution of stars with ages in the ranges 110 , 1535 , and 40100 myr , respectively . the top panel is the youngest group of stars , clearly showing that the stars in these regions are mostly distributed along the active star - forming region ( i.e. , associated with strong h@xmath11 emission ) in the spiral arms . the stars in the middle panel tend to be found slightly downstream of the spiral arms , while the older stars are still farther out in the inter - arm regions , as expected . these diagrams can also be used to examine whether there is evidence that most of the young stars form in clusters and clustered regions and then dissolve into the field . @xcite made a similar set of diagrams for the galaxy ngc1313 , which supported this interpretation . we find that the young star samples show strong clustering , while the older star samples are progressively more uniform . hence , these distribution maps of stars in the different age group support the idea that the stars form in clusters in spiral arms , and then most of the clusters dissolve , populating the field with stars @xcite . chandar et al . ( 2012 , in preparation ) will examine this subject more quantitatively in the future . color - magnitude diagrams and color - color diagrams of resolved stars from the multi - band @xmath7/wfc3 ers observations of m83 have been used to measure the ages of stellar populations in 50 regions of this well known face - on spiral galaxy . the diagrams show the presence of multiple stellar features , including recently formed ms , he - burning blue - loop stars , and shell - burning red giant stars . comparisons between our stellar age estimates , and a wide variety of other age estimators , allow us to investigate a number of interesting topics . the primary new results from this study are as follows : \(1 ) an innovative new technique using a combination of cmd and color - color diagrams has been developed in order to correct for the extinction towards each individual star and to age - date the stellar population . the mean extinction values for the 50 regions studied in this galaxy are 0.696 ( @xmath88 ) , 0.559 ( @xmath89 ) , 0.441 ( @xmath90 ) , 0.256 ( @xmath91 ) mag , respectively . \(2 ) the various age estimators ( stellar , integrated light , clusters within the region ) show fair correlations ( i.e. , between 4.6 @xmath32 and 5.4 @xmath32 ) . this is expected , since many of the regions have a mixture of different - aged populations within them , and each technique uses light from a different subset of the stars ( e.g. , the stellar age estimates use only bright stars , while integrated light includes all the light ) . a comparison with previous age estimates by @xcite shows a good agreement ( see 3.2 for details ) . ( 3 ) comparisons between the stellar ages and other parameters that are known to correlate with age show a range from very good correlations ( e.g. , with red - to - blue ratios : i.e. , 7.2 @xmath32 ) and fair correlations ( e.g. , h@xmath11 morphology with 5.8 @xmath32 ) to little or no correlation . this is expected for reasons similar to the ones discussed above . \(4 ) the regions with ages younger than 10 myr are generally located along the active star - forming regions in the spiral arm . the intermediate age stars tend to be found `` downstream '' ( i.e. , on the opposite side from the dust lane ) of the spiral arms , as expected based on density wave models . a more detailed comparison with models with various other triggering mechanisms ( e.g. , bars , density waves , tidal disturbance , and stochasticity ; see * ? ? ? * ) is in process by chandar and chien for m51 and m83 ( 2012 , in preparation ) . \(5 ) as discussed in the appendix a , the locations of wolf - rayet sources from @xcite are in broad agreement with the age estimates discussed in the current paper . the much better spatial resolution from @xmath7 shows that many of the wolf - rayet `` stars '' from ground - based observations are actually young star clusters . \(6 ) effects of spatial resolution on the measured colors , magnitudes , and the derived ages of stars in our m83 images are described in the appendix b. based on a numerical experiment using a star cluster ngc2108 in the lmc , we found that while individual stars can occasionally show measurable differences in the colors and magnitudes , the age estimates for entire regions are only slightly affected . + we thank zolt levay for making the color images used in figure [ fig : m83_reg50 ] . we also thank the anonymous referee for helpful comments . this project is based on early release science observations made by the wfc3 science oversight committee . we are grateful to the director of the space telescope science institute for awarding director s discretionary time for this program . support for program # 11360 was provided by nasa through a grant from space telescope science institute , which is operated by the association of universities for research in astronomy , inc . , under nasa contract nas 5 - 26555 . facilities : @xmath7(wfc3 ) in this paper we compared the ages resulting from a variety of age - dating methods for different star - forming regions in m83 . here , we extend our analysis by considering previously identified wolf - rayet stars in m83 , which are @xmath014 myr old @xcite , as another independent age estimate . @xcite identified 283 wolf - rayet sources in m83 from narrow - band imaging centered on the he ii @xmath924686 emission line , plus spectroscopic follow - up . seventeen of these wolf - rayet sources fall in our wfc3 field - of - view . figure [ fig : wr1 ] shows their locations in the @xmath7 image , and table 2 lists the coordinates of these sources , as well as the region in which they are located . the lower resolution of the ground - based @xcite study makes it difficult ( in many cases ) to uniquely identify the wolf - rayet sources in the @xmath7 image , although we can easily determine whether such a source is located within one of our 50 regions . figure [ fig : wr1 ] shows a color image of m83 ( center panel ) with five regions , labeled a through e , that contain all 17 wolf - rayet sources . here , the h@xmath11 and f814w images are shown in red , f555w in green , and f336w in blue . black and white images are the image cutouts in the f814w and the h@xmath11 filters of regions a e , as outlined in green in the central panel . the original finding chart of the region a taken from @xcite is shown in the top right panel in figure [ fig : wr1 ] . in the h@xmath11 image , the stellar continuum has been subtracted from the h@xmath11 image , leaving just the ionized gas . we matched the coordinates of the wolf - rayet sources and the @xmath7 images , by assuming that the wolf - rayet source # 99 is perfectly centered on a relatively isolated , compact h@xmath11 knot . we also compared the locations of the 17 matched wolf - rayet sources in the m83 @xmath7 images with the finding charts from @xcite to verify our source matching . the excellent spatial resolution of the @xmath7 image also makes it possible to check whether any of the wolf - rayet sources are actually compact star clusters . while the majority do appear to be individual stars , or a dominant star with a little `` fuzz '' which is likely a faint companion star , there are at least three cases where the counterpart appears to be a bright compact cluster ( i.e. , # 74 , 86 , 105 : circled objects in figure [ fig : wr2 ] ) . these three objects are brighter than @xmath93 mag with the concentration index ( ci ) larger than 2.3 , consistent with the cluster ci used in @xcite . there are several more cases where the counterpart is in a looser association of stars ( e.g. , # 78 , 102 , 103 ) . all the wolf - rayet sources , except the sources in cutout e and # 108 in cutout d , are located in the spiral arms . this is expected , since these are the sites of most recent star formation . nine of the 17 are in regions of strong h@xmath11 emission , seven more are in regions of faint h@xmath11 emission , and only one source ( # 107 in cutout d ) appears to have no h@xmath11 associated with it . our primary question is whether the wolf - rayet sources tend to be found in regions that we estimated to have young ages . in table 2 we match the wolf - rayet sources with the region where they are found in and find that their ages range from 1.1 myr ( region # 35 ) to 50.1 myr ( region # 41 ) . this shows that such a correlation does exist , with 6/8 ( 75 % ) of wolf - rayet sources in regions having @xmath81@xmath684.6 myr . similarly , @xcite found that five of their wolf - rayet sources are included in a compact star cluster catalog developed by @xcite , and three of these five clusters ( # 193 , 179 , 179 ) have estimated ages in the range 1.5 6 myr , consistent with ages expected for wolf - rayet stars . figure [ fig : wr2 ] shows the cmd and color - color diagram of the measured colors from our @xmath7 observations of the object that best matches the wolf - rayet sources from @xcite . we find that most of these sources are in the top left of the two - color diagram . we conclude that there is a strong tendency for wolf - rayet sources to be associated with the youngest regions of star formation . the good correlation provides general support for the accuracy of our cmd age estimates . spatial resolution affects virtually all studies of individual stars in nearby galaxies . as is typical , in this paper we have assumed that the point sources identified in section 2 are individual stars , although we also discovered a small fraction of pairs of nearly aligned stars , one red and one blue , that can be identified from their anomalous colors ( e.g. , `` pseudo - match '' figures 2 and 4 ) . here , we address more generally how spatial resolution affects the measured colors and magnitudes , and hence the derived ages , of stars in our m83 regions . we begin by answering a simple question : `` what would the orion nebula look like at the distance of m83 ? would we be able to distinguish the four central stars that make up the trapezium or would they appear as a single star ? '' the separation between the four stars in the trapezium ( @xmath94 at d@xmath95kpc ) would be roughly @xmath96 at the distance of m83 , approximately a factor of 40 smaller than a single pixel in our image , and hence these would appear as one point - like source . however , we note that the vast majority of the remaining bright stars in the orion nebula are much more widely separated , and would not suffer from this problem . we also note that our selection criteria are designed to avoid including stars in the crowded central regions of star clusters and associations , which also minimizes the problem . we perform a more quantitative analysis by degrading @xmath7/acs images of the compact star cluster ngc2108 in the large magellanic cloud ( lmc ) ( id:10595 ; pi : goudfrooij , see * ? ? ? * ) , which is at a distance of @xmath97kpc , by a factor of 100 ( i.e. , equivalent with the distance of @xmath98mpc ) , as roughly appropriate for m83 . the degraded images , created from a combination of rebinning and gaussian smoothing to approximately mimic the point spread function of acs , are then run through our normal object - finding software described in section 2 . the numbers of objects found in the degraded ngc2108 images are @xmath020 in the f555w and f814w acs images . figure [ fig : n2108 ] shows the f555w acs image of ngc2108 , along with the degraded images and a color image ( i.e. , a color composite of f435w , f555w , and f814w acs images ) . additionally , we then perform aperture photometry on the detected source positions in the non - degraded acs images using a radius 300 pixels to collect light from all stars that would fall within our aperture if ngc2108 was located at a distance of 5mpc ( i.e. , m83 ) . objects along the edge and within 1000 pixels of the center of the cluster in ngc2108 are discarded , as would be the case for the corresponding photometry in m83 . figure [ fig : circles ] shows the extent of the 300 pixel aperture for several sources in yellow , as well as the nearest dominant star in red . the magnitude of the dominant stars ( using a 3 pixel aperture and the appropriate aperture correction ) is considered the `` truth '' measurement while the magnitude in the corresponding 300 pixel aperture is the value that would be determined at the distance of m83 . we calculate the magnitude and color offsets , i.e. , @xmath99 and @xmath100 mag of the dominant stars , by taking the difference between the 3 pixel and 300 pixel aperture magnitudes . one complication is that our default sky subtraction method would remove the vast majority of faint and moderate brightness stars in the background annulus in the ngc2108 image , but these stars would not be detected individually at the distance of m83 and hence would be included in the sky measurement . we therefore adjusted the `` clipping '' parameters in the sky subtraction algorithm to remove only the bright stars in ngc2108 , thereby mimicking the case for m83 to the degree possible . in figure [ fig : m83_reg3 ] we show the original cmd for region 3 in m83 ( solid points ) , compared with a version where the magnitudes and colors have been ` perturbed ' by an amount @xmath99 and @xmath101 determined by matching to the closest values of @xmath28 from the ngc2108 data experiment ( open points ) . as expected , corrections for the degraded spatial resolution in m83 tend to make the corrected magnitudes slightly fainter and redder , as can be seen in figure [ fig : m83_reg3 ] . the mean difference in @xmath19 is @xmath102mag with a range from @xmath103 to @xmath104mag . in @xmath20 , the mean difference is @xmath105mag with a range from @xmath106 to @xmath107mag . we note that the largest difference is seen for object 10 , as shown in figure [ fig : circles ] , where several relatively bright red stars in the 300pixel aperture cause the largest correction ( i.e. , @xmath107mag in @xmath20 and redward by 0.63mag in @xmath28 ) . object 40 , dominated by a single very bright star , shows the opposite extreme , with a change of @xmath108mag in @xmath19 and @xmath109mag in @xmath20 . we also note that object 31 does not make it into our sample since the two bright stars are far enough apart that it is clear from the degraded image in figure [ fig : n2108 ] that this is not a single star . this object gets removed in the dophot photometry as described in 2 , and hence is not used in the experiment described above . we then rerun our age - dating software on the corrected cmd in figure [ fig : m83_reg3 ] . we find a mean age of 23.8myr compared to the original value of 21.9myr . hence , while a specific `` star '' may be affected by a sizeable amount , the overall effects of degraded spatial resolution are relatively minor for our study . to summarize , we have performed a numerical experiment using observations of an intermediate - age star cluster in the lmc ( ngc 2108 ) for our `` truth '' image . using a factor of 100 in spatial degradation , roughly appropriate for our measurements in m83 , we find that the typical corrections to our photometry are on the order of a few tenths of a magnitude , although much larger values are possible in specific cases . we conclude that while spatial resolution can result in measurable differences in the luminosities and colors of bright stars at the distance of m83 , it does not strongly affect the age estimates for the 50 regions studied in this paper . bastian , n. , et al . 2011 , mnras , 417 , l6 bresolin , f. , & kennicutt , r. c. 2002 , apj , 572 , 838 cardelli , j. a. , clayton , g. c. , & mathis , j. s. 1989 , aj , 345 , 245 chandar , r. , et al . 2010 , apj , 719 , 966 crowther , p. a. 2007 , ara&a , 45 , 177 dalcanton , j. j. , et al . 2009 , apjs , 183 , 67 dobbs , c. l. , & pringle , j. e. 2010 , mnras , 409 , 396 dopita , m. a. , et al . 2010 , apj , 710 , 964 faria , c. , bosch , g. l. , & barb , r. h. 2012 , aj , 143 , 43 freedman , w. l. , et al . 2001 , apj , 553 , 47 goudfrooij , p. , puzia , t. h. , kozhuria - platais , v. , & chandar , r. 2011 , apj , 737 , 3 hadfield , l. j. , crowther , p. a. , schild , h. , & schmutz , w. 2005 , a&a , 439 , 265 harris , j. , calzetti , d. , gallagher , j. s. , conselice , c. j. , & smith , d. a. 2001 , aj , 122 , 3046 karachentsev , i. d. , et al . 2007 , aj , 133 , 504 koekemoer , a. m. , fruchter , a. s. , hook , r. n. , & hack , w. 2002 , in the 2002 @xmath7 calibration workshop : hubble after the installation of the acs and the nicmos cooling system , ed . s. arribas , a. koekemoer , & b. whitmore ( baltimore , md : stsci ) , 337 lada , c. j. , & lada , e. a. 2003 , ara&a , 41 , 57 larsen , s. s. 2004 , a&a , 416 , 537 larsen , s. s. , et al . 2011 , a&a , 532 , a147 marigo , p. , girardi , l. , bressan , a. , groenewegen , m. a. t. , silva , l. , & granato , g. l. 2008 , a&a , 482 , 883 mihalas , d. , & binney , j. 1981 , galactic astronomy ( san francisco : freeman ) , 187 pellerin , a. , meyer , m. , harris , j. , & calzetti , d. 2007 , apj , 658 , l87 saha , a. , thim , f. , tammann , g. a. , reindl , b. , & sandage , a. 2006 , apjs , 165 , 108 saha , a. , et al . 2010 , aj , 140 , 1719 schechter , p. l. , mateo , m. , & saha , a. 1993 , pasp , 105 , 1342 schlegel , d. j. , finkbeiner , d. p. , & davis , m. 1998 , apj , 500 , 525 sirianni , m. , et al . 2005 , pasp , 117 , 1049 whitmore , b. c. , zhang , w. , leitherer , c. , fall , s. m. , schweizer , f. , & miller , b. w. 1999 , aj , 118 , 1551 whitmore , b. c. , et al . 2011 , apj , 729 , 14 whitmore , b. c. & the wfc3 soc team 2011 , in asp conf . 440 , up2010 : have observations revealed a variable upper end of the initial mass function ? m. treyer , t. k. wyder , j. d. neill , m. seibert , and j. c. lee ( san francisco , ca : asp ) , 161 windhorst , r. a. , et al . 2011 , apjs , 193 , 27 . the f336w image is shown in blue , the f502n ( [ o iii ] ) image in cyan , the f555w image in green , and the combined f814w and f657n ( @xmath17 ) image in red . the 50 selected regions are outlined in boxes of three different colors based on the values of @xmath71@xmath73 determined in this paper ( see 3.1 for details ) . the regions with ages in the range 110myr are outlined in blue , ages of 1020myr in yellow , and ages greater than 20myr in red . ] 2@xmath1213@xmath219 , 340pc@xmath12310pc ) of regions # 26 and # 27 in the f555w ( @xmath110 ) and the color - composite images ( @xmath111 ) . stars circled in red have @xmath1120.8 mag . the `` postage stamp '' images on the left are``pseudo '' matched candidates , while the images on the right appear to be heavily reddened ( see 2.2 for details ) . ] 3@xmath1216@xmath219 , 410pc@xmath12380pc ) of the area near region 9 as an example of a `` field '' region with old red giant stars , and a small population of isolated young blue stars . see 3.1 for for detail . , scaledwidth=50.0% ] ): correlations between the ages of stars ( @xmath71@xmath73 ) and clusters ( @xmath80 ) . ( @xmath113 ) : correlation between the ages of stars determined by @xmath71@xmath73 and the integrated light within the region ( @xmath81 ) . see 3.4 for details . the best linear fits are shown in black solid line with 5.4 @xmath32 @xmath114 and 4.6 @xmath32 @xmath115 correlations . the dashed line is the unity line . ] @xmath73 ) and the number ratio of red - to - blue stars @xmath114 , h@xmath11 morphology @xmath116 , and pixel - to - pixel fluctuations @xmath115 for the 50 selected regions . see 4 for details . the best linear fits are indicated in solid black line with 7.2 @xmath32 @xmath114 , 5.8 @xmath32 @xmath116 , and 1.1 @xmath32 @xmath115 correlations . ] emission ) in the spiral arms . the stars in the middle panel tend to be found slightly downstream of the spiral arms while the older stars are still farther out in the inter - arm regions , as expected . see 5 for details . ] wfc3 m83 images . the color image at the center is the @xmath117 composite and the image cut - outs are the h@xmath11 ( @xmath118 or @xmath110 ) and f814w ( @xmath119 or @xmath111 ) images for regions a e . the top right panel is an original finding chart ( * ? ? ? * @xmath924684 narrow - band fors2 image ) for the region a. numbers in blue are the ids of wolf - rayet sources identified by @xcite.,scaledwidth=100.0% ] . three sources ( # 78 , 86 , 105 ) circled in both diagrams are identified as clusters in our m83 wfc3 images . a twice solar metallicity bc09 cluster model ( bruzual & charlot 2009 , private communication ; see also bruzual & charlot 2003 ) is shown in the right panel ( dots ) , and padova stellar models @xcite are shown in both panels . no correction has been made for internal reddening , which appears to be minimal for all but one object . the arrow indicates the galactic reddening vector.,scaledwidth=100.0% ] /acs images ( 290@xmath120290@xmath121 , 68pc@xmath1268pc ) of the compact star cluster ngc2108 in the lmc : a color composite of the f435w , f555w , and f814w acs images ( @xmath118 ) , the f555w acs image ( @xmath122 ) , and the degraded f555w acs image ( @xmath119 ) by a factor of 100 ( from a distance of @xmath050kpc to a distance of @xmath05mpc ) . see appendix b for details.,scaledwidth=100.0% ] @xmath73 of this region changes from 21.9myr ( determined from the solid points ) to 23.8myr ( from the open points ) . the difference in the age estimates is within @xmath123 error listed in table 1 . see appendix b for details.,scaledwidth=80.0% ]

a gaas / algaas based two - qubit quantum device that allows the controlled generation and straightforward detection of entanglement by measuring a stationary current - voltage characteristic is proposed . we have developed a two - particle green s function method of open systems and calculate the properties of three - dimensional interacting entangled systems non - perturbatively . we present concrete device designs and detailed , charge self - consistent predictions . one of the qubits is an all - electric mach - zehnder interferometer that consists of two electrostatically defined quantum wires with coupling windows , whereas the second qubit is an electrostatically defined double quantum dot located in a second two - dimensional electron gas beneath the quantum wires . we find that the entanglement of the device can be controlled externally by tuning the tunneling coupling between the two quantum dots .

0710.3572

semiconductor based concepts for quantum information processing promise a high degree of scalability . however , the implementation of even a single qubit proves to be more difficult in semiconductors than in alternative approaches . @xcite the reasons lie in the short decoherence times and strong interactions between elementary excitations in solids . @xcite while a 7-qubit nmr quantum computer was realized already in 2001 , @xcite the first two - qubit quantum operation in semiconductors has been demonstrated with spin qubits only recently . @xcite at liquid helium temperatures and below , electrons in a gaas / algaas two - dimensional electron gas ( 2deg ) show mean free paths as well as phase - relaxation lengths of the order of @xmath0 which are remarkably long distances . @xcite indeed , bertoni _ _ et al.__@xcite and ionicioiu _ _ et al.__@xcite have proposed a scheme for quantum computation in semiconductors that exploits these long coherence lengths for electrons propagating through quantum wires ( qwrs ) . in this scheme , a single electron propagates through two parallel qwrs , that represent the qubit states @xmath1 and @xmath2 , respectively . a single - qubit rotation gate can be realized by an electronic directional coupler@xcite that may consist of a small window in the barrier between the two qwrs and is able to transfer the wave packet from one channel to the other . single - qubit structures have been studied theoretically by several authors@xcite who showed how the dimensions of the coupling window can be utilized to tailor the transfer process . this approach allows one , at least in principle , to apply multiple quantum gate operations to a qubit without exceeding the relevant coherence lengths . universal quantum computation requires not only one - qubit but also two - qubit gates @xcite that are responsible for the creation of entanglement between the qubits . to this end , two pairs of qwrs may be brought close to one another so that the electrons in the qwrs can interact with one another in a controlled way via their coulomb repulsion . this concept has been analyzed both for wave packets@xcite as well as for stationary states@xcite . the latter investigations used simple models that provide important proofs - of - principle , but do not provide quantitative predictions of realizable device structures that exhibit quantum gate operations . while two - qubit devices have only been addressed theoretically in the literature , one - qubit systems have also stimulated a number of experiments in both lateral@xcite as well as vertical@xcite nanostructures . however , quantum entanglement can clearly not be exposed in single - qubit systems as entanglement expresses the non - separability of a multi - qubit state . therefore , the following step towards the realization of a prototype quantum gate based on coupled qwrs consists of a two - qubit device . also , we should keep in mind that conceivable two - qubit experiments need realistic estimations of an entanglement witness that is experimentally accessible and a clear signature of the correct coherent functioning of the directional coupler . the aim of the present work is to theoretically study realistic single- and two - qubit quantum gates , by proposing conceptually simple and experimentally realizable novel semiconductor devices for quantum information processing . the system proposed is based on ballistic gaas / algaas qwrs and allows for the controlled generation and detection of entanglement between an all - electric mach - zehnder interferometer and an electrostatically defined single - electron double quantum dot . the mach - zehnder interferometer is realized by two electrostatically defined qwrs that are connected by two coupling windows . we model the electrons that are propagating through the interferometer by stationary scattering states , and our qubit state will be defined accordingly . in fact , the original proposal of bertoni _ et al . _ involves the injection of electron wave packets into the pair of qwrs . however , we chose here a time - independent approach since , ( i ) it is equivalent to the time - dependent one when the spatial dimension of the wave packet tends to be larger than the device ( indeed , this is the case in our devices due to the small source / drain biases leading to a well defined kinetic energy of the electron ) , ( ii ) the structures that implement the quantum gates operate in the same way , ( iii ) it allows us to compute realistic estimations of the charge - self - consistent ballistic @xmath3-@xmath4 characteristics of the device . in our calculations , we include the detailed charge - self consistent three - dimensional device geometry , material composition , doping profile and bias voltage . importantly , we demonstrate that the all - electric mach - zehnder interferometer can function as a fully controllable single - qubit gate for experimentally attainable parameters . in addition , we have developed a green s function method for the quantitative analysis of the entangled mach - zehnder and double quantum dot system that includes the coulomb interaction between the two qubits non - perturbatively . in order to gain better qualitative insight into the numerical results , we interpret them in terms of an analytical model that reproduces the computational results qualitatively . taking both results together , we are able to show that the degree of entanglement can be related to the dc @xmath3-@xmath4 characteristics of the interferometer and that the mach - zehnder double quantum dot device can be employed as an externally controllable two - qubit gate . this paper is organized as follows . in sec . [ method ] , we introduce a novel method for the quantum mechanical calculation of the ballistic current through an interacting two - particle system . this includes the calculation of the electronic structure and the determination of the ballistic transmission characteristics of the realistic three - dimensional nanostructure . in sec . [ numericaldetails ] , we discuss numerical details of the method . in sec . [ analyticalmodel ] we introduce a simplified analytical scheme able to gather the essential features of the two - qubit device . the results obtained with this approach will be compared , in the following section , with the calculations of the realistic device . [ results ] focuses on the results and the discussion of the ballistic current through the proposed gaas / algaas single- and two - qubit qwr devices . in the same section , the degree of entanglement between the two qubits is evaluated . the paper concludes with final remarks and a summary in sec . [ conclusions ] . in this section , we present a novel method for the calculation of the stationary ballistic current of two interacting particles in realistic three - dimensional nanostructures . we restrict ourselves to the case where only one of the two particles ( say , particle 1 ) contributes to the current while the second particle ( particle 2 ) is assumed to be bound . importantly , this method accounts for many - particle effects non - perturbatively which is an important prerequisite of studying entanglement and , consequently , predicting solid - state based quantum information processing . the presently developed scheme generalizes the contact block reduction ( cbr ) method for single - particle ballistic currents that we have developed previously . @xcite in analogy to the cbr method , we proceed in two steps that we will describe in this section . first , we determine the equilibrium electronic structure of a closed two - particle system , obtained by substituting the leads of our device with closed boundary conditions . then , in the second step , we calculate the current carrying states of the open two - particle system . in fact , the closed calculation only provides a convenient basis for the computation of the scattering states . for the sake of efficiency and without loss of generality , we perform the former calculation in the hartree approximation , by solving two coupled schrdinger and poisson equations . then , we use single - particle product states as the basis for the scattering states of the fully interacting two - particle device . we consider a single - particle , single - band , effective mass schrdinger equation for this electron , and include the electrostatic hartree potential in the absence of the second particle . the closed system is represented by a hermitian hamiltonian with von neumann boundary conditions for the wave functions at the contacts ( vanishing normal derivative ) . @xcite the electronic structure of the closed device is calculated self - consistently . to this end , we iteratively solve the schrdinger equation @xmath5 \langle\mathbf{x}|\alpha_{i}\rangle \notag \\ & = e_{i}^{0}\langle\mathbf{x}|\alpha_{i}\rangle\ , , \label{schroedingerequation}\end{aligned}\ ] ] and the non - linear poisson equation @xmath6\ , , \label{nonlinearpoissonequation}\ ] ] until the electrostatic potential @xmath7 and the total charge density @xmath8 have reached convergence . the effective mass @xmath9 , the conduction band offset @xmath10 , and the relative dielectric constant @xmath11 are position - dependent material parameters in a general three - dimensional nanostructure that may be composed of several different materials . the total charge density is composed of the electron density @xmath12 and the ionized donor density @xmath13 @xmath14 the latter results from the donor density @xmath15 , the degeneracy @xmath16 , and the donor energy @xmath17 according to the thomas - fermi approximation , @xmath18 the electron density is calculated quantum mechanically by occupying the electronic states according to the fermi - dirac statistics @xmath19 here , @xmath20 denotes the temperature , @xmath21 is boltzmann s constant , @xmath22 denotes the fermi level , and @xmath23 is the fermi distribution function . for the second particle , we will focus on an electron in a closed double quantum dot that can tunnel between these dots . the two lowest lying quantum states can be described by an effective hamiltonian , that takes into account the energy splitting @xmath24 between the ground states of the isolated quantum dots and the tunneling coupling @xmath25 between the two quantum dots . this two - by - two hamiltonian is given by latexmath:[\[h_{2}|y\rangle= \begin{pmatrix } -\frac{\delta}{2 } & -\frac{t}{2}\\ -\frac{t}{2 } & \frac{\delta}{2 } \end{pmatrix } values for @xmath24 and @xmath25 for lateral semiconductor quantum dots are of the order of @xmath27 . @xcite we use this model hamiltonian for the subsystem of the second particle and represent the charge distribution of the electron in the ground state of each ( isolated ) quantum dot by a point charge centered at @xmath28 and @xmath29 , respectively . the use of a model hamiltonian rather than a realistic three - dimensional device hamiltonian is not a principle limitation of the present method but is adequate for the concrete device geometry that we will study in detail in this paper . the eigenstates of @xmath30 are linear combinations of the ground states @xmath31 and @xmath32 of the two isolated quantum dots , @xmath33 with real valued coefficients @xmath34 ( @xmath35 , @xmath36 ) . here , @xmath37 is the bonding eigenstate and @xmath38 the anti - bonding eigenstate . the corresponding eigenenergies are @xmath39 for vanishing @xmath25 , the eigenstates @xmath37 and @xmath38 reduce to @xmath31 and @xmath32 . the coulomb interaction @xmath40 between particle @xmath41 and particle @xmath42 yields the following expression for the interaction matrix elements @xmath43 where @xmath44 is the product state of particle @xmath41 state @xmath45 , and particle @xmath42 state @xmath46 . in this second step , we develop a green s function method by extending the cbr method@xcite to the case of an open device that describes a system of two interacting particles . we stress that we will consider the two particles as distinguishable . this is not an approximation for the proposed two - qubit devices but comes directly from the system geometry . indeed , we choose the qwrs and the double dot to be well separated from each other so that no significant tunneling between the two structures can occur . in other words , the two wave functions namely the bound state in the double dot and the scattering state in the qwrs never overlap . this leads to the distinguishability of the two particles based on their spatial localization . in addition , we assume that the interaction is negligible outside the device , due to screening in the contacts and/or due to barriers . in the following , the term _ device _ denotes a finite three - dimensional region that is coupled to reservoirs by an arbitrary number of leads . the device may be under applied bias and contain some spatially varying potential . the total ( two - particle ) hamiltonian of the system , including the device and the leads , can be written in symbolic matrix form @xmath47 where @xmath48 represent the hamiltonian of lead @xmath49 , the hamiltonian @xmath50 corresponds to the device region , and @xmath51 is the coupling between the device and this lead ( @xmath52 ) . @xmath50 is composed of the single - particle hamiltonians @xmath53 and @xmath30 , corresponding to particle 1 and 2 , respectively , and the interaction term @xmath40 , @xmath54 the leads ( acting as reservoirs ) are semi - infinite and therefore , the total hamiltonian @xmath55 is infinite - dimensional . this infinite - dimensional hamiltonian can be reduced to a non - hermitian finite - dimensional hamiltonian @xmath56 that describes the _ open device _ exactly@xcite . in this formulation , the influence of the leads is included through a finite - dimensional operator @xmath57 . this is the sum of the complex contact self - energies @xmath58 that are nonzero only in the _ contact regions _ where the lead @xmath49 adjoins to the device . the self - energies couple only particle 1 to the leads since particle 2 is assumed to be bound . the hermitian hamiltonan @xmath50 represents the _ decoupled device _ , i.e. the device with no coupling to the leads . @xmath53 as well as @xmath30 are hamiltonians of the closed device . in the ballistic case , all observables of interest such as transmission functions and the current can be calculated from the retarded green s function @xmath59 of the open device . this is defined by @xmath60 the retarded green s function can be obtained from the green s function @xmath61 of the decoupled device by the dyson equation @xmath62 which in turn can be evaluated from its spectral representation @xmath63 @xmath64 the direct evaluation of @xmath59 according to eq . ( [ dysonequation ] ) requires the inversion of a huge matrix that is proportional to the number of grid points @xmath65 of the device . by contrast , the cbr method allows one to drastically reduce this effort by utilizing the following exact properties of @xmath59 that remain valid in the present many - particle case . we decompose @xmath59 into four blocks , @xmath66 where @xmath67 is a matrix proportional to the number @xmath68 of contact grid points and is called the _ contact block_. note that @xmath69 . for the contact block @xmath67 , the following equation holds : @xmath70 where @xmath71 , @xmath72 , and @xmath73 are the corresponding contact blocks of the unity matrix , @xmath61 , and @xmath74 , respectively . the key point is that this is a linear matrix equation of the order of @xmath68 , since @xmath74 is nonzero only in the contact region . as a consequence , this contact block is also sufficient to determine the transmission functions through the device . the ballistic current can then be calculated from the transmission functions according to the landauer - bttiker formalism . @xcite in the following , we represent @xmath67 in the mixed basis of position eigenstates of particle @xmath41 and energy eigenstates of particle @xmath42 , @xmath75 in this form , @xmath67 represents the probability amplitude for the propagation of particle @xmath41 from position @xmath76 to position @xmath77 , accompanied by a transition of particle @xmath42 from eigenstate @xmath78 to @xmath79 . for @xmath80 , the expression @xmath81 therefore yields the probability for the transmission of particle @xmath41 from lead @xmath82 to lead @xmath49 under the condition , that initially particle @xmath42 is in state @xmath78 and ends up in state @xmath79 . note that the trace in eq . ( [ cbrconditionaltransmission ] ) is taken only with respect to the position variables of particle @xmath41 . if the final state of particle @xmath42 is not measured , the corresponding transmission probability is obtained by summing over all final states of particle @xmath42 @xmath83 in our concrete calculations ( see sec . [ results ] ) , we found it most efficient to first determine the eigenstates of the single - particle hamiltonians @xmath53 and @xmath30 and then diagonalize the interaction @xmath40 in the product basis of these single - particle eigenstates . this procedure allows us to take advantage of another built - in efficiency of the cbr method , namely the fact that only energetically low lying eigenstates of the closed device couple to the incoming and outgoing lead states . @xcite consequently , it suffices to take into account only a reduced set of eigenstates of the single - particle hamiltonians @xmath53 and @xmath30 without affecting the transmission results noticeably . another quantity that can be readily obtained from the retarded green s function @xmath59 is the charge density of the current carrying states of particle @xmath41 . this requires one to know not only the contact block @xmath67 but also the submatrix @xmath84 . however , the calculation of the submatrix @xmath84 only requires one to evaluate the inverse of the small matrix @xmath85 , @xmath86 where the matrix @xmath87 corresponds to @xmath84 in the open device . @xcite according to eq . ( [ grcmixedbasis ] ) , the latter green s function reads in the mixed basis latexmath:[\[g_{dc}^{r}(\mathbf{x}_{i},\mathbf{x}_{j},y , z)= \sum_{\alpha_{i},\alpha_{j } } \langle\mathbf{x}_{i}|\alpha_{i}\rangle\langle\alpha_{i}y|g_{dc}^{r } charge density of the current carrying states of particle @xmath41 is consequently given by @xmath89 where the contribution @xmath90 results from the projection of the total charge density onto the eigenstate @xmath78 of the second particle . here , @xmath91 denotes the fermi level in lead @xmath49 . equations ( [ schroedingerequation ] ) and ( [ nonlinearpoissonequation ] ) are coupled partial differential equations in position space that we discretize and map onto a nonuniform tensor product grid . for the discretization of the all - electric mach - zehnder interferometer , a total of @xmath92 grid points have been used . we employ box integration finite differences in order to ensure flux conservation across boundaries with different material parameters . for the solution of the nonlinear system that results from the discretization of the nonlinear poisson equation ( [ nonlinearpoissonequation ] ) , newton s method with inexact line search is invoked . this remaps the problem into a sequence of linear solutions . we use the dupont - kendall - rachford preconditioned conjugate gradient method for the solution of the resulting linear systems . @xcite the discretization of the schrdinger equation ( [ schroedingerequation ] ) results in a large matrix eigenvalue system . however , since the occupation of electron states falls off exponentially with increasing energy distance from the fermi level , only about @xmath93 quantum states suffice for a converged calculation of the electron density . this allows for the use of the iterative arnoldi method that is implemented in the published arpack libraries . @xcite we found that the calculation of eigenvalues and eigenstates can be accelerated by an order of magnitude by invoking a chebyshev polynomial based spectral transformation that provides an efficient preconditioning of the linear system . @xcite finally , in order to solve the coupled poisson - schrdinger system , we employ an approximate quantum charge density inside of poisson s equation in order to estimate the dependence of the density on the potential through schrdinger s equation . this corrector - predictor scheme reduces the number of required diagonalizations and accelerates the convergence of the coupled system significantly . @xcite equation ( [ nonlinearpoissonequation ] ) requires the specification of boundary conditions in order to obtain unique solutions . for the potential in the poisson equation , we generally use von neumann boundary conditions . dirichlet boundary conditions are employed at schottky contacts and at the gaas / air interfaces . in the former case , the electrostatic potential at the boundary is set equal to the schottky barrier height , i.e. the difference between the conduction band edge and the quasi - fermi level . at the gaas / air interface , on the other hand , the electrostatic potential is determined by the experimentally known fermi level pinning . the schrdinger equation for the open quantum system is solved in terms of the cbr method that has been described in detail elsewhere . @xcite for the calculation of the transmission probabilities , we have taken into account the lowest @xmath93 eigenstates for the determination of the retarded green s function of the closed device and the lowest @xmath94 propagating modes in each of the four leads . the self - consistent solution of eqs . ( [ schroedingerequation ] ) and ( [ nonlinearpoissonequation ] ) , for given split gates voltages , takes approximately @xmath42 days on state - of - the - art pc hardware . converged self - consistent results require typically @xmath95 iterations in the two sets of equations . our novel method for the calculation of the ballistic transport properties through a system of two interacting particles requires the diagonalization of the two - particle hamiltonian including the interaction . this has to be done once for every bias point in the @xmath3-@xmath4 characteristics . the diagonalization is performed in the set of product states formed by subsets of the single - particle eigenstates of the two non - interacting particles . we found that a subset of @xmath96 lowest eigenstates of particle @xmath41 is sufficient to obtain convergent transmission functions in the mev energy range around the fermi level that is relevant for this device . before presenting and discussing the numerical results , we develop a simple analytical model of our two - qubit mach - zehnder double - dot system in the present section . the device is shown schematically in fig . [ schematicdevice3d ] . the purpose of this procedure is to be able to better grasp the physics behind the generation of entanglement before we present the numerical results of the fully three - dimensional self - consistent calculations in the next section . the good agreement between the two approaches will be a further confirmation of the correct functioning of the device : the results that we obtain for the entangled mach - zehnder double quantum dot device are indeed the result of a two - qubit quantum operation . obviously , the results of the two methods will only be in qualitative agreement with one another , since the analytical model developed in the this section can not account for the full complexity of the realistic device . applied between the source ( @xmath97 ) and the three drain contacts ( @xmath98 ) that leads to currents ( @xmath99 ) that are labeled accordingly . there are additional gates acting on the device that are shown in the following fig . [ heterostructureandmetalgates ] . for sake of clarity , the figure is not drawn to scale . ] we describe the device consisting of a mach - zehnder interferometer coupled to a double quantum dot , by two coupled qubits , w and d , respectively . the basis states @xmath100 and @xmath101 of qubit w are defined by an electron that propagates through either of the two qwrs . the basis states @xmath102 and @xmath103 of qubit d , on the other hand , are defined by the ground states of the isolated quantum dots . quantum gates can be described by unitary operators that map an initial state @xmath104 onto a corresponding final state @xmath105 . the unitary operator @xmath106 that represents our two - qubit quantum transport device can be decomposed into a sequence of unitary operators of less complex single- and two - qubit gates as follows@xcite : @xmath107 the operators @xmath108 and @xmath109 represent single - qubit gates that act solely on qubit w. in particular , @xmath108 represents a @xmath110 rotation gate and describes a coupling window , whereas @xmath109 represents a phase gate that adds a phase difference to the two components of the qubit . the latter can be realized through an additional split gate located on one of the qwrs , as explained in sec . [ results ] . the sequence @xmath111 therefore corresponds to a mach - zehnder interferometer . the operator @xmath112 represents a two - qubit gate and describes the coupling of the two qubits d and w due to their interaction . this decomposition allows us to specify the action of the single - qubit gates @xmath108 and @xmath109 directly in terms of unitary operators . we proceed by describing the entangled mach - zehnder double dot device as a stationary quantum - mechanical scattering problem . to this end , we represent the two electrons in the pair of qwrs and the double quantum dot by the following 8 basis states for given total energy @xmath113 and a given position @xmath114 , @xmath115 these states describe the two electrons as follows : the first electron propagates along the wire axis @xmath114 in either of the two states @xmath116 and @xmath101 , corresponding to the two qwrs . the electron momentum at position @xmath114 is denoted by @xmath117 . the second electron occupies one of the two quantum dots , represented by states @xmath118 and @xmath119 . in this representation , the operators in eq . ( [ sequenceofgates ] ) correspond to @xmath120 transfer matrices , @xmath121 this will allow us to finally determine the transmission function through the device . as a first step , we derive the transfer matrix @xmath122 of the two - qubit gate @xmath112 by solving a two - particle schrdinger equation . the non - interacting two - particle hamiltonian that describes the dynamics of the system is given by @xmath123 here , @xmath124 is the model hamiltonian that describes a pair of tunneling coupled quantum dots in terms of the bare splitting @xmath24 and the tunneling coupling @xmath25 in accordance with eq . ( [ modelhamiltonianquantumdots ] ) . for the coulomb interaction between the two electrons , we make the following ansatz @xmath125 this describes a localized interaction with interaction strength @xmath126 that is nonzero only for the two - particle state @xmath127 but zero for all other states . in addition , the interaction vanishes if the position of the qwr electron lies outside the spatial interval from @xmath128 to @xmath129 . the resulting quantum mechanical scattering problem @xmath130 can be solved analytically for the @xmath120 transfer matrix @xmath131 . since @xmath132 only acts on the state @xmath133 , @xmath122 has the following form , @xmath134 where @xmath135 is a @xmath136 submatrix that connects the states @xmath137 , i.e. states that contain @xmath138 . in order to determine this submatrix , we split the @xmath114-axis into 3 regions @xmath139 , @xmath140 , and @xmath141 . within each region , the hamiltonian @xmath142 can be diagonalized . requiring continuity and differentiability , we obtain the following explicit expression for the submatrix @xmath135 . in this expression , we have transformed the basis by using the bonding and antibonding dot states @xmath143 and @xmath144 from eqs . ( [ bondingstate ] ) , ( [ antibondingstate ] ) instead of the states @xmath118 and @xmath103 . in this basis , we obtain @xmath145 where the submatrices @xmath146 read for @xmath147 , @xmath148 @xmath149 @xmath150 where @xmath151 . the @xmath152 matrix @xmath153 is defined as follows : @xmath154 the coefficients @xmath155 , @xmath156 and the wave vectors @xmath157 , @xmath158 ( @xmath35 ) are given by ( @xmath159 ) @xmath160 @xmath161 @xmath162 @xmath163 where @xmath164 . note that the electron wave vectors @xmath157 arise from the region @xmath139 and from @xmath141 , whereas @xmath158 arises from the central region . the next step is to determine the transfer matrices @xmath165 and @xmath166 from the following representations of ideal rotation and phase gates , @xmath167 @xmath168 where @xmath7 is the phase gate angle . we assume that these gates do not introduce back - scattering . in this case , we can apply @xmath108 and @xmath109 separately for both directions of propagation . this leads to unitary matrices @xmath169 with the following nonvanishing elements @xmath170\quad(j=1,8).\end{aligned}\ ] ] this completes the calculation of the total transfer matrix @xmath20 . the final step is to determine the transmission function which requires us to set an initial condition for the scattering problem . we consider the situation where the propagating electron enters the pair of qwrs from the left in state @xmath116 . we further assume that the electron in the double quantum dot lies initially in the binding state @xmath144 . the eight final state coefficients represent the transmission amplitudes into the four states @xmath171 ( @xmath172 , @xmath35 ) for each of the two asymptotic propagation directions ( to the right or to the left ) . we denote these final state coefficients by @xmath173 and @xmath174 , respectively . by construction , the two rotation gates and the phase gate do not introduce any back - scattering . the amount of back - scattering due to the interaction @xmath132 depends on the ratio of the interaction strength @xmath126 and the length @xmath175 of the interaction region and can be made arbitrarily small . thus , the transmission amplitudes @xmath174 are negligible . in addition , we only need the transmission probabilities @xmath176 and @xmath177 that describe the total probability of an electron to arrive at @xmath116 or @xmath101 , irrespective of the state of the double dot electron , @xmath178 the resulting transmission probabilities @xmath176 and @xmath177 will be presented in the following section for different values of the tunneling coupling . they will be compared with the numerical results obtained with the realistic model described in sec . [ method ] . indeed , the above comparison will show that the two models are able to reproduce the basic physics of the device and that the device is able to produce an entangled state of the two qubits . the method presented in the sec . [ method ] allows us to calculate the ballistic current through entangled two - particle systems . concretely , we first present results for a double - qwr based mach - zehnder interferometer and show that this interferometer can be employed as a fully controllable single - qubit gate . based on this system , we then propose a novel two - qubit quantum transport device that allows the controlled generation and measurement of entanglement between the qwr qubit and the double quantum dot qubit . the measurement involves only the dc @xmath3-@xmath4 characteristics , no higher order current correlations or magnetic fields are required to detect entanglement . we would like to emphasize that the subject of this paper is to show that these devices are well suited to prepare arbitrary single - qubit and entangled two - qubit states . actual quantum computing applications , however , require additionally the performance of these operations in a time - resolved fashion with single electrons . a schematic view of the proposed two - qubit device is shown in fig . [ schematicdevice3d ] . it is based on a gaas / algaas heterostructure that consists of two vertically stacked 2degs . the 2degs are locally depleted by the application of negative voltages to appropriate metal top and bottom gates , that are not shown in this figure . for suitable gate voltages , the top gates create a pair of parallel qwrs in the upper 2deg that are connected to each other by two coupling windows . for specific energies , these coupling windows act as beam - splitters and the qwr network behaves as an electronic mach - zehnder interferometer . in contrast to electronic mach - zehnder interferometers based on quantum hall edge channels,@xcite no magnetic fields are employed . we therefore use the term _ all - electric mach - zehnder interferometer _ for this kind of device . we note that the relative phase shift between the electronic wave function in the two qwrs is controlled electrostatically by phase gates . these are additional top gates , but may be combined with the gates that are used to define the qwrs in the first place . the lower 2deg contains the two electrostatically defined tunneling - coupled single - electron quantum dots . an electron in the upper 2deg couples to the electron in the double quantum dot in the lower 2deg by means of the coulomb interaction . we have designed the entire device so that the consequences of entanglement become markedly evident in the observables , and we refer to it as _ entangled mach - zehnder double quantum dot device_. the mach - zehnder interferometer is operated under ballistic transport conditions . by applying a small dc bias voltage @xmath179 between the upper left source @xmath97 and the remaining three drain contacts @xmath180 , the device can be used in such a way that current flows predominantly from @xmath97 to @xmath181 ( denoted by @xmath182 ) or to @xmath183 ( denoted by @xmath184 ) , as indicated in fig . [ schematicdevice3d ] . first , we turn to the @xmath3-@xmath4 characteristics of the single - electron device , i.e. the all - electric mach - zehnder interferometer alone . the calculations have been based on the gaas / algaas heterostructure that is depicted schematically in fig . [ heterostructureandmetalgates ] . it consists of a @xmath185 thick gaas cap layer [ see vertical cross section in fig . [ heterostructureandmetalgates ] ( a ) ] , followed by a @xmath186 barrier of al@xmath187ga@xmath188as , and a @xmath189 gaas 2deg layer . beneath , there lies a @xmath190 al@xmath187ga@xmath188as substrate . a silicon @xmath191 doping layer with a concentration of @xmath192 is located @xmath193 below the surface . the electron sheet density in the 2deg layer has been calculated to be @xmath194 for the ungated sample at a temperature of @xmath195 . ga@xmath188as is shown in gray . the gaas cap layer is 5 nm thick . a silicon @xmath191 doping layer with a concentration of @xmath192 is located @xmath193 below the surface ( dashed line ) . ( b ) top view of the mach - zehnder interferometer that depicts the side gates ( black ) with voltage @xmath196 and the mid gates ( gray ) with voltage @xmath197 that define the two qwrs and the coupling windows between them . on top of the wires , there are two additional phase gates ( hatched ) at a bias of @xmath198 . ] figure [ heterostructureandmetalgates ] ( b ) shows a top view of the structure including the gates . the source and drain contacts ( cf . fig . [ schematicdevice3d ] ) are not shown in this figure . the device is @xmath199 wide and @xmath200 long . here we distinguish three types of gates : side gates ( black ) , mid gates ( gray ) , and phase gates ( hatched ) . each of the two side gates are , respectively , @xmath201 wide , @xmath200 long , and are biased at @xmath202 with respect to the source . the three mid gates are @xmath203 wide and @xmath204 , @xmath199 , and @xmath204 long , respectively . we apply a gate voltage of @xmath205 to the mid gates . they define two qwrs ( white ) with a nominal width of @xmath206 as well as two coupling windows , each with a nominal length of @xmath207 in the direction parallel to the wires . all of these structural parameters have been chosen to guarantee optimal device operation , while allowing for its fabrication with current technologies . the subband spacing of the two lowest subbands in each of these qwrs amounts to @xmath208 . it is important for the interferometer that the fermi wave length is close to the length of the coupling windows . indeed , our calculations yield a fermi wave length of @xmath209 in the lowest subband . in order to control the relative phase of the electron wave function in the two qwrs , a small gate voltage @xmath210 and @xmath211 is applied to the phase gates , respectively . the two phase gates are centered on top of the two qwrs and are each @xmath212 long and @xmath213 wide . a separating layer of resist could be used to reliably insulate the phase gates from the other gates . figure [ densityandpotential ] shows the equilibrium charge density ( upper panel ) and the corresponding potential ( lower panel ) in the 2deg layer of the mach - zehnder interferometer . the black and the white framed rectangles indicate the position of the metal top gates . the upper panel shows the 2deg to be fully depleted underneath the gates . in addition , it is evident from the figure that the qwrs and the coupling windows are formed indeed . the wires are strictly one - dimensional in the sense that only the lowest subband contributes to the density . the lower panel shows the potential barrier that separates the two qwrs . the energy scale is chosen so that the fermi level lies at zero @xmath214 . ( solid ) , @xmath215 ( dashed ) , and @xmath216 ( dotted ) through the mach - zehnder interferometer as a function of the phase gate voltage @xmath217 for a nominal length of the coupling windows of @xmath207 . the source - drain voltage @xmath218 is set to @xmath219 . ( b ) same as ( a ) , but for a coupling window length of @xmath220 . ] figure [ currentmachzehnder ] ( a ) shows the currents @xmath182 ( solid ) , @xmath184 ( dashed ) , as well as the back - scattered current @xmath221 ( dotted ) that flow from the source @xmath97 to @xmath222 , @xmath223 , and @xmath224 , respectively . the currents have been calculated by assuming a dc bias voltage of @xmath219 and a temperature of @xmath225 . the figure shows the dependence of the currents on the voltages @xmath226 and @xmath227 applied to the two phase gates , respectively . we first note that the back - current @xmath221 is smaller than @xmath228 for the entire relevant range of gate voltages and therefore negligible . in contrast to @xmath221 , the currents @xmath182 and @xmath184 strongly depend on the gate voltage and oscillate between @xmath229 and @xmath230 for more than @xmath231 oscillation periods . the currents @xmath182 and @xmath184 are seen to be phase - shifted relative to one another by 180 degrees which confirms that the device actually behaves as a mach - zehnder interferometer . note that this mach - zehnder interference pattern becomes damped out once the absolute gate voltage @xmath232 exceeds @xmath233 because this increases the amount of back - scattering . back - scattering can be suppressed , however , by increasing the length of the device , because the phase shift depends linearly on the length of the phase gate , whereas back - scattering is independent of the length but increases with the magnitude of @xmath234 . for comparison , fig . [ currentmachzehnder ] ( b ) depicts the calculated currents @xmath182 , @xmath184 , and @xmath221 for the same device but for a nominal length of the coupling windows that has been reduced by @xmath235 . the difference in results shows the sensitivity of the interference pattern to small changes in geometry , and implies very high demands on fabrication precision . figure [ densitiesopendevice ] shows the charge densities of the stationary current carrying states of the mach - zehnder interferometer for selected phase gate voltages . the three figs . [ densitiesopendevice ] ( a ) , ( b ) , and ( c ) correspond to gate voltages of @xmath236 , @xmath237 , and @xmath238 , respectively . the three graphs show that the quantum mechanical charge densities change predominantly near the drain contacts on the right hand side of the device . these changes reflect the redistribution of the total current between @xmath182 and @xmath184 . in fig . [ densitiesopendevice ] ( a ) , the charge density is zero near the upper right terminal but large near the lower right terminal . this agrees with fig . [ currentmachzehnder ] ( a ) that shows that the current @xmath182 is minimal ( @xmath184 maximal ) for the corresponding gate voltage @xmath239 . in fig . [ densitiesopendevice ] ( b ) , on the other hand , the charge density is seen to be almost equal near both right drain contacts , again in accord with fig . [ currentmachzehnder ] ( a ) for @xmath240 @xmath241 . finally , the charge density is large near the upper right terminal but zero at the lower right terminal in fig . [ densitiesopendevice ] ( c ) , again in correspondence with fig . [ currentmachzehnder ] ( a ) . of ( a ) 0 mev , ( b ) -3.6 mev , ( c ) -7.5 mev . in all cases , the source - drain voltage @xmath218 is set to @xmath219 . ] all of these effects are a consequence of the self - interference of the electron wave function . depending on the relative phase between the two qwrs , the interference at the second beam splitter leads to partial extinction either near the upper or the lower right terminal . the dc transfer characteristics shown in fig . [ currentmachzehnder ] ( a ) exhibits multiple pronounced maxima and minima that can be attributed to rotations of a qubit on the bloch sphere . before entering into the discussion of the results , let us give a rigorous definition of the mach - zehnder interferometer qubit state . to this aim , we depict the device schematically in fig . [ qwrqubitdef ] where we show the regions of the two coupling windows and the phase gate that perform the @xmath108 and @xmath109 transformations of eqs . ( [ eqrtransf ] ) and ( [ eqptransf ] ) , respectively . and one phase gate @xmath109 , and four other segments where the qubit state is well defined ( see table [ tableqwrqubit ] ) . ] we define the basis states @xmath31 and @xmath32 of the qwr qubit by an electron scattering state localized in the upper and lower qwr , respectively . this definition holds only in the regions where the two channels are well separated , as in the segments , , , and of fig . [ qwrqubitdef ] . by contrast , the qubit state is not well defined within the window regions @xmath108 , where the barrier between the qwrs is very small . as an example , the approximate qubit states for the four regions indicated in fig . [ qwrqubitdef ] are included in table [ tableqwrqubit ] , and specified for each of the three cases depicted in fig . [ densitiesopendevice ] . indeed , in the calculation of fig . [ densitiesopendevice ] , the split gate voltages have been chosen in order to obtain a qubit rotation angle @xmath242 and phase angles @xmath243 in ( a ) , ( b ) , ( c ) , respectively [ cf . ( [ eqrtransf ] ) and ( [ eqptransf ] ) ] . .quantum state of the mach - zehnder qubit in the four device segments indicated in fig . [ qwrqubitdef ] ( columns ) , corresponding to the three situations shown in fig . [ densitiesopendevice ] ( rows ) , respectively . [ cols= " < , < , < , < , < " , ] between source @xmath97 and drain @xmath222 as a function of kinetic energy in mev and as a function of the coupling window length @xmath175 in nm . the total width of the structure equals @xmath244 nm , whereas the width of the inner coupling windows amounts to @xmath245 nm . the darker the color , the smaller the value of @xmath246 . the dark lines therefore indicate minima in @xmath246 . the arrows point to crossings of resonances where @xmath246 vanishes . ] ( solid ) , @xmath247 ( dashed ) , and @xmath248 ( dotted ) , as a function of the kinetic energy of the electron in mev . the length @xmath175 of the coupling window equal has been set to @xmath249 . the other symbols have been defined in figure [ schematicdevice3d ] . ] since it is unlikely that one can fabricate the device discussed in fig . [ heterostructureandmetalgates ] precisely , it is imperative to understand the dependence of the mach - zehnder interferences on the various device geometry parameters such as width and length of the qwrs and the coupling windows not only quantitatively but also qualitatively . a physically transparent picture can be obtained by studying a model h shaped structure that consists of a single coupling window where the depletion potentials are hard - wall potentials as depicted in the inset of fig . [ transmissionhstructure ] . the coupling between the two qwrs causes the transmission to become resonant for specific energies that depend on the length @xmath175 and width @xmath250 of the coupling window . these energies are approximately equal to the energies of the bound states in a rectangular box with the dimensions of the coupling window , i.e.@xmath251 beam splitting , i.e. a crossover in the transmission from channel @xmath97-@xmath222 to @xmath97-@xmath223 , requires that the corresponding transmission functions obey the relation @xmath252 for some energy @xmath253 . in turn , this condition requires that there are 2 sets of values @xmath254 that correspond to the same energy @xmath113 , i.e. @xmath255 . thus , the crucial geometric requirement for quantum mechanical switching is to achieve an appropriate ratio of length @xmath175 and width @xmath250 that obeys this condition . in order to obtain a pronounced resonance , the values of @xmath256 should be of the order of unity additionally , since the resonances lie densely in energy for large integers . figure [ transmissionhstructure ] shows a contour graph of the transmission probability @xmath176 ( from source @xmath97 to drain @xmath181 ) as a function of the coupling window length @xmath175 and the kinetic energy of the electron . the arrows in this figure point to suitable values of the length @xmath175 where the energies of two eigenstates cross . to further illustrate this crossover in the transmissions , fig . [ transmissionresonance ] shows a cross - section through fig . [ transmissionhstructure ] for the length @xmath257 that is marked by the leftmost white arrow . this crossing corresponds to the switching from channel @xmath97-@xmath222 to @xmath97-@xmath223 and occurs near the degeneracy of the energy states @xmath258 and @xmath259 . in the following , we present the results for the entangled mach - zehnder double quantum dot device that has been schematically depicted in fig . [ schematicdevice3d ] . for these calculations , we have used a somewhat different set of geometry parameters for the mach - zehnder interferometer in order to optimize the entanglement . since the calculations of the entangled system are very demanding , we have defined the mach - zehnder interferometer by a hard - wall potential within a @xmath189 thick slab of gaas rather than performing a charge self - consistent calculation including all metal gates . we have taken the two qwrs to be @xmath260 wide , @xmath261 long , and the lateral distance between them has been set to @xmath262 . the coupling windows now have a length of @xmath263 . the resulting energy spacing of the two lowest subbands amounts to @xmath264 which is nearly twice as large as in the mach - zehnder device of sec . [ results ] a. the fermi level @xmath22 has been set to @xmath265 in the lowest subband . this fermi level causes the two coupling windows to act as almost perfect beam - splitters , with channel transmissions @xmath266 . for the calculation of the interaction between the mach - zehnder interferometer and the double quantum dot , we assume a vertical distance of @xmath267 between them . the larger of the two quantum dots with a lower ground state energy is located exactly underneath the center of the mach - zehnder interferometer whereas the smaller one lies @xmath268 away in the direction perpendicular to the qwrs . the double quantum dot system is modeled by the hamiltonian given in eq . ( [ modelhamiltonianquantumdots ] ) . the positions of the mach - zehnder interferometer and the double quantum dot relative to each other have a large influence on the operation characteristics of the device . the present configuration exploits the idea that the interaction between the electron in the qwrs and the electron in the larger quantum dot is identical for both wires . the electron in the smaller quantum dot , on the other hand , lies closer to the qwr between gates @xmath269 and interacts mostly with the electron in that wire . consequently , a phase difference in the electron wave function for the two qwrs is exclusively induced by the electron in the smaller dot . and @xmath184 in na as a function of the phase gate voltage @xmath217 in mv for two different quantum dot tunneling couplings @xmath25 in @xmath270 . the inset is a simplified version of figure [ schematicdevice3d ] . ] at first , we study the transfer characteristics of the device for a situation where the tunneling between the quantum dots is inhibited ( @xmath271 ) . the ground states of the larger and the smaller quantum dot are assumed to differ by @xmath272 . the resulting current in the qwrs is shown in fig . [ currententangler ] by the dashed curves . we obtain the typical interference pattern that we have also found in sec . [ results ] a for the all - electric mach - zehnder interferometer . the asymmetry with respect to the sign of the gate voltage is caused by the asymmetric position of the phase gate along only one of the two qwrs in combination with the additional repulsive potential due to the electron in the double quantum dot . , as a function of the tunneling coupling in @xmath270 . ] . the upper ( lower ) panel shows the single - particle electron density in the qwrs when the electron in the quantum dot occupies the bonding ( anti - bonding ) state . the quantum dot tunneling coupling amounts to @xmath273 . ( b ) same type of figure , but with the coulomb interaction between the electron in the wires and the electron in the quantum dot switched off . in case ( b ) , the two possible paths of the scattered electron can interfere with one another . in case ( a ) , however , the electron propagating in the lower wire changes the double - dot state and looses its capability to interfere with the other path , thus reducing the visibility ( see text ) . ] we now allow the electron in the double quantum dot to tunnel between the two dots ( @xmath273 ) . the currents @xmath182 and @xmath184 that correspond to this situation are shown in fig . [ currententangler ] by the solid curves . we still obtain the mach - zehnder interference pattern , but the visibility is markedly reduced . let us define the visibility by @xmath274 i.e. for zero phase gate voltage @xmath217 . figure [ visibilityvscoupling ] shows the visibility as a function of the tunneling coupling . the visibility is almost @xmath41 for vanishing tunneling coupling and decreases monotonously with increasing tunneling . the visibility is therefore uniquely related to the magnitude of the tunneling coupling . the above effect can be explained as follows . if the tunneling probability is high , the double - dot electron can easily change its state . in this regime , it is also easy for the double - dot to _ measure _ the path of the interferometer particle , thus causing decoherence and destroying the interference . for low tunneling probability , on the other hand , the electron that sits initially in the large dot is insensitive to the coulomb interaction with the interferometer . when the tunneling is completely suppressed , the state of the dot is fixed and the interferometer electron is unable to change the dot state . while the electrons are still interacting due to their charge in this case , the path of the interferometer particle is not measured by the double - dot . figures [ entangleddensity ] ( a ) and ( b ) show the electron density eq . ( [ projecteddensity ] ) in the entangled mach - zehnder double quantum dot device . to be precise , the figures depict the projection of the probability densities of the current - carrying two - particle states onto the eigenstates of the isolated double quantum dot for a tunneling coupling of @xmath273 and vanishing phase gate voltage @xmath217 . in particular , fig . [ entangleddensity ] ( a ) shows the interacting case , whereas fig . [ entangleddensity ] ( b ) shows the non - interacting case where the electron - electron interaction between the electrons in the mach - zehnder interferometer and the quantum dot electron has been switched off . thus , the two figures illustrate the effect of the coulomb interaction on the charge density . in each of the two figures , the upper panel corresponds to the ground state , the lower panel corresponds to the excited state of the double quantum dot , respectively . in the interacting case ( a ) , we find a finite probability of the electron in the double quantum dot to be in the excited state . this has to be compared to the case of vanishing interaction ( b ) as well as to the case of zero tunneling coupling . in each of the latter cases , the probability of the electron in the double quantum dot to be in the excited state is zero this is shown in fig . [ entangleddensity ] ( b ) . note that for cases ( a ) and ( b ) , the total probability densities that are given by the sum of the two projected probability densities are almost identical near the left contacts , but differ significantly near the right contacts . in ( a ) , the total probability density near the upper right contact is only partially extinct by interference , whereas in ( b ) , the total probability density vanishes near the upper right contact . this indicates that the interaction between the mach - zehnder interferometer and the double quantum dot destroys the self - interference of the qwr - electron wave function . the results that we have obtained for the entangled mach - zehnder double quantum dot device are indeed the result of a two - qubit quantum operation . in order to show this , we apply the analytical model developed in sec . [ analyticalmodel ] to our two - qubit device . figure [ transmissionanalytical ] shows the transmission probabilities @xmath176 and @xmath177 , computed from eqs . ( [ transmission1 ] ) and ( [ transmission2 ] ) , for two different values of the quantum dots tunneling . indeed , the comparison with fig . [ currententangler ] shows that the model is able to reproduce the basic physics of the device and to yield transmission probabilities @xmath176 and @xmath177 that qualitatively agree with the results obtained by the numerical simulations described in secs . [ method ] and [ numericaldetails ] . in particular , figs . [ currententangler ] and [ transmissionanalytical ] show a similar dependence of the visibility on the tunneling coupling . the reason for the suppression of the visibility is the entanglement of the two qubits . in fact , from the point of view of the individual qubit , its entanglement with the other one is nothing else but decoherence and this decoherence partially suppresses the interference of the electron wave function in the two qwrs . in other words , the reduced density matrix of each of the qubits , obtained by tracing out the degrees of freedom of the other qubit , represents a mixed state if and only if the two qubits are entangled . thus , the observed suppression of the visibility is both a direct measure of the degree of entanglement between two qubits and of the decoherence undergone by the qwrs qubit . we stress that our approach does not include other sources of decoherence as , for example , electron - phonon interaction or charge fluctuations in the metallic split gates that define the structure . therefore , our reduction of the visibility is obviously ascribed to the quantum entanglement of the two qubits . on the other hand , in real experiments the visibility will be reduced by the coupling of the interferometer with any external degree of freedom and not only with the other qubit . what makes our results functional to experiments is the estimation of the visibility decrease against tunable parameters of the realistic device , as , for example , the tunneling coupling of the second qubit . in fact , it is unlikely that the effect of the environment will be tuned by the same parameters that affect the amount of entanglement between the two qubits . the phonon effects on the interferometer , for example , are determined solely by the temperature , the structure composition and the energy of the traveling electrons . thus , our proposed device demonstrates from a modeling perspective a controlled generation of two - qubit entanglement . the predicted behavior of the visibility against various system parameters may serve as an entanglement witness in experimental realizations of the system . in the remainder of this section , we show that the relation between the visibility of our device , as defined in eq . ( [ visibility ] ) and depicted in figs . [ visibilityvscoupling ] and [ transmissionanalytical ] , and the entanglement can be quantified in terms of the von neumann entropy of the reduced density matrix @xcite that is a measure of the degree of entanglement . and @xmath247 as a function of the phase angle in rad for the analytical model of the entangled mach - zehnder double quantum dot device . the full line is for a finite tunneling coupling @xmath275 in units of @xmath276 , whereas the dashed curve is for vanishing tunneling coupling that corresponds to no entanglement . ] to simplify the derivation , we generally assume that back - scattering is negligible in the directional couplers and phase gates . this reduces all matrices to size @xmath136 and allows the analytical evaluation of the transmissions @xmath246 and @xmath247 in eqs . ( [ transmission1 ] ) and ( [ transmission2 ] ) . the density matrix of the final state is then given by @xmath277 the amount of entanglement of the final two - qubit state is given by the von neumann entropy of the reduced density matrix , namely @xmath278 where the reduced density matrix @xmath279 is obtained from @xmath8 by tracing out the electron in the double quantum dot . this yields the @xmath152 matrix @xmath280 where the so - called coherence @xmath281 is given by @xmath282 with the eigenvalues of the reduced density matrix , @xmath283 the von neumann entropy reads @xmath284 we now show that the eigenvalues @xmath285 of @xmath279 can be expressed in terms of the visibility @xmath286 , eq . ( [ visibility ] ) , of the interference pattern according to @xmath287 first , we determine the explicit expressions for the transmission functions as a function of the angle @xmath7 of the phase gate . to this end , we first apply the @xmath110 rotation gate @xmath108 and the two - qubit gate @xmath112 onto the initial state @xmath104 . the probabilities @xmath288 of the four two - particle basis states @xmath289 ( @xmath172 , @xmath35 ) of the resulting state are given by @xmath290 where @xmath291 and @xmath292 correspond to @xmath293 and @xmath137 , respectively , and @xmath294 and @xmath295 are angular constants that depend on the two - particle interaction @xmath112 . they describe the linear combination of the dot - electron states @xmath296 and @xmath297 within the two - particle wave function . the explicit expressions of @xmath294 and @xmath295 do not enter the final result of this section . the expressions eq . ( [ avisibility ] ) reflect the fact that the probability for the electron in the qwr to be either in state @xmath100 or in state @xmath101 are equal . the next step consists in applying the phase gate @xmath109 and the second @xmath110 rotation gate @xmath108 . this procedure yields the final state probability amplitudes @xmath298 note that the gates @xmath108 and @xmath109 , being single - qubit transformations , do not change the degree of entanglement of the two electrons . for the transmission probabilities @xmath176 and @xmath177 we finally obtain @xmath299 where the signs @xmath300 correspond to @xmath176 and @xmath177 , respectively . the phase angles are defined by @xmath301 and analogously for @xmath302 . by inserting the expressions for @xmath303 into eq . ( [ transmissionbasisstates ] ) , the transmission probabilities read @xmath304 , \ ] ] where @xmath305 and @xmath306 is a constant phase shift . this result shows that the transmission probabilities @xmath176 and @xmath177 yield a sinusoidal oscillation as a function of the phase angle @xmath7 . by comparing the definition eq . ( [ visibility ] ) of the visibility with this result for the transmission probabilities , we see that the visibility can be written in the form @xmath307 now , we evaluate the expression @xmath308 within eq . ( [ eigenvaluesreddensmatrix ] ) , and obtain @xmath309 inserting this result into eq . ( [ eigenvaluesreddensmatrix ] ) , we finally obtain the assertion eq . ( [ eigenvaluesvisibility ] ) . this result proves that the von neumann entropy of the reduced density matrix depends monotonously on the visibility . a visibility of @xmath310 corresponds to vanishing von neumann entropy ( zero degree of entanglement ) , while for a visibility of @xmath311 , we find a von neumann entropy of @xmath312 ( maximum degree of entanglement ) . while the relation eq . ( [ eigenvaluesvisibility ] ) between visibility @xmath286 and von neumann entropy @xmath97 may not hold rigorously for the realistic entangled mach - zehnder double quantum dot device , it is nevertheless interesting to calculate @xmath97 for this device as a function of the tunneling coupling , invoking the relation of eq . ( [ eigenvaluesvisibility ] ) . the result is presented in fig . [ entropyvscoupling ] and shows that the degree of entanglement of the realistic device monotonously increases with increasing tunneling coupling . c , as a function of the tunneling coupling in @xmath270 . ] we have theoretically analyzed semiconductor single- and two - qubit quantum gates based on electrostatically defined qwrs and quantum dots . we predict the detailed three - dimensional geometry , material composition , doping profile , and bias voltage of an all - electric mach - zehnder interferometer . our calculations of the electronic structure and ballistic transport properties of this device show that the proposed device is a fully controllable single - qubit gate for electrons that propagate in qwrs . the fabrication of the mach - zehnder interferometer is within the reach of present - day technology but full control of the electronic beam - splitters requires a very high fabrication precision in the range of a few @xmath313 . based on this all - electric mach - zehnder interferometer , we predict a two - qubit quantum transport device where the electrons in the interferometer couple to a single - electron double quantum dot by coulomb interaction . we have calculated the ballistic transport properties of the three - dimensional two - qubit device with an interacting two - particle quantum transport method and designed the device geometry for optimal entanglement . by means of an analytical model of the device , we have illustrated the qualitative physics of the two - qubit device and showed that the visibility is a faithful measure of the entanglement . in particular , we have found that the visibility can be controlled externally by tuning the tunneling coupling between the two quantum dots of the second qubit . the device realizes a non - trivial two - qubit gate that allows the controlled generation and straightforward detection of entanglement from dc current - voltage characteristics .

the herbig ae / be stars ( haebes ) allow an exploration of the properties of pre - main sequence(pms ) stars above the low - mass range ( ) and those bordering the high - mass range ( ) . this paper is the second in a series exploring accretion in 91 haebes with very large telescope / x - shooter spectra . equivalent width measurements are carried out on 32 different lines , spanning the uv to nir , in order to obtain their line luminosities . the line luminosities were compared to accretion luminosities , which were determined directly from measurements of an uv - excess . when detected , emission lines always demonstrate a correlation with the accretion luminosity , regardless of detection frequency . the average relationship between accretion luminosity and line luminosity is found to be . this is in agreement with the findings in classical t tauri stars , although the haebe relationship is generally steeper , particularly towards the herbig be mass range . since all observed lines display a correlation with the accretion luminosity , all of them can be used as accretion tracers . this has increased the number of accretion diagnostic lines in haebes tenfold . however , questions still remain on the physical origin of each line , which may not be due to accretion . [ firstpage ] stars : early - type stars : variables : herbig ae / be stars : pre - main sequence stars : formation accretion , accretion discs techniques : spectroscopic .

1610.09636

herbig ae / be stars , haebes , bridge an important mass range between high and low mass stars ; namely the intermediate masses of 28 . the haebes are pre - main sequence ( pms ) stars that display properties such as an infra - red excess due to a circumstellar disc @xcite , and complex spectral line profiles @xcite ; akin to their solar - mass counterparts , the classical t tauri stars , ctts ( see * ? ? ? * for a review ) . the haebes serve as excellent observational targets which can further our understanding of star formation , as they are readily observable at optical wavelengths unlike the massive young stellar objects , mysos , which are still heavily embedded in their natal clouds @xcite . theoretical models have long suggested that ctts accrete via magnetospheric accretion , ma @xcite . in this scenario the circumstellar disc is truncated at a given radius from the star by the stellar magnetic field . the magnetic field lines can then funnel disc material onto the star , at approximately free - fall velocities , via an accretion column . this in - falling material shocks the photosphere and gives rise to hotspots on the stellar surface . the result is an excess of ultra - violet ( uv ) energy which is detectable in addition to the regular photospheric emission . this phenomenon has been observed in numerous ctts to date @xcite , and has been shown to be present in haebes @xcite . provided the stellar parameters are known , the accretion luminosity can be straight - forwardly converted into a mass accretion rate , , one of the key astrophysical parameters in star formation . it has also been shown that the accretion luminosity ( ) in ctts , derived from an uv - excess or line veiling , is correlated with the luminosity of emission lines ( ) . this appears to hold true for a large number of lines , such as br@xmath4 , h@xmath5 , [ o]@xmath6 , and the ca ir - triplet @xcite . to date , the luminosity relationships for ctts have largely been shown to hold for haebes too , with only minor changes in intercept and gradient @xcite . this means that emission lines can be used as powerful accretion diagnostics and provide a method of inferring for stars where direct methods of measurement are difficult or impossible e.g. the uv - excess is often used to measure directly , but uv - observations are more difficult to obtain and interpret than simple emission lines ; emission lines are also available at a variety of different wavelengths making them a preferred choice over uv observations in highly extinct targets . therefore , by establishing correlations between the accretion luminosity and line luminosity for a large series of lines , possibilities to infer accretion rates when using data sets with a limited wavelength range are opened . changes in the accretion mechanism are suspected to be taking place towards the early - be stars , as evidenced by spectropolarimetric observations @xcite . it was also thought that the and relationship was breaking down towards hbes by @xcite . however , the observed change in the relationship can be explained by the adopted ma model which was for a single cool hae star . in fact , the observed trends between and have been explained by @xcite to be a consequence of the luminosity of the star . therefore , many observed relationships may not actually be directly due to accretion . however , any observed relationships can still be used to infer an accretion rate , meaning that they remain a valuable tool in this field . therefore , a large scale investigation into these relationships for haebes is required . in our previous paper in this series ( * hereafter ) , we used spectroscopic data from x - shooter , covering the uv , optical and near - infrared wavelengths , to determine the stellar parameters of a large sample of 91 haebe stars . out of this sample , it was possible to determine the uv - excess , and thereby the accretion luminosity and mass accretion rates , for 62 of them . the large wavelength range of the data allows for the detection and measurement of many emission lines ; observed simultaneously with the uv - excess . the large sample is therefore ideal to revisit the hitherto published line luminosities and accretion luminosity relationships , along with an investigation into new accretion diagnostics . the overall aim of this paper is to provide the measurement of numerous spectral lines across the entire sample of 91 haebes . the strengths of the lines are investigated , along with a determinations of line luminosities based on previous stellar properties and photometry . these are compared against the accretion rates determined in in order to provide a critical assessment of accretion luminosity versus line luminosity relationships in haebes . this paper is organised as follows . section [ sec : data ] provides details on the sample , data reduction , and techniques of emission line measurements and line luminosity calculations . section [ sec : results ] presents the measurements of the lines , calculated line luminosities , and the resulting correlations with the accretion luminosity . analysis of these relationships and the presentation of new accretion diagnostics are provided in this section too . finally , section [ sec : disc_and_conc ] provides a summary of the results and the concluding remarks . the sample , observations and basic data reduction processes are detailed in paper i. we give a brief summary of these here , while we discuss the correction for telluric absorption lines and go into detail on the measurement of line strengths . the stars were selected from the catalogues of haebe stars published by @xcite and @xcite . the main criterion to be included the sample was their observability in the southern hemisphere during the ( southern ) summer semester . in total the sample contains 91 stars , of which 37 have a mass greater than 3 , which we choose to classify as the herbig be stars . the remaining stars are all haes , with perhaps a few intermediate mass ctts which border the mass boundary . the observations were carried out with x - shooter mounted on the eso s vlt , ut3 , paranal , chile in the 2009 - 2010 season @xcite . the instrument provides complete coverage from 0.3 - 2.4 @xmath7 m over three separate arms ; the arms are split into the following : the uvb arm , 30005600 ; the vis arm , 550010200 @xmath8 ; and the nir arm , 10200 - 24800 @xmath8 . the observations were taken with the smallest possible slit widths resulting in a spectral resolution of 10,000 - 18,000 . the integration times were such that signal - to - noise ratios of 100 - 200 were typical . the bias subtraction , flat fielding and wavelength calibration were all performed using the eso pipeline reduction ( modigliani et al . when available , data were taken from the recent phase iii release with pipeline version 2.3.12 . for the small number where this was not the case , the reduced data from paper i ( pipeline version 0.9.7 ) are taken . in general , the quality of the reduced data in both cases is similar , except in the nir arm where the snr is better in the phase iii data . the spectra were not flux calibrated as a matter of course , and instead we use published photometry of all our sources when computing the final line luminosities ( see for details of photometry sources ) . since the typical errors in the flux determination from spectrophometric data amount to around 20% , this choice is warranted for objects that vary less than 20% ( or roughly 0.2 magnitude ) . it has been shown in pms - stars that the variability amplitude decreases with increasing wavelength e.g. for a change in the v - band of 0.4 mag changes in the k - band are generally less than 0.2 mag @xcite . spot - checks on our targets indicate that the variability is indeed at most of this order on average , validating the choice of using existing photometry . due to the large sample used any objects with large variability changes , that could produce anomalous results , are likely to be identified or their effects diluted by the sheer number of other targets used . before measurements of the lines are undertaken , telluric absorption features in the spectra must be removed . this is particularly necessary for the majority of the observable emission lines in the near - infrared ( nir ) arm , but also for a few cases in the vis arm covering the optical . a telluric standard star can be used for such a correction , but these stars contain their own spectral features , which also need to be corrected for . since many of the lines observed are hydrogen recombination lines , which are commonly seen in absorption in telluric standard stars , correction is instead performed using the eso molecfit telluric removal tool @xcite . there are a few additional issues that contributed to rejecting the telluric standard and using molecfit instead : airmass could differ by large amounts ( mostly uncommon ) , and short exposure times of the tellurics ( on average the snr was half that of the targets ) . molecfit provides an atmospheric transmission spectrum using a radiative transfer code based upon a combination of the observatory conditions at the time and the features observed in the target spectra . removal of the lines using both molecfit and telluric standards were compared and it was found that using molecfit yielded better quality spectra over using the telluric standards ; the snr of the latter contributes to a poorer reduction . therefore , molecfit correction is performed on all spectral orders in each arm that contains telluric line features , and will be used for the remainder of this paper ( see also * ? ? ? * where molecfit is used in the same way to study the co overtone emission at 2.3@xmath7 m ) . after telluric corrections , line strengths are recorded by measuring the equivalent widths , ew , of the lines after normalising the spectra of each star to the continuum level . these direct measurements of the ew will be referred to as . in order to obtain the true line strengths of the emission lines needs to be corrected for both the underlying photospheric absorption , , and the presence of any excess continuum emission , which would dilute the underlying absorption . the photospheric absorption depends upon the temperature , surface gravity , and metallicity of the stars . the temperature and surface gravity of each star were determined in and are used to select an appropriate spectral model atmosphere , from which the absorption line strengths are measured . in this work the munari set of kurucz - castelli models , kc - models , are used for lines with @xmath9 , due to their 1 sampling @xcite ; a metalicity of [ m / h]=0 is adopted . for longer wavelengths additional models are required , the older kc - models are used here computed by @xcite . for consistency , the same region of measurement used for is used for measuring . figure [ fig : ew_curves ] provides an example of the intrinsic absorption strength as a function of temperature for some of the lines ( log(g ) is fixed in this example ) . the net emission , which is stronger than the straightforward measurement of would imply , can now be corrected for the intrinsic absorption . the corrected equivalent width , , is calculated as follows : @xmath10 where @xmath11 is the strength of the continuum excess at the wavelength in question , measured in magnitudes . it is the difference between the observed , dereddened magnitude and the intrinsic magnitude of a star of the same spectral type . figure [ fig : ux_ori ] demonstrates the importance of correcting for intrinsic absorption . in the figure it can be seen that ux ori definitely has emission present . however , when measuring a positive value is obtained i.e. an absorption line . correcting for makes negative , and therefore in emission . the benefit of this approach is that the result encompasses all of the true emission . at near - infrared wavelengths excess emission is also present , due to thermal emission from circumstellar dust heated by the central star . i.e. there is a non - zero @xmath12 component , for the majority of stars , because of the ir - excess . in order to obtain @xmath13 the intrinsic magnitude is first obtained by scaling a kc - model of the same stellar parameters to match the dereddened visible photometry of the star and measuring the flux from the model at the wavelength in question ( the compiled photometry is listed in paper i ) . to avoid underestimating the continuum flux , which can be affected by the underlying absorption in the model , an interpolation of the continuum regions either side of the line is used to derive the continuum flux . this is then converted into a magnitude to give the expected intrinsic magnitude ( since this is at an arbitrary wavelength no convolutions with pass - bands are involved ) . next , this must be compared against the observed dereddened magnitude . this makes use of the @xmath14 photometric data from 2mass @xcite . the photometry is dereddened and converted into a flux using the @xmath15 values determined in , and then an interpolation is made between the photometric points either side of the wavelength in question ( and converted back into a magnitude ) . with both the observed and expected intrinsic magnitudes obtained , @xmath12 is found via the difference ; can now be calculated . a consideration of errors needs to be made in all the above steps . let us first start with the line measurement itself . since is a numerical calculation over a wavelength region , the error will be the extent of this region , @xmath16 divided by the snr of the continuum : @xmath17 . the error on is also complicated by stellar parameters , metallicity , and the adopted set of models . for this reason a generous error of 10% is adopted for all measured from the munari set of kc - models ; an error of 20% is adopted for the nir lines measured from the lower resolution models . errors in @xmath11 are low , due to the quality of the @xmath14 photometry , and the relatively low errors in the stellar parameters . since the ew errors dominate over @xmath11 only the and errors are considered in . with established , a line flux , , can now be calculated by = @xmath18 , where @xmath19 is the continuum flux corresponding to the central wavelength of the line . for all lines with line centres @xmath20 , @xmath19 is calculated from a kc - model in the same way the intrinsic magnitude was calculated earlier ( but omitting the final magnitude conversion ) . for the lines with line centres @xmath21 , an interpolation of the dereddened @xmath14 photometry is performed in the same manner as was done for obtaining the ir - excess magnitude earlier , except a further step is required of converting this into a flux . expanding upon this , a line luminosity , , is also calculated by : @xmath22 , where @xmath23 is the distance to star ( the values used are complied in ) . the value takes into account both errors on the distance and the stellar parameters used to calculate . a complete list of errors on are included in the online version of table [ tab : ew_sample ] . in total , , , and the corresponding line luminosities are calculated for a set of 32 different lines . the lines were selected based on being observed previously in the literature , or by having a high detection rate throughout the sample ; preferably both . these lines span the full wavelength range covered . to put the number of lines into perspective , we note that lines previously identified in the literature as accretion diagnostics for haebes stars are limited to h@xmath5 , [ oi]6300 and br@xmath4 @xcite ; hence the number of lines investigated in this manner has increased tenfold . the ew , , and data for the h@xmath5 line are provided in table [ tab : ew_sample ] . the table contains this information for all of the observed targets . a further 31 lines have been measured and are available online . the following sections investigate the empirical relationship between the line luminosities and the accretion luminosities , along with a comparison against ctts . l rrr rcc + name & & & & & & log ( ) + & & & & & ( wm@xmath24@xmath25 ) & [ ] + ux ori & 2.39 @xmath26 0.34 & 14.65 @xmath26 1.47 & - & -12.26 @xmath26 1.51 & 3.27 ( @xmath26 0.40 ) e -15 & -1.43 @xmath26 0.10 + pds 174 & -54.19 @xmath26 1.34 & 6.46 @xmath26 0.65 & - & -60.65 @xmath26 1.49 & 2.33 ( @xmath26 0.06 ) e -14 & -0.03 @xmath26 0.09 + v1012 ori & 4.74 @xmath26 0.55 & 16.36 @xmath26 1.64 & - & -11.62 @xmath26 1.73 & 1.42 ( @xmath26 0.21 ) e -15 & -2.06 @xmath26 0.11 + hd 34282 & 4.18 @xmath26 0.40 & 16.23 @xmath26 1.62 & - & -12.05 @xmath26 1.67 & 3.03 ( @xmath26 0.42 ) e -15 & -1.90 @xmath26 0.11 + hd 287823 & 10.18 @xmath26 0.50 & 15.58 @xmath26 1.56 & - & -5.40 @xmath26 1.64 & 1.64 ( @xmath26 0.50 ) e -15 & -2.23 @xmath26 0.16 + hd 287841 & 8.36 @xmath26 0.43 & 13.22 @xmath26 1.32 & - & -4.86 @xmath26 1.39 & 9.64 ( @xmath26 2.75 ) e -16 & -2.46 @xmath26 0.15 + hd 290409 & 0.35 @xmath26 0.55 & 15.00 @xmath26 1.50 & - & -14.65 @xmath26 1.60 & 3.18 ( @xmath26 0.35 ) e -15 & -1.58 @xmath26 0.10 + hd 35929 & 1.54 @xmath26 0.69 & 8.69 @xmath26 0.87 & - & -7.15 @xmath26 1.11 & 1.09 ( @xmath26 0.17 ) e -14 & -1.35 @xmath26 0.11 + hd 290500 & -1.46 @xmath26 0.25 & 13.14 @xmath26 1.31 & - & -14.60 @xmath26 1.33 & 1.26 ( @xmath26 0.11 ) e -15 & -1.04 @xmath26 0.10 + hd 244314 & -14.04 @xmath26 0.61 & 15.54 @xmath26 1.55 & - & -29.58 @xmath26 1.67 & 6.91 ( @xmath26 0.39 ) e -15 & -1.38 @xmath26 0.09 + hk ori & -61.86 @xmath26 0.72 & 15.79 @xmath26 1.58 & - & -77.65 @xmath26 1.74 & 1.51 ( @xmath26 0.03 ) e -14 & -1.04 @xmath26 0.09 + hd 244604 & 2.32 @xmath26 0.51 & 14.79 @xmath26 1.48 & - & -12.47 @xmath26 1.57 & 5.72 ( @xmath26 0.72 ) e -15 & -1.46 @xmath26 0.10 + uy ori & 4.98 @xmath26 0.26 & 15.28 @xmath26 1.53 & - & -10.30 @xmath26 1.55 & 4.81 ( @xmath26 0.72 ) e -16 & -1.80 @xmath26 0.11 + hd 245185 & -13.56 @xmath26 0.50 & 14.48 @xmath26 1.45 & - & -28.04 @xmath26 1.53 & 6.54 ( @xmath26 0.36 ) e -15 & -1.26 @xmath26 0.09 + t ori & -4.15 @xmath26 0.43 & 13.07 @xmath26 1.31 & - & -17.22 @xmath26 1.38 & 1.04 ( @xmath26 0.08 ) e -14 & -0.74 @xmath26 0.09 + v380 ori & -81.88 @xmath26 0.48 & 13.61 @xmath26 1.36 & - & -95.49 @xmath26 1.44 & 9.74 ( @xmath26 0.15 ) e -14 & -0.48 @xmath26 0.09 + hd 37258 & -0.33 @xmath26 0.39 & 15.00 @xmath26 1.50 & - & -15.33 @xmath26 1.55 & 4.88 ( @xmath26 0.49 ) e -15 & -1.56 @xmath26 0.10 + hd 290770 & -24.08 @xmath26 0.35 & 12.98 @xmath26 1.30 & - & -37.06 @xmath26 1.35 & 1.57 ( @xmath26 0.06 ) e -14 & -1.02 @xmath26 0.09 + bf ori & -0.02 @xmath26 0.46 & 14.70 @xmath26 1.47 & - & -14.72 @xmath26 1.54 & 5.35 ( @xmath26 0.56 ) e -15 & -1.36 @xmath26 0.10 + hd 37357 & 4.76 @xmath26 0.41 & 14.64 @xmath26 1.46 & - & -9.88 @xmath26 1.52 & 6.32 ( @xmath26 0.97 ) e -15 & -1.63 @xmath26 0.11 + hd 290764 & -2.38 @xmath26 0.43 & 13.31 @xmath26 1.33 & - & -15.69 @xmath26 1.40 & 5.03 ( @xmath26 0.45 ) e -15 & -1.46 @xmath26 0.10 + hd 37411 & -1.07 @xmath26 0.47 & 15.55 @xmath26 1.56 & - & -16.62 @xmath26 1.63 & 5.29 ( @xmath26 0.52 ) e -15 & -1.67 @xmath26 0.10 + v599 ori & 1.69 @xmath26 0.57 & 13.35 @xmath26 1.34 & - & -11.66 @xmath26 1.46 & 6.48 ( @xmath26 0.81 ) e -15 & -1.28 @xmath26 0.10 + v350 ori & 3.09 @xmath26 0.33 & 15.66 @xmath26 1.57 & - & -12.57 @xmath26 1.60 & 2.52 ( @xmath26 0.32 ) e -15 & -1.69 @xmath26 0.10 + hd 250550 & -48.83 @xmath26 0.39 & 10.03 @xmath26 1.00 & - & -58.86 @xmath26 1.07 & 2.01 ( @xmath26 0.04 ) e -14 & -0.22 @xmath26 0.09 + v791 mon & -88.08 @xmath26 0.25 & 7.94 @xmath26 0.79 & - & -96.02 @xmath26 0.83 & 4.22 ( @xmath26 0.04 ) e -14 & -0.26 @xmath26 0.09 + pds 124 & -13.83 @xmath26 1.37 & 14.13 @xmath26 1.41 & - & -27.96 @xmath26 1.97 & 2.02 ( @xmath26 0.14 ) e -15 & -1.30 @xmath26 0.09 + lkha 339 & -5.39 @xmath26 1.37 & 12.98 @xmath26 1.30 & - & -18.37 @xmath26 1.89 & 4.24 ( @xmath26 0.44 ) e -15 & -1.33 @xmath26 0.10 + vy mon & -17.81 @xmath26 1.76 & 8.45 @xmath26 0.84 & - & -26.26 @xmath26 1.95 & 6.94 ( @xmath26 0.52 ) e -14 & -0.38 @xmath26 0.09 + r mon & -114.51 @xmath26 1.76 & 9.35 @xmath26 0.93 & - & -123.86 @xmath26 1.99 & 4.22 ( @xmath26 0.07 ) e -14 & -0.07 @xmath26 0.09 + v590 mon & -60.10 @xmath26 0.52 & 9.63 @xmath26 0.96 & - & -69.73 @xmath26 1.09 & 3.46 ( @xmath26 0.05 ) e -15 & -0.49 @xmath26 0.09 + pds 24 & -25.47 @xmath26 1.79 & 12.98 @xmath26 1.30 & - & -38.45 @xmath26 2.21 & 1.17 ( @xmath26 0.07 ) e -15 & -1.00 @xmath26 0.09 + pds 130 & -31.17 @xmath26 0.70 & 11.41 @xmath26 1.14 & - & -42.58 @xmath26 1.34 & 2.75 ( @xmath26 0.09 ) e -15 & -0.58 @xmath26 0.09 + pds 229n & 7.41 @xmath26 0.82 & 9.63 @xmath26 0.96 & - & -2.22 @xmath26 1.26 & 1.72 ( @xmath26 0.97 ) e -16 & -1.99 @xmath26 0.26 + gu cma & -14.86 @xmath26 0.48 & 4.75 @xmath26 0.47 & - & -19.61 @xmath26 0.67 & 1.61 ( @xmath26 0.06 ) e -13 & 0.15 @xmath26 0.09 + ht cma & -20.94 @xmath26 0.35 & 11.87 @xmath26 1.19 & - & -32.81 @xmath26 1.24 & 1.75 ( @xmath26 0.07 ) e -15 & -0.84 @xmath26 0.09 + z cma & -63.55 @xmath26 0.99 & 9.97 @xmath26 1.00 & - & -73.52 @xmath26 1.41 & 7.72 ( @xmath26 0.15 ) e -13 & 1.43 @xmath26 0.09 + hu cma & -52.00 @xmath26 0.40 & 9.09 @xmath26 0.91 & - & -61.09 @xmath26 0.99 & 6.64 ( @xmath26 0.11 ) e -15 & -0.50 @xmath26 0.09 + hd 53367 & -7.62 @xmath26 0.52 & 4.02 @xmath26 0.40 & - & -11.64 @xmath26 0.66 & 2.15 ( @xmath26 0.12 ) e -13 & -0.11 @xmath26 0.09 + pds 241 & -8.36 @xmath26 0.29 & 4.21 @xmath26 0.42 & - & -12.57 @xmath26 0.51 & 4.32 ( @xmath26 0.17 ) e -15 & 0.06 @xmath26 0.09 + nx pup & -37.01 @xmath26 0.45 & 8.74 @xmath26 0.87 & - & -45.75 @xmath26 0.98 & 1.75 ( @xmath26 0.04 ) e -14 & -1.04 @xmath26 0.09 + pds 27 & -73.20 @xmath26 0.73 & 4.40 @xmath26 0.44 & - & -77.60 @xmath26 0.85 & 1.08 ( @xmath26 0.01 ) e -13 & 1.53 @xmath26 0.09 + pds 133 & -103.11 @xmath26 3.91 & 7.84 @xmath26 0.78 & - & -110.95 @xmath26 3.99 & 4.99 ( @xmath26 0.18 ) e -15 & -0.01 @xmath26 0.09 + hd 59319 & 5.86 @xmath26 6.14 & 7.13 @xmath26 0.71 & - & -1.27 @xmath26 6.18 & @xmath27 1.19 e -15 & @xmath27 -1.26 + pds 134 & -12.22 @xmath26 0.41 & 5.98 @xmath26 0.60 & - & -18.20 @xmath26 0.73 & 1.58 ( @xmath26 0.06 ) e -15 & 0.20 @xmath26 0.09 + hd 68695 & 0.48 @xmath26 0.48 & 16.51 @xmath26 1.65 & - & -16.03 @xmath26 1.72 & 4.18 ( @xmath26 0.45 ) e -15 & -1.81 @xmath26 0.10 + hd 72106 & 8.78 @xmath26 0.63 & 14.60 @xmath26 1.46 & - & -5.82 @xmath26 1.59 & 4.77 ( @xmath26 1.30 ) e -15 & -1.69 @xmath26 0.15 + tyc 8581 - 2002 - 1 & 11.05 @xmath26 0.42 & 13.61 @xmath26 1.36 & - & -2.56 @xmath26 1.42 & 3.50 ( @xmath26 1.95 ) e -16 & -2.05 @xmath26 0.26 + pds 33 & -3.10 @xmath26 0.48 & 15.83 @xmath26 1.58 & - & -18.93 @xmath26 1.65 & 7.90 ( @xmath26 0.69 ) e -16 & -1.67 @xmath26 0.09 + hd 76534 & -11.00 @xmath26 0.34 & 5.84 @xmath26 0.58 & - & -16.84 @xmath26 0.67 & 3.63 ( @xmath26 0.15 ) e -14 & -0.44 @xmath26 0.09 + pds 281 & 4.30 @xmath26 0.49 & 5.42 @xmath26 0.54 & - & -1.12 @xmath26 0.73 & 3.81 ( @xmath26 2.48 ) e -15 & -0.98 @xmath26 0.30 + pds 286 & -26.84 @xmath26 0.42 & 3.93 @xmath26 0.39 & - & -30.77 @xmath26 0.57 & 2.75 ( @xmath26 0.05 ) e -13 & 0.37 @xmath26 0.09 + pds 297 & 7.42 @xmath26 0.41 & 11.36 @xmath26 1.14 & - & -3.94 @xmath26 1.21 & 2.80 ( @xmath26 0.86 ) e -16 & -1.73 @xmath26 0.16 + hd 85567 & -50.25 @xmath26 0.47 & 6.76 @xmath26 0.68 & - & -57.01 @xmath26 0.83 & 1.09 ( @xmath26 0.02 ) e -13 & 0.45 @xmath26 0.09 + hd 87403 & 6.79 @xmath26 0.63 & 9.76 @xmath26 0.98 & - & -2.97 @xmath26 1.17 & 1.27 ( @xmath26 0.50 ) e -15 & -0.89 @xmath26 0.19 + pds 37 & -119.89 @xmath26 0.47 & 3.87 @xmath26 0.39 & - & -123.76 @xmath26 0.61 & 2.53 ( @xmath26 0.01 ) e -13 & 2.17 @xmath26 0.09 + hs 305298 & 0.01 @xmath26 0.41 & 3.25 @xmath26 0.32 & - & -3.24 @xmath26 0.52 & 9.93 ( @xmath26 1.59 ) e -16 & -0.45 @xmath26 0.11 + hd 94509 & -16.84 @xmath26 0.53 & 6.33 @xmath26 0.63 & - & -23.17 @xmath26 0.82 & 1.09 ( @xmath26 0.04 ) e -14 & 0.82 @xmath26 0.09 + lrrrrcc + name & & & & & & log ( ) + & & & & & ( wm@xmath24@xmath25 ) & [ ] + hd 95881 & -12.53 @xmath26 0.39 & 9.35 @xmath26 0.94 & - & -21.88 @xmath26 1.02 & 2.60 ( @xmath26 0.12 ) e -14 & 0.13 @xmath26 0.09 + hd 96042 & 3.20 @xmath26 0.80 & 3.95 @xmath26 0.39 & - & -0.75 @xmath26 0.89 & @xmath27 1.27 e -15 & @xmath27 -0.90 + hd 97048 & -24.43 @xmath26 0.34 & 13.54 @xmath26 1.35 & - & -37.97 @xmath26 1.39 & 8.06 ( @xmath26 0.30 ) e -14 & -1.13 @xmath26 0.09 + hd 98922 & -14.64 @xmath26 0.44 & 10.04 @xmath26 1.00 & - & -24.68 @xmath26 1.09 & 1.16 ( @xmath26 0.05 ) e -13 & -0.36 @xmath26 0.09 + hd 100453 & 5.27 @xmath26 0.77 & 10.23 @xmath26 1.02 & - & -4.96 @xmath26 1.28 & 9.50 ( @xmath26 2.45 ) e -15 & -2.35 @xmath26 0.14 + hd 100546 & -23.96 @xmath26 0.42 & 15.50 @xmath26 1.55 & - & -39.46 @xmath26 1.61 & 1.76 ( @xmath26 0.07 ) e -13 & -1.29 @xmath26 0.09 + hd 101412 & -0.15 @xmath26 0.69 & 15.28 @xmath26 1.53 & - & -15.43 @xmath26 1.68 & 8.38 ( @xmath26 0.91 ) e -15 & -1.62 @xmath26 0.10 + pds 344 & -22.23 @xmath26 0.45 & 7.80 @xmath26 0.78 & - & -30.03 @xmath26 0.90 & 7.69 ( @xmath26 0.23 ) e -16 & -0.74 @xmath26 0.09 + hd 104237 & -13.74 @xmath26 0.98 & 13.75 @xmath26 1.38 & - & -27.49 @xmath26 1.69 & 1.61 ( @xmath26 0.10 ) e -13 & -1.18 @xmath26 0.09 + v1028 cen & -101.61 @xmath26 0.52 & 7.01 @xmath26 0.70 & - & -108.62 @xmath26 0.87 & 2.22 ( @xmath26 0.02 ) e -14 & 0.37 @xmath26 0.09 + pds 361s & -3.97 @xmath26 0.45 & 5.35 @xmath26 0.53 & - & -9.32 @xmath26 0.70 & 8.10 ( @xmath26 0.60 ) e -16 & -0.31 @xmath26 0.09 + hd 114981 & -7.60 @xmath26 0.46 & 5.64 @xmath26 0.56 & - & -13.24 @xmath26 0.72 & 3.37 ( @xmath26 0.18 ) e -14 & -0.06 @xmath26 0.09 + pds 364 & -78.38 @xmath26 0.74 & 9.63 @xmath26 0.96 & - & -88.01 @xmath26 1.21 & 4.40 ( @xmath26 0.06 ) e -15 & -0.39 @xmath26 0.09 + pds 69 & -69.42 @xmath26 1.30 & 7.00 @xmath26 0.70 & - & -76.42 @xmath26 1.48 & 8.40 ( @xmath26 0.16 ) e -14 & 0.02 @xmath26 0.09 + dg cir & -47.98 @xmath26 0.83 & 13.10 @xmath26 1.31 & - & -61.08 @xmath26 1.55 & 6.27 ( @xmath26 0.16 ) e -15 & -1.00 @xmath26 0.09 + hd 132947 & 10.38 @xmath26 0.33 & 11.96 @xmath26 1.20 & - & -1.58 @xmath26 1.24 & 9.04 ( @xmath26 7.12 ) e -16 & -2.04 @xmath26 0.35 + hd 135344b & -4.60 @xmath26 0.75 & 5.66 @xmath26 0.57 & - & -10.26 @xmath26 0.94 & 1.35 ( @xmath26 0.12 ) e -14 & -2.08 @xmath26 0.10 + hd 139614 & 0.48 @xmath26 0.61 & 13.26 @xmath26 1.33 & - & -12.78 @xmath26 1.46 & 1.33 ( @xmath26 0.15 ) e -14 & -2.09 @xmath26 0.10 + pds 144s & -16.12 @xmath26 0.68 & 13.07 @xmath26 1.31 & - & -29.19 @xmath26 1.48 & 9.49 ( @xmath26 0.48 ) e -16 & -1.53 @xmath26 0.09 + hd 141569 & 5.13 @xmath26 0.57 & 15.55 @xmath26 1.56 & - & -10.42 @xmath26 1.66 & 3.36 ( @xmath26 0.54 ) e -14 & -1.88 @xmath26 0.11 + hd 141926 & -43.48 @xmath26 0.43 & 3.40 @xmath26 0.34 & - & -46.88 @xmath26 0.55 & 3.00 ( @xmath26 0.04 ) e -13 & 1.17 @xmath26 0.09 + hd 142666 & 5.14 @xmath26 0.68 & 11.70 @xmath26 1.17 & - & -6.56 @xmath26 1.35 & 9.17 ( @xmath26 1.89 ) e -15 & -2.22 @xmath26 0.12 + hd 142527 & -6.95 @xmath26 0.98 & 6.18 @xmath26 0.62 & - & -13.13 @xmath26 1.16 & 1.91 ( @xmath26 0.17 ) e -14 & -1.93 @xmath26 0.10 + hd 144432 & -0.98 @xmath26 0.68 & 11.65 @xmath26 1.17 & - & -12.63 @xmath26 1.35 & 1.76 ( @xmath26 0.19 ) e -14 & -1.85 @xmath26 0.10 + hd 144668 & -8.41 @xmath26 0.45 & 14.12 @xmath26 1.41 & - & -22.53 @xmath26 1.48 & 1.40 ( @xmath26 0.09 ) e -13 & -0.95 @xmath26 0.09 + hd 145718 & 8.34 @xmath26 0.52 & 14.53 @xmath26 1.45 & - & -6.19 @xmath26 1.54 & 6.86 ( @xmath26 1.71 ) e -15 & -2.34 @xmath26 0.14 + pds 415n & 3.10 @xmath26 1.30 & 5.04 @xmath26 0.50 & - & -1.94 @xmath26 1.39 & 2.45 ( @xmath26 1.76 ) e -16 & -3.53 @xmath26 0.32 + hd 150193 & -5.59 @xmath26 0.57 & 16.07 @xmath26 1.61 & - & -21.66 @xmath26 1.71 & 6.21 ( @xmath26 0.49 ) e -14 & -1.55 @xmath26 0.09 + ak sco & -0.82 @xmath26 0.95 & 5.06 @xmath26 0.51 & - & -5.88 @xmath26 1.08 & 4.81 ( @xmath26 0.88 ) e -15 & -2.60 @xmath26 0.12 + pds 431 & 1.27 @xmath26 0.50 & 10.49 @xmath26 1.05 & - & -9.22 @xmath26 1.16 & 4.41 ( @xmath26 0.56 ) e -16 & -0.94 @xmath26 0.10 + kk oph & -25.15 @xmath26 0.31 & 16.36 @xmath26 1.64 & - & -41.51 @xmath26 1.67 & 1.28 ( @xmath26 0.05 ) e -14 & -1.50 @xmath26 0.09 + hd 163296 & -3.63 @xmath26 0.36 & 16.02 @xmath26 1.60 & - & -19.65 @xmath26 1.64 & 7.82 ( @xmath26 0.65 ) e -14 & -1.60 @xmath26 0.09 + mwc 297 & -590.00 @xmath26 0.90 & 4.52 @xmath26 0.45 & - & -594.52 @xmath26 1.01 & 4.68 ( @xmath26 0.01 ) e -11 & 1.63 @xmath26 0.09 + in figure [ fig : lacc_vs_lline ] the relationship between the accretion luminosity , derived for 62 stars using the uv - excess , is shown plotted against the luminosity of six different emission lines . the selected lines are he@xmath28 , h@xmath5 , o@xmath29 , he@xmath30 , pa@xmath31 , and br@xmath4 as they cover a large wavelength range and a few different elemental species . correlations for other lines are available in the on - line version of this paper . the line luminosity accretion luminosity relationship can be approximated by a power law , and has been in numerous works on ctts and haebes @xcite , where : @xmath32 where @xmath33 and @xmath34 are constants describing the intercept and slope respectively . such a relationship is obtained across all lines for the herbig ae / be stars ; the slopes and intercepts are presented in table [ tab : lacc_lum ] . the best - fits were calculated using the idl _ mpfitexy _ package , which factors in errors in both the @xmath35 and @xmath36 axis . figure [ fig : lacc_vs_lline_append ] , in the appendix , displays the correlations for all lines observed ; table [ tab : lacc_lum ] provides details of the correlations . where possible the findings of @xcite on a set of ctts are shown in these figures for comparison . the same relationship between the line and accretion luminosities observed for ctts , and reported previously in haebes , extends to other lines as well . indeed , this trend is essentially observed for _ all _ lines . the strong correlation between line luminosity and accretion luminosity is the reason why emission lines are a useful tool for easily determining accretion luminosities ( and in turn determining the mass accretion rate ) . a comparison with previous determinations of the slope and intercept derived in the literature for herbig ae / be stars is made here . specifically the h@xmath5 , [ o]@xmath6 , and br@xmath4 relationships from @xcite are assessed . it is found that the relationships obtained here agree with previous work ; they lie within the errorbars of each other . the main difference here is that the errors , particularly on the intercept , are reduced here . this is presumably due to the increased sample size and simultaneity of the observations of the uv excess and lines . the relationship for br@xmath4 in this paper agrees well with @xcite . both of these relationships do not agree with the breakdown in br@xmath4 relationship implied by @xcite this can be attributed to the ma models used in the latter being associated with a relatively low mass haebe star , producing lower accretion luminosities than expected ; their model was based upon ux ori from @xcite , while this work focuses on multiple models from @xcite and . therefore , we can say that there remains a - relationship for haebes . the data on ctts from @xcite for the lines in common with this study are also included for comparison , along with their best fit in figure [ fig : lacc_vs_lline_append ] . overall , the plots appear to show a qualitative agreement in trends between the haebes and the ctts , where is increasing proportionally to , generally a 1:1 correlation . the placement of the ctts and haebes in the plots often agree with each where they transition between each other , suggesting similarity between the two groups . an exception is the ca ii triplet where there is an offset , likely due to blending with the paschen series . the best - fit relationship to the haebes is generally steeper than it is for the extrapolated ctts best - fit , suggesting a possible difference in the origin of the line luminosity . however , this is not true in all cases and the average of all relationships are in agreement with each other . the slope of the vs. relationship , the @xmath34 value , is found to have an average of @xmath37 in the haebes , while for the ctts it is @xmath38 . the values of the best - fit relationships for the ctts and the haebes are provided in table [ tab : lacc_lum ] . [ cols="^,^,<,^,^,^,^,^,^ " , ] in table [ tab : yes_no ] all 32 lines are presented along with details of the number of emission detections . the other columns of the table are split into different categories based upon emission detection and uv - excess detection . the total accretion rate detection is defined as all of the haebes for which an uv - excess was clearly detected in 62 stars match this criteria . this includes the 7 stars in which an accretion rate could not be determined within the context of ma ; it would require an unrealistically high filling factor of shocked material to reproduce their uv excess . the measured @xmath39 of these 7 objects is likely associated with accretion due to the overall properties of the stars , although the exact mechanism is not known ( an alternative could be boundary layer accretion for these hot objects ( see * ? ? ? * ; * ? ? ? the two categories of emission and accretion can be divided into two outcomes for each case , those with and those without . this gives a total of four possible categories for a star to be in e.g. a star may have a particular line in emission , but no @xmath39 was detected in the star . table [ tab : yes_no ] gives the number of stars split into each category , for each line , and is presented visually in figure [ fig : percentage ] . what the four different categories represent are now detailed : 1 . the first category contains objects for which both emission lines and a uv excess are detected , represented by the green segments in the diagram . this category measures if the line is a good one to one tracer of accretion i.e. excess is present and so is the emission line . examples of good tracers are h@xmath5 , br@xmath4 , and pa@xmath31 as their emission detections fall predominately in this category . 2 . the second category indicates how difficult the line is to detect i.e. how insensitive it is . this category is shown in red and contains those stars for which a clear uv excess is detected and for which a mass accretion rate could be derived , but in whose spectra no emission line is detected . an example of this is the oxygen 7773 line , for which the majority of @xmath39 detections have no corresponding emission line detection . the third category ( shown in yellow in the figure ) denotes the objects for which an emission line is detected , but no uv excess was detected . assuming that the emission line is sensitive to accretion , and more readily detectable than an excess , the lines in this category allow the vs. relationship to be used to infer an accretion luminosity . 4 . the fourth and final category , shown in orange , contains those objects with no obvious uv excess and no emission in the line of interest . these objects generally have h@xmath5 emission , but no sign of other emission lines . if they are actively accreting , the mass accretion rate is well below our detection limits ( the lowest log ( ) detected in was only -7.78 ) . the findings in the table and figure highlight that the already established accretion tracing relationships are good ones ; h@xmath5 , br@xmath4 , and pa@xmath31 are all predominately in the green . there are also lines which appear to be poor tracers of accretion . in particular the brackett series appears to get worse as an accretion tracer towards the higher order transitions . such a decline is likely because of the decreasing strength in emission towards the shorter wavelengths ; which is supported by the green / red ratio changing between successive orders ( with a slight deviation around br 12 - 4 ) . the balmer series demonstrates a similarity to this with the two lowest level transitions , h@xmath5 and h@xmath31 , being sensitive to accretion while the remainder are less sensitive . the explanation for this is likely to be the same as for the brackett series , however , another possible explanation for this could be the uv - excess veiling the lines . veiling of the lines occurs when their intrinsic absorption is filled in by excess emission , which decreases the observed absorption in them and gives then an closer to 0 . however , the brightest of haebes makes this scenario unlikely . the paschen series is an exception to the other two hydrogen series ; it maintains a constant level of sensitivity throughout the series . this is due to the location of the lines ; the visible region . the snr here is much higher than the uv and allows us to make a greater number of detections . while in the nir , due to telluric line correction and ir - excess correction , there is greater margin of error for detection than the visible . as for the ca ii line it is likely that the paschen blending results in more detections for the 8662 line than the other two , as it is blended with the stronger paschen lines . as for lines which lean more towards being poor tracers of accretion , like the o i 7773 and the he i 5876 lines . a possible explanation is that is measured as a sum over the entire line profile . this means that complex profiles , with multiple absorption and emission components , may be missing details of additional contributions to the lines . in extreme cases stars may be found to have no emission detection despite the clear presence of an inverse p - cygni profile , where the absorption is strong enough to make the measurement uncertain . therefore , when considering accretion in haebes analysis of line profiles should be used in combination with line strengths and uv - excesses ( where possible ) , to reach a consensus on detection and origin . such an example is presented by @xcite where their analysis of the lines profiles of the he i 10830 line demonstrates strong evidence of ma occurring in haes , and perhaps from hbes too . despite the convincing line profiles the detection rate of the line being in overall emission is low , making it a poor tracer in that regard . however , if detected in emission it can still be used to infer . despite information regarding line profiles being omitted in this work the - relationships hold for the majority of stars and can be used as a diagnostic to infer accretion . even the lines which have fewer detections correlate well with accretion and can be used if detected . one particularly interesting group of stars are those that show lines in emission , but have no detectable uv excess , @xmath39 . this opens up the possibility that the emission lines could be tracing accretion where the accretion rate is below the detection limits of the balmer excess method . this is now investigated here for the case of h@xmath5 . the line luminosity of h@xmath5 is converted into , using the relationship determined previously and tabulated in table [ tab : lacc_lum ] . for this test is calculated fo the stars in which only an upper limit could be placed on the balmer excess ( 25 stars , see ) . the derived is then converted into using the stellar parameters determined in . this , in turn , is used to predict the resulting @xmath39 values . the predicted excess can be calculated via the @xmath39 vs. curves seen in figure 9 of . the predicted @xmath39 agrees with the upper limits measured in for 21/25 of the objects . this agreement within previous limits for the majority of these stars suggests that the lines in this work are suitable as accretion tracers for inferring . for the remaining 4 objects a @xmath39 is predicted that is higher than the upper limits previously measured ; the values lie just 1@xmath40 away from the previous limits . considering that these fours results are only 1@xmath40 away this is not significant to detract from the previous paragraph ; the accretion diagnostic properties of the lines still holds in general . overall , we can conclude that the `` yes in emission , no in excess '' criteria serves as a positive indicator that a line can be used as an accretion diagnostic line i.e. the line luminosity can be reliably used to infer the accretion rate . this study provides the largest spectroscopic investigation into accretion rates of haebes to date . in addition , the spectral wavelength range covered in these stars is much larger than any other haebe investigation , which allows an investigation from the uv up to the nir at 2.5 . the combination of a large sample , 91 objects , and huge wavelength coverage allows for the most robust statistical investigation into the emission lines in haebes to be performed to date . line luminosities were obtained for 32 different lines . the line luminosities are observed to be correlated with the accretion luminosities by implication they are therefore correlated to the mass accretion rates . we focussed on expanding our understanding of relationships between and in haebes , and how these relate to similar relationships observed in ctts . the large size of the sample and large spectral coverage resulted in a tenfold increase in accretion line diagnostics for haebe stars compared to what was known in the literature . the following key points are found : * relationships are obtained between and for 32 different emission lines . in all cases a best fit is made to these lines , with the average correlation being @xmath41 . the fact that both uv excess and emission lines were measured simultaneously make these the most robust set hitherto published . * we find that all lines can be used as an accretion tracer . additionally , it was shown by extrapolation , for the case of h@xmath5 , that an emission line could be used to infer low mass accretion rates which can not normally be measured by an uv excess . this is particularly applicable for the hotter objects where their large intrinsic uv output makes the detection of weak uv excess emission difficult . * the relationship @xmath42 obtained for the haebes in this work agrees with the relationship observed in ctts by @xcite , where @xmath43 ; the relationship is steeper for the haebes . on an individual line basis , some variations are seen in the relationship between the two . notably the ca and he lines , lines with complex line profiles that behave differently for both types of sources and warrants further investigation . * an assessment of the sensitivity of each line as an accretion tracer has been performed though a comparison of respective emission line and uv - excess detection rates . we confirm that the well - known accretion tracers for ctts , such as h@xmath31 , h@xmath5 , pa@xmath31 and br@xmath4 , are also tracers of accretion in haebe stars . * the sheer number of objects and emission lines analysed here provide robust relationships between and for an exceptional wavelength coverage . in particular , known accretion tracing lines have been further verified and new observational windows have been opened up for measuring in haebes e.g. h@xmath31 in the blue @xmath34-band region , pa@xmath31 in the @xmath44-band region , and br(13 - 4 ) in the @xmath45-band region . in the visible the paschen series shows consistent sensitivity between lines . these new diagnostics will prove valuable in future observations . * additionally , some lines can be considered poor tracers of accretion due to a low rate incidence of detection of emission lines e.g. [ o]@xmath6 and he i 5876 . however , when these lines are detected in emission they are nearly always associated with a uv - excess i.e. associated with accretion , and the relationship holds . since this aspect is present for all lines and relationships tested here , it is likely there is a deeper physical connection . for example , in the case of [ o]@xmath6 it is suspected that photospheric uv photo - dissociates oh in upper layers of the disk , giving rise to this emission @xcite . * the high number of correlations is a topic that has been discussed recently by @xcite , where the authors found that the relationships observed between and appear to be a consequence of them being directly related to the star s own stellar luminosity ( see also * ? ? ? * ) . therefore , the line luminosities may not physically arise due to accretion onto the star , but they can still serve as a diagnostic for obtaining accretion rates . in order to gain knowledge on the physical origin of the emission lines , and their association to accretion , the line profiles themselves must be considered . work on line profiles so far has demonstrated notable differences between ctts and haebes so far @xcite . connecting these changes with both the relationships determined here , the underlying mathematical relationships @xcite , and the large number of lines obtained with x - shooter is the next challenge ; this will will be presented in a future paper by the authors . jrf gratefully acknowledges a studentship from the science and technology facilities council of the uk . jdi gratefully acknowledges support from the discsim project , grant agreement 341137 , funded by the european research council under erc-2013-adg . this work has made use of nasa s astrophysics data system ; it has also made use of the simbad database , operated at cds , strasbourg , france . the authors would like to extend their thanks to the anonymous referee who helped strengthen the manuscript . 50 natexlab#1#1 b. , van den ancker m. e. , dullemond c. p. , 2005 , , 436 , 209 j. m. et al . , 2014 , , 561 , a2 c. , 1989 , , 27 , 351 c. , basri g. , bouvier j. , 1988 , , 330 , 350 p. f. c. , djie h. r. e. t. a. , 2006 , , 456 , 1045 t. , catala c. , 1995 , , 301 , 155 n. , gullbring e. , 1998 , , 509 , 802 n. , muzerolle j. , briceo c. , hernndez j. , hartmann l. , saucedo j. l. , gordon k. d. , 2004 , , 128 , 1294 f. , kurucz r. l. , 2004 , new grids of atlas9 model atmospheres , arxiv astrophysics e - prints , arxiv : astro - ph/0405087 p. w. , johns - krull c. m. , 2014 , , 797 , 112 p. w. , johns - krull c. m. , 2015 , , 810 , 5 r. m. et al . , 2003 , 2mass all sky catalog of point sources . nasa / ipac infrared science archive s. e. , 2008 , , 136 , 521 b. , brittain s. , 2011 , , 141 , 46 c. et al . , 2001 , , 365 , 110 , c. et al . , 2002 , , 384 , 1038 j. r. , oudmaijer r. d. , mendiguta i. , ilee j. d. , van den ancker m. e. , 2015 , , 453 , 976 u. , mundt r. , 1984 , , 55 , 109 , jr . l. m. , 1978 , , 224 , 535 p. , lamb f. k. , 1979 , , 232 , 259 e. , calvet n. , muzerolle j. , hartmann l. , 2000 , , 544 , 927 e. , hartmann l. , briceno c. , calvet n. , 1998 , , 492 , 323 f. , persson s. e. , 1992 , , 82 , 247 f. , persson s. e. , 1992 , , 82 , 285 g. h. , 1960 , , 4 , 337 g. j. , hillenbrand l. a. , 2008 , , 681 , 594 j. d. , fairlamb j. , oudmaijer r. d. , mendiguta i. , van den ancker m. e. , kraus s. , wheelwright h. e. , 2014 , , 445 , 3723 l. et al . , 2013 , , 767 , 112 w. et al . , 2015 , , 576 , a78 a. , 1991 , , 370 , l39 r. l. , 1993 , synthe spectrum synthesis programs and line data . cambridge , ma : smithsonian astrophysical observatory g. , waters l. b. f. m. , bouwman j. , van den ancker m. e. , waelkens c. , malfait k. , 2001 , , 365 , 476 i. , calvet n. , montesinos b. , mora a. , muzerolle j. , eiroa c. , oudmaijer r. d. , mern b. , 2011 , , 535 , a99 i. et al . , 2013 , , 776 , 44 i. , oudmaijer r. d. , rigliaco e. , fairlamb j. r. , calvet n. , muzerolle j. , cunningham n. , lumsden s. l. , 2015 , , 452 , 2837 j. c. et al . , 2011 , , 525 , a149 u. , sordo r. , castelli f. , zwitter t. , 2005 , , 442 , 1127 j. , hartmann l. , calvet n. , 1998 , , 116 , 455 j. , dalessio p. , calvet n. , hartmann l. , 2004 , , 617 , 406 e. , natta a. , testi l. , randich s. , alcal j. m. , covino e. , stelzer b. , 2012 , , 548 , a56 f. , najita j. , ostriker e. , wilkin f. , ruden s. , lizano s. , 1994 , , 429 , 781 m. f. et al . , 2006 , , 131 , 1163 a. et al . , 2015 , , 576 , a77 p. s. , de winter d. , perez m. r. , 1994 , , 104 , 315 y. , shibata k. , 1985 , , 37 , 515 m. e. , bouwman j. , wesselius p. r. , waters l. b. f. m. , dougherty s. m. , van dishoeck e. f. , 2000 , , 357 , 325 j. et al . , 2011 , , 536 , a105 s. l. a. , corradi w. j. b. , alencar s. h. p. , mendes l. t. s. , torres c. a. o. , quast g. r. , guimares m. m. , da silva l. , 2003 , , 126 , 2971 j. s. , drew j. e. , harries t. j. , oudmaijer r. d. , 2002 , , 337 , 356 j. s. , drew j. e. , harries t. j. , oudmaijer r. d. , unruh y. , 2005 , , 359 , 1049 the sheer number of lines analysed here warrants the use of an appendix for easy reading . in here , figure [ fig : lacc_vs_lline_append ] displays the accretion luminosity versus line luminosity relationships for all 32 lines analysed ( an extension of figure [ fig : lacc_vs_lline ] ) . the best - fits to the data are present in the figure . where possible the data on ctts from @xcite is also included for comparison .

it has recently been demonstrated that the-ray emission spectrum of the egret - identified , central galactic source 2eg j1746 - 2852 can be well fitted by positing that these photons are generated by the decay of s produced in p - p scattering at or near an energizing shock . such scattering also produces charged pions which decay leptonically . the ratio of-rays to neutrinos generated by the central galactic source may be accurately determined and a well - defined and potentially - measurable high energy neutrino flux at earth is unavoidable . an opportunity , therefore , to detect neutrino oscillations over an unprecedented scale is offered by this source . in this paper we assess the prospects for such an observation with the generation of neutrino erenkov telescopes now in the planning stage . we determine that the next generation of detectors may well find an oscillation signature in the galactic center ( gc ) signal . 3gcm^-3 3cm^-3 # 1#1 #1#2#3#1 , a&a , * # 2 * , # 3 . # 1#2#3#1 , a&a _ ( letters ) _ , * # 2 * , # 3 . # 1#2#3#1 , a&as , # 2 , # 3 # 1#2#3#1 , aj , # 2 , # 3 # 1#2#3#1 , _ apj _ , * # 2 * , # 3 . # 1#2#3#1 , _ apj ( letters ) _ , * # 2 * , # 3 . # 1#2#3#1 , apjs , # 2 , # 3 # 1#2#3#1 , ara&a , # 2 , # 3 # 1#2#3#1 , baas , # 2 , # 3 # 1#2#3#1 , icarus , # 2 , # 3 # 1#2#3#1 , _ mnras _ , * # 2 * , # 3 . # 1#2#3#1 , _ nature _ , * # 2 * , # 3 . # 1#2#3#1 , pasj , # 2 , # 3 # 1#2#3#1 , pasp , # 2 , # 3 # 1#2#3#1 , qjras , # 2 , # 3 # 1#2#3#1 , science , # 2 , # 3 # 1#2#3#1 , soviet astr . , # 2 , # 3 # 1#2#3#1 , soviet astr . lett . , # 2 , # 3 # 1#2#3#4#1 , # 2 , # 3 , # 4

astro-ph9911292

the dominant radio emitting structures at the galactic center ( gc ) are the supernova remnant ( snr)-like shell sagittarius ( sgr ) a east , a three - armed spiral of ionized gas dubbed sgr a west , and , embedded at the center of sgr a west , the galactic dynamical nucleus , sgr a * , thought to be a massive ( @xmath2 ) black hole ( @xcite ; @xcite ; @xcite ) . sgr a east has a major axis length of about 10.5 pc and its center is located 2.5 pc from sgr a * in projection , and probably behind the latter ( @xcite ) . ( 1998 ) have recently determined the intrinsic size of sgr a * to be less than @xmath3 m at @xmath4 mm . also located at the gc is the egret - identified @xmath0-ray source 2eg j1746 - 2852 ( @xcite ) . it has been shown that the high energy ( @xmath5 ) @xmath0-ray emission spectrum of this source is very likely due to the decay of @xmath1 s ( @xcite ; @xcite ) . these pions are produced by p - p collisions which might plausibly take place at either of two shock regions : 1 ) the shock at sgr a * due to gas accretion from ambient winds , or 2 ) the shock produced by the expansion of the snr - like nonthermal shell of sgr a east into the ambient gas of the interstellar medium . thus , _ a priori _ , either or both sgr a * and sgr a east might be the source of the @xmath0-rays which constitute 2eg j1746 - 285 . it has recently been shown , however , that the identification of sgr a * with 2eg j1746 - 28 is disfavored because charged leptons produced in @xmath6 decays would emit too much synchrotron flux in sgr a * s intense magnetic field at ghz frequencies to be consistent with the well - studied radio spectrum of this object ( @xcite ; @xcite ) . on the other hand , given the physical conditions in sgr a east , the putative charged leptons generated there have a distribution that mimics a power - law with index @xmath7 . the synchrotron flux radiated by these charges is consistent with the radio spectrum of sgr a east observed with the vla . in fact , such relativistic electrons and positrons would also radiate by bremsstrahlung and undergo inverse compton scattering in such a way as to self - consistently explain the entire broadband emission spectrum of sgr a east , ranging from ghz frequencies all the way up to the tev energies observed by whipple ( @xcite ) . for the purposes of this paper , then , we shall take it that the egret source 2eg j1746 - 285 is identical with sgr a east ( @xcite ) . we note in passing that the maximum energy attained by the shocked protons at sgr a east , given the energy loss rate via collision in the shock , is @xmath8 ( @xcite ) . regardless of the ultimate identity of the egret source 2eg j1746 - 285 , given that the process producing the high energy emission is pionic , there should be an associated neutrino flux from the gc ( @xcite ) . these neutrinos are due both to direct pion decay ( @xmath9 ) and to the decay of muons to electrons and positrons ( @xmath10 ) , where we take @xmath11 to mean @xmath11 and @xmath12 here ( as we shall often do in the remainder of this paper ) . _ prima facie _ , then , we expect the flavor composition of the neutrino ` beam ' generated at the gc to be essentially @xmath13 @xmath14-like and @xmath15 @xmath16-like by nave channel counting ( c.f . atmospheric neutrinos in the gev energy range ) . note that there is a @xmath17 background produced at the source due to non - pionic processes like charmed hadron decay . this background is , however , small ; see later . of course , in the absence of neutrino flavor oscillations , one would expect to observe g.c . neutrinos at earth with the same flavor composition as that generated at the source . we do not distinguish between @xmath11 and @xmath18 , because present and planned terrestrial detectors do / will not distinguish between the two . there is one interesting proviso to this statement , however : a @xmath19 flux at @xmath20 can be detected by resonant @xmath21 boson production via @xmath22 with the electrons in the detector medium . the resonance energy , however , is just above that attained by neutrinos generated in the processes described above at the gc ( @xcite ; @xcite ; @xcite ) . given our detailed knowledge of the basic physical processes producing the gc @xmath0-rays , we are able to determine an expression for the total neutrino emission at the source , @xmath23 , in terms of the @xmath0-ray emission there , @xmath24 , the numerical power of the proton spectrum at the source , @xmath25 ( such as would result from shock acceleration at either sgr a east or sgr a * ) , and @xmath26 ( @xcite ) . the quantity @xmath25 has been empirically - determined to lie between 2.1 and 2.4 ( @xcite ; @xcite ) , using a procedure to fit the egret spectrum of 2eg j1746 - 2852 with a detailed calculation of the particle cascade using an extensive compilation of pion - multiplicity cross - sections . in the energy range between the @xmath27-resonance ( @xmath28 gev ) and the isr ( intersecting storage rings ) range ( @xmath29 gev ) , simple scaling ( @xcite ) does not adequately take into account the strong dependence of the cross section on the rapidity at lower energy , and the pion distribution is not adequately described by a power - law mimicking the injected relativistic proton distribution between @xmath30 and @xmath31 gev . instead , the distribution steepens in this region and is curved , which is consistent with the suggested spectral shape measured by egret . above about @xmath32 gev , however , the pion distribution settles into the ` asymptotic ' form suggested by scaling , where the power - law index is a direct reflection of the underlying relativistic protons . thus , an egret spectrum with an effective spectral index of @xmath33 below @xmath32 gev is produced by a pion distribution whose power - law index lies in the range @xmath34 above this energy . in other words , a relatively steep and curved @xmath0-ray spectrum below @xmath32 gev is consistent with a flatter neutrino spectrum at tev - energies . the relative normalization between the @xmath0-ray and neutrino distributions is effected at @xmath32 gev where the pions take on a power - law form . we take the neutrino spectrum at earth to be , in general , given by : @xmath35 normalizing to the observed @xmath0-ray flux at earth at @xmath32 gev , one arrives at the following values for the total neutrino flux here ( @xcite ) : @xmath36 for @xmath37 , and @xmath38 for @xmath39 , where we have taken the absolute upper bound to the energy spectrum of g.c . neutrinos to be given by the highest energy ( @xmath40 ) of the shocked protons . ( kinematical calculations show that neutrinos created by the decay of charged pions produced in scattering of a ` beam ' proton off a stationary ` target ' proton can attain energies very close to the ` beam ' proton . ) note that in the above we make the very reasonable assumption that high energy @xmath0 s and @xmath11 s travel to earth equally unimpeded by the ambient matter they encounter ( which has a column number density of barely @xmath41 ) . two factors improve the odds for the detection of the gc neutrino flux above the atmospheric neutrino background . these are 1 ) the effectively point - source nature of the gc , and 2 ) a gc neutrino spectrum that is significantly flatter than that of atmospheric neutrinos ( which goes as @xmath42 ) . if we preliminarily adopt an angular resolution of @xmath43 for the proposed large scale detectors ( @xmath44 effective detector area ) , the condition for the detection of the gc neutrino flux is @xmath45 , where @xmath46 is the solid angle corresponding to the angular resolution of the experiment and @xmath47 is the flux of atmospheric neutrinos per unit solid angle . this condition is fulfilled above a few tev , and the expected event rate from this preliminary analysis is @xmath48 for @xmath39 to @xmath49 for @xmath37 ( @xcite ) . note that a fuller analysis of event rates ( presented later ) must also consider the problems posed by the atmospheric _ muon _ background and earth neutrino opacity . we see therefore that preliminary calculations reveal that there is a well - determined and potentially observable neutrino flux at the earth from the galactic center . we now briefly list the motivations behind this work before going on to consider whether any sort of neutrino oscillation signature might be detectable in the gc signal . the main motivations behind the present work are : 1 . sgr a east is arguably the most thoroughly understood extra - solar astrophysical source of very high energy neutrinos identified to date . it is thus of fundamental importance for the embryonic science of neutrino astronomy . 2 . it is important for general scientific reasons to explore the neutrino oscillation phenomenon in a wide variety of regimes . because of the high energy scales and the very long baselines involved , astrophysical sources such as sgr a east provide a novel regime not investigated in previous and current solar , atmospheric , reactor and accelerator neutrino detection experiments . previous works to have considered propagation effect signatures in galactic and extra - galactic high energy neutrino signals include @xcite . 3 . given that solar and atmospheric neutrino observations have essentially established the existence of neutrino oscillations , it is important to incorporate this propagation effect when examining possible sources for study through neutrino astronomy . neutrino signals from astrophysical sources are an important complement to electromagnetic signals from same , and they will serve to improve our understanding of the dynamics of important astrophysical objects such as supernova remnants , gamma ray bursters and active galactic nucleii . for purposes of calculational expediency we take the neutrino source sgr a east to have a linear dimension of @xmath50 . this distance is relevant because we need to know how the neutrino oscillation lengths compare with the size of the emitting object to determine whether the neutrino source is flavor coherent . if the former are small compared to the latter , then , because neutrinos are emitted from all points within the source , the oscillations will be averaged out . alternatively , if the latter are large compared to the former , then no averaging due to the finite size of the source will be needed and the source is essentially flavor coherent for neutrinos of a given energy . note that two types of averaging generally need to be done : over distance , and over energy . thus far we have only considered distance averaging due to the finite size of the @xmath11 source . one also has to take into account distance ( and energy ) averaging due to the detector . for sgr a east the source distance scales involved are at least six orders of magnitude larger than those for the detector ( 1 a.u . @xmath51 m ) . detector - based distance averaging , then , will not impact on calculations concerning sgr a east . we do not address the issue of energy averaging due to the finite energy resolution of the detector in great detail in this paper . the distance between source and detector is about @xmath52 we consider only 2-flavor oscillation modes @xmath53 for simplicity and definiteness . suppose a beam of flavor @xmath25 is produced at @xmath54 . then at a point @xmath55 distant from the source the oscillation probability is @xmath56 whereas the `` survival probability '' is obviously @xmath57 the parameter @xmath58 is the ` mixing angle ' which determines the amplitude of the oscillations . the value @xmath59 , which leads to the largest possible amplitude , is termed ` maximal mixing ' . the parameter @xmath60 is the ` oscillation length ' and is given by @xmath61 in natural units @xmath62 . note that the oscillation length increases linearly with energy . this is important because the high energy scale under consideration @xmath63 stretches the oscillation length . the parameter @xmath64 is the squared - mass difference between the two mass eigenstate neutrinos . totally averaged oscillations see the second @xmath65 factor in equation ( 5 ) set equal to @xmath66 , leading to @xmath67 this , to reiterate , can be due to either distance or energy spread or both . given the poor statistics of the proposed neutrino telescopes , only modes with large mixing angles , @xmath58 , can be probed ( unless the msw phenomenon takes place see later ) . the atmospheric neutrino anomaly ( for @xmath68 s ) seen by superkamiokande and other experiments clearly indicates large angle vacuum oscillations , however ( @xcite ; @xcite ) further , the solar neutrino anomaly ( for @xmath69 s ) can be solved by large angle oscillations ( or by small angle oscillations through the msw effect ) ( @xcite ) . in summary , then , the atmospheric anomaly _ definitely _ requires a large mixing angle solution , while the solar problem _ can _ be solved by large angle oscillations . we now briefly review the various possible solutions to the atmospheric and solar neutrino problems , and then apply the various scenarios to the gc neutrino flux . superkamiokande detects a @xmath70 deficit of @xmath14-like atmospheric neutrinos coming up through the earth ( @xcite ) . they see no deficit of either upward- or downward - going @xmath16-like neutrinos . the lower energy downward - going @xmath14-like events are deficient , whereas their high - energy counterparts are not . these data can be explained by close - to - maximal @xmath71 oscillations with @xmath72 and @xmath73 or @xmath74 ( sterile ) . these two alternatives both require parameters in the range : @xmath75 ( to be strict , the @xmath76 ranges are a little different for the two possibilities because of the ` matter effect ' in the earth , but this will be irrelevant for us ( @xcite ; @xcite ) . ) superkamiokande currently favors oscillations to @xmath77 over oscillations to a sterile neutrino at the 2@xmath78 level ( though this is a very preliminary result ) ( @xcite ) . the solar neutrino problem can be solved by @xmath79 oscillations , where @xmath80 are all allowed , with one important proviso : if the los alamos lsnd experiment is correct , then @xmath81 oscillations , with parameters that can not solve the solar neutrino problem , have already been detected ( @xcite ) . so , if the still - controversial lsnd result is correct , then @xmath82 is ruled out . the miniboone and boone experiments at fermilab should eventually settle this issue ( @xcite ) . the precise oscillation parameter space required to account for the solar data depends on which of the solar neutrino experiments are held to be correct . the two parameter ranges defined below , however , are broadly consistent with all solar data ; 1 . @xmath79 with a _ small _ mixing angle ( sma ) @xmath83 is possible through the msw effect . if this pertains , then the oscillation amplitude will be far too small to affect sgr a east neutrinos . 2 . @xmath79 with a very large mixing angle ( lma ) @xmath84 is an interesting possibility for the range @xmath85 the immediate vicinity of @xmath86 defines ` just - so ' oscillations where the oscillation length for solar neutrinos is of order 1 a.u . @xmath87for larger @xmath88 values completely averaged oscillations , with a flux suppression factor of @xmath89 , result . maximal mixing explains almost all of the data with averaged oscillations ( excepting the homestake result ( @xcite ) , and the controversial superk spectral anomaly ) . values of @xmath90 are ruled out by the non - observation of @xmath19 disappearance from reactors ( chooz , palo verde experiments ( @xcite ; @xcite ) ) . in summary , for gc neutrinos the following are well motivated scenarios that are composed of 2-flavor subsystems : 1 . large angle @xmath91 + large angle @xmath92 ( scenario 1 ) . 2 . large angle @xmath91 + large angle @xmath93 ( scenario 2 ) . 3 . large angle @xmath94 + large angle @xmath95 ( scenario 3 ) . 4 . small angle @xmath79 + large angle @xmath96 ( scenario 4 ) . small angle @xmath79 + large angle @xmath97 ( scenario 5 ) . we are now in a position to perform a number of simple calculations for neutrino oscillations between the gc and the earth motivated by the above list of 2-flavor possibilities . note here that bimaximal ( @xcite ; @xcite ; @xcite ; @xcite ; @xcite ; @xcite ) and trimaximal ( @xcite ; @xcite ; @xcite ) mixing scenarios , which are intrinsically 3-flavor , will not be considered in this paper . using the atmospheric problem parameters , we see that the @xmath98 oscillation length is given by : @xmath99 therefore , the oscillation length is orders of magnitude less than the size of sgr a east for the entire neutrino spectrum ( which only reaches up to @xmath100 ) . this means that the oscillations will be distance averaged , and hence at earth we expect a @xmath101 mixture of @xmath68 and @xmath102 , where @xmath73 or @xmath74 depending on which solution to the atmospheric problem turns out to be the correct one . using the solar problem parameters one determines the @xmath79 oscillation length to be @xmath103 the reference @xmath88 is in the ` just - so ' range . the oscillation length of @xmath79 oscillations in this range , therefore , becomes larger than sgr a east for @xmath104 or so . this means that the more energetic component of the @xmath69 beam from the source is flavor - coherent . in principle , such coherence would evidence itself by an energy dependent spectral distortion ; the @xmath69 flux at a particular energy ( @xmath105 ) would depend on the part of the neutrino oscillation wave ( for that particular energy ) encountered by the earth at its distance from sgr a east , i.e. the neutrino flux at a particular energy might be anything from maximally suppressed to unsuppressed depending exactly on @xmath88 and the source - observation point distance . certainly , ranging over the expected energy spectrum ( and therefore ranging over @xmath106 ) , we should see the flux vary ( over and above the variation given by the spectral shape ) between maximally suppressed and unsuppressed . imagining , then , that we had both a neutrino detector able to determine the energy of an incoming neutrino to arbitrary accuracy , and that we had a very long time to accumulate statistics , we should be able to find an experimental signature of the flavor - coherence in the form of this spectral distortion ( and thus determine whether @xmath88 were in the ` just - so ' energy range able to lead to such coherence , and , if it were , exactly what value it takes ) . pragmatically , given the small statistics that will accrue from the gc source and the limited energy resolution expected to be achieved by any of the proposed neutrino telescopes , one expects no observational consequence of the flavor coherence . this is because the energy dependence of the flux suppression washes out with the inevitably large size of the energy bins particular neutrino events are accumulated into . the beam , therefore , is indistinguishable from one in the distance - averaged oscillation regime . note also that the @xmath79 oscillation length would become of the order of the gc - earth distance for @xmath107 for @xmath108 . the @xmath69 flux would , then , rise from being suppressed below @xmath109 to unsuppressed above @xmath109 if the @xmath69 attained this energy . of course given that the maximum energy of the shocked protons does not surpass @xmath110 this phenomenon does not occur for the gc source . at the opposite extreme of the acceptable parameter space , i.e. , @xmath111 , the oscillation length is @xmath112 this is back in the totally distance - averaged oscillation regime . in conclusion , for the entire allowable @xmath88 regime we pragmatically expect a situation similar to the muon - type neutrino case : totally averaged oscillations , i.e. , a @xmath101 mixture of @xmath69 and @xmath113 for maximal mixing . a brief calculation is sufficient to show that for the gc , matter effects ( refractive indices for neutrinos ) do not impinge significantly on the oscillation probabilities . the quantities that have to be compared are @xmath114 and @xmath115 , where @xmath116 is the fermi constant , and @xmath117 is the electron minus positron number density for the medium . right at the source we expect @xmath118 because of the equal production of electrons and positrons there . concerning propagation of the neutrinos from source to detector , we have that the interstellar medium consists of approximately 1 h atom per @xmath119 so that @xmath120 , converting the number density to natural units . this number works out to be about @xmath121 . the smallest @xmath76 we consider is @xmath122 and for the highest attainable neutrino energy of @xmath123 , we get @xmath124 , so we are 11 orders of magnitude away from having important matter effects due to the interstellar medium . we do not consider matter effects due to dense intervening objects between the gc and earth , since their covering fraction for sgr a east is trivially negligible . we consider now the observational consequences of scenarios 1 to 5 listed above in terms of the neutrino flux at earth ( we remind the reader that all @xmath68 mixing scenarios are lma ) . 1 . scenario 1 ( lma @xmath91 and @xmath92 ) : @xmath70 reduction of both @xmath69 and @xmath68 flux , and no @xmath77 appearance above background . 2 . scenario 2 ( lma @xmath91 and @xmath93 ) : @xmath70 reduction of @xmath69 flux , and equal @xmath68 and @xmath77 fluxes . scenario 3 ( lma @xmath94 and @xmath95 ) : equal @xmath69 and @xmath77 fluxes , and @xmath70 reduction of @xmath68 flux . 4 . scenario 4 ( sma @xmath79 and @xmath96 ) : unreduced @xmath69 flux , @xmath70 reduced @xmath68 flux , no @xmath77 appearance above background . scenario 5 ( sma @xmath79 and @xmath97 ) : unreduced @xmath69 flux , and equal @xmath68 and @xmath77 fluxes . the scenarios above imply the following ratios ( and ratios of ratios ) of neutrino flavor fluxes : [ cols="^,^,^,^,^,^,^",options="header " , ] the superscript ` @xmath125 ' denotes the flux ratios observed by a neutrino telescope , while ` @xmath126 ' denotes the ratio expected from the no - oscillation theoretical calculation . deviation away from the value predicted for the no oscillation case in any of the ratios defined above , beyond experimental uncertainty , would constitute a _ prima facie _ case for whatever neutrino oscillation scenario most closely predicts the experimental fluxes . deviation in the third last ratio would constitute the strongest evidence for oscillation because errors due to uncertainties in the determination of the total theoretical neutrino flux tend to cancel in taking the ratio of the theoretical @xmath69 and @xmath68 flavor ratios given that @xmath69 s and @xmath68 s are produced by the same mechanism at the source . on the other hand , there is considerable uncertainty concerning the @xmath77 background ( see later ) so that estimates of @xmath127 may not be particularly meaningful . for this reason , we do not list @xmath128 . as displayed above , though , in the absence of oscillations we still expect the @xmath17 flux to be considerably smaller than the other flavor fluxes in the absence of oscillations to this flavor type . further , deviation from 1 in the first two ratios defined could only provide strong evidence of oscillations if the uncertainties in the power of the neutrino spectrum , @xmath25 , and @xmath129 were both significantly reduced by future @xmath0-ray observations using instruments with better energy resolution and coverage . the glast mission may be the first to provide the necessary improvements over the next few years ( @xcite ) . note also that , unfortunately , none of the five scenarios considered here realistically exhibits the energy - dependent flux suppression ( within an appropriate energy range ) that would be the most telling signature of neutrino oscillations . further , even assuming that we possess a detector with near perfect neutrino identification capability , so that we can determine the ratios defined above and hence distinguish between the five broad scenarios , we still can not further pin down @xmath130 or @xmath88 than has already been achieved with the terrestrial solar , atmospheric , reactor and accelerator neutrino experiments . parameter space _ would _ have an experimental signature . ] the allowable mixing angle parameter space might only be constrained in the sense that the above ratios distinguish between a large and a small @xmath83 . in the next section we examine the prospects for determining the neutrino flux of each flavor at earth . in this work we consider only the erenkov neutrino telescopes now in planning and construction stages as observation platforms . other proposed astronomical neutrino detection methods tend to require neutrino energies in excess of that possessed by sgr a east neutrinos ( see appendix c of ( @xcite ) for a brief review ) . detection and identification , are limited to energies in excess of @xmath131 by the atmospheric background ( @xcite ) . radio detection of neutrinos will probably require energies at or in excess of the gc neutrino energy upper limit ( i.e. , @xmath132 ) because of signal - to - noise problems ( @xcite ; @xcite ; @xcite ) . ] the erenkov detectors are planned to operate through the instrumentation of very large volumes ( @xmath133 is thought to be optimal for astronomical neutrino detection ( @xcite ) ) of some transparent medium ( in practice water or ice ) with photomultiplier tubes ( pmts ) . these tubes will detect the erenkov light generated by superluminal charged leptons traversing the detector volume . the erenkov light is generated at a characteristic angle ( for the medium ) away from the direction of travel of the charged particle . note that only muons and extremely energetic tauons have path lengths through water and ice significant on the scales of the pmt separation of these detectors ( tens of meters ) . electrons are arrested very quickly ( within a meter or so ) , even at the highest energies we are considering : @xmath134 $ ] . lower energy tauons ( produced by @xmath77 primaries with @xmath135 decay within meters . we remark in passing that , at considerably higher energies still ( i.e. , @xmath136 ) , the landau - pomeranchuk - migdal ( lpm ) effect starts to cause a measurable reduction to the pair production and bremsstrahlung cross sections of the electron . this increases the radiation lengths of @xmath137 ( @xcite ) . erenkov neutrino telescopes of course encounter background generated by atmospheric ( i.e. , muons generated directly by cosmic ray interactions in the atmosphere ) as well as the atmospheric _ neutrino _ background . in fact , this background overwhelms the genuine neutrino signal due to any conceivable astronomical object at sea level and hence neutrino telescopes must be shielded somehow . this requirement ( as well as the requirement for sufficient clarity of the medium ) is what has driven all proposed sites for working neutrino telescopes deep ( few kilometers ) into the antarctic icecap or underwater . even at these depths , however , the atmospheric muon background is far from negligible . the simplest way to ensure exclusive selection of genuine neutrino - generated leptons against this background is to have a erenkov detector register only ` upcoming ' leptons ; those that arrive at a specified angle to the vertical somewhat below the horizontal . it is then almost assured that such charged leptons have been generated by neutrinos which traverse some large fraction of the earth s diameter and then subsequently undergo charged current ( cc ) interactions with the water / ice at the detector or the rock / water / ice fairly close to it . exclusive selection for upcoming leptons can be achieved through a combination of geometry ( simply situating all pmts so they face downwards ) and triggering ( which can discern up - going from down - going signal on the basis of fairly simple timing considerations ( @xcite ; @xcite ) ) . note that the highest energy muons might traverse a distance of ten kilometers water equivalent and still retain sufficient energy to produce a detectable erenkov signal ( @xcite ) . the effective volume , therefore , for @xmath68 detection is substantially larger than the ` instrumented ' volume . of course , such a simple triggering system ( and geometry ) means that one misses out completely on the signal from down - going neutrinos ( which also generate down - going leptons in cc interactions ) . this may seem like a reasonable compromise ( the effective detector area is greater from below , after all , because of the greater amount of material below the detector than above ) until one considers the fact that for high energy neutrinos earth ` shadowing ' or opacity becomes a significant effect . a neutrino is shadowed when its interaction length becomes smaller than the distance it must travel through the earth to reach a detector . at energies of @xmath138 earth opacity affects all neutrinos except those that reach a detector from an almost horizontal direction . it would seem , therefore , that with the scheme described above reject all down - going leptons and the unavoidable issue of earth opacity , ultra high energy neutrino telescopy is impossible , except for a tiny window on neutrinos which come from a practically horizontal direction . in order for ultra high energy neutrino astronomy to have a future , detector triggering must be designed that does something smarter than simply rejecting all down - going leptons ; it must be able to select something of the genuine down - going signal , at least at higher energies . for the moment , let us assume that some more discerning trigger can be instantiated . the question now is , does the gc neutrino source have a large enough flux at high energies to be seen against the muon background ( for reasonable values of telescope angular resolution ) , even in principle ? to make this calculation , we first adopt values for the high energy ( vertical ) fluxes of atmospheric muons at sea level given elsewhere ( @xcite ) and then convolve these fluxes with values for the probabilities of these muons to reach the particular water and ice depths of the proposed detectors ( @xcite ) then , to determine event rates in a detector due to high energy atmospheric muon background , we take these fluxes over a sensible range of telescope angular resolutions ( from @xmath139 to @xmath140 , say ) . subsequently , we compare these ( angular resolution dependent ) rates with the genuine , neutrino - generated event rate found by convolving the gc neutrino spectrum with reasonable estimates for the neutrino detection probability . the probability that a high energy muon - type neutrino is detected in a @xmath141-scale neutrino telescope depends on two factors , viz ; approximately inversely on the interaction length of the neutrino ( @xmath142 ) at that energy ( which , in turn , depends on the charged current cross - section ) and approximately directly on the radiation length of the muon ( @xmath143 ) produced in the interaction . ( we assume here that the linear dimension of the detector is small on the scale of @xmath143 . ) we can make a rough estimate of the effect of these factors by writing down a detection probability multiplier which goes as some power of the energy : @xmath144 halzen gives @xmath145 and @xmath146 for tev to pev energies , with @xmath147 measured in tev units ( @xcite ) . note here paranthetically that all proposed neutrino detectors will have an overburden depth less than the radiation length of muons with energies in the energy range we are considering ( see later ) . this means that the above method actually over - estimates the neutrino detection probability for downward - going neutrinos because the volume of material _ above _ a detector available for the neutrino to interact within ( and subsequently produce a muon which might then travel on to the detector volume ) is substantially less than that for upward going neutrinos . employing the above ( generous ) parameterization of the neutrino detection probability allows us to determine that the gc signal does not , in fact , emerge clearly from the background until well into the upper end of the neutrino energy spectrum ( even for a detector resolution as low as @xmath140 and the flattest empirically allowable neutrino spectrum , @xmath37 ) . it seems that even granted a detector able to trigger on neutrino - generated , down - going leptons , we can not , on the basis of our preliminary calculations , confidently conclude that we might see the gc source in such a way . we therefore restrict ourselves to consideration of the sgr a east signal to be found in upcoming leptons . an immediate consequence of this self - imposed restriction is that we are unable to reach many conclusions about the usefulness of the ( successor to the ) amanda neutrino telescope in regards to observing the gc . this is somewhat unfortunate because the amanda experiment , of all neutrino telescope projects , is probably the best currently placed to realize the desired @xmath141 status and thus evolve to ` icecube ' ( @xcite ) . this is because amanda / icecube s south polar location means that the gc is always overhead fom the detector . a detector specific monte carlo calculation will probably be needed to settle whether our particular source can be seen by icecube . we can predict , however , that icecube may well be able to detect the distinctive ` double bang ' signature of gc @xmath77 interactions above the background ; see later . we must , therefore , look to the northern hemisphere for @xmath68 observing platforms about which we can make more confident predictions regarding the gc source . there are currently four neutrino telescope projects under development there . the lake baikal project is a mature experiment , having run on and off since 1993 . this collaboration has achieved an effective , energy - dependent detector area of @xmath148 and has demonstrated the viability of large - scale , water - based erenkov technology . the collaboration is planning for a neutrino telescope of @xmath149 effective area . this will not be a large enough platform for the relatively high energy ( and low flux ) neutrino signal generated at sgr a east ( @xcite ) . three other projects , all based in the deep mediterranean , are currently in the design and prototype stage . they are antares , nestor and nemo . none of these projects is guaranteed of the funds to reach @xmath141 status , though this is the stated goal of all three collaborations . needless to say , these deep sea environment projects call for great ingenuity and considerable technical innovation . the antares collaboration has completed preliminary reconnaissance of its chosen site at a depth of 2400 m below the sea near toulon . they are also well into design of electronics and mechanics for the detector . the collaboration s current mid - term goal is to have 13 ` strings ' , with @xmath150 attached pmts , in place by 2003 . such a configuration would have an effective area of @xmath151 ( @xcite ) . the nestor collaboration plans for a deployment at @xmath152 depth off the south west grecian coast . this collaboration is at a similar stage of advancement to the antares group , having completed reconnaissance of their chosen site and preliminary field testing of crucial components . nestor also aims for a @xmath153 effective area detector in the near future ( @xcite ) . lastly , the nemo project is least advanced being in the early r&d stage . this collaboration is investigating the suitability of a site off the southern italian coast . they have conducted monte carlo studies of their proposed detector layout ( @xcite ) . in conclusion , one does not expect to see a @xmath141 neutrino telescope in the northern hemisphere within a decade , but within two decades the chances for such would seem to be quite good . neutral current ( nc ) interactions do not identify the incoming neutrino flavor and basically constitute a background to the more useful charged current interactions . energy determination for nc events is poor because of the missing final state neutrino . angular determination is also poor because the single hadronic shower produced is almost point - like on the scale of a typical detector s pmt spacing . nc interactions are only about one third as common as cc interactions . the best prospects for observing _ any _ neutrino flux from the gc source are offered by muon type neutrinos ; @xmath154 s and @xmath155 s ( we remind the reader that neutrino telescopes can not distinguish a neutrino from an anti - neutrino of the same flavor type ) . charged current interactions of a muon type neutrino in or near the detector volume result in a nearly point - like hadronic shower and a high energy muon ( @xmath156 ) which , we reiterate , might travel up to ten kilometers and still possess enough energy to be detected . certainly in the above-@xmath157 energy scales of relevance to this paper , muons will be ` uncontained ' in the sense that they can not be expected to be both generated and arrested within the @xmath141 detector volumes . such long tracks mean , of course , very good determination of the muon s direction of travel . on the other hand , the fact that the muons are necessarily uncontained leads to uncertainty in energy determination . we now discuss , in the context of the observation of neutrino - generated muons , the general issues of angular and energy determination in more detail . an accurate determination of the energy possessed by a muon neutrino ( which produces a muon observed by a detector ) is limited by three factors : uncertainty in the fraction of the neutrino s total energy imparted to the muon , ignorance of the energy loss by the muon outside the instrumented volume and , finally , the intrinsic energy resolution of the detector apparatus itself ( @xcite ) . regarding the first factor , it can be shown that the average energy imparted to the muon is half that of the neutrino in the cc interaction @xmath158 and three quarters in the interaction @xmath159 ( @xcite ) . a determination of an individual muon s energy , then , might only give us a minimum energy for the neutrino primary but this problem is not a limiting factor if a significant number of events can be accumulated and we take a statistical view . note that when the muon is uncontained and , hence , an accurate determination can not be achieved by measuring the length of the entire muon track , a rougher muon energy determination can be achieved for @xmath160 by measuring @xmath161 because at such energies , where energy loss is dominated by radiative processes , @xmath162 . it may also eventually be possible to glean some neutrino energy information from the hadronic shower resulting from the first cc interaction if this happens to be within the detector volume ( keeping in mind the difficulty posed by the relatively small size of such showers on the scale of a next - generation detector s pmt spacing ) . that we are dealing with uncontained muon tracks means that one can only arrive at a minimum original muon energy . that we can make some sort of energy determination from @xmath161 , though , means that we have a much better idea of the original energy of a _ totally _ uncontained muon than would be imparted by just assigning it a minimum energy enough to take it across the detector given all the above factors , the antares collaboration has judged on the basis of monte carlo simulations of their detector array that they can gauge a muon neutrino s energy to within a factor of three for @xmath163 ( @xcite ) . again three factors limit the determination of the primary neutrino s direction of travel . these are the uncertainty in the angle between the incoming @xmath154 and the resulting @xmath14 , the deviation of the @xmath14 away from its original direction of travel due to multiple scattering and , lastly , the detector s intrinsic angular resolution as determined by uncertainties in its exact geometry , etc . ( @xcite ) . of course , the severity of the first two problems decreases with increasing energy , but the relative severity of the two likely changes with energy . for example , the antares collaboration has determined from mc simulations that below @xmath164 total angular resolution is limited by detector effects whereas above @xmath165 it is limited by the unavoidable angular distribution of the neutrino interactions . they claim an angular resolution of @xmath166 is achievable ( @xcite ) . with such a resolution the gc signal is above atmospheric neutrino background for energies greater than a few @xmath167 . the amanda project ( which will hopefully evolve into icecube ) has to contend with the short scattering length of the erenkov light in ice , @xmath168 , as compared to seawater at @xmath169 . despite this , icecube will achieve an angular resolution less than one degree and perhaps as low as @xmath170 ( @xcite ) . we note parenthetically that such a resolution will mean that many southern sky sources ( i.e. sources of downgoing neutrinos ) _ will _ be able to be seen by icecube above atmospheric muon background . at @xmath171 the interaction lengths of all neutrino flavors become less than the earth s diameter . this means that , in particular , @xmath68 s are unlikely to reach a detector from a nadir angle of @xmath172 ( @xcite ) . ( the same is true for @xmath69 s but _ not _ @xmath77 s ; see later . ) the attenuation of the @xmath68 interaction length continues until at @xmath138 it is less than a very small fraction of the earth s diameter , so that this flavor is attenuated over all nadir angles , even those approaching the horizontal . at such high energies , then , the earth is said to be ` opaque ' to @xmath68 s ( and @xmath69 s ) ( @xcite ) . we must take both this effect and our self - imposed requirement that the gc neutrino source be below the horizon from the observation point ( in order to avoid the atmospheric muon background problem ) into account to generate a more realistic estimate of the event rate due to the gc source . let us assume the best case scenario for @xmath68 fluxes detector angular resolution of @xmath166 , and a neutrino spectrum that goes as @xmath173 to make a determination of the expected event rate in a hypothetical , @xmath141 detector located on the proposed antares site . note that with this revised angular resolution , the gc neutrino flux is above atmospheric neutrino background at an energy around an order of magnitude lower than previously : a few @xmath157 . also note that the gc is below the horizon about two thirds of the time from this latitude ( @xcite ) and , therefore , invisible at least one third of the time ( even if low enough detector resolution were achieved to unequivically avoid the atmospheric muon background problem , antares is being designed with downward pointing pmts ) . adopting the neutrino penetration coefficients calculated by naumov and perrone ( @xcite ) , we determine that the expected annual event rate from @xmath68 s generated at the gc is @xmath174 for the no - oscillation case and @xmath175 if oscillations do occur . for @xmath176 , but retaining an angular resolution of @xmath166 , we expect @xmath177 events without oscillation and @xmath178 with . clearly , then , we approach the lower end of statistical relevance with this value for @xmath25 . ( in these calculations we have not allowed for the regeneration effect due to nc interactions that affects all neutrino flavors . we expect this effect to be small ( @xcite ) . ) we note , in passing , one unavoidable source of @xmath68 background ; cc @xmath77 interactions can mimic cc @xmath68 interactions if 1 ) the @xmath77 energy is too low to effectively separate the original cc interaction vertex and the @xmath179 decay vertex , and 2 ) the @xmath179 decays muonically ( the branching ratio for this decay is @xmath180 % ( @xcite ) ) . in contrast to the case for muon neutrinos , the prospects for identifying electrons ( @xmath181 ) in a detector generated by @xmath182 s from the direction of the gc seem remote . quite a few of the significant problems with observing the @xmath69 signal can be related back to the relatively tiny propagation length ( @xmath183 meter ) of high energy electrons ( and positrons ) in matter . perhaps most significant is that , as with nc interactions , the hadronic and electromagnetic showers initiated by a @xmath182 in a cc interaction have almost point - like dimension on the scale of the proposed detectors and , hence , provide little directional information . thus , even if we grant that an electron signal in the appropriate energy range for the gc source might be identified , we can not actually identify the origin of the primary electron neutrino . a second problem is that electron neutrino initiated cc events are very difficult to conclusively identify . in principle a smoking gun for such events is presented by the coincident presence of both a hadronic shower ( from the disturbed nucleus ) _ and _ an electromagnetic shower from the quickly braked @xmath137 . it is very difficult , however , for the proposed , next - generation erenkov technology to distinguish between the two types of showers . both showers , we repeat , are essentially point - like on the scale of the typical detector s pmt spacing and , after all , are observed only indirectly through the erenkov flash they produce . thus , nc events , which produce a point - like hadronic shower , are difficult to distinguish from cc @xmath69 initiated events and provide a significant background problem . further , even imagining that we had some reliable technology to identify the presence of a high energy electron , the cc interactions of @xmath77 s can still mimic cc @xmath69 events if 1 ) the @xmath179 energy is not high enough to ensure that the hadronic shower from the cc interaction of the primary @xmath17 and the later decay are effectively separated on the scale of the detector , and 2 ) the @xmath179 decays electronically ( with a branching ratio of @xmath184 % ( @xcite ) ) . a yet further problem is the fact that the short path of the electron in matter means that one can only register contained @xmath69 cc events , dramatically reducing the effective volume monitored by the detector in comparison with @xmath68 events . altogether one can not but conclude that the chances for detecting gc @xmath69 s , at this stage , seem remote . although the chances for observing gc @xmath77 s seem more hopeful than those for gc @xmath69 s , there will still be considerable problems with this flavor . at least two unique signatures for the @xmath77 have been identified in the literature : 1 ) the ` double bang ' and 2 ) flat angular dependence of the signal or ` pile up ' ( @xcite ; @xcite ; @xcite ; @xcite ) . these both , however , tend to become significant on the higher energy side of the gc neutrino spectrum . in more detail , the ` double bang ' signal requires that a @xmath77 undergo a cc interaction in the detector volume to produce a @xmath179 . if the energy of this @xmath179 is high enough then the hadronic shower resulting from the initial interaction of the neutrino primary and the later hadronic shower resulting from the @xmath179 decay will be resolvable on the scale of the detector . exactly where the resolvability threshold is can probably only be determined by detector - specific mc simulations . the antares group believes the signal certainly _ can not _ be resolved for @xmath185 ( @xcite ) . at @xmath186 , towards the upper limit of the gc neutrino spectrum , the two bangs should be separated by about @xmath187 and clearly resolvable . the usefulness of this signature , then , will depend on detector specifics and the question of whether a statistically significant flux can be obtained from whatever part of the @xmath77 spectrum remains able to produce a signal . one also notes that earth opacity will significantly reduce the flux of @xmath77 s sufficiently energetic to produce the double bang signal if one is looking for the signal in upcoming neutrinos . it is in searching for the double bang signature from gc @xmath77 s , then , that we can predict that amanda ( or , more precisely , icecube the @xmath141 extension of amanda ) may well find employment in regards to this source ; gc neutrinos will not be affected by earth opacity when observed by icecube . ( the genuine gc @xmath77 flux is substantially above that of the atmospheric @xmath77 s due to ` prompt ' and conventional flux over an angular resolution even as bad as @xmath188 ( @xcite ) and @xmath188 is a pessimistic prediction for the icecube s angular resolution ( @xcite ) ) . assuming a best - case scenario for @xmath77 detection , viz , the flattest allowable gc spectrum ( @xmath37 ) , double bang resolvability all the way down to @xmath165 and @xmath189 oscillations , and assuming a double bang detection probability given by @xmath190 ( @xmath191 means @xmath192 water equivalent ) , we can arrive at an ( optimistic ) annual event rate prediction for icecube . we derive the double bang detection probability by employing similar logic to that which led to the @xmath68 detection probability presented previously . the difference here is that we assume the @xmath179 decay length is small on the scale of the linear dimension of the detector ( hence the 1 in the numerator ) , whereas previously we assumed that the @xmath14 radiation length is large in comparison to this scale . we employ a parameterization of the neutrino interaction length presented in graphic form ( fig . 11 ) in ( @xcite ) . using the detection probability described above , and the best case scenarios for the gc spectrum and double bang resolvability , we determine an event rate of 1 double bang signal per year . this is at the threshold of detectability . the idea behind the second @xmath77 signature the flat angular dependence which has recently received attention from halzen and saltzberg ( 1998 ) and iyer , reno , and sarcevic ( 1999 ) is to actually make positive use of the earth opacity previously mentioned . when @xmath193 climbs beyond @xmath194 the interaction length of the @xmath77 becomes , as for the @xmath69 and @xmath68 , less than the earth radius . but whereas @xmath16 s and @xmath14 s resulting from cc interactions are stopped in the earth , @xmath179 s from cc interactions decay back to @xmath77 s before being stopped , producing a neutrino with something around one quarter the energy of the original and traveling in much the same direction . this process can occur more than once , each iteration producing a progressively lower energy @xmath77 , ensuring that whatever the energy of the primary @xmath77 , a @xmath77 signal from a point source should reach a detector on the other side of the earth . this signal will exhibit a ` pile up ' just below the energy where the @xmath77 s interaction length becomes greater than the fraction of the earth s diameter subtended by a ray from the source to the detector . in other words , for @xmath195 energies in excess of @xmath196 , as the angle between a neutrino source and the nadir is decreased from @xmath197 , a critical angle will be reached where the @xmath195 flux will begin to be attenuated . this attenuation increases to reach a maximum at @xmath172 . further , as the energy of the @xmath195 signal increases , the flux attenuation sets in at increasingly large ( i.e. , increasingly horizontal ) angles . on the other hand , the @xmath77 flux , although shifted downward in energy , should still be the same . this results in the flat angular dependence of the @xmath77 part of the signal at high energies and , given a significant @xmath77 component of the total neutrino flux , a flatter than expected angular dependence of the total neutrino flux . one way to search for a @xmath77 signal , then , is through the decay chain @xmath198 ( branching ratio @xmath199 ( @xcite ) ) . given the above considerations , if we assume that a significant part of a neutrino signal is due to @xmath77 s , we expect an enhancement of the number of @xmath14 s coming from the direction of our source , below certain energies and nadir angles , over that expected from the ` raw ' @xmath68 and @xmath77 fluxes . in order to see this enhancement , however , we require that the @xmath11 energy spectrum not be too steep . otherwise the increase of the @xmath14 flux in some particular , lower energy ` bin ' will be insignificant on the scale of the number of events that would be recorded there anyway due to the raw @xmath68 and @xmath77 fluxes . iyer , reno , and sarcevic ( 199 ) have made calculations of the ` pile up ' enhancement for neutrino spectra which go as different negative powers : @xmath200 . for @xmath201 the enhancement is a noticeable effect , but for @xmath202 and greater the spectra are too steep for the effect to be discernible . for the sgr a east neutrino flux , with a best - case spectrum which has an @xmath117 of 2.1 , we must unfortunately conclude that the above diagnostic for the presence of a significant @xmath77 component in the total neutrino signal will not be useful . in summary for the @xmath77 case , we believe that the gc can produce @xmath77 s energetic enough to produce a double bang signal , but that the spectrum is too steep to evidence @xmath77 s with pile up . a preliminary calculation reveals a double bang signal at the threshold of detectability in icecube , but a confident indication that this signal will produce a statistically significant event rate requires a detector - specific study . as has been mentioned , we expect no @xmath77 flux from pion decay from p - p scattering at the gc in the absence of oscillations and , hence , observation of a @xmath77 flux of the order of the @xmath69 or @xmath68 flux constitutes _ prima facie _ evidence for exactly such neutrino oscillations . one must be concerned , however , about sources of background to the @xmath77 oscillation signal , both genuine @xmath77 flux from sources that have not been accounted for and false @xmath77 signals in the detector . one source of @xmath77 s that we can anticipate at higher energies at the production site is the decay of charmed mesons ( principally @xmath203 ) produced in p - p scattering through @xmath179 and @xmath77 production . it should be noted that the cross - sections for @xmath204 and @xmath205 production via p - p scattering are greatly uncertain in the energy range of interest , as are the fractional likelihood of @xmath206 and the branching ratio for @xmath207 ( @xcite ; @xcite ) . in comparison , however , with pion production processes leading to @xmath195 such charmed meson production is still greatly suppressed . the flux ratio @xmath208 can still , therefore , be expected to be a small number taking this process into account , although there might be considerable deviation from @xmath209 in @xmath210 ( if we assume large statistics ) without oscillations necessarily being implied . if we grant that the gc source will not produce @xmath69 s in an observational energy range , might produce @xmath17 s in an observational range and certainly will produce observational @xmath154 s , we are left with only one useful flux ratio that is certainly measurable : @xmath211 and two that may be measurable : @xmath212 as previously discussed , deviation from one in the first ratio , by itself , would provide only weak evidence for oscillations unless the empirical values of @xmath25 and @xmath129 were further constrained ( by future @xmath0-ray observations ) . even if this were achieved , however , given the indirectness of the @xmath213 measurement , there would have to be some doubt about whether the presence of oscillations had been conclusively demonstrated . with empirical determination of the values of all three ratios , only scenario 3 emerges with a unique signature . otherwise , we can only distinguish the @xmath96 scenarios ( 1 and 4 ) from the @xmath93 scenarios ( 2 and 5 ) , without being able to conclude anything about @xmath69 mixing . certainly , however , @xmath214 potentially offers very strong evidence of oscillations if it is found to deviate substantially from zero . given that the measurement of these ratios lies at least a decade into the future it is , in fact , not unlikely that the uncertainty regarding the @xmath69 and @xmath68 oscillation modes be largely dispelled by the time of such measurement , i.e. , other experiments will determine which of scenarios 1 to 5 ( or bimaximal or trimaximal oscillations or even one of the non - oscillation scenarios see below ) actually occurs in nature . the most interesting science that might be extracted from gc neutrino observations , then , may be an empirical determination of @xmath25 and @xmath129 independent of @xmath0-ray observations . oscillations _ do _ take place and , hence , the knowledge that @xmath215 must be @xmath66 , and from an empirical determination of the @xmath154 spectrum , one can work backwards to obtain @xmath213 and , thence , @xmath25 and @xmath129 . ] note in passing that determination of the flavor composition of the gc neutrino signal could certainly provide for stringent tests of various alternative , no - oscillation explanations to the solar and atmospheric neutrino anomalies if these are not ruled out in the near future . for instance , a large @xmath68 component in the gc neutrino spectrum would imply a much larger lower limit on the ` @xmath68 lifetime ' ( we should , strictly , consider the lifetimes of the mass eigenstates composing the @xmath68 ) than is required to explain the atmospheric neutrino anomaly . see , e.g. , ( @xcite ) . on the other hand , flavor changing neutral currents ( fcnc ) , invoked as explanations of the atmospheric anomaly ( @xcite ; @xcite ; @xcite ; @xcite ) , can not affect the sgr a east signal . this is because the column density encountered by neutrinos propagating from the gc to the earth is far too small to allow this mechanism to occur . fcnc explanations of the atmospheric anomaly , then , predict a gc neutrino event rate undiminished from the nave expectation and deviation from this would tell against such explanations . there are a number of sources of background to the gc neutrino signal . logically , we can break these down into the two general classes : 1 ) ` enshrouded sources ' and 2 ) terrestrial background . by the former we refer to any sources of genuine neutrino signal from the gc which are ` hidden ' in the sense that they are not correlated with the gc @xmath0-ray spectrum . by the latter we mean the atmospheric neutrino and muon backgrounds that are endemic . these two have already been addressed . we know of two potential sources of an enshrouded neutrino signal from the gc . one neutrino production via high energy cosmic ray scattering on the ambient material in the galactic plane is virtually assured ( @xcite ; @xcite ) . the other neutrino production via annihilation of wimps accumulated in the gravitational well at the gc is a possibility ( @xcite ; @xcite ) . note that the first background source is , like the sgr a east source , due to decay of pions produced in nucleon - proton scattering . the density of ambient matter in the galaxy is greatest , in general , in the galactic plane and greatest of all at the gc so we may expect a large background neutrino flux from this direction . of course , the pionic decay also leads to the production of @xmath0 s . that we consider this neutrino source enshrouded , then , is due to the relatively large angular resolution of the proposed neutrino telescopes ; the neutrino telescopes see neutrinos from a much larger area of sky than the @xmath0-defined size of sgr a east . detailed estimates have been made of the rate of neutrino production by the interaction of cosmic rays with the interstellar medium ( @xcite ) . this neutrino flux has been shown , however , to be below the atmospheric neutrino background for much of the energy range under consideration . even given that the gc background exceeds the atmospheric one above @xmath216 , the background from @xmath166 of sky ( as relevant for antares ) is still considerably below the signal . the exact flavor composition of the neutrino flux generated by wimp annihilation is model - dependent . it is conceivable , for instance , that a large @xmath77 component might be present in this signal , if it exists at all . there is a fairly robust and model - independent upper bound to the wimp mass of @xmath217 ( @xcite ) . neutrinos generated in wimp annihilation processes will have typically between one half to one third the wimp rest mass energy ( @xcite ) . we can not , therefore , strictly rule out the possibility that the sgr a east neutrino signal is polluted with neutrinos from wimp annihilation . we do not consider this possibility in any detail , however , because , most reasonable wimp candidates have maximum masses some orders of magnitude below this . the neutralino , for instance , can not be more massive than @xmath218 if it is to be a wimp candidate ( @xcite ) . neutrinos produced in its decay , therefore , can , at worst , be just below the energy cut - off of the part of the gc neutrino signal we are examining . the gc neutrino source should produce an observable oscillation signature . the strongest evidence for such would take the form of a @xmath77 flux attaining a significant fraction of the @xmath68 flux from from this source . such a @xmath77 flux may be inferred from the double bang signature at icecube , the @xmath141 successor to the amanda telescope . detector - specific simulations are required for a confident determination of whether the double bang event rate due to the gc will be statistically significant in the event that either of scenarios 2 , 3 or 5 is correct , but preliminary calculations reveal that this event rate may be just at the threshold of detectability . such simulations are also required to determine whether icecube might see the gc @xmath68 signal against the atmospheric _ muon _ background . a deviation from the expected @xmath68 flux determined from @xmath0-ray observations of the gc is guaranteed for all neutrino oscillation scenarios identified . observation of such deviation would , however , constitute more equivocal evidence for oscillation than a strong @xmath77 signal because of uncertainties in the total expected neutrino flux calculated on the basis of @xmath0-ray observations . certainly , the value of @xmath25 , the numerical power of the power - law proton spectrum at sgr a east , would have to be further constrained before the above became a useful diagnostic ( as would @xmath129 ) . the actual @xmath68 flux should be able to be inferred from the @xmath68 event rate experienced by a future , mediterranean - based @xmath141 erenkov neutrino detector . strong confirmation of the oscillation signature will require observation of @xmath69 flux from the gc to see whether the @xmath69 to @xmath68 ratio varies significantly from @xmath66 ( though , as discussed , if the small mixing angle solution to the solar neutrino problem is correct @xmath69 s will not oscillate on their way from the gc ) . the energetics of the gc ` beam ' , however , place it below the region where next - generation techniques and detectors are currently predicted to be able to identify a @xmath69 component . such confirmation , then , must lie some decades into the future . perhaps the best science that might be extracted from the gc neutrino spectrum as observed by a future @xmath141 erenkov neutrino detector , assuming that other experiments resolve the electron- and muon - type neutrino oscillation mode questions first , is an empirical determination of @xmath25 . by such a determination , a neutrino telescope would realize the aspiration expressed in its very name , that , at base , it is a device for investigating the nature of astronomical objects , not merely the radiation they emit . r.m.c . gratefully acknowledges useful discussion and correspondence with d. fargion , f. halzen , a. oshlack , w. rhodes , and m. whiting . r.r.v . would like to thank v. barger for a useful discussion ; he is supported by the australian research council . is supported by the commonwealth of australia . is partially supported by nasa under grant nagw-2518 at the university of arizona .

motivated by the recent detection of gravitational waves from the black hole binary merger gw150914 , we study the dynamical evolution of ( stellar mass ) black holes in galactic nuclei where massive star clusters reside . with masses of and sizes of only a few parsecs , nuclear star clusters are the densest stellar systems observed in the local universe and represent a robust environment where black hole binaries can dynamically form , harden and merge . we show that due to their large escape speeds , nuclear star clusters can retain a large fraction of their merger remnants . successive mergers can then lead to significant growth and produce black hole mergers of several tens of solar masses similar to gw150914 and up to a few hundreds of solar masses , without the need of invoking extremely low metallicity environments . we use a semi - analytical approach to describe the dynamics of black holes in massive star clusters . our models give a black hole binary merger rate of from nuclear star clusters , implying up to a few tens of possible detections per year with advanced ligo . moreover , we find a local merger rate of for high mass black hole binaries similar to gw150914 ; a merger rate comparable to that of similar binaries assembled dynamically in globular clusters . finally , we show that if all black holes receive high natal kicks , , then nuclear star clusters will dominate the local merger rate of binary black holes compared to either globular clusters or isolated binary evolution . # 1 # 1

1606.04889

on september 14 , 2015 the advanced ligo interferometer ( aligo ) detected the event gw150914 , which has been interpreted as the first direct observation of gravitational waves ( gws ) from the inspiral and merger of a pair of black holes ( bhs ; * ? ? ? * ) . the event gw150914 was produced by two bhs with masses of @xmath4 and @xmath5 ( in the source frame ) , at a redshift @xmath6 assuming standard cosmology @xcite . the detection of the gravitational - wave signal of gw150914 has provided the first direct evidence that black holes with mass @xmath7 exist and that they can reside in binary systems . assuming that the source - frame binary bh merger rate is constant within the volume in which gw150914 could have been detected , and that gw150914 is representative of the underlying binary bh population , the bh - bh merger rate is inferred to be @xmath8 in the comoving frame @xcite . @xcite reviews various channels for the formation of bh binaries that can coalesce within a hubble time thus becoming potentially detectable by aligo . these include dynamical formation in dense stellar environments ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) , and isolated binary evolution ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? while most of the former literature focused on bh binaries forming in globular clusters ( gcs ) , little attention has been devoted to the formation of such binaries in nuclear star clusters ( nscs ) . yet nscs have total stellar masses that are comparable to the whole stellar mass of the gc system for the galaxy , at least in the milky way , and are the densest and most massive star clusters observed in the local universe ( e.g. , * ? ? ? * ; * ? ? ? * ) , representing therefore a natural environment where dynamical processes can efficiently lead to the formation of bh binaries . in this paper we consider the dynamical formation scenario , and explore the contribution to the bh binary merger rate from nscs . in a stellar cluster , stellar mass bhs formed from the death of massive stars , quickly segregate to the center through dynamical friction @xcite . in these high - density environments , bhs can efficiently interact with each other and dynamically form new binaries . such binaries will subsequently harden through three - body interactions @xcite . via such dynamical processes , gcs can produce a significant population of bh binaries that , after being ejected from the cluster , will be able to merge in the local universe ( e.g. , * ? ? ? over the last years our understanding of the evolution of bhs in star clusters has improved considerably thanks to numerical efforts ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) . however , the role of nscs and their contribution to the bh binary merger rate in the local universe remains quite obscure . as discussed in what follows , nscs differ from lower mass gcs in at least three important ways , each of these can significantly enhance the bh merger rate and affect the properties of the merging binaries in nscs . \(i ) _ nscs retain most of their bhs_. while natal kicks can easily eject bhs from gcs , the natal kick magnitudes are unlikely to be large enough to eject a considerable number of bhs from nscs given the large escape speed in these latter systems . whether dynamically formed bh binaries will merge , and whether the merger will occur inside the cluster also depends on the cluster escape speed . the low escape speed ( @xmath9 ) from low mass clusters ( @xmath10 ) , implies that most bh binaries are ejected early after their formation with an orbital semi - major axis which is typically too large for gw emission to become efficient and drive the merger of the binary in one hubble time . the vast majority of dynamically formed bh binaries in gcs are also kicked out before merging , but they are able to merge in the local universe ( e.g. , * ? ? ? * ) . as argued by @xcite given that nscs have escape speeds that are several times those of globulars , they can retain most of their bh binaries . moreover , as we show below , even when accounting for the recoil kick due to anisotropic emission of gw radiation a large fraction of merger products is likely to be retained inside nscs . \(ii ) _ nscs contain young stellar populations . _ the common finding emerging from spectroscopic surveys is that nscs are characterized by complex star formation histories with a mixture of morphological components and different stellar populations spanning a wide range of characteristic ages ( from 10 myr to 10 gyr ) and metallicities @xcite . this implies that unlike gcs , nscs can still form fresh bhs and bh binaries at the present time . the presence of significant additional gas not found in old globular clusters could also result in differences in the black hole mass distribution , as well as the dynamics of the underlying black hole population @xcite . \(iii ) _ nscs reside at the center of galaxies . _ therefore , unlike gcs , nscs are not isolated . in time , newly formed star clusters could migrate by dynamical friction from the galaxy into the nsc itself replenishing bhs that have been kicked out by three - body processes or by gw recoil kicks . the orbital decay of massive star clusters through dynamical friction constitutes an additional source which can repopulate the bh binary population in the nuclei of galaxies @xcite . in this paper we study the dynamical formation of bh binary mergers in nscs , with particular focus on nscs which do not host a central massive black hole ( mbh ) which we define here as bhs having a mass of @xmath11 . our cluster models are based on a semi - analytical approach which describes the formation and evolution of bh binaries in static cluster models . although necessarily approximated , these models are shown to give reasonable results when compared to recent monte carlo models of massive gcs @xcite and previous bh binary merger rate estimates from nscs @xcite . we stress that although the mbh occupation fraction in nscs is largely unconstrained observationally , it has been long recognized that some nscs do not have mbhs ( e.g. , * ? ? ? * ; * ? ? ? we note that nscs with mbhs are very different , dynamically , than nscs without . if a mbh is present the velocity dispersion keeps growing towards the mbh , which means that no binary will be hard all the way to the center . here we make use of the semi - analytical galaxy formation models presented in @xcite to predict the occupation fraction of mbhs in nscs and the nsc initial mass function , which combined with the results of our cluster models allows an estimate of the aligo detection rate and properties of bh mergers forming in nscs . our results suggest that the bh merger event rate from nscs is substantial , with several tens of events per year detectable with aligo . in addition , we propose a new dynamical pathway to the formation of high mass bh binary mergers similar to gw150914 . this merger path is exclusive to nscs and to the most massive gcs . due to their large escape speeds , such massive clusters can keep a large fraction of their bh merger remnants while also evolving rapidly enough that the holes can sink back to the central regions where they can form a new binary , which will subsequently harden and merge . we find that this process can repeat several times and produce bh mergers of several tens of solar masses and up to a few hundreds of solar masses , without the need of invoking extremely low metallicity environments . the paper is organized as follows . in sections [ 3rec ] we discuss the processes leading to the formation and merger of bh binaries in the high density cores of gcs and nscs , focusing on the processes that can lead to the full ejection of bhs . in section [ sam ] we describe our semi - analytical approach and derive the expected merger rate of bh binaries in nscs . in section [ disc ] we discuss the implications of our results including the aligo detection rate and the contribution to the bh merger rate from nscs hosting central mbhs . finally , we summarize the main results of our study in section [ summ ] . due to to asymmetries in the mass ejection or in the neutrino flux during core - collapse supernovae the black holes might receive appreciable natal kicks which could eject them from the cluster . thus , before we discuss the dynamical processes that can lead to the formation and merger of bh binaries in star clusters it is useful to consider natal kicks as a phenomenon that can fully eject bhs from a star cluster thereby aborting the dynamical formation channel for bh mergers . the histograms in figure [ fig0 ] show the distribution of escape velocities from nscs and gcs . the escape velocities of gcs are central escape velocities calculated using the photometric data from the catalog by harris @xcite and using single - mass king models with a constant mass - to - light ratio @xmath12 . the nscs escape velocities ( from the cluster half - mass radius ) were computed from the expression ( e.g. , * ? ? ? * ) : @xmath13 where @xmath14 and @xmath15 are the cluster half - light radius and mass ; the coefficient @xmath16 takes into account the dependence of the escape velocity on the concentration of the cluster ( i.e. , @xmath17 , with @xmath18 and @xmath19 the cluster tidal and core radii ) . the cluster radii and masses were taken from the sample of late - type galaxies of @xcite . for more than half of the nscs in these galaxies @xcite find that a king profile with a high concentration index , @xmath20 , provides the best fit . this concentration parameter corresponds to @xmath21 @xcite this latter is the value of @xmath22 that we adopted in evaluating eq . ( [ vesc ] ) . figure [ fig0 ] shows that escape velocities from nscs are substantially larger than those from gcs although the two distributions somewhat overlap near @xmath23 where the two type of systems have similar structural properties ( e.g. , * ? ? ? * ; * ? ? ? the distributions of natal kicks ( @xmath24 ; blue lines ) in figure [ fig0 ] were taken from figure 3 of @xcite . these authors consider two different neutron star natal kick distributions . one is the @xcite distribution , the other is the bimodal distribution for neutron star kicks proposed by @xcite which has a lower peak at @xmath25 and the higher peak at @xmath26 . we also show two modified versions of these distributions ( blue dot - dashed lines ) , which were obtained by assuming that the momentum imparted on a bh is the same as the momentum given to a neutron star taken from the two former distributions . thus the kick velocities are reduced in these latter models by the neutron star to bh mass ratio @xcite . figure [ fig0 ] shows that bhs receiving natal kicks as large as @xmath27 will escape from gcs before they can dynamically interact , which will suppress the dynamical formation of bh mergers in these systems . however , from figure [ fig0 ] we also see that bhs will be easily retained in nscs even for natal kicks as large as a few @xmath28 . hence , if bhs receive natal kicks of @xmath29 , we expect that this will greatly reduce the bh merger rate from gcs ( e.g. , * ? ? ? * ) as well as that from isolated binary evolution ( e.g. , * ? ? ? * ) virtually to zero , but it will not significantly alter the merger rate of bh binaries formed dynamically in nscs unless the birth kick velocities are @xmath30 . as we will show in section [ bkick ] these basic predictions are in agreement with the results of our cluster models ; for now we note that the obvious consequence of the comparison shown in figure [ fig0 ] is that the nsc @xmath31 gc relative contribution to the bh merger rate will depend on the poorly constrained natal kick velocity distribution . in the following we assume that at least some bhs are retained inside the cluster and consider the subsequent formation and dynamical evolution of binary bhs . after few million years from the birth of a star cluster , the most massive stars explode in supernovae or collapse directly to form bhs . if the bhs are not ejected by their natal kicks , being more massive than a typical main - sequence star , they will migrate to the cluster center via dynamical friction in a process that is generally referred to as mass segregation . in the dense environment of the cluster core bhs can efficiently form binaries which will then harden and eventually merge . a useful reference time is the two - body relaxation timescale evaluated at the half - mass radius of the star cluster @xcite : @xmath32 with @xmath33 the coulomb logarithm . on a time @xmath34 , two - body gravitational interactions of stars are important in driving the dynamical evolution of the cluster . . such systems are still in the process of forming a bh subsystem . for the majority of the systems we considered , including most nscs and ucds , the bulk of the stellar mass bhs are likely to have already experienced significant mass segregation . systems that lie to the left the solid black line have @xmath35 . this is an indicative value of cluster escape velocity below which bh binaries will be ejected from the cluster before merging . in systems that are within the green hatched region the bhs are more likely to merge while still inside their host cluster . ] while low - mass nscs have short relaxation times , for some of the most massive nscs the half - mass relaxation time can exceed the hubble time . however , even in the most massive nscs the bhs can still segregate at the center on the much shorter dynamical friction timescale @xcite . more precisely , the bhs will decay to the cluster core on the timescale ( e.g. , * ? ? ? * ) : @xmath36 with @xmath37 and @xmath38 the mass of a typical cluster star and bh respectively . after a time @xmath39 the bhs will dominate the densities inside the cluster core . ) and eq . ( [ df ] ) are strictly valid only for a singular isothermal sphere model . ] in figure [ fig2 ] we plot the half - mass radius ( or effective radius ) versus the total stellar mass for various types of compact clusters : nscs , gcs and ultra compact dwarfs ( ucds ) . the dashed line delineates the region below which the dynamical friction timescale becomes shorter than @xmath40 suggesting that the bhs in these systems will sink to the center in much less than a hubble time . virtually all systems we considered but the most massive nscs and ucds ( @xmath41 ) can evolve rapidly enough so that the bhs will sink to the center where they can participate in dynamical interactions and swap into hard binaries . the formation of such binaries and their dynamical evolution is discussed in the following . after the bhs segregate to the cluster core , bh binary formation can efficiently occur through the processes described below . during core - collapse , if the densities of bhs become sufficiently high , bh binaries can be assembled through three - body processes in which a binary is formed with the help of a third bh , which carries away the excess energy needed to bound the pair @xcite . the timescale for three - body binary formation can be written as ( e.g. , * ? ? ? * ) : @xmath42 with @xmath43 the number density of black holes near the center . the constant @xmath44 in the previous expression parametrizes the departure of the cluster from equipartition and we have used the relation @xmath45 in order to express @xmath46 in terms of the cluster stellar velocity dispersion . in addition to three - body binary formation , bh binaries can potentially form through exchange interactions involving primordial stellar binaries @xcite . exchange interactions can lead to the efficient formation of bh binaries only if the cluster contains a number of hard binaries . thus , this channel might be somewhat suppressed in nscs because of the larger velocity dispersion than in gcs a larger fraction of binaries will be soft and will be quickly ionized in nscs . however , as argued in @xcite , the reduction is not going to be by a large fraction given that binaries are typically born with roughly equal probability per logarithmic interval of semi - major axis , @xmath47 , in the range @xmath48 @xcite . if , for example , we consider a nsc with velocity dispersion @xmath49 all binaries with semi - major axis @xmath50 will be hard . if we assume a constant probability per @xmath51 for @xmath52au , then the probability of finding a binary in the range of @xmath53 au is substantial , @xmath54 percent . when a bh gets within a couple of semi - major axis lengths of a binary , the binary will be broken apart and the bh will tend to acquire a companion . the characteristic timescale on which such exchange interaction occurs is @xmath55 , where @xmath56 $ ] is the interaction cross section for periapsis distances @xmath57 and @xmath58 is the total mass of the interacting objects . if the cluster core is dominated by stellar binaries , then the timescale for a bh to be capture into a binary is @xcite : @xmath59 where @xmath60 is the typical semi - major axis of hard binaries and @xmath61 is the core binary fraction . by comparing the previous equation to eq . ( [ t3bb ] ) we see that even under quite standard conditions ( but even more so during core - collapse ) , three - body binary formation likely dominates the initial dynamical formation of bh binaries in nscs @xcite . two - body binary formation can also occur through gravitational bremsstrahlung in which two initially unbound bhs become bound after a close encounter in which energy is dissipated through gravitational wave radiation . if a bh binary is formed in this manner it mergers almost immediately , without further interactions . however , @xcite showed that for velocity dispersions @xmath62 and numbers of bhs @xmath63 expected in the most massive star clusters we study here , the rate of binary formation from gravitational bremsstrahlung is much less than that of regular three - body binary formation . therefore , for our investigation , we do not account for binary formation though gravitational bremsstrahlung , but we caution that this process could become important in the most massive nscs . in addition , we assume that after bh binaries are formed , binary - single interactions dominate over binary - binary interactions , which will be the case unless the binary fraction is very high ( @xmath64 ; * ? ? ? after bh - binaries are formed they will dominate the dynamics inside the cluster core . assuming the interaction is now between three bhs each with mass @xmath65 , the typical timescale on which a three - body interaction occurs is : @xmath66 given that three - body encounters tend to pair the most massive bhs participating in the interaction , we expect that after a time @xmath67 the most massive bhs in the cluster will become part of a hard binary . after a hard binary is formed it will tend to harden at a constant rate ( e.g. , * ? ? ? * ) @xmath68 in this last expression @xmath69 is the local density of stars and bhs , @xmath70 is the binary hardening rate and we have assumed all equal mass interlopers . if after a single interaction with a cluster member of mass @xmath38 the semi - major axis of the binary decreases from @xmath71 to @xmath72 , then a binary with components of mass @xmath73 and @xmath74 will recoil with a velocity @xmath75 , where @xmath76 , @xmath77 , @xmath78 and @xmath79 . in deriving the previous expression we have assumed that in the interaction the binding energy of the binary increases by a fraction @xmath80 @xcite . the previous expressions can be used to derive the limiting semi - major axis below which a three body interaction will eject the binary from the system : @xmath81 the binary keeps hardening at a constant rate until either gw radiation takes over and drives its merger or it is ejected from the cluster . the time evolution of the binary semi - major axis due to gw radiation is described by the orbit averaged evolution equation @xcite : @xmath82 where @xmath83 is the binary eccentricity . the merger time for the two bhs is : @xmath84 comparing the above expression with the expression for @xmath85 demonstrates an important point : since the larger the escape velocity from the cluster the smaller @xmath85 , bh binaries that are produced in nscs will have shorter merger time and are therefore more likely to merge within one hubble time than bh binaries from lower - mass gcs . let @xmath86 be the semi - major axis at which gw radiation begins to dominate the energy loss from the binary . a reasonable choice is to set @xmath86 equal to the semi - major axis at which @xmath87 where @xmath88 . if @xmath89 merger happens before ejection . by comparing eq . ( [ agw ] ) with the expression for @xmath85 we see that bh binaries that are produced in nscs are less likely to be ejected from the cluster . the binary will continue to interact with other cluster members until it reaches a semi - major axis @xmath90 . after the binary has decayed to @xmath91 ( where it spends most of its lifetime ) the timescale between two consecutive interactions becomes ( e.g. , * ? ? ? * ) @xmath92 if we assume as before that each interaction removes a fraction @xmath93 of the binary binding energy@xcite , then the timescale required to decay to @xmath91 from a much larger separation is of order @xcite : @xmath94 during the hard interaction the interloper will recoil at the speed @xmath95 . thus the field bhs will start being ejected when @xmath96 , at this point the binary semi - major axis is @xmath97 . at a fractional hardening of @xmath80 per interaction , the mass ejected from the cluster required in order to shrink the binary semi - major axis from @xmath98 to @xmath91 is approximately : @xmath99 so that for low mass clusters @xmath100 and @xmath101 for high mass clusters . given that @xmath85 decreases with the cluster escape speed , the previous equation implies that the larger the cluster mass the fewer bh interlopers will be ejected , in addition to fewer binaries being ejected ; when @xmath102 the bh binary will merge without ejecting any of the field bhs . from the condition @xmath103 , we derive the critical cluster escape velocity above which binaries will merge before being ejected through hard scattering with surrounding stars : @xmath104 so that for @xmath105 we have @xmath106 . the solid line in figure [ fig2 ] shows the locus of points where the escape velocity from the clusters , @xmath107 ( see eq.([vesc ] ) above ) , is equal to @xmath108 . the bh binaries forming in clusters lying to the left of this line are likely to be ejected before merger . the approximate relation @xmath109 implies that only clusters with velocity dispersion @xmath110 will be able to retain their binaries . in many nscs and ucds stellar mass bh binaries will merge while still inside the cluster , while most bh mergers in gcs are expected to occur outside the cluster unless an initially already massive bh ( @xmath111 ) is present in the system@xcite . this result is consistent with monte carlo simulations of gc models where the vast majority of bh binary assembled dynamically through @xmath112-body interactions are found to merge after escaping from their host systems ( e.g. , * ? ? ? figure [ fig2 ] shows that in many nscs and ucds , the bh merger remnants are likely to be retained so they might form new bh - binaries , that will subsequently harden and merge . it is therefore possible that bhs in these massive star clusters will undergo a number of repeated mergers and grow considerably . in addition to the recoil kick due to three - body interactions , as two compact objects merge , asymmetric emission of gravitational radiation will also induce a recoil velocity which can eject the merger product from the system . in the next section we discuss this additional effect . the gw recoil velocity of a merged bh depends on the mass ratio and spins of the progenitor bhs . hence , in order to make predictions about the distribution of recoil velocities for dynamically formed bhs in star clusters we first define the pre - merger bh spin and mass distributions . the mass distribution of bh binaries is quite uncertain . here , we use the bh mass distribution of the dynamically formed merging black hole binaries from the monte carlo models of massive star clusters presented in @xcite . these distributions contain no bh binary with mass ratio less than @xmath113 . this is expected given that dynamical encounters in star clusters tend to pair and eject tight bh binaries with similar mass components binaries in dense stellar environments are prone to exchange components , preferentially ejecting lighter partners in favor of more massive companions @xcite . the distribution of bh spins is also very uncertain . if the bh inspiral is driven predominantly by random gravitational interactions with other bhs and stars we might expect the spin orientations to be close to random . we note that if there is significant coherent gas accretion , the spins might align in a way that might lead to low recoil kicks ( e.g. , * ? ? ? but stellar - mass bhs are unlikely to accrete enough mass from the interstellar medium for this process to be effective @xcite . for these reasons , in our computations the misalignment angle of each bh is chosen at random in cos@xmath114 . for the spin magnitudes of the bhs we consider two choices . the blue solid line in the left panel of figure [ fig1 ] corresponds to a `` uniform '' model in which the initial spin magnitudes are drawn uniformly from the range @xmath115 , where @xmath116 is the dimensionless spin of the bh ( @xmath117 , where @xmath118 is the spin angular momentum in units of @xmath119 ) . the blue dashed line corresponds to one additional `` high - spin '' model in which the spin magnitude is set to a fixed value , @xmath120 . we note here that our spin magnitude distributions differ for example from those of @xcite who adopted low - spinning bhs , leading to low merger kick velocities . our choice is motivated by observations : typical estimates of stellar - mass bh spins suggest high values , @xmath121 , in many cases ( for a review see * ? ? ? in addition , equal - mass non - spinning binaries produce a rotating ( kerr ) bh with final spin magnitude @xmath122 @xcite , so that bhs undergoing more than one merger inside the cluster will have a finite spin magnitude @xcite . however , we note that configurations leading to rapidly spinning bhs are rare . the dimensionless spin magnitude tends to decrease for a bh that engages in a series of mergers , if the lighter bhs with which it mergers have a constant mass @xcite . this will keep the growing bh safely in the cluster after the first few mergers : not only does the mass ratio get farther from unity , which decreases the kick , but the spin of the more massive black hole drops as well . after the pre - merger bh spin and mass distributions have been defined , we compute the recoil kick velocity from the following fitting formula based on the results from numerical relativity simulations of @xcite : @xmath123 @xmath124 @xmath125 @xmath126 \times \nonumber \\ where @xmath127 is the symmetric mass ratio ; @xmath128 and @xmath129 refer to vector components perpendicular and parallel to the orbital angular momentum , respectively , @xmath130 and @xmath131 are orthogonal unit vectors in the orbital plane , and @xmath132 . the values of @xmath133 km s@xmath134 , @xmath135 , @xmath136 km s@xmath134 , and @xmath137 are from @xcite and @xcite , and @xmath138 km s@xmath134 , @xmath139 km s@xmath134 , @xmath140 km s@xmath134 , and @xmath141 km s@xmath134are taken from @xcite . the angle @xmath142 is that between the in - plane component @xmath143 of the vector @xmath144 and the infall direction at merger . we take the phase angle @xmath145 of the binary to be random . the histograms in the left panel of figure [ fig1 ] show the escape velocities from nscs and gcs computed as described in section [ nk0 ] . the blue curves show the recoil velocity distributions for our models computed using eq . ( [ eqn : kick ] ) . the recoil velocity distribution in the uniform spin model is peaked at @xmath146 km s@xmath134 , while the high - spin model produces significantly larger kicks with typical velocities @xmath147 km s@xmath134 . note however that in both models there is a substantial fraction of systems that are accelerated with velocities @xmath148 km s@xmath134 . the left panel of figure [ fig1 ] suggests that only the most massive gcs have a finite probability of retaining a bh merger remnant formed inside the cluster . considering also that bh binaries in gcs are likely to be flung before merger due to three - body encounters , we conclude that the retention probability of bh merger remnants in gcs is small . the left panel of figure [ fig1 ] shows instead that the escape velocities of many nscs are high enough that a substantial number of mergers are expected to be retained inside these systems . in the right panel of figure [ fig1 ] we compute the probability of remaining in the cluster for our spin distributions as a function of the initial bh mass and assuming that the secondary bh mass is @xmath65 . for escape velocities @xmath149 ( typical of massive gcs ) the probability of remaining inside the cluster after a merger is essentially zero , unless the cluster contains initially a bh seed of mass @xmath150 . for escape velocities @xmath151 , which are more typical of nscs , the probability of retaining a bh merger remnant of initial mass @xmath152 is approximately @xmath153 or @xmath154 depending on the assumed spin distribution . this makes nscs excellent candidates for producing massive bh mergers that are potentially observable by aligo , because they can retain their bhs while also evolving rapidly enough that the bhs can sink back to the center and dynamically form new binaries which will subsequently merge . this merger channel is expected to occur quite naturally in massive stellar clusters such as nscs and ucds , while it is unlikely to happen in lower mass systems such as open clusters and gcs . in the next section we present a semi - analytical model that we use in order to make predictions about the mass distribution and rates of bh binary mergers forming in nscs . and @xmath155 of the ejected systems in these models . solid curve is the median of the mass distribution . , title="fig:",width=240 ] and @xmath155 of the ejected systems in these models . solid curve is the median of the mass distribution . , title="fig:",width=240 ] as argued above the dynamical evolution of nscs is of great interest as these systems could represent a important source of inspiraling bhs detectable by aligo . yet a good understanding of the dynamical evolution of massive clusters ( @xmath156 ) and their implications for aligo detections is still elusive . the main difficulty is the large number of particles comprising these systems which makes their treatment extremely challenging even for approximate monte carlo methods . here we adopt a semi - analytical approach which allows us to make predictions about the expected rate and properties of inspiraling bh binaries forming in nscs . first , we define the structural properties of our star clusters . we assign a total stellar mass @xmath15 to the cluster . for @xmath157 the half - mass radius is independent on the cluster mass and it is set to @xmath158 . in the nsc mass regime , @xmath159 , we adopt the fitted relation to the nscs in late type galaxies from @xcite : @xmath160 , with @xmath161 , @xmath162 , @xmath163 and @xmath164 . while sampling from the adopted distributions we also accounted for the scatter of the observed relations . the escape velocity from the cluster is then computed using the approximate eq . ( [ vesc ] ) above ; the cluster velocity dispersion is @xmath165 . the central number density of stars was computed as @xmath166 . this latter expression gives a central number density of stars for a milky way like nsc of @xmath167 and @xmath168 for a @xmath169 gc this is consistent with observed values @xcite . next we define the initial mass distribution and number density of bhs in our cluster models . we take the mass distribution of single bhs from figure6 of @xcite . these authors used the stellar evolution code bse @xcite improved with the stellar remnant prescription from @xcite and @xcite . our models adopt the update prescriptions for stellar winds and supernova fallback , in order to replicate the bh mass distribution of @xcite and @xcite . we consider two values of metallicity , @xmath170 and @xmath171 , defined below as low metallicity and high metallicity models . in our calculations we assume that all clusters formed @xmath172 ago regardless of their mass . while this is a good approximation for gcs , nscs are known to have complex star formation histories , including recent episodes of star formation . we neglect such complication in the following , noting that the bulk of the stellar population in nscs is also likely to be in old stars formed many gyrs ago ( e.g. , * ? ? ? initially , our cluster models have a total mass in bhs that is @xmath173 . this is the typical mass fraction in bhs expected for standard initial mass functions ( e.g. , * ? ? ? the total number of bhs is therefore @xmath174 , with @xmath175 the average bh mass in our models . then we consider natal kicks . for each bh in our fiducial model we compute a natal kick velocity from a maxwellian given by @xmath176 as commonly done for neutron stars @xcite , and assume that the natal velocity of a bh of mass @xmath38 is lowered by the factor @xmath177 . for any sufficiently massive bh progenitor ( @xmath178 ) , the fallback completely damps any natal kick , and the bh is retained in the cluster @xcite . bhs that receive a kick with velocity larger than the escape velocity from the cluster are removed from our models . however , since we only consider massive clusters with large escape velocities , a large fraction of bhs in our models is retained after experiencing a natal kick . this makes our conclusions less sensitive to the prescription we used for natal kicks , provided that the real kick magnitudes are not much larger than what we have adopted here . we discuss in more details the effect of varying the natal kick magnitudes below in section [ bkick ] . we assume that after a time @xmath179 the bhs have segregated to the cluster center . after this time , due to the high densities in the core , bh binaries will efficiently form though @xmath180-body binary formation @xcite and possibly through exchange interactions with stellar binaries @xcite . therefore we assume that after a time @xmath179 , a fraction @xmath181 of the bhs end up in hard bh binaries . although this fraction is quite uncertain , the value we adopted is typical for monte carlo models of massive star clusters with low binary fraction ( e.g. , see fig . 6 in after the bhs have segregated to the center and we have assigned a fraction of them to be in bh binaries we follow the evolution , ejection and formation of new binaries adopting the scheme described in what follows . we divide the bh cluster in @xmath182 sub - groups , each containing the same number of bhs . we find the two most massive bhs in each sub - group and assume that after a time @xmath183 they form a binary . thus , we assume that each sub - group always contains one binary and that this binary is always composed of the two most massive bhs in the sub - group . while each binary is assumed to evolve in the gravitational potential of the entire cluster the adopted numerical scheme allows us to simulate a scenario in which the number of bh binaries in the cluster is approximately constant with time . this , besides allowing us to greatly simplify our approach , appears to be reasonable when compared to the results of monte carlo models of massive clusters @xcite . moreover , we assume that the bh binaries are always composed of the two most massive black holes in each sub - group because during exchange encounters lighter partners are more likely to be ejected . this favors the formation of high mass binaries with similar mass components ( e.g. , * ? ? ? * ) . in our calculation we conservatively assume that the interactions occur between bhs and stellar binaries so that the timescale for binary formation is the longer timescale given by eq . ( [ 12-stars ] ) . the binary fraction in evaluating @xmath183 was computed taking a primordial binary fraction of @xmath184 and lowering this fraction by the number of soft binaries for a constant probability in @xmath185 @xcite . any binary forms with an initial semi - major axis @xmath186 . given the cluster velocity dispersion , its density and the mass of the binary , we compute ( i ) the semi - major axis , @xmath86 , below which gw radiation will start to dominate ( eq . [ [ agw ] ] ) , and ( ii ) the semi - major axis , @xmath85 , at which the binary will be ejected as a consequence of three body scatterings . if @xmath187 the binary will merge outside the cluster and will be ejected with a semi - major axis @xmath188 ; in this case we evaluate the timescale from the formation of the binary to its ejection , @xmath189 , using eq . ( [ mt ] ) so that the lifetime of the binary is @xmath190 . if @xmath191 the binary will merge inside the cluster ; in this latter case the total lifetime of the binary is @xmath192 . in the previous expressions the gw merger timescale , @xmath193 , was computed by sampling the binary eccentricity from a thermal distribution @xmath194 . if the bh binary is ejected from the cluster , then after a time @xmath183 we form a new binary and , as before , we take its components to be the next two most massive bhs in the sub - group . then the hardening timescale of the binary is evaluated as before and it is determined whether the new binary will merge inside the cluster , and , if it does , whether it will be retained inside the cluster after merging . if the binary merges inside the cluster ( i.e. , @xmath191 ) we assign the two progenitor bhs a spin magnitude and orientation from the spin models described in section [ gwr ] and compute the gw recoil speed through eq . ( [ eqn : kick ] ) . in order to account for the recoil kick due to the interaction with a third object we compute a total kick velocity as @xmath195 , with @xmath196 computed as in section [ mass - se ] ( note that @xmath197 typically ) . if @xmath198 the bh merger remnant is ejected from the cluster , otherwise it is retained . if the bh merger remnant is ultimately retained inside the cluster it will have another chance of interacting with new binaries and experience additional mergers . in this case , we place the bh remnant at a distance @xmath199 from the center and evaluate the dynamical friction timescale for the bh to reach the cluster core through eq.([df ] ) . if @xmath200 is greater than @xmath201 the bh is removed from the computation , otherwise after a time @xmath183 the bh forms a new binary with the next most massive bh in the sub - group . then , the hardening timescale of the new binary is evaluated as before and it is determined whether the new binary will merge inside the cluster , and if it does whether it will be ejected from the cluster after the recoil due to anisotropic emission of gw radiation . as the binary hardens we calculate the number of field bhs that are ejected through three - body encounters as @xmath202 where @xmath203 is given by eq . ( [ 3ej ] ) . if the number of ejected bhs becomes larger than the total initial number of bhs we stop the integration . the previous steps ( i.e. , binary formation , hardening , and ejection / merger ) are repeated for each sub - group for a time - step @xmath204 . after an interval of time @xmath204 the remaining bhs in all the sub - group are mixed together and the procedure described above is repeated recursively until either all bhs have been ejected from the cluster or the total integration time exceeds the hubble time . the mixing of the sub - groups every @xmath204 allows us to avoid suppressing exchange interactions between massive bhs that might grow in different sub - groups . in what follows we set @xmath205 , but found that values in the range @xmath206 all produced similar results . the main steps of our semi - analytical algorithm are also schematically illustrated in figure [ dgrm ] . we note that our prescriptions are oversimplified in many ways and that more accurate monte - carlo simulations will be needed in order to confirm our results . one basic simplifying assumption is that the cluster structural properties ( e.g. , central density , half - mass radius ) remain constant in time . we believe that this assumption is also justified in many cases , and especially in very massive clusters where the relaxation timescale is longer . for example , monte carlo simulations of moderately massive gcs find that @xmath14 increases with time , but often only by a factor @xmath207 throughout the cluster evolution @xcite . additionally , in our models we assume that the binary - single interactions rate is always dominant with respect to that of binary - binary interactions . this latter assumption is also reasonable , unless the cluster has a very large initial binary fraction ( @xmath208 ; * ? ? ? finally , we note that we do not follow the evolution of the bh spins through consecutive mergers but assume that the spins are always drawn from the assumed distributions . given that our prescriptions are simplified in many ways , we proceed here by testing the results of our models against the results from the monte carlo models of @xcite . in the upper panel of figure [ fig3 ] we show the total number of mergers per cluster for systems containing different numbers of stars and having different metallicities . in order to convert @xmath15 in number of stars we have taken a mean stellar mass of @xmath209 typical of old stellar populations @xcite . moreover , we select here the bh spins based on the uniform spin model described in section [ gwr ] . our semi - analytical models predict that the total number of mergers increases with cluster mass and so do the monte carlo models . the total number of inspirals over 12 gyr is nearly linearly proportional to the final cluster mass . this result is also in agreement with previous models of gcs and shows that this statement can likely be extrapolated up to numbers of stars of order of a few @xmath210 . our models also predict an inversion of this simple correlation showing that for @xmath211 the number of merging bhs flattens or even declines towards larger cluster masses . this is expected given that for such massive clusters with higher values of @xmath212 have a larger binary formation time @xmath183 . the most massive clusters in our integrations , which could represent nscs , produce up to a few thousand bh mergers per cluster . the lower panel of figure [ fig3 ] gives the median mass of the ejected bh binaries formed in 10 cluster models with mass @xmath213 and half - mass radius @xmath214 . these results are directly compared to those from figure 4 in @xcite which corresponds to a cluster model of initial mass @xmath213 and final half - mass radius @xmath215 . the good agreement between the results of our simplified approach and those of monte carlo simulations gives a high level of reliability to our semi - analytical models . figure [ fig4 ] and figure [ fig5 ] give the masses for each of the bh inspirals occurring in 10 star cluster models of metallicities @xmath216 and @xmath171 respectively . masses in the ranges @xmath217 ( upper panel ) and @xmath218 ( lower panel ) were considered . the overall structure of the plots agrees well with our understanding of the dynamics of bhs and their evolution in star clusters , and with the results of previous work ( e.g. , see fig . 7 of * ? ? ? after the formation of the cluster at high redshift , the bhs segregate to the center , the most massive bhs form binaries and the majority of them are ejected . the cluster processes through its population of bhs that merge and are ejected from most to least massive , so that only low - mass bhs are retained by the present epoch . more massive clusters , which could represent typical nscs , produce bh mergers in the local universe that are significantly more massive than mergers occurring in lower mass clusters . as also noted in @xcite the plateaus in the chirp mass and total mass distributions in figure [ fig5 ] are mainly a consequence of the maximum bh mass in the initial models , which is regulated by the wind - driven mass loss from the vink prescription . for the high metallicity models this produces a large population of @xmath219 bhs , which leads to the formation of a large population of equal - mass mergers with total mass of @xmath220 . more interestingly , we find that massive cluster models produce an additional collection of binaries at @xmath221 and @xmath222 which can be clearly seen at high redshift in figure [ fig5 ] . one of the two bhs in these binaries has experienced one and two earlier mergers with lower mass bhs respectively . for the the low metallicity models there is no apparent collection of sources as might be expected @xcite . the decreased efficiency of the stellar winds in the low metallicity models implies that a lower number of high - mass stars are converted into bhs with the maximum - mass set by the wind - driven mass loss prescription , resulting in a wider range of bh masses . in figure [ fig4 ] and figure [ fig5 ] we show the total and chirp mass and uncertainties associated with the recent detection of the bh binary merger gw150914 @xcite . the reported masses of gw150914 are consistent with the masses of black hole mergers from gcs in the local universe . however , even for the low metallicity models only @xmath223 percent of the total number of mergers in gcs produce a merger at low redshift with a total mass significantly larger than @xmath224 as required to match the total mass of the gw150914 event . in nscs this percentage is significantly larger , being @xmath225 percent of the total number of inspiraling binaries . in high metallicity clusters ( figure [ fig5 ] ) a smaller number of high mass bh mergers is produced at low redshift making these clusters less likely progenitors of gw150914-like events . in table 1 we report the mean number of mergers per cluster obtained from our models . nscs are defined here as clusters with masses in the range @xmath226 , while gcs have masses in the range @xmath227 . in order to obtain the mean rate of mergers we weighted the number of mergers from each of the cluster models by a cluster initial mass function ( cimf ) . for gcs we assume a power law cimf : @xmath228 . for nscs the initial mass function is largely unknown . here we take the imf of nscs directly from the mass distribution of nscs at @xmath229 from the galaxy formation models of @xcite ( their figure 10 ) . these models produce a mass distribution at @xmath230 that is consistent with the observed nsc mass distribution from @xcite . we note that here we might be underestimating the number of _ massive _ mergers from nscs occurring at low redshift because we have assumed that these systems are as old as galactic gcs . in fact , while most nscs appear to be dominated by old stellar components they are also known to have a complex star formation history and to contain young stellar populations which can produce high mass mergers also at later times ( we will come back to this point below ) . it is also possible that a large fraction of the nsc stars accumulated gradually in time by infalling globular clusters that decayed to the center through dynamical friction . if this process is the main mechanism for nsc formation , then nscs and gcs will comprise similar stellar populations @xcite . table 1 shows that our models predict a few thousands bh mergers per nsc over 12 gyr of evolution . this expectation also appears to be consistent with previous estimates @xcite . in addition , nscs produce between @xmath231 to @xmath232 bh mergers with high mass @xmath233 at @xmath234 depending on the bh spin magnitudes and assumed metallicities distribution of the underlining stellar population . our gc models produce only a few mergers per cluster within @xmath234 and total mass @xmath235 . these massive binaries are found to form only in the most massive gcs ( @xmath236 ) of low metallicity . the number of massive mergers at low redshift is also sensitive to the spin magnitude distribution we assume . for high spin models , a smaller number of bhs are retained in the clusters compared to the uniform spin models . consequently , high spin models produce fewer high mass bh mergers at low redshift compared to models that assume low spins . however , in either spin models a number of inspiraling bh binaries with mass @xmath237 is found to merge at low redshift . finally , table 1 gives the number of bh mergers that are retained inside the cluster . between 10 and @xmath238 percent of high mass ( @xmath239 ) mergers occurring in nscs at @xmath240 are retained inside the cluster enabling the formation of even more massive bh mergers . the results presented in this section suggest that nscs are a natural environment for producing bh mergers that are observable by aligo detectors . in addition to this , nscs can form high mass bh binaries , and mergers with mass consistent with that of gw150914 also in relatively high metallicity environments . the implications of our results are discussed in more detail in the following section . [ cols="<,<,<,<,<,<,<",options="header " , ] [ t1 ] our study shows that a multitude of bh binary mergers can be produced at the center of galaxies where nscs reside . in the following we derive an approximate expected detection rate from the results of our models and discuss some implications for possible aligo detections of these mergers over the next decade . finally we discuss the production of bh mergers in nscs hosting a central mbh , and the possibility of a continues supply of bh - binaries in nscs through episodic and/or continuous star formation . here we use results from semi - analytical galaxy formation models to derive an expected mbh occupation fraction in nscs and use this as well as the results of the cluster semi - analytical models presented in this paper to make predictions about the merger rate of bh binaries produced in nscs . we also consider the merger rates from gcs and compare these to estimates made in former studies . to compute the aligo merger rate of bh binaries per unit volume we use the following expression : @xmath241 where @xmath242 is the number density of galaxies , @xmath243 is the averaged merger rate of bh binaries per cluster that merge within the observable volume , and @xmath244 is the fraction of galaxies which host a nsc but do not have a mbh . while observations show that nscs and mbhs coexist in some galaxies , and that nscs exist in most galaxies , @xmath244 remains largely unconstrained . here we use the results of semi - analytical galaxy formation models that follow the cosmological evolution of galaxies , their mbhs and nscs . these galaxy formation models are described in @xcite , @xcite and @xcite . figure [ fig6 ] shows the fraction of galaxies in these models that contain a nsc but do not host a mbh . these models predict that the number of galaxies hosting a nsc but without a mbh is quite large , being @xmath245 for galaxies with total mass @xmath246 regardless of galaxy type . based on figure [ fig6 ] we adopt here a conservative value of @xmath247 , and adopt a number density of galaxies of @xmath248 ( e.g. , * ? ? ? * ; * ? ? ? assuming that bh - bh mergers can be seen by aligo out to a redshift of @xmath249 @xcite , which corresponds to an age of the universe of @xmath250 @xcite , then eq . ( [ r - f ] ) gives a merger rate of bhs in nscs of @xmath251 thus our calculation predicts a substantial number of detectable bh mergers from nscs . from table 1 we see also that between 10 and @xmath238 percent of the total number of merger remnants are retained inside nscs . adopting an aligo detection - weighted comoving volume of @xmath252 for full design sensitivity @xcite , the rate in eq . ( [ gnsc ] ) translates into a detection rate of @xmath253 . the merger rate of eq . ( [ gnsc ] ) can be directly compared to that from gcs : @xmath254 which was obtained from the number of mergers per gcs within @xmath234 in table 1 and assuming a number density of gcs equal to @xmath255 @xcite . note that eq . ( [ gcs ] ) is very well consistent with the rate previously derived by other authors @xcite . as before , taking an aligo detection - weighted comoving volume of @xmath256 for full design sensitivity , we obtain an aligo detection rate of bh mergers from gcs of @xmath257 . ) but no mbh as a function of _ total _ galaxy mass . results are from the galaxy formation models of @xcite and @xcite . dashed line is for galaxies with bulge - to - total - mass - ratio smaller than 0.7 . solid line is for early type galaxies which are defined here as galaxies with bulge - to - total - mass - ratio larger than 0.7 . ] now we only consider mergers with mass @xmath258 occurring at a redshift @xmath234 ( see table 1 ) . we define these as possible progenitors of the event gw150914 . the detection rate of high mass mergers from nscs in the local universe is in the range @xmath259 where the lower limit corresponds to high metallicity clusters and the high spin model , and the upper limit to low metallicity clusters and to a uniform bh spin distribution . interestingly , we find that between 20 and 50 percent of all massive mergers in nscs are produced by the consecutive merger channel discussed in this paper with a few percent having a mass @xmath260 . the corresponding merger rate of high mass bh binaries mergers in gcs is @xmath261 similar to that corresponding to high mass bh mergers produced in nscs . both rates of high mass mergers from gcs and nscs are marginally consistent with the rate of @xmath262 of gw150914-like mergers given by @xcite . from our rate computation we find that the detection rate of bh - bh binaries from nscs is substantial , and it is about one tenth of that from gcs . importantly , we also find that the nsc detection rate of high mass bh mergers similar to gw150914 ( @xmath263 ) is comparable to that from gcs , with many of the mergers being produced through the consecutive merger scenario discussed in this paper . finally , our results show that most gw150914-like mergers are more likely to be a product of dynamical interactions occurring in massive clusters of low metallicity , also in agreement with previous findings @xcite . as a function of the dispersion of the maxwellian distribution of natal kick velocities , @xmath264 , applied to all bhs independently of mass . red curves correspond to a metallicity @xmath265 and black curves to a metallicity @xmath266 . we have assumed a number of gcs per galaxy equal to 100 and a light travel time of @xmath267 at @xmath268 . at @xmath269 the merger rate of bh binaries from nscs is dominant compared to the corresponding merger rate from gcs . ] @xcite noted that `` _ _ for both dynamical formation in [ globular ] clusters and isolated binary evolution , the implication of bh binary existence is that bh natal kicks can not always be high ( @xmath270 ) , in order to avoid disrupting or widening the orbits too much , or ejecting the bhs from clusters before they can interact . _ _ '' for example , large natal kicks will widen the orbits of massive binary progenitors and quench the formation of binary bh systems that will merge within the age of the universe ( e.g. , * ? ? ? @xcite did not consider the possibility that bh binaries can be dynamically assembled inside nscs which we discuss here . motivated by the fact that the distribution of formation kicks for bhs is largely uncertain , even at the qualitative level , we assume here that the kick magnitude distributions are fully unconstrained and explore how the nsc and gc bh merger rates are affected by varying these distributions . in figure [ fig5b ] we plot the mean merger rate per galaxy at @xmath240 in models where the bhs are given a natal kicks taken from maxwellian distributions with dispersion @xmath264 , and the spin magnitude of the bhs is uniformly distributed in the range @xmath271 . unlike the models discussed above , here we assume that the bh natal kick distributions do not depend on the mass of the bhs . as @xmath264 increases above @xmath272 the merger rate from gcs decreases and becomes comparable to that of nscs for @xmath273 . above this value the merger rate of nscs begins to decrease as well because a fraction of bhs starts to be ejected from low mass nscs as well . however , for @xmath274 we see that the bh merger rate from nscs becomes significantly larger than the corresponding merger rate from gcs . we conclude that for high natal kicks , @xmath29 , nscs can dominate the merger rate of dynamically formed bh binaries that are detectable by aligo . recent theoretical studies suggest that the birth kicks may not be directly correlated with the bh mass @xcite , and that the distribution of black hole kick velocities could be similar to that of neutron stars @xmath275 ( but see * ? ? ? if this is the case , then the merger rate of bh binaries from nscs will greatly dominate over that from gcs . however , we also note that for @xmath276 the rate of bh mergers from nscs becomes significantly smaller than the rate of @xmath262 implied by the detection of gw150914 @xcite . thus very high values of natal kicks are excluded by our analysis when combined with the estimated aligo detection rate . so far we have considered the merger of bh binaries in stellar clusters which do not host a central mbh . this is certainly not the case for at least a handful of nscs which are known to have mbhs at their center ( e.g. , * ? ? ? one example is our own galaxy which contains a @xmath277 nsc and a central mbh of @xmath278 @xcite . more generally nscs and mbhs are known to co - exist in galaxies with masses @xmath279 ( e.g. , * ? ? * ; * ? ? ? * ) ; galaxies with masses lower than this value show clear evidence for nucleation but little evidence for a mbh . conversely galaxies with masses above @xmath280 are dominated by mbhs but generally show no evidence for nucleation @xcite . in nscs containing a central mbh , the merger rates of bh binaries are expected to be significantly different , although not necessarily smaller , than the rates given in table 1 . if a mbh is present it will inhibit core - collapse , causing the formation of a bahcall - wolf cusp instead @xcite . after the bhs have segregated to the center their densities will dominate over the stellar densities within a radius @xmath281 times the mbh influence radius @xcite note that this latter statement depends on the formation history of the nsc @xcite . at such small distances from the mbh all binaries will be effectively soft so that any interaction with a third bh or star will tend to make the binary internal orbit wider . in this situation , three - body interactions between binaries and field objects will not lead to bh mergers but rather to the `` evaporation '' of binaries . several mechanisms have been discussed in the literature which can produce bh mergers even in the extreme stellar environment of mbhs . below we briefly review two of these processes : ( i ) mergers of bh binaries due to lidov - kozai ( lk ) resonance induced by the central mbh @xcite ; ( ii ) single - single captures of compact objects due to gravitational wave energy loss @xcite . @xcite showed that near a mbh the dynamical evolution of binaries is dominated by perturbations from the central mbh . in particular , the lk mechanism exchanges the relative inclination of the inner bh binary orbit to its orbit around the mbh for the eccentricity of the bh binary @xcite . at the peak eccentricity , during a lidov - kozai cycle , gw emission can become efficient leading to a merger of the two bhs . more recently @xcite performed @xmath112-body simulations of small clusters of stars containing a central mbh , to estimate the rate of bh mergers induced by the lk mechanism . these authors found that this mechanism could produce mergers at a maximum rate of @xmath282 per myr per milky way equivalent galaxy . this rate appears to be somewhat comparable to the upper end of the expected detection rate from stellar clusters @xcite . @xcite showed that this rate could translates into a maximum rate per volume of @xmath283 . however , as also noted by @xcite their merger rate estimates are likely to be an overestimate of the true merger rate from the lk process as they used optimistic values for both the merger fraction as well as for the bh binary fraction . the efficiency of the lk process in inducing bh mergers in nscs with mbhs is sensitive to major uncertainties . the major challenge to this being that a continuous supply of bh binaries is needed in order to obtain a finite merger rate . in fact , a continuous formation of binaries is part of the assumptions made in the rate estimates of @xcite and @xcite . binaries well inside the influence radius of a mbh will be essentially all soft and will be disrupted over the typical timescale @xcite : @xmath284 because @xmath285 is much shorter than the lifetime of any nsc , we expect that most primordial binaries will be disrupted by now in these systems . @xcite describe various processes which may affect the replenishment rate of compact binaries and/or their progenitors in nscs with a mbh . @xcite suggested that disruption of triple stars could leave behind a binary in a close orbit around the mbh and could serve as a continuous source of replenished binaries close to the mbh . in addition , in - situ star formation can also repopulate the binary population of nscs . central star formation bursts may occur continuously ( or episodically ) throughout the evolution of the stellar cusp and lead to a steady population of massive binaries near the center . finally , nscs might result from the merger of stellar clusters in the inner galactic regions ( e.g. , * ? ? ? * ) . such clusters may harbor an inner core cluster of bhs that formed during the cluster evolution . if these bhs are retained in the cluster this mechanism may also contribute to the bh and bh - binary populations in nscs @xcite . @xcite showed that in the dense stellar environments such as those of nscs bh binaries can efficiently form out of gw radiation during bh - bh ( single - single ) encounters . interestingly they show that most of the mergers from this channel will have a finite eccentricity while they enter the @xmath286 frequency band of aligo . this processes could become important for sufficiently large cluster masses . however , the predicted rate of bh - bh mergers from this channel , although very uncertain , is estimated to be only @xmath287 @xcite , and therefore it is sub - dominant with respect to the rate from the other processes discussed in this paper . finally , we note that @xcite argued that even the rate of _ eccentric _ mergers from bh - bh scattering in nscs is likely to be much smaller to that of eccentric mergers from bh triples formed in gcs . in conclusion , the role of nscs containing mbhs in producing bh mergers is a subject of considerable interest , which will likely require high precision @xmath112-body simulations of large number of particles . for now , theoretical models suggest that the merger rate of bhs in these systems might be considerably lower than that from nscs without mbhs . as mentioned above , most nscs are known to have undergone a complex star formation history characterized by recurrent episodes of star formation ( e.g. , * ? ? ? thus nscs might still be forming bhs and bh binaries at the present epoch . this is different from gcs where all stars are old ( ages @xmath288 ) and many ( if not all ) bhs are expected to have been already ejected by now . as an example , our galactic center contains a large population of young massive stars , many of which reside in a stellar disk . these stars most likely originated in - situ following the fragmentation of a gaseous disk formed from an infalling gaseous clump ( e.g. , * ? ? ? such stars formation bursts are thought to occur episodically throughout the evolution of the central cusp . eclipsing and close binaries are observed among the galactic center young stars , suggesting star formation as an additional process which can repopulate the binary population ( including bh binaries ) in the nsc @xcite . observational studies of nscs in external galaxies , including high resolution spectroscopic surveys , have been used to characterize the star formation history and ages of nscs @xcite . the common finding emerging from these studies is that most nscs are characterized by a mixture of morphological components and different stellar populations spanning a wide range of characteristic ages from 10 myr to 10 gyr . observations also suggest that the ages of nscs and masses depend on the host galaxy hubble type , with nscs in early - type spirals being older and more massive than those of late - type spirals @xcite . more generally the growth of the nuclei might be a continuous and ongoing process occurring during and after most of the host galaxy was formed . as shown in figure [ fig4 ] and [ fig5 ] , most massive bh mergers ( @xmath289 ) occur at early times in the lifetime of a single stellar population star cluster . this is because the most massive objects segregate earlier and are ejected earlier through dynamical interactions . this will not be the case if the clusters form new stellar populations at later times . new episodes of star formations in nscs will lead to the formation of new bhs with a substantial contribution to the merger rate in the local universe . in - situ star formation could therefor contribute significantly to the detection rate of high mass mergers we previously derived ( c.f . , eq . [ [ ncs2 ] ] ) , although it will likely not affect the total detection rate of bh mergers in nscs . understanding the distribution of bhs at the centers of galaxies is crucial for making predictions about the expected event rate and source properties for high - frequency gravitational wave detectors . since the distribution of stellar bhs is not known , and monte carlo simulations of star clusters are still limited to a few @xmath290 particles , we opted here for a semi - analytical approach which we used in order to make predictions about the properties and rate of bh binary mergers that are dynamically assembled in nscs . in the future we plan to explore this topic using more accurate , although computationally more demanding , monte carlo simulations . the main conclusions of our work are summarized below . * nscs produce bh binary mergers at a realistic rate of @xmath291 . * bhs in nscs can experience a number of mergers and grow to masses up to a few hundred solar masses . although rare , such high - mass bh mergers at low redshift are unique to nscs , because these are the only clusters with sufficiently high escape velocities such that they can retain a large fraction of their merging bhs . * assuming that bhs receive low natal kicks , with an imparted momentum equal to the momentum imparted to neutron stars , then the nsc detection rate of high mass bh mergers similar to gw150914 ( @xmath292 ) is @xmath293 . this rate is comparable or larger tan the corresponding merger rate of dynamically formed bh binaries in gcs . * if bhs receive natal kicks as large as @xmath29 then bh binary mergers produced dynamically in nscs could dominate over the merger rate of similar sources produced either in gcs or through isolated binary evolution . we thank enrico barausse , sourav chatterjee , cole miller , carl rodriguez for useful discussions and the anonymous referee for useful suggestions . fa acknowledges support from a ciera postdoctoral fellowship at northwestern university . far acknowledges support from nsf grant ast-1312945 and nasa grant nnx14ap92 g , at northwestern university , and from nsf grant phy-1066293 through the aspen center for physics .. abbott , b. p. , abbott , r. , abbott , t. d. , et al . 2016 , arxiv:1602.03840 abbott , b. p. , abbott , r. , abbott , t. d. , et al . 2016 , arxiv:1602.03842 abbott , b. p. , abbott , r. , abbott , t. d. , et al . 2016 , , 818 , l22 antonini , f. , faber , j. , gualandris , a. , & merritt , d. 2010 , , 713 , 90

we present a study of outflow and feedback in the well - known seyfert 2 galaxy markarian 573 using high angular resolution long - slit spectrophotometry obtained with the _ hubble space telescope _ imaging spectrograph ( stis ) . through analysis of the kinematics and ionization state of a biconical outflow region emanating from the nucleus , we find that the outflow does not significantly accelerate the surrounding host - galaxy interstellar gas and is too weak to be a strong ionization mechanism in the extended emission regions . instead , the excitation of the extended regions is consistent with photoionization by the active nucleus . from energetics arguments we show that the nuclear outflow is slow and heavy and has a mechanical luminosity that is only% of the estimated bolometric luminosity of the system . the energy in the outflow is able to mildly shape the gas in the extended regions but appears to be insufficient to unbind it , or even to plausibly disrupt star formation . these results are at odds with the picture of strong agn feedback that has been invoked to explain certain aspects of galaxy evolution .

0902.0004

recent studies of galaxy evolution have invoked a significant component of energetic feedback from an active galactic nucleus ( agn ) to attempt to explain many of the observed properties of galaxies . specifically , agn feedback has been implicated as the primary cause behind the differences between observed and theoretical galaxy luminosity functions , the observed color bimodality of galaxies @xcite in which galaxies are divided into distinct populations of blue star - forming galaxies and red dead galaxies ( e.g. * ? ? ? * ) with a pronounced deficit of galaxies with intermediate color , and the very tight correlations observed between the masses of supermassive black holes and their host galaxy spheroid s velocity dispersion @xcite and luminosity @xcite . the relative lack of galaxies with intermediate star formation rates suggests that star formation ceases abruptly rather than gradually @xcite , while the tight correlations of black hole masses with the properties of their host bulges are taken as evidence that the two are strongly co - eval . in the various proposed feedback scenarios , the supermassive black holes grow primarily by accretion until they become sufficiently massive and energetic that thermal and/or radiative feedback from their activity either heats the surrounding interstellar gas or mechanically pushes it out of the host galaxy @xcite . this simultaneously deprives the host galaxy of raw material from which to form new stars @xcite and starves the black hole , truncating its growth @xcite . despite the considerable utility of agn feedback , there is scant observational data to inform us about how or even if agn feedback operates in actual galaxies . arguably some of the most promising cases are instances of mechanical feedback , such as the outflows that have been detected as blue - shifted uv and x - ray intrinsic absorbers in @xmath050% of seyfert 1 galaxies and quasars @xcite . while it is clear that nuclear outflows are common , it is still difficult to determine the mechanical luminosity ( critical for determining if the outflow can unbind circumnuclear gas ) and mass outflow rate to determine their impact . a good example of this difficulty is the work of @xcite on the time evolution of the ionization state of the x - ray absorbers relative to the x - ray continuum in the warm absorber of ngc4051 . from the absorption line variability they obtained good constraints on the density and location of the absorbers from the black hole . assuming a biconical geometry , they measured a low outflow velocity with respect to the escape speed from the black hole , and a corresponding low mass outflow rate with respect to the accretion rate of the black hole . from these values , they conclude that while the outflows might disrupt the hot ism , they are not capable of ejecting large amounts of interstellar gas from the host . @xcite reached a similar conclusion for the nearby seyfert galaxy markarian78 using visible - wavelength and radio data to measure the kinematics and ionization state of extended emission - line regions surrounding the active nucleus . they concluded that the biconical outflow in mrk78 is weak , slow and heavy , and hence insufficiently energetic to significantly heat or unbind surrounding host galaxy gas . the results for these two galaxies present an apparent problem for agn feedback : in both cases the measured agn outflows are too weak to thermally or mechanically disrupt star formation in the host galaxy . nevertheless , these are just two objects , and analyses of other agns are critical for further understanding the role of agn feedback in galaxy evolution . seyfert galaxies are of special interest for detailed studies of agn feedback as many are luminous enough to be accreting near their maximal eddington rate , yet close enough to study at relatively high angular resolution ( scales of 10100pc for nearby examples ) using the _ hubble space telescope _ and ground - based adaptive optics . a very promising candidate for detailed study is the nearby seyfert galaxy markarian573 , which is well - known for its extended , richly structured circumnuclear emission - line regions . mrk573 features a prominent ionization bicone and bright arcs and knots of emission - line gas @xcite that are strongly aligned and interacting with a kiloparsec - scale low - power radio outflow @xcite . the association between the radio lobes and the emission regions was studied in detail by @xcite , who proposed that the arcs result from gas cooling after passing through a radio - induced radiative bow shock . in contrast , @xcite argued that the arcs are instead a morphological artifact of dust lanes in the galaxy being illuminated by the nuclear ionizing continuum . this is supported by visible and near - infrared _ hst _ images that show the emission - line arcs as extensions of larger circumnuclear dust lanes that are illuminated when they pass into the ionization cone . circumnuclear dust lanes like these are commonly found in active and inactive galaxies alike @xcite , and in the specific case of mrk573 they appear to be shaped by the presence of a nuclear bar @xcite , not unlike what is seen in galaxies with inner bars regardless of the presence of nuclear activity . @xcite examined several possibilities for the origin of the arcs and extended emission - line regions using a combination of _ hst _ images and ground - based , integral - field spectroscopy . they modeled the excitation in the arcs as either shock features , linked to radio jets , or pre - existing structures photoionized by the nucleus . using standard emission - line diagnostics , they showed that the inner arcs are excited by the central continuum source rather than by fast , photoionizing shocks , citing as evidence that they detected no signs of a strong kinematic interaction through a radio - induced shock . the excitation of the outer arcs was not completely explained by nuclear photoionization ; these arcs require an external source of photons in addition to the nucleus to account for the excitation levels . however , their analysis suggested that both the inner and outer arcs show little evidence that they are shock excited . in this paper we present new long - slit spectrophotometry of mrk573 obtained with the hubble space telescope imaging spectrograph ( stis ) . the high angular resolution of stis , combined with good velocity resolution in the bright h@xmath1 + [ ] emission lines allows us to examine the excitation and kinematics in these regions in greater detail than possible in previous studies . our analysis of mrk573 follows a threefold approach . first , we examine the kinematics of the circumnuclear emission - line regions to separate kinematically - disturbed gas from quiescent gas in the rotating disk of the galaxy ( see sec.[sec : los_vel ] ) . the existence of these two kinematical components was clear from the earlier ground - based work , but the greater angular resolution provided by _ hubble _ lets us pinpoint the regions of disturbed and undisturbed gas on a scale of 10s of parsecs . next we measure the densities and temperatures in the emission - line regions using standard nebular diagnostic emission lines . we examine the excitation state of these regions , comparing them with the predictions of published photoionization and shock excitation models to clarify the nature of the ionization mechanism in the bright emission line areas ( see sec.[sec : spectrophot ] ) . finally , we undertake a quantitative analysis of the outflow energetics of mrk573 using techniques developed by @xcite to examine the degree of feedback on the host galaxy surroundings ( see sec.[sect : whittle_quant ] ) . throughout this paper we adopt a distance of 74mpc for mrk573 ( @xmath2 km s@xmath3 , @xmath4 km s@xmath3mpc@xmath3 ) , giving us a projected linear scale of @xmath0360pcarcsec@xmath3 . for reference , this is approximately half the distance of mrk78 ( @xmath5 km s@xmath3 ) . we acquired spectra of the nucleus of mrk573 on utc 2001 october 17 using the _ hubble space telescope _ and stis , with the 52@xmath60.2 aperture . our data were obtained in one observation period lasting two orbits with two spectral settings : the medium - dispersion g750 m grating centered on the h@xmath1 emission line , ranging from @xmath7 and the low - dispersion g430l grating providing coverage from @xmath8 . these configurations provide fwhm spectral resolutions for extended sources of 2.2 and 10.9 respectively , and angular resolution of 0051 pixel@xmath3 . the spacecraft roll angle was unconstrained in our observation planning as the data were acquired as part of a program to observe the nucleus proper , and an attempt to constrain the slit position angle to the apparent jet axis resulted in plan windows too short for practical scheduling . by pure good fortune , the slit position angle was @xmath9 , which is within @xmath10 of the radio - axis position angle of @xmath11but still passes through the bright part of the inner arcs , well within the ionization cone and the main region of the nuclear outflow seen by others . figure [ fig : em_regions ] shows the location of the slit relative to the emission - line regions on a contrast - enhanced `` structure map '' created from an _ hst _ f606w image ( see * ? ? ? the g750 m spectra were acquired as a sequence of three exposures with target exposure times of 1080 , 1080 , and 840s . the g430l observations consisted of two exposures of 805 and 840s duration . to avoid problems with hot pixels on the stis ccd , the observations were dithered by @xmath12025 ( 5 pixels ) along the slit relative to the first spectrum . wavelength calibration lamp spectra were taken during earth occultation . our data were acquired after the primary ( side-1 ) stis electronics failed on utc 2001 may 16 . since the side-2 electronics did not provide closed - loop temperature control of the stis ccd , the ccd temperature varies with the ambient temperature of the spacecraft when the thermoelectric cooler is run at a constant rate @xcite . the result is that the dark rate varies with temperature , and the standard dark calibration images used by the calstis data reduction pipeline are often a poor match to the actual dark rates in the hot pixels . to correct for this , we broke out of the calstis pipeline after the basic2d step and corrected the dark pixels individually before re - inserting them into the pipeline for the final wavelength calibration , flux calibration , and geometric rectification steps . in brief , a custom program scanned the data quality frame ( dqf ) for each individual image and created a list of hot pixels . this list was used to fix all pixels in the science frame that had either negative data values or that deviated from the median within a 7@xmath61-pixel box centered on that pixel by more than 2@xmath13 . for this last step , the long - axis of the analysis box is oriented along the dispersion axis of the detector . once this hot - pixel surgery was completed for the three g750 m observations , the second and third images in the set were then aligned with the first and co - added using the calstis ocrreject task to detect and remove cosmic rays . remaining cosmic rays missed by this step were removed by hand using the median filter routine tvzap in xvista . the final combined , cleaned , spectral image was then fed back into the calstis pipeline to produce the final calibrated long - slit spectrum . an analogous process was used for each of the two g430l spectra to create the final g430l long - slit spectrum , but because only two images were available , more hand cleaning of residual cosmic rays was required . this process yielded superior results by comparison with the default reduction pipeline . an analogous procedure is described in @xcite . an additional consequence of using the stis side-2 electronics is that our long - slit spectra are affected by low - amplitude ( 23adu peak - to - peak ) fixed - pattern noise with a @xmath03-pixel horizontal scale that takes the form of a regular herring - bone noise pattern when viewed at high contrast . attempts to remove this component using fast fourier - transform filtering only made things worse . its primary effect is to increase the effective readout noise of the device by about 1@xmath14pix@xmath3 for the gain=1 mode of our data . when extracting faint outer regions of the long - slit spectra , we took particular care to be wary of this fixed - pattern component in interpreting our spectra . the final , calibrated long - slit spectra are shown in figure[fig : lspec ] . we have adopted the nomenclature of @xcite for the specific emission - line regions we examine below . the extended emission - line region of mrk573 consists of a series of arcs and knots located northwest and southeast of the nucleus along the general direction of the radio outflow . our stis slit intersects the arcs but just misses a group of bright knots located along the main radio axis as shown in figure[fig : em_regions ] ) . we shall focus our attention on four specific regions : the se1 arc , se2 arc , the nucleus , and the nw1 arc @xcite . the arcs at regions se1 and nw1 are located at a projected angular distance of @xmath02 on either side of the nucleus , and the se2 arc is @xmath03 from the nucleus . there is another arc in the northwest region that is roughly symmetric with the se2 arc , but the emission lines from this region are too faint and diffuse for detailed analysis with our spectra . we extracted two 1d spectral data sets from the 2d spectra . the first set is integrated spectra of specific regions of interest for measuring the density , temperature , and excitation state of the line - emitting gas . for these we first created slit intensity profiles for the h@xmath1 + [ ] emission lines in the g750 m spectrum and for [ ] @xmath155007 in the g430l spectrum , and for adjacent line - free continuum regions bracketing these lines . the derived mean continuum profiles were then subtracted from the line profiles to produce pure emission - line profiles for each line . these profiles were then used in conjunction with the 2d spectra of the nuclear regions ( see figure[fig : lspec ] ) to precisely locate the windows for extracting 1d spectra of the regions we wished to examine . the second set of spectra were extracted from the g750 m data for making kinematic measurements - line centroids and velocity widths - using the h@xmath1 and [ ] emission lines . for these we extracted 1d spectra in contiguous increments along the slit moving radially outwards from the nucleus . in bright regions we extracted spectra from single - pixel apertures ( 02@xmath6005 ) , and as the regions became fainter at larger radii , we increased the width of the extraction window up to a limit of 4 pixels wide ( 02@xmath602 ) to improve the signal - to - noise ratio in the lines . this gives us nearly continuous spatial sampling along the slit for measuring line - of - sight velocities and emission - line velocity dispersions . after identifying individual emission lines , we measured the line parameters with the liner interactive line analysis program developed at ohio state . for each spectral line , we define a local continuum by averaging over adjacent line - free regions and then fit the lines with a single- or multi - component gaussian to resolve blends or multiple velocity components if present . we derive the line centroid , full - width at half maximum ( fwhm ) , and integrated line intensity for each emission - line component present . when fitting doublet lines of a single ionic species ( e.g. , [ ] @xmath166548,83 and [ ] @xmath166716,31 ) we fit both lines together , imposing physically - motivated constraints on the profile parameters to improve the fidelity of the fits . for the [ ] doublet , we constrain the relative line centroids in velocity space and match them in line width , but leave the relative intensities ( which are correlated with density ) unconstrained . for the [ ] doublet lines , which arise out of the same upper excited level , we constrain their relative line centroids and widths and further constrain their relative line strengths as dictated by the ratio of their radiative transition probabilities . this greatly improves the quality of the fits when deblending these lines from the h@xmath1 emission line that lies between them . some of the emission lines appear double - peaked , particularly in the kinematic spectral extractions . in these cases we model the two peaks as gaussians and apply similar rules for setting constraints on the physically related line pairs . finally , if the line of interest is very faint but clearly detectable and unblended , we extract the line intensity by direct integration without attempting to fit a line profile . the rms uncertainty in the continuum fits and a simple model for the pixel - to - pixel signal to noise ratio is used to estimate the overall quality of the fit and the uncertainties on the best - fit line parameters . we make additional checks of our fit - derived line parameters in a number of ways . the total line intensities are compared to a numerical integration of the blend itself without fitting to a particular line shape . line fwhms from the fits are compared to direct measurements of the fwhm using profile tracing algorithms that assume no underlying profile shape ( this works best for the brightest emission lines with the least blending and highest signal - to - noise ratio ) . finally , the gaussian fit derived line centroids for well - isolated lines are compared to intensity - weighted centroids derived from the central 5 pixels about the peak . all of these give us further estimates of the relative uncertainties , and help alert us to possible systematics that would be missed by simply examining the rms residuals . the measured fluxes for various emission lines are listed in table[tab : regional_feat_fluxes ] . in general , line centroids and fwhm are measured to @xmath120.1 or @xmath020 km s@xmath3 for the h@xmath1 and [ ] emission lines , with some degradation in fainter regions . measurement uncertainties for total line fluxes are typically @xmath1210% in the brighter lines , increasing to @xmath1220% in the faintest regions measured . we pursue two different analyses of our spectra . the first uses the kinematic spectra ( sec . [ sec : los_vel ] ) to derive outflow speeds and separate outflowing gas from ambient ism gas in orderly rotation in the disk . combining the outflow kinematics with the emission - line geometry seen in the direct images provides us with constraints on the configuration of the outflow , allowing us to deproject the observed radial velocities and thus determine the actual physical outflow velocities . the second line of analysis uses emission line diagnostics from the spectrophotometric set of spectra ( sec . [ sec : spectrophot ] ) to estimate the density , temperature , and excitation state of the gas in these regions . these are compared to published models of shock- and photo - ionized gas to determine the excitation mechanism in these regions . figures [ fig : halpha_rotcurve ] and [ fig : oiii5007_rotcurve ] show the radial velocities along our long slit for the h@xmath1 @xmath156563 and [ ] @xmath155007 emission lines @xmath154861 , [ ] @xmath17 , and [ ] @xmath164363 , 4959 lines show consistent results and are available upon request . ] . a number of distinct kinematic systems are present . the largest - scale pattern is a rotating disk extending out as far as we can trace ionized gas ( roughly @xmath124 ) , with the nw side receding and the se side approaching us and an amplitude of roughly @xmath12125 km s@xmath3 . in the inner 1 on either side of the nucleus the kinematics are dominated by a coherent system of double - valued radial velocities ( separated by nearly 300 km s@xmath3 ) with a red / blue splitting pattern characteristic of a biconical outflow ( e.g. * ? ? ? * ; * ? ? ? finally , in the inner 05 around the nucleus we see various additional h@xmath1 velocity components with no obvious pattern associated with the inner parts of the outflow region . the high - velocity components of the biconical outflow pattern and the large - scale rotation pattern are also visible in the [ ] @xmath155007 line - of - sight velocities ( figure[fig : oiii5007_rotcurve ] ) despite the lower dispersion of the g430l grating . in the biconical outflow models of @xcite and @xcite emission from the front of the cone ( the surface closest to the observer ) is blueshifted and the back of the cone is redshifted . when the bicone axis is at high inclination angles ( the cone axis near the line of sight ) there is a strong blue - to - red asymmetry in the radial velocities , while at low inclinations ( the axis near the sky plane ) , the radial velocity pattern is more symmetric @xcite . in the context of these models , our radial velocity maps indicate that the biconical outflow in mrk573 has a relatively low inclination with the nw cone directed toward us . using the velocities from our h@xmath1 emission line , the observed ionization cone opening angle of @xmath18 @xcite , and assuming a constant velocity , @xmath19 , of material in a shell along the cone and a slit roughly aligned with the cone axis , the @xcite outflow model gives projected velocities of @xmath20 for the red and blue sides of the cone , respectively . the cone opening angle ( @xmath21 ) is related to these velocites and the cone inclination angle ( @xmath22 ) by @xmath23 which gives a bicone inclination angle of @xmath24 and @xmath25 km s@xmath3 for the nuclear outflow in mrk573 . we model this outflow geometry in figure[fig : geometry ] . between about 115 from the nucleus there is very little emission line gas , and so we have no measurements in these regions . this region between the nucleus and the two close emitting arcs likely represents the nearly evacuated `` bubbles '' resulting from the biconical outflow . beyond about 15 on either side of the nucleus a coherent pattern of velocities consistent with a rotating disk of relatively undisturbed gas extends out to the limits of our long - slit data . this sense of rotation ( se approaching , nw receding ) is opposite the direction of the biconical outflow which helps to distinguish these components . at the location of the bright arcs ( se1 and nw1 ) , the kinematics are disturbed : the radial velocities split by @xmath0100 km s@xmath3relative to the rotating disk pattern , and the velocity dispersion jumps dramatically in the se1 arc to nearly 300 km s@xmath3 . this can be seen in the 1d radial velocity profile ( figure[fig : halpha_rotcurve ] , middle panel ) and in the 2d spectrum ( figure[fig : lspec ] ) where the h@xmath1 , [ ] and [ ] emission lines are all suddenly broader . the effect is less obvious in the fainter nw1 arc , although in the 2d spectrum there is faint but noticeable broadening of the h@xmath1 and [ ] @xmath156583 lines . these are the regions that @xcite identified with the working surface between the radio outflow and ambient gas in the galaxy . the se2 arc appears to have similar kinematic disturbance as se1 and nw1 in the [ ] velocity map ; however , it is not consistently double - valued in h@xmath1 , like the other two arcs ( see figure[fig : halpha_rotcurve],[fig : oiii5007_rotcurve ] ) . this implies that se2 is not significantly kinematically disturbed as se1 and nw1 are . throughout this analysis , we include calculations for se2 , but our primary focus is on the effect of an outflow as manifested in se1 and nw1 . the spectra of the nucleus and bright , off - nuclear regions are remarkably rich in emission lines , as can be seen in figure[fig : regions ] where we plot the spectra of the nucleus and the brightest off - nuclear regions . of particular interest are the lines of h@xmath1 , h@xmath26 , [ ] @xmath166548,6583 , [ ] @xmath164363,4959,5007 , and [ ] @xmath166716,31 which are used to estimate internal extinction , densities , and temperatures in the gas , and provide diagnostics of shock versus photoionization . in addition , we can observe a number of fainter high - excitation lines of interest , including [ ] @xmath156375 and the [ ] @xmath153426 line . the densities of the emission - line gas are estimated using the [ ] @xmath166716,31 doublet ratio and the theoretical calculations of @xcite . the gas temperatures are calculated using the [ ] @xmath164363,4959 , and 5007 emission lines following the method described by @xcite . because the [ ] emission - line ratio @xmath27 is reddening - sensitive , we estimate the effects of extinction by assuming a simple screen geometry and using the balmer decrement assuming case - b recombination . the small differences between the observed and emitted [ ] line ratios in table[tab : ne_temp_fluxes ] indicate that reddening in these regions is negligible . because the [ ] doublet density method is weakly temperature dependent and the [ ] temperature estimate is weakly density dependent ( see * ? ? ? * ) , we have adopted an iterative approach , whereby we first estimate the [ ] temperature in the low - density limit , evaluate the [ ] density at that temperature , and then use that density to re - evaluate the [ ] temperature , iterating until the solutions converge to within the measurement uncertainties . the densities we find range from 760 - 2500@xmath28 , and the temperatures range between 10000 - 14500k . the resulting temperature and density estimates for each region and their uncertainties are given in table[tab : ne_temp_fluxes ] . the arcs are high - surface brightness emission - line features , but the `` bubbles '' between se 1 , nw 1 , and the nucleus ( b@xmath29 and b@xmath30 respectively ) are very low surface brightness . as the bubbles have only very weak emission lines , we can not estimate the temperature and density using standard metal - line diagnostics . instead , we make an order - of - magnitude estimate of the mean density @xmath31 using the emission measure ( @xmath32 ) derived from the surface brightness of h@xmath26 ( @xmath33 ) , an estimate of the line - of - sight thickness of the bubbles , @xmath34 , and the volume filling factor @xmath35 : @xmath36 we assume that the temperature is roughly constant , and that the bubbles are roughly spherical ( i.e. , as thick as they are wide ) and have unit filling factor . a first estimate assumes a temperature of 15000k , slightly higher than the temperature of the nuclear region , to evaluate the effective recombination coefficient , giving a mean density of @xmath01@xmath28 in the bubbles ( see table[tab : ne_temp_bubbles ] ) . at this temperature and density the gas in the bubbles is out of thermal pressure balance with the gas in the emission - line arcs just outside them by a factor of @xmath0500 - 1500 . the bubbles may instead be more like supernova bubbles with internal temperatures of @xmath37k ; adopting @xmath38k gives emission - measure derived mean densities an order of magnitude larger ( see table[tab : ne_temp_bubbles ] ) , bringing the bubbles into rough thermal pressure balance . the latter temperature is consistent with thermalization of bulk motions with velocities as observed , i.e. @xmath39 km s@xmath3 . while the bubble regions have only very weak emission lines , the nuclear emission spectrum has many high ionization lines that are undetected in other emission - line regions , notably [ ] @xmath156375 , [ ] @xmath156435 , and [ ] @xmath404711 , 4740 ( see figure[fig : regions ] ) . the presence of these lines indicates that the photons emitted from the nucleus have a sufficiently hard spectrum to substantially ionize the surrounding material . some of these lines appear to be double peaked ; in particular [ ] @xmath156375 has a broad , flat top . this feature is best fit by two blended gaussian lines , at @xmath156483 and @xmath156486 , respectively , corresponding to line - of - sight velocities of @xmath4146 and + 92 km s@xmath3 . this line splitting likely represents highly excited gas at the base of the biconical outflow . shock - ionized gas will have different relative emission line strengths than gas photoionized by the active nucleus . collisionally excited uv resonance lines provide the best discriminants between shock and photoionization due to the higher temperatures expected in the shocked gas @xcite . unfortunately , we do not have near - uv spectra of mrk573 and therefore we must use visible - wavelength indicators . to evaluate the importance of shock excitation for the arcs , we compare the fluxes of our observed optical emission lines ( table[tab : regional_feat_fluxes ] ) to the predictions from models for ionizing shocks with precursors , photons traveling ahead of the shock front , and nuclear photoionization calculated by @xcite and @xcite . the shock models cover a range of shock speeds ( @xmath42 km s@xmath3 ) and magnetic field strengths , @xmath43g@xmath44 . we adopt two illustrative photoionization models from @xcite . the first is a power - law photoionization model for optically thick gas with a density of @xmath45@xmath28 and a nuclear continuum with a spectral index of @xmath46 ( where @xmath47 ) . this model covers photon energies from 50kev to 100@xmath48 m with an ionization parameter range of @xmath49 to @xmath50 . the second is the `` @xmath51 '' model that is similar to the first but includes a realistic dust component that accommodates changes in the internal structure of dusty clouds in response to the incident radiation field from @xcite . this second model is for a density of @xmath52@xmath28 , continuum spectral index of @xmath46 , and a metallicity twice the solar abundance . figure [ fig : diagram1 ] shows four emission - line diagnostic diagrams that provide the cleanest discrimination between shock and central photoionization for the lines available in our spectra . the upper two panels show two diagnostics using the [ ] @xmath53line : [ ] /h@xmath26 vs. [ ] / [ ] and [ ] /h@xmath26 vs. [ ] /h@xmath1 . in both cases the data are in best agreement with the simple , @xmath54 photoionization model and appear to disfavor the shock sequences and to a lesser extent the @xmath51 model . however , these results should be viewed with some caution as the [ ] line is difficult to model : in photoionization models the [ ] is formed in a partially ionized zone heated by x - ray photons , while shock models are known to overpredict the [ ] line strength by as much as an order of magnitude ( e.g. * ? ? ? the lower left panel of figure [ fig : diagram1 ] shows the [ ] /h@xmath26 vs. [ ] /h@xmath1 diagnostic in which the photoionization and shock models are essentially orthogonal to one another . here the nucleus and se2 region are in better agreement with the photoionization model , while the nw1 and se1 arc regions are ambiguous between photoionization and a relatively fast shock ( @xmath55 km s@xmath3 ) with a strong magnetic field ( @xmath56g@xmath44 ) , while they are in good agreement with the @xmath51 model . in the last panel we show the [ ] / [ ] vs. [ ] / [ ] diagnostic . these line ratio measurements are similar to the previous panel , in particular the agreement between the nw1 and se1 measurements and the shock models , while the nucleus and se2 region do not agree well with any of the models . one caveat with this diagram is that [ ] may be overpredicted by the models @xcite . while none of these line ratios are ideal discriminants between shock and photoionization , we nevertheless see consistent trends in our analysis that favor photoionization . one interesting aspect of these data is that the se1 and nw1 arc regions , which trace the most morphologically - obvious bowshock structures , are more consistent with shock models than either the nucleus or the se2 region . this also supports our assertion that se2 is not strongly interacting with a biconical outflow . additional far uv data could prove useful to further investigate the strengths of any shocks in these regions ; however , based on these and previous arguments we conclude that the emission line ratios in all four regions are more consistent with photoionization than with shocks . while they themselves are not a clear cut indicator of photoionization , they do support the findings from kinematics and analysis of energetics . our analysis thus far adds additional detail to the picture for the inner regions of mrk573 developed by previous work : a hot biconical outflow emerging from the nucleus that is running into the host galaxy at the location of the bright emission - line arcs . our kinematic data show that this outflow is mildly inclined with respect to the host galaxy disk ( @xmath57 ) , and there is a modest kinematic disturbance at the location of the arcs . the inner part of the outflow between the dense base of the bicone and the arcs has low density ( @xmath58@xmath28 ) the nature of the arcs from previous work has been equivocal . on the one hand , @xcite inconclusively modeled these same emission regions as either pre - existing structures or shock features , while on the other _ hst _ images show dust lanes that extend beyond the ionization cones which led @xcite to suggest that the arcs of mrk573 do not result from bow shocks . our analysis of the densities of the different emission line regions reveal that the arcs and nucleus have similar densities , implying that the arcs were not formed by a fast shock . a fast shock would have slammed into the surrounding material , making it more dense , which we do not observe . the density values support the morphological argument of @xcite that the central photons are ionizing pre - existing dust lanes . in turn , this supports the suggestion of @xcite that any shock from the central agn is too weak and slow to photoionize the surrounding material . analysis of the evacuated regions , the bubbles , compared to the emission line regions also supports central photoionization of dust lanes . if these bubbles were rapidly expanding and compressing the surrounding material , they would shock ionize the arcs . in this scenario we expect a higher pressure inside the bubbles than the arcs to cause the expansion . our calculation of the emission measures ( see sec.[sect : emission_measure ] ) reveals that the bubbles have @xmath0500 - 1500 times less thermal pressure than the arcs , assuming a temperature of 15000 k. when we assume a temperature of @xmath59k , motivated by plausible shock velocities , the thermal pressure in these regions is still only @xmath600.1 - 1 times the amount in the emission line regions ( see table[tab : ne_temp_bubbles ] ) . for the bubbles to be rapidly expanding into the surrounding material , bulldozing it , they need to have higher pressure than their surroundings . over a large range of temperatures , the pressure in the bubbles is simply not large enough for the amount of expansion that would result in arc creation and shock ionization . our spectrophotometric analysis of mrk573 indicates that the emission line regions are likely photoionized by the nuclear source . if there is a shock present , it is not strong enough to form the material into arcs and ionize it , agreeing with the conclusion of @xcite . we now complement these analyses with a detailed investigation of the emission - line region and outflow energetics , quantifying the outflow s interaction with the host galaxy ism . specifically , we estimate the energies and pressures of the emission - line regions ( section [ sec : pressure , etc ] ) and relate them to values calculated for the outflow ( section [ sec : jetprops ] ) . these analyses broadly follow the techniques applied by the @xcite study of mrk78 , and we shall adopt similar notation . by comparing the values of the different energy components of the outflow and emission line regions , we can confirm that the emission line regions are photoionized . the various energies also indicate the importance of mechanical feedback , and the overall strength of the outflow present ( section [ sec : jetstrength ] ) . we also compare our results to those of the analogous emission - line regions studied by @xcite , and demonstrate that mrk573 s emission - line regions are very similar to what is observed in mrk78 . to understand the relationship between ionized and neutral gas in the emission regions , we estimate the values of the various energy components . this is critical for determining what contributes to the kinematics , ionization , and structure , in addition to the relative importance of mechanical feedback on the host . we are able to calculate the energy of the various components by building off of determinations of the flux , mass and lifetimes of the emission line regions . we assume that the portion of a given emission - line region within the slit is representative of the larger region as a whole , and so multiply the flux within the slit by the ratio of the total region area ( @xmath61 ) to the slit region area to approximate the total flux . the slit regions are the width of the slit , 0.2 multiplied by the size in arcseconds each region covers in the pure emission - line profiles ( see sec.[sec : extract ] , table[tab : regional_feat_fluxes ] ) . the physical sizes of the regions that we use are listed in table[tab : jet_nums ] . our region sizes and fluxes are smaller than those measured by @xcite because we can better distinguish the sharp boundaries with our improved resolution ; @xcite had an angular resolution of @xmath60035 fwhm compared to our stis resolution of 01 fwhm . the ionized gas mass in each emission - line region ( @xmath62 ) may be estimated from the h@xmath26 emission - line luminosity and the electron density @xmath63 derived from the [ ] doublet ratio from recombination theory : @xmath64 in this and subsequent equations , the numerical subscript appended to a variable indicates a logarithmic scaling , e.g. @xmath65 represents density in units of @xmath52 @xmath28 . the interaction age of the region , @xmath66 , is estimated as the crossing time , @xmath67 , the amount of time required for a compact parcel of gas expanding outwards at @xmath68 to traverse the distance from the nucleus to the arcs , i.e. @xmath69 where @xmath70 is the angular separation between the nucleus and arcs in units of arcseconds . the emission - line masses for nw1 and se2 are both @xmath71m@xmath72 , while se1 is less massive , @xmath73m@xmath72 . se1 has an interaction age of @xmath74myr , and nw1 is @xmath75myr old . se2 has the largest interaction age , @xmath76myr . this is the expected age progression , in that se2 lies further from the nucleus than se1 and nw1 . however , we must note that se2 does not appear to be strongly interacting with the biconical outflow , based on its lack of kinematic disturbance in h@xmath1 ( see figure[fig : halpha_rotcurve ] ) . we use the interaction age of se2 to determine other properties of the emission - line region to better understand the basic energetics of the region , but it is merely an approximation . the age determinations for all of the regions are consistent with what is seen in mrk78 by @xcite from emission - line structures of comparable scale ( 0.48.2myr ) . the total energy in the emission - line regions , @xmath77 , may be estimated from the product of the interaction age and the luminosity in the [ ] @xmath155007 emission line which measures the nebular cooling power : @xmath78 where @xmath79 is the [ ] line flux in cgs units . each of the regions has total emission energy on the order of @xmath80ergs . there are three energy components of interest in the emission line regions associated with their mechanical and thermal energy content . the translational kinetic energy , @xmath81 , of the outflowing gas may be estimated from the mass of the region and its transverse velocity , @xmath68 , @xmath82 where we use the estimated bicone inclination angle @xmath22 to deproject the observed radial velocity . the internal ( turbulent ) kinetic energy , @xmath83 , depends on the emission - line mass and velocity full - width , @xmath84 : @xmath85 @xcite used the width of the [ ] @xmath86 emission line , but because we have only low - dispersion spectra in which these lines are either unresolved or marginally resolved at their broadest , we instead use the widths of the h@xmath1 and [ ] emission lines . finally the thermal energy , @xmath87 , is @xmath88 where we use the electron temperatures derived from the [ ] emission lines . all of these energies are listed in table[tab : jet_nums ] . in general , the translational kinetic energy in the regions is about 100 times larger than the thermal energy . this is consistent with our earlier suggestion that the gas is not significantly shocked , or if shocked it has had enough time to cool to the @xmath89k temperatures observed . this implies that the acceleration mechanism must be fairly slow and gentle ( the typical gas speeds are @xmath90 km s@xmath3 in these regions , about an order of magnitude larger than the sound speed for @xmath91k gas ) . the internal kinetic energy is roughly 10 times larger than the thermal energy , suggesting that this motion is not supersonic turbulence but instead is bulk motion of the gas along the line of sight that is unresolved at our scales , for example an expansion of order @xmath92 km s@xmath3 along the line of sight at the point of contact between the nuclear outflow and the emission - line gas in the se1 and nw1 arcs . our spectrophotometric analysis of mrk573 suggests that the emission - line regions are photoionized by the active nucleus rather than ionized by fast shocks . we investigate two sources of energy input into these regions . the first is the available photon energy , @xmath93 , from the active nucleus . we use the iras 60@xmath48 m and 100@xmath48 m flux from mrk573 to estimate the nuclear bolometric luminosity by assuming that all photons from the central source are absorbed and re - emitted in the infrared by surrounding dust , which underestimates the actual luminosity . each region s @xmath93 follows from the fraction of nuclear energy that is intercepted by the arc , the covering fraction @xmath94 , over its interaction age ( @xmath66 ) assuming that the regions have unity filling factor @xmath95 for the covering fraction , we assume that the depth of the cloud is approximately the arc thickness , @xmath96 . the depth could in fact be @xmath010 times this size , increasing the photon energy by this same factor . for our purposes , being conservative with our covering fraction is acceptable , as even with this small value , @xmath93 is significantly larger than all other energy sources in the emission regions . the second source of energy is relativistic energy , @xmath97 , stored in the radio outflow , which provides an estimate of the energy input due to expansion of the radio lobes into the circumnuclear gas . @xmath97 depends on the minimum magnetic field strength , @xmath98 , a lower limit found by assuming equipartition between the relativistic particles and the magnetic field @xmath99 @xmath100 where @xmath101 is the radio spectral index , and @xmath102 is the filling factor . the spectral indexes are observed to be @xmath103 for nw1 and @xmath104 for se1 @xcite . for the projected size of the radio outflow working surface we adopt @xmath105 and @xmath106 to be @xmath001 . we adopt a@xmath602 , where a represents the contribution of relativistic energy density by ions in proportion to the contribution by electrons , following @xcite and assume unity filling factor for the relativistic material ( @xmath107 ) . following our calculation of @xmath98 , the available relativistic energy , following @xcite , is thus @xmath108 the estimates for the nw1 and se1 regions are listed in table[tab : jet_nums ] . our calculations of the energies of the different components in the emission line regions and the potential sources indicates that @xmath93 dominates , being @xmath01000 times larger than @xmath97 and about 10 times larger than the total energy in the emission regions , @xmath77 . a complete discussion of the different energy components follows in sec.[sec : jetstrength ] . the energetics of the emission line region and its possible radiative and relativistic sources are important for understanding the host - agn interaction , and how it manifests itself . similarly , examination of the different pressures can reveal whether the relativistic pressure from the radio lobe expansion or the radiative pressure from the central photon source are responsible for the arc structures . following @xcite , we calculate the emission - line region thermal pressure ( @xmath109 ) , the relativistic pressure ( @xmath110 ) , and the radiation pressure ( @xmath111 ) , expressed in dynes @xmath112 : @xmath113 where @xmath114 is the distance from the region to the active nucleus . the thermal pressure , @xmath109 , is based on the temperatures and densities determined from our spectrophotometric analysis . the relativistic pressure , @xmath110 , which results from radio lobe expansion is based on the calculated magnetic field @xmath98 . lastly the radiation pressure from the central photons , @xmath111 , is based on the fir values , similar to our calculation of @xmath93 . we find @xmath110:@xmath109@xmath6015:1 , with @xmath111 contributing negligibly . in the next section we describe our methods for quantifying the properties of the outflow itself , rather than its effect on the emitting material . we combine these two analyses in section[sec : jetstrength ] , and use the results to determine how much influence the nuclear outflow of mrk573 has on its host . beyond analyzing the energies and pressures of the emission line regions , we can quantify parameters of the outflow itself , in particular the velocity and dynamical pressure . this allows us to determine the outflow s strength directly , rather than conjecturing it based upon the behavior of surrounding material . as the outflow does not appear to be affecting se2 , we only determine properties of it with respect to nw1 and se1 . the strength of the outflow is dependent upon its composition . a relativistic outflow , in the form of a jet , with speed @xmath115 approximately equal to the speed of light , is able to both bulldoze and shock - ionize material . for an outflow with both thermal and relativistic material , the thermal component literally weighs down the outflow , slowing it down to @xmath116 km s@xmath3@xcite . thus , outflow speed reflects the amount of thermal and relativistic material present . @xcite have suggested that the outflow of mrk78 is a heavy thermal jet , making it too weak to have significant influence over its surroundings . through our quantitative analysis of the outflow in mrk573 , we can determine whether or not it is also heavy and weak . the approximate speed of the outflow is estimated from the energy and momentum values . an outflow with both material components can contribute to the kinetic and relativistic energy of the emission line regions ; it can also affect the momentum of the emitting material , @xmath117 . thus , we use these values to determine @xmath115 , the speed of the outflow . following @xcite : @xmath118 where @xmath119 is the ratio of kinetic energy to relativistic energy . if we assume the outflow primarily consists of heavy thermal material , making @xmath120 , we calculate velocities of @xmath121 km s@xmath3 ( see table[tab : jet_nums ] ) , which is greater than the observed emission - line regions speeds by factors of @xmath010100 . this supports the suggestion that the outflow in mrk573 is primarily made up of thermal material . a primarily relativistic outflow would have velocities vastly more than @xmath010100 times the emission - line velocities , and we do not calculate speeds of this extent . the thermal component of the outflow drives dynamical pressure , @xmath122 , into its surroundings . following @xcite : @xmath123 where @xmath124 is the area of the outflow , which we measure from the radio maps of @xcite to be @xmath125arcsecond@xmath126 , and we assume that @xmath127 , the force of the outflow , is @xmath128 , the force in the emission line regions , calculated using the momentum and crossing times . we can then compare the ram pressure of the outflow to the pressures measured for the emission regions and evacuated bubbles , finding that it is significantly less than @xmath109 and @xmath110 . following this , we use the dynamical pressure calculation to determine the mach number of the outflow . @xmath129 as listed in table[tab : jet_nums ] , the estimated mach numbers , though uncertain , are very small , on the order of 0.01 , confirming that the outflow is slow and thus , likely dominated by thermal material , similar to the outflow in mrk78 @xcite . a final aspect of our outflow analysis is to determine the amount of mass the outflow transports . this value quantifies the influence of the outflow on its surroundings . the mass flux is : @xmath130 and is @xmath131m@xmath72s@xmath3 based on our estimates above . our spectrophotometric analysis suggested that the outflow in mrk573 is unable to bulldoze material into the emitting arcs . through analysis of the physical properties of the emission regions and the outflow , we can confirm this picture . as shown in section[sec : pressure , etc ] , the relative proportions of each of the energy components is + @xmath93:@xmath77:@xmath97:@xmath132:@xmath133:3000:10:1:0.01 ( see table[tab : jet_nums ] ) , similar to what was found in mrk78 by @xcite . the photon energy e@xmath134 clearly dominates . the emission - line region energy e@xmath135 is @xmath025% of e@xmath134 , confirming that central photoionization alone is sufficient to be the ionizing source of the emission regions . the relativistic and kinetic energies , while both substantially smaller than the emission - line region energy , are comparable to each other . this implies that the slow expansion of the radio source into the surrounding material is able to drive the small amount of kinematic activity in the emission - line regions . also note that se2 , where there is little evidence of interaction between the material and an outflow , has very similar energy magnitudes to se1 and nw1 . despite the fact that nw1 and se1 are in the direct path of the outflow , the regions appear quite similar to non - affected material , implying that an outflow has little influence on the emission - line regions properties . we further investigate the kinematics by looking at the different pressures in the regions . the radio outflow pressure @xmath110 is slightly larger than the ionized gas pressure , @xmath109 , indicating that the small kinematic motions in the emission regions are caused by the gentle expansion due to the radio outflows . our previous analysis of the densities and temperatures of the emission regions and the evacuated bubbles shows that the bubbles have comparable or less thermal pressure than the arcs over a wide range of temperatures . this supports the conclusion that any expansion of the outflows is gentle . the radiative pressure @xmath111 is markedly less than the emission - line region thermal pressure , and therefore makes no significant contribution to the emission - line region kinematics we observe . we also investigate the dynamic pressure contribution from the thermal component of the outflow , which has been suggested as a potentially significant pressure source @xcite . however for mrk573 the dynamic pressure is less than the relativistic pressure . the dynamic pressure would have to be significantly larger than the relativistic pressure or comparable to @xmath109 to be able to push around the host ism significantly . in this case , however , it appears that the expansion is caused primarily by the radio lobes , with the mechanical contribution of the outflow being supplementary at best , and likely negligible . our calculation of the amount of material transported by the outflow also indicates that it has little influence over the host . the outflow can transport @xmath136m@xmath72yr@xmath3 of material . over the calculated lifetimes of the outflows , they have carried at most a tenth of the emission mass . strong jets have been measured carrying as much as @xmath00.6m@xmath72yr@xmath3 @xcite , significantly more than we calculate here , once again pointing towards the general impotence of the outflow in mrk573 . the energy analysis shows that radio outflow contributes primarily to the kinetic energy of the ionized region but not to its ionization , while the pressure analysis shows that the outflow does not strongly expand into the surrounding material . while these estimates have significant uncertainties associated with them , the differences between the estimated energies , pressures , etc . are large enough that we can clarify the relationship between the outflow and host despite this uncertainty . taken together , the data suggest that the host galaxy of mrk573 is photoionized rather than shock - ionized , and that the radio outflow gently shapes the regions but does not otherwise contribute to its ionization or heating significantly . this is similar to what @xcite found for mrk78 . the extensive evidence that the emission regions of mrk573 are pre - existing structures photoionized by a nuclear source also support that the outflow itself is weak and likely thermal . an outflow made up of primarily thermal material is slower and therefore less able to bulldoze and strongly shock surrounding material . earlier work by @xcite on the outflow in ngc1068 has found similar results , suggesting that the radio jets in seyferts differ from those in radio galaxies and quasars primarily in that they are dominated by thermal gas instead of light relativistic material . our analysis of the outflow speeds confirm that mrk573 has a slow outflow , orders of magnitude slower than one dominated by relativistic material . this suggests that the outflow of mrk573 is similar to the weak heavy jet that is incapable of shocking material and being the primary ionization mechanism in mrk78 @xcite . combining the different facets of this quantitative analysis presents a consistent picture of a slow , heavy outflow capable of gently shaping the surrounding host ism into the arcs we see , but not fast enough to substantially heat , displace , or ionize that material . we have examined the influence of the agn in mrk573 on its host galaxy using high angular resolution spectrophotometry from _ hst _ stis . these spectra provide both kinematic and spectrophotometric measurements from which we derive the kinematics and physical state of the gas in the circumnuclear regions associated with an interaction between a nuclear radio outflow and spiral dust arms in the host galaxy . arguments from emission - line diagnostics and thermodynamics ( energy and pressure in the various components ) lead us to conclude that the outflow from the active nucleus in markarian573 does not strongly influence the surrounding ism other than gently sculpting the material into arcs . the dominant source of the heating and ionization in the gas can be entirely explained by photoionization by energetic photons from the active nucleus proper . this is similar to what has been seen in other , well - studied nearby agn outflows with sufficient data to make similar energetics arguments , particularly the work of whittle et al . on mrk78 . in many ways , mrk573 and mrk78 are very similar energetically and morphologically : they have hot , heavy outflows sculpting extended emission - line regions lit - up by ionizing photons from the nucleus . all of the outflow and photoionization energetics taken together suggest that while there is some feedback on the agn s host , it is relatively gentle and insufficient to unbind the host ism or otherwise shutdown star formation by many orders of magnitude . the outflows expand gently into the host and have transported at most a tenth of the mass in the emission regions over their lifetimes . the star formation regions in the inner few kiloparsecs of mrk573 , seen as wound spiral arms outlined by regions @xcite , appear to be undisturbed by the outflow . this has important consequences for agn feedback models in galaxies . specifically , this shows that outflows such as the one in mrk573 are not strong enough to remove material from the host galaxy ; in fact , analyses of low luminosity agn indicate that agn feedback has little influence over the host @xcite . as feedback is frequently invoked to explain different aspects of host - agn evolution , new models must take into account that feedback , at least for agn comparable to local seyferts , has a significantly smaller effect on the host than previously thought . we would like to thank mark whittle for useful discussions . support for this work was provided by nasa through grant number go-9143 from the space telescope science institute , which is operated by the association of universities for research in astronomy , inc . , under nasa contract nas5 - 26555 . lcccc aperture area ( arcsec@xmath137 ) & 0.051 & 0.061 & 0.041 & 0.041 + @xmath153426 & 5.59e-17 & 3.40e-15 & 4.04e-17 & 5.16e-17 + @xmath153727 & 3.11e-16 & 1.51e-15 & 7.09e-16 & 1.28e-16 + @xmath153868 & 1.17e-16 & 1.73e-15 & 2.42e-16 & 5.75e-17 + @xmath153968 & 3.76e-17 & 7.80e-16 & 1.37e-16 & 3e-17 + h@xmath138 @xmath154340 & 5.07e-17 & 8.48e-16 & 1.03e-16 & 3.98e-17 + @xmath154363 & 1.10e-17 & 4.22e-16 & 3.47e-17 & 5.90e-18 + @xmath154686 & 2.87e-17 & 8.96e-16 & 5.33e-17 & 1.63e-17 + h@xmath26 @xmath154861 & 1.08e-16 & 1.85e-15 & 2.53e-16 & 7.24e-17 + @xmath154959 & 4.70e-16 & 7.94e-15 & 9.58e-16 & 3.10e-16 + @xmath155007 & 1.29e-15 & 2.37e-14 & 2.80e-15 & 8.98e-16 + @xmath156300 & 5.32e-17 & 5.37e-16 & 1.02e-16 & 1.81e-17 + h@xmath1 @xmath156563 & 4.48e-16 & 7.14e-15 & 9.76e-16 & 2.58e-16 + @xmath156583 & 4.32e-16 & 3.57e-15 & 9.62e-16 & 1.21e-16 + @xmath156716 & 1.28e-16 & 6.29e-16 & 2.69e-16 & 4.81e-17 + @xmath156731 & 1.31e-16 & 8.64e-16 & 3.24e-16 & 4.3e-17 + ccccc @xmath139 & 0.471 @xmath120.13 & 0.458@xmath120.13 & 0.407@xmath120.11 & 0.549@xmath120.15 + @xmath140 & 160.42@xmath1254 & 74.96@xmath1225 & 108.36@xmath1236 & 204.90@xmath1269 + @xmath141 & 160.59@xmath1254 & 77.17@xmath1226 & 125.53@xmath1242 & - + @xmath142 & 0.979@xmath120.27 & 0.728@xmath120.20 & 0.828@xmath120.23 & 0.981@xmath120.27 + @xmath63 ( @xmath28 ) & 789@xmath1278 & 2437@xmath12243 & 1489@xmath12148 & 762@xmath1276 + @xmath143 ( k ) & 10814@xmath121081 & 14454@xmath121445 & 12445@xmath121244 & 10046@xmath121004 + cccc i(h@xmath26 ) ( erg s@xmath3 @xmath112 arcsec@xmath144 ) & - & 4.1e-16 & 2.4e-16 + depth ( arcsec ) & - & 1.6 & 1.6 + depth ( parsec ) & - & 576 & 576 + emission measure ( @xmath145 pc ) & 15000 & 826 & 483 + @xmath146 ( @xmath28 ) & 15000 & 1.2 & 0.9 + emission measure ( @xmath145 pc ) & @xmath59 & 36210 & 21197 + @xmath146 ( @xmath28 ) & @xmath59 & 7.9 & 6.1 + lccccc region angular size ( x ) & @xmath70 ( arcsec ) & 2.14@xmath120.1 & 0.67@xmath120.1 & 2.55@xmath120.1 & 3.72@xmath120.1 + region angular size ( y ) & @xmath96 ( arcsec ) & 0.31@xmath120.1 & 0.31@xmath120.1 & 0.26@xmath120.1 & 0.31@xmath120.1 + distance from nucleus & @xmath147 ( arcsec ) & 1.81@xmath120.1 & - & 1.76@xmath120.1 & 2.63@xmath120.1 + electron temperature & @xmath143 ( k ) & 10814@xmath121081 & 14454@xmath121445 & 12445@xmath121244 & 10046@xmath121004 + electron density & @xmath63 ( @xmath148 ) & 789@xmath1278 & 2437@xmath12243 & 1489@xmath12148 & 762@xmath1276 + h@xmath26 emission flux & @xmath149 ( erg s@xmath3 @xmath112 ) & -14.8@xmath120.1 & -14.0@xmath120.1 & -14.4@xmath120.1 & -14.9@xmath120.1 + [ ] @xmath155007 emission flux & @xmath150 ( erg s@xmath3 @xmath112 ) & -13.7@xmath120.1 & -12.9@xmath120.1 & -13.4@xmath120.1 & -13.9@xmath120.1 + emission - line width & @xmath151 ( km / s ) & 159@xmath1220 & 217@xmath1220 & 214@xmath1220 & 108@xmath1220 + emission - line velocity & @xmath152 ( km / s ) & 110@xmath1220 & - & -83@xmath1220 & -62@xmath1220 + emission mass & @xmath62 ( @xmath153)@xmath1222@xmath154 & 0.009 & 0.018 & 0.012 & 0.007 + emission region age & @xmath66 ( myr ) & 2.24@xmath120.65 & - & 3.53@xmath121.31 & 6.90@xmath122.50 + emission luminosity & @xmath155 ( erg / s ) & 41.1@xmath120.1 & 41.9@xmath120.1 & 41.4@xmath120.1 & 40.9@xmath120.1 + emission region energy & @xmath156 ( erg ) & 54.97@xmath120.15 & - & 55.46@xmath120.18 & 55.26@xmath120.18 + emission region momentum & @xmath157 ( gm cm s@xmath3 ) & 44.81@xmath120.15 & - & 44.81@xmath120.18 & 44.45@xmath120.18 + emission region force & @xmath158 ( dyne ) & 30.96@xmath120.22 & - & 30.76@xmath120.26 & 30.11@xmath120.26 + thermal energy & @xmath159 ( erg ) & 49.60@xmath120.10 & 50.03@xmath120.10 & 49.79@xmath120.10 & 49.46@xmath120.10 + transverse kinetic energy & @xmath160 ( erg ) & 52.06@xmath120.15 & - & 51.94@xmath120.18 & 51.45@xmath120.18 + internal kinetic energy & @xmath161 ( erg ) & 50.61@xmath120.10 & - & 51.00@xmath120.10 & 50.17@xmath120.10 + total mechanical energy & @xmath162 ( erg ) & 52.07@xmath120.18 & - & 51.99@xmath120.20 & 51.47@xmath120.20 + photon energy & @xmath163 ( erg ) & 55.6@xmath120.1 & - & 55.9@xmath120.2 & 56.1@xmath120.2 + relativistic energy & @xmath164 ( erg ) & @xmath6052@xmath120.17 & - & @xmath6053@xmath120.17 & - + emission region pressure & @xmath165 ( dyne @xmath112 ) & -8.92@xmath120.04 & -8.31@xmath120.04 & -8.59@xmath120.04 & -8.97@xmath120.04 + radiation pressure & @xmath166 ( dyne @xmath112 ) & -10.5@xmath120.06 & - & -10.5@xmath120.06 & -10.9@xmath120.04 + relativistic pressure & @xmath167 ( dyne @xmath112 ) & -8.0@xmath120.14 & - & -7.6@xmath120.14 & - + outflow ram pressure & @xmath168 ( dyne @xmath112 ) & -10.9@xmath12.22 & - & -11.1@xmath12.26 & - + outflow speed & @xmath169 ( km / s ) & 340@xmath12228 & - & 1700@xmath121207 & - + outflow mach number & @xmath170 & 0.03@xmath120.5 & - & 0.01@xmath120.5 & - + magnetic field strength & @xmath171 ( gauss ) & -3.25@xmath120.07 & - & -3.05@xmath120.07 & - + iras 60@xmath48 m flux & @xmath172 ( jy ) & & 1.27@xmath120.15 + iras 100@xmath48 m flux & @xmath173 ( jy ) & & 1.26@xmath120.16 +

_ hubble space telescope _ ( _ hst _ ) observations of the nearby ( 3.22 pc ) , k2 v star eridani have been combined with ground - based astrometric and radial velocity data to determine the mass of its known companion . we model the astrometric and radial velocity measurements simultaneously to obtain the parallax , proper motion , perturbation period , perturbation inclination , and perturbation size . because of the long period of the companion , eri b , we extend our astrometric coverage to a total of 14.94 years ( including the three year span of the _ hst _ data ) by including lower - precision ground - based astrometry from the allegheny multichannel astrometric photometer . radial velocities now span 1980.8 2006.3 . we obtain a perturbation period , p = 6.85 0.03 yr , semi - major axis mas , and inclination = 301 38 . this inclination is consistent with a previously measured dust disk inclination , suggesting coplanarity . assuming a primary mass , we obtain a companion mass _ m _ = 1.55 0.24 _ m_ . given the relatively young age of eri ( myr ) , this accurate exoplanet mass and orbit can usefully inform future direct imaging attempts . we predict the next periastron at 2007.3 with a total separation , = 03 at position angle , p.a . = -27 . orbit orientation and geometry dictate that eri b will appear brightest in reflected light very nearly at periastron . radial velocities spanning over 25 years indicate an acceleration consistent with a jupiter - mass object with a period in excess of 50 years , possibly responsible for one feature of the dust morphology , the inner cavity .

astro-ph0610247

@xmath0 eridani ( = hd 22049 = hip 16537 = hr 1084 = plx 742 ) , with a spectral type of k2v , is one of the nearest solar - type stars with a distance of about 3.2 pc . it is slightly metal - poor ( fe / h = -0.13@xmath1 0.04 , santos 2004 , laws 2003 ) . its proximity makes it a prime target for future extrasolar planet direct imaging efforts . the success of these efforts will depend on knowing exactly where to look , requiring accurate orbital elements for the companion . it will depend on the mass of the planetary companion , and will depend on the age of the system . younger and more massive gas giant planets are predicted to be brighter ( hubbard 2002 ) . if young enough , the intrinsic luminosity of @xmath0 eri b might be greater than its brightness in reflected light . however , planetary mass objects with the age of @xmath0 eri change intrinsic luminosity by a factor of nearly 100 between 1 @xmath8 and 7 @xmath8 . hence the need for a more precise companion mass . @xmath0 eri has been the subject of multiple radial velocity ( rv ) planet searches . walker ( 1995 ) , using measurements spanning 11 years , found evidence for @xmath9 10 yr variation with an amplitude of 15 m s@xmath10 . these results were substantiated by nelson & angel ( 1998 ) using an analysis of the same data set . cumming ( 1999 ) analyzed 11 years of rv data on this star taken at lick observatory and found significant variations with comparable amplitude but with a shorter period of 6.9 years . because of the high level of magnetic activity for @xmath0 eri ( inferred from chromospheric activity ) , these rv variations were largely interpreted as arising from a stellar activity cycle . the mcdonald observatory planet search program ( cochran & hatzes 1999 ) has monitored @xmath0 eri since late 1988 . the mcdonald results in combination with these other surveys , along with data from eso ( endl 2002 ) , confirmed the presence of long period rv variations and demonstrated that the most likely explanation for the observed rv variations was the presence of a planetary companion with a period , p = 6.9 y. details of this analysis are given in hatzes ( 2000 ) . we obtained three years of astrometry with _ hst _ with millisecond of arc precision which we combined with radial velocity data as we have in previous planetary mass studies @xcite . just as in the case of the hatzes ( 2000 ) radial velocity analysis , where less - precise data extended the observation span and allowed a companion detection , we anticipated that less - precise astrometry with a 14 year baseline from the allegheny multichannel astrometric photometer ( map ) astrometry project would improve the astrometric result . gatewood ( 2000 ) reported in a meeting abstract an inclination , i = 46@xmath1 17 and a companion mass @xmath11 obtained with the map data alone . in this study we combine the map data with the _ hst _ data , only to improve the determination of the proper motion of @xmath0 eri . the parallax and proper motion must be removed as accurately as possible to determine the perturbation orbit of @xmath0 eri , which , when combined with an estimate of the mass of @xmath0 eri , will provide the mass of the companion , @xmath0 eri b. this paper presents a mass of the planet orbiting @xmath0 eri discussed in hatzes ( 2000 ) , not the far longer period object inferred from dust - disk morphology @xcite . our mass is derived from combined astrometric and radial velocity data , continuing a series presenting accurate masses of planetary companions to nearby stars . previous results include the mass of gl 876b @xcite and of @xmath12 cancri d @xcite . in section 2 we briefly review the astrometers and discuss the data sets coming from each , and identify our many sources for radial velocities . in section 3 we present the results of extensive spectrophotometry of the astrometric reference stars , information required to correct relative parallax to absolute . in section 4 we briefly discuss our astrometric modeling and the quality of our results as determined by residuals . in section 5 we review our radial velocity data . in section 6 we derive an absolute parallax and relative proper motion for @xmath0 eri , those nuisance parameters that must be removed to determine the perturbation orbital parameters . we finally establish the perturbation orbital parameters and , combined with an estimate of the mass of @xmath0 eri , estimate a mass for @xmath0 eri b. we discuss system age , dust , and companion detectability in section 7 , and summarize our conclusions in section 8 . we used _ hst _ fine guidance sensor 1r ( fgs1r ) to carry out our space - based astrometric observations . nelan ( 2003 ) provides a detailed overview of fgs1r as a science instrument . benedict ( 2002b ) describe the fgs3 instrument s astrometric capabilities along with the data acquisition and reduction strategies used in the present study . we use fgs1r for the present study because it provides superior fringes from which to obtain target and reference star positions @xcite . table [ tbl - loof ] presents a log of _ hst _ fgs observations . epochs 24 contain multiple data sets acquired contiguously , the time span less than a day . each time is that of the first observation within each epoch . each distinct observation set typically contains five measurements of @xmath0 eri . the field was observed at multiple spacecraft roll values , and @xmath0 eri had to be placed in different , non - central locations within the fgs1r fov to accommodate the less than optimal distribution of reference stars . fgs photometric sensitivity depends on location within the fov ( _ e.g. _ benedict 1998 ) and depends on time , as the fgs pmt age . given the faintness of our reference stars , we could not use them to provide a high - precision flat field . hence , we could not extract millimagnitude photometry ( _ c.f . _ benedict 2000a ) to monitor @xmath0 eri stellar activity . the multichannel astrometric photometer ( map ) and associated observation and reduction procedures are described by gatewood ( 1987 ) . the observational program utilizing the map began in 1986 , but reluctance to observe at the low @xmath0 eri declination over the city of pittsburgh delayed initiation of its observation until january of 1989 . despite the reduced precision and rate of successful observation , the field remained on the map program until the installation of new instrumentation early in 2004 ( gatewood 2004 ) . map observations of the brightest stars utilize either a specially filtered 12th channel ( e.g. gatewood & han 2006 ) or a divide by 16 feature on channel 2 , which reduces the count sufficiently for the 16 bit counters . with an r magnitude of 3.0 @xmath0 eri could be placed on channel 2 without counter flooding . the other 10 channels were assigned to the map reference stars noted below . an observation consists of 4 x 11 minute sweeps of the ruling across the field with probe rotations and ruling rotations to reduce systematic error ( gatewood 1987 ) . thus each observation consists of approximately 22 minutes of integration on each axis . table [ tbl - loom ] presents a log of map observations . the radial velocity data used include all data described in hatzes ( 2000 ) , in addition to more recent data from mcdonald observatory and eso . these data now span over 25 years . all sources are listed in table [ tbl - rv ] . briefly , mcdonald phases i , ii , and iii are all data obtained with the 2.7 m smith telescope . the phases correspond ( i ) to early velocities referenced to atmospheric o@xmath13 , ( ii ) velocities obtained with an i@xmath13 cell , and ( iii ) velocities obtained with the mcdonald 2d - coude spectrograph ( tull 1995 ) and an i@xmath13 cell . the only new radial velocity data included in this new study are from the mcdonald observatory phase iii 2.7 m program . they are listed in table [ tbl - p3 ] . note that the errors associated with these data are larger than typically produced by this telescope / spectrograph combination @xcite due to high levels of @xmath0 eri stellar activity . any prior knowledge concerning the fifteen stars included in our reference frame eventually enters our modeling as observations with error , and yields the most accurate parallax and proper motion for the prime target , @xmath0 eri . these periodic and non- periodic motions must be removed as accurately and precisely as possible to obtain the perturbation inclination and size caused by @xmath0 eri b. figure [ fig-1 ] shows the distribution of the ten reference stars in the 130 map @xmath0 eri observation sets ( table [ tbl - loom ] ) . note that the areal coverage is approximately 06@xmath1406 , allowing for the use of relatively bright reference stars , well - distributed around the prime target @xmath0 eri , in contrast to the case for the _ hst _ fgs ( below ) . figure [ fig-2 ] shows the distribution in fgs1r pickle coordinates of the 52 sets of five reference star measurements for the @xmath0 eri field . the arcing pattern is enforced by the requirement that _ hst _ must roll to keep its solar panels fully illuminated throughout the year . to ensure access to all reference stars for every observation set , it was not possible to keep @xmath0 eri ( symbol , @xmath14 ) located in the center of the fgs1r fov . at each epoch we measured each reference stars 2 4 times , and @xmath0 eri 45 times . because the parallax determined for @xmath0 eri is measured with respect to reference frame stars which have their own parallaxes , we must either apply a statistically- derived correction from relative to absolute parallax ( * ? ? ? * yale parallax catalog , ypc95 ) , or estimate the absolute parallaxes of the reference frame stars . in principle , the colors , spectral type , and luminosity class of a star can be used to estimate the absolute magnitude , @xmath15 , and @xmath16-band absorption , @xmath17 . the absolute parallax is then simply , @xmath18 our bandpasses for reference star photometry include : @xmath16 ( from fgs1r ) , and @xmath19 from 2mass . the @xmath19 values have been transformed to the bessell & brett ( 1988 ) system using the transformations provided in carpenter ( 2001 ) . table [ tbl - ir ] lists @xmath20 photometry for the target and reference stars indicated in figures [ fig-1 ] and [ fig-2 ] . figure [ fig-3 ] contains a @xmath21 vs. @xmath22 color - color diagram with reference stars and @xmath0 eri labeled . schlegel ( 1998 ) find an upper limit @xmath17@xmath6 0.1 towards @xmath0 eri . in the following we adopt @xmath23 = 0.0 , but increase the error on reference star distance moduli by 0.1 magnitudes to account for absorption uncertainty . the derived absolute magnitudes are critically dependent on the assumed stellar luminosity , a parameter impossible to obtain for all but the latest type stars using only figure [ fig-3 ] . to confirm the luminosity classes we obtain ucac2 proper motions @xcite for a one - degree - square field centered on @xmath0 eri , and then iteratively employ the technique of reduced proper motion @xcite to discriminate between giants and dwarfs . the end result of this process is contained in figure [ fig-4 ] . we derive absolute parallaxes using our estimated spectral types and luminosity class and @xmath15 values from cox ( 2000 ) . our adopted input errors for distance moduli , @xmath24 , are 0.4 mag for all reference stars ( except ref-2 and -6 , as discussed below ) . contributions to the error are a small but undetermined @xmath17 and errors in @xmath15 due to uncertainties in color to spectral type mapping . we estimate a spectral type for reference star ref-6 only through its apparent magnitude , hence , the larger error in its distance modulus . ref-2 , which in figure [ fig-3 ] straddles the gap between giants and dwarfs , was finally typed k4v , because the @xmath25 ( from modeling of the reference frame ) significantly decreased , using that typing . its input parallax error was also increased . all reference star absolute parallax estimates are listed in table [ tbl - spp ] . individually , no reference star absolute parallax is better determined than @xmath26 = 18% . the average input absolute parallax for the reference frame is @xmath27 mas , a quantity known to @xmath28% ( standard deviation of the mean of fifteen reference stars ) . we compare this to the correction to absolute parallax discussed and presented in ypc95 ( sec . 3.2 , fig . 2 ) . entering ypc95 , fig . 2 , with the galactic latitude of @xmath0 eri , @xmath29 , and average magnitude for the reference frame , @xmath30 , we obtain a correction to absolute of 2.3 mas , considerably different . rather than apply a model - dependent correction to absolute parallax , we introduce our spectrophotometrically - estimated reference star parallaxes into our reduction model as observations with error . the @xmath0 eri reference frame contains fifteen stars and has been measured by two different astrometers , fgs1r and map . the only object in common is @xmath0 eri . from these positional measurements we determine the scale , rotation , and offset plate constants " relative to an arbitrarily adopted constraint epoch for each observation set . as for all our previous astrometric analyses , we employ gaussfit ( jefferys 1988 ) to minimize @xmath25 . the solved equations of condition for the @xmath0 eri field are : @xmath31 @xmath32 @xmath33 @xmath34 for fgs1r data and @xmath35 @xmath36 for the map data . identifying terms , @xmath37 and @xmath38 are the measured coordinates from _ hst _ and the map ; @xmath39 represents the @xmath39 color of each star , estimated from its spectral type , @xmath17 , and @xmath21 color listed in table [ tbl - ir ] ; and @xmath40 and @xmath41 are the lateral color corrections , applied only to fgs1r data . here @xmath42xfx and @xmath42xfy are the cross filter corrections in @xmath37 and @xmath38 , applied to the observations of @xmath0 eri in fgs1r . @xmath43 , @xmath44 , @xmath45 and @xmath46 are scale- and rotation plate constants , @xmath47 and @xmath48 are offsets ; @xmath49 and @xmath50 are proper motions ; @xmath51 is the epoch difference from the mean epoch ; @xmath52 and @xmath53 are parallax factors ; and @xmath54 is the parallax . we obtain the parallax factors from a jpl earth orbit predictor ( standish 1990 ) , upgraded to version de405 . orientation to the sky for the fgs1r data is obtained from ground - based astrometry ( 2mass catalog ) with uncertainties of @xmath55 . histograms of the map residuals ( figure [ fig - maph ] ) indicate per - observation precision of @xmath67 mas . because we are seeking the signature of a perturbation over three times smaller than that per - observation precision , the map data were only used to lower the errors on parallax and proper motion , not to establish any perturbation parameters . as for the fgs data , the optical field angle distortion calibration @xcite reduces as - built _ hst _ telescope and fgs1r distortions with magnitude @xmath56 to below 2 mas over much of the fgs1r field of regard . from histograms of the fgs astrometric residuals ( figure [ fig - fgsh ] ) we conclude that we have obtained correction at the @xmath57 mas level . the reference frame catalogs for map and fgs1r in @xmath58 and @xmath59 standard coordinates ( table [ tbl - pos ] ) were determined with @xmath60 and @xmath61 mas ( map ) , and @xmath62 and @xmath63 mas ( fgs ) . measurements from four planet search groups were included in our modeling . table [ tbl - rv ] lists the source , coverage , technique , number of observations and the rms deviation from the final orbit of these observations . the weighting of the rv data was carefully evaluated with independent modeling and significant outliers were filtered . for example - if five data points were taken in succession , all assigned with the same weight and one point was 50 m s@xmath10 offset from the others , that point was discarded for this solution . initially these points were merely re - weighted , but later discarded as spurious data . the total number of observations so discarded was less than 1% of the aggregate data . this improved our goodness of fit ( @xmath25 /degrees of freedom ) measurement of the modelling for the rv data set from 0.94 in the announcement paper to 0.30 in the current analysis . solving for relative parallax , proper motion , and orbital motion , using astrometry and radial velocities simultaneously , the model now becomes , @xmath64 @xmath65 where orbit is a function ( through thiele - innes constants ) of the traditional astrometric and radial velocity orbital elements listed in table [ tbl - orb ] . the period ( p ) , the epoch of passage through periastron in years ( t ) , the eccentricity ( e ) and the angle in the plane of the true orbit between the line of nodes and the major axis ( @xmath66 ) , are constrained to be equal for the radial velocity and astrometry portions of the model . only radial velocity provides information with which to determine the half - amplitudes ( @xmath67 ) and @xmath68 , the systemic velocity . combining radial velocity observations from different sources is possible with gaussfit , which has the ability to simultaneously solve for many separate velocity offsets ( because velocities from different sources are relative , having differing zero points ) , along with the other orbital parameters . we force a relationship between the astrometry and the radial velocity by a constraint from @xcite @xmath69 where quantities derived only from astrometry ( parallax , @xmath70 , primary perturbation orbit size , @xmath71 , and inclination , @xmath3 ) are on the left , and quantities derivable from both ( the period , @xmath72 and eccentricity , @xmath73 ) , or radial velocities only ( the radial velocity amplitude for the primary , @xmath67 ) , are on the right . combining rv measurements complete through 2006.3 ( table [ tbl - rv ] ) , all the astrometric measurements , and the equation 10 constraint , we solve for parallax , proper motion , and the semi - major axis , orbit orientation , and orbit inclination for the perturbation caused by the companion . for the parameters critical in determining the mass of @xmath0 eri we find a parallax , @xmath74 mas and a proper motion @xmath75 mas y@xmath10 in position angle 2690 @xmath1 06 . table [ tbl - sum ] compares values for the parallax and proper motion of @xmath0 eri from _ hst _ and _ hipparcos_. we note satisfactory agreement . our precision and extended study duration have significantly improved the accuracy and precision of the parallax and proper motion of @xmath0 eri . at this stage we can assess the reality of any @xmath0 eri perturbation by plotting residuals to a model that does not include an orbit . figure [ fig - orbxy ] shows the x and y components of only the higher - precision astrometry fgs residuals plotted as grey dots . the lower precision map data were not considered in the determination of the orbital parameters . we also plot normal points formed from those dots at nine epochs . finally , each plot contains as a dashed line the x and y components of the perturbation we find by including an orbit in our modeling . we find a perturbation size , @xmath76 mas , and an inclination , @xmath77= 301 @xmath1 38 . these , and the other orbital elements for the perturbation , are listed in table [ tbl - orb ] with 1-@xmath78 errors . errors generated by gaussfit ( jefferys 1988 ) come from a maximum likelihood estimation that is an approximation to a bayesian maximum a posteriori estimator with a flat prior ( jefferys 1990 ) . figure [ fig - pjconstemp ] illustrates the pourbaix and jorrisen relation ( equation 10 ) between parameters obtained from astrometry ( left - side ) and radial velocities ( right side ) and our final estimates for @xmath71 and @xmath3 . as seen in tables [ tbl - sum ] and [ tbl - orb ] , most of the errors of the terms in equation 10 are quite small . in essence , our simultaneous solution uses the figure [ fig - pjconstemp ] curve as a quasi - bayesian prior , sliding along it until the astrometric and radial velocity residuals are minimized . gross deviations from the curve are minimized by the high precision of many of the terms in equation 10 . figure [ fig - rvs ] contains all radial velocity measures and the predicted velocity curve from the simultaneous solution . compared to the typical perturbation radial velocity curve ( e.g. hatzes 2005 , mcarthur 2004 , cochran 2004 ) , figure [ fig - rvs ] exhibits far more scatter about the derived orbit . there are two reasons for this . the perturbation amplitude is small ( k1=18.5 m s@xmath10 ) , and @xmath0 eri is an active star , as discussed in hatzes ( 2000 ) . reiterating their conclusions , none of the activity cycles have periods commensurate with the planetary perturbation period . figure [ fig - orbresids ] presents the astrometric residuals and the derived perturbation orbit for the primary star , @xmath0 eri . stellar activity has even less of an effect on astrometry at our level of precision . a star spot covering 30% of the surface would induce a photocenter shift of less than 0.2 mas ( sozzetti 2005 ) . the astrometry confirms the existence of the companion . our analysis of the radial velocities ( now spanning over 25 years , all shown in figure [ fig - rvs ] ) included a linear drift term , a change in velocity as a function of time . this drift is clearly seen in the overplotted final radial velocity curve , and amounts to 0.32 @xmath1 0.05 m s@xmath10 yr@xmath10 . such a change can be caused by longer - period companions and/or secular acceleration @xcite . the secular acceleration expected for @xmath0 eri is ( van de kamp , 1967 ) quite small , 0.07 m s@xmath10 yr@xmath10 . the trend we find is over four times larger than the predicted secular acceleration , and is approximately the acceleration one might find for a planet similar in mass to @xmath0 eri b , but with a 50100 year period . figure 9 shows sufficient enough overlap among the many velocity data sets that the trend is unlikely an artifact due to a mismatch in the center of mass velocity offsets ( discussed above ) obtained for each set . typical offset random error is @xmath79 m s@xmath10 . while this acceleration is not a detection of the longer - period companion ( @xmath80 au ) invoked by quillan & thorndike ( 2002 ) and ozernoy ( 2000 ) to modify the dust distribution as discussed below in section 7 , it may ( with a semimajor axis 1020 au ) be at least partially responsible for the inner cavity in the dust disk distribution imaged by greaves ( 2005 ) . the astrometric motion over 15 years due to this possible tertiary would be of order 3 mas and difficult to separate from proper motion , e.g. , @xcite . the planetary mass depends on the mass of the primary star , for which we have adopted @xmath81 @xcite . for this @xmath82 we find @xmath83 . the companion is clearly an extrasolar giant planet . in table [ tbl - massprm ] the mass value , @xmath84 , incorporates the present uncertainty in @xmath82 . until @xmath0 eri b is directly detected , its radius is unknown . from a review of exoplanet masses and radii @xcite , a radius of @xmath85 seems reasonable . our eccentricity value , @xmath73 = 0.70 , allows for a significant difference in separation between star and exoplanet at apastron compared to periastron . at time of periastron passage , t@xmath86 = 2007.29 , we predict a separation 03 @xmath1 01 at a position angle of -27 . at the next apastron , to occur 2010.71 , the separation should be 18 @xmath1 04 at position angle of 153 . the dominant sources of error for the separations are the eccentricity ( 6% ) and the @xmath0 eri b planet mass ( 15% ) . our accurate mass and orbital parameters for this planetary companion have value for future direct imaging projects . we now know where near @xmath0 eri to look for @xmath0 eri b. we would now like to know when to look , what bandpass is best , and what we can expect to see . as stated previously system age , companion mass , and orbital geometry are critical parameters when estimating visibility . a high level of chromospheric activity is seen for @xmath0 eri ( e.g. gray & baliunas 1995 ) , and is consistent with a relatively young age ; @xmath87 1 gyr ( soderblom & dppen 1989 ) . saffe ( 2005 ) used the calibrations of donahue ( 1993 ) and rocha - pinto & maciel ( 1998 ) ( which corrected the age with an effect from stellar metallicity ) to estimate ages of 0.66 and 0.82 gyr , respectively . henry ( 1986 ) derived from caii lines a value of 0.8 gyr . song ( 2000 ) used li abundances with its position in the h - r diagram and kinematics to derive a value of 0.73 @xmath1 0.2 gyr . di folco ( 2004 ) estimated the age at 0.85 gyr , a value obtained through the measurement of the radius of @xmath0 eri by long - baseline interferometry . their modeling is consistent with a primary mass of @xmath81 , an estimate that weakly depends on measured metallicity , which ranges -0.13 @xmath87 [ fe / h ] @xmath87 -0.06 in the literature . hubbard ( 2002 ) predict the intrinsic luminosity of extrasolar giant planets as a function of mass and age . from their figure 11 , an age of 800 my @xcite , and our planetary mass , @xmath88 , we find for @xmath0 eri b , @xmath89.using the di folco ( 2004 ) t@xmath90=5135 k , their radius , @xmath91 , our parallax , @xmath92 mas , and a bolometric correction , b.c . = -0.27 , from flower ( 1996 ) , we find a difference in bolometric magnitude of @xmath0 eri compared to the sun of @xmath93 . hence , neglecting reflected light and orbital phase , @xmath0 eri b is 4.67@xmath94 fainter in bolometric luminosity than @xmath0 eri . sudarsky ( 2005 ) , dyudina ( 2005 ) , and burrows ( 2004 ) discuss exoplanet apparent brightness in reflected host star light as functions of orbit geometry , orbital phase , and cloud cover . burrows ( 2004 ) predict the full spectrum of @xmath0 eri b from 0.5 to 6@xmath95 , asserting that the planet is too young for its atmosphere to contain condensed ammonia clouds . however , @xmath0 eri b should exhibit h@xmath13o clouds . they predict a maximum planet / host star flux ratio , log@xmath96 , at @xmath97 with a secondary peak , log@xmath98 , at @xmath99 . dyudina ( 2005 ) predict that for @xmath100 , inclination , @xmath101 , @xmath73 = 0.5 , and a jupiter - like atmosphere , the planet / host star flux ratio is largest very shortly after periastron , late 2007 . however , the separation remains small ( @xmath603 ) . the inclination of the @xmath0 eri system , @xmath101 , is likely to decrease the flux ratio by approximately a factor of two ( sudarsky 2005 ) , compared to a @xmath102 edge - on orientation . given the orientation of the orbit of @xmath0 eri b ( its ascending node , @xmath103 = 254 ) , the disk of @xmath0 eri b is most fully illuminated at apastron , but is three times further away from its primary . the dusty rings or debris disks surrounding @xmath0 eri also suggest relative youth for the system . photometric measurements from the iras satellite ( aumann , 1988 ) provided the first hint of dust around @xmath0 eri . subsequently , submillimeter common - user bolometric array ( scuba ) measurements were made between 1997 and 2002 . these measurements determined that the dust , distributed in a ring , is located 65 au from the star @xcite . the sub - mm bolometer , simba , provided observational confirmation of this extended dust disk ( schutz 2004 ) . the stis ccd camera on _ hst _ took deep optical images around @xmath0 eri in an effort to find an optical counterpart for the sub - millimeter observations . these measurements did not provide clear evidence for the detection of that optical counterpart , but did place a limit on the optical surface brightness of the dust , that it could not be brighter than approximately 25 stmag arcsec@xmath104 , which places constraints on the nature and amount of the smallest dust grains ( proffit 2004 ) . observational and theoretical searches for the signature of planetary / brown dwarf objects in the structure of the dust disk around @xmath0 eri are underway . clumps seen in the ring are thought to come from the interaction between the disk and a massive planetary body ( holland 2003 ) . adaptive optics on the keck telescope were used to search for extrasolar planets . these studies found no evidence of brown dwarf or planetary companions down to 5 jupiter masses at the angular separations comparable to that of the dust rings ( macintosh 2003 ) . spitzer space telescope ( sst ) observations made with the multi - band imaging photometer ( mips ) and the infrred spectograph ( irs ) have confirmed the disk and provided evidence for asymmetries in the structure of the disk that may have been caused by the gravitational perturbation of sub - stellar companions ( marengo 2004 ) . two recent studies suggested that debris disks and long - period planets co - exist , with planetary bodies sculpting the disk . @xmath0 eri is the prototypical system . first , high - resolution modeling of the structure of the disk around @xmath0 eri predict an angular motion of the asymmetry of the disk of about 06 - 08 yr@xmath10 ( ozervnoy 2000 ) . secondly , quillen & thorndike ( 2002 ) carried out numerical simulations of dust particles captured in mean motions resonances with a hypothetical planet ( e = 0.3 , m = @xmath105 m@xmath106 , a = 40 au ) at periastron . these produced a dust distribution that agreed with the morphology of the dust ring around @xmath0 eri presented by greaves ( 1998 , 2005 ) . an investigation into the dynamics of the dust ring around @xmath0 eri ( moran , kuchner & holman , 2004 ) concluded that the eccentricity of the dust released in the inner ring ( @xmath8720 au ) could reveal patterns in the dust which could confirm the existence of the planet reported by hatzes ( 2000 ) . we determined an inclination of @xmath3=301 @xmath1 32 for @xmath0 eri b. our measured inclination is consistent with the previously measured dust disk inclination from 450 and 850 @xmath107 m maps of @xcite , @xmath108 . this suggests that the dust disk and plane of the orbit of @xmath0 eri b are coincident and that the dust distribution is nearly circular . this provides support for hiearchical accretion models for planet formation @xcite , where coplanar dust and a debris disk are expected remnants of planet formation ( tsiganis 2005 ) . lastly , @xmath0 eri b and the possible tertiary deduced from the linear trend in the radial velocities ( section 6 ) would most likely eject particles that would spiral inward , and recent scuba submillimeter observations have shown that the center of the disk is relatively excavated of dust , with half or less of the signals seen in the ring ( greaves 2005 ) . analyzing three years of _ hst _ fgs and over 14 years of allegheny observatory map astrometry , we find an independently determined parallax and proper motion for @xmath0 eri that agree within the errors with hipparcos . astrometric observations with _ hst _ fgs , combined with long - duration allegheny map astrometry and ground based radial velocities , have confirmed the existence of the planet orbiting @xmath0 eri , first suggested by walker ( 1995 ) , noted by cummings ( 1999 ) , and finally announced by hatzes ( 2000 ) . combining the astrometry with radial velocities from six different sources , spanning 25 years , and applying the pourbaix & jorrisen constraint between astrometry and radial velocities , we obtain for the perturbing object @xmath0 eri b a period , p=6.85 @xmath1 0.02 y , inclination , @xmath3=301 @xmath1 32 , and perturbation semimajor axis , @xmath76 mas . assuming for @xmath0 eri a stellar mass @xmath82 = 0.83@xmath10.05@xmath109 , we obtain a mass for @xmath0 eri b , @xmath84 = 1.55@xmath10.24@xmath8 . this companion inclination matches the disk inclination determined by greaves ( 2005 ) . our astrometry predicts for @xmath0 eri b periastron passage at t@xmath86 = 2007.29 , with a separation @xmath603 in position angle -27(a = 3.39 au ) . at the next apastron , to occur 2010.71 , the separation should be 17 in position angle 153 . the orbital geometry suggests that 2007.97 ( late december 2007 ) is the most favorable time for direct detection in reflected light . for an @xmath0 eri age @xmath6850 my and our determined mass , @xmath0 eri b will have an intrinsic luminosity @xmath89 , 4.67@xmath94 times fainter in bolometric luminosity than @xmath0 eri . radial velocities spanning 25 years indicate a long - term linear trend , an acceleration consistent with a jupiter - mass object with a period of 50100 years . this is a possible detection of a tertiary companion responsible for a major feature of the dust morphology , the central cavity . support for this work was provided by nasa through grants go-09167 , go-09347 , and go-09969 from the space telescope science institute , which is operated by the association of universities for research in astronomy , inc . , under nasa contract nas5 - 26555 . the allegheny observatory received support from the national science foundation through grant ast-0098552 and the national aeronautics and space administration through grant nag-510628 . this publication makes 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 . this research has made use of the simbad database , operated at cds , strasbourg , france ; the nasa / ipac extragalactic database ( ned ) which is operated by jpl , california institute of technology , under contract with the nasa ; and nasa s astrophysics data system abstract service . we thank an anonymous referee for a careful review and constructive criticism which improved this final version of the paper . mcarthur , b. , benedict , g. f. , jefferys , w. h. , & nelan , e. 2002 , the 2002 hst calibration workshop . edited by santiago arribas , anton koekemoer , and brad whitmore . baltimore , md : space telescope science institute , p 373 rlllrlll 1&51946.528&2001.101&96.9&24&52862.76878&2003.609&281.0 + 2&51982.209&2001.198&91.0&25&52865.68038&2003.617&280.0 + 3&52180.393&2001.741&280.0&26&52868.70565&2003.625&280.0 + 4&52309.062&2002.093&97.0&27&52871.75072&2003.634&280.0 + 5&52497.787&2002.610&280.0&28&52874.84206&2003.642&280.0 + 6&52500.124&2002.616&280.0&29&53007.24013&2004.005&106.0 + 7&52503.062&2002.624&280.0&30&53010.26367&2004.013&105.0 + 8&52506.667&2002.634&280.0&31&53014.10664&2004.024&105.0 + 9&52509.629&2002.642&280.0&32&52672.409&2003.088&105.0 + 10&52513.545&2002.653&280.0&33&52679.324&2003.107&105.0 + 11&52537.069&2002.718&280.0&34&52681.993&2003.114&105.0 + 12&52540.651&2002.727&280.0&35&52682.004&2003.114&105.0 + 13&52543.679&2002.736&280.0&36&52683.127&2003.117&105.0 + 14&52642.962&2003.007&105.0&37&52686.352&2003.126&105.0 + 15&52645.364&2003.014&105.0&38&52688.330&2003.132&105.0 + 16&52649.703&2003.026&105.0&39&52862.769&2003.609&281.0 + 17&52672.40928&2003.088&105.0&40&52865.680&2003.617&280.0 + 18&52679.32437&2003.107&105.0&41&52868.706&2003.625&280.0 + 19&52681.9927&2003.114&105.0&42&52871.751&2003.634&280.0 + 20&52682.00362&2003.114&105.0&43&52874.842&2003.642&280.0 + 21&52683.12669&2003.117&105.0&44&53007.240&2004.005&106.0 + 22&52686.35236&2003.126&105.0&45&53010.264&2004.013&105.0 + 23&52688.32956&2003.132&105.0&46&53014.107&2004.024&105.0 + + rllrll 1&47544.0781&1989.0474&66&53001.0760&2003.988 + 2&47772.4045&1989.6726&67&50362.2622&1996.763 + 3&47790.3559&1989.7217&68&50362.2948&1996.763 + 4&47822.2920&1989.8091&69&50362.3281&1996.763 + 5&47826.2622&1989.8200&70&50371.2538&1996.788 + 6&47829.2503&1989.8282&71&50371.2872&1996.788 + 7&47905.0385&1990.0357&72&50371.3184&1996.788 + 8&47923.0024&1990.0849&73&50719.3260&1997.741 + 9&47933.0142&1990.1123&74&50719.3594&1997.741 + 10&48209.2163&1990.8685&75&50736.3129&1997.787 + 11&48213.2052&1990.8794&76&50736.3448&1997.787 + 12&48234.1406&1990.9367&77&50741.3281&1997.801 + 13&48281.0559&1991.0652&78&50798.1108&1997.956 + 14&48290.0135&1991.0897&79&50799.1316&1997.959 + 15&48528.3413&1991.7422&80&50799.1649&1997.959 + 16&48547.2948&1991.7941&81&51079.3615&1998.727 + 17&48901.3066&1992.7633&82&51079.3934&1998.727 + 18&48915.2816&1992.8016&83&51079.4253&1998.727 + 19&49007.9990&1993.0554&84&51100.3219&1998.784 + 20&49022.0066&1993.0938&85&51100.3538&1998.784 + 21&49027.0066&1993.1075&86&51145.1372&1998.907 + 22&49268.2733&1993.7680&87&51145.1698&1998.907 + 23&49274.2691&1993.7844&88&51145.2024&1998.907 + 24&49285.2309&1993.8145&89&51209.0010&1999.081 + 25&49302.2240&1993.8610&90&51829.2788&2000.780 + 26&49312.1545&1993.8882&91&51829.3115&2000.780 + 27&49334.1448&1993.9484&92&51829.3434&2000.780 + 28&49372.9997&1994.0548&93&52183.3115&2001.749 + 29&49597.4087&1994.6692&94&52183.3448&2001.749 + 30&49600.4003&1994.6773&95&52185.2774&2001.754 + 31&49640.3052&1994.7866&96&52219.1969&2001.847 + 32&49653.2205&1994.8220&97&52219.2316&2001.847 + 33&49668.1733&1994.8629&98&52219.2628&2001.847 + 34&49708.0587&1994.9721&99&52219.3080&2001.848 + 35&49747.0052&1995.0787&100&52219.3413&2001.8476 + 36&49748.0476&1995.0816&101&52220.2115&2001.8500 + 37&50362.2622&1996.7632&102&52220.2455&2001.8501 + 38&50371.2538&1996.7878&103&52225.1816&2001.8636 + 39&50719.3260&1997.7408&104&52225.2149&2001.8637 + 40&50736.3129&1997.7873&105&52225.2483&2001.8638 + 41&50741.3281&1997.8010&106&52226.1899&2001.8664 + 42&50798.1108&1997.9565&107&52226.2233&2001.8665 + 43&50799.1316&1997.9593&108&52226.2573&2001.8665 + 44&51079.3615&1998.7265&109&52227.2226&2001.8692 + 45&51100.3219&1998.7839&110&52227.2552&2001.8693 + 46&51145.1372&1998.9066&111&52265.0691&2001.9728 + 47&51209.0010&1999.0815&112&52265.1024&2001.9729 + 48&51829.2788&2000.7797&113&52265.1351&2001.9730 + 49&52183.3115&2001.7490&114&52893.3497&2003.6929 + 50&52185.2774&2001.7544&115&52893.383&2003.6930 + 51&52219.1969&2001.8472&116&52924.2726&2003.7776 + 52&52219.3080&2001.8475&117&52924.3059&2003.7777 + 53&52220.2115&2001.8500&118&52925.2747&2003.7804 + 54&52225.1816&2001.8636&119&52925.3094&2003.7805 + 55&52226.1899&2001.8664&120&52925.342&2003.7805 + 56&52226.2573&2001.8665&121&52925.3774&2003.7806 + 57&52227.2226&2001.8692&122&52937.2087&2003.8130 + 58&52265.0691&2001.9728&123&52937.2413&2003.8131 + 59&52893.3497&2003.6929&124&52946.224&2003.8377 + 60&52924.2726&2003.7776&125&52946.2566&2003.8378 + 61&52925.2747&2003.7804&126&52946.2899&2003.8379 + 62&52925.3420&2003.7805&127&53000.058&2003.9851 + 63&52937.2087&2003.8130&128&53001.076&2003.9879 + 64&52946.2240&2003.8377&129&53001.1177&2003.9880 + 65&53000.0580&2003.9851&130&53001.1448&2003.9881 + + lllcc cfht & 1980.81 - 1991.88 & hf cell & 48 & 10.5 + lick & 1987.69 - 1998.99 & iodine cell & 54 & 11.5 + mcd @xmath110i & 1988.74 - 1994.81 & telluric & 27 & 15.2 + mcd @xmath110ii & 1990.78 - 1998.07 & iodine cell & 42 & 11.7 + eso & 1992.84 - 1998.02 & iodine cell & 36 & 9.6 + mcd @xmath110iii & 1998.69 - 2004.86 & iodine cell & 28 & 7.4 + & & total&235 & lrlr 51066.4339&-3.6@xmath16.4&52539.4226&-16.4@xmath16.4 + 51212.1671&13.6 5.0&52576.4378&2.0 5.5 + 51239.1133&1.7 8.6&52661.1445&-11.8 5.2 + 51449.4333&6.5 5.0&52931.3912&-13.9 6.4 + 51503.3574&14.7 5.1&52958.2449&-0.4 5.4 + 51529.1986&-3.5 7.4&52958.2481&0.0 6.0 + 51555.1645&-0.2 5.2&53016.2424&2.9 6.0 + 51775.4643&1.3 5.9&53016.2456&2.1 6.0 + 51809.4033&-2.7 4.8&53035.1757&-5.2 5.6 + 51917.2260&2.9 5.5&53075.0940&3.0 5.7 + 51984.0740&-2.8 5.0&53318.3124&-10.1 6.6 + 52142.4227&-0.7 6.2&53632.4520&-1.2 5.3 + 52142.4264&-0.4 6.1&53632.4550&0.4 5.5 + 52248.2921&-14.0 5.9&53689.3900&-3.0 5.2 + 52303.1266&-16.1 7.7&53745.2110&6.3 5.1 + 52328.1208&2.1 5.6&53809.0850&9.0 5.5 + 52330.1064&-0.9 5.7 & & + rlllll 1&@xmath0 eri&53.232961&-9.458295&3.82&03325591 - 0927298 + 2&&53.325845&-9.467569&15.61&03331820 - 0928032 + 3&&53.312638&-9.434314&15.58&03331503 - 0926035 + 4&&53.295306&-9.421752&16.12&03331087 - 0925183 + 5&&53.269681&-9.418811&16.41&03330472 - 0925077 + 6&&53.207917&-9.449667&16.48 & + 7&gen # + 6.10280864&53.374062&-9.504524&11.36&03332977 - 0930162 + 8&hd 22130&53.401866&-9.349848&9.52&03333644 - 0920594 + 9&bd-09 696 & 53.164328&-9.267778&10.69&03323943 - 0916040 + 10&gen # + 6.10280861 & 53.032108&-9.514245&11.60&03320770 - 0930512 + 11&bd-10 699 & 53.399013&-9.586081&10.24&03333576 - 0935098 + 12&hd 21951 & 53.001248&-9.385150&9.69&03320029 - 0923065 + 13&bd-10 695 & 53.320908&-9.696887&9.78&03331701 - 0941487 + 14&bd-10 700 & 53.466004&-9.641330&9.94&03335184 - 0938287 + 15&bd-09 699 & 53.429559&-9.232722&11.20&03334309 - 0913577 + 16&2mass 03320556 - 0945292&53.023188&-9.758118&11.15&03320556 - 0945292 + cccccc 1&3.82@xmath10.01&1.82@xmath10.05&0.40@xmath10.28&0.48@xmath10.07&2.00@xmath10.05 + 2&15.61@xmath10.03&13.01@xmath10.03&0.58@xmath10.04&0.72@xmath10.04&2.60@xmath10.04 + 3&15.58@xmath10.03&14.09@xmath10.06&0.41@xmath10.05&0.44@xmath10.07&1.50@xmath10.06 + 4&16.12@xmath10.03&13.88@xmath10.05&0.51@xmath10.04&0.59@xmath10.06&2.24@xmath10.06 + 5&16.41@xmath10.03&14.21@xmath10.08&0.46@xmath10.07&0.62@xmath10.09&2.20@xmath10.08 + 6&16.48 & & & & + 7&11.36@xmath10.03&9.93@xmath10.02&0.29@xmath10.03&0.33@xmath10.03&1.43@xmath10.04 + 8&9.52@xmath10.01&8.63@xmath10.02&0.21@xmath10.06&0.24@xmath10.03&0.89@xmath10.02 + 9&10.69@xmath10.03&8.85@xmath10.02&0.48@xmath10.04&0.54@xmath10.03&1.84@xmath10.04 + 10&11.60@xmath10.03&10.43@xmath10.02&0.33@xmath10.03&0.41@xmath10.03&1.17@xmath10.0 4 + 11&10.24@xmath10.03&8.73@xmath10.02&0.40@xmath10.06&0.43@xmath10.04&1.51@xmath10.04 + 12&9.69@xmath10.03&8.98@xmath10.02&0.13@xmath10.03&0.18@xmath10.03&0.71@xmath10.04 + 13&9.78@xmath10.01&7.87@xmath10.02&0.47@xmath10.03&0.55@xmath10.02&1.91@xmath10.02 + 14&9.94@xmath10.03&7.69@xmath10.03&0.57@xmath10.06&0.66@xmath10.04&2.25@xmath10.04 + 15&11.20@xmath10.03&8.13@xmath10.03&0.62@xmath10.05&0.74@xmath10.03&3.07@xmath10.04 + 16&11.15@xmath10.03&9.43@xmath10.02&0.45@xmath10.03&0.49@xmath10.03&1.72@xmath10.04 + lccccc 2&k4v&15.6&7.1&8.5@xmath11&1.9@xmath11.0 + 3&g8v&15.6&5.6&10.0@xmath10.4&1.0@xmath10.2 + 4&k2v&16.1&6.5&9.6@xmath10.4&1.2@xmath10.2 + 5&k2v&16.4&6.5&9.9@xmath10.4&1.0@xmath10.2 + 6&k2v&16.4&6.5&9.9@xmath12&1.0@xmath10.9 + 7&g0v&11.4&4.4&7.0@xmath10.4&4.1@xmath10.7 + 8&f5v&9.5&3.5&6.0@xmath10.4&6.3@xmath11.2 + 9&k0v&10.7&5.9&4.8@xmath10.4&11.0@xmath12.0 + 10&g5v&11.6&5.1&6.5@xmath10.4&5.0@xmath10.9 + 11&g8v&10.2&5.6&4.6@xmath10.4&11.8@xmath12.2 + 12&f0v&9.7&2.7&7.0@xmath10.4&4.0@xmath10.7 + 13&k0v&9.8&5.9&3.9@xmath10.4&16.7@xmath13.1 + 14&k0iii&9.9&0.7&9.2@xmath10.4&1.4@xmath10.3 + 15&k2iii&11.2&2.7&8.5@xmath10.4&2.0@xmath10.4 + 16&g8v&11.2&5.6&5.6@xmath10.4&7.8@xmath11.4 + + cccc 1&3.73&297.0383@xmath10.0001&35.6893@xmath10.0001 + 2&15.57&-25.6479@xmath10.0002&126.1290@xmath10.0001 + 3&15.57&0.0000@xmath10.0002&0.0000@xmath10.0002 + 4&16.09&52.8710@xmath10.0003&-55.1822@xmath10.0002 + 5&16.40&433.1267@xmath10.0003&-20.0260@xmath10.0003 + 6&16.37&140.6876@xmath10.0004&-81.2800@xmath10.0002 + 1&3.73&715.2729@xmath10.0002&201.7543@xmath10.0002 + 7&11.36&1214.0919@xmath10.0013&35.0491@xmath10.0014 + 8&9.52&1313.1141@xmath10.0005&591.8986@xmath10.0007 + 9&10.69&469.2508@xmath10.0011&887.5650@xmath10.0014 + 10&11.60&0.0000@xmath10.0021&0.0000@xmath10.0025 + 11&10.24&1302.5205@xmath10.0007&-258.4640@xmath10.0011 + 12&9.69&-109.9029@xmath10.0008&464.7282@xmath10.0008 + 13&9.78&1025.1161@xmath10.0005&-657.0894@xmath10.0007 + 14&9.94&1540.2424@xmath10.0005&-457.4732@xmath10.0006 + 15&11.2&1411.6967@xmath10.0011&1013.5480@xmath10.0018 + 16&11.15&-31.1988@xmath10.0015&-877.8886@xmath10.0023 + cccccc 2&15.57&0.0053@xmath10.0078&-0.0112@xmath10.008&0.0110@xmath10.0002&- 0.0076@xmath10.0002 + 3&15.57&0.0166@xmath10.0076&-0.0014@xmath10.0076&0.0072@xmath10.0003&- 0.0026@xmath10.0002 + 4&16.09&0.0044@xmath10.009&-0.0028@xmath10.0079&0.0080@xmath10.0004&- 0.0022@xmath10.0002 + 5&16.4&-0.0004@xmath10.004&-0.0044@xmath10.004&0.0070@xmath10.0004 & 0.0082@xmath10.0002 + 6&16.37&0.0064@xmath10.003&0.0083@xmath10.003&0.0014@xmath10.0004&-0.0030@xmath10 . 0003 + 7&11.36&0.0253@xmath10.005&0.0117@xmath10.005&0.0221@xmath10.0002&0.0083@xmath10 . 0003 + 8&9.52&0.0104@xmath10.005&-0.0025@xmath10.005&0.0088@xmath10.0001&- 0.0052@xmath10.0001 + 9&10.69&0.003@xmath10.005&-0.0178@xmath10.005&0.0035@xmath10.0002&- 0.0184@xmath10.0003 + 10&11.6&-0.0015@xmath10.005&-0.0074@xmath10.005&-0.0020@xmath10.0004&- 0.0029@xmath10.0004 + 11&10.24&0.0101@xmath10.005&-0.0094@xmath10.005&0.0126@xmath10.0001&- 0.0089@xmath10.0002 + 12&9.69&0.0095@xmath10.005&-0.0038@xmath10.005&0.0109@xmath10.0001&- 0.0074@xmath10.0002 + 13&9.78&0.0549@xmath10.005&-0.037@xmath10.005&0.0579@xmath10.0001&- 0.0326@xmath10.0001 + 14&9.94&-0.0003@xmath10.005&-0.0079@xmath10.005&-0.0017@xmath10.0001&- 0.0102@xmath10.0001 + 15&11.2&0.0117@xmath10.005&-0.0063@xmath10.005&0.0108@xmath10.0002&- 0.0042@xmath10.0003 + 16&11.15&0.0073@xmath10.005&-0.0136@xmath10.005&0.0052@xmath10.0003&- 0.0158@xmath10.0004 + rlll study duration & 2.92 y & 14.94 y & + number of observation sets & 46 & 130 & + reference star @xmath111 & 16.0 & 10.2 & 12.2 + reference star @xmath112 & & & @xmath113 + absolute parallax & & & 311.37 @xmath1 0.11 mas + relative proper motion & & & @xmath114 mas y@xmath10 + in pos . angle & & & 2690 @xmath1 06 + absolute parallax & & & 310.74 @xmath1 0.85 mas + _ hipparcos _ proper motion & & & 976.52 @xmath1 1.9 mas y@xmath10 + in pos . angle & & & 2711 @xmath1 38 + ll @xmath71&1.88 @xmath1 0.20 mas + @xmath115 & 3.02e-3 @xmath1 0.32e-3 au + p & 2502 @xmath1 10 d + p & 6.85 @xmath1 0.03 yr + t@xmath86 & 54207 @xmath17 mjd + t@xmath86 & 2007.29 @xmath1 0.02 y + e & 0.702 @xmath1 0.039 + i & 301 @xmath1 38 + @xmath103&74@xmath1 7 + @xmath66&47@xmath1 3 + @xmath67&18.5 @xmath10.2 m s@xmath10 + @xmath82 & 0.83@xmath10.05@xmath109 + lr ( au ) & 3.39 @xmath1 0.36 + @xmath103 & 254 + @xmath66 & 47 + mass function ( @xmath109 ) & 5.9e-10 @xmath1 1.0e-10 + @xmath116(m@xmath117 ) & 0.78 @xmath1 0.08 + m ( m@xmath117 ) & 1.55 @xmath1 0.22 + m ( m@xmath117 ) & 1.55 @xmath10.24 +

we extend and apply a model - independent analysis method developed earlier by daly & djorgovski to new samples of supernova standard candles , radio galaxy and cluster standard rulers , and use it to constrain physical properties of the dark energy as functions of redshift . similar results are obtained for the radio galaxy and supernova data sets , which rely upon completely independent methods , suggesting that systematic errors are relatively small for both types of distances ; distances to sz clusters show a scatter which can not be explained by the quoted measurement errors . the first and second derivatives of the distance are compared directly with predictions in a standard model based on general relativity . the good agreement indicates that general relativity provides an accurate description of the data on look - back time scales of about ten billion years . the first and second derivatives are combined to obtain the acceleration parameter , assuming only the validity of the robertson - walker metric , independent of a theory of gravity and of the physical nature of the dark energy . the data are analyzed using a sliding window fit and using fits in independent redshift bins . the acceleration of the universe at the current epoch is indicated by the sliding window fit analysis . the effect of non - zero space curvature on is explored ; for a plausible range of values of the effect is small and causes a to shift to the redshift at which the universe transitions from deceleration to acceleration . we solve for the pressure , energy density , equation of state , and potential and kinetic energy of the dark energy as functions of redshift assuming that general relativity is the correct theory of gravity . results obtained using a sliding window fit indicate that a cosmological constant in a spatially flat universe provides a good description of each of these quantities over the redshift range from zero to about one . we define a new function , the dark energy indicator , in terms of the first and second derivatives of the coordinate distance and show how this can be used to measure deviations of from and to obtain a new and independent measure of .

0710.5345

understanding of the physical nature of the dark energy which appears to be driving the accelerated expansion of the universe is among the most pressing and important topics in cosmology today . studies of the expansion history of the universe allow us to constrain the physical nature of its matter and energy constituents . one way that the expansion and acceleration history of the universe can be studied is through the use of a set of coordinate distances and redshifts for some standard set of objects . type ia supernovae provide a modified standard candle ( e.g. phillips 1993 , hamuy et al . 1995 ) that allow the distance modulus , luminosity distance , and coordinate distance to each source to be determined . the recent data sets presented by astier et al . ( 2006 ) , riess et al . ( 2007 ) , wood - vasey et al . ( 2007 ) , and davis et al . ( 2007 ) have been analyzed by these groups and compared with numerous models by other researchers . in a novel , largely model - independent approach to this problem , it was shown by daly & djorgovski ( 2003 ) that the first and second derivatives of the coordinate distance with respect to redshift could be obtained from the coordinate distances and combined to solve for the expansion rate @xmath5 and acceleration rate @xmath0 of the universe . the functions @xmath6 and @xmath7 are exact , that is , they are not obtained by expansions in terms of derivatives about some point . the only assumptions are that the universe is described by a robertson - walker metric and has zero space curvature . the results are independent of the contents of the universe and their physical properties , and even independent of whether general relativity provides an accurate description of the universe . here , we drop the assumption of zero space curvature ; it turns out that the deceleration parameter at a redshift of zero , @xmath8 , remains the same , independent of whether space curvature is zero or not . in this paper we expand on the previous analysis done by daly & djorgovski ( 2003 , 2004 ) . first , we use updated and expanded data sets , as described in section 2.1 . second , we introduce a more direct way to compare the model - independent results obtained from the data with predictions ; this is done by directly comparing the first and second derivatives of the coordinate distance with respect to redshift to predicted values in various models , as described in section 2.2 . third , we analyze the data using both a sliding window fit and fits in independent redshift bins . to solve for the physical properties of the dark energy as functions of redshift , a theory of gravity must be specified . to determine the properties of the dark energy , general relativity is taken to be the correct theory of gravity , allowing us to solve for the pressure , energy density , and equation of state of the dark energy as a function of redshift in section 2.4 . fourth , in section 2.4 , we introduce a way to solve for the potential and kinetic energy densities of the dark energy as functions of redshift . in addition , we define a new function , the dark energy indicator , which provides a measure of deviations of @xmath2 from @xmath3 and a new and independent measure of @xmath4 . fifth , in section 2.3 , these derivatives are combined to solve for the expansion and acceleration rates of the universe as functions of redshift for both zero and non - zero space curvature ; in our previous work we have not considered the effects of non - zero space curvature . the only assumption that must be made to obtain the functions @xmath5 and @xmath0 from the data are that the robertson - walker metric is valid in our universe . a discussion and conclusions follow in section 3 . we consider three types of distances : those determined from luminosity distances to supernova standard candles ( sn ) , those determined from the angular diameter distances to radio galaxies ( rg ) , and those determined to clusters of galaxies with sz measurements of angular diameter distances ( cl ) . the sn samples include those of davis et al . ( 2007 ) , riess et al . ( 2007 ) , and astier et al . ( 2006 ) ; these authors provide the pertinent details about their measurements . there is some overlap between these supernovae samples , and this comparison allows the effects of different samples and sub - samples to be seen . in addition , a comparison between the values of @xmath9 , @xmath10 , and @xmath11 for the different samples goes hand in hand with a comparison of the best fit parameter values obtained in different models for these same samples , described by daly et al . the 71 new supernovae presented by astier et al . ( 2006 ) are included in both the riess et al . ( 2007 ) and davis et al . ( 2007 ) samples , and the high - redshift hst supernovae of riess et al . ( 2007 ) are included in davis et al . ( 2007 ) sample which , otherwise , includes only essence supernovae and low - redshift supernovae . these comparisons can be quite helpful , as illustrated by the work on nesseris & perivolaropoulos ( 2006 ) . the dimensionless coordinate distances @xmath12 to supernovae can be obtained from the published distance moduli @xmath13 using the best fit value of @xmath14 and the relations @xmath15 $ ] and @xmath16 . there are several ways to determine @xmath17 for published data sets that do not indicate the effective value of @xmath18 adopted to obtain @xmath13 ; the different methods provide values of @xmath14 that are in good agreement . here , we use the best fit value of @xmath14 obtained by daly et al . ( 2007 ) . in the cases where a value of @xmath18 is included in the publication of the values of @xmath13 ( e.g. astier et al . 2006 ) the value of @xmath18 adopted by astier et al . ( 2006 ) is recovered to very high accuracy . we use the new rg sample of daly et al . ( 2007 ) ; eleven new radio galaxies were observed and analyzed , which increases the sample size to 30 radio galaxies with redshifts between zero and about 1.8 . finally , we also use the angular diameter distances to a sample of 38 clusters determined with the sz measurements by bonamente et al . the angular diameter distances @xmath19 obtained by bonamente et al . ( 2006 ) for the hydrostatic equilibrium model were used . to convert from the angular diameter distance to the dimensionless coordinate distance we need to remove the value of @xmath18 that was adopted by bonamente et al . ( 2006 ) , so we use their best fit value of @xmath18 of @xmath20 to obtain the dimensionless coordinate distance @xmath12 to each of their clusters using the well - known relations @xmath21 and @xmath22 . after a detailed comparison between the sn and rg samples , we combine the davis et al . ( 2007 ) supernovae sample with the daly et al . ( 2007 ) radio galaxy sample , and study the combined sample of 222 sources . dimensionless coordinate distances @xmath12 and their uncertainties @xmath23 are obtained for these samples as described by daly et al . ( 2007 ) , and are listed here in table 1 . we also study results obtained by adding the cluster sample of bonamente et al . ( 2006 ) to obtain a sample of 260 sources , and these distances are also included in table 1 . figure 1 shows a comparison of the distances for these three data sets relative to the standard lambda cold dark matter ( lcdm ) model with @xmath24 and @xmath25 . a comparison of each data set with the standard lcdm model indicates that both the sn and rg data sets provide reliable cosmology probes , while the sz cluster method perhaps need some refinement ; the reduced chi - square for the sn and rg is about one , as expected from the quoted measurement errors , whereas that for the sz clusters is greater than two , suggesting that the quoted errors substantially underestimate the true uncertainties of these distance measurements . however , to illustrate how our method can be applied to coordinate distances obtained using different methods , we consider the analysis of the full sample of 260 sources as well as the analysis of the sample of 222 sn and rg . the distances @xmath12 to each source are used to obtain the distance @xmath9 to any redshift within the redshift range of the sample , and first and second derivatives of the distance with respect to redshift , @xmath10 and @xmath11 and their uncertainties using the method of daly & djorgovski ( 2003 , 2004 ) . in previous work , we have used @xmath26 and @xmath27 to obtain @xmath28 and @xmath0 . we then compared our empirically determined functions @xmath29 and @xmath0 with predictions in different models . however , it is also possible to compare our empirically determined functions @xmath10 and @xmath11 directly with model predictions for these quantities . the predicted values of these quantities are labeled @xmath30 and @xmath31 . for a universe with non - relativistic matter with mean mass energy density @xmath32 , mean dark energy density @xmath33 , and space curvature @xmath34 in which einstein s equations apply , we have , in full generality , the predicted values of @xmath26 , @xmath30 , and @xmath27 , @xmath31 are given by @xmath35 and @xmath36 \over [ \omega_m(1+z)^3+\omega_{de}f(z)+\omega_k(1+z)^2 ] } \right)\ ] ] where @xmath37 , @xmath38 is the zero redshift value of the mean mass density of non - relativistic matter relative to the critical density , @xmath39 , and @xmath40 is obtained by numerically integrating eq . ( 1 ) . this derivation does not assume that @xmath41 , is constant , and allows for variable @xmath42 . the function @xmath43 describes the redshift evolution of the energy density of the dark energy ; in a quintessence model with constant equation of state @xmath44 , @xmath45 , and for a cosmological constant @xmath46 . the data were analyzed using a sliding window fit , described in section 2.2.1 , and using fits in independent redshift bins , described in section 2.2.2 . fits are done using the window @xmath47 throughout for the sn data , @xmath48 for the rg data , when considered individually , and @xmath47 for the joint samples using the method described by daly & djorgovski ( 2003 , 2004 ) . the width of the fitting window is driven by the need to obtain useful confidence intervals for the fits by including a sufficient number of data points . as the size of the available data sets increases in the future , this width could be correspondingly narrowed . in decreasing the window function width from 0.6 to 0.5 and 0.4 , the trends and overall results remain the same , the uncertainties increase ( because there are fewer data points in the window ) , and the trends become more noisy ( due to sparser sampling ) . in addition , to test whether the window function has an effect on the trends extracted from the data we created a mock data set with the same number and redshift distribution of points as in each data set and the same fractional error per point , but with @xmath12 values obtained from a standard lcdm cosmology with @xmath49 and @xmath25 and ran it through the programs to extract @xmath9 , @xmath10 , and @xmath11 . as expected , the uncertainties increase as the window width decreases due to the smaller number of data points in the window . for a window widths of 0.3 and 0.4 in redshift , @xmath9 , @xmath10 , @xmath11 , and @xmath0 match the input cosmology to very high precision . for widths of 0.5 and 0.6 , there is a very slight offset of @xmath11 and @xmath0 from the input cosmology that sets in above redshifts of about one , at a level that is very small compared with the uncertainties . thus , the values of @xmath9 , @xmath10 , @xmath11 , and @xmath0 , and quantities obtained by combining these quantities , provide reliable determinations to redshifts well above one . our completely model - independent determinations of @xmath9 , @xmath10 , and @xmath11 are shown in figures [ yofz ] through [ d2ydz2192 ] . in figures [ yofz ] , [ dydz ] , and [ d2ydz2 ] , they are compared with the predicted value in a spatially flat universe , described by general relativity with a cosmological constant @xmath25 , and non - relativistic matter @xmath49 . this provides a reasonable description of the data to redshift of about one or so for the supernovae and 1.5 or so for the radio galaxies . ( we note that this is not a model fit , but simply an illustration of its compatibility with the data . ) the values of @xmath10 and @xmath11 obtained for the davis et al . ( 2007 ) supernova sample are compared with the best fit model parameters obtained in a spatially flat quintessence model , a lambda model that allows for non - zero space curvature , and the standard lcdm model in figures [ dydz192 ] and [ d2ydz2192 ] using the best fit model parameters listed by daly et al . the best fit model parameters are obtained by fitting the data to a model , and each model is based upon general relativity . a comparison of the first and second derivatives of @xmath12 with those predicted in each of the models provides another way to see whether the model predictions provide a good description of the data . the models and best fit model parameters are shown for a lambda model that allows for non - zero space curvature , and a quintessence model . [ d2ydz2 ] shows @xmath27 for each of the samples . the canonical lcdm model provides a good description of the radio galaxy data over the redshift range from zero to 1.5 . the comparison between the values of @xmath26 and @xmath27 determined directly from the data with those predicted in a standard lcdm model that relies upon the equations of general relativity provides effectively a large - scale test of general relativity . the good agreement obtained indicates that general relativity provides a good description of the data on time scales ( and the corresponding length scales ) of about ten billion years . for the davis et al . ( 2007 ) sample of 192 supernovae , the best fit models track @xmath26 and @xmath27 to a redshift of about 1.2 or so , beyond which the data drops away from the model prediction ( see figs . [ dydz192 ] and [ d2ydz2192 ] ) . curves expected for the best fit parameters obtained in a quintessence model and cosmological constant model with space curvature by fitting @xmath9 , obtained by daly et al . ( 2007 ) , are shown as well as a standard lcdm model . for @xmath26 all three curves fit well , and some differences emerge for @xmath27 ; in particular , the curve with non - zero space curvature does not fit the data well at low redshift . our sliding window fit method produces fit values which are strongly correlated over the redshift range corresponding to the fitting window , and are thus indicative of trends , but can not be used simply to evaluate a statistical significance of any particular model . for that , we would need independent values at different redshifts . to this effect , we divided the data sample in a number of independent redshift bins ( note : the data are @xmath50 binned or averaged , the sample is divided in groups of points belonging to non - overlaping redshift bins ) . one drawback of this approach is that the numbers of data points in each bin are smaller , and thus the fit values are noisier . another drawback is that the boundary values of the fits are not constrained , allowing for discontinuities in the values of @xmath9 , @xmath51 , and @xmath52 at the bin boundaries , which physically makes no sense . this is the price of the statistical independence . the dimensionless coordinate distances to the 192 supernovae of davis et al . ( 2007 ) were combined with those to the 30 radio galaxies of daly et al . ( 2007 ) to obtain a sample of 222 sources with redshift between 0.016 and 1.79 . these data were divided into bins based on the redshifts of the points to be able to obtain independent determinations of @xmath26 and @xmath27 and their uncertainties . the data were divided into redshift bins with roughly equal numbers of points per bin . we considered 2 bins with 111 points each ; 3 bins with 74 points each ; 4 bins with 55 points in the first three bins and 57 points in the highest redshift bin ; and 6 bins with 37 points each . the bin , number of points per bin , median redshift of the points in the bin , and the minimum and maximum redshift of points within the bin are listed in table [ binresults ] . points in each bin were used to determine the values of @xmath26 and @xmath27 and their uncertainties at the median redshift of the bin . the results for @xmath26 are shown in figure [ yprimebin ] for 3 , 4 , and 6 bins ; the redshift range of the points that went into the determination of @xmath26 at the median redshift @xmath53 are indicated by the horizontal lines . three theoretical curves are included on the figure . the dotted line is the curve predicted by the standard lcdm model with @xmath49 , and the other two curves , which are nearly identical are those predicted in the cardassian model ( freese & lewis 2002 ) and the generalized chaplygin gas model ( kamenshchik , moschella , & pasquier 2001 ; bilic , tupper , & viollier 2002 ; and bento , bertolami , & sen 2002 ) are shown for the best fit parameters obtained by bento et al . ( 2005 ) assuming a spatially flat universe . the results obtained for @xmath27 are shown in figure [ yprimeprimebin ] for 2 and 3 bins . the full set of results are listed in table [ binresults ] . given the noise inherent in the data that is presently available , it is not possible to obtain meaningful results for @xmath27 with a larger number of independent bins . with more data , we will be able to obtain this quantity in a larger number of independent redshift bins . the dimensionless expansion rate @xmath54 and the deceleration parameter @xmath0 can be constructed directly from the fist and second derivatives of the coordinate distance @xmath26 and @xmath27 , as discussed in detail by daly & djorgovski ( 2003 ) . the relationship between @xmath29 and @xmath26 , and that between @xmath0 and @xmath26 and @xmath27 are exact ; they do not represent expansions about some point . the only requirement to derive these exact relationships is that the most the robertson - walker line element applies in our universe . with this assumption alone , it can be shown that @xmath55 ( e.g. weinberg 1972 ) , and @xmath56 ( daly & djorgovski 2003 ) , where @xmath57 , @xmath58 , @xmath59 , and @xmath34 is positive and @xmath1 is negative when space curvature is positive . thus , the zero redshift value of @xmath60 is independent of space curvature , as is @xmath61 since @xmath62 at @xmath63 . thus , the zero redshift values of @xmath29 and @xmath0 obtained from @xmath26 and @xmath27 are independent of space curvature . and , the zero redshift value of @xmath64 indicates whether the universe is accelerating at the current epoch and can be determined independent of @xmath1 . the data and analysis described in section 2.2.1 were used to obtain @xmath5 and @xmath0 using equations [ h ] and [ q ] . the resuls are shown in figures [ yesnprg ] and [ qsnprg ] for the combined sample of 222 sources described above . the results confirm that the universe is accelerating at the current epoch . for the davis et al . ( 2007 ) sample of 192 supernovae , we find a zero redshift value of @xmath64 are @xmath65 . for the radio galaxies , we find @xmath66 , consistent with the results obtained using supernovae . again , these results depend only upon the form of the robertson - walker line element and are independent of space curvature , whether general relativity is the correct theory of gravity , and the content or evolution of the contents of the universe . for @xmath67 , the redshift at which the universe transitions from acceleration to deceleration for the davis et al . ( 2007 ) sample of 192 sn @xmath68 . we investigated the effect of the size of the window function of the value of @xmath8 and the transition redshift for the combined sample of 192 supernovae and 30 radio galaxies . for a window function of width 0.6 in redshift , we obtain @xmath69 , and a transition redshift @xmath70 ; for a window function of width 0.5 in redshift , we obtain @xmath71 , and a transition redshift @xmath72 ; and for a window function of width 0.4 , we obtain @xmath73 , and @xmath74 . these numbers are all consistent , though the uncertainties increase as the size of the window function decreases since then fewer points are used to determine each quantity . it is important to increase the number of data points at all redshifts to that the size of the window function can be decreased . these transition redshifts are consistent with those obtained by melchiorri , pagano , & pandolfi ( 2007 ) . the data and analysis described in section 2.2.2 were used to obtain @xmath5 and @xmath0 using equations [ h ] and [ q ] . the results are shown in figures [ binh ] and [ binq ] , and are listed in table [ binresults ] . the values of @xmath75 are consistent with predictions in the standard lcdm model at about one sigma or better . for comparison , predictions obtained in a cardassian model and generalized chaplygin gas model , described in section 2.2.2 . are also displayed . the value of the deceleration parameter at the median redshift of the bin , @xmath76 , does not definitely show that the universe is accelerating today , using this approach . that is , we do not see that @xmath64 must be less than zero . if we consider the data split into two bins , each with 111 data points , and review the results for the lower redshift bin , we find that at a redshift of 0.02 the value of @xmath64 is @xmath77 . if we consider the data split into three bins , each with 74 data points , and review the results for the lowest redshift bin , we find that the value @xmath64 is consistent with zero for all of the data points in this bin . evidently , we do not yet have a sufficient density of data points even at a redshift less than about 0.5 to be able to definitely state at three sigma or better that the universe is accelerating today , using this method ( fits in independent redshift bins ) . it is only when we increase the number of data points that go into the determination of @xmath0 by using the sliding window function , that we can conclude that the universe is accelerating today using our model - independent analysis . of course , when it is concluded that the universe is accelerating today in the context of a quintessence model or other models , all of the data are being used in the context of that particular model . the quintessence and most other models implicitly assume that our universe is described by the robertson - walker line element , the equations of general relativity apply , and that a specific function for the redshift evolution of the dark energy is valid over the redshift interval under study . much more data are needed in order to test independently these ( perfectly reasonable ) assumptions . we can solve for the properties of the dark energy as functions of redshift if a theory of gravity is specified , as shown by daly & djorgovski ( 2004 ) . einstein s equations of general relativity are used to relate the pressure , @xmath78 , energy density , @xmath79 , equation of state @xmath44 , and potential , @xmath80 and kinetic @xmath81 energy densities as functions of redshift to the cosmic scale factor @xmath82 , and the first and second derivatives of @xmath82 with respect to time . since the robertson - walker line element has been used to relate the first and second derivatives of @xmath82 with respect to time to the first and second derivatives of the coordinate distance with respect to redshift , @xmath26 and @xmath27 , the equations of general relativity can be used to solve for the pressure , energy density , equation of state , and potential and kinetic energy density of the dark energy in terms of @xmath26 and @xmath27 , which are obtained directly from the data . here a value of @xmath67 is assumed , and the equations of daly & djorgovski ( 2004 ) are used to solve for the properties of the dark energy as functions of redshift . to obtain the pressure , we only need to specify that einstein s equations apply . @xmath83~. \label{p}\ ] ] as pointed out by daly & djorgovski ( 2004 ) , the zero redshift value of @xmath84 translates into a value of @xmath85 is a standard lcdm model since in this model @xmath86 , so @xmath87 . here , this implies that @xmath88 if the dark energy is due to a cosmological constant in a universe with zero space curvature . this value of @xmath85 , which is obtained from the first and second derivatives of the coordinate distance , is consistent with other measures . to obtain the energy density @xmath79 of the dark energy , @xmath4 must be specified , and a value of @xmath24 is adopted here . @xmath89 the equation of state @xmath2 is obtained by taking the ratio @xmath90 and is given by @xmath91 \over 1-\omega_m(1+z)^3~(y^{\prime})^{2}}~. \label{w}\ ] ] the potential energy density of a dark energy scalar field is given by @xmath92 so @xmath93 - 0.5 \omega_m(1+z)^3 \label{v}\ ] ] and the kinetic energy density is given by @xmath94 so @xmath95 - 0.5 \omega_m(1+z)^3~ , \label{k}\ ] ] since @xmath96 and @xmath97 . we define a new function , the dark energy indicator @xmath98 , which is given by @xmath99 this is motivated by the fact that , for zero space curvature , the predicted value of @xmath98 is @xmath100{\omega_{de}f(z ) \over \omega_m[1+z]^3}\right ) = -1.5 \omega_m \left(1+[1+w]{\rho_{de}\over \rho_m } \right ) ~\label{sp}\ ] ] as indicated by equations ( 1 ) and ( 2 ) ; this occurs because @xmath101 when @xmath102 . equation [ sp ] is derived assuming only that general relativity is valid , space curvature is zero , and there are two mass - energy components controlling the expansion of the universe , dark energy and non - relativistic matter including dark matter and baryons . it is not assumed that the equation of state of the dark energy , @xmath2 , is constant , and no functional form for the evolution of the energy density of the dark energy , @xmath43 , is assumed . if @xmath86 , then @xmath103 , and the value of @xmath98 provides a new and independent measure of @xmath4 . deviations of @xmath98 from a constant provide a measure of the deviations of @xmath2 from -1 ; the amount of the deviation also depends upon the ratio of the energy density of the dark energy @xmath104 to that of the mass density of non - relativistic matter @xmath105 , including dark matter and baryons , at a given redshift as discussed in more detail below . the analysis described in section 2.2.1 was applied to the combined sample of 192 supernovae from davis et al . ( 2007 ) and 30 radio galaxies from daly et al . the values of @xmath26 and @xmath27 were substituted into equations [ p ] , [ f ] , [ w ] , [ v ] , [ k ] , [ s ] , and [ s ] and the results are shown in figures [ pfw ] , [ kv ] , [ s ] , and [ s ] . the zero redshift values of these parameters are @xmath106 , @xmath107 , @xmath108 , @xmath109 , @xmath110 , and @xmath111 . the zero redshift value of @xmath98 indicates a value of @xmath112 if @xmath86 , and the zero redshift value of @xmath84 indicates a value of @xmath113 if @xmath86 . overall , the results are consistent with predictions in a standard lcdm model . as noted above , the dark energy indicator @xmath98 provides a measure of whether @xmath86 over the redshift range shown in figure [ s ] . to illustrate the signature of a value of @xmath2 that differs from @xmath3 , three curves are overlayed on the figure . each curve is obtained using equation [ sp ] assuming @xmath49 , @xmath114 , @xmath115 and @xmath2 is constant over the redshift range shown , with values of @xmath2 of @xmath116 ( short dashed curve ) , @xmath117 ( solid line ) , and @xmath118 ( long dashed curve ) . clearly , the curves with @xmath118 and @xmath119 do not provide a good description of the data . another way to look at this is that equation [ s ] and [ sp ] indicate that @xmath120 $ ] so the second part of this is a measure of the deviation of @xmath2 from -1 , and at zero redshift the ratio @xmath121 and @xmath49 , so @xmath98 places rather strong constraints on deviations of @xmath122 from -1 . since most of these functions involve combinations of the first and second derivatives of the coordinate distance , results obtained with independent redshift bins are quite noisy , so only values obtained in two independent redshift bins are listed in table 3 . space curvature , parameterized by @xmath1 , has only a modest effect on @xmath0 , as illustrated in figure [ kqofz192 ] . positive space curvature , with negative @xmath1 flattens the @xmath0 curve and pushes the redshift at which the universe transitions from an accelerating state to a decelerating state to higher redshift . this follows since @xmath26 is known to be positive in our universe since the universe is expanding . similarly , negative space curvature moves the redshift at which @xmath123 to lower redshift causing the universe the transition from an accelerating to a decelerating state at lower redshift . for @xmath67 , the redshift at which the universe transitions from acceleration to deceleration for the davis et al . ( 2007 ) sample is @xmath124 . for negative space curvature with @xmath125 this transition redshift is shifted to about 0.6 , and for positive space curvature with @xmath126 it shifts to about 0.8 . space curvature also affects the shape of @xmath5 as illustrated in figure [ keofz192 ] . positive values of @xmath1 cause @xmath127 to increase more steeply with redshift than negative values of @xmath1 , which tend to flatten out the @xmath127 curve . thus , the effect of space curvature on @xmath127 and @xmath0 is small relative to the overall uncertainties of their measurements , within a plausible range of the curvature parameter @xmath1 . we also consider the full sample of 260 sources including the 38 sz clusters of bonamente et al . ( 2006 ) , the 192 supernovae of davis et al . ( 2007 ) , and the 30 radio galaxies of daly et al . these results are shown in figure [ yetal260 ] ; a window function of width 0.6 was used to analyze the data . this analysis illustrates how this method can be applied to diverse data sets . however , while the sz cluster distances currently provide a useful tool for measurements of the far - field hubble parameter , it is probably premature to use them as standard rulers to probe the global geometry and kinematics of the universe at this time . gamma - ray bursts as standard candles have been analyzed in detail by schaefer ( 2007 ) , who gives other pertinent references . the values of @xmath13 listed by schaefer ( 2007 ) can be used to determine the dimensionless coordinate distance to each source if the value of @xmath18 relevant for the sample is known , as described in section 2.1 , and a value of @xmath128 was used ( schaefer , private communication ) . the dimensionless coordinate distances were analyzed to determine the functions @xmath9 , @xmath10 , and @xmath11 using a window function of width 2.0 in redshift . a first look at results for the gamma - ray bursts are shown in fig . [ grb ] , which suggest that these are a potentially promising tool to study cosmology at very large distances , and are broadly consistent with predictions in the canonical lcdm model , but the quality and the sparsity of the data are still not sufficient for the model - independent analysis as shown above for the sn+rg sample . the work presented here improves and extends our previous results . first , expanded and improved data sets are considered : three supernova samples and one radio galaxy sample . the radio galaxy data set has 11 new sources , increasing its size to 30 sources , and the supernovae data sets have increased substantially in size and quality . in addition , sz cluster distances and gamma - ray burst distances are considered . the dimensionless coordinate distances ( obtained directly from the data ) , and first and second derivatives of the distance are obtained as functions of redshift using a sliding window fit . the good agreement obtained using supernovae and radio galaxies , two completely independent methods , with sources that cover similar redshift ranges , suggests that neither method is strongly affected by systematic effects , and that each method provides a reliable cosmological tool . the first and second derivatives of the distance are combined to obtain the acceleration parameter @xmath0 , allowing for non - zero space curvature . it is shown that the zero redshift value of @xmath0 , @xmath129 , is independent of space curvature , and can be obtained from the first and second derivatives of the coordinate distance . thus , @xmath8 , which indicates whether the universe is accelerating at the current epoch , can be obtained directly from the supernova and radio galaxy data ; our determinations of @xmath0 only relies upon the validity of the robertson - walker line element , and is independent of a theory of gravity , and the contents of the universe . each of the supernova samples , analyzed using a sliding window fit , indicate that the universe is accelerating today independent of space curvature , independent of whether general relativity is the correct theory of gravity , and of the contents of the universe . the effect of non - zero space curvature on @xmath0 is to shift the redshift at which the universe transitions from acceleration to deceleration , moving this to lower redshift for negative space curvature and to higher redshift for positive space . the zero redshift values of @xmath64 obtained using a sliding window fit . for the davis et al . ( 2007 ) supernova sample is @xmath65 and that obtained for the radio galaxy sample of daly et al . ( 2007 ) is @xmath130 indicating that the universe is accelerating at the current epoch . the data were also binned so that only certain subsets of the data were used to solve for @xmath26 , @xmath27 , @xmath5 , and @xmath0 . the results for @xmath26 and @xmath5 indicate that the standard lcdm model provides a good description of the data . the results for @xmath27 and @xmath0 are consistent with the standard lcdm model , but do not independently confirm the model or the acceleration of the universe . in addition to the evaluation of the standard cosmological parameters , in an even more direct approach , we compared @xmath26 and @xmath27 obtained from the fits to the data to model predictions . comparisons of @xmath26 and @xmath27 with predictions based on general relativity indicate that general relativity provides an accurate description of the data on look - back time scales of about ten billion years , thus providing a very large scale test of general relativity . another new approach is that the data were analyzed using both a sliding window fit and fits in independent redshift bins . the fits in statistically independent redshift bins are broadly consistent with the sliding window fits , but are generally noisier ( as expected ) . we also explored the effects of non - zero space curvature on determinations of @xmath127 and @xmath0 . it is shown that the zero redshift value of @xmath64 , obtained by applying equation ( 4 ) to @xmath26 and @xmath27 , is independent of space curvature . this means that our method can be used to determine @xmath8 , and thus the degree to which the universe is accelerating at the current epoch , with only one assumption , that the robertson - walker line element is valid . in addition , it is found that the effect of space curvature on the shape of @xmath127 and @xmath0 is small , relative to the uncertainties arising from the measurement errors . after determining the expansion and acceleration rates of the universe as functions of redshift independent of a theory of gravity , we solve for the pressure , energy density , equation of state , and potential and kinetic energy of the dark energy as functions of redshift assuming that general relativity is the correct theory of gravity . we also define a new function , the dark energy indicator @xmath98 , which provides a measure of deviations of the equation of state of the dark energy @xmath2 from @xmath3 , and provides a new and independent measure of @xmath4 if @xmath86 . the results obtained using a sliding window fit indicate that a cosmological constant in a spatially flat universe provides a good description of each of these quantities over the redshift range from zero to one . the zero redshift values of these quantities obtained with the davis et al . ( 2007 ) supernovae sample are @xmath131 , @xmath132 , @xmath133 , @xmath134 , @xmath135 , and @xmath111 . in the standard lambda - cold dark matter model , @xmath136 , obtained using the first and second derivatives of the coordinate distance , provides an independent measure of @xmath85 . in addition , in this model , @xmath86 , so @xmath98 provides a measure of @xmath4 , and the value obtained here using the first and second derivatives of the coordinate distance , is @xmath112 . overall , the shapes of the pressure , energy density , equation of state , and other parameters as functions are redshift are consistent with those predicted in a standard lcdm model . there is a tantalizing hint that there may be divations from the standard model at high redshift ; more observations at high redshift will be needed to investigate this further . the results obtained using fits in independent redshift bins are consistent with the standard lcdm model , but do not independently confirm the model . we would like to thank the observers for their tireless efforts in obtaining the data used for this study . we would also like to thank the referee for helpful comments and suggesitons . this work was supported in part by u. s. national science foundation grants ast-0507465 ( r.a.d . ) and ast-0407448 ( s.g.d . ) , and the ajax foundation ( s.g.d . ) . davis , t. m. , mortsell , e. , sollerman , j. , becker , a. c. , blondin , s. , challis , p. , clocchiatti , a. , filippenko , a. v. , foley , r. j. , garnavich , p. m. , and 17 coauthors , submitted to apj , ( astro - ph/0701510 ) lllll sn & sn1994s & 0.016 & 0.0160 & 0.0016 + sn & sn2001v & 0.016 & 0.0145 & 0.0015 + sn & sn1996bo & 0.016 & 0.0135 & 0.0015 + sn & sn2001cz & 0.016 & 0.0155 & 0.0017 + sn & sn2000dk & 0.016 & 0.0161 & 0.0016 + sn & sn1997y & 0.017 & 0.0174 & 0.0018 + sn & sn1998ef & 0.017 & 0.0146 & 0.0016 + sn & sn1998v & 0.017 & 0.0160 & 0.0016 + sn & sn1999ek & 0.018 & 0.0155 & 0.0016 + sn & sn1992bo & 0.018 & 0.0190 & 0.0018 + sn & sn1992bc & 0.020 & 0.0200 & 0.0017 + sn & sn2000fa & 0.022 & 0.0205 & 0.0020 + sn & sn1995ak & 0.022 & 0.0187 & 0.0018 + sn & sn2000cn & 0.023 & 0.0226 & 0.0019 + sn & sn1998eg & 0.024 & 0.0248 & 0.0023 + sn & sn1994 m & 0.024 & 0.0239 & 0.0022 + sn & sn2000ca & 0.025 & 0.0239 & 0.0019 + sn & sn1993h & 0.025 & 0.0224 & 0.0019 + sn & sn1992ag & 0.026 & 0.0228 & 0.0023 + sn & sn1999gp & 0.026 & 0.0284 & 0.0021 + sn & sn1992p & 0.026 & 0.0281 & 0.0025 + sn & sn1996c & 0.028 & 0.0329 & 0.0027 + sn & sn1998ab & 0.028 & 0.0231 & 0.0019 + sn & sn1997dg & 0.030 & 0.0361 & 0.0032 + sn & sn2001ba & 0.031 & 0.0319 & 0.0023 + sn & sn1990o & 0.031 & 0.0309 & 0.0024 + sn & sn1999cc & 0.032 & 0.0310 & 0.0024 + sn & sn1996bl & 0.035 & 0.0350 & 0.0029 + sn & sn1994 t & 0.036 & 0.0338 & 0.0026 + sn & sn2000cf & 0.037 & 0.0395 & 0.0031 + sn & sn1999aw & 0.039 & 0.0429 & 0.0026 + sn & sn1992bl & 0.043 & 0.0417 & 0.0035 + sn & sn1992bh & 0.045 & 0.0505 & 0.0044 + sn & sn1995ac & 0.049 & 0.0431 & 0.0032 + sn & sn1993ag & 0.050 & 0.0541 & 0.0045 + sn & sn1990af & 0.050 & 0.0454 & 0.0042 + sn & sn1993o & 0.052 & 0.0553 & 0.0038 + sn & sn1998dx & 0.054 & 0.0503 & 0.0035 + rg & 3c 405 & 0.056 & 0.0514 & 0.0105 + sn & sn1992bs & 0.063 & 0.0695 & 0.0064 + sn & sn1993b & 0.071 & 0.0736 & 0.0064 + sn & sn1992ae & 0.075 & 0.0713 & 0.0069 + sn & sn1992bp & 0.079 & 0.0731 & 0.0050 + sn & sn1992br & 0.088 & 0.0718 & 0.0076 + sn & sn1992aq & 0.101 & 0.1145 & 0.0079 + sn & sn1996ab & 0.124 & 0.1174 & 0.0108 + cl & abell 1413 & 0.142 & 0.2283 & 0.0454 + cl & abell 2204 & 0.152 & 0.1801 & 0.0192 + sn & e020 & 0.159 & 0.1716 & 0.0229 + cl & abell 2259 & 0.164 & 0.1731 & 0.0806 + cl & abell 586 & 0.171 & 0.1561 & 0.0405 + cl & abell 1914 & 0.171 & 0.1321 & 0.0135 + cl & abell 2218 & 0.176 & 0.1990 & 0.0377 + sn & k429 & 0.181 & 0.1764 & 0.0138 + cl & abell 665 & 0.182 & 0.2000 & 0.0288 + cl & abell 1689 & 0.183 & 0.1971 & 0.0273 + cl & abell 2163 & 0.202 & 0.1602 & 0.0139 + sn & d086 & 0.205 & 0.1887 & 0.0261 + sn & h363 & 0.213 & 0.2103 & 0.0320 + sn & n404 & 0.216 & 0.2364 & 0.0338 + cl & abell 773 & 0.217 & 0.3057 & 0.0484 + sn & g005 & 0.218 & 0.2133 & 0.0255 + cl & abell 2261 & 0.224 & 0.2290 & 0.0518 + cl & abell 2111 & 0.229 & 0.2016 & 0.0583 + cl & abell 267 & 0.230 & 0.1892 & 0.0315 + cl & rx j2129.7 + 0005 & 0.235 & 0.1456 & 0.0301 + sn & e132 & 0.239 & 0.2146 & 0.0287 + cl & abell 1835 & 0.252 & 0.3434 & 0.0160 + cl & abell 68 & 0.255 & 0.2027 & 0.0563 + sn & 04d3ez & 0.263 & 0.2462 & 0.0238 + sn & n326 & 0.268 & 0.2509 & 0.0300 + sn & k425 & 0.274 & 0.2881 & 0.0371 + cl & abell 697 & 0.282 & 0.2892 & 0.0871 + sn & p455 & 0.284 & 0.2832 & 0.0378 + sn & 03d4ag & 0.285 & 0.2678 & 0.0160 + sn & m027 & 0.286 & 0.3447 & 0.0508 + cl & abell 611 & 0.288 & 0.2575 & 0.0594 + cl & zw 3146 & 0.291 & 0.2747 & 0.0066 + sn & 03d3ba & 0.291 & 0.2196 & 0.0324 + sn & g055 & 0.302 & 0.3192 & 0.0544 + sn & d117 & 0.309 & 0.3219 & 0.0400 + sn & n278 & 0.309 & 0.2856 & 0.0276 + cl & abell 1995 & 0.322 & 0.4033 & 0.0491 + cl & ms 1358.4 + 6245 & 0.327 & 0.3844 & 0.0323 + sn & 03d1fc & 0.331 & 0.2996 & 0.0290 + sn & e029 & 0.332 & 0.3297 & 0.0425 + sn & d083 & 0.333 & 0.2279 & 0.0147 + sn & 04d3kr & 0.337 & 0.3209 & 0.0251 + sn & g097 & 0.340 & 0.3354 & 0.0479 + sn & 04d3nh & 0.340 & 0.3463 & 0.0271 + sn & m193 & 0.341 & 0.2960 & 0.0313 + sn & d149 & 0.342 & 0.3459 & 0.0334 + sn & h364 & 0.344 & 0.2994 & 0.0234 + sn & 03d1bp & 0.346 & 0.3248 & 0.0374 + sn & h359 & 0.348 & 0.3881 & 0.0483 + sn & e136 & 0.352 & 0.3417 & 0.0425 + sn & 04d2fs & 0.357 & 0.3452 & 0.0350 + sn & 04d3fk & 0.358 & 0.3089 & 0.0299 + sn & d093 & 0.363 & 0.3566 & 0.0230 + sn & n263 & 0.368 & 0.3285 & 0.0257 + sn & 03d3ay & 0.371 & 0.3662 & 0.0388 + cl & abell 370 & 0.375 & 0.3807 & 0.0687 + sn & g052 & 0.383 & 0.3250 & 0.0329 + sn & g142 & 0.399 & 0.3862 & 0.0765 + sn & d085 & 0.401 & 0.3857 & 0.0391 + sn & k448 & 0.401 & 0.4594 & 0.0846 + rg & 3c142.1 & 0.406 & 0.3325 & 0.0607 + cl & macs j2228.5 + 2036 & 0.412 & 0.4416 & 0.0851 + sn & 04d2fp & 0.415 & 0.3999 & 0.0313 + sn & k485 & 0.416 & 0.4184 & 0.0751 + sn & g133 & 0.421 & 0.4286 & 0.0651 + sn & h342 & 0.421 & 0.4208 & 0.0310 + sn & f235 & 0.422 & 0.3498 & 0.0387 + sn & b013 & 0.426 & 0.3824 & 0.0405 + sn & e148 & 0.429 & 0.4321 & 0.0398 + sn & 04d2 gb & 0.430 & 0.3526 & 0.0292 + rg & 3c 244.1 & 0.430 & 0.3631 & 0.0671 + sn & d089 & 0.436 & 0.3922 & 0.0361 + sn & d097 & 0.436 & 0.4013 & 0.0314 + sn & 03d3aw & 0.449 & 0.3923 & 0.0434 + cl & rx j1347.5 - 1145 & 0.451 & 0.3571 & 0.0260 + sn & 04d3gt & 0.451 & 0.2812 & 0.0298 + sn & hst04yow & 0.460 & 0.4114 & 0.0625 + sn & 03d3cd & 0.461 & 0.3982 & 0.0330 + sn & 03d3cc & 0.463 & 0.4222 & 0.0331 + sn & m158 & 0.463 & 0.4914 & 0.0634 + sn & e108 & 0.469 & 0.4262 & 0.0314 + sn & 04d3df & 0.470 & 0.3796 & 0.0350 + sn & sn2002dc & 0.475 & 0.4091 & 0.0396 + cl & macs j2214.9 - 1359 & 0.483 & 0.5474 & 0.0950 + cl & macs j1311.0 - 0310 & 0.490 & 0.5271 & 0.1604 + sn & g160 & 0.493 & 0.4391 & 0.0526 + sn & h319 & 0.495 & 0.4426 & 0.0428 + sn & 03d1ax & 0.496 & 0.4283 & 0.0394 + sn & e149 & 0.497 & 0.4087 & 0.0489 + sn & h283 & 0.502 & 0.4592 & 0.0782 + sn & 03d1au & 0.504 & 0.4713 & 0.0391 + sn & p524 & 0.508 & 0.4449 & 0.0451 + sn & g120 & 0.510 & 0.4185 & 0.0405 + sn & d084 & 0.519 & 0.5612 & 0.0749 + rg & 3c172 & 0.519 & 0.6635 & 0.1366 + sn & 04d2gc & 0.521 & 0.4431 & 0.0673 + sn & 04d1ak & 0.526 & 0.4520 & 0.0604 + sn & n285 & 0.528 & 0.4814 & 0.0576 + sn & d033 & 0.531 & 0.5594 & 0.0438 + sn & 03d3af & 0.532 & 0.5289 & 0.0706 + sn & f011 & 0.539 & 0.4846 & 0.0558 + sn & f244 & 0.540 & 0.4979 & 0.0596 + cl & cl 0016 + 1609 & 0.541 & 0.5451 & 0.0869 + cl & macs j1149.5 + 2223 & 0.544 & 0.3166 & 0.0693 + cl & macs j1423.8 + 2404 & 0.545 & 0.5901 & 0.0178 + rg & 3c 330 & 0.549 & 0.3424 & 0.0652 + cl & ms 0451.6 - 0305 & 0.550 & 0.5642 & 0.0973 + sn & 04d4bq & 0.550 & 0.5016 & 0.0670 + sn & 04d3hn & 0.552 & 0.4035 & 0.0762 + sn & f041 & 0.561 & 0.4912 & 0.0385 + cl & macs j2129.4 - 0741 & 0.570 & 0.5352 & 0.1308 + sn & 03d4gf & 0.581 & 0.5149 & 0.0427 + sn & 03d1aw & 0.582 & 0.5828 & 0.0564 + cl & ms 2053.7 - 0449 & 0.583 & 1.0063 & 0.1725 + cl & macs j0647.7 + 7015 & 0.584 & 0.3126 & 0.0792 + sn & 03d4gg & 0.592 & 0.5161 & 0.0594 + sn & h323 & 0.603 & 0.5467 & 0.0554 + sn & 03d4dy & 0.604 & 0.4937 & 0.0728 + sn & 04d3do & 0.610 & 0.4987 & 0.0666 + sn & e138 & 0.612 & 0.5386 & 0.0446 + sn & 04d4an & 0.613 & 0.5611 & 0.1008 + sn & f231 & 0.619 & 0.5513 & 0.0432 + sn & 04d3co & 0.620 & 0.5904 & 0.0707 + sn & 03d4dh & 0.627 & 0.5216 & 0.0552 + rg & 3c 337 & 0.630 & 0.5064 & 0.0704 + sn & e140 & 0.631 & 0.5084 & 0.0421 + sn & n256 & 0.631 & 0.5575 & 0.0385 + sn & g050 & 0.633 & 0.4805 & 0.0398 + sn & 03d4at & 0.633 & 0.6021 & 0.0693 + rg & 3c169.1 & 0.633 & 0.6222 & 0.0708 + sn & sn2003be & 0.640 & 0.5246 & 0.0628 + sn & 04d3cy & 0.643 & 0.6209 & 0.0629 + sn & e147 & 0.645 & 0.5327 & 0.0442 + rg & 3c44 & 0.660 & 0.7621 & 0.0848 + sn & sn2003bd & 0.670 & 0.5597 & 0.0644 + sn & 03d1co & 0.679 & 0.6817 & 0.0848 + cl & macs j0744.8 + 3927 & 0.686 & 0.7261 & 0.1858 + sn & g240 & 0.687 & 0.5267 & 0.0485 + sn & h300 & 0.687 & 0.5390 & 0.0422 + sn & 03d1fl & 0.688 & 0.5486 & 0.0581 + rg & 3c34 & 0.690 & 0.5920 & 0.0644 + sn & 04d2iu & 0.691 & 0.6259 & 0.1124 + sn & p454 & 0.695 & 0.6569 & 0.0514 + sn & 03d4cz & 0.695 & 0.5414 & 0.0848 + rg & 3c441 & 0.707 & 0.5340 & 0.0667 + sn & 04d3is & 0.710 & 0.7074 & 0.1108 + rg & 3c 55 & 0.720 & 0.5852 & 0.0777 + sn & 04d1aj & 0.721 & 0.6264 & 0.0664 + sn & 04d3fq & 0.730 & 0.6556 & 0.0785 + sn & sn2002kd & 0.735 & 0.5265 & 0.0485 + sn & hst04rak & 0.740 & 0.5863 & 0.0621 + sn & 04d2ja & 0.741 & 0.6544 & 0.0723 + rg & 3c 247 & 0.749 & 0.5430 & 0.0682 + sn & 04d3ks & 0.752 & 0.5877 & 0.0622 + cl & ms 1137.5 + 6625 & 0.784 & 1.3033 & 0.2629 + sn & 03d4fd & 0.791 & 0.6880 & 0.0665 + rg & 3c41 & 0.794 & 0.6315 & 0.0714 + sn & 03d1fq & 0.800 & 0.7403 & 0.0920 + rg & 3c 265 & 0.811 & 0.5901 & 0.0793 + cl & rx j1716.4 + 6708 & 0.813 & 0.4833 & 0.2184 + rg & 3c114 & 0.815 & 0.6368 & 0.0695 + sn & 04d3nc & 0.817 & 0.6688 & 0.0647 + sn & 03d4cn & 0.818 & 0.8111 & 0.1009 + sn & 04d3lu & 0.822 & 0.6795 & 0.0688 + cl & ms 1054.5 - 0321 & 0.826 & 0.6225 & 0.1264 + rg & 3c54 & 0.827 & 0.7573 & 0.0835 + sn & 04d3cp & 0.830 & 0.6342 & 0.0496 + sn & hst05spo & 0.839 & 0.5729 & 0.0554 + sn & 04d4bk & 0.840 & 0.7110 & 0.0557 + sn & sn2003eq & 0.840 & 0.6336 & 0.0642 + rg & 3c6.1 & 0.840 & 0.7402 & 0.0861 + sn & hst04man & 0.854 & 0.7187 & 0.0993 + rg & 3c 325 & 0.860 & 0.7080 & 0.1292 + sn & 03d1ew & 0.868 & 0.7233 & 0.0566 + cl & cl j1226.9 + 3332 & 0.890 & 0.5232 & 0.1696 + sn & sn2003eb & 0.900 & 0.6052 & 0.0725 + sn & 03d4di & 0.905 & 0.6742 & 0.0497 + sn & 04d3gx & 0.910 & 0.7973 & 0.0661 + sn & 04d3ki & 0.930 & 0.8732 & 0.0764 + sn & sn2003xx & 0.935 & 0.6918 & 0.0956 + sn & 03d4cx & 0.949 & 0.7996 & 0.0626 + sn & 04d3ml & 0.950 & 0.7562 & 0.0522 + sn & sn2002dd & 0.950 & 0.6897 & 0.1112 + sn & sn2003es & 0.954 & 0.7975 & 0.1028 + sn & hst04tha & 0.954 & 0.6482 & 0.0836 + rg & 3c 289 & 0.967 & 0.5950 & 0.1097 + sn & hst04pat & 0.970 & 0.9380 & 0.1598 + rg & 3c 268.1 & 0.974 & 0.7458 & 0.1432 + sn & hst04omb & 0.975 & 0.7570 & 0.0941 + rg & 3c 280 & 0.996 & 0.6477 & 0.1221 + sn & 04d3dd & 1.010 & 0.9494 & 0.0743 + sn & hst05str & 1.010 & 0.9627 & 0.0887 + sn & hst05fer & 1.020 & 0.6688 & 0.0862 + sn & hst04eag & 1.020 & 0.8537 & 0.0786 + rg & 3c 356 & 1.079 & 0.8735 & 0.1842 + sn & hst05gab & 1.120 & 0.8716 & 0.0763 + sn & sn2002ki & 1.140 & 0.8795 & 0.1215 + sn & hst04gre & 1.140 & 0.7767 & 0.1145 + rg & 3c 267 & 1.144 & 0.7136 & 0.1396 + rg & 3c 194 & 1.190 & 1.0074 & 0.2047 + rg & 3c 324 & 1.210 & 1.0208 & 0.3089 + sn & hst05koe & 1.230 & 1.0432 & 0.1153 + sn & hst05lan & 1.230 & 0.9514 & 0.0920 + sn & sn2002fw & 1.300 & 0.9615 & 0.0930 + sn & sn2002hp & 1.305 & 0.7447 & 0.1063 + rg & 3c469.1 & 1.336 & 1.1364 & 0.3255 + sn & sn2003dy & 1.340 & 0.8860 & 0.1306 + sn & hst04mcg & 1.370 & 1.0091 & 0.1208 + sn & hst04sas & 1.390 & 0.8595 & 0.0792 + rg & 3c 437 & 1.480 & 0.9299 & 0.2697 + rg & 3c 68.2 & 1.575 & 1.5748 & 0.4873 + rg & 3c 322 & 1.681 & 1.3078 & 0.3400 + sn & sn1977ff & 1.755 & 0.9174 & 0.1521 + rg & 3c 239 & 1.790 & 1.3666 & 0.3382 + [ ytable ] lllllllll 1/6&37 & 0.025 & 0.016 & 0.052 & @xmath137&@xmath138&@xmath139&@xmath140 + 2/6&37 & 0.275 & 0.054 & 0.348 & @xmath141&@xmath142&@xmath143&@xmath144 + 3/6&37 & 0.430 & 0.352 & 0.504 & @xmath145&@xmath146&@xmath147& + 4/6&37 & 0.600 & 0.508 & 0.670 & @xmath148&@xmath149&@xmath150& + 5/6&37 & 0.790 & 0.679 & 0.905 & @xmath151&@xmath152&@xmath153& + 6/6&37 & 1.100 & 0.910 & 1.790 & @xmath154&@xmath155&@xmath156&@xmath157 + + 1/4&55 & 0.035 & 0.016 & 0.268 & @xmath158&@xmath159&@xmath160&@xmath161 + 2/4&55 & 0.400 & 0.274 & 0.502 & @xmath162&@xmath163&@xmath164&@xmath165 + 3/4&55 & 0.630 & 0.504 & 0.749 & @xmath166&@xmath167&@xmath168& + 4/4&57 & 0.965 & 0.752 & 1.790 & @xmath169&@xmath170&@xmath171&@xmath172 + + 1/3&74 & 0.050 & 0.016 & 0.348 & @xmath173&@xmath174&@xmath175&@xmath176 + 2/3&74 & 0.505 & 0.352 & 0.670 & @xmath177&@xmath178&@xmath179&@xmath180 + 3/3&74 & 0.905 & 0.679 & 1.790 & @xmath181&@xmath182&@xmath183&@xmath184 + + 1/2&111 & 0.275 & 0.016 & 0.504 & @xmath185&@xmath186&@xmath187&@xmath188 + 2/2&111 & 0.790 & 0.508 & 1.790 & @xmath189&@xmath190&@xmath191&@xmath192 + [ binresults ]

we introduce the use , monitoring , and enforcement of integrity constraints in trust management - style authorization systems . we consider what portions of the policy state must be monitored to detect violations of integrity constraints . then we address the fact that not all participants in a trust management system can be trusted to assist in such monitoring , and show how many integrity constraints can be monitored in a conservative manner so that trusted participants detect and report if the system enters a policy state from which evolution in unmonitored portions of the policy could lead to a constraint violation .

cs0503061

trust management @xcite ( tm ) is an approach to managing authorization in environments where authority emanates from multiple sources . authorization policy consists of statements issued by many participants , and resource sharing is facilitated by delegating authority from one principal to another . a particular authorization is decided by posing a query to the system . an evaluation procedure combines the statements issued by all relevant principals to derive the query s answer . by adding or removing a policy statement , a principal can potentially affect many authorizations of many principals . one of the difficulties of operating in such a context is that at present no system exists for monitoring unexpected consequences of policy changes made by other principals . basically , in present tm systems , delegating trust implies losing a great deal of control on the policy involved the delegation . let us first see three example of this . firstly , resources may become unavailable unexpectedly . consider for instance a team leader who needs to be informed if members of his team suffer interruption in their authorization for mission - critical resources . if the team s mission involves rapid response , the notification of interruption should not depend on team members attempting to access a critical resource and discovering its unavailability only because the attempt fails . what is needed is that the policy change triggers a procedure that pushes the notification to the team leader . secondly : properties such as mutual exclusion can not be guaranteed . while in the above example , the exceptional state involved someone losing authorization , having someone unexpectedly gain authorization can be just as important to detect . for instance , it should be possible to trigger an action if a principal becomes authorized for two mutually exclusive purposes . mutual exclusion is an approach often used , for instance in rbac systems @xcite , to enforce separation of duty , a classic device aimed at preventing fraud . by ensuring that no individual is authorized to complete all parts of a sensitive task , the technique ensures that only a colluding group could misuse the capability . because the participants in a trust management system are autonomous , it is in general not possible to prevent a principal being given two authorizations . however , cooperating principals should be able to prevent another principal from gaining two mutually exclusive authorizations under the control of the cooperating group . what is needed is a way to distribute the mutual exclusivity requirement and monitor policy evolution to ensure that control over the key authorizations is not delegated outside the cooperating group . thirdly : quality can not be monitored . consider the situation in which the principal @xmath0 states , for instance , that he considers expert anyone that @xmath1 considers an expert ( @xmath0 delegates to @xmath1 the definition of `` expert '' ) . in addition , @xmath0 expects experts to have a phd degree . now , @xmath0 has no way of controlling that all experts added by @xmath1 actually have doctorates . of course , @xmath0 could modify his policy as follows `` @xmath0 considers expert anyone _ holding a phd _ that @xmath1 considers an expert '' . however often it would be preferable for @xmath0 to know whether a non - phd had been added to the expert list because it might suggest to @xmath0 that an exception to @xmath0 s policy is acceptable , or that some other evolution of @xmath0 s policy should take place ( perhaps it is time to revoke the trust in @xmath1 s experts ) . thus , what @xmath0 needs is to be able to monitor whether @xmath1 ever decides that a non - phd is an expert . notice that this is what would happen in practice : before delegating to @xmath1 the definition of expert @xmath0 would normally put in place a monitoring activity to guarantee that @xmath1 s expert fulfill the quality criteria . unfortunately , present decentralized tm systems do not allow for such monitoring . summarizing , there is a need for a mechanism to monitor a tm system and to reveal when an exceptional state has been entered so that appropriate steps can be taken proactively . ideally , it would even be possible to enlist the assistance of others in preventing exceptional states from arising . the problem of providing such a monitoring system is aggravated by the fact that changes are made by autonomous principals that may not agree or be trusted to assist in the monitoring . in this paper we introduce a new trust management construct called a constraint , inspired by integrity constraints in database management systems ( see , e.g. @xcite ) , that provides system participants the ability to monitor the evolution of the policy . the author of a constraint receives notification when the constraint is violated . this is achieved by enlisting the assistance of principals to which authority is delegated and triggering constraint checks when those principals make relevant policy changes . the emphasis in this paper is on determining whether a policy change is relevant , or can be ignored . in addition we also consider the setting in which some principals are not trusted or willing to help monitoring a constraint . as mentioned above , in some environments , it is not appropriate to assume that all principals to whom one delegates authority will assist in monitoring one s constraints . by providing a sufficiently expressive constraint language , we show how to limit to an arbitrary , specified set those principals that are trusted to cooperate in monitoring a constraint . this is done by allowing a constraint to express a security analysis problem of the kind formulated by li et al . such a constraint quantifies over policy states that are reachable by policy changes made by untrusted principals asking whether a given query holds either in all reachable states ( universal quantification ) or in some reachable state ( existential quantification ) . by checking such a constraint each time the trusted principals make relevant policy changes , and committing their changes only if the constraint is satisfied , the trusted principals can ensure that a state violating the constraint is never entered , no matter what the untrusted principals do . they are able to do this because the untrusted principals are unable to affect the validity of the constraint . the technical contribution in this paper is a method to identify portions of the policy state that must be monitored in order to detect constraint violations . we do this first under the assumption that all principals in the system can be trusted to assist in monitoring the portion of the policy state under their control . we then relax this assumption by requiring only that a given portion of the policy can be reliably monitored . in this case , monitoring is carried out by using security analysis to assess the possibility of the constraint becoming violated by policy changes that can not be monitored directly . section [ sec : preliminaries ] discusses the tm policy language that we use . section [ sec : main ] identifies the portion of the policy state to be monitored for constraint violations , assuming all portions can be monitored . section [ sec : safety ] shows how to monitor constraints for potential violations when not all parts of the policy state can be monitored directly . section [ sec : relatedwork ] discusses related work . section [ sec : conclusion ] concludes . some proofs are reported in the appendix . trust management @xcite is an approach to access control in decentralized distributed systems with access control decisions based on policy statements issued by multiple principals . in trust management systems , statements that are maintained in a distributed manner are often digitally signed to ensure their authenticity and integrity ; such statements are sometimes called _ credentials _ or _ certificates_. this section presents the trust management language @xmath2 @xcite , which we use in this paper . a _ principal _ is a uniquely identified individual or process . principals are denoted by names starting with an uppercase , typically , @xmath0 , @xmath1 , @xmath3 . a principal can define a _ role _ , which is indicated by principal s name followed by the _ role name _ , separated by a dot . for instance @xmath4 , and @xmath5 are roles . for the sake of simplicity we assume that @xmath0 is the _ owner _ ( or the administrator ) of @xmath4 , though the results of this papers apply also in the case @xmath4 is owned by some other principal . we use names starting with a lowercase letter ( sometimes with subscripts ) to indicate role names . a role denotes a set of principals ( the principals that populate it , i.e. , the members of the role ) . to indicate which principals populate a role , @xmath2 allows the owning principal to issue four kind of _ policy statements _ : * _ simple member _ : @xmath6 + with this statement @xmath0 asserts that @xmath3 is a member of @xmath4 . * _ simple inclusion _ : @xmath7 + with this statement @xmath0 asserts that @xmath4 includes ( all members of ) @xmath8 . this represents a delegation from @xmath0 to @xmath1 , as @xmath1 may add principals to become members of the role @xmath4 by issuing statements defining ( and extending ) @xmath8 . * _ linking inclusion _ : @xmath9 + we call @xmath10 a _ linked role_. with this statement @xmath0 asserts that @xmath4 includes @xmath11 for every @xmath1 that is a member of @xmath12 . this represents a delegation from @xmath0 to all the members of the role @xmath12 . * _ intersection inclusion _ : @xmath13 + we call @xmath14 an _ intersection_. with this statement @xmath0 asserts that @xmath4 includes every principal who is a member of both @xmath15 and @xmath16 . this represents partial delegations from @xmath0 to @xmath17 and to @xmath18 . for any statement @xmath19 , @xmath4 is called the _ head _ and @xmath20 is called the _ body _ of the statement . we write @xmath21 . the set of statements having head @xmath4 is called the _ definition _ of @xmath4 . the definition of @xmath2given here is a slightly simplified ( yet expressively equivalent ) version of the one given in @xcite . a _ policy state _ ( _ state _ for short , indicated by @xmath22 ) is a set of policy statements . given a state @xmath22 , we define the following : @xmath23 is the set of principals in @xmath22 , @xmath24is the set of role names in @xmath22 , and @xmath25 . to express constraints , we need one last definition : _ positive roles expressions _ are defined by the following grammar : * sets of principals are positive role expressions , * roles are positive role expressions , * union and intersections of positive role expressions are positive role expressions . e.g. , @xmath4 , @xmath26 and @xmath27 . positive role expressions , and are denoted by greek letters , @xmath28 , and @xmath29 . a positive role expression containing no roles ( but only sets of principals ) is called _ static_. the semantics of a policy state is defined by translating it into a logic program . the _ semantic program _ , @xmath30 , of a state @xmath22 , is a prolog program has one ternary predicate @xmath31 . intuitively , @xmath32 means that @xmath3 is a member of the role @xmath4 . [ def : sp ] given a state @xmath22 , the _ semantic program _ @xmath30 for it is the logic program defined as follows : ( here symbols that start with `` @xmath33 '' represent logical variables ) * for each @xmath34 add to @xmath30 the clause + @xmath35 * for each @xmath36 , add to @xmath30 the clause + @xmath37 * for each @xmath38 add to @xmath30 the clause + @xmath39 * for each @xmath40 add to @xmath30 the clause + @xmath41 we can now define the semantics of a role in a state . given a state @xmath22 , the semantics of a role @xmath4 is defined in terms of atoms entailed by the semantic program : * @xmath42\!]_{\mathit{sp({\ensuremath{{\cal p}}\xspace } ) } } } } = \{z | { \ensuremath{\mathit{sp({\ensuremath{{\cal p}}\xspace})}}\xspace } \models m(a , r , z)\}$ ] we extend this semantics to positive role expressions in the natural way as follows : @xmath43\!]_{\mathit{sp({\ensuremath{{\cal p}}\xspace } ) } } } } & = & \{d_1,\ldots , d_n\ } \\ { \ensuremath{[\![\phi_1 \cup \phi_2]\!]_{\mathit{sp({\ensuremath{{\cal p}}\xspace } ) } } } } & = & { \ensuremath{[\![\phi_1]\!]_{\mathit{sp({\ensuremath{{\cal p}}\xspace } ) } } } } \cup { \ensuremath{[\![\phi_2]\!]_{\mathit{sp({\ensuremath{{\cal p}}\xspace } ) } } } } \\ { \ensuremath{[\![\phi_1 \cap \phi_2]\!]_{\mathit{sp({\ensuremath{{\cal p}}\xspace } ) } } } } & = & { \ensuremath{[\![\phi_1]\!]_{\mathit{sp({\ensuremath{{\cal p}}\xspace } ) } } } } \cap { \ensuremath{[\![\phi_2]\!]_{\mathit{sp({\ensuremath{{\cal p}}\xspace})}}}}\end{aligned}\ ] ] consider a state @xmath22 , which might change in time . we are interested in defining a _ constraint _ , which intuitively is a _ query _ that is intended to hold throughout the state changes . to this end , we focus on the class of constraints already considered for the purposes of security analysis in @xcite . these constraints express set containment . a _ constraint _ is an expression of the form @xmath44 , in which @xmath45 is a principal called the _ owner _ of the constraint , and @xmath46 and @xmath47 are positive role expressions . the following definition clarifies that @xmath48 represents set containment . let @xmath49 be a state and @xmath50 be the constraint @xmath44 , we say that * @xmath49 _ satisfies _ @xmath51 ( @xmath52 iff @xmath53\!]_{\mathit{sp({\ensuremath{{\cal p}}\xspace } ) } } } } \subseteq { \ensuremath{[\![{\ensuremath{\varrho}}]\!]_{\mathit{sp({\ensuremath{{\cal p}}\xspace})}}}}$ ] ( @xmath54 _ violates _ @xmath55 otherwise ) constraints of this form can capture many important and intuitive requirements . * consider @xmath56 . this constraint captures a safety requirement that bob must not become a member of @xmath4 . * the constraint @xmath57 captures the availability requirement that alice must be authorized for @xmath4 . * the constraint @xmath58 captures the mutual exclusivity requirement that no one must be authorized for both @xmath59manager and @xmath60controller . [ cols= " < , > " , ] + hence the thesis . theorem [ thm : main ] also shows that , as long as the changes to @xmath22 satisfy ( i ) and ( ii ) , we do not have to recompute the set @xmath61 or the support @xmath62 . technically , this is due to the fact that changes satisfying ( i ) and ( ii ) do not affect @xmath62 ( by lemma [ lem : sprop ] , @xmath62 is still a support of @xmath47 ) , and can only reduce the set @xmath61 ( by lemma [ lem : gprop ] ) . when statements defining roles in @xmath61 are issued , ( i ) is violated , and when statements defining roles in @xmath62 are revoked , ( ii ) is violated . at these times , the constraint must be checked and the sets @xmath61 and @xmath62 must be recomputed . the theorem indicates how a system for monitoring constraints should be deployed : the first step ( mentioned above ) is to check that @xmath22satisfies @xmath63 . while doing this , we can build an appropriate @xmath62 . secondly , we have to build @xmath61 . thirdly , we need to put in place _ monitoring _ of the roles in @xmath62 and in @xmath61 such that each time a statement defining a role in @xmath61 ( resp . @xmath62 ) is added to ( resp . deleted from ) @xmath22 , the constraint owner is warned . when the constraint owner receives a warning he has to ( a ) check whether the constraint still holds , and ( b ) recompute @xmath61 and @xmath62 . * returning to example [ ex : hazmat ] , to monitor @xmath64 , we must monitor revocation of definitions of roles in some @xmath22-support of each member of @xmath65\!]_{\mathit{sp({\ensuremath{{\cal p}}\xspace})}}$ ] for @xmath66.@xmath67 . in this example , @xmath68 is a @xmath22-support of each such member for @xmath66.@xmath67 . we must also monitor additions to @xmath69 , as discussed in example [ ex : gamma ] . if new statements are added defining other roles , no action has to be taken . similarly , if statement ( 10 ) , , were removed , no action would be necessary because @xmath70.@xmath71is not in @xmath62 . * consider now example [ exa : support ] ( ii ) , together with the query @xmath72 . to apply theorem [ thm : main ] , we have to choose one support of @xmath73 for @xmath4 ( the two candidate support are @xmath74 and @xmath75 ) and monitor the roles in it . suppose we choose @xmath76 . suppose we now remove the statement @xmath77 . this does not yield to a violation of the constraint , but we do have to recompute @xmath62 , which now becomes @xmath75 . * finally , it is also instructive to see that a change in @xmath61 might require recomputing @xmath62 , even if it does not entail a violation of the constraint . let @xmath49 be the following set of statements : @xmath78 together with the constraint @xmath79 . this constraint is satisfied and to monitor its evolution we have to monitor the roles in @xmath80 and @xmath81 . now if we add the statement @xmath82 then the constraint owner is warned that a change in @xmath61 has occurred . the constraint owner can check that the constraint is still satisfied in @xmath83 ; however @xmath62 has to be recomputed to take into account that it should be a @xmath84-support of @xmath73 too . the new @xmath62 is @xmath85 . we have defined the _ @xmath22-support _ @xmath62 to be a set of roles . alternatively , we could have defined @xmath62 to be a set of _ [ def : altsupport ] let @xmath4 be a role , @xmath3 be a principal , @xmath49 be a set of statements and and @xmath86 be a set of credentials * we say that @xmath62 is a _ @xmath22-support of _ @xmath3 _ for _ @xmath4 if @xmath87\!]_{\mathit{sp(\sigma)}}}}$ ] . * for @xmath88 , we say that @xmath62 is a _ @xmath22-support of _ @xmath89 _ for _ @xmath4 if @xmath87\!]_{\mathit{sp(\sigma)}}}}$ ] for every @xmath90 . * we say that @xmath62 is a _ @xmath22-support for _ @xmath4 if and only if it is a @xmath49-support of every @xmath87\!]_{\mathit{sp({\ensuremath{{\cal p}}\xspace})}}}}$ ] . monitoring constraint using this definition requires more machinery than using definition [ def : support ] , but it could yield to a more efficient implementation . with this definition one monitors the _ credentials _ and not the _ roles _ which might affect the right hand side of the constraint . therefore , to apply this definition one needs a mechanism for monitoring every single credential of @xmath62 ( which might be difficult ) . [ thm : altmain ] assume that @xmath49 satisfies the constraint @xmath44 . let @xmath62 be a @xmath49-support of @xmath53\!]_{\mathit{sp({\ensuremath{{\cal p}}\xspace})}}}}$ ] for @xmath91 ( according to definition [ def : altsupport ] ) , and let @xmath92 be a ( possibly multistep ) change from @xmath49 to @xmath84 . if 1 . @xmath93 , and 2 . @xmath94 then @xmath95 satisfies @xmath44 as well . the advantage of definition [ def : altsupport ] , is that the hypothesis of theorem [ thm : altmain ] hold more often than those of theorem [ thm : main ] . in other words , using definition [ def : altsupport ] one has to check whether the query still holds and to recalculate @xmath61 and @xmath62 less often than with definition [ def : support ] . the previous section showed how principals in a trust management system can monitor integrity constraints by monitoring changes in the definitions of certain roles . this section considers the problem of monitoring integrity constraints when not all principals in the system agree to assist in monitoring their roles . the idea is to make the assumption that the owners of a certain set of roles are trusted to monitor new statements added to their definitions . we call these the growth - trusted roles and denote them by @xmath96 . similarly , the owners of a set of shrink - trusted roles , denoted @xmath97 , are trusted to monitor statements removed from their definitions . the owners of these roles are trusted to test whether changes made to untrusted roles could violate the constraint and , if so , to signal that potential violation . we call the pair @xmath98 a _ role monitor _ because it indicates the roles that can be monitored with respect to growth and shrinkage . [ def : reachable ] in the presence of a role monitor @xmath99 , we say that @xmath84 is @xmath100-reachable from @xmath49 if @xmath84 can be obtained from @xmath22without adding any statements defining roles in @xmath96or removing any statements defining roles in @xmath97 . that is to say , @xmath101 and @xmath102 . the problem we address is to monitor whether the system ever enters a state @xmath22from which some reachable @xmath84 violates @xmath63 . this problem is closely related to the security analysis problem @xcite , which also is defined in terms of a role monitor @xmath98 , although in that context it is called a restriction rule . in security analysis , the definitions of roles in @xmath96are assumed not to grow and those of roles in @xmath97 , not to shrink ; the security analysis problem is to determine whether other changes to the policy state could cause a constraint to become violated . in @xcite it was shown that this problem is decidable ( @xmath103 ) for @xmath2 over the class of constraints we consider here , and that it is polynomial for an important subclass of those constraints . what has not been shown before , and what we show in this section , is how to identify subsets of @xmath96and @xmath97that need to be monitored so that security analysis can be used to maintain integrity constraints . in the rest of this section , we introduce alternative semantics that can be used to answer questions about policy states that are reachable through changes to the definitions of untrusted roles . we then formalize sets of roles that must be monitored and show that monitoring these roles is sufficient . finally , we provide a method for monitoring integrity constraints when not all principals in the system are trusted to assist the process . we now recall two non - standard semantics for a policy state @xmath22and role monitor @xmath100 . these were introduced @xcite for computing the lower and upper bounds on role memberships under the assumption that the definition of roles in @xmath96do not grow and the definition of roles in @xmath97do not shrink . we first recall the lower - bound program for a state @xmath22and a restriction @xmath100 ; this program enables one to compute the lower - bounds of every role . [ def : lb ] given @xmath22and @xmath100 , the _ lower - bound program _ for them , @xmath104 , is constructed as follows : * for each @xmath105 in @xmath106 , add + @xmath107(a , r , d ) * for each @xmath108 in @xmath109 , add + @xmath110 * for each @xmath111 in @xmath109 , add + @xmath112 * for each @xmath113 in @xmath109 , add + @xmath114 . we now recall the upper - bound program for a state @xmath22and a role monitor @xmath100 . this program enables one to simulate the upper - bound of any role . [ def : ub ] given @xmath22and @xmath115 , their upper - bound program , @xmath116 , is constructed as follows . ( @xmath117 is a special principal symbol not occurring in @xmath22 , @xmath100 , or any query @xmath55 . ) * add @xmath118 * for each @xmath119 , add + @xmath120 * for each @xmath105 in @xmath22 , add + @xmath121 * for each @xmath108 in @xmath22 , add + @xmath122 * for each @xmath111 in @xmath22 , add + @xmath123 * for each @xmath113 in @xmath22 , add + @xmath124 the rules @xmath125 to @xmath126 follow from the meanings of the four types of statements and are similar to the semantic program construction in definition [ def : sp ] . the rule @xmath127 means that for any role @xmath4 not in @xmath96 , the upper - bound of @xmath4 contains every principal . the rule @xmath128 means that for any role name @xmath129 , the upper - bound of @xmath130 contains every principal . this is so because @xmath117 does not appear in @xmath131 . the rule @xmath128 is needed because given @xmath132 , where @xmath133 and @xmath134 , we should ensure that the upper - bound of @xmath4 contains every principal . we define : @xmath135\!]_{\emph{ub}({\ensuremath{{\cal p}}\xspace } ) } } } & = & \{z \ | \ { \ensuremath{\mathit{ub}}\xspace}({\ensuremath{{\cal p}}\xspace } ) \models m(a , r , z)\}\\ { \ensuremath{[\![a.r]\!]_{\emph{lb}({\ensuremath{{\cal p}}\xspace } ) } } } & = & \{z \ | \ { \ensuremath{\mathit{lb}}\xspace}({\ensuremath{{\cal p}}\xspace } ) \models m(a , r , z)\}\end{aligned}\ ] ] and by definition we have that [ rem : semantics ] * if @xmath136 then @xmath42\!]_{\emph{lb}({\ensuremath{{\cal p}}\xspace } ) } } } = \emptyset$ ] . * if @xmath137 then @xmath42\!]_{\emph{ub}({\ensuremath{{\cal p}}\xspace } ) } } } = { \ensuremath{\sf principals({\ensuremath{{\cal p}}\xspace})}\xspace}{\ensuremath{\:\cup\:}}\ { \top \}$ ] . the next theorem gives the link between the two new semantics and the problem of checking that a constraint is satisfied in all reachable @xmath84 . [ thm : strongsemantics ] let @xmath100 be a role monitor , @xmath22be a state , and @xmath63 be a containment constraint . * if @xmath53\!]_{\emph{ub}({\ensuremath{{\cal p}}\xspace } ) } } } \subseteq { \ensuremath{[\![{\ensuremath{\varrho}}]\!]_{\emph{lb}({\ensuremath{{\cal p}}\xspace})}}}$ ] then @xmath138 for each @xmath84 reachable from @xmath22 , * if either @xmath139 or @xmath91 is _ static _ ( i.e. , it is a set of principals ) then @xmath138 for each @xmath84 reachable from @xmath22implies that @xmath53\!]_{\emph{ub}({\ensuremath{{\cal p}}\xspace } ) } } } \subseteq { \ensuremath{[\![{\ensuremath{\varrho}}]\!]_{\emph{lb}({\ensuremath{{\cal p}}\xspace})}}}$ ] then @xmath138 for each @xmath84 reachable from @xmath22implies that @xmath53\!]_{\emph{ub}({\ensuremath{{\cal p}}\xspace } ) } } } \subseteq { \ensuremath{[\![{\ensuremath{\varrho}}]\!]_{\emph{lb}({\ensuremath{{\cal p}}\xspace})}}}$ ] . ] . we now proceed as in the previous section , by identifying the roles we have to monitor . in the light of theorem [ thm : strongsemantics ] , given a state @xmath22 , a role monitor @xmath100 , and a role @xmath4 , we want to isolate a set @xmath140 of roles we have to monitor , as they might affect the _ growth _ of @xmath141\!]_{\emph{ub}({\ensuremath{{\cal p}}\xspace})}$ ] . one might think that when some roles are untrusted , we need only restrict @xmath142 to the @xmath131-roles ( or to check that @xmath143 ) . the following example shows that this is not adequate . consider the constraint @xmath79 , where @xmath4 is defined by @xmath144 @xmath4 depends on @xmath145 , @xmath146 and @xmath147 ( which are in @xmath142 ) , and , if we used the method of the previous section , we would have to monitor all three of them . we now make two observations about monitoring when it is not possible to monitor all three roles . first , if @xmath147 is not in @xmath96 , we can not monitor it . this implies that there is no point in monitoring @xmath146 either , as it directly depends on @xmath147 . second if @xmath146 is not in @xmath96 , there is no point in monitoring it nor in monitoring @xmath147 ( which can only influence @xmath4 via @xmath146 ) . to cope with this we now define the @xmath22-core of @xmath96 , which intuitively contains those role of @xmath96 which additionally do not fully depend on an untrusted role . let @xmath22be a state and @xmath96 be a set of roles . the _ @xmath22-core of _ @xmath96 , @xmath148 , is the maximal subset of @xmath96such that * if @xmath36 , and @xmath149 , then @xmath150 * if @xmath38 , and @xmath151 , then @xmath150 . * if @xmath38 , and @xmath152\!]_{\emph{ub}({\ensuremath{{\cal p}}\xspace})}}}$ ] such that @xmath153 , then @xmath150 . * if @xmath154 , and for every @xmath155 , @xmath156 , then @xmath157 . the following proposition is proved in the appendix . [ pro : noncore ] let @xmath22be a set of statements and @xmath96 be a set of roles . * if @xmath158 , then @xmath42\!]_{\emph{ub}({\ensuremath{{\cal p}}\xspace } ) } } } = { \ensuremath{\sf principals({\ensuremath{{\cal p}}\xspace})}\xspace}{\ensuremath{\:\cup\:}}\{\top \}$ ] . we now construct the set of roles that must be monitored for new definitions to detect growth in a role s membership . [ def : gpgr ] let @xmath159 be a role in @xmath148 , @xmath100 be a role monitor , and @xmath22be a state ; @xmath160 is the least set satisfying the following : * if @xmath161 , @xmath162 . * if @xmath163 , and @xmath164 , then @xmath165 . * if @xmath163 and @xmath166 , then @xmath167 and @xmath168 for all @xmath169\!]_{\emph{ub}({\ensuremath{{\cal p}}\xspace})}}}$ ] * if @xmath163 and @xmath170 , then , for each @xmath171 $ ] if @xmath172 , @xmath173 . it is easy to prove by a simple induction on the steps in the iterative construction of @xmath174 that @xmath175 we now have the counterpart of lemma [ lem : gprop ] . [ lem : gpropsafety ] assume @xmath176\!]_{\emph{ub}({\ensuremath{{\cal p}}\xspace})}}}$ ] . let @xmath99 be a role monitor , @xmath177 , where @xmath178 , then 1 . @xmath42\!]_{\emph{ub}({\ensuremath{{\cal p}}\xspace } ) } } } = { \ensuremath{[\![a.r]\!]_{\emph{ub}({\ensuremath{{\cal p}}\xspace}')}}}$ ] , and 2 . @xmath179 . moreover , if @xmath84 is obtained from @xmath22by ( a ) adding zero or more statements whose head is not in @xmath180 , and ( b ) removing zero or more statements , then 1 . @xmath42\!]_{\emph{ub}({\ensuremath{{\cal p}}\xspace } ) } } } \supseteq { \ensuremath{[\![a.r]\!]_{\emph{ub}({\ensuremath{{\cal p}}\xspace}')}}}$ ] , and 2 . @xmath181 . _ proof ( sketch ) . _ the result follows by using reasoning similar to that used for proving lemma [ lem : gprop ] . to handle the right hand side of the constraints we simply have to generalize lemma [ lem : sprop ] in the obvious way by taking into account the presence of the role monitor . the proof of this lemma is also identical to that of lemma [ lem : sprop ] [ lem : spropsafety ] let @xmath98 be a role monitor , @xmath4 be a role , @xmath3 be a principal , @xmath49 be a state and @xmath62 be a @xmath22-support of @xmath3 for @xmath4 _ such that @xmath182_. then 1 . @xmath87\!]_{\emph{lb}({\ensuremath{{\cal p}}\xspace})}}}$ ] moreover , if @xmath84 is obtained from @xmath22by ( a ) removing zero or more statements whose head is not in @xmath62 , and ( b ) adding zero or more statements , then 1 . @xmath62 is a @xmath84-support for @xmath4 , and therefore 2 . @xmath87\!]_{\emph{lb}({\ensuremath{{\cal p}}\xspace}')}}}$ ] . recall that by remark [ rem : semantics ] , if @xmath136 then we have that @xmath42\!]_{\emph{lb}({\ensuremath{{\cal p}}\xspace } ) } } } = \emptyset$ ] . consequently , it is easy to show that if @xmath87\!]_{\emph{lb}({\ensuremath{{\cal p}}\xspace})}}}$ ] , then there exists a @xmath22-support of @xmath3 for @xmath4 consisting of roles that are in @xmath183 . we can now prove the result we were aiming at . differently from the case in which all roles were trusted , we now want to check that @xmath63 holds in any @xmath100-reachable state @xmath84 . the additional problem here is we can not rely on the cooperation of the roles that are not in @xmath131 ( resp . @xmath183 ) in monitoring the constraint and telling the constraint owner when a statement defining a role in @xmath61 is added ( resp . a statement defining a role in @xmath62 is removed ) . because of this we refer to two `` pessimistic '' semantics , @xmath53\!]_{\emph{ub}({\ensuremath{{\cal p}}\xspace})}}}$ ] and @xmath184\!]_{\emph{lb}({\ensuremath{{\cal p}}\xspace})}}}$ ] , and we check if @xmath53\!]_{\emph{ub}({\ensuremath{{\cal p}}\xspace } ) } } } \subseteq { \ensuremath{[\![{\ensuremath{\varrho}}]\!]_{\emph{lb}({\ensuremath{{\cal p}}\xspace})}}}$ ] . if this does not hold , then , by theorem [ thm : strongsemantics ] the chance is high that in some reachable @xmath84 the constraint is violated . if @xmath53\!]_{\emph{ub}({\ensuremath{{\cal p}}\xspace } ) } } } \subseteq { \ensuremath{[\![{\ensuremath{\varrho}}]\!]_{\emph{lb}({\ensuremath{{\cal p}}\xspace})}}}$ ] does hold , then we can apply the following : [ thm : mainsafety ] let @xmath98 be a role monitor . assume that @xmath53\!]_{\emph{ub}({\ensuremath{{\cal p}}\xspace } ) } } } \subseteq { \ensuremath{[\![{\ensuremath{\varrho}}]\!]_{\emph{lb}({\ensuremath{{\cal p}}\xspace})}}}$ ] . let @xmath62 be a @xmath49-support of @xmath53\!]_{\emph{ub}({\ensuremath{{\cal p}}\xspace})}}}$ ] for @xmath91 such that @xmath185 , and let @xmath186 be a ( possibly multistep ) change from @xmath49 to @xmath84 . if 1 . @xmath187 , and 2 . @xmath188 then @xmath53\!]_{\emph{ub}({\ensuremath{{\cal p}}\xspace } ' ) } } } \subseteq { \ensuremath{[\![{\ensuremath{\varrho}}]\!]_{\emph{lb}({\ensuremath{{\cal p}}\xspace}')}}}$ ] . + _ proof_. take any @xmath189\!]_{\emph{ub}({\ensuremath{{\cal p}}\xspace}')}}}$ ] , by lemma [ lem : gpropsafety ] , @xmath189\!]_{\emph{ub}({\ensuremath{{\cal p}}\xspace})}}}$ ] . by assumption , @xmath190\!]_{\emph{lb}({\ensuremath{{\cal p}}\xspace})}}}$ ] , and by lemma [ lem : spropsafety ] , @xmath190\!]_{\emph{lb}({\ensuremath{{\cal p}}\xspace}')}}}$ ] . hence the thesis . because of theorem [ thm : mainsafety ] , in the presence of untrusted roles we can deploy a monitoring procedure very similar to that described after theorem [ thm : main ] . first we check that @xmath53\!]_{\emph{ub}({\ensuremath{{\cal p}}\xspace } ) } } } \subseteq { \ensuremath{[\![{\ensuremath{\varrho}}]\!]_{\emph{lb}({\ensuremath{{\cal p}}\xspace})}}}$ ] holds nor @xmath47 is static , it is possible that @xmath53\!]_{\mathit{sp({\ensuremath{{\cal p}}\xspace } ' ) } } } } \subseteq { \ensuremath{[\![{\ensuremath{\varrho}}]\!]_{\mathit{sp({\ensuremath{{\cal p}}\xspace}')}}}}$ ] for all @xmath84 reachable from @xmath22 . however , in general , for the class of constraints we consider , determining this is * pspace*-hard @xcite , i.e. , intractable . thus , our technique makes an efficient conservative approximation for the more general constraints we consider . ] . while doing this , we compute a @xmath22-support @xmath62 of @xmath53\!]_{\emph{ub}({\ensuremath{{\cal p}}\xspace})}}}$ ] for @xmath91this time a @xmath62 such that @xmath182 . second , we have to build @xmath191 . third , we monitor the roles in @xmath62 and in @xmath191 so that each time a statement defining a role in @xmath191 ( resp . @xmath62 ) is added to ( resp . deleted from ) @xmath49 , the constraint owner is warned . when the constraint owner receives a warning , he has to ( a ) check whether @xmath53\!]_{\emph{ub}({\ensuremath{{\cal p}}\xspace } ) } } } \subseteq { \ensuremath{[\![{\ensuremath{\varrho}}]\!]_{\emph{lb}({\ensuremath{{\cal p}}\xspace})}}}$ ] still holds , and ( b ) recompute @xmath191 and @xmath62 . reconsider again example [ ex : hazmat ] . suppose that @xmath192 is ( the only role ) not in @xmath131 , then we have that @xmath193 . therefore + @xmath194 + nonetheless , if @xmath195 we have that @xmath196\!]_{\emph{ub}({\ensuremath{{\cal p}}\xspace})}}}\\ \subseteq & { \ensuremath{[\![{\ensuremath{\mathit atf}\xspace}.{\ensuremath{\mathit hazmatdb}\xspace}]\!]_{\emph{lb}({\ensuremath{{\cal p}}\xspace } ) } } } \end{array}\ ] ] so by theorem [ thm : strongsemantics ] we know that the constraint @xmath197 is satisfied in all reachable @xmath84 . by theorem [ thm : mainsafety ] , if the two roles @xmath198.@xmath199 , and @xmath66.@xmath200 , prompt a warning when a statement defining one of them is added and the role @xmath66.@xmath67gives a warning when one of its statement is removed , then the constraint needs to be re - checked only when a warning is given . in that case , we also have to recompute @xmath62 and @xmath201 . theorem [ thm : mainsafety ] guarantees that no matter which changes are made to @xmath49 , until a warning is given , we still have that every reachablealso changes the reachability relation , i.e. , the set of reachable @xmath84s . ] @xmath84 satisfies the constraint . in database theory , an integrity constraint is a query that must remain _ true _ after the database has been updated . originally , integrity constraints were introduces to prevent incorrect updates and to check the database for integrity . nevertheless , integrity constraints have later been used for a number of purposes , ranging from query optimization to view updating . we refer to @xcite for illustrative examples of the uses of integrity constraints in deductive databases . in section [ sec : preliminaries ] , we listed several papers presenting various trust management systems . none of these incorporates a notion of integrity constrains . the work in trust management that is most closely related is @xcite . as we discussed at the beginning of section [ sec : safety ] , that work is complimentary to ours . it studies the problem of determining , given a state @xmath22 , a role monitor @xmath100 , and a constraint @xmath202 , whether there is a reachable state in which @xmath202 is violated . by contrast , we analyze the problem of which roles must have their definitions monitored to detect when such a @xmath22is entered . we introduce the use , monitoring , and enforcement of integrity constraints in trust management - style authorization systems . we consider the portions of the policy state that must be monitored to detect violations of integrity constraints . we also address the extra difficulty that not all participants in a trust management system can be trusted to assist in such monitoring , and show how many integrity constraints can be monitored in a conservative manner so that trusted participants detect and report if the system enters a policy state from which evolution in unmonitored portions of the policy could lead to a constraint violation . we thank pieter hartel and ha manh tran for their precious help and the anonymous referees for their comments . 10 k. r. apt . . prentice hall , 1997 . m. blaze , j. feigenbaum , j. ioannidis , and a. keromytis . the keynote trust - management system , version 2 . ietf rfc 2704 , september 1999 . m. blaze , j. feigenbaum , j. ioannidis , and a. keromytis . the role of trust management in distributed systems security . in j. vitek and c. jensen , editors , _ secure internet programming : security issues for mobile and distributed objects _ , volume 1603 of _ lncs _ , pages 185210 . springer - verlag , 1999 . m. blaze , j. feigenbaum , and j. lacy . decentralized trust management . in ieee computer society press , editor , _ proc . 1996 ieee symposium on security and privacy _ , pages 164173 , 1996 . d. clarke , j - e . elien , c. ellison , m. fredette , a. morcos , and r. rivest . certificate chain discovery in spki / sdsi . , 9(4):285322 , 2001 . s. k. das . . addison - wesley , 1992 . c. ellison , b. frantz , b. lampson , r. rivest , b. thomas , and t. ylonen . certificate theory . ietf rfc 2693 , september 1999 . s. etalle and w. h. winsborough . integrity constraints in trust management extended abstract . in g - j . ahn , editor , _ proc . 10th acm symp . on access control models and technologies ( sacmat ) _ , page to appear . acm press , 2005 . p. godfrey , j. grant , j. gryz , and j. minker . integrity constraints : semantics and applications . in j. chomicki and g. saake , editors , _ logics for databases and information systems _ , pages 265306 . kluwer academic , 1998 . c. gunter and t. jim . policy - directed certificate retrieval . , 30(15):16091640 , september 2000 . t. jim . : a trust management system with certified evaluation . in _ proceedings of the 2001 ieee symposium on security and privacy _ , pages 106115 . ieee computer society press , may 2001 . n. li , j. mitchell , and w. winsborough . design of a role - based trust - management framework . in _ proc . of the ieee symposium on research in security and privacy _ , pages 114130 . ieee computer society , technical committee on security and privacy , ieee computer society press , 2002 . n. li , j. mitchell , and w. winsborough . beyond proof of compliance : security analysis in trust management . , 2004 . to appear . n. li and j. c. mitchell . datalog with constraints : a foundation for trust management languages . in v. dahl and p. wadler , editors , _ 5th international symposium on practical aspects of declarative languages ( padl03 ) _ , volume 2562 of _ lncs _ , pages 5873 . springer - verlag , 2003 . n. li , w. winsborough , and j. mitchell . distributed credential chain discovery in trust management . , 11(1):3586 , 2003 . ninghui li , benjamin n. grosof , and joan feigenbaum . : a logic - based approach to distributed authorization . , 6(1):128171 , february 2003 . r. rivest and b. lampson . a simple distributed security infrastructure , october 1996 . available at http://theory.lcs.mit.edu/@xmath203rivest/sdsi11.html . r. sandhu , e. coyne , h. feinstein , and c. youman . role - based access control models . , 29(2):3847 , february 1996 . s. weeks . understanding trust management systems . in _ proceedings of 2001 ieee symposium on security and privacy _ , pages 94105 . ieee computer society press , 2001 . * lemma [ lem : gprop ] * let @xmath177 , where @xmath204 , then 1 . @xmath42\!]_{\mathit{sp({\ensuremath{{\cal p}}\xspace } ) } } } } = { \ensuremath{[\![a.r]\!]_{\mathit{sp({\ensuremath{{\cal p}}\xspace}')}}}}$ ] , and 2 . @xmath205 . moreover , if @xmath84 is obtained from @xmath22by ( a ) adding zero or more statements whose head is not in @xmath206 , and ( b ) removing zero or more statements , then 1 . @xmath42\!]_{\mathit{sp({\ensuremath{{\cal p}}\xspace } ) } } } } \supseteq { \ensuremath{[\![a.r]\!]_{\mathit{sp({\ensuremath{{\cal p}}\xspace}')}}}}$ ] , and 2 . @xmath207 . _ proof_. \(a ) let @xmath208 , and @xmath209 . first , summarize some logic - programming notation : we denote by @xmath210 the herbrand base of @xmath211 ( and @xmath212 ) , consisting of the set of all ground ( variable - free ) atoms . @xmath213 denotes the set of all ground instances of clauses in @xmath211 . the usual @xmath214 operator is defined as follows : let @xmath215 , then @xmath216 . as usual , we define @xmath217 , and @xmath218 . by well - known results ( see e.g. , @xcite ) , since @xmath211 contains no function symbols , for some @xmath219 we have that @xmath220 now we define the lp - counterpart of @xmath142 : @xmath221 and the complement @xmath222 . furthermore , let @xmath223 and @xmath224 be two sets of ground atoms such that @xmath225 , and @xmath226 . by the monotonicity of @xmath214 , we have that @xmath227 we now want to show that @xmath228 we proceed by contradiction and assume that there exists @xmath229 such that @xmath230 since @xmath231 , there exists a ground instance @xmath232 of a clause @xmath233 such that @xmath234 . since @xmath235 , @xmath233 . therefore @xmath236 . we now want to show that @xmath237 since @xmath238 , this will demonstrate that @xmath239 , and therefore that @xmath240 , contradicting ( [ eq : absurd ] ) . we distinguish two cases according to the kind of statement from which @xmath241 is generated . case 1 : @xmath241 is the lp - translation of a simple inclusion or intersection inclusion ( not a linking inclusion ) . then @xmath242 by definition [ def : gp ] . case 2 : @xmath241 is the lp - translation of a linking inclusion ( linked role ) . then @xmath232 has the form @xmath243 . by definition [ def : gp ] , @xmath244 . since @xmath238 , and @xmath245 , we have that @xmath246 . since @xmath226 , then @xmath247\!]_{\mathit{sp({\ensuremath{{\cal p}}\xspace})}}}}$ ] . therefore , again by definition [ def : gp ] , @xmath248 , proving ( [ eq : nextstep ] ) ( which in turn contradicts [ eq : absurd ] ) . now that we have proven ( [ eq : tpprimeingamma ] ) , since for each @xmath31 we have that @xmath249 , from ( [ eq : tpprimebigger ] ) , ( [ eq : tpprimeingamma ] ) and a straightforward inductive reasoning it follows that , for each @xmath31 , @xmath250 since the least model of @xmath212 and @xmath211 is the least fixpoint of these continuous operators on a finite lattice , this demonstrates that @xmath251 . since by definition @xmath252 it follows that @xmath42\!]_{\mathit{sp({\ensuremath{{\cal p}}\xspace } ) } } } } = { \ensuremath{[\![a.r]\!]_{\mathit{sp({\ensuremath{{\cal p}}\xspace}')}}}}$ ] . hence the thesis . \(b ) since @xmath204 , @xmath253 is not reachable from @xmath4 . so removing @xmath254does not alter the reachability from @xmath4 . \(c ) and ( d ) first notice that , by construction , @xmath255 now , suppose that we have a chain @xmath256 , where each @xmath257 is obtained from @xmath258 by either adding a statement whose head is not in @xmath206 or removing a statement . we now show by induction on @xmath155 that for each @xmath259 $ ] : @xmath42\!]_{\mathit{sp({\ensuremath{{\cal p}}\xspace } ) } } } } \supseteq { \ensuremath{[\![a.r]\!]_{\mathit{sp({\ensuremath{{\cal p}}\xspace}_{i})}}}}$ ] and @xmath260 , which imply the thesis . the base case is trivial , as @xmath261 , for the inductive case we have two subcases : case 1 . if @xmath257 is obtained from @xmath258 by adding a statement @xmath254such that @xmath262 , then by the inductive hypothesis @xmath263 , and , by statements ( a ) and ( b ) we have that @xmath42\!]_{\mathit{sp({\ensuremath{{\cal p}}\xspace}_{i } ) } } } } = { \ensuremath{[\![a.r]\!]_{\mathit{sp({\ensuremath{{\cal p}}\xspace}_{i+1})}}}}$ ] and @xmath264 , and the result follows from the inductive hypothesis . if @xmath257 is obtained from @xmath258 by removing a statement , then the result follows from the monotonicity of @xmath42\!]_{\mathit{sp({\ensuremath{{\cal p}}\xspace}_{i})}}}}$ ] ( [ eq : monotonicity ] ) , and ( [ eq : gdiminishes ] ) . * proposition [ pro : noncore ] * let @xmath22be a set of statements and @xmath96 be a set of roles . if @xmath158 , then @xmath42\!]_{\emph{ub}({\ensuremath{{\cal p}}\xspace } ) } } } = { \ensuremath{\sf principals({\ensuremath{{\cal p}}\xspace})}\xspace}{\ensuremath{\:\cup\:}}\{\top \}$ ] . + _ proof_. consider the following closure operator on sets of roles ( @xmath265 ) . let @xmath266 be a set of roles . @xmath267\!]_{\emph{ub}({\ensuremath{{\cal p}}\xspace } ) } } } \mbox { such that } b.r_2 \in \delta \ } \\ & & { \ensuremath{\:\cup\ : } } & \ { a.r \ | \ a.r { \ensuremath{\longleftarrow}}b_1.r_1 { \ensuremath{\:\cap\:}}\ldots b_n.r_n \in { \ensuremath{{\cal p}}\xspace}\mbox { and } \forall i \in [ 1,n]\ b_i.r_i \in \delta \ } \end{array}\ ] ] it is easy to see that @xmath148 is by construction exactly the least fixpoint of @xmath268 containing @xmath269 , the complement of @xmath131 . now , define @xmath270 , and @xmath271 . since @xmath268 is monotonically increasing , and since @xmath272 is finite , we have that , for some @xmath219 . @xmath273 now , by definition , for every @xmath274 , @xmath42\!]_{\emph{ub}({\ensuremath{{\cal p}}\xspace } ) } } } = { \ensuremath{\sf principals({\ensuremath{{\cal p}}\xspace})}\xspace}{\ensuremath{\:\cup\:}}\{\top \}$ ] . + by the definition of @xmath268 , it is straightforward to check that this implies that for every @xmath275 , @xmath42\!]_{\emph{ub}({\ensuremath{{\cal p}}\xspace } ) } } } = { \ensuremath{\sf principals({\ensuremath{{\cal p}}\xspace})}\xspace}{\ensuremath{\:\cup\:}}\{\top \}$ ] . + by iterating this reasoning it is straightforward to check that this implies that for every @xmath276 , @xmath42\!]_{\emph{ub}({\ensuremath{{\cal p}}\xspace } ) } } } = { \ensuremath{\sf principals({\ensuremath{{\cal p}}\xspace})}\xspace}{\ensuremath{\:\cup\:}}\{\top \}$ ] . + the thesis follows from ( [ eq : fixpoint ] ) . we now show how one can compute the support in bottom - up way . we do this by defining a semantics : @xmath277 for which it holds that if @xmath278 then @xmath62 is a minimal @xmath49-support of @xmath3 in @xmath4 . the construction is parametric wrt the partial order used to define minimality . xx = xx = xx = xx = xx = xx = xx = xx = xx = * init phase * + for each role @xmath4 , @xmath286 + * repeat * + for each role @xmath4 , do @xmath287 + for each @xmath288 do + if @xmath289 then + _ remove from @xmath290(a.r ) all triples subsumed by @xmath291 _ + @xmath292 + if @xmath293 then + for each @xmath294 do + if @xmath295 is not subsumed by any triple in @xmath296 then + _ remove from @xmath290(a.r ) all triples subsumed by @xmath297 _ + @xmath298 + if @xmath299 then + for each @xmath300 do + for each @xmath301 do + if @xmath302 is not subsumed by any triple in @xmath296 then + _ remove from @xmath290(a.r ) all triples subsumed by @xmath302 _ + @xmath303 + if @xmath304 then + for each @xmath305 do + if , for some @xmath306 , @xmath307 @xmath308 then + if @xmath302 is not subsumed by any triple in @xmath296 then + _ remove from @xmath290(a.r ) all triples subsumed by @xmath302 _ + @xmath303 + * until * for each role @xmath4 , @xmath309 _ proof . _ ( @xmath313 ) assume @xmath312 is a minimal set of roles such that @xmath314\!]_{\mathit{sp({\ensuremath{{\cal p}}\xspace}|_{\sigma_0})}}}}$ ] . we show by induction on the construction of @xmath315 that for all @xmath316 and for each @xmath317 , if @xmath318 , then at some stage in the execution of the algorithm , for some @xmath155 and @xmath62 , @xmath319 with @xmath320 . the desired result then follows by taking @xmath321 , by using the fact , shown below in the second part of the proof , that @xmath322 implies @xmath323 , and by using the minimality of @xmath312 . _ step . _ we assume the hypothesis holds for @xmath316 and show that it holds for @xmath325 . we proceed by case analysis of the clause used to add @xmath326 to @xmath327 . we show here only the case of linking inclusion ; the other cases are similar . _ case : _ @xmath328 . by definition of @xmath329 , there exists @xmath1 such that @xmath330 . so by induction hypothesis , there exist @xmath331 such that @xmath332 , @xmath333 , and @xmath301 by some stage in the execution . consider the first such stage . in the following iteration , either @xmath334 already contains a triple that subsumes @xmath335 , or else this triple is added . in either case , at the end of the iteration , @xmath334 contains a triple that subsumes @xmath336 , for all @xmath337 . ( note @xmath338 . ) ( @xmath339 ) we show by induction on @xmath155 that if @xmath340 , then @xmath341 . this direction of the theorem then follows because , by the other direction , all minimal @xmath22-support are in @xmath296 , and the algorithm removes all entries that are subsumed by other entries . _ we assume the hypothesis holds for all @xmath347 and show that it holds for @xmath348 . we proceed by case analysis of the statement used to add @xmath349 to @xmath296 . we show here only the case of linking inclusion ; the other cases are similar . @xmath111 . in this case there are @xmath350 , @xmath306 , @xmath351 , @xmath307 , and @xmath1 such that @xmath352 , @xmath301 , @xmath353 , and @xmath354 . by induction hypothesis , @xmath355 and @xmath356 . by monotonicity of @xmath329 in @xmath211 , it follows that @xmath357 . consider the first @xmath316 such that @xmath358 . because @xmath359 is in @xmath360 , it follows that @xmath361 , the latter being a subset of @xmath362 . it must be acknowledged that the algorithm given here may construct a value for @xmath290 whose size is combinatorial in the size of @xmath22 . in practice , a variant of this algorithm should be used in which a small constant number of entries in @xmath296 are stored for each @xmath314\!]_{\mathit{sp({\ensuremath{{\cal p}}\xspace})}}}}$ ] .

sn 1999aw was discovered during the first campaign of the nearby galaxies supernova search ( _ ngss _ ) project . this luminous , slow - declining ( ) type ia supernova was noteworthy in at least two respects . first , it occurred in an extremely low luminosity host galaxy that was not visible in the template images , nor in initial subsequent deep imaging . secondly , the photometric and spectral properties of this supernova indicate that it very likely was similar to the subclass of type ia supernovae whose prototype is sn 1999aa . this paper presents the _ bvri _ and _ jhk _ lightcurves of sn 1999aw ( through days past maximum light ) , as well as several epochs of optical spectra . from these data we calculate the bolometric light curve , and give estimates of the luminosity at maximum light and the initialni mass . in addition , we present deep _ bvi _ images obtained recently with the baade 6.5-meter telescope at las campanas observatory which reveal the remarkably low - luminosity host galaxy .

astro-ph0207409

type ia supernovae ( sne ia ) offer arguably the most precise method to measure cosmological distances . over the last ten years , these highly luminous explosions have been used to determine distances accurate to @xmath6 , despite the fact that we know little about their progenitors . these objects show considerable uniformity in their absolute _ b _ magnitudes at maximum light with an intrinsic dispersion of less than 0.4 mag ( hamuy et al . this scatter is greatly reduced by the application of empirical relations linking the luminosity at maximum to the width , or decay rate , of the lightcurve ( luminosity - width relations , or _ more luminous sne ia decline in brightness after maximum at slower rates than less luminous sne ia . the @xmath0@xmath1 relation ( phillips 1993 , hamuy et al . 1996a , phillips et al . 1999 ) , which is a measure of the decay in the _ b _ band lightcurve from peak to 15 days after peak , has reduced the scatter around the hubble law to less than 0.2 mag , making it a powerful tool in using sne ia at high redshifts to investigate cosmological parameters . including reddening corrections further decreases the dispersion in the hubble diagram to 0.14 mag ( phillips et al . an equally effective method developed by riess , press , & kirshner ( 1996 ) uses the lightcurve shapes in multiple passbands to simultaneously estimate the sn ia luminosity and amount of extinction / reddening . this mlcs method has demonstrated it can produce hubble diagrams with dispersions of only 0.12 mag ( 6% in distance ) . attention is now being directed to the possible systematic errors involved in using these objects as high redshift distance indicators . part of the challenge lies in untangling the dispersion of sne ia lightcurve widths from possible sub - populations of type ia supernovae . with more discoveries of nearby supernovae in various host environments , and the development of new spectroscopic and photometric techniques for isolating these sub - populations , we may soon be able to determine @xmath7 and @xmath8 with lower systematic uncertainties as well as hone in on the progenitors of sne ia . over the past three years , the nearby galaxies supernova search team _ ( ngss ) _ has conducted successful search campaigns for supernovae of all types using the kitt peak national observatory s 36-inch telescope and the mosaic north camera ( mosaic i ) . each of our 5 - 8 night campaigns have allowed us to search @xmath9 250 fields ( each nearly 1@xmath10 square ) along the celestial equator and out of the galactic plane to limiting magnitudes of @xmath11 . at the project s end , we had searched nearly 750 fields and discovered 42 supernovae . the goals of this project have been to understand supernova rates ( for all sne types ) in both field and galaxy cluster environments , to investigate correlations of sn type with host galaxy environments , and to increase knowledge of observationally rare and peculiar sn types through increased detailed photometric and spectroscopic observations . further information concerning the _ ngss _ project goals and methods will be described in a forthcoming paper ( strolger et al . 2003b , in preparation , hereafter referred to as paper 2 ) . sn 1999aw was one of the supernovae discovered during the first of the _ ngss _ campaigns ( feb . 24 , and mar . 04 - mar . 09 , 1999 ) which was conducted in co - operation with the supernova cosmology project ( _ scp _ ; aldering 2000 , nugent & aldering 2000 ) . initial discovery and confirmation images surprisingly showed a bright new object in a location which was devoid of galaxies on the template images taken only a few weeks earlier . spectroscopic and photometric evidence show sn 1999aw is a member of an intriguing subclass of type ia supernovae , characterized spectroscopically by sn 1999aa ( li et al . these unusual supernovae have lightcurve shapes that are slightly different from those of normal type ia ( strolger et al . 2003a , also in preparation ; see also section [ subsec : templates ] ) . these differences , although not yet completely understood , may be key in understanding not only the physical processes of this subclass , but of all sne ia , and may place limits on models of type ia progenitors . in section [ sec : discovery ] we discuss the discovery , confirmation , and classification of sn 1999aw . in section [ sec : spectroscopy ] we present several epochs of optical spectra and discuss their similarity to 1999aa - like supernovae . in section [ sec : lcurves ] we present the optical and infrared photometry , and the calibrations and corrections . in section [ sec : analysis ] we show the photometric similarity to 1991t/1999aa sne , determine the bolometric lightcurve , estimate the luminosity at maximum light , and estimate the initial @xmath5ni mass . in section [ sec : galaxy ] we present the photometry of the host galaxy and discuss the host environment . sn 1999aw was discovered in search images obtained on ut 1999 mar . 9 with j2000.0 coordinates r.a . = 11h01m36.37s , decl . = @xmath12060631.6 . no star was seen at its location in images obtained on ut 1999 feb . 23 ( see figure [ fig : discovery ] for the discovery image ) . our methods for the discovery of candidate supernovae are outlined in detail in paper 2 . to summarize , for a given field , a pair of second epoch images were taken two weeks to two months after a template image was obtained . these second epoch images were then aligned to the template image by stellar matching algorithms . the images from the epoch with the better seeing were convolved to match that of the worst , and then scaled to be at the same flux level . the template image was subtracted from the second epoch images to produce residual images that , in principle , contain only variable objects , transients , and cosmic rays , on a nearly zero - level background with noise . the residual images were then automatically searched for candidates , with search parameter ranges set to eliminate cosmic rays and moving objects by requiring consistency between the pairs of second - epoch images . as described in paper 2 , the routines used in the image subtractions for this first campaign were developed by the supernova cosmology project for use in their high - z supernova searches ( perlmutter et al . 1997 , 1999 ) . the candidate supernova was photometrically confirmed on ut 1999 mar . 10 in direct images obtained by r. covarrubias at the cerro tololo inter - american observatory ( ctio ) 0.9-meter telescope . the direct images showed that the object had not moved , within the limits of the seeing , in the time since its discovery . figure [ fig : locals ] shows sn 1999aw 11 days after discovery , and 3 days after maximum light . the candidate was identified as a type ia supernova near maximum light from spectra obtained by a. goobar , t. dahlen , and i. hook on ut 1999 mar . 16.1 at the 2.6-meter nordic optical telescope ( not ) using the andalucia faint object spectrograph and camera ( alfosc ) . wavelength coverage was @xmath13 with a resolution of 700 ( or 430 km / s ) . a day later , l .- g . strolger and r. c. smith also obtained spectra at the ctio 4.0-meter blanco telescope ( ut mar . 17.3 ) using the r - c spectrograph . the effective wavelength coverage was 3400 - 7500 , with an approximate resolution of 1000 ( or 300 km / s ) . from these spectra ( presented in figure [ fig : spectra ] ) , m. m. phillips confirmed the initial identification by goobar et al . , and based on the very small ratio of si ii@xmath145978 absorption relative to the si ii@xmath146355 line ( see nugent et al . 1995a ) , suggested that sn 1999aw was likely to be a luminous , slow - declining sn ia . although sne ia generally show similar spectral evolution ( e.g. , branch et al . 1993 ) , nugent et al . ( 1995a ) showed that a spectral sequence exists for sne ia which is analogous to the luminosity - decline rate relations . specifically , certain spectral features show a systematic variation as a function of @xmath0@xmath1 . nugent et al . argued that this spectral sequence is due primarily to the differences in effective temperature which are presumably correlated with the amount of @xmath5ni produced in the explosion . at maximum light , spectroscopically normal sne ia exhibit strong ca ii h and k features near 3750 , and strong si ii@xmath146355 absorption which is blue - shifted by the high velocity of the expansion to appear near 6150 in the rest frame of the event ( minkowski 1940 , pskovskii 1969 , branch & patchett 1973 ) . excellent examples of normal sne ia are sne 1981b , 1989b , 1992a , and 1994d . in general , there are two groups of the spectroscopically `` peculiar '' type ia sne , also characterized near maximum light : sn 1991t - like events , and sn 1991bg - like events . 1991bg - like supernovae show wide absorption from 4150 - 4400 due to ti ii , and an enhanced 5800 feature ( commonly attributed to si ii ) in relation to si ii@xmath146355 ( filippenko et al . 1992 , leibundgut et al . the 5800 feature has been recently shown to be dominated by ti ii absorption ( rather than si ii ) in sn 1991bg - like sne , and may even be considerably significant in spectroscopically normal sne ( garnavich et al . 2001 ) . alternatively , 1991t - like sne have prominent fe ii and fe iii features at maximum light , but little or no si ii , s ii , or ca ii ( phillips et al . 1992 ) . a week after maximum light , however , the si ii , s ii and ca ii features develop , and the spectra become virtually indistinguishable from normal type ia . nugent et al . ( 1995a ) showed that 1991bg - like events correspond to the low - temperature end of the sne ia spectroscopic sequence ( a fact supported by garnavich et al . 2001 ) , while 1991t - like events are associated with the highest effective temperatures . normal sne ia inhabit the middle range of the sequence , where ti ii absorption is weak , the ratio of si ii@xmath146355 to si ii@xmath145800 is relatively high , and the ca ii h and k trough is strong . hence it may be more precise to refer to 1991bg - like and 1991t - like events as the extremes in a spectroscopic sequence rather than as `` peculiar '' events . ( 2001 ) also document a variation of sn 1991t - like events for supernova spectra that resemble sne 1999aa , 1998es , and 1999ac . these sn 1999aa - like supernovae are similar to 1991t - like sne supernovae , but show _ some _ si ii@xmath146355 prior to maximum light ( stronger than in 1991t - like sne , but weaker than in normal type ia sne ) , and clearly present fe ii and fe iii lines near maximum light . they also exhibit strong ca ii h and k lines . the spectra of sn 1999aw obtained by goobar et al . 16.1 ) and strolger and smith ( mar . 17.3 ) are displayed in figure [ fig : spectra ] , and clearly show a sn ia at or near maximum light . in figure [ fig : comp ] , the mar . 17.3 spectrum is compared with maximum - light spectra of four other sne ia with similarly - slow decline rates . sn 1990o was observed by hamuy et al . ( 1996c ) , and in spectra as early as 8 days before maximum clearly showed strong si ii@xmath146355 absorption ( phillips et al . , in preparation ) . sn 1992p was also observed by hamuy et al . ( 1996c ) . although nothing is known about its spectrum a week or more before maximum , the spectrum plotted in figure [ fig : comp ] also shows strong si ii@xmath146355 absorption at maximum . sn 1999aa is the prototype of the 1999aa - like events described in the previous paragraph . overall , the spectrum of this supernova ( reproduced from fig . 5 of li et al . 2001 ) is very similar to those of sne 1990o and 1992p , except that the si ii@xmath146355 absorption is noticeably weaker . at the bottom of figure [ fig : comp ] we plot the spectrum of sn 1999aw obtained on mar . the similarity to the spectrum of sn 1999aa is striking , with the only significant differences being the somewhat broader lines and stronger ca ii absorption of sn 1999aw . in section [ sec : lcurves ] we shall give further evidence for such a link based on the optical and ir light curves . on ut 1999 apr . 23 , m. m. phillips obtained a spectrum of sn 1999aw using the dupont 2.5-m telescope at las campanas observatory ( see figure [ fig : spectra ] ) . the wfccd was used in spectroscopic mode at a resolution of 630 ( or 480 km / s ) and the data covered the approximate wavelength range of 3750@xmath129250 . the date of observation corresponds to an epoch of 37 days after maximum light . by this late epoch , sne ia have begun their nebular phase , where their photospheres have essentially disappeared and their spectra become montages of blended emission features ( wells et al . the peaks of these features , fortunately , do not change substantially with time , hence they can be used to determine the redshift of sn 1999aw relative to sne ia with well - determined host galaxy redshifts and late epoch spectra . we compared the ut 1999 apr . 23 spectrum to spectra of sne 1992bc , 1989b , and 1992a taken at epochs of + 37 days , + 36 days , and + 36 days past maximum light respectively , and determined the relative shift between them . using the host galaxy redshifts given in ned , we calculated the redshift for sn 1999aw of 0.0392 , 0.0372 , and 0.0380 , respectively . the mean result is @xmath15 which we will use throughout this paper . an additional late - time spectrum of sn 1999aw was obtained by m. hamuy on ut 1999 may 19 using the ntt 3.6-m telescope at european southern observatory ( see figure [ fig : spectra2 ] ) . the emmi was used in spectroscopic mode at a resolution of 600 ( or 500 km / s ) and an approximate wavelength range of 4600@xmath128600 . the date of this observation corresponds to an epoch of 64 days after maximum light , at which time the supernova was more `` nebular '' than in previous epochs , with more pronounced emission features . recalculating the redshift to sn 1999aw based on the centers of fe iii@xmath144658 and co iii@xmath145890 results in @xmath16 , which is consistent with the redshift determined from the ut 1999 apr . 23 spectrum . our search techniques provided bulk sne detections , and therefore we could schedule sufficient follow - up observations to produce well - sampled light curves . following the search campaign , we obtained several optical images using scheduled time on the ctio 0.9-meter , 1.5-meter , and 1.0-meter ( yalo ) telescopes . the scheduling was planned such that the sampling would produce at least one data point every three nights , in each of the _ bvri _ filters , for at least 30 days after the discovery . we also obtained scheduled observations from various other telescopes . data were reduced ( bias subtracted and flat - field corrected ) using standard iraf packages for reducing multiple and single amplifier data . a shutter correction was measured for ctio 0.9-meter , and 1.5-meter data , and applied to short exposures ( @xmath17 seconds ) . the brightness of sn 1999aw , and its lack of host galaxy - light contamination , allowed accurate aperture photometry to be performed , without the need for late - time galaxy - light subtraction . using the daophot ii package ( stetson 1992 ) , data from photometric nights were compared with stars from tabulated standard fields ( landolt 1992 ) . interactive iterative solutions were made to construct a growth curve and mean aperture correction for stars on each frame . solutions were then calculated for the coefficients of the linear color and airmass terms to convert from natural to standard apparent magnitudes ( see appendix , tables [ table : eqn ] and [ table : terms ] ) . a set of local field standards stars was then produced around sn 1999aw ( see figure [ fig : locals ] ) , tied to the landolt ( 1992 ) standard stars observed . table [ table : locals ] contains the sequence of local photometric standards , coordinate offsets in arcseconds from sn 1999aw ( based on the wcs ) , and the photometric indices of these standards along with the mean errors . data of the sn from all other nights were compared to the data from the photometric nights using these standardized local field stars . table [ table : observations ] contains the epochs of observation of sn 1999aw , and the _ bvri _ photometry along with the photometric errors . these optical lightcurves are plotted in figure [ fig : lightcurve ] . as expected from the identifying spectra , sn 1999aw did indeed dim after peak at a slower - than - average rate . we performed a monte carlo simulation to randomly vary each data point within the photometric error 300 times , each time least - squares fitting a 3rd order polynomial to the peak of the _ b _ band lightcurve . averaging the solutions to the simulations led to the _ b_-band peak of @xmath18 on jd 2451254.7@xmath19 ( ut mar . 17.2 ) , and a decline rate within the first 15 days after maximum light of @xmath0@xmath20 . similar monte carlo simulations were performed on the _ v , r _ and _ i _ lightcurves , the results of which are summarized in table [ table : peaks ] . using @xmath21 and assuming @xmath22 km / s / mpc , we derive the absolute magnitudes of @xmath23 , @xmath24 , @xmath25 , and @xmath26 , correcting only for galactic extinction in the direction of sn 1999aw using values of @xmath27 , @xmath28 , @xmath29 , and @xmath30 ( schlegel , finkbeiner , & davis 1998 ) . infrared imaging of sn 1999aw was obtained over a 50-day period , beginning at the epoch of _ b _ maximum , using the swope 1-meter and dupont 2.5-meter telescopes at las campanas observatory ( lco ) , and using the vlt at european southern observatory . these data were taken in the _ j@xmath3 , h , _ and _ k@xmath3 _ bandpasses . the lco infrared images were reduced with iraf using standard techniques . briefly , the steps consisted of 1 ) application of a correction for the slightly non - linear response of the nicmos detector , 2 ) subtraction of dark images of the same exposure time as the supernova images , 3 ) division by a twilight flat field , 4 ) subtraction from each individual exposure of a sky image created from the dithered images of the supernova , and 5 ) shifting and summing the individual images to create a final `` mosaic '' image of the supernova field . daophot ii was used to obtain psf photometry , construct growth curves , and to set up local standards using observations on photometric nights of the persson et al . ( 1998 ) standard stars . as the persson et al . standards were established using the same instrument / detector / filters employed for the swope telescope observations of sn 1999aw , no color corrections were applied to the supernova magnitudes . the dupont irc observations are probably also very nearly on the same system , but this may not be the case for the dupont cirsi and vlt data . it has been noticed that although the vlt / isaac transmission curves are fairly similar to the lco filters in _ h _ and _ k@xmath3 _ , the _ j@xmath3 _ filter is significantly narrower and therefore the _ j@xmath3_-band observations can be quite different from the lco system . we hope to eventually derive appropriate transformations for the latter data to the persson et al . photometric system ; in the absence of such information , no color corrections for these data are made in the present paper . table [ table : localir ] lists the local standard star photometry ( using the same numbering system as in the optical ) , table [ table : irobs ] contains the epochs of observation of sn 1999aw with photometric errors , and figure [ fig : lightcurve ] shows the resulting infrared light curves . at @xmath21 , the redshift of sn 1999aw is large enough that it is important to correct the flux received in a given filter to that which would have been received if the event occurred at nearly zero redshift . if spectroscopic data were available for each epoch a direct image was obtained , corrections could be made to the flux received in each passband , or k - corrections , by determining the change in brightness in the transmission curves of those passbands as the spectra are shifted from the observed reference frame to the rest frame . corrections would be applied as @xmath31 , where @xmath32 is the epoch from maximum light . as this extensive spectral data is unavailable , we employ a method similar to that of nugent et al . ( 2002 , in prep ) . here several epochs of spectra for sn 1999aw spanning the optical regime were used to produce the k - correction by manipulating the spectral energy distribution , producing synthetic photometry on a particular date to match the corresponding observed colors of sn 1999aw . we interpolate and extrapolate these few k - terms to derive k - terms for all epochs of photometric observation . the uncertainties from this method , based on the photometric data and the associated uncertainties for sn 1999aw , are less than 0.02 magnitudes for a given epoch . in at least _ b _ and _ v _ , the k - terms are in good agreement with the values produced in a pure application of the method of nugent et al . ( 2002 ) , and with the values tabulated in hamuy et al . ( 1993 ) for a sn ia at @xmath33 . however , in _ i _ there is a significant difference from yet unpublished k - terms determined from other nearby sne ia . this difference is caused by the delay ( or `` stretch '' ) in the _ i_-band lightcurve of sn 1999aw , which is consistent with a possible trend among bright slow - declining 1991t/1999aa - like sne ia in which the _ ijhk _ lightcurves evolve at a slower rate than other sne ia ( see section [ subsec : colorcurves ] ) . this difference in photometric ( and plausibly spectroscopic ) evolution exemplifies the danger in applying k - corrections calculated from sne ia with radically different decline rates . applying the derived k - corrections to the photometric data for sn 1999aw , and correcting the observed passage of time by a factor of @xmath34 for time dilation , produced a rest frame apparent magnitude lightcurve that was not that different from the observed lightcurve , with @xmath0@xmath35 . we have not attempted to make k - corrections for the near - ir data because we do not have ir spectra for sn 1999aw , nor are there yet sufficient libraries of ir spectra for other sne ia . the lightcurves of sn 1999aw are remarkable for how much slower they evolve compared to the template curves of a typical decline - rate sne ia ( see figure [ fig : lightcurve ] ) . this is evident not only in the slow initial decline rate of @xmath0@xmath36 ( the mean decline rate of sne ia is @xmath0@xmath37 ) , but in the delay of the second maximum in _ i , j@xmath3 , h , _ and _ k@xmath3_. the _ b_-band lightcurve is particularly interesting in that its `` shape '' is subtly different than that for typical sne ia . one parameter lwr relations such as the @xmath0@xmath1 suggest that a _ b_-band lightcurve template can be made to fit the _ b_-band lightcurve of a sn by applying a time delaying `` stretch '' factor to the epochs of observation , and the related magnitude offset ( perlmutter et al . 1997 ) . however , strolger et al . ( 2000 & 2003a ) show that _ b_-band template lightcurves built from well - sampled spectroscopically normal sne ia fit poorly to the _ b_-band lightcurves of 1991t/1999aa - like supernovae , even when stretched to fit the observed lightcurve in early epochs ( by `` normal '' , we mean those those sne ia which displayed strong si ii@xmath38 absorption at least 5 days before _ b_-band maximum ) . the strolger et al . ( 2000 & 2003a ) analysis was conducted on a few sne that were 1 ) spectroscopically identified at least 5 days before maximum light , and 2 ) frequently observed with well sampled _ b_-band lightcurves from just prior to maximum light to around the @xmath3980 day epoch . the low - order mean difference between the template curve and the supernovae lightcurves from the beginning of the exponential phase ( after the @xmath3925 day epoch ) was determined for each supernova in the sample : @xmath40 observations made at some epoch , @xmath41 , were compared to the time stretched template curve , @xmath42 , and then averaged over the number of observations ( @xmath43 ) from the @xmath3925 day to @xmath3980 day epochs . results of the analysis show that spectroscopically normal sne ia exhibit little to no difference from the template curve during the exponential phase , whereas 1991t/1999aa - like sne ia show a substantial overbrightness during this phase . the analysis , when performed on sn 1999aw , produced a result consistent with that of the 1991t/1999aa - like sne . figure [ fig : templates ] shows four sne spectroscopically similar to sne 1991 t and/or 1999aa ( including sn 1999aw ) , along with the normal sne ia template , stretched in time to fit the lightcurves within the first 15 days past maximum light . the template does not fit well to the data past around the @xmath44 day epoch , and the data are systematically brighter than the template from that epoch . in the future , with more examples of 1991t/1999aa - like sne , this analysis may lead to photometric method that , in addition to spectroscopy , indicate the possible 1991t/1999aa - like peculiarity of supernovae . sne ia show a impressive uniformity in their intrinsic colors in late epochs after maximum light . as lira ( 1995 ) and riess et al . ( 1996 ) independently showed , sne ia with @xmath45@xmath0@xmath1@xmath46 and little or no reddening from their host galaxies have very uniform @xmath47 color evolution from @xmath3930 to @xmath3990 days after maximum light . krisciunas et al . ( 2000 ) also observe uniformity in the _ v_@xmath12_near ir _ color evolution of spectroscopically normal sne ia with mid - range decline rates from @xmath129 days to @xmath3927 days past maximum light . figure [ fig : colors ] shows the evolution in multiple colors of sn 1999aw , corrected for galactic reddening assuming an excess of @xmath48 ( schlegel , finkerbeiner , & davis 1998 ) . the left - hand side panels of figure [ fig : colors ] show the optical color evolution , along with some example sne ( zero - reddening corrected ) for comparison . the solid line in the plot of the evolution of @xmath47 in this figure represents the zero - reddening least - squares fit derived by lira ( 1995 ) . sn 1992al is included to show the evolution of a typical sne ia , while sn 1992bc is a slow - declining sne ia ( @xmath0@xmath49 ) but with normal pre - maximum light spectra . sne 1991 t and 1999aa , the prototypical 1991t/1999aa - like sne , have color evolutions nearly parallel to the lira ( 1995 ) line , thus showing that the uniformity holds for _ some _ spectroscopically extreme sne ia . this color uniformity has proven useful as an indicator of host galaxy reddening , which has been important to revising the lwr relations ( phillips et al . we have used the recipes in phillips et al . ( 1999 ) to determine the host galaxy reddening for sn 1999aw . the analysis gave @xmath50 , @xmath51 , and @xmath52 , which when averaged , resulted in a negative color excess , implying zero host galaxy reddening . clearly the negative value of @xmath53 is heavily influenced by the anomalous @xmath47 color evolution of sn 1999aw ; specifically , for most of the period from @xmath3940 days to @xmath3960 days sn 1999aw was considerably bluer than the lira ( 1995 ) zero - reddening fit , and perhaps evolving with a steeper slope . as sn 1999aw was spectroscopically similar to sn 1999aa , we might have expected that its color evolution would be similar to sn 1999aa and/or sn 1991 t , both of which appeared generally bluer than sn 1992bc in @xmath54 and @xmath55 . in sn 1999aw , the bluer @xmath54 and @xmath55 evolution were caused by the delay in the appearance of the second lightcurve maximum , which was more clearly seen in redder passbands ( see figure [ fig : lightcurve ] ) , and therefore is accentuated in the @xmath54 and @xmath55 curves . sn 1999aw also appeared bluer than sn 1999aa and sn 1991 t in @xmath54 and @xmath55 , which perhaps was an effect of sn 1999aw having a much slower decline rate than both sn 1999aa and sn 1991 t . the @xmath56 and @xmath57 curves for sn 1999aw are also considerably bluer than the zero - reddening fits from krisciunas et al . ( 2000 ) in the period after @xmath3910 days ( right - hand side panels of figure [ fig : colors ] ) . the trend in the period prior to @xmath3910 days in the @xmath56 curve also seems to be different than the fit from krisciunas et al . it appears that one could apply a `` stretch '' to the krisciunas lines to force them to fit the sn 1999aw data , that is to say there is an apparent delay in the @xmath56 evolution . this again is due to the apparent delay of the second maximum in the _ i , j@xmath3 , h , _ and _ k@xmath3 _ lightcurves . note , however , that the @xmath56 , @xmath57 , and @xmath58 colors in this figure may change somewhat once k - corrections for the ir bandpasses become available . as nearly all of the bolometric luminosity of a typical type ia supernova is emitted in the range of 3000 to 10000 ( suntzeff 1996 ) , the integrated flux in the _ ubvrij@xmath3hk@xmath3 _ bandpasses provides a reliable and meaningful estimate of the bolometric luminosity , which is directly dependent on the amount of nickel produced in the explosion . the _ bvrij@xmath3hk@xmath3 _ data were used to calculate `` _ uvoir _ '' bolometric fluxes using the techniques described in suntzeff ( 1996 ) and suntzeff and bouchet ( 1990 ) . a table of _ ubvrij@xmath3hk@xmath3 _ data was made by linearly interpolating the _ j@xmath3hk@xmath3 _ data to the optical dates . for some of the missing optical and _ j@xmath3 _ and _ h _ data , we added photometry based on spline fitting of the data to the date of the missing data . we have added on _ u _ data because the optical ultraviolet adds significant flux to the early time bolometric light curve . there is little high - quality _ u _ data for type ia supernovae due to the generally poor ultraviolet sensitivity of the present generation of ccds . we have instead relied on @xmath59 photometry from photoelectric measurements . for dates past @xmath399 days from _ b _ maximum , we have used the @xmath59 data of sn 1972e from lee et al . ( 1972 ) and ardeberg & de groot ( 1973 ) . for dates before @xmath399 days from _ b _ maximum we have used the @xmath59 data for sn 1980n from hamuy et al . ( 1991 ) , and for sn 1981b compiled by cadonau & leibundgut ( 1990 ) . the @xmath59 for these supernovae were corrected to the reddening of sn 1999aw using the reddening values in phillips et al . ( 1999 ) and a value of @xmath60 from cardelli et al . the _ u _ photometry of sn 1999aw was then estimated from spline fits to the @xmath59 data of these supernovae combined with our _ b _ data of sn 1999aw . we then converted the broadband magnitudes to equivalent monochromatic fluxes at the effective wavelengths of vega ( bessell 1979 , 1990 ; bessell & brett 1988 ) . a magnitude scale of @xmath61 and @xmath62 was used for vega . the monochromatic fluxes were then scaled to the magnitude of the supernova , dereddened by @xmath63 using the reddening law of cohen et al . ( 1981 ) , and corrected to intrinsic fluxes using a distance modulus of 36.28 based on a hubble flow with a hubble constant of 63.3 km / s / mpc ( phillips et al . 1999 ) . these fluxes were then integrated using a simple trapezoidal integration . we added on a rayleigh - jeans extrapolation to zero frequency to the reddest flux point . we extrapolated to the ultraviolet by adding a flux point at 3000 with zero flux . we correct the derived bolometric lightcurves for time dilation effects , and plot them in figure [ fig : bolcurve ] . the _ ubvri _ and the _ ubvrij@xmath3hk@xmath3 _ integrations track each other well , except that the inflection points around days 20 - 45 are more pronounced in the _ ubvrij@xmath3hk@xmath3 _ integrations . similar inflection points were noted by suntzeff ( 1996 ) and are indicative of significant flux redistributions which may be related to rapid changes in the wavelength dependence of the opacities ( pinto & eastman 2000 a & b ) . the peak bolometric luminosity is about @xmath64 erg s@xmath65 . it is fairly straightforward to derive the nickel mass produced in the explosion from the bolometric luminosity at peak . at maximum light , photons escape the surface at a rate which is equal to the radioactive energy input produced primarily by the @xmath5ni decay , and thus it is also related to the @xmath5ni synthesized in the explosion ( arnett 1982 , nugent et al . 1995b , pinto & eastman 2000a ) . contardo et al . ( 2000 ) calculate the luminosity and nickel mass for several sne ia from _ ubvri _ bolometric peak fluxes . using the same method , and assuming a rise time of 17 days to bolometric peak for consistency with contardo , et al . , we derive an initial nickel mass of @xmath66 for sn 1999aw . this is brighter and more nickel massive than many of the normal type ia sne discussed in contardo , et al . , and it is comparable in brightness and nickel mass to sn 1991 t ( see table [ table : masses ] ) . however as they note , a number of the sne ia in their study do not have _ u_-band data available at peak , and therefore they have developed a correction curve based on data from the very well - sampled sn 1994d . although this sn was spectroscopically normal , it did have some rather unusual features , including an unusually blue @xmath59 color at maximum light . additionally , riess et al . ( 1999 ) have shown that the characteristic rise time of sne ia is @xmath67 days , and that brighter and slower declining sne ia have longer rise times . for the peak magnitude and decline rate observed in sn 1999aw , we expected a rise time of @xmath68 days . using this value , we derive an alternative initial nickel mass of @xmath69 . there are a number of additional uncertainties in both the calculated luminosity , and the derived nickel mass . the first is the assumption that more than 80% of the true bolometric light is emitted in the optical regime , that less than 10% can be expected in the uv below 3200 , and that no more than 10% ( in early epochs ) from _ jhk _ ( suntzeff 1996 , elias et al . 1985 , contardo et al . 2000 ) . in comparing our ubvrij@xmath3hk@xmath3 _ and _ ubvri _ bolometric fluxes , we find the ir contribution to be only a few percent ( @xmath70 at early epochs , @xmath71 after 35 days past @xmath72 ) . we have not accounted for the space ultraviolet flux . also , as we will discuss further in section [ sec : galaxy ] , we do not have much information about the host galaxy of sn 1999aw . although we have compensated for extinction and reddening due to our own galaxy , it difficult to do the same for the host galaxy . however , as the @xmath47 color of sn 1999aw is nearly zero at maximum light , as it would be for an unreddened type ia sn ( phillips et al . 1999 ; garnavich et al . 2001 , in figure 15 ) , we have assumed that the extinction due to the host must be negligibly small . another approach to understanding sn 1999aw can be to investigate its host galaxy . the metallicity of the progenitor may be related to the environment of the event ( hamuy et al . however , the apparent low luminosity of the galaxy has made it very difficult to study . as earlier stated , there was no obvious host galaxy in the template images nor in subsequent photometry . on ut 2000 feb . 12 , l. strolger , p. candia , j. seguel , and a. bonacic obtained deep images of the sn 1999aw field using the ctio 1.5-meter telescope . pairs of long exposures were taken in the same standard _ bvri _ filters as were used in the photometry of sn 1999aw ( the combined exposure times were 1200 seconds for b , v , and i , and 1440 for r ) . zero points in each filter were found by comparison of standards fields taken on the same evening , with tabulated standard magnitudes . no clear detection ( to a 3-@xmath73 level ) was made of a galaxy at or near the position of sn 1999aw in these deep images ( see figure [ fig : hostgxy ] ) . a second attempt to detect the host was made on ut 2001 mar . 17 by a. szentgyorgyi , and on ut 2001 mar . 18 by m. mateo , a. athey , and k. von braun , both using the baade 6.5-meter telescope at las campanas observatory and the magellan imaging camera ( magic ) . they obtained deep exposures for a combined 1600 sec in _ b _ , 1400 sec in _ v _ , and 900 sec in _ i _ of the sn 1999aw field . these deep images did reveal a resolved galaxy , with fwhm of 0.6 , only slightly broader than the seeing ( 0.42 ) . the central brightness ( within 1.5 radii ) were @xmath74 , @xmath75 , and @xmath76 magnitudes . the radial growth curve for the galaxy was very steep in all passbands , and heavily contaminated by background noise in the image . therefore , we assume the 1.5 magnitudes are the best estimates of the total light in the galaxy . absolute values of @xmath77 , @xmath78 , and @xmath79 were calculated assuming @xmath22 km / s / mpc and correcting only for galactic extinction . coincidentally , the low luminosity dwarf galaxy ic 4182 hosted type ia supernova sn 1937c , also with slow decline rate ( @xmath0@xmath80 ) . although branch et al . ( 1993 ) indicate that spectra from minkowski ( 1939 & 1940 ) of sn 1937c seem to show a normal type ia sn , there may be some question as to whether or not the si ii@xmath146355 feature was relatively weak at early epochs , possibly indicating a similarity to sn 1999aa . using the measured ccd photometry of ic 4182 from makarova ( 1999 ) , and the cepheid distance modulus for the galaxy calculated by saha , et al . ( 1994 ) and correcting for galactic extinction , we determined its absolute magnitude to be @xmath81 , @xmath82 , and @xmath83 . the sn 1999aw host galaxy is * significantly fainter * ( by 4.2 , 4.3 , and 5.5 magnitudes in _ b _ , _ v _ , and _ respectively ) . the error in these observations , especially in _ i _ , make it difficult to estimate color differences for the host galaxy , however , using the magnitudes above yields @xmath84 and @xmath85 . assuming @xmath86 better exemplifies the color trend , the galaxy would appear to be fairly blue in comparison to dwarf galaxies in the local group ( see table 4 and fig . 3 of makarova 1999 ) . initial studies of sne ia characteristics and their relation to their host galaxies have been conducted by ivanov , hamuy , & pinto ( 2000 ) and hamuy et al . their studies show : * the distribution of sne ia decline rates ( and therefore brightness ) changes considerably with host galaxy morphological type . earlier hubble type galaxies ( such as ellipticals and s0 spirals ) produce faint fast - declining sne ia , and late type galaxies ( sa@xmath12irr ) tend to produce bright slow - declining sne ia . * sne ia decline rates also change with @xmath47 color of the host galaxy . bright sne ia occur more frequently in bluer environments . * bright sne ia occur more frequently in less luminous galaxies . it is difficult , however , to determine how these color@xmath12luminosity correlations relate to age and metallicity effects . integrated galaxy luminosities seem to correlate with metallicity ( henry & worthy 1999 ) , and both the sample of galaxies from hamuy et al . ( 2000 ) and the sample of dwarf galaxies from makarova ( 1999 ) seem to indicate that the least luminous galaxies are also the bluest . as bluer galaxies are also younger , perhaps the brightness distribution of sne ia with @xmath47 color is an age effect , suggesting that younger environments produce bright slow - declining sne ia . alternatively , it is plausible that the sne ia brightness@xmath12color correlation is just a reflection of the sne ia brightness@xmath12host galaxy luminosity trend , which may be a metallicity effect indicating that metal - poorer environments produce the brightest sne ia . nonetheless , observations of sn 1999aw and its host galaxy are consistent with these trends in that the host was faint and blue [ among the faintest 10% in @xmath87 , and the bluest 15% in @xmath88 of the makarova ( 1999 ) sample ] , and that it produced a bright slow - declining sn ia . our photometric and spectroscopic study of sn 1999aw indicate that sn 1999aw was probably a 1999aa - like event . the light curve decline rate is among the slowest observed at @xmath0@xmath89 . at the redshift of @xmath90 , it is also among the brightest type ia , with @xmath91 , @xmath92 , @xmath93 , and @xmath94 . although luminous , these magnitudes are slightly less ( by a few tenths of a magnitude ) than one might expect from the trends of absolute magnitude versus decline rate or @xmath0@xmath95 ( see figure 11 of krisciunas et al . perhaps this suggests that there is some host galaxy extinction that needs to be accounted for . or it is possible that sn 1999aw is simply slightly less luminous than expected , and therefore may be indicating a possible downward curve to the relations in fig . 11 for the slowest - declining sne ia ( similar to the downward curve recently determined by garnavich et al . ( 2001 ) for fast - declining sn 1991bg - like sne ia ) . it is thought that the brightness and decline rate of type ia are dependent on the metallicity of the progenitor , and additionally on the opacity in the atmosphere of the event ( hoflich et al . 1998 , pinto and eastman 2000a & b , mazzali et al . 2001 ) . the derived luminosity of @xmath96 erg s@xmath65 , and @xmath5ni mass of @xmath97 for sn 1999aw are both relatively high , but are consistent with what can be expected for a sn ia with the observed decline rate , as inferred from the trend between @xmath0@xmath1 and @xmath5ni mass evident in table [ table : masses ] . the observations made at the baade 6.5-meter have provided some very interesting information on the host galaxy of sn 1999aw . the derived absolute magnitudes seem to indicate that it is among the intrinsically faintest and bluest dwarf galaxies observed . this may indicate that the galaxy consists of either a young population of stars , or is a fairly metal - poor environment . it will be necessary to eventually obtain deeper images with large telescopes such as the magellan 6.5-meter or the vlt to obtain sufficient signal - to - noise to put more accurate constraints on not only the colors of this galaxy , but its structure as well . the discovery of sn 1999aw exemplifies an advantage of magnitude limited field searches over galaxy - targeted searches . it is presently unclear how many supernova similar to sn 1999aw occur at low redshift , and with the biases associated with current targeted surveys we are less likely to detect them . as more magnitude limited field searches commence , relations between low luminosity ( and probably low metallicity ) galaxies and the supernova types that occur in them can be accurately determined . this will not only place constraints on type ia supernova progenitor models , but help to clarify the possibility of supernova `` evolution '' which may explain the difference in sn 1991t - like event rates between the low and high-@xmath98 surveys ( li et al . 2001 ) . this work is dedicated to the memory of robert a. schommer , whose contributions we valued , and whose friendship we will miss . ls is very grateful to cerro tololo inter - american observatory and its staff for providing a office , exceptional support , mentorship , and an unequivocal learning experience during his stay there from 1998 to 2000 . additional acknowledgements to b. schaefer , c. baiyln , and s. tourtellotte for the use of yalo and yalo data , and to a. block , m. block , p. challis , j. krick , h. mathis , and a. soderberg for their assistance is searching through countless images in search of supernovae . optical aperture photometry was performed using the daophot ii package ( stetson 1992 ) . 14 diameter apertures were used and were corrected by an iterative growth curve to a nearly infinite diameter . standard fields were observed on 8 photometric nights on the ctio 0.9-meter , and 7 photometric nights on the ctio 1.5-meter . caution was taken to observe fields at various airmasses in all filters . the data from 50 stars were compared to tabulated values ( landolt 1992 ) and solutions were found for the coefficients of the linear color and airmass terms in the photometric transformation equations ( see table [ table : eqn ] and [ table : terms ] ) . the equations were then used to transform the natural magnitudes of the local standard stars to standard magnitudes ( see table [ table : locals ] ) . nugent , p. , aldering , g. 2000 , `` the spring 1999 nearby supernovae campaign '' , the greatest explosions since the big bang : supernovae and gamma - ray bursts , space telescope science institute symposium , may 1999 , ed . m. livio , n. panagia , & k. sahu , baltimore : space telescope science institute . cccccccccccc 1&@xmath1213.5&0.6&17.138 ( 0.015)&1.346 ( 0.033)&0.861 ( 0.022)&0.846 ( 0.022)&1.707 ( 0.020)&23&20&21&21 + 2&49.7&@xmath1230.0&16.67 ( 0.014)&1.95 ( 0.023)&0.373 ( 0.020)&0.339 ( 0.023)&0.712 ( 0.022)&23&20&19&21 + 3&19.7&33.1&18.477 ( 0.028)&0.7 ( 0.049)&0.406 ( 0.038)&0.39 ( 0.038)&0.795 ( 0.039)&17&16&15&17 + 4&20.4&63.4&17.587 ( 0.021)&0.856 ( 0.035)&0.486 ( 0.027)&0.433 ( 0.024)&0.919 ( 0.027)&23&22&19&21 + 5&58.9&104.2&17.83 ( 0.019)&0.728 ( 0.035)&0.428 ( 0.028)&0.4 ( 0.029)&0.828 ( 0.028)&22&19&18&20 + 6&@xmath12154.8&103.7&18.178 ( 0.028)&0.691 ( 0.045)&0.394 ( 0.041)&0.378 ( 0.038)&0.772 ( 0.037)&19&18&17&17 + 7&@xmath12112.5&26.6&17.38 ( 0.018)&0.555 ( 0.028)&0.372 ( 0.026)&0.378 ( 0.025)&0.75 ( 0.024)&21&20&19&19 + 8&@xmath12157.6&16.1&15.65 ( 0.010)&0.961 ( 0.018)&0.547 ( 0.014)&0.476 ( 0.012)&1.024 ( 0.012)&17&16&15&17 + 9&@xmath12116.0&@xmath1262.1&17.488 ( 0.018)&0.938 ( 0.033)&0.526 ( 0.026)&0.457 ( 0.024)&0.983 ( 0.025)&20&19&18&18 + 10&@xmath12147.8&@xmath1283.4&18.201 ( 0.027)&0.59 ( 0.043)&0.358 ( 0.039)&0.351 ( 0.039)&0.708 ( 0.037)&21&20&19&19 + 11&@xmath12122.9&@xmath12196.0&15.632 ( 0.017)&1.019 ( 0.027)&0.411 ( 0.025)&0.377 ( 0.024)&0.788 ( 0.023)&12&11&10&10 + 12&@xmath1225.6&@xmath12129.3&18.529 ( 0.028)&0.773 ( 0.048)&0.463 ( 0.040)&0.442 ( 0.041)&0.905 ( 0.041)&19&18&17&17 + 13&51.6&@xmath12190.1&16.958 ( 0.016)&0.648 ( 0.025)&0.393 ( 0.024)&0.384 ( 0.025)&0.777 ( 0.023)&17&16&15&15 + 14&126.2&@xmath12104.2&18.092 ( 0.026)&0.45 ( 0.038)&0.326 ( 0.040)&0.321 ( 0.043)&0.647 ( 0.040)&19&16&17&17 + 15&160.3&@xmath1261.6&16.279 ( 0.012)&2.387 ( 0.021)&0.515 ( 0.021)&0.452 ( 0.024)&0.967 ( 0.021)&17&14&15&15 + 16&208.7&138.7&17.369 ( 0.015)&0.576 ( 0.023)&0.356 ( 0.026)&0.346 ( 0.029)&0.702 ( 0.025)&9&9&8&7 + 17&@xmath12139.3&33&15.052 ( 0.009)&0.559 ( 0.016)&0.334 ( 0.013)&0.333 ( 0.013)&0.667 ( 0.013)&21&20&19&19 + 18&14.0&@xmath1243.7&20.951 ( 0.061)&0.672 ( 0.093)&2.037 ( 0.073)&0.579 ( 0.062)&2.616 ( 0.077)&4&5&5&5 + 19&@xmath1256.1&@xmath1264.6&19.206 ( 0.043)&1.641 ( 0.067)&1.031 ( 0.050)&0.722 ( 0.040)&1.753 ( 0.052)&2&3&3&3 + 245.40 & kpno 0.9-m & ngss team&&&17.251 ( 0.010 ) & + 249.76 & yalo & service & 16.961 ( 0.016 ) & 16.980 ( 0.008 ) & 16.990 ( 0.009 ) & 17.229 ( 0.027 ) + 254.66 & yalo & service & 16.858 ( 0.023 ) & 16.806 ( 0.015 ) & 16.800 ( 0.007 ) & 17.272 ( 0.015 ) + 255.88 & lick 40-inch & quimby & 16.838 ( 0.023 ) & 16.856 ( 0.028 ) & 16.797 ( 0.020 ) & + 257.80 & ctio 0.9-m & strolger & 16.941 ( 0.007 ) & 16.855 ( 0.007 ) & & + 258.66 & yalo & service & 16.972 ( 0.005 ) & 16.809 ( 0.007 ) & 16.818 ( 0.009 ) & 17.418 ( 0.016 ) + 258.67 & ctio 1.5-m & aldering & & 16.864 ( 0.005 ) & 16.801 ( 0.005 ) & 17.421 ( 0.012 ) + 259.80 & lick 40-inch & quimby & 17.123 ( 0.041 ) & 16.905 ( 0.029 ) & 16.814 ( 0.024 ) & + 265.57 & yalo & service & 17.311 ( 0.009 ) & 17.108 ( 0.011 ) & 17.095 ( 0.012 ) & 17.783 ( 0.026 ) + 266.68 & ctio 0.9-m & strolger & 17.412 ( 0.013 ) & 17.210 ( 0.017 ) & 17.172 ( 0.019 ) & 17.875 ( 0.030 ) + 272.72 & ctio 0.9-m & germany & 17.905 ( 0.016 ) & 17.533 ( 0.054 ) & 17.521 ( 0.020 ) & 18.191 ( 0.049 ) + 274.73 & ctio 0.9-m & strolger & 18.146 ( 0.031 ) & 17.624 ( 0.032 ) & 17.610 ( 0.017 ) & 18.234 ( 0.035 ) + 275.71 & ctio 0.9-m & strolger & 18.245 ( 0.020 ) & 17.625 ( 0.048 ) & 17.615 ( 0.049 ) & 18.236 ( 0.045 ) + 275.82 & lick 40-inch & quimby & 18.197 ( 0.055 ) & & 17.628 ( 0.026 ) & + 276.72 & ctio 1.5-m & smith + strolger&18.298 ( 0.009 ) & 17.697 ( 0.008 ) & 17.637 ( 0.012 ) & 18.182 ( 0.020 ) + 277.69 & ctio 1.5-m & smith + strolger&18.365 ( 0.009 ) & 17.725 ( 0.008 ) & 17.640 ( 0.009 ) & 18.160 ( 0.025 ) + 280.87 & lick 40-inch & gates & 18.680 ( 0.171 ) & 17.798 ( 0.066 ) & 17.613 ( 0.062 ) & + 284.48 & lick 40-inch & quimby & 18.907 ( 0.051 ) & 17.948 ( 0.036 ) & & + 284.62 & yalo & service & 18.917 ( 0.019 ) & 17.957 ( 0.015 ) & 17.683 ( 0.018 ) & 17.947 ( 0.044 ) + 285.46 & jkt 1-m & mndez + blanc & & 17.991 ( 0.006 ) & & 17.893 ( 0.014 ) + 287.63 & yalo & service & 19.110 ( 0.008 ) & 18.096 ( 0.009 ) & 17.703 ( 0.005 ) & 17.923 ( 0.015 ) + 288.82 & lick 40-inch & gates & 19.219 ( 0.138 ) & & 17.746 ( 0.120 ) & + 291.68 & ctio 1.5-m & smith + strolger&19.378 ( 0.020 ) & 18.277 ( 0.035 ) & 17.855 ( 0.019 ) & 17.847 ( 0.021 ) + 292.61 & ctio 1.5-m & smith + strolger&19.396 ( 0.044 ) & 18.348 ( 0.022 ) & 17.889 ( 0.024 ) & 17.854 ( 0.060 ) + 297.46 & ctio 1.5-m & strolger & 19.629 ( 0.080 ) & 18.637 ( 0.080 ) & 18.209 ( 0.022 ) & 18.161 ( 0.106 ) + 303.67 & ctio 1.5-m & strolger & 19.766 ( 0.031 ) & 18.936 ( 0.100 ) & & + 304.65 & ctio 1.5-m & strolger & 19.743 ( 0.016 ) & 18.948 ( 0.011 ) & 18.494 ( 0.007 ) & 18.543 ( 0.036 ) + 313.60 & yalo & service & 19.856 ( 0.005 ) & 19.236 ( 0.007 ) & 18.891 ( 0.015 ) & 19.093 ( 0.081 ) + b&254.7(0.3)&16.88(0.01)&-19.45(0.11)&0.81(0.03 ) + v&255.6(0.3)&16.81(0.01)&-19.50(0.11)&0.66(0.04 ) + r&256.0(0.3)&16.79(0.01)&-19.38(0.11)&0.62(0.04 ) + i&251.8(4.2)&17.24(0.08)&-18.97(0.12)&0.61(0.09 ) + 2&15.520(0.015)&15.189(0.018)&15.182(0.035)&5&10&11 + 3&17.236(0.036)&16.854(0.038)&16.813(0.070)&5&10&1 + 4&16.095(0.017)&15.646(0.021)&15.678(0.048)&4&10&11 + 18&17.113(0.032)&16.455(0.031)&16.357(0.057)&4&10&10 + 19&15.671(0.016)&15.125(0.019)&14.863(0.032)&5&10&11 + 254.66&lco 1-m / c40irc&phillips , roth&17.719(0.035)&18.031(0.039)&17.525(0.075 ) + 255.68&lco 1-m / c40irc&roth&17.853(0.049)&&17.481(0.065 ) + 256.67&lco 1-m / c40irc&roth&17.824(0.033)&18.000(0.048)&17.566(0.197 ) + 257.66&lco 1-m / c40irc&roth&17.916(0.041)&18.305(0.053)&17.739(0.097 ) + 258.71&lco 1-m / c40irc&roth&17.914(0.035)&18.129(0.051)&17.735(0.092 ) + 259.67&lco 1-m / c40irc&roth&18.058(0.043)&18.193(0.045)&17.413(0.093 ) + 260.67&lco 1-m / c40irc&roth&18.235(0.043)&18.319(0.058)&17.645(0.078 ) + 262.68&lco 2.5-m / c100irc&galaz&18.361(0.045)&18.167(0.057)&17.589(0.060 ) + 264.57&lco 2.5-m / c100irc&galaz&18.495(0.041)&18.123(0.061)&17.558(0.050 ) + 270.62&lco 1-m / c40irc&muena&19.336(0.129)&18.263(0.091)&17.937(0.114 ) + 272.61&lco 2.5-m / cirsi&marzke , persson&19.118(0.038 ) & & + 273.53&lco 2.5-m / cirsi&marzke , persson&&18.406(0.033 ) & + 275.49&lco 2.5-m / cirsi&marzke , persson , phillips&18.965(0.072 ) & & + 288.59&lco 1-m / c40irc&hamuy&18.658(0.055)&18.051(0.063 ) & + 290.56&lco 1-m / c40irc&phillips&18.527(0.045)&18.177(0.082)&18.021(0.100 ) + 294.59&lco 1-m / c40irc&roth&18.551(0.056)&18.031(0.056 ) & + 301.47&lco 2.5-m / cirsi&mccarthy&&18.528(0.039 ) & + 302.5&lco 2.5-m / cirsi&mccarthy , phillips&19.422(0.047 ) & & + 304.48&vlt / isaac&hamuy&19.684(0.129)&18.722(0.123)&18.853(0.128 ) + 1989b&1.20&43.06&0.57 + 1991t&0.97&43.23&0.84 + 1991bg&1.85&42.32&0.11 + 1992a&1.33&42.88&0.39 + 1992bc&0.87&43.22&0.84 + 1992bo&1.73&42.91&0.41 + 1994d&1.46&42.91&0.41 + 1994ae&0.95&43.04&0.55 + 1995d&0.98&43.19&0.77 + * 1999aw*&*0.81*&*43.18*&*0.76 * + ccccc a&ctio 0.9-m&3.477 ( 0.027)&0.092 ( 0.005)&0.246 ( 0.029 ) + & ctio 1.5-m&2.416 ( 0.010)&0.094 ( 0.003)&0.205 ( 0.025 ) + b&ctio 0.9-m&3.142 ( 0.026)&@xmath120.015 ( 0.002)&0.125 ( 0.017 ) + & ctio 1.5-m&2.086 ( 0.020)&@xmath120.015 ( 0.002)&0.119 ( 0.027 ) + c&ctio 0.9-m&3.154 ( 0.026)&@xmath120.006 ( 0.004)&0.085 ( 0.022 ) + & ctio 1.5-m&2.062 ( 0.031)&@xmath120.005 ( 0.004)&0.071 ( 0.018 ) + d&ctio 0.9-m&3.923 ( 0.024)&@xmath120.016 ( 0.004)&0.048 ( 0.019 ) + & ctio 1.5-m&2.901 ( 0.011)&@xmath120.015 ( 0.004)&0.042 ( 0.027 ) +

we investigate the problem of shaping radially symmetric annular beams into desired intensity patterns along the optical axis . within the fresnel approximation , we show that this problem can be expressed in a variational form equivalent to the one arising in phase retrieval . using the uncertainty principle we prove various rigorous lower bounds on the functional ; these lower bounds estimate the error for the beam shaping problem in terms of the design parameters . we also use the method of stationary phase to construct a natural _ ansatz _ for a minimizer in the short wavelength limit . we illustrate the implications of our results by applying the method of stationary phase coupled with the gerchberg - saxton algorithm to beam shaping problems arising in the remote delivery of beams and pulses . beam shaping ; localized waves ; phase retrieval ; paraxial wave equation ; fresnel approximation

1307.6289

in many applications it is desirable to shape a beam or pulse of light so that it has specific properties along the optical axis . in particular , for applications in microscopic imaging @xcite , optical tweezers @xcite , laser micro - machining @xcite , dressing of optical filaments @xcite , filament formation @xcite and long - range laser ablation @xcite , to name a few , it is important to have a well controlled beam with a nearly uniform intensity along the optical axis . however , due to their wave nature localized packets of light will broaden spatially through diffraction . for example , gaussian beams of width @xmath1 and wavenumber @xmath2 double in spatial extent over the rayleigh range @xmath3 @xcite . in a linear , isotropic medium the electric field of a beam of light polarized in the @xmath4 direction can be modeled by a wave packet of the form @xmath5 , where @xmath6 denotes the spatial coordinate along the optical axis , @xmath7 are the cartesian coordinates transverse to the optical axis , @xmath8 is time and @xmath9 the carrier angular frequency of the wave . within the slowly varying _ ansatz _ on @xmath10 , the propagation of a beam with initial data @xmath11 at the @xmath12 plane is described by the paraxial wave equation : @xmath13 where @xmath14 is the transverse laplacian defined by @xmath15 @xcite . exact solutions to equation ( [ intro : governingequation ] ) can be expressed in terms of an integral transform of the initial profile @xmath16 ; see [ appendix : solution ] . this integral transform is in fact equal to the fresnel diffraction integral @xcite . constructing solutions to ( [ intro : governingequation ] ) in free space that resist diffraction over length scales much larger than the rayleigh range has attracted considerable interest within the optics community . diffraction free beams are solutions to equation ( [ intro : governingequation ] ) satisfying the condition that for all @xmath6 , @xmath17 . if we make the _ ansatz _ @xmath18 then diffraction free beams can be found by solving the two dimensional helmholtz equation . indeed , plane waves , bessel beams @xcite , mathieu beams @xcite and parabolic beams @xcite are examples of such solutions in various orthogonal coordinate systems . note , however , that the diffraction free beams described above are not realizable in practice since they carry infinite energy . one technique for obtaining finite energy approximations of such solutions is apodization at the input plane @xmath12 . that is , if @xmath19 solves ( [ intro : governingequation ] ) then apodized approximations to @xmath19 are found by solving ( [ intro : governingequation ] ) with the initial data @xmath20 where @xmath16 is a function of finite extent . in @xcite so called `` bessel - gauss '' beams were created by apodizing bessel beams with a gaussian function and a general framework for constructing gaussian apodized solutions to ( [ intro : governingequation ] ) was developed in @xcite and @xcite . bessel - gauss beams have received considerable interest because they can be produced experimentally , for example by a diffraction grating with concentric grooves @xcite . alternatively , a gaussian beam passing through an axicon lens produces a ring beam with a truncated gaussian cross - section that results from the division of the beam around the axicon tip @xcite . the effect of these optical elements is to create a linear superposition of gaussian beamlets whose individual wave vectors @xmath21 lie on a cone of aperture angle @xmath22 . the mutual interference of these beamlets generates the bessel pattern and due to influx of energy supplied by the oscillatory wings of the bessel function to the central core , these waves are robust to nonlinear losses and physical barriers @xcite . however , for a truncation of width @xmath23 the bessel - gauss beam only accurately approximates the ideal bessel beam in a `` bessel - zone '' of width @xmath24 near the tip of the axicon @xcite and then transitions to a ring - beam in the far - field @xcite . it was also shown by bagini et . @xcite that bessel - gauss beams can be created remotely by focusing ring beams . specifically , in cylindrical coordinates @xmath25 bessel - gauss beams can be generated remotely by solving equation ( [ intro : governingequation ] ) with the initial data @xmath26 where @xmath27 is the peak value of the electric field , @xmath28 is the radius of the ring beam , @xmath1 is the characteristic width of the ring , @xmath29 is the focal distance , @xmath30 is the zero order bessel function of the first kind and @xmath31 is a parameter related to the focusing of the beam @xcite . note that for this initial data the argument of the bessel function is complex valued and in particular if @xmath32 no focusing it reduces to a modified bessel function of the first kind . in figure [ fig : genbesselbeams ] we provide a contour plot of the @xmath33 intensity profile for such a `` generalized '' bessel - gauss beams and select four radial cross sections to highlight the transition from a ring - beam into a bessel - gauss beam and back to a ring beam . as can be seen from figure [ fig : genbesselbeams ] , the intensity profile along the optical axis ( @xmath34 ) is not uniform in @xmath35 within the bessel - zone and is in fact given by a lorentzian - gaussian type function @xcite . the non - uniformity of the on - axis intensity limits the utility of bessel - beams in some applications . in this paper we explore the problem of applying additional phase shaping to the ring beam in order to create beams with specific properties along the optical axis . specifically , we consider the problem of constructing a phase function @xmath36 at the input plane @xmath12 that focuses a ring beam intensity profile @xmath37 of radius @xmath38 and width @xmath1 into a target intensity profile @xmath39 of width @xmath40 along the optical axis centered at a target distance @xmath29 . in figure [ fig : besselsubfig2 ] we illustrate the geometry and notation for the problem we are considering . the constants @xmath41 denote peak values of the electric field and @xmath42 , @xmath43 are smooth , non negative compactly supported functions with a maximum value of @xmath44 supported on the intervals @xmath45 and @xmath46 defined by @xmath47 respectively . simultaneous amplitude and phase shaping of the beam will achieve the task and indeed this problem was studied in @xcite . however , in practice this is achieved through holographic techniques which are difficult to implement and have low power efficiency @xcite . in figure [ fig : besselsubfig1 ] we illustrate one possible method for how phase shaping of ring beams can be achieved . first , a ring beam of radius @xmath38 and width @xmath1 is created by passing a collimated beam through an axicon or a concentric diffraction grating . the phase function is then applied , perhaps through a lens with a radially dependent thickness , which focuses the beam into the desired profile . the effect of adding a shaper phase @xmath36 can be realized by solving equation ( [ intro : governingequation ] ) with the initial data @xmath48 . the exact solutions to equation ( [ intro : governingequation ] ) can be expressed in terms of a hankel transform and the field along the optical axis is given by the following relationship : @xmath49 ( see [ appendix : solution ] ) . in this paper we consider the problem of optimizing the on - axis profile of the beam in the sense that we are interested in minimizing the @xmath0 norm : @xmath50 over the space of measurable phase functions @xmath51 . note that while the optimal axis is defined by the half - line @xmath52 we have defined the @xmath0 norm over the entire real axis . we make this definition as it allows us to use the @xmath0 isometry between a function and its fourier transform . moreover , this definition of the functional penalizes the beam wasting light outside the support of the target intensity distribution . since we are assuming that @xmath53 is compactly supported away from zero it follows from the change of variables @xmath54 and @xmath55 that @xmath56(\omega),\ ] ] where @xmath57 denotes the fourier transform from @xmath58-space into @xmath59-space . the optimal design problem can then abstractly be posed as finding a measurable phase function @xmath60 that minimizes the functional @xmath61 $ ] defined by @xmath62=\left\| g(\omega)- |\mathcal{f } \left[g(s ) \exp(i \varphi(s))\right](\omega)\right|\|_{l^2},\ ] ] where @xmath63 and @xmath64 are positive compactly supported functions . this abstract formulation can be identified with the specific application we are interested in by setting @xmath65 in the two - dimensional setting similar functionals have been studied within the context of shaping beams into desired intensity patterns in the focal plane of a lens @xcite . the variational problem we are considering is closely related to the problem of phase retrieval from two intensity measurements , i.e. the problem of determining the complex argument of a function given both knowledge of the modulus of a function and the modulus of its fourier transform . phase retrieval arises in applications in various fields ranging from astronomy @xcite , ultrashort pulses @xcite , radial beam shaping @xcite , pulse shaping @xcite , microscopy @xcite , to image reconstruction @xcite . a detailed review of the history of this problem and a number of numerical techniques that have been developed to solve this problem can be found in @xcite . within the context of phase retrieval , the most common technique for optimizing @xmath66 is the alternating projection algorithm pioneered by gerchberg and saxton @xcite and its variants such as the hybrid input - output algorithm discovered by fienup @xcite . in the original gerchberg - saxton algorithm the phase is recovered as follows : 1 . choose an initial guess @xmath67 . 2 . define @xmath68\right)$ ] . 3 . define @xmath69\right)$ ] . 4 . loop through items 2 and 3 until @xmath70 $ ] is sufficiently close to zero . through plancherel s identity it can be shown that the gerchberg - saxton algorithm is an error reducing algorithm in the sense that @xmath71\leq i[\phi_{n}]$ ] @xcite . however , this property alone does not guarantee convergence of the algorithm . while projection algorithms converge on convex sets @xcite , the projections employed by the gerchberg - saxton algorithm are analogous to projections onto the boundary of the unit ball in @xmath72 which is clearly not convex this lack of convexity commonly leads to stagnation of the algorithm away from the global minimum which must be overcome by additional ad hoc means @xcite . indeed , additional techniques such as the hybrid input - output algorithm are commonly implemented along with the gerchberg - saxton to force the algorithm out of local minimum @xcite . other techniques for phase retrieval have also been developed including in the two diemnsional setting a variational principle for the transport of intensity distributions between two planes orthogonal to the optical axis @xcite . in phase shaping , however , the infimum of @xmath66 may be significantly bounded away from zero and therefore when applied to our problem convergence of the gerchberg - saxton algorithm can not be assessed . indeed , without any further _ a priori _ information about the minimum value of @xmath66 stagnation of the algorithm is the only indication of possible convergence . moreover , if the gerchberg - saxton stagnates at a large value of @xmath66 it may indeed be at the global minimum for the problem . the paper is organized as follows . in section 2 we use elementary fourier analysis to deduce some basic results concerning the functional @xmath66 and translate these results in terms of the practical parameters for the phase shaping problem we are interested in . using plancherel s identity we obtain as a necessary condition for accurate beam shaping a normalization for the peak target intensity @xmath73 in terms of the design parameters : @xmath74 we also prove the existence of a minimum for this problem and in doing so obtain the regularity result that the optimal phase is the argument of an analytic function . finally , in using the uncertainty principle we prove _ ansatz free _ lower bounds on the minimum value of the functional . indeed we prove the following necessary condition for accurate beam shaping : @xmath75 although we discovered this result independently , romero and dickey are to the best of our knowledge the first to make the connection between the uncertainty principle and limitations in beam shaping arising from diffraction @xcite . our work goes beyond their treatment in that we prove in a mathematically precise manner the exact range of parameters for which the minimum value is bounded away from zero . in section 3 we use the stationary phase method to a construct a phase @xmath36 such that solutions to equation ( [ intro : governingequation ] ) with the initial data @xmath76 satisfy @xmath77 the phase is constructed as follows : 1 . the following initial value problem is solved : @xmath78 2 . the phase is found by integrating : @xmath79 a similar approach was taken by friberg in the analysis of a logarithmic axicon lens @xcite . while friberg showed that the logarithmic axicon lens maps the support of @xmath53 into @xmath43 in the short wavelength limit ( @xmath80 ) , it fails to match the target intensity . our approach is more general in that for our construction the intensity distribution is exactly matched in the short wavelength limit . we would like to point out , however , that we are not the first to discover the use of the method of stationary phase for beam shaping . indeed , after we successfully extended friberg s analysis in @xcite to our setting we discovered that romero and dickey developed essentially the same algorithm in the two - dimensional setting in @xcite and further developed the theory in @xcite . in section 4 we explore how the method of stationary phase coupled with the gerchberg - saxton algorithm can be applied to practical problems . we consider three specific applications . we first study the problem of creating a near - uniform intensity along the optical axis down range from a source . specifically , we take the input beam to be a gaussian ring and the target profile to be a @xmath81-th order super - gaussian . we show that the stationary phase approximation yields a good initial guess that is further improved upon by the gerchberg - saxton algorithm . we also highlight the role of the uncertainty principle and show that accurate phase shaping is impossible for sufficiently small values of @xmath1 or @xmath40 . we also consider in this section the problem of creating beams with oscillatory intensity profiles . we show that that as a consequence of the uncertainty principle there is a cutoff in the period of these oscillations below which phase shaping can not adequately match the target intensity . we illustrate this by applying our algorithm to periodic intensity profiles with small scale oscillations . finally , we show in this section how the method of stationary phase coupled with linear temporal chirping can be used to form pulses with specified temporal widths at the target distance @xmath29 . specifically , by taking a separable spatio - temporal _ ansatz _ on the initial data we show that spatial phase shaping can be employed to correct for the additional spatial spreading of the beam arising from linear temporal chirping . we conclude with a summary and discussion of our results . in this section we use the abstract variational formulation of our problem to obtain quantitative information about our beam shaping problem . let @xmath82 denote the space of lebesgue measurable functions on @xmath83 and @xmath84 the space of smooth compactly supported functions on @xmath85 . recall from the introduction that we are interested in minimizing the functional @xmath86 defined by @xmath87=\left \|g-\left |\mathcal{f}\left[g\exp(i \varphi)\right ] \right| \right\|_{l^2},\ ] ] with @xmath88 having support bounded away from the origin . in terms of the particular beam shaping problem we are interested in the _ shaper phase _ is given by @xmath89 , the _ input profile _ of the electric field is given by @xmath90 and the _ target profile _ is given by @xmath91 . in particular we will assume that the supports of @xmath64 and @xmath63 are given by the intervals @xmath92 and @xmath93 defined by @xmath94 and @xmath95 respectively where @xmath96 . recall plancherel s theorem which states that @xmath97\right \|_{l^2}$ ] . the following lemma immediately follows from plancherel s identity and applications of the triangle inequality and reverse triangle inequality . [ form : thm : lowerbound1 ] let @xmath86 be the functional defined by equation ( [ form : deffunc ] ) with @xmath98 having supports @xmath92 and @xmath99 defined by equations ( [ form : sg ] ) and ( [ form : sg ] ) respectively . then , @xmath100 \leq \|g\|_{l^2 } + \sqrt{2\pi } \|g\|_{l^2}.\ ] ] if the lower bound in lemma [ form : thm : lowerbound1 ] is zero , i.e. @xmath101 , then in terms of the design parameters we are interested in we have that @xmath102 this provides a natural normalization for @xmath103 in terms of @xmath27 and the @xmath0 norms for the specific profiles of the input and target intensity patterns . it is important to note that in our initial formulation of the problem we did not specify any constraints on the target electric field . throughout the rest of this paper we will assume this normalization and hence that @xmath101 which will be a critical assumption in the stationary phase analysis as well . we can also use plancherel s identity to obtain functionals that are defined over a larger admissible set but have the same infimum as @xmath66 . these functionals are easier to analyze and we use them to prove the existence of a minimizer for our original problem . for @xmath88 , define the functional @xmath104 by @xmath105=\left \| g\exp(i \psi)- \mathcal{f}\left [ g\exp(i \varphi ) \right ] \right\|_{l^2}.\ ] ] letting @xmath106(\omega)\right)$ ] it follows for a fixed @xmath107 that @xmath108 $ ] is minimized by setting @xmath109 . therefore , @xmath110=i[\varphi]$ ] and consequently @xmath111=\inf_{\varphi \in \mathcal{m } } i[\varphi].\ ] ] let @xmath112 denote the unit ball in the @xmath113 norm . that is , @xmath114 define another functional @xmath115 by @xmath116=2 \re\left(\left < gw,\mathcal{f}\left[g\zeta\right ] \right>\right).\ ] ] since we are assuming @xmath101 it follows that @xmath117=2\left\|g\right\|_{l^2}+i^{\prime}[\exp(i\varphi),\exp(i\psi)].\ ] ] therefore , to study minimizers of @xmath118 it is equivalent to study minimizers of @xmath119 . furthermore , we can use the bilinearity of @xmath119 to prove the existence of a minimizer for our functional of interest . [ thm : existence ] let @xmath115 be the functional defined by equation ( [ form : deffunc3 ] ) with @xmath98 having supports @xmath92 and @xmath99 defined by equations ( [ form : sg ] ) and ( [ form : sg ] ) respectively . there exists @xmath120 such that for all @xmath121 @xmath122\leq i^{\prime}[\zeta , w].\ ] ] to prove the result we proceed in two steps . first we use the direct method to show that minimizers @xmath123 of @xmath119 over @xmath124 exist . second we show that these minimizers satisfy @xmath125 , @xmath126 @xmath127 and hence @xmath128 defined by @xmath129 , @xmath130 are the desired minimum . let @xmath131 be a minimizing sequence for @xmath119 over @xmath124 . that is , @xmath132=\inf_{\zeta,\omega\in b_1^{\infty}}i^{\prime}[\zeta,\omega].\ ] ] by the banach - alaoglu theorem @xmath112 is weak-@xmath133 compact and hence @xmath124 is also compact in the product topology . therefore , reindexing a subsequence if necessary , we can assume without loss of generality that there exists @xmath134 such that @xmath135 and @xmath136 @xcite . let @xmath137 $ ] and @xmath138 $ ] . from weak-@xmath133 convergence it follows that for fixed @xmath59 that @xmath139 furthermore , for all @xmath59 it follows from holder s inequality that @xmath140 therefore , by lebesgue s dominated convergence theorem it follows that @xmath141 finally , from weak-@xmath133 convergence of @xmath142 and the strong convergence of @xmath143 it follows that @xmath144=\lim_{n\rightarrow \infty}i^{\prime}[\zeta_n,\omega_n]=\inf_{\zeta,\omega\in b_1^{\infty}}i^{\prime}[\zeta,\omega].\ ] ] that is , @xmath145 minimizes @xmath119 over @xmath124 . since @xmath146 we deduce from the paley - wiener theorem that @xmath147 is analytic and hence has only isolated zeros @xcite . let @xmath148 and @xmath149 . therefore , @xmath144=-2\int_{\mathbb{r } } \left| h(\omega ) \right| \left| w(\omega ) \right| g(\omega)\cos\left ( \theta(\omega)-\psi(\omega)\right)\,d\omega\ ] ] is minimized if and only if @xmath150 a.e . and there exists an integer valued function @xmath151 such that @xmath152 . from parseval s equality a similar argument proves that @xmath153 a.e .. let @xmath154 be the functional defined by equation ( [ form : deffunc ] ) with @xmath98 having supports @xmath92 and @xmath99 defined by equations ( [ form : sg ] ) and ( [ form : sg ] ) respectively . there exists @xmath155 such that for all @xmath107 @xmath156 < i[\varphi].\ ] ] let @xmath157 be the functional defined by equation ( [ form : deffunc2 ] ) and let @xmath128 be the global minimizer for @xmath118 . if we let @xmath158\right)(\omega)\ ] ] then without loss of generality we can assume @xmath159 and hence @xmath156=\overline{i}[\overline{\varphi},\overline{\psi}]\leq \inf_{\varphi\in \mathcal{m}}i[\varphi].\ ] ] if @xmath154 is the functional defined by equation ( [ form : deffunc ] ) with @xmath98 having supports @xmath92 and @xmath99 defined by equations ( [ form : sg ] ) and ( [ form : sg ] ) respectively then [ cor : lowerbound1 ] @xmath160>0.\ ] ] let @xmath155 be the global minimizer of @xmath66 over @xmath82 . since @xmath146 the paley - weiner theorem implies that @xmath161 $ ] is analytic . therefore , @xmath161 $ ] can not be compactly supported on @xmath83 and since @xmath162 is compactly supported on @xmath99 it follows that @xmath163\geq \int_{\mathbb{r}\backslash s_g(z_d , w_t)}\left| \mathcal{f}[g\exp\left(i \overline{\varphi}\right)]\right|^2(\omega)\,d\omega>0.\ ] ] [ cor : regularity ] let @xmath154 be the functional defined by equation ( [ form : deffunc ] ) with @xmath98 having supports @xmath92 and @xmath99 respectively . if @xmath155 minimizes @xmath66 then @xmath164 is analytic . since @xmath165 minimizes @xmath66 over @xmath82 it follows from theorem [ thm : existence ] that there exists @xmath166 such that @xmath167 minimizes @xmath119 . consequently , from the same argument used in the proof of theorem [ thm : existence ] and parseval s theorem it follows that @xmath168(s)\right).\ ] ] since @xmath162 it follows from the paley - wiener theorem that @xmath169 is analytic . it follows from the previous corollary that the minimizers of @xmath66 are well behaved in the sense that they are smooth except possibly for jump discontinuities of @xmath170 . the uncertainty principle quantifies the impossibility of localizing both a function and its fourier transform @xcite . therefore , if the length of the support of @xmath63 and @xmath64 are both sufficiently small then the minimum value of @xmath66 will necessarily be significantly bounded away from zero . in particular , while corollary [ cor : lowerbound1 ] quantifies that the minimum value of @xmath66 is not zero the uncertainty principle can give us quantitative information about how large the minimum value can be . this is made precise by the following theorem . [ thm : unc1 ] let @xmath171 be the functional defined by equation ( [ form : deffunc ] ) with @xmath98 having supports @xmath92 and @xmath99 respectively . if @xmath172 , we have the lower bound @xmath160\geq \|g\|_{l^2}\left(1-\frac{1}{\sqrt{\pi}}\sqrt{\frac{2kw_tr_0w_0}{4z_d^2-w_t^2}}\right).\ ] ] by the reverse triangle inequality it follows that @xmath173&= \left\|g-\mathcal{f}\left[g\exp\left ( i \varphi\right)\right]\right\|_{l^2}\\ & \geq \left(\int_{s_g(z_d , w_t ) } \left(g(\omega)-\left|\mathcal{f}\left[g\exp\left ( i \varphi \right)\right](\omega)\right| \right)^2\,d\omega \right)^{\frac{1}{2}}\\ & \geq \left| \left(\int_{s_g(z_d , w_t ) } g^2(\omega)\,d\omega \right)^{\frac{1}{2 } } - \left(\int_{s_g(z_d , w_t ) } \left|\mathcal{f}\left[g\exp\left ( i \varphi \right)\right](\omega)\right|^2 \,d\omega \right)^{\frac{1}{2 } } \right|\\ & = \left| \|g\|_{l^2 } - \left(\int_{s_g(z_d , w_t ) } \left|\mathcal{f}\left[g \exp\left ( i \varphi \right)\right](\omega)\right|^2 \,d\omega \right)^{\frac{1}{2 } } \right|.\\\end{aligned}\ ] ] furthermore , by direct calculation , an application of the cauchy - schwarz inequality and lemma [ form : eqn : normalization ] we have that @xmath174(\omega)\right|^2 \,d\omega \right)^{\frac{1}{2 } } & \leq \sqrt { \frac{2k w_t}{4z_d^2-w_t^2}}\left \| \mathcal{f}\left[g \exp\left ( i \varphi \right)\right ] \right\|_{l^{\infty}}\\ & \leq \sqrt { \frac{2k w_t}{4z_d^2-w_t^2}}\|g(s)\|_{l^1}\\ & = \sqrt { \frac{2k w_t}{4z_d^2-w_t^2}}\int_{s_g(r_0,w_0)}\left|g(s)\right|\,ds\\ & \leq \sqrt { \frac{4k w_tr_0w_0}{4z_d^2-w_t^2}}\|g \|_{l^2}\\ & = \sqrt { \frac{4k w_tr_0w_0}{4z_d^2-w_t^2}}\frac{1}{\sqrt{2\pi}}\|g \|_{l^2}.\end{aligned}\ ] ] using the same notation as in @xcite we define the dimensionless parameter @xmath31 by @xmath175 it then follows from theorem [ thm : unc1 ] , that @xmath176\geq \|g\|_{l^2}\left(1-\sqrt{\frac{\beta}{\pi}}\right)}$ ] , for any choice of the phase function @xmath60 . in particular , if @xmath177 , there is no way of using phase shaping to match the target profile . we emphasize that the result is only a lower bound for the functional @xmath61 $ ] , and in particular it does not follow that if the parameters @xmath178 and @xmath29 are such that @xmath179 , then the infimum of @xmath66 is zero . also , this result is somewhat crude in that it is independent of the exact forms of @xmath63 and @xmath64 . below is a refinement of theorem [ thm : unc1 ] that takes into account the @xmath180 norms of @xmath64 and @xmath63 . [ thm : unc2 ] let @xmath171 be the functional defined by equation ( [ form : deffunc ] ) with @xmath98 having supports @xmath92 and @xmath99 defined by equations ( [ form : sg ] ) and ( [ form : sg ] ) respectively . 1 . if @xmath172 then @xmath160\geq \sqrt{2\pi}\|g\|_{l^2 } \left(1- \sqrt{\frac{2kw_t}{4z_d^2-w_t^2}}\frac{\|g\|_{l^1}}{\left\|g\right\|_{l^2}}\right).\ ] ] 2 . if if @xmath172 then @xmath160\geq \left\|g\right\|_{l^2}\left ( 1 - \sqrt{\frac{r_0w_0}{\pi}}\frac{\|g\|_{l^1 } } { \left\| g \right\|_{l^2}}\right).\ ] ] item 1 follows from the proof of theorem [ thm : unc1 ] and plancherel s identity . to prove item 2 we note that from plancherel s identity and the triangle inequality that @xmath181\geq \sqrt{2\pi } \left|\left(\int_{s_g(r_0,w_0 ) } \left| \mathcal{f}^{-1}[g\exp\left(i \psi \right)](s)\right|^2\,ds\right)^{\frac{1}{2}}-\|g\|_{l^2}\right|.\ ] ] following the same arguments as in the proof of theorem [ thm : unc1 ] it follows that @xmath182(s)\right|^2\,ds\right)^{\frac{1}{2}}\leq \frac{1}{2\pi } \sqrt{2r_0w_0}\|g \|_{l^1}.\ ] ] the result follows from applying plancherel s identity to @xmath183 . theorems [ thm : unc1 ] and [ thm : unc2 ] are estimates for the global error in the beam shaping problem in terms of the design parameters . the same techniques can be used to obtain local information as well . specifically , the uncertainty principle implies that if @xmath63 has sufficiently small scale features within its support then the infimum of @xmath66 will be bounded away from zero . to make this statement precise we look at the local error for an interval of width @xmath184 centered at @xmath185 on the optical axis specifically , for all @xmath186 satisfying @xmath187 define @xmath188 by @xmath189&=\left \|g-\left |\mathcal{f}\left[g\exp(i \varphi)\right ] \right| \right\|_{l^2(s_g(z_d^{\prime},w_t^{\prime}))}\\ & = \left(\int_{s_g(z_d^{\prime},w_t^{\prime})}\left(g(\omega)-\left |\mathcal{f}\left[g\exp(i \varphi)\right](\omega ) \right|\right)^2\,d\omega\right)^{\frac{1}{2}},\end{aligned}\ ] ] which is a local measure of the @xmath0 error on the interval @xmath190 along the optical axis . the following corollary follows from the same arguments used to prove theorem [ thm : unc1 ] . [ cor : unc1 ] let @xmath191 be the functional defined by equation ( [ form : func4 ] ) with @xmath98 having supports @xmath92 and @xmath99 defined by equations ( [ form : sg ] ) and ( [ form : sg ] ) respectively . suppose @xmath192 satisfy @xmath193 and @xmath194 . if @xmath172 and @xmath195 then @xmath196\geq \|g\|_{l^2(s_g(z_d^{\prime},w_t^{\prime}))}-\frac{1}{\sqrt{\pi}}\sqrt{\frac{2kw_t^{\prime}r_0z_d^{\prime}}{4\left(z_d^{\prime}\right)^2-\left(w_t^{\prime}\right)^2}}\|g\|_{l^2}.\ ] ] [ cor : unc2 ] let @xmath191 be the functional defined by equation ( [ form : func4 ] ) with @xmath98 having supports @xmath92 and @xmath99 defined by equations ( [ form : sg ] ) and ( [ form : sg ] ) respectively . suppose @xmath192 satisfy @xmath193 and @xmath194 . if @xmath172 and @xmath197 then @xmath196\geq \|g\|_{l^2(s_g(z_d^{\prime},w_t^{\prime}))}-\sqrt{\frac{2kw_t^{\prime}}{4\left(z_d^{\prime}\right)^2-\left(w_t^{\prime}\right)^2}}\|g\|_{l^1}.\ ] ] [ form : remark ] the local error @xmath198 also provides a lower bound for the global error @xmath66 . indeed , if we let @xmath199 then a sharper estimate for the lower bound on the global error can be found by considering : @xmath200&\geq \sup _ { ( z_d^{\prime } , w_t^{\prime})\in \mathcal{a } } \inf_{\varphi \in \mathcal{m}}i^{w_t^{\prime}}_{z_d^{\prime}}[\varphi]\\ & \geq \sup_{(z_d^{\prime},w_t^{\prime})\in \mathcal{a } } \left(\|g\|_{l^2(s_g(z_d^{\prime},w_t^{\prime}))}-\frac{1}{\sqrt{\pi}}\sqrt{\frac{2kw_tr_0z_d^{\prime}}{4\left(z_d^{\prime}\right)^2-\left(w_t^{\prime}\right)^2}}\|g\|_{l^2}\right ) . \nonumber\end{aligned}\ ] ] recall that the method of stationary phase can be used to obtain asymptotic expansions for integrals of the form @xmath201 where @xmath202 , @xmath203 is real , @xmath204 is a smooth real valued function and @xmath147 is a smooth but not necessarily analytic complex valued function . if @xmath204 has one critical point at @xmath205 , i.e. @xmath206 , and @xmath207 then @xmath208 as @xmath209 @xcite . in this section we show how the method of stationary phase can be used to approximate the on - axis profile of the electric field in the limit of short wavelengths and we show how this approximation can be used to construct initial guesses for the optimal phase . first , we must define precisely what we mean by short wavelengths in terms of the input parameters and the length scales in the parameter . using the geometry of the problem as in figure [ fig : besselsubfig1 ] , we define the dimensionless quantities @xmath210 equation ( [ intro : opticalaxis ] ) gives the on - axis electric field : @xmath211 where we used the fact that the input profile @xmath212 is only supported on @xmath213 $ ] . assuming @xmath214 , we expand the phase in powers of @xmath215 and @xmath216 to obtain @xmath217 note that the first three terms in the expansion of @xmath218 do not depend on both @xmath219 and @xmath220 jointly . consequently , they can be absorbed as phase factors in @xmath221 or @xmath222 as below , and they _ do not affect the intensity _ of the input @xmath221 or the on - axis profile @xmath222 . dropping the higher order terms in @xmath215 and @xmath216 , we get @xmath223 \\ & \times e^{i \left(\phi - \frac{k r_0 w_0 w_t \zeta \varrho } { z_d^2 } \right ) } ( r_0 + w_0 \varrho)\,d\varrho.\end{aligned}\ ] ] with @xmath219 . in order to use phase variations for optical beam shaping , it is clear that the optical path differences over the relevant region @xmath224 should be many times the wavelength , so that the beam can interfere with itself . this allows for the possibility of controlling the intensity of the beam as a function of position . since the variations in @xmath220 and @xmath219 are @xmath225 by construction , the preceding calculation determines the correct notion of the short wavelength limit in our problem , namely @xmath226 we include an additional factor of 2 in defining @xmath227 so that for @xmath228 , the dimensionless parameter @xmath31 , whose definition was motivated by theorem [ thm : unc1 ] _ is identical _ to the non - dimensional wave number @xmath227 arising from the analysis of beam shaping using phase variations . this argument therefore gives an alternative physical interpretation of the parameter @xmath31 it is the phase variation within the beam arising from the different optical path lengths for different parts of the beam in our geometry ( fig . [ fig : besselsubfig1 ] ) . we now assume @xmath229 , @xmath230 , @xmath231 and formally apply the method of stationary phase to approximate the on - axis electric field . define dimensionless quantities by @xmath232 then by equation ( [ intro : opticalaxis ] ) it follows that the on - axis profile of solutions to the paraxial wave equation is given by @xmath233 if @xmath234 is a given smooth function , then for a fixed value of @xmath235 stationary points of the phase can be found by solving the equation @xmath236 for @xmath237 . if @xmath234 is a monotone decreasing function in @xmath237 and @xmath238 is either strictly convex or concave then there exists one solution to this equation for each value of @xmath239 and we can implicitly define a function @xmath240 by @xmath241 therefore , if we let @xmath242 then by the method of stationary phase it follows that if @xmath243 then @xmath244 as @xmath245 . equation ( [ stat : approximation2 ] ) can be used to construct an accurate approximation to an optimal @xmath238 . formally , if we set @xmath251 then the following initial value problem must be satisfied : @xmath252 the phase corresponding to @xmath253 can then be found by integration : @xmath254 equations ( [ stat : zdifferentialequation ] ) and ( [ stat : phasedifferentialequation ] ) are a simple system of differential equations whose solution we will use as an approximation to the near optimal phase . solutions to ( [ stat : zdifferentialequation ] ) must also satisfy the additional physical constraint that @xmath255 , that is , the stationary phase approximation maps the support of @xmath53 to the support of @xmath43 . the following lemma shows that this condition is met if the normalization ( [ form : eqn : normalization ] ) is satisfied and again highlights the important role this normalization takes . if @xmath253 is a solution to the initial value problem ( [ stat : zdifferentialequation ] ) and @xmath103 and @xmath27 satisfy equation ( [ form : eqn : normalization ] ) then @xmath255 . suppose @xmath253 is a solution to the initial value problem ( [ stat : zdifferentialequation ] ) and @xmath256 . from the normalization ( [ form : eqn : normalization ] ) and the fact that @xmath257 solves ( [ stat : zdifferentialequation ] ) it follows that @xmath258 now , if @xmath259 then @xmath260 if @xmath261 then since @xmath249 is a continuous monotone increasing function of @xmath262 there exists @xmath263 such that @xmath264 . consequently , @xmath265 therefore it follows that @xmath266 . the full electric field near the bessel - zone @xmath267 can also be approximated in the focal region @xmath268 using the method of stationary phase . in the dimensionless coordinates we obtain through the method of stationary phase that : latexmath:[\[\label{stat : offaxis } that is , the beam essentially forms a bessel - beam in the focal region . however , the function @xmath30 is also a rapidly oscillating function and the above approximation is only valid for radii satisfying @xmath270 . this expression is not applicable outside the focal region , and of course does not can not capture the transition of the ring beam into a bessel - gauss beam and back into a ring beam . in this subsection we use the method of stationary phase coupled with the gerchberg - saxton algorithm to obtain approximations to the optimal phase when the input profile @xmath42 is gaussian and the target profile @xmath271 is super - gaussian . specifically , for @xmath81 a positive integer and @xmath272 we assume that @xmath273 since we are using input and target functions that decay exponentially fast we can use laplace s method @xcite to obtain accurate estimates for the normalization . directly applying laplace s method it follows that @xmath274 and explicitly calculating we have that @xmath275 therefore , by equation ( [ form : eqn : normalization ] ) it follows that @xmath276 where @xmath277 is bounded between @xmath278 and @xmath279 . equation ( [ bb : peakintensity ] ) gives the ratio of the peak intensity of the on - axis profile of the beam to the peak intensity of the input beam . the input and target functions defined above are not smooth bump functions and strictly speaking the theory and numerical algorithm we developed in the previous two sections does not apply . however , the functions given by equation ( [ bb : functions ] ) are schwartz class and they can be very accurately approximated by non - smooth functions with compact support which in turn can be approximated by smooth functions with compact support @xcite . in particular , we truncate these functions by forcing them to be zero for values in which the function is less than @xmath280 . this gives us the following estimates for the width of the supports @xmath281 with the above normalization , equations ( [ stat : zdifferentialequation ] ) and ( [ stat : phasedifferentialequation ] ) can be numerically solved to construct a guess for an optimal phase function @xmath51 . using @xmath51 to initialize the gerchberg - saxton algorithm , further corrections to the stationary phase construction can be obtained . to illustrate the potential applicability of our techniques to remote laser ablation , we applied the algorithm outlined in the previous section with 100 iterations of the gerchberg - saxton algorithm for various input and target widths with the following parameters fixed @xmath282 the exact profile along the optical axis is then obtained by applying the fast fourier transform ( fft ) to equation ( [ form : fouriertrans ] ) . in figure [ fig : besseltargetwidth ] we plot the normalized value of @xmath66 for both the stationary phase analysis and the gerchberg - saxton algorithm for @xmath283 and @xmath184 ranging from @xmath284 to @xmath285 . for this range of parameters , @xmath31 varies from @xmath286 to @xmath287 . the inset plot in figure [ fig : besseltargetwidth ] illustrates a subset of the profiles obtained for @xmath288 . in figure [ fig : besselinputwidth ] we plot the normalized value of @xmath66 for both the stationary phase analysis and the stationary phase analysis coupled with the gerchberg - saxton algorithm for @xmath289 ranging from @xmath290 to @xmath291 and @xmath292 . for this range of parameters , @xmath31 varies from @xmath293 to @xmath294 . the inset plot in figure [ fig : besselinputwidth ] illustrates the intensity profile obtained for @xmath295 . figures [ fig : besseltargetwidth ] and [ fig : besselinputwidth ] also illustrate the important role that @xmath31 not only plays in terms of the lower bound presented in theorem [ thm : unc1 ] but also the accuracy of the stationary phase approximation . for @xmath296 @xmath297 , the lower bound in theorem [ thm : unc1 ] is relevant and as is guaranteed by theorem [ thm : unc1 ] neither the stationary phase approximation nor the gerchberg - saxton algorithm , or indeed any other algorithm can yield accurate shaping of the beam . for @xmath184 ranging from @xmath298 to @xmath299 ( @xmath300 to @xmath301 ) theorem [ thm : unc1 ] no longer applies and the stationary phase analysis yields a guess that is further improved upon by the gerchberg - saxton algorithm . for @xmath302 ( @xmath301 ) , the stationary phase analysis yields a very accurate guess that is only slightly improved upon by the gerchberg - saxton algorithm . in figures [ fig : phaseplot1 ] and [ fig : phaseplot2 ] we plot surface and contour plots of the phase function found by coupling the stationary phase approximation with the gerchberg - saxton algorithm for the parameters given by equation ( [ bb : parameters ] ) with @xmath283 and @xmath184 ranging from @xmath303 to @xmath285 . specifically , we plot @xmath304 the term @xmath305 corresponds to a linear shift in the fourier variable @xmath59 and also has an interpretation in optics as the effect of a lens that focuses a beam at the distance @xmath29 @xcite . interestingly , when @xmath306 ( @xmath286 ) the phase @xmath307 is essentially constant . that is , except for the focusing term @xmath305 additional phase shaping can not improve upon the effect of a focusing lens . however , for @xmath308 ( @xmath287 ) there is a significant difference between the numerically obtained phase and the focusing term . from figures [ fig : statphaseplot1 ] and [ fig : statphaseplot2 ] it is again evident that the phase obtained from that gerchberg - saxton algorithm converges in the limit of @xmath309 to the one obtained using the stationary phase alone . in laser ablation while it is important to have a beam with a near uniform on - axis intensity profile , the amount of power delivered by the beam is also of critical importance . in particular , numerical experiments indicate that for critical collapse of a bessel - gauss beam to occur the power contained in the central lobe of the beam must be near the critical power of the townes profile @xcite . for a fixed value of @xmath35 the power delivered by a beam in a circular cross section of radius @xmath310 about the optical axis is proportional to the following integral : @xmath311 using the asymtotic formula ( [ stat : offaxis ] ) for the off - axis profile electric field , the power in the central core of the bessel beam at the target distance @xmath312 can be approximated in the short wavelength limit . if we assume that @xmath313 , then for @xmath312 the electric field is approximately zero at the value @xmath314 . consequently , the power delivered in the central core of the bessel beam at the distance @xmath29 is approximated by @xmath315 from the normalization ( [ form : eqn : normalization ] ) it follows that in the short wavelength limit we have that : @xmath316 where @xmath317 is the initial power of the beam . therefore , from equation ( [ bb : wtnorm ] ) it follows that @xmath318 this equation illustrates the trade - off between the amount of power delivered within the central lobe and the accuracy of the shaping . in particular , for @xmath319 and @xmath320 the shaping will be very accurate but the amount of power delivered by the beam in the central core will be much smaller than the initial power of the beam . for the range of parameters used to generate figure [ fig : besseltargetwidth ] , for example , this approximation for @xmath321 ranges from @xmath322 to @xmath323 . in this subsection we explore the possibility of using beam shaping to create intensity distributions with an oscillatory pattern for length scales that are applicable to micromachining . again we use a gaussian input beam : @xmath324 with @xmath325 and @xmath326 . for integer values of of @xmath327 we take the target profile to be : @xmath328 with @xmath329 , @xmath330 and we take @xmath331 . as before , we can then apply the method of stationary phase coupled with the gerchberg - saxton algorithm to numerically optimize @xmath66 . now , it follows that @xmath332 for these parameters and hence theorem [ thm : unc1 ] is not relevant . however , as the value of @xmath327 increases the target intensity develops fine scale oscillations of approximate width @xmath333 and therefore by corollary [ cor : unc1 ] we expect the accuracy of the beam shaping to decrease with increasing @xmath327 . as in section 2 define @xmath334 and for @xmath335 define the interval @xmath336 by @xmath337 to capture the influence of the small scale oscillations on the error we define the following local quantity : @xmath338 which is a local measure of the error in matching the target intensity within one period . from corollary [ cor : unc1 ] it follows that if @xmath339 then the small scale oscillations can not be matched by any phase function . in figure [ fig : bumps ] we plot the on - axis intensity profiles for the beams obtained by using the stationary phase as an initial guess for the gerchberg - saxton algorithm with @xmath340 iterations . we selected the values @xmath341 to illustrate the influence of @xmath327 on the accuracy of the shaping . as we can see , with increasing @xmath327 the method of stationary phase yields a guess that matches the support of the function but fails to capture the small scale oscillations . for these values of @xmath327 we have that @xmath342 , @xmath343 , @xmath344 and @xmath345 and indeed for @xmath346 , as is to be expected , the gerchberg - saxton algorithm can not match the target intensity . it is interesting to note , however , for large @xmath327 that although the gerchberg - saxton algorithm fails to capture the amplitude of the oscillations it does a decent job recovering the period . . the blue dashed - dotted curves are the on - axis intensities obtained through the stationary phase analysis alone , the solid red curves are the on - axis intensities obtained by coupling the stationary phase algorithm with the gerchberg - saxton algorithm and the green dashed curves are the target intensity profile . the function @xmath347 quantifies the accuracy of the beam shaping and in particular for @xmath339 the amplitude of the oscillations is not matched . ] in this subsection we show that for a linear medium and under the paraxial approximation the techniques we developed in this paper can be combined with linear temporal chirping to remotely deliver pulses with a desired temporal width at a target distance . the governing equation for such pulses is the temporal paraxial wave equation : @xmath348 where @xmath349 is the group velocity dispersion for a pulse with wavenumber @xmath2 @xcite . if we assume separable initial data of the form @xmath350 then the solution to equation ( [ eqn : temppw ] ) can be expressed in terms of the hankel transform of a convolution : @xmath351 again , we assume that the input spatial profile is given by @xmath352 , where @xmath353 , @xmath354 is supported on the interval @xmath355 and @xmath51 is a measurable function corresponding to a shaper phase . we assume that the input temporal profile is a linearly chirped gaussian pulse : @xmath356 where @xmath357 is a measure of the temporal width of the initial pulse and @xmath358 is a linear chirping parameter . with this initial data , the on - axis intensity is given by @xmath359 with @xmath360 a quadratic function defined by @xmath361 the function @xmath360 quantifies how the temporal width broadens or contracts as it propagates ; indeed the characteristic temporal width is @xmath362 . we assume that the target profile along the optical axis is also given by a gaussian pulse of the form @xmath363 where @xmath364 , @xmath365 with support @xmath366 and @xmath367 . to ensure that @xmath368 matches the desired temporal width at @xmath29 and the change in @xmath368 over the support of @xmath43 is as small as possible we must have that @xmath369 solving these equations yields the conditions : @xmath370 define @xmath371 by @xmath372 and as we did before define @xmath373 , @xmath374 . the problem of designing the optimal shaper phase with this temporal chirping then corresponds to minimizing the functional @xmath171 defined by @xmath61=\|g-|\mathcal{f}(g \exp(i \varphi))| \|_{l^2}$ ] which problem corresponds to the variational problem we considered before except for the additional term @xmath375 accounting for the additional modification to the spatial intensity arising from temporal broadening . in figure [ fig : temporalprofile ] we plot the on - axis spatial - temporal profile for a gaussian pulse propagating in air obtained by using the method of stationary phase . for this figure we again assumed that @xmath376 with @xmath377 @xmath378 following the experimental results in @xcite we took the group velocity dispersion of air for this wavelength of light to be @xmath379 and we assumed a target temporal width of @xmath380fs . for these values the initial temporal width is @xmath381fs and the chirping parameter is @xmath382fs . in this paper we studied the problem of phase shaping an annular beam into a desired intensity profile along the optical axis by analyzing minimizers of the functional @xmath383 defined by equation [ form : deffunc ] . formally , this functional is equivalent to the ones arising in phase retrieval from two intensity measurements . typically , due to its speed and ease of implementation , the gerchberg - saxton algorithm is used to construct approximate minimizers . while the gerchberg - saxton algorithm is an error reducing algorithm there are no known general convergence results for this algorithm . in particular the algorithm typically stagnates away from the global minimum which in phase retrieval can be overcome by additional modifications to the algorithm such as the hybrid input - output method @xcite or those in @xcite . moreover , in phase retrieval the infimum of @xmath66 is zero and convergence of the gerchberg - saxton algorithm and its modifications can be assessed by the proximity of @xmath66 to zero . however , as a consequence of the uncertainty principle the minimum value of @xmath66 may be significantly bounded away from zero . consequently it is not clear if modifications to the gerchberg - saxton algorithm are applicable to beam shaping problems nor is it clear what a sufficient convergence criterion of such an algorithm would be . the central result of our work is the identification of the dimensionless quantity @xmath384 and the associated lower bounds on the functional @xmath66 . we identified three scaling regimes for @xmath31 for which different quantitative and qualitative behaviors of the optimal profile occur . these regimes are as follows . 1 . for @xmath385 the uncertainty principle guarantees that the optimal profile will differ significantly from the target profile . this result has the consequence that neither the gerchberg - saxton algorithm nor any other numerical algorithm will yield accurate shaping of the beam . 2 . for @xmath231 the method of stationary phase yields an approximation to the problem that is very close to the optimal profile . this asymptotic regime can be considered within the geometrical optics setting in the sense that the light rays originating from the input plane can be accurately mapped to the target intensity profile through the method of stationary phase . indeed , for the problem of creating a beam with a nearly uniform on - axis intensity , the gerchberg - saxton algorithm coupled with the method of stationary phase yields nearly optimal profiles . in particular , the method of stationary phase can be used as a robust method for overcoming the stagnation issues of the gerchberg - saxton algorithm commented on in @xcite . 3 . for the intermediate regime where @xmath386 the phase produced by the method of stationary phase is significantly improved upon by the gerchberg - saxton algorithm . however , in this asymptotic regime the uncertainty principle no longer applies . moreover , the value of @xmath66 for @xmath51 obtained through the method of stationary phase is not close to zero . consequently , within this regime there is no clear convergence of the algorithm . following the reasoning in remark [ form : remark ] we could obtain more accurate lower bounds on the functional that account for the specific forms of the input and target intensity profile instead of simply the widths of their supports . however , in the regime @xmath387 a universal scaling law for the error in terms of the wavelength is not possible . the functional we considered was derived under the paraxial assumption and hence is applicable for linear beams with the same range of wavelengths and geometrical length scales as in the fresnel approximation . indeed , for the range parameters we considered in the applications , the fresnel number ranges from @xmath388 to @xmath389 . consequently , for these parameters the integral transform given by equation ( [ intro : opticalaxis ] ) very accurately approximates the propagation of the on - axis electric field for linear beams . therefore , the results of this paper and in particular the uncertainty principle are fundamental and place restrictions on the types of intensity profiles that can be achieved through linear phase shaping alone . for specific applications and in particular those requiring high power beams it may be necessary to consider the off - axis electric field and nonlinear interactions . equation ( [ stat : offaxis ] ) gives an asymptotic result indicating that the intensity pattern near the target distance will form a bessel - like function . numerical evidence indicates that for critical collapse of the beam to occur the power in the central core of the bessel function must be near that of the townes profile @xcite . for the applications we considered , it follows from equation ( [ form : eqn : normalization ] ) that the initial peak intensity of the ring beam @xmath390 is much smaller than the target peak intensity @xmath73 of the generated bessel beam and hence kerr nonlinearities will be negligible until the ring beam is focused into the bessel - like pattern . predicting the filamentation patterns of a beam after critical collapse is a challenging problem and perhaps the nonlinear geometrical optics method presented in @xcite could be used in conjunction with our methods near the target distance . the full effect of the kerr nonlinearities will be addressed in subsequent publications . additionally , for propagation of the beam over ranges of @xmath391 it will be necessary to consider the effects of turbulence in the atmosphere which is beyond the scope of this paper . consider the initial value problem with radially symmetric initial data : @xmath392 where @xmath53 is a smooth function with compact support . if we let @xmath393(k_{\perp})$ ] and @xmath394(r)$ ] denote the hankel transform and its inverse defined by @xmath395(k_{\perp})=\int_0^{\infty } g(r)j_0(rk_{\perp})\,dr \text { and } \mathcal{h}^{-1}[g](r)=\int_0^{\infty } g(k_{\perp})j_0(rk_{\perp})k_{\perp}dk_{\perp}\ ] ] then the solution to the initial value problem is given by @xmath396(k_{\perp } ) \exp\left(-\frac{i k_{\perp}^2 z}{2k } \right)j_0\left(rk_{\perp}\right)k_{\perp } dk_{\perp}.\ ] ] noting that @xmath397(r)=-\frac{ik}{z } \exp\left ( \frac{i k r^2}{2z}\right)\ ] ] it follows from the convolution theorem for hankel transforms @xcite that @xmath398 * acknowledgments : * the authors acknowledge funding support under the muri afosr grant fa9550 - 10 - 0561 . c.d . acknowledges funding support from afosr under the grant fa9550 - 10 - 1 - 0394 an s.v . acknowledges support from the nsf grant 0807501 . the authors wish to thank ewan wright for many useful discussions and bringing to our attention reference @xcite and we would like to thank colm dineen who constructed figure [ fig : besselsubfig1 ] . the authors would also like to thank the two anonymous referees who reviewed this paper . their comments greatly improved the organization and the presentation of our results . + * references * t. planchon , l. gao , d. e. milkie , m. w. davidson , j. a. galbraith , c. g. galbraith , e. betzig , rapid three - dimensional isotropic imaging of living cells using bessel beam plane illumination , nature methods 8 ( 2011 ) 417423 . k. stelmaszczyk , p. rohwetter , g. mjean , j. yu , e. salmon , j. kasparian , r. ackermann , j .- wolf , l. woste , long - distance remote laser - induced breakdown spectroscopy using filamentation in air , applied physics letters 85 ( 18 ) ( 2004 ) 39773979 . p. rohwetter , k. stelmaszczyk , l. wste , r. ackermann , g. mjean , e. salmon , j. kasparian , j. yu , j .- wolf , filament - induced remote surface ablation for long range laser - induced breakdown spectroscopy operation , spectrochimica acta part b : atomic spectroscopy 60 ( 7 ) ( 2005 ) 10251033 . t. graf , d. christodoulides , m. s. mills , j. moloney , s. c. venkataramani , e. m. wright , propagation of gaussian - apodized paraxial beams through first - order optical systems via complexcoordinate transforms and ray transfer matrices , j. opt . am . a 29 ( 9 ) ( 2012 ) 18601869 . j. wrzesinski , d. pestov , v. v. lozovoy , j. r. gord , m. dantus , s. roy , et al . , group - velocity - dispersion measurements of atmospheric and combustion - related gases using an ultrabroadband - laser source , opt . express 19 ( 6 ) ( 2011 ) 51635171 . n. gavish , g. fibich , l. t. vuong , a. l. gaeta , predicting the filamentation of high - power beams and pulses without numerical integration : a nonlinear geometrical optics method , physical review a 78 ( 4 ) ( 2008 ) 043807 .

in this paper we consider a large class of many - variable polynomials which contains generalisations of the classical hermite , laguerre , jacobi and bessel polynomials as special cases , and which occur as the polynomial part in the eigenfunctions of calogero - sutherland type operators and their deformations recently found and studied by chalykh , feigin , sergeev , and veselov . we present a unified and explicit construction of all these polynomials .

math-ph0703090

in this paper we discuss the construction of symmetric polynomials which arise as eigenfunctions of exactly solvable quantum many - body systems of calogero - sutherland type @xcite . in particular , we demonstrate that a particular construction method found and studied by us in simple special cases @xcite can be naturally extended to the full class of such model . our results provide a unified approach to many - variable generalisations of the classical hermite , lagueree , jacobi and bessel polynomials . we also show that it is natural to generalise our approach to a larger family of polynomials related to a deformation of the calogero - sutherland type systems found and studied by chalykh , feigin , sergeev , and veselov @xcite . this allows us to derive , for each such polynomial , a family of different representations labeled by a pair of non - negative integers @xmath0 . moreover , one can choose @xmath0 such that the representation is simplest , i.e. , a linear combination of certain explicitly given polynomials with the least number of terms . these results are non - trivial already for the non - deformed polynomials . to mention a specific example , we find that a jack polynomial @xcite has a @xmath0 representation if there exists a pair of non - negative integers @xmath1 such that the young diagram corresponding to this jack polynomial can be covered by the union of the two rectangular young diagrams of size @xmath2 ( i.e. , @xmath3 rows and @xmath4 columns ) and @xmath5 , respectively . moreover , the simplest such representation is obtained by minimizing @xmath6 . for example , the simplest possible representation of the jack polynomial corresponding to the young diagrams @xmath7 is for @xmath8 , @xmath9 , and @xmath10 , respectively , and in the first two cases our representation actually consists only of a single term . the main tool in our construction are certain identities which , to each ( deformed ) calogero - sutherland system , relates a system of the same type . in special cases , these identities are well known and related to a duality of the jack polynomials ( see , e.g. , @xcite ) , or , more generally , the algebra homomorphism which maps jack polynomials to so called super jack polynomials ( see , e.g. , @xcite ) . we deduce these identities in a unified manner , and thus unify , as well as extend , previously known results . while we mention these examples in the beginning to give an impression of the kind of results we obtain , the emphasis in our presentation is on our construction method . to avoid overloading our formulae we explain this method in detail for the non - deformed polynomials and the representations @xmath11 . the subsequent extension to the general case is then remarkably easy . as mentioned above , the polynomials studied in this paper play an important role in physics since they occur in prominent exactly solvable quantum mechanical models defined by partial differential operators of calogero - sutherland type . these latter operators play a central role in our construction , and we therefore proceed to explain their relation to the classical hermite , laguerre , jacobi and bessel polynomials . as elaborated below , to each such sequence of one - variable polynomials @xmath12 there exists a quantum mechanical model defined by a schrdinger operator @xmath13 with a particular potential function @xmath14 such that its eigenfunctions are of the form @xmath15 for particular @xmath16 functions @xmath17 and @xmath18 ( @xmath19 is a real variable , and we use the notation @xmath20 ) . many - variable generalisation of these polynomials appear in the following generalisation to an arbitrary number @xmath21 of particles : each one - particle schrdinger operator allows for a many - body generalisation @xmath22 with @xmath23 a particular two - body interaction potential such that @xmath24 has eigenfunctions @xmath25 which are labelled by integer vectors @xmath26 and are of the form @xmath27 @xmath28 where the functions @xmath29 are certain symmetric polynomials in @xmath21 variables @xmath30 generalising the polynomials @xmath31 . to avoid certain technicalities and since it includes the cases of main physical interests we will assume @xmath32 , even though many of our results actually hold true also for other values of @xmath33 . we mention already at this point that although the symmetry of the polynomials @xmath29 implies that a complete set of eigenfunctions can be labelled by partitions , i.e. , sequences @xmath34 of non - negative integers @xmath35 such that @xmath36 , it is in our approach not always natural to make this restriction , as will become evident below . these operators @xmath24 define exactly solvable quantum - many body systems of calogero - sutherland type @xcite ; see table [ table1 ] for a list of well - known examples ( note that we have added constants to the potentials @xmath14 to simplify certain formulae later on ) . a further interesting generalisation of the one - particle schrdinger operators is the following class of differential operators in two sets of variables @xmath37 and @xmath38 , where @xmath21 and @xmath39 are arbitrary non - negative integers : @xmath40 with a potential function @xmath41 of the same kind as @xmath14 but with different parameters , as specified in table [ table1 ] . these differential operators are natural generalisations of the schrdinger operator in that they also have polynomial eigenfunctions with corresponding eigenvalues that can be computed explicitly ; see section [ sec4 ] for a precise formulation of this result . in this paper we present an explicit construction of all these many - variable generalisations of the above mentioned classical polynomials . we obtain the results by a unified treatment making no reference to special cases , as explained below . & @xmath42 & @xmath43 & @xmath44 + & & & + i & @xmath45 & @xmath46 & @xmath47 + ii & 0 & @xmath48 & @xmath49 + & & & + iii & @xmath50 & @xmath51 & @xmath52 + & & & + iv & @xmath53 & @xmath54 & @xmath55 + & @xmath56 & & @xmath57 + & & & + v & @xmath58 & @xmath59 & @xmath60 + & & & + vi & @xmath61 & @xmath59 & @xmath62 + & @xmath63 & & @xmath64 + & & & + vii & @xmath65 & @xmath66 & @xmath67 + & @xmath68 & & @xmath69 + to explain the nature of our results we now discuss the standard calogero - sutherland cases @xmath70 in more detail . our approach is based on a set of remarkable identities which provide us with particular sets of symmetric polynomials on which the action of the differential operators is particularly simple . these latter polynomials can be defined by the following expansion : @xmath71 valid for @xmath72 . we mention that these polynomials were first defined in @xcite by a certain contour integral . however , using cauchy s theorem it is easy to verify that the two definitions are equivalent . we explicitly construct series representations for the polynomials @xmath29 in which are of the form @xmath73 and , in addition , we obtain certain completeness results ; see section 3 for the precise statements . we will also derive a partial generalisation of these results to the differential operators @xmath74 ; see section 4 . we thus generalise , unify , and extend various results which were known before only in special cases @xcite . we now explain in which sense our approach is unified and give a precise definition of the classical polynomials we consider : each such sequence consists of polynomials @xmath31 of order @xmath75 such that they are common eigenfunctions of a differential operator @xmath76 with @xmath77 for some real coefficients @xmath78 and @xmath79 ; see table [ table2 ] for the polynomials corresponding to our examples in table [ table1 ] . it is interesting to note that the schrdinger operator @xmath80 in with the eigenfunctions can be fully characterised by these polynomials @xmath81 and @xmath82 : it is straightforward to verify that @xmath83 ( the prime here and in the following indicates differentiation ) with @xmath18 a solution of the differential equation @xmath84 and @xmath85 with @xmath86 any solution of the differential equation @xmath87 moreover , the potential functions @xmath23 , ensuring that the differential operators @xmath24 in and @xmath74 in have polynomial eigenfunctions , are given by @xmath88 the constant @xmath89 is subtracted to simplify formulae later on , and @xmath41 is given by a formula as in but with @xmath90 replaced by @xmath91/\kappa.\ ] ] it is interesting to note that the exact eigenvalues of the differential operators @xmath24 in and @xmath74 in are determined by the leading coefficients of these polynomials @xmath92 and @xmath90 . for example , the exact eigenvalues of @xmath24 corresponding to the eigenfunctions are @xmath93 with @xmath94 this result , well - known in many special cases ( see , e.g. , @xcite ) , will naturally emerge from our construction in section [ sec3 ] . we are now in a position to state more precisely in which sense our approach is unified : in our construction information concerning the different cases will enter only via the two polynomials @xmath81 and @xmath82 . we also stress that the special cases listed in table [ table2 ] are only intended as examples , and that our approach is for arbitrary polynomials @xmath81 and @xmath82 of the form . & @xmath17 & @xmath18 & @xmath95 & @xmath92 & @xmath90 + & & & & & + i & @xmath96 & @xmath19 & @xmath97 & @xmath98 & @xmath99 + & & & ( hermite ) & & + ii & @xmath98 & @xmath100 & @xmath101 & @xmath102 & @xmath103 + & & & & & + iii & @xmath104 & @xmath105 & @xmath106 & @xmath107 & @xmath108 + & & & ( gegenbauer ) & & + iv & @xmath109 & @xmath110 & @xmath111 & @xmath112 & @xmath113 + & & & ( laguerre ) & & + v & @xmath114 & @xmath115 & @xmath116 & @xmath117 & @xmath118 + & & & ( gegenbauer ) & & + vi & @xmath119 & @xmath115 & @xmath120 & @xmath117 & @xmath121 + & & & ( jacobi ) & & @xmath122 + vii & @xmath123 & @xmath124 & @xmath125 & @xmath126 & @xmath127 + & & & ( gen . bessel ) & & + to put this general scheme into perspective we now discuss its well - known special cases summarised in tables [ table1 ] and [ table2 ] . the case @xmath128 and @xmath129 corresponds to the exactly solvable many - body generalisation of the quantum harmonic oscillator introduced by calogero @xcite ( case i in tables [ table1 ] and [ table2 ] ) , while the sutherland model @xcite , generalising the quantum model of a free particle on a circle , is obtained by setting @xmath130 and @xmath131 ( case ii ) . these models correspond to the hermite polynomials and the ordinary monomials @xmath132 , respectively . the many - body models associated with the legendre , gegenbauer and jacobi polynomials ( cases iv , v and vi ) are identical with olshanetsky and perelomov s @xmath133- and @xmath134-variants of the calogero- and sutherland models ; see @xcite and references therein . an exact solution to the many - body generalisation of the morse potential ( case vii ) , corresponding to the generalised bessel polynomials @xcite , was first given by inozemtsev and meshcheryakov @xcite . a further study of this case can be found in @xcite . there exist various other interesting special cases which we have not explored in detail , but they can all be transformed to the cases mentioned above by a rescaling and a translation of the variable @xmath135 . the deformations of the calogero - sutherland models defined by @xmath74 in were found and explored by chalykh , feigin , sergeev , and veselov ; see , e.g. , @xcite . we will refer to these differential operators @xmath74 as deformed calogero - sutherland operators . we now discuss the relation of our results to previous work in the literature . the many - variable polynomials corresponding to the eigenfunctions of the calogero - sutherland type models in cases i vi in table [ table1 ] have been extensively studied in the mathematics literature ; see , e.g. , @xcite and references therein . we mention , in particular , heckman and opdam s root system generalisation of the jacobi polynomials @xcite and the work of baker and forrester @xcite , van diejen @xcite , and lassalle @xcite , as well as of macdonald @xcite , on many - variable generalisations of the classical orthogonal hermite , laguerre and jacobi polynomials . a particularly well - studied case are the so - called jack polynomials corresponding to the sutherland model ( case ii ) . explicit formulae for the jack polynomials were recently obtained by lassalle and schlosser @xcite ( see also @xcite ) by inverting a so - called pieri formula . for very particular values of the integer vector @xmath26 or a low number of variables explicit series expansions of the jack polynomials were obtained by stanley @xcite . in addition , there exist expansions of the jack , as well as certain other many - variable classical orthogonal polynomials , which are of a combinatorial nature @xcite . these combinatorial results also include a formula resulting from the use of sutherland s original solution algorithm @xcite . we also mention operator solutions of the calogero- and sutherland models obtained in @xcite and @xcite , respectively , as well as integral representations of the jack polynomials ; see @xcite and references therein . our results in this paper provide an explicit construction of the many - variable polynomials @xmath29 . these results were recently announced in @xcite , and they generalise those previously obtained by us for the sutherland- @xcite and the calogero models @xcite . we mention that our unified treatment of all the cases in table [ table1 ] is different from the one based on root systems ( see , e.g. , @xcite ) , and that it has been previously used by gomez - ullate et.al . @xcite to obtain the energy eigenvalues in and by purely algebraic means . we also mention that our point of view is closely related to the theory of quasi - exactly solvable schrdinger operators ; see , e.g. , @xcite . moreover , the identities which are the key to our results are stated in corollaries [ cor2 ] and [ cor4 ] below , and they were known before only in special cases : an important special case of these identities for the sutherland model ( case ii ) is a consequence of a well - known result on jack polynomials which , to our knowledge , is due to stanley ( see proposition 2.1 in @xcite ) , and a generalisation of the latter to the deformed case and other non - deformed cases can be found in @xcite and @xcite , respectively . these identities relate schrdinger- or deformed calogero - sutherland operators with different parameter values and are natural quantum analogs of the bcklund transformations for the classical calogero - moser models first found by woijekowski @xcite , as discussed by kutznetsov and sklyanin @xcite . we obtain these identities as specialisations of a particular identity stated in proposition [ prop1 ] which has the natural physical interpretation of giving the groundstate of a generalisation of the schrdinger operator where the particles are allowed to have different masses @xmath136 and with specific mass dependent external potentials @xmath137 . this is a powerful result of independent interest , and to our knowledge it was previously known only in the special cases i , ii and iv in table [ table1 ] @xcite . we finally mention an integral representation of the jack polynomials recently obtained by kuznetsov et.al . @xcite using a separation - of - variables approach which is also based on the identity in corollary [ cor2 ] . in section [ sec2 ] we derive and discuss the identities which are the key to our solution method . section [ sec3 ] contains our explicit construction of series representation of all many - variable polynomials @xmath29 determining the eigenfunctions of the schrdinger operators . in section [ sec4 ] we extend our construction of polynomial eigenfunctions to the deformed calogero - sutherland operators . we conclude with a few remarks in section [ sec5 ] . we denote by @xmath138 , @xmath139 , @xmath140 and @xmath141 the sets of all integers , positive integers , non - negative integers and real numbers , respectively . we shall say that an integer vector @xmath142 is of length @xmath21 , denoted @xmath143 , if @xmath21 is the smallest non - negative integer such that the parts @xmath144 of @xmath145 are zero for @xmath146 . the symbols @xmath145 and @xmath147 will in most instances denote integer vectors , except in section [ sec2 ] where the symbols @xmath136 are used for real ` mass parameters ' rather than for parts of an integer vector @xmath147 . we will use the symbols @xmath148 , @xmath149 and @xmath150 to emphasise that a particular integer vector is a partition . we will also use the notation @xmath151 where @xmath152 and @xmath153 . moreover , for @xmath154 and @xmath155 we will write @xmath156 short for @xmath157 . in this section we present and prove a particular identity associated with the one - particle schrdinger operator . the identities underlying our construction of eigenfunctions of the differential operators in and are stated in section [ sec22 ] and obtained as special cases of this more general results . throughout the section , we assume the polynomials @xmath81 and @xmath82 in fixed , and @xmath18 and @xmath158 are as in and . the following identity can be interpreted as providing the exact groundstate of a generalisation of the schrdinger operator where the particles can have different masses @xmath136 . however , we will allow these parameters @xmath136 also to be negative , and we will use this identity as a source from which we obtain various other identities as special cases . to state this result we find it convenient to denote this generalised schrdinger operators as @xmath159 , the particle coordinates as @xmath160 , and the particle number as @xmath161 . [ prop1 ] let @xmath162 with @xmath136 arbitrary real and non - zero parameters and @xmath163 with @xmath164 where the prime indicates differentiation with respect to the argument @xmath135 and @xmath165 furthermore , let @xmath166 with @xmath167 and @xmath168 with @xmath169 any solution of the differential equation @xmath170 then @xmath171 with the constant @xmath172 where @xmath173 the @xmath4-dependence of the external potentials and one - particle groundstate eigenfunctions for our examples is given in table [ table3 ] with @xmath4-dependent parameters defined in the table captions . [ betaremark ] note that @xmath174 and @xmath175 always refers to the coefficients of the polynomial @xmath90 as defined in , and are not to be confused with @xmath176 for @xmath177 and @xmath178 . there should be no danger of confusion since we always write @xmath90 for @xmath176 if @xmath177 and assume @xmath179 . & @xmath180 & @xmath181 + & & + i & @xmath182 & @xmath183 + ii & @xmath184 & @xmath185 + iii & @xmath186 & @xmath187 + iv & @xmath188 & @xmath189 + v & @xmath190 & @xmath191 + vi & @xmath192 & @xmath193 + vii & @xmath194 & @xmath195 + we will show by straightforward computations that @xmath196 where @xmath197 since @xmath198 for all @xmath199 , this will prove the identity in . to prove we compute @xmath200 and thus @xmath201 with @xmath202 the one - body terms , @xmath203 the two - body terms , and @xmath204 the three - body terms ; we inserted the relations @xmath205 and @xmath206 and used the short hand notation @xmath207 . symmetrising the sum defining @xmath208 and inserting @xmath209 we can write @xmath210 and observing that the fraction on the r.h.s . is identical with @xmath211 and @xmath98 for @xmath212 and @xmath213 , respectively , we find that @xmath208 is equal to the constant @xmath214 symmetrising the sum defining @xmath215 and inserting @xmath216 , following from the second equation in , we obtain @xmath217 we now decompose @xmath218 into two parts as follows : @xmath219 where @xmath220 and @xmath221 recalling we find that the symmetric part gives us the two - body terms we want , up to a constant , @xmath222 inserting and a straightforward but somewhat tedious computation shows that the terms in the antisymmetric term add up to a constant , @xmath223 inserting the second equation in and a simple computation shows that the one - body terms are identical with @xmath224 with @xmath225 defined in . collecting all terms we obtain the identity in with the constant @xmath226 using the notation in we find by straightforward computations that this constant is identical with the one given in . the physical interpretation of @xmath159 in as schrdinger operator of a quantum many - body system requires that it defines a self - adjoint hilbert space operator bounded from below . this is the case under certain obvious restrictions on parameters . we now discuss this hilbert space structure for the different cases listed in table [ table1 ] , but our discussion will be brief since this aspect does not play any role for our construction in the following sections . in cases i , iii and vii the relevant hilbert space is @xmath227 , in case iv we have instead @xmath228 , while @xmath229 is associated to case ii , and @xmath230 to cases v and vi ( the weight function is in all these cases constant and equal to 1 ) . it is obvious in all these cases that @xmath231 in the proof above is the hilbert space adjoint of @xmath232 ( on suitable domains ) , and thus , if all @xmath136 are positive , that @xmath159 in ( [ i d ] ) defines a unique self - adjoint operator via the friedrichs extension ( see , e.g. , @xcite ) with @xmath233 as groundstate provided that @xmath233 is square integrable . in particular , the @xmath24 in ( [ hn ] ) define , under obvious restrictions on parameters , self - adjoint operators bounded from below . however , this is not the case for the deformed operators ( [ d ] ) . we proceed to discuss the special cases of proposition [ prop1 ] which underlies our construction of eigenfunctions of the differential operators and in sections [ sec3 ] and [ sec4 ] , respectively . setting @xmath234 , @xmath235 and @xmath236 for @xmath237 we obtain as a first special case the following : the function @xmath238 , as defined by and , is an eigenfunction of the schrdinger operator with corresponding eigenvalue @xmath239 . the remarkable identity underlying our solution method in section [ sec3 ] is obtained from proposition [ prop1 ] by choosing @xmath240 , @xmath241 , @xmath242 , @xmath243 , and @xmath244 for @xmath245 . we observe that @xmath159 in then becomes a difference of two schrdinger operators . denoting @xmath233 as @xmath246 and @xmath247 as @xmath248 we obtain the following : [ cor2 ] let @xmath249 be the operator in , @xmath250 and @xmath251 with @xmath252 , @xmath253 , and @xmath254 and @xmath255 as in and for @xmath256 . then @xmath257 with the constant @xmath258 it is important to note that also @xmath259 is a schrdinger operator of the same kind as @xmath24 , only the coupling parameters are different . for the convenience of the reader we give the modified parameters for our examples i and iii vii , @xmath260 the key property that makes the special case of proposition [ prop1 ] stated in corollary [ cor2 ] interesting as a tool for constructing the eigenfunctions of a schrdinger operator is that the variables @xmath261 of @xmath159 in are divided in two groups @xmath262 and @xmath263 such that the interaction terms involving variables belonging to different groups all vanish . it is interesting to note that there are other cases where this happens . for example , the number of particles with parameters @xmath241 and @xmath264 need not be the same . more generally , if we divide the variables @xmath261 into two groups where the ` mass parameters ' in the first group are @xmath241 or @xmath265 and in the second group @xmath266 or @xmath267 , then all interaction terms between particles in different groups vanish ( since @xmath268 in all these cases ) . we then obtain an identity as in but now involving two deformed calogero - sutherland operators which can have , in general , different particle numbers . such an identity involves four different kinds of particles , and the number of particles of each kind can be arbitrary . these identities will allow us to construct eigenfunctions of the deformed calogero - sutherland operators in section [ sec4 ] . the most general such identity corresponds to the following special case of proposition [ prop1 ] : @xmath269 with non - negative integers @xmath270 such that @xmath161 is non - zero , @xmath241 and @xmath242 for @xmath271 , @xmath272 and @xmath273 for @xmath274 , @xmath275 and @xmath276 for @xmath277 , and @xmath278 and @xmath279 for @xmath280 . [ cor4 ] let @xmath281 be the operator in with @xmath282 , @xmath283 where @xmath3 and @xmath284 are arbitrary non - negative integers , and @xmath285 with @xmath252 , @xmath253 , @xmath286 , @xmath287 , and @xmath288 and @xmath181 as in and , respectively , for @xmath256 and @xmath289 . then @xmath290 with the constant @xmath291 where @xmath292 using this identity we can straightforwardly generalise our solution method and obtain many different formulae for eigenfunctions of the deformed calogero - sutherland operators : for fixed particle numbers @xmath21 and @xmath39 one is free to choose @xmath3 and @xmath284 arbitrarily and each choice gives a family of eigenfunctions labelled by integer vectors @xmath293 . as discussed in section [ sec42 ] , we obtain in this way a number of series representations for each eigenfunction , a fact which is interesting already in the standard ( non - deformed ) case @xmath294 . in order to put the general discussion in the previous two sections into perspective we consider in some detail the special case ii ; see tables [ table1 ] and [ table2 ] . the schrdinger operator is in this case given by @xmath295 as we mentioned in the introduction , this schrdinger operator was introduced and studied by sutherland @xcite , and its eigenfunctions are given by the so - called jack polynomials . in this special case the identity in corollary [ cor2 ] is equivalent to a well - known and important identity in the theory of jack polynomials , and corollary [ cor4 ] gives a natural generalisation of this identity to the deformed case . more precisely , we have the following : let @xmath296 with @xmath297 and @xmath298 . then @xmath299 where the arguments indicate that the schrdinger operator acts in the variables @xmath262 and @xmath263 , respectively . as mentioned above , the identity , and in particular the function @xmath246 , is directly related with the theory of jack polynomials . in order to make this precise we observe that @xmath300 with the function @xmath301 given by @xmath302 it is well known that this latter function @xmath301 has the following expansion in the monic jack polynomials @xmath303 : @xmath304 for some ( explicitly known ) coefficients @xmath305 , and where @xmath306 ; see , e.g. , sections vi.4 and vi.10 in @xcite . we recall that the monic jack polynomials are usually denoted @xmath307 rather than @xmath303 . we have used the latter notation in order to avoid clashes with notation used in later sections . up to degeneracies in the spectrum of the schrdinger operator @xmath24 , this expansion can be deduced as a consequence of the identity . to establish this fact we let @xmath308 , and similarly for @xmath309 . it is well known that the functions @xmath310 are eigenfunctions of @xmath24 ; this can for example be inferred from results of stanley @xcite . since @xmath311 is invariant under the substitution of @xmath312 for @xmath313 which maps @xmath314 to @xmath315 , the functions @xmath316 are eigenfunctions of the same schrdinger operator @xmath24 , and with the same eigenvalues . we observe that @xmath317 is invariant under permutations of both the variables @xmath318 as well as the variables @xmath319 . it follows that it has an expansion in jack polynomials of the form @xmath320 where the functions @xmath321 are to be determined . since @xmath316 is an eigenfunction of @xmath24 , the identity implies that also @xmath322 is an eigenfunction of @xmath24 , and with the same eigenvalue . it follows that , unless this eigenvalue is degenerate , @xmath323 is proportional to @xmath324 . we mention that although such degeneracies do occur they are quite rare in this case . the identities obtained in the previous section thus suggest that there exist expansions , similar to the one given above , for each function @xmath246 corresponding to a schrdinger operator or a deformed counterpart . however , in most cases the schrdinger operator @xmath24 , as well as its deformed counterpart @xmath74 , is not invariant under a change of coordinates mapping @xmath314 to @xmath315 . this explains the appearance of the operator @xmath325 and the constant @xmath326 in corollary [ cor2 ] , as well as that of @xmath327 and @xmath328 in corollary [ cor4 ] . in this section we construct our series representations of the eigenfunctions of the schrdinger operators . these series representations are in terms of the functions @xmath329 defined in . as we will see , the identities deduced in the previous section play a central role in this construction . these series representation generalises those obtained in @xcite for the special case ii and in @xcite for i and iv ( c.f . table [ table1 ] ) . we fix the polynomials @xmath81 , @xmath82 in and the coupling constant @xmath32 and consider the resulting schrdinger operator @xmath24 as defined by , , and . we observe that the eigenfunctions in are completely determined by the functions @xmath29 in @xmath21 variables @xmath318 , and we will refer to the latter as _ reduced eigenfunctions _ of @xmath24 . we also observe that @xmath29 is a reduced eigenfunction of @xmath24 if and only if it is an eigenfunction of the differential operator @xmath330 where the last equality follows by a straightforward computation using the fact that @xmath238 is an eigenstate of @xmath24 with eigenvalue @xmath239 and the identities in , , and . using the identity in corollary [ cor2 ] it is now straightforward to compute the action of this differential operator on the polynomials @xmath329 . in order to state this action in a simple form , and also to facilitate a discussion of its implications , we first introduce some convenient notation . an important ingredient in our construction is the following partial ordering of integer vectors @xmath331 : @xmath332 that this is only a partial ordering is easily seen , e.g. , @xmath333 and @xmath334 are incomparable . the algebra of symmetric polynomials @xmath335 in @xmath21 variables @xmath336 with complex coefficients is denoted @xmath337 . the linear subspace consisting of the homogeneous symmetric polynomials of degree @xmath338 , together with the zero polynomial , is denoted by @xmath339 . we also denote as @xmath340 the standard basis in @xmath341 , i.e. , @xmath342 for all @xmath343 , and let @xmath344 as before , @xmath345 for simplicity of notation we shall to each @xmath346 associate the shifted integer vector @xmath347 with @xmath348 we are now ready to state and prove the following : [ actionlemma ] for each @xmath155 , @xmath349 note that the function @xmath246 in corollary [ cor2 ] can be written as follows : @xmath350 with @xmath351 the definition of the polynomials @xmath329 thus implies @xmath352 it follows from corollary [ cor2 ] that @xmath353 where @xmath354 using , , and , as well as the definition of @xmath259 , we find that @xmath355 with @xmath356 recalling we expand the interaction terms in laurent series and obtain @xmath357 valid for @xmath72 . it is now straightforward to compute the r.h.s . of , and by comparing the coefficients of @xmath358 on both sides of the resulting equation we obtain with @xmath359 + c_n.\ ] ] by straightforward computations we find that the latter coincides with @xmath360 in and . it is clear that the action of @xmath361 on the polynomials @xmath329 is triangular in the sense that @xmath362 is a linear combination of polynomials @xmath363 with @xmath364 . in fact , depending on the specific choice of @xmath81 and @xmath82 , the set of possible @xmath147 is in many cases significantly smaller . in any case , this triangular structure suggests that there exist eigenfunctions , with corresponding eigenvalues @xmath365 , of the differential operator @xmath361 which are of the form @xmath366 where the sum is over integer vectors @xmath367 . indeed , if we take this as an ansatz then we find that this is the case if and only if the coefficients satisfy the recursion relation @xmath368 since @xmath369 is fixed to one , it is clear that if @xmath370 for all @xmath147 which appear in , then this recursion relation uniquely determines the remaining coefficients @xmath371 . this condition of non - degeneracy on the eigenvalues is generically satisfied , but there exist special cases and specific integer vectors @xmath145 for which it fails to hold true ; see remark [ rmk32 ] for a further discussion of this point . in order to draw attention to some of the specific features of the action of @xmath361 on the polynomials @xmath329 we consider in some detail the special case ii ; see tables [ table1 ] and [ table2 ] . on the one hand this is in many ways the simplest non trivial case , and our construction of reduced eigenfunctions is therefore particularly simple in this case , but , on the other hand , it already contains many of the key features of the general case . for the special case ii we have @xmath130 and @xmath131 , and consequently that @xmath372 clearly , the right hand side now only contains polynomials @xmath363 such that @xmath373 . it is interesting to compare this expression with the corresponding action of @xmath361 on the monomial symmetric polynomials @xmath374 with @xmath148 a partition of length @xmath375 , and where the sum extends over all distinct permutations @xmath376 of the parts @xmath35 of @xmath148 . to this end we let @xmath377 denote the integer part of @xmath378 . in addition , for any @xmath379 we identify @xmath380 with the unique @xmath381 such that @xmath382 for some permutation @xmath376 . using the identity @xmath383 valid for any non - negative integers @xmath338 and @xmath4 such that @xmath384 , it is now easy to verify that @xmath385 we note , in particular , that the action of @xmath361 on the polynomials @xmath329 does not contain an explicit dependence on the integer vector @xmath145 . in contrast , the action on the symmetric monomials @xmath381 do contain an explicit dependence on the partition @xmath148 . it is also interesting to note that the corresponding recursion relation can in fact be solved to yield the following explicit series representation for the eigenfunctions of @xmath361 : @xmath386 see @xcite for further details . it is important to note that the fact that @xmath329 is non - zero only if @xmath387 ( see corollary [ cor31 ] ) implies that this series only contains a finite number of terms , and thus is well - defined . even though it can be done in principle , it seems harder to solve the recursion relations resulting from the use of the monomial symmetric polynomials . detailed discussions of the above construction of series representations for the reduced eigenfunctions of the schrdinger operator in the cases i and iv can be found in @xcite . we also mention that it is possible to write down explicit formulae for the reduced eigenfunctions of the schrdinger operator in the general case which , however , are somewhat involved . in section [ sec31 ] we constructed , under a certain condition of non - degeneracy on the corresponding eigenvalues , for each integer vector @xmath155 a reduced eigenfuntion @xmath29 of the schrdinger operator @xmath24 . in this section we show that if we restrict attention to integer vectors @xmath388 for some partition @xmath148 , then the corresponding reduced eigenfunctions form a linear basis for @xmath337 . the idea of the proof is to first establish that the expansion of the symmetric polynomials @xmath329 in terms of a particular set of symmetric polynomials @xmath389 , known to constitute such a basis , has a particular triangular structure . this implies that the @xmath390 , labelled by the partitions @xmath148 of length @xmath375 , form a linear basis for @xmath337 . we then obtain a similar result for the reduced eigenfunctions @xmath307 by showing that they are , in the same sense , triangular in the @xmath390 . we proceed to recall that the polynomials @xmath389 in question , sometimes referred to as the modified complete symmetric polynomials , can be defined by the expansion ( @xmath391 ) @xmath392 valid for @xmath393 , and where the sum extends over all partitions @xmath148 of length @xmath375 . it is well known that the @xmath389 constitute a linear basis for @xmath337 ; see , e.g. , section vi.10 in @xcite . using this fact we now establish the following relation between the @xmath329 and the @xmath389 , which , in particular , implies that the @xmath329 indeed are well - defined symmetric polynomials : [ ftrianglemma ] let @xmath155 . then @xmath394 for some coefficients @xmath395 and where the sum is over partitions @xmath396 such that @xmath397 . moreover , if @xmath388 for some partition @xmath148 of length @xmath375 , then @xmath398 . it follows from the definitions of the @xmath389 and @xmath329 that @xmath399 where @xmath400 denotes the unique partition obtained by permuting the components of the integer vector @xmath147 , e.g. , @xmath401 . comparing the coefficients of @xmath402 on both sides of this identity we find that @xmath403 we thus obtained a representation of @xmath329 as a linear superposition of polynomials @xmath404 where @xmath405 for @xmath406 with non - negative integers @xmath407 . obviously , @xmath408 for such @xmath149 . to complete the proof we thus only need to show that @xmath409 . for that we observe that @xmath364 . moreover , since @xmath149 is a partition , @xmath410 , and the definition of the ordering thus implies that @xmath411 for any permutation @xmath376 of the parts of @xmath412 and , in particular , @xmath413 . it is interesting to note that in the particular case @xmath414 the result in is equivalent to the so - called jacobi - trudi identity ( originally due to jacobi @xcite ) , which implies that the well - known schur polynomials @xmath415 are identical with the polynomials @xmath390 for @xmath414 . the proof of this fact can be obtained as a simple consequence of well - known arguments which can be found in macdonald s book @xcite ; see the proof of ( 3.4 ) and the preceding discussion in chapter i. the fact that each @xmath404 is a symmetric homogeneous polynomial of degree @xmath416 now implies the following : [ cor31 ] the function @xmath329 , @xmath155 , is non - zero only if @xmath417 . in that case it is a symmetric homogeneous polynomial of degree @xmath418 . let @xmath338 be a non - negative integer and consider the two sets of polynomials @xmath389 and @xmath390 which are both indexed by those partitions @xmath148 of @xmath338 which are of length @xmath375 . let @xmath419 denote the transition matrix from the @xmath390 to the @xmath389 , defined by the equalities @xmath420 . given a partition @xmath148 , lemma [ ftrianglemma ] implies that @xmath421 unless @xmath422 and that @xmath423 . it is readily verified that the inverse of such a matrix always exist and is of the same form ; see , e.g. , section i.6 in @xcite . it follows that @xmath424 where @xmath425 are the elements in the matrix inverse to @xmath426 and the sum is over partitions @xmath427 of @xmath338 . we thus obtain the following : [ basisprop ] for any non - negative integer @xmath338 , the polynomials @xmath390 , labelled by those partitions @xmath148 of @xmath338 which are of length @xmath375 , form a linear basis for @xmath339 . suppose now that the reduced eigenfunctions @xmath29 of the form exist for all integer vectors @xmath155 such that @xmath388 for some partition @xmath148 . from lemma [ ftrianglemma ] and we have that any such reduced eigenfunction has a series expansion of the form @xmath428 for some coefficients @xmath429 , and where the sum is over partitions @xmath430 . fix a partition @xmath148 of length @xmath375 and consider the transition matrix from the reduced eigenfunctions @xmath431 to the polynomials @xmath432 , both indexed by the partitions @xmath412 of length @xmath433 such that @xmath422 . observe that its entries are non - zero only if @xmath422 and that all diagonal entries are equal to one . it is clear from the arguments leading up to proposition [ basisprop ] that this transition matrix has a well - defined inverse of the same form . it follows that each polynomial @xmath390 is a linear combination of reduced eigenfunctions @xmath431 with @xmath434 . proposition [ basisprop ] thus implies the following : [ prop32 ] if the reduced eigenfunctions @xmath307 in exist for all partitions @xmath148 of length @xmath375 then they form a linear basis for @xmath337 . [ rmk32 ] as discussed after equation , a reduced eigenfunction of the form exist if @xmath370 for all @xmath147 which appear in . it is easy to see that this condition of non - degeneracy holds true if @xmath435 ( and @xmath436 ) , including the special cases i and iv ( see , e.g. , @xcite ) . it is also known to hold for all integer vectors @xmath155 such that @xmath388 for some partition @xmath148 in the special case ii ( see , e.g. , @xcite ) , assuming that @xmath32 . it is interesting to note that if @xmath437 and if one allow @xmath33 to be negative then there exist parameter values and integer vectors @xmath145 such that @xmath438 for some integer vector @xmath147 which appear in . we stress , however , that for generic parameter values and integer vectors @xmath145 this condition of non - degeneracy on the eigenvalues is satisfied . in this section we sketch the generalisation of our construction of eigenfunctions from the previous section to the deformed calogero - sutherland operators @xmath74 , emphasising the features which are new . just as the construction in the previous section was based on corollary [ cor2 ] , the discussion below is based on corollary [ cor4 ] . we recall that the identity in this latter corollary relates the operator @xmath74 to an operator of the same type , but with particle numbers @xmath3 and @xmath284 . we stress that these particle numbers can be chosen freely , and that , as we will discuss in section [ sec42 ] , the complexity of the resulting series representation of a given eigenfunction can be highly dependent on the specific choice . as described already in the beginning of section [ sec1 ] , this can be used to construct simpler explicit formulae for many of the eigenfunctions already in the special case @xmath70 . we proceed to describe the structure of the eigenfunctions of the deformed calogero - sutherland operators . following sergeev and veselov @xcite we let @xmath439 be the algebra of polynomials @xmath440 in two sets of independent variables @xmath336 and @xmath441 with complex coefficients such that they are separately symmetric in the variables @xmath318 and @xmath442 and , furthermore , obey the condition @xmath443 for all @xmath237 and @xmath444 . we will show that the deformed calogero - sutherland operators have eigenfunctions which are naturally labelled by partitions @xmath445 in the so - called fat @xmath446-hook , i.e. , @xmath447 , and are of the form @xmath448 with @xmath449 , @xmath450 , @xmath451 and where the @xmath307 are particular polynomials contained in the algebra @xmath439 . moreover , the corresponding eigenvalues are given by @xmath452 with @xmath453 to the best of our knowledge , this fact was previously known only in special cases ; see @xcite . at this point we fix the polynomials @xmath81 and @xmath82 in and the coupling constant @xmath32 and consider the resulting deformed calogero - sutherland operator @xmath74 . as we will see , the construction of explicit series representations for the polynomials @xmath307 is very similar to the one in the previous section , but there are a few important differences in definitions and notation which we now specify . we start by noting that @xmath454 is an eigenfunction of @xmath74 if and only if @xmath307 is an eigenfunction of the differential operator @xmath455 with corresponding eigenvalue @xmath456 ; this equivalence follows from a straightforward computations using that @xmath238 is an eigenstate of @xmath74 with eigenvalue @xmath239 ( corollary [ cor4 ] for @xmath457 ) and the identities , , and . we shall refer to the polynomials @xmath307 as reduced eigenfunctions of the deformed calogero - sutherland operators . furthermore , we let @xmath458 where @xmath459 and define the polynomials @xmath460 , @xmath461 , through the following expansion : @xmath462 valid in the region @xmath463 on these polynomials @xmath464 . in order to simplify such a computation it is useful to introduce the following ` parity ' function @xmath465 on the index set @xmath466 : @xmath467 we also find it convenient to define the following shifted integer vectors @xmath468 associated with each quantum number @xmath145 : @xmath469 proceeding in analogy with the proof of lemma [ actionlemma ] in the previous section it is now straightforward to obtain the following : [ actionlemmad ] for each @xmath470 , @xmath471 with @xmath472 similarly to the previous section , lemma [ actionlemmad ] implies that that there exist an eigenfunction of the form @xmath473 where the sum is over integer vectors @xmath474 such that @xmath367 , if and only if the coefficients @xmath371 satisfy the recursion relation @xmath475 provided that @xmath476 for all integer vectors @xmath147 which appear in this uniquely determines all coefficients @xmath371 . as previously mentioned , an important feature of our construction is that the non - negative integers @xmath3 and @xmath284 can be chosen freely , and for each choice we obtain a family of reduced eigenfunctions @xmath477 of @xmath74 which are labelled by integer vectors @xmath470 . this yields eigenfunctions of @xmath74 with corresponding eigenvalues @xmath478 given by and . a priori , it is not clear whether these eigenvalues coincide with the ones given in . in particular , it seems that the eigenvalues we obtain depend on the specific values of @xmath3 and @xmath284 , but , of course , the spectrum of @xmath74 does not . this apparent paradox can be resolved by exhibiting an explicit mapping from the eigenvalues @xmath478 to @xmath479 under certain restrictions on the integer vectors @xmath461 . to state this result we need the notion of the conjugate @xmath480 of a partition @xmath412 , obtained by interchanging rows and columns in the young diagram corresponding to @xmath412 , e.g. , @xmath481 see , e.g. , section i.1 in @xcite . [ enlemma ] let @xmath482 with @xmath483 and @xmath412 a partition of length @xmath484 . then @xmath485 with @xmath486 . we recall the following two well - known identities obeyed by all partitions @xmath412 : @xmath487 where the sums are over all non - zero parts @xmath488 and @xmath489 of @xmath412 and @xmath480 , respectively . this , together with the identity @xmath490 inserted into the second sum in , yields @xmath491 the statement now follows by substituting @xmath35 for @xmath136 and @xmath492 for @xmath493 . the construction described in the previous section gives reduced eigenfunctions @xmath494 for all integer vectors @xmath145 such that a certain condition of non - degenracy on the corresponding eigenvalues is satisfied , but unless @xmath145 satisfies the condition in for some partition @xmath148 in the fat @xmath446-hook this function ought to vanish identically . lemma [ enlemma ] gives , for each partition @xmath148 in the fat @xmath446-hook , such an integer vector @xmath145 provided one chooses @xmath495 , and the latter choice thus is natural if one is interested in a general formula which works for ( essentially ) all eigenfunctions . however , we stress that we have constructed a family of reduced eigenfunctions for each choice of non - negative integers @xmath3 and @xmath284 , and that there exist a simple relation between any two such families : for any two sets of non - negative integers @xmath496 and @xmath497 one can show that two reduced eigenfunctions @xmath498 and @xmath499 , labelled by integer vectors @xmath500 and @xmath501 which correspond to one and the same partition @xmath148 under the mapping defined in lemma [ enlemma ] , are equal up to normalisation ; see the first arxiv version ( v1 ) of @xcite . we can thus , in a simple manner , obtain different series representations for one and the same reduced eigenfunction by varying the values of @xmath3 and @xmath284 . since the complexity of this series representation is highly dependent on the values of @xmath3 and @xmath284 , one can choose the latter such that the complexity is minimised , in many cases substantially below that of the canonical choice @xmath502 . this is already evident in the case of the schrdinger operators , i.e. , for @xmath503 . for example , suppose that we are interested in the reduced eigenfunction of the standard sutherland model for @xmath504 and corresponding to the partition with diagram @xmath505 rather than setting @xmath506 and @xmath507 we can use the fact that the conjugate of this partition is given by the diagram @xmath508 and thus the sought after reduced eigenfunction can equally well be obtained by setting @xmath509 and @xmath510 , in the process decreasing the complexity of its series representation from the @xmath511-particle case to that of only @xmath512-particles . as a further illustrative example we consider the reduced eigenfunction corresponding to the partition with diagram @xmath513 for which we can set @xmath514 to minimise the complexity of its series representation . similar examples can be easily constructed also in the ` deformed ' case . in generally , it can be readily verified that , given a specific partition @xmath148 , the minimal value for @xmath515 is attained by setting @xmath516 and @xmath517 where @xmath199 denotes the row in the diagram of @xmath148 for which @xmath518 is minimal . this means , in particular , that we always can choose @xmath3 and @xmath284 such that their sum does not exceed the length @xmath519 of the partition @xmath148 in question , which may be less than @xmath520 . this observation is also reflected in the fact that @xmath521 for all non - negative values of @xmath426 as long as @xmath522 ; the latter is readily inferred from the definition of the polynomials @xmath464 . we conclude this section by verifying that the functions @xmath464 , and consequently also the reduced eigenfunctions @xmath477 , are contained in the algebra @xmath439 for all values of the non - negative integers @xmath3 and @xmath284 . let @xmath470 . then the function @xmath464 is non - zero only if @xmath387 , and in that case it is a homogeneous polynomial of degree @xmath418 in @xmath439 . the proof is by straightforward computations : expanding each factor on the r.h.s . in in the region @xmath523 ) in a binomial series one obtains a well - defined series representation of @xmath524 as a linear superposition of monomials @xmath402 . by inspection one can check that in this series only terms @xmath402 with @xmath387 appear , that each such term appear only a finite number of times , and that the coefficients of @xmath402 is a homogeneous polynomials of degree @xmath418 in the variables @xmath525 and @xmath526 . it thus follows from that @xmath464 is non - zero only if @xmath387 , and in that case it is a homogeneous polynomial of degree @xmath418 . to prove that the @xmath464 are contained in @xmath439 we observe that @xmath524 is separately symmetric in the variables @xmath525 and @xmath526 , and that @xmath527 for all @xmath237 and @xmath444 . the latter follows from by straightforward computations , @xmath528 and @xmath529 in this final section we briefly discuss the relation of our construction to the so called generalised classical orthogonal polynomials of lassalle and macdonald , present some further remarks of the effect of spectral degeneracies on our constructions of reduced eigenfunctions of the schrdinger operators and the deformed calogero - sutherland operators , and comment on the question of completeness in the case of the deformed calogero - sutherland operators . as is well known , the spectra of the schrdinger operators are degenerate . in particular , in cases i and iv the eigenvalue of a given reduced eigenfunction @xmath29 depends only on the weight @xmath418 of the integer vector @xmath145 . an interesting question is thus what the relation is between the basises for the reduced eigenfunctions we construct and previously known such basises . below we will sketch a proof of the fact that our basis coincide , up to normalisation , with the generalised classical orthogonal polynomials of lassalle @xcite and macdonald @xcite . we recall that these latter polynomials are reduced eigenfunctions of the schrdinger operator in case i ( hermite ) , case iv ( laguerre ) or case vi ( jacobi ) , and of the form @xmath530 for some coefficients @xmath531 , and where @xmath532 means that the young diagram of @xmath149 is contained in that of @xmath148 . we shall require the following partial order on the set of partitions : @xmath533 when restricted to the set of partitions of a fixed weight this coincides with the so called dominance order . it is a simple exercise to verify that if @xmath534 and @xmath535 then @xmath536 and vice versa . furthermore , comparing the expansions and , and using the triangular structure of the jack polynomials when expanded in monomial symmetric polynomials , one finds that @xmath537 for some coefficients @xmath538 such that @xmath539 ; see equation 10.16 in section vi.10 of @xcite for the definition of @xmath305 ( in doing so observe that the parameter @xmath540 ) . it is now readily inferred from lemma [ ftrianglemma ] that any reduced eigenfunction @xmath307 , as constructed in section [ sec31 ] , is of the form @xmath541 for some coefficients @xmath542 such that @xmath543 . since @xmath544 if @xmath532 and , up to normalisation , there can be only one eigenfunction of this form , it follows that the relation between the reduced eigenfunctions we construct and those of lassalle and macdonald is given by @xmath545 we stress once more the issue of degeneracies of the eigenvalues of the schrdinger operators and the deformed calogero - sutherland operators : since our construction of reduced eigenfunction of these operators only is valid under a certain non - degeneracy condition on the corresponding eigenvalues , this issue plays a decisive role in our construction . from our point of view , it is essentially only this issue which truly distinguishes the different special cases of our results listed in table [ table1 ] . it thus would be interesting to investigate this issue further . we mention that one can prove an analogue of proposition [ prop32 ] for our construction of reduced eigenfunctions of the deformed calogero - sutherland operators . in fact , proceeding similarly to section [ sec33 ] one can prove that the reduced eigenfunctions constructed in section [ sec41 ] have a triangular expansion in so - called super jack polynomials . we recall that the super jack polynomials are known to constitute a linear basis for the algebra @xmath439 . using these two facts above one can show that if @xmath546 and @xmath547 then the reduced eigenfunctions constructed in section [ sec41 ] are also a linear basis for @xmath439 . if either @xmath548 or @xmath549 then they span only a subspace of @xmath439 which , however , can be given a rather simple characterisation . a detailed discussion of these facts can be found in the first arxiv version ( v1 ) of @xcite . we are grateful to f. calogero , o. chalykh , a. fordy , p.forrester , y. suris , and a. veselov , for useful discussions . m.h.would also like to thank p. forrester for making available to him an unpublished manuscript by macdonald @xcite . we are also grateful to two anonymous referees for constructive criticism which helped us to improve the paper . this work was significantly influenced by discussions with v. kuznetsov . his experience , critical remarks , and interest were very valuable to us and are now sadly missed . this work was supported by the swedish science research council ( vr ) , the gran gustafsson foundation , and the european union through the fp6 marie curie rtn _ enigma _ ( contract number mrtn - ct-2004 - 5652 ) . e. langmann , _ a method to derive explicit formulas for an elliptic generalization of the jack polynomials _ , jack , hall - littlewood and macdonald polynomials ( v.b . kuznetsov and s. sahi , eds . ) , contemporary mathematics , vol . 417 , american mathematical society , 2006 . m. lassalle , _ a short proof of generalized jacobi - trudi expansions for macdonald polynomials _ , jack , hall - littlewood and macdonald polynomials ( v.b . kuznetsov and s. sahi , eds . ) , contemporary mathematics , vol . 417 , american mathematical society , 2006 .

in this article , the 2-iterated-appell family is introduced . certain 2-iterated-appell and mixed type-special polynomials are considered as members of this family . the numbers related to these polynomials are obtained . the determinant definitions for the 2-iterated-appell family and for the 2-iterated and mixed type-appell polynomials are established . the graphs of some 2-iterated-appell and mixed type-appell polynomials are drawn for different values of indices and the roots of these polynomials are also investigated for certain values of index by using matlab . finally , the approximate solutions of the real zeros of these polynomials are given . [ section ] [ thm]corollary [ thm]lemma [ thm]proposition [ thm]example [ thm]definition [ thm]remark + * subuhi khancorresponding author;e - mail : subuhi2006@gmail.com ( subuhi khan ) + second author ; e - mail : mumtazrst@gmail.com ( mumtaz riyasat ) ] and mumtaz riyasat * + department of mathematics + aligarh muslim university + aligarh , india + * _ keywords : _ * -appell polynomials ; 2-iterated-appell polynomials ;-bernoulli polynomials ;-euler polynomials . + * msc 2010 * : 33d45 , 33d99 , 33e20 .

1606.04265

the area of @xmath0-calculus has in the last twenty years served as a bridge between mathematics and physics . recently , there is a significant increase of activities in the area of @xmath0-calculus due to its applications in various fields such as mechanics , mathematics and physics . the definitions and notations of @xmath0-calculus reviewed here are taken from @xcite . + the @xmath0-analogue of the shifted factorial @xmath3 is defined by @xmath4 the @xmath0-exponential function is defined as : @xmath10_{q}!},~0 < |q|<1.\ ] ] the @xmath0-derivative @xmath11 of a function @xmath12 at a point @xmath13 is defined as : @xmath14 the @xmath0-analogue of taylor series expansion of an arbitrary function @xmath15 for @xmath16 is defined as : @xmath17 where @xmath18 is the @xmath19 @xmath0-derivative of the function @xmath12 at point @xmath5 . the class of appell polynomials is characterized completely by appell @xcite in 1880 . in 1954 , sharma and chak @xcite introduced a @xmath0-analogue for the family of appell polynomials and called this sequence of polynomials as @xmath0-harmonic . further , in 1967 al - salaam @xcite introduced the family of @xmath0-appell polynomials @xmath20 and studied some of its properties . the @xmath1-degree polynomials @xmath21 are called @xmath0-appell provided they satisfy the following @xmath0-differential equation : @xmath22_q ~a_{n-1,q}(x),~n=0,1,2,\ldots~;~q \in \mathbb{c};~0<q<1.\ ] ] it is to be noted that @xmath25 is an analytic function at @xmath26 and @xmath27 are the @xmath0-appell numbers . also , there exists a sequence of numbers @xmath28 , such that the sequence @xmath21 satisfies the following relation @xcite : @xmath29 based on appropriate selection for the function @xmath25 , different members belonging to the family of @xmath0-appell polynomials can be obtained . these members are listed in table 1 . + * table 1 . certain members belonging to the @xmath0-appell family * + + + al - salaam showed that the class of all @xmath0-appell polynomials is a maximal commutative subgroup of the group of all polynomial sets , _ i.e. _ the class of all @xmath0-appell sequences is closed under the operation of @xmath0-umbral composition of polynomial sequences . if @xmath30 and @xmath31 are sequences of @xmath0-polynomials , then the @xmath0-umbral composition of @xmath21 with @xmath32 is defined to be the sequence @xmath33 under this operation the set of all @xmath0-appell sequences is an abelian group and it can be seen by considering the fact that every @xmath0-appell sequence is of the form @xmath34_q ! } d_{q}^k\right ) x^n\ ] ] and that the umbral composition of @xmath0-appell sequences corresponds to multiplication of these formal @xmath0-power series in the operator @xmath35 . in 1982 , srivastava @xcite gave several characterizations for the well known appell polynomials and for their basic analogues : the @xmath0-appell polynomials . in 1985 , roman proposed an approach similar to the umbral approach under the area of nonclassical umbral calculus which is called @xmath0-umbral calculus @xcite . the @xmath0-appell polynomials are studied using determinantal and umbral approaches , see for example @xcite . obtaining determinant forms for the @xmath0-appell polynomials and their members is an important aspect of such study . the determinant forms can be helpful for computation purposes and can also be useful in finding the solutions of general linear interpolation problems . + in this article , a family of the 2-iterated @xmath0-appell polynomials is introduced . the determinant forms and other properties of this family and also for certain members belonging to this family are established . in section 2 , the 2-iterated @xmath0-appell polynomials are introduced by means of generating function and series definition . certain members belonging to the 2-iterated @xmath0-appell family and some mixed type @xmath0-special polynomials and related numbers are also considered . in section 3 , the determinant forms of the 2-iterated and mixed type @xmath0-appell polynomials are derived . in section , 4 , the roots of the 2-iterated @xmath0-appell and mixed type @xmath0-appell polynomials are investigated and their graphs are drawn for suitable values of indices . the 2-iterated @xmath0-appell polynomials are introduced by means of generating function and series definition . in order to introduce the 2i@xmath0ap , two different sets of @xmath0-appell polynomials @xmath36 and @xmath37 are considered . thus , from definitions ( 1.9 ) and ( 1.10 ) , it follows that @xmath38_{q}!},~0<q<1,\ ] ] where @xmath39_q!};~a_{n , q}^{i}:=a_{n , q}^{i}(0);~a_{0,q}^{i}=1;~a_q^{i}(t)\neq 0\ ] ] and @xmath40_{q}!},~0<q<1,\ ] ] where @xmath41_q!};~{a}_{n , q}^{ii}:={a}_{n , q}^{ii}(0);~{a}_{0,q}^{ii}=1;~a_q^{ii}(t)\neq 0,\ ] ] respectively . + the generating function for the 2i@xmath0ap is derived by using a different approach based on replacement techniques . for this , the following result is proved : the following generating function for the 2-iterated @xmath0-appell polynomials holds true : @xmath42}(x)\frac{t^n}{[n]_q!},~0<q<1.\ ] ] expanding the @xmath0-exponential function @xmath43 in the l.h.s . of equation ( 2.1 ) and then replacing the powers of @xmath44 , _ i.e. _ @xmath45 by the corresponding polynomials @xmath46 , @xmath47 in l.h.s . and replacing @xmath44 by the polynomial @xmath48 in the r.h.s . of the resultant equation , it follows that @xmath49_q!}+{a}_{2,q}^{ii}(x)\frac{t^2}{[2]_q!}+\ldots+{a}_{n , q}^{ii}(x)\frac{t^n}{[n]_q!}+\ldots\big]= \sum\limits_{n=0}^\infty a_{n , q}^{i}\{a_{1,q}^{ii}(x)\}\frac{t^n}{[n]_q!}.\ ] ] summing up the series in l.h.s . and then using equation ( 2.3 ) and denoting the resultant 2i@xmath0ap in the r.h.s . by @xmath50}(x)$ ] , that is @xmath51}(x)=a_{n , q}^{i}\{a_{1,q}^{ii}(x)\},\ ] ] assertion ( 2.5 ) is proved . + * remark 2.1 . * the generating function ( 2.5 ) for the 2i@xmath0ap @xmath50}(x)$ ] is derived by replacing the powers of @xmath44 by the polynomials @xmath52 in generating function ( 2.1 ) of the @xmath0-appell polynomials @xmath36 . the replacements of the powers of @xmath44 by the polynomials @xmath53 in generating function ( 2.3 ) of the @xmath0-appell polynomials @xmath37 , yields the same generating function . + * remark 2.2 . * it is important to remark that equation ( 2.7 ) is the operational correspondence between the 2i@xmath0ap @xmath50}(x)$ ] and @xmath0-appell polynomials @xmath21 . + the following series definition for the 2-iterated @xmath0-appell polynomials @xmath50}(x)$ ] holds true : @xmath51}(x)=\sum\limits_{k=0}^n { n \brack k}_q a_{k , q}^{i}~{a}_{n - k , q}^{ii}(x).\ ] ] using equations ( 2.2 ) and ( 2.3 ) in the l.h.s . of generating function ( 2.5 ) and then using cauchy - product rule in the l.h.s . of the resultant equation , such that @xmath54_q!}=\sum\limits_{n=0}^\infty a_{n , q}^{[2]}(x)\frac{t^n}{[n]_q!}.\ ] ] equating the coefficients of same powers of @xmath55 in both sides of equation ( 2.9 ) , assertion ( 2.8 ) follows . + * remark 2.3 . * in view of equations ( 1.14 ) and ( 1.15 ) , it is remarked that the 2-iterated @xmath0-appell polynomials @xmath50}(x)$ ] satisfy the following relation : @xmath51}(x)=\left(\sum\limits_{k=0}^\infty \frac{{a}_{k , q}^{i}}{[k]_q!}d_q^k\right)a_{n , q}^{ii}(x).\ ] ] certain members belonging to the @xmath0-appell family are given in table 1 . since , corresponding to each member belonging to the @xmath0-appell family , there exists a new special polynomial belonging to the 2-iterated @xmath0-appell family . thus , by making suitable choice for the functions @xmath56 and @xmath57 in equations ( 2.5 ) and ( 2.8 ) , the generating function and series definition for the corresponding member belonging to the 2-iterated @xmath0-appell family can be obtained . these resultant 2-iterated @xmath0-appell polynomials along with their notations , names , generating functions and series definitions are given in table 2 . + * table 2 . certain members belonging to the 2-iterated @xmath0-appell family * + + + the combinations of any two different members of the @xmath0-appell family in the 2-iterated @xmath0-appell family , yields a new mixed type @xmath0-special polynomial . thus , by making suitable selections for the functions @xmath56 and @xmath57 in equations ( 2.5 ) and ( 2.8 ) , the generating function and series definition of the resultant mixed type @xmath0-special polynomials are obtained . the possible combinations of the @xmath0-bernoulli , @xmath0-euler and @xmath0-genocchi polynomials ( table 1 ( i - iii ) ) are considered . the resultant mixed type @xmath0-special polynomials along with their notations , names , generating functions and series definitions are given in table 3 . + * table 3 . certain mixed type @xmath0-special polynomials * + + + * remark 2.4 . * from remark 2.1 and table 3 , it follows that @xmath58 where @xmath59 , @xmath60 and @xmath61 are the @xmath0-euler - bernoulli polynomials ( @xmath0ebp ) , @xmath0-genocchi - bernoulli polynomials ( @xmath0gbp ) and @xmath0-genocchi - euler polynomials ( @xmath0gep ) . this is due to the fact that the set of all @xmath0-appell sequences is closed under the operation of @xmath0-umbral composition of polynomial sequences and form an abelian group . + next , the numbers related to the 2i@xmath0ap @xmath50}(x)$ ] and also related to certain members of this family given in tables 2 and 3 are explored . as mentioned in definition ( 1.11 ) , the @xmath0-appell numbers are the values of the polynomials @xmath21 at @xmath62 and satisfy equation ( 1.12 ) . + taking @xmath62 in equation ( 2.8 ) , the following series definition for the 2-iterated @xmath0-appell numbers ( 2i@xmath0an ) @xmath50}:=a_{n , q}^{[2]}(0)$ ] is obtained @xmath51}=\sum\limits_{k=0}^n { n \brack k}_q a_{k , q}^{i}a_{n - k , q}^{ii},\ ] ] where @xmath63 and @xmath64 denote two different sets of @xmath0-appell numbers . + next , the numbers related to the 2i@xmath0bp @xmath65}(x)$ ] , 2i@xmath0ep @xmath66}(x)$ ] and 2i@xmath0gp @xmath67}(x)$ ] are explored . taking @xmath62 in series definitions of 2i@xmath0bp @xmath65}(x)$ ] , 2i@xmath0ep @xmath66}(x)$ ] and 2i@xmath0gp @xmath67}(x)$ ] given in table 2 ( i - iii ) and in view of notations given in table 1 , the 2-iterated @xmath0-bernoulli , @xmath0-euler and @xmath0-genocchi numbers are obtained . these numbers are given in table 4 . + * table 4 . 2-iterated @xmath0-numbers * + + + further , the numbers corresponding to the mixed type @xmath0-special polynomials given in table 3 are explored . taking @xmath62 in series definitions of the @xmath0bep @xmath68 , @xmath0bgp @xmath69 and @xmath0egp @xmath70 ( table 3 ( i - iii ) ) and in view of notations given in table 1 , the @xmath0-bernoulli - euler , @xmath0-bernoulli - genocchi and @xmath0-euler - genocchi numbers are obtained . these numbers are given in table 5 . + * table 5 . mixed type @xmath0-numbers * + + + * note . * from remark 2.4 and table 5 , it follows that @xmath71 where @xmath72 , @xmath73 and @xmath74 are the @xmath0-euler - bernoulli numbers ( @xmath0ebn ) , @xmath0-genocchi - bernoulli numbers ( @xmath0gbn ) and @xmath0-genocchi - euler numbers ( @xmath0gen ) , respectively . + in the next section , the determinant forms for the 2-iterated @xmath0-appell polynomials and mixed type @xmath0-special polynomials are established . keleshteri and mahmudov @xcite studied the @xmath0-appell polynomials from determinant point of view . the determinant forms of the special polynomials are important for the computational and applied purposes . this fact provides motivation to establish the determinant definitions of the @xmath0-special polynomials introduced in previous section . in order to define the 2i@xmath0ap @xmath50}(x)$ ] by means of determinant , the following result is proved : the following determinant form for the 2-iterated @xmath0-appell polynomials @xmath50}(x)$ ] of degree @xmath1 holds true : @xmath75}(x)=\frac{1}{\beta_{0,q } } , \\ a_{n , q}^{[2]}(x)=\frac{{(-1)}^n}{{(\beta_{0,q})}^{n+1}}\left|\begin{array}{ccccccc } 1 & { a}_{1,q}^{ii}(x ) & { a}_{2,q}^{ii}(x ) & \cdots & { a}_{n-1,q}^{ii}(x ) & { a}_{n , q}^{ii}(x)\\ \\ { \beta}_{0,q } & { \beta}_{1,q } & { \beta}_{2,q } & \cdots & { \beta}_{n-1,q } & { \beta}_{n , q}\\ \\ 0 & { \beta}_{0,q } & { 2 \brack 1}_q~{\beta}_{1,q } & \cdots & { n-1 \brack 1}_q~ { \beta}_{n-2,q } & { n \brack 1}_q~{\beta}_{n-1,q}\\ \\ 0 & 0 & { \beta}_{0,q } & \cdots & { n-1 \brack 2}_q~{\beta}_{n-3,q } & { n \brack 2}_q~{\beta}_{n-2,q}\\ . & . & . & \cdots & . & . & . & . & \cdots & . & . \\ 0 & 0 & 0 & \cdots & { \beta}_{0,q } & { n \brack n-1}_q~{\beta}_{1,q } \end{array } \right|,\end{array}\ ] ] where @xmath76 and @xmath77 are the @xmath0-appell polynomials of degree @xmath1 ; @xmath78 and @xmath79 let @xmath50}(x)$ ] be a sequence of the 2i@xmath0ap defined by equation ( 2.5 ) and @xmath80 , be two numerical sequences ( the coefficients of @xmath0-taylor s series expansions of functions ) such that @xmath81_{q}!}a_{1,q}^{i}+\frac{t^2}{[2]_{q}!}a_{2,q}^{i}+\cdots+\frac{t^n}{[n]_{q}!}a_{n , q}^{i}+\cdots,~ n=0,1,\ldots;~a_{0,q}^{i}\neq0,\ ] ] @xmath82_{q}!}\beta_{1,q}+\frac{t^2}{[2]_{q}!}\beta_{2,q}+\cdots+\frac{t^n}{[n]_{q}!}\beta_{n , q}+\cdots,~ n=0,1,\ldots;~\beta_{0,q}\neq0,\ ] ] satisfying @xmath83 then , according to the cauchy - product rule , it follows that @xmath84_q!},\nonumber\ ] ] which gives @xmath85 consequently , the following holds : @xmath86 multiplication of equation ( 2.5 ) by @xmath87 gives @xmath88}(x ) \frac{t^n}{n!},\nonumber\ ] ] which in view of equations ( 3.4 ) , ( 3.5 ) and ( 2.3 ) gives @xmath89_{q}!}=\sum\limits_{n=0}^\infty a_{n , q}^{[2]}(x)\frac{t^n}{[n]_{q } ! } \sum\limits_{k=0}^\infty \beta_{k , q}\frac{t^k}{[k]_{q}!}.\ ] ] again , multiplication of the series on the r.h.s . of equation ( 3.8 ) according to cauchy - product rule leads to the following system of infinite equations in the unknowns @xmath50}(x)~(n=0,1,\ldots$ ] ) : @xmath90}(x)\beta_{0,q}=1,\cr \\ a_{0,q}^{[2]}(x)\beta_{1,q}+a_{1,q}^{[2]}(x)\beta_{0,q}={a}_{1,q}^{ii}(x),\cr \\ a_{0,q}^{[2]}(x)\beta_{2,q}+{2 \brack 1}_q a_{1,q}^{[2]}(x)\beta_{1,q}+a_{2,q}^{[2]}(x)\beta_{0,q}={a}_{2,q}^{ii}(x),\cr \vdots\\ a_{0,q}^{[2]}(x)\beta_{n-1,q}+{n-1 \brack 1}_q a_{1,q}^{[2]}(x)\beta_{n-2,q}+\cdots+a_{n , q}^{[2]}(x)\beta_{0,q}={a}_{n-1,q}^{ii}(x),\cr \\ a_{0,q}^{[2]}(x)\beta_{n , q}+{n \brack 1}_q a_{1,q}^{[2]}(x)\beta_{n-1,q}+\cdots+a_{n , q}^{[2]}(x)\beta_{0,q}={a}_{n , q}^{ii}(x),\cr \vdots \end{cases}\ ] ] first equation of system ( 3.9 ) , proves the first part of assertion ( 3.1 ) . also , the special form of system ( 3.9 ) ( lower triangular ) allows to work out the unknowns @xmath50}(x)$ ] . operating with the first @xmath91 equations simply by applying the cramer s rule , it follows that @xmath92}(x)=\frac{\left|\begin{array}{cccccc } \beta_{0,q } & 0 & 0 & \cdots & 0 & 1\\ \\ \beta_{1,q } & \beta_{0,q } & 0 & \cdots & 0 & { a}_{1,q}^{ii}(x)\\ \\ \beta_{2,q } & { 2 \brack 1}_q~\beta_{1,q } & \beta_{0,q } & \cdots & 0 & { a}_{2,q}^{ii}(x)\\ \\ . & . & . & \cdots & . & . \\ \beta_{n-1,q } & { n-1 \brack 1}_q~\beta_{n-2,q } & { n-1 \brack 2}_q~\beta_{n-3,q } & \cdots & \beta_{0,q } & { a}_{n-1,q}^{ii}(x)\\ \\ \beta_{n , q } & { n \brack 1}_q~\beta_{n-1,q } & { n \brack 2}_q~\beta_{n-2,q } & \cdots & { n \brack n-1}_q~\beta_{1,q } & { a}_{n , q}^{ii}(x)\\ \end{array}\right|}{\left|\begin{array}{cccccc } \beta_{0,q } & 0 & 0 & \cdots & 0 & 0\\ \\ \beta_{1,q } & \beta_{0,q } & 0 & \cdots & 0 & 0\\ \\ \beta_{2,q } & { 2 \brack 1}_q~\beta_{1,q } & \beta_{0,q } & \cdots & 0 & 0 \\ . & . & . & \cdots & . & . \\ \beta_{n-1,q } & { n-1 \brack 1}_q~\beta_{n-2,q } & { n-1 \brack 2}_q~\beta_{n-3,q } & \cdots & \beta_{0,q } & 0\\ \\ \beta_{n , q } & { n \brack 1}_q~\beta_{n-1,q } & { n \brack 2}_q~\beta_{n-2,q } & \cdots & { n \brack n-1}_q~\beta_{1,q } & \beta_{0,q}\\ \end{array}\right|}\end{array},\ ] ] where , @xmath93 . + now , taking the transpose of the determinant in the numerator and expanding the determinant in the denominator , so that @xmath92}(x)=\frac{1}{(\beta_{0,q})^{n+1}}\left|\begin{array}{cccccc } \beta_{0,q } & \beta_{1,q } & \beta_{2,q } & \cdots & \beta_{n-1,q } & \beta_{n , q}\\ \\ 0 & \beta_{0,q } & { 2 \brack 1}_q~\beta_{1,q } & \cdots & { n-1 \brack 1}_q~\beta_{n-2,q } & { n \brack 1}_q~\beta_{n-1,q}\\ \\ 0 & 0 & \beta_{0,q } & \cdots & { n-1 \brack 2}_q~\beta_{n-3,q } & { n \brack 2}_q~\beta_{n-2,q}\\ \\ . & . & . & \cdots & . \\ 0 & 0 & 0 & \cdots & \beta_{0,q } & { n \brack n-1}_q~\beta_{1,q}\\ \\ 1 & { a}_{1,q}^{ii}(x ) & { a}_{2,q}^{ii}(x ) & \cdots & { a}_{n-1,q}^{ii}(x ) & { a}_{n , q}^{ii}(x ) \end{array}\right|,\end{array}\ ] ] which after @xmath1 circular row exchanges , that is after moving the @xmath94 row to the @xmath95 position for @xmath96 , yields second part of assertion ( 3.1 ) . + * remark 3.1 . * since the polynomials @xmath32 , @xmath97 and @xmath98 are particular members of the @xmath0-appell family @xmath21 . therefore , the determinant forms of @xmath32 , @xmath97 and @xmath98 can be obtained by giving suitable values to the coefficients @xmath99 and @xmath100 in the definition of @xmath21 . + taking @xmath101 , @xmath102_q}~(i=1,2,\cdots , n)$ ] in the determinant definition of the @xmath0-appell polynomials @xmath21 @xcite ) , the following determinant definition of the @xmath0-bernoulli polynomials @xmath32 is obtained : + * definition 3.1 . * the @xmath0-bernoulli polynomials @xmath32 of degree @xmath1 are defined by @xmath103_q } & \frac{1}{[3]_q } & \cdots & \frac{1}{[n]_q } & \frac{1}{[n+1]_q}\\ \\ 0 & 1 & { 2 \brack 1}_q~\frac{1}{[2]_q } & \cdots & { n-1 \brack 1}_q~\frac{1}{[n-1]_q } & { n \brack 1}_q~\frac{1}{[n]_q}\\ \\ 0 & 0 & 1 & \cdots & { n-1 \brack 2}_q~\frac{1}{[n-2]_q } & { n \brack 2}_q~\frac{1}{[n-1]_q } \\ . & . & . & \cdots & . & . \\ . & . & . & \cdots & . & . \\ 0 & 0 & 0 & \cdots & 1 & { n \brack n-1}_q~\frac{1}{[2]_q } \end{array}\right|,~~n=1,2,\cdots~.\end{array}\ ] ] the particular case of definition ( 3.12 ) , for @xmath104 is considered in @xcite . + next , taking @xmath101 , @xmath105 in the determinant definition of the @xmath0-appell polynomials @xmath21 @xcite ) , the following determinant definition of the @xmath0-euler polynomials @xmath97 is obtained : + * definition 3.2.*the @xmath0-euler polynomials @xmath97 of degree @xmath1 are defined by @xmath106 further , taking @xmath101 , @xmath107_q}~(i=1,2,\cdots , n)$ ] in the determinant definition of the @xmath0-appell polynomials @xmath21 @xcite ) , the following determinant definition of the @xmath0-genocchi polynomials @xmath98 is obtained : + * definition 3.3 . * the @xmath0-genocchi polynomials @xmath98 of degree @xmath1 are defined by @xmath108_q } & \frac{1}{2[3]_q } & \cdots & \frac{1}{2[n]_q } & \frac{1}{2[n+1]_q}\\ \\ 0 & 1 & { 2 \brack 1}_q~\frac{1}{2[2]_q } & \cdots & { n-1 \brack 1}_q~\frac{1}{2[n-1]_q } & { n \brack 1}_q~\frac{1}{2[n]_q}\\ \\ 0 & 0 & 1 & \cdots & { n-1 \brack 2}_q~\frac{1}{2[n-2]_q } & { n \brack 2}_q~\frac{1}{2[n-1]_q } \\ . & . & . & \cdots & . & . \\ . & & . & \cdots & . & . \\ 0 & 0 & 0 & \cdots & 1 & { n \brack n-1}_q~\frac{1}{2[2]_q } \end{array}\right|,~~n=1,2,\cdots~.\end{array}\ ] ] * remark 3.2 . * taking suitable values of the coefficients @xmath99 and @xmath100 in the determinant definition of the 2i@xmath0ap family , the determinant definitions for the 2-iterated @xmath0-members belonging to this family and certain mixed type @xmath0-special polynomials can be obtained . + first , the determinant definitions for the members of the 2i@xmath0ap given in table 2 are obtained . taking @xmath101 , @xmath102_q}~(i=1,2,\cdots , n)$ ] and @xmath109 in equation ( 3.1 ) , the following determinant definition of the 2i@xmath0bp @xmath65}(x)$ ] is obtained : + * definition 3.4 . * the 2-iterated @xmath0-bernoulli polynomials @xmath65}(x)$ ] of degree @xmath1 are defined by @xmath110}(x)=1,\\ b_{n , q}^{[2]}(x)=(-1)^n\left|\begin{array}{ccccccc } 1 & b_{1,q}(x ) & b_{2,q}(x ) & \cdots & b_{n-1,q}(x ) & b_{n , q}(x)\\ \\ 1 & \frac{1}{[2]_q } & \frac{1}{[3]_q } & \cdots & \frac{1}{[n]_q } & \frac{1}{[n+1]_q}\\ \\ 0 & 1 & { 2 \brack 1}_q~\frac{1}{[2]_q } & \cdots & { n-1 \brack 1}_q~\frac{1}{[n-1]_q } & { n \brack 1}_q~\frac{1}{[n]_q}\\ \\ 0 & 0 & 1 & \cdots & { n-1 \brack 2}_q~\frac{1}{[n-2]_q } & { n \brack 2}_q~\frac{1}{[n-1]_q } \\ . & . & \cdots & . & . \\ . & . & . & \cdots & . \\ 0 & 0 & 0 & \cdots & 1 & { n \brack n-1}_q~\frac{1}{[2]_q } \end{array}\right|,~~n=1,2,\cdots~,\end{array}\ ] ] where @xmath111 are the @xmath0-bernoulli polynomials of degree @xmath1 . + taking @xmath101 , @xmath105 and @xmath112 in equation ( 3.1 ) , the following determinant definition of the 2i@xmath0ep @xmath66}(x)$ ] is obtained : + * definition 3.5 . * the 2-iterated @xmath0-euler polynomials @xmath66}(x)$ ] of degree @xmath1 are defined by @xmath113}(x)=1,\\ e_{n , q}^{[2]}(x)={(-1)}^n\left|\begin{array}{ccccccc } 1 & e_{1,q}(x ) & e_{2,q}(x ) & \cdots & e_{n-1,q}(x ) & e_{n , q}(x)\\ \\ 1 & \frac{1}{2 } & \frac{1}{2 } & \cdots & \frac{1}{2 } & \frac{1}{2}\\ \\ 0 & 1 & \frac{1}{2}{2 \brack 1}_q & \cdots & \frac{1}{2}{n-1 \brack 1}_q & \frac{1}{2}{n \brack 1}_q\\ \\ 0 & 0 & 1 & \cdots & \frac{1}{2}{n-1 \brack 2}_q & \frac{1}{2}{n \brack 2}_q \\ . & . & . & \cdots & . & .\\ . & . & . & \cdots & . & . \\ 0 & 0 & 0 & \cdots & 1 & \frac{1}{2}{n \brack n-1}_q\\ \end{array}\right|,~n=1,2,\cdots,\end{array}\ ] ] where @xmath114 are the @xmath0-euler polynomials of degree @xmath1 . + taking @xmath101 , @xmath107_q}~(i=1,2,\cdots , n)$ ] and @xmath115 in equation ( 3.1 ) , the following determinant definition of the 2i@xmath0gp @xmath67}(x)$ ] is obtained : + * definition 3.6 . * the 2-iterated @xmath0-genocchi polynomials @xmath67}(x)$ ] of degree @xmath1 are defined by @xmath116}(x)=1,\\ g_{n , q}^{[2]}(x)=(-1)^n\left|\begin{array}{ccccccc } 1 & g_{1,q}(x ) & g_{2,q}(x ) & \cdots & g_{n-1,q}(x ) & g_{n , q}(x)\\ \\ 1 & \frac{1}{2[2]_q } & \frac{1}{2[3]_q } & \cdots & \frac{1}{2[n]_q } & \frac{1}{2[n+1]_q}\\ \\ 0 & 1 & { 2 \brack 1}_q~\frac{1}{2[2]_q } & \cdots & { n-1 \brack 1}_q~\frac{1}{2[n-1]_q } & { n \brack 1}_q~\frac{1}{2[n]_q}\\ \\ 0 & 0 & 1 & \cdots & { n-1 \brack 2}_q~\frac{1}{2[n-2]_q } & { n \brack 2}_q~\frac{1}{2[n-1]_q } \\ . & . & . & \cdots & . & . \\ . & . & . & \cdots & . & . \\ 0 & 0 & 0 & \cdots & 1 & { n \brack n-1}_q~\frac{1}{2[2]_q } \end{array}\right|,~~n=1,2,\cdots~,\end{array}\ ] ] where @xmath117 are the @xmath0-genocchi polynomials of degree @xmath1 . + next , the determinant definitions for the mixed type @xmath0-special polynomials given in table 3 are obtained . + taking @xmath101 , @xmath105 and @xmath109 in equation ( 3.1 ) , the following determinant definition of the @xmath0bep @xmath68 is obtained : + * definition 3.7 . * the @xmath0-bernoulli - euler polynomials @xmath68 of degree @xmath1 are defined by @xmath118 where @xmath111 are the @xmath0-bernoulli polynomials of degree @xmath1 . + taking @xmath101 , @xmath107_q}~(i=1,2,\cdots , n)$ ] and @xmath109 in equation ( 3.1 ) , the following determinant definition of the @xmath0bgp @xmath69 is obtained : + * definition 3.8 . * the @xmath0-bernoulli - genocchi polynomials @xmath69 of degree @xmath1 are defined by @xmath119_q } & \frac{1}{2[3]_q } & \cdots & \frac{1}{2[n]_q } & \frac{1}{2[n+1]_q}\\ \\ 0 & 1 & { 2 \brack 1}_q~\frac{1}{2[2]_q } & \cdots & { n-1 \brack 1}_q~\frac{1}{2[n-1]_q } & { n \brack 1}_q~\frac{1}{2[n]_q}\\ \\ 0 & 0 & 1 & \cdots & { n-1 \brack 2}_q~\frac{1}{2[n-2]_q } & { n \brack 2}_q~\frac{1}{2[n-1]_q } \\ . & . & . & \cdots & . & . \\ . & . & . & \cdots & . & . \\ 0 & 0 & 0 & \cdots & 1 & { n \brack n-1}_q~\frac{1}{2[2]_q } \end{array}\right|,~n=1,2,\cdots~,\end{array}\ ] ] where @xmath111 are the @xmath0-bernoulli polynomials of degree @xmath1 . + taking @xmath101 , @xmath107_q}~(i=1,2,\cdots , n)$ ] and @xmath112 in equation ( 3.1 ) , the following determinant definition of the @xmath0egp @xmath70 is obtained : + * definition 3.9 . * the @xmath0-euler - genocchi polynomials @xmath70 of degree @xmath1 are defined by @xmath120_q } & \frac{1}{2[3]_q } & \cdots & \frac{1}{2[n]_q } & \frac{1}{2[n+1]_q}\\ \\ 0 & 1 & { 2 \brack 1}_q~\frac{1}{2[2]_q } & \cdots & { n-1\brack 1}_q~\frac{1}{2[n-1]_q } & { n \brack 1}_q~\frac{1}{2[n]_q}\\ \\ 0 & 0 & 1 & \cdots & { n-1 \brack 2}_q~\frac{1}{2[n-2]_q } & { n \brack 2}_q~\frac{1}{2[n-1]_q } \\ . & . & . & \cdots & . & . \\ . & . & . & \cdots & . & . \\ 0 & 0 & 0 & \cdots & 1 & { n \brack n-1}_q~\frac{1}{2[2]_q } \end{array}\right|,~~n=1,2,\cdots~,\end{array}\ ] ] where @xmath114 are the @xmath0-euler polynomials of degree @xmath1 . + * remark 3.3 . * taking @xmath62 in determinant definitions ( 3.12)-(3.14 ) of @xmath32 , @xmath97 and @xmath98 and on expanding the determinants with respect to first row and using suitable notations from table 1 ( i - iii ) , the determinant definitions of the @xmath121 , @xmath122 and @xmath123 can be obtained . + * remark 3.4 . * taking @xmath62 in determinant definitions ( 3.15)-(3.20 ) of @xmath65}(x)$ ] , @xmath66}(x)$ ] and @xmath67}(x)$ ] , @xmath68 , @xmath69 and @xmath70 and then using suitable notations from tables 4 and 5 ( i - iii ) , the determinant definitions of the @xmath65}$ ] , @xmath66}$ ] , @xmath67}$ ] , @xmath68 , @xmath124 and @xmath125 can be obtained . + to give the applications of the operational correspondence introduced in section 2 , the following identities for the @xmath0-appell polynomials are considered @xcite . + @xmath126 the above examples illustrate that the operational correspondence established in this article can be applied to derive the results for the newly introduced @xmath0-special polynomials given in tables 2 and 3 from the results of the corresponding member belonging to the @xmath0-appell family . + in the next section , the shapes of the 2i@xmath0bp @xmath65}(x)$ ] , 2i@xmath0ep @xmath66}(x)$ ] , 2i@xmath0gp @xmath67}(x)$ ] , @xmath0bep @xmath68 , @xmath0bgp @xmath69 and @xmath0egp @xmath70 are displayed . the zeros of these polynomials are also computed by using matlab . there has been increasing interest in solving mathematical problems with the aid of computers . numerical investigation of the roots of certain @xmath0-polynomials are considered in @xcite . also , the shapes of a new class of @xmath0-bernoulli polynomials are explored in @xcite . here , the plots of the 2i@xmath0bp @xmath65}(x)$ ] , 2i@xmath0ep @xmath66}(x)$ ] , 2i@xmath0gp @xmath67}(x)$ ] , @xmath0bep @xmath68 , @xmath0bgp @xmath69 and @xmath0egp @xmath70 are drawn for @xmath132 and @xmath133 @xmath134 . + this shows the four plots combined into one for each of these polynomials . for this , the values of the first five @xmath121 , @xmath122 and @xmath123 are required . the values of first five @xmath121 , @xmath122 and @xmath123 @xcite are listed in table 6 . + * table 6 . values of first five @xmath121 , @xmath122 and @xmath123 * + + + the expressions of the first five @xmath32 , @xmath97 and @xmath98 are obtained by making use of the values of the first five @xmath121 , @xmath122 and @xmath123 in the series definitions given in table 1 ( i - iii ) . the expressions of first five @xmath32 , @xmath97 and @xmath98 are listed in table 7 . + * table 7 . expressions of first five @xmath32 , @xmath97 and @xmath98 * + + + by making appropriate substitutions from tables 6 and 7 in the series definitions given in tables 2 and 3 ( i - iii ) , the expressions of the first five 2i@xmath0bp @xmath65}(x)$ ] , 2i@xmath0ep @xmath66}(x)$ ] , 2i@xmath0gp @xmath67}(x)$ ] , @xmath0bep @xmath68 , @xmath0bgp @xmath69 and @xmath0egp @xmath70 for @xmath133 are obtained . these expressions are given in table 8 . + * table 8 . expressions of first five @xmath135}(x)$ ] , @xmath136}(x)$ ] , @xmath137}(x)$ ] , @xmath138 , @xmath139 and @xmath140 * + + + with the help of matlab and by using the expressions for the first five @xmath135}(x)$ ] , @xmath136}(x)$ ] , @xmath137}(x)$ ] , @xmath138 , @xmath139 and @xmath140 from table 8 for @xmath132 , the following graphs are drawn : + next , we find the real and complex zeros of the polynomials given in table 8 . the computation of these zeros by hand is too complicated , therefore , we use matlab to investigate these zeros . the investigation in this direction will lead to a new approach employing numerical methods in the field of these @xmath0-special polynomials to appear in mathematics and physics . + the real zeros of @xmath135}(x)$ ] , @xmath136}(x)$ ] , @xmath137}(x)$ ] , @xmath138 , @xmath139 and @xmath140 are computed by using matlab . these zeros are given in table 9 . + * table 9 . real zeros of @xmath135}(x)$ ] , @xmath136}(x)$ ] , @xmath137}(x)$ ] , @xmath138 , @xmath139 and @xmath140 * + + + again , with the help of matlab , we find the complex zeros of these polynomials . these complex zeros are given in table 10 . + * table 10 . complex zeros of @xmath135}(x)$ ] , @xmath136}(x)$ ] , @xmath137}(x)$ ] , @xmath138 , @xmath139 and @xmath140 * + + + * remark 4.1 . * from tables 9 and 10 , the following general relation is observed . the number of real zeros lying on the real plane @xmath141 , i.e. , @xmath142}(x)=n-\textrm{complex zeros of } ~a_{n , q}^{[2]}(x),\nonumber\ ] ] where @xmath1 is the degree of polynomial . + in order to make the above discussion more clear , we draw the combined graphs of shapes and zeros of the polynomials @xmath135}(x)$ ] , @xmath136}(x)$ ] , @xmath137}(x)$ ] , @xmath138 , @xmath139 and @xmath140 for @xmath104 . + it is to be noted that in figures 4.13 , 4.14 and 4.16 out of total two complex zeros only one , with positive imaginary part is visible , due to the absence of negative imaginary axis in these graphs . + the numerical results for approximate solutions of real zeros of @xmath135}(x)$ ] , @xmath136}(x)$ ] , @xmath137}(x)$ ] , @xmath138 , @xmath139 and @xmath140 for @xmath143 are displayed in table 9 . + also , we note that the real zeros of these polynomials displayed in table 9 are giving the numerical results for the approximate solutions of @xmath135}(x)=0 $ ] , @xmath136}(x)=0 $ ] , @xmath137}(x)=0 $ ] , @xmath144 , @xmath145 and @xmath146 for @xmath132 . + * note . * it is important to note that the 2-iterated @xmath0-appell polynomials introduced in this article are actually the @xmath0-appell polynomials , since their generating function is of the type @xmath147 , with a suitable choice for @xmath148 . the results established in this article for the 2-iterated and mixed type @xmath0-special polynomials can be used to solve the existing as well as new emerging problems of certain branches of mathematics , physics and engineering . + this work has been done under senior research fellowship ( award letter no . f1 - 17.1/2012 - 13 , manf - mus - utt-9243 ) awarded to the second author by the university grants commission , government of india , new delhi . + 99 w.a . al - salam , @xmath0-appell polynomials , ann . pura appl . * 4*(17 ) ( 1967 ) 31 - 45 . al - salam , @xmath0-bernoulli numbers and polynomials , math . * 17 * ( 1959 ) 239 - 260 . andrews , r. askey , r. roy , 71th special functions of encyclopedia of mathematics and its applications , cambridge university press , cambridge , 1999 . p. appell , sur une classe de polynmes , ann . @xmath149cole . norm . * 9*(2 ) ( 1880 ) 119 - 144 . t. ernst , @xmath0-bernoulli and @xmath0-euler polynomials , an umbral approach , int . j. diff . equ . 0973 - 6069 * 1*(1 ) ( 2006 ) 31 - 80 . keleshteri , n.i . mahmudov , a study on @xmath0-appell polynomials from determinantal point of view , appl . * 260 * ( 2015 ) 351 - 369 . keleshteri , n.i . mahmudov , a @xmath0-umbral approach to @xmath0-appell polynomials , arxiv:1505.05067 . mahmudov , on a class of @xmath0-bernoulli and @xmath0-euler polynomials , adv . difference equ . * 108 * ( 2013 ) 11 . s. roman , the theory of the umbral calculus i , j. math . anal . appl . 87 ( 1982 ) 58 - 115 . s. roman , more on the umbral calculus , with emphasis on the @xmath0-umbral calculus , j. math . appl . * 107 * ( 1985 ) 222 - 254 . ryoo , a note on @xmath0-bernoulli numbers and polynomials , appl . math lett , * 20 * ( 2007 ) 524 - 531 . ryoo , t. kim , r.p . agarwal , a numerical investigation of the roots of @xmath0-polynomials , int . * 83*(2 ) ( 2006 ) 223 - 234 . a. sharma , a.m. chak , the basic analogue of a class of polynomials , riv . parma * 5 * ( 1954 ) 325 - 337 . h. m. srivastava , some characterizations of appell and @xmath0-appell polynomials , ann . mat . pura appl . * 4*(130 ) ( 1982 ) 321 - 329 . w. wang , a determinantal approach to sheffer sequences , linear algebra appl . * 463 * ( 2014 ) 228 - 254 .

two groups are said to have the same nilpotent genus if they have the same nilpotent quotients . we answer four questions of baumslag concerning nilpotent completions . ( i ) there exists a pair of finitely generated , residually torsion - free - nilpotent groups of the same nilpotent genus such that one is finitely presented and the other is not . ( ii ) there exists a pair of finitely presented , residually torsion - free - nilpotent groups of the same nilpotent genus such that one has a solvable conjugacy problem and the other does not . ( iii ) there exists a pair of finitely generated , residually torsion - free - nilpotent groups of the same nilpotent genus such that one has finitely generated second homology and the other does not . ( iv ) a non - trivial normal subgroup of infinite index in a finitely generated parafree group can not be finitely generated . in proving this last result , we establish that the first betti number of a finitely generated parafree group of rank is . it follows that the reduced-algebra of the group is simple if , and that a version of the freiheitssatz holds for parafree groups .

1211.0493

if each finite subset of a group @xmath6 injects into some nilpotent ( or finite ) quotient of @xmath6 , then it is reasonable to expect that one will be able to detect many properties of @xmath6 from the totality of its nilpotent ( or finite ) quotients . attempts to lend precision to this observation , and to test its limitations , have surfaced repeatedly in the study of discrete and profinite groups over the last forty years , and there has been a particular resurgence of interest recently , marked by several notable breakthroughs . we take up this theme here , with a focus on the nilpotent completions of residually torsion - free - nilpotent groups . we begin by recalling some terminology . a group @xmath6 is said to be _ residually nilpotent _ ( resp . _ residually torsion - free - nilpotent _ ) if for each non - trivial @xmath7 there exists a nilpotent group ( resp . torsion - free - nilpotent group ) @xmath8 and a homomorphism @xmath9 with @xmath10 . thus @xmath6 is residually nilpotent if and only if @xmath11 , where @xmath12 , the @xmath13-th term of the _ lower central series _ of @xmath6 , defined inductively by setting @xmath14 and defining @xmath15 : x\in \g_n , y\in \g\>.$ ] we say that two residually nilpotent groups @xmath6 and @xmath16 have the same _ nilpotent genus _ if they have the same lower central series quotients ; i.e. @xmath17 for all @xmath18 . residually nilpotent groups with the same nilpotent genus as a free group are termed _ parafree_. in @xcite gilbert baumslag surveyed the state of the art concerning groups of the same nilpotent genus with particular emphasis on the nature of parafree groups . he concludes by listing a number of open problems that are of particular importance in the field . we will address problems 2 , 4 , and 6 from his list @xcite , two of which he raised again in @xcite where he emphasised the importance of the third of the problems described below . [ baumslag1 ] does there exist a pair of finitely generated , residually torsion - free - nilpotent groups of the same nilpotent genus such that one is finitely presented and the other is not ? [ baumslag2 ] does there exist a pair of finitely presented residually torsion - free - nilpotent groups of the same nilpotent genus such that one has a solvable conjugacy problem and the other does not ? [ baumslag3 ] does there exist a pair of finitely generated , residually torsion - free - nilpotent groups of the same nilpotent genus such that one has finitely generated second homology @xmath0 and the other does not ? [ baumslag4 ] let @xmath19 be a finitely generated parafree group and let @xmath20 be a finitely generated , non - trivial , normal subgroup . must @xmath21 be of finite index in @xmath19 ? in @xcite baumslag gives some evidence to suggest that pairs of groups as in problem [ baumslag2 ] might exist ; he had earlier proved that there exist finitely presented residually torsion - free nilpotent groups with unsolvable conjugacy problem @xcite . baumslag also proves a partial result in connection with problem [ baumslag4 ] ( see theorem 7 of @xcite ) . note that problem [ baumslag4 ] is well known to have a positive answer for free groups . in this paper we answer these questions . concerning the possible divergence in behaviour between groups in the same nilpotent genus , we prove : [ main ] 1 . there exist pairs of finitely generated , residually torsion - free - nilpotent groups @xmath22 of the same nilpotent genus such that @xmath23 is finitely presented and @xmath24 is not . 2 . there exist pairs of finitely presented , residually torsion - free - nilpotent groups @xmath25 of the same nilpotent genus such that @xmath6 has a solvable conjugacy problem and @xmath26 does not . 3 . there exist pairs of finitely generated , residually torsion - free - nilpotent groups @xmath27 of the same nilpotent genus such that @xmath28 is finitely generated but @xmath29 is not . the following result strengthens items ( 1 ) and ( 3 ) of the above theorem . [ t : h2 ] there exist pairs of finitely generated residually torsion - free - nilpotent groups @xmath30 that have the same nilpotent genus and the same profinite completion , but @xmath6 is finitely presented while @xmath31 is infinite dimensional . in the above theorems , the pairs of discrete groups that we construct are such that the inclusion map induces isomorphisms of both the profinite and pro - nilpotent completions . these results emphasise how divergent the behaviour can be within a nilpotent genus . our solution to problem [ baumslag4 ] , in contrast , establishes a commonality among parafree groups , and further commonalities are established in section 8 . [ t : n ] let @xmath19 be a finitely generated parafree group , and let @xmath20 be a non - trivial normal subgroup . if @xmath21 is finitely generated , then @xmath32 is finite . theorem [ t : n ] is proved in section 7 . as with our other results , the proof exploits the theory of profinite groups . it also relies on results concerning the @xmath1-betti numbers of discrete groups . a key observation here is that the first @xmath1-betti number of a finitely presented residually torsion - free - nilpotent group is an invariant of its pro-@xmath33 completion for an arbitrary prime @xmath33 ( see corollary [ c : samel2 ] ) . theorem [ t : n ] will emerge as a special case of a result concerning @xmath1-betti numbers for dense subgroups of free pro-@xmath33 groups ( theorem [ t : general ] ) . in section 8 we discuss some implications of our results . the non - vanishing of the first @xmath1-betti number for parafree groups of rank @xmath5 is shown to have several important consequences , two of which are gathered in the following theorem . we also apply our results to the study of homology boundary links and prove that a non - free parafree group can not be isomorphic to a lattice in a connected lie group . [ t : new ] if @xmath19 is finitely generated and parafree of rank @xmath5 then the reduced @xmath4-algebra of @xmath19 is simple , and every generating set for @xmath19 contains an @xmath2-element subset that generates a free subgroup of rank @xmath2 . the first part of the paper is organised as follows . in section 2 we recall some basic properties about profinite and pro - nilpotent completions of discrete groups and about the correspondence between the subgroup structure of the discrete group and that of its various completions . in section 3 we describe criteria for constructing distinct groups of the same nilpotent genus . in particular we prove that if a map of finitely generated discrete groups @xmath25 induces an isomorphism of profinite completions , then @xmath26 and @xmath6 have the same nilpotent genus . ( if ones assumes merely that @xmath34 , then the genus need not be the same . ) our proof of theorem [ main ] involves the construction of carefully crafted pairs of residually torsion - free - nilpotent groups @xmath35 such that @xmath36 is an isomorphism . theorem [ main](1 ) is proved in section 5 , and theorem [ main](2 ) is proved in section 4 . the proof of theorem [ t : h2 ] is more elaborate : it involves the construction of a finitely presented group with particular properties that may be of independent interest ( proposition [ p : newd ] ) . in relation to theorem a(1 ) , we draw the reader s attention to a recent paper of alexander lubotzky @xcite in which he constructs pairs of groups that have isomorphic profinite completions but have finiteness lengths that can be chosen arbitrarily . these examples are @xmath37-arithmetic groups over function fields ; they do not have infinite proper quotients and therefore are not residually torsion - free nilpotent , but they are residually finite - nilpotent . commenting on an earlier version of this manuscript , chuck miller pointed out that a non - constructive solution to problem 1.2 is implicit in his work with baumslag on the isomorphism problem for residually torsion - free nilpotent groups @xcite ; see remark [ r : chuck1 ] . he also suggested an alternative proof of theorem b ; see remark [ r : chuck2 ] . we thank him for these comments . we also thank an anonymous referee whose comments improved section 3 . * acknowledgement . * the authors thank the mittag leffler institute ( djursholm , sweden ) for its hospitality during the drafting of this paper . let @xmath6 be a finitely generated group . if one orders the normal subgroups of finite index @xmath38 by reverse inclusion , then the quotients @xmath39 form an inverse system whose inverse limit @xmath40 is the _ profinite completion _ of @xmath6 . similarly , the _ pro - nilpotent completion _ , denoted @xmath41 is the inverse limit of the nilpotent quotients of @xmath6 , and the _ pro-(finite nilpotent ) _ completion , denoted @xmath42 , is the inverse limit of the finite nilpotent quotients of @xmath6 . given a prime @xmath33 , the the _ pro-@xmath33 completion _ @xmath43 is the inverse limit of the finite @xmath33-group quotients of @xmath6 . 1 . the inverse limit topology on @xmath44 , on @xmath43 , and on @xmath45 makes them _ compact _ topological groups . the induced topologies on @xmath6 are called the profinite , pro-@xmath33 and _ pro-(finite nilpotent ) _ topologies , respectively . 2 . we do not view @xmath46 as a topological group , and the absence of a useful topology makes it a less interesting object . 3 . by construction , @xmath47 ( resp . @xmath48 ) is residually finite ( resp . nilpotent ) , while @xmath45 is residually finite - nilpotent , and @xmath49 is residually @xmath33 . the natural homomorphism @xmath50 is injective if and only if @xmath6 is residually finite , while the natural maps @xmath51 and @xmath52 are injective if and only if @xmath6 is residually nilpotent . @xmath53 is an isomorphism if and only if @xmath6 is nilpotent . the quotients @xmath54 of @xmath6 by the terms of its lower central series are cofinal in the system of all nilpotent quotients , so one can equally define @xmath41 to be the inverse limit of these . 7 . if @xmath55 then @xmath56 , but the converse is false in general since there exist finitely generated non - isomorphic nilpotent groups that have the same finite quotients @xcite . 8 . less obviously , if @xmath6 and @xmath16 are finitely generated and have the same nilpotent genus then @xmath57 . to see this note that having the same nilpotent genus is equivalent to the statement that @xmath6 and @xmath16 have the same nilpotent quotients , in particular the same finite nilpotent quotients . the conclusion @xmath58 follows from theorem 3.2.7 of @xcite . finite @xmath33-groups are nilpotent , so for every prime @xmath33 there is a natural epimorphism @xmath59 . finitely generated groups of the same nilpotent genus have isomorphic pro-@xmath33 completions for all primes @xmath33 . every homomorphism of discrete groups @xmath60 induces maps @xmath61 and @xmath62 and @xmath63 and @xmath64 . the image of the canonical map @xmath65 is dense regardless of whether @xmath6 is residually finite or not , so the restriction to @xmath6 of any continuous epimomorphism from @xmath66 to a finite group is onto . a deep theorem of nikolov and segal @xcite implies that if @xmath6 is finitely generated then _ every _ homomorphism from @xmath67 to a finite group is continuous . and the universal property of @xmath47 ensures that every homomorphism from @xmath6 to a finite group extends uniquely to @xmath47 . thus we have the following basic result in which @xmath68 denotes the set of homomorphisms from the group @xmath6 to the group @xmath8 , and @xmath69 denotes the set of epimorphisms . if one replaces @xmath47 by @xmath48 or @xmath70 in the following lemma , one obtains bijections for finite nilpotent groups @xmath8 . and for @xmath43 one obtains bijections when @xmath8 is a finite @xmath33-group . [ ns ] let @xmath6 be a finitely generated group and let @xmath71 be the natural map to its profinite completion . then , for every finite group @xmath8 , the map @xmath72 defined by @xmath73 is a bijection , and this restricts to a bijection @xmath74 . closely related to this , we have the following basic but important fact relating the subgroup structures of @xmath6 and @xmath42 and @xmath43 ( see @xcite proposition 3.2.2 , and note that the argument is valid for other profinite completions ) . + * notation . * given a subset @xmath75 of a pro - finite group @xmath19 , we write @xmath76 to denote the closure of @xmath75 in @xmath19 . [ correspondence ] let @xmath77 be the class of finite nilpotent groups or finite @xmath33-groups ( for a fixed prime @xmath33 ) . if @xmath6 is a finitely generated discrete group which is residually @xmath77 , then there is a one - to - one correspondence between the set @xmath78 of subgroups of @xmath6 that are open in the pro-@xmath77 topology on @xmath6 , and the set @xmath79 of all open subgroups in the pro-@xmath77 completion of @xmath6 . identifying @xmath6 with its image in the completion , this correspondence is given by : * for @xmath80 , @xmath81 . * for @xmath82 , @xmath83 . if @xmath84 and @xmath85 then @xmath86 = [ \overline{h}:\overline{k}]$ ] moreover , @xmath87 if and only if @xmath88 , and @xmath89 . let @xmath6 be a finitely generated group that is residually nilpotent , and for each @xmath90 let @xmath91 denote the intersection of all the normal subgroups @xmath92 of index @xmath93 such that @xmath94 is nilpotent . let @xmath95 be the closure of @xmath91 in @xmath96 . then @xmath95 is the intersection of all the normal subgroups of index @xmath93 in @xmath96 , and hence @xmath97 . we begin with a preliminary remark . it is proved in @xcite that for a finitely generated group @xmath6 , _ every _ subgroup of finite index in @xmath45 is open , and likewise in @xmath43 . ( see @xcite for the case of a general profinite group . ) thus , by proposition [ correspondence ] every normal subgroup of index @xmath98 in @xmath70 is the closure of a subgroup @xmath99 such that @xmath100 is nilpotent and @xmath101 , and every subgroup of index @xmath102 in @xmath43 is the closure of a subgroup of index @xmath103 in @xmath6 . in the light of this remark , it suffices to show that if @xmath104 are normal and @xmath105 and @xmath106 are finite and nilpotent , then @xmath107 . but @xmath108 is the kernel of the extension of @xmath109 to @xmath110 , while @xmath111 is the kernel of the map @xmath112 that one gets by extending each of @xmath113 and then taking the direct product . these maps coincide on @xmath6 , which is dense , and are therefore equal . the same argument establishes : [ c : exhaust - p ] let @xmath33 be a prime and let @xmath6 be a finitely generated group that is residually p. for each @xmath90 let @xmath114 denote the intersection of all the normal subgroups @xmath92 of @xmath33-power index less than @xmath98 . let @xmath115 be the closure of @xmath114 in @xmath96 . then @xmath115 is the intersection of all the normal subgroups of index @xmath93 in @xmath96 , and hence @xmath116 . a key feature of the subgroups @xmath91 [ resp . @xmath114 ] is that they form a fundamental system of open neighbourhoods of @xmath118 defining the pro-(finite nilpotent ) [ resp . pro-@xmath33 ] topology . if we had merely taken an exhausting sequence of normal subgroups in @xmath6 , then we would not have been able to conclude that the intersection in @xmath70 [ resp . @xmath120 of their closures was trivial . a key advantage of the class of @xmath33-groups over the class of finite nilpotent groups is that the former is closed under group extensions whereas the latter is not . the importance of this fact from our point of view is that it means that the induced topology on normal subgroups of @xmath33-power index @xmath121 behaves well . the following is a consequence of lemma 3.1.4(a ) of @xcite ( in the case where @xmath77 is the class of finite @xmath33-groups ) . [ l : same ] let @xmath33 be a prime and let @xmath6 be a finitely generated group that is residually-@xmath33 . if @xmath121 is a normal subgroup of @xmath33-power index , then the natural map @xmath122 is an isomorphism . the first betti number of a finitely generated group is @xmath123 ) \otimes_{\mathbb z } { \mathbb q}.\ ] ] given any prime @xmath33 , one can detect @xmath124 in the @xmath33-group quotients of @xmath6 , since it is the _ greatest integer @xmath125 such that @xmath6 surjects @xmath126 for every @xmath127 . _ we exploit this observation as follows : [ l : denseb1 ] let @xmath16 and @xmath6 be finitely generated groups and let @xmath33 be a prime . if @xmath16 is isomorphic to a dense subgroup of @xmath128 , then @xmath129 . for every finite @xmath33-group @xmath130 , each epimorphism @xmath131 will restrict to an epimorphism on both @xmath132 and @xmath16 ( since by density @xmath16 can not be contained in a proper closed subgroup ) . but the resulting map @xmath133 need not be surjective , in contrast to lemma [ ns ] . thus if @xmath6 surjects @xmath126 then so does @xmath16 ( but perhaps not vice versa ) . [ sameb1 ] if @xmath16 and @xmath6 are finitely generated and @xmath134 , then @xmath135 . the proof of the following proposition uses the lyndon - hochschild - serre ( lhs ) spectral sequence , which is explained on page 171 of @xcite . given a short exact sequence of groups @xmath136 , the lhs spectral sequence calculates the homology of @xmath19 in terms of the homology @xmath21 and @xmath8 . the terms on the @xmath137 page of the spectral sequence are @xmath138 , where the action of @xmath8 on @xmath139 is induced by the action of @xmath19 on @xmath21 by conjugation . the terms on the diagonal @xmath140 of the @xmath141 page are the composition factors of a series for @xmath142 . [ p : gpnilp ] let @xmath143 be a short exact sequence of groups and let @xmath144 be the homomorphism induced by @xmath145 . suppose that @xmath21 is finitely generated , that @xmath8 has no non - trivial finite quotients , and that @xmath146 . then @xmath147 is an isomorphism for all @xmath148 , and hence @xmath149 is an isomorphism . in particular , if @xmath6 is residually nilpotent then @xmath21 and @xmath6 have the same nilpotent genus . from the lower left corner of the lhs spectral sequence one obtains the following 5-term exact sequence ( see @xcite page 328 ) : @xmath150 we have assumed that @xmath151 , so the second arrow @xmath152 is an isomorphism . by definition , @xmath153 is the quotient of @xmath154 by the action of @xmath8 . but since @xmath155 is a finitely generated abelian group , @xmath156 is residually finite , whereas @xmath8 has no non - trivial finite quotients . thus , the action of @xmath8 on @xmath24 is trivial and the composition @xmath157 ( which is the map on @xmath158 induced by @xmath159 ) is an isomorphism . moreover , as @xmath8 is perfect and @xmath154 is a trivial @xmath8-module , @xmath160 . and since @xmath161 is also assumed to be zero , the spectral sequence has only one term @xmath162 with @xmath163 that might be non - zero , namely @xmath164 . it follows that the composition @xmath165 ( which is the map on @xmath0 induced by @xmath159 ) is an epimorphism . stallings @xcite proved that if a homomorphism of groups @xmath166 induces an isomorphism on @xmath158 and an epimorphism on @xmath0 , then @xmath144 is an isomorphism for all @xmath148 . we remind the reader that the _ fibre product _ @xmath167 associated to an epimorphism of groups @xmath168 is the subgroup @xmath169 . [ c : fpok ] under the hypotheses of proposition [ p : gpnilp ] , the inclusion @xmath170 of the fibre product induces an isomorphism @xmath171 for every @xmath172 and hence an isomorphism @xmath173 . we have inclusions @xmath174 . the quotient of @xmath26 by @xmath175 is @xmath8 and the quotient of @xmath176 by @xmath175 is @xmath177 . the knneth formula tells us that @xmath178 , and it is obvious that @xmath177 has no finite quotients . thus , by the proposition , both @xmath179 and @xmath180 induce isomorphisms modulo any term of the lower central series , and therefore @xmath181 does as well . the criterion established in proposition [ p : gpnilp ] will be enough for the demands of theorems a and b but we should point out immediately that ( with different finiteness assumptions ) one can prove a stronger result using similar ideas : and @xmath182 induce isomorphisms of profinite completions . the purpose of this subsection is to expand on this remark and prove that the statement about profinite completions is indeed stronger ( an observation that we feel is of independent interest ; cf . remark [ rr ] ) . let @xmath19 be a residually finite group and let @xmath183 be the inclusion of a subgroup @xmath26 . then @xmath184 is called a _ grothendieck pair _ if the induced homomorphism @xmath185 is an isomorphism but @xmath186 is not . ( when no confusion is likely to arise , it is usual to write @xmath187 rather than @xmath184 . ) grothendieck @xcite asked about the existence of such pairs of finitely presented groups and the first such pairs were constructed by bridson and grunewald in @xcite . the analogous problem for finitely generated groups had been settled earlier by platonov and tavgen @xcite . both constructions rely on versions of the following result ( cf . @xcite , @xcite theorem 5.2 and @xcite ) . [ l : pt ] let @xmath188 be a short exact sequence of groups with @xmath6 finitely generated and let @xmath26 be the associated fibre product . suppose that @xmath189 is finitely presented , has no proper subgroups of finite index , and @xmath146 . then 1 . @xmath190 is a grothendieck pair ; 2 . if @xmath21 is finitely generated then @xmath191 is a grothendieck pair . grothendieck pairs give rise to pro - nilpotent equivalences : [ groth - is - enough ] let @xmath35 be a pair of finitely generated , residually finite groups , and for each @xmath148 , let @xmath192 be the induced homomorphism . if @xmath184 is a grothendieck pair , then @xmath147 is an isomorphism for all @xmath148 , and hence @xmath193 is an isomorphism . in particular , if @xmath26 and @xmath6 are residually nilpotent , then they have the same nilpotent genus . before presenting the proof of this proposition we make a remark and establish a lemma . [ rr ] in this proposition , it is vital that the isomorphism between @xmath194 and @xmath47 is induced by a map @xmath195 of discrete groups . for example , as we remarked earlier , there exist finitely generated nilpotent groups that are not isomorphic but have the same profinite completion . a nilpotent group is its own pro - nilpotent completion , so these examples have the same profinite genus but different nilpotent genera . [ groth - bijection ] let @xmath35 be a pair of finitely generated , residually finite groups . if @xmath184 is a grothendieck pair , then for every finite group @xmath19 , the map @xmath196 defines a bijection @xmath197 . as in lemma [ ns ] , by restricting homomorphisms @xmath198 to @xmath199 we obtain a bijection @xmath200 . similarly , there is a bijection @xmath201 . and the isomorphism @xmath202 induces a bijection @xmath203 . given @xmath204 , the map @xmath196 described in lemma [ ns ] completes the commutative square @xmath205&\epi(p , g)\\ \epi(\wh{\g},g)\ar@{>}[r]\ar[u]&\epi(\wh{p},g)\ar[u ] } \ ] ] and hence is a bijection . _ proof of proposition [ groth - is - enough ] : _ if @xmath19 is nilpotent of class @xmath206 , and @xmath24 is any group , then every homomorphism from @xmath24 to @xmath19 factors uniquely through @xmath207 and hence there is a natural bijection @xmath208 . by combining two such bijections with the epimorphism @xmath196 of lemma [ groth - bijection ] , @xmath209 we see that if @xmath19 is finite , then @xmath210 defines a bijection @xmath211 . finitely generated nilpotent groups are residually finite , so for every @xmath212 and every non - trivial element @xmath213 , there is an epimorphism @xmath214 to a finite ( nilpotent ) group such that @xmath215 . the preceding argument provides @xmath216 such that @xmath217 , whence @xmath218 . thus @xmath147 is injective . @xmath147 is also surjective , for if @xmath219 were not in the image then using the subgroup separability of nilpotent groups @xcite , we would have an epimorphism @xmath220 to some finite group such that @xmath221 , contradicting the fact that @xmath222 is an epimorphism . recall that a finitely generated group @xmath19 is said to have a _ solvable conjugacy problem _ if there is an algorithm that , given any pair of words in the generators , can correctly determine whether or not these words define conjugate elements of the group . if no such algorithm exists then one says that the group has an unsolvable conjugacy problem . in the light of proposition [ groth - is - enough ] , the following theorem settles problem [ baumslag2 ] ( cf . remark [ r : chuck1 ] ) . this theorem will be proved by combining the techniques of @xcite with recent advances in the understanding of non - positively curved cube complexes and right - angled artin groups ( raags ) . we remind the reader that a raag is a group with a finite presentation of the form @xmath223=1 \forall ( i , j)\in e\rangle.\ ] ] much is known about such groups . for our purposes here , their most important feature is that they are residually torsion - free - nilpotent ( @xcite theorem 2.1 ) . [ solution1 ] there exist pairs of finitely presented , residually torsion - free - nilpotent groups @xmath35 such that @xmath186 induces isomorphisms of pro - nilpotent and profinite completions , but @xmath6 has a solvable conjugacy problem while @xmath26 has an unsolvable conjugacy problem . let @xmath224 be a short exact sequence of groups and let @xmath225 be the associated fibre product . the 1 - 2 - 3 theorem @xcite states that if @xmath21 is finitely generated , @xmath24 is finitely presented , and @xmath8 has a classifying space with a finite 3-skeleton , then the @xmath26 is finitely presented @xcite . a further result from @xcite states that if @xmath24 is torsion - free and hyperbolic , and @xmath8 has an unsolvable word problem , then the fibre product @xmath226 associated to any epimorphism will have an unsolvable conjugacy problem . on the other hand , the conjugacy problem in @xmath227 is always solvable in polynomial time ( cf . @xcite 1.5 ) . using ideas from @xcite , it is proved in @xcite that there exist groups @xmath8 with no finite quotients that are of type @xmath228 , have @xmath146 and are such that the word problem in @xmath8 is unsolvable . combining the conclusion of the preceding three paragraphs with proposition [ l : pt](1 ) ( or proposition [ p : gpnilp ] in the pro - nilpotent case ) , we see that the theorem would be proved if , given a finitely presented group @xmath8 with the properties described above , we could construct a short exact sequence @xmath229 with @xmath21 finitely generated and with @xmath24 hyperbolic and residually torsion - free - nilpotent . for then @xmath230 is the required pair of groups . in @xcite rips describes an algorithm that , given a finite presentation of any group @xmath8 will construct a short exact sequence @xmath231 where @xmath232 is finitely generated and @xmath24 is hyperbolic . there are many refinements of rips s construction in the literature . haglund and wise @xcite proved a version in which @xmath24 is constructed to be _ virtually special_. by definition , a virtually special group @xmath24 has a subgroup of finite index @xmath233 that is a subgroup of a raag @xcite , and as remarked upon above , raags are residually torsion - free - nilpotent by @xcite . consider the short exact sequence @xmath234 . if we take @xmath8 to be as in third paragraph of the proof , then @xmath235 , because @xmath8 has no proper subgroups of finite index . and @xmath236 , being of finite index in @xmath232 , is finitely generated . thus we have constructed a short exact sequence @xmath237 of the required form . 1 . in the preceding proof we quoted haglund and wise s version of the rips construction ; this was used to ensure that the group @xmath24 was virtually special and hence virtually residually torsion - free - nilpotent . less directly , it follows from recent work of agol @xcite and wise @xcite that the groups produced by the original rips construction @xcite are also virtually special . 2 . if one is willing to reduce the finiteness properties in the above theorem then one can simplify the proof : it is easy to prove that if @xmath8 has an unsolvable word problem and @xmath238 is an epimorphism from a hyperbolic group , then the conjugacy problem in @xmath239 is unsolvable . so if @xmath229 is constructed as in the above proof , then @xmath240 is a grothendieck pair of groups in the same nilpotent genus , but @xmath24 is hyperbolic and @xmath21 has an unsolvable conjugacy problem . [ r : chuck1 ] commenting on an earlier version of this manuscript , chuck miller pointed out to us that his work with baumslag on the isomorphism problem for residually torsion - free nilpotent groups @xcite contains an implicit solution to problem 1.1 . to prove the main theorem in that paper , the authors construct a sequence of finite presentations for residually torsion - free nilpotent groups @xmath241 indexed by words in the generators of an auxillary group @xmath24 that has an unsolvable word problem . @xmath241 is isomorphic to @xmath242 if and only if @xmath243 in @xmath24 . if @xmath244 in @xmath24 then @xmath241 has unsolvable conjugacy problem , whereas @xmath242 has a solvable conjugacy problem . it follows that some @xmath241 with @xmath244 must have the same nilpotent genus as @xmath242 , for if not then one could decide which @xmath241 were isomorphic to @xmath242 be running the following algorithm . given a finite presentation of @xmath245 , use the solution to the isomorphism problem for nilpotent groups to test if the quotients of @xmath241 by the terms of its lower central series are isomorphic to the corresponding quotients of @xmath242 . ( explicit presentations for these quotients are obtained by simply adding basic commutators to the given presentations of @xmath241 and @xmath242 . ) if the nilpotent genus of @xmath241 is different from that of @xmath242 , this process will eventually find non - isomorphic quotients . at the same time , search naively for an isomorphism from @xmath241 to @xmath242 . these parallel processes will together determine whether or not @xmath241 is isomorphic to @xmath242 _ unless _ @xmath241 and @xmath242 are non - isomorphic groups of the same nilpotent genus . notice that although this argument settles problem 1.2 , it does not prove theorem a(2 ) : it does not construct an explicit pair of groups @xmath242 and @xmath246 with the same nilpotent quotients , it just proves that such pairs exist . moreover , there is no homomorphism between the discrete groups inducing the isomorphism of pro - nilpotent completions . we present two constructions settling problem [ baumslag1 ] . both rely on the existence of finitely presented infinite groups @xmath8 with @xmath146 that have no non - trivial finite quotients . a general method for constructing such groups is described in @xcite . the first such group was discovered by graham higman : @xmath247 finitely generated free groups are residually torsion - free nilpotent @xcite , and hence so are subgroups of their direct products . thus the following proposition resolves problem [ baumslag1 ] . [ plat - tav ] let @xmath8 be a finitely presented infinite group with @xmath146 that has no finite quotients . let @xmath248 be a finitely generated free group , let @xmath249 be a surjection , and let @xmath250 be the associated fibre product . then : 1 . @xmath26 is finitely generated but not finitely presented ; 2 . @xmath251 induces an isomorphism of pro - nilpotent completions . let @xmath248 be the free group on @xmath252 . since @xmath8 is finitely presented , the kernel of @xmath249 is the normal closure of a finite set @xmath253 . it is easy to check that @xmath254 is generated by @xmath255 . but grunewald @xcite proves that @xmath26 is finitely presentable if and only if @xmath8 is finite . assertion ( 2 ) is a special case of corollary [ c : fpok ] . platonov and tavgen @xcite proved that @xmath251 also induces an isomorphism of profinite completions . again , let @xmath8 be a finitely presented infinite group with @xmath146 that has no non - trivial finite quotients . in the proof of theorem [ solution1 ] we used a version of the rips construction to obtain a short exact sequence @xmath256 with @xmath24 hyperbolic ( in particular finitely presented ) and residually torsion - free nilpotent , and with @xmath21 finitely generated and not free . a further property of the rips construction is that @xmath24 is small cancellation , in particular of cohomological dimension @xmath257 . bieri @xcite proved that a finitely presented normal subgroup of infinite index in a group of cohomological dimension @xmath257 must be free . thus if @xmath8 is infinite then @xmath21 is not finitely presented . but proposition [ p : gpnilp ] tells us that @xmath21 and @xmath6 have the same nilpotent genus ( and proposition [ l : pt ] tells us that @xmath258 is a grothendieck pair ) . in the previous section we settled the question of whether there exist finitely generated residually torsion - free - nilpotent groups of the same genus such that one is finitely presented and the other was not . in @xcite baumslag laid particular emphasis on a homological variant of this question : is there is a pair of finitely generated residually torsion - free - nilpotent groups of the same genus such that the schur multiplier @xmath259 of one is finitely generated , while that of the other group is not ? we shall settle this question by proving the following theorem . here , and in what follows , we take homology with coefficients in @xmath260 . when we discuss the _ dimension _ of a homology group , we mean its dimension as @xmath260-vector space . note that if @xmath142 is finitely generated then @xmath261 is finite dimensional . [ t : h2 ] there exists a pair of finitely generated residually torsion - free - nilpotent groups @xmath30 that have the same nilpotent genus and the same profinite completion , but @xmath6 is finitely presented while @xmath31 is infinite dimensional . our proof of the above theorem draws on the ideas in previous sections , augmenting them with a spectral sequence argument to control the homology of @xmath21 . the proof also relies on the construction of a group with particular properties that we regard as having independent interest see proposition [ p : newd ] . [ p : ss ] let @xmath262 be a short exact sequence of finitely generated groups . if @xmath263 is finite dimensional but @xmath264 is infinite dimensional , then @xmath31 is infinite dimensional . as in the proof of proposition [ p : gpnilp ] we use the lhs spectral sequence to calculate the homology of @xmath19 in terms of @xmath21 and @xmath8 . the terms on the @xmath137 page of the spectral sequence are @xmath265 . @xmath266 is infinite dimensional and @xmath267 is finite dimensional , so @xmath268 , which is the kernel of @xmath269 , is infinite dimensional . the kernel of @xmath270 is @xmath271 , which is a section of @xmath263 and hence is finite dimensional . so the image of the map to @xmath272 is infinite dimensional . but @xmath272 is a quotient of @xmath273 , which in turn is a quotient of @xmath31 . thus @xmath31 is infinite dimensional . recall that a group @xmath130 is termed _ acyclic _ ( over @xmath274 ) if @xmath275 for all @xmath276 . the higman group described in section [ s : b1 ] was the first example of a finitely presented acyclic group with no proper subgroups of finite index . further examples were constructed in @xcite , including , for each integer @xmath277 , @xmath278,\ , b_1^{-1}[a_2 , a_1^{-1}a_2a_1]\rangle.\ ] ] let @xmath130 be one of the above groups . the salient features of @xmath130 are that it is finitely presented , acyclic over @xmath274 , has no finite quotients , contains a 2-generator free group , @xmath248 say , and is torsion - free ( indeed it has a 2-dimensional classifying space @xmath279 , cf . @xcite p.364 ) . let @xmath280 be the double of @xmath281 along @xmath282 , where @xmath37 is the first stallings - bieri group , i.e. the kernel of a homomorphism @xmath283 whose restriction to each of the factors is surjective . the key features of @xmath37 are that it is finitely generated but @xmath284 is infinite dimensional ( see @xcite , or @xcite pp . 482 - 485 ) . [ delta ] @xmath285 is torsion - free , finitely presented , has no non - trivial finite quotients , and @xmath286 is infinite dimensional . the amalgamated free product of two finitely presented groups along a finitely generated subgroup is finitely presented , so @xmath285 is finitely presented . and an amalgam of torsion - free groups is torsion - free . the four visible copies of @xmath130 generate @xmath285 , and these all have trivial image in any finite quotient , so @xmath285 has no non - trivial finite quotients . we calculate @xmath286 using the mayer - vietoris sequence ( omitting the coefficient module @xmath260 from the notation ) : @xmath287 as @xmath130 is acyclic , so is @xmath281 , by the knneth formula . hence @xmath288 is infinite dimensional . recall that a group @xmath19 is termed _ super - perfect _ if @xmath289 . proposition [ l : pt ] explains our interest in this condition . @xmath285 is perfect but it is not super - perfect . [ h2d ] @xmath290 . a slight variant of the above mayer - vietoris argument shows that @xmath291 . the first homology of @xmath37 can be calculated using theorem a of @xcite : if @xmath292 is a subdirect product and we write @xmath293 and @xmath294 , then @xmath295 where @xmath296 . in our case , @xmath297 and @xmath298 , so @xmath299 and @xmath300 is cyclic . ( recall that @xmath301 is free of rank @xmath257 . ) to obtain a super - perfect group , we pass to the _ universal central extension _ @xmath302 . [ p : newd ] @xmath303 is torsion - free , finitely presented , super - perfect , has no non - trivial finite quotients , and @xmath304 is infinite dimensional . the standard theory of universal central extensions ( see @xcite , chapter 5 ) tells us that @xmath303 is perfect and that there is a short exact sequence @xmath305 since @xmath285 and @xmath306 are finitely presented and torsion - free , so is @xmath303 . since @xmath285 has no non - trivial finite quotients , @xmath306 would have to map onto any finite quotient of @xmath303 , which means that all such quotients are abelian . since @xmath303 is perfect , it follows that it has no non - trivial finite quotients ( cf . @xcite , p.369 ) . to see that @xmath304 is infinite dimensional , we consider the lhs spectral sequence associated to the above short exact sequence . as @xmath285 is finitely presented and @xmath307 is of type @xmath308 , all of the groups in the first three columns , @xmath309 with @xmath310 , are finite dimensional . on the other hand , @xmath311 is infinite dimensional . therefore @xmath312 and @xmath313 are infinite dimensional . but @xmath271 is a quotient of @xmath304 , so @xmath304 is also infinite dimensional . we have constructed a group @xmath303 that is super - perfect , finitely presented and has @xmath314 infinite dimensional . by applying a suitable version of the rips construction to @xmath303 ( as in the proof of theorem [ solution1 ] ) , we obtain a short exact sequence @xmath315 with @xmath21 finitely generated and @xmath24 is a 2-dimensional hyperbolic group that is virtually special . passing to a subgroup of finite index @xmath233 and replacing @xmath21 by @xmath316 , we may assume that @xmath24 is a subgroup of a raag , and hence is residually torsion - free - nilpotent . propositions [ p : gpnilp ] and [ l : pt](2 ) tell us that @xmath317 induces an isomorphism of pro - nilpoent and profinite completions . proposition [ p : ss ] tells us that @xmath31 is infinite dimensional . [ r : chuck2 ] we outline an alternative proof of theorem [ t : h2 ] suggested by chuck miller . let @xmath130 be a group satisfying the conditions of section 6.2 . we fix a finite presentation @xmath318 then augment if by adding a new generator @xmath319 and a new relation @xmath320 . let @xmath248 be the free group on @xmath321 , let @xmath322 be the natural epimorphism , and let @xmath323 be the associated fibre product . the results in section 3 tell us that @xmath26 and @xmath324 have the same nilpotent quotients . because @xmath325 is part of a free basis for @xmath326 , one can express @xmath26 as an hnn extension with stable letter @xmath319 and amalgamated subgroup @xmath327 . then , as in the proof of theorem 4 of @xcite , one can use the mayer - vietoris sequence for the hnn extension to prove that @xmath328 maps onto the first homology of the infinitely generated free group @xmath329 . we settled baumslag s first three questions by constructing groups of a somewhat pathological nature that lie in the same nilpotent genus as well - behaved groups . our solution to problem [ baumslag4 ] is of an entirely different nature : the point here is to prove that parafree groups share a significant property with free groups . correspondingly , the nature of the mathematics that we shall draw on is entirely different . our proof of the following theorem relies on a mix of @xmath1 betti numbers and profinite group theory that we first employed in our paper @xcite with m. conder . ( = theorem [ t : n ] ) [ maininfindex ] if @xmath132 is a finitely generated parafree group , then every finitely generated non - trivial normal subgroup of @xmath6 is of finite index . the standard reference for this material is lck s treatise @xcite . in what follows @xmath330 denotes usual first betti number of a group , and @xmath331 denotes the first @xmath1 betti number . we shall not recall the definition of the @xmath1 betti number as it does not inform our arguments . in the case of finitely presented groups , one can use _ lck s approximation theorem _ @xcite to give a surrogate definition of @xmath331 : suppose that @xmath6 is _ finitely presented _ and let @xmath332 be a sequence of normal subgroups , each of finite index in @xmath6 , with @xmath333 ; lck proves that @xmath334 } = b_1^{(2)}(\g).\ ] ] [ bf ] let @xmath248 be a free group of rank @xmath2 . euler characteristic tells us that a subgroup of index @xmath98 in @xmath248 is free of rank @xmath335 , so by lck s theorem @xmath336 . a similar calculation shows that if @xmath337 is the fundamental group of a closed surface of genus @xmath338 , then @xmath339 . if one assumes only that the group @xmath6 is _ finitely generated _ , then one does not know if the above limit exists , and when it does exist one does not know if it is independent of the chosen tower of subgroups . this is a problem in the context of theorem [ maininfindex ] because we do not know if finitely generated parafree groups are finitely presentable . thus we appeal instead to the weaker form of lck s approximation theorem established for finitely generated groups by lck and osin @xcite . [ luck ] if @xmath6 is a finitely generated residually finite group and @xmath340 is a sequence of finite - index normal subgroups with @xmath333 , then @xmath341 } \leq b_1^{(2)}(\g).\ ] ] another result about @xmath1 betti numbers that we will make use of is the following theorem of gaboriau ( see @xcite theorem 6.8 ) . [ gab ] suppose that @xmath342 is an exact sequence of groups where @xmath21 and @xmath16 are infinite . if @xmath343 , then @xmath344 . the key step in the proof of theorem [ maininfindex ] is the following result ( cf . @xcite proposition 3.2 ) . recall that we write @xmath43 to denote the pro-@xmath33 completion of a group @xmath6 , and @xmath345 to denote the closure of a subgroup in @xmath43 . [ boundl_2 ] let @xmath6 be finitely generated group and let @xmath248 be a finitely presented group that is residually-@xmath33 for some prime @xmath33 . suppose that there is an injection @xmath346 and that @xmath347 . then @xmath348 . for each positive integer @xmath98 let @xmath349 be the intersection of all normal subgroups of @xmath33-power index at most @xmath98 in @xmath248 . let @xmath350 . we saw in corollary [ c : exhaust - p ] that @xmath351 , hence @xmath352 . since @xmath6 and @xmath248 are both dense in @xmath353 , the restriction of @xmath354 to each of these subgroups is surjective , and therefore @xmath355 = [ \widehat{f}_p : \overline{n(d)}]= [ f : n(d)].\ ] ] @xmath356 is dense in @xmath115 and in lemma [ l : same ] we saw that @xmath357 , so lemma [ l : denseb1 ] implies that @xmath358 . we use the towers @xmath359 in @xmath6 and @xmath360 in @xmath248 to compare @xmath1-betti numbers , applying theorem [ luck ] to the finitely generated group @xmath6 and lck s approximation theorem to the finitely presented group @xmath6 : @xmath361 } } } \geq\displaystyle \limsup\limits_{d\rightarrow \infty } { { { b_1(n(d))}\over{[f : n(d ) ] } } } = \displaystyle \lim\limits_{d\rightarrow \infty } { { { b_1(n(d))}\over{[f : n(d ) ] } } } = b_1^{(2)}(f)\ ] ] as required . [ c : samel2 ] let @xmath16 and @xmath6 be finitely presented groups that are residually-@xmath33 for some prime @xmath33 . if @xmath362 then @xmath363 . by combining proposition [ boundl_2 ] with theorem [ gab ] we deduce : [ t : general ] let @xmath6 be a finitely generated group and let @xmath364 be a non - trivial normal subgroup . let @xmath248 be a finitely presented group that is residually-@xmath33 for some prime @xmath33 and suppose that there is an injection @xmath365 with dense image . if @xmath366 , then either @xmath367 or else @xmath368 . in particular , if @xmath21 is finitely generated then it is of finite index . * proof of theorem [ maininfindex ] . * suppose now that @xmath6 is a finitely generated parafree group , with the same nilpotent genus as the free group @xmath248 , say . finitely generated groups with the same nilpotent genus have isomorphic pro-@xmath33 completions for every prime @xmath33 ( cf . remark 2.1(9 ) ) , so @xmath369 . if @xmath248 is cyclic , then it is easy to see that @xmath132 must also be cyclic , so the conclusion of theorem [ maininfindex ] holds . so we assume that @xmath248 has rank @xmath370 . @xmath6 is residually-@xmath33 for all primes @xmath33 . indeed , since @xmath6 is residually nilpotent , every non - trivial element of @xmath6 has a non - trivial image in @xmath371 for some term @xmath372 of the lower central series , and the free nilpotent group @xmath373 is residually-@xmath33 for all primes @xmath33 ( see @xcite ) . combining this observation with the conclusion of the first paragraph , we see that the natural map from @xmath6 to @xmath374 is injective . in example [ bf ] we showed that @xmath375 . thus we are in the situation of theorem [ t : general ] and the proof is complete . we close with some further consequences of the arguments in the preceding section , and a discussion of related matters . recall that a group @xmath6 is said to be _ parafree of rank @xmath2 _ if @xmath6 is residually nilpotent and is in the same nilpotent genus as a free group of rank @xmath2 . as special cases of proposition [ boundl_2 ] and corollary [ c : samel2 ] we have : a group @xmath132 is _ @xmath4-simple _ if its _ reduced @xmath4-algebra _ @xmath377 is simple as a complex algebra ( i.e. has no proper two - sided ideals ) . by definition , the _ reduced @xmath4-algebra _ @xmath378 is the norm closure of the image of the complex group algebra @xmath379 $ ] under the left - regular representation @xmath380\to \mathcal l ( \ell^2(\gamma))$ ] defined for @xmath7 by @xmath381 for all @xmath382 and @xmath383 . a group @xmath132 is @xmath4-simple if and only if any unitary representation of @xmath132 which is weakly contained in @xmath384 is weakly equivalent to @xmath384 . we refer the reader to @xcite for a thorough account of the groups that were known to be _ @xmath4-simple _ by 2006 . the subsequent work of peterson and thom @xcite augments this knowledge . an important early result in the field is the proof by powers @xcite that non - abelian free - groups are @xmath4-simple . in contexts where one is able to adapt the powers argument , one also expects the canonical trace to be the only normalized trace on @xmath378 ( cf . appendix to @xcite ) . by definition , a linear form @xmath385 on @xmath386 is a _ normalised trace _ if @xmath387 and @xmath388 for all @xmath389 , and the _ canonical trace _ is uniquely defined by @xmath390 for every _ finite _ sum @xmath391 where @xmath392 and @xmath393 contains @xmath394 . corollary [ l2parafree ] tells us that @xmath396 . by definition , @xmath6 is residually torsion - free nilpotent , and therefore it satisfies condition @xmath397 of @xcite , section 4 , i.e. every non - trivial element of @xmath398 $ ] acts without kernel on @xmath399 . with these facts in hand , the present result follows immediately from corollary 4.6 of @xcite ( which in turn relies on @xcite ) . section 4 of peterson and thom @xcite contains a number of other results concerning the structure of finitely presented groups that satisfy their condition @xmath397 and have non - zero @xmath331 . in the context of parafree groups , the following consequence of our results and @xcite has a particular appeal as it articulates a further freeness property of parafree groups . [ freiheitssatz ] let @xmath132 be a finitely generated parafree group of rank @xmath395 . then every generating set @xmath400 has an @xmath2-element subset @xmath401 such that the subgroup of @xmath19 generated by @xmath402 is free of rank @xmath2 . with corollary [ l2parafree ] in hand , we can apply corollary 4.7 of @xcite where the conclusion of the corollary is shown to hold for any finitely generated group @xmath6 with @xmath403 that satisfies @xmath397 this result strengthens greatly magnus s observation that an @xmath2-generator parafree group of rank @xmath2 must be free of rank @xmath2 . also , the proof shows that if @xmath6 is an @xmath404-generator parafree group of rank @xmath2 ( not necessarily finitely presented ) then @xmath405 ( cf . corollary [ l2parafree ] ) . the _ parafree conjecture _ posits that @xmath406 for every finitely generated parafree group @xmath19 . if this were proved , then corollary [ freiheitssatz ] would be a consequence of a theorem of stallings @xcite which states that if @xmath19 is a finitely generated group with @xmath406 and @xmath407 then any @xmath408 that is independent in @xmath409 freely generates a free subgroup . if @xmath6 were a lattice then it would be finitely generated , so @xmath396 by corollary [ l2parafree ] . it is shown in @xcite that if @xmath6 is a lattice in a connected lie group and @xmath396 , then @xmath6 is commensurable with a lattice in @xmath410 ; i.e. the group is virtually free or virtually the fundamental group of a closed orientable surface of genus at least @xmath257 . now @xmath6 is torsion - free , with torsion - free abelianization , so in fact , in this case , @xmath6 is free or the fundamental group @xmath411 of a closed orientable surface of genus @xmath412 . the former is ruled out by assumption , and the latter is ruled out by the observation that @xmath413 ; see example [ bf ] . ( alternatively , it is straightforward to construct a finite nilpotent quotient of a free group of rank @xmath414 that can not be a quotient of the genus @xmath338 surface group . ) from @xcite , we have the presentation @xmath419=1,uzu^{-1}=v^{-1}zvz , l = v^{-1}uzu^{-1}vz\>,\ ] ] where @xmath186 is a meridian for the unknotted component and @xmath420 is a longitude for @xmath186 . the other peripheral subgroup ( i.e. the one corresponding to the square knot component ) is @xmath421 , with meridian @xmath422 and longitude @xmath423 . the two meridians generate @xmath424 . by setting @xmath425 we obtain a homomorphism from @xmath6 onto @xmath426 ; i.e. the free group of rank @xmath257 . this homomorphism @xmath427 induces an isomorphism @xmath428 and a surjection on higher homology groups ( since @xmath429 for all @xmath430 ) . an application of stallings theorem @xcite now establishes that @xmath6 has the same lower central series quotients as @xmath431 . note that the homomorphism @xmath427 can be realised topologically by performing @xmath432-surgery on both the unknotted component ( forcing @xmath433 ) and the knotted component ( forcing @xmath434 , hence @xmath425 ) . the closed manifold resulting from these surgeries is a connected sum of two copies of @xmath435 . corollary [ notlattice ] ( alternatively , remark [ r : ref1 ] ) tells us that @xmath132 is not residually nilpotent . in fact , it is not difficult to see that the longitudes described above lie in the intersection of the terms in the lower central series of @xmath6 . [ r : ref1 ] an anonymous referee pointed out to us that the lattice described above exemplifies the following more general phenomenon . let @xmath248 be a free group , let @xmath19 be a finitely generated group that is not free , and suppose that there is an epimorphism @xmath436 that induces an isomorphism @xmath437 . then @xmath19 is not residually nilpotent . to see that this is the case , note that since @xmath248 is free there is a homomorphism @xmath438 such @xmath439 . the map that @xmath440 induces on abelianisations is the inverse of @xmath441 ; in particular @xmath442 generates @xmath443 $ ] . but if a set generates a nilpotent group modulo the commutator subgroup , then it generates the nilpotent group itself . it follows that @xmath444 induces an isomorphism @xmath445 for all @xmath212 . therefore , @xmath446 ( which is assumed to be non - trivial ) is contained in the intersection of the terms of the lower central series of @xmath19 . the link depicted above is an example of a homology boundary link . recall that a link @xmath447 of @xmath448 components is called a _ homology boundary link _ if there exists an epimorphism @xmath449 where @xmath248 is a free group of rank @xmath450 . by alexander duality , @xmath451 . thus @xmath452 induces an isomorphism @xmath453 , and arguing as above we conclude that ( with the exception of trivial links ) the fundamental groups of homology boundary links are never residually nilpotent . as in the example above , @xmath452 induces also induces a surjection @xmath454 for all @xmath455 . therefore , by stallings theorem @xcite , @xmath452 induces an isomorphism between the lower central series quotients of @xmath456 and @xmath248 . as in the example above , the fundamental group of the complement of a homology boundary link ( that is not the trivial link of m components ) is not residually nilpotent . on the other hand , standard results guarantee that all lattices in @xmath415 are _ virtually _ residually @xmath33 for all but a finite number of primes @xmath33 ( and hence virtually residually nilpotent ) . recent work of agol @xcite implies that these lattices are also virtually residually torsion - free - nilpotent . one says that a group @xmath6 is _ pro - p good _ if for each @xmath457 , the homomorphism of cohomology groups @xmath458 induced by the natural map @xmath459 is an isomorphism , where the group on the left is in the continuous cohomology of @xmath67 . one says that the group @xmath6 is _ cohomologically complete _ if @xmath6 is pro-@xmath33 good for all primes @xmath33 . it is shown in @xcite that many link groups are cohomologically complete , and indeed it was claimed by hillman , matei and morishita @xcite that all link groups are cohomologically complete . counterexamples to the method of @xcite were given in @xcite , but @xcite left open the possibility that link groups might nevertheless be cohomologically complete . here we note that homology boundary links provide counterexamples . in particular there are hyperbolic links that provide counterexamples . [ nocohomcom ] let @xmath416 be a homology boundary link that is not the trivial link of @xmath450 components . then @xmath456 is not pro-@xmath33 good for any prime @xmath33 . in particular , @xmath456 is not cohomologically complete . let @xmath460 . since @xmath6 has the same nilpotent quotients as the free group of rank @xmath450 , that @xmath43 is isomorphic to a free pro-@xmath33 group . hence @xmath461 on the other hand , since @xmath416 is a non - trivial link with @xmath448 components , @xmath462 has dimension @xmath463 as an @xmath464-vector space . by way of contrast , we also note that the lattice @xmath6 described in 8.2 is _ good _ in the sense of serre ; i.e. for each @xmath457 and for every finite @xmath6-module @xmath465 , the homomorphism of cohomology groups @xmath466 induced by the natural map @xmath459 is an isomorphism between the cohomology of @xmath6 and the continuous cohomology of @xmath67 . note that goodness deals with the profinite completion and not the pro - p completions . theorem [ t : general ] can be usefully applied to cases where @xmath248 is not free ; for example , @xmath248 might be a non - abelian surface group or more generally non - abelian limit group . that these satisfy the condition on the first @xmath1 betti number can be seen in example 7.2 for surface groups and @xcite for limit groups . this leads to an analogue of theorem 7.1 for _ paralimit _ groups . note also that , for both surface groups and limit groups , condition @xmath397 of @xcite holds ( since these groups are residually torsion - free nilpotent ) . hence , groups that are paralimit or parasurface groups , in the above sense , also satisfy a version of freiheitssatz as in corollary [ freiheitssatz ] . g. baumslag , _ groups with the same lower central sequences _ , in guido s book of conjectures . a gift to guido mislin on the occasion of his retirement from e.t.h . collected by indira chatterji . monographies de lenseignement mathmatique * 40 * geneva , ( 2008 ) .

the first catalog of the rr lyrae stars ( rrls ) in the galactic halo by the quasar equatorial survey team ( quest ) has been searched for significant overdensities that may be debris from disrupted dwarf galaxies or globular clusters . these rrls are contained in a band wide in declination that spans in right ascension and lie to kpc from the sun . away from the major overdensities , the distribution of these stars is adequately fit by a smooth halo model , in which the flattening of the halo decreases with increasing galactocentric distance . this model was used to estimate the `` background '' of rrls on which the halo overdensities are overlaid . a procedure was developed for recognizing groups of stars that constitute significant overdensities with respect to this background . to test this procedure , a monte carlo routine was used to make artificial rrls surveys that follow the smooth halo model , but with poisson distributed noise in the numbers of rrls and , within limits , random variations in the positions and magnitudes of the artificial stars . the artificial surveys created by this routine were examined for significant groups in exactly the same way as the quest survey . these calculations provided estimates of the frequencies with which random fluctuations produce significant groups . in the quest survey , there are six significant overdensities that contain six or more stars and several smaller ones . the small ones and possibly one or two of the larger ones may be artifacts of statistical fluctuations , and they need to be confirmed by measurements of radial velocity and/or proper motion . the most prominent groups are the northern stream from the sagittarius dwarf spheroidal galaxy and a large group in virgo , formerly known as the `` 12.4 hr clump '' , which recently have shown contains a stellar stream ( the virgo stellar stream ) . two other groups lie in the direction of the monoceros stream and at approximately the right distance for membership . another group is related to the globular cluster palomar 5 .

astro-ph0604359

recent surveys of the galactic halo have shown that it contains streams of stars emanating from dwarf galaxies . the firmest evidence for this comes from the numerous detections of streams from the sagittarius ( sgr ) dwarf spheroidal galaxy ( hereafter dsph ) , which wrap around the sky @xcite . there is also strong evidence for another large stream , the monoceros stream @xcite , that forms a ring - like structure around the milky way @xcite . several authors have argued that the parent galaxy of the mon stream lies in the constellation canis major , but there is considerable controversy over this association and even the existence of the cma galaxy ( * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * among others ) . these on - going merger events may be only the most recent examples of a long history of mergers that build the milky way from an ensemble of smaller systems , as proposed by hierarchical picture of galaxy formation ( e.g. * ? ? ? while the evidence for ancient merger events are less spectacular than that for the recent ones , it is nonetheless compelling . it has long been speculated that the thick disk was produced by a merger of the milky way with a relatively large satellite galaxy soon after the formation of the first disk structure ( see review by * ? ? ? * ) . although the very unusual globular cluster @xmath5cen has not yet been directly linked to a merger event , its exceptional structure and internal ranges in metallicity and in age has fueled speculation that it is the nucleus of a now extinct nucleated dwarf galaxy ( * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * among others ) . simulations of the destruction of dwarf galaxies @xcite have shown that the tidal debris may be recognized in the halo as over - densities in space and as coherent structures in radial velocity space long after the merger event . by detecting this debris , modeling their orbits , and studying their ages and chemical compositions , the merger history of the milky way may be pieced together . in addition to documenting the importance of mergers to galactic evolution , this may help explain the large inconsistency between the number of satellite galaxies predicted by the hierarchical picture for the formation of a large galaxy and the relatively small number of satellite galaxies around the milky way , the `` satellite problem '' ( see * ? ? ? * ) . debris from satellite galaxies is not the only kind expected in the galactic halo . numerical simulations have shown that several processes can lead to the disruption of globular clusters , and there has been much speculation that the roughly 150 the globular clusters that are now identified in the galaxy are the survivors of a once much larger population ( e.g. * ? ? ? a few examples of tidal tails from globular clusters have been reported @xcite , the most spectacular of which are the streams that stretch several degrees from the globular cluster pal 5 @xcite . @xcite have detected some small groups of rr lyrae variables that have similar radial velocities and metallicities , precisely the properties expected of debris from disrupted globular clusters . in this paper we discuss the spatial distribution of the stars that were discovered in the first band of the quest rr lyrae survey @xcite . rr lyrae stars ( hereafter rrls ) were selected for this survey of halo substructure because they are easily detected and are excellent standard candles . our primary goal is to identify overdensities that may be debris from dwarf galaxies or from globular clusters . in a few structures identified here , the densities of the rrls are so much higher than the average densities of variables that there is little doubt that the feature is real and not a random fluctuation in the background . not all of the other features identified here may be real , and they require confirmation by determining if their member stars have similar motions . we will report in later papers our measurements of the radial velocities and the metallicities of the stars in some of these spatial groups . before discussing the quest survey , it is import to consider a bias that affects all rrls surveys . rrls are only found in the oldest ( age @xmath6 gyrs ) stellar populations that have the proper combination of [ fe / h ] and other factors ( `` 2nd parameters '' e.g. , age , cno / fe , he / h , and stellar density ) to enable horizontal branch ( hb ) stars to evolve into the instability strip . these complexities of hb morphology produce a wide range of rrls populations even among the outwardly similar old and metal - poor globular clusters in the galactic halo . some of these clusters contain several tens of rrls , others contain only a few rrls , and some contain none at all . clearly a search for the debris from destroyed globular clusters is limited by this effect , but does it also seriously impact a search for stellar streams from dwarf galaxies ? the rrls populations in globular clusters has been quantified by @xcite , who following @xcite , calculated the specific frequency of rrls which they defined as the number of rrls per unit absolute visual magnitude ( @xmath7 ) , normalized to @xmath8 ( we call this @xmath9 following the 2003 version of the @xcite catalogue ) . this fiducial @xmath7 is near the peak of the distribution of the milky way globular clusters . using measurements of @xmath9 for a large sample of globular clusters , @xcite demonstrated that there is a systematic variation in the frequencies of rrls with galactocentric distance ( @xmath10 ) in the sense that the metal - poor globular clusters in the outer halo ( @xmath11 ) have higher frequencies than do inner halo clusters of the same metallicity . this effect , which is due to the variation of the 2nd parameter with @xmath10 ( e.g. * ? ? ? * ) may explain the difference in mean metallicity between the field rrls lying in the inner and outer halos @xcite . we are interested here in examining @xmath9 in low - mass dwarf galaxies as well as in globular clusters . the dsph galaxies provide a convenient sample of dwarf galaxies covering the extreme low - mass range of the distribution of galaxies . because they are the most numerous type of satellite galaxy around the milky way and m31 , it is likely that dsph galaxies were the type most frequently destroyed in the past . the sgr dsph galaxy is of course undergoing tidal destruction at the present time . values of @xmath9 for 9 of the 10 known dsph companions of the milky way and for 4 of the 7 dsph companions of m31 have been computed and are listed in table [ tab - frequency ] ( insufficient data were available for the other dsph companions ) . with two exceptions , the values of @xmath7 in this table were taken from @xcite and @xcite for the milky way and the and systems , respectively . @xcite have shown that draco has a significantly larger radius than was previously measured . to estimate the @xmath7 of draco , we adopted @xmath12 for its integrated color , which is consistent with its old and very metal - poor stellar population . with the transformation equation in @xcite , this value yields @xmath13 , which we added to the value of @xmath14 @xcite . for the sgr dsph , we adopted the @xmath7 that @xcite derived , which lies in near the middle of the range quoted in the recent literature . the values of the mean [ fe / h ] of the galaxies were taken from @xcite for the milky way companions and from @xcite for the and galaxies . because dsph galaxies have significant internal dispersions in [ fe / h ] and because the rrls are among the oldest stars , the mean [ fe / h ] of a galaxy may be larger than the mean value of its rrls . the important point for our discussion is that the rrls in these systems probably span a range in [ fe / h ] that is not grossly different from the range observed among the rrls in the outer galactic halo ( * ? ? ? * @xmath15 } > -2.5 $ ] ) . it is possible then that the field rrls formed in similar galaxies that were later torn apart . in table [ tab - frequency ] , @xmath16 is the number of rrls that were observed in the fields covered in the variability searches , and @xmath17 is our estimate of the total number of rrls in the galaxies . for most of the galaxies , we calculated @xmath17 from @xmath16 by dividing @xmath16 by the fraction of total galaxy light that is included in the observed field . for the milky way systems , we computed these fractions from @xcite profiles , using the parameters given by @xcite . the ellipticities of the galaxies were taken into account , and it was assumed that the observed fields coincided with the centers of the systems . @xcite estimated the fractions of the luminosities of the and galaxies that were covered by their fields , which made the calculations of @xmath17 straightforward . for the sgr dsph galaxy , we adopted the estimate by @xcite that its main body contains 4200 type ab variables . we then used the observed ratio of the type c to type ab rrls to derive @xmath17 . for draco , we set @xmath18 because @xcite did not specify the size of their field . draco is so rich in variables that it matters little to our discussion that this procedure may have underestimated its @xmath9 . the standard deviations ( @xmath19 ) of the values of @xmath9 were calculated by assuming poisson statistics applies to @xmath17 and by adopting the uncertainties in @xmath7 given by their sources . because the @xmath17 values are only estimates , the @xmath19 s listed in table [ tab - frequency ] should be treated as lower limits . the variability studies probably missed some variables , and consequently , the values of @xmath9 in table [ tab - frequency ] are more likely to be too small than too large . this is particularly true for the fornax galaxy because @xcite state there are other variable stars of the right magnitude to be rrls for which they could not determine periods . in figure [ fig - nrr ] , @xmath9 for the globular clusters ( harris 2003 catalogue ) and the dsph galaxies are compared . one can see again the difference in @xmath9 among the metal - poor globular clusters in the inner and outer halos . the near zero values of @xmath9 among the metal - poor clusters of the inner halo suggest that rrls may not be a good tracer of its stellar populations . rrls may be better probes of the outer halo , where the mean value of @xmath9 for the clusters is well above zero . the dsph galaxies span a wide range in @xmath9 , but none has @xmath20 . draco has a remarkably high value that is comparable to the highest values in the sample of 90 globular clusters . several of the other dsph galaxies have values of @xmath9 that are comparable to the variable - rich outer halo clusters . this is somewhat remarkable because many of these galaxies contain substantial populations of intermediate - aged stars that contribute to the total luminosities of the galaxies but not of course to the samples of rrls . @xmath9 is plotted against @xmath7 for the galaxies in figure [ fig - nrrmv ] , where one can see a clear trend of decreasing @xmath9 with increasing galaxy luminosity . this is not surprising because the more luminous galaxies tend to have larger populations of intermediate - age stars . note that the milky way and and galaxies are indistinguishable in this plot . the small but still significantly above zero values of @xmath9 among the high luminosity galaxies indicate that their destruction will release large numbers of rrls . it is not surprising then that the tidal streams from the sgr dsph galaxy have been easily detected by rrls surveys ( see below and also * ? ? ? the higher @xmath9 among the low luminosity dsph galaxies partially offsets the greater difficulty that any probe of the halo will have detecting the smaller streams that are produced by the tidal destruction of low - luminosity systems . if the 13 galaxies in table [ tab - frequency ] are representative of the dwarf galaxies that merged with the milky way , then rrls surveys for stellar streams are probably not seriously handicapped by the population bias . they may also detect the debris from the destruction of the most luminous and variable rich globular clusters . because rrls are superior standard candles ( see below ) , rrls surveys may provide a better description of halo substructure than other surveys that employ probes that suffer less from population biases ( e.g. , k giants ) . the first band of the quest rr lyrae survey ( * ? ? ? * hereafter paper i ) identified 498 rrls in almost 380 deg@xmath21 of the sky that spans a wide range of galactic coordinates and in apparent magnitude ( @xmath22 ) . paper i describes in detail the techniques of the survey and its completeness . the survey covered a @xmath23 wide strip of the sky , centered at declination @xmath24 , from right ascencion ( @xmath25 ) @xmath26 to @xmath27 and from @xmath28 to @xmath29 . the span from @xmath27 to @xmath28 was not observed because it includes regions near the galactic plane . subsequent work on the individual stars in the catalogue has shown that 41 of them are not real variables . their apparent variability was produced in most cases by close neighbors not resolved by the quest instrumentation . the photometric pipeline that was used did not include a deblending algorithm . the regions most affected by blends were the ones closest to the galactic plane which present the most crowding . these regions were also the less observed in the quest survey and consequently there is a small number of observations per star . with few epochs available there was a greater chance that the variations in magnitude that were produced by the poor centering on a blended image mimicked a rrls light curve . only 3 cases of blends were found at high galactic latitudes ( @xmath30 ) , which represents only the @xmath31 of all the stars in the catalogue in that region.we eliminated all cases of blends before preforming the following analysis on the remaining 457 rrls . the distances of the rrls from the sun ( @xmath32 ) are estimated in the usual way from measurements of their mean v magnitudes , @xmath33 , and interstellar extinctions , @xmath34 , and by assuming an average value for their absolute magnitudes , @xmath7 : @xmath35 because our ability to detect small halo substructures depends on the precisions of the distances obtained , it is essential to estimate the errors of these measurements and to investigate the variation in @xmath7 among the rrls . the uncertainty in the magnitude system of the quest observations ( 0.02 mag , see paper i ) is sufficiently small that random errors dominate the error budget for most of the rrls . the values of @xmath33 for the type ab rrls ( rr_ab _ ) were determined by fitting template light curves from @xcite to the quest observations and then integrating these curves after they were transformed to intensity units ( see paper i ) . for the type c variables ( rr_c _ ) , the mean value of the individual v observations was adopted for @xmath33 . the errors in the values of @xmath33 obtained by these methods vary from 0.01 to 0.12 and average 0.05 . this variation is caused by a combination in the errors in the v measurements , which increase with increasing v , and the uneven coverage in phase of the observations . the interstellar reddenings of each rrls was obtained from the dust maps of @xcite and transformed to extinction using @xmath36 . figure [ fig - ext ] shows the value of @xmath34 for each star as a function of right ascension . the instellar extinction is low except in the directions approaching the galactic plane . for example , in the region between @xmath37 h , which includes portions of the dense molecular clouds of the orion star forming region , some stars have up to @xmath38 magnitudes of extinction . since it is very difficult to estimate the completeness of the survey in this part of the sky where the extinction is highly variable , this region is excluded when we consider the shape and density fall - off of the halo . it is examined , however , for density enhancements . most of our survey is at much higher galactic latitudes where @xmath39 . since @xcite estimate that the errors in their values of e(b - v ) are @xmath40 , the standard deviation of @xmath34 , @xmath41 , is probably @xmath42 for most stars . the @xmath7 of a rrls depends on its position on the evolutionary track from the zero age horizontal branch ( zahb ) and on its metallicity , which alters the evolutionary tracks and timescales and changes the @xmath7 of the zahb . it is convenient to summarize this as : @xmath43 , where @xmath44 is the mean absolute visual magnitude of the horizontal branch at the instability strip and @xmath45 is a star s deviation from this mean value due to the state of its evolution . the dispersion in these quantities are estimated in the following analysis . the star to star variation in @xmath33 among the rrls in a globular cluster , @xmath46 , provides a measure of the dispersion in @xmath45 at fixed [ fe / h ] because there is very little variation in [ fe / h ] among the stars in a typical globular cluster . it is essential to calculate @xmath46 for a number of clusters that span a range in [ fe / h ] because of the changes in track morphology makes @xmath46 larger for the metal rich clusters ( see * ? ? ? figure [ fig - sigmavhb ] shows the dependence of @xmath46 on [ fe / h ] , where we have plotted the values of @xmath46 that we calculated for the clusters m3 , m15 , m92 , ngc 3201 , 6171 , 6712 , 6723 and 6981 from the photographic photometry compiled by @xcite . we ignored sandage s data for @xmath5cen because its rrls vary in [ fe / h ] and for m4 because it has variable interstellar extinction across its face ( * ? ? ? * ; * ? ? ? * and references therein ) , which significantly increases the value of @xmath46 . accurate background subtraction is notoriously difficult in the crowded fields of globular clusters . since it is done differently by the techniques of photographic and ccd photometry , we measured a few values of @xmath46 for clusters that had been observed by ccd cameras . the ccd photometry of the rrls in ngc 6171 by @xcite yields a value of @xmath46 that agrees to within the errors with the photographic value , and the values that we obtained from the ccd photometry of m5 @xcite , ic 4499 @xcite , and m68 @xcite are consistent with the measurements from the photographic photometry of other clusters that have similar metallicities . in all cases , the variation in @xmath33 among the rrls in a cluster is approximately gaussian . because very few halo rrls are as metal rich as [ fe / h]@xmath47 @xcite , we adopt @xmath48 for the dispersion in @xmath45 . figure [ fig - sigmavhb ] shows that this value is clearly an overestimate for all but the most metal rich rrls . while it is well known that the @xmath44 varies with [ fe / h ] , there is much debate among different authors over the functional form of this variation and the values of @xmath7 at particular values of [ fe / h ] . this has been reviewed by @xcite , who obtained from a weighted average of ten methods in the literature the value of @xmath49 for @xmath44 at [ fe / h]@xmath50 . while there is considerable evidence that @xmath44 is a nonlinear function of [ fe / h ] , over the [ fe / h ] range of halo rrls ( @xmath51 } \lesssim -1 $ ] ) , most of the recently proposed @xmath52 [ fe / h ] relationships can be closely approximated by a linear dependence with a slope of @xmath53 @xcite . we have therefore adopted the relationship : @xmath54}+1.5)+(0.59\pm 0.03)\ ] ] using the above @xmath52 [ fe / h ] relationship , we can investigate the distance errors introduced by the adoption of one value of @xmath7 for all rrls . an estimate of the likely distribution of [ fe / h ] among the quest rrls is provided the spectroscopic observations by @xcite of 113 halo rrls lying at galactocentric distances greater than 8.5 kpc . the distribution of these measurements is approximately gaussian and has a mean value of [ fe / h]@xmath55 and a standard deviation , corrected for measuring errors , of 0.30 dex . these values are on the @xcite metallicity scale for globular clusters , which despite its age agrees well with recent determinations based on high dispersion spectroscopy @xcite . this metallicity dispersion and the uncertainties in the @xmath44 relation produce @xmath56 . when combined with our estimate of 0.08 for the dispersion in @xmath45 , we obtain @xmath57 . using the above estimates of the errors in the mean magnitudes and interstellar extinctions , we obtain 0.15 for the @xmath58 uncertainty in the true distance modulus for a typical rrls in the quest survey . this translates into a fractional error ( @xmath59 ) of only 0.07 . the distances to most other halo tracers have larger fractional errors , which for k giants @xcite m giants @xcite , and main - sequence stars @xcite are larger by factors of @xmath60 . only blue horizontal branch ( bhb ) stars , with fractional error estimates from @xmath61 @xcite , are similar to the rrls in the precision of their distance estimates . but bhb stars must be observed spectroscopically before they can be reliably separated from other types of blue stars . for determining the distances of the quest rrls , we have adopted , as we have previously ( e.g. * ? ? ? * ) , @xmath62 , which is consistent to 0.01 mag . with the above @xmath63[fe / h ] relation and the mean halo metallicity measured by @xcite . figure [ fig - sky ] shows a polar plot of the right ascension and extinction corrected mean magnitudes @xmath64 , of the rrls in the survey . the dotted circles indicate the values of @xmath64 corresponding to @xmath658 , 19 and 49 kpc . this simple plot shows that the distribution of rrls far from the sun is not uniform . particularly notable is the group of stars at @xmath66 , @xmath67 , which was reported earlier in @xcite . this feature , which is undoubtedly tidal debris from the sgr dsph galaxy , is described in more detail later along with other density enhancements . the positions of the rrls in a galactic cartesian system have been calculated from their galactic coordinates , @xmath68 and @xmath69 , and their heliocentric distances of the rrls using the equations : @xmath70 in this coordinate system , the @xmath71-plane coincides with the galactic plane , with the line from the galactic center ( gc ) to the sun defining the the @xmath72-axis . positive @xmath72 values are on the sun s side of the gc and , the positive @xmath73 axis is in the direction of @xmath74 . coordinate @xmath75 is positive towards the north galactic pole . a value of 8 kpc was adopted for @xmath76 , the distance of the sun from the gc @xcite . galactocentric distances are given by @xmath77 table [ tab - rrgc ] contains the galactic coordinates ( @xmath68 , @xmath69 ) , extinction corrected v magnitude ( @xmath64 ) , galactocentric coordinates ( @xmath72 , @xmath73 , @xmath75 ) and heliocentric and galactocentric distances , @xmath32 and @xmath10 . the i d numbers of the stars are the same as in table 2 of paper i. the full version of table [ tab - rrgc ] is available only on - line as a machine - readable table . although figure [ fig - sky ] and several other recent halo surveys ( see 1 ) have shown that the halo does not have smooth density contours , it is still useful to use this approximation to estimate the `` background '' of halo rrls upon which density enhancements , such as the one described above , are overlaid . in this spirit , we have performed the following analysis . any determination of the space density of rrls requires a careful consideration of the survey s completeness . in paper i we showed that the completeness of the rr_c _ variables is significantly lower than that of the rr_ab _ , especially at the faint end of the survey . the sample of rr_c _ may have also a small , but unknown , contamination from eclipsing binaries and variable blue stragglers ( see paper i ) . for these reasons , we have used only the rr_ab _ to estimate the density distribution . as we explained above , we do not use stars in the region with @xmath78 hrs because of the large interstellar extinction . finally , since the saturation limit of the different ccds in the quest camera is not uniform ( see paper i ) , we eliminated stars with @xmath79 . this ensures that all distances are observed through the same solid angle . the space density of rrls is calculated using a list of 334 rr_ab _ stars spread over @xmath80 deg@xmath21 of the sky . we followed @xcite for calculating the number density of rrls as a function of galactocentric distance @xmath10 . in this method , the total number of objects @xmath81 found in a solid angle @xmath5 is @xmath82 solving for the space density of rrls , @xmath83 , @xmath84 the quantity @xmath85 is estimated from a plot of the cumulative number of rrls versus @xmath32 . first , all stars are sorted by increasing @xmath32 . then , for each star @xmath86 at a distance @xmath87 , @xmath88 is the local slope of the curve , which is calculated by fitting a straight line to the 3 contiguous points centered at @xmath87 . finally , distances @xmath32 are transformed to galactocentric distances @xmath10 using equation ( [ eq - rgal ] ) . this way , it is possible to calculate a space density at the @xmath10 of each star ( except for the ones at the extreme @xmath32 s ) . as noted by @xcite this method works well for small solid angles , where a heliocentric distance corresponds to a unique galactocentric distance . our survey has a big solid angle which covers a large range in galactic latitude and longitude , and saha s method will not work for such a region . @xcite proposed a variation of this method which was applied to their driftscan survey , similar in shape to ours , to calculate space densities based on a model of a spherical halo or a ellipsoidal one . we are interested here in studying the density profile of the halo along different lines of sight in order to see differences as a function of location in the galaxy . consequently , we divided our survey into small sub - regions , where it was possible to use saha s procedure . the size of the sub - regions must be large enough to contain a sufficient number of stars for the calculations , but small enough to cover only a small range in galactic coordinates . we divided our @xmath89-wide strip in 19 pieces of equal area . each piece is 0.5 hr long in right ascension for an area of 16.5 deg@xmath21 ( the area covered by the gaps between ccds has been subtracted ) . the number of rr_ab _ in each of these slices of sky varies from 5 to 33 . figure [ fig - los ] shows the line of sights of each of our sub - regions on the sky . the length of the arrows indicate the range in distance from the sun covered in each sub - region which was calculated from the limits in v magnitude of the survey and the extinction in that particular direction . table [ tab - regions ] has the coordinates and range of @xmath10 covered in each sub - region . for comparison figure [ fig - los ] also shows , at the same scale , the lines of sight and depths of previous surveys of rrls aimed to calculate the space density of the halo : the lick survey @xcite , the palomar - groningen survey @xcite , saha s survey @xcite and hawkins survey @xcite . the area covered by each line of sight in those surveys ( all photographic ) varies from 16 to 43.6 deg@xmath21 and each one has different degrees of completeness . the details of each survey are summarized in @xcite ( see also table 4 - 2 in @xcite ) . one of the lick s fields , mwf 361 or rr i , which has been extensively studied @xcite , has partial overlap with our sub - region @xmath90 . because of its special shape and characteristics , figure [ fig - los ] does not include the cti survey of @xcite . the survey is similar to ours in the sense that it is a driftscan survey . it covers a very narrow strip ( @xmath91 ) over all right ascensions for a total area of 36 deg@xmath21 , up to a distance @xmath32 of 30 kpc . also , we do not show the rr lyrae candidate searchs made with data from the sdss @xcite because their completeness are relatively low . the line of sights searched by @xcite have similar direction than ours ( but longer ) in the range of @xmath92 . figure [ fig - los ] shows that our survey covers a wide area of the outer parts of the halo where few rrls have been discovered . of the previous surveys that identified rrls by light curve and period , only saha s and hawkins surveys ventured before in the @xmath93 kpc region of the halo . notice that previous surveys were mostly made toward 3 directions : the galactic center , the galactic anti - center and the north galactic pole . our survey provides a homogeneous sample of rrls in a wide range of galactic coordinates . in order to correct for incompleteness , the space density given by equation [ eq - rho ] was divided by the completeness of the survey in that part of the sky . the completeness for the rr_ab _ in the quest survey is high ( @xmath94 ) over most of the region but decreases toward the ends of the strip . the completeness was calculated using extensive simulations as a function of @xmath25 , separately for bright ( @xmath95 ) and faint stars , and it is shown in figure 10 of paper i. small corrections have been made to these completeness estimates because we removed the nonvariables mentioned in secton 2 the number density of rrls as a function of @xmath10 along the different lines of sight is shown in figure [ fig - density ] . we have plotted also the space densities of rrls in lick field `` rr i '' , as determined by @xcite , in the panel of @xmath96 . the fact that we get similar results in the range of distances of overlap give us confidence that our completeness estimates are correct . @xcite used a different method to estimate densities , which basically consists of calculating the volume occupied by a fixed number of stars of continuously increasing @xmath32 . the most striking feature of the density profiles of figure [ fig - density ] is the big over - density of rrls at @xmath97 kpc in the regions @xmath98 , which is due to the sgr tidal tail ( see also figure [ fig - sky ] ) . traditionally , the halo has been pictured as a region where the density contours smoothly vary with @xmath10 as a power law : @xmath99 where @xmath100 is the local space density of rrls . the exponent @xmath101 has been determined by @xcite by combining the results of several surveys of rrls and has a value of @xmath102 . they fitted the power law using data with @xmath10 from 0.6 to 80 kpc , with the caveat that only 9 objects with @xmath103 kpc were available . @xcite have determined a value of @xmath104 kpc@xmath105 with rr_ab _ stars from the first results of the rotse all - sky survey . the power law with the above values is shown in the density profiles of figure [ fig - density ] as a reference . in principle , we do not expect a perfect fit of the data to a power law not only because of statistical fluctuations but also because the outer halo may be filled with sub - structures , as predicted by hierarchical models of galaxy formation ( eg . @xcite ) . in fact there are several large regions of unexpectly high density in our survey , which are seen in figure [ fig - density ] as large departures from the solid lines at certain distances @xmath10 . ignoring these over - densities , our first model , a spherical halo with a power - law density fall - off , does a fair job describing the density profiles along the different lines of sight , except in the regions with @xmath106 h for which it over - estimates the space density of rrls . as seen in figure [ fig - los ] and table [ tab - regions ] these sub - regions have lines of sights approaching the galactic center and thus , they contain several objects at small @xmath10 . since this power law reproduces well the space densities at larger @xmath10 , there is either a problem with the parameters of the power law or the assumption of an spherical halo . results from previous surveys toward the galactic center direction led several groups to propose density contours that are flatter in the inner halo @xcite . @xcite constructed an empirical model of a halo with variable flattening which fit simultaneously rrls density profiles toward the north galactic pole and the gc . the model consists of isodensity contours described by ellipsoids of revolution of semi - major axis @xmath107 , whose flattening @xmath108 is given by @xmath109(a/\!a_u ) & \mbox{if $ a < a_u$ } \\ 1 & \mbox{if $ a > a_u$ } \end{array } \right . \label{eq - ca}\ ] ] with @xmath110 and @xmath111 kpc . consequently , at large distances from the gc , this model has spherical density contours . our sample of rrls provides a test of models with different halo shapes and whether they can simultaneously fit the data along all lines of sight . we tested 3 different models of the halo : a spherical one ( shown in figure [ fig - density ] ) , a flattened halo with constant @xmath112 , and @xcite s model with variable flattening . in the non - spherical models , the density contours are described by @xmath113 where @xmath114 to see which model best describes our data , we calculated the standard deviation of the data points about the power law in each sub - region . for these calculations we left out the the most obvious over - densities of rrls in figure [ fig - density ] . specifically , we did not include any star with @xmath115 kpc in the range @xmath116 , because they are known to belong to the sgr stream . we also took out some other stars causing notable over - densities : 3 stars at @xmath117 and @xmath118 ; 5 stars at @xmath119 and @xmath120 ; and the 5 known variables of the globular cluster pal 5 at @xmath121 and @xmath122 . except for the sgr and pal 5 stars , the elimination of the other over - densities made no significant differences in the results we describe here . with the parameters @xmath123 kpc@xmath105 and @xmath124 the three models produce the dispersions that are plotted as functions of right ascension in figure [ fig - model ] . as expected from figure [ fig - density ] , the spherical halo ( filled circles ) produces large deviations at @xmath106h . on the other hand , a halo with large , constant flattening ( open triangles ) yields a good fit of the data in the regions toward the gc ( @xmath125 ) , but it does not reproduce the observed density profile at high galactic latitudes ( @xmath126 , see figure [ fig - badmodel]a ) . in those regions , the power - law tends to underestimate the observed density of rrls . the model with varying flattening yields the best fit since it produces small and similar standard deviations over all lines of sight ( see figure [ fig - newdensity ] ) . in particular , this model provides a better description of the data than the spherical model at @xmath127 ( cf . figures [ fig - density ] and [ fig - newdensity ] ) . the values of @xmath101 and @xmath100 used so far were both determined by assuming a spherical halo , and they may not be optimal for the other models . for example , @xcite found a steeper power law ( @xmath128 ) when the variable flattening model was applied to their data . also , a slightly higher value of the local density ( @xmath129 kpc@xmath105 ) is found by counting rrls very close to the sun , instead of averaging the value of the density at @xmath130 in several directions of the galaxy @xcite . therefore , we repeated the experiment with these other plausible density power laws and the best result was always found with the variable flattening model . for example , a steeper power law ( @xmath131 ) does not describe well the region of @xmath132 with any model ( figure [ fig - badmodel]b ) unless the local density takes a much higher value , which is then unreasonable for all other lines of sight and inconsistent with published measurements of that quantity . we used the data over all the regions to derive the best parameters for a power law with all the tested models . the densities along all lines of sight were averaged in bins of equal size in @xmath133 ( or @xmath134 in the case of the spherical model ) and fit by the method of least squares . the results are presented in table [ tab - fit ] . on average , both the flattened halo model and the variable flattening one produced the best overall fit to our data ( lowest rms ) . however , we favored the variable flattening model because it best reproduces both the local density found by @xcite , and the slope @xmath101 from previous works , including other tracers ( @xmath135 for globular clusters @xcite and bhb @xcite ) . notice that 52 stars with @xmath103 kpc were used in the fit , @xmath1366 times more than the number used in previous studies @xcite . in figure [ fig - avden ] , the densities averaged over all lines of sight are compared with the best fitting profile of the variable flattening model . the fit is reasonable , but with a larger than expected number of points deviating by more than one standard deviation . this may be due to the failure of the model to account for the lumpy nature of the halo and the variations in the specific frequency of rrls . nonetheless , this model serves our purpose of providing a description of the `` background '' of rrls . there is no evidence for a steeper power law ( @xmath137 ) beyond 25 kpc as was suggested by @xcite . the underdensity that he observed in one of his fields may be simply due to the clumpy nature of the outer halo . this might also be the explanation for the observation by ivezic et al ( 2000 ) of an edge to the halo at 50 - 60 kpc ( see ivezic et al . we have performed the following analysis to identify overly dense regions that may escape detection by visually examining figure [ fig - sky ] . many of these feature need to be confirmed by radial velocity and/or proper motion measurements . we have restricted this analysis to stars with @xmath138 because at brighter magnitudes the differences in the saturation of limits of the ccds in the quest camera produce some variation in completeness . the first step in our analysis was the development of a computer code to recognize groups of rrls that may be statistically significant . for each star ( @xmath86 ) in the quest catalogue , the code computed the distances ( @xmath139 ) from @xmath86 to each of the other stars @xmath140 and kept as neighbors the stars within a distance limit ( @xmath141 ) . it then created a ranking of the @xmath139 values of the neighbors and computed the volumes centered on star @xmath86 that included in turn each of its more distant neighbors . because the quest survey is limited in @xmath142 to @xmath143 , these volumes are non - spherical , and we approximated them with cylinders with heights in the @xmath142 direction . each volume contains star @xmath86 , the star @xmath140 at distance @xmath139 and all stars with smaller distances from star @xmath86 , if any . the number of stars observed in a volume , @xmath144 , was then compared with the number expected in the absence of an overdensity . our model for the density contours ( see above ) provides the number density of rr_ab _ variables at the position of star @xmath86 . these densities were then multiplied by factors that take into account the incompleteness of survey in type ab and in type c variables , which vary with @xmath25 and @xmath33 ( paper i ) . the resulting number density for all types of rrls was then multiplied by the volume to yield @xmath145 , the expected number of rrls , which was rounded to the nearest integer . we then computed the probability , @xmath146 , that @xmath147 , given that @xmath81 is a random number following a poisson distribution of mean @xmath145 , @xmath148 star @xmath86 and each of its neighbors that occupy a volume where @xmath149 are considered members of a group . if none of these stars has been previously assigned to a group , then a new group number is assigned . otherwise , they are considered additional members of the previously assigned group . in addition to group number , star @xmath86 is tagged by the lowest @xmath149 that occurred in the calculations of the values of @xmath146 for the small volumes . the free parameters in this analysis are the choices for @xmath141 and @xmath150 . for @xmath141 , we chose @xmath151 , which is approximately @xmath152 . consequently , two stars that lie at the same distance and direction have a probability of @xmath153 of being considered neighbors in this analysis . increasing @xmath141 significantly above this value produces overlap between what may be separate over - densities . decreasing @xmath141 significantly may mean that some small over - densities are missed because of distance errors . for @xmath150 , we chose the value @xmath154 , which at first glance may appear smaller than necessary . because there are 457 stars in the revised quest catalogue and a minimum of a few stars within @xmath141 of each , thousands of calculations of @xmath146 are made . it is fairly common for random fluctuations alone to produce a few small groups with @xmath155 , as our monte carlo simulations indicate ( see below ) . in figure [ fig - overdensities ] , we have used different symbols to indicate the confidence that we attach to a star s membership in an overdensity . values of @xmath146 in the ranges @xmath156 , @xmath157 , and @xmath158 indicate with low , medium , and high confidence , respectively , that a star is considered part of an overdensity . we must emphasize that some of the stars identified as belonging to a group will turn out to be clearly nonmembers once their motions are measured because every overdensity is overlaid on a `` background '' of unrelated rrls . we are therefore identifying stars that are candidate members of the groups , and their @xmath146 values provide an estimate of relative likelihood of membership . the sparse globular cluster pal 5 lies within the survey region ( see figure [ fig - sky ] ) . not surprisingly , its five rrls , which were all detected by the quest survey , produce a very significant overdensity which is not of interest because the cluster has been known for decades . we therefore removed these 5 stars from the survey before performing the search for overdensities . one of the identified overdensities is in close proximity to pal 5 and is undoubtedly related ( see below ) . for each overdensity recognized by the above procedure , the quantity @xmath159 was calculated by averaging the @xmath146 values that were assigned each member of the group . the above limits on @xmath146 have also been applied to @xmath159 in order to identify groups of low , medium , and high confidence . the groups that seem most likely to be real and their confidence category are listed in table [ tab - overden ] , where also are listed the position of the group in figure [ fig - overdensities ] , the number of members ( @xmath144 ) , @xmath159 , and the frequency ( @xmath160 ) with which a similar group occurred by chance in the monte carlo calculations . @xmath160 is equal to the total number of groups with @xmath161 and average @xmath162 that were produced by the simulated surveys divided by the total number of artificial surveys ( @xmath163 ) . in the monte carlo simulations , the volume of the galactic halo covered by the quest survey was subdivided by @xmath25 into intervals of 0.25 hours ( @xmath164 ) and by @xmath165 into intervals of 0.25 mag . no subdivision was made by @xmath142 . the position in the galaxy of the center of each subdivision was computed using @xmath166 for rrls , and the number density of rr_ab _ at this position was calculated from our model for the background of type ab rrls . this number density was corrected for the incompleteness of the quest survey using the same factors that were used with the real survey . it was then multiplied by the volume delineated by the intervals in @xmath25 , @xmath142 , and @xmath165 . after rounding to the nearest integer , this yielded @xmath145 for the volume . to provide for statistical variation , we computed a random number ( @xmath81 ) from a poisson distribution ( a poisson deviate ) with mean equal to @xmath145 , which employed a random number as a seed variable . these @xmath81 artificial stars were then randomly distributed in @xmath25 , @xmath142 , and @xmath165 within the limits of the subdivision using random a number generator with different seed values for each quantity . the calculation then proceeded to the next subdivision where @xmath145 was again calculated for the new position and different random selections were made for the seed variables in the calculation of the poisson deviate and in the distribution of the stars in position and in magnitude . these calculations were continued until the area of the quest survey was covered . the program for identifying overdensities that we described above was then used to find and characterize any groups that were produced in this artificial survey . in order to see what size groups and their @xmath159 values that might have arisen by chance in the quest survey , a total of @xmath163 artificial surveys were produced by starting each monte carlo calculation with a different seed value . each of these surveys was fed to the program for finding overdensities and the values of the frequencies @xmath160 ( see above ) were computed . experiments with different sizes for the intervals in @xmath25 and @xmath165 showed that these monte carlo calculations were not particularly sensitive to their choices as long as the intervals were not so large to encompass a significant variation in number density and not so small that @xmath167 for many subdivisions . one of these artificial surveys , which is typical in terms of the number and sizes of overdensities , is displayed in figure [ fig - mc ] . the artificial stars that are in overdensities are labeled in exactly the same manner as the stars in rrls in the real survey . since the same model for the smoothed density contours was used both to generate the artificial survey and to identify the overdensities , the ones that appear in the artificial surveys are entirely due to statistical fluctuations . once can see from this figure that random variations alone can produce significant overdensities , even ones that meet our criteria for the high confidence category . not surprisingly , the majority of the randomly occurring overdensities contain very few stars . table [ tab - overden ] lists the 6 most significant overdensities in the quest survey according to our group finding routine . with the exception of group 2 , we will discuss each of these groups in order of their sizes . group 2 lies in the region of the survey where the interstellar extinction is high and variable . because our estimate of the completeness of the survey in this region is suspect , we consider group 2 to be only a marginal detection of halo substructure . the largest and by far the most significant over - density is group 5 , at a mean distance of 50 kpc , which was produced by the tidal destruction of the sgr dsph galaxy . according to our group finding routine , group 5 contains 104 candidate members , which is @xmath168 of the whole sample of quest rrls . its @xmath160 value ( see table [ tab - overden ] ) indicates that none of the @xmath4 monte carlo simulations produced a group of this size or larger with @xmath169 . this striking overdensity was discovered in the spatial distribution of a stars @xcite and rrls candidates @xcite . shortly thereafter , a main - sequence was found in the color - magnitude diagram of a small region @xcite . it has also been detected in the distributions of carbon stars @xcite , f type main - sequence stars @xcite , and m giants @xcite . our initial results from the quest survey @xcite provided additional evidence for a large overdensity of rrls . the distributions of these stars and the measurements of radial velocities for subsamples ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) provide conclusive proof that this feature is part of the leading stream from the sgr dsph galaxy . figures [ fig - sky ] and [ fig - overdensities ] show that there are very few quest rrls at large @xmath32 that are not part of group 5 . this observation and similar ones using the different types of stars mentioned above illustrate that the debris from the sgr dsph galaxy is a major contributor to the outer galactic halo . group 4 , which after the sgr stream is the most prominent feature in figures [ fig - sky ] and [ fig - overdensities ] , contains 42 potential members . the @xmath160 value of this group ( table [ tab - overden ] ) indicates that a few of the monte carlo simulations produced artificial groups of equal or larger size and significance . based on this information alone , there is a small , but not negligible probability , that this feature is an artifact . the eastern edge of this feature was identified as an overdensity in our first report of quest results @xcite . it became known as the `` 12.4 h clump '' when its large size was recognized in the completed quest survey @xcite . the survey of f type main - sequence stars in the sdss @xcite also revealed its large size . figure [ fig - overdensities ] shows that it spans @xmath170 and @xmath171 , which corresponds to a span of @xmath172 kpc at the mean of 17 kpc . the densest part of this feature is located at @xmath173 and @xmath174 kpc . very recently , @xcite have shown that a subsample of the quest rrls in this dense region and a subsample of the bhb stars in the same region of space from the @xcite survey have very similar radial velocities , indicating that they are part of a stellar stream . the rrls in the stellar stream have a low mean metallicity but a wide range in metallicity , which suggests that they are the debris from a low luminosity dsph galaxy . @xcite have suggested the name the virgo stellar stream ( vss ) for this feature of halo substructure , which they have traced over 106 deg@xmath21 of the sky . until more radial velocity measurements are obtained , it is not clear that majority the stars of group 4 belong to the vss . @xcite noted that in addition to the vss there is less significant evidence for two other moving groups containing smaller numbers of rrls and bhb stars , which may be a sign of that there is more to group 4 than the vss . group 4 also lies in the direction of the `` virgo overdensity '' that has been recently identified by @xcite in distribution of main - sequence stars in the latest photometry of the sdss . they estimate that this feature covers @xmath175 deg@xmath21 of the sky and lies in the range @xmath176 kpc . it therefore overlaps considerably with group 4 , and considering distance errors of @xmath177 for the main - sequence stars , it also overlaps with the known members of the vss , which lie at its distant edge . @xcite suggest that this feature is the tidal stream from a dwarf galaxy or possibly the galaxy itself . while the exact relationships between the vss , the virgo overdensity , and group 4 are unclear at this moment , there is no doubt that at least part of group 4 is a real halo substructure . the next two most significant groups , groups 1 and 3 , may be related to each other . they lie at similar values of @xmath32 on either side of the region skipped by the quest survey because of the severe crowding of star images near the galactic plane . the low @xmath160 values of these groups suggest that they are likely to be real . group 1 is the less certain of the two because it lies in the range of @xmath25 where it was difficult to estimate the completeness of the survey . spectroscopic observations have been obtained for many of the stars in these groups and will be discussed in a later paper . as we noted in a preliminary report of our results @xcite , groups 1 and 3 lie approximately in the same direction and at the same distance as the monoceros stream that was discovered by @xcite in the sdss photometry and confirmed by the spectroscopic observations of @xcite . this stream appears to be part of large - ring like structure that envelops the milky way ( see also * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? it is widely interpreted as the debris of tidally disrupted dwarf galaxy , but it is clearly not part of the sgr stream nor any of the other halo overdensities that we have discussed above . while the association of groups 1 and 3 with the mon stream appears likely , this is not certain . the measurements by @xcite of the radial velocities of a sample of rrls in the anticenter direction did not reveal any that clearly belong to the mon stream . as we noted above , we purposely removed the 5 rrls in the globular cluster pal 5 from the quest catalogue before searching for overdensities . group 6 lies very close to pal 5 and at exactly the same distance , which of course suggests a relationship with the cluster . consequently , we discount the relatively large @xmath160 value of group 6 , which suggests that the quest survey is expected to contain one or two groups of the size and significance of group 6 that are nothing more than statistical fluctuations . in the top diagram of figure [ fig - pal5 ] , we have plotted the 5 previously identified rrls in pal 5 , the 6 stars in group 6 , and other quest rrls in the region . the bottom diagram of figure [ fig - pal5 ] shows the distribution of the same stars on the sky along with a rough outline of pal 5 and its tidal tails , as revealed by the analysis of sdss photometry by @xcite . one can see from the upper diagram that all 6 stars in group 6 lie at the same @xmath32 as pal 5 to within the errors . two of the group 6 stars , rrls 403 and 405 , lie approximately 9.2 and 11.2 arcmin . , respectively , from the cluster center ( see lower diagram ) . the extensive photometry of @xcite shows that there are many pal 5 stars at similar angular distances ; consequently the association of these rrls with the cluster is nearly certain . another rrls ( 393 ) , which was not identified as member of group 6 because it lies too far away from the group s center is also likely to have been once a member of pal 5 . its position on the sky and distance are consistent with it being part of the southern tidal stream from pal 5 . it is interesting to note that pal 5 has been considered odd among globular clusters because all five of its previous known rrls are type c. two of the three probable members identified here are type ab , which partially removes the imbalance in the numbers of these types of variables . it is less certain that any of the other members of group 6 are related to pal 5 , for they are offset from the cluster to the east by significant amounts . however , as reported in @xcite , our preliminary measurements suggest that the two members of group 6 near @xmath178 have radial velocities and metallicities that are consistent with membership in the cluster . this possibility is being explored as the measurements are refined . one can see from figure [ tab - overden ] that the group finding routine has identified a number of other small groups . while the @xmath160 values of these groups suggest that they may be not be real , this can not be discounted entirely until their motions are measured . @xcite have shown that small groups of co - moving rrls exist in the halo , which they suggest may be the remains of disrupted globular clusters . our spectroscopy of few of the small groups in the quest survey will be reported later . our major result is that the galactic halo contains substructures over a spectrum of sizes . while much remains to be learned about the features identified here and by other surveys , there is strong evidence that at least some were produced by the merger of smaller galaxies with the milky way . these merger events have clearly deposited a large number of stars in the halo , as predicted by models of the hierarchical picture of galaxy formation ( e.g. * ? ? ? while this may seem so obvious now , it represents a large departure from the traditional picture of the galactic halo that could be characterized by smooth density contours and a power - law drop - off . this research project was partially supported by the national science foundation under grant ast-0098428 and ast 0507364 . akv thanks bruce carney for several interesting and helpful comments and suggestions . the quest survey for rr lyrae stars uses the facilities of the observatorio nacional de llano del hato , which is operated by cida for the ministerio de ciencia y tecnologa , venezuela . we thank the anonymous referee for encouraging us to discuss about the specific frequency of rrls in different populations . kinemuchi , k. , smith , h. a. , lacluyz , a. p. , clark , c. l. , harris , h. c. , silbermann , n. & snyder , l. a. 2002 , in radial and nonradial pulsations as probes of stellar physics , eds . c. aerts , t. r. bedding and j. christensen - dalsgaard , asp conf . series , 259 , 130 lccrrc carina & @xmath179 & -2.0 & 75 & 76 & @xmath180 + draco & @xmath181 & -2.0 & 268 & 268 & @xmath182 + fornax & @xmath183 & -1.3 & 515 & 660 & @xmath184 + leo i & @xmath185 & -1.5 & 74 & 74 & @xmath186 + leo ii & @xmath187 & -1.9 & 148 & 148 & @xmath188 + sagittarius & @xmath189 & -1.0 & 1700 & 5200 & @xmath190 + sculptor & @xmath191 & -1.8 & 226 & 500 & @xmath192 + sextans & @xmath193 & -1.7 & 111 & 260 & @xmath194 + ursa minor & @xmath195 & -2.2 & 82 & 84 & @xmath196 + and i & @xmath197 & -1.5 & 99 & 582 & @xmath198 + and ii & @xmath199 & -1.5 & 72 & 400 & @xmath200 + and iii & @xmath201 & -1.9 & 51 & 204 & @xmath202 + and vi & @xmath203 & -1.6 & 111 & 529 & @xmath204 + [ tab - frequency ] lrrcrrrrr 217 & 299.05 & 61.87 & 16.23 & 4.9 & -5.7 & 12.1 & 13.7 & 14.2 + 218 & 299.38 & 60.74 & 14.53 & 6.5 & -2.7 & 5.4 & 6.2 & 8.9 + 219 & 299.42 & 61.98 & 14.79 & 6.4 & -2.9 & 6.2 & 7.0 & 9.4 + 220 & 300.00 & 60.95 & 18.96 & -3.7 & -20.2 & 42.0 & 48.0 & 46.8 + 221 & 300.19 & 61.79 & 16.43 & 4.4 & -6.1 & 13.2 & 15.0 & 15.3 + [ tab - rrgc ] crrc 4.5 - 5.0 & 198.2 & -28.3 & 15 - 70 + 8.0 - 8.5 & 223.7 & 18.0 & 14 - 69 + 8.5 - 9.0 & 227.7 & 24.5 & 14 - 70 + 9.0 - 9.5 & 232.2 & 30.9 & 13 - 69 + 9.5 - 10.0 & 237.3 & 37.1 & 13 - 66 + 10.0 - 10.5 & 243.3 & 43.0 & 13 - 67 + 10.5 - 11.0 & 250.5 & 48.6 & 12 - 65 + 11.0 - 11.5 & 259.5 & 53.5 & 11 - 64 + 11.5 - 12.0 & 270.7 & 57.6 & 11 - 66 + 12.0 - 12.5 & 284.2 & 60.5 & 10 - 65 + 12.5 - 13.0 & 299.5 & 61.8 & 10 - 63 + 13.0 - 13.5 & 315.3 & 61.3 & 9 - 62 + 13.5 - 14.0 & 329.7 & 59.1 & 8 - 62 + 14.0 - 14.5 & 341.9 & 55.4 & 7 - 60 + 14.5 - 15.0 & 351.7 & 50.8 & 7 - 59 + 15.0 - 15.5 & 359.6 & 45.5 & 6 - 55 + 15.5 - 16.0 & 6.1 & 39.7 & 5 - 51 + 16.0 - 16.5 & 11.6 & 33.6 & 5 - 50 + 16.5 - 17.0 & 16.3 & 27.3 & 4 - 45 + [ tab - regions ] cccccccccclcl 1 & 6 & 72 & -1.0 & 199 & -28 & 15.6 & 10.2 & 17.5 & @xmath211 & high & 0.0090 & mon stream ? + 2 & 7 & 82 & -0.9 & 204 & -19 & 16.9 & 18.8 & 26.0 & @xmath212 & low & 1.1521 & + 3 & 12 & 128 & -1.3 & 226 & 22 & 15.7 & 10.5 & 16.8 & @xmath213 & medium & 0.0101 & mon stream ? + 4 & 42 & 189 & -0.8 & 296 & 62 & 16.7 & 17.0 & 17.2 & @xmath214 & low & 0.0032 & virgo feature + 5 & 104 & 214 & -1.1 & 342 & 55 & 19.0 & 50.2 & 46.5 & @xmath215 & high & 0.0000 & sgr stream + 6 & 6 & 230 & -0.5 & 2 & 45 & 17.3 & 22.3 & 17.6 & @xmath216 & low & 1.3848 & close to pal 5 + [ tab - overden ]

we show that anderson localization in quasi - one dimensional conductors with ballistic electron dynamics , such as an array of ballistic chaotic cavities connected via ballistic contacts , can be understood in terms of classical electron trajectories only . at large length scales , an exponential proliferation of trajectories of nearly identical classical action generates an abundance of interference terms , which eventually leads to a suppression of transport coefficients . we quantitatively describe this mechanism in two different ways : the explicit description of transition probabilities in terms of interfering trajectories , and an hierarchical integration over fluctuations in the classical phase space of the array cavities .

0802.0976

the interplay of quantum phase coherence and repeated random scattering is at the origin of many effects in mesoscopic physics.@xcite these effects include weak localization and universal conductance fluctuations , both of which are small but fundamental corrections with respect to the conductance obtained from drude - boltzmann theory . they culminate in anderson localization , the phenomenon that the resistance of a one or two - dimensional electron gas grows exponentially with system size if the system size is sufficiently large.@xcite originally , anderson localization and other mesoscopic effects were discovered in the context of disordered metals , in which electrons scatter off impurities with a size comparable to their wavelength . theoretically , quantum effects in disordered metals are described using the ` disorder average ' which deals with an ensemble of macroscopically equivalent but microscopically different impurity configurations . the presence of impurities is not essential for the existence of quantum effects , however . the same effects , with the same statistical properties , have been found to appear if the electron motion is ballistic and chaotic , the only source of scattering of electrons being specular reflection off the sample boundaries.@xcite besides being of theoretical interest for understanding the quantum properties of systems with chaotic classical dynamics,@xcite the case of ballistic electron motion is relevant experimentally for very clean artificially structured two - dimensional electron gases in semiconductor heterostructures , such as quantum dots or antidot lattices.@xcite unlike disordered metals , in which impurities scatter diffractively , electrons in ballistic conductors have a well - defined classical dynamics . many quantum properties of ballistic conductors with chaotic classical dynamics can be understood in terms of classical mechanics by making use of well - chosen semiclassical methods.@xcite a well - known example is the gutzwiller trace formula , which relates the density of states to properties of periodic orbits of the classical dynamics.@xcite for the conductance @xmath0 and its quantum corrections , there is a variant of gutzwiller s formula , which expresses @xmath0 as a double sum over classical trajectories @xmath1 and @xmath2 connecting the source and drain contacts,@xcite @xmath3 upon entrance and exit , the trajectories @xmath1 and @xmath2 have the same transverse momenta @xmath4 and @xmath5 , respectively , but the position at which they enter or exit the sample may be different . further , @xmath6 and @xmath7 are the so - called `` stability amplitude '' and the classical action of the trajectories . with eq . ( [ eq : gsemi ] ) as a starting point , the conductance , including its quantum corrections , can be calculated solely from knowledge of classical trajectories . quantum effects have been linked to the existence of families of classical trajectories that only differ near ` small - angle encounters ' of the trajectory with itself.@xcite this semiclassical approach has been very successful explaining quantum effects in the ` perturbative regime ' in which quantum interference provides a small correction to the classical mechanics.@xcite weak localization and universal conductance fluctuations are examples of such perturbative quantum effects . recently , heusler _ et al . _ showed that it is also possible to understand quantum effects in terms of classical trajectories outside the perturbative regime . they considered the spectral form factor @xmath8 of a ballistic cavity , the fourier transform of the two - point correlation function of the density of states . a direct semiclassical evaluation of @xmath8 using gutzwiller s trace formula was known to be possible in the perturbative regime @xmath9 only , where the heisenberg time @xmath10 , @xmath11 being the cavity s mean level spacing.@xcite motivated by the field - theoretical formulation of the problem , heusler _ et al . _ used a different way to express @xmath8 in terms of periodic orbits that allowed them to calculate @xmath8 for all times.@xcite in this communication , we consider anderson localization for ballistic electron gases , and show that it , too , can be understood in terms of interference of classical trajectories , with eq . ( [ eq : gsemi ] ) as a starting point . examples of systems that may exhibit such ` ballistic anderson localization ' are antidot lattices or arrays of chaotic cavities . although it is generally accepted that anderson localization exists irrespective of the details of the microscopic electronic dynamics,@xcite the similarity of the phenomena for ballistic and disordered electron systems should not obscure the vastly different starting points of the theories for the two cases . this difference not only pertains to the microscopic dynamics ( classical - deterministic vs. quantum - probabilistic ) , but also to the statistical assumptions of the theory . unlike theories of quantum transport in disordered metals , semiclassical theories of ballistic conductors are intended to describe one specific system.@xcite fluctuations appear solely from variations of the fermi energy ; no changes in the classical dynamics are invoked . for disordered metals , anderson localization is most prominent for a quasi - one dimensional geometry , with sample length @xmath12 much larger than the sample width @xmath13 . for quasi - one dimensional samples , a full theory of transport in the localized regime was developed by dorokhov@xcite and mello , pereyra , and kumar,@xcite using a stochastic approach , and by efetov and larkin,@xcite using a field - theoretic approach . our theory of ballistic anderson localization closely follows these approaches . the fact that the semiclassical theory for ballistic electrons follows the corresponding quantum mechanical theory for disordered metals is not special to the present case . it is also typical of the trajectory - based semiclassical theories in the perturbative regime , which have a structure that resembles the diagrammatic perturbation theory of quantum corrections in disordered metals.@xcite in addition to the ballistic electron gases considered here , anderson localization also occurs in certain dynamic systems with a periodic time - dependent hamiltonian , such as the kicked rotor.@xcite classically , the dynamics of the kicked rotor is chaotic , with a momentum that changes diffusively under the influence of the periodic kicks . quantum mechanically , it exhibits ` dynamic localization ' , anderson localization in momentum space rather than in real space.@xcite the phenomenology of dynamic localization is equal to that of anderson localization in disordered metals . this is confirmed by extensive numerical simulations,@xcite as well as a field theoretical analysis of the problem.@xcite the trajectory - based theory of ballistic anderson localization that we report here is constructed for a specific system , an array of chaotic cavities . a schematic drawing of such an array is shown in fig . [ fig:1 ] . we use this model system because the trajectory - based theory of transport through a single chaotic cavity is well established in the literature.@xcite arrays of cavities have been used as a starting point for a field - theoretic description of anderson localization using random matrix theory,@xcite but not with dorokhov s method . in sec.[sec : dis ] we show how dorokhov s theory can be adapted to this system if the cavities are disordered and random matrix theory can be used to describe transport through a single cavity . in sec . [ sec:2 ] we summarize the basic elements of the trajectory - based semiclassical formalism . this formalism is used to construct the trajectory - based semiclassical theory of ballistic localization in sec . [ sec : semi ] . in section [ sec : field_theory ] we approach the localization phenomenon from a different perspective . we describe the array in terms of a nonlinear @xmath14-model whose perturbative ( ` diagrammatic ' ) evaluation generates structures of paired feynman amplitudes similar to those appearing in the native semiclassical approach . alternatively , the dynamical structure of the theory can be analyzed to identify and successively eliminate hierarchies of different types of dynamical fluctuations in the system . in this way we obtain an effective low energy theory which turns out to be equivalent to the nonlinear @xmath14-model of diffusive quantum wires ; the latter model is known to predict exponential localization at large length scales . we conclude in sec . [ sec : concl ] . ( color online ) schematic drawing of an array of @xmath15 chaotic cavities . the cavities are connected via ballistic contacts with dimensionless conductance @xmath16 . the semiclassical theory requires the limit @xmath17 at a fixed ratio @xmath18 . ] we now describe how dorokhov s approach can be used to describe anderson localization in an array of disordered cavities or quantum dots . we take the disorder in each cavity to be weak ( cavity size much smaller than the localization length ) , so that transport through a single cavity is described by random matrix theory.@xcite a schematic drawing of the array of cavities is shown in fig.[fig:1 ] . the cavities are connected via ballistic contacts with dimensionless conductance @xmath16 . since we want to compare with a semiclassical theory for an array of ballistic cavities , we require @xmath19 . localization takes place if the conductance of the array is of order unity . this condition is met if the number @xmath15 of cavities in the array is comparable to @xmath16 . hence , in the calculations below we take the limit @xmath20 while keeping the ratio @xmath18 fixed . the same limit is taken in the field theoretical description of localization,@xcite where it is known as the `` thick wire limit '' . the transport properties of the array of cavities are encoded in its scattering matrix @xmath21 . ( the superscript `` q '' is used to distinguish the quantum mechanical scattering matrix from its semiclassical counterpart to be introduced in sec . [ sec:2 ] . ) the matrix indices of @xmath21 represent the two contacts @xmath22 at the far left and right of the array and the transverse modes in each contact , @xmath23 , @xmath24 , where @xmath25 is the number of channels in contact @xmath26 , @xmath22 . the matrix @xmath21 is a random quantity because it depends on the fermi energy and on the precise disorder configuration in each cavity . following the approach of dorokhov , mello , pereyra , and kumar,@xcite we consider the hermitian matrix @xmath27 and calculate its statistical distribution by expressing @xmath28 in terms of @xmath29 and proceeding recursively . the matrix @xmath30 is related to the dimensionless conductance @xmath31 of the array through the landauer formula , @xmath32 taking the scattering matrix of each individual cavity from the circular ensemble of random matrix theory,@xcite one then finds that the recursion relation for @xmath30 takes the form @xmath33 where the hermitian matrix @xmath34 is a random ( noise ) term with a gaussian distribution , @xmath35 where @xmath36 , @xmath37 or @xmath38 in the presence or absence of time - reversal symmetry , respectively , and @xmath39 the averaging brackets @xmath40 denote an average with respect to the disorder configuration in the @xmath15th cavity only . in the limit @xmath20 while keeping @xmath18 fixed , the stochastic recursion relation ( [ eq : ttdiff ] ) can be mapped to the dorokhov - mello - pereyra - kumar ( dmpk ) equation , which is a stochastic differential equation for the eigenvalues of @xmath30.@xcite the solution of the dmpk equation is known,@xcite which completes the theory of localization for an array of disordered cavities . alternatively , the stochastic recursion relation ( [ eq : ttdiff ] ) can be used to generate a coupled set of recursion relations for the disorder averages of the moments of @xmath41 , @xmath42 with @xmath43 ( the argument @xmath44 is suppressed on the right hand side of the second equality . ) the average in eq . ( [ eq : tgeneral ] ) is the full disorder average , applied to all cavities in the array . taking the limit @xmath20 at fixed @xmath18 , eq . ( [ eq : tgeneral ] ) is mapped to a coupled set of differential equations for the moments of @xmath45 which is identical to the corresponding set of differential equations for a disordered wire.@xcite a subset of these equations can be resummed into a partial differential equation for the generating function @xmath46 , with@xcite @xmath47 as shown in refs . , the resulting theory of localization in quasi - one dimension is formally equivalent to that obtained from the one - dimensional nonlinear sigma model.@xcite the central object in the trajectory - based semiclassical theory of localization in an array of ballistic chaotic cavities is a semiclassical representation of the scattering matrix @xmath48 . in the semiclassical representation , the discrete transverse momenta become continuous variables , so that the scattering matrix @xmath48 becomes a ` scattering kernel ' @xmath49 . following standard semiclassical approximations , this scattering kernel is then represented as a sum over classical trajectories @xmath1 connecting contact @xmath50 to contact @xmath26,@xcite @xmath51 such that the transverse momentum of @xmath1 upon entrance and exit equals @xmath52 and @xmath53 , respectively . further , @xmath54 is the classical action of @xmath1 , and @xmath55 the stability amplitude , @xmath56 where @xmath5 is the transverse position upon entrance into the sample . maslov indices and other additional phase shifts are included in @xmath57 . because the transverse modes in the quantum mechanical formulation are linked to the absolute value of the transverse momentum , not to the transverse momentum itself , the semiclassical counterpart of the products @xmath58 and @xmath59 consist of two contributions : one in which transverse momenta in contact @xmath50 are equal and one in which the transverse momenta are opposite , @xmath60({p'},{p } ) & = & \sum_{\pm } \int_{-p_f}^{p_f } \frac{dp''}{2 \pi \hbar } s_{ij}^{\vphantom{m}}({p'},p '' ) s_{kj}^{\dagger}({p } , \pm p '' ) , \nonumber\end{aligned}\ ] ] @xmath61 ( { p}',{p } ) & = & \sum_{\pm } \int_{-p_f}^{p_f } \frac{dp''}{2 \pi \hbar } s_{jk}^{\dagger}(\pm p'',{p } ' ) s_{ji}^{\vphantom{m}}(p'',{p } ) , \nonumber \\ \label{eq : sproduct}\end{aligned}\ ] ] together , eqs . ( [ eq : ssemi ] ) and ( [ eq : sproduct ] ) specify how products of the quantum - mechanical scattering matrix and its hermitian conjugate are expressed in terms of classical trajectories . [ the `` @xmath62 '' terms in the summations were omitted from the semiclassical expression for the conductance , eq . ( [ eq : gsemi ] ) above . ] for a theory of anderson localization , we are interested in the trace of a product of alternating factors @xmath63 and @xmath64 or in the product of such traces . using the semiclassical representation ( [ eq : ssemi ] ) , a polynomial function @xmath65 that involves the alternating product of @xmath15 factors @xmath63 and @xmath15 factors @xmath64 is written as a summation over @xmath66 classical trajectories @xmath67 and @xmath68 , one trajectory for each factor @xmath63 or @xmath64 , respectively . each configuration of classical trajectories is weighed by a phase factor @xmath69 with @xmath70 building on work by richter and sieber,@xcite haake and coworkers have identified a hierarchy of families of classical trajectories @xmath71, ,@xmath72 that contribute to the average @xmath73,@xcite where the average is taken with respect to variations of the fermi energy while keeping the classical dynamics ( i.e. , the shape of the cavities ) fixed . their identification is based on the recognition that families of trajectories contribute to @xmath74 only if their total action difference @xmath75 is of order @xmath76 systematically , which happens only if the trajectories @xmath77 are piecewise and pairwise identical to the trajectories @xmath78 , @xmath79 , up to classical phase space distances of order @xmath80.@xcite trajectories that are separated by larger phase space distances have typical action differences @xmath75 that are parametrically larger than @xmath76 , so that their contribution vanishes upon taking the average . the simplest choice for a family of trajectories for which @xmath81 is of order @xmath76 systematically is if each trajectory @xmath77 equals another trajectory @xmath78 for the full length of the trajectory . calculating @xmath74 from the contribution from such families of trajectories only is known as the `` diagonal approximation''.@xcite nontrivial families of trajectories emerge from the possibility of small - angle encounters between trajectories , at which more than two trajectories are within a phase space distance @xmath82.@xcite at such small - angle encounters , the pairing between the @xmath77 and @xmath78 can be changed so that now trajectories need to be piecewise identical only . the duration of a small - angle encounter is the `` ehrenfest time '' @xmath83 , where @xmath84 is the lyapunov exponent of the classical dynamics , @xmath85 the fermi momentum , and @xmath86 a characteristic length scale of the classical dynamics . the fundamental action integrals corresponding to each small - angle encounter are known , @xcite and the resulting theory takes the form of simple combinatorial rules with which any product of traces of products the scattering matrix and its hermitian conjugate can be calculated to arbitrary order in @xmath76 from the semiclassical representation of @xmath63 , provided that the ehrenfest time @xmath87 be much smaller than the sample s mean dwell time @xmath88.@xcite ( the case of finite @xmath89 is considerably more complicated , see , _ e.g. _ , refs . , but not relevant for a semiclassical theory of anderson localization . ) in the remainder of this text , we refer to a calculation of the energy - average @xmath74 using contributions from families of trajectories thus constructed as the `` trajectory - based semiclassical formalism '' . although there is no formal proof that this formalism is exact , _ i.e. _ , that there are no other contributions to @xmath74 than from families of piecewise paired classical trajectories , the formalism satisfies all known conservation rules and calculations based on trajectory - based semiclassics have been found to agree with fully quantum mechanical calculations whenever applicable.@xcite the present calculation can be viewed as another demonstration of the validity of trajectory - based semiclassics , by showing that the same formalism can serve as the starting point of a theory of localization . while we do not need the detailed results of the trajectory - based semiclassical formalism , there are two properties of ensemble averages calculated using that formalism that are particularly relevant for our calculations below : + ( i ) all averages are compatible with the condition of unitarity,@xcite @xmath90(p , p ' ) & = & \sum_{j } [ s^{\dagger}_{jk } s_{ji}^{\vphantom{m}}](p , p ' ) \nonumber \\ & = & \delta_{ik } \delta(p - p').\end{aligned}\ ] ] ( ii ) for a product @xmath91 or @xmath92 , the trajectory @xmath1 of the semiclassical representation for @xmath63 and the trajectory @xmath2 of the semiclassical representation of @xmath64 at contact @xmath50 satisfy @xmath93 in particular , this implies that there is no contribution from the second term in eq . ( [ eq : sproduct ] ) for a product of two scattering kernels.@xcite + in the perturbative regime , the semiclassical theories of quantum corrections to transport and to the density of states closely followed the corresponding theory for disordered metals . the semiclassical formalism described in the previous section was instrumental in formalizing the relation between the two types of theories . motivated by this correspondence , we now look for the possibility to adapt the theory of localization in an array of disordered cavities [ sec . [ sec : dis ] ] to the case of an array of ballistic cavities . thus , paraphrasing the arguments of sec . [ sec : dis ] , the goal of our calculation will be to find the full probability distribution of the function @xmath94({p'},{p})\ ] ] for an array of @xmath15 cavities . as shown in sec . [ sec : dis ] , there are two ways in which this can be accomplished : 1 . using a stochastic approach , in which one considers the stochastic evolution of the function @xmath95 as a function of @xmath15 , or through 2 . the construction of a set of recursion relations for all moments of @xmath41 . in both cases , the resulting theory is formally identical to the known theories of localization in quasi - one dimension . although technically simpler , the stochastic approach is at odds with the goals of a semiclassical theory for ballistic localization : the goal of a theory of localization in an array of ballistic cavities is to describe an array of cavities with a fixed shape , using variations of the fermi energy as the only source of statistical fluctuations . since the fermi energy is set globally , for all cavities at the same time , quantum corrections for different cavities are not independent , and a stochastic approach is ruled out a priori . a stochastic approach is possible , however , if one relaxes the goals of the theory , allowing for small variations of the shape of each cavity , or for variations of a `` gate voltage '' that sets the fermi energy of each cavity independently . below we first describe the stochastic approach . in sec.[sec : b2 ] we consider the case of broken time - reversal symmetry , which is technically simpler . the discussion of localization in the presence of time - reversal symmetry is given in sec . [ sec : time ] . in sec . [ sec : hierarchy ] we discuss how a hierarchy of recursion relations for the moments of @xmath96 can be constructed , where the average is taken with respect to variations of the fermi energy of the entire array only . the stochastic approach deals with ( statistical ) properties of the function @xmath41 before averaging . although the properties ( i ) and ( ii ) of the trajectory - based semiclassical formalism [ sec . [ sec:2 ] ] are satisfied for the average of _ any _ product of traces of products alternating factors @xmath63 and @xmath64 to arbitrary order in @xmath76 , they have not been shown to follow from the semiclassical scattering matrix ( [ eq : ssemi ] ) _ before _ averaging . however , since our goal is a statistical theory of the transport the final statements of the theory will refer to averaged quantities only we will accept these two properties on the level of the sample - specific semiclassical scattering kernel @xmath97 in the arguments that follow below . starting from eq . ( [ eq : ssemi ] ) , the kernel @xmath96 is expressed as a double sum over classical trajectories @xmath1 and @xmath2 that connect the entrance and exit contacts [ fig . [ fig:2 ] ] , @xmath98 here @xmath4 and @xmath5 are the transverse momenta of @xmath1 and @xmath2 upon entrance . the two trajectories have equal transverse momenta upon exit , and exit at positions a quantum uncertainty @xmath99 apart see property ( ii ) above . below , we express the difference @xmath100 in terms of classical trajectories . we first calculate the average @xmath101 , where the average is taken with respect to variations of the fermi energy of the @xmath15th cavity ( or of its shape ) , while keeping the fermi energy and shape of the other cavities fixed . after that , we calculate the variance of @xmath102 and the higher cumulants . ( color online ) example of a pair of trajectories @xmath1 and @xmath2 contributing to @xmath103({p'},{p})$ ] . upon entering the sample , the two trajectories have different transverse momenta @xmath4 and @xmath5 , and may enter at different transverse positions . upon exiting the sample , @xmath1 and @xmath2 have the same transverse momentum , and exit at transverse positions a distance @xmath104 apart . ] _ average of @xmath102 . _ when calculating the average @xmath105 to leading order in @xmath106 in the absence of time - reversal symmetry , it will be sufficient to calculate @xmath107 in the diagonal approximation by considering trajectories @xmath1 and @xmath2 that are `` paired '' in the @xmath15th cavity . note , however , that trajectories need not be paired in the first @xmath44 cavities , because no average is taken there . each trajectory is classified by the number of times @xmath108 , @xmath109 that it enters the @xmath15th cavity from the @xmath110th cavity . hence , for each trajectory there are @xmath108 and @xmath109 segments in the @xmath15th cavity , which we label as @xmath71 , , @xmath111 and @xmath112 , , @xmath113 . since trajectories are paired in the @xmath15th cavity , @xmath114 . examples of trajectory pairs @xmath1 and @xmath2 with @xmath115 , @xmath38 , and @xmath116 are shown in fig . [ fig:3 ] . since trajectories need to be paired upon exit , @xmath117 has to be paired with @xmath118 . while there are @xmath119 ways in which the remaining segments can be paired , we now show that the `` diagonal pairing '' , @xmath120 paired with @xmath78 , @xmath121 , gives the main contribution to @xmath101 , whereas all other pairings give contributions a factor @xmath106 smaller . cutting the trajectories @xmath1 and @xmath2 at the contact between the @xmath110st and @xmath15th cavity also separates the part of each trajectory that resides in the first @xmath44 cavities into @xmath122 segments . the first segment of each trajectory connects the entrance contact to the exit of the @xmath110st cavity ; all other segments connect the exit contact of the @xmath110st cavity to itself . since the electron dynamics in the @xmath15th cavity is fully ergodic , the positions and transverse momenta with which these segments cross the interface between the @xmath110st and @xmath15th cavity are fully random , without correlations between different segments . [ correlations can be ruled out down to quantum phase space distances @xmath123 because the dwell time @xmath124 . ] hence , these @xmath122 segments can be interpreted as the semiclassical representation of a product of @xmath125 , @xmath126 , and @xmath127 factors @xmath128 and @xmath129 , _ before _ taking any average . note that , although trajectories @xmath130 that are `` paired '' in the @xmath15th cavity have a phase space distance of order @xmath131 or smaller when they enter / exit the @xmath15th cavity from / to the @xmath110st cavity , such trajectory pairs are sufficient for the semiclassical calculation of the complete kernels @xmath96 because of property ( ii ) of sec . [ sec:2 ] . ( color online ) examples of trajectories @xmath1 and @xmath2 contributing to @xmath132 , where an average is taken over the fermi energy of the @xmath15th cavity only . the four panels show trajectories contributing for @xmath115 ( top left ) , @xmath133 ( top right ) , @xmath134 with diagonal pairing ( bottom left ) , and @xmath134 with non - diagonal pairing ( bottom right ) . the number @xmath122 counts the number of times @xmath1 and @xmath2 enter the @xmath15th cavity from the @xmath110st cavity . the three dashed lines in each panel represent the entrance contact , the contact between the @xmath110st cavity and the @xmath15th cavity , and the exit contact ( from left to right ) . in the @xmath15th cavity , the trajectories @xmath1 and @xmath2 are piecewise equal up to quantum uncertainties . ] for the diagonal pairing , the @xmath122 segments of @xmath1 and @xmath2 that reside in the first @xmath44 cavities generate a particularly simple product of factors @xmath63 and @xmath64 , @xmath135^{m-1}$ ] , where @xmath136(p , p')\ ] ] is the reflection coefficient for the first @xmath44 cavities , seen from the exit contact of the @xmath110st cavity , and @xmath137 using the trajectory - based semiclassical formalism to evaluate the diagonal trajectory sums in the @xmath15th cavity,@xcite one then easily finds that the diagonal pairing gives @xmath138^{m-1 } \nonumber \\ & = & \frac{{\cal t}(n-1 ) g_{\rm c}}{2 g_{\rm c } - \mbox{tr}\ , { \cal r}(n-1)}. \label{eq : x1p}\end{aligned}\ ] ] using unitarity , one has @xmath139 so that eq . ( [ eq : x1p ] ) can be rewritten as @xmath140 up to corrections of order @xmath141 . this is precisely the semiclassical equivalent of the recursion relation of eq . ( [ eq : ttdiff ] ) , averaged over the fermi energy or shape of the @xmath15th cavity . it remains to show that non - diagonal pairings , in which a segment @xmath120 is not paired with @xmath78 give a contribution to @xmath101 that can be neglected in the limit of large @xmath16 . hereto we first consider the case @xmath134 , for which the only possible non - diagonal pairing is @xmath142 , @xmath143 [ fig . [ fig:3 ] , bottom right ] . for this pairing , the three segments of the trajectories that reside in the first @xmath44 cavities generate the semiclassical representation of a product of six scattering matrices , @xmath144 . from unitarity , one has @xmath145 for comparison , the diagonal pairing for @xmath134 generates @xmath146 ^ 2 $ ] , which is a factor @xmath147 larger because @xmath148 . using the ergodic dynamics in the @xmath15th cavity , the non - diagonal pairing of segments in the @xmath15th cavity for @xmath134 gives a contribution to @xmath149 equal to @xmath150.\end{aligned}\ ] ] this is a factor @xmath151 smaller than the leading contribution ( [ eq : ttdiffavg ] ) to @xmath101 . the same arguments can be used for @xmath152 : non - diagonal pairings come at the expense of at least two factors @xmath153 and , hence , lead to contributions to @xmath101 that are at least a factor @xmath151 smaller than the contribution from diagonal pairing . these arguments can also be used to show that contributions to @xmath107 that involve small - angle intersections of the trajectories in the @xmath15th cavity are a factor @xmath106 smaller than the leading contribution considered above . _ fluctuations of @xmath102 . _ the fluctuations of @xmath102 are described by the covariance @xmath154 . we calculate @xmath155 from the identity @xmath156 where we used the shorthand notation @xmath157 , @xmath158 , and @xmath159 . the average @xmath101 is given by eq . ( [ eq : ttdiffavg ] ) above , so it remains to calculate @xmath160 . the product @xmath161 is represented as a sum over four classical trajectories , @xmath1 , @xmath2 , @xmath162 , and @xmath163 . as before , we introduce the numbers @xmath108 , @xmath109 , @xmath164 and @xmath165 that indicate how often each trajectory enters the @xmath15th cavity . since trajectories always enter or exit the @xmath15th cavity in pairs , one has @xmath166 . unlike the average @xmath101 , for which the only contribution came from the diagonal approximation in the @xmath15th cavity with diagonal pairing of the segments @xmath120 and @xmath78 , the average of the second moment @xmath167 has contributions from both non - diagonal pairings in the diagonal approximation and from trajectories beyond the diagonal approximation , which have small - angle encounters in the @xmath15th cavity . we first consider the diagonal approximation , for which each segment @xmath77 or @xmath168 is paired with another segment @xmath78 or @xmath169 . for the diagonal pairing of segments , @xmath170 , @xmath171 and @xmath172 , @xmath173 , we find @xmath174 this contribution to @xmath175 precisely cancels the second and third terms in eq.([eq : dtt ] ) . hence @xmath176 must be from non - diagonal pairings of the trajectory segments within the diagonal approximation , or from trajectory configurations beyond the diagonal approximation in the @xmath15th cavity . since the latter class of trajectories have small - angle encounters , we write @xmath177 ( color online ) examples of four trajectories @xmath1 , @xmath2 , @xmath162 , and @xmath163 contributing to @xmath178 for an array of chaotic cavities , where the average is taken with respect to the fermi energy of the @xmath15th cavity only . the examples shown here have a non - diagonal pairing with @xmath179 , @xmath180 ( left panel ) and @xmath181 ( right panel ) . in the @xmath15th cavity , the trajectories @xmath1 and @xmath162 are piecewise equal to @xmath2 and @xmath163 , up to quantum uncertainties . ] the leading non - diagonal pairing appears if one pairs the first @xmath182 segments of @xmath1 with the first @xmath182 segments of @xmath163 , and the last @xmath183 segments of @xmath1 with the last @xmath183 segments of @xmath2 , as well as the first @xmath184 segments of @xmath162 with the first @xmath184 segments of @xmath2 , and the last @xmath185 segments of @xmath162 with the last @xmath185 segments of @xmath163 , where @xmath186 and @xmath187 are integers . two examples , with @xmath188 , @xmath189 , @xmath190 , and @xmath191 , and with @xmath192 and @xmath193 , are shown in fig . [ fig:4 ] . one then finds @xmath194 + { \cal o}(1/g_{\rm c}^2 ) , \nonumber\end{aligned}\ ] ] where @xmath195 and @xmath196 , with @xmath197 other pairings give contributions of order @xmath141 or smaller . ( color online ) two examples of a set of four trajectories contributing to @xmath198 that have a small - angle encounter in the @xmath15th cavity . for the left panel the encounter is in the interior of the @xmath15th cavity ; for the right panel it touches the exit contact . ] the second contribution to the fluctuations of @xmath102 comes from trajectories with a small - angle encounter in the @xmath15th cavity . since we only need fluctuations of @xmath102 to leading order in @xmath106 , it is sufficient to consider trajectories with one small - angle encounter only . taking the small - angle encounter to be between the segments @xmath199 , @xmath200 , @xmath201 , and @xmath202 , with @xmath203 , and pairing the first @xmath204 segments of @xmath1 with the first @xmath204 segments of @xmath163 , and the last @xmath205 segments of @xmath1 with the last @xmath205 segments of @xmath2 , as well as the first @xmath206 segments of @xmath162 with the first @xmath206 segments of @xmath2 , and the last @xmath207 segments of @xmath162 with the last @xmath207 segments of @xmath163 , we find @xmath208 where the first line comes from encounters in the interior of the @xmath15th cavity and the second line comes from encounters that touch the exit contact.@xcite examples of the two terms for @xmath209 are shown in fig.[fig:5 ] . [ we do not need to consider encounters that touch the contact between the @xmath110st cavity and the @xmath15th cavity , because the contribution from such encounters is included in the ( products ) of the kernel @xmath96 of the first @xmath44 cavities . ] combining everything we have @xmath210 \nonumber \\ & & \mbox { } + { \cal o}(1/g_{\rm c}^2 ) . \label{eq : ddtresult}\end{aligned}\ ] ] equation ( [ eq : ddtresult ] ) is the semiclassical equivalent of eq . ( [ eq : caltstat ] ) . higher cumulants of @xmath102 can be calculated in the same way . for the @xmath211th cumulant , one finds that only pairings that involve trajectories out of all @xmath211 factors @xmath96 contribute . each additional factor @xmath96 involved in the pairing scheme contributes an additional factor @xmath106 , which is why all cumulants with @xmath212 are of sub - leading order in @xmath106 . together , eqs . ( [ eq : ttdiffavg ] ) and ( [ eq : ddtresult ] ) form the semiclassical equivalent of the stochastic recursion relation ( [ eq : ttdiff ] ) used in the fully quantum mechanical theory of localization in disordered quasi - one dimensional wires . in the limit @xmath20 at fixed @xmath18 , eqs.([eq : ttdiffavg ] ) and ( [ eq : ddtresult ] ) can be mapped to stochastic differential equation for the eigenvalues of @xmath96 , which will be formally equivalent to the dmpk equation . the dmpk equation , in turn , provides a full description of localization in quasi - one dimensional wires.@xcite in the presence of time - reversal symmetry both the average @xmath213 and the covariance @xmath214 are different . in both cases the difference appears because one can pair time - reversed trajectories when taking the ensemble average in the @xmath15th cavity . there are two additional contributions to the average @xmath101 . the first of these arises from the diagonal approximation for the average in the @xmath15th cavity . as before , we define the number @xmath215 as the number of times the trajectories @xmath1 and @xmath2 enter the @xmath15th cavity from the @xmath110st cavity . the first additional contribution to @xmath101 then involves the pairing of segments @xmath216 with the time - reversed of @xmath217 , @xmath218 , and @xmath219 , @xmath220 examples for trajectory pairs contributing to the first and second line in eq . ( [ eq : tt1 ] ) are shown in fig . [ fig:6 ] . the second contribution comes from small - angle encounters inside @xmath15th cavity involving the segments @xmath221 , @xmath222 , @xmath223 , and @xmath224 , where @xmath225 . for this contribution one finds @xmath226 examples of trajectory pairs contributing to @xmath227 are shown in fig . [ fig:7 ] . using @xmath228 and adding these two contributions to eq . ( [ eq : ttdiffavg ] ) one finds @xmath229 up to corrections of order @xmath141 . ( color online ) two examples of trajectory pairs contributing to @xmath101 in the presence of time - reversal symmetry . ] for the fluctuations of @xmath102 in the presence of time - reversal symmetry one finds an extra contribution from pairing @xmath120 with the time - reversed of @xmath230 , @xmath231 , or @xmath232 with the time - reversed of @xmath233 , while following diagonal pairing rules for all other segments . proceeding as before , we find @xmath234 \nonumber \\ & & \mbox { } + { \cal o}(1/g_{\rm c}^2 ) . \label{eq : ddtresultb1}\end{aligned}\ ] ] where @xmath235(p , p').\end{aligned}\ ] ] the stochastic process defined by eqs . ( [ eq : ttdiffavgb1 ] ) and ( [ eq : ddtresultb1 ] ) precisely mirrors the stochastic process ( [ eq : ttdiff ] ) for the quantum mechanical matrix @xmath236 . ( color online ) two examples of trajectory pairs contributing to @xmath237 in the presence of time - reversal symmetry . the trajectory pairs have a small - angle encounter in the @xmath15th cavity . ] the direct construction of recursion relations for the moments of @xmath96 is an alternative to the stochastic approach that avoids extending the use of the properties ( i ) and ( ii ) of sec . [ sec:2 ] to sample - specific quantities and the necessity to define a statistical ensemble by varying the fermi energy or shape of each cavity individually . the construction of recursion relations for moments of @xmath96 proceeds in the very same manner as the construction of the stochastic recursion relations for @xmath96 , with the additional requirement that trajectories are piecewise paired in all @xmath15 cavities , not only in the @xmath15th cavity . since the arguments of the preceding section did not rely on the structure of the trajectories in the first @xmath44 cavities , one immediately concludes that the recursion relations for the moments of @xmath96 derived this way are identical to the recursion relations for the moments of @xmath96 one obtains from the stochastic approach . starting from the stochastic recursion relations ( [ eq : ttdiffavg ] ) and ( [ eq : ddtresult ] ) or ( [ eq : ttdiffavgb1 ] ) and ( [ eq : ddtresultb1 ] ) one arrives at the same hierarchy of recursion relations ( [ eq : tgeneral ] ) derived for an array of disordered cavities . ( color online ) examples of trajectory pairs contributing to the full average @xmath238 in the absence of time - reversal symmetry . the trajectory pairs have @xmath115 ( top ) , @xmath133 ( center ) , and @xmath134 ( bottom ) . the left panel shows trajectory pairs without small - angle self intersections ; the right panel shows trajectory pairs with two small - angle self intersections in the first @xmath44 cavities . ] we illustrate this procedure for the recursion relation for the first moment @xmath239 in the absence of time - reversal symmetry . following the rules of the semiclassical formalism , the average @xmath240 is determined by trajectory pairs @xmath1 , @xmath2 that are piecewise paired throughout the entire array of cavities . the trajectories can have small - angle self encounters , at which the pairing between @xmath1 and @xmath2 can be changed . each pair of trajectories is classified by the number @xmath122 of times the trajectories enter the @xmath15th cavity from the @xmath110st cavity . the @xmath122 segments of each trajectory in the @xmath15th cavity are labeled @xmath241 and @xmath242 . as in sec . [ sec : b2 ] , it will be sufficient to consider pairs of trajectories in which each segment @xmath120 is paired with the corresponding segment @xmath78 , @xmath121 . trajectory pairs with self encounters in the @xmath15th cavity or with non - diagonal pairings in the @xmath15th cavity give contributions to @xmath243 that are a factor @xmath106 smaller than the leading contribution . for the diagonal pairing , the remaining @xmath122 segments of the trajectories @xmath1 and @xmath2 that reside in the first @xmath44 cavities precisely generate @xmath244^{m-1 } \rangle$ ] , where the averaging brackets refer to variations of the fermi energy for the entire array of cavities . hence , we find @xmath245^{m-1 } \right\rangle}{2^m g_{\rm c}^{m-1 } } \nonumber \\ & = & \left\langle \frac{g_{\rm c } \mbox{tr}\ , { \cal t}(n-1 ) } { 2 g_{\rm c } - \mbox{tr}\ , { \cal r}(n-1 ) } \right\rangle . \label{eq : tavgdiff}\end{aligned}\ ] ] a schematic representation of trajectory pairs contributing to eq.([eq : tavgdiff ] ) with up to two self - encounters in the first @xmath44 cavities are shown in fig . [ fig : avg ] . using unitarity to express @xmath246 in terms of @xmath96 and subtracting @xmath247 , one then finds @xmath248 this is the same equation as one obtains from taking the average of the increment @xmath249 in the stochastic approach . in this section , we will approach the localization phenomenon from a different perspective . we will use that the quantum dot array depicted in fig . [ fig:1 ] supports hierarchies of different types of field - fluctuations in a field - theoretic description . these fluctuations reflect the fate of density distributions in classical phase space under the dynamical evolution of the system . each of these fluctuations , thus , comes with a characteristic ` relaxation time ' , _ i.e. , _ a time scale on which the fluctuation decays . ( for example , fluctuations inhomogeneous in the sector of phase space describing an individual quantum dot will decay on a time scale comparable to the time of flight through the dot , etc . ) in the description of low energy phenomena such as the zero frequency ( dc ) conductance , modes operating at short time scales can be treated perturbatively . their feedback into the sector of long time scales then stabilizes a ` low energy theory ' . in the following , we will derive a theory that is minimal in that it contains information equivalent to that stored in the fokker - planck equation of localization . the strategy pursued here parallels one applied previously @xcite to the problem of dynamical localization in the quantum kicked rotor ( also known as the `` standard map '' . ) one difference is that the spectrum of different modes encountered in the present problem happens to be more complex . our logics also resembles that of ref . , where it had been shown that the relevant low energy theory of an _ ergodic _ quantum system contains the information otherwise stored in random matrix theory . technically , our discussion will be based on a formulation of the array in terms of the ballistic nonlinear @xmath14-model.@xcite a quadratic approximation in energetically high lying modes generates an effective low energy theory wherein each quantum dot is treated as a structureless ( ` ergodic ' ) entity . this theory will be equivalent to the celebrated nonlinear @xmath14-model of disordered quantum wires,@xcite a model that predicts anderson localization on large length scales . we will see that the parameters stabilizing the hierarchical mode integration are the same as those utilized in previous sections of this paper . conceptually , the hierarchical scheme is an alternative to an indiscriminate perturbative integration over all modes in one go . that latter scheme would be essentially equivalent ( see ref . @xcite for a discussion in the context of spectral statistics . ) to a semiclassical expansion in terms of paired trajectories . in this sense , the hierarchical mode integration processes the information stored in the statistics of trajectories by different means . our starting point will be the description of the quantum dot array in terms of the supersymmetric ballistic nonlinear @xmath14-model . this theory is obtained by averaging exact functional representations of green functions over an energy interval of width @xmath250 centered around the uniform fermi energy @xmath251 of the array . a subsequent saddle point approximation ( stabilized in the parameter @xmath252 ) then obtains a field theory in classical phase defined by @xmath253),\end{aligned}\ ] ] with @xmath254 & = & \frac { \beta\pi \hbar \nu}{2 } \int_\gamma ( d{{\bf x } } ) \,{\rm tr}\big(t\ast \lambda \ { h , t^{-1}\}\big ) \nonumber \\ & & \mbox { } + s_{\rm reg}[t].\end{aligned}\ ] ] the integration variables in ( [ eq:5 ] ) , @xmath255 are ( super ) matrix valued fields defined on shells @xmath256 of constant energy in classical phase space . further , @xmath257 where @xmath258 and @xmath259 are coordinates and momenta , respectively , @xmath260 is the hamiltonian function of the system , the integral over the energy - shell is normalized to the ( spatial ) volume of the system , @xmath261 , and @xmath262 is the single particle density of states per volume , @xmath263 . for time reversal and spin rotation invariant systems ( orthogonal symmetry class , @xmath37 ) , the ` internal ' structure of the matrices @xmath264 is described by a composite index @xmath265 , where @xmath266 discriminates between the advanced and retarded sector of the theory , @xmath267 discriminates between commuting and anticommuting sectors , and @xmath268 accounts for the operation of time reversal . time reversal symmetry reflects in the relation @xmath269 , where @xmath270 is a fixed matrix whose detailed structure will not be of concern throughout . for time reversal non - invariant systems ( unitary symmetry , @xmath271 ) no time - reversal structure exists and @xmath272 . in either case , the matrices @xmath273 carry a coset space structure in the sense that configurations @xmath273 and @xmath274 are to be identified if @xmath275=0 $ ] , where @xmath276 and `` ar '' stands for action in advanced / retarded space . finally , the regulatory action @xmath277\equiv \delta \int_\gamma ( d{{\bf x}})\ , { \rm str}\big ( \lambda t^{-1 } \ast \lambda t\big),\ \ \delta \downarrow 0\ ] ] determines the ` causality ' of field fluctuations , but will otherwise not be of much relevance throughout . the fluctuation behavior of the fields @xmath273 in ( [ eq:5 ] ) is governed by the classical liouville operator @xmath278 ( where @xmath279 is the poisson bracket . ) quantum mechanics enters the problem through the presence of moyal products `` @xmath280 '' in ( [ eq:5 ] ) . in essence , this product operation@xcite limits the maximum resolution of the theory in phase space to scales of the order of a planck cell . in the following sections we reduce the above ` bare ' theory to an effective low energy theory describing localization phenomena . before turning to the technicalities of the mode integration , let us describe the relevant hierarchies in qualitative terms : fluctuations inhomogeneous in the phase space sector representing individual dots are expected to relax on short time scales comparable to the time of flight across the dot , @xmath281.@xcite on the other hand , relative fluctuations in the configuration of different dots can survive up to time scales of the order of the dwell time @xmath282 . to describe this hierarchical decay profile within our field theory framework , we focus on a section of the array containing two neighboring quantum dots ( cf . [ coupling_phase_space ] ) . the fields @xmath283 representing phase space fluctuations in this subsystem may be decomposed as @xmath284 , where @xmath285 are ` slow ' and ` fast ' fluctuations , respectively . the slow fluctuations are ( i ) homogeneous within each dot separately . in particular , ( ii ) they do not vary in any spatial cross section transverse to the array , and ( iii ) do not depend on momentum . however , ( iv ) the weakness of the inter - dot coupling ( @xmath286 ) leaves room for gradual fluctuations of the slow modes as we pass from one dot into the other . this suggests a parameterization @xmath287 , where @xmath5 is the component of @xmath258 parallel to the longitudinal direction of the two - dot system . point ( i ) above means that @xmath288 and @xmath289 , where @xmath290 are the constant slow mode configurations of the left and the right dot , respectively . as we are going to check in a self consistent manner , ( v ) relative fluctuations between left and right dot are suppressed ( in a parameter of the order of the number of transverse quantum channels supported by the connector region ) so that a leading order expansion in relative fluctuations @xmath291 is sufficient . ( conceptually , this expansion is equivalent to the kramers - moyal expansion employed in the derivation of the fokker - planck equation above . ) the field @xmath292 encapsulates all other fluctuations , _ i.e. _ , fluctuations that do not meet the criteria ( i - v ) . generic fluctuations of this type think of fast fluctuations deep inside the phase space of one of the dots are strongly gapped and do not couple to the slow fluctuations . ( formally , this decoupling manifests itself in an effective ` orthogonality ' between the modes representing these fluctuations . ) however , there is one sector of phase space , @xmath293 , in which fast and slow fluctuations talk to each other . the domain @xmath293 includes all points in phase space which pass from one dot to the other in a time of order @xmath294 , much smaller than the typical dwell time ( cf . the shaded area in fig . [ coupling_phase_space ] ) . this region is special in that it overlaps with the domain of gradual variation of the slow fields ( cf . point ( iv ) above . ) as we will see , a perturbative integration over fast fields in @xmath293 effectively determines the slow field coupling between the two dots . to prepare the integration over the fast fields in @xmath293 , we need to bring the notation to a more explicit level : assuming that the bulk dynamics is ballistic , @xmath295 where @xmath296 is the unit vector in momentum space , and @xmath297 the fermi velocity . current conservation in the specular reflection at the system boundaries translates to the effective boundary condition @xmath298 , where @xmath299 is a phase function subject to the action of @xmath278 , @xmath300 a boundary point and @xmath301 the direction vector with flipped normal component . we next parameterize fluctuations as @xmath302 , where the field generators carry a block structure ( in advanced / retarded space ) , @xmath303 ( for further technical details on this representation we refer to ref . . ) we now substitute these generators into the action , expand to leading order in @xmath304 , and integrate . this leads to the effective action @xmath305= \frac{1}{2}\left \langle \left(s^{(1)}[t_s , w_f]\right)^2\right\rangle + s_{\rm reg}[t_s],\ ] ] where @xmath306)(\dots)$ ] , and @xmath307 $ ] is of @xmath15th order in @xmath304 . due to the isotropy of @xmath308 in momentum space and the linearity of the poisson bracket in @xmath309 , the action does not contain a contribution of zeroth oder in the fast fields . the dominant contribution to the coupling between fast and slow fluctuations is given by the linear term , @xmath310&= \beta \pi \hbar \nu v_f \int_{\gamma_c } d{{\bf q}}d{{\bf n}}\,n_\parallel \,{\rm str}\left[w_s({{\bf q}},{{\bf n } } ) \phi(q)\right ] \label{tsw}\end{aligned}\ ] ] where we have introduced the abbreviation @xmath311 , the integration over the direction of momentum is normalized as @xmath312 , @xmath313 is the component of @xmath309 parallel to the longitudinal direction and the moyal product has been omitted ( which is permissible due to the general relation @xmath314 for the integrated product of _ two _ functions presently , matrix elements of @xmath315 and @xmath13 in phase space . ) neglecting contributions of @xmath316 ( as compared to the @xmath317-terms above ) the quadratic @xmath318-action is given by @xmath319 & \simeq & s^{(2)}[\openone , w_f ] \nonumber \\ & = & \beta \pi \hbar \nu \int_{\gamma_c } d{{\bf q}}d{{\bf n}}\,{\rm str}(\bar b_f({{\bf q}},{{\bf n } } ) ( v_f { { \bf n}}\cdot \nabla + \delta ) b_f({{\bf q}},{{\bf n}})).\end{aligned}\ ] ] fast field fluctuations can now be integrated out according to the prescription @xmath320 , where @xmath321 specifically , the effective action is given by @xmath322 & = & ( \beta \pi \hbar \nu v_f)^2\int_{\gamma_c } d{{\bf q}}d{{\bf q}}'d{{\bf n}}\ , n_\parallel^2 \mbox{str}\ , \left(\phi_{21}(q ) \phi_{12}(q ) \right ) \pi({{\bf q}},{{\bf q}}';{{\bf n } } ) \end{aligned}\ ] ] so far , we have not made reference to the specific properties of the phase space region @xmath293 . we now assume that the corridor connecting the dots has wave - guide properties , in that it ( a ) does not contain significant backscattering , and ( b ) the restriction of the liouville operator to @xmath293 has plane - wave like eigenfunctions characterized by a conserved longitudinal / transverse momentum @xmath323 . we also assume that ( c ) the slow fields smoothly interpolate between @xmath324 and @xmath325 in a region centered around the longitudinal coordinate @xmath326 . the presumed proximity @xmath327 implies that @xmath328 can be linearized , where we suppressed the slow field index ` @xmath329 ' for notational transparency . under these circumstances , and noting that the integration over @xmath330 implies a projection onto the zero - momentum sector @xmath331 , we obtain @xmath322 & = -\beta \pi \hbar \nu s_c v^2_f\int_{-c}^c dq dq ' { \rm \,str}\left(\phi_{21}(q)\phi_{12}(q ' ) \right ) \int d{{\bf n}}\ , n^2_\parallel \int \frac { dk } { 2\pi } \frac{e^{i k ( q - q')}}{i v_f n_\parallel k+ \delta } = \\ & = -\beta \pi \hbar \nu s_c v_f\int_{-c}^c dq dq ' { \rm \,str}\left(\phi_{21}(q ) \phi_{12}(q ' ) \right ) \underbrace{\int d{{\bf n}}\ , |n_\parallel|\theta\left ( \frac{q - q ' } { n_\parallel}\right)}_{={\rm const.}}=\\ & = { \rm const.}\times \hbar \nu s_c v_f\ , { \rm str}((\bar b_r-\bar b_l)(b_r - b_l)),\end{aligned}\ ] ] where the indices refer to `` ar''-space , @xmath332 is a constant , @xmath333 the transverse cross section of the contact , and @xmath334 are the generators of the slow fields in the left and the right dot , respectively . noting that the density of states per volume , @xmath335 ( where @xmath336 is the fermi momentum ) and @xmath337 is proportional to the number of transverse channels , @xmath338 , supported by the connector region , the prefactor can be written as @xmath339 , a number which we assume large lest a semiclassical description of the contact becomes meaningless . we also note that the quadratic form may be replaced by its unique rotationally invariant generalization to the full field manifold , @xmath340 , where @xmath341 . while a quadratic expansion of the latter expression reproduces the bilinear term , the largeness of @xmath338 implies that typical values contributing to the @xmath342-integration , @xmath343 , which means that non - linear contributions become inessential in the limit of large channel numbers . we thus conclude that the coupling term can be rewritten as @xmath344= { \rm const . } \times n { \,\rm str}(q_l q_r).\ ] ] finally , the obvious generalization of the above two - dot construction to an entire quantum dot array reads as @xmath345= { \rm const . } \times \frac{1}{\tau_{\rm d } \delta } \sum_{m } { \rm str}(q_m q_{m+1}),\ ] ] where @xmath346 is the @xmath347-matrix representing the @xmath122-th dot and we used that the channel number @xmath348 . before turning to the discussion of localization properties , a few remarks on the construction above are in order . * conceptually , the above reduction programs involves three steps : 1 ) identification of ` low energy modes ' , _ i.e. _ , modes that decay on the largest time scales of the problem , 2 ) identification of ` high energy ' , or quickly decaying modes , and 3 ) perturbative integration over those fast modes that will conceivably couple to the slow modes . in principle , that integration can be explicated for any fluctuation in the problem . in practice , however , only few modes will effectively couple to the slow degrees of freedom , and these relevant fluctuations are best identified by semiclassical considerations : * semiclassically speaking , a ` mode ' represents the coherent propagation of a retarded and an advanced feynman amplitude along classical trajectories in phase space . locally , the semiclassical dynamics of such composites is described by the liouville operator , as is manifest in the action ( [ eq:5 ] ) . this trajectory interpretation helps in identifying the relevant fast modes . e.g. , in the system depicted in fig . [ coupling_phase_space ] , the coupling between the ergodic slow modes of each dot , @xmath346 , is dominated by trajectories swiftly propagating from one dot to the other , _ i.e. _ , modes emanating at phase space points @xmath349 . * although this identification of fast modes rests on specific model assumptions ( no backscattering in the contact region , etc . ) , the result ( [ qdotarray ] ) is reasonably universal . for example , a somewhat more elaborate construction will show that the same action describes connector regions containing chaotic scattering and momentum relaxation by themselves . ( the latter complication would manifest itself in an altered value of @xmath88 , though . ) generally speaking , the low energy physics of the system will reduce to @xmath350 , as long as the dots are isolated from each other in the sense @xmath351 . the effective action ( [ qdotarray ] ) is equivalent to a ` lattice version ' of the diffusive nonlinear @xmath14-model of quasi - one dimensional disordered wires . indeed , we may pass to a continuum limit @xmath352 where @xmath353 is replaced by a smooth field , @xmath354 , and @xmath355 the spacing between the dots . comparing to the standard form of the diffusive model,@xcite where the action is @xmath356 with @xmath357 the localization length , we are led to the identification @xmath358 . in the preceding sections we showed how a theory of anderson localization can be constructed for a sample in which the microscopic electron dynamics is ballistic , rather than disordered - diffractive . our theory of `` ballistic anderson localization '' paraphrases the scaling approach to localization in disordered quantum wires of dorokhov , mello , pereyra , and kumar,@xcite using the language of the trajectory - based semiclassical formalism . as noted in the introduction , the interest of constructing such a semiclassical theory is not the structure of the theory itself or the phenomena it explains . like most semiclassical theories of quantum corrections in the perturbative regime , the structure of the theory closely resembles the structure of its fully quantum mechanical counterpart for disordered metals , whereas the observed phenomena are the same in the ballistic and disordered cases . instead , the interest of the semiclassical theory is that it shows how quantum effects that were known from disordered metals arise if the electron dynamics is ballistic . on hindsight it should not come as a surprise that a theory of ballistic anderson localization can be constructed by adapting the derivation of the dorokhov - mello - pereyra - kumar equation for a disordered quantum wire . this point is best made by reconsidering dorokhov s original derivation.@xcite in this derivation , impurity scattering is treated in the born approximation . all quantum mechanical amplitudes are squared into quantum probabilities . hence , it is sufficient if one can replace the quantum mechanical probabilities by classical ones . such a replacement is a standard procedure when connecting quantum mechanical and semiclassical theories . its implementation for the array of chaotic cavities is what is done here . it is interesting to observe that , while small - angle encounters form a crucial link in our understanding of quantum interference corrections in ballistic conductors , their role is very limited in our description of localization in quasi - one dimension : they serve to cancel spurious contributions from trajectories that enter the last cavity of the array more than twice . ( in dorokhov s original approach such processes are excluded automatically because of the condition that the length of the wire is increased by an amount @xmath359 much smaller than the mean free path @xmath86.@xcite ) implicitly , small - angle encounters do play a much more important role in our theory , however , because they help to preserve unitarity in the semiclassical theory . unitarity is a key ingredient of both the quantum - mechanical derivation of the dmpk equation and the present semiclassical derivation . we note that unitarity has played an important role in other extensions of the semiclassical framework beyond its previously assumed domain of validity : it is used to relate weak localization and enhanced backscattering , thus enabling a semiclassical description of weak localization without reference to small - angle encounters,@xcite and it is used to obtain an alternative expression for the spectral form factor , allowing its calculation in the non - perturbative regime by considering periodic orbits of duration below the heisenberg time only.@xcite we have also shown that the dynamical information entering the semiclassical approach can be processed by different means to derive an effective low energy field theory of the system . this latter approach is based on the concept of ` modes ' , fluctuations in phase space decaying on parametrically different time scales . a successive integration over short lived modes stabilizes an effective action of the most persistent modes in the fluctuation spectrum . in the present context , that low energy limit turned out to be equivalent to the diffusive nonlinear @xmath14-model of disordered wires . while the field theory approach is arguably less explicit than the direct classification of trajectories , it enjoys the advantage of high computational efficiency , a paradigm previously exemplified on the problem of universal spectral correlations.@xcite for example , the above mentioned condition of unitarity , as well as the symmetries relating trajectories to their time reversed are built into the field theory approach from the outset ; there is no need for explicit bookkeeping in terms of encounter processes . the price to be payed for this compactness in the description is a higher level of abstraction , though . we thank fritz haake for bringing this problem to our attention . this work was supported by the packard foundation , the humboldt foundation , the nsf under grant no . dmr 0705476 , and by the sonderforschungsbereich sfb / tr 12 of the deutsche forschungsgemeinschaft .

de sitter space - time has a one complex parameter family of invariant vacua for the theory of a free , massive scalar field . for most of these vacua , in an interacting scalar theory the one loop corrections diverge linearly for large values of the loop momentum . these divergences are not of a form that can be removed by a de sitter invariant counterterm , except in the case of the euclidean , or bunch - davies , vacuum .

hep-th0306028

the importance of understanding quantum field theory in de sitter space , the space - time associated with a positive cosmological constant , has been heightened by recent observations of both the early and late universe . the dramatic results of wilkinson microwave anisotropy probe @xcite have provided further strong evidence that the universe underwent a rapid inflationary expansion . both the large - angle anti - correlation in the temperature - polarization cross - power spectrum and the nearly flat spectral index are consistent with the predictions of inflation . more surprisingly , the dimming of the type ia supernovae seen by the supernova search team @xcite and the supernova cosmology project @xcite , combined with other observations , is yielding a new standard picture for the contents of the universe , the largest component of which is a dark energy whose properties are consistent with a positive cosmological constant . a striking difference between de sitter and flat space - time is the richer vacuum structure of the former . for a free scalar field in a minkowski space , there exists an essentially unique poincar invariant vacuum state . in contrast , for a de sitter background , mottola @xcite and allen @xcite discovered an infinite family of vacua for the quantum theory of a free massive scalar field that are invariant under the isometries of de sitter space . these vacua can be parameterized by a single complex number , @xmath0 , and are usually called the @xmath0-vacua . most of these @xmath0-vacua have a host of peculiar features , such as a mixture of positive and negative frequency modes at short - distances and a non - thermal behavior that violates the principle of detailed balance . only one of these states , the euclidean or bunch - davies @xcite vacuum , behaves thermally when viewed by an unruh detector @xcite and reduces to the minkowski vacuum as we take the cosmological constant to zero . the assumption that the universe was at least approximately in the euclidean vacuum underlies the successful predictions of inflation for the calculation of the density fluctuations which produced the temperature anisotropies in the cosmic microwave background radiation . despite their unappealing features , the @xmath0-vacua are perfectly valid vacua for a free scalar field . if they can not be shown to be unphysical , then their existence would undercut some of the robustness of the inflationary paradigm we would need to explain how the epoch prior to inflation managed to place the universe in the euclidean vacuum rather than one of the other infinite family of @xmath0-vacua . for example , the regularization needed by the energy - momentum tensor even for the free theory in the @xmath0-vacuum is not generally compatible with that needed after inflation @xcite . a complication in formulating quantum field theory in de sitter space is its lack of a well - defined @xmath1-matrix . in an interacting theory we have two sources of time dependence for matrix elements one induced by any inherent time dependence of the background geometry and another introduced by the interactions . in such a system , it is therefore appropriate only to ask time dependent questions to study how a matrix element evolves from a given initial state . schwinger @xcite and keldysh @xcite developed a formalism to solve for this finite time evolution . in their approach , we specify the state of the system at an initial time and then evolve to a finite time later . here , both the ` in ' and ` out ' states correspond to the same state and are evolved together when we evaluate the expectation value of an operator in effect this formalism evaluates matrix elements between two ` in ' states . the schwinger - keldysh formalism is thus ideally suited for studying the behavior of the @xmath0-vacua in the presence of interactions . we place the system initially in an @xmath0-vacuum and then study whether a sensible evolution results . since the quantum field theory only is evolved over a finite interval , our results are relevant not only for the more formal question of the @xmath0-vacua in an eternal de sitter background but also for the phenomenological problem of a finite epoch of inflation . the methods established here can also be applied to any initial state , such as the ` truncated @xmath0-vacua ' of @xcite . in these vacua , the short - distance behavior of the @xmath0-vacua is modified either in accord with some specific theory , such as the stringy uncertainty relation of @xcite , or simply by truncating the @xmath0 mode functions above some energy scale to reflect our ignorance of the new physics @xcite . we address the formal case of a pure @xmath0-vacuum and shall study the truncated case later in @xcite . in this article , we show that an interacting scalar field theory in a general @xmath0-vacua contains linear divergences which can not be removed with a de sitter invariant renormalization prescription . these divergences appear in the one loop corrections and are present for arbitrarily weak interactions . the specific example we study is the expectation value of the number of euclidean particles in an @xmath0-vacuum . the divergences appear in the high momentum region of the loop integral . we show that they only vanish for the euclidean vacuum , which is completely renormalizable . the subject of the @xmath0-vacua for an interacting theory has also been recently investigated in related work @xcite . both of these works essentially studied the corrections to the two - point correlation function obtained between an ` in ' @xmath0 state and an ` out ' state given by the @xmath0 state at a later time . banks and mannelli @xcite found that the interacting theory in the @xmath0-vacuum required non - local counterterms while einhorn and larsen @xcite found pinched singularities in the loop corrections . these features provided highly suggestive evidence that the @xmath0-vacua are pathological in the presence of interactions . some attempts to modify the theory to avoid these problems appear in @xcite . we begin with a review of the de sitter invariant vacua for a free scalar field in sec . [ greens ] . this section also shows the form of the wightman functions in conformally flat coordinates . section [ sk ] derives the expectation value of an operator in an interacting theory based on the schwinger - keldysh formalism . in sec . [ number ] we calculate the change in the number of euclidean particles in an @xmath0-vacuum due to a cubic interaction and show that in the presence of this interaction , the expectation value is renormalizable for the euclidean vacuum while an unrenormalizable divergence appears for the @xmath0-vacuum . section [ discuss ] explores the origin of these divergences in the @xmath0-vacua in a more general setting . we derive the necessary conditions for these divergences to arise and show how they can appear in a general interacting scalar field theory . section [ conclude ] summarizes our results and suggests future applications for this formalism . in this section we review the rich vacuum structure of a free scalar field in de sitter space @xcite . we derive the form of the wightman function and eventually the feynman propagator in conformally flat coordinates . these green s functions will be used later for studying the interacting theory . the most straightforward method for demonstrating the existence of a family of de sitter invariant vacua is to evaluate the two - point wightman function for a free massive scalar field in an @xmath0-vacuum . for this purpose it is useful to use a coordinate system that covers the entire space - time . such coordinates are not , however , those best suited for more explicit calculations . therefore , throughout this article we shall study de sitter space using conformally flat coordinates , @xmath2 with @xmath3 $ ] which cover half of de sitter space @xcite . the other half of the space is covered by a set of coordinates with @xmath4 . these coordinates are simply related to the standard coordinates used in inflation , @xmath5 through @xmath6 . @xmath7 is the hubble constant and is related to the cosmological constant by @xmath8 . to an observer capable only of probing length scales on which the curvature of de sitter space is not apparent , the space - time appears approximately flat . for the high energy modes then , this observer can apply the same prescription for defining positive and negative frequency modes as in minkowski space . the vacuum state annihilated by the operators @xmath9 associated with these modes corresponds to the euclidean vacuum . the euclidean vacuum possesses many desirable properties in addition to matching with the minkowski vacuum at short distances or as @xmath10 . it corresponds to the unique state whose wightman function is analytic when continued to the lower half of the euclidean sphere . moreover , an unruh detector placed in the euclidean vacuum satisfies the principle of detailed balance as though it were immersed in a thermal system at the de sitter temperature , @xmath11 @xcite . if we denote the euclidean vacuum by @xmath12 , the euclidean wightman function for a free massive scalar field @xmath13 is defined by @xmath14 since the metric is spatially flat , the expansion of the scalar field @xmath13 in creation and annihilation operators @xmath15 for the vacuum state @xmath12 is @xmath16 . \label{phimodes}\ ] ] with the commutator normalized to be @xmath17 = ( 2\pi)^3 \delta^3(\vec k - \vec k ' ) , \label{aadagger}\ ] ] the euclidean wightman function in position space is @xmath18 where the momentum representation the wightman function is @xmath19 note that the mode functions only depend on the magnitude of the spatial momentum , @xmath20 . a free massive scalar field satisfies the klein - gordon equation , @xmath21 \phi(x ) = 0 , \label{kg}\ ] ] so that the mode functions solve the differential equation , @xmath22 u_k^e(\eta ) = 0 . \label{kgmodes}\ ] ] note that @xmath23 here represents the effective mass of the theory which includes any contribution from coupling the field to the curvature , @xmath24 . in de sitter space - time the curvature @xmath25 is constant so this coupling is of the same form as a mass term . the solutions to eq . ( [ kgmodes ] ) are linear combinations of bessel functions , @xmath26 with @xmath27 we shall assume hereafter that @xmath28 is real . the general form for the mode functions is applicable to both the euclidean vacuum and the @xmath0-vacuum . what distinguishes the former is that as the hubble constant is taken to vanish , @xmath10 , so that de sitter space becomes flat , we should recover only the positive frequency mode functions , @xmath29 . in the small @xmath7 limit , @xmath30 the leading time dependence of the modes is @xmath31 when @xmath32 up to the normalization factor , @xmath33 , the euclidean mode functions are given by @xmath34 \nonumber \\ & = & n_k\ , \eta^{3/2}\ , h_\nu^{(2)}(k\eta ) . \label{emodesunnorm}\end{aligned}\ ] ] @xmath35 represents a hankel function . we shall now choose the units such that @xmath36 . the normalization is fixed by the canonical equal - time commutation relation @xmath37 = -i \delta^3(\vec x - \vec x ' ) \label{equaltime}\ ] ] where the conjugate momentum is @xmath38 the equal - time commutation relation requires that the modes satisfy a wronskian condition of the form @xmath39 which determines the normalization of the modes to be @xmath40 therefore , the euclidean mode functions are given by @xmath41 while the de sitter invariance of the wightman function is not manifest from eq . ( [ emodes ] ) , it is possible to write @xmath42 as a function of the de sitter invariant distance between its arguments @xcite . in conformally flat coordinates , this invariant distance between @xmath43 and @xmath44 is @xmath45 although we shall state most of our results in terms of the mode functions for a general mass , it will be convenient to show the results for a particular case in which the mode functions simplify substantially . when @xmath46 , the hankel function in eq . ( [ emodes ] ) is proportional to an exponential , @xmath47 this case corresponds to a massless , conformally coupled scalar field for which the effective mass is @xmath48 . the euclidean wightman function is then @xmath49 and is finite provided we choose the appropriate @xmath50 prescription , @xmath51 here the appearance of the invariant distance @xmath52 establishes the de sitter invariance of the vacuum . the choice of the short distance behavior of the mode functions which determined the relative contributions of the two independent solutions to the klein - gordon equation is not the unique choice which leads to a de sitter invariant wightman function . mottola @xcite and allen @xcite observed that the vacuum state @xmath53 annihilated by a bogolubov transformation of the euclidean operators , @xmath54 , \label{aalphadef}\ ] ] also yields a de sitter invariant wightman function , @xmath55 here , @xmath56 and the normalization @xmath57 is chosen to preserve the normalization of the commutation relation in the @xmath0-vacua , analogous to eq . ( [ aadagger ] ) . note that the euclidean vacuum is itself among the @xmath0-vacua being obtained when @xmath58 . in proving that @xmath59 only depends on @xmath52 it is useful to use a coordinatization that covers the entire de sitter space - time . in such global coordinates , both a point @xmath60 and its antipode @xmath61 occur in the same coordinate system @xcite . it is then possible to chose euclidean mode functions @xmath62 such that @xmath63 so that the bogolubov transformation of eq . ( [ aalphadef ] ) gives @xmath64 . \label{globalbog}\ ] ] here @xmath65 labels the elements of a general basis of mode functions . in this form , the de sitter invariance of the euclidean wightman function and the fact that @xmath66 together imply that the @xmath0-wightman function only depends on the de sitter invariant distance between @xmath60 and @xmath67 @xcite . while it is helpful to use a coordinate system which contains the antipode of every point to establish this invariance , for explicit calculations it is not necessary to use global coordinates . equation ( [ aalphadef ] ) relates the mode functions of the @xmath0-vacuum to the mode functions of the euclidean vacuum and their complex conjugates we do not need to transform the conjugated mode function into a function of the antipode once we have established that @xmath59 is invariant . from the euclidean mode functions of eq . ( [ emodes ] ) we can now construct the mode functions for the @xmath0-vacua . expanding the scalar field in terms of @xmath0 creation and annihilation operators , @xmath68 , \label{phiamodes}\ ] ] and using eq . ( [ aalphadef ] ) yields @xmath69 \label{amodesshort}\ ] ] since the @xmath70 only depend on the magnitude of @xmath71 . thus the @xmath0-vacuum modes are @xmath72 . \label{amodes}\ ] ] inserting the @xmath0-mode expansion into eq . ( [ awight ] ) yields @xmath73 again , the spatial flatness makes it natural to use a momentum representation @xmath74 the additional complexity of the mode functions in the @xmath0-vacuum means that it is particularly helpful to have a case in which these functions simplify . for a massless , conformally coupled scalar field , @xmath75 . \label{amodesfornuhalf}\ ] ] and the wightman function becomes @xmath76 as in the euclidean case , the de sitter invariance is manifest in the above expression . to study the propagation of signals in a de sitter background , define the feynman propagator , @xmath77 so that it satisfies the klein - gordon equation with a point source , @xmath78 g(x , x ' ) = { \delta^4(x - x')\over\sqrt{-g(x ) } } . \label{kgfeyn}\ ] ] the propagator can only depend on the difference between the spatial positions of the points so that its fourier transform is @xmath79 where @xmath80 is the solution to @xmath81 g_k(\eta,\eta ' ) \nonumber \\ & & \qquad\qquad\qquad\qquad\qquad\qquad = \eta^2 \eta^{\prime 2 } \delta(\eta-\eta ' ) . \qquad \label{ftgreenseqn}\end{aligned}\ ] ] the solution to this equation which satisfies the correct boundary conditions at @xmath82 has the form @xmath83 here @xmath84 and @xmath85 are essentially the two - point wightman functions calculated earlier in eq . ( [ ftawightmodes ] ) , @xmath86 although propagation for a free field theory in the @xmath0-vacua contains some peculiar features , it is not otherwise ill - defined . the pathological features of quantum field theory in an @xmath0-vacuum only appear in an interacting field theory . the form of the @xmath0-wightman function already suggests that the interacting theory could be ill - defined , since the various terms in eq . ( [ awightnuhalfz ] ) contain different @xmath50 prescriptions . this property implies that in a standard approach to calculating the one loop corrections to the propagator , among the products of the propagators participating in the loop appear products of poles with the opposite @xmath50 prescription pinched singularities @xcite . for example , in the one - loop correction to the propagator appears a product of the green s functions given in eq . ( [ awightnuhalfz ] ) . however , these pinched singularities do not by themselves prove whether the @xmath0-vacuum is itself pathological or whether the standard methods for studying the quantized theory are inappropriate for a time - evolving background such as de sitter space . a significant difference between de sitter space , in most coordinatizations , and flat space is the explicit time - dependence of the metric . unlike a flat space - time where the generator of time - translations is a killing vector globally , de sitter space - time has no such global time - like killing vector . moreover , in a particular coordinate system such as inflationary coordinates the time derivative may not even generate an isometry locally . these properties suggest that rather than attempt to define an @xmath1-matrix between ` in ' and ` out ' states defined at different times , we should apply a quantization procedure that evolves an entire matrix element over a finite interval . it is also useful to be able to evolve a given state forward from a specified initial time @xmath87 , rather than to use a state in the asymptotic past . we can always take @xmath88 . an additional advantage of solving the evolution over finite intervals is that such an approach more immediately determines whether the @xmath0-vacuum is applicable for inflation , which does not require a de sitter space - time eternally , but only over a sufficient interval to generate the number of @xmath89-foldings needed to explain the flatness and the homogeneity of the universe . if the interacting @xmath0-vacuum shows its pathology even over a finite interval , then we can exclude the possibility that the universe was in a pure @xmath0-state during any epoch of inflation , regardless of the prior history of the universe . the closed time contour formalism developed by schwinger @xcite , keldysh @xcite and mahanthappa @xcite allows us to study the evolution of a quantum field theory over a finite interval after specifying the state at an initial surface . we review here their approach which leads to an expression for perturbatively evaluating the matrix element of an operator , which is given at the end of this section in eq . ( [ omatrix ] ) . in the interaction picture , the evolution of operators is given by the free hamiltonian , @xmath90 , while the evolution of states is given by the interactions , @xmath91 . if we let @xmath92 denote a general basis of states for the theory , then the behavior of the system is completely described by the density matrix , @xmath93 . \label{heisenberg}\ ] ] the advantage of the interaction picture is that fields evolve using the free hamiltonian , @xmath94 . \label{heisenbergphi}\ ] ] the time evolution of the field is precisely that given in the previous section since here the mode functions still are solutions to the free klein - gordon equation . to study the evolution introduced by the interactions , it is convenient to include a ` turning on ' function in the interacting part of the hamiltonian , @xmath95 here @xmath96 vanishes when @xmath97 and becomes one when @xmath98 is sufficiently large compared with @xmath87 . later we shall let this function be a @xmath99 step function . we shall often not write this function explicitly , absorbing it into @xmath91 . thus the state does not evolve before @xmath87 : @xmath100 for @xmath97 . once we have specified the state at a particular time , @xmath101 , then the eq . ( [ heisenberg ] ) allows us to determine the state at all subsequent times . to study the vacuum structure of de sitter space , the initial state will correspond to an @xmath0-vacuum . if we introduce a unitary , time - evolution operator @xmath102 that evolves the state , @xmath103 then from eq . ( [ heisenberg ] ) @xmath104 obeys @xmath105 with @xmath106 . the formal solution to this equation is given by dyson s equation in terms of the time - ordered exponential @xmath107 the evolution of the expectation value of an operator in this time - dependent background is given by @xmath108 \over { \rm tr } \bigl [ \rho(\eta ) \bigr ] } \nonumber \\ & = & { { \rm tr } \bigl [ \rho_0 u_i^{-1}(\eta,\eta_0 ) { \cal o } u_i(\eta,\eta_0 ) \bigr ] \over { \rm tr } \bigl [ \rho_0 \bigr ] } \label{oevolve}\end{aligned}\ ] ] since the state @xmath101 does not evolve before the interactions are turned on , we can insert the identity in the form @xmath109 with @xmath110 and commute one of these evolution operators with @xmath111 to obtain @xmath112 \over { \rm tr } \bigl [ \rho_0 \bigr ] } \label{oevolveii}\ ] ] inserting another factor of the identity , @xmath113 , with @xmath114 yields @xmath115 \over { \rm tr } \bigl [ \rho_0 \bigr ] } \label{oevolveiii}\ ] ] and finally we let @xmath116 and @xmath117 , which represent the infinite past and infinite future in conformal coordinates , so that @xmath118 \over { \rm tr } \bigl [ u_i(-\infty,0 ) u_i(0 , -\infty ) \rho_0 \bigr ] } \label{oevolveiv}\ ] ] reading the operators from right to left in the numerator of this equation , @xmath111 sets the initial state of the system which is then evolved along a time contour from @xmath119 to @xmath120 with an operator inserted at @xmath98 ; the final operator evolves back from @xmath120 to @xmath119 . the closed time contour which results is depicted in fig . [ ctp ] . at which time the interactions are turned on . we double the field content so that separate copies of the fields are used for the upper and lower parts of the contour.[ctp ] ] to evaluate eq . ( [ oevolveiv ] ) it is useful to group the evolution operators into a single time - ordered exponential . this is accomplished by formally doubling the field content of the theory , with a set of ` @xmath121 ' fields on the increasing - time contour and a set of ` @xmath122 ' fields on the decreasing - time contour . the arrows on the contour indicate time ordering of events so that events on the @xmath122 contour always occur after those on the @xmath121 contour . we can group the effects of both parts of the contour together by writing the interacting part of the action appearing in dyson s equation , as @xmath123 since the two terms differ only in the direction of the integral over the conformal time , we can write the action as a single lagrange density , @xmath124 . \label{action}\ ] ] the field doubling induced by the closed contour effectively doubles the number of vertices we must include when studying any process one set with fields on the @xmath121 branch and one with fields on the @xmath122 branch . from the eq . ( [ action ] ) , the latter will have couplings with the opposite sign . in evaluating matrix elements , wick contractions produce four propagators for the possible contractions of pairs of the two types of fields , @xmath125 the time - ordering of the contractions is determined by the direction along the contour , @xmath126 with the wightman functions given in eq . ( [ gkpmexpand ] ) . assembling the ingredients of the schwinger - keldysh formalism the general expression for an operator expectation value in eq . ( [ oevolveiv ] ) , dyson s equation ( [ dyson ] ) and eq . ( [ action])provides an explicit expression for the evolution of @xmath127 . if we let the initial density matrix be that for a pure @xmath0-vacuum , then eq . ( [ oevolveiv ] ) becomes @xmath128 } \right\ } | \alpha\rangle \over \langle\alpha | t \left\ { e^{-i\int_{-\infty}^0 d\eta\ , [ h_i(\phi^+ ) - h_i(\phi^- ) ] } \right\ } | \alpha\rangle } . \label{omatrix}\ ] ] here we have absorbed any ` turning on ' function in @xmath91in essence the time integrals begin at @xmath87 . the time ordering has allowed us to group the time evolution operators in eq . ( [ oevolveiv ] ) along the two contours into a single operator . this equation for the finite evolution of the expectation value of @xmath129 is the analogue of the standard @xmath1-matrix expression used in minkowski space . the virtue of the field doubling is that it removes any acausal behavior from the matrix element since the @xmath99-functions in the propagators combine to limit the upper end of the conformal time integrals to @xmath98 @xcite . this property will become clear in the calculation of a specific example . we now evaluate the expectation value of the number of euclidean particles in the @xmath0-vacuum using the schwinger - keldysh formalism @xcite . this number operator provides a good measure of whether a particular choice for the vacuum state becomes pathological in the presence of interactions . from the perspective of the euclidean vacuum , the @xmath0-vacuum is an excited state . we can determine the stability of this state when the interactions are turned on by following the time evolution of the change in the euclidean number operator evaluated in the @xmath0-vacuum . in the non - interacting theory , we should not encounter any infinite euclidean particle production in the @xmath0-vacuum . when the interactions are turned on , some further particle production will occur in the euclidean vacuum , but the rate per unit volume should be finite and any divergences which appear perturbatively must be renormalizable . what we shall discover is that in the interacting theory , the @xmath0-vacuum produces a new class of divergences that can not be removed with the usual set of counterterms . these divergences in the @xmath0-vacuum occur at each time in the integrand as we propagate from some initial state at @xmath87 to a finite time later so that the theory diverges even for an arbitrarily short time after the interactions are turned on . this behavior indicates that the @xmath0-vacuum is inappropriate even for a finite inflationary epoch . although we examine in detail the evolution of the euclidean number operator in this section , all of the divergences we find are generic to the one loop corrections to an arbitrary operator . furthermore , while we consider a scalar theory with a cubic interaction since it has a simple , non - trivial self - energy correction , similar loop integrals occur in any interacting scalar field theory . these more general cases are treated in the next section . we first show how the free hamiltonian can produce a non - trivial , but finite , time dependence . we then set up the calculation for the expectation value of the derivative of the number operator , expressing it in terms of the scalar field and its conjugate momentum . the corrections to this expectation value to leading , non - trivial order in the coupling are then calculated and we obtain general expressions for the one - loop corrections and the counterterms . in the interaction picture , the hamiltonian is divided into free and interacting parts , @xmath130 here we have included a @xmath99-function so that before @xmath87 the system evolves freely ; we are always free to take @xmath88 . for conformally flat de sitter coordinates , the free hamiltonian for a scalar field @xmath131 of mass @xmath132 is given by @xmath133 . \label{freeham}\ ] ] an important difference between de sitter space and flat space is that the free hamiltonian is not diagonal in terms of creation and annihilation operators , @xmath134 g_k^e(\eta ) \nonumber \\ & & \qquad\qquad + a_{\vec k}^e a_{-\vec k}^e f_k^e(\eta ) + a_{\vec k}^{e\dagger } a_{-\vec k}^{e\dagger } f_k^{e*}(\eta ) \biggr\ } \qquad \label{freehamilton}\end{aligned}\ ] ] where @xmath135 u_k^e u_k^{e * } \nonumber \\ f_k^e & = & \partial_\eta u_k^e \partial_\eta u_k^e + \left [ k^2 + { m^2\over\eta^2 } \right ] u_k^e u_k^e . \label{gkfka}\end{aligned}\ ] ] note that this property holds also for the @xmath0-vacuum , @xmath136 . the origin of the off - diagonal terms in the free hamiltonian lies in the fact that the surfaces of constant conformal time are not orthogonal to the generator of an isometry . this effect introduces an additional non - trivial source of time evolution which combines with that produced by the evolution of the state @xmath53 when the fields interact . for a time derivative of a generic operator , we formally have @xmath137 \nonumber \\ & = & i\ , { \rm tr}\ , \bigl [ - [ h_i , \rho ] { \cal o } + \rho [ h_0 , { \cal o } ] \bigr ] \nonumber \\ & = & i\ , { \rm tr}\ , \bigl [ \rho [ h , { \cal o } ] \bigr ] = i \langle [ h , { \cal o } ] \rangle . \label{odotdef}\end{aligned}\ ] ] the off - diagonal terms in the free hamiltonian induce an evolution in the number operator even in the free theory . since we wish to explore the effect of interactions on the @xmath0-vacuum , we construct the number operator from creation and annihilation operators , @xmath138 and @xmath139 , which satisfy @xmath140 at the moment the interactions are turned on . the subsequent evolution in the interaction picture is given by the solution to @xmath141 . \label{aheisen}\ ] ] the solution to this equation can be formally expressed as @xmath142 where @xmath143 is the time evolution operator for the free part of the theory , @xmath144 from the form for the free hamiltonian , a general solution to eq . ( [ aheisen ] ) is given by a bogolubov transformation of the time independent creation and annihilation operators , @xmath145 where the coefficients satisfy @xmath146 the standard normalization of the commutator of the transformed creation and annihilation operators also requires that @xmath147 the general solution to the coefficient equations ( [ bogolcoef ] ) is of the form @xmath148 where the constants @xmath149 and @xmath150 should satisfy @xmath151 from eq . ( [ bogolnorm ] ) . if we would like the number operator to count the number of euclidean particles at the moment we turn on the interactions , @xmath152 , then @xmath153 the number operator constructed from the transformed creation and annihilation operators , @xmath154 , has the correct evolution for an interaction picture operator . using eq ( [ odotdef ] ) , even the free hamiltonian induces some evolution in this number operator , @xmath155 = { i\over\eta^4 } d_{ab } \int d^3\vec x d^3\vec y\ , e^{i\vec k\cdot ( \vec x-\vec y ) } \phi(\eta_a,\vec x ) \phi(\eta_b,\vec y ) , \label{hocommute}\ ] ] where @xmath156 is the differential operator , @xmath157 here @xmath158 are only labels to indicate the functions on which the derivatives act ; in the end , @xmath158 are set equal to @xmath98 . the functions @xmath159 are @xmath160 | u_k^e(\eta_0 ) |^2 \nonumber \\ p_3(\eta ) & = & - { \eta^2\over\eta_0 ^ 2}\ , \left [ k^2 + { m^2\over\eta^2 } \right ] \partial_\eta |u_k^e(\eta_0)|^2 . \label{pthree}\end{aligned}\ ] ] evaluating eq . ( [ hocommute ] ) in the @xmath0-vacuum gives the tree level evolution of the number operator in this state , @xmath161 | \alpha\rangle = - { 1\over\eta^4 } v d_{ab}(\eta ) u_k^\alpha(\eta_a ) u_k^{\alpha * } ( \eta_b ) \label{ntree}\ ] ] @xmath162 is the spatial volume . we can divide the volume from both sides to yield the particle production rate per unit volume for which this tree contribution is completely finite , if non - zero . now consider a cubic interaction with its associated counterterms , @xmath163 . \label{hintdef}\ ] ] the cubic vertex will generally introduce one - loop corrections to the two - point functions , so we have included a mass counterterm , @xmath164 . since the @xmath165 interaction breaks the @xmath166 symmetry of the free theory , we also expect the interaction to generate graphs containing tadpole insertions which are cancelled with the correct choice for @xmath167 . to the order we shall study , no wave function renormalization is needed . the change in the number operator induced by these interactions is given by applying eq . ( [ omatrix ] ) , @xmath168 e^{-i\int_{\eta_0}^0 d\eta\ , [ h_i(\phi^+ ) - h_i(\phi^- ) ] } \right\ } | \alpha\rangle \over \langle\alpha | t \left\ { e^{-i\int_{\eta_0}^0 d\eta\ , [ h_i(\phi^+ ) - h_i(\phi^- ) ] } \right\ } | \alpha\rangle } . \nonumber\end{aligned}\ ] ] the evolution of the free field is simple in the interaction picture , so it is useful to write the creation and annihilation operators in terms of the field and its conjugate momentum , by expanding in the time independent operators , eq . ( [ bogol ] ) , and inverting the operator expansion in eq . ( [ phimodes ] ) , @xmath169 \nonumber \\ % a_{\vec k}^{e\dagger } & \ ! = \ ! & - i \int d^3\vec x\ , e^{i\vec k\cdot \vec x } \left [ u_k^e\ , \pi(\eta,\vec x ) - \eta^{-2 } \partial_\eta u_k^e \phi(\eta,\vec x ) \right ] . \nonumber \\ & & \label{invert}\end{aligned}\ ] ] the commutator with the free part of the hamiltonian is then as was given in eq . ( [ hocommute ] ) while that with the interacting part is @xmath170 & = & { i\over\eta^4 } j\ , ( 2\pi)^3 \delta^3(\vec k ) \int d^3\vec x\ , \bigl [ 2 |u_k^e(\eta_0)|^2 \pi(\eta,\vec x ) - \eta_0^{-2 } \partial_\eta |u_k^e(\eta_0)|^2 \phi(\eta,\vec x ) \bigr ] \label{hicommute } \\ % & & + { i\over\eta^4 } \delta m^2 \int d^3\vec x d^3\vec y\ , e^{i\vec k\cdot ( \vec x-\vec y ) } \bigl [ |u_k^e(\eta_0)|^2 \left [ \phi(\eta,\vec x ) \pi(\eta,\vec y ) + \pi(\eta,\vec x ) \phi(\eta,\vec y ) \right ] \nonumber \\ & & \qquad\qquad\qquad - \eta_0^{-2 } \partial_\eta |u_k^e(\eta_0)|^2 \phi(\eta,\vec x ) \phi(\eta,\vec y ) \bigr ] \nonumber \\ % & & + { i\over\eta^4 } { \lambda\over 2 } \int d^3\vec x d^3\vec y\ , e^{i\vec k\cdot ( \vec x-\vec y ) } \bigl [ |u_k^e(\eta_0)|^2 \left [ \phi^2(\eta,\vec x ) \pi(\eta,\vec y ) + \pi(\eta,\vec x ) \phi^2(\eta,\vec y ) \right ] \nonumber \\ & & \qquad\qquad\qquad - \eta_0^{-2 } u_k^e(\eta_0 ) \partial_\eta u_k^{e*}(\eta_0 ) \phi(\eta,\vec x ) \phi^2(\eta,\vec y ) - \eta_0^{-2 } u_k^{e*}(\eta_0 ) \partial_\eta u_k^e(\eta_0 ) \phi^2(\eta,\vec x ) \phi(\eta,\vec y ) \bigr ] . \nonumber \end{aligned}\ ] ] the expectation values of each of these commutators will be evaluated in the @xmath0-vacuum to order @xmath171 . before evaluating the expectation values of these commutators perturbatively , a large class of graphs , those containing a tadpole subgraph , are eliminated through the proper choice of the coefficient @xmath167 of the linear counterterm . to leading order in @xmath172 , this choice for @xmath167 is @xmath173 this cancellation is shown diagrammatically in fig . [ tadpole ] . note that while the loop integral in eq . ( [ jtadpole ] ) contains an apparent time dependence , it is in fact time independent , @xmath174 of the linear counterterm is chosen to cancel insertions of tadpoles . the dashed line represents a line in a general diagram . [ tadpole ] ] the form of the leading corrections to the expectation value of @xmath175 $ ] in the @xmath0-vacuum are simpler since the commutator is already itself of order @xmath172 . the order @xmath171 corrections to this commutator are shown diagrammatically in fig . [ loopi ] . the other corrections from the cubic interaction at this order contain tadpole subgraphs and are cancelled when eq . ( [ jtadpole ] ) is satisfied . to the expectation value of @xmath176 $ ] in the @xmath0-vacuum . the first represents a self - energy graph while the second is from the mass counterterm . [ loopi ] ] one of the subtleties in evaluating the evolution of the number operator is that it contains time derivatives . for example , the @xmath164 term of fig . [ loopi ] contains a term of the form @xmath177 which can produce schwinger terms if the time ordering of the operators is not treated carefully . a method for avoiding such terms , following @xcite , is to write the canonical momentum as @xmath178 and then to place the fields on the appropriate contours so that the time ordering naturally given along the contour preserves the correct ordering of the operators in eq . ( [ timepi ] ) , @xmath179 with this prescription , the mass counterterm in fig . [ loopi ] contributes @xmath180 at lowest order . the only non - trivial effect of the interactions on expectation value of the commutator with the interaction hamiltonian arises from the self - energy graph , @xmath181 \nonumber \\ % & & + { \lambda^2\over\eta^4 } { v\over 8\pi^3 } { \partial_\eta |u_k^e(\eta_0)|^2\over\eta_0 ^ 2 } \nonumber \\ & & \quad\times \int_{\eta_0}^\eta { d\eta'\over\eta^{\prime 4}}\ , { \rm i m } \bigl [ u_k^\alpha(\eta ) u_k^{\alpha * } ( \eta ' ) l_k^\alpha(\eta,\eta ' ) \bigr ] \qquad \label{aone}\end{aligned}\ ] ] where the loop integral is given by @xmath182 in addition to the corrections in eq . ( [ cone ] ) and eq . ( [ aone ] ) , the fact that the free hamiltonian is not the conserved quantity associated with a time - like generator of an isometry of de sitter space means that the expectation value of @xmath183 $ ] also contributes at order @xmath171 . the linear counterterm also cancels the graphs containing a tadpole subgraph and any vacuum to vacuum disconnected graphs are removed by the denominator of eq . ( [ matrix ] ) . the only remaining corrections then are those shown in fig . [ loop0 ] . corrections to the expectation value of @xmath184 $ ] in the @xmath0-vacuum . again , the first represents a self - energy graph while the second is from the mass counterterm . [ loop0 ] ] these graphs contribute @xmath185 \nonumber \\ & & \qquad\qquad\times { \rm i m } \bigl [ u_k^\alpha(\eta_b ) u_k^{\alpha * } ( \eta_2 ) l_k^\alpha(\eta_1,\eta_2 ) \bigr ] \qquad \label{anaught}\end{aligned}\ ] ] and @xmath186 \qquad \label{cnaught}\end{aligned}\ ] ] to the expectation value of the derivative of the number operator , @xmath187 . an important feature to note is that the mass counterterm insertions , in eq . ( [ cone ] ) and eq . ( [ cnaught ] ) , do not vanish in the euclidean limit ; they are needed to cancel a logarithmic divergence in the corresponding self - energy diagrams . the one loop corrections in the generic @xmath0-vacuum and in the euclidean limit differ in their divergence structure . both cases contain a logarithmic divergence which can be regularized and cancelled by the appropriate choice for the mass counterterm , @xmath164 . what distinguishes the general @xmath0-vacuum from the euclidean vacuum is the appearance of an additional class of terms that diverge linearly in the loop momentum . this divergence can not be renormalized . in this subsection , we summarize the renormalization of the logarithmic divergence for the case of the massless conformally coupled scalar field in the euclidean vacuum . the detailed calculation for the @xmath0-vacuum is left for the appendix [ dimen ] . formally , the evolution of the euclidean number operator to order @xmath171 in the euclidean vacuum is given by @xmath188 \label{eucllim}\end{aligned}\ ] ] the case of a massless , conformally coupled scalar field is most readily analyzed since the mode functions have a simple form given in eq . ( [ uehalfnu ] ) . the tree contribution in this case is given by @xmath189 \label{ntreenuhalf}\ ] ] which is finite for @xmath190 . the loop integral for @xmath46 , which occurs in @xmath191 , is @xmath192 so that @xmath193 \int_{\eta_0}^\eta { d\eta'\over\eta-\eta'}\ , \cos 2k(\eta-\eta ' ) \nonumber \\ & & - { \lambda^2\over k } { v\over 32\pi^2 } { \eta_0 ^ 2\over\eta^3 } \int_{\eta_0}^\eta { d\eta'\over\eta-\eta'}\ , \sin 2k(\eta-\eta ' ) . \label{aoneenuhalf}\end{aligned}\ ] ] the corresponding counterterm is @xmath194 . \label{coneenuhalf}\ ] ] the second integral in eq . ( [ aoneenuhalf ] ) is completely finite at all times @xmath195 . the first integral , however , contains a logarithmic divergence at the upper end of the @xmath196 integration . this term can be regularized as described in the appendix [ dimen ] . the pole as the regularization is removed , @xmath197 , is given by @xmath198 { 1\over\epsilon } + { \rm finite } , \label{aoneenuhalfpole}\ ] ] and is cancelled by choosing @xmath199 in eq . ( [ coneenuhalf ] ) . the analogous logarithmic divergence in @xmath200 is cancelled by the counterterm @xmath201 . the self - energy diagrams in the @xmath0-vacuum also contain a logarithmic divergence which can be removed by a suitable choice for the mass counterterm , @xmath202 which reduces to that for the euclidean case in eq . ( [ euclmct ] ) . the origin and regularization of this divergence is discussed more fully in appendix [ dimen ] . including the order @xmath171 corrections , the derivative of the number of euclidean particles in the @xmath0 vacuum is again given by the sum of the contributions shown in fig . [ loopi ] and fig . [ loop0 ] as well as the tree level contribution of eq . ( [ ntree ] ) , @xmath203 each self - energy graph contains a loop integral , eq . ( [ loopdef ] ) . unlike the euclidean case which is completely finite once we have established an @xmath50 prescription , the loop integral over @xmath0-mode functions diverges linearly in the spatial momentum . introducing a bound @xmath204 to remove the large momenta in the loop , @xmath205 , assuming @xmath206 , the divergent part of eq . ( [ loopdef ] ) is @xmath207 \nonumber \\ & & + \ { \rm finite } . \nonumber\end{aligned}\ ] ] the appearance of the factor @xmath208 shows why such a divergent term does not arise in the euclidean limit . unlike the logarithmic divergence , this divergence can not be removed by a momentum independent value for @xmath164 . for example , the divergent piece of the self - energy graph in fig . [ loopi ] is @xmath209 \nonumber \\ & & + e^{\alpha+\alpha^ * } n_\alpha^4 { \lambda^2\over k\eta^2 } { v\lambda\over 4\pi^2 } { \partial_\eta |u_k^e(\eta_0)|^2\over\eta_0 ^ 2 } \nonumber \\ & & \int_{\eta_0}^\eta { d\eta'\over\eta^{\prime 2}}\ , { \rm im}\ , \bigl\ { u_k^e(\eta ) u_k^{e*}(\eta ' ) \bigr\ } \nonumber \\ & & \qquad\times \left [ { \sin k(\eta-\eta')\over\eta-\eta ' } + { \sin k(\eta+\eta')\over\eta+\eta ' } \right ] + \cdots \nonumber \end{aligned}\ ] ] which can not be cancelled by a momentum - independent choice for @xmath164 in eq . ( [ cone ] ) . to observe this incompatibility of the momentum dependence of the divergences in the @xmath0-vacua loop corrections and the available counterterms , at least in the case of a massless , conformally coupled scalar field , it is sufficient to compare the limiting behavior in the external momentum of the loop corrections and the corresponding counterterms . in the @xmath210 limit the divergent parts of both @xmath191 and @xmath191 scale as @xmath211 . in contrast , the leading @xmath212-dependence of the counterterms scale as @xmath213 , at least when @xmath0 is real , which is required in any case for a @xmath214 invariant theory @xcite . thus , no choice of @xmath164 , which does not depend on @xmath212 , is possible such that @xmath215 and @xmath201 cancel the divergence in @xmath191 and @xmath216 as @xmath217 . note that the linear divergence is not present in the opposite quantity the expectation value of the number of @xmath0-particles in the euclidean vacuum . this quantity is given by expressions similar to those above except with the labels interchanged for the mode functions , @xmath218 , and some of the @xmath0-dependent coefficients are slightly altered in eqs . ( [ cone])([cnaught ] ) . the crucial difference is that the presence of the euclidean states in eq . ( [ omatrix ] ) leads to euclidean propagators so that in particular the loop integral is over euclidean modes , as in eq . ( [ euclloop ] ) , for which no linear divergence occurs . the linear divergence that arises from the one loop corrections in the @xmath0-vacuum is a uv effect . at arbitrarily short distances there exists an interference of the positive and negative frequency modes which cancels the rapidly oscillating phases among some of the terms within the loop integral . without such a cancellation , these phases could damp these high - momentum contributions through an appropriate @xmath50 prescription . this interference of phases is a specific feature of the propagator in the @xmath0-vacuum and does not occur in the euclidean case . in this section we shall discuss the origin of this divergence and determine the necessary conditions for it to arise . consider a loop containing @xmath65 vertices connected by @xmath65 internal propagators those through which the common loop momentum flows . since in de sitter space it is convenient to perform a fourier transform over only the spatial coordinates , each vertex has a time , @xmath219 for @xmath220 , associated with it . eventually , we integrate over all these times as they arise from the exponent of the time evolution operator in eq . ( [ omatrix ] ) . to determine whether a particular loop can produce a uv divergence , we must first count the powers of momentum in the high loop momentum region . let @xmath221 represent the wightman function within a loop propagator connecting the @xmath222 and @xmath223 vertices . the loop momentum is @xmath224 and @xmath225 denotes other momenta following through the @xmath222-th leg . in the uv limit @xmath226 . \nonumber\end{aligned}\ ] ] note that the propagator contains factors of both @xmath221 and @xmath227 . for the purpose of the power counting , it is important to note that , aside from the phases , in the uv limit @xmath228 . thus in integrating over a loop , @xmath229 we only encounter a possible uv divergence if @xmath230 . note that the @xmath231 case can be removed by a counterterm since the loop only depends on the loop momentum and not on any other momenta in the graph . the @xmath232 case can produce a linear divergence . since the divergence only depends on the form of the propagator and not the form of the interaction , such divergences generically occur in any interacting theory , for example in the processes shown in fig . [ loop0 ] and fig . [ loop4 ] . ) . this result follows from the structure of the propagator in the @xmath0-vacuum and not the form of the interactions . [ loop4 ] ] superficially , a divergence might seem possible even in the euclidean case if a product of wightman functions , @xmath233 , occurs in the loop integral . however , the schwinger - keldysh formalism is constructed to remove such terms and only factors of @xmath234 , and @xmath235 can appear in the loop integral in which the @xmath236-dependent phases do not cancel in the uv . the important difference in the @xmath0-vacuum is that each wightman function contains terms whose @xmath236-dependent phases have the opposite signs . thus even after the schwinger - keldysh formalism has been applied , a product of the two loop green s functions , @xmath237 + \cdots \nonumber\end{aligned}\ ] ] will have some phase cancellation as @xmath238 . what renders these terms unrenormalizable , however , is that although the @xmath236-dependent phases cancel in the divergent terms , they still contain a non - trivial dependence on the momentum entering the loop from the rest of the diagram . this analysis of the phase structure of the two - propagator loop does not necessarily show that such divergences can not be removed from the theory through a suitable renormalization prescription . this fact can only be established by summing all the contributions to this graph and demonstrating that the dependence of the resulting divergent term on the momenta external to the loop is incompatible with a counterterm insertion , as was done in the previous section . however , since all loop integrals containing only two propagators have essentially the same structure , given by eq . ( [ loopdef ] ) , we see that the @xmath0-vacuum can not be renormalized in any interacting theory , regardless of the form of the interaction . the power counting argument indicates that a logarithmic divergence can arise for a loop with three legs , such as the vertex correction graph shown in fig . [ vertex ] . the loop integral contains a product of three terms of the form of eq . ( [ highpofwight ] ) , or its complex conjugate , which generally contains terms where the high loop momentum dependence of the phase cancels . theory could generate such a divergence . [ vertex ] ] for example , in fig . [ vertex ] if the coordinates associated with the three vertices are @xmath239 , @xmath240 and @xmath241 , then in the product of three propagators occur terms such as @xmath242 in the high momentum region of the loop momentum , the phase factor will be independent of the integrated momentum and the integral will be logarithmically divergent . as with the self - energy case before , these arguments can only demonstrate under what conditions a divergence can occur . whether these logarithmic terms cancel among each other or whether the resulting divergence can be removed by a counterterm requires performing the full integration and summing all the relevant products of wightman functions . however , since we have already seen that the self - energy graphs exhibit a pathological behavior in the @xmath0-vacuum , we shall not study these vertex corrections further here . the preceding discussion shows that a class of linear divergences from loops with two propagators and logarithmic divergences from loops with three propagators generically appears in any interacting theory in an @xmath0-vacuum . these divergences arise from the form of the @xmath0-vacuum propagator , which is determined by the free field hamiltonian , and not on the detailed form of the interaction . what the schwinger - keldysh formalism allows is a precise statement of the problem of an interacting theory in an @xmath0-vacuum . with a high - momentum cut off , we can find all the terms that diverge linearly with this cut off and analyze their dependence on the momenta external to the loop . the resulting expressions are not cancelled by a set of de sitter invariant counterterms of the same form as those in the original lagrangian . the appearance of @xmath0-dependent prefactors also shows why such terms do not plague the euclidean vacuum . the fact that the divergence originates from high momentum modes and the form of the @xmath0-dependent prefactors provides the basis of a simple heuristic explanation for the divergence . the number density of euclidean particles per unit volume at the time that the interactions are turned on is @xmath243 since @xmath244 . from the perspective of the euclidean vacuum , the @xmath0-vacuum looks like a distribution whose occupation number is given by @xmath245 . note that @xmath246 is actually independent of @xmath71 . if we then replace the @xmath0-dependent terms in the propagators with the factors @xmath246 , maintaining the momentum labels , then among the many divergent terms contributing to @xmath191 occurs the expression , @xmath247 \nonumber \\ & & \quad\times \sin\bigl[[p+k-|\vec p - \vec k|](\eta-\eta ' ) \bigr ] \nonumber \\ & & + \cdots , \label{kinetic}\end{aligned}\ ] ] in the massless conformally coupled case . this expression resembles a ` gain minus loss ' process in the @xmath0-background for example , one part describes the creation of two particles from one while the other describes the creation of one from the annihilation of two . since @xmath248 is constant , nothing suppresses the large @xmath236 divergence . this divergence is only present in a true @xmath0-vacuum and not in a ` truncated @xmath0-vacuum'a state that is set equal to a euclidean vacuum above some scale @xmath249 @xcite . for a truncated @xmath0-vacuum , the @xmath248 s vanish above @xmath250 so integrals such as eq . ( [ kinetic ] ) become finite . the largest contribution to the change in the number operator scales as @xmath251 . these truncated @xmath0-vacua have no divergences , although exactly how the state evolves may depend on how the truncation is implemented . using our formalism , it becomes possible to study how one of these vacua evolves during inflation and the amount by which it would alter the appearance of angular power spectrum of the cosmic microwave background radiation @xcite . to conclude then , interactions destabilize the @xmath0-vacua . we have calculated a physical quantity , the conformal time rate of change of the number of euclidean mode particles in the @xmath0-vacuum , and found that it diverges . this divergence reflects a physical pathology of an interacting theory in a true @xmath0-vacuum . this work was supported in part by doe grant de - fg03 - 91-er40682 . we are grateful to dan boyanovsky for helpful discussions . r. h. thanks m. einhorn , f. larsen and jim cline for helpful discussions , as well as the aspen center for physics where this work was begun . the self - energy diagrams in the fig . [ loopi ] and fig . [ loop0 ] contain two classes of divergences in the @xmath0-vacuum . one class diverges linearly in the magnitude of the spatial loop momentum . the dependence of this divergence on the external momentum flowing through the graph is not of the same form as that appearing from the insertion of one of the available counterterms . in this sense , these divergences can not be renormalized and indicate a pathological feature of an interacting theory in the @xmath0-vacuum . however , these linearly divergent terms vanish in the euclidean limit . the second class of divergences exist in both the euclidean and the general @xmath0-vacua . these divergences depend logarithmically on the conformal time and can , at least in the massless conformally coupled case where the calculation simplifies , be removed by a constant mass counterterm . it is important to establish the renormalization of this type of divergence not only to show that the evolution of the number operator is finite for an interacting scalar field in the euclidean vacuum , but also in the @xmath0-case . if we wish to consider a ` truncated @xmath0-vacuum ' , one which is cut off at some high energy scale such as the planck mass @xcite , we thereby remove the linear divergence of eq . ( [ loopdivg ] ) , but the logarithmic divergence is still present in @xmath191 and @xmath216 . in this appendix , we demonstrate how to renormalize this divergence . in the massless conformally coupled case , loop integral over @xmath0-vacuum propagators of eq . ( [ loopdef ] ) yields @xmath252 { e^\alpha e^{ik(\eta_1+\eta_2 ) } + e^{\alpha^ * } e^{-ik(\eta_1+\eta_2 ) } \over \eta_1+\eta_2 } \nonumber \\ & & - { i\pi\over k } e^\alpha n_\alpha^4 \eta_1\eta_2 \sin k\eta_1 \left [ e^{ik\eta_2 } - e^{2\alpha^ * } e^{-ik\eta_2 } \right ] \nonumber \\ & & + { i\pi\over k } e^{\alpha^ * } n_\alpha^4 \eta_1\eta_2 \sin k\eta_2 \left [ e^{-ik\eta_1 } - e^{2\alpha } e^{2ik\eta_1 } \right ] \nonumber \\ & & + { 2\pi\over k } ( \lambda - k ) e^{\alpha+\alpha^ * } n_\alpha^4 ( \eta_1\eta_2)^2 \nonumber \\ & & \qquad\qquad\times \left [ { \sin k(\eta_1-\eta_2)\over \eta_1-\eta_2 } + { \sin k(\eta_1+\eta_2)\over \eta_1+\eta_2 } \right ] . \label{loopalpha } \end{aligned}\ ] ] the final term is the linearly divergent term . among the remaining terms , only the first produces any logarithmic divergence when integrated over the conformal time . in the case of either self - energy contribution , eq . ( [ aone ] ) or eq . ( [ anaught ] ) , we integrate @xmath253 from @xmath87 to @xmath254 ( @xmath255 for the @xmath191 graph ) and encounter a singularity from the @xmath256 denominator of imaginary part of the first term in eq . ( [ loopalpha ] ) , @xmath257 inserting this result in the self - energy contributions yields @xmath258 and @xmath259 \nonumber \\ & & + \cdots \label{logofanaught}\end{aligned}\ ] ] in both of these equations , the ellipses indicate terms which do not diverge logarithmically in the conformal time . both eq . ( [ logofaone ] ) and eq . ( [ logofanaught ] ) contain divergent integrals of the form @xmath260 where @xmath261 is @xmath212 , @xmath120 , or @xmath262 . changing to a dimensionless variable , @xmath263 eq . ( [ logdiverge ] ) becomes @xmath264 we can regularize this integral by inserting a factor of @xmath265 in the integrand and then extract the pole as @xmath197 , @xmath266 thus , @xmath267 applying this regularization to the self - energy graphs gives @xmath268 \nonumber \\ & & + \cdots \label{poleofaone}\end{aligned}\ ] ] and @xmath269 in deriving eq . ( [ poleofanaught ] ) we have used the fact that the operator @xmath156 is symmetric in @xmath270 and @xmath271 . both poles are removed by the appropriate mass counterterm graphs @xmath215 and @xmath201 given in eq . ( [ cone ] ) and eq . ( [ cnaught ] ) , respectively , when @xmath272 r. easther , b. r. greene , w. h. kinney and g. shiu , phys . d * 64 * , 103502 ( 2001 ) [ hep - th/0104102 ] ; r. easther , b. r. greene , w. h. kinney and g. shiu , phys . d * 67 * , 063508 ( 2003 ) [ hep - th/0110226 ] .

we have numerically investigated the doped- ladder using exact diagonalization . we have studied both the limit of strong inter - chain coupling and isotropic coupling . the ladder scales to the luther - emery liquid regime in the strong inter - chain coupling limit . in this strong coupling limit there is a simple picture of the excitation spectrum that can be continued to explain the behavior at isotropic coupling . at we have indications of a ferromagnetic ground state . at a large the ladder is phase separated into holes and a heisenberg ladder . at intermediate coupling the ground state shows hole pairing with a modified-wave symmetry . the excitation spectrum separates into a limited number of quasiparticles which carry charge and spin and a triplet magnon mode . at half - filling the former vanish but the latter evolves continuously into the magnon band of the spin liquid . at low doping the quasiparticles form a dilute fermi gas with a strong attraction but simultaneously the fermi wave vector , as would be measured in photoemission , is large . the dynamical structure factors are calculated and are found to be very similar to calculations on 2d clusters .

cond-mat9510150

the properties of strongly correlated electrons confined to a ladder ( or double chain ) and described by @xmath0-@xmath1 or hubbard models have been the subject of intensive investigation recently . the reason lies in the unusual spin liquid nature of the undoped parent system . another reason for especial interest is weakly coupled ladders compounds like @xmath7 and @xmath8.@xcite recent measurements of the magnetic susceptibility and the nuclear spin relaxation rate in these materials show the existence of a finite spin gap . the key question in the current study is the evolution of the finite gap in the spin excitation spectrum upon doping . the spin gap remains in other spin liquids systems and is a sign of strong superconducting fluctuations.@xcite a recent analysis of the @xmath0-@xmath1 ladder using a mean - field theory with gutzwiller renormalization of the matrix elements to account for the strong correlations , gave a continuous evolution of the spin gap with doping.@xcite the short range resonance valence bond ( rvb ) state evolves into a superconductor with modified @xmath4-wave symmetry within this mean - field approximation . a tendency towards modified @xmath4-wave superconductivity was also found in a bosonization approach@xcite and in a recent numerical study of the hubbard ladder.@xcite we have investigated @xmath0-@xmath1 ladders up to a size of @xmath9 sites using a lanczos diagonalization method . first results have been published in ref . . here we report in more detail our results for larger lattices including a detailed investigation of the excitation spectrum , a discussion of phase separation and the calculation of the superconducting order parameter and of the form factor of the cooper pairs . we find clear evidence of hole pairing and a modified @xmath4-wave rvb state in lightly doped systems in agreement with the mean - field theory . an interesting difference however is the discontinuous evolution of the excitation spectrum upon doping . quasiparticle " excitations appear carrying both charge and spin . these excitations are in addition to a band of magnons which evolve continuously away from the undoped spin liquid . this separation of the excitation spectrum into bound holon - spinon quasiparticles and collective magnon excitation contrasts with the full spin - charge separation found in a luttinger liquid . the @xmath0-@xmath1 ladder hamiltonian is @xmath10 where @xmath11 runs over @xmath12 rungs , and @xmath13 @xmath14 and @xmath15 @xmath16 are spin and leg indices . the @xmath0-@xmath1 ladder is sketched in fig . [ fig : geometry ] . the first two terms are the kinetic energies and the @xmath1 @xmath17 are exchange couplings along the ladder ( rungs ) . unless noted otherwise we set @xmath18 . the projection operator @xmath19 prohibits double occupancy of a site . periodic or antiperiodic boundary conditions ( pbc , apbc ) are used along the ladder . the wave vector @xmath20 is consequently well defined , @xmath21 and @xmath22 being the momenta along the ladder and rungs . the transverse momentum @xmath22 takes only the values @xmath23 and @xmath24 , corresponding to bonding and antibonding states . at half filling the @xmath0-@xmath1 ladder is equivalent to the heisenberg ladder , which was investigated in earlier publications . the ground state of the heisenberg ladder is a short range rvb state with a spin gap of @xmath25 at isotropic coupling , @xmath26 . the strong coupling limit @xmath27 is a good starting point to describe the system as there a simple description of the spectrum is available.@xcite in that limit , each eigenfunction of the total system can be written as a direct product of one - rung states , which are either spin singlets or one of the triplets , and the ground state is that with all singlets . the first excited multiplet consists of the states with one triplet rung . a small but finite value of @xmath1 lifts the degeneracy of these states . the one - magnon excitations then form a three - fold spin degenerate band with dispersion @xmath28 up to second order in @xmath1 . it has a minimum gap @xmath29 at @xmath30.@xcite the momentum perpendicular to the chains is @xmath31 . the higher excited states form a continuum of excited states and its minimum is at @xmath32 with energies slightly larger than twice the gap @xmath33 . with increasing @xmath1 the collective excitation branch crosses into the continuum , but the qualitative description is still valid . in this paper we study the effects of doping holes into such a ladder . although the isotropic case , @xmath34 , is of most interest , we also study the limit @xmath35 , which can be easily understood . in this limit the problem reduces to a system of weakly coupled rungs . the properties can be continuously followed down to the isotropic point @xmath36 . this paper is organized as follows . in sec . [ sec : ferro ] we briefly discuss the occurrence of ferromagnetism in the ladder doped with one or two holes at @xmath2 and discuss the relationship with the occurrence of ferromagnetism in two dimensions . next in sec . [ sec : pair ] we discuss the pairing of holes doped into the ladder and the occurrence of phase separation . to understand the excitation spectra we start from the single hole case in sec . [ sec : onehole ] and go on to the two - hole case in sec . [ sec : twohole ] . section [ sec : paircorr ] discusses long range correlations , in particular the interesting question of the symmetry of the pairs and the mapping to a luther - emery liquid . the single - particle excitations are discussed in sec . [ sec : chargexc ] . over all we find a remarkable similarity between the ladder and 2d clusters . the @xmath0-model ( @xmath0-@xmath1 model with @xmath2 ) is equivalent to the infinite-@xmath37 hubbard model . in single chains the ground state of the @xmath0-model is degenerate in the spin degrees of freedom . in two dimensions on the other hand the ground state of the @xmath0-model doped with one hole is ferromagnetic.@xcite this is called the nagaoka effect . the extension of the proof by nagaoka to finite hole doping in the thermodynamic limit proved to be difficult . actually the ground state of the two - dimensional ( 2d ) square - lattice @xmath0-model doped with _ two _ holes is _ not _ ferromagnetic.@xcite for finite densities in the thermodynamic limit there are contradicting results . variational estimates for the @xmath38 hubbard model indicate that the fully polarized ferromagnetic state is stable until a critical doping @xmath39.@xcite high temperature series expansions by putikka _ _ on the other hand show evidence that the fully polarized ferromagnetic ground state does not survive at any finite doping . instead they find evidence for a partially polarized ferromagnetic state at low hole doping . a fully polarized ferromagnetic state at finite doping was found only for @xmath40.@xcite in this context it is of interest to study the occurrence of ferromagnetism in the ladder models . while the proof by nagaoka @xcite can not be applied to the one - dimensional chain it is valid for the ladder . the proof relies on the existence of closed loops on the lattice . such loops exist in 2d lattices and on ladders , but can not be formed on single chains . the ground state of the ladder doped with one hole is thus ferromagnetic . we have numerically studied the @xmath0-ladder with @xmath41 rungs , doped with two holes . in fig . [ fig : fm_engy ] we show the ground state energies of the ladders for both pbc s and apbc s . we find that the ground state is always ferromagnetic for apbc and an even number of rungs and for pbc and an odd number of rungs . for the other boundary conditions the ground state is a spin singlet . an important point is that the ferromagnetic state always has the lower energy for a ladder with at least four rungs . the singlet state is very close in energy and deserves a more detailed investigation . in fig . [ fig : fm_corr ] we plot the real - space spin correlations @xmath42 of the lowest singlet state of the @xmath43 ladder . these spin correlations show that the singlet state actually consists of two ferromagnetic domains with opposite magnetization . the results show clear evidence for a ferromagnetic ground state of the @xmath0-ladder ( @xmath44 ) doped with two holes . in the thermodynamic limit however two holes is not a finite density . extrapolations of our small - cluster results at finite doping to the thermodynamic limit @xmath45 are hard to obtain . but one may speculate that the existence of a ferromagnetic ground state of the @xmath0-ladder with 2 holes and @xmath46 could indicate a ferromagnetic state for dopings @xmath47 . similar results were obtained by hirsch and mller - hartmann.@xcite in this section we will discuss the pairing of holes doped into the @xmath0-@xmath1 ladder and the occurrence of phase separation at large values of @xmath3 . we will start from the simple limit @xmath48 . in this limit the undoped ladder consists of weakly coupled rungs , as is sketched in fig . [ fig : strong_limit](a ) . in this limit two holes doped into the ladder will go onto the same rung in order to minimize the number of broken singlet bonds . this state is graphically shown in fig . [ fig : strong_limit](c ) . in order to study the occurrence of hole pairing at smaller values of @xmath49 and down to the isotropic point @xmath36 we calculate the binding energy and the hole - hole correlation function . we find that even at isotropic coupling the holes still form a bound pair , although the pair is more spread out there . the binding energy @xmath50 is defined as @xmath51 where @xmath52 is the ground state energy for @xmath53 electrons , the boundary conditions are chosen between pbc and apbc to give the lowest energy . in the large @xmath49 region the binding energy can easily be estimated . a single hole doped into a heisenberg ladder breaks one bond with energy loss @xmath49 , but can gain kinetic energy @xmath54 along the ladder ( see the next section for details ) and @xmath55 along the rung . it follows that @xmath56 . two holes on the same rung also break one bond , but the kinetic energy of such a bound pair is much smaller , of order @xmath57 , as will be calculated later . thus we estimate @xmath58 , and a binding energy : @xmath59 figure [ fig : gaps ] shows @xmath50 as a function of @xmath49 . it remains positive and thus shows binding down to the isotropic value , @xmath60 . the same holds for a larger @xmath61 . additional evidence for pairing is provided by the hole - hole correlation functions @xmath62 measured on the same leg @xmath63 and on different legs @xmath64 in the ground state . they are plotted in fig . [ fig : hole_corr ] for @xmath65 and @xmath34 and @xmath66 . for @xmath67 the two holes are predominantly on the same rung and the correlation function shows a clear exponential decay . at the isotropic point the pair is more extended . the maximum of the correlation function is now at a distance 1 along the legs and on different legs , but it again decays at large distances . we can calculate the size of the hole pair by fitting the inter - chain correlations to an exponential form @xmath68 for the two largest distances , @xmath69 and @xmath70 . the inset of fig . [ fig : hole_corr ] shows the size @xmath71 of the hole pair as a function of the inter - chain coupling @xmath72 . the pair is very tightly bound for @xmath73 . at the isotropic point the pair is still bound , with a diameter of about two lattice spacings . note the oscillation of the radius with respect to @xmath12 . the size seems to converge to a value in between the @xmath74 and the @xmath43 result at the isotropic point . we may say that the system belongs to the luther - emery universality class of 1d correlated systems,@xcite in the sense that the spin excitations acquire a finite gap while the charge excitations remain gapless . in the limit of large @xmath49 , however , the picture that tightly bound hole pairs are moving in a background of singlet rungs is more appropriate than weak coupling approaches like @xmath75-ology . considering these hole pairs as hard core bosons , we can determine the long - range correlations by a mapping to an effective boson model . the pair hopping matrix element to second order in perturbation theory is @xmath76 there is a weak attraction @xmath77 between two hole pairs on neighboring rungs , which again to second order takes the form @xmath78 where the first , attractive , term comes from the charge part of the @xmath1-term in the hamiltonian . as @xmath79 we can map the low - energy part of the @xmath0-@xmath1 ladder onto an effective hard - core boson model on a chain with nearest neighbor interaction : @xmath80 where the hard - core boson creation operator @xmath81 creates a hole pair at the rung @xmath82 and @xmath83 is its number operator . there is a hard - core repulsion since only one hole pair can be created on any given rung . our effective boson model is equivalent to the xxz - model in a magnetic field , which has been solved exactly by a bosonization approach and conformal field theory.@xcite for @xmath84 the system is phase separated . this is the case for @xmath85 , where @xmath86 again to second order perturbation theory . for physically reasonable values of @xmath3 phase separation occurs only at very large values of @xmath49 : @xmath87 for @xmath65 and @xmath88 for @xmath61 . note that the dominant attractive part of the interaction comes from the charge part @xmath89 of the @xmath1-term . next we will discuss the region where the system is not yet phase separated but @xmath49 is still large ( @xmath90 ) . there we can determine the dominant correlations from the effective boson model . the correlation exponents have been calculated indirectly by bethe ansatz.@xcite both the charge density wave correlations and the superconducting correlations show a power - law decay at large distances : @xmath91 the superconducting correlations @xmath92 are dominant if @xmath93 . this is the case for most of the phase diagram , except for the phase separation regime at @xmath94 . at quarter filling @xmath95 and for @xmath96 the system is in the ising - limit and shows a long range charge density wave ground state . at fillings close to that line and for large @xmath96 there is a small region where @xmath97.@xcite in our effective model we have @xmath98 and there @xmath99.@xcite we are thus always in the region of dominant superconducting correlations . even neglecting the attractive charge part of the @xmath1-term we are still in the superconducting regime where @xmath93 . in the limit of large @xmath49 the equivalence of the @xmath0-@xmath1 ladder with a luther - emery liquid can clearly be seen . going to isotropic coupling the spin gap remains finite and the only low - lying excitation is the collective charge mode , as we will show in the following sections . thus also at isotropic coupling the @xmath0-@xmath1 ladder is still a luther - emery liquid . in sec . [ sec : map ] , we will develop another approach which relates the long - range correlations to thermodynamic quantities for more general @xmath1 s , based on a bosonization of density fluctuations . finally we study the occurrence of phase separation at isotropic coupling @xmath36 . we estimate the onset of phase separation by determining the coupling @xmath1 at which the compressibility @xmath100 diverges . the compressibility per site can be calculated as usual @xmath101 where @xmath102 is the energy density per site of the ladder with a particle density per site @xmath103 . in a finite system usually the discrete version @xmath104 \label{eq : kappaf}\ ] ] is used , where @xmath105 is the ground state energy of the finite system with @xmath53 particles on the ladder with @xmath12 rungs ( volume @xmath106 ) . at small hole doping however this procedure may not be reliable due to finite size effects caused by frustration on small lattices . to see this let us consider the @xmath74 ladder doped with zero , two or four holes . in the undoped case there are 8 spins on each leg of the ladder . two holes doped into the ladder will predominantly go onto different legs and there will be seven spins per leg . thus the antiferromagnetic configuration on the legs will be frustrated . for four holes there will be six holes on each leg and the system is again not frustrated . conversely on an @xmath107 ladder the undoped ladder and the ladder doped with four holes will be frustrated , while the ladder doped with two holes will be non - frustrated . we have thus used a different formula to calculate the compressibility at small hole doping . we calculate the ground state energies for an @xmath74 ladder doped with @xmath108 and @xmath109 holes and for an @xmath107 ladder doped with @xmath110 and @xmath111 holes . in all these cases the ladder is not frustrated . then we estimate the compressibility from these energies using finite differences similar to the above eq . ( [ eq : kappaf ] ) . in the thermodynamic limit @xmath45 both formulas give the same result , as the frustration appears only on small lattices . while the finite size effects are quite small at low electron densities they are much larger at small hole dopings due to frustration mentioned before . the estimated errors on the phase separation line may thus be much larger there , about @xmath112 . a comparison with the results obtained with open boundary conditions ( obc ) confirms our results . only at small doping the obc results are not reliable since there the holes are trapped on the ends of the chain . figure [ fig : phasesep ] shows the phase separation line for the @xmath0-@xmath1 ladder for @xmath36 , in the @xmath1-@xmath113 plane . note that , opposite to the single chain case,@xcite the onset of phase separation at small hole doping is at lower values of @xmath3 than at small electron concentrations . this resembles the behavior in two dimensions,@xcite although the precise position of the phase separation line in two dimensions has not yet been established . in the previous section we have discussed the ground state of the ladder doped with two holes . in order to understand the low energy excitations of the ladder it is useful to study the one - hole problem first . as mentioned above the limit @xmath48 is a good starting point to explore the @xmath0-@xmath1 ladder . there are nine different states , depicted in fig . [ fig : nine_basis ] . a single electron goes either into the bonding or antibonding orbital @xmath114 with energy @xmath115 , respectively . two electrons on the rung are either in the singlet state with energy @xmath116 or in one of the three triplet states with energy @xmath23 . the singlet state expressed in bonding and antibonding orbitals is @xmath117 similarly the three triplets can be expressed as combinations of one bonding and one antibonding electron : @xmath118 figures [ fig : disp_1h](a ) and ( b ) show the one - hole spectra for @xmath74 for large inter - chain coupling @xmath119 , calculated by exact diagonalization for @xmath65 , @xmath120 and @xmath61 , @xmath121 respectively . a hole on a single rung can be either in the bonding or the antibonding orbital . one hole doped into the half filled ladder will thus be either in a bonding or antibonding state , depending on the parity symmetry of the total ladder [ see fig . [ fig : strong_limit](b ) ] . this hole can propagate along the ladder with a hopping matrix element @xmath122 in first order perturbation theory . thus the low energy states are two bands of holes in the bonding and anti - bonding orbitals . they are split by the energy difference @xmath123 of the bonding and antibonding states . these two bands can clearly be seen in the spectra [ figs . [ fig : disp_1h](a ) and ( b ) ] . the minimum of the bands is at @xmath30 , since the hopping matrix element for holes @xmath124 . the bandwidth of both bands is @xmath125 in the limit @xmath48 . at finite @xmath1 the bandwidth is reduced due to hybridization with the higher excited states . decreasing @xmath49 to the isotropic points @xmath36 changes the dispersion of these bands [ see figs . [ fig : disp_1h](c ) and ( d ) ] . at low energies we can still see the bands of holes in the bonding and antibonding orbitals . these bands evolve continuously from the large @xmath49 limit . the minima of the energy bands are not at @xmath126 or @xmath30 , but at a large momentum @xmath127 for the bonding and @xmath128 for the antibonding band . we can fit the low - lying hole bands to a dispersion of the form @xmath129 corresponding to nearest neighbor ( @xmath130 ) , next - nearest neighbor ( @xmath131 ) and third - nearest neighbor ( @xmath132 ) hopping . @xmath133 is the ground state energy of the undoped ladder and @xmath134 the shift in energy of the center of the band upon doping . in fig . [ fig : disp_fit ] we show the bands and the excellent fit . the parameters are shown in table [ tab : fitspec ] . the changes in the hole dispersion with decreasing @xmath49 are summarized as follows : \(i ) the center of the bands shifts downwards by @xmath135 , compared to the undoped ladder . the energy gain for one hole in the case of @xmath2 would be just the kinetic energy @xmath55 . when @xmath136 we lose magnetic energy by introducing the hole . the energy gain is therefore smaller at larger @xmath3 , as we can also see from the fit parameters . \(ii ) the hole bands are narrowed compared to the large @xmath49 limit . in that limit the bandwidth of the hole bands was @xmath137 . this bandwidth is renormalized by the stronger polarization effects at isotropic coupling , and it is now of the same order as the magnetic energy @xmath1 , instead of the kinetic energy @xmath137 . \(iii ) the dispersion changes as longer range hopping processes ( @xmath138 are introduced with decreasing @xmath49 , and the minima move away from @xmath139 . the minima of both bands are very close in energy , again in contrast to the strong coupling region where they are split by @xmath123 . in sec . [ sec : chargexc ] we will identify the minima @xmath140 and @xmath141 with the fermi points of the bonding and antibonding quasiparticle bands . another interesting question is the behavior of the free spin that is left over after one hole has been doped into the ladder . in the @xmath0-@xmath1 chain the spin and charge excitations are carried by different soliton excitations which are far apart in space from each other . this is a typical feature of spin - charge separation and such a system is called a luttinger liquid . in a fermi liquid on the other hand they are bound and the excitations are described by quasiparticles carrying both charge and spin . we have calculated the hole - spin correlations to answer the question if spin - charge separation occurs in the ladder . the real space correlations @xmath142 are shown in fig . [ fig : hole - spin ] . this correlation function is nonzero for the ground state in the subspace of @xmath143 since there remains one spin unpaired . the result shows that the hole is tightly bound to the remaining free spin . at strong inter - chain coupling @xmath48 it is again predominantly on the same rung . at isotropic coupling the spin - hole bound state is more extended . these spin - hole bound states thus carry both charge and spin . in this sense they are similar to the quasiparticles in a fermi liquid . this is in contrast to the spin - charge separation in the single chain . we will therefore call the single holes bound to the free spin `` quasi - particles '' , although the system has a spin gap . the ground state of the ladder doped with two holes is , as discussed above , a bound state of the two holes . this bound pair coherently propagates along the ladder , giving rise to the lowest - lying band . when @xmath73 , this band , spin - singlet charge excitations , is clearly seen in the numerical results as shown in fig . [ fig : disp_2h_strong ] . the higher energy excitations are again understood simply in the large @xmath49 limit . an essential difference from the lowest - lying singlet band is that two holes are now separate rather than forming a bound pair . being separate , they can gain a larger kinetic energy , but only in return for a even larger cost of exchange energy @xmath144 as one more singlet rung is broken . thus there are continua of scattering states of the two holes ( `` quasiparticles '' ) at higher energies . since the residual interactions between the two quasiparticles are weak , the energy is almost degenerate between the @xmath145 and @xmath146 spin subspaces . on the finite lattice we naturally do not see a continuum of scattering states , but only several discrete bands . these bands , and the fact that the energies of the triplet and singlet are nearly degenerate ( up to boundary effects ) can be seen in the spectra . there are various combinations of the two quasiparticle bands in the two quasiparticle continuum of states . the lowest are scattering states of two bonding quasiparticles , with @xmath147 . higher states are scattering states of one bonding and one anti bonding quasiparticle . having the same transverse momentum , @xmath31 these states hybridize with the one - magnon excitations in the spin background , resulting in a `` bound state '' below the two - particle continuum . this is nearly dispersionless and clearly seen in the spectrum . the continua of two antibonding quasiparticles is much higher in energy and not included in the figure . at higher energies there are spin excitations in the spin background . they are not described by the quasiparticles and will be discussed in more detail in the next section . the results in the isotropic case , @xmath34 , is shown in fig . [ fig : disp_2h_iso ] . the energy spectrum is more complicated but the above description still holds qualitatively . the ground state is still the bound hole pair . it moves coherently along the ladder , yielding a gapless band of singlet charge excitations . the band has a linear dispersion around @xmath32 compared with quadratic in the large @xmath49 case . an important point is that despite its complicated dispersion the low energy part is well separate from the other excitations . therefore also at isotropic coupling the only low - energy excitation is the collective charge excitations , and we may identify the isotropic @xmath0-@xmath1 ladder as luther - emery liquid . we will discuss the essential role of these charge fluctuations concerning superconductivity in sec . [ sec : map ] . in addition to the gapless band of charge fluctuations , there are various local minima at higher energies . however , they can be explained by taking account of the nonmonotonic dispersion of the one - hole spectra shown in fig . [ fig : disp_1h ] , and our quasiparticle picture still holds . more specifically , there are four local minima in the single - hole spectra at @xmath148 and @xmath149 , which have nearly the same energy . thus to construct low - energy two - quasiparticle excitations there are many possible combinations of different minima of one - particle states as discussed below . this is the origin of many local minima in the two - particle spectra and this is confirmed by the fact that the dependence on the boundary condition is consistent with this picture . the ground state of the two - hole spectrum can be constructed from holes near @xmath150 in the single - hole spectra . the other minima can be explained similarly . the minimum in the pbc spectrum at @xmath151 can be identified with the @xmath152 excitation , where a particle moves from one fermi point to the opposite one . ( @xmath153 ) . when using apbc we do not have the @xmath154-value of @xmath155 in the single - hole spectrum on an @xmath43 ladder . the closest @xmath21-points are @xmath156 and @xmath157 , leading to minima at @xmath158 and @xmath159 in the two - hole spectra . another feature that can be explained from the single hole spectra is the minimum at @xmath160 ( odd parity , @xmath30 ) . using pbc this state is obtained with one hole with @xmath161 and one with @xmath162 , leading to the minimum at @xmath163 . using apbc s we can combine one hole at @xmath164 with one at either @xmath165 or @xmath166 . as mentioned above these two states are higher in energy than the minimum at @xmath159 and very similar in energy . therefore we expect two states at @xmath167 and @xmath160 which are similar in energy but at a higher energy than the corresponding states with pbc . this is exactly what we observe . the odd parity states near @xmath168 can be constructed similarly . the qualitative features of the three - hole excitation spectrum can again be explained similarly . one of the most interesting properties of the @xmath0-@xmath1 ladder is that there are two distinct types of spin excitation . although it is most easily seen in the large @xmath49 limit , the qualitative distinction remains down to the isotropic point . the first type is the collective magnon excitations inherited from the undoped spin ladder . one of the electron - filled rungs is now excited to a spin triplet . this local excitation is what we call `` magnon '' and it propagates coherently along the ladder , leading to an energy dispersion with respect to @xmath21 . for a detailed investigation , we have examined the two - hole spectra in more detail and have calculated the spin - spin and spin - hole correlations of the low - lying triplet states . we find that the magnon excitations of the heisenberg ladder evolve continuously upon doping . however there discontinuously appears a new kind of spin - triplet excitation at lower energies , which is not present in the undoped ladder . the lowest excitation is a different type for which quasiparticles play an essential role . therefore , the spin gap , defined as the excitation energy to the lowest triplet , is a discontinuous function of the hole doping at @xmath169 . this new type of excitations consists of breaking a pair of holes into two separate quasiparticles , each carrying charge @xmath5 and spin @xmath170.@xcite when the two quasiparticles are both in the bonding orbital , their lowest energy , at @xmath32 , is lower than the lowest magnon excitation . the additional energy gain is easily understood , since the two separate holes have a larger kinetic energy of the order of @xmath0 while the gain of the magnon kinetic energy is the order of @xmath171 . as was shown in fig . [ fig : gaps ] , down to the isotropic point the lowest two - quasiparticle excitation is lower in energy than the lowest magnon excitation . the above picture is confirmed by comparing the correlation function between the two different states . the dynamical spin structure factor gives another confirmation . figure [ fig : corrs ] shows the equal - time correlations of the two holes , @xmath172 , and of spin and hole , @xmath173 , calculated for these two types of spin - triplet excitations . the latter quantity is nonzero since the states with @xmath174 are used in the calculation . when @xmath119 , the two holes are separate in space in the lowest state , while tightly bound in the other state . the position of magnetic excitation is , on the other hand , close to the hole position in the lowest state , while they are far apart from each other in the other state . these two behaviors are what is predicted by our picture explained above , and despite modification in small detail they are qualitatively consistent even at the isotropic point . our numerical results confirm that the lowest triplet state is the quasiparticle excitation where the bound hole pair breaks up . the two holes repel each other and the hole - hole correlations have the maximum at the largest distance @xmath69 . each of the holes is bound to a spin-@xmath170 , as can be seen from the hole - spin correlation function . this state is sketched in fig . [ fig : strong_limit](d ) . as a typical magnon excitation , we show the correlations for the state at @xmath160 that has the main spectral weight in the dynamical spin structure factor which will be discussed soon . the hole - hole correlations are similar to the ground state . the hole - spin correlations show that the triplet carrying spin current is far away from the hole pair . this state is shown in fig . [ fig : strong_limit](e ) . mean field theory@xcite predicts only this magnon excitation , which evolves continuously from the heisenberg ladder . in neutron scattering experiments the relevant quantity is the dynamical spin structure factor @xmath175 where @xmath176 is the complete set of eigenstates with energy @xmath177 , @xmath178 is the ground state with energy @xmath179 and @xmath180 figure [ fig : spin_0h ] shows @xmath181 calculated for the heisenberg ladder using the lanczos diagonalization combined with the continued fraction method.@xcite it can clearly be seen that the dominant contributions arise from the collective excitations near @xmath182 and there is very little weight in the continuum of spin excitations at higher energy . in the doped case the two types of spin excitations have different contributions to the dynamical spin structure factor . it can be seen in fig . [ fig : spin_2h ] that the continuum of spin excitations move towards lower energies . most of the weight is in the magnon excitations , consistent with the mean - field theory,@xcite and there is very little weight in the lowest triplet excitation consisting of the two separate quasiparticles . as in the large @xmath49 limit the lowest triplet excitation with @xmath183 is a bound state of a spin triplet and the hole pair . at @xmath182 this state has no spectral weight , while most of the weight is in the second excited state , which has the triplet separated from the bound pair . this is a finite size effect of the two - hole system . the reason is that at @xmath184 or @xmath24 we have an additional symmetry , reflection invariance in the ladder direction . the parity under these reflections is different for the ground state and the triplet - hole pair ground state , leading to the vanishing weight . a significant difference between the two types of spin excitations is that the largest number of quasiparticle " excitations is limited by the number of holes . we need at least two holes to create such an excitation . the number of possible excitations is thus proportional to the hole doping @xmath185 . on the other hand , the magnon excitations can be excited at any rung where there are no holes . the number of these excitations is thus proportional to @xmath186 instead , much larger for a small doping @xmath185 . therefore with decreasing temperature the susceptibility will show a large exponential drop at temperatures of the order of the gap of the undoped system ( @xmath187 ) , followed by a small drop at temperatures around the spin gap of the doped system . to summarize we can describe the spin excitations of the @xmath0-@xmath1 ladders by a simple picture : quasiparticles moving in a spin liquid background . in the ground state the quasiparticles are paired . in the excitation spectrum two types of excitations can be distinguished . the first corresponds to the breaking of a pair of quasiparticles . this excitation has the lowest energy , but its number is limited by the number of holes . of more importance for measurements of the susceptibility or inelastic neutron scattering experiments is the second type , which are magnon excitations in the spin liquid background . they evolve continuously from the undoped heisenberg ladder . although the gap for this type of excitation is larger it is more important since we can excite more of these excitations . also the weight in the dynamical spin structure factor is larger . similarly to the calculation of the dynamical spin structure factor we calculate the dynamical charge structure factor defined by @xmath188 where @xmath189 is the fourier transform of the local density fluctuation around the average density @xmath113 . the result of our calculations is shown in fig . [ fig : charge_2h ] . the main contribution arises from the coherent motion of the hole pairs . this leads to large peaks at low energies near @xmath190 . the rest of the weight is distributed incoherently over a wide region at rather high energies of order @xmath191 . it arises from interactions of a single hole with the surrounding spin background . this is similar to results obtained for two dimensions.@xcite recently tohyama et al . have calculated @xmath181 and @xmath192 for single chains and 2d clusters.@xcite in the single chain they find that , as expected from luttinger liquid theory , the charge and spin excitations are decoupled in the low energy region . the dynamical charge structure factor is very similar to that of spinless fermions , consisting of large peaks at the energies expected from the cosine band of the spinless fermions . in two dimensions they find different behavior . in the spin structure factor nearly all of the weight is in a few sharp peaks at low energies . in the charge structure on the other hand the main weight is at relatively large energies , of the order of several @xmath0 , and it strongly broadened . this is indicative of strong spin - charge interactions . only at some special @xmath193-points are there peaks at low energies . the dynamical charge structure factor , @xmath192 for the ladder shown in fig . [ fig : charge_2h ] differs a lot from the single chain but resembles the 2d cluster . at large @xmath194 , we see the peaks at large energies @xmath195 . at small @xmath194 in the @xmath184 sector , @xmath196 for the ladder is dominated by the collective mode of the hole pairs , and its behavior differs substantially from spinless fermions . another similarity to the 2d system is found in the dynamical spin structure @xmath181 where the main weight is in peaks with energies @xmath197 ( see fig . [ fig : spin_2h ] ) . we conclude from this comparison that the 2d clusters and the ladders are closely related - a fact which points towards a `` d - wave '' paired state for the 2d clusters . a more difficult question is to what extent this behavior of the finite 2d cluster is a consequence of the strong tendency of finite clusters to favor singlet ground states and to what extent it is representative of an infinite plane . a highly intriguing and much debated subject of the high-@xmath198 superconductors is the internal symmetry of the order parameter . in this section we study the internal structure of the pairs in the doped ladders . for the @xmath0-@xmath1 model usually only nearest - neighbor pairs have been considered except for a few cases . it is reasonable to assume that they are the dominant pair correlations . however more quantitatively we have chosen the optimal form for the pairs . using the lanczos algorithm we calculate the pairing correlation functions for different pairings . let us introduce the operator creating a singlet pair of electrons on sites @xmath199 and @xmath200 , @xmath201 using this definition of the pair operator we can calculate the superconducting order parameter @xmath202 and its fourier transform @xmath203 where @xmath204 is the fourier transform of the real space pair operator with zero total momentum . figure [ fig : pair_k ] shows the superconducting order parameter for the @xmath74 ladder with pbc s for several values of @xmath3 . the most obvious properties are that the sign is opposite for pairs with @xmath147 and @xmath31 and that the absolute values are very small near @xmath205 and @xmath163 . this is similar to @xmath4-wave pairing in a fully 2d system . the absolute value is largest near wave vectors which we have identified with the fermi points of the quasiparticles @xmath206 and @xmath207 . at large @xmath3 the order parameter is very similar to the simple @xmath208 structure of nearest neighbor @xmath4-wave pairs . consider now pairs with a form factor @xmath209 : @xmath210 where @xmath209 is normalized @xmath211 a simple calculation shows that the superconducting `` order parameter '' @xmath212 is maximized by the pair with the form factor @xmath213 ^{1/2 } } , \label{eq : formfactor}\ ] ] which is proportional to the order parameter @xmath214 . this optimal order parameter then is @xmath215 ^{1/2}. \label{eq : optimal}\ ] ] in fig . [ fig : op ] we show this maximal order parameter for a @xmath74 ladder as a function of @xmath3 . a clear increase as a function of @xmath3 can be seen , similar to single chains and 2d planes where the superconducting correlations are also enhanced at larger @xmath1 values . we note that the expectation value of the optimal order parameter is larger for four doped holes than for two holes . this is a consequence of the strong correlation effect , as any pairing order parameter or fluctuations should vanish as @xmath216 . the mean - field theory including the strong correlation effects predicts a @xmath185-linear dependence of the order parameter as @xmath216:@xcite i.e. , the superfluid density is proportional to the hole doping . our results show that the value for four doped holes is less than twice the value for two holes . this may be because that in the four hole case the doping , @xmath217 , is already large and out of the region of @xmath185-linear dependence . we need further study using larger lattices to clarify this point more quantitatively . in secs . [ sec : pair]-[sec : twohole ] , our numerical results have shown that all the spin excitations cost a finite energy and the only gapless excitations are charge fluctuations , i.e. , coherent propagation of bound hole pairs along the ladder direction . we may therefore say that the system is a luther - emery liquid@xcite from large @xmath49 limit down to the isotropic coupling . this observation indicates that the low - energy and long - wavelength properties of lightly doped @xmath0-@xmath1 ladders would be entirely described by an effective continuum hamiltonian in terms of charge degrees of freedom and this model correctly predicts long - range asymptotic behavior of correlation functions . the effective model is actually the bosonic gaussian model proposed by efetov and larkin,@xcite @xmath218 ^ 2 \over \pi k_{\rm b } } + \pi k_{\rm b } v_s^2 \bigl [ { \nabla \theta_b ( x ) \over \pi } \bigr ] ^2 \right\},\end{aligned}\ ] ] where @xmath219 represents the density of bound hole pairs at the rung @xmath220 with the average value @xmath221 , and @xmath222 is its conjugate phase obeying the canonical commutation relation , @xmath223 = i \delta ( x - x')$ ] . as shown by the fact that the @xmath224 mode of the first term in eq . ( [ efetov_larkin ] ) denotes the change in the ground state energy associated with the number of bound hole pairs , @xmath225 , the parameter @xmath226 is given by the compressibility of the hole pairs , which will be shown afterwards . the second term , on the other hand , describes the energy change associated with current , related with the sound velocity @xmath227 . instead of a direct calculation of the dispersion relation , the sound velocity can also be obtained numerically by applying an aharanov - bohm flux penetrating in the center of the ladder with pbc s , since the flux induces a finite current along the chain direction . once the two parameters , @xmath227 and @xmath226 , are determined in this way , it is straightforward to calculate the power - law exponents of correlation functions as follows . the efetov - larkin hamiltonian is actually identical to the single - component ( i.e. , spinless ) luttinger model diagonalized by bosonization , which was studied by mattis and lieb,@xcite luther and peschel,@xcite and haldane@xcite in details . the hamiltonian ( [ efetov_larkin ] ) is immediately solved by rewriting each fourier component with a boson operator , @xmath228 then the result is @xmath229 .\end{aligned}\ ] ] here @xmath230 is the total number of bound hole pairs and @xmath231 is the charge velocity and @xmath232 is the current velocity associated with the number of @xmath233 excitations , @xmath234 , obeying the universal relation , @xmath235 . here @xmath236 is the fermi wave number of spinless fermions transmuted from the bosonic hole pair operators by the jordan - wigner transformation . the last two terms in eq . ( [ ham_diag ] ) describe non - bosonic excitations accompanied with the change in the quantum numbers , @xmath237 and @xmath234 . their importance was first pointed out by haldane in his luttinger liquid concept.@xcite the propagator of bound hole pair , i.e. pairing correlation , is then obtained as in the calculation of the debye - waller factor , @xmath238 } \rangle = \overline{n}_b \ , e^ { - \langle [ \theta_b ( r ) - \theta_b ( 0 ) ] ^2 \rangle /2 } \nonumber\\ & & \propto r^{-1/k_{\rho}},\end{aligned}\ ] ] where the exponent is @xmath239 where @xmath113 is the electron density per site and @xmath100 is the electron compressibility . the relation for the ladder , @xmath240 , is used to obtain the second equality . this result differs from the one for the chain , which has the numerical factor @xmath241 instead . it is important to emphasize that this result is the luttinger liquid parameter , @xmath242 , for the single - component boson systems . density - density correlations are characterized by a power - law decay at the wave number @xmath233 . since @xmath243 is twice the average of the fermi wave number of the original electron bonding and antibonding bands , @xmath244 , the wave number @xmath233 is actually @xmath245 in the original picture . the exponent of the @xmath233=@xmath245 cdw fluctuations is calculated using a density - phase duality as was done in the original paper.@xcite the density operator has a short - range @xmath233 cdw order , but since the value of @xmath233 is locally determined by the density at each position , it is fluctuating around the average value @xmath246 associated with the density fluctuations . we therefore may write as [ density ] @xmath247 , \\ { \partial \theta_j \over \partial x } & & = \pi ( n_b - \overline{n}_b ) .\end{aligned}\ ] ] rewriting the hamiltonian ( [ efetov_larkin ] ) with @xmath248 and its conjugate operator @xmath249 , we again obtain a gaussian model , @xmath250 ^ 2 + v_n \bigl [ { \nabla \theta_j ( x ) \over \pi } \bigr ] ^2 \right\},\end{aligned}\ ] ] where @xmath251= i \delta ( x - x')$ ] . this duality of the charge and current operators was also emphasized in the haldane s paper.@xcite since the density - density correlation is written as an exponential of @xmath248 in this representation , its calculation can be done similarly as before , @xmath252 } \rangle \nonumber \\ & & \propto \cos ( 2k_f^{\rm sf } r ) \ , r^{-k_{\rho}}. \label{density_corr}\end{aligned}\ ] ] instead of calculating the sound velocity , the value of @xmath253 is more accurately obtained through the drude weight defined by @xmath254 where @xmath255 is the total energy of the ground state when the flux @xmath256 is penetrating . assuming the holes propagate in pair , the relation between @xmath257 and @xmath227 for the ladder is obtained and the result agrees with the one for the chain@xcite aside from its numerical factor , @xmath258 combined this with eq . ( [ exponent ] ) , we finally obtain the formula of the correlation exponent in terms of the compressibility and the drude weight : @xmath259 it is noted that this expression is identical to the one for the chain in terms of @xmath100 and @xmath257 , while they differ by factor 2 in terms of @xmath100 and @xmath227 . in this way , when the energy scale concerned is smaller than the spin gap , we can predict long range asymptotic behavior of the correlation functions based on the efetov - larkin effective model , and calculate correlation exponents once the two parameters , @xmath227 and @xmath226 ( or equivalently @xmath100 and @xmath257 ) , are numerically determined . figure [ fig : k_rho ] shows the correlation exponent calculated in this way for the isotropic ladder ( @xmath34 ) with @xmath260 and two holes as a function of @xmath1 . this corresponds to the electron density , @xmath261 . the necessary quantities are carefully calculated by using the lanczos diagonalization . the compressibility , @xmath100 , is determined from the ground state energy of unfrustrated systems in the sense explained before . the drude weight , @xmath257 , is then calculated by eq . ( [ eq : drude ] ) . combining these two , @xmath253 is obtained via eq . ( [ eq : exponent_2 ] ) . near the phase separation boundary , @xmath253 grows rapidly . this is owing to the divergence of the compressibility at this boundary . in other words , the collective charge excitations become softening and the superconducting fluctuations are enhanced correspondingly . recently hayward et al.@xcite directly calculated various correlation functions for the @xmath0-@xmath1 ladders by using the density - matrix renormalization - group method . at @xmath262 and @xmath263 , they found a power - law decay of the pairing correlations with the exponent close to unity . this exponent corresponds to @xmath264 , which is larger than our estimate at the same @xmath1 and @xmath49 , @xmath265.@xcite the discrepancy may be owing to the larger electron density in our calculation , @xmath266 , but it is not clear if this suffices to account for the difference until new calculation is carried out at the same density . hayward et al . also report data on the density - density correlation function@xcite but the exponent of the expected power - law decay of the oscillatory term can not be easily extracted . finally we discuss the one - particle green s function where we can see the quasiparticle excitations directly . the electron and hole parts of its spectral function are defined as @xmath267 where @xmath268 is an eigenstate for @xmath53 electrons with the energy @xmath269 and @xmath270 denotes the ground state with @xmath53 electrons . positive ( negative ) energies correspond to the electron ( hole ) part . the chemical potential is defined by @xmath271.\ ] ] the results for @xmath43 and the isotropic couplings , @xmath34 , are shown in figs . [ fig : gf ] for @xmath3=0.3 and 0.5 . the wave vectors along the ladder @xmath272 ( @xmath273 : integer ) are for pbc s and @xmath274 for apbc s . the ground state energy @xmath275 and the chemical potential @xmath276 in eq . ( [ eq : spe ] ) are the average over both boundary conditions . the results for @xmath43 are very similar to our @xmath74 results published before.@xcite the spectral function has large weights for the bonding ( _ b _ ) ( @xmath277 ) and antibonding ( _ a _ ) ( @xmath278 ) orbitals only near the fermi energy @xmath279 , and they seem to constitute quasiparticle bands . away from the fermi energy , the individual quasiparticle peaks are much less prominent and there is an incoherent part with an energy of the order of @xmath0 . the quasiparticle part of the spectrum is consistent with the mean - field theory based on the @xmath4-wave rvb state.@xcite the undoped ladder consists of local singlets on the rungs . such a singlet is the superposition of two electrons in the bonding orbital and two electrons in the antibonding orbital , @xmath280 . holes doped into the half - filled ladder will go predominantly into the anti - bonding orbitals to gain a larger kinetic energy along the rung direction . the bonding band is occupied by more electrons , while the antibonding band is occupied by less electrons . the quasiparticle with energy closest to @xmath281 has a wave vector nearest to the original fermi wave number , @xmath282 : ( @xmath283 for bonding and @xmath284 for antibonding ) . because of the band splitting , @xmath285 , but the luttinger sum rule is satisfied , @xmath286 . this means the fermi volume is large , proportional to the electron number rather than the hole number , and this is consistent with photo - emission experiments on cuprate superconductors.@xcite it is important to notice that the quasiparticle peaks near the fermi energy have their counterparts on the opposite side of the fermi energy . an electronic quasiparticle peak at energy @xmath287 has a shadow hole peak at energy around @xmath288 , and _ vice versa_. these peaks indicate that the quasiparticle excitations are those of the bogoliubov quasiparticles as in bcs theory , _ i.e. _ , mixture of an electron and a hole ( @xmath289 ) . the weights in the electron and hole parts are proportional to @xmath290 and @xmath291 . they are hole - like around @xmath168 and electron - like around @xmath292 for both the bonding and antibonding bands . there exists a finite energy gap in the quasiparticle spectra . the electron and hole branches both come close to the fermi energy at @xmath293 , but instead of passing through they move away from it . the energy gap for @xmath65 ( @xmath294 ) is @xmath295 ( @xmath296 ) at @xmath297 and @xmath298 ( @xmath299 ) at @xmath300 . this corresponds to a quasiparticle gap @xmath301 for @xmath65 and @xmath302 for @xmath61 . it is interesting to note that the calculations of @xmath303 in 2d clusters@xcite show similar behavior for @xmath304-points not along @xmath305 but no shadow peaks for @xmath306 , indicating @xmath307-pairing also . figure [ fig : dos ] shows the spectral function of the one - particle green s function of two holes in an @xmath43 ladder summed over all wave vectors : @xmath308 , and @xmath309 . this quantity is the local density of states to add and remove electrons . in a strongly correlated system , the sum rules on the weight ( i.e. , the integrated values ) of @xmath310 and @xmath311 are very different since the former is given by the number of empty sites and the latter by the number of filled sites ( or equivalently the number of holes and electrons , respectively ) . in a fermi liquid the values of @xmath310 and @xmath311 for small values of @xmath312 are determined by quasiparticle weight at the fermi energy and are continuous . it is interesting therefore to note that fig . [ fig : dos ] shows approximately similar values for @xmath313 and @xmath314 around @xmath315 , but the sum rule on the total weight is satisfied through the large weight in incoherent excitations in @xmath311 at higher energies , @xmath316 the strong correlation condition is reflected in the much smaller total weight in @xmath317 which comes about through an effective cut - off in energy on @xmath317 . in this respect the system in energy space is similar to a lightly hole doped band insulator although as we discussed earlier the location in @xmath304-space of the coherent quasiparticle peaks corresponds to a large fermi surface to add and remove electrons . the momentum distribution for electrons , @xmath318 , and for holes , @xmath319 , is shown in fig . [ fig : n_k ] . note because of the strong correlation condition , these do not add to one but instead their sum is given by @xmath320 . the strong correlation condition is also evident in the reduced magnitude of the variation of @xmath321 as a function of @xmath304 . nonetheless the presence of an apparent `` fermi surface '' in the center of the brillouin zone is clear , consistent with the dispersion relations of the coherent quasiparticles . the difference between the bonding ( @xmath147 ) and antibonding ( @xmath31 ) bands and the reduced occupation of the antibonding band arise from the energy gain in placing the doped holes preferentially in the antibonding band . the results of our lanczos diagonalizations confirm earlier studies which concluded that lightly doped two - leg ladders belong to a different universality class from single chains . the latter are tomonaga - luttinger liquids with gapless and separated spin and charge excitations . the ladder in contrast has a finite gap in the spin excitation spectrum and gapless excitations only in the charge sector . the low energy excitations evolve continuously from the limit of strong inter - chain exchange coupling ( @xmath322 ) and the simplicity of that limit allows a clear interpretation of our results . at large @xmath49 the dispersion relation of a single doped hole consists of two cosine bands corresponding to bonding and antibonding states on a rung . lowering @xmath49 to the isotropic limit ( @xmath34 ) and setting both @xmath323 changes the dispersion relation substantially . the coherent parts of both bands are centered at energies @xmath324 but the width is @xmath197 only . the spin and charge components are still bound but more loosely and the large magnetic polarizability of the spin background introduces longer - range hoppings . remarkably the form of the bands resembles the noninteracting band structure so that a photoemission experiment which removes electrons would measure in effect a large `` fermi surface '' with bonding and antibonding pieces . in this regime the quasiparticle propagation is strongly influenced by the coupling to magnetic excitations . when two holes are added they bind together on a single rung at large @xmath49 and remain bound although the size of the bound hole pair increases as @xmath49 approaches @xmath325 . moreover the qualitative features of the density - density @xmath192 and spin - spin structure factor @xmath181 , which are easy to understand at large @xmath49 , remain similar as @xmath49 approaches @xmath325 . @xmath192 near @xmath190 is dominated by the low energy mode associated with the motion of hole pairs . at large @xmath194 , @xmath192 has a broad peak at high energies ( @xmath326 ) similar to that found by ohta , eder and maekawa for 2d clusters.@xcite they interpreted this as local excitations of single holes in the magnetic cloud or spin bag.@xcite the dynamical spin structure factor , @xmath181 , also resembles 2d clusters and not 1d chains when we compare to the results of tohyama , horsch and maekawa.@xcite the major weight is at energies @xmath197 . the spin gap evolves discontinuously upon doping through the new quasiparticle excitations that can be made by breaking a hole pair into two separate single holes . however the major weight of the spin excitations remains in the collective magnon mode whose dispersion evolves continuously from the @xmath327 limit , although it is influenced by the continuum of quasiparticle excitations . we have also investigated the one - particle spectral functions to add and remove electrons from the two - hole ground state . these show clearly the unusual nature of this `` fermi liquid '' . when electrons are removed ( or holes added ) , the spectral weight is spread over a large energy region ( @xmath328 ) , but the coherent part is limited only to energies @xmath197 below the fermi energy @xmath276 . the energy dispersion relations show a large apparent fermi surface for the coherent quasiparticles and which matches onto a similar one for adding electrons at energies greater than @xmath276 . these @xmath304-space features resemble a metal with a large fermi surface . the property that resembles a lightly hole doped insulator is the energy dependence of the spectral weight to add an electron . this shows a low energy cut - off ( @xmath329 ) similar to a lightly hole doped band insulator . the result is an intriguing duality between metallic - like features in @xmath304-space and lightly hole doped insulating features in energy space . the overall properties of the lightly doped ladder place it in the luther - emery class rather than the tomonaga - luttinger class of 1d systems . the low energy properties of luther - emery liquids are described by interacting hard - core bosons as shown by efetov and larkin . in the present case the efetov - larkin bosons are bound hole pairs . two features distinguish the @xmath0-@xmath1 ladder from the usual luther - emery liquids arising from attractive interactions . one is the @xmath4-wave character of the pairing and the second is the presence of magnon excitations and limited quasiparticle excitations . note the magnon excitations can not be viewed as the collective mode of quasiparticles since the latter vanish as @xmath330 . the system is not a standard fermi liquid , but rather is a new and interesting mixture of a dilute attractive fermi gas in which the hole binding energy remains finite as @xmath330 , and a dense fermi liquid with an apparent large fermi surface in @xmath331-space . comparing the ladder with the results by tohyama et al . , we see that the ladder is very different from the single chain but similar to 2d clusters in many respects . both in ladders and in 2d clusters @xmath4-wave pairs are found down to small @xmath3 . the dynamical charge and spin structure factors look remarkably similar and the single - particle spectral functions indicate the existence of bogoliubov quasiparticles with a finite superconducting gap . thus we are lead to the conjecture that the @xmath0-@xmath1 model on 2d clusters is a doped rvb spin liquid showing @xmath4-wave pairing , similar to the ladder . we wish to thank m. sigrist , f.c . zhang , h. monien , r. noack , d. poilblanc , p. prelovsek , d.j . scalapino , s.r . white , and d. wrtz for helpful discussions . this work has been supported by the swiss national fund under grant number nfp-304030 - 032833 , by an internal grant of ethz and by the centro svizzero di calcolo scientifico cscs manno . the calculations have been performed on the cray y - mp/464 of eth zrich and on the nec sx-3/24r of cscs manno . d. v. khveshchenko and t. m. rice , phys . b * 50 * , 252 ( 1994 ) ; d. v. khveshchenko , _ ibid . _ * 50 * , 380 ( 1994 ) ; n. nagaosa and m. oshikawa , preprint cond - mat/9412003 ; h. j. schulz , preprint cond - mat/9412098 . w. van der linden and d.m . edwards , j. phys . condens . matter * 3 * , 4917 ( 1991 ) ; _ ibid . _ * 3 * , 7229 ( 1991 ) ; a.j . basile and v. elser , phys . b * 41 * , 4842 ( 1990 ) ; b.s . shastry , h.r . krishnamurthy , and p.w . anderson , phys . b * 41 * , 2375 ( 1990 ) .

design methods for paraunitary matrices from complete orthogonal sets of idempotents and related matrix structures are presented . these include techniques for designing non - separable multidimensional paraunitary matrices . properties of the structures are obtained and proofs given . paraunitary matrices play a central role in signal processing , in particular in the areas of filterbanks and wavelets .

1205.0703

a one - dimensional ( 1d ) paraunitary matrix over @xmath0 is a square matrix @xmath1 satisfying @xmath2 . here @xmath3 denotes complex conjugate transposed and @xmath4 denotes the identity matrix of the size of @xmath1 . in general a @xmath5-dimensional ( kd ) paraunitary matrix over @xmath0 is a matrix @xmath6 where @xmath7 is a vector of ( commuting ) variables @xmath8 such that @xmath9 and @xmath10 . over fields other than @xmath0 a paraunitary matrix is a matrix @xmath6 satisfying @xmath11 . paraunitary matrices are important in signal processing and in particular the concept of a paraunitary matrix plays a fundamental role in the research area of multirate filterbanks and wavelets . in the polyphase domain , the synthesis matrix of an orthogonal filter bank is a paraunitary matrix ; see for example @xcite . orthogonal filter banks may also be used to construct orthonormal wavelet bases @xcite ; see also references in @xcite . paraunitary matrices over finite fields have been studied for their own interest and for applications ; see for example @xcite . here general methods for constructing and designing such matrices from complete orthogonal sets of idempotents together with related matrix schemes are presented . this includes methods for designing non - separable multidimensional paraunitary matrices . construction methods for complete orthogonal sets of idempotents are included . group ring construction methods were the original motivation and from these more general methods evolved . a structure called the _ tangle _ of matrices is introduced ; this may have independent interest . in certain cases specialising the variables of the paraunitary matrices allows the construction of series of regular real or complex hadamard matrices . walsh - hadamard matrices , used extensively in the communications areas , are examples of such regular hadamard matrices . complex hadamard matrices arise in the study of operator algebras and in the theory of quantum computation . it is noted that the renowned building blocks for 1d paraunitary matrices over @xmath0 due to belevitch and vaidyanathan as described in @xcite are constructed from @xmath12 where @xmath13 is a complete orthogonal set of two idempotents in which @xmath14 has rank @xmath4 and @xmath15 has rank @xmath16 with @xmath17 the size of the matrices under consideration . see section [ tub ] below for details on this . connections between group rings , matrices and design of codes have been established in @xcite , @xcite and @xcite ; these are related but independent . designing non - separable multidimensional paraunitary matrices is deemed difficult as there is no multidimensional factorisation theorem corresponding to the 1d factorisation theorem of belevitch and vaidyanathan ( @xcite ) . for the finite impulse response ( fir ) case there seems to be only a few examples such as @xcite . see also @xcite for background and further discussion . in @xcite a factorization of a subclass of 2d paraunitary matrices is obtained ; these though involve iir ( infinite impulse response ) systems . in section [ grmat ] results are obtained on the ranks of the idempotents and on the determinants of the paraunitary matrices formed . the concept of a pseudo - paraunitary matrix is introduced in section [ pseudo ] and construction methods for these are given . these may also be considered as fir ( finite impulse response ) systems . the book @xcite is an excellent reference for background material on the algebraic structures used . now @xmath18 denotes a general field , @xmath19 denotes a general ring , @xmath0 denotes the complex numbers , @xmath20 denotes the real numbers and @xmath21 denotes the rational numbers . also @xmath22 denotes the finite field of @xmath23 elements , @xmath24 denotes the set of @xmath25 matrices with coefficients from @xmath19 and @xmath26 $ ] denote the polynomial ring with coefficients from @xmath19 in commuting variables @xmath27 . note that @xmath28_{n\ti m } = r_{n\ti m}[\bf z]$ ] . let @xmath19 be a ring with identity @xmath29 . ( in general @xmath4 will denote the identity of the system under consideration . ) a _ complete family of orthogonal idempotents _ is a set @xmath30 in @xmath19 such that + ( i ) @xmath31 and @xmath32 , @xmath33 ; + ( ii ) if @xmath34 then @xmath35 ; + ( iii ) @xmath36 . the idempotent @xmath37 is said to be _ primitive _ if it can not be written as @xmath38 where @xmath39 are idempotents such that @xmath40 and @xmath41 . a set of idempotents is said to be _ primitive _ if each idempotent in the set is primitive . various methods for constructing complete orthogonal sets of idempotents are derived below . such sets always exist in @xmath42 , the group ring over a field @xmath18 , when @xmath43 . see @xcite for properties of group rings and related definitions . these idempotent sets are related to the representation theory of @xmath42 . a mapping @xmath44 in which @xmath45 is said to be an _ involution _ on @xmath19 if and only if ( i ) @xmath46 , ( ii ) @xmath47 , and ( iii ) @xmath48 . we shall be particularly interested in the case where @xmath3 denotes complex conjugate transpose in the case of matrices over @xmath0 and denotes transpose for matrices over other fields . such a mapping @xmath3 on group rings is also defined below . an element @xmath49 is said to be _ symmetric _ ( relative to @xmath3 ) if @xmath50 and a set of elements is said to be symmetric if each element in the set is symmetric . @xmath51 denotes the tensor product of the matrices @xmath52 . as already noted @xmath53 is used to denote the complex conjugate transpose of a matrix @xmath54 . suppose @xmath19 is a ring with involution @xmath3 . then @xmath3 may be extended to matrices over @xmath19 as follows . let @xmath55 and define @xmath56 to be the matrix with each entry @xmath57 of @xmath58 replaced by @xmath59 . then define @xmath60 . this matrix @xmath61 has size @xmath62 . let @xmath63 be a matrix with polynomial in variables @xmath64 over some ring with involution @xmath3 . define @xmath65 to be @xmath66 . when @xmath54 is used for @xmath63 write @xmath53 to mean @xmath67 . ( in other words consider ` complex conjugate transposed ' of a variable @xmath68 to be @xmath69 ; this is consistent with group / group ring considerations . ) let @xmath19 be a ring with involution @xmath3 . for @xmath70 $ ] define @xmath71 . say @xmath72 is a _ paraunitary element _ in @xmath73 $ ] ( relative to @xmath3 ) if and only if @xmath74 . suppose @xmath75 and @xmath76 are rows of blocks of a matrix @xmath77 where each block is of the same size . then define the _ block inner product _ of @xmath78 and @xmath79 , written @xmath80 , to be @xmath81 . this is to include the case when @xmath82 are polynomial matrices and the @xmath83 are defined as above . the building methods using complete sets of orthogonal idempotents for the 1d paraunitary matrices in this section are generalised later in section [ multi ] below and following . the next section [ idems ] considers methods for designing such complete sets of orthogonal idempotents . [ first ] let @xmath84 be a complete orthogonal set of idempotents in a ring @xmath19 . . then @xmath86 . since @xmath30 is a complete set of orthogonal idempotents , @xmath87 . [ second ] if @xmath88 is symmetric then @xmath89 . thus @xmath90 is a paraunitary element when @xmath88 is a symmetric orthogonal complete set of idempotents . it is not necessary to use primitive idempotents . note also that if @xmath91 is a complete set of orthogonal idempotents then @xmath92 may be replaced by @xmath93 in @xmath94 and the result is ( still ) a complete set of orthogonal idempotents . this idea may be used to obtain real paraunitary matrices from ( complex ) complete orthogonal sets of idempotents in group rings . we single out the case @xmath95 for special mention . let @xmath96 be a complete symmetric set of orthogonal idempotents in the ring @xmath97 of @xmath98 matrices over @xmath18 . then @xmath99 is a paraunitary 1d @xmath100 matrix over @xmath18 where the @xmath101 are non - negative integers . in the group ring case a paraunitary element in @xmath42 with @xmath102 gives a paraunitary matrix in @xmath97 via the embedding of @xmath42 into @xmath97 as given for example in @xcite . suppose @xmath103 is an orthogonal symmetric complete set of idempotents in @xmath97 and that @xmath77 is a unitary matrix . then also @xmath104 is a symmetric complete orthogonal set of idempotents in @xmath97 . for our purposes say a paraunitary matrix @xmath77 is separable if it can be written in the form @xmath105 or @xmath106 where @xmath52 are paraunitary with @xmath107 ; otherwise say @xmath77 is non - separable . the following standard lemma is included for completeness and is not needed subsequently ; the proof is omitted . [ trans ] suppose @xmath108 is a paraunitary matrix . then @xmath109 and @xmath110 are paraunitary matrices . in proposition [ first ] the coefficients of the idempotents are @xmath111 times monomials . this can be extended in @xmath0 to coefficients with modulus @xmath4 times monomials . in @xmath20 and fields of finite characteristic define @xmath112 and then @xmath111 are the only elements which satisfy @xmath113 . suppose @xmath114 is a complete symmetric orthogonal set of idempotents in @xmath97 . define @xmath115 and then @xmath116 . here if @xmath117 , then @xmath118 , the complex conjugate of @xmath119 , and for other fields @xmath120 . use @xmath121 to mean @xmath122 for any field . therefore @xmath123 @xmath124 is a paraunitary matrix if and only if @xmath125 for each @xmath126 . if each @xmath127 then from ( * * ) @xmath128 . if on the other hand @xmath129 then multiplying ( * * ) through ( on right ) by @xmath130 gives @xmath131 from which it follows that @xmath125 . thus proposition [ first ] may be generalised as follows : [ second2 ] let @xmath132 be a complete symmetric orthogonal set of idempotents and @xmath115 , with @xmath133 and @xmath134 for each @xmath135 . then @xmath124 is a paraunitary matrix . now in @xmath0 , @xmath136 if and only if @xmath137 for real @xmath138 with @xmath139 and in @xmath20 , @xmath140 if and only if @xmath141 . in a field of characteristic @xmath142 , @xmath143 if and only if @xmath144 or @xmath145 . as expected unitary matrices are built from complete symmetric orthogonal sets of matrices as per proposition [ second2 ] : [ 7 ] @xmath146 is a unitary @xmath100 matrix over @xmath0 if and only if @xmath147 where @xmath148 is an orthonormal basis for @xmath149 and @xmath150 . further the @xmath151 are the eigenvalues of @xmath146 . suppose @xmath152 with @xmath148 an orthonormal basis and @xmath153 . then @xmath154 and so the @xmath151 are the eigenvalues of @xmath146 . it follows from proposition [ second2 ] that @xmath146 is unitary since @xmath155 is a complete symmetric orthogonal set of idempotents . suppose then @xmath146 is a unitary matrix . it is known that there exists a unitary matrix @xmath77 such that @xmath156 where @xmath157 is diagonal with entries of modulus @xmath4 . then @xmath158 where @xmath159 is an orthonormal basis ( of row vectors ) for @xmath149 and @xmath160 with @xmath153 and the @xmath151 are the eigenvalues of @xmath146 . then @xmath161 thus unitary matrices are generated by complete symmetric orthogonal sets of idempotents formed from the diagonalising unitary matrix . notice that the @xmath151 are the eigenvalues of @xmath146 . for example consider the real orthogonal / unitary matrix @xmath162 . this has eigenvalues @xmath163 and @xmath164 is a diagonalising unitary matrix . take the rows @xmath165 of @xmath77 and consider the complete orthogonal symmetric set of idempotents @xmath166 . then applying propositon [ 7 ] gives @xmath167 , which may be checked independently . a product of paraunitary matrices and the tensor product of paraunitary matrices are also paraunitary matrices . thus further paraunitary matrices may be designed using these products from those already constructed . paraunitary matrices are designed from complete symmetric sets of orthogonal idempotents in section [ 1d1 ] and also in later sections . here we concentrate on how such sets may be constructed . let @xmath168 . assume @xmath169 has an inner product so that the notion of orthonormal basis exists in @xmath170 and its subspaces . in @xmath171 and @xmath172 the inner product is @xmath173 for row vectors @xmath174 where @xmath3 denotes complex conjugate transpose ; in @xmath171 , @xmath175 , the transpose of @xmath176 . suppose now @xmath177 is any direct decomposition of @xmath170 . let @xmath178 denote the projection of @xmath170 to @xmath179 . then @xmath178 is a linear transformation on @xmath170 and ( i ) @xmath180 ; ( ii ) @xmath181 ; ( iii ) @xmath182 . thus @xmath183 is complete orthogonal set of idempotents . if each @xmath178 is an orthogonal projection then this set is a complete symmetric orthogonal set of idempotents . the matrix of @xmath178 may be obtained as follows when @xmath178 is an orthogonal projection . let @xmath184 be an orthonormal basis for @xmath179 and consider @xmath185 . then @xmath186 where @xmath187 and @xmath188 . here @xmath189 means omitting that term . then @xmath190 is given by @xmath191 . now @xmath192 . take the inner product with @xmath193 to get @xmath194 . hence @xmath195 . thus the matrix of @xmath178 is @xmath196 . on the other hand suppose @xmath197 is a complete symmetric orthogonal set of idempotents in @xmath97 . then @xmath178 defines a linear map @xmath198 by @xmath199 . let @xmath179 denote the image of @xmath178 . then it is easy to check that @xmath200 . the case when each @xmath179 has dimension @xmath4 is worth looking at separately . suppose @xmath201 is an orthonormal basis for @xmath169 . such bases come up naturally in unitary matrices . let @xmath178 denote the projection of @xmath169 to the space generated by @xmath202 . then @xmath203 is an orthogonal symmetric complete set of idempotents in the space of linear transformations of @xmath169 . it is easy to obtain the matrices of @xmath178 . the matrices @xmath178 may be combined and the resulting set is ( still ) a complete symmetric orthogonal sets of idempotents . for example @xmath204 ( @xmath205 ) is still idempotent and is the projection of @xmath169 to the space generated by @xmath206 ; replace @xmath207 by @xmath208 in @xmath77 and the new set is ( still ) an orthogonal symmetric complete set of idempotents . then @xmath209 also see lemma [ trrank ] below . for example @xmath210 is an orthonormal basis for @xmath211 . the projection matrices are respectively @xmath212 . thus @xmath213 is a complete symmetric orthogonal set of idempotents and each @xmath178 has rank 1 . set @xmath214 and then @xmath215 is a complete symmetric orthogonal set of idempotents also and @xmath216 . note that the inner product in @xmath172 is @xmath173 for row vectors @xmath174 where @xmath3 denotes _ complex conjugate _ transposed . for example @xmath217 is an orthonormal basis for @xmath218 . projecting then gives the complete orthogonal symmetric set of idempotents @xmath219 . let @xmath146 be a unitary or paraunitary @xmath100 matrix in variables @xmath220 say over @xmath19 . then the rows @xmath221 of @xmath146 satisfy @xmath222 . define @xmath223 for @xmath224 . then @xmath178 is an @xmath100 matrix of rank @xmath4 . @xmath225 is a complete symmetric orthogonal set of idempotents in @xmath226 $ ] . it is easy to check that @xmath227 . it is necessary to show that the set is complete . let @xmath228 . note that @xmath229 . then @xmath230 . thus @xmath54 has @xmath17 linearly independent eigenvectors corresponding to the eigenvalue @xmath4 . hence @xmath231 . in @xmath232 let @xmath233 denote the matrix with @xmath234 on the ( diagonal ) @xmath235 position and @xmath236 elsewhere . then @xmath237 is a complete symmetric orthogonal set of idempotent matrices . this is a special case of section [ joint ] but is worth mentioning separately ; paraunitary matrices have been designed from @xmath13 which , although not generally useful in themselves directly , may be combined with other designed paraunitary matrices with which they do not commute in general . group rings are a neat way with which to obtain complete orthogonal symmetric sets of idempotents . these systems have nice structures from which properties of the paraunitary matrices designed may be deduced . let @xmath238 be an element in the group ring @xmath42 and @xmath13 denotes the matrix of @xmath176 as defined in @xcite and @xcite . this matrix @xmath13 depends on the listing of the elements of @xmath239 and relative to this listing @xmath240 is an embedding of @xmath42 into the ring of @xmath100 matrices , @xmath241 , over @xmath18 where @xmath242 . transpose _ , @xmath243 , of @xmath176 is @xmath244 . note that the matrix of @xmath243 is then @xmath245 . over @xmath0 define @xmath246 where bar denotes complex conjugate . note then that for a group ring element @xmath176 with corresponding matrix @xmath13 the matrix of @xmath247 is indeed @xmath248 . say an element @xmath249 $ ] is a paraunitary group ring element if and only if @xmath250 and this happens if and only if the corresponding @xmath251 $ ] is a paraunitary matrix ( where @xmath252 ) . the @xmath253 obtained from @xmath72 in this case is termed a group ring paraunitary matrix . group rings are a rich source of complete sets of orthogonal idempotents and group rings have a rich structure within which properties of the paraunitary matrices so designed may be obtained . the theory brings representation theory and character theory in group rings into play . the orthogonal idempotents are obtained from the conjugacy classes and character tables , see e.g. @xcite . the orthogonal sets of idempotents depend on the field under consideration and classes of paraunitary matrices over different fields such as @xmath21 , @xmath20 or finite fields are also obtainable . the primitive central idempotents of the complex group algebra @xmath254 are given by @xmath255 where @xmath256 runs through the irreducible ( complex ) characters @xmath256 of @xmath239 , see @xcite , theorem 5.1.11 page 185 , where the @xmath37 are expressed as @xmath257 . the idempotents from group rings are automatically symmetric . [ thm : third ] for the group idempotents @xmath37 , @xmath258 . this is a matter of showing that the coefficients @xmath259 and @xmath260 in each @xmath37 are complex conjugates of one another . but this is immediate as it is well - known that @xmath261 , and thus the result follows from the expression for @xmath37 given above . let @xmath130 denote the matrix of @xmath37 as per an embedding of the group ring into the ring of matrices as for example in @xcite . [ idem3 ] let @xmath262 be a complete set of orthogonal idempotents in a group ring and define @xmath263 where the @xmath101 are non - negative integers . then @xmath1 is a paraunitary matrix . [ idem4 ] let @xmath262 be a complete set of orthogonal idempotents in a group ring over @xmath0 and define @xmath264 where the @xmath101 are non - negative integers and @xmath265 . then @xmath1 is a paraunitary matrix . the formula for the @xmath266 as given above ( taken from @xcite ) may be used to construct complete orthogonal sets of idempotents . the computer algebra packages gap and magma can construct character tables and conjugacy classes from which complete sets of orthogonal idempotents in group rings may be obtained . the literature contains other numerous methods for finding complete ( symmetric ) orthogonal sets of idempotents in group rings . in general the paraunitary matrices designed using orthogonal sets of idempotents in the group ring over @xmath0 have complex coefficients but specialising and combining idempotents allows the design so that the coefficients may be in @xmath20 , the real numbers , or in @xmath21 , the rational numbers . when the group ring of the symmetric group @xmath267 is used the paraunitary matrices derived by these methods all have coefficients automatically in @xmath21 and when the group ring of dihedral group @xmath268 is used the coefficients are in @xmath20 . in general idempotents occur in complex conjugate pairs and these may be combined to give real coefficients resulting in paraunitary matrices with real coefficients . most of the results hold in the case when the characteristic of @xmath18 does not divide the order of @xmath239 ; in this case this means that the characteristic of @xmath18 does not divide the size @xmath17 of the @xmath269 matrices under consideration . in these cases also it may be necessary to extend the field to include roots of certain polynomials . it is easy to check that the tensor product of paraunitary matrices is also a paraunitary matrix . if @xmath270 then @xmath271 when the products @xmath272 can be formed . complete orthogonal sets of idempotents may be designed using products of these sets . suppose @xmath273 is a complete orthogonal set of idempotents in @xmath97 and @xmath274 is a complete orthogonal set of idempotents in @xmath275 . then @xmath276 is a complete orthogonal set of idempotents in @xmath277 . here @xmath278 denotes tensor product . if both @xmath273 and @xmath274 are symmetric then so is the resulting tensor product set . the details are omitted . if @xmath279 and @xmath280 are complete orthogonal sets of idempotents within group rings @xmath281 respectively then @xmath282 is a complete orthogonal set of matrices in @xmath283 . suppose @xmath284 gives an embedding into matrices , then @xmath285 gives an embedding into @xmath283 ; this may be deduced from @xcite and details are omitted . the complete orthogonal symmetric systems of idempotents @xmath286 were obtained in section [ joint ] . then @xmath287 is a paraunitary matrix . 2 . let @xmath288 in @xmath13 in 1 . gives a unitary matrix , @xmath289 say . the rows of @xmath289 form an orthonormal basis for @xmath290 . these rows may then be used to form a complete symmetric orthogonal set of idempotents from which paraunitary matrices may be constructed . this process could be continued . 3 . in @xmath291 , where @xmath292 is the cyclic group of order @xmath293 , the orthogonal complete set of idempotents formed are @xmath294 where @xmath295 is a primitive 3rd root of unity . + then @xmath296 is a paraunitary matrix . 4 . give values of modulus @xmath4 to @xmath297 in @xmath298 above and get a unitary matrix @xmath19 . use the rows of @xmath19 to form a further complete symmetric set of idempotents from which paraunitary matrices may be formed . 5 . combine @xmath298 in 3 . with @xmath124 in 1 . to give for example the paraunitary matrix @xmath299 . we give some examples from orthogonal sets of idempotents derived from group rings . the group ring idea is used later as a prototype in which to extend the method for the design of non - separable paraunitary matrices . the complete orthogonal sets of idempotents obtained from group rings are automatically symmetric as noted in . recall that @xmath300 denotes the circulant @xmath100 matrix with first row @xmath301 . consider @xmath302 where @xmath303 is a cyclic group of order @xmath17 . 1 . when @xmath304 the ( primitive ) orthogonal set of idempotents consists of @xmath305 , where @xmath259 generates @xmath306 . thus paraunitary matrices may be formed from @xmath307 and @xmath308 giving for example @xmath309 . ( looks familiar ? ) 2 . these may be combined with paraunitary matrices formed from @xmath310 . note that @xmath311 do not commute with @xmath312 . for example the following is a paraunitary matrix : + @xmath313 + they may also be combined with paraunitary matrices formed from orthonormal bases as in section [ joint ] such as @xmath219 . + the determinant of these matrices which are powers of @xmath297 may be obtained from theorem [ det ] below . the primitive orthogonal idempotents for a cyclic group are related to the fourier matrix . 4 . in @xmath314 , for example , the orthogonal primitive idempotents are @xmath315 from which @xmath316 paraunitary matrices may be constructed . here @xmath295 is a primitive @xmath317 root of unity and in this case @xmath318 . + notice that @xmath319 as could be deduced from . combine the @xmath37 to get real sets of orthogonal idempotents . note that it is simply enough to combine the conjugacy classes of @xmath259 and @xmath260 . in this case then we get + @xmath320 , which can then be used to construct real paraunitary @xmath316 matrices . 6 . using @xmath321 gives different paraunitary matrices . here the set of primitive orthogonal idempotents consists of @xmath322 and the paraunitary matrices derived are all real . the paraunitary matrices produced from @xmath323 from @xmath321 and from @xmath324 may then be combined to produce further ( @xmath316 ) paraunitary matrices . so for example the following @xmath316 is a paraunitary matrix : @xmath325 again the determinant of the matrix may be obtained from theorem [ det ] . + the @xmath326 are derived from the @xmath327 ( as per @xcite ) so for example @xmath328 , @xmath329 . by combining complex conjugate idempotents in a complete orthogonal sets of complex idempotents , real paraunitary matrices may be obtained . we illustrate this with an example . suppose @xmath330 is the complete set of primitive idempotents in @xmath331 . here then @xmath332 where @xmath333 is a primitive @xmath334 root of unity and @xmath335 is generated by @xmath259 . then @xmath336 . let @xmath337 . note that @xmath338 . now combine @xmath339 with @xmath340 and @xmath341 with @xmath342 to get @xmath343 and @xmath344 this gives the real orthogonal complete set of idempotents @xmath345 from which real paraunitary matrices may be constructed . the ranks of the idempotents and determinants of the paraunitary matrices formed may be deduced from lemma [ trrank ] and theorem [ det ] . [ symm6 ] let @xmath268 denote the dihedral group of order @xmath346 . as every element in @xmath268 is conjugate to its inverse , the complex characters of @xmath268 are real . thus the paraunitary matrices obtained directly from the complete orthogonal set of idempotents in @xmath347 have real coefficients . the characters @xmath268 are contained in an extension of @xmath21 of degree @xmath348 and this is @xmath21 only for @xmath349 . let @xmath267 denote the symmetric group of order @xmath17 . representations and orthogonal idempotents of the symmetric group are known ; see for example @xcite . the characters of @xmath267 are rational and thus the paraunitary matrices produced directly from the complete orthogonal set of idempotents in @xmath350 have rational coefficients . the paraunitary matrices formed from different group rings ( with same size group ) may be combined to form further paraunitary matrices ; these in general will not commute . we present an example here from @xmath351 , the symmetric group on 3 letters . ( note that @xmath352 . ) now @xmath353 where these are cycles . we also use this listing of @xmath351 when constructing matrices . there are three conjugacy classes : @xmath354 ; @xmath355 ; @xmath356 . define + @xmath357 , + @xmath358 , + @xmath359 , and @xmath360 . then @xmath361 form a complete orthogonal set of idempotents and may be used to construct paraunitary matrices . the @xmath239-matrix of @xmath351 ( see @xcite ) is @xmath362 thus the matrices of @xmath363 are respectively @xmath364 note that @xmath365 have @xmath366 and that @xmath367 has @xmath368 ; the proof for the ranks of these @xmath130 in general is contained in lemma [ trrank ] . thus for example the following are paraunitary matrices : @xmath369 , @xmath370 , @xmath371 . the paraunitary matrices formed from these idempotent matrices may then be combined with paraunitary matrices formed using complete orthogonal idempotents obtained from @xmath331 and ones using @xmath372 . let @xmath373 be the orthogonal idempotents from @xmath331 . let @xmath374 . ( @xmath375 is the matrix of @xmath376 . ) then products of @xmath377 are paraunitary matrices . note that the @xmath378 do not commute . here we consider constructing examples of complete symmetric sets of orthogonal idempotents over finite fields . suppose @xmath379 is a orthogonal basis for @xmath380 under @xmath381 . thus @xmath382 for @xmath205 . suppose also @xmath383 . define @xmath384 which is a @xmath385 matrix . then @xmath386 and @xmath387 for @xmath205 . it also follows that @xmath388 . to see this consider @xmath389 . then @xmath390 as @xmath391 for @xmath205 and @xmath392 . hence @xmath393 . let @xmath394 . then @xmath395 is non - singular as @xmath379 is linearly independent . then @xmath396 and hence @xmath397 . note that in the above we do not need to take the square root of elements . another way could be to construct such sets over @xmath21 and when the denominators do not involve a prime dividing the order of the field it is then possible to derive complete symmetric orthogonal sets of idempotents over the finite field . for example the complete orthogonal symmetric systems of idempotents @xmath286 were obtained in section [ joint ] . over a field of characteristic @xmath398 these come to the trivial set @xmath399 of symmetric complete orthogonal set of idempotents . over the field @xmath400 of 5 elements they become ( note that here @xmath401 ) : @xmath402 . this is a complete symmetric orthogonal set of idempotents in @xmath400 which may be checked independently . the following are complete symmetric orthogonal sets of idempotents over @xmath403 : @xmath404 , @xmath405 . these different sets may be used to construct paraunitary matrices over @xmath403 and in a later section are used to show how to construct as an example a non - separable paraunitary matrix over a finite field . the great factorisation theorem of belevitch and vaidyanathan , see @xcite ( pp . 302 - 322 ) , is that matrices of the form @xmath406 , where @xmath407 is any unit column vector ( @xmath408 ) , are the building blocks for 1d paraunitary matrices over @xmath0 . consider @xmath409 where @xmath407 is a unit column vector and so @xmath410 . thus @xmath411 and so @xmath14 is an idempotent . hence @xmath412 is a complete symmetric orthogonal set of these ( two ) idempotents with @xmath413 and @xmath414 where the matrices have size @xmath100 ; see theorem [ trrank ] below for rank result . then @xmath415 and hence the paraunitary 1d matrices are built from complete symmetric orthogonal sets of two idempotents , one of which has @xmath366 and the other has @xmath416 . let @xmath18 be a field in which every element has a square root . suppose also an involution @xmath3 is defined on the set of matrices over @xmath18 . then @xmath77 is a symmetric ( with respect to @xmath3 ) idempotent of @xmath366 in @xmath97 if and only if @xmath417 where @xmath407 is a column vector such that @xmath418 . ( note that ` symmetric with respect to @xmath3 ' in the case of @xmath0 in which @xmath3 denotes complex conjugate transposed is normally termed ` hermitian ' . ) if @xmath419 with @xmath420 then @xmath77 is a symmetric idempotent of @xmath366 . suppose @xmath77 is a symmetric idempotent of @xmath366 in @xmath97 . since @xmath77 has rank 1 each row is a multiple of any non - zero row . suppose the first row is non - zero and that the first entry of this row is non - zero . proofs for other cases are similar . since @xmath77 is symmetric it has the form @xmath421 with @xmath422 . since @xmath77 is idempotent it follows that @xmath423 . . then @xmath425 and @xmath426 . it is necessary that square roots exist in the field . for example @xmath427 over @xmath428 is a symmetric idempotent matrix of rank @xmath4 but can not be written in the form @xmath429 ; however @xmath398 does not have a square root in @xmath428 . note that @xmath430 and that @xmath431 is a complete orthogonal set of idempotents in @xmath432 . over @xmath428 the following complete symmetric sets of idempotents are the building blocks for @xmath433 matrices : @xmath434 , @xmath435 . thus the paraunitary matrices @xmath436 matrices over @xmath428 are built from these sets using proposition [ second2 ] and products . these sets are not of belevitch and vaidyanathan form and their result does not apply here . the 1d building block result of belevitch and vaidyanathan can not be extended to multidimensions . group ring 1d paraunitary matrices and other 1d paraunitary matrices over @xmath0 constructed here can in theory then be obtained from this characterisation . group rings have special features and paraunitary matrices from these have nice structures . [ 1d ] [ there ] let @xmath437 be a complete symmetric orthogonal set of idempotents and define products of non - negative powers of variables by @xmath438 for @xmath439 where @xmath440 and @xmath441 are non - negative integers . define @xmath442 . then @xmath253 is a @xmath5-dimensional paraunitary element . since @xmath437 is a complete symmetric orthogonal set of idempotents , @xmath443 . as before for @xmath117 define @xmath444 , the complex conjugate of @xmath119 , and for other fields define @xmath120 . define @xmath445 in all cases . [ there1 ] let @xmath437 be a complete symmetric orthogonal set of idempotents and define products of non - negative powers of the variables by @xmath446 for @xmath439 where @xmath440 and @xmath441 are non - negative integers and @xmath125 . define @xmath442 and then @xmath253 is a paraunitary matrix . the @xmath253 so formed can be combined using products of matrices , or tensor products of matrices when appropriate , to form further paraunitary matrices . such paraunitary matrices formed from propositions [ there ] and [ there1 ] can however be shown to be separable but have uses of their own and will be used later to form ( constituents of ) non - separable paraunitary matrices . recall that if @xmath57 is a group ring element of @xmath42 with @xmath447 then @xmath146 ( capital letter equivalent ) denotes the matrix of @xmath57 under the embedding of @xmath42 into the ring @xmath97 of @xmath100 matrices over @xmath18 , see @xcite . let @xmath448 be a ( primitive ) orthogonal complete set of idempotents in @xmath291 , and thus @xmath449 is an orthogonal complete symmetric set of idempotents in @xmath450 . let @xmath451 be an orthogonal complete set of idempotents in @xmath314 . define @xmath452 and let @xmath453 be obtained from @xmath454 by replacing each @xmath37 by @xmath130 and each @xmath376 by @xmath375 . then @xmath453 is a paraunitary matrix . however here @xmath455 and so @xmath453 is separable . let @xmath456 . then @xmath453 is a paraunitary matrix . paraunitary matrices may also be constructed from matrices with blocks of complete orthogonal sets of idempotents . consider the following example . let @xmath457 . define @xmath458 . then @xmath459 as @xmath460 is an orthogonal symmetric complete set of idempotents . however @xmath13 is separable as @xmath461 . ( * * ) and each of the matrices on the right in ( * * ) is separable into 1d paraunitary matrices . here if we let @xmath462 then ( * * ) is a trivial product as the first matrix on the right is then the identity . if further @xmath463 this produces the matrix @xmath464 which is a common matrix used in quantum theory as non - separable . ( this matrix @xmath465 with fraction omitted is a hadamard regular matrix . ) thus non - separability of a paraunitary matrix is a stronger condition by comparison . let @xmath466 . @xmath467 . then @xmath13 is a paraunitary matrix . now letting the variables have complex values of modulus @xmath4 gives rise to complex hadamard regular matrices as for example @xmath468 . let @xmath449 be an orthogonal symmetric complete set of idempotents in @xmath469 . define @xmath470 . the variables of @xmath13 are @xmath471 which need not necessarily be distinct . then @xmath472 . for example in section [ joint ] the following complete set of symmetric idempotents was constructed in @xmath473 : @xmath474 . then @xmath475 is a paraunitary matrix . let @xmath476 be a complete symmetric orthogonal set of idempotents in @xmath97 . arrange these into a @xmath385 block matrix of with each row of blocks containing one of the blocks @xmath477 exactly once . now attach monomials to each @xmath130 ; the same monomial need not be used with each @xmath130 that appears . let @xmath13 be the resulting matrix . @xmath13 is a paraunitary matrix . take the block inner product of two different rows of blocks . the @xmath130 are orthogonal to one another so the result is @xmath236 . take the block inner product of any row of blocks with itself . this gives @xmath478 . hence @xmath479 . the @xmath13 is a paraunitary matrix in the union of the variables of the monomials . the condition that each row and column block contains each @xmath130 once and once only can be obtained by using the group ring matrix of any group of order @xmath5 ; see for example @xcite . so for example the @xmath130 could be arranged as a circulant block of matrices . different arrangements will in general give inequivalent paraunitary matrices . ( modifications in the construction of @xmath13 by attaching elements of modulus @xmath4 as coefficients will give paraunitary matrices but further conditions are necessary on these elements.these modifications are not included here . ) these are nice constructions for paraunitary matrices but can be shown to be separable . however they have uses in themselves and will prove useful later as parts of constructions of non - separable paraunitary matrices . they may also be used to construct special and regular types of hadamard real and complex matrices . these are illustrated in the following sections [ two ] and [ three4 ] . although the matrices of idempotents as constructed in sections [ gyt ] and [ grty ] produce separable paraunitary matrices these can be useful structures in themselves ; they may also be used in certain cases to produce ` regular ' hadamard matrices . ( walsh - hadamard matrices are regular hadamard matrices which have been used extensively in communications theory . ) if in a paraunitary matrix the entries are @xmath111 times monomials in the variables then substituting @xmath111 for the variable gives a hadamard matrix . if in a paraunitary matrix the entries are @xmath295 times monomials where @xmath295 is a complex number of modulus @xmath4 then substituting each variable by a complex number of modulus @xmath4 gives a complex hadamard matrix . for example use @xmath480 where @xmath295 is a primitive third root of unity gives the following matrix : @xmath481 . substituting a complex number of modulus @xmath4 for each of @xmath482 gives a complex hadamard matrix . butson - type hadamard matrices @xmath483 are complex hadamard @xmath100 matrices with entries which are @xmath484 roots of unity . with the above example substituting a third root of unity for the variables gives a hadamard @xmath485 matrix , that is a a matrix @xmath465 with entries which are third roots of unity so that @xmath486 . these matrices could then be used to produce hadamard @xmath487 matrices . this can be extended to @xmath488 matrices using the complete orthogonal set of idempotents for the cyclic group of order @xmath23 ; this will involve @xmath484 roots of unity and is related to the representation theory of the finite cyclic group . from this hadamard @xmath489 matrices can be produced and from these hadamard @xmath490 matrices can be produced and so on . it is noted that interchanging rows and/or columns in a paraunitary matrix results in a paraunitary matrix and thus interchanging blocks of rows and/or columns results in a paraunitary matrix . when using complete orthogonal sets of idempotents the blocks of row and/or columns of idempotents can be interchanged in the construction stage . the resulting paraunitary matrices can take a regular form . in certain cases the variables can be specialised to form regular hadamard matrices . here examples are given but details of the constructions are omitted . let @xmath491 be a complete symmetric orthogonal set of idempotents in @xmath492 . for example in @xmath493 these could be @xmath494 . define @xmath495 . then @xmath13 is a paraunitary matrix in the variables @xmath496 . interchanging blocks of rows and/or columns in @xmath13 will also give a paraunitary matrix which is equivalent to @xmath13 . by varying the signs other constructions of paraunitary matrices are obtained and these are not generally equivalent to one another . they may seem to be non - separable but by interchanging blocks of rows and blocks of columns it can be shown that they are separable . however they are interesting in themselves with interesting properties and can also be used to construct hadamard matrices of regular types . examples may also be interpreted as coming from the structure of group rings of various groups . then giving the values @xmath111 to the variables can result in regular hadamard matrices or giving the values @xmath497 to the variables can result in hadamard complex matrices . hadamard matrices with entries which are roots of unity may also be obtained from these constructions . the following is an example of this type . @xmath498 this matrix has the form @xmath499 where @xmath500 and has the structure of the group ring of @xmath501 . it is clear that @xmath502 may each be replaced by a monomial times a complex number of modulus @xmath4 in @xmath13 and a paraunitary matrix is obtained . values of @xmath111 may be given to the variables in @xmath13 and with the fraction omitted this gives a regular real hadamard matrix . values of modulus @xmath4 may be given to to the variables in @xmath13 and with the fraction omitted a regular complex hadamard matrix is obtained . other group ring structures can also arise in this manner . walsh - hadamard matrices have the structure of the group ring of @xmath503 . these examples may also be extended in a similar way . so for example constructions with the structure of the group ring @xmath504 may be made if paraunitary matrices with the structure of @xmath239 can be formed such as when @xmath505 or when @xmath506 and others . several methods have now been developed for constructing multidimensional paraunitary methods from complete orthogonal sets of matrices . the matrices produced are useful in many ways but are found to be separable although not appearing so initially . the methods use just one complete orthogonal set of idempotents in the various constructions . thus we are led to consider _ different _ complete orthogonal sets of variables and to ` tangle ' them up in order to construct non - separable paraunitary matrices . [ fty1 ] let @xmath507 be paraunitary matrices of the same size but not necessarily with the same variables over a field in which @xmath398 has a square root . then @xmath508 and @xmath509 are paraunitary matrices in the union of the variables in @xmath507 . suppose @xmath507 are @xmath100 matrices . then @xmath510 the case for @xmath395 can be considered similarly or it follows from lemma [ trans ] since @xmath395 is the transpose of @xmath13 . paraunitary matrices constructed by methods of previous sections may be used as input to proposition [ fty1 ] to construct paraunitary matrices . matrices constructed using the proposition can then also be used as input . the methods are fairly general and it is easy to produce examples for input using various complete orthogonal sets of idempotents . the result holds in general over any field which contains the square root of @xmath398 . if @xmath511 then @xmath13 in proposition [ fty1 ] is the tensor product @xmath512 where @xmath513 . if @xmath54 and @xmath514 are formed using the same complete symmetric orthogonal set of idempotent as in [ multi ] or [ gyt ] then @xmath13 can be shown to be separable . it would appear initially that proposition [ fty1 ] could / should be generalised to @xmath515 where @xmath516 are monomials or in the case of @xmath0 where @xmath516 are monomials times a complex number of modulus @xmath4 . however then @xmath13 is separable as a product @xmath517 . if @xmath518 where @xmath519 are matrices of the same size then @xmath519 are referred to as the _ blocks _ of @xmath13 and @xmath520 and @xmath521 are the _ row blocks _ of @xmath13 . similarly _ column blocks _ of @xmath13 are defined . suppose @xmath507 are matrices of the same size . then a _ tangle _ of @xmath522 is one of 1 . 2 . a matrix obtained from 1 . by interchanging rows of blocks and/or columns of blocks . 3 . the transpose of any matrix obtained in 1 . or 2 . a tangle of @xmath522 is not the same as , and is not necessarily equivalent to , a tangle of @xmath524 . note that interchanging any rows and/or columns of a paraunitary matrix results in an ( equivalent ) paraunitary matrix . thus in particular interchanging rows and/or columns of blocks also results in equivalent paraunitary matrices ; thus item 2 . gives equivalent paraunitary matrices to item 1 . the negative of a paraunitary matrix is a paraunitary matrix as is the @xmath3 of a paraunitary matrix . for example + @xmath525 are tangles of @xmath522 + and + @xmath526 are tangles of @xmath524 . proposition [ fty1 ] may be generalised as follows . [ fty ] let @xmath507 be paraunitary matrices of the same size but not necessarily with the same variables . then a tangle of @xmath522 or @xmath524 is a paraunitary matrix . use the expression ` @xmath54 is tangled with @xmath514 ' to mean that a tangle of @xmath522 or @xmath524 is formed . 1 . 1 . construct @xmath527 and @xmath528 . 2 . construct @xmath529 . then @xmath13 is a paraunitary matrix . 3 . similarly construct @xmath530 . 4 . tangle @xmath13 and @xmath395 to produce for example the paraunitary matrix @xmath531 . the process can be continued : matrices produced from ( d ) , with different variables , can be input to form further paraunitary matrices . 2 . 1 . construct as is [ finite ] the following complete symmetric sets of idempotents in @xmath532 matrices over @xmath403 : + @xmath533 , @xmath534 . 2 . form @xmath535 . tangle @xmath507 to form for example the following paraunitary matrix over @xmath403 : @xmath536 . 3 . 1 . construct , in @xmath493 , the following complete symmetric ( different ) sets of orthogonal idempotents @xmath491 and @xmath537 where : + @xmath538 @xmath539 2 . construct @xmath540 . 3 . construct @xmath508 . then @xmath13 is a paraunitary matrix of size @xmath316 with variables @xmath541 . 4 . 1 . construct different @xmath491 and @xmath537 complete symmetric orthogonal sets of idempotents in @xmath542 . 2 . construct @xmath543 . then @xmath13 is a paraunitary matrix . + it is essential that @xmath491 and @xmath537 are _ different _ complete orthogonal sets of idempotents in order for @xmath13 to be non - separable although the construction does not depend on this . 3 . clearly also the roles of @xmath491 and @xmath537 can be interchanged in @xmath13 and a paraunitary matrix is still obtained . changing the @xmath544 signs in such a way that the block inner product of any two rows of blocks is @xmath236 will give a different inequivalent paraunitary matrix . thus for example @xmath13 could be replaced by the following : 4 . construct @xmath545 . then this @xmath13 is also a paraunitary matrix . 5 . 1 . see section [ idems ] for methods for constructing complete orthogonal symmetric sets of idempotents . examples of such sets in @xmath546 are @xmath547 where these are given as in example 2 above another complete symmetric orthogonal set of idempotents in @xmath546 is the following : + @xmath548 . 3 . in example 2 . @xmath491 is ` tangled ' with @xmath537 . @xmath549 may similarly be combined ( ` tangled ' ) with either @xmath491 or @xmath550 to construct paraunitary matrices . 4 . using @xmath549 with @xmath491 produces paraunitary matrices of the form @xmath551 where the entries of @xmath77 are @xmath552 with @xmath139 . by specialising the variables , complex hadamard matrices may be obtained . for example the following is a paraunitary matrix : @xmath553 . 6 . by giving values of modulus @xmath4 to the variables , complex hadamard matrices are obtained for example letting all the variables have the value @xmath554 gives the following complex hadamard matrix : + @xmath555 . 6 . 1 . construct @xmath556 2 . construct the complete symmetric orthogonal set of idempotents obtained from the group ring @xmath291 of the cyclic group of order @xmath293 : @xmath557 where @xmath295 is a primitive cube root of @xmath4 ; 3 . construct @xmath558 and @xmath559 . 4 . construct @xmath508 . + then @xmath13 is a paraunitary matrix . it has size @xmath560 and @xmath561 variables ; variables can be equated . @xmath13 in the above could for example be replaced by @xmath562 where @xmath516 are variables and @xmath139 but as pointed out this is separable and may be constructed as a product . a complex hadamard matrix is a matrix @xmath465 of size @xmath100 with entries of modulus @xmath4 and satisfying @xmath563 . complex hadamard matrices arise in the study of operator algebras and the theory of quantum computation . by giving values which are @xmath564 of unity to the variables , with @xmath5 divisible by @xmath565 , in the above example 4 , special types of complex hadamard matrices which are called _ butson - type _ are obtained . a butson type hadamard @xmath483 matrix is a complex hadamard matrix of size @xmath100 all of whose entries are @xmath484 roots of unity . here is another example which uses group rings : . 1 . construct the complete symmetric set of orthogonal idempotents @xmath566 from the group ring @xmath331 of the cyclic group @xmath335 of order @xmath567 . this gives @xmath568 where @xmath295 is a primitive 6th root of unity . 2 . define @xmath569 . note that @xmath570 are real . 3 . let @xmath364 + be the complete symmetric set of orthogonal idempotents obtained from the group ring of @xmath571 as in section [ symm6 ] . 4 . define @xmath572 and @xmath559 . 5 . construct @xmath508 . + then @xmath13 is a paraunitary matrix which is real . it is a @xmath573 matrix with @xmath561 variables . 1 . construct different sets @xmath574 and @xmath575 of complete orthogonal symmetric of idempotents by the methods of section [ joint ] or section [ grring ] ( or otherwise ) in @xmath97 . ( @xmath18 is usually @xmath0 but other fields can also be used . ) 2 . construct a paraunitary matrix from @xmath574 by either the methods of section [ multi ] or the methods of section [ gyt ] . call this matrix @xmath54 . 3 . construct a paraunitary matrix from @xmath575 by either the methods of section [ multi ] or the methods of section [ gyt ] . call this matrix @xmath514 . the variables in @xmath507 can be different . 4 . construct a tangle of @xmath522 or @xmath524 . an example : 1 . construct @xmath576 and @xmath577 in @xmath578 where the @xmath178 and @xmath579 are obtained by construction methods of [ joint ] , [ grring ] . 2 . form @xmath580 . ( here the structure of @xmath323 is used and a circulant structure is obtained . ) form @xmath581 . 4 . let the matrix in 3 . be denoted by @xmath54 . form @xmath582 . ( here the structure of @xmath583 is used . ) let the matrix in 6 . be denoted by @xmath514 . form @xmath585 . 1 . input paraunitary matrices @xmath507 of the same size but not necessarily with the same variables . these may be formed from method of section [ sisty ] or from this algorithm . 2 . form a tangled product of @xmath522 or @xmath524 . the non - separable paraunitary matrices and separable paraunitary matrices can be combined when appropriate as products or as tensor products to construct further paraunitary matrices . these may then be input to algorithm of section [ further ] . let @xmath77 be a paraunitary @xmath100 matrix with variables @xmath64 over a field @xmath18 . then the rows @xmath586 of @xmath77 satisfy @xmath587 and @xmath588 for @xmath205 . note that @xmath589 means transpose conjugate over @xmath0 , and transpose over other fields , with the understanding that @xmath590 for a variable @xmath297 . let @xmath223 which are @xmath100 matrices of rank @xmath4 and involve the variables @xmath591 . then @xmath592 is a complete orthogonal symmetric set of idempotents in the polynomial ring @xmath593 $ ] . hence by the methods of sections [ idems ] , [ gyt ] , [ multi ] and [ nonsep ] paraunitary - type matrices may be formed ; for the method of section [ nonsep ] two such sets must be constructed . for example @xmath594 in variables @xmath595 satisfies @xmath596 . now @xmath13 is a matrix in the variables @xmath597 but can not be termed _ paraunitary_. call such a matrix a _ pseudo - paraunitary _ matrix . having constructed @xmath13 its rows may then be used to construct further pseudo - paraunitary matrices and so on . thus say @xmath598 $ ] is a _ pseudo - paraunitary _ matrix if @xmath596 . pseudo - paraunitary matrices may be constructed from paraunitary matrices and from pseudo - paraunitary matrices . here s an example . 1 . form @xmath599 . 2 . form @xmath600 3 . let @xmath601 . 4 . form @xmath602 . 5 . @xmath603 and @xmath604 . form @xmath605 . @xmath606 , and @xmath607 . 7 . the rows of @xmath13 can be used to form further pseudo - paraunitary matrices . consider @xmath608 as in this example above . define @xmath609 . define @xmath610 . then @xmath611 $ ] , and is a polynomial with @xmath612 . say @xmath613 $ ] is a _ pseudo - paraunitary _ matrix over @xmath614 $ ] if @xmath615 where @xmath142 is a monomial in @xmath616 . a pseudo - paraunitary matrix over @xmath617 $ ] may be used to construct a pseudo - paraunitary matrix over @xmath614 $ ] and vice versa . pseudo - paraunitary matrices in general may be constructed from paraunitary matrices and from pseudo - paraunitary matrices . here we consider properties of complete sets of idempotent matrices and ranks of the idempotents . it is known that @xmath620 for an idempotent matrix , see for example @xcite , and so @xmath621 for each @xmath126 . if @xmath622 is a set an orthogonal idempotent matrices so is @xmath623 . from this it follows that @xmath624 . now @xmath644 . thus each column of @xmath130 is an eigenvector of @xmath54 corresponding to the eigenvalue @xmath645 . thus there are at exist @xmath646 linearly independent eigenvectors corresponding to the eigenvalue @xmath645 . since @xmath647 , there are exactly @xmath646 linearly independent eigenvectors corresponding to the eigenvalue @xmath645 . let @xmath648 . let these @xmath649 linearly independent eigenvectors corresponding to @xmath645 be denoted by @xmath650 . do this for each @xmath126 . thus the set of vectors @xmath659 is linearly independent and form a basis for @xmath169 remember that @xmath660 . hence @xmath54 can be diagonalised by the matrix of these vectors and thus there is a non - singular matrix @xmath77 such that @xmath661 where @xmath157 is a diagonal matrix consisting of the @xmath645 repeated @xmath649 times for each @xmath662 . 99 m. n. do and m. vetterli , `` the contourlet transform : an efficient directional multiresolution image representation '' , ieee trans . on image processing , 14 , 12 , 2091 - 2106 , 2005 . oskar m. baksalary , dennis s. bernstein , gtz trenkler , `` on the equality between rank and trace of an idempotent matrix '' , applied mathematics and computation , 217 , 4076 - 4080 , 2010 . andreas klappenecker , `` on multirate filter bank structures '' , in _ transforms and filter banks , proc . of second international workshop on transforms and filter banks _ , r. creutzburg , j. astola ( eds . ) , ticsp 4 , pages 175 - 188 , 2000 . csar milies & sudarshan sehgal , _ an introduction to group rings _ , klumer , 2002 . gilbert strang and truong nguyen,_wavelets and filter banks _ , wesley - cambridge press , 1997 . xiqi gao , truong nguyen , gilbert strang , `` on factorization of @xmath58-channel paraunitary filterbanks '' , ieee trans . on signal processing , 49 , 7 , 1433 - 1446 , 2001 . vaidyanathan , _ multirate systems and filterbanks _ , prentice - hall , 1993 . c. w. curtis and i. reiner , _ representation theory of finite groups and associative algebras _ , interscience , new york , 1962 . paul hurley and ted hurley , `` codes from zero - divisors and units in group rings '' , int . j. inform . and coding theory , 1 , 57 - 87 , 2009 . paul hurley and ted hurley , `` block codes from matrix and group rings '' , chapter 5 , 159 - 194 , in _ selected topics in information and coding theory _ , eds . i. woungang , s. misra , s.c . misma , world scientific 2010 . ted hurley , `` group rings and rings of matrices '' , inter . j. pure & appl . , 31 , no.3 , 2006 , 319 - 335 . m. n. do and y. m. lu , _ multidimensional filter banks and multiscale geometric representations _ , submitted , 2011 . available on - line . f. delgosa , f. fekri , `` results on the factorization of multidimensional matrices for paraunitary filterbanks over the complex field '' , ieee trans . on signal processing , 52 , no.5 , 1289 - 1303 , 2004 . jianping zhou , minh n. do , and jelena kovaevi , `` special paraunitary matrices , cayley transform and multidimensional orthogonal filter banks '' , ieee trans . on image processing , 15 , no . 2 , 511 - 519 , 2006 . i. daubechies , `` orthonormal bases of compactly supported wavelets '' , comm on pure and appl . 41 , pp . 909996 , 1988 . f. fekri , r.m . mersereau , r.w . `` theory of paraunitary filter banks over fields of characteristic 2 '' , ieee trans . on information theory , 48 , 11 , 2964 - 2979 , 2002 . a. cohen and i. daubechies , `` nonseparable bidimensional wavelet bases '' , rev . mat . iberoamericana , vol . 9 , no . 1 , pp . 51137 , 1993 . j. kovaevi and m. vetterli , `` nonseparable multidimensional perfect reconstruction filter banks and wavelet bases for @xmath664 '' , ieee trans . on information theory , special issue on wavelet transforms and multiresolution signal analysis , vol . 38 , no . 2 , 533 - 555 , 1992

we resolve the question of whether fourier sampling can efficiently solve the hidden subgroup problem . specifically , we show that the hidden subgroup problem over the symmetric group can not be efficiently solved by strong fourier sampling , even if one may perform an arbitrary povm on the coset state . these results apply to the special case relevant to the graph isomorphism problem .

quant-ph0501056

many problems of interest in quantum computing can be reduced to an instance of the _ hidden subgroup problem _ ( hsp ) . we are given a group @xmath0 and a function @xmath1 with the promise that , for some subgroup @xmath2 , @xmath1 is invariant precisely under translation by @xmath3 : that is , @xmath1 is constant on the cosets of @xmath3 and takes distinct values on distinct cosets . we then wish to determine the subgroup @xmath3 by querying @xmath1 . for example , in simon s problem @xcite , @xmath4 and @xmath1 is an oracle such that , for some @xmath5 , @xmath6 for all @xmath7 ; in this case @xmath8 and we wish to identify @xmath5 . in shor s factoring algorithm @xcite @xmath0 is the group @xmath9 where @xmath10 is the number we wish to factor , @xmath11 for a random @xmath12 , and @xmath3 is the subgroup of @xmath9 whose index is the multiplicative order of @xmath13 . both simon s and shor s algorithms use the following approach , referred to as the _ standard method _ or _ fourier sampling _ @xcite . : : prepare two registers , the first in a uniform superposition over the elements of @xmath0 and the second with the value zero , yielding the state @xmath14 step 2 . : : query ( or calculate ) the function @xmath1 defined on @xmath0 and xor it with the second register . this entangles the two registers and results in the state @xmath15 step 3 . : : measure the second register . this puts the first register in a uniform superposition over one of @xmath1 s level sets , i.e. , one of the cosets of @xmath3 , and disentangles it from the second register . if we observe the value @xmath16 , we have the state @xmath17 where @xmath18 alternately , we can view the first register as being in a mixed state with density matrix @xmath19 where the sum includes one representative @xmath20 for each of @xmath3 s cosets . : : carry out the quantum fourier transform on @xmath21 and measure the result . ( note that in shor s algorithm , since @xmath22 is unknown , the fourier transform is performed over @xmath23 for some @xmath24 ; see @xcite or @xcite . ) in both simon s and shor s algorithms , the group @xmath0 is abelian ; it is not hard to see that , in this abelian case , a polynomial number ( i.e. , polynomial in @xmath25 ) of experiments of this type determine @xmath3 . in essence , each experiment yields a random element of the dual space @xmath26 perpendicular to @xmath3 s characteristic function , and as soon as these elements span @xmath26 we are done . while the _ nonabelian _ hidden subgroup problem appears to be much more difficult , it has very attractive applications . in particular , solving the hsp for the symmetric group @xmath27 would provide an efficient quantum algorithm for the graph automorphism and graph isomorphism problems ( see e.g. jozsa @xcite for a review ) . another important motivation is the relationship between the hsp over the dihedral group with hidden shift problems @xcite and cryptographically important cases of the shortest lattice vector problem @xcite . so far , algorithms for the hsp are only known for a few families of nonabelian groups , including wreath products @xmath28 @xcite ; more generally , semidirect products @xmath29 where @xmath30 is of polynomial size , and groups whose commutator subgroup is of polynomial size @xcite ; `` smoothly solvable '' groups @xcite ; and some semidirect products of cyclic groups @xcite . ettinger and hyer @xcite provided another type of result , by showing that fourier sampling can solve the hsp for the dihedral groups @xmath31 in an _ information - theoretic _ sense . that is , a polynomial number of experiments gives enough information to reconstruct the subgroup , though it is unfortunately unknown how to determine @xmath3 from this information in polynomial time . to discuss fourier sampling for a nonabelian group @xmath0 , one needs to develop the fourier transform over @xmath0 which relies on the group s _ linear representations_. for abelian groups , the fourier basis functions are homomorphisms @xmath32 such as the familiar exponential function @xmath33 for the cyclic group @xmath34 . in the nonabelian case , there are not enough such homomorphisms to span the space of all @xmath35-valued functions on @xmath0 ; to complete the picture , one introduces _ representations _ of the group , namely homomorphisms @xmath36 where @xmath37 is the group of unitary matrices acting on some @xmath35-vector space @xmath38 of dimension @xmath39 . it suffices to consider _ irreducible _ representations , namely those for which no nontrivial subspace of @xmath38 is fixed by the various operators @xmath40 . once a basis for each irreducible @xmath41 is chosen , the matrix elements @xmath42 provide an orthogonal basis for the vector space of all @xmath35-valued functions on @xmath0 . the quantum fourier transform then consists of transforming ( unit - length ) vectors in @xmath43}= \ { \sum_{g \in g } \alpha_g { \left| g \right\rangle } \mid \alpha_g \in { \mathbb{c}}\}$ ] from the basis @xmath44 to the basis @xmath45 where @xmath41 is the name of an irreducible representation and @xmath46 index a row and column ( in a chosen basis for @xmath38 ) . indeed , this transformation can be carried out efficiently for a wide variety of groups @xcite . note , however , that a nonabelian group @xmath0 does not distinguish any specific basis for its irreducible representations which necessitates a rather dramatic choice on the part of the transform designer . indeed , careful basis selection appears to be critical for obtaining efficient fourier transforms for the groups mentioned above . perhaps the most fundamental question concerning the hidden subgroup problem is whether there is always a basis for the representations of @xmath0 such that measuring in this basis ( in step 4 , above ) provides enough information to determine the subgroup @xmath3 . this framework is known as _ strong fourier sampling_. in this article , we answer this question in the negative , showing that natural subgroups of @xmath27 can not be determined by this process ; in fact , we show this for an even more general model , where we perform an arbitrary positive operator - valued measurement ( povm ) on coset states @xmath47 . we emphasize that the subgroups on which we focus are among the most important special cases of the hsp , as they are those to which graph isomorphism naturally reduces . [ [ related - work . ] ] related work . + + + + + + + + + + + + + the terminology `` strong fourier sampling '' @xcite was invented to distinguish this approach from the natural variant , called _ weak fourier sampling _ , where one only measures the name of the representation @xmath41 , and ignores the row and column information . weak fourier sampling is basis - independent , making it attractive from the standpoint of analysis ; however , it can not distinguish conjugate subgroups from each other , and hallgren , russell and ta - shma @xcite showed that it can not distinguish the trivial subgroup from an order-2 subgroup consisting of @xmath48 disjoint transpositions . specifically , they used character bounds to show that the probability distribution obtained on representation names for the trivial and order-2 subgroups are exponentially close in total variation distance : it requires an exponential number of such experiments to distinguish them . kempe and shalev @xcite have generalized this result to other conjugacy classes , and conjecture that one can do no better than classical computation with this approach . in an effort to shed light on the power of strong fourier sampling , grigni , schulman , vazirani and vazirani @xcite showed that , for groups such as @xmath27 , measuring in a _ random _ basis yields an exponentially small amount of information . this can be explained , roughly , by the fact that projecting a vector into a sufficiently high - dimensional random subspace results in tightly concentrated length . on the other hand , moore , rockmore , russell and schulman @xcite showed that for the affine and @xmath49-hedral groups , measuring in a well - chosen basis can solve the hsp ( at least information - theoretically ) in cases where random bases can not . [ [ our - contribution . ] ] our contribution . + + + + + + + + + + + + + + + + + in this paper we show that strong fourier sampling , in an arbitrary basis of the algorithm designer s choice , can not solve the hsp for @xmath27 . as in @xcite we focus on order-2 subgroups of the form @xmath50 where @xmath51 is an involution consisting of @xmath48 disjoint transpositions ; we remark that if we fix two rigid connected graphs of size @xmath48 and consider permutations of their disjoint union , then the hidden subgroup is of this form if the graphs are isomorphic and trivial if they are not . then we show that strong fourier sampling and more generally , arbitrary measurements of coset states cannot distinguish these subgroups from each other , or from the trivial subgroup , without an exponential number of experiments . we remark that our results do not preclude the existence of an efficient quantum algorithm for the hsp on @xmath27 . rather , they force us to consider _ multi - register _ algorithms , in which we prepare multiple coset states and subject them to entangled measurements . ettinger , hyer and knill @xcite showed that the hsp on arbitrary groups can be solved information - theoretically with a polynomial number of registers , and the authors have shown how to carry out such a measurement in the fourier basis @xcite . kuperberg @xcite devised a subexponential ( @xmath52 ) algorithm for the hsp on the dihedral group @xmath31 that works by performing entangled measurements on two registers at a time , bacon , childs , and van dam @xcite have determined the optimal multiregister measurement for the dihedral group , and the present authors have generalized this to all cases where the @xmath3 and @xmath0 form a gelfand pair @xcite . whether a similar approach can be taken to the symmetric group is a major open question . in a companion paper @xcite , the first two authors take a step towards answering this question by showing that if we perform arbitrary entangled measurements over pairs of registers , distinguishing @xmath53 from the trivial group in @xmath27 requires a superpolynomial number ( specifically , @xmath54 ) of experiments . we briefly discuss the elements of the representation theory of finite groups . our treatment is primarily for the purposes of setting down notation ; we refer the reader to @xcite for complete accounts . let @xmath0 be a finite group . a _ representation _ @xmath41 of @xmath0 is a homomorphism @xmath55 , where @xmath38 is a finite dimensional hilbert space and @xmath37 is the group of unitary operators on @xmath38 . the _ dimension _ of @xmath41 , denoted @xmath39 , is the dimension of the vector space @xmath38 . by choosing a basis for @xmath38 , then , each @xmath40 is associated with a unitary matrix @xmath56 $ ] so that for every @xmath57 , @xmath58 = [ \rho(g ) ] \cdot [ \rho(h)]$ ] where @xmath59 denotes matrix multiplication . fixing a representation @xmath55 , we say that a subspace @xmath60 is _ ( left)-invariant _ if @xmath61 for all @xmath62 ; observe that in this case the restriction @xmath63 , given by restricting each @xmath40 to @xmath64 , is also a representation . when @xmath41 has no invariant subspaces other than the trivial space @xmath65 and @xmath38 , @xmath41 is said to be _ irreducible_. when @xmath41 is irreducible , _ schur s lemma _ asserts that the centralizer of the subgroup @xmath66that is , the set of @xmath67 such that @xmath68 for all @xmath62consists only of the scalar matrices @xmath69 . we use this fact below . in the case when @xmath41 is _ not _ irreducible , then , there is a nontrivial invariant subspace @xmath70 and , as the inner product @xmath71 is invariant under the unitary maps @xmath40 , it is immediate that the dual subspace @xmath72 is also invariant . associated with the decomposition @xmath73 is the natural decomposition of the operators @xmath74 . by repeating this process , any representation @xmath75 may be decomposed into a direct sum of irreducible representations : we write @xmath76 and , for the @xmath77 appearing at least once in this decomposition , @xmath78 . in general , a given @xmath79 can appear multiply in this decomposition , in the sense that @xmath41 may have an invariant subspace isomorphic to the direct sum of @xmath80 copies of @xmath79 . in this case @xmath80 is called the _ multiplicity _ of @xmath77 in @xmath41 , and we write @xmath81 where @xmath82 . if two representations @xmath41 and @xmath79 are the same up to a unitary change of basis , we say that they are _ equivalent_. it is a fact that any finite group @xmath0 has a finite number of distinct irreducible representations up to equivalence and , for a group @xmath0 , we let @xmath83 denote a set of representations containing exactly one from each equivalence class . the irreducible representations of @xmath0 give rise to the fourier transform . specifically , for a function @xmath84 and an element @xmath85 , define the _ fourier transform of @xmath1 at @xmath41 _ to be @xmath86 the leading coefficients are chosen to make the transform unitary , so that it preserves inner products : @xmath87 there is a natural product operation on representations : if @xmath88 and @xmath89 are representations of @xmath0 , we may define a new representation @xmath90 by extending the rule @xmath91 . in general , the representation @xmath92 is not irreducible , even when both @xmath41 and @xmath79 are . this leads to the _ clebsch - gordan problem _ , that of decomposing @xmath92 into irreducibles . for a representation @xmath41 we define the _ character _ of @xmath41 , denoted @xmath93 , to be the function @xmath94 given by @xmath95 . as the trace of a linear operator is invariant under conjugation , characters are constant on the conjugacy classes of @xmath0 ; for a conjugacy class @xmath96 , we define @xmath97 . characters are a powerful tool for reasoning about the decomposition of reducible representations . in particular , when @xmath98 we have @xmath99 and , moreover , for @xmath100 , we have the orthogonality conditions @xmath101 then for an irreducible representation @xmath41 and representation @xmath79 , @xmath102 is equal to the multiplicity with which @xmath41 appears in the decomposition of @xmath79 . for example , since @xmath103 , the multiplicity of @xmath104 in @xmath92 is @xmath105 in general , we can consider subspaces of @xmath43}$ ] that are invariant under left multiplication , right multiplication , or both ; these subspaces are called _ left- _ , _ right- _ , or _ bi - invariant _ respectively . each @xmath85 corresponds to a @xmath106-dimensional bi - invariant subspace of @xmath43}$ ] , which can be broken up further into @xmath39 @xmath39-dimensional left - invariant subspaces , or ( transversely ) @xmath39 @xmath39-dimensional right - invariant subspaces . however , this decomposition is not unique . if @xmath41 acts on a vector space @xmath38 , then choosing an orthonormal basis for @xmath38 allows us to view @xmath40 as a @xmath107 matrix . then @xmath41 acts on the @xmath106-dimensional space of such matrices by left or right multiplication , and the columns and rows correspond to left- and right - invariant spaces respectively . more generally , each left - invariant subspace isomorphic to @xmath41 corresponds to a unit vector @xmath108 . in this section we show that starting with a random coset state , the optimal one - register measurement for the hidden subgroup problem is precisely an instance of strong fourier sampling ( possibly in an over - complete basis ) . this has been pointed out several times in the past , at varying levels of explicitness @xcite ; we state it here for completeness . everything we say in this section is true for the hidden subgroup problem in general . however , for simplicity we focus on the special case of the hidden subgroup problem called the _ hidden conjugate problem _ in @xcite : there is a ( non - normal ) subgroup @xmath3 , and we are promised that the hidden subgroup is one of its conjugates , @xmath109 for some @xmath62 . we may treat the states arising after step 3 of the procedure above as elements of the group algebra @xmath43}$ ] . we use the notation @xmath110}$ ] so that the vectors @xmath111 form an orthonormal basis for @xmath43}$ ] . given a set @xmath112 , @xmath113 denotes a uniform superposition over the elements of @xmath114 , @xmath115 . the most general type of measurement allowed in quantum mechanics is a _ positive operator - valued measurement _ ( povm ) . a povm with a set of possible outcomes @xmath116 consists of a set of positive operators @xmath117 subject to the completeness condition , @xmath118 since positive operators are self - adjoint , they can be orthogonally diagonalized , and since their eigenvalues are positive , they may be written as a positive linear combination of projection operators ( see e.g. @xcite ) . any povm may thus be refined so that each @xmath119 where @xmath120 is a projection operator and @xmath121 is positive and real . the result of this measurement on the state @xmath122 is a random variable , taking values in @xmath116 , that is equal to @xmath123 with probability @xmath124 . note that outcomes @xmath125 need not correspond to subgroups directly ; the algorithm designer is free to carry out a polynomial number @xmath126 of experiments , observing outcomes @xmath127 , and then apply some additional computation to find the most likely subgroup given these observations . if @xmath128 is chosen from @xmath0 uniformly so that the hidden subgroup is a uniformly random conjugate of @xmath3 , we wish to find a povm that maximizes the probability of correctly identifying @xmath128 from the coset state @xmath129 . ( of course , to identify a conjugate @xmath130 , we only need to specify @xmath128 up to an element of the normalizer of @xmath3 . ) since a random left coset of @xmath130 can be written @xmath131 for a random @xmath132 , the probability we observe outcome @xmath125 is @xmath133 ip @xcite observed that in the special case that each outcome @xmath125 corresponds to a subgroup , maximizing the probability that @xmath125 is correct subject to the constraint gives a semi - definite program . since such programs are convex , the optimum is unique and is a fixed point of any symmetries possessed by the problem . however , our proof relies on an elementary `` symmetrization '' argument . given a group element @xmath134 , let @xmath135 denote the unitary matrix corresponding to left group multiplication by @xmath7 . in particular , applying @xmath136 maps one left coset onto another : @xmath137 . writing @xmath138 we conclude that replacing @xmath120 for each @xmath125 with the symmetrization @xmath139 does not change the resulting probability distribution @xmath140 . since @xmath141 commutes with @xmath136 for every @xmath134 and provides exactly the same information as the original @xmath120 , we may assume without loss of generality that the optimal povm commutes with @xmath136 for every @xmath134 . it is easy to see that any projection operator that commutes with left multiplication projects onto a left - invariant subspace of @xmath43}$ ] , and we can further refine the povm so that each @xmath120 projects onto an _ irreducible _ left - invariant subspace . each such space is contained in the bi - invariant subspace corresponding to some irreducible representation @xmath41 , in which case we write @xmath142 . as discussed in section [ sec : representation - theory ] , a given irreducible left - invariant subspace corresponds to some unit vector @xmath143 in the vector space @xmath38 on which @xmath41 acts . thus we can write @xmath144 where @xmath145 acts within the left - invariant subspace . let @xmath146 ; then implies a completeness condition for each @xmath85 , @xmath147 and so @xmath148 is a ( possibly over - complete ) basis for @xmath38 . in other words , the optimal povm consists of first measuring the representation name @xmath41 , and then performing a povm on the vector space @xmath38 with possible outcomes @xmath148 . another way to see this is to regard the choice of coset as a mixed state ; then its density matrix is block - diagonal in the fourier basis , and so as kuperberg puts it @xcite measuring the representation name `` sacrifices no entropy . '' we note that in the special case that this povm is a von neumann measurement that is , when the @xmath148 is an orthonormal basis for @xmath38then it corresponds to measuring the column of @xmath41 in that basis , which is how strong fourier sampling is usually defined . ( as pointed out in @xcite , nothing is gained by measuring the row , since we have a random left coset @xmath149 and left - multiplying by a random element @xmath20 in an irreducible representation completely mixes the probability across the rows in each column . here this is reflected by the fact that @xmath120 is a scalar in each left - invariant subspace . ) however , in general the optimal measurement might consist of an over - complete basis , or _ frame _ , in each @xmath41 , consisting of the vectors @xmath150 and the weights @xmath121 . now that we know @xmath120 takes this form , let us change notation . given @xmath85 acting on a vector space @xmath38 and a unit vector @xmath108 , let @xmath151 denote the projection operator onto the left - invariant subspace corresponding to @xmath143 . then @xmath152 , and becomes @xmath153 we can write this as the product of the probability @xmath154 that we observe @xmath41 , times the conditional probability @xmath155 that we observe @xmath156 . note that by , @xmath157 is the projection operator onto the bi - invariant subspace corresponding to @xmath41 . then @xmath158 where @xmath159 note that @xmath155 depends on @xmath128 but @xmath154 does not , which is why weak sampling is incapable of distinguishing conjugate subgroups . now let us use the fact that @xmath160 is a superposition over a subgroup , and calculate @xmath154 and @xmath155 as defined in and . this will set the stage for asking whether we can distinguish different conjugates of @xmath3 from each other or even the trivial subgroup . fix an irreducible representation @xmath41 that acts on a vector space @xmath38 . then fourier transforming the state @xmath161 yields the coefficient @xmath162 where @xmath163 is a projection operator onto a subspace of @xmath38 . the probability that we observe @xmath41 is then the norm squared of this coefficient , @xmath164 and , as stated above , this is the same for all conjugates @xmath130 . the conditional probability that we observe the vector @xmath156 , given that we observe @xmath41 , is then @xmath165 in the case where @xmath3 is the trivial subgroup , @xmath166 and @xmath155 is given by @xmath167 we call this the _ natural distribution _ on the frame @xmath168 . in the case that @xmath148 is an orthonormal basis , @xmath169 and this is simply the uniform distribution . this probability distribution over @xmath148 changes for a conjugate @xmath130 in the following way . again ignoring left multiplication since the columns are left - invariant , the fourier transform becomes @xmath170 and we have @xmath171 where we write @xmath172 for @xmath173 . it is not hard to show that , for any @xmath108 , the _ expected _ value of @xmath174 , over the choice of @xmath62 , is @xmath175 . our primary technical contribution is a method for establishing concentration results for this random variable . in this section we focus on the case where @xmath53 for an element @xmath51 chosen uniformly at random from a conjugacy class @xmath176 $ ] of involutions . ( observe that order is preserved under conjugation so that if @xmath51 is an involution , then so are all elements of @xmath176 $ ] . ) given an irreducible representation @xmath55 and a vector @xmath108 , we bound the variance , over the choice of @xmath51 , of the probability @xmath177 that @xmath143 is observed given that we observed @xmath41 . our key insight is that this variance depends on how the tensor product representation @xmath178 decomposes into irreducible representations @xmath79 , and how the vector @xmath179 projects into the constituent orthogonal irreducibles . recall that , if a representation @xmath41 is reducible , it can be written as an orthogonal direct sum of irreducibles @xmath180 where @xmath80 is the multiplicity of @xmath79 . we let @xmath181 denote the projection operator whose image is @xmath182 , that is , the span of all the irreducible subspaces isomorphic to @xmath79 . [ lem : exp ] let @xmath41 be a representation of a group @xmath0 acting on a space @xmath38 and let @xmath108 . let @xmath51 be an element chosen uniformly from a conjugacy class @xmath176 $ ] of involutions . if @xmath41 is irreducible , then @xmath183)}{\dim \rho } { \left\| { { \mathbf{b}}}\right\|}^2 \enspace .\ ] ] if @xmath41 is reducible , then @xmath184)}{\dim \sigma } { \left\| \pi^\rho_\sigma { { \mathbf{b}}}\right\|}^2 \enspace .\ ] ] fix a particular element @xmath185 $ ] . let @xmath186)$ ] denote the average of @xmath187 over @xmath176 $ ] ; this is @xmath188 ) = \frac{1}{|[m]| } \sum_{m \in [ m ] } \rho(m ) = \frac{1}{|g| } \sum_{g \in g } \rho(g^{-1 } \mu g ) = \frac{1}{|g| } \sum_{g \in g } \rho(g)^\dagger \rho(\mu ) \,\rho(g ) \enspace .\ ] ] observe that @xmath186)$ ] commutes with @xmath40 for all @xmath62 and hence , by schur s lemma , its action on any irreducible subspace is multiplication by a scalar . note that for the scalar linear operator @xmath189 acting on a space of dimension @xmath190 , we have @xmath191 . in particular , if @xmath41 is irreducible then @xmath188 ) = \frac{\chi_\rho([m])}{\dim \rho } \,{\varmathbb 1}_{d_\rho}\ ] ] and so @xmath192 ) { { \mathbf{b}}}\rangle = \frac{\chi_\rho([m])}{\dim \rho } { \left\| { { \mathbf{b}}}\right\|}^2 \enspace .\ ] ] if @xmath41 is reducible , these same considerations apply to each irreducible subspace : @xmath188 ) = \sum_{\sigma \prec \rho } \frac{\chi_\sigma([m])}{\dim \sigma } \pi^\rho_\sigma\ ] ] and so @xmath192 ) { { \mathbf{b}}}\rangle = \sum_{\sigma \prec \rho } \frac{\chi_\sigma([m])}{\dim \sigma } \langle { { \mathbf{b } } } , \pi^\rho_\sigma { { \mathbf{b}}}\rangle = \sum_{\sigma \prec \rho } \frac{\chi_\sigma([m])}{\dim \sigma } { \left\| \pi^\rho_\sigma { { \mathbf{b}}}\right\|}^2 \enspace .\ ] ] turning now to the second moment of @xmath193 , we observe that @xmath194 where the action of @xmath51 on the vector @xmath179 is precisely given by the action of @xmath0 in the representations @xmath178 . this will allow us to express the second moment of the inner product @xmath195 in terms of the projections of @xmath179 into the irreducible constituents of the tensor product representation @xmath196 . [ lem : second ] let @xmath41 be a representation of a group @xmath0 acting on a space @xmath38 and let @xmath108 . let @xmath51 be an element chosen uniformly at random from a conjugacy class @xmath176 $ ] of involutions . then @xmath197)}{\dim \sigma } { \left\| \pi^{\rho \otimes \rho^*}_\sigma ( { { \mathbf{b}}}\otimes { { \mathbf{b}}}^ * ) \right\|}^2 \enspace .\ ] ] we write the second moment as a first moment over the product representation @xmath178 : as above , @xmath198 , so that @xmath199 and applying lemma [ lem : exp ] completes the proof . now let @xmath200 denote the projection operator given by @xmath201 for a given vector @xmath202 , we will focus on the expectation and variance of @xmath203 . these are given by the following lemma . [ lem : var ] let @xmath41 be an irreducible representation acting on a space @xmath38 and let @xmath108 . let @xmath51 be an element chosen uniformly at random from a conjugacy class @xmath176 $ ] of involutions . then @xmath204)}{\dim \rho } \right ) \label{eq : exp } \\ \operatorname{var}_{m } { \left\| \pi_m { { \mathbf{b}}}\right\|}^2 & \leq & \frac{1}{4 } \sum_{\sigma \prec \rho \otimes \rho^ * } \frac{\chi_\sigma([m])}{\dim \sigma } { \left\| \pi^{\rho \otimes \rho^*}_\sigma({{\mathbf{b}}}\otimes { { \mathbf{b}}}^ * ) \right\|}^2 \enspace . \label{eq : var } \end{aligned}\ ] ] for the expectation , @xmath205)}{\dim \rho } \right ) \end{aligned}\ ] ] where the last equality follows from lemma [ lem : exp ] . for the variance , we first calculate the second moment , @xmath206)}{\dim \rho } + \sum_{\sigma \prec \rho \otimes \rho^ * } \frac{\chi_\sigma([m])}{\dim \sigma } { \left\| \pi^{\rho \otimes \rho^*}_\sigma({{\mathbf{b}}}\otimes { { \mathbf{b}}}^ * ) \right\|}^2 \right)\end{aligned}\ ] ] where in the last line we applied lemmas [ lem : exp ] and [ lem : second ] and the fact that any character evaluated at an involution is real . then @xmath207)}{\dim \rho } { \left\| \pi^{\rho \otimes \rho^*}_\sigma ( { { \mathbf{b}}}\otimes { { \mathbf{b}}}^ * ) \right\|}^2 - { \left\| { { \mathbf{b}}}\right\|}^4 \left(\frac{\chi_\rho([m])}{\dim \rho } \right)^{\!2\ , } \right ] \label{eq : var1 } \enspace .\end{aligned}\ ] ] ignoring the second term , which is negative , gives the stated result . finally , we point out that since @xmath208 we have @xmath209)}{\dim \rho } \right ) \enspace , \ ] ] a fact which we will use below . in this section we record the particular properties of @xmath27 and its representation theory applied in the proofs of the main results . the irreducible representations of @xmath27 are labeled by young diagrams , or equivalently by integer partitions of @xmath10 , @xmath210 where @xmath211 and @xmath212 for all @xmath213 . the _ conjugate _ young diagram @xmath214 is obtained by flipping @xmath215 about the diagonal : @xmath216 where @xmath217 . in particular , @xmath218 . the number of such diagrams , equal to the number of conjugacy classes in @xmath27 , is the partition number @xmath219 , which obeys @xmath220 we denote these irreducibles @xmath221 , their characters @xmath222 , and their dimensions @xmath223 . in particular , @xmath224 is the trivial or parity representation if @xmath215 is a single row @xmath225 or a single column @xmath226 respectively , and @xmath227 is the ( tensor ) product of @xmath221 with the parity representation . the dimensions of the representations of the symmetric group are given by the remarkable _ hook length formula _ : @xmath228 where this product runs over all cells of the young diagram associated with @xmath215 and @xmath229 is the number of cells appearing in either the same column or row as @xmath20 , excluding those that are above or to the left of @xmath20 . for example , the partition @xmath230 is associated with the diagram shown in figure [ fig : young ] below . the hook associated with the cell @xmath231 in this diagram appears in figure [ fig : hook ] ; it has length 6 . a hook of length 6 . ] a hook of length 6 . ] the symmetric groups have the property that every representation @xmath221 possesses a basis in which its matrix elements are real , and so all its characters are real . however , in a given basis @xmath221 might be complex , so we will refer below to the complex conjugate representation @xmath232 ( not to be confused with @xmath227 ) . the study of the asymptotic properties of the representations of @xmath27 typically focuses on the _ plancherel _ distribution . ( see , e.g. , kerov s monograph @xcite for further discussion . ) for a general group @xmath0 , this is the probability distribution obtained on @xmath83 by assigning @xmath41 the probability density latexmath:[$d_\rho^2 / @xmath41 is proportional to its contribution , dimensionwise , to the group algebra @xmath43}$ ] ; this will allow us to reduce structural questions about representations chosen according to this distribution to questions about the regular and conjugation representation . note that in the context of the hidden subgroup problem , the plancherel distribution is exactly the one obtained by performing weak fourier sampling on the trivial hidden subgroup . in the symmetric groups a fair amount is known about representations chosen according to the plancherel distribution . in particular , vershik and kerov have shown that with high probability they have dimension equal to @xmath234 . [ thm : plancherel ] let @xmath221 be chosen from @xmath235 according to the plancherel distribution . then there exist positive constants @xmath236 and @xmath237 for which @xmath238 = 1\enspace.\ ] ] vershik and kerov have also obtained estimates for the _ maximum _ dimension of a representation in @xmath235 : [ thm : max - dimension ] there exist positive constants @xmath239 and @xmath240 such that for all @xmath241 , @xmath242 along with these estimates , we shall require some ( one - sided ) large - deviation versions of theorem [ thm : plancherel ] , recorded below . [ lem : lower ] let @xmath221 be chosen according to the plancherel distribution on @xmath235 . 1 . [ item : lower - sqrt ] let @xmath243 . then for sufficiently large @xmath10 , @xmath244 < e^{-\delta\sqrt{n}}$ ] . [ item : lower - linear ] let @xmath245 . then @xmath246 = n^{-\omega(n)}$ ] . for the first bound , setting @xmath247 and using @xmath248 , we have @xmath249 for the second bound , recalling sterling s approximation @xmath250 , we have @xmath251 finally , we will also apply roichman s @xcite estimates for the characters of the symmetric group : for a permutation @xmath252 , define the _ support _ of @xmath253 , denoted @xmath254 , to be the cardinality of the set @xmath255 \mid \pi(k ) \neq k \}$ ] . [ thm : roichman ] there exist constants @xmath256 and @xmath257 so that for @xmath258 , for every conjugacy class @xmath259 of @xmath27 , and every irreducible representation @xmath221 of @xmath27 , @xmath260 where @xmath261 for any @xmath262 . in our application , we take @xmath10 to be even and consider involutions @xmath51 in the conjugacy class of elements consisting of @xmath48 disjoint transpositions , @xmath263 . note that each @xmath264 is associated with one of the @xmath265 perfect matchings of @xmath10 things , and that @xmath266 . we consider the hidden subgroup @xmath267 , where @xmath51 is chosen uniformly from @xmath268 , the conjugacy class @xmath269 we assume throughout that @xmath10 is even . we start by measuring the name of an irreducible representation , yielding @xmath221 for a diagram @xmath215 . we remark that hallgren , russell , and ta - shma @xcite established that the probability distribution on @xmath215 is exponentially close to the plancherel distribution in total variation . we allow the algorithm designer to choose an arbitrary povm , with a frame @xmath270 and weights @xmath271 obeying the completeness condition . we will show that with high probability ( over @xmath51 and @xmath215 ) , the conditional distribution induced on the vectors @xmath148 is exponentially close to the natural distribution on @xmath148 . it will follow by the triangle inequality that it requires an exponential number of single - register experiments to distinguish two involutions from each other or , in fact , distinguish @xmath3 from the trivial subgroup . for simplicity , and to illustrate our techniques , we first prove this for a von neumann measurement , i.e. , where @xmath148 is an orthonormal basis for @xmath221 . in this case , we show that the probability distribution on @xmath148 ( or equivalently , on the columns of @xmath221 ) is exponentially close to the uniform distribution . [ thm : one ] let @xmath272 be an orthonormal basis for an irreducible representation @xmath221 . given the hidden subgroup @xmath53 where @xmath51 is chosen uniformly at random from @xmath273 , let @xmath274 be the probability that we observe the vector @xmath143 conditioned on having observed the representation name @xmath221 , and let @xmath275 be the uniform distribution on @xmath148 . then there is a constant @xmath276 such that for sufficiently large @xmath10 , with probability at least @xmath277 in @xmath51 and @xmath215 , we have @xmath278 first , recall from in section [ sec : optimal ] that the conditional distribution on @xmath148 is given by ( since @xmath169 ) @xmath279 our strategy will be to bound @xmath280 using lemma [ lem : var ] , and apply chebyshev s inequality to conclude that it is almost certainly close to its expectation . recall , however , that our bounds on the variance of @xmath203 depend on the decomposition of @xmath281 is into irreducibles and , furthermore , on the projection of @xmath179 into these irreducible subspaces . matters are somewhat complicated by the fact that certain @xmath282 appearing in @xmath281 may contribute more to the variance than others . while theorem [ thm : roichman ] allows us to bound the contribution of those constituent irreducible representations @xmath282 for which @xmath283 and @xmath284 are much smaller than @xmath10 , those which violate this condition could conceivably contribute large terms to the variance estimates . fortunately , we will see that the total fraction of the space @xmath281 , dimensionwise , consisting of such @xmath282 is small with overwhelming probability . despite this , we can not preclude the possibility that for a _ specific _ vector @xmath143 , the quantity @xmath285 is large , as @xmath143 may project solely into spaces of the type described above . on the other hand , as these troublesome spaces amount to a small fraction of @xmath281 , only a few @xmath143 can have this property , and we will see that this suffices to control the distance in total variation to the uniform distribution . specifically , let @xmath286 be a constant , and let @xmath287 denote the collection of young diagrams @xmath288 with the property that either @xmath289 or @xmath290 . we have the following upper bounds on the cardinality of @xmath291 and the dimension of any @xmath282 with @xmath292 : [ lem : lambda ] let @xmath219 denote the number of integer partitions of @xmath10 . then @xmath293 , and @xmath294 for any @xmath292 . for the first statement , note that removing the top row of a young diagram @xmath288 with @xmath295 gives a young diagram of size @xmath296 . the number of these is at most @xmath297 , and summing over all such @xmath283 gives @xmath298 . the case @xmath299 is similar , and summing the two gives @xmath293 . now let @xmath292 with @xmath295 . by the hook - length formula , since the @xmath213th cell from the right in the top row has @xmath300 , @xmath301 . the case @xmath299 is similar . as a result , the representations associated with diagrams in @xmath291 constitute a negligible fraction of @xmath235 ; specifically , from lemma [ lem : lower ] , part [ item : lower - linear ] , the probability that a @xmath215 drawn according to the plancherel distribution falls into @xmath291 is @xmath302 . the following lemma shows that this is also true for the distribution @xmath154 induced on @xmath235 by weak fourier sampling the coset state @xmath160 . [ lem : notinlambda ] let @xmath303 be a constant and let @xmath10 be sufficiently large . then there is a constant @xmath304 such that we observe a representation @xmath221 with @xmath305 with probability at least @xmath306 . the proof is nearly identical to that of lemma [ lem : lower ] , and follows an argument from @xcite . recall from in section [ sec : optimal ] that we observe an irreducible @xmath41 with probability @xmath307 . given @xmath308 , @xmath309 , and @xmath310 , we have @xmath311 . since the total number of irreducibles of @xmath27 is @xmath219 and @xmath312 , we have @xmath313 and setting @xmath314 completes the proof . on the other hand , for a representation @xmath282 with @xmath315 , theorem [ thm : roichman ] implies that @xmath316 for a constant @xmath317 . thus the contribution of such an irreducible to the variance estimate of lemma [ lem : var ] is exponentially small . in addition , let @xmath318 be the event that @xmath319 as in lemma [ lem : notinlambda ] ; then conditioning on @xmath318 and setting @xmath190 such that @xmath320 , lemma [ lem : lambda ] implies that @xmath321 , and equations and imply @xmath322 we turn now to the problem of bounding the multiplicities with which representations @xmath282 , for @xmath292 , can appear in @xmath323 . while no explicit decomposition is known for @xmath323 , the _ endomorphism representations _ of @xmath27 , we record a coarse bound below which will suffice for our purposes . recall that character of @xmath281 is @xmath324 as characters of @xmath27 are real . the multiplicity of the representation @xmath282 in @xmath281 is @xmath325 . however , this is equal to @xmath326 , the multiplicity of @xmath221 in the representation @xmath327 . counting dimensions , this is clearly no more than @xmath328 . hence the multiplicity of @xmath282 in @xmath281 is never more than @xmath329 ; we have @xmath330 let @xmath331 be the subspace consisting of copies of representations @xmath282 with @xmath292 , and let @xmath332 be the projection operator onto this subspace . by lemma [ lem : lambda ] , we have @xmath333 note by lemma [ lem : notinlambda ] with @xmath334 , @xmath335 is a vanishingly small fraction of @xmath223 . as @xmath148 is an orthonormal basis for @xmath221 , the vectors @xmath336 are mutually orthogonal in @xmath281 therefore , @xmath337 in particular , if @xmath338 denotes the set of @xmath143 with the property that @xmath339 then , conditioning on @xmath318 , @xmath340 can not be larger than @xmath341 . for a basis vector @xmath342 , we apply lemma [ lem : var ] to bound the variance as follows : assuming pessimistically that @xmath343 for all @xmath292 , and applying , gives @xmath344 \\ & \leq & \frac{1}{4 } \left [ e^{-\alpha n } + e^{-\alpha n } \sum_{s^\mu : \mu \notin \lambda } { \left\| \pi^\lambda_\mu({{\mathbf{b}}}\otimes { { \mathbf{b}}}^ * ) \right\|}^2 \right ] \;\ ; \leq \;\ ; \frac{1}{2 } \,e^{-\alpha n } \enspace . \end{aligned}\ ] ] in this case , by chebyshev s inequality , @xmath345 \leq e^{-\alpha n/3 } \enspace .\ ] ] we say that a basis vector is _ bad _ if this bound is violated , i.e. , @xmath346 let @xmath347 denote the subset of @xmath148 consisting of bad basis vectors ( observe that while @xmath348 depends only on the choice of @xmath215 , @xmath347 depends also on @xmath51 ) . then implies @xmath349 now let @xmath350 be the event that @xmath351 then by markov s inequality , @xmath350 occurs with probability at least @xmath352 . now let us separate @xmath353 into contributions from basis vectors outside and inside @xmath354 : @xmath355 the first sum is taken only over vectors @xmath143 for which @xmath356 conditioning on @xmath318 and recalling that @xmath357 , the rank estimate of gives @xmath358 it follows that @xmath359 is at most @xmath360 . therefore , since conditioning on @xmath318 and @xmath350 we have @xmath361 , the second sum in is at most @xmath362 then combining , and , @xmath363 with probability at least @xmath364 \geq 1 - n^{-\gamma n } - e^{-\alpha n/6 } \geq 1 - 2e^{-\alpha n/6}$ ] . we complete the proof by setting @xmath365 . we now generalize the proof of theorem [ thm : one ] to the case where the algorithm designer is allowed to choose an arbitrary finite frame @xmath366 of unit length vectors in @xmath221 , with a family of positive real weights @xmath367 , that satisfy the completeness condition @xmath368 ( note that this is simply where we have written @xmath143 and @xmath367 instead of @xmath156 and @xmath121 . ) [ thm : one - frames ] let @xmath272 be a frame with weights @xmath369 satisfying the completeness condition for an irreducible representation @xmath221 . given the hidden subgroup @xmath53 where @xmath51 is chosen uniformly at random from @xmath273 , let @xmath274 be the probability that we observe the vector @xmath143 conditioned on having observed the representation name @xmath221 , and let @xmath370 be the natural distribution on @xmath148 . then there is a constant @xmath276 such that for sufficiently large @xmath10 , with probability at least @xmath277 in @xmath51 and @xmath215 , we have @xmath371 recall from in section [ sec : optimal ] that the conditional distribution on @xmath148 is given by @xmath372 and the natural distribution is given by @xmath373 . the proof of theorem [ thm : one ] goes through with a few modifications . first , let us change some semantics : given a subset @xmath374 , we let @xmath375 denote the weighted size of @xmath376 , @xmath377 with this definition , the total probability that falls in @xmath376 under the natural distribution is @xmath378 . then we will use the following lemma : [ lem : framesum ] let @xmath379 be a subspace of @xmath281 and let @xmath332 be the projection operator onto @xmath379 . then @xmath380 first note that a vector @xmath381 has entries @xmath382 for @xmath383 . there is a unique linear operator @xmath384 on @xmath221 whose matrix entries are @xmath385 , and the inner product @xmath386 in @xmath281 can then be written as the bilinear form @xmath387 in @xmath221 . the frobenius norm of @xmath384 is @xmath388 . now let @xmath389 be an orthonormal basis for @xmath379 and let @xmath390 be the operator corresponding to @xmath391 . then @xmath392 = { \textbf{tr}\,}e_i^\dagger e_i = { \left\| { { \mathbf{e}}}_i \right\|}^2 = 1 \end{aligned}\ ] ] where we used the cauchy - schwartz inequality in the second line and completeness in the second . summing over the @xmath335 basis vectors @xmath391 then gives . we define @xmath291 and @xmath318 as before , and lemmas [ lem : lambda ] and [ lem : notinlambda ] still apply . as before , let @xmath331 be the subspace consisting of copies of representations @xmath282 with @xmath292 , and let @xmath338 denote the set of @xmath143 with the property that @xmath393 then lemma [ lem : framesum ] implies that @xmath394 where @xmath340 is defined as above . we again define @xmath347 as the set of @xmath395 such that @xmath396 and chebyshev s and markov s inequalities imply that the event @xmath350 , namely @xmath397 occurs with probability at least @xmath352 . we separate @xmath398 as we did @xmath353 before : @xmath399 since @xmath400 for all @xmath401 , and since @xmath402 , conditioning on @xmath318 and using bounds the first sum as follows , @xmath403 it follows that @xmath359 is at most @xmath404 . conditioning on @xmath318 and @xmath350 , we have @xmath405 . thus the second sum in is at most @xmath406 then combining , and , @xmath407 with probability at least @xmath364 \geq 1 - n^{-\gamma n } - e^{-\alpha n/6 } \geq 1 - 2e^{-\alpha n/6}$ ] . we complete the proof by setting @xmath365 as before . the preceding development focuses on the case where the hidden subgroup is distributed uniformly among the conjugates of the subgroup @xmath408 . as such , this shows that the canonical reduction of graph automorphism ( the problem of determining whether a given graph has a non - trivial automorphism ) to the hidden subgroup problem does not , in general , give rise to an efficient quantum algorithm via fourier sampling . however , the canonical reduction of graph isomorphism to the hidden subgroup problem induces a more structured set of involutions . as referred to in the introduction , fixing two rigid graphs @xmath409 and @xmath410 , each with @xmath10 vertices , the automorphism group of their disjoint union @xmath411 is nontrivial exactly when they are isomorphic , in which case it is generated by an involution @xmath51 with full support such that @xmath412 . identifying @xmath413 and @xmath414 with the sets @xmath415 and @xmath416 respectively , and letting @xmath417 denote the involution @xmath418 , the standard reduction to the hidden subgroup problem in @xmath419 then results in a hidden subgroup @xmath420 where @xmath51 is a conjugate involution @xmath421 . however , rather than @xmath422 being drawn from all of @xmath419 , it is an element of the young subgroup @xmath423 which fixes @xmath413 and @xmath414 : @xmath424 in other words , rather than considering all conjugates of @xmath51 in @xmath419 , it suffices just to consider conjugates in @xmath423 . _ a priori _ , it seems that this smaller set of possible hidden subgroups might be easier to identify . moreover , let @xmath30 be the subgroup generated by @xmath423 and @xmath417 : this is the wreath product @xmath425 , which can also be written as a semidirect product @xmath426 . then each such @xmath3 is contained in @xmath30 , and it seems that it might be more intelligent to fourier sample over @xmath30 rather than over all of @xmath419 . however , we can show that nothing is gained by this approach . first , note that the involutions described above form the ( @xmath30-)conjugacy class @xmath427 we remark the that the development of section [ sec : optimal ] is unchanged , and that the optimal measurement to find a hidden conjugate again consists of strong fourier sampling . now note that fourier sampling over @xmath419 and over @xmath30 are equivalent for the following reason : suppose we are trying to distinguish a set of hidden subgroups @xmath428 , all of which are contained in a subgroup @xmath429 . let @xmath430 be a set of representatives for the cosets of @xmath30 . then a random left coset of @xmath431 in @xmath0 is the product of a random left coset of @xmath431 in @xmath30 with a random element of @xmath430 . thus the mixed state describing a uniformly random coset of @xmath431 in @xmath0 can be written as the corresponding state over @xmath30 with the completely mixed state over @xmath430 . since this completely mixed state ( whose density matrix is the identity ) contains no information , nothing is gained ( or lost ) by sampling over all of @xmath0 rather than over @xmath30 . finally , we observe that one can obtain easy character bounds for this conjugacy class of @xmath30 . one determines @xmath30 s irreducible representations and their characters as follows . for two irreducible representations @xmath41 and @xmath79 of @xmath27 , let @xmath432 denote their tensor product as a representation of @xmath433 . we consider the _ induced representation _ @xmath434 ( see @xcite ) and denote its character @xmath435 . it is easy to see that @xmath436 as the notation suggests , this depends only on the multiset @xmath437 . observe that if @xmath376 and @xmath148 are two distinct multisets of representations of @xmath27 ( each of cardinality two ) then the characters @xmath438 and @xmath439 are orthogonal . furthermore , an easy computation shows that @xmath440 and hence that @xmath441 is irreducible if @xmath442 , while if @xmath443 then the induced representation is the direct sum of two irreducible representations , @xmath444 where @xmath445 and @xmath253 are the trivial and sign representations , respectively , of @xmath446 . each of these irreducible representations acts on @xmath447 , the vector space supporting the action of @xmath448 . both of them realize the element @xmath449 as the linear map @xmath450 , while @xmath451 and @xmath452 realize the element @xmath453 as the maps which send @xmath454 to @xmath455 and @xmath456 respectively . the characters of these representations can be determined by inspection ; in particular , @xmath457 where @xmath458 and @xmath459 denote the characters of @xmath451 and @xmath452 respectively . in light of this description of the irreducible characters of @xmath30 , one finds that @xmath460 when @xmath461 and that @xmath462 for @xmath463 . in particular , the absolute value of the normalized character at @xmath464 is @xmath465 . as before , the dimensions @xmath39 of the diagrams with large normalized character sum to very little , and the analysis of section [ sec : single ] can be undertaken _ mutatis mutandis_. remarkably , this can be carried out without use of roichman s estimates ! this work was supported by nsf grants ccr-0093065 , phy-0200909 , eia-0218443 , eia-0218563 , ccr-0220070 , and ccr-0220264 , and aro grant w911nf-04-r-0009 . we are grateful to denis thrien , mcgill university , and bellairs research institute for organizing a workshop at which this work began ; to dorit aharonov , daniel rockmore , and umesh vazirani for helpful conversations ; to chris lomont for pointing out several typos ; and to tracy conrad and sally milius for their support and tolerance . also thanks rosemary moore for providing a larger perspective . finally , we thank gorjan alagic for his comments on the structured involutions material . michelangelo grigni , leonard j. schulman , monica vazirani , and umesh vazirani . quantum mechanical algorithms for the nonabelian hidden subgroup problem . 33rd acm symposium on theory of computing _ , pages 6874 , 2001 . sean hallgren , alexander russell , and amnon ta - shma . normal subgroup reconstruction and quantum computation using group representations . 32nd acm symposium on theory of computing _ , pages 627635 , 2000 . _ asymptotic representation theory of the symmetric group and its applications in analysis_. translated by n. v. tsilevich . volume 219 in translations of mathematical monographs . american mathematical society , 2003 . cristopher moore , daniel rockmore , alexander russell , and leonard schulman . the value of basis selection in fourier sampling : hidden subgroup problems for affine groups . _ proc . 15th annual acm - siam symposium on discrete algorithms _ , pages 11131122 , 2004 . vershik and s. v. kerov . asymptotic behavior of the maximum and generic dimensions of irreducible representations of the symmetric group . funk . i prolizhen , 19(1):2536 , 1985 ; english translation , funct . appl . , 19:2131 , 1989 .

for a ( single - source ) multicast network , the size of a base field is the most known and studied algebraic identity that is involved in characterizing its linear solvability over the base field . in this paper , we design a new class of multicast networks and obtain an explicit formula for the linear solvability of these networks , which involves the associated coset numbers of a multiplicative subgroup in a base field . the concise formula turns out to be the first that matches the topological structure of a multicast network and algebraic identities of a field other than size . it further facilitates us to unveil _ infinitely many _ new multicast networks linearly solvable over gf( ) but not over gf( ) with , based on a subgroup order criterion . in particular , i ) for every , an instance in can be found linearly solvable over gf( ) but _ not _ over gf( ) , and ii ) for arbitrary distinct primes and , there are infinitely many and such that an instance in can be found linearly solvable over gf( ) but _ not _ over gf( ) with . on the other hand , the construction of also leads to a new class of multicast networks with nodes and edges , where is the minimum field size for linear solvability of the network . network coding , multicast , linear solution , multiplicative subgroup , coset , generalized cauchy - davenport theorem .

1510.02305

consider a communication network , which is a finite directed acyclic multigraph with each edge representing a noiseless transmission channel of unit capacity . there is a set of source nodes , each of which independently generates a set of data symbols belonging to a base field and transmits the data symbols simultaneously along the network . there is another set of receiver nodes , each of which attempts to recover a certain subset of source data symbols . a network is _ linearly solvable _ over a base field gf(@xmath1 ) if there is a linear network coding scheme , which encodes every outgoing symbol of a node as a gf(@xmath1)-linear combination of the incoming data symbols to this node , so that every receiver can recover its desired source data symbols at the same time . the algebraic structure of a base field is closely related to the linear solvability of a network . when a network is ( single - source ) multicast , _ i.e. _ , every receiver attempts to recover the same set of data symbols transmitted from a unique source , the fundamental theorem of linear network coding @xcite guarantees the existence of a linear solution when the base field is sufficiently large . since the connection of field size on the linear solvability of a multicast network was revealed , there have been extensive studies on the field size requirement and efficient construction of a linear solution [ 2 - 12 ] . among them , the best known explicit sufficient condition for a multicast network to be linearly solvable over gf(@xmath1 ) is @xmath16 , where @xmath17 is the number of receivers @xcite@xcite . moreover , the combination networks @xcite@xcite constitute the only investigated class of multicast networks with an explicit linear solvability characterization that matches algebraic identities of gf(@xmath1 ) and topological parameters of the network . specifically , an @xmath18-combination network , as depicted in fig . [ fig : combination_network ] , is known to be linearly solvable over gf(@xmath1 ) if and only if @xmath19 , where @xmath20 is the topological parameter for the number of layer-3 nodes . without the restriction of considering a multicast network , more algebraic identities inherited in a field other than size have been revealed to affect the linear solvability of a network . in @xcite , two networks were designed , one of which is linearly solvable only over a field with even characteristic , while the other is linearly solvable only over a field with odd characteristic . it is further unveiled in @xcite that for an arbitrary finite or co - finite set @xmath21 of prime numbers , a network can be constructed such that the network is linearly solvable over some field of characteristic @xmath7 for every @xmath7 in @xmath21 , whereas it is not linearly solvable over any field whose characteristic is not in @xmath21 . meanwhile , whether a set of polynomials with integer coefficients has common roots is connected to a network linear solvability problem . specifically , a general method was proposed in @xcite to associate a polynomial set with a network , such that the polynomials have common roots over a field if and only if the corresponding network has a linear solution over the same field . in @xcite , analogous equivalence was also established between the existence of common roots over a finite field for a polynomial set and the existence of a linear solution over the same field for an associated _ sum - network _ , in which all receivers attempt to recover the sum of all source data symbols . however , when attention is only paid to multicast networks , these results are not applicable anymore . first , since a multicast network is linearly solvable over a large enough field , it is linearly solvable over some field of characteristic @xmath7 for every prime @xmath7 . moreover , the general method proposed in @xcite and @xcite can not necessarily construct a multicast network from a polynomial set . if the field size were the only algebraic factor that affects the linear solvability of a multicast network , then all multicast networks would conceivably share a property @xmath22 , where @xmath23 refers to the minimum field size for the existence of a linear solution and @xmath24 refers to the maximum field size for the non - existence of a linear solution ( @xmath24 is set to 1 if the network is linearly solvable over every field . ) the first few exemplifying networks with the special property @xmath25 were not discovered until a rather recent work @xcite . as pointed out in @xcite , there are fundamental reasons underlying the intriguing observation that @xmath25 is possible for multicast networks : not only the field size @xmath1 , but also _ the order of proper subgroups in the multiplicative group _ gf(@xmath1)@xmath26 = gf(@xmath1)@xmath27 of gf(@xmath1 ) plays an important role in the linear solvability over gf(@xmath1 ) . however , even though it was further observed in @xcite that all these newly designed networks with @xmath25 share a common topological structure , the intrinsic impact of multiplicative subgroup orders of a field on the linear solvability of multicast networks remains unclear . in this paper , we explicate in depth how multiplicative subgroups in gf(@xmath1 ) can affect the linear solvability of multicast networks over gf(@xmath1 ) , as outlined as follows . * in sec . [ sec : general_framework ] , we construct a general class @xmath0 of layered multicast networks , which subsumes all multicast networks presented in @xcite with @xmath28 as special instances . after deriving a generalized cauchy - davenport theorem , we proceed to construct an explicit formula for the linear solvability of networks in @xmath0 over gf(@xmath1 ) . besides the topological parameters of @xmath0 , the concise formula involves the associated coset numbers of a multiplicative subgroup in gf(@xmath1 ) . it turns out to be the first that matches the topological structure of a multicast network and algebraic identities of a field other than size . * in sec . [ sec : subgroup_order_criterion ] , based on the general characterization of @xmath0 , we further formulate a subgroup order criterion for a pair of finite fields . as long as ( gf(@xmath1 ) , gf(@xmath2 ) ) satisfies the subgroup order criterion , where @xmath3 is possible , we can establish an instance in @xmath0 that is linearly solvable over gf(@xmath1 ) but not over gf(@xmath2 ) . [ sec : establish_instances ] , as an application of the subgroup order criterion , we are able to establish _ infinitely many _ multicast networks in @xmath0 with @xmath28 . as intriguing instances , for every @xmath29 , we can establish a multicast network with @xmath30 and @xmath31 for some @xmath32 . this proves that @xmath33 can tend to infinity for a multicast network . moreover , for _ arbitrary _ distinct primes @xmath7 and @xmath8 , there are _ infinitely many _ @xmath9 and @xmath10 such that an instance in @xmath0 can be found to be linearly solvable over gf(@xmath11 ) but _ not _ over gf(@xmath12 ) with @xmath13 . [ sec : construction_network_prescribed_q_min ] , based on the general characterization of @xmath0 , a new procedure is proposed to construct a multicast network with @xmath23 equal to any prescribed prime power @xmath15 , and the constructed network has a smaller size compared with the @xmath34-combination network , which has @xmath35 too . to summarize , we systematically develop a framework to reflect the intrinsic impact of multiplicative subgroups on the linear solvability of a multicast network . a number of new results on the comparison between @xmath23 and @xmath24 are also subsequently deduced . our findings suggest that a `` matching '' between the algebraic structure of a base field and the topological structure of a multicast network is necessary for the existence of a linear solution to the multicast network . a _ multicast network _ is modeled as a finite directed acyclic multigraph with a unique source node @xmath36 and a set @xmath37 of receivers . the out - degree of @xmath36 is denoted by @xmath38 and is assumed equal to the source dimension , which means the number of data symbols to be simultaneously transmitted by the source . for a node @xmath39 in the network , denote by @xmath40 the set of its incoming edges . for an arbitrary set @xmath41 of non - source nodes , denote by @xmath42 the maximum number of edge - disjoint paths starting from @xmath36 and ending at nodes in @xmath41 . every edge in the network is of unit capacity . a linear network code ( lnc ) over gf(@xmath1 ) is an assignment of a coding coefficient @xmath43 to every pair @xmath44 of edges such that @xmath45 when @xmath44 is not an adjacent pair . every lnc uniquely determines a coding vector @xmath46 , which is an @xmath38-dimensional column vector over gf(@xmath1 ) , for each edge @xmath47 in the network . a multicast network is linearly solvable over gf(@xmath1 ) if there is an lnc over gf(@xmath1 ) such that for each receiver @xmath48 , the @xmath49 matrix @xmath50_{e\in in(t)}$ ] over gf(@xmath1 ) is of full rank @xmath38 . such an lnc is called a linear solution over gf(@xmath1 ) for the multicast network . let @xmath23 be the minimum field size for the existence of a linear solution over gf(@xmath23 ) , and @xmath24 the maximum field size for the nonexistence of a linear solution over gf(@xmath24 ) . we set @xmath24 to @xmath51 if the network is linearly solvable over all finite fields . fig . [ fig : rank_3_networks ] reproduces two of the multicast networks discovered in @xcite , with @xmath28 . in both networks , the source dimension @xmath38 is equal to @xmath52 and there is a non - depicted receiver connected from every set @xmath41 of three grey nodes whenever the maximum flow from the source to the nodes in @xmath41 is @xmath52 . for example , there is a receiver connected from @xmath53 in fig . [ fig : rank_3_networks](a ) and from @xmath54 in fig . [ fig : rank_3_networks](b ) . it can be shown that @xmath55 , @xmath56 for the network in fig . [ fig : rank_3_networks](a ) and @xmath57 , @xmath58 for the network in fig . [ fig : rank_3_networks](b ) . in the course of characterizing the linear solvability of these networks , it has already been noted in @xcite that these exemplifying networks share a common topological structure , and a general 5-layer multicast network is correspondingly constructed and analyzed for unifying the proof . in this paper , we construct a more general class of multicast networks in the sense of involving more topological parameters as follows . [ alg : general_construction ] given a positive integer @xmath59 and an @xmath38-tuple @xmath60 of positive integers larger than 1 as input parameters , the procedure below constructs a multicast network @xmath61 with source dimension @xmath38 consisting of nodes on five layers , which are labeled 1 - 5 from upstream to downstream , and all edges are between adjacent layers . . create a source @xmath36 , which forms the unique node at layer 1 . . create @xmath38 layer-2 nodes , each of which is connected with @xmath36 by a single edge . sequentially label these nodes as @xmath62 . . create @xmath38 layer-3 nodes , labeled as @xmath63 . each node @xmath64 , @xmath65 , has 2 incoming edges , one leading from @xmath66 and the other from @xmath67 , where @xmath68 will represent @xmath69 . . for each layer-3 node @xmath64 , @xmath65 , create @xmath70 downstream layer-4 nodes @xmath71 , each of which has the unique incoming edge @xmath72 leading from @xmath73 . . for every set @xmath41 of @xmath38 layer-4 nodes with @xmath74 , create a layer-5 node connected from every node in @xmath41 by an edge . set all layer-5 nodes to be receivers . @xmath75 fig . [ fig : general_network ] depicts the topology of a general @xmath76 . the network in fig . [ fig : rank_3_networks ] can be constructed by algorithm [ alg : general_construction ] with respective parameters @xmath77 and @xmath78 . moreover , the general network considered in @xcite with parameters @xmath79 can also be regarded as a particular instance of @xmath61 with @xmath80 . the following equivalent conditions for the linear solvability of network @xmath61 can be readily derived . [ lemma : general_network_linear_solvability ] consider a network @xmath61 constructed by algorithm [ alg : general_construction ] . the followings are equivalent . a. the network is linearly solvable over a finite field gf(@xmath1 ) . b. there is a matrix completion of the matrix @xmath81 depicted in ( [ eqn : general_matrix_m_for_linear_solvability_analysis ] ) over gf(@xmath1 ) , that is , an assignment of values @xmath82 in gf(@xmath1 ) to all indeterminates @xmath83 such that the rank of every @xmath84 submatrix in @xmath81 is preserved . + @xmath85\ ] ] c. there exists a set @xmath86 of distinct nonzero elements in gf(@xmath1 ) for each @xmath65 such that @xmath87 moreover , when condition c ) holds for some @xmath88 , the assignment of @xmath89 is a matrix completion of matrix @xmath81 , and a linear solution is given by @xmath81 which represents the juxtaposed matrix of coding vectors for edges into layer-4 ( grey ) nodes . please refer to appendix[appendix_general_linear_solvability ] . the equivalent conditions in lemma [ lemma : general_network_linear_solvability ] implicitly match the topological parameters @xmath38 and @xmath90 in @xmath76 with gf(@xmath1 ) , while the inherent algebraic identity that affects the linear solvability of @xmath61 remains unveiled . in the remaining part of this section , we shall proceed to derive an explicit equivalent condition that matches the parameters @xmath38 and @xmath90 with not only the size , but also the multiplicative subgroup orders of gf(@xmath1 ) . recall that all nonzero elements in gf(@xmath1 ) can be represented as @xmath91 , @xmath92 for some element @xmath93 . such an element @xmath93 is called a _ primitive element _ in gf(@xmath1 ) . thus , @xmath94 . since @xmath95 , the set @xmath96 forms a cyclic group with generator @xmath93 and of order @xmath97 . this cyclic group is isomorphic to the additive group @xmath98 of integers modulo @xmath99 via the mapping of @xmath100 . when the field gf(@xmath1 ) has odd characteristic , @xmath101 . else , @xmath102 . thus , the network @xmath76 has a linear solution over gf(@xmath1 ) if and only if * there exist subsets @xmath103 of @xmath98 with respective cardinalities @xmath104 , such that the set @xmath105 does not include @xmath106 when @xmath1 is odd and does not include @xmath107 when @xmath1 is even . with a parallel shift on all members of @xmath108 if necessary , we have the following equivalent condition . [ prop : linear_solvability_exhaust_claim ] the network @xmath76 has a linear solution over gf(@xmath1 ) if and only if there exist subsets @xmath103 of additive group @xmath98 with respective cardinalities @xmath104 such that the set @xmath109 does not exhaust @xmath98 , that is , @xmath110 consider the network @xmath76 with parameters @xmath111 . if there is a positive integer @xmath112 that divides @xmath99 , then there is a subgroup @xmath113 of order @xmath114 in @xmath98 . if @xmath115 , then we can assign each @xmath116 , @xmath117 , to be arbitrary @xmath118 elements in @xmath113 and @xmath119 to be arbitrary @xmath120 elements in @xmath121 , _ i.e. _ , the complement of @xmath113 in @xmath98 . thus , @xmath122 and @xmath123 . this implies that the set @xmath124 does not exhaust the whole @xmath98 and hence the network is linearly solvable over gf(@xmath1 ) . this sufficient condition in terms of subgroup orders was applied in @xcite for the purpose of proving the existence of linear solutions over gf(@xmath1 ) . for instance , for the network @xmath76 with @xmath125 as depicted in fig . [ fig : rank_3_networks](b ) , it is linearly solvable over gf(@xmath126 ) since we can set @xmath127 and @xmath128 in @xmath129 such that @xmath130 does not exhaust @xmath129 . however , no handy necessary condition has been derived in @xcite for the nonexistence of linear solutions , and this is a key reason that only a few exemplifying networks with the special property @xmath28 were designed therein . for instance , the nonexistence of a linear solution over gf(17 ) for @xmath76 with @xmath125 is merely verified in @xcite by exhaustive enumeration on all possible subsets of nonzero elements in gf(@xmath131 ) for condition ( [ eqn : network_linear_solvability_rule ] ) in lemma [ lemma : general_network_linear_solvability ] . recall that when @xmath99 is prime , cauchy - davenport theorem asserts that for any two non - empty subsets @xmath132 and @xmath133 of @xmath98 , @xmath134 . thus , for arbitrary three - element subsets @xmath135 of @xmath136 , @xmath137 , that is , @xmath135 exhaust the whole @xmath136 . consequently , by applying proposition [ prop : linear_solvability_exhaust_claim ] to the network @xmath138 with @xmath139 , we conclude that the network depicted in fig . [ fig : rank_3_networks](a ) is not linearly solvable over gf(@xmath140 ) . we next generalize the cauchy - davenport theorem to be over @xmath98 , where @xmath99 can be a composite . this requires the concept of stabilizer group ( see , e.g. , @xcite ) . recall that the _ stabilizer group _ of a subset @xmath132 in a group @xmath141 is defined as @xmath142 . [ thm : cauchy - davenport - generalization](generalized cauchy - davenport theorem over @xmath143 ) let @xmath144 be nonempty subsets of @xmath145 and @xmath113 the stabilizer group of @xmath146 in @xmath145 . then @xmath147 . it can be checked that when @xmath148 and @xmath20 is a prime , equation ( [ eqn : cauchy - davenport - generalization ] ) in theorem [ thm : cauchy - davenport - generalization ] degenerates to the cauchy - davenport theorem . based on theorem [ thm : cauchy - davenport - generalization ] , we are able to further characterize the linear solvability of the network @xmath76 from the perspective of subgroup orders . [ thm : concise_linear_solvability_characterization ] consider a network @xmath61 constructed by algorithm [ alg : general_construction ] with parameters @xmath38 and @xmath149 . it is linearly solvable over gf(@xmath1 ) if and only if there is positive divisor @xmath114 of @xmath97 subject to @xmath150 or equivalently , @xmath151 as a consequence of proposition [ prop : linear_solvability_exhaust_claim ] , it is equivalent to show that there exist subsets @xmath103 of @xmath98 with respective cardinalities @xmath104 such that the set @xmath109 does not exhausting @xmath98 if and only if there is a divisor @xmath152 of @xmath97 subject to ( [ eqn : equivalence_subgroup_order_perspective ] ) . for the necessity part , let @xmath103 be subsets of @xmath98 with respective cardinalities @xmath104 subject to @xmath153 . according to theorem [ thm : cauchy - davenport - generalization ] , @xmath154 this implies that there exists at least one divisor @xmath152 of @xmath97 to make inequality ( [ eqn : equivalence_subgroup_order_perspective ] ) hold . for the sufficiency part , let @xmath152 be an arbitrary divisor of @xmath99 such that inequality ( [ eqn : equivalence_subgroup_order_perspective ] ) holds . thus , there is a subgroup of @xmath98 , to be denoted by @xmath113 , of order @xmath114 . for all @xmath155 , write @xmath156 for brevity . note that under condition ( [ eqn : equivalence_subgroup_order_perspective ] ) , every @xmath157 is not greater than @xmath158 . for each @xmath155 , let @xmath159 denote the union @xmath160 of @xmath157 cosets @xmath161 , @xmath162 , @xmath163 , @xmath164 of @xmath113 in @xmath98 . thus , @xmath165 and recursively @xmath166 since @xmath167 is in at most @xmath168 cosets of @xmath113 , @xmath169 where the last inequality holds under condition ( [ eqn : equivalence_subgroup_order_perspective ] ) . if for each @xmath155 , @xmath116 is assigned to contain arbitrary @xmath170 elements in @xmath159 , then @xmath171 such a selection of @xmath103 makes @xmath109 not exhaust @xmath98 as desired , so inequality ( [ eqn : equivalence_subgroup_order_perspective ] ) holds . inequality ( [ eqn : equivalence_coset_perspective ] ) is a simple variation of ( [ eqn : equivalence_subgroup_order_perspective ] ) . in view of the explicit characterization ( [ eqn : equivalence_coset_perspective ] ) in theorem [ thm : concise_linear_solvability_characterization ] , the more important algebraic identity that affects the linear solvability of @xmath76 is the multiplicative subgroup order in gf(@xmath1 ) rather than the field size . let @xmath114 be a divisor of @xmath99 . then there is a subgroup in gf(@xmath1)@xmath26 of order @xmath114 . the value @xmath172 in ( [ eqn : equivalence_coset_perspective ] ) represents the total number of cosets of @xmath141 in gf(@xmath1)@xmath26 . for each @xmath155 , @xmath170 represents the out - degree of layer-3 node @xmath173 in @xmath61 , and @xmath174 in ( [ eqn : equivalence_coset_perspective ] ) represents the minimum number of cosets of @xmath141 in gf(@xmath1)@xmath26 such that each outgoing edge of @xmath173 can be assigned a different value in these cosets . thus , @xmath61 is the first class of multicast networks ever discovered which has a linear solvability characterization that matches the topological parameters with multiplicative subgroups and the associated coset numbers in gf(@xmath1 ) . based on theorem [ thm : concise_linear_solvability_characterization ] , the fact that the network @xmath175 with @xmath125 as depicted in fig . [ fig : rank_3_networks](b ) is not linearly solvable over gf(17 ) can be easily verified without exhaustive enumeration . one application of the explicit linear solvability characterization of the network @xmath61 constructed by algorithm [ alg : general_construction ] is to systematically yield multicast network linearly solvable over gf(@xmath1 ) but not over gf(@xmath2 ) with @xmath3 , and hence @xmath176 . we next introduce a simple way to do so stemming from the following criterion on subgroup orders . [ def : subgroup_order_criterion ] a pair ( gf(@xmath1 ) , gf(@xmath2 ) ) of finite fields is said to satisfy the _ subgroup order criterion _ if there is such a proper subgroup @xmath141 of gf@xmath177 other than @xmath178 that for every proper subgroup @xmath179 of gf@xmath180 , at least @xmath181 or @xmath182 . , it is possible for @xmath181 and @xmath182 to hold at the same time . but when @xmath3 , only one of @xmath181 and @xmath182 can hold . ] the inequality @xmath182 in @xmath183 implies the smaller cardinality of @xmath184 than the one of @xmath185 . it can be verified that for all known multicast networks with @xmath25 , the corresponding ( gf(@xmath23 ) , gf(@xmath24 ) ) satisfies the subgroup order criterion . in particular , ( gf(7 ) , gf(8 ) ) and ( gf(16 ) , gf(17 ) ) can be shown to satisfy the criterion by respectively setting @xmath141 to be any proper subgroup of gf(@xmath186)@xmath26 other than @xmath178 and to be the subgroup of gf(@xmath126)@xmath26 of order @xmath187 . some simple sufficient conditions for the subgroup order criterion to hold are characterized below . [ prop : subgroup_order_criterion ] consider a pair ( gf(@xmath1 ) , gf(@xmath2 ) ) with @xmath99 being a composite . if @xmath188 or @xmath2 is equal to a prime plus 1 , then by setting @xmath141 to be an arbitrary proper subgroup of gf(@xmath1)@xmath26 other than @xmath178 , condition @xmath183 holds . thus , in both cases , the pair ( gf(@xmath1 ) , gf(@xmath2 ) ) satisfies the subgroup order criterion . when @xmath99 is a composite , gf(@xmath1)@xmath26 has a proper subgroup other than @xmath178 . for the case @xmath188 , let @xmath179 be any proper subgroup of gf(@xmath2)@xmath26 . if @xmath189 , then there is nothing to prove . if @xmath190 , then @xmath191 . condition @xmath183 thus holds . for the case that @xmath2 is a prime plus one , @xmath178 is the only proper subgroup of gf(@xmath2)@xmath26 . therefore , by setting @xmath141 to be any proper subgroup of gf(@xmath1)@xmath26 other than @xmath178 , condition ( @xmath192 ) naturally holds . based on the concept of subgroup order criterion on @xmath193 , parameters @xmath194 for the general network @xmath61 can be appropriately chosen such that it has a linear solution over gf(@xmath1 ) but no linear solution over gf(@xmath2 ) . [ thm : implication_subgroup_order_criterion ] consider a pair ( gf(@xmath1 ) , gf(@xmath2 ) ) subject to the subgroup order criterion . let @xmath141 be any proper subgroup of gf(@xmath1)@xmath26 satisfying @xmath183 other than @xmath178 . write @xmath195 . for the multicast network @xmath76 constructed by algorithm [ alg : general_construction ] , select @xmath59 to satisfy @xmath196 and set the @xmath38-tuple @xmath90 to be @xmath197 . then , @xmath76 is linearly solvable over gf(@xmath1 ) but not over gf(@xmath2 ) . moreover , @xmath35 for the considered @xmath61 . first recall that algorithm [ alg : general_construction ] requires @xmath59 for the construction of @xmath61 . since @xmath114 is the order of subgroup @xmath141 of gf(@xmath1)@xmath26 , @xmath114 divides @xmath97 as well as @xmath198 . thus , @xmath199 and condition ( [ eqn : equivalence_subgroup_order_perspective ] ) in theorem [ thm : concise_linear_solvability_characterization ] holds . the considered network is thus linearly solvable over gf(@xmath1 ) . let @xmath179 be an arbitrary subgroup of gf(@xmath2)@xmath26 . write @xmath200 . in the case @xmath201 , @xmath202 where the inequality holds due to the assumed value of @xmath38 . thus , condition ( [ eqn : equivalence_subgroup_order_perspective ] ) does not hold over gf(@xmath2 ) for @xmath203 in this case . in the case @xmath204 but @xmath205 , @xmath206 where the last inequality can be established by noting that @xmath205 and @xmath203 divides @xmath207 . thus , condition ( [ eqn : equivalence_subgroup_order_perspective ] ) does not hold over gf(@xmath2 ) for @xmath203 in this case either . it has been verified that condition ( [ eqn : equivalence_subgroup_order_perspective ] ) does not hold over gf(@xmath2 ) for any divisor @xmath208 of @xmath209 . theorem [ thm : concise_linear_solvability_characterization ] then asserts the considered network @xmath61 is not linearly solvable over gf(@xmath2 ) . now consider an arbitrary prime power @xmath210 and an arbitrary divisor @xmath211 of @xmath212 . in the case @xmath213 , similar to the analysis above , @xmath214 so condition ( [ eqn : equivalence_subgroup_order_perspective ] ) does not hold over gf(@xmath215 ) for @xmath216 . consider the case @xmath217 . this implies @xmath218 and hence @xmath219 . note that @xmath220 consequently , @xmath221 where the last inequality holds due to @xmath59 and @xmath219 . hence , condition ( [ eqn : equivalence_subgroup_order_perspective ] ) does not hold over gf(@xmath215 ) for @xmath216 in this case either . we conclude that the considered network is not linearly solvable over gf(@xmath215 ) and hence @xmath35 . in the theorem above , the bound ( [ eqn : omega_sharp_bound ] ) on the choice of @xmath38 to guarantee the network @xmath61 not linearly solvable over gf(@xmath2 ) is sharp , since if @xmath222 for some divisor @xmath203 of @xmath209 smaller than @xmath114 , then @xmath223 and thus theorem [ thm : concise_linear_solvability_characterization ] affirms that the network is linearly solvable over gf(@xmath2 ) . besides the sharp bound ( [ eqn : omega_sharp_bound ] ) on @xmath38 , another convenient bound of @xmath224 to guarantee the network @xmath61 not linearly solvable over gf(@xmath2 ) is @xmath225 where @xmath226 is the largest divisor of @xmath209 that is smaller than @xmath114 . condition ( [ eqn : omega_weak_bound ] ) implies ( [ eqn : omega_sharp_bound ] ) due to @xmath227 for any @xmath228 . when @xmath209 is a prime , that is , when @xmath51 is the only proper divisor @xmath209 , conditions ( [ eqn : omega_sharp_bound ] ) and ( [ eqn : omega_weak_bound ] ) become the same . in general , rule ( [ eqn : omega_weak_bound ] ) is not tight . for example , consider the pair ( gf(16 ) , gf(17 ) ) which satisfies the subgroup order criterion by setting @xmath141 to be the subgroup of gf(16)@xmath26 containing 5 elements . thus , the smallest @xmath38 subject to ( [ eqn : omega_sharp_bound ] ) is 2 , but ( [ eqn : omega_weak_bound ] ) requires @xmath229 . @xmath75 for a same pair ( gf(@xmath1 ) , gf(@xmath2 ) ) satisfying the subgroup order criterion , different choices of subgroups @xmath141 of gf(@xmath1)@xmath26 obeying @xmath183 will yield different multicast networks @xmath61 linearly solvable over gf(@xmath1 ) but not over gf(@xmath2 ) . for instance , for the pair ( gf(7 ) , gf(8 ) ) , both subgroups @xmath230 and @xmath231 satisfy condition @xmath183 . hence , according to theorem [ thm : implication_subgroup_order_criterion ] , not only the network @xmath138 with @xmath139 depicted in fig . [ fig : rank_3_networks](a ) , but also the network @xmath232 with @xmath233 depicted in fig . [ fig : rank_4_network](a ) can be constructed to be linearly solvable over gf(7 ) but not over gf(8 ) . based on the subgroup order criterion and theorem [ thm : implication_subgroup_order_criterion ] , a number of interesting new multicast networks @xmath61 linearly solvable over gf(@xmath1 ) but not over gf(@xmath2 ) can be accordingly designed . this will be illustrated in the next section . however , it is worth noting that _ not _ every @xmath61 with @xmath28 can be found in this manner . for instance , consider the pair ( gf(16 ) , gf(17 ) ) . though it satisfies the subgroup order criterion with respect to @xmath141 being the subgroup of order 5 in gf(@xmath126)@xmath26 , if we set @xmath141 to be the subgroup of order 3 in gf(@xmath126)@xmath26 , condition @xmath183 is no longer obeyed . on the other hand , one can check , based on theorem [ thm : concise_linear_solvability_characterization ] , that the network @xmath138 depicted in fig . [ fig : rank_4_network](b ) with @xmath234 is also linearly solvable over gf(16 ) but not over gf(17 ) . stemming from theorem [ thm : concise_linear_solvability_characterization ] , one might be able to formulate other criteria from the perspective of subgroup orders for appropriate selection of @xmath38 and @xmath38-tuple @xmath90 such that the corresponding network @xmath61 has @xmath28 . however , that is beyond the scope of the current paper , since as we shall see in the next section , the subgroup order criterion discussed in this section has already allowed us to unveil infinitely many new interesting instances with @xmath28 . in the following , we aim at finding new multicast networks with @xmath28 and answering open problems raised in @xcite . the tool is to establish prime power pairs ( gf(@xmath1 ) , gf(@xmath2 ) ) with @xmath3 subject to the subgroup order criterion introduced in the previous section . as pointed out in @xcite , one of their motivations to design the first few exemplifying multicast networks with @xmath28 is the matroid structure of _ free swirl_. correspondingly , a class of so - called _ swirl networks _ are constructed , which are linearly solvable over gf(@xmath23 ) with @xmath235 but not over gf(@xmath2 ) when @xmath236 and @xmath2 is equal to a prime plus 1 . as special instances in the present framework , they are the networks @xmath61 constructed by algorithm [ alg : general_construction ] with an arbitrary source dimension @xmath59 and the @xmath38-tuple @xmath237 . when the prime power @xmath238 is equal to a prime plus 1 , it must be in the form of @xmath239 with @xmath7 a prime and @xmath240 is known to be a mersenne prime . since there exist @xmath241 and @xmath242 such that @xmath243 and @xmath244 are composite while @xmath245 is a mersenne prime , the swirl network not only demonstrates that @xmath28 is possible for multicast networks , but also verifies a conjecture raised in @xcite that a multicast network linearly solvable over gf(@xmath246 ) and gf(@xmath247 ) is _ not even necessarily _ linearly solvable over gf(@xmath248 ) . however , although it is a popular conjecture that there are infinitely many mersenne primes in the literature , only 48 mersenne primes have been discovered so far ( see the gimps project @xcite ) . thus , at this moment , the class of swirl networks , as well as other exemplifying networks with @xmath28 , still fails to show the following fundamental problems on multicast networks : * can gf(@xmath23 ) , gf(@xmath24 ) have the _ same _ characteristic ? * can the gap @xmath33 tend to _ infinity _ ? * are there _ infinitely many _ prime power pairs @xmath249 with @xmath3 such that each @xmath249 is equal to @xmath250 for some multicast networks ? in the remainder of this section , we first design appropriate parameters @xmath38 and @xmath90 for the general network @xmath61 to affirm positive answers to _ all _ problems listed above . [ thm : multicast_networks_characteristic_2 ] let @xmath251 and @xmath252 , where @xmath9 is an arbitrary integer larger than 2 . set @xmath59 satisfying ( [ eqn : omega_sharp_bound ] ) and the @xmath38-tuple @xmath253 . then , the multicast network @xmath61 is linearly solvable over gf(@xmath1 ) but _ not _ over gf(@xmath254 ) . moreover , @xmath30 and @xmath255 for some @xmath256 . we shall first show that the subgroup order criterion of definition [ def : subgroup_order_criterion ] holds for ( gf(@xmath1 ) , gf(@xmath254 ) ) . since @xmath257 is divisible by 3 , in the multiplicative group @xmath258 , there is a subgroup @xmath141 of order @xmath259 . on the other hand , it is easy to check that 7 is the smallest integer larger than 1 that divides @xmath260 . let @xmath203 be the order of an arbitrary subgroup @xmath179 of @xmath261 . then , @xmath262 . because @xmath263 we conclude that condition @xmath183 in definition [ def : subgroup_order_criterion ] holds and the subgroup order criterion is satisfied . moreover , since @xmath38 is set to satisfy ( [ eqn : omega_sharp_bound ] ) and @xmath264 , theorem [ thm : implication_subgroup_order_criterion ] asserts the considered network is linearly solvable over gf(@xmath5 ) but not over gf(@xmath6 ) and @xmath30 . it remains to prove @xmath255 for some @xmath256 . let @xmath2 be an arbitrary prime power larger than @xmath6 and _ not _ in the form of @xmath265 . it is then equivalent to show that the network is linearly solvable over gf(@xmath2 ) . as a consequence of the equivalent condition ( [ eqn : equivalence_subgroup_order_perspective ] ) in theorem [ thm : implication_subgroup_order_criterion ] , it suffices to find a proper subgroup @xmath179 of @xmath266 such that @xmath267 ) holds as @xmath268 and theorem [ thm : concise_linear_solvability_characterization ] then implies the considered network is linearly solvable over gf(@xmath2 ) . assume that @xmath2 is odd . set @xmath179 to be the subgroup @xmath269 of @xmath266 , where @xmath93 is a primitive element of gf(@xmath2 ) . it can be checked that @xmath270 and hence condition ( [ eqn : prove_q*_max ] ) holds . assume that @xmath2 is even . since @xmath2 is assumed not in the form of @xmath265 , @xmath271 for some @xmath272 . then , @xmath209 is divisible by 3 and hence we can set @xmath179 to be the subgroup @xmath273 of @xmath266 . it can be checked @xmath274 so condition ( [ eqn : prove_q*_max ] ) holds for this case too . we have verified that @xmath24 must be in the form @xmath275 form some @xmath256 . the class of networks @xmath61 under the setting of @xmath38 and @xmath90 in theorem [ thm : multicast_networks_characteristic_2 ] not only turns out to be the first discovered in the literature with gf(@xmath23 ) , gf(@xmath24 ) having the _ same characteristic _ , but also subsequently answers two open questions raised in @xcite . [ coro : infi_even_prime_powers ] there are _ infinitely many even _ prime power pairs @xmath249 such that @xmath276 for some multicast network . moreover , the gap @xmath33 can tend to _ infinity_. consider the network @xmath61 under the setting of @xmath38 and @xmath90 in theorem [ thm : multicast_networks_characteristic_2 ] . when @xmath9 tends to infinity , @xmath277 tends to infinity too . theorem [ thm : multicast_networks_characteristic_2 ] uncovered infinitely many pairs ( gf(@xmath1 ) , gf(@xmath2 ) ) of finite fields with characteristic @xmath278 and @xmath3 such that there is a multicast network @xmath61 linearly solvable over gf(@xmath1 ) but not over gf(@xmath2 ) . we next further extend the result to the cases that gf(@xmath1 ) and gf(@xmath2 ) have _ arbitrary distinct _ characteristics . we need the following lemma , which is established based on the _ weyl s equidistribution theorem _ ( see , e.g. , @xcite . ) [ lemma : diophantine_approximation ] let @xmath279 be two coprime integers larger than 1 , and @xmath280 , @xmath281 , @xmath282 be real numbers with @xmath283 . there are infinitely many positive integers @xmath284 such that @xmath285 see appendix[appendix_diophantine ] . [ thm : p_vs_p_general_odd ] let @xmath286 be two arbitrary distinct prime numbers . there are _ infinitely many _ @xmath284 such that there is a multicast network @xmath61 linearly solvable over gf(@xmath23 ) with @xmath287 , but _ not _ over gf(@xmath12 ) with @xmath288 . ( sketch ) note that there must be such a prime power @xmath289 that @xmath97 has a divisor @xmath114 no smaller than 3 . then , @xmath290 is divisible by @xmath114 for all @xmath291 . based on lemma [ lemma : diophantine_approximation ] , it can be shown that there are infinitely many integers @xmath284 with @xmath292 such that the pair ( gf(@xmath293 ) , gf(@xmath294 ) ) satisfies the subgroup order criterion by setting @xmath141 to be the subgroup of gf(@xmath293)@xmath26 of order @xmath295 , and hence there is a multicast network @xmath61 linearly solvable over gf(@xmath23 ) with @xmath296 , but _ not _ over gf(@xmath294 ) according to theorem [ thm : implication_subgroup_order_criterion ] . a detailed proof can be found in appendix[appendix_general_p_vs_p ] . as a counterpart of theorem [ thm : multicast_networks_characteristic_2 ] , theorem [ thm : p_vs_p_general_odd ] unveils that for _ any _ two distinct primes @xmath7 and @xmath8 , we can make use of the subgroup order criterion to find infinitely many ( @xmath11 , @xmath12 ) with @xmath13 such that there is a multicast network @xmath61 linearly solvable over gf(@xmath11 ) but not over gf(@xmath12 ) . however , @xmath24 is not necessarily a power of @xmath8 for this established @xmath61 , which is weaker than the consideration in theorem [ thm : multicast_networks_characteristic_2 ] . however , we can still have the following partial generalization of corollary [ coro : infi_even_prime_powers ] , as a consequence of the above theorem . [ corollary : infi_general_prime_powers ] let @xmath7 be an arbitrary prime . there are _ infinitely many _ prime power pairs @xmath297 such that @xmath298 for some multicast networks . all above interesting results are established based on the general framework developed in section [ sec : general_framework ] and the subgroup order criterion formulated in section [ sec : subgroup_order_criterion ] . yet , this approach seems not helpful to find exemplifying multicast networks linearly solvable over gf(@xmath1 ) but not over gf(@xmath2 ) with @xmath3 when gf(@xmath1 ) and gf(@xmath2 ) have the _ same odd _ characteristic . a key reason is as follows . [ prop : framework_insufficiency ] if the network @xmath61 constructed by algorithm [ alg : general_construction ] is linearly solvable over gf(@xmath11 ) , where @xmath7 is odd , then it is linearly solvable over gf(@xmath299 ) for all @xmath256 . if the network @xmath61 is linearly solvable over gf(@xmath11 ) , then there is a divisor @xmath152 of @xmath300 subject to @xmath301 according to theorem [ thm : concise_linear_solvability_characterization ] . this implies that @xmath302 , and thus @xmath303 for all @xmath155 . let @xmath10 be an arbitrary integer larger than @xmath9 . since @xmath7 is odd , @xmath304 is divisible by 2 . write @xmath305 . since @xmath306 , we have @xmath307 for all @xmath155 . hence , @xmath308 theorem [ thm : concise_linear_solvability_characterization ] now in turn affirms that the network @xmath61 is linearly solvable over gf(@xmath299 ) . * problem*. is there a multicast network that is linearly solvable over gf(@xmath11 ) but not over gf(@xmath299 ) for an odd prime @xmath7 and @xmath309 ? let @xmath1 be an arbitrary prime power . the @xmath34-combination network is the best known network in the network coding literature with @xmath23 equal to the prescribed @xmath1 . another application of the general framework developed in section [ sec : general_framework ] is to construct new multicast networks with @xmath23 equal to @xmath1 . assume that @xmath1 is not equal to a prime plus 1 . let @xmath114 denote an arbitrary proper divisor of @xmath97 other than 1 . the network @xmath175 with @xmath310 constructed by algorithm [ alg : general_construction ] has @xmath35 . according to proposition [ prop : subgroup_order_criterion ] , ( gf(@xmath1 ) , gf(@xmath2 ) ) is subject to the subgroup order criterion for all prime powers @xmath188 . moreover , for every prime power @xmath188 , when @xmath38 is set to 3 , condition ( [ eqn : omega_sharp_bound ] ) in theorem [ thm : implication_subgroup_order_criterion ] holds . thus , theorem [ thm : implication_subgroup_order_criterion ] shows @xmath35 for the considered network @xmath175 with @xmath310 . the proposition above justifies a new way to construct multicast networks with @xmath23 equal to a prescribed prime power @xmath1 . in order to further reduce the network size while keeping @xmath35 for the constructed network , we are motivated to consider the network @xmath138 with @xmath311 , which can be regarded as a degenerate instance constructed by algorithm [ alg : general_construction ] in the sense that not every component in @xmath90 is larger than 1 . none of the previous discussions indicates the linear solvability of this degenerate case . actually , this network has @xmath312 because in this degenerate case , condition a ) becomes weaker than b ) and c ) in lemma [ lemma : general_network_linear_solvability ] . larger than 1 . ] we next refine the network by carefully resetting some receivers , to satisfy : ( i ) @xmath35 , and ( ii ) the resulting network has a smaller number of nodes than the @xmath34-combination network . [ alg : q-1-omega ] let @xmath1 be an arbitrary prime power no smaller than 5 . construct a network , to be denoted by @xmath313 , consisting of nodes on @xmath187 layers as follows . the nodes at the first 4 layers and the edges among them are identical to the degenerate network @xmath138 with @xmath311 constructed by algorithm [ alg : general_construction ] , and are redrawn in fig . [ fig : netowrk_q_1_3 ] . create bottom - layer nodes , which will be set as receivers , according to the following rule : * there is a node connected from @xmath314 as well as a node from @xmath315 ; * there is a node connected from @xmath316 as well as a node from @xmath317 for each @xmath318 ; * there is a node connected from @xmath319 for each @xmath320 , and a node connected from @xmath321 for each @xmath322 ; * there is a node connected from @xmath323 for each @xmath324 . thus , there will be edges directly connected from layer-2 nodes to layer-5 nodes , and the total number of layer-5 nodes is @xmath325 . the network @xmath326 constructed by algorithm [ alg : q-1-omega ] is linearly solvable over gf(@xmath2 ) with every @xmath327 . in an essentially same way to derive equivalent conditions b ) and c ) of linear solvability for network @xmath61 in lemma [ lemma : general_network_linear_solvability ] , it is straightforward to show that there is a linear solution over gf(@xmath2 ) for network @xmath326 if and only if there is a matrix completion of @xmath328\ ] ] over gf(@xmath2 ) , and consequently if and only if there exist @xmath329 , @xmath330 such that @xmath331 when @xmath332 , the number of elements in @xmath333 is smaller than @xmath334 , so that the condition above can not be satisfied . when @xmath335 , set @xmath336 , and assign arbitrary @xmath334 distinct elements in @xmath333 to @xmath337 . in this way , the condition above can be satisfied . we can then conclude that the network is linearly solvable over gf(@xmath338 ) for all @xmath335 while not linearly solvable over gf(@xmath338 ) for all @xmath332 . for the network @xmath326 constructed by algorithm [ alg : q-1-omega ] , there are @xmath339 nodes on the first 4 layers and then @xmath340 nodes in the whole network @xmath326 . it can also be counted that there are @xmath341 edges in the network . in comparison , in the @xmath34-combination network with @xmath35 , there are total @xmath342 nodes and @xmath343 edges . thus , the size of the network @xmath326 is smaller than the @xmath344-combination network in terms of the number of receivers as well as the number of nodes . although the @xmath344-combination network has a smaller number of edges , it has source dimension @xmath38 equal to 2 while the network @xmath326 has @xmath345 . it would be fairer for the network size comparison if we increase @xmath38 of the combination network to 3 . to do so , as what has been adopted in @xcite , simply create a new node with the unique incoming edge leading from the source , and connect it to every receiver by a new edge . in this extended network , there are @xmath346 nodes and @xmath347 edges . compared with this network , the network @xmath326 constructed by algorithm [ alg : q-1-omega ] has not only fewer receivers and nodes , but fewer edges as well . table [ table : network_size_comparison ] summarizes the sizes of the discussed three types of multicast networks with @xmath35 . .comparison of multicast networks with @xmath23 equal to a prescribed prime power @xmath1 [ cols="^,^,^,^ " , ] in this paper , we propose a particular class of multicast networks @xmath61 with topological parameters @xmath194 . by deriving a generalized cauchy - davenport theorem over the additive group @xmath143 , we obtain an explicit formula on the linear solvability of @xmath61 over a base field gf(@xmath1 ) , which connects @xmath38 and @xmath90 with the associated coset numbers of a multiplicative subgroup in gf(@xmath1 ) , rather than the conventional algebraic identity field size . stemming from the special linear solvability behavior of @xmath76 , we further formulate a subgroup order criterion for a pair of finite fields . for every pair ( gf(@xmath1 ) , gf(@xmath2 ) ) subject to the subgroup order criterion , an instance in @xmath61 can be found to be linearly solvable over gf(@xmath1 ) but not over gf(@xmath2 ) . subsequently , different classes of infinitely many instances in @xmath76 are established with the special property @xmath28 , where @xmath23 is the minimum field size for the existence of a linear solution and @xmath24 is the maximum field size for the non - existence of a linear solution . moreover , it is proved that the gap @xmath33 can tend to infinity . our findings suggest a new `` matching '' between the algebraic structure of a base field and the topological structure of a particular class of multicast networks , and this matching condition is both necessary and sufficient for the existence of a linear solution to the multicast networks . for a more general multicast network coding problem , it is interesting to explore and characterize similar matching conditions that are sufficient for the existence and nonexistence of a linear solution . this lemma can be proven in a similar way as that of theorem 8 in @xcite . here we outline the sketch of the proof . consider an lnc with all coding coefficients being indeterminates . without loss of generality , assume that the coding coefficients are set to @xmath51 for all those adjacent pairs @xmath348 where @xmath349 is the unique incoming edge to some node . assume the coding vector for the unique incoming edge to node @xmath350 is equal to the @xmath351 @xmath38-dimensional unit vector , and denote by @xmath352 the @xmath353 outgoing edge from node @xmath350 , @xmath155 , @xmath354 . then , the juxtaposition of coding vectors for edges @xmath352 can be represented as @xmath355\ ] ] where @xmath356 , @xmath357 represent coding coefficient indeterminates . since in @xmath61 there is a receiver connected from _ every _ set @xmath41 of @xmath38 layer-4 ( grey ) nodes with @xmath74 , there exists a linear solution over gf(@xmath1 ) if and only if there exists a matrix completion of @xmath358 over gf(@xmath1 ) . note that the coding vectors for @xmath359 form a full rank matrix @xmath360 $ ] , where @xmath361 is the only nonzero entry in the first row . thus , for every matrix completion of @xmath358 over gf(@xmath1 ) , @xmath361 must be set to a nonzero element in gf(@xmath1 ) . by a similar argument , all indeterminates @xmath356 and @xmath357 must be assigned nonzero elements in gf(@xmath2 ) too . it can be subsequently seen that there is a matrix completion of @xmath358 over gf(@xmath1 ) if and only if there is a matrix completion of matrix @xmath81 in ( [ eqn : general_matrix_m_for_linear_solvability_analysis ] ) over gf(@xmath1 ) , where @xmath362 must be assigned two distinct nonzero elements in gf(@xmath1 ) for all @xmath155 , @xmath363 . last , since coding vectors for @xmath364 , @xmath365 , form a full rank matrix @xmath366 $ ] , for every matrix completion of the matrix @xmath81 in ( [ eqn : general_matrix_m_for_linear_solvability_analysis ] ) over gf(@xmath1 ) , @xmath367 must be such assigned that @xmath368 . there is thus no difficulty to establish the equivalence between condition b ) and c ) in the lemma . @xmath75 first , note that if the elements in @xmath369 are in a same coset of the stabilizer group @xmath113 , then @xmath370 . moreover , if they are in @xmath371 different cosets of @xmath113 , then @xmath372 . since the elements in each @xmath373 , @xmath374 , belong to at least @xmath375 different cosets of @xmath113 , @xmath376 ) implies ( [ eqn : kneser_generalization ] ) . the kneser s additive theorem ( see , e.g. , theorem 3.3.2 in @xcite ) asserts that for any two nonempty finite subsets @xmath132 and @xmath133 in an abelian group @xmath141 , the inequality @xmath377 holds , where @xmath378 is the stabilizer group of @xmath379 . thus , the case of @xmath148 for inequality ( [ eqn : appendix_proof_1 ] ) degenerates to the kneser s theorem . by induction on @xmath9 , we may assume that the theorem holds when @xmath9 is substituted by @xmath380 . thus , latexmath:[\[\label{eqn : appendix_proof_2 } since @xmath113 is a group , @xmath382 . by applying the kneser s theorem again , we have @xmath383 is the stabilizer group of @xmath384 in @xmath145 . since @xmath385 , @xmath113 is contained in the stabilizer group @xmath386 . on the other hand , because @xmath387 where the first equality holds for @xmath386 is the stabilizer group of @xmath384 and the second equality holds for @xmath113 is the stabilizer group of @xmath388 , @xmath389 is contained in the stabilizer group @xmath113 too . thus , @xmath390 , and ( [ eqn : appendix_proof_3 ] ) becomes @xmath391 this , together with ( [ eqn : appendix_proof_2 ] ) implies ( [ eqn : appendix_proof_1 ] ) as desired , so condition ( [ eqn : kneser_generalization ] ) holds . since the order of the stabilizer group @xmath113 divides @xmath20 , inequality ( [ eqn : cauchy - davenport - generalization ] ) is a direct consequence of ( [ eqn : kneser_generalization ] ) . @xmath75 let @xmath392 be an arbitrary real number with @xmath393 . if @xmath394 , then set @xmath395 . otherwise , set @xmath396 . when @xmath397 , we have @xmath398 , and hence @xmath399 , that is , @xmath400 consequently , it suffices to show that there are infinitely many positive integers @xmath9 , @xmath10 such that @xmath401 if @xmath402 , then for every @xmath403 , there is an integer @xmath9 in the interval @xmath404 . assume that @xmath405 . since @xmath406 , @xmath407 are coprime , @xmath408 is an irrational number . thus , according to the weyl s equidistribution theorem ( see , e.g. , @xcite ) , @xmath409 is uniformly distributed over @xmath410 , where @xmath411 is the real number field . therefore , a fraction @xmath412 of all positive integers @xmath10 can make the interval @xmath404 contain an integer , that is , to make the inequality ( [ eqn : appendix_b_proof ] ) hold . let @xmath1 be a prime power @xmath413 such that @xmath97 has a divisor @xmath114 larger than 2 . thus , @xmath290 is divisible by @xmath114 for all @xmath291 . let @xmath414 be the smallest integer such that @xmath415 is divisible by @xmath114 , if there does not exist such one , then set @xmath416 . thus , @xmath417 is not divisible by @xmath114 for all @xmath403 . let @xmath284 be arbitrary positive integers subject to @xmath418 according to lemma [ lemma : diophantine_approximation ] , there are infinitely many such choices of @xmath9 and @xmath10 . we next show that the subgroup order criterion in definition [ def : subgroup_order_criterion ] holds for ( gf(@xmath293 ) , gf(@xmath419 ) ) by setting @xmath141 to be the subgroup in gf(@xmath293)@xmath26 of order @xmath295 . in the case that @xmath420 does not have a proper divisor no smaller than @xmath421 , the subgroup order criterion naturally holds for ( gf(@xmath293 ) , gf(@xmath419 ) ) as desired . otherwise , let @xmath203 be an arbitrary proper divisor of @xmath420 no smaller than @xmath421 . by the second inequality in ( [ eqn : inequality_p_vs_p_general_odd ] ) , it can be deduced that @xmath422 since the divisor @xmath114 of @xmath99 is selected larger than 2 , @xmath423 and consequently , @xmath424 moreover , since @xmath114 does not divide @xmath425 , @xmath426 by combining ( [ eqn : inequality_p_vs_p_general_odd1 ] ) and ( [ eqn : inequality_p_vs_p_general_odd2 ] ) , we obtain @xmath427 this implies that the subgroup order criterion in definition [ def : subgroup_order_criterion ] indeed holds for ( gf(@xmath293 ) , gf(@xmath419 ) ) as desired . consider the network @xmath61 with @xmath38 set to satisfy ( [ eqn : omega_sharp_bound ] ) and @xmath90 to be the @xmath38-tuple @xmath428 . theorem [ thm : implication_subgroup_order_criterion ] then shows that @xmath61 is linearly solvable over gf(@xmath23 ) with @xmath429 but not linearly solvable over gf(@xmath419 ) , where @xmath430 is due to the first inequality in ( [ eqn : inequality_p_vs_p_general_odd ] ) . @xmath75 s. jaggi , p. sanders , p. a. chou , m. effros , s. egner , k. jain , and l. tolhuizen , `` polynomial time algorithms for multicast network code construction , '' _ ieee trans . inf . theory _ , 51 , no . 6 , pp . 1973 - 1982 , jun . 2005

we present the results of measurements of the solar neutrino capture rate in gallium metal by the russian - american gallium experiment sage during slightly more than half of a 22-year cycle of solar activity . combined analysis of the data of 92 runs during the 12-year period january 1990 through december 2001 gives a capture rate of solar neutrinos with energy more than 233 kev of ( stat . ) ( syst . ) snu . this represents only slightly more than half of the predicted standard solar model rate of 128 snu . we give the results of new runs beginning in april 1998 and the results of combined analysis of all runs since 1990 during yearly , monthly , and bimonthly periods . using a simple analysis of the sage results combined with those from all other solar neutrino experiments , we estimate the electron neutrino flux that reaches the earth to be/(-s ) . assuming that neutrinos oscillate to active flavors the neutrino flux emitted in the solar fusion reaction is approximately/(-s ) , in agreement with the standard solar model calculation of/(-s )

astro-ph0204245

there has been outstanding progress in solar neutrino research in the last several years . the large water cherenkov detectors superkamiokande ( sk ) @xcite and the sudbury neutrino observatory ( sno ) @xcite have begun to measure the high - energy solar neutrinos from @xmath7b decay in real time with a high counting rate . the results from these two neutrino telescopes of a new generation add significant information to the existing results from the chlorine @xcite and gallium @xcite radiochemical experiments and the kamiokande experiment @xcite . the elastic scattering results of sk combined with the charged current results from sno indicate that , in addition to electron neutrinos , active neutrinos of other flavors are reaching the earth from the sun . the combined analysis of results of all these experiments gives convincing evidence that part of the electron neutrinos formed in thermonuclear reactions in the sun change their flavor on their way to the earth . further measurement of the details of flavor conversion for solar neutrinos requires a new generation of neutrino telescopes that will be sensitive to the spectrum below 2 mev , a region that contains the proton - proton @xmath8 and cno continua as well as the @xmath9be and @xmath10 lines . although there are many promising ideas for real - time low - energy neutrino detectors @xcite , at the present time only the radiochemical gallium experiments are able to measure and monitor this part of the solar neutrino spectrum . the 233-kev threshold of the reaction @xmath11ga@xmath12ge @xcite enables one to measure the @xmath2 neutrinos , the principal component of the solar neutrino spectrum . if we exclude exotic hypotheses , the rate of the @xmath2 reaction is directly related to the solar luminosity and is insensitive to alterations in the solar models that influence the subsequent reactions in the solar fusion chain . the neutrino capture rate in @xmath13ga predicted by the standard solar model ( ssm ) @xcite is @xmath14 snu , with the main contribution of 69.7 snu from the @xmath2 neutrinos atoms of the neutrino absorbing isotope . contributions by @xmath9be and @xmath7b neutrinos are 34.2 snu and 12.1 snu , respectively . the insensitivity of ga to variation in the solar model is seen in the independently calculated result of 127.2 snu @xcite for the total capture rate . from the results of the sno and sk experiments we now have a high accuracy measurement of the @xmath7b flux that reaches the earth and its electron neutrino component . in the near future it is expected that the kamland experiment @xcite will greatly restrict the range of possible electron neutrino mixing parameters . this data , combined with results from the borexino experiment @xcite , will give us a good measurement of the @xmath9be flux from the sun . by subtracting the @xmath9be and @xmath7b components from the total signal in the ga experiment , we will obtain a measurement of a fundamental astrophysical parameter the neutrino flux from the proton - proton fusion reaction . there will be a slight contamination from the cno and @xmath10 neutrinos but this can be removed with the information from the cl experiment . a rough estimate of the @xmath2 neutrino flux using the information that we have now is presented in the next to last section . since in the immediate future only the ga experiments provide this measurement , it is very important that both sage @xcite and gallex s successor gno @xcite continue to operate so as to improve the accuracy of their results . the experimental layout and procedures , including extraction of germanium from gallium , counting of @xmath13ge , and data analysis , are described in detail in our article `` measurement of the solar neutrino capture rate with gallium metal '' in physical review c @xcite . that article gives the sage results for the period january 1990 through december 1997 . since it contains a complete description of the experiment and the analysis techniques , we refer the reader who wishes further detail to that publication . here we briefly discuss the main principles of the experiment , give the statistical analysis of the data from 19982001 , present new results for some systematic uncertainties , and conclude with the current implications of the sage result for solar and neutrino physics . the sage experiment is situated in a specially built deep underground laboratory @xcite at the baksan neutrino observatory ( bno ) of the institute for nuclear research of the russian academy of sciences in the northern caucasus mountains . it is located 3.5 km from the entrance of a horizontal adit excavated into the side of mount andyrchi . the main chamber of the laboratory is 60 m long , 10 m wide , and 12 m high . the rock gives an overhead shielding equivalent to 4700 m of water and reduces the muon flux by a factor of @xmath15 . the measured muon flux is @xmath16/(@xmath4-s ) @xcite . to reduce neutron and gamma backgrounds from the rock , the laboratory is entirely lined with 60 cm of low - radioactivity concrete with an outer 6 mm steel shell . the flux of neutrons with energy ( 111 ) mev is less than @xmath17/(@xmath4-s ) @xcite . all facilities required by the experiment are in this underground area , with additional rooms devoted to chemistry , counting , and a low - background solid - state ge detector . subsidiary measurements are made in a general laboratory building outside the adit . the gallium target used at the present time for measurements is about 50 t. it is in the form of liquid metal and is contained in 7 chemical reactors . a measurement of the solar neutrino capture rate , which we call a `` run '' , begins by adding to the gallium a stable ge carrier . the carrier is a ga - ge alloy with a known ge content of approximately 350 @xmath18 g and is distributed equally among all reactors . the reactor contents are stirred thoroughly to disperse the ge throughout the ga mass . after a typical exposure interval of 4 weeks , the ge carrier and @xmath13ge atoms produced by solar neutrinos and background sources are chemically extracted from the ga using procedures described elsewhere @xcite . the final step of the chemical procedure is the synthesis of germane ( geh@xmath19 ) , which is used as the proportional counter fill gas with an admixture of ( 8090)% xe . the total efficiency of extraction is the ratio of mass of ge in the germane to the mass of initial ge carrier and is typically in the range of ( 8090)% . the systematic uncertainty in this efficiency is 3.4% , mainly arising from uncertainties in the mass of added and extracted carrier . each extraction dissolves @xmath200.1% of the gallium . ga metal is regenerated from the accumulated extracted solutions , purified , and subsequently will be returned to the target . the proportional counter is placed in the well of a nai detector that is within a large passive shield and is counted for a typical period of 46 months . @xmath13ge decays via electron capture to @xmath13ga with a half - life of 11.43 d. low energy @xmath21- and @xmath22-shell auger electrons and x rays from electron shell relaxation produce a nearly point - like ionization in the counter gas , resulting in a fast rise time for the pulse from the counter . in contrast , most background events leave an extended trail of ionization , producing a slower rise time . a candidate @xmath13ge event must thus have an energy within a window around an @xmath22 or @xmath21 peak and must also have a rise time consistent with point - like ionization . in addition , since @xmath13ge decays without gamma emission , this event must not be coincident with an event in the nai detector . l @ c c c c c c d c c c c & mean & exposure & ga + exposure & exposure & time & mass & extraction & counter & pressure & & operating & @xmath21-peak & @xmath22-peak & peak + date & date & ( days ) & ( t ) & efficiency & name & ( mm hg ) & & voltage & efficiency & efficiency & ratio + apr . 98 & 1998.225 & 44.9 & 48.05 & 0.85 & a13 & 695 & 37.0 & 1480 & 0.243 & 0.219 & 1.01 + may 98 & 1998.347 & 30.0 & 51.17 & 0.91 & ly4 & 690 & 29.5 & 1366 & 0.238 & 0.245 & 1.00 + july 98 & 1998.477 & 45.6 & 51.06 & 0.90 & a12 & 680 & 32.0 & 1414 & 0.235 & 0.237 & 1.00 + aug . 98 & 1998.611 & 45.7 & 50.93 & 0.89 & la51 & 660 & 27.0 & 1356 & 0.234 & 0.244 & 1.04 + oct . 98 & 1998.745 & 45.8 & 50.81 & 0.92 & a13 & 680 & 32.0 & 1404 & 0.244 & 0.212 & 1.00 + nov . 98 & 1998.883 & 45.8 & 50.68 & 0.92 & ly4 & 680 & 26.5 & 1322 & 0.238 & 0.244 & 1.00 + jan . 99 & 1999.014 & 44.7 & 50.54 & 0.92 & a12 & 700 & 30.0 & 1398 & 0.239 & 0.241 & 1.00 + feb . 99 & 1999.130 & 38.7 & 50.43 & 0.89 & la51 & 705 & 11.0 & 1194 & 0.248 & 0.234 & 1.05 + apr . 99 & 1999.279 & 51.7 & 50.29 & 0.89 & a13 & 665 & 13.5 & 1206 & 0.253 & 0.231 & 1.05 + june 99 & 1999.417 & 46.7 & 50.17 & 0.87 & ly4 & 670 & 11.0 & 1140 & 0.246 & 0.239 & 1.00 + july 99 & 1999.551 & 45.7 & 50.06 & 0.90 & la116 & 635 & 12.5 & 1164 & 0.243 & 0.244 & 1.03 + sep . 99 & 1999.685 & 45.7 & 49.91 & 0.91 & la51 & 660 & 11.5 & 1172 & 0.242 & 0.238 & 1.05 + oct . 99 & 1999.801 & 38.7 & 49.78 & 0.90 & a13 & 665 & 12.5 & 1186 & 0.254 & 0.202 & 1.01 + jan . 00 & 2000.035 & 28.8 & 49.59 & 0.91 & la51 & 700 & 13.5 & 1224 & 0.324 & 0.310 & 1.05 + feb . 00 & 2000.127 & 30.7 & 49.48 & 0.83 & ly4 & 646 & 10.4 & 1130 & 0.320 & 0.316 & 1.01 + mar . 00 & 2000.207 & 28.8 & 49.42 & 0.91 & a13 & 665 & 14.5 & 1206 & 0.332 & 0.329 & 1.10 + may 00 & 2000.359 & 30.7 & 49.24 & 0.92 & la116 & 705 & 14.0 & 1244 & 0.329 & 0.315 & 1.03 + june 00 & 2000.451 & 33.7 & 49.18 & 0.84 & la51 & 652 & 12.0 & 1160 & 0.317 & 0.314 & 1.03 + july 00 & 2000.540 & 32.0 & 49.12 & 0.92 & ly4 & 670 & 13.8 & 1182 & 0.321 & 0.316 & 1.01 + aug . 00 & 2000.626 & 31.3 & 49.06 & 0.73 & a13 & 707 & 9.5 & 1176 & 0.343 & 0.321 & 1.08 + sep . 00 & 2000.704 & 27.7 & 49.00 & 0.89 & a12 & 690 & 14.7 & 1224 & 0.324 & 0.312 & 1.00 + oct . 00 & 2000.796 & 30.7 & 48.90 & 0.84 & la116 & 734 & 9.4 & 1188 & 0.337 & 0.303 & 1.03 + nov . 00 & 2000.876 & 28.7 & 48.84 & 0.93 & la51 & 680 & 11.9 & 1196 & 0.345 & 0.330 & 1.03 + dec . 00 & 2000.958 & 30.7 & 48.78 & 0.93 & ly4 & 697 & 12.0 & 1174 & 0.327 & 0.312 & 1.02 + feb . 01 & 2001.122 & 29.8 & 41.11 & 0.87 & la116 & 687 & 9.2 & 1144 & 0.330 & 0.314 & 1.04 + mar . 01 & 2001.214 & 33.4 & 48.53 & 0.92 & la51 & 635 & 13.5 & 1180 & 0.314 & 0.317 & 1.02 + apr . 01 & 2001.290 & 22.7 & 48.43 & 0.90 & yct1 & 695 & 13.1 & 1210 & 0.344 & 0.333 & 1.00 + may 01 & 2001.373 & 31.7 & 48.37 & 0.88 & yct2 & 625 & 14.9 & 1178 & 0.332 & 0.342 & 1.00 + june 01 & 2001.469 & 31.7 & 48.27 & 0.92 & yct3 & 678 & 12.2 & 1190 & 0.342 & 0.334 & 1.00 + july 01 & 2001.547 & 23.7 & 48.17 & 0.93 & la116 & 690 & 12.7 & 1196 & 0.328 & 0.315 & 1.03 + aug . 01 & 2001.624 & 28.7 & 48.11 & 0.59 & a12 & 768 & 7.2 & 1148 & 0.340 & 0.302 & 1.00 + sep . 01 & 2001.701 & 27.7 & 48.06 & 0.90 & yct1 & 665 & 15.0 & 1204 & 0.338 & 0.337 & 1.00 + oct . 01 & 2001.793 & 30.7 & 47.96 & 0.88 & yct2 & 758 & 12.2 & 1210 & 0.354 & 0.326 & 1.00 + nov . 01 & 2001.887 & 34.8 & 47.91 & 0.92 & yct3 & 685 & 14.2 & 1210 & 0.342 & 0.335 & 1.00 + dec . 01 & 2001.955 & 22.8 & 47.86 & 0.86 & yct4 & 685 & 11.4 & 1176 & 0.344 & 0.333 & 1.00 + the data acquisition electronics have evolved over the course of sage . during the first two years a hardware measure of the pulse rise time was used the amplitude of the differentiated pulse ( adp ) technique . this method suffices well in the @xmath21 peak ( 10.4 kev ) but was found to be inadequate for the @xmath22 peak ( 1.2 kev ) , which is more sensitive to electronic drifts and has higher background . in 1992 an 8-channel counting system with a 1-ghz digital oscilloscope was implemented , which permits off - line analysis of the event waveforms . the digitized pulse is fit to a functional form @xcite which gives the energy deposited during the event and the time duration @xmath23 over which the ionization arrives at the anode wire . all @xmath22-peak results and the vast majority of @xmath21-peak results are obtained from this waveform analysis . after filling , counters are calibrated with an @xmath24fe source ( 5.9 kev ) through a window in the fe cathode . typically , they are again calibrated after 3 days of operation and approximately every two weeks subsequently . calibrations are also made with a @xmath25cd source whose gamma rays penetrate the counter wall and fluoresce the length of the fe cathode , thus giving the @xmath21 x - ray peak from fe at 6.4 kev . this allows correction of the @xmath13ge peak position due to the accumulation of polymer deposits on the anode wire which may occur after prolonged operation . a @xmath25cd+se source , which gives peaks at 1.4 kev and 11.2 kev , is also occasionally used to check linearity . the measure of energy is the integral of the pulse waveform for 800 ns after pulse onset . the @xmath13ge peak position is based on the @xmath24fe calibration adjusted for polymerization and the energy window ( two full widths at half maximum ) is set by the @xmath24fe resolution . if the calibration centroid shifts between two calibrations , the window for energy selection is linearly shifted in time between the two calibrations . typical gain shifts are of the order of a few percent which leads to a calculated uncertainty in the counting efficiency of no more than -3.1% . to determine the rise time windows for @xmath13ge events , proportional counters filled with @xmath11geh@xmath19 are measured in each of the counting channels . all events inside the energy windows of the @xmath21 and @xmath22 peaks are selected , and the rise time @xmath23 of each event is calculated @xcite . the rise time values are arranged in ascending order and limits are set such that 5% of the events are excluded . for a counter filled with 30% geh@xmath19 this leads to @xmath21- and @xmath22-peak event selection limits on @xmath23 of 0.0 ns to 18.4 ns and 0.0 ns to 10.0 ns , respectively . the upper limits are reduced to 13.2 ns and 9 ns for the @xmath21 and @xmath22 peaks , respectively , when the geh@xmath19 concentration is in the range of 7% to 15% . the variation with counter filling and electronics channel results in an uncertainty in the efficiency of about @xmath261% . the small loss in counting efficiency associated with this cut leads to a significant reduction in the number of background events . table [ runparameters ] gives the parameters of the 35 runs from 19982001 that are used for solar neutrino measurement . the selection criteria result in a group of events from each extraction that are candidate @xmath13ge decays . these events are fit to a maximum likelihood function @xcite , assuming that they originate from an unknown but constant - rate background and the exponentially decaying rate of @xmath13ge . two time cuts are made on the counting intervals to reduce the effect of radon and radon daughters , which can give pulses that mimic @xmath13ge . to minimize the effect of rn that enters the interior of the passive shield when it is opened for counter calibration , we delete the first 2.6 h of counting time in the @xmath21 and @xmath22 peaks after shield openings . another very dangerous background occurs if even a few atoms of @xmath27rn enter the counter during filling . most decays of rn inside the counter give slow pulses at a saturated energy beyond the @xmath13ge peaks , but approximately 8% of the pulses from rn and its daughters make fast pulses that are indistinguishable from those of @xmath13ge . since the radon decay chain takes on average 50 min to reach the long - lived isotope @xmath28pb , deleting 15 min before and 3 h after each saturated pulse removes most of these internal rn events . 0.75l @ d d d r @ d @ c d & & + exposure & & & & & & + date & & & & & & ( % ) + apr . 98 & 39 & 5.4 & 75 & 26 & 134 & 0.052 & 72 + may 98 & 23 & 3.4 & 44 & 10 & 88 & 0.051 & 68 + july 98 & 22 & 4.8 & 61 & 24 & 108 & 0.065 & 52 + aug . 98 & 33 & 3.6 & 46 & 5 & 97 & 0.039 & 84 + oct . 98 & 40 & 3.8 & 45 & 4 & 95 & 0.028 & 95 + nov . 98 & 32 & 5.9 & 67 & 28 & 116 & 0.101 & 30 + jan . 99 & 21 & 4.5 & 56 & 15 & 107 & 0.036 & 84 + feb . 99 & 16 & 1.6 & 24 & 0 & 67 & 0.114 & 28 + apr . 99 & 10 & 1.8 & 38 & 5 & 83 & 0.105 & 36 + june 99 & 14 & 12.9 & 172 & 123 & 232 & 0.048 & 80 + july 99 & 17 & 5.5 & 103 & 49 & 172 & 0.118 & 20 + sep . 99 & 20 & 7.1 & 93 & 43 & 154 & 0.099 & 28 + oct . 99 & 16 & 10.0 & 138 & 80 & 206 & 0.066 & 56 + jan . 00 & 24 & 5.4 & 63 & 23 & 111 & 0.060 & 59 + feb . 00 & 21 & 9.1 & 107 & 63 & 157 & 0.058 & 55 + mar . 00 & 19 & 10.1 & 117 & 78 & 165 & 0.046 & 79 + may 00 & 15 & 0.0 & 0 & 0 & 32 & 0.143 & 40 + june 00 & 17 & 1.4 & 23 & 0 & 75 & 0.179 & 17 + july 00 & 29 & 6.4 & 69 & 33 & 111 & 0.088 & 34 + aug . 00 & 14 & 5.2 & 74 & 39 & 117 & 0.086 & 33 + sep . 00 & 30 & 9.2 & 111 & 64 & 166 & 0.093 & 24 + oct . 00 & 14 & 3.0 & 37 & 8 & 75 & 0.020 & 99 + nov . 00 & 25 & 2.9 & 32 & 0 & 73 & 0.208 & 9 + dec . 00 & 27 & 7.6 & 81 & 43 & 127 & 0.062 & 68 + feb . 01 & 21 & 6.3 & 79 & 43 & 125 & 0.088 & 34 + mar . 01 & 18 & 3.8 & 44 & 14 & 80 & 0.120 & 24 + apr . 01 & 17 & 6.7 & 76 & 43 & 117 & 0.074 & 45 + may 01 & 21 & 11.9 & 127 & 90 & 171 & 0.088 & 31 + june 01 & 20 & 9.4 & 93 & 57 & 135 & 0.025 & 96 + july 01 & 9 & 2.0 & 24 & 0 & 58 & 0.033 & 92 + aug . 01 & 21 & 5.4 & 90 & 38 & 155 & 0.065 & 57 + sep . 01 & 10 & 2.1 & 22 & 0 & 53 & 0.139 & 18 + oct . 01 & 12 & 7.5 & 73 & 44 & 109 & 0.082 & 41 + nov . 01 & 15 & 2.6 & 23 & 0 & 54 & 0.084 & 38 + dec . 01 & 9 & 5.2 & 62 & 34 & 101 & 0.063 & 70 + combined & 711 & 191.8 & 67 & 60 & 74 & 0.080 & 42 + for each individual extraction , the best estimate of the @xmath13ge production rate is found by maximizing the likelihood function . the statistical uncertainty in the production rate is found by integrating the likelihood function over the background rate to give a function of production rate only , and then finding the minimum range in this rate that includes 68% of the area under the curve . this procedure is done separately for events in the @xmath22 and @xmath21 peaks . the small variation in the earth - sun distance ( no greater than 3% ) due to the eccentricity of the earth s orbit is taken into account . the best estimate for a combination of runs ( and also for the @xmath29 combination of a single run ) is obtained by multiplying the individual likelihood functions , requiring that the production rate per unit mass of ga be equivalent but allowing the background rate to be different for each component of the function . the global maximum of the product function is then found and the 68% confidence region for the production rate is set by finding where the function has decreased from its value at the maximum by the factor 0.606 , all other variables being maximized . the results of the combined analysis of recent extractions are given in table [ recentruns ] and the results of all runs of sage are plotted in fig . [ ratevstime ] . after publication of ref . @xcite , which reported measurements from january 1990 through december 1997 , it was found that a faulty data acquisition program was used in the period from june 1996 through december 1999 . at the beginning of this period it was necessary to change the hardware that recognized coincidences between the nai and proportional counters . the new hardware required a modification of the acquisition program which introduced an error in the trigger logic . as a result @xmath30% of triggers this error artificially reduced the results of individual runs counting during this period and slightly influenced the overall result . corrected results were given in @xcite . table [ systematics ] summarizes the systematic effects that may affect the measured solar neutrino capture rate . these uncertainties fall into three main categories : those associated with extraction efficiency , with counting efficiency , and with backgrounds . some of these effects were mentioned above and the others will be briefly discussed here . the counting efficiency and its uncertainty was determined by a series of measurements with gas fillings of @xmath13ge , @xmath31ar , and @xmath32ge . the uncertainty contains terms due to volume efficiency , end effects , and gas efficiency . adding each of these effects in quadrature gives a @xmath33 uncertainty due to the counters . there exist small contributions to the @xmath13ge signal by means other than solar neutrinos . limits on the creation of @xmath13ge through the @xmath34 reaction on @xmath13ga and by cosmic - ray muons were obtained by measurement in the ga chamber of both the fast neutron @xcite and muon fluxes @xcite . limiting values on the u and th concentration in the ga , which can also make @xmath13ge , have been determined by low background counting in a ge detector @xcite and by glow discharge mass spectrometry @xcite . the inferred @xmath13ge production rate from the combination of all of these processes is no more than 1 snu . rn decay in the vicinity of the proportional counters or inside the counters themselves can also give events that mimic @xmath13ge decay . to reduce such events rn is purged from the volume near the counters by a constant flow of gas from boiling liquid nitrogen and special anti - rn procedures are applied to purify the gas mixture at the time of counter filling . the influence of residual rn was quantified by studies in which rn was added inside the counter @xcite and the response of the counter to external gamma - rays was measured @xcite . the systematic uncertainties on the @xmath13ge production rate for rn that remains after the time cuts were determined to be @xmath350.2 snu for internal rn and 0.03 snu for external rn . l @ l c extraction efficiency & ge carrier mass & @xmath261.5 + & extracted ge mass & @xmath261.8 + & residual carrier ge & @xmath260.6 + & ga mass & @xmath260.2 + + counting efficiency & counter effects & @xmath261.3 + & gain shifts & + 2.2 + & resolution & -0.4,+0.5 + & rise time limits & @xmath260.7 + & lead and exposure times & @xmath260.6 + + backgrounds & neutrons & @xmath35 - 0.02 + & u and th & @xmath35 - 0.7 + & muons & @xmath35 - 0.7 + & internal radon & @xmath35 - 0.2 + & external radon & 0.0 + & other ge isotopes & @xmath35 - 0.7 + + total & & -3.2,+3.7 + @xmath36ge and @xmath32ge can be produced by background processes and their decays can be misinterpreted as @xmath13ge . @xmath36ge is of special concern as it decays solely by electron capture giving events that are identical to those from @xmath13ge . it is mainly produced by cosmic - ray muons and we can estimate the rate by scaling the predicted @xmath13ge rate of @xmath37 atoms / day in 60 t of ga @xcite by the measured cross section ratio for @xmath36ge to @xmath13ge production with 280-gev muons of @xmath38 @xcite . this gives @xmath39 atoms of @xmath36ge produced by muons per day in 50 t of ga . since the @xmath36ge half life is 271 d , considerably longer than the usual total counting time , these pulses will be distributed almost uniformly during the counting period , but there will be a small excess during the early part of counting . using typical values for exposure time of 30 d , ga mass of 50 t , extraction efficiency of 0.9 , counting efficiency of 0.6 in the sum of @xmath22 and @xmath21 peaks , 150 d of counting , and our average @xmath29 background rate of 0.175/day , simulations show that a true @xmath36ge production rate of 0.022/day leads to a false @xmath13ge production rate of 0.0085/day , which is equivalent to 0.05 snu . the isotope @xmath32ge can be produced from ga by several processes , including the interaction of cosmic rays , @xmath7b solar neutrinos , and protons on @xmath13ga , where the protons are secondaries either from @xmath40 particles from internal radioactivity or from neutrons from the surrounding rock . the total @xmath32ge production rate in 60 t of ga is estimated to be 0.21 atoms / day @xcite with an uncertainty of approximately 50% . because most @xmath32ge decays are accompanied by gamma rays which are rejected by the nai veto with an efficiency of 90% , and counting usually begins @xmath201.5 d after extraction , a time comparable with the @xmath32ge half life of 1.6 d , we calculate that only 0.045 @xmath32ge decays will be observed in a typical run @xcite , a factor of 100 less than the average number of detected @xmath13ge decays . the background effect of @xmath32ge is thus no more than 0.7 snu . these estimates of the counting rate of @xmath36ge and @xmath32ge can be checked by searching for these isotopes in the large set of @xmath13ge counting data that we now have from solar runs . the procedures and efficiencies for selecting candidate events are described in ref . the @xmath36ge production rate determined in this way is @xmath41 atoms / day in 60 t of ga . the central value is a factor of 7 more than the estimate given above , but within the large error range the rates determined by the two methods are in agreement . since the result from the cosmic - ray production rate of @xmath13ge and the cross section ratio has the smaller uncertainty , we use it for our estimate of the @xmath36ge background . there is some concern though that the predicted cosmic - ray production rate of @xmath13ge may be underestimated ; a more direct measurement would be desirable and this is considered briefly in @xcite . a similar search for @xmath32ge events shows that the production rate in 60 t of gallium is less than 0.49 atoms / day , in agreement with the value given above . the global best fit capture rate for all data from january 1990 to december 2001 ( 92 runs and 158 separate counting sets ) is @xmath42 snu , where the uncertainty is statistical only . in the windows that define the @xmath22 and @xmath21 peaks there are 1723 counts with 406.4 assigned by time analysis to @xmath13ge ( the total counting live time is 29.5 years ) . if one considers the @xmath22-peak and @xmath21-peak data independently , the results are @xmath43 snu and @xmath44 snu , respectively . the agreement between the two peaks serves as a strong check on the robustness of the event selection criteria . the total systematic uncertainty is determined by adding in quadrature all the contributions given in table [ systematics ] . our overall result is thus @xmath45 snu . for comparison , the latest result of the gno experiment ( including gallex ) is @xmath46 snu @xcite . if we combine the sage statistical and systematic uncertainties in quadrature , the result is @xmath47 snu . the validity of this result relies on the ability to chemically remove with a well known efficiency a few atoms of @xmath13ge produced by neutrino interactions from @xmath48 atoms of ga . to measure this efficiency about 350 @xmath18 g of stable ge carrier is added to the ga at the beginning of each exposure , but even after this addition , the separation factor of ge from ga is still 1 atom in 10@xmath49 . we have performed several auxiliary measurements which confirmed that the technology of our experiment has the capability to extract @xmath13ge at this level . an initial test was carried out in which ge carrier doped with a known number of @xmath13ge atoms was added to a reactor holding 7 t of ga . three successive extractions were made , and the number of @xmath13ge atoms in each extraction was determined by counting . the results @xcite showed that the extraction efficiency for the natural ge carrier and the @xmath13ge atoms did not differ . a second experiment addressed the concern that the @xmath13ge atom from inverse beta decay may be created in an excited or ionized state which results in the @xmath13ge being tied up in a chemical form from which we can not efficiently extract . a set of measurements designed to test directly this question was carried out by observing the beta decay of radioactive ga isotopes in liquid ga . these measurements @xcite showed that the expected isotopes are formed in the amounts anticipated at the 10% level . a test of all the experimental procedures including the chemical extraction , counting , and the analysis technique was performed using a 19.1 pbq ( 517 kci ) @xmath50cr neutrino source . the result , expressed as the ratio of the measured @xmath13ge production rate to that expected due to the source strength , is @xmath51 @xcite . this value provides strong verification that the experimental efficiencies are as claimed and validates the fundamental assumption in radiochemical experiments that the extraction efficiency of atoms produced by neutrino interactions is the same as that of the natural carrier . and @xmath21 peaks ( except may 1996 ) . the live time is 1169.9 days . the approximate expected location of the @xmath13ge @xmath22 and @xmath21 peaks as predicted by calibrations is shown darkened . lower panel shows the same histogram for all events that occurred during an equal live time interval beginning at day 100 after the time of extraction.,width=324 ] direct visual evidence that we are really observing @xmath13ge is in fig . [ 2d_hist ] which shows all events that survive the time cuts and that do not have a nai coincidence . the expected location of the @xmath13ge @xmath22 and @xmath21 peaks is shown darkened . these peaks are apparent in the upper panel but missing in the lower panel because the @xmath13ge has decayed away . and @xmath21 peaks . the solid line is a fit to the data points with the 11.4-day half - life of @xmath13ge plus a constant background . the vertical error bar on each point is proportional to the square root of the number of counts and is shown only to give the scale of the error . the horizontal error bar is @xmath265 d , equal to the 10-day bin size.,width=324 ] a major analysis hypothesis is that the time sequence of observed events for each run consists of the superposition of events from the decay of a fixed number of @xmath13ge atoms plus background events which occur at a constant rate . the quantity @xmath52 and the goodness of fit probability inferred from it provide a quantitative measure of how well the data fit this hypothesis . these numbers are evaluated for each data set and given in table [ recentruns ] . there are occasional runs with low probability of occurrence , but no more of these are observed than are expected due to normal statistical variation . this method can also be used to determine the goodness of fit for the combined time sequence of all @xmath22 plus @xmath21 events from any combination of runs . for all sage runs this test yields @xmath53 , with a goodness of fit probability of @xmath54 . a visual indication of the good quality of this fit is provided in fig . [ countrate ] which shows the count rate for all events in the @xmath22 and @xmath21 peaks vs time after extraction . an additional quantitative indication that @xmath13ge is being counted can be obtained by allowing the decay constant during counting to be a free variable in the maximum likelihood fit , along with the combined @xmath13ge production rate and all the background rates . the best fit half - life to all selected events in both @xmath22 and @xmath21 peaks is then @xmath55 days , in agreement with the measured value @xcite of 11.43 days . another analysis hypothesis is that the rate of @xmath13ge production is constant in time . by examination of fig . [ ratevstime ] , it is apparent that , within the large statistical uncertainty for each run , there is no substantial long - term deviation from constancy . another way to consider this question is to use the cumulative distribution function of the production rate @xmath56 , defined as the fraction of data sets whose production rate is less than @xmath57 . figure [ prod ] shows this distribution for all data sets and the expected distribution from simulation , assuming a constant production rate of 70.8 snu . the two curves parallel each other closely and can be compared by calculating the @xmath52 test statistic @xcite . this gives @xmath52 = 0.337 whose probability is 11% . l @ d d d d r @ d @ c d & & & + exposure & & & & & & & + interval & & & & & & & ( % ) + 1990 & 5 & 43 & 4.9 & 43 & 2 & 78 & 0.260 & 9 + 1991 & 6 & 59 & 25.5 & 112 & 82 & 145 & 0.120 & 17 + 1992 & 13 & 145 & 39.8 & 76 & 59 & 95 & 0.047 & 68 + 1993 & 15 & 97 & 33.2 & 84 & 62 & 105 & 0.199 & 6 + 1994 & 10 & 155 & 24.1 & 73 & 51 & 98 & 0.027 & 95 + 1995 & 13 & 210 & 37.7 & 102 & 77 & 129 & 0.041 & 82 + 1996 & 10 & 120 & 19.3 & 55 & 34 & 78 & 0.067 & 51 + 1997 & 16 & 183 & 35.7 & 62 & 48 & 78 & 0.057 & 62 + 1998 & 12 & 189 & 26.7 & 56 & 39 & 75 & 0.064 & 60 + 1999 & 14 & 114 & 40.8 & 87 & 66 & 110 & 0.068 & 33 + 2000 & 22 & 235 & 62.2 & 67 & 55 & 80 & 0.102 & 29 + 2001 & 22 & 173 & 64.4 & 65 & 55 & 76 & 0.050 & 70 + january & 11 & 129 & 24.8 & 58 & 37 & 80 & 0.082 & 35 + february & 12 & 101 & 25.5 & 60 & 44 & 77 & 0.045 & 74 + march & 9 & 129 & 34.5 & 102 & 79 & 127 & 0.043 & 78 + april & 9 & 80 & 16.9 & 54 & 37 & 73 & 0.072 & 39 + may & 12 & 114 & 34.7 & 75 & 59 & 94 & 0.051 & 62 + june & 11 & 101 & 33.6 & 79 & 58 & 102 & 0.175 & 5 + july & 15 & 176 & 26.6 & 52 & 37 & 69 & 0.091 & 35 + august & 15 & 161 & 38.7 & 78 & 60 & 96 & 0.058 & 51 + september & 20 & 220 & 48.4 & 68 & 54 & 84 & 0.035 & 91 + october & 17 & 169 & 40.3 & 73 & 57 & 91 & 0.080 & 45 + november & 15 & 196 & 37.5 & 59 & 44 & 75 & 0.034 & 82 + december & 12 & 147 & 46.4 & 105 & 84 & 127 & 0.040 & 89 + january+february & 23 & 230 & 50.5 & 59 & 46 & 73 & 0.095 & 34 + march+april & 18 & 209 & 49.2 & 75 & 61 & 91 & 0.026 & > 99 + may+june & 23 & 215 & 68.0 & 77 & 63 & 91 & 0.111 & 10 + july+august & 30 & 337 & 65.4 & 65 & 53 & 78 & 0.075 & 50 + september+october & 37 & 389 & 88.7 & 70 & 59 & 82 & 0.041 & 85 + november+december & 27 & 343 & 84.3 & 78 & 66 & 91 & 0.042 & 85 + february+march & 21 & 230 & 58.8 & 77 & 63 & 91 & 0.037 & 84 + april+may & 21 & 194 & 50.8 & 66 & 54 & 79 & 0.049 & 60 + june+july & 26 & 277 & 58.7 & 63 & 50 & 77 & 0.081 & 42 + august+september & 35 & 381 & 87.1 & 72 & 61 & 84 & 0.043 & 84 + october+november & 32 & 365 & 78.0 & 66 & 54 & 78 & 0.043 & 84 + december+january & 23 & 276 & 73.6 & 84 & 70 & 99 & 0.059 & 65 + february+november & 27 & 297 & 63.1 & 59 & 48 & 71 & 0.017 & 99 + march+october & 26 & 298 & 75.1 & 84 & 71 & 99 & 0.062 & 66 + april+september & 29 & 300 & 64.2 & 63 & 52 & 75 & 0.043 & 86 + may+august & 27 & 275 & 73.3 & 77 & 64 & 89 & 0.045 & 75 + neutrino oscillations can give a seasonal variation of the capture rate for some values of the mass and mixing angle parameters @xcite . other phenomena can also yield temporal variations ( see , e.g. , @xcite ) . we thus give in table [ sbyyears ] the results of combining the sage runs in various ways , monthly , bimonthly , and yearly . there is no compelling evidence for a temporal variation in any of these data divisions . the yearly results are plotted in fig . [ yearly ] which shows that the rate has been more or less constant during the data taking period . considering only the statistical errors , a @xmath58 test against the hypothesis of the constant rate of 70.8 snu yields @xmath59 , which , with 11 degrees of freedom , has a probability of 83% . day - night effects can also produce large differences in the capture rate between winter and summmer ( see , e.g. , @xcite ) . defining summer as the @xmath60-year interval centered on 21 june and winter as the rest of the year , the winter minus summer difference in sage capture rate is @xmath61 snu where the stated error is only statistical . in our method of data analysis @xcite we remove the known change in rate caused by the earth s orbital eccentricity , and thus in the absence of neutrino oscillations the expected value for @xmath62 is zero . ( if , rather than using the above solstice - based definition , we define summer as the @xmath60-year interval centered on 5 july , the time of the aphelion , then @xmath63 snu , also consistent with a null rate . ) one of the main purposes of the ga experiment is to provide information that leads to the experimental determination of the flux of @xmath2 neutrinos at the earth . in this section we indicate the present state of this measurement where we use only information from the various solar neutrino experiments and assume that their reduced capture rate compared to ssm predictions is due to neutrino oscillations . by combining the results of sage , gallex , and gno , the capture rate in the ga experiment is approximately @xmath65 snu . this rate is the sum of the rates from all the components of the solar neutrino flux , which we denote by [ @xmath2+@xmath66be+cno+@xmath67b@xmath68ga , exp ] , where `` exp '' indicates that this is a measured rate . ( we ignore the @xmath69 contribution , as it is a negligible 0.05% of the total in the ssm calculation @xcite . ) the only one of these flux components that is known is the @xmath7b flux , measured by sno to be [ @xmath70b@xmath68sno , exp ] = @xmath71 electron neutrinos/(@xmath4-s ) @xcite . since the shape of the @xmath7b spectrum in sno and superkamiokande is very close to that of the ssm above 5 mev and the cross section for ga rises steeply with energy , we can use the sno flux and the cross section for @xmath7b neutrinos with ssm shape ( @xmath72 @xmath4 [ see @xcite for the cross sections used here ] ) to conclude that the @xmath7b contribution to the ga experiment is [ @xmath70b@xmath68ga , exp ] = @xmath73 snu . subtracting this measured value from the total ga rate gives [ @xmath74be+cno+@xmath75ga , exp ] = @xmath76 snu . the measured capture rate in the cl experiment is [ @xmath66be+@xmath70b+cno+@xmath75cl , exp ] = @xmath77 snu @xcite . ( the @xmath69 contribution will again be neglected as it is only 0.5% of the total in the ssm . ) since the cross section in cl is dominated by neutrinos above 5 mev , we can again use the sno flux and the cross section calculated for the ssm ( @xmath78 @xmath4 ) , and deduce that the contribution of @xmath7b to the cl experiment is [ @xmath70b@xmath68cl , exp ] = @xmath79 snu . subtracting this component from the total leaves [ @xmath66be+cno+@xmath75cl , exp ] = @xmath80 snu , all of which is due to neutrinos of medium energy . neutrino oscillations have the effect of introducing an energy - dependent survival factor to the fluxes predicted by the ssm . for the medium - energy neutrinos this factor for the cl experiment can be approximated by the ratio of the measured rate to the ssm prediction of [ @xmath66be+cno+@xmath75cl , ssm ] = @xmath81 snu . if we assume that the survival factor varies slowly with energy , we find it to be given by [ @xmath66be+cno+@xmath75cl , exp]/[@xmath66be+cno+@xmath75cl , ssm ] = @xmath82 . since the @xmath9be contribution dominates , and it is at a single energy , the error in this factor due to the assumption that it is the same for all of these flux components can be estimated by considering the contribution of the non-@xmath9be components to the total in the ssm , which is 36% . we thus increase the error from 0.17 to @xmath83 the relative contributions to the capture rate of the medium - energy neutrinos are about the same in ga as in cl ( 75% from @xmath9be in ga compared to 64% in cl ) . thus it is reasonable to apply the survival factor determined for cl to the ga experiment , i.e. , [ @xmath66be+cno+@xmath75ga , exp ] = @xmath84 [ @xmath66be+cno+@xmath75ga , ssm ] = @xmath85 snu . we subtract this contribution from the rate above and get the result for the measured @xmath2 rate in the ga experiment [ @xmath86ga , exp ] = [ @xmath74be+cno+@xmath75ga , exp ] - [ @xmath66be+cno+@xmath75ga , exp ] = @xmath87 snu . since the cross section does not change appreciably over the narrow range of ga response to the @xmath2 neutrinos , ( 0.230.42 ) mev , we divide the capture rate by the ssm cross section for electron neutrinos of @xmath88 @xmath4 and obtain the measured electron neutrino @xmath2 flux at earth of @xmath89/(@xmath4-s ) . alternatively , if we divide the capture rate by the cross section multiplied by the survival probability for @xmath2 neutrinos , which is 60% for the favored lma solution @xcite assuming no transitions to sterile neutrino flavors , we receive the rate of @xmath2 neutrino emission in the solar fusion reaction of @xmath90/(@xmath4-s ) , in agreement with the ssm calculation of @xmath6/(@xmath4-s ) @xcite . the major component of the error in the @xmath2 flux measurement is due to the poor knowledge of the energy - dependent survival factor . several approximations were made in arriving at this value , whose nature can not be easily quantified , so perhaps the error given here is somewhat underestimated . nonetheless , as mentioned in the introduction , it will be possible to reduce the error in this flux greatly when the region of mass and mixing angle parameters is better determined , as should be done by the kamland experiment , and when the @xmath9be flux is directly measured , as anticipated by borexino @xcite . the dominant error should eventually be due to the inaccuracy of the ga measurement itself , and hence we are seeking to reduce our statistical and systematic errors . the methods and analysis of the sage experiment have been summarized and results for 92 extractions during 12 years of operation from january 1990 through december 2001 have been presented . the measured capture rate is @xmath91 snu where the uncertainty is statistical only . analysis of all known systematic effects indicates that the total systematic uncertainty is @xmath92 snu , less than the statistical error . finally we have examined the counting data and shown that there is good evidence that @xmath13ge is being counted , that the counting data fit the analysis hypotheses , and that the counting data are self - consistent . the sage result of 70.8 snu represents 55% of ssm predictions @xcite . given the extensive systematic checks and auxiliary measurements that have been performed , especially the @xmath50cr neutrino source experiment @xcite , this 6.0@xmath93 reduction of the solar neutrino capture rate compared to ssm predictions is very strong evidence that the solar neutrino spectrum below 2 mev is significantly depleted , as has been proven for the @xmath7b flux by the cl , kamiokande , and sno experiments . the sage result is even somewhat below the astrophysical minimum capture rate of @xmath94 snu @xcite . several recent phenomenology papers ( see , e.g. , @xcite ) discuss the combined fit of all solar neutrino experiments . their conclusion is that the electron neutrino oscillates into other species and the best fit is to the lma region of mikheyev - smirnov - wolfenstein ( msw ) oscillations . to more precisely determine the oscillation parameters in the solar sector will require additional data , especially from experiments sensitive to the low - energy neutrinos . in this vein , sage continues to perform regular solar neutrino extractions every four weeks with @xmath2050 t of ga and will continue to reduce its statistical and systematic uncertainties . we thank j. n. bahcall , m. baldo - ceolin , g. t. garvey , w. haxton , v. a. kuzmin , v. v. kuzminov , v. a. matveev , l. b. okun , r. g. h. robertson , v. a. rubakov , a. yu . smirnov , a. n. tavkhelidze , and many members of gallex and gno for their continued interest and for stimulating discussions . we greatly appreciate the work of our prior collaborators o. l. anosov , o. v. bychuk , m. l. cherry , r. davis , jr . , i. i. knyshenko , v. n. kornoukhov , r. t. kouzes , k. lande , a. v. ostrinsky , d. l. wark , p. w. wildenhain , and yu . i. zakharov . we acknowledge the support of the russian academy of sciences , the institute for nuclear research of the russian academy of sciences , the russian ministry of science and technology , the russian foundation of fundamental research under grant no . 96 - 02 - 18399 , the division of nuclear physics of the u.s . department of energy under grant no . defg03 - 97er41020 , and the u.s . civilian research and development foundation under award no . rp2 - 159 . this research was made possible in part by grant no . m7f000 from the international science foundation and grant no . m7f300 from the international science foundation and the russian government . s. fukuda _ et al . , _ hep - ex/0103032 , phys . lett . * 86 * , 5651 ( 2001 ) ; hep - ex/0103033 , phys . lett . * 86 * , 5656 ( 2001 ) . q. r. ahmad _ et al . , _ , nucl - ph/0106015 , phys . lett . * 87 * , 071301 ( 2001 ) . b. t. cleveland , t. j. daily , r. davis , jr . , j. r. distel , k. lande , c. k. lee , and p. s. wildenhain , astrophys . j. * 496 * , 505 ( 1998 ) . j. n. abdurashitov , v. n. gavrin , s. v. girin , v. v. gorbachev , t. v. ibragimova , a. v. kalikhov , n. g. khairnasov , t. v. knodel , i. n. mirmov , a. a. shikhin , e. p. veretenkin , v. m. vermul , v. e. yants , g. t. zatsepin , t . j. bowles , w. a. teasdale , d. l. wark , m. l. cherry , j. s. nico , b. t. cleveland , r. davis , jr . , k. lande , and p. s. wildenhain , s. r. elliott and j. f. wilkerson , astro - ph/9907113 , phys . c * 60 * , 055801 ( 1999 ) . w. hampel , j. handt , g. heusser , j. kiko , t. kirsten , m. laubenstein , e. pernicka , w. rau , m. wojcik , y. zakharov , r. v. ammon , k. h. ebert , t. fritsch , e. henrich , l. stielglitz , f. weirich , m. balata , m. sann , f. x. hartmann , e. bellotti , c. cattadori , o. cremonesi , n. ferrari , e. fiorini , l. zanotti , m. altmann , f. v. feilitzsch , r. mbauer , s. wnninger , g. berthomieu , e. schatzmann , i. carmi , i. dostrovsky , c. bacci , p. belli , r. bernabei , s. dangelo , l. paoluzi , m. cribier , j. rich , m. spiro , c. tao , d. vignaud , j. boger , r. l.hahn , j. k. rowley , r. w. stoenner , and j. weneser , phys . b * 447 * , 127 ( 1999 ) . y. fukuda _ et al . , _ phys . * 77 * , 1683 ( 1996 ) . see www.sns.ias.edu/@xmath20jnb/meetings/lownu/index.html and www-sk.icrr.u-tokyo.ac.jp/lownu/index.html . v. a. kuzmin , zh . eksp . teor . fiz . * 49 * , 1532 ( 1965 ) ( sov . jetp * 22 * , 1051 ( 1966 ) ) . j. n. bahcall , m. h. pinsonneault , and s. basu , astro - ph/0010346 , astrophys . j. * 555 * , 990 ( 2001 ) . a. s. brun , s. turck - chize , and p. morel , astro - ph/9806272 , astrophys . j. * 506 * , 913 ( 1998 ) . a. piepke , nucl . b ( proc . suppl . ) * 91 * , 99 ( 2001 ) . g. alimonti _ et al . , _ ( borexino collaboration ) , hep - ex/0012030 , astroparticle physics * 16 * , 205 ( 2002 ) . m. altmann , m. balata , p. belli , e. bellotti , r. bernabei , e. burkert , c. cattadori , g. cerichelli , m chiarini , m. cribier , s. dangelo , g. del re , k. h. ebert , f. v. feilitzsch , n. ferrari , w. hampel , j. handt , e. henrich , g. heusser , j. kiko , t. kirsten , t. lachenmaier , j. lanfranchi , m. laubenstein , d. motta , w. rau , h. richter , s. wnninger , m. wojcik , l. zanotti , hep - ex/0006034 , phys . lett . * * 4**90 , 16 ( 2000 ) . v. n. gavrin , v. n. kornoukhov , and g. t. zatsepin , institute for nuclear research of the academy of sciences of the ussr report no . p-0690 , 1991 . v. n. gavrin , v. e. gurentsov , v. n. kornoukhov , a. m. pshukov , and a. a. shikhin , institute for nuclear research of the academy of sciences of the ussr report no . p-0698 , 1991 . j. n. abdurashitov , v. n. gavrin , a. v. kalikhov , v. l. matushko , a. a. shikhin , v. e. yants , and o. s. zaborskaya , to be published in proceedings of the xith int . school on particles and cosmology , baksan , russia , april 2001 . j. n. abdurashitov , e. l. faizov , v. n. gavrin , a. o. gusev , a. v. kalikhov , t. v. knodel , i. i. knyshenko , v. n. kornoukhov , i. n. mirmov , a. m. psukhov , a. m. shalagin , a. a. shikhin , p. v. timofeyev , e. p. veretenkin , v. m. vermul , g. t. zatsepin , t. j. bowles , s. r. elliott , j. s. nico , w. a. teasdale , d. l. wark , j. f. wilkerson , b. t. cleveland , t. daily , r. davis , k. lande , c. k. lee , p. s. wildenhain , m. l. cherry , and r. t. kouzes , phys . lett . b * 328 * , 234 ( 1994 ) . s. r. elliott , nucl . instrum . methods phys . res . a 290 , 158 ( 1990 ) . b. t. cleveland , nucl . instrum . methods phys . res . 214 , 451 ( 1983 ) . b. t. cleveland , nucl . instrum . methods phys . res . a * 416 * , 405 ( 1998 ) . v. n. gavrin , proceedings of the xixth international conf . on neutrino physics and astrophysics , sudbury , canada , 1621 june 2000 , ed . by j. law , r. w. ollerhead , and j. j. simpson , nucl . b ( proc . suppl . ) * 91 * , 36 ( 2000 ) . v. n. gavrin , v. n. kornoukhov , and v. e. yants , institute for nuclear research of the academy of sciences of the ussr report no . p-0703 , 1991 . v. n. gavrin , s. n. danshin , a. v. kopylov , and v. e. cherekhovsky , institute for nuclear research of the academy of sciences of the ussr report no . p-0494 , 1986 . charles evans associates report ( unpublished ) . v. n. gavrin , v. v. gorbachev , and i. n. mirmov , yad . phys . ( subm . ) . v. n. gavrin and v. v. gorbachev , yad . ( subm . ) . v. n. gavrin and yu . i. zacharov , institute for nuclear research of the academy of sciences of the ussr report no . p-0560 , 1987 . m. cribier , b. pichard , j. rich , j. p. soirat , m. spiro , th . stolarczyk , c. tao , d. vignaud , p. anselmann , a. lenzing , c. schlosser , r. wink , and j. k. rowley , astropart . * 6 * , 129 ( 1997 ) . v. n. gavrin , v. v. gorbachev , t. v. ibragimova , and b. t. cleveland , yad . phys . * 65 * , 1309 ( 2002 ) ; phys . atomic nuclei * 65 * , 1276 ( 2002 ) . j. n. abdurashitov , v. n. gavrin , s. v. girin , v. v. gorbachev , t. v. ibragimova , a. v. kalikhov , n. g. khairnasov , t. v. knodel , v. n. kornoukhov , i. n. mirmov , a. a. shikhin , e. p. veretenkin , v. m. vermul , v. e. yants , g. t. zatsepin , t. j. bowles , j. s. nico , w. a. teasdale , d. l. wark , m. l. cherry , v. n. karaulov , v. l. levitin , v. i. maev , p. i. nazarenko , v. s. shkolnik , n. v. skorikov , b. t. cleveland , t. daily , r. davis , jr . , k. lande , c. k. lee , p. s. wildenhain , yu . s. khomyakov , a. v. zvonarev , s. r. elliott , and j. f. wilkerson , phys . lett . * 77 * , 4708 ( 1996 ) . j. n. abdurashitov , v. n. gavrin , s. v. girin , v. v. gorbachev , t. v. ibragimova , a. v. kalikhov , n. g. khairnasov , t. v. knodel , v. n. kornoukhov , i. n. mirmov , a. a. shikhin , e. p. veretenkin , v. m. vermul , v. e. yants , g. t. zatsepin , yu . s. khomyakov , a. v. zvonarev , t. j. bowles , j. s. nico , w. a. teasdale , d. l. wark , m. l. cherry , v. n. karaulov , v. l. levitin , v. i. maev , p. i. nazarenko , v. s. shkolnik , n. v. skorikov , b. t. cleveland , t. daily , r. davis , jr . , k. lande , c. k. lee , p. s. wildenhain , s. r. elliott , and j. f. wilkerson , hep - ph/9803418 , phys . rev . c. * 59 * , 2246 ( 1999 ) . w. hampel and l. remsberg , phys . c * 31 * , 666 ( 1985 ) . v. berezinsky , g. fiorentini , and m. lissia , hep - ph/9904225 , astropart . * 12 * , 299 ( 2000 ) . g. l. fogli , e. lisi , d. montanino , and a. palazzo , hep - ph/9910387 , phys . d * 61 * , 073009 ( 2000 ) . j. pulido and e. kh . akhmedov , hep - ph/9907399 , astropart . * 13 * , 227 ( 2000 ) . p. a. sturrock and j. d. scargle , astro - ph/0011228 , astrophys . j. * 550 * , l101-l104 ( 2001 ) . j. n. bahcall , hep - ph/0108148 , phys . c * 65 * 025801 ( 2002 ) . j. n. bahcall , hep - ph/9710491 , phys . c * 56 * , 3391 ( 1997 ) . j. n. bahcall _ et al . , _ nucl - th/9601044 , phys . c * 54 * , 411 ( 1996 ) . j. n. bahcall , m. c. gonzalez - garcia , and c. pea - garay , hep - ph/0111150 , jhep * 0204 * , 007 ( 2002 ) . after the appearance of our preprint we learned that similar arguments to ours have been used to make predictions for what will be measured by borexino . see s. m. bilenky , t. lachenmaier , w. potzel , and f. v. feilitzsch , hep - ph/0109200 , phys . lett . * 533 * , 191 ( 2002 ) . j. n. bahcall , m. c. gonzalez - garcia , and c. pea - garay , hep - ph/0106258 , jhep * 0108 * , 014 ( 2001 ) . p. i. krastev and a. yu . smirnov , hep - ph/0108177 , phys . d * 65 * , 073022 ( 2002 ) . gonzalez - garcia , m. maltoni , and c. pea - garay , hep - ph/0108073 .