Datasets:

scientific_papers

like 122

auger - electron spectroscopy ( aes ) and the complementary appearance - potential spectroscopy ( aps ) have become valuable tools for investigating the electronic structure of solids and solid surfaces @xcite . they represent highly element specific and non - destructive methods with a comparatively simple experimental set up . the auger line shape from a core - valence - valence ( cvv ) process yields information on the occupied part of the valence band , while the aps provides insight into the unoccupied valence states . however , much effort has been spent on the detailed interpretation of the spectra . + lander @xcite suggested that the spectrum obtained by aes ( aps ) is given as the self - convolution of the occupied ( unoccupied ) valence density of states ( dos ) . on the other hand , powell @xcite discovered the cvv auger line shape of ag to behave `` anomalously '' in the sense of lander s self - convolution model . these anomalous features are by now well known to be caused by correlation effects dominating the electronic properties of various solids . therefore , aes ( aps ) seems to be a useful technique to study electron - correlation effects , but it is doubtful whether it is able to compete with one - particle spectroscopies , such as photoemission ( inverse photoemission ) , in deriving the dos by deconvolution . + in the theoretical treatment of the cvv auger process , there are mainly two problems . the first one is to take into account the correlation effects . here one may distinguish between the _ direct _ and _ indirect _ correlations . the direct correlations describe the correlations of those electrons which participate in the auger process . they are responsible for the most prominent effects in the auger line shape as compared to the self - convolution . on the other hand , the indirect correlations among the electrons in the rest system manifest themselves in the quasi - particle density of states ( qdos ) as a renormalisation of the one - particle dos . + the second problem is the calculation of the transition - matrix elements for the auger process as well as the scattering of the outgoing auger electron ( cf . refs . ) . these effects will ( slightly ) modify the bare line shape and may become important for a refined interpretation of experimental data . within the present paper , however , we set aside this second problem and like to concentrate on electron - correlation effects in aes from ferromagnetic 3d - transition metals . + within the framework of the single - band hubbard model , correlation effects can be treated exactly for systems with completely filled or empty bands , as was first shown by cini and sawatzky @xcite . the generalisation to the case of degenerate bands was introduced in ref . and further analysed in ref . , for example . these results may also be extended to include the core - valence interaction @xcite . + considering the more general case of partially filled bands introduces several complications concerning the indirect as well as the direct correlations . for the indirect valence - valence correlations there is a number of approximation schemes applicable to a multi - band hubbard model . a method which reproduces the experimentally observed curie temperature quite well , especially for ni , is the spectral density approach @xcite . other approaches are , for example , the generalisation of the single - band modified perturbation theory @xcite to the multi - band model @xcite and quantum - monte - carlo simulations @xcite in connection with the dynamical mean - field theory @xcite . however , these methods suffer from some necessary restrictions concerning the completeness of the coulomb - matrix . this is not the case for the fluctuation exchange @xcite and the hubbard i approximation @xcite , for example . for a more detailed discussion on the indirect valence - valence correlations see ref . . + for the treatment of the direct correlations , one has the exact - diagonalisation method @xcite for small systems and the equation - of - motion method @xcite with its in general uncontrolled termination of the hierarchy of the equations of motion . another approximate solution is the valence - valence ladder ( vv ladder ) approximation @xcite and its generalisation to include the core - valence interaction @xcite ( cvv ladder ) . in particular one has to account for the broken translational symmetry in the initial state of aes , caused by the presence of the core hole and its screening due to the valence electrons . in the final state this interaction is responsible for the sudden response of the valence electrons due to the destruction of the core hole . in the limit of completely filled or empty bands the ladder approximations recover the above - mentioned exact solution . here the interaction strength is taken as the control parameter , which is correct in the weak - coupling regime . we are aware that this method has restrictions for the 3d - transition metals . however , in this work we prefer a common treatment of one - particle spectroscopies ( photoemission and inverse photoemission ) and two - particle spectroscopies like aes . to be concrete , we will use the second - order perturbation theory around the hartree - fock solution @xcite for the indirect valence - valence correlations . the direct correlations will be treated by applying two different methods , i.e. the vv ladder approximation and the first - order perturbation theory in the valence - valence and core - valence interaction . + within this approach it is possible to include a realistic one - particle input taken from tight - binding band - structure calculations . we do not only account for the degeneracy of the 3d - band but also for the hybridisation with the 4s and 4p states . the theory is formulated and evaluated for a non - orthogonal basis set where the states can be distinguished by the angular momentum quantum number and the cubic harmonic index . this facilitates the interpretation of the resulting spectra . furthermore , we do not restrict ourselves to correlations among the final - state holes only and include core - hole effects from the very beginning . this implies the necessity for a proper treatment of the initial state where the core - hole screening breaks the translational symmetry . the theory is implemented numerically and evaluated for ferromagnetic ni . the paper is organised as follows . in the next section we will introduce the model under consideration . in section [ sec : auger ] we give the expression for the auger intensity . section [ sec : indirect ] concentrates on the indirect and section [ sec : direct ] on the direct correlations . finally , section [ sec : sum ] concludes the paper . some details concerning the non - orthogonal basis set are given in the appendix [ sec : nobs ] . the hamiltonian @xmath0 is decomposed into a one - particle part @xmath1 and an interaction part @xmath2 . @xmath3 is the operator for the particle number . the one - particle part describes non - interacting valence and core electrons : @xmath4 the index @xmath5 refers to lattice sites , @xmath6 is the spin - index ( @xmath7 ) , @xmath8 is the orbital index with angular momentum quantum number @xmath9 and cubic harmonic index @xmath10 . @xmath11 ( @xmath12 ) denotes the creation ( annihilation ) operator of a valence electron at the lattice site @xmath5 with spin @xmath6 and orbital index @xmath13 while @xmath14 ( @xmath15 ) creates ( annihilates ) a core electron . the core states are assumed to be non - degenerate and dispersionless with the one - particle energy @xmath16 well below the chemical potential @xmath17 . the hopping integrals @xmath18 @xmath19 ( @xmath20 denotes the hamiltonian of the tight - binding band - structure calculation ) are taken from ref . as well as the overlap integrals @xmath21 @xmath22 @xmath18 and @xmath21 refer to a non - orthogonal basis set ( see appendix [ sec : nobs ] ) . contrary to an orthonormal basis set ( where the overlap matrix is replaced by @xmath23 ) , the basis states under consideration can be characterised by the orbital index @xmath8 . the construction operators likewise refer to the non - orthogonal basis and satisfy the following anti - commutation rules : @xmath24_+ & = & 0\nonumber\\ \left[c_{il\sigma},c^\dag_{i^\prime l^\prime\sigma^\prime}\right]_+ & = & \left(s^{-1}\right)_{ii^\prime}^{ll^{\prime } } \delta_{\sigma\sigma^\prime}\ , . \label{eq : acr}\end{aligned}\ ] ] it should be noted that the action of the creation operator on the vacuum state @xmath25 , in general , does not yield @xmath26 ( see eq . ( [ eq : constr ] ) of the appendix [ sec : nobs ] ) . to describe the correlations among the valence electrons ( vv ) as well as the correlations between valence and core electrons ( cv ) the interaction consists of two parts @xmath27 here the occupation number operator for valence electrons is @xmath28 and for core electrons @xmath29 . assuming a strong screening of the coulomb interaction , the interaction part is taken to be purely local . @xmath30 are the on - site coulomb - matrix elements for the valence electrons . + the electronic structure of the 3d - transition metals may be understood considering mainly two types of electronic orbitals : the 4s- and 4p - states which form broad free - electron like bands . they should be well described by the band - structure calculation . the other group are the well localised 3d - states which in the solid form relatively narrow bands positioned around the fermi energy . the localised nature of the 3d - electrons gives rise to important dynamic 3d-3d correlation effects which are believed to be responsible e.g. for the magnetic behaviour of the 3d - transition metals . these correlations may not be adequately taken into account within a mean - field picture . we thus treat them separately . + exploiting atomic symmetries , one is able to express all remaining coulomb - matrix elements for the 3d - electrons in terms of three effective slater integrals @xcite ( @xmath31 , @xmath32 , @xmath33 ) only . these integrals are connected to averaged values for direct @xmath34 and exchange interaction terms @xmath35 for 3d - elements one has to a good accuracy the ratio @xmath36 @xcite to be that of free ions @xcite . @xmath37 and @xmath38 are treated as free parameters to be fixed by comparison with experimental results ( see section [ sec : indirect ] ) . + the cv interaction part is necessary to describe the core hole effects in aes . @xmath39 in equation ( [ eq : hi ] ) are the coulomb - matrix elements between the valence and the core electrons which can be fixed by assuming complete screening of the core hole by the valence electrons ( see section [ sec : indirect ] ) . + to avoid a double counting of interactions , we subtract the correction @xmath40 which is to a good approximation the hartree - fock part of the respective interaction term @xcite . the auger process can be divided into two subprocesses . the first one is the creation of a core hole with spin @xmath41 at the lattice site @xmath42 by absorbing an x - ray quantum for example . the second subprocess is the radiationless decay of the core hole via ejecting an auger electron with spin @xmath6 and momentum @xmath43 . provided that the life time of the core hole is large compared to typical relaxation times of the valence electrons in the presence of the core hole , the two subprocesses become independent from each other @xcite ( two - step model ) . this implies the absence of any decay term in the hamiltonian . consequently @xmath44 is a good quantum number , @xmath45_-=0 $ ] . we can concentrate on the second subprocess . within the two - step model the initial state for the auger transition process is the ground state within the subspace @xmath46 of the hilbert space @xmath47 that is built up by all many - body states with @xmath48 . to perform thermodynamic averages in practice , one has to take into account this restriction by introducing an additional lagrange parameter . the transition process itself is described by the transition operator @xcite @xmath49 @xmath43 and @xmath6 denote the quantum - numbers of the auger electron , @xmath41 the spin of the core state involved . the ( intra - atomic ) auger - matrix elements are given by @xmath50 where @xmath51 is the valence orbital , @xmath52 the one - particle wave - function of the auger - electron and @xmath53 the core state . the bar denotes complex conjugation . + following ref . we consider the ( retarded ) three - particle green function , relevant for aes , which is defined as ( with the abbreviation @xmath54 and @xmath55 ) @xmath56 with @xmath57 @xmath58 refers to zubarev green functions @xcite . the aes intensity is then mainly given by the three - particle spectral - density @xmath59 : @xmath60 here @xmath61 is the dispersion of the auger - electron . in general the three - particle green function will be a ( complicated ) functional of one - particle green functions . this functional represents the direct correlations . in the following we concentrate on the indirect correlations first , i.e. on the determination of the relevant one - particle green functions we consider the ( retarded ) one - particle valence - band green - function @xmath62 . using a matrix notation with respect to the orbital index @xmath8 @xmath63 and defining a lattice - fourier transformation @xmath64 where @xmath65 is the number of lattice sites , we get dyson s equation in the form @xmath66 here the matrices @xmath67 and @xmath68 are the fourier - transformed hopping and overlap integrals of eqs . ( [ eq : hopping ] ) and ( [ eq : overlap ] ) , respectively . @xmath69 is the self - energy . the one - particle spectral density is given by @xmath70 the on - site terms of the spectral density are diagonal in the orbital index as a consequence of lattice symmetries , and we have for the orbital resolved qdos @xmath71 where we dropped the site index . the total qdos is obtained via @xmath72 to calculate the self - energy , we use a standard approximation scheme , the second order perturbation theory around the hartree - fock solution @xcite ( sopt - hf ) . it is known from the single - band hubbard model that the non - local terms of the sopt - hf self - energy rapidly decrease with increasing number of shells taken into account @xcite . furthermore , due to the band degeneracy there is a much weaker @xmath43-dependence of the sopt - hf self - energy compared to the single - band case @xcite . we may thus employ the local approximation @xmath73 as for the on - site green function , lattice symmetries require the on - site self - energy to be diagonal in the orbital index . the hartree - fock ( hf ) contribution to the self - energy reads as @xmath74 where @xmath75 denotes the expectation value in the full model and @xmath76 the expectation value of the band - structure calculation that stems from the double counting correction in equation ( [ eq : hi ] ) . approximating the self - energy by @xmath77 , corresponds to the lda+@xmath37 approach @xcite . here we additionally include the next order in the interaction . the second - order contribution ( soc ) to the self - energy reads as @xmath78 with the fermi function @xmath79 and where the local hf spectral - density @xmath80 is obtained by using the hf self - energy ( [ eq : hfse ] ) in equation ( [ eq : vgf ] ) . the sopt - hf self - energy @xmath81 determines the full green function via equation ( [ eq : vgf ] ) . let us now focus on the core hole screening in the initial state for aes . the cv interaction and the presence of the core hole introduce an additional ( hartree - like ) term @xmath82 to the valence - band self - energy which breaks the translational symmetry . this term represents the core - hole potential at the lattice site @xmath42 where the core hole was created ( the superscript `` e '' indicates averaging in the restricted hilbert - space @xmath46 ; see section [ sec : auger ] ) . it is responsible for the screening . the valence - band self - energy in the presence of the core hole then reads @xmath83 @xmath84 incorporates the vv - correlation effects and has the same structure as the self - energy ( [ eq : vvse ] ) for the translational invariant system . but in contrast to its translational invariant counterpart , @xmath85 is defined in terms of the green functions in the presence of the core hole . + the valence band green function @xmath86 in the presence of the core hole can be obtained by using dyson s equation in the form @xmath87 in general @xmath88 for a certain number of shells around the core - hole site because the vv - correlation effects depend on the occupation numbers , which as a consequence of the screening locally differ from the translational invariant ones . in the following we assume complete screening , i.e. charge neutrality at the site @xmath42 , which is reasonable especially for 3d - transition metals because the screening time scale is small compared to the life - time of the core hole @xcite . this implies @xmath89 to be small for all sites @xmath5 except for @xmath90 . neglecting the terms for @xmath91 one can solve equation ( [ eq : vgfcd ] ) @xmath92 for @xmath93 one obtains the local screened green - function at the core - hole site : @xmath94 the assumption of complete screening will be utilised as condition to fix the cv - interaction parameter , which is taken to be the same for s- , p- and d - orbitals ( @xmath95 ) . to take into account the cv interaction for the core green function @xmath96 one may calculate the core self - energy @xmath97 using e.g. the sopt - hf in the same way as for the valence - band green function . on the other hand it is believed that the core states are influenced by other and presumably more important effects , such as life - time effects @xcite . in fact , the core spectral density obtained within sopt - hf turns out to be dominated by a delta - peak that is shifted by about 1ev below @xmath98 . this does not affect the auger line shape . therefore , we assume for convenience the core self - energy to be zero . the spectral density becomes @xmath99 before we discuss the results for fcc - ni , we like to make a short remark concerning the numerical evaluation of the theory . the @xmath43-sum for the local green function in equation ( [ eq : vgf ] ) was performed on a mesh of 240 @xmath43-points within the irreducible part of the brillouin zone using the tetrahedron method @xcite generalised to complex band structures , similar to that presented in ref . . the evaluation of the total qdos ( [ eq : tdos ] ) as well as the qdos in the presence of the core hole was done in @xmath43-space . for the latter , equation ( [ eq : vgfcd ] ) has to be used to perform the fourier transformation . + the effective slater integrals or , equivalently , the averaged direct and exchange interaction parameters are chosen as @xmath100ev and @xmath101ev . this leads to a calculated magnetic moment per atom of @xmath102@xmath103 at @xmath104k which is the same as the measured moment @xcite . with the ratio @xmath105 we assume a typical value for the late 3d transition metals . the values given in the literature , for instance @xmath106ev , @xmath107ev @xcite and @xmath108ev , @xmath109ev @xcite , are of the same order of magnitude but slightly overestimate the magnetic moment within the present theory . + the `` free '' dos , used as starting point for our theory is shown on the left of figure [ fig : qdosju02 ] ( thin dotted line ) and corresponds to tight binding band structure calculations @xcite for paramagnetic fcc - ni . + the left - hand side of figure [ fig : qdosju02 ] shows the qdos per atom for the model parameters given above . as is known from the experiment , ni is a strong ferromagnet , i.e. the majority - spin states are fully occupied . the renormalisation effects of the a priori uncorrelated s- and p - states seen in figure [ fig : qdosju02 ] can be traced back to the hybridisation with the d - states . + taking into account the presence of the core hole and following the procedure to fix the cv interaction parameter as described above ( charge neutrality ) , leads to @xmath110ev . the corresponding `` screened '' qdos is plotted on the right - hand side of figure [ fig : qdosju02 ] . the structure has changed remarkably . spectral weight of the d - electrons from the upper band edge is transferred to lower energies , the s- and p - states are more populated , too . the screened magnetic moment at the site @xmath42 ( @xmath111 at @xmath104k ) is considerably decreased since the local occupation is increased ( see also figure [ fig : mcju02 ] ) . + the left side of figure [ fig : secju02 ] shows the self - energy @xmath112 . within an energy range of about 1ev above and below @xmath113ev one has @xmath114 . thus we have well - defined quasi - particles at the fermi energy and their weight @xmath115 is 0.887 for the @xmath116-@xmath117-states and 0.893 for @xmath118-@xmath119-states . for @xmath120-states we find 0.878 and 0.883 , respectively . for energies above @xmath121ev , where one finds clearly distinguishable structures , a significant band - narrowing caused by the real part of the self - energy is observed . while the imaginary part of the self - energy leads to strong damping effects in the qdos ( figure [ fig : qdosju02 ] , l.h.s . ) for energies below @xmath121ev . about @xmath122ev below the fermi energy where one expects the `` ni-6-ev - satellite '' @xcite we find the largest damping effects . however , we do not find the correlation - induced 6evsatellite . this is not surprising by applying the sopt - hf or any other finite - order perturbational approach @xcite . for different interaction parameters ( larger @xmath37 , smaller @xmath38 ) a small shoulder in the qdos of the d - states is visible as was also mentioned in ref . . + on the right - hand side of figure [ fig : secju02 ] the self - energy in the presence of the core hole @xmath123 is plotted . again there are well - defined quasi - particles , but with an enhanced weight compared to the case where the core hole is absent ( @xmath124 , @xmath125 , @xmath126 , @xmath127 ) . the screened case behaves less correlated than the unscreened case since here one is closer to the limit of completely filled bands . + finally , we show the local magnetic moment per atom as a function of temperature in figure [ fig : mcju02 ] . the magnetisation curves ( figure [ fig : mcju02 ] , l.h.s . ) have a brillouin - function - like form , except for the @xmath128-magnetisation which shows up a maximum at @xmath129k . this can be traced back to a transfer of charge carriers from the @xmath120-orbitals to the @xmath116-orbitals with increasing temperature . because ni is a strong ferromagnet the charge - carrier transfer leads to an increase of the @xmath130-magnetisation . contrary , the @xmath116-magnetisation is decreased in addition to the usual temperature - induced depolarisation . this leads to a temperature - dependent increase of the ratio @xmath131 as it is known from polarised neutron - scattering experiments @xcite . for @xmath104k this ratio is @xmath132 and in good agreement with the measured value of @xmath133 @xcite . as is observed experimentally , the s- and p - states couple antiferromagnetically to the d - states @xcite . the curie temperature turns out to be @xmath134k and is thereby about a factor 2.6 larger than the measured value of 624k @xcite . the large value for @xmath135 is probably due to the mean - field character of the sopt - hf . note , however , that a simple lda+@xmath37 ( hf ) calculation for the same parameters @xmath37 and @xmath38 yields a @xmath104 magnetisation 0.57@xmath103 and a curie temperature of approximately 2500k . + the local moment at the site @xmath42 as function of temperature is shown on the right - hand side of figure [ fig : mcju02 ] . its d - contribution is strongly reduced compared to the unscreened case while the s- and p - moment is increased . the strong reduction of the total magnetic moment will influence the spin polarisation of aes , since the `` screened '' qdos enters the auger green function [ eq:3tgf ] . note that the total magnetisation has to be calculated using equation ( [ eq : tdos ] ) , incorporating hybridisation with delocalised states . to express the auger intensity as a functional of the one - particle green functions we consider the diagrammatic expansion of the three - particle green function @xmath136 ( @xmath54 ) . here we can restrict ourselves to the direct diagrams and incorporate the exchange diagrams by introducing `` direct '' ( @xmath137 ) and `` exchange '' auger matrix elements ( @xmath138 ) . the auger intensity then reads as @xmath139 the auger matrix elements are taken to be constant . following ref . we set : @xmath140{rl } 1 & \mbox{for } l_1\leq l_2 \\ -1 & \mbox{for } l_1 > l_2\;. \end{array}\right . \label{eq : augerc}\ ] ] we thereby account for the singlet contributions , i.e. the holes in the final state have opposite spin ( @xmath141 ) , as well as for the triplet contributions ( @xmath142 ) to the auger intensity . the triplet contributions would be ignored if the auger matrix elements were chosen to be symmetric in the orbital index ( @xmath143 ) because the transition operator then vanishes , as can be seen from equation ( [ eq : transop ] ) . we consider two different approximation schemes for the treatment of the direct correlations . in the first approach , following refs . we neglect the direct cv correlations and treat the direct vv correlations by means of the vv ladder approximation , which becomes exact in the limit of completely filled or empty bands . a typical diagram contributing to the vv ladder is shown in figure [ fig:2ladder ] . the solid lines represent the renormalised one - particle propagators of the valence - band while the wiggly line is the one - particle core - propagator . the dashed line corresponds to the vv interaction . summing up all diagrams of this kind , yields the vv ladder approximation . the three - particle spectral density is given by @xmath144 where @xmath145 is the bose function . the two - particle valence - band spectral density @xmath146 is obtained from the corresponding two - particle green function . using a matrix notation with respect to @xmath147 and @xmath148 , the two - particle green function reads as @xmath149 with @xmath150 . @xmath151 is the two - particle green function which has to be calculated form the self - convolution of the partial qdos @xmath152 using the spectral representation : @xmath153 in equation ( [ eq : ladder ] ) we have applied the local approximation , i.e. only the on - site elements for @xmath90 are assumed to be non - zero . this approximation is analogous to the local approximation for the one - particle green function in section [ sec : indirect ] . a straightforward way to include the direct cv correlation on the same level as the direct vv correlations has been discussed in refs . . this leads to the ( `` three - particle '' ) cvv ladder approximation . for the limiting case of completely filled or empty bands the cvv ladder represents the exact solution and recovers the vv ladder but shifted energetically by @xmath154 due to the cv interaction @xcite . a typical diagram is shown in figure [ fig:3ladder ] where the dotted line represents the cv interaction . the cvv ladder approximation leads to a coupled set of fredholm integral equations @xcite . for a multi - band model , however , the numerical evaluation is beyond our present computational capacities . we therefore discuss a simpler approximation where only diagrams up to first order in the direct correlations are retained ( see figure [ fig : firstorder ] ) . the calculated auger spectra for ni resulting from different approximations are shown in figure [ fig : aest0 ] . the core hole is assumed to be unpolarised ( non - resonant @xcite process ) . the intensities for core spin @xmath41 and @xmath155 are added incoherently : @xmath156 . however , the auger intensity is still spin - dependent due to the ferromagnetic order in ni . + in figure [ fig : aest0 ] we plotted the total auger - intensity @xmath157 on the left and the spin asymmetry @xmath158 on the right . part ( a ) shows the result of the self - convolution model ( equations ( [ eq:3tsd ] ) and ( [ eq : sc ] ) inserted in ( [ eq : augerint ] ) ) , i.e. the self - convolution of the occupied qdos in figure [ fig : qdosju02 ] ( l.h.s . ) . direct correlations and core - hole screening are neglected altogether . part ( b ) corresponds to the vv ladder approximation starting from the ( unscreened ) qdos . taking additionally into account the screening effects introduced by the presence of the core hole in the initial state results in ( c ) . the spectrum obtained by the first order in the direct vv and cv correlations ( see figure [ fig : firstorder ] ) and by the screened qdos is plotted in part ( d ) . + as one can see in the plots on the left - hand side , the vv interaction is too weak to produce bound states , no sharp satellite appears , and the spectra appear to be band - like . compared to the self - convolution ( a ) , however , a considerable shift to lower energies is observed in ( b ) . this shift results from the direct correlations between the two final - state holes in the valence band . in ( c ) the main peak is shifted to still lower energies . this is an effect of the core - hole screening in the initial state and can be traced back to the redistribution of spectral weight in the screened qdos ( figure [ fig : qdosju02 ] ) . compared to ( a ) and ( b ) , the total aes intensity is clearly increased in ( c ) . again this is a consequence of the core - hole screening since the number of occupied states available for the auger process is increased ( figure [ fig : qdosju02 ] , r.h.s . ) . the spectrum shown in ( d ) not only includes the initial - state core - hole screening but the final - state effects due to the destruction of the core hole . compared to ( c ) , where the initial state is described in the same way , these effects result in a strong shift of the main peak to higher energies . this shift almost exactly compensates the shifts to lower energies that are due to direct vv correlations ( b ) and the core - hole screening ( c ) . however , a weak shoulder at about @xmath159ev remains in the spectrum ( d ) . + in all cases there is a high spin asymmetry ( up to @xmath160% ) for energies between approximately @xmath161ev and 0 . this is a consequence of the fact that ni is a strong ferromagnet . there are almost no @xmath117-electrons above about @xmath162ev ( see figure [ fig : qdosju02 ] ) that can participate in the auger - process . the main contribution to the intensity is therefore due to triplet configurations where the two final - state holes or , equivalently , core - hole and auger electron have spin@xmath119 . however , the intensity is very small in this energy region . by taking into account the screening of the core hole in the initial state ( compare ( b ) and ( c ) ) the spin asymmetry is reduced over the whole energy range which essentially is the same effect as the reduction of the local magnetic moment at @xmath42 caused by the presence of the core hole ( figure [ fig : mcju02 ] ) . the total spin - polarisation @xmath163 in the case ( d ) is 2.6% and 1.6% for ( c ) . both values are close to the experimental value @xcite of 2% for the @xmath164-process . the cases ( a ) and ( b ) with a polarisation of 8.7% and 9.3% , respectively , overestimate the total spin - polarisation compared with the experimental value . + for the calculation of the orbitally resolved contributions to the auger intensity we may restrict the summation in equation ( [ eq : augerint ] ) to orbital indices ( @xmath165 ) belonging to @xmath116(@xmath120)-character only . the resulting contributions are shown on the left ( right ) of figure [ fig : aespt0 ] . the contributions due to the remaining terms are plotted in the middle . + in all cases ( a)(b ) the @xmath116 contributions are clearly stronger compared with the @xmath120 contributions . the ratio between the @xmath116 and @xmath120 partial intensities corresponds to the different degeneracies . comparing the cases ( a)(d ) , we notice that there are essentially the same trends in the partial intensities as for the total intensities , and the discussion is the same as above . the line shape in case ( d ) , however , shows up some fine structure , especially in the @xmath116 partial intensity , which is not that pronounced in the total intensity . the shoulder at the low - energy tail of ( d ) is due to direct vv correlations and may be interpreted as a hint for the formation of a bound state of the final holes . except for this shoulder , a surprising similarity between ( d ) and ( a ) is noticed , even for the orbital resolved spectra . this might also be due to the small number of diagrams taken into account . however , the cancellation of effects according to different interactions ( vv and cv ) was also pointed out in ref . . in this paper we have investigated electron - correlation effects on the auger line shape of ni as an example of a ferromagnetic 3d - transition metal . the starting point is a realistic set of hopping and overlap parameters taken from tight - binding band - structure calculations . we additionally consider a strongly screened on - site coulomb interaction between the rather localised 3d electrons . the respective coulomb - matrix elements are expressed in terms of effective slater integrals . choosing a non - orthogonal basis , a distinction between the different angular - momentum characters of the valence orbitals is possible . this is necessary for the precise definition of the coulomb - interaction part in the ( multi - band hubbard ) hamiltonian and also facilitates the interpretation of the auger spectra . furthermore , we account for the core - valence interaction which is responsible for the screening of the core hole in the initial state of aes and for the sudden response of the valence electrons due to the destruction of the core hole in the final state . within a diagrammatic approach , the indirect and the direct correlations can be studied separately . the indirect correlations have been treated by second - order perturbation theory around the hartree - fock solution ( sopt - hf ) . the vv interaction parameters are fixed by assuming a ratio @xmath105 and by fitting the experimentally observed magnetic moment for @xmath104k , leading to @xmath100ev and @xmath101ev . this is equivalent to an intra - orbital interaction of @xmath166ev ( @xmath167 ) . the resulting curie temperature within the theory presented here is by a factor 2.6 larger than the experimentally observed one but considerably lower than the lda+@xmath37 ( hartree - fock ) value . + the core - valence interaction leads to a breakdown of the translational invariance in the initial state for aes . the interaction parameter @xmath110ev is fixed be requiring charge neutrality at the site @xmath42 where the core hole is created . the screening of the additional core - hole potential causes a transfer of spectral weight below the fermi energy and thus a considerable reduction of the local magnetic moment . however , the local magnetic moment is finite since the s- and p - electrons also contribute to the screening . to study the direct correlations we have used two different approaches . the first is the vv ladder approximation , which results in a band - like auger spectrum with a single maximum . in a second approach we have summed up the first - order diagrams with respect to the vv and cv interaction . the resulting spectrum shows up a shoulder at the lower tail due to the vv interaction . otherwise the line shape is very similar to that obtained by the self - convolution of the unscreened qdos . + as far as concerns the line shape , we conclude that the different correlation effects , vv and cv correlations in the initial and the final state , nearly cancel . however , a strong effect of electron correlations has been found in the orbitally resolved partial intensities and particularly for the spin - asymmetry . the calculated spin polarisation is in a good agreement with the measured one of the @xmath164-process @xcite . this corresponds to the excitation of a not too deep lying core level , i.e. the two - step model should be applicable . the reduced but finite spin polarisation of the @xmath164-process ( as compared to the band polarisation ) can be explained by effects of core - hole screening rather than by a core - hole polarisation , caused by a resonant excitation of the core electron into the valence band @xcite . future work will show whether these findings also apply to other 3d - transition metals . financial support of the _ deutsche forschungsgemeinschaft _ within the project no . 158/5 - 1 is greatfully acknowledged . the numerical calculations were performed on the crayt3e at the _ konrad - zuse - zentrum fr informationstechnologie berlin ( zib)_. there are several advantages for using the non - orthogonal basis set @xmath168 . the slater - koster parameters @xcite for the two - center approximation ( used here ) are much more accurate for a non - orthogonal basis set compared to an orthonormal one @xcite . secondly , the non - orthogonal basis ( lcao basis ) is built up from quasi - atomic orbitals . one therefore knows the behaviour of the basis states under symmetry operations belonging to the @xmath169-group which eventually results in the fact that local quantities , e.g. the on - site green function , are diagonal in the orbital index . furthermore , the coulomb matrix elements can be calculated in a highly symmetric way by using @xmath170-symbols in combination with a transformation from spheric to cubic harmonics . the unknown radial parts of the basis are parametrised by the effective slater integrals ( @xmath31 , @xmath32 , @xmath33 ) . finally , one can ensure that the coulomb interaction acts between 3d - electrons only . for the formalism of second quantisation , for the many - body and green - function theory , for the proof of wick s theorem and thus the development of the diagram technique , however , the use of an orthonormal set of one - particle basis states is necessary . therefore , it is convenient to derive all expressions that refer to the non - orthogonal basis , @xmath171 by a ( non - unitary ) lwdin transformation @xcite from a related set of orthogonal one - particle basis states : @xmath172 the overlap matrix @xmath173 is hermitian , @xmath174 indeed represents an orthonormal and complete basis set . the completeness relation can be written as @xmath175 annihilation ( and creation ) operators referring to the non - orthogonal basis may be defined as : @xmath176 it is instructive to see how the creation operator acts on the vacuum state : @xmath177 furthermore , one gets from the transformations ( [ eq : ob ] ) and ( [ eq : oc ] ) : @xmath178 thus , an operator in second quantisation has the same structure for both , the orthonormal and the non - orthogonal basis set . a one - particle operator @xmath179 , for example , reads as : @xmath180 the non - orthogonal green functions are defined as in the orthonormal case , e.g. @xmath181 for the one - particle green function . for example , using the non - orthogonal version of the fundamental anti - commutation rules ( @xmath182_+=s_{\alpha\alpha^\prime}^{-1}$ ] , see also eq . ( [ eq : acr ] ) ) and the equation of motion , the non - interacting green function turns out to be : @xmath183 where we used the matrix notation @xmath184 etc . for the interacting green function one has ( compare with eq . ( [ eq : vgf ] ) ) : @xmath185 10 [ 2]#2 [ 1]#1 [ 1]#1 , ( , london , ) , , . and , , ( ) . and , ( , amsterdam , ) , , . , ( , bristol , ) , . , , , , and , ( 2 ) , ( ) . , , , and , ( 18 ) , ( ) . , , ( ) . , , , , , , , and , ( 10 ) , ( ) . , , ( ) . , , ( ) . , , , and , ( 2 ) , ( ) . and , , ( ) . , , ( ) . , ( 8) , ( ) . and , ( 5 ) , ( ) . , ( 6 ) , ( ) . , , and , ( 11 ) , ( ) , , , and , , ( ) . , , , and , ( 7 ) , ( ) . , , and , , ( ) . , , and , , ( ) . and , , ( ) . and , ( 12 ) , ( ) . and , ( 3 ) , ( ) . and , , ( ) . and , , ( ) . and , ( 8) , ( ) . , , ( ) . and , , ( ) . , , ( ) . , , , and , , ( ) . and , , ( ) . , , ( ) . , , and , ( 22 ) , ( ) . , , , and , ( 19 ) , ( ) . , , and , , ( ) . , , , and , , ( ) . , , , and , , ( ) . , , and , ( , new york , ) . and , ( 12 ) , ( ) . , , and , ( 23 ) , ( ) . , ( , new york , ) . , , and , , ( , new york , ) . , , and , , ( ) . , , ( ) . , , ( , ) . and , , ( ) . and , , ( ) . and , ( 6 ) , ( ) . , , , , and , , ( ) . and , , ( ) . , , and , ( 6 ) , ( ) . , , , , , , , , and , , ( ) . , , , , , and , , ( ) . and , , ( ) , ( 10 ) , ( ) . , , and , ( ) , http://xxx.lan.gov / abs/. , , and , , ( ) . , , , and , , ( ) . , ( , berlin , ) , , chap . magnetic properties of metals . , , , and , ( 10 ) , ( ) . and , ( 6 ) , ( ) . , ( 3 ) , ( ) .

cvv auger electron spectra are calculated for a multi - band hubbard model including correlations among the valence electrons as well as correlations between core and valence electrons . the interest is focused on the ferromagnetic 3d - transition metals . the auger line shape is calculated from a three - particle green function . a realistic one - particle input is taken from tight - binding band - structure calculations . within a diagrammatic approach we can distinguish between the _ direct _ correlations among those electrons participating in the auger process and the _ indirect _ correlations in the rest system . the indirect correlations are treated within second - order perturbation theory for the self - energy . the direct correlations are treated using the valence - valence ladder approximation and the first - order perturbation theory with respect to valence - valence and core - valence interactions . the theory is evaluated numerically for ferromagnetic ni . we discuss the spin - resolved quasi - particle band structure and the auger spectra and investigate the influence of the core hole .

introduction model auger intensity indirect correlations direct correlations summary non-orthogonal basis-set

in this article we analyse the stability of the hyperboloidal initial value problem for a class of solutions to the vacuum einstein field equations the so - called _ purely radiative vacuum spacetimes_. purely radiative vacuum spacetimes describe gravitational radiation which comes from infinity , interacts with itself in a non - linear way and then disperses to infinity . in order to encode that the spacetime is made only of incoming gravitational radiation , one assumes that it admits a penrose conformal extension such that the null generators of future and past null infinity ( @xmath0 and @xmath1 ) are complete and that future and past timelike infinity are represented in the conformal extension by two points @xmath2 and @xmath3 . the points @xmath4 are required to be regular that is , the conformally rescaled unphysical spacetime admits a smooth extension which is regular at @xmath4 . more precisely , given a vacuum spacetime @xmath5 it will be assumed that there exists a conformally related spacetime @xmath6 such that @xmath7 where @xmath8 is a suitable conformal factor . null infinity , @xmath9 , is characterised as the set of points for which @xmath10 on the other hand , the points @xmath4 in a purely radiative spacetime are such that @xmath11 it is perhaps worth noting that the corresponding points @xmath4 in a spacetime with a black hole ( say , the schwarzschild spacetime ) do not satisfy these conditions . it has been know for a long time that the question of the existence of non - trivial examples of purely radiative solutions to the einstein field equations hinges very delicately on the structure of the spacetimes in a neighbourhood of the so - called spatial infinity , @xmath12 see e.g. @xcite . a way of getting around the so - called @xmath12-_problem _ is to consider initial value problems posed on _ hyperboloidal hypersurfaces _ , that is , spacelike hypersurfaces which can be thought of as intersecting null infinity in a transversal manner . if the initial data prescribed on the hyperboloid is suitably close to a _ canonical _ hyperboloidal data implied by the minkowski spacetime that is , the data is a _ perturbation _ of minkowski data then it has been proved that the resulting development has a conformal structure similar to that of minkowski spacetime @xcite . more precisely , the future development of initial data admits a conformal compactification such that the locus of points , @xmath0 satisfying the condition ( [ scri : condition ] ) has the topology @xmath13 , and that there is a single point , @xmath2 , satisfying the conditions ( [ i+:condition ] ) where the null generators of null infinity converge . the original proof of this result was carried out using the so - called _ conformal einstein field equations _ discussed in @xcite . these equations have the property of being regular even at the points where the conformal factor vanishes and are such that a solution to them implies a solution to the vacuum einstein field equations . a property of the evolution systems implied by these conformal field equations is that it includes the conformal factor as one of the unknowns , and thus any discussion of the structure of the conformal boundary has to be performed _ a posteriori_. the results of @xcite have been generalised to the case of the einstein field equations are coupled to maxwell and yang - mills fields @xcite . a more general system of conformal field equations has been introduced in @xcite . this _ extended conformal einstein field equations _ are expressed in terms of so - called _ weyl connections _ torsion free connections ( not necessarily levi - civita ) which preserve the conformal structure . thus , the field equations contain more gauge freedom . in particular , using the extended conformal equations it is possible to introduce gauges _conformal gaussian gauges_ which are based on certain type of conformal invariants the so - called _ conformal geodesics_. the advantages of these conformal gaussian systems are twofold : combined with the extended conformal field equations , they imply simpler systems of evolution equations ; and , if working in vacuum , they provide an _ a priori _ canonical conformal factor so that the location and properties of the conformal boundary are known and can be adjusted according to need before the propagation equations are solved . in @xcite , the extended conformal einstein field equations and conformal gaussian gauge systems have been used to provide a new version of the proof of the stability results for minkowski and de sitter spacetime . in particular , the analysis of the properties of the conformal boundary of the spacetimes obtained is much simpler and transparent . the proofs of the stability of perturbations of the minkowski spacetime in @xcite and in @xcite make use of an adaptation of a very general result on the existence and stability of symmetric hyperbolic systems by kato @xcite . the possibility of using these theorems to directly prove the global existence of solutions to the einstein field equations stems from the use of a conformally compactified picture where timelike infinity is at a finite position . as a result , kato s theorems allow to provide very compact global existence results which require almost no pde ( partial differential equations ) analysis . the latter should be contrasted , for example , with the analysis of the non - linear stability of the minkowski spacetime carried out by christodoulou & klainerman @xcite where precise asymptotic estimates of the solution are obtained . in the present article we generalise the methods and result of @xcite to reference solutions which are purely radiative spacetimes in the sense discussed in the first paragraphs of this introduction . a procedure to construct an infinite number of these purely radiative spacetimes has been given in @xcite . the possibility of generalising the existence and stability results of the minkowski spacetime using this class of solutions has been suggested in @xcite . this construction depends upon a remarkable connection between stationary solutions to the einstein field equations and spacetimes with vanishing mass . the reference spacetimes thus obtained are only known in an abstract sense and without much information about their geometry at hand . it is remarkable that although one has very little explicit information about the reference solutions , it is still possible to use the methods of @xcite to obtain a stability result . our main result can be summarised in the following form . given a sequence of multipole moments for a static solution to the einstein vacuum equations subject to a suitable convergence condition , there exists a hyperboloidal initial data set whose future development is a purely radiative spacetime admitting a conformal compactification with a future null infinity satisfying conditions ( [ scri : condition ] ) and a point representing timelike infinity satisfying conditions ( [ i+:condition ] ) . a small enough perturbation of the hyperboloidal initial data has a future development whose conformal boundary has the same properties as that of the unperturbed solution . in order to prove this theorem one starts by considering the asymptotic end of an initial data set for a vacuum static solution to the einstein field equations . this asymptotic end can be compactified to a 3-dimensional manifold , @xmath14 , such that the point at infinity corresponds to a certain point @xmath15 . the subsequent analysis will be restricted to a suitable neighbourhood of this point . the observations in @xcite allow to construct an initial data set for the conformal einstein field equations which is analytical in a neighbourhood of @xmath15 . this initial data has the peculiarity of having a vanishing adm mass . as our analysis is local to @xmath15 the later feature causes no problem . a local existence for this data follows directly . our interest is then concentrated to the region in the development which is included in @xmath16 ( the causal past of @xmath15 ) . as a consequence of our conformal setting , this region implies the existence a portion of a purely radiative spacetime . the latter can be regarded as the future development of some data for a purely radiative spacetime prescribed on a fiduciary hyperboloid @xmath17 . the point @xmath15 in @xmath14 corresponds to the point @xmath2 of the radiative spacetime . in order to prove the existence of the reference radiative spacetime one employs a conformal gaussian system based on a congruence of conformal geodesics leaving the hypersurface @xmath14 orthogonally . it should be noted that this congruence will , in general , intersect the fiduciary hyperboloid in a non - orthogonal manner . this feature hinders a direct application of kato s theorems in order to conclude the stability of the hyperboloidal data . the reason for this is that in order to make use of kato s theorems one has to make use of a gaussian system in which the conformal geodesics leave the initial hyperboloid with the same angle and orientation as the congruence in the reference solution meets the hyperboloid @xmath17 . it turns out that the data for such an evolution are not really intrinsic to the initial hyperboloid but contain information along the direction normal to the initial hypersurface . thus , one needs to show , first , local existence for the perturbed hyperboloidal data using , for example , a congruence of conformal geodesics which is orthogonal to the hyperboloid . once this local existence has been established , one can construct initial data for the evolution along the _ tilted _ congruence using a lorentz transformation . from here kato s theorems and the requirement of the perturbed data being close enough to the reference solution yields the required stability result . we start by setting our conventions in section [ conventions ] and recalling the physical and the conformal vacuum constraints . as we are working on @xmath18 and subsets thereof , we fix coordinates and vector fields on @xmath18 and briefly review the norm and the extension operator for sobolev spaces associated to @xmath18 . in section [ cefe ] we discuss the conformal tools , which we will be using later on . these include conformal geodesic , their associated conformal gaussian gauge and the conformal einstein field equations . we follow the standard approach by friedrich and derive an evolution system form the field equation by using space spinors . section [ section : reference : radiative : spacetime ] is concerned with the construction of a purely radiative reference spacetime from time symmetric static vacuum data with non - vanishing mass given on a cauchy surface passing through the point @xmath15 at spacelike infinity . in section [ section : hyperboloidal : reference : data ] the resulting radiative spacetime is analysed from the viewpoint of hyperboloidal data on a surface not passing through @xmath15 . we note at this stage that the initial data on such a hypersurface is not proper hyperboloidal data due to the fact that the conformal geodesic congruence does not intersect the hyperboloid orthogonally . thus we analyse the relationship between this so named tilted hyperboloidal data and the proper hyperboloidal data . section [ section : perturbed : radiative : spacetime ] is concerned with the existence and stability result for radiative spacetimes . the structure of the conformal boundary is analysed prior to the evolution of the data . in section [ section : stability : result ] we use a modified version of a theorem by t.kato employed in @xcite and @xcite to establish the existence and stability for purely radiative spacetimes in theorems [ katopart2 ] and [ katopart2a ] . we shall consider spacetimes @xmath5 satisfying the vacuum einstein field equations @xmath19 where @xmath20 denotes the ricci tensor of a lorentzian metric @xmath21 , ( @xmath22 ) with signature @xmath23 . the discussion of the solutions to equation ( [ efe ] ) will be carried out in terms of a conformally rescaled metric @xmath24 related to @xmath21 according to equation ( [ rescaled : metric ] ) . in what follows , let @xmath25 denote the levi - civita covariant derivative of @xmath24 , while @xmath26 , @xmath27 , @xmath28 , @xmath29 denote the associated riemann , weyl and ricci tensors and the ricci scalar of @xmath24 . in the stability analysis we will compare a spacetime @xmath30 , evolved from some chosen initial data , with a given reference spacetime . throughout the article , the latter will be denote by @xmath31 . we will also adapt this notation for all quantities related to the reference spacetime . we will make use of two types of hyperboloidal data . _ proper _ hyperboloidal data is given with respect to a frame consisting of three tangent vectors to the hypersurface and its normal , while _ tilted _ hyperboloidal data depends on a frame not intrinsic in the hypersurface . in our setup this frame will be oriented along the congruence . we will distinguish between the two frames and the corresponding frame components , by using an underscore for proper hyperboloidal data . the conformal factor on the spacetime will be denoted by @xmath8 , while @xmath32 is the conformal factor on an initial hypersurface . let @xmath33 denote a timelike hypersurface of the spacetime @xmath5 . let @xmath34 , @xmath35 denote , respectively , the first and second fundamental forms induced by the metric @xmath21 on @xmath33 ( @xmath36 ) . the vacuum einstein field equations ( [ efe ] ) imply the following constraint equations on @xmath33 @xmath37 where @xmath38 denotes the levi - civita and @xmath39 the ricci scalar of the metric @xmath34 . the first and second fundamental forms determined by the metrics @xmath24 and @xmath21 on @xmath33 are related by @xmath40 where @xmath41 and the function @xmath42 on @xmath33 denotes the derivative of @xmath8 in the direction of the future directed @xmath43-normal of @xmath33 . the traces @xmath44 and @xmath45 are related by @xmath46 the hamiltonian and momentum constraints for vacuum , equations ( [ hamiltonian : constraint ] ) and ( [ momentum : constraint ] ) , written in terms of the conformal fields @xmath32 , @xmath42 , @xmath47 and @xmath48 read @xmath49 where @xmath50 denotes the levi - civita connection and @xmath51 the ricci scalar of the metric @xmath47 . in particular if @xmath52 one has that @xmath53 in the sequel we will consider two different classes of solutions to equations ( [ conformal_hamiltonian ] ) and ( [ conformal_momentum ] ) . the first class will consist of initial data sets for the einstein vacuum field equations which are _ asymptotically euclidean _ , and thus , the relevant initial hypersurfaces are cauchy hypersurfaces of an asymptotically flat spacetime . further details will be discussed in section [ asympt : euclidean : regular ] . the second class of solutions to be considered are hyperboloidal data sets . in this case the initial hypersurfaces are not cauchy hypersurfaces of a globally hyperbolic spacetimes . further details will be given in section [ section : hyperboloidal : data ] . the character of the solutions to the conformal constraint equations ( asymptotically euclidean or hyperboloidal ) is determined through the boundary conditions which supplement equations ( [ conformal_hamiltonian ] ) and ( [ conformal_momentum ] ) . in the sequel we shall work with conformal spacetimes with time slices diffeomorphic to simply connected subsets of the 3-sphere , @xmath54 . therefore , it will be important to have a good coordinatisation of @xmath54 and to have a frame which is globally defined over this manifold . following the work in @xcite we regard @xmath54 as a submanifold of @xmath55 : @xmath56 the restrictions of the functions @xmath57 , @xmath58 on @xmath55 to @xmath54 will again be denoted by @xmath57 . if we view the point @xmath59 as the north pole and @xmath60 as the open upper half sphere , then for any subset @xmath61 of @xmath62 with @xmath63 the functions @xmath64 provide coordinates on @xmath61 centered at @xmath65 . the vector fields @xmath66 on @xmath55 are tangent to @xmath54 and provide a globally defined frame field on @xmath54 . given a simply connected 3-manifold @xmath67 with boundary @xmath68 , there exists a diffeomorphism @xmath69 , such that @xmath70 . since @xmath71 is a diffeomorphism we can map any tensor from @xmath67 to @xmath61 and vice versa . for convenience we denote tensors on @xmath67 and @xmath61 by the same symbols and the underlying manifold is to be understood from the context . in particular we can pull - back the functions @xmath57 to provide coordinates on @xmath67 and use the vectors @xmath72 , as given by ( [ v_field_c1])-([v_field_c3 ] ) , as a globally defined frame field on @xmath67 . onto @xmath54 . ] in the sequel it will be necessary to extend certain ( vector valued ) functions defined on a subset @xmath73 to the whole of @xmath54 . for this we follow the approach pursued in @xcite . on the spaces @xmath74 and @xmath75 of smooth @xmath76-valued functions on @xmath61 and @xmath54 define for @xmath77 the sobolev - like norm @xmath78 with @xmath79 . let @xmath80 and @xmath81 be the hilbert spaces obtained , respectively , as the completion of the spaces @xmath74 and @xmath75 on the norms given by formula ( [ norm ] ) . given @xmath82 , there exists a linear extension operator @xmath83 such that @xmath84 with @xmath85 a constant which is universal for fixed @xmath86 . it is important to notice that if , for example , @xmath87 is a solution to the conformal constraints , equations ( [ conformal_hamiltonian ] ) and ( [ conformal_momentum ] ) , on @xmath61 then @xmath88 will not satisfy the constraints in @xmath89 . when discussing the existence of solutions to the einstein field equations in terms of conformally rescaled spacetimes , it is convenient to introduce gauges based on conformally invariant structures . the particular conformal structures to be used in our analysis are _ conformal geodesics_. these curves are autoparallel with respect to _ weyl connections _ that is , torsion - free connections which preserve the conformal structure , but are not necessarily the levi - civita connection of a metric . conformal geodesics were first applied in the context of general relativity in @xcite . the idea of conformal gaussian coordinates based on congruences of conformal geodesics has been introduced in @xcite . it has been shown in @xcite that conformal gaussian coordinates can be used to cover even the strong field regions of the schwarzschild spacetime . in this section we present a brief discussion of the properties of gauges based on conformal gaussian coordinates , and of the associated propagation equations implied by the extended conformal einstein field equations . for further details we refer the reader to @xcite and to the references mentioned in the previous paragraph . given a spacetime @xmath5 , a conformal geodesic @xmath90 and its associated 1-form @xmath91 are solutions to the system of equations @xmath92 where @xmath93 denotes the _ schouten tensor _ of the levi - civita connection @xmath94 , given by @xmath95 the notation @xmath96 has been used . associated to the pair @xmath97 , there is a preferred family of conformal factors , @xmath8 , which are developed along the congruence using @xmath98 consistent with equation ( [ rescaled : metric ] ) , define @xmath99 , and denote the corresponding levi - civita connection by @xmath100 . in this conformal gauge the 1-form along the conformal geodesics is given by @xmath101 and satisfies @xmath102 . in the following , we will work with the unphysical metric @xmath24 and the 1-form @xmath103 the 1-forms @xmath104 and @xmath105 are used to define a weyl connection , @xmath106 , via @xmath107 in the gauge corresponding to the connection @xmath106 the equations ( [ cge1 ] ) and ( [ cge2 ] ) take the simpler form @xmath108 where @xmath109 is the schouten tensor of the weyl connection @xmath106 . it is given by @xmath110 note that generally one will have that @xmath111}\neq 0 $ ] . we transport an orthogonal frame @xmath112 , @xmath113 , such that @xmath114 , along the congruence of conformal geodesics according to @xmath115 which implies that the frame is automatically @xmath43-orthonormal . if @xmath116 is a vacuum metric , then the conformal factor , @xmath8 , can be written down explicitly by @xmath117 where @xmath118 , @xmath119 and @xmath120 are functions which are constant along a given conformal geodesic . they are subject to the constraint @xmath121 furthermore , along each conformal geodesic @xmath122 are constant for @xmath123 see e.g. @xcite . analogous to the case of metric geodesic congruences there exist deviation vector fields for congruences of conformal geodesics @xcite , referred to as _ conformal jacobi fields _ and denote them by @xmath124 . let @xmath125 . these fields satisfy @xmath126 we say the congruence has a conjugate point if @xmath127 vanish at that point . the coordinates @xmath128 described in section [ coordinates : s3 ] will be extended off a fiduciary hypersurface by dragging them along a congruence of conformal geodesics . hence , setting @xmath129 one obtains a coordinatisation of a slab of the fiduciary hypersurface . in these coordinates one has that @xmath130 let @xmath131 and @xmath132 denote , respectively , the components of @xmath133 and @xmath109 with respect to the frame @xmath112 propagated according to the rule ( [ propagation : frame ] ) . one then has that @xmath134 see e.g. @xcite . this choice of coordinates , frame field and conformal will be referred to as a _ conformal gaussian gauge system _ based on the fiduciary hypersurface . the appearance of conjugate points in the congruence of conformal geodesics implies a breakdown of the conformal gaussian gauge . let @xmath135 , and let @xmath136 be its spinorial counterpart . in terms of spinors , the gauge conditions ( [ gauss : gauge1 ] ) and ( [ gauss : gauge2 ] ) can be written as see @xcite for details of the definition of @xmath137 @xmath138 the frame vectors @xmath139 will be expressed with respect to the vectors @xmath140 where @xmath141 is the normal to the surfaces of constant @xmath142 and @xmath143 , @xmath144 and @xmath145 are given by ( [ v_field_c1])-([v_field_c3 ] ) . we will write @xmath146 in order to obtain a suitable set of evolution equations we consider a space - spinor formalism see e.g. @xcite as it allows us to obtain evolution equations involving only _ unprimed _ spinors . introduce a spin dyad @xmath147 such that @xmath148 we decompose the connection as follows @xmath149 note that @xmath150 is not the levi - civita connection of the surfaces of constant @xmath142 , but the so - called _ sen connection_. the frame fields @xmath151 are similarly decomposed using @xmath152 the fields @xmath153 are associated to spatial vectors , and hence , they satisfy the reality conditions @xmath154 contracting with @xmath155 and symmetrising , one decomposes the remaining connection and curvature spinors as follows . the space spinor @xmath156 is defined by @xmath157 one observes that @xmath158 and @xmath159 . define @xmath160 , which in turn , will be decomposed as @xmath161 the spinors in the latter equation possess the following symmetries @xmath162 in the sequel we will make use of propagation equations which are derived from the extended conformal field equations of @xcite see also @xcite . these equations are a generalisation of the original conformal equations which allow the use of weyl connections , as defined in ( [ weyl : connection ] ) . weyl connections make it possible to consider more general gauges in the derivation of systems of propagation equations out of the conformal field equations . in particular , it makes it possible to make use of the conformal gaussian gauge systems discussed in section [ section : conformal : gaussian : gauges ] . using the space spinor formalism and suitable contractions of the extended conformal field equations with @xmath163 , one finds that the conformal gaussian gauge system implies the following propagation equations for the unknowns @xmath153 , @xmath164 , @xmath165 , @xmath166 , @xmath167 and @xmath168 : @xmath169 where @xmath170 and @xmath171 denote , respectively , the electric and magnetic parts of @xmath172 and @xmath173 is the spinor representation of @xmath174 for further details see e.g. @xcite . note that the above equations involve an unspecified conformal factor @xmath8 and the space spinor representation of the 1-form @xmath175 ( @xmath176 ) introduced in ( [ d : decomposition ] ) . for @xmath8 we will use ( [ conformal factor ] ) and ( [ general : theta ] ) . the evolution equations for the spinor @xmath172 are derived from a space - spinor decomposition of the bianchi identity @xmath177 the resulting propagation equations are not unique as one can always add to them a multiple of the constraint equations implied by ( [ spinor : bianchi ] ) , @xmath178 . in @xcite the following bianchi propagations equations @xmath179 are considered to the so called _ standard system_. note that equation ( [ extended_cfe4 ] ) is expressed with respect to the levi - civita connection @xmath100 and not the weyl connection @xmath106 . in order to keep track of the possible appearance of conjugate points in the solutions to the propagation equations , we append to ( [ p1])-([p7 ] ) and ( [ extended_cfe4 ] ) an evolution equation for the jacobi field . the conformal jacobi field @xmath124 has a spinorial counterpart @xmath180 which can be split as @xmath181 conjugate points in the congruence of conformal geodesics arise if @xmath182 . equation ( [ jacobiequs ] ) implies that the fields @xmath183 , @xmath184 satisfy the propagation equations @xmath185 besides the above propagation equations ( [ p1])-([p7 ] ) , and ( [ extended_cfe4 ] ) one also obtains a set of equations referred to as the _ extended conformal constraint equations _ involving the operators @xmath186 and @xmath187 ( respectively @xmath150 ) . it should be noticed that despite their name , the constraint equations , are not intrinsic to the hypersurfaces of constant @xmath142 except , possibly , for an initial fiduciary slice as generically , they contain @xmath188 derivatives . the crucial observation for our purposes concerns their _ propagation properties _ as given by the following lemma proved in @xcite . let @xmath189 denote the future domain of dependence of @xmath190 see e.g. @xcite for definitions . [ lemma : propagation constraints ] assume that the constraint equations are satisfied on an initial hypersurface @xmath190 . if the conformal propagation equations are satisfied on @xmath189 , then the conformal constraint equations are also satisfied on @xmath189 . we discuss now some general structural properties of the equations ( [ p1])-([p7 ] ) , ( [ extended_cfe4 ] ) , ( [ j1])-([j2 ] ) which will be used systematically in the sequel . let @xmath191 where the subscript @xmath192 indicates that after symmetrisation @xmath15 indices are set to @xmath193 . introduce the notation @xmath194 where it is understood that @xmath195 contains only the independent components of the respective spinor which are obtained by writing linear combinations of irreducible spinors . in terms of @xmath195 and @xmath71 , the propagation equations ( [ p1])-([p7 ] ) and ( [ j1])-([j2 ] ) can be written as : @xmath196 where @xmath85 and @xmath197 denote , respectively , a linear constant matrix - valued function and a bilinear vector - valued function both with constant entries and @xmath198 is a linear matrix - valued function with coefficients depending on the coordinates . similarly equation ( [ extended_cfe4 ] ) can be written in the form @xmath199 where @xmath83 denotes the @xmath200 identity matrix and @xmath201 , @xmath202 , are @xmath200 matrices depending on the coordinates , while @xmath203 denotes a constant matrix - valued linear function of the connection coefficients @xmath204 . for later reference it is noted that @xmath205 and that @xmath206 with @xmath207 . from the reality condition ( [ reality : condition ] ) one has that @xmath208 so that in particular @xmath209 and @xmath210 are real . the hermitian matrices @xmath211 , @xmath212 imply real symmetric matrices for the propagation equations obtained from splitting the system ( [ upsilon : propagation])-([bianchi : propagation ] ) into real and imaginary parts of the propagation equations . these matrices are explicitly given in @xcite . let @xmath213 . the unknown @xmath214 takes values in @xmath76 for some @xmath215 . in this article @xmath214 will be defined over @xmath18 or subsets thereof . from equations ( [ upsilon : propagation ] ) and ( [ bianchi : propagation ] ) it follows that @xmath214 satisfies a system of quasilinear partial differential equations for @xmath214 of the form @xmath216 with @xmath217 denoting the vector fields ( [ v_field_c1])-([v_field_c3 ] ) acting on the unknown @xmath214 . given any @xmath218 , the matrix valued functions @xmath219 , @xmath220 have entries which are polynomial in @xmath221 . these polynomials are at most of degree one and have constant coefficients . the matrices are symmetric @xmath222 , @xmath223 . the matrix valued function @xmath224 with @xmath225 has entries which are polynomials in @xmath221 ( of at most degree 1 ) with coefficients which are analytic functions on @xmath226 . in what follows we will construct purely radiative vacuum spacetimes using the ideas of @xcite . these spacetimes are required to allow a conformal compactification such that the resulting conformal metric is smooth at null infinity including the point past or future timelike infinity . the purely radiative spacetimes in @xcite have been obtained from of static , asymptotically flat solutions to the einstein field equations . before discussing the results in @xcite which are crucial for our analysis , we introduce some definitions . asymptotically flat spacetimes , @xmath5 , can be obtained as the development of _ asymptotically euclidean _ initial data sets . the initial data @xmath227 will be said to be asymptotically euclidean if there is a compact subset of @xmath33 whose complement for some positive number @xmath228 is diffeomorphic to @xmath229 , where @xmath230 . in terms of the coordinates @xmath231 introduced by this identification a standard asymptotic flatness requirement is @xmath232 with @xmath86 a constant the adm mass of @xmath33 . an initial data set @xmath227 is said to be time symmetric if @xmath233 . it follows form the einstein field equations that its development will have a time reflexion symmetry with respect to the surface @xmath33 . as we shall be working with conformally compactified spacetimes , it will convenient to consider a compactified version of the initial hypersurface @xmath33 . to this end , it will be assumed that there is a 3-dimensional , orientable , smooth _ manifold @xmath234 , a point @xmath235 , a diffeomorphism @xmath236 and a function @xmath237 with the properties @xmath238 the last condition shall be , sloppily , written as @xmath239 that is , @xmath240 will be identified with @xmath33 . under these assumptions @xmath227 will be said to be _ asymptotically euclidean and regular_. suitable , punctured neighbourhoods of the point @xmath15 will be mapped into the asymptotic end of @xmath33 . in the sequel , we will consider initial data sets for the einstein field equations which satisfy the boundary conditions ( [ asymptotic:1])-([asymptotic:3 ] ) but have vanishing mass ( @xmath241 ) . it follows from the _ mass positivity theorem _ that if @xmath241 then either @xmath242 is data for the minkowski spacetime , or the initial data is singular somewhere in @xmath33 see e.g. @xcite . our discussion of initial data sets with vanishing mass will be concentrated on a small neighbourhood @xmath243 of the point @xmath15 where the relevant objects are suitably smooth . the construction of purely radiative spacetimes of @xcite can be summarised as follows . let @xmath244 be the norm of the timelike killing vector , @xmath245 , of some static , asymptotically flat , vacuum spacetime . assume that the spacetime has non - vanishing mass . in a static spacetime , there are coordinates @xmath246 for which the spacetime metric takes the form @xmath247 where @xmath248 denotes the metric of the hypersurfaces of constant @xmath249 , whereas @xmath250 is the metric of the quotient manifold in the case of static spacetimes the quotient manifold and any surface of constant @xmath249 can be identified . denote by @xmath33 one of these ( time symmetric ) hypersurfaces of constant @xmath249 . the metric @xmath251 satisfies on @xmath33 the time symmetric hamiltonian constraint @xmath252=0 $ ] , whereas @xmath250 and @xmath253 satisfy the _ vacuum static einstein field equations_. static spacetimes are usually analysed in terms of the quotient metric @xmath250 . the asymptotic behaviour of the quotient metric @xmath250 is best understood in terms the behaviour of a conformally rescaled quotient metric @xmath254 in a neighbourhood , @xmath255 , of the point at infinity @xmath15 . as before we have @xmath256 . using the so - called _ conformal static equations _ beig & simon @xcite have shown that there is a choice of @xmath257 and coordinates for which @xmath257 and @xmath258 are analytic in @xmath243 and satisfy the boundary conditions ( [ asymptotic:1])-([asymptotic:3 ] ) . the crucial observation in @xcite is that the conformally rescaled metric @xmath259 also satisfies the time symmetric hamiltonian constraint . the metric @xmath34 is asymptotically euclidean with vanishing mass . let @xmath260 the pair @xmath261 satisfies the time symmetric conformal hamiltonian constraint , equation ( [ conformal_hamiltonian ] ) , with @xmath262 . both @xmath263 and @xmath264 are analytic in @xmath243 and , furthermore , they satisfy the boundary conditions ( [ asymptotic:1])-([asymptotic:3 ] ) . the pair @xmath261 can be used to construct initial data for the conformal einstein field equations of section [ conformal : evolution : equations ] this data consists essentially of the value of the rescaled weyl tensor and the schouten tensor on the initial hypersurface . this initial data is obtained out of algebraic manipulations of the conformal constraint equations . although @xmath47 and @xmath32 are given in a gauge which makes them analytic around @xmath15 , the data for the conformal field equations need not be . it is a non - trivial result that the data constructed out of this procedure is indeed analytic around @xmath15 . the time evolution of this data yields using , say , the cauchy - kowalevska theorem in a 4-dimensional neighbourhood of @xmath15 an analytic spacetime metric @xmath24 and an analytic conformal factor @xmath8 . in the sequel this local existence problem will be solved by a different method . under these circumstances , the physical metric , @xmath265 , obtained on the timelike future @xmath266 of @xmath15 in this evolution is a solution of the einstein vacuum field equations with an analytic structure at past null infinity , for which the point @xmath15 represents a regular past timelike infinity @xmath3 . these solutions of the einstein field equations are _ radiative spacetimes _ in the sense discussed in the introduction . note that due to the time reflexion symmetry , the same holds for the pair @xmath267 where @xmath15 now represents the regular future timelike infinity @xmath2 . on the complement of @xmath268 these are the points in the development which are spacelike related to @xmath15 one also has a solution of einstein s field equations in a neighbourhood of @xmath15 , for which the point @xmath15 now represents the point spatial infinity @xmath12 . definitions of the causal sets @xmath269 and @xmath270 can be found in @xcite . the results of @xcite relevant for our analysis are presented in the following theorem . [ thm : reference : radiative : data ] given an analytic solution to the conformal static equations on @xmath255 , there exists a solution , @xmath261 , to the time symmetric conformal hamiltonian constraint on @xmath271 satisfying * it is asymptotically euclidean and regular in a neighbourhood @xmath243 , * it has vanishing adm mass , * the spinorial fields @xmath172 and @xmath272 constructed out of @xmath261 are analytic in @xmath243 . it was long conjectured that given a series of multipole moments satisfying an appropriate convergence condition , there exists a static solution of the einstein equations with precisely those moments . this conjecture was proved in full generality in @xcite see also @xcite for partial results and an alternative approach . these results show that the metric , once written in a certain gauge , is uniquely determined by the multipole moments . thus there is an infinite family of static metrics analytic in a neighbourhood of @xmath15 . this family is parametrised by the multipole moments . as a consequence of theorem [ thm : reference : radiative : data ] it follows that we have an infinite family radiative spacetimes constructed in the above mentioned way that will serve as reference spacetimes for our stability analysis . in this section we discuss how conformal gaussian gauge systems can be used to construct radiative spacetimes out of the cauchy initial data provided by theorem [ thm : reference : radiative : data ] . the original construction discussed in @xcite makes use of a propagation system derived from the `` original '' conformal field equations of @xcite . as mentioned before , due to the analyticity of the setting , existence of a unique analytic solution in a spacetime neighbourhood of the point @xmath15 follows from the cauchy - kowalewska theorem see e.g. @xcite . here we follow a different approach . our intention is to use the development of the data given by theorem [ thm : reference : radiative : data ] as a reference solution from which a stability result will be derived . as seen in @xcite , the discussion of the properties of the conformal boundary of radiative spacetimes simplifies with the use of a conformal gaussian system . thus , it is natural to discuss the structure and existence of the reference solution using the same type of gauges . in our approach , existence of a solution of the propagation equations implied by the extended conformal field equations and the conformal gaussian gauge follows from a variation of a theorem of kato see @xcite . let @xmath261 on @xmath14 be one of the solutions to the time symmetric conformal hamiltonian constraint , equation ( [ conformal_hamiltonian ] ) , given by theorem [ thm : reference : radiative : data ] . it is assumed that @xmath273 . in what follows all quantities derived from this initial set and from its time development will be distinguished with an overbar . although it is customary to make use of a conformal factor @xmath263 which is non - negative , it will be convenient to consider a non - positive conformal factor on @xmath274 the reason for this will be evident in section [ section : hyperboloidal : reference : data ] . accordingly , if one takes the conformal factor of theorem [ thm : reference : radiative : data ] with an extra factor of @xmath275 , so that it is negative at least in a neighbourhood away from @xmath15 . one has that @xmath263 satisfies the appropriate modification of the boundary conditions ( [ asymptotic:1])-([asymptotic:2 ] ) . namely , @xmath276 on a suitably small neighbourhood , @xmath277 , we use the coordinates @xmath278 centered at @xmath15 and consider the following initial data for a congruence of conformal geodesics : @xmath279 the coordinates @xmath280 are extended off @xmath14 by dragging along the congruence of conformal geodesics to obtain conformal gaussian coordinates as discussed in section [ section : conformal : gaussian : gauges ] . it follows that @xmath281 at @xmath15 and using @xmath282 , we get @xmath283 . the constraint ( [ theta : constraint ] ) applied to our situation gives the initial data for @xmath284 on @xmath285 , which must hence be positive there . the initial value at @xmath15 is obtained taking the corresponding limit . making use of normal coordinates centred in @xmath15 , one finds that since @xmath286 , we have @xmath287 and thus @xmath288 . hence @xmath289 everywhere on @xmath274 . it follows from equation ( [ general : theta ] ) that along each conformal geodesic the conformal factor @xmath290 is given by @xmath291 where @xmath292 define now the conformal boundary , @xmath293 , in a natural way as the locus of points in the development of the data on @xmath14 for which @xmath294 and @xmath295 it is easy to see that a conformal geodesic with data given by ( [ cg_initial_data ] ) passes through @xmath296 whenever @xmath297 . note that every conformal geodesic , except for the one passing through @xmath15 , intersects @xmath296 twice , once to the past of @xmath14 and once to the future . * remark . * we shall always work on a neighbourhood @xmath255 which is assumed to be sufficiently small to ensure @xmath298 in a neighbourhood of @xmath15 . we discuss now the existence of solutions to the propagation system ( [ sh : system ] ) with data prescribed on @xmath299 . the initial data for this initial value problem is obtained through the conformal constraint equations on @xmath14 @xcite and satisfies on @xmath255 @xmath300 , \\ & & \chi_{(ab)cd}=0 , \\ & & \theta_{abcd } = -\frac{1}{\bar{\omega } } d_{(ab } d_{cd ) } \bar{\omega } + \frac{1}{12}h_{abcd } , \\ & & \phi_{abcd } = \frac{1}{\bar{\omega}^2 } d_{(ab } d_{cd ) } \bar{\omega } + \frac{1}{\bar{\omega } } s_{abcd } , \\ & & \eta = 0 , \\ & & \eta_{ab } = \sigma^\er_{ab } , \label{reference : data9}\end{aligned}\ ] ] where @xmath301 and @xmath302 are the spatial infeld - van der waerden symbols . the spinorial fields @xmath303 $ ] and @xmath304 are the connection coefficients and the trace - free ricci curvature of the 3-metric @xmath264 . since @xmath243 is assumed to be a simply connected neighbourhood of @xmath15 , we can find a diffeomorphism @xmath305 onto @xmath306 . we extend the initial data to @xmath18 in the following way . the initial values for @xmath307 are constant and are extended with their respective values to all of @xmath54 . for @xmath308 we use the linear extension operator @xmath83 given in section [ section extending initial data ] to obtain fields defined over @xmath54 . the extended fields will be denoted on @xmath54 in the same way as in ( [ preextensionlabel ] ) . the vectorial unknown @xmath309 is viewed as a function of @xmath310 with values in the sobolev space @xmath311 . let @xmath312 denote the extended initial data on @xmath54 as described in the previous paragraphs . for @xmath313 , @xmath314 , @xmath315 , define set @xmath316 where @xmath317 denotes the matrix valued function defined by the symmetric hyperbolic system ( [ sh : system ] ) , and @xmath318 is the standard scalar product on @xmath76 . it can be verified that @xmath319 and hence @xmath320 for @xmath321 . one has the following local existence result . [ katopart1 ] suppose @xmath322 . there exists a @xmath323 , a subset @xmath324 and a unique solution @xmath325 of equation ( [ sh : system ] ) defined on @xmath326 $ ] with initial data @xmath327 on @xmath328 and such that @xmath329 furthermore , @xmath330 \times { \mbox{\sym s}}^3)$ ] and @xmath331 can be chosen such that @xmath325 has non - vanishing jacobi fields @xmath332 , so that the solution is free of conjugate points in @xmath333 $ ] . * proof . * the first part of the theorem follows from the generalisation given in @xcite see also @xcite of kato s existence and stability result for quasilinear symmetric hyperbolic systems @xcite . for the jacobi fields we observe that @xmath332 satisfies an ordinary differential equation along the curves of the congruence . hence there is a minimum interval for which the fields do not vanish . the lack of conjugate points then follows from our earlier discussion . @xmath334 * remark 1 . * it follows that on @xmath335 the solution , as given by the theorem , is of class @xmath336 * remark 2 . * let @xmath337 \times { \mbox{\sym s}}^3 $ ] and denote the metric on @xmath338 by @xmath339 . the spacetime @xmath340 will be used as the reference spacetime for our stability analysis later on . let @xmath341 denote the domain of dependence of @xmath243 see e.g. @xcite for definitions . the value of @xmath331 can be chosen small enough so that @xmath342 . the shaded region represents the physical spacetime . ] the finite reduction of @xmath331 is not a problem in our scheme as any @xmath323 corresponds to an infinite time interval in the physical description . as it will be seen in the sequel , it may be necessary to further reduce @xmath331 by a finite ( non - negative ) amount . the spacetime @xmath343 is conformally related to a vacuum spacetime , @xmath344 , with vanishing cosmological constant . the spacetime @xmath345 is a radiative spacetime as the conformal metric @xmath24 and associated fields extend smoothly through the conformal boundary @xmath296 and furthermore , there exists a point @xmath346 satisfying ( by construction ) the conditions to be a _ future timelike infinity _ cfr . the discussion in the introduction . as a result of lemma [ lemma : propagation constraints ] it follows that the conformal constraints are only propagated in @xmath341 . consequently , the region @xmath347 \times { \mbox{\sym s}}^3\setminus d(\mathcal{b}_a(i))$ ] is of no physical interest as it does not describe a vacuum spacetime . it is nevertheless relevant to use the whole of @xmath348 for the correct formulation of the stability part of our work . in the following , in a slight abuse of language , we will refer to @xmath349 as _ the purely radiative reference solution_. the conformal affine parameter @xmath350 defines , in a natural way , a foliation of the manifold @xmath351 . let @xmath352 denote the surfaces of constant @xmath350 . for fixed @xmath350 one has that @xmath352 is diffeomorphic to @xmath54 . let @xmath353 and define @xmath354 the set @xmath355 intersects null infinity in a _ hyperboloidal way_. more precisely , @xmath356 divides @xmath355 into two regions : one where @xmath357 and one where @xmath358 . the latter contains the origin and will be denoted by @xmath17 . hence @xmath359 the latter definition justifies the choice of a negative conformal factor @xmath263 made in the previous sections . we rewrite the conformal parameter @xmath350 and the conformal factor of the reference solution given by equation ( [ special : theta ] ) , so that their initial surface is @xmath355 . define @xmath360 so that @xmath361 on @xmath355 and @xmath362 the initial value of @xmath363 on @xmath355 will be denoted by @xmath364 . recalling that @xmath365 gives the location of null infinity , we can see that @xmath366 and hence @xmath364 is the correct boundary defining function on @xmath355 with @xmath367 . one can briefly verify that the new conformal factor @xmath363 gives the correct location of null infinity . if @xmath368 we get that @xmath369 whenever @xmath370 along the unique conformal geodesic @xmath371 passing through the point @xmath65 on @xmath54 at all times , we have @xmath372 and @xmath373 thus , along @xmath371 one has that @xmath374 only when @xmath375 . this corresponds to the point @xmath15 in the cauchy hypersurface @xmath376 . in what follows , let @xmath377 for consistency , we shall denote entries of @xmath378 , as defined above , also with @xmath379 . the entries of @xmath378 imply on @xmath355 data for the symmetric hyperbolic system ( [ sh : system ] ) . we shall write @xmath380 . the data @xmath381 are not truly hyperboloidal data . the reason for this is the following . the frame vector @xmath382 which on @xmath14 is normal to this surface , ceases to be normal to the surfaces @xmath352 , of constant @xmath350 if @xmath383 . this is a feature that differentiates conformal geodesics from standard geodesics . this phenomenon manifests itself in the fact that the components @xmath384 of the space spinor decomposition of @xmath382 according to formulae ( [ space : spinor : decomposition:1 ] ) and ( [ space : spinor : decomposition:2 ] ) , which are zero on @xmath14 , become , in general non - zero off @xmath14 . as a consequence , the spatial vectors of the frame @xmath385 , @xmath123 which span the tangent bundle to @xmath14 , @xmath386 , in general , pick up components off the tangent bundle @xmath387 . accordingly , the data @xmath381 , which is expressed in terms of the spatial frame @xmath385 ( @xmath388 ) is not intrinsic to @xmath355 . we shall call this type of data _ tilted hyperboloidal data_. in order to relate the tilted data @xmath381 to proper hyperboloidal data one proceeds as follows . denote the normal vector field to @xmath355 , the 3-metric induced by @xmath389 and the second fundamental form of @xmath355 by @xmath390 , @xmath391 and @xmath392 respectively . note that @xmath392 and @xmath393 are two conceptually different quantities the former is related to the hypersurface , the later to the conformal geodesic congruence . choose a @xmath394-orthonormal frame @xmath395 and set @xmath396 , so that @xmath397 , forms a @xmath398-orthonormal frame on @xmath355 . the frames @xmath399 and @xmath400 are related by a lorentz transformation @xmath401 , which will be regarded as a matrix valued function over @xmath54 , as follows @xmath402 the fields @xmath403 , @xmath404 , @xmath405 , @xmath406 are the frame components of spacetime quantities and the lorentz transformation induces transformed components @xmath407 , @xmath408 , @xmath409 , @xmath410 in the canonical way . for example @xmath411 the sl(2,)-spinorial counterparts of the above tensors can be obtained by suitable contractions with the constant infeld - van der waerden symbols . these , in turn , can be space - spinor split with respect to @xmath412 to obtain _ proper hyperboloidal data _ on @xmath17 . given one of the radiative spacetimes discussed in section [ section : reference : radiative : spacetime ] , it is natural to ask whether it is possible to establish for these spacetimes an analogue of the semiglobal stability results for the minkowski spacetime discussed in @xcite . in this section we establish such a result . this result generalises the aforementioned stability results to the case where the reference solutions are not flat . in order to discuss appropriately the notion of closeness of two hyperboloidal initial data sets , we make use of the notion of _ limit of spacetimes _ as discussed in e.g @xcite . assume that one has a smooth family of hyperboloidal initial data sets @xmath413 , parametrised by @xmath414 , such that @xmath415 where @xmath416 are simply connected 3-dimensional manifolds with boundary @xmath417 . the fields @xmath418 satisfy the vacuum constraints , equations ( [ conformal_hamiltonian ] ) and ( [ conformal_momentum ] ) , with @xmath419 as the manifolds @xmath416 are taken to be simply connected , there exists a family of diffeomorphisms @xmath420 with @xmath61 as in section [ coordinates : s3 ] . in particular , @xmath421 is a smooth family of diffeomorphisms between @xmath416 and @xmath17 , such that @xmath422 is the identity map . note that the locus of points for which @xmath423 coincides as sets on @xmath54 , by construction , for all @xmath414 . let @xmath424 , @xmath425 denote a smooth family of @xmath426-orthonormal frames such that @xmath427 with @xmath428 as given in the end of section [ section : hyperboloidal : reference : data ] . the levi - civita connection @xmath429 of the metric @xmath430 , together with the frame @xmath431 , imply the 3-dimensional ricci tensor @xmath432_{\alpha\beta}$ ] and spin connection coefficients @xmath433 , as well as their spinorial counterparts @xmath434 and @xmath435 . from these quantities and the tensor @xmath436 one can determine the @xmath437-components of the schouten tensor and @xmath438-electric and @xmath438-magnetic parts of the rescaled weyl tensor for the spacetime on @xmath416 see e.g. @xcite . using the frame @xmath437 and @xmath439 , their spinorial counterparts can be readily calculated . this procedure produces proper hyperboloidal data in the sense discussed in section [ section : hyperboloidal : reference : data ] . we note that so far the data is only given on @xmath440 , respectively , @xmath441 . we use the extension operator @xmath83 given in section [ section extending initial data ] to extend the proper hyperboloidal data to the whole of @xmath54 , bearing in mind that outside @xmath61 the data may violate the vacuum constraints ( [ conformal_hamiltonian ] ) and ( [ conformal_momentum ] ) . in the following , the parameter @xmath442 will be dropped from the expressions , in order to ease the presentation . from studying the proper hyperboloidal data on the hypersurface @xmath355 in the reference solution one can see that this data does not include all the quantities that are needed for the tilted hyperboloidal data used for ( [ sh : system ] ) . we are missing the data for @xmath443 , @xmath165 , @xmath164 , @xmath444 . in order to consider the construction of tilted hyperboloidal data for the conformal propagation equations we shall first develop a small part of the perturbed spacetime . for this we need a local existence result for the development of data constructed using an auxilary congruence of conformal geodesics which departs the initial hyperboloid orthogonally . the information thus obtained complements the proper hyperboloidal data constructed above . for the construction of the auxilary congruence we use the following initial setup . let @xmath445 , @xmath446 , @xmath447 $ ] , @xmath448 . as long as the data for @xmath449 is smooth and satisfies @xmath450 , i.e. @xmath451 it can be chosen arbitrarily , since we are only interested in local existience in some small neighbourhood of the initial surface . in particular , we can set @xmath452 again . we can now proceed to our local existence result for the unknown @xmath453 the underline in the above spinors indicates that they are expressed with respect to the frame @xmath437 . the proposition below is essentially very similar to theorem [ katopart1 ] . this is not a stability result , but purely a local existence for the perturbed data . for this we define @xmath454 analogously to ( [ positivedefiniteness : condition ] ) . [ lemma : local existence ] suppose @xmath322 . there exists a @xmath455 , a subset @xmath456 and a unique solution @xmath457 of equation ( [ sh : system ] ) defined on @xmath458 $ ] with initial data @xmath459 on @xmath460 and such that @xmath461 furthermore , @xmath462 \times { \mbox{\sym s}}^3)$ ] and @xmath457 has non - vanishing jacobi fields @xmath463 , so that the solution is free of conjugate points in @xmath464 $ ] . * remark : * the local existence theorem clearly applies to @xmath465 and the resulting spacetime coincides locally with the reference spacetime @xmath466 . one could then consider the derivation of a result similar to the stability of minkowski in @xcite . however it should be noted that there is no guarantee that this setup covers @xmath2 or @xmath467 of @xmath466 and its perturbations . thus a semiglobal existence and stability result can not be guaranteed . as a result of proposition [ lemma : local existence ] we have a spacetime slab @xmath468 \times { \mbox{\sym s}}^3 $ ] with metric @xmath469 which describes a vacuum spacetime on @xmath470 . for sufficiently small @xmath442 , one can regard the 3-manifold @xmath190 as a hypersurface of a spacetime @xmath471 . on this hypersurface we now want to give tilted hyperboloidal data for ( [ sh : system ] ) . for this we will evolve a conformal geodesic congruence , that departs @xmath190 in the same way as in the reference solution . thus we can locally obtain all the components of the unknown @xmath214 on @xmath472 and use this as initial data @xmath327 for ( [ sh : system ] ) . we know fix the initial data for the frame and the congruence . as before , let @xmath473 denote the unit normal to @xmath190 . then @xmath474 is an orthonormal tetrad on @xmath190 . from @xmath475 we construct a tilted orthonormal frame @xmath476 on @xmath477 via @xmath478 where @xmath479 is the _ same _ matrix valued function associated to the lorentz transformation via ( [ lorentz : transformation:1])-([lorentz : transformation:3 ] ) . it follows then that the timelike vector @xmath480 has the same projections on the orthonormal frame @xmath481 as @xmath382 has on @xmath482 . that is @xmath483 . note that @xmath382 is by construction tangent to a congruence of conformal geodesics . for discussions of the development of the _ perturbed initial data _ @xmath484 which go beyond local existence , it will be important to consider a congruence of conformal geodesics that departs the hypersurface @xmath67 in the same way as the congruence of conformal geodesics in the reference spacetime arrives at @xmath17 . for this reason we will use @xmath485 . ) normal to @xmath67 and the tilted congruence ( ) used to evolve the perturbed initial data . ] as in the case of the reference spacetime @xmath486 , we want to cover the development of the initial data set @xmath487 with a conformal geodesic coordinate system . the formulation of the evolution equations is dependent on a gauge for which @xmath488 . taking this as well as @xmath489 into account , we have @xmath490 assume that the perturbed hyperboloidal data @xmath487 satisfies the technical condition @xmath491 close to @xmath68 , where @xmath492 denote the value of the components of the 1-form @xmath175 for the reference solution and @xmath493 indicates that the quotient of the left hand side divided by @xmath32 goes to zero as @xmath32 goes to zero . initial data for the congruence of conformal geodesics on which the conformal gaussian system hinges will be prescribed such that : @xmath494 on @xmath67 . as a result of the technical condition ( [ technical : condition ] ) one has that that @xmath495 has a finite limit at @xmath496 and is thus well defined there . this particular choice of initial data for the congruence of conformal geodesics provides an easy analysis of the structure of the conformal boundary of the spacetime as will be seen in [ section : structure : of : the : conformal : boundary ] . initial values for @xmath118 , @xmath119 , @xmath120 which characterise the canonical conformal factor @xmath8 associated to the congruence of conformal geodesics defined by ( [ cg1])-([cg3 ] ) are given on @xmath67 by @xmath497 with the corresponding limits for points on @xmath496 . as usual , the subscript @xmath498 indicates that the value of the functions are extended of the initial hyperboloid by requiring them to be constant along a given conformal geodesic with initial data given by ( [ cg1])-([cg3 ] ) . it follows from the conformal hamiltonian constraint , equation ( [ conformal_hamiltonian ] ) that @xmath499 is null on @xmath68 . note that as a consequence of the conformal hamiltonian constraint it follows that when @xmath52 @xmath500 consistently with the choice of initial data for the conformal factor in ( [ cf : data ] ) . initial data for the jacobi field associated to the congruence of conformal geodesics arising from ( [ cg1])-([cg3 ] ) is set by @xmath501 finally , one needs to construct tilted data for the rescaled weyl ( @xmath502 ) and schouten ( @xmath503 ) tensors . for this , one starts with the standard hyperboloidal data for these tensors implied by the conformal constraint equations . as a consequence of the local existence result given by proposition [ lemma : local existence ] , one can regard @xmath67 rightfully as a hypersurface of a spacetime . let @xmath504 and @xmath505 denote the components of the weyl and schouten tensors with respect to the frame @xmath506 . the tilted initial data for the schouten and rescaled weyl tensors on @xmath67 is given by @xmath507 the required spinorial data is then obtained by means of a space spinor decomposition of the spinorial counterpart of the above tensor . in this subsection we analyse the structure of the conformal boundary of the development of hyperboloidal data which is close to hyperboloidal data of the reference radiative spacetime . attention is focused , in particular , on the location of timelike infinity @xmath508 . this investigation is possible as in the vacuum case the conformal factor can be calculated _ a priori _ from the initial data . along the congruence of conformal geodesics with initial data on @xmath67 given by ( [ cg1])-([cg3 ] ) , the canonical conformal factor , @xmath8 , takes the form @xmath509 with @xmath119 and @xmath120 given by the conditions in ( [ cf : data ] ) . as it is customary , denote by @xmath510 the locus of points in the development of the data @xmath511 for which @xmath512 . for @xmath513 to qualify as a candidate to be the timelike infinity of the development it needs to satisfy @xmath514 we recall that the 1-form @xmath515 and that @xmath516 hold . thus , for smooth @xmath105 and @xmath8 we observe that on @xmath510 , @xmath175 is null and can only vanish there if @xmath517 . therefore the only point on @xmath510 at which all @xmath518 can vanish , is at a candidate for @xmath2 . define the function @xmath519 and observe that @xmath520 if and only if @xmath521 . we recall here that @xmath522 , and hence @xmath523 , are constant along conformal geodesics see e.g. @xcite . now , by construction of the reference solution from the initial cauchy data given on @xmath14 and our choice of @xmath524 , there is a unique point in @xmath525 such that @xmath526 , namely @xmath15 itself . by our choice of coordinates , @xmath15 lies at the origin of our spatial coordinate chart and is identified with the north pole of @xmath54 . thus by theorem [ katopart1 ] and remark @xmath527 following it , the point @xmath528 is the unique point in @xmath341 satisfying the conditions for @xmath508 . this corresponds to @xmath529 in the @xmath355-adapted coordinates . furthermore the points @xmath530 are by construction the only points in @xmath341 where @xmath520 for the reference solution . given our choice for the initial data of @xmath175 in ( [ cg2 ] ) for the perturbed spacetime , it follows immediately that the only point at which @xmath523 vanishes on @xmath67 is the origin . therefore the only candidate for @xmath2 is the point , assuming it exists , where the conformal geodesic @xmath531 intersects @xmath510 . note that for general initial data it is possible for the development to contain a singularity before reaching @xmath2 as in the case of black hole spacetimes . the first condition in ( [ i+conditions ] ) implies @xmath532 while the second one can be rewritten as @xmath533 together with @xmath520 , both conditions imply that @xmath534 for sufficiently small perturbations , @xmath535 does not vanish and @xmath536 is finite . in fact later on the initial data will be restricted such that @xmath537 lies in the interval @xmath538 . note that for the reference solution we have @xmath539 and @xmath540 . by construction we have that @xmath541 is non - degenerate , and hence the @xmath2-candidate in the reference spacetimes can be regarded , rightfully , as future timelike infinity . the point @xmath542 satisfies the first two conditions of ( [ i+conditions ] ) . we now verify that the conditions discussed in the previous paragraphs imply that the hessian of @xmath8 is non - degenerate at @xmath543 . for this we make use of the following expression : @xmath544 using the conformal gaussian coordinates @xmath545 for our calculations and @xmath546 as well as ( [ dtheta1 ] ) we find that @xmath547 where the right hand side is given in terms of initial data and spatial derivatives . observe that @xmath548 are identical to the ones in the reference solution and thus the only change arises from the perturbation of @xmath549 . for data @xmath550 close enough to the reference solution , in the sense that at the spatial origin @xmath549 and @xmath551 are sufficiently close to their counterparts in the reference solution , the hessian will not be degenerate and thus @xmath543 can be identified as the timelike infinity @xmath508 of the perturbed spacetime . note that one may rescale the initial data such that @xmath42 remains fixed , while @xmath552 . this greatly simplifies the analysis of the effect of perturbations of @xmath549 on the hessian . we are now in the position of presenting our main result a stability result for purely radiative spacetimes arising from hyperbolidal data . the proof of this result follows the ideas of @xcite . let @xmath381 and @xmath327 denote , respectively , the tilted initial data for the radiative reference spacetime and for a perturbation thereof , as discussed in the previous section . using the diffeomorphism @xmath553 , we can think of @xmath381 and @xmath327 as a vector - valued function over @xmath54 . on the other hand , using @xmath71 , @xmath327 can only be regarded as a vector valued function over @xmath73 . on @xmath61 , let @xmath554 using the linear extension operator given in section [ section extending initial data ] , with @xmath555 satisfying ( [ norm : relationship ] ) , we set on @xmath54 @xmath556 consistently with the previous discussion of the initial data , one writes @xmath557 where we recall that both @xmath378 and @xmath558 are regarded as vector valued functions over @xmath559 . in particular , the quantities in @xmath560 describe the ( non - linear ) perturbations of the reference radiative spacetime . substituting the ansatz ( [ ansatz : perturbation ] ) into the evolution system ( [ sh : system ] ) one obtains a symmetric hyperbolic system for @xmath558 of the form @xmath561 with @xmath562 a matrix valued function with entries which are polynomials in @xmath558 of at most degree one and coefficients which are smooth functions on @xmath559 . as discussed earlier , the matrix valued functions @xmath219 are symmetric , have entries which are polynomial in @xmath221 of at most degree one and have constant coefficients . in contrast to the situation discussed in @xcite , the matrix @xmath563 is not diagonal . nevertheless , theorem [ katopart1 ] guarantees that there exists a neighbourhood @xmath564 such that @xmath565 . hence , @xmath566 the previous discussion allows us to formulate an existence and stability result for the solutions to equation ( [ sh : system : perturbation ] ) . [ katopart2 ] suppose @xmath322 . let @xmath567 be hyperboloidal initial data satisfying the technical condition ( [ technical : condition ] ) . given @xmath568 , there exists @xmath569 such that : * for @xmath570 there exist a unique solution @xmath571 to the conformal propagation equations ( [ p1])-([p7 ] ) and ( [ extended_cfe4 ] ) with minimal existence interval @xmath572 $ ] and @xmath573\times { \mbox{\sym s}}^3)$ ] . * the associated congruence of conformal geodesics contains no conjugate points in @xmath574 $ ] . * at the origin one has @xmath575 $ ] . * the hessian as given by ( [ hess1])-([hess3 ] ) is nondegenerate at @xmath576 . the solution @xmath577 on @xmath578 implies a @xmath579 solution @xmath580 to the vacuum einstein field equations with vanishing cosmological constant , where @xmath581 with @xmath8 given by ( [ quadratictheta ] ) . * by hypothesis , we know that @xmath582 . we thus satisfy the conditions of the variation of kato s theorem given in @xcite . parts ( i ) and ( ii ) follow directly from kato s theorem . parts ( iii ) and ( iv ) follow from the discussion in section [ section : structure : of : the : conformal : boundary ] . due to lemma [ lemma : propagation constraints ] on the propagation of the extended conformal constraints , one knows that the existence of a solution of the propagation system ( [ p1])-([p7 ] ) and ( [ extended_cfe4 ] ) implies a solution to the full extended conformal einstein field equations which in turns implies a solution to the vacuum einstein field equations with vanishing cosmological constant . @xmath583 one has the following corollary . [ katopart2a ] suppose the condition of theorem [ katopart2 ] are satisfied , then a solution @xmath580 , as given above , has a conformal boundary given by the set @xmath584 . the conformal boundary consists of the set @xmath510 , which represents future null infinity , and the point @xmath2 given by @xmath576 , which represents timelike infinity . hence the conformal boundary of radiative spacetimes is shown to be stable , subject to the conditions of theorem [ katopart2 ] . * remark . * note that the evolution of @xmath327 implies a unique solution @xmath214 and a spacetime , which is locally isometric with @xmath585 as obtained in proposition [ lemma : local existence ] . the conformal einstein field equations have been used to formulate an hyperboloidal initial value problem by means of which one can address the nonlinear stability of a class of purely radiative spacetimes . our analysis is particularly concerned with the structure of null infinity and the location of future timelike infinity , @xmath2 . the use of a conformal gaussian gauge system coordinates allows to calculate the location of the conformal boundary directly from the initial data , as well as analyse its structure . the results are a generalisation of previous results on minkowski - like spacetimes @xcite . in particular the conditions on the location of @xmath2 are given in a more general form , while the choice of initial data for the perturbed conformal geodesic congruence facilitates the stability analysis of the conformal boundary in comparison to @xcite . the reference purely radiative spacetimes are constructed from static initial data using procedure introduced by friedrich @xcite . recent results on the necessary and sufficient convergence conditions required on a series of _ multipolar data _ , one can conclude that there is an infinite family of reference purely radiative spacetimes . we observe that we were able to prove the stability of radiative spacetimes and study their conformal boundary without requiring any explicit knowledge of the reference spacetime . all our analysis has been carried out in the abstract only relying on general properties of these spacetimes . this raises the hope that similar results may be possible for other classes of spacetimes . cl would like to thank the leverhulme trust for a research project grant ( f/07 476/ai ) . javk is an epsrc advanced research fellow . we would like to thank thomas bckdahl for helpful discussions on multipole moments . h. friedrich , , in _ the geometric universe . science , geometry and the work of roger penrose _ , edited by s. a. huggett , l. j. mason , k. p. tod , s. t. tsou , & n. m. j. woodhouse , page 81 , oxford university press , 1999 .

an existence and stability result for a class of purely radiative vacuum spacetimes arising from hyperboloidal data is given . this result generalises semiglobal existence results for minkowski - like spacetimes to the case where the reference solution contains gravitational radiation . the analysis makes use of the extended conformal field equations and a gauge based on conformal geodesics so that the location and structure of the conformal boundary of the perturbed solutions is known _ a priori_.

introduction basics and conventions the conformal einstein field equations and conformal gaussian gauges purely radiative reference spacetimes radiative spacetimes from perturbed hyperboloidal data conclusions acknowledgements

markarian 501 ( mrk 501 ) is a bl lacertae ( bl lac ) object at _ z _ = 0.034 and was discovered as a @xmath0-ray source at @xmath6 350 gev by the whipple collaboration @xcite and subsequently confirmed by the hegra collaboration@xcite . in march 1997 the source went into a highly variable and strong emission state with flux doubling times of less than 0.5 day @xcite . the time period of the outburst coincided with the source visibility windows of several ground - based imaging atmospheric cerenkov telescopes ( iacts ) . thus almost continuous monitoring of mrk 501 in tev @xmath0-rays was possible with several iacts ( whipple , hegra , cat , tactic ) located in the northern hemisphere and they detected an integral flux of up to roughly 10 times that from the crab nebula above 1 tev@xcite . all sky monitor ( asm ) on board the _ rossi x - ray timing explorer _ ( rxte ) , also reported the high x - ray activity of the source which started in march 1997 and continued till october 1997 @xcite . the egret observations did not result in any detection of the source in high - energy @xmath0-ray range ( 30mev -30gev)@xcite in 1997 , though a year earlier , the same instrument had detected mrk 501 at 5.2 @xmath5 level for e@xmath10500 mev@xcite . there was an indication that the average optical u - band flux was higher in the month of the peak @xmath0-ray activity @xcite . observations of mrk 501 during 1998 - 1999 following the major 1997 outburst phase , reveal a mean emission level of @xmath11 of the crab nebula flux at 1 tev , a factor of 10 lower than the 1997 outburst . the vhe spectrum measured during this period is substantially softer ( by 0.44 @xmath12 0.1 in spectral index ) than the 1997 time - averaged spectrum @xcite . the celeste group has also detected a weak @xmath0-ray signal from mrk 501 ( 3.4@xmath13 ) using a low threshold system at themis during their 2000 observations@xcite . in 2003 - 04 hess observation of mrk 501 at large - zenith angle range for 1.8 hours , resulted in a marginal detection ( 3.1 @xmath5 ) of the source@xcite . the magic group has reported detection of two flaring episodes from the source on june 30 and july 09 , 2005 with flux values of and crab units respectively @xcite . among a number of interesting source related features this group has unveiled are the detection of the ic peak in the sed for the most active nights , flux variability increase with energy and an intriguing 4 minute shift in time ( on july 9 , 2005 ) of the higher energy flare compared to lower energies . + the tactic atmospheric cherenkov telescope is located at mt . abu ( 24.6@xmath14 n , 72.7@xmath14 e , 1300 m asl ) , a hill station in western india . it uses a tessellated light - collector of 9.5 @xmath15 area which is configured as a quasi - parabolic surface , yielding a measured spot - size of 0.3@xmath14 for on - axis parallel rays . the pc - controlled 2-axes drive system gives a pointing / tracking accuracy of better than @xmath123 arc - mins . the telescope deploys a 349-pixel , photomultiplier tube ( pmt ) based imaging camera with a uniform pixel resolution of @xmath16 0.3@xmath17 and a field of view of @xmath16 6@xmath14 @xmath1 6@xmath14 to record images of atmospheric cherenkov events produced by an incoming vhe cosmic - ray particle or a @xmath0-ray photon . present observations use the inner 225 pixels ( 15 @xmath1 15 matrix ) of the camera for the event - trigger generation , based on the nnp ( nearest neighbour pair ) topological logic . the data acquisition and control system of the telescope @xcite has been designed around a network of pcs running the qnx real - time operating system . the details of the telescope functioning are presented elsewhere @xcite . the telescope is sensitive to @xmath0-rays above 1 tev and can detect the crab nebula at 5@xmath5 significance level in 25 hours of on source observation . .observation log for mrk 501 [ cols="^,^,^,^,^ " , ] the differential energy spectrum obtained after applying the dynamic supercuts to the combined 2006 data is shown in fig . [ fig:501sp ] . a power law fit to the data ( @xmath7 ) in the energy range 1 - 11 tev yields @xmath8 and @xmath9 with a @xmath18 ( probability = 0.97 ) . again a low value of the reduced @xmath19 may be due to overestimation of the errors in the present measurements . the errors in the flux constant and spectral index are standard errors . we have also tried a power law with an exponential cutoff function of the form ( @xmath20 while fitting the observed source differential energy spectrum which , however , does not indicate any exponential cutoff feature in the energy range 1 - 11 tev of the source spectrum . one of the possible reasons for not having obtained exponential cutoff in the reported source energy spectrum could be the low statistics ( about 500 @xmath0-ray like events only ) detected with the tactic telescope . we have studied the bl lac object mrk 501 in vhe gamma - ray energy range with the tactic @xmath0-ray telescope during 2005- 06 . we do not find any evidence for the presence of a statistically significant vhe gamma - ray signal either in the overall data or when the data is analysed on the month to month basis or a night to night basis during 2005 observations . an upper limit of i(@xmath211 tev)@xmath22 photons @xmath3 s@xmath4 ( 28@xmath23 of the tactic detected crab nebula flux ) has been placed at a 3@xmath5 confidence level on the integrated @xmath0-ray flux . our results do not conflict with those of the magic group @xcite in 2005 as the two observations are not overlapping in time as mentioned earlier . these observations rather provide the source light curve prior to the magic detected two historical flares on june , 30 and july , 9 , 2005 . during 2006 observations we have detected a tev @xmath0-ray signal from the source direction with a statistical significance of 7.5@xmath5 in 66.8 hours of on - source observations . monthly spellwise analysis revealed that the source was relatively brighter during ii spell of 2006 observations as its statistical significance is close to 6@xmath5 in 27.71 hours of observations . data collected during three nights of march 26 , 27 and 28 , 2006 ( mjd 53820 , 53821 and 53822 ) do reveal hourly gamma - ray rate more than two times that of the crab nebula . excesses of gamma - ray like events obtained are 37.0 @xmath12 12.7 , 41.7 @xmath12 11.7 and 56.0 @xmath12 15.5 for 1.9 , 1.7 and 2.8 hours of observations respectively . these observations possibly suggest that the source has gone into a variable active state in the vhe region during the observations of aforementioned three nights . we have measured the source differential energy spectrum which fits well with the power law function of the form @xmath7 with @xmath8 and @xmath9 with a @xmath18 ( probability = 0.97 ) in the energy range 1 - 11 tev . we do not get any evidence for an exponential cutoff like feature when we use a power law function of the form @xmath20 while fitting the observed source differential spectrum in the aforementioned tactic energy range . in figure [ fig : comp ] we compare important source related vhe spectra results obtained during different epochs by various groups , whipple @xcite , hegra @xcite , cat@xcite and magic@xcite . as is clear from this figure , tactic 2006 results closely follow those obtained by the hegra group during the period 1998 - 99 , except for the exponential cutoff feature with @xmath24 = 2.61 tev . this source along with mrk 421 is very interesting as it is the second nearest bl lac object , which has been observed in highly variable states by aforementioned groups . in order to understand such highly violent extra galactic objects it is necessary to continue monitoring them for long durations with sensitive @xmath0-ray telescopes . the authors would like to convey their gratitude to all the concerned colleagues of the astrophysical sciences division for their contributions towards the instrumentation and observation aspects of the tactic telescope . we are thankful to dr . d. paneque of max - planck - institut fr physik , mnchen , germany , for providing useful inputs with respect to the magic mrk 501 work . we also thank the anonymous referees for their useful comments . aharonian , f. a. et al . 2001 , apj , 546 , 898 aharonian , f. a. et al . 2001 , a&a 366 , 62 aharonian f. et al . , 2005 a & a , 441 , 465 albert , j. et al . , 2007 , apj , 669 , 862 asm http://xte.mit.edu bradbury s. m. , et al . 1997 , a & a , 320 , l5 catanese , m . 1997 , apj , 487 , l143 catanese , m . 1998 , apj , 501 , 616 djannati - ata a. et al . 1999 , a & a , 350 , 17 hillas , a. m. , 1985 in proc 19th icrc , 3 , 445 kataoka , j. et al , 1999 , astroparticle physics , 11 , 149 koul , r. et al . , 2007 nucl . instrum . and meth . a 578 , 548 li , t.p . , & ma , y.q . 1983 , apj , 272 , 317 mohanty , g . 1998 , astroparticle physics , 9 , 15 protheroe r.j . , bhat , c.l . , fleury , p. , et al . , 1997 , 25th icrc , durban 1997 , astroph/9710118 quinn et al . 1996 , apj , 456 , l83 quinn et al . 1999 , apj , 518 , 693 reynolds , p.t.et al . , 1993 , apj , 404 , 206 . samuelson f.w . , et al . , 1998 , apj , 501 , l17 smith , d. a. et al . 2006 , a & a , 459 2,453 weekes t.c . , et al . , 1989 , apj , 342 , 379 yadav , k. k. et al . 2004 , nucl . instrum . and meth . a. , 527 , 411 yadav , k. k. et al . 2007 , astropart . phys . , 27 , 447

in this paper we report on the markarian 501 results obtained during our tev @xmath0-ray observations from march 11 to may 12 , 2005 and february 28 to may 7 , 2006 for 112.5 hours with the tactic @xmath0-ray telescope . during 2005 observations for 45.7 hours , the source was found to be in a low state and we have placed an upper limit of 4.62 @xmath1 10@xmath2 photons @xmath3 s@xmath4 at 3@xmath5 level on the integrated tev @xmath0-ray flux above 1 tev from the source direction . however , during the 2006 observations for 66.8h , detailed data analysis revealed the presence of a tev @xmath0-ray signal from the source with a statistical significance of 7.5@xmath5 above @xmath6 1 tev . the time averaged differential energy spectrum of the source in the energy range 1 - 11 tev is found to match well with the power law function of the form ( @xmath7 ) with @xmath8 and @xmath9 .

introduction experimental setup discussion and conclusions acknowledgements references

given a graph with preassigned edge lengths then a common problem is to determine if this weighted graph can be realized in @xmath1 . a _ graph _ @xmath2 is a pair @xmath3 where @xmath4 , known as the _ vertex set _ of @xmath2 , is a finite set , and @xmath5 , known as the _ edge set _ of @xmath2 , is a multiset whose elements are elements of @xmath6^{2}$ ] , the set of @xmath7-element subsets of @xmath4 . each edge @xmath8 is denoted @xmath9 in the sequel . in this note , graphs can have parallel edges but not loops . for further detail regarding graph theory , see @xcite . a length function on a graph @xmath2 is a function @xmath10 . a weighted graph is a pair @xmath11 where @xmath2 is a graph and @xmath12 is a length function on @xmath5 . given a weighted graph @xmath11 , then the _ configuration space _ @xmath13 of @xmath11 is defined as @xmath14 each @xmath15 contained in @xmath13 is called a _ realization _ of @xmath11 and if there exists a realization of @xmath11 , then the weighted graph @xmath11 is said to be _ realizable_. note that in the sequel , and particularly in figures , given a realization @xmath15 then @xmath16 is denoted @xmath17 . given a graph @xmath2 with vertex set @xmath4 then the group @xmath18 of _ orientation preserving isometries _ of @xmath1 acts on @xmath13 by @xmath19 for all @xmath20 given a weighted graph @xmath11 and the configuration space @xmath13 , then the _ moduli space _ @xmath21 of @xmath11 is the quotient space @xmath22 elements of a moduli space @xmath21 are equivalence classes and so are usually denoted by @xmath23 $ ] , however , whenever no confusion can arise , by a slight abuse of notation , the elements of @xmath21 are simply denoted @xmath15 in the sequel . + a subspace of a configuration space which is utilized in the sequel is now described . given a weighted graph @xmath11 , the vertices @xmath24 and @xmath25 in @xmath4 such that @xmath26 and that @xmath27 , then define @xmath28 note that @xmath29 and @xmath30 are different as sets but are homeomorphic topological spaces . observe that given a weighted graph @xmath11 then the space @xmath29 is homeomorphic to the moduli space @xmath21 . + the _ realizability problem _ for a weighted graph is the problem of establishing whether or not there exists a realization of @xmath11 and , in general , this problem is hard . note that this problem is sometimes referred to as the _ molecule problem _ and for further details on this see @xcite and @xcite . one of the simplest weighted graphs for which the realizability problem has been solved is @xmath31 , where @xmath32 is the complete graph on four vertices and this solution is now briefly outlined . consider @xmath31 , with vertex set @xmath33 and edge set @xmath34 . it is assumed throughout this section that the lengths assigned by @xmath35 are denoted as follows @xmath36 , @xmath37 , @xmath38 , @xmath39 , @xmath40 and @xmath41 . this notation is illustrated in fig . [ k1 ] . + @xmath42 it is well known , see @xcite for instance , that @xmath31 is realizable if and only if all cyclic permutations of the four inequalities @xmath43 , @xmath44 , @xmath45 and @xmath46 are satisfied and equation [ eqn1 ] holds . note that the determinant contained in equation [ eqn1 ] is known as the _ cayley - menger determinant . _ @xmath47 the fact that realizability conditions exist for the weighted graph @xmath31 appears to be something of a rarity as there does not appear to exist in the literature general realizability conditions , analogous to the @xmath31 case for other ( non - trivial ) weighted graphs . however , one recent development to this end , is a result contained in @xcite ( and will appear in @xcite ) which gives realizability conditions for weighted graphs where the graph is contained in the class of series - parallel graphs . + at this point the focus switches from the _ realizability problem _ to the following , more tractable , _ realizable extension problem_. given a realizable weighted graph @xmath48 where @xmath49 , then what conditions must an extension of @xmath50 , denoted @xmath12 , satisfy so that @xmath11 is realizable . observe that as every graph has a spanning tree ( or spanning forest if the graph is not connected ) then it is possible to state the following elementary existence result for such extensions . [ exten2 ] given a graph @xmath2 and a realizable weighted graph @xmath48 where @xmath49 , then it is possible to find an extension of @xmath50 , denoted @xmath12 , such that @xmath11 is realizable . the realizable extension problem is now examined in the case of the weighted graph @xmath0 . the reason for choosing @xmath0 is that this graph is essentially the simplest graph for which the realizable extension problem is non - trivial . with the exception of @xmath32 , for which the realizable extension problem is essentially trivial , all graphs smaller than @xmath51 are series - parallel and so the realizability problem and hence , the realizable extension problem , can be solved using the results of @xcite . + consider the weighted complete bi - partite graph @xmath52 , where @xmath53 @xmath54 and @xmath55 , @xmath56 . it is assumed throughout this section that the lengths assigned by @xmath57 are denoted @xmath58 , @xmath59 , @xmath60 , @xmath61 , @xmath62 , @xmath63 , @xmath64 , @xmath65 and @xmath66 . the values @xmath67 are not assumed to be fixed at this stage . this notation is illustrated in fig . + consider also the four specific subgraphs of @xmath51 which are defined as @xmath68 , @xmath69 , @xmath70 and @xmath71 which is a path . the former three of the aforementioned subgraphs of @xmath51 are shown in fig . + @xmath72 assuming that @xmath73 is given , thus fixing the edge lengths @xmath74 and @xmath75 , then determining conditions which the extensions @xmath76 and @xmath77 must satisfy so that @xmath78 , @xmath79 , @xmath80 and @xmath52 , respectively , are realizable , is the focus of the remainder of this section . [ lemmg0 ] given a weighted graph @xmath81 , as above , then @xmath78 is realizable if and only if @xmath82 assigns a value for @xmath83 such that @xmath84 $ ] as @xmath85 is a path then @xmath81 is always realizable . the graph @xmath86 is a cycle , and so @xmath87 is realizable if and only if the inequality @xmath88 , and all of the other five cyclic permutations of this inequality , are satisfied . choosing @xmath84 $ ] ensures all six inequalities are satisfied . given a realizable weighted graph @xmath78 , as above , and letting @xmath89 and @xmath90 , then @xmath79 is realizable if and only if @xmath91 assigns a value for @xmath92 such that @xmath93\cap [ \max\{0,\mu_{2}-(b+c+d)\},b+c+d]$ ] consider the paths @xmath94 and @xmath95 contained in @xmath96 with respective edge sets @xmath97 and @xmath98 . consider also the cycles @xmath99 and @xmath100 contained in @xmath101 with respective edge sets @xmath102 and @xmath103 . clearly @xmath104 is realizable if and only if both @xmath105 and @xmath106 are realizable _ and _ both @xmath107 and @xmath108 assign the same ( permissible ) value of @xmath92 to the edge @xmath109 . it now follows from lemma [ lemmg0 ] that @xmath93\cap [ \max\{0,\mu_{2}-(b+c+d)\},b+c+d]$ ] where @xmath89 and @xmath90 . before considering the weighted graph @xmath80 the concept of a _ workspace _ is introduced . for more details regarding workspaces see @xcite , @xcite or @xcite , where the concept first appears . given a weighted graph @xmath11 , then the _ workspace _ of a vertex @xmath110 with respect to the graph @xmath2 , the length function @xmath12 and an edge @xmath111 where @xmath27 , is defined as the image of the map @xmath112 i.e. @xmath113 where @xmath114 and @xmath115 is the restriction of @xmath12 induced by @xmath49 . + note that the moduli space @xmath116 is in fact a copy of @xmath117 . it is possible to construct an explicit homeomorphism @xmath118 as follows . for each @xmath23 \in m(h , l_{\mid h})$ ] , let @xmath119 be the unique realization in @xmath120 that satisfies @xmath121 , @xmath122 , and @xmath123=[p]$ ] in @xmath116 . it is now possible to define @xmath124 ) = q(v)$ ] . it is clear that @xmath125 is a homeomorphism . in the sequel , the map @xmath118 is used to identify the workspace of a vertex with a particular subset of @xmath126 . [ varphi ] given a realizable weighted graph @xmath79 , as above , then the subset of @xmath127 from which the value of @xmath128 can be chosen so that @xmath80 is realizable is an interval or the disjoint union of two intervals . given a weighted graph @xmath129 where @xmath130 is a cycle such that @xmath131 , then it is well known , see @xcite , that the image of @xmath132 has one of three types ; a circle @xmath133 with centre @xmath134 and radius @xmath135 , a contractible subset of @xmath133 or two disjoint contractible subsets of @xmath133 . all three of these subsets of @xmath133 are also symmetric about the @xmath136-axis i.e. @xmath137 where @xmath138 is the reflection in the @xmath136-axis . returning to the @xmath139 case at hand . consider the circle @xmath140 with centre @xmath141 and radius @xmath142 and the circle @xmath143 with centre @xmath144 and radius @xmath145 . observe that the images of @xmath146 and @xmath147 are subsets of circles @xmath140 and @xmath143 , respectively , and these images are denoted @xmath148 and @xmath149 , respectively , for the rest of this proof . the structure of the set @xmath150 is now determined . + consider the value @xmath151 and the value @xmath152 . a brief consideration of subsets of two circles ( centred on the @xmath136-axis ) which are symmetric about the @xmath136-axis leads to the conclusion that there is only one case where @xmath153 $ ] . this case is a special case of the instance where @xmath148 and @xmath149 are themselves two disjoint contractible subsets of @xmath140 and @xmath143 respectively . in order to describe this special case denote by @xmath154 the component of @xmath148 contained in the upper half - plane and denote by @xmath155 the component of @xmath148 contained in the lower half - plane . the components @xmath156 and @xmath157 of @xmath149 are defined similarly . now , consider the value @xmath158 and the value @xmath159 . the aforementioned special case occurs whenever @xmath160 and so the subset of @xmath127 from which the value of @xmath161 can be chosen so that @xmath80 is realizable is the disjoint union of two intervals @xmath162\sqcup[n , n]$ ] . consider fig . [ disjoint ] and note that the subsets @xmath163\sqcup[w_{3},w_{4}]$ ] and @xmath164\sqcup[w_{7},w_{8}]$ ] of the circles @xmath140 and @xmath143 , respectively , and let @xmath165 . + hence , the subset of @xmath127 from which the value of @xmath161 can be chosen so that @xmath80 is realizable is either an interval @xmath166 $ ] or the disjoint union of two intervals @xmath162\sqcup [ n , n]$ ] , where @xmath167 and @xmath168 are defined as above . given a realizable weighted graph @xmath80 , as above , then the subset of @xmath127 from which the value of @xmath169 can be chosen so that @xmath52 is realizable is an interval or the disjoint union of two , three or four intervals . consider a weighted graph @xmath170 , where @xmath171 and @xmath172 as shown in fig . [ k16b ] . observe that @xmath173 is homeomorphic to @xmath174 . if @xmath170 is realizable , then for every @xmath175 there exists a @xmath176 whose image is a reflection of the image of @xmath119 in the @xmath136-axis . observe that @xmath177 can have at most @xmath178 connected components . further motivation of this statement is provided in fig . the images of the realizations @xmath179 and @xmath180 of @xmath48 are shown , and there also exists four corresponding realizations @xmath181 and @xmath182 of @xmath48 in @xmath177 which are not shown . clearly it is possible to chose length functions @xmath183 and @xmath184 so that @xmath185 and @xmath186 and that the edge length @xmath187 is assigned an arbitrarily small length @xmath188 by @xmath183 . consider the image of the realization @xmath189 which is contained in fig . note that the incidence structure of the larger node , labeled @xmath190 , is shown in the detailed ( _ blown - up _ ) section contained in the circle on the right - hand - side of fig . + @xmath191 as a result of such length functions , then each connected component of the moduli space @xmath192 is a circle whereas each connected component of the moduli space @xmath173 is a point . the salient point here is that the connected components of @xmath173 and @xmath192 are in a one - to - one correspondence . + observe that if @xmath193 has @xmath194 connected components then these eight components must occur in pairs such that for any realization @xmath15 contained in one component of @xmath193 there exists a realization , denoted @xmath195 , in another component of @xmath193 such that the image of @xmath195 is a reflection of the image of @xmath15 in the @xmath136-axis . as a reflection is an isometry , then the distance @xmath196 must be equal to the distance @xmath197 , for each @xmath189 . hence , the subset of @xmath127 from which the value of @xmath169 can be chosen so that @xmath52 is realizable , can have at most four connected components . examples [ ex1 ] , [ ex2 ] , [ ex3 ] and [ ex4 ] are , respectively , occurrences of the subset of @xmath127 from which the value of @xmath198 can be chosen so that @xmath52 is realizable being one , two , three and four , disjoint intervals . this completes the proof . [ ex1 ] suppose that @xmath199 then the subset of @xmath127 from which the value of @xmath198 can be chosen so that @xmath52 is realizable is the set @xmath200 . see figure [ k13e ] . @xmath201 in examples [ ex2 ] , [ ex3 ] and [ ex4 ] the larger nodes labeled @xmath202 , @xmath203 , @xmath204 and @xmath205 are each analogous to the larger node contained in fig . [ k13a ] i.e. they possess the same incidence structure . it should also be noted that , in the interest of brevity , the images of the four realizations @xmath206 and @xmath182 of @xmath0 , ( whose images are the reflections in the @xmath136-axis of the images of @xmath15 , @xmath119 , @xmath207 and @xmath180 , respectively ) are omitted from fig . [ k13d ] , fig . [ k13c ] and fig . [ k13b ] . [ ex2 ] suppose that @xmath208 and @xmath209 then the subset of @xmath127 from which the value of @xmath169 can be chosen so that @xmath52 is realizable is @xmath210 $ ] @xmath211 @xmath212 $ ] where @xmath213 . see figure [ k13d ] . @xmath214 @xmath215 @xmath216 @xmath217 [ ex3 ] suppose that @xmath218 and @xmath219 then the subset of @xmath127 from which the value of @xmath169 can be chosen so that @xmath52 is realizable is @xmath220 $ ] @xmath211 @xmath221 $ ] @xmath211 @xmath222 $ ] where @xmath223 . see figure [ k13c ] . @xmath224 @xmath225 @xmath226 @xmath227 [ ex4 ] suppose that @xmath228 and @xmath229 then the subset of @xmath127 from which the value of @xmath169 can be chosen so that @xmath52 is realizable is @xmath230 $ ] @xmath211 @xmath231 $ ] @xmath211 @xmath232 $ ] @xmath211 @xmath233 $ ] where @xmath234 and @xmath235 are all distinct and @xmath236 for all distinct @xmath7-element subsets @xmath8 contained in @xmath237 . see figure [ k13b ] . @xmath238 @xmath239 @xmath240 @xmath241 this analysis of the @xmath52 case is now distilled into theorem [ k33thm ] . [ k33thm ] given the weighted graph @xmath81 , then the subset of @xmath127 from which the value of @xmath242 can be chosen so that @xmath78 is realizable is an interval ; having chosen @xmath243 and hence fixed @xmath83 , then the subset of @xmath127 from which the value of @xmath244 can be chosen so that @xmath79 is realizable is an interval ; having chosen @xmath245 and hence fixed @xmath92 , then the subset of @xmath127 from which the value of @xmath246 can be chosen so that @xmath80 is realizable is either an interval or the disjoint union of two intervals ; finally , having chosen @xmath247 and hence fixed @xmath161 , then the subset of @xmath127 from which the value of @xmath169 can be chosen so that @xmath52 is realizable is an interval or the disjoint union of two , three or four intervals . a nice corollary of the analysis of previous section is that it is possible to establish a result relating to the connectedness of the moduli space @xmath249 . given the weighted graph @xmath52 then the moduli space @xmath248 can only have one , two , four , six or eight connected components . as outlined above in relation to the weighted graph @xmath170 where @xmath250 contains three @xmath251-cycles , the moduli space @xmath248 can contain at most eight connected components . if @xmath0 is not realizable then the moduli space @xmath249 is the empty set and has , by definition , a single component . it is now required to show that the moduli space @xmath249 can not contain three , five or seven connected components . in a similar fashion to the proof of lemma [ varphi ] , let the image of @xmath252 be denoted @xmath253 for @xmath254 . clearly if a @xmath253 is connected then this does not imply that the corresponding moduli space is connected . however , if all @xmath253 , for @xmath254 , are connected then the corresponding moduli space must be connected . it follows that when @xmath249 is disconnected , there must exist at least one @xmath253 , for @xmath254 , which is disconnected . + each @xmath253 , for @xmath254 , is symmetric about the @xmath136-axis i.e. @xmath255 , where @xmath138 is the reflection in the @xmath136-axis . this means that if @xmath249 is disconnected then there exists some disconnected @xmath253 , for @xmath254 , such that the images of the realizations contained in the fibres of @xmath256 over @xmath257 are all reflections in the @xmath136-axis of the images of realizations contained in the fibres of @xmath258 over @xmath259 . hence , if @xmath249 is disconnected then the connected components of @xmath249 must occur in pairs where the images of realizations contained in these components differ by a reflection in the @xmath136-axis . as @xmath249 can have at most eight connected components , and as the empty set has one connected component , then @xmath249 can not contain three , five or seven connected components . + in example [ ex1 ] the moduli space @xmath248 is homeomorphic to the moduli space of an equilateral @xmath260-cycle @xmath261 . the moduli space @xmath262 is well known to be a connected space , see @xcite , hence the moduli space @xmath249 can have one ( non - empty ) component . in example [ ex2 ] there are three realizations @xmath263 and @xmath180 in which the length of @xmath198 is contained in the interval @xmath210 $ ] , with @xmath264 , such that there does not exist a continuous deformation between any two of the realizations @xmath263 and @xmath180 . this means that such a choice of @xmath198 results in the moduli space @xmath248 containing six connected components . in the same example , choosing @xmath198 to be contained in the interval @xmath265 $ ] , where @xmath266 , results in the moduli space @xmath248 containing two connected components . in example [ ex3 ] there does not exist a continuous deformation between realizations @xmath15 and @xmath207 . the length of @xmath198 is contained in the interval @xmath267 $ ] , where @xmath268 , and results in the moduli space @xmath248 containing four connected components . finally , example [ ex5 ] , below , illustrates that there exists a scenario where it is possible to choose a value for @xmath198 which results in the moduli space @xmath248 containing eight connected components . this completes the proof . [ ex5 ] suppose that @xmath269 and @xmath270 then the subset of @xmath127 from which the value of @xmath198 can be chosen so that @xmath0 is realizable is the interval @xmath271 $ ] where @xmath272 . observe that the moduli space @xmath248 has eight connected components . recall that @xmath248 is homeomorphic to @xmath273 . the images of realizations @xmath179 and @xmath180 which are each contained in distinct connected components of @xmath273 are shown in fig . again , the larger nodes labeled @xmath202 , @xmath203 , @xmath204 and @xmath205 are analogous to the larger node contained in fig . the reflection @xmath138 in the @xmath136-axis applied to the images of each of the realizations @xmath179 and @xmath180 results in the images of the four realizations @xmath206 and @xmath182 which are each contained in one of the remaining four distinct connected components of @xmath273 . ' '' '' the moduli space of a weighted cycle is a well understood object , see for example @xcite,@xcite , @xcite and @xcite . it may seem reasonable therefore that whenever a weighted graph @xmath11 contains weighted cycles that by determining realizability and/or connectedness results for certain weighted cyclic subgraphs of @xmath11 then these results may be extended to realizability and/or connectedness results relating to the weighted graph @xmath11 . this section contains two examples which show that properties of a moduli space @xmath274 are not necessarily possessed by the moduli space of @xmath21 where @xmath130 is a cyclic subgraph of @xmath2 . [ fourex2 ] consider the graph @xmath11 where @xmath275 @xmath276 , @xmath277 and @xmath12 assigns the lengths @xmath278 , @xmath279 , @xmath280 , @xmath281 , @xmath282 and @xmath283 . observe that such a length assignment results in the situation where the weighted graph @xmath11 is not realizable and so @xmath21 is empty . further justification of the fact that @xmath21 is empty can be found in fig . [ four8 ] which contains the weighted graph @xmath11 and two attempted realizations " of @xmath11 which are labeled @xmath284 and @xmath285 . observe that @xmath2 contains seven cyclic subgraphs and that all seven of these weighted cyclic subgraphs contained in @xmath11 are realizable . however , as the value of @xmath286 is always strictly positive in any attempted realization " of @xmath11 , for example @xmath284 and @xmath285 in fig . [ four8 ] , then @xmath11 is not realizable . this section contains an example which shows that even though the moduli spaces of some weighted cyclic subgraphs of a given @xmath11 are not connected , the moduli space @xmath21 may itself be connected . note that there exists a realization @xmath195 in the moduli space @xmath292 whose image is a reflection of the image of @xmath15 in the half - line containing the images @xmath293 and @xmath294 . given @xmath295 , whose images are illustrated in fig . [ five3 ] , then there does not exist a path @xmath296\to m(h , h)$ ] such that @xmath297 and @xmath298 , a path @xmath299\to m(h , h)$ ] such that @xmath300 and @xmath301,or a path @xmath302\to m(h , h)$ ] such that @xmath303 and @xmath304 . however , observe that there does exist a path @xmath305\to m(h , h)$ ] such that @xmath306 and @xmath307 , where @xmath308 is the realization whose image is the reflection of the image of @xmath119 in the line containing @xmath309 and @xmath310 . it follows that the moduli space @xmath292 has three connected components . + given @xmath48 as per fig . [ five3 ] , then consider the weighted graph @xmath11 where @xmath2 has vertex set @xmath311 , edge set @xmath312 and @xmath12 is an extension of @xmath50 which also assigns the lengths @xmath313 . + consider now the inclusion map @xmath315 . observe that @xmath316 is not surjective as neither @xmath15 nor @xmath195 , as per fig . [ five3 ] , are mapped onto by @xmath316 . note that the components of @xmath317 which contain the realizations @xmath15 and @xmath195 each contain just a single point . note also that there exists a path @xmath318\to m(g , l)$ ] such that @xmath319 and @xmath320 where @xmath321 is the realization of @xmath11 whose image is the reflection of the image of @xmath207 in the line containing @xmath322 and @xmath323 . it follows from these notes that @xmath21 is connected .

this note outlines the realizable extension problem for weighted graphs and provides results of a detailed analysis of this problem for the weighted graph @xmath0 . this analysis is then utilized to provide a result relating to the connectedness of the moduli space of planar realizations of @xmath0 . the note culminates with two examples which show that in general , realizability and connectedness results relating to the moduli spaces of weighted cycles which are contained in a larger weighted graph can not be extended to similar results regarding the moduli space of the larger weighted graph .

introduction the realizable extension problem for @xmath0 the moduli space @xmath248 moduli spaces of weighted cyclic subgraphs

measurements of the cosmic microwave background ( cmb ) anisotropies from the early days of the cosmic background explorer ( cobe ) satellite until the most recent observations of the wilkinson microwave anisotropy probe ( wmap)have consistently indicated a low value of the quadrupole , below the cosmic variance of the concordant @xmath3cdm cosmological model with a nearly scale invariant , primordial spectrum @xcite . while there has been a recurring debate about the statistical significance of the quadrupole and the other outliers ( notably , near the multipole moments @xmath4 and @xmath5 ) in the cmb angular power spectrum ( see refs . @xcite and references therein ) , there has also been constant activity to understand possible underlying physical reasons for the outliers ( see , for an inexhaustive list , refs . @xcite ) . given the cmb observations , different model independent approaches have been used to recover the primordial spectrum ( see , for example , refs . while all these approaches arrive at a spectrum that is nearly scale invariant at the smaller scales , most of them inevitably seem to point to a sharp drop in power at the scales corresponding to the hubble scale today . within the inflationary scenario , a variety of single and two field models have been constructed to produce such a drop in power at the large scales @xcite . however , in single field inflationary models , in order to produce such a spectrum , we find that many of the scenarios either assume a specific pre - inflationary regime , say , a radiation dominated epoch , or special initial conditions for the background scalar field , such as an initial period of fast roll @xcite . moreover , some of them impose the initial conditions on the perturbations when the largest scales are outside the hubble radius during the pre - inflationary or the fast roll regime @xcite . such requirements are rather artificial and , ideally , it would be preferable to produce the desired power spectrum during an inflationary epoch without invoking any specific pre - inflationary phase or special initial conditions for the inflaton . furthermore , though a very specific pre - inflationary phase such as the radiation dominated epoch may allow what can be considered as natural ( i.e. minkowski - like ) initial conditions for the perturbations even at super - hubble scales , we believe that choosing to impose initial conditions for a small subset of modes when they are outside the hubble radius , while demanding that such conditions be imposed on the rest of the modes at sub - hubble scales , can be considered unsatisfactory . it has long been known that power spectra with large deviations from scale invariance can be generated in inflationary models that admit one or more periods of departure from the slow roll phase ( see , for instance , refs . the degree of the deviation from a nearly scale invariant spectrum would be determined by the extent and duration of the departure , which are , in turn , controlled by the parameters of the model . a departure from slow roll affects the evolution of modes that leave the hubble radius just before the departure . rather than remaining constant , the curvature perturbations , say , @xmath6 , corresponding to these modes evolve at super - hubble scales , sourced by the intrinsic entropy perturbations of the inflaton field which , typically , exhibit a rapid growth during the fast roll regime @xcite . such an evolution on super - hubble scales results in dips or bursts of oscillations in the scalar power spectrum . usually , such a departure is induced by introducing a sharp feature in the potential of the inflaton field , such as a step or a sudden change in the slope @xcite . however , this is not necessary , and transitions to fast roll for brief periods can be generated even with smooth and better motivated effective potentials @xcite . our purpose in this paper is to present a simple model of inflation that supresses the power spectrum on large scales , a feature as we discussed above that seems to be necessary to fit the lower power in the quadrupole ( and , to some extent , in the deviant power at other lower multipoles such as the octopole and the multipole @xmath7 ) of the cmb angular power spectrum , using an effective potential of the canonical scalar field _ without _ introducing any ad hoc sharp feature . we find that the form of the potentials motivated by a class of certain minimal supersymmetric extensions of the standard model provide us with the desired behavior @xcite . these large field models allow a period of fast roll sandwiched between two stages of slow roll inflation . the first phase of slow roll inflation allows us to impose the standard bunch - davies initial conditions on the modes which exit the hubble radius during the subsequent fast roll regime , an epoch due to which the curvature perturbations on the super - hubble scales are suppressed . the second slow roll phase lasts for about @xmath8-@xmath9 @xmath10-folds , thereby allowing us to overcome the standard horizon problem associated with the hot big bang model . the advantages of our approach over other single field models mentioned earlier are twofold . firstly , we do not need to assume any specific pre - inflationary phase . the entire evolution of the inflationary era is described by a single inflaton potential and , therefore , is much simpler . secondly , the modes which exit the hubble radius during the fast roll regime are inside the hubble radius during the first stage of slow roll inflation and , hence , we do not have to impose any special initial conditions on the large scale modes . this paper is organized as follows . in sec . [ sec : bd ] , we shall review the essential features of the effective inflaton potential that we shall consider , and describe the background dynamics in situations of our interest . in sec . [ sec : ps ] , after an outline of the slow roll ` expectations ' of the scalar spectrum that can arise in such a background , we shall discuss the spectra that we obtain through numerical integration . in sec . [ sec : comparison ] , using the cosmological boltzmann code camb and the monte carlo code cosmomc , we shall compare the power spectra from the models we consider with the recent wmap 5-year data . finally , we shall close with sec . [ sec : discussion ] , wherein after a brief summary of our results , we shall discuss as to how the results from our model compare with those that have been obtained in another closely related single field model . in the discussions below , we shall set @xmath11 and @xmath12 as well as @xmath13 to unity . also , throughout , an overdot and an overprime shall denote differentiation with respect to the cosmic and the conformal times , respectively . the effective potential for the inflation field that we shall consider is described by two parameters @xmath14 and @xmath15 , and is given by v ( ) = ()^2 - ( ) ^n + ()^2(n-1 ) , [ eq : mssm - p ] where @xmath16 is an integer . such potentials are known to arise in certain minimal supersymmetric extensions of the standard model @xcite , and their role as an inflaton and its related effects have been studied recently @xcite . ( we should also hasten to add that the specific case of @xmath17 has been considered much earlier for reasons similar to ours , viz . producing certain features in the primordial spectrum @xcite . ) in the above potential , the coefficient of the @xmath18 term has been chosen in such a way that the potential has a point of inflection at @xmath19 ( i.e. the location where both @xmath20 and @xmath21 vanish ) , with @xmath22 given by _ 0 = ^. near this point of inflection , the potential exhibits a plateau with an extremely small curvature , which , as we shall discuss below , proves to be crucial for the desired evolution of the inflaton field . the potential ( [ eq : mssm - p ] ) for the case @xmath17 is depicted in fig . [ fig : mssm - p ] . -112 true pt -260 true pt @xmath23 100 true pt 15 true pt @xmath24 2 true pt note that the potential ( [ eq : mssm - p ] ) roughly behaves as @xmath25 recall that , the first potential slow roll parameter is given by @xcite _ 1 = ( ) ()^2 , and inflation ends as @xmath26 crosses unity . it is then clear that , in a power law potential of the form @xmath27 , slow roll inflation will occur ( i.e. @xmath28 ) when @xmath29 , and inflation will end when @xmath30 \sim \mathcal{o}(1)$ ] . thus , for a transition from slow roll to fast roll to occur , we need to choose the two parameters in the potential ( [ eq : mssm - p ] ) so that @xmath31 , i.e. of the order of the ( reduced ) planck scale . restarting inflation after the fast roll phase and the number of @xmath10-folds that can be achieved during the second phase of slow roll crucially depends on the value of @xmath32 . we rely on the numerics to choose this parameter carefully since the above slow roll estimate only provides a rough order of magnitude . choosing @xmath32 in such a way is actually fine tuning , but it seems to be inevitable if we are to achieve the desired slow - fast - slow roll transition as well as the required number of @xmath10-folds . once the point of inflection has been identified , we find that the normalization to the cmb angular power spectrum data provides the second constraint , thereby determining the value of the other free parameter @xmath14 . the equation of motion governing the scalar field described by the potential ( [ eq : mssm - p ] ) , when expressed as two first order equations for the coupled variables @xmath24 and @xmath33 , has one attractive fixed point located at the origin , i.e. at @xmath34 . for positive values of @xmath24 , we find that there exists an attractor trajectory towards which all other trajectories with arbitrary initial conditions on @xmath24 and @xmath33 quickly converge . for a suitably chosen @xmath32 , we find that the attractor trajectory exhibits two regimes of slow roll inflation sandwiching a period of fast roll . hence , if we start the evolution with @xmath35 , then the initial values of @xmath24 and @xmath36 prove to be irrelevant for the subsequent dynamics as they approach the attractor . this behavior is evident from fig . [ fig : pp ] where we have plotted the phase portrait for the @xmath37 case . -118 true pt -220 true pt @xmath33 107 true pt 75 true pt @xmath24 2 true pt once the field reaches close to @xmath32 , due to the extreme flatness of the potential ( [ eq : mssm - p ] ) , it relaxes and then moves very slowly , commencing the second stage of the slow roll inflation . this stage ends when the field finally rolls down towards the minima of the potential at @xmath38 . we had mentioned earlier that potentials of the type ( [ eq : mssm - p ] ) are encountered in the minimal supersymmetric standard model ( mssm ) @xcite , and that their role as an inflaton has been analyzed recently @xcite . at this point it is important that we highlight the differences between mssm inflation and the scenario we are considering . in mssm inflation , the point of inflection is located at sub - planckian values ( i.e. @xmath39 ) which can be an advantage as it avoids the problems associated with having super - planckian values for the field . in contrast , in our case , as emphasized above , the saddle point should be located around the planck scale ( i.e. @xmath40 ) , if we are to achieve the second period of slow roll before the end of inflation . however , in the mssm case , to have successful inflation , the initial values of @xmath24 and @xmath36 have to be finely tuned so that @xmath41 and @xmath42 . but , in our scenario , we do not require such fine tuning of the initial conditions on @xmath24 and @xmath36 . instead , we require for the location of @xmath32 . though the parameters of the potential that we work with are different from the mssm case , we nevertheless believe that it may be possible to realize the potential ( [ eq : mssm - p ] ) in theories beyond the standard model , such as , for instance , string theory ( in this context , see , for example , refs . @xcite ) . for example , it is known that the existence of a number of string axion fields can give rise to the following potential describing multi - field chaotic inflation @xcite : v(_i ) = _ i ( ) m_i^2 _ i^2 , with the initial field displacements smaller than unity . the dynamics and the inflationary predictions in such examples are surprisingly similar to the corresponding single field chaotic inflation models @xcite , due to the assisted inflation mechanism @xcite . similarly , with enough number of fields and with the non - renormalizable superpotential w = () ( ) , [ eq : sp ] and the corresponding @xmath43 term and the soft mass term , one might be able to build an inflation model that is effectively equivalent to the single field one described by the potential ( [ eq : mssm - p ] ) . ( note that , for clarity , we have temporarily restored @xmath44 in the expression ( [ eq : sp ] ) above . ) in this section , after providing general arguments for the form of the scalar spectra that we can expect from our model , we present the spectra evaluated numerically . let us begin by quickly summarizing the essential equations and the quantities that we are interested in @xcite . the curvature perturbation @xmath6 satisfies the differential equation _ k+2 ( ) r_k + k^2 r_k=0,[eq : derk ] where the quantity @xmath45 is given by z = (a/h ) . the quantity @xmath46 denotes the scale factor , @xmath24 the background inflaton , and @xmath47 is the conformal hubble factor given by @xmath48 . the scalar power spectrum @xmath49 is then defined as _ _ s(k ) = ( ) _ k^2 , with the amplitude of the curvature perturbation @xmath6 evaluated , in general , in the super - hubble limit . the tensor perturbation @xmath50 satisfies the equation _ k+2 ( ) u_k + k^2_k=0,[eq : deuk ] with the tensor power spectrum @xmath51 being given by _ _ t(k ) = ( ) _ k^2 , where , as in the scalar case , the tensor amplitude @xmath50 is evaluated at super - hubble scales . finally , the tensor - to - scalar ratio @xmath52 is defined as follows : r ( ) . before we evaluate the scalar spectra numerically , let us broadly try and understand the spectra that we can expect to arise in the slow - fast - slow roll scenario that we are interested in . consider modes that exit the hubble scale during an epoch of slow roll inflation . provided there is no deviation from slow roll soon after the modes leave the hubble radius , the amplitude of these modes will remain constant at super - hubble scales . therefore , their amplitude is determined by their value at hubble exit , and the scalar power spectrum corresponding to these modes can be expressed in terms of the potential as follows @xcite : _ _ s(k ) ( ) ( ) . however , if there is a period of deviation from slow roll inflation , then the asymptotic ( i.e. the extreme super - hubble ) amplitude of the modes that leave the hubble radius _ just before _ the deviation are enhanced when compared to their value at hubble exit @xcite . while modes that leave well before the deviation remain unaffected , it is found that there exists an intermediate range of modes whose amplitudes are actually _ suppressed _ at super - hubble scales @xcite . as a result , in the slow - fast - slow roll scenario of our interest , the scalar power spectrum is initially characterized by a sharp dip and a rise corresponding to modes that leave the hubble radius just before the transition to fast roll . then arises a regime of nearly scale invariant spectrum corresponding to modes that leave during the second stage of slow roll inflation . let us now understand the behavior of the tensor modes . in the case of the scalar modes , the quantity @xmath53 that appears in the differential equation ( [ eq : derk ] ) turns out to be negative during a period of fast roll , and it is this feature that proves to be responsible for the amplification or the suppression of the modes at super - hubble scales @xcite . in contrast , the coefficient of the friction term in the equation ( [ eq : deuk ] ) that describes the tensor modes , viz . @xmath54 , is a positive definite quantity . hence , we do not expect any non - trivial super - hubble evolution of @xmath50 . we find that , in the models that we consider , the tensor - to - scalar ratio @xmath52 remains smaller than @xmath55 over scales of cosmological interest , which is below the levels of possible detection by forthcoming missions such as planck @xcite . it is the background quantity @xmath53 that appears in the differential equation ( [ eq : derk ] ) for the curvature perturbation which essentially determines the form of the scalar power spectrum . the quantity @xmath56 can be expressed in terms of the first two hubble slow roll parameters , viz . @xmath57 and @xmath58 , with @xmath59 being the standard hubble parameter . it is given by @xcite ( ) = (1++ ) , and it is clear from this expression that , during slow roll inflation ( i.e. when @xmath60 and @xmath61 ) , the quantity @xmath56 will remain close to unity @xcite . in fig . [ fig : zpz ] , we have plotted the evolution of @xmath62 as a function of the number of @xmath10-folds @xmath63 for the cases of @xmath17 and @xmath64 in the potential ( [ eq : mssm - p ] ) . and , in fig . [ fig : ed ] , we have plotted the evolution of the field in the plane of the hubble slow roll parameters @xmath65 and @xmath66 for the @xmath17 case . -122 true pt -320 true pt 110 true pt 15 true pt 2 true pt -135 true pt -300 true pt @xmath66 125 true pt 5 true pt @xmath65 2 true pt it is manifest from these figures that the departure from slow roll occurs roughly between @xmath10-folds @xmath67 in the @xmath37 case and between @xmath10-folds @xmath68 for @xmath64 . we should also point out that inflation is actually interrupted for about a @xmath10-fold during the fast roll . in fig . [ fig : ps - n34 ] , we have plotted the corresponding scalar spectra evaluated numerically . -110 true pt -340 true pt @xmath49 -110 true pt -340 true pt @xmath49 105 true pt 10 true pt @xmath69 the broad arguments we had presented in the previous subsection are evidently corroborated by these two figures . note that , in plotting all these figures , we have chosen parameters that eventually provide the best fit to the wmap @xmath0-year data . also , in the inset in the top panel of fig . [ fig : ps - n34 ] , we have highlighted the difference between the scalar spectra in our model and the power law case ( i.e. when @xmath70 , with @xmath71 and @xmath72 ) . moreover , we should stress here that the standard sub - hubble , bunch - davies , initial conditions have been imposed on _ all _ the modes in arriving at these spectra . the scalar power spectrum with a drop in power at large scales is often approximated by an expression with an exponential cut off of the following form @xcite : _ _ s(k ) = a__s ( 1- ) k^n__s-1.[eq : ps - ecm ] in fig . [ fig : ps - n34 ] , we have also plotted this expression for values of @xmath73 , @xmath74 , @xmath75 and @xmath76 that closely approximate the spectra we obtain . it is useful to note that the spectra we obtain correspond to @xmath77 , @xmath78 , @xmath79 and @xmath80 when @xmath17 , while @xmath77 , @xmath81 , @xmath82 and @xmath83 in the @xmath64 case . we should emphasize here that we have arrived at these values for @xmath73 , @xmath74 , @xmath75 and @xmath76 by a simple visual comparison of the numerically evaluated result with the above exponentially cut off spectrum . in this section , we shall discuss as to how our model compares with the recent wmap @xmath0-year data . in the standard concordant cosmological model viz . the @xmath3cdm model with a power law inflationary perturbation spectrum six parameters are introduced when comparing the theoretical results with the cmb data ( see , for instance , ref . four of them are the following background parameters : the baryon density @xmath84 , the density of cold dark matter @xmath85 , the angular size of the acoustic horizon @xmath86 , and the optical depth @xmath87 , with @xmath88 denoting the hubble constant today ( viz . @xmath89 ) expressed in units of @xmath90 . the parameters that are introduced to describe the inflationary perturbation spectrum are the scalar amplitude @xmath73 and the scalar spectral index @xmath74 . the tensor - to - scalar ratio @xmath52 is also introduced as a parameter provided the ratio is sufficiently large , say , when @xmath91 . however , in the models we consider , the tensor - to - scalar ratio proves to be smaller than @xmath55 over the scales of cosmological interest . so , we completely ignore the contribution due to the gravitational waves in our analysis . we retain the standard background cosmological parameters , and we introduce the following three parameters to describe the inflationary perturbation spectrum : @xmath14 , @xmath22 and @xmath92 . while @xmath14 appears explicitly in the potential ( [ eq : mssm - p ] ) , @xmath22 has been chosen in place of @xmath15 . the quantity @xmath92 denotes the initial value of the scale factor ( i.e. at @xmath93 ) , and it basically determines the location of the cut - off in the power spectrum . thus , we have one additional parameter in comparison with the standard case . essentially , we have traded off the scalar amplitude @xmath73 for @xmath14 , and the scalar spectral index @xmath74 for @xmath22 . in tab . [ tab : priors ] , we have listed the ranges of uniform priors that we have imposed on the various parameters . .the priors on the various parameters describing the reference @xmath3cdm model with a power law primordial spectrum and our model . while the first four background cosmological parameters are common for both the models , the fifth and the sixth parameters describe the power law primordial spectrum of the reference model . as discussed in the text , in our model , we have traded off the scalar amplitude @xmath73 for @xmath14 and the spectral index @xmath74 for @xmath22 . the additional parameter in our model , viz . @xmath92 , represents the value of the scale factor at @xmath93 and it essentially identifies the location of the cut - off in the power spectrum . [ cols="^,^,^,^",options="header " , ] [ tab : bfv ] it is clear from figs . [ fig : lh - n3 ] and [ fig:2dc3 ] that the constraint on the parameter @xmath14 is prior dominated . in our model , it is the parameter @xmath14 that determines the amplitude of the power spectrum when it is nearly scale invariant . this amplitude , in turn , is essentially determined by the first peak of the cmb angular power spectrum . we should mention here that our choice of priors for the parameter @xmath14 has been arrived at by a simple visual fit to the first peak . in fig . [ fig : cl ] , we have plotted the angular power spectrum of the cmb temperature anisotropies for the best fit values of the parameters for the @xmath17 case . for comparison , we have also plotted the angular power spectrum for the best fit reference model . -165 true pt -340 true pt 50 true pt 10 true pt @xmath94 2 true pt it is immediately obvious from the figure that our model fits the lower multipoles much better than the reference model . as we have mentioned above , we obtain an improvement in @xmath1 of @xmath2 at the cost of introducing just one additional parameter when compared to the standard power law case . we should also emphasize here that the improvement in the fit that we have achieved is not only due to the cut - off in the scalar power spectrum , but also because of the presence of the oscillations at the top of the spectrum , just before it turns scale invariant . also , note the difference in the angular power spectrum for our model and the standard power law spectrum at the higher multipoles , which we have illustrated in the inset in fig . [ fig : cl ] . this disparity essentially arises due to the difference in the asymptotic spectral index in our model ( which proves to be about @xmath95 ) and the spectral index in the power law case ( which is about @xmath96 , cf . tab . [ tab : bfv ] ) . the planck mission @xcite is expected to provide more accurate data at these higher multipoles and , therefore , may aid us discriminate between these models better . in this section , after a quick summary of our results , we compare the results we obtain with those obtained in another single scalar field model that has been considered earlier . we emphasize the fact that the difference between these models has immediate observational consequences . in this work , we have investigated a two stage slow roll inflationary scenario sandwiching an intermediate period of deviation from inflation , driven by potentials that are similar in shape to certain mssm potentials @xcite . in the mssm case , inflation occurs when the field values are much smaller than the planck scale @xcite . however , in our case , since we demand two epochs of slow roll , it necessarily requires that the initial values of the field ( assuming , say , about @xmath9 @xmath10-folds of inflation ) be greater than @xmath97 . the period of fast roll period produces a sharp drop in the scalar power spectrum for the modes that leave the hubble radius just before the second slow roll phase . we choose our scales such that the drop in power corresponds to the largest cosmological scales observable today . we find that the resulting scalar power spectrum provides a much better fit to the recent wmap data than the canonical , nearly scale invariant , power law , primordial spectrum . at this stage , it is important that we compare our results with those obtained in another single field model that has been studied before . as we had mentioned in the introduction , an initial kinetic dominated ( i.e. fast roll ) stage preceding slow roll inflation , driven by a quadratic potential has been considered earlier to provide a sharp drop in the scalar power spectrum at large scales @xcite . at first glance , one may be tempted to conclude that the model we have studied here is equivalent to such a scenario if we disregard the first slow roll stage , since we have a kinetic dominated phase preceding a period of slow roll inflation . however , there are crucial differences between the two models which we have outlined below . to begin with , in the model we consider , there is no freedom to choose the type of fast roll ( say , the equation of state during the epoch of fast roll ) . it is fixed once we have chosen the parameters so as to fit the observations . secondly , in the scenario considered earlier , the modes which are outside the hubble radius during the kinetic dominated phase _ would have always remained so _ in the past @xcite . the authors assume that somehow there may have been a previous phase of inflation , during which they were inside the hubble radius and began life in the bunch - davies vacuum . while it is not impossible to think of situations where there may have been a previous inflationary epoch for instance , it can be achieved by invoking another scalar field @xcite the consequences can be quite different . in contrast , in the scenario that we have considered here , the standard , sub - hubble , bunch - davies , initial conditions have been imposed on _ all _ the modes . thirdly , it was argued that , since the suppression of power for the scalar spectrum proves to be sharper than that of the tensor , the tensor - to - scalar ratio @xmath52 displays a sharp rise towards large physical scales @xcite , a feature that may possibly be detected by upcoming missions such as , for instance , planck @xcite . however , in the models that we consider , the tensor amplitude on scales of cosmological interest proves to be too small ( @xmath98 ) to be detectable in the very near future . in conclusion , we would like to mention that a detection of the @xmath99 modes corresponding to , say , @xmath100 , can rule out the class of models that we have considered in this work . rkj and ls would like to thank the inter - university centre for astronomy and astrophysics ( iucaa ) , pune , india , and the korea institute for advanced study ( kias ) , seoul , korea , for hospitality , where part of this work was carried out . we would also like to acknowledge the use of the high performance computing facilities at the harish - chandra research institute , allahabad , india , as well as at iucaa and kias . we wish to thank anupam mazumdar , biswarup mukhopadhyaya and arman shafieloo for discussions . we also wish to thank hiranya peiris and alexei starobinsky for valuable comments on our results . we would also like to thank tuhin ghosh , minu joy , juhan kim , subharthi ray and arman shafieloo for help with the numerical codes at various stages . jg is partly supported by the korea research foundation grant krf-2007 - 357-c00014 funded by the korean government . finally , we acknowledge the use of the cosmomc package @xcite and the data products provided by the wmap science team @xcite . 99 c. l. bennett _ et . astrophys . j. * 436 * , 423 ( 1994 ) ; e. l. wright , g. f. smoot , c. l. bennett and p. m. lubin , arxiv : astro - ph/9401015 . e. l. wright , c. l. bennett , k. gorski , g. hinshaw and g.f . smoot , astrophys . j. * 464 * , l21 ( 1996 ) ; m. tegmark , astrophys . j. lett . * 464 * , l35 ( 1996 ) . h. v. peiris _ et al . _ , astrophys . j. suppl . * 148 * , 213 ( 2003 ) . g. hinshaw _ et al . _ , astrophys . j. suppl . * 170 * , 288 ( 2007 ) . m. r. nolta _ et al . _ , arxiv:0803.0593 [ astro - ph ] . odwyer _ et al . _ , astrophys . j. * 617 * , l99 ( 2004 ) . j. magueijo and r. d. sorkin , mon . not . * 377 * , l39 ( 2007 ) . park , c. park and j. r. gott iii , astrophys . j. * 660 * 959 ( 2007 ) . chiang , p. d. naselsky and p. coles , arxiv:0711.1860 [ astro - ph ] . a. a. starobinsky , sov . jetp lett . * 55 * , 489 ( 1992 ) . y. p. jing and l. z. fang , phys . lett . * 73 * , 1882 ( 1994 ) . b. feng and x. zhang , phys . lett . b * 570 * , 145 ( 2003 ) . g. efsthathiou , mon . not . r. astron . soc . * 343 * , l95 ( 2003 ) . j. p. luminet _ et al . _ , nature ( london ) * 425 * , 593 ( 2003 ) . a. hajian and t. souradeep , astrophys . j. lett . * 597 * , l5 ( 2003 ) ; a. hajian , t. souradeep and n. cornish , astrophys . j. * 618 * , l63 ( 2005 ) . m. kawasaki and f. takahashi , phys . lett . b * 570 * , 151 ( 2003 ) ; m. kawasaki , f. takahashi and t. takahashi , _ ibid . _ * 605 * , 223 ( 2005 ) . c. gordon and w. hu , phys . d * 70 * , 083003 ( 2004 ) . c. r. contaldi , m. peloso , l. kofman and a. linde , jcap * 0307 * , 002 ( 2003 ) . j. m. cline , p. crotty and j. lesgourgues , jcap * 0309 * , 010 ( 2003 ) . t. moroi and t. takahashi , phys . lett . * 92 * , 091301 ( 2004 ) . p. hunt and s. sarkar , phys . d * 70 * , 103518 ( 2004 ) ; _ ibid . _ * 76 * , 123504 ( 2007 ) . s. shankaranarayanan and l. sriramkumar , phys . rev . d * 70 * , 123520 ( 2004 ) . s. shankaranarayanan and l. sriramkumar , arxiv : hep - th/0410072 . l. sriramkumar and t. padmanabhan , phys . rev . d * 71 * , 103512 ( 2005 ) . r. sinha and t. souradeep , phys . d * 74 * , 043518 ( 2006 ) . l. covi , j. hamann , a. melchiorri , a. slosar and i. sorbera , phys . rev.d * 74 * , 083509 ( 2006 ) ; j. hamann , l. covi , a. melchiorri and a. slosar , phys . d * 76 * , 023503 ( 2007 ) . d. boyanovsky , h. j. de vega and n. g. sanchez , phys . d * 74 * , 123006 ( 2006 ) ; _ ibid _ * 74 * , 123007 ( 2006 ) . j. f. donoghue , k. dutta and a. ross , arxiv : astro - ph/0703455 . b. a. powell and w. h. kinney , phys . d * 76 * , 063512 ( 2007 ) . g. nicholson and c. r. contaldi , jcap * 0801 * , 002 ( 2008 ) . m. joy , v. sahni , a. a. starobinsky , phys . d * 77 * , 023514 ( 2008 ) ; m. joy , a. shafieloo , v. sahni , a. a. starobinsky , arxiv:0807.3334 [ astro - ph ] . c. destri , h. j. de vega and n. g. sanchez , phys . rev . d * 78 * , 023013 ( 2008 ) . s. l. bridle , a. m. lewis , j. weller and g. efstathiou , mon . not . soc . * 342 * , l72 ( 2003 ) . p. mukherjee and y. wang , astrophys . j. * 599 * , 1 ( 2003 ) . s. hannestad , jcap * 0404 * , 002 ( 2004 ) . a. shafieloo and t. souradeep , phys . rev . d 70 , 043523 ( 2004 ) ; a. shafieloo , t. souradeep , p. manimaran , p. k. panigrahi and r. rangarajan , _ ibid _ * 75 * , 123502 ( 2007 ) ; a. shafieloo and t. souradeep , _ ibid _ * 78 * , 023511 ( 2008 ) . d. tocchini - valentini , y. hoffman and j. silk , mon . not . * 367 * , 1095 ( 2006 ) . h. m. hodges , g. r. blumenthal , l. a. kofman and j. r. primack , nucl . b * 335 * , 197 ( 1990 ) . v. f. mukhanov and m. i. zelnikov , phys . lett . b * 263 * , 169 ( 1991 ) . t. souradeep , ph.d . thesis , university of pune ( inter - university centre for astronomy and astrophysics ) , pune , india ( 1995 ) . j. a. adams , g. g. ross and s. sarkar , nucl . phys . b * 503 * , 405 ( 1997 ) . j. lesgourgues , nucl . phys . b * 582 * , 593 ( 2000 ) . j. barriga , e. gaztanaga , m. santos and s. sarkar , mon . not . roy.astron . soc . * 324 * , 977 ( 2001 ) ; nucl . suppl . * 95 * , 66 ( 2001 ) . j. a. adams , b. cresswell , r. easther , phys . rev . d * 64 * , 123514 ( 2001 ) . gong , jcap * 0507 * , 015 ( 2005 ) . s. m. leach and a. r. liddle , phys . d * 63 * , 043508 ( 2001 ) ; s. m. leach , m. sasaki , d. wands and a. r. liddle , _ ibid . _ * 64 * , 023512 ( 2001 ) . r. k. jain , p. chingangbam and l. sriramkumar , jcap * 0710 * , 003 ( 2007 ) . r. allahverdi , j. garcia - bellido , k. enqvist and a. mazumdar phys . * 97 * , 191304 ( 2006 ) . r. allahverdi , k. enqvist , j. garcia - bellido , a. jokinen and a. mazumdar , jcap * 0706 * , 019 ( 2007 ) . j. c. b. sanchez , k. dimopoulos and d. h. lyth , jcap * 0701 * , 015 ( 2007 ) . r. allahverdi , a. mazumdar and t. multamaki , arxiv:0712.2031 [ astro - ph ] . l. a. kofman , a. d. linde and a. a. starobinsky , phys . b * 157 * , 361 ( 1985 ) . j. silk and m. s. turner , phys . d * 35 * , 419 ( 1987 ) ; m. s. turner , j. v. villumsen , n. vittorio , j. silk and r. juszkiewicz , astrophys . j. * 323 * , 423 ( 1987 ) . d. polarski and a. a. starobinsky , nucl . b * 385 * , 623 ( 1992 ) ; d. polarski , phys . d * 49 * , 6319 ( 1994 ) ; d. polarski and a. a. starobinsky , phys . b * 356 * , 196 ( 1995 ) . d. langlois , phys . d * 59 * , 123512 ( 1999 ) . s. tsujikawa , d. parkinson and b. a. bassett , phys . d * 67 * , 083516 ( 2003 ) . a. r. liddle and d. h. lyth , _ cosmological inflation and large - scale structure _ ( cambridge university press , cambridge , england , 1999 ) ; v. f. mukhanov , _ physical foundations of cosmology _ ( cambridge university press , cambridge , england , 2005 ) . h. kodama and m. sasaki , prog . suppl . * 78 * , 1 ( 1984 ) ; v. f. mukhanov , h. a. feldman and r. h. brandenberger , phys . * 215 * , 203 ( 1992 ) ; j. e. lidsey , a. liddle , e. w. kolb , e. j. copeland , t. barreiro and m. abney , rev . mod . phys . * 69 * , 373 ( 1997 ) ; b. bassett , s. tsujikawa and d. wands , rev . * 78 * , 537 ( 2006 ) . n. itzhaki and e. d. kovetz , jhep * 0710 * , 054 ( 2007 ) . a. ashoorioon and a. krause , arxiv : hep - th/0607001 ; a. ashoorioon , a. krause and k. turzynski , arxiv:0810.4660 [ hep - th ] . s. dimopoulos , s. kachru , j. mcgreevy and j. g. wacker , jcap * 0808 * , 003 ( 2008 ) . l. alabidi and d. h. lyth , jcap * 0605 * , 016 ( 2006 ) s. a. kim and a. r. liddle , phys . d * 74 * , 023513 ( 2006 ) . gong , phys . d * 75 * , 043502 ( 2007 ) . a. r. liddle , a. mazumdar and f. e. schunck , phys . d * 58 * , 061301 ( 1998 ) . see , http://www.rssd.esa.int/index.php?project=planck . a. lewis , phys . d * 78 * , 023002 ( 2008 ) . a. lewis , a. challinor and a. lasenby , astrophys . j. * 538 * , 473 ( 2000 ) . see , http://camb.info/. see , http://lambda.gsfc.nasa.gov/ a. lewis and s. bridle , phys . d * 66 * , 103511 ( 2002 ) . see , http://cosmologist.info / cosmomc/.

we investigate inflationary scenarios driven by a class of potentials which are similar in form to those that arise in certain minimal supersymmetric extensions of the standard model . we find that these potentials allow a brief period of departure from inflation sandwiched between two stages of slow roll inflation . we show that such a background behavior leads to a step like feature in the scalar power spectrum . we set the scales such that the drop in the power spectrum occurs at a length scale that corresponds to the hubble radius today a feature that seems necessary to explain the lower power observed in the quadrupole moment of the cosmic microwave background ( cmb ) anisotropies . we perform a markov chain monte carlo analysis to determine the values of the model parameters that provide the best fit to the recent wmap @xmath0-year data for the cmb angular power spectrum . we find that an inflationary spectrum with a suppression of power at large scales that we obtain leads to a much better fit ( with just one extra parameter , @xmath1 improves by @xmath2 ) of the observed data when compared to the best fit reference @xmath3cdm model with a featureless , power law , primordial spectrum .

introduction the inflaton potential and the background dynamics the scalar power spectrum comparison with the recent wmap @xmath0-year data summary and discussion references

it is known that the trivial connection contributions of chern simons perturbation theory ( @xcite etc . ) for homology 3-spheres give very fine topological invariants , which are fine enough to dominate all real - valued finite type invariants . in @xcite , t. ohtsuki defined a @xmath0-equivariant perturbative invariant of 3-manifolds with @xmath5 and proved that it is also very fine for 3-manifolds with @xmath5 . this suggests that the theory of perturbative quantum invariants is quite rich also for 3-manifolds with @xmath6 . shortly after ohtsuki s results appeared , c. lescop developed another theory of @xmath0-equivariant perturbative invariant using configuration spaces ( @xcite ) , which can be seen as a @xmath0-equivariant version of chern simons perturbation theory of homology 3-spheres . in @xcite , two kinds of interpretations of @xmath0-equivariant invariant from configuration spaces are given : 1 . as a @xmath0-equivariant invariant of knots in a 3-manifold . this takes values in a space @xmath7 of jacobi diagrams ( definition [ def : jacobi ] ) whose edges are colored by rational functions in a formal variable @xmath8 . 2 . as an invariant of 3-manifolds with @xmath5 by considering the equivalence class of the @xmath0-equivariant invariant of knots with respect to some equivalence relation in the target space . lescop proved in @xcite in a topological method that the invariant of ( 1 ) is universal among garoufalidis rozansky s finite type invariants of knots in an integral homology 3-sphere defined by null - claspers ( @xcite ) , which are related to the loop expansion of kontsevich s universal vassiliev invariant of knots . lescop s knot invariant can be considered as a configuration space version of kricker s rational invariant @xmath9 for knots defined combinatorially ( @xcite ) , which is universal among finite type invariants of @xcite . she also proved in @xcite that for @xmath10 with @xmath11 and with trivial alexander polynomial , the invariant of ( 2 ) coincides with the 2-loop part of the invariant of ohtsuki defined in @xcite for 3-manifolds with @xmath5 , up to normalization . in the definitions of lescop s invariants , a knot @xmath12 in a 3-manifold @xmath10 is fixed and a certain 3-cycle @xmath13 in the boundary of the equivariant configuration space @xmath14 of 2 distinct points in @xmath10 ( @xmath0-covering of the configuration space @xmath15 of 2 distinct points in @xmath10 , definition in [ ss : equiv - conf ] ) , which is associated to @xmath12 , is considered . then the 3-cycle has an extension to a 4-chain in @xmath14 with coefficients in the field of rational functions , which is called an equivariant propagator . for example , the term for the @xmath16-graph in lescop s equivariant invariant is given by the `` equivariant triple intersection '' in @xmath14 among equivariant propagators . in this paper , we apply lescop s construction to the explicit equivariant propagator of `` al - paths '' given in @xcite . namely , we introduced in @xcite the notion of al - paths in a 3-manifold @xmath10 fibered over @xmath2 ( definition in [ ss : al ] ) and gave an explicit 4-chain @xmath17 in @xmath14 by the moduli space of al - paths of a fiberwise gradient @xmath18 of a fiberwise morse function on @xmath10 . roughly , an al - path of @xmath18 is a piecewise smooth path consisting of segments each of which is either a part of a critical locus of @xmath18 ( vertical segment ) or a flow line of @xmath19 ( horizontal segment ) . then , for example , the equivariant triple intersection of ( parallel copies of ) @xmath17 is given by a generating function of counts of certain graphs in @xmath10 such that each edge is an al - path ( al - graphs , figure [ fig : al - graph - generic ] ) . we shall give two results . first , we take a fiberwise morse function on @xmath10 as an extra structure to define an @xmath7-valued invariant . we show that our invariant @xmath1 is an invariant of concordance classes of fiberwise morse functions on @xmath10 ( definition in [ ss : fmf ] ) and of spin structures on @xmath10 ( theorem [ thm : main ] ) , by a bifurcation argument similar to hatcher wagoner @xcite . roughly , concordance of fiberwise morse function is analogous to isotopy of closed braid in @xmath10 . though the concordance relation looks too restricted , it is enough for defining equivariant perturbative isotopy invariants for nullhomologous knots in @xmath10 . we will write in a subsequent paper @xcite about the knot invariants which count al - graphs with univalent vertices attached to a knot . next , concerning ( 2 ) , we modify our @xmath17 in order to utilize lescop s result because unlike lescop s equivariant propagator in @xcite the boundary of @xmath17 does not concentrate on a knot @xmath12 . by adding a bordism in @xmath20 between @xmath21 and a multiple of a 2-sphere bundle over @xmath12 , we give a combinatorial formula for lescop s 3-manifold invariant @xmath3 in @xcite for fibered 3-manifolds with @xmath22 as a generating function of counts of al - graphs ( theorem [ thm : q ] ) with some geometric correction terms . we remark that lescop s invariant is actually nontrivial , as shown in @xcite . by using the combinatorial formulas for @xmath1 and @xmath3 , we derive surgery formulas of the invariant @xmath1 ( theorem [ thm : surgery ] ) and of @xmath3 ( theorem [ thm : surgery2 ] ) . the surgery formula suggests that @xmath1 has a property similar to finite type invariant as does the perturbative invariant for homology 3-spheres ( @xcite ) . this gives another proof of lescop s surgery formula of @xmath3 in @xcite for special kind of 3-manifolds and surgeries , which is simple in the sense that the formula is obtained easily by counting certain graphs in a 3-manifold . in computing the values of @xmath1 and @xmath3 for concrete examples , we show that the homology of the mapping torus of a diffeomorphism of a closed surface is given by counting al - paths ( proposition [ prop : al - homology ] ) . the count of al - paths between fibers of a surface bundle gives the twisted tensor product of k. igusa ( @xcite ) , which in this setting computes the homology of the mapping torus . in this paper , manifolds and maps between them are assumed to be smooth unless otherwise noticed . by an @xmath23-dimensional _ chain _ in a manifold @xmath24 , we mean a _ finite _ linear combination of smooth maps from oriented compact @xmath23-manifolds with corners to @xmath24 . we understand a chain as a chain of smooth simplices by taking triangulations of manifolds . let @xmath25 denote the group of piecewise smooth chains in @xmath24 and let @xmath26 . we represent an orientation of a manifold @xmath24 by a non - vanishing section of @xmath27 and denote it by @xmath28 . we consider a coorientation @xmath29 of a submanifold @xmath30 of a manifold @xmath24 as an orientation of the normal bundle of @xmath30 and represent it by a differential form in @xmath31 . we identify the normal bundle @xmath32 with the orthogonal complement @xmath33 in @xmath34 , by taking a riemannian metric on @xmath24 . we equip orientation or coorientation of @xmath30 so that the identity @xmath35 holds , where we say that two orientations @xmath36 and @xmath37 are equivalent ( @xmath38 ) if they are related by multiple of a positive function . @xmath39 determines @xmath29 up to equivalence and vice versa . we orient boundaries of an oriented manifold by the inward normal first convention . for a sequence of submanifolds @xmath40 of a smooth riemannian manifold @xmath41 , we say that the intersection @xmath42 is _ transversal _ if for each point @xmath43 in the intersection , the subspace @xmath44 spans the direct sum @xmath45 , where @xmath46 is the orthogonal complement of @xmath47 in @xmath48 with respect to the riemannian metric . let @xmath49 $ ] denote the ring of laurent polynomials in @xmath23 variables . let @xmath50 denote the field of fractions of @xmath51 $ ] and we identify @xmath50 with a subset of the field of formal power series in @xmath8 with finitely many negative degree terms . in [ s : preliminaries ] , we review the definitions of relevant graphs , concordance of fiberwise morse functions , al - paths and equivariant propagator . in [ s : invariants ] , we give definitions of two invariants @xmath1 and @xmath3 , which are mainly based on lescop s construction in @xcite . we claim that one is an invariant of concordance classes of fiberwise morse functions on a fibered 3-manifold over @xmath2 . the other one gives a combinatorial formula for lescop s invariant for fibered 3-manifolds with @xmath5 in @xcite . in [ s : invariance ] , we prove that @xmath1 is invariant under a concordance of fiberwise morse functions by a bifurcation argument . in [ s : sformula ] , we give a surgery formula of @xmath1 . namely , we consider a special kind of torelli surgery and we give an explicit formula of the value of @xmath1 for an alternating sum of surgeries . in [ s : homology ] , we define a chain complex by counting al - paths . the homology of the chain complex is naturally isomorphic to the homology of the total space of the fibration . this is used to understand the surgery formula . by a _ graph _ , we mean a finite connected graph with each edge oriented , i.e. , an ordering of the boundary vertices of an edge is fixed . we allow multiple edges and self - loops . a _ labeled graph _ is a graph @xmath52 equipped with bijections @xmath53 and @xmath54 , where @xmath55 is the set of vertices of @xmath52 and @xmath56 is the set of edges of @xmath52 . we will identify @xmath55 and @xmath56 with the sets of labels through @xmath57 and @xmath58 respectively . let @xmath59 denote the subset of @xmath56 consisting of self - loop edges and let @xmath60 . an _ orientation _ of a graph is an orientation of the real vector space @xmath61 where @xmath62 is the two - element set of ` half - edges ' , namely @xmath63 $ ] and @xmath64 $ ] for an orientation preserving homeomorphism @xmath65 $ ] . a labeled graph @xmath66 gives a canonical orientation @xmath67 . vertex - orientation _ of a vertex @xmath68 in a graph is a cyclic ordering of the edges incident to @xmath68 . let @xmath69 be a commutative ring with 1 . for a trivalent graph @xmath52 , an _ @xmath69-coloring _ of @xmath52 is an assignment of an element of @xmath69 to every edge of @xmath52 . an @xmath69-coloring is represented by a map @xmath70 . the _ degree _ of a trivalent graph is defined as half the number of vertices . [ def : jacobi ] let @xmath71 $ ] and @xmath72 . let @xmath73 ( resp . @xmath7 ) be the vector space over @xmath74 spanned by pairs @xmath75 , where @xmath52 is a trivalent graph of degree @xmath23 with vertex - orientation and @xmath76 is a @xmath77-coloring ( resp . @xmath78-coloring ) of @xmath52 , quotiented by the relations as , ihx , orientation reversal , linearity , holonomy ( figure [ fig : relations ] ) and automorphisms of oriented graphs . it is known that there is a canonical correspondence between an orientation of a trivalent graph and a vertex - orientation to each vertex ( e.g. , @xcite ) . we denote a pair @xmath75 by @xmath79 or by @xmath80 . we say that a @xmath77-colored graph @xmath79 is a _ monomial _ if for each edge @xmath81 of @xmath52 , @xmath82 is a power of @xmath8 . in this case , we may consider @xmath76 as a map @xmath83 by identifying @xmath84 with @xmath85 . there is a bijective correspondence between the equivalence class of a monomial labeled graph @xmath79 modulo the holonomy relation and the homotopy class of a continuous map @xmath86 , or the cohomology class @xmath87\in h^1(\gamma;\z)=[\gamma , s^1]$ ] . let @xmath88 be a smooth fiber bundle with fiber diffeomorphic to a closed connected oriented 2-manifold @xmath89 . we equip @xmath10 with a riemannian metric . we fix a fiberwise morse function @xmath90 and its gradient @xmath91 along the fibers that satisfies the parametrized morse smale condition , i.e. , the descending manifold loci and the ascending manifold loci for @xmath18 are mutually transversal in @xmath10 . we consider only fiberwise morse functions that are _ oriented _ , i.e. , the bundles of negative eigenspaces of the hessians along the fibers on the critical loci are oriented . there always exists an oriented fiberwise morse function on @xmath10 ( e.g. , @xcite ) . the graph of critical values of @xmath92 forms a diagram in @xmath93 ( cerf s graphic ) , which consists of closed immersed curves . see figure [ fig : al - path ] . for a generic choice of @xmath94 , the intersection of curves in its graphic consists of transversal double points , which are called a _ level exchange points_. a _ generalized morse function ( gmf ) _ is a @xmath95 function on a manifold with only morse or birth - death singularities ( ( * ? ? ? * appendix ) ) . fiberwise gmf _ is a @xmath95 function @xmath90 whose restriction @xmath96 is a gmf for all @xmath97 . a _ critical locus _ of a fiberwise gmf is the subset of @xmath10 consisting of critical points of @xmath98 , @xmath97 . a fiberwise gmf is _ oriented _ if it is oriented outside birth - death loci and if birth - death pairs near a birth - death locus have incidence number 1 . it is known that for a pair of fiberwise morse functions @xmath99 , there exists a homotopy @xmath100}$ ] between @xmath101 and @xmath102 in the space of oriented gmf s on @xmath10 which gives an oriented fiberwise gmf on the surface bundle @xmath103\to s^1\times [ 0,1]$ ] . we say that the homotopy @xmath104 is a _ concordance _ if each birth - death locus of @xmath105 in @xmath106 $ ] projects to a closed curve that is not nullhomotopic in @xmath107 $ ] . there exists a pair @xmath108 of oriented fiberwise morse functions as above that are homotopic through a family of oriented gmf s but not concordant because there may be a birth - death locus for a homotopy with nullhomotopic projection in @xmath109 $ ] which can not be removed by deformation . in @xcite , we considered a fiberwise gradient of a fiberwise morse function as a kind of a limit of the nonsingular vector field @xmath110 . in this way , a concordance would correspond to a certain isotopy of such a vector field . a birth - death locus in a concordance would correspond to a birth or a cancellation of closed orbits . let @xmath111 be the @xmath0-covering associated to @xmath112 . let @xmath113 be the lift of @xmath112 , @xmath114 denote the @xmath0-invariant lift of @xmath94 , and @xmath115 denote the lift of @xmath18 . we say that a piecewise smooth embedding @xmath116\to m$ ] is _ descending _ if @xmath117 and @xmath118 , where @xmath119\to\wm$ ] is a lift of @xmath120 . we say that @xmath120 is _ horizontal _ if @xmath121 is included in a single fiber of @xmath112 and say that @xmath120 is _ vertical _ if @xmath121 is included in a critical locus of @xmath94 . [ def : al - path ] let @xmath122 be two points of @xmath10 . al - path from @xmath43 to @xmath123 _ is a sequence @xmath124 , where 1 . for each @xmath125 , @xmath126 is a descending embedding @xmath127\to m$ ] for some real numbers @xmath128 such that @xmath129 , 2 . for each @xmath125 , @xmath126 is either horizontal or vertical with respect to @xmath94 , 3 . if @xmath126 is horizontal , then @xmath126 is a flow line of @xmath19 , possibly broken at critical loci , 4 . @xmath130 , @xmath131 , 5 . @xmath132 for @xmath133 , 6 . if @xmath126 is horizontal ( resp . vertical ) and if @xmath134 , then @xmath135 is vertical ( resp . horizontal ) . we say that two al - paths are _ equivalent _ if they differ only by reparametrizations on segments . see figure [ fig : al - path ] for an example of an al - path . we remark that we do not allow @xmath126 to be a constant map . for an al - path @xmath124 , we write @xmath136 we recall the fulton macpherson type compactification of configuration spaces of real oriented manifolds by kontsevich . see @xcite for detail . for a closed @xmath137-dimensional manifold @xmath10 , the configuration space @xmath138 is the complement of the closed subset @xmath139 there is a natural filtration @xmath140 with @xmath141 the difference @xmath142 is a disjoint union of smooth submanifolds of @xmath143 . this property allows one to iterate ( real ) blow - ups along the filtration from the deepest one : first , one can consider the blow - up @xmath144 along the @xmath137-dimensional submanifold @xmath145 of @xmath146 with oriented normal bundle . recall that a blow - up replaces a submanifold with its normal sphere bundle . since the closure of @xmath147 in @xmath144 is also a disjoint union of smooth submanifolds ( with boundaries ) , one can apply another blow - up along it , and so on . after the blow - ups along all the strata of @xmath89 of codimension @xmath148 , one obtains a smooth compact manifold with corners @xmath149 . the space @xmath149 has a natural stratification corresponding to bracketings of the @xmath23 letters @xmath150 , e.g. , @xmath151 ( see @xcite ) . roughly speaking , a pair of brackets corresponds to a face created by one blow - up . for example , the face corresponding to @xmath151 is obtained by a sequence of blow - ups corresponding to a sequence @xmath152 . the codimension one ( boundary ) strata of @xmath149 correspond to bracketings of the form @xmath153 , with only one pair of brackets . the stratum @xmath154 of @xmath155 corresponding to the bracketings @xmath156 is the face created by the blow - up along the closure of the following submanifold of @xmath146 : @xmath157 in the result of the previous blow - ups . more precisely , @xmath154 can be naturally identified with the blow - ups of the total space of the normal @xmath158-bundle of @xmath159 along the intersection with the closures of deeper diagonals that correspond to deeper bracketings . the fiber of the normal @xmath158-bundle over a point @xmath160 is @xmath161 where the coordinate @xmath162 corresponds to @xmath163 ( where it makes sense ) via a local framing of @xmath164 . the stratum @xmath154 is a fiber bundle over @xmath165 . we denote the fiber of @xmath154 over a point of @xmath165 by @xmath166 . we identify @xmath166 with the subset of @xmath167 , as @xmath168 we denote by @xmath169 the closure of the image of the inclusion @xmath170 , which is compact . the base space @xmath165 is naturally diffeomorphic to @xmath171 and we denote by @xmath172 the projection @xmath173 . so @xmath154 has the structure of the pullback of the associated @xmath169-bundle of @xmath174 by @xmath175 : @xmath176 \ar[d ] & p\times_{so(d)}\bconf_j^\loc(\r^d ) \ar[d ] \\ \delta_j \ar[r]^-{\mathrm{pr}_j } & m } \ ] ] ( @xmath177 is the orthonormal frame bundle of @xmath174 . ) the definition of @xmath178 for general subset @xmath179 corresponding to the bracketing @xmath180 is similar . it will turn out that the faces of @xmath155 corresponding to coincidence of two points are special among the codimension one strata of @xmath149 . we denote by @xmath181 ( ` pri ' for principal ) the union of the faces corresponding to coincidence of two points and we denote by @xmath182 ( ` hi ' for hidden ) the union of all the faces corresponding to coincidence of at least three points . let @xmath52 be a labeled graph with @xmath23 vertices and @xmath183 edges . by the labeling @xmath53 , we identify @xmath56 with the set of ordered pairs @xmath184 , @xmath185 . let @xmath186 denote the set of tuples @xmath187 where @xmath188 and @xmath189 is the homotopy class of continuous maps @xmath190\to s^1 $ ] relative to the endpoints such that @xmath191 and @xmath192 . we consider @xmath186 as a topological space as follows . let @xmath193 be the space of continous maps @xmath194 equipped with the @xmath195-topology and let @xmath196 be the space that is the pullback in the following commutative square . \ar[d ] & c^0(\gamma , s^1 ) \ar[d]^-{\nu}\\ m^{n } \ar[r]^-{\kappa\times\cdots\times\kappa } & ( s^1)^{n } } \ ] ] here , @xmath198 assigns to each @xmath199 the images of the @xmath23 vertices of @xmath52 under @xmath76 . the fiberwise quotient map @xmath200 by the homotopy relation of edges gives @xmath186 the quotient topology . the forgetful map @xmath201 is a locally trivial fibration , which is a @xmath202-covering . there is a canonical bijection between the set of connected components in @xmath186 with @xmath203\approx \z^{1-\chi(\gamma)}$ ] . note that the covering @xmath186 may depend on @xmath112 if @xmath204 . let @xmath205 be the space obtained from @xmath186 by blowing - up along all the lifts of the diagonals in @xmath146 as in [ ss : fm ] . the forgetful map @xmath206 is a @xmath202-covering . since @xmath205 is naturally a @xmath202-space by the covering translation , the twisted homology @xmath207 where @xmath208 $ ] and @xmath209 ( tensor product of @xmath74-modules ) , is defined . here , @xmath210 acts on @xmath211 by @xmath212 . if @xmath52 is the complete graph @xmath213 with 2 vertices , @xmath214 is a @xmath0-covering . the associated twisted homology is @xmath215 . if @xmath52 is a labeled trivalent graph with @xmath216 vertices , we obtain a @xmath217-covering @xmath218 . there is a canonical bijection between the set of connected components in @xmath205 with @xmath219 . the associated twisted homology is @xmath220 , where @xmath221 $ ] and @xmath222 . following @xcite , we will consider below an intersection form among chains in the equivariant configuration space @xmath223 . it will be necessary to take a fundamental chain for the intersection called an _ equivariant propagator_. here , we take a special one from @xcite among equivariant propagators . let @xmath18 be the fiberwise gradient of an oriented fiberwise morse function on @xmath10 . we say that an al - path @xmath224 in @xmath10 with positive length with respect to the riemannian metric on @xmath10 is a _ closed al - path _ if the endpoints of @xmath224 coincide . a closed al - path @xmath224 gives a piecewise smooth map @xmath225 , which can be considered as a `` closed orbit '' in @xmath10 . we will also call @xmath226 a closed al - path . a closed al - path has an orientation that is determined by the orientations of descending and ascending manifolds loci of @xmath227 . see [ ss : coori ] for detail . then we define the sign @xmath228 and the period @xmath229 of @xmath224 by @xmath230,[\bar{\gamma}]\rangle|,\quad \ve(\gamma)=\frac{\langle[d\kappa],[\bar{\gamma}]\rangle}{|\langle[d\kappa],[\bar{\gamma}]\rangle|}.\ ] ] let @xmath231 be the subbundle of @xmath174 of unit tangent vectors . let @xmath232 be the pullback @xmath233 , which can be considered as a piecewise smooth 3-dimensional chain in @xmath234 . we say that two closed al - paths @xmath235 and @xmath236 are _ equivalent _ if there is a degree 1 homeomorphism @xmath237 such that @xmath238 . the indices of vertical segments in a closed al - path must be all equal since an al - path is descending . we define the index @xmath239 of a closed al - path @xmath224 to be the index of a vertical segment ( critical locus ) in @xmath224 , namely , the index of the critical point of @xmath92 for any @xmath97 that is the intersection of @xmath224 with @xmath240 . let @xmath241 and let @xmath242 be the normalization @xmath243 of the section @xmath19 . the closure @xmath244 in @xmath231 is a smooth manifold with boundary whose boundary is the disjoint union of circle bundles over the critical loci @xmath224 of @xmath18 , for a similar reason as ( * ? ? ? * lemma 4.3 ) . the fibers of the circle bundles are equators of the fibers of @xmath232 . let @xmath245 be the total space of the 2-disk bundle over @xmath224 whose fibers are the lower hemispheres of the fibers of @xmath232 which lie below the tangent spaces of the level surfaces of @xmath112 . then @xmath246 as sets . let @xmath247 this is a 3-dimensional piecewise smooth manifold . we orient @xmath248 by extending the natural orientation @xmath249 on @xmath250 induced from the orientation @xmath251 of @xmath10 . the piecewise smooth projection @xmath252 is a homotopy equivalence and @xmath248 is homotopic to @xmath253 . let @xmath12 be a knot in @xmath10 such that @xmath254,[k]\rangle=-1 $ ] . let @xmath255 be the set of all al - paths in @xmath10 . there is a natural structure of non - compact manifold with corners on @xmath255 . for a closed al - path @xmath224 , we denote by @xmath256 the minimal closed al - path such that @xmath224 is equivalent to the iteration @xmath257 for a positive integer @xmath258 and we call @xmath256 the _ irreducible _ factor of @xmath224 . this is unique up to equivalence . if @xmath259 , we say that @xmath224 is irreducible . we orient @xmath260 so that @xmath261=p(\gamma^\irr)[st(k)]$ ] . note that this may not be the one naturally induced from the orientation of @xmath256 but from @xmath262 . [ thm : propagator ] let @xmath10 be the mapping torus of an orientation preserving diffeomorphism @xmath263 of closed , connected , oriented surface @xmath89 . let @xmath18 be the fiberwise gradient of an oriented fiberwise morse function @xmath90 . 1 . there is a natural closure @xmath264 of @xmath255 that has the structure of a countable union of smooth compact manifolds with corners whose codimension 0 strata are disjoint from each other . 2 . let @xmath265 be the evaluation map , which assigns the pair of the endpoints of an al - path @xmath224 together with the homotopy class of @xmath266 relative to the endpoints . let @xmath267 denote the blow - up of @xmath264 along @xmath268 . then @xmath269 induces a map @xmath270 and it represents a 4-dimensional @xmath78-chain @xmath17 in @xmath14 that satisfies the identity @xmath271 where the sum is taken over equivalence classes of closed al - paths in @xmath10 . moreover , @xmath272 is a @xmath77-chain , where @xmath273 is the alexander polynomial of the fibration @xmath88 . 3 . suppose that @xmath112 induces an isomorphism @xmath274 . let @xmath12 be a knot in @xmath10 such that @xmath254,[k]\rangle=-1 $ ] . then the homology class of @xmath21 in @xmath275 is @xmath276=[s_{\xi}^*(m)]+\frac{t\zeta'_{\varphi}}{\zeta_{\varphi}}[st(k)],\ ] ] where @xmath277 is the lefschetz zeta function of @xmath278 . we recall the orientation convention for @xmath17 in theorem [ thm : propagator ] . as in @xcite , we give the orientations of the strata by coorientations in auxiliary spaces . the dimensions of strata in @xmath17 of al - paths having vertical segments of index 0 or 2 degenerates in @xmath14 . a nondegenerate ( codimension 0 ) stratum @xmath279 in @xmath17 consists of al - paths @xmath124 satisfying one of the following conditions . 1 . @xmath280 and both @xmath281 and @xmath282 are horizontal . 2 . @xmath283 and @xmath281 is horizontal . in the case ( 1 ) , @xmath224 may have several horizontal segments between critical loci of index 1 . each such horizontal segment is the transversal intersection of the descending manifold ( locus ) @xmath284 of a critical locus @xmath85 and the ascending manifold ( locus ) @xmath285 of another critical locus @xmath286 . such an intersection is called a _ @xmath287-intersection _ in @xcite . a @xmath287-intersection has a sign which is determined by the coorientations of the descending and the ascending manifolds by the identity @xmath288 where @xmath289 is the level surface locus of @xmath94 including @xmath43 and @xmath290 . let @xmath291 be the product of the signs of @xmath287-intersections on @xmath224 . then we define the coorientation @xmath292 at generic point as @xmath293)}\conf_{k_2}(m)=\twedge^\bullet t^*_xm\otimes \twedge^\bullet t^*_ym . \end{split}\ ] ] in the case ( 2 ) , suppose that @xmath224 goes from @xmath294 to @xmath295 . if @xmath296 is spanned by an orthonormal basis @xmath297 , then @xmath298 is spanned by @xmath299 , @xmath300 , @xmath301 , @xmath302 , where @xmath303 . let @xmath304 be the dual basis for @xmath297 and let @xmath305 be the pushforward . then we define @xmath306 and @xmath307 where @xmath308 is the hodge star operator . we define the 1-cycle @xmath309 in @xmath10 , where the sum is over equivalence classes of all closed al - paths @xmath224 for @xmath18 considered as oriented 1-cycles . this is an infinite sum but is well - defined as a @xmath78-chain . the orientation of a closed al - path @xmath224 is given by @xmath291 times the downward orientation on @xmath224 . let @xmath88 be a smooth fiber bundle with fiber a closed connected oriented 2-manifold . fix a compact connected oriented 2-submanifold @xmath89 of @xmath10 without boundary such that the oriented bordism class of @xmath89 in @xmath10 corresponds to @xmath312 $ ] via the canonical isomorphism @xmath313 . let @xmath52 be a labeled oriented 3-valent graph of degree @xmath23 . 1 . if @xmath314 , then let @xmath315 denote the projection that gives the endpoints of @xmath81 together with the associated curve in @xmath2 . take a compact oriented 4-submanifold @xmath316 in @xmath14 with corners . if @xmath317 , then let @xmath318 denote the projection that gives the unique endpoint of @xmath81 together with the associated curve in @xmath2 . take a compact oriented 1-submanifold @xmath316 in @xmath10 with boundary . note that in both cases @xmath319 is of codimension 2 . then we define @xmath320 which gives a compact 0-dimensional submanifold in @xmath223 if the intersection is transversal . we equip each point @xmath321 of @xmath322 with a coorientation ( a sign ) in @xmath205 by @xmath323 here , we identify a neighborhood of a point in @xmath223 with its image of the projection in @xmath324 . the coorientation gives a sign as the sign of @xmath325 in the equation @xmath326 , where @xmath327 is the standard orientation of @xmath324 . by this , @xmath322 represents a 0-chain in @xmath205 . this can be extended to tuples of codimension 2 @xmath74-chains by multilinearity . we will denote the homology class of @xmath322 ( integer ) by the same notation . we extend the form @xmath328 to tuples of codimemsion 2 @xmath77-chains @xmath329 in @xmath14 or in @xmath10 as follows . when @xmath330 , suppose that @xmath329 is of the form @xmath331 , where @xmath332 and @xmath333 is a compact oriented smooth 4-submanifold in @xmath334 $ ] with corners , where @xmath334 $ ] is the subspace of @xmath14 consisting of @xmath335 such that @xmath336 is represented by an arc in @xmath10 whose algebraic intersection with @xmath89 is @xmath337 . when @xmath338 , suppose that @xmath329 is of the form @xmath331 , where @xmath333 is a piecewise smooth path in @xmath10 transversal to @xmath89 and @xmath339 , @xmath340 . then we define @xmath341 which can be considered as a 0-chain in @xmath342 with coefficients in @xmath343 $ ] . this is multilinear by definition . next , we extend the form @xmath328 to tuples of codimension 2 @xmath78-chains in @xmath14 or @xmath10 as follows . let @xmath344 be codimension 2 @xmath78-chains in @xmath14 or @xmath10 depending on whether the corresponding edge is not a self - loop or a self - loop . then there exist @xmath345 such that @xmath346 is a @xmath77-chain for each @xmath125 . then we define @xmath347 which can be considered as a 0-chain in @xmath342 with coefficients in @xmath348 . this does not depend on the choices of @xmath349 and this is multilinear by definition . note that the multilinear form @xmath328 depends on the choice of @xmath89 . we define a @xmath74-linear map @xmath350 by setting @xmath351.\ ] ] this induces a linear map @xmath352 for the theta - graph @xmath16 in ( [ eq : theta ] ) , the definition of @xmath353 given above is equivalent to the equivariant triple intersection @xmath354 defined in @xcite . take three 4-dimensional @xmath77-chains @xmath355 in @xmath14 and represent as linear combinations of 4-chains in @xmath334 $ ] as follows . @xmath356 where @xmath357 are 4-submanifold of @xmath334 $ ] that are mutually transversal and the sums are finite . then we have @xmath358}\ , \tr_\theta(t_1^{\ell}\otimes t_2^{\ell'}\otimes t_3^{\ell''})\\ & = \sum_{\ell,\ell',\ell''}\bigl\langle f_x^{(\ell)},f_y^{(\ell')},f_z^{(\ell'')}\bigr\rangle_{\bconf_{k_2}(m)[0]}\ , \tr_\theta(1\otimes t_2^{\ell'-\ell}\otimes t_3^{\ell''-\ell})\\ & = \sum_{\ell , i , j\in\z}\bigl\langle f_x^{(\ell)},f_y^{(\ell+i)},f_z^{(\ell+j)}\bigr\rangle_{\bconf_{k_2}(m)[0]}\ , \tr_\theta(1\otimes t_2^i\otimes t_3^j)\\ & = \sum_{i , j\in\z}\bigl\langle f_x , t^{-i}f_y , t^{-j}f_z\bigr\rangle_{\bconf_{k_2}(m)}\ , \tr_\theta(1\otimes t_2^i\otimes t_3^j).\\ \end{split}\ ] ] this agrees with lescop s equivariant triple intersection in @xcite . let @xmath360 denote the set of concordance classes of oriented fiberwise morse functions for a fibration @xmath88 . let @xmath361 be fibrations isotopic to @xmath112 . let @xmath362 , @xmath363 , be oriented fiberwise morse functions for @xmath364 such that @xmath365 is concordant to a pair isotopic to @xmath366 . let @xmath367 be the fiberwise gradient of @xmath368 . let @xmath369 be the equivariant propagator in theorem [ thm : propagator ] for @xmath367 and let @xmath370 be as in [ ss : moduli_closed ] . choosing these data generically , we define @xmath371 where @xmath372 in the term for @xmath52 is @xmath369 or @xmath370 depending on whether the corresponding edge in @xmath52 is not a self - loop or a self - loop and the sum is over all labeled 3-valent graphs of degree @xmath23 for all possible edge orientations . we also denote by @xmath359 the homology class of @xmath359 in @xmath373 . we need a correction term to turn the homology class of @xmath359 into an invariant under a deformation . choose a spin structure @xmath374 on @xmath10 . recall that rohlin s @xmath325-invariant for @xmath375 is defined by @xmath376 where @xmath41 is a compact spin 4-manifold that spin bounds @xmath375 . melvin s @xmath377-invariant ( @xcite ) of @xmath375 is given by @xmath378 where @xmath379 . in @xcite , it is shown that if @xmath41 is a simply - connected spin 4-manifold that spin bounds @xmath375 , for example constructed by attaching 2-handles to @xmath380 along an even framed link , then @xmath381 ( mod 2 ) . in @xcite , the invariant @xmath377 was used to determine a canonical stable framing of @xmath382 that is compatible with @xmath374 . for simplicity , we assume the following condition . [ assum : mu0 ] @xmath383 ( mod 16 ) . if @xmath384 ( mod 16 ) , the correction term may still be defined as follows . consider the 16-fold cyclic covering @xmath385 which is the pullback of @xmath112 by the 16-fold covering @xmath386 . then @xmath387 induces a natural spin structure @xmath388 on @xmath389 and we have @xmath390 ( mod 16 ) . now define the correction term for @xmath375 to be @xmath391 of that for @xmath392 . this definition is consistent when @xmath383 ( mod 16 ) . under assumption [ assum : mu0 ] , one has @xmath393 or @xmath394 ( mod 4 ) , depending on whether @xmath395 is odd or even . if @xmath393 ( mod 4 ) , then by @xcite there is a unique stable framing @xmath76 such that @xmath396 where @xmath120 is the signature defect @xmath397 for @xmath41 as above and @xmath398 is the degree of the map @xmath399 which gives the direction of the framing @xmath400 of @xmath401 considered with respect to the trivialization @xmath76 . if @xmath402 ( mod 4 ) , then by @xcite there is a unique stable framing @xmath76 such that @xmath403 the latter is the same situation as @xmath0-homology sphere with canonical spin structure discussed in @xcite . [ lem : cframing ] under assumption [ assum : mu0 ] , we have the following . 1 . if @xmath395 is odd , then there is a spin 4-manifold @xmath41 that spin bounds @xmath375 and a unique framing @xmath278 of @xmath174 such that @xmath404 hence by ( * ? ? ? * lemma 2.3 ) , the stable framing @xmath405 extends to a framing of @xmath406 that is compatible with the spin structure . if @xmath395 is even , then there is a spin 4-manifold @xmath41 that spin bounds @xmath375 and a unique stable framing @xmath76 of @xmath407 such that @xmath408 hence by ( * ? ? ? * lemma 2.3 ) , the stable framing @xmath76 extends to a framing of @xmath406 that is compatible with the spin structure . we only give a proof for case ( 1 ) since the proof for case ( 2 ) is similar . by assumption [ assum : mu0 ] , there is a spin 4-manifold @xmath41 that spin bounds @xmath375 with @xmath409 ( mod 16 ) . since @xmath395 is odd , such a @xmath41 has even euler characteristic . let @xmath410 denote the kummer @xmath410 surface and @xmath411 be the 4-torus , both spinnable 4-manifolds . by connected sums of several @xmath410 or @xmath412 to @xmath41 , we may assume that @xmath413 , since @xmath414 . note that for a 4-manifold @xmath24 , one has @xmath415 , @xmath416 . hence we may assume that @xmath417 by connect summing several @xmath418 and @xmath411 , without changing the signature . now choose the canonical stable framing @xmath76 of @xmath407 with @xmath419 and @xmath420 , which is uniquely determined . for the @xmath41 as above , we have @xmath421 . by @xmath420 , there is a framing @xmath278 of @xmath174 such that @xmath76 is homotopic to @xmath422 . this completes the proof . choose @xmath41 and @xmath76 as in lemma [ lem : cframing ] . when @xmath395 is odd , one can find a 4-framing of @xmath406 and its sub 3-framing @xmath423 of @xmath406 that extends @xmath278 . the 3-framing @xmath423 spans a rank 3 subbundle @xmath424 of @xmath406 . we extend @xmath425 to a map @xmath426 , which restricts to a section on @xmath427 . we choose @xmath428 generic as in @xcite so that we may define @xmath429\in\cala_n(\widehat{\lambda}).\ ] ] see ( * ? ? ? * definition 2.7 ) for the definition of @xmath430 . roughly , @xmath430 is the moduli space of graphs in fibers in the vector bundle @xmath424 whose @xmath125-th edge is parallel to @xmath431 . one can prove that @xmath432 does not depend on the choices of @xmath41 as in lemma [ lem : cframing ] and of the extension @xmath433 ( ( * ? ? ? * proposition 2.12 ) ) . now we define @xmath434 when @xmath395 is even , let @xmath76 be the canonical stable framing of @xmath382 as in lemma [ lem : cframing ] ( 2 ) . this is in a similar situation as the case of @xmath0-homology spheres with canonical spin structure considered in @xcite . the stable framing @xmath76 is obtained from the stabilization @xmath422 of an honest framing @xmath278 of @xmath435 ( @xmath436 : base point ) by modifying it on a neighborhood of @xmath437 to a fixed stable framing near @xmath437 . then we may also define the correction term @xmath432 and @xmath310 as above . see @xcite for detail . [ thm : main ] @xmath310 is an invariant of @xmath438,[f])$ ] , where 1 . @xmath374 is a spin structure on @xmath10 , 2 . @xmath312\in h^1(m)$ ] is the homotopy class of a fibration @xmath88 , 3 . @xmath439\in \cale(\kappa)$ ] is the concordance class of an oriented fiberwise morse function @xmath90 . we will denote @xmath359 and @xmath1 by @xmath440,[f])$ ] and @xmath441,[f])$ ] respectively . proof of theorem [ thm : main ] is given in [ s : invariance ] . according to a result of laudenbach blank @xcite , if the nonsingular closed 1-forms @xmath442 and @xmath443 are cohomologous , then they are isotopic . by integrating the 1-parameter family of closed 1-forms , one sees that @xmath444 and @xmath445 are isotopic to each other . one can remove the dependence of @xmath1 on spin structures as follows . recall that for a compact 3-manifold @xmath10 , the number of all spin structures on @xmath10 is @xmath446 , which is finite ( ( * ? ? ? * lemma ) ) . one may consider the sum @xmath447,[f])=\frac{1}{|h_1(m;\z_2)|}\sum_{\mathfrak{s}\in\spin(m)}\wz_n(m,\mathfrak{s},[\kappa],[f])\ ] ] over all spin structures on @xmath10 . let @xmath448 and @xmath449 be as in [ ss : def_z ] . let @xmath52 be a labeled trivalent graph of degree @xmath23 . let @xmath450 . suppose that no @xmath287-intersections for @xmath367 is on @xmath89 . we define @xmath451 , @xmath452 , as the set of piecewise smooth maps @xmath453 such that 1 . the restriction of @xmath454 to the @xmath125-th edge is an al - path of @xmath367 , 2 . the algebraic intersection of the restriction of @xmath454 to the @xmath125-th edge @xmath455 with @xmath89 is @xmath456 . we call such maps _ al - graphs _ for @xmath457 of type @xmath458 . we define a topology on @xmath451 as the transversal intersection of smooth submanifolds of @xmath223 , as in [ ss : mlinearform ] . here the condition ( 2 ) implies that @xmath459 represents the cocycle @xmath460 , @xmath461 . let @xmath462 be the union of the preimages of all the vertical segments . let @xmath463 be the restriction of @xmath454 to the closure of @xmath464 in @xmath52 . then @xmath465 consists of finitely many `` horizontal '' components each of which consists only of horizontal segments in al - paths . we say that a component of @xmath465 is of _ generic type _ if it is a unitrivalent graph whose univalent vertices are mapped by @xmath454 to critical loci of index 1 . [ lem : transversality ] for generic choices of @xmath466 , the moduli space @xmath467 is a compact oriented 0-dimensional manifold for all @xmath458 and for all labeled oriented 3-valent graph @xmath52 of degree @xmath23 simultaneously . by a dimensional reason , we may assume that a horizontal component in @xmath465 does not have a vertex that is on a critical locus of index 0 or 2 , for a generic choice of @xmath468 . namely , every horizontal components in @xmath465 are of generic type . the transversality can be proved by an argument similar to @xcite or ( * ? ? ? * proposition 2.4 ) except that the descending ( resp . ascending ) manifolds of critical points of index 2 ( resp . index 1 ) is replaced with the descending ( resp . ascending ) manifold loci of critical loci of index 1 . for the compactness , we use the fact that the subset of @xmath469 of paths of bounded lengths consists of finitely many compact strata ( ( * ? ? ? * lemma 4.1 ) ) . namely , the lift @xmath470 $ ] , @xmath471=\pi^{-1}(i)$ ] , of @xmath89 in @xmath472 decompose @xmath472 into cobordisms : @xmath473 $ ] where @xmath474=\bar{\kappa}^{-1}[i , i+1]$ ] . by the definition of the moduli space of al - paths given in @xcite , the set of al - paths from a point of @xmath475 $ ] to @xmath476 $ ] forms a finite union of finite coverings over compact subsets of @xmath475\times m[k - n]$ ] , where the number of sheets in the covering is the number of al - paths from an output point in @xmath471 $ ] to an input point in @xmath477 $ ] . hence the moduli space of al - graphs of given type @xmath458 is compact . for a fixed @xmath23 and for a generic choice of @xmath478 , there are finitely many possiblities for horizontal components that may be a horizontal part of an al - graph of degree @xmath23 . thus the transversality for all the horizontal components implies that @xmath467 is a compact oriented 0-dimensional manifold for all @xmath458 and for all @xmath52 of degree @xmath23 . we may identify a point of @xmath467 with an oriented 0-manifold in @xmath205 . hence the moduli space @xmath467 can be counted with signs . the sum of the signs agrees with the sum of coefficients of the terms of @xmath479 in the power series expansion of @xmath480 . we denote the sum of signs by @xmath481 . lemma [ lem : transversality ] implies that for generic choices of @xmath482 , an al - graph @xmath483 consists of finitely many horizontal components of generic type and some al - paths connecting the univalent vertices of the horizontal components . we say that such an al - graph is of _ generic type_. the following proposition follows from theorem [ thm : propagator ] and from definition of @xmath359 . [ prop : f ] let @xmath449 be as in lemma [ lem : transversality ] . for a labeled trivalent graph @xmath52 , let @xmath484 be the generating function @xmath485 where @xmath486 is the count of al - graphs of type @xmath458 of generic type . then there exist laurent polynomials @xmath487 , @xmath363 , @xmath488 , such that @xmath489 considering this as an element of @xmath490 , we have @xmath491 the trace @xmath492 can not be directly applied to formal power series . for example , consider the formal power series @xmath493 when @xmath494 in ( [ eq : theta ] ) in [ ss : lescop - inv ] below , applying @xmath495 to each monomial @xmath496 corresponds to taking the equivalence class in @xmath497/(t_1t_2t_3 - 1)$ ] by the holonomy relation . however , the infinite sum of the equivalence classes is not well - defined since @xmath498 if @xmath499 . here , we assume that @xmath11 . let @xmath500 be a trivialization of @xmath174 and let @xmath501 be a section induced by @xmath502 that sends @xmath10 to @xmath503 for a fixed @xmath504 . we denote by @xmath505 the image of @xmath506 . suppose that @xmath507 agrees with the unit tangent vectors of @xmath508 . moreover , by choosing @xmath94 suitably , we may assume that @xmath12 agrees with a critical locus of index 0 . let @xmath509\to \bconf_2(m)$ ] be the map defined by @xmath510 and let @xmath511\to \bconf_{k_2}(m)$ ] be the lift of @xmath512 such that @xmath513 is taken to @xmath334 $ ] . originally , an equivariant propagator is defined in @xcite as a 4-dimensional @xmath78-chain @xmath514 in @xmath14 such that the chain level identity @xmath515 holds , where @xmath516 is a parallel of @xmath12 and @xmath517 @xmath273 is the alexander polynomial of @xmath10 normalized so that @xmath518 and @xmath519 , and such that @xmath520 lescop considered in @xcite such a @xmath514 to define an equivariant invariant of @xmath10 . since our equivariant propagator @xmath17 does not satisfy the identity ( [ eq : dq_tau ] ) ( compare to theorem [ thm : propagator ] ) , we extend it by adding a bordism in @xmath231 in order to use results of @xcite . for a closed al - path @xmath521 , we shall take a 2-dimensional bordism @xmath522 such that @xmath523 for an integer @xmath325 , as follows . let @xmath524 for a generic value @xmath97 and let @xmath525 be all the intersection points of @xmath89 with critical loci of an oriented fiberwise morse function @xmath94 . suppose that @xmath516 intersects @xmath89 transversally at @xmath526 with the intersection number @xmath527 . for each @xmath528 , choose a path @xmath529 on @xmath89 from @xmath526 to @xmath530 . let @xmath531 be the set of al - paths from @xmath532 to @xmath530 which do not intersect @xmath89 except the endpoints . then for @xmath533 , the 1-cycle @xmath534 is bordant to @xmath516 since @xmath22 . take a 2-dimensional bordism @xmath535 such that @xmath536 and such that @xmath537 , and let @xmath538 a closed al - path @xmath521 is cut by @xmath89 into segments : @xmath539 , @xmath540 . then the 2-dimensional bordism @xmath541 satisfies @xmath542 . [ lem : st(v ) ] let @xmath543 be the 2-dimensional @xmath78-chain @xmath544 where @xmath545 is the matrix @xmath546 and @xmath547 is the @xmath548-th entry of @xmath549 . then we have @xmath550 where the sum is over equivalence classes of all closed al - paths for @xmath18 , and @xmath551 is the lefschetz zeta function of the monodromy of the fibration @xmath112 . by an argument similar to ( * lemma 4.2 ) , we have @xmath552 indeed , by decomposing @xmath553 as ( [ eq : vsum ] ) , the right hand side can be rewritten as a linear combination of the chains @xmath554 where @xmath555 . the coefficient of @xmath554 in the right hand side is a power series in @xmath8 , which is the generating function of the numbers of irreducible closed al - paths that starts from @xmath556 , namely , base pointed closed al - paths . if @xmath533 , the coefficient of @xmath554 is @xmath557 by definition of @xmath558 , we have @xmath559 where the last equality follows from ( * ? ? * proposition 4.9 ) . the following lemma follows from ( * ? ? ? * proposition 2.12 , 4.5 ) . [ lem : u ] there exists a 4-dimensional chain @xmath560 in @xmath231 such that @xmath561 where @xmath562 is the genus of @xmath89 . [ rem : u ] the 4-chain @xmath560 can be taken more explicitly as follows . let @xmath563 , where @xmath564 is a nonnegative smooth function supported on a small tubular neighborhood of the union of all critical loci . then @xmath565 is a nonsingular vector field on @xmath10 and the image of the section @xmath566 can be arbitrarily close to @xmath248 with respect to the hausdorff distance . now one can take @xmath502 so that the restriction of @xmath567 to the complement of a small tubular neighborhood @xmath568 of @xmath569 is parallel to @xmath253 at every point @xmath570 . then @xmath560 can be taken as follows . 1 . the restriction of @xmath560 to each fiber @xmath571 of @xmath572 is the minimal geodesic between @xmath573 and @xmath574 , which is a constant path for @xmath570 . @xmath502 may be perturbed so that @xmath505 is arbitrarily close to @xmath248 in @xmath575 although they are singularly different on a small neighborhood of @xmath516 . the limit of the perturbation of @xmath505 restricts on @xmath576 to a chain that is homologous to @xmath577 in @xmath576 . namely , the limit of @xmath505 is homologous to @xmath578 and the difference of the restrictions on @xmath576 bound a 4-chain in @xmath576 . we may assume that in a small neighborhood of @xmath576 in @xmath231 the restriction of @xmath560 is given by this 4-chain . we put @xmath579 let @xmath16 denote the following labeled oriented graph : @xmath580 let @xmath581 be the minimal polynomial of @xmath582 normalized so that @xmath583 and @xmath584 . let @xmath585 be the subspace of @xmath586 spanned by the elements @xmath587 for all positive integers @xmath258 . let @xmath362 , @xmath588 , be oriented fiberwise morse functions for the fibration @xmath88 and let @xmath367 be the fiberwise gradient of @xmath368 . let @xmath589 be a small closed tubular neighborhood of @xmath12 in @xmath10 and let @xmath590 be three parallels of @xmath12 taken with respect to the trivialization @xmath591 . we may assume without loss of generality that @xmath12 is disjoint from all the closed al - paths of @xmath367 . let @xmath592 be three points that are close to the fixed point @xmath504 . then by replacing @xmath593 with @xmath594 in the definition of @xmath595 above , we obtain three 4-dimensional chains @xmath596 , @xmath588 . [ thm : q ] we have @xmath597 @xmath598 after perturbations of @xmath599 and @xmath600 along small inward normal vector fields on @xmath234 fixing on @xmath601 and @xmath602 for each @xmath125 , we may arrange that the expression @xmath603 is well - defined and the class of @xmath604 in @xmath605 is an invariant of @xmath10 and the homotopy class of @xmath502 . ( up to normalization , @xmath3 agrees with lescop s invariant in @xcite without framing correction term . ) for generic choices of @xmath606 and for @xmath600 chosen as in remark [ rem : u ] , we have @xmath607 . \end{split}\ ] ] the alexander polynomial of a fibered 3-manifold @xmath10 over @xmath2 is of the form @xmath608 , where @xmath609 and @xmath610 is the monodromy action of the standard generator of @xmath611 . this together with the formula @xmath612 gives the following identity of the logarithmic derivatives @xmath613 hence by lemma [ lem : u ] we have @xmath614 the vanishing of @xmath615 is immediate from remark [ rem : u ] , namely , since the restriction of @xmath367 on @xmath12 is horizontal everywhere , @xmath600 does not meet @xmath616 . since we assumed that @xmath12 is a critical locus of @xmath94 of index 0 , @xmath12 is induded in an ascending manifold locus of @xmath367 of index 0 , @xmath12 is transversal to fibers of @xmath364 and @xmath617 . moreover , @xmath618 implies that @xmath619 . this shows that @xmath3 agrees with lescop s invariant of 3-manifolds in @xcite without framing correction term . let @xmath620 be the union of the images of all the closed al - paths of @xmath367 and @xmath621 . by a dimensional reason , we may assume that @xmath622 if @xmath623 . this proves the vanishing of the terms like @xmath624 moreover , since @xmath625 is a 1-chain , we may also assume that @xmath626 is nullhomologous if @xmath627 are distinct . this implies the vanishing of the terms like @xmath628 before the perturbations of @xmath599 , the triple intersection @xmath629 is @xmath630 of @xmath631 as a set . but the intersection is not transversal and perturbations of @xmath599 are necessary . at each intersection point @xmath43 in the triple intersection @xmath631 , let @xmath632 be a small neighborhood of @xmath43 in @xmath10 . there is a local coordinate @xmath633 , @xmath634 small , on a small neighborhood of @xmath635 in @xmath15 , where @xmath635 corresponds to @xmath636 . then perturb @xmath599 to the level @xmath637 for a small positive number @xmath638 by using a cloche function supported on @xmath632 . if @xmath639 are mutually distinct , the result of the perturbation has empty intersection in @xmath640 . this shows the vanishing of the term ( [ eq : st3 ] ) . hence we obtain the formula of the statement . by ( an analogue of ) proposition [ prop : f ] the first term of the sum formula in theorem [ thm : q ] counts al - graphs of generic type . the rest in the formula of theorem [ thm : q ] have geometric descriptions as follows . [ prop : formula - linking ] for generic choices of @xmath606 and for @xmath600 chosen as in remark [ rem : u ] and theorem [ thm : q ] , we have the following . 1 . @xmath641 , where the sum is taken over closed al - paths for @xmath642 and @xmath643 is the linking number of two parallels of @xmath256 given by @xmath644 and @xmath645 defined with respect to @xmath502 . @xmath646 , where @xmath647 is the piecewise smooth link in @xmath10 defined as the projection of @xmath648 , with the orientation determined by the intersection in @xmath231 . @xmath649 , where @xmath650 is the linking number of @xmath647 and its parallel given by @xmath642 . since proposition [ prop : formula - linking ] is not necessary in the rest of this paper , we omit the proof of proposition [ prop : formula - linking ] . [ lem : indep_sigma ] the term @xmath651 of @xmath359 does not depend on the choice of @xmath89 within its oriented bordism class . there are finitely many possibilities for horizontal components in al - graphs for @xmath478 . a bordism of @xmath89 is decomposed into a sequence of homotopies and attachings of 1-handles that are disjoint from all the horizontal components and all the critical loci of @xmath367 s . since the 1-handle attach does not change the value of @xmath359 , it suffices to check the lemma for a homotopy of @xmath89 . suppose that a horizontal component @xmath652 is a part of an al - graph of type @xmath458 and has at least one trivalent vertex . if @xmath653 is a set of horizontal components for @xmath478 having trivalent vertices such that the total number of trivalent vertices in the graphs of @xmath279 is exactly @xmath216 and if @xmath52 is a labeled trivalent graph , then the generating function @xmath654 of counts of al - graphs from @xmath52 whose set of horizontal components agrees with @xmath279 is a rational function in @xmath655 . the reason for this is that there are finitely many ways of joining legs of @xmath656 s to form the uncolored 3-valent graph @xmath52 . in each joining of a pair of legs , the weighted number of ways of joining the pair by al - paths is a rational function , as shown in ( * ? ? ? * lemma 4.2 ) or in lemma [ lem : st(v ) ] above . hence @xmath654 is a finite sum of rational functions . moreover , there are finitely many possibilities for the set @xmath279 that contain @xmath652 . therefore , @xmath657 is a rational function in @xmath655 . now suppose that a homotopy of @xmath89 crosses a trivalent vertex @xmath658 . then @xmath659 may change under the homotopy and the terms that may change at the crossing are the terms in @xmath660 . more precisely , if the three edges incident to @xmath68 are labeled @xmath627 , then all the terms in @xmath660 get multiplied by @xmath661 , @xmath662 ( depending on the edge orientations ) . see figure [ fig : holonomy ] . namely , the change of @xmath659 under the homotopy at the crossing is @xmath663 , whose trace vanishes by the holonomy relation for @xmath78-colored graph ( figure [ fig : relations ] ) . this completes the proof . let @xmath88 be a fiber bundle with fiber diffeomorphic to a connected oriented closed surface . a concordance of an oriented fiberwise morse function gives a 2-parameter family @xmath664\to \r$ ] of oriented gmf s on a surface parametrized over @xmath107 $ ] . [ lem : arrange - bd ] after a perturbation of the concordance @xmath105 fixing the endpoints , we may assume that there is a sequence @xmath665 such that 1 . @xmath666 is an oriented fiberwise morse function if @xmath667 , and 2 . at @xmath668 , there is exactly one locus of @xmath669-singularities ( birth - death locus ) for @xmath670 which forms a finite covering over @xmath671 by @xmath672 . the proof is an analogue of the beak lemma in ( * ? ? ? 3 ) . since @xmath106 $ ] is 4-dimensional , a birth - death locus is 1-dimensional , and a critical locus of @xmath105 is 2-dimensional and unknotted , one may take a homotopy of a birth - death locus into a slice @xmath673 that is disjoint from critical loci of @xmath105 and from other birth - death loci . for a birth locus , we may take a homotopy which pulls back the locus to a minimal parameter of @xmath674 on the locus . for a death locus , we may take a homotopy which pushes forward the locus to a maximal parameter of @xmath674 on the locus . then the result is as desired . to prove the invariance of @xmath1 , we take a concordance @xmath105 as in lemma [ lem : arrange - bd ] and decompose @xmath675 $ ] as @xmath676\cup[s_1-\ve , s_1+\ve]\cup [ s_1+\ve , s_2-\ve]\cup \cdots\cup [ s_r-\ve , s_r+\ve]\cup [ s_r+\ve,1],\ ] ] for a small @xmath677 , and show that @xmath1 is invariant at each piece . the restriction of @xmath105 on @xmath678 $ ] gives a concordance through oriented fiberwise morse functions . let @xmath679}$ ] be the fiberwise gradient of @xmath105 . we say that a parameter @xmath680 $ ] is a _ bifurcation _ for @xmath681}$ ] if @xmath682 does _ not _ satisfy one of the following conditions , which were assumed in the definition of the moduli space @xmath264 . 1 . level - exchange points and @xmath287-intersections occur at different levels for @xmath112 . 2 . transversality of curves in the graphic . @xmath682 satisfies parametrized morse smale condition . in @xcite , generic 2-parameter families of smooth functions are studied . according to @xcite , a generic homotopy between generic loops in the space of morse functions is as follows . [ lem : generic - ho ] let @xmath683 be as above . for a generic choice of a fiberwise gradient @xmath227 of @xmath104 on @xmath684 $ ] , we may arrange that the possible bifurcations in the family @xmath685 on @xmath686 $ ] are of the following forms : 1 . a nondegenerate critical point on a level exchange curve . 2 . a point where three critical loci have the same value . a crossing between two level exchange curves . a nondegenerate critical point on a @xmath287-intersection curve . a crossing between two @xmath287-intersection curves ( not successive ) a crossing between two @xmath287-intersection curves ( successive ) a crossing between a level exchange curve and a @xmath287-intersection curve . ( see figure [ fig : bifurcations ] and [ fig : bifurcations2 ] ) . moreover , we may assume that no two bifurcations occur simultaneously in each time in @xmath686 $ ] . in the rest of this section , let @xmath687}$ ] be a generic concordance arranged as in lemma [ lem : arrange - bd ] , [ lem : generic - ho ] . the moduli space @xmath264 is defined by taking an auxiliary decomposition of @xmath10 into small pieces @xmath688 , which are called _ cells _ ( @xcite ) . the decomposition arose from the graphic . see figure [ fig : cellular - diag ] for an example . the decomposition is done roughly as follows . first , partition @xmath10 by ( horizontal ) level surfaces @xmath240 for several values of @xmath689 so that each piece @xmath690 satisfies either of the following . 1 . @xmath690 contains exactly one level exchange pair of critical loci and no @xmath287-intersections . @xmath690 contains no level exchange pair of critical loci but may contain several @xmath287-intersections . then partition each piece @xmath690 further by ( vertical ) level surfaces of @xmath94 so that each piece contains one or two components of critical loci and that it contains two if and only if the two loci are the pair forming a crossing in the graphic . by using the decomposition , we may identify each al - path @xmath224 in @xmath264 with a point of the direct product of vertical level surfaces on which @xmath224 intersects . thus one sees that @xmath264 is locally modeled on the direct product of the vertical level surfaces . namely , embed a small neighborhood of @xmath691 into the product of the vertical level surfaces and take the closure in the product . then glue together such local models for the closure suitably to obtain @xmath264 . we orient @xmath692 by @xmath693 . [ lem : nobifurcations ] let @xmath694\subset [ 0,1]$ ] be an interval such that the restriction of @xmath685 on @xmath692 does not have bifurcations . then @xmath695 since there are no bifurcations in the restriction @xmath696 of @xmath685 on @xmath692 , the graphics of @xmath697 and @xmath698 with the arrows of 1/1-intersections are isotopic through those in the 1-parameter family . hence , the auxiliary decompositions of @xmath10 for @xmath699 and @xmath700 are extended to a decomposition of @xmath692 , which is isomorphic to the product decomposition with @xmath701 . by using the product structure , we may define a moduli space @xmath702 , which gives a cobordism between @xmath703 and @xmath704 diffeomorphic to @xmath705 . by replacing @xmath703 with @xmath702 in the definition of @xmath359 , we may define the 1-chain @xmath706 in @xmath707 , where @xmath708 is the corresponding bordism between @xmath709 and @xmath710 . the coorientation of @xmath711 ( resp . @xmath712 ) is given so that its restrictions on @xmath713 is equivalent to that of @xmath709 ( resp . @xmath714 ) . for @xmath715 , let @xmath716 be the vector fields defined by @xmath717 . putting @xmath718 , we define @xmath719,\ ] ] where @xmath720 is the moduli space of linear graphs in the rank 3 vector bundle @xmath721 over @xmath692 . we shall check that the homology class of @xmath722 is given by @xmath723 this is an analogue of an identity in the proof of ( * ? ? ? * lemma 10.1 ) . by assuming this , the proof will complete as follows . @xmath724=0 , \end{split}\ ] ] where @xmath725 and @xmath726 are the tuples of sections of @xmath424 that extend @xmath727 and @xmath728 respectively . since @xmath729 is the intersection form among relative cycles in @xmath730 , the boundary of @xmath729 that are not on @xmath731 consists of configurations for al - graphs that are in the preimage of @xmath732 . according to the description of the strata of @xmath149 in [ ss : fm ] , each of such al - graphs @xmath453 is one of the following forms . 1 . ( principal face ) for an edge @xmath733 , the image @xmath734 of @xmath81 collapses into a point . ( hidden face ) a subgraph @xmath735 in a horizontal component @xmath736 of @xmath737 collapses into a point . note that it is not necessary to consider an al - graph with a non self - loop edge forming a closed al - path in @xmath10 since such an al - graph and an al - graph with one 4-valent vertex do not occur simultaneously in a generic homotopy . if @xmath52 is the dumbbell graph @xmath738-.3mm-.3mm@xmath738 , there may be an al - graph formed by exactly two closed al - paths sharing a point , which is treated in case ( 1 ) . in a generic homotopy , there does not exist an al - graph formed by three closed al - paths sharing one point . the contributions of the case ( 1 ) are canceled each other out in @xmath739 by the ihx relation , for a similar reason as in chern simons perturbation theory for homology 3-spheres ( e.g. , @xcite ) . more precisely , for each labeled trivalent graph @xmath52 , the contributions in @xmath740 having such collapsed edge , say , the one labeled @xmath258 , is a rational function with no terms of nonzero exponents of @xmath741 , as in the proof of lemma [ lem : indep_sigma ] . thus the ihx relation for @xmath78-colored graphs can be applied to prove that the changes are canceled each other out without any difficulty . see ( * ? ? ? * lemma 10.1 ) for detail . the contributions of the case ( 2 ) for @xmath742 vanish by an analogue of kontsevich s lemma ( * ? ? ? * lemma 2.1 ) ( see also ( * ? ? ? * lemma 6.3 , 6.4 , 6.5 ) ) . roughly , the contribution of the case ( 2 ) is the product of the count of al - graphs from @xmath743 in @xmath744 and the count of linear graphs from @xmath735 in @xmath745 . @xmath735 has a univalent vertex or a bivalent vertex . if @xmath735 has a univalent vertex , then the moduli space of linear graphs in @xmath746 is generically empty by a dimensional reason . if @xmath735 has a bivalent vertex , then there is an orientation reversing involution on @xmath745 , which exchanges the moduli space of @xmath735 and that of another graph @xmath747 , which is @xmath735 with different labels and different edge - orientations . the cancellation between the two graphs @xmath735 and @xmath747 is examined in ( * ? ? ? * lemma 6.3 ) . the contributions of the case ( 2 ) for @xmath748 , which are not covered by the previous paragraph , are those correspond to the collapse of whole graph . they contribute as @xmath749 . [ lem : case(a ) ] let @xmath750 $ ] be a parameter at which the bifurcation of type ( a ) in figure [ fig : bifurcations ] occurs in @xmath683 . for a small number @xmath677 , @xmath751 let @xmath752 $ ] be a point such that the projections of two critical loci of @xmath753 in the graphic are tangent to each other ( see figure [ fig : bifurcations2 ] ( a ) ) . after a small perturbation of @xmath227 which does not change the critical loci , we may assume that the moduli space @xmath754 is empty on @xmath755 $ ] for all @xmath52 . hence we may assume that the anomaly correction term @xmath756 does not change through @xmath755 $ ] . take a neighborhood @xmath757 of @xmath758 in @xmath109 $ ] so small that there are no @xmath287-intersections in @xmath759 $ ] . let @xmath760 $ ] and suppose that @xmath761 is sufficiently small so that @xmath762\times j\subset u_{(u_0,s_0)}$ ] . moreover , we assume without loss of generality that the curve of the level exchange parameters in @xmath757 intersects the segment @xmath763 $ ] in two points , as in the left hand side of figure [ fig : bifurcations ] ( a ) . we may decompose @xmath764\times j)$ ] into small pieces as in the proof of lemma [ lem : nobifurcations ] since @xmath227 has no bifurcation there . the rest @xmath765\times j)$ ] can be decomposed as follows . for the critical loci of @xmath766}$ ] that are not involved in the bifurcation , we may take product cells as in the proof of lemma [ lem : nobifurcations ] each of which includes exactly one critical locus . we consider the closure @xmath767 of the complement in @xmath692 of the union of all the cells taken above as one big cell , which includes exactly two critical loci . see figure [ fig : partition ] . now we have a decomposition of @xmath692 into cells . applying a similar method as in the definition of @xmath768 with the decomposition of @xmath692 given above , we obtain a moduli space @xmath702 of al - paths in @xmath692 , which restricts to @xmath768 and @xmath769 at @xmath770 , without any difficulty . by replacing @xmath703 with @xmath702 in the definition of @xmath359 , we may define the 1-chain @xmath706 in @xmath707 . proof of the identity @xmath771=&z_n(\xi_{s_0-\ve},\xi_2,\ldots,\xi_{3n})-z_n(\xi_{s_0+\ve},\xi_2,\ldots,\xi_{3n } ) \end{split}\ ] ] is the same as lemma [ lem : nobifurcations ] . [ lem : case(b ) ] let @xmath750 $ ] be a parameter at which the bifurcation of type ( b ) or ( c ) in figure [ fig : bifurcations ] occurs in @xmath683 . for a small number @xmath677 , @xmath751 the proof is the same as lemma [ lem : case(a ) ] . note that for ( c ) , there are two big cells at @xmath772 in the decomposition of @xmath10 into cells . [ lem : case(d ) ] let @xmath750 $ ] be a parameter at which the bifurcation of type ( d ) in figure [ fig : bifurcations ] occurs in @xmath683 . for a small number @xmath677 , @xmath751 let @xmath752 $ ] be a point such that the bifurcation of type ( d ) occurs . after a small perturbation of @xmath227 which does not change the bifurcation parameters , we may assume that the moduli space @xmath754 is empty on @xmath773 $ ] for all @xmath52 . hence we may assume that the anomaly correction term @xmath756 does not change through @xmath755 $ ] . take a neighborhood @xmath757 of @xmath758 in @xmath109 $ ] so small that there are no other @xmath287-intersections nor level exchange bifurcations in @xmath759 $ ] . let @xmath760 $ ] and suppose that @xmath761 is sufficiently small so that @xmath762\times j\subset u_{(u_0,s_0)}$ ] . we assume without loss of generality that the curve of the @xmath287-intersection parameters in @xmath757 intersects the segment @xmath763 $ ] in two points , as in the left hand side of figure [ fig : bifurcations ] ( d ) . then we may take a decomposition of @xmath692 into pieces as in the proof of lemma [ lem : nobifurcations ] with no trouble . by the parametrized morse smale condition for @xmath227 , we may apply a similar method as in the definition of @xmath768 with the decomposition of @xmath692 given above . then we obtain a moduli space @xmath702 of al - paths in @xmath692 , which restricts to @xmath768 and @xmath769 at @xmath770 . by replacing @xmath703 with @xmath702 in the definition of @xmath359 , we may define the 1-chain @xmath706 in @xmath707 . proof of the identity @xmath771=&z_n(\xi_{s_0-\ve},\xi_2,\ldots,\xi_{3n})-z_n(\xi_{s_0+\ve},\xi_2,\ldots,\xi_{3n } ) \end{split}\ ] ] is the same as lemma [ lem : nobifurcations ] . [ lem : case(e ) ] let @xmath750 $ ] be a parameter at which the bifurcation of type ( e ) or ( g ) in figure [ fig : bifurcations ] occurs in @xmath683 . for a small number @xmath677 , @xmath751 the proof is the same as lemma [ lem : case(d ) ] . [ lem : case(f ) ] let @xmath750 $ ] be a parameter at which the bifurcation of type ( f ) in figure [ fig : bifurcations ] occurs in @xmath683 . for a small number @xmath677 , @xmath751 after a small perturbation of @xmath227 which does not change the bifurcation parameters , we may assume that the moduli space @xmath754 is empty on @xmath755 $ ] for all @xmath52 . hence we may assume that the anomaly correction term @xmath756 does not change through @xmath755 $ ] . let @xmath760 $ ] . as in the proof of lemma [ lem : case(d ) ] , we may define a moduli space @xmath702 of al - paths in @xmath774 , @xmath775 , which restricts to @xmath768 and @xmath769 at @xmath770 . this gives rise to the 1-chain @xmath706 in @xmath707 . note that the the boundary of the chain + @xmath729 consist only of the forms given in the proof of lemma [ lem : nobifurcations ] although the trajectory spaces for the @xmath287-intersections involved in the bifurcation may have boundaries at @xmath772 . the al - graphs in @xmath776 that may arrive at the boundary at @xmath772 are those with an edge @xmath81 including both the horizontal segments labeled @xmath377 and @xmath325 in the left side of figure [ fig : bifurcations2 ] ( f ) . they can be paired with the al - graphs in @xmath777 with an edge @xmath778 including the horizontal segment labeled @xmath779 in the right side of figure [ fig : bifurcations2 ] ( f ) . @xmath780 since these edges have the same sign , the boundaries at @xmath772 are cancelled with each other . hence we have @xmath771=&z_n(\xi_{s_0-\ve},\xi_2,\ldots,\xi_{3n})-z_n(\xi_{s_0+\ve},\xi_2,\ldots,\xi_{3n } ) . \end{split}\ ] ] [ lem : bd ] let @xmath750 $ ] be a parameter at which a birth - death bifurcation occurs in @xmath683 . for a small number @xmath677 , @xmath751 by the symmetry between a birth and a death bifurcation with respect to the parameter @xmath674 , it is enough to check the lemma only for a birth bifurcation . let @xmath781 $ ] be the birth locus that occurs at @xmath782 and suppose that a pair @xmath783 of critical loci of @xmath105 of index 1 and 2 appear after @xmath782 . see figure [ fig : bd - cancel ] ( a ) . the case where @xmath784 and @xmath785 is symmetric to this case . moreover , we assume for simplicity that the image of @xmath224 in @xmath107 $ ] spans @xmath786;\z)=\z$ ] since the proof of other cases are the same . take a small neighborhood @xmath787 of @xmath788 in @xmath107 $ ] , a section @xmath789 that is a smooth extension of @xmath224 , and a fiberwise small tubular neighborhood @xmath632 of @xmath790 in @xmath106 $ ] . choose a level surface locus @xmath736 of @xmath105 in @xmath632 that is disjoint from @xmath791 in @xmath632 and that lies just below @xmath792 . by the normal form lemma for @xmath669-singularities in @xcite , we may assume that for each point @xmath793 on @xmath224 , there is a local coordinate @xmath794 around @xmath795 such that @xmath105 agrees with @xmath796 for a smooth function @xmath797 , where it makes sense . moreover , we may assume for a fixed pair @xmath795 that the euclidean local coordinate is isometric near @xmath795 . the intersection of the descending manifold locus @xmath798 of @xmath224 with @xmath736 forms a bundle over @xmath799 with fiber a small closed interval . by perturbing @xmath227 in a small neighborhood of @xmath736 , the closed interval can be made arbitrarily small in each fiber . then by smoothness of @xmath227 , the intersection of @xmath800 in each fiber of @xmath632 with @xmath736 can be made arbitrarily small too . then for a small number @xmath677 , @xmath801 forms a 1-parameter family of thin half - disks which flow down to a critical locus of index 0 for a generic parameter , or which may go through another critical locus of index 1 as @xmath8 increases . see figure [ fig : bd - cancel ] ( b ) . first , we consider the limit of al - graphs as @xmath802 from below . if @xmath227 is generic , the moduli spaces @xmath803 for @xmath804 forms a finite covering over @xmath805 . as seen from the result in @xcite , the limit consists of al - graphs that do not intersect @xmath224 and those with a broken edge that intersect @xmath224 . the al - graphs at @xmath772 that do not intersect @xmath224 extends smoothly over @xmath806 $ ] as a covering over @xmath755 $ ] . the al - graphs that intersect @xmath224 stop at @xmath772 as boundaries of a 1-dimensional moduli space and do not extend over @xmath806 $ ] . we shall check that @xmath227 can be perturbed within the space of 1-parameter families of gradient - like vector fields for the given a 2-parameter family @xmath105 of gmf s so that all the boundaries of the moduli spaces at @xmath772 disappear . long broken edge : suppose that the limit of a 1-parameter family @xmath807 of al - graphs at @xmath782 has an edge @xmath81 with positive length that is broken at the birth locus @xmath224 . let @xmath778 be the horizontal segment in @xmath81 that is broken at @xmath224 . in this case , by perturbing @xmath227 in small neighborhoods of two points on @xmath778 that are not close to @xmath224 , we may assume that @xmath81 is disjoint from @xmath224 . this is possible since we are assuming that the descending manifold locus @xmath798 forms a bundle of arbitrarily thin half - disks . @xmath808 collapsed broken edge : suppose that the limit of a 1-parameter family @xmath807 of al - graphs at @xmath782 has an edge that collapses to a point on the birth locus @xmath224 . this is the contribution of @xmath234 . the limit can be described as follows . let @xmath760 $ ] . we may define the moduli space @xmath702 of al - paths by gluing the closures in the auxiliary spaces ( @xmath809 ) as in the definition of @xmath264 in @xcite . the evaluation map @xmath810 represents a 5-dimensional @xmath78-chain @xmath811 . the boundary of @xmath811 on @xmath812 is described as follows . let @xmath813 and let @xmath814 be the normalization @xmath815 of the section @xmath816 . the closure @xmath817 in @xmath818 is a smooth manifold with boundary whose boundary at @xmath234 is the disjoint union of circle bundles over the critical loci @xmath120 of @xmath819 ( including the birth - death locus @xmath224 ) , for a similar reason as ( * ? ? ? * lemma 4.3 ) . the fibers of the circle bundles are equators of the fibers of @xmath820 . let @xmath821 be the total space of the 2-disk bundle over @xmath120 whose fibers are the lower hemispheres of the fibers of @xmath820 which lie below the tangent spaces of the level surfaces of @xmath112 . then @xmath822 as sets . let @xmath823 this is a 4-dimensional piecewise smooth manifold . we orient @xmath824 by extending the natural orientation @xmath825 on @xmath826 . the contribution of @xmath234 in the boundary of @xmath811 is of the following form : @xmath827 see @xcite for the reason of the signs in this formula . we could describe the sign @xmath828 and the orientations of @xmath829 precisely , but it is not necessary here . the point is that one can define @xmath830 using the restriction of @xmath811 at @xmath772 and that there is no bifurcation for the anomaly correction term @xmath756 around @xmath772 if @xmath819 is generic , by a dimensional reason . this completes the proof of vanishing of the contributions of the al - graphs with an edge degenerate at the birth locus . next , we consider the limit of al - graphs as @xmath802 from above . in this side , there may be other kind of al - graphs with an edge broken at the birth point , namely , those with an edge that visits the critical locus @xmath57 ( of index 1 ) . let @xmath831 be a generic al - graph whose first edge @xmath832 has a vertical segment @xmath833 included in the critical locus @xmath834 . there must be a horizontal segment @xmath289 in @xmath832 next to @xmath833 . let @xmath652 be a horizontal component in @xmath454 that contains @xmath289 . when @xmath652 has a trivalent vertex , let @xmath835 be the graph obtained from @xmath652 by removing the edge @xmath289 . then the moduli space of al - graphs @xmath836 near the inclusion @xmath837 is 1-dimensional and the locus of the bivalent vertex of @xmath835 forms a local section @xmath838 of @xmath839 , where @xmath632 is a small open interval in @xmath2 , @xmath444 is the first of @xmath840 ( [ ss : def_z ] ) and @xmath841 . we define the descending manifold @xmath842 and the ascending manifold @xmath843 of @xmath844 as follows . @xmath845 then the segment @xmath289 corresponds to a transversal intersection of @xmath846 with @xmath847 . the observation in the previous paragraph implies that @xmath289 goes along one side of a thin half - disk in @xmath801 . by transversality of @xmath846 and @xmath847 , the other side of the half - disk intersects @xmath847 with the opposite orientation in a fiber close to @xmath289 this implies that there is a unique al - graph @xmath848 that is close to @xmath454 and whose first edge is nearly parallel to that of @xmath454 . moreover , the signs of @xmath454 and @xmath849 are opposite . hence , no terms of the al - graphs whose first edge visits @xmath57 survive in the sum @xmath850 . the equality for other graphs can be proved similarly as lemma [ lem : case(a ) ] . when @xmath652 does not have a trivalent vertex , then @xmath652 is a single flow - line between a pair of critical loci , say , from @xmath57 to @xmath851 . in this case , replace @xmath847 with @xmath852 in the previous paragraph so that similar cancellation occurs . the theorem follows as a corollary of lemmas [ lem : generic - ho ] , [ lem : nobifurcations ] , [ lem : case(a ) ] , [ lem : case(b ) ] , [ lem : case(d ) ] , [ lem : case(e ) ] , [ lem : case(f ) ] and [ lem : bd ] . let @xmath853\subset s^1 $ ] be a small interval which does not contain any integers . let @xmath854 , @xmath855 and @xmath856 . for an oriented fiberwise morse function @xmath90 and its fiberwise gradient @xmath18 , let @xmath556 be the restriction of @xmath18 on @xmath857 . suppose that there are no @xmath287-intersections in @xmath857 for @xmath556 . for simplicity , we assume assumption [ assum : mu0 ] for a spin structure @xmath374 on @xmath10 . choose the canonical stable framing @xmath76 as in lemma [ lem : cframing ] that is compatible with @xmath374 . let @xmath858 be a small disk that is disjoint from critical loci of @xmath18 and put @xmath859 . the negative gradient of @xmath112 induces a diffeomorphism @xmath860 . the orbit of @xmath861 under the flow of @xmath862 forms a cylinder @xmath863 in @xmath857 . put @xmath864 and @xmath865 . now we shall define a surgery of @xmath866 along @xmath857 . take an orientation preserving diffeomorphism @xmath867 such that @xmath868 , an oriented fiberwise morse function @xmath869 and its fiberwise gradient @xmath870 . let @xmath871\to \kappa^{-1}[\alpha,\alpha+\eta]$ ] ( @xmath872 small ) be a local trivialization of @xmath112 that extends @xmath562 , let @xmath873\to \kappa^{-1}[\beta-\eta,\beta]$ ] be a local trivialization of @xmath112 that extends @xmath874 , and let @xmath875 be the local trivialization of the disk bundle @xmath876 such that @xmath877 is the intersection of the gradient line of @xmath862 from @xmath878 with @xmath879 and such that @xmath880}=\bar{g}|_{b'\times [ \alpha,\alpha+\eta]}$ ] and @xmath881}=\bar{h}|_{b'\times [ \beta-\eta,\beta]}$ ] . let @xmath882 be the set of critical points of @xmath883 and let @xmath884 be the set of critical points of @xmath885 such that @xmath886 and @xmath887 are the endpoints of a critical locus of @xmath888 for each @xmath125 . let @xmath889 . [ def : adapted ] we say that @xmath890 is _ adapted to @xmath891 _ if the following conditions ( 1)(4 ) are satisfied . 1 . on a collar neighborhood of @xmath892 in @xmath893 , @xmath465 and @xmath870 agree with @xmath894 and @xmath895 respectively . 2 . on a collar neighborhood of @xmath896 in @xmath893 , @xmath465 and @xmath870 agree with @xmath897 and @xmath898 respectively . 3 . on @xmath899 , @xmath465 and @xmath870 agree with @xmath900 and @xmath901 respectively . the number of al - paths of @xmath870 from @xmath886 to @xmath902 counted with signs is the kronecker delta @xmath903 . the condition ( 4 ) is an analogue of pure braid with zero linking numbers . according to ( * ? ? ? * lemma 4.10 ) , the action of al - paths agrees with the action of cobordism on homology . this together with the condition ( 4 ) implies that the diffeomorphism @xmath904 represents an element of the torelli group . for @xmath905 that is adapted to @xmath891 , we define the _ adapted surgery _ @xmath906 as the 3-manifold obtained from @xmath10 by removing @xmath907 and pasting @xmath893 back by the diffeomorphism @xmath908 by the conditions ( 1 ) , ( 2 ) and ( 3 ) for the adaptedness , @xmath94 and @xmath18 on @xmath909 extends smoothly over @xmath906 by @xmath465 and @xmath870 on @xmath893 . we denote by @xmath910 and @xmath911 the resulting function and vector field on @xmath906 respectively . let @xmath912 be the projection obtained by gluing @xmath913 and @xmath914 . we say that @xmath915 is obtained from @xmath891 by adapted surgery with respect to @xmath502 . [ exa : borromean ] suppose that @xmath916 is of genus 3 and the restriction of @xmath94 on @xmath917 is a minimal morse function , i.e. , the numbers of critical points of index @xmath918 are @xmath919 respectively . suppose that there is a level curve @xmath920 of @xmath921 such that * three of the 6 critical points of @xmath921 of index 1 lie below @xmath289 and the rest of those of index 1 lie above @xmath289 and * @xmath922 $ ] has four boundary components , i.e. , a disk with three holes . then a borromean 3-strand braid @xmath923=(\sigma_1\sigma_2^{-1})^3 $ ] gives rise to a relative diffeomorphism of @xmath690 relative to @xmath924 , which can be realized by a sequence of 1-handle - slides below @xmath289 . one has a 1-parameter family of morse functions on @xmath690 that induces the handle - slides , where the handle structures for the endpoints of the path may be assumed to be equal . extending such 1-parameter family by trivial family above @xmath289 , one obtains a 1-parameter family of morse functions on @xmath917 that gives rise to an adapted 5-tuple @xmath890 . it is known that such a surgery may yield topologically different 3-manifold . suppose that a triple @xmath925 is obtained from @xmath891 by a small perturbation . then we perturb @xmath926 as follows . put @xmath927\tcoprod [ \beta-\eta,\beta])\cup ( b'\times i).\ ] ] we consider the pullback @xmath928 where @xmath929 are the maps fixed above by using @xmath891 . then we deform @xmath930 slightly to @xmath931 so that @xmath932 agrees with @xmath933 on @xmath30 . similarly , we deform @xmath465 and @xmath870 slightly to @xmath934 and @xmath935 so that they agree with the pullbacks @xmath936 and @xmath937 on @xmath30 . after surgery by @xmath502 with @xmath938 , we get a triple @xmath939 on @xmath906 . we say that such a triple @xmath939 is obtained by an _ adapted perturbation _ from @xmath915 with respect to @xmath925 . given a spin structure @xmath374 on @xmath10 , one can choose a spin structure on @xmath906 as follows . as in @xcite , we consider a spin structure on an @xmath23-manifold @xmath24 as an element @xmath940 , where @xmath941 is the orthonormal frame bundle of @xmath34 , such that the restriction of @xmath374 to each fiber is non trivial in @xmath942 . for a manifold @xmath24 , let @xmath943 denote the set of spin structures on @xmath24 . let @xmath24 be a closed oriented manifold obtained from two compact oriented spinnable manifolds @xmath944 and @xmath945 with connected boundaries @xmath946 respectively , by identifying the boundaries by an orientation reversing diffeomorphism @xmath947 . suppose that the set @xmath948 is not empty . then @xmath24 is spinnable and the restriction map @xmath949 is injective . ( in such case , for each @xmath950 , there is a unique spin structure @xmath951 on @xmath24 that restricts to the given pair . ) for example , if @xmath905 is adapted to @xmath891 , then in particular @xmath952 induces the identity on @xmath953 , as guaranteed by the condition ( 4 ) for the adaptedness , and @xmath952 acts trivially on the spin structure on @xmath917 . hence the restriction of @xmath374 to @xmath954 can be extended to a spin structure @xmath955 on @xmath906 . we fix one such for each @xmath502 . note that by ( * ? ? ? * lemma 5 ) , one has @xmath956 for disjoint surgeries @xmath957 . let @xmath958 be a disjoint collection of small intervals that are disjoint from @xmath959 . let @xmath960 be a sequence of 5-tuples on @xmath961 as in definition [ def : adapted ] that are adapted to @xmath891 . suppose that there are no @xmath287-intersections in @xmath962 for @xmath18 for all @xmath125 . let @xmath963 and put @xmath964 = \sum_{s\subset t}(-1)^{|s|}(m_s,[\kappa_s],[f_s])\ ] ] where @xmath965 denotes @xmath966 for each object @xmath967 . we define @xmath968 ) & = \sum_{s\subset t}(-1)^{|s|}z_n(m_s,[\kappa_s],[f_s]),\\ \wz_n([m , t],\mathfrak{s}_t ) & = \sum_{s\subset t}(-1)^{|s|}\wz_n(m_s,\mathfrak{s}_s,[\kappa_s],[f_s ] ) . \end{split}\ ] ] let @xmath969 be the vector space over @xmath74 spanned by alternating sums @xmath970 $ ] with @xmath971 where @xmath736 is as in [ ss : alternatingsum ] . then the sequence @xmath972 forms a descending filtration , i.e. , @xmath973 let @xmath974 $ ] denote the alternating sum of @xmath975)$ ] that is obtained from @xmath970 $ ] by forgetting the classes of fiberwise morse functions . let @xmath976 be the vector space over @xmath74 spanned by alternating sums @xmath974 $ ] with @xmath971 where @xmath736 is as in [ ss : alternatingsum ] . again , @xmath977 forms a descending filtration . we consider the following problem . for each @xmath978 , determine the structure of the quotient space 1 . @xmath979 2 . @xmath980 although we do not have a solution to the problem , the invariants @xmath1 and @xmath3 are helpful for understanding of the quotient spaces . we do not know whether the filtrations are the right ones that correspond to the space @xmath73 , but we think that there should be a close connection of the filtrations to @xmath73 . let @xmath981 , @xmath362 , @xmath367 ( @xmath363 ) be as in [ ss : moduli_hori ] . suppose that for each @xmath125 the triple @xmath982 is close to @xmath891 with respect to the @xmath95-topology so close that they are obtained from @xmath891 by a small isotopy in @xmath10 . let @xmath983 be as above that are adapted to @xmath891 and let @xmath963 . let @xmath984 , @xmath363 , @xmath985 , be triples obtained by adapted perturbations from @xmath986 with respect to @xmath982 . moreover , replacing @xmath987 with @xmath988 , we obtain triples @xmath989 . put @xmath990 $ ] , @xmath991 and @xmath992 . let @xmath993 be the critical locus of @xmath94 that intersects @xmath994 and @xmath995 at @xmath996 and @xmath997 respectively . the isotopy which takes @xmath891 to @xmath982 also takes critical loci @xmath993 of @xmath94 to those of @xmath368 . let @xmath998 be the critical locus of @xmath368 corresponding to @xmath993 by the isotopy . then let @xmath999 and @xmath1000 . for a fixed number @xmath1001 and for a triple @xmath1002 , take critical points @xmath1003 , @xmath1004 , @xmath1005 of index 1 in @xmath10 . then we define the _ @xmath1006-graph _ @xmath1007 as the @xmath1006-shaped unitrivalent graph such that the three univalent vertices are labeled @xmath1008 respectively . the order of @xmath1009 determines a vertex - orientation of the trivalent vertex of @xmath1007 . we impose the relation @xmath1010 for @xmath1011 . let @xmath1012 be the moduli space of al - graphs in @xmath1013 for @xmath1007 . the coorientation of the moduli space is determined as the wedge product of the coorientations of the loci of the descending and the ascending manifolds of critical loci considered at the trivalent vertex . we take the wedge product of the coorientations using the vertex - orientation . put @xmath1014 note that the product @xmath1015 does not depend on the choice of vertex - orientation . we define @xmath1016 by the same formula as @xmath1017 but restricting the sum @xmath1018 to triples @xmath1009 consisting only of points of @xmath1019 or only of points of @xmath1020 . for a pair @xmath1021 and a number @xmath1022 , take critical points @xmath1023 and @xmath1024 in @xmath1025 and @xmath1026 respectively of index 1 . suppose that @xmath122 are such that if @xmath1027 then @xmath1028 and if @xmath1029 then @xmath1030 . then we define a _ chord _ @xmath1031 as the connected univalent graph @xmath1032 with vertices labeled by @xmath43 and @xmath123 respectively . let @xmath1033 be the moduli space of al - paths in @xmath1034 going from @xmath43 to @xmath123 . this gives a 0-dimensional chain in @xmath14 , whose twisted homology class is represented by a rational function in @xmath78 . put @xmath1035 where @xmath1036 . we define a product of several @xmath1006-graphs by the disjoint union and extend it to the @xmath74-linear combinations of products of @xmath1006-graphs by @xmath74-multilinearity . we also define a product of several @xmath78-colored chords by the disjoint union and extend it by @xmath74-multilinearity . then we define a non - symmetric pairing @xmath1037 , where @xmath545 is a product of @xmath216 @xmath1006-graphs and @xmath690 is a product of @xmath1038 @xmath78-colored chords , as follows . we set @xmath1039 unless the labels of the @xmath1040 legs in @xmath545 and @xmath690 match , i.e. , unless the sets of the labels of the @xmath1040 univalent vertices in @xmath545 and @xmath690 agree . if the labels of the @xmath1040 legs in @xmath545 and @xmath690 match , we set @xmath1041 to be the @xmath78-colored graph obtained by joining all pairs of legs which have the same label . we extend the pairing to @xmath74-linear combinations by @xmath74-linearity . for example , @xmath1042.\ ] ] [ thm : surgery ] let @xmath736 be as in [ ss : alternatingsum ] and let @xmath1043 . 1 . if @xmath1044 , then @xmath1045,\mathfrak{s}_t)=0.\ ] ] 2 . if @xmath1046 , then @xmath1045,\mathfrak{s}_t ) = ( 2n)!\,\bigl\langle \prod_{i=1}^{2n}z_{n , y}^t(i),\prod_{k=1}^{3n}z_c^t(k)\bigr\rangle \in\cala_n(\widehat{\lambda}).\ ] ] theorem [ thm : surgery ] shows that the values of @xmath1 for @xmath970 $ ] with @xmath1047 does not depend on the choice of spin structures @xmath1048 for @xmath988 , hence @xmath1 induces a well - defined linear map @xmath1049 we shall only give an outline of the proof since similar argument has been given previously by several authors ( e.g. , @xcite ) for different purposes . we assume that for each @xmath125 the triple @xmath1050 is close to @xmath891 as above . for each @xmath988 and for each @xmath125 such that @xmath1051 , we put @xmath1052 . if @xmath1053 , an al - graph @xmath1054 in @xmath1055 , @xmath988 , having a trivalent vertex in the complement of @xmath1056 does not survive in the alternating sum @xmath1057)$ ] since in such a case either of the following occurs . 1 . there is @xmath528 such that no trivalent vertex of @xmath1054 belongs to @xmath1058 . 2 . there is a unique graph @xmath1059 in @xmath1060 that represents the same monomial @xmath77-colored graph as @xmath1054 , that agrees with @xmath1054 outside @xmath1061 and that has no trivalent vertex in @xmath1062 . in the case ( a ) , @xmath1054 collapses to an al - graph in @xmath1063 that represents the same monomial @xmath77-colored graph as @xmath1054 . by the condition ( 4 ) of definition [ def : adapted ] , the counts of both graphs are equal . such a pair cancels each other out in @xmath1064,[f_s])-z_n(m_{s\setminus\{\tau_j\}},[\kappa_s],[f_{s\setminus\{\tau_j\}}])$ ] . in the case ( b ) , similar cancellation occurs in + @xmath1065,[f_{s\cup\{\tau_j\}}])-z_n(m_s,[\kappa_s],[f_s])$ ] . in the case @xmath1044 , either of ( a ) or ( b ) always occurs and hence the claim ( 1 ) holds . in the case @xmath1046 , the only surviving terms in the alternating sum @xmath1057)$ ] are those for al - graphs @xmath1054 in @xmath1066 whose set of @xmath216 trivalent vertices intersects @xmath1062 for every @xmath1067 . this means that for each @xmath1067 , there is a horizontal tree component of @xmath1054 with one trivalent vertex , i.e. , a @xmath1006-graph , in @xmath1062 . the sum of counts of such @xmath1054 gives the right hand side of the formula of ( 2 ) . the reason for the coefficient @xmath1068 is that for each al - graph @xmath1054 , the same contribution is counted for @xmath1068 different labelings for the trivalent vertices of @xmath1054 . finally , we should check that the alternating sum of the correction terms vanishes , if @xmath1069 . for simplicity , we assume assumption [ assum : mu0 ] for @xmath375 and @xmath1070 for all @xmath125 . if @xmath1070 does not satisfy assumption [ assum : mu0 ] , we have only to replace @xmath1071 with the 16 iterations of @xmath1071 when defining the correction term and to divide the result by 16 . by assumption [ assum : mu0 ] for @xmath375 and @xmath1070 , there exist spin cobordisms @xmath1072 with @xmath1073 ( mod 16 ) that spin bound @xmath1074 . then one can find sequences @xmath1075 of sections of @xmath1076 extending given ones on the boundaries , and the correction term in @xmath1077,[f_s])$ ] is @xmath1078 the total alternating sum is as follows . @xmath1079\\ = & \sum_{j=1}^m -\bigl(\#\{s;\,\tau_j\in s,\ , |s|\mbox { is even}\}-\#\{s;\,\tau_j\in s,\ , |s|\mbox { is odd}\}\bigr)\,z_n^\anomaly(\vec{\rho}_{w_j})\\ = & \sum_{j=1}^m -(1 - 1)^{m-1}\,z_n^\anomaly(\vec{\rho}_{w_j})=0 . \end{split}\ ] ] we have another surgery formula analogous to theorem [ thm : surgery ] for @xmath3 . the following theorem is a special case of lescop s lagrangian preserving surgery formula of @xmath3 in @xcite . [ thm : surgery2 ] let @xmath736 be as in [ ss : alternatingsum ] and let @xmath1043 . 1 . if @xmath1080 , then @xmath1081)=0.\ ] ] 2 . if @xmath1082 , then @xmath1081 ) = 2!\,\bigl\langle z_{1,y}^t(1)_0z_{1,y}^t(2)_0,z_c^t(1)z_c^t(2)z_c^t(3)\bigr\rangle \in\cala_1(\widehat{\lambda})/o_\delta.\ ] ] the proof of theorem [ thm : surgery2 ] is almost parallel to theorem [ thm : surgery ] . we remark that for @xmath1083 , no contribution of @xmath234 survive in the alternating sum . hence so are the correction terms involving @xmath600 . theorem [ thm : surgery2 ] ( 1 ) shows that @xmath3 induces a well - defined linear map @xmath1084 and ( 2 ) gives an explicit formula for this map . let @xmath1085 , @xmath1086 . let @xmath90 be an oriented fiberwise morse function for the fibration @xmath112 and let @xmath18 be its gradient along the fibers . suppose that the @xmath287-intersections in @xmath18 are disjoint from both @xmath1087 and @xmath1088 . put @xmath1089 $ ] and @xmath1090 $ ] . for @xmath97 , put @xmath1091 , @xmath1092 and @xmath1093 and let @xmath1094 denote the set of critical points of @xmath98 . let @xmath1095 denote the set of critical points of @xmath98 of index @xmath125 . now we define a chain complex @xmath1096 using al - paths . for @xmath1097 , let @xmath1098 be the subset of @xmath264 consisting of al - paths from @xmath85 to a point of @xmath1099 that does not intersect @xmath1100 . as paths in @xmath1101 , we consider a _ reverse _ al - path , which is defined by using @xmath1102 instead of @xmath112 in the definition of al - path . let @xmath1103 be the ( closed ) moduli space of reverse al - paths in @xmath10 . for @xmath1097 , let @xmath1104 be the subset of @xmath1103 consisting of reverse al - paths from @xmath85 to a point of @xmath1101 that does not intersect @xmath1105 . we orient @xmath1104 and @xmath1098 by using the orientations of descending and ascending manifold loci and the orientations of level surfaces determined by @xmath862 , as in [ ss : coori ] or @xcite . put @xmath1106 the assignment of the terminal endpoint of a path defines a continuous map @xmath1107 for @xmath1108 , let @xmath1109 denote the descending manifold of @xmath286 for @xmath1110 and let @xmath1111 be its compactification to the space of possibly broken flow lines ( see e.g. , @xcite ) . let @xmath1112 denote the continuous map that extends the inclusion @xmath1113 . roughly , the set of maps @xmath269 and @xmath1114 gives a cell - like structure on @xmath10 whose degree @xmath125 part is generated by @xmath1115 for @xmath1116 and @xmath1117 for @xmath1118 . the incidence coefficients are determined as follows . for @xmath1119 , let @xmath1120 be the subspace of @xmath1121 or @xmath1122 consisting of al - paths ( or reverse al - paths ) from @xmath85 to @xmath286 . if @xmath1123 and @xmath1124 satisfies the conditions ( 1 ) , ( 2 ) , ( 3 ) below and if @xmath18 is generic , then @xmath1120 is a compact 0-dimensional manifold . 1 . @xmath1125 , @xmath1126 and @xmath1127 . @xmath1125 , @xmath1108 and @xmath1128 . @xmath1129 , @xmath1108 and @xmath1127 . moreover , in such cases @xmath1120 has a natural orientation ( sign ) as follows . if @xmath1130 satisfies ( 1 ) or ( 3 ) above , then we orient @xmath1120 as usual , namely , comparing @xmath1131 ( @xmath1132 , @xmath1133 ) with the orientation of the level curve @xmath1134 of @xmath98 including @xmath43 . here , the orientation of @xmath1135 is given by @xmath1136 . if @xmath1130 satisfies ( 2 ) above , then we orient @xmath1120 by the signs of al - paths from @xmath85 to @xmath286 . now we put @xmath1137 let @xmath1138 , @xmath1139 be the boundary operators of morse complexes @xmath1140 respectively . namely , @xmath1141 let @xmath1142 be the operator defined by @xmath1143 put @xmath1144 and define @xmath1145 by @xmath1146 this definition is motivated by the twisted tensor product in @xcite . [ prop : al - homology ] the pair @xmath1147 forms a chain complex and its homology is canonically isomorphic to @xmath1148 . since @xmath1149 and @xmath1150 , we have @xmath1151 thus the condition @xmath1152 is equivalent to the condition that @xmath1153 is a chain map . it is enough to see that @xmath1154 is given by the boundaries of the 1-dimensional moduli spaces @xmath1120 for @xmath1155 , @xmath1156 . from the definition of the moduli space of al - paths given in @xcite , it follows that @xmath1157 consists of al - paths between @xmath85 and @xmath286 of the forms @xmath1158 , where @xmath235 goes from @xmath85 to some @xmath1159 and @xmath236 goes from @xmath1160 to @xmath286 . this implies that @xmath1161 it suffices to check that this differs by a multiple of @xmath1162 from the coefficient of @xmath286 in @xmath1163 . we assume for simplicity that the bundle @xmath112 is trivial and that @xmath94 is a trivial 1-parameter family of a morse function on @xmath89 since the mechanism of inducing boundary orientations for general case is the same except that the signs of al - paths with @xmath287-intersections are multiplied . moreover , we have only to check the claim for a trivial family over a geometric 1-simplex @xmath1164 $ ] ( replacing @xmath1165 with @xmath1085 , @xmath1166 as elements of @xmath675 $ ] respectively ) . since the orientations of @xmath1167 and @xmath1168 induced from that of @xmath1169 are opposite , the orientation of @xmath1120 induces opposite signs in the boundary operators on @xmath1167 and @xmath1168 . hence the coefficient of @xmath286 in @xmath1170 is @xmath1171 . since @xmath1153 is a chain map , the system @xmath1172 satisfies the axiom of igusa s @xmath1173-functor ( @xcite ) . as mentioned above , @xmath1174 agrees with igusa s twisted complex for the @xmath1173-functor , which gives the homology of @xmath10 . ( one may use mayer vietoris argument and ( * ? ? ? * lemma 4.10 ) to see this directly . ) [ cor : zynontrivial ] suppose that the fiberwise morse function @xmath94 is such that @xmath1175 is minimal , i.e. , @xmath1176 and @xmath1177 . suppose moreover that @xmath1178 . if @xmath982 are sufficiently close to @xmath891 and generic , then the element @xmath1179\wedge [ p_2]\wedge [ p_3]\in \textstyle\bigwedge^3 h_2(m;\q)\ ] ] where the sum is over @xmath1180 , @xmath1181 , @xmath1182 , is well - defined and is a multiple of the dual to the triple cup product on @xmath1183 . hence @xmath1017 ( or @xmath1016 ) is nontrivial if the triple cup product @xmath1184 of the mapping torus of the surgery @xmath1071 is nontrivial . let @xmath1185 denote an oriented surface of genus @xmath562 with @xmath258 boundary circles and let @xmath1186 be a diffeomorphism such that @xmath1187 and such that the induced map @xmath1188 is represented by the matrix @xmath1189 . since @xmath1190 is a product of elementary matrices , @xmath1191 can be realized by 1-handle - slides in the standard handle presentation of @xmath1192 in figure [ fig : spine - transverse ] ( b ) . by boundary connected sums between the 0-handles , we obtain a diffeomorphism @xmath1193 such that the induced map on @xmath1194 is represented by @xmath1195 . let @xmath1196 be an extension of @xmath1197 by the identity . let @xmath10 be the mapping torus of @xmath278 and let @xmath88 be the projection . since @xmath1198 is invertible in @xmath0 , we have @xmath22 ( see e.g. , ( * ? ? ? * lemma a.1 ) ) . let @xmath90 be an oriented fiberwise morse function that restricts to a minimal morse function on a fiber , which exists . then @xmath94 satisfies the assumption of example [ exa : borromean ] . let @xmath1199 be small disjoint intervals and put @xmath1200 . we identify @xmath1201 with @xmath1202 . consider two disjoint @xmath1006-surgeries @xmath1203 on @xmath1204 respectively as in example [ exa : borromean ] , as follows . a @xmath1006-surgery can be considered as a replacement of @xmath1205 where @xmath1206 is a ( thickened ) ribbon graph on @xmath1207 of genus 0 with four boundary components . assume that the spines of @xmath1194 and of @xmath1208 intersect transversally as in figure [ fig : spine - transverse ] ( a ) . from corollary [ cor : zynontrivial ] , one sees that @xmath1209 and @xmath1210 are of the following forms : @xmath1211 where @xmath1212 and @xmath1213 , @xmath1214 and `` permutations '' are the contributions of the corresponding triples of critical loci with the labels of the edges permuted . the coefficients in @xmath1215 are given by matrix entries of @xmath1216 where @xmath1217 . hence for @xmath1203 , we have @xmath1218)\\ & = 2!\,3!\,\delta\ve\,\tr_\theta\bigl(\frac{1-t_1}{d(t_1)}\otimes\frac{1-t_2}{d(t_2)}\otimes\frac{1-t_3}{d(t_3)}+\frac{t_1(1-t_1)}{d(t_1)}\otimes\frac{t_2(1-t_2)}{d(t_2)}\otimes\frac{t_3(1-t_3)}{d(t_3)}\bigr)\\ & = 24\,\delta\ve\,\tr_\theta\bigl(\frac{1-t_1}{1 - 3t_1+t^2_1}\otimes\frac{1-t_2}{1 - 3t_2+t^2_2}\otimes\frac{1-t_3}{1 - 3t_3+t^2_3}\bigr ) \end{split}\ ] ] by theorem [ thm : surgery2 ] . in this way , one may find many nontrivial elements in @xmath1219 and in @xmath1220 , where we must use @xmath1017 for @xmath1220 in place of @xmath1221 . a _ fatgraph _ ( or a ribbon graph ) is a vertex - oriented graph . a fatgraph @xmath1054 gives an oriented surface @xmath1222 with spine @xmath1223 , which is a 2-dimensional handlebody consisting of 0- and 1-handles . a handle - slide in such a handlebody corresponds to a local move on a fatgraph . a sequence @xmath1224 of such local moves with @xmath1225 gives rise to a ( possibly unorientable ) fiberwise morse function for a surface bundle @xmath1226 . it is known ( e.g. , @xcite ) that the mapping class group @xmath1227 of punctured surface @xmath1228 can be represented as the set of sequences of `` whitehead moves '' @xmath1229 on ( bordered ) trivalent fatgraphs with @xmath1225 , modulo the relations called involutivity , commutativity and pentagon . a whitehead move can be realized as a sequence of handle - slides . in this way , @xmath1227 can be represented as the set of sequences of handle - slides on fatgraphs modulo the corresponding relations rewritten in terms of handle - slides . the relations can be realized by 2-parameter families of gmf s . it can be checked that the pentagon relation may not be realized by a concordance of oriented fiberwise morse functions . thus , to get an invariant of 3-manifolds , the method of this paper needs to be improved , although we do not know whether the pentagon relation changes the value of @xmath1 or not . the author is supported by jsps grant - in - aid for young scientists ( b ) 26800041 . c. lescop , _ invariants of knots and 3-manifolds derived from the equivariant linking pairing _ , chern simons gauge theory : 20 years after , ams / ip stud . adv . math . * 50 * ( 2011 ) , amer . math . soc . , providence , ri , 217242

we apply lescop s construction of @xmath0-equivariant perturbative invariant of knots and 3-manifolds to the explicit equivariant propagator of `` al - paths '' given in @xcite . we obtain an invariant @xmath1 of certain equivalence classes of fiberwise morse functions on a 3-manifold fibered over @xmath2 , which can be considered as a higher loop analogue of the lefschetz zeta function and whose construction will be applied to that of finite type invariants of knots in such a 3-manifold . we also give a combinatorial formula for lescop s equivariant invariant @xmath3 for 3-manifolds with @xmath4 fibered over @xmath2 . moreover , surgery formulas of @xmath1 and @xmath3 for alternating sums of surgeries are given . this gives another proof of lescop s surgery formula of @xmath3 for special kind of 3-manifolds and surgeries , which is simple in the sense that the formula is obtained easily by counting certain graphs in a 3-manifold . ( 2015 )

*introduction* *preliminaries* *the invariants @xmath310 and @xmath3* *proof of invariance of @xmath1* *surgery formulas of @xmath1 and @xmath3* *al-paths and the homology of @xmath10* *concluding remarks* *acknowledgments.*

one of the most profound observations in theoretical physics is the relationship between finite - temperature quantum theories and zero - temperature quantum theories which are compactified on a circle . indeed , the fundamental idea behind this so - called `` temperature / radius correspondence '' is that the free - energy density of a theory at finite temperature @xmath2 can be reformulated as the vacuum - energy density of the same theory at zero temperature , but with the euclidean time dimension compactified on a circle of radius @xmath3 . this connection between temperature and geometry is a deep one , stretching from quantum mechanics and quantum field theory all the way into string theory . this extension to string theory is truly remarkable , given that the geometric compactification of string theory gives rise to numerous features which do not , at first sight , have immediate thermodynamic analogues or interpretations . for example , upon spacetime compactification , closed strings accrue not only infinite towers of kaluza - klein `` momentum '' states but also infinite towers of winding states . while the kaluza - kelin momentum states are easily interpreted in a thermal context as the matsubara modes corresponding to the original zero - temperature states , it is not _ a priori _ clear what thermal interpretation might be ascribed to these winding states . likewise , as a more general ( but not unrelated ) issue , closed - string one - loop vacuum energies generally exhibit additional symmetries such as modular invariance which transcend field - theoretic expectations . while the emergence of modular invariance is clearly understood for zero - temperature geometric compactifications , the need for modular invariance is perhaps less obvious from the thermal perspective in which one would simply write down a boltzmann sum corresponding to each string state which survives the gso projections . both of these issues tended to dominate the earliest discussions of string thermodynamics in the mid-1980 s . historically , they were first flashpoints which seemed to show apparent conflicts between the thermal and geometric approaches which had otherwise been consistent in quantum field theory . however , it is now well understood that there are ultimately no conflicts between these two approaches @xcite . indeed , modular invariance emerges naturally upon relating the integral of the boltzmann sum over the `` strip '' in the complex @xmath4-plane to the integral of the partition function over the fundamental domain of the modular group @xcite . likewise , thermal windings emerge naturally as a consequence of modular invariance and can be viewed as artifacts arising from this mapping between the strip and the modular - group fundamental domain . there is , however , one additional feature which can generically arise when a theory experiences a geometric compactification : because of the topologically non - trivial nature of the compactification , it is possible for a non - zero gauge flux to pierce the compactification circle . in other words , the compactification might involve a non - trivial wilson line . viewed from the thermodynamic perspective , this corresponds to nothing more than the introduction of a chemical potential . however , as we shall see , this is ultimately a rather unusual chemical potential : it is not only imaginary but also temperature - dependent . such chemical potentials have occasionally played a role in studies of finite - temperature field theory ( particularly finite - temperature qcd @xcite ) . however , with only a few exceptions , such chemical potentials ( and the wilson lines to which they correspond ) have not historically played a significant role in discussions of finite - temperature string theory . at first glance , it might seem reasonable to hope ( or simply postulate ) that wilson lines should play no role in discussions of finite - temperature string theory . however , wilson lines play such a critical role in determining the allowed possibilities for self - consistent geometric compactifications of string theory that it is almost inevitable that they should play a significant role in finite - temperature string theories as well . indeed , the temperature / radius correspondence essentially guarantees this . thus , it is natural to expect that theories with non - trivial wilson lines will be an integral part of the full landscape of possibilities for string theories at finite temperature , that they will be part of the full `` thermal string landscape '' . a heuristic argument can be invoked in order to illustrate the connection that might be expected between wilson lines and string theories at finite temperature . as we know , thermal effects treat bosons and fermions differently and thereby necessarily break whatever spacetime supersymmetry might have existed at zero temperature . in string theory there are tight self - consistency constraints which relate the presence or absence of spacetime supersymmetry to the breaking of the corresponding gauge symmetry , and these connections hold even at zero temperature . for example , the @xmath1 heterotic string in ten dimensions is necessarily supersymmetric , and it is inconsistent to break this supersymmetry without simultaneously introducing a non - trivial wilson line ( or in this context , a gauge - sensitive orbifold twist ) which also breaks the @xmath1 gauge group . indeed , the two are required together . even for the @xmath0 gauge group , a similar situation arises : although there exist two @xmath0 heterotic strings , one supersymmetric and the other non - supersymmetric , the @xmath5 orbifold which relates them to each other is not simply given by the susy - breaking action @xmath6 , where @xmath7 is the spacetime fermion number . rather , the required orbifold which twists the supersymmetric @xmath0 heterotic string to become the non - supersymmetric @xmath0 heterotic string is given by @xmath8 where @xmath9 is a special non - zero wilson line which acts non - trivially on the gauge degrees of freedom . this example will be discussed further in sect . indeed , such a wilson line is needed even though we are not breaking the @xmath0 gauge symmetry in passing from our supersymmetric original theory to our final non - supersymmetric theory . such examples indicate the deep role that wilson lines play in zero - temperature string theory , and which they might therefore be expected to play in a finite - temperature context as well . in this paper , we shall undertake a systematic examination of the role that such wilson lines might play in string thermodynamics . we shall concentrate on the zero - temperature heterotic and type i strings in ten dimensions , and survey the possible wilson lines which might be introduced in their finite - temperature extensions . as we shall see , this gives rise to a rich structure of possible thermal string theories , and we shall demonstrate that these new thermal string theories can be interpreted as extrema of a continuous thermal free - energy `` landscape '' . in fact , some of these new thermal theories even have non - traditional hagedorn temperatures , an observation which we shall discuss ( and explain ) in some detail . our analysis will also uncover a unique finite - temperature extension of the heterotic @xmath0 and @xmath1 strings which involves a non - trivial wilson line , but which like the traditional finite - temperature extension without wilson lines is metastable in this thermal landscape . such new theories might therefore play an important role in describing the correct thermal vacuum of our universe . this paper is organized as follows . in order to set the stage for our subsequent analysis , in sect . ii we provide general comments concerning string theories at finite temperature and in sect . iii we discuss the possible role that wilson lines can play in such theories . we also discuss the equivalence between such thermal wilson lines and temperature - dependent chemical potentials . in sect . iv , we then survey the specific wilson lines that may self - consistently be introduced when constructing our thermal theories , concentrating on the two supersymmetric heterotic strings in ten dimensions as well as the supersymmetric type i string in ten dimensions . in sect . v , we demonstrate that non - trivial wilson lines can also affect the hagedorn temperatures experienced by these strings , and show how such shifts in the hagedorn temperature can be reconciled with the asymptotic densities of the zero - temperature bosonic and fermionic string states . then , in sect . vi , we extend our discussion in order to consider continuous thermal wilson - line `` landscapes '' for both heterotic and type i strings . it is here that we discuss which wilson lines lead to `` stable '' and/or `` metastable '' theories . finally , in sect . vii , we conclude with some general comments and discussion . an appendix summarizes the notation and conventions that we shall be using throughout this paper . we begin by discussing the manner in which a given zero - temperature string model can be extended to finite temperature . this will also serve to establish our conventions and notation . because of its central role in determining the thermodynamic properties of the corresponding finite - temperature string theory , we shall focus on the calculation of the one - loop string thermal partition function @xmath10 . the situation is slightly different for closed and open strings , so we shall discuss each of these in turn . in order to begin our discussion of closed strings at finite temperature , we begin by reviewing the case of a one - loop partition function for a closed string at zero temperature . our discussion will be as general as possible , and will therefore apply to all closed strings , be they bosonic strings , type ii superstrings , or heterotic strings . for closed strings , the one - loop partition function is defined as z_model ( ) tr ( -1)^f ^h_r q^h_l [ pf ] where the trace is over the complete fock space of states in the theory , weighted by a spacetime statistics factor @xmath6 . here @xmath11 where @xmath4 is the one - loop ( torus ) modular parameter , and @xmath12 denote the worldsheet energies for the right- and left - moving worldsheet degrees of freedom , respectively . note that in general , @xmath13 is the quantity which appears in the calculation of the one - loop cosmological constant ( vacuum - energy density ) of the model : ^(d ) -m^d _ f d^2 ( i m ) ^2 z_model ( ) [ cosconstdef ] where @xmath14 is the number of uncompactified spacetime dimensions , where @xmath15 is the reduced string scale , and where : |re| , im>0 , ||1[fdef ] is the fundamental domain of the modular group . of course , the quantity in eq . ( [ cosconstdef ] ) is divergent for the compactified bosonic string as a result of the physical bosonic - string tachyon . given the general form for the zero - temperature one - loop string partition function in eq . ( [ pf ] ) , it is straightforward to construct its generalization to finite temperature . as is well known in field theory , the free - energy density @xmath16 of a boson ( fermion ) in @xmath14 spacetime dimensions at temperature @xmath2 is nothing but the zero - temperature vacuum - energy density @xmath17 of a boson ( fermion ) in @xmath14 spacetime dimensions , where the ( euclidean ) timelike dimension is compactified on a circle of radius @xmath18 about which the boson ( fermion ) is taken to be periodic ( anti - periodic ) . we shall refer to this observation as the `` temperature / radius correspondence '' . this correspondence generally extends to string theory as well @xcite , state by state in the string spectrum . however , for closed strings there is an important extra ingredient : we must include not only the `` momentum '' matsubara states that arise from the compactification of the timelike direction , but also the `` winding '' matsubara states that arise due to the closed nature of the string . indeed , both types of states are necessary for the modular invariance of the underlying theory at finite temperature . as a result , a given zero - temperature string state will accrue not a single sum of matsubara / kaluza - klein modes at finite temperature , but actually a _ double sum _ consisting of the matsubara / kaluza - klein momentum modes as well as the matsubara winding modes . the final expressions for our finite - temperature string partition functions @xmath19 must also be modular invariant , satisfying the constraint @xmath20 . because our thermal theory necessarily includes two groups of momentum quantum numbers ( namely those with @xmath21 as well as those with @xmath22 ) which are treated separately ( corresponding to spacetime bosons and fermions respectively ) , modular invariance turns out to imply that winding numbers @xmath23 which are even will likewise be treated separately from those that are odd . as a result , the most general thermal string - theoretic partition function will take the form @xcite z_string(,t ) & = & z^(1 ) ( ) _ 0(,t ) + z^(2 ) ( ) _ 1/2(,t ) + & & + z^(3 ) ( ) _ 0(,t ) + z^(4 ) ( ) _ 1/2(,t ) . [ eomix ] here @xmath24 and @xmath25 represent the thermal portions of the partition function , namely the double sums over appropriate combinations of thermal momentum and winding modes @xcite . specifically , the @xmath24 functions include the contributions from even winding numbers @xmath26 along with either integer or half - integer momenta @xmath27 , while the @xmath25 functions include the contributions from odd winding numbers @xmath26 with either integer or half - integer momenta @xmath27 . these functions are defined explicitly in the appendix . likewise , the terms @xmath28 ( @xmath29 ) represent the traces over those subsets of the zero - temperature string states in eq . ( [ pf ] ) which accrue the corresponding thermal modings at finite temperature . for example , @xmath30 represents a trace over those string states in eq . ( [ pf ] ) which accrue even thermal windings @xmath31 and integer thermal momenta @xmath32 , and so forth . modular invariance for @xmath33 as a whole is then achieved by demanding that each @xmath28 transform exactly as does its corresponding @xmath34 function . in the @xmath35 limit , it is easy to verify that @xmath36 and @xmath37 each vanish while @xmath38 with @xmath39 . as a result , we find that z_string(t ) t as t0 . [ preorigmodel ] the divergent prefactor proportional to @xmath40 in eq . ( [ preorigmodel ] ) is a mere rescaling factor which reflects the effective change of the dimensionality of the theory in the @xmath35 limit . specifically , this is an expected dimensionless volume factor which emerges as the spectrum of surviving matsubara momentum states becomes continuous . however , we already know that @xmath13 in eq . ( [ pf ] ) is the partition function of the zero - temperature theory . as a result , we can relate eqs . ( [ pf ] ) and ( [ eomix ] ) by identifying z_model = z^(1 ) + z^(2 ) . [ origmodel ] we see , then , that the procedure for extending a given zero - temperature string model to finite temperature is relatively straightforward . any zero - temperature string model is described by a partition function @xmath13 , the trace over its fock space . the remaining task is then simply to determine which states within @xmath13 are to accrue integer modings around the thermal circle , and which are to accrue half - integer modings . those that are to accrue integer modings become part of @xmath30 , while those that are to accrue half - integer modings become part of @xmath41 . in this way , we are essentially decomposing @xmath13 in eq . ( [ origmodel ] ) into separate components @xmath30 and @xmath41 . once this is done , modular invariance alone determines the unique resulting forms for @xmath42 and @xmath43 . the final thermal partition function @xmath10 is then given in eq . ( [ eomix ] ) . in complete analogy to eq . ( [ cosconstdef ] ) , we can then proceed to define the @xmath44-dimensional vacuum - energy density ^(d-1 ) - m^d-1 _ f d^2 _ 2 ^ 2 z_string(,t ) [ reducedlambda ] ( where @xmath45 ) , whereupon the corresponding @xmath14-dimensional free - energy density @xmath46 is given by f(t ) = t ^(d-1 ) . [ freeenergydef ] as we see from this discussion , the only remaining critical question is to determine how to decompose @xmath13 as in eq . ( [ origmodel ] ) into the pieces @xmath30 and @xmath41 , to determine which states within @xmath13 are to accrue integer thermal modings ( and thereby be included within @xmath30 ) , and which are to accrue half - integer modings ( and thereby be included within @xmath41 ) . however , this too is relatively simple . in general , a given string model will give rise to states which are spacetime bosons as well as states which are spacetime fermions . in making this statement , we are identifying `` bosons '' and `` fermions '' on the basis of their spacetime lorentz spins . ( by the spin - statistics theorem , this is equivalent to identifying these states on the basis of their bose - einstein or fermi - dirac quantizations . ) as a result , we can always decompose @xmath13 into separate contributions from spacetime bosons and spacetime fermions : z_model = z_boson + z_fermion . [ bosfermdecomp ] however , the temperature / radius correspondence instructs us that bosons should be periodic around the thermal circle , and fermions should be anti - periodic around the thermal circle . in the absence of any other effects , a field which is periodic around the thermal circle will have integer momentum quantum numbers @xmath32 , while a field which is anti - periodic will have half - integer momentum quantum numbers @xmath22 . thus , given the decomposition in eq . ( [ bosfermdecomp ] ) , the standard approach which is taken in the string literature is to identify z^(1)= z_boson , z^(2)= z_fermion . [ boltzmann ] this makes sense , since @xmath30 corresponds to the @xmath47 sector which accrues integer thermal matsubara modes @xmath32 while @xmath41 corresponds to the @xmath48 sector which accrues half - integer thermal matsubara modes @xmath49 . indeed , the choice in eq . ( [ boltzmann ] ) is the unique choice which reproduces the standard boltzmann sum for the states in the string spectrum . we can illustrate this procedure by explicitly writing down the standard thermal partition functions for the ten - dimensional supersymmetric @xmath0 and @xmath1 heterotic strings at finite temperature . at zero temperature , both of these string theories have partition functions given by z_model = z^(8)_boson ( _ v-_s ) l [ pfs ] where @xmath50 denotes the contribution from the eight worldsheet bosons and where the contributions from the right - moving worldsheet fermions are written in terms of the barred characters @xmath51 of the transverse @xmath52 lorentz group . these quantities are defined in the appendix . by contrast , @xmath53 denotes the contributions from the left - moving ( internal ) worldsheet degrees of freedom . written in terms of products @xmath54 of the unbarred characters @xmath55 of the @xmath56 gauge group , these left - moving contributions are given by = [ pfs2 ] states which are spacetime bosons or fermions contribute to the terms in eq . ( [ pfs ] ) which are proportional to @xmath57 or @xmath58 , respectively . the standard boltzmann prescription in eq . ( [ boltzmann ] ) therefore leads us to identify z^(1 ) = z^(8)_boson _ v l , z^(2 ) = -z^(8)_boson _ s l , [ boltzii ] whereupon modular invariance requires that z^(3 ) = -z^(8)_boson _ c l , z^(4 ) = z^(8)_boson _ i l . [ dictated ] we therefore obtain the thermal partition functions z(,t ) = z^(8)_boson _ v _ 0 - _ s _ 1/2 - _ c _ 0 + _ i _ 1/2 l . [ fake ] this is indeed the standard result in the string literature @xcite . we now turn to the case of type i strings . such strings , of course , have both closed and open sectors . because type i strings are unoriented , their one - loop vacuum vacuum energies receive four separate contributions : those from the closed - string sectors have the topologies of a torus and a klein bottle , while those from the open - string sectors have the topologies of a cylinder and a mbius strip . we therefore must consider four separate partition functions : @xmath59 , @xmath60 , @xmath61 , and @xmath62 . at zero temperature , both @xmath59 and @xmath60 are traces over the closed - string states in the theory : z_t ( ) tr ( -1)^f ^h_r q^h_l , z_k ( ) tr ( -1)^f ^h_r q^h_l [ torklein ] where @xmath63 is the orientation - reversing operator which exchanges left - moving and right - moving worldsheet degrees of freedom . thus , taken together , the sum @xmath64 represents a single trace over those closed - string states which are invariant under @xmath63 , as appropriate for an unoriented string . note that because of the presence of the orientifold operator @xmath63 within @xmath60 , the klein - bottle contribution @xmath60 can ultimately be represented as a power series in terms of a single variable @xmath65 where @xmath4 represents the modulus for double - cover of the torus , as given in eq . ( [ taucases ] ) . likewise , corresponding to this are the traces over the open - string states in the theory : z_c ( ) tr ( -1)^f q^h , z_m ( ) tr ( -1)^f q^h , [ cylmob ] where @xmath66 is the open - string worldsheet energy and where @xmath65 with @xmath4 defined in eq . ( [ taucases ] ) . the presence of @xmath63 within @xmath62 guarantees that @xmath67 represents a single trace over an orientifold - invariant set of open - string states . extending these contributions to finite temperatures is also straightforward . the extension of the torus contribution @xmath59 to finite temperatures proceeds exactly as discussed above for closed strings , ultimately leading to an expression of the same form as in eq . ( [ eomix ] ) , with four different thermal sub - contributions @xmath68 . the corresponding finite - temperature klein - bottle contributions can be derived from the finite - temperature torus contribution by implementing the orientifold projection in the finite - temperature trace , ultimately leading to an expression which can be recast in the form z_k(,t ) = z^(1)_k ( ) ( , t ) + z^(2)_k ( ) (,t ) where the thermal functions @xmath69 and @xmath70 are defined in eq . ( [ eeprime ] ) and serve as the open - string analogues of the closed - string thermal @xmath24 functions . likewise , the open - string sector extends to finite temperatures in complete analogy with the closed - string sector , by associating certain states with @xmath69 and others with @xmath70 : z_c(,t ) & = & z^(1)_c ( ) ( , t ) + z^(2)_c ( ) (,t ) + z_m(,t ) & = & z^(1)_m ( ) ( , t ) + z^(2)_m()(,t ) . note that @xmath71 and @xmath72 become equal as @xmath35 . it therefore follows that @xmath73 for @xmath74 . once these four partition functions are determined , the corresponding free - energy density is easily calculated . the contribution from the torus to the free - energy density is given by f_t(t ) = -t ^9 _ z_t(,t ) , [ un ] in complete analogy with eqs . ( [ reducedlambda ] ) and ( [ freeenergydef ] ) . by contrast , the remaining contributions to the free - energy density are each given by f_x(t ) = -t ^9 _ 0^ z_x(,t ) where x , c , m . [ deux ] the total free - energy density of the thermal string model is then given by f(t ) = f_t(t ) + f_k(t ) + f_c(t)+ f_m(t ) . [ trois ] thus , just as for closed strings , we see that the art of extending a given zero - temperature type i string theory to finite temperatures ultimately boils down to choosing the manner in which the zero - temperature partition functions @xmath59 and @xmath61 are to be decomposed into the separate thermal contributions @xmath75 and @xmath76 respectively . once these choices are made , the rest follows uniquely : modular invariance dictates @xmath77 , and orientifold projections determine @xmath78 and @xmath79 . moreover , just as for closed strings , it turns out that the traditional boltzmann sum is reproduced in the finite - temperature theory by making the particular choices for @xmath75 and @xmath76 such that the spacetime bosonic ( fermionic ) states within @xmath75 and @xmath76 are associated with @xmath80 ( @xmath48 ) and @xmath81 ( @xmath82 ) respectively . to illustrate this procedure , let us consider the single self - consistent zero - temperature ten - dimensional type i string model which is both supersymmetric and anomaly - free : this is the @xmath0 type i string @xcite . note that this string can be realized as the orientifold projection of the ten - dimensional zero - temperature type iib superstring , whose partition function is given by z_iib = z_boson^(8 ) ( _ v-_s)(_v-_s ) . [ typeiibpartition ] here @xmath50 denotes the contribution from the eight worldsheet coordinate bosons , just as for the heterotic strings discussed above , and the contributions from the left - moving ( right - moving ) worldsheet fermions are written in terms of the holomorphic ( anti - holomorphic ) characters @xmath83 ( @xmath84 ) of the transverse @xmath52 lorentz group . implementing the orientifold projection is relatively straightforward , and leads to the type i contributions : & z_t ( ) & = z_boson^(8 ) ( _ v -_s)(_v-_s ) + klein : & z_k ( ) & = z_boson^(8 ) ( _ v -_s ) + cylinder : & z_c ( ) & = n^2 z_boson^(8 ) ( _ v -_s ) + mobius : & z_m ( ) & = --0.04 truein n_boson^(8 ) ( _ v -_s ) , [ typeifourcontribs ] where we have used the notation and conventions defined in the appendix . tadpole anomaly cancellation ultimately requires that we take @xmath85 , thereby leading to the @xmath0 gauge group . note that while the cylinder contribution scales as @xmath86 [ representing the sum of the dimensionalities of the symmetric and anti - symmetric tensor representations of @xmath0 ] , the contribution scales only as @xmath88 ( representing their difference ) . given the results for the zero - temperature @xmath0 type i theory in eq . ( [ typeifourcontribs ] ) , it is straightforward to construct their finite - temperature extension . within the torus contribution in eq . ( [ typeifourcontribs ] ) , we recognize that the states which are spacetime bosons are those which contribute to @xmath89 , while those that are spacetime fermions contribute to @xmath90 . following the standard boltzmann description , we thus identify z_t^(1 ) & = & z_boson^(8 ) ( _ v_v+_s_s ) + z_t^(2 ) & = & - z_boson^(8 ) ( _ v_s+_s_v ) . similar reasoning for the cylinder contribution in eq . ( [ typeifourcontribs ] ) also leads us to identify z_c^(1 ) & = & n^2 z_boson^(8 ) _ v + z_c^(2 ) & = & - n^2 z_boson^(8 ) _ s . given these choices , the remaining terms in the total thermal partition function are determined through modular transformations and orientifold projections , leading to the final finite - temperature result : & z_t(,t ) & = z_boson^(8 ) _ v_v + _ s_s _ 0 + & & - _ v_s + _ s_v _ 1/2 + & & + _ i_i + _ c_c _ 0 + & & - _ i_c + _ c_i _ 1/2 + klein : & z_k(,t ) & = z_boson^(8 ) ( _ v -_s ) + cylinder : & z_c(,t ) & = n^2 z_boson^(8 ) ( _ v -_s ) + mobius : & z_m(,t ) & = --0.02 truein n _ boson^(8 ) ( _ v -_s ) . [ typeithermal ] this , too , is the standard result in the string - thermodynamics literature . as we have seen in the previous section , there is a simple procedure by which a given zero - temperature string model can be extended to finite temperatures . indeed , because of constraints coming from modular invariance and/or orientifold projections , we have relatively little choice in how this is done . for closed strings , the only freedom we have is related to how our ( torus ) partition function @xmath13 is decomposed into @xmath30 and @xmath41 , into the separate contributions that determine which of the zero - temperature states in the theory are to receive integer modings @xmath21 around the thermal circle , and which states are to receive half - integer modings @xmath49 . likewise , for type i strings , we have an additional freedom which concerns how the same choice is ultimately made for the open - string sectors . however , once those choices are made , all of the resulting thermal properties of the theory are completely determined . as discussed in the previous section , the standard prescription is to identify those states which are spacetime bosons with integer modings @xmath32 around the thermal circle , and those which are spacetime fermions with half - integer modings @xmath22 . indeed , this is ultimately the unique choice for which the resulting string partition functions correspond to the standard field - theoretic boltzmann sums for each string state ( a fact which is most directly evident after certain poisson resummations are performed , essentially transforming our theory from the so - called @xmath91-representation we are using here to the so - called @xmath92-representation in which the modular invariance of the torus contributions is not manifest ) . given this observation , it might seem that there is therefore no choice in how our zero - temperature string theories are extended to finite temperatures . however , this is not entirely correct . it is certainly true that the temperature / radius correspondence instructs us to treat bosonic states as periodic around the thermal circle and fermionic states as anti - periodic . however , this does not necessarily imply that all bosonic states will correspond to integer momentum modings @xmath21 , or that all fermionic states will correspond to half - integer momentum modings @xmath22 . indeed , in the presence of a non - trivial wilson line , this result can change . in order to understand how this can happen , let us first recall how the standard `` temperature / radius correspondence '' is derived ( see , , ref . @xcite ) . as is well known , this correspondence is most directly formulated in quantum field theory ( as opposed to string theory ) and ultimately rests upon the algebraic similarity between the free - energy density of a thermal theory and the vacuum - energy density of the zero - temperature theory in which the euclidean timelike direction is geometrically compactified on a circle . this similarity can be demonstrated as follows . let us begin on the thermal side , and consider the thermal ( grand - canonical ) partition functions corresponding to a single real @xmath14-dimensional bosonic field and a single @xmath14-dimensional fermionic field of mass @xmath27 : z_b , f(t ) = _ ( 1 e^-e_p / t)^1 [ pfs3 ] with @xmath93 . in eq . ( [ pfs3 ] ) , the products are over all @xmath44-dimensional spatial momenta @xmath94 . given these thermal partition functions , the corresponding @xmath14-dimensional free - energy densities are given by f_b , f(t ) -t z_b , f(t ) = t ( 1e^-e_/t ) . however , thanks to certain infinite - product representations for the hyperbolic trigonometric functions , it is an algebraic identity that ( 1e^-e / t ) = _ n = -^ log + ... [ iden ] where @xmath95 and @xmath96 . in writing eq . ( [ iden ] ) , we have followed standard practice and dropped terms beyond the infinite products as well as terms which compensate for the dimensionalities of the arguments of the logarithms . we therefore find that f_b , f(t ) = _ n = -^ log + ... [ freees ] on the zero - temperature side , by contrast , we can consider the zero - point one - loop vacuum - energy density corresponding to a single real quantum field of mass @xmath27 in @xmath14 uncompactified dimensions : ( -1)^f ( p^2+m^2 ) . [ lambdadef ] here @xmath6 indicates the spacetime statistics of the quantum field ( @xmath97 for a bosonic field , @xmath98 for a fermionic field ) . moreover , if we imagine that the time dimension is compactified on a circle of radius @xmath99 ( so that the integral over @xmath100 can be replaced by a discrete sum ) , and if the quantum field in question is taken to be periodic ( p ) or anti - periodic ( a ) around this compactification circle , then eq . ( [ lambdadef ] ) takes the form _ p , a = 12r ( -1)^f _ n=-^ [ lambdanext ] where @xmath101 , @xmath102 . given these results , it is now possible to make the `` temperature / radius correspondence '' : comparing eq . ( [ freees ] ) with eq . ( [ lambdanext ] ) , we see that we can identify the free - energy density @xmath16 of a boson ( fermion ) in @xmath14 spacetime dimensions at temperature @xmath2 with the zero - temperature vacuum - energy density @xmath103 of a boson ( fermion ) in @xmath14 spacetime dimensions , where a ( euclidean ) timelike dimension is compactified on a circle of radius @xmath18 about which the boson ( fermion ) is taken to be periodic ( anti - periodic ) . given this derivation of the temperature / radius correspondence , it may at first glance seem that the identification of bosons and fermions with integer and half - integer modings around the thermal circle is sacrosanct . however , let us consider what happens when we repeat this derivation in the presence of a non - trivial gauge field @xmath104 on the geometric ( zero - temperature ) side . when we calculate the vacuum energies of our bosonic or fermionic quantum field in the presence of a non - trivial gauge field @xmath104 , we must use the kinematic momenta @xmath105 where @xmath106 is the charge ( expressed as a vector in root space ) of the field in question . of course , if the field @xmath104 is pure - gauge ( , with vanishing corresponding field strength ) and our spacetime geometry is trivial , then this change in momenta from @xmath107 to @xmath108 will have no physical effect . _ however , if we are compactifying on a circle , there is always the possibility that our compactification encloses a gauge - field flux . _ as in the aharonov - bohm effect , this then has the potential to introduce a non - trivial change in modings for fields around this circle , even if the gauge field @xmath104 is pure - gauge at all points along the compactification circle . indeed , such a flat ( pure - gauge ) background for the gauge field @xmath104 is nothing but a wilson line . to be specific , let us first consider the situation in which our compactification circle of radius @xmath99 completely encloses a @xmath109 magnetic flux of magnitude @xmath110 which is entirely contained within a radius @xmath111 . at all points along the compactification circle , this then corresponds to a @xmath109 gauge field @xmath104 whose only non - zero component is the component @xmath112 along the compactified dimension . because of the non - trivial topology of the circle , we then find that the shift from @xmath107 to @xmath108 for a state with @xmath109 charge @xmath113 induces a corresponding shift in the corresponding modings : nr + 12r . while this result holds for @xmath109 gauge fields , it is easy to generalize this to the gauge fields of any gauge group @xmath114 . for any gauge group @xmath114 , we can describe a corresponding gauge flux in terms of the parameters @xmath115 for each @xmath116 , where @xmath117 is the rank of @xmath114 . collectively , we can write @xmath118 as a vector in root space . likewise , the gauge charge of any given state can be described in terms of its cartan components @xmath119 for @xmath116 ; collectively , @xmath106 is nothing but the weight of the state in root space . we then find that the modings are shifted according to nr + 12r . as a result , complex fields which are chosen to be periodic ( p ) or anti - periodic ( a ) around the compactification circle will have vacuum energies given by _ p , a = 12r ( -1)^f _ n=-^ [ lambdanext2 ] where @xmath120 . note that in each case , the underlying periodicity properties of the field are unaffected . rather , it is the _ manifestations of these periodicities in terms of the modings _ which are affected by the appearance of the wilson line . this , then , explains how a non - trivial wilson line can produce unexpected modings due to the non - trivial compactification geometry . however , we still wish to understand the appearance of such a wilson line thermally . what is the thermal analogue of the non - trivial wilson line ? or , phrased somewhat differently , what effect on the thermal side can restore the temperature / radius correspondence if a non - trivial wilson line has been introduced on the geometric side ? it turns out that introducing a non - trivial wilson line on the geometric side corresponds to introducing a non - zero chemical potential on the thermal side . in fact , this chemical potential will be imaginary . to see this , let us reconsider the partition functions of complex bosons and fermions in the presence of a non - zero chemical potential @xmath121 where @xmath122 . in general , a complex bosonic field will have a grand - canonical partition function given by z_b(t ) = _ * p * [ parfb ] where the two factors in eq . ( [ parfb ] ) correspond to particle and anti - particle excitations respectively . the corresponding free energy @xmath123 then takes the form f_b(t ) & = & t + + & = & t2 _ n=-^ + + & = & t2 _ n=-^ + & = & t2 _ n=-^ + & = & t2 _ n=-^ + & = & t2 _ n=-^ + + & = & t _ [ manysteps ] in eq . ( [ manysteps ] ) , the second equality follows from the algebraic identities in eq . ( [ iden ] ) while the final equality results upon exchanging @xmath124 in the second term . thus , comparing the result in eq . ( [ manysteps ] ) with the result in eq . ( [ lambdanext2 ] ) , we see that the free energy of a bosonic field at temperature @xmath2 is equal to the vacuum energy of a periodically - moded field on a circle of radius @xmath99 , where @xmath18 and where = ( ) t = i ( ) t . [ chempotential ] a similar result holds for complex fermions and anti - periodic fields , with the same chemical potential . we thus conclude that the introduction of a non - trivial wilson line on the geometric side corresponds to the introduction of an imaginary , temperature - dependent chemical potential on the thermal side . this result is well known in field theory @xcite , and has also recently been discussed in a string - theory context @xcite . before concluding , we should remark that the above discussion has been somewhat field - theoretic . indeed , if we define the wilson - line parameter 2 = -a2 t , then our primary result is that a non - trivial wilson line @xmath125 induces a shift in the momentum quantum number of the form m m + [ momentumshift ] for a state carrying charge @xmath106 with respect to the gauge field constituting the wilson line . this result is certainly true in quantum field theory , and also holds by extension for open - string states . however , closed - string states can carry not only momentum quantum numbers @xmath27 but also winding numbers @xmath26 which parametrize their windings around the thermal circle . this is important , because in the presence of a non - zero winding mode @xmath26 , a non - trivial wilson line shifts not only the momentum @xmath27 but also the charge vector @xmath106 of a given state , so that eq . ( [ momentumshift ] ) is generalized to @xcite [ shiftnumbershet ] it is clear that eq . ( [ shiftnumbershet ] ) reduces to eq . ( [ momentumshift ] ) for @xmath126 . we have already seen in sect . ii that the manner in which a zero - temperature string theory is extended to finite temperature depends on the choice as to which zero - temperature states are to be associated with integer momenta @xmath32 around the thermal circle , and which are to be associated with half - integer momenta @xmath49 . once this decision is made , the thermal properties of the resulting theory are completely fixed . moreover , we have seen in sect . iii that there is considerable freedom in making this choice , depending on whether ( and which ) wilson lines might be present . indeed , in principle , each choice of wilson line leads to an entirely different thermal theory . while all of these thermal theories necessarily reduce back to the starting zero - temperature theory as @xmath35 , they each represent different possible finite - temperature extensions of that theory which correspond to different possible chemical potentials which might be introduced into their corresponding boltzmann sums . indeed , viewed from this perspective , we see that the traditional boltzmann choices merely correspond to one special case : that without a wilson line , for which the corresponding chemical potential vanishes . as an example , let us consider the supersymmetric @xmath0 heterotic string . at zero temperature , the partition function of this theory is z_model = z^(8)_boson ( _ v-_s ) ( _ i^2 + _ v^2 + _ s^2 + _ c^2 ) , [ start ] and without a wilson line we would normally decompose this into the separate thermal contributions @xmath30 and @xmath41 by making the associations z^(1 ) & = & z^(8)_boson _ v ( _ i^2 + _ v^2 + _ s^2 + _ c^2 ) + z^(2 ) & = & -z^(8)_boson _ s ( _ i^2 + _ v^2 + _ s^2 + _ c^2 ) . [ nowilson ] indeed , this is precisely the decomposition discussed in sect . ii , which leads to the standard boltzmann sum . however , there are in principle other ways in which the zero - temperature partition function in eq . ( [ start ] ) might be meaningfully decomposed . for example , let us consider an alternate decomposition of the form z^(1 ) & = & z^(8)_boson _ v ( _ i^2 + _ v^2 ) - _ s ( _ s^2 + _ c^2 ) + z^(2 ) & = & z^(8)_boson _ v ( _ s^2+_c^2 ) - _ s ( _ i^2+_v^2 ) . [ alternate ] unlike the standard boltzmann decomposition , this alternate decomposition treats spacetime bosonic and fermionic states in ways which are also dependent on their corresponding gauge quantum numbers . specifically , while vectorial representations of the @xmath0 gauge group are treated as expected , with spacetime bosons having integer momentum modings @xmath32 and spacetime fermions having half - integer momentum modings @xmath49 , the spinorial representations of the left - moving @xmath0 gauge group have the opposite behavior , with spacetime bosons associated with modings @xmath49 and spacetime fermions associated with modings @xmath32 . [ note , in this connection , that the @xmath127 character combination @xmath128 is nothing but the @xmath0 character @xmath129 , and likewise @xmath130 is nothing but @xmath131 . ] however , as we have seen in sect . iii , such `` wrong '' modings can be easily understood as the effects of a non - trivial wilson line . indeed , looking at eq . ( [ shiftnumbershet ] ) , we see that the results in eq . ( [ alternate ] ) are obtained directly if our wilson line @xmath125 is chosen such that @xmath132 ( mod @xmath133 ) for states in spinorial representations of @xmath0 , while @xmath134 ( mod @xmath133 ) for states in vectorial representations of @xmath0 . given that @xmath135 for vectorial representations of @xmath0 and @xmath136 for spinorial representations of @xmath0 , we see that a simple choice such as @xmath137 can easily accomplish this . however , at this stage , we have no knowledge as to whether or not such a wilson line represents a legitimate choice for the @xmath0 heterotic string . for example , we have no idea whether such a wilson - line choice is compatible with a worldsheet interpretation in which the possible choices of wilson lines are tightly constrained by numerous string self - consistency constraints . moreover , along the same lines , we do not know what other wilson lines might also be available . in order to explore all of the potential possibilities , we shall therefore proceed to survey the set of all possible wilson lines which might be self - consistently introduced when attempting to extend a given zero - temperature string theory to finite temperatures . as we shall see , however , the situation is somewhat different for closed strings and type ii strings . we shall therefore consider these two cases separately . in general , there are two classes of closed strings which are supersymmetric and hence perturbatively stable : type ii superstrings and heterotic strings . in ten dimensions , however , the type ii superstrings lack gauge symmetries ; thus no possible wilson lines can exist in their extensions to finite temperatures . for this reason , when discussing closed strings , we shall concentrate on the ten - dimensional supersymmetric @xmath0 and @xmath1 heterotic strings . note , however , that in lower dimensions , all of the closed strings will accrue additional gauge symmetries as a result of compactification indeed , this holds for type ii strings as well as heterotic . thus , in lower dimensions , the sets of allowed wilson lines in each case are likely to be much more complex than we are considering here . in general , the temperature / radius correspondence provides us with a powerful tool to help determine the allowed wilson lines that may be introduced when forming our thermal theory : we simply replace the temperature @xmath2 with @xmath138 and consider the corresponding problem of introducing a wilson line into the geometric compactification of our original zero - temperature theory . for example , if we are seeking the set of allowed wilson lines that can be introduced into the construction of the finite - temperature ten - dimensional @xmath0 heterotic theory , we can instead investigate the allowed wilson lines that may be introduced upon compactifying the zero - temperature @xmath0 theory to nine dimensions . in principle , the latter problem can be studied through any number of formalisms having to do with the construction of self - consistent zero - temperature string models such model - building formalisms are numerous and include various orbifold constructions , narain lattice constructions , constructions based on free worldsheet bosons and fermions , and so forth . however , for closed strings , it turns out that t - duality leads to a significant simplification : while the @xmath139 ( or @xmath35 ) limit reproduces our original string model in the original @xmath14 spacetime dimensions , and while taking @xmath140 leads to a string model in @xmath141 spacetime dimensions , the formal @xmath142 ( or @xmath143 ) limit actually yields a new string theory which is back in @xmath14 spacetime dimensions ! moreover , the structure of the finite - temperature string partition function in eq . ( [ eomix ] ) guarantees that this new @xmath14-dimensional theory is nothing but a @xmath5 orbifold of our original @xmath14-dimensional theory ; indeed , while the original theory in the @xmath144 limit has the partition function @xmath145 , the final theory in the @xmath142 limit has the partition function @xmath146 . in some sense , the thermal theory in @xmath44 dimensions _ interpolates _ between the original @xmath14-dimensional theory at @xmath147 and a different @xmath14-dimensional theory as @xmath143 , these two @xmath14-dimensional theories being @xmath5 orbifolds of each other . thus , the allowed wilson lines that may be introduced into the finite - temperature extension of a given zero - temperature closed string theory are in one - to - one correspondence dimensions , one may construct the corresponding thermal theory through a specific sequence of steps : first , one compactifies the zero - temperature theory on a circle of radius @xmath148 , and then one orbifolds the resulting theory by the @xmath5 action @xmath149 where @xmath7 is the spacetime fermion number , where @xmath150 denotes a half - shift around the thermal circle , and where @xmath9 ( the wilson line ) indicates an additional specific orbifold action which is sensitive to the gauge quantum numbers of each state . the resulting @xmath44-dimensional thermal theory then has the property that the original @xmath14-dimensional theory is reproduced as @xmath35 , and that a new @xmath14-dimensional theory emerges in the formal @xmath143 limit . moreover , it can also be shown that the new theory which emerges in the @xmath143 limit is a @xmath5 orbifold of the original theory , where the @xmath5 orbifold in this case is nothing but @xmath8 . thus , for each wilson line @xmath9 which is involved in construction of the thermal theory in @xmath44-dimensions , there is a corresponding orbifold @xmath8 which directly relates the two `` endpoint '' @xmath14-dimensional theories to each other . ] with the set of allowed @xmath5 orbifolds of that theory , the set of @xmath5 orbifolds which reproduce another self - consistent string theory in @xmath14 dimensions . this correspondence provides us with exactly the tool we need , because the complete set of self - consistent heterotic string theories in ten dimensions is known . indeed , these have been classified in ref . @xcite , and it turns out that in addition to the supersymmetric @xmath0 and @xmath1 heterotic theories , there are only seven additional heterotic theories in ten dimensions . these are the tachyon - free @xmath127 string model @xcite as well as six tachyonic string models with gauge groups @xmath0 , @xmath151 , @xmath152 , @xmath153 , @xmath154 , and @xmath155 . however , not all of these models can be realized as @xmath5 orbifolds of the original supersymmetric @xmath0 or @xmath1 models . indeed , of the seven non - supersymmetric models listed above , only four are @xmath5 orbifolds of the supersymmetric @xmath0 string ; likewise , only four are @xmath5 orbifolds of the @xmath1 string . these @xmath5 orbifold relations are shown in fig . [ orbifoldfig ] . it is important to note that there also exists a non - trivial @xmath5 orbifold relation which directly relates the supersymmetric @xmath0 and @xmath1 strings to each other . however , it is easy to see that this orbifold must be excluded from consideration . on the thermal side , we know that finite - temperature effects necessarily treat bosons and fermions differently and will therefore necessarily break whatever spacetime supersymmetry might have existed at zero temperature . this implies that we must restrict our attention to those @xmath5 orbifolds which project out whatever gravitino might have existed in our original @xmath14-dimensional model . the @xmath5 orbifold relating the supersymmetric @xmath0 and @xmath1 strings to each other does not have this property . likewise , there also exists a non - trivial @xmath5 orbifold [ specifically @xmath6 ] which maps the supersymmetric @xmath0 and @xmath1 heterotic strings to chirality - flipped versions of _ themselves_. this somewhat degenerate orbifold actually corresponds to the situation _ without _ a wilson line , and has thus already been implicitly considered in eq . ( [ fake ] ) . 6.5 truein given these results , we see that there are only four non - trivial candidate wilson - line choices for the finite - temperature @xmath0 heterotic string . likewise , there are only four candidate wilson - line choices for the finite - temperature @xmath1 heterotic string . for each of these wilson - line choices , we can then construct the corresponding finite - temperature theory . it is straightforward to write down the partition functions of these finite - temperature theories , some of which have already appeared in various guises in previous work ( see , , refs . @xcite ) . in each case , we shall follow the exact notations and conventions established in the appendix . however , for convenience , we shall also establish one further convention . although the anti - holomorphic ( right - moving ) parts of these partition functions will always be expressed in terms of the ( barred ) characters @xmath157 of the transverse @xmath52 lorentz group , it turns out that we can express the holomorphic ( left - moving ) parts of each of these partition functions in terms of the ( unbarred ) characters @xmath158 associated with the group @xmath127 . indeed , it turns out that such a rewriting is possible in each case regardless of the actual gauge group @xmath114 of the ten - dimensional model that is produced by the @xmath5 orbifold . of course , if @xmath127 is a subgroup of @xmath114 , then such a rewriting is meaningful and the characters which appear in the resulting partition function correspond to the actual gauge - group representations which appear in spectrum of the model . by contrast , if @xmath127 is not a subgroup of @xmath114 , then such a rewriting is merely an algebraic exercise ; the @xmath127 characters then have no meaning beyond their @xmath159-expansions , and can appear with non - integer coefficients . in all cases , however , these expressions represent the true partition functions of these thermal theories as far as their @xmath159-expansions are concerned . we shall therefore follow these conventions in what follows . let us begin by considering the zero - temperature supersymmetric @xmath0 heterotic string , which has the partition function given in eq . ( [ start ] ) . for this string , our four possible finite - temperature extensions are then as follows . in each case we shall label each of the possibilities according to the @xmath143 model produced by the corresponding @xmath5 orbifold . the partition function of the thermal model associated with the @xmath5 orbifold producing the non - supersymmetric @xmath0 heterotic string is given by z_so(32 ) = z^(8)_boson & & _ v ( _ i^2 + _ v^2 ) - _ s ( _ s^2 + _ c^2 ) _ 0 + & + & _ v ( _ s^2+_c^2 ) - _ s ( _ i^2+_v^2 ) _ 1/2 + & + & _ i ( _ i_v+_v_i ) - _ c ( _ s_c+_c_s ) _ 0 + & + & _ i ( _ s_c+_c_s ) - _ c ( _ i_v+_v_i ) _ 1/2 , [ a1 ] while the partition functions of the thermal models associated the @xmath5 orbifolds that produce the the @xmath151 , @xmath160 , and @xmath127 models are respectively given by z_so(8)so(24 ) = z^(8)_boson & & _ v ( _ i^2 + _ v^2 + _ s^2 ) - _ s ( _ s^2 + _ v^2 + _ c^2 ) _ 0 + & + & _ v ( _ s^2 + _ v^2 + _ c^2 ) - _ s ( _ i^2 + _ v^2 + _ s^2 ) _ 1/2 + & + & _ i ( _ i_v + _ v_c ) - _ c ( _ s_c + _ i_s ) _ 0 + & + & _ i ( _ s_c + _ i_s ) - _ c ( _ i_v + _ v_c ) _ 1/2 , [ a2 ] z_u(16 ) = z^(8)_boson & & _ v ( _ i^2 + _ v^2 + _ s^2 ) - _ s ( _ s^2 + _ v^2 + _ c^2 ) _ 0 + & + & _ v ( _ s^2 + _ v^2 + _ c^2 ) - _ s ( _ i^2 + _ v^2 + _ s^2 ) _ 1/2 + & + & _ i ( _ i_v + _ v_c ) - _ c ( _ s_c + _ i_s ) _ 0 + & + & _ i ( _ s_c + _ i_s ) - _ c ( _ i_v + _ v_c ) _ 1/2 , [ a3 ] and z_so(16)so(16 ) = z^(8)_boson & & _ v ( _ i^2 + _ s^2 ) - _ s ( _ v^2+_c^2 ) _ 0 + & + & _ v ( _ v^2 + _ c^2 ) - _ s ( _ i^2+_s^2 ) _ 1/2 + & + & _ i ( _ v_c+_c_v ) - _ c ( _ i_s+_s_i ) _ 0 + & + & _ i ( _ i_s+_s_i ) - _ c(_v_c+_c_v ) _ 1/2 . [ firstso16 ] note that as @xmath35 , each of these expressions reduces to the partition function of the zero - temperature supersymmetric @xmath0 heterotic string in eq . ( [ start ] ) , as required . as is easy to verify , these four different thermal extensions of the supersymmetric @xmath0 heterotic string correspond to the wilson lines so(32 ) : & = & ( ( 1)^n ( 0)^16-n ) for n2 + 1 + so(8)so(24 ) : & = & ( ( ) ^4 ( 0)^12 ) or ( ( ) ^4 ( 0)^12 ) + so(16)so(16 ) : & = & ( ( ) ^8 ( 0)^8 ) or ( ( ) ^8 ( 0)^8 ) + u(16 ) : & = & ( ( ) ^16 ) for , , , . [ hetcases ] indeed , because our original supersymmetric @xmath0 heterotic theory contains only vectorial and spinorial representations of @xmath0 , each of the individual components of the wilson line @xmath125 is defined only modulo 2 . a similar situation exists for the zero - temperature @xmath1 heterotic string , which has partition function z^(8)_boson ( _ v -_s ) ( _ i + _ s)^2 [ e8pf ] the partition function of the thermal extension of this model associated with the @xmath5 orbifold producing the non - supersymmetric @xmath153 model is given by z_so(16)e_8 = z^(8)_boson & & _ v_i - _ s_s _ 0 + & + & _ v_s - _ s_i _ 1/2 + & + & _ i_v - _ c_c _ 0 + & + & _ i_c - _ c_v _ 1/2 ( _ i+_s ) , [ b1 ] while the partition functions of the thermal models associated with the @xmath161 , @xmath155 , and @xmath127 orbifolds are respectively given by z_(e_7)^2su(2)^2 = z^(8)_boson & & _ v ( _ i^2 + _ i _ s + _ s^2 ) - _ s ( _ s^2 + _ i _ s ) _ 0 + & + & _ v ( _ s^2 + _ i _ s ) - _ s ( _ i^2 + _ i _ s + _ s^2 ) _ 1/2 + & + & _ i ( _ i_v + _ v_s ) - _ c ( _ s_s + _ i_s ) _ 0 + & + & _ i ( _ s_s + _ i_s ) - _ c ( _ i_v + _ v_s ) _ 1/2 , [ b2 ] z_e_8 = z^(8)_boson & & _ v ( _ i^2 + _ i _ s + _ s^2 ) - _ s ( _ s^2 + _ i _ s ) _ 0 + & + & _ v ( _ s^2 + _ i _ s ) - _ s ( _ i^2 + _ i _ s + _ s^2 ) _ 1/2 + & + & _ i ( _ i_v + _ v_s ) - _ c ( _ s_s + _ i_s ) _ 0 + & + & _ i ( _ s_s + _ i_s ) - _ c ( _ i_v + _ v_s ) _ 1/2 , [ b3 ] and z_so(16)so(16 ) = z^(8)_boson & & _ v ( _ i^2 + _ s^2 ) - _ s ( _ i_s + _ s_i ) _ 0 + & + & _ v ( _ i_s+_s_i ) - _ s ( _ i^2+_s^2 ) _ 1/2 + & + & _ i ( _ v_c+_c_v ) - _ c ( _ v^2 + _ c^2 ) _ 0 + & + & _ i ( _ v^2+_c^2 ) - _ c ( _ v_c+_c_v ) _ 1/2 . [ secondso16 ] once again , using the identities listed in the appendix , it is straightforward to verify that each of these expressions reduces to eq . ( [ e8pf ] ) as @xmath35 . moreover , the expressions in eqs . ( [ firstso16 ] ) and ( [ secondso16 ] ) are actually equal as the result of the further identity on @xmath56 characters given by _ i _ s + _ s _ i = _ v^2 + _ . [ identy ] this is ultimately the identity which is responsible for the fact that the two expressions within eq . ( [ pfs2 ] ) are equal at the level of their @xmath159-expansions , , that the ten - dimensional supersymmetric @xmath0 and @xmath1 heterotic strings have the same bosonic and fermionic state degeneracies at each mass level . 3.5 truein 3.5 truein as an aside , it is interesting to note that all of these thermal functions can be written in a common form parametrized by a single integer @xmath162 : z^(8)_boson & & _ v ( _ i^2 + _ v^2 + _ s^2 ) - _ s ( _ c^2 + _ s^2 + _ v^2 ) _ 0 + & + & _ v ( _ c^2 + _ s^2 + _ v^2 ) - _ s ( _ i^2 + _ v^2 + _ s^2 ) _ 1/2 + & + & _ i ( _ i_v + _ v_i + _ v_c + _ c_v ) + & & -_c ( _ s_c + _ c_s + _ v^2 + _ c^2 ) _ 0 + & + & _ i ( _ s_c + _ c_s + _ v^2 + _ c^2 ) + & & -_c ( _ i_v + _ v_i + _ v_c + _ c_v ) _ 1/2 . [ singleinterpolation ] in particular , the values @xmath163 correspond to the partition functions in eqs . ( [ a1 ] ) , ( [ b1 ] ) , ( [ a2 ] ) , ( [ b2 ] ) , ( [ a3 ] ) , ( [ b3 ] ) , and ( [ firstso16 ] ) [ or ( [ secondso16 ] ) ] respectively , where eq . ( [ identy ] ) has been used wherever needed . it is also instructive to examine the free - energy densities @xmath46 associated with each of these possible wilson - line choices . as we have seen , for each thermal partition function @xmath19 listed above , the corresponding free - energy density @xmath46 is given by eq . ( [ freeenergydef ] ) . following this definition , we then obtain the results shown in fig . [ ffig ] . we observe from fig . [ ffig ] that the non - trivial wilson line which minimizes the free - energy density in each case is the one which breaks the gauge group minimally . for the @xmath0 string , this is the wilson line associated with the non - supersymmetric @xmath0 orbifold , while for the @xmath164 heterotic string , this is the wilson line associated with the non - supersymmetric @xmath153 orbifold . it is tempting to say , therefore , that these particular non - trivial wilson lines are somehow `` preferred '' in some dynamical sense over the others . however , this assumption presupposes the existence of a mechanism by which these wilson lines can smoothly be deformed into each other with finite energy cost . given that these wilson lines ultimately correspond to fluxes which are not only constrained topologically but also presumably quantized , such wilson - line - changing transitions would require exotic physics ( such as might occur on a full thermal landscape ) . we shall discuss the structure of such a landscape in sect . we also note that for both of our supersymmetric heterotic strings , there remains the traditional option of constructing a thermal theory _ without _ a non - trivial wilson line . it turns out that the free energies corresponding to these choices are numerically almost identical ( but ultimately slightly smaller ) than those of the non - supersymmetric @xmath0 and @xmath153 cases plotted in fig . these features will be discussed further in sect . we see , then , that have been able to construct four new thermal theories for the supersymmetric @xmath0 heterotic string as well as four new thermal theories for the supersymmetric @xmath1 heterotic string . each of these theories has the novel feature that a non - trivial wilson line has been introduced when constructing the finite - temperature extension , or equivalently that a non - trivial temperature - dependent chemical potential has been introduced into the boltzmann sum . each of these theories reduces to the correct supersymmetric theory as @xmath35 , and moreover each is modular invariant for all temperatures @xmath2 . even more importantly , the temperature / radius correspondence guarantees that in each case , the temperature variable @xmath2 like the radius variable @xmath99 to which it corresponds is a bona - fide _ modulus _ of the theory , able to be freely changed without disturbing the worldsheet self - consistency of the string . despite the fact that the non - trivial wilson lines we have introduced in each case have led to certain unorthodox modings for our string states around the thermal circle , none of these theories violates any spin - statistics relations . indeed , the spin - statistics theorem relates the spacetime lorentz spin of a given quantum field to its thermal statistics , and the temperature / radius correspondence relates such thermal statistics to the periodicity of such a field around the thermal circle . indeed , it is only the relation between this periodicity and the resulting algebraic moding which is altered as a result of the non - trivial wilson line . we now turn our attention to the corresponding situation for type i strings . in ten dimensions , there is a single self - consistent type i string model which is both supersymmetric and anomaly - free : this is the @xmath0 type i string @xcite . our goal is therefore to survey the possible wilson lines which can be introduced when formulating its thermal extension . as discussed in sect . ii , the ten - dimensional zero - temperature @xmath0 type i string has a partition function given in eq . ( [ typeifourcontribs ] ) , and its extension to finite temperature _ without _ wilson lines is given in eq . ( [ typeithermal ] ) . however , just as for the heterotic strings , we expect that new thermal possibilities can be constructed when non - trivial wilson lines are introduced @xcite . in general , as discussed in sect . ii , there are two kinds of wilson lines which might be introduced for type i theories . first , there are wilson lines that might be introduced into the closed - string sectors of such theories , much along the lines we have already discussed for the heterotic strings . however , the closed - string sectors of type i strings are essentially type ii superstrings ( indeed , these are the strings from which the type i strings can be obtained by orientifolding ) , and in ten dimensions the perturbative states of such type ii strings do not carry gauge charges . thus , for the ten - dimensional type i string , it is not possible to introduce a non - trivial wilson line in the closed - string sector . this guarantees that the results for @xmath165 and @xmath166 given in eq . ( [ typeithermal ] ) will remain invariant regardless of what happens in the open - string sector . the question then boils down to determining the allowed wilson lines that might be introduced in the _ open - string _ sector of the ten - dimensional type i string . indeed , because the states contributing to the cylinder and partition functions carry @xmath0 gauge charges , their modings in eq . ( [ typeithermal ] ) are potentially affected by the presence of an @xmath0 wilson line . fortunately , thanks to the temperature / radius correspondence , this problem can be mapped to the purely geometric issue of determining the allowed wilson lines that can be introduced when compactifying the type i string to nine dimensions on a circle indeed , the fact that we continually refer to `` wilson lines '' and `` thermal circles '' already implicitly presupposes that this can be done ! it turns out that the allowed wilson lines fall into two distinct classes . the first class consists of wilson lines of the form = ( , , , ... , 0,0,0 , ... ) [ wilsonclass1 ] where the number of non - zero components is given by @xmath26 , with @xmath167 . the @xmath126 special case corresponds to the case without a wilson line , and in general we shall define @xmath168 and @xmath169 . for wilson lines of this form , the contribution in eq . ( [ typeithermal ] ) turns out to be independent of @xmath26 , and thus remains the same as in eq . ( [ typeithermal ] ) for all @xmath170 : : z_m(,t ) = - _ open^(8 ) ( n_1+n_2 ) ( _ v -_s ) . however , we find that the cylinder contribution in eq . ( [ typeithermal ] ) now takes the form : z_c(,t ) = z_open^(8 ) & & ( n_1 ^ 2 + n_2 ^ 2 ) _ v - 2 n_1 n_2 _ s + & - & ( n_1 ^ 2 + n_2 ^ 2 ) _ s - 2 n_1 n_2 _ v . [ newcylinder ] it is easy to demonstrate that as a result of the shift induced by this wilson line , the gauge group of the resulting model is broken to @xmath171 . [ in the t - dual picture , the choice of the wilson line in eq . ( [ wilsonclass1 ] ) indicates that we have simply moved @xmath172 of the original 32 d8-branes in this theory to the opposite side of the thermal circle . ] note , however , that the appearance of this wilson line has also induced states with the `` wrong '' thermal modings to appear in eq . ( [ newcylinder ] ) . specifically , we see from eq . ( [ newcylinder ] ) that we now have spacetime spinors accruing integer thermal momentum modes within @xmath81 , while we also have spacetime vectors accruing half - integer thermal modes within @xmath82 . the second `` class '' of wilson lines we shall consider consists of a single wilson line of the form = ( , , , ... , ) . [ wilsonclass2 ] for this wilson line , the cylinder and partition functions in eq . ( [ typeithermal ] ) now take the form : z_c(,t ) & = & z_open^(8 ) 2 n _ v - ( n^2 + ^2 ) _ s + & & - 2 n _ s - ( n^2 + ^2 ) _ v + mobius : z_m(,t ) & = & - z_open^(8 ) ( n+ ) ( -_s + _ v ) [ newcylmob ] where @xmath173 . in this case , the wilson line has deformed the gauge group of our original @xmath0 theory to @xmath152 . note that the alternate wilson line @xmath174 produces the same theory . in either case , however , we once again observe that the wilson line has induced states to appear in eq . ( [ newcylmob ] ) with the `` wrong '' thermal modings . it should be stressed that when discussing the possible `` allowed '' wilson lines , we are not enforcing the possible open - string ns - ns tadpole - anomaly constraints for all temperatures ( as might normally be done within a more general type i model - building framework ) . indeed , only the @xmath127 and @xmath152 cases outlined above satisfy these constraints and completely avoid ns - ns tadpole divergences at all temperatures ; in all other cases , these constraints are satisfied only for temperatures below the hagedorn temperature . however , this approach is justified in this context because we are not seeking to avoid the possible emergence of open - string tachyons . in fact , such tachyons and the divergences they induce are both desired and expected , since these are precisely the features which ultimately trigger the hagedorn transition for type i strings . it should also be stressed that there are many different ways of obtaining the models discussed in this section . while one approach involves compactifying the ten - dimensional supersymmetric type i string on the thermal circle in the presence of various wilson lines , it is also possible to compactify the type ii string directly on the thermal circle , implementing the orientifold projection only after this compactification is performed ( see , , ref . the different allowed choices for open - string sectors in this orientifold projection then yield the models we have constructed here . regardless of the approach taken , however , we see that there are only a finite set of self - consistent possibilities which are available as potential finite - temperature extensions of the zero - temperature type i string . 3.5 truein 3.5 truein given the set of wilson lines outlined above , we can now examine their corresponding free energies @xmath46 . as discussed in sect . ii , for type i string models the corresponding free - energy density receives separate contributions from the torus , the klein bottle , the cylinder , and the amplitudes ; these are shown in eqs . ( [ un ] ) , ( [ deux ] ) , and ( [ trois ] ) . in particular , several particular wilson lines will interest us , such as those which yield the gauge groups we have considered for heterotic strings : so(32 ) : & = & ( ( 0)^16 ) + so(8)so(24 ) : & = & ( ( ) ^4 ( 0)^12 ) + so(16)so(16 ) : & = & ( ( ) ^8 ( 0)^8 ) + u(16 ) : & = & ( ( ) ^16 ) or ( ( ) ^16 ) . [ type1cases ] the results are shown in fig . [ type1plot ] . note that several of these results have also appeared in a different context in ref . @xcite . it is straightforward to understand the general features shown in fig . [ type1plot ] . first , we recall from the above discussion that all four of these possible finite - temperature extensions share the same torus and klein - bottle contributions to the free - energy density : as shown in fig . [ type1plot](a ) , the torus contribution is relatively small and negative for @xmath175 , remaining finite until it diverges discontinuously at the critical temperature @xmath176 , while the klein - bottle contribution actually vanishes as a result of the identity @xmath177 which holds for the characters of the transverse @xmath52 lorentz group . thus , as expected , it is the cylinder and contributions which are responsible for the relative differences between these different wilson - line choices . let us first consider the wilson - line cases following from the choice in eq . ( [ wilsonclass1 ] ) . as indicated above , these cases all share the same contribution as well . like the torus contribution , the contribution is also relatively small for @xmath175 ; however , unlike the torus contribution , we see from fig . [ type1plot](a ) that it is positive rather than negative and does not diverge , even at @xmath176 . indeed , if these were the only three contributions in the @xmath178 wilson - line cases , we would obtain the total curve which is shown in fig . [ type1plot](b ) for @xmath127 . in this case , the discontinuous divergence at @xmath179 is solely due to the _ closed - string _ tachyon coming from the torus amplitude in fig . [ type1plot](a ) . however , for all other cases with @xmath180 , we also have a relatively huge cylinder contribution which is negative for @xmath175 , with @xmath181 smoothly as @xmath182 . in fact , recalling the @xmath52 character identity @xmath177 , we see from eq . ( [ newcylinder ] ) that the overall general magnitude of this contribution is proportional to n_1 ^ 2 + n_2 ^ 2 - 2 n_1 n_2 = ( n_1-n_2)^2 = ( 32-n)^2 . thus the @xmath126 case ( , the case with vanishing wilson line ) has the most negative cylinder contribution and correspondingly the most negative total free - energy density from amongst the choices with gauge groups @xmath171 . of course , it still remains possible that the @xmath152 case in eq . ( [ wilsonclass2 ] ) might yield a free - energy density which is even more negative . however , since @xmath183 when @xmath184 , we see that the cylinder contribution in eq . ( [ newcylmob ] ) actually vanishes for this case . the contribution remains small but switches sign , becoming negative , but it continues to remain finite , even at @xmath179 . [ indeed , as discussed above , only in the @xmath152 and @xmath127 cases are all open - string tachyons avoided . ] as a result , the free - energy density in the @xmath152 case remains small , diverging discontinuously at @xmath179 only because of the _ closed - string _ tachyon in the torus amplitude . comparing fig . [ type1plot ] for the type i string with the analogous plot for the @xmath0 heterotic string in fig . [ ffig ] , we see certain superficial similarities . for example , the set of permitted gauge groups is similar in each case , and moreover wilson - line choice in each case that leads to the minimum free energy is the one that breaks the gauge group minimally . however , we stress that despite such superficial similarities , there remains one fundamental distinction between the heterotic and type i cases : except for the cases involving the particular @xmath127 and @xmath152 wilson lines , the type i cases lead to free - energy densities which actually diverge smoothly as the hagedorn temperature @xmath185 is approached , , @xmath186 as @xmath187 , while the corresponding heterotic free - energy densities actually remain finite as @xmath187 . this feature is already well known in the string - thermodynamics literature : it is a direct result of the open - string tachyon at @xmath185 , and suggests that the hagedorn temperature is actually a limiting temperature for type i strings rather than the location of a phase transition . as we have seen , one of the most prominent aspects of thermal string theories is the existence of a hagedorn transition at which the string free - energy density diverges . however , the introduction of a non - trivial wilson line can actually change the temperature at which this transition takes place . this is particularly true for heterotic strings , and we have already seen evidence of this fact in fig . [ ffig ] : the free - energy densities corresponding to the different possible wilson - line choices all diverge at critical temperatures which differ from the hagedorn temperature @xmath188 normally associated with the heterotic string without wilson lines . in some sense , it is to be expected that the introduction of non - trivial wilson lines can affect the resulting hagedorn temperature , since we have seen that such wilson lines affect the thermal partition function as a whole . however , we can equivalently associate the hagedorn temperature with the asymptotic densities of bosonic and fermionic states in the original zero - temperature theory , and the zero - temperature theory is clearly independent of the introduction of a non - trivial thermal wilson line . our goal in this section is to explain these different perspectives , and to show how they can ultimately be reconciled with each other in the presence of a non - trivial wilson line . for concreteness , we shall focus on the case of the @xmath0 and @xmath1 heterotic strings , and examine the consequences of moving from the standard thermal theories in eq . ( [ fake ] ) which do not involve non - trivial wilson lines to the new thermal theories [ such as those in eqs . ( [ a1 ] ) through ( [ firstso16 ] ) for @xmath0 , and those in eqs . ( [ b1 ] ) through ( [ secondso16 ] ) for @xmath1 ] which do . we begin with several preliminary remarks concerning the hagedorn transition and its dual uv / ir nature . the hagedorn transition is one of the central hallmarks of string thermodynamics . originally encountered in the 1960 s through studies of hadronic resonances and the so - called `` statistical bootstrap '' @xcite , the hagedorn transition is now understood to be a generic feature of any theory exhibiting a density of states which rises exponentially as a function of mass . in string theory , the number of states of a given total mass depends on the number of ways in which that mass can be partitioned amongst individual quantized mode contributions , leading to an exponentially rising density of states @xcite . thus , string theories should exhibit a hagedorn transition @xcite . originally , it was assumed that the hagedorn temperature is a limiting temperature at which the internal energy of the system diverges . however , later studies demonstrated that for closed strings the internal energy actually remains finite at this temperature . this then suggests that the hagedorn temperature is merely the critical temperature corresponding to a first- or second - order phase transition . there has been much speculation concerning possible interpretations of this phase transition , including a breakdown of the string worldsheet into vortices @xcite or a transition to a single long - string phase @xcite . it has also been speculated that there is a dramatic loss of degrees of freedom at high temperatures @xcite . over the past two decades , studies of the hagedorn transition have reached across the entire spectrum of modern string - theory research , including open strings and d - branes , strings with non - trivial spacetime geometries ( including ads backgrounds and @xmath190-waves ) , strings in magnetic fields , @xmath191 strings , tensionless strings , non - critical strings , two - dimensional strings , little strings , matrix models , non - commutative theories , as well as possible cosmological implications and implications for the brane world . a brief selection of papers in many of these areas appears in refs . @xcite . in general , determining the hagedorn temperature associated with a given finite - temperature thermal partition function is relatively straightforward . given this thermal partition function , the one - loop free - energy density @xmath46 is given by the modular integral in eq . ( [ freeenergydef ] ) , whereupon the full panoply of thermodynamic quantities such as the internal energy @xmath192 , entropy @xmath193 , and specific heat @xmath194 then follow from the standard definitions @xmath195 , @xmath196 , and @xmath197 . in string theory , the hagedorn transition is usually associated with a divergence or other discontinuity in the free energy @xmath46 as a function of temperature . it turns out that are only two ways in which such a divergence may arise within the expression in eq . ( [ freeenergydef ] ) . first , of course , is the possibility of a divergence or discontinuity due to the well - known exponential rise in the degeneracy of string states which contribute to @xmath10 . this may be considered an ultraviolet ( uv ) divergence because it is triggered by the behavior of the extremely massive portion of the string spectrum . however , it turns out that this rise in the state degeneracies ultimately does not cause @xmath46 to diverge . to understand why , we may expand @xmath19 in the form @xmath198 where @xmath199 , where @xmath200 describe the right- and left - moving worldsheet energies ( with thermal contributions included ) , and where @xmath201 describe the corresponding degeneracies of bosonic minus fermionic states . although the degeneracies @xmath201 indeed experience exponential growth of the generic form @xmath202 where @xmath203 are positive coefficients , the contribution of each such state to the modular integrand in eq . ( [ freeenergydef ] ) is suppressed according to @xmath204 $ ] . for all @xmath205 and sufficiently large @xmath200 , this exponential suppression easily overwhelms the exponential rise in the degeneracy of states . as a result , the integrand in eq . ( [ freeenergydef ] ) remains convergent everywhere except as @xmath206 . however , this dangerous uv region is explicitly excised from the fundamental domain @xmath207 in eq . ( [ fdef ] ) . thus , we conclude that the expression in eq . ( [ freeenergydef ] ) does not suffer from any uv divergences resulting from the exponential growth in the asymptotic degeneracies of states . on the other hand , the expression in eq . ( [ freeenergydef ] ) may experience a divergence due to on - shell states within @xmath10 which may become massless or tachyonic at specific critical temperatures . for example , as the temperature increases , there may exist a critical temperature @xmath208 at which certain states which were massive for @xmath209 become massless at @xmath210 and ultimately tachyonic for @xmath211 . this can therefore be considered an infrared ( ir ) divergence . since such on - shell tachyons correspond to states with worldsheet energies @xmath212 , their contributions to the modular integral in eq . ( [ freeenergydef ] ) _ grow _ as @xmath213 . the contributions from the ( infrared ) @xmath214 region of the fundamental domain then lead to a divergence for @xmath46 . thus , a study of the hagedorn transition in string theory essentially reduces to a study of the _ tachyonic _ structure of @xmath10 as a function of temperature . before proceeding further , however , we caution that we have reached this conclusion only because we have chosen to work in the so - called @xmath207-representation for @xmath46 given in eq . ( [ freeenergydef ] ) . by contrast , utilizing poisson resummations and modular transformations @xcite , we can always rewrite @xmath46 as the integration of a different integrand @xmath215 over the strip : |re| , im>0 . [ sdef ] in such an @xmath216-representation , the ir divergence as @xmath214 is transformed into a uv divergence as @xmath206 . this formulation thus has the advantage of relating the hagedorn transformation directly to a uv phenomenon such as the exponential rise in the degeneracy of states . however , both formulations are mathematically equivalent ; indeed , modular invariance provides a tight relation between the tachyonic structure of a given partition function and the rate of exponential growth in its asymptotic degeneracy of states @xcite . in the following , therefore , we shall utilize the @xmath207-representation for @xmath46 and focus on only the tachyonic structure of @xmath10 , but we shall comment on the connection to the asymptotic degeneracy of states in sect . so what then are the potential tachyonic states within @xmath10 , and at what temperatures @xmath208 do they first arise ? note that we are concerned with states whose masses are temperature - dependent : positive at temperatures below a certain critical temperature , zero at the critical temperature , and tachyonic at temperatures immediately above the critical temperature . the sudden appearance of such new `` thermally massless '' states at a critical temperature @xmath208 is the signal of the appearance of the long - range order normally associated with a phase transition , and the fact that such states generally become tachyonic immediately above @xmath208 reflects the instabilities which are also normally associated with a phase transition . as a result , in order to derive the hagedorn temperature of a given theory , it is sufficient to search for states within the thermal partition function @xmath10 whose masses decrease as a function of temperature , reaching ( and perhaps even crossing ) zero at a certain critical temperature . we shall refer to such states as `` thermally massless '' at the critical temperature . since thermal effects always provide a positive contribution to the squared masses of any states , such states must intrinsically be tachyonic at zero temperature . in other words , for such thermally massless states , masslessness is achieved at the critical temperature @xmath208 as the result of a balance between a tachyonic non - thermal mass contribution ( arising from the characters @xmath217 within @xmath33 ) and an additional positive temperature - dependent thermal mass contribution ( arising from the thermal @xmath218 functions ) . we can quantify this mathematically as follows . a given state with worldsheet energies @xmath12 will contribute a term of the form @xmath219 to the characters @xmath217 within @xmath33 . likewise , as evident from their definitions in the appendix , the thermal @xmath218 functions will make an additional , thermal contribution to these energies which is given by [ h_r , h_l ] = [ ( m a -n / a)^2 , ( m a + n / a)^2 ] [ thermalmasscontribution ] where @xmath220 are respectively the momentum and winding quantum numbers around the thermal circle and where @xmath221 . the conditions for thermal masslessness then become h_r + ( ma - n / a)^2 = 0 , h_l + ( ma+n / a)^2 = 0 , [ masslessness ] which together imply the useful relation @xmath222 . since the thermal contributions in eq . ( [ thermalmasscontribution ] ) are strictly non - negative ( and are not zero , according to our assumption of _ thermal _ masslessness ) , we see that the possibility of obtaining a thermally massless state requires that either @xmath223 or @xmath224 ( or both ) must be negative , and neither can be positive . in other words , the zero - temperature state contributing within the characters @xmath217 within @xmath33 must be a tachyon which is either on - shell ( if @xmath225 ) or off - shell ( if @xmath226 ) ; this tachyonic mode is then `` dressed '' with specific thermal contributions in order to become massless at the critical temperature @xmath227 . moreover , if our solution to eq . ( [ masslessness ] ) has non - zero @xmath26 , then such a state will be massive for all temperatures below this critical temperature , as desired . it will also usually be tachyonic for temperatures immediately above this critical temperature . given these observations , our procedure for determining the hagedorn temperature corresponding to a given thermal partition function @xmath10 is then fairly straightforward . first , we identify any zero - temperature states which are tachyonic ( either on- or off - shell ) contributing to the characters appearing within @xmath10 . for each such state , we then attempt to solve the conditions in eq . ( [ masslessness ] ) , subject to the constraints that @xmath220 are restricted to the values which are appropriate for the corresponding thermal function ( , @xmath32 or @xmath228 and @xmath31 or @xmath229 ) . if such a solution exists and has non - zero @xmath26 , then we have succeeded in identifying a massive state in the full thermal theory which will become massless at the corresponding critical temperature @xmath227 . this then signals a hagedorn transition . in situations where multiple thermally massless states exist , the hagedorn temperature is identified as the lowest of the corresponding critical temperatures , since the presumed existence of a phase transition at that temperature invalidates any analysis based on @xmath33 at temperatures above it . let us now calculate the hagedorn temperatures corresponding to the heterotic partition functions @xmath10 in sect . we focus first on the standard heterotic results without wilson lines , as given in eq . ( [ fake ] ) . for both the @xmath0 and @xmath1 cases , we find that the sector @xmath230 is the sector which is capable of providing thermally massless states at the lowest possible temperature . indeed , solving the conditions for masslessness in eq . ( [ masslessness ] ) , we see that the @xmath231 off - shell tachyon within @xmath232 dressed with the thermal excitations @xmath233 within @xmath37 becomes thermally massless at the critical temperature @xmath234 . this , of course , is nothing but the traditional hagedorn temperature associated with the @xmath0 and @xmath1 heterotic strings . by contrast , let us now examine the thermal partition functions for the @xmath0 string which are constructed using non - trivial wilson lines . for example , if we concentrate on the partition function in eq . ( [ a1 ] ) , we now find that the term @xmath235 is the one which gives rise to thermally massless level - matched states at the lowest possible temperature . indeed , the @xmath127 character @xmath236 gives rise to 32 on - shell @xmath237 tachyons , and these are nothing but the 32 tachyons of the non - supersymmetric @xmath0 heterotic string which serves as the @xmath143 endpoint of the corresponding wilson - line orbifold . moreover , we find that the @xmath238 thermal excitations of these states are massless at @xmath185 , massive below this temperature , and tachyonic above it . indeed , there are no other tachyonic sectors within eq . ( [ a1 ] ) which could give rise to other phase transitions at lower temperatures . thus the hagedorn temperature associated with eq . ( [ a1 ] ) is actually given by @xmath185 , not @xmath239 , and agrees with the locations of the divergences indicated in fig . remarkably , this new temperature is exactly the same as the hagedorn temperature of the type i and type ii strings . the same is true for eqs . ( [ a2 ] ) and ( [ a3 ] ) as well : each of these thermal partition functions corresponds to @xmath185 , not @xmath239 . this makes sense , since each of these wilson lines corresponds to a non - supersymmetric heterotic model containing on - shell tachyons with worldsheet energies @xmath237 . indeed , the only exception is the partition function in eq . ( [ firstso16 ] ) . this too makes sense , since the wilson line in this case corresponds the @xmath127 heterotic string model . although non - supersymmetric , this string model is tachyon - free . indeed , for the partition function given in eq . ( [ firstso16 ] ) , we find that off - shell tachyons with @xmath240 arise within the term @xmath241 ; these , when dressed with the @xmath242 thermal excitations within @xmath37 , become massless at @xmath243 . this is the lowest temperature at which such thermally massless states appear , which identifies this as the hagedorn temperature corresponding to the @xmath127 wilson line . a similar situation exists for the possible thermal extensions of the @xmath1 string . examining eq . ( [ b1 ] ) , we see that only the sector @xmath244 is capable of giving rise to thermally massless level - matched states ; once again , these are the tachyons with energies @xmath237 within @xmath245 , dressed with the @xmath238 thermal excitations within @xmath36 . these states are massless at @xmath185 , massive below this temperature , and tachyonic above it . thus , we see that @xmath185 emerges as the hagedorn temperature following from eq . ( [ b1 ] ) as well . indeed , the same is also true for eqs . ( [ b2 ] ) and ( [ b3 ] ) , while we find that @xmath243 for eq . ( [ secondso16 ] ) . these results are precisely in one - to - one correspondence with those for the @xmath0 string . we conclude , then , that the existence of non - trivial wilson lines in the formulation of finite - temperature heterotic strings has , in most cases , shifted the corresponding heterotic hagedorn temperature from @xmath188 to @xmath185 . remarkably , this is the same hagedorn temperature as that associated with type ii strings . it is easy to understand why this is the case . without a non - trivial wilson line , our thermal heterotic theories are described by eq . ( [ fake ] ) , and the lowest mode contributing within @xmath232 is the ( tachyonic ) ground state of the heterotic theory , with non - level - matched vacuum energies @xmath231 . however , as we have seen , turning on the wilson lines leading to eqs . ( [ a1 ] ) , ( [ a2 ] ) , and ( [ a3 ] ) in the @xmath0 case , or to eqs . ( [ b1 ] ) , ( [ b2 ] ) , and ( [ b3 ] ) in the @xmath1 case , effectively projects this non - level - matched state out of the finite - temperature theory and leaves behind only the `` next - deepest '' tachyon with @xmath237 within @xmath246 . thus , with these wilson lines turned on , this new tachyon becomes the effective ground state of the theory . however , this `` next - deepest '' tachyon has exactly the same worldsheet energies @xmath237 as the ground state of the type ii string . thus it is not surprising that the presence of the non - trivial wilson line shifts the corresponding heterotic hagedorn temperature in such a way that it now matches the type ii value . as discussed in sect . v.a , our analysis of the hagedorn temperature has thus far been based on an analysis of the tachyonic structure of our thermal partition functions . yet we know that there is a tight relation between the hagedorn temperature of a given theory and the exponential rate of growth of its asymptotic degeneracies of bosonic and fermionic states . specifically , if @xmath247 denotes the number of string states with mass @xmath248 , then the thermal partition function is given by @xmath249 . however , if @xmath250 as @xmath251 , then @xmath252 diverges for @xmath253 . this appears to provide a firm link between the hagedorn temperature and the asymptotic degeneracy of states . of course , @xmath254 is not a proper string - theoretic partition function . however , even when we utilize a proper string - theoretic partition function @xmath10 and calculate a proper string - theoretic amplitude as in eq . ( [ fdef ] ) in the @xmath216-representation , the same basic argument continues to apply . we are thus left with the critical question : _ how can we justify a shifted hagedorn temperature @xmath185 for heterotic strings , given that the zero - temperature bosonic and fermionic densities of heterotic states are apparently unchanged ? _ this question is particularly urgent , given that the feature which is inducing this shift in the hagedorn temperature namely the introduction of a non - trivial thermal wilson line does not affect the zero - temperature theory in any way . specifically , an increase in the hagedorn temperature of the heterotic string from the traditional value @xmath255 to a new , higher value @xmath185 would seem to require a corresponding decrease in the exponential rate of growth of the asymptotic density of heterotic string states . in what sense can we understand such a decrease ? to answer this question , let us look again at the original partition function of the zero - temperature ten - dimensional @xmath0 heterotic string model in eq . ( [ start ] ) . recall that @xmath57 and @xmath58 indicate the transverse @xmath52 lorentz spins of the different states which contribute in this theory . as a result of spacetime supersymmetry , this partition function vanishes identically , all of its level - degeneracy coefficients are identically zero . there is no exponential growth here at all . but of course one does not examine the _ total _ partition function in order to derive a hagedorn temperature ; one instead looks at its separate bosonic and fermionic contributions . ordinarily , these contributions would be identified on the basis of the lorentz spins of these states as z^(bosonic)_so(32 ) = z^(8)_boson _ v ( _ i^2 + _ v^2 + _ s^2 + _ c^2 ) , z^(fermionic)_so(32 ) = -z^(8)_boson _ s ( _ i^2 + _ v^2 + _ s^2 + _ c^2 ) , [ usualcomponents ] and indeed each of these expressions separately exhibits an exponential rise in the degeneracy of states which is consistent with the traditional heterotic hagedorn temperature . but what do we really mean by `` bosonic '' and `` fermionic '' in this context ? for most purposes , we would identify states as `` bosonic '' or `` fermionic '' based on their lorentz spins , as above . moreover , by the spin - statistics theorem , this is equivalent to identifying states as bosonic or fermionic based on their zero - temperature quantization statistics . however , _ for the purposes of computing a hagedorn temperature _ , we should really be focused on a _ thermodynamic _ definition of `` bosonic '' and `` fermionic '' wherein we identify states as bosons or fermions on the basis of their matsubara frequencies , , on the basis of their modings around the thermal circle . of course , under normal circumstances , all three of these identifications are equivalent . however , we have already seen in sect . iii that this chain of equivalences is modified in the presence of a non - trivial wilson line : certain states which are `` bosonic '' in terms of their lorentz spins and zero - temperature quantization statistics can nevertheless have half - integer modings @xmath22 around the thermal circle , while other states which are `` fermionic '' in terms of their lorentz spins and zero - temperature quantization statistics can nevertheless have integer modings @xmath32 around the thermal circle . thus , _ in the presence of a non - trivial wilson line , _ certain bosonic states can behave as fermions from a thermodynamic standpoint , and certain fermionic states can behave as bosons . we emphasize that this is _ not _ a violation of the spin - statistics theorem . indeed , the spin - statistics theorem is believed to hold without alteration in string theory , providing a connection between the lorentz spin of a state and its zero - temperature quantization statistics @xcite . rather , as discussed in sect . iii , the effect of the wilson line is to modify the _ thermodynamic manifestation _ of these properties as far as their matsubara modings are concerned . for issues pertaining to zero - temperature physics , these thermodynamic manifestations may be of little consequence . however , when we seek to understand the thermal properties of a theory , these modifications are critical . therefore , if we seek to understand the spectra of bosonic and fermionic states in the heterotic string _ for thermodynamic purposes _ , we should return to the partition function in eq . ( [ start ] ) and separate this expression into individual contributions from bosonic and fermionic states _ on the basis of their matsubara modings around the thermal circle_. it is here where the wilson line comes into play . let us begin by considering the case of the @xmath0 string . without a wilson line , we know that bosonic states will correspond to integer matsubara modings @xmath32 and fermionic states will correspond to half - integer modings @xmath22 . this corresponds to the traditional identifications in eq . ( [ usualcomponents ] ) , and these lead to the traditional hagedorn temperature @xmath188 . however , under the influence a non - trivial wilson line , these identifications can change . for example , let us consider the case of the wilson line corresponding to eq . ( [ a1 ] ) . in this case , the `` bosonic '' contributions to eq . ( [ start ] ) must be identified as those which multiply the thermal sum @xmath47 in eq . ( [ a1 ] ) , while the `` fermionic '' contributions to eq . ( [ start ] ) must be identified as those which multiply the thermal sum @xmath48 in eq . ( [ a1 ] ) . in other words , we replace the identifications in eq . ( [ usualcomponents ] ) with z^(bosonic)_so(32 ) & = & z^(8)_boson _ v ( _ i^2 + _ v^2 ) - _ s ( _ s^2 + _ c^2 ) , + z^(fermionic)_so(32 ) & = & -z^(8)_boson _ s ( _ i^2+_v^2 ) - _ v ( _ s^2+_c^2 ) . [ newcomponents ] in the presence of the non - trivial wilson line , it is therefore _ these _ expressions which serve to define our separate bosonic and fermionic contributions to eq . ( [ start ] ) , and indeed their sum z^(bosonic)_so(32 ) + z^(fermionic)_so(32 ) correctly reproduces the expression in eq . ( [ start ] ) . given these results , we can now calculate the exponential rates of growth in the degeneracies of the states contributing to @xmath256 and @xmath257 in eq . ( [ newcomponents ] ) . we find that while each individual term within @xmath256 and @xmath257 in eq . ( [ newcomponents ] ) continues to exhibit the traditional rate of growth associated with the traditional hagedorn temperature for heterotic strings , _ the minus signs within @xmath256 and @xmath257 have the net effect of cancelling this dominant exponential behavior , leaving behind only a smaller exponential rate of growth for the state degeneracies corresponding to @xmath256 and @xmath257 . _ moreover , as expected , this smaller exponential rate of growth precisely matches the rate of growth that corresponds to the new ( increased ) heterotic hagedorn temperature @xmath185 . similar results also hold the wilson lines associated with eqs . ( [ a2 ] ) and ( [ a3 ] ) . it may seem strange that two terms , each exhibiting a dominant exponential growth rate , can be subtracted and leave behind a sub - dominant exponential growth rate . yet this phenomenon is well known for modular functions such as these , and has been shown to operate in other string - theoretic contexts @xcite . we emphasize that this subtraction is relevant only in the sense that it deforms the exponential growth rate when we count @xmath256 and @xmath257 separately . each string state still continues to appear with positive unit weight in the string fock space , as always , and still contributes to the overall partition function with unit weight and an appropriate sign ( positive for spacetime bosons , negative for spacetime fermions ) . similar results hold for the @xmath1 heterotic string . without a wilson line , the usual identification of bosonic and fermionic states is nothing but z^(bosonic)_e_8e_8 = z^(8)_boson _ v ( _ i+ _ s)^2 , z^(fermionic)_e_8e_8 = -z^(8)_boson _ s ( _ i+ _ s)^2 . [ usualcomponentse8 ] however , in the presence of the wilson line associated with eq . ( [ b1 ] ) , we find that our new thermal identification of bosonic and fermionic states is given by z^(bosonic)_e_8e_8 & = & z^(8)_boson ( _ v_i - _ s_s)(_i+_s ) + z^(fermionic)_e_8e_8 & = & - z^(8)_boson ( _ s_i - _ v_s)(_i+_s ) . [ e8newbosferm ] just as with the @xmath0 string , the minus signs within eq . ( [ e8newbosferm ] ) lead to state degeneracies which show a reduced exponential growth one which is precisely in accordance with the new , increased heterotic hagedorn temperature associated with this wilson line . similar results also hold the wilson lines associated with eqs . ( [ b2 ] ) and ( [ b3 ] ) . this , then , is the essence of the manner in which the asymptotic density of states is ultimately reconciled with the modified hagedorn temperature for heterotic strings . the presence of the non - trivial wilson line `` deforms '' the thermal identification of bosonic and fermionic states , trading states between the separate sets of bosonic and fermionic states in such a way that the net exponential rate of growth for the asymptotic state degeneracy of each set is reduced . there are also other ways to understand this result . for example , one might argue on general conformal - field - theory ( cft ) grounds that such a change in the hagedorn temperature should not be possible . after all , there exists a general result which relates the hagedorn temperature of a given closed - string theory to the central charges @xmath258 of its underlying worldsheet cft s : t_h = ( + ) ^-1 . [ thag ] for type ii strings , we have @xmath259 , while for heterotic strings , we have @xmath260 . however , in deriving eq . ( [ thag ] ) , there is only place in which the central charges enter : this is in setting the ground - state energies @xmath261 . moreover , as we have seen in sect . v.b , the wilson line has effectively projected the true heterotic ground state with @xmath231 out of the spectrum , leaving behind only the `` next - deepest '' tachyon with @xmath237 to serve as the effective ground state of the theory . thus , in the presence of the wilson line , the effective central charges of the theory become @xmath259 , just as for type ii strings . we see , then , that the introduction of a non - trivial wilson line induces both a shift in the vacuum energy of the effective ground state and a shift in the asymptotic rates of growth for the state degeneracies . these shifts , of course , are flip sides of the same coin , deeply related to each other through modular transformations . indeed , these are nothing but equivalent uv / ir descriptions of the same phenomenon , all arising due to the existence of the non - trivial wilson line . under poisson resummation , a half - integer shift in the moding of a given set of string states around the thermal circle translates into an overall @xmath5 phase ( , a minus sign ) in front of the corresponding character in the partition function . thus the wilson line , which shifts the apparent thermal modings of certain states in the theory , necessarily induces a corresponding change in the asymptotic state degeneracies and a corresponding shift in the hagedorn temperature of the theory . we thus conclude that the introduction of a non - trivial wilson line has the potential to change the hagedorn temperature of the resulting thermal theory , and in many cases actually shifts this temperature from its traditional heterotic value to a new value which is the same as that associated with type i and type ii strings . indeed , although the heterotic string would navely appear to have a slightly lower hagedorn temperature than the type ii string due to its non - level - matched ground state , we see that the introduction of a non - trivial wilson line has the potential to eliminate this discrepancy . such wilson lines deform the effective worldsheet central charges of the heterotic theory as far as its thermal properties are concerned , and lead to new , effective ground states for the theory as well as modified rates of exponential growth for the corresponding bosonic and fermionic densities of states . both effects then alter hagedorn temperature of the heterotic string , and potentially bring it into agreement with the type i and type ii value . as we have seen in sect . v.b , the thermal heterotic theories without wilson lines have the hagedorn temperature @xmath262 ; indeed , this is the case because the relevant partition functions in eq . ( [ fake ] ) each contain a term of the form @xmath263 . indeed , such a term encapsulates the contribution of the @xmath231 tachyon , and this is the state which , when dressed with the thermal excitations @xmath233 within @xmath37 , is massive at low temperatures but becomes massless at @xmath239 . by contrast , we found that this state is projected out of the thermal partition function when non - trivial wilson lines are introduced . for example , for the wilson lines leading to eqs . ( [ a1 ] ) and ( [ b1 ] ) , we found that the `` next - deepest '' remaining tachyons have worldsheet energies @xmath237 ; indeed , these are the tachyons which contribute to the terms @xmath244 which appear within eqs . ( [ a1 ] ) and ( [ b1 ] ) . these tachyons , when dressed with the @xmath238 thermal excitations within @xmath36 , are massless at @xmath185 , massive below this temperature , and tachyonic above it . thus , @xmath185 emerges as the hagedorn temperature corresponding to these choices of wilson lines . we found that this result also holds for the wilson lines leading to eqs . ( [ a2 ] ) , ( [ a3 ] ) , ( [ b2 ] ) , and ( [ b3 ] ) . by contrast , we found that the hagedorn temperature is @xmath243 for the wilson lines leading to eqs . ( [ firstso16 ] ) and ( [ secondso16 ] ) . indeed , in these cases , even the @xmath237 tachyons are projected out of the spectrum , so that the lowest possible remaining tachyons have @xmath240 and contribute to the @xmath37 sector . given these results , it is natural to wonder whether other hagedorn temperatures might also be possible . indeed , one might even wonder whether there exist wilson lines for which the hagedorn transition might be avoided completely ! of course , neither of these options is possible for the supersymmetric heterotic strings in ten dimensions , for we have presented a complete classification of all self - consistent wilson lines in such cases , and our results are quoted above . however , in lower dimensions , this might no longer be the case . in fact , in lower dimensions , even the type ii superstrings can have non - trivial wilson lines which are associated with the gauge symmetries that emerge upon compactification . thus , these questions become relevant for type ii strings as well as heterotic . towards this end , we shall now provide a general classification all of the hagedorn temperatures which can ever be realized for closed strings with non - trivial wilson lines . our analysis will apply to all closed strings , both type ii and heterotic . for concreteness , we shall restrict our attention to theories built from only @xmath5 orbifolds , so that @xmath264 are quantized in half - integer values . given the heterotic constraints @xmath265 and @xmath266 ( which also subsume the type ii constraints @xmath267 ) , we then find that there are only eight different terms which could possibly appear in @xmath10 and trigger a hagedorn transition . these are listed in table [ hagedornia ] , along with their corresponding thermal excitations and hagedorn temperatures [ obtained by solving eq . ( [ masslessness ] ) ] . it is interesting to note the mathematical fact that these terms come in `` dual '' pairs under which @xmath268 and @xmath269 . roughly speaking , this duality corresponds to exchanging the direction of the corresponding `` interpolating '' thermal partition functions , exchanging the @xmath147 and @xmath270 endpoints . although this `` thermal duality '' phenomenon has played a significant role in other work @xcite ) , it will not be critical for the following discussion . we can therefore view the emergence of this duality within table [ hagedornia ] as a mere mathematical curiosity . .complete set of possible terms ( labeled a through h ) which can potentially trigger a hagedorn transition for string models built with @xmath5 orbifolds . as discussed in the text , case a is responsible for the traditional heterotic hagedorn transition , while case c with @xmath271 is responsible for the traditional type ii hagedorn transition as well as the `` shifted '' heterotic hagedorn transition . cases b and d can only arise in theories which are already tachyonic ( and hence unstable ) at zero temperature , while case h is guaranteed to arise for all heterotic strings which are supersymmetric at zero temperature . observe that all of these possibilities come in `` dual '' pairs under which @xmath268 and @xmath269 . thus the two possibilities within cases a and b are dual to each other , while cases c and g are dual to cases d and h respectively ( and vice versa ) . by contrast , cases e and f are each self - dual . note that cases a , g , and h are unique to heterotic strings , while all other cases can in principle arise in both heterotic and type ii strings . [ cols="^,^,^,^,^,^",options="header " , ] these two sets of theories are summarized in tables [ so32table ] and [ e8table ] , and it is readily apparent that these theories share close similarities with each other . in each case , the ` a ' theories correspond to thermal theories without wilson lines , while the ` b ' theories correspond to the wilson line @xmath272 . both options lead to thermal theories which are locally stable within the thermal landscape , and indeed these are the _ only _ theories which have this property . moreover , as remarked in sect . iv for the case of the @xmath0 string , the free - energy difference between the ` a ' theory and the corresponding ` b ' theory is extremely small indeed , such differences would not even be visible on the plots in fig . [ ffig ] . consequently , even though the ` b ' theories are technically only metastable , there is very little dynamical `` force '' which would cause our universe to flow from the ` b ' state to the corresponding ` a ' state . ( phrased more precisely , an instanton analysis of the transitions from the ` b ' vacua to the ` a ' vacua would lead to a very small decay width or equivalently an extremely long lifetime for the ` b ' theories . ) as a result , it is quite possible that our universe , if somehow `` born '' in the ` b ' state , might reside there essentially forever . however , as briefly mentioned in sect . iv , even this notion presupposes the existence of transitions between theories with different wilson lines something which is not at all obvious , given the quantized topological nature of the wilson lines themselves . we are therefore faced with a situation in which both the ` a ' theories and the ` b ' theories may be considered as legitimate `` ground '' states for our thermal heterotic strings . indeed , as we have explicitly seen in the case of the @xmath0 landscapes in sect . vi.b , both the @xmath273 and @xmath274 thermal heterotic theories have equal claims to be considered as the legitimate heterotic analogue of the @xmath0 type i thermal theory . it is quite remarkable that the heterotic string gives rise to such a situation . however , this phenomenon ultimately rests upon the existence of a unique wilson line which simultaneously has all of the properties needed in order to endow the resulting ` b ' theories with these critical features . as we have seen , the type i string _ a priori _ has a richer set of allowed wilson lines , yet none of these has the required properties . given the existence of these ` b ' theories , many questions naturally arise . for example , it is well known that the zero - temperature supersymmetric @xmath0 type i and heterotic strings are related to each other under strong / weak coupling s - duality relations . one naturally wonders , therefore , whether such s - duality relations extend to finite temperatures . however , as we have seen , the `` landscape '' of the @xmath0 heterotic string at finite temperature includes not only the @xmath273 theory but also the @xmath274 theory . it would therefore be interesting to understand how this `` doubling '' phenomenon can be reconciled with the existence of a unique thermal @xmath0 theory on the type i side @xcite . another important question concerns whether there might be some other way ( , through dynamical means , or perhaps through self - consistency arguments ) in order to develop a thermal `` vacuum selection '' criterion and thereby assert that only one of these theories is preferred or allowed , and the other excluded . this question will be examined in refs . it is well known , for example , that the @xmath273 theory has a number of unexpected and disturbing features , the least of which is the fact that the @xmath143 limit of this theory is again supersymmetric . this is also true of the @xmath275 theory . this property is completely surprising , given our expectation that finite - temperature effects should treat bosons and fermions differently , and is very different from what occurs for type i and type ii thermal theories . indeed , this behavior leads to several disturbing features ( see , , the discussion in ref . by contrast , the @xmath274 theory is relatively natural from this point of view : the @xmath143 limit of this theory lacks spacetime supersymmetry , but preserves the underlying @xmath0 gauge symmetry . even in the @xmath1 case , the ` b ' theory has a non - supersymmetric @xmath143 limit yet breaks the gauge symmetry as minimally as possible . thus , if we could somehow argue that the ` b ' theories are the unique correct thermal heterotic theories , we would then reach the remarkable conclusion that all string theories in ten dimensions , whether open or closed , whether type i or type ii or heterotic , actually have a unique hagedorn temperature . in other words , we would be in the aesthetically pleasing situation in which we could assert a existence of a single hagedorn temperature for string theory as a whole . this issue is discussed further in refs . in this paper , we investigated the consequences of introducing a non - trivial wilson line ( or equivalently a non - trivial chemical potential ) when formulating string theories at finite temperature . we focused on the heterotic and type i strings in ten dimensions , and surveyed the possible wilson lines which might be introduced when extending these strings to finite temperature . we found a rich structure of resulting thermal string theories , and showed that in the heterotic case , some of these new thermal theories even have hagedorn temperatures which are shifted from their usual values . remarkably , these shifts in the hagedorn temperature are not in conflict with the densities of bosonic and fermionic states which are exhibited by their zero - temperature counterparts . we also demonstrated that our new thermal string theories can be interpreted as extrema of a continuous thermal free - energy `` landscape '' . finally , as part of this study , we also uncovered a pair of unique finite - temperature extensions of the heterotic @xmath0 and @xmath1 strings which involve a non - trivial wilson line , but which are nevertheless metastable in this thermal landscape . as we have argued , these new thermal theories ( the so - called ` b ' theories discussed in sect . vi.c ) represent _ bona fide _ alternatives to the traditional thermal heterotic strings , and may be viewed as equally legitimate candidate finite - temperature extensions of the zero - temperature @xmath0 and @xmath1 heterotic strings in ten dimensions . indeed , as we have seen , the analysis in this paper also illustrates that the @xmath273 and @xmath274 heterotic theories are on equal footing as potential heterotic analogues of the thermal @xmath0 type i string . clearly , many outstanding questions remain . for example , in this paper we have found that for each of the supersymmetric heterotic strings in ten dimensions , there exists a unique non - trivial wilson line which leads to a metastable theory . however , it would be interesting to understand more generally for which string theories this will be the case . likewise , we have found that these new wilson lines lead to an increased hagedorn temperature . it is therefore natural to wonder whether there might exist situations in which metastable wilson lines manage to avoid the hagedorn transition entirely . it is also important to realize that in the type i case , our analysis has essentially focused on those wilson lines associated with the open - string sector . in ten dimensions , this was a legitimate restriction , as one can not introduce wilson lines in the closed - string sector of ten - dimensional type i strings . however , it would clearly be of great interest to examine the situation in lower dimensions . in lower dimensions , perturbative type ii strings can have non - trivial gauge groups which are generated by compactification . as a result , there can be non - trivial wilson lines that are associated with such type ii compactifications , and therefore the potential type i orientifolds of such models can quickly become quite numerous . in particular , the set of candidate thermal extensions of a given type i string model might include models which are distinct in terms of their closed - string sectors as well as their open - string sectors . of course , our analysis in this paper is subject to a number of important caveats . first , we have been dealing with one - loop string vacuum amplitudes , and likewise considering only the tree - level ( non - interacting ) particle spectrum . thus , we are neglecting all sorts of particle interactions . gravitational effects , in particular , can be expected to change the spectrum quite dramatically , and have recently been argued to eliminate the hagedorn transition completely by deforming the resulting spectrum away from the expected exponential rise in the degeneracy of states . however , the purpose of this paper has been to show that even within the non - interacting string theory which has been studied for more than two decades , the introduction of non - trivial wilson lines can have significant effects on the resulting thermal theories . this work has clearly focused on the thermal behavior of string theories at temperatures _ below _ the hagedorn transition . as such , it is not clear that these results will shed any light on that feature which remains the most mysterious aspect of string thermodynamics : the nature of the hagedorn transition itself . however , the involvement of non - trivial wilson lines may eventually have ramifications in this regard that are not yet apparent . in this paper we have restricted our attention to the thermal properties of _ perturbative _ type i and heterotic strings , especially when non - trivial wilson lines are introduced into the mix . however , it would clearly be of great interest to consider the implications of these results at the non - perturbative level , once d - branes and other non - perturbative structures are included . in particular , it would be interesting to study the possible implications of these results for the thermodynamics of d@xmath276-branes @xcite as well as for the cosmological applications of finite - temperature d - branes @xcite . in this connection , it would also be interesting to understand the thermal consequences of our observations within the recent brane - world scenarios , as these frameworks involve a subtle dynamical interplay between bulk ( closed - string ) physics and brane ( open - string ) physics . likewise , it would also be interesting to extend our results to non - flat backgrounds in order to address important questions such as the thermodynamics of black holes , the ads / cft correspondence , and so forth . continuing this line , it would also be interesting to understand the implications of these results for the existence of strong / weak coupling duality relations at finite temperature . it is certainly aesthetically pleasing that the heterotic and type i hagedorn temperatures can be brought into agreement through the introduction of non - trivial wilson lines , and this immediately raises the question whether the zero - temperature heterotic / type i dualities can be extended to finite temperature . in particular , although the perturbative type i string lacks states transforming as @xmath0 spinors which is ultimately the reason why the @xmath273 and @xmath274 theories are equivalent on the type i side such states do emerge non - perturbatively @xcite . thus we can expect that the non - perturbative type i theory will have a thermal behavior which is even closer to that of the perturbative heterotic string , and for which a non - trivial wilson line might be examined as well . these issues will be examined in more detail in ref . @xcite . but above all , perhaps the most pregnant issue is the existence of the so - called ` b ' theories , and the roles these theories may ultimately play in the general structure of string theory at finite temperature . given that the ` a ' and the ` b ' theories appear to be equally compelling as finite - temperature extensions of the traditional zero - temperature heterotic strings , it remains to investigate whether there might be a thermal vacuum selection principle that favors one over the other ( see , , refs . @xcite ) . moreover , if no such selection principle exists , it will be important to study the different physics to which they each lead , and the possibility of phase transitions between them . these issues clearly warrant further study . we are happy to thank e. dudas for discussions . this work was supported in part by the department of energy under grant de - fg02 - 04er-41298 , and by the national science foundation through its employee ir / d program . ml also wishes to acknowledge the ecole polytechnique in france and the perimeter institute and mcmaster university in canada , where portions of this work were done . in these latter institutions , ml was supported in part by anr grant anr-05-blan-0079 - 02 ( france ) , rtn contracts mrtn - ct-2004 - 005104 and mrtn - ct-2004 - 503369 ( france ) , cnrs pics # 2530 , # 3059 , and # 3747 ( france ) , european union excellence grant mext - ct-2003 - 509661 , and the natural sciences and engineering research council ( nserc ) of canada . the opinions and conclusions expressed herein are those of the authors , and do not represent either the department of energy or the national science foundation . in this appendix , we collect the mathematical expressions which are used in this paper for the traces over relevant string fock spaces . these results also serve to define our notations and conventions . for any temperature @xmath2 , we define the corresponding dimensionless temperature @xmath277 where @xmath278 . we also define the associated thermal radius @xmath279 . a field compactified on a circle with this radius then accrues integer matsubara momentum and winding modes around this thermal circle , resulting in left- and right - moving spacetime momenta of the forms p_r = 1 ( ma - n / a ) , p_l = 1 ( ma+n / a ) . here @xmath27 and @xmath26 respectively represent the momentum and winding quantum numbers of the field in question . the contribution to the partition function from such thermal modes then takes the form of the double summation z_circ(,t ) = _ m , n ^ p_r^2/2 q^ p_l^2/2 = _ m , n ^(ma - n / a)^2/4 q^(ma+n / a)^2/4 [ zcircdef ] where @xmath65 and where @xmath280 respectively denote @xmath281 and @xmath282 . note that @xmath283 as @xmath284 , while @xmath285 as @xmath286 . the trace @xmath287 is sufficient for compactifications on a thermal circle . however , in this paper we are interested in compactifications on @xmath5 orbifolds of the thermal circle . towards this end , we introduce @xcite four new functions @xmath24 and @xmath25 which are the same as the summation in @xmath287 in eq . ( [ zcircdef ] ) except for the following restrictions on their summation variables : _ 0 & = & m , n even + _ 1/2 & = & m+ , n even + _ 0 & = & m , n odd + _ 1/2 & = & m+ , n odd . [ eofunctions ] note that these functions are to be distinguished from a related ( and also often used ) set of functions with the same names in which the roles of @xmath27 and @xmath26 are exchanged . under the modular transformation @xmath288 , the first three functions are invariant while @xmath37 picks up a minus sign ; likewise , under @xmath289 , these functions mix according to ( -1/ ) = ( ) . [ smixing ] in the @xmath284 limit , @xmath36 and @xmath37 each vanish while @xmath290 ; by contrast , as @xmath291 , @xmath48 and @xmath37 each vanish while @xmath292 . clearly , @xmath293 . the thermal @xmath294 functions are primarily of relevance for closed strings , since such strings have both momentum and winding modes . for open strings , by contrast , we instead define the thermal functions _ m p_m , _ m p_m+1/2 , [ eeprime ] where @xmath295 , with @xmath296 . note that @xmath81 is the open - string analogue of @xmath47 , while @xmath82 is the open - string analogue of @xmath48 . indeed , the remaining closed - string functions @xmath25 do not have open - string analogues because they both involve non - trivial winding modes , and winding modes do not survive the sorts of direct orientifold projections that we implement in this paper in order to construct our thermal type i string models . we begin by recalling the standard definitions of the dedekind @xmath298 and jacobi @xmath299 functions : ( q ) & & q^1/24 _ n=1^ ( 1-q^n ) = _ n=-^ ( -1)^n q^3(n-1/6)^2/2 + _ 1(q ) & & 2_n=0^(-1)^n q^(n+1/2)^2/2 + _ 2(q ) & & 2 q^1/8 _ n=1^(1+q^n)^2 ( 1-q^n ) = 2_n=0^q^(n+1/2)^2/2 + _ 3(q ) & & _ n=1^(1+q^n-1/2)^2 ( 1-q^n ) = 1 + 2_n=1^q^n^2/2 + _ 4(q ) & & _ n=1^(1-q^n-1/2)^2 ( 1-q^n ) = 1 + 2_n=1^(-1)^n q^n^2/2 . [ etathetadefs ] these functions satisfy the identities @xmath300 and @xmath301 . note that @xmath302 has a vanishing @xmath159-expansion and is modular invariant ; its infinite - product representation has a vanishing coefficient and is thus not shown . this function is nevertheless included here because it plays a role within string partition functions as the indicator of the chirality of fermionic states , as discussed below . the partition function of @xmath26 free bosons is given by z^(n)_boson _ 2 ^ -n/2 ( ) ^-n . [ bosons ] by contrast , the characters of the level - one @xmath297 affine lie algebras are defined in terms of both the @xmath298- and the @xmath303-functions . recall that at affine level one , the @xmath297 algebra for each @xmath304 has four distinct representations : the identity ( @xmath305 ) , the vector ( @xmath306 ) , the spinor ( @xmath193 ) , and the conjugate spinor ( @xmath307 ) . in general , these representations have conformal dimensions @xmath308 , and their characters are given by _ i & = & ( ^n + ^n)/^n = q^h_i - c/24 ( 1 + n(2n-1)q + ... ) + _ v & = & ( ^n - ^n)/^n = q^h_v - c/24 ( 2n + ... ) + _ s & = & ( ^n + i^-n _ 1^n)/^n = q^h_s - c/24 ( 2^n-1 + ... ) + _ c & = & ( ^n - i^-n _ 1^n)/^n = q^h_c - c/24 ( 2^n-1 + ... ) [ chis ] where the central charge is @xmath309 at affine level one . the vanishing of @xmath310 implies that @xmath129 and @xmath311 have identical @xmath159-expansions ; this is a reflection of the conjugation symmetry between the spinor and conjugate spinor representations . when @xmath297 represents a transverse spacetime lorentz group , the distinction between @xmath193 and @xmath307 is equivalent to relative spacetime chirality ; the choice of which spacetime chirality is to be associated with @xmath193 or @xmath307 is a matter of convention . note that the special case @xmath52 has a further triality symmetry under which the vector and spinor representations are indistinguishable . thus , for @xmath52 , we find that @xmath177 , an identity already given below eq . ( [ etathetadefs ] ) in terms of @xmath303-functions . the above results are primarily of relevance for closed strings , where @xmath4 is the complex torus modular parameter and where @xmath312 . however , with only small modifications , these functions can also be used to describe the partition - function contributions in type i strings . in general , the one - loop type i partition function includes not only a closed sector with contributions from a torus and a klein bottle , but also an open sector with contributions from a cylinder and a mbius strip . it turns out that all four of these contributions can be written in terms of the above cft characters @xmath313 , which are strictly defined as functions of @xmath312 . indeed , all that changes is the definition of @xmath4 : for the torus , @xmath4 will continue to represent the complex modulus , while for the klein bottle , cylinder , and mbius strip , @xmath4 will instead represent the modulus of the double - covering torus : [ taucases ] likewise , the contribution from worldsheet bosons will also change from that in eq . ( [ bosons ] ) to z_boson^(n ) = _ 2 ^ -n/2 ^-n . finally , in this paper we shall also define the hatted characters @xmath314 , where @xmath315 are the conformal weights of the corresponding primary fields @xmath115 in the underlying cft . thus , the hatted characters are explicitly real . these characters are particularly useful for expressing the contributions from the mbius sectors . for an introduction , see : + m. b. green , j. a. schwarz and e. witten , _ superstring theory , vols . i and ii _ ( cambridge university press , 1987 ) ; + j. polchinski , _ string theory , vols . i and ii _ ( cambridge university press , 1998 ) . e.g. _ : a. roberge and n. weiss , nucl . b * 275 * , 734 ( 1986 ) ; + m. g. alford , a. kapustin and f. wilczek , phys . d * 59 * , 054502 ( 1999 ) [ hep - lat/9807039 ] ; + a. hart , m. laine and o. philipsen , phys . lett . b * 505 * , 141 ( 2001 ) [ hep - lat/0010008 ] ; + p. de forcrand and o. philipsen , nucl . b * 642 * , 290 ( 2002 ) [ hep - lat/0205016 ] ; + m. delia and m. -p . lombardo , phys . d * 67 * , 014505 ( 2003 ) [ hep - lat/0209146 ] ; + p. h. damgaard , u. m. heller , k. splittorff , b. svetitsky and d. toublan , phys . d * 73 * , 105016 ( 2006 ) [ hep - th/0604054 ] ; + m. delia , f. di renzo and m. p. lombardo , phys . rev . d * 76 * , 114509 ( 2007 ) [ arxiv:0705.3814 [ hep - lat ] ] ; + y. sakai , k. kashiwa , h. kouno and m. yahiro , phys . d * 77 * , 051901 ( 2008 ) [ arxiv:0801.0034 [ hep - ph ] ] ; c. lehner , m. ohtani , j. j. m. verbaarschot and t. wettig , phys . d * 79 * , 074016 ( 2009 ) [ arxiv:0902.2640 [ hep - th ] ] . m. mcguigan , phys . d * 38 * , 552 ( 1988 ) ; + c. kounnas and b. rostand , nucl . b * 341 * , 641 ( 1990 ) ; + i. antoniadis and c. kounnas , phys . b * 261 * , 369 ( 1991 ) ; + i. antoniadis , j. p. derendinger and c. kounnas , nucl . b * 551 * , 41 ( 1999 ) [ arxiv : hep - th/9902032 ] . c. angelantonj , m. cardella and n. irges , phys . b * 641 * , 474 ( 2006 ) [ arxiv : hep - th/0608022 ] ; + c. angelantonj , c. kounnas , h. partouche and n. toumbas , nucl . b * 809 * , 291 ( 2009 ) [ arxiv:0808.1357 [ hep - th ] ] . i. antoniadis , e. dudas and a. sagnotti , nucl . b * 544 * , 469 ( 1999 ) [ arxiv : hep - th/9807011 ] . c. angelantonj and a. sagnotti , phys . rept . * 371 * , 1 ( 2002 ) [ erratum - ibid . * 376 * , 339 ( 2003 ) ] [ arxiv : hep - th/0204089 ] . j. a. harvey and j. a. minahan , phys . b * 188 * , 44 ( 1987 ) ; + p. horava , nucl . b * 327 * , 461 ( 1989 ) ; + g. pradisi and a. sagnotti , phys . b * 216 * , 59 ( 1989 ) ; + a. sagnotti , arxiv : hep - th/0208020 . r. hagedorn , nuovo cim . suppl . * 3 * , 147 ( 1965 ) . s. fubini and g. veneziano , nuovo cim . a * 64 * , 811 ( 1969 ) ; + k. huang and s. weinberg , phys . lett . * 25 * , 895 ( 1970 ) ; + s. frautschi , phys . rev . d * 3 * , 2821 ( 1971 ) ; + r. d. carlitz , phys . d * 5 * , 3231 ( 1972 ) ; + n. cabibbo and g. parisi , phys . b * 59 * , 67 ( 1975 ) ; + l. susskind , phys . d * 20 * , 2610 ( 1979 ) . p. g. o. freund and j. l. rosner , phys . lett . * 68 * , 765 ( 1992 ) ; + j. r. cudell and k. r. dienes , phys . lett . * 69 * , 1324 ( 1992 ) [ erratum - ibid . * 69 * , 2311 ( 1992 ) ] [ arxiv : hep - ph/9207242 ] ; phys . lett . * 72 * , 187 ( 1994 ) [ arxiv : hep - th/9309126 ] . e.g. _ : + m. j. bowick and l. c. r. wijewardhana , phys . lett . * 54 * , 2485 ( 1985 ) ; + s. h. h. tye , phys . b * 158 * , 388 ( 1985 ) ; + b. sundborg , nucl . phys . b * 254 * , 583 ( 1985 ) ; + e. alvarez , nucl . phys . b * 269 * , 596 ( 1986 ) ; + e. alvarez and m. a. r. osorio , phys . d * 36 * , 1175 ( 1987 ) ; physica * a 158 * , 449 ( 1989 ) [ erratum - ibid . a * 160 * , 119 ( 1989 ) ] ; phys . b * 220 * , 121 ( 1989 ) ; + m. axenides , s. d. ellis and c. kounnas , phys . d * 37 * , 2964 ( 1988 ) ; + y. leblanc , phys . d * 38 * , 3087 ( 1988 ) ; + b. a. campbell , j. r. ellis , s. kalara , d. v. nanopoulos and k. a. olive , phys . b * 255 * , 420 ( 1991 ) . m. j. bowick and s. b. giddings , nucl . phys . b * 325 * , 631 ( 1989 ) ; + s. b. giddings , phys . lett . b * 226 * , 55 ( 1989 ) ; + n. deo , s. jain and c. i. tan , phys . d * 40 * , 2626 ( 1989 ) ; + d. a. lowe and l. thorlacius , phys . d * 51 * , 665 ( 1995 ) [ arxiv : hep - th/9408134 ] . f. englert and j. orloff , nucl . b * 334 * , 472 ( 1990 ) ; + m. hellmund and j. kripfganz , phys . b * 241 * , 211 ( 1990 ) ; + m. a. r. osorio and m. a. vazquez - mozo , phys . b * 280 * , 21 ( 1992 ) [ arxiv : hep - th/9201044 ] ; phys . d * 47 * , 3411 ( 1993 ) [ arxiv : hep - th/9207002 ] ; + m. laucelli meana , m. a. r. osorio and j. puente penalba , phys . b * 400 * , 275 ( 1997 ) [ arxiv : hep - th/9701122 ] . m. a. vazquez - mozo , phys . b * 388 * , 494 ( 1996 ) [ arxiv : hep - th/9607052 ] ; + j. l. f. barbon and m. a. vazquez - mozo , nucl . b * 497 * , 236 ( 1997 ) [ arxiv : hep - th/9701142 ] ; + s. chaudhuri and d. minic , phys . b * 433 * , 301 ( 1998 ) [ arxiv : hep - th/9803120 ] ; + k. r. dienes , e. dudas , t. gherghetta and a. riotto , nucl . b * 543 * , 387 ( 1999 ) [ arxiv : hep - ph/9809406 ] ; + s. a. abel , j. l. f. barbon , i. i. kogan and e. rabinovici , jhep * 9904 * , 015 ( 1999 ) [ arxiv : hep - th/9902058 ] ; + i. i. kogan , a. kovner and m. schvellinger , jhep * 0107 * , 019 ( 2001 ) [ arxiv : hep - th/0103235 ] ; + s. chaudhuri , arxiv : hep - th/0502141 . m. mcguigan , phys . d * 42 * , 2040 ( 1990 ) ; + d. a. lowe and a. strominger , phys . d * 51 * , 1793 ( 1995 ) [ arxiv : hep - th/9410215 ] ; + g. grignani , m. orselli and g. w. semenoff , jhep * 0111 * , 058 ( 2001 ) [ arxiv : hep - th/0110152 ] ; + j. l. f. barbon and e. rabinovici , jhep * 0203 * , 057 ( 2002 ) [ arxiv : hep - th/0112173 ] ; + h. liu , arxiv : hep - th/0408001 . l. a. pando zayas and d. vaman , phys . d * 67 * , 106006 ( 2003 ) [ arxiv : hep - th/0208066 ] ; + b. r. greene , k. schalm and g. shiu , nucl . b * 652 * , 105 ( 2003 ) [ arxiv : hep - th/0208163 ] ; + r. c. brower , d. a. lowe and c. i. tan , nucl . b * 652 * , 127 ( 2003 ) [ arxiv : hep - th/0211201 ] ; + g. grignani , m. orselli , g. w. semenoff and d. trancanelli , jhep * 0306 * , 006 ( 2003 ) [ arxiv : hep - th/0301186 ] ; + s. j. hyun , j. d. park and s. h. yi , jhep * 0311 * , 006 ( 2003 ) [ arxiv : hep - th/0304239 ] ; + f. bigazzi and a. l. cotrone , jhep * 0308 * , 052 ( 2003 ) [ arxiv : hep - th/0306102 ] . j. g. russo , phys . b * 335 * , 168 ( 1994 ) [ arxiv : hep - th/9405118 ] ; + j. g. russo and a. a. tseytlin , nucl . b * 454 * , 164 ( 1995 ) [ arxiv : hep - th/9506071 ] ; + j. g. russo , nucl . b * 602 * , 109 ( 2001 ) [ arxiv : hep - th/0101132 ] . s. d. odintsov , phys . b * 237 * , 63 ( 1990 ) ; phys . b * 274 * , 338 ( 1992 ) [ sov . j. nucl . phys . * 55 * , 440 ( 1992 ) ] [ yad . fiz . * 55 * , 795 ( 1992 ) ] ; + a. a. bytsenko and s. d. odintsov , phys . b * 243 * , 63 ( 1990 ) [ yad . fiz . * 52 * , 1495 ( 1990 sjnca,52,945 - 948.1990 ) ] . t. harmark and n. a. obers , phys . b * 485 * , 285 ( 2000 ) [ arxiv : hep - th/0005021 ] ; + m. berkooz and m. rozali , jhep * 0005 * , 040 ( 2000 ) [ arxiv : hep - th/0005047 ] ; + d. kutasov and d. a. sahakyan , jhep * 0102 * , 021 ( 2001 ) [ arxiv : hep - th/0012258 ] . s. s. gubser , s. gukov , i. r. klebanov , m. rangamani and e. witten , j. math . phys . * 42 * , 2749 ( 2001 ) [ arxiv : hep - th/0009140 ] ; + j. l. f. barbon and e. rabinovici , jhep * 0106 * , 029 ( 2001 ) [ arxiv : hep - th/0104169 ] . s. a. abel , nucl . b * 372 * , 189 ( 1992 ) ; + s. a. abel , k. freese and i. i. kogan , jhep * 0101 * , 039 ( 2001 ) [ arxiv : hep - th/0005028 ] ; + s. abel , k. freese and i. i. kogan , phys . lett . b * 561 * , 1 ( 2003 ) [ arxiv : hep - th/0205317 ] . k. r. dienes , nucl . b * 429 * , 533 ( 1994 ) [ arxiv : hep - th/9402006 ] ; nucl . b * 611 * , 146 ( 2001 ) [ arxiv : hep - ph/0104274 ] ; + k. r. dienes , m. moshe and r. c. myers , phys . lett . * 74 * , 4767 ( 1995 ) [ arxiv : hep - th/9503055 ] . k. r. dienes and m. lennek , arxiv : hep - th/0312173 ; phys . d * 70 * , 126005 ( 2004 ) [ arxiv : hep - th/0312216 ] ; phys . d * 70 * , 126006 ( 2004 ) [ arxiv : hep - th/0312217 ] . s. chaudhuri , phys . d * 65 * , 066008 ( 2002 ) [ arxiv : hep - th/0105110 ] ; arxiv : hep - th/0409301 . k. r. dienes and m. lennek , hep - th/0505233 ; hep - th/0507201 . a. a. tseytlin , nucl . b * 524 * , 41 ( 1998 ) [ arxiv : hep - th/9802133 ] ; + j. ambjorn , yu . makeenko , g. w. semenoff and r. j. szabo , phys . d * 60 * , 106009 ( 1999 ) [ arxiv : hep - th/9906134 ] ; + c. angelantonj , c. kounnas , h. partouche and n. toumbas , nucl . b * 809 * , 291 ( 2009 ) [ arxiv:0808.1357 [ hep - th ] ] . m. a. vazquez - mozo , phys . b * 388 * , 494 ( 1996 ) [ arxiv : hep - th/9607052 ] ; + s. lee and l. thorlacius , phys . b * 413 * , 303 ( 1997 ) [ arxiv : hep - th/9707167 ] ; + s. a. abel , j. l. f. barbon , i. i. kogan and e. rabinovici , jhep * 9904 * , 015 ( 1999 ) [ arxiv : hep - th/9902058 ] . j. l. f. barbon and m. a. vazquez - mozo , nucl . b * 497 * , 236 ( 1997 ) [ arxiv : hep - th/9701142 ] ; + j. ambjorn , y. m. makeenko and g. w. semenoff , phys . b * 445 * , 307 ( 1999 ) [ arxiv : hep - th/9810170 ] . k. r. dienes , e. dudas , t. gherghetta and a. riotto , nucl . b * 543 * , 387 ( 1999 ) [ arxiv : hep - ph/9809406 ] ; + s. alexander , r. h. brandenberger and d. easson , phys . d * 62 * , 103509 ( 2000 ) [ arxiv : hep - th/0005212 ] ; + s. a. abel , k. freese and i. i. kogan , jhep * 0101 * , 039 ( 2001 ) [ arxiv : hep - th/0005028 ] ; + m. majumdar and a. christine - davis , jhep * 0203 * , 056 ( 2002 ) [ arxiv : hep - th/0202148 ] ; + r. brandenberger , d. a. easson and a. mazumdar , phys . d * 69 * , 083502 ( 2004 ) [ arxiv : hep - th/0307043 ] .

according to the standard prescriptions , zero - temperature string theories can be extended to finite temperature by compactifying their time directions on a so - called `` thermal circle '' and implementing certain orbifold twists . however , the existence of a topologically non - trivial thermal circle leaves open the possibility that a gauge flux can pierce this circle _ i.e. _ , that a non - trivial wilson line ( or equivalently a non - zero chemical potential ) might be involved in the finite - temperature extension . in this paper , we concentrate on the zero - temperature heterotic and type i strings in ten dimensions , and survey the possible wilson lines which might be introduced in their finite - temperature extensions . we find a rich structure of possible thermal string theories , some of which even have non - traditional hagedorn temperatures , and we demonstrate that these new thermal string theories can be interpreted as extrema of a continuous thermal free - energy `` landscape '' . our analysis also uncovers a unique finite - temperature extension of the heterotic @xmath0 and @xmath1 strings which involves a non - trivial wilson line , but which like the traditional finite - temperature extension without wilson lines is metastable in this thermal landscape . .7ex .7ex # 1*#1 * = cmss10 = cmss10 at 7pt caption#1[#2]#3 @=12 epsf

introduction and motivation strings at finite temperature wilson lines and imaginary chemical potentials surveying possible wilson lines wilson lines and the hagedorn temperature discussion and conclusions acknowledgments useful trace formulae

the semiconductor microcavity structure studied here is a planar , strain compensated 2@xmath103 gaas microcavity with embedded ingaas quantum wells ( qws ) . strain compensation was achieved by alas@xmath104p@xmath105/gaas dbr layers instead of the thin alp inserts in the alas layers used in ref . [ ] as their effective composition could be better controlled . the bottom dbr consists of 26 pairs of gaas and alas@xmath104p@xmath105 while the top has 23 of these pairs , resulting in very high reflectance ( > 99.9@xmath106 ) in the stop - band region of the spectrum . the average density of hatches along the [ @xmath107 direction was estimated from transmission imaging to be about 6/mm , while no hatches along the [ @xmath108 direction were observed . three pairs of 6 nm in@xmath109ga@xmath110as qws are embedded in the gaas cavity at the anti - nodes of the field as well as two additional qws at the first and last node to serve as carrier collection wells . the large number of qws was chosen to increase the rabi splitting and keep the exciton density per qw below the mott density @xcite also for sufficiently high polariton densities to achieve polariton condensation under non - resonant excitation . the strong coupling between the exciton resonance and the cavity mode is observed with a vacuum rabi - splitting of @xmath111mev . a wedge in the cavity thickness allows access to a wide range of exciton - cavity detuning . all measurements reported here are taken at @xmath112mev . the measured q - factor is @xmath113 , while the calculated bare cavity q - factor , neglecting in - plane disorder and residual absorption , is @xmath114 . as the emission energy of the ingaas qws is lower than the absorption of the gaas substrate we can study the photoluminescence of the sample both in reflection and transmission geometry . the transmission geometry , which is not available for gaas qws , allows to filter the surface reflection of the excitation , and has been widely utilized to probe the features of polariton fluids @xcite under resonant excitation of polaritons . using real and reciprocal space spectroscopic imaging under non - resonant optical excitation , polariton condensation , and a second threshold marking the onset of photon lasing , i.e. the transition from the strong to the weak - coupling regime has been studied in this microcavity @xcite . in the experiments described here the sample was held in a cold finger cryostat at a temperature of @xmath115k . continuous wave excitation is provided by a ti : sapphire laser . we use non - resonant excitation from the epi side , and detect the emission from the substrate side , so that the excitation is filtered by the absorption of the gaas substrate . the optical excitation , for all the measurements reported in this work , is at the first reflectivity minimum above the cavity stop band . the spatial profile of the excitation beam is modulated to a graph with gaussian profiles at each vertex of approximatelly equal in diameter spots using a reflective spatial light modulator ( slm ) . we use a high numerical aperture microscope objective ( na = 0.65 ) to focus the spatially modulated beam to @xmath1161 - 2@xmath117 in diameter at full width at half maximum ( fwhm ) excitation spots . the photoluminescence from the sample is collected in transmission geometry with @xmath118 collection angle , by a 0.42 na microscope objective . fourier ( dispersion ) imaging is performed by projecting the fourier - space at the slit of a @xmath119 mm spectrophotometer coupled a cooled charge coupled ( ccd ) device and using a @xmath120 grooves / mm with 50 @xmath16ev energy - resolution . the real - space tomography images are acquired with sub - micron optical resolution using a ccd camera imaging configuration through a tunable fabry - perot etalon with @xmath79 20 @xmath16ev fwhm bandwidth . the condensate wavevector , @xmath60 , of the polariton ising chains is measured using two dimensional fourier - space tomography utilising a tunable fabry - perot etalon with @xmath79 20 @xmath16ev fwhm bandwidth and a ccd camera imaging configuration . figure 1(a , b ) in supp . mat . shows the false - colour normalised photoluminescence intensity of the two dimensional cross - section of the fourier - space at the energy of the condensate from the ising chain configuration of fig.2(b ) and fig.2(c ) respectively at condensation threshold . the outer ring in both images corresponds to @xmath60 , whereas the inner fringes correspond to self - diffraction from the ising chain . for the anti - ferromagnetic ising chain of fig.2(a ) , @xmath1211.79@xmath122 and for the ferromagnetic ising chain of fig.2(c ) , @xmath1211.67@xmath122 . the expression for the coupling coefficients @xmath6 given by eq . ( 8) can be estimated based on the width of the hankel transformation of the wavefunction of an individual condensate given by @xmath123j_0(kr)r\,dr . \label{a1}\ ] ] the density of the hankel transformation , @xmath124 , peaks at @xmath125 with the width , characterized by @xmath126 , inversely proportional to the width of the condensate density @xmath127 , which is set by the width of the pumping profile @xmath128 . we , therefore , approximate @xmath124 by @xmath129 we integrate eq . ( 8) using eq . ( [ a2 ] ) to get @xmath130 . \label{jj } \end{aligned}\ ] ] in the limit of @xmath131 we recover our @xmath132function approximation @xmath133 . the finite width of the hankel transformation of the condensate wavefunction , as seen from eq . [ jj ] , induces a phase shift , so the criterion for the phase switching can be approximated by the sign switching of @xmath134 , where @xmath83 is the system parameter dependent term . we find the global minimum of the xy hamiltonian directly for the sample configurations considered in our paper . for the lattice sites arranged in a square the phases relative to one fixed phase that we set equal to zero , @xmath135 , minimize the xy hamiltonian @xmath136 where we denoted @xmath137 to be the ratio of the coupling of the diagonal cites to the coupling of the neighboring sites . the coupling strength decays with the distance between sites , therefore , @xmath138 . if all couplings are ferromagnetic , @xmath139 , the minimum of @xmath140 is for @xmath141 . if @xmath142 , there is a @xmath59 phase difference between the neighboring sites @xmath143 even for nonzero @xmath137 ( as long as @xmath138 is satisfied ) . for a rhombus , consisting of two equilaterial triangles , the xy hamiltonian becomes @xmath144 where we associated @xmath135 with one of the sites along the shorter diagonal . @xmath137 in this case represents the ratio of the coupling along the long diagonal to that between the neighbours . while for an equilaterial triangle the xy hamiltonian @xmath145 , @xmath142 is minimized by @xmath146 , the xy hamiltonian ( [ hrhombus ] ) is minimized by @xmath147 , @xmath148 as in the case of the square . for three equilaterial triangles non - trivial winding around sites is again possible , since the xy hamiltonian @xmath149 for @xmath142 reaches its minimum at @xmath150 , therefore , creating an alternating winding around each of the equilaterial triangles . here we associated @xmath135 with the site that is connected to all other sites and neglected the interactions along two long diagonals . if the distances are close to the switching points between ferro- and antiferro- couplings the small deviation in the position of the sites may lead to an even more complex configurations as is illustrated on fig . 5 of the main text . in our numerical simulation we used a gaussian pumping profile that produces the same width of the condensate as in experiment ( fwhm @xmath151 ) and choose the pumping intensity to obtain the correct outflow wavenumber for a single condensate . the common integration parameters used for all numerical simulations are , therefore , @xmath152 the numerical simulations were performed for various geometries and distances as the main text shows . for fig . 3 we varied the distances between the nearest neighbors , so that @xmath153 for figs . 3a , b , c , d respectively . for fig . 4 we used @xmath154 for figs . 4a , b , c respectively . for fig.5 we used @xmath155 .

several platforms are currently being explored for simulating physical systems whose complexity increases faster than polynomially with the number of particles or degrees of freedom in the system . defects and vacancies in semiconductors or dielectric materials @xcite , magnetic impurities embedded in solid helium @xcite , atoms in optical lattices @xcite , photons @xcite , trapped ions @xcite and superconducting q - bits @xcite are among the candidates for predicting the behaviour of spin glasses , spin - liquids , and classical magnetism among other phenomena with practical technological applications . here we investigate the potential of polariton graphs as an efficient simulator for finding the global minimum of the @xmath0 hamiltonian . by imprinting polariton condensate lattices of bespoke geometries we show that we can simulate a large variety of systems undergoing the u(1 ) symmetry breaking transitions . we realise various magnetic phases , such as ferromagnetic , anti - ferromagnetic , and frustrated spin configurations on unit cells of various lattices : square , triangular , linear and a disordered graph . our results provide a route to study unconventional superfluids , spin - liquids , berezinskii - kosterlitz - thouless phase transition , classical magnetism among the many systems that are described by the @xmath0 hamiltonian . many properties of strongly correlated spin systems , such as spin liquids and unconventional superfluids are difficult to study as strong interactions between @xmath1 particles become intractable for @xmath1 as low as @xmath2 @xcite . feynman envisioned that a quantum simulator a special - purpose analogue processor could be used to solve such problems @xcite . it is expected that quantum simulators would lead to accurate modelling of the dynamics of chemical reactions , motion of electrons in materials , new chemical compounds and new materials that could not be obtained with classical computers using advanced numerical algorithms @xcite . more generally , quantum simulators can be used to solve hard optimization problems that are at the heart of almost any multicomponent system : new materials for energy , pharmaceuticals , and photosynthesis , among others @xcite . many hard optimisation problems do not necessitate a _ quantum _ simulator as only recently realised through a network of optical parametric oscillators ( opos ) that simulated the ising hamiltonian of thousands of spins @xcite . the ising model corresponds to the @xmath3 case of the @xmath1-vector model of classical unit vector spins @xmath4 with the hamiltonian @xmath5 , where @xmath6 is the coupling between the sites labelled @xmath7 and @xmath8 . for @xmath9 the @xmath1-vector hamiltonian becomes the @xmath0 hamiltonian @xmath10 , where we have parameterized unit planar vectors using the polar coordinates @xmath11 . since @xmath12 is invariant under rotation of all spins by the same angle @xmath13 the xy model is the simplest model that undergoes the @xmath14 symmetry - breaking transition . as such , it is used to emulate berezinskii - kosterlitz - thouless phase transition and the emergence of a topological order @xcite , topological quantum information processing and storage @xcite , and to study quantum phase transitions , unconventional superfluids , quantum spin models , spin - liquid phases and high-@xmath15 superconductivity . the @xmath0 hamiltonian has been simulated on a triangular lattice of atomic condensates investigating a variety of magnetic phases and frustrated spin configurations @xcite . whereas optical lattices offer a scalable platform , they are likely to reach a local rather than global minimum of the hamiltonian and are limited to sub-@xmath16k temperatures @xcite . in this article , we propose and experimentally demonstrate the use of polariton graphs as a scheme for finding the global minimum of the @xmath0 hamiltonian . polaritons are the mixed light - matter quasi - particles that are formed in the strong exciton - photon coupling regime in semiconductor microcavities @xcite . under non - resonant optical excitation , rapid relaxation of carriers and bosonic stimulation result in the formation of a non - equilibrium polariton condensate characterized by a single many - body wave - function @xcite . polariton condensates can be imprinted into any two - dimensional graph by spatial modulation of the pumping source , offering the scalability matched only by optical lattices @xcite . optically injected polariton condensates can potentially be imprinted in multi - site configurations with arbitrary polarisation and density profiles offering the possibility to control the separation distance between sites . such flexibility allows for unprecedented control of the interaction between neighbouring sites . due to the finite cavity lifetime , polaritons decay in the form of photons ( through quasi - mode coupling @xcite ) that carry all information of the corresponding polariton state ( energy , momentum , spin and phase ) . the continuous coupling of polaritons to free photons allows for the in - situ characterisation of static polariton graphs , but more importantly it also allows for the dynamic control of an arbitrary set of sites , whilst measuring in real time the kinetics and phase configuration of the modulated polariton graph . in a graph of two or more coupled polariton vertices , with increasing excitation density , polariton condensation occurs at the state with the phase - configuration that carries the highest polariton occupation @xcite . this is due to the bosonic character of the condensate formation : the probability of a particle to relax in a particular state grows with the population of that state . just above condensation threshold a macroscopic coherent state is formed described by the wavefunction @xmath17 . @xmath17 can be written as a superposition of the wavefunctions @xmath18 at the sites @xmath19 with phase @xmath20 ; that is @xmath21 $ ] . below we will show that the system of an arbitrary polariton graph condenses into the global minimum of the @xmath0 hamiltonian : @xmath22 where @xmath23 is the phase difference between two sites , @xmath24 and @xmath6 is the corresponding coupling strength ; the latter depends on the density of the sites @xmath7 and @xmath8 , the distance between them , @xmath25 , and the outflow condensate wavenumber @xmath26 , which under non - resonant optical excitation depends on the pumping intensity and profile . the bottom - up approach for the search of the global minimum of the @xmath0 hamiltonian is achievable within the linewidth of the corresponding state similarly to a network of time - multiplexed opos @xcite that guarantees a phase - transition to the global minimum of the ising hamiltonian . this is an advantage over classical or quantum annealing techniques , where the global ground state is reached through transitions over metastable excited states ( local minima ) , with an increase of the cost of the search with the size of the system . the interactions are either ferromagnetic ( solid - blue lines ) or anti - ferromagnetic ( dashed - red lines ) . at each vertex @xmath27 of the graph polaritons have a local phase @xmath28 that is mapped to a classical vector spin @xmath29 . ( b ) the vertices ( blue solid - circles ) and edges of the polariton density map depicted in ( a ) , showing the sign of the coupling and the spin vector @xmath4 of each vertex.,width=325 ] _ modelling the phase coupling : _ we model the phase coupling in polariton graphs using the complex ginzburg - landau equation ( cgle ) with a saturable nonlinearity and energy relaxation @xcite : @xmath30 where @xmath31 is the condensate wavefunction , @xmath32 is the density profile of the hot exciton reservoir , @xmath33 is the polariton effective mass , @xmath34 and @xmath35 are the strengths of effective polariton - polariton interaction and the blue - shift due to interactions with non - condensed particles , respectively , @xmath36 is the rate at which the exciton reservoir feeds the condensate , @xmath37 is the decay rate of condensed polaritons , @xmath38 is the rate of redistribution of reservoir excitons between the different energy levels , @xmath39 is the energy relaxation coefficient specifying the rate at which gain decreases with increasing energy , and @xmath40 is the pumping into the exciton reservoir . in eq . ( [ initial reservoir equation ] ) we neglected the diffusion of the reservoir as well as density - density repulsion with the condensate in the view of the large mass of the hot exciton as compared to the mass of the polariton ( five orders of magnitude ) . we non - dimensionalize these equations using @xmath41 and introducing the notations @xmath42 @xmath43 , @xmath44 and @xmath45 we choose @xmath46 and consider the stationary states . by using the madelung transformation @xmath47 $ ] in the dimensionless eqs . ( [ initial gl equation],[initial reservoir equation ] ) , where @xmath48 , @xmath49 is the velocity , @xmath50 is the phase and separating the real and imaginary parts we obtain the mass continuity and the integrated form of the bernoulli equation which we write for a steady state , and , therefore , introduce the chemical potential @xmath16 @xmath51 first , we consider a single pumping spot with a radially symmetric pumping profile . asymptotics at large distances from the center of the pump gives the velocity @xmath52 and @xmath53r^{-1}.$ ] from eq . ( [ systemm1 ] ) at infinity , therefore , we obtain @xmath54 . we can estimate the chemical potential for a wide pumping spot so that the quantum pressure term @xmath55 and @xmath56 are insignificant at the pumping center . under this assumption @xmath57 and @xmath58 in @xcite we established experimentally for the pulsed excitation that the coupling between two pumping spots ( apolariton dyad " ) can be either in - phase or with a @xmath59 phase difference depending on the outflow wavenumber @xmath60 and the distance between the spots . below , in the steady state excitation regime , we obtain a general criterion for the switching between the relative phases . we start by considering the wavefunction of the condensate as the sum of the wavefunctions of individual condensates located at @xmath61 , where @xmath62 with the phase difference @xmath23 : @xmath63 the total number of condensed polaritons can be found in fourier space as @xmath64,\ ] ] where @xmath65j_0(kr)r\,dr $ ] is the hankel transform of the wavefunction of an individual condensate . we conclude that @xmath66 which in the case of @xmath1 condensates generalizes to @xmath67 the oscillating behaviour of the bessel function , @xmath68 , brings about the sign change in the coupling constants @xmath6 depending on the distance @xmath69 . when @xmath6 is positive the coupling is said to be ferromagnetic and when @xmath6 is negative the coupling is said to be anti - ferromagnetic . the state with the phase configuration that carries the highest number of particles in eq . ( [ nnn ] ) corresponds to the solution that minimises the @xmath0 hamiltonian , @xmath70 . between any two polariton nodes the polariton wavefunction forms a standing wave with the density @xmath71 $ ] , where @xmath72 is the coordinate along the line that connects the two nodes separated by a distance @xmath69 and @xmath73 . between two polariton nodes the density oscillates as @xmath74 , from which the phase difference @xmath23 of a single shot realization can be extracted directly . in fig . 1(a ) we plot the density of a polariton graph , where for simplicity we have annotated the sign of the coupling for some of the edges of the graph . depending on the separation distance between the vertices and the outflow wavevector @xmath60 the interactions are either ferromagnetic ( solid - blue lines ) or anti - ferromagnetic ( dashed - red lines ) . at each vertex @xmath27 of the graph polaritons have a local phase @xmath28 , which in the following we map to a classical vector spin @xmath29 . in fig . 1(b ) we show the vertices of the polariton graph , the edges of fig.1(a ) depicting the sign of the coupling and the spin vector @xmath4 of each vertex as calculated from the minimisation of the @xmath0 hamiltonian . , of a polariton condensate dyad formed under incoherent pumping of two nodes as the function of the product @xmath75 between the nodes obtained by numerical integration of the cgle for a fixed @xmath60 . the solid black line corresponds to the maximum number of particles in the in - phase ferromagnetic configuration and the dashed black line to the @xmath59-phase difference anti - ferromagnetic configuration . the switching occurs with the periodicity @xmath76 as the superimposed graph of @xmath77 illustrates in red , where @xmath78 . ( b , c ) experimental realization of an ising chain of five equidistant polariton nodes with lattice constants of @xmath79 6.7@xmath80 and @xmath799.6@xmath80 respectively . the false - grey scale images show the normalised photoluminescence intensity of the real - space tomography at the energy of the condensate ; ( c ) is saturated at 0.5 to increase the visibility of the low intensity fringes between the nodes . the corresponding @xmath75 ( 11.9 and 16.04 ) are shown by two solid circles on ( a ) . , title="fig:",width=325 ] + _ the ising polariton chain : _ we theoretically describe and experimentally address the minimization of the @xmath0 hamiltonian for the simple case of a linear polariton chain with equal spacing @xmath81 between neighbours . for a given @xmath60 with increasing separation distance the coupling between the neighbors , @xmath6 , oscillates between negative and positive values . we approximate the switching of the coupling sign with @xmath82 , where @xmath83 is fixed by the system parameters ( see supp . mat . for the derivation ) . in the steady state excitation regime , we can calculate the maximum particle number of a polariton dyad as a function of the separation distance @xmath84 by numerically integrating the cgle to find the solutions of eqs . ( [ systemm1]-[systemm2 ] ) for a given pumping profile @xmath85 $ ] of a characteristic width @xmath86 ; the results are shown in fig . 2(a ) . the relative phases that realise the maximum particle number switch periodically between @xmath87 and @xmath59 with the period @xmath76 as shown by superimposing the function @xmath88 in fig.2(a ) ; we have used the experimental parameters for the pumping profile and @xmath60 as described in `` wavevector tomography '' in supp . mat . where the coupling is ferromagnetic ( anti - ferromagnetic ) the graph of the maximum number of particles is plotted with a solid ( dashed ) line . we experimentally address the ising chain by injecting a linear chain of five equidistant polariton nodes through non - resonant , continuous wave and spatially modulated optical excitation of a multiple ingaas quantum well semiconductor microcavity that allows for detection of the polariton photoluminescence in the transmission geometry ( for the sample description read the microcavity sample " and for the description of the excitation / detection scheme read the experimental setup " in supp . mat . ) . figures 2(b , c ) show the real - space tomography of the photoluminescence intensity at the energy of the condensate from the linear chain with lattice constants of @xmath79 6.7@xmath80 and @xmath799.6@xmath80 respectively at condensation threshold . the relative phase difference realised between neighbours in the chain is either @xmath59 or zero . the patterns are clearly distinguishable by the number of fringes ( density maxima ) between the sites : zero or even for anti - ferromagnetic and odd for ferromagnetic coupling . in fig.2(a ) we have annotated the abscissa with solid circles for each of the two separation distances from which the expected sign of coupling is depicted showing good agreement with the experiment . the observed phase configurations realise the ferromagnetic and anti - ferromagnetic ising spin chain of the @xmath0 model . , the real - space energy tomography of the experimental realisations , and the averaged condensate densities of the numerically simulated condensate wavefunctions for several realizations are shown on the left , central and right columns respectively . solid and dashed blue lines on the spin vector diagrams ( left column ) indicate ferromagnetic and anti - ferromagnetic coupling , respectively . the false - grey scale images of the middle column show the normalised photoluminescence intensity of the real - space tomography at the energy of the condensate ; ( c ) is saturated at 0.5 to increase the visibility of the low intensity fringes between the vertices . the configurations shown are some elementary building blocks of square lattices such as ( a , c ) anti - ferromagnetic , ( b ) ferromagnetic , ( d ) @xmath89-compass . the centers of the pumping spots are shown by white dashed circles on the numerical density profiles ( right column ) . the parameters of the numerical simulations of eqs . ( [ initial gl equation],[initial reservoir equation ] ) are listed in the supp . mat.,width=325 ] _ equidistant vertices across a circle : _ we consider a geometry of @xmath1 incoherently pumped equidistant polariton vertices positioned on the circumference of a circle . for equal separation distances @xmath81 between adjacent sites the @xmath0 hamiltonian to minimise becomes @xmath90 where @xmath91 , the summation is cyclic and we took into account only nearest neighbour interactions . if @xmath92 is positive , then all sites lock in phase ( @xmath93 ) . if @xmath92 is negative , the minimum of @xmath12 occurs for @xmath94 , when @xmath1 is even and for @xmath95 when @xmath1 is odd ( @xmath96 ) . we experimentally access these two regimes through incoherent injection of polaritons at the vertices of a square ; figure 3(a , b , c ) show the spin configuration , experimental results of the real - space tomography of the photoluminescence intensity at the energy of the condensate at condensation threshold and numerical simulation for a square with lattice constants that lead to anti - ferromagnetic , ferromagnetic and the next anti - ferromagnetic coupling respectively . similar to the ising polariton chain the type of coupling is clearly distinguishable by the number and symmetry of fringes between the vertices : zero or even for anti - ferromagnetic ( fig.3(a , c ) ) and odd for ferromagnetic coupling ( fig.3(b ) ) . these observations are in agreement with the @xmath59 phase difference reported in ref.@xcite . we can thus summarise in the case of the square lattice cell that for ferromagnetic coupling polaritons at the vertices lock with zero phase difference and for anti - ferromagnetic coupling polaritons at neighbouring vertices lock with a @xmath59 phase difference . _ @xmath89 compass model : _ in the context of topological quantum computing apart from the trivial all ferromagnetic or all anti - ferromagnetic coupling configurations in a square geometry , more complex coupling configurations are of interest . examples of such configurations are the compass models , where the coupling between the internal spin components is inherently directionally dependent . such compass - type coupling appears in various physical systems , where the interactions are sensitive to the spatial orientation of the involved orbitals . in polariton graphs the compass models with direction dependent coupling or spin glassy models with random couplings can be realised by changing the pumping intensity and preserving the square geometry , or alternatively , tuning the separation distances so that each vertex has one ferromagnetic and one anti - ferromagnetic coupling with its nearest neighbours . in fig . 3(c ) we have realised the @xmath89 compass model , where each vertex has one ferromagnetic and one anti - ferromagnetic coupling with its neighbours as it is clearly distinguishable by the number of fringes between nearest vertices . the @xmath89 compass , where both ferro- and anti - ferromagnetic coupling appear across the two orthogonal diagonals here , has been proposed as a model to mott insulators with orbital degrees of freedom and frustrated magnets @xcite . other compass - type models accessible through polariton graphs include the plaquette orbital model , where the ferromagnetic and anti - ferromagnetic coupling alternate along each direction @xcite and the orbital compass model on a checkerboard lattice @xcite . fully random couplings in the square lattice describes the thermodynamic behaviour of several disordered systems , such as magnetic systems with random dzyaloshinskii- moriya interactions @xcite , disordered josephson junction arrays @xcite , disordered substrates @xcite , and vortex glasses in high-@xmath15 cuprate superconductors @xcite . it should be possible to address these systems by considering a polariton lattice composed of the individual compass elements analogous to the one we realised here . _ triangular lattice : _ the @xmath0 hamiltonian has been simulated on a triangular lattice of atomic condensates discovering variety of magnetic phases and frustrated spin configurations @xcite . in the case of an anti - ferromagnetically coupled polariton triad , arranged at the vertices of an equidistant triangle , the energy flux that minimizes the @xmath0 hamiltonian corresponds to @xmath97 winding ( @xmath98 phase difference between the condensates ) @xcite . here , we experimentally realise an equidistant triangular lattice of two lattice cells ( rhombus configuration ) under incoherent injection of polaritons in the bespoke lattice configurations . figure 4(a , b , c ) show the spin configuration , experimental results of the real - space tomography of the photoluminescence intensity at the energy of the condensate at condensation threshold and numerical simulation for a rhombus with lattice constants that lead to anti - ferromagnetic , ferromagnetic and the next anti - ferromagnetic coupling respectively . in the case of ferromagnetic coupling between nearest neighbours and neglecting opposite neighbours interaction across the long diagonal axis of the rhombus , the @xmath0 hamiltonian is minimised at @xmath99 when all polariton sites lock in phase , as shown in fig . 4(b ) . similarly , in the case of anti - ferromagnetic coupling between nearest neighbours the @xmath0 hamiltonian is minimised at @xmath100 when there is @xmath101 phase difference between the outer edges of the rhombus . this configuration forces the rhombus in a frustrated state wherein opposite vertices have the same phase . this type of frustrated spin configuration is experimentally realised in fig . 4(a , c ) . the corresponding states in figs . 4(a , b , c ) are shown in the order of the increasing distance between the sites , therefore , the anti - ferromagnetic states of figs . 4(a ) and 4(c ) belong to two different bands of anti - ferromagnetic regions separated by a ferromagnetic band ( the alternating anti - ferromagnetic / ferromagnetic couplings bands are shown in fig . 2(a ) ) . the measured density profiles show some clear differences : the local minimum at the center of the rhombus along the long diagonal in fig . 4(a ) is replaced by a local maximum in fig . 4(c ) . _ random polariton graph : _ beyond the minimization of the @xmath0 hamiltonian of polariton condensates on regular lattices we test our platform on a disordered polariton graph of five vertices . we took a graph initially consisting of three equidistant triangular unit cells for a lattice constant that leads to anti - ferromagnetic coupling , but with one spot breaking the symmetry . this is achieved experimentally by slightly displacing one spot on the graph . figure 5 shows the spin configuration , experimental results of the real - space tomography of the photoluminescence intensity at the energy of the condensate at condensation threshold and numerical simulations that correspond to this graph . for the symmetric configuration of three equidistant triangular cells and considering only nearest neighbours interactions , the @xmath0 hamiltonian is minimised at @xmath102 with an alternating winding around each cell slightly deviating from @xmath98 difference reported for a single equilateral triangle ( see supp . mat . for details ) . breaking the symmetry leads to a different phase distribution , while maintaining the winding around each cell . the analysis of the fringes on the experimental image ( with the different rows of local maxima along the two long diagonals ) shows that the symmetry is explicitly broken . in conclusion , we propose polariton graphs as an analog platform for minimizing the @xmath0 hamiltonian and experimentally demonstrate its experimental implementation for simple building blocks of lattices . we demonstrated that the search for the global ground state of a polariton graph is equivalent to the minimisation of the @xmath0 hamiltonian @xmath22 . polariton graphs offer the scalability of optical lattices , together with the potential to study disordered systems , and to control both the sign and the strength of the coupling for each edge independently . similar to networks of time - multiplexed opos , phase transitions in polariton graphs occur at the global ground state . furthermore , polariton graphs offer the potential to quench either all or individual vertices on variable time scales and study the complex relaxation dynamics . with the recent advances in the field of polariton condensates , such as room temperature operation @xcite and condensation under electrical pumping @xcite , polariton graph based simulators offer unprecedented opportunities in addressing the @xmath0 hamiltonian and therefore topological quantum information processing and the study of exotic phase transitions . 00 pla , j. j. _ et al . _ high - fidelity readout and control of a nuclear spin qubit in silicon . _ nature _ * 489 * , 541 - 545 ( 2013 ) . hanson , r. & awschalom , d. d. coherent manipulation of single spins in semiconductors . _ nature _ * 453 * , 1043 - 1049 ( 2008 ) . lemeshko , m. _ et al . _ controllable quantum spin glasses with magnetic impurities embedded in quantum solids . _ phys . rev . b _ * 88 * , 014426 ( 2013 ) . saffman , m. , walker , t. g. & molmer , k. quantum information with rydberg atoms . _ rev . mod . phys . _ * 82 * , 2313 ( 2010 ) . simon , j. _ et al . _ atomic physics : a route to quantum magnetism . _ nature _ * 472 * , 307 - 312 ( 2011 ) . northup , t. e. & blatt , r. quantum information transfer using photons . _ nature photonics _ * 8 * , 356 ( 2014 ) . kim , k. _ et al . _ quantum simulation of frustrated ising spins with trapped ions . _ nature _ * 465 * , 590 - 593 ( 2010 ) . lanyon , b. p. _ et al . _ universal digital quantum simulation with trapped ions . _ science _ * 334 * , 57 , ( 2011 ) . corcoles , a. d. _ et al . _ demonstration of a quantum error detection code using a square lattice of four superconducting qubits . _ nature commun . _ * 6 * , 6979 ( 2015 ) . sandvik , a. w. ground states of a frustrated quantum spin chain with long - range interactions . _ phys . rev . lett . _ * 104(13 ) * , 137204 ( 2010 ) . feynman , r. simulating physics with computers . _ int . j. theor . phys . _ * 21 * , 467488 ( 1982 ) . lloyd , s. universal quantum simulators . _ science _ * 273 * , 1073 ( 1996 ) . boixo , s. _ et al . _ evidence for quantum annealing with more than one hundred qubits . _ nature physics _ , * 10 * , 218 ( 2014 ) . utsunomiya , s. , takata , k. & yamamoto , y. mapping of ising models onto injection - locked laser systems . _ opt . express _ * 19 * , 18091 ( 2011 ) . marandi , a. , wang , z. , takata , k. , byer , r. l. & yamamoto , y. network of time - multiplexed optical parametric oscillators as a coherent ising machine . _ nature photonics _ * 8 * , 937 - 942 ( 2014 ) . berezinskii , v. l. destruction of long - range order in one - dimensional and two - dimensional systems having a continuous symmetry group i. classical systems . _ sov . . jetp _ * 32 * , 493 ( 1971 ) . kosterlitz , j. m. & thouless , d. j. ordering , metastability and phase transitions in two - dimensional systems . _ j. phys . c _ * 6 * , 1181 ( 1973 ) . nayak , c. , simon , s. h. , stern , a. , freedman , m. & sarma , s. d. non - abelian anyons and topological quantum computation . _ rev . mod . phys . _ * 80 * , 1083 ( 2008 ) . struck , j. quantum simulation of frustrated classical magnetism in triangular optical lattices . _ et al . _ _ science _ * 333 * , 996 ( 2011 ) . georgescu , i. m. , ashhab , s. & nori , f. quantum simulation . _ rev . mod . phys . _ * 86 * , 153 ( 2014 ) . weisbuch , c. , nishioka , m. , ishikawa , a. & arakawa , y. observation of the coupled exciton - photon mode splitting in a semiconductor quantum microcavity . _ phys . rev . lett . _ * 69 * , 3314 ( 1992 ) kasprzak , j. _ et al . _ bose - einstein condensation of exciton polaritons . _ nature _ * 443 * , 409 ( 2006 ) . amo , a. _ et al . _ light engineering of the polariton landscape in semiconductor microcavities . _ phys . rev . b _ * 82 * , 081301(r ) ( 2010 ) . ciuti , c. , schwendimann , p. , deveaud , b. & quattropani , a. theory of the angle - resonant polariton amplifier . _ phys . rev . b _ * 62 * , r4825(r ) (2000 ) . ohadi , h. _ et al . _ non - trivial phase coupling in polariton multiplets . in press by phys . rev . x. wouters , m. & carusotto , i. excitations in a nonequilibrium bose - einstein condensate of exciton polaritons . _ phys . rev . lett . _ * 99 * , 140402 ( 2007 ) . keeling , j. & berloff , n. g. spontaneous rotating vortex lattices in a pumped decaying condensate . _ phys . rev . lett . _ * 100 * , 250401 ( 2008 ) . tosi , g. _ et al . _ geometrically locked vortex lattices in semiconductor quantum fluids . _ nat . comm . _ , * 3 * , 1243 ( 2012 ) . nussinov , z. & van den brink , j. compass models : theory and physical motivations . _ rev . mod . phys . _ * 87 * , 1 ( 2015 ) . biskup , m. & kotecky , r. true nature of longrange order in a plaquette orbital model . _ j. stat .. mech . _ * 11 * , 11001 ( 2010 ) . nasu , j. and ishihara , s. orbital compass model as an itinerant electron system . _ europhysics letters _ * 97 * , 27002 ( 2012 ) . rubinstein , m. , shrainam , b. & nelson , d. r. two - dimensional xy magnets with random dzyaloshinskii - moriya interactions . _ phys . rev . b _ * 27 * , 1800 ( 1983 ) . granato , e. & kosterlitz , j. m. disorder in josephson - junction arrays in a magnetic field . _ phys . rev . lett . _ * 62 * , 823 ( 1989 ) . cha , m.c . & fertig , h. a. orientational order and depinning of the disordered electron solid . _ phys . rev . lett . _ * 73 * , 870 ( 1994 ) . gingras , m. j. p. & huse , d. a. topological defects in the random - field xy model and the pinned vortex lattice to vortex glass transition in type - ii superconductors . _ phys . rev . b _ * 53 * , 15193 ( 1996 ) . lucas , a. ising formulations of many np problems . _ frontiers in physics _ * 2 * , 5 ( 2014 ) . plumhof , j. d. , stferle , t. , mai , l. , scherf , u. & mahrt , r. f. room - temperature bose - einstein condensation of cavity exciton - polaritons in a polymer . _ nat . mat . _ * 13 * , 247 ( 2014 ) . schneider , c. _ et al . _ an electrically pumped polariton laser . _ nature _ * 497 * , 348 ( 2013 ) . * supplemental material : realizing the @xmath0 hamiltonian in polariton simulators *

microcavity sample experimental setup wavevector tomography finding the expression for the coupling coefficients minimization of the xy hamiltonian for sample configurations parameters of the numerical simulations

it has been observed that the momentum distribution of particles created in heavy ion collisions is not azimuthally symmetric in the plane perpendicular to the beam direction . this dominant asymmetry , the elliptic flow , is the result of several factors , the main one being the anisotropy of the initial configuration after the collision due to the non - zero impact parameter . the elliptic flow is typically characterized using the second coefficient of the fourier expansion of the momentum distribution , @xmath0 . it was found in experiments @xcite that the @xmath0 parameter as a function of the transverse momentum , @xmath1 , scales with the number of constituent quarks , @xmath2 , in a hadron . that is , if the @xmath3 curves are re - scaled according to the constituent quark number of the considered hadron species , and @xmath4 is plotted as function of @xmath5 for each type of hadron with mass @xmath6 , the curves will coincide . later results showed that the scaling is more precise if @xmath0 is plotted as a function of the transverse kinetic energy , @xmath7 , instead of the transverse momentum @xmath8 . this experimentally found scaling law is remarkably simple , and indicates that the elliptic flow develops before the quarks recombine into hadrons . therefore , understanding the factors influencing the measurable @xmath0 can provide information about the state of the quak gluon plasma ( qgp ) . the fluid dynamical ( fd ) model describes the dynamical development of the qgp from the ( already thermalized ) initial state until the break - down of the equilibrium , where first the chemical equilibrium among quarks and anti - quarks ceases . initially the plasma has only two flavors , u and d , and thus the fd model assumes two flavors . however , the flavor equilibration in the plasma is a rapid process with a timescale of the order of @xmath9 , so by the end of the fd development a flavor equilibrium among three flavors , u , d , and s , is reached . our usual fd evolution does not take into account this chemical change of flavors with an additional rate equation , instead this process is taken into account at the final stage of the fd development , when the subsequent eos is already assumed to have three equilibrated flavors . energy and momentum conservation and the requirement of non - decreasing entropy is enforced in the transition , from the ideal qgp state to the state where quark and anti - quark numbers are frozen out . as a consequence , the mass change of the quarks starts in the initial qgp with two flavors , and we use this approximation to estimate in section ii.b the final boundary where the initial fd stage of the evolution ends , with light quark masses . we consider a model of hadronization and investigate the constituent quark number scaling ( qns ) of the @xmath0 parameter . in this model , a gas of quarks and anti - quarks expands in a background field represented by the bag constant @xmath10 , which depends on density and temperature of the expanding fireball . initially this @xmath10-field includes the energy of the deconfined perturbative vacuum and of the gluon fields . as the system expands the deconfinement starts and the average @xmath10 decreases . furthermore , as the chiral symmetry breaking starts , the quarks gain mass . the quark mass is calculated as a function of the temperature and density of the matter . the quark gas expands rapidly while the quark mass increases . this process can be considered as a simple representation of the breaking chiral symmetry and deconfinement in a dynamical transition crossing the quarkyonic phase @xcite . the point during the expansion when the quarks recombine into hadrons is determined from the condition that the average hadron energy is equivalent to @xmath11 , as found from the systematics of experimental data @xcite . at recombination , the thermal and flow equilibrium between particles is broken . finally , the @xmath0 parameter is determined using simple two- and three - source models of the elliptic flow and the particle distributions obtained from the hadronization model . the paper is organized as follows : in section [ a ] the density and temperature dependence of constituent quark mass is considered . the energy , entropy , momentum and chemical potential of each individual source and of the total source is given in section [ b ] . in section [ c ] we describe the applied non - equilibrium model of rapid hadronization . details of calculation of elliptic flow parameter @xmath0 are presented in section [ d ] . a summary is given in section [ e ] . to present our arguments we consider the nambu - jona - lasinio model ( njl ) which is motivated by quantum chromodynamics ( qcd ) and is basically a purely quark - quark - interaction theory ; a comprehensive overview about njl has been presented in @xcite . the extended njl - model contains three types of quark - quark - interaction and is given by the lagrangian @xmath12 \nonumber\\ \nonumber\\ & & -\frac{g_v}{2}\sum\limits_{f = u , d , s}\sum\limits_{a=1}^{8}\left[\left(\overline{q}_f\,\gamma_{\mu}\,\lambda^a q_f\right)^2 + \left(\overline{q}_f \,\gamma_5\,\gamma_{\mu}\,\lambda^a q_f\right)^2 \right ] \nonumber\\ \nonumber\\ & & + \frac{g_d}{2 } \left [ { \rm det } \left ( \overline{q}_f ( 1 + \gamma_5 ) q_f \right ) + { \rm det } \left ( \overline{q}_f ( 1 - \gamma_5 ) q_f \right ) \right ] , \label{eq_12}\end{aligned}\ ] ] where @xmath13 , @xmath14 are the dirac matrices , @xmath15 is the current quark mass of flavor @xmath16 , @xmath17 is the coupling constant of the scalar - current interaction , @xmath18 is the coupling constant of the vector - current interaction terms and @xmath19 is the coupling constant of the determinantal flavor - mixing term ( determinant in flavor space ) . the gell - mann matrices are @xmath20 where @xmath21 with @xmath22 , where @xmath23 is the unit matrix . this lagrangian is used to derive the relation between constituent quark masses and chiral condensates , both for vacuum and for finite densities and temperatures . first , we consider the constituent quark mass in vacuum . for three flavors , @xmath24 , the constituent quark masses in vacuum @xmath25 where @xmath26 , are related to the ciral condensates in vacuum as follows @xcite @xmath27 here @xmath28 and so on , @xmath29 and @xmath30 are the current quark masses of the u - quark , d - quark and s - quark , respectively . typical values of the chiral condensates in vacuum are given by , e.g. @xcite and references therein : @xmath31 typical values for the coupling constants are @xmath32 , @xmath33 and @xmath34 @xcite . we note that in relations ( [ eq_23 ] ) - ( [ eq_25 ] ) the magnitude of terms proportional to @xmath19 are small compared to the terms proportional to @xmath17 . let us now consider the case of quarks in a hot and dense medium , at early times of fd evolution . at this stage we assume to have two flavors . then , the relations ( [ eq_23 ] ) - ( [ eq_25 ] ) for two flavors @xmath35 are reduced to @xmath36 the suffix @xmath37 at chiral condensate denotes the gibbs average over hadron states and meson states @xcite and references therein . note that for isospin - symmetric matter there is no difference in the density and temperature dependence of u - quark and d - quark condensate , that means @xmath38 . in the limit of high densities and temperatures the constituent quark mass , @xmath39 , approaches the current quark mass , @xmath15 . the well - known model - independent linear density dependence of the chiral condensate has widely been applied in many investigations , e.g. @xcite . the temperature dependence of the chiral condensate has been determined up to order @xmath40 in @xcite . in order to determine the density _ and _ temperature dependence of the chiral condensate we follow the arguments of refs . @xcite , where the first leading terms in the low - density low - temperature expansion have been obtained : @xmath41 the temperature and density dependence of chiral condensate has been plotted in ref . @xcite . the baryonic density in terms of quark degrees of freedom is given by @xmath42 where @xmath43 ( @xmath44 ) is the quark ( anti - quark ) density . the baryonic density in a given volume @xmath45 is related to the conserved baryon number @xmath46 by @xmath47 ; note the relation @xmath48 . for the logarithmic scale we take @xmath49 , for details see @xcite . the pion mass in vacuum is @xmath50 and the pion decay constant in vacuum is @xmath51 . the numerical value of the quark - sigma - term is @xmath52 ( see e.g. ref . @xcite ) , which is three times smaller than the nucleon - sigma - term @xmath53 . according to eq . ( [ eq_70 ] ) . ] by combining eq . ( [ eq_50 ] ) with eq . ( [ eq_60 ] ) , we obtain the expression for the in - medium mass of constituent - quarks : @xmath54 the temperature and density dependence of the constituent quark mass is plotted in fig . [ condensate ] . according to eq . ( [ eq_70 ] ) , for sufficiently high temperatures and densities the constituent quark mass will coincide with the current - quark - mass @xmath15 . of course , the applicability of eq . ( [ eq_70 ] ) is restricted to low densities and temperatures . in any case , the most upper limit for density and temperature is given by the condition that constituent quark mass has to be positive . the bag consists of quarks , anti - quarks , and the gluon fields . non - perturbative effects in the qcd vacuum and the energy density of gluon fields are described by the bag constant , @xmath10 . in our model , we subdivide the bag into @xmath55 individual cells , each of which moves with an individual flow velocity @xmath56 . the total energy of the bag in the center - of - mass frame of the colliding nuclei is given by the volume integral of the `` 00 '' component of the energy momentum tensor of each cell @xmath57 , @xmath58 . considering that our eos is given as a sum of the energy and pressure of the ideal quark and anti - quark gas plus the uniform bag energy density @xmath10 , that means @xmath59 and @xmath60 , the total energy of all cells of the bag in the center - of - mass frame of the colliding nuclei is given by @xmath61 where the sum runs over the number of all cells of the bag . here , we assume a uniform bag - field energy density @xmath10 over the whole volume , @xmath45 ; @xmath10 depends on density and temperature of the bag , thus on the time of the evolution of the fireball . we have used the notation @xmath62 for the invariant scalar , rest - frame energy density , and @xmath63 for the scalar pressure , @xmath64 is the volume of cell @xmath57 in the center - of - mass frame , and @xmath65 where @xmath66 is the 3-velocity of cell @xmath57 . in order to determine the rest - frame energy or proper energy of the cells , we assume a boltzmann - jttner distribution @xcite for the particles inside each individual cell : @xmath67 and @xmath68 is the normalization of the jttner distribution , where @xmath69 is the number of particles of a given type inside cell @xmath57 under consideration . the proper energy density is a lorentz invariant quantity by definition and , therefore , can be evaluated in any frame , e.g. in the local rest frame . accordingly , we obtain the following expression for the energy density of quarks and anti - quarks for each cell of the bag : @xmath70 , \label{eq_80 } \\ { \overline e}_q^i & = & \frac{1}{8\;\pi^2}\;\sum\limits_{f = u , d}\ , m_f^3\,t_i\,\exp \left(\frac{{\overline \mu}_q^i}{t_i}\right ) \nonumber\\ & & \times \left[k_1 \left(\frac{m_f } { t_i}\right ) + 3 k_3 \left(\frac{m_f } { t_i}\right)\right ] . \label{eq_85}\end{aligned}\ ] ] here , the density and temperature dependent constituent quark mass @xmath39 is given in ( [ eq_70 ] ) , and @xmath71 and @xmath72 are the baryonic density and temperature , respectively , of the cell @xmath57 . the initial value of the bag constant is abbreviated by @xmath73 , and we use the numerical value @xmath74 ; see also considerations in ref . @xcite . the total entropy in the center - of - mass lorentz frame ( cm ) is given by @xmath75 where the sum runs over all cells of the bag . the invariant scalar entropy density , @xmath76 , can be evaluated in any frame . the total entropy density consists of entropy density of the quarks and anti - quarks , and for the assumed jttner distribution is given by : @xmath77 , \label{entropy_11 } \\ { \overline s}_q^i & = & \frac{1}{2\pi^2 } \sum_{f = u , d } m_f^2 \exp\left(\frac{{\overline \mu}_q^i}{t_i}\right ) \nonumber\\ & & \hspace{-2.3 cm } \times \left[m_f\ , k_1\left(\frac{{\ensuremath{m_f}}}{t_i}\right ) + t_i \left(4 - \frac{{\overline \mu}_q^i}{t_i}\right ) k_2 \left(\frac{{\ensuremath{m_f}}}{t_i}\right ) \right ] . \label{entropy_12}\end{aligned}\ ] ] in what follows , we will take the adiabatic limit , which implies the total entropy to be a constant . we note that there is no bag constant in the expression for the total entropy because the bag - field is uniform , i.e. the entropy of bag - field vanishes . the total momentum of the bag in the center - of - mass lorentz - frame is given by @xmath78 where the sum runs over all cells of the bag . the scalar pressure of each individual cell of the bag is given by @xmath79 we note , that there is no bag constant in the expression for the total momentum ( [ total_momentum ] ) because it cancels out in the sum @xmath80 according to eqs . ( [ eq_75 ] ) and ( [ eq_96 ] ) . assuming a jttner distribution , the density @xmath81 of quarks having flavor @xmath16 and the density of anti - quarks @xmath82 having flavor @xmath16 in a given cell @xmath57 is determined by the following equations : @xmath83 in mechanical ( pressure ) , thermal and chemical equilibrium , the number of quarks depends only on baryonic chemical potential @xmath84 and temperature @xmath85 . especially , if the quarks and anti - quarks are in chemical equilibrium with each other , then @xmath86 and @xmath87 . in order to determine the baryonic chemical potential @xmath84 we have to use the total baryonic number density @xmath88 given in eq . ( [ baryonic_density_1 ] ) . then , according to the expressions ( [ chemical_potential_5 ] ) and ( [ chemical_potential_10 ] ) , we obtain a definition of the chemical potential @xmath89 of cell @xmath57 as @xmath90 where we have used @xmath91 . we assume that these thermodynamic relations hold in the ideal qgp until the final stage of the expansion starts . then the net baryon number in each cell @xmath92 and the number of quarks and anti - quarks , @xmath93 and @xmath94 , respectively , are given . while @xmath95 and thus @xmath96 and @xmath97 are changing during the expansion , the numbers of quarks and anti - quarks , @xmath98 and @xmath99 , are fixed at this point in each cell and do not change afterwards . this is the chemical freeze out in the model . the volumes , baryonic and quark densities and temperatures of each individual cell change during the expansion . the chemical equilibrium between the baryon density and quark- and anti - quark densities breaks down at this point . consequently , the chemical potentials for the single quarks are related to the single quark ( and anti - quark ) densities only and are not connected to the net baryon density . although we need the chemical potentials for the calculation of the energy density , pressure and entropy , these must be calculated from the given quark and anti - quark densities in each cell , by using eqs . ( [ chemical_potential_5 ] ) and ( [ chemical_potential_10 ] ) , which determine @xmath100 and @xmath101 . the net baryon number @xmath88 is conserved and is determined by eq . ( [ baryonic_density_1 ] ) using eqs . ( [ chemical_potential_5 ] ) and ( [ chemical_potential_10 ] ) . only in case of chemical equilibrium eq . ( [ baryonic_density_2 ] ) is used to determine the chemical potential @xmath84 . that is , the @xmath102 and @xmath103 ratio freezes out first at the initialization of our model calculation for the final expansion stage , where quarks start to gain mass , the background field @xmath10 starts to disappear , but the constituent quark numbers do not change . the flow does not evolve , the cells expand while coasting with negligible pressure . still considerable changes happen regarding the masses and temperatures of the system . a first order phase transition in chemical and thermal equilibrium with homogeneous nucleation would take a long time@xcite , longer than two particle correlation measurements indicate . if the homogeneous nucleation can not support the required rapid transition then the transition becomes delayed and freeze out and hadronization will happen rapidly and simultaneously from a supercooled qgp . thus , a rapid process must be out of equilibrium , at least of chemical equilibrium @xcite . possible detailed mechanisms of this out of equilibrium transition are addressed recently in several works @xcite . in the framework of the present model this transition is represented by the mass gain and coalescence of constituent quarks , and the simultaneous disappearance of the bag - field . these two processes are treated phenomenologically , while we enforce conservation laws and our model constraints . such a rapid , out of equilibrium process must result in additional entropy production @xcite from the latent heat of the transition ( just like at sudden condensation of supercooled vapour ) . we start with our model at the line where we have quarks with current quark masses , and the process ends when we reach the empirical hadron freeze out line , where quarks have constituent quark mass . it was found by cleymans et al . @xcite that , in a wide range of beam energies , detected hadrons freeze out and reach the detectors when the average energy of hadrons ( estimated in a thermal , statistical equilibrium fireball model ) is between @xmath104 . just at the freeze out moment the particles do not interact any longer , so the ideal hadron gas mixture is a good approximation . furthermore , if we plot the fo points on the temperature chemical potential @xmath105 $ ] plane , the @xmath106 constant specific energy contours fall on a continuous line indicating that the final state is representing a statistical thermal equilibrium state of hadrons or a state which is close to it . at the fo state with vanishing interactions and dissipation the expansion of the fireball is adiabatic to a good approximation . we can also approximate the final fo hadron state by assuming a jttner distribution to all hadrons . that means , we use a relation between chemical potential @xmath84 and temperature @xmath85 by means of average energy per hadron @xmath107 where hadronic energy density and hadronic particle density are given by @xmath108 here , @xmath6 is the mass of the hadron , @xmath109 is the degeneracy factor , and for baryons we have @xmath110 while for mesons @xmath111 . the boltzmann - jttner gas approach is a good approximation to the full fluid dynamics and boltzmann equation statistical result : in the medium @xmath105$]-range the fo contour lines are shown in diagram ( a ) of fig . [ fig : cleymans - lines ] ; we have taken into account the lightest @xmath112 hadrons , i.e. @xmath113 . $ ] -plane , corresponding to mean energy @xmath114 and @xmath115 per hadron . the solid lines were computed assuming that the hadrons have a jttner distribution , see eq . ( [ average_energy ] ) ( upper solid curve for @xmath115 , lower solid curve for @xmath116 ) . the dotted lines are from the statistical thermal model @xcite ( upper dotted curve for @xmath115 , lower dotted curve for @xmath116 ) . diagram ( b ) shows the same curves transformed to the @xmath117 $ ] plane.,title="fig : " ] $ ] -plane , corresponding to mean energy @xmath114 and @xmath115 per hadron . the solid lines were computed assuming that the hadrons have a jttner distribution , see eq . ( [ average_energy ] ) ( upper solid curve for @xmath115 , lower solid curve for @xmath116 ) . the dotted lines are from the statistical thermal model @xcite ( upper dotted curve for @xmath115 , lower dotted curve for @xmath116 ) . diagram ( b ) shows the same curves transformed to the @xmath117 $ ] plane.,title="fig : " ] in this transition the expansion of the quark gas continues until the average energy per hadron reaches about @xmath11 . in a statistical thermal equilibrium model this can be represented on the @xmath118 $ ] plane . in the present non - equilibrium model the thermal equilibrium is kept , but @xmath84 is not representative of the hadron multiplicity . however , we can approximate the cleymans - line by estimating the hadron multiplicity or density after coalescence of our constituent quarks into baryons and mesons . the freeze - out curve on the @xmath119 $ ] plane is plotted in diagram ( b ) of fig . [ fig : cleymans - lines ] . the cold uniform nuclear matter has an equilibrium density of @xmath120 @xmath121 , similar to the central density of large nuclei . at higher temperatures a considerable number of mesons are also present in statistical equilibrium , thus the _ effective _ baryon density for the same energy density decreases with increasing temperature . this enables us to construct an effective cleymans freeze - out line on the @xmath122$]-plane for the purposes of our non - equilibrium model . we have assumed that our rapid hadronization starts at the @xmath123$]-contour where our quarks have current quark mass of about @xmath124 . this contour is plotted in fig . [ initial_condition ] . the chemical equilibrium then ceases among quarks and anti - quarks , and we neglect quark and anti - quark annihilations in the final expansion stage . thus , the total number of quarks and anti - quarks , @xmath125 and @xmath126 , remains constant ( in contrast to the statistical thermal equilibrium model ) , while the baryon charge , @xmath127 , remains also constant as required by baryon conservation . ) and ( [ eq_60 ] ) . this corresponds to temperature and density conditions where constituent quark mass equals current quark mass @xmath128 according to eq . ( [ eq_70 ] ) . ] at the final point of the non - equilibrium expansion quarks will coalesce into mesons , baryons and anti - baryons , based on the phase space arguments used in @xcite . the rate of recombination is given by @xmath129 here , @xmath130 ( @xmath131 ) denotes quarks ( anti - quarks ) , @xmath132 denotes mesons , @xmath133 ( @xmath134 ) denotes baryons ( anti - baryons ) , @xmath135 and @xmath136 are mesonic and baryonic degeneracy factors , and the coefficients @xmath137 and @xmath138 should be determined by including phase space factors and an additional normalization constant to satisfy the following relations . notice that the same statistical phase - space factors appear in coalescence rates and in the statistical thermal model if the overlap integrals of the coalescing constituents are similar . eventually other channels via di - quark formation can also be taken into account , and the time integrated solution will provide the final hadron abundances . on the other hand the baryon charge is the excess number of baryons over anti - baryons , and this leads to the normalization requirements for the final hadrons : @xmath139 thus , a fraction `` @xmath140 '' of the anti - quarks may form anti - baryons : @xmath141 and the rest of quarks form mesons @xmath142 the only coefficient , @xmath140 , should arise from the coalescence factors above . initially from eqs . ( [ r2 ] ) - ( [ r3 ] ) the ratio of formed baryons and anti - baryons is @xmath143 , where @xmath144 . assuming that the initial recombination is dominant the conservation laws then yield @xmath145 the full integration of the rate equations may lead to a change of @xmath140 , which would modify the ratio of anti - baryons to mesons to a smaller extent . in @xcite the same coalescence model explained the constituent quark number scaling of the flow parameter , @xmath0 , from a weak elliptic asymmetry of the coalescing quark distributions . in this way at any stage of expansion , from an initial volume @xmath146 to a point at time @xmath147 with a given volume @xmath148 , we can get all the meson and baryon densities and masses , as well as the total energy from energy conservation ( neglecting the mechanical work done by the negligible pressure ) . from the baryon and meson densities @xmath88 and @xmath149 and the hadronic energy density , @xmath150 , we can calculate the baryon and meson chemical potentials ( the meson density may exceed the thermal equilibrium value ) , @xmath84 and @xmath151 , and the respective temperatures @xmath152 and @xmath153 . notice that with the broken chemical equilibrium the equality of chemical potentials is broken . in this rapid hadronization process the thermal , pressure and flow equilibrium will also break down . thus the details how this happens should be described in terms of extensive variables as we will see in the next section . let us consider now the initial conditions for all these parameters . the applicability of eq . ( [ eq_70 ] ) is restricted by the condition that the constituent quark mass has to be positive . accordingly , we will take the initial conditions such that in the initial state the in - medium chiral condensate @xmath154 vanishes and quarks have a current quark mass . this condition implies , that the density and temperature dependent terms in eq . ( [ eq_70 ] ) are equal to @xmath155 . the corresponding curve in the temperature - density plane is shown in fig . [ initial_condition ] . we used for the current quark mass @xmath156 , ignoring the difference between @xmath157 and @xmath158 quarks . in this model it is assumed that the chemical equilibrium between the quarks and anti - quarks breaks precisely on this curve , and the quark and anti - quark numbers are conserved separately during further expansion . at the earlier stage , the quark and anti - quark numbers , @xmath159 and @xmath126 , are computed from the temperature @xmath85 and baryon number density @xmath88 using eqs . ( [ baryonic_density_2 ] ) , ( [ chemical_potential_5 ] ) and ( [ chemical_potential_10 ] ) with chemical potentials @xmath86 and @xmath87 . but after crossing the line on fig . [ initial_condition ] the numbers of quarks and anti - quarks , @xmath125 and @xmath126 , are assumed to stay constant . during further expansion @xmath160 and @xmath161 change separately and can be obtained numerically from the quark and anti - quark densities , @xmath162 and @xmath163 , by means of eqs . ( [ chemical_potential_5 ] ) and ( [ chemical_potential_10 ] ) . since this is a non - equilibrium model , all numerical calculations are carried out on the @xmath119 $ ] plane , using the densities @xmath88 , @xmath162 , @xmath163 and the temperature @xmath85 ( and not on the @xmath164 $ ] plane ) . the expansion of the gas of quarks and anti - quarks is considered to be adiabatic , i.e. at constant entropy . the initial total entropy density of the gas , @xmath165 , can be calculated from the initial temperature , @xmath166 , and initial baryon charge density , @xmath167 , assuming chemical equilibrium between the quarks and anti - quarks , i.e. @xmath168 , and using eq . [ baryonic_density_2 ] to get the chemical potential and substituting it into eqs . ( [ entropy_10 ] ) , ( [ entropy_11 ] ) , and ( [ entropy_12 ] ) . after the initial moment , the quark and anti - quark densities decrease inversely proportionally to the volume @xmath45 of the system : @xmath169 , @xmath170 , @xmath171 , where @xmath146 is the initial volume . then the total entropy of the system can be expressed simply as a function of volume , temperature and numerically obtained chemical potentials ( or temperature and baryon density ) using eqs . ( [ entropy_10 ] ) , ( [ entropy_11 ] ) and ( [ entropy_12 ] ) . using the condition that the total entropy is constant , i.e. the entropy density also decreases as @xmath172 , the expansion trajectories on the @xmath119 $ ] plane can be calculated numerically . these trajectories are plotted in fig . [ fig : adiabatic - trajectories ] . the trajectories corresponding to constant energy expansion ( iso - ergic ) were calculated , in a similar way , using eqs . ( [ eq_80 ] ) and ( [ eq_85 ] ) . this case corresponds to a dissipative expansion . it is apparent from fig . [ fig : adiabatic - trajectories ] that the adiabatic expansion where @xmath173 leads to the fastest temperature decrease than the iso - ergic one where @xmath174 . . beyond that curve we take @xmath128 . the dots on the thick solid line indicate the initial temperature and initial baryonic density for each curve of adiabatic and iso - ergic expansion . the thin solid lines represent the trajectories of the adiabatic expansion of the gas of quarks and anti - quarks . the trajectories of the iso - ergic dissipative expansion ( dotted lines ) are plotted for comparison . ] due to confining forces , the quarks will finally recombine into hadrons . we assume that this happens rapidly at the point of recombination when the average energy per hadron ( including the background field ) decreases to @xmath175 , a value which is still above the values of @xmath176 obtained by cleymans et al . this corresponds nearly to the energy per hadron of the empirically observed freeze - out . the endpoints of the expansion curves where the recombination happens are shown in fig . [ fig : final - points ] . the thick solid line is the same as in fig . [ initial_condition ] . ] the quarks are assumed to recombine into three types of particles : baryons , anti - baryons and mesons , which contain three quarks , three anti - quarks , or a quark and an anti - quark , respectively . all hadrons are assumed to have a mass that is the sum of the masses of their constituent quarks at the freeze - out , i.e. @xmath177 further differences between the various hadron species are disregarded . most of the anti - quarks will pair with quarks to form mesons , but a small fraction , @xmath140 , will form anti - baryons . this ratio , @xmath140 , can be estimated based on the recombination rates given in @xcite . thus the baryon , anti - baryon and meson densities @xmath178 , @xmath179 and @xmath149 are calculated from the quark and anti - quark densities ( @xmath162 and @xmath163 ) using eqs . ( [ eq : baryon - number ] ) , ( [ eq : meson - number ] ) and ( [ eq : recomb - a - param ] ) . ( diagram ( a ) ) and transverse energy @xmath180 ( diagram ( b ) ) , for the case of an adiabatic expansion . the initial state for this calculation was @xmath181 and @xmath182.,title="fig : " ] ( diagram ( a ) ) and transverse energy @xmath180 ( diagram ( b ) ) , for the case of an adiabatic expansion . the initial state for this calculation was @xmath181 and @xmath182.,title="fig : " ] , but re - scaled according to constituent quark number @xmath2 ( @xmath183 for baryons , @xmath184 for mesons ) . diagram ( a ) shows the scaled distribution @xmath185 as function of re - scaled transverse momentum @xmath186 . diagram ( b ) shows the re - scaled distribution @xmath187 as function of re - scaled transverse energy @xmath188 . these re - scaled distributions almost coincide for baryons and mesons ; they are on top of each other so no difference can be seen here in the diagrams . the initial state for this calculation was @xmath181 and @xmath182.,title="fig : " ] , but re - scaled according to constituent quark number @xmath2 ( @xmath183 for baryons , @xmath184 for mesons ) . diagram ( a ) shows the scaled distribution @xmath185 as function of re - scaled transverse momentum @xmath186 . diagram ( b ) shows the re - scaled distribution @xmath187 as function of re - scaled transverse energy @xmath188 . these re - scaled distributions almost coincide for baryons and mesons ; they are on top of each other so no difference can be seen here in the diagrams . the initial state for this calculation was @xmath181 and @xmath182.,title="fig : " ] we assume that at the point of recombination the flow freezes out too , and both the thermal and chemical equilibria cease . the hadrons are assumed to have jttner distribution after the freeze out , but the temperature parameter in this distribution will be different for baryons and mesons . the parameters of the distribution after recombination are calculated from the condition of energy conservation : the thermal energy of each hadron type will be equal to the energy of their constituent quarks , @xmath189 . due to the different masses of baryons and mesons , their temperature parameters will be different . the temperature ratio @xmath190 will correspond to the mass ratio @xmath191 . the distributions of baryons and mesons for a calculation done with initial state @xmath181 and @xmath182 is shown in fig . [ fig : pt - distribution_a ] and fig . [ fig : pt - distribution_b ] . here the final baryon and meson temperatures are @xmath192 and @xmath193 and @xmath194 , while the final constituent quark mass is 308 mev . the scaled @xmath8 and @xmath180 distributions become identical under this condition , however this is not enough to reproduce the ncq scaling of @xmath195 indicating that the recombination influences the flow velocities of the final hadrons . this concept was already pointed out in @xcite based on the properties of the collision integral . it is important to point out that the transport theoretical treatment and the collision integral are applicable also at situations when the local equilibrium has ceased to exist . the elliptic flow parameter , @xmath0 , can be calculated from the final , post freeze - out distribution by the cooper - frye formula . assuming an isochronous fo hypersurface , we obtain simple expressions for final measurables @xcite . in this section we mainly follow the arguments of refs . the kinematic average of a quantity @xmath196 is given by @xmath197 where @xmath198 is the one - particle distribution function . especially , we are interested in the elliptic flow @xmath0 , that is the following kinematic average @xmath199 according to eqs . ( [ eq_1 ] ) and ( [ eq_5 ] ) , we obtain @xmath200 now we subdivide the system into @xmath55 fireballs or cells , each of which has a given volume @xmath95 and contains a given number of particles @xmath69 , which are distributed by a given distribution function @xmath201 inside cell @xmath57 . since in a given cell the elliptic flow parameter @xmath0 does not depend on coordinate @xmath202 , we can take the integral over @xmath203 and obtain @xmath204 where we have used the relation @xmath205 with @xmath206 being the transverse momentum . in our study we consider the mid - rapidity particles , ( @xmath207 or the rapidity @xmath208 ) , i.e. @xmath209 and there is no integration for the longitudinal momentum or rapidity . furthermore , we are interested in the momentum dependence of elliptic flow parameter @xmath0 . thus we obtain from ( [ eq_15 ] ) the expression : @xmath210 where the sum runs over all individual cells @xmath211 , and we have used the fact that for mid - rapidity particles @xmath212 . let us study the elliptic flow of a single type of particle . we assume a jttner distribution in each individual cell of the bag , see eq . ( [ juttner_a ] ) , and obtain @xmath213 where @xmath214 is the component of the momentum parallel to the cell velocity @xmath215 , and @xmath216 is the chemical potential of the given particle type in the cell @xmath57 ; here we have also used @xmath217 and @xmath218 . since the elliptic flow parameter is calculated after chemical freeze - out has happened , each particle type has their own chemical potential @xmath219 where @xmath220 is the density and @xmath221 is the mass of the given particle type . for particles at cm rapidity the zeroth component of four - momentum equals the transverse mass , i.e. it is just the energy of one particle @xmath222 then , according to eqs . ( [ eq_16 ] ) and ( [ juttner ] ) , we obtain the following expression for the elliptic flow : @xmath223 where the sum runs over all cells of the bag . now let us insert eq . ( [ juttner ] ) into expression ( [ elliptic_flow_a ] ) . then , by means of eq . ( [ eq : juttner - density ] ) we obtain the following expression for the @xmath0 parameter : @xmath224 where @xmath225 denotes @xmath226 if the direction of the velocity of cell @xmath57 relative to axis @xmath227 is denoted by @xmath228 , then @xmath214 can be written as @xmath229 , and the following expression is obtained : @xmath230 @xmath231 and @xmath232 denote the zeroth and second order modified bessel functions of the first kind @xcite . notice that if the source temperatures are all equal or if the freeze out temperature of all fluid cells is the same , than one observes a linear increase of @xmath0 as a function of @xmath1 or @xmath5 . this is frequently quoted as a linear hydrodynamical increase , observed already in the first model calculations , e.g. @xcite . such linear increase leads trivially to a constituent quark number scaling . it is easy to see that this feature is just a consequence of the equal temperature assumption . if for example we assume a large , hotter , static central source , the high @xmath1 behaviour of @xmath0 will be dominated by this hot source and thus @xmath0 will decrease at high @xmath1 as pointed out in ref . @xcite . if we assume that the temperatures of all cells are the same , @xmath233 , then the temperature- and mass - dependent parts of @xmath234 ( see eq . ( [ eq : n - tilde ] ) ) cancel from the numerator and denominator of eq . ( [ eq : v2-full ] ) , and we get ( note at fo the constituent quark mass does not depend on density and temperature anymore ) @xmath235 in the special case of four cells moving into the four directions ( @xmath236 ) , eq . ( [ eq : v2-simple ] ) reduces to the expression given in @xcite . the simplest configuration that can approximate elliptic flow is dividing the system into two cells that move in opposite directions with the same velocity @xmath237 . for this two - cell " model case the @xmath0 parameter is expressed as @xmath238 a slightly more complicated possibility is having one larger non - moving central cell , and two smaller side cells moving in opposite directions , as shown in fig . [ fig : three - cells ] . in this three - cell " model case @xmath3 can be expressed as @xmath239 here @xmath240 denotes the particle number of the central cell , while @xmath241 denotes the particle number of the identical side - cells . this configuration corresponds to a flow with less pronounced asymmetry . the large central cell has a spherical momentum distribution , while the smaller side cells introduce a slight asymmetry to this . as a first approximation , the @xmath0 parameter was calculated using a simple model of elliptic flow where the system is divided into two droplets moving in opposite directions with the same velocity . for this model , @xmath0 is given by eq . ( [ eq : v2-twocell ] ) . the baryons and mesons were given different flow energies , such that the ratio of flow energy per quark is @xmath242 . this leads to different flow velocities for baryons and mesons , and reproduces the constituent quark number scaling of the elliptic flow parameter . in this model the scaling according to the transverse momentum @xmath8and transverse energy @xmath180 are tied to each other , i.e. the scaling is either present or not for both these variables . by this point of the evolution of our system , all energy in the background field is exhausted and the internal , excitation and random kinetic energies of the hadrons reach the fo value @xmath243 . the @xmath3 curve obtained with the initial state parameters discussed in the previous section is shown in fig . [ fig : v2-pt ] . parameter as a function of @xmath8 , calculated from the two - cell model , eq . ( [ eq : v2-twocell ] ) . the initial state used is the same as in fig . [ fig : pt - distribution_a ] and fig . [ fig : pt - distribution_b ] . the dotted curve represents the baryons , while the solid curve represents the mesons . the cell velocities for baryons and mesons are @xmath244 and @xmath245 , corresponding to a flow - energy ratio of @xmath246 of the constituent quarks of the two different particle types ( calculated relativistically ) . ] , as a function of @xmath5 ( diagram ( a ) ) and @xmath247 ( diagram ( b ) ) , calculated from the two - cell model . the dotted curves represent the baryons while the solid curves represent the mesons . the curves coincide for low @xmath8 value , i.e. the constituent quark number scaling is reproduced for the low @xmath8 region.,title="fig : " ] , as a function of @xmath5 ( diagram ( a ) ) and @xmath247 ( diagram ( b ) ) , calculated from the two - cell model . the dotted curves represent the baryons while the solid curves represent the mesons . the curves coincide for low @xmath8 value , i.e. the constituent quark number scaling is reproduced for the low @xmath8 region.,title="fig : " ] the elliptic flow parameter , re - scaled according to constituent quark number is shown in fig . [ fig : v2-cqn ] . in this figure , the baryon- and meson - curves coincide for low @xmath8 , i.e. constituent quark number scaling of @xmath0 is reproduced for small values of the transverse momentum , but not for @xmath248 mev . these results of qns can be further improved by considering a three - cell model , as we will see in the next subsection . the asymmetry in the two - cell model described in the previous subsection is very strong . in @xcite it is shown that the constituent quark number scaling is more precise if the @xmath0 coefficient is small , and the higher harmonic coefficients , @xmath249 ( @xmath250 ) are negligible . therefore , a three - cell model with one large stationary central cell and two moving side cells was also studied . the schematic scheme of this arrangement of cells is shown on fig . [ fig : three - cells ] . , or two smaller moving side cells and a stationary central cell . the schematic of the latter configuration is shown here . ] parameter as a function of @xmath8 , calculated from the three - cell model , eq . ( [ eq : v2-threecell ] ) . the initial state used is the same as in fig . [ fig : pt - distribution_a ] and fig . [ fig : pt - distribution_b ] . the dotted curve represents the baryons , while the solid curve represents the mesons . the cell velocities for baryons and mesons are @xmath244 and @xmath245 , corresponding to a flow - energy ratio of @xmath246 of the constituent quarks of the two different particle types ( calculated relativistically ) . a particle ratio @xmath251 has been assumed . ] , as a function of @xmath5 ( diagram ( a ) ) and @xmath247 ( diagram ( b ) ) , calculated from the three - cell model . the dotted curves represent the baryons while the solid curves represent the mesons . the curves overlap nearly completely , and the constituent quark number scaling is reproduced for a much wider range of @xmath8 than in the case of the two - cell model . the velocities of the side sources are the same as in fig . [ fig : v2-pt ] . a particle ratio @xmath251 has been assumed.,title="fig : " ] , as a function of @xmath5 ( diagram ( a ) ) and @xmath247 ( diagram ( b ) ) , calculated from the three - cell model . the dotted curves represent the baryons while the solid curves represent the mesons . the curves overlap nearly completely , and the constituent quark number scaling is reproduced for a much wider range of @xmath8 than in the case of the two - cell model . the velocities of the side sources are the same as in fig . [ fig : v2-pt ] . a particle ratio @xmath251 has been assumed.,title="fig : " ] the elliptic flow parameter of the three - cell model is given by eq . ( [ eq : v2-threecell ] ) and the result is shown in fig . [ fig : v2-pt - three ] . the side cells were assumed to have the same velocities as in the case of the two - cell model , and the temperature of all cells was the same . the particle number ratio of the central cell to the side cells was set to @xmath251 . it should be noticed , that the elliptic flow parameter , shown in fig . [ fig : v2-pt - three ] , is not insensitive on such choice , e.g. a ratio of @xmath252 would enlarge the value of @xmath0 . the elliptic flow parameter , re - scaled according to constituent quark number is shown in fig . [ fig : v2 - 3-cqn ] . we have found that qns is insensitive on the chosen particle number ratio @xmath251 , e.g. a ratio @xmath252 yields very similar results as shown in fig . [ fig : v2 - 3-cqn ] according to the results in fig . [ fig : v2 - 3-cqn ] , the three - cell model is able to reproduce the constituent quark number scaling of @xmath0 for a wider range of @xmath8 values compared to the two - cell model . it must be noted that this is not only a result of the reduction of the momentum distribution asymmetry compared to the two - cell model . however , the choice of flow velocities for baryons and mesons is still relevant . a model of rapid hadronization was considered in which the effective constituent quark mass depends on density and temperature . in this model the evolution starts from an ideal qgp in chemical , thermal and mechanical equilibrium . then , one after the other the chemical , thermal and mechanical equilibrium break down rapidly , while the quarks build up constituent quark mass , and the background gluon field ( bag constant ) breaks up and vanishes . this model can be considered as a simple representation of the breaking chiral symmetry and deconfinement in a dynamical transition crossing the quarkyonic phase @xcite . the elliptic flow parameter @xmath0 was calculated for the final hadron distributions obtained from the model , and the constituent quark number scaling was partially reproduced . we assume the following stages of hadronization : 1 ) the chemical equilibrium between the quarks and anti - quarks breaks and the chiral symmetry breaks at the same time . 2 ) the quark and anti - quark numbers are assumed to be conserved during further expansion , i.e. quark , anti - quark pair creations and annihilations are ceased . 3 ) as the quark gas expands and cools , the quarks gain mass according to eq . ( [ eq_70 ] ) , and the field associated with the bag constant @xmath10 decreases . 4 ) the thermal freeze - out and recombination into hadrons are completed when the mean energy per hadron reaches the fo value of @xmath253 . the created hadrons are not in local thermal and flow equilibrium with each other . we considered only three types of hadrons : baryons , anti - baryons and mesons , ignoring the differences between different hadron species , and the momentum distribution of hadrons was calculated . using the obtained momentum distributions , the elliptic flow parameter @xmath0 was computed from a simple multi - source model of the elliptic flow . two cases were considered . in the first case , the flow asymmetry is approximated by two fireballs moving in opposite directions . although this leads to a very strong asymmetry in the momentum distribution , it leads to a @xmath3 curve that scales with the constituent quark number for small @xmath8 values only . in the second case , we added a large central fireball to reduce the asymmetry and approximate pure elliptic flow better ( i.e. reduce the higher @xmath249 ( @xmath250 ) harmonic coefficients ) . this three - source model was able to reproduce the constituent quark number scaling for a wide range of @xmath8 values . the present model is highly simplified and attempts to provide an insight to the rapid hadronization and freeze out process in view of the quark number scaling . we considered the features arising from breaking down of equilibrium in this process in terms of thermo and fluid dynamical parameters which are applicable to partial components of the matter . we intend to implement these concepts in more complex models ( like hybrid models ) and to search for more fundamental reasons for the observed freeze out features in terms of the partial extensives . despite the simplicity of the model , it is capable of reproducing the constituent quark number scaling of the @xmath0 elliptic flow parameter . the presence of constituent quark number scaling in experimental data suggests that the elliptic flow develops in the quark - gluon plasma phase , before the quarks recombine into hadrons . therefore understanding the origin the elliptic flow can provide insight into the quark phase of matter . enlightening discussions with prof . daniel d. strottman are gratefully acknowledged . the authors also would like to thank dr . kalliopi kanaki , yun cheng , dr . csaba anderlik and dr . etele molnar for useful hints and advices . sven zschocke thanks for the sincere hospitality of the physics department of university of bergen , and magne hvg is acknowledged for kind computer assistance . this work was supported by the alexander von humboldt foundation , by the meltzer fund of the university of bergen and by the computational subatomic physics project at uni - research of the research council of norway . igor n. mishustin acknowledges partial support from dfg grant 436rus 113/957/0 - 1 ( germany ) and grants ns-7235.2010.2 and rfbr09 - 02 - 91331 ( russia ) . s. s. adler et al . ( phenix collaboration ) , phys . lett . * 91 * , 182301 ( 2003 ) ; j. adams et al . ( star collaboration ) , phys . lett . * 92 * , 052302 ( 2004 ) . l. mclerran , r.d . pisarski , nucl . phys . a * 796 * , 83 ( 2007 ) . j. cleymans , d. elliott , a. keranen , e. suhonen , phys . c * 57 * , 3319 ( 1998 ) ; j. cleymans , h. oeschler , k. redlich , phys . c * 59 * , 1663 ( 1999 ) ; a. andronik , p. braun - munzinger , j. stachel , nucl . a * 772 * , 167 ( 2006 ) ; j. cleymans , h. oeschler , k. redlich , s. wheaton , acta phys . polonica b proc . suppl . * 3 * , 533 ( 2010 ) . k. rajagopal and f. wilczek , nucl . b * 404 * , 577 ( 1993 ) ; i.n . mishustin , phys . lett . * 82 * , 4779 ( 1999 ) ; i.n . mishustin , nucl . a * 681 * , 56 ( 2001 ) ; o. scavenius , a. dumitru , e.s . fraga , j.t . lenaghan , and a.d . jackson , phys . d * 63 * , 116003 ( 2001 ) ; j. randrup , phys . lett . * 92 * , 122301 ( 2004 ) ; g. torrieri , b. tomasik , i.n . mishustin , phys . c * 77 * , 034903 ( 2008 ) ; a. bessa , e.s . fraga , and b.w . mintz , phys . d * 79 * , 034012 ( 2009 ) ; j. randrup , phys . c * 82 * , 034902 ( 2010 ) .

the constituent quark number scaling of elliptic flow is studied in a non - equilibrium hadronization and freeze - out model with rapid dynamical transition from ideal , deconfined and chirally symmetric quark gluon plasma , to final non - interacting hadrons . in this transition a bag model of constituent quarks is considered , where the quarks gain constituent quark mass while the background bag - field breaks up and vanishes . the constituent quarks then recombine into simplified hadron states , while chemical , thermal and flow equilibrium break down one after the other . in this scenario the resulting temperatures and flow velocities of baryons and mesons are different . using a simplified few source model of the elliptic flow , we are able to reproduce the constituent quark number scaling , with assumptions on the details of the non - equilibrium processes .

introduction constituent quark mass expanding quarkanti-quark bag non-equilibrium expansion stage elliptic flow summary acknowledgements

the band theory of itinerant magnetism has been much developed in the past half a century on the basis of the density functional theory ( dft ) @xcite with local spin density approximation ( lsda ) @xcite or the generalized gradient approximation ( gga ) @xcite . the dft explains quantitatively the ground - state magnetism of 3@xmath3 transition metals and alloys . the first - principles band calculations with use of the lsda , for example , yield the magnetization per atom 2.15 @xmath4 for fe and 0.59 @xmath4 for ni , which are in good agreement with the experimental values @xcite 2.22 @xmath4 and 0.62 @xmath4 , respectively . the lsda stabilizes artificially the fcc fe instead of the bcc fe with the volume contraction by 15 % . the gga has solved the problem on the stability of structure in fe @xcite . on the other hand , understanding of the finite - temperature magnetism in transition metals is still far from the final goal of quantitative description . the stoner theory based on the dft band calculations yields the curie temperatures ( @xmath1 ) , 6000 k for fe and 3000 k for ni @xcite , which are 6 or 5 times as large as the observed values ( 1040 k for fe @xcite and 630 k for ni @xcite ) . because of a large discrepancy between the theory and the experiment , many theories which take into account spin fluctuations at finite temperatures have been developed @xcite . hubbard @xcite and hasegawa @xcite proposed a single - site spin fluctuation theory ( ssf ) on the basis of the functional integral method @xcite and the coherent potential approximation ( cpa ) @xcite . they explained the curie - weiss susceptibility in 3@xmath3 transition metals and obtained the curie temperatures in fe and ni which qualitatively agree with the experiments . the ssf , however , reduces to the hartree - fock theory at @xmath5 because it is based on a high - temperature approximation . therefore , the theory does not take into account electron correlations at the ground state as discussed by gutzwiller @xcite , hubbard @xcite , and kanamori @xcite . kakehashi and fulde @xcite proposed a variational theory at finite temperatures which takes into account the gutzwiller - type electron correlations at the ground state and reduces to the ssf at high temperatures . they found that the curie temperatures in fe and ni are reduced by @xmath6 due to correlated motion of electrons . hasegawa @xcite also proposed a similar theory on the basis of the slave - boson functional integral method . these theories are not suitable for systematic improvements of the theory because of the difficulty in finding a suitable temperature dependence of the effective potential being projected onto the static fields . in order to make a systematic improvements possible , we proposed the dynamical cpa @xcite which fully takes into account the single - site spin and charge fluctuations self - consistently , and clarified the basic properties of the theory with use of a monte - carlo method . in the next paper @xcite which we refer to i , we proposed more analytic method to the dynamical cpa using the harmonic approximation ( ha ) @xcite . towards quantitative calculations , it is indispensable to take into account realistic band structures in solids . in our recent paper @xcite which we refer to ii , we proposed the first - principles dynamical cpa combining the dynamical cpa+ha with the lda+u scheme @xcite based on the tight - binding linear muffin - tin orbitals ( tb - lmto ) @xcite . quite recently we extended the calculations taking into account the dynamical corrections up to the 4th order in coulomb interactions , and have shown that the single - particle excitation spectra obtained by the first - principles dynamical cpa explain systematic change of the x - ray photoemission spectroscopy ( xps ) and the bremsstrahlung isochromat ( bis ) data from sc to cu at high - temperatures @xcite . in our numerical calculations , we adopted the coulomb interactions @xmath7 obtained by anisimov _ et . @xcite , and by bandyopadhyay and sarma @xcite . the exchange interaction energies @xmath0 were taken from the values obtained by the hartree - fock atomic calculations @xcite . recently , considerable efforts to obtain the first - principles @xmath7 and @xmath0 have been made at the ground state . aryasetiawan and coworkers @xcite obtained the coulomb and exchange interactions for @xmath3 orbitals which are screened by @xmath8 electrons , using the random phase approximation ( rpa ) and the lmto method . the same type of calculations has been performed by using the first - principles wannier orbitals by miyake and aryasetiawan @xcite . these results indicate that irrespective of transition metal elements the exchange energy integral @xmath0 are reduced by about @xmath9 as compared with the atomic ones . although the methods have been applied to many other compounds @xcite , there is no systematic investigation for the influence of the screened @xmath0 on magnetic properties . we investigate in this paper the ferromagnetic properties of fe , co , and ni on the basis of the first - principle dynamical cpa using both the atomic exchange interaction energies ( atomic @xmath0 ) and the exchange interaction energies screened by 30% ( screened @xmath0 ) . we examine the role of the exchange energy parameters in the magnetic properties at finite temperatures and clarify their validity to the magnetic properties . we will demonstrate that use of the screened @xmath0 does not necessary improve theoretical results at finite temperatures , though we find some improvements at low temperature regime . as we have proven in our previous papers @xcite , the dynamical cpa is equivalent to the many - body cpa @xcite in disordered alloys , the dynamical mean - field theory ( dmft ) @xcite in the metal - insulator transitions , and the projection operator cpa @xcite for excitation problem in solids . the first - principles dmft calculations for fe and ni have been performed at the ground state by miura and fujiwara @xcite within the iterative perturbation method . the finite - temperature dmft calculations for fe and ni have been performed by using the hamiltonian without transverse spin fluctuations @xcite . in both cases , atomic exchange interaction energy @xmath10 ry was adopted for fe and ni . we present here the finite - temperature results obtained by the hamiltonian with transverse spin fluctuations and by using both the atomic @xmath0 and the screened @xmath0 values . the outline of the paper is as follows . in the following section , we summarize the first - principles dynamical cpa based on the ha . in sec . iii , we present numerical results of calculations for the densities of states as the excitation spectra , magnetization vs temperature curves , curie temperatures , paramagnetic susceptibilities , effective bohr magneton numbers , as well as the amplitudes of local magnetic moments in fe , co , and ni . by comparing these quantities calculated by the atomic @xmath0 with those obtained by the screened @xmath0 , we will clarify the role of the screening effects on the exchange energy @xmath0 , and discuss the availability of the screened @xmath0 . furthermore , we discuss the validity of the screened @xmath0 by comparing calculated magnetic properties with those in experiments . in sec . iv , we present a summary of our works , and discuss possible reasons for the disagreement with experiments when the screened @xmath0 are applied . in the first - principles dynamical cpa , we adopt the tb - lmto hamiltonian combined with a lda+u coulomb interactions as follows @xcite . @xmath11 @xmath12 @xmath13 \ . \label{h1}\end{aligned}\ ] ] we assumed here a transition metal with an atom per unit cell . @xmath14 in eq . ( [ h0 ] ) is an atomic level on site @xmath15 and orbital @xmath16 , @xmath17 is the chemical potential , @xmath18 is a transfer integral between orbitals @xmath19 and @xmath20 . @xmath21 denotes @xmath22 , @xmath23 , and @xmath3 orbitals . @xmath24 ( @xmath25 ) is the creation ( annihilation ) operator for an electron with orbital @xmath16 and spin @xmath26 on site @xmath15 , and @xmath27 is a charge density operator for electrons with orbital @xmath16 and spin @xmath26 on site @xmath15 . the coulomb interaction term @xmath28 in eq . ( [ h1 ] ) consists of the on - site interactions between @xmath3 electrons ( @xmath29 ) . @xmath30 ( @xmath31 ) and @xmath0 denote the intra - orbital ( inter - orbital ) coulomb and exchange interactions , respectively . @xmath32 ( @xmath33 ) with @xmath29 is the charge ( spin ) density operator for @xmath3 electrons on site @xmath15 and orbital @xmath34 . note that the atomic level @xmath14 in @xmath35 is not identical with the lda atomic level @xmath36 ; @xmath37 . here @xmath38 is the charge density at the ground state , @xmath39 is a lda functional to the intra - atomic coulomb interactions @xcite . in the dynamical cpa , we transform in the free energy the interacting hamiltonian @xmath28 into a one - body hamiltonian with dynamical potential @xmath40 for time - dependent random charge and exchange fields , using the functional integral method @xcite . introducing a site - diagonal uniform medium , _ i.e. _ , a coherent potential @xmath41 into the potential part , we expand the correction @xmath42 with respect to sites in the free energy . the zeroth term in the expansion is the free energy for a uniform medium , @xmath43 $ ] . the next term is an impurity contribution to the free energy . the dynamical cpa neglects the higher - order terms associated with inter - site correlations . the free energy per atom is then given by @xcite @xmath44 - \beta^{-1 } { \rm ln } \ , \int \big [ \prod_{\alpha } \sqrt{\dfrac{\beta \tilde{j}_{\alpha}}{4\pi } } d \xi_{\alpha } \big ] \ , { \rm e}^{\displaystyle -\beta e_{\rm eff}(\mbox{\boldmath$\xi$ } ) } . \label{fcpa2}\end{aligned}\ ] ] here @xmath45 is the inverse temperature , @xmath46j$ ] , @xmath47 , and @xmath48 is a static field variable on a site . the effective potential @xmath49 in eq . ( [ fcpa2 ] ) consists of the static term @xmath50 and the dynamical correction term @xmath51 as follows . @xmath52 the static term is given as @xmath53 \nonumber \\ & & + \dfrac{1}{4 } \big [ - ( u_{0}-2u_{1}+j ) \sum_{m } \tilde{n}_{l}(\boldsymbol{\xi})^{2 } - ( 2u_{1}-j ) \tilde{n}_{l}(\boldsymbol{\xi})^{2 } + \tilde{j}^{2}_{\bot } \xi^{2}_{\bot } + \tilde{j}^{2}_{z } \xi^{2}_{z } \big ] . \label{est2}\end{aligned}\ ] ] here @xmath54 , and @xmath55 . @xmath56 is a static potential given by @xmath57/2 - \tilde{j}_{z}\xi_{z}\sigma/2 $ ] , @xmath58 is the coherent potential for matsubara frequency @xmath59 . the electron number @xmath60 for a given @xmath61 is expressed by means of an impurity green function as @xmath62 and @xmath63 . the impurity green function @xmath64 has to be determined self - consistently . the explicit expression will be given later ( see eq . ( [ gimp ] ) ) . the coherent green function @xmath65 in eq . ( [ est2 ] ) is defined by @xmath66_{il\sigma il\sigma } . \label{fls}\end{aligned}\ ] ] here @xmath67 is the one - electron hamiltonian matrix for the noninteracting hamiltonian @xmath35 , and @xmath68 . the dynamical potential @xmath51 in eq . ( [ eeff ] ) has been obtained within the harmonic approximation ( ha ) @xcite . it is based on an expansion of @xmath69 with respect to the frequency mode of the dynamical potential @xmath70 , where @xmath71 . the ha is the neglect of the mode - mode coupling terms in the expansion . we have then @xmath72 . \label{edyn1}\end{aligned}\ ] ] here the determinant @xmath73 is a contribution from a dynamical potential @xmath70 with frequency @xmath74 , and the upper bar denotes a gaussian average with respect to the dynamical charge and exchange field variables , @xmath75 and @xmath76 ( @xmath77 ) . the determinant @xmath73 is expressed as @xcite @xmath78 , \label{dnu2}\end{aligned}\ ] ] @xmath79 note that 1 in the above determinant denotes the @xmath80 unit matrix , @xmath81 is a @xmath80 matrix in the spin space , which are defined by @xmath82 @xmath83 @xmath84_{\sigma\sigma^{\prime } } \ . \label{gst}\end{aligned}\ ] ] here @xmath85 ( @xmath86 ) are the pauli spin matrices . @xmath87 , @xmath88 @xmath89 ) , and @xmath90 . @xmath91 is the impurity green function in the static approximation , @xmath92 , and @xmath93 is defined by @xmath94 . the determinant @xmath95 defined by eq . ( [ dnukm ] ) is expanded with respect to the dynamical potential as follows . @xmath96 @xmath97 here the subscripts @xmath98 and @xmath99 take 4 values @xmath100 , @xmath101 , @xmath102 , and @xmath103 , and @xmath104 @xmath105 note that the subscript @xmath106 of @xmath107 in eq . ( [ dnnukm ] ) denotes a set of @xmath108 . substituting eq . ( [ dnukm2 ] ) into eq . ( [ dnu2 ] ) and taking the gaussian average , we reach @xmath109 and @xmath110 . \label{dnubarn}\end{aligned}\ ] ] here each element of @xmath111 has a value of zero or positive integer . @xmath112 takes one of 4 cases @xmath100 , @xmath101 , @xmath102 , and @xmath103 . @xmath113 denotes the @xmath113-th member of the @xmath114 block with @xmath115 elements . p denotes a permutation of a set @xmath116 ; @xmath117 . @xmath118 means a rearrangement of @xmath119 according to the inverse permutation p@xmath120 . the coefficient @xmath121 in eq . ( [ dnubarn ] ) is a coulomb interaction defined by @xmath122 the frequency dependent factors @xmath107 in eqs . ( [ dnnukm ] ) and ( [ dnubarn ] ) consist of a linear combination of @xmath123 products of the static green functions . their first few terms have been given in appendix a of our paper ii @xcite . in the calculations of the higher - order dynamical corrections @xcite @xmath107 , we note that the coupling constants @xmath124 are considerably smaller than @xmath125 and @xmath126 because @xmath30 and @xmath127 . thus we neglect the transverse potentials , @xmath128 and @xmath129 . the approximation implies that @xmath130 . the determinant @xmath95 in eq . ( [ dnu2 ] ) is then written by the products of the single - spin components as @xmath131 here @xmath132 is defined by eq . ( [ dnukm ] ) in which the @xmath80 unit matrices have been replaced by 1 ( _ i.e. _ , @xmath133 unit matrices ) , and the @xmath80 matrices @xmath134 have been replaced by the @xmath133 matrices @xmath135 . the latter is given by @xmath136 @xmath137 here @xmath138 , and we used a notation @xmath139 for simplicity . in order to reduce these summations , we make use of an asymptotic approximation @xcite . @xmath140 where @xmath141 . the approximation is justified in the high - frequency limit where @xmath142 is written as @xmath143 in the asymptotic approximation , we obtain @xmath144 here we wrote the subscript at the r.h.s . ( right - hand - side ) explicitly to avoid confusion . note that the values of @xmath98 and @xmath99 are limited to @xmath100 or @xmath103 in the present approximation . the spin - dependent quantities are given by @xcite @xmath145 @xmath146 @xmath147 + \sum^{l-1}_{i=0 } \dfrac{(-)^{l - i } l!}{i ! ( l - i ) ! } \big [ \prod^{i-1}_{j=-(l - i ) } \tilde{g}_{l\sigma}(j\nu+k ) \big ] \big [ 1 + \dfrac{l - i}{\overline{q}_{\nu}^{\,i } } \tilde{g}_{l\sigma}(i\nu+k ) \big ] \ . \label{bls}\end{aligned}\ ] ] here @xmath148 . @xmath149 is the number of @xmath150 pairs such that @xmath151 among the @xmath152 pairs . equation ( [ bls ] ) reduces to the result of the zeroth asymptotic approximation in our paper i @xcite when there is no orbital degeneracy . in the actual applications we make use of the exact form up to a certain order of expansion in @xmath153 ( see eq . ( [ dnukm2 ] ) ) , and for higher - order terms we adopt an asymptotic form ( [ dnag2 ] ) . in this way , we can take into account dynamical corrections systematically starting from both sides of the weak - interaction limit and the high - temperature limit . the coherent potential can be determined by the stationary condition @xmath154 . this yields the dynamical cpa equation as @xcite @xmath155 here @xmath156 at the l.h.s . ( left - hand - side ) is a classical average taken with respect to the effective potential @xmath49 . the impurity green function is given by @xmath157 the first term at the r.h.s . is the impurity green function in the static approximation , which is given by eq . ( [ gst ] ) . the second term is the dynamical corrections , and @xmath158 . solving the cpa equation ( [ dcpa3 ] ) self - consistently , we obtain the effective medium . the electron number on each orbital @xmath16 is then calculated from @xmath159 the chemical potential @xmath17 is determined from the condition @xmath160 . here @xmath161 denotes the conduction electron number per atom . the magnetic moment is given by @xmath162 in particular , the @xmath29 component of magnetic moment is expressed as @xmath163 the above relation implies that the effective potential @xmath49 is a potential energy for a local magnetic moment @xmath164 . in the numerical calculations , we took into account the dynamical corrections up to the second order ( @xmath165 ) exactly , and the higher - order terms up to the fourth order within the asymptotic approximation . summation with respect to @xmath166 in eqs . ( [ edyn2 ] ) and ( [ gimp ] ) was taken up to @xmath167 for @xmath168 and @xmath169 , and up to @xmath170 for @xmath171 . when we solve the cpa equation ( [ dcpa3 ] ) , we adopted a decoupling approximation to the thermal average of impurity green function @xcite for simplicity , _ i.e. _ , @xmath172 here we wrote the static exchange field @xmath61 as @xmath173 in order to make the decoupling approximation clearer . the approximation is correct up to the second moment ( _ i.e. _ , @xmath174 ) and arrows us to describe the thermal spin fluctuations in a simpler way . on the other hand , we adopted a diagonal approximation @xcite to the coherent green function at the r.h.s . ( [ dcpa3 ] ) . @xmath175 here @xmath176 is the local density of states for the lda band calculation , and @xmath177 . the approximation partly takes into account the effects of hybridization between different @xmath152 blocks in the nonmagnetic state , but neglects the effects via spin polarization . the cpa equation ( [ dcpa3 ] ) with use of the decoupling approximation ( [ gapprox ] ) yields an approximate solution to the full cpa equation . for the calculations of the single - particle densities of states , we adopted the following average @xmath178-matrix approximation @xcite ( ata ) after we solved eq . ( [ dcpa3 ] ) with use of the decoupling approximation ( [ gapprox ] ) . @xmath179 here the coherent potential in the decoupling approximation is used at the r.h.s . , but the full average @xmath156 of the impurity green function is taken . the ata is a one - shot correction to the full cpa . the coherent potential @xmath180 on the real axis @xmath181 is then calculated by using the pad numerical analytic continuation method @xcite . here @xmath182 is an infinitesimal positive number . the densities of states ( dos ) as the single - particle excitations , @xmath183 are calculated from the relation , @xmath184 spin - resolved @xmath3 densities of states ( dos ) of fe at @xmath185 for atomic @xmath0 ( dashed curve for the up - spin dos , dotted curve for the down - spin dos ) and screened @xmath0 ( solid curve for the up - spin dos , dot - dashed curve for the down - spin dos ) . ] in the numerical calculations , we adopted the lattice constants used by andersen _ et . @xcite , and performed the lda calculations with use of the barth - hedin exchange - correlation potential to make the tb - lmto hamiltonian ( [ h0 ] ) . for fe and ni , we adopted average coulomb interaction parameters @xmath186 and the average exchange interactions @xmath187 used by anisimov _ et . @xcite , and for co we adopted @xmath186 obtained by bandyopadhyay _ et . al . _ @xcite and @xmath187 obtained by the hartree - fock atomic calculations @xcite ; @xmath188 ry and @xmath189 ry for fe , @xmath190 ry and @xmath191 ry for fcc co , and @xmath192 ry and @xmath189 ry for ni . the exchange interaction energies mentioned above are basically atomic ones . as for the screened exchange interaction energy , we adopted 70% values @xcite of these atomic @xmath187 ; @xmath193 ry for fe , @xmath194 ry for fcc co , and @xmath193 ry for ni . the intra - orbital coulomb interaction @xmath30 , inter - orbital coulomb interaction @xmath31 , and the exchange interaction energy parameter @xmath0 were calculated from @xmath186 and @xmath187 as @xmath195 , @xmath196 , and @xmath197 , using the relation @xmath198 . we present in fig . 1 the @xmath3 densities of states ( dos ) in the ferromagnetic fe . in the case of atomic @xmath10 ry , the dos for up - spin electrons consists of two peaks , the main peak due to e@xmath199 electrons at @xmath200 ry and the second peak due to t@xmath201 electrons at @xmath202 ry . we also find small humps at @xmath203 ry and 0.0 ry . these humps are caused by both e@xmath199 and t@xmath201 electrons . the dos for down - spin electrons also show the structure consisting of the main peak at @xmath204 ry mainly due to e@xmath199 electrons and the second peak at @xmath205 ry mainly due to t@xmath201 electrons . we find a small hump at @xmath206 ry due to t@xmath201 electrons . when we adopt the screened @xmath207 ry , the peaks of the up - spin dos shift up towards the fermi level by about 0.05 ry , and those of the down - spin dos shift down . it indicates the reduction of the magnetization . the hump at @xmath203 ry shifts up and loses its weight , while the hump at @xmath206 ry disappears . accordingly the spectral weights of the two peaks are enhanced . the disappearance of the humps implies that the humps originates in the multiplet excitations due to strong @xmath0 . the resulting total dos are shown in fig . the screened @xmath0 enhances the main peak and shifts it to the fermi level . the dos obtained by the screened @xmath0 seems to explain better the xps data at room temperature @xcite . total dos of fe for atomic @xmath0 ( dashed curve ) and screened @xmath0 ( solid curve ) at @xmath185 . the xps @xcite and bis @xcite data at room temperature are shown by dotted curves . ] solving the cpa equation ( [ dcpa3 ] ) self - consistently at each temperature , we obtain the magnetization vs temperature ( @xmath208 ) curves for fe as shown in fig . 3 . the ground - state magnetizations @xmath209 obtained by an extrapolation of the @xmath208 curves are 2.58 @xmath4 for @xmath10 ry and 2.39 @xmath4 for @xmath210 ry . the latter is in better agreement with experimental value @xcite 2.22 @xmath4 , though it is still somewhat overestimated . the reduction of @xmath209 with decreasing @xmath0 is explained by the fact that the hund - rule coupling @xmath0 which parallels the spins of electrons in each atom competes with the disordering due to electron hopping and thus @xmath209 is not saturated . we also find the reduction of the amplitude of local moment @xmath211 by 0.066 @xmath4 when the screened @xmath0 is used , as seen in fig . magnetization vs temperature ( @xmath208 ) curves , inverse susceptibilities @xmath212 , and amplitudes of local moments @xmath211 of fe for atomic @xmath0 ( dashed curves ) and screened @xmath0 ( solid curves ) . corresponding curves in the static approximation are drawn by dotted curves . the curve with higher ( lower ) @xmath1 corresponds to the screened ( atomic ) @xmath0 . experimental data @xcite of @xmath213 are shown by @xmath214 . ] with increasing temperature , the magnetization for screened @xmath0 shows less temperature dependence at low temperatures and rapidly decreases near @xmath1 . resulting @xmath1 is higher than that obtained by the atomic @xmath0 . basically the curie temperature @xmath1 is determined by the ratio of the magnetic energy to the magnetic entropy . the enhancement of @xmath1 with reducing @xmath0 in fe is attributed to the reduction of the magnetic entropy with the collapse of the hund rule coupling . further reduction of @xmath0 should decrease @xmath1 because of the reduction of magnetic energy . calculated susceptibilities of fe follow the curie - weiss law as shown in fig . we find the effective bohr magneton numbers @xmath215 @xmath4 for @xmath10 ry and @xmath216 @xmath4 for @xmath207 ry , which are compared with the experimental value @xcite 3.2 @xmath4 . the screened @xmath0 underestimates the effective bohr magneton number . we have also calculated the effective bohr magneton number for fcc fe at high temperatures ( @xmath217 k ) , and obtained @xmath218 @xmath4 for @xmath10 ry and @xmath219 @xmath4 for @xmath207 ry after having included the volume effects on the susceptibilities . these values should be compared with the experimental value @xcite 7.0 @xmath4 . the results again show that the screened @xmath0 value underestimates @xmath220 . present results indicate that the high - temperature properties of fe are explained better by the atomic @xmath0 , while the low - temperature ones are explained by the screened @xmath0 . total dos of fcc co for atomic @xmath0 ( dashed curve ) and screened @xmath0 ( solid curve ) at @xmath221 . the xps @xcite and bis @xcite data at room temperature for hcp co are shown by dotted curves . ] in the case of fcc co , the @xmath3 dos for up - spin electrons shift up by 0.04 ry and those for down - spin electrons shift down by 0.01 ry when the atomic @xmath0 is replaced by the screened @xmath0 . the total dos shifts towards the fermi level and the peak at @xmath222 ry is considerably enhanced as shown in fig . it should be noted that the xps @xcite and bis @xcite data are obtained for the hcp co at room temperature , thus they are not able to be compared directly with the present results for fcc co. nevertheless , we find the peak position of the bis data is in accordance with the hump position in the dos . the band width of the dos seems to be larger than that expected from the xps data . magnetization vs temperature ( @xmath208 ) curves , inverse susceptibilities @xmath212 , and amplitudes of local moments @xmath211 of fcc co for atomic @xmath0 ( dashed curves ) and screened @xmath0 ( solid curves ) . corresponding curves in the static approximation are drawn by dotted curves . the curve with higher ( lower ) @xmath1 corresponds to the atomic ( screened ) @xmath0 . experimental data @xcite of @xmath213 for fcc co are shown by open squares and dashed line . ] we present the magnetization vs temperature curves of fcc co in fig . the ground - state magnetizations obtained by an extrapolation are 1.82 @xmath4 for the atomic @xmath223 ry and 1.72 @xmath4 for the screened @xmath224 ry , respectively ; the latter obtained by the screened @xmath0 is in better agreement with the experimental one @xcite ( 1.74 @xmath4 ) . the amplitudes of local moments decrease by about 0.007 @xmath4 with the reduction of the hund - rule coupling by 0.021 ry . the magnetization for the screened @xmath0 decreases in the same way as in the atomic @xmath0 with increasing temperature ( see fig . 5 ) . the calculated @xmath1 is 2075 k for the screened @xmath225 ry , and is smaller than @xmath226 k for atomic @xmath223 ry . the former ( latter ) is 1.5 ( 1.8 ) times as large as the experimental value of fcc co ( @xmath227 k ) @xcite . we also calculated the susceptibilities . the inverse susceptibilities follow the curie - weiss law but show an upward convexity to some extent . calculated effective bohr magneton numbers are 3.0 @xmath4 in both cases being in good agreement with the experimental value @xcite 3.15 @xmath4 . they hardly change due to the reduction of @xmath0 . these results indicate that effects of the hund - rule coupling on the local magnetic moment are less significant as compared with the case of fe because the number of @xmath3 holes is rather small ( about 2.36 ) . total dos of ni for atomic @xmath0 ( dashed curve ) and screened @xmath0 ( solid curve ) at @xmath228 . the xps @xcite and bis @xcite data are shown by dotted curves . ] the dos for ni are presented in fig . they are shown for paramagnetic state because calculated @xmath1 of ni is lower than those in fe and co by a factor of 3 or 5 , and the calculations of dos at low temperatures below @xmath1 are not easy in the present method with use of the numerical analytic continuation . the paramagnetic dos are not sensitive to the choice of @xmath0 since the hund - rule coupling is not effective for the system with small number of @xmath3 holes ( @xmath229 ) . calculated @xmath3-band widths seem to be somewhat larger than those expected from the xps data . the same tendency was found also in fcc co ( see fig . the results suggest that the higher - order dynamical corrections are desired for the quantitative description of the dos in these systems . the magnetization vs temperature curves for ni are presented in fig . the ground - state magnetization obtained by an extrapolation is 0.64 @xmath4 for atomic @xmath10 ry and @xmath230 @xmath4 for screened @xmath207 ry . the experimental value @xcite 0.62 @xmath4 is in - between . the amplitude of local moment for @xmath10 ry is 1.93 @xmath4 , for example , at @xmath231 . the amplitude hardly changes even if the screened @xmath0 is adopted . this is because the hund - rule coupling is negligible for the system with small number of @xmath3 holes . magnetization vs temperature ( @xmath208 ) curves , inverse susceptibilities @xmath212 , and amplitudes of local moments @xmath211 of ni for atomic @xmath0 ( dashed curves ) and screened @xmath0 ( solid curves ) . corresponding curves in the static approximation are drawn by dotted curves . the curve with higher ( lower ) @xmath1 corresponds to the atomic ( screened ) @xmath0 . experimental data @xcite of @xmath213 are shown by @xmath214 . ] the curie temperatures in ni are strongly reduced by the dynamical effects as shown in fig . we obtained @xmath232 k for the atomic @xmath0 ( = 0.066 ry ) and 355 k for the screened @xmath0 (= 0.046 ry ) . the former is in good agreement with the observed @xmath233 k @xcite . the screened @xmath0 underestimates the curie temperature . the results indicate that the value of @xmath0 is significant for the spin polarization and the inter - site ferromagnetic coupling in ni . in order to examine the ambiguity of the coulomb interaction energy parameters , we performed the same calculations with use of different sets of @xmath234 : @xmath235 ry and ( 0.25 , 0.051 ) ry , where @xmath236 ry was obtained by bandyopadhyay _ et . @xcite and @xmath237 ry was taken from the hartree - fock atomic calculations @xcite . we obtained @xmath238 k for the former ( , _ i.e. _ , atomic @xmath0 ) and 325 k for the latter ( , _ i.e. _ , screened @xmath0 ) ; the results are not so sensitive to the choice of @xmath7 . calculated inverse susceptibilities are upwards convex , being in agreement with the experimental data @xcite for ni . effective bohr magneton numbers calculated at @xmath239 k are 1.6 @xmath4 for both the screened and the atomic @xmath0 . the results quantitatively agree with the experimental value @xcite 1.6 @xmath4 . we have calculated the magnetic properties of fe , co , and ni at finite temperatures on the basis of the first - principles dynamical cpa+ha , and investigated the effects of screened exchange interaction energy @xmath0 on the magnetic properties in these ferromagnets . when we apply the atomic @xmath0 , we obtained from the magnetization vs temperature curves the curie temperatures 1930k for fe , 2550k for fcc co , and 620k for ni . although calculated @xmath1 in ni is close to the experimental value @xcite 630k , those in fe and co are overestimated by a factor of 1.8 as compared with the experimental values , 1040k ( fe ) @xcite and 1388k ( co ) @xcite , respectively . in the present calculations using atomic @xmath0 , the ground state magnetizations obtained by an extrapolation are 2.58 @xmath4 ( fe ) , 1.82 @xmath4 ( fcc co ) , and 0.64 @xmath4 ( ni ) . calculated magnetizations for fe and fcc co are considerably larger than the experimental values , 2.22 @xmath4 ( fe ) @xcite and 1.74 @xmath4 ( co ) @xcite , while that for ni agrees well with the experiment ( 0.62 @xmath4 ) . the results indicate that the atomic @xmath0 overestimate the ferromagnetism of fe and fcc co at low temperatures . we have calculated the paramagnetic susceptibilities . calculated effective bohr magneton numbers , 3.0 @xmath4 ( fe ) , 3.0 @xmath4 ( co ) , and 1.6 @xmath4 ( ni ) , quantitatively explain the experimental data . the reduction of the exchange interaction energy by 30% due to screening by _ sp _ electrons weakens the hund - rule coupling which builds up the local magnetic moment on each atom . we found that it reduces the ground - state magnetization @xmath209 and the amplitudes of local moments in fe and co , especially the ground - state magnetizations 2.39 @xmath4 ( fe ) and 1.72 @xmath4 ( fcc co ) are in good agreement with the experiments . the screened @xmath0 reduces the exchange splitting between the up and down dos . the reduction of the splitting shifts the main peak in the spin - resolved dos to the fermi level and enhances the peak in case of fe and co. the screened @xmath0 reduces the humps in the dos below @xmath1 which originate in the multiplet - type of excitations associated with @xmath0 . the screened @xmath0 also weakens the inter - site magnetic couplings . it reduces the magnetization in ni , where the on - site hund - rule coupling is not significant because of a small number of @xmath3 holes . with reducing @xmath0 , the magnetic entropy and the magnetic energy decrease in general we found that the 30% screened @xmath0 enhances the curie temperature @xmath1 by 140 k in fe because of the reduction of the magnetic entropy , while it reduces @xmath1 in fcc co by 480 k because of the reduction of the magnetic energy . the curie temperatures of fe and co are overestimated by a factor of @xmath240 irrespective of the choice of @xmath0 in the present calculations . it should be attiributed to the magnetic short range order which is not taken into account in the present theory . in the case of the susceptibilities , the screened @xmath0 excessively reduces the effective bohr magneton number in fe where the hund - rule coupling builds up the on - site magnetic moment . by comparing both results of atomic @xmath0 and screened @xmath0 with the experimental data , we found that the screened @xmath0 explains better the ground - state magnetization @xmath209 and the dos at low temperatures in case of fe and co. but it yields the underestimate of @xmath209 in ni . the atomic @xmath0 seems to be better for the explanation of @xmath209 in ni . concerning the finite - temperature properties , we found that the screened @xmath0 improves the curie temperature in co , while the atomic @xmath0 yields better @xmath1 in case of fe and ni . the effective bohr magneton numbers are quantitatively explained by the atomic @xmath0 . the screened @xmath0 underestimates @xmath220 for both the bcc and fcc fe . the above - mentioned results indicate that only for fcc co the screened @xmath0 seems to improve both the low- and the high - temperature properties . in the case of fe , the high - temperature properties seem to be explained better by the atomic @xmath0 , while the low - temperature ones by the screened @xmath0 . in ni , both the low- and the high - temperature properties are explained better by the atomic @xmath0 . although we have to improve further the theory of dynamical cpa+ha at low - temperatures in order to obtain more solid conclusion , the present results obtained in the high - temperature region should be reasonable because the theory is an approach from the high - temperature limit and the magnetic short range order is not significant at high temperatures . the results for the effective bohr magneton numbers indicate that one has to take into account the temperature dependence of @xmath0 for the system like fe where the hund - rule coupling builds a large local magnetic moment but competes with the kinetic energy of @xmath3 electrons . it is plausible that spin fluctuations at finite temperatures break the screening of @xmath0 . the present results also suggest that the 30% screening of @xmath0 is an overestimate for ni . the range of applications of screened @xmath0 and its validity have to be examined further in the future investigations with use of more advanced theories . the present work is supported by grant - in - aid for scientific research ( 22540395 ) . numerical calculations have been partly carried out with use of the hitachi sr11000 in the supercomputer center , institute of solid state physics , university of tokyo . 99 p. hohenberg and w. kohn , phys . 136 , b864 ( 1964 ) . w. kohn and l.j . sham , phys . 140 , a1133 ( 1965 ) . parr and w. yang , _ density functional theory of atoms and molecules _ ( oxford university press , oxford , 1989 ) . u. von barth and l. hedin , j. phys . c * 5 * , 1629 ( 1972 ) . perdew and y. wang , phys . b * 33 * , 8800 ( 1986 ) . moruzzi , j.f . janak , and a.r . williams , _ calculated electronic properties of metals _ ( pergamon press , 1978 ) . h. danan , a. herr , and a.j.p . meyer , j. appl . * 39 * , 669 ( 1968 ) . besnus , a.j.p . meyer , and r. berninger , phys . lett . * 32 a * , 192 ( 1970 ) . p. bagno , o. jepsen , and o. gunnarsson , phys . rev . b 40 , 1997 ( 1989 ) . o. gunnarson , physica b * 91 * , 329 ( 1977 ) . staunton and b.l . gyorffy , phys . lett . * 69 * , 371 ( 1992 ) . a. arrott and j.e . noakes , phys . * 19 * , 786 ( 1967 ) . noakes , n.e . tornberg and a. arrott , j. appl . * 37 * , 1264 ( 1966 ) . y. kakehashi , adv . in phys . * 53 * , 497 ( 2004 ) ; phil . mag . * 86 * , 2603 ( 2006 ) . j. hubbard , phys . b * 19 * , 2626 ( 1979 ) ; * 20 * , 4584 ( 1979 ) ; * 23 * , 5974 ( 1981 ) . h. hasegawa , j. phys . * 46 * , 1504 ( 1979 ) ; * 49 * , 178 ( 1980 ) . r. l. stratonovich , dokl . sssr * 115 * , 1097 ( 1958 ) [ sov . phys . * 2 * , 416 ( 1958 ) ] . j. hubbard , phys . * 3 * , 77 ( 1959 ) . w. e. evenson , j. r. schrieffer , and s. q. wang , j. appl . * 41 * , 1199 ( 1970 ) ; j. r. schrieffer , w. e. evenson , and s. q. wang , j. phys . ( paris ) colloq . * 32 * , c1 - 19 ( 1971 ) . g. morandi , e. galleani dagliano , f. napoli , and c. f. ratto , adv * 23 * , 867 ( 1974 ) . p. soven , phys . rev . * 156 * , 809 ( 1967 ) . h. ehrenreich and l. m. schwarz , solid state physics , edited by h. ehrenreich , f. seitz , and d. turnbull ( academic , new york , 1976 ) , vol . 31 , p.150 . gutzwiller , phys . rev . lett . * 10 * , 159 ( 1963 ) . gutzwiller , phys . rev . * 134 * , a923 ( 1964 ) gutzwiller , phys . rev . * 137 * , a1726 ( 1965 ) . j. hubbard , proc . london a * 276 * , 238 ( 1963 ) . j. hubbard , proc . london a * 281 * , 401 ( 1964 ) . j. kanamori , prog . * 30 * , 275 ( 1963 ) . y. kakehashi and p. fulde , phys . b * 32 * , 1595 ( 1985 ) . h. hasegawa , j. phys . condensed matter * 1 * 9325 ( 1989 ) . y. kakehashi , phys . rev . b * 45 * , 7196 ( 1992 ) . y. kakehashi , phys . b * 65 * , 184420 ( 2002 ) . d.j . amit and c.m . bender , phys . b * 4 * , 3115 ( 1971 ) ; d.j . amit and h. keiter , j. low temp . * 11 * , 603 ( 1973 ) . dai xianxi , j. phys . condens . matter . * 3 * , 4389 ( 1991 ) . y. kakehashi , j. phys . jpn * 77 * , 094706 ( 2008 ) . anisimov , a.i . poteryaev , m.a . korotin , a.o . anokhin , and g. kotliar , j. phys . matter * 9 * , 7359 ( 1997 ) . o.k . andersen , o. jepsen , and g. krier , in _ methods of electronic structure calculations _ ed . by v. kumar , o.k . andersen , and a. mookerjee ( world scientific pub . , singapore , 1994 ) p. 63 . y. kakehashi , m.a.r . patoary , and t. tamashiro , j. phys . jpn . * 78 * , 093705 ( 2009 ) . y. kakehashi , m. atiqur r. patoary , and t. tamashiro , phys . b * 81 * , 245133 ( 2010 ) . anisimov , f. aryasetiawan , and a.i . lichtenstein , j. phys . matter * 9 * , 767 ( 1997 ) . t. bandyopadhyay and d.d . sarma , phys . b * 39 * , 3517 ( 1989 ) . mann , los alamos scientific laboratory rep . lasl-3690 ( 1967 ) . f. aryasetiawan , k. karlsson , o. jepsen , and schnberger , phys . b * 74 * , 125106 ( 2006 ) . t. miyake and f. aryasetiawan , phys . b * 77 * , 085122 ( 2008 ) . see for example , m. imada and t. miyake , cond - mat/1009.3851v1 ( 2010 ) . y. kakehashi , phys . b * 66 * , 104428 ( 2002 ) . s. hirooka and m. shimizu , j. phys . . jpn . * 43 * , 70 ( 1977 ) . e. mller - hartmann , z. phys . b * 74 * , 507 ( 1989 ) . m. jarrell , phys . * 69 * , 168 ( 1992 ) ; m. jarrell and h.r . krishnamurthy , phys . b * 63 * , 125102 ( 2001 ) . f.j . ohkawa , phys . b * 46 * , 9016 ( 1992 ) . a. georges and g. kotliar , phys . b * 45 * , 6479 ( 1992 ) ; a. georges and w. krauth , phys . b * 48 * , 7167 ( 1993 ) . a. georges , g. kotliar , w. krauth , and m. j. rozenberg , rev . phys . * 68 * , 13 ( 1996 ) . y. kakehashi and p. fulde , phys . b * 69 * , 045101 ( 2004 ) . o. miura and t. fujiwara , phys . b * 77 * , 195124 ( 2008 ) . a.i . lichtenstein and m.i . katsnelson , and g. kotliar , phys . lett . * 87 * , 067205 ( 2001 ) . y. kakehashi , j. phys . jpn . , * 50 * , 1505 ( 1981 ) ; j. phys . jpn . , * 50 * , 2251 ( 1981 ) . s. kirkpatrick , b. velick , and h. ehrenreich , phys . b * 1 * , 3250 ( 1970 ) . j. korringa , j. phys . solids * 7 * , 252 ( 1958 ) . h.j . vidberg and j.w . serene , j. low temp . * 29 * , 179 ( 1977 ) . a.g . narmonev and a.i . zakharov , phys . met . metallogr . * 65 * , 315 ( 1988 ) . w. speier , j.c . fuggle , r. zeller , b. ackermann , k. szot , f.u . hillebrecht , and m. campagna , phys . b * 30 * , 6921 ( 1984 ) . potter , proc . london * a 146 * , 362 ( 1934 ) . m. fallot , j. de phys . v , 153 ( 1944 ) . y. barnier , r. pauthenet , and l. nel , cobalt no . 21 ( 1963 ) ( cobalt information center ) . myers and w. sucksmith , proc . soc . * a 207 * , 427 ( 1951 ) . colvin and s. arajs , j. phys . solids * 26 * , 435 ( 1965 ) . p. weiss and r. forrer , ann . paris * 5 * , 153 ( 1926 ) .

we have investigated the magnetic properties of fe , co , and ni at finite temperatures on the basis of the first - principles dynamical coherent potential approximation ( cpa ) in order to clarify the role of the exchange interaction energy ( @xmath0 ) screened by _ sp _ electrons and its applicability to finite - temperature magnetism . with use of the atomic @xmath0 , we obtained the curie temperatures ( @xmath1 ) 1930 k for fe and 2550 k for fcc co , which are overestimated by a factor of 1.8 as compared with the experimental values , while we obtained @xmath2 k for ni being in good agreement with the experiment . calculated effective bohr magneton numbers also quantitatively agree with the experimental values . by comparing the results with those obtained by the screened @xmath0 and by comparing them with the experiments , we found that the screened @xmath0 , which are reduced by 30% as compared with the atomic ones , improve the ground - state magnetizations and densities of states at low temperatures in fe and fcc co , as well as @xmath1 in fcc co. but the screened @xmath0 yield worse results for finite temperature properties of fe and underestimate both the ground - state magnetization and @xmath1 in case of ni . we discuss possible origins for these inconsistencies .

introduction first-principles dynamical cpa numerical results summary

galaxies in which the star - forming gas has a metallicity smaller than a tenth of the solar metallicity are known as extremely metal - poor galaxies . since metals are primarily produced by stars , the star - forming gas in xmps is chemically primitive . they represent less than 0.1% of the objects in galaxy catalogs ( e.g. , * ? ? ? xmps can be divided into two types , according to their absolute magnitude . the first type is ultra - faint late - type galaxies , with @xmath2 . since galaxies follow a well known luminosity - metallicity relation ( e.g. , * ? ? ? * ) , those with @xmath2 should be xmps @xcite . such xmps are particularly rare in surveys ( e.g. , * ? ? ? * ; * ? ? ? the second type , with @xmath3 , corresponds to low metallicity outliers of the luminosity - metallicity relation . with high surface brightness , they dominate the catalogs of xmps , and they are the subject of our study . the evolutionary pathways of the two types probably differ . in this paper the term xmp refers exclusively to the second type . in addition to having chemically primitive star - forming regions , xmps also show signs of being structurally primitive , as if they were disks in early stages of assembly , with massive off - center starbursts providing them with a characteristic tadpole shape , with slowly rotating bodies , and with large spatially - unresolved turbulent motions . reinforcing such a primitive character , their chemical composition is often non - uniform , with the lowest metallicities mostly in the regions of intense star formation . the existence of chemical inhomogeneities is particularly revealing because the timescale for mixing in disk galaxies is short , of the order of a fraction of the rotational period ( e.g. , * ? ? ? * ; * ? ? ? * ) . it implies that the metal - poor gas in xmps was recently accreted from a nearly pristine cloud , very much in line with the expected cosmic cold - flow accretion predicted to build disk galaxies @xcite , but which has been so difficult to test observationally . the gas expected in cosmological gas accretion events is described in detail by . thus , xmps seem to be local galaxies that are growing through the physical process that created and shaped disk galaxies in the early universe . they open up a gateway to study this fundamental process with unprecedented detail at low redshift . the presence of chemical inhomogeneities in most xmps is the touchstone of the whole scenario , since it unequivocally shows the star formation to be feeding from external metal - poor gas . in order to put the presence of these inhomogeneities on a firm observational basis , we measured the oxygen abundance along the major axes of ten xmps ( * ? ? ? * paper i ) , using a method based on models consistent with the direct method @xcite . in nine out of the ten cases , sharp metallicity drops were found , leaving little doubt as to the existence and ubiquity of the drops . the present paper follows up on paper i , and measures the kinematic properties of these xmps . the spectra used to infer metallicities in paper i do not possess enough spectral resolution to carry out kinematic studies . therefore , a new set of observations was obtained . for this reason , the overlap with the original xmp sample is not complete . however , most of the original galaxies are examined here and show evidence for the star - forming regions having coherent motions , which also hints at the external origin of the galaxy gas . in addition , the h@xmath1 line presents multiple side lobes that seem to trace feedback processes of the star formation on the ism ( interstellar medium ) of the galaxies . since we use h@xmath1 , our measurements characterize the properties of the ionized gas near the sites of star formation . the paper is organized as follows : sect . [ observations ] describes the observations and reduction , emphasizing the aspects required to obtain the spectral resolution needed for the kinematic analysis . [ explaining_rules ] describes how physical parameters are obtained from the calibrated spectra . it is divided into various parts ; one for the main h@xmath1 signal ( sect . [ main ] ) , and three for the secondary lobes ( sects . [ lobes ] , [ mass_loss_rate ] , and [ sec : mass_bh ] ) . the overall properties , such as rotation , turbulent motions , and global expansion , are presented in sect . [ sect : rot_turb ] . in this section , we also study the uniformity of the n / o ratio . the properties of the multiple secondary components bracketing the main h@xmath1 component are presented and interpreted in sect . [ multiple_comp ] . they appear to trace gas swept by the star formation process , and in sect . [ sec : mydiscussion ] we predict the properties when this gas reaches the cgm ( circum - galactic medium ) . our results are summarized and discussed in sect . [ conclusions ] . the emission expected from an expanding dusty shell of gas is formulated in appendix [ appendix ] . for the study of the kinematics of xmps , we selected a sample of 9 galaxies , 8 of which were previously studied in paper i ( see table [ table_obs ] ) , and thus have spatially resolved measurements of oxygen abundance . this time the galaxies were observed with the isis instrument on the 4.2 m william herschel telescope ( wht ) , using a high resolution grating . the slit of the spectrograph was positioned along the major axis of the galaxies ( see fig . [ fig : all_images ] ) , overlaying the slit used in paper i. the slit width was selected to be 1 arcsec , according to the seeing at the telescope and also matching the value of our previous observation of these galaxies in paper i. the spectrograph has two arms , the red and the blue arm , each one with its own detector . all galaxies were observed in both arms , except for hs0822 , where we only have the red spectrum . a @xmath4 binning of the original images was performed to reduce oversampling . after binning , the dispersion turns out to be 0.52 pix@xmath0 for the red arm and 0.46 pix@xmath0 for the blue arm . similarly , the spatial scale is 0.44 arcsecpix@xmath0 for the red arm and 0.40 arcsecpix@xmath0 for the blue arm . the wavelength coverage for the blue arm is 940 centered at h@xmath5 ( from 4390 to 5330 ) , whereas the read arm covers 1055 around h@xmath1 ( from 6035 to 7090 ) . the total exposure time per target was approximately two hours ( table [ table_obs ] ) , divided in snapshots of 20 minutes as a compromise to simultaneously minimize cosmic - ray contamination and readout noise . lcccccccc our i d & galaxy name & obs . date & seeing@xmath6 & @xmath7@xmath8 & @xmath9@xmath10 & scale@xmath10 & exp . time&@xmath11 + & & dd / mm / yyyy & [ arcsec ] & [ kms@xmath0 ] & [ mpc ] & [ pcarcsec@xmath0 ] & [ s]&[m@xmath13 + ml4 & j@xmath14 & 16/08/2012 & 0.5 & 38.1 @xmath15 2.7 & 125 @xmath15 9 & 608 & 3600 & 8.33 @xmath15 0.33 + hs0822 & j@xmath16 & 02/02/2013 & 1.4 & 37.8 @xmath15 1.6 & 9.6 @xmath15 0.7 & 47 & 7200 & 6.04 @xmath15 0.03 + sbs0 & j@xmath17 & 15/03/2015 & 0.9 & 41.1 @xmath15 3.2 & 23.1 @xmath15 1.6 & 112 & 4800 & 7.05 @xmath15 0.05 + sbs1 & j@xmath18 & 15/03/2015&1.2 & 40.4 @xmath153.6 & 22.5 @xmath15 1.6 & 109 & 7200 & 7.53 @xmath15 0.48 + ml16 & j@xmath19 & 15/03/2015 & 1.0 & 40.1 @xmath15 4.0 & 23.6 @xmath15 1.7 & 114 & 7200 & 6.71 @xmath15 0.07 + um461 & j@xmath20 & 15/03/2015 & 1.2 & 40.6 @xmath15 4.0 & 12.6 @xmath15 0.9 & 61 & 6400 & 6.7 @xmath15 1 + n241@xmath21 & j@xmath22 & 15/03/2015 & 1.0 & 40.3 @xmath15 4.0 & 10.2 @xmath15 0.7 & 49 & 6000 & 6.52 @xmath15 0.07 + sbs2104 & j@xmath23 & 16/08/2012 & 0.5 & 39.0 @xmath15 2.2 & 20.1 @xmath15 1.7 & 97 & 7200 & 6.19 @xmath15 0.05 + ml32 & j@xmath24 & 16/08/2012 & 0.5 & 42 @xmath15 5 & 138 @xmath15 10 & 666 & 7200 & 8.39 @xmath15 0.33 + [ cols= " < " , ] [ tab : summary_lobes ] the velocities and the ratio between the secondary components and the central component , @xmath25 , are provided directly by the multi - gaussian fit described in sect . [ lobes ] . in the case of paired components , the velocity is the mean of the absolute value of the two velocities , and @xmath25 is the total @xmath25 of the pair . the estimates also depend on the temperature and electron density of the emitting plasma , as well as on the relative thickness of the shell , @xmath26 , and the filling factor , @xmath27 . the electron temperature is taken from the data used to determine oxygen abundance in paper i. specifically , we use the mean over the main star - forming region of the galaxy . the electron density has been inferred from the ratio of the emission lines [ sii]@xmath286716 and [ sii]@xmath286731 , following the standard procedure ( e.g. , * ? ? ? * ) , using the mean value over the whole star - forming region . these two lines are present within the spectral range observed with the wht . electron densities and temperatures are listed in table [ tab : mean_data ] . as for the thickness of the shell , we use @xmath29 ( see sect . the filling factor , @xmath27 , is assumed to be one . the velocities of the weak secondary components are typically in excess of 100 km s@xmath0 ( see table [ tab : summary_lobes ] ) . these velocities are very large compared with the rotational and turbulent velocities of the galaxies ( sect . [ sect : rot_turb ] ) , and , hence , are almost certainly larger than the escape velocity of the galaxy , expected to be of the order of the rotational plus turbulent velocity in the disk ( e.g. , * ? ? ? assuming that the components in the wings correspond to outflows in expanding shells , it is still unclear whether or not the involved mass will finally escape from the galaxy . it all depends on the ism of the host galaxy , and the pressure it exerts on the expanding shell . the issue is analyzed in sect . [ sec : interpretation ] , with the conclusion that the gas will likely escape . the mass involved in these motions is given in table [ tab : summary_lobes ] and summarized in fig . [ fig : mass_loading ] . . the color code and symbols have the same meaning as in the previous figure , which is indicated in the inset . , scaledwidth=50.0% ] the masses are in the range between @xmath30 and @xmath31 , and they represent a few percent of the total gas mass in the hii region ( the parameter @xmath25 in table [ tab : summary_lobes ] gives the flux ratio between the secondary and the main emission lobe , and it has been taken as a proxy for the mass ratio ; see eqs . [ [ epsilon ] ] and [ [ massloss2 ] ] ) . these masses are very small compared to the stellar masses and the dynamical masses of the entire galaxies ( tables [ table_obs ] and [ table_rotation ] ) . figure [ fig : mass_loading ] shows that the mass loss rate scales with the gas mass , since the velocities and radii of the different components and galaxies are similar . the weak components have associated mass loss rates between @xmath32 and 1 m@xmath33 yr@xmath0 . they are very significant compared with the sfrs , as is attested in fig . [ fig : mass_loading2 ] , which shows the mass loading factor , @xmath34 . most mass loading factors are in excess of one , meaning that for every mass unit transformed into stars , several mass units are swept away . whether this mass escapes the hii region , and later on , the galaxy , we can not know from our observations . however , cosmological numerical simulations predict large mass loading factors in low mass galaxies ( 10 is not uncommon ; see , e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) . ) versus gas mass , for the weak components observed in the wings of h@xmath1 . all factors are larger than one , meaning that for every unit mass transformed in stars , several of units are swept away by the expansion . the color code and symbols have the same meaning as in previous figures , which is indicated in the inset . , scaledwidth=50.0% ] mass loading factors observed in high redshift objects are generally lower than the ones we infer ( say , of the order of 2 , e.g. , * ? ? ? * ; * ? ? ? * ) , likely because they refer to massive galaxies ( @xmath35 ) . however , it is not uncommon to find factors up to 10 in local dwarfs . a caveat is in order , though . we are assuming the density of the shell to be the same as the mean density of the star - forming region . if the expanding shell builds up mass by sweeping gas around the explosion site , its density is expected to be higher than the mean density of the region . this increase of density would reduce our mass loading factor estimates ( see eq . [ [ massloss2 ] ] ) . figure [ fig : kinetic ] shows the kinetic energy associated with the weak components . as with the mass loss rate , it scales with the mass of gas that is involved in the motion . figure [ fig : kinetic ] also shows the energy typically associated with a core - collapse sn explosion , of the order of @xmath36ergs ( e.g. , * ? ? ? the energies involved in the observed motions are usually around this value , albeit smaller ; ml32 is the only exception . therefore , energy - wise , the components may be tracing shells produced by individual sn explosions . ml32 would require some 10 sn exploding simultaneously . . the line of @xmath36erg corresponds to the typical kinetic energy released by a core - collapse sn , characteristic of massive stars . the color code and symbols have the same meaning as in previous figures , which is indicated in the inset . , scaledwidth=50.0% ] depending on the imf ( initial mass function ) , population synthesis models predict a sn rate between 0.02 and 0.005 yr@xmath0 for a constant sfr=1 m@xmath33 yr@xmath0 @xcite . using a value in between these two extremes , say 0.01 yr@xmath0/@xmath37 , the sfrs of our star - forming regions ( table [ tab : mean_data ] ) allow us to estimate the expected sn rates assuming the observed sfrs to be constant . the predicted sn rate turns out to be between @xmath38 and @xmath39 , which corresponds to a time - lag between sn explosions ( @xmath40 , defined as the inverse of the sn rate ) from 1000 yr to 0.1 myr . these rates suffice to maintain the observed signals . figure [ fig : size_age ] provides @xmath41 for individual shells . @xmath42 in half of the cases , which implies expecting more than one sn signal at a time , as we observe . ) versus @xmath43 of the shell relative to the expected time - lag between sn explosions ( @xmath40 ) . the color code and symbols have the same meaning as in previous figures , which is indicated in the inset . , scaledwidth=50.0% ] ages are estimated from the measured velocity and the estimated radius of the shell ( eq . [ [ eq : age_shell ] ] ) . these radii are small compared to the size of the star - forming regions . figure [ fig : size_age ] shows the radius relative to the effective radius of the region , with the typical ratio being around 10% or smaller . therefore , it is important to keep in mind , for interpreting the weak emission components in the wings of h@xmath1 , that there are one or several ( probably non - concentric ) expanding shells embedded in a large star - forming region . this substructure is common in giant hii regions , such as 30 doradus in the large magellanic cloud ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? shells and holes appear in hubble space telescope ( hst ) images of the large starburst in kiso 5639 , a tadpole galaxy very similar to the xmps studied here @xcite . we do not see a dependence of the properties of the weak emission - line components on the mean metallicity of the starburst , except perhaps for the mass loading factor . figure [ fig : metal1 ] shows the scatter plot of metallicity versus mass loading factor , and it hints at an increase of the factor with increasing metallicity . the trend may be coincidental , due to the small number of points in the plot . however , it may also indicate the contribution of stellar winds to the energy and momentum driving the observed expansion . the winds of massive stars can provide a total kynetic energy comparable to the energy released during their sn explosions ( e.g. , * ? ? ? * ) , and these radiation - pressure driven winds have mass loss rates that increase with increasing metallicity ( e.g. , * ? ? ? * ) . the metallicities represented in fig . [ fig : metal1 ] correspond to the metallicity averaged over the starburst and , thus , due to the variation of metallicity across the star - forming region , their value is slightly higher than the values quoted in paper i. ml16 , the point of highest metallicity in fig . [ fig : metal1 ] , is also the galaxy with no detected metallicity inhomogeneity in paper i. one can think of several alternatives to explain the existence of multiple paired emission - line components in the wings of h@xmath1 . the previous section analyzes the possibility of being produced by spatially - unresolved expanding shells driven by sn explosions . here we discuss other options , such as ( 1 ) spatially - unresolved rotating structures , ( 2 ) spatially - resolved expanding shells , ( 3 ) bipolar outflows , ( 4 ) blown out spherical expansion leading to expanding rings , and ( 5 ) spatially - unresolved expanding shells driven by bh accretion feedback . ( 1 ) _ spatially - unresolved rotating disk - like structures . _ rotation naturally produces double - horned profiles ( e.g. , * ? ? ? therefore , one of the possibilities to explain the existence of blue and red paired emission peaks is the presence of a gaseous structure rotating around a massive object . the mass of the required central object can be estimated from the size of the rotation structure and the velocity ( sect . [ sec : mass_bh ] ) . figure [ fig : bh_mass ] shows that the mass of the central object needs to be around three orders of magnitude larger than the gas mass involved in the motions , with central masses , @xmath44 , ranging from @xmath45@xmath46m@xmath33 . this mass turns out to be of the order of the stellar mass of the whole galaxy ( table [ table_obs ] ) , which rules out the stellar nature of the required central concentration . if , on the other hand , it were a bh , then it would have to have @xmath47 , which is completely out of the so - called magorrian - relationship between the central bh mass and the stellar mass of the galaxy , for which @xmath48 . thus , the exotic nature of the required central mass disfavors the interpretation of the weak emission in the wings of h@xmath1 as caused by a disk orbiting around a massive central object , such as a bh . \(2 ) _ spatially - resolved expanding shells . _ the inner part of a spatially resolved expanding shell also produces double peak emission - line profiles , since the outer parts of the shell , contributing to the line core , are excluded ( e.g. , * ? ? ? we often see several pairs of emission peaks ( fig . [ fig_lobes]a ) , therefore , this interpretation requires the existence of several resolved concentric shells , which is unrealistic since the first shell would have already swept the ism around the center of the expansion . moreover , the size of the shell is estimated to be smaller than the typical spatial resolution , so that the expanding regions can not be spatially resolved . \(3 ) _ bipolar outflows . _ insofar as they comply with the energetics required to explain the observed motions , bipolar flows can not be easily discarded . they create two - horned profiles , and several of them coexisting in the resolution element could explain the presence of multiple paired components . in fact , bipolar flows are extreme cases of non - spherically - symmetric shells , where double - horned line profiles appear naturally ( e.g. , * ? ? ? ( 4 ) _ blown out spherical expansion leading to expanding rings . _ if the radius of the shell exceeds the scale height of the disk of the galaxy , it may break up , leading to a ring - like expanding structure that produces two - horned profiles when observed edge - on . this possibility can be discarded since the ring would have to be as large as the disk thickness , which is incompatible with the small size of the expanding structures ( table [ tab : summary_lobes ] ) . ( 5 ) _ spatially - unresolved expanding shells driven by bh accretion feedback . _ energywise , intermediate mass bhs of @xmath49 can also power the expanding shells that we characterize in sect . [ sec : mult1 ] ( see eq . [ [ eq : bh_mass2 ] ] ) . however , the observations also disfavor this option for a number of reasons . it is unclear how a primordial bh seed can coincide spatially with a star - forming region unless there are many such bhs lurking in the galaxy , or the bh is born in situ . @xmath49 bhs can be produced by direct collapse in the early universe , but this explanation can be discarded , since they are very rare ( e.g. , * ? ? ? * ) and many of them are needed . on the other hand , @xmath50 bhs are expected to be more common , left as remnants from population iii stars formed in the early universe or even today ( e.g. , * ? ? ? . however , this alternative is also very unlikely , since hii regions are short - lived ( a few tens of myr ; e.g. , * ? ? ? * ) , and a bh requires several hundred myr of continuous feeding to grow from @xmath51 to @xmath49 . finally , accreting bhs are expected to be luminous x - ray sources . five of our xmps were observed with the x - ray satellite chandra : hs0822 , sbs1 , n241 , sbs2109 , and um 461 @xcite . only one of them shows a signal above the noise , namely , sbs1 , with an x - ray luminosity around @xmath52 . this level of emission is consistent with young sne . independently of the details on whether the shells are symmetric , or if they have internal extinction , the expanding shell model appears to be the best explanation to reproduce the weak emission features observed in the h@xmath1 wings . one of the consequences of this interpretation is that the mass loading factor that we infer may be reflecting the feedback of the star formation process on the medium in , and around , the galaxy . the measured mass loading factor is very large , of the order of ten or larger ( fig . [ fig : mass_loading2 ] ) , which implies that for every solar mass of gas transformed into stars , at least ten times more gas is swept away from the region . how far this swept gas goes is impossible to tell from the data alone . however , a significant fraction may escape from the galaxy to the cgm and even the intergalactic medium ( igm ) . the speeds , larger than 100 kms@xmath0 and up to 400 kms@xmath0 ( table [ tab : summary_lobes ] ) , likely exceed the escape velocity of the galaxy . the final fate of the gas depends on whether the gas pressure in the ism of the galaxy suffices to slow down and confine the expansion . even though we do not know the gas pressure of the ism , it must be similar to the turbulent pressure in the star - forming region , so that the region is neither expanding nor shrinking considerably ( the moderate global expansion observed in some cases is analyzed in the paragraph at the end of the section ) . the turbulent pressure scales with the square of the turbulent speed , and so does the pressure exerted by the shell on the ism . the measured turbulent velocities are of the order of 25 kms@xmath0 ( see table [ tab : mean_data ] , and keep in mind that the quoted values are fwhm velocities rather than velocity dispersions ) , hence , provided that the hii region and the shell have similar densities , the ism will not be able to confine a expanding shell with a speed of a few hundred kms@xmath0 . a significant part of the matter in the shell may eventually escape from the galaxy . @xcite made a qualitative estimate of the conditions for starburst - driven outflows to escape , finding that the gas flows out from galaxies with rotation smaller than 130 kms@xmath0 , i.e. , from our xmps . under the hypothesis of expanding shell , it is possible to compute the column density to be expected if the shell material reaches the cgm . assuming mass conservation , and if the shell geometry is maintained , the typical column density when the shell expands to a radius @xmath53 is , @xmath54 ^ 2 , \label{eq : column_density}\ ] ] where the electron density , @xmath55 , and the radius , @xmath56 , are those measured in the star - forming regions . note that for the typical sizes and densities of our shells , and with expansion factors of the order of @xmath57 that transform pc - size shells into intergalactic size structures ( from @xmath58pc to @xmath59kpc ) , the column densities predicted by eq . ( [ eq : column_density ] ) are of the order of @xmath60 . this gas would be undetectable with current instrumentation , even if it were fully neutral . for example , a high - sensitivity survey such as hipass sets the limit at @xmath61 @xcite , and values of @xmath62 will be achieved by ska after a 1000 hour integration @xcite . moreover , at these low column densities the intergalactic ultraviolet ( uv ) background will keep the gas fully ionized , which makes it even more elusive . how long would it take for the gas to reach the cgm ? assuming that the original velocities are maintained , the time to reach the cgm is the expansion factor , @xmath63 , times the @xmath43 in eq . ( [ eq : age_shell ] ) . given the @xmath43 in table [ tab : summary_lobes ] and @xmath64 , the time - scale is of the order of 10 myr , which is short compared to the time - scale for galaxy evolution . in summary , the shells that we observe in the star - forming regions are expected to reach the cgm , and may do it in a reasonable time - scale . unfortunately , they would be undetectable with current instrumentation unless the expansion is more moderate than expected according to eq . ( [ eq : column_density ] ) . we show in sect . [ sect : rot_turb ] that some of the star - forming regions may be undergoing a global expansion , with a moderate velocity of around 15 kms@xmath0 . the mass loss rate associated with this global expansion , @xmath65 , can also be estimated using eq . ( [ massloss ] ) , except that this time the global mass of the starburst , @xmath66 , and the radius of the starburst , @xmath67 , have to be considered . a simple substitution renders @xmath65 in terms of the mass loss rate inferred for one of the weak components in the wings of h@xmath1 , i.e. , @xmath68 where we have assumed @xmath69 , and @xmath70 . using typical values for these parameters , @xmath71 ( table [ tab : summary_lobes ] ) , @xmath72 ( table [ tab : summary_lobes ] ) , and @xmath73 ( figure [ fig : size_age ] ) , one finds that the two mass loss rates are similar , i.e. , @xmath74 one can repeat the exercise for the kinetic energy involved in the global expansion ( eq . [ [ eq : kinetic ] ] ) , which scales with the mass ratio and the square of the expansion velocities . using the typical values given above , the two kinetic energies turn out to be similar too . this rough estimate leads to the conclusion that the mass loss rate and the kinetic energy involved in the global expansion of the star - forming region are similar to the values inferred for the emission features observed in the wings of h@xmath1 . the question arises as to whether the global expansion unbounds the gas in the starburst , and so may contribute to the global mass loss of the xmps . as was the case with the small expanding shells discussed above , the answer to the question is very uncertain , and can not be addressed resorting to our observational data alone . the fate of the gas depends on the unknown value of the ism pressure . in contrast with the small shells , the expansion velocities are significantly smaller than the expected escape velocity . moreover , the turbulent velocity of the galaxies is typically 4050 kms@xmath0 , and , hence , comparable , but larger , than the turbulent velocity in the star - forming region ( see fig . [ fig : dispersion ] ) . only if the density in the ism of the xmps is much lower than the density in the starburst , the starburst may be overpressureized and the mild observed expansion can drive gas through the ism to large distances . the xmp galaxies analyzed in paper i are characterized by having a large star - forming region with a metal content lower than the rest of the galaxy . the presence of such chemical inhomogeneities suggests cosmological gas accretion as predicted in numerical simulations of galaxy formation . cosmological gas accretion induces off - center giant star - forming clumps that gradually migrate toward the center of the galaxy disks @xcite . the star - forming clumps may be born in - situ or ex - situ . in the first case , the accreted gas builds up the gas reservoir in the disk to a point where disk instabilities set in and trigger star formation . in the second case , already formed clumps are incorporated into the disk . they may come with stars and dark matter , and thus , they are often indistinguishable from gas - rich minor mergers @xcite . since they feed from metal - poor gas , the star - forming clumps are more metal - poor than their immediate surroundings @xcite . in order to complement the chemical analysis carried out in paper i , this second paper studies the kinematic properties of the ionized gas forming stars in the xmp galaxies . we study nine objects ( table [ table_obs ] ) with a spectral resolution of the order of 7500 , equivalent to 40 kms@xmath0 in h@xmath1 . most xmps have a measurable rotation velocity , with a rotation amplitude which is only a few tens of kms@xmath0 ( fig . [ fig : rot_curve ] ) . all rotation curves present small - scale velocity irregularities , often with an amplitude comparable to the velocity gradient across the whole galaxy . on top of the large - scale motion , the galaxies also present turbulent motions of typically 50 kms@xmath0 fwhm ( see fig . [ fig : dispersion ] and table [ tab : mean_data ] ) , and , therefore , larger than the rotational velocities . observations suggest that the main star - forming region of the galaxy ( fig . [ fig : all_images ] ) moves coherently within the host galaxy . the velocity is constant ( fig . [ fig : disp ] ) , so that the region moves as a single unit within the global rotation pattern of the galaxy . in addition , the velocity dispersion in some of these large star - forming clumps increases toward the center - side of the galaxy . the excess velocity dispersion on the center - side may indicate an intensification of the turbulence of the gas that collides with the ism of the host galaxy , as if the clumps were inspiraling toward the galaxy center . this migration to the galaxy center is expected from tidal forces acting upon massive gas clumps , and the large starbursts in our xmps may be going through the process at this moment . in some other cases , the velocity dispersion has a local maximum at the core of the star - forming region , which suggests a global expansion of the whole region , with a moderate speed of around 15 kms@xmath0 . sometimes the velocity dispersion presents a minimum at the core , which may reflect a past global expansion that washed out part of the turbulent motions . we find no obvious relationship between the kinematic properties and the metallicity drops , except for the already known correlation between the presence of a starburst and the decrease in metallicity . we do find , however , a significant lack of correlation between the metallicity and the n / o ratio ( fig . [ fig : oh_no ] ) , which is consistent with the gas accretion scenario . if the accretion of metal - poor gas is fueling the observed starbursts , then the fresh gas reduces the metallicity ( i.e. , o / h ) , but it can not modify the pre - existing ratio between metals ( e.g. , * ? ? ? * ; * ? ? ? * ) , forcing the n / o ratio to be similar inside and outside the starburst . the h@xmath1 line profile often shows a number of faint emission features , doppler - shifted with respect to the central component . their amplitudes are typically a few percent of the main component ( fig . [ fig_lobes ] ) , with the velocity shifts being between 100 and 400 kms@xmath0 ( table [ tab : summary_lobes ] ) . the components are often paired , so that red and blue peaks , with similar amplitudes and shifts , appear simultaneously . the red components tend to be slightly fainter , however . assuming that the emission is produced by recombination of h , we have estimated the gas mass in motion , which turns out to be in the range between 10 and @xmath49 ( table [ tab : summary_lobes ] and fig . [ fig : mass_loading ] ) . assuming that the doppler shifts are produced by an expanding shell - like structure , we infer a mass loss rate between @xmath32 and a few @xmath75 . given the observed sfr , the mass loss rate yields a mass loading factor ( defined as the mass loss rate divided by the sfr ) typically in excess of 10 . large mass loading factors are indeed expected from numerical simulations ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) , and they reflect the large feedback of the star formation process on the surrounding medium . numerical simulations predict that low mass galaxies are extremely inefficient in using their gas , most of which returns to the igm without being processed through the stellar mechanism . since the measured expansion velocities exceed by far the rotational and turbulent velocities , we argue in sect . [ sec : interpretation ] that the gas involved in the expansion is bound to escape from the galaxy disk , reaching the cgm and possibly the igm . therefore , the large mass loading factors that we measure support and constrain the predictions from numerical simulations . we have estimated the h column density to be expected when the shell material reaches the cgm , and it turns out to be around @xmath76 @xmath77 . even if all the h were neutral , this is undetectably small with present technology . moreover , at these low column densities , all h is expected to be ionized by the cosmic uv background . even if the igm of the xmps were filled by expanding shells from previous starbursts , they would be extremely elusive observationally . from an energetic point of view , the observed motions involve energies between @xmath78 and @xmath79 erg , which are in the range of the kinetic energy released by a core - collapse sn ( @xmath80 erg ) . the low - end energy requires that only part of the sn kinetic energy drives the observed expansion , whereas the high - end energy requires several sne going off quasi - simultaneously . the observed sfrs allow for the required number of sne , and the estimated age of the shells ( @xmath81@xmath45 yr ; table [ tab : summary_lobes ] ) is also consistent with the expected sn rate . other alternatives to explain the weak emission peaks in the wings of h@xmath1 , like bh driving the motions or the expansion , can be discarded using physical arguments ( sect . [ sec : interpretation ] ) . some of the data suggest a moderate global expansion of the star - forming regions ( sect . [ sec : mydiscussion ] ) . even if the expansion velocity is small ( around 15 kms@xmath0 ) , the motion involves the mass of the whole region , and , hence , the associated mass loss rate and kinetic energy turn out to be similar to those carried away by the fast - moving secondary components . thus , global expansion may also be important in setting up the mass and energy budgets associated with star formation feedback . high redshift clumpy galaxies are probably growing through a cosmological gas accretion process similar to the one experienced by xmps . however , the physical conditions at high and low redshifts differ ( e.g. , the gas accretion rates are larger at high redshifts ) , so do the observational biases affecting xmps and high - redshift clumpies ( e.g. , low - mass galaxies are undetectable at high redshift ) . therefore , addressing the issue of whether xmps are low - mass analogues of high - redshift clumpies is complicated , and so goes beyond the scope of our work . however , because of their probable common physics , it is interesting to point out some of the similarities and differences of the two samples . as in the case of xmps , some clumpy galaxies show metallicity drops associated with enhanced star - formation activity . high redshift galaxies are significantly more massive than xmps . using as reference for high redshift galaxies ( 1.32.6 ) the sins sample @xcite , the stellar mass of the high - redshift objects ( @xmath82@xmath83@xmath84 ) is three orders of magnitude larger than the mass of xmps ( @xmath45@xmath85@xmath84 ; table [ table_obs ] ) . this difference in stellar mass shows in dynamical mass too , with rotational velocities between 100 and 260kms@xmath0 at high redshift . we can not directly compare the rotational velocities of xmps with these values since we can not corrected for inclination , however , the uncorrected values go from zero to 30kms@xmath0 ( see fig . [ fig : rot_curve ] ) . the velocity dispersion in xmps ranges from 10 and 20kms@xmath0 ( corresponding to fwhm between 25 and 50kms@xmath0 ; see fig . [ fig : dispersion ] ) , whereas it is between 50 and 200kms@xmath0 at high - redshift @xcite . the high redshift objects have a ratio between rotational velocity and velocity dispersion between 0.5 and one @xcite . xmps seem to have similar ratios , even though we can not be more precise since the measured rotation in xmps is affected by inclination . the total sfrs are very different ( 10 500@xmath84yr@xmath0 at high redshift and 0.001 0.1@xmath84yr@xmath0 in xmps ) , but the surface star - formation rates are not ; they go from 0.01 to 10@xmath84yr@xmath0kpc@xmath86 at high redshift ( e.g. , * ? ? ? * ) and from 0.005 to 0.8@xmath84yr@xmath0kpc@xmath86 in xmps ( paper i ) . mass loading factors are one order of magnitude larger in xmps than those inferred for high - redshift clumpy galaxies ( sect . [ multiple_comp ] ; see also * ? ? ? * ) , a difference that we attribute to their very different masses . however , the differences of size between the expanding regions of clumpies and xmps may also play a role ( see eq . [ [ massloss ] ] ) . local uv bright clumply galaxies @xcite are often xmps with metallicity drops @xcite , and we know that their star - forming clumps resemble those in high - redshift cumplies in terms of total sfr , clump mass , and stellar mass surface density @xcite . thanks are due to r. amorn , y. ascasibar , a. b. morales - luis , p. papaderos , and j. vlchez , that participated in the target selection and in some of the observing time proposals that rendered the spectra analyzed in the paper . we also thank an anonymous referee whose comments help improving the readability of the paper . this article is based on observations made with the wht operated by the ing in the observatorio del roque de los muchachos , la palma , spain . this work has been partly funded by the spanish ministry of economy and competitiveness , projects _ estallidos _ aya201347742c0402p and aya201347742c0401p . aog thanks fundacin la caixa for financial support in the form of a phd contract . mef is partly funded by the _ estallidos _ project , as well as by the _ fundao para a cincia e tecnologia_ ( fct portugal ) , through the research grant sfrh / bpd/107801/2015 . jma acknowledges support from the european research council starting grant ( sedmorph ; p.i . v. wild ) . 107 natexlab#1#1 , s. , albareti , f. d. , allende prieto , c. , anders , f. , anderson , s. f. , anderton , t. , andrews , b. h. , armengaud , e. , aubourg , . , bailey , s. , & et al . 2015 , , 219 , 12 , r. , prez - montero , e. , contini , t. , vlchez , j. m. , bolzonella , m. , tasca , l. a. m. , lamareille , f. , zamorani , g. , maier , c. , carollo , c. m. , kneib , j .- p . , le fvre , o. , lilly , s. , mainieri , v. , renzini , a. , scodeggio , m. , bardelli , s. , bongiorno , a. , caputi , k. , cucciati , o. , de la torre , s. , de ravel , l. , franzetti , p. , garilli , b. , iovino , a. , kampczyk , p. , knobel , c. , kova , k. , le borgne , j .- f . , le brun , v. , mignoli , m. , pell , r. , peng , y. , presotto , v. , ricciardelli , e. , silverman , j. d. , tanaka , m. , tresse , l. , vergani , d. , & zucca , e. 2015 , , 578 , a105 , r. , prez - montero , e. , vlchez , j. m. , & papaderos , p. 2012 , , 749 , 185 , r. o. , prez - montero , e. , & vlchez , j. m. 2010 , , 715 , l128 , m. , grevesse , n. , sauval , a. j. , & scott , p. 2009 , , 47 , 481 , n. , emsellem , e. , kissler - patig , m. , & maraston , c. 2006 , , 445 , 471 , d. a. , skillman , e. d. , marble , a. r. , van zee , l. , engelbracht , c. w. , lee , j. c. , kennicutt , jr . , r. c. , calzetti , d. , dale , d. a. , & johnson , b. d. 2012 , , 754 , 98 , j. , & tremaine , s. 2008 , galactic dynamics : second edition ( princeton university press ) , y. , & dekel , a. 2003 , , 345 , 349 , s. , heyvaerts , j. , falgarone , e. , perault , m. , & puget , j. l. 1987 , , 172 , 293 , f. , daddi , e. , elmegreen , b. g. , elmegreen , d. m. , nesvadba , n. , vanzella , e. , di matteo , p. , le tiran , l. , lehnert , m. , & elbaz , d. 2008 , , 486 , 741 , f. , perret , v. , renaud , f. , dekel , a. , elmegreen , b. g. , elmegreen , d. m. , teyssier , r. , amram , p. , daddi , e. , duc , p .- a . , elbaz , d. , epinat , b. , gabor , j. m. , juneau , s. , kraljic , k. , & le floch , e. 2014 , , 780 , 57 , a. , zaragoza - cardiel , j. , beckman , j. e. , font , j. , garca - lorenzo , b. , erroz - ferrer , s. , & amram , p. 2015 , , 447 , 3840 , j. a. , clayton , g. c. , & mathis , j. s. 1989 , , 345 , 245 , s. , marconi , a. , maiolino , r. , balmaverde , b. , brusa , m. , cano - daz , m. , cicone , c. , comastri , a. , cresci , g. , fiore , f. , feruglio , c. , la franca , f. , mainieri , v. , mannucci , f. , nagao , t. , netzer , h. , piconcelli , e. , risaliti , g. , schneider , r. , & shemmer , o. 2015 , , 580 , a102 , d. , dekel , a. , & bournaud , f. 2010 , , 404 , 2151 , d. , snchez almeida , j. , muoz tun , c. , dekel , a. , elmegreen , b. g. , elmegreen , d. m. , & primack , j. 2016 , , 457 , 2605 , s. 1951 , proceedings of the royal society of london series a , 210 , 18 , r. , & terlevich , r. 1994 , line profiles in compact supernova remnants and active galactic nuclei . , ed . g. tenorio - tagle , g. , mannucci , f. , maiolino , r. , marconi , a. , gnerucci , a. , & magrini , l. 2010 , , 467 , 811 , j. j. , & stilp , a. m. 2010 , , 721 , 547 , r. , finlator , k. , & oppenheimer , b. d. 2012 , , 421 , 98 , m. a. , & mac low , m .- m . 2002 , , 581 , 1047 , w. j. g. , fraternali , f. , heald , g. h. , adams , e. a. k. , bosma , a. , koribalski , b. s. , & the hi science working group . 2015 , arxiv e - prints , a. , birnboim , y. , engel , g. , freundlich , j. , goerdt , t. , mumcuoglu , m. , neistein , e. , pichon , c. , teyssier , r. , & zinger , e. 2009 , , 457 , 451 , v. v. , & gruszko , j. 2012 , , 419 , 1515 , b. g. , bournaud , f. , & elmegreen , d. m. 2008 , , 688 , 67 , b. g. , elmegreen , d. m. , snchez almeida , j. , muoz - tun , c. , dewberry , j. , putko , j. , teich , y. , & popinchalk , m. 2013 , , 774 , 86 , b. g. , zhang , h .- x . , & hunter , d. a. 2012 , , 747 , 105 , d. m. , elmegreen , b. g. , snchez almeida , j. , muoz - tun , c. , putko , j. , & dewberry , j. 2012 , , 750 , 95 , d. m. , elmegreen , b. g. , sanchez almeida , j. , munoz - tunon , c. , mendez - abreu , j. , gallagher , j. s. , rafelski , m. , filho , m. , & ceverino , d. 2016 , arxiv e - prints , d. 1996 , interpreting astronomical spectra , 472 , e. e. 1964 , , 69 , 521 , a. , salvadori , s. , yue , b. , & schleicher , d. 2014 , , 443 , 2410 , k. m. , burkert , a. , ntormousi , e. , fierlinger , p. , schartmann , m. , ballone , a. , krause , m. g. h. , & diehl , r. 2016 , , 456 , 710 , m. e. , fraternali , f. , markoff , s. , nagar , n. m. , barthel , p. d. , ho , l. c. , & yuan , f. 2004 , , 418 , 429 , m. e. , winkel , b. , snchez almeida , j. , aguerri , j. a. , amorn , r. , ascasibar , y. , elmegreen , b. g. , elmegreen , d. m. , gomes , j. m. , humphrey , a. , lagos , p. , morales - luis , a. b. , muoz - tun , c. , papaderos , p. , & vlchez , j. m. 2013 , , 558 , a18 , d. b. , glazebrook , k. , damjanov , i. , abraham , r. g. , obreschkow , d. , wisnioski , e. , bassett , r. , green , a. , & mcgregor , p. 2016 , , n. m. , genzel , r. , bouch , n. , cresci , g. , davies , r. , buschkamp , p. , shapiro , k. , tacconi , l. j. , hicks , e. k. s. , genel , s. , shapley , a. e. , erb , d. k. , steidel , c. c. , lutz , d. , eisenhauer , f. , gillessen , s. , sternberg , a. , renzini , a. , cimatti , a. , daddi , e. , kurk , j. , lilly , s. , kong , x. , lehnert , m. d. , nesvadba , n. , verma , a. , mccracken , h. , arimoto , n. , mignoli , m. , & onodera , m. 2009 , , 706 , 1364 , j. , king , a. , & raine , d. j. 2002 , accretion power in astrophysics : third edition , 398 , c. d. , & obrien , t. j. 1999 , , 307 , 677 , o. , masegosa , j. , mrquez , i. , guerrero , m. a. , & dultzin - hacyan , d. 2006 , , 460 , 45 , r. , muoz - tun , c. , & mndez - abreu , j. 2016 , arxiv e - prints , j. l. , rix , h .- w . , & bernstein , g. m. 2001 , , 121 , 683 , g. , ecuvillon , a. , rebolo , r. , garca - lpez , r. , bonifacio , p. , & molaro , , , 421 , 649 , y. i. , guseva , n. g. , fricke , k. j. , & papaderos , p. 2009 , , 503 , 61 , y. i. , thuan , t. x. , & guseva , n. g. 2012 , , 546 , a122 , b. l. , koposov , s. , stark , d. p. , belokurov , v. , pettini , m. , & olszewski , e. w. 2015 , , 448 , 2687 , p. , schmitt , j. , & gorski , m. 2011 , , 741 , 10 , d. , & woosley , s. e. 2009 , , 703 , 2205 , c. , vlchez , j. m. , prez - montero , e. , iglesias - pramo , j. , brinchmann , j. , kunth , d. , durret , f. , & bayo , f. m. 2015 , , 801 , l28 , b. w. , wadsley , j. , benincasa , s. m. , & couchman , h. m. p. 2014 , , 442 , 3013 , r. c. , & evans , n. j. 2012 , , 50 , 531 , jr . , r. c. 1998 , , 36 , 189 , d. , katz , n. , weinberg , d. h. , & dav , r. 2005 , , 363 , 2 , j. , & ho , l. c. 2013 , , 51 , 511 , d. , & stlin , g. 2000 , , 10 , 1 , k. r. 1999 , astrophysical formulae , c. , schaerer , d. , goldader , j. d. , delgado , r. m. g. , robert , c. , kune , d. f. , de mello , d. f. , devost , d. , & heckman , t. m. 1999 , , 123 , 3 , e. m. , berger , e. , soderberg , a. m. , & chornock , r. 2011 , , 739 , 23 , r. , gallerani , s. , neri , r. , cicone , c. , ferrara , a. , genzel , r. , lutz , d. , sturm , e. , tacconi , l. j. , walter , f. , feruglio , c. , fiore , f. , & piconcelli , e. 2012 , , 425 , l66 , n. , dekel , a. , ceverino , d. , tweed , d. , moody , c. e. , & primack , j. 2014 , , 443 , 3675 , b. r. 1971 , statistics for physicists , c. l. 1999 , , 513 , 156 , c. l. , shapley , a. e. , coil , a. l. , kornei , k. a. , bundy , k. , weiner , b. j. , noeske , k. g. , & schiminovich , d. 2012 , , 760 , 127 , f. 2015 , in sf2a-2015 : proceedings of the annual meeting of the french society of astronomy and astrophysics , ed . f. martins , s. boissier , v. buat , l. cambrsy , & p. petit , 343348 , j. 1984 , , 211 , 521 , m. j. , zwaan , m. a. , webster , r. l. , staveley - smith , l. , ryan - weber , e. , drinkwater , m. j. , barnes , d. g. , howlett , m. , kilborn , v. a. , stevens , j. , waugh , m. , pierce , m. j. , bhathal , r. , de blok , w. j. g. , disney , m. j. , ekers , r. d. , freeman , k. c. , garcia , d. a. , gibson , b. k. , harnett , j. , henning , p. a. , jerjen , h. , kesteven , m. j. , knezek , p. m. , koribalski , b. s. , mader , s. , marquarding , m. , minchin , r. f. , obrien , j. , oosterloo , t. , price , r. m. , putman , m. e. , ryder , s. d. , sadler , e. m. , stewart , i. m. , stootman , f. , & wright , a. e. 2004 , , 350 , 1195 , a. b. , snchez almeida , j. , aguerri , j. a. l. , & muoz - tun , c. 2011 , , 743 , 77 , n. , & chang , p. 2015 , , 804 , 44 , s. f. , genzel , r. , frster - schreiber , n. m. , shapiro griffin , k. , mancini , c. , lilly , s. j. , renzini , a. , bouch , n. , burkert , a. , buschkamp , p. , carollo , c. m. , cresci , g. , davies , r. , eisenhauer , f. , genel , s. , hicks , e. k. s. , kurk , j. , lutz , d. , naab , t. , peng , y. , sternberg , a. , tacconi , l. j. , vergani , d. , wuyts , s. , & zamorani , g. 2012 , , 761 , 43 newville , m. , stensitzki , t. , allen , d. b. , & ingargiola , a. 2014 , lmfit : non - linear least - square minimization and curve - fitting for python , b. d. , dav , r. , katz , n. , kollmeier , j. a. , & weinberg , d. h. 2012 , , 420 , 829 , d. e. 1974 , astrophysics of gaseous nebulae , f. , ferrara , a. , grazian , a. , fiore , f. , giallongo , e. , & puccetti , s. 2016 , , 459 , 1432 , p. , guseva , n. g. , izotov , y. i. , & fricke , k. j. 2008 , , 491 , 113 , p. , izotov , y. i. , guseva , n. g. , thuan , t. x. , & fricke , k. j. 2006 , , 454 , 119 , m. s. , & shankar , f. 2011 , , 417 , 2962 , e. 2014 , , 441 , 2663 , e. , & contini , t. 2009 , , 398 , 949 , e. , contini , t. , lamareille , f. , maier , c. , carollo , c. m. , kneib , j .- p . , le fvre , o. , lilly , s. , mainieri , v. , renzini , a. , scodeggio , m. , zamorani , g. , bardelli , s. , bolzonella , m. , bongiorno , a. , caputi , k. , cucciati , o. , de la torre , s. , de ravel , l. , franzetti , p. , garilli , b. , iovino , a. , kampczyk , p. , knobel , c. , kova , k. , le borgne , j .- f . , le brun , v. , mignoli , m. , pell , r. , peng , y. , presotto , v. , ricciardelli , e. , silverman , j. d. , tanaka , m. , tasca , l. a. m. , tresse , l. , vergani , d. , & zucca , e. 2013 , , 549 , a25 , w. h. , flannery , b. p. , & teukolsky , s. a. 1986 , numerical recipes . the art of scientific computing , a. h. , tsantaki , m. , zezas , a. , jackson , f. , roberts , t. p. , foltz , r. , linden , t. , & kalogera , v. 2013 , , 769 , 92 , m. e. , peek , j. e. g. , & joung , m. r. 2012 , , 50 , 491 , j. i. , iorio , g. , agertz , o. , & fraternali , f. 2016 , , 462 , 3628 , b. , & goldreich , p. 2012 , , 750 , l31 , e. , anderson , j. , lennon , d. j. , van der marel , r. p. , aloisi , a. , boyer , m. l. , cignoni , m. , de marchi , g. , de mink , s. e. , evans , c. j. , gallagher , iii , j. s. , gordon , k. , gouliermis , d. a. , grebel , e. k. , koekemoer , a. m. , larsen , s. s. , panagia , n. , ryon , j. e. , smith , l. j. , tosi , m. , & zaritsky , d. 2013 , , 146 , 53 , p. , lapi , a. , tonini , c. , gentile , g. , yegorova , i. , & klein , u. 2007 , , 378 , 41 , j. , elmegreen , b. g. , muoz - tun , c. , & elmegreen , d. m. 2014 , , 22 , 71 , j. , elmegreen , b. g. , muoz - tun , c. , elmegreen , d. m. , prez - montero , e. , amorn , r. , filho , m. e. , ascasibar , y. , papaderos , p. , & vlchez , j. m. 2015 , , 810 , l15 ( paper i ) , j. , & et al . 2016b , on the dearth of ultra - faint xmps , in preparation , j. , morales - luis , a. b. , muoz - tun , c. , elmegreen , d. m. , elmegreen , b. g. , & mndez - abreu , j. 2014 , , 783 , 45 , j. , muoz - tun , c. , elmegreen , d. m. , elmegreen , b. g. , & mndez - abreu , j. 2013 , , 767 , 74 , j. , prez - montero , e. , morales - luis , a. b. , muoz - tun , c. , garca - benito , r. , nuza , s. e. , & kitaura , f. s. 2016 , , 819 , 110 , r. , fraternali , f. , oosterloo , t. , & van der hulst , t. 2008 , , 15 , 189 , f. 1979 , , 91 , 149 , e. d. , kennicutt , r. c. , & hodge , p. w. 1989 , , 347 , 875 , m. , cayrel , r. , plez , b. , hill , v. , spite , f. , depagne , e. , franois , p. , bonifacio , p. , barbuy , b. , beers , t. , andersen , j. , molaro , p. , nordstrm , b. , & primas , f. 2005 , , 430 , 655 , g. , munoz - tunon , c. , & cid - fernandes , r. 1996 , , 456 , 264 , t. a. , & krumholz , m. r. 2016 , , 455 , 334 , p. , maiolino , r. , sommariva , v. , cresci , g. , mannucci , f. , marconi , a. , meneghetti , m. , grazian , a. , cimatti , a. , fontana , a. , nagao , t. , & pentericci , l. 2014 , , 563 , a58 , p. g. 2001 , , 113 , 1420 , s. , volonteri , m. , walker , m. g. , & gair , j. r. 2010 , , 408 , 1139 , s. , cecil , g. , & bland - hawthorn , j. 2005 , , 43 , 769 , m. 2010 , , 18 , 279 , c .- c . , & krumholz , m. 2012 , , 758 , 48 , b. 1987 , in spectroscopy of astrophysical plasmas , ed . a. dalgarno & d. layzer , 185209 the emission of a uniform shell produces a top - hat line profile that is not consistent with the two - horned profiles observed in the wings of h@xmath1 ( fig . [ fig_lobes ] ) . however , if the shell has some internal extinction , then it preferentially absorbs photons emerging from the regions where the line - of - sight ( los ) tends to be parallel to the surface of the shell . those are the regions of small doppler shifts that contribute to the line center , thus creating a two - horned profile . and width @xmath87 . ( a ) projection of the shell in the y - z plane , with z in the direction along the los . the velocity vector , @xmath88 , in the direction set by @xmath89 is shown as a blue solid - line arrow . ( b ) zoom - in of the region enclosed by the dotted - line rectangle in ( a ) . the length along the los traversed by the photons emitted at coordinates ( @xmath89 , @xmath53 ) is denoted as @xmath90 , with @xmath91 . , title="fig : " ] and width @xmath87 . ( a ) projection of the shell in the y - z plane , with z in the direction along the los . the velocity vector , @xmath88 , in the direction set by @xmath89 is shown as a blue solid - line arrow . ( b ) zoom - in of the region enclosed by the dotted - line rectangle in ( a ) . the length along the los traversed by the photons emitted at coordinates ( @xmath89 , @xmath53 ) is denoted as @xmath90 , with @xmath91 . , title="fig : " ] the effect can be readily shown assuming an expanding shell of radius @xmath56 and width @xmath87 , as shown in fig . [ cartoon_shell]a . the velocity along the los , @xmath92 , only depends on the inclination angle , @xmath89 , @xmath93 therefore , the line profile , @xmath94 , is just the total energy emitted with velocity between @xmath92 and @xmath95 , divided by @xmath96 . because of eq . ( [ app : eq1 ] ) , it is proportional to the energy emitted by the ring of the shell with inclination angles between @xmath89 and @xmath97 , i.e. , @xmath98 where @xmath56 is the radius of the shell of width @xmath87 , @xmath99 is the azimuth in spherical coordinates , and @xmath100 stands for the emission per unit volume . assuming the shell to be small enough ( @xmath101 ) , then @xmath102 according to the definition of @xmath94 , @xmath103 so eqs . ( [ app : eq1 ] ) , ( [ app : eq3 ] ) and ( [ app : eq4 ] ) yield , @xmath104 which holds only for @xmath105 . outside this range of velocities , @xmath106 is equal to zero , and so is @xmath94 . in the case that @xmath100 is constant , then @xmath94 is also constant ( eq . [ [ app : eq5 ] ] ) , leading to the top - hat line profiles expected from an optically thin expanding shell . we will consider the case where the emission is constant , @xmath107 , but the shell absorbs part of the emitted photons due to internal extinction , @xmath5 , i.e. , @xmath108 with @xmath109 as shown in fig . [ cartoon_shell]b . for the moment , we have assumed that the emission occurs in the upper cap only . using eqs . ( [ app : eq5 ] ) , ( [ app : extintion ] ) , and ( [ app : eq6 ] ) , @xmath110 , \label{app : eq7}\ ] ] @xmath111 with @xmath112 the emitted intensity if there were no extintion ( @xmath113 ) , @xmath114 the case of the lower cap is similar , except that one has to consider the additional absorption produced by the upper cap . in this other case , @xmath115\,\exp(-\beta\delta r/|\cos\theta| ) , \label{app : eq8}\ ] ] with @xmath116 the typical values of the extinction coefficient at h@xmath5 in xmps are between @xmath117 and @xmath118 @xcite . these values were inferred from the balmer decrement in sloan digital sky survey ( sdss ) spectra of xmps . if we use them to represent the total extintion in the shell , then @xmath119 , where we have considered the differential extintion between h@xmath1 and h@xmath5 using the extinction law in the milky way by @xcite . the dashed lines in fig . [ app : example ] show examples of the line profiles resulting from eqs . ( [ app : eq7 ] ) and ( [ app : eq8 ] ) . they represent extinctions that cover the range of values mentioned above . we show the original profiles ( dashed lines ) plus their convolution with a gaussian that represents the finite resolution of the observation ( solid lines ) . the width of the gaussian has been chosen so that , if it represents our wht observations ( fwhm @xmath120 km s@xmath0 ) , then the expansion velocities of the shells are similar to the speeds of the multiple components analyzed in sect . [ multiple_comp ] . one can see the two - horned shape , with a redshifted lobe that is smaller than the blushifted lobe ; the redshifted photons come from the lower cap of the shell , so that they are also extincted by the upper cap ( we have flipped the sign of the abscissa axis in fig . [ app : example ] , so that positive corresponds to redshift , as usual ) . ; profiles in red ) , and significant internal extinction ( @xmath121 , profiles in black ) . red and black lines also differ in the expansion speed @xmath122 , as shown in the inset . we include the profiles predicted by eqs . ( [ app : eq7 ] ) and ( [ app : eq8 ] ) ( dashed lines ) , together with these profiles convolved with a gaussian to represent the finite spectral resolution of the spectrograph ( solid lines ) . we assume the fwhm of the gaussian to be 0.35 in the dimensionless velocity scale . if this value is ascribed to the resolution of our observation ( @xmath120 km s@xmath0 ) , then the expansion speeds for the shells giving rise to the red and black profiles are 110 km s@xmath0 and 220 km s@xmath0 , respectively . abscissae show @xmath123 rather than @xmath92 so that the positive values correspond to redshifts . ] the model in eqs . ( [ app : eq7 ] ) and ( [ app : eq8 ] ) also predicts the fraction of emitted light lost by internal absorption . for the two profiles shown in fig . [ app : example ] , the fraction of light that emerges is 74% and 24% of the emitted light for the red and black profiles , respectively .

the extremely metal - poor ( xmp ) galaxies analyzed in a previous paper have large star - forming regions with a metallicity lower than the rest of the galaxy . such a chemical inhomogeneity reveals the external origin of the metal - poor gas fueling star formation , possibly indicating accretion from the cosmic web . this paper studies the kinematic properties of the ionized gas in these galaxies . most xmps have rotation velocity around a few tens of kms@xmath0 . the star - forming regions appear to move coherently . the velocity is constant within each region , and the velocity dispersion sometimes increases within the star - forming clump towards the galaxy midpoint , suggesting inspiral motion toward the galaxy center . other regions present a local maximum in velocity dispersion at their center , suggesting a moderate global expansion . the h@xmath1 line wings show a number of faint emission features with amplitudes around a few percent of the main h@xmath1 component , and wavelength shifts between 100 and 400 kms@xmath0 . the components are often paired , so that red and blue emission features with similar amplitudes and shifts appear simultaneously . assuming the faint emission to be produced by expanding shell - like structures , the inferred mass loading factor ( mass loss rate divided by star formation rate ) exceeds 10 . since the expansion velocity exceeds by far the rotational and turbulent velocities , the gas may eventually escape from the galaxy disk . the observed motions involve energies consistent with the kinetic energy released by individual core - collapse supernovae . alternative explanations for the faint emission have been considered and discarded .

introduction observations and data reduction the fate of the swept out material conclusions emission of an expanding dusty shell of gas

tidal disruption of stars in the gravitational potential of massive black holes in the centers of galaxies has been recognized long ago as a way to prove the existence of massive black holes in otherwise non - active galaxies . during the tidal disruption about 50% of the mass of the star are stripped off and get unbound from the star ( mostly independent of impact parameter , masses and velocities ; ayal 2000 ) . this material can eventually be accreted by the black hole , leading to a flare of optical to x - ray emission lasting a few months to years . while originally proposed to explain the activity in agn ( hills 1975 ) , a first calculation of the frequency , temporal behaviour and spectral signatures of tidal disruption events ( frank & rees 1976 , young 1977 , kato & hoshi 1978 , lidskii & ozernoi 1979 , gurzadyan & ozernoi 1980 , gurzadyan & ozernoi 1981 ) revealed both , the potential of such events to learn more about the inner regions of galaxies as well as the problems in observing these rare events ( rees 1988 ) . over the recent few years , tidal disruption has attracted renewed attention , both theoretically and observationally . loeb & ulmer ( 1997 ) have studied the case that a surrounding gaseous envelope could reprocess most of the radiation from the accretion process into a thermal spectrum of 10@xmath0 k , thus being easily detectable in optical surveys rather than in x - ray surveys . magorrian & tremaine ( 1999 ) have recalculated the stellar disruption rates in detailed dynamical models of real galaxies , taking into account the refilling of the loss cone of stars on disruptable orbits by two - body relaxation and tidal forces in non - spherical galaxies . ayal ( 2000 ) have performed simulations of the long - term evolution of tidally disrupted stars and find that only about 10% of the stellar mass are actually accreted by the central black hole . on the observational side , the discovery of time - variable uv or x - ray emission in distant , otherwise non - active galaxies may provide the first indirect evidence of tidal disruption events . besides the early report on a uv flare in ngc 4552 ( renzini 1995 ) , a few very good candidates have been detected with rosat during the recent years , such as the @xmath1 erg / s outburst in ngc 5905 ( bade 1996 , komossa & bade 1999 ) , the @xmath2 erg / s flare in rxj1242.61119 ( komossa & greiner 1999 ) , the @xmath3 erg / s flare in rxj1624.9 + 7554 ( grupe et al . 1999 ) , or the @xmath3 erg / s flare in rxj1331.93243 ( reiprich & greiner 2000 ) . interestingly , in all these rosat discoveries the x - ray spectrum of the flares is extremely soft , of the order of 50 - 100 ev effective temperature . the time scales of these events , though not in all cases well constrained , are of the order of months to years . all these properties , together with the fact that none of the corresponding galaxies shows any optical sign of activity have led to the tentative identification as being associated with disruption events . here we report on the discovery of another transient x - ray source and the optical and x - ray follow - up observations . the search for supersoft x - ray sources in the rosat pspc all - sky - survey ( greiner 1996 ) revealed a few sources for which neither confirmation through a rosat follow - up observation nor a reasonable optical counterpart could be found . one of these sources is @xmath4 1rxs j142024.4 + 533403 @xmath4 rbs 1376 ( schwope 2000 ) . it was detected during the pspc all - sky - survey between december 58 , 1990 at a count rate of 0.28 cts / s , resulting in the collection of a total of 255 counts . there is no significant short - term variability within the 3 day scanning coverage . the source position has been determined using only photons above 0.25 kev ( to avoid position deterioration by ghost images ) to be : ra ( 2000.0 ) = 1420242 , decl . ( 2000.0 ) = + 533411 with an error of @xmath520 . the spectrum is extremely soft as indicated by the hardness ratio : @xmath6 which is defined as ( h s)/(h+s ) , with h ( s ) being the counts above ( below ) 0.4 kev over the full pspc range of 0.12.4 kev . source photons were extracted with a radius of 4 . the background was chosen at the same ecliptic longitude at @xmath71 distance , corresponding to background photons collected typically 15 sec before or after the time of the source photons . standard corrections were applied using the dedicated exsas software package ( zimmermann 1994 ) . the spectrum fitted with a blackbody model is shown in fig . [ xspec ] , and tab . [ xpar ] summarizes the fit parameters , also including fits with a power law and an accretion disk model . follow - up pointed rosat observations were performed on july 10/11 , 1995 and again on june 3 , 1997 . in both cases the hri was used . in the 1995 observation we detect three x - ray photons at the position of , corresponding to a likelihood of 6.2 ( 2.5 @xmath8 ) . because of this marginal significance , we do not claim a detection , but rather use these detected photons as an upper limit ( fig . [ xlc ] ) on the count rate : 5.5@xmath910@xmath10 hri cts / s . the mean position of these photons , ra ( 2000.0 ) = 1420236 , decl . ( 2000.0 ) = + 533416 with an error of @xmath510is shown in fig . [ fc ] for comparison purposes . the 1997 observation , with less than half of the 1995 exposure , results in an upper limit of 1.2@xmath910@xmath11 hri cts / s . is serendipituously in the field of view of a rosat pspc pointed observation performed during the calibration phase , though at a very large offaxis angle of 49 . no source was detected ( using standard exsas commands ) , giving a 3@xmath8 upper limit of 1.8@xmath910@xmath11 pspc cts / s ( fig . [ xlc ] ) . investigation of the x - ray source position on the poss revealed only one bright optical object ( labelled `` a '' in fig . [ fc ] ) , at ra ( 2000.0 ) = 1420244 , decl . ( 2000 ) = 53 34 12 . we obtained a low - resolution spectrum of this r @xmath7 18.8 mag source using the 3.5 m telescope at calar alto on june 13 , 1993 . we used the cassegrain spectrograph with a 1024@xmath9640 rca chip ( pixel size 15 @xmath12 m ) . a grating with 240 / mm was used , allowing the range 40008500 to be observed with a fwhm resolution of 10 . the spectrum was bias and flat - field corrected and calibrated in wavelength using standard midas reduction packages . bd + 254655 was used for flux calibration . the optical spectrum is characterized by strong absorption lines of nai 5175 , mgi 5890 and h@xmath13 ( though it is near the atmospheric a band ) , typical of an elliptical or early spiral type . using the mgi , nai and h@xmath13 line we derive a redshift of @xmath14 . the balmer absorption lines and the strong drop of the flux beyond the caii h / k break argue against a classification as bl lac object . no agn - like forbidden emission lines , like [ oiii]@xmath155007 , are detected . in order to search for optical variability , we embarked on a more extensive photometry programme . observations were done at three locations , the sonneberg observatory ( germany ) , the special astrophysical observatory ( russia ) and at the wiyn telescope at kitt peak ( usa ) . [ log ] summarizes the logistic details . throughout these observations we did not found any significant optical variability . the imaging revealed a second , much fainter ( @xmath16 mag ) source within the x - ray error circle ( labelled `` b '' in fig . [ fc ] ) . during some of the wiyn observations , at very good seeing conditions of 09 , this source is readily seen to be extended ( see inset of fig . moreover , the intensity distribution over the galaxy disk is very flat . since active galaxies typically have a very peaked intensity profile ( dominated by a bright central source ) , this flat distribution argues for a non - active galaxy . the shape and radial profile suggest an elliptical or early spiral subtype . size and luminosity arguments then suggest that its distance is about a factor 35 larger than that of source `` a '' . a lower limit of z@xmath170.07 can be placed by assuming @xmath18 mag , corresponding to the faint end of the galaxy luminosity distribution . certainly , a spectrum of this galaxy would be very helpful to verify these suggestions . lllcr teles- & date & filter/ & d@xmath19 & t@xmath20 + cope@xmath21 & & wavelength & ( hrs ) & ( sec ) + + r 150046p & 1990 jul 1922 & 0.12.4 kev & 77.7 & 11475 + r survey & 1990 dec 58 & 0.12.4 kev & 77.8 & 910 + r 201999p & 1995 jul 10/11 & 0.12.4 kev & 30.7 & 5560 + r 300577h & 1997 jun 3 & 0.12.4 kev & 0.6 & 2150 + + ca 3.5 m & 1993 jun 13 & 40008500 & & 3600 + so 0.6 m & 1994 mar 30 & white & 2.5 & 30 + so 0.6 m & 1994 apr 14 & white & 2.9 & 60 + so 0.6 m & 1994 apr 19 & white & 3.2 & 30 + so 0.6 m & 1994 may 27 & white & 3.0 & 60 + @xmath22wiyn 3.5 m & 1996 may 25 & r / b & 1.0 & 270 + @xmath22wiyn 3.5 m & 1996 may 26 & b & 0.3 & 500 + @xmath22wiyn 3.5 m & 1996 may 27 & b & 0.1 & 500 + sao 1.0 m & 1996 jul 14 & v & 1.5 & 300 + @xmath21 the abbreviations mean : r = rosat , so = sonneberg observatory ( germany ) , ca = calar alto ( spain ) , sao = special astrophysical observatory ( russia ) . + @xmath19 duration of the coverage . [ log ] [ ospec ] llccc model@xmath21 & fit parameters@xmath19 & @xmath23 & flux@xmath24 ( 0.1 - 2.4 kev ) & flux@xmath24 ( bol . ) + bbdy , free fit & @xmath25 = 28@xmath510 @xmath26 = 2.0@xmath50.1 & 0.96 & 7.9@xmath910@xmath27 & 1.6@xmath910@xmath28 + bbdy , @xmath26 = @xmath29 & @xmath25 = 38@xmath510 & 0.90 & 8.8@xmath910@xmath30 & 4.1@xmath910@xmath27 + powl , free fit & @xmath311.5 @xmath26 = 2.8@xmath51.0 & 0.96 & 2.4@xmath910@xmath27 & + powl , @xmath26 = @xmath29 & @xmath320.5 & 1.07 & 8.9@xmath910@xmath30 & + dibb , free fit & = 0.01@xmath50.005 @xmath33 = 10@xmath34 ( fix ) @xmath26 = 6.6@xmath51.5 & 0.94 & 5.0@xmath910@xmath35 & 1.8@xmath910@xmath36 + dibb , @xmath26 = @xmath29 & = 2.6@xmath50.1 @xmath33 = 10@xmath34 ( fix ) & 0.90 & 9.2@xmath910@xmath30 & 6.2@xmath910@xmath27 + dibb , @xmath26 = @xmath29 & = 1.8@xmath50.1 @xmath33 = 7@xmath910@xmath37 norm@xmath38=2.9@xmath910@xmath35 ( fix ) & 0.90 & 9.2@xmath910@xmath30 & 6.4@xmath910@xmath27 + @xmath21 the galactic column density is @xmath29 = 1.2@xmath39 @xmath40 ( dickey & lockman 1990 ) . the model abbreviations have the following meaning : bbdy = black body , powl = power law , dibb = disk black body model . + @xmath19 units are ev for temperatures , for masses , @xmath41 @xmath40 for @xmath26 . + @xmath24 observed fluxes in units of erg @xmath40 s@xmath42 ( absorption - corrected ) . + @xmath38 normalization is chosen such that distance is equal to @xmath43 = 588 mpc ( and inclination assumed to be face - on ) . we have presented evidence for a transient x - ray source with an amplitude of a factor of @xmath44150 within 6 months . the only optical counterparts down to 23rd magnitude are two galaxies , the brighter of which is at a redshift of z=0.147 , and the fainter one probably a factor 35 more distant . if the x - ray flare occurred in the brighter galaxy , the unabsorbed , bolometric x - ray luminosity is 2.5@xmath910@xmath45 erg / s ( using the disk blackbody model with only galactic foreground absorption and a normalization which fits the z=0.147 distance ; see last line in tab . [ xpar ] ) . it should be mentioned that this is a lower limit since most probably we did not catch the peak of the flare , and the x - ray emission may have suffered additional intrinsic absorption . if the flare occurred in the fainter galaxy ( b ) then the luminosity would probably be even larger , but certainly not lower than 6@xmath910@xmath46 erg / s ( if at z@xmath170.07 ) . though it is impossible to claim with certainty that this evidence proves a tidal disruption origin , we will in the following , by the principle of exclusion , discuss possible alternatives ( for a more extensive discussion of alternatives see komossa & bade 1999 ) . ( 1 ) an active galaxy as the most common origin of large x - ray variability is very unlikely given our optical observations ( see above section 2.3 ) . ( 2 ) x - ray afterglows of gamma - ray bursts ( grb ) may appear as new , bright x - ray sources on the sky . however , all of the observed afterglows have faded rapidly ( with time@xmath47 ; see greiner 2000 ) . the rosat all - sky survey data suggests a rather flat light curve over about 3 days , inconsistent with a grb afterglow . ( 3 ) this flat , 3-day long light curve also excludes neutron star merger events which are predicted to reach peak luminosities of 10@xmath45 erg / s , but should fade by a factor of @xmath173 within @xmath172 days ( li & paczynski 1998 ) . thus , a tidal disruption scenario seems a logical consequence . moreover , it is even plausible : spectral fitting of a simple disk blackbody model ( shakura & sunyaev 1973 ) gives a consistent picture concerning x - ray temperature , mass of the central object ( @xmath1710@xmath34 ) , luminosity and distance ( see last row of tab . [ xpar ] ) . we are grateful to e. pavlenko ( crimean observatory ) and w. wenzel ( sonneberg ) for initial optical observations in 1994 , as well as to n. borisov for help in observations conducted at sao . we appreciate enlightening discussions with m. salvato ( aip ) on the radial profiles of galaxies . jg was partly supported by the german bundesministerium fr bildung , wissenschaft , forschung und technologie ( bmbf / dlr ) under contract 50qq96023 . the project was supported by bmbf / dlr and the max - planck - society .

we have discovered a transient x - ray source , , which displays a rosat flux variation of 150 between the all - sky - survey in 1990 and a preceding pointed rosat observations in july 1990 . optical observations suggest a non - active galaxy as the only visible counterpart . we therefore tentatively identify as a tidal disruption event in a non - active galaxy .

introduction observational results discussion

since their discovery in the mid-1980 s , collimated jets from young stars have played an increasingly important role in clarifying how stars form . jets provide a means for protoplanetary disks to shed angular momentum as disk material accretes onto the star , and recent observations of rotation in jets provide a direct observational link to the accretion disk that ultimately powers these flows @xcite . most stars are members of binaries , and jets provide crucial information about disks in such systems through measurements of precession @xcite , and by tracing separate flows from both components of close binaries @xcite . jets evacuate large cavities that pierce the surrounding molecular cloud and power large - scale bipolar molecular outflows which help to support molecular clouds against gravitational collapse @xcite . because outflows are driven by accretion , knots in jets give a fossil record of the accretion history of young stars . collimated outflows also accompany accretion disks in compact galactic sources @xcite , and in extragalactic nuclei @xcite ; jets from young stars are a particularly good place to study the connection between accretion and outflow because shock waves in the flow cool radiatively , so images and spectra reveal where the flows are heated and trace how they propagate . the excellent spatial resolution provided by the hubble space telescope has greatly advanced our understanding of jet dynamics . cooling zones behind shock waves in stellar jets are typically near the limit of what can be resolved from the ground , but these zones are clearly resolvable with _ hst_. shock waves in nearby jets move @xmath4 1 arcsecond per decade , but also change their spatial structure on that timescale . hence , the subarcsecond spatial resolution provided by _ hst _ makes it possible to detect real differential motions between jet knots , and thereby observe the complex dynamics of working surfaces , merging bow shocks , and fluid instabilities in real time . it is possible to observe the true density structure and calculate mass loss rates directly when jets are irradiated by ultraviolet light @xcite . the three bright stellar jets ( hh 1 , hh 34 , and hh 111 ) we have observed to date in this manner each reveal something new and unexpected about the dynamics of jets from young stars . a portion of the large bow shock hh 34 @xcite appears to be in the process of forming a rayleigh - taylor instability , and we observed new shock waves form as dense knots emerged from the source . the proper motions in hh 111 @xcite clearly show the aftermath of merging bow shocks , and faint emission wings along the shocks in the flow demonstrate that the surrounding material plays no role in collimating the jet once the jet emerges from the source . because hh 111 is in the plane of the sky , the proper motions are the true space velocities , so here we can observe the actual velocity variations in the flow and thereby test models of radiative shocks in jets . a remarkable area of strong shear exists along the edges of the bow shock hh 1 , while the flow is much more complex and less ordered on the opposite side of the flow in hh 2 @xcite . discovered by @xcite , the famous hh 46/47 bipolar outflow is in many ways an ideal object to study jet dynamics . the jet is bright , and shows a classic structure of a collimated flow with several large bow shocks . the brightest bow shock , hh 47a , is one of the best examples of a working surface , and h@xmath0 images of the system taken with _ hst _ * hereafter h96 ) reveal a distinct bow shock that accelerates downstream material and a mach disk that decelerates incident jet gas . the star formation in the region is relatively simple the jet emanates from a resolved ( @xmath5 ) low mass binary system surrounded by a circumstellar disk @xcite that has formed in the isolated bok globule eso 210 - 6a at the edge of the gum nebula . this binary is the only known young stellar system in the globule . the globule is clearly compressed to the north and west toward the source of ionizing radiation ( figure 1 of h96 ) . the nascent stellar system and circumstellar disk are thus transitioning from being embedded in the dense globule to being exposed to the uv radiation from the nearby massive stars @xmath6 pup and @xmath7 vel through photoevaporation of the surrounding envelope @xcite . the hh 46/47 outflow has been studied extensively across the electromagnetic spectrum at uv , optical , and infrared wavelengths ( h96 ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? the optically bright hh 46/47 jet is part of the northeast flow , and terminates in the hh 47a bow shock @xmath8 from the ir driving source . hh 47a moves outward in the wake of the larger , fainter bow shock hh 47d . a faint counterjet in the redshifted southwest lobe @xcite extends toward the bow shock hh 47c @xcite . assuming a distance of 450 pc @xcite , the visible outflow from hh 47d to hh 47c extends @xmath40.57 pc . the outflow velocities of @xmath4300 km@xmath2s@xmath3 derived from proper - motion and radial - velocity studies @xcite imply a kinematical age for the main optical / near - ir outflow of @xmath9 years , with large mass ejections occuring every 400 years or so @xcite , perhaps driven by fu orionis - type eruptions @xcite . however , as occurs with other hh flows ( e.g. hh 34 * ? ? ? * ) , deep wide field images reveal that hh 46/47 extends at least 2.6 pc from the irs , and therefore must be @xmath10 years old @xcite . in this paper we continue our investigation of stellar jet dynamics by presenting _ hst _ proper motions of the optically emitting shocks in the hh 46/47 jet . the new observations were taken about 5 years after those in h96 , long enough for knots in the jet to move significantly , but short enough that morphological changes remain small for most of the knots . in addition to being able to quantify precession and differential motions in the jet with high precision , the new images make it possible to study the dynamics of a working surface in exquisite detail for the first time . we imaged the hh 47 jet through the f656n and the f673n filters with the wfpc-2 camera on _ hst _ on 26 february 1999 , a time interval of 4.9 years after the first epoch of images described by h96 . the images include the entire bright jet on one wf chip , with the extended bow shock hh 47d spread across the pc ( fig . 1 ) . three exposures each for h@xmath0 and [ s ii ] facilitated cosmic - ray removal . total exposure times for the h@xmath0 and [ s ii ] images in the 1999 epoch were 4100 seconds and 4000 seconds , respectively . we reduced the new images and aligned them with the previous epoch using the same procedures described in h96 . a difference image between the two epochs ( fig . 2 ) shows clearly that the jet and bow shocks move away from hh 47-irs . to measure proper motions we used a method described by @xcite , explained in more detail in @xcite the algorithm shifts two images to minimize the square of the difference of the counts over a region that defines the object . we examined each image for features that retained their overall shape between the two epochs and applied the procedure separately for h@xmath0 and for [ s ii ] , because hh objects often have different morphologies when imaged in these two filters . figs . 3 and 4 show objects in the jet , while figs . 5 and 6 cover the hh 47a bow shock . variable features within hh 47a are emphasized in fig table 1 compiles proper motions of all the objects identified in the figures . coordinates in the table were obtained using the stsdas task xy2rd . errors in the proper motions come from imperfect alignment of the images between epochs , uncertainties in the centroiding routine that determines the spatial shifts , and errors in the distance to hh 47 . in the hh 47 images there are many field stars available to align the epochs , so uncertainties in the registration are small . using the proper motion code on stars after the images were aligned shows the rms error in the image registration is about 0.1 pixel . formal statistical errors in centroiding features are typically 0.05 pixel . however , an additional systematic error of @xmath4 0.1 pixel arises from using different buffer sizes to search for the centroid ( e.g. h96 ) . the distance to hh 47 derives from estimates of the distance to ngc 2547 ( 450 pc * ? ? ? * ) and to a group of coeval stars within about 5 degrees of the vela pulsar ( 425 pc * ? ? ? the projected distance of hh 47 from ngc 2547 is about 3 degrees , or 25 pc at a distance of 450 pc . we adopt 25 pc as the uncertainty in the distance , which introduces a 6% error in the observed tangential velocities . altogether , internal errors in the proper motions over the 4.9 year interval between epochs are @xmath4 5 km@xmath2s@xmath3 , for all objects except the large , diffuse bow shock hh 47d , where the uncertainty is 40 km@xmath2s@xmath3 . the 6% uncertainty in the distance is a multiplicative constant that affects all the measurements in the same way . proper motions in hh 47d derive from the sum of the h@xmath0 and [ s ii ] images to improve the signal - to - noise . when combined with existing radial velocity information , our new proper motion measurements make it possible to calculate the orientation of the flow , its true space velocity , and to assess the degree to which clumpy flows and variations in the intrinsic velocity and in the ejection angles affect the jet dynamics . initial results from the analysis of these data were reported by @xcite . radial velocities along the jet observed by meaburn & dyson ( 1987 , their fig . 2 ) increase from hh46 irs out to the hh 47a bow shock , and we observe the same behavior in the proper motions shown in fig . this correlation between radial velocities and proper motions implies that the jet varies intrinsically in velocity . the alternative explanation , that changes in the ejection angles cause the radial velocities and proper motions to change , is ruled out because in this model proper motions increase as radial velocities decrease and vice - versa for a constant velocity flow . the relatively poor spatial resolution of the ground - based radial velocity measurements makes it difficult to compare these directly with proper motions of specific emission features observed by _ hst_. however , it is easy to identify the maximum radial velocity of the jet with respect to the cloud as being 200 @xmath1 10 km@xmath2s@xmath3 before the jet encounters hh 47a @xcite . we measure the proper motion of knot ah1 to be 300 @xmath1 18 km@xmath2s@xmath3 ( where the error is dominated by uncertainties in the distance ) , so this feature moves at an angle of 34 @xmath1 2 degrees from the plane of the sky , with a velocity of 360 @xmath1 28 km@xmath2s@xmath3 . because the jet is not straight , a long slit may miss some of the line emission in specific features and thereby give misleading estimates for the orientation angles . to obtain radial velocities for the entire emitting region one can use a slit map or observe with a fabry - perot , as was done by @xcite . using the datacube from that paper , the average radial velocity of the jet with respect to the cloud is 180 @xmath1 10 km@xmath2s@xmath3 . the average proper motion of the jet in fig . 8 is 235 km@xmath2s@xmath3 , with an intrinsic scatter of 10 km@xmath2s@xmath3 and a systematic error of 14 km@xmath2s@xmath3 from the distance . hence , the average orientation angle of the jet with respect to the plane of the sky is 37.5 @xmath1 2.5 degrees , in good agreement with the above measurement of ah1 and with previous estimates from ground - based proper motions @xcite . as the last column of table 1 shows , the outer bow shock hh 47d is about 1300 years old , and hh 47a was ejected a little less than 800 years ago , assuming no significant deceleration of these objects has occurred . the average proper motion is @xmath4 235 km@xmath2s@xmath3 in the jet , @xmath4 200 km@xmath2s@xmath3 in hh 47a and @xmath4 180 km@xmath2s@xmath3 in hh 47d , so if this decline is caused by deceleration rather than source variability , then the amount of deceleration over the length of the jet is rather small compared with the velocity of the flow . the youngest knots in the jet are only a few decades old , and these knots have significantly higher proper motions than those ejected 150 @xmath11 200 years ago . the hh 47 flow is an excellent place to observe how a jet with a variable ejection angle clears out a larger cavity . the inner 20 arcseconds of the hh 47 jet curves to the left in figs . 1 and 2 , but the proper motions are not along the jet . instead , the knots in the jet move away from the source radially , as expected for a jet with a variable ejection angle the current direction of the jet , defined by the motion of knots jh1 , jh2 , and js2 , is more northerly than its average over the past millennium . the reflection nebula along the cavity is also brighter on the northern side , as if the space between the irs and the northern cavity wall has been mostly cleared of dust , and fig . 2 shows that this wall brightened between 1994 and 1999 . there is direct evidence that the jet currently collides with stationary material on this side of the cavity . a morphologically variable , but stationary , linear h@xmath0 feature noted by h96 has the spatial position and orientation expected if the jet were impacting the northern edges of the cavity obliquely . it is now possible to study deflection shocks in laboratory experiments that propagate jets into crosswinds @xcite . the hh 46/47 deflection shock is present only in h@xmath0 , and is marked in fig . shocks in stellar jets are caused by faster material overtaking slower material ( e.g. * ? ? ? typical velocity variations along the hh 47 jet between adjacent knots are @xmath4 20 km@xmath2s@xmath3 , with a few cases up to 40 km@xmath2s@xmath3 , in agreement with shock models of the emission line ratios @xcite . two places in the flow will experience higher shock velocities in the future proper motions in knot ah1 exceed those of hh 47a by about 100 km@xmath2s@xmath3 , so the true impact velocity should be @xmath4 125 km@xmath2s@xmath3 when this knot enters the working surface in about the year 2040 . this impact velocity should be large enough to generate [ o iii ] line emission . the other area where proper motions change markedly is close to the source , where knots jh1 and jh2 have proper motions @xmath4 100 km@xmath2s@xmath3 higher than jh3 and other knots ahead in the flow . the first dynamical consequence of these large proper motions should be in about 30 years , when these knots encounter the current position of the deflection shock ( fig . hh 47a appears as a bow shock with a well - defined mach disk ( ah4 in fig . 5 ) , and models of hh 47a have attempted to predict the emission line ratios and line widths using this scenario . based on the lack of [ o iii ] @xmath125007 , the shock velocities in hh 47a are @xmath13 90 km@xmath2s@xmath3 @xcite , and the bow shock model that fits the line widths and the entire optical / uv spectrum the best is one where the bow shock has a velocity of 60 km@xmath2s@xmath3 and the mach disk 40 km@xmath2s@xmath3 @xcite . while these shock velocities are generally consistent with the new proper motion data , it is clear that the dynamics within hh 47a is more complex than a simple bow shock / mach disk structure . figs . 5 , 6 , and 8 show that individual clumps within hh 47a have different proper motions , and spread @xmath1 40 km@xmath2s@xmath3 or so relative to the mean motion of hh 47a . for example , knot ah8 clearly moves faster than the adjacent knot ah10 ( fig . 7 ) , and table 1 confirms a relative velocity of @xmath4 40 km@xmath2s@xmath3 for these two knots . arrows in fig . 7 mark two areas on the mach disk that have changed over the five years between the epochs . the changes could result from some sort of fluid instability , or could simply arise when a small dense blob like ah8 plunges through the mach disk . the area between the bow shock and the mach disk is particularly variable in fig . 7 . in general , h@xmath0 , where the flux is determined largely by collisional excitation at the shock and so is proportional to the mass flux across the shock at any given moment , traces morphological changes more quickly than does [ s ii ] , which radiates in a spatially extended zone that averages changes out over a cooling time of tens of years . we have measured precise proper motions from two epochs of _ hst _ images of the prototypical stellar jet hh 47 . uncertainties of the relative proper motions between knots are typically @xmath1 5 km@xmath2s@xmath3 , considerably lower than variations of the internal motions of the knots in the jet , so we can assess how the fluid dynamics operates in this system accurately for the first time . taking the source to be at a distance of 450 pc , the orientation angle of the flow is 37.5 @xmath1 2.5 degrees with respect to the plane of the sky . the bright portion of the jet is about 800 years old , and has an average space velocity of 300 km@xmath2s@xmath3 . although the jet curves , the motion is primarily directed radially away from the source . the ejection angle of the jet varies with time by @xmath1 15 degrees or so , and currently impacts the northern side of the reflection nebula , where an oblique shock redirects the supersonic flow . proper motions in the jet increase with distance along the bright section of the jet in front of the bow shock hh 47a , similar to the behavior of radial velocities . this correlation implies that the source varies intrinsically in velocity , and it is these velocity variations that produce shocks that heat the jet internally . the observed variations in proper motion along the jet are consistent with shock models that were designed to explain the observed emission line ratios in the jet . the bright bow shock hh 47a has a clear mach disk and bow shock , but the object is more complex than a dual shock structure , and has many clumps whose differential motions dominate much of the dynamics in this object . the region between the bow shock and mach disk has a highly variable morphology , and the mach disk also changed structure significantly over the five year interval between images . the individual knots within hh 47a are easiest to explain as small bullets that pass from the jet through the mach disk and working surface to emerge as bumps in the bow shock . understanding hh flow dynamics will require systematic work that spans decades , but the rewards are great because these objects make it possible to test our basic understanding of the time dependences of shocked supersonic flows , a ubiquitous phenomenon in astrophysics . additional epochs of high spatial resolution imaging of hh 47 will be able to follow the dynamics of this fascinating object in real time , and thereby provide ideal constraints for numerical models of working surfaces within collimated supersonic flows . cccrrrrrr jh1 & 8:25:44.009&@xmath1151:00:33.49&117 & 36 & 249 & 76 & 261 & 19 + jh2 & 8:25:44.154&@xmath1151:00:32.27&124 & 32 & 264 & 68 & 273 & 32 + jh3 & 8:25:44.826&@xmath1151:00:25.10 & 80 & 22 & 170 & 46 & 176&166 + jh4 & 8:25:45.556&@xmath1151:00:22.40 & 89 & @xmath111 & 191 & @xmath113 & 191&229 + jh5 & 8:25:45.816&@xmath1151:00:23.60 & 95 & 6 & 203 & 13 & 204&228 + jh6 & 8:25:45.863&@xmath1151:00:21.42&100 & 14 & 214 & 30 & 217&231 + jh7 & 8:25:46.005&@xmath1151:00:19.93&102 & 8 & 219 & 17 & 219&247 + jh8 & 8:25:46.224&@xmath1151:00:17.81&103 & 1 & 220 & 2 & 220&275 + jh9 & 8:25:46.629&@xmath1151:00:14.06 & 96 & 6 & 205 & 12 & 205&350 + jh10 & 8:25:47.415&@xmath1151:00:13.17&111 & 15 & 237 & 31 & 239&359 + jh11 & 8:25:47.215&@xmath1151:00:08.11 & 85 & 35 & 181 & 75 & 196&456 + jh12 & 8:25:48.423&@xmath1150:59:59.35&129 & 10 & 275 & 20 & 276&436 + ah1 & 8:25:49.783&@xmath1150:59:52.42&139 & 16 & 297 & 34 & 299&505 + @xmath14ah@xmath15&8:25:50.262&@xmath1150:59:52.51 & 93 & @xmath112 & 199 & @xmath114 & 199&797 + ah2 & 8:25:50.146&@xmath1150:59:53.65 & 90 & @xmath1112&192 & @xmath1126&194 & + ah3 & 8:25:49.978&@xmath1150:59:50.69&103 & 10 & 221 & 22 & 222 & + ah4 & 8:25:50.156&@xmath1150:59:52.45 & 96 & @xmath117 & 206 & @xmath1116&206 & + ah5 & 8:25:50.260&@xmath1150:59:52.87 & 79 & 6 & 169 & 12 & 169 & + ah6 & 8:25:50.466&@xmath1150:59:53.37 & 94 & @xmath1114&201 & @xmath1131&203 & + ah7 & 8:25:50.208&@xmath1150:59:50.34 & 89 & @xmath118 & 189 & @xmath1116&190 & + ah8 & 8:25:50.307&@xmath1150:59:51.08&115 & @xmath114 & 244 & @xmath118 & 245 & + ah9 & 8:25:50.360&@xmath1150:59:51.84 & 85 & @xmath115 & 182 & @xmath1111&183 & + ah10 & 8:25:50.304&@xmath1150:59:50.42 & 96 & @xmath1110&204 & @xmath1121&205 & + ah11 & 8:25:50.424&@xmath1150:59:51.51 & 91 & @xmath118 & 195 & @xmath1117&195 & + ah12 & 8:25:50.409&@xmath1150:59:50.77 & 98 & @xmath119 & 210 & @xmath1118&211 & + js1 & 8:25:44.017&@xmath1151:00:33.83&125 & 32 & 267 & 68 & 276 & 16 + js2 & 8:25:44.816&@xmath1151:00:25.12 & 88 & 20 & 187 & 43 & 192&151 + js3 & 8:25:45.011&@xmath1151:00:24.17&114 & 18 & 243 & 38 & 246&135 + js4 & 8:25:45.312&@xmath1151:00:23.18&115 & 6 & 245 & 13 & 245&159 + js5 & 8:25:45.445&@xmath1151:00:21.19 & 96 & 3 & 204 & 7 & 204&215 + js6 & 8:25:45.532&@xmath1151:00:22.39 & 97 & @xmath112 & 207 & @xmath113 & 207&210 + js7 & 8:25:45.811&@xmath1151:00:23.53 & 94 & 5 & 200 & 10 & 200&232 + js8 & 8:25:45.688&@xmath1151:00:21.16&102 & @xmath117 & 219 & @xmath1115&219&217 + js9 & 8:25:45.870&@xmath1151:00:21.43&102 & 15 & 218 & 32 & 220&227 + js10 & 8:25:46.031&@xmath1151:00:19.25&108 & 8 & 231 & 17 & 231&240 + js11 & 8:25:46.319&@xmath1151:00:17.95&108 & 2 & 230 & 5 & 230&269 + js12 & 8:25:46.697&@xmath1151:00:15.18&111 & @xmath112 & 236 & @xmath115 & 236&303 + js13 & 8:25:46.951&@xmath1151:00:15.06&114 & @xmath112 & 244 & @xmath115 & 244&311 + js14 & 8:25:47.274&@xmath1151:00:11.89&107 & 10 & 229 & 21 & 230&370 + js15 & 8:25:47.284&@xmath1151:00:08.15&116 & 3 & 248 & 7 & 248&364 + js16 & 8:25:47.655&@xmath1151:00:08.53&122 & 15 & 260 & 32 & 262&365 + js17 & 8:25:47.544&@xmath1151:00:06.02&118 & 5 & 251 & 12 & 251&387 + js18 & 8:25:48.048&@xmath1151:00:02.95&118 & 4 & 252 & 9 & 252&434 + js19 & 8:25:48.482&@xmath1151:00:00.36&118 & 8 & 251 & 16 & 252&476 + js20 & 8:25:49.335&@xmath1150:59:55.84&133 & 1 & 283 & 2 & 283&492 + as1 & 8:25:49.783&@xmath1150:59:52.42&140 & @xmath111 & 299 & @xmath111 & 299&505 + @xmath14as@xmath15&8:25:50.254&@xmath1150:59:52.57 & 95 & @xmath115 & 203 & @xmath1112&203&779 + as2 & 8:25:50.199&@xmath1150:59:56.11 & 76 & @xmath114 & 162 & @xmath118 & 162 & + as3 & 8:25:50.316&@xmath1150:59:56.31 & 92 & 5 & 196 & 11 & 197 & + as4 & 8:25:50.284&@xmath1150:59:54.73 & 83 & 3 & 176 & 6 & 176 & + as5 & 8:25:50.390&@xmath1150:59:55.87 & 92 & @xmath111 & 197 & @xmath112 & 197 & + as6 & 8:25:50.379&@xmath1150:59:54.87 & 90 & @xmath117 & 192 & @xmath1115&193 & + as7 & 8:25:50.550&@xmath1150:59:54.54 & 91 & @xmath111 & 195 & @xmath112 & 195 & + as8 & 8:25:50.477&@xmath1150:59:56.62 & 83 & @xmath117 & 178 & @xmath1115&179 & + as9 & 8:25:50.448&@xmath1150:59:54.45 & 98 & @xmath116 & 210 & @xmath1112&210 & + as10 & 8:25:50.475&@xmath1150:59:53.65 & 93 & @xmath118 & 198 & @xmath1118&199 & + as11 & 8:25:50.064&@xmath1150:59:50.49&102 & 6 & 218 & 13 & 218 & + as12 & 8:25:50.187&@xmath1150:59:49.76 & 86 & 3 & 184 & 6 & 184 & + as13 & 8:25:50.165&@xmath1150:59:54.43 & 94 & @xmath114 & 201 & @xmath119 & 201 & + as14 & 8:25:50.118&@xmath1150:59:53.67&108 & 6 & 230 & 13 & 230 & + as15 & 8:25:50.269&@xmath1150:59:52.92 & 85 & 8 & 182 & 17 & 183 & + as16 & 8:25:50.380&@xmath1150:59:51.87 & 86 & @xmath1113&184 & @xmath1128&187 & + as17 & 8:25:50.417&@xmath1150:59:51.27 & 99 & @xmath114 & 211 & @xmath118 & 211 & + as18 & 8:25:50.328&@xmath1150:59:51.04&111 & @xmath116 & 236 & @xmath1112&236 & + as19 & 8:25:50.314&@xmath1150:59:50.40&101 & @xmath1111&216 & @xmath1124&217 & + as20 & 8:25:50.215&@xmath1150:59:51.59 & 67 & 10 & 143 & 21 & 144 & + d & 8:25:53.64 & @xmath1150:59:32 & 84 & 0 & 179 & 0 & 179&1333 +

we present a proper motion study of the shock waves within the classic stellar jet hh 47 based on hubble space telescope h@xmath0 and [ s ii ] images of the region taken over two epochs . individual knots within the jet and in the bow shock / mach disk working surface of hh 47a move significantly in the five years that separate the images , and the excellent spatial resolution of hst makes it possible to measure the proper motions with enough precision to easily observe differential motions throughout the flow . the bright portion of the jet emerges at 37.5 @xmath1 2.5 degrees from the plane of the sky with an average velocity of 300 km@xmath2s@xmath3 . dynamical ages of the shock waves in the jet range from a few decades for knots recently ejected by the source to @xmath4 1300 years for the faint extended bow shock hh 47d . the jet curves , but motions of knots in the jet are directed radially away from the exciting source , and velocity variability in the flow drives the shock waves that heat the jet internally . the jet orientation angle varies with time by about 15 degrees , and currently points to the northwestern portion of a cavity outlined by a reflection nebula , where a quasi - stationary shock deflects the jet . the major working surface hh 47a is more complex than a simple bow shock / mach disk , and contains numerous clumps that move relative to one another with velocities of @xmath4 @xmath1 40 km@xmath2s@xmath3 . small clumps or instabilities affect the mach disk , and dense clumps may move all the way through the working surface to cause the bumpy morphology seen at the bow shock . a localized area between the bow shock and mach disk varies significantly between the two sets of images .

introduction observations and data reduction proper motions, variability, clumpy flow, and the cavity of the hh 47 jet conclusions

in recent years the field of radio transients has seen much interest , most of which has been focused on @xmath0 1 ghz radio emission @xcite . only a handful of blind searches have been carried out below 100 mhz @xcite leaving this a relatively unexplored region of the spectrum . a recent study with the lwa1 yielded two promising transients below 40 mhz @xcite . this search was focused on placing limits on prompt emission from gamma ray bursts and therefore only the times shortly after the occurrence of grbs were searched . in total @xmath1 200 hours of data were analyzed . the two detected transients were not associated with grbs , but at the time of publication their origins were unknown . however the need to conduct further investigation was evident . in this letter we present an analysis of over 11,000 hours of all - sky images recorded by the lwa1 . including the two previously mentioned events , a total of 49 transients have been detected in this data and there is strong correlation to large meteors known as fireballs . the lwa1 is a radio telescope located in central new mexico , operating between 10 and 88 mhz . it consists of 256 cross - dipole antennas spread over an ellipse of 100 @xmath2 110 m. there are also 5 additional antennas at distances ranging from 200 to 500 m from the center of the array @xcite . the telescope is collocated with the karl jansky very large array ( vla ) , at a latitude of 34.070@xmath3 n and a longitude of 107.628@xmath3 w. pasi , a backend to the lwa1 , correlates the signals from all antennas in real time , integrating for 5 s with 75 khz bandwidth tunable to a center frequency anywhere within the 78 mhz over which the lwa1 operates . pasi also produces dirty images of the entire @xmath4 sr sky above the lwa1 @xcite , and in april 2012 began saving the images to a permanent archive . this archive now contains 11,000 hours of all - sky images at various center frequencies . the vast majority of the data has been recorded at center frequencies of 37.8 , 37.9 , 52.0 , and 74.0 mhz , with 30% , 31% , 17% , and 13% recorded at those frequencies , respectively . image subtraction algorithms have been developed to search for transients in the full 11,000 hours of pasi data . the underlying method is that from every image , the image 15 seconds prior is subtracted . pixels above 6 @xmath5 of the image noise are reported as candidate events and the equatorial coordinates are calculated . all four stokes parameters and total spectra are saved and used to check if the candidate event is radio frequency interference ( rfi ) . the coordinates of events are also checked against known bright sources which can be focused by the ionosphere at low frequencies @xcite . this method has resulted in the discovery of 22 transients at 37.8 mhz , 20 at 37.9 mhz , 1 at 29.9 mhz , 1 at 25.6 mhz , and none at either 52.0 mhz or 74.0 mhz . the majority of these transients display a fast rise and exponential decay , lasting for tens of seconds to up to a few minutes , and have flux densities ranging from 500 to 3500 jy ( fig . they also show constant power across the 75 khz band and contain low polarization levels consistent with instrumental leakage of unpolarized sources . most of the transients appear as point sources , meaning they are limited to @xmath6 at 38 mhz . however some are extended over several degrees across the sky , most of which span less than ten degrees . in one case on january 21 2014 , a source leaves a trail covering 92@xmath3 in less than ten seconds ( fig . the trail then slowly recedes to the end point which glows for @xmath790 seconds . the only known source that could cover this distance across the sky in less than 10 seconds and leave a persistent trail is a fireball . to investigate this further , the transients were compared with the detections from nasa s all sky fireball network . the network consists of 12 all - sky cameras , two of which are situated in southern new mexico and share a portion of the sky with the lwa1 . the cameras are used to determine the 3 dimensional position , speed , absolute magnitude , and mass of the of the fireballs . while the two new mexico stations are too far south east of the lwa1 to have detected the extended transient candidate , 5 of the other 44 transients correlate in both space and time to fireballs . the fireball network could not have seen the remaining 39 events because they were either too far north west or occurred during the day . figure 3 shows a histogram of the events , demonstrating that groups of transient detections appear around the times of meteor showers . this implies that a large fraction of the events not seen by the network are most likely meteors as well . the mean delay between an event and the nearest previous major meteor shower in figure 3 is 2.3 time bins . in order to quantify the probability of this distribution with respect to the meteor showers , we performed a monte carlo simulation . random dates were generated and checked if they occurred during actual pasi observing time at 25.6 , 37.8 , or 37.9 mhz . if they did occur during times at these frequencies , they were saved into an array , otherwise they were discarded . once 44 events were compiled , the average delay between the simulated events and the closest of the four major meteor showers was calculated . this was performed 10@xmath8 times and a normalized histogram of the results are shown in figure 3 . the actual data lies between 2 and 3 bins which has a probability of 0.2% . older events from the fireball network were used to search locations and times for our oldest data , which was taken at 25.6 mhz and was only saved as movie files which are not searchable with the image subtraction algorithm . in this data an additional five extremely bright correlated events were found , one of which covered @xmath7 65@xmath3 passing through zenith and was of comparable brightness to the galactic plane . these additional events bring the total number to 49 , 10 of which are correlated directly to fireballs . the fireballs themselves are of the more energetic variety seen by the fireball network . the velocities of all 10 were above 50 km s@xmath9 , with an average velocity of 68 km s@xmath9 . the observed optical emission of each event lasted for @xmath7 1 s , and was followed shortly ( usually within 1 pasi integration ) by the onset of the radio emission . the four seen near 38 mhz had peak absolute visual magnitudes of @xmath7 @xmath104.8 , @xmath106.1 , @xmath106.3 , and @xmath106.4 . the six seen at 25.6 mhz had peak absolute visual magnitudes of @xmath7 @xmath103.3 , @xmath104.1 , @xmath104.3 , @xmath104.7 , @xmath106.6 , and @xmath107.1 . meteors with visual magnitudes brighter than @xmath104 occur about once every 20 hours @xcite . the chance overlap in space and time of 5 events found randomly in pasi data with fireballs brighter than magnitude @xmath104 within the 11,000 hours of data is about 1 in 10@xmath11 . therefore it is clear that these correlations are not mere coincidence but , indeed , connected to fireballs . figure 4 shows images of two fireballs observed simultaneously by the fireball network and the lwa1 . the ionized trails left by meteors have long been known to scatter radio waves , and since the 1940s the use of radar echoes became a popular method for observing their trails allowing the detection of very dim as well as daytime meteors unobservable by optical instruments . a large fraction of these observations have been of specular trails , which occur when a line perpendicular to the meteor s path satisfies the reflection requirement that the angle to the transmitter equals the angle to the receiver @xcite . the majority of these reflections last for @xmath1 1 s , have electron line densities of @xmath12 cm@xmath9 , and are known as underdense meteors . a less common variety of specular trail last for several seconds , have electron line densities of @xmath14 cm@xmath9 , and are known as overdense meteors . these are associated with larger meteors , the largest of which are associated with optical fireballs @xcite . a smaller subset of meteor trail echoes are non - specular ( i.e. due to scattering ) , which are not yet well understood . they most often occur when the radar is pointed perpendicular to the geomagnetic field , but do not necessarily satisfy the specular reflection requirement . these types of trails can last from seconds to several minutes , but are typically much weaker than specular trails and therefore high power large - aperture radars are required to detect them @xcite . meteor trail echoes are a well studied phenomenon , and their characteristics are well documented . moreover a recent study by @xcite using 55.25 mhz analog tv broadcasting stations for meteor scatter provides a direct example of how to detect meteor echoes with the lwa1 and how to identify them in the all - sky data . in comparison , several key differences arise between the transients reported in this paper and what is expected from trail echoes . the differences are as follows : first , typical transmitters are strongly polarized , resulting in reflections that are strongly polarized @xcite . however no significant amount of linear nor circular polarization has been detected from the observed transients . secondly , a large portion of the rfi seen by lwa1 is narrower than the 75 khz pasi band and is easily identifiable by its spectra @xcite . yet none of the observed transients contain any spectral features . thirdly , the light curves of the observed transients are consistent with each other , ranging from 30 to 150 s , showing a linear rise and a long exponential decay , and and otherwise show a smooth evolution ( figure 1 ) . a typical reflection from an overdense trail reaches maximum brightness in just a few seconds , maintains a relatively constant average brightness while undergoing sporadic dimming and rebrightening . it then quickly decays away once it expands to the point that the density reaches the underdense criteria @xcite . the observed transients are also inconsistent with light curves from non - specular echoes , which vary greatly from one to the next , following no particular pattern . more importantly , however , non - specular echoes are weaker and more rare than overdense specular echoes @xcite . therefore if the lwa1 were seeing non - specular reflections it should also see many more bright specular reflections scattering from the same transmitters . finally , the observed transients have azimuths and elevations consistent with the uniform distribution convolved with the lwa1 power pattern ( figure 5 ) . this distribution implies that the sources appear in random locations with no preferable sky position . this is inconsistent with what is expected from specular echoes of man - made radio frequency interference ( rfi ) , which should increase towards the horizon due to the increased number of incident angles with distant transmitters required for forward scattering @xcite . the observed pattern could be consistent with nearby transmitters . however , because the signal strength depends on the inverse - cube of the distance to the meteor , there should be many very bright nearby rfi sources on the horizon but these are not observed . this pattern is also inconsistent with non - specular reflections , which are preferentially located in a relatively small region of the sky that satisfies the requirement that the pointing vector is perpendicular to the geomagnetic field @xcite . for these reasons it seems unlikely that forward scattering is responsible for the signals detected from fireballs . it is therefore our conclusion that fireball trails radiate at low frequencies . since pasi can only record at one center frequency at a time , there is limited spectral information for each event . nevertheless the similarity of the light curves of fireballs recorded at 25.6 , 29.9 , 37.8 , and 37.9 mhz implies that this emission is broad band . it also appears that they are brighter at 25.6 than at 38 mhz , suggesting a spectral slope favorable to lower frequencies . also , there have been no detections at 52.0 or 74.0 mhz which may be the result of a sharp cutoff at higher frequencies or a steep spectrum . in either case it is clear from these observations that the emission is non - thermal , yet the exact mechanism is currently unknown . if a magnetic field of 10 to 15 g were present within the trail , it follows that cyclotron radiation would be emitted at the observed frequencies by the electrons in the plasma . however , the surface geomagnetic field is only 0.5 g , so this would require the generation of a strong magnetic field by a fireball , an effect that has never been observed . brightness temperature for a radio source can be calculated using the equation : @xmath15 where @xmath16 is the flux density of the source , @xmath17 is the observed wavelength , @xmath18 is the boltzmann constant , and @xmath19 is the solid angle of the source . the solid angle of the emitting portion of the fireball is estimated by using typical size scales reported for bright meteors trails . a typical fireball observed by the lwa1 has a peak flux density of 1 kjy , 20 s after first light , and most are point sources with a measured beam size of @xmath20 at 37.8 mhz . meteor plasma trails are typically modeled as long cylinders with radii much smaller than their length . assuming a height of 110 km and zenith angle of 40@xmath3 , an angular size of @xmath20 corresponds to 12 km , which is a reasonable value for the length of the emitting part of the plasma trail . however for large plasma trails a typical initial radius is 10 m , which quickly expands due to diffusion . it is estimated that at 20 s after entry , the radius of the expanding trail would be @xmath7100 m @xcite . given these dimensions the estimated brightness temperature is @xmath21 k. this temperature is two orders of magnitude higher than the typical peak temperature of a fireball trail , and this provides further evidence that the emission is non thermal . while the spectral slope has yet to be fully measured , the total energies of the radio emission are estimated by assuming a flat spectrum from 10 to 50 mhz and heights of 110 km . these assumptions yield total radio energy estimates ranging from 10@xmath22 to 10@xmath23 j , which is one part in @xmath24 to @xmath25 of the kinetic energy of a typical fireball . further observations are in progress to better characterize the spectral properties of this emission , and to get better estimates of the total amount of energy radiated at radio frequencies . decametric radio emission from meteors has not been previously detected , but this is not the first time its existence has been discussed . @xcite conducted a search for radio emission , but reported only upper limits with a 5 @xmath5 sensitivity of @xmath26 jy at 30 mhz for 1 s bursts . it is also interesting to note that in the last several decades detections of extremely low frequency ( elf ) and very low frequency ( vlf ) emission have been reported coincident with large meteors @xcite . the physical mechanism responsible for this emission is not well understood , but might be related to our detections of higher frequency emission . given the vast range in energies and size scales of meteors and their corresponding plasma trails , this emission may exist at a wide range of frequencies , timescales , and energies . investigating this emission further will yield new insights into the physics of meteors and their interaction with our atmosphere . moreover fireballs are now a known radio transient foreground source and need to be taken into account when searching for cosmic transients . it is interesting to note that transient atmospheric phenomena , unknown to emit at radio frequencies , have been proposed as the possible source of perytons and perhaps even fast radio bursts @xcite . in fact katz ( 2014 ) suggests meteors as a possible source of perytons . in this paper we reported observations of high energy meteors generating bright radio emission . construction of the lwa1 has been supported by the office of naval research under contract n00014 - 07-c-0147 . support for operations and continuing development of the lwa1 is provided by the national science foundation under grants ast-1139963 and ast-1139974 of the university radio observatory program . 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 . beech , m. , brown , p. , & jones , j. , 1995 , earth , moon , and planets , 68 , 181 bourdillon , a. , haldoupis , c. , hanuise , c. , le roux , y. , & menard , j. , 2005 gephysical research letters , 32 bronshten v.a . 1983 , physics of meteoric phenomena , d. reidel publishing company burke - spolaor , s. , bailes , m. , ekers , r. , maquart , j. , & crawford , f. 2011 , apj , 727 , 5 ceplecha , z. , borovicka , j. , elford , w.g . , et al . 1998 space science reviews , 84 , 327 close , s. , kelley , m. , vertatschitsch , l. et al . 2011 , jgr , 116 , a1 cohen , a. s. , lane , w.m . , cotton , w.d . 2007 , aj , 134 , 1245 cutchin , s.e . 2011 , ph.d . thesis , virginia polytechnic institute and state university ellingson , s.w . , craig , j. , dowell , j. , taylor , g.b . , & helmboldt , j.f . 2013 , eprint arxiv:1307.0697 frail , d.a . , kulkarni , s.r . , ofek , e. o. , bower , g.c . , & nakar , e. 2012 , apj , 747 , 1 , 12 guha , a. , de , b.k . , choudhury , a. , & roy , r. 2012 , astrophysics and space science , 341 , 287 halliday , i. , griffin , a.a . , blackwell , a.t . 1996 meteoritics and planetary science , 31 , 185 hawkins , g.s . 1958 , nature , 181 , 1610 helmboldt , j.f . , ellingson , s.w . , hartman , j.m . 2014 , radio science kardashev , n.s , soglasnov , v.a . , saveleva n.a . 1977 , soviet astronomy , 21 , 1 katz , j.i . 2014 , arxiv:1403.0637 keane , e.f . , stappers , b.w . , kramer , m. , & lyne , a.g . 2012 , mnras , 425 , l71 keay , c.s.l . 1980 , science , 210 , 11 kulkarni , s.r . , ofek , e.o . , neill , j.d . , zheng , z. , & juric , m. 2014 , 2014arxiv1402.4766 lazio , t . j.w . , clarke , t.e . , lane , w.m . 2010 , aj , 140 , 1995 lorimer , d.r . , bailes , m. , mclaughlin , m.a . , & narkevic , d.j . , crawford , f. 2007 science , 318 , 777 mckinley , d.w.r . 1961 , meteor science and engineering , mcgraw - hill mcmullin , j. p. , waters , b. , schiebel , d. , young , w. , & golap , k. 2007 , astronomical data analysis software and systems xvi ( asp conf . 376 ) , ed . r. a. shaw , f. hill , & d. j. bell ( san francisco , ca : asp ) , 127 obenberger , k.s . hartman , j.m . , taylor , g.b . 2014 , apj , 785 , 1 , 7 obenberger , k.s & dowell , j. 2011 lwa1 rfi survey memo 183 , lwa memo series http://www.ece.vt.edu/swe/lwa/memo/lwa0183.pdf sugar , g. , oppenheim , m.m . , bass , e. , chau , j.l . , jgrsp , 115 , a12 taylor , g.b . , ellingson , s.w . , kassim , n.e . , et al . 2012 , astronomical instrumentation , 1 , 1 thornton , d. , stappers b. , bailes , m. , et al . 2013 , science , 341 , 53 wayth , r.b . , tingay , s.j . , deller , a.t . 2012 , apj , 753 , 2 , l36 wislez , j.m . 1995 , proceedings of the international meteor conference , roggemans , p. & knofel , a. , 83

we present the findings from the prototype all - sky imager ( pasi ) , a backend correlator of the first station of the long wavelength array ( lwa1 ) , which has recorded over 11,000 hours of all - sky images at frequencies between 25 and 75 mhz . in a search of this data for radio transients , we have found 49 long ( 10s of seconds ) duration transients . ten of these transients correlate both spatially and temporally with large meteors ( fireballs ) , and their signatures suggest that fireballs emit a previously undiscovered low frequency , non - thermal pulse . this emission provides a new probe into the physics of meteors and identifies a new form of naturally occurring radio transient foreground .

introduction observations discovery of transients and correlations with fireballs reflection vs emssion physical constraints on the emission discussion acknowledgments

the firs release of _ planck _ data improves the quality of cosmological data extraordinary @xcite . it allows us to give a tighter constraint to the cosmological parameter space . and the cold dark matter plus a cosmological constant @xmath4 , the so - called @xmath4cdm model , can almost agree with the most recent cosmic observations which include the type - ia supernovae ( sn ) , the baryon acoustic oscillation ( bao ) and the cosmic microwave background radiation ( cmb ) successfully at large scales . however , it has several potential problems on smaller scales @xcite . how to explain the discrepancies on large and small scales is currently still under debate @xcite . the warm dark matter has been proclaimed as a potential solution to the small scale difficulties of cold dark matter @xcite . it leaves some space to an alternative to the cold dark matter model . the focus point is whether it is cold or warm . actually , the hot dark matter was ruled out due to the difficulty in forming the observed large scale structure . to characterize the properties of warm dark matter , the equation of state @xmath0 is an important indicator in a fluid perspective . and the value of @xmath7 should be determined by the cosmic observations . a significant nonzero value of @xmath0 indicates the dark matter is warm rather than cold . in the literature , the equation of state of wdm was constrained by many groups under the assumption of a constant @xmath0 , a time variable @xmath0 with a cosmological constant or @xmath8 dark energy . for an example , please see @xcite and references therein . see also ref . @xcite , in which the difficulties of the small scale behavior of warm dark matter were pointed out . if the dark matter is really warm the model parameters would be different even it were to have the same background evolution history as that of the @xmath4cdm model . a different large scale structure would form due to the different perturbation evolutions . therefore , the large scale structure information will be important to break degeneracies between model parameters and to determine the properties of dark matter which is the main part of the large scale structure of our universe . we would also like to mention the particle side of dark matter ; please see ref . @xcite for brief reviews . thanks to the measurements of wigglez dark energy survey , a total @xmath9 galaxies in the redshift range @xmath10 were measured . these galaxies were split into four redshift bins with ranges @xmath11 , @xmath12 , @xmath13 and @xmath14 . the corresponding power spectrum in the four redshift bins was measured ; for details , please see @xcite . we estimate the nonlinear growth from a given linear growth theory power spectrum based on the principles of the halo model ; actually the halofit formula will be used in this paper @xcite . at the scale of halo , the growth of halos depends on the local physics , and not on the details of precollapse matter and the large scale distribution of matter . thus , in the nonlinear regime the growth depends only on the nonlinear scale , the slope and curvature of the power spectrum @xcite . in this paper , we loosen the constraint to a zero equation of state and investigate the simplest model for dark matter , i.e. the one with a constant @xmath0 . therefore , the main information is stored in the matter power spectrum . also we assume the halofit formula is still suitable for this case , though the formula would be modified due to the free - streaming of warm dark matter @xcite . by a combination of cmb , sdss bao , sn and wigglez , the eos of dark matter will be tested . as results , we did not find significant deviation from @xmath15 in @xmath6 regions . actually , this simplest form dark matter was already constrained by using @xmath16 sn , cmb shift parameter @xmath17 and bao distance parameter @xmath18 @xcite ( see also in refs . @xcite ) , in which a time variable equation os state of dark matter was also considered . the authors of ref . @xcite discussed the degeneracies between model parameters extensively and found the equation os state of dark matter in the range of @xmath19 regions @xmath20 by using the data combination of wmap5+sdss+snls . it implies the cold dark matter @xmath15 is compatible in the @xmath6 region . as a revisit to the work of ref . @xcite and a comparison to the work of ref . @xcite , here we will use the full information of cmb , which includes the recently released _ planck _ data sets , which include the high - l tt likelihood ( _ camspec _ ) up to a maximum multipole number of @xmath21 from @xmath22 ; the low - l tt likelihood ( _ lowl _ ) up to @xmath23 ; and the low - l te , ee , bb likelihoods up to @xmath24 from wmap9 ; the data sets are available on line @xcite . for the sn data points as `` standard candles '' , the luminosity distances will be employed . in this paper , we keep to use the snls3 which consists of @xmath25 sn calibrated by sifto and salt2 ; for details , please see ref . although the photometric calibration of the snls and the sdss supernova surveys were improved @xcite , they are still unavailable publicly . for the bao data points as `` standard ruler '' , we use the measured ratio of @xmath26 , where @xmath27 is the comoving sound horizon scale at the recombination epoch , and @xmath28 is the `` volume distance '' which is defined as @xmath29^{1/3},\ ] ] where @xmath30 is the angular diameter distance . here the bao measurements from wigglez are not included , as they come from the same galaxy sample as the @xmath31 measurement . because the full information of cmb and wigglez power spectrum is employed , a tiger constraint is expected . our results will show that it is indeed the case . the plan of this paper is as follows : in sec . [ sec : bpes ] , we present the main background evolution and perturbation equations for dark matter with an arbitrary equation os state . in sec . [ sec : results ] , the constrained results are presented via the markov chain monte carlo ( mcmc ) method . section [ sec : con ] is the conclusion . the equation os state of dark matter is given as @xmath32 the friedmann equation for a spatially flat frw universe reads @xmath33,\label{eq : ehfe}\ ] ] where @xmath34 are the present dimensionless energy densities for the radiation , the baryon , the dark matter and the cosmological constant respectively , where @xmath35 is respected for a spatially flat universe . in this paper , the dark matter is taken as a perfect fluid with a constant equation os state , and then in the synchronous gauge the perturbation equations of density contrast and velocity divergence for the dark matter are written as @xmath36 following the notations of ma and bertschinger @xcite , in which the definition of the adiabatic sound speed @xmath37 is used . here the equations are also ready for a general form of dark matter . when the equation os state of a pure barotropic fluid is negative , the imaginary adiabatic sound speed can cause instability of the perturbations . to overcome this problem , one can introduce an entropy perturbation and assume a positive or null effective speed of sound . following the work of ref . @xcite , the nonadiabatic stress or entropy perturbation can be separated out @xmath38 which is gauge independent . in the rest frame of dark matter , the entropy perturbation is specified as @xmath39 where @xmath40 is the effective speed of sound . transforming into an arbitrary gauge @xmath41 gives a gauge - invariant form for the entropy perturbations . by using eqs ( [ eq : entropyper ] , ) ( [ eq : restframe ] ) and ( [ eq : gaugetrans ] ) , one can recast eqs . ( [ eq : continue ] ) , and ( [ eq : euler ] ) into @xmath42,\label{eq : wdmdelta}\\ \dot{\theta}_{dm}&=&-\mathcal{h}(1 - 3c^2_{s , eff})\theta_{dm}+\frac{c^2_{s , eff}}{1+w_{dm}}k^2\delta_{dm}-k^2\sigma_{dm}\label{eq : wdmv}. \end{aligned}\ ] ] for the dark matter , we assume the shear perturbation @xmath43 and the adiabatic initial conditions . actually , the effective speed of sound @xmath40 is another freedom to describe the microscale property of dark matter in addition to the equation os state . and , we should take it as another free model parameter . the sound speed determines the sound horizon of the fluid via the equation @xmath44 . the fluid can be smooth or cluster below or above the sound horizon @xmath45 respectively . if the sound speed is smaller , the perturbation of the fluid can be detectable on large scale . in turn the clustering fluid can influence the growth of density perturbations of matter , large scale structure and evolving gravitational potential which generates the integrated sachs - wolfe effects . the dark matter is responsible to formation of the large scale structure of our universe , so we assume the effective speed of sound @xmath46 in this work . to obtain the equation os state of dark matter from the currently available cosmic observations , we use the mcmc method which is efficient in the case of more parameters . we modified the publicly available * cosmomc * package @xcite to include the perturbation evolutions of dark matter with a general form of equation os state according to the eqs . ( [ eq : wdmdelta ] ) and ( [ eq : wdmv ] ) . we adopted the seven - dimensional parameter space @xmath47\}\ ] ] the priors for the model parameters are summarized in table [ tab : results ] . furthermore , the hard - coded prior on the comic age @xmath48 is also imposed . also , the new hubble constant @xmath49 @xcite is adopted . the pivot scale of the initial scalar power spectrum @xmath50 is used in this paper . after running eight chains for every cosmological model with different values of @xmath0 on the _ computing cluster for cosmos _ , we show the obtained results in table [ tab : results ] . this result is compatible with the previous results as shown in refs . @xcite ; however a tighter constraint was obtained due to the high quality of current cosmic observational data points and the inclusion of wigglez measurements of the power spectrum . the @xmath51 and @xmath52 contours for @xmath53 , @xmath54 and @xmath0 were plotted in figs . [ fig : wnowigglecontoure ] and [ fig : wcontoure ] , where the degeneracy between model parameter @xmath53 and @xmath0 can be understood easily : the amount of @xmath53 at the decoupling epoch is fixed by the cmb power spectra ; to maintain the same early time values of @xmath53 , lower present values of @xmath53 need larger values of @xmath0 . this implies that the determination of @xmath53 is vital to pin down the equation os state of dark matter . the inclusion of wigglez measurements of power spectrum decreases the values of @xmath0 and @xmath54 and shrinks the bounds mildly . for comparison , by using the same data sets combination , the model parameters of the @xmath4cdm model were also obtained as shown in table [ tab : results ] and fig . [ fig : lcdmcontoure ] . the difference of the minimum @xmath3 between the @xmath4cdm and @xmath4wdm models is @xmath5 for one extra model parameter @xmath0 . although the currently available cosmic observations favor the @xmath4wdm mildly , no significant deviation from @xmath4cdm model is found in @xmath6 regions . .the mean values with @xmath55 errors and the best fit values of model parameters for general relativity and modified gravity theory , where snls3 , bao , _ planck_+wmap9 with and without wigglez measurements of matter power spectrum are used . [ cols="^,^,^,^,^,^,^,^",options="header " , ] c.l . , @xmath56 c.l . for the @xmath4wdm model by using cmb+bao+sn data points.,width=321 ] , but with the wigglez measurements of the power spectrum . , width=321 ] , but with the wigglez measurements of the power spectrum for the @xmath4cdm model.,width=321 ] to show the effects of @xmath0 on the matter power spectra , we fix the other relevant model parameters to their mean values as given in table [ tab : results ] and vary the values of @xmath0 from the negative to the positive . their effects on the matter power spectrum due to different values of @xmath0 are shown in fig . [ fig : mp ] , where the redshift is fixed to @xmath57 . the constrained result showed the values of @xmath0 is very close to the cold dark matter @xmath15 ; then the linear matter power spectra of them will not be too different as expected . as shown in fig . [ fig : mp ] , larger negative values of @xmath0 move the matter and radiation equality to later times and oscillate the matter power spectra ( blue , dotted line ) . and large positive values of @xmath0 move the matter and radiation equality to earlier times and increase the matter power spectrum . therefore it can be easily understood that hot dark matter is ruled out due to the significant difference on the observations of the large scale structure of our universe . for different values of the eos of dark matter @xmath0 where the other relevant model parameters are fixed to their mean values as shown in table [ tab : results].,width=377 ] in this paper , we have investigated the constraints to the equation of state of dark matter by using the currently available cosmic observational data sets , which include the cmb of the first 15.5 months from _ planck _ , snls3 , sdss bao and wigglez measurements of power spectrum . the previous results were updated . we have found that the latest data provide the constraints @xmath58 at @xmath56 c.l .. this result is compatible with the previous results , but a relative tighter constraint was obtained due to the high quality of the currently available data points . the difference of the minimum @xmath3 between the @xmath4cdm and @xmath4wdm models is @xmath5 for one extra model parameter @xmath0 . although the currently available cosmic observations favor the @xmath4wdm mildly , no significant deviation from the @xmath4cdm model is found in the @xmath6 region .

in this paper , we loosen the zero equation of state of dark matter to a constant @xmath0 . by using the currently available cosmic observations which include the type - ia supernovae , the baryon acoustic oscillation , the wigglez measurements of matter power spectrum and the cosmic microwave background radiation from the firs release of _ planck _ dat through the markov chain monte carlo method , we found the equation of state of dark matter in @xmath1 regions : @xmath2 . the difference of the minimum @xmath3 between the @xmath4cdm and @xmath4wdm models is @xmath5 for one extra model parameter @xmath0 . although the currently available cosmic observations favor the @xmath4wdm mildly , no significant deviation from the @xmath4cdm model is found in @xmath6 regions .

introduction background and perturbation equations constrained results conclusion

in the quote above , made at the first workshop on particle correlations and femtoscopy in the czech republic , professor biaas was expressing a frustration felt periodically by those of us who labor to understand deeply the fascinating features of soft - scale qcd as manifest in the quark - gluon plasma , a bulk thermodynamic system of deconfined colored partons as degrees of freedom . every time we gain a deeper insight into the physics and phenomenology of this system ( the dream ) , more detailed theories ( the ants ) or experimental observations make clear that the system is more complicated than we thought . new advances often raise more questions than they answer . professor biaas made this statement with a smile on his face , however . he clearly considers himself an ant in the spirit of the quote from proverbs : a worker with a mission much larger than himself , destined to build , piece by piece over the course of his life , an edifice in pursuit of that mission . he clearly relishes this role . it turns out that this symposium is held shortly before a milestone birthday of my own , and i found myself contemplating my own much less impressive anthills . one topic i have returned to repeatedly in various forms is the timescale of the system formed in a heavy ion collision . here , i discuss previous studies ( and one unpublished analysis ) to show the development of our understanding of these timescales as measured with two - particle intensity interferometry . to the general public , the field of heavy ion physics resembles high energy particle physics . the accelerators , the collaborations , and the detectors are mammoth . papers are written by committee , and talks are selected according to the bylaws set by councils and led by elected management teams . students are well - versed in the particle zoo ( often much more so , than their professors who grew up as nuclear physicists ) . the origins of the field , however , lie more in the realm of nuclear physics . concepts were developed and people trained in heavy ion experiments at facilities like ganil , sis / gsi , and the nscl / msu cyclotron facility . pions were rare and almost exotic . students were relatively well - versed in nuclear physics . however , heavy ion physics is a field of its own neither nuclear physics ( which strives to understand the nucleus in its seemingly infinite complexity ) nor particle physics ( which attempts to bypass the complexity of all interactions to study symmetries manifest as particles ) . in heavy ion physics , we seek to create and study a new _ system_. ideally , it will be a nearly thermalized system , so that we may study its equation of state . at lower energies , the equation of state of highly compressed , cold matter provides information relevant to the cores of neutron stars @xcite . in ultrarelativistic energies , the equation of state of colored matter near the deconfinement transition probes qcd under the most extreme conditions @xcite . the hot system is self - interacting and characterized by detailed flow fields . its femtoscopic substructure is dynamic and rich , with long lifetimes , anisotropic shapes , correlations between momentum and space - time , and whirling vortices . to understand the evolution of this substructure , it is important to obtain measures of the timescales involved . figure [ fig : evolutioncartoon ] identifies two of them , for the case of ultrarelativistic collisions . unfortunately , they are often conflated , using the ambiguous term `` lifetime ; '' however , they are different , and it is best to keep the distinction clear . the evolution timescale @xmath0 refers to the time between initial interpenetration and particle freezeout . ( particles `` freeze out '' when they cease interacting with each other and the system . ) meanwhile , @xmath1 refers to the duration of the freezeout process itself . the evolution of an ultra - relativistic heavy ion collision is sketched to indicate two relevant timescales , corresponding to the evolution of the entire system and the duration of the freezeout process . see text for details.,scaledwidth=80.0% ] the technique of two - particle intensity interferometry is a well - developed tool to extract spatio - temporal information from dynamic subatomic sources . also known as femtoscopy , it exploits the fact that , given the observation of one particle , the conditional probability to measure a second particle depends on the relative momentum ( measured ) and the relative space - time position ( inferred , by measuring the conditional probability ) of the pair . for details and compilations of results , i refer the reader to reviews at both low @xcite and high @xcite energy collisions . in principle , information about both space and time scales may be extracted by studying multi - dimensional correlation functions in the `` out - side - long '' ( or , for low energies , the `` longitudinal - transverse '' ) system of bertsch and pratt @xcite . here , the `` out '' ( or , for low energies , the `` longitudinal '' ) direction is parallel to the direction of motion of the particles , while the `` side '' ( or `` transverse '' ) is perpendicular to it . a long emission duration ( @xmath1 ) will generate a particle distribution extended in the direction of particle motion , and the resulting correlation will be less if the relative momentum is oriented in this direction . emission duration measurements are of particular interest , because a first - order phase transition from a deconfined to a confined state , is expected to extend the emission time @xcite . through the early nineties , no emission duration greater than @xmath22 fm / c had been observed in the correlation data ; indeed , most extracted timescales were consistent with zero . for the newly available collisions at the multi - gev scale @xcite , this was a disappointing development , though perhaps not shocking . however , at non - relativistic energies available at nscl and ganil , this was surprising indeed . at these lower energies , the collision and evolution dynamics were believed to be better understood . repeated reports of vanishing timescales from correlation measurements led some experts at the time to wonder whether the femtoscopic technique itself was sufficiently well understood . could femtoscopy really measure timescales , after all ? two publications @xcite on proton correlations answered this important question with a resounding affirmative . in one , near multifragmentation energies , a lifetime greater than 10 fm / c was finally extracted from femtoscopic data . in the other , at compound nucleus energies , a lifetime greater than 1000 fm / c ( ! ! ) was reported . both timescales were of the order of theoretical expectations . why had all previous published results reported no difference between longitudinal and transverse correlation functions , and hence emission timescales consistent with zero ? the reason turned out to be simple : at least in the u.s . , we had all been looking in the wrong frame . two - proton correlation functions at small relative momenta probe the space - time geometry of the emitting system , because the magnitude of nuclear and coulomb final - state interaction and antisymmetrization effects depends on the spatial separation of the emitted particles @xcite . the attractive s - wave nuclear interaction leads to a pronounced maximum in the correlation function at relative momentum @xmath3 mev / c . this maximum decreases for increasing source dimensions and/or emission time scales . the coulomb interaction and antisymmetrization produce a minimum at @xmath4 . nonspherical phase - space distributions , predicted for long - lived emission sources , can lead to a dependence of the two - proton correlation function on the direction of the relative momentum @xcite . until 1993 , however , such directional dependences had not yet been observed unambiguously . the first observation was published in 1993 @xcite . the experiment was performed at the national superconducting cyclotron laboratory at michigan state university ( msu ) . a beam of ar ions at e / a=80 mev incident energy and intensity @xmath5/sec bombarded a sc target of areal density 10 mg / cm . charged particles were measured in the msu @xmath6 array , which consisted of 209 plastic @xmath7 phoswich detectors covering polar angles between @xmath8 in the laboratory frame . one of the hexagonal modules of the @xmath6 array , located at @xmath9 in the laboratory frame , was replaced by a 56-element high - resolution hodoscope . each @xmath7 telescope of the hodoscope consisted of a @xmath10-thick si detectors backed by a 10-cm - long csi(tl ) detector and subtended a solid angle of @xmath11 msr . the energy resolution was about 1% for 60 mev protons ; this is important for measuring large source sizes . the problem of identifying finite emission duration is illustrated in figure [ fig : lisa1993fig1 ] . it depicts phase - space distributions in the laboratory rest frame of protons emitted with fixed laboratory velocity @xmath12 towards the detector at @xmath13 for a source at rest in the laboratory ( a ) and for a source at rest in the center - of - momentum system of the projectile and target ( @xmath14 ) . we assumed a spherical source of 7 fm diameter and 70 fm / c lifetime emitting protons of momentum 250 mev / c . schematic illustration of phase - space distributions at a time @xmath15 fm / c , seen by a detector at @xmath13 , for a spherical source of radius @xmath16 fm and lifetime @xmath17 fm / c emitting protons of momentum 250 mev / c . ( a ) source at rest in the laboratory . ( b ) source moves with @xmath14 . in the phase - space distributions , the laboratory velocities of the emitted particles ( @xmath12 ) are depicted by small arrows , and the directions perpendicular and parallel to @xmath12 are depicted by the large double - headed arrows . in ( a ) and ( b ) , @xmath12 is kept constant , and @xmath18 is different ; therefore , the elongations along @xmath18 are different . from @xcite . , scaledwidth=50.0% ] measured longitudinal and transverse correlation functions for protons emitted in central @xmath19 collisions at @xmath20 mev . the correlation functions are shown for proton pairs of total laboratory momentum @xmath21 mev / c detected at @xmath22 . longitudinal and transverse correlation functions ( solid and open points , respectively ) correspond to @xmath23 and @xmath24 , respectively . solid and dashed curves represent longitudinal and transverse correlation functions predicted for emission from a spherical gaussian source with @xmath25 fm and @xmath26 fm / c , moving with @xmath14 . upper panel : @xmath27 and @xmath28 are defined in the rest frame of the presumed source . lower panel : @xmath27 and @xmath28 are defined in the laboratory frame . from @xcite . , scaledwidth=50.0% ] for emission from a source at rest , the phase - space distribution of particles moving with fixed velocity @xmath29 towards the detector exhibits an elongated shape oriented parallel to @xmath12 . a source of lifetime @xmath1 appears elongated in the direction of the proton momentum by an incremental distance @xmath30 . correlation functions for relative momenta @xmath31 reflect a stronger pauli suppression , and hence a reduced maximum at @xmath32 mev / c , than those for @xmath33 . cuts on the relative orientation of @xmath34 and @xmath27 are sensitive to the motion of the source , since the direction of the total momentum depends on the rest frame , while the direction of the relative momentum at least in the nonrelativistic limit does not . ( note : the key to this result is not even a relativistic boost , but simply a galilean one ! ) previous analyses compared the shapes of the correlation functions selected by cuts on the relative angle @xmath35 between @xmath34 and @xmath36 , where @xmath37 and @xmath38 are the laboratory momenta of the two protons and @xmath34 is the momentum of relative motion . such analyses are optimized to detect emission duration effects of sources stationary in the laboratory system , but they can fail to detect such effects for nonstationary sources . for the specific case illustrated in fig . [ fig : lisa1993fig1](b ) , the source dimensions parallel and perpendicular to @xmath39 are very similar , and no significant differences are expected for the corresponding longitudinal and transverse correlation functions . for a source of known velocity , the predicted lifetime effect is detected most clearly if longitudinal and trans- verse correlation functions are selected by cuts on the angle @xmath40 , where the primed quantities are defined in the rest frame of the source . in the frame of the source , the phase - space distribution is always elongated in the direction of @xmath18 . hence , in fig . [ fig : lisa1993fig1](b ) , the source dimensions should be compared in directions parallel and perpendicular to @xmath18 . such analyses can only be carried out for emission from well - characterized sources . figure [ fig : lisa1993fig2 ] corroborates this reasoning with experimental data . it shows longitudinal and transverse two - proton correlation functions for central ar+ sc collisions at e / a=80 mev selected by appropriate cuts on the total transverse energy detected in the @xmath6 array . in a geometrical picture , the applied cuts correspond to reduced impact parameters of @xmath41 . longitudinal ( solid points ) and transverse ( open points ) correlation functions were defined by cuts on the angle @xmath23 and @xmath24 , respectively . the normalization constant c in eq . ( 1 ) is independent of @xmath28 . to maximize lifetime effects and reduce contributions from the very early stages of the reaction , the coincident proton pairs were selected by a low - momentum cut on the total laboratory momentum , @xmath21 mev / c . the top panel shows correlation functions for which the angle @xmath28 was defined in the center - of - momentum frame of projectile and target ( @xmath42 ) ; for central collisions of two nuclei of comparable mass , this rest frame should be close to the rest frame of the emitting source . the bottom panel shows correlation functions for which the angle @xmath28 was defined in the laboratory frame . consistent with the qualitative arguments presented in figure [ fig : lisa1993fig1 ] , a clear difference between longitudinal and transverse correlation functions is observed for cuts on @xmath43 ( top panel of figure [ fig : lisa1993fig2 ] ) but not for cuts on @xmath44 ( bottom panel of figure [ fig : lisa1993fig2 ] ) . the clear suppression of the transverse correlation function with respect to the longitudinal correlation function observed in the top panel in figure [ fig : lisa1993fig2 ] is consistent with expectations for emission from a source of finite lifetime . the solid and dashed curves in the top and bottom panels of figure [ fig : lisa1993fig2 ] depict calculations for emission from a spherical gaussian source comoving with the center - of - momentum frame of the projectile and target . the calculations were performed for the radius and lifetime parameters @xmath25 fm and @xmath45 fm / c the calculations corroborate the qualitative arguments illustrated in figure [ fig : lisa1993fig1 ] . the data in figure [ fig : lisa1993fig2 ] represent the first clear experimental evidence of this predicted lifetime effect . for a more quantitative analysis , we performed calculations assuming a simple family of sources of lifetime and spherically symmetric gaussian density profiles , moving with the center - of - momentum frame of reference . energy and angular distributions of the emitted protons were selected by randomly sampling the experimental yield @xmath46 . specifically , the single particle emission functions were parametrized as @xmath47 contour diagram of @xmath48 determined by comparing theoretical correlation functions to the data shown in the upper panel of figure [ fig : lisa1993fig2 ] . the fit was performed in the peak region of the correlation function @xmath49 mev / c . from @xcite . , scaledwidth=70.0% ] in equation [ eq : gausssource1 ] , @xmath50 , @xmath51 , and @xmath52 are understood as being in the rest frame frame of the source . phase - space points generated in the rest frame of the source were lorentz boosted into the laboratory frame , and the two - proton correlation function was obtained by convolution with the two - proton relative wavefunction . transverse and longitudinal correlation functions were calculated for the range of parameters @xmath53 fm and @xmath54 fm / c . for each set of parameters , the agreement between calculated and measured longitudinal and transverse correlation functions was evaluated by determining the value of @xmath48 in the peak region , @xmath49 mev / c . a contour plot of @xmath48 as a function of @xmath55 and @xmath56 is given in figure [ fig : lisa1993fig3 ] . good agreement between calculations and data is obtained for source parameter values of roughly @xmath57 fm and @xmath58 fm / c . these extracted emission time scales are qualitatively consistent with those predicted by microscopic transport calculations . the measurement discussed in the previous section provided the first unambiguous observation of long emission durations with femtoscopy . it thus validated the technique source lifetimes ( emission durations ) _ can _ be measured . for years , the problem had been that we were looking at longitudinal and transverse cuts in the wrong ( laboratory ) frame . dynamical models for symmetric systems with beam energies @xmath59 mev predict lifetimes @xmath60 fm / c , consistent with data , as we ve seen . but _ really _ long lifetimes are predicted at lower excitation energies , where a compound nucleus is briefly formed and cools by nucleon emission . a study of two - proton correlation functions in the inverse kinematics reaction xe+al at @xmath61 mev , reported @xcite no difference between longitudinal and transverse correlation functions , although a very long lifetime ( @xmath62 fm / c ) would be expected . with the newfound insight on the importance of analyzing the data in the `` right '' ( source ) frame , we decided to extract the raw data from storage and perform a re - analysis . the results are shown in figure [ fig : lisa1994fig3 ] . when we repeated the analysis of @xcite , we found no difference when cutting on @xmath44 , in agreement with the original published result . however , when we selected on the angle between @xmath34 and @xmath27 in the center - of - mass frame , a significant difference was observed @xcite . the curves in figure [ fig : lisa1994fig3 ] correspond to a spherical source , moving in the lab at @xmath63 ( the system center - of - momentum velocity ) , with radius and lifetime parameters @xmath64 fm / c and @xmath65 fm / c , respectively . figure [ fig : lisa1994fig2 ] quantifies the sensitivity of the parameter extraction , through contours of the chi - square per degree of freedom , analogous to that of figure [ fig : lisa1993fig3 ] . this result is virtually unknown in the relativistic heavy ion community , which is unfortunate 1300 fm / c ! this value , which is precisely what one expects for an evaporating compound nucleus at this energy , remains the longest timescale ever measured with multidimensional intensity interferometry in subatomic physics . comparison of measured ( points ) and calculated ( curves ) correlation functions . the calculations were performed for emission from a schematic source with radius and lifetime parameters @xmath64 fm and @xmath65 fm / c . from @xcite . , scaledwidth=70.0% ] contour plot of @xmath48 evaluated by comparing measured longitudinal and transverse correlation functions ( over the range @xmath66 to those predicted for emission from a schematic source with radius and lifetime parameters @xmath67 and @xmath56 . from @xcite . , scaledwidth=70.0% ] there have long been predictions @xcite that a first - order transition from a deconfined state ( quark - gluon plasma ) to a confined ( hadronic ) final state , may lead to an increase in the system emission time . the expectation @xcite is that this increase should occur just at the threshold energy for which a deconfined state is formed . at lower energies there is no transition at all , whereas at higher energies the system is exploding too quickly to form a `` burning log '' scenario . the threshold energy samples the `` softest point '' in the qcd equation of state . at the relativistic collision energies where this phenomenon might occur , studies have used multi - dimensional pion interferometry @xcite , where the relative momentum components ( and corresponding `` hbt radii '' ) are identified in the `` bertsch - pratt '' decomposition @xcite . referring to figure [ fig : lisa1993fig1 ] , @xmath68 measures the length scale of the pion cloud in the direction of the particle motion , and @xmath69 quantifies the length scale perpendicular to this motion . ( at relativistic energies , where the dynamics in the beam direction are substantially different from those in the transverse direction , @xmath68 and @xmath69 are forced to be perpendicular to the beam direction , and a third radius , @xmath70 quantifies the length scale along the beam . at the lower energies discussed in sections [ sec : arsc80 ] and [ sec : xeal31 ] , where compound nucleus formation occurs , this distinction is not made in the `` longitudinal - transverse '' decomposition . ) in the hypothetical case where the system is not flowing , these radii are related to the emission duration @xmath1 as @xmath71 where @xmath72 is the pion speed in the transverse direction . relativistic heavy ion collisions , however , are dominated by transverse flow , so equation [ eq : lifetimeapprox ] is only a crude approximation @xcite ; indeed , @xmath68 can be less than @xmath69 at high @xmath73 @xcite . a review of femtoscopic results in 2005 @xcite concluded that there was no evidence for the burning log signature in the two decades of relativistic heavy ion measurements at the ags , sps and rhic . since that review , another decade has passed , and many more measurements have been done . figures [ fig : allpiondata1 ] and [ fig : allpiondata2 ] contain the world dataset of pion hbt radii from collisions of the heaviest nuclei ( au+au in the u.s . and pb+pb in europe ) . two - pion femtoscopy has been measured in central heavy ion collisions over three orders of magnitude . above , hbt radii from @xmath74 gev collisions are plotted versus the transverse mass of the pair . figure [ fig : allpiondata2 ] shows analogous data up to @xmath75 gev.black datapoints originate from experiments at the ags ; red datapoints originate from experiments at rhic ; blue datapoints originate from experiments at the sps ; pink datapoints originate from experiments at the lhc.yellow panels identify measurements done after a 2005 review @xcite.,scaledwidth=90.0% ] two - pion femtoscopy has been measured in central heavy ion collisions over three orders of magnitude . above , hbt radii from @xmath74 gev collisions are plotted versus the transverse mass of the pair . figure [ fig : allpiondata2 ] shows analogous data up to @xmath75 gev.black datapoints originate from experiments at the ags ; red datapoints originate from experiments at rhic ; blue datapoints originate from experiments at the sps ; pink datapoints originate from experiments at the lhc.yellow panels identify measurements done after a 2005 review @xcite.,scaledwidth=90.0% ] two - pion femtoscopy has been measured in central heavy ion collisions over three orders of magnitude . above , hbt radii from @xmath76 gev collisions are plotted versus the transverse mass of the pair . figure [ fig : allpiondata1 ] shows analogous data down to @xmath77 gev.black datapoints originate from experiments at the ags ; red datapoints originate from experiments at rhic ; blue datapoints originate from experiments at the sps ; pink datapoints originate from experiments at the lhc.yellow panels identify measurements done after a 2005 review @xcite.,scaledwidth=90.0% ] two - pion femtoscopy has been measured in central heavy ion collisions over three orders of magnitude . above , hbt radii from @xmath76 gev collisions are plotted versus the transverse mass of the pair . figure [ fig : allpiondata1 ] shows analogous data down to @xmath77 gev.black datapoints originate from experiments at the ags ; red datapoints originate from experiments at rhic ; blue datapoints originate from experiments at the sps ; pink datapoints originate from experiments at the lhc.yellow panels identify measurements done after a 2005 review @xcite.,scaledwidth=90.0% ] datapoints in the yellow panels correspond to measurements that were performed in last decade . alice measurements @xcite at the lhc extend the measured energy range to three orders of magnitude . more important , however , are the measurements at rhic and the sps at energies _ below _ the maximum energy of the machine . these data were taken in `` energy scan '' programs , motivated by the increasing realization that some of the most important phenomena in hot qcd physics might only be revealed by a careful , systematic study of heavy ion collisions as the system conditions are gradually changed . the versatility of the rhic collider is clear from the fact that the rhic data ( red data points ) extend to low energies well below traditional sps energy of 17.3 gev . star has collected data in `` fixed - target mode , '' in which one low - energy rhic beam struck a gold foil placed toward the edge of the beam pipe at one end of the star detector . despite the fact that star is designed for midrapidity measurements at a 200-gev collider , the data taken were good , and hbt radii , fully in line with data at similar energies , have been measured . at this moment , these results are unavailable for release ; however , they are firm , and i could not resist a placeholder in figure [ fig : allpiondata1 ] indicating that rhic has now extended measurements into the ags energy range . the energy scan at rhic may have finally revealed the burning log signature , as shown in figure [ fig : phenixburninglog ] . a clear peak in @xmath78 ( or @xmath79 @xcite ) is observed around @xmath80 gev , an energy region where other intriguing phenomena have been reported @xcite . this figure includes only data from rhic and lhc collider experiments ; these have all been performed with similar techniques and acceptances . experiments at the cern sps have acceptances which vary with @xmath81 , making them not ideal for searching for subtle changes as collision energy changes ; femtoscopic results fluctuate significantly and disagree experiment - to - experiment . furthermore , sps measurements are performed with a variety of methods to handle the coulomb effect ; this can affect hbt radii significantly @xcite . rhic and lhc experiments all use the so - called bowler - sinyukov @xcite approach , explicitly including coulomb effects in the fits to the correlation functions . it is increasingly important that hydrodynamic theory address the rhic beam energy scan range in detail . while calculations at lhc energies are technically easier to perform ( due to approximate boost invariance , a simple equation of state , and low viscosity ) , the lower energies are more important . qcd has a scale , after all . just as solid state physicists study superconductivity around the transition point , heavy ion studies must focus on the energy region set by qcd physics . the difference ( or ratio ) of @xmath68 and @xmath69 is related to the emission duration of the collision . as discussed in the text , a generic expectation from a first - order phase transition is a rise and fall of this difference , with collision energy . data from the rhic beam energy scan appears to validate this prediction . figure from @xcite . , scaledwidth=90.0% ] thus far , i have discussed measurements of the emission duration @xmath1 of the hot system generated in a heavy ion collision . an estimate of the evolution timescale , @xmath0 , is also crucial for a detailed understanding of the system s dynamics . at the very low energies discussed in sections [ sec : arsc80 ] and [ sec : xeal31 ] , it is unclear how to distinguish the system evolution time from the emission duration . however , as sinyukov and collaborators pointed out @xcite , in ultrarelativistic collisions , the strong longitudinal flow generates a nearly boost - invariant system in which the longitudinal hbt radius and evolution time are related by @xcite @xmath82 where @xmath83 is the system temperature at freezeout , and @xmath84 is the transverse mass of the particles . figure [ fig : rlongmercedes ] shows fits of formula [ eq : sinyukov ] to longitudinal radii measured @xcite by the star collaboration for collisions at @xmath85 gev at various centralities . the fit is reasonable . evolution timescales extracted from star @xcite and alice @xcite are shown in figure [ fig : evolutiontimes ] . longitudinal hbt radii for @xmath85 gev au+au collisions of varying centrality , measured by the star collaboration . from @xcite . , scaledwidth=70.0% ] estimate of the system evolution time based on sinyukov fits ( equation [ eq : sinyukov ] ) to measured longitudinal hbt radii @xmath70 . , scaledwidth=50.0% ] in section [ sec : evolutiontimerlong ] , i outlined the `` traditional '' way to estimate the evolution time , based on the @xmath84 dependence of the longitudinal hbt radius . here , i provide an independent cross - check from another direction . in non - central heavy ion collisions , the hot system is initially anisotropic relative to the reaction plane ( spanned by the impact parameter vector and the beam direction ) of the collision . the response of the system to this coordinate - space anisotropy generates a corresponding momentum - space anisotropy , in which more ( and faster ) particles are emitted in the reaction plane than perpendicular to it . this is the well - known `` elliptic flow '' phenomenon , often quantified by a momentum - space anisotropy parameter @xmath86 @xcite . the preferentially in - plane expansion will tend to reduce ( or perhaps reverse ) the anisotropy of the initial state ; i.e. the system will become more round in coordinate space . if the system retains some anisotropy , the transverse hbt radii will oscillate as a function of azimuthal angle relative to the reaction plane @xcite . figure [ fig : starashbt ] shows femtoscopic radii measured @xcite by the star collaboration for mid - central au+au collisions at @xmath85 gev . fourier coefficients , the space - time analog to @xmath86 , may be extracted from the oscillations and used to estimate the spatial anisotropy of the source at freeze - out @xcite , defined as @xmath87 where @xmath88 ( @xmath89 ) is the root - mean - square extent of the source in ( out of ) the event plane . pion hbt radii from mid - central au+au collisions at @xmath85 gev , as a function of azimuthal angle relative to the event plane . from @xcite . , scaledwidth=60.0% ] spatial anisotropies have been extracted for heavy ion collisions over the entire available energy range and are plotted in figure [ fig : allashbt ] . for all energies , the system retains its out - of - plane extension in coordinate space , though it is reduced from its initial value of 0.25 ( estimated from glauber calculations ) . the spatial anisotropy , defined in equation [ eq : epsilon ] , for mid - central au+au ( pb+pb ) collisions from e895/ags @xcite , ceres / sps @xcite , star / rhic @xcite , phenix / rhic @xcite and alice / lhc @xcite . calculations @xcite with hydrodynamic and transport models are shown for comparison . , scaledwidth=100.0% ] the evolution of the hot system produced in a heavy ion collision is complex , but toy models can be useful to check whether disparate measurements may be understood in a single simple scenario . in this spirit , i construct a blast - wave @xcite inspired model and ask whether the evolution times plotted in figure [ fig : evolutiontimes ] are reasonably consistent with the reduction in coordinate - space anisotropy seen in figure [ fig : allashbt ] . let @xmath90 be the angle - averaged rms size of the source at @xmath91 , and @xmath92 be its anisotropy . further , let @xmath93 and @xmath94 be the average flow velocities in and out of the reaction plane . assuming constant anisotropic free - streaming evolution of the source , its spatial anisotropy after evolving for @xmath0 is @xmath95 based on glauber calculations , @xmath96 fm and @xmath97 . the evolution time @xmath0 for the 10 - 30% central collisions were extracted from @xmath70 for that centrality . average transverse flow velocities are related to blast - wave @xcite flow parameters according to @xmath98 using blast - wave fit parameters extracted by the phenix collaboration @xcite , @xmath99 and @xmath100 . while one expects higher flow velocities at higher energies , the same values for @xmath93 and @xmath94 were used for all @xmath81 , since other blast - wave fits were not readily unavailable . however , it turns out that these velocities vary little with @xmath81 , so these values should serve for a test . the estimate from this toy model is compared to data in figure [ fig : toymodel ] . considering its crudity and not tinkering with parameters , the agreement is remarkable . both the magnitude and the @xmath81-dependence of @xmath101 seem to be consistent with an evolution timescale extracted in the `` traditional '' way , using @xmath70 , as discussed in section [ sec : evolutiontimerlong ] . measurements of the final freeze - out eccentricity ( from figure [ fig : allashbt ] are compared with calculations of a toy model based on an initially out - of - plane extended source evolving with preferential in - plane expansion . the grey band indicates the initial anisotropy based on glauber calculations , and the downward - facing arrows indicate the evolution of the shape . the terminus of the arrow corresponds to the shape size at the time @xmath0 extracted by sinyukov fits ( equation [ eq : sinyukov ] to the longitudinal radii . see equation [ eq : toymodel ] and text for details . , scaledwidth=100.0% ] to understand the dynamics of a heavy ion collision , it is important to have an estimate of the timescales associated with its evolution . i have discussed experimental measurements of the emission and evolution timescales based on particle intensity interferometry measurements . when the analysis was performed in the correct reference frame , two - proton correlation functions at low collision energies revealed long lifetimes , consistent with theoretical expectations . these observations were important , as they put to rest troubling doubts about our understanding of intensity interferometry overall . in ultra - relativistic heavy ion collisions , the long - sought `` burning log '' signature of a softening of the equation of state was found , but only after a systematic scan of the collision energy . this is one of several interesting signatures at energies around @xmath102 gev that have been revealed in the rhic beam energy scan program . the evolution time of a collision is found to grow with collision energy , and the traditional estimate based on the longitudinal hbt radius was found to be consistent with a toy model describing the evolution of the spatially anisotropic source as estimated by azimuthal oscillations of the transverse hbt radii . these timescale estimates serve as important input to theoretical studies of the dynamics of the collision . such studies are crucial , if the field is to generate lasting physics contributions to our understanding of qcd . while dynamic modeling of the highest - energy collisions ( e.g. at the lhc ) are much easier , it is much more important to focus on lower energies around @xmath103 gev , where nontrivial phenomena associated with the qcd equation of state may appear .

the study of high energy collisions between heavy nuclei is a field unto itself , distinct from nuclear and particle physics . a defining aspect of heavy ion physics is the importance of a bulk , self - interacting system with a rich space - time substructure . i focus on the issue of timescales in heavy ion collisions , starting with proof from low - energy collisions that femtoscopy can , indeed , measure very long timescales . i then discuss the relativistic case , where detailed measurements over three orders of magnitude in energy reveal a timescale increase that might be due to a first - order phase transition . i discuss also consistency in evolution timescales as determined from traditional longitudinal sizes and a novel analysis using shape information . the slowly crawling ants will eat our dreams . + _ - andrzej biaas , musing on words of andre breton _ + go to the ant , thou sluggard ; consider her ways , and be wise . + _ - proverbs vi.6 _

preface introduction can femtoscopy measure long emission durations? evidence for a burning log the evolution time of the system summary

random geometric network models @xcite comprise a collection of entities called nodes embedded in region of typically two or three dimensions , together with connecting links between pairs of nodes that exist with a probability related to the node locations . they appear in numerous complex systems including in nanoscience @xcite , epidemiology @xcite , forest fires @xcite , social networks @xcite , and wireless communications @xcite . such networks exhibit a general phenomenon called _ percolation _ @xcite , where at a critical connection probability ( controlled by the node density ) , the largest connected component ( cluster ) of the network jumps abruptly from being independent of system size ( microscopic ) to being proportional to system size ( macroscopic ) . percolation phenomena are closely related to thermodynamic phase transitions where the number of nodes @xmath0 goes to infinity and the critical percolation density @xmath1 is largely independent of the system size , shape , and of the microscopic details of the model ; the phenomenon of universality . at the critical point , conformal invariance in two dimensional networks leads to detailed expressions for the probability of a connection across general regions @xcite and more general connections with conformal field theory @xcite and schramm - loewner evolution @xcite . here , we take a different approach and are concerned with finite networks and with questions related to percolation , but fundamentally different : what node density ensures a specified probability @xmath2 that the entire network is a single connected component ( cluster ) , that is , _ fully connected _ ? how is this probability affected by the shape of the network domain ? these questions are crucial for many applications , including for example the design of reliable wireless mesh networks . these consist of communication devices ( the nodes ) that pass messages to each other via other nodes rather than a central router . this allows the network to operate seamlessly over a large area , even when nodes are moved or deactivated . a fully connected network means that every node can communicate with every other node through direct or indirect connections . mesh networks have been developed for many communication systems , including laptops , power distribution ( `` smart grid '' ) technologies , vehicles for road safety or environmental monitoring , and robots in hazardous locations such as factories , mines and disaster areas @xcite . for many applications of random geometric networks including those above , _ direct connection _ between two nodes @xmath3 and @xmath4 can be well described by a probability @xmath5 , a given function of the distance between the nodes @xmath6 . often , the nodes are mobile or otherwise not located in advance , hence we assume @xmath0 uniformly distributed nodes confined in a specified @xmath7-dimensional region @xmath8 with area ( @xmath9 ) or volume ( @xmath10 ) denoted by @xmath11 . the node density is then defined as @xmath12 . for reference , we will later take @xmath13 $ ] , where @xmath14 is a relevant length scale , and @xmath15 determines the sharpness of the cut - off . note that when @xmath16 a step function corresponding to the popular _ unit disk _ deterministic model @xcite is obtained , where connections have a fixed range @xmath17 . our derivation however is completely general and only requires that @xmath18 is sufficiently short - ranged compared to system size . using this as a basis , we find that contrary to common belief and practice , the geometrical details of the confined space boundaries ( _ corners , edges _ and _ faces _ ) dominate the properties of the percolation transition . moreover , the short - range nature of @xmath18 allows us to separate individual boundary components and obtain analytic expressions for @xmath2 at high densities as a sum over their contributions . we confirm this through computer simulations and argue that the substantial improvement offered by our main result eq . [ e : bdy ] can be used to predict , control , optimize or even set benchmarks for achieving full network connectivity in a wide variety of suitable models and applications involving finite size geometries . isolated nodes shown as black balls concentrate at the boundaries of the domain and particularly near corners at higher densities . nodes are placed randomly in a cube , with lighter colors indicating a higher probability of being in the largest connected component . we use @xmath19 , while the side length of the cube is @xmath20 . there are 500 nodes in ( a ) and 700 nodes in ( b ) . ] as in conventional continuum percolation theory @xcite , we start by utilizing a cluster expansion approach @xcite to derive a systematic perturbative method for determining the full connection probability @xmath2 as a function of density @xmath21 . formulation of the expansion can be summarized as follows . the probability of two nodes being connected ( or not ) leads to the trivial identity @xmath22 . multiplying over all links expresses the probabilities @xmath23 of all @xmath24 possible graphs @xmath25 , @xmath26\equiv \sum_{g}\mathcal{h}_{g},\ ] ] collecting terms according to largest cluster size we get @xmath27 where @xmath28 is the set of all possible graphs with largest cluster of size @xmath29 . the first term on the right hand side is the probability of being fully connected given a specific configuration of nodes . the average over all random configurations @xmath30 of this quantity is thus the overall probability of obtaining a fully connected network @xmath2 . moreover , the main idea conveyed by eq . ( [ 2 ] ) is that at high densities , full connectivity is most likely to be broken by a single isolated node ( the @xmath31 term ) ; this is sufficient detail for most applications . further corrections incorporate the probability of several isolated single nodes and smaller clusters of nodes , for which a systematic expansion can be developed @xcite . averaging eq . ( [ 2 ] ) over all configurations and noting that to leading order the @xmath32 cluster is fully connected , and that all nodes are identical , the first order approximation becomes @xmath33 where the `` connectivity mass '' accessible from a node placed at @xmath34 is given by @xmath35 assuming that the volume @xmath36 for any @xmath34 , which is reasonable if the system is significantly larger than @xmath14 at moderate densities and that the number of nodes @xmath0 is large , eq.[first ] simplifies to @xmath37 this equation is equivalent to eq . ( 8) in mao and anderson @xcite which was derived for the specific case of a square domain . following numerous studies by probabilists and engineers @xcite , these authors however assumed an exponential scaling of system size @xmath11 with @xmath21 which essentially renders boundary effects negligible . scaling the system in such a way is a common approach as it corresponds to the limit of infinite density at fixed connection probability , however in practice this limit is approached only for unphysically large volumes . in contrast , we do not assume exponential growth of @xmath11 , and also consider far more general geometries in any dimension @xmath38 . without an exponentially growing volume @xmath11 , the behavior of the full connection probability at high densities is qualitatively different : it is controlled by the exponential in eq . ( [ e : ma ] ) , and hence node positions @xmath34 where the connectivity mass is small , that is , near the boundary of @xmath8 . thus in contrast to the usual situation in statistical mechanics , the boundaries ( and in particular corners ) are important , and we will see they in fact dominate . we illustrate this in fig . [ f : balls ] where nodes are placed randomly inside a cube and an average over a large number of possible graphs gives the connectivity of each node . notice that isolated and hard - to - connect nodes shown as dark balls concentrate near the boundaries of the domain and particularly near corners at higher densities . this observation forms the basis of our work , and has led to a radically different understanding of connectivity in confined geometries which we now detail further . the contributions to the integrals in eq . ( [ e : ma ] ) come from @xmath34 at _ boundary components _ @xmath39 of dimension @xmath40 , for example the bulk , the faces , and right angled edges and corners of a cube , with @xmath41 and @xmath42 respectively . the short - range nature of @xmath18 allows us to isolate each boundary component , whilst to leading order the connectivity mass splits into independent radial and angular integrals , depending only on the local geometry of @xmath43 and hence @xmath44 where @xmath45 is the angle ( @xmath9 ) or solid angle ( @xmath10 ) subtended by @xmath43 . for example , if @xmath46 is near a corner of the cube then @xmath47 , while near an edge @xmath48 , near faces @xmath49 and @xmath50 for the bulk interior . hence , from eq . ( [ e : ma ] ) we see that corner contributions to @xmath2 as a function of @xmath21 are exponentially larger than edge contributions which are themselves exponentially larger than face contributions etc . this simple argument shows that the dominant contribution to @xmath2 at high densities comes from the `` pointiest '' corners . expanding @xmath51 about @xmath52 near the corresponding boundary component we obtain a next to leading order expansion for @xmath53 which we can then use to approximately evaluate the integral in eq . ( [ e : ma ] ) . ignoring exponentially smaller correction terms and combining all boundary contributions we arrive at our main result @xmath54 where @xmath55 is the @xmath40-dimensional `` volume '' of each component ( equal to one in the case of a @xmath42-dimensional corner and @xmath11 when @xmath56 ) , @xmath57 is a geometrical factor depending on @xmath43 and implicitly on @xmath58 and @xmath59 is as in ( [ e : sgf ] ) ; see examples below . notice that eq . ( [ e : bdy ] ) is completely general as we have only assumed a short - ranged @xmath18 and not used its specific form . moreover , it also does not depend on using euclidean distance and holds in any dimension @xmath60 and geometry where the lack of connectivity is dominated by a situation involving an @xmath32 cluster and a single disconnected node . hence eq . ( [ e : bdy ] ) is a powerful and useful multi - purpose tool for analyzing full network connectivity at high densities in a wide variety of suitable models and applications involving finite size geometries . for example , in the context of single input single output ( siso ) wireless communication channels and a rayleigh fading model @xcite , information theory predicts @xmath61 $ ] with @xmath15 an environment and wavelength dependent decay parameter equal to @xmath62 for free propagation , increasing to @xmath63 for a cluttered environment , while @xmath14 depends on the minimum outage rate threshold . for nodes confined to a cube of side length @xmath64 and @xmath19 we find @xmath65 , @xmath66 , and @xmath67 with contributions from each of the eight corners , twelve edges , six faces and bulk . however the derivation is general : once @xmath57 and @xmath59 have been evaluated for these boundary components ( right angled edges etc . ) by standard asymptotic analysis of the relevant integrals , they apply to any geometry with these features and length scales significantly larger than @xmath17 . this independence on the large scale geometry follows from the short - range nature of @xmath18 and is a type of universality allowing for the calculation of @xmath2 in complex high dimensional geometries without increased difficulty . ( a ) comparison of the full analytic prediction of eq . ( [ e : bdy ] ) ( solid curve ) with direct numerical simulation of the random network in a cube of side length @xmath68 ( jagged curve ) . the dashed line corresponds to the bulk contribution ( previous theory ) . ( b ) contributions from the bulk ( dotted blue , left ) , faces ( red ) , edges ( yellow ) and corners ( green , right ) , together with the total ( solid blue ) and numerical simulation ( black jagged curve ) , showing the dominance of the corners at the highest densities and good agreement between theory and simulation at moderate to high densities . here it is convenient to plot the outage probability @xmath69 . ] the substantial improvement offered by eq . ( [ e : bdy ] ) becomes clear when compared with the `` bulk '' contribution corresponding to current conventional wisdom shown in fig . [ f : cube]a ) for a network confined to a cube . [ f : cube]b ) further demonstrates the inaccuracy of the bulk model as well as the benefits of including boundary effects when analyzing network connectivity in confined geometries . corner contributions in triangles with equal area and perimeter : comparison of theory with direct simulation , as in fig . [ f : cube ] . the red triangle has side lengths of 26.88 , 15.44 and 15.44 in units of the connectivity length scale @xmath17 , while the blue triangle has side lengths of 8.40 , 24.68 and 24.68 . the black dashed lines correspond to the equal bulk ( left curve ) and bulk@xmath70edge ( right curve ) contributions while neglecting corner contributions . the colored curves give the total ( including crucial corner ) contributions for each triangle . both theory and simulation are plotted , showing excellent agreement with the numerical simulations ( jagged curves ) which cover them completely for @xmath71 . ] we can go beyond simple geometries restricted to right - angled corners . consider the case of a two dimensional triangle with general angles @xmath72 . the relevant integrals for this case come to @xmath73 , with @xmath74 for the corners and @xmath75 for the edges and bulk and can be generalized easily to higher dimensions . [ f : tri ] shows two triangles chosen to have identical perimeter and area ; the connectivity at a given density differs only due to the corner angles and agrees perfectly with the full theory of eq . ( [ e : bdy ] ) . a bulk theory , even supplemented with edge contributions , is clearly incapable of explaining the difference between the connectivities of networks in these two triangles . moreover , such a situation motivates inverse problems , similar to hearing the shape of a drum " @xcite by attempting to determine the size and shape details of an unknown domain containing a random network . an important aspect of the theory presented here is how it affects the design of real life random geometric networks . for wireless mesh networks , the lack of connectivity near the boundaries can be mitigated by increasing the signal power , the number of spatial channels , or by constructing a hybrid network with a regular array of fixed nodes along the boundaries as well as randomly placed nodes in the interior . in each of these cases , the design can now be analyzed given information about the cost and connectivity function @xmath76 and of course the desired connectivity region . conversely , boundary effects can be harnessed to avoid full connectivity if desired . for example in the case of forest fires @xcite we have a prediction for the number of unburnt regions as a function of the geometric landscape and environment parameters ( for example angles between fire - lanes and/or natural boundaries ) , again given a specific model for connectivity that depends on the type of vegetation , temperature , moisture content etc . similar models could be devised for the spread of epidemics @xcite or mobile phone viruses @xcite where boundaries are embedded in a more complex ( possibly non - euclidean ) space yet @xmath18 is still short - ranged . we examined connectivity in confined geometries and illustrated the importance of the often neglected boundary effects . we then derived a general high density expansion eq . ( [ e : bdy ] ) for the probability of full connectivity @xmath2 assuming only a short - ranged connectivity function relative to system size and showed that it displays universal features allowing for its easy calculation in complex geometries . this we have confirmed through computer simulations and argued that our approach is well placed to facilitate efficiency in design in a variety of physical applications ranging from wireless networks to forest fire - lanes . appropriate modifications of our theory can aid the understanding of small boundary - dominated systems such as for example the electrical conduction through carbon nanotubes in a polymer matrix @xcite but possibly larger systems such as highly connected social and financial networks @xcite . the authors thank the directors of the toshiba telecommunications research laboratory for their support , and charo del genio , jon keating and mark walters for helpful discussions . 999 m. penrose , _ random geometric graphs _ , ( oxford university press , 2003 ) . m. franceschetti and r. meester , _ random networks for communication , _ ( cambridge university press , 2007 ) . a. v. kyrylyuk , m. c. hermant , t. schilling , b. klumperman , c. e. koning and p. van der schoot , nature nanotech . * 6 * , 364 ( 2011 ) . j. c. miller , j . soc . interf . * 6 * , 1121 ( 2009 ) . l. danon , a. p. ford , t. house , c. p. jewell , m. j. keeling , g. o. roberts , j. v. ross , and m. c. vernon , interdisc . diseases * 2011 * , 284909 ( 2011 ) . s. pueyo , , p. m. l. d. a. graa . r. i. barbosa , r. cots , e. cardona , and p. m. fearnside , ecol . lett . * 13 * , 793 ( 2010 ) . g. palla , a .- barabsi , and t. vicsek , nature * 446 * , 664 ( 2007 ) . r. parshani , s. buldyrev , and s. havlin , proc . sci . * 108 * , 1007 ( 2011 ) . m. haenggi , j. g. andrews , f. baccelli , o. dousse o. , and m. franceschetti , ieee j. select . area . * 27 * , 1029 ( 2009 ) . j. li , l. andrew , c. foh , m. zukerman , and h. chen , sensors , * 9 * , 7664 ( 2009 ) . p. wang , m. c. gonzlez , c. a. hidalgo , and a. l. barabsi , science * 324 * , 1071 ( 2009 ) . d. s. callaway , m. e. j. newman , s. h. strogatz , and d. j. watts , phys . * 85 * , 5468 ( 2000 ) . b. bollobs and o. riordan , _ percolation , _ ( cambridge university press , 2006 ) . j. l. cardy , j. phys . a : math . gen . , * 25 * , l201 ( 1992 ) . j. fuchs , i. runkel , and c. schweigert , j. math . phys . * 51 * , 015210 ( 2010 ) . , _ in preparation _ , ( 2011 ) . g. mao , and b. anderson , infocom , 2011 proc . ieee , 631 ( 2011 ) . d. tse , and p. viswanath , _ fundamentals of wireless communication _ ( cambridge university press , 2005 ) . , eur.j.phys , * 73 * , 1 , ( 1966 ) .

many complex networks exhibit a percolation transition involving a macroscopic connected component , with universal features largely independent of the microscopic model and the macroscopic domain geometry . in contrast , we show that the transition to full connectivity is strongly influenced by details of the boundary , but observe an alternative form of universality . our approach correctly distinguishes connectivity properties of networks in domains with equal bulk contributions . it also facilitates system design to promote or avoid full connectivity for diverse geometries in arbitrary dimension .

introduction full connection probability boundary effects discussion acknowledgments

in many physics analyses , one has non - interfering background events which can not be cleanly separated from the desired signal . this is of particular concern when the problem being studied is multi - dimensional , _ i.e. _ there are kinematic correlations in the signal which must be preserved . one typical way to handle these situations is with the _ side - band subtraction _ method . distributions constructed using events outside the signal region are subtracted from those using events inside the signal region to create distributions in which the backgrounds have been removed . while this method can be effective in some situations , implementing it can become problematic if the kinematics of the background region are different than those of the signal or if the problem is sufficiently multi - dimensional such that binning the data is severely limited by statistical uncertainties . in this paper , we describe a procedure for assigning each event in a data sample a quality factor ( @xmath0-factor ) which gives the chance that it originates from the signal sample . this @xmath0-factor can then be used as an event weight in subsequent analysis procedures to gain access to the signal distribution . for example , the @xmath0-factors can be used to perform event - based unbinned maximum likelihood fits on the data to extract physical observables . eliminating the need to bin the data is highly desirable for the case of multi - dimensional problems . we note here that this procedure is not appropriate for situations where there are quantum mechanical interference effects between the signal and background . consider a data set composed of @xmath1 total events , each of which is described by @xmath2 coordinates @xmath3 ( @xmath4 ) . the coordinates can be masses , angles , energies , _ etc_. furthermore , the data set consists of @xmath5 signal and @xmath6 background events . both the signal and background distributions are functions of the coordinates , @xmath7 and @xmath8 respectively . for this procedure , we need to know _ a priori _ the functional dependence ( possibly with unknown parameters ) of the signal and background distributions in terms of one of the coordinates . we will refer to this coordinate as the reference coordinate and label it @xmath9 . as an example , consider the case where the reference coordinate is a mass . the functional dependence of the signal , in terms of @xmath9 , might be given by a gaussian or breit - wigner distribution . the background may be well represented by a polynomial . in both cases , there could be unknown parameters ( _ e.g. _ the width of the gaussian ) ; these are permitted when using this procedure . no other _ a priori _ information is required concerning the dependence of @xmath7 or @xmath8 on any of the other coordinates . the aim of this procedure is to assign each event a quality factor , or @xmath0-factor , which gives the chance that it originates from the signal sample . we first need to define a metric for the space spanned by @xmath3 ( excluding @xmath9 ) . a reasonable choice is to use @xmath10 where @xmath11 is the maximum possible difference between the coordinates @xmath12 of any two events in the sample . using this metric , the distance between any two events , @xmath13 , is given as @xmath14 ^ 2,\ ] ] where the sum is over all coordinates except @xmath9 . for each event , we compute the distance to all other events in the data set , and retain the @xmath15 _ nearest neighbor _ events , including the events itself , according to ( [ eq : dist ] ) . the value of @xmath15 , which varies depending on the analysis , is discussed below . the @xmath15 events are then fit using the unbinned maximum likelihood method to obtain estimators for the parameters , @xmath16 , in the probability distribution function @xmath17d\xi_r},\ ] ] where @xmath18 and @xmath19 describe the functional dependence on the reference coordinate , @xmath9 , of the signal and background respectively . the @xmath0-factor for each event is then calculated as @xmath20 where @xmath21 is the value of the event s reference coordinate and @xmath22 are the estimators for the parameters obtained from the event s fit . if one wants to bin the data , the signal yield in a bin is obtained as @xmath23 where @xmath24 is the number of events in the bin . for example , to construct a histogram ( of any dimension ) of the signal , one would simply weight each event s contribution by its @xmath0-factor . it is also important to extract the uncertainties on the individual @xmath0-factors so that we can obtain error estimates on measurable quantities . the full covariance matrix obtained from each event s fit , @xmath25 , can be used to calculate the uncertainty in @xmath0 as @xmath26 when using these values to obtain errors on the signal yield in any bin , we must consider the fact that the nature of our procedure leads to highly - correlated results for each event and its @xmath15 nearest neighbors ; thus , simply adding the @xmath27 values in quadrature would certainly underestimate the true error . the actual degree of correlation of the @xmath0-factors would depend on the population of the bins . a safe choice is to assume 100% correlation ; thus , @xmath28 which provides an overestimate of the true uncertainty inherent in the procedure . in addition to the uncertainties associated with the fits , there will also be a purely statistical error associated with the signal yield in each bin , given by poisson statistics . for large values of @xmath29 , the signal yield obtained using ( [ eq : sig - yield ] ) , this can be taken to be @xmath30 ; however , for smaller @xmath29 the upper limit is better approximated by @xmath31 @xcite . the total uncertainty on the signal yield in any bin is then obtained by adding the fit errors , calculated using ( [ eq : q - err ] ) , in quadrature with the statistical errors discussed above . as an example , we will consider the reaction @xmath32 in a single @xmath33 bin , _ i.e. _ a single center - of - mass energy and production angle bin ( extending the example to avoid binning in production angle , or @xmath34 , is discussed below ) . the @xmath35 decays to @xmath36 about 90% of the time ; thus , we will assume we have a detector which has reconstructed @xmath37 events . of course , there are production mechanisms other than @xmath32 which can produce this final state and there is no cut which can be performed to separate out events that originated from @xmath32 . below we will construct a toy - model of this situation by generating monte carlo events for both signal , _ i.e. _ @xmath35 events , and background , _ i.e. _ non-@xmath35 @xmath36 events ( 10,000 events were generated for each ) . the goal of our model analysis is to extract the @xmath35 polarization observables known as the _ spin density matrix elements _ , denoted by @xmath38 ( discussed below ) . in terms of the mass of the @xmath36 system , @xmath39 , the @xmath35 events were generated according to 3-body phase space weighted by a voigtian ( a convolution of a breit - wigner and a gaussian , see ( [ eq : voigt ] ) ) to account for both the natural width of the @xmath35 and detector resolution . for this example , we chose to use @xmath40 mev / c@xmath41 for the detector resolution ( see figure [ fig : m3pi - gen ] ) . the goal of our analysis is to extract the three measurable elements of the spin density matrix ( for the case where neither the beam nor target are polarized ) traditionally chosen to be @xmath42 , @xmath43 and @xmath44 . these can be accessed by examining the distribution of the decay products ( @xmath36 ) of the @xmath35 in its rest frame . mass of the @xmath36 system ( gev / c@xmath41 ) for all generated events ( unshaded ) and for only the background ( shaded ) . , scaledwidth=45.0% ] for this example , we chose to work in the helicity system which defines the @xmath45 axis as the direction of the @xmath35 in the overall center - of - mass frame , the @xmath46 axis as the normal to the production plane and the @xmath47 axis is simply given by @xmath48 . the decay angles @xmath49 are the polar and azimuthal angles of the normal to the decay plane in the @xmath35 rest frame , _ i.e. _ the angles of the vector @xmath50 . the decay angular distribution of the @xmath35 in its rest frame is then given by @xcite @xmath51 which follows directly from the fact that the @xmath35 is a vector particle ; it has spin - parity @xmath52 . we chose to use the following @xmath38 values for this example : [ eq : rho - gen ] @xmath53 @xmath54 @xmath55 the resulting generated decay distribution is shown in figure [ fig : decay - angles - gen ] . for the background , we chose to generate it according to 3-body phase space weighted by a linear function in @xmath39 and @xmath56 in the decay angles . figure [ fig : m3pi - gen ] shows the @xmath36 mass spectrum for all generated events and for just the background . the generated decay angular distributions for all events , along with only the signal and background are shown in figure [ fig : decay - angles - gen ] . there is clearly no way to separate out the signal events through the use of a cut . to obtain the @xmath0-factors , we first need to identify the relevant coordinates , _ i.e. _ the kinematic variables in which we need to separate signal from background . the @xmath36 mass will be used as the reference coordinate , @xmath57 . the stated goal of our analysis is to extract the @xmath38 elements . we will do this using ( [ eq : schil ] ) ; thus , only the angles @xmath49 are relevant . other decay variables , such as the distance from the edge of the @xmath36 dalitz plot , are not relevant to this analysis though , they would be in other analyses ( see section [ section : example : full - pwa ] ) . using the notation of section [ section : method ] , @xmath58 and the distance between any two points , @xmath13 , is given by @xmath59 the functional dependence of the signal and background on the reference coordinate , @xmath39 , are @xmath60 @xmath61 where @xmath62 gev / c@xmath41 , @xmath63 mev , @xmath64 mev is the simulated detector resolution , @xmath65 are unknown parameters and @xmath66,\end{aligned}\ ] ] is the convolution of a gaussian and non - relativistic breit - wigner known as a voigtian ( @xmath67 is the complex error function ) . as stated above , the number of nearest neighbor events required depends on the analysis . specifically , it depends on how many unknown parameters there are , along with the functional forms of @xmath18 and @xmath19 . for this relatively simple case , the value @xmath68 works well . for each simulated event , we then find the @xmath15 closest events ( containing both signal and background ) and perform an unbinned maximum likelihood fit , using the cernlib package minuit @xcite , to determine the estimators @xmath69 . the @xmath0-factors are then calculated from ( [ eq : q - factor ] ) and the uncertainties are straightforward to calculate using section [ section : errors ] . ( color online ) mass of the @xmath36 system ( gev / c@xmath41 ) for all generated events ( unshaded ) , only generated background events ( shaded ) and all generated events weighted by @xmath70 ( dashed - red ) . , scaledwidth=45.0% ] figure [ fig : m3pi - fit ] shows the comparison of the extracted and generated background @xmath39 distributions integrated over all decay angles . the agreement is quite good ; however , we are looking for more than just global agreement . figure [ fig : decay - angles - fit ] shows the extracted angular distributions for the signal and background . the agreement with the generated distributions is excellent ( see figure [ fig : decay - angles - gen ] ) . we can also compare the @xmath0-factors extracted by the fits to the theoretical distributions from which our data was generated . figure [ fig : q - compare ] shows that the extracted values are in very good agreement with the generated values . we conclude this section by discussing the importance of quality control in the fits . for this example , we performed 20,000 independent fits to extract the @xmath0-factors . to avoid problems which can arise due to fits not converging or finding local minima , each unbinned maximum likelihood fit was run with three different sets of starting values for the parameters @xmath16 : ( 1 ) 100% signal ; ( 2 ) 100% background ; ( 3 ) 50% signal , 50% background . in all cases , the fit with the best likelihood was used . the @xmath15 events were then binned and a @xmath71 was obtained . in about 2% of the fits the @xmath71 was very large , a clear indicator that the fit had not found the best estimators @xmath69 . for these events , a binned @xmath72 fit was run to obtain the @xmath0-factor . as discussed in section [ section : errors ] , the covariance matrix obtained from each event s fit can be used to obtain the uncertainty in @xmath0 , @xmath73 , using ( [ eq : sigma - q ] ) . the nature of our procedure leads to a high degree of correlation between neighboring event s @xmath0-factors . this means that adding the uncertainties in quadrature would definitely underestimate the true error . in section [ section : errors ] , we argued that a better approach was to assume 100% correlation which provides an overestimate of the true error . to examine the error bars in our toy example , we chose to project our data into a one - dimensional distribution in @xmath74 . this was done to avoid bin occupancy issues which arise in the two - dimensional case due to limitations in statistics . figure [ fig : compare - q - err](a ) shows the comparison between the generated and calculated @xmath74 distributions . the agreement is excellent . the error bars on the calculated points were obtained using ( [ eq : q - err ] ) . for this study , we ignore the poisson statistical uncertainty in the yield due to the fact that the number of generated events is known . in a real world analysis , these should be included in the quoted error bars . we can examine the quality of the error estimation by examining the difference between the generated and calculated yields in each bin , @xmath75 . figure [ fig : compare - q - err](b ) shows the comparison between @xmath75 , @xmath76 obtained assuming 100% correlation and @xmath76 obtained assuming no correlation , _ i.e. _ adding the individual uncertainties in quadrature . as expected , the correlated errors provide an overestimate of @xmath75 in every bin , while the uncorrelated errors greatly underestimate @xmath75 in the majority of bins . it is not surprising that the errors obtained using ( [ eq : q - err ] ) are better estimates of the true uncertainty when the bin occupancy is closer to @xmath15 . we could come up with a scheme to scale the size of the error bars according to bin occupancy ; however , recall that in this example we fit the background with the functional form to which it was generated . this typically will not be the case for real data . in a real analysis , fitting with different functional forms and examining the extracted signal distributions to estimate systematic effects due to the choice of background is generally required . in this case , one should simply consider that the uncorrelated error bars are overestimates for each choice of background function when estimating the systematic uncertainty in the signal yields . the goal of our model analysis is to extract the spin density matrix elements . when studying multi - dimensional problems , binning the data is often undesirable due to limitations in statistics . the @xmath0-factors obtained for each event above can be used in conjunction with the unbinned maximum likelihood method to avoid this difficulty . if we could cleanly separate out the signal events from the background , then the likelihood function would be defined as @xmath77 where @xmath78 is the decay angular distribution defined in ( [ eq : schil ] ) . we could then obtain estimators for the spin density matrix elements by minimizing @xmath79 in this example , it is not possible to separate the signal and background event samples ; however , we can use the @xmath0-factors to achieve the same effect by rewriting ( [ eq : log - l ] ) as @xmath80 where the sum is now over all events ( signal and background ) . thus , the @xmath0-factors are used to weight each event s contribution to the likelihood . using the @xmath0-factors obtained in section [ section : example : apply ] , minimizing ( [ eq : log - l - q ] ) yields @xmath81 @xmath82 @xmath83 where the uncertainties are purely statistical ( obtained from the fit covariance matrix ) . thus , the values extracted for the spin density matrix elements are in excellent agreement with the values used to generate the data given in ( [ eq : rho - gen ] ) . to extend this example to allow for the case where the data is not binned in production angle , we would simply need to include @xmath84 or @xmath34 in the vector of relevant coordinates , @xmath3 . to perform a full partial wave analysis on the data , we would also need to include any additional kinematic variables which factor into the partial wave amplitudes , _ e.g. _ the distance from the edge of the @xmath36 dalitz plot ( typically included in the @xmath35 decay amplitude ) . we would then construct the likelihood from the partial waves and minimize @xmath85 using the @xmath0-factors obtained by applying our procedure including the additional coordinates . an example of this can be found in @xcite . in this paper , we have presented a procedure for separating signals from non - interfering backgrounds by determining , on an event - by - event basis , a quality factor ( @xmath0-factor ) that a given event originated from the signal distribution . we have shown that this @xmath0-factor can be used as an event weight in subsequent analysis procedures to allow more direct access to the true signal distribution . though this procedure may be computationally expensive , in principle it only needs to be performed once for each event in the data sample . this work was supported by grants from the united states department of energy no . de - fg02 - 87er40315 and the national science foundation no . 0653316 through the `` physics at the information frontier '' program . 99 v. h. regener , phys . rev . * 84 * , 161 ( 1951 ) . k. schilling , p. seyboth and g. wolf , nucl . b15 * , 397 ( 1970 ) . f. james , cern program library d506 ( 1998 ) . m. williams , carnegie mellon university ph.d . thesis , ( 2007 ) .

a common situation in experimental physics is to have a signal which can not be separated from a non - interfering background through the use of any cut . in this paper , we describe a procedure for determining , on an event - by - event basis , a quality factor ( @xmath0-factor ) that a given event originated from the signal distribution . this @xmath0-factor can then be used as an event weight in subsequent analysis procedures , allowing one to more directly access the true spectrum of the signal . , and 29.85.-c , 29.85.fj

[section:intro]introduction [section:method]quality factor determination [section:errors]error estimation [section:example]example application [section:conclusions]conclusions

the domino effect has received considerable attention because it is spectacular and seems to be governed by simple physics @xcite . however , careful experiments @xcite are not so easy . measuring the propagation speed of the order of a meter per second of a train of falling dominoes , is already too difficult to observe by elementary means . equally , the theory of toppling dominoes is not as simple as the phenomenon looks like and can not be reduced to the straightforward application of the conservation laws of energy and angular momentum . the discussion of a row of falling dominoes involves two aspects : the tumbling motion and the collisions . the first to realize that the domino effect is a collective phenomenon was d. e. shaw @xcite . he noted that the dominoes lean on each other after collision . consequently the collision of dominoes is fully inelastic and this is the main source of energy losses during the process . it is tempting to invoke conservation of angular momentum at the collision as in @xcite , but this does not apply , since colliding dominoes topple around different axes @xcite . the reason to come back on the domino effect is a question posed at the dutch science quiz 2012 reading : `` _ how many dominoes does one need to topple a domino as tall as the domtoren ? _ '' ( a tower of 112 meter high ) . a restriction was that the dominoes have the same aspect ratios as that of standard dominoes . this restriction is essential for an unique answer , since the thinner the domino ( with fixed height ) the easier it topples . the idea behind the question is that every successive domino is a factor @xmath0 larger than the preceding one . the dominoes grow in size as a geometric series @xmath1 and the correct answer is the @xmath2 that makes @xmath1 times the height of a standard domino equal to 112 meter . whitehead @xcite describes a demonstration with a magnification factor @xmath3 . in order to turn this problem into a scientific problem , one has to make further specifications . one is a rule for the distance between successive dominoes . although not specified in the problem , the logical choice is to scale the distance with the size . to let the distance grow with the size of the domino is reasonable , the larger the domino , the more space one needs between them . moreover , there will be an optimal distance in each cycle and this optimum will scale with the size of the dominoes . so we restrict ourselves to a scale invariant model . in order to make the mechanical model precise , a few more conditions have to be imposed : 1 . the dominoes may not slide with respect to the ground . it is clear that sliding is an energy loss that impedes the domino effect . infinite friction with the floor is favorable and can basically be reached in practice . thus the only possible motion is to topple over . so the state of domino is given by a single angle @xmath4 measuring the deviation from the normal to the ground . 2 . the dominoes stay in contact with each other after the collision @xcite . this makes the collisions fully inelastic . in practice this condition is fulfilled as the movie of the the demonstration with wooden dominoes shows @xcite . inelastic collisions are the main energy drain . steel dominoes would do better , but are too costly to produce in large sizes . we take this condition as a constraint on the maximum possible @xmath0 . 3 . after hitting a new domino , the dominoes slide frictionless over each other . it is of course not realized with wooden dominoes , but it would be not difficult to smooth the surfaces and minimize the friction by lubrication . inclusion of friction is quite well possible @xcite , leading to a substantial complication in the calculation , while adding little to the understanding of the mechanism . we start the discussion with the elementary process of one domino colliding with another at distance @xmath5 and tumbling together downwards . this provides the setting for the longer train of falling dominoes . the dimensions of the dominoes are denoted by height @xmath6 , the thickness @xmath7 and the width @xmath8 . we will consistently work with dimensionless distances to ease the calculations . the parameter @xmath9 hardly enters in the calculation . the width is only co - determinant for the mass @xmath10 of the dominoes . as in all gravitational phenomena , the total mass of the object drops out of the equations , but not the mass distribution . we consider two ( extreme ) cases : @xmath11 , referring to massive dominoes where the mass is proportional to the volume and @xmath12 , referring to hollow dominoes , with a mass proportional to the surface . the latter case seems curious , because standard dominoes are massive . however in an attempt to topple a very large domino @xcite , the larger dominoes have to be hollow for practical reasons . as mentioned the dominoes and their mutual distance all have the same aspect ratios : @xmath13 and @xmath14 , which is taken equal to the ratio of standard dominoes . each new domino is @xmath0 times larger than its predecessor . one could consider varying @xmath5 and @xmath0 ; it would make the discussion considerably more involved and not so much richer , since we investigate the largest possible magnification factor and it turns out that this occurs at a well defined value of @xmath5 . so deviating from the optimal @xmath5 and @xmath0 makes the domino effect less effective . restricting ourselves to the scale invariant case allows to discuss only one cycle in the domino effect , all the others are similar . a dimensionless time is somewhat more delicate . the natural time unit reads @xmath15 where @xmath16 is the moment of inertia of the foremost domino . this is the appropriate time scale for one cycle , the next is a factor @xmath17 larger , the previous a factor @xmath17 smaller . in the formulas we will not express time in terms of the unit @xmath18 , as this complicates the connection between different cycles . the description of the train of falling dominoes centers around the tilt angle @xmath4 of the foremost domino as function of the time @xmath19 . it starts at @xmath20 and ends at @xmath21 at collision angle @xmath22 , where the foremost domino looses its role as head of the falling train of dominoes . @xmath23 is given by @xmath24 after @xmath23 , the domino continues to fall , but it becomes a slave , just as the previous domino was of the foremost . we number the dominoes with respect to the foremost falling domino , which gets the number 0 , or by default no number . so @xmath25 is the tilt angle of the foremost domino , @xmath26 that of its predecessor and so on . the calculation of @xmath21 gives the propagation speed of the domino effect . it is of no concern here . we only note that @xmath21 will be proportional to @xmath18 defined in ( [ a1 ] ) and therefore @xmath21 will increase with a factor @xmath17 in every cycle . this slowing down of each cycle is the beauty of the record attempt @xcite , where one can see _ ad oculos _ the details of a cycle . the derivative of @xmath4 with respect to @xmath19 is the angular velocity @xmath27 as long as @xmath28 is positive the domino train proceeds . so the question is whether @xmath28 remains positive till the foremost collides with the next one . in order to calculate @xmath28 we have to understand what happens during a collision and what is the equation of motion between collisions . the central function in this calculation is the relation between the angle @xmath26 of domino 1 as function of the angle @xmath4 of the leading domino 0 . we start the discussion with two dominoes 0 and 1 . to ease the notation we replace from now on @xmath29 by @xmath5 and @xmath30 by @xmath7 . consider domino 1 , freely rotating towards the still upright domino 0 . in order that domino 1 keeps moving , the initial push must be large enough to overcome the point of highest potential energy , occuring at the angle @xmath31 , given by @xmath32 here we assume that domino 1 reaches this point before that it hits domino 0 . so we must have @xmath33 . with ( [ a2 ] ) and ( [ b1 ] ) this implies @xmath34 provided that @xmath5 fulfils this criterion , we find for the maximum potential energy @xmath35 under marginal circumstances , domino 1 has a vanishing rotation at this point and thus is its total energy also @xmath36 . while falling the potential energy decreases and reaches at the collision angle @xmath23 the value @xmath37 the energy difference is converted into kinetic energy . at the collision the kinetic energy amounts @xmath38 where @xmath39 is the moment of inertia of domino 1 . so the angular velocity at the collision is given by ( with @xmath40 the time scale ( [ a1 ] ) for domino 1 ) @xmath41 we consider the collision as instantaneous . domino 0 gets an impuls @xmath42 from 1 , changing the angular velocity of domino 0 abruptly from 0 to @xmath43 , given by @xmath44 where @xmath45 is the arm of the torque exerted on domino 0 . through the principle action is reaction , domino 0 gives an impuls @xmath46 to domino 1 , such that its angular velocity changes from @xmath47 instantaneously to @xmath43 ( as we will see in a moment ) . so we have the equation @xmath48 where @xmath49 is the moment arm of the torque on domino 1 with respect to its turning point . for the computation of @xmath45 and @xmath49 we have to rely on the goniometry between touching dominoes which is sketched in fig . [ dominoes ] . .goniometric relations referring to fig . [ gonio ] [ cols="<,^ " , ] the various goniometric relations are explained in table [ gonio ] . the basic relation between the angle @xmath26 and @xmath4 follows by equating the given expression for bc and for bc as bd@xmath50 @xmath51 it is a direct consequence of the condition that the dominoes lean on each other after a collision . ( [ b8 ] ) defines @xmath26 as a function @xmath52 of @xmath4 in the interval @xmath53 . the moment arms @xmath45 and @xmath49 , as defined in table [ gonio ] , are obviously equal at collision where @xmath54 . the angular velocity of domino 1 just after collision , equals that of domino 0 . to see this we have to determine the relation between @xmath55 and @xmath28 , when the motion of domino 1 is a slave of domino 0 . generally holds @xmath56 the prime denotes differentiation with respect to the argument and is used for shortness . differentiation of ( [ b8 ] ) with respect to @xmath4 gives @xmath57 the last equality is demonstrated in table [ gonio ] . using this relation for @xmath54 , where @xmath58 , leads to @xmath59 , showing that the two angular velocities are equal at collision . we use arguments for the angle @xmath4 and subscripts for corresponding time . so @xmath47 , in ( [ b5 ] ) , is the angular velocity of domino 1 just before the collision with 0 and @xmath60 is its value just after the collision has taken place . the equality @xmath58 ( at collision ) makes it easy to eliminate the impuls @xmath42 , yielding the following relation between @xmath43 and @xmath47 @xmath61 here we see that the ratio of the moments of inertia of the two dominoes is the important ingredient for the collision . it equals @xmath62 with @xmath11 for massive dominoes and @xmath12 for hollow dominoes . now @xmath43 must be large enough to get the combination of the two sliding dominoes over the maximum of their combined potential energy , which equals @xmath63.\ ] ] note that domino 1 weighs by a factor @xmath64 less in the sum . we get a condition for the domino effect by requiring that the kinetic energy of the pair 1 and 0 must be larger than the potential barrier . with @xmath65 the condition reads @xmath66,\ ] ] where @xmath67 is the tilt angle where the maximum of @xmath68 is reached . it will be close the angle @xmath31 where domino 0 reaches its maximum potential energy , but the contribution of domino 1 is decreasing at this point and so @xmath67 is ( slightly ) smaller than @xmath31 ( as the contribution of domino 1 only counts by a factor @xmath64 less ) . we now have sufficient equations to determine maximum factor @xmath0 still allowing the domino 1 to topple over domino 2 : the expression ( [ b5 ] ) for @xmath47 , the relation ( [ b11 ] ) between @xmath43 and @xmath47 and the condition ( [ b13 ] ) for @xmath43 . @xmath40 is related to @xmath18 by @xmath69 the above equations impose on the parameter @xmath0 a condition for given @xmath5 . the boundary , for which the domino effect becomes marginal , is given by the equality in ( [ b13 ] ) . although the equations can written out explicitly , their solution is too complicated to discuss analytically , because the angle @xmath67 can not be found analytically . however it is simple to do it numerically . start with a value of the separation @xmath5 ( well beyond the restriction ( [ b0 ] ) ) and a low value for @xmath0 for which we are sure that the condition ( [ b13 ] ) is fulfilled . then determine @xmath67 by slowly raising @xmath4 from @xmath54 and see whether the maximum @xmath70 is reached . once the @xmath71 starts to decrease , one is beyond the maximum . it may happen right away at @xmath54 and then the maximum is @xmath72 , where ( [ b13 ] ) is surely fulfilled . otherwise check whether ( [ b13 ] ) is fulfilled and if so raise @xmath0 . if not , one is beyond the boundary , which then can be determined more accurately by interpolating beteen the present @xmath0 and the previous @xmath0 . the outcome of this exercise is shown for hollow dominoes in fig . [ marginal ] . the flank of the curve at the large - separation side is trivial : the domino must be large enough to hit the next domino . the discussion of the previous section suffers from the fact that we have assumed that the initial push to domino 1 is marginal , i.e. just sufficient to get it over its own potential barrier . the bound following from this condition is too restrictive in practice . we now assume that the initial push is large enough to set the train into motion and ask the question whether the train can sustain itself , i.e. whether the total kinetic energy after collision with a new domino , is large enough to get the enlarged collection over its potential barrier . in every cycle the potential barrier can only shift to smaller angles @xmath4 , since at the tail , one more domino is in the train , which is already ( long ) over its potential maximum . once the maximum occurs at @xmath54 , there is no question anymore whether the new train will make it , since it runs only downhill . the analysis of this problems runs to a large extent along the same lines as in the previous section . we divide out the mass and are concerned about the total height contribution to the potential energy of a train of @xmath73 dominoes leaning on each other . it follows as the sum @xmath74.\ ] ] likewise the total kinetic energy can be represented by @xmath75 ^ 2.\ ] ] here @xmath76 is the tilt angle of domino @xmath77 as function of the tilt angle @xmath4 of the foremost and @xmath78 is its derivative . in appendix [ rotation ] we derive that conservation of energy between collisions implies relation ( [ d9 ] ) or @xmath79 this enables to calculate @xmath80 for given @xmath81 . ( [ c3 ] ) also yields the initial value @xmath82 if we combine it with the collision equation and the scale invariance . in appendix [ coltrain ] we show that the angular velocity @xmath43 of the foremost domino , just after being hit , is related to the angular velocity @xmath47 of the hitting domino as @xmath83 so far this is just a generalization from 2 dominoes to @xmath73 dominoes in the train . the new element is that we require that the cycles are self - similar . @xmath84 the three equations ( [ c3]-[c5 ] ) determine the values of @xmath81 . @xmath85 } , \\*[4mm]\ ] ] here @xmath86 is the `` fuel '' of the domino effect : the difference between the potential energy of an upright domino and a fallen domino . @xmath87 the value of @xmath88 is given in ( [ a18 ] ) of appendix [ geometry ] . in ( [ c6 ] ) we have eliminated the value @xmath89 using relation ( [ d8 ] ) . with these equations one can check whether the kinetic energy of the train after collision , exceeds the potential increase from @xmath54 to its maximum at @xmath67 @xmath90,\ ] ] we now summarize the steps to find the maximum @xmath0 with a domino effect as function of @xmath5 . in the appendices the derivations and additional expressions are given . 1 . compute for a given @xmath5 and @xmath0 the initial value of @xmath91 as given by ( [ a18 ] ) or ( [ a16 ] ) , given in appendix [ geometry ] , depending on the condition ( [ a17 ] ) . 2 . start the iteration scheme for the @xmath92 and @xmath78 , as described in appendix [ geometry ] , for the angle @xmath54 . this yields the value of @xmath93 . compute with ( [ c7 ] ) the value of the fuel @xmath86 and with ( [ c6 ] ) the initial angular velocity of a cycle @xmath82 . 3 . then search with equation ( [ c1 ] ) for the value @xmath67 for which @xmath94 is maximal and find the corresponding minimal @xmath95 with ( [ c3 ] ) . 4 . if @xmath95 is still positive , raise the value of @xmath0 and repeat the previous steps 2 - 4 in order to find again the minimal angular velocity @xmath95 . keep raising @xmath0 till the minimum becomes zero . this gives the maximum magnification factor for the chosen value of @xmath5 5 . do this story for @xmath96 the result of the calculation is shown in fig . [ growth ] for hollow and massive dominoes . the curve consists of a rising part for @xmath97 an a drop - off as @xmath98 for higher values . the latter part of the curve reflects the condition the foremost domino must be hit by the previous in order to be toppled . this limit is fairly unrealistic , but follows from our condition that the dominoes can not slip over the ground but only topple . near that line , the foremost domino hits the next at a low point , such that the tendency to slip is larger than to topple , as the moment arm becomes very small . this appendix discusses the geometric relations between the row of falling dominoes that lean on each other . in section [ two ] we gave in equation ( [ b8 ] ) the relation between the tilt angle @xmath26 of domino 1 and the @xmath4 of domino 0 . the successively higher numbered dominoes follow from the same relation as domino @xmath77 is related in exactly the same way to @xmath99 as 1 to 0 . in general @xmath100 thus equation ( [ a2 ] ) gives all the tilt angles of the followers in terms of @xmath4 . rather then using the explicit relation @xmath101),\ ] ] we construct the relation in terms of a rotation in carthesian coordinates @xmath102 with the definitions @xmath103 , \quad y_i = \cos ( \theta_{i+1 } -\theta_i)= \sqrt{1 - x^2_i},\ ] ] the rotation equations read ( using @xmath104 ) @xmath105 these equations are the same for every pair of successive dominoes . note that this scheme contains only multiplications and one square root and not any goniometric function , which speeds up the iteration . the function @xmath106 is the central ingredient for the calculations . so we list a few more of its properties . as @xmath54 corresponds to the collision angle @xmath23 for domino 1 we have @xmath107 by differentiating ( [ a2 ] ) with respect to @xmath4 , we construct a recursion relation between the derivatives @xmath108 for the first factor we can use ( [ b10 ] ) in the form @xmath109 with the convention the indices 0 apply to the foremost . we cast the combination of ( [ a10 ] ) and ( [ a9 ] ) in the convenient recursive form @xmath110 as last useful formula we write ( [ a2 ] ) as @xmath111 and differentiate with respect to @xmath4 , yielding for the derivative @xmath112 we will use this expression for @xmath54 with the result , using ( [ a7 ] ) and ( [ b10 ] ) @xmath113 the derivatives of the @xmath114 are employed in the expression of the angular velocities @xmath115 in terms of @xmath28 @xmath116 this completes the discussion of the properties of the falling dominoes in terms of the parameters @xmath4 and @xmath28 of the foremost . the transformations ( [ a6 ] ) generate for a given @xmath4 the sequence of @xmath114 or rather the sets @xmath117 and @xmath118 and with these values we can iterate simultaneously the angular velocities @xmath115 in terms of @xmath28 . in principle we would have to iterate @xmath119 many steps , but the tilt angles rapidly approach the stacking angle @xmath120 , following as @xmath121 the stacking angle is the angle in which a static row of dominoes leans on each other for a given separation @xmath5 . so a limited number of iteration steps suffices . in fact for rather wide separation the stacking angle is not realized , because the previous domino is too small to lean on the next . this happens when @xmath122 the last tilt angle @xmath88 which enables leaning is given by @xmath123 as soon as @xmath124 exceeds @xmath91 , one must end the iteration , because for larger angles the domino does not lean anymore , but falls free and does not contribute to the mechanics of the domino effect . in case condition ( [ a17 ] ) is fulfilled , we take @xmath125 . the domino effect is a succession between tumbling and colliding . in this appendix we consider the phase of tumbling . as the dominoes slide frictionless over each other , the motion is completely determined by the conservation of energy . energy is lost in the collissions . the energy has a kinetic and a potential part . the potential energy of domino @xmath77 is given by @xmath126 the power of @xmath127 comes in the expression because the mass goes down with a factor @xmath128 and the size with @xmath129 in every step backwards down the train . the kinetic energy of domino @xmath16 reads @xmath130 with the moment of inertia @xmath131 @xmath132 /3= m \ , [ h^2 + d^2 ] \ , r^{-(q+1)i}/ 3= i \ , r^{-(q+1)i}.\ ] ] the @xmath133 power of the magnification factor follows from the mass and the size squared . the total energy of the falling train then equals @xmath134 we make the energy dimensionless by dividing it by @xmath135 yielding @xmath136 with the total dimensionless potential energy @xmath137 @xmath138\ , r^{-qi}\equiv h(\theta)\ ] ] and the total dimensionless moment of inertia is @xmath139 ^ 2 \equiv j(\theta).\ ] ] for a given value of @xmath4 all the ingredients for the functions @xmath94 and @xmath140 are well defined by the recursion relations of the previous section . we need in section [ general ] the values of @xmath140 for the angles @xmath54 and @xmath141 . using ( [ a14 ] ) one finds the relation @xmath142.\ ] ] conservation of energy implies that @xmath143 is independent of time of the tilt angle , such that we have @xmath144 this allows to find @xmath80 for a given @xmath81 . the stages of rotational motion are connected by collisions . in this appendix we compute the initial value @xmath81 from the final value @xmath145 of the previous cycle . as we mentioned in the introduction , we assume that the two colliding dominoes stick together after the collision . that means that in the in the center of mass system the relative kinetic energy is completed dissipated . in other words the collision is fully inelastic . we assume that we have a fully developed domino train such that the cycles are the similar ( upon a factor @xmath17 ) : before the collision the hitting domino has an angular velocity @xmath146 and the upright domino gets a velocity @xmath81 after the collision . the idea is that during the collision , forces are exerted in a very short time span , such that the angles do not change during the collision . instead the angular velocities make a jump . when domino @xmath147 hits @xmath148 , its own angular velocity is suddenly reduced and that of @xmath148 jumps to the non - zero value @xmath82 . the jumps in the angular velocity propagate downwards in magnitude , in order to keep the dominoes in contact . therefore the impulses have to propagate downwards in order to realize these jumps . the jump of the foremost domino is from @xmath149 before the collision to @xmath82 after the collision @xmath150 the second equality gives the integrated torque during the collision , which is the impuls @xmath151 exerted by domino 1 on domino 0 times the arm @xmath152 with respect to the rotation axis of 0 . likewise the domino 1 feels the impuls @xmath153 from domino 0 and @xmath154 from domino 2 . @xmath155 = f_1 a_1 - f_0 b_0.\ ] ] the first term gives the value of @xmath55 as calculated from the foremost domino 0 . the second term is the angular velocity of 1 before the collision when domino 1 was the foremost with angle @xmath23 . the factor @xmath156 is there for reason of generality and the index 1 in @xmath157 is to indicate that here domino 1 is seen as the foremost . in general domino @xmath77 receives @xmath158 from @xmath99 and @xmath159 from @xmath160 . as general equation we get @xmath161=f_i a_i - f_{i-1 } b_{i-1}.\ ] ] we see that impuls @xmath159 occurs in two equations : for domino @xmath160 with a minus sign and for domino @xmath77 with a plus sign . we can eliminate the impulses by multiplying equation for domino @xmath77 with the factor @xmath162 and add them up . if @xmath162 has the property @xmath163 the impulses drop out of the sum . we take @xmath164 . comparing this recursion for the @xmath162 with the recursion ( [ a11 ] ) for the derivatives of the tilt angles we may identify @xmath165 the argument @xmath54 occurs in this formula , because the moment arms are taken in the situation where the foremost domino is hit and thus still has a tilt angle @xmath54 . using ( [ e5 ] ) the coefficient of @xmath81 becomes @xmath166 ^ 2 = i \ , j(0 ) .\ ] ] the coefficient of @xmath157 reads with ( [ a14 ] ) @xmath167 ^ 2 = i\ , ( j(0)-1),\ ] ] such that we obtain the collision law @xmath168 99 robert b. banks , _ towing icebergs , falling dominoes and other adventures in applied mechanics _ , princeton university press , 1998 . c. j. efthimiou and m. d. johnson , domino waves , siam review * 49 * ( 2007 ) 111 - 120 . jearl walker , scientific american , august 1984 . b. g. mclachlan , g. beaupre , a. b. cox and l. gore , falling dominoes , siam review * 25 * ( 1983 ) 403 - 404 . w. j. stronge and d. shu , the domino effect : successive destabilization by cooperative neighbours proc . lond . a * 418 * ( 1988 ) 155 - 163 . r. larham , validation of a model of the domino effect , arxiv:0803.2898v1 ( 2008 ) . d. e. shaw , mechanics of a chain of dominoes am . j. phys . * 46 * ( 1978 ) 640 - 642 . j. m. j. van leeuwen , the domino effect , arxiv : physics/0401018v1 ( 2004 ) . j. m. j. van leeuwen , am . j. phys . * 78 * ( 2010 ) 721 - 727 . at http://www.lorentz.leidenuniv.nl/@xmath169jmjvanl/friction.cc a program for calculating the velocity of the domino effect can be found and at http://www.lorentz.leidenuniv.nl/@xmath169jmjvanl/ideal.cc one for the frictionless sliding . l. a. whitehead , am . j. phys , * 2 * ( 1983 ) 182 . see ( in dutch ) http://www.wetenschap24.nl/programmas/nwq/answers12/antwoord-14.html .

the conditions are investigated under which a row of increasing dominoes is able to keep tumbling over . the analysis is restricted to the simplest case of frictionless dominoes that only can topple not slide . the model is scale invariant , i.e. dominoes and distances grow in size at a fixed rate , while keeping the aspect ratios of the dominoes constant . the maximal magnification factor for which a domino effect exist is determined as a function of the mutual separation .

introduction two dominoes a general domino train relations between the falling dominoes the motion between collisions the collision equation

a well - known surface model is the one of helfrich , polyakov and kleinert ( hpk ) @xcite . the phase structure of hpk model is connected to string models and membranes @xcite and therefore has long been investigated theoretically @xcite and numerically @xcite . membranes are considered to be two - dimensional surfaces in the conventional hpk model , and hence homogeneity is assumed in the model . however , membranes are not always homogeneous ; cell membranes are known to be supported by skeletons . hop diffusion of membrane proteins or lipids was experimentally observed recently @xcite . the free diffusion of molecules is prohibited , and the diffusion is localized in some domains on the cell membranes . the origin of this is considered to be due to the cytoskeletons . artificial membranes are considered to have skeletons because they are partly polymerized @xcite . the phase structure of skeleton models should therefore be studied as a statistical mechanical problem . skeleton models for the cytoskeleton were already investigated in @xcite . a hard - wall and hard - core potential was assumed on the polymer chains with junctions , and the responses to some external stress and the compression modulus were obtained @xcite . giant fluid vesicles coated with skeletons was experimentally investigated , and the mechanical properties were reported @xcite , where the actin filaments introduce an inhomogeneous structure in homogeneous artificial membranes . a compartmentalized surface model was reported to undergo a first - order transition @xcite , where the compartment boundary prevents the vertices from the free diffusion . in @xcite , the phase structure of a surface model with skeleton was investigated , and it was reported that the model has a first - order transition between the smooth phase and the crumpled phase . the interaction of the model in @xcite is described by one - dimensional bending energy for linear chains ( or bonds ) and two - dimensional bending energy for junctions . the two - dimensional gaussian bond potential is also assumed in the hamiltonian . thus , the interaction of the model in @xcite seems a little bit more complicated than that of hpk model . one of the reasons of this is because the two - dimensional elasticity is assumed at the junctions . therefore , it is interesting to see whether the transition of @xcite occurs in more simplified skeleton models . one possible model is obtained from the model of @xcite by replacing the elastic junction with a rigid junction . in this letter , we study the rigid junction model by using the canonical monte carlo ( mc ) simulations and see how the transition of @xcite occurs in such a simplified model . we must note that the rigid junction model is not included in the elastic junction models . in fact , there are two types of elasticity at the elastic junctions ; one is the out - of - plane elasticity and the other is the in - plane elasticity . the former elasticity can be rigid in the elastic junction model of @xcite in the limit of infinite bending rigidity @xmath0 , however , the latter one can not be controlled in the elastic junction model . therefore , the rigid junction model in this letter and the elastic junction model in @xcite are considered to be two different models . the model is defined on the triangulated surfaces , which are characterized by @xmath1 the total number of vertices including the junctions , @xmath2 the total number of vertices on the chains , @xmath3 the total number of junctions , and @xmath4 the length of chains between junctions . the junctions are assumed as rigid plates ; twelve of them are pentagon and the others are hexagon . it should be noted again that @xmath3 is included in @xmath1 ; a junction is counted as a vertex . figures [ fig-1](a ) and [ fig-1](b ) show surfaces of size @xmath5 and @xmath6 , respectively . thick lines denote the chains terminated at the junctions , and the chains together with junctions form the compartment boundary . all the vertices can fluctuate only locally on the chains as well as inside the compartments , and they are prohibited from the diffusion because of the fixed connectivity nature of the lattice . 0.1 in ( 0,0 ) ( 10,10 ) ( 18,8.5)(0,0)(a ) @xmath7 ( 39,8.5)(0,0)(b ) @xmath8 and ( b ) @xmath6 , where @xmath1 is the total number of vertices including the total number of junctions , @xmath2 is the total number of vertices on the chains , @xmath3 is the total number of junctions , and @xmath4 is the length of chains between junctions . thick lines denote the chains , which terminate at the junctions . , width=396 ] the lattice of @xmath5 in fig.[fig-1](a ) corresponds to that of @xmath9 in fig.[fig-1](a ) of @xcite . the reason why @xmath10 of the surface in fig.[fig-1](a ) is smaller than @xmath11 of the one in @xcite is because the junctions are rigid objects in this letter while they are elastic ones in @xcite . one hexagonal junction reduces @xmath1 by @xmath12 , and one pentagonal junction also reduces @xmath1 by @xmath13 if they were assumed as rigid objects . thus , we can check that @xmath14 , where @xmath15 and @xmath16 are the total number of pentagonal junction and that of hexagonal junctions , respectively . the triangulated surfaces such as the ones shown in figs . [ fig-1](a ) and [ fig-1](b ) are constructed as follows : firstly , we process the icosahedron by dividing every edge into @xmath17-pieces of uniform length , and obtain a triangulated surface of size @xmath18 (= the total number of vertices ) , where @xmath17 is the number of division of an edge in the icosahedron . the obtained surfaces are thus characterized by @xmath19 and @xmath20 , where @xmath21 is the total number of vertices with co - ordination number @xmath22 . those @xmath23-vertices and @xmath24-vertices respectively form pentagonal junctions and hexagonal junctions together with their nearest neighbor vertices , in the next stage . secondly , compartmentalized structures are obtained by dividing @xmath17 further into @xmath25-pieces ( @xmath26 ) , and we have uniform chains of length @xmath27 . those chains terminate at the above - described pentagonal or hexagonal junctions . these junctions include @xmath12 ( or @xmath28 ) vertices and @xmath29 ( or @xmath16 ) bonds according to whether they are pentagonal or hexagonal , and hence each junction has many degrees of freedom at this stage . thirdly , the junctions are assumed as rigid objects on the triangulated surfaces : the pentagonal ( hexagonal ) junction is identified with a regular pentagonal ( regular hexagonal ) plate . as a consequence , the @xmath12 ( or @xmath28 ) vertices together with the @xmath29 ( or @xmath16 ) bonds are identified with a pentagon ( or a hexagon ) . thus , we have compartmentalized surfaces with rigid junctions as those shown in figs . [ fig-1](a ) and [ fig-1](b ) . on the surfaces shown in figs . [ fig-1](a ) and [ fig-1](b ) , we have @xmath30 , @xmath31 and @xmath32 , @xmath33 , respectively . we also note that the compartmentalized structures shown in figs . [ fig-1](a ) and [ fig-1](b ) are identical to those of the model in @xcite except the fact that the junctions are rigid . the total number @xmath34 of the compartments depends on the surface size @xmath1 . consequently , @xmath34 is increased with increasing @xmath1 just as in the surfaces in @xcite . we choose the chain length @xmath4 between the junctions as constant so that it is independent of @xmath1 in the model . we fix the chain length @xmath4 such that @xmath35 which respectively correspond to the values @xmath36 , @xmath37 , the total number of vertices inside a compartment @xcite . the reason why we fix @xmath38 is because the size of compartment is considered to be finite in the cell membranes , and also because it is expected that total number of lipids in the compartment remains finite in the cell membranes . the hamiltonian of the model is given by a linear combination of the two - dimensional gaussian bond potential @xmath39 , the one - dimensional bending energy @xmath40 , which are defined by @xmath41 in these expressions , @xmath42 in @xmath39 denotes the sum over all the bonds @xmath43 connecting the vertices @xmath44 and @xmath45 , and @xmath42 in @xmath40 denotes the sum over bonds @xmath44 and @xmath45 , which contain not only bonds in the chains but also _ virtual bonds _ that connect the center of a rigid junction and the neighboring vertices on the chains . the symbol @xmath46 in @xmath40 is the angle between the bonds @xmath44 and @xmath45 , which include the virtual bonds described just above . the gaussian potential @xmath39 is defined not only on the chains but also on all other bonds . as a consequence , the model is considered to be a surface model , although the mechanical strength is maintained by one - dimensional elastic skeletons joined to each other at the rigid junctions . figure [ fig-2 ] shows a hexagonal rigid junction connected to chains , where the angle @xmath46 is defined not only at the vertices on the chains but also at the corners ( = _ virtual vertices _ ) of the junction . the triangular lattices attached to the chains were eliminated from the schematic drawing in fig . [ fig-2 ] for simplicity . in @xmath40 is defined not only at the vertices on the chains but also at the corners ( = virtual vertices ) of the junction . the triangular lattices attached to the chains were eliminated from the schematic drawing for simplicity . , width=396 ] we must comment on the size of the junctions . the junctions are two - dimensional objects and therefore have their own size to be fixed . the size of junction can be specified by the edge length @xmath47 ; the perimeter length of the pentagonal ( hexagonal ) junction is therefore expressed by @xmath48 ( @xmath49 ) . in this letter , we fix the size of the junctions such that @xmath50 the value @xmath51 is quite smaller than that of the elastic junctions in @xcite , where the edge length squared is @xmath52 because of the relation @xmath53 satisfied in the model of @xcite . since the junctions are two - dimensional rigid objects , it is expected that large - sized junctions influence equilibrium properties of the surface . therefore , in order to reduce such unclear size - effects we choose the size @xmath47 relatively small such as in eq.([junctionsize ] ) . we must note that the junctions in the snapshots of figs . [ fig-1](a),(b ) were drawn larger than that expected from eq.([junctionsize ] ) ; the junction size in the snapshots is fixed according to @xmath54 . in the following section , we discuss how do we fix the size @xmath47 to the assumed value in eq.([junctionsize ] ) in the mc simulations . the partition function @xmath55 of the model is defined by @xmath56,\quad s(x)=s_1 + b s_2,\ ] ] where @xmath57 is the bending rigidity corresponding to the one - dimensional bending energy , and @xmath58 denotes that the center of the surface is fixed . the integration @xmath59 is a product of the integration over vertices and that of junctions such that @xmath60 where @xmath61 is the integration over the degrees of freedom for three - dimensional translations and rotations . the bending rigidity @xmath57 has unit of @xmath62 , where @xmath63 is the boltzmann constant , and @xmath64 is the temperature . the surface tension coefficient @xmath65 of @xmath39 is fixed to @xmath66 ; this is always possible because of the scale invariant property of the model . in fact , in the expression @xmath67 we immediately understand that @xmath66 is possible , because the factor @xmath65 of @xmath39 can be eliminated due to the scale invariance of the partition function . since the unit of @xmath65 is @xmath68 , the length unit of the model is given by @xmath69 . we use the unit of length provided by @xmath70 in this letter , although @xmath65 is arbitrarily chosen to be fixed . it must also be noted on the relation of the size @xmath47 in eq.([junctionsize ] ) to the thermodynamic limit @xmath71 . as discussed in the last part of the previous section , the limit @xmath71 is taken in our model so that the compartment size remains constant as in the model of @xcite . as a consequence , @xmath47 is also remained constant in the limit @xmath71 . the integration @xmath59 in the partition function is done by the canonical metropolis technique . the update of @xmath72 in mc can be divided into two steps , which are corresponding to the integrations @xmath73 and @xmath74 in eq.([integration ] ) . the first is the update of @xmath72 for the vertices including those on the chains : @xmath72 are shifted so that @xmath75 , where @xmath76 is randomly chosen in a small sphere . the new position @xmath77 is accepted with the probability @xmath78 $ ] , where @xmath79 . the second is the update of the position of the junctions as three - dimensional rigid objects . this can be further divided into two processes : the first is a random three - dimensional translation , and the second is a random three - dimensional rotation . all of these mc processes are independently performed under about @xmath80 acceptance rate , which is controlled by small numbers fixed at the beginning of the simulations . we introduce the lower bound @xmath81 for the area of triangles . no lower bound is imposed on the bond length . the junction size @xmath47 is fixed to @xmath51 in eq.([junctionsize ] ) during the thermalization mcs . the initial value of @xmath47 is given by @xmath82 on the surfaces such as those shown in figs . [ fig-1](a ) and [ fig-1](b ) . thus , we reduce @xmath47 from @xmath83 to @xmath51 by @xmath84 at every @xmath85 mcs in the first @xmath86 mcs . because of this forced reduction of the junction size , the equilibrium statistical mechanical condition seems to be violated in the first @xmath86 mcs . therefore , relatively many thermalization ( @xmath87 or more ) mcs is performed after the reduction . further thermalization mcs should be performed , if necessary . we use surfaces of size @xmath88 , @xmath89 , @xmath90 , and @xmath91 for the length @xmath30 , and @xmath92 , @xmath93 , and @xmath94 for the length @xmath32 . a random number sequence called mersenne twister @xcite is used in the simulations . snapshots of surfaces are shown in figs.[fig-3](a)[fig-3](d ) . figure [ fig-3](a ) is a surface of size @xmath95 obtained in the crumpled phase at @xmath96 , and fig.[fig-3](b ) is the one obtained in the smooth phase at @xmath97 . the surface sections of figs . [ fig-3](a ) and [ fig-3](b ) are shown in figs . [ fig-3](c ) and [ fig-3](d ) , respectively . these figures are drawn in the same scale . we immediately see the surface in the smooth phase is actually swollen , while the surface is collapsed in the crumpled phase . 0.1 in ( 0,0 ) ( 10,10 ) ( 19.5,47.5)(0,0)(a ) collapsed surface at @xmath96 ( 43,47.5)(0,0)(b ) smooth surface at @xmath97 ( 17.2,8.5)(0,0)(c ) the surface section ( 40,8.5)(0,0)(d ) the surface section obtained in the crumpled phase at ( a ) @xmath96 and in the smooth phase at ( b ) @xmath97 , both of which are close to the transition point . the figures are drawn in the same scale.,width=396 ] vs. @xmath57 obtained on the surfaces of ( a ) @xmath30 and ( b ) @xmath32 . @xmath98 slightly deviates from @xmath54 . the curves are drawn by the multihistogram reweighting technique.,width=396 ] the gaussian bond potential @xmath98 is shown against @xmath57 in figs . [ fig-4](a ) and [ fig-4](b ) , which correspond to the conditions @xmath30 and @xmath32 , respectively . the solid curves are drawn by the multihistogram reweighting technique @xcite . the values of @xmath98 in the figures slightly deviate from @xmath53 , which is satisfied on the surface without the rigid junctions or with rigid junctions of negligible size . the reason of this discrepancy is because the surface includes the rigid junctions of finite size . a vertex is the zero - dimensional point , while the rigid junction is the two - dimensional plate and hence shares some area on the surface . we find a gap or a jump in @xmath98 of the @xmath95 surface in fig.[fig-4](b ) , which can be viewed as a sign of a discontinuous transition . against @xmath57 obtained on the surfaces of ( a ) @xmath30 and ( b ) @xmath32 . the curves are drawn by the multihistogram reweighting technique.,width=396 ] the mean square size is defined by @xmath99 where @xmath100 is the center of the surface . the crumpling transition is conventionally understood as the one of surface fluctuations accompanied by surface collapsing phenomena , which can be seen in our surface model ; we have seen a collapsed surface and a swollen surface in figs.[fig-3](a)[fig-3](d ) . therefore , we expect that @xmath101 reflects the crumpling transition on spherical surfaces . figures [ fig-5](a ) and [ fig-5](b ) are plots of @xmath101 against @xmath57 obtained under @xmath30 and @xmath32 . we find that the variation of @xmath101 becomes sharp against @xmath57 as @xmath1 increases . thus , it is expected that the variation of @xmath101 has a jump at intermediate value of @xmath57 in either case of @xmath4 . the crumpling transition is originally understood as the one for surface fluctuation phenomena . therefore , the bending energy @xmath102 is expected to reflect the transition , where @xmath103 is given by @xmath104 @xmath102 is the bending energy per vertex , because @xmath103 is the total number of vertices where @xmath40 is defined . @xmath103 includes virtual vertices which are the corners of the junctions ( see also fig.[fig-2 ] ) . those virtual vertices are not counted as the vertices and hence are not included in @xmath2 . total number of virtual vertices are @xmath105 , because the hexagonal junction includes 6-virtual vertices , and the total number of pentagonal junction is @xmath16 . thus , we have eq.([bendingenegryvertices ] ) for @xmath103 , and therefore we have @xmath106 , @xmath107 , @xmath108 , and @xmath109 for the surfaces of length @xmath30 and size @xmath88 , @xmath89 , @xmath90 , and @xmath91 ; and @xmath110 , @xmath111 , and @xmath112 for the surfaces of length @xmath32 and size @xmath92 , @xmath93 , and @xmath94 . figures [ fig-6](a ) and [ fig-6](b ) are plots of @xmath102 against @xmath57 obtained under @xmath30 and @xmath32 . we find the expected behavior in @xmath102 under both conditions @xmath30 and @xmath32 ; @xmath102 has a gap ( or a jump ) at intermediate @xmath57 . this clearly shows that the crumpling transition is of first order . vs. @xmath57 obtained on the surfaces of ( a ) @xmath30 and ( b ) @xmath32 . , width=396 ] against @xmath57 obtained on the surfaces of ( a ) @xmath30 and ( b ) @xmath32 . @xmath113 is defined by eq.([two - dim - bending - energy ] ) and is not included in the hamiltonian . @xmath114 is the total number of bonds where @xmath113 is defined.,width=396 ] the transition can also be reflected in the two - dimensional extrinsic curvature , which is defined by @xmath115 where @xmath116 is the unit normal vector of the triangle @xmath44 . in @xmath113 , @xmath117 denotes the summation over all nearest neighbor triangles @xmath44 and @xmath45 that have the common bond @xmath118 , which includes bonds belonging to the skeleton chains . we denote the total number of bonds , where @xmath113 is defined , by @xmath114 , which is given by @xmath119 . note that @xmath114 includes _ virtual edges _ , which are the one - dimensional edges of the rigid junctions . in fact , we define @xmath113 on the virtual edges , because it is reasonable to define extrinsic curvature on those edges . it is also noted that @xmath113 is not included in the hamiltonian , and therefore @xmath113 gives no mechanical strength to the surface . the two - dimensional extrinsic curvature @xmath120 is plotted in figs . [ fig-7](a ) and [ fig-7](b ) against @xmath57 , which are corresponding to the conditions @xmath30 and @xmath32 . @xmath114 is the total number of bonds described above . we find that the dependence of @xmath120 on @xmath57 shown in fig.[fig-7 ] is almost identical to that of @xmath102 in fig.[fig-6 ] . the gap ( or jump ) seen in @xmath120 also supports that the transition is of first order . the specific heat corresponding to the one - dimensional bending energy @xmath40 is defined by @xmath121 which can also reflects phase transitions if it has an anomalous behavior . figures [ fig-8](a ) and [ fig-8](b ) show @xmath122 against @xmath57 obtained under @xmath30 and @xmath32 . solid curves drawn in the figures were obtained by the multihistogram reweighting technique , and those curves clearly show an expected anomalous behavior indicating that @xmath122 is divergent when @xmath123 ( or equivalently @xmath124 ) . for @xmath40 against @xmath57 obtained on the surfaces of ( a ) @xmath30 and ( b ) @xmath32 . @xmath122 is defined by eq.([specific - heat-1 ] ) . the error bars on the symbols are the statistical error , which is obtained by the binning analysis.,width=396 ] for @xmath113 against @xmath57 obtained on the surfaces of ( a ) @xmath30 and ( b ) @xmath32 . @xmath125 is defined by eq.([specific - heat-2 ] ) . the error bars on the symbols are the statistical error , which is obtained also by the binning analysis.,width=396 ] the specific heat corresponding to the extrinsic curvature @xmath113 in eq.([two - dim - bending - energy ] ) can also be defined by @xmath126 which reflects the transition as @xmath122 does . curvature coefficient for @xmath125 was assumed to be @xmath127 , because @xmath113 is not included in the hamiltonian and therefore the curvature coefficient is not . figures [ fig-9](a ) and [ fig-9](b ) show @xmath125 against @xmath57 obtained under @xmath30 and @xmath32 . we can see in @xmath125 the same anomalous behavior as in @xmath122 . against @xmath103 and ( b ) @xmath128 against @xmath129 obtained on the surfaces of @xmath30 and @xmath130 . the straight lines are drawn by fitting the largest three data of @xmath131 and @xmath128 to eq.([scaling - exponents ] ) . the peak values and the statistical errors for the fittings were obtained by multihistogram reweighting.,width=396 ] in order to see the anomalous behaviors in @xmath122 and @xmath125 in more detail , we plot the peak values of them in figs . [ fig-10](a ) and [ fig-10](b ) in log - log scales against @xmath103 and @xmath1 , respectively . the straight lines were drawn by fitting the data to @xmath132 where @xmath133 , @xmath134 are critical exponents . largest three data were used in the fitting in the case @xmath30 . thus , we have @xmath135 @xmath136 for @xmath30 is inconsistent with the fact that the transition is of first - order , however , @xmath137 is consistent with that . @xmath138 and @xmath139 under @xmath32 support the discontinuous transition . a surface model with skeletons has been investigated by the canonical monte carlo simulations . the skeletons are composed of one - dimensional linear chains and rigid junctions , whose size is chosen sufficiently small compared to the mean chain - length . the surface is a triangulated sphere and divided into a lot of compartmentalized domains , whose boundary corresponds to the skeletons . the mechanical strength of the surface is given by the skeletons . there is no two - dimensional curvature energy in the hamiltonian , while one - dimensional bending energy is defined on the chains connected to each other at the junctions . the two - dimensional gaussian bond potential is included in the hamiltonian just as in the standard surface model of helfrich , polyakov and kleinert . the surface model in this letter is considered to be different from the one with elastic junctions , because the rigid junctions can not be identified with the elastic junctions due to the property on the in - plane elasticity . the surface is characterized by @xmath140 , which are respectively the total number of vertices including the junctions , the total number of vertices on the chains , the total number of junctions , and the length of chains between junctions . the length of chains was fixed to @xmath30 and @xmath32 , which correspond to @xmath36 and @xmath37 the total number of vertices in a compartment . we found that the surface undergoes a first - order crumpling transition between the smooth phase and the crumpled phase . the one - dimensional bending energy @xmath40 has a gap ( or a jump ) at intermediate bending rigidity @xmath57 , and the two - dimensional extrinsic curvature @xmath113also has a gap at that point . these imply that the surface fluctuations are considered to be a first - order transition . moreover , it is found that the mean square size @xmath101 also has a gap at the transition point . this implies that the surface - collapsing phenomenon can be viewed as a first - order transition . the results in this letter together with those in @xcite show that the first - order crumpling transition can be seen in the spherical surface model even when the mechanical strength is maintained only by skeletons , which are composed of linear chains joined to each other at the junctions . moreover , the order of transition is independent of whether the junction is elastic or rigid . it has been reported in @xcite that the transition can be seen in a compartmentalized surface model , whose mechanical strength is maintained by the two - dimensional curvature of the surface in contrast to the model in this letter and that in @xcite . therefore , we can also conclude that the first - order transition occurs in the compartmentalized surface model independently whether the surface is mechanically supported by the skeleton (= the compartment boundary ) or by the surface . fluidity of lateral diffusion of vertices can be considered by using dynamically triangulated mc technique as in @xcite . vertices freely diffuse inside each compartment on the fluid surfaces supported by the skeletons . it is interesting to see how fluidity influences the transition of the skeleton - supported model . many interesting problems remain to be studied on the surface model with skeletons . this work is supported in part by grant - in - aid of scientific research , no . 15560160 and no . 18560185 . m. bowick , s. catterall , m. falcioni , g. thorleifsson , and k. anagnostopoulos , j. phys . i france 6 ( 1996 ) 1321 ; + m. bowick , s. catterall , m. falcioni , g. thorleifsson , and k. anagnostopoulos , nucl . 47 ( 1996 ) 838 ; + m. bowick , s. catterall , m. falcioni , g. thorleifsson , and k. anagnostopoulos , nucl . suppl . 53 ( 1997 ) 746 . m. matsumoto and t. nishimura , `` mersenne twister : a 623-dimensionally equidistributed uniform pseudorandom number generator '' , acm trans . on modeling and computer simulation vol . 1 , january ( 1998 ) pp.3 - 30 . wolfhard janke , _ histograms and all that _ , in : computer simulations of surfaces and interfaces , nato science series , ii . mathematics , physics and chemistry - vol . 114 , proceedings of the nato advanced study institute , albena , bulgaria , 9 - 20 september 2002 , edited by b. dunweg , d.p . landau , and a.i . milchev ( kluwer , dordrecht , 2003 ) , pp . 137 - 157 .

a surface model with skeletons is investigated by using the canonical monte carlo simulations . the skeleton is composed of linear chains , which are joined to each other at the rigid junctions . a one - dimensional bending energy is defined on the linear chains , and no two - dimensional curvature energy is assumed on the surface . the model undergoes a first - order transition between the smooth phase and the crumpled phase . we conclude that the first - order transition of the surface model with skeletons is independent of whether the junctions are elastic or rigid . , , , phase transition , extrinsic curvature , elastic membranes 64.60.-i , 68.60.-p , 87.16.dg

introduction triangulated surfaces model monte carlo technique results summary and conclusion acknowledgment

the snrs consisting of two partially overlapping nonthermal shells represent a significant fraction of the known shell - like snrs . several explanations were put forward to explain the unusual morphology of these objects : _ i _ ) superposition of two separate snrs ; _ ii _ ) collision of two separate snrs ( e.g. uyaniker et al . 2002 ) ; _ iii _ ) breakout phenomenon in a region with a density discontinuity ( e.g. velzquez et al . 2001 ) . we propose an alternative explanation based on the idea that the two - shell snrs can be products of an off - centered supernova ( sn ) explosion in a preexisting bubble created by the wind of a moving massive star . massive stars ( the progenitors of most of sne ) are the sources of strong stellar wind , which creates extended ( tens of parsecs ) bubbles in the interstellar medium . in the absence of stellar proper motion , the wind bubbles blown - up in the homogeneous medium are spherically - symmetric with the star at the center of symmetry . however , after correction for galactic rotation , all stars have a proper motion . most of the massive stars have a peculiar velocity of a few km / s ; some of them have much larger velocities . the stellar proper motion could result in a considerable offset of the sn explosion site from the center of the wind - driven bubble , while the initially spherical shape of the bubble could be significantly distorted if the star reaches the edge of the bubble and the stellar wind starts to interact directly with the ambient interstellar medium ( weaver et al . 1977 ; see fig . 1 for schematic illustration of this effect ) . it is likely that just this situation takes place in the case of the progenitor star of the sn 1987a . wang et al . ( 1993 ) suggested that the large - scale structure to the southeast of sn 1987a ( the dark bay in their fig . 2 ) is the wind - driven bubble created by the moving sn progenitor star during the main - sequence ( ms ) stage , while the napoleon s hat nebula originates due to the interaction of the post - ms winds with the ambient interstellar medium ( see , however , sugerman et al . 2005 ) . after the massive star exploded as a sn , the sn blast wave takes on the shape of the preexisting cavity ( e.g. r@xmath0yczka et al . 1993 ; brighenti & dercole 1994 ; gvaramadze & vikhlinin 2003 ) and the resulting snr becomes considerably non - spherically - symmetric . 2 shows the @xmath1 image ( smith 1997 ) of rcw 86 a shell - like snr with a peculiar protrusion to the southwest . we believe ( gvaramadze & vikhlinin 2003 ) that rcw 86 is an older version of the sn 1987a and that the southwest protrusion is the remainder of a bow shock - like structure created in the interstellar medium by the post - ms winds . given the youth of the snr , we expect that the stellar remnant should still be within the protrusion . motivated by these arguments we searched for a stellar remnant using the _ chandra _ archival data and discovered a neutron star candidate just in the `` proper place '' ( gvaramadze & vikhlinin 2003 ) . we propose that a two - shell snr is a natural consequence of an off - centered cavity sn explosion of a moving massive star , which ended its evolution near the edge of the ms wind - driven bubble . this proposal implies that one of the shells is the former ms bubble reenergized by the sn blast wave . the second shell , however , could originate in two somewhat different ways , depending on the initial mass of the sn progenitor star . it could be a shell swept - up by the sn blast wave expanding through the unperturbed ambient interstellar medium if the massive star ends its evolution as a red supergiant ( rsg ) , i.e. if the star evolves through the sequence : ms - rsg - sn . or it could be the remainder of a preexisting shell ( adjacent to the ms bubble ) swept - up by the fast progenitor s wind during the late evolutionary phases if after the rsg phase the star evolves through the wolf - rayet ( wr ) phase ( i.e. ms - rsg - wr - sn ) . in both cases the resulting ( two - shell ) snr should be associated only with one ( young ) ns ( cf . gvaramadze 2006 ) . we note several distinctions characterizing the second case : ( a ) the birth - place of the stellar remnant could be significantly offset from the center of the preexisting wr shell due to the proper motion of the sn progenitor star , ( b ) the sn explosion site could be marked by a nebula of thermal x - ray emission ( see sect.4.2 ) , and ( c ) the preexisting wr shell causes the rapid evolution of the sn blast wave from the adiabatic phase to the radiative one . the polarized intensity image of the cygnus loop by uyaniker et al . ( 2002 ) shows a prominent shell - like structure encompassing the `` break - out '' region in the south of this snr ( see fig . the geometric center of the shell nearly coincides with a neutron star candidate discovered in x - rays by miyata et al . ( 2001 ; indicated in fig . 3 by a cross ) . uyaniker et al . ( 2002 ) believe that the cygnus loop is actually two individual snrs interacting with each other . an alternative possibility is that the cygnus loop is the result of sn explosion near the south edge of a cavity blown up by the sn progenitor during the ms stage and that the sn precursor was a rsg star ( gvaramadze 2003 ) . this implies that the north ( well - known ) shell of the cygnus loop corresponds to the former ms bubble reenergized by the sn blast wave , while the newly - discovered ( south ) shell is created by the interaction of the sn blast wave with the unperturbed interstellar medium . accordingly , we expect that only one stellar remnant should be associated with both shells . the snr 3c 400.2 consists of two circular radio shells with the centrally - filled thermal x - ray emission peaked on the region where the radio shells overlap each other ( fig . 4 ; dubner et al . 1994 ; yoshita et al . 2001 ) and therefore belongs to the class of mixed - morpology snrs ( i.e. shell - like in the radio and centrally - filled in the x - ray ; rho & petre 1998 ) . the origin of mixed - morphology snrs is usually treated in the framework of white & long s ( 1991 ) model of evaporation of embedded interstellar cloudlets . an alternative explanation of the origin of the centrally - filled thermal x - ray emission is that it is due to the evaporation of dense circumstellar clumps ( produced by the interaction of the fast wr wind with the preceding slow rsg one ; gvaramadze 2001 , 2002 ) . we therefore suggest that the sn precursor was a wr star and that the northwest shell of 3c 400.2 is the former wr shell . this suggestion can be supported by the fact that the northwest shell has a bilateral appearance with the bilateral axis parallel to the galactic plane ( see gvaramadze 2004 ) . if our suggestion is correct , then one can expect that the stellar remnant associated with 3c 400.2 should be within the northwest shell .

the proper motion of massive stars could cause them to explode far from the geometric centers of their wind - driven bubbles and thereby could affect the symmetry of the resulting diffuse supernova remnants ( snrs ) . we use this fact to explain the origin of snrs consisting of two partially overlapping shells ( e.g. cygnus loop , 3c 400.2 , etc . ) .

introduction precursors of two-shell snrs two-shell snrs two examples

he purpose of this paper is to formulate a volume integral equation for the determination of the natural modes of helical structures . low - frequency applications are of particular importance where it can be anticipated the existence of twisted modes of a particularly simple structure . the presented problem formulation is largely motivated by the need of being able to accurately model the field distribution and losses inside twisting three - phase high - voltage power cables at 50@xmath0 , see _ e.g. _ , @xcite . the approach could also potentially be useful for analyzing the wave propagation characteristics of the so called litz wires . helical waveguide structures have been treated previously such as _ e.g. _ , with helical sheaths @xcite , and approximations for wire helices@xcite . presently , there are also very promising numerical techniques being developed that are based on finite element modeling @xcite and the method of moments @xcite . however , to our knowledge there has not been any general presentation regarding analytical modeling of the natural modes of helical structures . it is the aim of this paper to fill in this gap . on the other hand , there is a large body of literature on the general dispersion properties of open waveguides , see _ e.g. _ , @xcite , as well as on the general properties of the electromagnetic volume integral equations , see _ e.g. _ , @xcite . in particular , it is well known that the volume integral operators in electromagnetics are strongly singular and that many questions regarding their spectral theory remain largely open @xcite . nevertheless , it can be shown in very general settings that modes of open waveguides exist and can be interpreted in terms of poles of a meromorphic fourier transform and that these poles depend continuously on the model data ( except for points where poles coalesce or at the boundary of the domain of meromorphicity , _ i.e. _ , at infinity and at the branch - point corresponding to the wavenumber of the exterior domain ) @xcite . when there are sources present , the modes ( the discrete set of eigenfunctions ) can be obtained as the residues of the poles and the non - discrete set is manifested as an integration along the branch - cut @xcite . in practical circumstances , the branch - cut contribution can often be neglected @xcite . in this paper , a general helical waveguide structure is treated by using classical analytic function theory and fourier techniques ( in particular the convolution theorem ) in connection with the electromagnetic volume integral formulation and a cylindrical vector wave expansion of the related dyadic green s functions . the floquet modes are defined as the poles of the corresponding integral operators , and the analytic periodic green s functions are derived by employing the classical poisson summation formula . two independent approaches ( with and without explicit sources ) are used to derive the resulting infinite system of one - dimensional integral equations . these equations can then be discretized by truncation and by using a standard collocation method . as a useful byproduct , the periodic dyadic green s functions provide a simple direct calculation of the quasi - static fields generated by thin helical wires . consider a straight helical waveguide structure of radius @xmath1 consisting of a twisting inhomogeneous and anisotropic material . the twisting means that the cross - section of the guide is rotating along the longitudinal direction of the structure . the waveguide may consist of several layers with different twist , but it is assumed that the material has a smallest common period @xmath2 in the longitudinal direction . the waveguide is assumed to be lossy , and it constitutes an open structure placed in a surrounding homogeneous and isotropic free space . let @xmath3 , @xmath4 , @xmath5 and @xmath6 denote the permeability , the permittivity , the wave impedance and the speed of light in vacuum , respectively , and where @xmath7 and @xmath8 . the wavenumber of vacuum is given by @xmath9 where @xmath10 is the angular frequency and @xmath11 the frequency . the relative permeability and permittivity of the surrounding free space are denoted by @xmath12 and @xmath13 , respectively , and the corresponding wavenumber is given by @xmath14 . the cylindrical coordinates are denoted by @xmath15 , the corresponding unit vectors @xmath16 , the transversal coordinate vector @xmath17 and the radius vector @xmath18 . let @xmath19 and @xmath20 denote the electric and magnetic fields , respectively , where the time - harmonic factor @xmath21 has been suppressed . further , let @xmath22 and @xmath23 denote the imposed electric and magnetic sources , respectively , and which are assumed to be constrained to a finite region @xmath24 inside the waveguide structure . maxwell s equations @xcite for this situation are given by @xmath25 where @xmath26 and @xmath27 are the complex valued relative permeability and permittivity dyadics of the material , respectively . the maxwell s equations are also supplemented with a radiation condition providing a unique solution with fields vanishing at infinity @xcite . the equations can be reformulated in terms of the surrounding free space as @xmath28 where @xmath29 and @xmath30 are the equivalent magnetic and electric sources @xmath31 where @xmath32 and @xmath33 are the magnetic and electric susceptibility ( or contrast ) dyadics , respectively , defined by @xmath34 and where @xmath35 is the identity dyadic . the solution to can now be expressed in terms of the following integral equation of the second kind @xmath36 where the integrals extend over the support of the material @xmath37\times[0,2\pi]\times[-\infty,+\infty]\}$ ] , and the finite source region @xmath24 , respectively , and where @xmath38 , _ cf . . here , @xmath39 and @xmath40 are the electric and magnetic dyadic green s functions for the surrounding free space , respectively , defined by @xmath41 and where @xmath42 is the corresponding scalar green s function , see _ e.g. _ , @xcite . in cylindrical coordinates , the electric dyadic green s function @xmath39 can be expressed as @xmath43 where @xmath44 is the part of @xmath39 which can be expanded in transverse ( solenoidal ) cylindrical vector waves for @xmath45 , @xmath46 the three - dimensional delta distribution and @xmath47 the corresponding source - point dyadics @xcite . here , the expansion in cylindrical vector waves is given by @xmath48 where the dyadic @xmath49 is defined in in the appendix [ sect : modeexp ] . similarly , the magnetic dyadic green s function @xmath40 is expanded in cylindrical vector waves based on the dyadic @xmath50 defined in . note that @xmath40 does not contain any source - point dyadics at @xmath51 @xcite . to simplify the description below , a non - magnetic material is assumed where @xmath52 and @xmath53 , and only the first equation is needed in . a generalization to the full system will be straightforward . extracting the contribution from the source - point as defined in , the electric field integral equation in can now be expressed as @xmath54\cdot{{\boldsymbole}}({{\boldsymbolr } } ) \\ -k^2\int_{v_{\rm m}}{{\boldsymbolg}}_{\rm e}^0({{\boldsymbolr}},{{\boldsymbolr}}^\prime , k)\cdot{{\boldsymbol\chi}}({{\boldsymbolr}}^\prime)\cdot{{\boldsymbole}}({{\boldsymbolr}}^\prime){\operatorname{d}\!}v^\prime = { { \boldsymbolf}}({{\boldsymbolr } } ) , \end{gathered}\ ] ] where @xmath38 , and the source vector function @xmath55 is given by @xmath56 at each radial distance @xmath57 the material is periodic in the @xmath58-coordinate with period @xmath59 and in the @xmath60-coordinate with period @xmath2 . the material dyadic @xmath61 can hence be represented by the following two - dimensional fourier series @xmath62 where @xmath63 are the corresponding dyadic fourier series coefficients . consider the following fourier representation of the electric field @xmath64 and similarly for the sources @xmath22 and @xmath23 and the source vector function @xmath55 . the material dyadic is furthermore represented as @xmath65 where @xmath66 and where the coefficients @xmath63 have been defined in . based on the convolution theorem the fourier transformation of yields the following integral equation in fourier space @xmath67 where @xmath68 $ ] , @xmath69 and @xmath70 . the fourier transformation of the source vector function is given by @xmath71 by further exploiting the distributional property given in , the integral equation becomes @xmath72 the integral operator defined on the left - hand side of can be continued to the complex @xmath73-plane taking the branch - cut ( defined by the square root @xmath74 ) into proper account , and hence defines an analytic operator - valued function @xmath75 where @xmath76 . the solution to can be expressed as the sum of discrete modes ( residues at poles ) , plus an integration along the branch - cut @xcite . in practice , the branch - cut contribution can often be neglected @xcite . it is assumed that the sources @xmath22 and @xmath23 are supported only for @xmath77 , and hence that the source vector function @xmath78 defined in is an analytic function in the upper half - plane @xmath79 . according to the floquet theorem @xcite , there will be floquet modes in the source - free region @xmath80 , and where the electric field can be expressed as @xmath81 where @xmath82 is the complex valued propagation constant of the floquet mode and where @xmath83 . the fourier transform @xmath84 can hence be written as @xmath85 with poles at @xmath86 . here , @xmath87 is the fourier transform of the left - sided part of @xmath19 which is supported in @xmath77 , and hence @xmath87 is analytic in the upper half - plane @xmath79 . the residue theorem can be stated here as @xmath88 where @xmath89 is a small closed contour enclosing the point @xmath86 , @xmath90 is a function which is analytic inside @xmath89 and @xmath91 denotes the kronecker delta . by applying the residue theorem to , the following integral equation is obtained for @xmath92 @xmath93 where @xmath68 $ ] and @xmath94 are integers . maxwell s equations for the free space with a periodically modulated plane wave excitation can be written @xmath95 where the magnetic source is given by @xmath96 with a @xmath2-periodic part @xmath97 , and similarly for the electric source @xmath98 , and where @xmath82 is the floquet wavenumber with @xmath99 . in the transverse plane the sources are assumed to be constrained to the cross - sectional area @xmath100 with radius @xmath1 . the maxwell s equations are also supplemented with a radiation condition in the transverse plane . following the standard derivation based on vector potentials as in _ e.g. _ , @xcite , the solution to can be written @xmath101 where the integration is over one unit cell @xmath102 $ ] } and @xmath103 , and where @xmath104 and @xmath105 are the periodic electric and magnetic green s dyadics , respectively , defined by @xmath106 and where @xmath107 is the scalar periodic green s function defined in in appendix [ sect : scalarperiodicg ] . the periodic electric green s dyadic can furthermore be factorized as @xmath108 , where @xmath109 is periodic in @xmath110 with period @xmath2 , and similarly for the periodic magnetic green s dyadic @xmath105 . the spectral representation of the electric green s dyadic is given by @xmath111 where @xmath112 and which are based on and in appendix [ sect : modeexp ] . from the definition it follows that @xmath39 can also be represented as @xmath113 where @xmath114 is the scalar two - dimensional green s function defined in in appendix [ sect : scalarperiodicg ] . a comparison of and yields immediately that @xmath115 and where both sides can be extended analytically into the complex @xmath73-plane . based on the poisson summation formula for the scalar periodic green s function given in , an analytic expression for the periodic electric green s dyadic can now be derived as follows @xmath116 finally , based on the periodic electric green s dyadic is given by @xmath117 where @xmath118 similarly , the following spectral representation of the magnetic green s dyadic follows from @xmath119 and hence @xmath120 an analytic expression for the periodic magnetic green s dyadic can then be derived as follows @xmath121 consider now the source - free maxwell s equations as in and ( with @xmath122 ) and with equivalent sources @xmath29 and @xmath30 defined as in , and where the periodic material dyadics are defined as in . for simplicity , a non - magnetic material is assumed here with @xmath52 and @xmath53 . the fields are assumed to have a floquet wavenumber @xmath82 with @xmath99 and can be expressed as @xmath123 with a @xmath2-periodic part given by @xmath124 the general solution then leads to the following integral equation for the electric field @xmath125\cdot{{\boldsymbole}}({{\boldsymbolr } } ) \\ -k^2\int_{s}\int_{0}^{p}{{\boldsymbolg}}_{\rm ep}^0({{\boldsymbolr}},{{\boldsymbolr}}^\prime , k,\beta)\cdot{{\boldsymbol\chi}}({{\boldsymbolr}}^\prime)\cdot{{\boldsymbole}}({{\boldsymbolr}}^\prime){\operatorname{d}\!}s^\prime { \operatorname{d}\!}z^\prime = { { \boldsymbol0 } } , \end{gathered}\ ] ] where the integration is over one unit cell @xmath126 and @xmath103 , and where the contribution from the source - point has been extracted by evaluating the periodic distribution in . by writing @xmath127 , the integral equation can also be written @xmath128\cdot\widetilde{{{\boldsymbole}}}({{\boldsymbolr } } ) \\ -k^2\int_{0}^{a}\int_{0}^{2\pi}\int_{0}^{p}\widetilde{{{\boldsymbolg}}}_{\rm e}^0({{\boldsymbolr}},{{\boldsymbolr}}^\prime , k,\beta)\cdot{{\boldsymbol\chi}}({{\boldsymbolr}}^\prime ) \cdot\widetilde{{{\boldsymbole}}}({{\boldsymbolr}}^\prime)\rho^\prime{\operatorname{d}\!}\rho^\prime{\operatorname{d}\!}\phi^\prime { \operatorname{d}\!}z^\prime \\ = { { \boldsymbol0 } } , \end{gathered}\ ] ] where @xmath103 . here , @xmath129 is defined by and and can hence be written as @xmath130 by employing the convolution theorem for two - dimensional fourier series , the integral equation can finally be written in fourier space exactly as in . assume that the cross - section of the waveguide rotates along the longitudinal direction and that there are @xmath131 radial regions @xmath132 with distinct rotation ( twist ) for @xmath133 . in each region , the period in the azimuthal direction is given by @xmath134 where @xmath135 . it is furthermore assumed that there is a smallest common period @xmath2 in the longitudinal direction such that the longitudinal period @xmath136 in each radial region is given by @xmath137 where @xmath138 , and where the positive ( negative ) sign indicates a left ( right ) handed twist . the twist in each radial region @xmath132 is thus characterized increases with the radius @xmath57 . ] by the twist direction @xmath139 , and the non - zero fourier coefficients @xmath63 defined in can only be found at the following points in the reciprocal ( fourier ) space @xmath140 where @xmath141 and @xmath133 . an example with three distinct ( one left and two right handed ) twist regions is illustrated in fig . [ fig : helicalfig2 ] . ( 50,100 ) ( 100,0)(50,90 ) in fourier space . here , @xmath142 , @xmath143 and @xmath144 denote three twist - regions with @xmath145 , @xmath146 and @xmath147 . the filled bullets correspond to a possible truncation.,title="fig:",width=302 ] ( 230,48)@xmath148 ( 128,90)@xmath149 ( 227,7)@xmath142 ( 227,-3)@xmath143 ( 227,81)@xmath144 the primary interest here is with the twisted modes of three - phase power cables at quasi - static ( 50@xmath0 ) conditions . here , the excitation is governed by a single symmetric component @xcite with positive phase progression , _ i.e. _ , an azimuthal floquet - mode with factor @xmath150 . the corresponding field twist - mode @xmath92 is defined by the excitation point @xmath151 together with the conditions of , and which yields the following set of feasible fourier indices @xmath152 where @xmath141 and @xmath133 . note that this is simply a one - step shift to the right in comparison to the illustration in fig . [ fig : helicalfig2 ] . a useful application of the periodic green s dyadics expressed in is with the calculation of the electromagnetic fields produced by a helical current distribution . as an example , consider the periodic magnetic field produced by a thin helical wire carrying the stationary current @xmath153 under quasi - static conditions . the radius of the helix is @xmath1 and the period is @xmath2 . here , @xmath154 where @xmath155 and @xmath156 is a curve constituting one period of the helix defined in cartesian coordinates @xmath157 as @xmath158 where @xmath159 $ ] and @xmath160 is an off - set parameter of the helix . in cylindrical coordinates it is seen that @xmath161 , @xmath162 and @xmath163 . by using , and for @xmath164 , the expression yields the result @xmath165 where the integration over one period has been performed as @xmath166 a discretization based on the collocation method @xcite is devised as follows . it is assumed that the material region of the helical structure is given by the radial domain @xmath167 where @xmath168 . an @xmath169-point discretization is defined where @xmath170 and @xmath171 denotes the corresponding lagrange basis consisting of linear splines with the interpolation property @xmath172 for @xmath173 , see _ e.g. _ , @xcite . the fourier components of the electric field defined in or is now expanded as @xmath174 where @xmath175 is a vector valued coefficient with constant cylindrical components . let @xmath176 denote the radial vector @xmath177 evaluated at the point @xmath178 for @xmath179 . due to the interpolation property of the lagrange basis it is seen that @xmath180 . the integral equation evaluated at the interpolation points @xmath178 now yields the discrete system @xmath181 where @xmath182 denotes the support region of @xmath171 . the fourier series is truncated using @xmath183 terms based on the assumed twist - modes that are at hand , _ cf . _ , section [ sect : twistmodes ] . a multi - index notation is introduced where @xmath184 , @xmath185 and where @xmath186 . the following definitions are made @xmath187 @xmath188 and @xmath189 where @xmath190 and @xmath173 . the integration in is highly regular with at most some points of discontinuity in either of the terms @xmath191 and @xmath192 . it is therefore convenient to place any possible points of discontinuity of the material function @xmath193 at the grid points @xmath178 of the linear interpolation , and to employ an efficient quadrature rule based on interior points such as the gauss - legendre quadrature @xcite to evaluate the integral in . the system can now be written in the more convenient form @xmath194 which can be readily interpreted in terms of @xmath195 block matrices and with the row and column multi - indices @xmath196 and @xmath197 , respectively . let @xmath198 denote the finite matrix corresponding to the linear system defined in . the propagation constant can then be computed from a numerical residue calculus based on @xmath199 where @xmath200 is the minimum ( modulus ) eigenvalue of the matrix @xmath198 , and where the closed loop @xmath201 is circumscribing the true zero @xmath202 in such a way that there are no other zeros or branch - points of @xmath200 inside the loop . appropriate windowing techniques can be applied directly in the fourier domain to reduce the gibbs phenomena in the truncation of the ( material ) fourier coefficients @xmath63 . as an example is illustrated in fig . [ fig : matfig1 ] a comparison of the truncated fourier series expansions of a square pulse with and without a one - dimensional coefficient smoothing based on the kaiser window @xcite . consider a homogeneous and isotropic cylindrical region with relative permittivity @xmath13 , relative permeability @xmath12 and wavenumber @xmath14 . the solenoidal ( source - free ) cylindrical vector waves are defined here by @xmath203 where @xmath204 is a bessel function or a hankel function of the first kind , each of order @xmath148 , _ cf . _ , @xcite . here , @xmath73 is the longitudinal wavenumber and @xmath74 the transversal wavenumber where the square root is chosen such that @xmath205 and hence @xmath206 . it can be shown by direct calculation that @xmath207 . the following curl properties are thus obtained @xmath208 for @xmath209 , and where @xmath210 denotes the complement of @xmath211 ( @xmath212 and @xmath213 ) . the following notation will be used @xmath214 where @xmath209 , and the vectors @xmath215 are given explicitly in cylindrical coordinates as @xmath216 and where the @xmath217 denotes differentiation with respect to the argument . let the regular and the outgoing ( radiating ) cylindrical vector waves @xmath218 and @xmath219 be defined as in ( [ eq : cylvecdef ] ) by using the regular bessel functions and the hankel functions of the first kind , @xmath220 and @xmath221 , respectively . the electric dyadic green s function defined in can be expanded in cylindrical vector waves as @xmath222 where @xmath223 and @xmath224 denote the vector in @xmath225 having the smallest and largest radial coordinate , respectively , _ i.e. _ , @xmath226 and @xmath227 , _ cf . _ , @xcite . the dagger @xmath228 refers to a sign - shift in the exponentials in the definition , and which can be placed on any of the two vector waves @xmath218 and @xmath219 in the expression , _ cf . _ , @xcite . note that the expression also contains a delta distribution taking the source point into account , _ cf . _ , @xcite . by employing the notation defined in , the electric green s dyadic can now be expressed as @xmath229 where @xmath230 @xmath231 and where the dyadic @xmath49 is given by @xmath232 where the dagger @xmath228 has been placed on the vector with primed coordinates , and where @xmath233 . the magnetic dyadic green s function can be obtained as @xmath234 @xcite , and it follows from and that it can be expanded in cylindrical vector waves as @xmath235 where @xmath236 and where the dyadic @xmath50 is given by @xmath237 the scalar green s function @xmath42 satisfies the inhomogeneous helmholtz equation @xmath238 and can be represented by the fourier integral @xmath239 where the two - dimensional green s function @xmath114 satisfies @xmath240 and where @xmath241 is the two - dimensional transverse laplace operator , @xmath74 ( @xmath206 ) and @xmath242 the transverse delta distribution . the function @xmath114 is given by @xmath243 and which can be expanded as @xmath244 see _ e.g. _ , @xcite and the graf s and gegenbauer s addition theorem @xcite . the periodic green s function @xmath107 for plane wave excitation @xmath245 satisfies @xmath246 together with a radiation condition in the transverse plane , and where @xmath2 is the period and @xmath99 . the periodic green s function can furthermore be factorized as @xmath247 where @xmath248 is periodic in @xmath110 with period @xmath2 . to derive an analytic expression for the periodic green s function @xcite , the following poisson summation formula @xcite can be employed @xmath249 where @xmath82 is real valued and where the fourier transform is defined by @xmath250 the periodic samples of @xmath251 along the @xmath252-coordinate are given by @xmath253 and it follows from the poisson summation formula that @xmath254 it is noted that the first sum to the left in converges only conditionally ( not absolutely ) when @xmath82 is real valued and it diverges when @xmath255 , whereas the sum on the right - hand side of converges exponentially for all @xmath82 with @xmath99 . by using the property which is valid with complex valued wavenumbers @xmath256 , it is readily seen that the expression in the right - hand side of satisfies when @xmath82 is complex valued . a. bostrm , g. kristensson , and s. strm . transformation properties of plane , spherical and cylindrical scalar and vector wave functions . in v. v. varadan , a. lakhtakia , and v. k. varadan , editors , _ field representations and introduction to scattering _ , acoustic , electromagnetic and elastic wave scattering , chapter 4 , pages 165210 . elsevier science publishers , amsterdam , 1991 . u. r. patel , b. gustavsen , and p. triverio . an equivalent surface current approach for the computation of the series impedance of power cables with inclusion of skin and proximity effects . , * 28*(4 ) , 24742482 , 2013 . sven nordebo received the m.s . degree in electrical engineering from the royal institute of technology , stockholm , sweden , in 1989 , and the ph.d . degree in signal processing from lule university of technology , lule , sweden , in 1995 . he was appointed docent in signal processing 1999 . since 2002 he is a professor of signal processing at the department of physics and electrical engineering at the linnaeus university . during 2009 - 2012 he was a guest professor of signal processing at the department of electrical and information technology , lund university . his research interests are in statistical signal processing , estimation theory , electromagnetics , antennas and propagation , inverse problems and imaging , optimization , dispersion modeling and analysis , direct and inverse problems in applied waveguide theory . mats gustafsson received the m.sc . degree in engineering physics 1994 , the ph.d . degree in electromagnetic theory 2000 , was appointed docent 2005 , and professor of electromagnetic theory 2011 , all from lund university , sweden . he co - founded the company phase holographic imaging ab in 2004 . his research interests are in scattering and antenna theory and inverse scattering and imaging with applications in microwave tomography and digital holography . he has written over 65 peer reviewed journal papers and over 85 conference papers . gustafsson received the best antenna poster prize at eucap 2007 , the ieee schelkunoff transactions prize paper award 2010 , and the best antenna theory paper award at eucap 2013 . he serves as an ieee ap - s distinguished lecturer for 2013 - 15 . gerhard kristensson was born 1949 , he received his b.s . degree in mathematics and physics in 1973 , and the ph.d . degree in theoretical physics in 1979 , both from the university of gteborg , sweden . in 1983 he was appointed docent in theoretical physics at the university of gteborg . during 19771984 he held a research position sponsored by the national swedish board for technical development ( stu ) and he was lecturer at the institute of theoretical physics , gteborg from 19801984 . in 19841986 he was a visiting scientist at the applied mathematical sciences group , ames laboratory , iowa state university . he held a docent position at the department of electromagnetic theory , royal institute of technology , stockholm during 19861989 , and in 1989 he was appointed the chair of electromagnetic theory at lund institute of technology , sweden . kristensson is the author of 4 textbooks and the editor of 3 scientific books . he has written 12 chapters in scientific books , and he is the author of over 70 peer reviewed journal papers and over 70 reviewed contributions in conference proceedings . kristensson s major research interests are focused on wave propagation in inhomogeneous media , especially inverse scattering problems . high frequency scattering methods , asymptotic expansions , optical fibers , antenna problems , and mixture formulas are also of interest , as well as radome design problems and homogenization of complex materials . brje nilsson received the m.s . degree in mathematics and physics from gteborg university , sweden , in 1971 , and the ph.d . degree in theoretical physics from the same university in 1980 . he has 13 years industrial experience in acoustics and has been visiting professor in engineering acoustics from 1988 to 1994 at the royal institute of technology , stockholm , sweden . he was appointed docent in engineering acoustics 1994 . since 2005 he is a professor of mathematical physics at the department of mathematics at the linnus university . his main research interest is mathematical modelling of wave phenomena . primarily electromagnetic , acoustic and quantum waves are considered . of special interest are inverse problems and scattering theory in waveguide systems . alexander i. nosich ( m94-sm95-f04 ) was born in kharkiv , ukraine , in 1953 . he received his m.s . , ph.d . , and d.sc . degrees in radio physics from the kharkiv national university in 1975 , 1979 , and 1990 , respectively . since 1979 , he has been with the institute of radio physics and electronics of the national academy of science of ukraine , kharkov , where he is currently a professor and principal scientist . since 2010 , he also heads initiated by him laboratory of micro and nano optics at this institute . since 1992 , he has held a number of guest fellowships and professorships in the eu , japan , singapore , and turkey . his research interests include the method of analytical regularization , propagation and scattering of waves , simulation of lasers , open waveguides , and antennas , and the history of microwaves . nosich was one of the initiators of the international conference series on mathematical methods in electromagnetic theory ( mmet ) held in ukraine in 1990 - 2014 . in 1995 , he organized ieee east ukraine joint chapter , the first in the former ussr . from 2001 to 2003 , he represented ukraine in the european microwave association . since 2008 , he represents ukraine in the european association on antennas and propagation . daniel sjberg received the m.sc . degree in engineering physics and the ph.d . degree in engineering , electromagnetic theory , from lund university , lund , sweden , in 1996 and 2001 , respectively . in 2001 , he joined the electromagnetic theory group , department of electrical and information technology , lund university , where he was appointed docent in electromagnetic theory in 2005 , and is presently a professor and the director of studies of the department of electrical and information technology . his research interests include electromagnetic properties of materials , composite materials , homogenization , periodic structures , numerical methods , radar cross section , wave propagation in complex and nonlinear media , and the inverse scattering problem .

natural modes of helical structures are treated by using the periodic dyadic green s functions in cylindrical coordinates . the formulation leads to an infinite system of one - dimensional integral equations in reciprocal ( fourier ) space . due to the twisted structure of the waveguide together with a quasi - static assumption the set of non - zero coefficients in reciprocal space is sparse and the formulation can therefore be used in a numerical method based on a truncation of the set of coupled integral equations . the periodic dyadic green s functions are furthermore useful in a simple direct calculation of the quasi - static fields generated by thin helical wires . high - voltage power cables , helical waveguides , dispersion relations , open waveguides , volume integral equations .

introduction natural modes with finite sources natural modes with periodic excitation applications numerical method mode expansions of the dyadic greens functions scalar periodic greens function

assemblies of dynamical elements coupled with each other are widely seen in nature . simplified models of such systems , e.g. coupled limit cycle oscillators or chaotic maps , have played important roles not only in modeling such systems realistically , but also in understanding the varieties of possible behavior of systems far from equilibrium . many important concepts , such as pattern formation or spatio - temporal chaos , have been extracted from the detailed studies on such models . the interaction between the elements are usually assumed to be attractive , and of mean - field type in a wide sense ; each element feels the mean amplitude of its neighboring elements , and driven by the difference between its amplitude and the mean amplitude , in such a way that the amplitude differences between the elements decrease . the diffusive coupling is a representative limiting case . each element interacts strongly with its nearest neighbors , so that the amplitude field of the system is always continuous and smooth . it is well known that some diffusively coupled systems of dynamical elements , such as the complex ginzburg - landau equation , exhibit spatio - temporal chaos@xcite . the opposite limiting case is the global coupling , or the mean field coupling in the narrow sense . each element feels the mean field of the entire system , and is thus coupled to all the elements with equal strength . the amplitude field becomes statistically spatially homogeneous , and the notion of space is lost . it is known that systems with global coupling generally show some typical behavior , e.g. clustering and collective chaos@xcite . in ref.@xcite , kuramoto introduced an intermediate system between the above two limiting cases , namely a system of non - locally coupled elements . our subsequent numerical simulations of one - dimensional non - locally coupled systems with various elements revealed that such systems generally exhibit anomalous spatio - temporal chaotic behavior , which can not be seen in the above two limiting cases . in this chaotic regime , the amplitude field becomes fractal , and the spatial correlation of the amplitude field shows power - law behavior on small scales . furthermore , the fractal dimension and the exponent of the spatial correlation vary continuously with the coupling strength . later , we developed a theory that can explain the fractality of the amplitude field and the power - law behavior of the spatial correlation based on a simple multiplicative stochastic model@xcite . such a model is frequently employed in describing the noisy on - off intermittency phenomena@xcite found in many physical systems , and this implies that our system should also exhibit this type of temporal intermittency . induced by the temporal intermittency , our system is also spatially intermittent . in order to study this , we expanded our analysis of the amplitude field into more general @xmath0-th structure functions , and found that the amplitude field shows multi - affinity . we also introduced multi - fractal analysis of the difference field of the original amplitude field , inspired by its seeming similarity to the intermittent energy dissipation field in fluid turbulence@xcite . all our previous studies have been done in one - dimensional systems . however , our previous theory does not require the systems to be one - dimensional , and spatio - temporal chaos with power - law structure functions is also expected in higher dimensions . in this paper , we study a system of non - locally coupled complex ginzburg - landau oscillators in two dimensions for the first time , and investigate its anomalous spatio - temporal chaotic regime . some attention is paid to the multi - scaling properties of the intermittent amplitude and difference fields . as proposed by kuramoto@xcite , the non - local coupling naturally appears in the following plausible situation . consider an assembly of spatially distributed dynamical elements , e.g. cells . each element is assumed to interact indirectly with other elements through some ( e.g. chemical ) substance , which diffuses and decays much faster than the dynamics of the element . such a situation will be described by the following set of equations : @xmath1 where @xmath2 is the amplitude of the element , @xmath3 is the dynamics of the amplitude , and @xmath4 is the concentration of the substance with decay rate @xmath5 and diffusion rate @xmath6 . the substance @xmath4 is generated at a rate proportional to the amplitude @xmath2 , and the amplitude @xmath2 is affected by the substance @xmath4 with a coupling matrix @xmath7 . the parameter @xmath8 determines the ratio of time scale of the elements to that of the substance , and is assumed to be very small . namely , the dynamics of the substance is much faster than that of the elements . now , let us consider the @xmath9 limit and eliminate the dynamics of @xmath4 adiabatically . putting the left - hand side of eq.([eq:02 ] ) to @xmath10 , we can solve the equation for @xmath4 as @xmath11 where @xmath12 is a kernel that satisfies @xmath13 by inserting eq.([eq : a ] ) into eq.([eq:01 ] ) , we obtain the following system of non - locally coupled dynamical elements : @xmath14 the kernel @xmath12 can be solved as @xmath15}{\eta + d |{{\mbox{\boldmath$q$}}}|^2},\ ] ] where @xmath16 is the dimension of the space . when the system is isotropic , the kernel @xmath17 becomes a function of the distance @xmath18 , and is expressed as @xmath19 where @xmath20 is the modified bessel function . the constant @xmath21 gives the inverse of the coupling length , and is calculated as @xmath22 each @xmath23 must satisfy the normalization condition @xmath24 . since we treat a two - dimensional system , we use eq.([eq : kernel2d ] ) for @xmath23 hereafter . as elements , we use complex ginzburg - landau oscillators . they are the simplest limit cycle oscillators that can be derived by the center - manifold reduction technique from generic oscillators near their hopf bifurcation points@xcite . the corresponding non - locally coupled system is given by the following equation for the complex amplitude @xmath25 : @xmath26 where @xmath27 is the coupling strength , @xmath28 are real parameters , and the non - local mean field @xmath29 is given by @xmath30 this is the non - local complex ginzburg - landau equation introduced by kuramoto@xcite as the first concrete example of non - locally coupled systems . in the numerical simulations presented here , we assume the total system to be a square lattice with both sides unit length long . the elements are placed on the lattice sites , and periodic boundary conditions are assumed . we use @xmath31 elements , and fix the coupling length @xmath32 at @xmath33 . the non - local mean field is easily calculated by using the fft technique , since it is simply a convolution of the amplitude field with the kernel ( [ eq : kernel2d ] ) . we fix the parameters @xmath34 and @xmath35 at @xmath36 and @xmath37 respectively . these are the standard values already used in our previous one - dimensional simulations . in fig.[fig:01]-[fig:03 ] , typical snapshots of the real part @xmath38 of the complex variable @xmath39 are shown for three different values of coupling strength @xmath27 . since we obtain similar figures for the imaginary part @xmath40 by symmetry , we use @xmath41 in the following analysis and call it amplitude field . the amplitude field at @xmath42 is continuous and smooth , while at @xmath43 it seems to be discontinuous and disordered , although not completely random . the amplitude field at the intermediate coupling strength @xmath44 looks somewhat more complex and intriguing ; it is composed of intricately convoluted smooth and disordered patches of various length scales . this is the anomalous spatio - temporal chaotic regime that our interest is focused on . = 0.4 = 0.4 = 0.4 let us examine the spatial correlation function first . figures [ fig:04](a)-(c ) show spatial correlation functions @xmath45 corresponding to the amplitude fields shown in figs.[fig:01]-[fig:03 ] . each correlation function is clearly rotationally symmetric due to the isotropy of the system . as the amplitude field becomes disordered , the correlation function becomes steep , and the center of the graph , which corresponds to the self - correlation @xmath46 , becomes peaked . in ref.@xcite , the anomalous spatio - temporal chaotic regime was characterized by the power - law behavior of the spatial correlation function in small distance : @xmath47 where @xmath48 are constants and @xmath49 is a non - integer parameter - dependent exponent . to confirm if this power - law behavior also holds in two dimension , we calculated the radial correlation function @xmath50 ( @xmath51 ) along a straight line in a certain direction or the @xmath52 direction , but results are independent of the direction . ] , and estimated the best fitting parameters @xmath53 and @xmath54 . figure [ fig:05 ] shows @xmath55 vs. @xmath56 $ ] for several values of the coupling strength @xmath27 . for each coupling strength , the experimental data are almost on a straight line , and the power - law behavior is evident . the exponent @xmath49 of the power law varies continuously with the coupling strength . although not shown in the figure , the correlation function @xmath57 is continuous at the origin @xmath58 for @xmath59 , but discontinuous at @xmath60 . there appears a finite gap between the self - correlation @xmath61 and the correlation between the nearest - neighbor elements @xmath62 . this means that the individual motion of the element becomes so violent that the amplitude field is no longer continuous statistically . in ref.@xcite , this transition point was identified with the blowout bifurcation point in the on - off intermittent dynamics of the amplitude difference between nearby elements . thus , the anomalous spatio - temporal chaos in two dimension is also characterized by power - law behavior of the spatial correlation function with a parameter - dependent exponent . = 0.4 = 0.4 in ref.@xcite , we clarified that the scaling behavior of the spatial correlation is a consequence of underlying multiplicative processes of amplitude differences between neighboring elements . we described this process by a multiplicative stochastic model , and related the exponent @xmath49 of the spatial correlation with the fluctuation of the finite - time lyapunov exponent of the element . such a model is essentially identical with those used in describing noisy on - off intermittency@xcite , which indicates that our system also exhibits this type of temporal intermittency . actually , the finite - time lyapunov exponent of the complex ginzburg - landau oscillator can fluctuate between positive and negative values , and neighboring oscillators are subjected to only slightly different non - local mean field . therefore , the conditions for the appearance of noisy on - off intermittency are satisfied in our system . now , let us confirm this in our system numerically . the coupling strength is set at @xmath44 . figure [ fig:06 ] shows typical time sequences of amplitude differences @xmath63 and @xmath64 . the distance between the elements is @xmath65 for @xmath63 , and @xmath66 for @xmath64 , respectively . strong intermittency of the signals is apparent . it can be seen that @xmath64 shows more frequent bursts than @xmath63 , reflecting that @xmath64 is subjected to larger fluctuations than @xmath63 . we can confirm that these intermittent signals are actually noisy on - off intermittent by calculating the laminar length distribution . the laminar phase is defined as a successive duration where the absolute value of the difference does not exceeds a certain threshold . here we choose @xmath67 as the threshold value . figure [ fig:07 ] shows laminar length distributions @xmath68 obtained from @xmath63 and @xmath64 . the characteristic shape of the distribution @xmath68 , i.e. , the power - law dependence on @xmath69 with slope @xmath70 for small @xmath69 , and the exponential shoulder seen in the large @xmath69 region , clearly indicates that the signals are actually noisy on - off intermittent . the shoulder reflects broken scale invariance due to the additive noise . as expected , the shoulder of @xmath64 appears at a smaller value of @xmath69 than that of @xmath63@xcite . = 0.4 = 0.4 the notion of multi - scaling , i.e. , multi - affinity and multi - fractality , have been employed successfully in characterizing complex spatio - temporal behavior of various phenomena , such as velocity and energy dissipation fields in fluid turbulence@xcite , rough interfaces in fractal surface growth@xcite , nematic fluid electro - convective turbulence@xcite , financial data of currency exchange rates@xcite , and even in natural images@xcite . in ref.@xcite , we introduced multi - scaling analysis to our system for the one - dimensional case , inspired by the seeming similarity of the amplitude and difference fields in our system to the velocity and energy dissipation fields in fluid turbulence . here , we attempt the multi - scaling analysis for the two - dimensional case . first , we introduce the difference field @xmath71 as @xmath72 which emphasizes the edges of the original amplitude field @xmath73 . with sufficiently small @xmath74 , and this is important to observe the multi - scaling behavior@xcite . ] this quantity is an analogue of the energy dissipation field in fluid turbulence . figure [ fig:08 ] shows a typical snapshot of the difference field @xmath75 at @xmath44 , corresponding to the amplitude field shown in fig.[fig:02 ] . the intermittency underlying the original amplitude field is now apparent . = 0.4 we then introduce the following quantities as measures for the amplitude field @xmath73 and the difference field @xmath71 : @xmath76 where @xmath51 , and the domain of integration @xmath77 is a square of size @xmath78 placed at @xmath79 . the first quantity is a difference of the amplitude field @xmath73 between two points separated by a distance of @xmath78 , and the second quantity is a volume enclosed by the difference field @xmath71 and the square @xmath80 . according to the multi - fractal formalism , two types of partition functions are defined as @xmath81 where @xmath82 is calculated along a certain straight line in some direction as in the case of the previous spatial correlation function , while @xmath83 is calculated over the whole system . @xmath84 is either the position of the line segment or the position of the square . @xmath85 is the number of line segments of length @xmath78 that are needed to cover the entire line , and @xmath86 is the number of squares of size @xmath78 that are needed to cover the whole system . the function @xmath82 is called structure function in the context of fluid turbulence@xcite . when the measures have scaling properties , the partition functions are expected to scale with @xmath78 as @xmath87 and @xmath88 . furthermore , if these exponents @xmath89 and @xmath90 depend nonlinearly on @xmath0 , the corresponding measures @xmath91 and @xmath92 are called multi - affine and multi - fractal , respectively@xcite . for the one - dimensional case , we already know that the amplitude field is multi - affine and the difference field is multi - fractal moreover , our previous theory predicts the following form for the scaling exponent @xmath89 of the amplitude field : @xmath93 where @xmath94 is a positive constant determined by the fluctuation of the finite - time lyapunov exponent of the element , and is related to the slope of the probability distribution of @xmath91@xcite . this is the simplest form of multi - affinity , and sometimes called bi - fractality@xcite . the same form of the scaling exponent @xmath89 is also expected in two dimension , since our previous theory imposed no restriction on the dimensionality of the system . for the scaling exponent @xmath90 of the difference field , we have not been able to develop a satisfactory theory yet . numerical results in one - dimensional systems suggest that @xmath90 also depends nonlinearly on @xmath0 , and the difference field is multi - fractal with a rather simple functional form for @xmath90@xcite . however , the scaling exponent @xmath90 for the two - dimensional system may be different from that for the one - dimensional case , since @xmath83 is defined depending on the dimensionality of the system , while @xmath82 is always defined along a one - dimensional line . = 0.4 = 0.4 let us proceed to the numerical results now . the coupling strength is fixed at @xmath44 hereafter , where the system is fully in the anomalous spatio - temporal chaotic regime . figure [ fig:09 ] shows the partition function @xmath82 obtained for several values of @xmath0 . each curve depends on @xmath78 in a power - law manner for small @xmath78 , and its exponent increases with @xmath0 . the dependence of the scaling exponent @xmath89 on @xmath0 is shown in fig.[fig:11 ] . the @xmath89 curve is a strongly nonlinear function of @xmath0 , and the multi - scaling property of the amplitude field is evident . furthermore , the @xmath89 curve has a bi - linear form as expected in eq.([eq : bifractal ] ) , although a sharp transition is absent due to the limited number of oscillators . from the large @xmath0 behavior of the exponent , we can roughly estimate the value of @xmath94 as @xmath95 . thus , the amplitude field turns out to be multi - affine , and the behavior of the scaling exponent is the same as that for the one - dimensional case . figure [ fig:10 ] shows the partition function @xmath83 for several values of @xmath0 . it is clear that each curve shows a power - law dependence on @xmath78 . the width of the region where the power law holds seems much wider than the previous case . the scaling exponent @xmath90 is plotted with regard to @xmath0 in fig.[fig:12 ] . the corresponding generalized dimension @xmath96 is also shown in the inset . the @xmath90 curve is again a nonlinear function of @xmath0 , but its dependence on @xmath0 does not seem to be so simple as that of the @xmath89 curve . however , as we conjectured numerically for the one - dimensional case@xcite , asymptotic linearity of the @xmath90 curve seems to hold . correspondingly , the @xmath97 curve seems to saturate to a horizontal line @xmath98 quickly . but we can not observe a clear transition to the horizontal line as in the previous one - dimensional case , which may be due to the limited number of oscillators , or the two - dimensionality of the system . thus , the difference field also turns out to be multi - fractal . the behavior of the scaling exponent somewhat resembles that in the one - dimensional case , but is not completely the same . = 0.4 = 0.4 the multi - scaling properties of amplitude and difference fields are consequences of intermittency underlying the system . in order to analyze this intermittency in more detail , we study here the probability distribution functions ( pdf ) of both measures at each length scale . let us consider the pdfs of the measures @xmath91 and @xmath92 . it is convenient to use rescaled measures @xmath99 , @xmath100 and corresponding rescaled pdfs @xmath101 , @xmath102 . with this rescaling , the peaks and the widths of the pdfs become relatively close . we show our results in these rescaled variables . the scaling exponents @xmath89 and @xmath90 are fully determined by the dependence of these pdfs on the scale @xmath78 . in ref . @xcite , we approximated the tails of the pdfs by certain functional forms and extracted the scaling exponents asymptotically in the small @xmath78 limit . here we present the numerical results only briefly for the purpose of emphasizing the intermittency of the amplitude and difference fields . figure [ fig:13 ] shows the rescaled pdf @xmath101 of the rescaled amplitude difference @xmath103 for several values of @xmath78 in log - log scales . each pdf has a characteristic truncated lvy - like shape , as we already obtained previously for the one - dimensional case ; it is composed of a constant region near the origin , a power - law decay in the middle , and a sharp cutoff . each curve roughly collapses to a scale - invariant curve in the constant and power - law regions , while the cut - off of the tail moves to the right with decreasing of @xmath78 due to the intermittency . more precisely , the cut - off position of the pdf ( defined in some suitable way ) is proportional to @xmath104 . this gives rise to the bi - fractal behavior of the @xmath89 curve , see ref . @xcite for detail . our previous theory predicts that the slope of the power - law decay is given by @xmath105 with the constant @xmath94 from eq . ( [ eq : bifractal ] ) . this is confirmed in fig.[fig:13 ] , where the slope of the power - law decay can be read off as @xmath106 , roughly in agreement with the previously obtained value @xmath107 from the scaling exponent @xmath89 . the inset of fig . [ fig:13 ] shows the pdf of the amplitude difference @xmath91 ( without taking the absolute value ) rescaled by the standard deviation in linear - log scales , in order to further emphasize the intermittency of the amplitude field . the pdf evolves from nearly gaussian into intermittent power - law with the decrease of scale @xmath78 . figure [ fig:14 ] shows the pdfs @xmath102 of the rescaled volume @xmath108 . their shapes are not so simple as those for @xmath109 . they are also qualitatively different from the pdf for the difference field in the previous one - dimensional case@xcite , which is responsible for the slightly different behavior of the scaling exponent @xmath90 from the one - dimensional case . but we can still see that both tails extend , and the distribution widens with the decrease of scale @xmath78 . namely , as we decrease the observation scale @xmath78 , largely deviated events appear more frequently . it is obvious that this intermittency effect gives rise to the nonlinearity of the scaling exponent @xmath90 , although the precise functional form is difficult to obtain . the inset shows the pdf of the measure @xmath92 rescaled by the standard deviation in linear - log scales for several values of @xmath78 . the pdf gradually gets steeper with the decrease of @xmath78 due to the intermittency . thus , the pdfs of the measures reveal the intermittency in our system clearly . especially , the pdf for the amplitude difference @xmath91 has the same shape as that already obtained in the one - dimensional case . = 0.4 = 0.4 we numerically analyzed a two - dimensional system of non - locally coupled complex ginzburg - landau oscillators . as in the previous one - dimensional case , we found an anomalous spatio - temporally chaotic regime characterized by a power - law behavior of the spatial correlation function . as expected from our previous theory , the amplitude difference between neighboring elements exhibits noisy on - off intermittency , giving a microscopic dynamical origin for the power - law spatial correlations . we performed multi - scaling analysis in that regime , and found that the amplitude and difference fields are indeed multi - affine and multi - fractal , indicating strong intermittency underlying the system . by studying the pdfs of the measures at each length scale , the intermittency was clearly observed as scale - dependent deviations of the pdfs in their tails . multi - scaling properties are also known in various phenomena such as turbulence or fractal surface growth . the appearance of similar multi - scaling properties in many different systems implies some underlying common statistical law leading to such behavior . further study of the intermittency in our system will give more insights and hints for the understanding of the multi - scaling properties observed in complex dissipative systems . the author gratefully acknowledges helpful advice and continuous support by prof . yoshiki kuramoto , and thanks dr . axel rossberg for a critical reading of the manuscript . he also thanks the yukawa institute for providing computer resources , and the jsps research fellowships for young scientists for financial support .

a two - dimensional system of non - locally coupled complex ginzburg - landau oscillators is investigated numerically for the first time . as already known for the one - dimensional case , the system exhibits anomalous spatio - temporal chaos characterized by power - law spatial correlations . in this chaotic regime , the amplitude difference between neighboring elements shows temporal noisy on - off intermittency . the system is also spatially intermittent in this regime , which is revealed by multi - scaling analysis ; the amplitude field is multi - affine and the difference field is multi - fractal . correspondingly , the probability distribution function of the measure defined for each field is strongly non - gaussian , showing scale - dependent deviations in the tails due to intermittency . * assemblies of mutually interacting dynamical units are ubiquitous in nature . as models for them , systems of coupled simple dynamical elements , e.g. limit cycle oscillators or chaotic maps , have been studied extensively . recently , a new class of coupled systems , namely , a system with non - locally coupled dynamical elements , was introduced and found to exhibit anomalous spatio - temporal chaos characterized by power - law spatial correlations . the preceding studies on this system have been done in one dimension , but the mechanism for the appearance of this spatio - temporal chaos seems to be universal , and it is also expected in higher dimensions . in this paper , a two - dimensional system of non - locally coupled oscillators is investigated for the first time . as in the one - dimensional case , the system is found to exhibit anomalous spatio - temporal chaos , accompanied by several distinctive features specific to this chaotic regime , i.e. , power - law spatial correlations , noisy on - off intermittency , and multi - scaling properties . *

introduction model anomalous spatio-temporal chaos spatial correlation function noisy on-off intermittency multi-scaling analysis probability distributions of the measures conclusion

an elementary information - theoretic argument shows that @xmath8 bits are necessary to represent integers between 0 and @xmath3 ( @xmath9 denotes binary logarithm throughout the paper ) . counting some interesting objects in a data set thus takes logarithmic space . certain applications need to be more economical because they need to maintain many counters simultaneously while , say , tracking patterns in large data streams . notable examples where memory becomes a limiting factor include analyses of the webgraph @xcite . numerous bioinformatics studies also require space - efficient solutions when searching for recurrent motifs in protein and dna sequences . these frequent sequence motifs are associated with mobile , structural , regulatory or other functional elements , and have been studied since the first molecular sequences became available @xcite . some recent studies have concentrated on patterns involving long oligonucleotides , i.e. , `` words '' of length 1640 over the 4-letter dna alphabet , revealing potentially novel regulatory features @xcite , and general characteristics of copying processes in genome evolution @xcite . hashtable - based indexing techniques @xcite used in homology search and genome assembly procedures also rely on counting in order to identify repeating sequence patterns . in these applications , billions of counters need to be handled , making implementations difficult in mainstream computing environments . the need for many counters is aggravated by the fact that the counted features often have heavy - tailed frequency distributions @xcite , and there is thus no `` typical '' size for individual counters that could guide the memory allocation at the outset . as a numerical example , consider a study @xcite of the 16-mer distribution in the human genome sequence , which has a length surpassing three billion . more than four billion ( @xmath10 ) different words need to be counted , and the counter values span more than sixteen binary magnitudes even though the average 16-mer occurs only once or twice . one way to greatly reduce memory usage is to relax the requirement of exact counts . namely , approximate counting to @xmath3 is possible using @xmath11 bits with probabilistic techniques @xcite . the idea of probabilistic counting was introduced by morris @xcite . in the simplest case , a counter is initialized as @xmath12 . the counter is incremented by one at the occurrence of an event with probability @xmath1 . the counter is meant to track the magnitude of the true number of events . more precisely , after @xmath3 events , the expected value of @xmath13 is exactly @xmath14 . a generalization of the binary morris counter is the so - called _ @xmath15-ary counter _ with some @xmath16 and @xmath17 . in such a setup , the counter is incremented with probability @xmath18 . the actual event count is estimated as @xmath19 , using the transformation @xmath20 the function @xmath21 yields an unbiased estimate , as @xmath22 after @xmath3 probabilistic updates . the accuracy of a probabilistic counting method is characterized by the variance of the estimated count . for the @xmath15-ary counter , @xmath23 which is approximately @xmath24 for large @xmath3 and @xmath25 . the parameter @xmath25 governs the tradeoff between memory usage and accuracy . the counter stores @xmath0 ( with @xmath26 ) using @xmath27 bits ; larger @xmath25 thus increases the accuracy at the expense of higher storage costs . the main goal of this study is to introduce a novel algorithm for approximate counting . our _ floating - point counter _ is defined with the aid of a design parameter @xmath28 , where @xmath5 is a nonnegative integer . as we discuss later , @xmath29 determines the tradeoff between memory usage and accuracy , analogously to parameter @xmath25 of the @xmath15-ary counter . the procedure relies on a uniform random bit generator @xmath30 . algorithm @xmath31 below shows the incrementation procedure for a floating - point counter , initialized with @xmath12 . notice that the first @xmath29 updates are deterministic . 0.9 @xmath32 // _ returns new value of @xmath0 _ set @xmath33 // _ bitwise right shift by @xmath34 positions _ * while * @xmath35 * do * * if * @xmath36 * then * * return * @xmath0 set @xmath37 * return * @xmath38 the counter value @xmath39 , where @xmath40 denotes the lower @xmath34 bits , is used to estimate the actual count @xmath41 . the counter thus stores @xmath0 using @xmath42 bits . the estimate s standard deviation is @xmath43 where @xmath44 fluctuates between about @xmath45 and @xmath46 asymptotically ( see corollary [ cor : accuracy ] for a precise characterization ) . notice that a @xmath15-ary counter with @xmath47 has asymptotically the same memory usage , and a standard deviation of about @xmath48 ( see eq . ) . our algorithm thus has similar memory usage and accuracy as @xmath15-ary counting . the floating - point counter is more advantageous in two aspects . first , the first @xmath29 updates are deterministic , i.e. , small values are exactly represented with convenience . second , the counter can be implemented with a few elementary integer and bitwise operations , whereas a @xmath15-ary counter works with irrational probabilities . the random updates in the floating - point counter occur with exact integer powers @xmath49 , and such random values can be generated using an average of 2 random bits . specifically , the @xmath31 procedure uses an expected number of @xmath50 calls to the random bit generator @xmath30 . in contrast , a @xmath15-ary counter needs a uniform random number in the range @xmath51 to produce a random event with probability @xmath52 . the rest of the paper is organized as follows . in order to quantify the performance of floating - point counters , we found it fruitful to develop a general analysis of probabilistic counting , which is of independent mathematical interest . section [ sec : general.thm ] presents the main results about the accuracy of probabilistic counting methods . first , theorem [ tm : f ] shows that every probabilistic counting method has a unique unbiased estimator @xmath21 with @xmath53 after @xmath3 probabilistic updates . second , theorem [ tm : variance ] shows that the accuracy of any such method is computable directly from the counter value . finally , theorem [ tm : accuracy ] gives relatively simple upper and lower bounds on the asymptotic accuracy of the unbiased estimator . the proofs of the theorems are given in section [ sec : general.proofs ] , which can be safely skipped on first reading . section [ sec : fp ] presents floating - point counters in detail , and mathematically characterizes their utility by relying on the results of section [ sec : general.thm ] . section [ sec : fp ] further illustrates the theoretical analyses with simulation experiments comparing @xmath15-ary and floating - point counters . for a formal discussion of probabilistic counting , consider the markov chain formed by the successive counter values . [ def : counter ] a _ counting chain _ is a markov chain @xmath54 with [ eq : counter.general.def ] @xmath55 where @xmath56 are the transition probabilities defining the counter . it is a classic result associated with probabilities in pure - birth processes @xcite that the _ @xmath3-step probabilities _ @xmath57 are computable by a simple recurrence ( see equations ( [ eq : pcnt.rec.0][eq : pcnt.rec.k ] ) later ) . in case of probabilistic counting , we want to infer @xmath3 from the value of @xmath58 alone through a computable function @xmath21 . a given probabilistic counting method is defined by the transition probabilities and the function @xmath21 . as we will see later ( theorem [ tm : f ] ) , the transition probabilities determine a unique function @xmath21 that gives an unbiased estimate of the update count @xmath3 . [ def : f ] a function @xmath59 is an _ unbiased count estimator _ for a given counting chain if and only if @xmath60 holds for all @xmath61 . in the upcoming discussions , we assume that the probabilistic counting method uses an unbiased count estimator @xmath21 . the merit of a given method is gauged by its accuracy , as defined below . [ def : accuracy ] the _ accuracy _ of the counter is the coefficient of variation @xmath62 the theorems below provide an analytical framework for evaluating probabilistic counters . theorem [ tm : f ] shows that the unbiased estimator is uniquely defined by a relatively simple expression involving the transition probabilities . theorem [ tm : variance ] shows that the uncertainty of the estimate can be determined directly from the counter value . theorem [ tm : accuracy ] gives a practical bound on the asymptotic accuracy of the counter . [ tm : f ] the function [ eq : gen.estimator ] @xmath63 uniquely defines the unbiased count estimator @xmath21 for any given set of transition probabilities @xmath64 . thus , for any given counting chain , we can determine efficiently an unbiased estimator . theorem [ tm : f ] confirms the intuition that the transition probabilities must be exponentially decreasing in order to achieve storage on @xmath65 bits . otherwise , with subexponential @xmath66 , one would have @xmath67 , leading to @xmath68 . the next definition provides a computable function for quantifying the uncertainty of @xmath19 . [ def : g ] the _ variance function _ for a given counting chain is defined by [ eq : g.def ] @xmath69 theorem [ tm : variance ] below shows that the accuracy is computable directly from the counter value for any counting chain . the statement has a practical relevance ( since count estimates can be coupled with the variance function s value ) , and the variance function is used to evaluate the asymptotic accuracy of any counting chain ( see theorem [ tm : accuracy ] ) . [ tm : variance ] the variance function @xmath70 of definition [ def : g ] provides an unbiased estimate for the variance of @xmath21 from theorem [ tm : f ] . specifically , @xmath71 holds for all @xmath72 , where the moments refer to the space of @xmath3-step probabilities . theorem [ tm : accuracy ] is the last main result of this section . the statement relates the asymptotics of the variance function , the unbiased count estimator , and the counting chain s accuracy . [ tm : accuracy ] let @xmath73 be the accuracy of definition [ def : accuracy ] , and let @xmath74 let @xmath75 . suppose that @xmath76 ( and , thus , @xmath77 ) . then [ eq : accuracy.bound ] @xmath78 [ [ example ] ] * example * + + + + + + + + + consider the case of a @xmath15-ary counter , where @xmath79 with some @xmath80 . theorem [ tm : f ] automatically gives the unbiased count estimator @xmath81 theorem [ tm : variance ] yields the variance function @xmath82 in order to use theorem [ tm : accuracy ] , observe that @xmath83 therefore , we obtain the known result @xcite that @xmath84 . in what follows , we use the shorthand notation @xmath85 for the @xmath3-step probabilities . by , @xmath86 , and the recurrences [ eq : pcnt.rec ] @xmath87 hold for all @xmath88 . [ lm : estimator.unique ] the unbiased estimator is unique . since @xmath89 is imposed , and @xmath90 with certainty , @xmath91 . for all @xmath3 , @xmath92 , so @xmath93 thus , for all @xmath94 , @xmath95 which shows that @xmath96 is uniquely determined by @xmath97 and the @xmath3-step probabilities . define the durations @xmath98 , i.e. , the number of times @xmath99 for @xmath100 . define also @xmath101 . clearly , @xmath102 . by the linearity of expectations , @xmath103}}{\mathbb{p}}\{x_n\le k\}\\ * & = { \mathbb{e}}l_k(n ) + { \mathbb{p}}\{x_n\le k\ } { \mathbb{e}}l_k,\end{aligned}\ ] ] where we used the memoryless property of the geometric distribution in the last step . consequently , @xmath104 now , @xmath105 since @xmath106 , we have @xmath60 for all @xmath3 . by lemma [ lm : estimator.unique ] , no other function @xmath21 has the same property . by , for all @xmath72 , @xmath107 since @xmath108 , @xmath109 by and , @xmath110 by , @xmath111 holds for all @xmath72 . since @xmath112 , @xmath113 holds for all @xmath3 . define @xmath114 let @xmath115 be an arbitrary threshold . by the definition of @xmath116 , there exists @xmath117 such that @xmath118 for all @xmath119 . therefore , @xmath120 since @xmath121 for all @xmath122 , @xmath123 for all @xmath122 . consequently , @xmath124 . as @xmath125 , there exists @xmath126 such that @xmath127 since @xmath128 , @xmath129 by , @xmath130 for all @xmath131 . so , @xmath132 since @xmath133 is arbitrarily small and @xmath134 , @xmath135 the lower bound is proven analogously . let @xmath115 be an arbitrary threshold . let @xmath117 be such that @xmath136 for all @xmath119 . so , @xmath137 for @xmath3 large enough , @xmath138 holds . since @xmath133 is arbitrarily small , and @xmath139 , @xmath140 the counting chain for a floating - point counter is defined using a design parameter @xmath28 with some nonnegative integer @xmath34 : [ eq : transz ] @xmath141 , where @xmath142 are the most significant digits and @xmath143 is the number of trailing zeros for the true count . ] figure [ fig : machine ] illustrates the states of the floating - point counter . the counter s designation becomes apparent from examining the binary representation of the counter value @xmath122 . write @xmath144 with @xmath145 i.e. , @xmath40 corresponds to the lower @xmath146 bits of @xmath122 , and @xmath143 corresponds to the remaining upper bits . theorem [ tm : f ] applies with @xmath147 , leading to the following corollary . [ cor : f ] the unbiased estimator for @xmath144 is @xmath148 in other words , @xmath149 is essentially a floating - point representation of the true count @xmath3 , where @xmath143 is the exponent , and @xmath40 is a @xmath5-bit significand without the hidden bit for the leading ` 1 . ' theorem [ tm : variance ] yields the following corollary . [ cor : variance ] the variance function for the floating - point counter is @xmath150 combining corollaries [ cor : f ] and [ cor : variance ] , we get the following bounds . [ cor : accuracy ] the accuracy of the floating - point counter is asymptotically bounded as @xmath151 by equations and , we have @xmath152 considering the extreme values at @xmath153 and @xmath154 , respectively : @xmath155 plugging these limits into theorem [ tm : accuracy ] leads to the corollary . for large @xmath28 , the bounds of corollary [ cor : accuracy ] become @xmath156 the accuracy is thus comparable to the accuracy of a @xmath15-ary counter with @xmath157 , which is approximately @xmath158 . the memory requirements of the two counters are equivalent : in order to count up to @xmath159 , @xmath160 bits are necessary . figures [ fig : trajectory ] and [ fig : estimate ] compare the performance of the floating - point counters with equivalent base-@xmath15 counters in simulation experiments . the equivalence is manifest on figure [ fig : trajectory ] that illustrates the trajectories of the estimates by the different counters . figure [ fig : estimate ] plots statistics about the estimates across multiple experiments : the estimators are clearly unbiased , and the two counters display the same accuracy . -ary counters ( * bottom * ) . each trajectory follows the the appropriate counting chain in a random simulated run . the lines trace the relative error @xmath161 for floating - point counters with @xmath34-bit mantissa , and comparable @xmath15-ary counters with @xmath162 where @xmath163 . the shaded areas indicate a relative error of @xmath164 . the dots at the end of the trajectories denote the final value for @xmath165 . ] -ary counters ( * bottom * ) . each trajectory follows the the appropriate counting chain in a random simulated run . the lines trace the relative error @xmath161 for floating - point counters with @xmath34-bit mantissa , and comparable @xmath15-ary counters with @xmath162 where @xmath163 . the shaded areas indicate a relative error of @xmath164 . the dots at the end of the trajectories denote the final value for @xmath165 . ] -ary counter ( * bottom * ) . each plot depicts the result of 1000 experiments , in which a floating - point counter with @xmath166-bit mantissa , and a @xmath15-ary counter with @xmath167 were run until @xmath168 . the dots in the middle follow the averages ; the black segments depict the standard deviations ( for each @xmath169 , they are of length @xmath169 spaced at @xmath169 from the average ) , and grey dots show outliers that differ by more than @xmath170 from the average . the shading highlights the asymptotic relative accuracy of the @xmath15-ary counter ( @xmath171 ) . ] -ary counter ( * bottom * ) . each plot depicts the result of 1000 experiments , in which a floating - point counter with @xmath166-bit mantissa , and a @xmath15-ary counter with @xmath167 were run until @xmath168 . the dots in the middle follow the averages ; the black segments depict the standard deviations ( for each @xmath169 , they are of length @xmath169 spaced at @xmath169 from the average ) , and grey dots show outliers that differ by more than @xmath170 from the average . the shading highlights the asymptotic relative accuracy of the @xmath15-ary counter ( @xmath171 ) . ] i am very grateful to philippe flajolet for valuable suggestions on improving previous versions of the manuscript . mikls csrs , laurent no , and gregory kucherov , _ reconsidering the significance of genomic word frequencies _ , trends in genetics * 23 * ( 2007 ) , no . 11 , 543546 , preprint available as q - bio/0609022 at arxiv.org . isidore rigoutsos , tien huynh , kevin miranda , aristotelis tsirigos , alice mchardy , and daniel platt , _ short blocks from the noncoding parts of the human genome have instances within nearly all known genes and relate to biological processes _ , proceedings of the national academy of sciences of the usa * 103 * ( 2006 ) , no . 17 , 66056610 . suzanne s. sindi , brian r. hunt , and james a. yorke , _ duplication count distributions in dna sequences _ , physical review e ( statistical , nonlinear , and soft matter physics ) * 78 * ( 2008 ) , no . 6 , 061912

memory becomes a limiting factor in contemporary applications , such as analyses of the webgraph and molecular sequences , when many objects need to be counted simultaneously . robert morris [ _ communications of the acm _ , 21:840842 , 1978 ] proposed a probabilistic technique for approximate counting that is extremely space - efficient . the basic idea is to increment a counter containing the value @xmath0 with probability @xmath1 . as a result , the counter contains an approximation of @xmath2 after @xmath3 probabilistic updates stored in @xmath4 bits . here we revisit the original idea of morris , and introduce a binary floating - point counter that uses a @xmath5-bit significand in conjunction with a binary exponent . the counter yields a simple formula for an unbiased estimation of @xmath3 with a standard deviation of about @xmath6 , and uses @xmath7 bits . we analyze the floating - point counter s performance in a general framework that applies to any probabilistic counter , and derive practical formulas to assess its accuracy .

introduction probabilistic counting proofs floating-point counters acknowledgment

`` mass segregation '' describes the effect that high - mass stars are preferentially found in the center of clusters , while the outskirts are preferentially occupied by lower - mass stars . ( primordial ) mass segregation is studied observationally in a number of clusters . some examples are : the orion nebula cluster ( @xcite ) , 6 lmc clusters ( @xcite ) . dynamical mass segregation is also found from n - body simulations ( e.g. @xcite ) , caused by two - body encounters and energy equipartition . a number of authors also point at the preferential mass loss of low - mass stars when the cluster evolves in a tidal field ( e.g. @xcite ) , as mass segregation populates the cluster outskirts preferentially with low - mass stars where they are most easily stripped from the cluster potential . the resulting changes in the overall stellar mass function inside the cluster are quantified by @xcite . in we use the results from @xcite to incorporate the changing mass function slope in a simplified manner into the galev code ( see and references therein ) to calculate its impact on the integrated photometry of star clusters . our main findings are : * at 0 40 per cent of the cluster s lifetime : the cluster colours are comparable to standard models ( i.e. without preferential loss of low - mass stars ) * at 40 80 per cent of the cluster s lifetime : a cluster appears too blue / young ( compared to standard models ) due to the loss of lower main - sequence stars * at 80 100 per cent of the cluster s lifetime : a cluster appears too red / old ( compared to standard models ) due to the loss of main sequence turn - off stars * when interpreting photometry of mass - segregated clusters , that have preferentially lost low - mass stars , with standard photometric models ( without taking mass - segregation effects into account ) the derived ages can be wrong by 0.3 0.5 dex here we propose a benchmark test for comparison of n - body codes : 1024 equal - mass particles , plummer sphere model , without primordial binaries , external tidal field or stellar / binary evolution . we performed simulations using the same input files ( 10 individual runs ) for nbody4 ( see e.g. @xcite ) and starlab ( see e.g. @xcite ) , following the setup from the proposed benchmark test . we find stochastic effects to be of importance for 10 runs , hence we supplemented the starlab results with 20 additional runs ( equivalent nbody4 runs are in preparation ) . some results are shown in fig . [ fig1 ] . for most parameters studied the results are virtually indistinguishable . some questions remain for the kinetic energy and the parameter distributions of dynamically formed binaries . any differences there could be originating from the different treatment of binaries , or due to stochastical effects . further tests ( e.g. with primordial binaries ) are in preparation .

two aspects of our recent n - body studies of star clusters are presented : + 1 ) what impact does mass segregation and selective mass loss have on integrated photometry ? + 2 ) how well compare results from n - body simulations using nbody4 and starlab / kira ?

selective mass loss and integrated photometry benchmark test for n-body codes: comparing nbody4 with starlab/kira

the observation of a metal - insulator transition in high - mobility two - dimensional electron systems at zero magnetic field has sparked renewed interest in this type of transition @xcite . in these systems , electron - electron interactions represent the largest energy scale in the problem @xcite . further evidence for the crucial role of electronic correlations has come from complementary experiments @xcite reporting a substantial mass enhancement close to the metal - insulator transition . taken together these experimental results stress the importance of the ill - understood effects of disorder in strongly correlated electronic systems @xcite . in one of the first studies of its kind , tanaskovi _ et al . _ have investigated the interplay of strong correlations and disorder using a dynamical mean field theory ( dmft ) @xcite approach @xcite . the dmft approach to disordered systems treats correlations on a local level by solving the embedded - atom strongly correlated anderson impurity problem in the self - consistently determined _ fixed _ bath of the other electrons . one is thus forced to consider an _ ensemble _ of single - impurity actions , one for each lattice site . remarkably , tanaskovi _ et al . _ found that very strong site disorder screening emerges precisely in the vicinity of the mott metal - insulator transition @xcite . this effect can be traced back to the pinning of the single - impurity kondo resonances to the fermi level , which acts to suppress the effective randomness . motivated by this striking result , we extended their work to finite dimensions__. _ _ to treat both strong correlations and disorder in a non - perturbative fashion at @xmath0 , we use a generalization of the dmft , the statistical dmft ( _ _ statdmft _ _ ) @xcite , implemented using a slave boson impurity solver @xcite . this approach is equivalent to the description of the effects of disorder through a gutzwiller - type wave function @xcite , which in the clean case realizes the brinkman - rice scenario of the mott transition @xcite . the statdmft method retains the local treatment of electronic correlations . however , in contrast to the infinite - dimensional dmft approach , here each strongly correlated site sees a _ different bath of electrons _ , reflecting the strong spatial fluctuations of their immediate environment . we find that , concomitant to a strong pinning effect , additional effective disorder is also generated through an increasingly broader distribution of quasiparticle weights . their interplay leads to a non - trivial landscape in energy space : the proximity to the mott transition acts to suppress density of states fluctuations close to the fermi level , while at the same time _ enhancing _ them at higher energies . we focus on the disordered hubbard model @xmath1 where @xmath2@xmath3 is the creation ( annihilation ) operator of an electron with spin projection @xmath4 on site @xmath5 , @xmath6 is the nearest - neighbor hopping amplitude , @xmath7 is the on - site hubbard repulsion , @xmath8 is the number operator , and the site energies @xmath9 are uniformly distributed in the interval @xmath10 $ ] , where @xmath11 is the disorder strength . we work at half filling ( chemical potential @xmath12 ) on an @xmath13 square lattice with periodic boundary conditions . all energies will be expressed in units of the clean fermi energy ( the half - bandwidth @xmath14 ) @xmath15 . we treat the hamiltonian of eq . [ eq : hubbard_disordered ] in its paramagnetic phase within the statdmft @xcite . this theory is exact in the non - interacting limit and reduces to the standard dmft in the absence of disorder . unlike the dmft , however , it incorporates anderson localization effects . we start by writing an effective action ( in imaginary time ) for a given site @xmath5 , with the simplification that we neglect all non - quadratic terms in the local fermionic operators except for the local @xmath7-term@xmath16 the site @xmath5 is connected with the rest of the lattice through the bath ( or `` cavity '' ) function @xmath17 , which in statdmft ( but in contrast to dmft ) varies from site to site and thus exhibits strong spatial fluctuations . the effective action in eq . [ eq : action_stat_dmft ] is precisely the action of an anderson impurity model @xcite embedded in a sea of conduction electrons described by @xmath17 . therefore , this approach maps the original hubbard hamiltonian in eq . [ eq : hubbard_disordered ] onto an _ ensemble _ of single - impurity anderson hamiltonians @xcite . the local @xmath5-site green s function , calculated under the dynamics dictated by the effective action in eq . [ eq : action_stat_dmft ] , can be written as ( @xmath18 is a matsubara frequency ) @xmath19 which also serves as a definition of @xmath20 , the @xmath5-site self - energy . it is important to point out that within statdmft the electronic self - energy @xmath20 is still local , albeit site - dependent . the bath function @xmath21 can be viewed as the weiss field of this mean field theory , here elevated to a full function of frequency or time . it is determined through a self - consistency condition that demands that the green s function @xmath22 obtained from the effective action in eq . [ eq : action_stat_dmft ] be equal to the diagonal ( local ) part of the full lattice green s function @xmath23_{ii},\label{eq : g_lattice}\ ] ] where @xmath24 and @xmath25 are site - diagonal matrices @xmath26_{ij}=\sigma_{i}\left(i\omega\right)\delta_{ij}$ ] , @xmath27_{ij}=\varepsilon_{i}\delta_{ij}$ ] and @xmath28 is the clean ( @xmath29 ) and non - interacting ( @xmath30 ) lattice hamiltonian . in general , this step involves the inversion of the frequency - dependent matrix within brackets in eq . ( [ eq : g_lattice ] ) . it is worthwhile to point out that the statdmft approach requires a massive numerical effort since it is necessary to solve a single impurity problem _ for every lattice site _ as well to perform the inversion implied by the self - consistency condition . on the other hand , it provides access to entire distribution functions and accounts for spatial correlations between local quantities . to find @xmath20 , we need to solve the auxiliary single impurity problems for a given set of @xmath21 . for this task , we have used the four - boson mean - field theory of kotliar and ruckenstein @xcite at @xmath0 , which is equivalent to the well - known gutzwiller variational approximation . in practice , we need to solve a pair of non - linear equations for the site - dependent kotliar - ruckenstein slave boson amplitudes @xmath31 and @xmath32 @xcite @xmath33 where @xmath34 and @xmath35 eqs . ( [ eq : sb1]-[eq : sb2 ] ) involve the local @xmath5-site green s function @xmath22 . in the kotliar - ruckenstein theory , the @xmath5-site self - energy @xmath20 is given by @xmath36 which , when plugged into eq . ( [ eq : g_loc_imp ] ) yields@xmath37 it is clear that @xmath38 has the physical interpretation of a quasi - particle weight ( wave function renormalization ) . it also renormalizes the hybridization function , thus setting the local kondo temperature @xcite . furthermore , from eq . ( [ eq : self_energy_sb4 ] ) we see that @xmath39 can be viewed as a renormalized on - site disorder potential @xmath40 we numerically solved the disordered hubbard model with statdmft , using the kotliar - ruckenstein theory as the impurity solver , for several lattice sizes up to @xmath41 . for every @xmath42 pair we typically generated around forty realizations of disorder . we carefully verified that for such large lattices , all our results are robust and essentially independent of the system size ( see , e.g. , the inset of fig . [ fig : p(v ) ] ) . strength of the renormalized site energy disorder , as given by the standard deviation @xmath43 of @xmath44 , normalized by its non - interacting value ( see text for definition of @xmath45 ) . close to the mott transition , the disorder screening remains strong even for moderate values of disorder . results are shown for @xmath46 . in the inset we show the typical ( @xmath47 ) and average ( @xmath48 ) values of the local quasiparticle weight @xmath38 as functions of the interaction @xmath7 . the mott transition is identified by the vanishing of @xmath47 . we note that @xmath48 is finite at @xmath49 indicating that a fraction of the sites remains nearly empty or doubly occupied . results are shown for @xmath46 and @xmath50 . ] according to the scaling theory of localization , any amount of disorder drives a system with dimension equal to or smaller than two to an insulating phase @xcite . however , this result was obtained in the absence of electron - electron interactions . ln the last few years , strong numerical evidence has been obtained indicating that interactions can act to enhance the conducting properties of two - dimensional electronic systems @xcite . moreover , the stability of a @xmath51 metal with respect to weak - localization corrections has been investigated by punnoose and finkelstein @xcite in very careful recent work . these authors have demonstrated that any @xmath51 metal remains stable with respect to sufficiently weak disorder due to additional ( anti - localizing ) interaction corrections . note , however , that both the weak - localization and the corresponding interaction corrections are manifested only through a very weak , logarithmic dependence on the system size . such subtle finite size effects are not visible for the very weak ( renormalized ) disorder we deal with in this work . nevertheless , based on the very convincing considerations of punnoose and finkelstein , the stability of a @xmath51 metal is beyond immediate doubt and these issues are not of relevance for the questions we focus on in this paper . the low energy behavior on the metallic side of the transition is characterized by the distribution of the local quasiparticle weight @xmath38 . following previous studies of the disordered mott transition @xcite , we choose the typical value of @xmath38 , here defined through the geometrical average @xmath52 , as the order parameter of the transition . this quantity vanishes linearly at a critical value of interaction @xmath53 at which the mott transition takes place , marking the transmutation of most itinerant electrons into local magnetic moments ( see the inset of fig . [ fig : screening ] ) . because random site energies tend to push the local occupation away from half filling , @xmath54 is an increasing function of the disorder strength @xmath11 @xcite . we also stress that the average value of the quasiparticle weight @xmath48 is small yet finite at the mott transition ( inset of fig . [ fig : screening ] ) , indicating that some sites remain either empty ( @xmath55 ) ou doubly occupied ( @xmath56 ) and do not give rise to localized magnetic moments . real ( dashed lines ) and imaginary ( full lines ) parts of the hybridization function @xmath57 . although the average value of @xmath57 is particle - hole symmetric , for an arbitrary site @xmath57 is locally particle - hole asymmetric . results are shown for @xmath46 and @xmath58 . ] using the fact that in the current statdmft approach the lattice problem is mapped onto an _ ensemble _ of auxiliary anderson impurity problems , we can characterize the approach to the mott transition by a steady decrease of the local kondo temperature @xmath59 @xmath60 . moreover , the renormalized disorder potential @xmath39 can be thought of as the position of the local kondo resonance energy . in the dmft limit @xmath61 , each site has many neighbors and thus is embedded in the same ( self - averaged ) environment described by @xmath62 . in this regime , as long as @xmath62 is particle - hole symmetric , we find _ perfect _ disorder screening close to the mott transition @xcite . this happens because , as we increase @xmath7 towards @xmath63 , we approach the kondo limit , @xmath64 , causing @xmath39 ( the kondo resonance ) to be `` pinned '' to the fermi energy @xcite . however , within the _ _ statdmft _ _ approach , @xmath57 fluctuates strongly from site to site and is not locally particle - hole symmetric ( see fig . [ fig : hibfxw ] ) . for this reason , we expect that @xmath39 will also have a contribution which is proportional to @xmath65 $ ] . therefore , we have no guarantee that a similar mechanism of disorder screening will persist close to the critical point in @xmath51 . surprisingly , however , for small and moderate disorder @xmath66 , we do get a very strong disorder screening close to the mott metal - insulator transition ( see figs . [ fig : screening ] and [ fig : p(v ) ] ) . probability distribution function @xmath44 of the renormalized site energy @xmath45 . as we move towards the mott transition , disorder screening takes place and @xmath44 becomes increasingly narrower . results are shown for @xmath41 and @xmath50 . the inset illustrates how for such large lattices our results for @xmath44 are essentially independent of the system size : results are shown for @xmath46 , full line , and @xmath41 , dashed line at @xmath50 and @xmath67 . ] spatial distribution of the of the local susceptibility @xmath68 normalized by its typical value @xmath69 . from bottom to the top we have @xmath70 . the color scale is logarithmic in order to stress that , as we approach the critical point , we have the formation of regions in which @xmath71 ( localized magnetic moments ) and @xmath72 ( anderson insulator droplets ) . results are shown for @xmath50 and @xmath41 . ] the site - to - site fluctuations in the hybridization function encode _ spatial correlations _ between the physical quantities , implying , for example , that the value of the local quasi - particle weight at a given site depends on the quasi - particle weight values at the neighboring sites . therefore , due to rare disorder configurations , it is possible to find regions containing sites in which @xmath73 ( red regions in fig . [ fig : zz_spatial ] ) . since the approach to the mott insulator corresponds to @xmath74 , such regions with @xmath75 should be recognized as `` almost localized '' mott droplets ( consisting of localized magnetic moments @xcite ) inside the strongly correlated metallic host . within our brinkman - rice picture , each local region provides @xcite a contribution @xmath76 to the spin susceptibility or the sommerfeld coefficient , respectively . the local regions with the smallest @xmath38 thus dominate the thermodynamic response and will ultimately give rise to an electronic griffiths phase in the vicinity of the disordered mott metal - insulator transition @xcite . the mott droplets have a direct influence on the probability distribution function @xmath77 , since they give rise to a low-@xmath78 tail , contributing to the fact that @xmath77 actually broadens as we approach the mott metal - insulator transition for a _ fixed _ @xmath11 , as shown in figs . [ fig : zz_spatial ] and [ fig : p(z_typ ) ] . surely , there is also a contribution to the broadening of @xmath77 coming from the high-@xmath78 tail originating from those sites with @xmath79 . such sites retain a finite @xmath38 at the transition and are nearly empty or doubly occupied , giving rise to anderson insulating regions ( blue regions in fig . [ fig : zz_spatial ] ) . the coexistence of localized magnetic moments , and nearly empty or doubly occupied sites close to the critical point is characteristic of a two - fluid behavior @xcite . nevertheless , this interesting phenomenon is not relevant for the present analysis , since the major contribution to the energy resolved inhomogeneities , discussed in the next section , as well as to the thermodynamic response , comes from the mott droplets . distribution of the quasiparticle weight divided by its typical value for @xmath50 and @xmath80 . as we approach the critical point , @xmath77 becomes increasingly broader , even though the bare disorder strength @xmath11 is kept fixed . results are shown for @xmath46 . ] from the above discussion , it is clear that the behaviors of @xmath44 and @xmath77 near the mott metal - insulator transition are quite distinct . while the former exhibits strong disorder screening ( fig . [ fig : p(v ) ] ) , the latter suggests that the disorder is actually increasing ( fig . [ fig : p(z_typ ) ] ) . this dichotomy gives rise to an _ energy - dependent effective disorder _ , which manifests itself , for example , in the spatial structure of the local density of states . the local density of states is defined as @xmath81 $ ] , where the lattice green s function was given in eq . [ eq : g_lattice ] , and has the following expression within our brinkman - rice picture @xmath82_{ii},\label{eq : g_lattice_sb4}\ ] ] where @xmath83 and @xmath84 are site - diagonal matrices such that @xmath85_{ij}=z_{i}\delta_{ij},\,\left[\mathbf{v}\right]_{ij}=v_{i}\delta_{ij}$ ] . therefore , the frequency - dependent effective disorder potential `` seen '' by the quasiparticles at energy @xmath86 can be defined as @xmath87 spatial distribution of the local density of states normalized by its clean and non - interacting value for one given realization of disorder and two distinct values of interaction : _ ( a ) _ @xmath88 and _ ( b ) _ @xmath89 , where @xmath90 . we have @xmath91 in the bottom figures and @xmath92 in the top ones . at the fermi energy , the local density of states distribution becomes homogeneous as we approach the mott transition . conversely , if we move even slightly away from the fermi energy , the distribution becomes in fact more inhomogeneous close to @xmath49 . results are shown for @xmath41 , @xmath50 . ] in our brinkman - rice scenario of the mott transition , the quasi - particle bandwidth is reduced by @xmath47 as @xmath93 . to monitor the behavior within the quasi - particle band , we therefore introduce a rescaled frequency @xmath94 , which we will keep constant as we approach the transition . in fig . [ fig : ldos_spatial ] , we show topographic maps of the local density of states for one specific realization of disorder . because of strong disorder screening , @xmath95 close to the critical point . thus , if the system is examined at the fermi energy ( @xmath91 ) , it becomes more and more homogeneous as the transition is approached . at higher energies ( @xmath96 ) , however , the fluctuations in @xmath97 come into play . since they are very pronounced close to the mott transition , we instead find a strong enhancement of the spatial inhomogeneity . this result is surprisingly reminiscent of recent spectroscopic images on doped cuprates @xcite . our theory , which focuses on local ( kondo - like ) effects of strong correlations ( while neglecting inter - site magnetic correlations ) and does not include any physics associated with superconducting pairing , strongly suggests that such energy - resolved inhomogeneity is a robust and general feature of disordered mott systems . indeed , recent results corroborate this picture of an energy - resolved strong - correlation driven disorder screening @xcite . we have discussed the results of a brinkman - rice approach to the disordered mott transition . a striking feature that is apparent in this scenario is the different behaviors of the effective site disorder and the quasiparticle weights as the transition is approached . whereas randomness in the former is suppressed , the latter becomes extremely singular and broad on the way to the mott insulator . the end result of this dichotomy is a non - trivial energy - space landscape , which is signaled by an energy - dependent inhomogeneity . this work was supported by fapesp through grants 04/12098 - 6 ( eca ) and 07/57630 - 5 ( em ) , capes through grant 1455/07 - 9 ( eca ) , cnpq through grant 305227/2007 - 6 ( em ) , and by nsf through grant dmr-0542026 ( vd ) .

we investigate the effects of weak to moderate disorder on the @xmath0 mott metal - insulator transition in two dimensions . our model calculations demonstrate that the electronic states close to the fermi energy become more spatially homogeneous in the critical region . remarkably , the higher energy states show the opposite behavior : they display enhanced spatial inhomogeneity precisely in the close vicinity to the mott transition . we suggest that such energy - resolved disorder screening is a generic property of disordered mott systems .

introduction strongly correlated theory the disordered mott transition particle-hole symmetry and strong disorder screening mott droplets energy-resolved inhomogeneities conclusions acknowledgments

single electron devices made up of a confined nanostructure located in the vicinity of electrodes that act as electron baths constitute a viable alternative to traditional field effect transistors which are about to reach their physical minimum size in a few years @xcite . therefore , detailed study of these devices has accelerated recently aided by the tremendeous advances in nanotechnology during the past two decades . to this end , it is of critical importance to simulate the switching characteristics of these prototypes in the presence of realistic electron - electron and electron - phonon interactions to assess their suitability for large electrical circuits . early studies largely focused on the influence of the many - body kondo resonance on the transient electrical conductance and discovered that varios time scales @xcite emerge after sudden switching of the gate or bias voltage @xcite . these timescales are inherently related to the spin fluctuations that govern the amount of time required for full formation of the kondo resonance @xcite . moreover , it turns out that the van hove singularities in the electrode density of states can induce dramatic effects in transient conductance . damped oscillations arising from interference between the emerging kondo resonance and the singularities in the contact density of states have been initially reported using prototypical density of states for an asymmetrically coupled system @xcite . subsequent studies undertaken to make the theoretical predictions more realistic used the electrode density of states calculated via ab initio methods as an input in many body calculations and exposed beating phenomena in transient electrical conductance @xcite . kondo effect is probably the most elegant and exotic example of many body physics whose signature has been originally traced in magnetic impurities embedded in bulk metals as resistance enhancement at low temperatures @xcite . it took some four decades to unravel its existence in confined nanostructures like quantum dots @xcite and molecules @xcite . the underlying physical mechanism for the kondo effect is the spin flip processes taking place at low temperatures between an isolated spin at dot energy level and the fermi sea of electrons at electrodes . when the dot energy level lies below the fermi level , a sharp resonance develops around the fermi level and it causes an enhancement in current due to the opening of an extra transport channel . the linewidth of this resonance is on the order of an energy scale called the kondo temperature @xmath0 given by @xmath1 where @xmath2 is the half bandwidth of the conduction electrons and @xmath3 is the total coupling of the dot to the electrodes . in this paper , we will take @xmath4 . introducing a temperature gradient across the electrodes provides another intriguing and realistic path for investigation . the existence of kondo resonance can be checked directly via a measurement of the thermopower ( seebeck coefficient ) @xmath5 and the thermoelectric figure of merit @xmath6 which is directly related to the thermopower is a key indicator of the device s performance . some experiments in molecular junctions have recently probed this by adjusting the alignment of the impurity orbital with respect to the fermi level of the contacts and this has been found to result in a change of sign in thermopower @xcite . behaviour of the thermopower and figure of merit has been elucidated as a function of temperature both in kondo @xcite and mixed valence regimes @xcite . these studies were largely in line with the experimental work and confirmed the critical role the impurity energy level plays in the sign of thermopower . nevertheless , they were by no means sufficient to shed light on the thermal switching behaviour of these devices since they were performed in steady state . thermal switching will be required in practical applications to be able to operate the device at different thermoelectric figure of merit values . therefore , it is imperative to understand the behaviour of the thermal switch in transient region . first analysis of the thermopower in a time - dependent situation showed that the the inverse of the saturated decay time of thermopower to its steady state value is equal to the kondo temperature when kondo resonance emerges as a result of the sudden motion of the dot level to a position nearby the fermi level of the electrodes @xcite . later , ac driving of the dot level sinusoidally revealed complex fluctuations in instantaneous thermopower whose extreme values are quite sensitive to the amplitude of the oscillation @xcite . these studies utilized parabolic density of states for contacts and the effect of singularities on the instantaneous thermopower and figure of merit in kondo regime remains to be seen . it is our aim in this paper to fill this gap and investigate the behaviour of instantaneous thermopower and figure of merit for a quantum dot whose energy level is shifted to a position close to the fermi level of contacts . this gives rise to the gradual appearance of the kondo resonance slightly above the fermi level . we will assume that the quantum dot energy levels are filled completely until its highest occupied molecular ( homo ) level with two electrons of opposite spins . we will represent the homo level with a single discrete spin degenerate energy level @xmath7 coupled to electrodes which are free electron baths . this physical system can be mapped to the anderson hamiltonian which can be written as @xmath8 , \label{anderson}\end{aligned}\ ] ] where @xmath9 ( @xmath10 ) and @xmath11 ( @xmath12 ) with @xmath13=l , r create ( destroy ) an electron of spin @xmath14 in the dot energylevel and in the left(l ) and right(r ) electrodes respectively . the @xmath15 and @xmath16 are occupancies of the dot energy level and the electrode @xmath13 . @xmath17 are the tunneling amplitudes between the electrode @xmath13 and the quantum dot . this model is an ideal venue to study kondo correlations since it can be mapped directly to kondo model via schrieffer - wolf transformations . the coulomb repulsion energy @xmath18 in this model represents the charging energy of the dot . this corresponds to the energy gap between the homo level and the lowest unoccupied molecular ( lumo ) level . we take the charging energy to be infinite . this is in line with the experiments which indicate that the typical charging energy of quantum dots overwhelm the other tunable parameters in the kondo regime . this choice results in single occupancy of the dot energy level with up or down spin . moreover , we will be using atomic units where we take @xmath19 . we redefine the electron operators on the dot by introducing a massless ( slave ) boson operator and a pseudofermion operator @xcite as @xmath20 these new operators must also satisfy @xmath21 which ensures one to one mapping between the anderson hamiltonian and the transformed hamiltonian in @xmath22 limit by limiting the dot energy level occupancy to single electron . we can safely omit the hubbard term from the transformed hamiltonian which turns out to be @xmath23 , \label{slave}\end{aligned}\ ] ] where @xmath24 and @xmath25 with @xmath13=l , r are creation(destruction ) operators for an electron with spin @xmath14 and a slave boson on the dot energy level . taking time independent hopping matrix elements , the total coupling of the quantum dot to the electrodes is given by @xmath26 , where @xmath27 is a constant given by @xmath28 and @xmath29 is the density of states function of the electrodes . both left and right electrodes are taken to be made up of the same material in conjunction with experiments implying that @xmath30=@xmath31=@xmath32 consequently , we obtain @xmath33 where @xmath34 and @xmath35 . graphene is a a highly resilient and versatile material which has been manufactured in large quantities quite recently thanks to the scotch tape method @xcite . its unique features largely stem from the fact that it can be considered both a semiconductor with zero bandgap and a metal whose density of states is zero at the fermi level @xcite . the first type of contact we choose in fig . [ schematic ] mimics the density of states of graphene @xmath32 except that we added an infinitesimal thick cylindirical feature inside the conic structure which results in nonzero value at the fermi level to enable the proper development of the kondo resonance . in this case , the van hove singularity lies at the bottom of the band . since we assume that the dot is weakly coupled to the electrodes , the band cutoff at @xmath4 lies at a sufficiently low energy . this enables us to safely ignore deviations from the linear dispersion relation at high energies and the dirac cone approximation is still valid . the second type of contact depicted in fig . [ schematic ] is a prototypical metal and of triangular shape . once again we added a cylindirical feature with infinitesimal thickness inside the triangle to enable nonzero value at the fermi level . the van hove singularity lies in the middle of the band in this latter case . in both cases , we showed only the valence bands in fig . [ schematic ] for simplicity . the conduction bands are just the mirror image of valence bands across the fermi level . furthermore , we made sure the number of electrons is the same in both cases by integrating the area from the bottom of the band to the top . the latter one is a necessary step to facilitate direct comparison between the two structures . in particular , the density of states function for graphene in fig . [ schematic]a is given by @xmath36 where the prefactor corresponds to the inverse of the fermi velocity in atomic units and @xmath37 is an infinitesimal number . on the other hand , the density of states function for the metal shown in fig . [ schematic]b is @xmath38 our technique to tackle the slave boson hamiltonian in eq . ( [ slave ] ) relies on the double time kadanoff - baym green functions which have been adapted to the quantum impurity problems @xcite . in particular , the retarded green function is given by @xmath39 where the curly brackets denote anticommutation . after performing the slave boson transformation described above , the retarded green function can be redefined in terms of the slave boson and pseudofermion green functions @xcite as @xmath40.\end{aligned}\ ] ] of contacts to which the quantum dot is attached . since both left and right electrodes are identical , only one of them is shown . , width=355,height=257 ] the task then boils down to calculating these double time green functions . we handle this by solving coupled dyson equations in a two - dimensional cartesian grid whose dimensions correspond to the time arguments of the green function . care must be taken in this procedure to build the memory effect into the calculation . this is accomplished by performing a convergence test via increasing size of the matrix in which the values of the green function are stored . in order to solve the coupled dyson equations , the pseudofermion and slave boson self energies must be approximated somehow since they are critical inputs to these equations . we omit higher order corrections and thus are content with the non - crossing approximation(nca ) @xcite for this purpose . as crude as it looks , it is known to give remarkably correct results for dynamical quantities at zero magnetic field outside the fermi liquid regime which corresponds to temperatures much smaller than the kondo energy scale . we will stay away from the fermi liquid regime in this paper for this reason . once the green functions are obtained in this method , the matrix that stores their values is propagated in time in discreet steps diagonally . this is how the instantaneous values of the physical quantities are determined . the physical quantities we will be interested are electrical conductance @xmath41 , thermal conductance @xmath42 and thermopower @xmath43 . in linear response , their instantaneous values can be expressed in terms of the onsager coefficients given by @xmath44 where @xmath45 is the fermi - dirac distribution function given by @xmath46 with @xmath47 . in terms of these coefficients , the physical quantities can be cast as @xmath48 on the other hand , the figure of merit of a nanodevice is a dimensionless quantity which is an indicator of its efficiency . hence , it is a benchmark on its performance in a circuit . in terms of the previously defined quantities , it is given by @xmath49 the figure of merit always acquires a positive value and larger figure of merit values imply better performance . we will be ignoring the phonon contribution to the thermal conductance . this assumption is justified because we will be carrying out our calculations around the kondo temperature of the device which is only a few kelvins . phonon contribution to the thermal conductance can be safely ignored in this regime compared to the electronic one in contrast with some previous work where the system is outside kondo regime and the ambient temperature is several hundred kelvins @xcite . in this paper , we will be concerned with the behaviour of instantaneous thermopower and figure of merit for a quantum dot whose energy level is shifted suddenly to a position close to fermi levels of contacts such that the kondo resonance can develop in this final state . we will have the chance to make a direct comparison between the effect of metal and graphene type electrodes on transient values of these quantities . this constitutes a fairly realistic model to simulate the thermal switching behaviour of single electron devices since electron - electron interactions are captured as well as the structure of the electrode material . hence , our investigation should shed light on design and implementation of these devices experimentally . our only limitation will be that our results will pertain to linear response only due to the validity of onsager coefficients in this regime . this is also realistic because infinitesimal temperature gradient between the electrodes can be achieved easily by illuminating one of the contacts with a faint laser beam . in this section , we will present our numerical results on the instantaneous thermopower and figure of merit of a single quantum dot whose energy level is abruptly moved from @xmath50=-5@xmath3 to @xmath50=-2@xmath3 via a gate voltage . this motion triggers a transition from a state where the kondo resonance is completely absent to a state where the kondo resonance gradually emerges . this transition takes place due to a drastic change in kondo temperature @xmath0 since it critically depends on the position of the dot energy level @xmath50 as seen in eq . ( [ tkondo ] ) . immediately after the dot energy level has been moved to its final position for ambient temperatures of @xmath51 ( red dashed ) , @xmath52 ( blue dot dashed ) and @xmath53 ( orange solid ) in linear response for metal electrodes shown in fig . [ schematic]b . , width=328,height=226 ] we will start our analysis with the instantaneous thermopower results . [ fig2 ] depicts the instantaneous thermopower for three different ambient temperatures immediately after the dot level has been switched to its final location for metal electrodes shown in fig . [ schematic]b . the kondo temperature in the final state is inferred to be @xmath0=0.0021@xmath3 . consequently , the higher two temperatures are above @xmath0 while the lowest temperature is slightly lower . we see in this figure that damped sinusoidal oscillations start developing at the onset of kondo timescale with the same frequency for all temperatures . kondo timescale , during which the kondo resonance starts emerging slightly above the fermi level , takes place after the short timescale which corresponds to charge fluctuations . short timescale roughly takes place between 0@xmath5410 . this result is quite interesting and different from conductance oscillations which appear much later at @xmath5580 @xcite . another novelty is the insensitivity of these oscillations to asymmetry of the junction defined as @xmath56 . changing the asymmetry factor @xmath57 does not effect the frequency and amplitude of the oscillations . moreover , the final steady state thermopower ( i.e. @xmath58 ) also remains unchanged when @xmath57 is altered in conjunction with other studies investigating steady state thermopower@xcite . the only resemblance these results bear to instantaneous conductance data is the saturation of the amplitudes below the kondo temperature . we did not show the curves well below @xmath0 in fig . [ fig2 ] as they overlap with higher temperature ones and obscure the oscillations . the oscillation frequency is the same at all temperatures and the oscillations persist deep into the long timescale with @xmath55300 when @xmath59 . immediately after the dot energy level has been moved to its final position for ambient temperatures of @xmath51 ( red dashed ) , @xmath52 ( blue dot dashed ) and @xmath53 ( orange solid ) in linear response for graphene electrodes shown in fig . [ schematic]a . , width=328,height=226 ] fig . [ fig3 ] displays our numerical results for the instantaneous thermopower using the same parameters used in fig . the major difference is that the quantum dot is attached to graphene electrodes shown in fig . [ schematic]a . the oscillations starting in the kondo timescale are still present here . however , their amplitudes are suppressed dramatically compared to metal electrodes . on the other hand , the steady state thermopower values are still the same as fig . [ fig2 ] but the oscillation frequency doubled . these results obviously suggest that graphene is a more desirable material to be used as electrode in single electron devices because it enables to minimize if not completely suppress the undesirable fluctuations during the switching of these devices . this is a great advantage in integrating these devices into larger circuits . = 200 that occurs in thermopower oscillations at ambient temperature of @xmath53 ( orange dashed ) and @xmath52 ( blue dotted ) as a function of the position of the van hove singularity in the contact density of states . solid dots represent the actual data points . , width=328,height=226 ] we want to dwell on the underlying microscopic mechanism for these oscillations . the numerical results presented above show that the frequency of oscillations is proportional to the energy separation between the fermi level of the contacts and the sharp feature in the density of states of the contacts . that feature happens to be in the edge of the band for graphene therefore it generates the highest frequency . more importantly , the amplitude of the oscillations decreases as the energy separation between the fermi level and the sharp feature increases resulting in the minimum amplitude for graphene . in order to generalize and verify this conclusion , we performed additional calculations with the same parameters in fig . [ fig2 ] and fig . [ fig3 ] by using metal electrodes whose density of states is triangular like fig . [ schematic]b except the fact that the van hove singularity is located at -0.75@xmath2 . we calculated the amplitude of the nearest peak or trough occurring around @xmath60=200 for all three electrodes . the amplitude has been determined by taking the average of the two amplitudes corresponding to the two adjacent peak - trough pairs . this step is necessary because the oscillations are damped . the results are shown in fig . [ fig4 ] for two ambient temperatures as a function of the position of the van hove singularity and it is obvious from here that the graphene electrodes where the van hove singularity is at -@xmath2 produce minimum amplitude at both temperatures . the instant @xmath60=200 is arbitrary and same calculation at other instants yields similar results . if the dot gets attached to electrodes made up of two dimensional electron gas with flat density of states and the same bandwidth @xmath2 , the oscillation frequency is the same as graphene because the van hove singularity still lies at the edge of the band . however , the oscillation amplitudes turn out to be larger than graphene presumably due to the large difference in the value of the density of states at fermi level . immediately after the dot energy level has been moved to its final position for ambient temperatures of @xmath51 ( red dashed ) , @xmath52 ( blue dot dashed ) and @xmath53 ( orange solid ) in linear response for metal electrodes shown in fig . [ schematic]b . , width=328,height=226 ] these results suggest the presence of an interference process . the components of this interference should be the two sharp features . these are nothing but the van hove singularity in the density of states of the electrodes and the kondo resonance in the dot density of states . this interference is enabled due to two factors . first , the density of states of the dot and the electrodes are mixed due to the hybridization term in the hamiltonian therefore the kondo resonance and the van hove singularity can interact even though they are not located in the same density of states . second , the development of the kondo resonance is gradual and it is fully formed only at the end of the long timescale @xcite implying that it is a dynamical time - dependent feature . on the other hand , the van hove singularity at the contact density of states is static and it does not evolve in time . consequently , the interference can take place between a dynamic and a static feature . it dies out once they both become static , hence we observe damped oscillations . the thermopower is related to the dot density of states at low temperatures via sommerfeld expansion which is given by @xmath61 in atomic units . @xmath62 corresponds to the value of the dot density of states at fermi level of the contacts and @xmath63 is its derivative . the formation of the kondo resonance has been previously shown to take place slightly above the fermi level @xcite giving rise to a positive derivative at fermi level . consequently , the overall sign of the thermopower in the final state is always negative . as the kondo resonance starts growing in the final state , the derivative begins to increase in magnitude . when the kondo resonance reaches its full form in the long timescale , the derivative stabilizes . this causes the thermopower to reach its steady state . the fundamental difference between the oscillations in the thermopower and the conductance reported previously lies in their start of occurrence and sensitivity to asymmetry . the thermopower oscillations start much earlier than the conductance oscillations because tiny fluctuations in the dot density of states at fermi level due to this interference are enhanced as a result of the derivative in sommerfeld expansion and thus manifest themselves in the thermopower at the onset of kondo timescale prominently . the insensitivity to asymmetry originates from the constructive interference of the right and left electrodes with the kondo resonance magnifying the effect even further . the destructive interference between the left and right electrodes diminishes the conductance oscillations gradually with increasing @xmath57 @xcite . we also want to go over the instantaneous figure of merit results . we use the same parameters in fig . [ fig2 ] and fig . [ fig3 ] for this purpose . [ fig5 ] displays the instantaneous figure of merit immediately after the dot level has been switched to its final position for metal electrodes . a first glance at this figure reveals that the figure of merit is nonzero in initial state on contrary to the thermopower shown in fig . [ fig2 ] and fig . this can be understood by simplifying the figure of merit using the wiedemann - franz law . the ratio of the thermal conductivity to electrical conductivity is given by the wiedemann - franz law which can be expressed as @xmath64 where @xmath65 is a constant called the lorentz number . it can be cast as @xmath66 in atomic units . inserting this identity into eq . ( [ merit ] ) gives @xmath67 this obviously implies that the thermoelectric figure of merit can never be negative and is proportional to the square of the instantaneous thermopower which we just investigated . however , the main subtlety is that the value of the lorentz number deviates from its real value in mesoscopic systems including quantum dots outside the fermi liquid state . fermi liquid state occurs when @xmath68 . incidentally , our calculations invoking nca is valid outside the fermi liquid state too . in the initial dot energy level , @xmath69 and our numerical results reveal a much larger thermoelectric figure of merit than the final dot energy level owing to the suppression of the lorentz number greatly in line with previous steady state results @xcite . consequently , the instantaneous figure of merit exhibits a sharp jump upon switching the dot level abruptly at t=0 . immediately after the dot energy level has been moved to its final position for ambient temperatures of @xmath51 ( red dashed ) , @xmath52 ( blue dot dashed ) and @xmath53 ( orange solid ) in linear response for graphene electrodes shown in fig . [ schematic]a . , width=328,height=226 ] the behaviour of the instantaneous figure of merit curves in the final state generally follows the square of the respective instantaneous thermopower curves . the oscillation frequency is again the same in this case . besides these obvious points , there is a much more intricate issue that is worth pointing out here . we deliberately chose ambient temperatures that are around @xmath0 for our calculations . an earlier investigation showed that the value of the lorentz number declines with a sharp slope upon passing through @xmath0 as a function of temperature @xcite . this rapid change of the value of the lorentz number plays an important for our time dependent calculations . this role stems from the fact that the instantaneous figure of merit is proportional to the inverse of the lorentz number as seen in eq . ( [ ztnew ] ) . the @xmath70 serves as an enhancement factor for the instantaneous figure of merit . the amplitudes of the oscillations taking place at or above the kondo temperature @xmath0 are enhanced more than the amplitudes of the oscillations occurring at ambient temperatures below @xmath0 as a result of this slope around @xmath0 . this can be seen in fig . [ fig5 ] . on the other hand , fig . [ fig6 ] shows the instantaneous figure of merit results for the same parameters used in fig . [ fig5 ] when the quantum dot is attached to graphene electrodes . we see in this figure that the graphene electrodes largely suppress the amplitude of the oscillations compared to fig . [ fig5 ] due to the larger energy gap between the van hove singularity at the edge of the band and the fermi level . this also doubles the oscillation frequency compared to fig . main advantage of this suppression is the prevention of the sudden drop of the device performance at trough points . another interesting peculiarity occurring in fig . [ fig6 ] due to the sharp variation of the value of the lorentz number across @xmath0 is that the value of the instantaneous thermopower at an ambient temperature above @xmath0 can temporarily exceed the value of the instantaneous thermopower below @xmath0 . this can be seen between 40@xmath54140 . this is clearly a transient phenomena and can not be accessed with a steady - state formalism . in conclusion , we studied in linear response both the instantaneous thermopower and figure of merit for a quantum dot whose energy level is suddenly moved into a position such that the many body kondo resonance develops in the final position . we modeled graphene and metal type contacts for this purpose . we uncovered damped sinusoidal oscillations starting at the onset of the kondo timescale and persisting until the long timescale regardless of the asymmetry factor . this behaviour is distinct and different from the instantaneous conductance oscillations which were investigated previously in great detail . the frequency of these oscillations turns out to be proportional to the separation between the fermi level of the contacts and the van hove singularity in the contact density of states . the amplitudes more or less saturate around the kondo timescale suggesting that the interference between the kondo resonance and the van hove singularities is the underlying physical mechanism for these oscillations . this interference gives rise to small fluctuations in the dot density of states at fermi level whose effects are amplified in the thermopower through its derivative in the sommerfeld expansion . this results in fairly large oscillation amplitudes for metallic electrodes . however , graphene contacts generate the smallest amplitude since their van hove singularity lies farthest away from the fermi level minimizing the interference . this renders graphene as an ideal electrode candidate for future single electron devices . we also investigated the instantaneous figure of merit for the same set up . violation of the wiedeman - franz law causes peculiarities here . a sharp decline of the value of the lorentz number around the kondo temperature results in an enhancement of the oscillations above the kondo temperature more than the ones occurring below the kondo temperature . due to this subtlety , the value of the instantaneous thermopower at an ambient temperature above @xmath0 can temporarily exceed the value of the instantaneous thermopower below @xmath0 in the kondo timescale . as a final note , we would like to comment on the experimental feasibility of our work . graphene sheets can be manufactured routinely via scotch tape method and quantum dots are produced via molecular beam epitaxy quite easily . the small temperature gradient between the contacts can be obtained by illuminating one of the contacts with a faint laser beam . moreover , recent remarkable progress in ultrafast pump - probe methods @xcite should make measuring thermopower and thermal conductance in real time possible . the authors thank t@xmath71bitak for generous financial support via grant 111t303 .

we investigate the effect of contact geometry on the instantaneous thermoelectric response of a quantum dot pushed suddenly into the kondo regime via a gate voltage using time dependent non - crossing approximation and linear response onsager relations . we utilize graphene and metal contacts for this purpose . instantaneous thermopower displays sinusoidal oscillations whose frequency is proportional to the energy separation between the van hove singularity in the contact density of states and fermi level for both cases regardless of the asymmetry factor at the onset of kondo timescale . the amplitude of the oscillations increases with decreasing temperature saturating around the kondo temperature . we also calculate the instantaneous figure of merit and show that the oscillations taking place at temperatures above the kondo temperature are enhanced more than the onesoccurring at lower temperatures due to the violation of the wiedemann - franz law . graphene emerges as a more promising electrode candidate than ordinary metals in single electron devices since it can minimize these oscillations .

introduction theory results and discussion conclusions acknowledgments

the need of a precise knowledge of the atmospheric production of secondary particles induced by the cosmic ray ( cr ) flux of charged particles , became a major stake for the scientific community with the recently reported evidence for atmospheric neutrino oscillations pointing to a non - zero neutrino mass , which evidence was based on a comparison of the measured data with secondary atmospheric neutrino flux calculations . in this context , the rapidly increasing amount and statistical significance of the data collected by underground neutrino detectors @xcite , also made precise atmospheric neutrino flux calculations highly desirable as a potential tool to test calculations and models . one - dimensional calculations of the neutrino flux have been considered for long as a good approximation of the flux at earth , until some first attempts of three dimension calculations were performed @xcite , which rapidly became the new standard approach . a survey of the different approaches can be found in @xcite . a precise calculation of the neutrino flux relies on a precise knowledge of the hadronic production cross sections , of nucleons and mesons at the top of the decay chains leading to neutrinos in the final state . these cross sections however are not known in general with a high level of accuracy . a three - dimensional ( 3d ) simulation describing the cr induced cascade in the atmosphere , particle propagation in the geomagnetic field , and interactions with the medium , was developed by the authors for the interpretation of the ams01 data . the code was successfully used to reproduce the proton , electron - positron , and helium 3 flux data @xcite measured by ams and their respective dependence on the geomagnetic coordinates . the lack of a 3d calculation of the neutrino flux at this time led the authors to be the first to investigate this important issue either , with some first results reported in @xcite . a complete report on the calculated muon and neutrino flux was published in @xcite . this paper reports the results of a further investigation of the issue , with the calculations improved by several respects . one improvement consisted in the use of variance reduction techniques ( see section [ sec : vrt ] below ) which allowed , together with some code optimization , to appreciably increase the statistics of the simulated events sample , and thereby improve the monte - carlo precision ( statistical accuracy ) of the calculation . a second improvement was based on the observation that the particle production cross sections are the main source of uncertainty in the secondary particle flux calculations . the reliability of the neutrino flux simulation is usually tested on the capacity of the calculations to reproduce the atmospheric muon and proton flux measurements . the latter can be as well used to directly constrain the secondary particle production cross sections , and to reduce the simulation uncertainties on the calculated neutrino flux . this prospect has been explored in details and it is shown ( see section [ sec : atmfit ] ) that the accuracy of the flux calculations can be significantly improved by this method . the paper is organized as follows . the method and models used in the calculations are introduced in section [ sec : simu ] . section [ sec : results ] is devoted to the results on the proton , muon and neutrino flux in the atmosphere . summary and conclusions are given in section [ sec : conc ] . the calculation proceeds by means of a full 3d - simulation monte carlo simulation . in this section the main features of the program are recalled for convenience . see refs @xcite for other details . for the primary flux , the 1998 ams measurement of cr proton and helium flux @xcite and the fit from @xcite for heavier elements ( up to iron ) , are used . the kinetic energy range of incident crs covered in the simulation is @xmath0 $ ] gev / n . for each periods of the solar cycle , the incident cosmic flux are corrected for the different solar modulation effects using the simple force field approximation @xcite . incident cosmic rays are generated on a virtual sphere at a finite distance in the earth neighborhood . the flux at any point inside the volume of this virtual sphere is isotropic provided the differential element of the zenith angle distribution of the particle direction generated on the sphere is proportional to @xmath1 , @xmath2 being the zenithal angle of the particle . the geomagnetic cut - off is applied by backtracing the particle trajectory in the geomagnetic field ( using the method described in [ propagation ] ) , keeping in the sample only those particles reaching a backtracing distance of 10 earth radii . the altitude of the virtual sphere has to be chosen outside the atmosphere to ensure a backtracing process in a region free of interaction . in the present case it was chosen at 2000 km . for cosmic ray nuclei , the superposition model @xcite was used . in this approximation a nucleus of mass number @xmath3 and charge number @xmath4 is replaced by @xmath5 neutrons and @xmath4 protons with the same velocity as the parent nucleus . the particles are then processed like real protons and neutrons , excepted that the effective charge used for the propagation in the geomagnetic field is set to @xmath6 to keep the same rigidity ( i.e. the same trajectory ) as the parent nucleus . proton invariant triple differential cross section measured ( open square ) in the @xmath7 at 19 gev / c reaction @xcite compared with the fitted parametrization as a function of the proton kinetic energy and for different production angles : 0.7@xmath8 , 1.1@xmath8 ( scaled by a factor @xmath9 ) , 1.7@xmath8 ( @xmath10 ) , 2.3@xmath8 ( @xmath11 ) , 2.9@xmath8 ( @xmath12 ) , 3.4@xmath8 ( @xmath13 ) , 4.0@xmath8 ( @xmath14 ) . , scaledwidth=40.0% ] the model used to simulate the atmosphere is the msise-90 model available from @xcite which describes the temperature and densities in the earth s atmosphere from ground to thermospheric altitudes . the geomagnetic model used in this version of the simulation is igrf9 @xcite which is the ninth generation of a mathematical model of the earth s main field and its annual rate of change ( secular variation ) adopted by the international association of geomagnetism and aeronomy ( iaga ) . each particle is propagated in the geomagnetic field and interacts with nuclei of the local atmospheric density . the particles trajectory is built by numerically integrating the equation of motion using the fifth order runge kutta method with adaptive step - size control from ref @xcite . at each step of the propagation , in addition to the three coordinates and the three components of the momentum , the grammage crossed by the particle is computed . this allows the integration step size to be adapted to the local value of the atmosphere density and then insure a smooth penetration in the atmosphere and a precise calculation of the interaction probability and of the ionization energy loss , and the latter be computed for each step along the trajectory . every secondary particles are processed the same way as their parent particle , leading to the generation of an atmospheric cascade , more or less extensive , for each event . nucleons , pions and kaons are produced for each interaction with their respective triple differential cross sections . the cross sections used to produce secondary particles ( @xmath15 , @xmath16 , @xmath17 , @xmath18 ) are a very important input in the calculation , and the main source of uncertainty in the atmospheric flux estimation . in the present approach the particle production cross sections are obtained from fits to the data using an approach based on the kalinovsky - mokhov - nikitin ( kmn ) parametrization of the inclusive hadronic cross sections @xcite . in the present work , a wide set of experimental data have been used to constrain the parameters of a modified kmn analytical formula @xcite . the set of data included in the fits have been selected to cover as far as possible , the kinematics of interest for atmospheric secondary particles production . an interest of this method is that together with the best fit parameter set for each reaction channel , the errors on the parameters can also be estimated . an illustrative example is shown on figure [ fig : allaby ] with the comparison between the data from @xcite ( @xmath19 at 19 gev / c ) and the fit results ( solid line ) . for all the studied samples of each reaction channel , the reduced @xmath20 obtained from fits were found larger than 1 , typically between 1 and 5 . this could be expected since the statistical and systematic errors on the cross section measurements are usually approximate or poorly known . to account for this uncertainty , a scale factor on the experimental error equal to 1/@xmath21 has been included to evaluate the 95% confidence interval on the cross sections . this procedure is strictly valid only in case of purely statistical errors . it is thus not completely satisfying from this point of view but it should be rather conservative however . the gray band on the figure corresponds to this 95% confidence interval obtained for the proton production cross section ( from the overall proton data analysis ) . the obtained confidence intervals on the cross sections were then used to compute the uncertainty on the secondary particles flux , and ultimately , on the neutrino flux . for the proton induced kaon production and the pion induced pion production the original kmn parametrizations were used @xcite . these cross sections are less critical since their contribution to the final neutrino flux are rather small , namely less than 25% and 5% respectively up to 30 gev incident energy . in the simulation , the ( @xmath22 ) spectra from muon decay , were generated according to the fermi theory , and the muon polarization was taken into account as in the previous work by the authors . for the kaon decay , the dalitz plot distribution given in @xcite was used . to increase the efficiency of the simulation several techniques could be used to reduce the variance of the estimator produced by the simulation @xcite . the technique used here was the important sampling method which consists simply of increasing arbitrarily the incident flux of particles , where the latter produce secondaries which contribute dominantly to the final differential flux to be evaluated , in a particular kinematical domain . for instance , in the computation of the high energy neutrino flux , the high energy galactic flux of charged cosmic rays provides the dominant contribution . the natural abundance of this generic flux is extremely small however . the above mentioned technique is then used to get around this difficulty and the he cr production probability is enhanced by a factor increasing the production yield of the secondaries of interest . the incident particle and all the particles produced in this event will then simply be weighted with the inverse enhancement factor used to boost the particle production . this weight is then taken into account when computing the estimator from the simulation ( see next section ) . another classical use of this method is for computing the flux of a given particle in a given area at a given altitude . the primary particle flux generated vertically downwards around the zenith of the fiducial virtual detector area will have a much larger contribution to the calculated final flux , compared to crs produced far away from this geographical region . in this case a natural way to improve the simulation efficiency is to increase production probability for particles generated near the detection region , and to correct for the induced bias by appropriate statistical weighting of the corresponding events , as described above . the advantage of the method is that all contributing particles are taken into account in the generation process , none of them being excluded or neglected . a bad choice in the enhancement factor used to increase a certain class of primary particles would thus increase the variance of the estimators but it would not bias the results . each particle was traced by the program , and his history , trajectory parameters and kinematics , were recorded . the event file was then analyzed separately to generate the various distributions of interest . to compare the simulated and the measured flux , a virtual surface of detection was defined at the detector altitude . it had to be large enough to ensure a large counting statistics and a precise estimation of the physics observables , but small enough to preserve the accuracy of the calculation . in practice the size of area for each computation was chosen to provide an accuracy below 1% . the simulated differential particle flux is calculated ( @xmath23 ) from the sum of the number of particles detected in the chosen area times , for each particle , the geometrical efficiency in the given particle direction of the considered experiment , times the weight obtained from the sampling method discussed in section [ sec : vrt ] to compensate the enhancement factor . this latter number is then divided by the ( local ) surface of the particle collection area , the integral of detector acceptance over the full solid angle , energy bin size , and equivalent sampling time of the cr flux . the equivalent sampling time of cr flux is obtained from the total event number generated in the simulation run(s ) divided by the surface of the generation sphere times the integrated ( @xmath24 weighted ) solid angle ( @xmath25 ) times the energy integrated flux . open squares : proton flux measured by the bess experiment at mountain altitude 2770 m above sea level , data from @xcite . light gray band : 95 % confidence interval from the simulation . the estimation of the confidence interval includes only errors from the production cross sections . the dark gray band corresponds to the 95 % confidence interval obtained by fitting the cross section on atmospheric data ( see section [ sec : atmfit].,scaledwidth=35.0% ] the ability of the simulation to account for the proton production in the atmosphere can be probed by comparing the simulation results with the recent @xmath15 flux measurements performed by the caprice experiment between sea level and high altitude ( @xmath26 40 km ) @xcite , and by the bess experiment at mountain altitude @xcite . figure [ bessproton ] shows the flux measured by bess @xcite compared to the calculated flux . the bess measurements were performed at mt norikura , japan ( geographical location 36n,137.5e ) at 2770 m of altitude above sea level ( corresponding to an atmospheric depth of 742 g / cm2 ) in 1999 . the virtual detector area for this computation was chosen as the geographical region defined by @xmath27 rad where @xmath28 is the cgm latitude @xcite , thus consisting of two bands ( one in each hemisphere ) centered on plus or minus the experiment cgm latitude ( 0.509 rad ) . as in the measurements , the geometrical efficiency is set to one for zenithal angular range @xmath2 with @xmath29 . figure [ bessproton ] shows the confidence interval on the accuracy of the simulated flux ( with a 95 % level ) . the estimation of the confidence interval includes only the error associated to the cross section and were computed using the confidence interval associated to the cross section parametrization ( see section [ sec : param ] ) . the large width of the confidence interval in this estimation is due to the large number of collisions between the primary particle and the detected proton at low altitude . any uncertainty in the proton production cross section is then amplified by a large factor ( approximatively 8 on the average , see @xcite ) . figure [ capproton ] shows the flux measured by caprice for different altitudes in the atmosphere during the 1998 flight from fort summer , n. m. , ( 34.28 n , 105.14 w , cgm latitude 0.76 rad ) @xcite and the corresponding simulation results . the virtual detector area for this computation was chosen as the region defined by @xmath30 rad . the geometrical efficiency used here was taken from @xcite to reproduce the detector acceptance . at high altitude , the effect of the geomagnetic cutoff is clearly observed in all spectra . at this cgm latitude the mean geomagnetic cutoff is 6 gev for protons . above the cutoff the flux is dominated by the primary cosmic component whereas below the geomagnetic cutoff the flux consists only of secondary protons produced by cosmic rays above the cutoff . the primary flux is increasingly absorbed in the atmosphere with the decreasing altitude , and the secondary component becomes dominant over the entire energy range . the confidence interval of the calculated flux is seen to increase with the decreasing altitude , i.e. , with the mean number of interactions as it could be expected @xcite . it can be observed that at low altitudes , the uncertainty of the simulation results ( which includes only the production cross section uncertainties ) is much larger than the uncertainty from the flux measurements . therefore , the precise proton flux data at low altitudes can be used to constrain the parametrization of the secondary particles production more tightly and more accurately than the available nuclear data . atmospheric muons are produced in the same decay chain as neutrinos . their spectra are thus an essential ground to probe the reliability of the neutrino flux calculated in the same framework . several experiments , caprice , heat , bess , have made precise flux measurements at various altitudes in the atmosphere @xcite which can be used for this purpose . figures [ fig : bessmountain ] and [ fig : besstsukuba ] show the calculated positive and negative muon flux and the corresponding flux ratios compared with the data measured by the bess experiment at mountain altitude ( mt norikura , location 36n,137.5e , altitude 2770 m ) @xcite and at sea level ( tsukuba , location 36.2 @xmath8n,140.1 @xmath8e , altitude 30 m ) @xcite . for these two calculations , the virtual detectors were chosen as the region defined by @xmath31 rad at the experiment altitude , and with the geometrical efficiency set to one for zenithal angles @xmath2 with @xmath32 . the 95 % confidence interval is also shown on the figure . it includes the uncertainty originating from the parametrization of the secondary protons , neutrons and pions production cross sections . note that for the muon charge ratio the uncertainty results only from the uncertainty on the pion production cross section . figure [ fig : cap98muon ] shows the calculated positive and negative muon flux compared with the data measured by the caprice experiment during the 1998 flight @xcite at fort summer . a good agreement is found between data and the calculation for the positive and negative fluxes at different altitude going from the top of atmosphere to the ground level . it is clearly seen on the figures [ bessproton ] to [ fig : cap98muon ] that , at least for the low altitude measurements , the statistical precision achieved by the atmospheric experiments is significantly better than the precision obtained from the simulation , taking into account the uncertainty on secondary particles production cross section . this results was obtained although all the relevant available nuclear data have been collected to constrain the parametrized cross sections used in these calculations . therefore , the atmospheric flux measurements are in this sense more accurate that the nuclear data . in other works on neutrino flux calculations , the calculated atmospheric flux of charged particles , used to check the reliability of the calculations , were all based on existing hadronic monte carlo generator @xcite , whereas in the present approach , in account of the good accuracy of the available measurements , the atmospheric data flux can be used to constrain the hadronic cross sections to ultimately improve the final accuracy of the calculated neutrino flux . in this work , the proton and muon measurements at the mt norikura altitude and the muon measurements in tsukuba , from the bess experiment , were used to constrain both the proton and pion production cross sections . the obtained 95 % interval of confidence is shown ( dark gray ) on figure [ bessproton],[fig : bessmountain ] and [ fig : besstsukuba ] . the reduced @xmath20 obtained from this fit was 4.7 , and the same procedure as described in section [ sec : param ] was thus used to estimate the confidence interval . it must be noted that these fitted cross sections may absorb systematic errors originating from other sources , like the primary cr fluxes , although the uncertainty on the latter are much smaller than on the cross sections . however , this does not affect the ultimate improvement of the neutrino flux uncertainty , since the deduced interval of confidence accounts for both sources of uncertainty . the calculated energy distributions of the atmospheric neutrino flux around the superkamiokande ( sk ) detector have been computed with the same simulation program . a virtual detector center was defined at the sk geographical coordinates ( 36@xmath8n,137@xmath8e ) with a size of 8@xmath8 in latitude and 18@xmath8 in longitude ( or @xmath26900@xmath331600 km@xmath34 ) . the virtual detector size has been chosen so as not to change the calculated flux by more than 1 % . figure [ fig : flux ] shows the average flux over the full @xmath35 range of solid angle for @xmath36 and @xmath37 ( left ) and the flux ratio @xmath38 , @xmath39 and @xmath40 ( right ) . the light gray area represents the 95 % confidence level corresponding to the uncertainty due to the particle production cross sections inaccuracies . the uncertainty for the absolute neutrino flux is of the order of 10 % and , as expected , these error contributions are largely reduced for the ratio of flux @xmath40 ( insensitive to proton / neutron production cross sections ) and for @xmath39 ( insensitive to proton / neutron production cross sections but also to pion production cross sections for low energy neutrinos ) and vanished for the @xmath41 flux ratio ( insensitive to proton / neutron and pion production cross sections ) . the dark gray band correspond to the 95 % confidence interval obtained by fitting the cross section on atmospheric data discussed in the section [ sec : atmfit ] . using this method the uncertainty for the absolute flux is reduced to the order of 3 % . the zenith angle distribution of the flux and of the flavor ratios are shown on figure [ zenith].left for 5 energy bins between 0.1 and 30 gev . at low energy the zenith angle distribution displays an enhancement for directions close to horizontal , in qualitative agreement with @xcite . this enhancement originates from the isotropisation of the neutrino flux resulting from cosmic ray showers . it is expected to be maximum for a purely isotropic neutrino emission and nil for a perfectly collimated neutrino production . the maximum is then naturally obtained for the lowest energy bin ( about twice the downward or upward going flux out of the horizontal plane ) . it is decreasing with the increasing energy . at low energy the zenith angle distribution for downward and upward neutrinos , outside the @xmath42 peak , is approximatively flat , this isotropy naturally originating from the cr isotropy . the small structures observed arise from geomagnetic effects on the primary flux . at high energies the distributions become sensitive to the muon decay probability which depends on the muon pathlength in the atmosphere and on its energy . as expected the @xmath43 dependence is becoming steeper with the increasing energy . the effect is enhanced for electronic neutrino which originate exclusively from muon decay . this is shown on figure [ zenith].right where the flavor ratio @xmath44 is seen to depend on the zenith angle , especially for high energy neutrinos , for the same reason as given previously . on these figures , the light gray area represents the 95% confidence level corresponding to the particle production uncertainty . here again we see that the uncertainty is reduced at high energy , this is mainly due to the decrease of the mean number of interactions between the primary cosmic ray impinging the atmosphere and the final neutrino detection . as in figure [ fig : flux ] , for the flavor ratio @xmath44 the error contributions from particle production uncertainty vanished . the dark gray band represents the 95 % confidence interval obtained by fitting the cross section on atmospheric data as described in section [ sec : atmfit ] . the azimuth angle distribution of the flux and of the flavor ratios are shown on figure [ azimuth ] for 5 energy bins between 0.1 and 30 gev . on this figure northward and westward going particles correspond to azimuth angles 0 , @xmath45 respectively . the non - flatness of the low energy bin originates from the geomagnetic cutoff on primary cosmic rays @xcite . the east - west ( ew ) effect of the primary flux ( more eastward going than westward going cosmic rays ) results in a ew asymmetry in the produced secondary particles and then in the neutrino flux at low energy . with the increasing energy , neutrinos are increasingly produced in atmospheric showers initiated by higher energy cosmic rays which are less and less sensitive to the geomagnetic cutoff ( the cosmic ray flux becomes isotropic at a rigidity of 60 gv ) . this results in the simulated azimuthal distribution becoming flat for the last energy bin in figure [ azimuth].left ( neutrino energy between 10 gev and 30 gev ) . for the azimuth angle distributions of the flavor ratios ( figure [ azimuth].right ) , it can be noted that the @xmath46 ratio is strongly ew asymmetric while the opposite is observed for the @xmath47 ratio . the @xmath48 ratio ( not shown ) is found to be almost structureless at all latitudes . this additional ew asymmetry featured by the @xmath46 and @xmath49 ratio can be explained by the muon bending in the geomagnetic field @xcite . in the production chain @xmath50 , the muon will propagate with the same bending as the proton one . an enhancement of ew effect originating from the geomagnetic cutoff is then expected for the neutrino @xmath51 and @xmath52 produced in the @xmath53 decay . in the @xmath54 production chain , the muon will propagate with the opposite bending and a reduction of the ew asymmetry is expected for the @xmath55 decay products @xmath56 and @xmath57 . this explains the ew features seen in the @xmath46 and the @xmath47 ratio . in addition , because muonic neutrinos are also produced in the pion decay , the ew asymmetry is found to be higher for @xmath58 than for @xmath59 . these results confirm the qualitative prediction from @xcite : @xmath60 where @xmath3 stands for the neutrino ew asymmetry . in figure [ azimuth ] the light and dark gray area represents respectively the 95 % confidence level corresponding to the particle production cross section uncertainty and the 95 % confidence interval obtained by fitting the cross section on atmospheric data as described in section [ sec : atmfit ] . the uncertainty are largely reduced for the ratio of flux @xmath40 ( insensitive to the proton / neutron production cross sections ) and for @xmath61 ( insensitive to proton / neutron production cross sections but also to pion production cross sections for low energy ) . a full three - dimensional simulation has been used to compute the secondary particles production in the atmosphere . in these calculations the statistics of the simulated sample has been significantly increase with respect to our previous work , resulting in a negligible contribution of the statistical uncertainty . the dominant error then appeared to come from the uncertainty on the particle production cross sections . in this work this systematic uncertainty has been studied quantitatively . the simulation has been used to reproduce the atmospheric measurements of the proton and muon flux , in order to test the reliability of the calculations . these atmospheric measurements , in account of their good accuracy , were also used to constrain the secondary particle production cross sections , which experimental values from direct measurements on accelerators were of lesser accuracy . the results for the absolute neutrino fluxes and their zenithal and azimuthal angular distributions together with the flavor ratio of these flux at the super - kamiokande location , have been presented . the high statistics accumulated allowed to investigate the detailed structure of these distribution and to discuss them . the uncertainty of the calculations has been estimated . it has been shown first , that this uncertainty is at the level of 10 % for the absolute flux , and second that it can be reduced to the level of 3 % by using the atmospheric data on the muon and proton flux to constrain the production cross sections . these results provide an accurate basis for the comprarison to future experimental data . 99 super - kamiokande collaboration , y. fukuda , et al . 81 , 1562(1998 ) , b433 , 9(1998 ) , b436 , 33(1998 ) . the soudan 2 collaboration , w. w. m. allison , et al . b449 , 137(1999 ) . w. w. m. allison , et al . b391 , 491(1997 ) . j. wentz , phys . d 67 , 073020 ( 2003 ) . m. honda et al . , phys . rev . d 70 , 043008(2004 ) . g. battistoni et al . 19 , 269(2003 ) . barr et al . , phys . rev . d 70 , 023006(2004 ) . l. derome et al . , phys . b489 , 1(2000 ) . l. derome , m. bunerd , and yong liu , phys . lett . b515 , 1(2001 ) . l. derome , and m. bunerd , phys . b521 , 139(2001 ) . y. liu , l. derome , m. bunerd , phys . d 67 073022(2003 ) . b. baret , l. derome , and m. bunerd , phys . d68 , 053009(2003 ) . y. liu , l. derome and m. bunerd , 27th int . ray conf . , hamburg , aug 8 - 13 , 2001 . the ams collaboration , j. alcaraz et al . b472 , 215(2000 ) , phys . b490 , 27(2000 ) , phys . b494 , 19(2000 ) . b. wiebel - sooth et al . , astronomy and astrophysics 330 , 389(1998 ) j.s . perko , astro . 184(1984)119 ; see m.s . potgieter , proc . icrc calgary , 1993 , p213 , for a recent review of the subject . j. engel , t. k. gaisser , t. stanev and p. lipari , phys . d46 , 5013 ( 1992 ) . allaby et al . , cern yellow report 70 - 12 , ( 1970 ) . a. e. hedin , j. geophys . 96 , 1159(1991 ) . + msise model available at ` http://modelweb.gsfc.nasa.gov/atmos/ ` the igrf models are available at ` http://www.ngdc.noaa.gov/iaga/vmod/home.html ` w. h. press , s. a. teukolsky , w. t. vetterling , numerical recipes in c++ : the art of scientific computing , cambridge university press ( 2002 ) l. derome and m. buenerd , proc . 29th icrc , pune ( india ) , 2005 , he.2.1 a. n. kalinovsky , n. v. mokhov and y. p. nikitin , passage of high - energy particles through matter , new york , usa : aip 1989 . s. eidelman et al . ( particle data group ) , physics letters b592 , 1(2004 ) . a. haghighat and jc . wagner , prog . energy 42 , 25(2003 ) . t. sanuki et al . , phys . b577 , 10(2003 ) . e. mocchiutti , phd thesis , royal institute of technology , stockholm , 2003 gustafsson , g. , n. e. papitashvili , and v. o. papitashvili , j. atmos . 54 , 1609(1992 ) . bess collaboration , t. sanuki et al . b541 , 234(2002 ) . bess collaboration , s. haino et al . b594 , 35(2004 ) . bess collaboration , m. motoki et al . , astropart . 19 , 113(2003 ) . caprice collaboration , m. boezio et al . d 67,072003 ( 2003 ) . caprice collaboration , j. kremer , et al . 83 , 4241(1999 ) . heat collaboration , s. coutu , et al . d62 , 032001(2000 ) . p. lipari , astropart . phys . * 14 * , 153(2000 ) p. lipari , astropart * 14 * , 171(2000 )

a precise evaluation of the secondary particle production and propagation in the atmosphere is very important for the atmospheric neutrino oscillation studies . the issue is addressed with the extension of a previously developed full 3-dimensional monte - carlo simulation of particle generation and transport in the atmosphere , to compute the flux of secondary protons , muons and neutrinos . recent balloon borne experiments have performed a set of accurate flux measurements for different particle species at different altitudes in the atmosphere , which can be used to test the calculations for the atmospheric neutrino production , and constrain the underlying hadronic models . the simulation results are reported and compared with the latest flux measurements . it is shown that the level of precision reached by these experiments could be used to constrain the nuclear models used in the simulation . the implication of these results for the atmospheric neutrino flux calculation are discussed .

introduction simulation model results summary and conclusion

at present methods for construction of special solutions of nonintegrable systems in terms of elementary ( more precisely , degenerated elliptic ) and elliptic functions are actively developed @xcite ( see also @xcite and references therein ) . some of these methods are intended for the search for elliptic solutions only @xcite , others allow to find either solutions in terms of elementary functions only @xcite or both types of solutions @xcite . note that the methods @xcite allow to find multivalued solutions as well . elliptic and degenerate elliptic functions are single - valued functions , therefore , such solutions of a nonintegrable system exist only if there exist the laurent series solutions of it . such local solutions can be constructed by means of the ablowitz ramani segur algorithm of the painlev test @xcite ( see also @xcite ) . moreover for a wide class of dynamical systems using this method one can find all possible laurent series expansions of solutions . in this way one obtains solutions only as formal series , that is sufficient , because really only a finite number of coefficients of these series is used . examples of construction of such solutions are given in @xcite . the laurent series solutions give the information about the global behavior of differential systems and assist to look for both exact solutions @xcite and the first integrals @xcite . the laurent series solutions can be used to prove the nonexistence of elliptic solutions @xcite as well . in @xcite r. conte and m. musette have proposed a new method for construction of single - valued special solutions of nonintegrable differential equations . a key idea of this method is the use of the laurent series solutions to transform the initial differential equation into a nonlinear system of algebraic equations . using this method one can in principal find all elliptic and degenerate elliptic solutions . unfortunately if the initial differential equation includes the large number of numeric parameters , then it is difficult to solve the obtained nonlinear system of algebraic equations . the goal of this paper is to propose a modification of the conte musette method , which allows to seek elliptic solutions only . we show that in this case it is possible to fix some parameters of the initial differential system and therefore simplify the resulting system of algebraic equations . to do this we use the hone s method , which has been proposed to prove the non - existence of elliptic solutions @xcite . note that using our approach one can find in principal all elliptic solutions . in @xcite the authors have transformed the initial system of two coupled ordinary differential equations into the equivalent single differential equation and only after this have constructed the laurent series solutions . in our paper we demonstrate that the analysis of the system of differential equations may be more useful than the analysis of the equivalent differential equation . moreover in this paper we show that if a system of differential equations includes a few functions it is possible to find the analytic form of a function which satisfies this system even without knowledge of other functions in the analytic form and without elimination of them from the system . in @xcite r. conte and m. musette have proposed a way for searching of elliptic and degenerate elliptic solutions to a polynomial autonomous differential equation . in this section we reformulate this method for a system of such equations : @xmath0 where @xmath1 and @xmath2 . it is known that any elliptic function ( including degenerate one ) is a solution of some first order polynomial autonomous differential equation . the classical results of p. painlev , l. von fuchs , c.a.a . briot and j .- c . bouquet allow one to construct the suitable form of such an equation , whose general solution is a meromorphic function with poles of order @xmath3 ( see details in @xcite ) ) runs over nonnegative integer @xmath4 that are less than or equal to @xmath5 : @xmath6 in which @xmath7 is a positive integer number and @xmath8 are constants to be determined . the general solution of @xmath9 is either an elliptic function , or a rational function of @xmath10 , @xmath11 being some constant , or a rational function of @xmath12 . note that the third case is a degeneracy of the second one , which in turn is a degenerate case of the first one . the conte musette algorithm is the following @xcite ( see also @xcite ) : 1 . choose a positive integer number @xmath7 , define the form of eq . ( [ subequ ] ) and calculate the number of unknown coefficients @xmath13 . 2 . construct solutions of system ( [ f ] ) in the form of the laurent series . if such solutions do not exist or they correspond to known exact solutions , then no unknown single - valued solutions exist . note that , since system ( [ f ] ) is autonomous , the coefficients of the laurent series do not depend on the position of the singular point . they may depend on values of the numerical parameters of ( [ f ] ) . in addition , some of these coefficients ( the number of which is less than the order of system ( [ f ] ) ) may take arbitrary values and have to be considered as new numerical parameters . one should compute more coefficients of the laurent series than the number of numerical parameters in the laurent series plus the number of @xmath13 . 3 . choose a laurent series expansion for some function @xmath14 and substitute the obtained laurent series coefficients into eq . ( [ subequ ] ) . this substitution transforms ( [ subequ ] ) into a linear and overdetermined system in @xmath8 with coefficients depending on numerical parameters . eliminate coefficients @xmath8 and get a nonlinear system in parameters . solve the obtained nonlinear system . r. conte and m. musette note that a computer algebra package is highly recommended for using of their method @xcite . steps of this algorithm can be implemented in computer algebra systems separately . for the given system it is easy to calculate the laurent series solutions to any accuracy . these computations base on the painlev test , which has been implemented in the most popular computer algebra systems @xcite . note that when one has computed a sufficient number of the laurent series coefficients he can forget about the system of differential equations and work only with coefficients of the obtained series . the first package of computer algebra procedures , which realize the third and the fourth steps of the algorithm , has been written in amp @xcite by r. conte . one can also use our maple and reduce packages of procedures , which are accessible in internet @xcite and are described in @xcite . so , one passes the first four steps of algorithm without any difficulties . at the fifth ( last ) step one should solve an overdetermined system of nonlinear algebraic equations . the standard method for solving of such systems is the construction of a lexicographically ordered grbner basis @xcite . the buchberger algorithm @xcite guarantees a construction of a grbner basis in a finite number of steps . however , appropriate estimates of the number of steps required for constructing of this basis do not exist . the required computer memory depends , in the general case , exponentially on the number of the unknowns . therefore , this number should be made as small as possible . the purpose of this paper is to show that we can essentially simplify the algebraic system of equations , which we have to solve on the last step of the conte musette method , if we search the elliptic solutions only . in @xcite r. conte and m. musette have used their method to find wave solutions of the complex cubic ginzburg landau equation ( cgle3 ) . the nonexistence of elliptic travelling and standing wave solutions of the cgle3 has been proved in @xcite and @xcite respectively . in section 4 we seek the elliptic solutions of the quintic complex ginzburg landau equation ( cgle5 ) using our modification of the conte musette method . note that both the cgle3 and the cgle5 have only one - parameter laurent - series solutions in the wave form and there exist only a finite number of such solutions . our approach can be effectively used in investigation of any system ( [ f ] ) or a single differential equation , for which only a finite number of different laurent - series solutions exist . note that for wide class of such differential equations it has been proved that all their meromorphic solutions are elliptic ( maybe degenerated ) functions @xcite . let us recall some definitions and theorems . the function @xmath15 of the complex variable @xmath16 is a doubly - periodic function if there exist two numbers @xmath17 and @xmath18 with @xmath19 , such that for all @xmath20 @xmath21 by definition a double - periodic meromorphic function is called an elliptic function @xcite . these periods define the period parallelograms with vertices @xmath22 , @xmath23 , @xmath24 and @xmath25 , where @xmath26 and @xmath27 are arbitrary natural numbers and @xmath22 is an arbitrary complex number . the fundamental parallelogram of periods is called a parallelogram of period , which does not include other parallelogram of periods , that corresponds to @xmath28 . the classical theorems for elliptic functions @xcite prove that * if an elliptic function has no poles then it is a constant . * the number of elliptic function poles within any finite period parallelogram is finite . * the sum of residues within any finite period parallelogram is equal to zero ( * the residue theorem * ) . * if @xmath15 is an elliptic function then any rational function of @xmath15 and its derivatives is an elliptic function as well . * for each elliptic function @xmath15 there exist such @xmath7 ( @xmath29 ) and such coefficients @xmath30 that @xmath15 is a solution of eq . ( [ subequ ] ) . an elliptic function can not have two poles with the same laurent series expansions in its fundamental parallelogram of periods . * let some elliptic function @xmath31 has two poles in points @xmath32 and @xmath33 , which belong to the fundamental parallelogram of periods . the corresponding laurent series are the same and have the convergence radius @xmath34 . then the function @xmath35 is an elliptic function as a difference between two elliptic functions with the same periods . at the same time for all @xmath36 such that @xmath37 @xmath38 , therefore , @xmath39 and @xmath40 and @xmath41 is a period of @xmath31 . it contradicts to our assumption that both points @xmath32 and @xmath33 belong to the fundamental parallelogram of periods . the one - dimensional quintic complex ginzburg landau equation ( cgle5 ) is a generalization of the one - dimensional cubic complex ginzburg landau equation @xcite ( cgle3 ) , which is one of the most - studied nonlinear equations ( see @xcite and references therein ) . moreover , the cgle5 is a generic equation which describes many physical phenomena , for example , the behaviour of travelling patterns in binary fluid convection @xcite and the large - scale behavior of many nonequilibrium pattern - forming systems @xcite . the cgle5 is as follows @xmath42 where subscribes denote partial derivatives : @xmath43 , @xmath44 , @xmath45 and @xmath46 . one of the most important directions in the study of the cgle5 is the consideration of its travelling wave reduction @xcite : @xmath47 substituting ( [ travwave ] ) in ( [ eqcgl5 ] ) and multiplying both sides of this equation on @xmath48 we obtain @xmath49 { } + 2c\psi-2p\psi^2 + 2\mathrm{i}p\psi ' \bigr)m^2 + 4qm^3 + 4rm^4=0 , \end{array } \label{equ1}\ ] ] where @xmath50 , @xmath51 . equation ( [ equ1 ] ) is a system of two equations : both real and imaginary parts of its left - hand side have to be equal to zero . dividing ( [ equ1 ] ) by @xmath3 and separating real terms from imaginary ones , we obtain the following system @xmath52 \displaystyle m\tilde{\psi } ' + \tilde{\psi}\left(m ' - \tilde{c}m\right)-g_rm + d_i m^2 + u_im^3=0 , \end{array } \right . \label{system}\ ] ] where new real variables are as follows @xmath53 @xmath54 and @xmath55 system ( [ system ] ) includes seven numerical parameters : @xmath56 , @xmath57 , @xmath58 , @xmath59 , @xmath60 , @xmath61 and @xmath62 . note that to obtain ( [ system ] ) from ( [ equ1 ] ) we assume that the functions @xmath63 and @xmath64 are real . the standard way to construct exact solutions for system ( [ system ] ) is to transform it into the equivalent third order differential equation for @xmath65 . we rewrite the first equation of system ( [ system ] ) as @xmath66 where @xmath67 from ( [ phi2 g ] ) it follows that @xmath68 multiplying the second equation of ( [ system ] ) on @xmath69 and substituting ( [ phi2 g ] ) and ( [ dphi2 g ] ) in it , we express @xmath69 in terms of @xmath65 and its derivatives : @xmath70 and obtain the third order equation for @xmath65 : @xmath71 below we consider the case @xmath72 which corresponds to the condition @xmath73 . in this case ( [ gequ ] ) is not integrable @xcite and its general solution ( which should depend on three arbitrary integration constants ) is not known . using the painlev analysis @xcite it has been shown that single - valued solutions of ( [ system ] ) can depend on only one arbitrary parameter . system ( [ system ] ) is autonomous , so this parameter is @xmath32 : if @xmath74 is a solution , then @xmath75 , where @xmath76 has to be a solution . special solutions in terms of elementary functions have been found in @xcite . all known exact solutions of ( [ system ] ) are elementary ( rational or hyperbolic ) functions . the full list of these solutions is presented in @xcite . the purpose of this section is to find at least one elliptic solution of ( [ system ] ) . system ( [ system ] ) is invariant under the transformation : @xmath77 therefore we can assume that @xmath78 without loss of generality . moreover , using scale transformations : @xmath79 we can always put @xmath80 . let us construct the laurent series solutions to system ( [ system ] ) . we assume that in a sufficiently small neighborhood of the singularity point @xmath32 the functions @xmath69 and @xmath65 tend to infinity as some powers of @xmath81 : @xmath82 where @xmath83 and @xmath84 are negative integer numbers and , of course , @xmath85 and @xmath86 . substituting ( [ leadterms ] ) into ( [ system ] ) we obtain that two or more terms in the equations of system ( [ system ] ) balance if and only if @xmath87 and @xmath88 . in other words in this case these terms have equal powers and the other terms can be ignored as @xmath89 . we obtain values of @xmath90 and @xmath91 from the following algebraic system : @xmath92 \displaystyle 2a - b^2=0 . \end{array } \right.\ ] ] system ( [ absystem ] ) has four nonzero solutions : @xmath93 @xmath94 @xmath95 and @xmath96 therefore , system ( [ system ] ) has four types of the laurent series solutions . denote them as follows : @xmath97 where @xmath98 . let @xmath63 is a nontrivial elliptic function . note that if @xmath69 is a constant , then from the second equation of system ( [ system ] ) it follows that @xmath65 can not be a nontrivial elliptic function . therefore , using ( [ psi ] ) , we conclude that @xmath99 has to be a nontrivial elliptic function as well . let us consider the fundamental parallelogram of periods for the function @xmath63 and define a number of its poles in this domain . let @xmath65 has a pole of type @xmath100 , hence , according to the residue theorem , it should has a pole of type @xmath101 ( it can not have a pole of type @xmath102 because @xmath60 is a real parameter ) . so @xmath69 has poles with the laurent series @xmath103 and @xmath104 . as an elliptic function it should have a pole of type @xmath105 or @xmath106 as well . it means that the function @xmath63 should have a pole of type @xmath107 and , hence , a pole of type @xmath102 . so @xmath63 should have at least four different poles in its the fundamental parallelogram of periods . using * lemma 1 * , we obtain that the function @xmath63 can not have the same poles in the fundamental parallelogram of periods . therefore , @xmath63 has exactly four poles in its fundamental parallelogram of periods . in this case by means of the residue theorem for @xmath69 we obtain @xmath108 we obtain that the cgle5 with @xmath109 has no elliptic solution in the wave form . in the case @xmath110 possible elliptic solutions should have four simple poles in the fundamental parallelogram of periods , and , therefore , has the following form @xcite : @xmath111 where the function @xmath112 is an integral of the weierstrass elliptic function multiplied by @xmath113 : @xmath114 @xmath115 and @xmath116 are constants to be defined . we also should define periods of the weierstrass elliptic function . to obtain restrictions on other parameters , we use the hone s method @xcite and apply the residue theorem to the functions @xmath117 , @xmath118 , and so on . the residue theorem for the function @xmath117 gives the equation : @xmath119 the values of @xmath120 are as follows ( @xmath110 ) : @xmath121{27}d_i-15\sqrt[4]{3}d_r\right),\ ] ] @xmath122{27}d_i+15\sqrt[4]{3}d_r\right),\ ] ] @xmath123{27}d_i- 15\sqrt [ 4]{3}d_r\right)\right),\ ] ] @xmath124{27}d_i- 15\sqrt [ 4]{3}d_r\right)\right).\ ] ] substituting @xmath125 and @xmath126 in ( [ psi2equ ] ) , we obtain @xmath127 therefore @xmath128 . for the function @xmath118 the residue theorem gives : @xmath129 the values @xmath130 are as follows ( we put @xmath128 ) : @xmath131 @xmath132 equation ( [ psi3equ ] ) is equivalent to @xmath133 the parameters @xmath58 and @xmath59 should be real , therefore , @xmath134 and @xmath135 . so , consideration of @xmath117 and @xmath118 gives three restrictions : @xmath136 the residue theorem for @xmath137 gives the restriction @xmath138 considering @xmath139 and @xmath140 , we do not obtain new restrictions on coefficients . taking into account ( [ ur0 ] ) and ( [ ristrict2 ] ) we obtain system ( [ system ] ) in the following form : @xmath141 \displaystyle \tilde{\psi}'m + \tilde{\psi } m'-g_rm+m^3=0 . \end{array } \right . \label{system27}\ ] ] to find elliptic solutions to system ( [ system27 ] ) we use the conte musette method . the function @xmath99 can have not four but two different laurent series expansions , whereas the functions @xmath63 should have four different laurent series expansions , so it is easier to find @xmath99 than @xmath63 . equation ( [ subequ ] ) with @xmath142 has no elliptic solution . let @xmath99 satisfies eq . ( [ subequ ] ) with @xmath143 : @xmath144 substituting in ( [ equpsi ] ) the laurent series of @xmath69 , which begins from @xmath145 ( more exactly we use the first ten coefficients ) , we obtain the following solution @xmath146 for an arbitrary value of the parameter @xmath147 and @xmath148 : @xmath149 a few solutions with @xmath148 and @xmath150 and no solution for @xmath151 . the straightforward substitution of the functions @xmath152{3}}{t},\ ] ] or @xmath153{3}}{t}\ ] ] in ( [ system27 ] ) with @xmath150 and @xmath148 proves that they are exact solutions . the coefficients of the laurent - series solutions do not include arbitrary parameters , so the obtained solutions are unique single - valued solutions and system ( [ system27 ] ) has no elliptic solution for these values of parameters . in the case of solutions ( [ solc1 ] ) the function @xmath99 satisfies the equation @xmath154 the polynomial in the right hand side of ( [ equel ] ) has four different roots , therefore @xmath69 is a non - degenerate elliptic function @xcite . surely we do not rigorously prove the existence of elliptic solutions to the cgle5 . more precisely , we calculate fifty coefficients of the laurent series of the function @xmath69 , which satisfies ( [ system27 ] ) with @xmath148 and find that they coincide with corresponding coefficients of the laurent series of the exact solution to eq . ( [ equel ] ) . for rigorous proof we should find the function @xmath63 and check that this function is a solution of @xmath155 . the function @xmath63 in a parallelogram of periods has four different laurent series expansions , so we should choose the parameter @xmath7 such that solutions of eq . ( [ subequ ] ) have four poles in its fundamental parallelogram of periods . minimal possible value of @xmath7 is equal to 4 . the general form of ( [ subequ ] ) for @xmath156 and @xmath157 is the following : @xmath158 substituting the laurent series @xmath159 from ( [ laura ] ) , we transform the left hand side of ( [ equm ] ) into the laurent series , which has to be equal to zero . therefore , we obtain the algebraic system in @xmath30 and @xmath56 . the first algebraic equation , which corresponds to @xmath160 is @xmath161 where @xmath162 is defined by ( [ ab1])([ab4 ] ) . if we would attempt to find all elliptic and degenerate elliptic solutions then we should use only one of @xmath162 and can express , for example , @xmath163 via @xmath164 , @xmath165 and @xmath166 . we seek only elliptic solutions , so we know that all @xmath162 have to satisfy ( [ equm1 ] ) and can consider eq . ( [ equm1 ] ) as the following system : @xmath167 using the explicit values of @xmath162 from ( [ ab1])([ab4 ] ) , we obtain that @xmath168 taking into account ( [ h ] ) , from other equations of the algebraic system we obtain @xmath169 all other @xmath30 are equal to zero . so , the equation for @xmath65 has the form @xmath170 equation ( [ gequ ] ) at @xmath80 , @xmath110 , @xmath128 , @xmath134 , @xmath135 and @xmath148 has the form : @xmath171 we multiply eq . ( [ gequ0 ] ) on @xmath172 and use eq . ( [ equm4 ] ) to express all derivatives of @xmath63 in terms of the function @xmath63 . the straightforward calculation shows that any solution of ( [ equm4 ] ) satisfies ( [ gequ0 ] ) . so , we obtain elliptic wave solutions of the cgle5 . if @xmath173 these solutions are the standing wave solutions , in the opposite case ( @xmath174 ) the solutions can have an arbitrary speed @xmath175 . note that we obtain ( [ gequ ] ) from ( [ equ1 ] ) using the condition that @xmath63 is a real function . for @xmath176 and any initial value of @xmath65 we obtain real solutions . in the case @xmath177 there exists the minimal possible initial value of @xmath65 for which real solutions exist and only particular solutions of ( [ equm4 ] ) are suitable elliptic solutions to the cgle5 . the function @xmath63 has the form ( [ mmm ] ) , the values of constants can be determined from ( [ equm4 ] ) . summing up we can conclude that our modification of the conte musette method allows us to get two results : we obtain new elliptic wave solutions of the cgle5 , and we prove that these solutions are unique elliptic solutions for the cgle5 with @xmath147 . from ( [ gigr0 ] ) it follows that elliptic solutions can exist if @xmath56 or @xmath57 is equal to zero . for all nonzero values of @xmath56 and zero @xmath57 we have found elliptic solutions . in the case @xmath178 there is no elliptic solution . in the case of zero @xmath56 and nonzero @xmath57 we substitute the obtained laurent series solutions @xmath159 into eq . ( [ subequ ] ) with @xmath179 and do not obtain neither elliptic functions nor degenerate elliptic solutions . we hope that the more detail analysis of this case allows us to find all elliptic solutions for the cgle5 . the approach , which we have considered in the previous section , can be applicable to many nonintegrable systems . the best applicability is nonintegrable systems of autonomous nonlinear ordinary differential equations with so - called finiteness property @xcite : there is only a finite number of formal laurent series that satisfy the system . for such systems we can propose the following way to the search for elliptic solutions : 1 . calculate a few first terms of all solutions of system ( [ f ] ) in the form of the laurent series . 2 . choose the function @xmath14 , which should be elliptic . check should other functions be elliptic or not . 3 . using the residue theorem define values of numeric parameters at which the solution @xmath14 can be an elliptic function . define a minimal number of poles for candidates to elliptic solutions . using this number , choose a positive integer @xmath7 , define the form of eq . ( [ subequ ] ) and calculate the number of unknown coefficients @xmath13 . calculate the sufficient number of coefficients for all laurent series of @xmath14 and substitute the obtained coefficients into eq . ( [ subequ ] ) . this substitution transforms ( [ subequ ] ) into a linear and overdetermined system in @xmath8 with coefficients depending on numerical parameters . eliminate coefficients @xmath8 and get a nonlinear system in parameters . solve the obtained nonlinear system . we restrict ourselves to the search of elliptic solutions only . this makes possible to consider not one laurent series , but as much as possible . this is a key idea of our modification of the conte musette method . using the conte musette method in the original form , one in principal would be able to find the elliptic solutions of the cgle5 , but calculations would be cumbersome , because he should reobtain all known solutions in terms of elementary functions to obtain new elliptic solutions . one has to construct and to solve an algebraic system , which includes at least @xmath180 equations , which are linear in @xmath181 unknowns @xmath30 and nonlinear in @xmath182 parameters . in our approach , we seek only elliptic solutions and , using the hone s method , reduce the number of arbitrary numerical parameters to @xmath183 . consideration of a few laurent series , instead of one also simplifies the obtained system of algebraic equations . we conclude that the use of the conte musette method in combination with the hone s method is very effective . in contrast to @xcite we consider a system of differential equations instead of the equivalent single differential equation and demonstrate that the system is more convenient for analysis than a single equation for @xmath65 . if we analyse only the laurent series expansions of the function @xmath65 , then we should also consider the case when @xmath65 has not four , but only two different laurent series expansions , beginning from @xmath184 and @xmath185 or @xmath186 and @xmath187 . the consideration of the laurent series expansions of two functions : @xmath188 and @xmath65 , allows us to reject this possibility . note that in our opinion if one seek degenerate elliptic solutions using the laurent series then a system of equations is more convenient for analysis than the equivalent single equation as well . moreover , in contrast to traditional methods when we use the conte musette method we can choose a function , which analytic form should be found . instead of the functions @xmath188 and @xmath65 we can consider some combination , for example , a polynomial , of this functions and seek this combination in the analytic form . note that we have no need of any differential equation for this combination . so , we can conclude that the use of the laurent series solutions gives new insight on problem of construction of exact solutions for nonintegrable systems . let us compare our approach with other methods for construction of special solutions of nonintegrable systems ( see @xcite as a review of these methods ) . without the use of the laurent series solutions it would be very difficult to find the elliptic solutions of the cgle5 , because the form of eq . ( [ equm ] ) is very complex . for example , the fan s technology @xcite allows to find solutions of nonintegrable systems , which are polynomials of the function @xmath189 , that satisfies the following equation : @xmath190 where @xmath191 and @xmath192 are constants to be determined . the function @xmath188 is a solution of a similar equation with @xmath193 , but , using the fan s approach , it is impossible to find @xmath194 without knowledge of @xmath63 , which satisfies the more complex equation ( [ equm ] ) . the kudryashov s method @xcite as well as the methods @xcite , proposed to search both elliptic and elementary solutions , attempt to find solutions for the initial differential system at all values of the numeric parameters such that single - valued solutions can exist . the use of the residue theorem and the hone s method allows us to fix some of these parameters and to simplify calculations without loss of elliptic solutions . note that we not only find elliptic solutions for the cgle5 , but also we prove that there are no other elliptic solutions for @xmath195 . using the standard methods one can say nothing about uniqueness of the obtained special elliptic solutions . to find solutions for the cgle5 the authors of papers @xcite put some restrictions on dependence between @xmath188 and @xmath65 . the use of laurent solutions allows us to search @xmath188 without any restrictions and without eliminating @xmath65 from system ( [ system ] ) . in this paper we propose a new approach for the search of elliptic solutions to systems of differential equations . the proposed algorithm is a modification of the conte musette method @xcite . we restrict ourselves to the search of elliptic solutions only . a key idea of this restriction is to simplify calculations by means of the use of a few laurent series solutions instead of one and the use of the residue theorem . the application of our approach to the quintic complex one - dimensional ginzburg landau equation ( cgle5 ) allows to find elliptic solutions in the wave form . let us point out that the obtained solutions are the first elliptic solutions for the cgle5 . using the investigation of the cgle5 as an example , we demonstrate that to find elliptic solutions the analysis of a system of differential equations is more preferable than the analysis of the equivalent single differential equation . we also find restrictions on coefficients , which are necessary conditions for the existence of elliptic solutions for the cgle5 . to do this we develop the hone s method @xcite . we show that this method is useful not only to prove the nonexistence of elliptic solutions , but also to find new elliptic solutions . note that the hone s method and , therefore , our approach , are so effective in the case of the cgle5 , because coefficients of the laurent series solutions depend only on parameters of equations , i.e. they do not include additional arbitrary parameters ( have no resonances ) . it is an important problem to generalize the hone s method on the laurent series solutions with resonances . another way for future investigations is the generalization of the conte musette method on the case of multivalued solutions . some results in this direction have been obtained in @xcite . this work has been supported in part by russian federation president s grant nsh1685.2003.2 . ablowitz , a. ramani , h. segur , a connection between nonlinear evolution equations and ordinary differential equations of p - type . i @xmath196 ii , _ j. math . * 21 * ( 1980 ) 715721 , @xmath196 10061015 . d. baldwin , w. hereman , symbolic software for the painleve test of nonlinear ordinary and partial differential equations , _ j. nonlinear math . * 13 * ( 2006 ) no . 1 , 90110 ; nlin.si/0505004 . b. buchberger , theoretical basis for the reduction of polynomials to canonical forms , _ sigsam bulletin _ * 39 * ( 1976 ) 1929 . b. buchberger , some properties of grbner bases for polynomial ideals , _ sigsam bulletin _ * 40 * ( 1976 ) 1924 . r. conte , m. musette , solitary waves of nonlinear nonintegrable equations , chapter in the book `` dissipative solitons '' ( lect . notes phys . , vol . 661 ) , eds . n.n . akhmediev and a. ankiewicz , springer , berlin , 2005 ; nlin.ps/0407026 v.p . gerdt , yu.a . blinkov , involutive bases of polynomial ideals , _ math . * 45 * ( 1998 ) 519542 ; math.ac/9912027 ; minimal involutive bases , _ math . * 45 * ( 1998 ) 543560 ; math.ac/9912029 . v.p . gerdt , involutive algorithms for computing grbner basis , in `` computational commutative and non - commutative algebraic geometry '' , s. cojocaru , g. pfister , v. ufnarovski ( eds . ) , nato science series , iop press , 2005 , pp . 199225 ; math.ac/0501111 . ginzburg , l.d . landau , on the theory of superconductors , _ jetp ) _ * 20 * ( 1950 ) 10641082 ; english translation in l.d . landau , _ collected papers _ , oxford , pergamon press , 1965 , p. 546 . kudryashov , nonlinear differential equations with exact solutions expressed via the weierstrass function , nlin.cd/0312035 . kudryashov , simplest equation method to look for exact solutions of nonlinear differential equations , nlin.si/0406007 . n.a . kudryashov , painlev property and the first integrals of nonlinear differential equations , nlin.si/0408041 . m. musette , r. conte , analytic solitary waves of nonintegrable equations , _ physica d _ * 181 * ( 2003 ) 7079 ; nlin.ps/0302051 . nikitin , t.a . barannyk , solitary wave and other solutions for nonlinear heat equations , math - ph/0303004 . w. van saarloos , p.c . hohenberg , fronts and pulses in the complex ginzburg landau equation near a subcritical bifurcations , _ phys . * 64 * ( 1990 ) 749752 w. van saarloos , p.c . hohenberg , fronts , pulses , sources and sinks in generalized complex ginzburg landau equations , _ physica d _ * 56 * ( 1992 ) 303367 , erratum * 69 * ( 1993 ) 209 g.s . santos , application of finite expansion in elliptic functions to solve differential eduations , _ j. of the physical society of japan _ * 58 * ( 1989 ) 43014310 . c. scheen , implementation of the painlev test for ordinary differential equations , _ theor . comp . * 187 * ( 1997 ) 87104 . m. tabor , chaos and integrability in nonlinear dynamics , wiles , new york , 1989 . timoshkova , a new class of trajectories of motion in the hnon heiles potential field , _ rus . j. _ * 76 * ( 1999 ) 470475 , \{in russian } , _ astron . rep . _ * 43 * ( 1999 ) 406411 , \{in english}. s.yu . vernov , e.i . timoshkova , on two nonintegrable cases of the generalized hnon heiles system , _ physics of atomic nuclei ( yadernaya fizica ) _ * 68 * ( 2005 ) 19471955 \{in english } ; 20082016 \{in russian } ; math - ph/0402049 . vernov , constructing solutions for the generalized hnon heiles system through the painlev test , _ theor . ( tmf ) _ * 135 * ( 2003 ) 409419 \{in russian } , 792801 \{in english } ; math - ph/0209063 . vernov , construction of special solutions for nonintegrable systems , _ j. of nonlin . * 13 * ( 2006 ) no . 1 , pp . 5063 . s.yu . vernov , packages of procedures for maple 9 ( ellipso ) and reduce 3.7 , 2005 , http://theory.sinp.msu.ru/~svernov/programs/ s.yu . vernov , construction of single - valued solutions for nonintegrable systems with the help of the painlev test , in : _ proceedings of the international conference `` computer algebra in scientific computing '' _ ( st . petersburg , russia , 2004 ) , v.g . ganzha , e.w . mayr , e.v . vorozhtsov ( eds . ) , technische universitat , munchen , garching , 2004 , pp . 457465 ; nlin.si/0407062 . vernov , interdependence between the laurent - series and elliptic solutions of nonintegrable systems , in : _ the lecture notes of computer science ( springer ) lncs _ , v. * 3718 * ( the proceedings of international conference `` computer algebra in scientific computing '' ( casc 2005 , kalamata , greece ) ) , editors : v.g . ganzha , e.w . mayr and e.v . vorozhtsov , springer verlag , berlin , germany , 2005 , pp . vernov , proof of the absence of elliptic solutions of the cubic complex ginzburg landau equation , _ theor . ( teor . mat . fiz . ) _ * 146 * ( 2006 ) 131139 \{in english } , 161171 \{in russian } ; preprint version : on elliptic solutions of the cubic complex one - dimensional ginzburg landau equation , nlin.ps/0503009 . vernov , construction of exact particular solutions of nonintegrable systems by means of formal laurent and puiseux series , _ programming and computer software _ , * 32 * ( 2006 ) no . vladimirov , e.v . kutafina , exact travelling wave solutions of some nonlinear evolutions equations , _ rep . phys . _ * 54 * ( 2004 ) 261271 . xu , z.b . li , symbolic computation of the painlev test for nonlinear partial differential equations using maple , _ comp . * 161 * ( 2004 ) no . ( 1 - 2 ) , pp .

the conte musette method has been modified for the search of only elliptic solutions to systems of differential equations . a key idea of this a priory restriction is to simplify calculations by means of the use of a few laurent series solutions instead of one and the use of the residue theorem . the application of our approach to the quintic complex one - dimensional ginzburg landau equation ( cgle5 ) allows us to find elliptic solutions in the wave form . we also find restrictions on coefficients , which are necessary conditions for the existence of elliptic solutions for the cgle5 . using the investigation of the cgle5 as an example , we demonstrate that to find elliptic solutions the analysis of a system of differential equations is more preferable than the analysis of the equivalent single differential equation . standing wave , the elliptic function , the laurent series , the residue theorem , the quintic complex one - dimensional ginzburg landau equation 05.04.-a , 02.30.-f , 02.70.wz , 47.27.-i

introduction the contemusette method for the system of differential equations properties of the elliptic functions construction of elliptic solutions construct of elliptic solutions for nonintegrable systems comparison with standard methods conclusions acknowledgements

the local white dwarf population preserves detailed information on the galactic star formation history . nearly 95% of all stars will end their lives as white dwarfs , and most @xmath1 stars ever formed are now stellar remnants . white dwarf cooling is well understood in terms of the underlying physics @xcite , and has been used to estimate the age of the galactic disc ( e.g. @xcite ) , individual open clusters ( e.g. @xcite ) , and the halo @xcite . broader application of this method has been prevented so far by the small and incomplete samples of known white dwarfs . because of their small radii , white dwarfs are intrinsically faint , requiring moderately large telescope apertures for their study . consequently the currently available luminosity functions contain at best a few thousand stars @xcite , based on white dwarfs identified by sdss . however , incompleteness and selection biases severely limit the constraints that can be drawn from these studies . @xcite demonstrated the potential of a complete and well - characterised white dwarf sample to derive the local star formation history and initial mass function , using only the 117 white dwarfs known within 20pc of the sun , though obviously limited by small number statistics . combining parallax , apparent magnitude , proper motion and bp / rp colour , gaia will overcome all the traditional limitations in the discovery of white dwarfs . the @xmath2 white dwarfs that gaia will identify ( fig.[f : gaia_wds ] ) will be 100% complete within @xmath3pc , and @xmath3% out to 300pc @xcite . the sample of gaia white dwarfs will be sufficiently large to reconstruct the stellar formation history and initial mass function for the thin / thick disc and halo separately and assess the ages of these three components individually . however , while gaia will identify the entire local white dwarf population , it will not provide the data that is necessary to determine accurate masses and temperatures , which are the key parameters for measuring the white dwarf cooling ages . the spectral resolution of the bp / rp spectra is too low to apply the standard method for measuring temperatures and surface gravities from the stark - broadened balmer line profiles ( fig.[f : sdss1228 ] ; e.g. @xcite ) . this implies that ground - based follow - up is essential to unlock the diagnostic potential that the gaia white dwarfs have for investigating the galactic star formation history . to break the degeneracy between temperature and surface gravity requires spectroscopy covering the higher balmer lines down to @xmath4 nm ( e.g. * ? ? ? * ) , which are most sensitive to surface gravity , and hence mass . several of the forthcoming wide - area mos instruments are perfectly suited for this goal , including desi , weave , and 4most . gaia will also not obtain radial velocities of white dwarfs , as they typically have no spectral features in the rvs wavelength range , and are too faint for rvs spectroscopy anyway . to establish the galactic orbits , and hence thin / thick disc or halo membership , radial velocities of the gaia white dwarfs have to be obtained from the ground . this is most efficiently done by intermediate - resolution ( @xmath5 ) spectroscopy of the sharp nlte core in h@xmath6 @xcite . _ the initial - to - final mass ( ifm ) relation . _ the physics of mass loss on the rgb / agb is still poorly understood , and lacks stringent observational constraints . yet , it is of crucial importance for the galactic lifecycle of matter , and for the chemical evolution of galaxies ( e.g. * ? ? ? traditionally , the ifm relation has been investigated using white dwarfs in open clusters @xcite , and gaia will significantly add to the known population of white dwarfs in open clusters . gaia , coupled with ground - based spectroscopy will also fully enable the potential of wide binaries to constrain the low - mass end of the ifm relation @xcite . _ the mass - radius ( m - r ) relation . _ despite its fundamental importance , from predicting white dwarf masses to understanding the nature of snia , this relation is very poorly constrained by observations @xcite . spectroscopic observations of white dwarfs can not directly determine both mass and radius , and therefore parallaxes are needed to independently determine the radius . currently only a handful of white dwarfs have parallaxes accurate to better than 5% , but gaia will improve this situation dramatically , both in terms of raw numbers of white dwarf parallaxes , as well as their accuracy . spectroscopic follow - up covering the higher balmer lines is essential to reach a few percent precision on both mass and radius . the internal stratification of white dwarfs can then be constrained by comparing the observed m - r relation with theoretical relations assuming different interior compositions ( 35% effect on the m - r relation ) . accurate knowledge of the core compositions will also vastly improve the derived cooling ages , which are essential for the galactic archeology applications . _ rare white dwarfs . _ sdss has thoroughly demonstrated the _ huge _ discovery potential of a large spectroscopic survey of white dwarfs . follow - up of the gaia white dwarfs will be essential to improve our understanding of rare evolution channels such as c / o - atmosphere white dwarfs , that may be remnants of s - agb stars ( e.g. @xcite ) , strongly magnetic white dwarfs that can be used as laboratories for physics under extreme conditions @xcite , and dynamically active evolved planetary systems ( e.g. @xcite , see fig.[f : sdss1228 ] ) . throughout the last decade , it has become clear that white dwarfs serve not only as tracers of stellar evolution and star formation , but provide equally important information on the formation , structure , and evolution of planetary systems . many white dwarfs ( @xmath7% ) show signatures of planetary systems in the form of metal - pollution of their otherwise gravitationally settled pure - h / he atmospheres ( e.g. @xcite ) . these metals , accreted from disrupted planetary bodies , provide the only method to directly measure the bulk composition of extra - solar planetary systems , including the detection of water - rich planetesimals ( e.g. @xcite ) . gaia follow - up spectroscopy will be essential to increase the known sample of metal - polluted white dwarfs . the strongest tracer of planetary debris is the cah / k doublet , again requiring blue coverage down to @xmath4 nm at intermediate - resolution to detect the most polluted objects . _ snia progenitors . _ although double - degenerate white dwarf binaries are one of the likely populations to produce snia , the current census of these stars is utterly inadequate to test population models and predictions of snia rates and delay time distributions @xcite . the largest dedicated search for double - degenerates covered only a few hundred stars @xcite , and found a close binary fraction of @xmath8% . time - resolved spectroscopy of the gaia white dwarfs using desi / weave/4most sub - spectra has the potential to identify 1000s of close double - degenerates , spanning the entire parameter space in mass , mass ratio , and orbital period . the key goal is to obtain low resolution ( @xmath5 ) spectroscopy of the entire local white dwarf population identified by gaia to maximize the number of thick disk and halo white dwarfs , as well as massive ( @xmath9 ) white dwarfs in the thin disk . massive white dwarfs make up @xmath10% of the population ( e.g. @xcite ) , and a luminosity function of @xmath11 of these stars will determine the star formation rate of short - lived a / b stars to within @xmath12% throughout the age of the thin disk with a time resolution of a few 100myr @xcite . the necessary number of fibres is small , @xmath13 ( fig.[f : gaia_wds ] ) , only @xmath14% of the fibres available in each desi , weave , or 4most field . white dwarfs are intrinsically faint , and hence to zeroth order isotropically distributed , and should be multiplexed into all wide - area spectroscopic surveys to maximise the total survey area / number of observed targets . the bulk of the gaia white dwarf population will have @xmath15 . given the proximity of most white dwarfs , even at the faint end , the gaia parallaxes will still be spectacularly good ( 5% at 100pc for the faintest white dwarfs in the sample ) . based on our experience with sdss ( 2.5 m aperture ) , a mos on a 4 m aperture should deliver a signal - to - noise ratio of @xmath16 in a typical 1h exposure , which is amply sufficient for the stellar parameter determination . also based on sdss ( @xmath17 ) , we expect a radial velocity uncertainty from the @xmath18 line cores measured from desi , weave , or 4most low - res spectra to be @xmath19km / s , well sufficient to determine the galactic population membership . the gaia - based selection of white dwarfs relies on the second data release ( positions , proper motions , parallaxes , integrated xp photometry , and @xmath20-magnitudes ) expected in early 2017 . because of their small radii ( @xmath21 ) , white dwarfs will be extreme outliers in the gaia hertzsprung - russell diagram , and their identification will be trivial and free of contaminants . should there be a delay in this data release , we will fall back onto an efficient multi - colour plus reduced proper motion selection that works very well in the sdss footprint @xcite . we are in the process of extending this selection to the southern hemisphere making use of the vst / atlas survey , as well as panstarrs 3pi and skymapper in the near future . higher - resolution ( @xmath22 ) weave and 4most spectroscopy of the brightest gaia white dwarfs ( @xmath23 ) would help to increase the radial velocity resolution obtained from h@xmath6 , and to improve the sensitivity to metal pollution in cah / k , or mgii 447 nm ( fig.[f : sdss1228 ] ) .

gaia will identify several @xmath0 white dwarfs , most of which will be in the solar neighborhood at distances of a few hundred parsecs . ground - based optical follow - up spectroscopy of this sample of stellar remnants is essential to unlock the enormous scientific potential it holds for our understanding of stellar evolution , and the galactic formation history of both stars and planets .

white dwarfs as tracers of the galactic star formation history additional science goals survey requirements

surveys in the far - infrared ( fir ) and sub - millimetre ( submm ) region from future space missions will provide crucial information on star formation and galaxy formation and evolution . if agn were in place roughly at the same time as galaxies , it is natural to expect that these surveys will shed light on the agn role in the early evolution of galaxies @xcite . both agn and star formation are characterized by a strong fir emission both in the continuum and in molecular and atomic lines , because of the dusty and gaseous environment where they release their energy . an early dust - enshrouded phase of agn is expected , but is missed in current surveys based on other techniques . the optical agn phase could occur late in the process of qso formation , corresponding to the exhaustion of the gaseous - dusty shroud inside which the host galaxy formed . the original 12@xmath0 m active galaxy sample @xcite provided a complete and largely unbiased sample of local active galaxies . it is relatively free from the strong selection effects of optical - uv surveys , which are biased in favor of type 1 agn , and far - ir surveys , which are biased in favor of dusty type 2 agn and starburst galaxies . it is one of the largest homogeneously - selected samples of seyfert galaxies available , and appears to contain representative populations of galaxies with type 1 and type 2 optical spectra ( see also tran 2001 , thean et al . , 2001 ) . the selection was done at 12@xmath0 m because this band contains a constant fraction of bolometric flux for active galaxies ( @xmath2 1/7 ) . the 12@xmath0 m selection is therefore the best approximation so far available to a selection at a bolometric flux limit , for the different types of active galaxies . the resulting active galaxy sample contains 53 seyfert type 1 galaxies ( hereafter seyfert 1 s ) , 63 seyfert type 2 galaxies ( hereafter seyfert 2 s ) , two blazars and 38 high - luminosity non - seyfert galaxies out of a sample of 893 galaxies ( rms ) . little is known about the fir properties of type 1 higher redshift objects . because of their faintness in this energy domain they lack homogeneous samples and information is still scant @xcite . type 2 agn are much more difficult to find , and are missed in classical searches . in this work we make use of a homogenous data sample of optically selected type 1 agn @xcite which was observed at far - ir wavelengths . we make predictions of number counts and luminosities that future fir and submm space missions sirtf ( space infrared telescope facility , nasa ) and the herschel space observatory ( herschel , esa)will be able to detect . for local galaxies we adopted the spectral energy distributions ( seds ) of homogenous subsamples extracted from the 12@xmath0 m galaxy sample : seyfert 1 s , seyfert 2 s , starburst and normal galaxies for which infrared space observatory ( iso ) photometry was collected @xcite . for higher redshift type 1 agn , we chose the optically selected sample at a mean redshift of @xmath3=0.7 observed in the fir by andreani et al . as it is well known , number counts of extragalactic objects are a fundamental tool for investigating evolution , since they are related to the luminosity function of the population , as a function of redshift . the differential number counts , @xmath4 , ( sources / unit flux interval / unit solid angle ) can be expressed as an integral of the epoch - dependent luminosity function , @xmath5 : where @xmath7 and @xmath8 are the effective upper and lower limits of the redshift distribution , and @xmath9 is the differential comoving volume . the integrated counts are found by integrating eq.[eq : diffcoun ] to different flux limits , @xmath10 . to predict the differential and integral number counts of active galaxies and quasars we need to compute equation [ eq : diffcoun ] with the following assumptions : * the geometry of the universe : we use two different models both with h@xmath11=75 km / s / mpc : ( 1 ) @xmath12 , @xmath13 and ( 2 ) @xmath12 , @xmath14 and @xmath15 . they have been chosen because their predictions represent two extreme cases . * the seds of the populations : we take the empirical optical - far - ir seds of active galaxies and qsos ( see [ sec : sed ] ) and assume that they are luminosity - independent . the only difference between normal and starburst galaxies is the luminosity break . this simplified assumption avoids more complicated situations where the sed continuously changes with luminosity . * the local luminosity function ( llf ) : we consider different llfs as empirically determined for the different classes : agn , starburst and normal galaxies ( see [ sec : lf ] ) . * the form of the lf evolution : different evolution laws are used for each class ( see [ sec : ev ] ) . in the following sections we discuss each of these assumptions . based on the results of spinoglio , andreani and malkan ( 2002 ) , figure [ fig : agnsed ] shows the observed average sed of the different classes of galaxies in the local universe , normalized to 12@xmath0 m . seyfert 1 s have the flattest sed throughout the entire wavelength range , with the relatively weakest fir emission . seyfert 2 s show two pronounced peaks : at 100@xmath0 m and 1.6@xmath0 m . starburst galaxies show the brighter peak at 60@xmath0 m and the weaker one at 1.6@xmath0 m , showing that recent and ongoing star formation detected in the fir outshines the old stellar populations , which peak in the h band . normal galaxies show again the same two peaks . for this work we need to assume an evolutionary law for agn , but seyfert 1 s lack of a good empirical determination of their redshift evolution . fortunately , the evolution of their higher redshift analogs type 1 qsos is well determined . their far - ir sed is however less well defined . in this work we take the composite sed derived from optical , near - ir and fir photometry of an optically selected sample of qsos . figure [ fig : sedcomp ] shows the average spectrum in the qso restframe , normalized at @xmath16 m . it has a well defined ir component peaking at 10 - 30 @xmath17 and dropping steeply towards 100@xmath0 m . the errorbars shown are those related to the averages , but also include photometric uncertainties . seyfert 1 s and qsos belong to the same class of objects , type 1 agn , with seyfert galaxies populating the low luminosity end of the luminosity function . figure [ fig : sedcomp ] shows also a comparison of the observed seyfert 1 s sed with that of qsos . within the errorbars , no large differences are present , both below a few microns and above 100@xmath0 m . the qso thermal spectrum has a steeper turnover in the fir range , and more power between 10 and 30@xmath0 m compared to the seyfert 1 s sed . qso sed shows higher average dust temperature ( 100 - 150 k ) , with respect to seyfert 1 s ( 22 - 70k ) @xcite . because of their selection criteria , this difference could be ascribed to a luminosity effect . we lack high - z object data to disentangle this effect from an evolutionary behaviour . note that the qso sed is quite different from seyfert 2 s . this latter has a large far - ir emission shifted to longer wavelengths , and a second bump in the near - ir , which is not present in qsos ( see figure [ fig : agnsed ] ) . to compute the k - corrections , i.e. @xmath18 in equation [ eq : diffcoun ] , we use for each class its appropriate composite spectrum as shown in figures [ fig : agnsed ] and [ fig : sedcomp ] . for the local luminosity function ( llf ) , @xmath19 , we adopt the 12@xmath0 m llf for seyfert , starburst and normal galaxies ( rms ) . there is some uncertainty about the space densities at the low luminosity end , because no explicit correction for the over - concentration of local galaxies in virgo was applied . fang et al . ( 1998 ) and xu et al . ( 1998 ) provide slightly different parameters , but all 3 of those 12@xmath0 m llfs are in reasonable agreement around l@xmath20 galaxies , which dominate the counts . we use the same llf shape with the different parameters given by rms for the different classes : seyfert 1 s and 2 s , and non - seyferts ( starburst and normal galaxies ) . this 12@xmath0 m seyfert 1 lf is in principle supposed to include quasars in it , as it was computed for a sample that includes 3c 273 . as a practical limitation , though , it is too insensitive and thus covers too small a volume to define the highest end of the sy1/qso lf . it also does not take into account the agn evolution which is , on the contrary , included in the qso lf . therefore , to extend the sy1 lf to higher luminosities , and to include the luminosity - number density evolution of these objects we use the b - band luminosity function computed by boyle et al . ( 2001 ) ( see also la franca and cristiani , 1998 ) . we assume that a constant relation exists between the b - band lf and the far - ir one . this latter has been applied to both type 1 agn ( seyfert 1 s and qso ) . the seyfert 1 s and qso used in this work were selected in different ways , and the corresponding lf differs slightly in shape . in figure [ fig : lf ] the qso lf is shown ( solid line ) superposed on that of the seyfert 1 s ( dotted line ) . both lfs are computed at a redshift of 0.3 , the minimun redshift for which a reliable qso lf exists . the seyfert 1 llf was evolved from redshift 0 to a redshift of 0.3 , according to the prescriptions described in [ sec : ev ] . as shown in figure [ fig : lf ] the bright end of the qso lf is slightly steeper ( the slope exponent being -2.5 ) than seyfert 1 s ( -2.2 ) . however , within the errorbars this difference is not significant . the knee at l@xmath20 differs and that of qsos falls at higher luminosity ( @xmath21 l@xmath2210.5 ) . at the faintest luminosities , the two lf slopes coincide but the amplitudes differ by a factor of 10 . this high luminosity knee and the lower numbers at lower luminosities for qsos can be explained since nearly all qso samples are based on integrated properties of stellar - looking objects . when the host galaxy starts becoming important , the object is no longer classified as a qso , but rather a seyfert1 . the best approximation of the type 1 luminosity function would be to adopt ( see figure [ fig : lf ] ) the qso lf above @xmath23 and then the relatively steep seyfert 1 s below that , where they cross over . thus , for the sake of completeness we use , when computing the expected number counts for seyfert 1 , two lfs : that of qso and that given by rms for seyfert 1 . [ cols="^,^,^,^,^",options="header " , ] [ fit_coef ] . evolution from one class of galaxies to another has not been considered , as well as other speculative assumptions , because we want to keep this work as much as possible on a firm empirical basis . * in this paper we predict , on an empirical basis , active and normal galaxy counts which future space missions will be able to measure in the far - ir . the observed seds of different types of objects ( seyfert 1 and 2 , starburt , normal galaxies and qsos ) are evolved backwards in time , using our present knowledge of their local luminosity functions and their evolution . * the number of seyfert 1 and 2 , starbursts and normal galaxies that should be detected in future surveys are shown in tables [ tab : counts1 ] and [ tab : counts2 ] and figures [ fig : counts ] and [ fig : diffcounts ] . in the most conservative case ( a flat universe @xmath13 , @xmath12 ) , sirtf and herschel will be able to detect from thousands to tens of thousands of active and starburst galaxies in a field of 0.3 and 1 deg@xmath24 , respectively . our results are affected by the following uncertainties : * * type 1 qsos : their evolution has a solid foundation since the b - band lf is well known at least up to redshift 3 . their far - ir sed is however very uncertain . we use the composite spectrum shown in figure [ fig : sedcomp ] whose photometric uncertainty is large ( see andreani et al . , 2003 for details ) . * * nearby seyfert and starburst galaxies on the contrary have a well - defined average sed , but the knowledge of their evolution is poor . future far - ir space missions , mainly sirtf , will eliminate this uncertainties . * colour - colour diagrams are constructed from the seds shown in figures [ fig : sedcomp ] and [ fig : agnsed ] , without evolution taken into account . curves are very uncertain for qsos , because of their poorly determined far - ir sed . colour - colour diagrams with the selected frequency bands will be a powerful tool for discriminating among different object types : quasars from seyfert 1 s , seyfert 2 s and starburst galaxies . redshift measurements ( or estimates ) can be used to resolve the degeneracy present in some cases . * the confusion - limiting flux tends to be high compared with the limiting sensitivity of both sirtf and herschel instruments . instrument teams are indeed seriously considering it for very long exposures . * sirtf will make fundamental discoveries about agn evolution , exploiting mainly the 8 and 24@xmath0 m bands . these bands are critical to selecting unbiased samples of type 1 and 2 agn . pa acknowledges the infrared group of the max - planck institut fr extraterrestrische physik of garching ( germany ) for hospitality during the first stage of this work . the italian space agency ( asi ) has funded part of this work under the contract asi - i - r-037 - 01 . andr p. , ward - thompson d. , barsony m. , 2000 protostars & planets iv , eds . v. mannings , a. boss , s. russell ( tucson : the univ ersity of arizona press ) , 59 . et al . , 2003 , 125 , 444 bianchi s. , cristiani s. , tae - sum k. , 2001 , 376 , 1 boyle b.j . et al . , 2001 , 317 , 1014 chapman s.c . , smail i. , ivison r.j . , helou g. , dale d.a . , and lagache g. , 2002 , 573 , 66 de zotti g. , franceschini a. , toffolatti a. , mazzei p. , danese l. , 1996 , astr . lett . com . 35 , 289 fang f. , shupe d.l . , xu c. , hacking p.b . , 500 , 693 ( san francisco ) franceschini a. , toffolatti a. , danese l. , de zotti g. , 1989 , 344 , 35 franceschini , a. 2000 , in `` galaxies at high redshift '' , f.sanchez , i.perez-fournon , m.barcells , f.moreno-insertis eds . , cambridge university press , in press ( astro - ph/0009121 ) granato g.l . , danese l. , 1994 , 268 , 235 granato et al . , 2001 , 324 , 757 hooper e.j . et al . , 1999 , in proc . of `` the universe as seen by iso '' , ed . m.kessler , esa - sp 427 lagache g. , dole h. , puget j.l . , in press , astro - ph/0209115 la franca f. , & cristiani s. , 1998 , 113 , 1517 , and 115 , 1688 malkan m.a . & stecker f.w . 1998 , 496 , 13 malkan m.a . & stecker f.w . 2001 , 555 , 641 martn e.l . , brandner w. , jewitt d.c . , simon t. , wainscoat r. , connelley m , 2001 , pasp 113 , 529 nielbock m. , chini r. , jtte m. , manthey e. , 2001 , 377,273 polletta m. , courvoisier t.j .- l . , hooper e.j . , wilkes b.j . , 2000 , 362 , 75 rush b. , malkan m.a . & spinoglio l. , 1993 , apjs 89,1 ( rms ) silk j. and rees m. , 1998 , 331 , l1 spinoglio l. and malkan , m.a . 1989 , 342 , 83 spinoglio l. et al . 1995 , 453,616 spinoglio l. , andreani p. & malkan , m.a . 2002 , 572 , 105 thean a. et al 2001 , 325 , 737 tran h.d . 2001 , 554 , l19 visnen p. , tollestrup e.v . , fazio g.g , 2001 , wilkes b.j . et al .. , 1999 , in proc . of `` the universe as seen by iso '' , ed . m.kessler , esa - sp 427 xu c. , hacking p. , fang f. et al . 1998 , 508 , 576

averaged spectral energy distributions ( seds ) of _ active and starburst galaxies _ from the 12@xmath0 m sample in the local universe and _ quasars _ , from an optically selected sample at a mean redshift @xmath1=0.7 are built from optical , near - ir and far - ir ( iras and iso ) photometric observations . these seds are then used to predict at various redshifts the number of seyfert type 1 and type 2 , starburst , normal galaxies , and quasars , that will be detected by future space missions dedicated to far - infrared and submillimeter astronomy , like sirtf and herschel . these predictions are then compared with the expected capabilities and detection limits of future deep far - ir surveys . possible ways to identify agn candidates on far - ir colour - colour plots for follow - up observations are then explored .

introduction expected counts of type 1 agn and active galaxies conclusions

the assembly of colloidal building blocks with designer engineered size , shape , and chemical anisotropy seems to be the next step in the fundamental and applied colloid science and science of soft materials , vigorously pursued by many researchers @xcite . moving beyond the traditional systems implies the creation of designer colloidal building blocks such as colloidal molecules , janus spheres , and other patchy particle motifs that have very different propensities for self - assembly @xcite . in order to realize this goal one needs detailed control over various aspects of colloid geometry at the nanoscale ( _ e.g. _ , aspect ratio , faceting , branching , roughness ) and microscale ( _ e.g. _ , chemical ordering , shape gradients , unary and binary colloidal `` molecules '' ) . the ability to spatially modify the surface structure of colloids with designed chemical heterogeneity , _ e.g. _ , to form a patchy surface structure , seems to be becoming a realistic goal . fabrication of stable anisotropic microcapsules was recently accomplished by the layer - by - layer polyelectrolyte adsorption technique combined with particle lithography technique to produce anisotropic polymer microcapsules with a single nanoscale patch @xcite . granick and co - workers @xcite recently reported a highly scalable synthetic pathway for creating bipolar janus spheres , _ i.e. _ , particles that consist of oppositely charged hemispheres . such colloidal animals " which are probably the simplest example of patchy colloids , exhibit orientation - dependent interactions that go together with localized patches of like / unlike charges or hydrophobic / hydrophilic regions . this heterogeneous interaction landscape promotes the formation of larger colloidal molecules and clusters that are themselves patchy . these advances in the nano- and microscopic tailoring of ( charged ) colloids motivated various approaches to generalizations of the existing theories of electrostatic interactions in charged colloids by explicitly including the structure of the counterions as embodied by their multipolar moments . the inclusion of structured counterions of a dipolar @xcite and quadrupolar @xcite type into the theory of electrostatic colloidal interactions has brought fourth some of the salient features of the counterion structure effects , which on face value appear to be quite distinct from the standard poisson - boltzmann framework . in view of these advances in the study of interactions between charged colloids it thus seems appropriate to explore the ramifications of the emerging paradigm if applied to these more complicated structured colloidal molecules . especially the non - spherically symmetric charge distribution of microscopically tailored colloidal particles might lead to some unexpected properties of electrostatic interactions in this type of systems . with this in mind we thus embark on a thorough examination of the consequences of multipolar charge distribution of mobile counterions that mediate interactions between charged ( planar ) macroions . in what follows we will consider a system of fixed macroions of surface charge distribution @xmath0 in an aqueous solution , described as a dielectric continuum with a dielectric constant @xmath1 and temperature @xmath2 , containing @xmath3 neutralizing counterions . counterions are assumed to be pointlike particles but they do posses a rigid internal structure described by a charge distribution @xmath4 that we assume can be written as a standard multipolar expansion @xcite @xmath5 for the @xmath6-th counterion located at @xmath7 with @xmath8 being the orientational variables specifying the angular dependence of the counterion charge density . the monopolar moment of each counterion is @xmath9 , where @xmath10 is the charge valency and @xmath11 the elementary charge , @xmath12 is the dipolar moment and @xmath13 the quadrupolar moment with director unit vector @xmath14 . we must emphasize here that the quadrupolar expansion in ( [ chaden ] ) is not general but adequately describes only a uniaxial counterion , _ e.g. _ , a charged particle of rod - like structure , fig . [ geom ] . one possible implementation of counterions possesing just monopolar and quadrupolar moment is a uniformly charged rod with charge @xmath9 and length @xmath15 . counterion s quadrupolar moment is then @xmath16 . another possibility is a negative charge ( @xmath17 ) in the center and two positive charges ( @xmath18 ) located at both ends of the rod leading to monopolar and quadrupolar moments of @xmath19 and @xmath20 , respectively . counterions interact _ via _ a coulomb interaction potential @xmath21 so that the interaction energy of two given counterions @xmath6 and @xmath22 at set positions and orientational configurations is obtained by integrating the coulomb interaction over their internal orientational degrees of freedom as @xmath23 although our formalism in this section is in general applicable to macroions of arbitrary shape and charge distribution , we shall primarily focus on the case of two charged planar surfaces located at @xmath24 with the charge distribution @xmath25 we also consider a dielectric inhomogeneity between the bounding surfaces and the ionic solution , such that the ionic solution is described with dielectric constant @xmath1 and the bounding surfaces with @xmath26 . in this geometry the coulomb interaction potential is composed of the direct interaction @xmath27 and the ( electrostatic ) image interaction @xmath28 , so that @xmath29 the green s function in planar geometry can be expressed as a sum of image charge contributions , _ viz . _ @xmath30 . \label{green}\end{aligned}\ ] ] in both sums the index @xmath31 runs through negative and positive integer values . the term corresponding to @xmath32 represents the direct green s function of the particle in a space of a uniform dielectric constant , namely @xmath33 . the unit vector @xmath34 points in the @xmath35 direction . the relative dielectric jump at the two bounding surfaces is quantified as @xmath36 . for most relevant situations the outer permittivity is smaller than the inner one , @xmath37 , so that the dielectric jump is positive , @xmath38 and vanishes in the homogeneous case with @xmath39 . the canonical partition function @xmath40 of this coulomb fluid composed of @xmath3 charged particles is given by @xmath41 \bigr),\end{aligned}\ ] ] with the configurationally dependent ionic interaction energy given by @xmath42 = \nonumber\\ & & \frac{1}{2}\sum_{i\ne j}\int\!\!\!\!\int \rmd{{\mathbf r } } \rmd{{\mathbf r } } ' \hat \rho({{\mathbf r } } ; { { { \mathbf r}}}_i , { \mbox{\boldmath$\omega$}}_i)\,u({{\mathbf r}},{{\mathbf r}}')\ , \hat \rho({{\mathbf r } } ' ; { { { \mathbf r}}}_j , { \mbox{\boldmath$\omega$}}_j ) + \nonumber\\ & & + \sum_i \int\!\!\!\!\int \rmd{{\mathbf r } } \rmd{{\mathbf r } } ' \hat \rho({{\mathbf r } } ; { { { \mathbf r}}}_i , { \mbox{\boldmath$\omega$}}_i)\,u({{\mathbf r}},{{\mathbf r}}')\rho_0({{\mathbf r } } ' ) + \nonumber\\ & & + \frac{1}{2 } \int\!\!\!\!\int \rmd{{\mathbf r } } \rmd{{\mathbf r } } ' \rho_0({{\mathbf r}})\ , u({{\mathbf r}},{{\mathbf r}}')\,\rho_0({{\mathbf r}}'),\end{aligned}\ ] ] where @xmath43 . the three terms in @xmath44 $ ] correspond to direct electrostatic interaction between counterions , electrostatic interactions between counterions and fixed charges and between fixed charges on the walls themselves , respectively . we will furthermore make the standard assumption that the system is overall electroneutral , implying that the mobile countercharge exactly compensates the charge on the surfaces . we now proceed in the netz @xcite fashion and perform the hubbard - stratonovitch transformation of the partition function @xcite , where the configurational integral over counterion positions is transformed into a functional integral over a fluctuating auxiliary electrostatic potential @xmath45 . in this way the grand - canonical partition function is obtained straightforwardly in the form @xcite @xmath46\,\rme^{-\beta h[\phi({{\mathbf r } } ) ] } , \label{funint}\ ] ] where @xmath47 is the bare fugacity which is the exponential of the chemical potential . the prefactor @xmath48 above is the functional determinant of the inverse coulomb kernel @xmath49 while the `` action '' of the functional integral is given by @xmath50&=&\!\!\int\!\!\rmd{{\mathbf r } } \bigg[{\textstyle\frac{1}{2}}\varepsilon({{\mathbf r}})\varepsilon_0 ( { \mbox{\boldmath$\nabla$}}\phi)^{2 } -\rmi\rho_0({{\mathbf r}})\phi({{\mathbf r } } ) - \nonumber\\ & & \hspace{-4ex}-\frac{\lambda'}{\beta}\!\int\ ! \rmd { \mbox{\boldmath$\omega$}}~\omega({{\mathbf r}})\exp{\left(\ ! \rmi\beta \!\!\int \!\ ! \rmd{{\mathbf r}}'\hat \rho({{\mathbf r } } ' ; { { \mathbf r } } , { \mbox{\boldmath$\omega$}})\phi({{\mathbf r}}')\!\right ) } \bigg ] . \label{action}\end{aligned}\ ] ] here we have introduced the renormalized fugacity as @xmath51 , where @xmath52 is the electrostatic direct self - energy of a single counterions with multipolar charge distribution . @xmath53 is the geometric characteristic function of the counterions , being equal to unity in the slab between the bounding surfaces and zero otherwise . the partition function in the above form can not be evaluated explicitly except in the one - dimensional case @xcite where the partition function evaluation is reduced to a solution of a schr " odinger - like equation . nevertheless the field - theoretical representation of the grand canonical partition function allows one to use quite powerful analytical approaches that eventually lead to an explicit evaluation of the partition function in two well defined and complementary limits @xcite that we shall address later in this paper . these limits retain the relevance also in the case of structured counterions . one may obtain a dimensionless representation for the present system by rescaling all length scales with a given characteristic length scale . recall that the characteristic distance at which two unit charges interact with thermal energy @xmath54 is known as the bjerrum length @xmath55 ( in water at room temperature , the value is @xmath56 nm ) . if the charge valency of counterions is @xmath10 then the aforementioned distance scales as @xmath57 . similarly , the distance at which a counterion interacts with a macromolecular surface of surface charge density @xmath58 with an energy @xmath54 is called the gouy - chapman length , defined as @xmath59 . a competition between ion - ion and ion - surface interactions can thus be quantified by the ratio ( @xmath60 ) of these characteristic lengths , that is @xmath61 which is referred to as the electrostatic coupling parameter . in what follows , we may rescale the length scales with the gouy - chapman length , _ i.e. _ , @xmath62 ; hence , the surface separation will be rescaled as @xmath63 or the rescaled half - distance as @xmath64 . other dimensionless quantities can be defined as follows . the dimensionless multipolar moments are defined as @xmath65 and the dimensionless pressure as @xmath66 for simple structureless counterions the functional integral ( [ funint ] ) can be rescaled @xcite yielding @xmath67\,\rme^{- h'[\phi({{\mathbf r}})]/\xi},\ ] ] with dimensionless @xmath68 $ ] . as will be shown later an identical representation can be obtained also for structured counterions , described on the level of a multipolar expansion of their charge density . in the absence of a general approach that would cover thoroughly all the regions of the parameter space one has to take recourse to various partial formulations that take into account only this or that facet of the problem @xcite . the traditional approach to these one - component coulomb fluids has been the mean - field poisson - boltzmann ( pb ) formalism applicable at weak surface charges , low counterion valency and high temperature @xcite . the limitations of this approach become practically important in highly - charged systems where counterion - mediated interactions between charged bodies start to deviate substantially from the mean - field accepted wisdom @xcite . one of the most important recent advances in this field has been the systematization of these non - pb effects based on the notions of _ weak _ and _ strong _ coupling approximations . they are based on the field - theoretical representation of grand canonical partition function ( [ action ] ) , whose behavior depends on a single dimensionless coupling parameter @xmath60 @xcite ( see below ) . the weak - coupling ( wc ) limit of @xmath69 coincides exactly with the mean - field poisson - boltzmann theory @xcite and is based on a collective description of the counterion density . the strong coupling ( sc ) limit of @xmath70 is diametrically opposite and corresponds to a single particle description . it has been pioneered by rouzina and bloomfield @xcite , elaborated later by shklovskii _ _ @xcite and levin _ et al . _ @xcite , and eventually brought into a final form by netz _ _ @xcite . these two limits are distinguished by the pertaining values of the coupling parameter @xmath60 which can be introduced in the following way . the regime of @xmath71 is the case of low counterionic valency and/or weakly charged surfaces , and is referred to as the weak - coupling limit . it is characterized by the fact that the width of the counterion layer @xmath72 is much larger than the separation between two neighbouring counterions in solution and thus the counterion layer behaves basically as a three - dimensional gas . each counterion in this case interacts with many others and the collective mean - field approach of the poisson - boltzmann type is completely justified . on the other hand in the strong coupling regime @xmath73 , which is true for high valency of counterions and/or highly charged surfaces . in this case the mean distance between counterions , @xmath74 , is much larger than the layer width ( _ i.e. _ , @xmath75 ) , indicating that the counterions are highly localized laterally and form a strongly correlated quasi - two - dimensional layer next to a charged surface . in this case , the weak - coupling approach breaks down due to strong counterion - surface and counterion - counterion correlations . since counterions can move almost independently from the others along the direction perpendicular to the surface , the collective many - body effects that enable a mean - field description are absent , necessitating a complementary sc description @xcite . these two approximations allow for an explicit and exact treatment of charged systems at two disjoint limiting conditions whereas the parameter space in between can be analysed only approximately and is mostly accessible solely _ via _ computer simulations . as will become clear in what follows the wc and sc limits remain valid also for structured counterions described with a multipolar charge distribution . the mean - field limit , @xmath69 , is defined _ via _ the saddle - point configuration of the hamiltonian ( [ action ] ) @xcite and is valid for a weakly charged system . this leads to the following generalization of the poisson - boltzmann equation @xmath76 } \label{pb}\end{aligned}\ ] ] for the real - valued potential field @xmath77 . note that the dielectric discontinuity at the boundaries is irrelevant within the mean - field theory @xcite since in the planar geometry this is effectively one - dimensional theory . taking now the counterion density function as a sum of the monopolar , dipolar and quadrupolar terms , where the mean - field depends only on the transverse coordinate @xmath78 , the above equation can be written in dimensionless form as @xmath79 where we have defined dimensionless potential @xmath80 with corresponding derivatives @xmath81 and @xmath82 and dimensionless multipolar moments @xmath83 and @xmath84 . we defined the orientational variable @xmath85 , where @xmath86 is the angle between the @xmath78-axis , assume to coincide with the normal to the bounding surfaces , the director of the uni - axial counterion is @xmath14 and the integral over this variable gives the orientational average . we have also introduced @xmath87 which is obviously the local orientationally dependent number density of the counterions and @xmath88 is the corresponding orientationally averaged charge density of the counterions . in the case of higher multipoles the number density and the charge density of the counterions are not proportional . expressions @xmath89 are simply the orientationally averaged multipolar charge densities , _ i.e. _ , @xmath90 for monopolar charge , @xmath91 for dipolar charge etc . , that are simply proportional to the three terms in the integrand of ( [ quadpb1 ] ) . the corresponding prefactor is simply obtained by appropriate normalisation of @xmath92 to satisfy electro - neutrality condition of the system . the above pb equation has to be supplemented by an appropriate boundary condition at @xmath24 corresponding to the electroneutrality of the system by taking into account surface charges , @xmath58 . the constant @xmath93 is set by these boundary conditions , as is the case in the standard pb theory . in dimensionless units these boundary conditions read @xmath94 . obviously in the case of counterions with only monopolar charge distribution the above set of equations reduces to the standard poisson - boltzmann theory . the characteristic function @xmath95 for the parallel plane geometry simply excludes the counterion configurations that would penetrate the bounding walls and depends on the geometric form of the counterions . since in what follows we will not be interested in steric effects , we will assume that all counterions are pointlike , and so @xmath96 . steric effects have been studied elsewhere @xcite . because of the similarity with the standard pb equation for monopolar charges one is led to believe that the pressure in an inhomogeneous system of multipolar countercharge can be derived in the same way as in the standard pb theory , _ via _ the s.c . contact value theorem @xcite . indeed this can be proven exactly , namely the mean - field pressure can be obtained from the first integral of the pb equation ( [ quadpb1 ] ) . after some manipulations its first integral is obtained in the form @xmath97 . \label{p1}\ ] ] here @xmath98 is the dimensionless equilibrium pressure in the system , defined as ( [ p_dimless ] ) . the r.h.s . of the above identity can be calculated at any arbitrary point @xmath99 as its value is independent of @xmath100 . the choice of the actual point is governed by the symmetry of the system , as is usual also in the standard pb theory . since we delve only on the symmetrical solution , with inversion symmetry centered on @xmath101 , the derivative of the mean potential at the mid - point must vanish , _ i.e. _ , @xmath102 . thus in this case @xmath103 while the first - vant hoff - term corresponds to repulsive interactions of the standard pb type , the second term in this particular geometry and symmetry entails attractive interactions between the bounding surfaces . note that @xmath104 . let us take a closer look at the above expression . for a polyelectrolyte chain , _ i.e. _ , an extended and flexible counterion , one can derive a similar type of pressure formula @xcite , with a negative term that contributes an attractive part to the force equilibrium . this attractive part in the case of polyelectrolytes is due to polyelectrolyte bridging interactions @xcite that stem from the connectivity and flexibility of the polyelectrolyte chain . does equation ( [ p3 ] ) have a similar physical content ? this interpretation certainly does not seem likely for point counterions , for which the above pressure formula was derived . the structure of the first integral of the pb equation ( [ p3 ] ) seems to be saying that apart from the ideal contribution to the equilibrium pressure , a term proportional to the midpoint number density , one also finds an electrostatic contribution that is due to the interaction between the monopolar and the quadrupolar part of the counterion charge density across the midplane and is proportional to the square of the midpoint density , see ( [ quadpb1 ] ) . the attractive part thus does not look like bridging which has its origin in the connectivity and flexibility of the polyelectrolyte chain , but more like a virial expansion in terms of the multipolar interactions . this is of course only true for point - like counterions . for extended counterions a bridging interpretation would be more appropriate as was already clear in the early studies of structured counterions @xcite . the above pb equation is a fourth order highly non - linear integro - differential equation and therefore difficult to handle with conventional numerical procedures . since here we are not particularly interested in the weak coupling results , we want only to show that the effect of quadrupolar moments , @xmath84 , on the interaction pressure is to induce a small attractive contribution . in order to show this we expand ( [ quadpb1 ] ) to the first order in @xmath84 and put @xmath105 , which then reads @xmath106 this equation can be solved much more easily with conventional numerical methods . although we must be aware that it is valid only for small @xmath84 . the corresponding formula for interaction pressure in the 1st order in @xmath84 expansion reads @xmath107 due to a convenient cancellation the formula contains no explicit @xmath84-dependence , but the pressure still depends on @xmath84 implicitly , due to the @xmath84 dependent potential @xmath108 . as can be shown numerically for small @xmath84 the correction to the standard pb pressure depends linearly on @xmath84 . here we are not interested in the details of the mean - field results they were analyzed in detail before @xcite but list them for completeness anyhow . in what follows we delimit ourselves to counterions that posses monopolar and quadrupolar charge , so we do not take into account any dipoles ( @xmath105 ) . ( a ) 0.25 cm ( b ) 0.25 cm ( c ) the density profile on fig . [ figpb](a ) corresponds to monopolar density @xmath109 , ( [ quadpb1 ] ) . counterions are localized preferentially in the vicinity of both surfaces due to mutual repulsion that prevents them from being localized at the center . what we then observe is that with increase of the quadrupolar moment , @xmath84 , counterions concentrate at surfaces even more . this can be easily explained . the potential energy of every quadrupolar particle in electrostatic potential is @xmath110 since the second derivative of the mean - field potential @xmath111 in the symmetric case considered here is a concave function of the coordinate @xmath78 , this means that the quadrupolar force , @xmath112 , acts away from the center toward both surfaces . the insight into the orientation of quadrupolar moments can be obtained through the average of the square of variable @xmath113 , namely @xmath114 , @xmath115 it is instructive to define an orientational order parameter @xmath116 this order parameter can have values in the range from @xmath117 to @xmath118 . in a totally disordered system where quadrupoles point randomly in all possible directions @xmath119 , and the orientational order parameter vanishes , @xmath120 . in the completely ordered case , @xmath121 and @xmath122 , so that all quadrupolar moments are parallel with the @xmath78-axis and perpendicular to the walls , @xmath123 . the other extremum , @xmath124 , corresponds to @xmath125 so that all quadrupolar moments are perpendicular to @xmath78-axis and parallel with the walls , @xmath126 . [ figpb](b ) shows the numerically obtained profile of the orientational order parameter . it increases with increasing quadrupolar strength @xmath84 and is larger at both surfaces than at the center . the order parameter @xmath127 indicates that quadrupolar moments are preferentially aligned parallel to the @xmath78-axis . according to ( [ wt ] ) the electrostatic energy can be minimized by increasing @xmath128 , since @xmath104 . so @xmath129 really is preferably increased above the value of @xmath130 . the quadrupolar contribution to the pressure for @xmath131 turns out to be to be attractive , fig . [ figpb ] , as was shown before @xcite . again this attractive contribution is obtained here on the mean - field level where the inter - counterionic correlations , essential in the strong coupling limit , play no role . it is indeed the _ intra - counterion correlations _ ( entering here _ via _ the rigid structure assumed for counterions ) that lead to such attractive contributions . at this juncture it does not make much sense to us in continuing the multipolar expansion to yet higher orders , that would become relevant for even more aspherical and elongated charge distributions that would obviously entail also some molecular flexibility . in that case one could use the well developed theory of polyelectrolyte mediated interactions with much better confidence @xcite . in the strong coupling limit , @xmath132 , the system is highly charged and counterions are highly correlated . the partition function ( [ funint ] ) can be approximated by the two lowest order terms in the virial expansion with respect to the ( renormalized ) fugacity @xcite @xmath133 the second term corresponds to a one - particle partition function as the sc limit is effectively a single particle theory . @xmath134 and @xmath135 are then given by @xmath136 , \label{funint - sc-0}\ ] ] which is the exponential of the interaction between bare surface charges , and @xmath137 , \label{funint - sc-1}\end{aligned}\ ] ] which is the partition function of all possible ( single ) counterion configurations . the interaction potential in this part of the partition function is composed of the direct and image electrostatic interactions ( [ decomp - u ] ) . in the case of two charged surfaces with uniformly smeared surface charge density ( [ sigmas ] ) , the surface charge electrostatic potential does not depend on the @xmath78-coordinate ; it is spatially homogeneous and is given by @xmath138 the corresponding potential energy of a counterion in this surface electrostatic potential is given by @xmath139 since all the terms in density operator ( [ chaden ] ) , except the first one , depend on the gradients , _ i.e. _ , spatial derivatives , the counterion energy in a homogeneous external electrostatic potential depends only on the first , monopolar term ( [ funint - u-0 ] ) . that means that higher multipoles do not interact directly with planar surface charge , but we must emphasize that this is not the case in inhomogeneous potential at curved surfaces , _ e.g. _ , at cylindrical or spherical surfaces or inhomogeneously charged surfaces . as for the self - energy contribution , the second term in the exponent of ( [ funint - sc-1 ] ) , it only picks up terms from the @xmath78-dependent parts of the image self - interaction , @xmath140 . the direct self - interaction , @xmath141 , does not depend on coordinates so it can be discarded . the self - image energy is then @xmath142 self - image contributions are among monopolar , dipolar and quadrupolar moments of the counterions interacting with their own electrostatic images of monopolar , dipolar and quadrupolar moments . summing up all these contributions we remain with nine terms of which only six are different . writing them up _ in extenso _ we obtain @xmath143 here , the first expression , @xmath144 , corresponds to the interaction between a monopole and its own monopole image , therefore it is proportional to @xmath145 . the second term , @xmath146 , corresponds to the monopole - dipole image interaction which is obviously the same as the dipole - monopole image interaction , @xmath147 , and so on for all the higher order terms . these image self - interaction terms can be expressed in a dimensionless form as @xmath148 the above expressions were derived for the planar geometry . similar expressions but with different image potentials can be derived also in other geometries with dielectric inhomogeneities . note that if there is no dielectric discontinuity , so that @xmath149 , they are all identically zero , @xmath150=0 ! thus one can conclude at this point that in the dielectrically homogeneous case the higher order multipoles are completely irrelevant on the sc level . only the monopolar term survives in the partition function . this is probably one of the most important conclusions of this work , so let us reiterate it : in a dielectrically homogeneous case in plan - parallel geometry the structure of the counterions as codified by their multipolar moments plays absolutely no role in the strong coupling limit ! for our case of two charged planar surfaces we evaluate the above expressions by using the green s function ( [ green ] ) , which was derived for planar geometry with a step - function dielectric profile at the two boundaries @xmath151 . we get the following somewhat cumbersome expressions for the dimensionless self - image interactions defined above @xmath152 ^ 2},\nonumber\\ \tilde w_3(\tilde z , x)&=&-\frac{1}{8}\xi t(1 - 3x^2)\left\ { \sum_{n\textrm { even}}\frac{\delta^n}{(n\tilde a)^3}+\sum_{n\textrm { odd}}n\tilde a\frac{(n\tilde a)^2 + 3\tilde z^2}{[(n\tilde a)^2-\tilde z^2]^3}\,\delta^n\right\},\nonumber\\ \tilde w_5(\tilde z , x)&=&\frac{1}{8}\xi p^2\left\{(1 - 3x^2)\sum_{n\textrm { even}}\frac{\delta^n}{(n\tilde a)^3}+(1+x^2)\sum_{n\textrm { odd}}n\tilde a \frac{(n\tilde a)^2 + 3\tilde z^2}{[(n\tilde a)^2-\tilde z^2]^3}\,\delta^n\right\},\nonumber\\ \tilde w_6(\tilde z , x)&=&\frac{3}{4}\xi p t(1+x^2)x\,\tilde z\sum_{n\textrm { odd } } n\tilde a\frac{(n\tilde a)^2+\tilde z^2}{[(n\tilde a)^2-\tilde z^2]^4}\delta^n,\nonumber\\ \tilde w_9(\tilde z , x)&=&\frac{3}{32}\xi t^2\left\{(3 - 30x^2 + 35 x^4)\sum_{n\textrm { even}}\frac{\delta^n}{(n\tilde a)^5}+(3 + 2x^2 + 3x^4)\sum_{n\textrm { odd}}n\tilde a \frac{(n\tilde a)^4 + 10(n\tilde a\tilde z)^2 + 5\tilde z^4}{[(n\tilde a)^2-\tilde z^2]^5}\,\delta^n\right\}.\nonumber\end{aligned}\ ] ] all indices @xmath31 here run only through positive values . note again that self - image contributions , @xmath153 , vanish when @xmath154 . taking all this into account , the strong coupling interaction free energy , @xmath155 , _ i.e. _ , the part of the free energy that depends on the intersurface separation , can be written in dimensionless form as @xmath156 just as in the wc case we have again assumed that the counterions are point - like particles so that the characteristic function @xmath157 does not depend on their coordinate @xmath158 and therefore the integration goes from @xmath159 to @xmath160 . this means that we also disregard the possible entropic effects due to the finite size and anisotropy in the shape of counterions . these entropic contributions to the partition function are relevant only for intersurface separation on the order of the size of the counterion or smaller . in that regime of separations other , much stronger effects would come into play and compete with ionic finite size effects , thus these type of effects are not the focus of this paper . the corresponding dimensionless pressure is simply obtained by taking the derivative of the free energy ( [ freeen ] ) with respect to wall separation , @xmath161 , @xmath162 it is again obvious from here that in the dielectrically homogeneous case , _ i.e. _ , the case with no electrostatic images , where @xmath163 , the strong coupling limit is given _ exactly _ by the monopolar term ( the first term in ( [ funint - sc-1 ] ) ) . thus without the images we remain with the same form of the strong coupling interaction free energy as in the case of monopolar point charges . it is given by @xmath164 with the corresponding pressure @xmath165 the higher order multipoles thus make _ no _ direct contribution to the strong coupling interaction free energy or forces between the bounding charged surfaces without dielectric inhomogeneities . this is a very powerful result whose ramifications in fact impose rather stringent limits on the significance of the multipolar expansion of counterionic charge . it appears that for highly charged counterions , that are in fact the only ones where the multipolar expansion really makes sense , most of the electrostatics is properly captured by the monopolar term . higher order multipoles simply do not contribute to the pressure in the system in the sc limit . of course all of this is valid in the limit of homogeneous dielectric properties for planar surfaces without any image effects . the counterion density profile can be extracted from ( [ freeen ] ) as an integrand in the second term , @xmath166 in the case without the dielectric mismatch the density profile is simply homogeneous , @xmath167 . dielectric images induce an additional repulsion between the counterions and the surface charges pushing them towards the midplane region between the two charged bounding surfaces . orientational order of quadrupoles can be inspected _ via _ averaged @xmath114 which is here defined as @xmath168 we can now use the same definition of order parameter @xmath169 as in wc , ( [ eqs ] ) . we now present some numerical results for the sc limit . without any dielectric mismatches the density profile of the monopolar counterions is homogeneous , _ i.e. _ , it does not depend on @xmath78 , which is a well known result @xcite . in the case of multipoles with dielectric images this is no longer true , but the counterion density does remain an even function of @xmath78 . ( a ) 0.25 cm ( b ) 0.25 cm ( c ) on fig . [ figsc ] we show the numerical results for coupling parameter strength @xmath170 , dielectric jump @xmath171 and @xmath172 . contrary to the wc situation , here counterions are localized in the vicinity of the midpoint between the two bounding surfaces due to image repulsion . even without quadrupolar contributions ( @xmath173 ) the monopolar interaction , @xmath144 , is repulsive and long ranged , @xmath174 @xcite , compressing the counterions to the midpoint . when the quadrupolar parameter @xmath84 , is increased a bit the midpoint peak first starts to widen . this is due to the attractive monopole - quadrupole image contribution , represented by the @xmath175 and @xmath176 terms which are proportional to @xmath84 . as @xmath84 increases further ( approximately for @xmath177 for large @xmath178 ) the repulsive quadrupole - quadrupole image contribution , @xmath179 , which is proportional with @xmath180 , takes over and confines counterions to the center even more so that the density peak sharpens up . for the value @xmath178 large enough , so that density profile becomes bell - shaped , the sc density profile can be approximated by a gaussian curve with a variance , _ i.e. _ , peak width , of the form @xmath181 as can be seen , small @xmath84 widens the density peak , but further increasing @xmath84 over the value @xmath177 the peak narrows . also by increasing parameters @xmath60 , @xmath182 the midpoint density peak narrows . this also means that the repulsive pressure between the walls increases . as the density peak narrows and the density of the counterions at the midpoint plane increases , the counterions become more ordered in the quasi two - dimensional midpoint layer , implying also that the correlations are stronger and therefore the sc limit should _ a fortiori _ be even more appropriate . fig . [ figsc](b ) shows also the orientational order parameter for the sc limit . contrary to the wc results , here the orientational order parameter is negative , implying that quadrupoles are preferably aligned perpendicular to @xmath78-axis . the parameter reaches its extremum @xmath117 at both walls where quadrupoles are perfectly perpendicular to @xmath78-axis and parallel to the walls . this is due to the strong quadrupole - quadrupole image repulsion , @xmath179 , which has its minimum at @xmath183 and is not a purely steric effect as in simulations of long charged stiff rods @xcite . since monopole - monopole image interaction , @xmath144 , has the longest range , we can expect that for large separations between the walls , @xmath184 , this interaction will dominate and higher multipoles can be again neglected . [ figsc](c ) shows the pressure difference @xmath185 , which is the difference between the pressure with finite @xmath84 and the pressure without the quadrupolar contributions , @xmath173 . as expected this difference indeed goes to zero at large separations . at moderate separations the quadrupolar contribution becomes negative , which is caused by monopole - quadrupole image attractions , @xmath175 and @xmath176 , that are shorter ranged than @xmath144 . at even smaller separations quadrupole - quadrupole image repulsion , @xmath179 , becomes dominant , since it has the shortest range and varies approximately as @xmath186 . the larger the @xmath84 the smaller is thus the maximal quadrupolar attraction at larger distance since it is overwhelmed by repulsive contributions . there appear some similarities between the polyelectrolyte bridging interaction @xcite and the attraction seen with strongly charged counterions . recent simulations and density functional results @xcite in fact do accentuate a close connection between polyelectrolyte bridging and strong - coupling electrostatics , a line of reasoning that we will explore here in finer detail . turesson _ et al . _ @xcite indeed find out that for flexible polyelectrolytes bridging attraction appears to be the dominant source of attractive interaction . on the other hand , for stiff charged rods the strong adsorption to the bounding charged walls prevents bridges to form . instead a very strong correlation attraction is apparent at shorter separations . flexibility thus extends the range of attractive interactions as their nature changes from short range correlation attraction to longer ranged polyelectrolyte bridging . similar conclusions have been also reached in @xcite . here we will formulate the exact correspondence between the intuitive notion of `` bridging '' in the case of multivalent counterions and the definition of `` bridging interaction '' in the case of charged polymers . let us analyse first the partition function for two disparate systems : a counterion coulomb fluid confined between two strongly charged walls , and a flexible polyelectrolyte chain confined in the same geometry with weakly charged walls . for point counterions between two charged surfaces the grand canonical partition function in the sc limit @xcite can be written to the first order in fugacity as @xmath187 where @xmath188~ \exp\biggl(&-&{\ul12 } \beta\varepsilon_0 \int \varepsilon({\bf r})({\mbox{\boldmath$\nabla$}}\phi({\bf r}))^2 \rmd{\bf r } + \nonumber\\ & + & \rmi \beta \int \rho_0({\bf r } ) \phi({\bf r } ) \rmd{\bf r } \biggr),\end{aligned}\ ] ] and @xmath189 \exp \biggl(\!\!-{\ul12 } \beta \varepsilon_0 \!\!\int \!\!\varepsilon({\bf r})({\mbox{\boldmath$\nabla$}}\phi({\bf r}))^2 \rmd{\bf r } + \nonumber\\ & + & \rmi \beta \!\int \!\rho_0({\bf r } ) \phi({\bf r } ) \rmd{\bf r } + \rmi \beta qe_0 \ ! \int\ ! \rho({{\mathbf r } } - { { \mathbf r } } ) \phi({\bf r } ) \rmd{\bf r}\biggr ) . \label{comp-1}\end{aligned}\ ] ] the above partition function is of course exactly the same as the one in ( [ funint - sc ] ) after one evaluates the functional integrals indicated above explicitly . here @xmath190 is again the external fixed charge on the two bounding surfaces located at @xmath191 and the shorthand @xmath192 is introduced for @xmath193 . on the other hand the partition function for a flexible polyelectrolyte chain is given by @xcite @xmath194 ~\exp\biggl(\!\!-{\ul12 } \beta \varepsilon_0\!\ ! \int \!\ ! \varepsilon({\bf r})({\mbox{\boldmath$\nabla$}}\phi({\bf r}))^2 \rmd{\bf r } + \\ & + & \rmi \beta\ ! \rho_0({\bf r } ) \phi({\bf r } ) \rmd{\bf r } + { \rm ln}{\int\!\!\!\!\int { \cal g}_{\phi}({{\mathbf r } } , { { \mathbf r } } ' ; n ) \,\rmd{\bf r}\rmd{\bf r'}}\biggr).\nonumber\end{aligned}\ ] ] where @xmath195 is the green function of a polyelectrolyte chain in external electrostatic field @xmath196 given by @xcite @xmath197\times\\ & & \hspace{-12ex}\times\exp\biggl(\!- { \textstyle{\frac{3}{2 \ell^2 } } } \!\int_0^n \!\ ! \left ( \frac{\rmd{{\mathbf r}}(n)}{\rmd n}\right)^2 \rmd n + \rmi\beta qe_0\ ! \int_0^n \!\!\phi({{\mathbf r}}(n ) ) \rmd n\biggr),\nonumber\end{aligned}\ ] ] where @xmath198 is the coordinate of a polyelectrolyte segment @xmath31 and the boundary condition is set as @xmath199 . representing the green function _ via _ a corresponding edwards equation one can derive that the polyelectrolyte monomer density @xmath200 is given by a functional derivative of @xmath201 with respect to the fluctuating electrostatic potential @xmath45 @xcite . the polyelectrolyte monomer density has to satisfy the condition @xmath202 in the planar geometry with two bounding surfaces , one can center the monomer density at the origin of the coordinate system , whose @xmath78 axis is directed along the surface normal and has the origin at the midplane . denoting @xmath203 one can thus write @xmath204 . in the limit of small electrostatic potentials , the lowest order ( _ i.e. _ , the first - order ) solution of the polyelectrolyte green function can be written as @xcite @xmath205 by construction this solution clearly corresponds to a weak coupling of the polyelectrolyte to the external field generated by the fixed charges at the boundary of the system . for large values of this charge higher orders in the expansion ( [ green-2 ] ) would certainly have to be taken into account . in the case of weak coupling the partition function of a polyelectrolyte chain is then given approximately by @xmath206 = \label{comp-2}\\ & = & { \mathcal c } \int { \cal d}[\phi({{\mathbf r } } ) ] ~\exp\biggl(-{\ul12 } \beta\varepsilon_0 \int \varepsilon({\bf r})({\mbox{\boldmath$\nabla$}}\phi({\bf r}))^2 \rmd{\bf r } + \nonumber\\ & + & \rmi \beta \!\int \!\rho_{0}({\bf r } ) \phi({\bf r } ) \rmd{\bf r } + \rmi \beta qe_0 \!\int \!\rho_{\phi=0 } ( { { \mathbf r } } - { { \mathbf r}}_0 ; n ) \phi({{\mathbf r } } ) \rmd{{\mathbf r } } \biggr).\nonumber\end{aligned}\ ] ] if we now compare the expressions ( [ comp-1 ] ) and ( [ comp-2 ] ) we can establish the following identity @xmath207,\ ] ] with the simple ( first order sc ) counterion partition function on the l.h.s . and the polyelectrolyte ( zeroth order wc ) partition function on the r.h.s . this `` duality '' relation connects the partition function of a strongly coupled point counterion system with the partition function of a weakly coupled flexible polyelectrolyte system . since @xmath208 for finite @xmath3 describes polyelectrolyte bridging interactions it is clear that the above formula establishes a formal connection between strong coupling point counterion attractive interactions and weak coupling polyelectrolyte bridging interactions . using field - theoretic methods we derived a description of the counterion - mediated electrostatic interaction when the counterions posses internal degrees of freedom such as a rotational axis . we concentrate on the symmetrically charged planar surfaces with a dielectric mismatch on both sides and counterions with quadrupolar moments in between . we analyse this system in two different regimes , namely in the weak coupling ( mean - field ) and the strong coupling limit . the former one describes the case of low surface charge and low counterion valency , whereas the latter one describes high surface charge and high counterion valency . in the mean - field limit , we derived the poisson - boltzmann equation for counterion density in the case that the counterions posses dipolar as well as quadrupolar moments . since the equation is highly non - linear , we solved it numerically only in the first order approximation , which should be valid for small quadrupolar moments , @xmath84 . as is already known the dielectric mismatches play no role in the mean - field theory and have therefore no effect on quadrupoles @xcite . one could add those corrections in by hand @xcite , but such an approach does not strictly corresponds to the mean - field analysis . quadrupolar interaction affects the counterion density distribution as well as interactions between the charged surfaces . counterions are depleted from the central region and concentrated in the vicinity of both surfaces . the orientation of quadrupoles is preferentially parallel to the @xmath78-axis as the moment @xmath84 is increased . the highest alignment is reached right next to both bounding surfaces . the quadrupolar contribution to the interaction pressure is attractive . in the strong coupling limit higher order multipoles also play an important role in the intersurface interactions but only for a dielectrically inhomogeneous case . without any dielectric discontinuities in the system the higher multipoles simply do not matter in plan - parallel case and the monopolar part of the partition function captures all the strong coupling effects . this is a completely general result for plan - parallel surfaces and should be of some importance in assessing the electrostatically mediated forces between strongly charged planar surfaces . as already mentioned the multipoles can have effects near curved surfaces even without dielectric discontinuities . if compared to the case of monopolar counterions , higher multipoles interact only _ via _ the orientationally dependent part of the image self - interaction . in our analysis we focused on the counterions with monopolar and quadrupolar moments , so that the interaction is composed of three contributions . these are the monopole - monopole image interaction , which is long ranged and repulsive , the monopole - quadrupole image as well as quadrupole - monopole image interactions that are shorter range and attractive . finally , very short ranged quadrupole - quadrupole image contribution is repulsive and plays a dominant role at very small distances . we are at present involved in extensive monte carlo simulations in order to explore the validity of the formal developments described above . preliminary results completely vindicate the theoretical analysis . r.p . would like to acknowledge the financial support by the slovenian research agency under contract nr . p1 - 0055 ( biophysics of polymers , membranes , gels , colloids and cells ) . m.k . would like to acknowledge the financial support by the slovenian research agency under the young researcher grant . a.n . would like to acknowledge financial support by the institute for research in fundamental sciences ( ipm ) , tehran . this research was supported in part by the national science foundation under grant no . phy05 - 51164 ( while at the kitp program _ the theory and practice of fluctuation induced interactions _ , ucsb , 2008 ) . is grateful to m.w . kim for useful discussions . y.s.j . and p.a.p . were supported in part by the national science foundation ( grants dmr-0503347 , dmr-0710521 ) and mrsec nsf dmr-0520415 . y.s.j . and p.a.p . were supported by a korea science and engineering foundation ( kosef ) grant funded by the korean government ( mest ) ( grant code : r33 - 2008 - 000 - 10163 - 0 ) and the brain korea 21 projects by the korean government . van blaaderen a 2006 _ nature _ * 439 * 545 hong l , cacciuto a , luijten e and granick s 2006 _ nano lett . _ * 6 * 2510 glotzer s c and solomon m j 2007 _ nature materials _ * 6 * 557 jerri h a , dutter r a and velegol d 2008 _ soft matter _ * 5 * 827 abrashkin a , andelman d and orland h , 2007 _ phys . lett . _ * 99 * 077801 maset s and bohinc k 2007 _ j. phys . _ * 40 * 11815 bohinc k , igli a and may s 2004 _ eur . phys . _ * 68 * 494 kim y w , yi j and pincus p a 2008 _ phys lett . _ * 101 * 208305 deraad l l , milton k a , tsai w y , schwinger j s , _ classical electrodynamics _ , the advanced book program , perseus books group , ( 1998 ) netz r r 2001 _ eur . j. e _ * 5 * 557 ; moreira a g and netz r r 2002 _ eur . phys . j. e _ * 8 * 33 podgornik r 1989 _ j. chem . phys . _ * 91 * 5840 dean d s , horgan r r , naji a and podgornik r 2009 _ j. chem . phys . _ * 130 * 094504 naji a , jungblut s , moreira a g and netz r r 2005 _ physica a _ * 352 * 131 attard p , mitchell d j and ninham b w 1988 _ j. chem . _ * 88 * 4987 boroudjerdi h , kim y w , naji a , netz r r , schlagberger x and serr a 2005 _ phys . rep . _ * 416 * 129 oosawa f 1968 _ biopolymers _ * 6 * 1633 ohnishi t , imai n and oosawa f 1960 _ j. phys . . japan _ * 15 * 896 jho y s , park g , chang c s , pincus p a and kim m w 2006 _ phys . * 73 * 021502 jho y s , park g , chang c s , pincus p a and kim m w , 2007 _ phys . rev . e _ * 76 * , 011920 podgornik r and ek b 1988 _ j. chem . soc . , faraday trans 2 _ * 5 * 611 ; podgornik r 1990 _ j. phys . * 23 * 275 rouzina i and bloomfield v a 1996 _ j. phys . chem . _ * 100 * 9977 grosberg a y , nguyen t t and shklovskii b i 2002 _ rev . phys . _ * 74 * 329 nguyen t t , grosberg a y and shklovskii b i 2000 _ j. chem . phys . * 1110 nguyen t t and shklovskii b i 2001 _ phys . rev . e _ * 64 * 041407 levin y 2002 _ rep . phys . _ * 65 * 1577 netz r r and joanny j f 1999 _ macromolecules _ * 32 * 9013 netz r r and joanny j f 1999 _ macromolecules _ * 32 * 9026 moreira a g and netz r r 2000 _ eur . _ * 52 * 705 kandu m and podgornik r 2007 _ eur . j e _ * 23 * 265 ; jho y s , kandu m , naji a , podgornik r , kim m w and pincus p a 2008 _ phys . lett . _ * 101 * 188101 poon w c k and andelman d 2006 _ soft condensed matter physics in molecular and cell biology _ , ( taylor & francis ) . podgornik r 1992 _ j. phys . chem . _ * 96 * 884 podgornik r and lier m 2006 _ curr . op . * 11 * 273 podgornik r 2004 _ j. pol . part b : polymer physics _ * 42 * 3539 miklavic s j and woodward c e 1990 _ j. chem . phys . _ * 93 * 1369 podgornik r 2004 _ j. polymer sci . part b : polymer physics _ * 42 * issue 19 3539 turesson m , forsman j and kesson t 2006 _ langmuir _ * 22 * 5734 * 352 * 131 ( 2005 ) . derouchey j , netz r r and r " adler j o 2005 _ eur . phys . j. _ netz r r 2001 _ eur . j. e _ * 5 * 557 podgornik r 1991 _ j. phys . chem . _ * 95 * 5249 doi m and edwards s f 1986 _ the theory of polymer dynamics _ , oxford university press , usa onuki a 2006 _ phys . e _ * 73 * 021506

we present general arguments for the importance , or lack thereof , of the structure in the charge distribution of counterions for counterion - mediated interactions between bounding symmetrically charged surfaces . we show that on the mean field or weak coupling level , the charge quadrupole contributes the lowest order modification to the contact value theorem and thus to the intersurface electrostatic interactions . the image effects are non - existent on the mean - field level even with multipoles . on the strong coupling level the quadrupoles and higher order multipoles contribute additional terms to the interaction free energy only in the presence of dielectric inhomogeneities . without them , the monopole is the only multipole that contributes to the strong coupling electrostatics . we explore the consequences of these statements in all their generality .

introduction model dimensionless representation strong and weak coupling dichotomy mean-field limit strong coupling limit multipoles, correlations, bridging and all that jazz conclusions acknowledgement

with the recent progress in high performance computational techniques , it is now possible to carry out accurate calculations on heavy atomic systems using powerful relativistic many - body theories @xcite . it is not possible to treat all the heavy atomic systems by a single many - body theory due to the differences in their electronic structures . the relativistic coupled - cluster ( rcc ) theory has been successfully applied earlier to a wide range of atomic properties @xcite including the lifetime calculations of pb@xmath1 @xcite and alkaline - earth ions @xcite . the role of electron correlation effects is dramatic in some of those cases . in the present work , we have carried out the lifetime calculations of the lowest @xmath2 and @xmath3 states in sc iii and y iii ions . and @xmath4 states in sc@xmath0 and y@xmath0 . the @xmath5 and @xmath6 represent the principal quantum numbers of the ground and metastable states where @xmath7 and @xmath8 for sc@xmath0 and @xmath9 and @xmath10 for y@xmath0 .,width=302 ] generally , the forbidden transition lines find numerous interesting applications in astrophysics which includes the determination of the elemental abundances in different celestial objects and also in the laboratory astrophysics and plasma physics studies . the spectral lines of many alkaline - earth ions including y and its doubly charged ions have been observed in the spectra of many hot stars like ap stars @xcite and these ions seem to be over abundant in hg - mn stars compared to their normal stellar abundances . this anomaly observed over abundances than that expected from the theoretical estimations based on the s - process of neutron capture is studied extensively in the literature @xcite . thus , the study of adjacent elements of the same row of the periodic table like , the sr - y - zr triad of elements , can be used to understand the nucleo - synthetic origin of these elements and it can reveal the post - main - sequence evolution of the chemically peculiar stars , such as hg - mn stars @xcite . the y iii and sc iii ions can also be used for the diagnostics and modeling of the stellar plasma @xcite . recently , brage et . @xcite have used the multi - configuration hartree - fock ( mchf ) method in calculating the lifetimes of the 5p , 5d and 6s states of y iii which were measured earlier by maniak et . al . @xcite . however the lifetimes of the @xmath11 and @xmath12 states in y iii seem to be unreported so far . though there are a large number of results available for the transition wavelengths of sc iii , it seems that there are no measurements on either the lifetimes or the transition probabilities of the @xmath13 and @xmath14 states . for this reason , we have carried out the calculations of the lifetimes of the lowest @xmath2 and @xmath3 metastable states in sc iii and y iii employing a powerful many - body theory known as the rcc theory . as the experimental wavelengths are known to high accuracy we have used them in our calculations in order to minimize the errors in our lifetime estimations . the low - lying excited states and their electromagnetic decay channels are shown in the figure [ fig1 ] . even though sc iii and y iii belong to k- and rb- iso - electronic sequence , respectively , the ground states of the former two are @xmath15 states ( @xmath7 for sc iii and @xmath9 for y iii ) unlike the latter two for which the ground states are @xmath16 . again though the considered systems are the doubly charged ions , they have the same ground states as that of their neutral counterparts . thus the ions considered in this work are interesting from the theoretical and as well as astrophysical point of view . we analyze the role of correlation effects in these calculations and compare their trends with the previously studied lifetimes of the @xmath2 states @xcite in alkali atoms and alkaline earth metal ions @xcite . the transition probabilities for the spontaneous emissions due to e2 and m1 electromagnetic transitions from the state @xmath17 to the state @xmath18 are given by , @xmath19 \lambda^5 } s^{\text{e2 } } \label{eqn1 } \\ a^{\text{m1}}_{i \rightarrow f } & = & \frac { 2.69735 \times 10^{13 } } { [ j_i ] \lambda^3 } s^{\text{m1}},\label{eqn2 } \end{aligned}\ ] ] where , @xmath20=2j_i+1 $ ] is the degeneracy of the state @xmath17 , @xmath21 is the transition line strength in atomic unit ( au ) for the corresponding transition operator @xmath22 and wavelength @xmath23 ( ) . as shown in the figure [ fig1 ] , the @xmath2 state can decay to the @xmath24 state in the considered systems via e2 and m1 channels ; hence the total transition probability for the @xmath2 state can be expressed as , @xmath25 similarly , the @xmath3 state can decay to the @xmath24 state in these systems via e2 and m1 channels and decay to the @xmath2 state via e2 channel ; the resulting total transition probability is @xmath26 in the above expressions , @xmath5 and @xmath6 represent the principal quantum numbers of the @xmath27 and @xmath28 states ; where , @xmath7 and @xmath8 for sc iii and @xmath9 and @xmath10 for y iii . the single particle reduced matrix elements of e2 and m1 operators used in our calculations can be found from @xcite . the lifetimes of these states are determined by taking the inverse of their total probabilities . we use the following rcc ansatz to construct the wave functions for one valence states ( @xmath29 ) in the considered systems ; @xmath30 with the reference state @xmath31 for the valence electron @xmath32 and @xmath33 is the dirac - fock ( df ) state of the closed - shell system constructed by the dirac - coulomb ( dc ) hamiltonian . this is the relativistic method of the universal valence coupled - cluster theory proposed by mukherjee et al @xcite and later was compactly formulated by lindgren @xcite . in the above rcc wave function expression , we define @xmath34 and @xmath35 as the closed - shell and open - shell excitation operators which excite core electrons and valence with core electrons , respectively . we have considered all single , double and a subset of important triple excitations approximation in our calculations which is known as the ccsd(t ) method in the literature @xcite . it has been found in many studies in the similar systems as considered here that the ccsd(t ) method is powerful enough to account the correlation effects to evaluate various properties to high accuracy @xcite . this is primarily because of the fact that this approach considers the most important correlation effects like core - polarization and pair - correlation effects to all - orders @xcite . the transition matrix elements for the operator @xmath22 are calculated in this approach using the following expression ; @xmath36 where , @xmath37 . since the above expression contains many non - linear terms those will be computationally expensive if they are calculated directly , we follow the extended wick s theorem @xcite to break these terms in simple form as discussed below . although these procedures are given in detail earlier @xcite , we mention them briefly here . first , we divide @xmath38 as the effective one - body , two - body etc . operators and later these terms are sandwiched between the necessary @xmath35 operators . to save the computational time , we calculate and save only the one - body terms which are later used where other terms are computed directly in the final property calculations . it has to be noticed that the fully contracted @xmath38 terms do not contribute in the present calculations . contributions from the normalization of the wave functions ( norm . ) are accounted by , @xmath39 where , @xmath40 for the corresponding valence electrons @xmath41 . in table [ tab1 ] , we present the wavelengths and transition line strengths of the low - lying transitions from the lowest @xmath2 and @xmath3 states to the ground states of sc iii and y iii . the transition wavelengths for sc iii are taken from the critically evaluated atomic data base of national institute for the standard and technology ( nist ) @xcite and the transition wavelengths for y iii are taken from @xcite for the given transitions . the calculated transition strengths are used to determine the corresponding transition probabilities which are given in the last column of the above mentioned table . lcccc transition & @xmath23 & transition & strength & @xmath42 + ( @xmath43 ) & ( in ) & channel & ( au ) & @xmath44 + + & & & & + + @xmath45 & 505970 & m1 & 2.3751 & 8.2434e-5 + & & e2 & 2.7164 & 1.5291e-11 + & & & + @xmath46 & 3916 & m1 & 7.9495e-7 & 1.786e-4 + & & e2 & 12.9157 & 7.8584 + @xmath47 & 3946 & e2 & 19.5007 & 11.413 + & & & + + @xmath48 & 138093 & m1 & 2.3822 & 4.0667e-3 + & & e2 & 9.9573&3.7011e-8 + & & & + @xmath49 & 13392 & m1 & 1.1136e-7 & 6.2544e-7 + & & e2 & 37.6554 & 4.895e-2 + @xmath50 & 14830 & e2 & 57.7644 & 4.5090e-2 + [ tab1 ] the calculated total transition probabilities for the lowest @xmath2 and @xmath3 states are given in table [ tab2 ] . the lifetimes are determined by taking the reciprocal of these quantities for the corresponding states . the m1 transition probabilities between the fine structure levels , i.e. the @xmath51 transitions are 5 - 6 orders of magnitude larger than the e2 transitions , where as , the m1 transition probabilities for the @xmath52 transitions are 5 - 6 orders smaller than the corresponding e2 transitions ( cf . table [ tab1 ] ) for both the ions . thus , the m1 transitions play crucial roles in determining the lifetimes of the @xmath2 states but it is the other way around for the @xmath3 states as the e2 transitions play crucial roles in their lifetime determinations . as it can be seen from table [ tab2 ] that the lifetime of the @xmath13 state in sc iii is very large implying that it is a highly forbidden state . however , the lifetime of the @xmath14 state is a few fraction of a second . on the other hand , the lifetime of the @xmath11 state in y iii is @xmath53 seconds which again is a long lived metastable state , where as , the @xmath12 state has a lifetime of only @xmath54 seconds . this also indicates that the correlation effects in the calculations of m1 transition amplitudes of the @xmath55 transitions are very small and hence these results can be ascribed as relativistic in origin . the transition probabilities of the states given in table [ tab2 ] will be very useful in astrophysics for the accurate determination of the abundances of these elements . we estimate that our lifetime results can have an error bar of less than 2% . the source of errors includes the truncation of the coupled - cluster series at the double excitations and partial inclusion of triple excitations , the omission of breit and qed effects in our calculations . lcc state & transition probability & lifetime + & ( sec@xmath56 ) & ( sec ) + + & & + + @xmath57 & 8.2434e-5 & 12130.86 + @xmath58 & 19.2715 & 0.0519 + & & + + @xmath59 & 4.0668e-3 & 245.89 + @xmath60 & 9.4041e-2 & 10.63 + [ tab2 ] .reduced matrix elements of the m1 and e2 transition operators from @xmath2 states to the ground states in au in the considered systems . [ cols="<,^,^,^,^ " , ] [ tab5 ] to understand the importance of the correlation effects , we present the contributions from the individual terms of eq . ( [ eqn14 ] ) in table [ tab3 ] to table [ tab5 ] . the contribution given by the operator @xmath22 is the dirac - fock term and the difference between @xmath61 and @xmath22 represents the pure core - valence ( cv ) correlation effects . the remaining large contributing terms @xmath62 and @xmath63 with their conjugate terms represent the all - order pair - correlation and core - polarization effects , respectively , for the given valence electron . the reduced matrix elements of the m1 and e2 transition operators for the @xmath51 transitions in sc@xmath0 and y@xmath0 are given in table [ tab3 ] . the transition amplitudes due to e2 are slightly larger than the amplitudes due to m1 transitions , but the latter totally decide the lifetimes of the @xmath2 states ( see table [ tab1 ] ) . it is because , the transition probabilities due to e2 transitions depend on fifth power of their wavelengths in the denominator , where as , the m1 transition probabilities depend on third power of their wavelengths in the denominator . the total contribution of the correlation effects , which can be evaluated as the difference between the df and the ccsd(t ) results , in the calculation of m1 transition amplitudes are around 0.5% for both the systems , however , they are more than 10% in the case of the e2 transition amplitudes . the reduced matrix elements of the m1 and e2 transition operators of the transition , @xmath64 in sc@xmath0 and y@xmath0 are given in table [ tab4 ] . the transition amplitudes due to e2 are almost 3 - 4 orders of magnitude larger than the m1 amplitudes and the former give largest contributions to the lifetimes of the @xmath3 states . the correlation effects are very large in both the m1 and e2 transitions in sc@xmath0 , where as , they are large only in the case of m1 transition in y@xmath0 . in table [ tab5 ] , we present the e2 transition amplitudes of the @xmath65 transitions . the trend of the contributions from different rcc terms are similar to the @xmath66 transitions of the corresponding iso - electronic alkali metal atoms @xcite and alkaline earth metal ions @xcite , but these amplitudes are smaller in the considered systems . the correlation effects in these calculations are just around 12% and 9% in sc iii and y iii , respectively . again , the correlation effects in m1 transition amplitudes of the @xmath55 transitions are very small compared to the @xmath67 transition amplitude in pb@xmath1 @xcite , where we observe similar behavior for the e2 transition amplitudes . the determination of the lifetimes of the @xmath3 states depend crucially on the accurate calculation of the e2 transition amplitudes of both @xmath68 and @xmath65 transitions , where as , the m1 amplitudes of the former transitions play negligible roles . we have determined the lifetimes of the lowest @xmath2 and @xmath3 metastable states in sc@xmath0 and y@xmath0 which seem to be unavailable in the literature . the transition strengths and the transition probabilities of these systems find interesting applications in astrophysics in understanding the post - main sequence evolution of the cool giant stars . these doubly ionized ions may also be suitable for other ion trapping experiments as they are reported to have long lifetimes , which need to be explored by the experimentalists . we are thankful to the authority of c - dac teraflop supercomputer facilities , bangalore , india for providing their computational resources . one of the authors ( bks ) thanks professor peter fulde for many useful discussions and for the hospitality at mpipks . 26 g. c. rodrigues , m. a. ourdane , j. bieron , p. indelicato and e. lindroth , phys . a * 63 * , 012510 ( 2000 ) sergey g. porsev and andrei derevianko , phys . a * 73 * , 012501 ( 2006 ) bijaya k. sahoo , rajat chaudhuri , p. p. das and debashis mukherjee , phys . lett . * 96 * , 163003 ( 2006 ) b. p. das , k. v. p. latha , b. k. sahoo , c. sur , r. k. chaudhuri and d. mukherjee , j. of theoretical and computational chemistry * 4 * , 1 ( 2005 ) b. k. sahoo , s. majumder , r. k. chaudhuri , b. p. das and d. mukherjee , j. phys . b * 37 * , 3409 ( 2004 ) b. k. sahoo , md . r. islam , b. p. das , r. k. chaudhuri and d. mukherjee , phys . a 74 , 062504 ( 2006 ) a. redfors , astron . astrophys . * 249 * , 589 ( 1991 ) b. n. g. guthrie , astrophysics and space sciences * 10 * , 156 ( 1971 ) s. j. adelman , mnras * 239 * , 487 ( 1989 ) m. s. dimitrijevic and s. sahal - brechot , journal of applied spectroscopy * 65 * , 492 ( 1998 ) + m. s. dimitrijevic and s. sahal - brechot , joint european and national meeting ( 1997 ) tomas brage , glen m. wahlgren , s. g. johansson , d. s. leckrone and c. r. proffitt , the astrophysical journal , * 496 * , 1051 ( 1998 ) s. t. maniak , l. j. curtis , c. e. theodosiou , r. hellborg , s. g. johansson , i. martinson , r. e. irving and d. j. beideck , astron . astrophys . * 286 * , 978 ( 1994 ) m. s. safronova , w. r. johnson and a. derevianko , phys . a * 60 * , 4476 ( 1999 ) w. r. johnson , d. r. plante and j. sapirstein , advances in at . mol . and optical physics * 35 * , 255 ( 1995 ) d. mukherjee , r. moitra and a. mukhopadhyay , mol . physics * 33 * , 955 ( 1977 ) i. lindgren , _ a coupled - cluster approach to the many - body pertubation theory for open - shell systems _ , in per - olov lwdin and yngve hrn , editors , atomic , molecular , and solid - state theory , collision phenomena , and computational methods , international journal of quantum chemistry , quantum chemistry symposium * 12 * , pages 33 - 58 , john wiley & sons , march 1978 . b. k. sahoo , b. p. das , r. k. chaudhuri and d. mukherjee , phys . a * 75 * , 032507 ( 2007 ) b. k. sahoo , t. beier , b. p. das , r. k. chaudhuri and d. mukherjee , j. phys . b * 39 * , 355 ( 2006 ) i. lindgen and j. morrison , _ atomic many - body theory _ , edited by g ecker , p lambropoulos , and h walther ( springer - verlag , berlin , 1985 ) see : http://physics.nist.gov/physrefdata/handbook/element@xmath69name.htm gabriel l. epstein and joseph reader , j. of opt . soc . of america * 65 * , 310 ( 1975 )

lifetimes of the first two metastable states in sc@xmath0 and y@xmath0 are determined using the relativistic coupled - cluster theory . there is a considerable interest in studying the electron correlation effects in these ions as though their electronic configurations are similar to the neutral alkali atoms , their structures are very different from the latter . we have made a comparative study of the correlation trends between the above doubly ionized systems with their corresponding neutral and singly ionized iso - electronic systems . the lifetimes of the excited states of these ions are very important in the field of astrophysics , especially for the study of post - main sequence evolution of the cool giant stars . 1.0 cm

introduction theory and method of calculation results and discussions conclusion acknowledgment

we thank prof . j. natowitz for stimulating discussions . 99 b.a . li , l.w . chen and c.m . ko , g. giuliani , h. zheng and a. bonasera , v. baran , m. colonna , v. greco and m. di toro , a.w . steiner , m. prakash , j.m . lattimer and p.j . ellis , j.m . lattimer and m. prakash , a. bonasera , m. bruno , c.o . dorso and p.f . mastinu , j.b . _ , a. keli , j.b . natowitz and k.h . schmidt , m.b . tsang , f. zhu , w.g . lynch , a. aranda , d.r . bowman , r.t . de souza , c.k . gelbke , y.d . kim , l. phair , s. pratt , c. williams , h.m . xu and w.a . friedman , s. albergo , s. costa , e. costanzo and a. rubbino , s. das gupta , j. pan and m.b . tsang , h. xi , w.g . lynch , m.b . tsang , w.a . friedman and d. durand , l. qin , k. hagel , r. wada , j.b . natowitz , s. shlomo , a. bonasera , g. rpke , s. typel , z. chen , m. huang , j. wang , h. zheng , s. kowalski , m. barbui , m.r.d . rodrigues , k. schmidt , d. fabris , m. lunardon , s. moretto , g. nebbia , s. pesente , v. rizzi , g. viesti , m. cinausero , g. prete , t. keutgen , y. ei masri , z. majka and y.g . ma , k. hagel , r. wada , l. qin , j.b . natowitz , s. shlomo , a. bonasera , g. rpke , s. typel , z. chen , m. huang , j. wang , h. zheng , s. kowalski , c. bottosso , m. barbui , m.r.d . rodrigues , k. schmidt , d. fabris , m. lunardon , s. moretto , g. nebbia , s. pesente , v. rizzi , g. viesti , m. cinausero , g. prete , t. keutgen , y. ei masri and z. majka , r. wada , k. hagel , l. qin , j.b . natowitz , y.g . ma , g. rpke , s. shlomo , a. bonasera , s. typel , z. chen , m. huang , j. wang , h. zheng , s. kowalski , c. bottosso , m. barbui , m.r.d . rodrigues , k. schmidt , d. fabris , m. lunardon , s. moretto , g. nebbia , s. pesente , v. rizzi , g. viesti , m. cinausero , g. prete , t. keutgen , y. ei masri and z. majka , a. mekjian , a.z . mekjian , t.c . awes , g. poggi , c.k . gelbke , b.b . back , glagola , h. breuer and v.e . viola , jr , l.p . csernai and j.i . kapusta , g. rpke , s. shlomo , a. bonasera , j.b . natowitz , s.j . yennello , a.b . mcintosh , j. mabiala , l. qin , s. kowalski , k. hagel , m. barbui , k. schmidt , g. giulani , h. zheng and s. wuenschel , s. wuenschel _ et al . _ , h. zheng and a. bonasera , h. zheng and a. bonasera , h. zheng , g. giuliani and a. bonasera , h. zheng , g. giuliani and a. bonasera , h. zheng , g. giuliani and a. bonasera , m. papa , t. maruyama and a. bonasera , m. papa , g. giuliani and a. bonasera , j. mabiala , a. bonasera , h. zheng , a.b . mcintosh , z. kohley , p. cammarata , k. hagel , l. heilborn , l.w . may , a. raphelt , g.a . souliotis , a. zarrella and s.j . yennello , b.c . stein , a. bonasera , g.a . souliotis , h. zheng , p.j . cammarata , a.j . echeverria , l. heilborn , a.l . keksis , z. kohley , j. mabiala , p. marini , l.w . may , a.b . mcintosh , c. richers , d.v . shetty , s.n . soisson , r. tripathi , s. wuenschel and s.j . yennello , j. esteve _ et al . _ , t. mller _ et al . _ , c. sanner _ et al . _ , westbrook , p. marini _ et al . _ in preparation . r. j. charity , computer code gemini , see http://www.chemistry.wustl.edu/@xmath31rc j.o . newton , d.j . hinde , r.j . charity , j.r . leigh , j.j.m . bokhorst , a. chatterjee , g.s . foote and s. ogaza , r.j . charity , m.a . mcmahan , g.j . wozniak , r.j . mcdonald , l.g . moretto , d.g . sarantites , l.g . sobotka , g. guarino , a. panteleo , l. fiore , a. gobbi and k. hildenbrand , d.r . bowman , g.f . peaslee , n. colonna , r.j . charity , m.a . mcmahan , d. delis , h. han , k. jing , g.j . wozniak , l.g . moretto , w.l . kehoe , b. libby , a.c . mignerey , a. moroni , s. angius , i. iori , a. pantaleo and g. guarino , r.j . charity , r.j . charity , l.g . sobotka , r.j . charity , j.tke and w.u . schrder , r.j . charity , w.d . tian , y.g . cai , d.q . fang , w. guo , w.q . shen , k. wang , h.w . wang and m. veselsky , p. zhou , w.d . tian , y.g . cai , d.q . fang , and h.w . wang , l. landau and f. lifshits , statistical physics ( pergamon , new york ) 1980 . k. huang , statistical mechanics ( j. wiley and sons , new york ) 1987 , second ed .. l.g . moretto , j.b . elliott , l. phair and p.t . lake , j.b . elliott , p.t . lake , l.g . moretto and l. phair , j.b . _ , j. pochodzalla _ et al . _ , a. bonasera , z. chen , r. wada , k. hagel , j. natowitz , p. sahu , l. qin , s. kowalski , t. keutgen , t. materna and t. nakagawa , j. mabiala _ et al . _ in preparation .

different methods to extract the temperature and density in heavy ion collisions are compared using a statistical model tailored to reproduce many experimental features at low excitation energy . the model assumes a sequential decay of an excited nucleus and a fermi gas entropy . we first generate statistical events as function of excitation energy but stopping the decay chain at the first step . in such a condition the ` exact ' model temperature is determined from the fermi gas relation to the excitation energy . from these events , using quantum and classical fluctuation methods for protons and neutrons , we derive temperature and density ( quantum case only ) of the system under consideration . additionally , the same quantities are also extracted using the double ratio method for different particle combinations . a very good agreement between the `` exact '' temperatures and quantum fluctuation temperatures is obtained , the role of the density is discussed . classical methods give a reasonable estimate of the temperature when the density is very low as expected . the effects of secondary decays of the excited fragments are discussed as well . # 1#2#3#4#1 * # 2 * , # 3 ( # 4 ) . the nuclear equation of state ( neos ) is one of the most challenging open problems today in particular the access to the symmetry energy part which carries relevant information especially for the nuclear ( astro)physics domain@xcite . a feasible way to experimentally constrain the neos is through the use of heavy ion collisions ( hic ) at intermediate energies involving nuclei with a large range of n / a concentrations . the systems created in such conditions are dynamical and strongly influenced by coulomb , angular momentum and other effects , thus the determination of ` quasi - equilibrium ' conditions is rather challenging . the neos can be determined if we are able to extract the temperature , density and pressure , or free energy from the hic data . several methods can be found in the literature to determine such quantities from available experimental data . classical approaches include slope temperature from the kinetic energy distribution of emitted ions @xcite , excited level energy distributions @xcite and double isotopic ratios @xcite . in particular the last method provides information on the density @xmath0 of the system at the time when fragments are emitted from a source at a given excitation energy per particle @xmath1 . a coalescence approach @xcite can also be used to estimate the density of the system . even though such a method is derived from classical physics , the obtained densities are higher than that of the double ratio method @xcite . this might be due to the introduction of a new parameter ( the coalescence radius @xmath2 ) which is determined from experimental data . such a parameter might in fact mimic important quantum effects @xcite which result in a generally higher density , for a given temperature @xmath3 , as compared to the one obtained from the double ratio method . more recently , a different method to extract @xmath3 and @xmath0 from the data has been proposed in @xcite based on fluctuations . the method was first used for quadrupole fluctuations within a classical approach , in order to obtain the temperature of the system . later on this method was applied to quantum systems for which the temperature is naturally connected to the density . in such a scenario , particle multiplicity fluctuations are used in order to pin down @xmath3 and @xmath0 from experimental data and modeling @xcite . an important first distinction between fermions and bosons is necessary in order to evidence important quantum effects such as fermion quenching ( fq ) and bose - einstein condensation ( bec ) @xcite . a correction due to coulomb effects for bosons and fermions was also introduced in refs . @xcite . are used in the left and right hand side panels . the exact value of @xmath3 obtained from the fermi gas relation , eq . ( [ energy ] ) , is given by the full ( red ) lines . available experimental data from the current literature are given by the ( blue ) dashed line @xcite and open stars @xcite . open ( full ) circles and open ( full ) squares refer to the classical ( quantum ) fluctuation method for protons and neutrons respectively . the full upward triangles refer to the total density . the dotted lines refer to the total densities estimated from level density for different @xmath4 , eq . ( [ leveldens ] ) . ] in order to distinguish among different approaches and test their region of validity , we have applied the double ratio method , the classical and quantum fluctuation methods to analyze ` events ' obtained from a commonly used statistical model dubbed as gemini @xcite . similar studies using the slope temperature have been reported in refs . @xcite . the model assumes a sequential statistical decay of a hot source of mass ( @xmath5 ) and charge ( @xmath6 ) , with @xmath1 and a given total angular momentum @xmath7 , which we assume equal to zero for simplicity in this work . we fix @xmath8 and @xmath9 also in order to compare to many calculations based on the constrained molecular dynamical model ( comd ) which we have performed before @xcite . the statistical model assumes that a hot source decays into a small fragment @xmath10 and a daughter nucleus @xmath11 . in general both fragments can be still excited and decay again into other fragments and so on until all excitation energy is transformed into kinetic energy of fragments and the q - value determined from the initial source and the final fragments . at each decay step , the probability of the process is determined from the entropy which is assumed to be that of a fermi gas @xcite : @xmath12 corresponding to an excitation energy @xmath13 both equations can be derived from a simple low temperature approximation of a fermi gas . the level density parameter @xmath14 in such approximation is given by : @xmath15 for ground state density @xmath16 mev . in order to take into account experimental observations , the parameter @xmath4 in the model is adjusted to a smaller value which could depend on excitation energy as well @xcite . for our purposes we will use two fixed @xmath4 values of 7.3 mev and 15 mev since our goal is to test different methods to determine @xmath3 and @xmath0 from the model data . in fact , in the model , the temperature can be derived from eq . ( [ energy ] ) if we stop the simulation after the first decay step . the following steps take into account a decreasing temperature due to particle emissions in previous decays , thus the @xmath3 determination becomes ` tricky ' and we will discuss it later on in the paper . the model does not implicitly assume any density , even though one might naively think that the density of the system is that of the excited nucleus in its ground state density . this is not however correct since the fermi gas relation is assumed . in particular the level density is connected to density by the relation : @xmath17 . ( top panels ) the asterisks and crosses refer to the @xmath3 obtained from dr using dth@xmath18 and pnth combinations respectively . ( bottom panels ) open diamonds , open crosses and full downward triangles refer to p , n and total densities obtained from the dth@xmath18 dr . open circles , squares and upward triangles refer to the same quantities from the pnth dr . ] thus @xmath16 mev implies a nuclear ground state density , while @xmath19 mev results in @xmath20 . we stress here that other effects can be invoked to justify the smaller value of @xmath4 , such as an effective mass or momentum dependence of the neos @xcite ; however in the simple fermi gas used in the model , the eq . ( [ leveldens ] ) is the natural explanation . this has an important consequence and in fact , we would naively expect that the different approaches to determine the density will give values compatible to the estimate above , at least asymptotically i.e. at high @xmath3 where different effects , such as coulomb barriers and q - values become less important . at low @xmath3 , these effects are important and will determine the probability of decay into a given channel rather than another . we anticipate that the decay probability at low @xmath3 will effectively decrease the values of the densities obtained from all methods , which might correspond to the fact discussed in comd simulation , that the emitted particles at low @xmath3 are located in a low density region of the nuclear surface : a low @xmath3 will correspond to a lower @xmath0 . and [ fig2 ] . ( bottom panels ) the full triangles , open triangles and asterisks refer to qf , dth@xmath18 and pnth dr respectively . ] we also point out that the model assumes a barrier penetration of particles at various excitation energies @xcite . an effective radius is assumed for the system @xmath21 . for a nucleus of mass @xmath8 this gives an effective radius @xmath22 which is equivalent to a system having a density less than one third of the normal nuclear density . of course this assumption might be in contrast with the density value obtained from the fermi gas relation . however our goal is not to modify the model but to use it as a test bench , keeping in mind that these assumptions might be justified at low excitation energies ( for which the model was proposed ) and not at higher ones where fragmentation will dominate . for all steps model simulations . ] using the gemini code available in the literature @xcite , we have generated one million events for each excitation energy ( or initial @xmath3 ) . first we discuss the results for the simulations stopped at the first decay step where the relation of the excitation energy and temperature is given by eq . ( [ energy ] ) . in figure [ fig1 ] we plot the temperature @xmath3 ( top panels ) and the density @xmath23 ( bottom panels ) as function of the excitation energy per particle of the intial hot system . two different values of the @xmath4 fermi gas parameter are used in the left and right hand side panels . the exact value of @xmath3 obtained from the fermi gas relation , eq . ( [ energy ] ) , is given by the full ( red ) lines . available experimental data from the current literature are given by the dashed ( blue ) lines @xcite and open stars @xcite , which we have reported for reference purposes only . the quantum fluctuation ( qf ) method , both for neutron and proton , agrees rather well with the exact result as expected since the basic assumption in the method and in the gemini model is the same , i.e. a nucleus made of fermions . the classical fluctuation method ( cf ) agrees with the exact method especially for the neutron case and at low excitation energies for both protons and neutrons @xcite . the reason for this behavior could be explained from the bottom part of figure [ fig1 ] . the densities estimated only from the qf method ( the cf does not determine a density since the multiplicity fluctuation are equal to one classically ) are very low especially for neutrons . we expect that at low densities and relatively high @xmath3 , classical and quantum methods should give similar results . as we will show the density obtained using the double ratio method is even smaller than the one obtained from qf . in figure [ fig1 ] , the proton density is given by full circles and the neutron density by full squares , while the total density is given by full triangle symbols . all densities have been normalized to their respective ground state values , i.e. @xmath24 , @xmath25 and @xmath26 . the reason why we have extended the model to such high excitation energies where it is not necessarily justified , is because we wanted to show that the estimated total density tends asymptotically to the value estimated from the fermi gas relation , eq . ( [ leveldens ] ) , using the respective @xmath4 values which are given by the dotted horizontal lines in figure [ fig1 ] . at low excitation energies , different q - values for particle emissions and barrier penetrations modify the fluctuations given by the fermi gas entropy , eq . ( [ entropy ] ) , which results in lower densities as displayed . if these effects would be turned off , then fluctuations would arise from the fermi gas entropy as for the quantum fluctuation method which is based on the same fermi gas assumption @xcite . the discussion above can be extended to the double ratio ( dr ) method @xcite and reported in figure [ fig2 ] . the asterisks and crosses refer to the @xmath3 obtained from dr using deuteron , triton,@xmath27 , @xmath18 ( dth@xmath18 ) and p , n , t , @xmath27 ( pnth ) combinations , respectively . the celebrated plateau of the caloric curve is observed in figure [ fig2 ] ( top panels ) especially for the dth@xmath18 combination mostly used in the literature @xcite . notice the peculiar and maybe surprising result obtained using the pnth dr for which @xmath3 goes down for increasing excitation energy ( top panels in figure [ fig2 ] ) . a similar opposite behavior for the two cases is obtained for the densities ( bottom panels in figure [ fig2 ] ) . in particular the pnth densities decrease for increasing excitation energy . all densities estimated from dr are much smaller than those from qf and do not asymptotically tend to the value obtained from the fermi gas relation , eq . ( [ leveldens ] ) . of course this is not surprising since we are using classical physics to estimate quantities obtained from a model based on a quantum ( fermi - gas ) system . for all steps model simulations . ] . ] another way to visualize the results is by plotting density ( top panels ) and the difference of neutron and proton density ( bottom panels ) as function of @xmath3 as reported in figure [ fig3 ] . now the surprising differences in the densities obtained from the pnth and dth@xmath18 cases in figure [ fig2 ] are not observed : the two dr methods agree which simply tells us that the control parameter is @xmath3 and not the excitation energy as it should be in a statistical model . the densities differ greatly in the qf and dr methods as observed before . equation ( [ leveldens ] ) and the available data support higher densities @xcite . furthermore the qf results tend asymptotically to the value expected from the fermi gas and the used @xmath4 parameters . notice the large difference between @xmath28 and @xmath29 densities as obtained in different approaches ( bottom panels in fig . [ fig3 ] ) . in particular all different methods fail to reproduce the initial source value of zero @xmath30 which should be recovered at high @xmath3 . this is however a failure of the statistical model which we have pushed at high excitation energies where it is not justified . we notice that in a two - component system phase transition , the quantity plotted in the bottom part of figure [ fig3 ] , could be considered as an order parameter @xcite . for completeness in figures [ fig4 ] and [ fig5 ] we display the results obtained when all steps in the statistical decay model are taken into account . these figures should be compared to figures [ fig1 ] and [ fig2 ] respectively . we observe generally a decrease of @xmath3 and an increase of @xmath0 compared to the first step results . this result implies that , if the general assumption of a sequential decay is correct , then the derived @xmath3 and @xmath0 estimated in the different methods are effective values influenced by the secondary decay . within the fluctuation method , it seems that the initial @xmath3 is somehow between the classical and the quantum cases , while the dr method fail in all cases . however , as we have seen in figure [ fig3 ] , plots of @xmath0 and @xmath3 as function of excitation energy might be misleading as in the pnth and dth@xmath18 cases . in figure [ fig6 ] we plot @xmath0 as function of @xmath3 obtained from different assumptions both at the first decay and all decay steps . as we see in the figure all results roughly collapse in single curves , especially results from the dr method , which suggests that indeed the values of @xmath3 might shift down due to the secondary decay , however the corresponding density is also modified in such a way to collapse in a single curve . this result should be compared to similar calculations using comd , see fig . ( 26 ) in ref . @xcite . in conclusion , in this paper we have compared different proposed methods to extract density and temperature using a statistical sequential model . we have shown that the model observables are better reproduced by the quantum fluctuations method since the same physical ingredient , the fermi gas , is used . double ratios fail because of the classical assumptions as it should be . however , the feature that different ratios give different @xmath3 and @xmath0 as function of excitation energy is misleading . an agreement of the different particle ratios is observed when the temperature is used as a control parameter as it should be in a statistical environment . secondary decays support again the qf method as compared to the dr and differences might be highlighted by plotting densities as function of the control parameter @xmath3 .

acknowledgement

chemical properties of the galactic bulge and inner disk should intersect in a given region . in this region bulge characteristics such as diversity in chemical abundances should meet disk properties such as the inner edge of the radial gradient of chemical abundances . the main goal of this work is to separate intermediate mass populations from the bulge and inner disk . this was made through the analysis of the abundance distribution of planetary nebulae ( pn ) located in the inner galaxy . new abundances for a pn sample were derived , then added to data from the literature for other objects . these results were combined and then statistically analyzed , in order to estimate a galactocentric distance where bulge and disk characteristics better separate . ultimately , we intend to provide an additional constraint to chemical evolution models for the galaxy . we selected 56 objects in the inner galaxy based on their positions with respect to the galactic center ( @xmath0 , @xmath1 ) , radio fluxes at 5 ghz and optical angular diameters ( @xmath2 ) , according to these criteria : i ) @xmath3 and @xmath4 ; ii ) f(5 ghz ) @xmath5 100 mjy ; iii ) @xmath6 12 " . figure [ fig1 ] indicates the position of the objects for which new abundances were derived ( in red ) , and those from the literature ( in blue ) . the figure also shows the contours of the galactic bulge as seen from the infrared observations of the cobe / dirbe satellite ( ( * ? ? ? * weiland et al . 1994 ) ) . observations were performed with the brazilian 1.60 m telescope located at pico dos dias in southeast brazil , with a boller & chivens cassegrain spectrograph . each object was observed at least twice . for all objects we used a 1.5 arcsec width slit . flux calibration was secured through the observation of spectrophotometric standards ( at least three each night ) . a 300 l / mm grating allowed a dispersion of 2.4 / pixel . details for the observations , derived physical parameters and abundances for the sample are given by ( * ? ? ? * cavichia et al . ( 2008 ) ) . we used the derived chemical abundances to build distance - independent correlations between @xmath7-elements . since they are produced basically by type - ii supernovae , their abundances reflect the chemical evolution of the galaxy . these correlations are shown in fig.[fig2 ] . mean errors for each element are given in the upper left corner of each graph . it is easy to see that @xmath7-element abundances are correlated , in the sense that older objects , formed in a poorer interstellar medium , present lower abundances of all these elements . the same effect appears both for our sample and abundances collected from the literature . therefore these values reflect the abundances of the interstellar medium at the progenitor formation epoch , indicating the chemical evolution of the galaxy . * statistical distances were selected for all objects in the sample using the statistical scales of ( * ? ? ? * zhang ( 1995 ) ) and ( * ? ? ? * cahn et al . ( 1992 ) ) . since these scales are internally consistent but can present large differences for a given object , we chose to use them separately , avoiding mixing distances from different scales . * once the individual galactocentric distances were defined , we divided the sample into two distance groups : one with distances smaller than a given limit , which we called group i , and other with distances larger than this limit , which we called group ii . for each group average abundances were calculated for the available elements . we considered limits from 0.1 to 3.6 kpc , in 0.7 kpc steps and , for each step , average abundances were calculated . the results are shown in figs.[fig3 ] and [ fig4 ] . figure [ fig3 ] was built using the ( * ? ? ? * cahn et al . ( 1992 ) ) distance scale , and fig.[fig4 ] using the ( * ? ? ? * zhang ( 1995 ) ) scale . in each figure , groups i and ii are represented respectively in black and red . each pair of black / red points represents the average abundance for that element adopting a given limit for the bulge - disk interface , and therefore each plot shows how the differences between both groups evolve adopting distinct limits to define them . examining fig.[fig3 ] we see that for the @xmath7-elements oxygen , sulfur and argon the differences between both samples reach a minimum between 2.5 and 3 kpc . for argon this effect is not clear , but already expected since the ionization correction factor used to derive its abundance can lead to uncertainties larger than those for other elements . for nitrogen and helium , whose abundances reflect the chemical enrichment processes occurring during the evolution of the progenitor star , this minimum difference at 2.5 kpc does not appear , as expected . figure [ fig4 ] shows the same plots using the ( * ? ? ? * zhang ( 1995 ) ) scale . examining again the @xmath7-elements , we see now that a minimum difference appears around 1.5 or 2 kpc , depending on the element . in order to ascertain a value for the galactocentric distance of the bulge - disk interface based on the intermediate mass population represented by this planetary nebulae sample , we performed a kolmogorov - smirnov test for groups i and ii at each step of the procedure described above . this test allowed to check which step in galactocentric distances separates more clearly both subsamples and resulted in a value of 2.9 kpc using the ( * ? ? ? * cahn et al . ( 1992 ) ) scale . using the ( * ? ? ? * zhang ( 1995 ) ) scale , the same test resulted in a better value of 1.5 kpc to the interface , however we adopted the 2.9 kpc value based on additional evidences that also point out to a distance around 3 kpc . according to ( * ? ? ? * tiede & terndrup ( 1999 ) ) , the abundance distribution in the inner galaxy points to a discontinuity at 3 kpc . at the same distance the gas density decreases rapidly ( ( * ? ? ? * portinari & chiosi 2000 ) ) . the galaxy evolution models also point out to a severe decrease in the stellar formation rate for galactocentric radii smaller than 3 kpc ( ( * ? ? ? * portinari & chiosi 1999 ) ) . combining these evidences with our results , we can point out to a bulge disk interface for the intermediate mass population , marking therefore the outer border of the bulge and inner border of the disk , at 3 kpc . finally , table [ tab1 ] shows the average abundances for bulge and inner disk objects , adopting the 2.9 kpc interface . these abundances are similar , taking into account the large dispersion found in the bulge , but it can be seen that the bulge abundances of @xmath7-elements are systematically lower than the inner disk abundances , so that the separation between the two populations can be clearly defined .

we report the spectrophotometric investigation of a planetary nebula sample located at the disk - bulge interface of the milky way . the main goal of this work was to determine the galactocentric distance where , according to the intermediate mass population , bulge and disk properties separate . in order to derive such distance , new abundances were derived for a sample of pn located at this region , and the results were combined with additional data from the literature . the abundance analysis indicates a chemical abundance distribution similar to that derived from bulge stars , as already pointed out by other authors . statistical distance scales were then used to study the distribution of chemical abundances across the disk - bulge connection . a kolmogorov - smirnov test was used to find the distance in which the chemical properties of these regions better separate , resulting in a best value of 2.9 kpc to define the inner limit of the disk .

introduction sample and observations results

the aim of weather forecasting is to give a good prediction of the future states of the atmosphere on the basis of present observations and mathematical models describing the dynamics ( physical behaviour ) of the atmosphere . these models consist of sets of non - linear partial differential equations which have only numerical solutions . the problem with these numerical weather prediction models is that the solutions highly depend on the initial conditions which are always in a way or in another not fully accurate . a possible solution to address this problem is to run the model with different initial conditions and produce ensembles of forecasts . with the help of ensembles one can estimate the distribution of future weather variables which leads us to probabilistic weather forecasting @xcite . the ensemble prediction method was proposed by @xcite and since its first operational implementation @xcite it became a widely used technique all over the world . however , despite e.g. the ensemble mean gives a better estimate of a meteorological quantity than most or all of the ensemble members , the ensemble is usually under - dispersive and in this way , uncalibrated . this phenomena was observed at several operational ensemble prediction systems , for an overview see e.g. @xcite . the bayesian model averaging ( bma ) method for post - processing ensembles in order to calibrate them was introduced by @xcite . the basic idea of bma is that to each ensemble member forecast corresponds a conditional probability density function ( pdf ) that can be interpreted as the conditional pdf of the future weather quantity provided the considered forecast is the best one . then the bma predictive pdf of the future weather quantity is the weighted sum of the individual pdfs corresponding to the ensemble members and the weights are based on the relative performances of the ensemble members during a given training period . in @xcite the bma method was successfully applied to obtain 48 hour forecasts of surface temperature and sea level pressure in the north american pacific northwest based on the 5 members of the university of washington mesoscale ensemble @xcite . these weather quantities can be modeled by normal distributions , so the predictive pdf is a gaussian mixture . later @xcite developed a discrete - continuous bma model for precipitation forecasting , where the discrete part corresponds to the event of no precipitation , while the cubic root of the precipitation amount ( if it is positive ) is modeled by a gamma distribution . in @xcite the bma method was used for wind speed forecasting and the component pdfs follow gamma distribution . finally , using von mises distribution to model angular data @xcite introduced a bma scheme to predict surface wind direction . in the present work we apply the bma method for calibrating ensemble forecasts of wind speed produced by the operational limited area model ensemble prediction system ( lameps ) of the hungarian meteorological service ( hms ) called aladin - huneps @xcite . aladin - huneps covers a large part of continental europe with a horizontal resolution of 12 km and it is obtained by dynamical downscaling ( by the aladin limited area model ) of the global arpege based pearp system of mto france @xcite . the ensemble consists of 11 members , 10 initialized from perturbed initial conditions and one control member from the unperturbed analysis . this construction implies that the ensemble contains groups of exchangeable forecasts ( the ensemble members can not be distinguished ) , so for post - processing one has to use the modification of bma as suggested by @xcite . as it was mentioned in the introduction , bma post - processing of ensemble predictions was applied for wind speed data obtained from the hms . the data file contains 11 member ensembles ( 10 forecasts started from perturbed initial conditions and one control ) of 42 hour forecasts for 10 meter wind speed ( given in m / s ) for 10 major cities in hungary ( miskolc , szombathely , gyr , budapest , debrecen , nyregyhza , nagykanizsa , pcs , kecskemt , szeged ) produced by the aladin - huneps system of the hms , together with the corresponding validating observations , for the period between october 1 , 2010 and march 25 , 2011 . the forecasts are initialized at 18 utc , the startup speed of the anemometers measuring the validating observations is @xmath0 m / s . the data set is fairly complete , since there are only two days ( 18.10.2010 and 15.02.2011 ) when three ensemble members are missing for all sites and one day ( 20.11.2010 ) when no forecasts are available . figure [ fig : fig1 ] shows the verification rank histogram of the raw ensemble , that is the histogram of ranks of validating observations with respect to the corresponding ensemble forecasts . this histogram is far from the desired uniform distribution , in most of the cases the ensemble members either underestimate , or overestimate the validating observations ( the ensemble range contains the observed wind speed only in @xmath1 of the cases ) . hence , the ensemble is under - dispersive and in this way it is uncalibrated . to obtain a probabilistic forecast of wind speed the modification of bma gamma model of @xcite for ensembles with exchangeable members @xcite was used . the first idea is to have two exchangeable groups : one contains the control denoted by @xmath2 , the other one the 10 ensemble members corresponding to the different perturbed initial conditions which are denoted by @xmath3 , respectively . in this way we assume that the probability density function ( pdf ) of the forecasted wind speed @xmath4 equals : @xmath5 where @xmath6 $ ] , and @xmath7 is the conditional pdf corresponding to the ensemble members . as we are working with wind speed data , @xmath8 is a gamma pdf with mean @xmath9 and standard deviation @xmath10 . here we restrict both the mean and the standard deviation parameters to constant values for all ensemble members , which reduces the number of parameters and simplifies calculations . mean parameters @xmath11 are estimated with the help of linear regression , while weight @xmath12 and standard deviation parameters @xmath13 , by maximum likelihood method , using training data consisting of ensemble members and verifying observations from the preceding @xmath14 days ( training period ) . in order to handle the problem that the wind speed values under 0.1 m / s are considered to be zero , the maximum likelihood ( ml ) method for gamma distributions suggested by @xcite is applied , while the maximum of the likelihood function is found with the help of em algorithm @xcite . for more details see @xcite . once the estimated parameters for a given day are available , one can use either the mean or the median of the predictive pdf as a point forecast . based on a more careful look on the ensemble members there are some differences in the generation of the ten exchangeable ensemble members . to obtain them only five perturbations are calculated and then they are added to ( odd numbered members ) and subtracted from ( even numbered members ) the unperturbed initial conditions @xcite . figure [ fig : fig2 ] shows the plume diagram of ensemble forecast of 10 meter wind speed for debrecen initialized at 18 utc , 22.10.2010 . ( solid line : control ; dotted line : odd numbered members , dashed line : even numbered members ) . this diagram clearly illustrates that the behaviour of ensemble member groups @xmath15 and @xmath16 really differ from each other . therefore , in this way one can also consider a model with three exchangeable groups : control , odd numbered exchangeable members and even numbered exchangeable members . this idea leads to the following pdf of the forecasted wind speed @xmath4 : @xmath17 where for weights @xmath18 $ ] we have @xmath19 , while pdf @xmath7 and parameters @xmath20 are the same as for the model . obviously , both the weights and the parameters can be estimated in the same way as before . to 9 truecm ( a ) ( b ) to 9 truecm ( c ) ( d ) as an illustration we consider the data and forecasts for debrecen for two different dates 30.12.2010 and 17.03.2011 for models and . figures [ fig : fig3]a and [ fig : fig3]b show the pdfs of the two groups in model , the overall pdfs , the median forecasts , the verifying observations , the first and last deciles and the ensemble members . the same functions and quantities can be seen on figures [ fig : fig3]c and [ fig : fig3]d , where besides the overall pdf we have three component pdfs and three groups of ensemble members . on 30.12.2010 the spread of the ensemble members is quite fair and the ensemble range contains the validating observation ( 3.2 m / s ) . in this case the ensemble mean ( 3.5697 m / s ) overestimates , while bma median forecasts corresponding to the two- and three - group models ( 3.2876 m / s and 3.2194 m / s , respectively ) are pretty close to the true wind speed . a different situation is illustrated on figures [ fig : fig3]b and [ fig : fig3]d , where the spread of the ensemble is even higher , but all ensemble members underestimate the validating observation ( 6.1 m / s ) . obviously , the same holds for the ensemble mean ( 3.2323 m / s ) and due to the bias correction the bma median forecasts corresponding to models and also give bad results ( 3.3409 m / s and 3.0849 m / s , respectively ) . to check the performance of probabilistic forecasts based on models and and the corresponding point forecasts , as a reference we use the ensemble mean and the ensemble median . we compare the mean absolute errors ( mae ) and the root mean square errors ( rmse ) of these point forecasts and also the mean continuous ranked probability scores ( crps ) @xcite and the coverages and average widths of @xmath21 and @xmath22 prediction intervals of the bma predictive probability distributions and of the raw ensemble . we remark that for mae and rmse the optimal point forecasts are the median and the mean , respectively @xcite . further , given a cumulative distribution function ( cdf ) @xmath23 and a real number @xmath4 , the crps is defined as @xmath24 the mean crps of a probability forecast is the average of the crps values of the predictive cdfs and corresponding validating observations taken over all locations and time points considered . for the raw ensemble the empirical cdf of the ensemble replaces the predictive cdf . the coverage of a @xmath25 prediction interval is the proportion of validating observations located between the lower and upper @xmath26 quantiles of the predictive distribution . for a calibrated predictive pdf this value should be around @xmath27 . data analysis provided below was performed with the help of the ensemblebma package of r @xcite . as a first step the length of the appropriate training period was determined , then the performances of the bma post - processed ensemble forecasts corresponding to models and were analyzed . according to the results of e.g. @xcite to determine the length of the training period to be used we compare the mae values of bma median forecasts , the rmse values of bma mean forecasts , the crps values of bma predictive distributions and the coverages and average widths of @xmath22 and @xmath21 bma prediction intervals for training periods of length @xmath28 calendar days . in order to ensure the comparability of the results we consider verification results from 02.12.2010 to 25.03.2011 ( 114 days ) . consider first the two - group model . on figure [ fig : fig4 ] the average widths and coverages of @xmath21 and @xmath22 bma prediction intervals are plotted against the length of the training period . the average widths of the prediction intervals show an increasing trend , so shorter training periods yield sharper forecasts . coverages of @xmath21 and @xmath22 prediction intervals are not monotonously increasing , too . for short training periods the coverage of the @xmath21 prediction interval oscillates around the correct @xmath21 , but for training periods not shorter than 17 days it stays above this level . the coverage of the @xmath22 prediction interval stabilizes above the correct @xmath22 for training periods longer than 24 days . hence , to have calibrated forecasts , one should choose a training period not shorter than 25 days . figure [ fig : fig5 ] shows crps values of bma predictive distribution , mae values of bma median forecasts and rmse values of bma mean forecasts as functions of the training period length . crps and rmse both take their minima at 28 days , the corresponding values are @xmath29 and @xmath30 , respectively . mae takes its minimum of @xmath31 at 30 days , while the second smallest value ( @xmath32 ) is obtained with a training period of length 28 days . this means that for model a 28 days training period seems to be reasonable and training periods longer than 30 days can not be taken into consideration . similar conclusions can be drawn from figures [ fig : fig6 ] and [ fig : fig7 ] for the three - group model . in this case the @xmath21 and @xmath22 prediction intervals are slightly narrower than the corresponding intervals of model , their coverages stabilize above the correct @xmath21 and @xmath22 for training periods longer than 17 and 24 days , respectively . crps and mae plotted on figure [ fig : fig7 ] both reach their minima of @xmath33 and @xmath34 , respectively , at 30 days , while values @xmath35 and @xmath36 corresponding to training period of length 28 days are both the fourth smallest ones . rmse takes its minimum of @xmath37 at 27 days , and increases afterwards . the fourth smallest value ( @xmath38 ) again corresponds to 28 days , while the rmse corresponding to 30 days is significantly larger ( @xmath39 ) . moreover , @xmath21 and @xmath22 prediction intervals corresponding to 28 days are sharper than the appropriate prediction intervals calculated using training period of length 30 days ( @xmath40 and @xmath41 vs. @xmath42 and @xmath43 ) . hence , we suggest the use of a training period of length 28 days for both bma models . according to the results of the previous subsection , to test the performance of bma post - processing on the 11 member aladin - huneps ensemble we use a training period of 28 calendar days . in this way ensemble members , validating observations and bma models are available for 146 calendar days ( on 20.11.2010 all ensemble members are missing ) . first we check the calibration of bma post - processed forecasts with the help of probability integral transform ( pit ) histograms . the pit is the value of the bma predictive cumulative distribution evaluated at the verifying observations @xcite . the closer the histogram to the uniform distribution , the better is the calibration . on figure [ fig : fig8 ] the pit histograms corresponding to two- and three - group bma models and are displayed . compared to the verification rank histogram of the raw ensemble ( see figure [ fig : fig1 ] ) one can observe a large improvement with the use of calibration . however , these pit histograms are still not perfect , e.g. kolmogorov - smirnov test rejects uniformity both for the two- and for the three - group model . the corresponding @xmath44-values are @xmath45 and @xmath46 , respectively , so the pit of the two - group model is slightly better . .coverage and average width of prediction intervals . [ cols="<,^,^,^,^,^ " , ] finally , on figure [ fig : fig10 ] common bias parameters @xmath47 of both bma models investigated and standard deviation parameters @xmath13 of the two - group model are plotted , together with the differences in standard deviation parameters of three- and two - group models . bias parameters are rather stable , the relative standard deviations of @xmath48 and @xmath49 are @xmath50 and @xmath51 , respectively . hence , the bma mean forecast of a particular day is mainly determined by the corresponding ensemble forecasts . the standard deviation parameters show more variability , for @xmath52 and @xmath53 the relative standard deviations are equal to @xmath54 and @xmath55 for model , and @xmath56 and @xmath57 for model . in the present study the bma ensemble post - processing method is applied for the 11 member aladin - huneps ensemble of the hms to obtain 42 hour predictions for 10 meter wind speed . two different bma models are investigated , one assumes two groups of exchangeable members ( control and forecasts from perturbed initial conditions ) , while the other considers three ( control and forecasts from perturbed initial conditions with positive and negative perturbations ) . for both models a 28 days training period is suggested . the comparison of the raw ensemble and of the probabilistic forecasts shows that the mean crps values of bma post - processed forecasts are considerably lower than the mean crps of the raw ensemble . further , the mae and rmse values of bma point forecasts ( median and mean ) are also lower than the maes and rmses of the ensemble median and of the ensemble mean . the calibrations of bma forecasts are nearly perfect , the coverages of @xmath21 and @xmath58 prediction intervals are very close to the right values . the three - group bma model slightly outperforms the two - group one and in almost all cases yields a real mixture of gamma distributions . in this way one can conclude that bma post - processing of ensemble forecasts of wind speed data of the hms significantly improves the precision and calibration of the forecasts , its operational application is worth considering . * acknowledgments . * research was supported by the hungarian scientific research fund under grants no . otka t079128/2009 and otka nk101680 and by the tmop-4.2.2.c-11/1/konv-2012 - 0001 project . the project has been supported by the european union , co - financed by the european social fund . the authors are indebted to tilmann gneiting for his useful suggestions and remarks and to mt mile and mihly szcs from the hms for providing the data . 99 bao , l. , gneiting , t. , raftery , a. e. , grimit , e. p. and guttorp , p. ( 2010 ) bias correction and bayesian model averaging for ensemble forecasts of surface wind direction . rev . _ * 138 * , 18111821 . descamps , l. , labadier , c. , joly , a. and nicolau , j. ( 2009 ) ensemble prediction at mto france ( poster introduction by olivier riviere ) _ 31st ewglam and 16th srnwp meetings , _ 28th september 1st october , 2009 . available at : http://srnwp.met.hu/annual_meetings/2009/download/sept29/morning/posterpearp.pdf fraley , c. , raftery , a. e. , gneiting , t. and sloughter , j. m. ( 2009 ) ensemblebma : an r package for probabilistic forecasting using ensembles and bayesian model averaging . _ technical report _ 516r , department of statistics , university of washington . hornyi , a , kertsz , s. , kullmann , l. and radnti , g. ( 2006 ) the arpege / aladin mesoscale numerical modeling system and its application at the hungarian meteorological service . _ i d ojrs _ * 110 * , 203227 .

prediction of various weather quantities is mostly based on deterministic numerical weather forecasting models . multiple runs of these models with different initial conditions result ensembles of forecasts which are applied for estimating the distribution of future weather quantities . however , the ensembles are usually under - dispersive and uncalibrated , so post - processing is required . in the present work bayesian model averaging ( bma ) is applied for calibrating ensembles of wind speed forecasts produced by the operational limited area model ensemble prediction system of the hungarian meteorological service ( hms ) . we describe two possible bma models for wind speed data of the hms and show that bma post - processing significantly improves the calibration and precision of forecasts . _ key words : _ bayesian model averaging , gamma distribution , continuous ranked probability score .

introduction data the model and diagnostics results conclusions

the dynamical study of the lyness difference equation ( @xcite ) and its generalizations to higher order lyness type equations ( @xcite ) , or to difference equations with periodic coefficients ( @xcite ) , has been the focus of an active research activity in the last two decades . in more recent dates , lyness type equations have also been approached using different points of view : from algebraic geometry ( @xcite ) to the theory of discrete integrable systems ( @xcite ) . this paper deals with the problem of the integrability and non - integrability of non - autonomous planar lyness difference equations of the form @xmath5 where @xmath1 is a cycle of @xmath2 positive numbers , i.e. @xmath6 for all @xmath7 being @xmath2 the primitive period and we consider positive initial conditions @xmath8 and @xmath9 . as we will see , the behavior of the sequences @xmath4 can be essentially different according to whether @xmath3 , @xmath2 is a multiple of @xmath10 or it is not . in this section we summarize our main results on in terms of @xmath2 . we also give an account of the tools that we have developed for this study that we believe might be interesting by themselves . we start by introducing the notations and definitions used in the paper . given a periodic sequence @xmath1 of primitive period @xmath2 we will say that its _ rank _ is @xmath11 if @xmath12 the values @xmath13 will be usually called _ parameters_. in our context the recurrence is called _ persistent _ if for any sequence @xmath4 there exist two real positive constants @xmath14 and @xmath15 which depend on the initial conditions , such that for all @xmath16 for each @xmath2 , the _ composition maps _ are @xmath17 where each @xmath18 is defined by @xmath19 and @xmath20 are the @xmath2 elements of the cycle . when there is no confusion , for the sake of shortness , we also will use the notation @xmath21}:=f_{a_k,\ldots , a_2,a_1}. $ ] note that these maps are birational maps and are always well - defined in the open invariant set @xmath22 moreover @xmath23 and in general , @xmath24}(x_1,x_{2})=(x_{k+1},x_{k+2}).\ ] ] there are two concepts coexisting in this context , _ the non - autonomous invariants _ and _ the first integrals _ , that we will use in this paper . given a difference equation of the form , a non - autonomous invariant is a function @xmath25 , such that @xmath26 for all initial conditions and all @xmath27 on the other hand , when the difference equation has @xmath2-periodic coefficients a first integral is a function @xmath28 , which is a first integral for the discrete dynamical system generated by @xmath29},$ ] that is @xmath30}(x , y))=h(x , y),$ ] for all points in an open set . in terms of the recurrence @xmath31 for all initial conditions @xmath32 . we will relate both concepts in section [ se:3 ] . two analytic functions @xmath33 are said to be coprime if the points of the set @xmath34 are isolated . a function @xmath35 with @xmath36 and @xmath37 coprime , will be called a _ meromorphic function_. a _ meromorphic first integral _ of an analytic map @xmath38 is a meromorphic function @xmath39 such that @xmath40 observe that from this definition @xmath41 for all points of @xmath42 for which both terms of this last equality are well - defined . when @xmath36 and @xmath37 are polynomials then it is said that @xmath28 is a _ rational first integral_. similarly we can talk about _ meromorphic or rational invariants _ , and in this sense we will talk about _ rational _ or _ meromorphic integrability_. finally , we will say that a planar map @xmath43 has _ structurally stable numerical chaos ( ssnc ) _ when studying numerically several of its orbits , we observe that it presents all the features of a non - integrable perturbed twist map , that is : many invariant curves and , between them , couples of orbits of @xmath44-periodic points ( for several values of @xmath44 ) , half of them of elliptic type and the other half of hyperbolic saddle type . moreover the separatrices of these hyperbolic saddles intersect transversally , for instance see ( * ? ? ? * chapter 6 ) . this subsection collects the general outlines of all our results about the recurrence , in terms of @xmath45 figure 1 shows some typical behaviors of the orbits of @xmath29}$ ] . in fact we consider some maps @xmath46}$ ] , which are conjugate to @xmath29}$ ] , because the pictures are much more clear . see lemma [ gg ] for the definition of @xmath46}.$ ] * cases @xmath47 and other concrete integrable cases*. for @xmath48 it is already known that the recurrences are persistent . moreover , either each sequence @xmath4 is periodic , with period a multiple of @xmath2 or it densely fills at most @xmath2 disjoint intervals of @xmath49 , see @xcite . a key point for the proof is the existence of a rational first integral for @xmath29}$ ] . when @xmath50 we can also prove the existence of a similar first integral ( see corollary [ fi ] ) and the persistence of recurrence . moreover we are confident that the same characterization of the sequences @xmath4 holds but we have only been able to prove the result when @xmath51}$ ] has a unique fixed point in the first quadrant , see lemma [ lk=5 ] and proposition [ c6 ] . it is also satisfied that for any @xmath52 there are values @xmath13 , with primitive period @xmath2 and high rank , satisfying the property that all the sequences @xmath4 given by are either periodic , with period a multiple of @xmath2 , or they densely fill at most @xmath2 disjoint intervals , see theorem [ main ] . + all @xmath2 @xmath53 @xmath54 figure 1 : different possible behaviors of the orbits of @xmath46}$ ] , according to @xmath2 . other behaviors are possible for @xmath2 being a multiple of @xmath10 . * cases @xmath55 being a multiple of @xmath56*. when @xmath2 is a multiple of 5 ( from now on denoted by @xmath57 ) , apart from the behaviors described above there appear others for an open set of values of @xmath58 and initial conditions . for instance we can find sequences @xmath4 such that @xmath59 and others such that their adherence consists of @xmath2 points , see theorem [ propo5 ] . the existence for @xmath60 of values of @xmath61 for which the sequence @xmath4 has the first behavior has been already established in previous works , see ( * ? ? ? * example 5.43.1 ) or @xcite , but only for very concrete initial conditions and parameters @xmath62 . moreover in this case we can prove that for most values of the parameters the map @xmath29}$ ] has no meromorphic first integral ( see theorem [ nfi ] ) . furthermore the phase portrait of the map @xmath29}$ ] does not always coincide with the ones found in all the rational integrable cases , see for instance the second picture in figure 1 . in this case , apart of the celebrated lyness map @xmath63 which satisfies @xmath64 , there are values of the parameters @xmath13 such that @xmath29}^{m_k}=\operatorname{id}$ ] ( see corollary [ gp ] ) and other for which the number of fixed points of the maps is a 1-dimensional manifold , or @xmath65 or @xmath66 points , see lemma [ lk=5 ] . finally , when @xmath67 there are cases presenting ssnc , see section [ num ] . * cases @xmath68*. when @xmath69 , for some values of the parameters , @xmath13 , we have numerically found ssnc , see section [ num ] . in fact we prove in lemma [ chaos ] that based on these examples we can obtain values @xmath13 , all different , with a similar behavior for all the remaining values of @xmath2 . so for all these values of @xmath2 there are situations for which the sequence @xmath4 can have different behaviors to those given in the above situations . for instance there appear sequences which fill more than @xmath2 intervals . some concrete examples for @xmath70 are shown in section [ num ] . observe that as an application of our results we can show an interesting and curious phenomena that can be understood as a kind of chaos regularization " : consider a map @xmath71},$ ] which is a rationally integrable map , and a map @xmath72}$ ] which has chaotic behavior , then both maps @xmath73 and @xmath74 are rationally integrable because are of type @xmath51}$ ] . thus @xmath75 regularizes @xmath28 . finally note that the above results show that the only cases for which recurrence can have a rational invariant for all values of the parameters @xmath13 are @xmath76 in this subsection we present several results that we have obtained which we believe that are interesting by themselves . other technical results will be given in section [ pr ] . the first result is a necessary condition for the meromorphic integrability of planar maps near a fixed point . our approach follows the guidelines of poincar when he studied the same problem for ordinary differential equations , see @xcite and the references there in for the approach to ordinary differential equations . in section [ k5 ] we will apply the result below to study the case @xmath53 . [ th - int ] let @xmath77 an analytic map defined in @xmath42 , an open neighborhood of the origin , such that @xmath78 and @xmath79 is diagonalizable with eigenvalues @xmath80 and @xmath81 . assume that @xmath43 has a meromorphic first integral @xmath28 in @xmath82 . a. if @xmath83 then there exists @xmath84 @xmath85 , such that @xmath86 . b. if @xmath87 and @xmath88 then there exists @xmath89 such that @xmath90 when the map @xmath91 is real valued and of class @xmath92 the proof of theorem [ th - int ] can be adapted following the same steps . taking into account that in this case , when @xmath93 is an eigenvalue of @xmath79 , then @xmath94 is also , and we have to deal with the resonant condition @xmath95 we obtain the following result : let @xmath96 be a @xmath92 map such that @xmath97 and @xmath79 is diagonalizable , with eigenvalues @xmath80 and @xmath81 . assume that @xmath43 has a meromorphic first integral @xmath28 in @xmath82 . a. if @xmath98 , @xmath99 then there exists @xmath100 such that @xmath101 . b. if @xmath102 ( hence @xmath103 ) , then either @xmath104 or @xmath105 and there exists @xmath106 such that @xmath107 . c. if @xmath87 and @xmath88 then there exists a @xmath89 such that @xmath90 the above results will be applied to prove the meromorphic non - integrability of many cases when @xmath108 see theorem [ nfi ] . on the other hand , next result will be the key point to prove the existence of rational integrable cases for all @xmath109 see theorem [ main ] . for the recurrence ( [ eq ] ) we look for non - autonomous invariants of the form @xmath110 where @xmath111 with all the sequences of positive numbers . this method is introduced in @xcite and the special form of @xmath112 is inspired by this paper and the known invariant of the lyness recurrences , see @xcite . we prove : [ noautonom ] if the recurrence has an invariant of the form then @xmath113 and @xmath114 where @xmath115 satisfies that @xmath116 and @xmath117 [ fi ] ( i ) the non - autonomous @xmath2-periodic recurrence has invariants of the form if and only if @xmath3 . \(ii ) the first integrals of the maps @xmath29}$ ] , for @xmath118 corresponding to the invariants given in theorem [ noautonom ] are : @xmath119 our proof of theorem [ noautonom ] does not require the sequence of parameters @xmath1 to be periodic . observe that @xmath120 is the usual first integral ( invariant ) of the map @xmath121 associated to the classical lyness recurrence , see for instance @xcite . it is already known , see @xcite , that in the two and three periodic cases the functions @xmath122 and @xmath123 are first integrals of the maps @xmath124 and @xmath125 respectively . these first integrals play a crucial role for the understanding of the recurrence when @xmath126 , see again @xcite . to the best of our knowledge the existence of a first integral for the general non - autonomous 6-periodic case was not known . in section [ ri ] we use it to describe the dynamics in this case . also observe that , due to the form of the invariants , all the maps @xmath29}$ ] , for @xmath118 preserve a foliation of the plane given by biquadratic curves , which are elliptic except for a finite number of level sets . in fact these maps are particular cases of the celebrated qrt family of planar maps . perhaps a further algebraic - geometric approach , like the one presented in @xcite , by studying the maps induced by each @xmath29}$ ] on the corresponding elliptic surface could give more information about the reason why the cases @xmath3 are special . it is also interesting to notice that other integrable qrt maps with periodic coefficients have been found recently @xcite , as well as another major family of maps : the hirota - kimura - yahagi type ones @xcite . as we have already commented , it is known that for very concrete values of @xmath62 and suitable initial conditions , the behavior of @xmath4 is different to the ones appearing when @xmath48 and in particular is non - persistent , see ( * ? ? ? * example 5.43.1 ) or @xcite . this behavior can also be seen considering @xmath127 since @xmath128 it is clear that for @xmath129 the orbits of the points of the form @xmath130 with @xmath131 are unbounded . our next result allows to establish , when @xmath53 , the non - persistence of the recurrence for many values @xmath132 [ propo5 ] consider recurrence for @xmath53 . set @xmath133 if for all @xmath134 , @xmath135 and @xmath136 then is non - persistent . in fact , for an open set of initial conditions , @xmath137 and @xmath138 . in the above result , the open set of initial conditions for which the result holds is sometimes the whole first quadrant @xmath139 . for instance , this is the case when @xmath60 and @xmath140 and @xmath141 , when @xmath129 and @xmath142 , because @xmath143 is a linear map with a stable node at the origin . the rest of the paper is organized as follows . in section [ pr ] we introduce some preliminary results , while in section [ se:3 ] we prove the main tools described in section [ mt ] . section [ ri ] is devoted to the cases for which we find rational integrability , while in section [ k5 ] we prove the non - integrability results when @xmath2 is a multiple of five . finally , in section [ num ] we present some numerical evidence of chaos . this section contains some technical preliminary results and other known results that we will use in the proofs given in subsequent sections . the next result will be useful for our numerical simulations . as we will see , some new variables allows to `` observe '' much better the numerical non - integrability studied in section [ num ] . its proof is straightforward . [ gg ] the sequence in the variables @xmath144 becomes @xmath145 and the corresponding maps @xmath29}$ ] are conjugate to @xmath46}$ ] , where @xmath146}=g_{a_k , a_{k-1}\ldots , a_2,a_1 } , $ ] @xmath147 and each @xmath148 is defined on the whole plane , @xmath149 . observe that the maps @xmath46}$ ] are area preserving . in fact , it is easy to see that the maps @xmath29}$ ] and @xmath46}$ ] satisfy the following properties : [ mus ] for every choice of positive numbers @xmath150 a. the map @xmath29}$ ] preserves the measure @xmath151 or , in other words , it preserves the symplectic form @xmath152 . in consequence , it holds that @xmath153}(x , y))=\det(df_{[k ] } ( x , y))\,\mu(x , y),\ ] ] where @xmath154 b. the map @xmath46}$ ] preserves the lebesgue measure @xmath155 that is , it preserves the canonical symplectic form @xmath156 . hence , it holds that @xmath157 } ( x , y))\equiv1 $ ] . a related issue in connection with the above lemma is the fact ( ( * ? ? ? 1 ) ) that the group of symplectic birational transformations of the plane ( which is the group of birational transformations of @xmath158 which preserve the differential form @xmath159 ) is generated by compositions of the lyness map @xmath63 ( the @xmath10-periodic case of @xmath160}$ ] , with @xmath161 ) ; an scaling ; and a map of the form @xmath162 where the matrix @xmath163 . this was conjectured by usnich in @xcite and recently proved by blanc in the above mentioned reference . the next result allows to know the dynamics of each @xmath29}$ ] when the map has a smooth first integral , its applicability to characterize the dynamics of any integrable @xmath29}$ ] is guaranteed by lemma [ mus ] . [ teor ] let @xmath164 be an open set and let @xmath165 be a diffeomorphism such that it has a smooth regular first integral @xmath166 and there exists a smooth function @xmath167 such that for any @xmath168 @xmath169 then the following holds : a. if a level set @xmath170 is a simple closed curve invariant under @xmath43 , then the map @xmath43 restricted to @xmath171 is conjugate to a rotation . b. if @xmath171 is diffeomorphic to an open interval curve and invariant under @xmath43 , then the map @xmath43 restricted to @xmath171 is conjugate to a translation . the next result proves that all the fixed points @xmath172 of some map @xmath29},$ ] in @xmath149 such that @xmath173 , for all @xmath174 , is well defined are resonant . as we will see in proposition [ final ] , the hypothesis on the maps @xmath173 is unavoidable because for @xmath175 there appear some cancelations that make that the maps @xmath176}$ ] have as fixed point @xmath177 and this point can be of saddle type with arbitrary eigenvalues . in fact this property will be the key point for our proof of non - existence of meromorphic first integrals for most @xmath29 } , k=\dot{5}$ ] , see theorem [ nfi ] . [ determinantesu ] let @xmath178 be a fixed point of a composition map @xmath29}$ ] and such that @xmath173 is well defined for all @xmath174 . then @xmath179}(\mathbf{p}))=1.\ ] ] from lemma [ mus ] we know that @xmath46}$ ] is a symplectic map that preserves the canonical form , hence @xmath157})=1 $ ] . on the other hand , since @xmath29}$ ] and @xmath46}$ ] are conjugated maps ( lemma [ gg ] ) , the jacobian matrices of these maps at their corresponding fixed points have the same eigenvalues and the result follows . as a consequence of the above proposition we have : [ corolofftopic ] let @xmath178 be a @xmath11-periodic point of a composition map @xmath29}$ ] such that @xmath173 is well defined for all @xmath180 . then @xmath181}^m(\mathbf{p}))=1.\ ] ] the following lemma studies the number of fixed points of @xmath29}$ ] for @xmath182 [ lk=5 ] a. there is a unique fixed point of @xmath183 in @xmath139 and it satisfies @xmath184 b. there are either 0,1,2 or a continuum of fixed points of @xmath185 in @xmath139 and they satisfy @xmath186 c. let @xmath187 be the set of fixed points of map @xmath188 and let @xmath189 be the set of singular points of its first integral @xmath190 given in corollary [ fi ] . then @xmath191 and both sets coincide with the set of points of @xmath139 satisfying @xmath192 moreover @xmath193 \(i ) the set of fixed points of @xmath183 is exactly the set of points satisfying @xmath194 but it is not easy to handle these two equations . on the other hand , the equivalent condition @xmath195 lead us to the system of the statement . clearly both parabolas meet at a unique point in @xmath139 . \(ii ) studying the condition @xmath196 we obtain the system of the statement . the @xmath197-coordinate of a fixed point has to satisfy the quadratic equation @xmath198 from this equation we easily obtain the result . notice that simple cases having infinitely many fixed points appear for instance when @xmath199 and @xmath200 \(iii ) the two conditions given by @xmath201 directly lead to the system of the statement . the set of singular points of @xmath190 is formed by the points satisfying @xmath202 the two equations describing the above set exactly coincide again with the two equations given in the statement . so @xmath191 . that @xmath203 can be seen studying the behavior of the functions , @xmath204 and @xmath205 , near @xmath66 and @xmath206 . finally , we will also use the following results : [ lemanou ] the map @xmath207 is conjugate to the lyness map @xmath208 . observe that @xmath209 and @xmath210 . if we consider the linear map @xmath211 it holds that @xmath212 , as we wanted to prove . a nice consequence of the 5-global periodicity of the lyness map @xmath213 and the above lemma is the following result : [ gp ] recurrence with @xmath70 and @xmath214=[1/c^2,1/c , c , c^2,1/c^2,1/c,\ldots]$ ] is globally 20-periodic , i.e. @xmath215 for all this section is devoted to proving theorems [ th - int ] , [ noautonom ] and [ propo5 ] . write @xmath217 where @xmath218 and @xmath219 are homogeneous polynomials with degrees @xmath220 and @xmath221 , respectively , and @xmath222 denotes terms of order at least @xmath2 . firstly we prove that it is not restrictive to assume that @xmath223 and that if @xmath224 then @xmath225 is not constant . notice that if @xmath28 is a first integral then @xmath226 is also . hence we can assume that @xmath223 . if @xmath227 and @xmath228 for some @xmath229 , take @xmath230 . clearly @xmath231 is a new first integral of the form @xmath232 as we wanted to see . it is also clear that in a neighborhood of the origin we can assume that @xmath233 \(i ) we start studying the case @xmath83 . by imposing that @xmath28 is a first integral of @xmath43 in @xmath42 we have that @xmath234 by taking the lower order terms of the above equality we obtain @xmath235 define @xmath236 where @xmath237 are suitable non - negative integers . by using the homogeneity of @xmath238 and @xmath239 , equation becomes @xmath240 where notice that we have canceled the common factor of @xmath241 and @xmath219 . by introducing the polynomials in one variable @xmath242 @xmath243 with respective maximum degrees @xmath244 and @xmath11 , and @xmath245 @xmath246 equation becomes @xmath247 where we know that @xmath248 and @xmath249 are not identically zero and have no common root . notice that equality implies that if @xmath250 is a root of @xmath248 then @xmath251 is also , and hence @xmath252 for any @xmath253 , is a root of @xmath248 . since @xmath248 has at most @xmath244 roots , if @xmath254 we have that @xmath255 for some @xmath256 proving the theorem , because @xmath257 . a similar reasoning can be done for @xmath249 . hence it only remains to study the cases @xmath258 for some complex numbers @xmath259 and @xmath260 , @xmath261 remember that we know that both polynomials have no common roots . so , at least one of the two numbers @xmath262 or @xmath263 has to be zero . in any case , the equation becomes @xmath264 giving @xmath265 as we wanted to prove . \(ii ) when @xmath87 and @xmath88 equation , after dropping the common factor of @xmath218 and @xmath219 , becomes @xmath266 notice that @xmath267 , @xmath268 , and @xmath269 because , otherwise @xmath238 and @xmath239 would have @xmath270 as a common factor . hence @xmath271 and so @xmath261 therefore @xmath272 . since @xmath238 and @xmath239 have no common factor , we get that @xmath273 and so @xmath274 hence this last equality becomes @xmath275 then @xmath276 giving @xmath277 as we wanted to prove . to illustrate the above result in the next remark we present some examples of maps having ( or not having ) meromorphic first integrals . \(i ) the linear maps @xmath278 , with @xmath80 and @xmath81 satisfying the resonant condition @xmath86 are the simplest maps with meromorphic first integrals @xmath279 \(ii ) the map @xmath280 , with an eigenvalue 0 , has the first integral @xmath281 \(iii ) for maps with identically zero linear part we can have existence or not of meromorphic first integrals . for instance the map @xmath282 has the first integral @xmath283 on the other hand , by using the same tools that in our proof of theorem [ th - int ] , we can prove that the map @xmath284 has no meromorphic first integral . the condition that a function @xmath25 of the form ( [ formainvariant ] ) is a non autonomous invariant of the recurrence ( [ eq ] ) becomes @xmath285 for all @xmath286 and all @xmath287 . imposing that the each one of the coefficients of the 31 monomials @xmath288 vanishes identically and playing a little bit with these conditions we obtain that holds and moreover that @xmath289 from the first equation we get that @xmath290 plugging this equation in the second one we obtain that @xmath291 clearly the above equation holds if either @xmath115 is a 6-periodic sequence or @xmath292 in the first situation let us prove that if @xmath115 is a @xmath44-periodic sequence , @xmath293 then @xmath294 also has to be @xmath44-periodic . assume for instance that @xmath295 then using we have @xmath296 as we wanted to see . the other cases follow similarly . in the second situation we have that @xmath292 using this equation and equality we have that @xmath297 which is a well - known 6-periodic recurrence , as we wanted to see . finally , using six times we get that @xmath298 . before proving corollary [ fi ] , we explain here how non - autonomous invariants and first integrals are related . consider a recurrence with @xmath2-periodic coefficients and having a non - autonomous invariant @xmath25 that satisfies @xmath299 . then @xmath300 is a first integral of @xmath29}$ ] . conversely , if @xmath301 is a first integral then @xmath302 is a non - autonomous periodic invariant of the recurrence . these relations are used in the next corollary for constructing the first integrals of @xmath29},$ ] @xmath303 using the invariant found in the above theorem . notice also that there is another way to relate both concepts . indeed , it is possible to replace the non - autonomous @xmath2-periodic recurrence by an autonomous map on an enlarged phase space of dimension @xmath304 , simply considering the map @xmath305 in this case , any non - autonomous invariant is just an ordinary first integral for the above map . this approach is not used in this paper . \(i ) this result is proved along the proof of theorem [ noautonom ] , above . \(ii ) we only give the details for @xmath50 . the other cases follow similarly . we introduce the following notations for the non - autonomous 6-periodic recurrences : @xmath306 from the relations @xmath307 we obtain that @xmath308 hence , with the notations of theorem [ noautonom ] we get that @xmath309 hence @xmath310 is a first integral of @xmath51}$ ] , as we wanted to prove . we first prove theorem [ propo5 ] for @xmath311 [ lema5 ] consider recurrence with @xmath60 and @xmath312 for @xmath313 and satisfying @xmath314 then the recurrence is non - persistent . moreover for an open set of initial conditions @xmath315 and @xmath316 . a computation shows that @xmath317}(x , y)=\left(p_{1}(x , y),p_{2}(x , y)\right)$ ] where @xmath318 and @xmath319 first observe that , contrary to what it happens for @xmath29 } , k<4 $ ] , @xmath317}$ ] can be extended to a neighborhood of @xmath139 . note also that @xmath320 is a fixed point and @xmath321}(0,0)=\left ( \begin{array}{cc } \frac{1}{a_1 } & 0 \\ 0 & a_5 \end{array } \right).\ ] ] so , under our hypotheses , the origin is a hyperbolic fixed point of @xmath317}$ ] . arguing similarly with the shifted maps @xmath322 , @xmath323 , @xmath324 , and @xmath325 , we obtain that the origin is also a hyperbolic fixed point for these maps , with jacobian matrices @xmath326 respectively . condition ( [ minmax ] ) implies that at least there exist a parameter with a value less than one , and other grater that one . since there are no parameters with value equal to one and the sequence is cyclic we can choose two contiguous parameters and such that @xmath327 and @xmath328 . this implies that the origin is an attractive fixed point for some of the five shifted maps . for example , suppose that @xmath329 and @xmath330 , then the origin is an stable node for @xmath323 . taking an initial condition @xmath331 , with positive coordinates and in the basin of attraction of the origin for the corresponding shifted map we obtain that @xmath332 for the solution of equation ( [ eq ] ) with initial condition @xmath333 and @xmath334 . recall @xmath335 @xmath336 . thus the fact that @xmath316 follows because if some @xmath337 tends to @xmath320 then the second component of @xmath338 tends to @xmath206 . set @xmath29}$ ] for @xmath339 . we can write @xmath21}=f_{a_{5m},\ldots , a_{5m-4}}\circ\ldots\circ f_{a_{5},\ldots , a_{1 } } , $ ] so @xmath29}$ ] can be extended to a neighborhood of @xmath139 . furthermore , observe that @xmath340}(0,0)=df_{a_{5m},\ldots , a_{5m-4}}(0,0)\circ\ldots\circ df_{a_{5},\ldots , a_{1}}(0,0)=\left ( \begin{array}{cc } \frac{1}{\phi_2 } & 0 \\ 0 & \phi_1 \end{array } \right).\ ] ] similarly the jacobian matrices of the shifted maps have the form @xmath341 arguing as in proposition [ lema5 ] , the relation ( [ relaciophi ] ) implies that at least there exists a couple @xmath342 such that one of the values is greater than one and the other less than one . so the origin is an attractive fixed point for some of the @xmath11 shifted maps . now the proof follows again as in proposition [ lema5 ] . as we have already explained in subsection [ smr ] the cases @xmath343 are very similar and totally understood . from corollary [ fi ] we can prove in the next proposition a similar result when @xmath344 before stating the result we introduce the following notation : @xmath345 @xmath346 [ c6 ] for @xmath50 the recurrence is persistent . moreover if @xmath347 , any sequence @xmath4 generated by is either periodic , with period a multiple of @xmath348 , or it densely fills at most @xmath348 disjoint intervals of @xmath49 . we follow the same steps as in the proof of ( * ? ? ? 1 ) . to prove the persistence of , it suffices to show that each level curve @xmath349 is bounded . since @xmath350 we know that @xmath351 and the persistence follows . by lemma [ lk=5].(iii ) , under our hypotheses , the set of fixed points of @xmath51}$ ] and the set of singular points of @xmath352}$ ] coincide and consists of a single point . following again the same guidelines of the proof of ( * ? ? ? 1 ) , which in turn is based on ( * ? ? ? 2.1 ) , we prove that all the level curves of @xmath352}$ ] in @xmath139 , apart from the fixed point , are diffeomorphic to circles . hence by using lemma [ mus ] and theorem [ teor].(i ) the proposition follows . [ rc6 ] we believe that @xmath353 is the whole of @xmath354 but we have not been able to prove this equality . in any case it is easy to find sufficient conditions to ensure that some @xmath355 belongs to @xmath356 for instance , since @xmath357 when both numerators have no positive real roots the point @xmath355 is in the set , because the functions that we have derived are both increasing and so the curves defined by system cut in a single point . next result collects our integrability results for any @xmath358 [ main ] * for any @xmath359 , there exist sequences @xmath1 of prime period @xmath2 and rank @xmath2 such that @xmath21}=f_{a_k,\ldots , a_2,a_1 } $ ] is rationally integrable and the corresponding recurrence ( [ eq ] ) is persistent . * for any @xmath360 there exist sequences @xmath1 of prime period @xmath2 with the ranks as in table 1 , such that @xmath29}$ ] is rationally integrable and the corresponding recurrence ( [ eq ] ) is persistent . * moreover it is possible to take in all the above cases parameters @xmath361 such that each sequence @xmath4 is either periodic , with period a multiple of @xmath2 , or it densely fills at most @xmath2 disjoint intervals of @xmath49 . + [ cols="^,^,^,^,^,^,^,^,^,^,^,^",options="header " , ] + table 1 . possible ranks for integrable @xmath29}$ ] . we start by introducing some notation . given @xmath362 and @xmath363 positive , we consider the sets @xmath364 . assume that @xmath362 and @xmath363 are such that @xmath365 and consider @xmath366 where we have used expression . notice that @xmath367 . given any natural number @xmath368 we also consider @xmath11 sets @xmath369 , and define @xmath370}=\phi_m\circ \phi_{m-1}\cdots\circ \phi_1 $ ] . then @xmath371}(x , y)=\left(\frac{x}{\prod\limits_{i=1}^{m}a_i},\frac{y}{\prod\limits_{i=1}^m a_i c_i}\right).\ ] ] when @xmath273 we consider @xmath372}(x , y)=(x , y).$ ] finally , choosing the values of the parameters such that @xmath373 and @xmath374 , we obtain that for all @xmath375 , @xmath370}(x , y)=(x , y)$ ] . moreover , for these values of @xmath11 , the parameters @xmath362 and @xmath363 can be chosen such that @xmath376 . observe also that when @xmath377 it is not possible to choose @xmath378 and @xmath379 . when @xmath380 it is again possible but with @xmath381 . now we can start the proof of the theorem . first we study the cases @xmath382 and @xmath383 consider @xmath384 with @xmath385 and @xmath273 or @xmath375 . now , taking @xmath386 with suitable values of @xmath259 , @xmath260 and @xmath14 , we obtain that the orbits of @xmath24}(x , y):=\psi\circ \phi^{[m]}(x , y)=\psi(x , y)\ ] ] are like the ones of @xmath387 and the @xmath388 then by using the known results for @xmath343 and lemma [ lemanou ] the result follows for @xmath389 and @xmath382 . when @xmath50 the result is proved in proposition [ c6 ] . finally for @xmath390 we consider @xmath391 and @xmath392 for @xmath393 with @xmath394 and @xmath395 suitably chosen . all these @xmath29}$ ] have rank @xmath396 , as we wanted to prove . our main result is the following theorem : [ nfi ] for @xmath53 and most values of @xmath1 the map @xmath29}$ ] has no meromorphic first integral . its proof is a consequence of the following result : [ final ] for @xmath53 let @xmath397 @xmath134 be as in theorem [ propo5 ] , @xmath398 then if @xmath399 the map @xmath29}$ ] has no meromorphic first integral . for simplicity , we prove result for the case @xmath400 being the proof in the general case similar . note that the above condition reads as @xmath401 if @xmath317}=f_{e , d , c , b , a}$ ] has a meromorphic first integral , the same holds for all the other maps @xmath402 , @xmath403 , @xmath404 , and @xmath405 . these five maps have the origin @xmath320 as a fixed point and are analytic in its neighborhood ( see the proof of proposition [ lema5 ] ) . moreover the corresponding couples of eigenvalues of their linear parts at zero are @xmath406 @xmath407 ; @xmath408 ; @xmath409 and @xmath410 , respectively . hence , applying theorem [ th - int ] , we obtain the following necessary conditions for the existence of a meromorphic first integral @xmath411 for some @xmath412 , @xmath134 . from these equalities we get that @xmath413 . so the result follows . a simple corollary of the above result is : the map @xmath414 has a meromorphic first integral if and only if @xmath415 of course , all the known rationally integrable cases , like @xmath317}=f_a^5,$ ] @xmath416}=f_{ba}^5,$ ] @xmath417}=f_{c , b , a}^3 $ ] and @xmath418}=f_{{1}/{a},c , ac , a}^5 $ ] satisfy @xmath419 . our simulations show that there are examples of maps @xmath29}$ ] exhibiting ssnc when @xmath420 these behaviors can be seen by plotting some orbits of the corresponding conjugated maps @xmath46}$ ] for the @xmath2-periodic sequences of parameters @xmath421 . more visual examples can be obtained by studying the maps @xmath422 , @xmath423 , @xmath424 and @xmath425 and @xmath426 with @xmath427 see some pictures in figures 1 and 2 . figure 2 : some orbits of @xmath428 and a zoom with much more orbits . in fact we prove : [ chaos ] if there exist maps of the form @xmath429}$],@xmath430}$ ] and @xmath431}$ ] exhibiting ssnc then , for any @xmath432 @xmath433 there exist maps of the form @xmath29}$ ] with @xmath434 having also ssnc . for @xmath435 , the maps @xmath436}&:=f^m_{1,1,1,1,1}\circ f_{[11]}=f_{[11]},\\ f_{[5m+7]}&:=f^m_{1,1,1,1,1}\circ f_{[7]}=f_{[7]},\\ f_{[5m+8]}&:=f^m_{1,1,1,1,1}\circ f_{[4]}\circ f_{[4]}=f_{[4]}\circ f_{[4]},\\ f_{[5m+4]}&:=f^m_{1,1,1,1,1}\circ f_{[4]}=f_{[4]},\\ f_{[5m+20]}&:=f^m_{1,1,1,1,1}\circ f_{[4]}^5= f_{[4]}^5,\end{aligned}\ ] ] will also have ssnc . note that , since @xmath437 , @xmath438 , @xmath439 , @xmath440 and @xmath441 @xmath442 , the above maps cover all the values of @xmath2 given in the statement . since one of the features of these maps is the existence of transversal homoclinic points , which is a structurally stable property , we can perturb each of the corresponding @xmath443 by @xmath444 , with all the @xmath445 sufficiently small , to obtain @xmath2-periodic sequences of parameters having ssnc and @xmath434 , as we wanted to prove . we have ( only numerically ) shown the existence of ssnc for @xmath446 , but note that the above lemma allows to reduce all the other cases to these four ones . although the complicated behavior of the maps , as in figure 2 , leads one to believe that even there may be no upper bound for the number of intervals given by the adherence of a sequence ( formed by the sequence itself and its accumulation points ) , we only present here some simple examples . concretely , for @xmath70 we give a map @xmath29}$ ] and two sets of initial conditions such that the adherence of the sequences @xmath4 generated by consists of more than @xmath2 intervals . we have that for @xmath447 and @xmath448 , * the sequence starting at @xmath449 is formed by 20 intervals ; * the sequence starting at @xmath450 is formed by 7 intervals . for instance , this last assertion can be seen by making the phase portraits of the orbit of @xmath451 starting at @xmath452 , which is formed by 5 islands , together with their images through @xmath453 , @xmath454 and @xmath455 and their projections in the @xmath197-axis , see figure 3 . the property of the existence of sequences @xmath4 generated by such that their adherence consists of more that @xmath2 intervals should be true for all the values of @xmath2 given in lemma [ chaos ] . we also want to comment that , for these values of @xmath2 , the initial conditions lying on the stable manifolds of the @xmath456-periodic saddle points of @xmath46}$ ] , also have a curious behavior : the adherence of @xmath4 is the sequence itself together with @xmath456 more points , corresponding to the saddle points . in general @xmath456 also is greater than @xmath45 on the other hand , the most complicated orbits , that is the ones between two big invariant curves , whose adherence seems to fill a region of positive measure give rise to a single interval when we consider their projections given by the sequence @xmath457 after the first version of this paper was finished , one of the authors ( a. cima ) and s. zafar have obtained a proof of the non rational integrability of generic maps @xmath429}$ ] by computing its dynamical degree , @xcite . we thank joan carles tatjer for his suggestion to introduce the maps @xmath46}$ ] for a better approach to the numerical study of the maps @xmath29}$ ] ; john a.g . roberts who gave us very useful information about the interest of classical lyness recurrences in the context of discrete integrability . g. bastien , m. rogalski . _ results and conjectures about order @xmath456 lyness difference equation @xmath464 in @xmath465 , with a particular study of the case @xmath466 _ advances in difference equations 2009 ( 2009 ) , article i d 134749 . e. camouzis , g. ladas . dynamics of third - order rational difference equations with open problems and conjectures . advances in discrete mathematics and applications vol . 5 . chapman & hall , boca raton fl , 2008 . a. cima , a. gasull , v. maosa . _ on 2 and 3 periodic lyness difference equations _ , j. difference equ . doi : 10.1080/10236198.2010.524212 a. cima , a. gasull , v. maosa . _ non - autonomous @xmath467-periodic gumovski - mira difference equations_. to appear in int . j. bifurcations & chaos .

this paper studies non - autonomous lyness type recurrences of the form @xmath0 , where @xmath1 is a @xmath2-periodic sequence of positive numbers with primitive period @xmath2 . we show that for the cases @xmath3 the behavior of the sequence @xmath4 is simple ( integrable ) while for the remaining cases satisfying this behavior can be much more complicated ( chaotic ) . we also show that the cases where @xmath2 is a multiple of 5 present some different features . _ 2000 mathematics subject classification : _ ` 37c55 , 39a11 , 39a20 ` _ keywords : _ integrability and non - integrability of discrete systems , numerical chaos , periodic difference equations , qrt maps , rational and meromorphic first integrals .

introduction and main results preliminary results proof of the main tools rational integrability and associated dynamics meromorphic non-integrability for the case @xmath53 numerical evidences of chaos acknowledgements

for macroscopic systems , the second law of thermodynamics prescribes ever - increasing entropy . in fact , decreases of entropy are permitted on short time scales . when the dynamics of small quantum systems on short time scales is studied , the flow of information between the environment and the system can be important [ @xcite ] . in a microscopic theory , there are three major effects of the environment ( or bath ) on the system . the first two are dissipation , which removes excess energy from the system , and fluctuations , which supply energy . these two effects are related through the fluctuation dissipation theorem , which assures that the correct finite temperature equilibrium state is reached . the third one , which is less well known outside the fields of nonlinear optics and nmr , is the presence of entanglement between system and environment states . this third effect plays a major role if the system bath interaction is strong , or if the characteristic time scale of the noise induced by the environment is slow compared to typical system time scales . it is the origin of a rephasing signal in photon echo and nmr echo measurements . the dynamics of a quantum system in contact with a bath is described theoretically by deriving an equation of motion for the reduced density matrix , which includes only the system degrees of freedom . it is often assumed that the characteristic time scale of motion in the environment is much shorter than anything that happens in the system . this approximation is convenient , because it allows the derivation of a closed equation of motion for the reduced density matrix . in particular , no information about the history of the dynamics can be stored in the bath , and the equation of motion is local in time . the requirements that the trace of the density matrix must be preserved , and that the diagonal matrix elements must be positive in any basis then lead to a master equation in the lindblad form [ @xcite ] . the dissipation operators that appear in this equation can be specified by modeling the environment and its interaction with the system . because of the fast bath approximation , commonly used in association with secular approximations [ @xcite ] , the master equation approach does not properly include the fluctuations caused by the bath . on the other hand , stochastic liouville equations , which do treat the fluctuations correctly , neglect the dissipation [ @xcite ] . the restriction to an environment with much faster dynamics than the system breaks down for many systems studied in ultrafast nonlinear optics and nmr . in addition , it is not valid at low temperature , when a correct quantum mechanical description of the bath introduces additional time scales determined by the matsubara frequencies . obviously , the dynamics becomes more complex in this situation . although strong system bath interaction can be included [ @xcite ] , the key difference with the lindblad formalism is the presence of memory . an environment which is not infinitely fast ( compared to typical time scales in the system ) can store information about the past . this information can subsequently flow back into the system , influencing the dynamics . such memory effects can be included in master equations . these can , however , usually not describe a second key effect of a slower environment : the presence of correlated superpositions of the system and the environment in the initial state [ @xcite ] . because such correlations introduce a second source of memory , they can not in general be ignored . the idea that the flow of information from the environment back to the system can be used to quantify the extent of memory in a non - markovian quantum process has been developed by @xcite and @xcite . in a memoryless situation , two system states that are initially a certain distance apart , will only get closer during the time evolution . therefore , when states are found that grow farther apart , the time evolution can be called non - markovian . by introducing distance measures on the space of system states , these ideas can be made precise , resulting in a measure for non - markovianity that depends only on system degrees of freedom . once such theoretical measures have been introduced , the question arises how they can be measured in experiment . in principle , quantum state tomography yields the complete quantum state of the system [ @xcite ] . once this measurement has been performed , any quantity that is a functional of the reduced density matrix can be calculated . however , this process is rather cumbersome and indirect . more straightforward methods to quantify concepts such as entanglement and non - markovianity from experiments are welcome [ @xcite ] . experiments that can achieve this goal are found in the field of nonlinear optics . observables such as the photon echo and two - dimensional optical spectroscopy depend on multiple time intervals , and are sensitive to memory effects that extend over several of these intervals [ @xcite ] . in this paper , we show how correlations between the system and the environment are of critical importance for the non - markovianity . in particular , for a simple model environment the dynamics is completely markovian if the initial state does not include such correlations , while it becomes non - markovian if correlations are allowed to be present . we describe the generation of initial correlations during a preparation time and discuss the close connection with the nonlinear optical response . for a classical stochastic process , the meaning of `` markovian '' is clear [ @xcite ] : the future depends only on the present state , and not on the past . in the case of a gaussian process , its correlation function must be exponential to have markovianity . in the quantum case , we will define markovianity following @xcite , although other approaches have been proposed as well [ @xcite ] . as explained in the introduction , the definition is based on the distance between a pair of quantum states . a convenient measure is given by the trace distance @xmath0 between two density matrices @xmath1 and @xmath2 , which is defined as @xmath3 or half the sum of the square root of the eigenvalues of @xmath4 . the subscript s in the trace indicates that it is taken over system degrees of freedom , in contrast with trace operations over the environment which we will encounter later . in an ergodic system , any initial state will evolve in time until it reaches a single well - defined equilibrium . if there is no memory in the bath , the dynamics can only bring the system closer to equilibrium . because no information can flow from the environment to the system , the distance between a pair of initial states will decrease with time . this is the case in memory - less approaches such as the lindblad master equation . memory in the bath means that the bath stores information about the system at a previous point in time , which affects the dynamics . the extra information opens the possibility of temporary time evolution in the unnatural direction . this suggest that non - markovianity can be measured by studying how much two states move away from each other . in the definition given by @xcite , this quantity is studied by defining the change in the trace distance @xmath5 which is integrated over time to define the non - markovianity @xmath6 the maximum is taken over all combinations of initial states @xmath7 and @xmath8 . this measure is based on the distinguishability of the two trajectories . although the trace distance can be defined for two density matrices of any shape , it takes a particularly simple form for a two - level system . in this case , the matrix elements of a density matrix are written in a given basis as @xmath9 the trace distance between @xmath1 and @xmath2 is found as @xmath10 although the ( statistical ) state of the system s degrees of freedom at any point in time is completely described by the reduced density matrix , this is nevertheless not the complete story . during the time evolution , the system gets entangled with the environment [ @xcite ] . if the time scale of the environment is not very short , this will influence the state of the system at later points in time . therefore , the presence of classical or quantum mechanical correlations between system and environment affects dynamic measures like the non - markovianity , although they are not explicitly present in the reduced description . although correlations are automatically produced during the time evolution , leading to non - markovianity , they can also be present in the initial state . these initial correlations contribute to the non - markovianity as well and should be included in a proper description . to see this explicitly , we denote the complete density matrix including all system as well as bath degrees of freedom as @xmath11 . its matrix elements in the system subspace are still operators on the bath degrees of freedom . because the system and bath taken together form a normal quantum system , the complete density matrix evolves coherently in time , as dictated by the complete hamiltonian @xmath12 . we can define a propagator @xmath13 which propagates the density matrix as @xmath14 , which is given by @xmath15 . the notation @xmath16 $ ] denotes the commutator . the reduced density matrix , which operates only on the hilbert space of the system , is found by taking the partial trace over the bath , @xmath17 . for a factorized initial state , @xmath18 , the time evolution can be written as a dynamical map @xmath19 . however , in the case of a slow environment , it is not clear why initial correlations between system and bath states can be neglected and the factorization assumption may break down . this means that the complete density matrix can not be written in the form @xmath20 , where @xmath21 is a density matrix in the hilbert space of the bath . instead , each matrix element in the system space may depend on the bath in its own way . the difference between uncorrelated and fully correlated equilibrium density matrices , which are given by @xmath22 , has been studied recently by @xcite and is readily observable in the optical response [ @xcite ] . to study the effect of more general initial correlations , we introduce a preparation time . this method allows us to interpolate between an uncorrelated state and the properly correlated equilibrium . an initially uncorrelated state is allowed to evolve for a time @xmath23 , during which correlations are formed . the dynamics of the thus obtained correlated state at time zero is then followed during a time @xmath24 . for @xmath25 , correlations between system and bath are absent , while for a long enough preparation time maximum correlation is reached . the coherent time evolution of the system is given by a hamiltonian @xmath26 . we employ a commonly used model for the environment that includes the complete quantum mechanical behaviour of bath modes , yet is flexible enough to be solved to a certain degree . in this model , the bath modes are harmonic oscillators , which couple linearly to the system . the hamiltonian for the bath and its coupling to the system is given by @xmath27 here , @xmath28 indexes the bath modes , which have coordinates @xmath29 , momenta @xmath30 and masses @xmath31 . @xmath32 denotes any operator on the hilbert space of the system , which couples to the bath modes with strength @xmath33 . all necessary information about the system bath interaction is contained in the spectral density @xmath34 and the temperature @xmath35 . the correlation function can be written as the inverse fourier transform of the spectral density as @xmath36 its real part corresponds to the fluctuations , which are a function of the inverse temperature @xmath37 ( @xmath38 is the boltzmann constant ) , whereas the imaginary part is the dissipation . because the statistics for linear coupling to a harmonic bath is the same as for a gaussian process , multi - point correlations functions are redundant ( they can be evaluated using wick s theorem ) , and are not necessary for the calculation of the propagator . in the classical limit , the bath can be modeled by a stochastic process . the time evolution is then given by a stochastic liouville equation , which can include initial correlations [ @xcite ] and is suitable for the calculation of the nonlinear response [ @xcite ] . to simplify the analytical treatment , we will describe the situation where the system - bath interaction commutes with the system hamiltonian , @xmath39 = 0 $ ] , such that the exact dynamics becomes second order in the system - bath interaction [ @xcite ] . for an overdamped brownian oscillator spectral density , the non - commuting case can be handled efficiently using the hierarchy of equations of motion approach [ @xcite ] . although the system hamiltonian can be chosen freely , we will for definiteness focus on a two - level system . in the basis of its eigenstates , the system hamiltonian is diagonal , with matrix elements @xmath40 and @xmath41 . the system - bath interaction causes dephasing in the excited state , and @xmath42 has matrix elements @xmath40 and @xmath43 , where the fact that @xmath44 is an operator on the bath degrees of freedom is indicated explicitly by the notation @xmath45 . in a linear response experiment , the system is brought out of equilibrium by an external pulse , and the subsequent time evolution is probed . non - markovianity during the evolution time can occur in two ways . firstly , it can be caused by memory in the system bath interaction during the evolution time . a second source of non - markovianity are initial correlations between the system and the environment , which are present at the time the impulsive force interacts with the system . such correlations can be studied in detail using nonlinear experiments , involving multiple pulses . when a system is initially in a factorized state , the only source of non - markovianity is the buildup of system bath correlations during the time evolution . suppose that two initial density matrices are given by @xmath46 and @xmath47 . we make the usual assumption ( which we want to relax later ) that the system can be separated from the bath , and that the bath is in thermal equilibrium . the complete density matrix is then written as the direct product of a system part and a bath part , @xmath48 , where the reduced density matrix is @xmath49 . similar relations are written for @xmath47 . such factorized initial conditions are typically found in electronic resonant spectroscopy , where the thermal energy is much smaller than the electronic excitation energy . the equilibrium density matrix then only contains population in the ground state , given by @xmath50 . from this state , one can create any factorized initial state by applying an impulsive external interaction @xmath51 , where @xmath52 denotes a liouville operator . the time evolution of the complete density matrix is given by coherent evolution @xmath53 we assume that the system hamiltonian commutes with the system - bath interaction . in the interaction picture with respect to the bath hamiltonian , the time evolution of the reduced density matrix then becomes @xmath54 where @xmath55 and @xmath56 denotes the time ordered exponential . the evaluation of the average over the bath is a standard calculation , giving @xmath57 which for the coherences reduces to @xmath58 , while the populations are constant in time . the dephasing function is given by @xmath59 with the correlation function @xmath60 given by fourier transforms of the spectral density according to equation [ cf ] [ @xcite ] . for a two - level system , the trace distance between two density matrices which are initially prepared as @xmath7 and @xmath8 can be readily evaluated using equation [ trd2ls ] . it is found to be @xmath61 because we study the pure dephasing case , the populations @xmath62 and @xmath63 are constant in time , while the coherences evolve according to the dephasing function @xmath64 . while the dephasing function contains an imaginary ( dissipative ) part , which causes a time - dependent shift in the effective frequency , only the real part appears in the trace distance . if two state are prepared at time zero with initial populations , the trace distance simplifies to @xmath65 in this case , the trace distance is directly related to the dephasing function . it is important to notice that the non - markovianity only depends on the fluctuation part of the bath contribution , represented by the real part of @xmath64 . thus , one can not reveal this effect from lindblad - like quantum master equations , which only includes the dissipative part of bath contribution properly . however , stochastic liouville equations may be useful for the study of non - markovianity . we are now in a position to analyse the conditions for which the dynamics is non - markovian . according to the definition of the non - markovianity , the dynamics is non - markovian only if the trace distance between two density matrices increases with time . the time derivative of the trace distance is found to be @xmath66 . because the exponential of a real number is always positive , the time derivative can be positive only if @xmath67 . from the definition in equation [ gt ] , the time derivative is @xmath68 . we see that the trace distance can only increase if the real part of the correlation function is negative , and sufficiently negative . although the relation between the trace distance and the dephasing function is more complex in the general case where @xmath69 , the populations do not influence the question whether the dynamics is markovian . non - markovian time evolution is found if the trace distance increases at a certain point in time . because the trace distance is a positive quantity , its derivative is given by a positive constant times @xmath70 also in the case of different populations . the previous analysis therefore applies , even though the value of the non - markovianity will be different . however , because we started from initial states where the system and bath are factorized , this treatment does not include initial correlations between the system and the bath . to study their effect , we next study the non - markovianity after an initial preparation time . to include initial correlations , we consider a preparation time . starting from an state that factorizes into system and bath parts , which can be created as @xmath71 in optical experiments , the sample evolves during a time @xmath23 . during this time correlations between the system and the bath form . an impulsive external force @xmath72 is then applied to the system , after which time evolution takes place during an interval @xmath24 . the time variables are illustrated in figure [ fig : schema ] . the non - markovianity during the time @xmath24 can now be caused by two effects : correlations that build up during @xmath24 , as well as initial correlations present at the moment the external force interacts with the system , which are the result of the preparation . the density matrix after evolution during two times is given by @xmath73 , where we assume factorized conditions at time @xmath74 . @xmath72 denotes a liouville operator that models the second impulsive external force . the matrix product in the system liouville space can be worked out explicitly by choosing a basis . we order the basis states as @xmath75 , @xmath76 , @xmath77 , @xmath78 and denote the matrix elements of @xmath72 in this basis as @xmath79 . assuming that @xmath80 = 0 $ ] , we find @xmath81 where @xmath82 are still operators on the bath degrees of freedom . the trace over the bath degrees of freedom can now be calculated analytically , using cumulant expansion or path integral methods . it results in dephasing functions , which depend only on a single time when the average over either @xmath83 or @xmath84 is taken , but explicitly on both times for the average of products of two @xmath85 functions . because we are interested in the effect of initial correlations , these terms , which can not be factorized into separate contributions depending on @xmath23 and @xmath24 only , are the most relevant to our treatment . they contain the effect of memory that extends over the externally applied force . from equation [ prop ] , we see that these interesting terms multiply the matrix elements of the external force that operate on the coherences @xmath86 and @xmath87 . there are four such terms , two which leave the coherence unchanged , and two which interchange the two coherences . to focus clearly on the effect of initial correlations , we choose an operation that flips the coherence , while leaving the populations unaffected . such a force is given by a liouville operator with matrix elements @xmath88 , and all other elements zero . writing out the matrix elements in equation [ prop ] , the density matrix is then given by @xmath89 starting from two density matrices @xmath7 and @xmath8 , with equal initial populations , the trace distance between them evolves in time as @xmath90 . as in the previous case of a single time interval , the restriction to equal populations only changes the value of the trace distance , but not the question whether the dynamics is markovian . using the cumulant expansion , we find @xmath91.\ ] ] this expression enables the straightforward evaluation of the trace distance and the non - markovianity for any spectral density . the term @xmath92 indicates the effect of initial correlations present at the time of interaction with the impulsive force @xmath72 . such correlations , which extend across the the excitation , can not be treated by the conventional reduced equation of motion approach , which includes redfield and lindblad equations . this has been pointed out in the calculations of nonlinear optical observables by @xcite . as a simple example , which allows a more detailed analytical treatment , we will discuss the case of an overdamped brownian oscillator . the spectral density is given by @xmath93 , which gives the dephasing function [ @xcite ] @xmath94 for this model , the imaginary ( dissipative ) part of the correlation and dephasing functions depends only on the single time scale @xmath95 . as can be seen from the first line of equation [ gtbo ] , the real ( fluctuation ) part includes additional time scales dictated by the matsubara frequencies @xmath96 . while the imaginary part of @xmath64 becomes constant due to the ohmic nature of @xmath97 for @xmath98 , the real part is time dependent as long as @xmath99 is small . this indicates that the fluctuation part of the bath noise can not be delta correlated , even if this approximation is valid for the dissipation part . at high temperature _ compared to the time scale of the bath _ , @xmath100 , these quantum fluctuation terms can be dropped , and the dephasing function simplifies @xmath101 in the high temperature case given above , the time derivative of the dephasing function is clearly always positive , and , consequently , the non - markovianity vanishes for a single time interval . however , if we account for initial correlations by allowing them to form during a preparation time , the dynamics can become non - markovian . this can be seen explicitly using the overdamped brownian oscillator spectral density . for @xmath25 , the trace distance varies with time as @xmath102 , and we recover the result found earlier . because @xmath103 for all times , the trace distance is strictly decreasing . thus , the dynamics is markovian for an exponential correlation function , in agreement with the classical definition . on the other hand , if we allow system bath correlations to form during the preparation time @xmath23 , the trace distance can increase , and the measure for non - markovianity is nonzero . this effect is shown in figure [ fig : trd ] , where we compare a factorized initial state , corresponding to @xmath25 , to a state that contains correlations , created by setting @xmath104 . for a nonzero preparation time , the trace distance increases during a certain time interval , which shows that the dynamics is non - markovian . it is clear that memory in the bath that extends over the pulse is crucial for this effect . the time evolution of the system after the application of the pulse is influenced by its state before the pulse , as can be seen from the presence of the @xmath92 term . for a bath that contains memory , this term does not factorize into functions of @xmath23 and @xmath24 only . the memory effect can be seen clearly by looking at the trace distance as a function of both times , as plotted in figure [ fig : trd2 t ] . the time evolution of a quantum system during two intervals , separated by an external impulsive force , is closely related to nonlinear optical experiments . in these experiments , an initial pulse excites the system out of the ground state . correlations between the system and the bath form during the following propagation time . after applying another pulse , the effect of these correlations can be observed . as mentioned before , the initial state @xmath105 can be prepared by applying a pulse on the equilibrium state , @xmath106 . in the case of optical experiments , the excitation energy is typically much larger than the thermal energy , and the equilibrium distribution only contains population in the ground state . the optical field couples to the dipole of the system , described by the dipole operator @xmath107 . if we choose the operator @xmath52 as the commutator of the dipole with the density matrix , @xmath108 $ ] , and furthermore set @xmath109 , the time evolution operators evaluated in the previous section correspond to the standard third - order nonlinear response functions with zero population time [ @xcite ] . the observable in nonlinear optical experiments is the trace of the dipole operator multiplied with the density matrix , @xmath110 . by choosing the wave vectors of the incident pulses properly , it is possible to select pathways that are sensitive directly to the coherence flip described in the previous section . the resulting photon echo signal is given by [ @xcite ] @xmath111.\ ] ] using heterodyne detection , both the real and imaginary parts of this response function are observable , while homodyne detection directly yields the absolute value . the connection with the non - markovianity in the previous section is immediately clear : the photon echo is sensitive to exactly the memory effects that are responsible for non - markovian dynamics . a photon echo experiment can be used to prepare a state in which the system and environment are correlated , and to subsequently probe the time evolution . plotting the thus obtained response function directly answers the question whether the dynamics is markovian or not , according to the definition given by @xcite . we have studied the non - markovianity in quantum mechanical time evolution . this concept of markovianity can be made precise by looking for states that become more distinguishable during time evolution . if such states are present , the process is clearly non - markovian , which is the basic idea of the measure for non - markovianity proposed by @xcite . non - markovian time evolution corresponds to the presence of memory effects . only the ( temperature dependent ) bath induced fluctuations , and not the dissipation , enter the non - markovianity . we have treated the dynamics without making markovian or rotating wave approximations , and thereby fully included the correlations between system and bath states , which influence the dynamics at a later point in time . not only the correlations that are formed during the evolution , but also those present in the initial state can cause memory effects . by forming a correlated initial state during a preparation time , this effect can be studied for more general initial states than the equilibrium with respect to the complete hamiltonian . we have shown that a process that is markovian without initial correlations can become non - markovian when such correlations are present . conventional master equations , which can not include the preservation of system bath entanglement across a pulse can not be used to analyse this situation . clearly , commonly used approximations such as a delta correlated bath or secular system - bath interaction do nt hold either . because the procedure of preparing correlations during an initial time , and subsequently measuring their effect following an external impulse , the non - markovianity is directly observable in nonlinear optical experiments such as the photon echo . future work should consider the three time photon echo , and the closely related two - dimensional optical spectra . in these experiments , population dynamics can be studied during a waiting time , allowing for more general measures of non - markovian time evolution . generalizations of the current work to more general system hamiltonians , multiple baths , the case where the system hamiltonian and the system - bath interaction do not commute and low temperature are possible using the hierarchy of equations of motion [ @xcite ] . agd acknowledges the japan society for the promotion of science for support in the form of a postdoctoral fellowship for foreign researchers . y.t . was supported by a grant - in - aid for scientific research b2235006 from the japan society for the promotion of science . engel , g. s. , calhoun , t. r. , l. , r. e. , ahn , t .- k . , mancal , t. , cheng , y .- c . , blankenship , r. e. & r. , f. g. 2007 evidence for wavelike energy transfer through quantum coherence in photosynthetic systems . _ nature _ , * 446 * , 782 . kuah , a .- m . , modi , k. , rodrguez - rosario , c. a. & sudarshan , e. c. g. 2007 how state preparation can affect a quantum experiment : quantum process tomography for open systems . a _ , * 76 * , 042113 . ( dashed line ) and @xmath112 ( solid line ) . the environment is modeled as an overdamped brownian oscillator at high temperature with parameters @xmath113 and @xmath114 . the dotted lines show the same quantities calculated with 100 low temperature correction terms . increase of the trace distance with time shows non - markovian evolution . [ fig : trd],width=302 ] and a detection time @xmath24 . parameters for the environment are the same as in figure [ fig : trd ] . correlations between the time evolution during @xmath23 and @xmath24 are clearly present . [ fig : trd2t],width=302 ]

non - markovianity , initial correlations , nonlinear optics by extending the response function approach developed in nonlinear optics , we analytically derive an expression for the non - markovianity [ laine , et al , phys . rev . a 81 , 062115 ( 2010 ) ] in the time evolution of a system in contact with a quantum mechanical bath , and find a close connection with the directly observable nonlinear optical response . the result indicates that memory in the bath induced fluctuations rather than in the dissipation causes non - markovianity . initial correlations between states of the system and the bath are shown to be essential for a correct understanding of the non - markovianity . [ firstpage ]

introduction trace distance and non-markovianity model results conclusion

electric fields exert forces on charge - stabilized colloidal particles both directly through their coupling to the particles charges and also indirectly through their influence on the surrounding electrolyte . neighboring particles also interact electrostatically and hydrodynamically with each other as they move . the resulting cooperative motions in many - body suspensions can be quite complex , and our understanding of such electrokinetic phenomena remains incomplete despite more than a century of study @xcite . the tendency of colloidal particles to form oriented chains in oscillatory ( ac ) electric fields was noted nearly a century ago @xcite and was correctly ascribed to field - induced dipolar interactions . more recently , ac fields also have been found to organize colloidal spheres into circulating chevron bands oblique to the field @xcite , glassy bands perpendicular to the field @xcite , and even highly ordered colloidal crystals @xcite . still other patterns form when electric fields are used to drive charged colloid against electrodes surfaces , with two - dimensional fractal aggregates appeared in monodisperse suspensions @xcite and planar superlattices forming in binary colloids @xcite . more recently , a colloidal model system has been introduced whose phase diagram can be tuned with salt concentration and ac electric field @xcite . even constant ( dc ) electric fields give rise to complex dynamics . for example , charged spheres driven onto electrodes surfaces by dc fields self - organize into epitaxial colloidal crystals @xcite . dc fields also have been observed to induce complex labyrinthine patterns in nonaqueous nanocolloids @xcite . most of these cooperative electrokinetic effects have been explained on the basis of electrohydrodynamic coupling among the spheres mediated by field - induced fluxes of ions . others , such as the electrokinetic compression of colloidal crystals @xcite have so far resisted explanation . in this paper , we report the remarkably diverse set of patterns that charged colloidal particles can form as sedimenting under gravity in a vertical dc electric field . preliminary results have been published in ref . @xcite . beautiful , highly organized dynamic patterns with spatial periods between 20 and 200 micrometers form at biases near the threshold for electrolysis . these are supplanted by macroscopic ( spatial period @xmath0 ) patterns at higher biases . we describe how such patterns form through the interplay of electrohydrodynamic coupling driven by reaction - diffusion of ions and the uniform body force provided by gravity . we further show that the microscopic patterns result from many - particle interactions rather than an underlying convective instability of the electrolyte . the experimental apparatus is shown schematically in fig . [ fig : schematic ] . our samples consist of aqueous dispersions of silica spheres 3.0 @xmath1 in diameter ( bangs laboratories , lot number 4181 ) confined to slit pore between a glass microscope slide and a coverslip . the inner glass surfaces were coated with 10 thick gold electrodes on 10 thick titanium or chromium wetting layers before assembly . while still optically thin , the electrodes have a resistivity of less than 50 ohms per square and allow us to apply uniform vertical electric fields to the confined suspension . conductive epoxy is used to make contact to the electrodes once the cell is assembled . in most cases , glass spacers were used to set the electrode separation , @xmath2 , in the range @xmath3 across the @xmath4 observation area . glass access tubes bonded to holes drilled through the slides were left open to air so that the suspension equilibrates to a ph of roughly 5.5 through the dissolution of carbon dioxide . under these conditions , the silica spheres acquire a surface charge density of roughly @xmath5 @xcite . as the density of colloidal silica is roughly @xmath6 , the 3.0 @xmath1diameter spheres sediment rapidly onto the the lower electrode . consequently , we use the areal coverage @xmath7 to report the concentration of particles , with @xmath8 corresponding to a close - packed monolayer and @xmath9 corresponding to a bilayer . colloidal patterns are imaged with ccd ( charge - coupled device ) camera through an inverted optical microscope . the images and videos reported in this paper were obtained looking upward through the lower electrode , in the direction opposite to gravity . we define the applied dc voltage to be positive if the upper electrode is positive so that the electric field is directed downward , but the electrostatic forces on negatively charged particles point upward . in the inter - electrode gap @xmath2 . the arrows indicate the vortex s observed circulation direction . ] the electrochemical processes in a parallel - plate water electrolysis cell are described in most standard references , such as refs . @xcite and @xcite , and are summarized in fig . [ fig : schematic ] . decomposing a water molecule requires an input of 1.23 , independent of the salt concentration or the ph value of the solution . in practice , however , the redox potential is at least 0.6 higher due to a voltage drop across the electrode - electrolyte interface known as the overvoltage ( overpotential ) . once the threshold for hydrolysis is reached , the electrodes rapidly become coated with insulating layers of @xmath10and @xmath11 . these barriers soon stop the electrode reactions unless the evolved gases form bubbles that escape the electrodes . a substantial nucleation barrier impedes the evolution of @xmath10and @xmath11bubbles large enough to expand and escape the electrodes . consequently , a bias as large as 2.5 to 3.0 typically is required to maintain hydrolysis in quasi - steady - state . the range of threshold biases is due to inevitable variations in the surface condition of the electrodes . our sample cells vacuum - deposited thin film gold electrodes yield threshold biases between 2.6 and 3.2 . visible bubbles begin to appear between a minute and an hour after the bias is applied , depending on the voltage . the onset of bubble evolution sets our observation time , and decreases with increasing voltage or salt concentration . although the redox potential for gold in water is only 1.5 , the overpotential prevents the anode from etching , so that a negligible concentration of gold ions , @xmath12 , is released into solution . indeed , no visible or electrochemical change in the electrodes is seen even after several hours operation at a steady bias . even in the absence of added salt , however , the electrolyte in our samples has total ionic concentrations of roughly @xmath13 and @xmath14 . a 0.1 @xmath1thick debye - hckel screening layer forms at each electrode at these concentrations , beyond which the field would not penetrate at equilibrium . during hydrolysis , however , the electrodes act like sources and sinks of @xmath15and @xmath16ions whose currents establish both an electric field and a ph gradient through the cell , with the lower half volume becoming acidic and the upper half basic under positive bias . as a function of applied voltage in samples of deionized water in three cells of thickness @xmath17 , 200 and 370 @xmath1 . a single cycle of increasing and decreasing bias is shown for each sample . values were recorded after 130 sec relaxation at each voltage . the intersection of the extrapolated linear trends yields the threshold for hydrolysis . ] to quantitatively characterize hydrolysis in our cell , we measured the current density @xmath18 under the simplest conditions : the parallel - plate cell filled with deionized water without colloid . we performed sweeping voltammetry by increasing the applied bias in discrete steps of 0.2 from 0 to 3.6 and then back down to 0 , waiting between 2 to 5 at each voltage level . figure [ fig : vi3 + - ] shows the voltage dependence of the steady - state current in cells of different inter - electrode separations , @xmath2 . each curve consists of two linear parts , one at low bias in which little current flows , and another ohmic regime at higher biases . their intersection yields the threshold for water decomposition @xcite . below the decomposition voltage , the small but nonvanishing current is due to the continuous dissolution of @xmath10and @xmath11molecules . this is known as the residual current and also increases with the applied voltage . the current was measured after 130 relaxation at each step in an increasing sequence of applied biases . ] the higher conductivity of the high - bias regime is fed by the additional ions from the decomposition of water . adding nacl increases the conductivity but does not affect the threshold , as shown in fig . [ fig : vi200salt ] . for this reason , nacl is referred to as indifferent electrolyte . from the slopes of the data in fig . [ fig : vi3 + - ] , we estimate the equilibrium conductivity of our electrolyte to be @xmath19 . adding colloidal spheres to the deionized water increases the ionic concentration and thus the conductivity . adding roughly @xmath8 of silica or polystyrene sulfate ( ps ) colloid increases the conductivity to @xmath20 and @xmath21 , respectively . from these values , we estimated the ph for these systems to be 5.0 , 4.7 and 4.2 , respectively , with the negatively charged colloidal spheres acting as weak acids . the field - driven fluxes of ions and charged particles are coupled to flows in the surrounding electrolyte and can give rise to electrohydrodynamic instabilities . the most commonly observed manifestation is the formation of colloidal crystals on the electrodes @xcite . charged colloidal spheres are driven toward the oppositely charged electrode , and might be expected to diffuse independently on the surface . instead , the particles rapidly coalesce into close - packed epitaxial crystals @xcite . the strong and long - ranged interparticle attraction responsible for this crystallization results from an attractive hydrodynamic coupling induced by the spheres distortion of the local fluxes of simple ions . such electrohydrodynamically bound crystals also can be induced by an ac field @xcite . at high enough frequencies , however , the dielectric spheres induced dipole moments mediate an in - plane repulsion strong enough to and the interfacial crystals dissociate destabilize the interfacial crystals @xcite . field - induced surface crystals have been observed in silica , ps , gold and even neutral ps colloidal spheres with a wide range of diameters , and on both indium tin oxide ( ito ) and gold electrodes . the threshold voltage for crystallization can be very low if the spheres diffusion is slow and if their inherent electrostatic and dipolar repulsions are relatively weak . in our system , we have observed that the threshold voltage for field - induced crystallization depends on the particles mobility and charge as shown in fig . [ fig : crystalv ] , but does not depend on the thickness of the cell and can be lower than the 1.23 redox potential of water . crystals were formed on the lower electrode with a negative dc bias . larger spheres crystallize at lower voltages both because they are less diffusive and also because the larger distortions they cause in the surrounding ionic fluxes give rise to stronger attractions . nominally neutral hydroxyl - terminated polystyrene spheres reproducibly crystallize at the lowest voltages of the particles we studied , presumably because of their substantially weaker electrostatic repulsions . in this section , we examine some more complex phenomena that arise under a more restricted set of conditions , and which provide new insights into the interplay of electrohydrodynamic and other forces in macroionic systems . applying a positive bias in our cell tends to drive negatively charged silica spheres upward into the bulk electrolyte . the field is uniform in our parallel plate geometry , so that the upward electrophoretic force on spheres of constant charge therefore should be independent of height @xmath22 in the cell . under such conditions , spheres dense enough to sediment under gravity should remain near the lower electrode at low biases , and should rise directly to the upper electrode at higher biases where electrohydrodynamics forces exceed gravity . this effect should be accentuated in charge - regulating silica spheres because the magnitude of their surface charge increases with ph , and therefore should increase with @xmath22 . indeed , individual isolated silica spheres exhibit hysteresis in their promotion to the upper electrode with increasing bias , and relaxation to the lower electrode with decreasing bias , consistent with these considerations . more concentrated dispersions , on the other hand , exhibit substantially more interesting collective behavior . at biases between 2.2 and 4 , which only slightly exceed the threshold for hydrolysis , the background electrolyte is stable against convective instabilities . nevertheless , initially quiescent collections of spheres develop several classes of striking many - body convective patterns that highlight a previously unsuspected role for cooperativity in charge - stabilized dispersions subjected to electric fields . unlike the electrohydrodynamically stabilized surface crystals described in previous reports , these quasi - steady - state patterns form in the bulk of the electrolyte and reflect the interplay of electrohydrodynamic forces and gravity . diameter silica spheres at @xmath23 . ( a ) @xmath24 : equilibrium monolayer , @xmath25 . ( b ) @xmath26 : diffuse tumbling clouds . ( c ) @xmath27 : circulating flower - like cluster levitated to @xmath28 . ( d ) @xmath29 : compact circulating cluster , also at @xmath28 . ] figure [ fig : vortexring ] shows the typical sequence of dynamic patterns that forms when an upward bias is applied to dispersions of 3.0 @xmath1diameter silica spheres in water . positively biasing the upper electrode by as much as @xmath30 causes no discernible out - of - plane motion . abruptly applying a larger bias in the range @xmath31 causes a transient in which spheres jump off the lower surface and then settle back to the bottom . this transient reflects the establishment of static ionic gradients within the water that screen out the field . the spheres collective behavior passes through distinct regimes as they are driven out of equilibrium . just above threshold , the monolayer of spheres breaks into diffuse billowing clouds occupying the lower half of the sample cell , as shown in fig . [ fig : vortexring](b ) . pushing the system further from equilibrium might be expected to yield increasingly chaotic behavior ; but quite the opposite occurs . increasing the bias beyond 2.6 v coalesces the itinerant clouds into extraordinary flower - like clusters such as the example in fig . [ fig : vortexring](c ) , all floating with their midplanes at @xmath28 above the lower wall . each cluster forms around a rapidly circulating toroid whose spheres travel downward along the inner surface and return upward along the outside , completing one cycle in a few seconds . the sense of this circulation is indicated schematically in fig . [ fig : schematic ] . because these clusters structure and motion are reminiscent of hydrodynamic vortex rings , we will refer to them as colloidal vortex rings . colloidal vortex rings formed at higher biases tend to consist of fewer clusters , each containing a larger number of spheres at higher density . the dense yet vigorously circulating cluster in fig . [ fig : vortexring](d ) was observed at @xmath29 . increasing the bias once clusters have already formed causes each to shrink and its circulation to accelerate . such over - driven clusters merge with their neighbors until they achieve the size appropriate to the final bias . the clusters height , @xmath22 , does not change appreciably with voltage . by around @xmath32 , some clusters density increases enough that they become jammed and stop circulating . these compactified clusters still dissociate immediately once the bias is removed . at higher voltages , the electrolyte itself becomes unstable against electroconvection . the resulting patterns span the electrochemical cell , with the colloid acting principally as passive tracers . we will discuss this distinct range of conditions elsewhere . most often , a densely packed colloidal vortex ring is surrounded by a diffuse circulating corona that extends outward for tens of micrometers . [ fig : rings ] shows alternate forms consisting of compact circulating toroidal clusters without coronas . such bare colloidal vortex rings form most often at slightly lower voltages , 2.6 versus 2.8 , and can coexist with fully dressed colloidal vortex rings . these clusters also are levitated to @xmath33 in a system of thickness @xmath23 . diameter silica spheres in @xmath34 cell . ( a ) compact toroidal cluster enclosing a domain of two - dimensional colloid crystal at @xmath35 . the toroid completes one cycle of circulation in about 1.5 . ( b ) compact toroidal cluster with empty core . ] although colloidal vortex rings somewhat resemble conventional laminar vortex rings @xcite , they are driven by quite different mechanisms @xcite and also have other distinctive features . perhaps their most remarkable characteristic is the presence of close - packed colloidal crystals at their centers . the example in fig . [ fig : rings](a ) is particularly noteworthy because much of its core consists of a static crystalline _ monolayer_. these close - packed spheres are not flocculated , and disperse immediately once the driving field is turned off . consequently , the crystals formation and stability suggests the existence of a stagnation plane along each cluster s midplane , quite unlike the streaming flow within a purely hydrodynamic vortex ring . the stagnation plane must end at the edge of the circulating toroid , and indeed particles from the core sometimes are swept into the circulating flow . while the crystalline domains tend to be stable for the duration of an experimental run ( up to ten minutes at @xmath36 ) , some are incorporated into the surrounding toroid after a few minutes , leaving hollow circulating rings such as that in fig . [ fig : rings](b ) . in most cases , the crystalline monolayer core evolves from a layered fluid - like state as shown in fig . [ fig : shrinking ] . as the ring s diameter decreases with increasing bias , the concentration of spheres in the core increases to the freezing point , and a static crystal results . diameter silica spheres in a layer of thickness @xmath23 . the sequence of photographs show a cluster shrinking as the bias increases from 2.4 to 4.0 . an ac component ( 0.7 square wave at 10 ) is applied to further stabilize the crystal . the circulation period of the toroid is about 1.5 . ( a ) : 0 . ( b ) : 0.5 . ( c ) : 1.0 . ( d ) : 1.5 . ( e ) : 2.0 . ( f ) : 3.0 . ] individual colloidal vortex rings sometimes develop breathing - mode instabilities with periods of a few seconds , with one cycle of a typical structural oscillation appearing in fig . [ fig : oscillate ] . the usual circulation continues throughout the oscillation . remarkably , the cluster s crystalline core is destroyed and reformed with each cycle . neighboring clusters oscillations do not become phase locked , nor do they necessarily pulsate at the same frequency . indeed , oscillating clusters can coexist with steadily circulating or compactified clusters . silica spheres in @xmath23 cell at 2.8 constant bias . the cluster s core remains crystalline throughout the cycle . ] within the range of biases for which flower - like colloidal vortex rings reproducibly form ( about 2.6 to 3.4 ) , we also observe a variety of variant vortex ring structures that either coexist with or even replace the morphologies represented in figs . [ fig : vortexring ] and [ fig : rings ] . some of these are shown in fig . [ fig : vortexvariants ] . diameter silica spheres in @xmath34 cells . ( a ) 3.0 : colloidal vortex rings with sharply defined coronas . ( b ) 2.6 : sharply bounded vortex ring without central core . ( c ) 2.6 : central crystal stabilized by a 0.1 square wave at 1.0 develops scalloped boundary . ( d ) 3.0 : rare variant in which the central core also is convective , @xmath37 . ] figure [ fig : vortexvariants](a ) shows colloidal vortex rings whose coronas have sharply defined boundaries , rather than the diffuse coronas evident in fig . [ fig : vortexring](c ) . these clusters also sometimes form without a crystalline core as can be seen in fig . [ fig : vortexvariants](b ) , and the two types can coexist . unlike the diffusely bounded clusters that undergo breathing mode instabilities the corona in the diffuse morphology can undulate like a jellyfish . figure [ fig : vortexvariants](c ) shows an extreme example of this circumferential instability . in all of these cases , the vigorously circulating vortex ring appears to be virtually decoupled from the central crystalline core . figure [ fig : vortexvariants](d ) shows a relatively rare case in which some of the particles in the central core are swept into the vortex ring s circulation . in roughly half of our experiments , we observed vortex rings coexisting with two - dimensional surface crystals @xcite that had nucleated on the upper electrode . the spheres in these crystals are driven to the upper electrode by the transient ionic current burst when the bias is initially applied . particles that reach the upper electrode during these transients tend to stay there and subsequently form crystalline clusters through mechanisms that have been previously reported @xcite . indeed , such surface crystals begin to form at biases below the onset for vortex ring formation . above 2.6 v , clusters hovering in the bulk coexist with crystals trapped on the upper surface . decreasing the bias to 1.4 v , which is the threshold for forming crystals on the upper electrode , causes all particles in crystals to sink to the bottom . not all colloidal vortices adopt a ring - like configuration , as the examples in fig . [ fig : wormlike ] show . lower concentration suspensions tend to favor elongated vortex rings , such as those in fig . [ fig : wormlike](a ) , that can extend into linear worm - like structures such as those in fig . [ fig : wormlike](b ) . these clusters continue to circulate , but as linear rolls rather than vortex rings . as a result , worm - like colloidal vortices interact more strongly that colloidal vortex rings , and can organize themselves into large - scale labyrinthine patterns , as in fig . [ fig : wormlike](c ) . as for the other patterns we have described , worm - like colloidal vortices are levitated to roughly @xmath38 in an @xmath23 cell . diameter silica spheres in @xmath39 cell at 2.6 . ( b ) the same sample , after increasing the bias to 2.8 . ( c ) labyrinths of 3.0 @xmath1diameter silica spheres in @xmath23 cell at 2.6 . ( d ) the same sample , after increasing the bias to 3 . ] although the patterns described in the previous section form reproducibly , they occasionally are replaced by qualitatively distinct labyrinthine and island - like patterns such as those shown in fig . [ fig : interfacial ] . unlike the vigorously circulating free - floating colloidal vortices , these slow - moving clusters form near the bottom electrode under the same conditions . these patterns also are distinct from the previously reported interfacial crystals , which form on the upper electrode under positive bias . once sedimented interfacial clusters nucleate , the choice between forming islands , as in fig . [ fig : interfacial](a ) and [ fig : interfacial](b ) , or labyrinths , such as fig . [ fig : interfacial](c ) and [ fig : interfacial](d ) , depends principally on the concentration of particles . two - dimensional islands form at surface coverages of @xmath40 , and two - dimensional labyrinths at @xmath41 . three - dimensional labyrinths form at higher concentrations . islands and labyrinths can coexist at intermediate sphere concentrations . the spheres in islands and labyrinths usually flow fluidly as the patterns form , in a manner that resembles the circulation in bulk colloidal vortices . unlike bulk colloidal vortices , whose circulation retains its sense over time , labyrinthine domains sometimes flip circulation direction as they coarsen . unlike rayleigh - bnard convection or bulk electroconvection , moreover , neighboring domains circulation directions generally are not correlated . larger labyrinthine domains , however , develop double - roll structures , with particles rising along the edges and sinking along the center line . after a few minutes of circulation , islands and labyrinths often become jammed . increasing the bias can fluidize them again , and coarsen their features . this can be seen in the transitions from fig . [ fig : interfacial](a ) to fig . [ fig : interfacial](b ) and from fig . [ fig : interfacial](c ) to fig . [ fig : interfacial](d ) . spheres in an @xmath23 cell . ( a ) a typical vortex ring at 3.4 . ( b ) the same cluster at 2.8 . ] increasing the bias still further causes coarsened blobs such as those in fig . [ fig : interfacial](d ) to circulate faster and develop into interfacial toroidal vortex rings , an example of which appears in fig . [ fig : smokering](a ) . these surface - hugging clusters differ from the free - floating vortex rings of the previous section in that they respond irreversibly to changes in bias . reducing the bias on circulating blobs causes them to open out into structures resembling smoke rings , fig . [ fig : smokering](b ) . this morphology can be formed only by first increasing and then decreasing the bias , and clearly inherits its topology from the progenitor toroidal cluster . the islands and labyrinths we have observed somewhat resemble the interfacial crystals described in refs . @xcite and @xcite , even though the electric field is reversed . we conjecture that gravity maintains the particles on the lower electrode , while the in - plane electrohydrodynamic attraction discovered in refs . @xcite and @xcite draws them together into islands ( fig . [ fig : interfacial](a ) ) and labyrinths ( fig . [ fig : interfacial](c ) ) . once formed , such close - packed clusters would have substantially reduced hydrodynamic drag coefficients @xcite , and so would remain sedimented at biases that would levitate a single sphere . applying the bias rapidly appears to favor the formation of levitated clusters , presumably by disrupting this stabilizing mechanism , while slowly increasing the bias improves the chances to form interfacial clusters . we have observed free - floating colloidal vortex rings coexisting with interfacial labyrinths and islands . this demonstrates that the two types of structures can be generated under the same conditions , and therefore are not distinguished by variations in the experimental procedure . silica spheres after a 4 bias is removed . ( a ) 0 : particles sedimented onto the lower electrode retain the basic profile of their cluster s toroidal geometry . left in this state , these particles would require several minutes to relax to their normal diffusive state . ( b ) applying a positive bias of 0.8 levitates the particles slightly and allows them to relax to a uniform density . ( c ) returning to 0 eliminates any diffusive motion and causes the spheres to form sedimented crystalline domains . ( d ) after 3 minutes , the spheres begin to diffuse and relax to a homogeneously random state . ] turning off the driving bias after a pattern has formed frees the spheres to revert to their equilibrium field - free behavior . in most cases , particles sediment into a two - dimensional brownian fluid on the lower electrode . on other occasions , and particularly after prolonged runs at the upper range of biases , the sedimented particles no longer diffuse freely on the lower electrode , but rather sit motionlessly where they land . this is the case for the particles in fig . [ fig : m4](a ) , which had been organized into a colloidal vortex ring before the field was switched off . these particles reached the lower electrode rapidly enough that their toroidal configuration is still clearly visible , and remained unchanged for several minutes . after this period , however , the particles began to diffuse normally , suggesting that they had not deposited onto the lower electrode . this conclusion is supported by the observation that a small positive bias suffices to levitate the motionless spheres , as shown in fig . [ fig : m4](b ) . once the positive bias removed , however , the particles sediment again to the lower electrode and form small crystals as they settle , a typical result appearing in fig . [ fig : m4](c ) . the spheres in these self - organized crystals still display no brownian motion for several minutes , after which the revert to their equilibrium behavior , as shown in fig . [ fig : m4](d ) . this bizarre transient behavior appears not to have been described before , and its origin is unknown . similar patterns and trends emerge if the bias voltage is applied abruptly or increased gradually enough that diffuse tumbling clouds form before coalescing into stable clusters . the typical number of particles in a steady - state cluster tends to increase with the bias . slightly changing the bias once clusters have formed usually does not affect this number because merging and splitting clusters to achieve the new optimal cluster size imposes a substantial kinetic barrier . large bias changes , however , can break up all the clusters and reorganize them into clusters with the appropriate number of particles . for example , groups of three clusters can merge then split into cluster pairs with an increase in bias . reducing the applied bias reverses the sequence of transitions , without obvious hysteresis . the inter - electrode spacing similarly also not to affect electrokinetic pattern formation over quite a large range , with similar patterns appearing in sample cells at @xmath17 , 200 , and 390 @xmath1 , all with similar crossover voltages . bulk clusters are always observed to float at about @xmath42 , as shown in fig . [ fig : height ] . interfacial patterns form at a fraction of this height . patterns do not form in thinner cells , however . for example , we observed that particles in a cell @xmath43 thick were pushed directly to the upper electrode and formed no bulk clusters . adding salt slows the particles motions and favors the formation of tumbling clouds instead of quasi - stationary clusters . patterns are blurred in 0.1 nacl solution , and disappear altogether in 0.5 solution . for example , 3.0 @xmath1silica spheres dispersed in 0.2 nacl solution failed to form any quasi - stationary structures , but rather entered an irregular cycle of collapsing into concentrated domains and then bursting like fireworks as shown in fig . [ fig : firework ] . silica spheres in 0.2 mm nacl solution . @xmath34 . 4 v bias . particles repeated clump into dense clusters then burst out like fireworks . ] unlike the previously reported interfacial crystals , which form for a wide variety of particle sizes and compositions , the rich panoply of patterns we have described arises appears only for particles meeting specific criteria of size and weight . to clarify this , we compared the electrokinetic pattern formation behavior in charge - stabilized dispersions of colloidal silica spheres , ps spheres and brominated ps spheres whose densities are 1.96 , 1.05 and 1.6 @xmath44 , respectively . the surface charge densities of ps and brominated ps are similar and do not depend on the ph . both are more highly charged than the silica particles we have discussed so far . silica s charge density , moreover , varies with ph , with an isoelectric point at ph 2 . the behavior of 2.9 @xmath1diameter silica spheres is very similar to that of the 3.0 @xmath1diameter spheres we have discussed to this point . these also can form various bulk and low - floating patterns . substantially larger silica spheres never leave the lower electrode at biases below the threshold for bulk electroconvection . substantially smaller , lighter and more diffusive silica spheres create less well organized patterns . for example , 1.58 @xmath1silica particles usually only form unstable tumbling clouds at comparatively low biases before being pushed to the upper electrode at higher biases . from previous studies @xcite , it is known that electrokinetic interactions are weaker for smaller spheres . this would tend to reduce the forces binding patterns such as colloidal vortex rings together . the smaller spheres more vigorous diffusion also would tend to hinder the localized clusters formation . it also suggests a possible role for gravity , which can be highlighted by considering spheres comparable in size to the pattern - forming silica particles that are composed of different materials . unlike silica spheres , the nearly neutrally - buoyant ps spheres fail to form any bulk patterns at any accessible bias over the entire diameter range from 0.3 @xmath1to 7.0 @xmath1 . these highly charged spheres should have comparably strong electrokinetic interactions to their silica counterparts . their diffusivity also should be similar . their failure to form bulk patterns demonstrates that the gravity plays a central role in the pattern formation we have described , a conclusion consistent with the observation that no bulk patterns form at negative bias . dense brominated ps spheres 1.6 @xmath1 in diameter can form tumbling clouds similar to those formed by comparably sized silica spheres . whereas clouds of 1.58 @xmath1silica spheres tumble vigorously , however , clouds of brominated ps particles have no detectable internal convection . such quiescent itinerant clouds also appear for dispersions of 3 @xmath1silica spheres but only at much lower concentrations . in both cases , each drifting cloud is about 50 @xmath1across and its constituent particles are separated by about 10 @xmath1 . stable clusters also do not form at very high or very low particle concentrations . dispersions at @xmath45 form chaotic clouds at low bias and jammed clusters at high bias . the absence of cyclic motion in low - concentration dispersions demonstrates that the flow patterns we have observed do not arise in the electrolyte alone under the relevant range of conditions , but rather require the spheres presence . the patterns concentration dependence further highlights the role of cooperative electrohydrodynamic interactions in establishing the distinct patterns . isolated particles behavior provides insights into these forces that help to clarify how such patterns form . to best highlight the role of collective behavior in electrohydrodynamic pattern formation , we studied the simplest case : a single silica sphere in an electrolysis cell . a single charged colloidal sphere of buoyant mass @xmath46 and charge @xmath47 in a constant electric field not only experiences a downward gravitational force @xmath48 and an electrostatic force @xmath49 but also more complicated electrohydrodynamic and osmotic forces due to the electrolyte s response to the field . when the lower electrode is negative , the positive screening cloud of the sphere flows downward . these flowing ions , including @xmath16and @xmath15created by hydrolysis , entrain a flow of water . the friction between this electrohydrodynamic flow and the sphere drags the sphere downward . the sphere s cloud of screening ions is deformed not only by the electric field , but also by the hydronium flux . this downward osmotic force and the downward electrohydrodynamic drag are second - order responses to the electric field , and therefore should be smaller than the electrostatic levitating force @xmath49 . straightforward considerations allow us to estimate the bias needed to levitate a particle against gravity . a single sphere sediments through water at speed @xmath50 , where @xmath51 is the density of water and @xmath52 is its viscosity . for 3 @xmath1diameter silica spheres , @xmath53 . the same particle s electrophoretic velocity saturates at @xmath54 , where @xmath55 is the electrolyte s debye - hckel screening length , and @xmath56 is the proton charge @xcite . if we choose the surface charge density @xmath57 @xcite and @xmath58 , these two velocity scales would be equal at about @xmath59 . thus , considering the neglected electrohydrodynamic and osmotic drags , the threshold for levitation should be at least 0.1 higher than the water decomposition voltage . this is consistent with our experiments . on a sphere tracks variations in ph . ( c ) the electric field @xmath60 is independent of height outside of the debye - hckel screening layers . ( d ) even so , the charge - regulating spheres experience an upward force that increases with height in the cell . this force opposes the force due to gravity . ] in fact , the ph is not uniform across the cell . [ fig : eqmg ] schematically shows how the profiles of ph , sphere charge @xmath47 , electric field @xmath60 , and electric force @xmath61 vary with height in the cell . even though the electric field is uniform in the bulk , the electrokinetic force levitating a charge - regulating silica sphere increases with height . when this force competes with gravity , therefore , it is possible to obtain two equilibrium heights , one near the lower electrode when gravity dominates , and another near the upper electrode where the electrokinetic force dominates . the situation is more complicated for two spheres because hydrodynamic forces between the spheres and bounding surfaces are long - ranged . reference @xcite shows the numerically calculated streamlines for one or two electrophoretically levitated charged spheres . two particles sinking towards the lower electrode create a back flow which tends to drive them apart @xcite . conversely , two electrically levitated particles will attract each other . this electrohydrodynamic interaction at least heuristically explains why an initially homogeneous colloidal fluid breaks up into discrete clusters as it is driven away from the lower electrode . we observed that the threshold for levitating a single sphere is lower than that for levitating an entire monolayer . this most likely reflects the reduced hydrodynamic drag coefficient for many spheres in a monolayer , as compared to that for a single sphere @xcite . the increased hydrodynamic attraction generated by the intensified ionic fluxes in the more concentrated dispersions would tend to draw these sedimented spheres together into clusters . this , however , tends to reduce their overall hydrodynamic drag @xcite , causing them to rise less than they otherwise might have . to compare our system s behavior with these predictions , we performed experiments on 3 @xmath1diameter silica spheres from a monolayer with areal density below @xmath62 , _ i.e. _ fewer than three particles per frame . abruptly applying a bias causes these isolated spheres to jump off the lower electrode during the transient in which vertical ionic concentration gradients are established . below a threshold voltage , the particles eventually return to the lower wall . above this threshold , they rise to the top wall and remain suspended . [ fig : single ] shows how the single sphere s height changes with applied voltage as the voltage is first increased and then decreased . the point at which a single sphere is electrolevitated to the upper wall is hysteretic : levitated particles only return to the lower wall when the voltage is lowered substantially . the explanation for this hysteresis also can be found in fig . [ fig : eqmg ] . at the critical bias , @xmath63 for all @xmath64 in the bulk . by contrast , a particle can only fall from the upper electrode , when the field falls below threshold at the upper electrode @xmath65 . therefore , the two thresholds are different . the inset to fig . [ fig : single ] shows this hysteresis more clearly at three different salt concentrations . such hysteresis supports the contention that the levitating force acting on silica spheres increases with the height @xmath22 . for salt concentration above 1 , a single particle can never be forced to the upper electrode . this is consistent with the observation that particles move less vigorously and cooperative structures are harder to form at high ionic concentrations . diameter in a @xmath23 cell as a function of applied bias . the voltage is increased and then decreased step by step . inset : the threshold voltage required to levitate spheres . ] attractions between particles are detectable only at relatively low particle concentrations ( _ e.g. _ @xmath66 ) and high applied voltages ( _ e.g. _ 2 ) . in such cases , particles attract each other as they levitate and so form tumbling clouds . the typical inter - particle separation in the most diffuse clouds is about 10 @xmath1and decreases with increasing bias . these field - induced attractions are surprisingly long - ranged , extending to roughly 50 @xmath1 . at biases above roughly 5 , even isolated spheres travel in circular orbits whose diameters range from 50 to 150 @xmath1 in the vertical plane . this motion is due to electroconvection in the underlying electrolyte . it almost certainly plays no role in forming the highly organized and substantially smaller - scale patterns discussed above . generally speaking , single particles are not stably levitated into the bulk of the electrolyte , and they certainly do not trace out the complex trajectories characterizing microscopic patterns such as those in figs . [ fig : vortexring ] - [ fig : firework ] . these observations help to confirm that such patterns do not arise from electroconvection in the electrolyte alone . we are left to conclude that the spheres play an active role in dynamical pattern selection . we have demonstrated that many distinctive patterns reproducibly emerge from the interplay of gravity and electrohydrodynamic forces in charge - stabilized colloid . at biases above the threshold for hydrolysis but below the onset of electroconvection , only fluxes of @xmath15and @xmath16ions flow through the system . electroneutrality favors overlap of these fluxes ; however , `` friction '' between opposite flows is minimized if these flows are separated . the optimal balance between these effects separates the fluxes by a characteristic lengthscale comparable to the debye - hckel screening length , or roughly 0.1 @xmath1 . this is much smaller than the extent of the experimentally observed clusters whose diameters extend to roughly 100 @xmath1 . electroneutrality suppresses the formation of such large - scale convective structures in the absence of the charged spheres . in other words , are not simply passive tracers for a pre - existing flow in the electrolyte , but rather they and their charges are essential for establishing the convective flows at the heart of the observed patterns . there are three general features about the dynamics of all clusters . ( 1 ) they have the same circulating direction as shown in fig . [ fig : schematic ] . ( 2 ) the circulation speed increases with the bias as clusters tend to become more compact . ( 3 ) the number of particles in each cluster increases with the bias . the bias - induced attraction of the second feature is crucial to understand the clusters formation . the breathing mode in figs . [ fig : oscillate ] and [ fig : firework ] may result from competition between this as - yet unexplained attraction and a combination of screened - coulomb repulsions and induced dipole repulsions among spheres . while the sequence of dynamical transitions from sedimented layer to tumbling clouds to highly organized clusters is reproducible , the crossover potentials depend sensitively on the changing electrochemical properties of the electrodes , and therefore can vary by as much as @xmath67 from run to run . we have reported a large family of previously unrecognized self - organized colloidal patterns that form in constant vertical electric fields at biases just above the decomposition voltage of water . our simple electrolysis cell gives rise to a complicated coupled system of hydrodynamic flows and ionic fluxes that are shepherded by the very colloidal particles they transport . we have discussed the electrohydrodynamic forces on a single particle in this system , and suggested how many - body coupling might give rise to the various dynamical structures we have observed . at low voltages , colloidal spheres in a specific range of charges , densities and mobilities are found to form various microscopic quasi - steady - state clusters through the competition of gravity and electrohydrodynamic levitating forces . these patterns falls into two categories : clusters levitated into the bulk and clusters localized near the lower electrode . their structures and evolutions are different , but all share some general features , such as persistent circulation and coarsening with increasing voltage . the simple behavior of a single particle in the field demonstrates that these microscopic clusters are formed cooperatively . patterns dependence on such control parameters as the composition and number density of particles , cell thickness @xmath2 , salt concentration and applied voltage range have been discussed .

applying an electric field to an aqueous colloidal dispersion establishes a complex interplay of forces among the highly mobile simple ions , the more highly charged but less mobile colloidal spheres , and the surrounding water . this interplay can induce a wide variety of visually striking dynamical instabilities , even when the applied field is constant . this article reports on the highly organized patterns that emerge when electrohydrodynamic forces compete with gravity in thin layers of charge - stabilized colloidal spheres subjected to low voltages between parallel plate electrodes . depending on the conditions , these spheres can form into levitating clusters with morphologies ranging from tumbling clouds , to toroidal vortex rings , to writhing labyrinths .

introduction experimental system colloidal electrokinetic pattern formation discussion summary

synchrotron emission from relativistic cosmic ray electrons accelerated in the galactic magnetic field dominates the diffuse radio continuum at frequencies below 1 ghz . it is an important foreground contaminant for measurements of the cosmic microwave background radiation , and also serves to probe the galactic magnetic field and cosmic ray distributions . measurements of the synchrotron frequency spectrum are thus of interest to several areas in astrophysics . an isotropic distribution of relativistic electrons at a single energy @xmath2 propagating in a uniform magnetic field @xmath3 has emissivity @xmath4 where @xmath5 is the pitch angle between the magnetic field and the line of sight , and @xmath6 is defined in terms of the modified bessel function of order @xmath7 with variable @xmath8 and @xmath9 @xcite . for a power - law distribution of electron energies @xmath10 propagating in a uniform magnetic field , the synchrotron emission is also a power law , @xmath11 where @xmath12 is antenna temperature , @xmath13 is the radiation frequency , and @xmath14 @xcite . measurements of the synchrotron spectral index provide important input for models of cosmic ray propagation . solar modulation reduces the local cosmic ray electron density for electron energies below a few gev so that synchrotron emission provides the most direct probe of low - energy cosmic rays . measurements of the cosmic ray spectrum above a few gev in turn inform models of the high - frequency synchrotron spectrum . energy losses from cosmic ray propagation steepen the cosmic ray spectrum , increasing @xmath15 toward higher energies . the observed steepening from @xmath16 at 5 gev to @xmath17 at 50 gev predicts a corresponding steepening in the synchrotron spectrum from @xmath18 at 1 ghz to @xmath19 at 100 ghz @xcite . comparison of the cosmic ray spectra to the predicted synchrotron spectrum is complicated by confusion from competing radio emission sources . the diffuse radio continuum is a superposition of the cosmic microwave background , synchrotron emission , free - free emission from the warm ionized interstellar medium , and emission from interstellar dust . a number of authors have attempted to disentangle the various emission sources to determine the synchrotron spectral index ( for a recent review see appendix a of @xcite ) . despite some discrepant results , the general trend shows a steepening of the synchrotron spectrum from @xmath20 at 22 mhz to @xmath21 above 23 ghz , in rough agreement with the observed cosmic ray spectra . several factors contribute to the observed scatter in estimates of the synchrotron spectral index . most estimates , particularly those below 23 ghz , assume a power - law spectrum for synchrotron and do not explicitly model spectral steepening . comparisons between closely - separated frequencies more accurately reflect the local synchrotron spectrum , but have larger uncertainties from competing emission sources or measurement offsets . analyses with broader frequency coverage reduce foreground and offset uncertainties but average over any spectral steepening . two additional effects are important for analyses including data from the wilkinson microwave anisotropy probe ( wmap ) at frequencies 23 to 94 ghz . a growing body of evidence suggests that a substantial fraction of the diffuse continuum near 23 ghz consists of electric dipole radiation from a population of small , rapidly spinning dust grains @xcite . spinning dust emission is expected to peak at frequencies near 20 ghz @xcite . analyses that ignore this component to attribute the observed emission only to synchrotron radiation tend to over - predict the synchrotron amplitude at frequencies near 23 ghz , biasing the derived spectral index to flatter values when comparing to lower frequencies and steeper values when comparing to higher frequencies . a second systematic error can result from improper treatment of offsets in the data . measurements from radio surveys at frequencies below 20 ghz generally include the absolute intensity ( zero level ) of the sky . the wmap differential radiometers are insensitive to any constant ( monopole ) intensity on the sky ; the zero level of the wmap sky maps is set so that the map intensity in the galactic polar caps matches a cosecant fit to the mid - latitude sky @xcite . analyses that directly compare low - frequency radio surveys to the wmap data without subtracting a monopole component from the radio data will miss the corresponding emission in the wmap bands , biasing the derived spectral index to steeper values . , width=278 ] the zero level bias can be significant . the @xcite survey at 408 mhz is commonly used to model synchrotron emission . the north galactic pole has measured temperature @xmath22 k at 408 mhz , while a @xmath23 fit to the same 408 mhz map predicts a polar contribution of only @xmath24 k. similar results apply to the southern hemisphere , where the measured polar cap temperature of @xmath25 k significantly exceeds the value @xmath26 k obtained from a @xmath23 fit . only 2.7 k of the difference can be attributed to emission from the cosmic microwave background , leaving a large residual . figure [ toy_model_fig ] illustrates the bias induced by including this residual at 408 mhz but excluding it from the wmap data . we take the 408 mhz map , remove the 2.7 k cmb monopole , and scale the remaining ( radio ) emission to 23 ghz using a power - law index @xmath27 to produce a toy model of the radio sky at 23 ghz . following the wmap processing , we then remove a monopole from the scaled map so that the map temperature in the south polar cap matches the @xmath23 fit . we then compute the bias in apparent spectral index by comparing the 408 mhz map to the scaled 23 ghz map before and after removing the scaled monopole . figure [ toy_model_fig ] shows the bias in spectral index at 23 ghz , binned by galactic latitude . dis - similar treatment of the map zero level creates a spatially varying bias @xmath28 , comparable to the total spectral steepening predicted by the measured cosmic ray spectra . the bias is largest in regions where the sky brightness is faintest , at high latitudes or away from the galactic center . c c c c frequency & calibration & offset & relative + ( ghz ) & uncertainty & uncertainty ( k ) & uncertainty@xmath29 + 0.022 & 0.05 & 5000 & 0.15 + 0.045 & 0.10 & 250 & 0.11 + 0.408 & 0.10 & 3.0 & 0.17 + 1.420 & 0.05 & 0.5 & 0.63 + 3.20 & 0.001 & 0.011 & 0.10 + 3.41 & 0.001 & 0.006 & 0.07 + 7.98 & 0.001 & 0.036 & 0.89 + 8.33 & 0.001 & 0.042 & 2.64 + 9.72 & 0.001 & 0.003 & 0.34 + 10.49 & 0.001 & 0.002 & 0.27 + + + measurement uncertainties in the absolute level of the sky brightness can also introduce bias in estimates of the synchrotron spectral index . many of the low - frequency surveys have uncertainty in the measured zero level approaching 30% of the polar cap brightness . as with the toy model above , such measurement errors introduce spatially dependent biases that are largest where the sky brightness is faintest . minimizing uncertainty in the derived synchrotron spectrum requires a combination of measurements with good sky coverage and good control of offset uncertainty at frequencies where competing emission sources are faint . no such ideal data set yet exists . in this paper , we model the synchrotron spectral index and curvature using low - frequency radio surveys with high sky coverage but large offset uncertainty , combined with higher - frequency measurements with limited sky coverage but still useful offset uncertainty . we model synchrotron emission using radio surveys at 22 mhz @xcite , 45 mhz @xcite , 408 mhz @xcite , and 1420 mhz @xcite . these surveys have full or nearly - full sky coverage at frequencies where galactic radio emission is significant , with gain and zero - level systematics controlled at the 1020% level . we supplement the radio surveys with sky maps from the absolute radiometer for cosmology , astrophysics , and diffuse emission ( arcade 2 ) instrument at 3 , 8 , and 10 ghz @xcite . the arcade 2 data observe both the galactic plane and mid - latitude regions ( @xmath30 ) with sufficient control of zero - level uncertainty to constrain the synchrotron curvature relative to the lower - frequency radio surveys . table 1 summarizes the input sky maps . the increase in the offset uncertainty at low frequency is compensated by a corresponding increase in sky brightness . the final column shows the relative measurement uncertainty for a mid - latitude region , defined as the ratio of the combined offset and calibration uncertainty to the measured brightness at @xmath31 after removing the cmb monopole . the selected maps provide roughly uniform relative sensitivity to synchrotron emission over 2.5 decades of frequency . we convert all maps to units of antenna temperature and subtract the cmb monopole at ( thermodynamic ) temperature 2.725 k from the measured sky temperatures . we then convolve each map to the 116 angular resolution of the arcade 2 instrument . at frequencies of 10 ghz and below , both thermal dust emission and spinning dust emission are negligible . free - free emission , however , can still be appreciable . we correct the convolved maps by scaling the wmap 7-year maximum entropy free - free model @xcite to each frequency using spectral index @xmath32 , convolving the scaled model to 116 angular resolution , and subtracting the convolved model from each sky survey . the resulting maps are dominated by synchrotron emission . , width=336 ] the input sky maps define a data set @xmath33 sampled at discrete pixel directions @xmath34 and 10 discrete frequencies @xmath13 ranging from 22 mhz to 10 ghz . we restrict the analysis to the 8% of the sky observed at all 10 frequencies . figure [ sky_coverage_fig ] shows the resulting sky coverage . within the common sky coverage , we model synchrotron emission as a modified power law @xmath35 with spectral index @xmath36 and curvature @xmath37 defined with respect to reference frequency @xmath38 = 310 mhz . the adopted value for @xmath38 minimizes covariance between the fitted amplitude @xmath39 and spectral index @xmath36 , simplifying extrapolation to other frequencies . for each pixel @xmath34 we define a 10 by 10 data covariance matrix @xmath40 with diagonal elements determined by the instrument noise , calibration and offset uncertainty . the input maps are not all linearly independent . measurements at 22 mhz used the 408 mhz map to determine the declination dependence of the gain . we estimate the resulting correlation of spatial structure in the two maps at 50% . the arcade 2 maps have independent instrument noise but share a fraction of the offset uncertainty related to absolute thermometry uncertainty and ground glint @xcite . all 10 maps share a common model for free - free emission . we conservatively estimate the uncertainty in the free - free correction at 30% of the free - free amplitude ; however , the results do not change significantly as the model free - free amplitude is varied by as much as 50% . off - diagonal elements in @xmath40 include these effects . diameter patch on the galactic plane centered on the the brightest pixels in the arcade 2 sky coverage . for clarity , the data are plotted relative to the best - fit power - law model ( @xmath41 ) with zero curvature ( dashed line ) . all fits include the non - trivial covariance between individual data points . the data are well described by a model with spectral index @xmath42 and curvature @xmath43 . [ raw_spectra_fig ] , width=249 ] for each pixel , a least - squares minimization determines the best - fit parameters @xmath39 , @xmath36 , and @xmath37 . figure [ raw_spectra_fig ] shows the measured temperature and best - fit model for the brightest galactic plane region @xmath44 within the common sky coverage . the data show evidence for spectral curvature , with best - fit values @xmath42 and @xmath43 evaluated at @xmath45 mhz . the spectral curvature @xmath37 for this region is significant at approximately 3 standard deviations compared to the baseline model with @xmath46 . we may extrapolate the spectral models to compare the results at frequencies 10 ghz and below to independent determinations of the spectral index using wmap data at frequencies 23 ghz and above . we compute the antenna temperature of the modeled spectra to derive the effective power - law index for frequencies near 23 ghz . note that this is not equivalent to evaluating eq . [ synch_model ] at @xmath47 ghz , which would yield the scaling from 310 mhz to 23 ghz but not the power - law index at 23 ghz . the mean for all 258 pixels in the common sky coverage is @xmath48 with standard deviation @xmath49 . the extrapolated value compares well with independent determinations of the spectral index above 23 ghz . @xcite analyze wmap polarization data to derive synchrotron spectral index @xmath50 averaged over the full sky . @xcite use a bayesian analysis of polarization data and find the mean spectral index @xmath51 with pixel - to - pixel standard deviation @xmath52 over the high - latitude sky . @xcite use template fitting techniques to derive spectral index @xmath53 between 23 and 33 ghz . the mean spectral index derived from the 10 low - frequency radio surveys agrees with the value derived from wmap data at higher frequencies . much of the scatter in the extrapolated spectral indices results from pixels at high latitude where the emission is faintest . the extrapolated index in these pixels can reach unphysical values . we reduce the scatter in the fitted spectral parameters by applying additional constraints using wmap data at 23 ghz and above . the simplest such constraint , adding wmap temperature data to the multi - frequency fit , is problematic . not only would the procedure need to include additional free parameters to account for emission from thermal dust or spinning dust ( both negligible at lower frequencies ) , but each of the low - frequency maps would require a correction to remove the monopole contribution missing from the wmap data . although the wmap zero level is clearly defined by a @xmath23 fit to mid - latitude data , the astrophysical interpretation of a similar procedure applied to low - frequency radio surveys is less clear . the coldest region of the radio sky is not at the galactic poles , but at mid - latitudes above the galactic anti - center . subtraction of too large a monopole can leave unphysical negative residuals . limited sky coverage exacerbates this problem . , width=259 ] we avoid these problems by using a constraint based on the spectral index derived solely from wmap data . for each pixel , we use the radio data ( table [ sky_map_table ] ) to fit the synchrotron amplitude @xmath54 over a 2-dimensional grid in the spectral parameters @xmath36 and @xmath37 ( eq . [ synch_model ] ) . at each grid point , we compute the @xmath55 value @xmath56 where @xmath57 is the inverse covariance matrix and @xmath58 is the difference vector between the measured and modeled temperatures . we then use the spectral parameters @xmath36 and @xmath37 to evaluate the power - law index at 23 ghz and compare the resulting value to a prior . we use the difference between the extrapolated spectral index and the prior to augment the @xmath55 at each grid point , @xmath59 where @xmath60 is the model spectral index evaluated at 23 ghz , and @xmath61 is the prior at 23 ghz . the minimum @xmath55 over the entire grid then defines the best - fit model at that pixel . this allows inclusion of the spectral information derived from frequencies above 23 ghz without confusion from either additional emission components ( thermal or spinning dust ) above 23 ghz or the missing zero level in the wmap data . corresponds to a steepening of the local spectral index by an amount @xmath1 every octave in frequency . [ curv_hist ] , width=249 ] figure [ param_maps ] shows the best - fit spectral parameter maps , while figures [ beta_hist ] and [ curv_hist ] show the distribution of the best - fit values for the spectral index and curvature . including the constraint at 23 ghz , the best - fit spectral index has mean value @xmath62 and standard deviation @xmath63 at reference frequency @xmath45 mhz . the best - fit curvature has mean value @xmath64 with standard deviation @xmath65 . the corresponding spectral index at 23 ghz is @xmath66 with standard deviation 0.05 . inclusion of the prior at 23 ghz does not induce a significant shift in the mean for the extrapolated spectral index , but does significantly reduce the pixel - to - pixel scatter . comparison of the parameter distributions with and without the spectral constraint at 23 ghz demonstrates that the addition of the spectral constraint mainly affects the fitted curvature values ( figs . [ beta_hist ] and [ curv_hist ] ) . the adopted value for the spectral constraint comes from independent analysis of wmap data and is weighted toward regions of higher synchrotron intensity . we test whether this creates a bias in the fitted curvature values by splitting the observed sky coverage into two subsets of equal area , defined by the brightest and faintest 50% of the fitted amplitudes at 310 mhz . within each subset , we compare the mean and standard deviation of the fitted curvature derived from the 10-frequency fit or the enhanced fit including the constraint at 23 ghz . the `` bright '' subset shows no shift in the mean spectral curvature when the 23 ghz constraint is added ( although the scatter is significantly reduced ) . the ` faint' subset shows both a reduction in scatter and a modest shift in the mean value , with curvature parameter steepening from -0.034 to -0.049 when the 23 ghz constraint is included . this shift is less than one standard deviation : given the limited sky coverage , the available radio data do not yet provide significant evidence for spatial variation in the synchrotron curvature . radio data show statistically significant steepening of the synchrotron spectrum from 22 mhz to 10 ghz . the nearly uniform relative uncertainty of the selected data minimizes dependence of the fitted parameters on offset or calibration errors at any one frequency . we test whether the best - fit parameters are particularly sensitive to any one input map by repeating the analysis after dropping one or two maps from the fit . we select either one radio survey ( 22 mhz , 45 mhz , 408 mhz , or 1420 mhz ) or a pair of arcade frequency channels ( 3 ghz , 8 ghz , or 10 ghz ) and repeat the fit after deleting the corresponding elements from the data vector @xmath67 and covariance matrix @xmath40 . the resulting shift in either the spectral index or curvature parameters is smaller than the pixel - to - pixel standard deviation using all 10 frequency channels . systematic errors in the offset or temperature calibration do not appear to dominate the multi - frequency analysis . the steepening of the synchrotron spectrum is broadly consistent with models of cosmic ray propagation in the galactic magnetic field . @xcite combine radio observations with the galprop cosmic ray propagation code to model synchrotron emission on the galactic plane . they find a power - law index @xmath68 from 408 mhz to 2.3 ghz and @xmath69 from 2.3 ghz to 23 ghz . the corresponding values for the spectral steepening model presented here are @xmath70 from 408 mhz to 2.3 ghz and @xmath71 from 2.3 ghz to 23 ghz , in good agreement with the cosmic ray model . we may use the best - fit values in each pixel to predict the synchrotron spectrum at higher frequencies where the emission is fainter and competing sources stronger . previous attempts to disentangle competing emission from free - free , synchrotron , thermal dust , and spinning dust emission using wmap data have suffered from degeneracy between the synchrotron and spinning dust emission , both of which are falling at frequencies above 33 ghz ( see , e.g. , the discussion in @xcite ) . extending the synchrotron curvature observed at lower frequencies into the millimeter band reduces confusion between the spinning dust and synchrotron spectra and may facilitate characterization of both the spatial distribution and frequency spectrum of spinning dust emission in the interstellar medium . c c frequency & power - law + ( ghz ) & index + 0.022 & -2.36 + 0.045 & -2.44 + 0.408 & -2.67 + 3.3 & -2.89 + 23 & -3.09 + 33 & -3.13 + 41 & -3.15 + 61 & -3.19 + 94 & -3.24 + table [ beta_table ] shows the local power - law index @xmath72 at selected frequency bands . the modeled spectrum steepens by @xmath1 every octave in frequency , from @xmath73 at 408 mhz to @xmath74 at 94 ghz . note , however , that the spectral steepening observed at low frequencies can not continue indefinitely . _ fermi _ measurements of the cosmic ray energy spectrum are consistent with a single power law from energy 7 gev to 1 tev @xcite . if anything , the _ fermi _ data suggest a modest flattening of the cosmic ray energy spectrum at higher energies , which would induce a positive curvature to the synchrotron spectrum at frequencies above 23 ghz . , width=259 ] @xcite combine radio data at 22 , 45 , 150 , 408 , and 1420 mhz with wmap data at 23 through 94 ghz and _ fermi _ large area telescope cosmic ray measurements and the galprop code to estimate the magnetic field intensity and synchrotron spectrum from 22 mhz to 94 ghz . figure [ synch_index_comparison ] compares the resulting synchrotron spectral index ( diffusion model with injection index 1.3 ) to the curvature model from this paper . the cosmic ray model analyzes a limited latitude range @xmath75 , which we follow in figure [ synch_index_comparison ] by excluding pixels at latitudes @xmath76 . the two methods agree for frequencies below 408 mhz ( where they share common radio data ) but differ at higher frequencies . the cosmic ray model shows an increased synchrotron curvature from 408 mhz to a few ghz , followed by a spectral break to near - constant index @xmath19 at higher frequencies . the arcade 2 data at 310 ghz do not reproduce these features , but are instead consistent with constant spectral curvature from 22 mhz to 10 ghz ( fig [ raw_spectra_fig ] ) . both models agree at frequencies near 23 ghz . the spectral index at 23 ghz derived from the 10-frequency radio data _ without _ the 23 ghz prior is consistent with independent measurements of the index above 23 ghz and with the full radio fit including the 23 ghz prior . the radio data , taken alone , do not support a break in the synchrotron spectrum at ghz frequencies . comparison of the radio fit to the cosmic ray model suggests that any spectral break in the synchrotron spectrum must occur at frequencies above 23 ghz . direct confirmation of the synchrotron spectrum above 23 ghz remains a challenge . radio data are consistent with a synchrotron spectrum that steepens with frequency from 22 mhz to 10 ghz . direct comparison of low - frequency radio surveys with the wmap data at 23 to 94 ghz is complicated both by the presence of additional emission components at higher frequencies and by the subtraction of a substantial monopole component of sky emission by the differential wmap instrument . the synchrotron spectral index at 23 ghz , derived using only lower - frequency radio surveys , is consistent with the value derived independently using only data at higher frequencies . we extend the radio data by comparing the extrapolated index at 23 ghz to a prior based on higher - frequency data . the combined data have mean spectral index @xmath0 and curvature @xmath77 at reference frequency 0.31 ghz . the measured spectrum steepens by an amount @xmath1 every octave in frequency . comparison of the radio data to models including the cosmic ray energy spectrum suggests that any break in the synchrotron spectrum must occur at frequencies above 23 ghz . this research is based upon work supported by the national aeronautics and space administration through the science mission directorate under the astronomy and physics research and analysis suborbital program .

we combine surveys of the radio sky at frequencies 22 mhz to 1.4 ghz with data from the arcade-2 instrument at frequencies 3 to 10 ghz to characterize the frequency spectrum of diffuse synchrotron emission in the galaxy . the radio spectrum steepens with frequency from 22 mhz to 10 ghz . the projected spectral index at 23 ghz derived from the low - frequency data agrees well with independent measurements using only data at frequencies 23 ghz and above . comparing the spectral index at 23 ghz to the value from previously published analyses allows extension of the model to higher frequencies . the combined data are consistent with a power - law index @xmath0 at 0.31 ghz , steepening by an amount @xmath1 every octave in frequency . comparison of the radio data to models including the cosmic ray energy spectrum suggests that any break in the synchrotron spectrum must occur at frequencies above 23 ghz .

introduction sky maps analysis discussion conclusions

in 1989 , my team began to work on developing a robotic telescope for ccd imaging of relatively faint objects . the history of the project is discussed in several papers ( e.g. , filippenko 2001 ; richmond , treffers , & filippenko 1993 ) , and several prototypes were used over the years . in 1996 , we achieved first light with our present instrument , the 0.76-m katzman automatic imaging telescope ( kait ) at lick observatory on mt . hamilton , california . it took the better part of another year to eliminate most of the remaining bugs in the system , and useful scientific results started appearing in 1997 . absolutely vital contributions to the programming and to the observing strategy were made by dr . weidong li , who joined my group in 1997 . kait is a fully robotic instrument whose control system checks the weather , opens the dome , points to the desired objects , acquires guide stars ( in the case of long exposures ) , exposes , stores the data , and manipulates the data automatically , all without human intervention . we reach a limit of @xmath0 mag ( @xmath1 ) in 25-s unfiltered , unguided exposures , while 5-min guided exposures yield @xmath2 mag . kait acquires well - sampled , long - term light curves of sne and other variable or ephemeral objects projects that are difficult to conduct at other observatories having a large number of users with different interests . one of our main goals is to discover nearby sne to be used for a variety of studies . special emphasis is placed on finding them well before maximum brightness . although the original sample of our lick observatory supernova search ( loss ; li 2000 ; filippenko 2001 ) had only about 5000 galaxies , in the year 2000 we increased the sample to @xmath3 galaxies ( most with redshift @xmath4 km s@xmath5 ) , separated into three subsets ( observing baselines of 2 days for about 100 galaxies , 36 days for @xmath6 galaxies , and 714 days for @xmath7 galaxies ) . we are able to observe @xmath8 galaxies per night in unfiltered mode . our software automatically subtracts new images from old ones ( after registering , scaling to account for clouds , convolving to match the point - spread - functions , etc . ) , and identifies sn candidates ( fig . 1 ) which are subsequently examined and reported to the central bureau for astronomical telegrams by numerous undergraduate research assistants in my group , working with weidong . * fig . 1 . * kait images of ngc 523 before _ ( left ) _ and after _ ( center ) _ sn 2001en appeared . the difference image _ ( right ) _ shows the sn ( circled ) , along with a cosmic ray ( boxed ) . a web page describing loss is at http://astro.berkeley.edu/@xmath9bait/kait.html . loss found its first supernova in 1997 sn 1997bs , which might not even be a `` genuine '' sn ( van dyk 2000 ) . in 1998 , mostly during the second half of the year , loss discovered 20 sne , thereby breaking the previous single - year record of 15 held by the beijing astronomical observatory supernova search . in 1999 , loss doubled this with 40 sne . in 2000 , loss found 38 sne , even though we spent a significant fraction of the observing time expanding the database of monitored galaxies rather than searching for sne . with this expanded database , loss discovered 68 sne in 2001 and 52 sne in 2002 through july . we discovered sn 2000a and sn 2001a , and hence the first supernova of the new millennium , regardless of one s definition of the turn of the millennium ! during the past few years , kait has discovered _ about half _ of all nearby sne reported world - wide , from all searches combined and through july 2002 it accounted for well over half ( 52/86 ) of them . thus , kait / loss is currently the world s most effective search engine for nearby sne . to further increase the sample of sne , my group recently decided to coordinate with michael schwartz of the tenagra observatory ( 0.60 m and 0.75 m telescopes ) , using largely complementary sets of galaxies ( the overlap between galaxy lists is larger during those months when either observatory often has bad weather ) . our `` lick observatory and tenagra observatory supernova search '' ( lotoss ; schwartz 2000 ) should discover almost all of the nearby sne over the accessible areas of the sky . already , the tenagra telescopes have discovered several sne ( e.g. , schwartz & li 2001 , 2002 ) . also , this strategy leaves kait with more time to conduct multicolor follow - up photometry of sne . at lick and keck observatories , we spectroscopically confirm and classify nearly all of the sne that other observers havent already classified . thus , the sample suffers from fewer biases than most . the distribution of types through sn 2002dy is 93 sne ia , 76 sne ii , 12 sne iin , 1 sn iib , 12 sne ib , 16 sne ic , and 6 unknown . we have started to determine the hubble types of the host galaxies of the sne ( van den bergh , li , & filippenko 2002 ) , as a first step in the calculation of rates of various types of sne . already , our observations and monte - carlo simulations have shown that the rate of spectroscopically peculiar sne ia is considerably larger than had previously been thought ( li 2001a ) . follow - up observations for the discovered sne are emphasized during the course of loss . our goal is to build up a multicolor database for nearby sne . because of the early discoveries of most loss sne , our light curves usually have good coverage from pre - maximum brightening to post - maximum decline . moreover , all loss sne are automatically monitored in unfiltered mode as a byproduct of our search ; these can sometimes be useful for other studies ( e.g. , matheson 2001 ) . the positions of sne in kait early - time images were used to identify the same sne at very late times in _ hubble space telescope _ images ( li 2002 ) , allowing us to determine the late - time decline rates . loss also discovers novae in nearby galaxies ( e.g. , m31 ) , cataclysmic variable stars , and occasionally comets ( e.g. , li 1998 ) . although it records many asteroids , we do nt conduct follow - up observations , so most of them are lost . high - quality observations of nearby sne ia provide valuable information about their progenitor evolution and the relevant physics . analyses of samples of well - observed , nearby sne ia enable observers to study the differences among sne ia , empirical correlations , and possible environmental effects ( e.g. , hamuy 2000 ) . it is thus important to expand the sample of well - observed nearby sne ia . thus , a substantial fraction of kait s time is devoted to follow - up photometry of bright sne ia . moreover , studies of high - redshift sne ia have revealed a surprising cosmological result , that the expansion of the universe is currently accelerating , perhaps due to a nonzero cosmological constant ( e.g. , riess 1998 , 2001 ; perlmutter 1999 ) . this result , however , is based on the assumption that there are no significant differences between sne ia at high redshift and their low - redshift counterparts . in particular , we rely on the luminosity / light - curve correlation , as quantified in a number of ways ( e.g. , riess 1998 ) ; phillips 1999 ; perlmutter 1999 ) , to standardize " the luminosities of different sne ia . but what if some sne ia do nt conform with this correlation ? if there are more of them at high redshifts than at low redshifts , systematic errors may creep into the analysis . we need to find and investigate such objects at low redshifts . sn 2000cx in the s0 galaxy ngc 524 is a case in point . it was discovered and confirmed by loss in july 2000 , and at became the brightest sn of the year 2000 . a follow - up program of multicolor photometry and spectroscopy was established at lick observatory ; photometry of sn 2000cx was also gathered at the wise observatory in israel . the results are described in detail by li ( 2001b ) ; here i summarize the main points . a very peculiar object , sn 200cx is , indeed , unique among all known sne ia . the light curves can not be fit well by any of the fitting techniques currently available ( e.g. , mlcs and the stretch method ) ; see figure 2 . there is an apparent asymmetry in the rising and declining parts of the @xmath10-band light curve , while there is a unique `` shoulder''-like evolution in the @xmath11-band light curve . the @xmath12-band and @xmath13-band light curves have relatively weak second maxima . in all @xmath14 passbands the late - time decline rates are relatively large compared to other sne ia . * * the mlcs fit ( riess 1998 ; _ left panel _ ) and the stretch method fit ( perlmutter 1999 ; _ right panel _ ) for sn 2000cx . the mlcs fit is the worst we had ever seen through 2000 . for the stretch method fit , the solid line is the fit to all the data points from @xmath15 = @xmath168 to 32 days , the dash - dotted line uses only the premaximum datapoints , and the dashed line only the postmaximum datapoints . the three fits give very different stretch factors . from li ( 2001b ) . sn 2000cx has the reddest @xmath17 color before @xmath18 7 days among several sne ia , and it subsequently has a peculiar plateau phase where @xmath17 remains at 0.3 mag until @xmath15 = 15 days . the late - time @xmath17 evolution of sn 2000cx is found to be rather blue , and is inconsistent with the fit proposed by lira ( 1995 ) and phillips ( 1999 ) . sn 2000cx also has very blue @xmath19 and @xmath20 colors compared with other sne ia . our earliest spectrum of sn 2000cx ( @xmath21 days ) reveals remarkable resemblance to those of sn 1991t - like objects , with prominent fe iii lines and weak si ii lines . as in the case of sn 1991 t , si ii lines strengthened around the time of maximum brightness . however , the subsequent spectral evolution of sn 2000cx is quite different from that of sn 1991 t . the fe iii and si ii lines remain strong , and the fe ii lines remain weak , in the spectra of sn 2000cx until @xmath18 20 days , indicating that the excitation stages of iron - peak elements change relatively slowly in sn 2000cx compared with other sne ia , and suggesting that the photosphere of sn 2000cx stays hot for a long time . both iron - peak and intermediate - mass elements are found to be moving at very high velocities in sn 2000cx . the @xmath22 measured from the si ii @xmath236355 line shows a peculiar ( nearly constant ) evolution . we find that the delayed detonation model dd3 ( woosley & weaver 1994 ) investigated by pinto & eastman ( 2001 ) accounts for the observations of sn 2000cx rather well . this model suggests that sn 2000cx is similar to sn 1991 t , but with a larger @xmath24ni production and a higher kinetic energy ( i.e. , greater expansion velocity for the ejecta ) . we emphasize that because of uncertainties in the current theoretical models for sne ia , various views should be considered . for example , the big difference between sn 2000cx and sn 1991 t in their @xmath25 and @xmath13 light curves may suggest that they are two very different objects . but sn 2000cx is not the end of the story , when it comes to peculiar sne ia . a more recent , even stranger sn ia was sn 2002cx , many of whose properties are the opposite of those of sn 2000cx ! sn 2002cx was discovered in may 2002 by wood - vasey ( 2002 ) with the oschin 1.2-m telescope at palomar observatory in unfiltered images . it host galaxy is cgcg 044 - 035 , at a redshift of @xmath26 km s@xmath5 ( determined from h ii region emission lines ) . an optical spectrum ( matheson 2002 ) identified the sn as a peculiar sn 1991t - like event at about a week before maximum brightness , but the object is very _ underluminous _ ( instead of somewhat overluminous ) compared with normal sne ia . the si ii @xmath236355 and ca ii h & k lines are extremely weak or absent , but the fe iii lines at 4300 and 5000 are present and indicate very low expansion velocity only half that of normal sne ia . recognizing the uniqueness of sn 2002cx shortly after its discovery , we established a follow - up program of multicolor photometry lick observatory . spectra of the sn were obtained with the fred l. whipple observatory 1.5-m telescope and also with the keck 10-m telescopes . li ( 2003 ) discuss the results in detail ; here i provide only a brief summary . besides being subluminous by @xmath27 mag at all optical wavelengths relative to normal sne ia ( implying that only a small amount of @xmath24ni was produced ) , sn 2002cx has peculiar photometric evolution ( fig . 3 ) . in the @xmath10 band it has a decline rate of @xmath28 mag , similar to those of sn 1994d and sn 1999ac , but it is less luminous by @xmath29 mag than sn 1994d and sn 1999ac . the @xmath12 band has a broad peak , and the @xmath13 band has a unique plateau that lasts until about 20 days after @xmath10 maximum . the late - time decline is rather slow in all @xmath14 bands . the @xmath30 color evolution is nearly normal , but the @xmath31 and @xmath32 colors are very red . * * comparison ( li 2003 ) between the _ b , v , r , i _ light curves of sn 2002cx and those of sn 1991 t ( lira 1998 ) , sn 2000cx ( li 2001b ) , sn 1994d ( richmond 1995 ) , sn 1991bg ( filippenko 1992 ; leibundgut 1993 ) , and sn 1999ac ( li , in preparation ) . all light curves are shifted in time and peak magnitude to match those of sn 2002cx . the premaximum spectrum of sn 2002cx resembles those of sn 1991t - like objects , but with extremely low expansion velocities , the lowest ever measured for a sn ia . the spectral evolution is dominated by fe - group element lines , with very weak intermediate - mass element features . the nebular phase was reached unprecedently soon after maximum , despite the low velocity of the ejecta , implying that the ejected mass is low . the nebular - phase spectrum is also quite different from those of other sne ia ( fig . 4 ) ; there are mysterious emission lines near 7000 around 3 weeks after maximum brightness , and other differences as well . at late times , the spectrum is dominated by very narrow fe ii and co ii lines , and the object is very red . sn 2002cx is inconsistent with the observed sn ia decline rate vs. luminosity relation , or the spectral vs. photometric sequence . no existing theoretical model successfully explains all observed aspects of sn 2002cx , though the pulsating delayed detonation of a chandrasekhar - mass white dwarf or the he detonation of a sub - chandra white dwarf have some promising characteristics and should be pursued further . * the spectrum of sn 2002cx at @xmath15 = + 20/25 d , shown with spectra of other sne ia at older ages ( li 2003 ) . the _ upper panel _ shows the line identifications and the comparison of the spectra . the pairs of short vertical lines above the sn 2002cx spectrum mark possible double peaks , " while these below the sn 2002cx spectrum mark possible additional resolved lines ( compared with other sne ia ) . the lower panel shows the comparison between the @xmath15 = + 25 d spectrum of sn 2002cx after convolving with a gaussian function with @xmath33 = 2,500 km s@xmath5 , and the day + 47 spectrum of sn 1997br . note that although the double peaks " are gone , additional features seem to be present around 7000 in the spectrum of sn 2002cx . we have also used kait to obtain excellent light curves of sne ii , with complementary spectra obtained at lick observatory and elsewhere . these are being used to study the physical properties of sne ii , and also to derive distances through the expanding photosphere method ( epm ) , a variant of the baade ( 1926 ) method used to measure distances to variable stars . in two detailed studies , we derived epm distances to sn 1999em ( @xmath34 mpc ; leonard 2002a ) and sn 1999gi ( @xmath35 ; leonard 2002b ) . in addition to its cosmological use , knowing the epm distance to sn 1999gi allowed us to set constraints on the upper mass limit of its progenitor star of @xmath36 , through the analysis of prediscovery images . this is substantially less restrictive than the upper mass limit ( @xmath37 ) recently found in the same manner by smartt ( 2001 , 2002 ) . the increased upper limit results mainly from the larger distance derived through epm than was assumed by the smartt ( 2001 , 2002 ) analyses , which relied on less precise ( and less recent ) distance measurements to ngc 3184 . we have also obtained high signal - to - noise ratio spectropolarimetry of some sne ii - p with the keck 10-m and lick 3-m telescopes ( leonard 2001 ) . at early times , sne ii - p appear to be polarized very little , suggesting that any departures from spherical symmetry are small . this is encouraging news for those who attempt to derive epm distances for sne ii - p : unlike the empirically based method used to measure distances to sne ia , distances derived to sne ii - p rely on the assumption of a spherically symmetric flux distribution during the early stages of development ( i.e. , the plateau ) . we plan to obtain spectropolarimetry of additional sne ii - p , in order to much more thoroughly test the fundamental assumption of spherical symmetry in epm . however , multi - epoch spectropolarimetry shows that the polarization increased with time ( fig . 5a ) , implying a substantially spherical geometry at early times that becomes more aspherical at late times when the deepest layers of the ejecta are revealed . in addition , our data on other core - collapse sne indicates large polarizations for objects that have lost much of their envelope prior to exploding ( e.g. , sn ic 2002ap , fig . 5b ; see below , and leonard 2002c ) . for core - collapse events , then , it seems that the closer we probe to the heart of the explosion , the greater the polarization and , hence , the asymmetry . the current speculation is that the presence of a thick hydrogen envelope dampens the observed asymmetry . although core - collapse sne present a wide range of spectral and photometric properties , there is growing consensus that much of this variety is due to the state of the progenitor star s hydrogen and helium envelopes at the time of explosion . those stars with massive , intact envelopes produce type ii - plateau sne , those that have lost their entire hydrogen envelope ( perhaps through stellar winds or mass transfer to a companion ) result in sne ib , and those that have been stripped of both hydrogen and most ( or all ) of their helium produce sne ic ; see filippenko ( 1997 ) for a general review . * ( a ) _ ( left ) _ the temporal increase in the polarization of the type ii - p sn 1999em suggests greater asphericity deeper into the ejecta . ( right ) _ polarization level ( _ thin noisy lines _ ) of the peculiar sn ic 2002ap , with relative flux ( _ thick smooth lines _ ) overplotted for comparison of features ( leonard 2002c ) . recently , a new subclass of objects has emerged whose members generically resemble sne ic ( no hydrogen or obvious helium spectral features ) , but , unlike traditional sne ic , have spectra characterized by unusually broad features at early times , indicating velocities in excess of @xmath38 km s@xmath5 . a few also possess inferred kinetic energies exceeding that of `` normal '' core - collapse sne by more than a factor of 10 ( see , e.g. , nomoto 2001 ) . these objects are colloquially referred to as `` hypernovae , '' although not all of them are clearly more luminous or energetic than normal sne ic . intense interest in hypernovae has been sparked not only by their peculiar spectral features , but also by the strong spatial and temporal association between the brightest and most energetic of these events , sn 1998bw , and the @xmath39-ray burst ( grb ) 980425 ( e.g. , galama 1998 ) . there are only a few generally accepted members of this rare class ( e.g. , sn 1997dq , sn 1997ef ) . a related subclass of sne exhibits many of the characteristics of these objects , but with hydrogen present in the spectra ; the clearest examples are sn 1997cy and sn 1999e ( germany 2000 ; turatto 2000 ; filippenko 2000 , and references therein ) , and they , too , are sometimes called hypernovae . the hydrogen emission probably comes from the interaction of relatively hydrogen - poor ejecta with circumstellar gas previously expelled by the progenitor star ( and richer in hydrogen than the remaining parts of the progenitor ) . sn 2002ap in m74 ( mazzali 2002 ) is one of the most recent examples of a peculiar sn ic of this kind . so far , it has been the brightest supernova of the year 2002 . though not discovered by loss ( it went off between two kait observations , but was discovered during that interval by yoji hirose [ iauc 7810 ] ) , kait s observations set a useful limit on the explosion date . figure 6 shows the spectrum of sn 2002ap obtained after the sn reappeared following solar conjunction , about 5 months after maximum brightness . it is characterized by strong emission lines of intermediate - mass elements superimposed on a weak continuum , implying that the sn has entered the nebular phase . unusual narrow lines are visible on top of some of the broad - line profiles , including especially those of [ o i ] @xmath406300 , 6364 and mg i ] @xmath234571 . * 6 . * sn 2002ap ( _ top _ ) in the nebular phase , compared with the sn iib / ib / ic 1985f ( _ middle _ ) and the `` ordinary '' type ic sn 1987 m ( _ bottom _ ) at similar epochs ; the estimated day since explosion is indicated . the exact spectral classification of sn 1985f is unknown , since it was discovered long after maximum brightness ( see filippenko & sargent 1986 ; filippenko 1997 ) , but it provides the closest match we could find to the spectrum of sn 2002ap . the spectra are scaled so that the height of the [ ca ii ] @xmath407319 , 7324 blend is approximately the same in all three cases . although there is no guarantee that the three objects were at the same stage of their development ( they may evolve at different physical rates , even if they are approximately the same calendar age ) , one can see in figure 6 that relative to sn 1987 m , a typical sn ic , sn 2002ap has _ much _ stronger [ o i ] and mg i ] emission . the only sn ib / ic we have found comparable to sn 2002ap is sn 1985f , as shown . however , sn 2002ap exhibits a larger mg i ] @xmath234571 to [ o i ] @xmath406300 , 6364 ratio than that of sn 1985f , suggesting that we are seeing even closer to the o - ne - mg layer in sn 2002ap . qualitatively , this supports the hypothesis that sn 2002ap ( and perhaps other peculiar examples of the sn ic subclass ) have progenitors that are even more highly stripped than normal sne ic . there might be other factors to consider as well , but the sequence ii - p @xmath41 iib @xmath41 ib @xmath41 ib / c @xmath41 ic @xmath41 ic - pec may fundamentally be one dominated by the degree to which the envelope of the progenitor has been stripped . identifying the massive progenitor stars that give rise to core - collapse sne is one of the main pursuits of supernova and stellar evolution studies . using ground - based images of recent , nearby sne obtained primarily with kait , astrometry from the two micron all sky survey , and archival images from the _ hubble space telescope _ , we have attempted the direct identification of the progenitors of 16 type ii and type ib / c sne ( van dyk 2002 ) . we may have identified the progenitors of the type ii sne 1999br in ngc 4900 , 1999ev in ngc 4274 , and 2001du in ngc 1365 as supergiant stars with @xmath42 mag in all three cases . we may have also identified the progenitors of the type ib sne 2001b in ic 391 and 2001is in ngc 1961 as very luminous supergiants with @xmath43 to @xmath44 mag , and possibly the progenitor of the type ic sn 1999bu in ngc 3786 as a supergiant with @xmath45 mag . additionally , we have recovered at late times sne 1999dn in ngc 7714 , 2000c in ngc 2415 , and 2000ew in ngc 3810 , although none of these had detectable progenitors on pre - supernova images . in fact , for the remaining sne only limits can be placed on the absolute magnitude and color ( when available ) of the progenitor . the detected type ii progenitors and limits are consistent with red supergiants as progenitor stars , although possibly not as red as we had expected . our results for the sne ib / c do not strongly constrain either wolf - rayet stars or massive interacting binary systems as progenitors . i am grateful to the committee on research ( u.c . berkeley ) and the conference organizers for providing partial travel support to attend this meeting . my recent research on sne has been financed by the us national science foundation , most recently through grant ast-9987438 , as well as by nasa grants ar-8754 , go-9114 , go-9428 , and ar-9529 from the space telescope science institute , which is operated by aura , inc . , under nasa contract nas5 - 26555 . kait and its associated science have been made possible with funding or donations from nsf , nasa , the sylvia and jim katzman foundation , sun microsystems inc . , lick observatory , the hewlett - packard company , photometrics ltd . , autoscope corporation , and the university of california . i thank r. foley , d. c. leonard , and w. d. li for assistance with the figures .

i review recent optical observations of supernovae ( sne ) conducted by my group . the lick observatory supernova search with the 0.76-m katzman automatic imaging telescope is currently the world s most successful search for nearby sne . we also use this telescope to obtain multicolor light curves of sne . one of the more interesting sne we discovered is sn 2000cx , which differs from all previously observed sne ia . another very strange sn ia that we studied is sn 2002cx , many of whose properties are opposite those of sn 2000cx . extensive data on sne ii - p 1999em and 1999gi were used to derive distances with the expanding photosphere method . results from spectropolarimetry suggest that the deeper we peer into the ejecta of core - collapse sne , the greater the asphericity . we are using _ hubble space telescope _ data to identify , or set limits on , the progenitors of core - collapse sne .

the lick observatory supernova search (loss) sn 2000cx: a very weird sn ia sn 2002cx: an even weirder sn ia studies of type ii supernovae the peculiar sn ic 2002ap the progenitors of core-collapse sne acknowledgments

graphene has been used as an analog case to study relativistic phenomena in heavy atoms because the critical charge for the dirac fermions in graphene is of the order of unity @xcite . upon introduction of a single charge impurity into an otherwise perfect lattice , a bound state appears within the mass gap that separates the original dirac cones electronicstruct , zhou07 . as one tunes the coupling strength of this charge impurity to larger values , eventually a critical charge condition is achieved whereby the energy of the bound states passes into the continuum . in gapped graphene , this crossover has implications for the screening properties of the electron gas.@xcite the wave function of such a bound state decays exponentially with the distance from the charge . as the coupling strength becomes supercritical , the screening effect is sufficiently strong that the observed effective charge is reduced by almost 4 elementary units compared with the unscreened case.@xcite this phenomenon is similar to what happens in atomic physics when the elementary charge is around @xmath0.@xcite the shape of the cloud of screening charge closely follows the shape of the so called critical state " just before it merges with the continuum of the valence band . a couple of questions arise : i ) is there a change in the electronic structure that depends on the range of potential ? ii ) what are the properties that define whether the impurity potential strength is critical or not ? the paper is organized as follows . in the next section we obtain the predictions of the long - wavelength dirac formalism in the case of one or few short - range impurities . in the third section , after we briefly review the green functions on the lattice pertinent to the honeycomb lattice , we obtain analytical results for one and two impurities on the graphene lattice . in the fourth section , we generalize our results for multiple short - ranged impurities , and then we close with the conclusion . the motion of an electron with a fixed full ( pseudo - spin plus orbital ) angular momentum in a circularly symmetric potential @xmath1 is : @xmath2where , for definiteness , @xmath3 , where @xmath4 is a heaviside step function , @xmath5 is the radius of the well , and @xmath6 is negative ( positive ) for a well ( barrier ) . the full wave function is @xcite : @xmath7 here j is an eigenvalue of angular momentum @xmath8 , and @xmath9 can refer to either the @xmath10 or @xmath11 sublattice . to solve eq . ( dirac ) , one can express @xmath12 in terms of @xmath13 : @xmath14 in what follows , we will discuss the case when the energy level is at either gap edge ; for negative ( positive ) @xmath6 this is the lower ( upper ) edge . the solutions inside the well are : @xmath15the solutions outside the well are : @xmath16where @xmath17 is a bessel function of the first kind , @xmath18 is a modified bessel function of the second kind chosen to satisfy the boundary condition at infinity . the equation for the energy levels is : @xmath19for a solution to exist , the terms on the left - hand side and on the right - hand side must be either pure imaginary or pure real numbers . this means , that for @xmath20 , and with @xmath21 , then we require that @xmath22 for a solution . we are primarily interested in negative values of @xmath6 . we are interested also in what sequence the levels will merge into the continuum , depending on their angular momentum . one can numerically analyze eq . ( [ secular ] ) but for the sake of clarity we will analyze the effective schrodinger equation for @xmath13 , obtained from the system ( dirac ) : @xmath23a(r)= 0\ ] ] this is a wave equation for a particle in a potential with functional form @xmath24 at zero energy . obviously , for states with a higher value of @xmath25 , the potential curve is higher , particularly near the origin ; states with a higher value of @xmath26 will merge into the continuum for larger values of @xmath27 . another way of looking at the properties of the wave function as these solutions merge into the continuum is to find asymptotic solutions outside the well when @xmath28 : @xmath29note that the wave function disappears on a - sites and @xmath30 becomes non - normalizable when @xmath31 . as we will see from the following , the dirac approximation for graphene gives correct predictions for electronic properties at the critical coupling strength , because the energies involved are right at the boundary of the valence band , where the linear dispersion is most accurate . the hamiltonian of a free electron in the two - dimensional graphene lattice , using the tight binding model is , @xmath32where @xmath33 is the creation operator of an electron on the a - atom site labeled @xmath26 in the honeycomb lattice , and @xmath34 represents the annihilation of an electron on the neighboring b - atom site labeled @xmath35 . here @xmath36 denotes the three vectors that connect an a - atom site to its three nearest neighboring b - atom sites . the parameters @xmath37 and @xmath38 represent the nearest neighbor hopping probability and the mass differentiating the @xmath10 and @xmath11 sublattices , respectively . the hamiltonian in k - space can be written as : @xmath39where we have adopted the standard spinor notation for the @xmath10 and @xmath11 sublattice components of the wave function . here , @xmath40 , where @xmath5 is the distance between neighboring atoms . the eigenvalues are @xmath41 where @xmath42 and @xmath43 . the green functions in k - space can be obtained straightforwardly as @xmath44,\]]where @xmath45 ( with @xmath46 an integer ) are the fermion matsubara frequencies ( @xmath47 is the temperature ) , and each component is given by @xmath48 we have added @xmath49 , the chemical potential , for completeness , and the superscript ` 0 ' serves to remind us that these green functions are applicable to the clean lattice , i.e. without impurity scattering . the lattice green functions in real space can be obtained by fourier transform from the above green s functions:@xmath50 , \label{greenarray_real}\]]where @xmath51 @xmath52 the remaining components are readily obtained through the relations @xmath53 and , for the off - diagonal components , @xmath54^{\ast } $ ] . through the analytic continuation , @xmath55 , we obtain the green functions slightly above the real axis , corresponding to the retarded green function . for the particular case of @xmath56 ( an on - site green function ) , we can obtain the diagonal components analytically . from now on we set t=1 , which means that all energies are measured in units of the hopping energy . the result for @xmath57 is ( we set @xmath58 for simplicity , and use the definition @xmath59 ) : \(i ) for @xmath60 , @xmath61 & = & -\frac{(\omega + m)}{\pi } \frac{2}{% \sqrt{\left ( \sqrt{e^{2}+1}\right ) ^{3}\sqrt{e^{2}+9 } } } \\ & & \times f(\frac{\pi } { 2},\frac{1}{2}\sqrt{\frac{-(e^{4}+12e^{2}-6)}{\left ( \sqrt{e^{2}+1}\right ) ^{3}\sqrt{e^{2}+9}}+2 } ) , \\ \mathrm{im}[g_{aa}^{0}(l , l,\omega ) ] & = & 0.\end{aligned}\]](ii ) for @xmath62@xmath61&=&-\frac{(\omega + m)}{\pi } \frac{4}{% \sqrt{3-e}\left [ \sqrt{\left ( e+1\right ) } \right ] ^{3 } } \\ & & \times f(\frac{\pi } { 2},\sqrt{\frac{\left [ 3+e\right ] \left ( 1-e\right ) ^{3}}{\left [ 3-e\right ] \left ( e+1\right ) ^{3}}}),\end{aligned}\]]@xmath63&=&-\frac{2(\omega + m)}{\pi } \frac{1}{% \sqrt{3-e}\left [ \sqrt{\left ( e+1\right ) } \right ] ^{3 } } \\ & & \times f(\frac{\pi } { 2},\sqrt{\frac{16e}{\left [ 3-e\right ] \left ( e+1\right ) ^{3}}}).\end{aligned}\]](iii ) for @xmath64@xmath61 & = & \frac{(\omega + m)}{2\pi } \frac{1}{% \sqrt{e } } \times f(\frac{\pi } { 2},\sqrt{\frac{\left ( e+3\right ) \left ( e-1\right ) ^{3}}{16e } } ) , \\ \mathrm{im}[g_{aa}^{0}(l , l,\omega ) ] & = & -\frac{(\omega + m)}{2\pi } \frac{1}{% \sqrt{e}}f(\frac{\pi } { 2},\sqrt{\frac{\left [ 3-e\right ] \left ( e+1\right ) ^{3}}{16e}}).\end{aligned}\]](iv ) for @xmath65 @xmath61&=&\frac{2(\omega + m)}{\pi } \frac{1}{% \sqrt{\left ( e+3\right ) \left ( e-1\right ) ^{3 } } } \\ & & \times f(\frac{\pi } { 2},\sqrt{\frac{16e}{\left ( e+3\right ) \left ( e-1\right ) ^{3 } } } ) , \\ \mathrm{im}[g_{aa}^{0}(l , l,\omega ) ] & = & 0.\end{aligned}\ ] ] in these expressions we have used @xmath66 , which is a complete elliptic integral of the first kind ( also denoted as @xmath67).@xcite we start with the equation @xmath68 where @xmath69 denotes a matrix where different rows ( and columns ) correspond to different lattice sites ( an explicit example below will make this clearer , see also ref . ) . as an example , we consider the specific case with two on - site impurities located at the sites labeled @xmath70 and @xmath71 ( without loss of generality , we number the first atom on the @xmath10 sublattice as @xmath70 , and we number the first atom on the @xmath11 sublattice as @xmath71 ; since the lattice is bi - partite , @xmath10-atoms are denoted by even numbers , and @xmath11-atoms are denoted by odd numbers ) . the @xmath72 matrix is then written explicitly as @xmath73{5}\end{pmatrix},\ ] ] where now the subscripts refer to the two site indices ( previously written as arguments in , say , eq . ( [ g_int1 ] ) ) , and @xmath6 is the strength of the impurity potential at both sites . then @xmath74 where @xmath75 is a cofactor in the matrix ( [ i - g_0v ] ) , and @xmath76 is the determinant of @xmath77 . the factor @xmath78 is @xmath79 times the determinant of the original matrix excluding the @xmath80-th row and @xmath26-th column . ( [ g ] ) can be expressed as : @xmath81 { \large { / } \delta}. \notag\end{aligned}\ ] ] here the number @xmath82 is given by the number of sites occupied by an impurity . for @xmath83 ( away from the impurities ) , @xmath84 in eq . ( [ g_exp ] ) . in this case ( [ g_exp ] ) becomes : @xmath85 { \large { / } \delta}.\ ] ] the cofactor @xmath86 is equal to @xmath76 , and therefore @xmath87 { \large { / } \delta}.\ ] ] note that when @xmath88 , @xmath89 . when there is only one impurity at any @xmath10-atom site , the hamiltonian is given by @xmath90 where @xmath91 . the corresponding green functions corresponding to the two hamiltonians are @xmath92 and @xmath93 by using the t - matrix expansion , we obtain the green function in the presence of a single impurity : @xmath94 the local density of states ( ldos ) at any position on the graphene lattice is defined by the imaginary part of green function : @xmath95 and we have restored the explicit frequency dependence for clarity . the local density of states at the impurity site is @xmath96 the position of the bound state in the gap is determined by the solution of the equation @xmath97 . at the lower band edge where @xmath98 @xmath99 . therefore , inspection of the above equation suggests that no solution exists , unless @xmath100 . this observation implies that for any @xmath6 the bound state will not merge into the lower continuum , i.e. no bound state energy crosses the edge at @xmath101 . for a single impurity on a @xmath11-atom site , the same remarks apply for a positive impurity potential , and the upper band edge plays the role previously played by the lower band edge . to understand how the bound state approaches the continuum band edge , we use the asymptotic expansion of the complete elliptic integral of the first kind @xcite to get @xmath102where @xmath103 by expanding the green function near @xmath104 we obtain a pole with spectral weight @xmath105 where @xmath106 is the solution of@xmath107 it is clear that as @xmath108 a solution will only occur as @xmath109 , and the residue corresponding to that solution approaches zero . at the other extreme , for a very weak ( negative ) impurity potential , a similar expansion near @xmath110 gives a bound state energy asymptotically approaching the upper band edge ( let @xmath111 ) : @xmath112 the spectral weight approaches zero here as well , as @xmath113 , which also goes to zero as the upper band edge is approached . to summarize the results of this section , we showed that , as the ( negative ) impurity potential decreases from zero towards negative infinity , the frequency of the pole migrates from @xmath114 ( upper band edge ) to @xmath115 ( lower band edge ) . as this occurs , the spectral weight first starts from zero , grows to some maximum , and then decreases again to zero , as the strength of the potential varies from zero to negative infinity . we now consider the two - impurity case , with one on an @xmath10-site , @xmath116 , and the second on a @xmath11-site * i*(i@xmath117,i@xmath118 ) . the hamiltonian is @xmath119where @xmath120 as in the single impurity case . the green functions @xmath121 , @xmath122 and @xmath123 correspond to @xmath124 and @xmath125 , respectively . the t - matrix for this case is @xmath126therefore , the green function becomes @xmath127 in fact , for the many - impurity case , the t - matrix method can be used in a recursive way , @xmath128 where@xmath129 to compute the local density of states at site @xmath116 , we need @xmath130 ( for simplicity we suppress the @xmath131 ) : @xmath132 + v_{2}|g_{01}^{0}(\omega ) |^{2}}{\left [ 1-v_{2}g_{11}^{0}(\omega ) \right ] % \left [ 1-vg_{00}^{0}(\omega ) \right ] -vv_{2}|g_{01}^{0}(\omega ) |^{2 } } } \nonumber \\ & = & \frac{f(\omega ) } { ( \omega -\omega _ { 0})+i\delta } , \label{3.6}\end{aligned}\]]where @xmath133 is the energy of the pole , and @xmath134 accounts for the remaining ( non - singular ) frequency dependence . the actual pole position is the solution of @xmath135 \left [ 1-vg_{00}^{0}(\omega _ { 0})\right ] = vv_{2}|g_{01}^{0}(\omega _ { 0})|^{2 } , \label{3.7}\ ] ] and the spectral weight is given by @xmath136 . then the green function in the case @xmath137 is given by @xmath138keeping the leading term for @xmath139 , we find @xmath140@xmath141where the quantity @xmath142 is finite near @xmath143 , and we get the result near the bottom of the upper band : @xmath144the definitions of @xmath145 are the same as in the case of single impurity scattering discussed in the previous section . the conclusion is the same : as the strength of the impurity interaction increases , the pole moves towards the top of the bottom band , but never crosses it . instead , the residue associated with the pole decreases to zero . the key difference with the coulomb case is that these are short range impurities , and this leads to qualitatively different behaviour . in this subsection we generalize to some extent the results obtained in the previous sections . we find the long - range asymptotic behavior of the green function in the case of multiple impurities located inside a finite area of the graphene sheet . as a particular case this discussion includes the circular well , discussed in the dirac approximation , at the beginning of the article in section ii . knowing these green functions asymptotes we show that the spectral weight of the state near the band edge ( on the verge of entering into the lower continuum ) is zero and the screening charge is not significantly reshaped by this state.@xcite we rewrite eq . ( [ g_exp2.5 ] ) : @xmath146 { \large { / } \delta}. \label{4.1}\ ] ] here we introduced the following notation : @xmath147 is the distance from the center of the area in which the impurities are confined ; we will call this area the potential well " ; * r * corresponds to the site index outside the well ; @xmath5 is the radius of the well ; @xmath148 and @xmath149 are the distances inside the circle of radius @xmath5 , and * r * is the index of the site inside the well . the second term in eq . ( [ 4.1 ] ) represents the change induced by the impurity and is responsible for the spectral weight of the bound state . assuming @xmath150 , the following conclusions can be made about the second term . @xmath76 does not depend on @xmath147 , but @xmath151 depends on @xmath147 , and @xmath152 , as the determinant @xmath151 contains only one row with @xmath153 . considering the sum over @xmath154 in @xmath155 we see that the @xmath148 dependency comes only from the phase @xmath156 in the exponent of the integrands ( [ g_int1 ] , [ g_int2 ] ) . at the energies near the band edge @xmath157 , the part @xmath158 can be neglected , as the small @xmath80 s produce most of the integral value and @xmath159 does not vary significantly near the dirac point . therefore , @xmath160 the spatial dependance of the second term in @xmath161 is determined by the @xmath162 , where @xmath163 denotes some ( arbitrary chosen ) site in the impurity - occupied area . to determine the asymptotic behaviour of @xmath164 , we use the method of stationary phase ( see problem 5.2 in @xcite ) , and get @xmath165where @xmath166 . thus @xmath167 , in qualitative agreement with the asymptotic behavior predicted by the dirac equation ( [ solo_asymp ] ) . as we can see from the standard definition of the green function @xcite : @xmath168 when @xmath169 the @xmath148-dependency of @xmath170 coincides with @xmath171 and we can judge if the state that is potentially crossing the band edge into the continuum is normalizable . in our case as @xmath172 the sum of @xmath161 over @xmath173 in the plane diverges , and so does the sum @xmath174 in the infinite lattice for any finite normalizing factor . hence the state merging into the continuum can be called non - normalizable or extended and as such has zero spectral weight ( in the thermodynamic limit ) . this confirms our conclusion that there is no such phenomenon like supercritical screening in the case of a localized potential ( as opposed to a coulomb potential ) in graphene . our results are in agrement with common intuition developed in the physics of shallow states in semiconductors . in particular it is not only a feature peculiar to gapped graphene that properties of the states near the band edge are strongly dependent on the long range tail " of the impurity potential.@xcite in general , the effective mass approach ( mostly determined by long - range properties of a system ) is in good agreement with exact numerical methods for the energies near the band edge . it is exactly near the band edge where the coulomb tail becomes important while it is negligible in computations related to deep levels.@xcite as illustrated in section ii , in the continuum limit a potential barrier emerges in the effective potential at large distances , due to the squaring of the coulomb potential . because of the two sublattices , a similar ` squaring ' occurs when the problem is solved on a lattice , and lattice green functions are utilized.@xcite

we show that the changes in the electronic density of states ( dos ) in graphene induced by localized impurities ( single , double and multiple ) are significantly different from those caused by the long - range coulomb potential . we focus on gapped graphene ; a bound state is present within the gap , with a certain amount of spectral weight . as the coupling to the impurity increases , the state lowers in energy and approaches the lower continuum valence band . the spectral weight of this state does not transform into a resonance state in the valence band , so no unusual screening effects related to a redistribution of dos in the continuum is observed . in terms of the continuous dirac limit , this phenomenon is a consequence of the absence of the potential bump " at infinity , which is present in the potential of the effective schrdinger equation for graphene with long - range coulomb impurity potential . the states induced by short - range impurity scattering in graphene , therefore , have distinctly different properties compared with the long - range potential case . these properties closely resemble the case of a short - range single impurity in other bipartite lattices , such as square , body - centered cubic , and simple cubic lattices . for these bipartite lattices , there is always a localized bound state with energy in the band gap for the entire range of on - site coupling strengths . in all cases the energy of these states asymptotically approaches the edge of the valence band as the magnitude of the coupling strength increases , but never crosses it .

introduction dirac equation with spherical well analytical results for one and two impurities on the lattice two or more impurity scattering conclusions

under physiological conditions , proteins undergo everlasting motions , ranging from atomic thermal fluctuation , side - chain rotation , residue swiveling , to domain swirling . protein motion strongly correlates with protein functions , including molecular docking @xcite , drug binding @xcite , allosteric signaling @xcite , self assembly @xcite and enzyme catalysis @xcite . the range of protein motions in a cellular environment depends on the structure s local flexibility , an intrinsic property of a given protein structure . protein flexibility is reflected by the debye - waller factor ( b - factor ) , i.e. , the atomic mean - square displacement , obtained in structure determination by x - ray crystallography , nmr , or single - molecule force experiments @xcite . however , the b - factor can not absolutely quantify flexibility : it also depends the crystal environment , solvent type , data collection condition and structural refinement procedure @xcite . the flexibility of a biomolecule can be assessed by molecular dynamics ( md ) @xcite , normal mode analysis ( nma ) @xcite , graph theory @xcite and elastic network models ( enms ) @xcite , including gaussian network model ( gnm ) @xcite and anisotropic network model ( anm ) @xcite . nma can be regarded as time - independent molecular dynamics ( md ) @xcite and diagonalizes the md potential to obtain a set of eigenvalues and eigenvectors , where first few eigenvectors predict the collective and global motions , which are potentially relevant to biomolecular functionality . nma with only the elasticity potential , which is a leading term in the md potential , was introduced by tirion @xcite , and was extended to the network setting in anm @xcite . here network refers to the connectivity between particles regardless of their chemical bonds @xcite . yang et al . @xcite demonstrated that due to its network setting , gnm is about one order more efficient than most other flexibility approaches in computational complexity . in terms of b - factor prediction , gnm is typically more accurate than anm @xcite . therefore , gnm has been widely used in the study of biomolecular structure , dynamics and functions @xcite . it has demonstrated its utilities in stability analysis @xcite , docking simulation @xcite , viral capsids @xcite and domain motions of hemoglobin @xcite , f1 atpase @xcite , chaperonin groel @xcite and the ribosome @xcite . in traditional elasticity network models , i.e. , enm , gnm and anm , the connectivity is determined by a fixed cutoff distance . all atoms within the cutoff distance are treated equally with no consideration given to effects which scale with distance . in this manner , the behavior of these methods typically depends on the cutoff distance used . many modified models are proposed to improve the hard cutoff distance practice by the incorporation of the distance information @xcite . hinsen has changed the spring constant to a distance - dependent gaussian function @xcite . riccardi et al . have used an inverse 6th power function of distance as the spring coefficient in their elastic network model @xcite . a parameter - free anm has been introduced by using the inverse 2nd power square distance for spring constant @xcite . a common feature of all the aforementioned approaches is that , they all depend on the mode decomposition of the potential matrix , which typically has the computational complexity of @xmath0 , where @xmath1 is the order the potential matrix . to bypass the matrix diagonalization , researchers explore the flexibility properties using the local packing density . many elegant methods and algorithms have been proposed , including the local density model ( ldm ) by halle @xcite , the local contact model ( lcm ) by zhang et al @xcite , weighted contact number ( wcn ) by @xcite , and others @xcite . recently , we have proposed a few mode - decomposition free methods for flexibility analysis , namely , molecular nonlinear dynamics @xcite , stochastic dynamics @xcite and flexibility - rigidity index ( fri ) @xcite . among them , the fri method is of @xmath2 in computational complexity and has been accelerated to @xmath3 by using the cell lists algorithm without loss of accuracy @xcite . the essential idea of the fri method is to evaluate the rigidity index or the compactness of the biomolecular ( network ) packing by the total correlation , a function of inter - atomic distance . then the flexibility index is defined as the inverse of the rigidity index . although the original motivation for fri comes from the design of continuum elasticity with atomic rigidity ( cewar ) , fri shares a similarity with the ldm , lcm and wcn . to be more specific , all of these methods make use of local packing information and are free from matrix diagonalization . however , significant distinction exists between our fri methods and other local density based methods and it can be summarized as the following . firstly , other than the discrete flexibility index and discrete rigidity index , our fri methods delivers a continuous flexibility function and a continuous rigidity function @xcite . the continuous rigidity function , which can be regarded as the density distribution function ( density estimator ) of a biomolecule , plays many important roles beyond the scope of flexibility study @xcite . for instance , it can be used to generate biomolecular surface representations @xcite , which reduce to the gaussian surface if an appropriate kernel is used . in fact , rigidity function can be applied to decipher the atomic information from the experimental electron density data @xcite . secondly , protein multiscale collective motions can be captured by using multiple kernels in our fri method , called multiscale fri or multikernel fri ( mfri ) @xcite . this approach significantly improve the accuracy of fri b - factor predictions . thirdly , we proposed an anisotropic fri ( afri ) model for the evaluation of biomolecular global motions . different from traditional normal mode analysis or anm , our afri allows adaptive representations , ranging from a completely global description to a completely local representation @xcite . the objective of the present work is twofold . first , we propose a unified framework to construct generalized gnms ( ggnms ) . we reveal that the gnm kirchhoff matrix can be constructed from the ideal low - pass filter ( ilf ) , which is the limiting case of admissible fri correlation functions . we demonstrate that fri and ggnm are asymptotically equivalent when the cutoff value in the kirchhoff matrix or the scale value in the correlation function is sufficiently large . this finding paves the way for understanding the connection between the gnm and fri methods . to clarify this connection , we introduce a generalized kirchhoff matrix to provide a unified starting point for the ggnm and fri methods , which elucidates on the similarity and difference between ggnm and fri . based on this new understanding of the ggnm working principle , we propose infinitely many correlation function based ggnms . we show that ggnm outperforms the original gnm for the b - factor prediction of a set of 364 proteins . both ggnm and fri deliver almost identical results when the scale parameter is sufficiently large . our approach sheds light on the construction of efficient ggnms . additionally , we propose two new methods , multiscale gnm ( mgnm ) and multiscale anm ( manm ) , to account for the multiscale features of biomolecules . most biomolecules , particularly large macromolecules and protein complexes , have multiple characteristic length scales ranging from covalent bond , residue , secondary structure and domain dimensions , to protein sizes . even for small molecules , due to the influence of crystal structure , multiscale effects play a significant role in atomic thermal fluctuations . consequently , gnm and anm , which are typically parametrized at a single cutoff distance , often do not work well in characterizing the flexibility of molecules involving multiscale behaviors . our essential idea is to generalize original gnm and anm into a multikernel setting so that each kernel can be parametrized at a given characteristic length . this generalization is achieved through the use of an fri assessment , which predicts the involvement of different scales , followed by an appropriate constructions of multikernel gnm or multikernel anm . this approach works because for a diagonally dominant matrix , the direct inverse of the diagonal element is essentially equivalent to the diagonal element of the inverse matrix . we demonstrate that the proposed mgnm and manm are able to successfully capture the multiscale properties of the protein and significantly improve the accuracy in protein flexibility prediction . the rest of this paper is organized as the follows . section [ sec : method ] is devoted to methods and algorithms . we first propose a concise formulation of ggnms using fri correlation functions in section [ sec : gnm ] . we show that there are infinitely many new ggnms that reduce to the original gnm at appropriate limits of their parameters . to establish notation , we further present a brief review of our mfri formalism in section [ sec : mfri ] . based on the connection between fri and gnm , we propose mgnm in section [ sec : mgnm ] . specifically , parameters learned from mfri are used to construct the multiscale kirchhoff matrix in mgnm . we discuss two types of realizations of mgnms . as an extension of our mgnm , the manm method is introduced in section [ sec : manm ] . we validate the proposed ggnm , mgnm and manm by extensive numerical experiments in section [ sec : result ] . we illustrate that the intrinsic multiscale properties of biomolecules are successfully captured in our mgnm and manm . finally , in section [ sec : application ] , we demonstrate the utility of mgnm and manm for protein flexibility analysis , protein domain separation and collective motion study . the present work offers a new strategy for the design and construction of accurate , efficient and robust methods for biomolecular flexibility analysis . this paper ends with a conclusion . to establish notation and facilitate new development , let us present a brief review of the gnm and fri methods . consider an @xmath1-particle coarse - grained representation of a biomolecule . we denote @xmath4 the coordinates of these particles and @xmath5 the euclidean space distance between @xmath6th and @xmath7th particles . in a nutshell , the gnm prediction of the @xmath6th b - factor of the biomolecule can be expressed as @xcite @xmath8 where @xmath9 is a fitting parameter that can be related to the thermal energy and @xmath10 is the @xmath6th diagonal element of the matrix inverse of the kirchhoff matrix , @xmath11 where @xmath12 is a cutoff distance . the gnm theory evaluates the matrix inverse by @xmath13_{ii}$ ] , where @xmath14 is the transpose and @xmath15 and @xmath16 are the @xmath17th eigenvalue and eigenvector of @xmath18 , respectively . the summation omits the first eignmode whose eigenvalue is zero . the fri prediction of the @xmath6th b - factor of the biomolecule can be given by @xcite @xmath19 where @xmath9 and @xmath20 are fitting parameters , @xmath21 is the @xmath6th flexibility index and @xmath22 is the @xmath6th rigidity index . here , @xmath23 is an atomic number depended weight function that can be set to @xmath24 for a c@xmath25 network , and @xmath26 is a real - valued monotonically decreasing correlation function satisfying the following admissibility conditions @xmath27 where @xmath28 is a scale parameter . delta sequences of the positive type @xcite are good choices . many radial basis functions are also admissible @xcite . commonly used fri correlation functions include the generalized exponential functions @xmath29 and generalized lorentz functions @xmath30 a major advantage of the fri method is that it does not resort to mode decomposition and its computational complexity can be reduced to @xmath3 by means of the cell lists algorithm used in our fast fri ( ffri ) @xcite . in contrast , the mode decomposition of nma and gnm has the computational complexity of @xmath0 . [ cols="^ " , ] gnm is an isotropic model which quantifies the atomic scalar fluctuations in molecule . in contrast , anm is designed to describe the anisotropic properties , such as collective motions of a molecule near the equilibrium . typically , the first six modes , corresponding to six zero ( or near zero ) eigenvalues , represent the trivial translational and rotational modes of a complex biomolecule . global modes that are unique to the biomolecular structure are described by eigenvectors associated with the nonzero ( next smallest ) eigenvalues . due to its simplicity , anm is widely used to study the dynamics of biomolecules . in the present work , we have designed our manm to maintain the aforementioned properties . to validate our manm for anisotropic mode analysis , we use two test proteins , i.e. , 1gru ( chain a ) and 1urp ( chain a ) . the protein 1gru is chaperonin groel , a benchmark test for anm @xcite . we employ our manm with two gaussian kernels ( @xmath31 ) with @xmath32 and @xmath33 . we compute eigenvectors associated with the first three nonzero eigenvalues . as illustrated in figure [ fig : manm_1gru ] , our manm results are in an excellent agreement with those of anm for chaperoin groel @xcite . to further validate our manm , we examine another test case , 1urp . it is a ribose - binding protein and its anisotropic motions have been studied in the literature @xcite . we utilize the same set of parameters described above . figure [ fig : manm_1urp ] demonstrates manm results . our results are in a close consistence with the traditional anm analysis @xcite . gaussian network model ( gnm ) and anisotropic network model ( anm ) are popular methods for macromolecular flexibility analysis . alternative methods , flexibility - rigidity index ( fri ) @xcite and anisotropic flexibility - rigidity index ( afri ) @xcite , have been introduced to achieve better accuracy and more adaptivity in our recent work . most recently , we have further proposed multiscale flexibility - rigidity index ( mfri ) @xcite to capture the multiscale behavior in macromolecules . our mfri utilizes multiple kernels which are parametrized at different scales to describe macromolecular multiscale connectivity . we have shown that mfri is about 20% more accurate than gnm in the b - factor prediction of a set of 364 protein @xcite . motivated by these achievements , we propose a few fri based generalizations of gnm and anm in this work . first , we construct a series of generalized gaussian network models ( ggnms ) . we show that the original kirchhoff matrix used in gnm can be constructed by using the ideal low - pass filter ( ilf ) , which is a special case of a family of admissible correlation kernels ( or functions ) used in fri . based on this connection , we propose a unified framework to construct generalized kirchhoff matrices for both gnm and fri . more specifically , the inverse of the generalized kirchhoff matrices leads to infinitely many ggnms and the direct inverse of the diagonal terms gives rise to fri . we reveal the identical behavior between ggnm and fri at a large cutoff distance or characteristic scale for b - factor protein predictions . additionally , we propose multiscale gaussian network models ( mgnms ) based on the relationship of gnm and fri . essentially , we develop a two - step procedure to construct mgnms . in the first step , we utilize mfri to come up with an optimal combination of multiscale kernels . in our second , we try to implement the same combination of multiscale kernels in the generalized kirchhoff matrices for mgnms . however , this step is not unique because for a given kirchhoff matrix , gnm and fri are connected only through diagonal elements . two types schemes , type-1 mgnm and type-2 mgnm , are proposed in this work . moreover , we propose multiscale anisotropic network models ( manms ) based on the similarity between anm and gnm and the connection between gnm and fri . since anm is typically less accurate than gnm in b - factor prediction @xcite , its main utility is for collective motion analysis . we therefore have developed manms to maintain the physical connectivity of protein atoms in the kirchhoff matrix . we have carried out intensive numerical experiments to validate the proposed ggnm , mgnm and manm methods for b - factor predictions . the ggnm method is examined over a set of 364 proteins . it is found that the proposed ggnm is about 10% more accurate than gnm in b - factor prediction . for mgnm , we use only a set of 362 protein due to limited computer resource . we show that mgnm can achieve about 13% improvement over gnm . similarly , the proposed manm is about 11% more accurate than its counterpart , anm , in b - factor prediction over a set of 300 proteins . further , we consider three types of applications of the proposed mgnm and manm methods . one type of application is to analyze the flexibility of proteins that fail the original gnm method in various ways . we employ four proteins to demonstrate the advantage of the proposed mgnm in flexibility analysis . another application is the study of protein domain separations . the first nontrivial eigenmode of the multiscale kirchhoff matrix is used . we found from the analysis of two proteins that type-1 mgnm does a good job in domain analysis while type-2 mgnm does not work for this purpose . the other application concerns the protein collective motions . our manm is found to offer similar results as those of the original anm method . in the future , we will further apply our manm to study the anisotropic b - factor @xcite and conformational change @xcite . it is worth to pointing that our mgnm and manm methods are not unique . how to design optimal new mgnm and manm methods is still an open problem . essentially , we hope these new methods are efficient , accurate and robust . more specifically , high accuracy in b - factor prediction is a main criterion . additionally , having the ability to provide correct protein domain analysis is a desirable property as well . for manm , the capability of offering correct motion analysis is a major requirement . the quality of both domain and motion analyses depends on how to design non - diagonal matrix elements so as to properly reflect the physical connectivity among particles . this work was supported in part by nsf grants iis-1302285 , and dms-1160352 , and nih grant r01gm-090208 . the authors thank jianyu chen and michael feig for useful discussions . h. y. li , z. x. cao , l. l. zhao , and j. h. wang . analysis of conformational motions and residue fluctuations for escherichia coli ribose - binding protein revealed with elastic network models . , 14(5):1055210569 , 2013 . f. tama , m. valle , j. frank , and c. k. brooks iii . dynamic reorganization of the functionally active ribosome explored by normal mode analysis and cryo - electron microscopy . , 100(16):9319 9323 , 2003 . e. eyal , c. chennubhotla , l. w. yang , and i. bahar . anisotropic fluctuations of amino acids in protein structures : insights from x - ray crystallography and elastic network models . , 23(13 ) , i175-i184 , 2007 .

gaussian network model ( gnm ) and anisotropic network model ( anm ) are some of the most popular methods for the study of protein flexibility and related functions . in this work , we propose generalized gnm ( ggnm ) and anm methods and show that the gnm kirchhoff matrix can be built from the ideal low - pass filter , which is a special case of a wide class of correlation functions underpinning the linear scaling flexibility - rigidity index ( fri ) method . based on the mathematical structure of correlation functions , we propose a unified framework to construct generalized kirchhoff matrices whose matrix inverse leads to ggnms , whereas , the direct inverse of its diagonal elements gives rise to fri method . with this connection , we further introduce two multiscale elastic network models , namely , multiscale gnm ( mgnm ) and multiscale anm ( manm ) , which are able to incorporate different scales into the generalized kirchkoff matrices or generalized hessian matrices . we validate our new multiscale methods with extensive numerical experiments . we illustrate that ggnms outperform the original gnm method in the b - factor prediction of a set of 364 proteins . we demonstrate that for a given correlation function , fri and ggnm methods provide essentially identical b - factor predictions when the scale value in the correlation function is sufficiently large . more importantly , we reveal intrinsic multiscale behavior in protein structures . the proposed mgnm and manm are able to capture this multiscale behavior and thus give rise to a significant improvement of more than 11% in b - factor predictions over the original gnm and anm methods . we further demonstrate benefit of our mgnm in the b - factor predictions on many proteins that fail the original gnm method . we show that the present mgnm can also be used to analyze protein domain separations . finally , we showcase the ability of our manm for the simulation of protein collective motions . key words : multiscale elastic network model , multiscale gaussian network model , multiscale anisotropic network model , flexibility - rigidity index , debye - waller factor .

introduction methods and algorithms conclusion acknowledgments

a basic issue in social dynamics is to understand how opinion diversity arises when interactions between individuals are primarily `` ferromagnetic '' in character . many kinetic spin models have been proposed to address this general question @xcite . an important example of this genre is the appealingly simple axelrod model @xcite , which accounts for the formation of distinct cultural domains within a population . in the axelrod model , each individual is endowed with a set of features ( such as political leaning , music preference , choice of newspaper , _ etc . _ ) , with a fixed number of choices for each feature . evolution occurs by the following voter - model - like update step @xcite . a random individual is picked and this person selects an interaction partner ( anybody in the mean - field limit , and a nearest - neighbor for finite spatial dimension ) . for this pair of agents a feature is randomly selected . if these agents have the same state for this feature , then another feature is picked and the initial person adopts the state of this new feature from the interaction partner . this dynamics mimics the feature that individuals who share similar sentiments on lifestyle issues can have a meaningful interaction in which one will convince the other of a preference on an issue where disagreement exists . depending on the number of traits and the number of states per trait , a population may evolve to global consensus or it may break up into distinct cultural domains , in which individuals in different domains do not have any common traits @xcite . this diversity is perhaps the most striking feature of the axelrod model . an even simpler example with a related phenomenology is the bounded compromise model @xcite . here , each individual possesses a single real - valued opinion that evolves by compromise . in an update step , two interacting individuals average their opinions if their opinion difference is within a pre - set threshold . however , if this difference is greater than the threshold , there is no interaction . these steps are repeated until the system reaches a final state . for a sufficiently large threshold the final state is consensus , while for a smaller threshold the system breaks up into distinct opinion clusters ; these are analogous to the cultural domains of the axelrod model . in spite of the simplicity of these models , most of our understanding stems from simulation results . as a first step toward analytic insight , a discrete three - state version of the bounded compromise model was recently introduced @xcite . this is perhaps the simplest opinion dynamics model that includes the competing features of consensus and incompatibility . this model was found to exhibit a variety of anomalous features in low dimensions , including slow non - universal kinetics and power - law spatial organization of single - opinion domains . in this work , we focus on the ultimate fate of this system in the mean - field limit . we determine the exact probability that the final state of the system is either consensus or a frozen mixture of leftists and rightists as a function of the initial composition of the system . we also compute the time required for the system to reach its ultimate state . in the next section , we define our model , outline its basic properties , and determine the ultimate fate of the system in terms of an equivalent first - passage process . the solution to this problem is used to obtain the probabilities of reaching either a frozen final state ( sec . iii ) or consensus ( sec . iv ) . in sec . v , we compute the mean time until the final state is reached . in the discussion section , we generalize to non - symmetric interactions and also show how our results can be adapted to determine the ultimate fate of a version of the axelrod model . calculational details are presented in the appendices . we consider a population of @xmath0 individuals , or agents , that are located at the nodes of a graph . each agent can be in one of three opinion states : leftist , centrist , or rightist . as shown in fig . [ steps ] , we represent these states as @xmath1 , and @xmath2 , respectively . in a single microscopic event an agent is selected at random . we consider the mean - field limit in which the neighbor of an agent can be anyone else in the system . if the two agents have the same opinion , nothing happens . if one is a centrist and the other is an extremist ( or vice versa ) , the initial agent adopts the opinion of its neighbor that is , each individual can be viewed as having zero self - confidence and merely adopts the state of a compatible neighbor ; this kinetic step is the same as in the classical voter model @xcite . however , if the two agents are extremists of opposite persuasions , they are incompatible and do not influence each other . as a result of this incompatibility constraint , the final state of the system can be either consensus of any of the three species , or a frozen mixture of leftist and rightist extremists , with no centrists . in a mean - field system with @xmath3 leftists , @xmath4 rightists , and @xmath5 centrists ( with @xmath6 ) , the probability of selecting a pair @xmath7 ( @xmath8 ) equals @xmath9 $ ] , where @xmath10 is the number of agents of type @xmath11 . then the elemental update steps and their respective probabilities are : @xmath12 while the probability for no change is @xmath13 . . the heavy dots denote consensus states and the heavy line denotes the frozen final states where no centrists remain . typical random walk trajectories are shown , along with their projections ( dashed ) on the @xmath14 plane . when a random walk reaches the lines @xmath15 or @xmath16 , the random walk subsequently must remain on this line until consensus is reached . the corresponding hopping probabilities on the @xmath14 plane are also indicated.,scaledwidth=40.0% ] eventually the system reaches a static final state that is either consensus of one of the three species or a frozen mixture of leftists and rightists . monte carlo simulations showed that the nature of the final state has a non - trivial dependence on the initial densities of the species @xcite . we now analytically determine this final state probability as a function of the initial population composition by solving an equivalent first - passage problem . as shown in fig . [ 3d - triangle ] , the state of the system corresponds to a point in the space of densities @xmath17 , @xmath18 , and @xmath19 . since @xmath20 , this constraint restricts the system to the triangle @xmath21 shown in the figure . when two agents interact , the state of the system may change and we can view this change as a step of a corresponding random walk on the triangle @xmath21 , with single step hopping probabilities given in eq . ( [ hops ] ) . when the walk reaches one of the fixed points @xmath22 , @xmath23 , or @xmath24 ( consensus ) , or any point on the fixed line @xmath25 ( frozen extremist mixture ) , the system stops evolving . this set represents absorbing boundaries for the effective random walk . thus to find the probability of reaching a given final state , we compute the first - passage probability for the random walk to hit a given absorbing boundary . we may simplify the problem by using the fact that only two of the densities @xmath26 are independent . we thus choose @xmath27 and @xmath28 as the independent variables by projecting the effective random walk trajectory onto the @xmath14 plane ( fig . [ 3d - triangle ] ) . according to eq . ( [ hops ] ) , this two - dimensional random walk jumps to its nearest neighbors in the @xmath27 and @xmath28 directions with respective probabilities @xmath29 and @xmath30 , and stays in the same site with probability @xmath13 . here @xmath31 , and we consider the limit @xmath32 . we now determine the probability @xmath33 for the system to reach a frozen state when the initial densities are @xmath34 and @xmath35 . this coincides with the first - passage probability for the equivalent random walk to hit the line @xmath36 when it starts at a general point @xmath37 in the interior of the triangle . there are also absorbing boundaries on the sides @xmath15 and @xmath16 , where the probability of reaching the frozen state is zero , and @xmath38 on the line segment @xmath36 . for notational simplicity , we write @xmath39 . in the equivalent random walk process , this first - passage probability obeys the recursion @xcite @xmath40\nonumber \\ & + & p_y[f(x , y-\delta)+f(x , y+\delta ) ] \nonumber \\ & + & [ 1 - 2(p_x+p_y)]f(x , y ) . \end{aligned}\ ] ] that is , the first - passage probability @xmath41 equals the probability of taking a step in some direction ( the factor @xmath42 ) times the first - passage probability from this target site to the boundary . this product is then summed over all possible target sites after one step of the walk . in the continuum limit ( @xmath43 ) we expand this equation to second order in @xmath44 and obtain @xmath45 supplemented with the boundary conditions @xmath46 the solution to eq . ( [ fp_eq ] ) is not straightforward because of the mixed boundary conditions . instead of attacking the problem directly , we transform to the coordinates @xmath47 , @xmath48 to map the triangle to a quarter - circle of unit radius and then apply separation of variables in this geometry . the transformed differential equation is @xmath49 because of the circular symmetry of the boundary conditions in this reference frame , it is now convenient to use the polar coordinates @xmath50 , where @xmath51 and @xmath52 . this transforms eq . ( [ uv_eq ] ) to @xmath53 we seek a product solution @xmath54 . substituting this into eq . ( [ pc ] ) leads to the separated equations @xmath55 where @xmath56 is the separation constant . ( [ r ] ) is equidimensional and thus has the general power - law form @xmath57 where @xmath58 are constants . to solve eq . ( [ theta ] ) , we proceed by eliminating the first derivative to give a schrdinger - like equation . thus we define @xmath59 and find the function @xmath60 that eliminates this first derivative term . substituting @xmath61 in eq . ( [ theta ] ) gives @xmath62u ' \nonumber \\ & + & \left[f''+\left(\tan\theta-\cot\theta\right)f'+m^2f \right]u=0 , \label{sch0}\end{aligned}\ ] ] and the coefficient of @xmath63 is zero if @xmath64 satisfies @xmath65 whose solution is @xmath66 substituting this expression for @xmath64 in eq . ( [ sch0 ] ) , we obtain @xmath67u=0 . \label{sch1}\ ] ] details of the solution to this schrdinger equation are given in appendix a. the final result is @xmath68 where @xmath69 is the associated legendre function . we transform back to the original @xmath70 coordinates through @xmath71 finally , identifying @xmath72 with @xmath73 , the solution in the original cartesian coordinates is @xmath74 this gives the probability of reaching a final frozen state when the population has initial densities @xmath27 and @xmath28 of leftists and rightists , respectively . notice that this probability is symmetric , @xmath75 , because @xmath69 is an even function for @xmath76 odd and an odd function for @xmath76 even . this reflects the obvious physical symmetry that the probability @xmath73 is invariant under the interchange of leftists and rightists . for equal initial densities of leftists and rightists , namely @xmath77 , eq . ( [ sol1 ] ) becomes a function of the initial density of centrists @xmath78 only @xmath79 as shown in appendix b , this can be simplified to the closed - form expression @xmath80 with @xmath81 . this solution is shown in fig . [ solution ] along with monte - carlo simulation results . 0.05 in , and extremist consensus , @xmath82 , as a function of the initial density of centrists @xmath78 for the case of equal initial densities of leftists and rightists , @xmath83 ( top ) and for the case @xmath84 ( bottom).,title="fig:",scaledwidth=42.0%]0.2 in , and extremist consensus , @xmath82 , as a function of the initial density of centrists @xmath78 for the case of equal initial densities of leftists and rightists , @xmath83 ( top ) and for the case @xmath84 ( bottom).,title="fig:",scaledwidth=42.0% ] has magnetization @xmath56 in the frozen final state , for the cases @xmath85 , 0.1 , and 0.2 ( progressively broadening curves).,scaledwidth=40.0% ] in addition to reaching a frozen final state , the system can also reach global consensus either leftist , rightist , or centrist . again , the probabilities for these latter three events depends on the initial composition of the system . we thus define @xmath86 , @xmath87 , and @xmath88 , as the respective probabilities to reach a leftist , rightist , or centrist consensus as a function of the initial densities . the probability for centrist consensus can be obtained by elementary means . we merely map the 3-state constrained system onto the 2-state voter model by considering the leftist and rightist opinions as comprising a single extremist state , while the centrist opinion state maintains its identity . the dynamics of this system is exactly that of the classical 2-state voter model . since the overall magnetization of the voter model is conserved @xcite , the probability that the system reaches centrist consensus , for a given @xmath27 and @xmath28 is simply @xmath89 to find the probability for leftist and rightist consensus , we use normalization of the final state probability @xmath90 and conservation of the global magnetization to write @xmath91 here @xmath92 is the probability of ending in a specific frozen state with density @xmath93 of @xmath94 spins and @xmath95 of @xmath2 spins as a function of the initial densities @xmath27 and @xmath28 . this function therefore satisfies the normalization condition @xmath96 . additionally the integral in eq . ( [ mag - norm ] ) is the final magnetization in the frozen state . from the exact solution for @xmath92 given in eq . ( [ fx ] ) , we thereby obtain as a byproduct the magnetization distribution in the final frozen state ( fig . [ magnet ] ) . as expected , for a small initial density of centrists @xmath78 , there is little evolution before the final state is reached and the magnetization distribution is narrow . using eq . ( [ p0 ] ) and the normalization condition for @xmath92 , we recast eqs . ( [ norm ] ) and ( [ mag - norm ] ) as @xmath97 and @xmath98 subtracting these equations , we obtain @xmath99 now the first - passage probability @xmath92 to a frozen state with a specified density @xmath93 of @xmath94 spins obeys the same differential equation as @xmath73 [ eq . ( [ fp_eq ] ) ] , but with the boundary conditions @xmath100 the last condition states that the first - passage probability to the point @xmath101 on the boundary @xmath36 is zero unless the effective random walk starts at @xmath101 . the first - passage probability @xmath92 has the general form given in eq . ( [ sol0 ] ) , but with the coefficients @xmath102 now determined from the boundary conditions in eq . ( [ bc2 ] ) . from the solution for @xmath92 given in appendix a , the probability of @xmath94 consensus as a function of the initial densities @xmath27 and @xmath28 is @xmath103 the probability of @xmath2 consensus can be obtained by using the fact that @xmath104 . for the special case of @xmath77 , the probabilities @xmath105 can be obtained either by setting @xmath106 in eq . ( [ p+1 ] ) and summing the series , or , more simply , by using eq . ( [ final - sol ] ) for @xmath107 and the probability conservation equation @xmath108 . by either approach , we find the following closed expression for the probability of extremist consensus as a function of @xmath78 @xmath109 this result is also shown in fig . [ solution ] . of fig . [ 3d - triangle ] . the triple points , where the final state probabilities of three of the four phases are equal , are located at @xmath110 and @xmath111.,scaledwidth=35.0% ] from our results for @xmath33 , @xmath87 , @xmath86 , and @xmath88 , we may define a `` phase diagram '' shown in figure [ phase - diag ] . each region in the figure corresponds to the portion of the composition triangle where the probability of ultimately ending up in the labeled state is greater than all other first - passage probabilities . the least likely outcome is the achievement of extremist consensus while the most likely result is to get stuck in the frozen mixed state . in addition to the probability of reaching a particular final state , we also study the mean time until the final state is reached as a function of the initial composition of the system . the simplest such quantity is the _ unconditional _ mean time @xmath112 to reach _ any _ of the four possible final states extremist consensus ( @xmath2 or @xmath94 ) , centrist consensus , and mixed frozen state , as a function of the initial densities @xmath27 and @xmath28 . this first - passage time can again be obtained trivially by considering an equivalent two - state system where we lump leftists and rightists into a single extremist state . the system stops evolving when the density of centrists reaches @xmath14 or @xmath113 , with the former corresponding either to extremist consensus or to a frozen mixed state . to find this first - passage time to reach the final state , note that in a single event the effective one - dimensional random walk that corresponds to the state of the system can either jump to one of its two nearest neighbors with probability @xmath114 or stay at the same site with probability @xmath115 , where @xmath116 . the time interval for each event is @xmath117 , corresponding to each person being selected once on average every @xmath0 update steps . then the mean time to reach the final states @xmath14 or @xmath113 as a function of the initial density @xmath78 obeys the recursion @xcite @xmath118+p_z[t(z-\delta ) + dt ] \\ & + & ( 1 - 2p_z)[t(z)+dt].\end{aligned}\ ] ] this formula has a similar form and a similar explanation as the equation for the first - passage probability [ eq . ( [ fxy ] ) ] . starting from @xmath78 , the mean time to reach the final state equals the probability of taking a single step ( the factors @xmath114 and @xmath115 ) multiplied by the time needed to reach the boundaries via this intermediate site . this path - specific time is just the mean first - passage time from the intermediate site plus the time @xmath119 for the initial step . in the large-@xmath0 limit this recursion reduces to @xmath120 where the diffusion coefficient is @xmath121 . since @xmath122 , we have @xmath123 . ( [ tz ] ) is subject to the boundary conditions @xmath124 , corresponding to immediate absorption if the random walk starts at the boundary . the solution to this equation is @xmath125 . \label{time}\ ] ] this is simply the mean consensus time of the 2-state voter model in the mean - field limit , in which the initial density of the two species are @xmath78 and @xmath126 . for the cases @xmath83 ( top ) and @xmath127 ( bottom ) . the curves for @xmath128 and @xmath129 are based on exact formulae , while the other are based on simulations . the latter results are based on @xmath130 realizations for each initial state . , title="fig:",scaledwidth=40.0%]0.2 in for the cases @xmath83 ( top ) and @xmath127 ( bottom ) . the curves for @xmath128 and @xmath129 are based on exact formulae , while the other are based on simulations . the latter results are based on @xmath130 realizations for each initial state . , title="fig:",scaledwidth=40.0% ] it is more interesting to consider the _ conditional _ first - passage time to a specific final state as a function of the initial condition . for example , consider the conditional time @xmath131 , the mean time to reach centrist consensus when the initial centrist density is @xmath78 . this is the mean time for the equivalent random walk to hit the point @xmath113 , _ without _ hitting @xmath14 . in the large @xmath0 limit , this conditional exit time obeys the differential equation ( see sec . 1.6 in @xcite ) @xmath132= -p_0(z ) . \label{t+ } \end{aligned}\ ] ] integrating eq . ( [ t - cond ] ) , using @xmath133 , and the boundary conditions @xmath134 at @xmath14 and at @xmath113 , we obtain @xmath135 , from which the conditional time to reach centrist consensus is @xmath136 similarly , the mean time to reach all other final states of the system is @xmath137 in principle , we can find the conditional times to reach the extremist consensus states and the frozen final state as a function of the initial condition . this involves solving the two - dimensional analogue of eq . ( [ t - cond ] ) in the density triangle , subject to the appropriate boundary conditions . because of the tedious nature of the calculation , we have instead resorted to numerical simulations to compute these conditional exit times . these results are shown in fig . [ times ] . while all exit times are of the order of @xmath0 , it is worth noting that @xmath138 and @xmath139 are typically the longest times . this stems from the fact that reaching extremist consensus is a two - stage process . first , the extremists of the opposite persuasion must be eliminated , and then there is a subsequent first - passage process in which the centrists are also eliminated . we determined basic properties of the final state in a simple opinion dynamics model that consists of a population of leftists , rightists , and centrists . centrists interact freely with extremists ( either leftist or rightist ) , so that one agent adopts the opinion of its interaction partner , while extremists of opposite persuasions do not interact . these competing tendencies lead either to ultimate consensus or to a frozen final state that consists of non - interacting leftists and rightists . while this model is clearly an oversimplification of opinion evolution in a real population , it provides a minimalist description for how consensus or distinct cultural domains can be achieved . our calculations are based on the mean - field limit in which each individual interacts with any other individual with equal probability . it is worth emphasizing that the dependence of the final state probabilities on the initial composition of the population is very close to the corresponding quantities on finite - dimensional systems @xcite . this coincidence is a reflection of the conservation of the global magnetization by the dynamics . indeed , for the classical voter model , the conservation of the magnetization immediately leads to the fact that the probability of @xmath2 consensus is equal to the initial density of @xmath2 spins in any dimension . there are two extensions of the model that are worth mentioning . first , it is natural that extremists have a stronger conviction than centrists . this suggests the generalization where a centrist adopts an extremist s opinion with probability @xmath140 while an extremist adopts a centrist s opinion with probability @xmath141 in an elementary update step . for this generalization , the densities of each species evolve by the following rate equations in the mean - field limit @xmath142 where @xmath143 quantifies the bias . for @xmath144 , the solution to these equations are @xmath145(e^{v t}-1)},\end{aligned}\ ] ] and similarly for @xmath146 , while @xmath147 . in the long time limit and for @xmath148 , we find @xmath149 while @xmath150 . thus if there is an innate bias favoring extremism , a frozen final state with no consensus is inevitable . on the other hand , if centrists are dominant ( @xmath151 ) , there centrist consensus is the final result . our results can also be adapted to determine the final states probabilities in the mean - field limit of the axelrod model for the case of two features and two states per feature . the four possible states of an individual can be represented by @xmath152 we define the dynamics to be that if an interaction pair has one common feature , then the other feature where disagreement exists is picked and the initial agent takes on the state of the interaction partner . there are four possible consensus states : all @xmath153 , @xmath154 , @xmath155 , and @xmath156 , and two types of frozen states : mixtures of @xmath153 and @xmath156 , and mixtures of @xmath154 and @xmath155 . we can identify the states @xmath153 and @xmath156 in this axelrod model with the @xmath94 and @xmath2 states , respectively , in our spin-1 ising model , and the states @xmath154 and @xmath155 taken together with the 0 state of the spin-1 ising model . then , using our previous results , if the initial densities of @xmath153 and @xmath156 states are @xmath27 and @xmath28 respectively , the probabilities of @xmath153 and @xmath156 consensus and the probability of frozen mixture of @xmath153 and @xmath156 are given by the expressions @xmath86 , @xmath87 and @xmath33 respectively ( eqs . ( [ p+1 ] ) and ( [ sol1 ] ) ) . by similar arguments , the probabilities of @xmath154 and @xmath155 consensus and the probability of @xmath154 and @xmath155 frozen mixture are given by the above expressions where @xmath27 and @xmath28 are now the initial densities of @xmath154 and @xmath155 states respectively . unfortunately , this exact mapping onto an equivalent ising spin system does not work when there are more than two features and/or more than two states per feature . we thank paul krapivsky for useful discussions and helpful advice . we also thank nsf grant dmr0227670 for financial support of this research . 99 s. galam , j. stat . phys . * 61 * , 943 ( 1990 ) ; physica a * 238 * , 66 ( 1997 ) . b. latan and a. nowak , in _ progress in communication science _ , pp . 43 , eds . g. a. barnett and f. j. boster ( 1997 ) . w. weidlich , _ sociodynamics ; a systematic approach to mathematical modelling in the social sciences _ ( harwood academic publishers , 2000 ) . k. sznajd - weron and j. sznajd , int . c * 11 * , 1157 ( 2000 ) . p. l. krapivsky and s. redner , phys . lett . * 90 * , 238701 ( 2003 ) . r. axelrod , j. conflict resolution * 41 * , 203 ( 1997 ) ; r. axelrod , _ the complexity of cooperation _ , ( princeton university press , princeton , 1997 ) . c. castellano , m. marsili , and a. vespignani , phys . 85 * , 3536 ( 2000 ) ; d. vilone , a. vespignani , and c. castellano , eur . j. b * 30 * , 399 ( 2002 ) . t. m. liggett , _ interacting particle systems _ ( springer - verlag , new york , 1985 ) . g. deffuant , d. neau , f. amblard and g. weisbuch , adv . complex syst . * 3 * , 87 ( 2000 ) ; g. weisbuch , g. deffuant , f. amblard , and j .- nadal , complexity * 7 * , 2002 ; g. deffuant , f. amblard , g. weisbuch and t. faure , journal of artificial societies and social simulations * 5 * , issue 4 , paper 1 ( jasss.soc.surrey.ac.uk ) ( 2002 ) . r. hegselmann and u. krause , journal of artificial societies and social simulation * 5 * , issue 3 , paper 2 ( jasss.soc.surrey.ac.uk ) ( 2002 ) and physica a , in press ( 2004 ) . e. ben - naim , p. l. krapivsky , and s. redner , physica d * 183 * , 190 ( 2003 ) . f. vazquez , p. l. krapivsky , and s. redner , j. phys . a * 36 * , l61 ( 2003 ) . s. redner , _ a guide to first - passage processes _ , ( cambridge university press , new york , 2001 ) . siegfried flgge ( 1994 ) _ practical quantum mechanics _ , ( springer - verlag , berlin , 1994 ) . i. s. gradshteyn and i. m. ryzhik , _ table of integrals , series and products _ , ( academic press , san diego , 1980 ) . a similar ultimate outcome also occurs in a cyclic 4-state voter model ; see l. frachebourg , p. l. krapivsky , and e. ben - naim , phys . e * 54 * , 6186 ( 1996 ) . we present the solution to eq . ( [ sch1 ] ) . this turns out be the schrdinger equation for the _ pschl - teller potential hole _ @xcite , for which the generic form of the equation is @xmath157u(\theta)=0 , \label{sch}\ ] ] with @xmath158 , @xmath159 , @xmath160 , and @xmath161 . this last equation has the roots @xmath162 and @xmath163 . the eigenvalues and eigenvectors of eq . ( [ sch ] ) are ( see , for example , @xcite page 92 ) : @xmath164 with @xmath165 , and where @xmath166 is the hypergeometric function . for @xmath167 , @xmath168 diverges at @xmath169 and @xmath170 . thus we take the solution for @xmath171 only and obtain the eigenfunctions and eigenvectors of eq . ( [ sch1 ] ) : @xmath172 + the solutions for the angular function in eq . ( [ theta ] ) therefore have the form @xmath173 with @xmath56 given by eq . ( [ eme ] ) . finally , the generic form of the product solution @xmath174 to eq . ( [ pc ] ) is @xmath175 the coefficient @xmath176 must be zero because @xmath72 is zero at the origin . thus @xmath177 and the general solution to eq . ( [ pc ] ) is @xmath178 we need the set of hypergeometric functions @xmath166 to form an orthogonal set to obtain the coefficients @xmath102 from the boundary condition : @xmath179 fortunately , the functions @xmath166 are related , for certain specific integer parameters , to the associated legendre polynomials ( that are known to form an orthogonal set ) via @xcite : @xmath180 if we take @xmath181 , @xmath182 and @xmath183 we obtain : @xmath184 then the general solution eq . ( [ sol ] ) can be written as @xmath185 where we absorbed the factor @xmath186 into the coefficient @xmath102 . to determine @xmath73 , we now apply the boundary condition @xmath187 at @xmath188 and obtain @xmath189 and using the orthogonality - normalization of the @xmath69 s we obtain the @xmath102 s : @xmath190 substituting these coefficients in eq . ( [ sol0 ] ) , we obtain eq . ( [ f - polar ] ) . to determine @xmath82 , we first need the first - passage probability @xmath92 for the effective random walker to hit the specific point @xmath101 on the locus @xmath36 , when starting from @xmath37 . this first - passage probability also satisfies the basic differential equation ( [ fp_eq ] ) , but with the boundary conditions given in eq . ( [ bc - fx ] ) . in polar coordinates the last of these boundary conditions is @xmath191 . thus we now determine the coefficients @xmath102 in eq . ( [ sol0 ] ) by @xmath192 inverting this relation and using the orthogonality of the @xmath69 s [ eq . ( [ on ] ) ] , we obtain @xmath193 using this result and transforming back to @xmath70 coordinates , we obtain the first - passage probability @xmath194 we then substitute this expression for @xmath92 into eq . ( [ p+ ] ) to determine @xmath87 . in so doing , we encounter the integral @xmath195 to evaluate this integral , we first make the variable change @xmath196 , use the identity @xmath197 , and integrate by parts to give @xmath198 where we have used @xmath199 , @xmath200 , as well as @xmath201 using the results of eq . ( [ fx ] ) and eq . ( [ ident ] ) in eq . ( [ p+ ] ) , we obtain the solution quoted in ( [ p+1 ] ) . to reduce the series representation ( [ sol1 ] ) for @xmath73 to a closed form , we start with the identity @xmath202 and make the variable change @xmath203 with @xmath204 integer to obtain @xmath205 this infinite series can be summed by the following algebraic manipulation . we define @xmath206 . the factor @xmath207 can be rewritten as @xmath208 , so that eq . ( [ sol_ro ] ) becomes @xmath209 where we used the taylor series expansion @xmath210 this then gives closed expression for the probability of reaching the mixed state as a function of @xmath78 quoted in eq . ( [ p - closed ] ) .

we determine the ultimate fate of individual opinions in a socially - interacting population of leftists , centrists , and rightists . in an elemental interaction between agents , a centrist and a leftist can become both centrists or both become leftists with equal rates ( and similarly for a centrist and a rightist ) . however leftists and rightists do not interact . this interaction step between pairs of agents is applied repeatedly until the system can no longer evolve . in the mean - field limit , we determine the exact probability that the system reaches consensus ( either leftist , rightist , or centrist ) or a frozen mixture of leftists and rightists as a function of the initial composition of the population . we also determine the mean time until the final state is reached . some implications of our results for the ultimate fate in a limit of the axelrod model are discussed .

introduction the model frozen final state consensus final states mean exit time discussion acknowledgments solution to equivalent schrdinger equation reduction of @xmath73

neutrinos are the only high energy particles able to propagate undeflected and unattenuated from the furthest reaches of the universe . extragalactic uhe @xmath1-ray astronomy falters for energies greater than a few tens of tev due to interactions with infrared and cosmic microwave background photons . the information carried by the neutrino messengers from distant , unexplored regions of the universe may help to unravel longstanding mysteries associated with the origin of the highest energy cosmic rays . several models predict high energy charged particle and neutrino emission ; the diffuse @xmath2-flux predictions by these models are constrained by the observed cosmic ray fluxes . amanda , the first neutrino telescope constructed in transparent ice , is deployed between 1500 m and 2000 m beneath the surface at the geographic south pole . it is designed to search for neutrinos that originate in the most violent phenomena in the observable universe . galactic objects like supernova remnants ( snr ) and extragalactic objects such as active galactic nuclei ( agn ) are expected to be the powerful engines accelerating protons and nuclei to the highest energies , which eventually interact to generate neutrinos . we have searched for point sources in the northern sky and for a diffuse flux of uhe ( e@xmath3 @xmath4 ev ) neutrinos of cosmic origin . we have performed the analysis using data collected in 2003 with a new data acquisition system based on transient waveform recorders ( twr - daq ) , which we compared to data collected by the standard amanda readout system ( @xmath5-daq ) . the upgraded amanda data acquisition electronics with transient waveform recorders , here displayed in a system of 2 racks with 72 twr modules.,scaledwidth=70.0% ] the data acquisition electronics of the amanda - ii detector ( fig . [ fig1 ] ) was upgraded in 2003 to read out the complete digitized waveform of the photomultiplier tubes ( pmts ) using transient waveform recorders ( twr ) . the transition was made to run the @xmath5-daq and twr - daq in parallel to verify that the performance of the latter is as good or better . to compare the two systems , the data from 2003 has been analyzed with data sets from both readout systems . the performance of the twr - daq is verified by comparing the results for the absolute rate of atmospheric neutrinos and the @xmath6 distribution with the results from the standard @xmath5-daq . extending the analysis tools to include twr data required several new developments : ( 1 ) the twr - daq measures the integrated charge @xmath7 from the full waveform of the pmt pulses , in contrast , the @xmath5-daq only measures the maximum amplitude of the pmt pulse in a 2 @xmath5s window . ( 2 ) various time offsets were taken into account for the twr - system , none of which impacted the @xmath5-daq . a time resolution of few nanoseconds was extracted . ( 3 ) the size of the twr - daq data set ( 20 tb for 2003 ) by far exceeds that of the @xmath5-daq ( @xmath81 tb ) , requiring an improved data handling tools . the data information of the two systems has been merged according to gps time and the fraction of overlapping pmts participating in the event in both systems . for the point source analysis we restricted the capabilities of the twr - daq system to mimic the features of the @xmath5-daq system as closely as possible . the timing and amplitude information extracted from waveforms has been used as input parameters to perform pmt - pulse cleaning , tot ( time - over - threshold ) and crosstalk cleaning . ( left ) time difference between the pmt pulses recorded by the two acquisition systems @xmath9 . ( right ) calibrated amplitude normalized to 1 photo - electron ( p.e . ) value for @xmath5-daq and twr - daq data . see text for details.,title="fig:",scaledwidth=49.0% ] ( left ) time difference between the pmt pulses recorded by the two acquisition systems @xmath9 . ( right ) calibrated amplitude normalized to 1 photo - electron ( p.e . ) value for @xmath5-daq and twr - daq data . see text for details.,title="fig:",scaledwidth=49.0% ] a gaussian fit of the distribution for @xmath9 yields @xmath10 ns ( fig . [ fig2 ] ( left ) ) , which is dominated by the systematic error of the time jitter between independent flash adc clocks of the twr system . the twr - daq timing calculations are relative to the values measured by the @xmath5-daq . the timing of the twr - daq system includes two sources of jitter , the digitization window of 10 ns and the time fluctuation of the flash adc clocks within the same 10 ns interval . this accounts for the @xmath11 ns . amplitudes are also calibrated by extracting the number of photo - electrons ( npe ) detected from the peak adc of the @xmath5-daq and charge @xmath7 of the twr system and normalized to 1-pe amplitude . by integrating the charge from pulses in the waveform , the dynamic range of the twr - daq extends to npe of @xmath12 , a factor of three higher than the @xmath5-daq . [ fig2 ] ( right ) shows the reconstructed amplitude of the twr - daq compared to the @xmath5-daq , which indicates a stable power law distribution extending up to 100 npe , while the @xmath5-daq system shows a `` knee '' around 10 npe . the knee is due to the amplitude saturation of channels read out by optical fibers , which comprise @xmath840% of the amanda channels . after cleaning , the muon track is reconstructed from the pulse times of the pmts . details on the reconstruction techniques can be found in . [ tab1 ] table [ tab1 ] summarizes the passing rates from the raw data level to the final sample of atmospheric neutrinos . the event selection criteria used in this analysis follows the method described in . [ fig4 ] shows that the angular mismatch for the final neutrino sample @xmath13 where @xmath14 . this value is expected from studies of the precision of the global minimizer in the reconstruction program , by reconstructing the same event sample twice with a 32-iteration algorithm . the azimuth angular mismatch @xmath15 , which is consistent with a @xmath16 of @xmath17 for the @xmath18 distribution to account for the zenith dependence on the azimuth angle . ( left ) the zenith @xmath19 difference distribution between the @xmath5-daq and the twr - daq systems , ( right ) the azimuthal @xmath20 difference distribution.,title="fig:",scaledwidth=49.0% ] ( left ) the zenith @xmath19 difference distribution between the @xmath5-daq and the twr - daq systems , ( right ) the azimuthal @xmath20 difference distribution.,title="fig:",scaledwidth=49.0% ] from the analysis based on twr - daq data , 1112 neutrinos are observed compared to 1026 neutrinos from the @xmath5-daq data analysis . the small differences in the event rate are compatible with the small differences in analysis procedures described in section [ sec2 ] . distribution of cos@xmath21 after final selection representing the atmospheric neutrino sample observed from the twr - daq and the @xmath5-daq analyses.,scaledwidth=99.0% ] sky plot displayed in equatorial coordinates , right ascension ( ra ) and declination ( dec ) , of the final sample of 1112 atmospheric neutrinos observed from the twr - daq analysis.,scaledwidth=100.0% ] fig . [ fig6 ] shows the cos@xmath21 distribution of the atmospheric neutrino sample extracted from the twr - daq and @xmath5-daq data analysis . satisfactory agreement can be seen for the cos@xmath21 distribution of the atmospheric neutrinos samples obtained by the different analyses . [ fig7 ] displays the sky map plotted in equatorial coordinates of the 1112 atmospheric neutrino candidates observed from the twr - daq analysis . all events are distributed in the norther hemisphere with very few events reconstructed at the horizon . in order to distinguish if the observed events follow a random distribution as expected by atmospheric neutrinos , or are an indication of localized event cluster as expected by a source of sky map plotted in coordinates of right ascension and declination of the 1112 atmospheric neutrino candidates from the twr - daq analysis . the scale on the right reflects excess or deficit in terms of the positive or negative deviation with respect to the mean background events.,scaledwidth=100.0% ] neutrinos of extraterrestrial origin , an analysis is required to estimate the statistical significance of the observed events . a full search of the northern sky was performed to look for any localized event cluster . the full scan is extended to @xmath22 in declination , since the limited statistics in the polar bin prevents an accurate estimate of the background . [ fig8 ] shows the sky map of the calculated significance for all observed 1112 events in terms of the a positive or negative deviation with respect to the mean background events . all observed regions with the highest significance are compatible with the background hypothesis . the highest significance observed shows a positive deviation of 4.3 @xmath16 , which corresponds to a probability of 23% for a search performed on events with randomized right ascension . ( left ) muon neutrino effective area as a function of e@xmath23 for point source searches displayed for different declinations @xmath24 . ( right ) average upper limit as a function of @xmath25 on @xmath26 for a e@xmath27 neutrino flux spectrum from the twr - daq and the @xmath5-daq analyses.,title="fig:",scaledwidth=49.0% ] ( left ) muon neutrino effective area as a function of e@xmath23 for point source searches displayed for different declinations @xmath24 . ( right ) average upper limit as a function of @xmath25 on @xmath26 for a e@xmath27 neutrino flux spectrum from the twr - daq and the @xmath5-daq analyses.,title="fig:",scaledwidth=49.0% ] numerous studies have been performed to implement the best possible description of the twr system in the monte carlo simulation . in particular , a correct description of the waveform hardware response is now available and proper tdc time windows as well as amplitude calculations have been now implemented . therefore a calculation of the neutrino effective area for point source search using twr data is now feasible . [ fig9 ] ( left ) shows the effective muon neutrino area as a function of e@xmath23 computed at the final selection of this analysis for four separated ranges of the declination . the a@xmath28 extends over six order of magnitudes as @xmath2-energy increases , while a@xmath28 decreases as declination increases due to the neutrino absorption in the earth . the average upper limit on @xmath26 as a function of declination is shown in the right panel of fig . the average neutrino flux upper limit is determined from the ratio of the average feldman and cousins@xcite upper limit @xmath29 computed according to the expected mean background and observed events , and the number of the expected signal events for a neutrino flux e@xmath30d@xmath31de@xmath32 gev s@xmath33cm@xmath27 from a point source at declination @xmath24 . together with the average upper limit extracted from the twr - daq data , the limit from the @xmath5-daq data of year 2003@xcite is also represented for comparison . these results are comparable , however the twr - daq limit is @xmath8 10% worse than the @xmath5-daq limit , because conservative event selection criteria have been applied that do not use the full waveform information in an optimal way . the preliminary average upper limit from the twr - daq analysis on the muon neutrino flux with spectrum d@xmath31de@xmath3 @xmath35 e@xmath36 is e@xmath30d@xmath31de@xmath3 @xmath37 gev s@xmath33cm@xmath27 at 90% confidence level , in the energy range 1.26 tev @xmath38 e@xmath39 @xmath38 1.6 pev . this upper limit does not include systematic uncertainties . amanda has submitted a publication on a 5-year search for point sources ( 2000 - 2004 ) based on the @xmath5-daq data@xcite , which improves the limit by @xmath867% compared to the results from this analysis . ( left ) one of the 1112 atmospheric neutrino event detected in the experimental array . ( right ) a high - energy mc simulated event . the size of the circles represent energy deposited in the detector , the colors represent the time profile of photons propagating in the ice which have scattered to the pmts.,title="fig:",scaledwidth=49.0% ] ( left ) one of the 1112 atmospheric neutrino event detected in the experimental array . ( right ) a high - energy mc simulated event . the size of the circles represent energy deposited in the detector , the colors represent the time profile of photons propagating in the ice which have scattered to the pmts.,title="fig:",scaledwidth=49.0% ] the search for uhe neutrinos differs from the point source mainly in two aspects . first , the search for point sources is restricted to upward going events of energy within the tev scale , the uhe analysis extends the search beyond pev energies . second , while point sources are localized in a small region of the sky , uhe diffuse neutrinos are expected to come from the entire visible sky ( 2@xmath40 sr ) , with a higher probability to originate from the horizon . fig . [ fig11 ] shows one of the 1112 upward going neutrino events and an almost horizontal uhe simulated neutrino event with energy e @xmath41 ev . simulated events with energy above one pev deposit a substantial amount of light which is recorded by almost all pmts . in order to better separate background events from signal events , we developed new variables which exploit the information of the full waveforms . contour plots displaying the capability of correctly reconstructing the total number of photon - electron injected into the detector ( left ) for the twr - daq and ( right ) for the @xmath5-daq system.,title="fig:",scaledwidth=49.0% ] contour plots displaying the capability of correctly reconstructing the total number of photon - electron injected into the detector ( left ) for the twr - daq and ( right ) for the @xmath5-daq system.,title="fig:",scaledwidth=49.0% ] the capability of improved photo - electron reconstruction of the twr - daq system can be seen in fig . [ fig13 ] , which displays a contour plot in log@xmath42-scale of the sum of the reconstructed number of pe , npe@xmath43 versus the npe@xmath44 generated by the signal mc simulation . while the @xmath5-daq system quickly saturates on the order of few thousands of pe s , the twr - daq , by using the after - pulse information , can see the millions of pe s which are typical for uhe events triggered by single muons . [ fig15 ] summarizes the results of several amanda searches for diffuse neutrinos at different energy ranges using @xmath5-daq data . the experimental limits assume a 1:1:1 ratio of neutrino flavors at the earth due to oscillation . the dotted and dashed lines represent a sample of model predictions , adjusted for oscillation if necessary . the horizontal solid lines represent limits on the integrated neutrino flux from the diffuse analyses with amanda - b10@xcite , with amanda - ii@xcite , from the cascade analysis@xcite , and from the uhe analysis@xcite , with the nt200 neutrino telescope@xcite , and at the highest energies from anita - lite@xciteand rice@xcite . these are the most stringent flux limits at 90% c.l . for an e@xmath27 spectrum to date . the amanda limits have been determined by analysing data collected with the @xmath5-daq system . however , as it has been shown above , by performing a new analysis approach with twr - daq data , it is possible to develop new techniques and to improve the current experimental limits for the uhe neutrino search . experimental limits on the integrated neutrino flux and detector sensitivity for a neutrino flux of e@xmath27 spectrum . the curves represents predictions from theoretical models , the solid lines show current amanda ( all - flavor limits ) , nt200 , anita and rice experimental limits.,scaledwidth=100.0% ] the point source search provides the first detailed evaluation of the performance of the twr - daq system . this analysis demonstrated that the twr - daq produces similar event rates and angular distribution as the data collected by the standard @xmath5-daq system . no statistically significant excess has been observed after performing a full search of the northern hemisphere of the sky for localized event clusters , therefore a flux limit based on twr - daq data is calculated . the amanda twr readout is now being incorporated into the icecube data acquisition system . by exploiting the information of the full waveforms from the twr - daq system , it is possible to develop new analysis techniques for an improved search for diffuse uhe neutrinos . currently the amanda experiment has placed the most stringent neutrino flux limits to date , which can be further improved by analyses performed with twr - daq data . the author acknowledges support from the u.s . national science foundation ( nsf ) physics division , the nsf - supported teragrid systems at the san diego supercomputer center ( sdsc ) and the national center for supercomputing applications ( ncsa ) , the phi beta kappa alumni in southern california for providing travel grants , the astrophysics associates , inc . , italian ministry of education , european physical society , ettore majorana foundation , and the electron tubes ltd . for providing the full scholarship at erice .

the antarctic muon and neutrino detector array ( amanda ) has been taking data since 2000 and its data acquisition system was upgraded in january 2003 to read out the complete digitized waveforms from the buried photomultipliers ( pmts ) using transient waveform recorders ( twr ) . this system currently runs in parallel with the standard amanda data acquisition system . once amanda is incorporated into the 1 km@xmath0 detector icecube , only the twr system will be kept . we report results from a first atmospheric neutrino analysis on data collected in 2003 with twr . good agreement in event rate and angular distribution verify the performance of the twr system . a search of the northern hemisphere for localized event clusters shows no statistically significant excess , thus a flux limit is calculated , which is in full agreement with previous results based on the standard amanda data acquisition system . we also update the status of a search for diffusely distributed neutrinos with ultra high energy ( uhe ) using data collected by the twr system .

introduction the twr system analysis search for point sources of neutrinos search for uhe neutrinos conclusion acknowledgments

the quantum ising model defined in terms of pauli matrices @xmath12 @xmath13 is the basic model to study quantum phase transitions at zero - temperature @xcite . on a hypercubic lattice in dimension @xmath2 , the pure model with the same transverse field @xmath14 on all sites , and the same ferromagnetic coupling @xmath15 between nearest - neighbors is well understood via the equivalence with the classical ising model in dimension @xmath16 , i.e. the time plays the role of an extra space - dimension @xcite , and the dynamical exponent is @xmath17 . in particular , the quantum model in @xmath18 corresponds to the exactly solved 2d classical ising model , the quantum model in @xmath4 corresponds to the 3d classical ising model , and the quantum model in dimension @xmath19 is characterized by the standard mean - field exponents . from the point of view of block - renormalization for quantum models , there exists a special self - dual procedure introduced by fernandez - pacheco @xcite , which is able to reproduce the exact critical point @xmath20 and the exact correlation length exponent @xmath21 . various generalizations of this procedure to higher dimensions @xmath1 have been studied @xcite , as well as extensions to other quantum models like the potts and the ashkin - teller models @xcite . in the disordered case , where the transverse fields @xmath22 and the couplings @xmath23 are random variables , many exact results have been obtained in @xmath18 by daniel fisher @xcite via the asymptotically exact strong disorder renormalization procedure ( for a review , see @xcite ) . in particular , the transition is governed by an infinite disorder fixed point and presents unconventional scaling laws with respect to the pure case . in dimension @xmath1 , the strong disorder renormalization procedure has been studied numerically with the conclusion that the transition is also governed by an infinite - disorder fixed point in dimensions @xmath24 @xcite . these numerical renormalization results are in agreement with the results of independent quantum monte - carlo in @xmath4 @xcite . the strong disorder renormalization is thus a very powerful method , but it leads to a complicated renormalized topology for the surviving clusters as soon as @xmath1 . in particular , a large number of very weak bonds are a priori generated during the rg , that will eventually not be important for the forthcoming rg steps . this is why recent numerical implementations of strong disorder rg rules are based on algorithms avoiding this proliferation of weak generated bonds @xcite . a natural question is whether infinite disorder fixed points can be also reproduced by more standard block - renormalization . recently , miyazaki and nishimori @xcite have proposed to generalize the block - renormalization of fernandez - pacheco @xcite concerning the pure model in @xmath18 , and their extension for the pure model in @xmath25 @xcite to the random case : their results are in agreement with the infinite disorder fixed point scalings . however in dimension @xmath1 , their procedure has the drawback that the various directions are treated inequivalently , so that they need to re - symmetrize afterward @xcite . in addition , in the random case , they have used some pool method that does not keep all the generated correlations between renormalized parameters , and the only critical exponent that they measure is the finite - size correlation length exponent @xmath26 , whereas it is interesting to measure also the activated exponent @xmath27 and the typical correlation length exponent @xmath28 in the present paper , we thus propose another generalization in dimensions @xmath4 and @xmath29 of the self - dual one - dimensional block - renormalization of fernandez - pacheco , that we study analytically in the pure case and numerically in the disordered case . the paper is organized as follows . in section [ sec_rgrule ] , we describe the elementary renormalization rule that will be applied throughout the paper . the application to the one - dimensional quantum ising model of eq . [ hqising ] , both pure and random , is recalled in section [ sec_d1 ] . in section [ sec_d2 ] , we derive the block renormalization rules for the two - dimensional case . the applications of these two - dimensional rules are discussed respectively in section [ sec_d2pure ] for the pure ferromagnetic model , in section [ sec_d2randomferro ] for the random ferromagnetic model , and in section [ sec_d2spinglass ] for the spin - glass model . our conclusions are summarized in section [ sec_conclusion ] . the extension to dimension @xmath29 is presented in appendix [ sec_d3 ] . in this section , we consider the following elementary hamiltonian involving @xmath30 spins @xmath31 with @xmath32 @xmath33 where @xmath34 plays the special role of the master , and the @xmath35 spins @xmath31 with @xmath36 are the slaves. for each eigenvalue @xmath38 of @xmath39 , one has to diagonalize independently the @xmath35 hamiltonians concerning the single quantum spin @xmath31 in the transverse field @xmath22 and in the effective magnetic field @xmath40 @xmath41 the two eigenvalues read @xmath42 with the following corresponding eigenvectors @xmath43 where @xmath44 in summary , the hamiltonian @xmath37 of eq . [ hintra ] has two degenerate ground states labeled by the two values @xmath38 . the ground - state energy is simply obtained by adding the @xmath35 contributions @xmath45 @xmath46 the two corresponding ground - states are obtained by the tensor products @xmath47 the projector onto the two ground - states reads @xmath48 it is thus convenient to define the renormalized spin @xmath49 from these two ground - states @xmath50 with the corresponding operators @xmath51 to evaluate @xmath53 , we have to consider the action of the operator @xmath54 on each ground state @xmath55 this yields the following trivial projection rule in terms of the renormalized operator @xmath56 of eq . [ sigmar ] @xmath57 to evaluate @xmath59 , we need to compute the action of the operator @xmath60 on each ground state @xmath61 using @xmath62 one obtains the following projection rule in terms of the renormalized operator @xmath56 of eq . [ sigmar ] @xmath63 to compute @xmath65 , we have to consider the action of the operator @xmath66 on each ground state @xmath67 using @xmath68 one obtains the following projection rule in terms of the renormalized operator @xmath69 of eq . [ sigmar ] @xmath70 \sigma_{r0}^x \label{projsigmax0}\end{aligned}\ ] ] the physical meaning of the procedure derived above is thus very simple : the renormalized spin @xmath71 represents the master spin @xmath72 dressed by its b slave spins @xmath73 . for each slave @xmath36 , we may consider the two limiting cases : \(a ) if @xmath74 , then the slave spin @xmath73 is ferromagnetically locked to its master @xmath72 ( eq . [ projsigmazi ] becomes @xmath75 to be compared with eq . [ projsigmaz0 ] ) , so that the flipping of the master spin is affected by the small ratio @xmath76 ( eq . [ projsigmax0 ] ) . \(b ) if @xmath77 , then the slave spin @xmath73 is mostly disordered and only weakly polarized by its master @xmath72 ( eq . [ projsigmazi ] becomes @xmath78 ) , so that the flipping of the master spin is unchanged ( eq . [ projsigmax0 ] ) . the projection rules above are thus compatible with the strong disorder rg rules in the two limits @xmath74 and @xmath77 , but can also apply to cases where @xmath79 . so they can be used to analyze both pure and random quantum ising models , as recalled in the following section for @xmath18 . the hamiltonian of the quantum ising chain reads @xmath80 in a block renormalization rule , one wishes to replace each block of two spins @xmath81 by a single renormalized spin @xmath82 . it is thus convenient to rewrite eq . [ h1d ] as @xmath83 \label{h1di}\end{aligned}\ ] ] the idea of fernandez - pacheco @xcite , written here for the random case @xcite , is the following choice of the intra - block hamiltonian @xmath84 since @xmath85 has the form the hamiltonian of eq . [ hintra ] analyzed in section [ sec_rgrule ] , the two spins @xmath86 can be renormalized via a single renormalized spin @xmath87 . the application of the projection rules of eqs [ projsigmaz0 ] , [ projsigmazi ] and [ projsigmax0 ] read for the present case @xmath88 as a consequence , the projection of the remaining part of the hamiltonian @xmath89 p_{intra}^{(1 ) } = \sum_{i } h_{i}^{r } \nonumber \\ h_{i}^{r } = & & - h^r ( 2i ) \sigma_{r(2i)}^x - j_{2 \vec x}^r ( 2i-2 ) \sigma^z_{r(2i-2 ) } \sigma^z_{r(2i ) } \label{h1drfin}\end{aligned}\ ] ] has the same form as the initial hamiltonian of eq . [ h1d ] , in terms of \(i ) the renormalized transverse fields on the remaining even sites @xmath90 \(ii ) the renormalized couplings between the remaining even sites @xmath91 if the initial parameters are @xmath92 on the whole chain , one obtains after one rg step the renormalized parameters ( eqs [ rgh1d ] and [ rgj1d2x ] ) @xmath93 so that the ratio @xmath94 evolves according to the simple rule @xmath95 the disordered attractive fixed point @xmath96 and the ferromagnetic attractive fixed point @xmath97 are separated by the unstable fixed point @xmath98 characterized by the correlation length exponent @xmath0 obtained by @xmath99 . the fact that both @xmath100 and @xmath26 are in agreement with the exact solution @xcite shows that the fernandez - pacheco choice @xcite of the intra - block hamiltonian of eq . [ h1intra1d ] is better than other choices @xcite . since the quantum ising chain represents the anisotropic limit of the two - dimensional classical ising model and is the the same universality class , it is interesting to compare with all the real - space renormalization procedures concerning classical spin models , from the early migdal - kadanoff schemes to the more recent tensor networks formulations ( see the recent review @xcite and references therein ) . for the two - dimensional ising model , whenever the two directions are treated on the same footing , the various real - space rg procedures that have been proposed are able to produce very good approximations of the exponent @xmath26 ( see for instance the table i of the review @xcite ) but never yield exactly @xmath0 , in contrast to the fernandez - pacheco quantum procedure described above . so it seems presently that the only way to obtain exactly @xmath0 for the 2d ising model is by defining a renormalization procedure _ for the transfer matrix _ @xcite in order to inherit the exactness of the exponent @xmath26 of the fernandez - pacheco quantum procedure . in terms of the ratios @xmath101 the rg rules of eq . [ rgh1d ] and [ rgj1d2x ] reads @xmath102 and thus corresponds to a simple addition in log - variables @xmath103 so after @xmath104 rg steps corresponding to a length @xmath105 , the renormalized ratio @xmath106 reads in terms of the initial variables @xmath107 \label{inffp}\end{aligned}\ ] ] the central limit theorem yields the asymptotic behavior @xmath108 + l^{1/2 } \sqrt { \left [ var[\ln j_{\vec x } ] + var[\ln h ] \right ] } u \label{inffpclt}\end{aligned}\ ] ] where @xmath109 is a gaussian random variable . the first term yields that the critical point corresponds to the condition @xmath110 and that the typical correlation length exponent is @xmath111 outside criticality , the competition between the first and the second term shows the finite - size correlation exponent is @xmath112 at criticality where the first term vanishes , the second random term of order @xmath113 corresponds to an infinite disorder fixed point of exponent @xmath114 all these conclusions of eqs [ criti1d ] , [ ntyp1d ] , [ nav1d ] , [ psi1d ] obtained via the application of the fernandez - pacheco renormalization to the random quantum ising chain @xcite , are in agreement with the fisher strong disorder renormalization exact results @xcite . it is thus interesting to look for an appropriate generalization of the fernandez - pacheco renormalization in higher dimensions , and first of all in dimension @xmath4 . the initial quantum ising hamiltonian defined on the square lattice of unit vectors @xmath115 reads @xmath116 \label{h2d}\end{aligned}\ ] ] various generalizations of the one - dimensional fernandez - pacheco renormalization procedure have been already proposed in dimension @xmath4 , both for the pure case @xcite and for the disordered case @xcite , with the drawbacks recalled in the introduction . in this section , we thus introduce another procedure where the two directions are considered on the same footing . we wish to define a block renormalization rule , where each block of four spins @xmath117 will be replaced by a single renormalized spin @xmath118 , after two elementary renormalization steps . it is convenient to start by rewriting eq . [ h2d ] as @xmath119 \nonumber \\ & & - \sigma_{(2i,2j-1)}^z [ j_{\vec x}(2i,2j-1 ) \sigma_{(2i+1,2j-1 ) } ^z + j_{\vec y}(2i,2j-1 ) \sigma_{2i,2j}^z \nonumber \\ & & \ \ \ \ \ \ \ \ \ \ \ \ \ \ + j_{\vec x}(2i-1,2j-1)\sigma_{(2i-1,2j-1)}^z + j_{\vec y}(2i,2j-2 ) \sigma_{(2i,2j-2)}^z ] \nonumber\end{aligned}\ ] ] for the first renormalization step , we choose the following intra - block hamiltonian @xmath120 since @xmath121 has the form the hamiltonian of eq . [ hintra ] analyzed in section [ sec_rgrule ] , the three spins @xmath122 can be renormalized via a single renormalized spin @xmath123 , whereas the spin @xmath124 that is not involved in @xmath121 remains unchanged . the application of the projection rules of eqs [ projsigmaz0 ] , [ projsigmazi ] and [ projsigmax0 ] read for the present case @xmath125 as a consequence , the projection of the remaining part of the hamiltonian reads @xmath126 p_{intra}^{(1 ) } = \sum_{(i , j ) } h_{i , j}^{r } \nonumber \\ h_{i , j}^{r } = & & - h ( 2i-1,2j-1 ) \sigma_{(2i-1,2j-1)}^x - h^r ( 2i,2j ) \sigma_{r(2i,2j)}^x \nonumber \\ & & - j^r_{2 \vec x } ( 2i-2,2j ) \sigma^z_{r(2i-2,2j ) } \sigma^z_{r(2i,2j ) } - j^r_{2\vec y}(2i,2j-2 ) \sigma^z_{r(2i,2j-2 ) } \sigma^z_{r(2i,2j ) } \nonumber \\ & & - j^r _ { \vec x- \vec y } ( 2i,2j ) \sigma^z_{r(2i,2j ) } \sigma_{(2i+1,2j-1 ) } ^z - j^r _ { -\vec x+ \vec y } ( 2i,2j ) \sigma^z_{r(2i,2j ) } \sigma_{2i-1,2j+1}^z \nonumber \\ & & - j^r _ { \vec x+ \vec y } ( 2i-1,2j-1 ) \sigma_{(2i-1,2j-1)}^z \sigma^z_{r(2i,2j ) } \label{h2drfin}\end{aligned}\ ] ] in terms of \(i ) the renormalized transverse fields of even - even sites @xmath127 \(ii ) the renormalized couplings along the horizontal directions at distance two @xmath128 \(iii ) the renormalized couplings along the vertical directions at distance two @xmath129 \(iv ) the renormalized couplings along the diagonal directions @xmath130 and @xmath131 @xmath132 \(v ) the renormalized couplings along the diagonal direction @xmath133 within each block @xmath134 for the hamiltonian @xmath135 of eq . [ h2drfin ] , we choose the following intra - block hamiltonian @xmath136 it has the form the hamiltonian of eq . [ hintra ] analyzed in section [ sec_rgrule ] , so that the two spins @xmath137 can be replaced by a single renormalized spin @xmath118 . the application of the projection rules of eqs [ projsigmaz0 ] , [ projsigmazi ] and [ projsigmax0 ] read for the present case @xmath138 ^ 2 } } \sigma^z_{rr(2i,2j ) } \nonumber \\ p_{intra}^{(2 ) } \sigma^x_{r(2i,2j ) } p_{intra}^{(2 ) } & & = \frac{h ( 2i-1,2j-1)}{\sqrt{h^2(2i-1,2j-1)+[j^r_{\vec x+\vec y } ( 2i-1,2j-1)]^2 } } \sigma_{rr(2i,2j)}^x \label{proj2dp2}\end{aligned}\ ] ] as a consequence , the projection of the remaining part of the hamiltonian reads @xmath139 p_{intra}^{(2 ) } = \sum_{(i , j ) } h_{i , j}^{rr } \nonumber \\ h_{i , j}^{rr } = & & - h^{rr } ( 2i,2j ) \sigma_{rr(2i,2j)}^x % \nonumber \\ & & - j^r_{2 \vec x } ( 2i-2,2j ) \sigma^z_{rr(2i-2,2j ) } \sigma^z_{rr(2i,2j ) } - j^r_{2\vec y}(2i,2j-2 ) \sigma^z_{rr(2i,2j-2 ) } \sigma^z_{rr(2i,2j ) } \label{h2drrfin}\end{aligned}\ ] ] i.e. it has the same form as the initial hamiltonian on the square lattice , in terms of \(i ) the renormalized transverse fields @xmath140 ^ 2+[j^r_{\vec x+\vec y}(2i-1,2j-1)]^2 } } \label{rghrr}\end{aligned}\ ] ] \(ii ) the renormalized couplings along the horizontal directions at distance two @xmath141 ^ 2 + [ j^r _ { \vec x+ \vec y}(2i-1,2j-1)]^2 } } \label{rgj2ddrrh}\end{aligned}\ ] ] \(ii ) the renormalized couplings along the vertical directions at distance two @xmath142 ^ 2 } } \label{rgj2ddrrv}\end{aligned}\ ] ] in the following sections , we discuss the application of these renormalization rules to the pure case , to the random ferromagnetic case , and to the spin - glass case . if we start from the pure model of parameters @xmath92 , the renormalization rules of eqs [ rghrr ] [ rgj2ddrrh ] [ rgj2ddrrv ] reduce to @xmath143 ^ 2 } } = \frac{h^4 } { \sqrt{h^2+j^2 } \sqrt{h^2(h^2+j^2)+4 j^4 } } \nonumber \\ j^{rr } & & = \frac{j^2}{\sqrt{h^2+j^2 } } \left [ 1 + \frac { 2 j^2 } { \sqrt{h^2(h^2+j^2)+ 4j^4 } } \right ] \label{purrg2d}\end{aligned}\ ] ] in terms of the ratio @xmath94 , the renormalization rule reads @xmath144 \equiv \phi(k ) \label{xevol2d}\end{aligned}\ ] ] we are now interested into the critical point satisfying the fixed point equation @xmath145 ( between the ferromagnetic fixed point @xmath146 and the disordered fixed point @xmath147 ) : eliminating the square - root yields the fourth degree equation @xmath148 the only positive root reads ( the other three roots can be disregarded since one is negative , and the two others are imaginary ) @xmath149 the correlation length exponent @xmath26 can be obtained from @xmath150 and the corresponding numerical value @xmath151 is very close to the numerical estimate @xmath5 for the 3d classical ising model . the two - dimensional procedure that we have proposed is thus much simpler that the miyazaki - nishimori - ortiz procedure that needs re - symmetrization between the two directions @xcite and yields a very good approximation for the correlation length exponent @xmath26 , better than other real - space renormalization procedures @xcite . the extension to the pure model in @xmath29 is described in the appendix a , and we now turn to the random models in dimension @xmath4 . in order to compare with the strong disorder renormalization numerical results @xcite , we have adopted the standard choice of a flat distribution of couplings between @xmath152 and @xmath153 @xmath154 and a flat distribution of transverse fields between @xmath155 and @xmath156 @xmath157 so that the control parameter of the zero - temperature transition is @xmath158 we have applied numerically the renormalization rules derived above to @xmath159 disordered two - dimensional samples of linear size @xmath160 ( containing @xmath161 spins ) with periodic boundary conditions . the renormalization procedure is stopped at the scale @xmath162 , where there remains only four sites and eight links in each sample . as a consequence for this largest length @xmath162 , the statistics is over @xmath163 random fields and over @xmath164 random couplings . at each renormalization step corresponding to the lengths @xmath165 with @xmath166 we have analyzed the statistical properties of the renormalized transverse fields and of the renormalized couplings . more precisely , we have measured the rg flows of the typical values defined by @xmath167 and of the widths of the probability distributions @xmath168 as a function of the length @xmath169 for 28 values of the control parameter @xmath170 of eq . [ control ] . the linear size @xmath171 and the statistics over @xmath159 are thus of the same order of those used in recent strong disorder renormalization studies in @xmath4 @xcite , but of course the implementation is much simpler here since the spatial structure remains a square lattice upon rg instead of an evolving non - trivial topology . another difference is that we analyze the statistics over samples at fixed control parameter @xmath170 and fixed size @xmath169 , whereas strong disorder renormalization studies of ref . @xcite are based on the determination of the pseudo - critical parameter for each sample . on fig . [ figflowh ] , we show in log - log scale the rg flows of the typical renormalized transverse field @xmath172 of eq . [ deftyp ] @xmath173 and of the width @xmath174 of eq . [ defwidth ] @xmath175 as a function of @xmath176 : + ( i ) ordered phase @xmath177 : the asymptotic slope is @xmath4 , as shown here for @xmath178 ( squares ) and @xmath179 ( diamond ) + ( ii ) disordered phase @xmath180 : the asymptotic slope is @xmath181 , as shown here for @xmath182 ( triangles up ) and @xmath183 ( triangles right ) + ( iii ) critical point @xmath184 ( circles ) : the asymptotic slope is @xmath185 . + ( b ) @xmath186 as a function of @xmath187 : + ( i ) ordered phase @xmath177 : the asymptotic slope is @xmath188 , as shown here for @xmath178 ( squares ) and @xmath179 ( diamond ) + ( ii ) disordered phase @xmath180 : the asymptotic slope is @xmath181 , as shown here for @xmath182 ( triangles up ) and @xmath183 ( triangles right ) + ( iii ) critical point @xmath184 ( circles ) : the asymptotic slope is @xmath185 . , title="fig:",height=226 ] as a function of @xmath176 : + ( i ) ordered phase @xmath177 : the asymptotic slope is @xmath4 , as shown here for @xmath178 ( squares ) and @xmath179 ( diamond ) + ( ii ) disordered phase @xmath180 : the asymptotic slope is @xmath181 , as shown here for @xmath182 ( triangles up ) and @xmath183 ( triangles right ) + ( iii ) critical point @xmath184 ( circles ) : the asymptotic slope is @xmath185 . + ( b ) @xmath186 as a function of @xmath187 : + ( i ) ordered phase @xmath177 : the asymptotic slope is @xmath188 , as shown here for @xmath178 ( squares ) and @xmath179 ( diamond ) + ( ii ) disordered phase @xmath180 : the asymptotic slope is @xmath181 , as shown here for @xmath182 ( triangles up ) and @xmath183 ( triangles right ) + ( iii ) critical point @xmath184 ( circles ) : the asymptotic slope is @xmath185 . , title="fig:",height=226 ] at the critical point @xmath189 , the rg flows of the typical value and of the width display the same activated scaling of an infinite disorder fixed point @xmath190 where @xmath191 is some @xmath192 random variable . the values obtained here for the location of the critical point @xmath193 and the activated exponent @xmath185 turn out to be different from the estimations @xmath194 and @xmath195 obtained via the strong disorder renormalization ( see @xcite and references therein ) : the origin of these differences is not clear to us . in the ordered phase , the logarithm of the typical renormalized transverse field grows extensively with respect to the volume @xmath196 ( with @xmath4 here ) @xmath197 where the length scale @xmath198 represents the characteristic size of finite disordered clusters within this ordered phase . from our numerical data concerning the ordered phase , the asymptotic behavior of eq . [ hlorder ] allows to measure @xmath198 and its divergence near criticality @xmath199 in the disordered phase , the asymptotic typical value @xmath200 diverges with an essential singularity as a function of the control parameter @xmath201 in the critical region , the finite - size scaling form is governed by some finite - size scaling correlation length exponent @xmath202 @xmath203 the matching with the behavior in the ordered phase ( eqs [ hlorder ] and [ nuh ] ) and in the disordered phase ( eq . [ defkappa ] ) yields the relations @xmath204 and @xmath205 the previous numerical measures of @xmath27 , @xmath206 and @xmath207 yield the estimate @xmath208 as shown on fig . [ figfss ] , this value of @xmath202 gives satisfactory finite - size scaling plots of the numerical data of fig . [ figflowh ] . corresponding to the sizes @xmath209 with the values @xmath184 for the critical point and @xmath210 for the finite - size correlation length exponent : + ( a ) @xmath211 as a function of @xmath212 . + ( b ) @xmath213 as a function of @xmath212 . , title="fig:",height=226 ] corresponding to the sizes @xmath209 with the values @xmath184 for the critical point and @xmath210 for the finite - size correlation length exponent : + ( a ) @xmath211 as a function of @xmath212 . + ( b ) @xmath213 as a function of @xmath212 . , title="fig:",height=226 ] this value for @xmath202 ( eq . [ nufss ] ) agrees with the estimations obtained via the strong disorder renormalization ( see @xcite and references therein ) , and with the asymmetric block renormalization of ref @xcite . as a function of @xmath176 : + ( i ) ordered phase @xmath177 : the flow is non - monotonic , as shown here for @xmath178 ( squares ) and @xmath179 ( diamond ) : the asymptotic behavior ( not visible on this plot appropriate to the critical region ) corresponds to the classical random ferromagnet growth @xmath214 + ( ii ) disordered phase @xmath180 : the asymptotic slope is @xmath188 , as shown here for @xmath182 ( triangles up ) and @xmath183 ( triangles right ) + ( iii ) critical point @xmath184 ( circles ) : the asymptotic slope is @xmath185 . + ( b ) @xmath215 as a function of @xmath187 : + ( i ) ordered phase @xmath177 : the asymptotic slope is @xmath216 , as shown here for @xmath178 ( squares ) + ( ii ) disordered phase @xmath180 : the asymptotic slope is @xmath217 , as shown here for @xmath182 ( triangles up ) + ( iii ) critical point @xmath184 ( circles ) : the asymptotic slope is @xmath185 . , title="fig:",height=226 ] as a function of @xmath176 : + ( i ) ordered phase @xmath177 : the flow is non - monotonic , as shown here for @xmath178 ( squares ) and @xmath179 ( diamond ) : the asymptotic behavior ( not visible on this plot appropriate to the critical region ) corresponds to the classical random ferromagnet growth @xmath214 + ( ii ) disordered phase @xmath180 : the asymptotic slope is @xmath188 , as shown here for @xmath182 ( triangles up ) and @xmath183 ( triangles right ) + ( iii ) critical point @xmath184 ( circles ) : the asymptotic slope is @xmath185 . + ( b ) @xmath215 as a function of @xmath187 : + ( i ) ordered phase @xmath177 : the asymptotic slope is @xmath216 , as shown here for @xmath178 ( squares ) + ( ii ) disordered phase @xmath180 : the asymptotic slope is @xmath217 , as shown here for @xmath182 ( triangles up ) + ( iii ) critical point @xmath184 ( circles ) : the asymptotic slope is @xmath185 . , title="fig:",height=226 ] on fig . [ figflowj ] , we show in log - log scale the rg flows of the typical renormalized coupling @xmath218 of eq . [ deftyp ] @xmath219 and of the width @xmath220 of eq . [ defwidth ] @xmath221 at the critical point @xmath193 , the rg flows of the typical value and of the width display the same activated scaling of infinite disorder fixed point as in eq . [ hlcriti ] @xmath222 where @xmath223 is some @xmath192 random variable . in the disordered phase , the typical renormalized coupling @xmath224 decays exponentially with the size @xmath169 @xmath225 where @xmath226 represents the typical correlation length . from our numerical data concerning the disordered phase , the asymptotic behavior of eq . [ jldisordertyp ] allows to measure @xmath227 and its divergence near criticality @xmath228 the compatibility with the finite - size scaling form analogous to eq . [ fsshtyp ] @xmath229 implies the relation @xmath230 it is satisfied by the previously quoted estimates of @xmath231 , @xmath232 and @xmath27 . to study the two - dimensional spin - glass quantum ising model , we have replaced the probability distribution of eq . [ pjflat ] concerning the random ferromagnetic case by the flat distribution of zero - mean @xmath233 we have then repeated exactly the same numerical analysis as in the previous section [ sec_d2randomferro ] : the critical point is now found at @xmath234 , and the critical exponents @xmath9 , @xmath235 , @xmath10 , @xmath11 are the same as for the random ferromagnet presented in the previous section , as expected for infinite disorder fixed points @xcite , and as found within the miyasaki - nishimori asymmetric scheme @xcite . in this paper , we have proposed a simple generalization in @xmath1 of the self - dual block renormalization procedure of fernandez - pacheco @xcite , that we have tested for pure and random quantum ising models , with the following conclusions . for the pure models , where the fernandez - pacheco procedure is known to reproduce the exact correlation length exponent @xmath236 @xcite , we have obtained @xmath237 ( to be compared with the 3d classical ising model exponent @xmath5 ) and @xmath238 ( to be compared with the 4d classical ising model mean - field exponent @xmath7 ) . for the random models , where the fernandez - pacheco procedure is known to reproduce exactly the location of the critical point and the critical exponents @xmath239 , @xmath240 and @xmath241 of the infinite disorder fixed point @xcite , we have applied numerically the renormalization rules to two - dimensional samples of linear size @xmath242 , with either random ferromagnetic disorder or spin - glass disorder , both types of disorder leading to the same infinite disorder fixed point : the finite - size correlation exponent @xmath11 coincides with strong disorder renormalization result ( see @xcite and references therein ) , and with the asymmetric block renormalization of ref @xcite , but the activated exponent @xmath9 turns out to be somewhat higher than strong disorder renormalization estimate @xmath243 ( see @xcite and references therein ) . the origin of this difference remains to be clarified . we have also analyzed the rg flows in the disordered and ordered phases , in order to extract the typical correlation length exponent @xmath10 and the analog @xmath235 , and tested the finite - size scaling . in summary , the generalization in @xmath1 of the self - dual block renormalization procedure of fernandez - pacheco @xcite is able to reproduce both the conventional scaling of pure critical points and the activated scaling of infinite disorder fixed points . it would be thus interesting to develop such methods in models governed by strong ( not infinite ) disorder fixed points like the quantum ising model with long - ranged interactions @xcite or the superfluid - insulator transition @xcite , as well as in models where the transition at weak disorder could be in another universality class @xcite . the initial quantum hamiltonian on the cubic lattice reads @xmath244 \label{h3d}\end{aligned}\ ] ] we wish to define a block renormalization rule , where each block of @xmath245 spins ( @xmath246 ; @xmath247 ; @xmath248 ; @xmath249 ; @xmath250 ; @xmath251 ; @xmath252 ; @xmath253 ) will be replaced by a single renormalized spin @xmath254 , after three elementary renormalization steps . it is thus convenient to start by rewriting eq . [ h3d ] as @xmath255 \nonumber \\ & & - \sigma_{(2i,2j-1,2k)}^z [ j_{\vec y}(2i,2j-1,2k ) \sigma_{(2i,2j,2k ) } ^z + j_{\vec x}(2i,2j-1,2k ) \sigma_{(2i+1,2j-1,2k)}^z + j_{\vec z}(2i,2j-1,2k ) \sigma_{(2i,2j-1,2k+1)}^z \nonumber \\ & & + j_{\vec y}(2i,2j-2,2k ) \sigma_{(2i,2j-2,2k)}^z + j_{\vec x}(2i-1,2j-1,2k ) \sigma_{(2i-1,2j-1,2k)}^z + j_{\vec z}(2i,2j-1,2k-1 ) \sigma_{(2i,2j-1,2k-1)}^z ] \nonumber \\ & & - \sigma_{(2i,2j,2k-1)}^z [ j_{\vec z}(2i,2j,2k-1 ) \sigma_{(2i,2j,2k ) } ^z + j_{\vec x}(2i,2j,2k-1 ) \sigma_{(2i+1,2j,2k-1)}^z + j_{\vec y}(2i,2j,2k-1 ) \sigma_{(2i,2j+1,2k-1)}^z \nonumber \\ & & + j_{\vec z}(2i,2j,2k-2 ) \sigma_{(2i,2j,2k-2)}^z + j_{\vec x}(2i-1,2j,2k-1 ) \sigma_{(2i-1,2j,2k-1)}^z + j_{\vec y}(2i,2j-1,2k-1 ) \sigma_{(2i,2j-1,2k-1)}^z ] \nonumber \\ & & - \sigma_{(2i-1,2j-1,2k-1)}^z [ j_{\vec x}(2i-1,2j-1,2k-1 ) \sigma_{(2i-1,2j-1,2k ) } ^z + j_{\vec y}(2i-1,2j-1,2k-1 ) \sigma_{(2i-1,2j,2k-1)}^z \nonumber \\ & & + j_{\vec z}(2i-1,2j-1,2k-1 ) \sigma_{(2i-1,2j-1,2k)}^z + j_{\vec x}(2i-2,2j-1,2k-1 ) \sigma_{(2i-2,2j-1,2k-1)}^z \nonumber \\ & & + j_{\vec y}(2i-1,2j-2,2k-1 ) \sigma_{(2i-1,2j-2,2k-1)}^z + j_{\vec z}(2i-1,2j-1,2k-2 ) \sigma_{(2i-1,2j-1,2k-2)}^z ] \label{h3dbox}\end{aligned}\ ] ] in the first renormalization step , we choose the following intra - block hamiltonian @xmath256 it has the form the hamiltonian of eq . [ hintra ] analyzed in section [ sec_rgrule ] , so that the four spins ( @xmath246 ; @xmath257 ; @xmath248 ; @xmath258 ) are replaced by a single renormalized spin @xmath259 , whereas the four other spins @xmath260 that are not involved in @xmath261 remain unchanged . as a consequence , the projection of the remaining part of the hamiltonian reads @xmath263 p_{intra}^{(1 ) } = \sum_{(i , j ) } h_{i , j , k}^{r } \nonumber \\ & & h_{i , j , k}^{r } = - h^r ( 2i,2j,2k ) \sigma_{r(2i,2j,2k)}^x - h ( 2i-1,2j-1,2k-1 ) \sigma_{(2i-1,2j-1,2k-1)}^x \nonumber \\ & & - h ( 2i-1,2j-1,2k ) \sigma_{(2i-1,2j-1,2k)}^x - h ( 2i,2j-1,2k-1 ) \sigma_{(2i,2j-1,2k-1)}^x - h ( 2i-1,2j,2k-1 ) \sigma_{(2i-1,2j,2k-1)}^x \nonumber \\ & & - \sigma^z_{r(2i,2j,2k ) } [ j_{2\vec x}^r(2i-2,2j,2k ) \sigma_{r(2i-2,2j,2k)}^z + j_{2 \vec y}^r(2i,2j-2,2k ) \sigma_{r(2i,2j-2,2k)}^z + j_{2\vec z}^r(2i,2j,2k-2 ) \sigma_{r(2i,2j,2k-2)}^z \nonumber \\ & & + j^r _ { \vec x- \vec y } ( 2i,2j,2k)\sigma_{2i+1,2j-1,2k}^z + j^r _ { -\vec x+ \vec y } ( 2i,2j,2k ) \sigma_{2i-1,2j+1,2k}^z \nonumber \\ & & + j^r _ { \vec x- \vec z } ( 2i,2j,2k ) \sigma_{2i+1,2j,2k-1}^z + j^r _ { -\vec x+ \vec z } ( 2i,2j,2k ) \sigma_{2i-1,2j,2k+1}^z \nonumber \\ & & + j^r _ { \vec y- \vec z } ( 2i,2j,2k)\sigma_{2i,2j+1,2k-1}^z + j^r _ { -\vec y+ \vec z } ( 2i,2j,2k)\sigma_{2i,2j-1,2k+1}^z \nonumber \\ & & + j^r _ { \vec x+ \vec y } ( 2i-1,2j-1,2k ) \sigma_{2i-1,2j-1,2k}^z + j^r _ { \vec x+ \vec z } ( 2i-1,2j,2k-1 ) \sigma_{2i-1,2j,2k-1}^z \nonumber \\ & & + j^r _ { \vec y+ \vec z } ( 2i,2j-1,2k-1 ) \sigma_{2i,2j-1,2k-1}^z ] \nonumber \\ & & - \sigma_{(2i-1,2j-1,2k-1)}^z [ j_{\vec x}(2i-1,2j-1,2k-1 ) \sigma_{(2i-1,2j-1,2k ) } ^z + j_{\vec y}(2i-1,2j-1,2k-1 ) \sigma_{2i-1,2j,2k-1}^z \nonumber \\ & & + j_{\vec z}(2i-1,2j-1,2k-1 ) \sigma_{2i-1,2j-1,2k}^z + j_{\vec x}(2i-2,2j-1,2k-1 ) \sigma_{(2i-2,2j-1,2k-1)}^z \nonumber \\ & & + j_{\vec y}(2i-1,2j-2,2k-1 ) \sigma_{(2i-1,2j-2,2k-1)}^z + j_{\vec z}(2i-1,2j-1,2k-2 ) \sigma_{(2i-1,2j-1,2k-2)}^z ] \label{h3drfin}\end{aligned}\ ] ] with for the second renormalization step , we choose the following intra - block hamiltonian @xmath274 it has the form the hamiltonian of eq . [ hintra ] analyzed in section [ sec_rgrule ] , so that the four spins ( @xmath275 ; @xmath276 ; @xmath277 ; @xmath278 ) are replaced by a single renormalized spin @xmath279 . the application of the projection rules of eqs [ projsigmaz0 ] , [ projsigmazi ] and [ projsigmax0 ] read for the present case @xmath280 ^ 2 } } \sigma^z_{rr(2i,2j,2k ) } \nonumber \\ & & p_{intra}^{(2 ) } \sigma^z_{2i-1,2j,2k-1 } p_{intra}^{(2 ) } = \frac { j^r_{\vec x+\vec z } ( 2i-1,2j,2k-1 ) } { \sqrt { h^2(2i-1,2j,2k-1)+[j^r_{\vec x+\vec z } ( 2i-1,2j,2k-1)]^2 } } \sigma^z_{rr(2i,2j,2k ) } \nonumber \\ & & p_{intra}^{(2 ) } \sigma^z_{2i,2j-1,2k-1 } p_{intra}^{(2 ) } = \frac { j^r_{\vec y+\vec z } ( 2i,2j-1,2k-1 ) } { \sqrt { h^2(2i,2j-1,2k-1)+[j^r_{\vec y+\vec z } ( 2i,2j-1,2k-1)]^2 } } \sigma^z_{rr(2i,2j,2k ) } \nonumber \\ & & p_{intra}^{(2 ) } \sigma^x_{r(2i,2j,2k ) } p_{intra}^{(2 ) } = \frac{h ( 2i-1,2j-1,2k)}{\sqrt{h^2(2i-1,2j-1,2k)+[j^r_{\vec x+\vec y } ( 2i-1,2j-1,2k)]^2 } } \nonumber \\ & & \frac{h ( 2i-1,2j,2k-1)}{\sqrt{h^2(2i-1,2j,2k-1)+[j^r_{\vec x+\vec z } ( 2i-1,2j,2k-1)]^2 } } \frac{h ( 2i,2j-1,2k-1)}{\sqrt{h^2(2i,2j-1,2k-1)+[j^r_{\vec y+\vec z } ( 2i,2j-1,2k-1)]^2 } } \sigma_{rr(2i,2j)}^x \nonumber\end{aligned}\ ] ] as a consequence , the projection of the remaining part of the hamiltonian reads @xmath281 p_{intra}^{(2 ) } = \sum_{(i , j ) } h_{i , j , k}^{rr } \nonumber \\ & & h_{i , j , k}^{rr } = - h^{rr } ( 2i,2j,2k ) \sigma_{rr(2i,2j,2k)}^x - h ( 2i-1,2j-1,2k-1 ) \sigma_{(2i-1,2j-1,2k-1)}^x \nonumber \\ & & - j_{2\vec x}^{rr}(2i-2,2j,2k ) \sigma_{rr(2i-2,2j,2k)}^z \sigma^z_{rr(2i,2j,2k ) } - j_{2 \vec y}^{rr}(2i,2j-2,2k ) \sigma_{rr(2i,2j-2,2k)}^z \sigma^z_{rr(2i,2j,2k ) } \nonumber \\ & & - j_{2\vec z}^{rr}(2i,2j,2k-2 ) \sigma_{rr(2i,2j,2k-2)}^z \sigma^z_{rr(2i,2j,2k ) } \nonumber \\ & & - j^{rr}_{\vec x-\vec y-\vec z } ( 2i,2j,2k ) \sigma^z_{rr(2i,2j,2k ) } \sigma_{(2i+1,2j-1,2k-1)}^z -j^{rr}_{-\vec x+\vec y-\vec z } ( 2i,2j,2k ) \sigma^z_{rr(2i,2j,2k ) } \sigma_{(2i-1,2j+1,2k-1)}^z \nonumber \\ & & - j^{rr}_{-\vec x-\vec y+\vec z } ( 2i,2j,2k ) \sigma^z_{rr(2i,2j,2k ) } \sigma_{(2i-1,2j-1,2k+1)}^z \nonumber \\ & & - j^{rr}_{\vec x+\vec y+\vec z } ( 2i-1,2j-1,2k-1 ) \sigma_{(2i-1,2j-1,2k-1)}^z \sigma^z_{rr(2i,2j,2k ) } \label{h3drrfin}\end{aligned}\ ] ] in terms of \(i ) the renormalized transverse fields @xmath282 ^ 2 } } \nonumber \\ & & \frac{h ( 2i-1,2j,2k-1)}{\sqrt{h^2(2i-1,2j,2k-1)+[j^r_{\vec x+\vec z}(2i-1,2j,2k-1)]^2 } } \frac{h ( 2i,2j-1,2k-1)}{\sqrt{h^2(2i,2j-1,2k-1)+[j^r_{\vec y+\vec z}(2i,2j-1,2k-1)]^2 } } \label{rghrr3d}\end{aligned}\ ] ] \(ii ) the renormalized couplings along the lattice directions at distance two @xmath283 ^ 2 + [ j^r _ { \vec x+ \vec y}(2i-1,2j-1,2k)]^2 } } \nonumber \\ & & + j^r _ { \vec x- \vec z } ( 2i-2,2j,2k ) \frac{j^r _ { \vec x+ \vec z } ( 2i-1,2j,2k-1)}{\sqrt{[h^r(2i-1,2j,2k-1)]^2 + [ j^r _ { \vec x+ \vec z}(2i-1,2j,2k-1)]^2 } } \nonumber \\ j^{rr}_{2 \vec y } ( 2i,2j-2,2k ) & & = j^r_{2 \vec y}(2i,2j-2,2k ) \nonumber \\ & & + j^r _ { -\vec x+ \vec y } ( 2i,2j-2,2k ) \frac{j^r _ { \vec x+ \vec y } ( 2i-1,2j-1,2k ) } { \sqrt{[h^r(2i-1,2j-1,2k)]^2+[j^r _ { \vec x+ \vec y}(2i-1,2j-1,2k)]^2 } } \nonumber \\ & & + j^r _ { \vec y- \vec z } ( 2i,2j-2,2k ) \frac{j^r _ { \vec y+ \vec z } ( 2i,2j-1,2k-1 ) } { \sqrt{[h^r(2i,2j-1,2k-1)]^2+[j^r _ { \vec y+ \vec z}(2i,2j-1,2k-1)]^2 } } \nonumber \\ j^{rr}_{2 \vec z } ( 2i,2j,2k-2 ) & & = j^r_{2 \vec z}(2i,2j,2k-2 ) \nonumber \\ & & + j^r _ { -\vec x+ \vec z } ( 2i,2j,2k-2 ) \frac{j^r _ { \vec x+ \vec z } ( 2i-1,2j,2k-1 ) } { \sqrt{[h^r(2i-1,2j,2k-1)]^2+[j^r _ { \vec x+ \vec z}(2i-1,2j,2k-1)]^2 } } \nonumber \\ & & + j^r _ { -\vec y+ \vec z } ( 2i,2j,2k-2 ) \frac{j^r _ { \vec y+ \vec z } ( 2i,2j-1,2k-1 ) } { \sqrt{[h^r(2i,2j-1,2k-1)]^2+[j^r _ { \vec y+ \vec z}(2i,2j-1,2k-1)]^2 } } \label{rgj3ddrrdiagneg}\end{aligned}\ ] ] \(iii ) the renormalized couplings along @xmath284 ; @xmath285 ; @xmath285 @xmath286 ^ 2 } } \nonumber \\ j^{rr}_{-\vec x+\vec y-\vec z } ( 2i,2j,2k ) & & = j_{\vec y}(2i-1,2j,2k-1 ) \frac { j^r_{\vec x+\vec z } ( 2i-1,2j,2k-1 ) } { \sqrt { h^2(2i-1,2j,2k-1)+[j^r_{\vec x+\vec z } ( 2i-1,2j,2k-1)]^2 } } \nonumber \\ j^{rr}_{-\vec x-\vec y+\vec z } ( 2i,2j,2k ) & & = j_{\vec z}(2i-1,2j-1,2k ) \frac { j^r_{\vec x+\vec y } ( 2i-1,2j-1,2k ) } { \sqrt { h^2(2i-1,2j-1,2k)+[j^r_{\vec x+\vec y } ( 2i-1,2j-1,2k)]^2 } } \label{proj3djrrdiagneg}\end{aligned}\ ] ] \(iv ) the renormalized couplings along @xmath287 @xmath288 ^ 2 } } \nonumber \\ & & + j_{\vec y}(2i-1,2j-1,2k-1 ) \frac { j^r_{\vec x+\vec z } ( 2i-1,2j,2k-1 ) } { \sqrt { h^2(2i-1,2j,2k-1)+[j^r_{\vec x+\vec z } ( 2i-1,2j,2k-1)]^2 } } \nonumber \\ & & + j_{\vec z}(2i-1,2j-1,2k-1)\frac { j^r_{\vec x+\vec y } ( 2i-1,2j-1,2k ) } { \sqrt { h^2(2i-1,2j-1,2k)+[j^r_{\vec x+\vec y } ( 2i-1,2j-1,2k)]^2 } } \label{proj3drrdiagpos}\end{aligned}\ ] ] for the third renormalization step , we choose the following intra - block hamiltonian @xmath289 it has the form the hamiltonian of eq . [ hintra ] analyzed in section [ sec_rgrule ] , so that the two spins @xmath290 can be replaced by a single renormalized spin @xmath291 . the application of the projection rules of eqs [ projsigmaz0 ] , [ projsigmazi ] and [ projsigmax0 ] read for the present case @xmath292 ^ 2 } } \sigma^z_{rrr(2i,2j,2k ) } \nonumber \\ p_{intra}^{(3 ) } \sigma^x_{rr(2i,2j,2k ) } p_{intra}^{(2 ) } & & = \frac { h ( 2i-1,2j-1,2k-1 ) } { \sqrt { h^2(2i-1,2j-1,2k-1)+[j^{rr}_{\vec x+\vec y+\vec z } ( 2i-1,2j-1,2k-1)]^2 } } \sigma_{rrr(2i,2j,2k)}^x \label{proj3dp3}\end{aligned}\ ] ] the projection of the remaining part of the hamiltonian reads @xmath293 p_{intra}^{(3 ) } = \sum_{(i , j , k ) } h_{i , j , k}^{rrr } \nonumber \\ h_{i , j , k}^{rrr } = & & - h^{rrr } ( 2i,2j,2k ) \sigma_{rrr(2i,2j,2k)}^x \nonumber \\ & & - j_{2\vec x}^{rrr}(2i-2,2j,2k ) \sigma_{rrr(2i-2,2j,2k)}^z \sigma^z_{rrr(2i,2j,2k ) } - j_{2 \vec y}^{rrr}(2i,2j-2,2k ) \sigma_{rrr(2i,2j-2,2k)}^z \sigma^z_{rrr(2i,2j,2k ) } \nonumber \\ & & - j_{2\vec z}^{rrr}(2i,2j,2k-2 ) \sigma_{rrr(2i,2j,2k-2)}^z \sigma^z_{rrr(2i,2j,2k ) } \label{h3drrrfin}\end{aligned}\ ] ] i.e. it has the same form as the initial hamiltonian , in terms of @xmath295 ^ 2 } } \nonumber \\ j^{rrr}_{2 \vec y } ( 2i,2j-2,2k ) & & = j^{rr}_{2 \vec y}(2i,2j-2,2k ) \nonumber \\ & & + j^{rr}_{-\vec x+\vec y-\vec z } ( 2i,2j-2,2k ) \frac { j^{rr}_{\vec x+\vec y+\vec z } ( 2i-1,2j-1,2k-1 ) } { \sqrt { h^2(2i-1,2j-1,2k-1)+[j^{rr}_{\vec x+\vec y+\vec z } ( 2i-1,2j-1,2k-1)]^2 } } \nonumber \\ j^{rrr}_{2 \vec z } ( 2i,2j,2k-2 ) & & = j^{rr}_{2 \vec z}(2i,2j,2k-2 ) \nonumber \\ & & + j^{rr}_{-\vec x-\vec y+\vec z } ( 2i,2j,2k-2 ) \frac { j^{rr}_{\vec x+\vec y+\vec z } ( 2i-1,2j-1,2k-1 ) } { \sqrt { h^2(2i-1,2j-1,2k-1)+[j^{rr}_{\vec x+\vec y+\vec z } ( 2i-1,2j-1,2k-1)]^2 } } \label{rgj3ddrrrv}\end{aligned}\ ] ] if we start from the pure model of parameters @xmath92 , the renormalization rules of eqs [ rghrrr ] and [ rgj3ddrrrv ] reads for the ratio @xmath94 to @xmath296 \equiv \phi(k ) \label{evolk3d}\end{aligned}\ ] ] the critical point satisfying @xmath145 is found to be @xmath297 the correlation length exponent @xmath26 given by @xmath298 @xmath299 is very close to the mean - field value @xmath300 of the 4d classical ising model . e. altman , y. kafri , a. polkovnikov and g. refael , phys . 100 , 170402 ( 2008 ) ; + e. altman , y. kafri , a. polkovnikov and g. refael , phys . rev . b 81 , 175528 ( 2010 ) ; + s. iyer , d. pekker and g. refael , phys . rev . b 85 , 094202 ; + g. refael and e. altman , comptes rendus physique 14 , 725 ( 2013 ) .

for the quantum ising chain , the self - dual block renormalization procedure of fernandez - pacheco [ phys . rev . d 19 , 3173 ( 1979 ) ] is known to reproduce exactly the location of the zero - temperature critical point and the correlation length exponent @xmath0 . recently , miyazaki and nishimori [ phys . rev . e 87 , 032154 ( 2013 ) ] have proposed to study the disordered quantum ising model in dimensions @xmath1 by applying the fernandez - pacheco procedure successively in each direction . to avoid the inequivalence of directions of their approach , we propose here an alternative procedure where the @xmath2 directions are treated on the same footing . for the pure model , this leads to the correlation length exponents @xmath3 in @xmath4 ( to be compared with the 3d classical ising model exponent @xmath5 ) and @xmath6 ( to be compared with the 4d classical ising model mean - field exponent @xmath7 ) . for the disordered model in dimension @xmath4 , either ferromagnetic or spin - glass , the numerical application of the renormalization rules to samples of linear size @xmath8 yields that the transition is governed by an infinite disorder fixed point , with the activated exponent @xmath9 , the typical correlation exponent @xmath10 and the finite - size correlation exponent @xmath11 . we discuss the similarities and differences with the strong disorder renormalization results . # 1#2 # 1#2

introduction elementary renormalization rule reminder on the application in dimension @xmath18 block renormalization rules in dimension @xmath4 application to the pure two-dimensional quantum ising model application to the random ferromagnetic two-dimensional quantum ising model application to the two-dimensional spin-glass quantum ising model conclusion renormalization rules in @xmath29

this paper is about a geometric view of the relationship between continued fractions and approximation of real numbers by rationals . whenever we speak of a rational @xmath1 we mean that @xmath2 and @xmath3 are coprime integers and @xmath3 is positive . given a real number @xmath0 we follow khinchin ( * ? ? ? * section 6 ) in describing a rational @xmath4 as a _ best approximation of the second kind _ of @xmath0 provided that , for each rational @xmath5 such that @xmath6 , we have @xmath7 with equality if and only if @xmath8 . khinchin also defines _ best approximation of the first kind _ , but that concept does not concern us here . a _ continued fraction _ is an expression of the form @xmath9 where @xmath10 is an integer and the other @xmath11 are positive integers . either the sequence @xmath12 is infinite , in which case the continued fraction is said to be _ infinite _ , or there is a final member @xmath13 of this sequence , in which case the continued fraction is said to be _ finite_. each of our finite continued fractions with @xmath14 is assumed to satisfy @xmath15 . convergents _ of @xmath0 are the rationals @xmath16 where , @xmath17 the _ value _ of a finite continued fraction is the value of the final convergent @xmath18 ( which is a rational number ) and the _ value _ of an infinite continued fraction is the limit of the sequence @xmath16 ( which is an irrational number ) . to each real number @xmath0 there corresponds a unique continued fraction with value @xmath0 . all these facts about continued fractions can be found in @xcite , as can the next theorem ( ( * ? ? ? * theorems 16 and 17 ) ) . [ h ] a rational @xmath19 , which is not an integer , is a convergent of a real number @xmath0 if and only if it is a best approximation of the second kind of @xmath0 . it is convenient to assume that @xmath19 is not an integer in theorem [ h ] , and later on , to avoid tiresome discussions of this trivial case . theorem [ h ] fails when @xmath19 is an integer , because if @xmath20 , for some integer @xmath21 , then @xmath21 is a convergent of @xmath0 , but not a best approximation of the second kind of @xmath0 . a version of theorem [ h ] including the possibility that @xmath19 is an integer can be found in ( * ? ? ? * theorem 1 ) . classic proofs of theorem [ h ] , such as that given in @xcite , are algebraic . irwin proves theorem [ h ] using plane lattices in @xcite . our aim is to give an illuminating geometric proof based on the theory of ford circles . ford circles , developed by ford in @xcite , are objects most naturally associated with hyperbolic geometry , and our proof has undertones of hyperbolic geometry . we now give a brief description of ford circles and their relationship to continued fractions ( full details can be found in @xcite ) . the _ ford circle _ @xmath22 of a rational number @xmath23 is the circle in the complex plane with centre @xmath24 and radius @xmath25 . this circle is tangent to the real axis at @xmath19 , and otherwise lies in the upper half - plane . a selection of ford circles are shown in figure [ s ] . two ford circles @xmath22 and @xmath26 , where @xmath23 and @xmath27 , are tangent if and only if @xmath28 , and if they are not tangent then they are wholly external to one another ( ford circles do not overlap ) . we define the _ continued fraction chain _ of a real number @xmath0 to be the sequence of ford circles @xmath29 , where @xmath16 are the convergents of @xmath0 . since @xmath30 we see that each pair of consecutive circles in the continued fraction chain of @xmath0 are tangent . also , the @xmath31 are given by the recurrence relation @xmath32 , @xmath33 , and @xmath34 for @xmath35 , which means that the sequence @xmath36 of positive integers is increasing . consequently , the sequence of radii @xmath37 of the circles from the continued fraction chain is decreasing . finally , the members of the continued fraction chain alternate from the left to the right side of @xmath0 , because @xmath38 the first few ford circles from a continued fraction chain are shown in figure [ t ] ( in black ) . a circle @xmath39 that is tangent to the real axis at a point @xmath40 , and that otherwise lies in the upper half - plane , is called a _ horocircle_. we denote the radius of @xmath39 by @xmath41 $ ] , and describe the point @xmath40 as the _ base point _ of @xmath39 . in order to state our geometric version of theorem [ h ] , we introduce a new definition . in this definition we use the simple fact that , given a real number @xmath40 and a horocircle @xmath39 , there is a unique horocircle @xmath42 that is tangent to @xmath39 and has base point @xmath40 . if @xmath39 has base point @xmath40 then we consider @xmath42 to have radius @xmath43 . given a real number @xmath0 we say that a ford circle @xmath22 is _ nearby _ to @xmath0 if , for each ford circle @xmath44 other than @xmath22 with @xmath45{\geqslant}\textnormal{rad}[c_x]$ ] , the radius of the unique horocircle tangent to @xmath44 and with base point @xmath0 is larger than the radius of the unique horocircle tangent to @xmath22 and with base point @xmath0 . when @xmath0 is rational , @xmath46 is nearby to @xmath0 , but no ford circle with equal or smaller radius than @xmath46 is nearby to @xmath0 . [ u ] let @xmath0 be a real number . given a rational @xmath19 , which is not an integer , the following are equivalent : 1 . @xmath19 is a convergent of @xmath0 ; 2 . @xmath22 is a member of the continued fraction chain of @xmath0 ; 3 . @xmath19 is a best approximation of the second kind of @xmath0 ; 4 . @xmath22 is nearby to @xmath0 ; 5 . there is a ford circle @xmath26 tangent to @xmath22 such that @xmath47 > \textnormal{rad}[c_y]$ ] , and either @xmath48 or @xmath0 lies in the open interval bounded by @xmath19 and @xmath49 . statement ( v ) is illustrated in figure [ z ] . ( v ) . ] statement ( ii ) of theorem [ u ] is merely a geometric reformulation of statement ( i ) , and in the next section we see that statement ( iv ) is a geometric reformulation of statement ( iii ) . the equivalence of statements ( i ) and ( iii ) yields theorem [ h ] . the key idea in this paper is about explaining best approximation of the second kind in terms of ford circles so that , using ford s continued fraction chains , we can prove theorem [ h ] geometrically . our key idea is encapsulated in the next proposition . [ d ] given a real number @xmath0 , the rational @xmath19 is a best approximation of the second kind of @xmath0 if and only if @xmath50 is nearby to @xmath0 . to prove proposition [ d ] we need the next lemma and corollary . [ a ] two horocircles @xmath39 and @xmath42 with radii @xmath51 and @xmath52 and distinct base points @xmath19 and @xmath49 , which intersect in at most one point , satisfy @xmath53 , with equality if and only if @xmath39 and @xmath42 are tangent . . ] let @xmath54 be the distance between the centres of the two horocircles . then @xmath55 , with equality if and only if @xmath39 and @xmath42 are tangent . we can calculate @xmath54 by applying pythagoras s theorem to the triangle with vertices @xmath56 , @xmath57 , and @xmath58 , and the result follows immediately . [ b ] the radius of the horocircle that is tangent to the ford circle @xmath59 , and has base point @xmath0 , is @xmath60 this corollary follows from lemma [ a ] , because @xmath59 has radius @xmath25 . for each rational @xmath40 , let @xmath61 denote the unique horocircle with base point @xmath0 that is tangent to @xmath44 . now , a rational @xmath23 is a best approximation of the second kind of @xmath0 if and only if for each rational @xmath27 distinct from @xmath19 and such that @xmath6 we have @xmath62 . equivalently , using corollary [ b ] , for each ford circle @xmath26 distinct from @xmath22 and such that @xmath63{\geqslant}\textnormal{rad}[c_x]$ ] we have @xmath64 > \textnormal{rad}[d_x]$ ] . in other words @xmath19 is a best approximation of the second kind of @xmath0 if and only if @xmath22 is nearby to @xmath0 . this section contains two elementary lemmas about basic properties of ford circles . [ x ] let @xmath22 and @xmath26 be tangential ford circles . if a rational @xmath40 lies strictly between @xmath19 and @xmath49 then @xmath44 has smaller radius than both @xmath22 and @xmath26 . . ] let @xmath22 , @xmath26 , and @xmath44 have radii @xmath65 , @xmath66 , and @xmath67 . by lemma [ a ] @xmath68 hence @xmath69 and similarly @xmath70 . [ q ] let @xmath22 and @xmath26 be tangential ford circles such that @xmath47>\textnormal{rad}[c_y]$ ] , and suppose that a real number @xmath0 lies strictly between @xmath19 and @xmath49 , and a rational @xmath40 lies strictly outside the interval bounded by @xmath19 and @xmath49 . then the radius of the horocircle that is tangent to @xmath22 and has base point @xmath0 is smaller than the radius of the horocircle that is tangent to @xmath44 and has base point @xmath0 . . ] let @xmath22 , @xmath26 , and @xmath44 have radii @xmath65 , @xmath66 , and @xmath67 . denote the radius of the horocircle that is tangent to @xmath22 , and has base point @xmath0 , by @xmath71 , and denote the radius of the horocircle that is tangent to @xmath44 , and has base point @xmath0 , by @xmath72 . by lemma [ a ] we have @xmath73 from which it follows that @xmath74 by lemma [ a ] we also have @xmath75 depending on the order of @xmath19 , @xmath49 , and @xmath40 we either have @xmath76 or @xmath77 in both cases we obtain @xmath78 combining and we conclude that @xmath79 . [ y ] let @xmath22 and @xmath26 be tangential ford circles such that @xmath47>\textnormal{rad}[c_y]$ ] , and suppose that a real number @xmath0 lies strictly between @xmath19 and @xmath49 . then @xmath22 is nearby to @xmath0 . choose a ford circle @xmath44 distinct from @xmath22 with @xmath45{\geqslant}\textnormal{rad}[c_x]$ ] . by lemma [ x ] , @xmath40 lies strictly outside the interval bounded by @xmath19 and @xmath49 . by lemma [ q ] , the radius of the horosphere based at @xmath0 and tangent to @xmath22 is smaller than the radius of the horosphere based at @xmath0 and tangent to @xmath44 . hence @xmath22 is nearby to @xmath0 . we can now prove theorem [ u ] . in our proof we denote the convergents of @xmath0 by @xmath80 , and we define @xmath81 to be the radius @xmath82 of @xmath83 , so that @xmath84 is a strictly decreasing sequence . statements ( i ) and ( ii ) of theorem [ u ] are equivalent by the definition of a continued fraction chain . statements ( iii ) and ( iv ) are equivalent because of proposition [ d ] . we proceed to prove that ( i ) implies ( v ) , ( v ) implies ( iv ) , and ( iv ) implies ( i ) . it is convenient to first dismiss the two cases when @xmath0 is rational , and @xmath19 is either the last or penultimate convergent of @xmath0 ( namely @xmath18 or @xmath85 ) . if @xmath86 then ( i ) and ( iv ) are satisfied by definition , and ( v ) is also satisfied by choosing any ford circle @xmath26 tangent to @xmath22 such that @xmath63<\textnormal{rad}[c_x]$ ] . suppose that @xmath87 . again , ( i ) holds by definition . let @xmath88 and @xmath89 . this pair are coprime because @xmath90 , and because @xmath91 and @xmath15 , we see that @xmath92 . let @xmath93 . notice that @xmath94 lies strictly between @xmath19 and @xmath49 . therefore ( v ) is satisfied , and ( iv ) is satisfied by corollary [ y ] . henceforth we assume that , when @xmath0 is rational , @xmath95 and @xmath96 . now we show that ( i ) implies ( v ) . suppose that @xmath97 for some @xmath98 . define @xmath99 . by , @xmath0 lies strictly between @xmath19 and @xmath49 , and since @xmath100 is decreasing we have that @xmath47>\textnormal{rad}[c_y]$ ] . next we show that ( v ) implies ( iv ) . suppose that ( v ) holds . the case @xmath101 has already been dealt with , hence @xmath0 lies strictly between @xmath19 and @xmath49 , and it follows from corollary [ y ] that @xmath22 is nearby to @xmath0 . last we show that ( iv ) implies ( i ) . suppose that @xmath22 is nearby to @xmath0 . the decreasing sequence @xmath84 has limit @xmath43 if @xmath0 is irrational , and limit @xmath102 $ ] if @xmath0 is rational . in the latter case , as @xmath101 has been considered already , we have @xmath47>\textnormal{rad}[c_\alpha]$ ] . since @xmath103 the greatest possible radius of a ford circle either ( a ) @xmath104 and there is a unique integer @xmath105 such that @xmath106 > r_{n+1}$ ] , or ( b ) @xmath107 and there is a unique integer @xmath108 such that @xmath106 > r_{n+1}$ ] . also , since @xmath19 is neither the last nor the penultimate convergent of @xmath0 , we see that @xmath0 lies strictly between @xmath16 and @xmath109 . if @xmath110 then , by lemma [ x ] , @xmath19 lies outside the closed interval bounded by @xmath16 and @xmath109 ; however , this can not be , because the assumption that @xmath22 is nearby to @xmath0 then contradicts lemma [ q ] . hence @xmath97 ( and case ( b ) with @xmath111 can not arise because @xmath19 is not an integer ) . let @xmath112 denote the point @xmath113 in three - dimensional euclidean space . the _ ford sphere _ @xmath114 of @xmath23 , where @xmath115 and @xmath116 are coprime gaussian integers , is the sphere with centre @xmath117 and radius @xmath118 . ford spheres share many properties with ford circles , and they can be used in the study of gaussian integer continued fraction expansions of complex numbers . a brief account can be found at the end of @xcite . it would be of interest to investigate whether the techniques of this paper can be applied to ford spheres and gaussian integer continued fractions to give results on approximation of complex numbers by quotients of gaussian integers .

we give an elementary geometric proof using ford circles that the convergents of the continued fraction expansion of a real number @xmath0 coincide with the rationals that are best approximations of the second kind of @xmath0 .

introduction best approximation of the second kind ford circles proof of theorem[u] concluding remarks

a rational motion is a motion with only rational trajectories . in the dual quaternion model of @xmath0 , the group of rigid body displacements , ( 9 ) it is described by a rational curve on the study quadric @xcite . in this article we construct a rational motion of minimal degree in the dual quaternion model with a given rational curve as trajectory , and we show that this motion is unique up to coordinate changes . this is an interesting result in its own right but it also has a certain potential for applications in computer graphics , computer aided design or mechanism science . usually , one defines the degree of a rational motion as the maximal degree of a trajectory @xcite . with this concept of motion degree , our problem becomes trivial as the curvilinear translation along the curve is already minimal . as we shall see , it is also minimal with respect to the dual quaternion degree if the prescribed trajectory is generic . the situation changes , however , if the trajectory is _ circular _ , that is , it intersects the absolute circle at infinity . in this case , the minimal achievable degree in the dual quaternion model is the curve degree minus half the number of conjugate complex intersection points with the absolute circle at infinity ( the curve s _ circularity _ ) . we will see that twice the circularity of a trajectory equals the trajectory degree minus the degree defect in the spherical component of the minimal motion . this leads to the rather strange observation that generic rational motions ( without spherical degree defect ) have very special ( entirely circular ) trajectories . conversely , the minimal motion to generic ( non - circular ) curves are curvilinear translations which are special in the sense that their spherical degree defect is maximal . we continue this article with an introduction to the dual quaternion model of rigid body displacements in . there we also introduce motion polynomials and their relation to rational motions . our results are formulated and proved in . the proof of the central result ( ) is constructive and can be used to actually compute the minimal rational motion by a variant of the euclidean algorithm . we illustrate this procedure by two examples . in this article , we work in the dual quaternion model of the group of rigid body displacements . this model requires a minimal number of parameters while retaining a bilinear composition law . moreover , it provides a rich algebraic and geometric structure . it is , for example , possible to use a variant of the euclidean algorithm for computing the greatest common divisor ( gcd ) of two polynomials . this section presents the necessary theoretical background on dual quaternions . the set @xmath1 of dual quaternions is an eight - dimensional associative algebra over the real numbers . it is generated by the base elements @xmath2 and the non - commutative multiplication is determined by the relations @xmath3 as important sub - algebras , the algebra of dual quaternions contains the real numbers @xmath4 , the complex numbers @xmath5 , the dual numbers @xmath6 , and the quaternions @xmath7 ( angled brackets denote a linear span ) . a dual quaternion may be written as @xmath8 where @xmath9 , @xmath10 are quaternions . the conjugate dual quaternion is @xmath11 and conjugation of quaternions is done by multiplying the coefficients of @xmath12 , @xmath13 , and @xmath14 with @xmath15 . it can readily be verified that the dual quaternion norm , defined as @xmath16 , equals @xmath17 . it is a dual number . the non - invertible dual quaternions @xmath8 are precisely those with vanishing primal part @xmath18 . an important application of dual quaternions is the modelling of rigid body displacements . the group of dual quaternions of unit norm modulo @xmath19 is isomorphic to @xmath0 , the group of rigid body displacements . a unit dual quaternion @xmath8 acts on a point @xmath20 in the three dimensional real vector space @xmath21 according to @xmath22 note that @xmath23 because of the unit norm condition . it is convenient and customary to projectivise the space @xmath1 of dual quaternions , thus arriving at @xmath24 , the real projective space of dimension seven . then , the unit norm condition can be relaxed to the non - vanishing of @xmath25 and the vanishing of @xmath26 . in a geometric language , @xmath0 is isomorphic to the points of a quadric @xmath27 , defined by @xmath28 , minus the points of a three - dimensional space , defined by @xmath18 . the quadric @xmath29 is called the _ study quadric . _ in this setting , the map becomes @xmath30 the action of @xmath8 with @xmath31 , @xmath28 can be extended to real projective three - space @xmath32 , modelled as projective space over @xmath33 . the point @xmath20 represented by @xmath34 is mapped according to @xmath35 this is a convenient representation for studying rational curves as trajectories of rational motions . in the projective setting , a rational motion is simply a curve in the study quadric @xmath29 that admits a parameterisation by a polynomial @xmath36 with dual quaternion coefficients @xmath37 . the non - commutativity of @xmath1 necessitates some rules concerning notation and multiplication : polynomial multiplication is defined by the convention that the indeterminate commutes with all coefficients , the ring of these polynomials in the indeterminate @xmath38 is denoted by @xmath39 $ ] , the subring of polynomials with quaternion coefficients is @xmath40 $ ] . we always write coefficients to the left of the indeterminate @xmath38 . this convention is sometimes captured in the term `` left polynomial '' but we will just speak of a `` polynomial '' . evaluating @xmath41 at different values @xmath42 gives points of a rational curve in @xmath24 . we also define @xmath43 in order to obtain the complete curve as image of @xmath44 under the map @xmath45 . a similar convention is used for rational curves in @xmath32 . the conjugate to the polynomial is @xmath46 it can readily be verified that the _ norm polynomial _ @xmath47 has dual numbers as coefficients . if @xmath47 has even _ real _ coefficients and the leading coefficient @xmath48 is invertible , we call @xmath41 a _ motion polynomial . _ this is motivated by the observation that the curve parameterised by a motion polynomial is contained in the study quadric , that is , it parameterises indeed a rigid body motion . writing @xmath49 with @xmath50 , @xmath51 $ ] , motion polynomials are characterised by the vanishing of @xmath52 . the trajectory of the point @xmath53 is @xmath54 this is a curve parameterised by polynomial functions in projective coordinates . it is also called a _ rational curve _ because it is always possible to clear denominators of rational functions . all motions in @xmath0 with only rational trajectories can be parameterised by motion polynomials @xcite . a rational parameterised equation @xmath55 with @xmath56 , @xmath57 , @xmath58 , @xmath59 $ ] is called _ reduced _ if the greatest common divisor @xmath60 of @xmath56 , @xmath57 , @xmath58 , @xmath61 has degree zero . the degree of @xmath20 is defined as the maximal degree of @xmath56 , @xmath57 , @xmath58 , @xmath61 . if @xmath20 is not reduced , we may divide it by @xmath60 to obtain an equivalent parameterised equation which describes the same rational curve , but possibly with fewer parameterisation singularities . given a reduced parameterised equation @xmath20 of degree @xmath62 , it is our ultimate aim to find a motion polynomial @xmath63 $ ] of minimal degree such that the trajectory of one point @xmath9 is parameterised by @xmath20 . an obvious example of a motion polynomial with trajectory @xmath20 is the curvilinear translation along @xmath20 and it will turn out that this is already the solution to our problem in generic cases . however , for trajectories which are non - generic , in a sense to be made precise below , a lesser degree can be achieved . a motion polynomial @xmath49 is called _ reduced , _ if both @xmath50 and @xmath64 have no common real factor . the real factor of the primal part @xmath50 with maximal degree is uniquely defined up to a constant scalar factor . we call it _ maximal real polynomial factor _ and abbreviate it by `` mrpf '' . it accounts for a difference in degrees between the rational motion , parameterized by @xmath49 , and its spherical component , parameterised by @xmath50 . hence , we also call the degree of the maximal real polynomial factor of @xmath50 the _ spherical degree defect _ of @xmath41 . we only consider reduced rational motions but do not exclude motions with positive spherical degree defect . it will turn out that the spherical degree defect and circularity of trajectories are closely related : [ def:1 ] the _ circularity _ of a rational curve with reduced rational parameterised equation @xmath65 is defined as @xmath66 the curve is called _ entirely circular _ if it is of maximal circularity @xmath67 . geometrically , circularity is half the number of ( conjugate complex ) intersection points of @xmath20 and the absolute circle at infinity , counted with their respective algebraic multiplicities . hence , it is always an integer ( this also follows algebraically from the fact that @xmath56 , @xmath57 , @xmath58 , @xmath61 are real and prime ) and does not depend on the chosen parameterisation , as long as it is reduced . [ th:1 ] let @xmath41 be a reduced motion polynomial of degree @xmath68 and spherical degree defect @xmath69 . then a trajectory of @xmath41 is of degree @xmath70 and circularity @xmath71 . [ rem:1 ] in , the strictly less and strictly greater cases can occur . one example of this is the general darboux motion @xcite where @xmath72 , @xmath73 , @xmath74 and hence @xmath75 ( but @xmath76 ) . another example is the circular translation motion @xcite with @xmath77 , @xmath78 . here @xmath79 but @xmath80 . [ lem:1 ] assume @xmath81 $ ] . if a monic quadratic irreducible real polynomial @xmath64 is a factor of the product @xmath82 , then @xmath64 either divides @xmath83 or @xmath64 divides @xmath84 or there is a unique quaternion @xmath85 and two quaternion polynomials @xmath86 and @xmath87 such that @xmath88 , @xmath89 and @xmath90 . it is sufficient to prove the third case under the assumption @xmath91 where @xmath92 , @xmath93 . there exist uniquely determined quaternions @xmath94 , @xmath95 such that * @xmath96 for @xmath97 and @xmath98 , * @xmath99 , @xmath100 where @xmath101 , @xmath102 , and * @xmath103 . this follows from the quaternion version of the factorisation theorem for motion polynomials ( * ? ? ? * theorem 1 ) . as @xmath64 divides @xmath82 , we get that @xmath64 divides @xmath104 . but then @xmath64 also divides @xmath105 . this can happen only if two neighbouring linear factors are conjugate and , because of @xmath106 , implies @xmath107 whence the claim follows with @xmath108 . let @xmath109 be the translation which moves a point @xmath110 to the origin @xmath111 and let @xmath112 . then the orbit of @xmath20 with respect to @xmath41 is equal to the orbit of the origin @xmath111 with respect to @xmath113 . because we have @xmath114 and the spherical degree defect of @xmath41 equals the spherical degree defect of @xmath113 , it is sufficient to prove the statement for @xmath113 and the origin @xmath111 ( instead of @xmath41 and @xmath20 ) . from now on , we use the symbol @xmath41 to abbreviate @xmath113 . let @xmath115 , set @xmath116 , @xmath117 , @xmath118 and use to compute a parametric equation @xmath119 for the trajectory of the origin @xmath111 : @xmath120 where we write `` @xmath121 '' for equality in the projective sense , modulo scalar ( or real polynomial ) multiplication . the trajectory s degree is not larger than @xmath122 . with @xmath123 and @xmath124 , the degree of the trajectory @xmath119 is @xmath125 because we have @xmath126 . if @xmath127 and @xmath128 are polynomials , the circularity is half the degree of @xmath129 this would then imply that the circularity is not less than @xmath130 , which equals @xmath131 for @xmath132 . thus , we have to show that @xmath133 divides both , @xmath134 and @xmath135 . moreover , it is sufficient to prove the claim only for a reduced @xmath136 as the claim in the non - reduced case follows easily . set @xmath137 where @xmath138 and @xmath139 . it is a useful fact that @xmath140 has no real linear factor because such a factor would also divide @xmath141 ( because @xmath142 $ ] has no real factor and hence @xmath143 has no real linear factor either ) and @xmath144 ( because it divides @xmath145 and @xmath146 has no real factor ) . but this contradicts the reducedness of @xmath147 . now we proceed by induction on the degree @xmath148 of @xmath140 . note that @xmath148 is zero or an even positive integer . in case of @xmath149 , @xmath140 is a real constant and the claim is clear . let @xmath150 . then there is a monic quadratic irreducible real polynomial @xmath151 which divides @xmath140 . set @xmath152 . we claim that there always exists a quaternion @xmath153 and two quaternion polynomials @xmath154 and @xmath155 such that @xmath156 , @xmath157 and @xmath158 . we handle the proof of this claim by distinguishing three cases : 1 . @xmath151 divides @xmath141 . then @xmath151 can not divide @xmath159 because of the reducedness of @xmath41 . since @xmath151 is a divisor of @xmath140 which divides @xmath145 , @xmath151 must divide @xmath145 . since @xmath146 has no real polynomial factor , by , there is a unique quaternion @xmath153 and two polynomials @xmath154 and @xmath155 such that @xmath156 , @xmath157 and @xmath158 . 2 . @xmath151 does not divide @xmath141 but divides @xmath159 . then @xmath151 must divide @xmath143 . since @xmath146 has no real polynomial factor , by , there is a unique quaternion @xmath153 such that @xmath156 , @xmath157 . then we set @xmath160 which is a polynomial and satisfies @xmath161 . 3 . @xmath151 divides neither @xmath141 nor @xmath159 . by , there is a unique quaternion @xmath153 and two polynomials @xmath154 and @xmath155 such that @xmath156 , @xmath157 and @xmath158 . in all three cases , we have a new reduced motion polynomial @xmath162 , such that @xmath163 by induction hypothesis , @xmath164 divides @xmath165 and @xmath166 . by the derivation of @xmath154 and @xmath155 , namely , @xmath157 and @xmath158 , we have that @xmath140 divides @xmath143 and @xmath167 . this concludes the induction proof and also the proof of the theorem . calling a rational motion _ generic _ if the primal part has no real factors one interpretation of is as follows : a trajectory of a generic rational motion can be entirely circular ( actually , only few exceptions are known ) . a similar statement for algebraic planar motions can be found in ( * ? ? ? there , also a more detailed discussion on the circularity of trajectories can be found . for planar rational curves , the geometric circularity conditions there imply the algebraic circularity conditions of . as one consequence of , among all rational curves of fixed degree , curves of high circularity can be generated by rational motions of low degree . in particular , we obtain the desired bound on the degree of a rational motion with a given rational trajectory : [ cor:1 ] if a rational curve of circularity @xmath168 and degree @xmath62 is a trajectory of the rational motion @xmath49 , the degree of @xmath41 is not less than @xmath169 . if it is of degree @xmath169 , the degree defect of the spherical motion component equals @xmath170 . we will see below in that the bound of is sharp . let us illustrate above results by simple examples from literature . in case of @xmath171 , the only generic rational motions are rotations with fixed axis . their generic trajectories ( trajectories that maximise the degree among all trajectories ) are circles and , of course , entirely circular . rational motions with @xmath172 are generated by reflecting a moving frame in one family of rulings on a quadric @xmath173 , see @xcite . the generic case is obtained if @xmath173 is a hyperboloid . in this case , generic trajectories are of degree four . their entirely circularity has already been observed in ( * ? ? ? * ch.ix , 7 ) . the non - generic trajectories are circles . points with non - generic trajectories lie on two skew lines that coincide if @xmath173 is a hyperboloid of revolution . finally , if @xmath173 is a hyperbolic paraboloid , generic trajectories are just of degree three and circularity one @xcite . now we turn to the task of computing a rational motion @xmath49 of minimal degree @xmath174 with a given rational parameterised curve @xmath55 as trajectory . denote the degree of the trajectory by @xmath62 and its circularity by @xmath168 . by we have @xmath175 . we will prove that , up to coordinate changes , exactly one solution of minimal degree @xmath176 exists and we also provide a procedure for its computation . [ th:2 ] let @xmath55 be a reduced rational parametric equation of a rational curve such that @xmath177 . then there exists a unique monic rational motion polynomial @xmath49 of minimal degree , such that @xmath178 and the trajectory of @xmath111 is parameterised by @xmath20 . a key ingredient in our proof of is a version of the euclidean algorithm to compute the _ left gcd _ of two polynomials @xmath179 $ ] . this is a well - known concept in the theory of polynomials over rings , see @xcite . call the polynomial @xmath180 $ ] a _ left factor _ or _ left divisor _ of @xmath181 if there exists @xmath51 $ ] such that @xmath182 . the polynomial @xmath183 $ ] is called _ left gcd _ of @xmath181 and @xmath184 , if @xmath185 is a left divisor of @xmath181 and @xmath184 and any left divisor @xmath186 of @xmath181 and @xmath184 also left divides @xmath185 . the left gcd is unique up to right multiplication with a non - zero quaternion . the euclidean algorithm in this context is based on polynomial _ right _ division . given @xmath187 $ ] , there exist @xmath188 $ ] such that @xmath189 and @xmath190 . note that the order of factors in the product @xmath191 is important . if @xmath185 is a left divisor of @xmath192 and @xmath193 , then @xmath185 is also a left divisor of @xmath191 and of @xmath194 . conversely , a left divisor of @xmath194 and @xmath193 also left divides @xmath192 . hence the left gcd of @xmath192 and @xmath193 equals the left gcd of @xmath193 and @xmath194 . assuming @xmath195 , we also have @xmath196 and we may recursively define sequences @xmath197 and @xmath198 of polynomials with strictly decreasing degree by @xmath199 the recursion ends as soon as @xmath200 whence the left gcd is @xmath201 . using polynomial long division , an algorithmic implementation of this algorithm is straightforward . [ lem:2 ] if @xmath202 $ ] and @xmath203 $ ] is an irreducible ( over @xmath204 ) quadratic factor of @xmath47 , then @xmath205 and @xmath41 have a left gcd of positive degree . use polynomial right division to compute @xmath206 $ ] with @xmath207 and @xmath208 . then @xmath209 if @xmath210 , then @xmath205 itself is a left gcd of @xmath41 and @xmath205 . otherwise , @xmath205 is a left factor of @xmath211 which implies @xmath212 . hence , @xmath213 is a left factor of @xmath205 and also of @xmath41 . [ lem:3 ] for any polynomial @xmath202 $ ] of positive degree and a monic divisor @xmath214 $ ] of @xmath47 with @xmath215 , there exists a unique monic quaternion polynomial @xmath50 ( the left gcd of @xmath41 and @xmath213 ) and a unique quaternion polynomial @xmath64 such that @xmath216 and @xmath217 . the polynomial @xmath213 has no real linear factor because such a factor necessarily is a divisor of @xmath41 which contradicts @xmath218 . using the euclidean algorithm , we can compute the unique monic left gcd @xmath219 of @xmath41 and @xmath213 , i.e. , @xmath220 with @xmath221 $ ] . we claim that @xmath219 and @xmath222 both have no real polynomial factor : * a real polynomial factor of @xmath219 is also a real polynomial factor of @xmath41 and @xmath213 and contradicts @xmath218 . * a real polynomial factor of @xmath222 must have a real quadratic factor @xmath205 which is irreducible over @xmath204 ( because @xmath213 has no real linear factor ) . that is , for some @xmath223 $ ] we have @xmath224 . we can write @xmath225 where @xmath226 is real . from this we infer @xmath227 . because @xmath228 has no real polynomial factor , the real polynomial @xmath229 divides @xmath226 and @xmath230 divides @xmath213 but also @xmath231 ( a multiple of @xmath213 ) . this implies that @xmath205 divides @xmath135 and , by , @xmath64 and @xmath205 have a left gcd @xmath232 with @xmath233 . but then , by , @xmath234 is a left divisor of @xmath41 and @xmath213 and @xmath235 . this contradicts the fact that @xmath219 is left gcd of @xmath41 and @xmath213 . now we claim that @xmath236 . left multiplying the second equation in with @xmath228 we obtain @xmath237 . we have a real polynomial factor @xmath213 on the left and a real polynomial factor @xmath229 on the right . because neither @xmath219 nor @xmath222 have real polynomial factors , we have @xmath238 and @xmath236 follows . thus , with @xmath239 , we have existence . as to uniqueness , we observe that @xmath50 is necessarily the unique monic left gcd of @xmath41 and @xmath213 . uniqueness of @xmath64 follows from uniqueness of @xmath50 . denote the degree of the rational parametric equation @xmath20 by @xmath62 and its circularity by @xmath168 . the condition @xmath177 implies @xmath240 for @xmath241 and it is no loss of generality to assume that @xmath56 is monic . there exists a monic polynomial @xmath60 of degree @xmath242 and two relatively prime polynomials @xmath243 ( monic ) , @xmath244 $ ] with @xmath245 , @xmath246 , such that @xmath247 denote by @xmath248 a ( yet unknown ) motion polynomial of minimal degree that parameterises the sought rational motion . the parametric trajectory of @xmath111 equals @xmath20 if @xmath249 with @xmath250 , we have @xmath251 by , there exist polynomials @xmath252 $ ] and @xmath253 $ ] such that @xmath254 with @xmath255 and @xmath256 we then have @xmath257 this shows that @xmath50 and @xmath64 solve the first of the two equations in . the second equation follows from the vanishing of the scalar part of @xmath185 : @xmath258 . thus , the polynomial @xmath49 with @xmath50 , @xmath64 computed as above is one solution to our problem and it is of minimal degree @xmath259 . conversely , assume that @xmath260 is another solution of minimal degree @xmath261 . again , we have @xmath262 and @xmath263 is monic . by the first of the two equations in we get @xmath264 for some monic real polynomial @xmath265 of degree @xmath170 . by , , and we get @xmath266 if we divide both sides by @xmath60 and multiply with @xmath265 , we get @xmath267 as @xmath243 and @xmath119 are relatively prime , we get that @xmath243 divides @xmath265 . but @xmath268 and @xmath243 and @xmath265 are monic . thus , we have @xmath269 left - multiplying the second of two equations in by @xmath270 and using , we get @xmath271 as @xmath243 and @xmath119 are relatively prime and @xmath243 is monic , @xmath272 must be a monic polynomial . we have @xmath273 because of @xmath274 and because the left gcd of @xmath60 and @xmath185 is unique , this implies @xmath275 and @xmath276 . this concludes the proof of uniqueness . in case of a rational curve of circularity zero , the primal part @xmath50 of the rational motion @xmath49 is a real polynomial and the trivial solution consists of the curvilinear translation along the given curve . in all other cases , guarantees solutions of lesser degree than curvilinear translations . our formulation of requires a special moving point , the origin @xmath111 , to generate the rational curve and produces a motion polynomial @xmath41 with the special property @xmath277 . these conditions are just coordinate dependencies and are convenient for our algebraic formulation and proof . if we allow coordinate changes , we may translate @xmath111 to an arbitrary moving point ( this amounts to replacing @xmath41 by @xmath278 with a suitable translation , as in the proof of ) and orient the coordinate axis arbitrarily ( this amounts to right - multiplying @xmath41 with @xmath279 ) . thus , we can re - formulate as [ cor:2 ] there is a unique ( up to coordinate changes ) rational motion of minimal degree in the dual quaternion model of rigid body displacements with a prescribed rational trajectory . if the trajectory is of degree @xmath62 and circularity @xmath168 , the minimal motion is of degree @xmath176 . frequently , the degree of a rational motion is defined as maximal degree of one of its trajectories . with this concept of degree , is no longer true . a counter example is a circle which occurs as trajectory of a circular translation and of a rotation about its centre . for both motions the maximal degree of a trajectory is two . let us illustrate by two examples : the rational curve given by the parameterised equations @xmath280 is of degree five . from the factorisation @xmath281 we see that its circularity is two . by , it is trajectory of a rational motion of minimal degree three whose primal part , by , has precisely one linear factor . with @xmath119 , @xmath243 and @xmath60 defined by @xmath282 we have @xmath283 setting @xmath284 and using the euclidean algorithm we compute , according to , the left gcd @xmath219 of @xmath185 and @xmath60 : with @xmath285 and @xmath286 we have @xmath287 , @xmath288 . then we set @xmath289 and @xmath290 . as expected , @xmath50 has a real polynomial factor of degree @xmath111 . the motion polynomial we want is @xmath291 . [ ex:3 ] the viviani curve @xcite is given by @xmath292 in order to meet the requirements of , we translate it by the vector @xmath293 and obtain @xmath294 this curve lies on the unit sphere and is entirely circular . as in the previous example , we compute @xmath295 the left gcd of @xmath250 and @xmath60 is @xmath296 the right quotient of @xmath185 and @xmath219 is @xmath297 and the minimal motion to the curve is @xmath298 we can simplify by conjugating with @xmath299 whence we get @xmath300 this shows that @xmath301 is the composition of two rotations about the second coordinate axis and the third coordinate axis with equal angular velocities , i.e. , the motion generated by the rolling of a spherical circle of radius @xmath302 on a spherical circle of the same radius . this is illustrated in . there , the moving frame is rigidly attached to the rolling circle . this article unveiled some relations between rational motions and their trajectories . a rational curve occurs as trajectory of a unique ( up to coordinate changes ) rational motion of minimal degree in the dual quaternion model . this was a surprise to the authors as a mere trajectory seems to leave a lot of freedom for the construction of a suitable rational motion . apparently , the requirement for minimality is rather restrictive . calling a rational curve `` generic '' if its circularity is zero and a rational motion `` generic '' if its primal part has no real factors ( the spherical motion component has full degree ) , we may also say that rational motions of minimal quaternion degree to a generic trajectory are non - generic . conversely , the trajectories of a generic rational motion are entirely circular and hence non generic . in conjunction with the factorisation of generic rational motions @xcite and its extension @xcite to non - generic motion polynomials , our results contribute to a variant of kempe s universality theorem ( * ? ? ? * section 3.2 ) for rational space curves . via motion factorisation , it is possible to construct linkages to generate a given rational motion and hence also a given rational trajectory . the dual quaternion degree of the generating motion is directly related to the number of links and joints in the mechanism . as a consequence , rational curves can be generated by spatial linkages with much fewer links and joints than those implied by the asymptotic bounds for algebraic space curves given by @xcite . for circular curves , the numbers of links and joints are even less . a precise formulation and a rigorous proof will be worked out in a future paper . our main result in corollary [ cor:2 ] raises questions . is a similar uniqueness statement true for algebraic curves and algebraic motions ? what rational motions are not minimal for any of their trajectories ? the darboux motion and the circular translation of remark [ rem:1 ] are examples . finally , determine simple ( classes of ) curves with simple minimal motion as in example [ ex:3 ] and exploit their low degree and rationality in a cagd or engineering context . this work was supported by the austrian science fund ( fwf ) : p 26607 ( algebraic methods in kinematics : motion factorisation and bond theory ) . zijia li , josef schicho , and hans - peter schrcker , _ 7r darboux linkages by factorization of motion polynomials _ , accepted for publication in the proceedings of the 14th iftomm world congress 2015 , 2015 , http://arxiv.org/abs/1501.07365[arxiv cs.ro/1501.07365 ] .

we give a constructive proof for the existence of a unique rational motion of minimal degree in the dual quaternion model of euclidean displacements with a given rational parametric curve as trajectory . the minimal motion degree equals the trajectory s degree minus its circularity . hence , it is lower than the degree of a trivial curvilinear translation for circular curves .

introduction the dual quaternion model of rigid body displacements rational curves as trajectories of rational motions discussion of results acknowledgements

in dealing with systems of equations having a large number of degrees of freedom one frequently adopts an approximation which leads to a nonlinear partial differential equation . for instance , the mean - field approach in the statistical physics is based on the introduction of an order parameter governed by a nonlinear equation . for example , the mean - field theory of the bose - einstein condensate of a degenerate quantum gas is based on the gross - pitaevskii equation @xcite . the gross - pitaevskii equation is the nonlinear schrdinger equation with an external potential . it describes the so - called `` matter waves '' the nonlinear collective modes of the degenerate quantum gas below the condensation temperature . in many other cases the approximate nonlinear equations , appearing as the leading order description of nonlinear waves in various branches of modern physics , for instance , in nonlinear optics , plasma , and on water , are of the nonlinear schrdinger class ( see , for instance , ref . @xcite and the references therein ) . in physics one is especially interested in the stationary solutions ( a.k.a . the stationary points ) of the governing equations , in particular , in the stationary points which minimize the energy . only in some exceptional cases the solution can be obtained analytically and one has to rely on numerical simulations . if the nonlinear equation in question possesses the lagrangian formulation then the stationary points can be found by a nonlinear optimization method . we propose a new optimization method for nonlinear partial differential equations , in particular , for the equations of the nonlinear schrdinger ( nls ) type . the method consists of a numerical minimization of the rayleigh functional . in a broader perspective , we introduce the concept of rayleigh functional for nonlinear systems and show how it can be employed in both numerical and analytical analysis of equations of physical importance . in our case , the optimization problem is to find the ground state of a nonlinear system , i.e. the stationary solution which minimizes the energy . we work in the space of smooth functions @xmath0 of @xmath1 spatial variables @xmath2 and time variable , which have bounded @xmath3-norm : @xmath4 . thus we consider only bounded stationary solutions which decay to zero as @xmath5 and correspond to finite energies . we adopt the following energy functional : @xmath6 and assume that it has the phase invariance symmetry @xmath7 ( we use the complex conjugate functions @xmath8 and @xmath9 as independent variational variables instead of the real and imaginary parts of @xmath8 ) . the phase invariance symmetry results in conservation of the @xmath3-norm . our method also works for the generalization of the energy functional ( [ eq1 ] ) to several variables of variation : @xmath10 , @xmath11 ( this case is discussed below ) , and to the higher order derivatives : @xmath12 , @xmath13 . we adopt the notations used in statistical physics , an important area for applications . hence , the dependent variable @xmath14 will be referred to as the order parameter . the well known gross - pitaevskii functional for the order parameter of a bose - einstein condensate , @xmath15 belongs to the class specified by equation ( [ eq1 ] ) . here @xmath16 , @xmath17 is the number of atoms in the condensate , @xmath18 is the interaction coefficient due to the s - wave scattering of the atoms and @xmath19 is the trap potential ( created by a magnetic field or non - resonant laser beams ) . the order parameter is normalized to 1 in equation ( [ eq2 ] ) . let us briefly recall the basics of the nonlinear optimization . one is interested in the stationary state , given as @xmath20 , where @xmath21 is the chemical potential . the euler - lagrange equation corresponding to the energy functional ( [ eq1 ] ) reads @xmath22 here ( and below ) we denote @xmath23 the variational derivative of a functional @xmath24 with respect to @xmath9 ( thus we part with the usual notation of the latter by using the symbol @xmath25 , while @xmath25 is reserved for the usual gradient of a function ) . the stationary state satisfies the equation @xmath26 . the idea of the imaginary time relaxation method is based on the fact that the variational derivative of a functional is an analog of the gradient of a function . thus , by introducing the `` imaginary time '' @xmath27 in equation ( [ eq3 ] ) one forces the order parameter to evolve in the direction of maximum decrease in the energy . the attractor of such evolution is , hopefully , a local minimum ( in general , just a stationary point ) . the imaginary time evolution does not conserve the @xmath3-norm and one must normalize the order parameter directly during the evolution . in other words , one allows the order parameter to evolve in the space of arbitrary functions but normalizes the solution after each step by the prescription @xmath28 . this can be formulated in a single equation : @xmath29 the method of lagrange multipliers can be used to take into account the constraints imposed on the variables of variation ( for instance , the fixed norm ) . it consists of a numerical minimization of an augmented functional which is the energy functional plus the constraint with a lagrange multiplier . the combined functional has the stationary solution as its extremal point and the constrained minimization problem is converted into an unconstrained one . in this case , the time evolution of equation ( [ eq4 ] ) can be substituted by a finite - step minimization scheme , such as the method of steepest descent or the conjugate gradient method , supplemented by an appropriate line - search algorithm ( consult , for instance , refs . @xcite ) . for instance , the following two functionals are used @xmath30 where the variable of variation is arbitrary ( has arbitrary @xmath3-norm ) . indeed , the first of these functionals evidently has the stationary solutions as its extremals , while the second has the variation @xmath31 . setting @xmath32 we get the stationary point by equating the variation of @xmath33 to zero . the use of the functional @xmath33 in the search for the energy minimizers was advocated recently in ref . the point is that the trivial solution @xmath34 , being an extremal , frequently makes reaching other solutions difficult with the use of the functional @xmath35 , while functional @xmath33 is free from such a flaw . the simplest numerical realization of the minimization method is given by the steepest descent scheme : @xmath36 where the parameter @xmath37 is selected by an appropriate line - search algorithm . this paper is organized as follows . in section [ ray ] we introduce the rayleigh functional for the energy given by equation ( [ eq1 ] ) and discuss its properties as a variational functional . the generalization to several order parameters is also considered . in section [ num ] we present examples of the nonlinear optimization based on the rayleigh functional . as an application of the rayleigh functional in the analytical approach , in section [ criticalnls ] we relate an identity satisfied by the stationary solutions of the critical nls equations to the scale invariance symmetry of the rayleigh functional finally , in section [ concl ] the advantage of the numerical optimization method based on the rayleigh functional is discussed . before formulating our nonlinear optimization method , it would be instructive to recall how the eigenfunctions of a linear operator can be obtained numerically . consider , for instance , the textbook problem of finding the eigenvalues and eigenfunctions of the hamiltonian operator @xmath38 for a quantum particle in a potential well ( we use @xmath39 and @xmath40 ) . this problem can be reformulated as an optimization problem by employing the well - known rayleigh quotient @xcite @xmath41 here the function of variation @xmath8 is arbitrary , i.e. not normalized . the eigenfunctions are the stationary points of the rayleigh quotient . indeed @xmath42 where the eigenvalue is given as @xmath43 , i.e. it is the value of @xmath44 at the eigenfunction . note that the rayleigh quotient allows one to cast the constrained minimization problem ( with the constraint @xmath45 ) into an unconstrained one . stationary solutions of nonlinear systems , the so - called nonlinear modes , can be considered as the generalization of the eigenfunctions . they can be found in a way similar to the solution of the above eigenvalue problem . in contrast to the linear systems , a _ nonlinear _ one usually possesses continuous families of stationary solutions , each solution corresponding to a different chemical potential ( an analog of the energy level in the nonlinear case ) . in general , the chemical potential takes its values from an infinite interval of the real line . by fixing the @xmath3-norm we introduce it as a parameter in the energy functional . thus , the chemical potential , being a continuous function of the norm , is also fixed ( in general , we get away with a finite number of _ distinct _ values lying on the different branches of the function @xmath46 ) . the only exception is the so - called critical case when a continuous family of stationary solutions corresponds to the same @xmath3-norm ( see also section [ criticalnls ] , where we discuss the critical case ) . therefore , apart from the critical case , by fixing the norm we select a finite number of solutions . among them there is the energy minimizer ( for a given value of the conserved @xmath3-norm ) . we can reconstruct the continuous family of the stationary solutions by using different values of the norm . in the case of bose - einstein condensates , for instance , this corresponds to fixing the number of atoms and looking for the corresponding ground state solution . it is assumed that the nonlinear equation in question has a phase invariance resulting in the @xmath3-norm conservation . in the following we will distinguish between the normalized and not - normalized functions , denoting the former by @xmath47 and the latter by @xmath0 . for convenience , we set the @xmath3-norm of the stationary solution equal to 1 , @xmath48 . this normalization is performed by using a scale transformation of the function of variation and results in the explicit appearance of the value of @xmath3-norm in the energy functional ( note that the chemical potential may also be scaled appropriately , as in equations ( [ eqincoh ] ) and ( [ eqcoh ] ) of section [ num ] ) . for example , in equation ( [ eq2 ] ) we have the coefficient @xmath17 at the nonlinear term , after we set the @xmath3-norm of the solution to 1 , and the value of the energy functional is , in fact , the energy of the condensate per atom . for the energy given by equation ( [ eq1 ] ) the following functional can serve as the nonlinear generalization of the rayleigh quotient : @xmath49 here the @xmath3-norm @xmath50 is not fixed but is a functional of @xmath8 and @xmath9 ( since the norm is constant _ function _ of @xmath51 we can take it out of the gradient operator @xmath25 ) . note that @xmath44 is a compound functional : it depends on the complex conjugate functions @xmath52 and @xmath53 through the normalized ones , @xmath54 and @xmath55 , used in the energy functional @xmath56 . this simple functional turns out to be very helpful in the search for the energy minimizers . let us discuss the properties of functional ( [ eq9 ] ) . first , its extremals are stationary points of the corresponding nonlinear equation , equation ( [ eq3 ] ) , similar as in the case of the rayleigh quotient ( [ eq7 ] ) and equation ( [ eq8 ] ) . ( the unique stationary point is selected by the value of the @xmath3-norm , appearing explicitly in the energy functional as a _ parameter _ ) . this follows from a simple calculation : @xmath57 @xmath58 @xmath59 where we have used that @xmath60 by interchanging the order of integration in the double integral in formula ( [ eq10 ] ) and gathering the terms with @xmath61 we obtain the variational derivative @xmath62 from equation ( [ eq11 ] ) it is quite clear that the stationary points of equation ( [ eq3 ] ) are extremals of the rayleigh functional and vice versa . the corresponding chemical potential is equal to the integral in the second term on the r.h.s . of equation ( [ eq11 ] ) , i.e. @xmath63 which fact can be verified by direct integration of equation ( [ eq3 ] ) , i.e. @xmath64 , and taking into account the normalization of @xmath65 . equation ( [ eq12 ] ) plays here the role of the expression for the energy level @xmath66 in the above problem of a quantum particle . the rayleigh functional also distinguishes the local minima among the stationary points . to see this let us compute its second variation . assuming that @xmath8 is a stationary point , we get @xmath67 @xmath68 where we have used the fact that @xmath65 satisfies the stationary equation @xmath69 . integrating by parts and taking into account the normalization of @xmath65 , i.e. @xmath70 , we arrive at @xmath71 the subscript in the first term on the r.h.s . of this equation means that the variable of variation of the energy functional is , in fact , @xmath54 . taking this into account , one immediately recognizes on the r.h.s . of equation ( [ eq13 ] ) the second variation of the functional @xmath72 defined in equation ( [ eq5 ] ) and evaluated in the space of normalized functions . therefore , the minima of the lagrange modified energy functional ( for a fixed nonzero @xmath3-norm ) are also minima of the rayleigh functional and vice versa . importantly , the rayleigh functional does not contain the trivial solution @xmath73 among its extremals , in contrast to the lagrange modified energy functional . finally , let us discuss the functional of ref . @xcite , which was proposed as another possible generalization of the rayleigh quotient to the equations of the nonlinear schrdinger type . in the latter work in the numerical search for the stationary solutions of two - dimensional gross - pitaevskii equation the following functional was employed @xmath74 this functional can be expressed as follows @xmath75 here the rayleigh functional is given by @xmath76 with @xmath77 and @xmath17 is the number of atoms corresponding to the stationary state . note that the first variation of the rayleigh functional is zero on a stationary solution and the last two terms in formula ( [ eqnull ] ) have nonzero variation unless the norm @xmath50 is kept fixed ( @xmath78 ) . hence , the functional @xmath79 introduced in ref . @xcite can not be employed for _ unconstrained _ minimization . now let us show that , in fact , the euler - lagrange equation for the functional @xmath79 defined by equation ( [ cras ] ) is self - contradictory ( in other words , the first variation of @xmath79 is never zero ) and , hence , does not lead to any stationary solutions at all . the euler - lagrange equation for the functional ( [ cras ] ) reads @xmath80 where @xmath79 is the value of the functional on the solution . on the other hand , by multiplying the above equation and integrating we get @xmath81 hence @xmath82 and we have arrived at a contradiction unless @xmath83 . q.e.d . the rayleigh functional was introduced above for a nonlinear system described by a single ( complex - valued ) order parameter @xmath52 . it can be easily generalized to the case several order parameters . the definition of the rayleigh functional will depend on the number of the conserved @xmath3-norms . this number is determined by the type of phase invariance of the energy functional . we will concentrate mainly on the two broad cases : the incoherent and coherent coupling of the order parameters . for simplicity , consider the case of two order parameters : @xmath84 . in the case of incoherent coupling , there are two independent constraints corresponding to the two conserved @xmath3-norms ( the normalization is defined independently for each order parameter : @xmath85 , @xmath86 ) . on the other hand , if the coupling is coherent , then there is only one constraint corresponding to conservation of the total @xmath3-norm ( @xmath87 ) . we also give an example of application of the rayleigh functional to a nonlinear system which does not belong to either of the above cases ( see section [ num ] ) . an example of the incoherent coupling is the gross - pitaevskii functional for a two - species mixture of degenerate quantum gases in an external trap below the condensation temperature ( see , for instance , ref . @xcite ) : @xmath88 on the other hand , the coupled - mode approximation for a bose - einstein condensate in the three dimensional parabolic trap modified in one dimension by a laser beam with creation of a central barrier illustrates the coherent coupling : @xmath89 ( for derivation of the coupled - mode system and the applicability conditions consult ref . the coupling coefficient @xmath90 is proportional to the tunnelling rate through the central barrier of the resulting double - well trap . it is easy to see that functional ( [ eq14 ] ) admits two independent phase invariance transformations , @xmath91 , while functional ( [ eq15 ] ) admits only the simultaneous phase invariance transformation of both variables with the same @xmath92 . the rayleigh functionals for the two cases corresponding to equations ( [ eq14 ] ) and ( [ eq15 ] ) are defined as follows : @xmath93 where @xmath94 . it should be stressed that to define the rayleigh functional one should employ the normalization which follows from the phase invariance in its general form . for instance , the normalization by the total @xmath3-norm in the incoherent coupling case would allow one to find only the stationary solutions with the same chemical potential for each component of the order parameter , i.e. a very limited class of solutions . the variational derivative of the rayleigh functional @xmath95 ( [ eq16 ] ) is derived by a mere repetition of the steps which led to equation ( [ eq11 ] ) . we have @xmath96 where , as usual , the @xmath65-variables are substituted in the energy functional instead of @xmath8-ones , i.e. @xmath97 . now it is easy to see that , thanks to the equation @xmath98 , the integral on the r.h.s . of formula ( [ eq17 ] ) coincides with the corresponding chemical potential @xmath99 . thereby the extremals of the rayleigh functional @xmath95 are stationary points of the corresponding nonlinear system of equations . the variational derivative of the functional @xmath100 ( [ eq16 ] ) can be most easily obtained from formula ( [ eq11 ] ) by noticing that this rayleigh functional depends on the function @xmath101 through @xmath102 and also through the total norm @xmath103 in all other @xmath65-variables . therefore , each variable of the energy functional leads to an integral term similar to the second term in equation ( [ eq11 ] ) but multiplied by @xmath104 . hence the final result : @xmath105 where @xmath106 . to show that the extremals of the rayleigh functional @xmath107 are also stationary points one has to use both stationary equations , i.e. @xmath108 and @xmath109 , multiplied by the complex conjugate functions @xmath110 . we get @xmath111 let us now find the second variation of the rayleigh functional at a stationary point in the above two cases . the derivation is similar to that for a single order parameter : one has to use the stationary equations and the normalization constraints for the @xmath65-variables . we get : @xmath112 @xmath113 in both cases the second variation of the rayleigh functional coincides with that of the lagrange modified energy functional considered in the space of normalized functions . therefore , the minima of the lagrange modified energy functional ( for the fixed values of the conserved @xmath3-norms ) are also minima of the rayleigh functional and vice versa . concluding this section we note that by introducing the concept of rayleigh functional we have converted the task of finding the stationary points to an unconstrained optimization problem which also distinguishes between the local minima and the saddle points . in the next section we give examples of application of this optimization method to the numerical search for the stationary states of bose - einstein condesates . by analogy with the use of the rayleigh quotient in the numerical search for eigenfunctions , the nonlinear optimization can be based on the rayleigh functional . leaving a mathematically rigorous analysis for future research , let us nevertheless give an heuristic argument . the rayleigh functional is a compound functional : the energy functional depends on @xmath54 and @xmath55 , while these are functionals of @xmath52 and @xmath53 . the numerical nonlinear optimization of a functional reduces to that of a smooth function ( though of a large number of variables ) . but a continuous function on a finite - dimensional sphere ( which in the numerics plays the role of the variable @xmath54 ) always has a minimum . since the second variation of the rayleigh functional is actually the second variation of the lagrange modified energy functional , evaluated in the space of normalized functions , the optimization method based on the rayleigh functional has the capacity to return the absolute minimum . generally , it converges to a local minimum . an additional argument in favor of the nonlinear optimization based on the rayleigh functional is provided already by the imaginary time relaxation method a widely used approach in the computational physics . indeed , it is precisely the rayleigh functional which is minimized in this method . let us recall the usual formulation given by equation ( [ eq4 ] ) of section [ intro ] . first of all , this formulation does not reflect the total change of @xmath8 , since after each numerical step one performs the direct normalization of the updated function . this means that the variational derivative @xmath114 does not tend to zero , since the norm of the order parameter is not forced to approach a constant value as @xmath115 . however , one can reformulate the relaxation method for the variable @xmath116 which has a fixed @xmath3-norm . we have @xmath117 @xmath118 where we have used equation ( [ eq4 ] ) to substitute for @xmath119 . we see that if the imaginary time relaxation converges , i.e. if @xmath120 , then it converges necessarily to an extremal of the rayleigh functional . equation ( [ eq21 ] ) , besides revealing a new side of the imaginary time evolution method , manifests its strong drawback : the order parameter evolves continuously and , hence , quite slowly . but , since it is actually a way to minimize the rayleigh functional , a discrete minimization scheme can be adopted with the order parameter performing a series of finite steps to a local minimum . we have naturally arrived at the steepest descent formulation : @xmath121 where the parameter @xmath122 is selected by some line - search algorithm . more advanced gradient methods , for example , the conjugate gradient method , can be employed for the finite - step minimization . however , we have found that the steepest descent algorithm performs very well in the nonlinear optimization based on the rayleigh functional . in selecting the step size @xmath123 we have adopted the barzilai - borwein two - point method @xcite ( see also ref . there are two closely related barzilai - borwein methods : [ eq24 ] @xmath124 both methods are fairly equivalent in terms of performance . the generalization of the numerical scheme ( [ eq23])-([eq24 ] ) to several order parameters is given by the substitution of the scalar function @xmath52 by a vectorial one @xmath125 defined similar to the vectorial function @xmath126 used in formulae ( [ eq17 ] ) and ( [ eq18 ] ) ( in the vector case , the inner product in equations ( [ eq24a ] ) and ( [ eq24b ] ) contains , besides the integral , the hermitian inner product in the finite - dimensional target vector space ) . in the numerical implementation we have used the spectral collocation method for the spatial grid based on the fourier modes ( see refs . the accuracy of the obtained solution was checked by direct numerical substitution into the governing equations . in all cases a rapid convergence to the stationary point was observed . we have cut off the allowed values of @xmath127 by a small positive number ( in our case @xmath128 ) to have the iteration step always in the direction of decrease of the rayleigh functional . _ a. incoherent coupling_. our first example of the numerical optimization based on the rayleigh functional is the calculation of the ground state in the two - species bose - einstein condensate of rubidium in the quasi two - dimensional geometry ( the pancake trap ) . the derivation of the governing equations and further discussion can be found in ref . the energy functional for the two - species bose - einstein condensate is given by the two - dimensional version of equation ( [ eq14 ] ) . the parameters @xmath129 and @xmath130 account for the different lande magnetic factors . for the two isotopes of rubidium we have : @xmath131 ( for @xmath132rb ) and @xmath133 ( for @xmath134rb ) , while the interaction coefficients are @xmath135 , @xmath136 , and @xmath137 . the variational derivative is given by equation ( [ eq17 ] ) . to simplify the task , we note that the first term of the variational derivatives is the r.h.s of the stationary euler - lagrange equations written for the normalized functions : [ eqincoh ] @xmath138 here the chemical potentials @xmath139 are defined as follows : @xmath140 , @xmath141 , with @xmath142 being the chemical potential for the order parameter @xmath143 normalized to the number of atoms @xmath144 . a nontrivial feature of the two - species condensate is that the ground state suffers from the symmetry breaking transformation , if for instance , for a fixed number of atoms of the @xmath132rb isotope the number of atoms of the @xmath134rb isotope increases @xcite . in figure [ fg1 ] we show the symmetry preserved ( axially symmetric ) ground states , while the symmetry - broken one is illustrated in figure [ fg2 ] . the axially symmetric states were found by the optimization method projected on the space of axially symmetric functions and formulated in polar coordinates . the polar radius grid contained 128 points , while the two - dimensional grid was @xmath145 ( we have used the @xmath146-grid to produce figure [ fg2 ] ) . the initial condition in both cases was the gaussian profile ( modulated by a symmetry - breaking perturbation in case of figure [ fg2 ] ) . our numerical simulations were performed in the matlab . the typical number of the iterations to reach the @xmath147 convergence was between 400 - 1000 ( on a personal computer with an amd 1ghz - processor the iterations have taken up to 5 seconds in the axially - symmetric case and up to 30 seconds in the 2d case ) . _ b. coherent coupling_. consider the following functional @xmath148 @xmath149 which appears in the description of the stationary states of a pair of repulsive and attractive ( @xmath150 ) two - dimensional condensates trapped in an asymmetric double - well potential @xcite ( in this case @xmath17 is proportional to the actual number of condensate atoms with the coefficient of proportionality of the order @xmath151 ) . here @xmath66 is the zero - point energy difference between the two wells of the external trap and @xmath17 is the total number of atoms . the euler - lagrange equations for the energy functional ( [ eq25 ] ) read : [ eqcoh ] @xmath152 where @xmath21 coincides with the chemical potential for the order parameter normalized to the number of atoms @xmath17 . as in the previous example of the incoherent coupling , we have observed a rapid convergence of the iterations to the ground state solution . the numerical minimization was performed in polar coordinates . the characteristic times to reach the deviation of about @xmath147 are the same as in the previous example ( formulated in polar coordinates ) . to verify that the solutions obtained with the use of the rayleigh functional method are indeed the ground states , we have used another numerical approach which enables one to find all possible stationary states ( consult for more details refs . the deformation of the ground state with the variation of the zero - point energy difference is illustrated in figure [ fg3 ] , where the condensate of vanishing atomic interaction , @xmath153 , is tunnel - coupled to an attractive condensate . _ c. arbitrary coupling_. the minimization method based on the rayleigh functional can be applied to systems of equations coupled by a different way than the coherent and incoherent coupling . as an example , we consider here the bose - einstein condensate in a pancake trap ( in the @xmath154 plane ) , but in contrast to the above considered cases , we take into account the transverse dimension ( @xmath155 ) by using the gaussian variational ansatz with the @xmath156-dependent parameters , @xmath157 for the order parameter in the gross - pitaevskii functional ( [ eq2 ] ) . a similar ansatz was already used with success for the condensate in a cigar shaped trap @xcite . here the parameter @xmath158 accounts for the transverse motion of the condensate . we are interested in the stationary solutions which correspond to @xmath159 and are given as @xmath160 and @xmath161 . integrating over the transverse coordinate in the gross - pitaevskii functional ( [ eq2 ] ) we get the energy functional for the stationary states : @xmath162 here @xmath163 the transverse oscillator length , @xmath164 is the external potential in 2d . note that @xmath165 as @xmath166 . the pancake geometry corresponds to @xmath167 . the equations satisfied by the stationary states read : @xmath168 @xmath169 where @xmath170 is the @xmath171-wave scattering length . assuming that the transverse width @xmath172 is of the order of @xmath173 ( which is supported by the numerics ) and estimating the operator @xmath174 we see that the term @xmath175 is much less than @xmath176 for @xmath177 . if this small term is omitted from the energy functional ( [ eqn2 ] ) , the equations for the stationary state become equivalent to the two - dimensional non - polynomial nls equation ( 2d npse ) of ref . ( to compare with the latter reference , though , one should take into account that the coefficient 1/2 is missing there at @xmath178 in the equation which is equivalent to equation ( [ eqn3 ] ) ) . the essential difference of the energy functional given by equation ( [ eqn2 ] ) from the above considered energy functionals lies in the fact that it has two variables of variation , @xmath179 and @xmath172 , but only one phase invariance exists and is associated with the normalization of @xmath179 . however , the minimization method based on the rayleigh functional defined with the substitution @xmath180 into the energy functional ( [ eqn2 ] ) works in this case also . to explain this , we note that one can think of the `` chemical potential '' corresponding to the variable @xmath172 as being equal to zero , thus the variation of the rayleigh functional with respect to @xmath172 coincides with that of the energy functional . the solutions of equations ( [ eqn3 ] ) and ( [ eqn4 ] ) were searched for in polar coordinates . as the system similar to ( [ eqn3])-([eqn4 ] ) was proposed in ref . @xcite as an improvement to the nls equation in two dimensions , we aimed to find out when does one need to use this complicated system for the pancake condensate instead of the much simpler 2d nls equation ? we have found that only in the study of collapse instability for @xmath181 the system ( [ eqn3])-([eqn4 ] ) gives visible improvement over the 2d nls equation . in all other cases , if one is not interested in the transverse profile of the condensate , the 2d nls equation gives satisfactory results for the stationary states . ( this , however , may not be so for the arbitrary time - dependent evolution , which is not captured by the 2d npse system of ref . @xcite since their @xmath182 identically . ) to illustrate this , in fig . [ fg4 ] we give the relative difference between the solutions @xmath183 of the system ( [ eqn3])-([eqn4 ] ) and the corresponding 2d nls equation for several values of the nonlinearity strength in the case of repulsive condensate ( @xmath184 ) . the solution is given in the units of @xmath185 , @xmath186 , the radius @xmath187 in terms of @xmath188 , and @xmath21 in terms of @xmath189 . the dimensionless nonlinearity strength is defined as @xmath190 . we have set @xmath191 which corresponds to the ratio @xmath192 . the transverse width @xmath172 , given in terms of @xmath173 , is shown in fig . [ fg5 ] . the maximal difference from @xmath173 is of the order of 10% . though in the case of attractive condensate the relative difference between the solutions of the system ( [ eqn3])-([eqn4 ] ) and the 2d nls equation is also small , the critical value for collapse differs noticeably . the critical value of the number of atoms is defined by @xmath193 , i. e. it is the endpoint coordinate of the curves in fig . the system ( [ eqn3])-([eqn4 ] ) is an improvement over the nls equation in this case , since the critical value differs by only 2% for @xmath194 from the actual value of the 3d condensate obtained in ref . the 2d nls value is achieved as the asymptotic limit when @xmath195 . we have argued that the minimization based on the rayleigh functional is always suitable for obtaining the ground state solutions . however , the possibility of obtaining excited states was not excluded at all . in contrast to the linear eigenvalue problem of section [ ray ] in the nonlinear case there is no general approach which would guarantee computation of all excited states , though there are some advances in this direction @xcite . however , a proper selection of the initial condition for minimization may lead to convergence to the `` nearest '' excited state . we illustrate this on the multi - vortex solutions of a nonlocal gross - pitaevskii equation , which in dimensionless form , in the frame of reference rotating with the frequency @xmath196 , reads ( see the details in ref . @xcite ) @xmath197 here @xmath198 , @xmath199 is the macdonald function , @xmath200 is the effective range of atomic interaction , @xmath201 is the angular momentum projection on the @xmath155-axis : @xmath202 , where @xmath92 is the polar angle , and @xmath196 is the rotation frequency . for simplicity , we consider the periodic boundary conditions along the @xmath155-axis , @xmath203 . we searched for new non - axial vortex solutions , involving combinations of vortices and antivortices , which represent the excited states of the system , since in the nonlocal gp equation ( [ gp3d ] ) , similar as the in local one , the stationary vortex solutions involving combinations of vortices and antivortices are never the energy minimizers . the energy is minimized by tkachenko lattices comprised of vortices only @xcite . in figures [ fg7 ] and [ fg8 ] we give two examples of excited states comprised of vortex and anti - vortex combinations ( the right panel ) and the respective energy minimizers ( the left panel ) . these solutions were found by using the initial conditions with the phase resembling that of the multi - vortex solution in quest . the ground state solution of a nonlinear partial differential equation is degenerate if fixing the @xmath3-norm does not select a unique chemical potential . in the critical case , when there are infinitely many stationary points with equal energy and the same @xmath3-norm , the nonlinear optimization method based on the rayleigh functional may not converge at all . to clarify this let us consider the critical nls equation . the ground state of the focusing critical nls equation is infinitely degenerate . it is well - known ( see , for instance , refs . @xcite ) that the two - dimensional critical nls equation has a family of soliton solutions having the same @xmath3-norm @xmath204 and a specific member @xmath205 satisfying the stationary equation @xmath206 is called the townes soliton . the stationary nls equation @xmath207 admits the scale invariance given by @xmath208 and @xmath209 . easy to see that , in two spatial dimensions , the scale invariance preserves the @xmath3-norm of the solution , while the energy of the stationary solution is exactly zero ( see below ) . this is precisely why one can not obtain the townes soliton by the nonlinear optimization method based on the rayleigh functional the method fails to converge . this drawback of the rayleigh functional in the numerical approach turns out into an advantage for the analytical analysis . indeed , for the critical nls equation it immediately leads to the well - known identity @xcite @xmath210 which is satisfied by any stationary solution . let us show this . first of all , it is easy to see that the energy functional of the nls equation @xmath211 has a scale invariance property : @xmath212 for @xmath213 . the invariance property is transferred to the rayleigh functional : @xmath214 . differentiation of the latter with respect to the scale invariance factor @xmath215 gives @xmath216 at the stationary point . hence , the second equality in equation ( [ eq29 ] ) follows . the first equality is derived by integrating the stationary equation multiplied by @xmath9 and using the already proven equality . in the above derivation we have explicitly related identity ( [ eq29 ] ) to the scale invariance property . the scale invariance is responsible for the exact balance of energies in any stationary solution of the critical nls equation : the kinetic energy is balanced by the energy due to the self - interaction . the identity ( [ eq29 ] ) is well - known . however , the point we try to make here that its relation to the critical scaling makes such an identity universal , i.e. it appears in connection to any nonlinear equation for which the energy functional has a scaling invariance for a family of solutions with the same @xmath3-norm . such is also the one - dimensional critical nls equation ( in the stationary form ) @xmath217 in this case the scale invariance is @xmath218 and @xmath219 leads to the energy scaling @xmath220 , with @xmath221 , while the @xmath3-norm @xmath222 remains unchanged . hence the rayleigh functional scales as follows @xmath214 and , by repetition of the above arguments , we get a similar identity satisfied by the family of stationary solutions : @xmath223 note that one can not use the energy functional in the above argument : the stationary solution is _ not _ an extremal of the energy functional . using the lagrange modified functional will not help either due to the explicit appearance of @xmath21 . the use of the rayleigh functional is indispensable in this short derivation . numerical search for stationary points of nonlinear partial differential equations is a difficult problem . to tackle such a problem one is left to try various methods and choose the one which gives a better performance and accuracy . in this paper we have proposed a new numerical method , the nonlinear optimization method based on the rayleigh functional . this method is a natural generalization of poincar s minmax principle for linear equations , formulated with the use of the rayleigh quotient . it turns out that the imaginary time relaxation method , a widely used method in the computational physics , is just a special case of nonlinear optimization based on the rayleigh functional . we have used the gradient scheme for minimization of the rayleigh functional , but this is not essential : one can use the minimization schemes which do not require use of the gradient . the simplicity of the rayleigh functional and its universality for nonlinear equations is one of the advantages of the method . moreover , the method can distinguish between the local minima and saddle points , since the second variation of the rayleigh functional is equal to the second variation of the lagrange modified energy functional , if the latter is evaluated in the space of normalized functions . the rayleigh functional takes care of the normalization constraint of the stationary solution . other constraints , however , have to be treated in the usual way . there are , however , some exceptional cases when the nonlinear optimization based on the rayleigh functional may fail . the principal cause of the failure is the infinite degeneracy of the ground state solution . still , a failure in the numerics is partially compensated by the fact that the very degeneracy allows one to get an important analytical insight relating the scale invariance and an identity satisfied by the stationary solutions in the critical case . this work was supported by cnpq and fapeal of brazil . the initial part of this work was done during the author s visit of the instituto de fisica terica , universidade estadual paulista , in so paulo , brazil , which was supported by the fapesp . we are grateful to jianke yang for helpful suggestions . 99 l. pitaevskii and s. stringari , _ bose - einstein condensation _ ( clarendon press , oxford , 2003 ) . c. sulem and p. l. sulem , _ the nonlinear schrodinger equation : self - focusing and wave - collapse _ , applied mathematical sciences , * 139 * , springer - verlag , 1999 . r. fletcher , _ practical methods of optimization _ ( 2nd . edition , jonh wiley & sons , new york , 1987 ) . j. nocedal , _ theory of algorithms for unconstrained optimization _ , acta numerica * 199 * , 242 ( 1992 ) . j. j. garca - ripoll and v. m. prez - garca , _ optimizing schrdinger functionals using sobolev gradients : applications to quantum mechanics and nonlinear optics _ , siam j. sci . * 23 * , 1315 ( 2001 ) . l. d. landau and e. m. lifshitz , _ quantum mechanics _ ( 3rd edition , pergamon press , oxford , 1987 ) . l. c. crasovan et al , phys . a * 66 * , 036612 ( 2002 ) . v. s. shchesnovich , s. b. cavalcanti , _ stable stationary states in two tunnel - coupled two - dimensional condensates with the scattering lengths of opposite signs _ , to appear in j. phys . a : math . gen . ( 2005 ) . v. s. shchesnovich , s. b. cavalcanti , and r. a. kraenkel , _ solitons in tunnel - coupled repulsive and attractive condensates _ a * 69 * , 033609 ( 2004 ) . v. s. shchesnovich , a. m. kamchatnov , and r. a. kraenkel , _ mixed - isotope bose - einstein condensates in rubidium _ , a * 69 * , 033601 ( 2004 ) . j. barzilai and j. m. borwein , _ two - point step gradient methods _ , i m a j. numer . analys . * 8 * , 141 ( 1988 ) . m. raydan , _ the barzilai and borwein gradient method for the large scale unconstrained minimization problem _ , siam j. optim . * 7 * , 26 ( 1997 ) . r. fletcher , _ on the barzilai - borwein method _ , research report department of mathematics , university of dundee , 2001 . b. fornberg , _ a practical guide to pseudospectral methods _ , ( cambridge , uk : cambridge university press ) 1996 . j. p. boyd , _ chebyshev and fourier spectral methods _ , ( second ed . , new york : dover publications inc . ) 2000 . l. n. trefethen , _ spectral methods in matlab _ , ( philadelphia : siam ) 2000 . v. s. shchesnovich , b. a. malomed , and r. a. kraenkel , _ solitons in bose - einstein condensates trapped in a double - well potential _ , _ physica d _ * 188 * , 213 ( 2004 ) . l. berge , phys . rep . * 303 * , 259 ( 1998 ) . a. m. kamchatnov , and v. s. shchesnovich , phys . a * 70 * , 023604 ( 2004 ) . l. salasnich , a. parola , and l. reatto , phys . a * 65 * , 043614 ( 2002 ) . a. gammal , t. frederico , and l. tomio , phys . a * 64 * , 055602 ( 2001 ) . v. s. shchesnovich and r. a. kraenkel , _ vortices in nonlocal gross - pitaevskii equation _ , j. phys . a : math . gen . * 37 * , 6633 ( 2004 ) . the axially symmetric ground state of the two - isotope bose - einstein condensate mixture of rubidium . the dashed and dotted lines give the order parameters ( normalized to the actual number of atoms ) of @xmath132rb and @xmath134rb isotopes , respectively . panel ( a ) @xmath224 , @xmath225 and @xmath226 ; panel ( b ) @xmath227 , @xmath228 , @xmath229 and @xmath230 ; panel ( c ) @xmath231 , @xmath232 and @xmath233 ( in dimensionless units ) . ] the symmetry breaking ground state of the bose - einstein condensate mixture of the isotopes of rubidium . the left and right panels give the shapes of the order parameters of @xmath132rb and @xmath134rb isotopes , respectively . here @xmath234 , @xmath235 , @xmath236 and @xmath237 ( in dimensionless units ) . ] deformation of the ground state of two bose - einstein condensates trapped in an asymmetric double - well potential with variation of the zero - point energy difference . the picture corresponds to the energy functional ( [ eq25 ] ) . here @xmath153 , @xmath238 and @xmath239 , the solid line gives @xmath240 , while the dashed one @xmath241 . ] the stationary vortex solutions for @xmath243 , @xmath244 and @xmath245 . the left panel shows the 3-vortex solution and the right one the 5-vortex solution comprised of 4 vortices and 1 antivortex ( in the center ) . ] the stationary vortex solutions for @xmath243 , @xmath244 and @xmath246 . the left panel shows the 8-vortex solution and the right one the 16-vortex solution comprised of 12 vortices and 4 antivortices . ]

we introduce rayleigh functional for nonlinear systems . it is defined using the energy functional and the normalization properties of the variables of variation . the key property of the rayleigh quotient for linear systems is preserved in our definition : the extremals of the rayleigh functional coincide with the stationary solutions of the euler - lagrange equation . moreover , the second variation of the rayleigh functional defines stability of the solution . this gives rise to a powerful numerical optimization method in the search for the energy minimizers . it is shown that the well - known imaginary time relaxation is a special case of our method . to illustrate the method we find the stationary states of bose - einstein condensates in various geometries . finally , we show that the rayleigh functional also provides a simple way to derive analytical identities satisfied by the stationary solutions of the critical nonlinear equations .

introduction definition and properties of the rayleigh functional numerical search for stationary points rayleigh functional in the analytical approach conclusion acknowledgements figures

chaos is now one of the most important ideas used to explain various non - linear phenomena in nature . since the research on the three body problem by poincar , many studies about chaos in celestial mechanics and astrophysics have been done and revealed the important role of chaos in the universe@xcite,@xcite . although we know many features of chaos in newtonian dynamics , we do not know , so far , so much about those in general relativity . if gravity is strong , e.g. , a close binary system or a particle near a black hole , we have to use einstein s theory of gravitation . because the gravitational field in general relativity is non - linear , we may find a new type of chaotic behavior in strong gravitational fields , which do not appear in newtonian dynamics@xcite-@xcite . in a previous paper@xcite , we studied a criterion for chaos of a test particle motion around an @xmath5-black hole system ( or an @xmath5-naked singularity system ) and found that a local instability determined by the riemann curvature tensor provides us a sufficient test for chaos . we also found that the existence of an unstable circular orbit , which guarantees the existence of a homoclinic or heteroclinic orbit , plays a crucial role for chaos . however , the relativistic systems analyzed so far by several authors@xcite-@xcite , in which chaotic behavior of a test particle is found , are rather unrealistic@xcite-@xcite , except for the perturbed spacetimes of the schwarzschild black hole solution@xcite,@xcite . as for the other interesting cases of systems , for example , the @xmath5-extreme black hole system@xcite-@xcite is unstable , the existence of a strong uniform magnetic field around a black hole@xcite is not likely , and naked singularities@xcite may not exist . we may wonder whether any realistic relativistic system can be chaotic and when chaos may play an important role in such a relativistic astrophysical phenomena . in astrophysics , rotation of a system plays a quite important role . the angular momentum or spin may completely change the evolution of the system . in a dynamical system , rotation or spin is one of the most important elements and it may sometime cause chaotic behavior . for example , the hnon - heiles system , which describes the motion of a star moving in the potential of a rotating galaxy , is chaotic@xcite . we also know that some spin - orbit interaction induces chaos in newtonian gravity . this may also be true in a relativistic system such as the evolution of a binary system . the motion of coalescing binary systems of neutron stars and/or black holes is very important to study because they are promising sources of gravitational waves , which we are planning to detect by large - scale laser interferometric gravitational observatories , such as us ligo@xcite . if we will detect the signal of gravitational waves emitted from these systems and compare it with theoretical templates , we may be able to determine a variety of astrophysical parameters of the sources such as their direction , distance , masses , spin , and so on@xcite . in order to extract exact information about such sources from the observed signal we need the exact theoretical templates of the gravitational waveforms . to make such templates , it is very important to know the exact motion of sources . hence , the equations of motion in the post - newtonian expansion in terms of a small parameter @xmath6 have been studied by many authors@xcite . those can be written schematically as @xmath7 where the subscripts @xmath8 and @xmath9 denote newtonian , post - newtonian , spin - orbit coupling , 2nd post newtonian , spin - spin coupling , and radiation reaction terms , respectively@xcite ; and the superscript corresponds to the order of expansion in @xmath10 . to make sufficient templates , we may need at least the 3rd - order post newtonian contribution to obtain the s / n ratio required from the observation@xcite-@xcite . but this is still under investigation in the world . the spin effect is also important . the spin terms in eq.([1.1 ] ) , @xmath11 and @xmath12 , induce a precession of the orbital plane through the spin - orbit or spin - spin coupling , resulting in modulation of the gravitational waveforms @xcite,@xcite . in @xcite , it is also shown that the orbital plane may behave very strangely due to the spin effects . we can not verify whether or not any chaotic behavior occurs in their system . but from the studies on spin effects in newtonian dynamics , we know that a spin effect can make a motion chaotic . we then expect that a relativistic system such as a coalescing binary pulsar may also show the similar non - linear phenomena . the gravitational waveform from the system with chaotic motion will be different from a system with a regular motion , for example , a regular precession of the orbital plane as shown in @xcite . the chaos might be too strong to make a complete template of gravitational waves , or rather it might give us new information about astrophysical parameters from a time series of the observed waveforms . we will discuss this problem for a coalescing binary system with highly spinning bodies elsewhere . thus we believe that a study about spin effects on the orbital evolution of a relativistic system and its gravitational waveform is very important from the viewpoint of observations as well as of academic interest . in this paper , to clarify the spin effect on the orbital motion , especially the spin - orbit interaction , we study the motion of a spinning test particle around a schwarzschild black hole . so far , studies about a spinning test particle in relativistic spacetime have been done by many authors since the basic equations were derived by papapetrou@xcite and reformulated by dixon@xcite . corinaldesi and papapetrou already discussed a spinning test particle in schwarzschild spacetime@xcite . but , apart from the supplementary condition , from which they adopted a different equation from the present standard one , they presented the basic equations and discussed some terms with physical interpretations . they did not analyze the orbits in detail from a viewpoint of the dynamical system . kerr or kerr - newman spacetime was also analyzed by several authors@xcite-@xcite . in @xcite,@xcite and @xcite , the effective potential of the spinning particle is given and the spin effects on the binding energy are discussed . in @xcite and @xcite , the gravitational waves produced by a spinning particle falling into a kerr black hole or moving circularly around it is discussed and the energy emission rate from those systems is calculated . but in those papers they discussed only the case of the orbit in the equatorial plane or on the symmetric axis of the black hole . since we are interested in chaotic motion induced by spin - orbit coupling here , we have to discuss the most generic situation , i.e. , the orbital motion off the equatorial plane . this paper is organized as follows . in section 2 we shall briefly review the basic equations , i.e. , the equations of motion for a spinning test particle in relativistic spacetime , a supplementary condition and some constants of motion . we specify the background spacetime to be a schwarzschild black hole , then we write down those equations and introduce a sort of effective potential " , which enables us to classify the particle behavior . in section 3 , performing numerical integrations , we show that chaos occurs for a highly spinning test particle . summary and some remarks follow in section 4 . throughout this paper we use units @xmath13 . we define the signature of the metric as @xmath14 . the equations of motion of a spinning test particle in a relativistic spacetime were first derived by papapetrou@xcite and then reformulated by dixon@xcite . those are a set of equations : @xmath15 where @xmath16 and @xmath17 are an affine parameter of the orbit , the 4-velocity of the particle , the momentum , and the spin tensor , respectively . @xmath18 is chosen as the proper time of the particle in this paper , then @xmath19 . the multipole moments of the particle higher than mass monopole and spin dipole are ignored . it is called the pole - dipole approximation . we need a supplementary condition which gives a relation between @xmath20 and @xmath21 , because @xmath21 is no longer parallel to @xmath20 in the present case . the r.h.s . of eq.([eqn : pdot ] ) denotes a spin - orbit coupling through a strong gravitational field . we adopt the following condition@xcite , @xmath22 this condition is related to how to choose the center of mass in an extended body , and this choice gives a consistent condition@xcite . using ( [ eqn : ps0 ] ) we can write down the relation between @xmath20 and @xmath21 explicitly , that is , @xmath23 , \label{eqn : p - v}\ ] ] where @xmath24 and @xmath25^{-1/2}\ ] ] is a normalization constant fixed by @xmath26 . @xmath27 is a unit vector parallel to the momentum @xmath28 , where the mass of the particle @xmath2 is defined by @xmath29 this system has several conserved quantities . regardless of the symmetry of the background spacetime , it is easy to show that @xmath2 and the magnitude of spin @xmath30 , defined by @xmath31 are constants of motion@xcite . if a geometry possesses some symmetry described by a killing vector @xmath32 associated with the symmetry , we can show that @xmath33 is also conserved@xcite . as for the background spacetime , we assume a schwarzschild spacetime , i.e. , @xmath34 where @xmath35 with @xmath3 being the mass of the black hole . because the spacetime is static and spherically symmetric , there are two killing vector fields , @xmath36 and @xmath37 . from ( [ eqn : killing ] ) , we find the constants of motion related with those killing vectors as @xmath38 @xmath39 and @xmath40 are interpreted as the energy of the particle and the @xmath41 component of the total angular momentum , respectively . because the spacetime is spherically symmetric , the @xmath42 and @xmath43 components of the total angular momentum are also conserved . then we have two additional constants of motion as @xmath44 @xmath45 because the background is spherically symmetric , without loss of generality we can choose the @xmath41 axis in the direction of total angular momentum as @xmath46 where @xmath47 . three constraint equations ( [ eqn : jz])@xmath48([eqn : jy ] ) with eq . ( [ eqn : total ] ) are reduced to @xmath49 @xmath50 are fixed from eq . ( [ eqn : ps0 ] ) with eqs . ( [ eqn : jx2])@xmath48([eqn : jz2 ] ) as @xmath51 the energy conservation eq.([eqn : energy ] ) is now @xmath52 the procedure to give initial conditions for calculating the orbital evolution of the particle is as follows . first , we give constants of motion @xmath30 , @xmath4 and @xmath39 with which the motion of the particle is bounded to a compact region of the spacetime . we will discuss how to do it in detail in the next subsection . we set the particle on the equatorial plane @xmath53 with @xmath54 and @xmath55 . next , we give the spatial components of the spin tensor @xmath56 . @xmath57 vanishes at @xmath53 from ( [ eqn : jx2 ] ) , which means that the initial spin is perpendicular to the radial direction . the last parameter we need to choose initially is , @xmath58 which determines the direction of the spin . note that @xmath59 and @xmath60 denote the spin anti - parallel and parallel to the positive @xmath41-direction , respectively . in this case the orbit of the particle is always restricted to the equatorial plane and chaos never occurs . we assume @xmath61 , which corresponds to the case that the @xmath41 component of spin points to the same direction as that of the total angular momentum . otherwise , the particle can not get into any relativistic region and chaos never occurs , as we will see later . from eq . ( [ eqn : spin ] ) with eqs . ( [ eqn : jx2])@xmath62 ( [ eqn : angle ] ) , we find that @xmath63 inserting eqs . ( [ eqn : jy2 ] ) and ( [ eqn : jz2 ] ) with these equations into eq . ( [ eqn : energy2 ] ) ( at @xmath64 ) , we get the quadratic equation for @xmath65 . solving it and inserting the result into eq . ( [ eqn : spin3 ] ) , we find the initial values for @xmath66 . @xmath67 and @xmath68 are determined from eqs . ( [ eqn : jy2 ] ) @xmath48 ( [ eqn : stph ] ) and @xmath69 is given by use of eq . ( [ eqn : mass ] ) . since we are interested in chaotic behavior of a particle orbit , we analyze a test particle which does not escape to infinity and does not fall into a black hole . therefore we have to choose appropriate parameters of the particle , @xmath39 , @xmath4 and @xmath30 . if we have an effective potential , such a choice is easy . for example , for a spinless particle traveling around a schwarzschild black hole with the orbital angular momentum @xmath70 , whose motion is restricted to a plane ( e.g. the equatorial plane @xmath71 , the effective potential of the particle is given as @xmath72 since the region where the particle with the energy @xmath39 can move is given by @xmath73 , it is easy to choose the energy @xmath39 such that the particle will move in a compact region . in the case of a spinning particle , because the spin - orbit coupling ( [ eqn : pdot ] ) is not a potential force and an additional dynamical variable , i.e. , the direction of the spin , exists , we can not find any effective potential in the 2-dimensional @xmath74-@xmath75 plane . to know the region where the particle can move , however , we do not need to find an effective potential itself . rather we need to know only the boundary of such a region , i.e. , a curve in the @xmath74-@xmath75 plane where both @xmath76 and @xmath77 vanish . from ( [ eqn : jy2 ] ) , @xmath78 also vanishes there , which means that the spin lies in the meridian plane and there remains no freedom of the spin direction ( @xmath59 or @xmath60 ) . from ( [ eqn : mass ] ) with @xmath79 , we can set @xmath80 where @xmath81 is an unknown function . inserting this into eq . ( [ eqn : spin ] ) with eqs . ( [ eqn : jx2 ] ) @xmath48 ( [ eqn : jz2 ] ) , we find the equation for @xmath82 as @xmath83 finally , from eq.([eqn : energy ] ) , we obtain the equations for such a curve as @xmath84 where @xmath85 , \label{eqn : effectivepotential}\ ] ] and @xmath86^{\frac{1}{2}}.\ ] ] the subscript @xmath87 corresponds to the direction of the spin , i.e. , if the @xmath41 component of the spin is the same direction as the total angular momentum ( @xmath88 , i.e. , @xmath89 ) , we take @xmath90 , while if it is the opposite ( @xmath91 , i.e. , @xmath92 ) , @xmath93 should be applied . note that @xmath94 never changes its signature during the evolution . imposing @xmath95 and @xmath96 , @xmath97 is reduced to the conventional effective potential for a spinless particle in a schwarzschild black hole , i.e. , eq . ( [ eqn : schwapot ] ) . the derivation also shows that the particle with energy @xmath39 can move in the region of the @xmath74-@xmath75 plane such that @xmath98 . then we shall call @xmath99 the effective potential " of a spinning particle in schwarzschild spacetime . the typical shape of the potential on the equatorial plane is shown in fig.1 . from this figure , we can see that the particle in the potential @xmath90 can move closer to the event horizon @xmath100 than that in the potential @xmath93 . this is because the spin - orbit interaction is a repulsive force if the spin is parallel to the orbital angular momentum and it will balance with the gravity , so that a stable periodic orbit closer to the event horizon becomes possible without falling into the black hole . if the direction of the spin is opposite , we find the reverse . such a spin - orbit interaction is induced through the gravitational interaction as the r.h.s of eq . ( [ eqn : pdot ] ) , which breaks the integrability of the particle motion . since we are interested in chaotic behavior , a large spin - orbit interaction may be more interesting . hence , in this paper , we study only particle motion in a strong gravitational field , i.e. , in the potential @xmath90 . in what follows , we drop the subscript @xmath101 . once @xmath4 and @xmath30 are given , we can depict a contour map of @xmath99 , with which we find an allowed region where a particle with @xmath39 can move . we find that the effective potential " @xmath102 is classified into four types depending on @xmath4 and @xmath30 . these are given in fig.2 . in fig.2a , the potential has one saddle and one minimal point on the equatorial plane , which appears when @xmath30 is small compared with @xmath4 . since the spin effect is small , the shape of the potential is similar to that for a spinless particle . we call this potential type ( b1 ) . the orbit in this case never becomes chaotic . fig.2b shows the potential which has two saddle points off the equatorial plane and one minimal point on the equatorial plane . the saddle points are located symmetrically on opposite sides of the equatorial plane . we call this potential type ( b2 ) . for fixed @xmath4 , as @xmath30 gets larger , type ( b1 ) changes into type ( b2 ) . this happens because a repulsive spin - orbit interaction is angle dependent . effective potential"s shown in both figs.2c and 2d have no bound region , although the latter case has a saddle point on the equatorial plane . the particle will eventually fall into a black hole . we call those potentials type ( u1 ) and ( u2 ) , respectively . these potentials appear when @xmath4 is small enough , i.e. , the centrifugal force is too small to balance with the gravity . the type ( u1 ) potential is similar to that for a spinless particle . in the case of type ( u2 ) , however , the situation is slightly different . because the saddle point marked by a cross in fig.2d is minimal in the @xmath74 direction and maximal in the @xmath75 direction , a particle will gradually depart from the equatorial plane and fall into a black hole . there is a potential barrier on the equatorial plane from the repulsive spin - orbit interaction . we show in fig.3 what values of @xmath4 and @xmath30 belong to which types . because types ( b2 ) and ( u2 ) never appear in the case of a spinless particle , we conclude that such strange behavior in those potentials is induced by a spin effect . as we will show later , we find that it is only for the type ( b2 ) that the particle motion can be chaotic . note that the value of @xmath30 at the bottom end of the ( u2 ) region is slightly smaller than @xmath103 , i.e. , at @xmath104 . hence the type ( u2 ) still appears with physically meaningful values of @xmath30 and @xmath4 , as we will discuss soon . using the bulirsch - stoer method @xcite , we integrate the equations of motion numerically for various values of parameters , @xmath39 , @xmath4 and @xmath30 and various initial configurations of @xmath105 and @xmath106 . as for the dynamical variables , we solve all components of the position vector @xmath107 , of the momentum @xmath108 , and of the spin tensor @xmath17 , and use the constraint equations ( [ eqn : ps0]),([eqn : mass]),([eqn : spin ] ) , ( [ eqn : energy ] ) @xmath48 ( [ eqn : jy ] ) to check the accuracy of our numerical integration . after @xmath109 orbital periods , which is the end of our numerical calculation , the relative errors are smaller than @xmath110 for each constraint . the methods with which we analyze the chaotic behavior are the poincar map and the lyapunov exponent . to make the poincar map , we adopt the equatorial plane ( @xmath111 ) as a poincar section and plot the point @xmath112 when the particle crosses the poincar section with @xmath113 . if the motion is not chaotic , the plotted points form a closed curve in the 2-dimensional @xmath74-@xmath114 plane , because a regular orbit will move on a torus in the phase space and the curve is a cross section of the torus . if the orbit is chaotic , some of those tori will be broken and the poincar map does not consist of a set of closed curves but the points will be distributed randomly in the allowed region . from the distribution of the points in poincar map , we can judge whether or not the motion is chaotic . the lyapunov exponent is another method to judge the occurrence of chaos and it also gives a naive estimation of the strength of chaos quantitatively . this denotes how fast the close orbits in the phase space will diverge in future . the lyapunov exponent @xmath115 is defined by @xmath116 where @xmath117 is the distance at time @xmath118 between two neighboring points in the phase space . if the orbit is chaotic , then @xmath115 will converge to some positive value , which means that the distance will diverge exponentially with a typical time scale @xmath119 . for a particle moving along a geodesic , we can calculate the lyapunov exponent by integration of the equation of geodesic deviation ; @xmath120 where @xmath121 is a deviation vector . in the present case , however , because a spinning test particle does not move along a geodesic as discussed in section 2 , we need another way to estimate the lyapunov exponent . here , we have adopted the method developed by sano and sawada@xcite to estimate the lyapunov exponent . we prepare a time series of @xmath122 , @xmath123 , @xmath124 and @xmath125 , where @xmath126 and @xmath41 are defined by @xmath127 and @xmath128 , and calculate the lyapunov exponent from such a set of data . the way to calculate the lyapunov exponent is as follows . let @xmath129 denote a set of points on some orbit in the 4-dimensional phase space , i.e. , @xmath130 where @xmath131 is a small time interval . a set of `` deviation vectors '' @xmath132 at a point @xmath133 is defined by @xmath134 where @xmath10 is a small constant chosen appropriately . after the evolution of a time interval @xmath135 , the orbital point @xmath133 will proceed to @xmath136 and neighboring points @xmath137 to @xmath138 . the deviation vectors @xmath139 are thereby mapped to @xmath140 which transformation is described by a matrix @xmath141 . then the lyapunov exponent @xmath115 is calculated by @xmath142 where @xmath143 is a set of arbitrary unit vectors . if the system is chaotic , @xmath115 will be positive and will not depend on choice of @xmath131 , @xmath10 , @xmath144 and @xmath145 . fig.4 shows the poincar maps for the total angular momentum @xmath1 and for several values of the spin @xmath30 = @xmath146 by @xmath147 . as for the spin of a particle with mass @xmath2 , we usually expect that @xmath148 . in fact , for a rotating kerr black hole with mass @xmath3 and angular momentum @xmath4 , we have the inequality @xmath149 , where the equality holds in the extreme limit . therefore , for the present case , @xmath150 should be much smaller than unity for a test particle , for which we assume that @xmath151 . however , if we would extend our analysis to the case of a relativistic binary system with masses @xmath152 and @xmath153 , we may find that @xmath154 as follows . @xmath2 and @xmath3 should be regarded as the reduced and total masses , i.e. , @xmath155 , and for the spin of the first relativistic star 1 , @xmath156 . then we have @xmath157 for the case of @xmath158 . therefore @xmath159 can be large as unity . because the value of @xmath160 is not a mathematical special bound@xcite , we will analyze the case of @xmath161 as well in order to see the spin effect more clearly . from figs.4a and 4b , which correspond to the type ( b1 ) potential , we find that no tori are broken . chaos never occurs . on the other hand , in the case of the type ( b2 ) potential , which is shown in figs.4c-4f , we see that some tori are broken , which means that chaos occurs for such orbits . this is the first example of chaotic behavior in the motion of a test particle in a schwarzschild spacetime . the sea of chaos spreads in these poincar sections as @xmath30 increases . we conclude that this chaotic behavior is induced by the spin effect . fig.5 shows the orbit in the 2-dimensional configuration space corresponding to each torus ( ( 1 ) @xmath48 ( 3 ) ) in fig.4d ( @xmath162 ) . the orbit ( 1 ) is almost perpendicular to the equatorial plane . the initial position is near the minimum point of the effective potential " . the orbit ( 2 ) is chaotic and the particle approaches the saddle points . the orbit ( 3 ) is constrained in the very narrow area near the equatorial plane contrary to the orbit ( 1 ) . the initial positions are located near the edge of the effective potential " . in fig.6 , we present the lyapunov exponents @xmath115 for orbits ( 1 ) @xmath62 ( 3 ) in fig.5 . @xmath163 is also given , where @xmath164 is the mean orbital period averaged after @xmath165 rotations around a black hole . @xmath163 gives a naive estimation for how many rotations we expect before the chaotic behavior becomes distinct . we may justify that the orbit ( 2 ) is strongly chaotic because of the large positive value of its lyapunov exponent . the reason why the lyapunov exponents of the orbits ( 1 ) and ( 3 ) converge to positive values may be either that the time series we prepared to calculate the lyapunov exponent are not sufficient or that the system is not integrable even through we can not see any break down of the tori in the poincar map . is there any critical value of the spin for occurrence of chaos ? if it is determined by the potential type , we obtain @xmath166 for @xmath1 from fig.3 . to confirm this , we plot the lyapunov exponent @xmath115 in terms of the value of the spin @xmath30 in fig.7 . after @xmath30 reaches the critical value @xmath167 , @xmath115 increases remarkably , which supports our conclusion . from our numerical investigation , we find several conditions for the occurrence of chaos when setting up the initial data . * a particle must move in the type ( b2 ) potential . * the particle must be bounded in a compact region , i.e. , @xmath168 , where @xmath169 is the potential energy at the saddle point @xmath170 . * the particle must have energy @xmath171 and the position @xmath105 , and the angle of the spin @xmath106 must be appropriately chosen in order for the particle to approach the saddle point . here we shall briefly discuss a mechanism for the occurrence of chaos . fig.7 shows the poincar map of five orbits with @xmath172 , @xmath160 and @xmath173 . note that @xmath174 for this case . we see four different types of orbits . * type ( i ) : the stable orbits which form a closed curve near the center of the poincar map . * type ( ii ) : the stable orbits which form a closed curve at the edge of the poincar map . * type ( iii ) : the stable orbits which form a closed curve in the upper or lower part of the poincar map . * type ( iv ) : the separatrix of the above three types of orbits . two hyperbolic fixed points in this poincar section exist . it is plausible that the type ( iv ) orbit plays an important role for the occurrence of chaos . the unstable manifold departing from one hyperbolic fixed point will join smoothly to the stable manifold of another hyperbolic fixed point . such an orbit is called a `` heteroclinic orbit '' . as @xmath39 increases from this state , the sea of chaos appears around the hyperbolic fixed points and spreads into whole poincar map . its behavior is shown in fig.8 . the chaos observed here may be caused by a `` heteroclinic tangle '' , although we need further detailed analysis . the next question is what causes the existence of type ( iv ) orbits . although we do not have a definite answer now , we find some correlation between an appearance of such an orbit and the type of effective potential " . as @xmath30 decreases , the type ( ii ) , and then types ( iii ) and ( iv ) will vanish and only the type ( i ) orbit remains . this transition is seen in fig.4 . for the case of @xmath175 ( fig.4c ) , we can not find the type ( ii ) orbit , and in the cases of @xmath176 and @xmath177 ( figs.4a and 4b ) , all orbits belong to the type ( i ) . the latter case belongs to the type ( b1 ) potential . hence , it seems that there exists some relation between the type of effective potential " and an appearance of types ( ii ) , ( iii ) and ( iv ) orbits . we may find some criterion for chaos using the effective potential " . in this paper , using the pole - dipole approximation , we study the motion of a spinning test particle near a schwarzschild black hole to clarify its dynamical properties such as a chaos . we find the motion of the particle can be chaotic under some appropriate conditions . because the motion of a spinless particle in this spacetime is never chaotic because of its integrability , this chaotic behavior is purely induced by the spin - orbit interaction . the effective potential " of the particle is also introduced to classify the dynamical behaviors . effective potential"s are classified into four different types depending on the total angular momentum @xmath4 and the spin @xmath30 . when @xmath4 is large , some orbits are bounded and the effective potential"s are classified into two types : for type ( b1 ) one saddle point ( unstable circular orbit ) and one minimal point ( stable circular orbit ) on the equatorial plane exist for small spin ; and for ( b2 ) two saddle points bifurcate from the equatorial plane and one minimal point remains on the equatorial plane for large spin . if @xmath4 is small , no bound orbits exist and the potentials are classified into another two types : for type ( u1 ) no extremal point is found for small spin ; and for type ( u2 ) one saddle point appears on the equatorial plane , which is unstable in the direction perpendicular to the equatorial plane , for large spin . the types ( b1 ) and ( u1 ) are the same as those for a spinless particle , but the types ( b2 ) and ( u2 ) are new potentials which appear through a spin - orbit coupling . the chaotic behavior is found only in the type ( b2 ) potential . we believe that the appearance of saddle points is important , because we find chaos only for the orbits which approach the saddle points . the critical value of the spin beyond which chaos will occur is @xmath178 for @xmath179 . we also present the lyapunov exponent , which increases rapidly after the spin @xmath30 gets larger than the above critical value . this supports the use of the `` effective potential '' as a criterion for chaos . the typical value of the lyapunov exponent is about several orbital periods of the particle , which may become important in some relativistic astrophysical phenomena . in a real astrophysical system such as a binary system , the symmetry of the system is lower than the present case . there may be other important effects in addition to the spin - orbit interaction , which make the motion more complicated . then we may expect that chaos occurs even in the real system , or that the other effects stabilize the system and chaos will never be found . we need further analysis taking into account the other effects such as the spin - spin interaction or a force due to multipole moments , even if we adopt a test particle analysis@xcite . as for the spin - spin interaction , we can analyze a spinning test particle in a rotating kerr black hole . in our preliminary analysis by the effective potential " near the equatorial plane , the critical value of the spin for chaos gets smaller as the angular momentum of the black hole becomes larger . the detailed analysis is under investigation . another important point is whether or not such a chaotic behavior , if it exists , affects any realistic astrophysical or physical phenomena . one of the important targets for such an investigation is a coalescing binary system , where we need general relativity , in particular relativistic dynamics of compact objects as we mentioned in section 1 . to examine it , since we have studied here only the condition for occurrence of chaos , then we next have to know the evolution of the system including emission of gravitational waves . the particle traveling around a black hole emits the gravitational waves , extracting the energy and the angular momentum from the system . this will tell us whether or not the evolutionary path will get into the region of parameter space where chaos will take place . we then have to calculate the emission rates of the energy and the angular momentum , @xmath180 and @xmath181 for the present system@xcite,@xcite and follow the evolution . we would like to thank m. sasaki , m. shibata , y. sota , h. tagoshi and t. tanaka for useful discussions and thank p. haines for reading the paper carefully . ss also acknowledges k. imafuku a stimulating discussion . this work was supported partially by the grant - in - aid for scientific research fund of the ministry of education , science and culture ( no . 06302021 and no . 06640412 ) , and by the waseda university grant for special research projects . in the text , we have used a spin tensor @xmath17 . however , it may be sometimes more convenient or more intuitive to describe the basic equations by use of a spin vector @xmath182 , which is defined by @xmath183 which gives the following constraint : @xmath184 the equations of motion ( [ eqn : xdot ] ) , ( [ eqn : pdot ] ) , and ( [ eqn : sdot ] ) are now @xmath185 where @xmath186 the relation between the 4-velocity and the momentum is @xmath187 where @xmath188 and @xmath5 is the normalization constant determined from @xmath189 . the supplementary condition to fix the center of mass is @xmath190 the spin vector is perpendicular to the 4-velocity as we expected . note that @xmath191 . equation ( [ eqn : spin ] ) is just @xmath192 in what follow , we assume a schwarzschild black hole as the background spacetime . we write down explicitly the relation ( [ def : spin ] ) between @xmath17 and @xmath193 as @xmath194 \label{eqn : sst}\\ s_r & = & r^2 \sin \theta \left[u^t s^{\theta\phi } -u^\theta s^{t \phi } + u^\phi s^{t \theta}\right ] \label{eqn : ssr}\\ s_\theta & = & r^2 \sin \theta \left[u^t s^{\phi r } -u^\phi s^{t r } + u^r s^{t \phi}\right ] \label{eqn : ssth}\\ s_\phi & = & r^2 \sin \theta \left[u^t s^{r \theta } -u^r s^{t \theta } + u^\theta s^{t r}\right ] \label{eqn : ssph}\end{aligned}\ ] ] conversely , eqs . ( [ eqn : sst ] ) @xmath48 ( [ eqn : ssph ] ) with eqs.([eqn : ps0 ] ) , ( [ eqn : sst ] ) give @xmath195 where @xmath196 using eqs.([eqn : jx2])@xmath48([eqn : stph ] ) , eqs.([eqn : sst])@xmath48([eqn : ssph ] ) are now @xmath197 \label{eqn : ssr2}\\ \tilde{s}_\theta & = & \frac{r}{u_t}\left[j\left ( \sin\theta - fu_r \frac{u_\theta}{r}\cos\theta\right ) -\frac{p_\phi}{\sin\theta}\right ] \label{eqn : ssth2}\\ \tilde{s}_\phi & = & \frac{r}{u_t}\left(p_\theta\sin\theta -j f u_r\frac{u_\phi}{r } \cos\theta\right ) \label{eqn : ssph2}\\ s_t & = & j\left(-f u_r \cos\theta + \frac{u_\theta}{r}\sin\theta\right ) \label{eqn : sst2}\end{aligned}\ ] ] this expression already includes the constants of motion for the angular momentum . as for the angle of the spin @xmath106 , we find that @xmath200 using this definition with ( [ eqn : spin2 ] ) , we have @xmath201 @xmath202 are then some functions of @xmath65 through eqs . ( [ eqn : sth3 ] ) and ( [ eqn : sph3 ] ) . inserting them into the energy conservation equation ( [ eqn : energy2 ] ) , we obtain the quadratic equation for @xmath65 . we also have eq . ( [ eqn : mass ] ) for @xmath203 . solving them , we can finally set up the initial data of @xmath204 and @xmath205 . setting @xmath206 and using eqs . ( [ eqn : mass ] ) and ( [ eqn : spin2 ] ) , we find the effective potential " ( [ eqn : effpot ] ) . after giving @xmath207 and @xmath39 and setting initial data , we can solve the dynamical equations ( [ eqn : xdot2 ] ) and ( [ eqn : pdot2 ] ) with algebraic equations ( [ eqn : v - u2 ] ) , ( [ eqn : ssr2 ] ) @xmath48 ( [ eqn : sst2 ] ) , although here we have solved for all variables using the dynamical equations in order to estimate the accuracy by the constraint equations . 99 j. moser , math . intelligencer , * 1 * , 65 ( 1978 ) . j. wisdom , proc . . london . * a413 * , 109 ( 1987 ) . c. w. misner , phys . * 22 * , 1071 ( 1969 ) . v. a. belinskii , i. m. khalatnikov and e. m. lifshitz , adv . * 19 * , 525 ( 1970 ) . j. d. barrow , phys . rep . * 85 * , 1 ( 1982 ) . y. sota , s. suzuki and k. maeda , class . ( 1996 ) , to be published . g. contopoulos , proc . a431 * , 183 ( 1990 ) ; ibid * a435 * , 551 ( 1991 ) . c. p. dettmann , n. e. frankel and n. j. cornish , phys . rev . * d50 * , r618 ( 1994 ) . u. yustsever , phys . * d52 * , 3176 ( 1995 ) . v. karas and d. vokrouhlick , gen . * 24 * , 729 ( 1992 ) . h. varvoglis and d. papadopoulos , astron . astrophys . * 261 * , 664 ( 1992 ) . l. bombelli and e. calzetta , class . * 9 * , 2573 ( 1992 ) . r. moeckel , commun . phys . * 150 * , 415 ( 1992 ) . m. hnon and c. heiles , astrophys . j. , * 152 * , 25 ( 1968 ) . a. abromovici , w. e. althouse , r. w. p. drever , y. grsel , s. kawamura , f. j. raab , d. shoemaker , l. sievers , r. e. spero , k. s. thorne , r. e. vogt , r. weiss , s. e. whitcomb and m. e. zucker , science , * 256 * , 325 ( 1992 ) . c. cutler , t. a. apostolatos , l. bildsten , l. s. finn , e. e. franagan , d. kennefick , d. m. markovic , a. ori , e. poisson , g. j. sussman and k. s. thorne , phys . * 70 * , 2984 ( 1993 ) . for a review , see , e.g. , t. damour , in_300 years of gravitation _ , edited by s. w. hawking and w. israel ( cambridge university press , cambridge , 1987 ) , p.128 . c. m. will , prepreint wugrav-94 - 3 ( gr - qc/9403033 ) . c. cutler and e. e. franagan , unpublished work in progress . h. tagoshi and t. nakamura , phys . rev . * d49 * , 4016 ( 1994 ) . h. tagoshi and m. sasaki , prog . phys . * 92 * , 745 ( 1994 ) . kidder , phys . d*52 * , 821 ( 1995 ) . t. a. apostlatos , c. cutler , g. j. sussman and k. s. thorne , phys . d. * 49 * 6274 ( 1994 ) . a. papapetrou , proc . * a209 * 248 ( 1951 ) . w. g. dixon , proc . a314 * , 499 ( 1970 ) ; _ ibid _ * a319 * , 509 ( 1970 ) ; journ . gen . rel . grav . * 4 * , 193 ( 1973 ) ; phil . trans . soc . * 277 * , 59 ( 1974 ) ; _ in isolated gravitating systems in general relativity _ , + edited by j. ehlers , ( north - holland , amsterdam , 1979 ) , p 156 . e. corinaldesi and a. papapetrou , proc . a209 * , 259 ( 1951 ) . s. n. rasband , phys . lett . * 30 * , 111 ( 1973 ) . k. p. tod and f. de felice , nuovo cim . * b34 * , 365 ( 1976 ) . r. hojman and s. hojman , phys . rev . * d15 * , 2724 ( 1977 ) . m. a. abramowicz and m. calvani , mon . not . . soc . * 189 * , 621 ( 1979 ) . e. wald , ann . 83 * , 548 ( 1974 ) . e. wald , phys . rev . * d6 * , 406(1972 ) . y. mino , m. shibata and t. tanaka , phys . d53 * , 622 ( 1996 ) . t. tanaka , y. mino , m. sasaki and m. shibata , preprint ou - tap-27 ( gr - qc/9602038 ) . r. rdiger , proc . soc . * a375 * , 185 ( 1981 ) ; _ ibid _ * a385 * , 229 ( 1983 ) . w. beiglbck , commun . * 5 * , 106 ( 1967 ) . w. press , b. p. flannery , s. teukolsky and w. t. vetterling , _ numerical recipies in c _ ( cambridge university press , cambridge , 1986 ) . m. sano and y. sawada , phys . * 55 * , 1082 ( 1985 ) . to find a non - singular `` effective potential '' , @xmath30 should be smaller than @xmath208 . recently , it has been pointed out that deformation of a neutron star by a tidal force may change the innermost stable orbit in a binary system ( see lai , f. a. raiso and s. l. shapiro , apj . * 420 * , 811 ( 1994 ) ) . it is one of the most important effects on observation of gravitational waves .

.15 in we study the motion of a spinning test particle in schwarzschild spacetime , analyzing the poincar map and the lyapunov exponent . we find chaotic behavior for a particle with spin higher than some critical value ( e.g. @xmath0 for the total angular momentum @xmath1 ) , where @xmath2 and @xmath3 are the masses of a particle and of a black hole , respectively . the inverse of the lyapunov exponent in the most chaotic case is about three orbital periods , which suggests that chaos of a spinning particle may become important in some relativistic astrophysical phenomena . the effective potential " analysis enables us to classify the particle orbits into four types as follows . when the total angular momentum @xmath4 is large , some orbits are bounded and the effective potential"s are classified into two types : ( b1 ) one saddle point ( unstable circular orbit ) and one minimal point ( stable circular orbit ) on the equatorial plane exist for small spin ; and ( b2 ) two saddle points bifurcate from the equatorial plane and one minimal point remains on the equatorial plane for large spin . when @xmath4 is small , no bound orbits exist and the potentials are classified into another two types : ( u1 ) no extremal point is found for small spin ; and ( u2 ) one saddle point appears on the equatorial plane , which is unstable in the direction perpendicular to the equatorial plane , for large spin . the types ( b1 ) and ( u1 ) are the same as those for a spinless particle , but the potentials ( b2 ) and ( u2 ) are new types caused by spin - orbit coupling . the chaotic behavior is found only in the type ( b2 ) potential . the `` heteroclinic orbit '' , which could cause chaos , is also observed in type ( b2 ) . .15 in

introduction basic equations for a spinning test particle numerical results summary and discussion

adaptive optics ( ao ) systems on large telescopes like naco at the vlt ( brandner et al . 2002 ) overcome the limitations introduced by earth s turbulent atmosphere in terms of image degradation and allow imaging and spectroscopy at the diffraction limit of these telescopes ( see beckers 1993 for a review ) . for example an 8m - class telescope offers an angular resolution of @xmath050 mas at 1.65 @xmath1 m . ao therefore enables the study of extragalactic targets at highest spatial resolution and e.g. directly allows the comparison of more distant galaxies with nearby ones , observed without ao in the same wavelength domain , at the same spatial resolution . the near infrared ( nir ) is a sensitive tracer of the mass dominating ( older ) stellar populations in galaxies . at the same time the nir is less affected by extinction , but still sensitive to the distribution and contribution of ( warm ) dust . studying the detailed morphology , dynamics and composition of the sources described below is therefore ideally done in the nir since especially the circum - nuclear regions of the targets are expected to be extincted . the x - ray background ( xrb ) in the 0.5 - 10 kev regime has been resolved into discrete sources by recent work on x - ray surveys like rosat , chandra , and xmm ( e.g. hasinger 1998 ; miyaji , hasinger , & schmidt 2000 ; mushotzki et al . 2000 ; giacconi et al . 2001 ; brandt et al . however , the astrophysical nature of these sources still remains unknown . especially two subjects are of interest in this context . * the hardness of the x - ray spectra : * it is found that the cosmic xrb is much harder than the x - ray emission of unobscured ( type i ) agn in the local universe . therefore the existence of a substantial obscured agn ( type ii ) population is required . these are found preferentially at lower redshifts ( @xmath2 ) , in contrast to predictions of xrb models ( gilli 2003 and references therein ) . an important question in understanding the nature of sources with hard x - ray spectra is whether they are hardened due to a large amount of intrinsic ( circum - nuclear ) absorption or whether the spectrum of the central engine itself is intrinsically hard . it is quite likely that the x - ray spectra are of composite nature , i.e. both the emission and absorption mechanisms are of importance . in this case it has to be determined how the nuclear properties are correlated with the properties of the corresponding hosts . * evolution of x - ray active agn : * a recent result ( tacconi et al . 2002 ) is that the more ( spectroscopically ) passive x - ray - bright , early - type galaxies may originate from a population of ulirg galaxies that contain qso - like active galactic nuclei . some of the ulirgs resemble local qsos in their nir and bolometric luminosities because they are very efficiently transforming dust and gas into stars and/or feed their central engine . however , ulirgs have smaller effective radii and velocity dispersions than the local qso / radio galaxy population . this then implies that their corresponding host and black hole masses must be smaller . smaller black holes are found in local seyfert galaxies . indeed , based on optical sdss spectra , the sources in our sample resemble those of liners or seyferts ( fig . it is therefore likely that they do not evolve into optically bright qsos but rather into quiescent field ellipticals or in a still active state into x - ray - bright , early - type galaxies that can be found in rosat and chandra based samples . if this finding is correct then our x - ray based sample should reveal host galaxy properties quite similar to agn - dominated and star formation dominated ulirgs . in other words they should fall close to the fundamental plane of early - type galaxies . the question is : how do they compare to l@xmath3 rotating ellipticals , giant ellipticals , optically / uv - bright , and low - z qsos / radio galaxies , i.e. what is their parent population ? the sloan digital sky survey ( sdss , e.g. york et al . 2000 ) with its multicolor photometric and spectroscopic data presents a rich source to search for astrophysically interesting objects . to address the above stated questions , using a cross - correlation of sdss optical and rosat x - ray data provided by the early data release ( edr , stoughton et al . 2002 ) and the rosat all sky survey ( rass , voges et al . 1999 ) , we searched for extragalactic , x - ray active objects brighter than @xmath4 with a natural guide star ( ngs ) brighter than @xmath5 in their vicinity . these magnitude units provide reasonable ao performance on 8m - class telescopes . recent ao implementations allow ngs / science target angular separations up to @xmath6 . these criteria resulted in a set of 27 galaxy / ngs pairs . to only select the best optical x - ray counterparts we rejected those objects fulfilling one ore more of the following conditions : + ( 1 ) _ angular separation between optical and x - ray position larger than @xmath6 . _ the sdss database identifies rosat sources with galaxies / agn if the separation between both positions is less than @xmath7 . + ( 2 ) _ hardness ratio of -1 or a stellar optical color @xmath8 . _ these are indicative for x - ray active stars , the former typical for white dwarfs . in this case the x - ray emission might be dominated by the ngs rather than the galaxy . + ( 3 ) _ x - ray emission which shows excess extend over the rosat psf width . _ here the x - ray flux is likely to be dominated by emission from an extended hot gas component at the center of a galaxy cluster ( cf . voges et al . 1999 ) . the final sample consists of 12 galaxy / ngs pairs in the redshift range @xmath9 . their x - ray luminosities ( calculated , assuming a power - law index of -2 and an average galactic absorption column density of @xmath10@xmath11 ) range from @xmath12 erg s@xmath13 to @xmath14 erg s@xmath13 ( fig . 1 ) . of these sources five objects have sdss spectra , one of which appears as a passive elliptical , although its x - ray luminosity indicates nuclear activity ( cf . comastri et al . 2002 ) . for the remaining galaxies we used their photometric redshifts . another source also shows a red nucleus ( source no . 5 in right inset of fig . 1 ) contained in the 2mass survey , indicating the presence of warm dust , heated by star formation and/or the central engine . the sdss and its cross - correlations with other surveys provide means to find suitable extragalactic sources for follow - up nir observations at highest angular resolution and sensitivity of 8m - class telescopes . large telescopes are also necessary , because the optical counterparts of deep x - ray sources become very faint . nir observations give complementary information to the optical and x - ray data in terms of highest resolution images of the detailed structure of the host galaxies . nir colors and spectra give information on the stellar content of the host galaxies and the importance of the presence of dust . furthermore dynamical information from spectra give estimates for host masses and possibly black hole masses , since the nir emission is a more accurate tracer of the mass in galaxies . beckers , j. m. 1993 , , 31 , 13 brandner , w et al . 2002 , the messenger , 107 , 1 brandt , w. n. et al . 2001b , , 122 , 2810 comastri , a. et al . 2002 , , 571 , 771 giacconi , r. et al . 2001 , , 551 , 624 gilli , r. 2003 , arxiv astrophysics e - prints , 3115 hutchings , j. b. , & neff , s. g. 1997 , , 113 , 550 hasinger , g. 1998 , an , 319 , 37 miyaji , t. , hasinger , g. , & schmidt , m. 2000 , , 353 , 25 mushotzky , r. f. et al . 2000 , nature , 404 , 459 richards , g. t. et al . 2001 , , 121 , 2308 stoughton , c. et al . 2002 , , 123 , 485 strateva , i. et al . 2001 , , 122 , 1861 tacconi , l. j. et al . 2001 , , 580 , 73 voges , w. et al . 1999 , , 349 , 389 york , d. g. et al . 2000 , , 120 , 1579

recent x - ray surveys have now resolved most of the x - ray background ( xrb ) into discrete sources . while this represents a breakthrough in the understanding of the xrb , the astrophysical nature of these sources still remains mysterious . in this article we present a sample of x - ray / optically selected extragalactic objects which are suitable for adaptive optics observations in the near infrared ( nir ) at highest angular resolution . the sample is based on a cross - correlation of the sloan digital sky survey and the rosat all sky survey . the nir properties can help to disentangle the nature of the x - ray bright , partially absorbed and spectroscopically passive background objects and their hosts . # 1_#1 _ # 1_#1 _ = # 1 1.25 in .125 in .25 in

introduction the rosat/sdss based sample conclusions

in the process known as internal diffusion limited aggregation ( idla ) one constructs for each integer time @xmath8 an * occupied set * @xmath9 as follows : begin with @xmath10 , and for each @xmath0 let @xmath11 be @xmath2 plus the first point at which a simple random walk from the origin hits @xmath12 . idla was first proposed by meakin and deutch @xcite as a model of industrial chemical processes such as electropolishing , corrosion and etching . after discussing these applications , they write `` it is also of some fundamental significance to know just how smooth a surface formed by diffusion limited processes may be . '' their numerical findings indicated that such surfaces are astonishingly smooth : in two and higher dimensions , the fluctuations appeared to be at most logarithmic in the number of particles . in the more than two decades since , this finding has resisted mathematical proof . diaconis and fulton @xcite identified idla as a special case of a `` smash sum '' operation on subsets of @xmath13 . lawler , bramson and griffeath @xcite proved that the asymptotic shape of the domain @xmath2 is a disk , and lawler @xcite showed that the fluctuations from circularity are at most of order @xmath14 up to logarithmic factors , where @xmath4 is the radius of the disk of area @xmath0 . later work @xcite related internal dla with multiple sources to variational problems in pde . however , the basic issue raised by meakin and deutch and again by lawler ( `` a more interesting question ... is whether the errors are @xmath15 for some @xmath16 . '' @xcite ) stood until asselah and gaudillire @xcite ( with an appendix by blachre ) improved the bound of @xmath14 in dimensions @xmath17 to @xmath18 . our main result shows that in dimension @xmath19 , the fluctuations from circularity are at most order @xmath20 . in fact we show somewhat more , namely that the probability that the fluctuations at time @xmath21 exceed @xmath22 decays like @xmath23 ( where @xmath24 can taken to be arbitrarily large , but the constant @xmath25 may depend on @xmath24 ) . let @xmath26 be the set of lattice points lying in the open disk of radius @xmath27 centered at @xmath28 . for convenience , we define the idla cluster @xmath29 for all @xmath30 by setting @xmath31 , where @xmath32 is the greatest integer @xmath33 ( and @xmath34 if @xmath35 ) . [ thm : logfluctuations ] for each @xmath24 there exists an @xmath36 such that for all sufficiently large @xmath27 , @xmath37 the proof will show that @xmath38 can be taken to be an affine function of @xmath24 . taking @xmath39 for large enough @xmath24 , theorem [ thm : logfluctuations ] in particular implies ( by borel - cantelli ) that the fluctuations from circularity are @xmath40 almost surely : @xmath41 this is the first part of a three part article . in the second part , entitled _ internal dla in higher dimensions _ @xcite , we show that internal dla in @xmath42 for @xmath43 has even smaller fluctuations than in dimension @xmath19 : the higher - dimensional analogue of holds with @xmath20 replaced by @xmath44 . we believe that these orders ( @xmath20 in dimension @xmath19 and @xmath44 in dimensions @xmath17 ) are the best possible . we discuss a possible approach to proving matching lower bounds , as well as several other open questions , in section [ s.conclusion ] . using methods rather different from ours , asselah and gaudillire @xcite have recently announced an independently obtained upper bound of order @xmath45 in all dimensions @xmath46 . in the third part , entitled _ internal dla and the gaussian free field _ @xcite , we show that the fluctuations from circularity have a weak limit which is a variant of the gaussian free field . while both this paper and the sequel address fluctuations of idla , the results are of a rather different nature : the present paper is concerned with fluctuations at the fine scale of individual lattice points . the sequel addresses average fluctuations , and will show that in a certain sense the fluctuation from circularity _ averaged _ over a constant fraction of the boundary ( here inner and outer fluctuations cancel each other out ) is only _ constant _ order . in fact , the fluctuations from circularity converge in law to a particular gaussian random distribution on the circle , whose time evolution is related to the gaussian free field . top : idla cluster @xmath2 for @xmath47 . early points ( see section [ sec : basecase ] ) are colored red , and late points are colored blue . bottom : closeup of a portion of the boundary of the cluster.,title="fig : " ] + top : idla cluster @xmath2 for @xmath47 . early points ( see section [ sec : basecase ] ) are colored red , and late points are colored blue . bottom : closeup of a portion of the boundary of the cluster.,title="fig : " ] we begin in section [ sec : nothintentacles ] with a lemma that rules out `` thin tentacles '' in the idla cluster ( lemma [ lem : thickening ] ) . more precisely , it is unlikely that @xmath48 and less than a constant fraction @xmath49 of lattice sites in the disk @xmath50 belong to @xmath2 . section [ sec : basecase ] establishes a rather rough a priori bound on the probability of certain types of very large fluctuations . section [ sec : exittimes ] gives some large deviation results for the exit time of brownian motion from an interval . these three sections comprise the background results needed for our argument . in order to control more precisely the deviation from circularity near a point @xmath51 , we define in section [ sec : hzetaproperties ] a discrete harmonic function @xmath52 which approximates the harmonic function @xmath53 where @xmath54 and @xmath55 are viewed as complex variables . ( discrete harmonicity will fail for @xmath56 only at a few points near @xmath55 . ) the function @xmath57 approximates the discrete poisson kernel for the disk @xmath58 , that is , the probability that simple random walk started at @xmath54 first exits @xmath58 at @xmath55 . we may extend the function @xmath56 linearly to the grid @xmath59 of horizontal and vertical line segments joining vertices of @xmath13 . if we begin with a collection of particles on the grid @xmath59 , and we allow some of them to move according to brownian motions on @xmath59 ( see section [ sec : martingale ] ) , according to determined starting and stopping rules , then the sum of @xmath56 over these particle locations is a continuous martingale ( provided that no particle continues to move after reaching a point where discrete harmonicity fails ) . level sets of @xmath60 . note that @xmath61 on @xmath62 and @xmath63.,width=192 ] by the martingale representation theorem ( see ( * ? ? ? * theorem v.1.6 ) ) , we can represent any continuous martingale as a time change of a standard brownian motion . using the representation theorem and large deviations for brownian exit times , we can control the likelihood that this martingale is larger or smaller than its mean by estimating its quadratic variation . we will ultimately show that large deviations of @xmath2 from circularity are unlikely without large deviations of the martingale for some @xmath55 . for example , the fact that thin tentacles are unlikely ( lemma [ lem : thickening ] ) means that with high probability , a point near some @xmath55 can not belong to @xmath2 unless many points near @xmath55 belong to @xmath2 . since @xmath64 is large for @xmath54 close to @xmath55 , this can be used to show that ( when @xmath55 is outside the typical range of @xmath2 ) the martingale is likely to be large if a point near @xmath55 belongs to @xmath2 . similarly , we can show that if @xmath55 fails to be hit ( long after we expect it to have been hit ) then the martingale has to be small at that time . ( for this step we actually use a modified version of the process in which particles are frozen when they first exit a set called @xmath65 that approximates @xmath58 , see figure [ fig : hzeta ] . ) the martingale is a brownian motion parameterized by its quadratic variation time , but the amount of quadratic variation time elapsed while @xmath2 is constructed is random . note that the conditional expected amount of time elapsed when depositing one particle ( given the locations of all previous particles ) is the conditional variance of the value of @xmath56 at the position that the particle is deposited . this value is easy to approximate on the event that @xmath2 is approximately circular . if , as a heuristic , we _ assume _ that @xmath2 is approximately circular for all @xmath0 , then a back of the envelope calculation shows that the total time lapse in constructing @xmath66 should be of order @xmath67 in dimension two ( and of constant order in all higher dimensions ) . on the event that the total time lapse is about right , the probability of having a martingale fluctuation of @xmath68 standard deviations should decay like @xmath69 . if @xmath5 is large enough then we should be able to sum over all @xmath55 and show that we are unlikely to see fluctuations of this magnitude anywhere . this heuristic actually suggests the basic strategy of our argument : we assume some amount of regularity for the growth of @xmath2 and use that assumption to show that the quadratic variation time lapse is unlikely to be large , hence large martingale fluctuations are unlikely , and hence the growth of @xmath2 is _ even more regular _ than we assumed . this strategy is pursued in sections [ sec : earlyimplieslate ] and [ sec : lateimpliesearly ] , which contain the heart of proof . using the ideas above , we will show that the presence of an `` early point '' ( a point @xmath54 visited at time less than @xmath70 ) implies , with high probability , the presence of a comparably late point at a previous time ( lemma [ lem : earlyimplieslate ] ) ; similarly , the presence of a late point implies , with high probability , the presence of a _ significantly earlier _ point at a previous time ( lemma [ lem : lateimpliesearly ] ) . by iterating these two lemmas , we will see that if there were a substantial probability to have even a slightly early or late point ( off by more than @xmath71 ) , then there would be a substantial probability of having a much larger fluctuation , which would contradict our a priori estimate . in general , we use @xmath72 to denote large constants and @xmath73 to denote small constants . unless explicitly specified otherwise , these are all positive absolute constants . let @xmath2 be the internal dla cluster formed from @xmath0 particles started at the origin in @xmath13 . for @xmath74 and a positive real number @xmath75 , let @xmath76 where @xmath77 is the euclidean norm . the following lemma is a variant of ( * ? ? ? * lemma 5.12 ) . the basic estimates involved are similar to those used by lawler , bramson and griffeath in their proof of the outer bound ; see in particular ( * ? ? ? * lemma 5b ) . [ lem : thickening ] there are positive absolute constants @xmath78 , @xmath79 , and @xmath80 such that for all real numbers @xmath81 and all @xmath74 with @xmath82 , @xmath83 in other words , it is unlikely that @xmath54 belongs to the cluster @xmath2 but less than a constant fraction of sites in the ball @xmath50 belong to the cluster . figure [ fig : tentacle ] depicts this unlikely scenario . a thin tentacle.,width=192 ] in fact , the exponent on the right side of can be improved to order @xmath84 ( and to order @xmath85 in dimensions @xmath43 ) . in appendix [ sec : tentacle ] , we prove these stronger bounds in all dimensions . we anticipate using these bounds for @xmath17 in our sequel paper _ internal dla in higher dimensions_. for purposes of the present paper , we will use only the bound in dimension @xmath86 . we expect a point @xmath74 to first join the idla cluster @xmath2 at time about @xmath87 . this motivates the following two definitions , which are illustrated in figure [ fig : earlylate ] . [ def : early ] for a positive real number @xmath75 , we say that @xmath88 is * @xmath75-early * if @xmath89 . for a time @xmath90 , let @xmath91 : = \bigcup_{z \in a(n ) } \left\{z \in a(\pi(|z|-m)^2 ) \right\}\ ] ] be the event that some point in @xmath92 is @xmath75-early . [ def : late ] for a positive real number @xmath93 , we say that @xmath88 is * @xmath93-late * if @xmath94 . for a time @xmath90 , let @xmath95 = \bigcup_{z \in \b_{\sqrt{n/\pi}-\ell } } \left\ { z \notin a(\pi ( |z|+\ell)^2 ) \right\}\ ] ] be the event that some point in @xmath96 is @xmath93-late . illustration of an idla cluster @xmath2 ( union of the black squares ) , an @xmath75-early point and an @xmath93-late point . the early point belongs to @xmath2 despite lying outside the disk @xmath97 , where @xmath98 ; the late point does not belong to @xmath2 despite lying inside the disk @xmath99.,width=480 ] the next lemma gives the a priori estimate on the probability of very late points that we will use to get the proof started in sections [ sec : earlyimplieslate ] and [ sec : lateimpliesearly ] . it follows easily from lemma 6 of @xcite ; for completeness , we include a proof in appendix [ sec : aprioriproof ] . [ lem : aprioriestimates ] there are absolute constants @xmath100 such that for all real @xmath101 , @xmath102 ) \le c_0 e^{-c_0\ell}.\ ] ] to avoid referring to too many unimportant constants , we take @xmath79 large enough and @xmath80 small enough so that the same @xmath79 and @xmath80 work in lemmas [ lem : thickening ] and [ lem : aprioriestimates ] . to prove theorem [ thm : logfluctuations ] we will show that for each @xmath24 there is a constant @xmath103 such that if @xmath27 is sufficiently large and @xmath104 and @xmath21 , then @xmath105 the next lemma is an easy observation which shows that implies theorem [ thm : logfluctuations ] . for all @xmath106 we have [ lem : earlylatealternatedefinition ] @xmath91^c = \bigcap_{n \leq n } \left \ { a(n ) \subset \b_{\sqrt{n/\pi}+m } \right \}\ ] ] @xmath95^c = \bigcap_{n \leq n } \left\ { \b_{\sqrt{n/\pi}-\ell } \subset a(n ) \right\}.\ ] ] taking complements in the definition , we obtain @xmath107^c & = \bigcap_{z \in a(n ) } \left\{z \not\in a(\pi(|z|-m)^2 ) \right\ } \\ & = \bigcap_{z \in a(n ) } \ ; \bigcap_{n \leq \min(n,\pi(|z|-m)^2 ) } \ { z \not\in a(n ) \ } \\ & = \bigcap_{n \leq n } \ ; \bigcap_{z \,:\ , |z| \geq \sqrt{n/\pi}+m } \{z \not\in a(n ) \ } \\ & = \bigcap_{n \leq n } \left\ { a(n ) \subset \b_{\sqrt{n/\pi}+m } \right\}. \end{aligned}\ ] ] likewise , @xmath108^c & = \bigcap_{z \in \b_{\sqrt{n/\pi}-\ell } } \left\{z \in a(\pi(|z|+\ell)^2 ) \right\ } \\ & = \bigcap_{z \in \b_{\sqrt{n/\pi}-\ell } } \ ; \bigcap_{\pi ( |z|+\ell)^2 \leq n \leq n } \ { z \in a(n ) \ } \\ & = \bigcap_{n \leq n } \ ; \bigcap_{z \,:\ , |z| \leq \sqrt{n/\pi}-\ell } \{z \in a(n ) \ } \\ & = \bigcap_{n \leq n } \left\ { \b_{\sqrt{n/\pi}-\ell } \subset a(n ) \right\}. \qed \end{aligned}\ ] ] for @xmath109 , write @xmath110 for the expectation on a probability space supporting a one - dimensional standard brownian motion @xmath111 started at @xmath112 . for @xmath113 , let @xmath114 \}\ ] ] be the first exit time of brownian motion from the interval @xmath115 $ ] . [ lem : exittime ] let @xmath116 and @xmath117 . if @xmath118 , then @xmath119 let @xmath120 , and for @xmath121 $ ] let @xmath122 by translating the interval @xmath115 $ ] to @xmath123 $ ] , it suffices to show that if @xmath124 , then @xmath125 the function @xmath126 obeys the heat equation in the strip @xmath127 $ ] , hence @xmath128 since @xmath129 is even and @xmath130 , we conclude that @xmath131 since @xmath132 is concave in the interval @xmath133 $ ] , we have @xmath134 which gives @xmath135 since @xmath136 for @xmath137 $ ] , we conclude that for @xmath138 $ ] , @xmath139 which establishes . the following well - known estimate appears as proposition ii.1.8 in @xcite . [ largedeviation ] let @xmath111 be a standard brownian motion with @xmath140 . then for any @xmath141 @xmath142 } b(s ' ) \ge ks \right\ } \le e^{-k^2 s/2}.\ ] ] fix a lattice point @xmath51 . in this section we define the discrete harmonic function @xmath56 resembling @xmath60 in , which we will use to detect early points near @xmath55 . for @xmath74 , let @xmath143 be the probability that an @xmath0-step simple random walk started at the origin in @xmath13 ends at @xmath54 . the _ recurrent potential kernel _ for simple random walk in @xmath13 is defined by @xmath144 this sum is finite for all @xmath74 , and its discrete laplacian @xmath145 vanishes for @xmath146 and equals @xmath6 if @xmath147 . here we have identified @xmath13 with @xmath148 . fukai and uchiyama @xcite and kozma and schreiber @xcite develop an asymptotic expansion for @xmath149 . we will use only the following estimate : there are absolute constants @xmath150 and @xmath79 ( rendered explicit in @xcite ; see also @xcite ) such that @xmath151 we will use this sharp asymptotic estimate of the potential kernel to show that the discrete harmonic function @xmath56 defined below is close to the continuum harmonic function @xmath60 ( lemma [ hsize]c ) , which in turn gives a discrete mean value property : the average value of @xmath56 on the discrete disk @xmath3 is very close to @xmath152 ( lemma [ hlevel]c ) . evidently , @xmath153 and @xmath154 is symmetric under the dihedral symmetries generated by the reflections across the axes @xmath155 , and the diagonal , @xmath156 . moreover , @xmath157 for all @xmath158 . exact values of @xmath154 near the origin can be computed using the mccrea - whipple algorithm , described in @xcite . in particular , @xmath159 ( we will use these exact values for convenience in specifying constants . but what we really need is not the actual values , but the property that @xmath160 , which is not surprising since @xmath161 is farther from the origin than @xmath6 . ) we construct the function @xmath56 for @xmath162 in the sector @xmath163 . ( for @xmath164 , we choose a dihedral symmetry @xmath165 such that @xmath166 , and define @xmath167 . ) let @xmath168 and write @xmath169 with @xmath170 and @xmath171 . the property we will use about the coefficients @xmath172 for @xmath173 is that they are positive and bounded . indeed @xmath174 . define @xmath175.\ ] ] note that ( using the symmetries ) @xmath176 > 0 \nonumber \\ & h_\zeta(\zeta + 1 ) = \frac{\pi}{2 } [ \alpha_2 g(1 ) - ( \alpha_1 + \alpha_2 ) g(1 ) ] < 0 \\ & h_\zeta(\zeta + 1+i ) = \frac{\pi}{2 } [ \alpha_1 g(1 ) - ( \alpha_1 + \alpha_2 ) g(1+i ) ] < 0 . \nonumber\end{aligned}\ ] ] the last inequality follows from the fact that @xmath177 . illustration of the set @xmath178 for @xmath179 . the function @xmath56 is discrete harmonic except at the three points @xmath55 , @xmath180 , and @xmath181 . its value is positive at @xmath55 and negative at @xmath180 and @xmath181 . for general @xmath182 , we have @xmath183 where @xmath184 is the first of these three points hit by a brownian motion in @xmath59 started at @xmath54 . large green dots on the edges of @xmath59 represent points @xmath182 for which @xmath185 . together with @xmath55 , these points form the boundary of @xmath178 . small black dots indicate points in @xmath186 . for @xmath187 , the value @xmath188 is proportional to the probability that a brownian motion in @xmath59 started at @xmath54 first exits @xmath178 at @xmath55.,width=384 ] extend @xmath56 linearly along each edge of the grid @xmath189 ( by an edge of @xmath59 we mean a line segment of length @xmath6 whose endpoints lie in @xmath13 . ) now define @xmath190 as the connected component of the origin in the set @xmath191 recall the discrete laplacian @xmath192 on @xmath13 , defined in . we say that a function @xmath129 defined on a subset @xmath193 is _ grid - harmonic _ if it is continuous and * for each edge @xmath194 of @xmath59 , if @xmath195 is an interval , then @xmath129 is linear on @xmath196 . * for each vertex of @xmath1 in the interior of @xmath197 , the sum of the slopes of the four line segments starting at @xmath54 ( directed away from @xmath54 ) is zero . for example , this holds if @xmath198 and @xmath129 is discrete harmonic on vertices and linear on edges . [ hsize ] * @xmath56 is grid - harmonic on @xmath190 , and @xmath199 on @xmath178 . * @xmath200 , and for all @xmath201 we have @xmath202 . * there is an absolute constant @xmath203 such that @xmath204 in particular , @xmath205 * @xmath206 . * by we have @xmath207 and @xmath208 , so the points @xmath209 and @xmath210 are not in @xmath178 . furthermore by definition @xmath211 . since @xmath56 is discrete harmonic at all other lattice points , its linear extension is grid - harmonic in @xmath178 . consequently , the inequality @xmath212 follows from the maximum principle . * this is immediate from the definition of @xmath178 . * the approximation for @xmath213 is trivial ( choosing @xmath214 sufficiently large ) . for @xmath215 , the bound implies @xmath216 by the fundamental theorem of calculus , @xmath217 * using the exact values @xmath218 and @xmath219 , we have @xmath220 with the constraints @xmath221 and @xmath222 hence @xmath223 and @xmath224 thus @xmath225 . for the lower bound , one checks that @xmath226 and hence @xmath227 is a concave function of @xmath228 . at the extreme points @xmath229 and @xmath230 , the values of @xmath231 are @xmath232 and @xmath233 . therefore , @xmath234 . @xmath235 a corollary of lemma [ hsize ] is the following . [ hlevel]there is an absolute constant @xmath236 such that * @xmath237 , where @xmath238 is the disk of radius @xmath27 in @xmath239 . * @xmath240 for @xmath241 and @xmath242 . * for all @xmath243 , @xmath244 * recall that lemma [ hsize ] ( c ) says @xmath245 we will show that @xmath246 @xmath247 since the level curve @xmath248 is the circle @xmath249 , this confirms part ( a ) with @xmath250 . + to prove and , note first that @xmath251 indeed , by the cauchy - riemann equations , this is the modulus of the ( complex ) derivative of @xmath252 . any curve from @xmath253 to @xmath254 has length at least @xmath5 . in particular following the integral curves vector field @xmath255 we see that @xmath60 changes by at least @xmath256 for @xmath54 along the curve . these integral curves are the circles through @xmath55 centered on the tangent line to @xmath253 at @xmath55 , and @xmath257 is nearly constant along the portion of the integral curve between @xmath253 and nearby concentric circles . therefore the change in @xmath60 is bounded as in , and similarly for . * the lower bound @xmath258 holds for all @xmath259 . as we saw in ( a ) , the level curves of @xmath60 differ from level sets of @xmath56 by at most a fixed distance @xmath260 . the supremum of @xmath60 in @xmath238 is @xmath261 . * first suppose that @xmath262 . by lemma [ hsize](c ) , @xmath263 + let @xmath264 be the union of unit squares around lattice points of @xmath3 . since @xmath265 , we have ( writing @xmath266 for area measure ) @xmath267 next , @xmath268 moreover , @xmath269 combining all these inequalities , we have shown that @xmath270 is within @xmath271 of @xmath272 the last equation holds because @xmath60 is harmonic in @xmath273 . + we now turn to the cases of @xmath178 and @xmath3 with @xmath274 . this follows from the case @xmath275 , which we have already proved . in fact , these other sums differ only on lattice sites in @xmath276 , and the sum over these sites is majorized the same way as in , @xmath277 one can replace @xmath56 with harmonic functions for which lemma [ hlevel](a ) and ( b ) hold with @xmath238 replaced by shapes other than disks . however , the approximate mean value property lemma [ hlevel](c ) relies on @xmath238 being a disk . the main ingredient is the continuum mean value property , which can be rephrased as saying that harmonic measure on @xmath278 from the origin is the same as from a uniform point in @xmath238 . accordingly , the place where the approximate mean value property lemma [ hlevel](c ) is used marks a crucial step in the proof : the bound for @xmath3 is used in the proof of lemma [ lem : earlyimplieslate ] , and the bound for @xmath178 in the proof of lemma [ lem : lateimpliesearly ] . for each integer @xmath0 , recall that the set @xmath2 consists of @xmath0 lattice points with @xmath280 growing according to the idla scheme . for @xmath51 , we will define a modified cluster @xmath281 in which particles are stopped on exiting @xmath178 ; moreover , we will define @xmath282 for all @xmath283 , not just for integer values . to define the modified cluster , recall first the notion of brownian motion @xmath284 on the grid @xmath285 to construct @xmath284 , take a standard brownian motion @xmath286 starting at @xmath28 at time @xmath287 until the first time @xmath288 that @xmath289 . for each open interval of the set @xmath290 choose a direction ( up , down , left , right ) with equal probability @xmath291 , independently . define a corresponding motion of a particle @xmath292 in an interval in which right was chosen ( and @xmath293 in an interval in which up was chosen , and similarly for left and down ) . on each interval except the final one , the particle returns to the origin . on the final interval it reaches a lattice point adjacent to the origin for the first time . we continue the process ( centered at this new lattice point ) until it hits a point adjacent to that one , etc . let @xmath294 be independent brownian motions on the grid @xmath59 , and for each integer @xmath295 , let @xmath296 be the first time that @xmath297 visits either a lattice point outside the cluster @xmath281 , or any point outside the set @xmath178 . then for @xmath298 , define @xmath299 where @xmath300 ( and by definition , @xmath301 ) . the choice of @xmath302 is not important here ; for our purposes , any increasing function @xmath303 satisfying @xmath304 and @xmath305 as @xmath306 would suffice in its place . effectively , the particle moves at an ever - increasing rate so that it _ _ visits the set @xmath307 in less than unit time . the points of @xmath308 where ( eventually ) many particles accumulate include @xmath309 , but are typically non - lattice points on the grid @xmath59 . we regard @xmath282 as a multiset : it may have points of multiplicity greater than one on the boundary @xmath310 , or at interior vertices at times when the moving particle is on top of an occupied site . when summing a function over @xmath282 , we take these multiplicities into account ; thus for a function @xmath129 defined on @xmath311 and @xmath312 , the notation @xmath313 means , @xmath314 recall the function @xmath56 defined in section [ sec : exittimes ] . because @xmath56 is grid - harmonic in @xmath190 ( that is , discrete harmonic at lattice points , and linear on edges ) , the process @xmath315 ( summing with multiplicity ) is a continuous - time martingale adapted to the filtration @xmath316 we will refer to @xmath317 as the martingale with pole at @xmath55 . by the martingale representation theorem of dambis and dubbins - schwarz ( see , e.g. ( * ? ? ? * theorem v.1.6 ) ) , @xmath318 where @xmath319 is a standard brownian motion and @xmath320 is the quadratic variation of @xmath317 . note that @xmath321 is a stopping time with respect to the filtration @xmath322 , where @xmath323 ; and @xmath319 is adapted to this filtration . by the strong markov property ( ( * ? ? ? * theorem 2.16 ) ) , for any @xmath0 , the process @xmath324 is a standard brownian motion that is independent of @xmath325 . it follows that we can write @xmath326.\ ] ] where @xmath327 are independent standard brownian motions started at @xmath28 . for @xmath328 , let @xmath329 \}\ ] ] be the first exit time of @xmath330 from the interval @xmath115 $ ] . [ lem : boundedbyexittime ] fix @xmath51 . let @xmath331 @xmath332 then @xmath333 by the maximum principle , since @xmath56 is harmonic on @xmath334 , we have @xmath335 for all @xmath336 . hence for all @xmath337 $ ] , we have @xmath338 that is , for all @xmath339 $ ] , we have @xmath340 that is , for @xmath341 $ ] , we have @xmath342 hence the first exit of @xmath343 from the interval @xmath344 $ ] occurs after time @xmath345 , which gives @xmath346 . fix @xmath347 . to detect the presence of @xmath75-early points in the aggregate @xmath29 , we will use the martingales @xmath317 for @xmath55 lying just outside the aggregate . the next lemma helps control the elapsed time @xmath348 on the event that there are no @xmath349-early points . define @xmath350 [ lem : earlyquadraticvariation ] for all @xmath351 and all @xmath51 such that @xmath352 , @xmath353^c } \right ] \leq t^{40}.\ ] ] fix @xmath354 , and abbreviate @xmath355 . on the event @xmath356^c$ ] , we have for all @xmath357 @xmath358 hence by lemma [ hlevel](b ) , on the event @xmath356^c$ ] we have @xmath359 for all positive integers @xmath357 and all @xmath336 . we obtain from lemma [ lem : boundedbyexittime ] , @xmath360^c } \le \tau_n \left(-a , b_n \right).\ ] ] where @xmath361 summing over @xmath362 yields @xmath363^c } \leq e^{s_\zeta(t ) \one_{\early_{m+1}[t]^c } } \leq e^{\sum_{n=1}^{t } \tau_n(-a , b_n)}.\ ] ] the brownian exit times @xmath364 for @xmath365 are independent . using lemma [ lem : exittime ] with @xmath366 , we obtain @xmath367 setting @xmath368 we have @xmath369 hence @xmath370 . we conclude that @xmath371^c } \right ] \leq ( r_0/c_2)^{10 \pi } \leq t^{40}. \qed\ ] ] in order to study late points in the cluster @xmath29 , we will need a variant of lemma [ lem : earlyquadraticvariation ] which helps control the elapsed time @xmath348 when @xmath372 is slightly smaller than @xmath373 . in this case the quadratic variation of @xmath317 is larger due to particles that accumulate on the boundary @xmath310 . [ lem : latequadraticvariation ] fix @xmath374 and @xmath375 . fix @xmath51 , and let @xmath376 . if @xmath377 , then @xmath378^c } \right ] \leq t^{40 } e^{700 m}.\ ] ] let @xmath379 , and suppose first that @xmath380 . let @xmath381 . then @xmath382 , so by lemma [ lem : earlyquadraticvariation ] , @xmath383^c } \right ] \leq \ee \left [ e^{s_\zeta(t_0 ) } \one_{\early_{m+1}[t_0]^c } \right ] \leq t_0^{40 } \leq t^{40}.\ ] ] ( here we have used the trivial inclusion @xmath384^c \subset \early_{m+1}[t_0]^c.$ ] ) furthermore , because @xmath385 on all of @xmath310 , we have by lemma [ lem : boundedbyexittime ] for @xmath386 @xmath387 hence , using the inclusion @xmath388^c \subset \early_m[t_0]^c$ ] , @xmath389^c } \leq e^{s_\zeta(t)\one_{\early_{m}[t_0]^c } } \leq e^{s_\zeta(t_0)\one_{\early_{m}[t_0]^c } } \prod_{n = t_0 + 1}^{t } e^{\tau_n(-\frac{1}{2\rho } , 2)}.\ ] ] by lemma [ lem : exittime ] , we have @xmath390 for @xmath391 , hence @xmath392 in the last inequality we have used that @xmath393 and @xmath394 . since the exit times @xmath395 for @xmath396 are independent of one another and of @xmath397 , we obtain @xmath398^c } \right ] & \leq \log \ee \left [ e^{s_\zeta(t_0)\one_{\early_{m}[t_0]^c } } \right ] + \sum_{n = t_0 + 1}^{t } \log \ee e^{\tau_n } \\ & \leq 40 \log t + 400m.\end{aligned}\ ] ] it remains to consider the case @xmath399 . in this case we take @xmath400 , obtaining @xmath401 now we begin the main part of the proof of theorem [ thm : logfluctuations ] . fix @xmath402 . with constants @xmath78 , @xmath79 and @xmath80 given by lemma [ lem : thickening ] , and @xmath403 given by lemma [ hsize ] , we set @xmath404 ( in particular , if @xmath405 and @xmath406 , then @xmath407 . ) recall the events @xmath408 = \big \ { \mbox{some $ z \in a(t)$ is $ m$-early } \big\}\ ] ] and @xmath409 = \big \ { \mbox{some $ z \in \b_{\sqrt{t/\pi}-\ell}$ is $ \ell$-late } \big\}\ ] ] defined in section [ sec : basecase ] . ( early points imply late points ) [ lem : earlyimplieslate ] fix @xmath402 , @xmath410 and @xmath411 . if @xmath412 , then @xmath413 \cap \late_{\ell}[t]^c ) \leq t^{-(\gamma+1)}.\ ] ] in other words , with high probability either there is no @xmath75-early point by time @xmath414 , or there is a @xmath415-late point by time @xmath414 . the proof consists of two main estimates . first , if @xmath74 is the first @xmath75-early point and joins the cluster at time @xmath354 , then for a suitably chosen point @xmath55 near @xmath54 , the quadratic variation @xmath348 is small : equation below . second , if in addition there are no @xmath93-late points by time @xmath354 , then the martingale @xmath279 is large : equation below . to finish the proof we use the large deviation bound for brownian exit times to show that these two events ( small @xmath348 and large @xmath279 ) are unlikely both to occur . without loss of generality we assume that @xmath416 . note that @xmath417 $ ] is empty if @xmath418 , so we may also assume @xmath419 . for each integer @xmath420 and each lattice point @xmath54 , let @xmath421^c\ ] ] be the event that @xmath54 first joins the cluster at time @xmath354 and @xmath54 is @xmath75-early , but no previous point is @xmath75-early . since @xmath422 , @xmath423.\ ] ] as in lemma [ lem : earlyquadraticvariation ] , we write @xmath424 fix a lattice point @xmath425 , and consider the unit square with lattice point corners containing @xmath426 . define @xmath427 to be the corner of that square farthest from the origin . then @xmath428 by lemma [ lem : earlyquadraticvariation ] , we have for @xmath429 @xmath430^c } \leq t^{40}.\ ] ] hence by markov s inequality , letting @xmath431 , we have @xmath432^c \cap \ { s_{\zeta}(t ) > s \ } ) \leq t^{-(2\gamma+100 ) } \ee e^{s_\zeta(t ) } \one_{\early_{m+1}[t]^c } \leq t^{-(2\gamma+60)}.\ ] ] on the event @xmath433 , there are no @xmath75-early points before time @xmath354 ; moreover , @xmath54 is at unit distance from some point of @xmath434 , so @xmath54 is not @xmath349-early . therefore , @xmath435^c.\ ] ] and hence @xmath436 next , we show that for @xmath437 , @xmath438^c \cap q_{z , t } ) < t^{-(\gamma+5)}.\ ] ] in other words , on the event @xmath439^c \cap q_{z , t } $ ] it is likely that @xmath440 . on the event @xmath441 , no @xmath442 is @xmath349-early , so we have @xmath443 for all @xmath444 . in other words , @xmath445 hence lemma [ hlevel](a ) implies that @xmath446 does not meet @xmath310 . in other words , on the event @xmath433 we have @xmath447 partition @xmath282 into three disjoint sets as follows . @xmath448 on the event @xmath449 , the ball of radius @xmath450 is filled by @xmath282 , so @xmath451 and since @xmath282 has @xmath354 points , @xmath452 has at most @xmath453 points . in particular , @xmath454 and @xmath455 since @xmath456 is non - negative on @xmath178 and , by lemma [ hsize](c ) , @xmath152 is close to @xmath457 . furthermore , lemma [ hlevel](c ) implies @xmath458 where in the last step we have used that @xmath459 and @xmath460 . on the other hand , since @xmath461 , it follows from lemma [ lem : thickening ] that @xmath462 next , recall that on the event @xmath433 , the point @xmath54 is @xmath75-early but not @xmath349-early , so that @xmath463 . moreover , @xmath55 is at distance roughly @xmath464 from @xmath465 along the ray @xmath28 to @xmath465 . @xmath466 , and using lemma [ hsize](c ) we find that @xmath467 for all @xmath468 . furthermore , @xmath469 and @xmath470 , so @xmath471 . therefore , on the event @xmath472 , we have @xmath473 summing the contributions to @xmath474 from @xmath475 , @xmath476 and @xmath477 , we find that on the event @xmath472 we have @xmath478 . hence @xmath479 intersecting with @xmath449 , we obtain @xmath480 which proves . recall that @xmath481 . since @xmath411 we have @xmath482 . using the large deviation bound for the maximum of brownian motion , lemma [ largedeviation ] , we obtain @xmath483 } b_\zeta(s ' ) \geq s \right\ } \nonumber \\ & \leq e^{-s/2 } = t^{-(\gamma+50)}. \label{eq : nolargemean } \end{aligned}\ ] ] finally , we assemble these ingredients to bound @xmath484 \cap \late_\ell[t]^c ) = \sum_{t = 1}^t \sum_{z \in \b_t } \pp(q_{z , t } \cap \late_\ell[t]^c).\end{aligned}\ ] ] split each term in the sum on the right into three pieces , with @xmath437 , as indicated below and majorize using , , and , respectively . we obtain with @xmath485 , @xmath486^c ) & \le \pp ( q_{z , t } \cap \{s_\zeta(t ) > s \ } ) \\ & \qquad+ \pp ( q_{z , t } \cap \{m_\zeta(t ) < ( b/10)m \ } \cap \late_\ell[t ] ^c ) \\ & \qquad \qquad + \pp ( \{s_\zeta(t ) \le s \ } \cap \ { m_\zeta(t ) \ge ( b/10 ) m \ } ) <3 t^{-(\gamma+5)}\end{aligned}\ ] ] summing over @xmath487 and @xmath488 , since @xmath410 we conclude that @xmath489 \cap \late_\ell[t]^c ) \leq ( 2\pi t^2 ) \cdot t \cdot 3t^{-(\gamma+5 ) } < t^{-(\gamma+1)}.\ ] ] this completes the proof of lemma [ lem : earlyimplieslate ] . fix @xmath402 , and let @xmath490 where @xmath403 is the constant appearing in lemma [ hlevel ] , and @xmath79 and @xmath80 appear in lemma [ lem : aprioriestimates ] . ( late points imply early points ) [ lem : lateimpliesearly ] if @xmath491 and @xmath492 , then @xmath493 \cap \early_m[t]^c ) \leq t^{-(\gamma+1)}.\ ] ] in other words , with high probability , either there is no @xmath93-late point by time @xmath414 , or there is an @xmath494-early point by time @xmath414 . without loss of generality we take @xmath495 . note that since @xmath496 this ensures @xmath497 . let @xmath498 be such that @xmath499 , and set @xmath500 note that @xmath501 . denote by @xmath502 = \{\zeta \not\in a(t_1 ) \ } $ ] the event that @xmath309 is @xmath93-late . then @xmath409 = \bigcup_{|\zeta| \leq \sqrt{t/\pi}-\ell } l[\zeta].\ ] ] we will use the martingale @xmath503 with pole at @xmath309 . we first show that on the event @xmath504 $ ] , @xmath505 to prove this , note first that by the maximum principle , the smallest value of @xmath56 on @xmath178 is @xmath506 attained on the boundary @xmath308 . this boundary value is achieved everywhere on @xmath310 except at the point @xmath309 , at which @xmath56 has a much larger value . on the event @xmath502 $ ] , no particle reaches @xmath309 by time @xmath507 , so the values of @xmath56 at all boundary points reached at time @xmath508 are @xmath506 , which is smaller than the interior values . therefore , @xmath509 is maximized if all the sites of @xmath190 are occupied , i.e. , @xmath510 . by lemma [ hlevel](a ) , we have @xmath511 , so @xmath512 since @xmath496 , the right side is at least @xmath513 . hence on the event @xmath502 $ ] , we have @xmath514 by lemma [ hsize](c ) , we have @xmath515 , which shows that the first term on the right side of is less than @xmath516 . by lemma [ hlevel](c ) , the second term is at most @xmath517 . this concludes the proof of . now suppose that @xmath518 . recall that @xmath519 and @xmath520 , so @xmath521 . by lemma [ lem : latequadraticvariation ] , @xmath522^c } \leq t^{40 } e^{700 m } \leq e^{740m}.\ ] ] letting @xmath523 , we have by markov s inequality @xmath524^c ) \leq e^{-s } \ee e^{s_\zeta(t_1 ) } \one_{\early_m[t_1]^c } \leq e^{-10m}.\ ] ] now by , @xmath525 \cap \early_m[t]^c ) & \leq \pp ( \early_m[t]^c \cap \{s_\zeta(t_1)>s \ } ) + \pp ( \{s_\zeta(t_1 ) \leq s \ } \cap l[\zeta ] ) \\ & \leq e^{-10 m } + \pp \big\{s_\zeta(t_1 ) \leq s , \ , m_\zeta(t_1 ) \leq -\ell \big\}. \end{aligned}\ ] ] we use lemma [ largedeviation ] to bound the second term . since @xmath526 , where @xmath527 is a standard brownian motion with @xmath528 , we have @xmath529 } b_\zeta(s ' ) \leq -\ell \right\ } \leq e^{- \ell^2 /2s } . \end{aligned}\ ] ] we conclude that for all @xmath51 such that @xmath530 , @xmath531^c \cap l[\zeta ] ) \leq e^{-10 m } + e^{-\ell^2 / 1500m}.\ ] ] since @xmath532 and @xmath533 by hypothesis , each term on the right side is at most @xmath534 . hence for @xmath535 , we have @xmath536^c \cap l[\zeta ] ) \leq 2t^{-(\gamma+3)}.\ ] ] to take care of the case @xmath537 , we use lemma [ lem : aprioriestimates ] , which gives @xmath538\right ) = \pp(\late_\ell [ 4\pi \ell^2 ] ) \le c_0 e^{-c_0\ell } < t^{-(\gamma+3)}.\ ] ] summing over @xmath539 yields @xmath540^c \cap \late_\ell[t ] ) & \leq \pp \left(\bigcup_{|\zeta| < \ell } l[\zeta]\right ) + \sum_{\ell \leq |\zeta | \leq \sqrt{t/\pi}-\ell } \pp(\early_m[t]^c \cap l[\zeta ] ) \\ & < t^{-(\gamma+3 ) } + t^{-(\gamma+2)}. \end{aligned}\ ] ] this completes the proof of lemma [ lem : lateimpliesearly ] . the proof of the main result from lemmas [ lem : earlyimplieslate ] and [ lem : lateimpliesearly ] is a routine iteration . we use the lemmas in their contrapositive direction ( no early point implies no late point and no late point implies no early point ) for progressively smaller values of @xmath75 and @xmath93 , starting from . fix @xmath402 and @xmath541 . . then imples @xmath543 ) \leq c_0 e^{-c_0 \ell_0}.\ ] ] if @xmath414 is sufficiently large , say @xmath544 , then the right side is less than @xmath545 . now define @xmath546 lemma [ lem : earlyimplieslate ] implies that if @xmath547 , then @xmath548\cap \late_{\ell_0}[t]^c ) \leq t^{-(\gamma+1)}.\ ] ] so that @xmath548 ) < 2 t^{-(\gamma+1)}.\ ] ] for @xmath549 , define @xmath550 and @xmath551 recursively by @xmath552 by lemma [ lem : lateimpliesearly ] , if @xmath553 , then @xmath554\cap \early_{m_0}[t]^c ) \leq t^{-(\gamma+1)}\ ] ] hence @xmath554 ) < 3 t^{-(\gamma+1)}.\ ] ] by induction on @xmath555 , we find that @xmath556 ) < ( 2k+1 ) t^{-(\gamma+1)}\ ] ] and @xmath557 ) < ( 2k+2 ) t^{-(\gamma+1)}\ ] ] provided @xmath558 and @xmath559 . the recursion gives @xmath560 and @xmath561 where @xmath562 . let @xmath563 . then @xmath564 for @xmath565 , we have @xmath566 therefore , if we choose @xmath555 largest so that @xmath567 , then @xmath568 thus , in our final step we may use @xmath569 to obtain @xmath570 ) < ( 2k + 3 ) t^{-(\gamma+1 ) } < \frac12 t^{-\gamma } \\ \pp(\early_{m}[t ] ) < ( 2k + 4 ) t^{-(\gamma+1 ) } < \frac12 t^{-\gamma}\end{aligned}\ ] ] for all sufficiently large @xmath414 . let @xmath571 and @xmath572 . then @xmath573 . by lemma [ lem : earlylatealternatedefinition ] , we conclude that @xmath574 ) + \pp(\early_m[t ] ) \\ & < t^{-\gamma } < r^{-\gamma}. \end{aligned}\ ] ] this completes the proof of theorem [ thm : logfluctuations ] . we mention here a few open questions and possible extensions of our results , along with a brief overview of our forthcoming sequels . let @xmath575^d \subset \mathbb z^d$ ] and let @xmath576 be an instance of the discrete gaussian free field on @xmath197 with zero boundary conditions . that is 1 . write @xmath577 2 . let @xmath456 be the @xmath578-dimensional vector space of real - valued functions on @xmath197 that vanish on the boundary @xmath579 , 3 . and choose @xmath576 from the measure @xmath580 ( normalized to be a probability measure ) , where @xmath581 is lebesgue measure on @xmath456 . @xmath43 + order of @xmath582 at typical @xmath583 & @xmath584 & @xmath585 & @xmath6 + order of @xmath586 & @xmath584 & @xmath587 & @xmath588 + order of @xmath589 & @xmath584 & @xmath6 & @xmath590 + table [ table : gfforder ] gives the orders of various observables associated to @xmath576 . ( see @xcite for recent results on the maximum of @xmath576 in two dimensions , and @xcite for a general survey on the gff . ) perhaps surprisingly , the orders in the last two rows do not change if we replace @xmath197 with a fixed co - dimension one subset of @xmath197 . this shows that weighted averages of @xmath576 are remarkably well concentrated ( and become more concentrated in higher dimensions ) . appropriately normalized , they scale to gaussian random variables with an asymptotic variance that depends @xmath165 . next , if we let @xmath591 represent the `` lateness '' of the vertex @xmath54 , then we expect @xmath592 and its higher dimensional analogues to behave similarly to @xmath576 in all dimensions . ( this is trivial for dimension @xmath593 . ) our sequel paper , _ internal dla and the gaussian free field _ @xcite addresses the bottom row ( for certain test functions @xmath165 ) and shows that @xmath592 has a variant of the continuum gaussian free field as a scaling limit . the present paper established the upper bound of @xmath587 for the fluctuations of @xmath592 corresponding to the middle square of table [ table : gfforder ] . simulations @xcite indicate that fluctuations of this order are really present , and proving this is a very natural open problem . one possible approach would be to use the methods of this paper to show that for small enough @xmath594 , the _ expected _ number of aberrations of size @xmath595 ( or @xmath596 in dimension @xmath43 ) in @xmath2 grows like a _ positive _ power of @xmath0 and then using a second moment estimate to show that infinitely many such aberrations occur almost surely . the main new ingredient required would be to control the correlations between fluctuations in two different directions . the @xmath17 sequel to this paper @xcite uses variants of the arguments presented here . the function @xmath56 is replaced by a discrete green s function for a ball in @xmath42 ( with zero boundary conditions ) . the obvious analogue of @xmath56 ( constructed from differences of @xmath42 green s functions ) does not have spherical level surfaces . this causes some extra complication in the estimates involved in lemmas [ hsize ] and [ hlevel ] . it becomes necessary to estimate the discrete green s function on the ball directly , instead of building on known estimates for the @xmath42 green s function . the argument requires a few other modifications as well ( for example , taking @xmath55 to be a point inside the ball , instead of on the boundary ) . the bound of order @xmath20 in theorem [ thm : logfluctuations ] measures the `` fluctuations in the worst direction . '' another interesting quantity is the fluctuation in a specific direction , say along positive @xmath597-axis , as measured by the differences @xmath598 and @xmath599 based on the analogy with the discrete gaussian free field , we believe that these random variables have standard deviation of order @xmath44 in two dimensions , and that their higher dimensional analogues have @xmath600 standard deviation . these orders match those in the top line of table [ table : gfforder ] . in numerical experiments on a related quantity , meakin and deutch @xcite found order @xmath44 fluctuations in dimension @xmath19 and @xmath600 fluctuations in dimension @xmath601 . we have not been able to adapt our bootstrapping argument to prove bounds of this type on fluctuations in a fixed direction . an @xmath75-early point where @xmath602 contributes only @xmath44 to the martingale @xmath317 , which is swamped by the logarithmic error term in the discrete mean value property , lemma [ hlevel](c ) . james propp around 2001 proposed a growth process called _ rotor - router aggregation _ , in which particles started at the origin in @xmath13 perform deterministic analogues of random walks until reaching an unoccupied site . large - scale simulations show that the aggregates produced in this way are substantially smoother than idla , with even smaller fluctuations from circularity @xcite . indeed , part of propp s motivation was to study how much of the random fluctuation can be removed by passing to the deterministic analogue . the data indicate that the deviation from circularity in rotor - router aggregation may even be bounded independent of the size of the cluster ; that is , @xmath603 for an absolute constant @xmath5 . yet the current best upper bounds ( order @xmath20 on the inside and @xmath604 on the outside , @xcite ) remain distant from this goal . in trying to adapt our methods to rotor - router aggregation , we can define a deterministic analogue of the martingale @xmath279 by summing the function @xmath56 over the rotor - router aggregate @xmath29 . at this point an interesting question arises : what should take the place of the martingale representation theorem for this deterministic process ? external dla , defined by witten and sander @xcite , produces intricate branching structures instead of a smooth spherical surface . typically it is defined with particles starting from infinity , but there is also a version in which particles start at the origin . in this version , the initial occupied set @xmath605 is the set of sites outside some extremely large origin - centered ball . at time @xmath0 we start a random walk at the origin and stop when it first reaches a site @xmath606 _ adjacent to _ @xmath2 . then we set @xmath607 . both this paper and the sequel ( about the gaussian free field ) use a similar technique : namely , they both use geometric arguments to control the quadratic variation of continuous martingales obtained by summing the values of discrete harmonic functions over sets of particle locations . one intriguing aspect of this work is that all of the martingales we utilize are still martingales for external dla ( grown from the origin ) . indeed , kesten s upper bound on the growth rate of external dla @xcite uses a closely related martingale . because the geometry of external dla is much more intricate , however , it becomes more difficult to control the quadratic variation of the martingales . in our argument , the interlinked inner and outer estimates indeed , even the notions of early and late points rely on the fact that the cluster fluctuates around a predictable asymptotic shape ( the disk ) . in external dla , on the other hand , even the large - scale geometry is random . this appendix is devoted to proving the following lemma , which strengthens lemma [ lem : thickening ] . since we anticipate using this result to bound the fluctuations of internal dla in higher dimensions as well , we treat all dimensions @xmath608 . for each dimension @xmath43 , there are positive absolute constants @xmath78 , @xmath79 , and @xmath594 ( depending only on @xmath611 ) such that for all real numbers @xmath81 and all @xmath74 with @xmath609 , @xmath612 note that even the thinnest possible tentacle , a path of length @xmath75 , can be formed using only @xmath613 random walk steps : for @xmath614 , the @xmath615-th particle to enter @xmath50 takes @xmath615 steps along the path and occupies the next site on the path . this shows that the exponent @xmath85 on the right side of is best possible . instead of working with the ball @xmath50 it will be slightly more convenient to use the box @xmath616^d \subset \mathbb z^d$ ] . since the volume of a ball is ( up to a @xmath611 dependent constant factor ) the same as that of a box of the same radius , lemma a holds as stated if and only if it holds with @xmath50 replaced by @xmath617 . fix @xmath618 , and for @xmath619 let @xmath620 be the set of points @xmath621 such that @xmath622 the sets @xmath620 partition @xmath617 into @xmath623 concentric shells , indexed from the outside inward . since @xmath624 is disconnected , at least one site in shell @xmath620 must join the cluster @xmath2 before any site in @xmath625 can join . it is enough to show that and hold if we replace @xmath2 with the set @xmath626 , where @xmath627 , and @xmath628 is defined to be the smallest @xmath0 for which @xmath48 . now write @xmath629 for @xmath630 . the cluster @xmath631 intersects each shell @xmath620 , so @xmath632 for all @xmath633 . given @xmath75 , the number of possibilities for the sequence @xmath634 is at most @xmath635 , which is much smaller than the reciprocal of the right side of and . it thus suffices enough to show that , for each _ fixed _ sequence @xmath636 with @xmath632 and @xmath637 , and hold when the left side is replaced by the probability of the event @xmath638 we emphasize that @xmath79 and @xmath594 are still absolute constants ; in particular , they do not depend on the sequence @xmath639 or on @xmath75 . we will henceforth assume that such a sequence @xmath639 has been fixed , and bound the probability @xmath640 . on the left : a possible value for @xmath626 ( black , with @xmath54 in red at the center ) . here @xmath641 and the sequence @xmath642 ( giving number of black squares in the square shell @xmath620 , indexed from outside inward ) is given by @xmath643 . taking @xmath644 , the corresponding sequence @xmath645 is given by @xmath646 , and a corresponding tower of squares is illustrated on the right . the sizes of the squares are @xmath647 , @xmath648 , @xmath649 , @xmath650 . the grey shells indicate the @xmath651 , where @xmath652 , @xmath653 , @xmath654 , and @xmath655 . since @xmath644 , the number of black squares in the `` annulus '' formed by the union of the @xmath620 crossing a given square on the right ( other than the innermost square ) is always at least twice the number black of squares in the corresponding annulus on the left . , width=499 ] 1 . @xmath656 is roughly the kind of sequence @xmath634 that one would see if @xmath626 were a tower of cubes . more precisely , let @xmath658 be positive integers summing to @xmath75 , representing the side lengths of the cubes . for @xmath659 , let @xmath660 . for @xmath661 , let @xmath662 be the unique integer @xmath663 such that @xmath664 . then set @xmath665 note that @xmath666 . 2 . there is a small constant @xmath667 such that for each @xmath668 ( but possibly not for @xmath669 ) @xmath670 once we have fixed @xmath667 , we can construct the sequence @xmath656 from @xmath634 in a deterministic way by induction : given the cube side lengths @xmath671 , take @xmath672 to be the smallest possible such that the second condition holds . next we make explicit our choice of @xmath667 . for @xmath673 , write @xmath674 for the probability measure associated to simple random walk @xmath675 in @xmath42 with @xmath676 . by the local central limit theorem @xcite , if @xmath354 is even , then for each site @xmath677 of even parity in the ball @xmath678 , we have @xmath679 for a constant @xmath680 depending only on @xmath611 . let @xmath681 , where @xmath682 is the volume of the unit ball in @xmath683 . by ( 2.1.2 ) , there is a constant @xmath684 depending only on @xmath611 , so that @xmath685 now take @xmath686 in the above construction of the sequence @xmath656 from @xmath634 . this defines a sequence of shells @xmath651 as in figure [ shells ] . according to the next lemma , if @xmath687 is any subset of @xmath42 satisfying @xmath688 for @xmath689 , then simple random walk started from any point in the @xmath663-th shell @xmath651 has at least a constant @xmath690 probability of exiting @xmath687 before reaching @xmath691 . [ lem : trialsuccessrate ] let @xmath687 be any set for which @xmath692 for each @xmath689 . fix @xmath693 and @xmath694 . for simple random walk in @xmath42 started at @xmath597 , let @xmath414 be the first hitting time @xmath691 , and let @xmath695 be the first hitting time of @xmath696 . then @xmath697 let @xmath698 , and let @xmath354 be the greatest even integer @xmath699 . let @xmath700 . by the local central limit theorem , for each site @xmath677 of even parity in the ball @xmath701 , we have @xmath702 suppose first that @xmath703 . let @xmath704 be the set of points of even parity in the intersection @xmath705 note that @xmath706 . hence by our choice of @xmath667 , @xmath707 hence @xmath708 by , we have @xmath709 we conclude that @xmath710 during the formation of the idla cluster @xmath626 , each time a particle hits a shell @xmath651 ( @xmath713 ) for the first time , we begin a `` trial '' which ends either when the particle stops walking ( due to exiting the cluster and occupying a previously unoccupied site ) or when it next hits @xmath691 . in the former case , we call the trial a `` failure . '' otherwise , we call it a `` success . '' let @xmath714 be the event that the @xmath715-th trial fails , and let @xmath716 be the @xmath717-field generated by all random walks generating the idla cluster up until the start of the @xmath715-th trial . let @xmath718 be the occupied cluster at the time when the @xmath715-th trial starts , and let @xmath719 . by lemma [ lem : trialsuccessrate ] , we have @xmath720 in other words , on the event @xmath721 , once the @xmath715-th trial begins ( and we condition on the past of the process in particular on all previous trials ) the trial fails with probability at least probability @xmath722 . since each failure adds a new site to @xmath626 , there are at most @xmath723 failures . next we estimate the total number of trials . each particle that makes it to @xmath651 has to first pass through all the @xmath724 for @xmath725 . thus the process of successful growing a set @xmath626 , such that @xmath726 for all @xmath615 , necessarily includes at least @xmath727 trials . hence on the event @xmath728 , there are at least @xmath414 trials , at most @xmath729 of which fail ; and conditional on all past trials , each trial fails with probability at least @xmath722 . before we put an explicit bound on @xmath414 , we d like to find a way to reduce the problem to the case that @xmath730 ( i.e. , failure is required to occur at less than half the expected rate ) . in other words , we d like to reduce the case that @xmath731 is larger ( by a fixed constant factor @xmath732 ) than @xmath729 . this will certainly be the case , for example , if more than a fraction @xmath291 of the sum @xmath729 comes from values of @xmath615 for which @xmath733 . let @xmath734 be a fixed integer greater than @xmath735 . let @xmath196 be the smallest positive integer such that @xmath736 then consider the smaller box we get by replacing @xmath75 by @xmath737 . we can take @xmath78 small enough so that @xmath738 is at most an arbitrarily small constant times @xmath739 , and hence @xmath740 is at most some arbitrarily small constant times @xmath75 . indeed , since @xmath734 is just a constant , we can take @xmath78 small enough so that @xmath741 is at most an arbitrarily small constant times @xmath75 . since the sequence @xmath742 decreases by a factor of at least @xmath743 each time @xmath663 increases by @xmath6 ( up until @xmath744 ) , we also obtain that the sequence @xmath745 is also bounded above by a sequence that decreases exponentially in @xmath663 , and hence ( by making @xmath78 small enough ) we can arrange so that @xmath746 is less than @xmath747 . then if we replace @xmath75 with @xmath737 ( and replace @xmath626 by its intersection with the corresponding smaller box ) , we reduce the problem to one for which holds and @xmath75 has decreased by at most a factor of @xmath19 . now assuming , the usual large deviations bound ( cramer s theorem ) implies that the probability of seeing @xmath414 trials with only @xmath748 or fewer failures is exponentially unlikely in @xmath414 . note that @xmath749 is up to a multiplicative constant ( between @xmath750 and @xmath667 ) the same as @xmath751 . hence the proof of lemma [ lem : strongnotentacle ] is completed by the following lower bound on @xmath752 . the first two conditions imply that @xmath761 . by holder s inequality , @xmath762 where @xmath763 is such that @xmath764 . hence @xmath765 if @xmath766 , then @xmath767 and the right side is at least a constant times @xmath768 . in the case @xmath86 , we have @xmath769 and the right side is at least a constant times @xmath770 . in this section we prove lemma [ lem : aprioriestimates ] . lawler , bramson and griffeath give an exponential bound on the probability of @xmath93-late points in the case that @xmath93 is comparable to the radius of the cluster . specifically , they show ( see lemma 6 of @xcite and the displayed equation immediately following it ) that for any @xmath771 there is a constant @xmath772 depending on @xmath773 such that for all @xmath774 @xmath775 fix @xmath776 . taking @xmath777 and @xmath778 in , we have @xmath779 and @xmath780 , so for all @xmath781 , @xmath782 where @xmath594 is a constant depending on @xmath773 . c. moore and j. machta , internal diffusion - limited aggregation : parallel algorithms and complexity , j. stat . phys.*99*(3 - 4 ) : 661690 , 2000 .

let each of @xmath0 particles starting at the origin in @xmath1 perform simple random walk until reaching a site with no other particles . lawler , bramson , and griffeath proved that the resulting random set @xmath2 of @xmath0 occupied sites is ( with high probability ) close to a disk @xmath3 of radius @xmath4 . we show that the discrepancy between @xmath2 and the disk is at most logarithmic in the radius : i.e. , there is an absolute constant @xmath5 such that with probability @xmath6 , @xmath7

introduction logarithmic fluctuations concluding remarks appendix: proof of lemmas[lem:thickening] and[lem:aprioriestimates]

typical simulations of lattice field theories have grave difficulties determining features of the theories continuum limits . one needs accurate simulation data within the theory s scaling region in order to measure critical indices with good control . finite size effects must be monitored and either kept small so that infinite volume results can be read off finite volume data , or a series of lattice volumes must be simulated so that finite size scaling can be used to determine scaling laws . carrying out such a program for fermion field theories is particularly daunting . fermion algorithms typically require more computer time than purely bosonic models and dealing with almost massless fermions , as must be done in the search for continuum limits of lattice models , is also very difficult . however , it was noticed long ago that simulation studies of nambu@xmath3jona - lasinio models tend to be much more quantitative than those of other fermion field theories @xcite . in particular , the logarithmic triviality of these models has been demonstrated , although determining logarithmic singularities decorating mean field scaling laws is a numerical challenge . the reason for this success lies in the fact that when one formulates these four - fermi models in a fashion suitable for simulations , one introduces an auxiliary scalar field @xmath4 in order to write the fermion terms of the action as a quadratic form . in this formulation @xmath5 , to be denoted @xmath4 in much that follows , then acts as a chiral order parameter which receives a vacuum expectation value , proportional to the chiral condensate @xmath6 , in the chirally broken phase . most importantly , the auxiliary scalar field @xmath7 acts as a site dependent dynamical mass term in the fermion propagator . in the chiral symmetry broken phase , the dirac operator is now not singular for quarks with vanishing bare mass and its inversion @xcite , @xcite is accurate and very fast . the algorithm for nambu@xmath3jona - lasinio models is smart it incorporates a potential feature of the solution of the field theory , chiral symmetry breaking and a dynamical fermion mass , into the field configuration generator . the good features of the simulation algorithm for the nambu@xmath3jona - lasinio model can be generalized to lattice qcd @xcite and noncompact qed @xcite by incorporating a weak four - fermi term into their actions . these generalized models now depend on two couplings , the familiar gauge coupling and a new four - fermi coupling . by choosing the four - fermi coupling small we can be confident that all the dynamics resides in the gauge and fermi fields and the four - fermi term just provides the framework for an improved algorithm which allows us to simulate the chiral limit of massless quarks directly . in the context of qcd , this approach is being used to determine the universality class of the finite temperature transition of the isodoublet version of the model @xcite . the method also has the promise of producing light quark mass spectrum calculations of qcd , a physical situation which is not practical in other approaches . in the context of qed , this approach has been used to show the logarithmic triviality of textbook qed @xcite . this is the noncompact version of the model , one free of monopoles and other topological field configurations that a lattice model can support @xcite . our interest here lies in the compact version of the model , its magnetic monopoles and how they might effect the model s continuum limit . the compact version of qed , periodic qed , with dynamical fermions has not been studied as well as the noncompact version of the model . we will be applying the standard algorithm , the hybrid molecular dynamics algorithm ( hmd ) @xcite , in the version in which the chiral limit of massless quarks is accessible . this is the @xmath0qed formulation as reviewed above . this formulation is particularly relevant because we are interested in a continuum theory with light physical fermions . the compact version of the four flavor qed has interactions coming from photons and magnetic monopoles coded into the periodic gauge fields . by adding a four - fermi interaction into this model which preserves a piece of the chiral symmetry of the massless model , we can study light fermions directly on the lattice and see how fermion charge screening coming from light fermion loops effects the dynamics . the four - fermi interactions also gives us the opportunity to separate the chiral transition of the model from its confinement / deconfinement transition which is controlled by monopole condensation . these issues will be discussed more fully in the context of this paper . it is crucial in all of this that fermion screening be accounted for accurately and realistically . it is inappropriate to make any simplifications here because the character of fermion screening is not understood outside perturbation theory . some unbiased simulation studies are called for to see if , for example , the anti - screening due to magnetic monopoles can balance the screening due to fermion loops and lead to ultra - violet stable fixed points @xcite . this need further motivates us to consider the @xmath0qed version of the model because then we are guaranteed that the fermions remain light even on the strongly cutoff lattice , accounting quantitatively for fermion charge screening . in this formulation of the model , there are light fermions even on the cutoff system , so we see the effects of fermion screening throughout the entire calculation and not just in the subtle continuum limit . our lattice model has two couplings : the conventional gauge coupling of compact qed and the four - fermi coupling of the nambu jona - lasinio term . the uncoupled versions of periodic qed and the nambu jona - lasinio models have been studied by lattice gauge theory methods and considerable quantitative information is available to guide this study . we shall find two lines of transitions in the two dimensional coupling constant parameter space of the compact u(1)-gauged nambu@xmath3jona - lasinio model . one line is associated with monopole condensation and the other with chiral symmetry breaking . long runs on fairly large lattices , @xmath2 , indicate that the monopole condensation transition is first order . a finite size scaling analysis of this transitions is necessary to state this result with absolute certainty and such a study must await more computer resources . it is well known that monopole condensation studies in compact lattice qed experience severe correlations in monte carlo time and very long runs are needed . in fact , many past simulations have been indecisive in establishing the order of the transition and some studies have produced data that has been fit with power behaved equations of state , characteristic of a second order transition . we will show that our data can not be fit sensibly with such a hypothesis and , in fact , a first order transition where the monopole concentration jumps discontinuously to zero is strongly favored numerically . the line of chiral transitions is distinct from the line of monopole transitions as long as the four - fermi term is nonzero . one point along the line of chiral transitions is studied in detail to determine the character of the chiral transition . this work parallels recent studies of the noncompact version of the model which concluded that the chiral transition was logarithmically trivial @xcite . conventional wisdom would suggest the same result here , since the monopole concentration vanishes on the line of chiral transitions . however , this result is far from clear in our numerical work . the best fits to the chiral condensate and its associated susceptibility are compatible with power laws of a second order transition , but the critical indices are far from the expected mean field values . unfortunately , the immediate vicinity of the critical point can not be well studied on this size lattice , @xmath2 , because of finite size effects , so the simulations reported here might not be registering the real continuum scaling laws of the theory . simulations on larger lattices closer to the critical point are required and studies on @xmath8 and @xmath9 lattices are planned . this paper is organized as follows . in the next section we present the formulation of the lattice action and discuss its symmetries and general features . in the third section we sketch the phase diagram and in the fourth section , the heart of the paper , we examine several points in the phase diagram in detail and come to preliminary quantitative conclusions about the first order character of the monopole condensation transition and the second order character of the chiral transition . in the fifth section we comment on the puzzling nature of the results , present a physical picture of a nontrivial chiral transition in which anti - screening due to entwined loops of magnetic monopoles and anti - monopoles balance screening due to fermion anti - fermion pairs and suggest further research . to begin , consider the abelian - gauged nambu jona - lasinio model with four species of fermions . the lagrangian for the continuum gauged nambu@xmath3jona - lasinio model is , @xmath10 the lagrangian has an electromagnetic interaction with continuous chiral invariance ( @xmath11 , where @xmath12 is the appropriate flavor matrix ) and a four - fermi interaction with discrete ( @xmath13 ) chiral invariance ( @xmath14 ) . the mass term @xmath15 breaks the chiral symmetries and will be set to zero in much of the work that folows . the pure nambu@xmath3jona - lasinio model has been solved at large @xmath16 by gap equation methods @xcite , and an accurate simulation study of it has been presented @xcite . the discrete ( @xmath13 ) chiral invariant action produces a particularly efficient algorithm . full chiral symmetry should be restored naturally in the continuum limit in those regions of the paramter space where the four - fermi term proves to be irrelevant . the action with just the @xmath13 chiral symmetry is preferable over models with continuous chiral symmetry because they are not as easily and efficiently simulated due to massless modes in the strongly cutoff theory . it is useful to introduce an auxiliary random field @xmath4 by adding @xmath17 to the lagrangian . this makes the lagrangian a quadratic form in the fermion field so it can be analyzed and simulated by conventional methods . the model is then discretized by using staggered fermions . the lattice action reads , in the case where the gauge symmetry is interpreted as a compact local @xmath18 symmetry , following wilson s original proposal @xcite : @xmath19 where @xmath20 where @xmath4 is an auxiliary scalar field defined on the sites of the dual lattice @xmath21 @xcite , and the symbol @xmath22 denotes the set of the 16 lattice sites surrounding the direct site @xmath23 . the factors @xmath24 are the gauge connections and @xmath25 are the staggered phases , the lattice analogs of the dirac matrices . @xmath26 is a staggered fermion field and @xmath27 is the bare fermion mass , which will be set to 0 . note that the lattice expression for @xmath28 is the circulation of the lattice field @xmath29 around a closed plaquette , the gauge field couples to the fermion field through compact phase factors to guarantee local gauge invariance and @xmath30 enters the action to make it periodic . it will often prove convenient to parametrize results with the inverse of the four - fermi coupling , @xmath31 , and the inverse of the square of the gauge coupling , @xmath32 . the global discrete chiral symmetry of the action reads : @xmath33 where @xmath34 is the lattice representation of @xmath35 . as we mentioned , interesting limiting cases of the above action are the pure @xmath13 nambu@xmath3jona - lasinio model ( @xmath36 ) , which has a phase transition at @xmath37 @xcite and the pure lattice pqed ( @xmath38 ) limit , whose first order chiral phase transition is coincident with its first order monopole condensation transition near @xmath39 for four flavors @xcite . the lattice simulation code is very similar to others in this program . the systematic step size errors , those varying as @xmath40 , the discretization of the molecular dynamics evolution equations in hmd time @xmath41 , have been studied in the past and are understood @xcite . taking @xmath42 produced chiral condensates whose systematic errors were considerably smaller than their statistical errors . other algorithmic problems , such as finite size effects , tunnelling , long correlation times were monitored carefully in the runs and will be discussed below when appropriate . we scanned the 2 dimensional parameter space ( @xmath43,@xmath44 ) using the hybrid molecular dynamics algorithm tuned for four continuum fermion species @xcite and measured the chiral condensate , its susceptibility and the monopole concentration and its susceptibility as a function of @xmath43 and @xmath44 on a @xmath1 lattice . these observables have been discussed extensively in literature and we refer the reader to @xcite for background . for very small @xmath44 the chiral transition and the monopole condensation transition were coincident to the accuracy and resolution of our survey . abrupt jumps in the order parameters were measured and were interpreted as signalling first order transitions . this is shown in fig . [ fig : pqedz2 ] as the thick line of transitions extending to the @xmath43 axis . earlier studies of the @xmath38 version of this lattice model @xcite presented evidence for first order coincident chiral and monopole condensation transitions on the axis in good agreement with this simulation . as we follow the line of first order transitions into the phase diagram fig.[fig : pqedz2 ] , there comes a point near @xmath45 and @xmath46 where the line of chiral transitions and the line of monopole condensation transitions become clearly distinct . we will show measurements on the vertical line @xmath47 on a @xmath2 lattice which will make this point very clear . in particular , setting @xmath47 and then increasing @xmath43 in small steps from the strong coupling region of @xmath48 , we shall see good evidence for an abrupt transition in the monopole concentration @xmath49 at @xmath50 . @xmath49 is large , approximately @xmath51 at @xmath52 and @xmath53 , while it is small , approximately @xmath54 at @xmath55 and @xmath53 . throughout this region of the phase diagram , the chiral condensate @xmath4 is distinctly nonzero , so the system resides in the chirally broken phase even though @xmath49 vanishes . the chiral order parameter @xmath4 does , however , experience a jump from @xmath56 at @xmath52 and @xmath53 to @xmath57 at @xmath55 and @xmath53 , indicating that free monopoles do contribute to chiral symmetry breaking , as expected , but the gauge coupling is sufficiently large in this case that chiral symmetry breaking occurs even without the active participation of free monopoles . in fact as we proceed up the vertical line @xmath47 , @xmath4 falls only slowly and it is not until @xmath58 that it vanishes . the vanishing appears to be continuous , so the upper line of transitions , the dashed line in the phase diagram fig.[fig : pqedz2 ] , appears to be second order . there is much more to discuss about the chiral transition and that will occur below . however , it is eminently clear that the chiral transition occurs at a smaller gauge coupling @xmath59 at fixed @xmath47 , and the two transitions , whatever their orders , are separate and distinct . the dashed line of chiral symmetry breaking transitions continues to larger @xmath44 and higher @xmath43 . the four - fermi coupling alone breaks chiral symmetry at strong coupling , @xmath60 , and the dashed line in fig.[fig : pqedz2 ] connects to that well - studied case @xcite . those studies of the pure four - fermi model were done with care and the logarithmic triviality of the nambu jona - lasinio model was confirmed by the numerical lattice simulation @xcite . the first order line of monopole condensation transitions extends across the phase diagram as shown in fig . [ fig : pqedz2 ] . in fact , we will study it in detail by simulating the two - coupling model on the horizontal line @xmath61 and variable @xmath44 and find evidence for a first order monopole condensation line at @xmath62 where @xmath49 jumps discontinuously to zero . we will see that @xmath4 is large and stable in the vicinity of this monopole condensation point , on both sides of it . therefore , the fermions have an effective mass comparable to the cutoff energy and should decouple from continuum physics , if there is any , in the model . in any case , the model should not be significantly different from the pure compact periodic lattice qed model and its monopole condensation should be essentially the same as in the pure gauge field theory . the order of the monopole condensation transition has always been controversial in this case because of extraordinarily long correlation times . our simulations show such effects , but our runs were long enough , we believe , that our configurations have achieved thermalization at each coupling and artificial rounding of the transition due to inadequate statistics has been avoided . we support the view that the monopole condensation transition , which also signals the confinement - deconfinement transition in the pure gauge theory case , is first order . recent simulations of the pure peridic qed model may be converging @xcite . consider the observables , the chiral condensate and its susceptibility and the monopole concentration and its susceptibility , as recorded in table i. in this case the strength of the four - fermi interaction is set to a constant @xmath63 and the gauge coupling decreases as the simulation moves up the phase diagram fig . [ fig : pqedz2 ] . at strong gauge coupling , chiral symmetry is spontaneously broken and and the monopole condensate is large and near its saturation value of unity . both the chiral condensate and the monopole concentrations are hardly fluctuating at @xmath43 in the vicinity of @xmath64 since their respective susceptibilities are relatively tiny there . however , there is a clear jump in the monopole concentration to zero at @xmath65 and the chiral condensate experiences a discontinuity here also , but it does not vanish . the chiral condensate will be discussed below . here we concentrate on the monopole physics first . conventional wisdom would state that there is a confinement / deconfinement phase transition along the @xmath66 line at @xmath65 . on the strong coupling side of this point there is a monopole condensate which causes confinement in the sense that all the physical states are neutral and the electric charge is a good quantum number . in such a phase one expects and finds chiral symmetry breaking . the character of the phase transition , whether it is first or second order , is crucial . in fig . [ fig : pqedm ] we show the monopole concentration as a function of @xmath43 and suggest that there is a discontinuity in the graph . 4.0 in 3.0 in attempts to fit the concentration data with a power law that would be indicative of a steep , but continuous transition failed badly . for example , a power law fit of the standard variety , @xmath67 , to the points between @xmath68 and @xmath69 , had a @xmath70 per degree of freedom in excess of @xmath71 and an exponent @xmath72 . the fit strongly suggests that a step function discontinuity is preferred as shown in the figure . why have other studies of this transition in the pure gauge theory produced power law fits ? since the effective mass of the fermion is comparable to the cutoff in this part of the phase diagram , there should be no significant difference between the model simulated here and the simpler pure compact qed model . the pure gauge part of the action is the same in both cases and the effect of the heavy fermions should be accommodated by a tiny shift in the coupling constants . we noticed in our runs , however , a point discussed extensively in the literature : extremely long relaxation times are needed to find the equilibrium state of the pure compact model . if we had stopped our simulations with half the statistics we ended with , there would have been some artificial rounding of the transition shown in the figure and a power law fit may have been possible and misleading . tens of thousands of hmd time units are necessary to relax the system in the vicinity of the phase transition . various authors have suggested that since the model is expressible as the sum over closed loops of conserved monopole currents @xcite , local monte carlo algorithms are particularly ill equipped to relax a system with such long range topological structures constrained by global and local conservation laws . additional qualitative evidence for the first order character of the transition comes from the monopole susceptibility . we see in the table that @xmath73 is highly spiked at the transition , jumping from @xmath74 at @xmath52 to @xmath75 at @xmath69 and then falling to @xmath76 on the weak coupling side of the transition at @xmath55 . this suggests that the real susceptibility is a delta function in the large volume thermodynamic limit . to really establish this claim , one would need a finite size study to monitor the size of the transition region and the height of the susceptibility curves as a function of volume . now consider the chiral transition along the @xmath63 line . this is the major focus of this paper . the raw data is shown in table i. and is plotted in fig . [ fig : pqedpbp ] 5.0 in 4.0 in at strong coupling chiral symmetry is strongly broken and the chiral condensate is large and does not fluctuate severely . as in past studies , it proves convenient to monitor chiral symmetry breaking through the vacuum expectation value of the auxiliary field @xmath4 @xcite . the quark@xmath3anti - quark bilinear was also calculated in the simulation and is propotional to @xmath4 , in accordance with the theory s equations of motion @xcite . the first interesting feature of this figure is the jump of the chiral condensate at @xmath77 , at the same point where the monopole concentration fell to zero . the fact that fig . [ fig : pqedpbp ] has a jump at @xmath77 is further support for the first order character of the deconfinement transition there . it is also interesting that the chiral condensate does not vanish on the weak coupling side of the deconfinement transition . apparently , the strong interactions in the theory are sufficient to cause chiral symmetry breaking without confinement . we expect that the theory has a rich spectrum of bound states on the weak coupling side of the transition at @xmath77 and @xmath63 , in accord with the physical picture of chiral symmetry breaking in ref . it might be informative to do spectrum calculations in this lattice theory in light of this result . it is very significant and nontrivial that the chiral condensate does not vanish on the weak coupling side of the deconfinement transition at @xmath77 and @xmath63 . chiral symmetry breaking observed here is not due to the four - fermi interaction alone . we know from ref . @xcite that a much stronger four - fermi interaction , @xmath78 , is needed to break chiral symmetry in the absence of gauge interactions . in addition , simulations of the pure gauge theory with four species of staggered fermions @xcite produced chiral condensates on the weak coupling side of the transition which were consistent with zero . physical mechanisms that could be causing substantial chiral symmetry breaking in this region of the phase diagram will be discussed in the last section of this paper . returning to fig.[fig : pqedpbp ] , we see that the chiral condensate falls essentially linearly in the gauge coupling @xmath79 until a chiral symmetry restoration transition is found in the vicinity of @xmath80 and @xmath63 . as recorded in table i. , the monte carlo statistics in this region of the phase diagram were quite considerable on this @xmath2 lattice . nonetheless , as the critical point was approached , there were the usual troubles with with critical slowing down and tunnelling that hamper the predictive power of these results . critical slowing down causes the rising statistical error bars recorded in the table as @xmath43 approaches the transition . tunnelling between @xmath13 vacua also became significant for @xmath81 and limited our ability to simulate too close to the critical point . notice that the four - fermi term in the extended action is chosen to have a @xmath13 chiral symmetry which is spontaneously broken when @xmath4 develops a vacuum expectation value . this symmetry was chosen for the four - fermi term for several reasons . first , it is particularly easy to simulate and extract observables . in particular , as long as the theory resides in a particular chiral vacuum , @xmath4 and @xmath82 can be measured directly , without the need for inserting a small bare fermion mass into the theory to pick out a unique vacuum state . this is a great advantage because a bare fermion mass breaks chiral symmetry explicitly and leads to rounding of the transition which makes quantitative studies of the transition and its universality class very difficult . in addition , it is not necessary to measure observables like the absolute value of the order parameter rather than the order parameter itself in order to simulate in a unique , specified vacuum state . this is in contrast to using a @xmath83 four - fermi term as is done in qcd applications of this approach @xcite . in that case the four - fermi term has a continuous chiral symmetry , so even in the broken phase the order parameter @xmath4 will average to zero as the phase of the vacuum state rotates over the unit circle . in the qcd case , we use the magnitude of two auxiliary fields , @xmath84 , as an effective order parameter to deal with this problem . since fluctuations make the effective order parameter non - zero everywhere , rounding of the transition is encountered . in the qcd case , it is important to use the @xmath83 four - fermi term instead of the @xmath13 term used here , because one needs a massless pion in the system to have the correct light degrees of freedom characteristic of the low energy spectroscopy of continuum qcd . by contrast , the emphasis in qed is on the short distance behavior of the effective charge and other couplings , and the low energy spectroscopy of the model is presumably irrelevant to the issue of the existence of the theory defined at the ultra - violet stable fixed point . the tunnelling between the two @xmath13 vacua , @xmath85 , limits our approach to the critical point and appears to be the most damaging finite size effect we must deal with . simulations are planned on larger lattices to lessen it . of course , the simulations of this study on a @xmath2 lattice with high statistics represent a serious first step in this program . examining table i. , we note that the error bars in the @xmath4 and @xmath0 data are particularly large for @xmath43 between @xmath86 and @xmath87 , reflecting the occasional , every thousand or so time intervals , tunnelling between @xmath13 vacua . the data for these points will not be used in the fits quoted here . unfortunately , omitting these data means that this simulation may not be sensitive to the real critical behavior of the model . for example , the span of couplings from @xmath88 to @xmath89 used in the fits here might be outside the real critical region of the theory , which might extend only between @xmath86 and @xmath87 . it might be that we need data in this region to confirm that this model has the expected logarithmically trivial scaling laws , like the non - compact model with four fermi terms @xcite . this possibility and other issues of conventional wisdom will be reviewed in the section on conclusions below . at this point we will just do what we can do and plot and fit the order parameter and susceptibility data where the finite size effects appear to be under control . in fig.[fig : pqedpbp ] we show the data on the chiral condensate fitted with a simple power law , @xmath90 . the fit takes the data at @xmath43 ranging from @xmath91 through @xmath89 and finds @xmath92 and @xmath80 . the confidence level of the fit is excellent , @xmath93 percent , corresponding to a @xmath70 per degree of freedom of @xmath94 . the central value for the magnetic critical exponent @xmath95 is unchanged by taking wider ranges of couplings and even approaching the apparent critical point at @xmath80 more closely , but the confidence levels deteriorate . we had expected a logarithmically improved mean field fit here with the magnetic critical exponent near @xmath96 , as was found in the non - compact qed case in @xcite , but there is no sign of that behavior . a value of @xmath92 suggests a nontrivial interacting theory , and is very perplexing , as will be discussed in the concluding section below . we also accummulated the fluctuations in the order parameter , the susceptibility @xmath0 , as shown in fig . [ fig : pqedchi ] . the values of @xmath0 in the immediate vicinity of @xmath80 are certainly not reliable because of finite size effects and tunnelling . however , the trend for the susceptibility to grow rapidly from @xmath97 to @xmath89 is clear and finite size effects appear to be under control on the @xmath2 lattice over this limited range of couplings . therefore , we attempted power law fits to the data and in the next figure , fig . [ fig : pqedchiinv ] , we show the reciprocal of the susceptibility , @xmath98 , fitted to the data over the range @xmath99 through @xmath86 , @xmath100 . 4.0 in 3.0 in 5.0 in 4.0 in the fit is quite good , having confidence level of @xmath101 percent , @xmath70/d.o.f . the critical index for the susceptibility is predicted to be @xmath103 and the critical coupling is again found to be @xmath104 . it is interesting that @xmath105 is far from its mean field value of unity , although our reservations about this data are the same as our reservations for the order parameter data and its fit . the possible physical significance of this result will be discussed in the section on conclusions below . since susceptibilities typically are more sensitive to finite size effects than the order parameter , it would be particularly informative to repeat this simulation on a larger lattice and attempt to track the height of the susceptibility peak as a function of lattice size to determine @xmath106 , where @xmath107 is the correlation length exponent , by finite size scaling methods . we also investigated the character of the monopole condensation transition along the horizontal line in the phase diagram fig . [ fig : pqedz2 ] . we set @xmath97 , on the basis of past studies of the pure gauge theory , and scanned in @xmath108 until we found the deconfinement transition . in fact , we started the simulations at strong four - fermi couplings , large @xmath44 or small @xmath109 , as given in table ii . for example , at @xmath110 , the monopole concentration @xmath49 is large , @xmath111 and its associated susceptibility is small @xmath112 . in addition , the chiral order parameter @xmath4 is large , @xmath113 , at this parameter set and its susceptibility is modest , @xmath114 , indicating that the fermions are irrelevant to the long distance dynamics in the model . throughout the entire @xmath108 range , from @xmath115 through @xmath64 presented in the table on the horizontal line , the fermions remain very massive . however , the monopole concentration experiences a deconfinement transition near @xmath116 . as shown in fig . [ fig : mon16b1.0 ] , the transition appears to be a discontinuity , indicating a first order transition . in fact , power law fits to the monopole concentration data near @xmath117 of the form , @xmath118 produced very small indices , @xmath119 and very poor confidence levels , @xmath120d.o.f . this result strongly suggests that this transition is actually first order , with a step discontinuity , in agreement with the results found at weaker four - fermi coupling , @xmath63 and @xmath65 , discussed in the previous section . 4.0 in 3.0 in note from table ii . that huge statistics , more than @xmath122 monte carlo time units , were accummulated near the transition to deal with the slow relaxation of monopole loops using a local , small change algorithm . the fact that both horizontal and vertical scans of the confinement / deconfinement line give first order transitions is a good test of our statistics and algorithm . when we began this study we believed that the monopole transitions would be very abrupt , perhaps first order , and that the chiral transitions would be described by mean field theory , decorated by the logarithms of triviality , as found in the noncompact theory , @xcite . our expectations for the monopole transition held true , although it took orders of magnitude more computer time to come to a decisive conclusion . we believed that the chiral transition would be described by logarithmically improved mean field theory , as found for the noncompact theory , because the monopole concentration would be vanishingly small in the vicinity of the chiral transition , the periodicity of the action would become irrelevant , and the gauge field dynamics would reduce to photon exchanges , just like in the non - compact model . only if we could somehow tune the two couplings to that region of the phase diagram where the first order confinement / deconfinement transition met the continuous chiral transition , did we believe there was any possibility of interesting physics . so we are left with a puzzling result : the chiral transition apppears to have critical indices far from mean field theory . just to illustrate and emphasize this point , assume hyperscaling and replace our numerical measurements of the critical indices with integer predictions @xmath123 and @xmath124 they are consistent with . then the remaining critical indices would be @xmath125 , @xmath126 , @xmath127 and @xmath128 . are such large deviations from mean field behavior possible ? when this model was studied in the limit of vanishing gauge couplings , toward the upper end of the dashed line of chiral transitions shown in the phase diagram , fig . [ fig : pqedz2 ] , the chiral transition was shown , both analytically @xcite , and numerically @xcite , to be described by logarithmically improved mean field theory . the algorithm used here but for vanishing gauge coupling gave results in fine agreement with @xmath129 analyses . even the exponents of the logarithms of triviality were compatible with theoretical expectations , even though logarithms that decorate power law scaling laws are notoriously difficult to pin down . the extreme stability and accuracy of the algorithm that uses the fermion dynamical mass to regulate it and guide it was cited as a reason for this numerical success . if the results of this paper are correct , then we must conclude that the critical indices of the chiral transition vary along the line of chiral transitions in fig . [ fig : pqedz2 ] and the four - fermi interaction is not irrelevant in this unfamiliar ground of strong coupling . conventional wisdom , based on perturbation theory , would say that fermion vacuum polarization would always screen the gauge couplings to zero leaving behind a nambu jona - lasinio model that has no interactions in a relativistic continuum limit . what could the explanation of the simulation results of this study be ? the obvious one is just that these results are wrong , in the sense that they are not indicative of the true continuum limit of the theory . perhaps if we could simulate on larger lattices , closer to the critical point , we would find that the true critical behavior is , indeed , logarithmically improved mean field theory as was found by exactly the same methods for the noncompact gauged nambu jona - lasinio theory @xcite . perhaps the region in coupling in fig.[fig : pqedpbp ] from @xmath130 to @xmath131 is outside the real scaling region and is strongly effected by irrelevant but large nonlinearities in the wilson action for the compact @xmath18 gauge fields . it may be that much larger lattices and much larger correlation lengths are needed to find the true continuum behavior in this model due to unusually large corrections to scaling for this particular action . we believe that these issues should be decided and our only tool that avoids uncontrollable approximations is numerical simulations . this is a pity . even numerical methods are sorely taxed by this problem . nonetheless , simulations on larger lattices , closer to the continuum limit are planned . now that we know the interesting regions of the phase diagram , we can focus in and , hopefully , get to the heart of the matter more efficiently than in this exploratory , but time consuming study . early analytic studies of the gauged nambu jona - lasinio model within a framework which included only ladder feynman diagrams @xcite and which explicitly excluded fermion vacuum polarization predicted a line of nontrivial chiral transitions in the two - coupling phase diagram . it can not be stressed too strongly , however , that this calculation was meant as a model of technicolor interactions and was not a solution of a field theory . it did not even include those effects , fermion loops and vacuum polarization , that are expected to render the theory trivial . however , other approximate approaches to this model which may account for screening to some degree have found a nontrivial line of chiral transitions @xcite . the reliability of this newer approximate approach is doubtful , however , because it predicts , contrary to the simulation results of @xcite , that even the noncompact gauged nambu jona - lasinio model is nontrivial . the work reported here and in reference @xcite indicate that the noncompact model is logarithmically trivial while the compact model may not be , contrary to @xcite . let us end this discussion with some speculations which could guide the next generation of simulations planned for this model . suppose that the preliminary results presented here are basically correct . what sort of physical excitations and interactions could support these results and how could they be discovered in the course of a numerical study ? the middle region of the phase diagram , region ii in fig . 1 , needs clarification . in this region @xmath132 is nonzero while the monopole concentration @xmath49 vanishes . the chiral transition at the point @xmath80 and @xmath63 lies on the boundary between regions ii and iii in which @xmath132 also vanishes . the confinement / deconfinement transition at the point @xmath133 and @xmath63 lies on the boundary between regions ii and i in which @xmath49 becomes nonzero . conventional wisdom suggests that the confinement / deconfinement transition between regions ii and i is a four dimensional generalization of the kosterlitz - thouless transition @xcite which describes the two dimensional planar spin model . in the two dimensional kosterlitz - thouless transition a state of vortex anti - vortex ionize and form a plasma of vortices and anti - vortices . in four dimensional pure compact qed , the transition from region ii to region i should be driven by the ionization of strands of bound monopole anti - monopole loops into a plasma of unbound individual loops of monopoles and loops of anti - monopoles that cause confinement through the formation of electric flux tubes @xcite . for the lattice action used here , this transition appears to be first order and does not lead to an interesting continuum field theory . a visualization of random loops is shown in fig.[fig : mmbarloops ] and it is tantalizing to call this the rna phase of the theory , and a visualization of double helix monopole anti - monopole strands is shown in fig . [ fig : mmbardna ] and it is tantalizing to call this the dna phase of the theory . perhaps these pictures convey why it has proved so difficult to decide the order of the confinement / deconfinement transition in compact qed . one must iterate a local monte carlo alogorithm enough so that line singularities , not the point singularities of more familiar transitions , can bind up into long , closed pairs . this is a very demanding requirement of such an alogorithm , especially in four dimensions . if the region ii of the phase diagram really consists of double helix monopole anti - monopole strands , a dna phase of compact qed , then we have a hint how the short distance properties of this theory can be qualitatively different from noncompact qed . in particular , the monopole anti - monopole pairs , which can exist in the compact model but not the noncompact one , provide a medium which anti - screens electric charge . in classical electrodynamics , such an environment raises the fundamental electric charge of an impurity @xmath59 to @xmath134 , where @xmath135 is the permittivity of the vacuum . for a dilute background of monopole anti - monopole dipoles , each having a mean magnetic dipole moment of @xmath136 , one finds that @xmath137 , where @xmath138 is a positive constant and @xmath139 is the density of the magnetic dipoles . so , if these closed dna strands remain relevant through region ii until the critical point on the boundary to region iii is reached , there is a candidate mechanism in place for the cancellation between the screening provided by the light fermion anti - fermion pairs and the anti - screening provided by the dna strands of monopole anti - monopole pairs . as the critical point at @xmath80 and @xmath63 is approached from region ii , the fermions are becoming lighter ( @xmath4 is decreasing ) and their contribution to screening is increasing , thereby diminishing the renormalized electric charge . finally , the renormalized electric charge would be sufficiently small at the critical point that chiral symmetry would be restored and the theory would enter region iii . so , the critical point at @xmath80 and @xmath63 might be nontrivial because the anti - screening due to the topology of the @xmath18 gauge field action balances the screening due to the fermion anti - fermion pairs of conventional vacuum polarization . other ideas concerning screening and anti - screening in compact gauge theories , such as collapse of the wavefunction and catalysis of symmetry breaking @xcite , ideas inspired by monopole - induced proton decay @xcite , should be considered in this framework again and might be ingredients in a successful quantitative implementation of the monopole anti - monopole rna - dna scenario suggested here . apparently there is still much to learn in this difficult subject of strongly coupled gauge theories . luckily , advances in computer simulation power make many of these issues testable in the next round of investigations . in particular , we plan measurements of the vacuum permittivity @xmath135 , the renormalized electric charge and the renormalized four - fermi coupling , the monopole anti - monopole spatial distribution in the vicinity of an external charge etc . measurements of the chiral condensate will be supplemented with measurements of the eigenvalue spectrum of the dirac operator and the reliability of the @xmath0qed algorithm will be studied in greater detail . perhaps some analytical progress can also be made . the dirac quantization condition , electric - magnetic duality transformations and other ingredients of quantum electro - magneto dynamics @xcite might be considered in this framework of chiral symmetry breaking . jbk was partially supported by nsf under grant nsf - phy01 - 02409 and cgs was supported by the leverhulme trust grant . the simulations were done at npaci and nersc . we especially thank nersc for their longterm , consistent support throughout this long ordeal . s. kim , a. koci@xmath140 and j.b . kogut , nucl . phys . * b429 * , 407 ( 1994 ) . s. duane , a.d . kennedy , b.j . pendleton and d. roweth , phys . b195 * , 216 ( 1987 ) . s. duane and j.b . kogut , phys . lett . * 55 * , 2774 ( 1985 ) . s. gottlieb , w. liu , d. toussaint , r.l . renken and r.l . sugar , phys . rev . * d35*,2531 ( 1987 ) . kogut , j .- f . lagae , and d.k . sinclair , phys . rev . * d58 * , 34004 ( 1998 ) . kogut , and d.k . sinclair , hep - lat/0104011 . phys . rev . * d64 * , 034508 ( 2001 ) . s. kim , j.b . kogut and m .- lombardo , phys . * b55 * , 2774 ( 2001 ) ; phys . rev . * d65 * , 054015 ( 2002 ) . a. koci@xmath140 , j. b. kogut and s. hands . nucl . phys . * b357 * , 467 ( 1991 ) . b. rosenstein , b. warr and s. park , phys . rep . * 205 * , 497 ( 1991 ) . wilson , phys . * d14 * , 2455 ( 1974 ) . y. cohen , s. elitzur and e. rabinovici , nucl . phys . * b220 * , 102 ( 1983 ) . e. dagotto and j.b . kogut , phys . * 59 * , 617 ( 1987 ) . s. hands and r. wensley , phys . lett . * 63 * , 2169 ( 1989 ) . g. arnold , b. bunk , th . lippert and k. schilling , hep - lat/0210010 . r. myerson , t. banks and j.b . b129 * , 493 ( 1977 ) . a. casher , phys . lett . * b83 * , 395 ( 1979 ) . a. koci and j. kogut , nucl . b * 422 * , 593 ( 1994 ) . c.n leung , s.t . love and w.a . bardeen , nucl . phys.*b327 * , 649(1986 ) . v. azcoiti , g. di carlo , a. galante , a.f . grillo , and v. laliena , phys . b416 * , 409 ( 1998 ) . e. dagotto , a. koci@xmath140 and j.b . lett . * 62 * , 1001 ( 1989 ) . v.a.rubakov , nucl . phys . b * 203b * , 311 ( 1982 ) . j. m. kosterlitz and d. j. thouless , j. phys . * c6 * , 118 ( 1973 ) . m. blagojevic and p. senjanovic , phys . rep . * 157 * , 233 ( 1988 ) . cccccc @xmath43 & @xmath4 & @xmath141 & @xmath49 & @xmath73 & @xmath142 + @xmath143 & 0.6556(10 ) & 1.43(20 ) & 0.99067(2 ) & 0.0451(2 ) & 1500 + @xmath144 & 0.6153(10 ) & 1.61(23 ) & 0.96809(5 ) & 0.1787(4 ) & 2550 + @xmath145 & 0.5614(10 ) & 2.05(40 ) & 0.8890(2 ) & 0.809(2 ) & 2500 + @xmath146 & 0.5519(18 ) & 6.11(50 ) & 0.8708(2 ) & 0.991(3 ) & 3350 + @xmath147 & 0.4674(10 ) & 3.21(52 ) & 0.263(1 ) & 27.6(3 ) & 4350 + @xmath148 & 0.4416(15 ) & 6.23(55 ) & 0.0179(1 ) & 16.52(4 ) & 5100 + @xmath149 & 0.4182(18 ) & 2.49(47 ) & 0.0423(6 ) & 16.0(1 ) & 1000 + @xmath150 & 0.4039(21 ) & 3.35(50 ) & 0.0661(5 ) & 18.6(2 ) & 1100 + @xmath151 & 0.3822(10 ) & 4.32(53 ) & 0.0214(1 ) & 9.00(1 ) & 8200 + @xmath152 & 0.3244(15 ) & 4.52(37 ) & 0.0001(1 ) & 6.082(6 ) & 2700 + @xmath153 & 0.2798(10 ) & 7.80(56 ) & 0.0001(1 ) & 5.352(3 ) & 14150 + @xmath154 & 0.2346(15 ) & 11.2(1.0 ) & 0.0001(1 ) & 6.17(2 ) & 6600 + @xmath155 & 0.1856(15 ) & 18.37(1.6 ) & 0.0001(1 ) & 4.751(2 ) & 10900 + @xmath156 & 0.1630(21 ) & 25.58(2.4 ) & 0.0001(1 ) & 4.678(4 ) & 7100 + @xmath157 & 0.1394(18 ) & 23.41(2.1 ) & 0.0001(1 ) & 4.601(3 ) & 8400 + @xmath158 & 0.1281(27 ) & 28.51(2.9 ) & 0.0001(1 ) & 4.568(4 ) & 6000 + @xmath89 & 0.1153(27 ) & 41.67(4.6 ) & 0.0001(1 ) & 4.539(4 ) & 5200 + @xmath159 & 0.0997(24 ) & 37.03(3.7 ) & 0.0001(1 ) & 4.511(4 ) & 4600 + @xmath160 & 0.0897(25 ) & 40.0(4.3 ) & 0.0001(1 ) & 4.501(5 ) & 5400 + @xmath161 & 0.092(5 ) & 50.0(6.5 ) & 0.0001(1 ) & 4.487(4 ) & 5300 + @xmath162 & 0.042(5 ) & 35.7(4.1 ) & 0.0001(1 ) & 4.413(6 ) & 4000 + @xmath163 & -0.021(5 ) & 35.6(3.7 ) & 0.0001(1 ) & 4.356(6 ) & 5000 + @xmath164 & -0.024(7 ) & 29.5(5.1 ) & 0.0001(1 ) & 4.26(1 ) & 2200 + @xmath165 & -0.005(5 ) & 33.6(4.2 ) & 0.0001(1 ) & 4.194(9 ) & 5850 + @xmath166 & -0.005(5 ) & 31.5(3.7 ) & 0.0001(1 ) & 4.17(2 ) & 2400 + cddddd @xmath167 & @xmath4 & @xmath141 & @xmath49 & @xmath73 & @xmath142 + @xmath168 & 2.909(2 ) & 5.35(75 ) & 0.8480(3 ) & 1.24(1 ) & 650 + @xmath169 & 2.2450(18 ) & 3.61(50 ) & 0.8341(4 ) & 1.39(1 ) & 550 + @xmath170 & 1.8271(10 ) & 3.75(51 ) & 0.7967(4 ) & 1.857(7 ) & 2200 + @xmath171 & 1.5370(6 ) & 3.35(35 ) & 0.7378(4 ) & 2.762(8 ) & 12200 + @xmath172 & 1.4760(7 ) & 2.76(35 ) & 0.6824(3 ) & 3.75(1 ) & 9650 + @xmath173 & 1.4180(5 ) & 2.64(34 ) & 0.6162(3 ) & 5.19(1 ) & 10200 + @xmath174 & 1.3839(7 ) & 3.15(50 ) & 0.0350(3 ) & 31.6(2 ) & 6650 + @xmath175 & 1.3592(6 ) & 3.17(45 ) & 0.0532(5 ) & 32.8(1 ) & 11200 + @xmath176 & 1.3327(6 ) & 2.75(45 ) & 0.0351(3 ) & 30.3(1 ) & 8700 + @xmath177 & 1.3180(5 ) & 2.83(36 ) & 0.0372(6 ) & 27.5(2 ) & 10600 + @xmath178 & 1.3086(6 ) & 3.58(35 ) & 0.085(1 ) & 22.2(2 ) & 6100 + @xmath179 & 1.1233(8 ) & 2.34(55 ) & 0.0198(2 ) & 18.54(7 ) & 1700 + @xmath143 & 0.9784(8 ) & 3.85(45 ) & 0.0150(2 ) & 15.33(7 ) & 3800 + @xmath144 & 0.8485(7 ) & 3.05(48 ) & 0.0149(2 ) & 13.37(6 ) & 2600 +

compact lattice quantum electrodynamics ( qed ) with four species of fermions is simulated with massless quarks by using the @xmath0qed scheme of adding a four - fermi interaction to the action . simulations directly in the chiral limit of massless quarks are done with high statistics on @xmath1 , and @xmath2 lattices , and the phase diagram , parameterized by the gauge and the four - fermi couplings , is mapped out . the line of monopole condensation transitions is separate from the line of chiral symmetry restoration . the simulation results indicate that the monopole condensation transition is first order while the chiral transition is second order . the challenges in determining the universality class of the chiral transition are discussed . if the scaling region for the chiral transition is sufficiently wide , the @xmath2 simulations predict critical indices far from mean field values . we discuss a speculative scenario in which anti - screening provided by double - helix strands of monopole and anti - monopole loops are the agent that balances the screening of fermion anti - fermion pairs to produce an ultra - violet fixed point in the electric coupling . -1 truecm -1 truecm

introduction formulation phase diagram. small lattice survey. @xmath2 simulations of the vertical @xmath63 line. @xmath2 simulations of the horizontal @xmath97 line. conclusions.

in a _ network creation game _ ( ncg ) , we are given @xmath0 players identified as the nodes of a graph , and each player attempts to connect itself to all the other players . in such a decentralized process , each player aims to selfishly optimize a certain _ routing _ cost towards the other players . thus , its action consists of choosing a suitable subset of players , which are then made adjacent through the activation of the corresponding set of incident edges . unavoidably , activating a link incurs a cost to the player , and so the overall _ building _ cost should be strategically balanced with the aforementioned routing cost . due to their generality , it is in clear evidence that ncgs can model very different practical situations , depending on how all the build - up ingredients are mixed . in the very classic formulation of the game @xcite , each player has no limitations in choosing a subset of adjacent players , its routing cost is a function of the sum of distances to all the other players ( i.e. , the so called _ sum cost _ ) , and activating a link has a fixed cost @xmath14 . not surprisingly , this model was devised by the economists , which were mainly interested in understanding whether the attainment of an equilibrium status ( i.e. , a status in which players are not willing to move from ) for a mutual - relationships social system is compatible with the behavior of the players , which tend to establish selfishly their personal contacts . with the recent advent of the algorithmic game theory , the interest on ncgs has been reawakened . this is especially due to the fact that ncgs are fit to model the decentralized construction of _ communication _ networks , in which the constituting components ( e.g. , routers and links ) are activated and maintained by different owners , as in the internet . according to its performance measurement philosophy , computer scientists put a new special emphasis on the challenge of understanding how the social utility for the ( very large ) system as a whole is affected by the selfish behavior of the players . this trend originated from the paper of fabrikant _ et al . _ @xcite , and was then followed by a sequel of papers , as detailed in the following . [ [ previous - work . ] ] previous work . + + + + + + + + + + + + + + as said before , the canonical form of a ncg , also known as sumncg , is as follows : we are given a set of @xmath0 players , say @xmath15 , where the strategy space of player @xmath16 is the power set @xmath17 . given a combination of strategies @xmath18 , let @xmath19 denote the underlying undirected graph whose node set is @xmath15 , and whose edge set is @xmath20 . then , the _ cost _ incurred by player @xmath21 under @xmath22 is @xmath23 where @xmath24 is the distance between nodes @xmath25 and @xmath21 in @xmath19 . thus , the cost function implements the inherently antagonistic goals of a player , which on one hand attempts to buy as little edges as possible , and on the other hand aims to be as close as possible to the other nodes in the resulting network . these two criteria are suitably balanced in ( [ cost ] ) by making use of the parameter @xmath1 . consequently , the _ _ nash equilibria _ _ ( ne ) space of the game is a function of it . actually , if we characterize such a space in terms of the _ price of anarchy _ ( poa ) , then this has been shown to be constant for all values of @xmath5 except for @xmath26 , for any @xmath27 ( see @xcite ) . a first natural variant of sumncg was introduced in @xcite , where the authors redefined the player cost function as follows @xmath28 this variant , named maxncg , received further attention in @xcite , where the authors improved the poa of the game on the whole range of values of @xmath5 , obtaining in this case that the poa is constant for all values of @xmath5 except for @xmath29 . besides these two basic models , many variations on the theme have been defined . in an effort of defining @xmath5-free models , in @xcite the authors proposed a variant in which a player , when forming the network , has a limited budget to establish links to other players . this way , the player cost function will only describe the goal of the player , namely either the maximum distance or the total distance to other nodes . since in @xcite links and hence the resulting graph are seen as directed , a natural variant of the model was given in @xcite , where the undirected case was considered . afterwards , in @xcite the authors proposed a model complementing the one given in @xcite . more precisely , they assumed that the cost function of each player now only consists of the number of bought edges ( without any budget on them ) , and a player needs to connect to the network by satisfying the additional constraint of staying within a given either maximum or total distance to the rest of players . then , in @xcite the authors proposed a further variant , called basicncg , in which given some existing network , the only improving transformations allowed are _ edge swap _ , i.e. , a player can only modify a _ single _ incident edge , by either replacing it with a new incident edge , or by removing it . recently , this model has been extended to the case in which edges are oriented and players can swap only outleading edges @xcite . notice that , differently from the previous models , the problem of computing a best - response strategy of a player in basicncg is _ not _ -hard . this inspired a subsequent model @xcite in which a player can swap , add , or delete a single edge . here the best response strategy is still computable in polynomial time while the player has more freedom to act . generally speaking , in all the above models the obtained results on the poa are asymptotically worse than those we get in the two basic models , and we refer the reader to the cited papers for the actual bounds . [ [ our - results . ] ] our results . + + + + + + + + + + + + in this paper we concentrate on a seemingly underplayed generalization of ncgs , namely that in which for each player the set of possible adjacent nodes is constrained by a given connected , undirected graph @xmath30 , which in the end will host the created network . this finds its practical motivations in the physical limitations of constructing links , and was originally studied in @xcite for sumncg , where it is shown that the poa is upper bounded by @xmath31 and @xmath32 for @xmath33 and @xmath10 , respectively , and lower bounded by @xmath34 . here , we focus on the max - distance version , that we call maxncg@xmath35 , and we show that also in this case the poa is fairly bad , ) as it was done in @xcite , in order to incentivize the players to converge to connected equilibria . alternatively , this can be obtained by assuming that initially the players sit on a connected network ( embedded in the host graph ) , and they move ( non - simultaneously ) with a myopic best / improving response dynamics . ] even when the host graph is restricted to some basic standard layout patterns . more precisely , we show that the poa is @xmath36 for the classes of @xmath6-regular ( with any constant @xmath7 ) and 2-dimensional grid host graphs . this lower bound is asymptotically tight for @xmath9 , since we can prove a general upper bound of @xmath37 , where @xmath38 is the radius of @xmath30 . moreover , on general host graphs , we exhibit a lower bound of @xmath39 for @xmath40 . quite surprisingly , this implies that the poa is @xmath4 even when the players can build edges for free . on the positive side , if @xmath10 , we show the poa is at most 2 ( this is a direct consequence of the fact that in this case any equilibrium is a tree ) . finally , in the meaningful practical case in which the host graph is sparse ( i.e. , @xmath41 , with @xmath42 ) , we prove that the poa is @xmath43 , and so for very sparse graphs , i.e. @xmath44 , we obtain that the poa is constant . preliminarily to the above study , we also provide some results concerning the computational and dynamic aspects of the game . first , we prove that computing a best response for a player is -hard for any @xmath45 , thus extending a similar result given in @xcite for maxncg when @xmath46 . then , we prove that maxncg@xmath35 is not a potential game , by showing that an improving response dynamic does not guarantee to converge to an equilibrium , even if we assume a minimal _ liveness _ property that no player is prevented from moving for arbitrarily many steps . this implies that an improving response dynamic may not converge for the maxncg game as well ( after relaxing such a liveness property ) . a deeper discussion on dynamics in ncgs can be found in @xcite . the paper is organized as follows : in section 2 we analyze the computational / convergence aspects of the game , while sections 3 and 4 discuss the upper and lower bounds to the poa , respectively . first of all , we observe that , as for the sum - distance version of the problem studied in @xcite , it is open to establish whether maxncg@xmath35 always admits an equilibrium . this problem is particularly intriguing , since the topology of @xmath30 could play a discriminating role on that . we conjecture an affirmative answer to this question for any @xmath47 ( for @xmath48 it is trivially true as any strategy profile @xmath22 such that @xmath49 is a ne ) . as a first step towards this direction , observe that given any @xmath30 , a breadth - first search tree rooted at a center of @xmath30 , and in which each node owns the edges towards its children , is an equilibrium whenever @xmath50 , where @xmath38 is the radius of @xmath30 . indeed , notice that each vertex @xmath25 has no unactivated edges towards the vertices of its subtrees . this immediately implies that @xmath25 can not improve its cost by changing to a strategy having at most the same number of bought edges . moreover , in any other strategy , @xmath25 must incur a cost of at least @xmath51 to buy some additional edges while its routing cost , which is actually less than or equal to @xmath52 can not decrease by more than @xmath38 since it can never be less than @xmath38 . besides that , and similarly to other ncgs , we also have the bad news that the problem of computing the best response of player is -hard , as stated in the following theorem . [ th : best response ] for every constant @xmath53 and for every @xmath45 , the problem of computing a best response strategy of a player is -hard . the reduction is from the -complete _ 3-exact 3-cover _ problem ( 3x3c for short ) which , given ( i ) a set @xmath54 of @xmath55 objects , with @xmath56 , and ( ii ) a set @xmath57 of subsets of @xmath54 each having cardinality equal to 3 and such that every @xmath58 is a member of at most three sets in @xmath57 , asks for determining whether there exists a subset @xmath59 of cardinality equal to @xmath6 that covers @xmath54 , i.e. , @xmath60 . the reduction is the following . let @xmath61 . for a given instance of the 3x3c problem with @xmath55 objects we build a host graph @xmath30 having @xmath62 vertices . more precisely , @xmath30 contains an object vertex @xmath63 for every object @xmath64 , and a set vertex @xmath65 for every @xmath66 . for every set @xmath66 and for every @xmath64 , @xmath30 contains a path of length @xmath67 having @xmath63 and @xmath65 as endpoints iff @xmath68 . all paths are vertex - disjoint except for at most their endpoints . finally , @xmath30 contains an additional path @xmath69 of length @xmath70 which is vertex - disjoint w.r.t . all the other paths , and an additional vertex @xmath71 linked to all the set vertices and to one endpoint of @xmath69 , that we denote by @xmath72 . let @xmath22 be any strategy profile such that @xmath19 contains all edges of @xmath30 except those incident to @xmath71 . we claim that any best response strategy of player @xmath71 has cost equal to @xmath73 iff @xmath57 contains a subset of cardinality @xmath6 that covers @xmath54 . to prove one direction , observe that if @xmath59 has size @xmath6 and covers @xmath54 , then , by buying the edge towards @xmath72 and all the @xmath6 edges @xmath74 for every @xmath75 , @xmath71 would incur a cost equal to @xmath76 . in fact , it is easy to see that each vertex of @xmath77 which is neither @xmath71 nor a vertex in @xmath69 is at distance of at most @xmath67 from some object vertex . furthermore , each object vertex is at distance of at most @xmath67 from some set vertex @xmath65 such that @xmath75 . to prove the other direction , let @xmath78 be the strategy profile obtained from @xmath22 by changing @xmath71 s strategy with one of its best response strategies and assume that @xmath79 . let @xmath80 be the subset of @xmath57 containing set @xmath81 iff @xmath71 is buying the edge towards @xmath65 in @xmath78 . first of all , observe that if @xmath80 does not cover @xmath54 , or @xmath71 has not bought the edge towards @xmath72 , then the routing cost of @xmath71 is greater than or equal to @xmath82 . consequently , @xmath71 has bought the edge towards @xmath72 and @xmath80 covers @xmath54 . therefore @xmath83 . as the distance from @xmath71 to the endpoint of @xmath69 different from @xmath72 is @xmath84 , and since @xmath85 , it follows that @xmath86 . notice that the result stated in theorem [ th : best response ] holds for any @xmath45 , and so this extends the -hardness proof given in @xcite which holds for complete host graphs and @xmath46 . we now discuss a negative result about the convergence of the improving response dynamics . to the best of our knowledge , this is the first result showing that an improving response dynamics on maxncg might not converge to an equilibrium . a deeper discussion about dynamics in ncgs can be found in @xcite . [ thm : lb_poa ] for every value of @xmath87 , maxncg@xmath35 is not a potential game . moreover , if @xmath88 , an improving response dynamics may not converge to an equilibrium . we prove the non existence of a potential function by showing a cyclic sequence of strategy profiles where , at the end of each cycle , the total cost of the moving players has decreased by a positive constant . let @xmath89 be an integer satisfying @xmath90 and consider a host graph @xmath30 similar to the one shown in figure [ fig : not_potential ] . @xmath30 is composed by a cycle of @xmath91 nodes labelled from @xmath92 to @xmath91 , by a path of @xmath93 edges having @xmath71 and @xmath72 as endpoints , and by all the edges between @xmath71 and the nodes of the cycle . the strategy profile @xmath94 being played is shown by using a graphical notation explained in the caption . . on the left side , the initial configuration , where directed edges exit from their respective owner node , dashed edges are non - bought edges of @xmath30 , and the spline denotes a path between @xmath72 and @xmath71 , whose edges are arbitrarily owned . ] in such a status , player @xmath92 is paying @xmath95 , whilst changing the strategy to @xmath96 by removing the edge @xmath97 yields a cost of @xmath98 , thus saving @xmath5 . observe that now @xmath99 is @xmath100 , and so @xmath71 has interest in swapping the edge @xmath101 with the edge @xmath102 , thus obtaining the strategy @xmath103 and saving @xmath92 . in such configuration @xmath104 has increased to @xmath105 , therefore player @xmath92 can buy back the edge @xmath97 , as shown in strategy @xmath106 , thus reducing its cost to @xmath107 , i.e. , saving @xmath108 . notice how configuration @xmath106 is similar to @xmath94 , with the only difference being the edge bought by @xmath71 . since @xmath90 , by repeating @xmath91 times these strategy changes , every node in the cycle @xmath109 will play a move at least once , and the resulting configuration is identical to @xmath94 , hence the players will cycle . to prove the latter part of the claim it suffices to note that after each cycle : ( i ) for @xmath88 each strategy change is an improving response , and ( ii ) that the nodes in the path from @xmath71 to @xmath72 other than @xmath71 , can never change their strategy . actually , the above proof shows that the improving response dynamics may not converge even if the minimal liveness property that each player takes a chance to make an improving move every fixed number of steps is guaranteed . indeed , as observed in the proof , the players sitting on the path appended to @xmath71 do not move just because they can not . therefore , we can extend the above proof to the case in which @xmath30 is a complete graph if the players sitting on the path appended to @xmath71 can never move . this shows that the improving response dynamics may not converge on complete host graphs as well , i.e. , for the classic ncg . in this section we prove some upper bounds to the poa for the game . in what follows , for a generic graph @xmath77 , we denote by @xmath110 and @xmath111 its radius and its diameter , respectively , and by @xmath112 the eccentricity of node @xmath21 in @xmath77 . moreover , we denote by @xmath113 the _ social cost _ of a generic strategy profile @xmath22 ( i.e. , the sum of players individual costs ) , and by @xmath114 a strategy profile minimizing the social cost . then [ lm : ub_aux ] let @xmath115 be a ne , and let @xmath116 . then , we have that @xmath117 . let @xmath25 be a center of @xmath77 , and let @xmath118 be a shortest path tree of @xmath77 rooted at @xmath25 . clearly , the diameter of @xmath118 is at most @xmath119 . now , for every node @xmath21 , let us denote by @xmath120 the number of edges of @xmath118 bought by @xmath21 in @xmath22 . the key argument is that if a node @xmath21 bought only the @xmath120 edges of @xmath118 , its eccentricity would be at most @xmath121 . hence , since @xmath22 is a ne , we have that @xmath122 . by summing up the inequalities over all nodes , we obtain @xmath123 now , since @xmath124 , we have @xmath125 as an immediate consequence , we obtain the following : for @xmath116 , the @xmath126 is @xmath127 . another interesting consequence of lemma [ lm : ub_aux ] concerns sparse host graphs : [ th : sparse host graph ] if the host graph @xmath30 has @xmath128 edges , and @xmath42 , then the @xmath126 is @xmath129 . let @xmath115 be an equilibrium network . since @xmath77 must be connected , we have that @xmath130 . this is sufficient to provide an upper bound to the diameter of @xmath77 . indeed , in @xcite it is shown that the diameter of a connected graph obtained from a supergraph of diameter @xmath131 by deleting @xmath132 edges is at most @xmath133 . this implies that in our case @xmath134 . now , the claim follows from lemma [ lm : ub_aux ] . theorem [ th : sparse host graph ] implies that , for very sparse host graphs @xmath30 , i.e. , @xmath41 and @xmath44 , we have that the poa is @xmath11 , for any @xmath5 . next theorem shows that the poa is upper bounded by 2 when @xmath5 is at least @xmath0 . a similar result has already been proved in @xcite for the case in which @xmath30 is a complete graph . in fact , it turns out that the same proof extends to any host graph @xmath30 . for @xmath135 , the @xmath126 is at most @xmath136 . we end this section by showing that when either @xmath5 is small , or the host graph has small diameter , every stable tree ( if any ) is a good equilibrium . this generalizes a result in @xcite given for complete host graphs , which states that the social cost of every acyclic equilibrium is @xmath11 times the optimum . [ lm : ecc(u)-ecc(v)<=1+a ] let @xmath137 be an edge of the host graph . then , for every stable graph @xmath77 , we have @xmath138 . w.l.o.g . assume @xmath139 . if @xmath140 , then the claim trivially holds . otherwise , if @xmath25 buys the edge @xmath141 then its eccentricity will decrease at least by @xmath142 , while its building cost will increase by @xmath5 . since @xmath77 is stable , we have @xmath143 , and the claim follows . [ crl : ecc(u)-ecc(v)<=1+a dh ] for every @xmath144 and for every stable graph @xmath77 , it holds that @xmath145 . [ lm : ecc(u)-ecc(v ) - > o(1+a ) ] let @xmath115 be a stable graph . if there are two nodes @xmath144 such that @xmath146 with @xmath147 and @xmath148 , then @xmath149 . we have @xmath150 moreover , from corollary [ crl : ecc(u)-ecc(v)<=1+a dh ] , we have @xmath151 from which , we obtain @xmath152 and hence the claim . we are now ready to give the following let @xmath22 be a ne such that @xmath115 is a tree . then , @xmath153 . let us consider a center @xmath25 of @xmath77 , and let @xmath21 be a node in the periphery of @xmath77 , namely @xmath154 . since @xmath77 is a tree , we have @xmath155 . now , using lemma [ lm : ecc(u)-ecc(v ) - > o(1+a ) ] and lemma [ lm : ub_aux ] , the claim follows . in this section we prove some lower bounds to the poa of the game , as summarized in table [ lb ] . .obtained lower bounds to the poa . [ cols="<,^,^,^",options="header " , ] [ lb ] before getting to the technical details , let us discuss the significance of the above bounds . first of all , we notice that the lower bound for @xmath156 is tight , due the upper bound given in the previous section . moreover , observe that we can obtain such a lower bound for two prominent classes of host graphs , namely for @xmath6-regular graphs ( for any constant @xmath7 ) and for 2-dimensional grids . we view this as a meaningful result , due to the practical relevance of such host topologies . concerning the case @xmath157{n } ) \cap o(\sqrt{n})$ ] , we notice that the lower bound still holds for the classes of @xmath6-regular graphs ( for any constant @xmath158 ) and 2-dimensional grids , but now it is not tight . finally , the lower bound of @xmath159 for @xmath160{n})$ ] is quite because it implies a very large lower bound of @xmath4 for tiny values of @xmath5 . summarizing , we point out that we get a polynomial lower bound for @xmath161 , for any @xmath162 , in strong contrast with the almost everywhere constant upper bound to the poa of maxncg . if the host graph is a 2-dimensional square grid , then the poa is @xmath163 . let @xmath164 where @xmath165 is an odd number , and let @xmath30 be a 2-dimensional square grid of @xmath166 vertices . in the rest of the proof , we assume that the vertex in the @xmath167-th row and @xmath168-th column of the grid is labeled with @xmath169 , where @xmath170 . for every @xmath171 , let @xmath172 be the path in @xmath30 which spans all the vertices of the @xmath168-th column of @xmath30 . let @xmath173 . let @xmath174 be the set of edges linking vertex @xmath175 with vertex @xmath176 iff @xmath168 is even and let @xmath177 be the set of edges linking vertex @xmath178 with @xmath179 iff @xmath168 is odd . when the host graph @xmath30 is a square grid of @xmath0 vertices . ] let @xmath77 be the subgraph of @xmath30 whose edge set is @xmath180 ( see also figure [ fig : grid ] ) . observe that @xmath77 is a tree of radius greater than or equal to @xmath181 . observe also that @xmath182 and @xmath183 are the two centers of @xmath77 . let @xmath182 be the root of @xmath77 and let @xmath184 be the directed version of @xmath77 where all the root - to - leaf paths are directed towards the leaves . finally , let @xmath22 be the strategy profile induced by @xmath184 , i.e. , each player @xmath21 is buying exactly the edges in @xmath184 outgoing from @xmath21 . clearly , @xmath185 . we prove that @xmath22 is a ne by showing that every vertex @xmath169 , with @xmath186 and @xmath187 , is playing a best response strategy . indeed , if we show that @xmath169 is playing a best response strategy , then , by symmetry , also @xmath188 is playing a best response strategy . let @xmath167 and @xmath168 be two fixed integers such that @xmath189 and @xmath171 and let @xmath190 be the number of edges bought by @xmath169 in @xmath22 . since @xmath77 is a tree and since @xmath182 and @xmath183 are the two centers of @xmath77 , there exists a vertex @xmath191 , with @xmath192 and @xmath193 , such that the distance in @xmath77 from @xmath169 to @xmath191 is exactly equal to the eccentricity of @xmath169 in @xmath77 . observe also that the ( unique ) path in @xmath77 from @xmath169 to @xmath191 traverses the root as well as the vertex @xmath183 . let @xmath191 be any vertex such that @xmath192 and @xmath193 . first of all , observe that if we add to @xmath77 all the edges adjacent to @xmath169 in @xmath30 , then the distance from @xmath169 to @xmath191 decreases by at most @xmath5 . since the cost of activating new links is at least @xmath5 , @xmath169 can not improve its cost function by buying more than @xmath190 edges . now we prove that @xmath169 can not improve its cost function by buying at most @xmath190 edges . first of all , observe that @xmath190 is the minimum number of edges @xmath169 has to buy to guarantee connectivity . moreover , to guarantee connectivity , @xmath169 has to buy an edge towards some vertex of every subtree of @xmath77 rooted at any of its @xmath190 children . since the subtree of @xmath77 rooted at @xmath169 does not contain @xmath191 when @xmath169 is not the root , @xmath169 can not improve its eccentricity , and thus its cost function , by buying an edge towards some vertex of every subtree of @xmath77 rooted at any of its @xmath190 children . furthermore , if @xmath169 is the root of @xmath77 , then @xmath169 can not improve its eccentricity , and thus its cost function , by buying an edge towards some vertex of every subtree of @xmath77 rooted at any of its @xmath190 children as @xmath169 is already buying the unique edge of @xmath30 linking it to the subtree of @xmath77 rooted at @xmath194 . to complete the proof , observe that @xmath195 is upper bounded by the social cost of @xmath30 , i.e. , @xmath196 . since @xmath197 , we have that @xmath198 we now show that a similar lower bound holds also when the host graph is @xmath6-regular . [ thm : lb_poa_k_regular ] if the host graph is @xmath6-regular , with @xmath199 , then the poa is @xmath200 . first of all , observe that for @xmath201 and @xmath202 the claim trivially holds since the lower bound becomes @xmath203 . therefore , we consider the case @xmath204 and @xmath205 . for the sake of readability , we provide the complete proof for even values of @xmath6 and we only sketch it for odd values of @xmath6 as the construction is very similar . let @xmath91 be the greatest integer such that @xmath206 , and let @xmath67 be a large value such that @xmath207 . notice that if the number of players @xmath0 is sufficiently large , then @xmath208 . we will use a host graph @xmath30 composed by : ( i ) a path @xmath69 of @xmath67 nodes , numbered from @xmath209 to @xmath210 , ( ii ) a set of shortcut edges on @xmath69 ( as described in the following ) , and ( iii ) a set of gadgets appended to @xmath69 and used to increase to @xmath6 the degree of its vertices ( as described in the following ) . concerning the shortcut edges , let @xmath211 be the node on @xmath69 numbered @xmath212 for @xmath213 , where @xmath214 . then , a shortcut edge connects @xmath211 to @xmath215 , for @xmath216 . notice that any node on @xmath217 has now degree @xmath136 while the degree of all the nodes @xmath218 is equal to 4 . concerning the gadgets , for each node @xmath25 on @xmath69 that has degree @xmath219 , we augment @xmath30 as follows : * we add a complete , loop - free , graph @xmath220 on @xmath221 vertices to @xmath30 ; * we remove @xmath222 vertex - disjoint edges from @xmath220 . so , @xmath223 vertices of @xmath220 have degree @xmath6 while the other @xmath224 vertices have degree @xmath225 ; * we connect @xmath25 to the @xmath224 nodes of @xmath220 with degree @xmath225 . at the end of this process the resulting host graph @xmath30 is @xmath6-regular . consider now a strategy profile @xmath22 such that : * all the edges of the path @xmath69 are bought ( arbitrarily ) by vertices other than @xmath211 , @xmath226 ; * each vertex of a gadget that has an edge towards a node on @xmath69 buys it ; * the remaining vertices of the gadgets buy a single edge towards a vertex adjacent to a node of @xmath69 . an example of the resulting configuration for @xmath227 along with the edges of the host graph is shown in figure [ fig : lb_poa_k_regular](b ) . for ( b ) @xmath227 , and ( c ) @xmath228 . ] we show that @xmath22 is a ne . indeed , every node @xmath211 can only change its strategy by buying either one or two edges , but this can decrease its eccentricity by at most @xmath98 , while increasing its building cost of at least @xmath229 . moreover , the remaining nodes in @xmath69 can not change their strategy , as doing so will cause the disconnection of the graph . finally , the nodes of the gadgets buy just a single edge , and no other choice can decrease their eccentricity . clearly @xmath230 , as @xmath19 is a tree with radius @xmath231 . let now @xmath232 be the graph obtained by adding to @xmath19 the shortcut edges of @xmath30 . the number of edges of @xmath232 is @xmath233 , and its diameter is bounded by @xmath234 , as @xmath235 can take advantage of the shortcut edges . as a consequence , with a small abuse of notation , we have @xmath236 . using the relations @xmath237 , @xmath238 , and @xmath62 , we have that @xmath239 from which the claim easily follows . if @xmath6 is odd , then a host graph similar to the one shown in figure [ fig : lb_poa_k_regular](c ) ( for the case @xmath228 ) is considered . notice that the shortcut edges are now vertex - disjoint , and each node incident to them has degree @xmath240 , but for @xmath235 and @xmath241 that have degree 2 . we first append a gadget to every node @xmath71 with degree @xmath136 , in order to obtain a @xmath240-regular graph . this gadget is a similar to a clique on @xmath242 vertices where an edge @xmath243 between two adjacent vertices has been replaced by two vertices going from the endpoints of @xmath243 to a new node @xmath25 . the vertex @xmath25 is then connected to @xmath71 using a new edge ( bought by @xmath25 ) . if @xmath228 we are done , otherwise we can append a gadget similar to the one shown for the even case to each node of the graph we just constructed , in order to increase the degree of each node to @xmath6 . notice that we can extend the previous lower bound to outerplanar and series - parallel graphs just by considering the host graph composed by the path plus the shortcut edges , without any additional gadget , as shown in figure [ fig : lb_poa_k_regular](a ) . the poa is @xmath163 , even when the host graph is a an outerplanar or a series - parallel graph . we end this section by proving a non - constant lower bound to the poa when @xmath205 . remarkably , our lower bound implies a non - constant lower bound to the poa for every small value of @xmath5 , i.e. , players buy edges for free . our lower bounding construction is a non - trivial modification of the 2d - torus - rotated-45@xmath244 construction used in @xcite to prove a lower bound for basicncg . [ lm : torus ] for @xmath205 , the poa is @xmath245 . let @xmath246 and let @xmath247 be an edge - weighted 2d - torus - rotated-45@xmath244 consisting of @xmath248 vertices that we call junction vertices . for every pair of integers @xmath249 , with @xmath250 even , there is exactly one vertex of @xmath247 labeled with @xmath169 . we treat the two integers of a vertex label as modulo @xmath251 . each vertex @xmath169 has exactly four neighbors in @xmath247 : @xmath252 . all edge weights are equal to @xmath253 . for every pair of integers @xmath249 , let @xmath254 . the properties satisfied by @xmath247 are the following : 1 . @xmath247 is vertex transitive , i.e. , any vertex can be mapped to any other by a vertex automorphism , i.e. , a relabeling of vertices that preserves edges ; 2 . the distance between two vertices @xmath169 and @xmath191 in @xmath247 is equal to @xmath255 , where @xmath256 ; 3 . the eccentricity of each vertex in @xmath247 is equal to @xmath257 ; 4 . for every @xmath258 , the distance from every vertex @xmath259 to vertex @xmath260 is equal to @xmath257 ; 5 . for every edge @xmath243 of @xmath247 , the eccentricity of both endpoints of @xmath243 in @xmath261 is greater than or equal to @xmath262 ; 6 . for every edge @xmath243 of @xmath247 and for every vertex @xmath169 , the distance from @xmath169 and the closest endpoint of @xmath243 is less than or equal to @xmath263 . it is easy to see that ( i ) holds and it is also easy to see that ( iv ) holds once ( ii ) has been proved . to prove ( ii ) , it is enough to observe that each label can change by @xmath264 each time we move from one vertex to any of its neighbors . to prove ( iii ) , we use ( i ) and the fact that the distance from vertex @xmath191 to @xmath265 , which is equal to @xmath266 , is maximized for @xmath267 . to prove ( v ) , we first use ( i ) to assume that , w.l.o.g , @xmath243 is the edge linking @xmath265 with @xmath268 . next , we observe that any path in @xmath269 going from @xmath265 to @xmath270 must traverse a neighbor @xmath21 of @xmath265 in @xmath269 and the distance between @xmath21 and @xmath270 in @xmath30 is equal to @xmath257 because one of the two integers in the label of @xmath21 is equal to @xmath221 . finally , to prove ( vi ) , we first use ( i ) to assume that , w.l.o.g . , @xmath271 , i.e. , @xmath169 is @xmath265 , and the two endpoints of @xmath243 are , respectively , @xmath191 and @xmath272 , where @xmath273 . using ( ii ) , it is easy to see that the distance from @xmath265 to @xmath272 is less than or equal to @xmath274 . let @xmath77 be an unweighted graph obtained from @xmath247 by replacing each edge of @xmath247 with a path of length @xmath132 via the addition of @xmath275 new vertices per edge of @xmath247 . let @xmath30 be the host graph obtained from @xmath77 by adding an edge between @xmath169 and every vertex in @xmath276 , for every junction vertex @xmath169 ( see also figure [ fig : torus ] ) . notice that the number of vertices of @xmath30 is @xmath277 . in what follows , we call the vertices in @xmath30 which are not in @xmath247 path vertices . on the left side , the host graph @xmath30 is depicted . for the sake of readability , only junction vertices are visible and not all the edges are shown . the white vertices of row @xmath278 are copies of the vertices of row @xmath209 while the white vertices of column @xmath278 are copies of the vertices of column @xmath209 . the solid edges are paths of length @xmath132 , while the dashed edges are all the other edges adjacent to vertex @xmath265 . on the right side , the stable graph @xmath77 is depicted . ] let @xmath22 be any strategy profile such that @xmath185 and all edges of @xmath19 are bought by players sitting on path vertices , i.e. , no edge of @xmath19 is bought by some player sitting on junction vertices . we prove that @xmath22 is a ne . we start proving that players sitting on junction vertices are playing a best response strategy . let @xmath169 be a junction vertex . observe that @xmath169 is not buying any edge , therefore it suffices to show that @xmath169 can not improve its cost function by buying edges . first of all , observe that by ( iii ) and ( vi ) , the eccentricity of @xmath169 in @xmath77 is equal to @xmath257 . indeed , if @xmath21 is a path vertex of a path @xmath69 corresponding to edge @xmath243 of @xmath247 , then the distance from @xmath169 to the closest endpoint of @xmath69 ( which corresponds to the closest endpoint of @xmath243 ) is less than or equal to @xmath263 . therefore , the distance from @xmath169 to @xmath21 is less than or equal to @xmath257 . to prove that @xmath169 is in equilibrium , simply observe that if we add to @xmath77 all edges of @xmath30 incident to @xmath169 , i.e. , all edges linking @xmath169 to vertices in @xmath276 , then by ( iv ) the distance from @xmath169 to @xmath260 is still @xmath257 . now , we prove that players sitting on path vertices are playing a best response strategy . let @xmath21 be a path vertex . first of all , the eccentricity of @xmath21 in @xmath77 is less than or equal to @xmath279 by ( vi ) . indeed , if @xmath21 is a vertex of a path @xmath69 corresponding to edge @xmath243 of @xmath247 , then the distance from any junction vertex to the closest endpoint of @xmath69 ( which corresponds to the closest endpoint of @xmath243 ) is less than or equal to @xmath263 . therefore , the distance from @xmath21 to every junction vertex is less than or equal to @xmath257 and the distance from @xmath21 to every other path vertex is less than or equal to @xmath279 . now , observe that @xmath77 already contains all edges of @xmath30 incident to @xmath21 and , moreover , the degree of @xmath21 in @xmath77 is equal to 2 . therefore , @xmath21 might improve its cost function by removing exactly one edge incident to it , i.e. , by buying fewer edges than those it is buying in @xmath22 . however , if @xmath21 removes any of its incident edges in @xmath77 , thus saving an @xmath5 factor from its building cost , then by ( v ) the eccentricity of the unique junction vertex closest to @xmath21 becomes greater than or equal to @xmath262 and thus , the eccentricity of @xmath21 also becomes greater than or equal to @xmath280 . since @xmath281 , @xmath21 does not improve its cost function by buying fewer edges than those it is buying in @xmath22 . to complete the proof , we have to show that poa is @xmath245 . first of all , observe that the radius of @xmath30 is @xmath282 . let @xmath118 be a breadth - first - search tree rooted at @xmath265 . clearly the radius of @xmath118 is also @xmath283 . furthermore , the social cost of @xmath284 is upper bounded by the social cost of @xmath118 , i.e. , @xmath285 . as @xmath286 , we have that @xmath287 in this paper , we studied the maxncg in the scenario in which the strategy space of all the players is constrained by a host graph . our game is interesting for two reasons . first of all , it models the practical situation in which not all the links can be constructed due to physical limitations . furthermore , the -hardness barrier of computing a best response strategy of a player in the maxncg , which clearly carries over into our game , can be easily broken if we restrict our game to the class of host graphs of ( constant ) bounded degree . in our paper , we proved a strong lower bound of @xmath288 to the poa for two meaningful graph classes , namely bounded - degree host graphs and 2-dimensional square grids . our lower bound asymptotically matches the general upper bound we provided for every @xmath9 . since the poa in the classical maxncg is mostly constant ( see @xcite ) , we therefore have that the drawback of having a polynomial - time computability of best response strategies is the existence of stable networks of large social cost . finally , we concluded our paper by proving another lower bound of @xmath159 to the poa . observe that all our lower bounds are never smaller than the lower bound of @xmath289 known for the corresponding sum version of the game ( see @xcite ) . this is mainly due to the fact that reducing the routing cost of a player in our game is costlier , in terms of the building cost incurred by that player , than in the corresponding sum version of the game . d. bil , l. gual , and g. proietti , bounded - distance network creation games , _ proceedings of the 8th international workshop on internet & network economics ( wine12 ) _ , lecture notes in computer science 7695 , springer , 7285 , 2012 . s. ehsani , m. fazli , a. mehrabian , s.s . sadeghabad , m. saghafian , s. shokatfadaee , and m. safari , on a bounded budget network creation game , in _ proceedings of the 23rd acm symposium on parallelism in algorithms and architectures ( spaa11 ) _ , acm press , 207214 , 2011 . a. fabrikant , a. luthra , e. maneva , c.h . papadimitriou , and s. shenker , on a network creation game , in _ proceedings of the 22nd symposium on principles of distributed computing ( podc03 ) _ , acm press , 347351 , 2003 . n. laoutaris , l.j . poplawski , r. rajaraman , r. sundaram , and s .- h . teng , bounded budget connection ( bbc ) games or how to make friends and influence people , on a budget , in _ proceedings of the 27th acm symposium on principles of distributed computing ( podc08 ) _ , acm press , 165174 , 2008 . p. lenzner , on dynamics in basic network creation games , in _ proceedings of the 4th international symposium on algorithmic game theory ( sagt11 ) _ , lecture notes in computer science 6982 , springer , 254265 , 2011 . p. lenzer , greedy selfish network creation , in _ proceedings of the 8th international workshop on internet and network economics ( wine12 ) _ , lecture notes in computer science 7695 , springer 142155 , 2012 . a. mamageishvili , m. mihalk , and d. mller , tree nash equilibria in the network creation game , in _ 10th international workshop on algorithms and models for the web graph ( waw13 ) _ , lecture notes in computer science 8305 , springer , 118129 , 2013 . m. mihalk and j.c . schlegel , asymmetric swap - equilibrium : a unifying equilibrium concept for network creation games , in _ proceedings of the 37th international symposium on mathematical foundations of computer science ( mfcs12 ) _ , lecture notes in computer science 7464 , springer , 693704 , 2012 . schoone , h.l . bodlaender , and j. van leeuwen , improved diameter bounds for altered graphs , in _ proceedings of the 12th workshop on graph - theoretic concepts in computer science ( wg86 ) _ , lecture notes in computer science 246 , 227236 , 1986 .

in this paper we study a generalization of the classic _ network creation game _ in the scenario in which the @xmath0 players sit on a given arbitrary _ host graph _ , which constrains the set of edges a player can activate at a cost of @xmath1 each . this finds its motivations in the physical limitations one can have in constructing links in practice , and it has been studied in the past only when the routing cost component of a player is given by the sum of distances to all the other nodes . here , we focus on another popular routing cost , namely that which takes into account for each player its _ maximum _ distance to any other player . for this version of the game , we first analyze some of its computational and dynamic aspects , and then we address the problem of understanding the structure of associated pure nash equilibria . in this respect , we show that the corresponding price of anarchy ( poa ) is fairly bad , even for several basic classes of host graphs . more precisely , we first exhibit a lower bound of @xmath2 for any @xmath3 . notice that this implies a counter - intuitive lower bound of @xmath4 for very small values of @xmath5 ( i.e. , edges can be activated almost for free ) . then , we show that when the host graph is restricted to be either @xmath6-regular ( for any constant @xmath7 ) , or a 2-dimensional grid , the poa is still @xmath8 , which is proven to be tight for @xmath9 . on the positive side , if @xmath10 , we show the poa is @xmath11 . finally , in the case in which the host graph is very sparse ( i.e. , @xmath12 , with @xmath13 ) , we prove that the poa is @xmath11 , for any @xmath5 . * keywords : * network creation games , pure nash equilibrium , price of anarchy , host graph .

introduction preliminary results upper bounds lower bounds conclusions

the rapid development of ground - based high - energy astronomy during the last ten years @xcite is closely associated with the success of the atmospheric cherenkov imaging technique @xcite developed with the whipple observatory 10 m telescope @xcite and elsewhere . cosmic rays and @xmath0-rays entering the atmosphere generate cascades of particles radiating cherenkov light . at ground level , the cherenkov light front extends over more than @xmath1 from the shower axis . thus , a single imaging telescope achieves an effective collection area of more than @xmath2 . the shape of the image can be used to preferentially select @xmath0-ray events over the much more frequent cosmic - ray events . furthermore , the elongated image of a @xmath0-ray shower points back to the source on the sky , providing information on its arrival direction . most of the present generation experiments ( cangaroo , hess , magic and veritas ) consist of arrays of imaging telescopes . following earlier pioneering studies @xcite , the power of the stereoscopic technique @xcite , was plainly illustrated with the hegra experiment @xcite and more recently with the hess experiment @xcite which achieved unprecedented levels of @xmath0-ray discrimination . in these arrays , each telescope records an image of the same shower . for a @xmath0-ray primary , the major axis of each image points back to the source so that superposing the images produces the source location on the sky . furthermore , the position of the image around the source corresponds to the position of the shower around the line from the telescope to the source . combining this information from two or more telescopes provides the position and orientation of the shower axis . in other words , each telescope can be associated with a plane that contains the shower axis and the telescope . as the shower axis is the line of intersection of the planes associated with each of the array telescopes , only two telescopes are required to extract this axis . since accurate reconstruction is difficult or impossible when the impact point is closely aligned with the two telescopes , a minimum of three telescopes is preferable . the stereoscopic technique described above requires that each image contains a sufficient number of pixels for the major axis to be clearly identified . however , this is generally not the case for events close to the detection threshold and for which some or even all of the images can consist of only a few pixels significantly above noise fluctuations . only the image positions are then available for a reconstructive analysis . in this paper , we study the possibility of reconstructing @xmath0-ray events using only the shower image centroid position from each telescope . such a minimalist analysis might be particularly important for the lowest energy events . the next section describes the method and in the following section , we use monte - carlo simulations to present an illustration of the method . finally , we discuss approaches for discriminating against cosmic - rays when the image positions are the only available significant information . consider a plane perpendicular to the shower axis . let @xmath3 be the intersection point of the shower axis in this plane . project the location of each telescope along the direction of the shower axis to the plane ; label these points @xmath4 . locate the source , @xmath5 , and the image centroid , @xmath6 , on the sky . we assume the image centroid to lie on the shower axis projected on the sky so that @xmath4 , @xmath3 , @xmath6 and s are coplanar as shown on figure [ geom ] . as a result the lines , @xmath7 and @xmath8 , are parallel . expressing this result using convenient coordinate systems on the sky and in the plane perpendicular to the source direction yields : @xmath9 where i ranges from 1 though the number of telescopes , n. since there are four unknown quantities , n must generally be at least equal to four for the system to have a discrete number of solutions . a minimum of four telescopes is necessary for reconstructing the shower axis from the image centroids . in fact , the minimal telescope number could in principle be reduced to three or even two through the geometrical connection between the distance @xmath7 and the impact parameter to telescope @xmath10 . however , this relationship , by relying on the physics as well as the geometry of the atmospheric shower , depends on the primary energy and is more susceptible to fluctuations than our assumption that the image centroid lies on the shower axis image . we prefer not to use such a complicated relationship at this stage . because of the projection of the telescope locations onto the plane perpendicular to the primary direction , @xmath11 and @xmath12 depend on the primary arrival direction . however , we note that the field of view of atmospheric cherenkov imaging telescopes remains relatively small , less than 2 or 3 degrees in radius . thus , to a good approximation , we can perform the projection along the tracked direction instead of the arrival direction . for a typical array with distances between telescopes of the order of 100 m , this approximation will affect the relative positions of the telescopes in the projection plane by much less than the physical size of the telescopes . we can now directly use , for example , the first two equations in [ equat ] to express @xmath11 and @xmath12 as functions of @xmath13 and @xmath14 . these expressions can then be substituted into the remaining equations . for more than four telescopes , the system is over - constrained . in the case of a four - telescope array , the system consists of two coupled quadratic equations in @xmath13 and @xmath14 with no apparent analytical solution . we thus use a numerical approach to find the source and impact point by minimizing the function : @xmath15 . since the equations for telescopes 1 and 2 are used to express @xmath11 and @xmath12 , the sum extends from @xmath16 up to @xmath17 , the total number of telescopes . this function reaches a minimum at the source position . in practice , by further manipulating the equations , the minimization becomes one - dimensional . for an array of more than four telescopes ( @xmath18 ) , two approaches are possible : a full four - dimensional minimization or one - dimensional minimizations for all groups of three telescopes . in the latter case , the solution will depend on the specific order of equation choices in reducing the problem to a one - dimensional optimization . the @xmath19 solutions have to be combined into a single solution . our examples only use this one - dimensional minimization with a four - telescope - array . figure [ stereoprincip ] illustrates the geometry and the @xmath20 function for a simulated vertical 500 gev shower and a four - telescope array ( @xmath21 ) arranged in a centered equilateral triangle . because of the high primary energy , the major axes of the images precisely point toward the source position and impact point . the function @xmath20 reaches a minimum near @xmath5 because the lines connecting @xmath5 to the various image centroids are approximately parallel to the lines connecting @xmath22 to the various telescopes . in the following section , we present the capabilities of this method for this array . in the case of a four - telescope array , the function @xmath20 reaches zero in two points . on figure [ stereoprincip ] , one solution is close to the actual source position ; the second solution is near @xmath23 . a fifth telescope could in principle select the proper solution , but we can also include additional , as yet unused , information . for example , each image centroid should lie on the the impact position side of the source point as indicated by the spatial position of the impact point with respect to the corresponding telescope . unfortunately , this criterion does not always select only one of the two the solutions . as previously discussed , we have not incorporated the distance @xmath7 from the source to the image centroid . for @xmath0-rays , this distance is approximately linearly correlated with the impact parameter @xmath24 . the slope of the relationship depends on the elevation and primary energy . hence , we can require that the ratio , @xmath25 have approximately the same value for each telescope . in practice , when the first criterion does not select a unique solution , we compute the standard deviation of the four distance ratios for both solutions and retain the solution with the smallest standard deviation . to test and illustrate this method , we used the grisu simulation package @xcite to simulate vertical 100-gev @xmath0 rays with a veritas - like array of four 12m - telescopes . each shower image is polluted by a typical level of night sky background . the individual telescope trigger conditions demanded that three channels coincidently exceed a threshold corresponding to four photoelectrons . all four telescopes must trigger to record an event . these are detailed monte - carlo simulations generated for a generic act array ; hence , they do not include the level of specificity required to characterize a particular instrument . figure [ perform ] shows the error distribution of the source position and impact - point reconstruction . the standard deviation from the actual source position is @xmath26 and the standard deviation from the actual impact point is @xmath27 . this analysis only uses the image centroids . we realize that an analysis using the distribution of light in the image , when the image is sufficiently bright , might give more accurate results . however , it is important to note that no threshold nor cosmic - ray discrimination criteria have been applied yet . in the case of point sources , the angular resolution provides an important factor of discrimination against cosmic - rays . further discrimination is possible using only the image centroid locations and the image brightness . a more detailed image analysis would be more efficient at rejecting cosmic rays , especially the high energy ones . here we just concentrate on the exercise of only using the centroid position information and total image brightness . the ratio @xmath28 , described in the previous section can now be used as a discrimination tool against cosmic rays . as is often the case , different parts of a cosmic - ray shower are seen by different telescopes . thus , the almost linear relationship between the @xmath7 and @xmath8 distances breaks down and the @xmath28 values will exhibit larger fluctuations between telescopes for a given cosmic ray shower . on the left panel of figure [ discri ] we compare the distributions of the relative standard deviation , @xmath29 , of the @xmath30 values for simulated 100-gev @xmath0-rays and 100-gev to 10-tev protons . a cut on the parameter @xmath29 can preferentially select @xmath0 rays . additionally , we have not used the quantity of light @xmath31 in each image . the amount of cherenkov light received by each telescope from a @xmath0-ray shower should follow the well - known cherenkov light - pool density profile @xcite . for vertical showers the profile is relatively flat for impact parameters less than @xmath32 . on the contrary , cosmic rays produce cherenkov light pools on the ground that are more patchy and irregular . on the right panel of figure [ discri ] , we compare the distribution of the relative standard deviation of the quantity of light @xmath33 . here again a cut on this parameter permits some discrimination against cosmic rays . as an example , selecting @xmath34 and @xmath35 preserves @xmath36 of our simulated @xmath0 rays while rejecting @xmath37 of the protons . for an array of more than four telescopes ( @xmath18 ) , the reconstruction process provides a method for rejecting cosmic rays . with @xmath18 the minimum of @xmath20 will generally not be zero . the value of the minimum will depend on how well the optimal source position and impact point account for the image centroid positions in all the n telescopes . events for which the @xmath7 lines are not all precisely parallel to the @xmath38 lines , will have larger minimal values of @xmath20 . as each telescope collects cherenkov light from different parts of the cosmic - ray showers , cutting on the minimized value of @xmath20 provides one more discrimination criterion . by using only the image centroid positions from imaging atmospheric telescopes , the axis of @xmath0-ray showers can be reconstructed for arrays of at least four telescopes . for a four - telescope array , we have verified that the analysis provides reasonable precision . applied to events close to detector threshold , this method can provide a good starting point for more sophisticated and much more computation - demanding methods involving , for example , detailed models of the cherenkov images @xcite . we have identified criteria useful for discriminating against cosmic rays while still using only the first - order image parameters , the image centroids and brightnesses . these criteria could be combined with more sophisticated image analysis , such as producing a global image width from the light distribution across the line connecting the image centroid to the reconstructed source position . since the analysis presented here uses the minimal amount of information from the recorded chenenkov images , the results must provide the lowest analysis energy threshold for a given experiment . this analysis , in comparison to analysis techniques using higher - order image parameters , could result in lower energy thresholds for modern air cherenkov imaging telescope arrays . this technique could also be used in the design of future detectors and suggests that coarser pixels with perhaps smaller light collectors than in hess and veritas could be cost saving and still produce accurate results in future very large arrays . we thank dirk putzfeld for his help and we acknowledge support by the department of energy high energy physics division . 999 chadwick , p. m. , et al . , 1996 , space science rev . , 75 , 153 de naurois , m. et al . , cosmic ray conf . , tsukuba ( 2003 ) , univ . academy press , tokyo , duke , c. & lebohec , s. , http://www.physics.utah.edu/gammaray/grisu ) grindlay , j.e . , et al . , 1975 , apj 201 , 82 hillas , a. m. , 1996 , space science rev . , 75,17 hinton , j. a. , 2004 , new astronomy rev . , 48,331 kohnle , a. , et al . , 1996 , astropart.phys . , 5,119 krennrich , f. , et al . , 1998 , astropart.phys . , 8 , 213 lebohec , s. , et al . , 1998 , nucl.instrum.meth . a416 , 425 pulhofer , g. , et al . 2003 , astropart.phys . , 20,267 weekes , t.c . 2003 , very high - energy gamma ray astronomy , series in astronomy and astrophysics , ed . institute of physics publishing weekes , t.c . 1989 , apj , 342 , 379

the trajectory of a primary @xmath0-ray detected with an array of at least four atmospheric cherenkov imaging telescopes can be reconstructed from the shower image centroid positions and geometrical considerations independent of the primary energy . using only the image centroid positions some cosmic - ray discrimination is also possible . this minimal approach opens the possibility of pushing the analysis threshold to lower values , close to the hardware threshold .

introduction stereoscopy with minimal imaging example application conclusion acknowledgment

type ia supernovae ( sne ia ) are considered to be one of the most precise tools for determining astronomical distances @xcite . because of their high luminosity and relatively small dispersion at the maxima of their bolometric light curves , they are commonly utilized as cosmological `` standardizable candles '' . the accelerated expansion of the universe and the presence of dark energy were discovered through sne ia used as standardizable candles @xcite . the effectiveness of sne ia as distance indicators and standard candles is hampered by the systematic uncertainties related to their explosion mechanism and progenitor systems , and more importantly , the line - of - sight extinction . the distance @xmath4 measured in parsec to a sn is @xmath5 , where @xmath6 and @xmath7 are its apparent and absolute magnitudes at wavelength @xmath8 , and @xmath9 is the extinction . as it is not easy to directly measure @xmath9 , one often measures the color excess ( or reddening ) @xmath10 , where @xmath11 is the extinction in the v - band ( centered around @xmath12 ) . sn reddening is often measured by comparing the observed sn colors to a zero - reddening locus . @xcite ( 1989 ; ccm ) found that the galactic extinction curves ( or extinction laws ) the wavelength dependencies of the extinction can be closely parametrized by the total - to - selective extinction ratio @xmath13 , where the b - band centers around @xmath14 ( also see * ? ? ? * hereafter ftz ) . astronomers often derive @xmath2 for sne ia by fitting the observed @xmath15 with the @xmath2-based ccm formula . once @xmath2 is determined , one can apply the ccm - formula ( or some other parameterizations ) to derive @xmath9 . however , we caution that the ccm- and ftz - parameterizations have been derived for galactic sightlines with @xmath16 , and may not be valid for external galaxies . note that the ccm formula is not even applicable to the large and small magellanic clouds ( lmc , smc ; * ? ? ? sne ia are so rare that nearby sne ia ( @xmath17 ) are detected only about once a decade . sn 2014j , discovered in the nearby starburst galaxy m82 at a distance of @xmath18 @xcite , is the nearest sn ia seen in the last three decades . its proximity offers an unprecedented opportunity to study the extinction and reddening toward a sn ia . the aim of this letter is to derive @xmath2 and @xmath9 by fitting the reddening curve obtained by amanullah et al . ( 2014 ) during the epoch range of @xmath19 $ ] days around its peak brightness ( [ sec : color_curve ] ) using the silicate - graphite model ( [ sec : model ] ) . the results are presented in 3 , discussed in 4 , and summarized in 5 . ( a ) comparison of the color - excess curve of f14 ( red solid line ) with that of a14 ( blue triangles ) . the blue dashed line is the best - fit ftz model of a14 , while the green dotted line is the best - fit two - component model ( i.e. , a combination of interstellar reddening and circumstellar scattering ) for all epochs ( `` csmd all '' ) of sn 2014j of f14 . ( b ) comparison of the color - excess curves of f14 ( red solid line ) and a14 ( blue triangles ) with our model predictions ( magenta dashed line for f14 , and cyan dot - dashed line for a14 ) . ( c ) residuals between our model color - excess @xmath20 for a14 and ( i ) the data points of a14 ( blue triangles ) , ( ii ) the ftz fit of a14 ( blue dashed line ) , ( iii ) the color - excess curve of f14 ( red solid line ) , ( iv ) the csmd fit of f14 ( green dotted line ) , and ( v ) our model color - excess for f14 ( magenta dashed line ) . , width=345 ] various studies have been carried out to determine the @xmath2 value for the sightline toward sn 2014j ( e.g. , see * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? more specifically , based on the uv to near - ir photometry of sn 2014j obtained with the wfc3 filters aboard the _ hubble space telescope _ ( hst ) and ground - based telescopes , @xcite ( 2014 , hereafter a14 ) determined the reddening curve @xmath15 for sn 2014j over a wide wavelength range by comparing the colors of sn 2014j with that of sn 2011fe , a reddening - free sn ia . they derived @xmath21 by fitting the observationally determined @xmath15 with three template extinction laws : an mw - type law as parameterized by ccm or ftz ( see figure [ fig : f14]a ) , a third - order polynomial salt2 law ( see * ? ? ? * ) , and a power - law parameterization @xmath22 which was shown to be a good approximation for multiple scattering scenarios @xcite . we note that a14 shifted the effective wavelengths of the hst / wfc3 filters ( see their figure 3 ) . in the following , we adopt the wavelength - shifted color - excess data of a14 . foley et al . ( 2014 , hereafter f14 ) determined the extinction curve toward sn 2014j at @xmath23 by comparing the uv / near - ir sed of sn 2014j with that of sn 2011fe . as shown in figure [ fig : f14 ] , the color - excess curve of f14 closely resembles that of a14 . the difference between the f14 curve and that of a14 , on average , is @xmath240.19@xmath25 ( see figure [ fig : f14]c ) . using ftz and ccm models , f14 derived @xmath26 and @xmath27 for sn 2014j , respectively . f14 also argued that a two - component model ( csmd ) with both a circumstellar scattering component of @xmath28 and an ftz interstellar reddening component of @xmath29 could best account for the observed properties of sn 2014j . we consider the silicate - graphite grain model that consists of amorphous silicate and graphite @xcite . we adopt the same exponentially cutoff power - law size distribution for both components : @xmath30 for the size range of @xmath31 , where @xmath32 is the spherical radius of the dust , @xmath33 is the number density of h nuclei , @xmath34 is the number density of dust of type @xmath35 with radii in the interval [ @xmath32 , @xmath32@xmath36@xmath37 , @xmath38 and @xmath39 are , respectively , the power index and exponential cutoff size , and @xmath40 is a constant related to the total amount of dust of type @xmath35 . the total extinction at wavelength @xmath8 is given by @xmath41 where the summation is over the two grain types , @xmath42 ( @xmath33 ) is the h column ( number ) density , and @xmath43 is the extinction cross section of grain type @xmath35 of size @xmath32 at wavelength @xmath8 calculated from mie theory using the optical constants of @xcite . for a given set of parameters @xmath38 and @xmath39 , we derive the constant @xmath40 from the abundances of the dust - forming elements . let @xmath44_{\rm ism}}$ ] be the total interstellar abundance of element x ( i.e. , fe , mg , si , o , and c ) relative to h in the interstellar medium ( ism ) of m82 , @xmath44_{\rm gas}}$ ] be the amount of x in the gas phase , @xmath44_{\rm dust}}$ ] be the amount of x contained in dust ( obviously , @xmath44_{\rm dust}}$]=@xmath44_{\rm ism}}-{\left[{\rm x / h}\right]_{\rm gas}}$ ] ) , and @xmath45 be the atomic weight of x. let @xmath46 and @xmath47 , respectively , be the mass density of amorphous silicate and graphite . for a chosen set of dust depletions , we derive @xmath40 from the dust size distributions : @xmath48_{\rm dust}}+ { \mu_{\rm mg}}{\left[{\rm mg / h}\right]_{\rm dust}}+ { \mu_{\rm si}}{\left[{\rm si / h}\right]_{\rm dust } } } { \int da \left(4\pi/3\right ) a^3\,{\rho_{\rm sil}}\ , a^{-\alpha } \exp\left(-a / a_{b}\right ) } \\ & \quad \frac{4\times \mu_{\rm o}{\left[{\rm si / h}\right]_{\rm dust } } } { } ~~ , \\ \end{split}\ ] ] @xmath49_{\rm dust } } } { \int da \left(4\pi/3\right ) a^3\,{\rho_{\rm gra}}\ , a^{-\alpha } \exp\left(-a / a_{b}\right ) } ~~,\ ] ] where we assume a stoichiometric composition of mg@xmath50fe@xmath51sio@xmath52 for amorphous silicate . m82 is a prototypical starburst galaxy , experiencing a major star formation episode in its nuclear region , with strong superwind and sn activity . @xcite obtained the stellar abundances in the nuclear region of m82 , and compared them with those of the hot gas derived from the nuclear x - ray spectra . compared with the solar abundance of @xcite , both the cool stars and the hot gas in m82 suggest a reduction of fe / h by @xmath53 dex ( i.e. , @xmath54_{\rm ism}}\approx 14.1{\,{\rm ppm}}$ ] ) and an overall @xmath55 and @xmath56 enhancement by @xmath240.4 and 0.5 dex , respectively ( i.e. , @xmath57_{\rm ism}}\approx 35.5{\,{\rm ppm}}$ ] , @xmath58_{\rm ism}}\approx 44.7{\,{\rm ppm}}$ ] ) . oxygen is enhanced by @xmath240.3 dex in stars and reduced by @xmath240.2 dex in the hot gas._{\rm dust}}= 4{\left[{\rm si / h}\right]_{\rm dust}}$ ] for a silicate composition of mg@xmath50fe@xmath51sio@xmath52 . ] the stellar abundance of c derived by @xcite is only @xmath241/4 of solar ( i.e. , @xmath59_{\rm ism}}\approx 83.2{\,{\rm ppm}}$ ] ) . similar to the galactic ism , we assume in m82 that fe , mg and si are all locked up in silicate dust ( i.e. , @xmath54_{\rm dust}}\approx{\left[{\rm fe / h}\right]_{\rm ism}}\approx 14.1{\,{\rm ppm}}$ ] , @xmath58_{\rm dust}}\approx{\left[{\rm mg / h}\right]_{\rm ism}}\approx 44.7{\,{\rm ppm}}$ ] , @xmath57_{\rm dust}}\approx{\left[{\rm si / h}\right]_{\rm ism}}\approx 35.5{\,{\rm ppm}}$ ] ; @xcite ) . for carbon , it is less clear . in the galactic ism , a substantial fraction ( @xmath60)_{\rm ism}}\approx{\left[{\rm c / h}\right]_{\odot}}\approx 331{\,{\rm ppm}}$ ] @xcite . ] of the total carbon abundance is in the gas phase ( @xmath59_{\rm gas}}\approx140{\,{\rm ppm}}$ ] , @xcite ) . for carbon dust ( i.e. , graphite ) in m82 , we will consider three cases @xmath59_{\rm dust}}= { \left[{\rm c / h}\right]_{\rm ism}}\approx 83.2{\,{\rm ppm}}$ ] ( i.e. , all c is locked up in dust ) , @xmath59_{\rm dust}}= 1/2 { \left[{\rm c / h}\right]_{\rm ism}}\approx 41.6{\,{\rm ppm}}$ ] , and @xmath59_{\rm dust}}= 0{\,{\rm ppm}}$ ] ( i.e. , all c is in the gas phase and the dust model only consists of amorphous silicate ) . with @xmath59_{\rm dust}}= 41.6{\,{\rm ppm}}$ ] , the model has a silicate - to - graphite mass ratio of @xmath61 . this is close to that of the smc ( @xmath62 , li et al . 2006 ) . to facilitate a direct comparison with the color excesses @xmath63 derived by a14 for sn 2014j , we first calculate @xmath9 from eq.[eq : ext ] and then convert to reddening @xmath64 . for simplicity , the model color - excess has not been convolved with the hst / wfc3 filters . we evaluate the goodness of fitting by @xmath65^{2 } /\sigma(\lambda_j)^2 } { n_{\rm obs } - n_{\rm para } } ~~,\ ] ] where @xmath66 is the observed color excess toward sn 2014j at wavelength @xmath67 derived by a14 , @xmath68 is the uncertainty of @xmath66 , @xmath69 is the model color excess at @xmath67 , @xmath70 is the number of observational data points , and @xmath71 is the number of adjustable parameters . fitting the color excesses between @xmath72 and @xmath73 days from the @xmath74 maximum of sn 2014j ( blue triangles ; a14 ) with the silicate - graphite model ( red solid lines ) , assuming all ( a ) , half ( b ) , and none ( c ) of the carbon elements are depleted in graphite . the dotted lines plot the ccm reddening curves with the @xmath2 values corresponding to that of the models ( a : @xmath2=1.7 , b : @xmath2=1.8 , c : @xmath2=1.7 ) . for comparison , the galactic average of @xmath2=3.1 ( dashed line ) and ccm reddening curve of @xmath2=1.4 ( dot - dashed line ) are also shown . , width=307 ]

type ia supernovae ( sne ia ) are powerful cosmological `` standardizable candles '' and the most precise distance indicators . however , a limiting factor in their use for precision cosmology rests on our ability to correct for the dust extinction toward them . sn 2014j in the starburst galaxy m82 , the closest detected sn ia in three decades , provides unparalleled opportunities to study the dust extinction toward an sn ia . in order to derive the extinction as a function of wavelength , we model the color excesses toward sn 2014j , which are observationally derived over a wide wavelength range in terms of dust models consisting of a mixture of silicate and graphite . the resulting extinction laws steeply rise toward the far ultraviolet , even steeper than that of the small magellanic cloud ( smc ) . we infer a visual extinction of @xmath0 , a reddening of @xmath1 , and a total - to - selective extinction ratio of @xmath2@xmath31.7 , consistent with that previously derived from photometric , spectroscopic , and polarimetric observations . the size distributions of the dust in the interstellar medium toward sn 2014j are skewed toward substantially smaller grains than that of the milky way and the smc .

introduction color-excess curves of sn2014j dust model

the formulation of qcd , or any other field theory , at finite temperature is well - established and several equivalent approaches exist . nevertheless , since finite temperature systems are determined not only by their ground states but also by all the excited states , the study of finite temperature field theory really involves different physics with respect to the zero temperature case and different concepts need to be introduced @xcite . in the existing literature the real - time dynamics is mainly discussed in a perturbative context . however , we are often interested in non - perturbative real - time physics . typical non - perturbative methods for qcd calculation at finite @xmath0 are the operator product expansion and lattice simulations . both these approaches are intrinsically euclidean and limited to static properties . the possibility of extracting real - time information from these euclidean quantities needs to be addressed . due to the lack of experimental data , we do not have enough intuition to guide our approaches . therefore , it is particularly useful to have some exact results which could provide us with insights into such matter . however , demanding exact solutions often implies a sacrifice in the direct phenomenological relevance . in this paper , we address these issues in the u(1 ) chirally symmetric gross - neveu model @xcite in @xmath3 dimensions and in the large-@xmath1 limit : this soluble model allows us to perform exact calculations both for real - time and screening dynamics . this model is often used in the literature @xcite whenever exact results are required and , besides , some relevance to qcd is also desired . in fact , the gross - neveu model has several features in common with qcd . for instance , this model undergoes a chiral phase transition at finite @xmath0 , as it is expected to happen in qcd , providing a good place for a qualitative modeling of the temperature dependence of real - time dynamics . however , some caution is needed : the gross - neveu model is not confining and , therefore , not every physical interpretation of the singularities associated to quark and antiquark can be literarily applied to qcd . more specifically , we solve the real - time bound state and study its properties as functions of temperature . we calculate its mass , its coupling constant , its decay constant and its spectral function . in addition , we carefully characterize the size of this bound state by calculating its on - shell form factor . similarly , we also solve the corresponding screening state by calculating the screening mass , coupling , decay constant , and the screening bethe - salpeter amplitude . the choice of screening observables , and of the corresponding real - time ones , has been suggested from the available lattice calculations @xcite . eventually , we contrast the real - time observables and the corresponding screening ones . while some of the results presented here can be found in the existing literature , and we include them only to make the presentation of our points coherent and self - contained , others are less well - known . among these new , or insufficiently discussed , topics we would like to emphasize the following . + ( 1 ) due to the lack of lorentz invariance at finite @xmath0 a general amplitude @xmath4 has often very different functional dependences on @xmath5 and @xmath6 . an obvious consequence of this non - uniformness is , for example , the non - covariant energy - momentum dispersion relation for elementary excitations . a less obvious but more interesting consequence is related to how conservation laws due to some internal symmetries manifest themselves in various physical processes . the goldberger - treiman relation and the effective charge of the pion are used as illustrations . + ( 2 ) to characterize the structure of a bound state at finite @xmath0 in a physical way we introduce the on - shell form factor in the elastic limit . the peculiarity of the relevant kinematic condition and the existence of an additional form factor are carefully addressed . then we isolate the singularity structure that characterizes the on - shell form factor as a function of the spatial momentum transfer and , therefore , determine the size of the bound state . + ( 3 ) we derive an exact sum rule , which states that the zeroth moment of the pseudoscalar spectral function is temperature independent . + ( 4 ) the screening bethe - salpeter amplitudes in scalar and pseudoscalar channels are calculated explicitly . + ( 5 ) we obtain the asymptotic formula for the screening mass in the high-@xmath0 limit and show that this result demonstrates the failure of dimensional reduction in this model . the reason of this failure is also explained . strictly speaking , the 2 + 1 dimensional gross - neveu model with a continuous symmetry has no finite-@xmath0 phase transition beyond the leading order in @xmath7 , due to the severe infrared singularity associated with massless goldstone bosons . one might question the relevance of our results at leading order in @xmath7 . this problem can be avoided by working in @xmath8 spatial dimensions ( @xmath9 ) , where the infrared catastrophe is absent . since the limit @xmath10 is smooth at the leading order in @xmath7 , our results presented in this paper should be understood as such . we could have also addressed the same questions in the 3 + 1 dimensional nambu - jona - lasinio model @xcite . however , since our present interest is more related to answering conceptual questions rather than finding direct phenomenological applications , we prefer to use the 2 + 1 gross - neveu model because of its renormalizability . moreover , we believe that the qualitative physics associated with the chiral restoration at finite @xmath0 is similar in all these models , independently of their dimensionality . in any case , we shall take notice of those results that are directly linked to the specific dimensionality . the paper is organized as follows . first , we give , in sec . [ realt ] , a brief pedagogical review of how real - time dynamics is formulated at finite @xmath0 . then , in sec . [ gn ] , we introduce the model and collect all those results that are useful in the rest of the paper . the real - time pseudoscalar bound state and its properties are calculated in sec . [ bs ] . in this same section we also discuss the goldberger - treiman relation , the on - shell form factor and derive a sum rule for the spectral function . the screening state is discussed in sec . [ screen ] , where we also contrast minutely the real - time bound state and the screening state . the last subsection of sec . [ screen ] is dedicated to the high temperature limit of the screening state and to the implications that this result has for dimensional reduction . we reserve sec . [ concl ] to a summary of our work and to conclusions . in this section we review the connection between correlation functions , which are the typical output of theoretical calculations , and the real - time response of a system to external perturbations . the following material is most likely well - known to experts @xcite ; nonetheless , since we are specifically concerned with the proper interpretation of euclidean correlation functions , we find useful to include it here as a convenient reminder . an experimentalist perturbs a system with an external probe , and then measures the effect of this perturbation with some detector . in a given theory we describe this external perturbation using an interaction hamiltonian of the form @xmath11 where @xmath12 is the external perturbation , which is switched on at @xmath13 , and couples to the system through the current @xmath14 ( the hat indicates operators ) . the system is described by a state , or ensemble of states of the hamiltonian , and the measurement of the response to the perturbation is performed through another current , which can be the same as the one that couples to the perturbation . for example , we may perturb the system with an external electric field , which couples to the charge density , and measure the resulting change in the same charge density . therefore , we typically measure the change of the expectation value of a given current due to the presence of the perturbation : @xmath15 where @xmath16 is the expectation value of the operator @xmath17 in the state ( or ensemble of states ) describing the system in the presence of the external perturbation , and @xmath18 the corresponding expectation value in the unperturbed system . here and in the following we drop the spatial label for simplicity . if the perturbation is weak , we can expand in the perturbation , and keep only the term linear in the external field . then the linear response of the system to the perturbation is simply proportional to the retarded correlation function : @xmath19 where the retarded correlation is defined as @xmath20\right ) } { \text{tr } \left ( e^{-\hat{h}_0/t}\right ) } \,\ ] ] and @xmath21 , @xmath22 is the unperturbed hamiltonian , and we consider a system in thermal equilibrium at temperature @xmath0 . the physical content of this equation is known as the fluctuation - dissipation theorem , i.e. correlation functions of a system describe not only the correlations ( fluctuations ) of the system in a given channel , but also its linear response ( dissipation ) to weak external perturbations . we say that the system has a ( real - time ) elementary excitation in the channel described by a given current , if we get a resonant ( very large ) response to the corresponding external perturbation at a given frequency . if we rewrite eq . ( [ linres ] ) in frequency - momentum space ( restoring the spatial dependence ) @xmath23 we can see that the appearance of a pole in the correlator as a function of @xmath24 gives this kind of resonant response , which will be called the bound state . it is clear from eq . ( [ freres ] ) that the resonance frequency depends on spatial momentum @xmath25 : @xmath26 . the function @xmath27 is the so - called dispersion relation for the excitation . alternatively , we can use a time - independent external perturbation ( @xmath28 ) , and study the static response of the system . in this case , we are interested in the modification of the system in spatial distribution , e.g. how the effect of the perturbation dies out with distance ( screening ) . in momentum space there can still exist purely imaginary poles in the retarded correlator , but now they are interpreted as pure exponential terms in the decaying response . at zero temperature only , due to euclidean invariance for relativistic theories , there is a one to one correspondence between these poles in the imaginary momentum and the ones in frequency . the resonant frequencies of the system ( poles of the propagator ) are important characteristics of the system , but they are not the only information we get from experiment ( and from the propagator ) . the strength with which the external probe couples to the system is also important . for instance , excitations of the system that do not couple to the external probe may not be relevant . in the propagator , the information about the coupling strength is carried by the residue of the pole at the resonant frequency . a nice example of the concepts we have just reviewed is the classic electron plasma probed through the charge density : @xmath29 . the response of the plasma to a static point charge is just the well - known debye screening @xmath30 and the corresponding pole in the static propagator , @xmath31 is the debye mass : @xmath32 ( @xmath33 and @xmath34 are the electron density and charge , respectively ) . but the plasma also possess real - time excitations , plasmons , with characteristic frequency @xmath35 , where @xmath36 is the electron mass . these excitations are described , near the plasmon pole , by the green s function @xmath37 where @xmath38 is the damping rate which gives a finite width to the plasmon . the frequency - momentum ( or dispersion ) relation for plasmons is @xmath39 . from the preceding example it is quite clear that in general real - time excitations and screening states describe different physics . nevertheless , there is a connection between measurements in minkowski and euclidean space . in general , causality and unitarity make possible the following dispersion relation for the correlation function @xmath40 where the spectral function @xmath41 is proportional to the total cross - section in the given channel . it is important to realize that the dispersion relation at finite @xmath0 is in frequency only and the spatial momentum is treated as a parameter . in the plasmon example , if the plasmon were the only excitation of the system , the spectral density @xmath41 would just be the on - shell condition for the plasmon : @xmath42 the power of the dispersion relation is that the physical content of the correlation is entirely embodied in the spectral function @xmath41 , which determines the correlation in the whole complex plane , apart from a possible additive polynomial in @xmath24 . for instance , if we analytically continue the correlation function to imaginary frequencies , we obtain @xmath43 this analytically continued correlation function calculated at the matsubara frequencies @xmath44 ( for fermions ) is the discrete fourier transform of the correlator calculated in the euclidean region ( finite temperature correlator ) : @xmath45\right ) } { \text{tr } \left ( e^{-\hat{h}_0/t}\right ) } \,\ ] ] where @xmath46 . this formula is valid in the interval @xmath47 , while for @xmath48 outside this interval we can use the periodicity condition , which for fermions reads @xmath49 ( it follows from the periodicity of the trace and the anticommutation property of the fermionic fields ) . vice versa , if we know the correlation function at the discrete set of euclidean points specified by the matsubara frequencies , the analytic continuation @xmath50 provides us with the retarded / advanced correlation for continuous real - time frequency . this continuation is unique only when the expression of the correlation function is explicitly free of singularity in the entire @xmath24-plane except on the real axis and obeys certain convergence property at @xmath51 @xcite . finally , due to the lack of lorentz invariance at finite @xmath0 , a general green s function @xmath52 often has very different functional dependence on @xmath5 and on @xmath6 . for example , @xmath53 strongly depends on the order in which we take the limits @xmath54 and @xmath55 , i.e. @xmath56 is usually different from @xmath57 . ( the innocent looking function @xmath58 in the limit of @xmath59 illustrates the relevant subtlety . ) of course , the physics of the specific process selects which of the two order is relevant . in particular , @xmath60 is intrinsically minkowskian , because it represents a process whose characteristic time is much shorter than the heat - bath response - time . we shall then call this kind of processes fast processes correspondingly , @xmath61 is intrinsically euclidean and the related processes are called slow processes , since the thermal environment has enough time to respond to the external perturbation . as we will see later , conservation laws due to internal symmetries only hold in fast processes . in slow processes the apparent violation of conservation laws should be understood in the sense of the grand canonical ensemble , not as the violation of the fundamental physics laws . in this section we introduce the gross - neveu model @xcite in @xmath3 dimensions , and present its main features at finite temperature . corresponding formulae for the model in @xmath62 dimensions can be found in ref . @xcite . the gross - neveu model is defined by the lagrangian density [ lagr ] @xmath63 , \ ] ] where @xmath64 is a 4-component dirac spinor with the color indices implicit . equivalently , we can write @xmath65 where @xmath66 and @xmath67 are the auxiliary scalar and pseudoscalar bosonic fields , respectively . we study this model in the limit @xmath68 with the coupling constant @xmath69 fixed . this lagrangian is invariant under a continuous @xmath70 chiral symmetry that is dynamically broken at low @xmath0 . the standard approach of the effective potential at finite temperature @xcite yields the critical temperature and the dynamically generated mass . to the leading order in @xmath7 , the effective potential for @xmath66 and @xmath67 fields is given by the one - loop expression @xmath71,\ ] ] where @xmath44 . up to an irrelevant constant the sum over @xmath72 can be done yielding @xmath73\right\}\ , .\ ] ] in this last and the following equations we have used the symmetry of the lagrangian , and rotated the @xmath74 field in the @xmath66 direction making @xmath75 . alternatively , we can read @xmath66 in the following formulae as representing @xmath76 . the integral is divergent , and a high momentum cutoff @xmath77 is required . we make @xmath78 finite by adding a counterterm of the form @xmath79 i.e. we have defined the bare coupling constant @xmath80 , with @xmath81 the renormalized coupling constant . then , the renormalized effective potential is @xmath82\right\ } \ , , \ ] ] where @xmath83 , which is independent of the renormalization scale @xmath84 , will turn out to be the dynamical quark mass at @xmath85 . the quantum theory is determined by the fluctuations around the minimum of the effective potential , which we find from the stationary condition @xmath86\right\}=0\ , .\ ] ] when @xmath87 ( high @xmath0 phase ) , the expression in braces is positive definite , and the only solution to eq . ( [ minpot ] ) is @xmath88 ( symmetric phase ) . the critical @xmath89 is the value of @xmath0 for which the expression in braces is zero with @xmath88 , i.e. @xmath90 . when @xmath91 , the expression in braces is zero for the following value of @xmath92 : @xmath93\ , , \ ] ] which is the absolute minimum of the effective potential , eq . ( [ pot ] ) , while the other stationary point , @xmath94 , is a maximum for these values of @xmath0 . consequently , the symmetry is dynamically broken and the corresponding quark mass is @xmath95 . in fig . [ fig1 ] we plot the dynamically generated mass as a function of temperature , i.e. eq . ( [ dynmass ] ) : it goes exponentially to @xmath96 in the zero temperature limit , while it approaches zero as @xmath97 for @xmath0 approaching @xmath89 from below . for later convenience we define the following function of @xmath0 , which is proportional to the first derivative of @xmath98 with respect to @xmath99 evaluated at @xmath100 , @xmath101=\left\ { \begin{array}{c } 0 \\ \mu\left(1-{t\over t_c}\right)\end{array } \quad\begin{array}{l } \mbox{if $ t\le t_c$ } \\ \mbox{if $ t > t_c$}\end{array}\right.\ , .\ ] ] in this model , the only one - particle irreducible loop graph ( in the large @xmath1 limit ) that is relevant for our purpose is the pseudoscalar bubble graph @xmath102 . in order to extract the bound state and screening states , we need this graph in minkowski space for general @xmath103 . we first use the imaginary - time formalism to perform the loop integral : @xmath104 where @xmath105 , @xmath106 , @xmath107 with @xmath108 , and @xmath109 with @xmath110 . the sum over frequencies can be performed by using standard contour integral techniques @xcite , and we obtain : @xmath111 where @xmath112 . the above equation is infinite and can be made finite by adding the contribution from the counterterm @xmath113 . we then analytically continue @xmath24 into the entire complex plane . for calculating the bound state mass , it is sufficient to have the expression for zero external momentum in the region @xmath114 for the pseudoscalar channel : @xmath115 -{\omega^2\over4}\int_m^\infty de\ , { 2\over 1+e^{e / t}}\,{1\over e^2-\omega^2/4}\right\}\ , . \label{psbbl2}\end{aligned}\ ] ] in the region @xmath116 , which is needed for the spectral function , the bubble develops an imaginary part , and the corresponding formulae are : [ psbblri ] @xmath117 & = & { n\over g^2 } + { n\over\pi}\left\{r(t)+{\omega\over4 } \,\ln\left[{\omega + 2 m \over \omega - 2m}\right ] -{\omega^2\over4 } p\!\!\!\!\!\int_m^\infty \!\ ! de\ , { 2\over 1+e^{e / t}}\,{1\over e^2-\omega^2/4}\right\ } \\ \text{im}\left[i\pi_p(\omega)\right ] & = & { n\over 4}\omega\tanh{\frac{\omega}{4 t } } \ , , \end{aligned}\ ] ] where @xmath118 indicates the principal part of the integral . for studying the screening phenomena we need the pseudoscalar bubble , eq . ( [ bblgr ] ) , in the euclidean region for @xmath28 , i.e. in the static limit . it is convenient to expand the hyperbolic tangent by using the formula @xmath119 ^ 2}\,\ ] ] and perform the @xmath25-integral . the resulting expression is @xmath120\ , , \label{bubble_p}\ ] ] where @xmath121 . to the leading order in @xmath7 the retarded two - point function for pseudoscalar current involves only a sum over a geometric series of the pseudoscalar bubble graph : @xmath122\rangle_t = \frac{i\pi_p(\omega,\bbox{p})}{1-(ig^2/n)\pi_p(\omega,\bbox{p})}\ , , \label{twopoint}\ ] ] where @xmath123 , and we have renormalized the two - point function in such a way that its lowest order coincides with the renormalized bubble . when the above equation has a pole in real frequency @xmath24 at a given spatial momentum @xmath124 , this pole is identified as the elementary excitation or the bound state . in the static limit ( @xmath28 ) , on the other hand , the lowest singularity for @xmath125 , denoted by @xmath126 , corresponds to the screening mass . in this section the real - time dynamics is studied . we first solve the bound state in the pseudoscalar channel , including the bound state mass , with its dispersion relation for small momenta , the coupling and decay constants . the goldberger - treiman relation is examined carefully . we then calculate the on - shell form factor and the charge radius . finally , the spectral function and its sum rule are derived . as discussed in section [ realt ] , a physical excitation in a given channel is signaled by a pole in the corresponding correlation function . therefore , a bound state in the pseudoscalar channel ( there is no elementary pseudoscalar excitation in the lagrangian ) exists , if there is a solution to the equation obtained by equating to zero the denominator of eq . ( [ ps2pf ] ) @xmath127 -{m_\pi^2\over 4}\int_m^\infty de\ , { 2\over 1+e^{e / t}}\,{1\over e^2-m_\pi^2/4}\ , , \label{bseq}\end{aligned}\ ] ] where we have used the expression for the pseudoscalar bubble , eq . ( [ psbbl2 ] ) , valid for @xmath114 . this equation has always the solution @xmath128 , which is lower than the energy of the unbound particles @xmath129 , as long as @xmath91 ( @xmath130 ) . the study of the residue associated with this solution can also help in determining the fate of the bound state when @xmath131 . as long as we are only interested in the bound state mass , the bubble calculated at zero momentum @xmath124 is sufficient . however , it is also instructive to calculate the energy as a function of momentum . in fact the system at finite temperature has a preferred reference frame , the heat bath , and we expect an explicit loss of covariance for the energy - momentum dispersion relation . to this end , we perform the calculation of the bound state energy retaining terms up to the first order in @xmath6 and @xmath5 ( @xmath5 is also small in the limit of small @xmath6 ) . the bubble in this limit can be written as : @xmath132 where we have dropped terms of order @xmath133 , @xmath134 , @xmath135 and higher . it should be emphasized that when obtaining eq . ( [ full - bubble ] ) the bubble is expanded in powers of @xmath6 first , and then the corresponding coefficients are expanded in powers of @xmath5 . again by solving the bound state equation @xmath136 , we find @xmath137 where the function @xmath138 has the following limits : @xmath139 , as it should , and @xmath140 . the fact that @xmath141 implies that the speed of the pion is reduced by the thermal environment , even though the pion is still massless . as stated in section [ realt ] , the bound state coupling constant to a quark - antiquark pair is given by the residue of the full pion propagator on the mass - shell @xmath142 which is also proportional to the residue of the pole of the correlation function , eq . ( [ ps2pf ] ) , at @xmath143 . the dependence of @xmath144 on temperature is shown in fig . [ fig2 ] . when @xmath145 , @xmath146 , and @xmath147 ( zero temperature limit ) . when @xmath148 , @xmath149 , and @xmath150 . in this last case , the coupling @xmath144 approaches zero logarithmically : @xmath151^{-1}$ ] . in contrast , the corresponding coupling in the four dimensional nambu - jona - lasinio model in the chiral limit approaches a finite constant at @xmath152 , and then it jumps to zero for @xmath87 @xcite . the bound state solution in the pseudoscalar channel for @xmath153 has the same energy of two free quarks , and one might wonder about the fate of the bound state in the limit @xmath154 . the coupling constant result gives us a clear answer suggesting that the bound state disappears at the phase transition . more precisely , we should conclude that the pseudoscalar meson decouples from its constituents at the phase transition . we can also introduce a pion decay constant defined as the residue of the axial current correlation function at the pion pole ( this definition at @xmath85 yields the usual expression @xmath155 ) . so we need to calculate the axial - pseudoscalar bubble graph @xmath156 , defined by replacing in eq.([bblgr ] ) one of the @xmath157 with @xmath158 . it is easy to verify that @xmath159 with @xmath160 where @xmath109 is the bosonic matsubara frequency . the above equation can not be naively used to analytically continue @xmath161 into the entire complex plane , since @xmath161 has singularities off the real axis . however , for calculating the pion decay constant we only need @xmath17 at @xmath162 . then the @xmath163-integral can be done , while the sum over @xmath72 can be carried out by the standard contour integral technique , yielding , @xmath164 which now can be trivially continued and then put on mass - shell . hence , the pion decay constant is given by @xmath165 which is valid in the broken phase . since there is no pion state in the symmetric phase , @xmath166 is not defined when @xmath87 . the vanishing of @xmath167 at @xmath153 merely reflects the fact that the chiral symmetry is restored . using the explicit expression of @xmath144 , the above equation can be rewritten as @xmath168 which is the goldberger - treiman relation at finite @xmath0 . notice that @xmath169 identically in the large-@xmath1 limit . if we were not careful in using @xmath170 to define @xmath166 we would have got an expression of @xmath166 not satisfying the goldberger - treiman relation . for example , if we put the axial - pseudoscalar bubble graph on the mass - shell before taking the limit @xmath55 , we would have effectively used @xmath171 in eq . ( [ fpirt ] ) . the physical reason that the pion decay should be regarded as a fast process is that the decay happens instantaneously and the thermal environment does not have enough time to respond . at this point it is appropriate to comment on the fact that it is not trivial how , at finite temperature , the conservation laws that are the consequence of symmetries of a theory manifest themselves in the real - time dynamics . ward identities associated with global symmetries can be straightforwardly generalized from the @xmath85 case to finite @xmath0 in the euclidean formalism . however , it is very subtle how to subsequently analytically continue these relations between euclidean green s functions to corresponding relations between real - time quantities and at which stage one should put the external lines on their mass - shells . therefore , one can not simply assume that euclidean ward identities immediately apply also to the on - shell quantities , as was done in ref . it appears that the validity of the conservation laws depends on the specific mathematical prescriptions , but this dependence is not arbitrary : it reflects the nature of the formulation of field theories at finite @xmath0 in terms of ensemble averages . only in fast processes , where the heat bath does not have time to respond , conservation laws hold explicitly . whereas in slow processes , where a measurement always involves the feedback of the heat bath , conservation laws are no longer manifest . we will encounter an example of such a `` violation '' in the next subsection . at any temperature the on - shell form factor in the elastic limit provides a physical measure of the size of an elementary excitation . in fact , even at finite @xmath0 , we can imagine the following experiment . in the heat - bath frame , we scatter a lepton off the pion . just as in the usual zero temperature scattering , we select those leptons that have scattered off an on - shell pion rather than off something else by using the appropriate kinematic conditions . in our case , this is possible because the pion is massless , while an unbound quark - antiquark is massive . the elasticity here is important , since the kinematics selects out slow processes in which the scattered lepton sees not only the pion but also the thermal cloud around it . the size defined this way provides a snap shot of the elementary excitation immersed in an equilibrated thermal environment . the on - shell form factor has also been studied in ref . however , the author apparently did not realize several complications that arise when defining on - shell form factors at finite @xmath0 , such as the proper on - shell condition at finite @xmath0 and the appearance of additional form factors . on general ground , a typical three - point function with two pion lines and one photon line , as shown in fig . [ fig4 ] , has the following structure in the heat - bath frame @xmath172 + \delta_{\mu 0}g[p^2,p'^2,p\cdot p']\ , . \label{f_mu}\ ] ] the additional form factor @xmath173 $ ] vanishes identically at @xmath85 and it is related to the heat bath . to minimize the environmental effect and hence to emphasize the intrinsic structure of the pion it is natural to use the the invariant function @xmath174 to define its size ; to this end , we can select the spatial components @xmath175 , which receive contribution only from @xmath174 , i.e. we consider a `` magnetic scattering process '' . for low energy elastic scattering , the momentum transfer is @xmath176 , i.e. @xmath177 and @xmath178 . due to the breaking of the explicit lorentz invariance at finite temperature , the on - shell condition for the pion does not imply @xmath179 . therefore , there exist two independent variables in the on - shell form factors @xmath174 and @xmath180 , even in the elastic limit . we choose these two independent variables to be @xmath181 and @xmath182 . then the radius of the pion at finite temperature is defined as @xmath183\over\partial\bbox{q}^2 } \biggr\}\ , .\ ] ] the factor of 4 , rather than the usual 6 , is due to the fact that we are in two spatial dimensions . this definition coincides with the usual definition at zero temperature . the overall normalization of @xmath184 $ ] , which depends on @xmath0 because the charge of the pion gets screened ( @xmath185\le 1 $ ] in general ) , does not affect the charge distribution . the limit of @xmath186 is to ensure that the radius is measured in the heat - bath frame . this condition need not be imposed at zero temperature , due to the lorentz covariance . in case one has trouble to imagine a charge radius for a massless particle , we could have included a small quark mass in the lagrangian such that @xmath187 . this formal process is not really necessary since the limit @xmath188 is smooth . the graph that gives the leading @xmath7 contribution to @xmath184 $ ] is shown in fig . the advantage of probing the pion by a photon instead of a scalar , the @xmath66 , is that meson - pole dominance is avoided , consistent with our intention to emphasize the intrinsic structure . using the pion - quark - antiquark coupling derived in the last section , we have @xmath189[(k - p)^2-m^2][(k - p')^2-m^2]}\ , , \label{ff - define}\ ] ] where @xmath190 is the mass gap . in the elastic limit the momenta are @xmath191 , @xmath192 and @xmath193 with @xmath194 . since we are dealing with a slow process , we should enforce the mass - shell condition at the very beginning . we can take advantage of the facts that the pion is massless and that the scattering is elastic and enforce the on - shell condition by setting all external matsubara frequencies to zero . this step is perfectly legal , because the euclidean point @xmath195 with @xmath196 happens to lie on the real axis and coincides with the point @xmath28 . then , the calculation of the on - shell form factor from eq . ( [ ff - define ] ) involves the following steps : ( 1 ) combine the denominators using the feynman parameter representation ; ( 2 ) carry out the spatial momentum integral ; ( 3 ) eliminate all momentum variables in terms of the two independent variables @xmath197 and @xmath181 ; ( 4 ) extract @xmath184 $ ] from @xmath175 using eq . ( [ f_mu ] ) ; ( 5 ) take the limit @xmath198 . without going through the details , which are tedious but straightforward , we directly give the final result @xmath199= { g^2_\pi(t)\over 2\pi}t\sum_{n=-\infty}^\infty\int_0 ^ 1 d\alpha { 1\over \omega_n^2+m^2+\alpha(1-\alpha)\bbox{q}^2}\ , . \label{singularity}\ ] ] the matsubara frequency sum can be readily carried out , yielding @xmath199= g^2_\pi(t){\tanh(m/2t)\over 4\pi m } \biggl\{1-{\bbox{q}^2\over 12m^2}\bigl[1-{(m / t)\over\sinh(m / t)}\bigr ] \biggr\}+{\cal o}\bigl(\bbox{q}^4\bigr)\ , .\ ] ] using the explicit expression of @xmath200 in eq . ( [ g_pi ] ) we immediately find that the effective charge @xmath201\le 1 $ ] , with the equality sign only at @xmath85 . the fact that @xmath184 $ ] vanishes in the limit @xmath202 merely reflects the fact that the cross section for a lepton to hit the pion also vanishes in the same limit , consistent with the decoupling of the pion from its constituents at the chiral restoration point . as another consistency check , @xmath201 $ ] can also be calculated from the full pion propagator @xmath203 $ ] through the ward identity , in the limit of @xmath204 @xmath205\ , = \,g_\pi^2(t)\ , { \partial\over\partial\bbox{p}^2}s_\pi^{-1}[\omega,\bbox{p}^2 ] \bigg|_{\omega=\omega(\bbox{p})}\ , , \ ] ] which is a consequence of the residual static gauge invariance at finite temperature . in the above equation one must use the full pion propagator in the limit of slow processes , i.e. taking @xmath28 before sending @xmath6 to zero as it was done in eq . ( [ bubble_p ] ) ; this limit differs from the one used to produce eq . ( [ full - bubble ] ) which is defined for fast processes . as we anticipated , the charge conservation that guarantees @xmath206=1 $ ] at @xmath85 no longer holds at finite @xmath0 , due to the well - known charge screening . the effective charge @xmath206 $ ] is the pion charge plus the charge in its thermal cloud . the explicit charge conservation only can be expected in a deeply inelastic process , in which the on - shell condition is enforced at the end of the form factor calculation . of course , the total charges of the test particle and the heat bath is definitely conserved . however , in a slow scattering process , the lepton only sees the charges localized near the test particle , but not those charges far away at the spatial boundary . it would be interesting to calculate directly the thermal - cloud charge distribution induced by a pion . however , this calculation appears to be beyond the scope of the linear response theory . the reason is that , according to the linear response theory , the induced thermal cloud is proportional ( in momentum space ) to the charge distribution of the test particle , and the temperature dependence appears only in the proportionality function , i.e. the retarded correlation function . in our case , the test particle itself has a non - trivial internal structure , which is subject to change under the thermal environment . therefore , there is back - reaction between the induced thermal cloud and the structure of the pion , which is not an external probe . this intricate entangling of the test particle and the thermal environment necessarily prompted us to the calculation of the on - shell form factor . now the radius of the pion in the heat - bath frame can be obtained easily @xmath207\ , . \label{radius}\ ] ] the complete curve of @xmath208 as a function of @xmath0 is displayed in fig . [ fig5 ] , which makes sense only in the symmetry broken phase where the pion exists . at zero temperature @xmath209 , which can also be verified directly from the covariant calculation at @xmath85 , where @xmath210=[4m^2/(-q^2)]^{1/2}\sin^{-1}[(-q^2)/(4m^2-q^2)]^{1/2}$ ] , with @xmath211 . near critical temperature @xmath212 . although the pion s size decreases as temperature increases from zero to the critical point by about factor of 3 , the characteristic size scale of the pion remains to be @xmath213 . this is related to the fact that the pion , once it forms , is always a tightly bound state when @xmath91 . it is also interesting to point out that , in general , the radius @xmath208 , defined through slow processes , can not be simply estimated by the threshold of the triangle graph , which is a standard practice at @xmath85 . in fact , if one did so , one would obtain the wrong estimate @xmath214 , which diverges when @xmath202 . the phase space integral , which produces the factor in the brackets in eq . ( [ radius ] ) , can play an important role at finite temperature . as can be seen clearly from eq . ( [ singularity ] ) , the actual singularity structure of @xmath215 $ ] in @xmath197 is always controlled by the lowest threshold @xmath216 , which stays of order @xmath217 in the entire broken phase . a physical interpretation can be given for eqs . ( [ singularity ] ) and ( [ radius ] ) : the thermal environment tends to wash out coherence beyond the thermal wave length ( @xmath218 ) in slow processes . at this point the physical picture is clear . at low temperature , the pion state is very much the usual pion state with exponentially small correction from the thermal environment . as temperature increases the quark and antiquark become less and less likely to bind together to form the elementary excitation . however , once they bind together , the elementary excitation remains qualitatively similar to its zero temperature counterpart . above the critical temperature the pion completely resolves into its constituents and no longer exists . so the phase structure of the system is strongly reflected from its spectral content in minkowski space . in sharp contrast , the screening bethe - salpeter amplitude defined in the euclidean space does not share this feature , since it is insensitive to the real - time singularity structure , as we shall demonstrate in the next section . nonetheless , the screening bethe - salpeter amplitude indeed provides a qualitative measure of the size of the corresponding bound state , though only in the phase where the bound state can be identified as an elementary excitation . this is due to the fact that the on - shell form factor and the screening bethe - salpeter amplitude have thresholds proportional to @xmath219 . can the result in eq ( [ radius ] ) be an accident of the 2 + 1 dimensional gross - neveu model ? to settle this question we need to study the on - shell form factor in 3 + 1 dimensions , i.e. in the nambu - jona - lasinio model . in 3 + 1 dimensions , we find that the resulting formula is not very different from the one in 2 + 1 dimensions , eq . ( [ singularity ] ) : apart from some trivial factors , which are independent of the momentum transfer , the only modification is that the integrand of the @xmath163-integral is raised to the power 1/2 , as it is obvious from dimensional counting . therefore , the lowest singularity in @xmath197-plane is still given by @xmath220 . although now the form factor requires a cutoff in the matsubara frequency sum , the charge radius is independent of this cutoff . this implies that the charge radius in 3 + 1 dimensions has qualitatively the same behavior as in the 2 + 1 dimensional case . our result , a pion charge radius that remains finite at the critical temperature both in the 2 + 1 gross - neveu model and in the 3 + 1 nambu - jona - lasinio model , is in sharp contradiction with the result found in ref . @xcite , a pion charge radius that diverges at the critical temperature in the nambu - jona - lasinio model . since the result of ref . @xcite is not expressed through a simple formula such as our eqs . ( [ singularity ] ) and ( [ radius ] ) , but is obtained by means of a numerical integration of an expression that involves principal value definitions , we have not been able to pin down the source of this discrepancy . however , it is worth mentioning that schulze used the temporal component of the triangle graph to define the charge radius , which we believe is less desirable due to the contamination from the second form factor @xmath180 associated with the heat bath . it is also not clear at what exact stage he enforced the on - shell condition , and hence it is not clear whether his form factor is defined for fast or slow processes . either of these two points could lead to very different results and , more importantly , to different interpretations in terms of physical measurements . we now calculate the spectral function in the pseudoscalar channel . we give the result for @xmath162 , where the cut contribution near the origin , due to the scattering of the current and thermal particles , vanishes . below @xmath89 we have a contribution from the continuum ( @xmath221 ) , and a pole contribution from the bound state : @xmath222\ , , \ ] ] where the bound state mass @xmath223 , and the coupling @xmath224 is given in eq . ( [ g2pi ] ) . above @xmath89 only the contribution from the continuum survives . the continuum part of the spectral function is related to the real and imaginary parts of the bubble graph , see eqs . ( [ psbblri ] ) , in the following way @xmath225 } { \left(1-(g^2/n)\text{re}\left[i\pi_p(\omega)\right]\right)^2 + \left((g^2/n)\text{im}\left[i\pi_p(\omega)\right]\right)^2}.\ ] ] we show the spectral function at different temperatures in fig . [ fig4 ] . the most remarkable feature is that near the phase transition the shape of the spectral function is very different from its shape at low temperature . a large peak develops just above threshold . the peak actually diverges right at the phase transition , though its integrated strength stays finite . at the same time , as expected , the pole becomes weaker and the threshold goes to zero . these qualitative features are not artifacts of the particular model we are considering here . they actually are the characteristics of a continuous phase transition and critical phenomena . for example , the strong peak near the threshold is a reflection of the fact that the corresponding susceptibility diverges at @xmath89 . these characteristics would still persist to a certain extent when the phase transition is weekly first order . furthermore , as we argued in ref . @xcite , qcd shares all these qualitative features . the strong peak near the origin in the spectral function , when @xmath0 is close to @xmath89 , indicates a resonance - like excitation . this resonance corresponds to a complex pole of the two - point function on the second riemann sheet of the @xmath5-plane , with a very small imaginary part . using explicit formulae in sec . [ gn ] , one can analytically continue the two - point function onto the second riemann sheet . then , it is not hard to find that the real and imaginary parts of the pole in the symmetric phase are proportional to @xmath226 and @xmath227 , respectively , where @xmath228 . as @xmath0 gradually increases from @xmath89 this resonance pole moves almost parallel to the real axis from the origin to the right initially ; and it eventually marches into the first quadrant . when @xmath0 is outside the scaling region , the imaginary part of this complex pole becomes so big such that it does not makes sense anymore to call it a resonance excitation . it appears that the diminishing of the pion strength and the magnification of the quark - antiquark continuum strength in the spectral function near the critical region provides a possible physical picture for the failure of the standard @xmath66-model scenario @xcite in predicting the nature of the chiral - symmetry - restoration phase transition @xcite in the gross - neveu model in 2 + 1 dimensions . to substantiate this statement a detailed investigation is necessary . the reason is that the @xmath66 field in the relevant @xmath66-model does not necessarily correspond to any real - time bound state pole , but more likely to the resonance pole mentioned above . to quantitatively characterize the weakening of the pole and the growing of the peak above threshold we derive an exact sum rule for the temperature dependent part of the zeroth moment of the spectral function . to derive the sum rule we need to find out the asymptotic behaviors of the two - point function in the deep euclidean region and of the spectral function in the large-@xmath5 limit . using the expression in eq . ( [ psbbl ] ) it is easy to obtain the pseudoscalar bubble graph in the euclidean region @xmath229 and its asymptotic form at large-@xmath230 @xmath231 where @xmath232 stands for the thermal average @xmath233 for example , @xmath234 , in the symmetric phase where @xmath235 . the corresponding two - point function , defined in eq . ( [ twopoint ] ) , is @xmath236\biggr\ } \nonumber \\ & + & { \cal o}(q^{-5})\ , . \label{j5_asym}\end{aligned}\ ] ] on the other hand , the spectral function has the following asymptotic form at large @xmath5 , as can be derived from eqs . ( [ psbblri ] ) and ( [ sfcon ] ) , @xmath237 to study the temperature dependence we make the subtraction , following ref . @xcite , @xmath238 with @xmath239 . since the leading term in @xmath240 is independent of @xmath0 , as seen from eq . ( [ rho_asym ] ) , we immediately have @xmath241 . this implies that the zeroth moment of @xmath242 exists . using eq . ( [ j5_asym ] ) , and hence @xmath243 we arrive at the following expression appropriate to derive the sum rule @xmath244 where @xmath245 . taking the limit @xmath246 we get the desired sum rule @xmath247 interchanging the order of the dispersion integral and the @xmath246 limit is allowed since the zeroth moment is finite . we have proved that the very same sum rule also holds in the 1 + 1 dimensional gross - neveu model and qcd @xcite , though we have used a totally different derivation in those cases and the corresponding sum rules would only converge logarithmically . it appears that the temperature dependence of the zeroth moment of the spectral function often shows qualitatively similar behaviors , no matter what is the underlying microscopic physics . another interesting point , which was especially clear in the derivation we used for the 1 + 1 dimensional gross - neveu model and qcd @xcite , is that the zeroth moment of the spectral function could be related , in asymptotically free theories , to expectation values of appropriate operators via the operator product expansion . that result can not be immediately applied to the 2 + 1 gross - neveu model , since this model is not asymptotically free and possesses a nontrivial ultra - violet fixed point . therefore , it would be extremely interesting to investigate whether a formalism similar to the operator product expansion @xcite can be extended to this case . as discussed in section [ realt ] , screening phenomena are associated with responses to time - independent external perturbations . therefore , we only need to consider static green s functions and purely spatial correlation functions . static green s functions are essentially euclidean and hence , in contrast to the time - dependent responses we have discussed in the preceding section , well - suited for the lattice formulation . in this section we study the static two - point and three - point functions that relate to screening masses and screening bethe - salpeter amplitudes , respectively , since these quantities are studied on the lattice @xcite . the pseudoscalar screening mass @xmath248 is the lowest solution to the equation @xmath249 when @xmath130 ( @xmath91 ) , the solution is obvious : @xmath250 . in the high temperature phase @xmath251 , the solution can only be obtained numerically . contrary to the naive expectation that higher than lowest matsubara modes decouple at high @xmath0 , we point out that the frequency sum in eq . ( [ screenp ] ) can not be truncated without introducing an error in the screening mass that is not of order @xmath218 . this fact suggests that dimensional reduction does not take place in this model @xcite . details on this point will be presented later . we can define a coupling constant for the screening state analogous to the one defined for the real - time bound state case @xmath252 then , near the screening - mass pole , the static pseudoscalar propagator behaves like @xmath253 . in fig . [ fig2 ] @xmath254 is plotted in dashed line . even though @xmath255 agrees with @xmath256 at @xmath85 , due to the euclidean invariance , it becomes drastically different from @xmath256 as temperature increases . in particular , @xmath257 does not vanish in the symmetric phase , but rather grows linearly with @xmath0 in the high-@xmath0 limit . in the broken phase @xmath258 has a particularly simple form : @xmath259 $ ] . as in the real - time case , an analogous `` decay constant '' for the `` on - shell '' screening pion can be defined . again it is easy to verify @xmath260 which also vanishes in the symmetric phase . it is interesting to notice that the `` goldberger - treiman '' relation holds exactly in the screening case @xmath261 we believe that this result is not accidental , because the screening state in the static limit can be regarded as a bound state in some 1 + 1 dimensional theory at zero temperature with the same symmetry as the original theory . this euclidean `` goldberger - treiman '' relation is in agreement with the results of ref . @xcite . in order to explicitly see the manifestation of the chiral symmetry also in the screening masses , it is instructive to calculate the scalar screening mass . the scalar bubble graph ( in the static limit ) is easily related to the pseudoscalar counterpart @xmath262\ , .\ ] ] similarly to the pseudoscalar screening mass , the scalar screening mass is found by solving the equation @xmath263 when @xmath91 , @xmath130 and the solution is again easily obtained @xmath264 ( notice that @xmath265 is not a solution in this channel ) . for @xmath87 the dynamical quark mass @xmath266 vanishes , and the scalar screening mass equation becomes identical to the one for the pseudoscalar screening mass . in fig . [ fig6 ] we display numerical results for the screening masses in both channels , pseudoscalar and scalar . in the low temperature phase , chiral symmetry is broken , and the pion is the relevant goldstone boson : this fact is also reflected in the screening mass . in the high temperature phase , chiral symmetry is restored , and we find the expected parity doubling , i.e. the degeneracy of the screening masses in the parity mirrored channels . we already know that the screening state is very different from the real - time bound state . one might , however , naturally ask whether the screening state be somehow related to the resonance excitation near @xmath89 . in particular , one might want to introduce some non - trivial external momentum dependence in the screening state and hope that , by making this external spatial momentum complex , the screening pole might be continued into the resonance pole . while this option is in principle possible , we think that a more relevant question is whether the knowledge of the static screening state as a function of temperature is sufficient by itself , i.e. when we do not have complete knowledge of the analytic structure of the two - point function , to provide enough direct information on the resonance state . we think that the answer to this latter question is unlikely to be affirmative . the reason is that the functional dependences of the two - point function on frequency and spatial momemtum at finite @xmath0 are genuinely different , due to the explicit lack of lorentz invariance . therefore , it is not possible to probe , in general , the full information carried by the frequency variable by varying a spatial component of the momentum . in addition , the following three points also support this general view . first , the screening pole is purely imaginary for any @xmath0 , whereas the resonance pole is complex in general . second , the screening state is well defined for any @xmath0 , whereas the resonance is only identifiable near the critical region . the third point is more specific to the model we have consider : when @xmath267 , the screening mass squared is proportional to @xmath268 , with @xmath269 , whereas the resonance pole is proportional to @xmath268 with logarithmic corrections ( see section iv.d ) . another observable that is calculated on the lattice at finite temperature @xcite is the screening bethe - salpeter amplitude ( sbsa ) . our definition of this amplitude in the pseudoscalar channel is @xmath270 this definition differs from the usual lattice definition by an extra integration over @xmath48 . we perform this extra integration for computational convenience , since it allows us to consider only the zero frequency mode . in terms of physical content of the sbsa , this difference should be irrelevant . to the leading order in @xmath7 , the feynman graphs contributing to the sbsa are those shown in fig . then the fourier transform of @xmath271 has the following form @xmath272 where @xmath273 is the off - shell screening bs amplitude @xmath274_{p=(i\omega\rightarrow 0,p_1,p_2)}\ , .\ ] ] an explicit calculation yields @xmath275 therefore , the complete sbsa is @xmath276 typically , we are interested in the dependence of the sbsa on @xmath277 for large @xmath278 , so that we project on the lowest `` screening state '' . in the limit @xmath279 the @xmath280integral picks out the lowest singularity , @xmath281 , and the sbsa becomes @xmath282 the scale that controls the size of the screening wave function is clearly @xmath283 , as one can show by deforming the integration contour to the imaginary axis . since @xmath284 , and @xmath223 at low temperature , in the low @xmath0 limit this scale is @xmath285 . at high temperature @xmath235 , and @xmath286 , therefore in this limit the scale is set by @xmath287 . as we have argued in the preceding section , a better measure of the spatial distribution of the quark and antiquark inside a meson is provided by its on - shell form factor . the sbsa qualitatively measures the size of the elementary excitation , or bound state , in the entire broken phase . however , the sbsa above the phase transition does not yield any information about the nature of the corresponding ( nonexistent ) real - time bound state . we also calculate the sbsa in the scalar channel using exactly the same procedure we have used in the pseudoscalar channel , and we find very similar results . the relevant mass scale is now given by @xmath288 , which has the same high temperature limit as the scale for the pseudoscalar channel . in the broken phase ( @xmath91 ) , since @xmath289 , and @xmath284 , @xmath290 . this last result is consistent with the fact that there is no binding energy in the scalar channel , and the only screening is due to the thermal mass . now we have all the ingredients to make an explicit comparison between the real - time bound state and the screening state . _ mass _ : when @xmath91 ( the chiral symmetry is spontaneously broken ) , both the real - time and the screening pion masses vanish , @xmath291 , while the scalar mass @xmath292 . however , when @xmath87 ( chirally symmetric phase ) , the real - time pion decouples and the spectral function has contributions only from the quark - antiquark continuum , whereas the screening pion mass @xmath293 becomes degenerate with the screening scalar mass @xmath294 and both of them asymptotically grow linearly with @xmath0 ( see fig . [ fig6 ] ) . _ coupling _ : the real - time coupling of the pion to the quark - antiquark state , @xmath144 , and the corresponding screening coupling , @xmath254 , have very different behaviors as functions of temperature , as it is clearly shown in fig . [ fig2 ] , and they only coincide at @xmath85 . in particular , the screening coupling @xmath254 does not provide any signal for the fact that the real - time excitation decouples when @xmath295 ; in fact @xmath254 does not show any distinctive feature near the critical region . _ decay constant _ : both @xmath166 and @xmath296 start at the same value at @xmath85 and gradually decrease to zero at @xmath153 . the vanishing of @xmath166 and @xmath296 when @xmath297 is due to the fact that the axial current decouples from the pseudoscalar current when the chiral symmetry is restored . this constraint forces a similar behavior for both @xmath166 and @xmath296 , even if their numerical values between @xmath85 and @xmath153 are different . above the phase transition , @xmath167 loses its meaning , and @xmath296 remains zero . _ size _ : as seen from fig . [ fig5 ] , the sizes defined through the on - shell form factor in the elastic limit , @xmath208 and the screening bethe - salpeter amplitude , @xmath298 , have qualitatively similar behaviors in the broken phase . the reason for this similarity is that they share the same singularity structure in low spatial momentum transfer . nevertheless , we find again that , while @xmath208 loses its meaning above @xmath89 , the size defined through the screening bethe - salpeter amplitude does not give us any signal of this disappearance of the real - time state . we believe that most of the above - mentioned qualitative features will also be present in 3 + 1 dimensions , although some of the details presented here are certainly specific to this model . therefore , we can safely conclude that the screening observables , in general , do not necessarily reflect the behavior of the corresponding real - time observables . these two kinds of observables are often qualitatively different . the present model study can also offer us very interesting information about the possibility of dimensional reduction for screening green functions at high temperature . in the symmetric phase , equations ( [ screenp ] ) and ( [ screens ] ) coincide and can be rewritten as @xmath299 where @xmath300 . in the high-@xmath0 limit we can solve the above equation iteratively by setting @xmath301 . in particular , the first two coefficients can be easily found by solving @xmath302 and @xmath303^{3/2 } } \tan^{-1}{c_0\over\sqrt{4(2n-1)^2-c_0 ^ 2 } } \right\ } \ , . \label{c1}\ ] ] therefore , the asymptotic form for the screening masses in the high-@xmath0 limit is , @xmath304 numerically , @xmath305 and @xmath306 . ( [ msasymp ] ) is plotted in dashed line in fig . a salient feature of this asymptotic behavior is that the high temperature limit of the screening mass is @xmath307 , where the numerical value of @xmath308 has contributions from all the matsubara frequency . if dimensional reduction took place in the high-@xmath0 limit , we would find that only the lowest ( or at most a finite number ) modes would contribute to @xmath308 . in fact , any truncation in the frequency sum in eq . ( [ screenp ] ) , and then in eq . ( [ c0 ] ) , would give a different @xmath308 : the resulting error in the screening mass would not be suppressed in the high-@xmath0 limit . the reason for the failure of the dimensional reduction picture is that the 2 + 1 dimensional gross - neveu model lacks the necessary scale hierarchy : there is no elementary bosonic particle whose zero mode could dominate over other modes of order @xmath0 , and the coupling constant can not provide a second lower scale compared to @xmath0 , as it happens in asymptotically free theory @xcite , since this model has a nontrivial ultraviolet fixed point . in fact , the high - momentum behavior of the dimensionless coupling @xmath309 has the form @xmath310 , with @xmath311 ( the value of @xmath312 is scheme dependent ) . this nontrivial fixed point implies that this dimensionless coupling `` runs '' , in the high temperature limit , to a finite value ( since @xmath313 @xcite ) . therefore the effective interaction strength at high @xmath0 between all the matsubara modes ( @xmath314 , once the fields have been appropriately re - scaled ) is strong and its strength in units of @xmath0 becomes independent of temperature ( @xmath315 ) . in this regard the gross - neveu model in 1 + 1 dimensions is very different from the one in 2 + 1 dimensions . the 1 + 1 dimensional model is asymptotically free and @xmath316 runs to zero logarithmically , providing the scale hierarchy between @xmath0 and @xmath317 @xcite . in this model dimensional reduction takes place and screening quantities are well reproduced by the lowest modes that are weakly interacting : the leading contribution to the screening mass is given by the free theory , @xmath318 , the first correction to this non - interacting behavior is of the order @xmath316 and it is correctly reproduced by the reduced theory . since @xmath316 runs to zero the reduced theory becomes exact asymptotically . finally , we should point out that the concept of dimensional reduction that we have discussed in this subsection and in refs . @xcite , is different from the concept of dimensional reduction employed for predicting the nature of phase transitions at finite @xmath0 in the @xmath66-model scenario @xcite . our criterion for dimensional reduction is more stringent and applies to a different temperature regime . we require that the static green s functions of the original theory at high-@xmath0 limit be matched by the corresponding green s functions in the reduced theory to a given accuracy , whereas in the @xmath66-model scenario one requires that the reduced theory be able to match the thermodynamical singularities near the critical region and hence give the same critical exponents . therefore , the success / failure of one of the two dimensional reduction pictures does not directly implies the success / failure of the other . we have shown in a specific example , the gross - neveu model in 2 + 1 dimensions , that the direct connection between the real - time bound state and the corresponding screening state is lost at finite temperature in general , especially near and above phase transition . in particular , the real - time pion disappears after the phase transition , as it is signaled by the vanishing of its bounding energy and of the coupling to its constituents @xmath144 . in spite of this , the corresponding screening state is still strongly `` bound '' in the high temperature limit , and the screening coupling @xmath258 grows linearly with @xmath0 . the screening mass approaches the asymptotic value of @xmath319 , which is much smaller than the sum of the masses of two noninteracting fermions , @xmath320 . moreover , the screening bethe - salpeter amplitude yields sizes of order @xmath218 both in the scalar and pseudoscalar channels at high @xmath0 , when the real - time bound state no longer exists . therefore , no relevant information about the nature of the real - time pion can be inferred from this screening amplitude . on the contrary , we can give a natural definition of the size of the real - time pion ( below the phase transition ) by means of the on - shell form factor in the elastic limit , since the on - shell form factor has direct information of the relevant coupling @xmath144 . furthermore , we show that the associated size can not be estimated solely from simple threshold considerations , reflecting the fact that the thermal environment tends to wash out coherence beyond thermal wavelength in slow ( compared to the equilibration time of the thermal bath ) processes . we have also computed the exact pseudoscalar spectral function of this model . the most important features of this spectral function are the diminishing of the strength of the pion pole and the magnification of the strength of the quark - antiquark continuum near the phase transition . in addition , we have derived an exact sum rule for this spectral function : the integrated strength of the spectral function is independent of @xmath0 . many of the features of the 2 + 1 dimensional gross - neveu model that we have listed above are dominated by the chiral symmetry and should have their analogues in the real world . for instance , very similar results were found in the 1 + 1 dimensional gross - neveu model @xcite , and in the nambu - jona - lasinio model @xcite in 3 + 1 dimensions . on the other hand , certain specific aspects certainly depend on the details of the theory . for example , we find that the 2 + 1 dimensional gross - neveu model does not undergo dimensional reduction in the high-@xmath0 limit . but the possibility of using a lower dimensional theory to describe static correlations of the original theory when @xmath0 is large is strictly tied to the short distance behavior , which is very different in this model from the one in qcd @xcite . finally , we also find that it is instructive to classify physical processes according to their characteristic time scales relative to the time scale required by the thermal environment to respond to an external perturbation . we have explicitly illustrated , in the cases of the goldberger - treiman relation and of the effective charge of the pion , that conservation laws due to internal symmetries are manifest only in fast processes but not in slow ones . it is our pleasure to thank a. kocic and j. kogut for their valuable comments on the manuscript , for bringing several relevant references to our attention and for very instructive discussions on topics related to this work . we would also like to thank t. hatsuda and a. ukawa for several useful discussions . one of us ( ml ) gratefully thanks the department of energy s institute for nuclear theory at the university of washington , where this work was initiated , for their hospitality and partial support during the workshop on `` phenomenology and lattice qcd '' . this work was supported in part by funds provided by the u.s . department of energy ( doe ) under contract number de - fg06 - 88er40427 and cooperative agreement de - fc02 - 94er40818 . t. a. degrand and c. e. detar , phys . d * 34 * , 2469 ( 1986 ) ; c. e. detar and j. kogut , phys . lett . * 59 * , 399 ( 1987 ) ; s. gottlieb _ et al . * 59 * , 1881 ( 1987 ) ; y. koike , m. fukugita and a. ukawa , phys . lett . b * 213 * , 497 ( 1988 ) ; k. m. bitar _ et al . _ , d * 43 * , 2396 ( 1991 ) ; k. d. born _ et al . lett . * 67 * , 302 ( 1991 ) ; c. bernard _ et al . _ , phys . 68 * , 2125 ( 1992 ) ; t. h. hansson and i. zahed , nucl . * b374 * , 227 ( 1992 ) ; v. koch , e. v. shuryak , g. e. brown and a. d. jackson , phys . d * 46 * , 3169 ( 1992 ) ; s. schramm and m. -c . chu , phys . d * 48 * , 2279 ( 1993 ) .

we discuss the interpretation of euclidean correlation functions at finite temperature ( @xmath0 ) and their relationship with the corresponding real - time green s functions . the soluble 2 + 1 dimensional gross - neveu model in the large-@xmath1 limit is used throughout as a working example . first , the real - time bound state , identified as an elementary excitation at finite @xmath0 , is solved . the bound state mass , the dispersion relation at low momenta , the coupling constant and decay constant are calculated . to characterize the structure of the bound state the on - shell form factor is carefully introduced and calculated . then we examine the corresponding screening state and contrast the screening mass , coupling constant , decay constant and the screening bethe - salpeter amplitude with the real - time quantities . we find that , although they can be used as qualitative indicators in the low-@xmath0 regime , the screening states at finite @xmath0 in general do not reflect the properties of the corresponding real - time bound states . besides , other relevant issues , such as the subtlety of the real - time manifestation of conservation laws due to some internal symmetries at @xmath2 , the temperature dependence of the pseudoscalar spectral function and its sum rule , and the high-@xmath0 limit of the screening state and its implications to the dimensional reduction , are also discussed in detail .

introduction general interpretation the gross-neveu model real-time dynamics screening phenomena conclusion

a quandle , which was introduced by joyce @xcite , is an algebraic object whose axioms are motivated by knot theory and conjugation in a group . in @xcite , carter , jelsovsky , kamada , langford and saito introduced a quandle homology theory , and they defined the quandle cocycle invariants for classical knots and surface knots . the quandle homology is defined as the homology of a chain complex generated by cubes whose edges are labeled by elements of a quandle . on the other hand the group homology is defined as the homology of a chain complex generated by tetrahedra whose edges are labeled by elements of a group . so it is natural to ask a relation between quandle homology and group homology . this question also arises from the fact that the quandle cocycle invariants were defined as an analogue of the dijkgraaf - witten invariants , which are defined using group cocycles . in @xcite , the authors defined a simplicial version of quandle homology and constructed a homomorphism from the usual quandle homology to the simplicial quandle homology . the important point is that this homomorphism gives a triangulation of a knot complement in algebraic fashion . this enables us to relate the quandle homology to the topology of knot complements . in this paper , we apply the results of @xcite to construct quandle cocycles from group cocycles . first , we demonstrate how to give a quandle cocycle of the dihedral quandle @xmath1 from a group cocycle of the cyclic group @xmath2 for any integer @xmath0 in section [ sec : dihedral ] . we will show that a generator of @xmath6 gives rise to a non - trivial quandle 3-cocycle of @xmath7 . when @xmath3 is an odd prime , since @xmath4 , our quandle 3-cocycle is equal to a constant multiple of the mochizuki 3-cocycle @xcite up to coboundary . then we generalize the construction to wider classes of quandles . let @xmath8 be a group and @xmath9 be an element of @xmath8 , then the set @xmath10 forms a quandle by conjugation . ( it is known by joyce @xcite that any faithful homogeneous quandle has such a presentation . ) when some obstruction in second cohomology vanishes , we will construct a quandle cocycle of @xmath11 from a group cocycle of @xmath8 . dually , we relate the quandle cycle associated with an arc and region coloring ( shadow coloring ) of a knot @xmath5 to a group cycle represented by a cyclic branched covering branched along @xmath5 . let @xmath12 be a diagram of @xmath5 . we can define the notion of arc and region colorings of @xmath12 by a quandle @xmath11 . a pair of an arc and a region coloring is called a _ shadow coloring_. we can associate a cycle of a quandle homology group to a shadow coloring of @xmath12 . using the homomorphism constructed in @xcite , we will construct a group cycle of @xmath8 represented by a cyclic branched covering branched along @xmath5 . this reveals a close relationship between the shadow cocycle invariant of a knot and the dijkgraaf - witten invariant of the cyclic branched cover . we remark that hatakenaka and nosaka defined an invariant of 3-manifolds called the _ 4-fold symmetric quandle homotopy invariant _ @xcite , based on the fact that any 3-manifold can be represented as a 4-fold simple branched covering of @xmath13 along a link . as an application , they showed that the shadow cocycle invariant of a link for the mochizuki @xmath14-cocycle is equal to a scalar multiple of the dijkgraaf - witten invariant of the double branched cover along the link . this paper is organized as follows . in section [ sec : group_homology ] , we recall the definition of group homology and show how to represent a group cycle by a triangulation with a labeling of its @xmath15-simplices . we give a presentation of the fundamental group of a cyclic branched covering branched along a knot in section [ sec : pres_of_cyc_bra_cov ] , which is independent from the other sections . in section [ sec : cyc_rep_by_cyc_bra_cov ] , we construct a group cycle represented by a cyclic branched cover . we recall the definition of quandles and their homology theory in section [ sec : quandle_and_homology ] . we review some results from @xcite in section [ sec : main_map ] and apply them to construct quandle cocycles of the dihedral quandle @xmath1 in section [ sec : dihedral ] . the reader who is interested in the form of the @xmath14-cocycle , consult ( [ eq : the_cocycle ] ) ( and ( [ eq : formula_of_d ] ) ) . we will generalize the construction to wider classes of quandles in section [ sec : gen_const ] . in section [ sec : quandle_cyc_as_cyc_bra_cov ] , we construct a group cycle represented by a cyclic branched covering from the quandle cycle associated with a shadow coloring . the reader who is only interested in the construction of quandle cocycles from group cocycles may skip sections [ sec : group_homology ] [ sec : cyc_rep_by_cyc_bra_cov ] except [ subsec : group_homology ] . * acknowledgement * the author thanks takefumi nosaka for useful discussions . he also thanks the referees for helpful comments to improve the exposition . the author is supported by jsps research fellowships for young scientists . in this section , we collect basic facts on group homology . in [ subsec : group_homology ] , we review the definition of group homology . we refer to @xcite for details . the material discussed in [ subsec : cycles_represented_by_triangulations ] and [ subsec : group_cycles_and_representations ] was developed in @xcite . let @xmath8 be a group . let @xmath16 be the free @xmath17$]-module generated by @xmath18-tuples @xmath19 $ ] of elements of @xmath8 . define the boundary map @xmath20 by @xmath21 ) = & g_1[g_2|\dots|g_n ] \\ & + \sum_{i=1}^{n-1}(-1)^i [ g_1|\dots|g_ig_{i+1}|\dots|g_n ] + ( -1)^n[g_1|\dots|g_{n-1 } ] . \end{split}\ ] ] we remark that the chain complex @xmath22 is acyclic , where @xmath23 \to \mathbb{z}$ ] is the augmentation map . so the chain complex @xmath24 gives a free resolution of @xmath25 . let @xmath26 be a right @xmath17$]-module . the homology of @xmath27 } c_n(g)$ ] is called the _ group homology _ of @xmath26 and denoted by @xmath28 . in other words , @xmath29}(m , \mathbb{z})$ ] . let @xmath30 be the free @xmath25-module generated by @xmath31 . then @xmath30 is a left @xmath17$]-module by the action @xmath32 . define the boundary operator of @xmath30 by @xmath33 @xmath24 and @xmath34 are isomorphic as chain complexes . in fact , the following correspondence gives an isomorphism : @xmath35 & \leftrightarrow ( 1,g_1,g_1g_2 , \dots , g_1 \cdots g_n ) \\ ( \quad g_0[g_0^{-1}g_1 | g_1^{-1}g_2| \dots | g_{n-1}^{-1 } g_n ] & \leftrightarrow ( g_0,\dots , g_n ) \quad ) \end{split}\ ] ] the notation using @xmath36 is called _ homogeneous _ and the one using @xmath37 $ ] is called _ inhomogeneous_. factoring out @xmath16 by the degenerate subcomplex , that is generated by @xmath37 $ ] such that @xmath38 for some @xmath39 , we obtain the _ normalized _ chain complex and its homology group . it is known that the group homology using the normalized chain complex coincides with the homology using the unnormalized one . in the homogeneous notation , we factor out @xmath30 by the subcomplex generated by @xmath40 such that @xmath41 for some @xmath39 to define the normalized chain complex . for a left @xmath17$]-module @xmath42 , the group cohomology @xmath43 is defined as the cohomology of the cochain complex @xmath44}(c_n(g),n)$ ] . let @xmath45 be an abelian group . a cocycle of @xmath46 in the homogeneous notation is a function @xmath47 satisfying the following conditions : 1 . @xmath48 , 2 . @xmath49 for any @xmath50 . ( left invariance ) if @xmath51 also satisfies 1 . @xmath52 if @xmath53 for some @xmath39 , then @xmath51 is a normalized @xmath18-cocycle . we can show that any @xmath18-cocycle is cohomologous to a normalized @xmath18-cocycle . let @xmath54 be an @xmath18-dimensional simplex . we label the vertices of @xmath54 by @xmath55 . a face of @xmath54 is presented by a subset of @xmath56 . let @xmath57 be the face spanned by @xmath58 with a vertex ordering given by @xmath59 . any face inherits a vertex ordering from the vertex ordering of @xmath54 i.e. @xmath57 with @xmath60 . let @xmath61 be a cw - complex obtained by gluing a finite number of @xmath18-dimensional simplices along their @xmath62-dimensional faces in pairs by simplicial homeomorphisms . we denote the @xmath63-skeleton of @xmath61 by @xmath64 . we assume that the gluing maps preserve the vertex orderings of the faces . then @xmath65 is homeomorphic to a topological @xmath18-manifold ( not orientable in general ) . when @xmath65 is oriented , we call @xmath61 an _ ordered @xmath18-cycle_. consider an @xmath18-cycle @xmath66 of @xmath67 . then @xmath66 is represented by a sum @xmath68\ ] ] where @xmath69 and @xmath70 . for each @xmath71 $ ] , take an @xmath18-simplex @xmath72 . then label the edge @xmath73 of @xmath72 by @xmath74 for @xmath75 . in particular , we have : @xmath76 we denote the label of @xmath77 by @xmath78 . for @xmath79 , label the oriented edge @xmath77 by @xmath80 . for any 2-dimensional face @xmath81 , we have @xmath82 . rewriting them in the homogeneous notation , we assign labels to the vertices of @xmath72 as @xmath83 up to the left action of @xmath8 ( figure [ fig : labeling ] ) . since @xmath84 , @xmath62-dimensional faces cancel in pairs . gluing @xmath72 s along their faces according to such pairings , we obtain an @xmath18-cycle @xmath61 . at any @xmath62-simplex of @xmath61 , there exist exactly two adjacent @xmath18-simplices . the labelings of the @xmath62-simplex derived from these two @xmath18-simplices coincide . thus we have a well - defined labeling of @xmath15-simplices @xmath85 satisfying , 1 . [ item : circle ] @xmath82 for any 2-dimensional face @xmath81 , 2 . [ item : inverse ] @xmath86 . we call a labeling of @xmath15-simplices satisfying the conditions ( [ item : circle ] ) and ( [ item : inverse ] ) _ @xmath8-valued @xmath15-cocycle_. orient @xmath72 positive if @xmath87 and negative if @xmath88 . since these orientations agree on face pairings , thus we have an orientation on @xmath61 . therefore @xmath61 is an ordered @xmath18-cycle with a @xmath8-valued @xmath15-cocycle @xmath89 . in general , @xmath61 may not be connected , but we assume that @xmath61 is connected because we can treat each connected component separately in our arguments . conversely any ordered @xmath18-cycle @xmath61 with a @xmath8-valued @xmath15-cocycle @xmath89 represents an @xmath18-cycle of @xmath67 . suppose a cycle @xmath90 is represented by an ordered @xmath18-cycle @xmath61 with a @xmath8-valued @xmath15-cocycle @xmath89 . then @xmath66 induces a homomorphism from @xmath91 to @xmath8 as follows . let @xmath92 be the universal covering of @xmath61 and @xmath93 be the covering map . then the simplices of @xmath61 lift to simplices of @xmath92 and each lift has an induced vertex ordering compatible with adjacent @xmath18-simplices . the @xmath8-valued @xmath15-cocycle @xmath89 of @xmath61 induces a @xmath8-valued @xmath15-cocycle of @xmath92 . consider a `` fundamental domain '' of @xmath61 , that is a contractible subcomplex @xmath12 of @xmath92 such that * @xmath94 , * @xmath95 , for any @xmath96 , where we regard @xmath91 as the group of deck transformations . by definition , the number of @xmath18-simplices in @xmath12 coincides with the number of @xmath18-simplices in @xmath61 , in particular finite . we fix a base point @xmath97 in the interior of @xmath12 . each @xmath62-simplex on @xmath98 is glued to another @xmath62-simplex on @xmath98 . we denote such pair of faces by @xmath99 . let @xmath100 be a path in @xmath61 which starts at @xmath101 , traverses @xmath102 in the direction from @xmath103 to @xmath104 and ends at @xmath105 . these paths form a system of generators of the fundamental group @xmath106 . the relations are given at any @xmath107-simplices , around which there are a finite number of @xmath102 s . fix an @xmath18-simplex @xmath54 in @xmath12 and a labeling of vertices of @xmath54 derived from @xmath89 . then @xmath18-simplices adjacent to @xmath54 inherit labelings of vertices from @xmath89 . in this way , all vertices of @xmath12 are labeled by elements of @xmath8 . now consider the labeling of vertices of @xmath108 and @xmath109 . since these reduce to the same labeling of edges , they coincide up to left multiplication . therefore there exists an element of @xmath8 which sends the labeling of vertices of @xmath109 to the one of @xmath108 . denote the element by @xmath110 . this @xmath111 induces a homomorphism @xmath112 . conversely if we have an ordered @xmath18-cycle @xmath61 and a homomorphism @xmath112 , we can construct a @xmath8-valued @xmath15-cocycle @xmath89 and then a cycle of @xmath67 up to boundary as follows . since @xmath111 induces a map @xmath113 , we obtain a labeling of @xmath15-simplices @xmath89 of @xmath61 . this gives rise to a @xmath8-valued @xmath15-cocycle @xmath89 and a homology class in @xmath114 . the @xmath8-valued @xmath15-cocycle @xmath89 is well - defined up to the coboundary action . a map @xmath115 acts on a @xmath8-valued @xmath15-cocycle @xmath89 as a coboundary action by @xmath116 we can show that the homology class obtained from @xmath89 does not change under the coboundary action . for any @xmath50 , the cycle corresponding to the representation @xmath117 is obtained from @xmath89 by the coboundary action by @xmath118 . as a result , the homology class obtained from @xmath111 depends only on the conjugacy class of @xmath111 . for a closed oriented @xmath18-manifold @xmath26 and a representation @xmath119 , we have a homology class defined by the image of the fundamental class @xmath120 $ ] under the map @xmath121 . when @xmath26 is homeomorphic to an ordered @xmath18-cycle @xmath61 , the homology class is represented by a @xmath8-valued @xmath15-cocycle @xmath89 of @xmath61 associated with @xmath111 . in this situation , we say that the homology class defined by @xmath26 and @xmath111 is _ represented _ by @xmath61 and @xmath89 . in this section , we give a presentation of the fundamental group of a cyclic branched covering from the wirtinger presentation . let @xmath5 be a knot in @xmath13 and @xmath12 be a diagram of @xmath5 . then @xmath122 is presented by generators and relations , called the wirtinger presentation . let @xmath123 be the generators of the wirtinger presentation , that correspond to the arcs of @xmath12 . each crossing ( figure [ fig : wirtinger_relation ] ) gives rise to a relation @xmath124 . for any integer @xmath125 , let @xmath126 be the @xmath127-fold cyclic covering of @xmath5 , i.e. the manifold corresponding to the kernel of @xmath128 putting back the knot @xmath5 to @xmath126 , we obtain the @xmath127-fold cyclic branched covering @xmath129 of @xmath5 . [ prop : presentation_of_pi_one ] @xmath130 has the following presentation : @xmath131 the inclusion map @xmath132 is given by @xmath133 if we take appropriate base points . by adding a relation @xmath134 , we obtain a presentation of @xmath135 . a method for obtaining a presentation of the fundamental group of a branched covering is given in @xcite . but we give a proof here because some techniques will be used to construct a group cycle represented by @xmath129 later . first , we construct a handle decomposition of the knot complement associated with the wirtinger presentation ( figure [ fig : trefoil ] ) . then we lift the handle decomposition to a handle decomposition of @xmath126 ( figure [ fig : lifted_splitting ] ) . after that , we read the relations given by attaching 2-handles . let @xmath136 be a regular neighborhood of @xmath5 . we shall give a handle decomposition of @xmath137 . we represent @xmath13 by the one point compactification of @xmath138 . let @xmath139 and @xmath140 . we denote the equatorial sphere of @xmath13 by @xmath141 . put @xmath5 in a position such that the projection to @xmath142 has only double points . let @xmath18 be the number of the crossings of this projection . we deform @xmath5 in the @xmath143-direction so that @xmath5 intersects @xmath142 at @xmath144 points and each of @xmath145 consists of @xmath18 arcs . we call @xmath146 over - crossing arcs and @xmath147 under - crossing arcs . index the arcs of @xmath148 by @xmath100 @xmath149 . now @xmath150 is homeomorphic to a handlebody of genus @xmath18 ( see figure [ fig : trefoil ] ) . projecting the over - crossing arcs to @xmath142 , we obtain a meridian disk system of the handlebody . we denote the meridian disk corresponding to @xmath100 by @xmath151 . for each under - crossing arc , attach a 2-handle @xmath152 along @xmath153 to @xmath154 ( see figure [ fig : trefoil ] ) . then the resulting manifold is homeomorphic to @xmath155 where @xmath156 is a 3-ball . attaching @xmath156 along the boundary to @xmath157 , the resulting manifold is homeomorphic to @xmath137 . so we have a handle decomposition of @xmath137 into one 0-handle , @xmath18 1-handles , @xmath18 2-handles and one 3-handle . in this handle decomposition , 1-handles correspond to the wirtinger generators and 2-handles to the wirtinger relations . we denote the set of @xmath39-handles by @xmath158 and @xmath159 . next we consider the preimage of @xmath160 in @xmath126 . cut @xmath160 along the meridian disks @xmath151 , and denote the resulting manifold by @xmath161 , which is homeomorphic to a 3-ball . let @xmath105 be a point in @xmath162 . we take a loop in @xmath160 which starts at @xmath105 , then intersects the meridian disk @xmath151 and ends at @xmath105 . orient the loop so that it corresponds to the generator of @xmath163 . by abuse of notation , we also denote this loop by @xmath100 . take a lift @xmath164 of @xmath161 and denote the preimage of @xmath165 in @xmath166 by @xmath167 . then there exists a unique lift @xmath168 of the loop @xmath100 starting at @xmath167 . since @xmath100 corresponds to the generator of @xmath169 , @xmath168 ends at another lift of @xmath161 , we denote it by @xmath170 . similarly the lift of @xmath100 starting at @xmath170 ends at another lift , we denote it by @xmath171 . continuing this , we see that @xmath172 and all lifts of @xmath161 will appear . therefore the preimage of @xmath160 is decomposed into @xmath127 3-balls @xmath173 . the intersection of @xmath174 and @xmath175 consists of @xmath18 disks each of which is a lift of a meridian disk @xmath151 . we denote this lifted disk by @xmath176 ( figure [ fig : lifted_splitting ] ) . it is easy to check that @xmath177 forms a meridian disk system of the preimage of @xmath160 . denote the lift of @xmath100 starting at @xmath174 ( ending at @xmath175 ) by @xmath178 . then the generator of @xmath179 corresponding to the meridian disk @xmath176 is given by @xmath180 we denote this element by @xmath181 . to give a simple presentation of @xmath179 , we add extra generators @xmath182 corresponding to @xmath183 respectively and relations @xmath184 . finally we consider the relations given by the lifts of 2-handles . we see that the relation @xmath124 lifts to @xmath185 see figure [ fig : lifted_splitting ] . since the generator @xmath181 is represented by ( [ eq : generator_in_cyclic_cover ] ) , the inclusion map @xmath186 is given by @xmath187 . this proves the second statement . by adding a 2-handle to @xmath126 along @xmath188 and capping off the resulting sphere , we obtain a manifold homeomorphic to the cyclic branched covering @xmath129 . therefore a presentation of the cyclic branched covering is obtained by adding a relation @xmath134 . for a representation @xmath189 , we have the restriction map @xmath190 given by @xmath191 if @xmath192 , it reduces to a representation @xmath193 and there is a group cycle given by @xmath194 and @xmath195 . in this section , we construct an explicit ordered @xmath14-cycle and its @xmath8-valued @xmath15-cocycle representing the homology class given by @xmath194 and @xmath195 . first we give a triangulation of @xmath137 associated with the wirtinger presentation . ll ( a ) & ( b ) + & + & + ( c ) & ( d ) + & + ( e ) & + let @xmath5 be a knot and fix a diagram of @xmath5 . as in the proof of proposition [ prop : presentation_of_pi_one ] , we define @xmath196 and give a heegaard splitting of @xmath154 with the meridian disks @xmath151 . cutting the handlebody @xmath154 along the meridian disks @xmath151 , the result is a ball with @xmath144 2-cells on the boundary . we denote the resulting 3-ball by @xmath161 and the pair of 2-cells corresponding to @xmath151 by @xmath108 and @xmath104 so that the wirtinger generator corresponding to @xmath100 runs from @xmath108 to @xmath109 ( figure [ fig : cut_heegaard](a ) ) . now consider the attaching regions of the two handles corresponding to under - crossing arcs , which consist of annuli @xmath197 on @xmath198 . each annulus is divided by @xmath151 s into 4 rectangles on @xmath199 . we define a graph on @xmath199 with vertices consisting of @xmath99 and edges consisting of these rectangles ( figure [ fig : cut_heegaard](b ) ) . each vertex of this graph has valency at least four . we can make all vertices of the graph into trivalent vertices ( figure [ fig : cut_heegaard](c ) ) by adding extra @xmath15-handles and @xmath200-handles ( stabilizations of the heegaard splitting , see figure [ fig : stabilization ] ) . we denote the new vertices by @xmath201 , @xmath202 , ... which originally belonged to @xmath99 . the dual of the graph gives a triangulation of @xmath199 ( figures [ fig : cut_heegaard](d ) and [ fig : cut_heegaard](e ) ) . by abuse of notation , we denote the triangles dual to the vertices @xmath201 , @xmath202 , ... by the same symbols . taking a cone from an interior point of @xmath161 , we obtain a triangulation @xmath61 of @xmath161 into @xmath203 tetrahedra . regluing the triangles @xmath204 , @xmath205 , ... to @xmath206 , @xmath207 , ... , we obtain a triangulation of @xmath137 into @xmath203 tetrahedra , which was explained in @xcite . we remark that this is not a triangulation in the usual sense : it is not a simplicial complex , moreover the link of some @xmath208-simplex is not homeomorphic to the @xmath200-sphere . actually there exists only three @xmath208-simplices , one is the cone point in @xmath161 ( north pole in @xcite ) , second is the @xmath208-simplex corresponding to the complementary regions of the diagram ( south pole ) , the last one is a @xmath208-simplex whose small neighborhood is homeomorphic to the cone over the torus @xmath209 . using this triangulation , we construct a triangulation of the cyclic branched covering @xmath129 . let @xmath210 be @xmath127 copies of @xmath161 and @xmath211 be the triangulation of @xmath174 we have constructed . we denote the triangles @xmath201 , @xmath202 , ... on @xmath212 by @xmath213 , @xmath214 , ... respectively . by abuse of notation , we regard @xmath213 , @xmath214 , ... simply as @xmath215 . glue @xmath211 s along their boundary triangles by the following identification maps @xmath216 denote this triangulation by @xmath217 . we define an ordering of each tetrahedron by assigning @xmath208 to the interior vertex of @xmath174 , @xmath15 to the vertex corresponding to complementary region of the diagram , @xmath200 to the vertex corresponding to under - crossing arc , @xmath14 to the vertex corresponding to over - crossing arc , respectively ( figure [ fig : ordering ] ) . the orderings are compatible under the gluing maps . so @xmath217 is an ordered @xmath14-cycle . here @xmath217 is a triangulation of @xmath129 except near the @xmath208-simplex corresponding to @xmath5 , whose small neighborhood is homeomorphic to a cone over a torus . we can resolve this singularity by inserting a suspension of a @xmath218-gon around @xmath5 ( figure [ fig : prism ] ) . as a result , we obtain an ordered @xmath14-cycle homeomorphic to @xmath129 . ( this procedure is called `` blowing up '' at a @xmath208-simplex in @xcite . ) we construct a group cycle given by @xmath129 and a representation @xmath219 using the triangulation @xmath217 . here @xmath195 is given by the set @xmath220 satisfying @xmath221 give a labeling of vertices on @xmath211 for each @xmath222 . let @xmath223 be the labeling of vertices on @xmath224 and @xmath225 be the labeling of vertices on @xmath226 . if these are related by @xmath227 then we obtain a @xmath8-valued @xmath15-cocycle @xmath89 on @xmath217 by gluing @xmath211 using @xmath195 . to obtain a group cycle given by @xmath129 and @xmath195 , we insert a suspension over a @xmath218-gon to @xmath217 at the @xmath208-simplex corresponding to @xmath5 . we give an ordering on the vertices of each tetrahedron of the suspension compatible with the ordering on the @xmath218-gon , e.g. order the central @xmath208-simplex maximal . these tetrahedra inherit a vertex labeling by @xmath8 on the boundary faces of the suspension . we assign any labeling at the central @xmath208-simplex of the suspension . then upper and lower tetrahedra have the same labelings with different orientations . so these cancel out in pairs and give no contribution to the group cycle represented by @xmath129 and @xmath195 . therefore the homology class given by @xmath217 and @xmath89 represents the homology class given by @xmath129 and @xmath195 . in this section , we review the definitions of quandles , rack ( co)homology and quandle ( co)homology . our treatment of quandle ( or rack ) homology follows that of @xcite . in [ subsec : shadow ] and [ subsec : cocycle_invariant ] , we recall the notions of colorings and quandle cocycle invariants defined in @xcite . a _ quandle _ @xmath228 is a set with a binary operation @xmath105 satisfying the following axioms : * @xmath229 for any @xmath230 , * the map @xmath231 defined by @xmath232 is a bijection for any @xmath233 , * @xmath234 for any @xmath235 . we denote the inverse of @xmath236 by @xmath237 . for a quandle @xmath228 , we define the _ associated group _ @xmath238 by @xmath239 a quandle @xmath228 has a right @xmath238-action in the following way . let @xmath240 be an element of @xmath238 where @xmath241 and @xmath242 . define @xmath243 . one can easily check that this is a right action of @xmath238 on @xmath228 . so the free abelian group @xmath244 $ ] generated by @xmath228 is a right @xmath245$]-module . let @xmath246 be the free left @xmath245$]-module generated by @xmath247 . we define the boundary map @xmath248 by @xmath249 figure [ fig : boundary_map ] shows a graphical picture of the boundary map . let @xmath250 be the @xmath245$]-submodule of @xmath246 generated by @xmath251 with @xmath53 for some @xmath39 . now @xmath250 is a subcomplex of @xmath246 . let @xmath252 . for a right @xmath245$]-module @xmath26 , we define the _ rack homology _ of @xmath26 by the homology of @xmath253 } c^r_n(x)$ ] and denote it by @xmath254 . we also define the _ quandle homology _ of @xmath26 by the homology of @xmath255 } c^q_n(x)$ ] and denote it by @xmath256 . the homology @xmath257 , here @xmath25 is the trivial @xmath245$]-module , is equal to the usual quandle homology @xmath258 . let @xmath259 be a set with a right @xmath238-action . for any abelian group @xmath45 , the abelian group @xmath260 $ ] freely generated by @xmath259 over @xmath45 is a right @xmath245$]-module . the homology group @xmath261)$ ] is usually denoted by @xmath262 ( @xcite ) . let @xmath42 be a left @xmath245$]-module . we define the _ rack cohomology _ @xmath263 by the cohomology of @xmath264 } ( c^r_n(x ) , n)$ ] . the _ quandle cohomology _ @xmath265 is defined in a similar way . for a set @xmath259 with a right @xmath238-action and an abelian group @xmath45 , we let @xmath266 be the left @xmath245$]-module generated by functions @xmath267 , where the action is defined by @xmath268 for @xmath269 and @xmath270 . the cohomology group @xmath271 is usually denoted by @xmath272 ( @xcite ) . let @xmath228 be a quandle . let @xmath273 be an oriented link in @xmath13 and @xmath12 be a diagram of @xmath273 . an _ arc coloring _ of @xmath12 is an assignment of elements of @xmath228 to arcs of @xmath12 satisfying the following relation at each crossing , where @xmath274 . by the wirtinger presentation of the knot complement , an arc coloring determines a representation of @xmath275 into the associated group @xmath238 . this is obtained by sending each meridian to its color . let @xmath259 be a set with a right @xmath238 action . region coloring _ of @xmath12 is an assignment of elements of @xmath259 to regions of @xmath12 satisfying the relation for any pair of adjacent regions , where @xmath276 and @xmath230 . a pair @xmath277 is called a _ shadow coloring_. if we fix a color of a region of @xmath12 , then colors of other regions are uniquely determined . therefore there always exists a region coloring compatible with a given arc coloring . we define a cycle @xmath278 of @xmath279)$ ] for a shadow coloring @xmath280 . assign @xmath281 for a positive crossing and @xmath282 for a negative crossing colored by respectively . then define @xmath283),\ ] ] where @xmath284 . we can easily check that [ prop : not_depend_on_region_coloring ] @xmath278 is a cycle and the homology class @xmath285 $ ] is invariant under reidemeister moves . moreover it does not depend on the choice of the region coloring if the action of @xmath238 on @xmath259 is transitive . so the homology class @xmath285 $ ] is an invariant of the arc coloring @xmath286 in many cases . there are two important sets with right @xmath238-action : one is when @xmath259 consists of one point @xmath287 and the other is when @xmath288 . eisermann showed that the cycle @xmath285 $ ] for @xmath289 is essentially described by the monodromy of some representation of the knot group along the longitude ( @xcite , @xcite ) . so we concentrate on the invariant @xmath285 $ ] in the case of @xmath288 from now on . let @xmath228 be a quandle with @xmath290 . let @xmath45 be an abelian group and @xmath51 be a cocycle of @xmath291 . we define the _ ( shadow ) quandle cocycle invariant _ by @xmath292.\ ] ] here the sum is finite because there are only a finite number of shadow colorings of @xmath12 . this is an invariant of oriented knots by proposition [ prop : not_depend_on_region_coloring ] . if @xmath238 acts on @xmath228 transitively ( i.e. @xmath228 is connected ) , @xmath293 does not depend on the choice of a region coloring @xmath294 by proposition [ prop : not_depend_on_region_coloring ] , thus @xmath295 coincides with the quandle cocycle invariant , where @xmath294 is a region coloring compatible with @xmath286 . we can regard @xmath296 as an element of @xmath297 by @xmath298 this gives a homomorphism @xmath299 . therefore a quandle @xmath14-cocycle @xmath300 gives rise to a quandle cocycle invariant . explicitly , the cocycle invariant has the following form : @xmath301\ ] ] where @xmath302 are given by @xmath286 and @xmath303 are given by @xmath294 . let @xmath228 be a quandle . let @xmath306 . we define the boundary operator of @xmath307 by @xmath308 since @xmath228 has a right action of @xmath238 , the chain complex @xmath307 has a right action of @xmath238 by @xmath309 . let @xmath26 be a left @xmath245$]-module . we denote the homology of @xmath310 } m$ ] by @xmath311 . for any abelian group @xmath45 , we can also define the cohomology group @xmath312 in a similar way . let @xmath313 be the set consisting of maps @xmath314 . we let @xmath315 denote the cardinality of the set @xmath316 . for each generator @xmath317 of @xmath318)$ ] , where @xmath319 , we define @xmath320 for any @xmath321 . fix an element @xmath322 . for each @xmath323 , we define a homomorphism @xmath324 ) \longrightarrow c^{\delta}_{n+1}(x ) \otimes_{\mathbb{z}[g_x ] } \mathbb{z}\ ] ] by @xmath325 for example , in the case @xmath326 ( see figure [ fig : main_map ] ) , @xmath327 and in the case @xmath328 , @xmath329 the map @xmath330 ) \longrightarrow c^{\delta}_{n+1}(x ) \otimes_{\mathbb{z}[g_x ] } \mathbb{z}$ ] is a chain map . therefore @xmath331 induces a homomorphism @xmath332 ) \to h^{\delta}_{n+1}(x;\mathbb{z } ) $ ] . we remark that the induced map @xmath332 ) \to h^{\delta}_{n+1}(x;\mathbb{z } ) $ ] does not depend on the choice of @xmath322 ( @xcite ) . in general , it is easier to construct cocycles of @xmath333 from group cocycles of some group related to @xmath228 than those of @xmath334)$ ] . if we have a function @xmath51 from @xmath335 to some abelian group @xmath45 satisfying 1 . @xmath336 , 2 . @xmath337 for any @xmath233 , 3 . @xmath338 if @xmath53 for some @xmath39 , then @xmath51 is a cocycle of @xmath339 and @xmath340 is a cocycle of @xmath341 . moreover , @xmath340 can be regarded as a cocycle in @xmath342 by @xmath343 we will construct functions satisfying these three conditions from group cocycles . for any integer @xmath344 , let @xmath1 denote the cyclic group @xmath2 with quandle operation defined by @xmath345 . actually this operation satisfies the quandle axioms . the quandle @xmath1 is called the _ dihedral quandle_. in this section , we construct quandle cocycles of @xmath1 from group cocycles of @xmath346 . in the next section , we will propose a general construction of quandle cocycles from group cocycles . let @xmath8 be the cyclic group @xmath2 ( @xmath3 is an integer greater than @xmath200 ) . the first cohomology @xmath347 is generated by the 1-cocycle @xmath348 defined by @xmath349 the connecting homomorphism @xmath350 of the long exact sequence corresponding to @xmath351 maps @xmath348 to a generator of @xmath352 , and the reduction @xmath353 maps it to a generator @xmath354 of @xmath355 . explicitly we have @xmath356 where @xmath357 is an integer @xmath358 such that @xmath359 . cup products of @xmath348 s and @xmath354 s are also cocycles . moreover , when @xmath3 is an odd prime , it is known that any element of @xmath360 can be presented by a cup product of @xmath348 s and @xmath354 s , see e.g. ( * ? ? ? * proposition 3.5.5 ) . we remark that @xmath348 and @xmath354 and their products are normalized cocycles . for an integer @xmath361 , let @xmath51 be a normalized @xmath63-cocycle of @xmath362 . regarding @xmath1 as @xmath363 , we obtain a map @xmath364 satisfying 1 . @xmath336 , 2 . @xmath338 if @xmath53 for some @xmath39 , by using the homogeneous notation ( [ subsec : group_homology ] ) . if @xmath51 also satisfies the condition 1 . @xmath337 for any @xmath365 , then @xmath51 gives rise to a quandle @xmath63-cocycle of @xmath366 by the construction of section [ sec : main_map ] . define @xmath367 by @xmath368 then @xmath369 satisfies the condition ( 2 ) by the left invariance of the homogeneous group cocycle . it is easy to check that @xmath369 also satisfies the condition ( 1 ) and ( 3 ) . so we obtain a quandle @xmath63-cocycle . we give an explicit presentation of the @xmath14-cocycle arising from @xmath370 . let @xmath371 then @xmath372 is a 2-cocycle . ( we can check that @xmath372 is cohomologous to @xmath373 . ) then by the definition ( equation ( [ eq : average_of_cocycle ] ) ) , @xmath374 is given by @xmath375 \mapsto x \cdot d(y , z).\ ] ] by definition we have @xmath376 and @xmath377 we remark that the cocycle @xmath372 can be easily understood geometrically . identify @xmath378 with the complex number @xmath379 where @xmath380 . then @xmath381 if @xmath382 is counterclockwise , @xmath383 if @xmath382 is clockwise and @xmath384 if @xmath382 is degenerate ( figure [ fig:2-cocycle ] ) . this interpretation and the equation ( [ eq : symmetric ] ) make various calculations easy . the quandle @xmath14-cocycle arising from @xmath370 has the following presentation : @xmath385 this is a non - trivial quandle 3-cocycle of @xmath1 with @xmath2 coefficient . in ( [ eq : main_map ] ) , since the map @xmath386 does not depend on the choice of @xmath387 , we let @xmath388 . then we have @xmath389 for @xmath390 . rewriting in inhomogeneous notation , this is equal to @xmath391 - [ 2y - x|x - y|z - y ] \\ & - [ 2z - x| x - y|y - z ] + [ 2z-2y+x| y - x|y - z ] . \\ \end{split}\ ] ] the evaluation of @xmath374 on this cycle is @xmath392 we will see that this cocycle is non - trivial because the evaluation on the cycle given by a shadow coloring @xmath280 of the @xmath393-torus link ( figure [ fig:(2,p)-torus ] ) is non - zero . color two arcs by @xmath394 as in the figure [ fig:(2,p)-torus ] . then other arcs must be colored by @xmath395 by the relations at the crossings . let @xmath396 be the color of the central region . then we have @xmath397 we assume that @xmath398 and @xmath399 . evaluation of the cocycle on @xmath278 is equal to @xmath400 by the next lemma , this is equal to @xmath401 . let @xmath361 be an integer . for @xmath402 , we have @xmath403 when @xmath404 , the left hand side is @xmath208 . when @xmath404 , this is straightforward since @xmath405 for any @xmath406 . when @xmath407 is a unit in @xmath2 , by ( [ eq : formula_of_d ] ) we have @xmath408 this is equal to @xmath409 when @xmath3 is odd . when @xmath3 is even , @xmath410 since @xmath411 is a unit in @xmath2 . when @xmath407 is not a unit in @xmath2 , let @xmath412 be the greatest common divisor of @xmath407 and @xmath3 then @xmath413 since @xmath414 is a unit in @xmath415 , this reduces to the previous case . since @xmath200 is divisible in @xmath2 when @xmath3 is odd , we have when @xmath361 is an odd number , @xmath416 is a non - trivial quandle 3-cocycle of @xmath1 with @xmath2 coefficient . when @xmath3 is prime , it is known that @xmath4 . therefore our cocycle is a constant multiple of the mochizuki 3-cocycle @xcite . we remark that when @xmath3 is prime , @xmath417 was calculated by nosaka @xcite for any @xmath18 , moreover he gave a system of generators of @xmath418 . when @xmath3 is an odd integer , nosaka showed in @xcite that @xmath419 . since @xmath420 is zero , we have @xmath421 . this means that there exists a non - trivial quandle 3-cocycle of @xmath1 with @xmath2 coefficient . in this section , we generalize the construction in the previous section to wider classes of quandles . we will construct a quandle cocycle of a faithful homogeneous quandle @xmath228 from a group cocycle of @xmath422 when an obstruction living in the second cohomology of @xmath422 vanishes . [ [ section ] ] let @xmath8 be a group . fix an element @xmath423 . let @xmath424 . now @xmath11 has a quandle operation by @xmath425 . in this section , we shall construct a quandle cocycle of @xmath11 from group cocycle of @xmath8 . first we shall show that this class of quandles is not so special . let @xmath228 be a quandle . we denote the group of the quandle automorphisms of @xmath228 by @xmath422 . we regard an automorphism acts on @xmath228 from the right . for @xmath230 , let @xmath426 be the map which sends @xmath407 to @xmath427 . by the axioms ( q2 ) and ( q3 ) , @xmath426 is a quandle automorphism . @xmath228 is called _ faithful _ if @xmath428 is injective . a quandle @xmath228 is _ homogeneous _ if @xmath422 acts on @xmath228 transitively . the following lemma was essentially shown in theorem 7.1 of @xcite , but we include a proof for completeness . every faithful homogeneous quandle @xmath228 is represented by @xmath11 with some group @xmath8 and @xmath423 . for @xmath230 and @xmath429 , we have @xmath430 . in fact , @xmath431 for any @xmath233 . let @xmath432 and fix an element @xmath433 . put @xmath434 . because @xmath228 is homogeneous , for any @xmath230 , there exists @xmath50 such that @xmath435 . so we have @xmath436 , namely @xmath437 . therefore we obtained a homomorphism @xmath438 . this is surjective since @xmath439 , and injective since @xmath228 is faithful . let @xmath440 be the centralizer of @xmath9 in @xmath8 . a map @xmath441 is well - defined and bijective . let @xmath442 . then @xmath443 , so @xmath444 and @xmath445 . this means that @xmath446 and @xmath447 belong to the same right coset . therefore the map is well - defined . by a similar argument , we can show the injectivity . surjectivity is trivial by definition . from now on we study the quandle structure on @xmath448 and construct a section of the projection @xmath449 . the quandle operation on @xmath11 induces a quandle operation on @xmath448 : @xmath450 this quandle operation on @xmath448 lifts to a quandle operation @xmath97 on @xmath8 by @xmath451 we can easily check that @xmath97 satisfies the quandle axioms ( the inverse operation is given by @xmath452 ) and the projection map @xmath453 is a quandle homomorphism . we remark that the quandle operation given by ( [ eq : lift_of_the_operation ] ) has been already studied by joyce @xcite and eisermann @xcite . let @xmath454 be a section of @xmath455 , i.e. a map ( not a homomorphism ) satisfying @xmath456 . since @xmath457 and @xmath458 are in the same coset in @xmath448 , there exists an element @xmath459 satisfying @xmath460 [ lem : section ] if @xmath461 is an abelian group , @xmath462 is a quandle 2-cocycle . if the cocycle @xmath412 is cohomologous to zero , we can change the section @xmath222 to satisfy @xmath463 . for @xmath464 and @xmath465 , we have @xmath466 therefore @xmath467 and @xmath468 for any @xmath469 . comparing ( [ eq : ditribution_1 ] ) and ( [ eq : ditribution_2 ] ) , we have @xmath470 by @xmath471 , we also have @xmath472 . if @xmath412 is cohomologous to zero , then there exists a map @xmath473 satisfying @xmath474 . put @xmath475 , then @xmath476 satisfies @xmath477 . the @xmath200-cocycle @xmath412 has already appeared in @xcite in a similar context . let @xmath8 be the dihedral group @xmath478 where @xmath3 is an odd number greater than @xmath200 . then we have @xmath479 and @xmath480 . we can identify @xmath481 with @xmath482 . define a section @xmath454 by @xmath483 then we have @xmath484 therefore @xmath485 for any @xmath486 . [ [ section-1 ] ] let @xmath8 be a group . fix @xmath423 with @xmath487 ( @xmath125 ) . in the following we assume : [ asm : vanishing ] @xmath461 is abelian and the 2-cocycle corresponding to @xmath488 is cohomologous to zero . under the assumption , we can take a section @xmath454 satisfying @xmath463 by lemma [ lem : section ] . let @xmath489 be a normalized group @xmath63-cocycle of @xmath8 in the homogeneous notation , where @xmath45 is an abelian group . then @xmath51 satisfies 1 . @xmath336 , 2 . @xmath490 for any @xmath50 ( left invariance ) , 3 . @xmath338 if @xmath53 for some @xmath39 . since it is convenient to use a right invariant function in the following construction , we modify the condition ( 2 ) by 1 . @xmath491 for any @xmath50 ( right invariance ) , by replacing @xmath492 by @xmath493 . define @xmath494 by @xmath495 for @xmath496 . the function @xmath369 satisfies the conditions ( 1 ) , ( 2 ) and ( 3 ) of section [ sec : main_map ] . therefore @xmath369 gives rise to a @xmath63-cocycle of @xmath497 . it is clear that ( 1 ) and ( 3 ) are satisfied from the conditions on a normalized group cocycle in the homogeneous notation . we only have to check the second property . @xmath498 combining with the arguments of section [ sec : main_map ] , we have if @xmath461 is abelian and the 2-cocycle corresponding to @xmath488 is cohomologous to zero , then there is a homomorphism @xmath499 for any abelian group @xmath45 . [ [ section-2 ] ] we return to the case of @xmath1 discussed in the previous section . we assume that @xmath3 is an odd integer greater than @xmath200 . let @xmath8 be the dihedral group @xmath478 . consider the short exact sequence @xmath500 we regard @xmath501 as @xmath502 by taking coset representatives in @xmath503 . then @xmath501 acts on @xmath2 by @xmath504 . this induces the restriction map @xmath505 we can show that this homomorphism is an isomorphism ( proposition iii.10.4 of @xcite ) . to obtain a group cocycle of @xmath503 from a group cocycle of @xmath2 , we need the inverse map , which is called the _ transfer_. the transfer map is described as follows ( see also @xcite ) . let @xmath396 be the map @xmath506 defined by @xmath507 and @xmath508 . for a cocycle @xmath51 of @xmath509 , the image @xmath510 of the transfer is given by @xmath511 in the homogeneous notation . when restricted to the image of @xmath454 , this is equal to the map defined in ( [ eq : average_of_cocycle ] ) . applying our construction for this group cocycle , we obtain a quandle 3-cocycle of @xmath1 , which is twice the cocycle constructed in the previous section . [ [ subsec : quandle_to_group ] ] we end this section by giving another homomorphism from @xmath512 to @xmath513 arising from more general context . let @xmath228 be a quandle and @xmath26 be a right @xmath245$]-module . we can construct a map from the rack homology @xmath514 to the group homology @xmath515 . the following lemma is well - known , e.g. see lemma 7.4 of chapter i of @xcite . let @xmath516 be a chain complex where @xmath517 are projective ( e.g. free ) . let @xmath518 be an acyclic complex . any homomorphism @xmath519 can be extended to a chain map from @xmath520 to @xmath521 . moreover such a chain map is unique up to chain homotopy . so there exists a unique chain map from @xmath522 to @xmath523 up to homotopy . this map induces @xmath524 } c^r_*(x ) \to m \otimes_{\mathbb{z}[g_x ] } c_*(g_x)$ ] and then @xmath525 . using normalized chains in group homology , we can also construct a map @xmath526 . we give an explicit chain map . let @xmath527 be a generator of @xmath246 . we define a map @xmath528 by @xmath529\ ] ] where @xmath530 is defined for a permutation @xmath66 and @xmath531 as follows . let @xmath532 be the maximal set of numbers satisfying @xmath533 . then define @xmath534 a graphical picture of this map is given in figure [ fig : quandle_cycle_to_group_cycle ] . for example , when @xmath328 , @xmath535 - [ x|z|y*z]+[y|z|(x*y)*z]-[y|x*y|z ] \\ & + [ z|x*z|y*z]-[z|y*z|(x*y)*z ] , \end{split}\ ] ] for @xmath536 . dually , we also have a map @xmath537 . we apply this map for @xmath11 . since there exists a natural homomorphism from the associated group @xmath538 to @xmath8 , we have a homomorphism @xmath539 fenn , rourke and sanderson defined the _ rack space _ @xmath540 in @xcite . since @xmath541 is isomorphic to @xmath238 , there exists a unique map , up to homotopy , from @xmath540 to the eilenberg - maclane space @xmath542 which induces the isomorphism between their fundamental groups . this map induces a homomorphism @xmath537 , which is equal to the map we have constructed in this subsection . clauwens showed in proposition 25 of @xcite that this map vanishes under some conditions on @xmath228 and @xmath26 . in particular , when @xmath3 is odd prime , @xmath543 vanishes for @xmath544 . in this section , we study the dual of the previous construction . we will show that the cycle @xmath278 associated with a shadow coloring @xmath277 of a knot @xmath5 gives rise to a group cycle represented by a cyclic branched covering along @xmath5 and the representation induced from the arc coloring @xmath286 . let @xmath228 be a quandle . let @xmath12 be a diagram of a knot @xmath5 . for a shadow coloring @xmath545 of @xmath12 whose arcs and regions are colored by @xmath228 , define @xmath546 and @xmath547 for @xmath548 by @xmath549 by the axiom ( q3 ) , @xmath550 is also a shadow coloring . in the following , we assume that @xmath551 for some group @xmath8 and @xmath423 and satisfies assumption [ asm : vanishing ] . as in the previous section , let @xmath552 be a section satisfying @xmath553 for @xmath554 . let @xmath555 composing with @xmath556 ) \longrightarrow c^{\delta}_{n+1}(x)$ ] , we have a map @xmath557 ) \xrightarrow{\varphi } { } c^{\delta}_{3}(x ) \xrightarrow{\iota } { } c_{3}(g ; \mathbb{z}).\ ] ] for a shadow coloring @xmath280 , @xmath558 is not a cycle of @xmath559 in general . but we can show [ thm : main_branched ] let @xmath560 be a shadow coloring of a diagram @xmath12 of a knot @xmath5 by @xmath11 . let @xmath561 be the color of an arc of @xmath12 . if @xmath562 , @xmath563 is a group cycle represented by the @xmath127-fold cyclic branched covering @xmath129 along the knot @xmath5 and the representation @xmath564 induced from the arc coloring @xmath286 . let @xmath565 be the arcs of the diagram @xmath12 such that @xmath566 . we denote the color @xmath567 by @xmath568 ( @xmath569 ) . the arc coloring @xmath286 induces a representation @xmath570 . using @xmath571 , @xmath111 is given by @xmath572 we have @xmath573 for any @xmath574 . so we have @xmath575 since @xmath576 by ( [ eq : covering_representation ] ) , we have @xmath577 let @xmath578 be copies of the triangulation of a @xmath14-ball with a vertex ordering on each tetrahedra constructed in section [ sec : cyc_rep_by_cyc_bra_cov ] ( figure [ fig : cut_heegaard ] and [ fig : ordering ] ) . then we define a labeling of vertices of @xmath211 by @xmath579 ( see figure [ fig : labeling_of_crossing ] ) . the vertices of the face @xmath580 is labeled by @xmath581 and the vertices of the face @xmath582 is labeled by @xmath583 where @xmath396 is the color of the region as indicated in figure [ fig : labeling_of_crossing ] . by ( [ eq : gluing_relation ] ) the labelings of the face @xmath584 and @xmath585 are related by @xmath586 therefore this gives a @xmath8-valued @xmath15-cocycle on @xmath217 as constructed in section [ sec : cyc_rep_by_cyc_bra_cov ] . thus the chain given by ( [ eq : cycle_represented_by_cyclic_branched_cover ] ) is a cycle represented by the cyclic branched cover @xmath129 and the representation induced from the arc coloring @xmath286 . we end this section by comparing the shadow cocycle invariant of the @xmath393-torus knot ( @xmath3 : odd ) for the @xmath14-cocycle obtained in section [ sec : dihedral ] and the dijkgraaf - witten invariant of the lens space @xmath587 . let @xmath8 be a finite group and @xmath588 be a cocycle of @xmath589 where @xmath45 is an abelian group . dijkgraaf and witten defined an invariant of closed oriented 3-manifolds for each @xmath588 . for an oriented closed manifold @xmath26 , it is defined by @xmath590 \rangle \in \mathbb{z}[a]\ ] ] where @xmath591 is the pull - back of @xmath588 by the classifying map @xmath592 corresponding to @xmath111 . since @xmath587 is a double branched cover along the @xmath393-torus knot , the cycle obtained from a shadow coloring by @xmath1 gives rise to a group @xmath14-cycle of @xmath503 represented by @xmath587 by theorem [ thm : main_branched ] . since every representation of @xmath593 into @xmath503 reduces to a representation into @xmath2 , it is natural to ask whether the shadow cocycle invariant for our quandle @xmath14-cocycle coincides with the dijkgraaf - witten invariant for @xmath594 . as we remarked in the introduction , these invariants coincide up to some constant by the result of hatakenaka and nosaka @xcite when @xmath3 is prime . let @xmath280 be the shadow coloring indicated in figure [ fig:(2,p)-torus ] . since the homology class @xmath285 $ ] does not change under the action of @xmath1 on the shadow coloring @xmath280 ( see lemma 4.5 of @xcite ) , we assume that @xmath399 . as computed in [ subsec:3_cocycle_of_dihedral ] , the evaluation of the @xmath278 at the @xmath14-cocycle derived from @xmath595 is equal to @xmath401 . therefore the shadow cocycle invariant is @xmath596/ ( t^p-1 ) \cong \mathbb{z}[\mathbb{z}/p].\ ] ] we compute the dijkgraaf - witten invariant of the lens space @xmath597 for the group @xmath14-cocycle @xmath594 . although this was computed in @xcite , we give a proof based on a triangulation . we represent @xmath597 by an ordered @xmath14-cycle and give a labeling of @xmath15-simplices as indicated in figure [ fig : lens ] . here the triple @xmath598 , @xmath599 and @xmath412 must satisfy @xmath600 and @xmath601 . using multiplicative notation , the evaluating of @xmath595 on this cycle is @xmath602 ) = \sum_{i=0}^{p-1 } a \cdot b_2(ia+b , -q a ) = -q a^2.\ ] ] the second equality follows form the fact that @xmath603 runs over @xmath2 and the equation ( [ eq:2_cocycle_of_cyclic_group ] ) when @xmath598 is a unit , and also when @xmath598 is not a unit by a similar argument . therefore the dijkgraaf - witten invariant of @xmath597 is equal to @xmath604/ ( t^p-1 ) \cong \mathbb{z}[\mathbb{z}/p].\ ] ] therefore this coincides with the shadow cocycle invariant up to a constant . 99 d. j. benson , _ representations and cohomology i : basic representation theory of finite groups and associative algebras _ , cambridge studies in advanced mathematics , 30 . cambridge university press , cambridge , 1991 . xii+224 pp . s. kamada , _ quandles with good involutions , their homologies and knot invariants _ , intelligence of low dimensional topology 2006 , 101108 , ser . knots everything , 40 , world sci . hackensack , nj , 2007 .

we give a construction of quandle cocycles from group cocycles , especially , for any integer @xmath0 , quandle cocycles of the dihedral quandle @xmath1 from group cocycles of the cyclic group @xmath2 . we will show that a group 3-cocycle of @xmath2 gives rise to a non - trivial quandle 3-cocycle of @xmath1 . when @xmath3 is an odd prime , since @xmath4 , our 3-cocycle is a constant multiple of the mochizuki 3-cocycle up to coboundary . dually , we construct a group cycle represented by a cyclic branched covering branched along a knot @xmath5 from the quandle cycle associated with a colored diagram of @xmath5 .

introduction group homology cyclic branched covering cycle represented by cyclic branched covering quandle homology @xmath304 and the map @xmath305) \to h^{\delta}_{n+1}(x;\mathbb{z})$] cocycles of dihedral quandles general construction quandle cycle and branched cover

the main motivation for studying @xmath0-photoproduction on the deuteron is to obtain information on the elementary process on the neutron . the evident advantages of using a deuteron target are ( i ) the small binding energy of nucleons in the deuteron , which from the kinematical point of view provides the case of a nearly free neutron target , and ( ii ) minimal number of nucleons , reducing the influence of a nuclear environment on the elementary production process and allowing a careful microscopic treatment . the interest in the elementary process on the neutron is closely connected with the question concerning the isotopic structure of the @xmath3(1535 ) photoexcitation amplitude . a combined analysis of @xmath4 , @xmath5 and inclusive @xmath6 reactions @xcite have confirmed its predominantly isovector character as predicted by quark models @xcite . furthermore , very recently exclusive measurements of the ( @xmath7 ) reaction on the deuteron for quasifree kinematics with detection of forward emerging nucleons in coincidence with etas were performed @xcite . this experiment , providing direct information on the ratio @xmath8 , is expected to be quite precise , since the experimental systematic errors as well as different model corrections will largely cancel . from the data of @xcite the ratio @xmath9 of the isoscalar to the proton amplitude is extracted as @xmath10 . at the same time , if one wants to find agreement between the data on the coherent @xmath11 cross section of @xcite and conventional impulse approximation calculations @xcite , one needs a considerably larger parameter @xmath12 . if one assumes that the value @xmath13 is more reliable , one is lead to the conclusion that the impulse approximation fails in describing the coherent reaction . thus one either would need to search for new mechanisms dominating the coherent process or one would need to question the accuracy of the experimental data . it should be noted , that the method of extracting the information about the single nucleon processes from the deuteron data is based on the so - called spectator - nucleon model , in which the pure quasifree production is considered as the only mechanism for the knock - out @xmath14 reaction . therefore , a careful investigation of the validity of this approximation , i.e. , the estimation of possible other effects is needed . the main goal of the present work is to study the role of rescattering in the final state of the @xmath15 process . one can expect that such effects become important close to the reaction threshold . in this region the smallness of the excitation energy in the final @xmath16-system and the large momentum transfer ( which is about the @xmath0 mass in the @xmath17 c.m . frame ) lead to a kinematical situation , where two final nucleons move primarily together with a large total , but small relative momentum . for this kinematics , the spectator model is expected to give a very small cross section since the momenta of both nucleons are large and , on the other hand , the corrections due to the strong @xmath18-interaction may be significant . furthermore , as has been shown in @xcite , the @xmath0-rescattering can also visibly change the @xmath15 cross section near threshold . in section [ section2 ] we present the isobar model for the elementary @xmath0-photoproduction process . the treatment of the @xmath15 amplitude , based on time - ordered perturbation theory including @xmath19 and @xmath18-rescattering , is developed in section [ section3 ] . a similar approach was adopted for the analysis of coherent @xmath20-photoproduction on the deuteron @xcite . in our notation , we follow mainly this paper . in this section , we also discuss the effects of rescattering on cross sections and polarization observables and compare our results with available experimental data . finally , some attention is paid to the interpretation of experimental results of ref . @xcite on the coherent process . as a starting point we consider the elementary process @xmath21 which we need as input for studying the @xmath0-photoproduction on the deuteron . the four momenta of the participating particles are denoted by @xmath22 and @xmath23 stands for the photon polarization vector . the particle energies are @xmath24 where @xmath25 and @xmath26 are the masses of nucleon and @xmath0 meson , and @xmath27 , @xmath28 , @xmath29 , and @xmath30 denote the absolute values of the respective three - momenta . past studies @xcite of the @xmath31 process for photon lab energies from threshold up to 800 mev have clearly shown the dominant role of the @xmath3(1535 ) resonance which decays with a relatively large probability into the @xmath19 channel . for this reason we have chosen a nonrelativistic isobar model , in which the @xmath0-photoproduction proceeds exclusively through the electric dipole transition to the @xmath3(1535 ) ( hereafter denoted by @xmath32 ) . possible background contributions , such as nucleon pole and vector meson exchange , are neglected because of their insignificance in the energy region under consideration . within the mentioned model the elementary @xmath0-photoproduction amplitude in an arbitrary frame of reference reads @xmath33 where @xmath34 is the @xmath35 invariant mass . the nonrelativistic current for the electromagnetic vertex @xmath36 is given by @xmath37[16 ] where @xmath38 is the resonance mass and @xmath39 stands for the relative photon - nucleon momentum @xmath40 the vertex constant @xmath41 can be expressed in terms of the helicity amplitude @xmath42 of the @xmath32 photoexcitation @xmath43 furthermore , we introduce the isoscalar @xmath44 and isovector @xmath45 coupling constants in accordance with the isotopic structure of the photoproduction amplitude for an isoscalar meson @xmath46 for the hadronic vertex @xmath47 we use the pseudoscalar coupling @xmath48 finally , the @xmath32 propagator @xmath49 involves the energy dependent decay width @xmath50 the partial @xmath19 and @xmath51 widths are related to the corresponding coupling constants in our nonrelativistic approach @xmath52 where @xmath53 is the momentum of the respective particle in the @xmath35 c.m . system . following ref . @xcite we treat the @xmath32 width for the @xmath54 channel purely phenomenologically , as a constant above @xmath54 threshold . thus the full @xmath0-photoproduction amplitude may be written in the form @xmath55 in the actual calculation we use the following set of @xmath32 parameters @xmath56 which gives for the total and partial decay widths at the resonance position @xmath57 @xmath58 this parametrization is reasonably consistent with the values following from the 1996 pdg listings @xcite . the amplitude ( [ 25 ] ) gives quite a good description of the @xmath4 total cross section data @xcite ( see fig . [ fig1 ] ) if we take for the proton helicity amplitude the value @xmath59gev@xmath60 . now we turn to incoherent @xmath0-photoproduction on the deuteron @xmath61 the momenta of all particles are defined similarly to those in section [ section2 ] . we restrict ourselves to the small region of incident photon energies from threshold up to 720 mev . all calculations refer to the laboratory frame , @xmath62 , unless otherwise noted . the coordinate system is chosen to have a right - hand orientation with @xmath63-axis along the photon momentum @xmath64 and @xmath65-axis parallel to @xmath66 . we work in the isospin formalism and consider the two outgoing nucleons symmetrically , not specifying their isotopic states . the ratio @xmath9 is taken to be 0.11 , which , as will be discussed below , produces the best agreement with experimental data . the laboratory cross section for the reaction ( [ 31 ] ) with unpolarized particles reads @xmath67 where @xmath68 denotes the reaction amplitude @xmath69 here @xmath70 stands for the deuteron spin projection , and the final two nucleon state is specified by the nucleon momenta @xmath71 , @xmath72 as well as by spin @xmath73 and isospin @xmath74 quantum numbers . the corresponding fully antisymmetric @xmath18-function can formally be written as @xmath75 the deuteron wave function in ( [ 33 ] ) is noncovariantly normalized to unity . the fourier transform of its internal part separating the spin part has the familiar form @xmath76 for the radial functions @xmath77 we use the parametrization of the bonn - potential model ( obepq - version ) @xcite . the main features of the process ( [ 31 ] ) will be investigated by considering the partially integrated cross sections @xmath78 and @xmath79 , which are obtained from the fully exclusive cross section @xmath80 is expressed in terms of relative @xmath81 and total @xmath82 momenta of the two final nucleons . we prefer this choice of variables , because in this case the kinematic factor does not have any singularity on the boundary of the available phase space , when @xmath830 @xcite . with respect to polarization observables , we consider in this work only the tensor asymmetries @xmath84 of the deuteron target , which , as will be discussed , is most sensitive to the effects of rescattering . for this observable , we use the definition similar to that given in ref . @xcite for deuteron photodisintegration @xmath85 where @xmath86 thus , for example , the asymmetry @xmath87 is given by the expression @xmath88 then , the cross section can be expressed in terms of the unpolarized cross section and the tensor target asymmetry as @xmath89,\ ] ] where @xmath90 is the degree of deuteron tensor polarization with respect to the orientation for which the deuteron density matrix is diagonal and which is characterized by the angles @xmath91 and @xmath92 in the chosen coordinate system . the transition matrix elements @xmath68 are calculated in the frame of time - ordered perturbation theory , using the electromagnetic and hadronic vertices , introduced in section [ section2 ] . as full amplitude we include besides the pure impulse approximation ( ia ) one - loop @xmath18- and @xmath19-rescattering terms as shown in fig . [ fig2 ] @xmath93 another possible two - step @xmath0-production mechanism in the two - nucleon system @xmath94 gives a negligible contribution as was shown in @xcite . we now will consider successively all three terms and investigate their role in the reaction ( [ 31 ] ) . the matrix element given by the diagramm ( a ) in fig . [ fig2 ] reads @xmath95 the effective one - particle operators for the hadronic vertex @xmath96 and current @xmath97 are defined using ( [ 15 ] ) through ( [ 20 ] ) as @xmath98 with arguments , referring to nucleon 1 and 2 . the propagator for the @xmath99 intermediate state is @xmath100 with @xmath34 being the energy available for the @xmath32 excitation @xmath101 in the deuteron lab system we have @xmath102 . taking plane waves for all three outgoing particles , the expression ( [ 41 ] ) can be reduced to the form @xmath103|\,\,m_d\rangle\,,\ ] ] with upper index @xmath104 referring to the isoscalar @xmath105 or the isovector @xmath106 amplitude ( [ 25 ] ) for @xmath107 and @xmath108 , respectively . the energy @xmath109 is defined in ( [ 44 ] ) with @xmath110 for i=1(2 ) . the amplitude ( [ 45 ] ) corresponds to the simple spectator - nucleon model . the elementary operator @xmath111 produces the ( @xmath112 ) process on the nucleon , which takes the whole photon energy available after meson production . since in our approach the spectator nucleon is taken to be on - shell , the active nucleon is off - shell as determined by the energy and momentum conservation at the corresponding vertex @xmath113 in order to qualitatively explain the approximations concerning the rescattering terms , discussed below , we would like to demonstrate here some features of the reaction amplitude keeping in ( [ 37 ] ) only the impulse approximation . due to the large transferred momentum associated with the large @xmath0 mass , the @xmath0-angular distribution shows a fast decrease with increasing angle , so that the major part of the cross section is concentrated in the forward direction . therefore , we choose for the discussion the forward angle region . the differential cross section @xmath79 calculated within the spectator model is shown in fig . [ fig3 ] as a function of the lab photon energy , where in addition the separate contribution from the various important partial waves in the initial and final states are shown . several conclusions can be drawn : \(i ) because of the large velocity of both outgoing nucleons near threshold , the high momentum components of the deuteron wave function become significant . as a consequence , a few mev above threshold a rather large contribution from the deuteron @xmath114-wave arises leading to a strong distructive interference between @xmath115 and @xmath116 states which reduces sizeably the pure @xmath117-wave cross section . with increasing photon energy , the role of the @xmath114-wave decreases rapidly , e.g. , at @xmath118 mev this reduction amounts to only 8@xmath119 . \(ii ) the separate contributions from the @xmath120 and @xmath116 states of the final @xmath18-system clearly show the dominance of the singlet over the triplet @xmath121-wave . the strong suppression of the @xmath116 state is due to the small isoscalar part of the @xmath0-photoproduction amplitude ( in our calculation @xmath122 0.12@xmath123 ) so that the transition from the deuteron bound state ( @xmath107,@xmath124 ) to the continuum @xmath18-state ( @xmath108,@xmath125 ) is most probable . if one neglects the @xmath114-wave of the deuteron , the following approximate relation holds @xmath126 this strong dominance of the singlet @xmath121-wave state is slightly reduced by the presence of the @xmath114-component , in particular near threshold . the diagram ( b ) in fig . [ fig2 ] describes the @xmath18-rescattering whose amplitude is given by @xmath127 the free nonrelativistic propagator of the two nucleons in terms of the relative @xmath18-momenta @xmath81 and @xmath128 has the form @xmath129 the @xmath18 dynamics in the final state is determined by the half - off - shell nucleon - nucleon + @xmath130-matrix @xmath131 , which we evaluate for the same obepq bonn potential @xcite as for the deuteron wave function . we restrict the @xmath18-rescattering to the @xmath120 state . this approximation is justified by the previous observation that due to the spin - isospin selection rules the contribution from the triplet @xmath116 state is at least one order of magnitude smaller than that from the @xmath120 state . as for the higher partial waves , their distortion is expected to be of minor importance compared to the resonant @xmath120-wave , since in the kinematical region considered in this work the relative kinetic energy in the @xmath18-system is less than 72 mev ( see also @xcite ) . within this approximation , a straightforward calculation yields finally @xmath132 . since , as was noticed above , the deuteron @xmath114-wave component plays a non - negligible role in the ia , it was also taken into account in the matrix element ( [ 55 ] ) . the influence of @xmath18-rescattering is demonstrated in fig . [ fig4 ] for the @xmath0-momentum distribution at a given angle @xmath133 . as expected , the strong interaction between the final nucleons in the @xmath120 state changes drastically the cross section for large @xmath0 momentum values . when @xmath28 reaches its maximum , the excitation energy @xmath134 in the @xmath16-pair vanishes , and thus the resonant @xmath120 state appears as a rather narrow peak . the same effect is noted for charged @xmath135-photoproduction on the deuteron @xcite as well as for deuteron electrodisintegration @xcite . in principle , the experimental observation of this peak in the high @xmath0-momentum spectrum may serve as another evidence for the isovector nature of the @xmath32 photoexcitation . in the hypothetical case @xmath136 , the low - energy @xmath18-rescattering would be dominated by the @xmath116 state , which does not exhibit any resonant behaviour at @xmath137 0 . in conclusion , one sees that the role of @xmath18-rescattering is quite important , especially for small photon energies . at higher energies , the main part of the cross section is dominated by the ia , which gives a rather broad quasifree bump , where the role of @xmath18-rescattering is expected to be of minor importance . the contribution from the deuteron @xmath114-wave is only significant close to threshold . its effect is a destructive interference between the @xmath138 and the dominant @xmath139 transitions ( see fig . [ fig4 ] ) . in fig . [ fig5 ] we depict also the tensor target asymmetries for @xmath140 at two different photon energies . one readily notes that the asymmetry @xmath87 is strongly affected by @xmath18-rescattering . the reason for this is that @xmath87 is directly defined by the difference between the cross section on the deuteron with spin parallel and perpendicular to the photon beam direction ( see ( [ t20 ] ) ) . because of the spin - transverse form of the elementary operator ( [ 25 ] ) , the transition @xmath139 may occur in the @xmath0-photoproduction process with only opposite directions of photon and deuteron spin . consequently , the strong @xmath120 interaction between the final nucleons , enhancing the probability of this transition , increases the contribution from the matrix elements @xmath68 in ( [ t20 ] ) with @xmath1411 . the other target asymmetries , @xmath142 and @xmath143 which are not shown here , appear not to be sensitive to the @xmath18-rescattering contribution . for the photon asymmetry @xmath144 we have a trivial relation @xmath145 , which follows from our neglect of other terms besides the @xmath32-excitation amplitude in the elementary operator . the @xmath19-rescattering amplitude given by the the diagram ( c ) in fig . [ fig2 ] reads @xmath146 where @xmath147 . similarly to the @xmath18-case ( [ 53 ] ) , we use the nonrelativistic propagator for the relative @xmath19-motion @xmath148 in the spirit of our isobar approach , we consider the @xmath19-interaction to proceed only via the @xmath32 formation , which gives the @xmath130-matrix in the form @xmath149 where @xmath150 is the @xmath19 invariant mass and other parameters are defined in section [ section2 ] . collecting the various pieces , we obtain for the amplitude ( [ 61 ] ) @xmath151 \frac{d^3q_{\eta n}^{\,\prime}}{(2\pi)^3 } @xmath152 and @xmath153 @xmath154 . the effect of @xmath19-rescattering is demonstrated in fig . [ fig6 ] for the @xmath0-angular distribution in the @xmath17 c.m . frame , where we also compare our results with experimental data for the inclusive reaction @xmath6 @xcite . in view of the small isoscalar part @xmath155 of the elementary amplitude and the large momentum mismatch , the contribution from the coherent @xmath11 process is expected to be negligible and , thus , the inclusive @xmath0-spectrum is dominated by the deuteron break - up channel . in the backward direction , the increase from @xmath19-rescattering is in part kinematically enhanced . at these angles , the nucleons leave the interaction region with large momenta . therefore , the spectator model gives a very small cross section understimating the data by roughly a factor of 3 for @xmath156 mev . in this situation , the @xmath19-rescattering mechanism , allowing the large transferred momentum to be shared between two participating nucleons , becomes much more effective . the resonant character of the @xmath19-interaction appears more pronounced at forward angles . although its strong inelasticity decreases the cross section at high photon energies , close to threshold this effect is more than compensated by the attraction in the @xmath19-system . finally , we show in fig . [ fig7 ] the total @xmath15 cross section . one readily notes , that the simple spectator approach can not describe the experimental data close to the threshold ( see also @xcite ) . as has been discussed in the introduction , at small photon energies the spectator model considers the @xmath15 process to proceed mainly through the high deuteron fourier component , which appears only with small probability . the final state interaction provides a mechanism for bypassing this suppression . our model predicts quite a significant contribution from @xmath18-rescattering . it turns out to be dominant in the vicinity of the threshold and still increases the spectator results by about 10@xmath119 at @xmath156 mev . the @xmath19-interaction is relatively less important , but is also significant mainly through the constructive interference between @xmath18- and @xmath19-rescattering contributions . with inclusion of the rescattering effects , we are able to reproduce the experimental cross section with @xmath157 . at this point , we would like to discuss briefly the possibility of extracting the ratio @xmath158 from the corresponding ratio of exclusive @xmath159 and @xmath160 yields as has been done in ref . @xcite . strictly speaking , the simple relation @xmath161 consistent with a pure spectator approach is not valid away from quasifree kinematics , in particular , when final state interaction is taken into account . this is due to the fact , that rescattering `` mixes '' the single proton and neutron @xmath0-photoproduction processes , as is illustrated schematically in fig . [ fig8 ] , taking as an example the @xmath18-interaction term . thus , in the region where the rescattering effects are dominant , the right hand side of ( [ 71 ] ) approximates unity in contrast to the elementary ratio . in general , we can conclude from the above investigations , that the relation ( [ 71 ] ) works well for photon energies not too close to the threshold . as a confirmation , we have plotted in fig . [ fig9 ] the theoretical ratio @xmath162 of the differential cross sections for the reaction @xmath14 without and with rescattering versus the photon lab energy @xmath163 and compare it to the corresponding value for free nucleons . the difference between the ratio for free nucleons and that for the deuteron , obtained within the spectator model , is due to the fact , that to the integration over nucleon momenta not only fast moving active nucleons but also slow ones partly contribute . one can see , that the additional final state interaction encreases even more the ratio when approaching the threshold . however , for the energies considered in @xcite , its role is negligible , and from this point of view the validity of ( [ 71 ] ) is fully justified . the ratio of the neutron to the proton cross sections @xmath164 , corresponding to @xmath157 , which we have used in this work and which reproduces well the @xmath6 data of @xcite , slightly understimates the one presented in ref . @xcite @xmath165 . finally , we would like to make a short comment on the coherent photoproduction process on the deuteron . we first note , that the sensitivity of the break - up cross section to a variation of @xmath166 is much less pronounced than for the coherent reaction , which is proportional to @xmath167 . the present study within the isobar model has shown , that the experimental uncertainity in the @xmath15 total cross section for photon energies between threshold and 720 mev @xcite fixes the parameter @xmath166 within the limits of @xmath168 . in fig . [ fig10 ] we demonstrate the strong sensitivity on @xmath166 of the @xmath0-angular distribution of the coherent reaction , which we have calculated within the conventional impulse approximation . in addition to the @xmath3(1535 ) resonance , the contributions from @xmath169(1520 ) as well as the nucleon pole term and @xmath163-meson exchange in the t - channel were included in the elementary operator ( for further detailes see @xcite ) . we do not touch here upon the question about the role of meson rescattering in the coherent reaction , which was investigated in @xcite . however , one can expect , that due to its two - particle character , this effect will not strongly affect the coherent process in the region of forward @xmath0-angles which is associated with relatively small momentum transfer . as can be seen from fig . [ fig10 ] even the largest permissible value @xmath170 gives a cross section , which is by more than a factor three below the data . only with a rather large isoscalar amplitude corresponding to @xmath171 the impulse approximation could describe the data well . this seeming inconsistency between the coherent and incoherent reaction with respect to the precise value of the parameter @xmath166 certainly needs further experimental and theoretical studies . we have studied incoherent @xmath0-photoproduction on the deuteron in the near threshold region . the elementary photoproduction amplitude , based on a pure isobar model , for which the @xmath172 and @xmath173 vertices have been fitted , gives a good description of the experimental @xmath31 cross section data . as for the deuteron amplitude , we have considered in addition to the spectator - nucleon approach also @xmath18-rescattering in the most important @xmath120 channel and @xmath19-rescattering . while the pure spectator model produces only a small fraction of the observed cross section near threshold , the strong attractive @xmath18-interaction in the resonant @xmath120 state enhances it considerably . the role of @xmath19-rescattering is found to be less important . with inclusion of both rescattering effects we obtain quite a satisfactory agreement with the available data on the @xmath6 differential and total cross sections , using as input for the ratio of isoscalar to proton amplitude @xmath157 . this value is roughly consistent with @xmath174 , extracted from the recently measured ratio @xmath175 . furthermore , we have shown , that this extraction can be justified within the frame of the pure spectator - nucleon model , since the role of final state interaction under the corresponding kinematical conditions is negligible . a. f. is grateful to b. krusche , r. schmidt , m. schwamb and p. wilhelm for fruitful discussions . he would also like to thank the theory group of professor h. arenhvel as well as professor h. strher , monika baumbusch and many scientists of the institut fr kernphysik at the mainz university for the very kind hospitality .

incoherent @xmath0-photoproduction on the deuteron is studied for photon energies from threshold up to 720 mev . the elementary @xmath1 amplitude is constructed in the frame of a conventional isobar model . effects of final state interaction are investigated and their role in the cross section as well as in polarization observables is found to be important . inclusion of such effects leads to a satisfactory agreement with experimental data in the near threshold region . in addition , recent experimental results on the coherent reaction @xmath2 are discussed .

introduction the @xmath1 amplitude the @xmath15 reaction conclusion

a symmetry can be seen as an equivalence of different physical situations @xcite . such an equivalence in quantum theory entails the invariance of a certain set of observables and can be formalised in terms of commutation relations between these operators and a given hamiltonian . the existence of good quantum numbers , also those having no classical counterpart , is a direct consequence of symmetries . it extends the amount of information accessible for researchers studying quantum systems . it is possible to explain an unusual system s behaviour , its properties and dynamics by means of symmetries . selection rules or kramers degeneracy @xcite may serve as a good example here . symmetries not only deepen our understanding of quantum systems but also can be included to engineer their physical realisation more effectively @xcite . in general , the more symmetries recognised ( together with related conserved quantities ) the more different approaches to study the system s dynamics are at our disposal . only in extreme cases , one can meet analytically solvable models ( such as harmonic oscillator , jaynes cummings model or hydrogen atom ) where symmetries can be found easily . in this paper , we consider a quantum model consisting of a two level system ( qubit ) interacting with a single mode bosonic field ( electromagnetic radiation ) with frequency @xmath1 . the hamiltonian of that system is assumed to be of the following form @xmath2 where @xmath3 and @xmath4 are the creation and annihilation operators of the bosonic field . mathematically , this means that @xmath5=\mathbb{i}$ ] . for an experimental characterisation of these operators see @xcite . @xmath6 and @xmath7 denote the two pauli spin matrices . the term @xmath8 stands for the unperturbed energy of the qubit with possible eigenenergies @xmath9 . tunnelling between the corresponding energy levels in the absence of the bosonic field ( spontaneous transition ) is described by @xmath10 . finally , the coupling constant @xmath11 reflects the strength of the interaction between the systems . the above hamiltonian is the well known rabi model @xcite probably the most influential model describing fully quantized interaction between matter and light . although the model originates from quantum optics @xcite , its applications range from molecular physics @xcite , solid state ( see refs . in @xcite ) to the recent experiments involving cavity and circuit qed @xcite . the rabi model can be implemented by means of rich variety of different setups such as josephson junctions @xcite , trapped ions @xcite , superconductors @xcite or semiconductors @xcite , to name a few . despite its simplicity , the hamiltonian of the rabi model can not be diagonalized exactly when @xmath12 . although some progress has been reported recently @xcite , exact analytical formulas for the eigenvalues and corresponding eigenfunction of the hamiltonian ( [ rabi ] ) are still missing . there is a wide spectrum of available approximation techniques including rotating wave approximation @xcite ( leading to the famous jaynes cummings model @xcite ) which allow the eigenproblem to be approached from many different directions . at this point , a question concerning the existence of symmetries in the rabi model ( together with related constants of motion ) arises naturally . provided that @xmath13 , the hamiltonian ( [ rabi ] ) remains unchanged when @xmath14 and @xmath15 ( hence @xmath16 ) . the symmetry operator @xmath17 that generates this transformation ( _ e.g. _ fulfills @xmath18=0 $ ] ) reads @xmath19 , where @xmath20 is the bosonic parity @xcite . this is the well known result : still being unsolvable , the rabi model possesses a discrete symmetry if @xmath13 . when @xmath21 , on the other hand , we can still leave @xmath22 unaffected after changing @xmath14 , @xmath15 if we change the sign of @xmath23 as well ( _ i.e. _ @xmath24 ) . this instantly raises a question : what does the corresponding generator of such transformation , @xmath25 , look like ? unfortunately , this question has not been answered so far . moreover , it was quite recently conjectured @xcite that the rabi model does not possess any symmetry at all ( except the trivial one related to the total energy conservation ) as long as @xmath21 . if that were true , the only self adjoint operator @xmath25 such that @xmath26}=0 $ ] would be the hamiltonian @xmath22 itself . on the basis of the results reported here , we prove that this conjecture is false . in particular , we show how one can find a self adjoint involution @xmath25 , that is @xmath27 , such that @xmath28 . also , we discuss the possibility of the exact diagonalization of the rabi hamiltonian ( [ rabi ] ) . it is worth mentioning that symmetry groups of the time evolution generator ( the hamiltonian @xmath22 in our case ) are larger than those of the corresponding equation of motion ( schrdinger equation : @xmath29 ) . in particular , we could consider the existence of a symmetry @xmath30 which does not necessarily commute with @xmath22 but still assures the same time evolution for two different states : @xmath31 , @xmath32 . of course , this is possible if @xmath33 $ ] . the idea of such dynamical symmetries is interesting by itself , yet it is beyond the scope of our current considerations and wo nt be pursued any further in this work . let us begin with formal rewriting of the rabi hamiltonian ( [ rabi ] ) as a matrix with operator entries : @xmath34 customarily , the parameters @xmath35 and @xmath23 denote @xmath36 and @xmath37 respectively . @xmath38 stands for the identity on the bosonic hilbert space @xmath39 . the matrix representation of the rabi model given in ( [ bom ] ) is established via a natural isomorphism @xmath40 . usually , such an identification is invoked in order to simplify purely algebraic calculations ( see _ e.g. _ this is not the reason why we use this idea here . instead , we are going to attack the problem in question by using a concept of block operator matrix @xcite in conjunction with its relation to an operator riccati equation @xcite . first however , we would like to clarify some technical aspects concerning the rabi matrix ( [ bom ] ) ( _ e.g. _ its domain @xmath41 ) . one should mention that this is not a primary issue in many papers addressing physical aspects of the rabi model . needless to say , one can not take the advantage of very powerful existential mathematical theorems ( _ e.g. _ the famous banach fixed point theorem @xcite ) in such cases simply because it is not known whether the premises of these statements are met . in a first step toward constructing @xmath25 , we define domains @xmath42 on which both operators @xmath43 are self adjoint . since the off diagonal elements of @xmath22 are bounded , we have @xmath44 on @xmath45 . as both @xmath3 and @xmath4 are unbounded , the canonical commutation relation holds only on some ( dense ) subspace @xmath46 of @xmath47 . let us assume that @xmath48 is a dense set on which @xmath3 and @xmath4 are adjoint to each other _ i.e. _ , @xmath49 and @xmath50 . at this point , it is not obvious that the subspaces having the desired properties exist at all . an interested reader can find the detailed construction of @xmath51 _ e.g. _ in @xcite . here , we briefly summarise what was covered therein . we have @xmath52 where @xmath53 is the canonical ( orthonormal ) basis in @xmath54 ( @xmath55 ) . considering the fact that @xmath3 , @xmath4 and @xmath56 ought to produce normalizable states , the above definitions seem natural . having ( [ dom ] ) in place , we define @xmath57 it follows immediately from ( [ aa ] ) that @xmath58 and @xmath59 . interestingly , the latter relations serve as the very definition of the creation and annihilation operators in most textbooks on quantum mechanics . a definition like this may be well motivated physically , yet it has at least one serious mathematical drawback . namely , it introduces closeable operators which are not closed . this leads to a variety of technical difficulties typical for such classes of operators . one can avoid them by taking the closures ( [ aa ] ) as proper definitions , instead . a basic result from operator theory ( see _ e.g. _ , theorem 4.2.7 in @xcite ) states that if @xmath60 is closed on @xmath61 then @xmath62 is positive , self - adjoint and its domain is a core of @xmath60 ( _ i.e. _ @xmath60 is the closure of its restriction @xmath63 ) . on @xmath64 , the operators @xmath43 can be written as @xmath65 and as a result , they are both self adjoint and their common domain @xmath64 is a core of both @xmath3 and @xmath4 . in conclusion , the rabi hamiltonian ( [ bom ] ) is well defined and self adjoint on @xmath66 . after discussing technical nuances concerning the rabi model , we introduce a quadratic second order operator equation , known as the riccati equation , which has the following form @xmath67 many of the relevant problems related to the rabi model ( [ rabi ] ) , including its exact diagonalization , can be reduced to the mathematical questions concerning solvability of this equation . there is more than one notion of a solution when equations with operator coefficients are involved . in hilbert spaces , one can define a solution in terms of the scalar product ( weak solution ) . on the other hand , one may require for operators to be equal when they produce the same results while acting on the same states . these kind of solutions , which are of great importance in quantum mechanics , are known as strong ones . let us briefly clarify these two notions for the riccati equation in question . a bounded operator @xmath68 acting on a hilbert space @xmath69 is called a weak solution of the riccati equation ( [ ricc ] ) if @xmath70 a bounded operator @xmath68 acting on @xmath69 such that @xmath71 and @xmath72 is a strong solution of ( [ ricc ] ) . of course , a strong solution is also a weak solution . it is often easier to prove the existence of a weak rather than a strong solution . however , strong solutions , especially in quantum mechanics , are those which we are interested in . fortunately , the two notions are in fact equivalent @xcite . nevertheless , there is no general method of finding either weak or strong solutions to a particular riccati equation . for this reason , the following theorem , which provides criteria of solvability , is of great importance to us . [ mt ] let @xmath43 be ( possibly unbounded ) self adjoint operators acting on domains @xmath73 in a separable hilbert space @xmath69 . let us also assume that @xmath74 and @xmath75 are bounded operators on @xmath69 . if the spectra @xmath76 are disjoint , i.e. , @xmath77 and if @xmath78 , @xmath75 satisfy the ` smallness assumption ' @xmath79 then the riccati equation @xmath80 has a unique weak solution @xmath68 in the ball @xmath81 satisfying an estimate @xmath82 in particular , if @xmath83 then @xmath68 is a strict contraction , that is , @xmath84 . an elegant and compact proof of this statement , based on the banach fixed point theorem , can be found in @xcite . now , let us prove our main result . first , we show that the existence of a solution of the riccati equation ( [ ricc ] ) implies the existence of an operator generating a symmetry in the system ( [ bom ] ) . second , we argue that under certain conditions imposed on the parameters @xmath35 , @xmath23 , @xmath1 this equation is weakly solvable . [ symm ] let us assume that there exists a weak ( and hence strong ) solution @xmath68 of the riccati equation ( [ ricc ] ) . then there also exists a self adjoint involution @xmath25 such that @xmath85 where @xmath22 is given by ( [ bom ] ) . moreover , the generator @xmath25 in terms of @xmath68 reads @xmath86 let @xmath87 be the graph of @xmath68 , that is @xmath88 @xmath68 is a strong solution of ( [ ricc ] ) thus @xmath89 ( by definition ) and @xmath90 for @xmath91 . therefore , @xmath92 that is @xmath93 . making use of the same arguments , one can verify that @xmath94 , which is given by @xmath95 is @xmath96invariant as well . @xmath68 is bounded and thus its graph forms a closed subspace of @xmath97 and hence the decomposition @xmath98 holds true . therefore , each state @xmath99 of the composite system can be uniquely decomposed @xmath100 where @xmath101 and @xmath102 . let @xmath103 be a projection onto @xmath87 . then it follows that @xmath104 and @xmath105 . assuming for a moment that @xmath106 , we obtain @xmath107 therefore , @xmath108 for all @xmath109 . the inverse @xmath110 exists and it is a bounded self adjoint operator on @xmath39 . thus , @xmath111 can be expressed as @xmath112 it is a matter of straightforward calculations to see that ( [ pp ] ) indeed projects onto @xmath87 . due to the fact that @xmath113 , the only question which we need to address to conclude the proof is whether @xmath114 is again in @xmath41 for @xmath99 . because @xmath68 is a weak ( and hence strong ) solution of ( [ ricc ] ) , we have @xmath115 . moreover , the function @xmath116 is continuous on @xmath64 for every @xmath117 . indeed , it follows from ( [ weak ] ) that @xmath118 as a result , @xmath119 , _ i.e. _ @xmath120 and therefore @xmath121 . @xmath122 is invertible , hence @xmath123 . in summary , @xmath124 which concludes the proof . let us assume that @xmath23 , @xmath1 , @xmath12 satisfy the following conditions @xmath125 then there exists a unique weak ( hence strong ) solution of the riccati equation ( [ ricc ] ) such that @xmath84 . as a result , there is a @xmath0 symmetry with respect to which the rabi model is invariant . the generator of this symmetry is given by ( [ symm ] ) . @xmath126 is self adjoint for @xmath127 thus the unitary weyl operator @xmath128 is well defined . moreover , @xmath129 and therefore @xmath130 by virtue of @xmath131 ( keep in mind that @xmath132 ) , we have @xmath133 if @xmath134 is not multiple of @xmath1 then the distance @xmath135 therefore , the spectra @xmath76 are disjoint _ i.e. _ , the condition ( [ d ] ) holds true . in addition , both the smallness assumption ( [ small ] ) and ( [ cont ] ) imposed on the off diagonal elements are satisfied as long as @xmath136 . according to lemma [ mt ] , there is exactly one solution of the riccati equation ( [ ricc ] ) and it is a strict contraction ( @xmath137 ) . the second statement of the theorem follows immediately from theorem [ symm ] . we begin with the @xmath13 case in which the spectra ( [ spec ] ) overlap and hence the separability condition ( [ d ] ) is not satisfied . therefore , one can not invoke lemma [ mt ] to establish the existence of a solution to the riccati equation ( [ ricc ] ) . however , the spectra @xmath76 in that particular case are identical and @xmath43 can be transformed one into another by the same bosonic parity operator that generates the symmetry @xmath17 . this is not an accidental coincidence as @xmath138 is a solution of the riccati equation ( [ ricc ] ) . indeed , @xmath139 from which it follows immediately that @xmath138 is bounded and @xmath140 . note , if @xmath141 are square summable , @xmath142 , so are @xmath143 . in the light of ( [ aa ] ) , we obtain @xmath144 as well as @xmath145 and finally @xmath146 . and because @xmath138 is a self adjoint involution , it solves the riccati equation ( [ ricc ] ) as stated . at this point , we would like to make some remarks . first and foremost , @xmath138 is not a unique solution of the riccati equation ( [ ricc ] ) . for instance , @xmath147 also satisfies this equation . second , the symmetry generator @xmath148 from theorem [ symm ] reads @xmath149 when @xmath150 as one may expect . if the conditions ( [ ass ] ) are met , in particular for @xmath21 , the spectra @xmath43 are separated and the riccati equation ( [ ricc ] ) possesses exactly one solution @xmath68 . according to theorem [ symm ] , this solution corresponds to a symmetry generator @xmath25 . the only problem is that @xmath68 is unknown . one can attempt to simplify the problem by putting @xmath151 , where @xmath152 + 2\beta{\text{y}}_{\beta}-\alpha{\text{p}}=0,\ ] ] and @xmath153 is redefined so that it reads ( [ compact ] ) for @xmath13 . this equation becomes trivial and its solution reads @xmath154 when @xmath13 . on the other hand , as long as @xmath155 , under ( [ ass ] ) , the premises of lemma [ mt ] are satisfied . hence , a unique @xmath156 exists and @xmath157 . moreover , if the inverse @xmath158 exists as well then @xmath159 + 2(-\beta){\text{y}}_{\beta}^{-1}-\alpha{\text{p}}=0,\ ] ] and therefore @xmath160 . although we can not solve ( [ nricc ] ) either , the latter equality indicates the class which @xmath156 belongs to . one can also verify that the operator @xmath156 is not self adjoint provided it is a function of @xmath153 and it can not be anti self adjoint ( @xmath161 ) indeed , if @xmath153 such that @xmath162 exists , ( [ nricc ] ) would imply the following separation into a self adjoint and anti self adjoin part @xmath163=0.\ ] ] both these equations can be solved separately , but the solutions do not agree with each other unless @xmath13 . similar arguments show that the condition @xmath161 is necessary for @xmath164 . this contradicts ( [ nricc ] ) even when @xmath13 . solutions of the riccati equation ( [ ricc ] ) can also be used to obtain the eigenfunctions and corresponding eigenvalues of the rabi hamiltonian . let us briefly discuss the idea . both @xmath87 and @xmath94 are @xmath22invariant . thus , if @xmath165 is an energy eigenstate then either @xmath166 or @xmath167 . actually , we can say more than that . let @xmath168 and @xmath169 be defined on @xmath64 . together with ( [ include ] ) , this gives @xmath170 provided @xmath171 is in @xmath87 . also , one can verify that all eigenstates from @xmath94 are of the form : @xmath172 it can be proven that @xmath173 are self adjoint on hilbert spaces @xmath174 and @xmath175 , respectively @xcite . moreover , @xmath176 and the following similarity relation holds @xmath177 the above block diagonal form of @xmath22 extends the notion of the parity chains introduced in @xcite . we have recognised a symmetry of the rabi hamiltonian and constructed its generator @xmath148 . although this symmetry is nonlocal ( unlike _ e.g. _ @xmath178 ) , it is a self adjoint involution . therefore , it can be considered as a generalised parity of the rabi model . invoking physical nomenclature , the rabi model is invariant with respect to this parity or it has an _ unbroken _ @xmath0 symmetry . in literature , the latter terminology is often used in a different ( local ) context where it is stated that the @xmath21 case corresponds to a broken @xmath0 symmetry ( because @xmath179\not=0 $ ] ) . our aim was to generalise the local parity combined by the parity operators of the individual subsystems : @xmath7 and @xmath180 to the nonlocal one for @xmath21 . our results are not of purely existential character . by means of a solution to an operator riccati type equation , we have derived an explicit formula for the generator @xmath25 and formulated conditions ( range of parameters ( [ ass ] ) ) guaranteeing its existence . the question whether the generator @xmath25 can exist under conditions other than ( [ ass ] ) remains open . this problem is a subject of our current intensive investigation . at this point one should mention that usually the existence of a discrete symmetry in a quantum system is not enough by itself to fully understand its dynamics . also , there is no obvious and direct guideline suggesting usefulness of symmetries given by discrete operators , especially nonlocal ones , in construction of solutions to the equations of motion of composite systems . however , discrete symmetries , local or not , allow the decomposition of the system hilbert space into two subspaces with states having certain properties . one can then seek for the solution to the equation of motion in the individual subspaces ( and then try to combine the results to obtain a full solution ) . for the rabi model , in the case of local parity , this idea can be realised in terms of so called parity chains @xcite . the generalisation to the nonlocal case can by accomplished by means of block diagonalization according to ( [ pch ] ) . the latter formula may also serve as a good starting point for developing new analytical approximations or numerical treatment of the eigenproblem @xcite . moreover , nonlocal discrete symmetries can help in classification and grouping of known solutions @xcite . they can also be used in constructing new solutions from the ones which are already known such as juddian solutions @xcite or so called quasi exact solutions @xcite . symmetries of the type presented here can also serve as a tool helping to verify certain conjectures concerning solutions of the rabi model such as the celebrated reik conjecture @xcite . we would like to emphasise that there is always a physical context ( beyond mathematics ) of studying symmetries ( both local and nonlocal ) in physics . for instance , there is a connection between symmetries of a quantum system and good quantum numbers in that system @xcite . any measurement confirming conservation of such numbers confirms , at least partially , correctness of the model ( _ i.e. _ , whether a given choice of the hamiltonian properly describes the system ) . as ` quantum phenomena do not occur in a hilbert space , they occur in a laboratory ' @xcite , the more symmetries to our disposal the more tests can be performed . this ultimately verifies our understanding of quantum systems and their behaviour . it seems that an inability to solve the riccati equation when @xmath21 is the core reason why the symmetry ( [ j ] ) hasnt been recognised earlier . although the solution of this equation exists as we have proved , it may not be expressible by standard ( well known ) operators . in that case , it is very unlikely to find the explicit form of @xmath148 also by means of different methods regardless of their nature . on the other hand , the riccati equation can easily be solved in terms of the well - known bosonic parity when @xmath13 . as one may expect , the corresponding generator @xmath181 has been known all along . the solvability problem of the riccati equation can also be related to the question regarding diagonalization of the rabi model . in this paper , we have investigated the possibility of finding the eigenvalues and eigenvector of the rabi hamiltonian . we have not offered full resolution , yet compact and exact expressions have been derived that , to some extent , simplify the problem . although our analysis was mainly focused on the rabi model , the presented scheme of diagonalization can be extended to general qubit environment models . 10 url # 1#1urlprefix[2][]#2 peres a 1993 _ quantum theory : concepts and methods _ ( kluwer academic publishers , london ) roberts b w 2012 _ phys . rev . _ a http://dx.doi.org/10.1103/physreva.86.034103[*86*(3 ) 034103 ] crespi a , longhi s and osellame r 2012 _ phys . lett . _ http://dx.doi.org/10.1103/physrevlett.108.163601[*108*(16 ) 163601 ] kumar r , barrios e and kupchak c _ et al _ 2013 _ phys . lett . _ http://dx.doi.org/10.1103/physrevlett.110.130403[*110*(13 ) 130403 ] rabi i i 1936 _ phys . rev . _ http://dx.doi.org/10.1103/physrev.49.324[*49*(4 ) 324328 ] rabi i i 1937 _ phys . _ http://dx.doi.org/10.1103/physrev.51.652[*51*(8 ) 652654 ] vedral v 2006 _ modern foundations of quantum optics _ ( imperial college press , london ) thanopulos i , paspalakis e and kis z 2004 _ chem . phys . lett . _ http://dx.doi.org/10.1016/j.cplett.2004.03.129[*390 * 228 235 ] irish e k 2007 _ phys . lett . _ http://dx.doi.org/10.1103/physrevlett.99.173601[*99*(17 ) 173601 ] englund d , faraon a and fushman i _ et al _ 2007 _ nature _ http://dx.doi.org/doi:10.1038/nature06234[*440 * 857861 ] niemczyk t , deppe f and huebl h _ et al _ 2010 _ nature physics _ * 772776 ] sornborger a t , cleland a n and geller m r 2004 _ phys . _ a http://dx.doi.org/10.1103/physreva.70.052315[*70*(5 ) 052315 ] leibfried d , blatt r and monroe c _ et al _ 2003 _ rev . phys . _ http://dx.doi.org/10.1103/revmodphys.75.281[*75*(1 ) 281324 ] johansson j , saito s and meno t _ et al _ 2006 _ phys . http://dx.doi.org/10.1103/physrevlett.96.127006[*96*(12 ) 127006 ] hennessy k , badolato a and winger m _ et al _ 2007 _ nature _ http://dx.doi.org/doi:10.1038/nature05586[*445 * 896899 ] braak d 2011 _ phys . lett . _ http://dx.doi.org/10.1103/physrevlett.107.100401[*107*(10 ) 100401 ] ziegler k 2012 _ j. phys . a : math . theor . _ http://dx.doi.org/10.1088/1751-8113/45/45/452001[*45 * 452001 ] romanelli a 2009 _ phys . _ a http://dx.doi.org/10.1103/physreva.80.014302[*80*(1 ) 014302 ] gardas b 2011 _ j. phys . a : math . theor . _ http://dx.doi.org/10.1088/1751-8113/44/19/195301[*44 * 195301 ] chen q h , wang c and he s _ et al _ 2012 _ phys . _ a http://dx.doi.org/10.1103/physreva.86.023822[*86*(2 ) 023822 ] langer h and tretter c 1998 _ j. operator theory _ * 39 * 339359 egoriv a i 2007 _ riccati equations _ ( pensoft publishers , bulgaria ) gardas b 2010 _ j. math . http://dx.doi.org/10.1063/1.3442364[*51 * 062103 ] gardas b 2011 _ j. math . phys . _ http://dx.doi.org/10.1063/1.3574889[*52 * 042104 ] gardas b and puchaa z 2012 _ j. math . _ http://dx.doi.org/10.1063/1.3676309[*53 * 012106 ] reed m and simon b 1980 _ method of modern mathematical physics _ ( academic press , london ) berezin f a 1966 _ method of second quantization _ vol 24 ( academic press , ny ) bratteli o and robinson w 1997 _ operator algebras and quantum statistical mechanics 2 _ ( springer - verlag , berlin ) szafraniec f h 1998 _ contemp . math . _ * 212 * 269276 blank j , exner p and havliek m 2008 _ hilbert space operators in quantum physics _ ( springer , ny ) albeverio s and motovilov a k 2011 _ trans . moscow math . soc . _ * 72 * 4577 albeverio s , makarov k a and motovilov a k 2003 _ can . j. math . _ * 55*(3 ) 449503 he s , wang c and chen q h _ et al _ 2012 _ phys . _ a http://dx.doi.org/10.1103/physreva.86.033837[*86*(3 ) 033837 ] albert v v , scholes g d and brumer p 2011 _ phys _ a http://dx.doi.org/10.1103/physreva.84.042110[*84*(4 ) 042110 ] judd b r 1977 _ j. chem . * 11741179 ] emary c and bishop r f 2002 _ j. phys . a : math . and gen . * 8231 ] r , koca m and ttncler h 2002 _ j. phys . a : math . and gen . _ http://dx.doi.org/doi:10.1088/0305-4470/35/44/311[*35 * 9425 ] reik h g and doucha m 1986 _ phys . lett . _ http://dx.doi.org/10.1103/physrevlett.57.787[*57*(7 ) 787790 ] gardas b and dajka j 2012 initial states of qubit - boson models leading to conserved quantities ( _ preprint _ ) , _ accepted _

it is recognised that , apart from the total energy conservation , there is a nonlocal @xmath0 and a somewhat hidden symmetry in this model . conditions for the existence of this observable , its form , and its explicit construction are presented .

introduction main result discussion summary

the cleo detector is undergoing a major upgrade to phase iii in parallel with a significant luminosity increase of the cesr electron - position collider@xcite . these improvements will make it possible to make precision measurements , especially of cp violation and rare b decays@xcite . one of the main goals of the cleo iii upgrade is to have excellent charged particle identification . in order to achieve this goal , we have designed and are constructing a ring imaging cherenkov ( rich ) detector , based on the proximity focusing approach@xcite . it needs no optical focusing elements , and requires that the radiator is relatively thin compared to the distance of expansion gap between the photon detector and radiator . the detector can be designed flat and compact to fit in limited space . our design is based on the pioneer work done by the fast - rich group@xcite . when a charged particle passes through a medium , it radiates photons if the velocity of the particle is faster than light speed in the medium . the direction of the radiated photons with respect to track is given by @xmath0 , where @xmath1 is the index of the medium , and @xmath2 . to distinguish between hypotheses for a track at momentum @xmath3 , the formula @xmath4 is applicable . in a medium of @xmath5 , a very fast particle ( @xmath6 ) radiates cherenkov photons at angle about 840 mrad . for cleo iii rich design , our goal is to separate charged pions and kaons ( @xmath7 ) at @xmath3 = 2.65gev / c , the highest momentum that need be considered at cleo for b decays . the cherenkov angle difference of @xmath8 and @xmath9 at this momentum is 14.4 mrad . in order to obtain very low fake rate with high efficiency , we would like to achieve a @xmath10 separation between @xmath8 and @xmath9 at all the momenta relevant for b decays with the addition of a @xmath11 de / dx contribution . this requires that the rich provides 4.0 mrad cherenkov resolution per track . thus our benchmark is average 12 photoelectrons per track each with a resolution of @xmath1214 mrad . the overall structure of the cleo iii rich detector is cylindrical as shown in fig . [ richdetector ] . it consists of compact photon detector modules at the outer radius and radiator crystals at the inner radius , separated by an expansion gap . the whole detector is divided into 30 sectors in the azimuthal direction . the rich resides between the drift chamber and the electromagnetic calorimeter . the detector occupies a radial space between 80 and 100 cm in radius , is 2.5 m long , and covers 81% of the solid angle @xcite . -.2 cm the choice of the material for the radiator and detector windows is driven by the choice of triethylamine ( tea ) as the photo - sensitive material . the @xmath13 gas mixture has a finite quantum efficiency in the vuv region ( 135 - 165 nm ) , which requires the use of fluoride crystals to insure transparency . the cherenkov angle resolution is dominated by chromatic dispersion in the emission of light @xcite . we choose lif as radiator for its smaller chromatic dispersion in vuv region , of all the fluorides . the photon detector windows are made of @xmath14 . the lif radiators will be mounted on an inner carbon fiber cylinder as a@xmath15 array , each radiator is @xmath16 and 10 mm thick . the normal choice of radiator shape is a flat plate . however , as the total reflection angle of 150 nm photon in lif ( @xmath17 at 150 nm ) is about @xmath18 and the cherenkov angle is close to @xmath19 , the photons radiated by track at normal incidence will be trapped as shown in fig . [ radiator ] . instead of tilting the plates to allow the light to get out near the center of the detector , we developed radiators with a sawtooth pattern on its outer surface , which allows the cherenkov photons to escape the radiator@xcite . we will use such sawtooth radiators for the center four of the 14 rings . -.2 cm the expansion gap ( 16 cm ) is filled with high purity nitrogen gas . the volume is well sealed to minimize the contamination of @xmath20 and @xmath21 , which absorb vuv photons . the photon detector is multi - wire proportional chamber with cathode pad readout . it is filled with @xmath22 gas bubbled through liquid tea at @xmath23 ( 7% ) . the entrance windows are 2 mm thick @xmath14 , coated with 100 @xmath24 wide silver traces to act as an electrode . photons with a wavelength between 135 nm and 165 nm can generate single photoelectrons in the @xmath13 mixture . the photoelectron drifts toward 20 @xmath25 diameter au - w anode wires , where avalanche multiplication takes place . charge is induced on an array of @xmath26 cathode pads , allowing a precise reconstruction of the position of the cherenkov photon at the detector plane . the chamber is asymmetric , with the wire to pad distance of 1 mm and the wire to window distance of 3.5 mm , to improve the wire cathode pad coupling . the choice of the readout electronics for cleo iii is governed by several considerations @xcite . first , the charge induced by a single photoelectron avalanche at moderate gain follows an exponential distribution as shown in fig . [ charge ] . as the most likely charge is zero , it is necessary to have a very low noise system so that high efficiency is achieved . furthermore , we want to reconstruct charged tracks accurately . we expect the pulse height from the charged track to be at least a factor 20 higher than the mean gain for a single photoelectron . in addition , high resolution analog readout allows a more effective suppression of cluster overlaps . lastly , in the final system , we will have 230,400 channels to be readout . an effective zero suppression algorithm is necessary to have a manageable data size . -.2 cm we use the va_rich chip as the front end processor . it is a custom - designed 64-channel vlsi chip based on the va series developed for si microchip readout @xcite . it features low noise and large dynamic range . with a 2 pf input capacitor , its equivalent noise charge is about 150 electrons . linearity is excellent up to @xmath27 input . two chips are wire bonded to one hybrid circuit . sixty hybrids are mounted on each photon detector . the signal travels over a 6 m long cable from the detector to a vme databoard . the databoard includes receivers , 12-bit differential adcs , bias circuitry for the va_rich chips , a circuit generating the timing sequence needed by the va_rich chip , and vme interface . in the final version , the databoards will include a dsp based algorithm for common mode subtraction that will make the system less vulnerable to coherent noise sources . for the beam test , 8 prototype databoards in one crate are used . these databoards were not equipped with coherent noise subtraction , and all the channels were readout to monitor and correct for coherent noise fluctuations . in order to test our design of the rich detector , a comprehensive beam test was performed . the beam test setup is shown in fig . [ beamtest ] . -.2 cm the first two photon detectors built for the final cleo iii rich were mounted in an aluminum box . in addition , one plane and two sawtooth lif radiators were mounted in the box at the same distance from the photon detectors as designed for cleo iii rich . the box was sealed and flushed with pure nitrogen . a schematic of the box can be found in ref . @xcite . the beam test was performed in a muon halo beam in the meson east area of fermilab . apart from the rich itself , the system consisted of trigger scintillator counters , and a tracking system . we used two sets of mwpcs to measure the track position and angle . they provided 0.7 mm spatial resolution per station and a track angle resolution of 1 mrad . to simulate different track incidence angles , the rich detector box was rotated at various polar and azimuthal angles . during the three week beam test , the photon detectors were run at an average gain of @xmath28 . the total electronic noise was about 1000 electrons . after coherent noise subtraction , the residual incoherent noise was 400 electrons , providing an average signal - to - noise ratio for single photon signal of 100:1 . an important effect that we discovered with the beam test was the temperature sensitivity of the analog + 2v and 2v , necessary to bias the va_rich . [ volt_t ] shows this effect . this , in turn , affected the bias configuration of the chip and its pedestal . the @xmath29 voltage regulator design has been changed to eliminate this problem . fig . [ event ] shows the displays of two single events from the plane radiator at @xmath30 track incidence and from a sawtooth radiator at normal incidence . for the plane radiator , the image is a single arc as shown . while for the sawtooth radiator , two arcs on opposition sides of the charged track are visible , with the lower one truncated due to acceptance . the acceptance for images from plane radiators is about 85% of the maximum possible acceptance . with the necessary chamber mounts and window frames , 85% is the maximum possible acceptance . the acceptance for sawtooth images was about 50% in this setup , and should be about 85% in the final system . -.2 cm channels recording pulse height above a threshold of @xmath31 noise are selected for further analysis . the first step in the analysis is clustering . the center of the cluster is treated as the location of a photoelectron . about 2.2 pads per cluster are found , with 1.1 photoelectrons per cluster , due to some unresolved photon overlaps . for each photoelectron , the trajectory is optically traced back through all media , with the assumption that it originates from the mid - point of radiator . from this propagation path , the cherenkov angle is calculated . we extract the number of photoelectrons per track by fitting the single photon spectrum taking into account the background . we invoke a @xmath32 cut so as not to include non - gaussian tails . the numbers of photoelectron per track at different incident angle are shown in fig . [ photon ] ( a ) . for plane radiator , the average number of photoelectrons per track is between 12 and 15 , varying with incident angle . for the sawtooth radiator the number is lower due to limited acceptance . an extrapolation made to estimate the performance of the final system predicts 17 - 20 photoelectrons . [ photon ] ( b ) shows the measured cherenkov angle resolution per photon @xmath33 . the resolution is between 11 and 15 mrad . -.2 cm the parameter that shows the particle identification power of this system is the cherenkov angle per track @xmath34 . the resolutions are summarized in fig . [ track ] for both plane and sawtooth radiators . the measured @xmath34 increases with track angle@xcite . a monte carlo study is performed to simulate the resolution . the monte carlo represents the plane radiator data well . for the sawtooth radiator , the data show a worse resolution than the simulation . there are several sources in the simulation that could account for this discrepancy , such as the tracking error , beam background , and imperfect knowledge of the sawtooth shape . -.2 cm the monte carlo study also shows that the contribution to @xmath34 from the tracking errors is significant , especially for large incident track angles . this error comes in due to the photon emission point error . for the plane radiator at @xmath35 , the contribution is about 4 mrad , while the total error is about 5.8 mrad . in the final cleo iii system , the tracking system is expected to be substantially better than the system used in the beam test . in order to have approximate projection of the performance of the cleo iii rich , we extrapolate the resolution from the beam test results , with the assumption of no tracking error and the acceptance in final system . this resolution is shown as the filled circles in fig . [ track ] . the beam test results indicate that the specifications of cleo iii rich design are fulfilled . in particular , the cherenkov angle resolution per track reaches 4 mrad , which will provide @xmath36 separation in cleo iii @xcite . the rich detector is on the final stage of construction . we expect the completion of the project in summer 1999 . we would like to thank fermilab for providing us with necessary infrastructure and dedicated beam time . we give particular thanks to chuck brown and other colleagues from e866 for their hospitality . more details can be found in r.j . mountain et al . , the cleo - iii ring imaging cherenkov detector , " invited talk by at the 3rd international workshop on ring imaging cherenkov detectors , " a research workshop of the israel science foundation , ein - gedi , dead - sea , israel , nov . 15 - 20 , 1998 , ( hep - ex/9902023 ) .

we are constructing a ring imaging cherenkov detector ( rich ) for the cleo iii upgrade for precision charged hadron identification . the rich uses plane and sawtooth lif crystals as radiators , mwpcs as photon detectors with tea as the photo - sensitive material , and low - noise viking readout electronics . results of a beam test of the first two out of total 30 sectors are presented .

introduction detector design and construction electronics beam test setup beam test electronics performance results summary and outlook acknowledgments

the cause of the long - term variability in quasars is still a matter of debate . unlike the short time - scale variations ( on the order of days ) , which are adequately described in terms of relativistic beaming effects , the variations at much longer time - scales ( years to decades ) are less understood . current scenarios under consideration are ranging from source intrinsic variations due to active galactic nucleus ( agn ) accretion disk instabilities ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) , and possible bursts of supernovae events close to the nucleus ( e.g. , * ? ? ? * ; * ? ? ? * ) , to source extrinsic variations due to micro - lensing events along the line - of - sight to the quasar ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? see also the review article by @xcite . determining which of the various proposed mechanisms actually dominates quasar variability is best done by studying it toward the longest possible time - baselines . depending on the mechanism , each has markedly different variability `` power '' at the longer time - scales ( e.g. , * ? ? ? this means that if one would have a quasar monitoring sample that is both large enough , and covers a large enough time - baseline , one could address these issues adequately . unfortunately , given the nature of monitoring programs , this is not something that can be started overnight . the longest quasar light - curve monitoring programs are on the order of 20 years ( e.g. , * ? ? ? * ) , and will take a long time before they are expanded significantly in time - baseline . the way around this is by using historic photographic plate material , in combination with a recent survey . like in our previous paper ( * ? ? ? * hereafter paper i ) , we chose to use the sloan digital sky survey ( sdss ) , data release 2 ( dr2 ) , in combination with the historic second generation guide star catalog ( gsc2 , * ? ? ? * ) and the palomar optical sky survey ( poss , * ? ? ? * ) . this allows for photometric information on the quasars spanning up to 50 years . the downside is that , unlike the monitoring programs , we have typically a very sparse light - curve sampling _ per quasar_. however , since we will have a very large number of them , the sampling across the complete database will be very good . this obviously only works if the variability of the quasars is due to a mechanism common to all quasars . we proved the validity of this concept in paper i , and recently a similar approach has been taken by @xcite . the paper is outlined as follows : in [ calib ] , we introduce the quasar sample , and we will argue that it can be considered a representative sample of the overall quasar distribution . section [ photometriccalibration ] goes through the careful calibration steps needed before one can properly start interpreting the results . the method outlined is in principle the same as in paper i , but since the sample is much larger , it does allow for some enhanced corrections . in [ introstructurefunction ] , we will introduce the variability diagnostic used throughout the paper : the structure function ( hereafter sf ) . this measure has been used extensively in the literature , and allows for easy and direct comparison with long - term variability studies based on the monitoring of individual quasars ( e.g. , * ? ? ? in addition , @xcite modeled sf behavior depending on the intrinsic variability mechanism . clear differences in the sf curves are expected depending on whether the dominant variability is due to either bursts of supernovae close to the nucleus , instabilities in the accretion disk , or intervening micro - lensing events . throughout the paper we will refer back to the predictions made in @xcite . section [ stellarstructurefunction ] describes the results of the calibration on the stellar sf . the purpose of this detailed section is twofold : first , it reflects the level of data - quality we have attained with the calibration method , and secondly , it identifies subtle effects on the data that may have gone unnoticed by just focusing on the quasar sf . among other things , the clear differences in photometric data - quality between the poss and gsc2 surveys only shows up significantly in the stellar sf . also , the malmquist bias signal is clearly seen in the stellar sf , whereas it is masked ( and indeed washed out ) in the quasar sf by the light - curve asymmetry signal ( cf . [ lcasymmetry ] ) . section [ qsf ] and its subsections detail the results we obtained for our quasar sample . these are discussed in light of the existing literature data and our last section on sf modeling in [ modeling ] . in order to significantly improve on our work in paper i ( using 3791 quasars from the sdss early data release ) , we had to wait until the later releases would increase the quasar sample by a large amount . this was accomplished by the two subsequent data releases , which expanded the database first to 16908 , and then to 35165 quasars ( data release 2 ) . this dr2 is described in detail in @xcite . in addition , we added all the 2df quasars @xcite that are covered by the dr2 , but are not among the 35165 in their quasar database . this increases the sample size to 41578 quasars , all of which have accurate and recent sdss photometric information . however , since we are interested in historic variability , we had to remove the 187 quasars that were not included in either the poss or the gsc2 catalogs . this leaves us with the final sample of 41391 quasars . table 1 has the exact break - down of photometric information on this sample . [ sampletable ] .quasar sample [ cols=">,<",options="header " , ] figure [ brightdist ] shows the @xmath1-band magnitude of the sample as function of redshift . the faintest quasars are about @xmath1=21 , except the very highest redshift quasars ( @xmath2 ) , which clearly have been selected using different criteria . the lower redshift sources ( @xmath3 ) also show a correlation between their redshift and optical magnitude . at these redshifts the quasar host galaxy is contributing significantly to the overall luminosity , and progressively more so with decreasing redshift . this non - variable host galaxy component will lower the relative variability amplitude of the agn . if , for instance , the agn varies intrinsically by 10% , placing it inside a galaxy with the same magnitude will lower the variability of the combined system to 5% . the optical variability of an individual quasar is not just a function of agn luminosity relative to its host galaxy luminosity , it also depends on the redshift of the source . first , the contrast between the agn and its host galaxy increases dramatically toward the restframe blue and uv wavelengths ( as probed by the passbands even at moderate redshifts ) . second , the @xmath4 cosmological surface brightness dimming factor affects the extended galaxy more than the point - source agn contribution , again increasing the contrast between the two components . both these redshift dependent trends diminish the unwanted variability - lowering effect by the host galaxy , and based on fig [ brightdist ] , it does not appear to contribute beyond @xmath5 . the bulk of our quasars ( 36 802 ) have redshifts between 0.4 and 4.0 ( marked in fig . [ brightdist ] ) , whereas 4424 ( or about 10% of the sample ) are at redshifts below 0.4 and might potentially be affected by their host galaxies . a direct comparison between the results with and without the low redshift data - set did not yield any significant differences , except at the longest time - lags in the @xmath1-band in particular ( see [ galcont ] ) . the quasars added from the 2df survey have a different brightness distribution ( cf . [ brightdist ] , light - colored points ) than the sdss quasars , mainly because of different selection criteria . this difference remains , even if we put the 2df and dr2 overlap quasars back into the 2df sample , as is clearly illustrated in fig . [ brighthist ] . both the sdss and 2df use multi - band photometric criteria to preselect for quasar candidates . in addition , the sdss sample has been augmented by targeting first and rosat counterparts as well @xcite , and in general uses a less restrictive color cut . the fraction of radio- and x - ray loud sources is relatively low ( 2692 first counterparts within 2 , and 479 rosat counterparts within 15 ) , so it does not significantly alter the distribution . figure [ samplediff ] illustrates the differences between the sdss and 2df selected quasars . the sources have been plotted on the @xmath6 and @xmath7 plane , which is one of the color - color diagrams used in selecting sdss quasar candidates @xcite . the 2df quasar candidates were selected from scanned uk schmidt telescope ( ukst ) photographic plates , with magnitudes ranging from @xmath8 . in addition , the candidates had to satisfy one of the following criteria ( see * ? ? ? * ) : @xmath9 ; @xmath10 ; or @xmath11 . this results in a markedly different color and ( @xmath12-band ) magnitude distribution from the sdss quasars . however , the important similarity is that the redshift distribution between @xmath13 is fairly uniform as function of @xmath12-band magnitude ( cf . figs . [ brightdist ] and [ brighthist ] ) . so , even though the quasars have been selected differently , and actually populate the color - color diagram of fig . [ samplediff ] differently , we feel that there is no a - priori bias in either sample with respect to variability in general , and variability on select time - scales in particular . calibration of historic photographic plate material can be achieved by virtue of using large numbers of random field stars around the ( quasar ) position of interest . plate - to - plate variations in emulsion quality , and even variations within a single plate can contribute significantly to measurement uncertainties . so , even though the poss i and gsc2 catalogs have been calibrated carefully ( as a whole ) , and brought up to ccd photometric standards , there is still a lot of improvement to be made by recalibrating the photometry . we basically follow the same procedure as outlined in paper i by using all the available photometry for the field stars within 5 of the quasar position . this typically amounts to ( depending on the epoch ) anywhere between 50 to 500 stars . we like to stress that this `` local '' calibration is to be preferred over complete plate corrections due to the potential inhomogeneities inherent to photographic plates ( e.g. , * ? ? ? * ; * ? ? ? we will go over the calibration process step by step , but we will refer to the calibration sections in paper i where appropriate . most of the next discussion will highlight the improvements we were able to make on the old procedure , mainly due to the much larger data - set . the first step is to calculate the best passband transformations for each quasar individually , using the nearby field stars . the transformations involved are , for the poss i : @xmath14 to @xmath15 , and @xmath16 to @xmath1 . note that the @xmath14 and @xmath16 magnitudes are already transformed to the johnson passbands from their photographic @xmath17 and @xmath18 emulsions ( see reid et al . 1991 , and monet et al . 2003 for the @xmath14 and @xmath16 transformations ) . for the gsc2 plates , the relevant color transformations are @xmath19 to @xmath15 , and @xmath20 to @xmath1 . in principle , our transformation will take care of the proper passband corrections , possible plate / weather variations , and the fact that sdss uses ab magnitudes whereas the catalogs are on the vega system . however , an important caveat we like to emphasize here is that the `` best transformation '' is defined as the particular transformation that results in the smallest color rms _ for the stars_. as we have explained in paper i , this does not necessarily translate into the best calibration for the quasars , which after all , is the transformation we are interested in . there are two main contributors to this stellar - quasar disparity : their optical spectrum is completely different ( cf . fig . 3 of paper i ) , and quasars typically have powerful emission lines which depending on their redshift , may , or may not , be present in the passband . these emission lines can account for upward of a few tenths of a magnitude of the total brightness . both these differences between the field stars and the quasars render the stellar transformation less than ideal . it is something we can correct for , however . figure [ comptemplate ] illustrates this calibration best . the bottom panel shows the median @xmath1@xmath21 color for our quasar sample as function of redshift . ideally , these residuals should be close to zero after calibration in the absence of emission lines and quasar - stellar spectral differences . this is clearly not the case . however , these color excursions ( of up to more than 0.2 in magnitude ) are closely matched by what one would expect using quasar and stellar template spectra ( top panel ) . as explained in paper i , we get the best agreement between the actual residuals and the theoretical ones by assuming a mean stellar template of a k2v star . this stellar type is consistent with expectations based on population models for our galaxy ( e.g. , * ? ? ? * ; * ? ? ? so , the fact that these color excursions are well understood in terms of quasar emission lines moving in and out of the observing passband , makes it clear that we have to correct for it . if left `` untreated '' it will affect the variability sf directly by artificially inflating the rms values at certain time - lags . given the epoch distribution of the observations ( time - lags preferentially at @xmath22 , @xmath23 , and @xmath24 years ) , the redshift maps more or less directly onto a particular time - lag . the strong excursion at @xmath25 , for example , would skew the sf signal preferentially at @xmath26 , and @xmath27 years . in this paper we opted to use the actual median , as calculated across bins with a width of 0.05 in redshift units , over the modeled offsets . this accounts much better for the low ( @xmath28 ) redshift quasars which are increasingly more contaminated ( with decreasing redshifts ) by their host galaxies . all of the other passband transformations are treated similarly . the result is that each historic passband has been brought onto their sdss counterpart ( either @xmath29 or @xmath12 ) , with the important distinction that the color distributions are centered around @xmath30 as a function of redshift . this method improves significantly over the procedure outlined in paper i. there , bulk corrections have been applied to the color distributions ( irrespective of redshift ) . figure 2 of paper i can therefore be considered a projection of fig . [ comptemplate ] onto the @xmath31-axis . only with the large increase in sample size were we able to actually correct for the redshift dependence in a meaningful way . after all the photometric data have been transformed onto the sdss passbands , the measurements for each individual quasar are permutated among each other , resulting in about 4 time - lag measurements per band per quasar . obviously , none of the individual quasars have been sampled photometrically anywhere near enough to produce a meaningful structure function for each quasar individually . the combined data - set , however , allows for detailed variability studies provided one assumes that the underlying cause of quasar variability is the same for all of them . we will get back to this issue in [ modeling ] . this final data - set ( one each for the @xmath15- and @xmath1-bands ) contains sorted pairs of time - lag ( in years ) and magnitude difference . for our sample of 41391 quasars , this amounts to 170102 individual measurements for the @xmath15-band , and 131123 for the @xmath1-band . this exceeds the total number of permutations in paper i by more than an order of magnitude . the difference in the totals between the bands is because some gsc2 photometry for the quasars have been repeated more in the @xmath19 than in the @xmath20 band , boosting the permutation numbers for the @xmath15-band . the actual data for the @xmath15-band have been plotted in fig . [ permcheck ] . the bottom panel shows the magnitude differences as function of intrinsic time - lag . since the quasars are quite spread out in redshift space ( cf . fig.[brightdist ] ) , we have to bring the actual time separation between the observations onto the reference frame of the quasar itself ( by dividing it by a @xmath32 factor ) . this has the additional advantage of smoothing out the time - lag distribution . so even though the observing campaigns were well separated in time ( 1950 s , 1990 s , and @xmath33 ) , resulting in time - lags clustering around a few , @xmath34 , and @xmath35 years , the @xmath32 redistributing factor results in a pretty smooth distribution up to time - lags of @xmath36 years ( cf . [ permcheck ] , bottom panel ) . the top four panels of fig . [ permcheck ] provide a direct picture of the increase in fwhm ( and hence the rms ) of the magnitude difference distribution as time - lags increase . this is actually the definition of the sf ( see next section ) . since the total number of time - lag measurements decreases with increasing time - lag , this is not immediately obvious looking at the point - cloud in the lower panel . the numbers of data - points for the current 10-year time - lag binning are : 86795 , 50974 , 22627 , and 8520 permutations respectively . while the numbers do decline , they are still large enough to assess the fwhm of the distribution very accurately . the last bin alone already contains 30% of the total number of permutations used for paper i. our analysis of paper i , and the current paper , will utilize the sf as the tool to characterize the quasar variability . sf s are not very sensitive to aliasing problems due to discrete and/or sparse time sampling ( e.g. , * ? ? ? * ) , which make them well suited for our purpose . as before , we define the sf as : @xmath37 ^ 2 } \right)^{\frac{1}{2}}\ ] ] with the summation over all the combinations of measurements for which @xmath38 . in our case we group all the @xmath39 permutations into bins which contain at least 200 measurements . the sf value for each bin is then given by the rms of the magnitude permutations . this results in @xmath01500 bins , which are then binned again onto a fixed grid in log time - lag space ( running from @xmath40 to 1.55 in 0.06 dex bins for a total of 42 ) . this facilitates easy comparison between model and actual sf curves . it also allows us to approximate the error on a particular sf point by calculating the rms of the @xmath41 values inside each bin . this is basically the same method as we employed in paper i. the presented error - bars reflect therefore accurately the actual local sf uncertainties . it should be stressed that , unlike a well monitored sf of a _ single _ source for which all of the bins are cross - correlated with each other and an objective error estimate is hard to give , our bins are essentially _ independent_. out of the 150000 or so time - lag measurements ( per band ) only measurements for a single quasar ( about 4 ) are correlated with each other . in other words , each of the sf bins contains a virtually completely different set of quasars . this bin - independence also allows us to quantify sf similarities in terms of their offset distributions . assuming two sf curves , labeled @xmath42 and @xmath14 , both of which are binned to the same @xmath43 bins specified above , we can define : @xmath44 @xmath45 after substituting @xmath46 . the quantities @xmath47 and @xmath48 represent the mean sf offset and its @xmath49 uncertainty , respectively . we will use this metric in particular for our sf asymmetry part of the paper . the sf for the calibration stars serves multiple purposes . if we assume that stars , _ on average _ , are not variable , then the sf derived from it should not exhibit any correlation with time - lag . in other words , it should be parallel to the @xmath50-axis ( in plots like fig . [ sfstars ] ) . this was indeed found to be the case for the stars in paper i , which clearly illustrated the significant differences between the sf behavior of stars and quasars . however , given the much smaller sample sizes for paper i ( note that the number of calibration stars is linked to the number of quasars ) , the overall stellar sf was rather noisy . it just served to make the point that constructing an sf from a random sample of stars resulted in a non - variable sf curve , but it clearly was not good enough to go beyond that . the current sample , however , is large enough . in the next few sections , we will discuss the stellar sf in more detail . in the same way spectral differences between the average stellar spectrum and a quasar spectrum lead to slightly different passband corrections , and therefore , additional noise to the variability measure , spectral differences among stars themselves will inflate its sf variability signal as well . this has to be considered in the construction of the stellar sf . the reason we can use stars to calibrate the _ quasars _ at all , is that the _ mean _ of the stellar color distribution does not change that much going from one sightline to another . the stellar population therefore does not change a lot across the sky covered by dr2 . in order to limit the stellar spectral range allowed for our template sf , we only included stars within a magnitude range ( @xmath51 ) , and an @xmath52 color within 0.2 magnitudes of the typical stellar color of @xmath53 ( cf . this color cut effectively limits the allowed range of stellar colors , and improves the passband calibrations accordingly . the resulting time - lag permutation database contains 2.1 million data - points ( over both bands ) , an order of magnitude larger than the quasar permutation database . the net effect is a lowering of the sf , especially for the gsc data ( below time - lags of 10 years ) . the poss i data , plotted separately in fig . [ sfstars ] , retains a slightly higher noise plateau , mainly due to the photometric data quality differences between the gsc and poss surveys . because of this data quality difference , we have separated out the gsc and poss contributions to the stellar sf in fig . [ sfstars ] . linear least squares fits to the data - points have been made , and their slopes are : @xmath54 ( gsc-@xmath29 ) , @xmath55 ( poss-@xmath29 ) , @xmath56 ( gsc-@xmath12 ) , and @xmath57 ( poss-@xmath12 ) . with the possible exception of the @xmath29-band sf for the gsc data , none of the slopes differ significantly from zero ( i.e. , no correlation with time - lag ) . even the gsc-@xmath29 case is only weakly increasing with time - lag with the highest sf value within @xmath49 of the lowest . it should also be noted in this respect that the small sf excursions from the mean in the gsc-@xmath12 case are not statistically significant . the main result from this exercise is twofold . first , there is no significant correlation between the sf for stars and the time - lag for both the gsc and poss data . this implies that any signal we detect in the _ quasars _ beyond the stellar sf curve must be due to the quasars themselves . also , it is not clear how much of the stellar sf signal is due to intrinsic color scatter , and how much can be attributed to measurement noise . recently , @xcite estimated the photometric error in the gsc2 and poss i to be 0.10 and 0.15 magnitude , respectively . they used individual sdss plates ( 100 square arcminutes ) to correct the poss i and gsc2 photometric catalogs . taken at face value , these uncertainties translate into a white noise signal in the sf at levels of @xmath58 and @xmath59 , using eqn . 3 from paper i. both these levels are lower than the plateaus we measure in our stellar sf . how much of this difference can be attributed to the different calibration method , and how much can still be improved upon by even more carefully designed color - cuts , is unknown . as a consequence , we can not `` correct '' the quasar sf by subtracting a measurement - noise component , and instead , like in paper i , we will have to include a white noise term in our monte carlo models . figure [ sfstars ] presents the final corrected sf for both the @xmath15- and @xmath1-bands . as measured from the figure , the final sf levels , separated into time - lags less and more than 10 years , are : @xmath60 and @xmath61 for @xmath15 , and @xmath62 and @xmath63 for @xmath1 . the corresponding formal noise levels are : 0.17 , 0.21 and 0.16 , 0.22 magnitudes . the poss i levels are only slightly lower than found in paper i , but the gsc2 levels are suppressed by about 0.10 ( in sf units ) . leading up to the construction of the sf is an intermediate step in which we take care to center the distribution of variations around @xmath30 , as function of time - lag ( see fig . [ permcheck ] for the quasar case ) , without changing the _ shape _ of the distribution . if there is any asymmetry present in the distribution , we can test for it by looking at the positive and negative sides of the distribution separately . the variations are defined with respect to the newest epoch , so positive variations imply that the source is brighter ( i.e. , had a lower magnitude ) than it was before , and negative variations imply that the source is fading with time . this definition is consistent with the one used in @xcite . for a perfectly symmetric variation distribution , the positive and negative sf curves should be identical . this , however , is clearly not the case for our calibration stars . figure [ sfposneg ] shows the sf for the positive ( light - gray triangles ) and negative variations ( dark - gray triangles ) separately . since the sf curves are not identical , it indicates an asymmetric variation distribution . in the next few paragraphs we will argue that this asymmetry is due to the malmquist bias in our stellar sample . it basically means that for the subset of our stars that have intrinsic variability , the ones that were a lot fainter at the earlier epochs ( i.e. , the positive variations ) would not make it into either the poss i or gsc2 catalogs ( given their brightness limits ) , thereby reducing the rms / sf signal . variations in the other direction are not affected by this since there is no upper brightness limit to the catalogs . the net effect is a skewing in the variation distribution . there are some important points to be made based on the sf curves in fig . [ sfposneg ] . first , both sf curves are the same , except for a constant offset in log ( which translates into a @xmath64 increase in rms ) . this implies two things : 1 ) the malmquist bias does not depend on time - lag ( and it should not ) , nor does it depend on the quality of the photometry . the poss i measurements are noisier than the gsc2 ones , but there is no evidence for a different offset between the positive and negative sf curves . 2 ) the magnitude of the stellar variability is not correlated to intrinsic time - lag because we are not sensitive to their time - scales , quite unlike the variations for the quasars ( see [ qsf ] ) . the second observation we like to make is that , on top of the uncertainty about the exact contribution of measurement noise to the sf ( cf . [ stellartypedep ] ) , this asymmetry induced sf signal further compounds the problem of disentangling the sf into its various contributions . therefore , we can not use the stellar sf to improve or correct the quasar sf . @xcite quote a variable star fraction ( with variability exceeding 0.2 magnitude ) of at least 1% of the population . the rms difference between the positive and negative sf curves ( from fig . [ sfposneg ] ) is actually not that large : for the shorter timescales , we measure rms = 0.148 and rms = 0.177 mag for the positive and negative sf curves . since this 0.029 magnitude difference is not that large , a small fraction of variable stars might be enough to skew the distribution . for sufficiently large distributions ( where @xmath65 ) , the rms of two distributions , each with its own @xmath66 , can be combined as follows : @xmath67 with @xmath68 , @xmath69 the total number of items in each distribution . in our case , we assume the following for the _ non - variable _ part of the stellar distribution : @xmath70 , and @xmath71 ( i.e. , only 1% of the sample is variable ) . we also have to assume that this @xmath72 scatter is symmetric around @xmath30 , and is small enough not to be affected by the malmquist bias . it is in effect a constant contribution to both of the sf curves . all of the malmquist signal therefore has to be ascribed to the variable subset of stars . we can make the observations agree with our simple distribution model by assuming that half of the variable stars vary with on average 0.5 magnitude , and the other half varies by 1.5 magnitude . since we are looking out of the galactic plane , and do not make any a - priori assumptions about the stars ( other than that they are found close to a quasar ) , the main constituents of the variable star population are rr - lyrae and asymptotic giant branch ( agb ) stars . neither the magnitude of their variation , nor their relative fractions are inconsistent with the values we assume here ( see , e.g. , * ? ? ? * ) . it is the high variability part that drops out on the positive variation side of the distribution ( i.e. , the half with the 1.5 magnitude variability ) . this results in the following rms values : @xmath73 and @xmath74 magnitude . by applying eqn . [ rmses ] , we arrive at total rms values of 0.148 and 0.179 , very close to the actual values . obviously , none of these values are constrained to any degree , they just serve to demonstrate that we can actually explain the observed stellar sf curves by the malmquist bias due to a small percentage of variable stars . we will come back to the asymmetry issue in the next section , where we discuss the sf of our quasar sample . our quasar sf curve is presented in fig . [ sfquasars ] by the dark - gray squares . it is immediately clear that this sf curve is a vast improvement over the one presented in paper i ( fig . 7 ) , with a much smaller scatter of the points . the individual error - bars are actually small enough to allow for detailed modeling , something that the data quality did not allow for in paper i. before we continue , however , we like to establish the reality of the quasar sf curve by comparing it to currently the best long - term one available in the literature . this paper s approach to long - term quasar variability is quite different from the monitoring approach , in which one repeatedly measures the brightnesses of a fixed sample of quasars . given our statistical method , which uses archival photometry data and a less than straightforward calibration path , one might have concerns about the results . we therefore compare our resulting quasar sf to the highest quality available long - term quasar sf from the literature ( taken from hawkins 2002 ) . this sf was constructed based on @xmath75 year monitoring data on a sample of 401 quasars , and is plotted with our sf in fig . [ sfhawkins ] . the hawkins sf , which has not been corrected for time - dilation , is represented by the dark - gray squares . since our data have been corrected by the @xmath32 term , we have to shift the hawkins sf over toward shorter time - scales . the light - gray squares have been shifted leftward by 0.3 dex ( corresponding to a sample mean redshift of 1 ) . this shifting does not affect the slope of the sf , as mentioned in , e.g. , @xcite . note that , with the exception of the shortest two time - lag data points , all of the hawkins data can be made to fall within @xmath49 of our curve . the offset at short time - scales is most likely due to the higher photometric accuracy ( and hence a correspondingly lower noise plateau ) of the hawkins data compared to our data ( for which these data points will be shifted upward a bit , see fig . 6 from paper i ) . it is reassuring to see that these two different approaches , each with their own set of clearly distinct potential systematic problems , produce sf curves that are so alike . we do like to remind our reader that , even though the curves are similar , there is a key difference : our data bins are almost completely independent with different quasars contributing to different bins , whereas in the hawkins sf , most bins contain data ( permutations ) from the same set of quasars . this is a big advantage when one tries to understand the sf errors ( cf . [ errorest ] ) and the potential differences between positive- and negative - variation sf curves ( cf . [ lcasymmetry ] ) . there are a few things we like to discuss based on our quasar sf in fig . [ sfquasars ] . first of all , there is no indication that the sf curve is turning over , consistent with the results from hawkins ( 2002 ) . what we are looking for in the sf curve is the presence of a consistent plateau beyond a certain time - scale ( cf . [ sfqmodel ] , and @xcite ) , and not necessarily an actual `` peak '' in the sf curve . a significant drop in the sf signal beyond a particular time - scale usually indicates problems with adequate time sampling at those time - lags . in our case , the sf curve might be affected beyond @xmath36 years due to the decrease in available time - lags ( remember that the intrinsic time - scale are shortened by the @xmath32 time - dilation factor ) . we do not see such a drop , however , and the possible leveling off seen in the last few bins is not significant enough to claim we have detected a preferred variability time - scale . this is not to say that there is no upper bound to the variability time - scale , just that we do not have any sensitivity to it . the second clear trend in our sf is that the @xmath15 sf curve has more signal ( i.e. , is more variable at any given time - lag ) than the @xmath1 sf curve . this is consistent with quasars being intrinsically more variable at shorter wavelengths . this is not a new result ( it was also present , albeit at rather low significance , in paper i ) , but is now very clearly detected . @xcite measured a @xmath76 magnitude rms difference between b- and r - band variability in their sample of 42 palomar green ( pg ) quasars . @xcite found variability shifts of these magnitudes between the blue and red to be consistent over a range of samples . wavelength variations in the near - ir tend to be smaller , and are possibly too small to be measured @xcite . possible mechanisms for these spectral variations include nuclear star - bursts / supernovae which are predominantly blue ( e.g. , * ? ? ? * ; * ? ? ? * ) , and instabilities in the nuclear accretion disk ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? figure [ timelagvsz ] shows the sf color offset as a function of time - lag . in the top panel , we have plotted the observed rms offset between the @xmath15- and @xmath1-band sf curves , as function of time - lag in the quasar restframe . it appears that the increase in color offset correlates with time - lag , but it is actually the redshift that the color offset depends on . in the bottom panel we have plotted the mean redshift of the quasars contributing to a particular time - lag . since the time - lags have been converted into the restframe of the associated quasar , a @xmath32 factor has shortened the epoch separation . this implies that the very longest time - lags ( @xmath77 years ) can only include data from the lowest redshift quasars , a trend clearly seen in the plot . now , assuming the quasar variability introduces a spectral slope change ( i.e. , it is bluer during an outburst , cf . * ; * ? ? ? * ) the biggest color contrast is attained when the observed passbands are furthest apart . this would be the case for redshift zero objects . at higher redshifts , one starts to probe progressively bluer parts of the spectrum , and the spectral separation between the @xmath15- and @xmath1-bands becomes smaller and smaller . this effect of decreasing color contrast with increasing sample mean redshift is accurately portrayed in fig . [ timelagvsz ] . even the sudden increase in mean redshift around the 10-year time - lag bins is reflected by the drop in the relative rms change . the spearman rank coefficient for the correlation between the mean redshift and the color offset ( beyond time - lags of 1 year ) is @xmath78 . this translates for the 23 degrees of freedom into a less than 0.1% likelihood that the correlation is by chance . this good correlation between the two suggests an _ intrinsic _ origin to the variability . however , this does not rule out the per definition extrinsic micro - lensing scenario . it is possible that by assuming a @xmath32 time - dilation correction in the first place , we have introduced some correlation between mean redshift and color offset . given the fact that we rely on the @xmath32 term to smooth out our time - lag coverage , we can not produce a similar plot _ without _ such a @xmath32 correction to test this . nevertheless , a stronger argument against a lensing scenario is the presence of a light - curve asymmetry signal in the sf curve ( cf . [ lcasymmetry ] ) . the predicted slopes for the starburst ( sb ) and accretion disk instability ( di ) models are @xmath79 and @xmath80 , respectively @xcite . note that both these slopes are significantly steeper than our measured value of @xmath81 , which is far closer to the modeled micro - lensing slopes ( @xmath82 ) , and a bit shallower than the @xmath83 slope of the hawkins ( 2002 ) data . it was this inconsistency between the measured quasar sf slope and the predicted sb and di slopes that led hawkins to propose a micro - lensing origin of long - term quasar variability . our data , based on the slope of the sf alone , seems to support this contention . it also implies that , in the case micro - lensing is ruled out , that either or both of the other models need significant modification to explain the observed slope . as we pointed out in paper i ( fig . 6 ) , the measurement noise does have a direct effect on the slope of the sf . the larger the noise , the shallower the slope . since our data is noisier than the photometric observations used by @xcite for their modeling , we will come back to the slope issue in [ intslope ] where we construct a noise - less sf . variability in quasars is associated with their agn . a large , non - varying , host galaxy might therefore limit the relative variation if one uses integrated magnitudes . since we are not attempting any galaxy / agn decomposition , and just use the total magnitude , this might be somewhat of a concern . the sharp upturn in residual @xmath84 colors toward low redshifts ( cf . fig . [ comptemplate ] , or fig . 1 of paper i ) already hinted that this might be at play at low redshifts . in this section we will quantify this effect on the sf curve . figure [ timelagvsz ] , bottom panel , shows that the low - redshift sources dominate the longer time - lag bins , and that the largest effects are to be expected here . this is confirmed in fig . [ sfhostgalcont ] , which plots the long time - lag part of the sf , with different low redshift thresholds . the apparent turnover at the extreme end of the bottom sf curve ( which includes all data ) , disappears if one removes all the quasars below redshifts of 0.6 . evidently , the host galaxy contribution is enough to lower the variability signal at these redshifts . clearly , if one wants to study variability of agn using a nearby sample , this has to be taken into account . in our case , a simple removal of the nearby quasars is enough , since _ none _ of the other sf bins are affected ( again thanks to our bin independence ) . it should also be noted that this turnover is not present ( or at least has not been detected ) in the @xmath15-band . this is consistent with the notion that the agn is more dominant at shorter wavelengths , whereas a galaxy is usually much redder . in further discussions about the @xmath1-band sf curve , we have removed these quasars ( 10% of the total ) from the sample . the @xmath15-band data - set is unaffected . asymmetries between the rising and falling parts of the light - curve can be investigated by separating the variations in positive and negative variations only ( cf . [ malmquistsf ] ) . @xcite model various scenarios of variability , each with different sf signatures . their starburst ( sb ) model has a very short rise time , followed by a long exponential decay ( cf . their fig . 2 ) . they also consider an accretion disk instability ( di ) model for which the variations rise slowly , but fall off rapidly ( basically in a saw - tooth like pattern , cf . their fig . 5 ) . the sf curves derived from these light curves both have significant asymmetries between the positive and negative variations . they are , predictably , of opposite nature : the fast - rise , slow - decline of the sb model results in more sf signal in the positive variations , whereas the slow - rise , fast - decline di model has more signal in the negative variations . this behavior can be intuitively understood in terms of the sf having typically more variability signal at the fast changing part of the light curve compared to the slowly changing part , for a given time - lag . in a sense the sf mirrors the derivative of the light curve . figure [ sfqposneg ] shows this asymmetry for our data . we only show data for the @xmath15-band because the sf offset between the @xmath15- and @xmath1-bands is of the same order of magnitude as the asymmetry signal ( cf . [ sfquasars ] ) , so combining both bands is not helpful . the positive variations ( light - gray symbols ) have more sf signal than the corresponding negative ones ( they are binned the same way in time - lag ) . this is a strong indication that the typical quasar variations are not symmetric , and behave in a fast - rise , slow - decline way . this is consistent with the modeled light curves we used in paper i ( see fig . 5 ) and the sb - like light curves , but appears at odds with the inferred behavior of the di models . furthermore , the asymmetry effect we see in the calibration stars ( cf . [ malmquistsf ] ) , and which we interpret as a sign of malmquist bias , works in the _ opposite _ sense . this increases the significance of the disparity between the positive and negative variations for our quasar sample . indeed , based on modeled sf curves for which we force the light - curves to be time - symmetric , we consistently measure a mean offset between the positive and negative variations of @xmath85 ( based on eqns . [ oline1 ] and [ oline2 ] ) . for our actual quasar sample these values are @xmath86 , which makes it significant at the 6.4@xmath66 level . another trend that does not appear in our data is for the two sf curves to merge beyond the typical time - scale of the variation . both the sb and di models of @xcite display this behavior , whereas our sf curves remain offset and parallel as function of time - lag . this is an indication that there does not appear to be a preferred time - scale in our quasar sample . in [ modelasym ] , we will actually argue for a continuum of variation time - scales . if variations are asymmetric in time , and either spend the least time rising or declining in brightness , offsets will appear in the sf between the positive and negative variations ( e.g. , kawaguchi et al . 1998 , and [ modelasym ] ) , given enough of an asymmetry . this asymmetry signal persists under arbitrary time - dilation corrections : a @xmath32 correction merely compresses the time - scales , it does not alter the intrinsic shape of the light - curve . neither does it change the slope of the sf ( e.g. , * ? ? ? * ; * ? ? ? the same statements are true for symmetric variations : there is no particular @xmath32 correction that will induce an sf asymmetry signal if the variations are intrinsically symmetric . this implies that the asymmetry we detect in our sf is not an artifact of the applied @xmath32 correction . in order to construct the sf in fig . [ sfqposneg ] , we used the quasar redshift itself to correct for time - dilation , with the side benefit of improving our time - sampling . however , in the case of micro - lensing , we would have to use the redshift of the lens , which is unknown . this effectively prevents us from ever creating an sf that is properly time - dilation corrected in the frames of the lenses . but _ if _ we are solely interested in whether there is an asymmetry signal in the variations or not , _ we do not have to_. our asymmetry signal persists , regardless of the location of the lenses and its associated proper @xmath32 correction . this asymmetry signal is at odds with the micro - lensing scenario , since these events have to be symmetric in time ( see for instance @xcite who specifically modeled lensing of the agn accretion disk ) . this strongly suggests that micro - lensing can not be the dominant cause of long - term variability in quasars . it quite likely contributes at some level , but not enough to define the sample average behavior . this contention is partly supported by @xcite , who found that micro - lensing can not account completely for the observed long term variability , based on its inability to explain the high number of large amplitude events and the mean variability amplitude at low redshifts ( where lensing is less likely ) . our results do constrain the level of micro - lensing a bit more though . in the rest of the paper we assume that all the variations are intrinsic to the quasar , and hence the @xmath32 time - dilation correction has to be applied . if one wants to consider both the time - dilation corrected and uncorrected cases , one introduces another level of degeneracy between redshift , absolute luminosity , and restframe spectral variability @xcite , which needlessly complicates matters . based on the results in the last section , we feel confident that the variations are indeed source related . this brings us to the first of the possible degeneracies : redshift and restframe spectral variability . it is clear that sources vary more in the blue than in the red ( cf . [ sfquasars ] , or * ? ? ? * ; * ? ? ? * ; * ? ? ? any trend in which higher redshift quasars become intrinsically less variable can be offset against the increase of variability as the observed spectral range shifts toward the blue with increasing redshift . disentangling the redshift and spectral variability contributions will be difficult , provided the higher redshift sf curve lies above the lower redshift one . however , as is shown in fig . [ sfhighlowz ] , it is clear that this is not the case : the high-@xmath87 ( light - gray symbols ) sf curve lies _ below _ the low-@xmath87 one ( dark - gray symbols ) . this provides us with a solid lower limit on the trend that high-@xmath87 quasars are less variable than their low-@xmath87 counterparts . we measure a mean offset along the @xmath31-axis of @xmath88 , for a 2.2@xmath66 significance . again , given the intrinsically bluer part of the spectrum that is probed for the high redshift bin ( @xmath89 , 18426 quasars ) compared to the low redshift bin ( @xmath90 , 18335 quasars ) , one expects the former bin to be more variable . that this is not the case only increases the significance of the assessment that low redshift quasars are more variable than high redshift ones . the observed magnitude range for the quasars between @xmath91 is roughly @xmath92 ( cf . [ brighthist ] ) , and rather uncorrelated with redshift . this in turn implies that the absolute @xmath1-band luminosity ( after taking a k - correction into account ) of the sample increases as a function of redshift . the two redshift bins of fig . [ sfhighlowz ] are therefore also separated in absolute luminosity . since absolute luminosity is an intrinsic property of the quasar ( whereas redshift is not ) , it makes more sense to investigate the variability as function of absolute luminosity . if one assumes that the quasar variations are similar in an _ absolute _ sense , then the _ relative _ variations are smaller for the intrinsically brighter objects . figure [ sfhighlowl ] shows this trend . the subsample of the quasars with redshifts @xmath91 has been divided up into two bins : all the quasars fainter , and brighter than @xmath93 . this value is the mean absolute luminosity of the subsample=71 km s@xmath94 mpc@xmath94 , @xmath95 , and @xmath96 throughout this paper . ] note that for the low - luminosity quasars ( the dark - gray symbols ) the photometry quality degrades toward higher redshifts , which expresses itself as an artificial increase in the sf signal ( especially for the time - lags between 10 to 20 years ) . the gsc2 photometry is less affected by this , resulting in a nicer signal . the sf curves are offset significantly ( and more so than in fig . [ sfhighlowz ] ) between time - lags of about a year to @xmath97 years . based on figs . [ sfhighlowz ] and [ sfhighlowl ] , we conclude that high - luminosity quasars vary less than low luminosity quasars . this supports earlier similar findings by , e.g. , @xcite , but contradicts for instance , @xcite . these authors argued that the absence of the trend for higher luminosity quasars to be less variable points toward a scenario in which the object varies as a single ( coherent ) source , rather than a flaring sub - unit . given our results , we have to conclude the opposite , namely that quasars tend to vary incoherently , and do so with a limited magnitude range of the flaring sub - units . @xcite tried to explain this within a supernova context ( e.g. , * ? ? ? * ) in which the time separation between outburst ( with total energies of up to @xmath98 ergs ) are distributed in a poissonian way . however , the resulting sf curves based on supernovae are too steep ( cf . * ? ? ? * ; * ? ? ? * ) , and are effectively ruled out . in paper i , we applied a similar shot - noise outburst distribution to model the sf curve , but used a broader ( more generalized ) exponential light - curve . in [ modeling ] , we will get back to this issue . in this section we will expand on the modeling done in paper i. the limited sample size and the resulting noisy sf curve did not allow for accurate fitting in that paper . with the present data , however , we are in a position to investigate this further . the main modeling result from paper i was that we could approximate the observed sf curve by assuming the following : 1 ) the typical quasar undergoes periodic outbursts with a decaying time - scale of @xmath99 years , and 2 ) these outbursts occur on a typical time - scale of @xmath100 years . as we will see in the next few sections , this simple picture is not correct . first , let us begin describing the modeling setup we used . like in paper i , we assume a typical outburst can be described by a canonical exponential function ( eqn . [ lceqnasym ] ) . curves of these types can be used to model various possible outburst scenarios , ranging from short time - scale supernovae to longer time - scale accretion disk instabilities ( see , e.g. , * ? ? ? * and references therein ) . in addition to paper i , we also introduce a symmetric version of eqn . [ lceqnasym ] , given in eqn . [ lceqnsym ] . we need both curves to adequately investigate the observed sf asymmetry ( cf . [ modelresults ] ) . for each equation , @xmath42 represents the amplitude of the outburst ( in magnitudes ) , @xmath101 is the normalization constant so that the peak has an amplitude @xmath42 , and @xmath102 is the exponential half - life time ( in years ) . the time parameter @xmath103 has been chosen such that the peak of the outburst occurs at time @xmath104 . the normalized light - curves have been plotted in fig . [ modelcurves ] to make this a little clearer . the asymmetric curve ( plotted in dark gray ) is the exact same as in paper i , and the symmetric case is identical to the asymmetric one for times @xmath105 . @xmath106 @xmath107 note that eqn . [ lceqnasym ] is only valid for times @xmath108 . the quantities @xmath42 and @xmath102 are free parameters in the model . for each of these points in phase - space , a quasar composite light - curve is constructed covering a large time period ( typically 10000 years ) . this period is filled up with outbursts based on @xmath42 and @xmath102 , separated in time by a characteristic outburst time - scale @xmath109 . like in paper i , the probability of an outburst _ not occurring _ within time @xmath103 is given by @xmath110 which is also known as a shot noise model of variability ( see , e.g. , * ? ? ? once we have constructed such a canonical quasar light - curve ( based on the values of @xmath42 , @xmath102 , and @xmath109 ) , we generate a database of measurements which is identical to the actual one for our quasars . so we use the very same redshifts ( all 41391 ) , and with on average 4 measurements per quasar ( from this canonical light - curve ) , this results in about 250000 permutations . this is close to the actual values ( 170000 for the @xmath15-band , and 130000 for the @xmath1-band ) . to each , randomly sampled , `` measurement '' we add white noise that matches the actual measurement uncertainties . we adopt the conservative values of @xmath111 , @xmath112 , and @xmath113 magnitudes . the error in the magnitude differences are then the rms - values of the appropriate @xmath66 s . the first two values are in part set by the sf noise plateau at very short time - scales ( which do not contain any poss data ) , and the @xmath114 value is based on the stellar sf plateau in [ stellartypedep ] , which can be considered an upper limit ( @xcite quote a @xmath115 mag ) . the artificial sf curve which is based on these data can then directly be compared to the actual one . the model sf curve is sampled on the exact same time - lag binning as the actual one , allowing for a simple comparison . there is , however , a complication . the parameter combination of @xmath42 and @xmath109 turns out to be degenerate . if one increases the amplitude @xmath42 , and at the same time makes outbursts rarer by increasing the typical time - scale @xmath109 , it results in the same sf curve ( which is still dependent on @xmath102 ) as for smaller values of @xmath42 and @xmath109 . the sole thing that discriminates between the two scenarios is that the error - bars in the high @xmath42 , high @xmath109 case are much larger than in the low @xmath42 , low @xmath109 case . this can be understood in terms of what defines the typical quasar behavior : in the high @xmath42 , high @xmath109 case , most quasars will not vary , except for a small subset . this introduces large sf variations , depending on whether such measurements are present in the bin or not . in the low @xmath42 , low @xmath109 case , a lot of quasars vary , but do so at low amplitude . the relative sample variations are small , and most time - lag bins are made up of more homogeneous measurements . remember , the error - bars in the sf curves are based on the rms within a set of time - lag bins ( we bin bins , see paper i for more details ) . so , instead of a simple @xmath116 value , our goodness - of - fit is defined by the sum of the @xmath116 value and the rms in the difference between the sizes of the model and actual error - bars . their relative values are normalized such that for a good fit each contribution has about equal weight . the best fitting values of @xmath42 , @xmath102 , and @xmath109 ( or sets thereof in cases where we fit multiple components ) are found by exploring the phase - space extensively . since this space typically does not have any steep gradients , we opted for a slow - cooling simulated annealing code , specifically developed for this purpose . it consistently converged to the same solutions , provided we cooled slow enough . in the first subsection , we will restrict the modeling to the symmetric light - curve functions ( eqn . [ lceqnsym ] ) . after that we are also considering the asymmetric light - curves of eqn . [ lceqnasym ] . in paper i we were able to fit the observed sf curve adequately by a single set of @xmath117 values , suggesting a possible characteristic variability time - scale / mechanism common to all quasars . the data quality , however , was such that this could not be put on sure footing . with the current high quality data , this is any easy thing to test . as can be seen in fig . [ sfqmodel ] , we can not fit the observed sf curve with a single component . the two model curves , one fitting the short ( @xmath118 years ) time - lags , and one fitting the long ( @xmath119 years ) time - lags illustrate the problem . in order to generate enough signal at the short time - lags , the model light - curve has to have a small half - life @xmath102 . the actual values for the fit are : @xmath120 , with @xmath42 in magnitudes , and @xmath102 , @xmath109 in years . however , it is also clear that the sf curve belonging to this light - curve does not have any signal increase beyond a few @xmath102 ; it effectively levels off after @xmath121 years and starts deviating from the actual sf curve significantly . on the other hand , sf model curves that do fit the longer time - lags ( based in this case on light - curves with a half - life @xmath122 years , @xmath123 , and @xmath124 ) , do not have any significant sf signal on time - scales much shorter than their @xmath102 value . we therefore need at least two variability components . this refutes the notion from paper i that there is a single preferred variability time - scale common to all quasars . it also implies that there may be a continuum of variability time - scales . our best fitting two component fit ( to the sf curve plotted in fig . [ sfqmodel ] ) has values of : @xmath125 , @xmath126 , @xmath127 , and @xmath123 , @xmath128 , @xmath129 . again , due to the degeneracy between @xmath42 and @xmath109 , their values are not very well constrained . longer periods @xmath109 are not excluded . however , we did try to make the error - bars on the model sf curve as much like the actual data as possible . even fitting with more components did not alter the shortest time - scales . most of the shortest @xmath102 values grouped around 0.2 year . our data have some sensitivity to these short time - lags , but not an awful lot . about 2% of the available time - lag measurements are shorter than 0.2 years . this half - life of up to a few months is comparable to high - redshift supernovae timescales ( see , e.g. , * ? ? ? the longer time - scales needed to fit the sf curve beyond a few years are more in line with disk instability models of large accretion disks around supermassive black holes ( e.g. , * ? ? ? it is therefore not immediately clear what , if any , mechanism dominates . as we have seen in [ lcasymmetry ] , our quasar sf curve displays significant asymmetry , with a mean offset of @xmath86 . this is not compatible with our symmetric fits from the previous section , which consistently resulted in offsets of @xmath130 , to within the @xmath131 uncertainty ( cf . [ sfmposneg ] , top curve ) . hence , we need to introduce some measure of light - curve asymmetry into our modeling . for this , we use eqn . [ lceqnasym ] as the functional form of the light - curve . the fitting method is the same as for the symmetric light - curve case , and the results are reasonably comparable . given the shape of the light - curves ( cf . [ modelcurves ] ) , one expects to find larger values of @xmath102 for the asymmetric case than for the symmetric one . this is because the latter light - curve has variations at longer time - scales than the asymmetric one for identical values of @xmath102 . we find for the asymmetric 2 component case : @xmath132 , @xmath133 , @xmath127 , and @xmath123 , @xmath134 , @xmath129 , which has indeed larger values of @xmath102 than the symmetric case . it also carries the asymmetry signal we are interested in . a typical resulting sf is shown in fig . [ sfmposneg ] , separated into negative- and positive - only variations . the sf curves are clearly separated ( mean offset @xmath135 for this case ) . however , there are a few things that are not entirely consistent with the observed data . aside from the slightly smaller value of the asymmetry signal @xmath47 , it appears that for time - lags somewhat shorter than @xmath102 , the sf does not have an asymmetry signal . this is not observed in the actual data ( cf . [ sfqposneg ] ) . since these nulls are associated with the few discrete time - scales @xmath102 , adding in more intermediate @xmath102 s will smooth out and remove the nulls . this suggests that the intrinsic quasar variations have a more continuous distribution of @xmath102 s . indeed , our models with more than four components tend not to have these nulls . the data , and our modeling effort , do not allow for a precise characterization of the magnitude of the light - curve asymmetry . it also does not allow for the isolation of particular variability time - scales , and is more compatible with a scenario in which quasars can vary with a continuous distribution of time - scales . one thing that is clear , though , is that the variations are asymmetric , and are of the fast - rise , slow - decline type . the theoretical slope calculations in @xcite do not include a white noise component , as their data quality for the individual sources was good enough . we , however , do have to include a white noise term in our modeling to make it agree with the actual data better . this does have the side - effect of lowering the sf slope ( as noted in paper i ) , which potentially might render the slope comparison suspect . to this end , we have generated `` noise - less '' sf curves derived from a well fitting `` noise - included '' sf curve . the results have been plotted in fig . [ sfnonoise ] . the top set of data - points are the actual sf with the best fitting ( 3 component in this case ) sf model . the slope for both curves is measured to be @xmath136 , which , as discussed in [ sfslope ] , is much shallower than any of the @xcite values . if we remove the measurement noise components , we end up with the two bottom sets of sf curves ( a symmetric and an asymmetric one ) . note that even though there is no measurement noise present , there will always be a range of magnitude differences for a fixed time - lag ( unless there is no variation ) , which explains the presence of error - bars . the slope for these curves is measured to be @xmath137 , which indeed is steeper . the fall - off below time - lags of a year is due to the fact that the actual sf ( top curve in the figure ) does not have any sensitivity to these short time - scales , as they are effectively masked by the measurement noise . since we just de - noised the fit to the actual sf , this lack of short time - scale signal becomes apparent . the quoted slope , therefore , is only for the sf curve beyond time - lags of one year . while the slope of the sf has indeed steepened a bit in the noise - less case , it is not enough to change the assessment of [ sfslope ] in a significant way . it is still too shallow for either the sb or di models , but is now marginally steeper than the micro - lensing slope . our results on the quasar sf strongly suggest , for the first time , that most of the long - term variations are intrinsic to the quasar itself . micro - lensing by objects along the line - of - sight to the quasar , or even in the quasar host galaxy itself , is not a viable explanation of the long - term variability in general . this result is mainly based on the observed asymmetry in the sf , indicative of a fast - rise , slow - decline type of variability ( cf . [ lcasymmetry ] ) . the significance of the observed asymmetry is enhanced by the malmquist signal , which works in the opposite sense . if our measured asymmetry would have had the same sign as the malmquist bias , one would be hard pressed to ascribe any of that asymmetry to the intrinsic quasar variability behavior , and the data would have been consistent with the symmetric variations needed in the micro - lensing scenario . this is , however , not the case . the formal statistical significance of the asymmetry is @xmath138 . we have put some other results on a more secure footing as well . first , no obvious turnover has been detected in the quasar sf , which indicates that there is no upper preferred variability time - scale ( smaller than a few decades ) . our results are consistent with a continuum range of variation time - scales . this is based on the absence of a turnover , as well as on the near constant offset between positive- and negative - only variations in fig . [ timelagvsz ] . if we model asymmetry based on a few preferred time - scales @xmath102 , we find almost no asymmetry signal at timescales of a few times @xmath102 ( cf . [ sfmposneg ] ) , something clearly absent in the real data . a more or less constant offset in the model can be achieved with a larger number of components ( 4 or more ) , all with time - scales @xmath102 less than 10 years . this opens up the door to any number of components , since as far as we know there is no observed turnover beyond @xmath040 years which would set an upper bound . second , the magnitude of the quasars variability is a clear function of wavelength : variability increases toward the blue part of the spectrum . this confirms previous observations by , e.g. , @xcite , and is consistent with both the starburst ( sb ) and the disk instability ( di ) model of variability . however , based on the model sf slopes by @xcite , the sf slopes for the sb model @xmath139 ) are even more inconsistent with our `` noise - less '' modeled value of @xmath140 than the di values of @xmath141 . clearly , some modifications need to be made before the models can adequately describe the observations . third , high - luminosity quasars vary less than low - luminosity quasars . this is consistent with a scenario in which variations have a limited absolute magnitude , and variations are due to sub - components instead of coherent variation of the agn ( e.g. , * ? ? ? these sub - components can be interpreted as either individual supernovae , or discrete flares due to disk instabilities . in summary , all the data presented here lead to the conclusion that quasar variability is intrinsic to the source , is caused by chromatic outbursts / flares with limited luminosity range and varying time - scales , and with an overall asymmetric light - curve . currently , the model that best explains this observed behavior is based on accretion disk instabilities . however , given the existing discrepancies between the sf slopes of the model and observations , some reservations are still in place . we like to thank the referee for comments that helped improve the paper . wdvs work was performed under the auspices of the u.s . department of energy , national nuclear security administration by the university of california , lawrence livermore national laboratory under contract no . . the authors also acknowledge support from the national radio astronomy observatory , the national science foundations ( grants ast 00 - 98259 and ast 00 - 98355 ) , the space telescope science institute , and microsoft . mclean , b. j. , hawkins , g. , spagna , a. , lattanzi , m. , lasker , b. m. , jenkner , h. , & white , r. l. 1998 , `` the second guide star catalogue '' , in `` new horizons from multi - wavelength sky surveys '' , proceedings of iau symposium 179 , eds . b. mclean , d. golombek , j. hayes , & h. payne , p. 431

in our second paper on long - term quasar variability , we employ a much larger database of quasars than in de vries , becker & white . this expanded sample , containing 35165 quasars from the sloan digital sky survey data release 2 , and 6413 additional quasars in the same area of the sky taken from the 2df qso redshift survey , allows us to significantly improve on our earlier conclusions . as before , all the historic quasar photometry has been calibrated onto the sdss scale by using large numbers of calibration stars around each quasar position . we find the following : ( 1 ) the outbursts have an asymmetric light - curve profile , with a fast - rise , slow - decline shape ; this argues against a scenario in which micro - lensing events along the line - of - sight to the quasars are dominating the long - term variations in quasars ; ( 2 ) there is no turnover in the structure function of the quasars up to time - scales of @xmath040 years , and the increase in variability with increasing time - lags is monotonic and constant ; and consequently , ( 3 ) there is not a single preferred characteristic outburst time - scale for the quasars , but most likely a continuum of outburst time - scales , ( 4 ) the magnitude of the quasar variability is a function of wavelength : variability increases toward the blue part of the spectrum , ( 5 ) high - luminosity quasars vary less than low - luminosity quasars , consistent with a scenario in which variations have limited absolute magnitude . based on this , we conclude that quasar variability is intrinsic to the active galactic nucleus , is caused by chromatic outbursts / flares with a limited luminosity range and varying time - scales , and which have an overall asymmetric light - curve shape . currently the model that has the most promise of fitting the observations is based on accretion disk instabilities .

introduction sample selection and calibration structure function modeling the structure function discussion and summary

model selection aims at choosing , from a set of candidates , a mathematical model that strikes a balance between simplicity and adequacy to the observed data . traditionally , it is performed by optimizing some information criteria ( ics ) . in particular , the bayesian information criterion ( bic , @xcite ) , aic @xcite , and the minimum message length ( mml ) principle @xcite , are widely used in different statistical model selection problems . these criteria have a discrete feasible domain . their optimization usually involves exhaustive search over all possible models , which is computationally intensive . when the model is very complex or the space of candidate models is very large , a brute force testing of all possible models causes very high computational costs and becomes impractical . to tackle this problem , a lot of efforts have been made to adjust the model complexity continuously . for instance , for the linear regression problem , lasso @xcite applies the @xmath0 penalty on the coefficients which could shrinking some coefficients to zero . various approaches , including adaptive lasso ( alasso , @xcite ) , scad @xcite , and first @xcite make use of similar but different ways of parameter shrinkage . for finite mixture models , entropic prior " @xcite or the dirichlet prior @xcite for the mixing weights could produce sparsity of the mixing weights and hence perform model selection . however , in these methods , how to select the penalization parameter is usually a crucial issue . moreover , asymptotic properties of these methods have been well studied , but less attention was paid to their performance on finite samples . it would be very useful if one can find their relationship to the ic - based approach for finite samples . we aim to develop an efficient model selection approach which is based on the continuous penalized likelihood and approximately coincides with model selection based on ics , such as bic . we call this approach quick information criterion - like ( quick - ic ) model selection . our contributions are mainly two fold . first , for regular models , we establish a bridge between the penalized likelihood of alasso and ics , and propose to approximate the latter with the former , resulting in convenient model selection ; this can also be considered as a way to directly determine the penalization parameter in alasso to perform ic - like model selection , which avoids the search for the penalization parameter and would save a lot of computation , especially when iterative procedures are needed to find alasso solutions . specifically , in sec . [ sec : alasso ] , we give the intuition that the penalty term for each parameter in alasso is closely related to an indicator function showing if this parameter is active . consequently , one can approximate the number of free parameters in ics in terms of such penalty terms and find continuous approximators to the ics . this inspires the proposed approach quick - ic in sec . [ sec : equivalence ] , which is shown to select exactly the same model as the corresponding ic does in the case with a diagonal fisher information matrix . general cases are also briefly discussed . the theoretical claims are verified by simulation studies in section 4 . second , in sec . [ sec : make_continuous ] , we extend quick - ic to non - regular and complex models , such as factor analysis , the gaussian mixture model and the mixture of factor analyzers @xcite , whose model selection is traditionally very difficult due to the large candidate model space . by making use of logarithm penalties with data - dependent weights , we provide continuous approximators to the ics suitable for model selection of these models , and make their model selection easy and efficient . this illustrates the good applicability of the proposed approach . in this section we assume that the model under consideration satisfies some regularity conditions including identification conditions for the parameters @xmath1 , the consistency of the maximum likelihood estimate ( mle ) @xmath2 when the sample size @xmath3 tends to infinity , and the asymptotic normality of @xmath2 . the penalized likelihood can be written as @xmath4 where @xmath5 is the log - likelihood , @xmath1 is the parameter vector , @xmath6 is the penalty , and @xmath7 is the penalization parameter . the maximum penalized likelihood estimate is @xmath8 . @xmath9 gives the @xmath0-norm penalty . the @xmath0 penalty produces sparse and continuous estimates @xcite , and it has been shown to outperform other penalties in some scenarios @xcite . however , it also causes bias in the estimate of significant parameters , and it could select the true model consistently only when the data satisfy certain conditions @xcite . certain methods , including stability selection with the randomized lasso @xcite and alasso @xcite , were proposed to overcome such disadvantages of the @xmath0 penalty . in particular , alasso uses @xmath10 , with @xmath11 , where @xmath12 , and @xmath2 is a ( initial ) mle of @xmath1 . consequently , the strength for penalizing different parameters depends on the magnitude of their estimate . under some regularity conditions and the condition @xmath13 and @xmath14 ( the subscript @xmath3 is used in @xmath15 to indicate the dependence of @xmath7 on the sample size @xmath3 ) , the alasso estimator is consistent in model selection . we are more interested in its behavior on finite samples . the result of alasso depends on the penalization parameter @xmath7 . for very simple models , one may use least angle regression ( lars , @xcite ) to compute the entire solution path , which gives all possible solutions as @xmath7 changes . among these solutions , the best model can then be selected by cross - validation or based on some ics @xcite . ( the latter approach is compared with our approach in sec . [ sec : compare ] , and one can see that it may give very different results from the corresponding ic . ) however , for complex models , especially when iterative algorithms are used to find the solution corresponding to a given @xmath7 , it is computationally very demanding and impractical to find the solution path . one then needs to select the penalization parameter in advance . in the next section we show that one can simply determine this parameter , while the model selection result approximately coincides with that based on ics . let us focus on the case @xmath16 , meaning that @xmath17 for such parameters . on the other hand , with suitable @xmath15 , the alasso estimator @xmath18 is also consistent @xcite ; roughly speaking , significant parameters are then expected to be changed little by the penalty , when @xmath3 is not very small . consequently , at convergence , @xmath19 _ approximately indicates whether the parameter @xmath20 is active or not_. suppose that the parameters are not redundant . @xmath21 is then an approximator of the number of active parameters , denoted by @xmath22 , in the resulting model . recall that the traditional ics whose minimization enables model selection can be written as @xmath23 the bic and aic criteria are obtained by setting the value of @xmath24 to @xmath25 respectively . relating ( [ eq : ic ] ) to the penalized likelihood ( [ eq : vs ] ) , one can see that in alasso , by setting @xmath26 ( @xmath24 may be @xmath27 , @xmath28 , etc . ) , the maximum penalized likelihood is closely related to the ic ( [ eq : ic ] ) . this will be rigorously studied next , and in fact _ @xmath29 _ ( instead of @xmath26 ) _ gives interesting results_. can we make the model selection results of alasso exactly the same as those based on the ics ? in fact , if the fisher information matrix is diagonal , this can be achieved by simply setting @xmath7 in alasso to @xmath30 , i.e. , maximizing the following penalized likelihood @xmath31 selects the same model as the ic ( [ eq : ic ] ) does , as seen from the following proposition . [ prop1 ] suppose that the following conditions hold the log - likelihood @xmath5 is quadratic around the mle @xmath2 , with a non - singular observed fisher information matrix @xmath32 . @xmath32 is diagonal . then the non - zero parameters selected by maximizing ( [ eq : proposed_pl ] ) are exactly those selected by minimizing the ic ( [ eq : ic ] ) . since the mle @xmath2 maximizes @xmath5 , we have @xmath33 . let @xmath34 . under the assumptions made in the proposition , the log - likelihood becomes @xmath35 the penalized likelihood ( [ eq : proposed_pl ] ) then becomes @xmath36 it is easy to show that the solutions maximizing @xmath37 are latexmath:[\[\nonumber \hat{\theta}_{i , al } = \textrm{sgn } ( \hat{\theta}_{i})\cdot \big(|\hat{\theta}_{i}| - 2\lambda_{ic } /(\mathbf{h}_{ii}\cdot @xmath18 estimated by maximizing ( [ eq : proposed_pl ] ) is non - zero if and only if @xmath39 on the other hand , the model selected by minimizing the criterion ( [ eq : ic ] ) has @xmath40 free parameters if @xmath41 and @xmath42 . according to ( [ eq : ic ] ) , we then have @xmath43 the least change in @xmath5 caused by eliminating a particular parameter has been derived in the optimal brain surgeon ( obs ) technique @xcite . here , due to the simple form of ( [ eq : like_approx ] ) , the least change in @xmath5 caused by eliminating @xmath20 , denoted by @xmath44 , can be seen directly : @xmath45_{ii } = \frac{1}{2 } [ \mathbf{h}]_{ii } \cdot \hat{\theta}_i^2.\ ] ] note that @xmath46 in ( [ eq : l_change ] ) is the minimum of @xmath44 for all @xmath40 parameters in the current model . therefore , one can see that model selection based on the ic ( [ eq : ic ] ) selects @xmath20 if and only if @xmath47 , which is equivalent to the constraint ( [ eq : criterion_pl ] ) . that is , under the assumptions made in the proposition , non - zero parameters produced by maximizing the penalized likelihood ( [ eq : proposed_pl ] ) are exactly those selected by the corresponding ic ( [ eq : ic ] ) . @xmath48 this proposition indicates that ( [ eq : proposed_pl ] ) can be considered as a continuous approximator to the ics ( [ eq : ic ] ) , which enables quick - ic ; one can see that the continuous approximator is obtained by simply replacing the maximum likelihood @xmath49 by the data likelihood @xmath5 , and replacing the number of free parameters , @xmath22 , by @xmath50 . the condition in proposition [ prop1 ] is rather restrictive ; in the linear regression scenario , it corresponds to the orthogonal design case . in the more general case , where @xmath32 is usually not diagonal , the condition for the parameters @xmath20 to be selected by maximizing ( [ eq : proposed_pl ] ) becomes more complex , and ( [ eq : ic ] ) and ( [ eq : proposed_pl ] ) are usually not exactly equivalent . we give some results on the relationship between quick - ic and model selection based on ics . [ prop2 ] suppose that condition 1 in proposition [ prop1 ] holds . assume that both the ic approach ( [ eq : ic ] ) and quick - ic ( [ eq : proposed_pl ] ) perform model selection in the backword elimination manner , i.e. , the penalization parameter is gradually increased to the target value , such that insignificant parameters are set to zero one by one . further assume that once a parameter is set to zero , it will not become non - zero again . let @xmath51 , where @xmath52 and is assumed to be nonsingular . then the ic approach ( [ eq : ic ] ) selects @xmath20 if and only if @xmath53_{ii } < \frac{1}{2\lambda_{ic}}$ ] , while quick - ic ( [ eq : proposed_pl ] ) does so if and only if @xmath53_{i\cdot}\mathbb{i } < \frac{1}{2\lambda_{ic}}$ ] , where @xmath53_{i\cdot}$ ] donotes the @xmath54th row of @xmath55 and @xmath56 is the vector of 1 s . from the proof of proposition [ prop1 ] or @xcite , one can see that the ic approach selects @xmath20 if and only if @xmath57_{ii } } > \lambda_{ic}$ ] , which is equivalent to @xmath53_{ii } < \frac{1}{2\lambda_{ic}}$ ] . let @xmath58 . on the other hand , due to condition 1 in proposition [ prop1 ] , the penalized likelihood with the penalization parameter @xmath7 is @xmath59 clearly , if @xmath7 is very small such that none of @xmath20 is set to zero , @xmath60 is maximized when @xmath61 , i.e. , @xmath62 which is equivalent to @xmath63 . consequently , we have @xmath64_{i\cdot}\mathbb{i}.\ ] ] when @xmath7 is gradually increased such that @xmath65_{j.}\mathbb{i } = 1 $ ] , @xmath66 , or equivalently @xmath67 , is set to zero . finally , when @xmath7 is increased to @xmath30 , the non - zero parameters @xmath20 selected by quick - ic satisfy @xmath53_{i\cdot}\mathbb{i } < \frac{1}{2\lambda_{ic}}$ ] . @xmath48 although in practice one may not adopt backword elimination , the above proposition helps us understand the similarity and difference between the ic approach and quick - ic . for example , if @xmath68_{ij } = 0 $ ] for all @xmath54 ( which includes proposition [ prop1 ] as a special case ) , the two approaches give the same results . of course , for finite samples , in the general case it is theoretically impossible to make parameter shrinkage - based quick - ic exactly identical to the ic approach . however , their empirical comparisons in various situations presented in sec . [ sec : compare ] suggest that they usually give the same model selection results for various sample sizes . we give the following remarks on the proposed model selection approach . firstly , the result of the proposed approach depends on @xmath2 . when the model is very large , @xmath2 may be too rough , and it is useful to update @xmath2 using a consistent estimator sometime when a smaller model is derived . minimization methods ( see , e.g. , @xcite ) . in the reweighted methods , in each iteration the penalized estimate given in the previous iteration is used to form the new weight ; in this way , the reweighted alasso penalty provides an approximator to the logarithm penalty , since @xmath69 can be locally approximated by @xmath70 plus some constant , about point @xmath71 . ] secondly , in sec . [ sec : make_continuous ] the idea of quick - ic is further applied to more complex models , by using data - dependent weights for suitable penalization functions and approximating the number of effective parameters . for example , in some cases one needs to resort to the logarithm penalty to produce sparsity of parameters , and we suggest using the corresponding data - adaptive penalty @xmath72 with @xmath73 , where @xmath74 is a very small positive number , as the penalty term , as discussed in section [ sec : make_continuous_mml ] . correspondingly , to obtain the continuous approximator of the ics , one just simply replaces the number of effective parameters in @xmath20 with @xmath75 . the proposed approach in section [ sec : equivalence ] directly applies to model selection of simple models such as regression and vector auto - regression ( var ) . var provides a convenient way for granger causality analysis @xcite , and has a lot of applications in economics , neuroscience , etc . unfortunately it usually involves quite a large number of parameters , making the ic approach impractical , while quick - ic gives efficient model selection . in this section we use simulations to investigate the performance of quick - ic . to verify the results in sec . [ sec : equivalence ] , we consider the simple linear regression problem @xmath76 , where @xmath77 is the target , @xmath78 contains predictors , and @xmath74 is the gaussian noise . we take bic as an example , i.e. , we compare bic - like quick - ic ( or quick - bic , with @xmath79 in ( [ eq : proposed_pl ] ) ) with the original bic ( [ eq : ic ] ) . we also compare them with the approach of alasso followed by the bic criterion ( alasso+bic ) : one first finds the solution path of alasso using lars , and then selects the best " model by evaluating the bic criterion with the maximum likelihood @xmath49 replaced by the likelihood of the parameter values on the solution path ( @xcite , sec . for this reason , alasso+bic is different from bic . in quick - bic , the noise variance was estimated from the full model . for bic , we searched the prediction number between 4 and 8 . 20 predictors @xmath80 were used , i.e. , @xmath81 . 14 entries of @xmath1 were set to zero . the magnitudes of the others were randomly chosen between 0.2 and 2.5 , and the signs were arbitrary . we considered three cases . case i corresponded to an orthogonal design , i.e. , all predictors are uncorrelated . in case ii , the pairwise correlation between @xmath80 and @xmath82 was set to be @xmath83 . in the last case , the covariance matrix of @xmath84 was randomly generated as @xmath85 with entries of the square matrix @xmath86 randomly sampled between @xmath87 and @xmath88 . in all cases we normalized the variance of each @xmath80 . the noise variance was @xmath88 . to see the sample size effect , we varied the sample size @xmath3 from 100 to 300 . the simulation was repeated for 100 random trials . table [ tab1 ] reports the frequency of the differences in the selected predictor numbers given by different methods . one can see that in case i , all the three methods almost always select the same number of predictors . in cases ii and iii , quick - bic still gives rather similar results to bic ; in particular , as the sample size increases , their results tend to agree with each other quickly . alasso+bic produces different models with a surprisingly noteworthy chance for both sample sizes , especially in case iii . however , it seems to be still statistically consistent in model selection , like bic ; we found that when @xmath89 , for 56 times it gave the same model as bic . as for the computational loads , bic took more than 550 times longer than quick - bic as well as alasso+bic . [ tab1 ] [ cols="^,^,^,^,^,^,^,^,^,^,^,^,^,^,^,^,^ " , ] below we focus on other frequently - used statistical models , especially some complex ones , and give continuous approximators to the ics for their model selection by extending quick - ic . we also give empirical results to illustrate the applicability and efficiency of quick - ic . for regular statistical models , under a set of regularity conditions , the asymptotic normality of @xmath90 holds . the @xmath0 penalty used in lasso can then produce sparsity of the parameters and hence perform model selection @xcite . the asymptotic properties of the variable selection techniques established in the linear regression scenario also approximately hold for regular models . for some non - regular models , it is still possible to do so . if the gradient of the log - likelihood changes slowly around @xmath90 , these penalties will successfully push insignificant parameters to zero . otherwise , one may apply penalization on suitable transformations of the parameters , instead of the original parameters . in practice , the parameters in a model often naturally belong to groups , i.e. , they are selected or discarded simultaneously @xcite . one can formalize this by introducing functions @xmath91 which allow computation of the penalties for groups of variables . generally speaking , the information criterion of the form ( [ eq : ic ] ) can be approximated by the negative penalized likelihood : @xmath92 where @xmath93 are suitable transformations of the parameters ( or selected parameters ) controlling the complexity of the model , and @xmath94 are the numbers of independent parameters associated with the group @xmath93 . minimization of the negative penalized likelihood ( [ eq : npl ] ) enables simultaneous model selection and parameter estimation . when a particular @xmath93 is pushed to zero , @xmath94 free parameters disappear , and the model complexity is reduced . how to choose @xmath93 and to calculate @xmath94 depends on the specific model . let us first consider model selection of the factor analysis ( fa ) model . in fa , the observed @xmath95-dimensional data vector @xmath96 is modeled as @xmath97 , where @xmath98 is the factor loading matrix , @xmath99 the vector of @xmath100 underlying gaussian factors , and @xmath101 the vector of uncorrelated gaussian errors with the covariance matrix @xmath102 . the factors @xmath103 and the errors @xmath104 are also mutually independent . here , we have assumed that @xmath84 is zero - mean and that the factors @xmath103 are normally distributed with zero mean and identity covariance matrix . given the factor number @xmath100 and a set of observations @xmath105 , the fa model can be fitted by maximum likelihood ( ml ) using the expectation - maximization ( em ) algorithm @xcite . but ml estimation could not determine the optimal factor number @xmath106 , since the ml does not consider the complexity of the model and it increases as @xmath100 grows . a suitable factor number gives the fa model enough capacity and avoids over - fitting . when the unconditional variances of @xmath107 are fixed , model selection of fa can be achieved by shrinking suitable columns of @xmath98 to zero . so entries in each column of @xmath98 are grouped . denote by @xmath108 the @xmath54th column of @xmath98 . note that @xmath109 is singular at @xmath110 , so penalization on @xmath111 can remove unnecessary columns in @xmath98 and consequently perform model selection . the negative penalized likelihood for approximating bic is @xmath112 where @xmath113 denotes the number of free parameters in the column of @xmath98 which is to be removed , and @xmath27 is given in ( [ eq : lambda_alasso ] ) . due to the rotation indeterminacies of the factors @xmath107 , the total number of free parameters in @xmath98 is @xmath114 . the proposed method removes columns of @xmath98 one by one . if one insignificant column of @xmath98 is shrinked to zero , the total number of free parameters in @xmath98 reduces from @xmath114 to @xmath115 . therefore , @xmath116 can be evaluated to equal @xmath117 , as the change of the number of free parameters in @xmath98 when a certain column disappears . once a column of @xmath98 is removed , @xmath100 is updated accordingly . the em algorithm for minimizing the negative penalized likelihood ( [ eq : pll_fa ] ) can be derived analogously to the derivation of that for the fa model @xcite . following @xcite , we use the local quadratic approximation ( lqa ) to approximate the penalties @xmath118 . as a great advantage , it admits a closed - form solution for @xmath98 in the m step . we would like to address the following advantages of adopting the negative penalized likelihood based on alasso , instead of the original bic criterion , for model selection . the negative penalized likelihood is easy to minimize . if the log - likelihood function is concave in the neighborhood of the maximum likelihood estimator ( like in the linear regression problem ) , the negative penalized likelihood is _ convex _ , and its minimization does not suffer from multiple local minima . the gaussian mixture model ( gmm ) models the density of the @xmath95-dimensional variable @xmath84 as a weighted sum of some gaussian densities : @xmath119 , where @xmath120 are gaussian densities with mean @xmath121 and covariance matrix @xmath122 , and @xmath123 are nonnegative weights that sum to one . bic is not suitable for model selection of mixture models , since not all data are effective for estimating the parameters specifying an individual component . instead , the mml - based model selection criterion is preferred @xcite . the message length to be minimized for model selection of gmm is @xmath124 where @xmath125 denotes the number of non - zero - probability components , and the number of free parameters in each component is @xmath126 . minimization of the above function is troublesome since it involves the discrete variable @xmath125 . below we develop an approximator to ( [ eq : mml_gmm ] ) which is continuous in @xmath123 . gmm is a typical non - regular statistical model . the expected complete - data log likelihood of gmm ( see @xcite for its formulation ) , which gives an approximation of the true data likelihood , involves @xmath127 . hence , its gradient w.r.t . @xmath123 grows very fast when @xmath128 . consequently , the @xmath0 penalty could not push insignificant @xmath123 to zero . fortunately , one can then naturally exploit the @xmath129 penalty to produce sparsity of @xmath123 . the @xmath129 penalty on @xmath123 also has the advantage of admitting a closed - form update equation for @xmath123 in the em algorithm . to avoid the discontinuity of the objective function when a component with @xmath130 vanishes , we use @xmath131 as the penalty , where @xmath132 is a small enough positive number ( we chose @xmath133 in experiments ) . let @xmath134 $ ] . inspired by the idea of adaptive weights in alasso , we can let the penalty term be @xmath135 $ ] . @xmath125 could then be approximately by 2@xmath136}$ ] . consequently , ( [ eq : mml_gmm ] ) is approximated by @xmath137\sum_{i=1}^m\hat{w}_i\log\big(\frac{\pi_i+\epsilon}{\epsilon } \big ) - l^{gm}.\end{aligned}\ ] ] the em algorithm for minimizing the function above is the same as that for maximizing the gmm likelihood , except that the update equation for @xmath123 is changed to @xmath138}{n-0.5md_f^{gm}- 0.5 [ \log(n/12 ) + d_f^{gm}+1]\sum_{j=1}^m\hat{w}_j } \big\},\end{aligned}\ ] ] where @xmath139 denotes the posterior probability that the @xmath140th point comes from the @xmath54th component . when @xmath123 becomes very small , say smaller than @xmath141 , we drop the @xmath54-th component . in practice , if the initialized model is very far from the desired one , as the model complexity reduces , it is better to occasionally update @xmath142 with the corresponding maximum likelihood estimator . now consider the mixture of factor analyzers ( mfa , @xcite ) , which has a lot of applications in pattern recognition . it assumes that the @xmath95-dimensional observations @xmath84 can be modeled as @xmath143 , where @xmath121 is the mean of the @xmath54th factor analyzer , and local factors @xmath144 , which follow @xmath145 , are independent from @xmath146 , which follow @xmath147 with @xmath148 . the factor number @xmath149 may vary for different @xmath54 . following @xcite , one can find the message length for mfa ( with some constant terms dropped ) : @xmath150 + \frac{1}{2}\sum_{i=1}^{m_{nz}}d_{fi}^{f } % \\ \label{eq : mml_mfa } & & ~~ + \frac{m_{nz}}{2}\big[\log\big(\frac{n}{12}\big)+1\big ] - l^{mfa},\end{aligned}\ ] ] where @xmath151 denotes the number of free parameters specifying the @xmath54-th factor analyzer , i.e. , @xmath152 . this function involves integers @xmath125 ( the number of factor analyzers ) and @xmath149 , @xmath153 , ( the number of factors in each factor analyzer ) . its optimization is computationally highly demanding due to the large search space of @xmath154 . using the @xmath0 and @xmath129 penalties with data - adaptive weights , we can approximate @xmath155 with the following function : @xmath156 -l^{mfa},\end{aligned}\]]where @xmath157 and @xmath158 is the alasso - based approximator to the number of free parameters in @xmath159 . after some derivations , one can see that a reasonable approximator is @xmath160 - d_f \sum_{j=1}^{k_i } \big(1 - p_\lambda(||\mathbf{a}_{i,(j)}|| ) \big ) = \frac{k_i(k_i-1)}{2 } + ( d - k_i+1)\sum_{j=1}^{k_i } p_\lambda(||\mathbf{a}_{i,(j)}||)$ ] . one can verify that @xmath158 changes very slightly when @xmath149 is reduced by shrinking columns of @xmath161 . similar to @xcite , one can derive the em algorithm for minimizing ( [ eq : npl_mml_mfa ] ) . we note that the proposed model selection methods for gmm and mfa generate new components or split any large component . for very complex problems , they may converge to local optima . if necessary , one can perform the split and merge operations @xcite after certain em iterations to improve the final results . we generated the data according to the fa model , and compared four model selection schemes , which are bic - like quick - ic given in section [ sec : fa ] ( or quick - bic ) , bic , aic , and fivefold cross - validation ( cv ) . the true factor number was @xmath162 , and the data dimension was @xmath163 . elements in @xmath98 were randomly generated between @xmath164 and @xmath165 , and the error variances @xmath166 were random numbers between 0 and 1 . when using bic , aic , or cv , we let @xmath167 and @xmath168 , while quick - bic was initialized with 8 factors . to investigate the sample size effect , we let the sample size @xmath3 be 40 and @xmath169 . in each case , we repeated all methods for 100 random trials . when @xmath170 , bic , aic , quick - bic , and cv approximately took 4 , 4 , 1.5 , and 20 seconds , respectively , for each trial . clearly quick - bic is most computationally appealing , as expected . [ fig : fa_exp ] plots the histogram of the factor numbers found by the four methods . when @xmath170 , bic ( as well as quick - bic ) seems to over - penalize the model complexity and results in a smaller factor number . but when @xmath3 is increased to 100 , its performance becomes almost the best . on the contrary , aic seems to under - penalize the complexity . in both cases , quick - bic is always similar to bic . also considering its light computational load , quick - bic is preferred among the four methods . we also tested the case @xmath171 , and found that quick - bic and bic give clearly the best results . ( 100 trials ) . the true value is 5 . top / bottom : @xmath3=40/100 . , title="fig:",width=499,height=96 ] + ( a ) bic ( b ) aic ( c ) quick - bic ( d ) cv + ( 100 trials ) . the true value is 5 . top / bottom : @xmath3=40/100 . , title="fig:",width=499,height=96 ] + ( e ) bic ( f ) aic ( g ) quick - bic ( h ) cv + we compared the approach quick - ic which minimizes the continuous version of mml ( [ eq : mml_gmm_continuous ] ) with the mml - based method proposed in @xcite ( denoted by fj s method ) , in terms of the chances of finding the preferred component number and the cpu time . here we present the results on two data sets . for each data set , we repeated each method for 100 trials . in each trial , the data were randomly generated , and for initialization , the mean of each gaussian component was randomly chosen from the data points . the results on the shrinking spiral " data set @xcite are given in fig . [ fig : gmm_exp](a - c ) , and fig . [ fig : gmm_exp](d - f ) shows the results on the triangle data " , which were obtained by rotating and shifting three sets of bivariate gaussian points following @xmath172 . for the first data set , we set @xmath173 for both methods and @xmath174 for fj s method . the cpu time taken by quick - ic and fj s method was @xmath175 and @xmath176 seconds , respectively . for the second data set , we let @xmath177 and @xmath178 for fj s method . the cpu time was about @xmath179 and @xmath180 seconds for the two methods . [ fig : gmm_exp](b , e ) and ( c , f ) give the histograms of the component numbers obtained by fj s method and quick - ic . one can see that they give similar results . however , fj method seems to produce less robust ( more disperse ) results for the spiral data . we conjecture that it is caused by the annihilation " process in fj s method @xcite : fj s method annihilates the least probable component ( with the smallest mixing weight @xmath181 ) to obtain a smaller model . this process is discontinuous , and simply uses the magnitude of the mixing weight to indicate the significance of the corresponding component . in fact the significance of a particular component also depends on its relationship to other components . as a consequence , when a component that has the least weight but is actually significant is removed , the message length @xmath182 may increase , resulting in a sub - optimal model . + ( a ) spiral data ( b ) fj s ( c ) quick - ic + + ( d ) triangle data ( e ) fj s ( f ) quick - ic + quick - ic uses the mml approximator ( [ eq : npl_mml_mfa ] ) to determine both the number of factor analyzers ( @xmath183 ) and the local factor numbers ( @xmath149 ) in the mfa model . we tested the spiral data ( figure [ fig : gmm_exp]a ) , and repeated the experiments with 50 trials . @xmath183 was initialized as @xmath184 and all of @xmath149 were initialized as @xmath185 . the number of factor analyzers learned by our approach is always between 10 and 13 ( with the chances 10 : 8/50 , 11 : 21/50 , 12 : 14/50 , and 13 : 7/50 ) . in the resulting model , most factor analyzers have 1 factor , and occasionally there is one factor analyzer with 2 factors ( with one dominating the other ) or with no factor . this is consistent with the previous results with @xmath149 _ a prior _ set to 1 @xcite . we then constructed another synthetic data set in which the local factor number varies for different factor analyzers . fig . [ fig : mfa_guy](a ) plots the data points without noise , and ( b ) shows the observed noisy data . the sample size was 5390 . quick - ic was compared with the variational bayesian method ( vbmfa , @xcite ) . we repeated both methods for 20 trials with different initializations . quick - ic and vbmfa took about @xmath165 and @xmath186 minutes for each run , and produced @xmath187 and @xmath188 factor analyzers , respectively . note that the data are clearly non - gaussian , so some factor analyzers may overlap to some extent to model the data well . fig . [ fig : mfa_guy](c ) and ( d ) show the results of the two method in one run . since quick - ic does not generate new local factor analyzers , it can not separate two factor analyzers which are initialized together . this can be alleviated by using a large @xmath183 for initialization . on the other hand , sometimes vbmfa may split one factor analyzer into two ; we found that in 2 trials vbmfa divided the `` arm '' or `` leg '' into two segments . . ( a ) noiseless points . ( b ) noisy data for analysis . ( c ) a structure learned by quick - ic , with @xmath189 . ( d ) that by vbmfa , with @xmath189 . note that ( c ) plots columns of loading matrices , while ( d ) depicts the contour of the gaussian distribution of each factor analyzer . , title="fig:",width=299 ] + ( a ) ( b ) + . ( a ) noiseless points . ( b ) noisy data for analysis . ( c ) a structure learned by quick - ic , with @xmath189 . ( d ) that by vbmfa , with @xmath189 . note that ( c ) plots columns of loading matrices , while ( d ) depicts the contour of the gaussian distribution of each factor analyzer . , title="fig:",width=288 ] + ( c ) ( d ) + we showed that under some conditions , the penalty used in adaptive lasso , which is the @xmath0 penalty with a data - dependent weight , resembles an indicator function showing if this parameter is active . this motivated us to approximate the traditional model selection criterion by the penalized likelihood with a fixed penalization parameter . the latter is continuous in the parameters , greatly facilitating the model selection procedure . we formulated this idea as the quick - ic approach . we showed that for finite samples , quick - ic produces exactly the same model as the corresponding information criterion when the fisher information matrix is diagonal . we also investigated more general cases . furthermore , for some complex and non - regular models , we provided continuous approximators to their model selection criteria , by using suitable penalty forms and data - adaptive weights . we have demonstrated that for these models , our simple approach is computationally very efficient in model selection , and that its results are similar to those produced by the corresponding ic . one line of our future research is to investigate the theoretical properties of quick - ic for non - regular models such as finite mixture models .

model selection based on classical information criteria , such as bic , is generally computationally demanding , but its properties are well studied . on the other hand , model selection based on parameter shrinkage by @xmath0-type penalties is computationally efficient . in this paper we make an attempt to combine their strengths , and propose a simple approach that penalizes the likelihood with data - dependent @xmath0 penalties as in adaptive lasso and exploits a fixed penalization parameter . even for finite samples , its model selection results approximately coincide with those based on information criteria ; in particular , we show that in some special cases , this approach and the corresponding information criterion produce exactly the same model . one can also consider this approach as a way to directly determine the penalization parameter in adaptive lasso to achieve information criteria - like model selection . as extensions , we apply this idea to complex models including gaussian mixture model and mixture of factor analyzers , whose model selection is traditionally difficult to do ; by adopting suitable penalties , we provide continuous approximators to the corresponding information criteria , which are easy to optimize and enable efficient model selection . model selection , parameter shrinkage , information criterion , adaptive lasso , factor analysis , gaussian mixture , mixture of factor analysizers

introduction relating adaptive lasso to information criteria: intuition basic approach for quick-ic model selection numerical studies extensions: quick-ic by approximating various information criteria conclusion and future work

the * platonic solids * the convex polyhedra with equivalent faces composed of congruent convex regular polygons have been known and studied by mathematicians and philosophers alike for millennia . more broadly , another class of convex polyhedra which are also well - known to mathematicians are the * archimedean solids * : the semi - regular convex polyhedra composed of two or more types of regular polygons meeting in identical vertices , with no requirement that faces be equivalent . there are three categories of such archimedean solids : ( i ) the platonic solids ; ( ii ) two infinite series solutions - the prisms and anti - prisms ; and ( iii ) fourteen further exceptional cases . in @xcite , a novel approach to the archimedean solids was taken by interpreting the graphs of these solids as clean * dessins denfants * in the sense of grothendieck . we recall that a dessin is a bipartite graph drawn on a riemann surface and that _ clean _ is the criterion that all the nodes of one of the two possible colours have valency two @xcite . the underlying riemann surface we will choose throughout will be simply the sphere @xmath1 onto which we can embed the solids ; furthermore , we will draw them in a planar projection . now , the planar graph of any polytope can be interpreted as a clean dessin by inserting a black node into every edge of the graph , and colouring every vertex white . given any such clean dessin , one can extract a remarkable array of group theoretic and algebraic properties . indeed , the motivation of @xcite was to investigate some of these properties for the interesting class of highly symmetric dessins which correspond to the archimedean solids . in a parallel vein , trivalent clean dessins can be naturally associated to conjugacy classes of subgroups of the modular group in a manner we will soon describe ; this further extends to a correspondence with k3 surfaces @xcite and to gauge theories @xcite . this paper constitutes an codification of the above directions into a unified outlook . while @xcite focused on finding the * belyi maps * corresponding to each of the dessins for the archimedean solids , here we shall approach these dessins from the group - theoretic angle discussed in @xcite , by considering their connections to subgroups of the famous * modular group * , and a more general class of groups known as * hecke groups*. recall that the modular group @xmath2 is the group of linear fractional transformations @xmath3 , with @xmath4 and @xmath5 . the presentation of @xmath6 is @xmath7 . a natural generalisation is to consider groups with presentation @xmath8 : such groups , which we shall denote @xmath9 , are the hecke groups . clearly , @xmath10 is the modular group . with this in mind , every clean dessin denfant can be associated to the conjugacy class of a subgroup of a certain hecke group , or free product of hecke groups , in the following way : black nodes are associated with elements of the cyclic group @xmath11 , while @xmath12-valent white nodes are associated with elements of the cyclic group @xmath13 . for a dessin with white nodes with multiple valencies , the different valency white nodes are associated with the different @xmath14 , so that a clean dessin with white nodes of a single valency is associated with a conjugacy class of subgroups of a certain hecke group , while a clean dessin with white nodes of multiple valencies is associated with a conjugacy class of subgroup of a group that is isomorphic to the free product of the relevant hecke groups . this association made , each dessin can be interpreted as the * schreier coset graph * of the corresponding hecke group @xcite , which is the cayley graph of the hecke group quotiented by some finite index subgroup . from this coset graph , the permutation representation for the associated conjugacy class of subgroups can be extracted . in this paper , we shall use these data to compute the generators for a representative of the conjugacy class of hecke subgroups corresponding to each of the exceptional archimedean solids as explicit @xmath15 matrices . once we have obtained these results , we shall moreover investigate whether any of these subgroups are so - called _ congruence _ subgroups . the structure of this paper is as follows . first , in section 2 we present some technical details regarding hecke groups and dessins denfants . we show that it is possible to interpret each clean dessin as the schreier coset graph for a conjugacy class of subgroups of a certain hecke group , or free product of hecke groups . we then show that the belyi map associated to the particular clean dessin in question is the map from the generalisation of the * modular curve * associated to each conjugacy class of subgroups of a hecke group in question to @xmath16 . in section 3 , we find the permutations for the conjugacy classes of subgroups of hecke groups corresponding to every archimedean solid , and provide explicit generating sets of matrices for representatives of those conjugacy classes of subgroups in the exceptional cases . finally , we remark on the congruence properties of these subgroups . in this section , we first recall some essential details regarding both hecke groups and clean dessins denfants . we begin by considering the modular group @xmath17 , as although this is isomorphic to only one particular hecke group , it is by far the most well - studied , and we shall draw upon the presented results at many points in the ensuing discussion . subsequently , we discuss hecke groups more generally , before moving on to consider clean dessins and their associated belyi maps . with these well - known results in hand , we then describe how every such clean dessin is isomorphic to the schreier coset graph for a certain conjugacy class of subgroups of a hecke group , or free product of hecke groups ( henceforth we may refer to such conjugacy classes as _ classes of hecke subgroups _ ; where there is no ambiguity we may use @xmath18 to denote such classes ) . we then take a short digression in order to resolve some tangential questions regarding the construction of these schreier coset graphs which have yet to be presented explicitly in the literature . next , we describe the connection between the belyi maps associated to the clean dessins in question and the algebraic curves corresponding to precisely the same classes of hecke subgroups as arose in the discussion of coset graphs . to begin our discussion , we recall some essential details regarding the modular group @xmath6 . this is the group of linear fractional transformations @xmath19 , with @xmath4 and @xmath5 . it is generated by the transformations @xmath20 and @xmath21 defined by : @xmath22 the presentation of @xmath6 is @xmath23 , and we will later discuss the presentations of certain modular subgroups . the @xmath24 matrices for @xmath21 and @xmath20 are as follows : @xmath25 letting @xmath26 and @xmath27 denote the elements of order 2 and 3 respectively , we see that @xmath6 is the free product of the cyclic groups @xmath28 and @xmath29 . it follows that @xmath24 matrices for @xmath30 and @xmath31 are : @xmath32 with these basic details in hand , we should consider some notable subgroups of @xmath6 . the most important subgroups of @xmath6 are the so - called _ congruence _ subgroups , defined by having the the entries in the generating matrices @xmath21 and @xmath20 obeying some modular arithmetic . some conjugacy classes of congruence subgroups of particular note are the following : * principal congruence subgroups : @xmath33 * congruence subgroups of level @xmath34 : subgroups of @xmath6 containing @xmath35 but not any @xmath36 for @xmath37 ; * unipotent matrices : @xmath38 * upper triangular matrices : @xmath39 * the congruence subgroups @xmath40 for certain choices of @xmath34 , @xmath41 , @xmath42 , @xmath43 . we note here that : @xmath44 in section 3 of this paper , we shall remark on the connections between some specific conjugacy classes of congruence modular subgroups and the archimedean solids . we can now extend our discussion of the modular group @xmath17 to the more general hecke groups @xmath9 . the hecke group @xmath9 has presentation @xmath8 , and is thus the free product of cyclic groups @xmath11 and @xmath13 . @xmath9 is generated by transformations @xmath20 and @xmath21 now defined by : @xmath45 where @xmath46 is some real number to be determined . the @xmath24 matrices for these @xmath21 and @xmath20 are : @xmath47 letting @xmath26 and @xmath27 as in our discussion of @xmath6 , we see that @xmath24 matrices for @xmath30 and @xmath31 are : @xmath48 for @xmath9 , we clearly have @xmath49 , thereby constraining @xmath46 for a given @xmath12 @xcite . diagonalizing @xmath31 to compute @xmath50 explicitly places a constraint which allows for a solution of @xmath51 , we find that the following general expression , as well as some important values for small @xmath12 : @xmath52 in particular , @xmath51 are algebraic numbers . by way of extension of the above discussion of subgroups of the modular group @xmath6 , it is useful to consider congruence subgroups of hecke groups . as stated in @xcite , the hecke groups are discrete subgroups of @xmath53 ; in fact , the matrix entries are in @xmath54 $ ] , the extension of the ring of integers by the algebraic number @xmath46 . note that : @xmath55\right ) \ .\ ] ] however , unlike the special case of the modular group , this inclusion is strict . with this point in mind , we can define the congruence subgroups of hecke groups in the following way @xcite . let @xmath56 be an ideal of @xmath54 $ ] . we then define : @xmath57,i\right)=\left\ { \begin{pmatrix}a & b\\ c & d \end{pmatrix}\in\mathrm{psl}\left(2,\mathbb{z}\left[\lambda_{n}\right]\right)\;;\;a-1,b , c , d-1\in i\right\ } \ .\ ] ] by analogy , we also define : @xmath58,i\right ) & = \left\ { \begin{pmatrix}a & b\\ c & d \end{pmatrix}\in\mathrm{psl}\left(2,\mathbb{z}\left[\lambda_{n}\right]\right)\;;\;a-1,c , d-1\in i\right\ } \\ \mathrm{psl^{0}}\left(2,\mathbb{z}\left[\lambda_{n}\right],i\right ) & = \left\ { \begin{pmatrix}a & b\\ c & d \end{pmatrix}\in\mathrm{psl}\left(2,\mathbb{z}\left[\lambda_{n}\right]\right)\;;\;c\in i\right\ } \ .\end{aligned}\ ] ] then we can define the congruence subgroups @xmath59 , @xmath60 and @xmath61 of @xmath9 as follows : @xmath62,i\right)\cap h_{n } \ ; \\ { \nonumber}h_{n}^{1}\left(i\right)&=\mathrm{psl^{1}}\left(2,\mathbb{z}\left[\lambda_{n}\right],i\right)\cap h_{n } \ ; \\ h_{n } \ .\end{aligned}\ ] ] as noted in @xcite , we clearly have : @xmath63 by analogy with our discussion of the modular group , we define congruence subgroups of level @xmath34 of the hecke group @xmath9 as subgroups of @xmath9 containing @xmath59 but not any @xmath64 for @xmath65 @xcite . with these details regarding hecke groups and their subgroups in hand , we can now consider their connections to clean dessins denfants . a * dessin denfant * in the sense of grothendieck is an ordered pair @xmath66 , where @xmath67 is an oriented compact topological surface and @xmath68 is a finite graph * * satisfying the following conditions @xcite : 1 . @xmath69 is connected . 2 . @xmath69 is _ bipartite _ , i.e. consists of only black and white nodes , such that vertices connected by an edge have different colours . @xmath70 is the union of finitely many topological discs , which we call the _ faces_. we can interpret any polytope as a dessin by inserting a black node into every edge , and colouring all vertices white . this process of inserting into each edge a bivalent node of a certain colour is standard in the study of dessins denfants and gives rise to so - called _ clean _ dessins , i.e. , those for which all the nodes of one of the two possible colours have valency two . an example of this procedure for the cube is shown in figure [ f : dessineg ] . now recall that there is a one - to - one correspondence between dessins denfants and * belyi maps * @xcite . a belyi map is a holomorphic map to @xmath16 ramified at only @xmath71 , i.e. for which the only points @xmath72 where @xmath73 are such that @xmath74 . we can associate a belyi map @xmath75 to a dessin via its _ ramification indices _ : the order of vanishing of the taylor series for @xmath75 at @xmath72 is the ramification index @xmath76 at that @xmath77th ramification point @xcite . to draw the dessin from the map , we mark one white node for the @xmath77th pre - image of 0 , with @xmath78 edges emanating therefrom ; similarly , we mark one black node for the @xmath79th pre - image of 1 , with @xmath80 edges . we connect the nodes with the edges , joining only black with white , such that each face is a polygon with @xmath81 sides @xcite . the converse direction ( from dessins to belyi maps ) is detailed in e.g. @xcite . as already discussed , the hecke group @xmath9 has presentation @xmath8 , and is thus the free product of cyclic groups @xmath11 and @xmath13 . given the free product structure of @xmath9 , we see that its cayley graph is an infinite free @xmath12-valent tree , but with each node replaced by an oriented @xmath12-gon . now , for any _ finite index _ subgroup @xmath82 of @xmath9 , we can quotient the cayley graph to arrive at a finite graph by associating nodes to right cosets and edges between cosets which are related by action of a group element . in other words , this graph encodes the permutation representation of @xmath9 acting on the right cosets of @xmath82 . this is called a * schreier coset graph * , sometimes also referred to in the literature as a _ schreier - cayley coset graph _ , or simply a _ coset graph_. there is a direct connection between the schreier coset graphs and the dessins denfants for each class of hecke subgroups @xcite : the dessins denfants for a certain conjugacy class of subgroups of a hecke group @xmath9 can be constructed from the schreier coset graphs by replacing each positively oriented @xmath12-gon with a white node , and inserting a black node into every edge . conversely , the schreier coset graphs can be constructed from the dessins by replacing each white node with a positively oriented @xmath12-gon , and removing the black node from every edge . let @xmath83 and @xmath84 denote the permutations induced by the respective actions of @xmath30 and @xmath31 on the cosets of each subgroup . we can find a third permutation @xmath85 by imposing the following condition , thereby constructing a * permutation triple * : @xmath86 the permutations @xmath83 , @xmath84 and @xmath85 give the * permutation representations * of @xmath9 on the right cosets of each subgroup in question . as elements of the symmetric group , @xmath83 and @xmath84 can be easily computed from the schreier coset graphs by following the procedure elaborated in @xcite , i.e. , by noting that the doubly directed edges represent an element @xmath30 of order 2 , while the positively oriented triangles represent an element @xmath31 of order @xmath12 . since the graphs are connected , the group generated by @xmath30 and @xmath31 is transitive on the vertices . clearly , @xmath83 and @xmath84 tell us which vertex of the coset graph is sent to which , i.e. which coset of the modular subgroup in question is sent to which by the action of @xmath9 on the right cosets of this subgroup . consider now a clean dessin denfant with white nodes with @xmath34 different valencies ( in the above , we assumed that all the white nodes of the dessin had the same valency ) . it is still possible to interpret this dessin as a schreier coset graph , though in this case the coset graph will correspond to a conjugacy class of subgroups of a group isomorphic to the free product of cyclic groups @xmath87 , where the @xmath88 denote the different valencies of the white nodes of the dessin . since , to recall , @xmath89 , we see that dessins with _ multiple _ white node valencies correspond to conjugacy classes of subgroups of free products of hecke groups . as detailed above , schreier coset graphs are built from simple edges @xmath90 and @xmath12-gons @xmath91 which are assumed to be positively oriented . reversal of @xmath12-gon orientation corresponds to applying the automorphism of @xmath9 that inverts each of the two generators @xmath30 and @xmath31 . at this point though , it is worth answering an intriguing tangential question : to what conjugacy classes of subgroups do the coset graphs correspond when _ arbitrary _ @xmath12-gons are reversed in orientation ? to answer this question , first consider the schreier coset graphs for the conjugacy classes of congruence subgroups of the modular group @xmath92 and @xmath93 , as shown in figure [ f : flip ] ( these coset graphs were originally presented in @xcite ; the associated dessins denfants are drawn in @xcite ) . blue lines correspond to permutations @xmath83 generated by @xmath30 , while red triangles correspond to permutations @xmath84 generated by @xmath31 , and we assume initially that all such triangles are positively oriented . as stated in @xcite , these two coset graphs are not isomorphic , as they have different circuits . + + what happens if we swap the orientation of an arbitrary triangle in one of these coset graphs ? for example , what happens if we choose the third - from - left triangle in the coset graph for @xmath92 to be _ negatively _ oriented rather than positively oriented ? such a modified coset graph is drawn in figure [ f : gamma0 - 8 ] . to answer this question , we read the relevant permutations off of the original and altered coset graph . in the original case , the corresponding conjugacy class of modular subgroups @xmath92 is generated by the permutations : @xmath94 in the second , altered case , the corresponding conjugacy class of modular subgroups is that generated by the permutations : @xmath95 note that the triple @xmath96 has changed to @xmath97 : this is the only change . now , at this point we can use the @xcite algorithm presented in section 3.2 of @xcite to compute a generating set of matrices for a representative of the conjugacy class of subgroups corresponding to our original and modified coset graphs , taking the above permutations as input . of course , we would need to modify the algorithm in order to study classes of hecke subgroups more generally ; we present this modified code explicitly in the appendix . in the original case , such a generating set is : @xmath98 from the definition of @xmath99 , we see that these are generators for a representative of the conjugacy class of congruence subgroups @xmath100 , as expected . in the modified case , a generating set is given by : @xmath101 brief reflection on these matrices suffices to make apparent that these are generators of a representative of the conjugacy class of congruence subgroups @xmath93 . by reversing the orientation of the chosen triangle in the coset graph for @xmath100 , we have obtained the permutations for @xmath93 . hence , this modified coset graph no longer corresponds to @xmath100 , but rather to @xmath93 . the reasons for this are easily stated in general terms : swapping @xmath12-gon orientations in schreier coset graphs changes the permutations defining the associated conjugacy class of subgroups of the relevant hecke group . these permutations need not define the same conjugacy class of subgroups as the original permutations . in fact , in general they will _ not _ , and will instead define the subgroup whose schreier coset graph with all _ positively _ oriented @xmath12-gons has the same cycles as the modified coset graph , with arbitrary @xmath12-gon orientations . this final point is nicely viewed in geometrical terms for our specific example : by choosing the third - from - left triangle in the coset graph for @xmath100 to be negatively oriented , we obtain a coset graph which has the same cycles as that for @xmath93 when all the triangles are chosen to be positively oriented . let @xmath102 be prime and let @xmath103 be the subgroup of matrices congruent to ( plus or minus ) the identity modulo @xmath102 . then @xmath104 acts on the completed upper half - plane @xmath105 , and the quotient @xmath106 can be given the structure of a riemann surface , known as a * modular curve*. defining the subgroup @xmath107 , the @xmath79-map @xmath108 is a galois cover ramified ( with suitable normalisation ) at @xmath71 . @xmath79 is a belyi map from @xmath109 to @xmath16 associated to a dessin drawn on the riemann surface @xmath109 . generalising now from @xmath17 to all hecke groups @xmath9 , we first note that , as detailed in @xcite , the quotient space @xmath110 is a sphere with one puncture and two elliptic fixed points of order 2 and @xmath12 . therefore , all hecke groups @xmath9 can be considered as triangle groups , and the hecke surface @xmath110 is a riemann surface @xcite . clearly , we are interested in the class of algebraic curves defined in analogy to the modular curve above , with the property that there exists a hecke subgroup @xmath111 such that the map @xmath112 is a galois cover ramified at @xmath71 . with this in mind , let @xmath67 be a compact riemann surface which admits a regular tessellation by hyperbolic @xmath12-gons . such a surface admits reconstruction from a torsion - free subgroup @xmath82 of finite index in a hecke group @xmath9 in the following way . the compact surface @xmath113 obtained from the group @xmath82 has finitely many _ cusps _ added to the surface @xmath114 , one for each @xmath82-orbit of boundary points at which the stabiliser is non - trivial . the holomorphic projection @xmath115 induced from the identity map on the universal covering extends to a belyi map @xmath116 from the compactification , with each cusp of @xmath67 mapped by @xmath117 to the single cusp @xmath118 of @xmath9 , which compactifies @xcite the plane @xmath110 to @xmath119 . the connection to dessins is then as follows : choose any particular clean dessin denfant . we have already described how this dessin can be interpreted as a schreier coset graph for a particular hecke group @xmath9 ( or free product of hecke groups ) . in addition though , the above results demonstrate that the belyi rational map associated to the dessin is precisely the map @xmath120 , with each cusp of @xmath67 mapped by @xmath117 to the single cusp @xmath118 of @xmath9 . with the results of the previous section in hand , we can proceed to apply the theory to the so - called _ archimedean solids_. in this section , we first give a technical definition of these geometrical entities . we then identify the hecke group for which each of the archimedean solids has a corresponding conjugacy class of subgroups , before giving the permutation representations @xmath83 and @xmath84 for each corresponding class . we then ( in every case where possible ) find a representative of these classes of subgroups associated to each archimedean solid by specifying a generating set for that representative , following the procedure in @xcite . finally , we follow up some residual questions from @xcite , investigating whether the trivalent archimedean solids , when interpreted as dessins denfants , correspond in general to conjugacy classes of _ congruence _ subgroups of the modular group @xmath6 . the platonic solids are well - known to us , and are the regular , convex polyhedra . they are the tetrahedron , cube , octahedron , dodecahedron and icosahedron . in order to introduce the wider class of so - called _ archimedean solids _ , consider planar graphs without loops , and with vertices of degree @xmath121 . following @xcite , let us call the list of numbers @xmath122 , where the @xmath123 are the number of edges of the adjacent faces taken in the counter - clockwise direction around the vertex , the _ type _ of that vertex . two such lists are _ equivalent _ if one can be obtained from the other by ( a ) making a cyclic shift and ( b ) inverting the order of the @xmath123 . a solid is called * archimedean * if the types of all its vertices are equivalent @xcite . we emphasize a subtlety here . as stated in the introduction , one informal way of defining the archimedean solids is as the semi - regular convex polyhedra composed of two or more types of regular polygons meeting in identical vertices , with no requirement that faces be equivalent . _ identical vertices _ is usually taken to mean that for any two vertices , there must be an isometry of the entire solid that takes one vertex to the other . sometimes , however , it is only required that the faces that meet at one vertex are related isometrically to the faces that meet at the other . on the former definition , the so - called _ pseudorhombicuboctahedron _ , otherwise known as the _ elongated square gyrobicupola _ , is not considered an archimedean solid ; on the latter it is . the formal definition of the archimedean solids above corresponds to the second , broader definition of identical vertices ; following @xcite , it is this definition which we shall use . for a more extended discussion of this point , the reader is referred to @xcite . [ cols="^,^,^,^,^,^,^,^,^,^",options="header " , ] it turns out the the solids which satisfy the above archimedean condition are : * the five platonic solids ; * two infinite series ( the prisms and anti - prisms ) ; and * fourteen exceptional solutions , and furthermore to arnold s simple surface singularities which have no deformations@xcite . it is curious that the next order generalization of the simple singularities which has exactly 1 deformation modulus , have precisely 14 exceptional cases . we conjecture , therefore , some connection to the exceptional archimedean solids . ] . all these solids are listed in table [ archtable ] , and are drawn in figures [ f : type - i ] to [ f : type - iii ] ( for the @xmath12-prisms and @xmath12-antiprisms , we give examples for small @xmath12 ) . all these images were constructed using @xcite . for completeness , we also give the symmetry groups of each of these solids in the column labelled `` sym '' in this table . these are given in standard so - called _ schnflies notation_. we recognize the standard tetrahedral , octahedral and icosahedral groups @xmath124 as well as the dihedral group @xmath125 . the extra subscript @xmath41 and @xmath126 denote extra symmetries about a horizontal ( h ) mirror plane or a diagonal ( d ) plane . tetrahedron cube octahedron dodecahedron icosahedron 3-prism 4-prism 5-prism 6-prism 7-prism 3-antiprism 4-antiprism 5-antiprism 6-antiprism 7-antiprism truncated tetrahedron truncated cube truncated octahedron truncated icosahedron truncated dodecahedron truncated cuboctahedron truncated icosidodecahedron cuboctahedron icosidodecahedron rhombicuboctahedron rhombicosidodecahedron pseudorhombicuboctahedron snub cube snub dodecahedron clearly , we can interpret all the archimedean solids ( and indeed , as mentioned , every other polytope ) as clean dessins denfants by inserting a black node into every edge , and colouring every vertex white . we are not the first to interpret the archimedean solids in this way : in @xcite , the authors do precisely this , computing the associated belyi map in each case ; thus the work in this section can be viewed as complimentary material to that work . by the correspondences detailed in the previous section , we should be able to associate a class of hecke subgroups to every archimedean solid . this task shall be undertaken in the following subsection . can the archimedean solids all be associated with conjugacy classes of subgroups of the modular group , and are those subgroups congruence ? the answer to the first question is straightforwardly _ no _ : only the trivalent solids are such that , interpreted as dessins , they are subgroups of the modular group , which is isomorphic to the free product of cyclic groups @xmath127 and @xmath128 , as detailed above . hence , only those archimedean solids labelled ` yes ' in the column ` trivalent ? ' of table [ archtable ] can be associated with subgroups of the modular group . the rest _ can _ be associated with conjugacy classes of subgroups of other hecke groups , however . for example , the cuboctahedron is 4-valent @xcite , and can thus be associated with a conjugacy class of subgroups of the hecke group @xmath129 . the hecke group related to each of the archimedean solids in this way is presented in the final column of table [ archtable ] . in the remainder of this subsection , we give the permutations @xmath83 and @xmath84 for the conjugacy class of subgroups of the relevant hecke group for each archimedean solid . we also find explicitly a representative of the conjugacy class of subgroups of the relevant hecke group to which each of the archimedean solids corresponds ( bar the two infinite series cases ) . the @xcite algorithm to achieve this task is presented in the appendix of this paper . since the platonic solids have already been partly studied in e.g. @xcite , we divide the task into three sections , based upon the three categories of archimedean solid presented in the previous subsection . interpreted as clean dessins denfants , the tetrahedron , cube , and dodecahedron correspond to the conjugacy classes of principal congruence subgroups @xmath130 , @xmath131 and @xmath132 , respectively . this can be verified explicitly from the dessins denfants presented in @xcite , where we can also see that the index of the associated class of modular subgroups is twice the number of edges of the platonic solid in question . for convenience , these results are tabulated in table [ t : platonics ] . now , reflection on table [ archtable ] suffices to show that the remaining two platonic solids - the octahedron and the icosahedron - will not correspond to subgroups of @xmath6 , but rather to @xmath133 and @xmath134 , respectively . what are these subgroups ? we can answer this question by finding a generating set for a representative of the corresponding class in each case . [ [ octahedron ] ] octahedron : + + + + + + + + + + + permutations @xmath83 and @xmath84 for the conjugacy class of subgroups corresponding to the octahedron are : @xmath135 using the procedure detailed in the appendix , generators for a representative of the conjugacy class of subgroups of @xmath133 for the octahedron are : @xmath136 comparison with our results concerning congruence subgroups of hecke groups in section 2.2.1 above suffices to show that the octahedron corresponds to the principal congruence subgroup @xmath137 . we remark here that , as stated in @xcite , the matrices of @xmath133 are of the following two types : @xmath138 the elements of the first type form a subgroup of index 2 in @xmath133 ; all the subgroups of @xmath133 corresponding to the archimedean solids will turn out to be subgroups of this subgroup . [ [ icosahedron ] ] icosahedron : + + + + + + + + + + + + the elements of @xmath134 need not fall into such simple categories as those for @xmath133 as given in equation ( 3.4 ) above . the consequence is that the elements of the matrices of the respesentiative generating set for the icosahedron , snub cube and snub dodecahedron ( to be discussed below ) , which correspond to conjugacy classes of subgroups of @xmath134 , are too complicated to present on paper , though they are given in the document which accompanies this paper . for this reason , in these three cases we resort in this paper proper to providing just the permutations @xmath83 and @xmath84 . permutations @xmath83 and @xmath84 for the conjugacy class of subgroups corresponding to the icosahedron are : @xmath139 the prisms and antiprisms form infinite series of archimedean solids . since all prisms are trivalent , all correspond to a conjugacy class of subgroups of @xmath6 ; since all antiprisms are 4-valent , all correspond to a conjugacy class of subgroups of @xmath133 . given that the prisms and antiprisms form infinite series , we can not write down a generating set of a representative of the conjugacy class of subgroups for each . nevertheless , we _ can _ construct general expressions for the permutations @xmath83 and @xmath84 for the prisms and antiprisms . these expressions can be used to find the permutations @xmath83 and @xmath84 for any particular ( anti)prism of interest . these permutations can in turn be input into the code presented in the appendix in order to compute a generating set for a representative of the relevant class of hecke subgroups for any particular ( anti)prism of interest . [ [ prisms ] ] prisms : + + + + + + + call the @xmath12-gon faced prism the @xmath12-prism . the @xmath12-prism has the following permutations @xmath83 and @xmath84 for @xmath30 and @xmath31 , respectively : @xmath140 this renders it a trivial matter to find a generating set for a representative of the relevant conjugacy class of modular subgroups for any prism , and moreover to check whether any @xmath12-prism corresponds to a conjugacy class of congruence subgroups of @xmath6 ( a topic to which we shall return in the following subsection ) . [ [ antiprisms ] ] antiprisms : + + + + + + + + + + + call the @xmath12-gon faced anti - prism the @xmath12-antiprism . the @xmath12-antiprism has the following permutations @xmath83 and @xmath84 for @xmath30 and @xmath31 , respectively : @xmath141 again , this renders it straightforward to find a generating set for a representative of the relevant conjugacy class of subgroups for any antiprism . in addition to the platonic solids and the prisms and antiprisms , there remain fourteen exceptional archimedean solids , as given in table [ archtable ] . here we give the permutations and compute a generating set for a representative of the corresponding class of hecke subgroups . [ [ truncated - tetrahedron ] ] truncated tetrahedron : + + + + + + + + + + + + + + + + + + + + + + permutations @xmath83 and @xmath84 for the conjugacy class of subgroups corresponding to the truncated tetrahedron are : @xmath142 generators for a representative of the conjugacy class of subgroups of @xmath6 for the truncated tetrahedron are : @xmath143 [ [ truncated - cube ] ] truncated cube : + + + + + + + + + + + + + + + permutations @xmath83 and @xmath84 for the conjugacy class of subgroups corresponding to the truncated cube are : @xmath144 generators for a representative of the conjugacy class of subgroups of @xmath6 for the truncated cube are : @xmath145 [ [ truncated - octahedron ] ] truncated octahedron : + + + + + + + + + + + + + + + + + + + + + permutations @xmath83 and @xmath84 for the conjugacy class of subgroups corresponding to the truncated octahedron are : @xmath146 generators for a representative of the conjugacy class of subgroups of @xmath6 for the truncated octahedron are : @xmath147 [ [ truncated - icosahedron ] ] truncated icosahedron : + + + + + + + + + + + + + + + + + + + + + + permutations @xmath83 and @xmath84 for the conjugacy class of subgroups corresponding to the truncated icosahedron are : @xmath148 generators for a representative of the conjugacy class of subgroups of @xmath6 for the truncated icosahedron are : @xmath149 [ [ truncated - dodecahedron ] ] truncated dodecahedron : + + + + + + + + + + + + + + + + + + + + + + + permutations @xmath83 and @xmath84 for the conjugacy class of subgroups corresponding to the truncated dodecahedron are : @xmath150 generators for a representative of the conjugacy class of subgroups of @xmath6 for the truncated dodecahedron are : @xmath151 [ [ truncated - cuboctahedron ] ] truncated cuboctahedron : + + + + + + + + + + + + + + + + + + + + + + + + permutations @xmath83 and @xmath84 for the conjugacy class of subgroups corresponding to the truncated cuboctahedron are : @xmath152 generators for a representative of the conjugacy class of subgroups of @xmath6 for the truncated cuboctahedron are : @xmath153 [ [ truncated - icosidodecahedron ] ] truncated icosidodecahedron : + + + + + + + + + + + + + + + + + + + + + + + + + + + + permutations @xmath83 and @xmath84 for the conjugacy class of subgroups corresponding to the truncated icosidodecahedron are : @xmath154 generators for a representative of the conjugacy class of subgroups of @xmath6 for the truncated icosidodecahedron are : @xmath155 [ [ cuboctahedron ] ] cuboctahedron : + + + + + + + + + + + + + + permutations @xmath83 and @xmath84 for the conjugacy class of subgroups corresponding to the cuboctahedron are : @xmath156 generators for a representative of the conjugacy class of subgroups of @xmath133 for the cuboctahedron are : @xmath157 [ [ icosidodecahedron ] ] icosidodecahedron : + + + + + + + + + + + + + + + + + + permutations @xmath83 and @xmath84 for the conjugacy class of subgroups corresponding to the icosidodecahedron are : @xmath158 generators for a representative of the conjugacy class of subgroups of @xmath133 for the icosidodecahedron are : @xmath159 [ [ rhombicuboctahedron ] ] rhombicuboctahedron : + + + + + + + + + + + + + + + + + + + + permutations @xmath83 and @xmath84 for the conjugacy class of subgroups corresponding to the rhombicuboctahedron are : @xmath160 generators for a representative of the conjugacy class of subgroups of @xmath133 for the rhombicuboctahedron are : @xmath161 [ [ rhombicosidodecahedron ] ] rhombicosidodecahedron : + + + + + + + + + + + + + + + + + + + + + + + permutations @xmath83 and @xmath84 for the conjugacy class of subgroups corresponding to the rhombicosidodecahedron are : @xmath162 generators for a representative of the conjugacy class of subgroups of @xmath133 for the rhombicosidodecahedron are : @xmath163 [ [ pseudorhombicuboctahedron ] ] pseudorhombicuboctahedron : + + + + + + + + + + + + + + + + + + + + + + + + + + permutations @xmath83 and @xmath84 for the conjugacy class of subgroups corresponding to the pseudorhombicuboctahedron are : @xmath164 generators for a representative of the conjugacy class of subgroups of @xmath133 for the pseudorhombicuboctahedron are : @xmath165 [ [ snub - cube ] ] snub cube : + + + + + + + + + + permutations @xmath83 and @xmath84 for the conjugacy class of subgroups corresponding to the snub cube are : @xmath166 [ [ snub - dodecahedron ] ] snub dodecahedron : + + + + + + + + + + + + + + + + + + permutations @xmath83 and @xmath84 for the conjugacy class of subgroups corresponding to the snub dodecahedron are : @xmath167 we have thus identified the class of hecke subgroups to which each of the archimedean solids , interpreted as clean dessins , correspond . moreover , where appropriate , we have explicitly provided generating sets of matrices for representatives of those classes . in the three cases of subgroups of @xmath168 , the snub cube , snub dodecahedron and the icosahedron , the matrices are too large to present in the text . we include these three in the accompanying file ` generators_for_h5_subgroups.txt ` . we have seen that , interpreted as clean dessins denfants , the tetrahedron , cube , and dodecahedron correspond to the conjugacy classes of principal congruence subgroups @xmath130 , @xmath131 and @xmath132 , respectively @xcite . in addition to these three arising as the dessins corresponding to certain congruence subgroups of the modular group , the dessins for the 33 conjugacy classes of genus zero , torsion - free congruence subgroups ( discussed in detail in @xcite ) reveal that _ other _ archimedean solids also correspond to certain conjugacy classes of congruence subgroups of the modular group . specifically , the truncated tetrahedron is associated with the modular subgroup @xmath169 , while the 8-prism is associated with the subgroup @xmath170 , as shown in figure [ f : tet_8prism ] . + + [ f : tet_8prism ] given these results , the question arises as to whether the other trivalent archimedean solids can also be associated with congruence subgroups of the modular group . in order to answer this question , we use the @xcite command ` is_congruence ( ) ` to determine whether that conjugacy class of modular subgroups , defined by permutations @xmath83 and @xmath84 , is congruence . below we give two examples of this procedure : one for the truncated tetrahedron ( which , for the reasons explained above , we expect to be identified as congruence ) , and one for the 3-prism ( the congruence properties of which are unknown ) . the clean dessin denfant of the truncated tetrahedron is drawn above . as discussed , this can be interpreted as a schreier coset graph by replacing every white node with a positively oriented triangle . we can then extract the permutations @xmath83 and @xmath84 for @xmath30 and @xmath31 . these permutations are given in the previous subsection . with these in hand , we can then proceed to check whether the modular subgroup associated to the truncated tetrahedron is congruence by implementing the following code in : ` sage : g = arithmeticsubgroup_permutation ( ` + ` s2="(1,32)(3,4)(2,7)(5,9)(6,24)(8,10)(11,13)(12,18)(15,16 ) ` + ` ( 17,19)(14,28)(29,31)(33,34)(30,36)(26,35)(20,25)(23,27)(21,22 ) " , ` + ` s3="(1,2,3)(4,5,6)(7,8,9)(10,11,12)(13,14,15)(16,17,18 ) ` + ` ( 19,20,21)(22,23,24)(25,26,27)(28,29,30)(31,32,33)(34,35,36 ) " ) ` + ` sage : g.is_congruence ( ) ` + ` true ` hence we see that the 3-prism is an archimedean solid whose corresponding modular subgroup _ is _ congruence . this is to be expected : we already know that the truncated tetrahedron corresponds to the congruence subgroup @xmath169 @xcite . permutations @xmath83 and @xmath84 for this dessin corresponding to the 3-prism , obtained from our general formula presented in are : @xmath171 we can now check whether the modular subgroup associated to the 3-prism is congruence by implementing the following code in : ` sage : g2 = arithmeticsubgroup_permutation ( ` + ` s2=(1,6)(4,9)(7,3)(2,10)(5,13)(8,16)(11,15)(14,18)(17,12 ) , ` + ` s3=(1,2,3)(4,5,6)(7,8,9)(10,11,12)(13,14,15)(16,17,18 ) ) ` + ` sage : g2.is_congruence ( ) ` + ` false ` hence we see that the 3-prism is an archimedean solid whose corresponding modular subgroup is _ not _ congruence . our result for the 3-prism above demonstrates that the prisms in general do not correspond to conjugacy classes of congruence subgroups of the modular group , when interpreted as clean dessins denfants . however , we know that _ some _ prisms _ do _ correspond to conjugacy classes of congruence subgroups : we have seen that the 8-prism corresponds to the conjugacy class of subgroups @xmath170 . we hence conclude that the prisms do not _ in general _ correspond to conjugacy classes of congruence subgroups of the modular group , though there are exceptions . repeating the working presented in the examples above for the remaining trivalent , exceptional archimedean solids gives the following result : the truncated tetrahedron is the _ only _ trivalent , exceptional archimedean solid which corresponds to a conjugacy class of congruence subgroups of the modular group . in this paper , we have elaborated the connections between clean dessins denfants and conjugacy classes of subgroups of hecke groups , or free products of hecke groups . we are not the first to notice this connection : though framed in terms of maps rather than dessins , the link between hecke groups and finite graphs was discussed in depth in @xcite . this current work extends the connections between dessins denfants , group theory , and algebraic curves - which has previously focused largely on the modular group @xmath17 - to a much wider class of groups . when it comes to applying this theory , we have first shown how , through interpretation as clean dessins denfants , each of the archimedean solids can be associated with a conjugacy class of subgroups of hecke groups . we have explicitly computed the matrix generators of these corresponding hecke subgroups . moreover , we have presented a number of results regarding the congruence properties of these subgroups . this work is useful not only as an illustration of the connections between hecke groups and clean dessins , but also because it develops one more area of study of the archimedean solids , and extends the work in this area already carried out in @xcite . possible developments of this work abound . for example , one might be interested in establishing whether the conjugacy classes of subgroups of hecke groups for the archimedean solids which correspond to subgroups of @xmath133 and @xmath134 are congruence . no general algorithm for checking the congruence properties of a hecke group currently exists . there is , however , an algorithm for determining whether a subgroup of @xmath6 is congruence @xcite , which we have employed ( via implementation in @xcite ) in this paper . to extend this algorithm to general hecke groups is clearly an interesting task for future pursuit . from an algebro - geometrical point of view , the subgroups of the modular group can be quotiented , upon adjoining cusps , to give modular curves which are riemann surfaces of some genus . the tetrahedron , cube and dodecahedron correspond to @xmath172 for @xmath173 , which are all genus 0 , torsion free subgroups of @xmath6 . it would be interesting to explore the geometry of all the riemann surfaces obtainable from the list of archimedean solids . indeed , physical interpretations of such dessins in relation to the modular curves have been studied in @xcite , to further investigate this direction is clearly of interest . we gratefully acknowledge many helpful comments from john mckay . yhh would like to thank the science and technology facilities council , uk , for an advanced fellowship and for stfc grant st / j00037x/1 , the chinese ministry of education , for a chang - jiang chair professorship at nankai university , the city of tian - jin for a qian - ren scholarship , the us nsf for grant ccf-1048082 , as well as city university , london , the department of theoretical physics and merton college , oxford , for their enduring support . jr is indebted to oriel college , oxford and trinity college , cambridge for their support . we can find the generators for a representative of all the conjugacy classes of subgroups of interest using @xcite . first , we use the permutation data @xmath83 , @xmath84 obtained from each of the schreier coset graphs ( in turn obtained from each of the dessins ) to find the group homomorphism by images between the relevant hecke group and a representative of the conjugacy class of subgroups of interest . we then use this to define the representative in question . finally , we use the command ` isomorphismfpgroup(g ) ` , which returns an isomorphism from the given representative to a finitely presented group isomorphic to that representative . this function first chooses a set of generators of the representative and then computes a presentation in terms of these generators . to give an example , consider the clean dessin for the octahedron . we can find a set of generators as @xmath24 matrices for a representative of the associated class of subgroups by implementing the following code in : ` gap > f:=freegroup(x , y ) ; ` + `` + ` gap > h4:=f/[f.1 ^ 2,f.2 ^ 4 ] ; ` + `` + ` gap > hom:=grouphomomorphismbyimages(h4,group ( ` + ` ( 1,5)(6,11)(4,12)(2,13)(3,23)(8,14)(7,18)(10,19)(9,22)(16,24)(15,17)(20,21 ) , ` + ` ( 1,2,3,4)(5,6,7,8)(9,10,11,12)(13,14,15,16)(17,18,19,20)(21,22,23,24 ) ) , ` + ` generatorsofgroup(h4 ) , ` + ` ( 1,5)(6,11)(4,12)(2,13)(3,23)(8,14)(7,18)(10,19)(9,22)(16,24)(15,17)(20,21 ) , ` + ` ( 1,2,3,4)(5,6,7,8)(9,10,11,12)(13,14,15,16)(17,18,19,20)(21,22,23,24 ) ) ] ) ; ` + ` x , y ] - > ` + ` ( 1,5)(2,13)(3,23)(4,12)(6,11)(7,18)(8,14)(9,22)(10,19)(15,17)(16,24)(20,21 ) , ` + ` ( 1,2,3,4)(5,6,7,8)(9,10,11,12)(13,14,15,16)(17,18,19,20)(21,22,23,24 ) ] ` + ` gap > octahedron_group:=preimage(hom , stabilizer(image(hom),1 ) ) ; ` + ` group(<fp , no generators known > ) ` + ` gap > iso:=isomorphismfpgroup(octahedron_group ) ; ` + ` < [ [ 1 , 1 ] ] |y*x*y*x^-1*y*x^-1 > , ` + ` , ` + ` , ` + ` , ` + ` , ` + ` , ` + `` + ` - > [ f1 , f2 , f3 , f4 , f5 , f6 , f7 ] ` we define the relevant hecke group ( here @xmath133 ) in the third line . then , the only input in each case are the permutation data @xmath83 , @xmath84 . once the output has been obtained , we see the generators ( here seven : @xmath174 $ ] ) in the final line , as functions of the @xmath30 and @xmath31 for the hecke group in question . now , returning to the explicit matrices for these @xmath30 and @xmath31 ( presented for @xmath175 , @xmath133 and @xmath134 in section 2 of this paper ) , the only thing left to do is to multiply together the matrices and their inverses as indicated . this will produce the generators for each representative , as desired . yang - hui he , john mckay and james read , `` modular subgroups , dessins denfants and elliptic k3 surfaces '' , lms journal of computation and mathematics 16:271318 , 2013 . fritz beukers and hans montanus , `` explicit calculation of elliptic fibrations of k3-surfaces and their belyi - maps , '' number theory and polynomials , 3351 , lms lecture note series , 352 , cup , 2008 . y. -h . he and j. mckay , `` @xmath176 gauge theories : congruence subgroups , coset graphs and modular surfaces , '' j. math . phys . * 54 * , 012301 ( 2013 ) [ arxiv:1201.3633 [ hep - th ] ] . y. -h . he and j. mckay , `` eta products , bps states and k3 surfaces , '' arxiv:1308.5233 [ hep - th ] . alexandra fritz , `` platonic stars,construction of algebraic curves and surfaces with prescribed symmetries and singularities '' diplom , innsbruck , + ` http://homepage.univie.ac.at/herwig.hauser/master_diss/diplom_fritz.pdf ` .

we show that every clean _ dessin denfant _ can be associated with a conjugacy class of subgroups of a certain hecke group , or free product of hecke groups . this dessin is naturally viewed as the schreier coset graph for that class of subgroups , while the belyi map associated to the dessin is the map between the associated algebraic curve and @xmath0 . with these points in mind , we consider the well - studied archimedean solids , finding a representative the associated class of hecke subgroups in each case by specifying a generating set for that representative , before finally discussing the congruence properties of these subgroups . ^1^__department of mathematics , city university , london , _ _ + _ northampton square , london ec1v 0hb , uk ; _ + _ school of physics , nankai university , tianjin , 300071 , p.r . china ; _ + _ merton college , university of oxford , ox1 4jd , uk _ + _hey@maths.ox.ac.uk_ ^2^__oriel college , university of oxford , ox1 4ew , uk ; _ _ + _ trinity college , university of cambridge , cb2 1tq , uk _ + _jamr2@cam.ac.uk_ + +

introduction dessins denfants and hecke groups archimedean solids conclusions acknowledgements algorithm for computing the generators for the heck subgroups

the concept of universality makes possible to understand and classify the various and complicate critical behavior of a number of equilibrium models @xcite . for nonequilibrium critical phenomena , extensive studies during past decades revealed that various nonequilibrium systems also exhibit universal behavior characterized by several features @xcite . hence , it has been an important issue to identify nonequilibrim universality classes by finding common physical features . among nonequilibrium critical phenomena , absorbing phase transitions ( apt s ) from fluctuating active states into absorbing states in which the system is trapped forever have been a field of growing interest during last decades @xcite . recent theoretical and numerical studies show that apt s exhibit universality and it can be classified according to conservation laws , dimensionality of systems and symmetries of absorbing states @xcite . however only a few universality classes have been identified so far . directed percolation ( dp ) @xcite and parity conserving ( pc ) @xcite class are well studied classes among others . dp class includes systems with no special attributes except the time reversal symmetry , so that most systems studied so far belong to this class . as a research direction to search for further unknown universality classes , coupled systems have been studied recently @xcite . a coupled system is a multi - species system in which each species is coupled to the others in certain ways such as bidirectional and unidirectional coupling in linear or quadratic ways . however the coupled systems do not always exhibit new critical behavior . for bidirectionally coupled systems , the critical behavior depends on the manner of the coupling . for instance , quadratically coupled dp systems still belong to dp class despite their complex behavior @xcite . however , linearly coupled systems belonging to dp or pc class exhibit mean - field or non - trivial critical behavior @xcite . linearly and unidirectionally coupled systems exhibit new critical behavior at multicritical point where all sub - systems are critical @xcite . among single species systems , pair contact process with diffusion ( pcpd ) can be regarded as a two species system . pcpd has been extensively studied during last years due to its nontrivial critical behavior ( see @xcite for review ) . however , in spite of extensive theoretical and numerical studies , the critical behavior is not clearly uncovered yet . pcpd is a stochastic reaction - diffusion model , which evolves by the competition of two processes , fission ( @xmath12 ) and annihilation ( @xmath13 ) . in addition , each particle performs isotropic diffusion . without diffusion , the model is so - called pair contact process ( pcp ) belonging to dp class @xcite . since the reactions involve pairs , diffusing solitary particles are not engaged in the binary reactions . however when two solitary particles form a pair , the reactions take place . on the other hand , solitary particles are created from pairs by diffusion . hence , pcpd can be regarded as a bidirectionally coupled two species system in which the order of the coupling is linear in the direction from pairs to solitary particles and quadratic in the opposite direction . this observation leads to the cyclically coupled dp and pair annihilation which exhibits the similar type of critical behavior to that of pcpd @xcite . in this paper , we investigate the domain structure of pcpd . when a spreading domain is formed from localized initial activities , the quadratic coupling from solitary particles to pairs allows the pair - free region in which only solitary particles are present . we call the pair - free region so - called _ uncoupled _ region . on the other hand , the linear coupling from pairs to solitary particles results in the so - called _ coupled _ region in which pairs and solitary particles coexist . the pair - free region encloses the coupled region as shown in fig . [ pattern ] . hence , the spreading domain is divided into two regions , coupled and uncoupled region . this kind of double domain structure was found in unidirectionally coupled two level hierarchies @xcite . as shown in previous studies on unidirectionally coupled systems @xcite , the measurements of critical exponents are very difficult due to long - time drift of the exponents . the one reason of the drift is the generic feature of the unidirectional coupling @xcite , the other is the double domain structure @xcite . in measuring critical exponents , one can overcome the latter effect by measuring quantities in each region separately . ( color online ) the double domain structure of pcpd . the region with size @xmath2 is the coupled region in which pairs ( blue ) and solitary particles ( red ) coexist . the region with size @xmath3 is the uncoupled region in which only solitary particles exist . ] to explain the effect of the domain structure on the critical behavior , let us consider the unidirectionally coupled dp - pc coupling @xcite . in the dp - pc coupling , pc process ( @xmath14 ) is unidirectionally coupled with dp process ( @xmath15 ) via @xmath16 ( linear ) . since dp process spreads faster than pc process , the coupling changes the spreading behavior of pc process . there are no interactions between @xmath15 and @xmath14 species . starting with a single @xmath15 particle on @xmath15-level , a @xmath14-domain is form and spreads faster than a @xmath15-domain does due to the coupling . as a result , the uncoupled region where no @xmath15 particles exist is formed @xcite . the coupled region is just the region of @xmath14-domain overlapped with the @xmath15-domain where the coupling exists . hence , the size of the @xmath14-domain ( @xmath17 ) is the sum of the size of coupled ( @xmath18 ) and uncoupled region ( @xmath3 ) , @xmath19 . at the multicritical point where both species are critical , the size of each domain increases in power - law such as @xmath20 , @xmath21 , @xmath22 and @xmath23 . since @xmath24 by definition , the scaling of @xmath17 is given as @xmath25 , @xmath26 and @xmath27 are constants . a recent study showed @xmath28 for the fast spreading @xmath15-domain ( source ) @xcite . hence , @xmath17 scales as @xmath29 asymptotically for faster spreading source @xmath15 , and @xmath3 plays the correction to the scaling of @xmath17 . since the difference @xmath30 is sufficiently small in the dp - pc coupling , the existence of the uncoupled region makes precise estimates of @xmath31 difficult within moderate simulation time . for the dp - pc coupling , the estimate of @xmath32 is @xmath33 which is less than the expectation @xmath34 @xcite . on the other hand , the estimate of @xmath35 is @xmath36 which is compatible with the leading scaling . as in unidirectionally coupled systems , it is expected that the double domain structure of pcpd also makes it difficult to identify the critical spreading behavior precisely . the aim of this paper is to investigate the effect of the double domain structure on the critical spreading of pcpd and to estimate critical exponents more precisely . we consider the pcpd of ref . @xcite . in this model , a randomly selected particle attempts to hop to one of the nearest neighbor sites with an equal probability . if the target site is empty , the attempt is accepted . however if the target site is occupied , ( i ) two particles annihilate with probability @xmath37 or ( ii ) the hopping attempt is rejected and the pair ( the chosen particle and one at the target site ) tries to create a particle at the randomly chosen nearest neighbor site of the pair . when the target site is occupied , the branching attempt is rejected . the critical point of this model is @xmath38 in one dimension @xcite . pcpd has three sectors in configuration spaces according to the existence of pairs(@xmath39 ) and solitary particles(@xmath40 ) . the one is the configurations in which both pairs and solitary particles are present ( @xmath41-ensemble ) . we call configurations with at least one pair ( two solitary particles ) @xmath39-ensemble ( @xmath40-ensemble ) . in @xmath39-ensemble ( @xmath40-ensemble ) , solitary particles ( pairs ) may be present or not . the @xmath39-ensemble is the reactive subspace of ref . a conventional ensemble includes configurations with at least two particles which can be either two solitary particles or one pair . we call the conventional ensemble @xmath42-ensemble . since solitary particles are effectively linearly coupled to pairs , the existence of pairs implies the existence of solitary particles . hence , @xmath41-ensemble should coincide with @xmath39-ensemble asymptotically . however , solitary particles transform into pairs by collisions , the coupling in this direction is quadratic . hence , the existence of solitary particles does not always guarantee the presence of pairs due to the long life time of solitary particles . so @xmath40-ensemble coincide with @xmath42-ensemble . as a result , there are two distinct ensembles in pcpd . we define the size of each domain as follows ( see fig . [ pattern ] ) . the size of the whole domain at time @xmath49 ( @xmath50 ) is defined as the distance between the leftmost and the rightmost particle . when both pairs and solitary particles exist simultaneously , we can define the size of the coupled region ( @xmath51 ) and the size of the uncoupled region ( @xmath52 ) . since solitary particles are linearly coupled to pairs , we define the size of the coupled region ( @xmath53 ) as the spreading distance of pairs ( @xmath2 ) defined as the distance between the leftmost and the rightmost pair . then , @xmath52 is given as @xmath54 . hence , we have three different lengths , @xmath4 , @xmath2 and @xmath3 in pcpd . to take into account the three length scales at the same time , one should use @xmath41-ensemble in which only two lengths , @xmath2 and @xmath3 are the fundamental length scales of pcpd due to @xmath5 . at @xmath59 , we perform defect monte carlo simulations with a pair on one dimensional empty lattice . we run simulations up to @xmath60 time steps using @xmath61 independent runs . we measure the squared sizes , @xmath43 , @xmath62 and @xmath63 for surviving @xmath41-ensemble . the pcpd has two absorbing states , vacuum and states with one diffusing solitary particle . hence , we stop the simulations when the total number of particles @xmath64 is less then two . at the criticality , the squared sizes scale as @xmath65 , @xmath66 and @xmath67 . fig . [ range ] shows the scaling plots of the squared sizes , @xmath68 . we obtain the best scaling plot with @xmath69 , @xmath70 and @xmath71 respectively . the errors of our estimates should be larger due to the error of @xmath72 . within the numerical errors at the criticality , our estimate of @xmath73 agree with the previous studies @xcite , especially @xmath74 @xcite . since @xmath8 , the total spreading distance @xmath75 should scale as @xmath76 , and @xmath3 plays the role of the correction to the scaling as in unidirectionally coupled systems . hence , we conclude @xmath77 which is the smallest value among the estimates of previous studies @xcite . for reference , we also measure @xmath43 using @xmath42-ensemble . the difference of @xmath42-ensemble from @xmath41-ensemble or @xmath39-ensemble is that @xmath42-ensemble includes configurations without pairs . since solitary particles spread diffusively , it is expected that @xmath4 averaged over @xmath42-ensemble scales differently from that of @xmath41-ensemble . from the scaling plot of @xmath78 , we estimate @xmath79 for @xmath42-ensemble which is larger than that of @xmath41-ensemble ( not shown ) . the slow spreading of the whole domain in @xmath42-ensemble results from the diffusive motions of solitary particles in configuration without pairs . since @xmath4 scales as @xmath80 , the correction to the @xmath4 is @xmath55 which decays as @xmath81 with @xmath82 . to see how the correction decays slowly in time , we plot @xmath83 in fig . [ ratio ] . when @xmath84 , the correction is negligible . however , as shown in the inset , @xmath83 is still comparable to one even at @xmath60 . we obtain the best scaling plot of @xmath56 with @xmath85 . as @xmath83 decays with very small @xmath86 , it is practically impossible to reach the asymptotic scaling regime of @xmath87 . as a result , one should take the double domain structure into account for the more precise measurement of the dynamic exponent @xmath73 . ( color online ) the scaling plot of @xmath92 and @xmath93 . the main plot shows the scaling plot of @xmath94 ( red ) and @xmath95 ( black ) with @xmath96 . we add a constant @xmath97 to the scaling plot of @xmath93 for the better presentation . the inset shows the double logarithmic plots . ] in addition to the sizes , we also measure the number of pairs ( @xmath88 ) and of solitary particles ( @xmath98 ) averaged over all samples , and the survival probability of @xmath41-ensemble ( @xmath92 ) and @xmath42-ensemble ( @xmath93 ) . as solitary particles are linearly coupled to pairs , @xmath98 is proportional to @xmath88 . at criticality , @xmath88 scales as @xmath99 . fig . [ eta ] shows the scaling plot @xmath89 . we obtain the best scaling plot with @xmath100 . the inset show the ratio of @xmath91 , which converges to one as expected . @xmath101 and @xmath102 decay in power - law as @xmath103 and @xmath104 at the criticality . [ p ] shows the scaling plots of @xmath105 and the double logarithmic plots of @xmath92 and @xmath93 . with @xmath106 , we obtain the best scaling plots . the system escapes from the non - reactive sector in which only diffusing solitary particles are present via pair annihilations of solitary particles . as an naive argument for the equality of @xmath107 , the escaping probability from non - reactive sector within time @xmath108 may scale as @xmath109 which is the death probability of two diffusing particles undergoing the reaction @xmath110 within @xmath108 . hence , the contribution of non - reactive sector to @xmath93 is negligible due to the fast escaping probability , which leads to the same scaling behavior of @xmath93 as @xmath111 . in summary , we investigate the domain structure of pcpd . we numerically confirm the existence of the double domain structure in pcpd . this double domain structure comes from the linear and quadratic bidirectional couplings . the structure intrinsically makes the serious correction to the scaling of the critical spreading of a domain . starting with a pair , a domain grows and spreads in vacuum via fission ( @xmath112 ) and spontaneous annihilation of pairs ( @xmath113 ) in pcpd . in addition to the binary reactions , each particle diffuses isotropically , which leads to the bidirectional coupling between solitary particles and pairs . the coupling from pairs to solitary particles is linear , while the opposite coupling is quadratic . the difference of the coupling ways results in the double domain structure of the whole domain , the coupled and the uncoupled region respectively . as a result , the size of the whole domain ( @xmath4 ) is given as the sum of the size of the coupled region ( @xmath2 ) and of the uncoupled region ( @xmath3 ) . hence , @xmath2 and @xmath3 are the basic length scales characterizing the scaling behavior of the spreading domain in pcpd . we numerically find that @xmath2 and @xmath3 scale as @xmath6 and @xmath7 with @xmath114 at criticality . hence , @xmath4 should asymptotically scale as @xmath115 with @xmath116 and @xmath3 plays the role of the correction to the scaling . however , the direct measurement of @xmath4 leads to the underestimate of the asymptotic value of @xmath73 because the correction @xmath11 decays with very small exponent . since it is practically impossible to reach the asymptotic scaling region of @xmath117 , it is important to take the domain structure into account in simulations for more precise estimate of the dynamic exponent @xmath73 of pcpd . we classify particle configurations into four ensembles , which are finally reduced to two distinct ensembles , @xmath39-ensemble and @xmath42-ensemble respectively . @xmath42-ensemble includes configurations without pairs , while @xmath39-ensemble does not . the survival probabilities of two ensembles decay with the same exponent . however , the whole domain appears to spread more slowly in @xmath42-ensemble than in @xmath39-ensemble due to the diffusive motions of solitary particles in configurations without pairs . hence , in addition to the domain structure , the diffusive motions of solitary particles in @xmath42-ensemble raise another correction to the scaling of the total spreading distance @xmath4 which does not appear in @xmath39-ensemble . as the linear - quadratic bidirectional coupling is the common feature of various pcpd studied so far , the double domain structure should appear in other pcpd variants . among pcpd variants , we investigate the domain structure of the bosonic pcpd with soft - constraint of ref . @xcite which is known to exhibit the clear power - law decays at the criticality . for this model , we also confirm the existence of the double domain structure and the critical spreading behavior similar to that of pcpd studied in this paper . hence , the double domain structure is a common feature of pcpd variants . as shown in recent studies , as pcpd extremely slowly approaches to its asymptotic scaling region , it is very difficult to identify the critical behavior . the bidirectional coupling should be the main reason in itself . on the other hand , the double domain structure naturally appears due to the linear - quadratic bidirectional couplings , which enhance the long - time drift of dynamic exponent @xmath73 . one can overcomes the latter effect by considering the scaling behavior of sub - domains separately as in unidirectionally coupled systems . h. hinrichsen , adv . phys . * 49 * , 815 ( 2000 ) . g. dor , rev . phys . * 76 * , 663 ( 2004 ) . p. grassberger and a. de la torre , ann . ( ny ) * 122 * , 373 ( 1979 ) . i. jensen , j. phys . a * 32 * , 5233 ( 1999 ) . i. jensen , phys . e , * 50 * , 3623 ( 1994 ) . s. kwon and h. park , phys . e * 52 * , 5955 ( 1995 ) . j. l. cardy and u. tuber , phys . lett . * 77 * , 4780 ( 1996 ) . j. kochelkoren and h. chat , phys . * 90 * , 125710 ( 2003 ) . p. grassberger , f. krause , and t. von der twer , j. phys . a * 17 * , l105 ( 1984 ) ; p. grassberger , j. phys . a * 22 * , l1103 ( 1989 ) ; m. h. kim and h. park , phys . lett . * 73 * 2579(1994 ) ; h. hinrichsen , phys . e * 55 * , 219 ( 1997 ) . w. hwang , s. kwon , h. park , and h. park , phys . e * 57 * , 6438 ( 1998 ) . o. al hammal , h. chate , i. dornic , and m. a. munoz , phys . lett . * 94 * , 230601 ( 2005 ) . h. k. janssen , phys . lett . * 78 * , 2890 ( 1997 ) . s. kwon , j. lee and h. park , phys . lett . * 85 * , 1682 ( 2000 ) . s. kwon and h. park , j. korean phys . soc . * 38 * , 490 ( 2001 ) ; g. dor , phys . rev . e * 63 * , 021113 ( 2001 ) . ; s. kwon and h. park , phys . e * 69 * , 066125 ( 2004 ) . j. noh and h. park , phys . * 94 * , 145702 ( 2005 ) u. alon , m. r. evans , h. hinrichsen and d. mukamel , phys . lett . * 76 * , 2746 ( 1996 ) . d. richardson , proc . cambridge philos . soc . * 74 * , 515 ( 1973 ) ; n. goldenfeld , j. phys . a * 17 * , 2807 ( 1984 ) ; j. m. kim and j. j. kosterlitz , phys . lett . * 62 * , 2571 ( 1989 ) ; n. rajewsky , d. e. wolf , and j. kertsz , physica a * 164 * , 81 ( 1990 ) ; a. toom , j. stat . phys . * 74 * , 91 ( 1994 ) ;

we investigate the domain structure of pair contact process with diffusion ( pcpd ) . pcpd is a stochastic reaction - diffusion model which evolves by the competition of two binary reactions , @xmath0 and @xmath1 . in addition , each particle diffuses isotropically , which leads to the bidirectional coupling between solitary particles and pairs . the coupling from pairs to solitary particles is linear , while the opposite coupling is quadratic . the spreading domain formed from localized activities in vacuum consists of two regions , the coupled region of size @xmath2 where pairs and solitary particles coexist and the uncoupled region of size @xmath3 where only solitary particles exist respectively . as the size of the whole domain @xmath4 is given as @xmath5 , @xmath2 and @xmath3 are the basic length scales of pcpd . at criticality , @xmath2 and @xmath3 scale as @xmath6 and @xmath7 with @xmath8 . we estimate @xmath9 and @xmath10 . hence , the correction to the scaling of @xmath4 , @xmath11 extremely slowly decays , which makes it practically impossible to identify the asymptotic scaling behavior of @xmath4 . in addition to the generic feature of the bidirectional coupling , the double domain structure is another reason for the extremely slow approach to the asymptotic scaling regime of pcpd .

introduction

this work is supported in part by dfg and nsfc ( crc 110 ) , by the national natural science foundation of china ( grant no . 11475186 ) and by the australian research council through grants dp140103067 , dp150103164 ( dbl ) , fl0992247 and dp151103101 ( awt ) and through the arc centre of excellence for particle physics at the terascale . j. ashman _ et al . _ [ european muon collaboration],phys . lett . b * 206 * , 364 ( 1988 ) . c. a. aidala , s. d. bass , d. hasch andg . k. mallot , rev . phys . * 85 * , 655 ( 2013 ) [ arxiv:1209.2803 [ hep - ph ] ] . f. myhrer and a. w. thomas , j . g * 37 * , 023101 ( 2010 ) [ arxiv:0911.1974 [ hep - ph ] ] . r. l. jaffe , aip conf . * 588 * , 54 ( 2001 ) [ hep - ph/0102281 ] . s. d. bass , rev . * 77 * , 1257 ( 2005 ) [ hep - ph/0411005 ] . m. burkardt , c. a. miller and w. d. nowak , rept . phys . * 73 * , 016201 ( 2010 ) [ arxiv:0812.2208 [ hep - ph ] ] . j. r. ellis and m. karliner , [ hep - ph/9601280 ] . g. m. shore , [ hep - ph/9812355 ] . s. d. bass and a. w. thomas , j. phys . g * 19 * , 925 ( 1993 ) . b. w. filippone and x. d. ji , adv . * 26 * , 1 ( 2001 ) [ hep - ph/0101224 ] . j. ashman _ et al . _ [ european muon collaboration ] , phys . b * 206 * , 364 ( 1988 ) . b. adeva _ et al . _ [ spin muon collaboration ] , phys . d * 58 * , 112002 ( 1998 ) . e. s. ageev _ et al . _ [ compass collaboration ] , phys . b * 612 * , 154 ( 2005 ) [ hep - ex/0501073 ] . v. y. alexakhin _ et al . _ [ compass collaboration ] , phys . b * 647 * , 8 ( 2007 ) [ hep - ex/0609038 ] . a. airapetian _ et al . _ [ hermes collaboration ] , phys . d * 75 * , 012007 ( 2007 ) [ hep - ex/0609039 ] . k. v. dharmawardane _ et al . _ [ clas collaboration ] , phys . b * 641 * , 11 ( 2006 ) [ nucl - ex/0605028 ] . b. i. abelev _ et al . _ [ star collaboration ] , phys . lett . * 97 * , 252001 ( 2006 ) [ hep - ex/0608030 ] . a. adare _ et al . _ [ phenix collaboration ] , phys . d * 76 * , 051106 ( 2007 ) [ arxiv:0704.3599 [ hep - ex ] ] . v. w. hughes , nucl . a * 518 * , 371 ( 1990)and references therein . m. g. alekseev _ et al . _ [ compass collaboration ] , phys . b * 690 * , 466 ( 2010 ) [ arxiv:1001.4654 [ hep - ex ] ] . r. d. carlitz , j. c. collins and a. h. mueller , phys . b * 214 * , 229 ( 1988 ) . g. altarelli and g. g. ross , phys . b * 212 * , 391 ( 1988 ) . s. d. bass , b. l. ioffe , n. n. nikolaev and a. w. thomas , j. moscow . * 1 * , 317 ( 1991 ) . a. v. efremov and o. v. teryaev , phys . b * 150 * , 383 ( 1985 ) . p. chen and x. ji , phys . b * 660 * , 193 ( 2008 ) [ hep - ph/0612174 ] . c. adolph _ et al . _ [ compass collaboration ] , phys . d * 87 * , no . 5 , 052018 ( 2013 ) [ arxiv:1211.6849 [ hep - ex ] ] . e. r. nocera _ et al . _ [ nnpdf collaboration ] , nucl . b * 887 * , 276 ( 2014 ) [ arxiv:1406.5539 [ hep - ph ] ] . f. e. close and r. g. roberts , phys . b * 316 * , 165 ( 1993 ) [ hep - ph/9306289 ] . r. l. jaffe and a. manohar , nucl . b * 337 * , 509 ( 1990 ) . g. ratcliffe , czech . j. phys . * 54 * , b11 ( 2004 ) [ hep - ph/0402063 ] . s. d. bass and a. w. thomas , phys . b * 684 * , 216 ( 2010 ) [ arxiv:0912.1765 [ hep - ph ] ] . a. w. schreiber and a. w. thomas , phys . b * 215 * , 141 ( 1988 ) . a. chodos , r. l. jaffe , k. johnson , c. b. thorn and v. f. weisskopf , phys . d * 9 * , 3471 ( 1974 ) . f. myhrer and a. w. thomas , phys . d * 38 * , 1633 ( 1988 ) . r. d. young , d. b. leinweber , a. w. thomas and s. v. wright , phys . d * 66 * , 094507 ( 2002 ) [ hep - lat/0205017 ] . a. w. thomas , phys . * 101 * , 102003 ( 2008 ) [ arxiv:0803.2775 [ hep - ph ] ] . r. l. jaffe , phys . b * 193 * , 101 ( 1987 ) . d. b. leinweber , phys . d * 69 * , 014005 ( 2004 ) [ arxiv : hep - lat/0211017 ] . p. wang , d. b. leinweber , a. w. thomas and r. d. young , phys . d * 75 * , 073012 ( 2007 ) [ hep - ph/0701082 ] . e. e. jenkins , m. e. luke , a. v. manohar and m. j. savage , phys . b * 302 * , 482 ( 1993 ) [ phys . b * 388 * , 866 ( 1996 ) ] [ hep - ph/9212226 ] . d. b. leinweber _ et al . _ , phys . rev . lett . * 94 * , 212001 ( 2005 ) [ arxiv : hep - lat/0406002 ] . p. wang , d. b. leinweber , a. w. thomas and r. d. young , phys . c * 79 * , 065202 ( 2009 ) [ arxiv:0807.0944 [ hep - ph ] ] . m. gensini , [ hep - ph/9512440 ] . s. a. larin , phys . b * 303 * , 113 , 334 ( 1993 ) [ hep - ph/9302240 ] . p. wang , d. b. leinweber and a. w. thomas , phys . d * 89 * , no . 3 , 033008 ( 2014 ) [ arxiv:1312.3375 [ hep - ph ] ] . j. m. m. hall , d. b. leinweber and r. d. young , phys . d * 89 * , no . 5 , 054511 ( 2014 ) [ arxiv:1312.5781 [ hep - lat ] ] . d. b. leinweber , a. w. thomas and r. d. young , phys . * 92 * , 242002 ( 2004 ) [ hep - lat/0302020 ] . wang and a. w. thomas , phys . d * 81 * , 114015 ( 2010 ) [ arxiv:1003.0957 [ hep - ph ] ] . c. r. allton , w. armour , d. b. leinweber , a. w. thomas and r. d. young , phys . b * 628 * , 125 ( 2005 ) [ hep - lat/0504022 ] . w. armour , c. r. allton , d. b. leinweber , a. w. thomas and r. d. young , nucl . a * 840 * , 97 ( 2010 ) [ arxiv:0810.3432 [ hep - lat ] ] . j. m. m. hall , d. b. leinweber and r. d. young , phys . d * 88 * , no . 1 , 014504 ( 2013 ) [ arxiv:1305.3984 [ hep - lat ] ] .

proton spin is investigated in chiral effective field theory through an examination of the singlet axial charge , @xmath0 , and the two non - singlet axial charges , @xmath1 and @xmath2 . finite - range regularization is considered as it provides an effective model for estimating the role of disconnected sea - quark loop contributions to baryon observables . baryon octet and decuplet intermediate states are included to enrich the spin and flavour structure of the nucleon , redistributing spin under the constraints of chiral symmetry . in this context , the proton spin puzzle is well understood with the calculation describing all three of the axial charges reasonably well . the strange quark contribution to the proton spin is negative with magnitude 0.01 . with appropriate @xmath3 evolution , we find the singlet axial charge at the experimental scale to be @xmath4 , consistent with the range of current experimental values . adp-15 - 48/t950 in 1988 the european muon collaboration ( emc ) published their polarized deep inelastic measurement of the proton s spin dependent structure function @xmath5 . their result suggested that the quark spins summed over the up , down and strange quark flavors contribute only a small fraction of the proton s spin @xcite . the emc data shocked the particle physics community , because it was thought to be contradictory to the apparently successful , naive quark model descriptions of proton structure where the constituent quarks carry the total proton spin . it inspired a vigorous global program of experimental and theoretical developments to understand the internal spin structure of the proton extending for nearly three decades . for reviews of the spin structure of the proton , see for example refs . . the experimental efforts at cern @xcite , desy @xcite , jlab @xcite , rhic @xcite and slac @xcite have been impressive . a summary of the status and recent experimental results on the spin structure of the nucleon can be found in ref . @xcite . unlike the early emc result which suggested that the quark spin contribution , @xmath6 , might be consistent with zero ( @xmath7 @xcite ) , today the experimental measurements indicate the nucleon s flavor - singlet axial charge measured in polarized deep inelastic scattering is @xmath8 at @xmath9 gev@xmath10 . this tends to about one - third of the total spin @xmath11 as @xmath12 @xcite . the matrix elements of the non - singlet axial current @xmath13 and the singlet axial current @xmath14 are defined as follows @xmath15 where @xmath16 are generators of the flavor group and @xmath17 is a vector in flavor space . the singlet axial current is not conserved due to the adler - bell - jackiw anomaly . as a result , the flavor - singlet matrix element can receive an additional contribution from gluon polarization @xcite . this led to the early idea that the measured singlet component @xmath0 receives an important contribution from the gluon polarization @xmath18 , _ i.e. _ @xmath19 the polarized gluon distribution function @xmath18 was estimated to be less than 0.3 at a scale of 1 gev@xmath20 in the mit bag model @xcite . from the extensive experimental studies one finds that the absolute value is of the order @xmath21 0.2 - 0.3 for @xmath3 = 3 gev@xmath20 @xcite . this amount of gluon polarization , by itself , is far too small to resolve the problem of the small value of @xmath6 through the axial anomaly . another explanation for the small value of @xmath6 draws on the strange quark contribution to the proton spin . the non - singlet axial charge @xmath2 extracted from hyperon beta - decays under the assumption of @xmath22 flavour symmetry is @xmath23 @xcite . if the strange quark contribution to the proton spin were around @xmath24 , the proton spin , @xmath6 , expressed as @xmath25 @xmath26 , would be close to the experimental data . however , the uncertainty of @xmath2 could be as large as @xmath27 @xcite . a recent re - evaluation of the nucleon s axial - charges in the cloudy bag model , taking into account the effect of the one - gluon - exchange hyperfine interaction and the meson cloud , led to the value @xmath28 @xcite . in this case , @xmath29 was found to be of order 0.01 in magnitude ( and negative ) , with the small value of @xmath2 a consequence of su(3 ) breaking . soon after the release of the emc data it was realized that the effect of the pion cloud of the nucleon , associated with chiral symmetry breaking , would be to lower the quark spin content of the nucleon @xcite . this is because pion emission tends to flip the nucleon spin and hence the spins of the quarks in it , while the quarks in the pion necessarily carry orbital angular momentum but no spin . this effect was calculated in the cloudy bag model and the effect of the pion cloud together with the relativistic motion of the light quarks in the bag @xcite reduced @xmath6 to around 0.5 . an alternative approach to the problem recognised that , given the standard spin dependent one - gluon - exchange correction to the energy of the nucleon , there must be a corresponding exchange current correction to the proton spin @xcite . this too reduces the proton spin by around 0.15 below the naive bag model result of 0.65 . it is only recently that studies of the @xmath30 nucleon mass splitting in lattice qcd @xcite provided the justification for combining the pion cloud and one - gluon exchange effects @xcite . this led to a theoretical result in the range @xmath31 to @xmath32 , which is compatible with the aforementioned experimental value and , after qcd evolution , with the results of lattice qcd for the angular momentum carried by the quarks in the proton @xcite . the scale dependence of @xmath0 presents another consideration in understanding the fraction of the proton spin carried by quarks @xcite . consideration of the genral features of qcd evolution long ago led to the conclusion that the natural scale at which to match a quark model to qcd is quite low , so that most of the momentum of the proton is carried by valence quarks and one can think of the gluons as having been integrated out of the theory . in jaffe s scenario , the small value of the experimental proton spin is due to differences in the energy scale of the experimental result and the quark model results . since the anomalous dimension of the singlet axial current is nontrivial , its matrix element @xmath0 is scale dependent . with the @xmath3 evolution , it is possible that the large proton spin at low @xmath3 will be reduced through @xmath3 evolution to the large @xmath3 of the experimental result . as mentioned in ref . @xcite , it is difficult to get a reliable evolution at low-@xmath3 because perturbative qcd is not applicable . more specifically , one can not determine which evolution line presented in ref . @xcite is correct . one needs a direct calculation of @xmath6 at the low energy scale . in this paper , we will investigate the proton spin carried by the quarks in the framework of effective field theory , assuming that at the corresponding low scale the gluons have been integrated out , with the only residue being a spin - dependent effective interaction between quarks . @xmath6 , and the non - singlet axial charges @xmath1 and @xmath2 will be calculated simultaneously in the chiral effective field theory . in this approach the proton structure is enhanced through the dressing of the proton by octet - meson and both octet and decuplet baryon intermediate states . these processes enrich the spin and flavour structure of the nucleon , redistributing spin under the constraints of chiral symmetry . as we will see , this formalism is able to describe all three axial charges in a reasonable manner . we consider heavy baryon chiral perturbation theory and include octet and decuplet intermediate - state baryons . the lowest - order chiral lagrangian used in the calculation of the nucleon spin distribution function is expressed as @xmath33 -i\overline{t}_{v}^{\mu } \left ( { v\cdot { \cal d}}\right ) { t_{v\mu } } \nonumber \\ & & + { \cal c}\left ( { \overline{t}_{v}^{\mu } { a_{\mu } } { b_{v}}+\overline{{b}}_{v}{a_{\mu } } t_{v}^{\mu } } \right ) \ , , \end{aligned}\ ] ] where @xmath34 is the covariant spin operator defined as @xmath35here , @xmath36 is the nucleon four velocity . in the rest frame , we have @xmath37 ) . @xmath38 , @xmath39 and @xmath40 are the standard @xmath22-flavour coupling constants . according to the lagrangian , the one - loop feynman diagrams , which contribute to the quark spin fraction of the proton , are plotted in fig . 1 . working with the chiral coefficients of full qcd @xcite , the contribution of the doubly - represented @xmath41-quark sector of the proton to the proton spin , described by diagram ( a ) of fig . 1 , is expressed as @xmath42s_{u } \ , , \end{aligned}\ ] ] where the first through fourth terms in the bracket are the contributions from the @xmath43 , @xmath44 , the @xmath45 transition and the @xmath46 intermediate states , respectively . the @xmath41-quark contribution with a @xmath47 intermediate state vanishes . the coefficients , @xmath48 , of the integrals , @xmath49 , are expressed as @xmath50 these coefficients reflect the @xmath22-flavour symmetry considered in obtaining the meson - baryon couplings ( proportional to @xmath39 and @xmath38 ) , the angular - momentum composition of the intermediate meson - baryon intermediate states , and the @xmath51-spin - flavour wave function of the intermediate state baryon , considered in assigning a quark - sector spin contribution . the latter is discussed in further detail below . with the above coefficients , one can write the @xmath52 quark - sector contribution to the proton spin of fig . 1a as @xmath53 s_{d}.\end{aligned}\ ] ] similarly , the strange quark contribution to the proton spin from diagram ( a ) of fig . 1 is written as @xmath54 s_s \ , , \ ] ] where @xmath55 in the above equations , the low energy coefficients @xmath56 ( @xmath57 ) describe the tree - level quark contribution to the baryon spin . for example , for the intermediate proton and neutron , their spins are expressed as @xmath58 in the naive quark model , the value of @xmath56 is 1 . however , it is smaller than 1 due to the relativistic and confinement effects @xcite . diagram ( b ) of fig . 1 illustrates decuplet baryon intermediate states . the @xmath41-sector contribution to the proton spin from this diagram is @xmath59\ , s_u \ , , \ ] ] where the coefficients @xmath60 and @xmath61 are @xmath62 the @xmath52 and @xmath63 quark - sector contributions are @xmath64\ , s_d,\ ] ] and @xmath65 in deriving these equations , the tree - level quark contributions to the spin of decuplet baryons are used . for example @xmath66 these contributions will also be reduced upon taking relativistic and confinement effects into account . diagrams ( c ) and ( d ) of fig . 1 provide contributions from intermediate states involving an octet - decuplet transition . the @xmath41 quark - sector contribution to the proton spin from these diagrams is expressed as @xmath67 \nonumber \\ & & \times s_u \ , , \end{aligned}\ ] ] where @xmath68 the @xmath52 and @xmath63 quark - sector contributions are @xmath69\ , s_{d}\ , , \\ % \delta s^{c+d}&=&-\frac{6}{5}\ , c_{\sigma \sigma ^{\ast } k}\ , i_{5k}^{n\sigma \sigma ^{\ast } } \ , s_{s}\ , .\end{aligned}\ ] ] the integrals in the above equations , @xmath70 , @xmath71 and @xmath72 are defined in ref . @xcite . including the tree - level contribution , the total @xmath41- , @xmath52- and @xmath63-quark sector contributions to the spin of the proton are @xmath73 here @xmath74 is the wave - function renormalization constant calculated from the standard diagrams corresponding to those of fig . 1 . values are listed in table i. in the numerical calculations , the @xmath22-flavour couplings are @xmath75 , @xmath76 . the decuplet coupling @xmath77 @xcite . the regulator in the integrals is chosen to be of a dipole form @xmath78 with @xmath79 gev . this prescription is known to model the contributions of disconnected sea - quark loop contributions well @xcite . the final quark spin contributions are related to the low - energy coefficients @xmath80 , @xmath81 and @xmath82 . these are the tree - level values of the quark spin and are unity in the naive constituent - quark model . relativistic and confinement effects associated with light quarks suppress this value . we begin by assuming @xmath83 and treat @xmath56 as a parameter constrained by the axial charge @xmath84 . with @xmath85 gev , the central value of @xmath56 that we find is @xmath86 , less than one as expected . since the strange quark is expected to be less relativistic , this value may be an overestimate of the spin suppression in that case . however , the strange quark contribution to the proton spin is small and so the approximation is adequate for this purpose . with @xmath86 , the @xmath41 , @xmath52 and @xmath63 quark contributions to the proton spin are @xmath87 the axial charge @xmath88 and @xmath89 . before considering the necessary @xmath3 evolution to the value @xmath9 gev@xmath10 relevant to the experimental data , it is interesting to consider other improvements to our use of @xmath51-spin - flavour wave functions in attributing quark spin to intermediate meson - baryon states . although it lies outside the framework of chiral effective field theory , the effect of one - gluon - exchange ( oge ) is particularly important for spin dependent quantities . hogaason and myhrer @xcite showed that the incorporation of the exchange current correction arising from the effective one - gluon - exchange ( oge ) force shifts the tree - level non - singlet charge , @xmath1 , from @xmath90 to @xmath91 , where @xmath92 is about 0.05 . thus , if one were to include the oge correction , @xmath56 would be somewhat larger at @xmath93 if one chose it to reproduce the axial charge @xmath94 . for the charges , @xmath0 and @xmath2 , the oge correction shifts their tree - level values from @xmath56 to @xmath95 @xcite . in this case , @xmath96 and @xmath97 . correspondingly , the quark contributions to the proton spin are @xmath98 the results show that the strange quark contribution to the proton spin is very small relative to the @xmath41 and @xmath52 contributions . the axial charge , @xmath99 , is intermediate between the value extracted under the assumption of su(3 ) symmetry from hyperon @xmath100-decay , @xmath101 @xcite , and that obtained in the cloudy bag model , @xmath102 @xcite . to provide an estimate of the uncertainty in these results , we vary the regulator parameter , @xmath47 , governing the size of meson cloud contributions to proton structure . considering @xmath103 gev , the uncertainties in the quark contributions to proton spin are @xmath104 the axial charges with the corresponding error bars are @xmath105 the non - singlet axial current is conserved in the limit of massless quarks and the anomalous dimension for the non - singlet axial current vanishes . therefore , the non - singlet matrix elements @xmath1 and @xmath2 are scale independent . however , the anomalous dimension of the singlet axial current is nontrivial , and @xmath0 is a scale dependent quantity . consistent with the idea that at a sufficiently low scale the valence quarks dominate and the gluons have been effectively integrated out of the theory , we set @xmath106 at that scale . then to compare the result for @xmath0 calculated within chiral effective field theory to experiment , @xmath107 is obtained through nnlo qcd evolution to @xmath9 gev@xmath10 . evolution of the singlet axial charge of eq . ( [ eq : a0a8final ] ) , @xmath108 for the proton . the upper , middle and lower lines are for the values @xmath109 , 0.8 and 1 gev , respectively and provide insight into the role of the meson cloud and the sensitivity of our result at @xmath85 to variations in the size of the meson - cloud dressing of the proton.,width=321 ] the @xmath3 evolution equation has the form @xcite @xmath110 where @xmath111 . after integrating in @xmath112 from a normalization scale of @xmath113 to @xmath3 , one obtains @xcite @xmath114 \ , , \label{evolution}\end{aligned}\ ] ] with the nnlo calculation of the anomalous dimension , @xmath115 , taken from ref . @xcite . in fig . 2 we illustrate the @xmath3 evolution of @xmath116 commencing with our result of eq . ( [ eq : a0a8final ] ) attributed to the scale @xmath117 gev@xmath20 as in ref . @xcite . initially , @xmath116 decreases rapidly with increasing @xmath3 raising concerns about the application of an nnlo calculation for @xmath118 . however , in the context of the model uncertainty presented in fig . 2 the present @xmath3 evolution will suffice . at @xmath119 gev@xmath20 , our calculation of the proton spin can be compared with experiment . our model provides @xmath120 which agrees with the experimental measurement of @xmath121 at @xmath9 gev@xmath10 . [ cols="^,^,^,^,^,^,^,^,^,^",options="header " , ] in summary , we have examined the proton spin fractions carried by quarks using a model in which the meson - cloud dressings of the proton are characterized by chiral effective field theory , regularized through a regulator characterizing the nontrivial size of the source of the meson cloud . finite - range regularization provides an effective model for estimating the role of disconnected sea - quark loop contributions to baryon observables @xcite . both baryon octet and decuplet intermediate states are included to enrich the spin and flavour structure of the nucleon , redistributing spin under the constraints of chiral symmetry . drawing on extensive experience , the preferred regulator parameter is @xmath85 gev . to gain insight into the role of the meson cloud and uncertainties associated in determining the size of the meson - cloud contributions , we have varied @xmath47 from 0.6 gev to 1 gev . the coefficient @xmath56 , which takes relativistic and confinement effects into account is constrained by the experimental axial charge @xmath122 . the one - gluon - exchange correction to the axial charges is also taken into consideration . because each quark - sector contribution is calculated separately , the non - singlet charges , @xmath1 and @xmath2 , and the singlet charge @xmath0 are obtained simultaneously . the results are summarized in table i. our model provides significant insight into the the proton spin puzzle . the main conclusions are : 1 . at low energy scales the total quark spin contribution to the proton spin , @xmath123 , is only of order one half in the valence region . 2 . as indicated in table [ tab1 ] , all three of the quark spin contributions @xmath124 , @xmath125 and @xmath29 decrease in value as one increases the size of the meson - cloud contribution by increasing @xmath47 . as a result the net spin carried by the quarks , @xmath6 , diminishes with increasing meson - cloud contributions . this is in accord with the increased role of orbital angular momentum @xcite between the odd - parity mesons and the even - parity baryons of the proton s meson cloud considered herein . 3 . the parameter @xmath56 reflecting the role of relativistic and confinement effects and constrained by @xmath1 is around 0.82 , smaller than 1 as expected but larger than the typical `` ultra - relativistic '' value of 0.65 . again , increasing the size of the meson - cloud contributions diminishes this value . for example , at @xmath126 gev , @xmath127 . 4 . the non - singlet charge @xmath128 lies between the value extracted from the hyperon @xmath100-decays under the assumption of su(3 ) symmetry , @xmath129 , and the value @xmath102 obtained in the cloudy bag model @xcite . because the experimental value of @xmath0 extracted from dis data depends on this quantity , further work to pin down the extent of su(3 ) breaking would be valuable . 5 . the strange quark contribution to the proton spin is negative and its absolute value is of the order 0.01 . larger @xmath47 values admit stronger hyperon contributions which act to increase this magnitude . 6 . the experimental value of the @xmath0 at 3 gev@xmath20 is reproduced through a combination of the chiral correction and @xmath3 evolution of @xmath6 from a scale of 0.5 gev@xmath20 @xcite . we find @xmath130 ( 3 gev@xmath20 ) is @xmath131 which agrees with the experimental measurement of @xmath132 . future work should explore the role of higher order terms in the @xmath3 evolution of @xmath0 and explore non - perturbative treatments that can provide further insight into the connection between models of hadron structure and modern experimental results .

acknowledgments

in financial markets , it seems very natural to believe that large price changes are caused by large transaction volumes @xcite . farmer et al . has , however , proposed an entirely different story of large price changes in the markets driven by continuous double auctions @xcite . they has argued that large returns are not caused by large orders , while the large gaps between the occupied price levels in the orderbook lead to large price changes in each transaction , and actually showed that the gap distribution closely matches the return distribution , based on the analysis of the orderbook as well as the transaction records on the london stock exchange . they have also shown that the virtual market orders of a constant size reproduces the actual distribution of returns . then , we arrive at the next question , that is , what really causes large gaps . maslov has introduced and studied a simple model of markets driven by continuous double auctions @xcite . in his model , traders choose one from the two types of orders at random . one is a market order which is a order to sell or buy a fixed amount of shares immediately at the best available price on the market on the time . the other is a limit order which is a order to buy or sell a fixed amount of shares with the specification of the limit price ( the worst allowable price ) . the relative price of a new limit order to the most recent market price is a stochastic variable drawn from a uniform distribution in a given interval . numerical simulations of the model show that it has such realistic features to a certain extent as the power - law tail of the distribution of returns , and the long range correlation of the volatility . however the price evolution has some essentially different statistical properties from the actual one on the market of large investors . first the exponent @xmath0 of the power - law tail of the distribution of returns is inside the levy stable region ( @xmath1 ) , while the actual value is close to 3 @xcite . second the hurst exponent h = 1/4 is unrealistic in wide enough time windows . in actual market , we have h = 1/2 for long - term , which is the value for free diffusion . challet and stinchcombe have proposed a model with non - constant occurrence rates of various types of orders as a model in the same class as maslov s @xcite . the price of their model is over diffusive ( @xmath2 ) for short - term and diffusive ( h=1/2 ) for long - term . in this paper , we propose a stochastic model with a novel feature in the same class as the above models . we take into account the mimetic behavior of traders , when they place limit order , then our virtual markets shows the behaviors which are widely recognized as the stylized facts in financial markets , that is , the power - law tail of the distribution of returns with the exponent outside the levy stable region @xcite , the short memory of returns and the long memory of volatilities @xcite . the hurst exponent of our model is asymptotically 1/2 . an explanation is also given for the profile of the autocorrelation function , which is responsible for the value of the hurst exponent we give the definition of our model here . we introduce the cancellation of orders to maintain the number of orders stored in orderbook as in the model of smith et al @xcite . and in other models @xcite . we have , therefore , three types of orders in total on both sides of trade , namely sell / buy market order , sell / buy limit order and the cancellation of sell / buy limit order . the conservation law indicates @xmath3 for i = sell or buy , where the parameters @xmath4 , @xmath5 and @xmath6 denote the occurrence rate of limit order , market order and cancellation respectively . for simplicity , we also assume that traders always place a fixed size of order . the prices at which new limit orders are placed reflect traders strategies . in maslov s paper @xcite , they are determined by offsetting the most recent market price by a random amount drawn from a uniform distribution in a given interval . challet and stinchcombe assume the gaussian distribution from which random variables for the relative prices of sell / buy limit order to ask / bid are drawn , and put the exclusion hypothesis that if there is already a order at the price , the price is not available @xcite . in the model of smith et al . @xcite , traders place the sell(buy ) limit order at any price in the semi - infinite range above bid ( below ask ) with uniform probability . in this paper , we take a little bit more strategic order placement than previous works . we assume the mimetic behavior of traders , when they place limit orders . traders in our model sometimes ( or a part of traders always ) leave the decision of the limit price to the others . they believe the majority , and should be so patient that they can wait in line the execution of their orders for long time . the assumption is implemented here as follows : with probability @xmath7 , a limit price of sell(buy ) order is chosen with the probability proportional to the size of orders stored on the price , and with probability 1-@xmath7 , a price is randomly chosen between [ bid+1 , ask+1 ] ( [ bid-1 , ask-1 ] ) with uniform probability . parameter @xmath7 is crucial for the model . the choice of the limit price in our model follows the preferential attachment dynamics in the growth model of scale free networks @xcite . `` rich gets richer '' is a concept common to both models . however , our model is not a growth model , but the total amount of limit orders is loosely fixed , because the number of coming limit orders are balanced with market orders and cancellations . instead , the result of simulations will show that the distribution of the fluctuation of the gaps between occupied price levels has a power - law tail owing to the unequal attractive powers of each of the prices . the models of continuous double auctions such as maslov s and ours have a difficulty to solve analytically due to the existence of free boundaries : sell ( buy ) limit prices acceptable for traders who place the orders are bounded on the best price of the opposite side , namely the bid ( ask ) price on the time . slanina @xcite and smith et al . @xcite have formulated a mean field approximation of the master equations to have good results . we adopt , however , a numerical method , and leave the analytical examination for the coming papers . we generate the each types of orders in the ratio of @xmath8 ( limit order ) , @xmath9 ( market order ) and @xmath10 ( cancellation ) for @xmath11 either equal to sell or buy . for the several values of @xmath7 , we perform 1,000 times runs of 10,000 step iterations with different initial conditions . we place a unit of shares at each price 1 , 2 , -1 , -2 as the ask , the second best sell limit price , the bid and the second best buy limit price respectively , and also 198 units of sell ( buy ) limit orders at random in the range between 1 ( -1 ) and the random integer drawn from a uniform distribution in the interval [ 2,201 ] ( [ -201,-2 ] ) as a initial condition . here we present the most interesting results obtained through the numerical simulations . 1 shows the cumulative distribution functions of price shifts , the gaps between ask and the second best sell limit price and spreads . price shifts and the gaps are sampled after every buy market order . the results for sell market order are omitted here in order to avoid a redundancy , because we take the symmetric values of the parameters and the initial conditions . spreads are sampled after every sell and buy market orders . all the three distributions become broader when the parameter @xmath7 becomes larger . the power law tails appear in all the graphs when @xmath7 beyond 0.4 . the distributions for the parameter @xmath7 beyond 0.5 are very broad , but steep falls are observed in the tail . we pick up some points from the interval @xmath12 , and roughly estimate that the power law exponent of the tails have minima in the interval @xmath13 , and the values are near 3 at @xmath7=0.475 as given in the caption of fig . 1 . we use the hill estimator of the largest @xmath14 data , where n is the size of sample . we see from fig . 2 that the relative limit price , namely the distance at which new limit orders are placed away from the current best price , is broadly distributed when @xmath7 becomes large . we want to demonstrate , however , that the broadness itself of the distribution of the relative limit price does not create the fat tail of the price shift distribution . first of all , for the purpose , we collect the data of the limit order by a numerical simulation of the model . then we shuffle the order of arrivals of limit orders , and perform a simulation using the surrogate data instead of the limit price generated by the original rule of our model . the comparison of the resultant probability distribution of price shift of the surrogate data with that of the original data is shown in fig . 3 . the tail of the distribution does not show a power law behavior , though the original data does . this experiment reveals that the information of orderbook plays a essential role in the decision of the price at which new orders are placed in our model . a similar role of the orderbook will be expected even in real markets , though the style of the reference to the orderbook is possibly different from that assumed here . we derive the autocorrelation functions of price shift and of the absolute value of the price shift obtained by the numerical simulations of the model . the results are given in the panels of fig . 4 . including comparison with the autocorrelation functions of the surrogate data mentioned in the previous section . in those panels , the unit of time increment corresponds to a buy market order . the autocorrelation function of price shift almost vanishes except the value of time lag @xmath15 for both data . the values of the autocorrelation at time lag @xmath15 are -0.41 and -0.46 respectively . those values are close to -0.5 , and are explainable by the mean field approximation of the autocorrelation function as follows : let @xmath16 and @xmath17 denote the mean square root of price shift normalized by the standard deviation . the value @xmath16 measures the price shift across the spread , corresponding to the case that the side of the trade changes from bid to ask or from ask to bid . the value @xmath17 corresponds to the case that the side of the trade remains the same . if we assume that the four cases occurs with the same probability 1/4 , the mean field approximation of autocorrelation functions gives the equation @xmath18 . from the normalization condition @xmath19 and the inequality @xmath20 ( because spread always exists , while the trade successively occurred on the same side do not necessarily move the price ) , we have the result @xmath21 . the profile of the autocorrelation function is responsible for the value of the hurst exponent h through the equation @xmath22 . in such case of short memory as our model , we have the equation @xmath23 for @xmath24 . in our case , the diffusion constant @xmath25 is quite small owing to the equation @xmath21 , and h=1/2 for large t. an empirical study of the price diffusion is presented in fig . we see from the panel ( b ) of fig . 4 that the autocorrelation functions of the absolute value of price shift ( empirical volatility ) have long memory . both data plotted there are well fitted by power laws . the original data , however , hold the memory of volatility stronger than the surrogate data does . taking the strategy leaving the decision of the limit price to the others in the stochastic model of financial markets driven by continuous double auction , the virtual market shows the power - law tail of the distribution of returns with the exponent near 3 according to the parameter which determines the ratio of the mimetic limit order . the short memory of returns and the long memory of volatilities are also reproduced by the model . the hurst exponent h of our model is asymptotically 1/2 . the mean field approximation explains the profile of the autocorrelation function , which is responsible for the value of the hurst exponent h. the strategy assumed here are effective in holding the memory of market volatility strong . x. gabaix , p. gopikrishnan , v. plerou , h. e. stanley , a theory of power law distributions in financial market fluctuations , nature 423 ( 2003 ) 267 . j. d. farmer , l. gillemot , f. lillo , s. mike , a. sen , what really causes large price change ? , quantitative finance 4 ( 2004 ) 383 . s. maslov , simple model of a limit order - driven market , physica a 278 ( 2000 ) 571 . p. gopikrishnan , m. meyer , l. a. n. amaral and h. e. stanley , inverse cubic law for the distribution of stock price variations , eur . j. b 3 ( 1998 ) 139 . v. plerou , p. gopikrishnan , l. a. n. amaral , m. meyer , h. e. stanley , scaling of the distribution of price fluctuations of individual companies , phys . e 60 , ( 1999 ) 6519 . d. challet , r. stinchcombe , non - constant rates and over - diffusive prices in a simple model of limit order market , quantitative finance 3 ( 2003 ) 155 . y. liu , p. gopikrishnan , cizeau , meyer , peng , h. e. stanley , statistical properties of the volatility of price fluctuations , phys . e 60 , ( 1999 ) 1390 . e. smith , j. d. farmer , l. gillemot , s. krishnamurthy , statistical theory of the continuous double auction , quantitative finance 3 ( 2003 ) 481 . barabasi , r. albert , emergence of scaling in random networks , science 286 ( 1999 ) 509 . r. albert , a .- barabasi , statistical mechanics of complex networks , rev . of mod . 74 ( 2002 ) 47 . f. slanina mean - field approximation for a limit order driven market model , phys . e 64 ( 2001 ) 056136 .

we give a stochastic microscopic modelling of stock markets driven by continuous double auction . if we take into account the mimetic behavior of traders , when they place limit order , our virtual markets shows the power - law tail of the distribution of returns with the exponent outside the levy stable region , the short memory of returns and the long memory of volatilities . the hurst exponent of our model is asymptotically 1/2 . an explanation is also given for the profile of the autocorrelation function , which is responsible for the value of the hurst exponent .

introduction model numerical simulations of the model autocorrelation functions of the model conclusions

in many stock markets , price limit rules are set expecting to reduce remarked swings by cooling off traders irrational mania . a stable stock market has lower risks and thus attracts more people to participate . this is certainly increase the resource reallocation function of stock markets and benefits the economies . price limit rules constrain intraday prices to move within a preset price interval embraced by a price up limit and a price down limit . usually , the limit prices are determined by fixed fluctuation percentages in reference to the closing price of the previous day . in most stock markets , the fluctuation percentages for up - limit and down - limit are symmetric . however , there are also examples for asymmetric price limits especially in certain market states . after the price reaches its limit , a circuit breaker may be triggered causing trading halt in some markets , while in other markets the traders can continue to trade shares @xcite . the effectiveness of the price limit rules is controversial . it is expected to have a cooling - off effect to reduce the volatility of stocks @xcite . on the contrary , it may also cause a magnet effect , which refers to the phenomenon that the price limit acts as a magnet to attract more trades leading to higher trading intensity and price volatility and increases the probability of price rise or fall when the price is closer to the limit price @xcite . the magnet effect occurs when the traders fear of the lack of liquidity and possible position lock caused by imminent price limit hits , and the traders are thus eager to protect themselves through submitting aggressive sub - optimal orders , which usually induces large price variations and heavy trading volumes . since the cooling effect or the magnet effect takes place at the intraday level , studies of the presence of either effect , the evolution of pre - hit dynamics and the performance of price limit rules are of great interest to academics , investors and regulators to gain a better understanding of the mechanisms of how the market structure and the investors trading behavior affects price discovery . the study of price limits started on futures markets @xcite . empirical analysis has been carried out for different markets at different time periods . there is no consensus on the presence of a magnet effect or a cooling - off effect . arak and cook investigated if price behavior is infected by price limits on the treasury bond futures market and found no evidence of a magnet effect but rather a reversal effect @xcite . berkman and steenbeek compared the price formation processes under different price limits between osaka securities exchange and nikkei 225 index on the singapore international monetary exchange and found no significant arbitrage opportunities between the two markets @xcite . in recent years , empirical studies about the magnet effect concentrated on stock markets . cho et al . studied the 5-min return time series of 345 stocks traded on the taiwan stock exchange from 1998/01/03 to 1999/03/20 and reported a statistically and economically significant magnet effect for stock prices to accelerate towards the up - limit and weak evidence of acceleration towards the down - limit @xcite . hsieh et al . analyzed the transaction data of 439 stocks traded on the taiwan stock exchange in 2000 using logit models and found evidence of the magnet effect on both up - limt and down - limit @xcite . du et al . also observed the magnet effect when the stock prices are approaching the price limits in the korean market @xcite . the chinese stock markets also set price limits , which varied over time . the current @xmath0 price limits were fixed since 1996/12/16 for a - share common stocks . there are several studies conducted on the presence of the magnet or cooling - off effect . however , empirical results lead to controversial conclusions @xcite . according to the data released by the china securities regulatory commission , by june 2014 , there are 2540 listed companies and the total market capitalization is about 24.412 trillion chinese yuan . due to its huge capitalization and representativeness as an emerging market , research on chinese stocks is of great importance and remarkable interest . in this work , we will perform detailed analyses on the statistical properties of variables associated with price limit hits and the pre - hit dynamics of important financial variables before limit hits in the chinese stock markets . these issues are less studied in previous works . to obtain conclusive results about the presence of a magnet or cooling - off effect in the chinese stock markets , one needs to adopt different methods proposed in the literature and consider possible evolution of the effect ( if present ) per se . we leave this topic in a future work . there are two stock exchanges in mainland china . the first market for government approved securities was founded in shanghai on 1990/11/26 and started operation on 1990/12/19 under the name of the shanghai stock exchange ( shse ) . shortly after , the shenzhen stock exchange ( szse ) was established on 1990/12/01 , and started its operations on 1991/07/03 . there are two separate markets for a - shares and b - shares on both exchanges . a - shares are common stocks issued by mainland chinese companies , subscribed and traded in chinese currency renminbi ( rmb ) , listed on mainland chinese stock exchanges , bought and sold by chinese nationals and approved foreign investors . b - shares are issued by mainland chinese companies , traded in foreign currencies and listed on mainland chinese stock exchanges . b - shares carry a face value denominated in rmb . the b - share market was launched in 1992 and was restricted to foreign investors before 2001/02/19 . it has been open to chinese investors since . the microstructure of the two markets has been changed on several aspects , such as the daily price up / down limit rules imposed since 1996/12/16 . the price limits are @xmath0 for common stocks and @xmath1 for specially treated ( st and st * ) stocks . we note that , before 1996/12/16 , there were also periods with different intervals of price limits or without price limits . on each trading day , the trading time period is divided into three parts : opening call auction , cooling periods , and continuous double auction . the market opens at 9:15 a.m. and enters the opening call auction until 9:25 a.m , during which the trading system accepts order submissions and cancelations , and all matched transactions are executed at 9:25 a.m. this is followed by a cooling period from 9:25 to 9:30 a.m. during the cooling period , the exchanges are open to order routing from members , but does not accept the cancelation of orders . all matched orders are executed in real time . however , the information is not released to trading terminals during the cooling period and is publicly available at the end of the cooling period . the continuous double auction operates from 9:30 to 11:30 and from 13:00 to 15:00 ( for szse , 14:57 - 15:00 is a closing call auction period to form the close price ) and transaction occurs automatically by matching due to price and time priority . the time interval between 11:30 a.m. and 13:00 p.m. is a trade halt period . outside these opening hours , unexecuted orders will be removed by the system . our data sets were provided by resset ( http://resset.cn/ ) , which is a leading financial data provider supporting academic research . the data sets contain all common a - share stocks traded on the shanghai stock exchange and shenzhen stock exchange . the price limits for these stocks are @xmath0 . specially treated stocks with price limits of @xmath1 are not included in our analysis . the sample covers the period from 2000/01/04 to 2011/12/30 , totally 12 years . because the stocks have different initial public offering dates , the lengthes of stocks are not fixed . the quote frequency is about 5 seconds before 27 june 2011 and 3 seconds afterwards . due to different liquidities of the stocks , the quote frequencies of different stocks can be lower . for each stock , we have a unique stock code that is a sequence of six digital numbers , the trading time , the trading price , the trading volume , and the prices and standing volumes at the three best levels before 5 december 2003 or 5 levels afterwards on both the buy and sell sides of the limit order book . the records of stocks do not contain any indicator of price hits . hence , we need to identify when the price of a stock hits the up - limit or the low limit . because the tick size of all stocks is one cent ( 0.01 chinese yuan ) , no matter they have high prices or low prices , we are able to identify price limit hits . indeed , for each stock , we can determine the price up limit and price down limit for each trading day . let @xmath2 denotes the closing price of stock @xmath3 on day @xmath4 . the up - limit @xmath5 and the down - limit @xmath6 of stock @xmath3 on day @xmath7 are determined as follows , @xmath8}/100 , \label{eq : p+-}\ ] ] where @xmath9 $ ] is a round operator of @xmath10 such that the daily price limits are rounded to the tick size according to the _ trading rules of shanghai stock exchange _ ( 2003 , 2006 ) and the _ shenzhen stock exchange trading rules _ ( 2003 , 2006 ) . a representative measure of the status of the chinese stock markets is the shanghai stock exchange composite ( ssec ) index . figure [ fig : ssec : bull : bear ] illustrates the evolution of the ssec index from 1999 to 2011 . the ssec index rose from 1406 on 2000/01/04 to its historical intraday high of 2242.4 on 2001/06/13 and since plummeted 32.2% to 1520.7 on 2001/10/22 . the chinese stock markets entered a bearish antibubble state since june of 2011 @xcite , which did not synchronize the u.s . antibubble after the burst of the so - called new economy bubble in 2000 @xcite . the ssec index hit its all - time intraday low at 998.23 on 2005/06/04 during the time period investigated in this work . after that , a huge bubble formed and the index reached its historical intraday high at 6124.04 on 2007/10/16 , which was followed by a severe crash @xcite and the index plummeted to 1666.93 on 2008/10/28 as a all - time low since the crash . although the whole individual sentiment was bearish in the last six years , there was a mediate - size bubble started after 2008/10/28 and the index reached 3478.01 on 2009/08/04 . in summary , the time period under investigation in this work can be divided into alternating periods of bullish and bearish states . the stock market was bullish during the three time periods : 2000/1/4 - 2001/6/13 , 2005/06/04 - 2007/10/16 , and 2008/10/28 - 2009/08/04 . other periods are recognized as being bearish . * evolution of the shanghai stock exchange composite index from january 2000 to december 2011 . * the data shown are the closing prices . the historical highs and lows usually occurred intraday . the index has been divided into alternating bullish and bearish periods . the red parts stand for bullish periods and the green parts correspond to bearish periods.,width=302,height=226 ] in this work , we will compare the statistics of price limit hits in bullish and bearish market states . it is true that individual stocks may evolve differently . however , it is hard to recognize bullish and bearish states for individual stocks and it is not irrational to make such a comparison based on the ssec index . we leave this more complicated stock - by - stock classification of bulls and bears in future research . the empirical distribution of the number of limit hits for individuals is truncated on the left . we find that most of the distributions investigated in this work can be fitted by the left truncated normal distribution : @xmath11 both ordinary least squares and maximum likelihood estimation are applied in curve fitting . when we use maximum likelihood estimation to estimate parameters such as mean and variance , we can not use the sample mean and variance to substitute the whole population . a feasible method is the following @xcite : @xmath12 where @xmath13 and @xmath14 are respectively the density function and the cumulative density function of the standard normal distribution . denoting @xmath15 the mean of the truncated normal distribution , one has @xmath16\ } \end{array } \right.\ ] ] denoting @xmath17 the variance of the truncated normal distribution , one obtains @xmath18\}}}.\ ] ] defining @xmath19 , @xmath20 , one gets @xmath21\div\sqrt{1-\frac{{f(r)}}{{f(r)}}\left[r + \frac{{f(r)}}{{f(r)}}\right]}.\ ] ] in practical applications , we find that the function including @xmath22 on the right side of eq . ( [ eq : r : result ] ) is monotonically increasing . hence the equation has a unique solution . we can determine @xmath22 firstly , and then @xmath23 and @xmath24 . finally , we use @xmath23 and @xmath24 to estimate the population distribution . we provide statistical properties of the numbers of trading days with different types of limit hits . the variables are the following . @xmath25 is the total number of trading days with limit hits . @xmath26 is the average number of limit - hitting days for individual stocks . @xmath27 is the number of limit - hitting days with continued next - day opening prices . it contains up ( down ) limit hitting days with the opening prices on next trading days being higher ( lower ) than the closing prices . @xmath28 is the number of limit - hitting days with price reversal on the next day . it contains up ( down ) limit hitting days with the opening prices on next trading days being lower ( larger ) than the closing prices . @xmath29 , @xmath30 , @xmath31 and @xmath32 are respectively the numbers of days with limit hits occurred in the opening call auction ( 9:15 , 9:30 ] , in the continuous double auction session ( 9:30,11:30 ] in the morning , in the continuous double auction session [ 13:00 , 15:00 ] in the afternoon and at the closure of the trading days . @xmath33 and @xmath34 are the numbers of trading days that closed at limit prices and the price continued rising up or falling down on the successive trading days . it is possible that for some stocks there are both up- and down - limit hits within the same day . in this case , the first limit hit is used in the calculation of different numbers . we further delete ipo days and ex - dividend days . since trading halt may trigger after a price limit - hitting day and thus there is no followup open price , we do not count these days in @xmath27 , @xmath28 , @xmath33 , and @xmath34 . we also partition evenly all stocks with limit hits on a given day into six portfolios based on their capitalizations ( portfolio 1 with the smallest capitalizations and portfolio 6 with the largest capitalizations ) and count the defined numbers for each portfolio . the numbers are determined for the whole period , and the bullish and bearish periods as well . the capitalization of a stock is calculated as the product of the amount of shares times the price . the basic statistics of limit hits in the whole sample period and in bullish and bearish periods are presented in table [ tb : statistics : manystocks ] . it is trivial to observe that @xmath35 where the subscripts @xmath36 and @xmath37 represent respectively price up limit hit and price down limit hit , and @xmath38 stands for the portfolio serial number . cccccccccccccccccccccccccccccc + & & & & & & & & & & & & & & + & & up & down & & up & down & & up & down & & up & down & & up & down & & up & down & & up & down + @xmath25 & & 62346 & 44216 & & 10393 & 7369 & & 10393 & 7369 & & 10390 & 7368 & & 10389 & 7368 & & 10387 & 7368 & & 10394&7374 + @xmath26 & & 20.58 & 14.60 & & 3.43 & 2.43 & & 3.43 & 2.43 & & 3.43 & 2.43 & & 3.43 & 2.43 & & 3.43 & 2.43 & & 3.43 & 2.43 + @xmath27 & & 39696 & 33601 & & 6595 & 6242 & & 6819 & 5706 & & 6662 & 5569 & & 6538 & 5514 & & 6443 & 5324 & & 6639 & 5246 + @xmath28 & & 22627 & 10603 & & 3798 & 1127 & & 3574 & 1662 & & 3724 & 1798 & & 3848 & 1852 & & 3937 & 2040 & & 3746 & 2124 + @xmath29 & & 6718 & 2126 & & 1444 & 441 & & 1445 & 332 & & 1115 & 334 & & 898 & 347 & & 886 & 316 & & 930 & 356 + @xmath30 & & 33204 & 15694 & & 5740 & 3082 & & 5831 & 2460 & & 5673 & 2551 & & 5522 & 2674 & & 5434 & 2496 & & 5004 & 2431 + @xmath31 & & 29157 & 28525 & & 4653 & 4287 & & 4562 & 4909 & & 4720 & 4818 & & 4871 & 4695 & & 4959 & 4873 & & 5392 & 4943 + @xmath32 & & 40752 & 22213 & & 6207 & 3203 & & 6948 & 3745 & & 7137 & 3882 & & 7004 & 3857 & & 6871 & 3791 & & 6585 & 3735 + @xmath33 & & 32340 & 19067 & & 5120 & 2986 & & 5610 & 3252 & & 5612 & 3316 & & 5480 & 3278 & & 5321 & 3147 & & 5197&3088 + @xmath34 & & 8394 & 3138 & & 1087 & 217 & & 1338 & 493 & & 1521 & 566 & & 1522 & 577 & & 1544 & 641 & & 1382 & 644 + + & & & & & & & & & & & & & & + & & up & down & & up & down & & up & down & & up & down & & up & down & & up & down & & up & down + @xmath25 & & 32593 & 17898 & & 5432 & 2983 & & 5432 & 2983 & & 5432 & 2983 & & 5432 & 2983 & & 5431 & 2983 & & 5434 & 2983 + @xmath26 & & 26.52 & 14.56 & & 4.42 & 2.43 & & 4.42 & 2.43 & & 4.42 & 2.43 & & 4.42 & 2.43 & & 4.42 & 2.43 & & 4.42 & 2.43 + @xmath27 & & 22344 & 13888 & & 3501 & 2563 & & 3713 & 2397 & & 3690 & 2315 & & 3787 & 2278 & & 3775 & 2190 & & 3878 & 2145 + @xmath28 & & 10249 & 4010 & & 1931 & 420 & & 1719 & 586 & & 1742 & 668 & & 1645 & 705 & & 1656 & 793 & & 1556 & 838 + @xmath29 & & 2908 & 986 & & 660 & 161 & & 508 & 134 & & 421 & 175 & & 432 & 183 & & 440 & 162 & & 447 & 171 + @xmath30 & & 15785 & 6730 & & 2925 & 1117 & & 2657 & 1005 & & 2650 & 1168 & & 2626 & 1191 & & 2590 & 1173 & & 2337 & 1076 + @xmath31 & & 16809 & 11168 & & 2507 & 1866 & & 2775 & 1978 & & 2782 & 1815 & & 2806 & 1792 & & 2842 & 1810 & & 3097 & 1907 + @xmath32 & & 20814 & 8778 & & 3035 & 1278 & & 3511 & 1480 & & 3644 & 1559 & & 3652 & 1540 & & 3577 & 1493 & & 3395 & 1428 + @xmath33 & & 17394 & 8011 & & 2541 & 1195 & & 2923 & 1371 & & 2977 & 1407 & & 3062 & 1406 & & 2986 & 1352 & & 2905 & 1280 + @xmath34 & & 3420 & 767 & & 494 & 83 & & 588 & 109 & & 667 & 152 & & 590 & 134 & & 591 & 141 & & 490 & 148 + + & & & & & & & & & & & & & & + & & up & down & & up & down & & up & down & & up & down & & up & down & & up & down & & up & down + @xmath25 & & 29753 & 26318 & & 4961 & 4386 & & 4960 & 4386 & & 4956 & 4385 & & 4957 & 4385 & & 4958 & 4385 & & 4961 & 4391 + @xmath26 & & 16.53 & 14.62 & & 2.76 & 2.44 & & 2.76 & 2.44 & & 2.75 & 2.44 & & 2.75 & 2.44 & & 2.75 & 2.44 & & 2.76 & 2.44 + @xmath27 & & 17352 & 19713 & & 3062 & 3678 & & 3034 & 3299 & & 2810 & 3247 & & 2785 & 3231 & & 2824 & 3162 & & 2837 & 3096 + @xmath28 & & 12378 & 6593 & & 1899 & 708 & & 1926 & 1086 & & 2141 & 1137 & & 2167 & 1152 & & 2130 & 1219 & & 2115 & 1291 + @xmath29 & & 3810 & 1140 & & 935 & 282 & & 931 & 189 & & 600 & 155 & & 437 & 165 & & 409 & 163 & & 498 & 186 + @xmath30 & & 17419 & 8964 & & 3004 & 1960 & & 3231 & 1434 & & 2984 & 1404 & & 2843 & 1445 & & 2747 & 1365 & & 2610 & 1356 + @xmath31 & & 12348 & 17357 & & 1957 & 2426 & & 1730 & 2952 & & 1977 & 2982 & & 2118 & 2941 & & 2214 & 3021 & & 2352 & 3035 + @xmath32 & & 19938 & 13435 & & 3244 & 1924 & & 3424 & 2261 & & 3432 & 2324 & & 3344 & 2322 & & 3333 & 2311 & & 3161 & 2293 + @xmath33 & & 14946 & 11056 & & 2621 & 1788 & & 2644 & 1871 & & 2502 & 1900 & & 2423 & 1857 & & 2435 & 1824 & & 2321 & 1816 + @xmath34 & & 4974 & 2371 & & 623 & 136 & & 780 & 390 & & 925 & 424 & & 917 & 463 & & 895 & 484 & & 834 & 474 + [ tb : statistics : manystocks ] panel a of table [ tb : statistics : manystocks ] shows the results for the whole period from 2000 to 2011 . there are more limit - up hits than limit - down hits , which is partially caused by the rapid growth of china s economy and the absence of short mechanism . the limit - down days are more likely to have next - day price continuation than limit - up days since @xmath39 and @xmath40 . in addition , a limit hitting day is more likely to have price continuation than price reversal because @xmath41 . for all other numbers concerning all stocks and different portfolios , we also observe more limit - up days than limit - down days . for limit - down days , the probability of next - day price continuation decreases with increasing average capitalization of the portfolio , while the probability of next - day price reversal increases with the capitalization . for limit - up days , the probabilities of next - day price continuation and price reversal do no have any clear trend . for @xmath42 and @xmath43 , we observe decreasing trends with capitalization . for @xmath44 , we observe increasing trends with capitalization . for @xmath45 , they increase first and then decrease . in all the seven cases ( all stocks and the six portfolios ) , limit - up events are more likely to occur in the morning than in the afternoon , while limit - down events are more likely to happen in the afternoon than in the morning . limit - up events have higher probability ( 65.3% ) to close at limit price than limit - down events ( 51.0% ) according to @xmath32 . when a trading day closes at the up - limit or the down - limit , the next - day opening price will rise with a very high probability ( 79.4% for limit - up events and 85.8% for limit - down events ) . in addition , @xmath46 and @xmath47 increase first and then decrease . for the bullish periods in panel b of table [ tb : statistics : manystocks ] , in all seven cases ( all stocks and the six portfolios ) , there are more limit - up days for @xmath25 , @xmath29 , @xmath30 and @xmath32 . for the bullish periods in panel c of table [ tb : statistics : manystocks ] , there are more limit - up days than limit - down days for @xmath25 , @xmath29 , @xmath30 , and @xmath32 , with an exception that there are more limit - down days than limit - up days in the afternoon ( @xmath48 ) . however , the occurrence difference between limit - up and limit - down days is smaller in the bearish periods than in the bullish periods . for bullish periods in panel b , @xmath49 and @xmath50 decrease with increasing capitalization , @xmath51 increases with capitalization , and no evident trends have been found in @xmath52 , @xmath53 , @xmath54 and @xmath45 . for bearish periods in panel c , @xmath42 and @xmath43 decrease with capitalization , @xmath51 and @xmath54 increase with capitalization , and @xmath55 and @xmath56 do not exhibit any clear trend with respect to capitalization . for both bullish and bearish periods , price continuation is more likely to occur than price reversal for both limit - up and limit - down days , because @xmath57 and @xmath58 . in addition , the probability of price continuation is higher for limit - down days than for limit - up days , that is @xmath59 and @xmath60 . for bullish periods , the probability of price continuation increases with capitalization for limit - up days ( @xmath61 ) and decreases with capitalization for limit - down days ( @xmath62 ) . these trends are absent for @xmath45 . for bearish period , we observe that @xmath62 , @xmath63 and @xmath64 all have a decreasing trend with respect to the capitalization . these findings have potential applications for practitioners in the chinese stock markets , keeping in mind that one can not short and there is a @xmath7 trading rule . when the price of a stock hits the up - limit , no matter in the intraday or at the close , she can buy right before market closure at 15:00 and sell with the opening price on the next trading day . regardless of the transaction cost , the return will have high probability to be positive . the probability of earning money is even high if the price is locked to the up - limit and when the capitalization of the stock is low . certainly , in this case , the liquidity is usually very low and it is hard to buy shares . on the contrary , if one holds a stock whose price experiences intraday down - limit hits , it is better to sell it to reduce losses . for each stock @xmath3 traded in @xmath65 days in our sample , we identify @xmath66 trading days on which the prices have hit either the up - limit or the down - limit at least once . we denote the set of limit - up days ( trading days that have up - limit hits ) of stock @xmath3 as @xmath67 , where @xmath68 is the number of limit - up trading days . similarly , the set of limit - down days of stock @xmath3 is denoted as @xmath69 , where @xmath70 is the number of limit - down trading days . the intersection of @xmath71 and @xmath72 is not necessary to be empty , because , although very rare , it is possible that stock @xmath3 hits its up - limit and down - limit on the same day . in other words , @xmath73 . the percents of limit - hitting days are calculated as follows @xmath74 where @xmath75 , @xmath76 and @xmath77 can be regarded as the empirical probabilities that stock @xmath3 will hit either the upper or the down - limit , the up - limit , and the down - limit on a trading day , respectively . figure [ fig : magneteffect : prob : hit : per : day ] shows the empirical distributions of limit - hitting probability per day for individual stocks . as shown in fig . [ fig : magneteffect : prob : hit : per : day](a ) , for most of the stocks , the daily limit - hitting probability is less than 6% , with the most probability value around 2.5% . however , there are also a few stocks having very large limit - hitting probability up to 17% . such seemingly outliers include 002606 , 002632 , 002635 , 002643 , 002644 , and 002646 . these stocks have relatively large numbers of price down limit hits , as shown in fig . [ fig : magneteffect : prob : hit : per : day](c ) . the large values of @xmath77 are mainly caused by the small values of @xmath65 because these stocks come into the market for a short time period . for most stocks , the daily up - limit hit is less than 4% according to fig . [ fig : magneteffect : prob : hit : per : day](b ) and the daily dow - limit hit is less than 3% according to fig . [ fig : magneteffect : prob : hit : per : day](c ) . this finding is consistent with the observation that @xmath78 in table [ tb : statistics : manystocks ] . it seems that the curves obtained by the ols method fit the empirical data better than those by the mle method . , @xmath79 and @xmath80 of daily price limit hitting probabilities @xmath81 , @xmath82 and @xmath83 for individual stocks . * the dots are empirical data , the black dash - dotted curves are the mle fits to the truncated normal distribution in eq . ( [ eq : normal : left ] ) , and the black dash - dotted curves are the ols fits to the truncated normal distribution . ( a ) all limit hits ; ( b ) price up limit hits ; ( c ) price down limit hits.,scaledwidth=96.0%,scaledwidth=25.0% ] consider a limit - hitting day @xmath84 ( or @xmath85 ) of stock @xmath3 . the price may hit the price limit for @xmath86 ( or @xmath87 ) times at intraday moments @xmath88 ( or @xmath89 ) with @xmath90 ( or @xmath87 ) . we define the average number of limit hits of stock @xmath3 as follows , @xmath91 where the denominators are not the total number of trading days , @xmath65 , but the number of price up limit hitting days and of price down limit hitting days . according to fig . [ fig : magneteffect : pdf : num : hits](a ) and fig . [ fig : magneteffect : pdf : num : hits](b ) , the distributions of @xmath86 and @xmath87 decrease sharply . about 30% of limit - hitting days have only one up - limit hit or down - limit hit . however , there are also limit - hitting days with very large numbers of limit hits in one day . for instance , stock 600863 hit the up - limit for 151 times on 2009/07/14 , while stock 600102 hit the down - limit for 149 times on 2009/04/27 , which has been delisted from the shanghai stock exchange due to it poor performance . surprisingly , the two distributions @xmath92 and @xmath93 are not monotonically decreasing functions . instead , they can be well fitted by the truncated normal distribution . speaking differently , if the price limit ( 10% or -10% ) of a stock has been reached on certain trading day , the number of limit hits on that day is usually more than once . and @xmath94 of @xmath86 and @xmath87 . ( c , d ) probability density functions @xmath92 and @xmath93 of @xmath95 and @xmath96 for individual stocks . , scaledwidth=66.0%,scaledwidth=52.0% ] when stock @xmath3 hits its price limit on day @xmath84 ( or @xmath85 ) , the @xmath97-th limit - up ( or limit - down ) started at @xmath88 ( or @xmath89 ) may be opened at @xmath98 ( or @xmath99 ) . hence , we have a sequence of limit - hitting durations @xmath100 ( or @xmath101 ) . the total duration on that day can be calculated as follows : @xmath102 considering only the limit - hitting days of stock @xmath3 , we can also define the average limit hit duration as follows : @xmath103 we stress that these two quantities are defined for individual stocks and the denominators are not @xmath65 . plots ( a ) and ( d ) of fig . [ fig : magneteffect : pdf : duration ] show the empirical distributions of the durations @xmath100 and @xmath101 of individual limit hits for all stocks . the two distributions exhibit a similar `` l '' shape . in each distribution , there are three obvious peaks at @xmath104s , 10800s and 12600s , which correspond respectively to 2 hours , 3 hours and 3.5 hours . these peaks are mainly caused by limit hits at 13:00 p.m. , 10:30 a.m. and 10:00 a.m. with the prices remaining at the limit prices till market closure . these patterns contain significant information contents . the peak around 13:00 reflects the fact that the majority of the traders or some informed traders may obtain cumulated important information about the holding stock during the market closure at noon ( 11:30 - 13:00 ) and form a collective behavior to buy or sell their stock to push the price to its limit . this interpretation applies certainly to the peak at 14400s , corresponding to the case that the stock opens at its limit price due to overnight information and does not fluctuate during the whole trading day . the peak at 10800s are caused by the one - hour opening trading halts in which the stock usually bears abnormal information pit and traders trading habits and round number preference . the peak at 12600s is less significant , which is very likely caused by traders trading habits and round number preference . and @xmath105 of the individual limit hit durations @xmath100 and @xmath101 . ( b , e ) probability density functions @xmath106 and @xmath107 of the daily limit hit durations @xmath108 and @xmath109 . ( c , f ) probability density functions @xmath110 and @xmath111 of the average daily limit hit durations @xmath112 and @xmath113 . the unit of the variables is second.,scaledwidth=96.0%,scaledwidth=52.0% ] plots ( b ) and ( e ) of fig . [ fig : magneteffect : pdf : duration ] show the empirical distributions of the daily limit hit durations @xmath108 and @xmath109 . the overall shapes of these two distributions are very similar with those in fig . [ fig : magneteffect : pdf : duration ] ( a ) and ( d ) . the main differences are that the distributions of daily limit hit durations have lower heights at the two edges and larger values in the bold parts . plots ( c ) and ( f ) of fig . [ fig : magneteffect : pdf : duration ] present the empirical distributions of the average daily durations @xmath112 and @xmath113 for individual stocks . the distributions are bimodal with an extra peak close to @xmath114s . we further define the total intraday limit hit duration on trading day @xmath115 , which is the time elapse from the moment of the first limit hit to the last moment that prices stay at limit in a limit - hitting day : @xmath116 by definition , we have @xmath117 . we also define the average of the total intraday limit - hitting duration for individual stocks as follows , @xmath118 similarly , we have @xmath119 . the upper panel of fig . [ fig : magneteffect : pdf : totalduration ] shows the empirical distributions of the total intraday limit hit durations @xmath120 and @xmath121 for all stocks . the distribution in fig . [ fig : magneteffect : pdf : totalduration](a ) has two local maxima around @xmath122s and 10800s as in fig . [ fig : magneteffect : pdf : duration](b ) for @xmath108 . the distribution in fig . [ fig : magneteffect : pdf : totalduration](b ) has a similar overall shape . however , it is less smoother . the lower panel of fig . [ fig : magneteffect : pdf : totalduration ] shows the empirical distributions of @xmath123 and @xmath124 . it is hard to find a suitable function form to fit the data . and @xmath125 of the total intraday limit hit durations @xmath120 and @xmath121 . ( c , d ) probability density functions @xmath126 and @xmath127 of the averages of total intraday limit hit durations @xmath123 and @xmath124 . the unit of the variables is second . , scaledwidth=66.0%,scaledwidth=52.0% ] as a complementary to the distributions in fig . [ fig : magneteffect : pdf : num : hits ] to fig . [ fig : magneteffect : pdf : totalduration ] , we investigate the maxima , the medians and the means of the three kinds of daily variables , the daily numbers of limit hits @xmath128 , the daily limit hit durations @xmath129 and the total intraday limit hit durations @xmath130 . we compare the results for the whole time period , the bullish periods and the bearish periods . we also divide all the stocks into six portfolios as defined in table [ tb : statistics : manystocks ] to investigate the impacts of stock capitalization . the results are presented in table [ tb : intraday : statistics : individualstocks ] . cccccccccccccccccccccccccccccccccccccccccccccc + & & & & & & & & & & & & & & + & & up & down & & up & down & & up & down & & up & down & & up & down & & up & down & & up & down + @xmath131 & & 151 & 149 & & 88 & 87 & & 114 & 102 & & 113 & 96 & & 125 & 119 & & 133 & 99 & & 151 & 149 + @xmath132 & & 7.47 & 8.27 & & 5.74 & 6.67 & & 7.13 & 7.86 & & 7.70 & 8.62 & & 7.95 & 8.55 & & 8.06 & 8.83 & & 8.23 & 9.09 + @xmath133 & & 4 & 5 & & 3 & 4 & & 4 & 5 & & 4 & 6 & & 4 & 6 & & 4 & 6 & & 4 & 6 + @xmath134 & & 4391 & 2189 & & 4285 & 2014 & & 4723 & 2036 & & 4696 & 2291 & & 4542 & 2347 & & 4300 & 2247 & & 3801 & 2202 + @xmath135 & & 2096 & 815 & & 1720 & 552 & & 2386 & 745 & & 2397 & 906 & & 2282 & 911 & & 2113 & 866 & & 1815 & 917 + @xmath136 & & 5740 & 3888 & & 5553 & 3793 & & 6241 & 3661 & & 6059 & 4035 & & 5822 & 4054 & & 5590 & 3924 & & 5175 & 3857 + @xmath137 & & 4118 & 1937 & & 3496 & 1785 & & 5020 & 1796 & & 4740 & 2035 & & 4324 & 2077 & & 4004 & 1953 & & 3408 & 1982 + + & & & & & & & & & & & & & & + & & up & down & & up & down & & up & down & & up & down & & up & down & & up & down & & up & down + @xmath131 & & 151 & 149 & & 88 & 87 & & 97 & 102 & & 114 & 96 & & 113 & 72 & & 125 & 97 & & 151 & 149 + @xmath132 & & 7.34 & 7.76 & & 6.42 & 7.16 & & 7.19 & 7.67 & & 7.71 & 8.18 & & 7.64 & 7.83 & & 7.46 & 7.90 & & 7.61 & 7.83 + @xmath133 & & 4 & 5 & & 3 & 4 & & 4 & 5 & & 4 & 5 & & 4 & 5 & & 4 & 5 & & 4 & 5 + @xmath134 & & 4064 & 2403 & & 3956 & 1757 & & 4276 & 1991 & & 4327 & 2643 & & 4288 & 2858 & & 4011 & 2719 & & 3525 & 2452 + @xmath135 & & 1778 & 790 & & 1290 & 408 & & 1901 & 603 & & 2070 & 956 & & 2093 & 1064 & & 1850 & 1016 & & 1580 & 965 + @xmath136 & & 5251 & 4011 & & 5123 & 3566 & & 5410 & 3521 & & 5524 & 4307 & & 5493 & 4388 & & 5225 & 4286 & & 4736 & 3997 + @xmath137 & & 3415 & 2018 & & 2689 & 1539 & & 3675 & 1605 & & 3982 & 2187 & & 3877 & 2264 & & 3495 & 2213 & & 2853 & 2089 + + & & & & & & & & & & & & & & + & & up & down & & up & down & & up & down & & up & down & & up & down & & up & down & & up & down + @xmath131 & & 133 & 125 & & 65 & 59 & & 88 & 71 & & 108 & 73 & & 133 & 119 & & 131 & 99 & & 124 & 125 + @xmath132 & & 7.61 & 8.62 & & 4.84 & 6.30 & & 7.23 & 8.02 & & 7.96 & 8.99 & & 8.28 & 9.05 & & 8.55 & 9.58 & & 8.81 & 9.78 + @xmath133 & & 3 & 6 & & 2 & 4 & & 4 & 6 & & 4 & 6 & & 4 & 6 & & 4 & 7 & & 4 & 7 + @xmath134 & & 4750 & 2044 & & 4912 & 2193 & & 5227 & 2042 & & 5065 & 2013 & & 4764 & 1988 & & 4468 & 1979 & & 4064 & 2049 + @xmath135 & & 2520 & 829 & & 2848 & 707 & & 3093 & 878 & & 2640 & 855 & & 2541 & 817 & & 2305 & 819 & & 2043 & 885 + @xmath136 & & 6275 & 3804 & & 6460 & 3950 & & 7179 & 3717 & & 6470 & 3830 & & 6104 & 3825 & & 5838 & 3753 & & 5597 & 3747 + @xmath137 & & 5030 & 1871 & & 5509 & 1919 & & 7172 & 1917 & & 5480 & 1869 & & 4905 & 1814 & & 4305 & 1845 & & 4002 & 1885 + [ tb : intraday : statistics : individualstocks ] for the maximal number of limit hits , we find that @xmath138 for all cases , except for portfolio 2 in bullish periods . however , for the means , we have @xmath139 in all the cases . on average , down - limit hits are less stable than up - limit hits , indicating that the fear sentiment of traders fluctuates more than the greed sentiments in the chinese stock markets . similarly , we observe that @xmath140 in all the cases . however , the difference @xmath141 is larger in the bearish periods than in the bullish periods . it indicates that the fighting between long positions and short positions is more severe when the price hits the down - limit than the up - limit , because trades can make profits when the market rises up while they can only deduce losses when the market falls down . no clear correlation is found between @xmath132 and capitalization in the bullish periods . in contrast , both @xmath142 and @xmath143 are positively correlated with capitalization in the bearish periods . for the other variables , we find that the quantities for up - limits are greater than their counterparts for down - limits . specifically , we observe that @xmath144 , @xmath145 , @xmath146 , and @xmath147 for different periods and different portfolios . together with the results for @xmath132 , we find that the average duration of individual up - limit hits is about twice longer than that of individual down - limit hits . this is probably caused by the no - short trading rule because traders can make money only when the market rises . there are also traders known as `` dare - to - die corps for up - limit hits '' , who are actively engaged in pushing prices to the up - limits . it is also found that these quantities have a bell - like shape or monotonically decrease with respect to the capitalization . we divide the continuous double auction period in each trading day into @xmath25 intervals of equal length of @xmath148 minutes . if an interval starts at @xmath149 and ends at @xmath150 , we can count the number of occurrences of up - limit hits in this interval as @xmath151 } , \label{eq : c : u : i}\ ] ] and similarly the number of occurrences of down - limit hits @xmath152 } , \label{eq : c : d : i}\ ] ] where @xmath153 is an indicator function of event such that the value of @xmath153 is 1 if the event @xmath10 is true and 0 otherwise . we can further calculate the following quantities @xmath154 where @xmath155 is the number of stocks in the investigated sample . the intraday patterns of the occurrence of limit hits in intervals of size @xmath156 min are illustrated in fig . [ fig : interval : c : ud ] . for each intraday interval , we have @xmath157 . the occurrence of up - limit hits is extremely high at the opening of the market around 9:30 a.m. , while the occurrence of down - limits is relatively high at the opening . there are also other local peaks around 09:30 , 09:45 , 10:00 , 10:30 , 11:30 , 13:30 , 14:30 and 14:45 . up - limit hits have different intraday patterns from down - limit hits . however , the intraday patterns do not differ much when comparing the bullish periods with the bearish periods . in all the six cases , limit hits are more frequent in the last hour , especially for down - limit hits . for up - limits , the occurrence number is quite stable over the period from 10:30 a.m. to 14:20 p.m. , except for the afternoon opening of the market . is 5 min.,scaledwidth=96.0%,scaledwidth=52.0% ] limit hits are rare events . the dynamics of financial variables before limits can enrich our understanding of trading activities of investors around such extreme events . we investigate here the evolution of several important financial variables before stock prices hit the up - limits and down - limits . limit hits occurred at the opening of the market are excluded from analysis , because one can not trace the pre - event dynamics of any financial variables . we first study the average velocity of price change for all limit - hitting events in isometric intervals of price changes during bullish periods and bearish periods . for each limit - hitting event @xmath3 , we consider the cases when the price rises above 5% or falls below -5% , compared with the close price of the previous trading day . we divide each of the two intervals @xmath158 $ ] and @xmath159 into 10 subintervals @xmath160 $ ] and @xmath161 , where @xmath162 . let @xmath163 denotes the time duration for the stock price rising from @xmath164 to @xmath165 $ ] or dropping from @xmath166 to @xmath167 $ ] . we consider four classes of limit hits : up - limit hits in bullish periods , up - limit hits in bearish periods , down - limit hits in bullish periods , and down - limit hits in bearish periods . for each class of limit hits , we define the dimensionless velocity of price change as follow , @xmath168 where @xmath169 is the number of limit hits of the class under investigation . if @xmath163 is independent of @xmath97 for all @xmath3 , the velocity is constant such that @xmath170 . if the average @xmath171 over @xmath3 increases ( decreases ) with @xmath97 , @xmath172 decreases ( increases ) when the price approaches the price limit , showing evidence of a cooling off ( magnet ) effect . as shown in fig . [ fig : prehit : price : acceleration ] , when stock price approaches the limit price , the price movement velocity decreases , except that the velocity does not change much when the price rise is less than 9% for up - limit hits in the bullish periods . it suggests that there exists cooling - off effects when stock price gets close to the limit price . in this sense , the price limit rule is effective in the chinese stock markets . the cooling - off effect is more remarkable in bearish periods than in bullish periods before up - limit hits and down - limit hits . however , the effectiveness of the cooling - off effect is mixed between up - limit hits and down - limit hits . the asymmetry of the cooling - off effect between bullish and bearish periods is probably caused by the no - short trading rule and the special lift forces in the bullish periods , such as the `` dare - to - die corps for up - limit hits '' . * cooling - off effect of the price limit rule . * ( a ) comparison of price movement velocity @xmath173 before up - limit hits in the bullish and bearish periods . ( b ) comparison of price movement velocity @xmath174 before down - limit hits in the bullish and bearish periods.,scaledwidth=96.0%,scaledwidth=36.0% ] we now study the evolution of another four financial measures associated with the last 100 trades till hitting the price limits . we first investigate the evolution of sizes of buyer - initiated trades and seller - initiated trades on the last 100 trades including the one pushing the price to its limit . the last trade causes the price to hit its limit , labelled as the 100th trade . if the @xmath175-th trade was initiated by a buyer before an up - limit hit @xmath3 or by a seller before a down - limit hit @xmath3 , we denote @xmath176 as its size and , in this case , @xmath177 . alternatively , if the @xmath175-th trade was initiated by a seller before an up - limit hit @xmath3 or by a buyer before a down - limit hit @xmath3 , we denote @xmath178 as its size and @xmath179 . hence , @xmath176 and @xmath178 are respectively the sizes of same - direction ( momentum ) and opposite - direction ( contrarian ) trades that move the price towards and away from the limit price . we compare up - limit hits and down - limit hits . we also consider separately bullish periods and bearish periods . the average logarithmic trade sizes of the last 100 transactions are calculated as follows , @xmath180 , \label{eq : prehit : tradesizes}\ ] ] where @xmath169 is the number of limit hits in one of the four cases ( up - limit in bullish periods , down - limit in bullish periods , up - limit in bearish periods , and down - limit in bearish periods ) . since the last trade is by definition the same - direction trade that push the price to the limit , we have @xmath181 in all cases . for simplicity , we use the notation @xmath182 instead of @xmath183 in the following . figure [ fig : prehit:100trades:4quantities](a ) illustrates the evolution of the same - direction trade sizes @xmath184 and the opposite - direction trade sizes @xmath185 for different cases . the same - direction trade sizes @xmath184 have similar evolutionary trajectories in all the four cases , so do the opposite - direction trade sizes @xmath185 . we find that @xmath184 increases from @xmath186 to reach the local maximum at @xmath187 and then decreases till @xmath188 followed by a very large @xmath189 pushing the price to the limit , while @xmath185 decreases from @xmath186 to reach the local minimum at @xmath187 and then increases till @xmath188 , ended with @xmath182 . for all @xmath175 , we have @xmath190 , indicating that the pushing force of the same - direction traders is stronger than that of the opposite - direction traders . the trade size difference @xmath191 continuously increases and reaches its local maxima at @xmath187 and then decreases till @xmath188 . these findings also provide clues about the cooling down of the last three trades right before limit hits . the trade sizes @xmath184 are larger when the price is close to the up - limit than to the down - limit , which is true for both bullish and bearish periods . the trade sizes @xmath184 for the case of down - limit hits in bearish periods are relatively the lowest . the relative position of the four @xmath184 curves provides evidence of a well - known trait of chinese traders that they tend to buy rising stocks and hate to sell holding shares when the stock price drops . * evolution of four important financial quantities along the last 100 trades right before limit hits . * ( a ) evolution of the logarithmic same - direction trade sizes @xmath184 ( the upper bundle ) and the logarithmic opposite - direction trade sizes @xmath185 ( the lower bundle ) for up - limit hits and down - limit hits in the bullish periods and bearish periods . ( b ) evolution of the average trade - by - trade return @xmath192 . ( c ) evolution of the average trade - by - trade volatility @xmath193 . ( d ) evolution of the average bid - ask spread right before individual trades . for each financial quantity , we consider four cases , that is , up - limit hits in bullish periods , down - limit hits in bullish periods , up - limit hits in bearish periods , and down - limit hits in bearish periods.,scaledwidth=80.0%,scaledwidth=60.0% ] as shown in fig . [ fig : prehit:100trades:4quantities](a ) , the average size @xmath189 of the last trade that pushed the price to hit the limit is extraordinary larger than the size of any preceding trades . we show the empirical distributions of @xmath189 for the four cases . all the distributions are unimodal and have similar shapes . however , they have different peak heights . the peak is the highest for the case of up - limit hits in the bullish periods , the second largest for the case of up - limit hits in the bearish periods . we denote @xmath194 the transaction price of the @xmath175-th trade before the @xmath3-th limit hit , @xmath195 and @xmath196 the prices and standing volumes at the first @xmath197 price levels of the sell - side limit order book right before the @xmath175-th trade , and @xmath198 and @xmath199 the prices and standing volumes on the first @xmath197 price levels of the buy - side limit order book right before the @xmath175-th trade . the second financial variable investigated is the average trade - by - trade return which is defined as follows , @xmath200 figure [ fig : prehit:100trades:4quantities](b ) shows the evolution of average trade - by - trade return @xmath192 before price limit hits . it is found that @xmath192 increases superlinearly before @xmath175 close to 100 and then decreases sharply before up - limit hits in both bullish and bearish periods . almost symmetrically , @xmath192 decreases superlinearly before @xmath175 close to 100 and then increases sharply before down - limit hits in both bullish and bearish periods . these patterns are consistent with the behaviors of the average trade sizes , because large trade sizes usually cause large price movements @xcite . the third variable is the average trade - by - trade volatility which is defined as follows , @xmath201 figure [ fig : prehit:100trades:4quantities](c ) presents the evolution of average trade - by - trade volatility @xmath202 before price limit hits . the volatility @xmath202 before up - limit hits for both bullish and bearish periods is relatively stable when @xmath175 is less than about 50 , then increases rapidly to reach a maximum one or two trades before limit hits , and finally drops to some extent . about five trades before up - limit hits , the volatility is higher during bullish periods than in bearish periods . the volatility about 70 trades before down - limit hits is higher than that before up - limit hits and exhibits a mild decreasing trend . the volatility increases afterwards and decreases again around @xmath203 . the fourth variable investigated is the average bid - ask spread right before the @xmath175-th trade which is defined as follows , @xmath204 } , ~~ k= 1,2,3,\cdots,100 . \label{eq : prehit:100trades : spread}\ ] ] as illustrated in fig . [ fig : prehit:100trades:4quantities](d ) , the four curves of the average bid - ask spread @xmath205 decrease before @xmath206 . the two spread curves for the bullish periods have very similar shapes . after the initial decrease , they increase during the next twenty trades or so and decrease again before the limit hits . for the curve associated with up - limit hits in bearish periods , the spread increases continuously after @xmath207 . the comparison of the four curves is also quite intriguing . the spread before up - limit hits is narrower than that before down - limit hits , indicating higher liquidity before up - limit hits . on average , the spread is the narrowest before up - limit hits in bearish periods and the widest before down - limit hits in bearish periods . stock markets are complex systems in which humans interact by buying and selling shares . the evolutionary trajectories of stock markets are fully determined by the behaviors of human being . it is widely accepted that people in emerging markets are less skilled and more irrational and thus these markets are much riskier . due to different factors such as imitation , global news , as well as collusive manipulation , traders may herd to push the price rise up or drop down rapidly in very short time intervals through positive feedback loops . such kind of collective behaviors might be caused partially by the somewhat reciprocity among a small amount of traders through price manipulation , which is reminiscent of the cooperation phenomena among human beings @xcite . to cool off traders intraday mania to avoid the price deviating much from its fundamental value , a number of stock exchanges pose price limit rules . empirical evidence is controversial about the presence of a magnet effect or a cooling off effect even for the chinese stock markets @xcite . our preliminary result on the price movement velocity favors the presence of an intraday cooling - off effect for both up - limits and down - limits in both bullish and bearish periods . the evolution of other financial variables such as trade size , trade - by - trade return , trade - by - trade volatility and bid - ask spread seems to support the conclusion that the traders are cooled off right before stock prices hit the price limits , as shown by the anti - trend behavior just before hitting the price limits . it is well documented that the limit order books are thin near price limits @xcite and thus the liquidity is worse . when the price is pushed towards its limit , traders submit larger and larger orders and the price changes enlarge . however , the population of traders who place opposite orders to realize their gains or act as bottom fishers also increases . when the price is close to its limit , the force of opposite traders reverses the trends . in this sense , the price limit rule works in the chinese stock markets . we have investigated the statistical properties of characteristic variables of up - limit and down - limit hits in bullish and bearish periods . we also uncovered nonlinear impacts of stock capitalization on price limit hits by comparing six portfolios sorted due to stocks capitalization on the daily level . it is intriguing to find that price continuation occurs more frequently than price reversal on the next trading day after a limit - hitting day . this effect is more significant for down - limit hits . the empirical probability of next - day price continuation is thus far greater than 50% . our empirical findings have potential practical values for market practitioners . for instance , it will be probably profitable to buy shares at the close price in a up - limit hitting day and sell the share at the opening of the next trading day , or to sell the shares one holds in a down - limit hitting day .

price limit trading rules are adopted in some stock markets ( especially emerging markets ) trying to cool off traders short - term trading mania on individual stocks and increase market efficiency . under such a microstructure , stocks may hit their up - limits and down - limits from time to time . however , the behaviors of price limit hits are not well studied partially due to the fact that main stock markets such as the us markets and most european markets do not set price limits . here , we perform detailed analyses of the high - frequency data of all a - share common stocks traded on the shanghai stock exchange and the shenzhen stock exchange from 2000 to 2011 to investigate the statistical properties of price limit hits and the dynamical evolution of several important financial variables before stock price hits its limits . we compare the properties of up - limit hits and down - limit hits . we also divide the whole period into three bullish periods and three bearish periods to unveil possible differences during bullish and bearish market states . to uncover the impacts of stock capitalization on price limit hits , we partition all stocks into six portfolios according to their capitalizations on different trading days . we find that the price limit trading rule has a cooling - off effect ( object to the magnet effect ) , indicating that the rule takes effect in the chinese stock markets . we find that price continuation is much more likely to occur than price reversal on the next trading day after a limit - hitting day , especially for down - limit hits , which has potential practical values for market practitioners .

introduction materials and methods results discussion

knot polynomials like the jones polynomial or the homfly polynomial are very good at detecting chirality . as the conway polynomials of a knot and its mirror image are the same , at first glance the conway polynomial seems to be less powerful . nonetheless , the conway polynomial was known to satisfy some interesting properties : if a knot is achiral , then its determinant is a sum of squares ( @xcite ) . moreover , if we assume that the achiral knot @xmath3 is alternating then the leading coefficient @xmath4 is , up to sign , a square of an integer ( @xcite , s ) . kawauchi ( @xcite ) conjectured for arbitrary achiral knots : the conway polynomial @xmath0 of an achiral knot has the splitting property , i.e. @xmath1 for a polynomial @xmath2 with integer coefficients . hartley and kawauchi established in @xcite this conjecture for strongly @xmath5achiral knots and for two - bridged knots . later hartley ( @xcite ) showed that the conjecture holds for arbitrary @xmath6achiral knots . the kawauchi conjecture was also verified for all prime achiral knots of @xmath716 crossings ( see remark 4.3 in @xcite ) and for a large class of hyperbolic achiral knots tested in @xcite . despite these results , we show that the kawauchi conjecture is not true . + in section 2 , we give a counterexample to the kawauchi conjecture . this counterexample is an alternating @xmath8achiral knot of order 4 ( theorem 2.2 ) . + in section 3 , we will show how the conway polynomial of an alternating achiral knot can be related to the decomposition of the knot into its arborescent and polyhedral parts . for this purpose we will use the decomposition of a knot projection into jewels and twisted band diagrams by haseman ( conway ) circles , as described by bonahon - siebenmann . in the appendix we recall the main lines of that decomposition which is carried out in detail in ( @xcite and @xcite ) : to each alternating knot @xmath3 we associate its * structure tree * @xmath9 : if @xmath3 is achiral , the symmetry induces an automorphism @xmath10 on its structure tree @xmath9 which has exactly one fixed point ( @xcite ) . an achiral alternating knot @xmath3 is * quasi - polyhedral * if the fixed point of @xmath11 is either a jewel or a haseman circle adjacent to two jewels . the achiral alternating knot @xmath3 is * quasi - arborescent * if the fixed point of @xmath11 is a haseman circle adjacent to two twisted band diagrams . we show that achiral quasi - arborescent knots always satisfy the splitting property ( theorem 3.3 ) . hence counterexamples such as the one given in section 2 belong to the class of quasi - polyhedral knots . more generally one can be interested in non - necessarily alternating knots and raise the question : what is among such knots , the nature of counterexamples to kawauchi conjecture . let us recall the definitions of @xmath12 achirality of knots . a knot @xmath3 is _ achiral _ if there is a diffeomorphism @xmath13 of @xmath14 such that @xmath15 and @xmath13 reverses the orientation of @xmath14 . + @xmath3 is _ positively achiral _ if @xmath13 preserves the orientation of @xmath3 . it is _ negatively achiral _ if @xmath13 reverses the orientation of @xmath3 . define a polynomial @xmath16 with coefficients in a unique factorization domain @xmath17 to be even if @xmath18 . assume that char@xmath17 @xmath19 . to be even is equivalent to having only monomials of even order . let @xmath20 $ ] be even . if we substitute @xmath21 for @xmath22 in its decomposition in irreducible elements , we see that this decomposition must be ( up to a unit of @xmath17 ) as follows : @xmath23 where : + 1 ) the polynomials @xmath24 and @xmath25 are irreducible ; + 2 ) @xmath26 if @xmath27 and @xmath28 if @xmath29 ; + 3 ) the polynomials @xmath24 are even and @xmath25 non even . + it is known that the conway polynomial of a knot @xmath30 is even with constant term equal to @xmath31 ( @xcite ) . hence one has the following proposition . the conway polynomial @xmath0 of a knot satisfies the kawauchi conjecture if and only if all the exponents @xmath32 in its decomposition into irreducible factors are even . [ figure 1 ] there exists an alternating achiral knot @xmath3 in @xmath14 such that its conway polynomial is equal to : @xmath33 where @xmath34 and @xmath16 is irreducible over @xmath35 . once the irreducibility of @xmath16 is established , the knot @xmath3 is clearly a counterexample to the kawauchi conjecture . \1 ) let us consider the alternating knot @xmath3 pictured in figure 1 . it is rather easy to realize that @xmath3 is positively achiral ( the alternating projection of @xmath3 is invariant by a rotatory reflection of order 4 ) . by knotscape , its conway polynomial is : @xmath36 2 ) it remains to show that the polynomial @xmath37 is irreducible over @xmath35 . + first we decompose @xmath16 over the field @xmath38 . this decomposition gives information on the possible irreducible factors of @xmath16 , from which we can deduce that @xmath16 is indeed irreducible over @xmath35 . on @xmath38 the polynomial @xmath16 is equal to : @xmath39 it is easy to see that over @xmath38 : @xmath40 + where @xmath41 is irreducible over @xmath38 . + hence @xmath16 either is irreducible or decomposes as a product @xmath42 where @xmath43 is irreducible non even and with reduction modulo 3 equal to @xmath44 . we show that the latter case does not occur . + write @xmath45 . the comparison of @xmath42 with @xmath16 produces the following equations for @xmath46 : @xmath47 a little computation shows that the coefficients @xmath46 are all divisible by 4 . write @xmath48 , @xmath49 , @xmath50 . + then the equations ( 2.1 ) , ( 2.2 ) and ( 2.3 ) give rise to the following equations : @xmath51 + the equation ( 2.6 ) implies that @xmath52 is even and hence this contradicts ( 2.5 ) . the above counterexample satisfies the following conjecture of j.conant ( @xcite ) . let @xmath3 be an achiral knot and @xmath0 its conway polynomial . then there exists a polynomial @xmath53 $ ] such that : @xmath54\ ] ] let us briefly recall some relations between the conway polynomial and some topological properties of knots . + 1 ) arborescence is not detected by the conway polynomial . let us recall that if @xmath3 is a knot then its conway polynomial is an even polynomial with constant term equal to @xmath31 . under these constraints , every polynomial can be realized as the conway polynomial of a knot , or even that of an arborescent ( and hyperbolic ) knot . [ figure 2 t ] ( @xcite ) . for a polynomial @xmath55 , there exists an arborescent knot which admits @xmath0 as its conway polynomial . let us consider the arborescent knot which has its weighted planar tree depicted in figure 2 . one can show that its conway polynomial is equal to : @xmath56 + the proof is done by induction . \2 ) the fact that a knot is alternating is reflected in the alternating form of its alexander polynomial ( another version of the conway polynomial ) : all of its coefficients are non zero and alternate in signs . however there exist alternating alexander polynomials which are not associated to an alternating knot ( @xcite ) + 3 ) it is not known whether there is an achiral knot associated to each conway polynomial satisfying the splitting property . however for the special case where @xmath57 and @xmath2 is an even polynomial with @xmath58 , e. flapan has shown that there exists a prime strongly + achiral knot which has @xmath0 as conway polynomial ( @xcite ) . + let us now focus on the class of achiral alternating knots . surprisingly the conway polynomial proves to be more powerful than one would expect . for this purpose , we will use our definitions and results in ( @xcite ) ( see also the appendix of this present paper ) . let @xmath3 be an achiral quasi - arborescent knot . then there exists a polynomial @xmath2 with integer coefficients such that @xmath0 is equal to @xmath59 if k is @xmath6achiral , by @xcite one has the result . if @xmath3 is @xmath8achiral and quasi - arborescent , then @xmath3 can be muted in a quasi - arborescent knot which is @xmath6achiral ( in fact with a minimal projection of type @xmath60 ) . as mutants have the same conway polynomial ( see for instance [ @xcite ) , the theorem is proved . in the class of alternating knots , counterexamples to the kawauchi conjecture are necessarily quasi - polyhedral . \1 ) the dasbach - hougardy knot ( @xcite ) is an arborescent knot which has a mutant which is @xmath6achiral as described by theorem 5 in @xcite . although these two knots have the same conway polynomial , they are distinguished by the fact that the dasbach - hougardy knot does not satisfy the kauffman conjecture as shown in @xcite . + 2 ) are @xmath8achiral alternating knots which satisfy the splitting property for their conway polynomial quasi - arborescent ? + 3 ) can the arborescence and polyhedral structures in the class of alternating achiral knots be detected by other polynomial invariants ? following bonahon - siebenmann , we decompose canonically in ( @xcite ) any link projection @xmath61 which is connected and prime into diagrams called jewels and twisted band diagrams . the decomposition is performed along haseman ( conway ) circles , i.e. circles which intersect @xmath61 transversely in four points and which satisfy an incompressibility condition . the decomposition of @xmath61 is partially coded by a tree . its vertices represent the diagrams of the decomposition and its edges the haseman circles . any minimal projection of a prime alternating knot @xmath3 satisfies the required conditions . moreover the tree depends only on @xmath3 and not on a particular minimal projection . we call it the * structure tree * of @xmath3 and denote it by @xmath9 . an alternating knot is called * arborescent * if all vertices of @xmath9 are twisted band diagrams and is called * polyhedral * if all vertices are jewels . if @xmath3 is alternating and achiral , the symmetry induces an automorphism @xmath62 we prove in @xcite that : + 1 ) @xmath10 has exactly one fixed point ; + 2 ) the fixed point corresponds either to an invariant jewel or to a haseman circle @xmath63 ; + 3 ) in the latter case , @xmath63 bounds either two jewels or two twisted band diagrams . these jewels or twisted band diagrams are exchanged by @xmath10 . in the case 3 ) , one has theorem 5 in @xcite : let @xmath3 be an oriented @xmath64achiral knot . suppose that @xmath3 has a projection with an invariant haseman circle @xmath63 . then up to a global change of orientation , @xmath3 admits a minimal projection of type @xmath65 or type @xmath60 as shown in figures 5 and 6 . we thank le fonds national suisse de la recherche scientifique for its support . + we thank franoise michel for useful conversations . f. bonahon , l. siebenmann : new geometric splittings of classical knots , and the classification and symmetries of arborescent knots . available on internet http://www-ref.usc.edu/fbonahon/research/preprints . j. conant : chirality and the conway polynomial . spring topology and dynamical systems conference . topology proc . 30 ( 2006 ) , no . 1 , 153 - 162 . p. cromwell : knots and links . cambridge university press , cambridge ( 2004 ) . o. dasbach , s. hougardy : a conjecture of kauffman on amphicheiral alternating knots jour . knot theory and its ramif . 5(1996 ) , 629 - 635 . n. ermotti , c.v . quach hongler , c.weber : a proof of tait s conjecture on alternating @xmath6achiral knots . arxiv:1103.3203 [ math.gt ] . e. flapan : a prime strongly positive amphicheiral knot which is not slice . phil . soc . ( 1986 ) , no . 100 , 533 - 537 . l. goeritz : knoten und quadratische formen . z. 36 ( 1933 ) , 647 - 654 . r. hartley : invertible amphicheiral knots . math . ( 1979/1980 ) , no . 2 , 103 - 109 . r. hartley , a. kawauchi : polynomials of amphicheiral knots . 243(1979 ) , 63 - 70 . l. kauffman : the conway polynomial . ( 1981 ) , 101 - 108 . a. kawauchi : h - cobordism . i. the groups among three dimensional homology handles . osaka j. math . 13 ( 1976 ) , no . 3 , 567 - 590 . l. lyubich , k. murasugi : on zeros of the alexander polynomial of an alternating knot . arxiv:1102.0701 [ math.gt ] . quach hongler : links with unlinking number one and conway polynomials . j. knot theory ramifications 8 ( 1999 ) , no . 7 , 887 - 900 . c.v . quach hongler , c. weber : on the topological invariance of murasugi special components of an alternating link . cambridge philos . ( 2004 ) , no . 1 , 95 - 108 . c.v . quach hongler , c. weber : link projections and flypes . acta math . 33(2008 ) , 433 - 457 . a. stoimenow : square numbers and polynomial invariants of achiral knots . math . z. 255 ( 2007 ) , no . 4 , 703 - 719 .

we give a counterexample to the kawauchi conjecture on the conway polynomial of achiral knots which asserts that the conway polynomial @xmath0 of an achiral knot satisfies the splitting property @xmath1 for a polynomial @xmath2 with integer coefficients . we show that the bonahon - siebenmann decomposition of an achiral and alternating knot is reflected in the conway polynomial . more explicitly , the kawauchi conjecture is true for quasi - arborescent knots and counterexamples in the class of alternating knots must be quasi - polyhedral .

introduction a counterexample the conway polynomial of achiral alternating knots remarks and questions appendix: the structure tree and its automorphism acknowledgments

as long as the intrinsic bulk - inversion asymmetry of the zinc - blende crystal structure is not taken into account , the six - band kane model hamiltonian reads @xmath86 with the expression of the hamiltonian subblocks @xmath87 listed in table [ tab : table1 ] . they are expressed in terms of the usual pauli matrices @xmath88 , the @xmath89 angular momentum matrices @xmath90 , @xmath91 , @xmath92 with @xmath93 the identity matrix and the following @xmath94 matrices : @xmath95 the parameters @xmath96 describe the coupling to remote bands and are , as well as @xmath97 , material - specific parameters whose values are reported in table [ tab : table1 ] we have considered for simplicity the axial approximation @xmath98 with the warping parameter @xmath99 in order to make the bulk band structure isotropic in the @xmath100 plane . .expressions of the kane model in the axial approximation . here @xmath101 denotes the anticommutator for the @xmath102 operators , @xmath103 cyclic permutations of the preceding term and we defined @xmath104 with @xmath105 the free electron mass . we also list the band structure parameters for hgte of ref.@xcite . [ cols="^,^,^",options="header " , ] [ tab : table2 ] effects of strain can be taken into consideration by applying the formalism of bir and pikus . they lead to additional terms in the eight - band kane model hamiltonian proportional to the strain tensor @xmath106 and expressed in terms of the conduction and valence band deformation potentials @xmath107 [ see table[tab : table2 ] ] . at the @xmath19 , @xmath20 , and @xmath21 surfaces . ] we recall that the ( first ) chern number of a generic multi - band system is defined as an integral over a pseudo-2-form @xmath108 where @xmath109 is the index of a non - degenerated band , @xmath110 the related projector , @xmath111 is the exterior derivative , @xmath112 is the wedge product and @xmath113 is the momentum manifold , which corresponds to the one - point compactified infinite momentum plane @xmath114 for the present long - wavelength continuum theory . in the presence of a full band gap in which the fermi energy lies , the sum over all occupied bands @xmath115 is a well defined integer topological quantity even in the presence of degeneracies of the bands over the manifold @xmath113 . at the mirror invariant planes @xmath46 the kane model hamiltonian commutes with the corresponding mirror operators . the corresponding hamiltonians for the states with @xmath70 mirror parity can be expanded in terms of the @xmath116 matrices as @xmath117 where @xmath118 are the @xmath119 gell - mann matrices with @xmath120 the identity . in addition the @xmath111 s define an eight - dimensional vector in momentum space . the projector for each band can be represented by @xmath121 where @xmath109 enumerates the bands from @xmath122 to @xmath123 . the appearing @xmath124 defines a point on the unit sphere in an eight - dimensional real euclidean space restricted by the condition @xmath125 where @xmath126 is the @xmath116 star product . the vector @xmath124 lives in a four dimensional orbit of a @xmath127 sphere which corresponds to an element of the complex projective plane @xmath128 . the relation between @xmath124 and the normalized @xmath129 is found out to be @xmath130 where @xmath131 eqs.(5),(6),(7 ) provide an explicit expression for the mirror chern numbers of the kane model hamiltonian .

the opening of a band gap due to compressive uniaxial strain renders bulk hgte a strong three - dimensional topological insulators with protected gapless surface states at any surface . by employing a six - band @xmath0 model , we determine the spin textures of the topological surface states of strained hgte using their close relations with the mirror chern numbers of the system and the orbital composition of the surface states . we show that at surfaces with @xmath1 point group symmetry an increase in the strain magnitude triggers a topological phase transition where the winding number of the surface state spin texture is flipped while the four topological invariants characterizing the bulk band structure of the material are unchanged . [ [ introduction ] ] introduction + + + + + + + + + + + + + + + topological insulators ( tis ) are new quantum states of matter whose theoretical prediction and experimental verification has had a tremendous impact in the field of fundamental condensed matter physics @xcite , and for potential applications in spintronics and quantum computation @xcite . time - reversal ( tr ) invariant tis are insulating in the bulk but they do possess gapless surface states topologically protected by tr symmetry @xcite . these metallic surface states are spin - momentum locked : surface electrons with opposite spins counterpropagate at the sample boundaries . for three - dimensional ( 3d ) tis , bi@xmath2se@xmath3 @xcite , bi@xmath4rh@xmath3i@xmath5 @xcite , and @xmath6-hgs @xcite to name but a few , the existence of these topological surface states ( tss ) can be directly inferred from the four @xmath7 indices characterizing the bulk band structure of a 3d tr invariant insulator @xcite . however , both the strong @xmath8 index and the three weak @xmath9 indices make no assertion on the nature of the spin textures of the surface states which , realizing a vortex structure in momentum space , can be characterized topologically by the winding number ( the topological charge of the vortex ) of the planar unit spin @xmath10 . it is defined by @xmath11,\ ] ] where @xmath12 is a closed loop in momentum space encircling the essential degeneracy point of the topological surface state , guaranteed by tr invariance . generally speaking , the two winding numbers @xmath13 [ c.f . fig . [ fig : fig0 ] ] are equally compatible for linear dirac cones , and the specific value is independent of the @xmath7 topological indices of the bulk band structure . in many strong 3d tis with a single dirac cone on the surface , however , additional point group symmetries at the surfaces pin the spin texture winding number to @xmath14 . this occurs at the high @xmath15 @xcite symmetry surfaces of materials with a rhombohedral crystal structure such as bi@xmath2se@xmath3 , or at the @xmath16 point group symmetric surfaces @xcite of cubic materials such as @xmath6-hgs @xcite . for surfaces where the symmetry is lowered , a similar assertion can not be made . . in the right panel @xmath17 . ] the aim of this letter is to show that for @xmath18 point group symmetric surfaces , the surface state spin textures of compressively strained bulk hgte a strong 3d ti whose non - trivial topological properties have been experimentally verified by quantum hall measurements @xcite have a topological charge that can be flipped from @xmath14 to @xmath17 by continuously increasing the strain magnitude . we will use the close relations @xcite connecting the winding number of the spin textures , the mirror chern numbers of the system and the orbital composition of the topological surface states , and thereby demonstrate , within a six - band @xmath0 kane model , that for a uniaxial strain along the @xmath19 direction , the orbital character of the topological surface states at the @xmath20 and @xmath21 surfaces depends sensitively upon the strain magnitude . this ultimately leads to a change in the nature of the surface state spin textures at a critical strain magnitude . [ [ topological - surface - state - dirac - points ] ] topological surface state dirac points + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + pristine hgte is a zero gap semiconductor with the fermi energy in the middle of the fourfold degenerate light - hole ( lh ) and heavy - hole ( hh ) @xmath22 states at the bz center @xcite . the topological nature of the electronic states in this material can not be inferred from these @xmath23 atomic levels @xcite but rather follows from the inverted band ordering at the zone center of the lh @xmath22 band , which is particle - like , and the @xmath24 @xmath25-band , which is hole - like . in normal semiconductors , such as cdte , the @xmath24 band forms the conduction band while the lh @xmath22 band represents one of the valence bands . this inverted band ordering , which is an immediate consequence of the strong spin - orbit coupling of hg , establishes this material to be topologically non - trivial since two bands of opposite parities have level crossed with respect to the normal band ordering . by externally applying a compressive uniaxial strain , the fourfold degeneracy of the @xmath22 states at the zone center is lifted and thus a gap at the fermi energy opens up @xcite . in addition , the parity eigenvalues of the occupied bands are unchanged , which thereby establishes compressively strained hgte as a strong 3d ti . the bulk - boundary correspondence @xcite then guarantees the existence of tss with a conical dispersion at any surface , and the surface kramer s doublet the dirac point sitting at the surface bz center . to verify this , we rely on an effective low - energy theory based upon a @xmath0 expansion of the lowest energy bands around the @xmath26 point of the bz . this approach has successfully described the quantum spin hall effect in hgte / cdte quantum wells @xcite . we thus employ the six - band kane model for the @xmath27 bands @xcite , where the influence of a compressive uniaxial strain , which , without loss of generality , we assume along the @xmath28 direction , is taken into account via the bir - pikus hamiltonian [ see the supplemental material ] . the luttinger and @xmath0 parameters are based on the @xmath29 band structure of pristine hgte @xcite . we can establish the presence and the electronic characteristic of the surface kramer s doublet at the @xmath19 , @xmath20 and @xmath21 surfaces by solving the @xmath0 model at the surface bz center in the half - infinite space @xmath30 , @xmath31 , and @xmath32 respectively , using the general method outlined in ref . . at the @xmath19,@xmath20 , and @xmath21 surfaces ( red lines ) . we also show the conduction and valence band edges as well as the @xmath24 band edge ( gray lines ) . at the @xmath19 surface , the surface dirac point is buried within the hh valence band but resides in the bandgap of the lh-@xmath24 ti bulk . ] fig . [ fig : fig1](a ) shows the behavior of the surface dirac point energy @xmath33 as a function of the uniaxial strain magnitude @xmath34 renormalized by the factor @xmath35 where @xmath36 is the distance of the @xmath24 band edge from the direct bz center midgap while @xmath37 is the deformation potential of hgte . at the @xmath19 surface , the surface kramer s doublet is buried within the hh valence band , while on the two other surfaces it resides in the indirect bulk gap of the system . this termination dependence is also reflected in the behavior of the penetration depth of the surface states [ see the supplemental material ] . specifically , the surface states at the @xmath20 and @xmath21 surfaces are characterized by a diverging decay length in the @xmath38 limit , which implies that at these planes the tss penetrate more deeply into the bulk as compared to the @xmath19 tss . these different electronic characteristics can be attributed to the different nature of the dirac wavefunction of the surface bz center . at the @xmath19 plane , indeed , the surface state dirac wavefunction is all made of lh and @xmath24 states . a uniaxial strain along the @xmath28 direction preserves the axial rotation symmetry in the plane , and thus at the surface @xmath26 point with momentum @xmath39 the total angular momentum @xmath40 is a good quantum number @xcite . this , in turn , implies the absence of any mixing between the @xmath41 hh states and the @xmath42 lh and @xmath24 states . henceforth , the hh bands play the role of inserted parasitic " bands @xcite on top of the lh-@xmath24 ti bulk , in the bandgap of which the surface dirac point resides [ c.f . fig . [ fig : fig1 ] ] . this does not hold true at the @xmath20 and @xmath21 planes where the uniaxial strain along the @xmath28 direction breaks the in - plane rotation symmetry , thereby leading to an effective hybridization between the @xmath43 states with @xmath44 hh states . the surface dirac wavefunction becoming a superposition of @xmath27 localized states is then pushed out of the hh bulk bandwidth @xcite and remerges in the full bandgap of the system , in agreement with the features encountered in the fano model @xcite . direction at the @xmath19 ( top panels ) , @xmath20 ( middle panels ) and @xmath21 ( bottom panels ) crystal planes . ( b ) physical spin textures for strain smaller than the critical one @xmath45 . for larger strain the physical spin textures corresponds to the pseudospin ones . ] [ [ mirror - chern - numbers - and - pseudospin - textures ] ] mirror chern numbers and pseudospin textures + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + to proceed further , we now introduce the notion of mirror chern numbers ( mcn ) . the mcn are topological invariants , which are protected by mirror symmetries . in the absence of strain and neglecting the bulk inversion asymmetry of the zincblende crystal structure , hgte has nine mirror planes and correspondingly the kane model hamiltonian is invariant under these symmetry operations . the presence of an uniaxial strain along the @xmath19 direction reduces the number of mirror planes but preserves the mirror symmetry with respect to the @xmath19 , @xmath20 , and @xmath21 planes . since the kane model hamiltonian commutes with the corresponding mirror symmetry operations at the three planes @xmath46 , all eigenstates can be classified according to their @xmath47 mirror parity . this allows to define two time - reversal related chern numbers @xmath48 whose sum vanishes but with a difference @xmath49 , which is an integer @xmath50 topological invariant and defines the mcn . we have computed the mcns of the full six - band kane model hamiltonian using its decomposition at the mirror planes in terms of the nine gell - mann matrices , and subsequently employed the elegant formulation of ref . to derive the chern numbers for the corresponding continuum models . this allows us to avoid an effective two - band modelling which can only be introduced _ ad hoc_. we find a mcn @xmath51 at the @xmath52 planes , whereas @xmath53 at the @xmath54 plane . and indeed , under proper coordinate transformations , the continuum @xmath0 hamiltonians at the @xmath52 planes can not be adiabatically transformed into the @xmath54 one without closing the bulk band gap . the three mcns indicated above allow us to immediately derive a _ pseudospin _ texture for the tss at the @xmath19 , @xmath20 and @xmath21 planes , as explained below . at the @xmath19 surface , the projection of the two unbroken mirror planes @xmath20 , @xmath21 define two mirror invariant lines where the tss can be classified according to their mirror eigenvalues . we can thus define a pseudospin vector @xmath55 with components related to the mirror operators @xmath56 by @xmath57 . the dispersion of the tss can be then written in terms of this pseudospin as @xmath58 where the sign of the two fermi velocities @xmath59 is uniquely determined by the mcns of the system via the bulk - edge correspondence for the mirror invariant planes @xmath52 . specifically we have @xmath60 which yields the pseudospin texture shown in the top panel of fig . [ fig : fig2](a ) . it exhibits an helical structure with a left - handed helicity for the surface state conduction band , and a right - handed one for the valence band , in perfect agreement with density functional theory studies @xcite . at the opposite @xmath61 surface the sign of both two fermi velocities are flipped , which changes the helicity of the pseudospin texture but still preserves the pseudospin texture winding number @xmath62 . a similar analysis at the @xmath20 and @xmath21 surface terminations yields the pseudospin textures shown in fig . [ fig : fig2](a ) . at these surfaces the different values of the two mcn for the mirror invariant planes yield a pseudospin texture with an opposite winding number @xmath17 . in the tss of strained hgte as a function of the strain magnitude @xmath34 . the tss at the @xmath20 and @xmath21 planes exhibit an anisotropic behavior since the uniaxial strain along the @xmath28 direction lower the point group symmetry at that surfaces to @xmath63 . ] fig . [ fig : fig3 ] shows the behavior of the magnitude of the tss fermi velocities @xmath64 as a function of the strain magnitude @xmath34 . we find that for the tss at the @xmath19 plane , the two fermi velocities @xmath65 have equal magnitudes . therefore , the tss display a global @xmath66 rotational symmetry similarly to the case of for instance bi@xmath2se@xmath3 @xcite . as a two - dimensional @xmath67 for the pseudospin one - half surface kramer doublet explicitly shows @xcite , this is an immediate consequence of the fourfold rotational symmetry along the @xmath28 axis . and indeed , at the @xmath20 and @xmath21 planes where the uniaxial strain along the @xmath21 direction lowers the surface point group symmetry from @xmath68 to @xmath63 , we find the fermi velocities of the tss to differ , with an anisotropy that is enhanced by increasing the strain magnitude . [ [ spin - textures ] ] spin textures + + + + + + + + + + + + + + + the pseudospin textures of the tss have a close relation to the physical spin textures . this can be found introducing the projector operator @xmath69 onto the subspaces of the kane model hamiltonian with mirror parity @xmath70 , and considering their effect on the physical spin operators @xmath71 at the mirror planes . it can be shown that @xmath72 for @xmath73 , which simply states that at the surface mirror invariant lines , the physical spin can be either parallel or antiparallel to the pseudospin . for smb@xmath74 , the physical spin was found to be always parallel to the pseudospin , and thus the knowledge of the mcns provides us a robust classification of the topological surface state spin textures @xcite . this , however , does not hold true for hgte . the projected spin operator of the @xmath27 bands has both positive and negative eigenvalues , and thus the relation between the pseudospin and the physical spin depends on the orbital composition of the surface states . by evaluating the projected spin operator eigenvalues of the tss along the mirror invariant lines , we find that the physical spin is always parallel to the pseudospin except at the @xmath75 and @xmath76 mirror invariant lines for the @xmath20 and @xmath21 surface terminations respectively . at these mirror invariant lines , the physical spin is indeed antiparallel to the pseudospin below a critical strain magnitude @xmath77 , and parallel above it . this implies that contrary to the tss pseudospin textures at the @xmath20 and @xmath21 surface terminations , which are characterized by a @xmath17 winding number independent of the strain magnitude , the physical spin textures exhibit a right - handed @xmath14 helical structure for small strain [ c.f . fig . [ fig : fig3](b ) ] . the @xmath14 spin textures equal to the pseudospin textures are then restored for larger strain values . ) for the @xmath20 [ @xmath21 ] plane at the mirror invariant line @xmath75 [ @xmath76 ] . for small strain , the physical spin is antiparallel to the pseudospin , while for large strain they are parallel . ] to gain more insight into this spin texture topological phase transition , we have computed the orbital resolved projected spin eigenvalue of the tss along the mirror invariant lines @xmath52 for the @xmath21 and @xmath20 surface terminations respectively , by varying the strain magnitude [ c.f . fig . [ fig : fig4 ] ] . the tss with positive pseudospin , _ i.e. _ mirror parity eigenvalue @xmath70 , is an admixture of @xmath80 @xmath24 , @xmath81 hh and @xmath82 lh states . for small strain , the tss has a predominant lh orbital character which implies that the physical spin is antiparallel to the pseudospin . by continuously increasing the strain magnitude , the tss starts to acquire a sizeable @xmath24 and hh character which ultimately reverse the spin direction to be parallel to the pseudospin . [ [ conclusions ] ] conclusions + + + + + + + + + + + + + we have classified the topological surface state spin texture of bulk uniaxially strained hgte using the close relations among spin texture winding number , mirror chern numbers and orbital character of the topological surface states . we have shown that assuming a strain along the @xmath28 direction , the spin texture at the @xmath83 point - group symmetric @xmath19 plane exhibit a conventional left - handed helical structure , while considering the @xmath18 symmetric @xmath20 and @xmath21 surface terminations the nature of the spin texture strongly depend on the orbital character of the topological surface states . the topological charge of the spin texture vortex structure can be indeed flipped by increasing the strain magnitude . this phenomenon is entirely due to the strain dependence of the orbital character of the tss and occurs without any bulk bandgap closing - reopening point or change in the bulk mcn values , making such a topological phase transition very different from the ones proposed in for instance smb@xmath74 @xcite and hgte@xmath84s@xmath85 @xcite . we acknowledge the financial support of the future and emerging technologies ( fet ) programme within the seventh framework programme for research of the european commission under fet - open grant number : 618083 ( cntqc ) . this work has been supported by the deutsche forschungsgemeinschaft under grant no . or 404/1 - 1 and sfb 1143 . jvdb acknowledges support from the harvard - mit center for ultracold atoms . 9 c. l. kane and e. j. mele , phys . rev . lett . * 95 * , 146802 ( 2005 ) . c. l. kane and e. j. mele , phys . rev . lett . * 95 * , 226801 ( 2005 ) . l. fu and c. l. kane , phys . rev . b * 74 * , 195312 ( 2006 ) . b. a. bernevig , t. l. hughes , and s .- c . zhang , science * 314 * , 1757 ( 2006 ) . c. wu , b. a. bernevig , and s .- c . zhang , phys . rev . lett . * 96 * , 106401 ( 2006 ) . m. knig , s. wiedmann , c. brne , a. roth , h. buhmann , l. w. molenkamp , x .- l . qi , and s .- c . zhang , science * 318 * , 766 ( 2007 ) . l. fu and c. l. kane , phys . rev . b * 76 * , 045302 ( 2007 ) . l. fu , c. l. kane , and e. j. mele , phys . rev . lett . * 98 * , 106803 ( 2007 ) . j. e. moore and l. balents , phys . rev . b * 75 * , 121306 ( 2007 ) . h. zhang , c .- x . liu , x .- l . qi , x. dai , z. fang , and s .- c . zhang , nat phys * 5 * , 438 ( 2009 ) . d. hsieh , d. qian , l. wray , y. xia , y. s. hor , r. j. cava , and m. z. hasan , nature * 452 * , 970 ( 2008 ) . y. xia , d. qian , d. hsieh , l. wray , a. pal , h. lin , a. bansil , d. grauer , y. s. hor , r. j. cava , and m. z. hasan , nat phys * 5 * , 398 ( 2009 ) . y. l. chen , j. g. analytis , j .- h . chu , z. k. liu , s .- k . mo , x. l. qi , h. j. zhang , d. h. lu , x. dai , z. fang , s. c. zhang , i. r. fisher , z. hussain , and z .- x . shen , science * 325 * , 178 ( 2009 ) . d. hsieh , y. xia , l. wray , d. qian , a. pal , j. h. dil , j. osterwalder , f. meier , g. bihlmayer , c. l. kane , y. s. hor , r. j. cava , and m. z. hasan , science * 323 * , 919 ( 2009 ) . b. rasche , a. isaeva , m. ruck , s. borisenko , v. zabolotnyy , b. bchner , k. koepernik , c. ortix , m. richter , and j. van den brink , nat mater * 12 * , 422 ( 2013 ) . m. z. hasan and c. l. kane , rev . mod . phys . * 82 * , 3045 ( 2010 ) . x. -l . qi and s. -c . zhang , rev . mod . phys . * 83 * , 1057 ( 2011 ) . a. r. akhmerov , j. nilsson , and c. w. j. beenakker , phys . rev . lett . * 102 * , 216404 ( 2009 ) . f. virot , r. hayn , m. richter , and j. van den brink , phys . rev . lett . * 106 * , 236806 ( 2011 ) . f. virot , r. hayn , m. richter , and j. van den brink , phys . rev . lett . * 111 * , 146803 ( 2013 ) . l. fu , phys . rev . lett . * 103 * , 266801 ( 2009 ) . c .- x . liu , x .- l . qi , h. zhang , x. dai , z. fang , and s .- c . zhang , phys . rev . b * 82 * , 045122 ( 2010 ) . c. ortix , j. w. f. venderbos , r. hayn , and j. van den brink , phys . rev . b * 89 * , 121408(r ) ( 2014 ) . in the remainder , the effect of the bulk inversion asymmetry in the zincblende crystal structure of mercury chalcogenides will be considered as a small perturbation @xcite , and thus neglected . c. brne , c. x. liu , e. g. novik , e. m. hankiewicz , h. buhmann , y. l. chen , x. l. qi , z. x. shen , s. c. zhang , and l. w. molenkamp , phys . rev . lett . * 106 * , 126803 ( 2011 ) . m. legner , a. regg , and m. sigrist , phys . rev . lett . * 115 * , 156405 ( 2015 ) . r. -l . chu , w. -y . shan , j. lu , and s. -q . shen , phys . rev . b * 83 * , 075110 ( 2011 ) . r. winkler , _ spin - orbit coupling effects in two - dimensional electron and hole systems _ ( springer , berlin , 2003 ) . e. g. novik , a. pfeuffer - jeschke , t. jungwirth , v. latussek , c. r. becker , g. landwehr , h. buhmann , and l. w. molenkamp , phys . rev . b * 72 * , 035321 ( 2005 ) . d. l. bergman and g. refael , phys . rev . b * 82 * , 195417 ( 2010 ) . u. fano , phys . rev . * 124 * , 1866 ( 1961 ) . r. barnett , g. r. boyd , and v. galitski , phys . rev . lett . * 109 * , 235308 ( 2012 ) . q .- z . wang , s .- c . wu , c. felser , b. yan , and c .- x . liu , phys . rev . b * 91 * , 165435 ( 2015 ) . t. rauch , s. achilles , j. henk , and i. mertig , phys . rev . lett . * 114 * , 236805 ( 2015 ) .

kane model hamiltonian evaluation of the mirror chern numbers

alternatively , we can measure the global many - body purity , and observe a static , high purity . this is in stark contrast to the vanishing global purity of the canonical thermal ensemble , yet this same ensemble accurately describes the local number distribution we observe . * ( c ) * to measure the atom number locally , we allow the atoms to expand in half - tubes along the @xmath0-direction , while pinning the atoms along @xmath1 . in separate experiments , we apply a many - body beam splitter by allowing the atoms in each column to tunnel in a projected double - well potential . the resulting atom number parity , even or odd , on each site encodes the global and local purity . ] for our experiments , we utilize a bose - einstein condensate of @xmath2rb atoms loaded into a two - dimensional optical lattice that lies at the focus of a high resolution imaging system @xcite . the system is described by the bose - hubbard hamiltonian , @xmath3 where @xmath4 , @xmath5 , and @xmath6 are the bosonic creation , annihilation , and number operators at the site located at @xmath7 , respectively . atoms can tunnel between neighboring lattice sites at a rate @xmath8 and experience a pairwise interaction energy @xmath9 when multiple atoms occupy a site . we have independent control over the tunneling amplitudes @xmath10 and @xmath11 through the lattice depth , which can be tuned to yield @xmath12 to @xmath13 . in addition to the optical lattice , we are able to superimpose arbitrary potentials using a digital micromirror device ( dmd ) placed in the fourier plane of our imaging system @xcite . to initiate the experiment , we isolate a @xmath14 plaquette from a larger low - entropy mott insulator with unity filling as shown in figure [ fig : protocol]a @xcite . at this point , each system is in a product state of single - atom fock states on each of the constituent sites . we then suddenly switch on tunneling in the @xmath1-direction while the @xmath0-direction tunneling is suppressed . each chain is restricted to the original six sites by introducing a barrier at the ends of the chains to prevent tunneling out of the system . these combined steps quench the six - site chains into a hamiltonian for which the initial state represents a highly excited state that has significant overlap with an appreciable number of energy eigenstates . each chain represents an identical but independent copy of a quenched system of six particles on six sites , which evolves in the quenched hamiltonian for a controllable duration . in the data that follow , we realize measurements of the quantum purity and on - site number statistics ( figure [ fig : protocol]c ) . for measurements of the former , we append to the quench evolution a beam splitter operation that interferes the two identical copies by freezing dynamics along the chain and allowing for tunneling in a projected double - well potential for a prescribed time @xcite . in the last step for both measurements , a potential barrier is raised between the two copies and a one - dimensional time - of - flight in the direction transverse to the chain is performed to measure the resulting occupation on each site of each copy . the ability to measure quantum purity is crucial to assessing the role of entanglement in our system . tomography of the full quantum state would typically be required to extract the global purity , which is particularly challenging in the full 462-dimensional hilbert space defined by the itinerant particles in our system . furthermore , while in spin systems global rotations can be employed for tomography @xcite , there is no known analogous scheme for extracting the full density matrix of a many - body state of itinerant particles . the many - body interference described here , however , allows us to extract quantities that are quadratic in the density matrix , such as the purity @xcite . after performing the beam splitter operation , we can obtain the quantum purity of the full system and any subsystem simply by counting the number of atoms on each site of one of the six - site chains ( figure [ fig : protocol]c ) . each run of the experiment yields the parity @xmath15 , where @xmath16 is iterated over a set of sites of interest in copy-@xmath17 . the single - site parity operator @xmath18 returns 1 ( -1 ) when the atom number on site-@xmath16 is even ( odd ) . it has been shown that the beam splitter operation yields @xmath19 , where @xmath20 is the density matrix on the set of sites considered for each copy @xcite . because the preparation and quench dynamics for each copy are identical , yielding @xmath21 , the average parity reduces to the purity : @xmath22 . when the set of sites considered comprises the full six - site chain , the expectation value of this quantity returns the global many - body purity , while for smaller sets it provides the local purity of the respective subsystem . by studying measurements with and without the beam splitter , our data immediately illustrates the contrast between the global and local behavior and how thermalization is manifest ( figure [ fig : protocol]b ) . we observe that the global many - body state retains its quantum purity in time , affirming the unitarity of its evolution following the quench . this global measurement also clearly distinguishes the quantum state we produce from a canonical thermal ensemble with orders of magnitude smaller purity . yet , we observe that the number statistics locally converge to a distribution of thermal character , which can be faithfully modeled by that same thermal ensemble . in what follows , we experimentally explore the question suggested by this observation : how does a pure state that appears globally distinct from a thermal ensemble possess local properties that mirror this thermal state ? the growth of entanglement following a quench is key to understanding how entropy forms within the subsystems of a pure quantum state , thereby facilitating thermalization @xcite . when two parts of a system are entangled , the full quantum state @xmath23 can not be written in a separable fashion with respect to the hilbert spaces of the subsystems @xcite . as has been shown theoretically @xcite and recently observed experimentally @xcite , this causes the subsystems @xmath24 and @xmath25 to be in an entropic mixed state even though the full many - body quantum state is pure @xcite . the mixedness of the subsystem can be quantified by the second - order rnyi entropy @xmath26)$ ] , which is the logarithm of the purity of the subsystem density matrix . while the von neumann entropy is typically used in the context of statistical mechanics , both quantities grow as a subsystem density matrix becomes mixed and increasingly entropic . in the rnyi case , the purity in the logarithm quantifies the number of states contributing to the statistical mixture described by the density matrix . we first study the dynamics of the entanglement entropy immediately following the quench , for varying subsystem sizes ( figure [ fig : eedyn ] ) . initially , we observe an approximately linear rise in the entropy , with similar slope among the subsystems considered ( figure [ fig : eedyn ] inset ) @xcite . after an amount of time that depends on the subsystem size , the entanglement entropy saturates to a steady - state value , about which there are small residual temporal fluctuations . the presence of residual fluctuations are , in part , attributable to the finite - size of our system . an exact numerical calculation of the dynamics with no free parameters shows excellent agreement with our experimental measurements . crucially , the data indicate that while the subsystems acquire entropy in time ( figure [ fig : eedyn],a - c ) , the full system entropy remains constant and is small throughout the dynamics ( figure [ fig : eedyn]d ) . the high purity of the full system allows us to conclude that the dynamical increase in entropy in the subsystems originates in the propagation of entanglement between the system s constituents . the approximately linear rise at early times ( figure [ fig : eedyn ] inset ) is related to the spreading of entanglement in the system within an effective light cone @xcite . furthermore , in analogy to the growth of thermodynamic entropy in an equilibrating classical mechanical system , such as a gas in a closed container , we observe the growth of local entropy in a closed quantum mechanical system . in the quantum mechanical case , however , the mechanism responsible for entropy is entanglement , which is absent from a system modeled by classical mechanics . when a system thermalizes , we expect that the saturated values of local observables should correspond to the predictions of a statistical ensemble . by analogy , if the entanglement entropy plays the role of thermal entropy , then in a thermalized pure state we expect extensive growth in the entanglement entropy with subsystem volume . when the entanglement entropy in a quantum state grows linearly with the size of the subsystem considered , it is known as a volume law . ground - breaking theoretical work using conformal field theory has shown that indeed , at long times , a volume law is expected for a quenched , infinite , continuous system , while only an area law with a log correction is expected for the ground state @xcite . characterizing the large amount of entanglement associated with a volume law is particularly challenging because it results in nearly every entry of the density matrix having small , but importantly non - zero magnitude . using the techniques outlined in this work , we show measurements displaying a near volume law in the entanglement entropy ( figure [ fig : volume]a ) . a linear growth with volume in the entanglement entropy occurs when each subsystem incoherently populates a number of states that scales with the size of the subsystem hilbert space . this is because , for the bose - hubbard model , the hilbert space is approximately exponential in the lattice size , which results in a linear growth in @xmath27)$ ] . furthermore , the exact slope of the entanglement entropy versus subsystem volume depends on the average energy of the thermalized pure state @xcite . by contrast , we can prepare the ground state of the quenched hamiltonian by adiabatically reducing the lattice depth . here , the superfluid ground state of the bose - hubbard model has suppressed entanglement , which is predicted to incur slow logarithmic growth in the entanglement entropy with subsystem volume @xcite . our measurements clearly distinguish the two cases . the back - bending of the entanglement entropy as the subsystem surpasses half the system size indicates that the state is globally pure . in the quenched state , the high global purity is striking in a state that locally appears completely dephased , which is behavior often associated with environmentally - induced decoherence or other noise sources . we further observe near quantitative agreement between the exact dependence of the entanglement entropy with subsystem volume and the prediction of a thermal ensemble . we make this comparison by computing a canonical thermal ensemble @xmath28 with an average energy that is the same as the quenched quantum state produced experimentally @xcite . the gray line in figure [ fig : volume]a is the rnyi ( thermal ) entropy as a function of subsystem size for this calculated thermal state . although our limited system size prevents comparison over a large range of subsystem sizes , the initial rise of the entanglement entropy with subsystem volume mimics that of the thermal entropy . despite their similarity , it is worth emphasizing the disparate character of the thermal and entanglement entropy . the entanglement entropy ( either the rnyi or von neumann ) is instantaneously present in the pure quantum state after coherent unitary evolution , arising from the non - separability of the quantum state between the subsystem and traced out degrees of freedom . on the other hand , the von neumann thermal entropy within a subsystem of a mixed thermal state is the thermodynamic entropy in statistical mechanics , which could be extracted from irreversible heat flow experiments on the subsystem @xcite . therefore , the similarity of the rnyi entropies we observe points to an experimental equivalence between the entanglement and thermodynamic entropy @xcite . the behavior of the entanglement entropy provides a clean framework for understanding the entropy within thermalizing , closed quantum systems . however , one of the most famous features of entanglement , the presence of non - local correlations , appears inconsistent with what one expects of thermalized systems . in particular , the massive amount of entanglement implied by a volume law suggests a large amount of correlation between disparate parts of the system , while a key feature of a thermal state is the very absence of these long - range correlations . a useful metric for correlations , both classical ( statistical ) and quantum , between two subsystems @xmath29 and @xmath30 is the mutual information @xmath31 @xcite . the mutual information demonstrates that the amount of correlation in the presence of a volume law is vanishing for subsystem volumes that sample less than half the full system , which is where the entropy growth is nearly linear ( figure [ fig : volume]b , c ) . furthermore , even though the thermalized quantum state carries more entanglement entropy than the ground state , small subsystems display smaller correlations than in the superfluid ground state . once the subsystem volume is comparable to the system size , which is where the entanglement entropy deviates from the volume law , the quantum correlations entailed by the purity of the full system become apparent ( figure [ fig : volume]c ) . the mutual information therefore illustrates how the volume law in the entanglement entropy yields an absence of correlations between sufficiently local observables , even though the quantum state retains a large amount of entanglement . , the saturated average particle number on each site ( density ) is nearly equal among the sites of the system , which resembles a system at thermal equilibrium . by comparison , the ground state for the same bose - hubbard parameters has significant curvature . * ( b ) * in measuring the probabilities to observe a given particle number on a single site , we can obtain the local , single - site density matrix and observe the approach to thermalization . using two different metrics , we compare the mixed state observed to the mixed state derived from the subsystem of a canonical thermal ensemble , after a quench to @xmath32 . the trace distance provides an effective distance between the mixed states in hilbert space , while the fidelity is an overlap measure for mixed states . the two metrics illustrate how the pure state subsystem approaches the thermal ensemble subsystem shortly after the quench . the starting value of these quantities is given by the overlap of the initial pure state with the thermal mixed state . solid lines connect the data points . ] our comparisons between the entanglement entropy and thermal entropy suggest that the pure quantum state we study possesses thermalized properties . we can further examine the presence of thermalization by performing a series of measurements of local observables against which we compare the predictions of various thermal ensembles . as with the entanglement entropy , we also contrast our observations of the quenched thermalized state with the adiabatically prepared ground state . in figure [ fig : local]a , we display the in - situ number density distribution on the six sites for the saturated quenched state and the ( superfluid ) ground state . while the ground state exhibits significant curvature , the quenched state exhibits a flat density distribution . this flat density distribution is consistent with a picture in which the constituents of the many - body system collectively thermalize , so that each site is in equilibrium with its neighbors and physically similar . we can perform a more rigorous test of single - site thermalization by comparing the measured density matrix of each site with the reduced density matrix of a canonical thermal ensemble @xmath33 ( figure [ fig : local]b ) . our measurements of probabilities to observe a given particle number on a site completely characterize that single - site density matrix , because there are no coherences between different number states due to super - selection rules . with this measured density matrix , we can perform a quantitative comparison to a thermal ensemble using the trace distance ( @xmath34 ) and quantum fidelity ( @xmath35 ) , both of which quantify the similarity of two mixed quantum states . after a short time , we see a quantum fidelity exceeding @xmath36 and a trace - distance that fluctuates between @xmath37 and @xmath38 , indicating the similarity between the local density matrix of a verified pure state with the local density matrix of a thermal state . the correspondence between the observables of a pure state and thermal state depends on the equivalence of their reduced density matrices within the hilbert space sampled by the observable . the measurement of figure [ fig : local]b therefore shows that observables for the single - site hilbert space should agree with the predictions of thermal ensembles . we now focus on direct comparisons of observables with various thermal ensembles , and the theoretical justification for doing so . while we have focused on the role of entanglement entropy in producing thermal characteristics , the eigenstate distribution resulting from a quench ( figure [ fig : eth]a ) determines the dynamics of observables , as well as their subsequent saturated values . it follows then that these populated eigenstates should clarify the origin of thermalization , which is the goal of eth . the underlying explanation for eth is that thermalizing , non - integrable systems possess excited eigenstates that look like nearly random vectors , or , equivalently , are described by a hamiltonian that approximately conforms to random matrix theory @xcite . that is , for most bases , each eigenvector projects onto each basis vector with random quantum amplitude . the diffuse probability distribution of the eigenstates in most bases , such as the fock state basis , is analogous to the chaotic dynamics of a closed classical mechanical system passing through every allowed point of phase space , and in the quantum case this has several consequences . surprisingly , this chaotic assumption can be adapted to explain the saturation of measurement observables , the agreement of these saturated observables with thermal ensembles , and the presence of a volume law in the entanglement entropy @xcite . and so , while in classical mechanical systems it is the chaos in the temporal dynamics that leads to entropy maximization and thermalization within thermodynamic constraints , in quantum thermalizing systems it is chaos in the energy eigenstates that generates the analogous behavior in the entanglement entropy , and , in turn , thermalization . in figure [ fig : eth]c , d , we compare our measurements to the predictions of thermal ensembles that are illustrated in figure [ fig : eth]b . we also compare our results to a grand - canonical ensemble truncated to our total atom number @xcite : this ensemble perhaps most closely models how well the many - body state can act as a reservoir for its constituent subsystems . for each single - site and three - site observable , we show the atom number distributions for two different effective temperatures of @xmath39 and @xmath40 , which are achieved by quenching to @xmath41 and @xmath42 , respectively . the data is averaged in the saturated regime over times between @xmath43 and @xmath44 , and the error bars are the standard deviation in the measured probabilities . the consistency within the the error bars indicates that in this temporal range our observations remain near the thermal predictions despite the presence of temporal fluctuations . for the single site subsystem , the data is in good agreement with all the ensembles considered . despite the fact that the quenched state is in a large distribution of eigenstates , surprisingly , we find favorable agreement for the case of a single eigenstate ensemble : this illustrates a key principle of eth , which holds that the reduced density matrix , and associated observables , vary slowly from eigenstate to eigenstate and are therefore relatively insensitive to breadth of the distribution of populated states from the quench . we perform the same comparisons to the three - site case in the bottom two panels . here we also observe agreement with most ensembles , though , interestingly , there is relatively less agreement with the single eigenstate and grand - canonical ensembles , particularly for the lower temperature quench . this variation in agreement may suggest that these ensembles are more sensitive to the relative size of the traced out reservoir compared to the subsystem , which indicates directions of further experiments @xcite . the above measurements were on specific subsystems , but our measurements also allow extraction of the average global interaction energy ( figure [ fig : eth]d ) . since the interaction term in eq . [ bhm ] is diagonal in the fock - state basis , we can use our measurements of the final particle configurations to compute the expectation value @xmath45 . for the @xmath46 data , we show a time scan indicating the initial growth in this quantity , which starts at zero since the initial state is a single particle per site . these observations , at long times , are in near agreement with the canonical prediction . interestingly , this measurement is sensitive to the entire six - site system as opposed to some subset of sites , which might suggest that it is global and unlikely to thermalize . yet , @xmath45 undergoes thermalization because it is a sum of local operators , each of which thermalizes and is insensitive to the global purity of the full system . the observed agreement is consistent with the idea that only a small set of operators , such as the global purity we measure or other specific fine - tuned state projectors , can truly distinguish the pure state we produce from a thermal state . our observations speak to a natural translation between thermalizing quantum mechanical and classical mechanical systems composed of itinerant particles . classical statistical mechanics relies on a fundamental assumption : a system in thermal equilibrium can be found in any microstate compatible with the thermodynamic constraints imposed on the system , and , as such , is described by an ensemble of maximal entropy @xcite . although it is vastly successful , classical statistical mechanics does not itself justify this entropy maximization for closed systems @xcite , and an open systems approach only defers the question of thermalization to the union of the bath and system @xcite . while ergodicity and time - averaging can provide a justification for entropy maximization in closed classical mechanical systems , ergodicity is not applicable on the same scale that statistical mechanics is successful , and time - averaging can require exponentially long times @xcite . the latter also obscures the fact that there is in reality only one system , which , nevertheless , is well - modeled by an entropic ensemble @xcite . our studies , and beautiful recent theoretical work @xcite , hint towards a microscopic origin for entropy maximization in a single quantum state , namely that induced by the entanglement we measure . quantum mechanics does not require time - averaging or thorny issues therein : a single quantum state obtains thermalized local observables , and these observables can not distinguish this thermalized pure state from a mixed thermal ensemble of the same thermodynamic character . our measurements open up several avenues for further investigation . instead of operating with fixed total system size , we can study how thermalization and fluctuations depend on the size of the system considered @xcite . conversely , studying integrable hamiltonians where thermalization fails @xcite , and the structure of the associated eigenstate spectrum of such systems , could allow direct tests of the relationship between conserved quantities and thermalization of a quantum state . lastly , applying these tools for characterizing the presence of thermalization and entanglement entropy could be powerful in studying many - body localization , where one of the key experimental signatures is the logarithmic growth of entanglement entropy at long times and suppression of precisely the thermalization we measure here @xcite . * acknowledgments * we acknowledge helpful discussions with soonwon choi , susannah dickerson , jens eisert , michael foss - feig , daniel greif , matthew headrick , david huse , maxim olshanii , cindy regal , johannes schachenmayer , and michael wall . we are supported by grants from the nsf , gordon and betty moore foundations epiqs initiative ( grant gbmf3795 ) , an air force office of scientific research muri program , an office of naval research muri program , and an nsf graduate research fellowship ( m.r . ) . r. geiger , t. langen , i. e. mazets , and j. schmiedmayer , _ local relaxation and light - cone - like propagation of correlations in a trapped one - dimensional bose gas _ , new journal of physics * 16 * , 053034 ( 2014 ) .

the concept of entropy is fundamental to thermalization , yet appears at odds with basic principles in quantum mechanics . statistical mechanics relies on the maximization of entropy for a system at thermal equilibrium . however , an isolated many - body system initialized in a pure state will remain pure during schrdinger evolution , and in this sense has static , zero entropy . the underlying role of quantum mechanics in many - body physics is then seemingly antithetical to the success of statistical mechanics in a large variety of systems . here we experimentally study the emergence of statistical mechanics in a quantum state , and observe the fundamental role of quantum entanglement in facilitating this emergence . we perform microscopy on an evolving quantum system , and we see thermalization occur on a local scale , while we measure that the full quantum state remains pure . we directly measure entanglement entropy and observe how it assumes the role of the thermal entropy in thermalization . although the full state remains measurably pure , entanglement creates local entropy that validates the use of statistical physics for local observables . in combination with number - resolved , single - site imaging , we demonstrate how our measurements of a pure quantum state agree with the eigenstate thermalization hypothesis and thermal ensembles in the presence of a near - volume law in the entanglement entropy . when an isolated quantum system is significantly perturbed , for instance due to a sudden change in the hamiltonian , we can predict the ensuing dynamics with the resulting eigenstate distribution induced by the perturbation or so - called quench " @xcite . at any given time , the evolving quantum state will have amplitudes that depend on the eigenstates populated by the quench , and the energy eigenvalues of the hamiltonian . in many cases , however , such a system can be extremely difficult to simulate , often because the resulting dynamics entail a large amount of entanglement @xcite . yet , surprisingly , this same isolated quantum system can thermalize under its own dynamics unaided by a reservoir ( figure [ fig : conceptual ] ) @xcite , so that the tools of statistical mechanics apply and challenging simulations are no longer required . in this case , a quantum state coherently evolving according to the schrdinger equation is such that most observables can be predicted from a thermal ensemble and thermodynamic quantities . strikingly , even with infinitely many copies of this quantum state , these same observables are fundamentally unable to reveal whether this is a single quantum state or a thermal ensemble . in other words , a globally - pure quantum state is apparently indistinguishable from a mixed , globally - entropic thermal ensemble @xcite . ostensibly the coherent quantum amplitudes that define the quantum state in hilbert space are no longer relevant , even though they evolve in time and determine the expectation values of observables . the dynamic convergence of the measurements of a pure quantum state to the predictions of a thermal ensemble , and the physical process by which this convergence occurs , is the experimental focus of this work . . subsystems of the full quantum state appear pure , as long as the entanglement ( indicated by grey lines ) between subsystems is negligible . if suddenly perturbed , the full system evolves unitarily , developing significant entanglement between all parts of the system . while the full system remains in a pure , and in this sense zero - entropy state , the entropy of entanglement causes the subsystems to equilibrate , and local , thermal mixed states appear to emerge within a globally pure quantum state . ] on - going theoretical studies over the past three decades @xcite have , in many regards , clarified the role of quantum mechanics in statistical physics . the conundrum surrounding the agreement of pure states with extensively entropic thermal states is resolved by the counter - intuitive effects of quantum entanglement . a canonical example of this point is the bell state of two spatially separated spins : while the full quantum state is pure , local measurements of just one of the spins reveals a statistical mixture of reduced purity . this local statistical mixture is distinct from a superposition , because no operation on the single spin can remove these fluctuations or restore its quantum purity . in such a way , the spin s entanglement with another spin creates local entropy , called entanglement entropy . entanglement entropy is not a phenomenon restricted to spins , but exists in all quantum systems that exhibit entanglement . and while probing entanglement is a notoriously difficult experimental problem , this loss of local purity , or , equivalently , the development of local entropy , establishes the presence of entanglement when it can be shown that the full quantum state is pure . in this work , we directly observe a globally pure quantum state dynamically lose local purity to entanglement , and in parallel become locally thermal . recent experiments have demonstrated analogies between classical chaotic dynamics and the role of entanglement in few - qubit spin systems @xcite , as well as the dynamics of thermalization within an ion system @xcite . furthermore , studies of bulk gases have shown the emergence of thermal ensembles and the effects of conserved quantities in isolated quantum systems through macroscopic observables and correlation functions @xcite . we are able to directly measure the global purity as thermalization occurs through single - particle resolved quantum many - body interference . in turn , we can observe microscopically the role of entanglement in producing local entropy in a thermalizing system of itinerant particles , which is paradigmatic of the systems studied in statistical mechanics . in such studies , we will explore the equivalence between the entanglement entropy we measure and the expected thermal entropy of an ensemble @xcite . we further address how this equivalence is linked to the eigenstate thermalization hypothesis ( eth ) , which provides an explanation for thermalization in closed quantum systems @xcite . eth is typically framed in terms of the small variation of observables ( expectation values ) associated with eigenstates close in energy @xcite , but the role of entanglement in these eigenstates is paramount @xcite . indeed , fundamentally , eth implies an equivalence of the local reduced density matrix of a single excited energy eigenstate and the local reduced density matrix of a globally thermal state @xcite , an equivalence which is made possible only by entanglement and the impurity it produces locally within a global pure state . the equivalence between these two seemingly distinct systems , the subsystems of a quantum pure state and a thermal ensemble , ensures thermalization of most observable quantities after a quantum quench . through parallel measurements of the entanglement entropy and local observables within a many - body bose - hubbard system , we are able to experimentally study this equivalence at the heart of quantum thermalization .

experimental protocol entanglement entropy dynamics and saturation local observables in the thermalized pure state discussion

the study on quantum chromodynamics ( qcd ) phase structure is recently extended to finite isospin density @xcite . the physical motivation to study qcd at finite isospin density and the corresponding pion superfluidity is related to the investigation of compact stars , isospin asymmetric nuclear matter and heavy ion collisions at intermediate energies . while the perturbation theory of qcd can well describe the properties of new qcd phases at extremely high temperature and density , the study on the phase structure at moderate temperature and density depends on lattice qcd calculation and effective models with qcd symmetries . the lattice simulation at finite isospin chemical potential @xcite shows that there is a phase transition from normal phase to pion superfluidity phase at a critical isospin chemical potential which is about the pion mass in the vacuum . the qcd phase structure at finite isospin density is also investigated in low energy effective models , such as the nambu jona - lasinio ( njl ) model @xcite applied to quarks @xcite which is simple but enables us to see directly how the dynamic mechanism of isospin symmetry breaking operates . near the phase transition point , the chiral and pion condensates calculated in this model are in good agreement with the lattice simulation @xcite . in a pion superfluid at zero baryon chemical potential , the quark and antiquark of a condensed pair have the same isospin chemical potential and in turn the same fermi surface . when a nonzero baryon chemical potential is turned on , it can be regarded as a fermi surface mismatch between the quark and antiquark . the pion superfluidity in baryonic matter is recently discussed at mean field level in the njl model in chiral limit @xcite and in real case with finite current quark mass @xcite . the pion superfluid can exist when the baryon density is not very high , otherwise the system will be in normal phase without pion condensation because of the too strong mismatch . inside the pion superfluid , the condensed state is separated into two phases . at small isospin chemical potential @xmath0 , the homogeneous and isotropic sarma phase @xcite is free from the sarma instability @xcite and magnetic instability @xcite due to the strong coupling and large enough effective quark mass , it is therefore the stable ground state . at large @xmath0 , while the sarma instability can be cured via fixing baryon density @xmath1 to be nonzero , its magnetic instability implies that the inhomogeneous and anisotropic larkin - ovchinnikov - fudde - ferrell ( loff ) phase @xcite is favored than the sarma phase . in the intermediate @xmath0 region , the stable ground state is the sarma phase at higher @xmath1 and loff phase at lower @xmath1 . the bose - einstein condensation bardeen - cooper - shriffer ( bec - bcs ) crossover at finite baryon and isospin chemical potentials is investigated in the njl model @xcite . the pion condensation undergoes a bec - bcs crossover when the isospin chemical potential increases . the point here is that the crossover is not triggered by increasing the strength of attractive interaction among quarks but driven by changing the isospin density . it is found that the chiral symmetry restoration at finite temperature and density plays an important role in the bec - bcs crossover . most of the work in the njl model is mainly based on the mean field approximation to the quark mass and on the random phase approximation ( rpa ) for the bethe - salpeter equation for the meson masses @xcite . if one examines the thermodynamic potential in the mean field approximation , one sees immediately the deficit of this approach , viz , that only the quarks contribute to the thermodynamic potential with mesons playing no role whatsoever . this is clearly inadequate and unphysical , since one expects at least that the pionic degrees of freedom should dominate the system at low temperature , while the quark degrees of freedom should be relevant only in the chiral symmetry restoration phase . as such , this indicates that calculations in the njl model must be performed beyond the mean field approximation . in ref . @xcite , the thermodynamics of a quark - meson plasma is calculated to order @xmath2 in an expansion in the inverse number of colors , and pions as goldstone particles corresponding to spontaneous chiral symmetry restoration do control the thermodynamic functions at low temperature and density . a characteristic feature of the sarma phase is the intermediate temperature superfluidity @xcite : the superfluidity happens at finite temperature but disappears at zero temperature . since the mean field treatment is a good approximation only at zero temperature @xcite , a careful study on the sarma phase needs to go beyond the mean field . as for the bec - bcs crossover induced by the change in density , the description on the bec phase at low density should be closely related to whether the meson fluctuations are included or not . in this paper , we investigate the pion superfluidity in the frame of the njl model beyond mean field approximation . we will focus on the effect of meson fluctuations on the sarma phase and the bec - bcs crossover at finite temperature and baryon and isospin density . the paper is organized as follows . in section [ s2 ] we present the thermodynamics of the pion superfluidity and the gap equations for the chiral and pion condensates in the njl model in and beyond mean field approximation . in section [ s3 ] we calculate the phase diagram and see the meson effect on the sarma phase and bec - bcs crossover . we summarize and conclude in section [ s4 ] . the two flavor @xmath3 njl lagrangian density is defined as @xmath4\ ] ] with scalar and pseudoscalar interactions corresponding to @xmath5 and @xmath6 excitations , where @xmath7 is the quark field , @xmath8 the current quark mass , @xmath9 the coupling constant with dimension ( gev)@xmath10 , and @xmath11 the quark chemical potential matrix in flavor space @xmath12 with @xmath13 and @xmath0 being baryon and isospin chemical potential . the lagrangian density has the symmetry @xmath14 corresponding to baryon number symmetry , isospin symmetry and chiral symmetry , respectively . however , at nonzero isospin chemical potential , the isospin symmetry @xmath15 breaks down to @xmath16 global symmetry with the generator @xmath17 which is related to the condensation of charged pions . at zero baryon chemical potential , the fermi surfaces of @xmath18 and anti-@xmath19 quarks coincide and hence the condensate of @xmath20 and anti-@xmath21 quarks is favored at sufficiently high @xmath22 and the condensate of @xmath21 and anti-@xmath20 quarks is favored at sufficiently high @xmath23 . we introduce the chiral condensate , @xmath24 and the pion condensate , @xmath25 with @xmath26 . a nonzero condensate @xmath5 means spontaneous chiral symmetry breaking , and a nonzero condensate @xmath27 means spontaneous isospin symmetry breaking . in mean field approximation the thermodynamic potential includes the condensation part and the quark part , @xmath28 and the quark part can be evaluated as a summation of four quasiparticle contributions @xcite , @xmath29 where @xmath30 are the dispersions of the quasiparticles , @xmath31 with the definitions @xmath32 and the function @xmath33 is defined as @xmath34 . the effective quark mass @xmath35 is controlled by the chiral condensate , @xmath36 . the gap equations to determine the condensates @xmath5 ( or quark mass @xmath35 ) and @xmath27 can be obtained by the minimum of the thermodynamic potential @xmath37 , @xmath38 from the first order derivatives , we have @xmath39 considering the relations between @xmath40 and @xmath41 , @xmath42 and @xmath43 , the baryon and isospin density @xmath44 and @xmath45 can be expressed in terms of the @xmath20 and @xmath46 quark density @xmath47 and @xmath48 , @xmath49 it is easy to see that @xmath1 plays the role of density asymmetry for pion condensation . for isospin symmetric matter with @xmath50 the only possible homogeneous and isotropic pion condensed state is the bcs state . the sarma state appears only in isospin asymmetric matter with @xmath51 . the two gap equations ( [ gap2 ] ) and two number equations ( [ number1 ] ) determine self - consistently @xmath52 and @xmath13 as functions of @xmath53 and @xmath1 at mean field level . we now consider the meson contribution to the thermodynamics of the system . the meson modes are regarded as quantum fluctuations above the mean field in the njl model and can be calculated in the frame of rpa @xcite . for the mean field quark propagator with off - diagonal elements in flavor space , @xmath54 we must consider all possible channels in the bubble summation in rpa . in the pion superfluidity region , @xmath5 and charged pions are coupled to each other and the uncharged pion is decoupled from them . using matrix notation for the meson polarization function @xmath55 @xcite , @xmath56 with the quark bubbles @xmath57,\ \ \ \ j , k=\sigma,\ \pi_+,\ \pi_-,\ \pi_0\ ] ] where the trace @xmath58 is taken in color , flavor and dirac spaces and the meson vertexes are defined as @xmath59 the meson masses @xmath60 are determined by @xmath61=0\ ] ] with meson chemical potentials @xmath62 . when the contribution from the meson fluctuations is taken into account , the total thermodynamic potential to order @xmath2 in an expansion in the inverse number of colors becomes @xmath63 where the mean field part @xmath37 is shown in ( [ omf ] ) and the meson part @xmath64 is expressed in terms of the polarization function @xcite , @xmath65.\ ] ] as is expected physically , the mesonic or collective degrees of freedom play a dominant role at low temperature , while the quark degrees of freedom are most relevant at high temperature @xcite . in the chiral symmetry restoration phase at high temperature and/or high density , mesons are not stable bound states , but rather resonant states . they will decay into their quark - antiquark pairs . as a consequence , the determinant in the logarithm of ( [ ofl ] ) is a complex function in the meson energy plane and the imaginary part can be expressed as a scattering phase shift associated with quark - antiquark scattering . from the calculation in the njl model with only chiral dynamics @xcite , the meson width is small around the critical temperature but becomes remarkable when the meson mass is much larger than two times the quark mass , and correspondingly , the contribution from the phase shift to the thermodynamics is negligible at low temperature but significant when the temperature is high enough . for our calculation in the pion superfluidity phase , it can be estimated that the phase shift will be important in the bcs state at high density but its contribution is weakened in the bec state at low density . since we focus in this paper the sarma phase and the bec state which exist at low isospin density , we take pole approximation and neglect the scattering phase shift to simplify the numerical calculations . in pole approximation , the meson contribution can be greatly simplified as a summation of four quasiparticles , @xmath66\nonumber\end{aligned}\ ] ] with meson energies @xmath67 . while mesons do not change the baryon density of the system , the charged pions modify the isospin density when the meson contribution to the thermodynamics is included , @xmath68 where @xmath69 and @xmath70 are the @xmath71 and @xmath72 density . up to this point , the order parameters @xmath35 for chiral phase transition and @xmath27 for pion superfluidity have been regarded as the values minimizing @xmath73 , and @xmath74 has been evaluated at these mean field values , @xmath75 and @xmath76 . while this is a correct perturbative expansion above the mean field , we may ask the questions : what is the feedback from the mesonic degrees of freedom to the order parameters and whether we could improve on these mean field values by regarding @xmath35 and @xmath27 as variational parameters of the total thermodynamic potential @xmath77 ? we now perform this procedure and see what the difference between the new and mean field condensates is . taking the first order derivatives of the total thermodynamic potential with respect to the unknown quark mass @xmath35 and pion condensate @xmath27 , we obtain the following modified gap equations , @xmath78 in comparison with the mean field gap equations ( [ gap2 ] ) , the fluctuation part @xmath74 in the thermodynamic potential leads to a new minimum at @xmath79 and @xmath80 that now differs from the mean field one at @xmath81 and @xmath82 . it is easy to see that the structure of the new gap equations guarantees the two phase transitions . from the second gap equation for pion superfluidity , the trivial solution @xmath83 corresponds to normal quark matter , while the nonzero solution from the zero of the bracket corresponds to the energetically favoured pion condensed state . in the chiral limit , there are also two solutions of the first gap equation corresponding respectively to the chiral symmetry breaking and restoration phase . we now expand the fluctuation part of the thermodynamic potential around the mean field minimum , @xmath84 and , to further simplify the calculation , we consider the expansion only to the first order derivatives . inserting the expansion into the new gap equations yields the following gap equations @xmath85 with two effective coupling constants @xmath86 and @xmath87 defined by @xmath88 in comparing this group of coupled gap equations with the mean field one ( [ gap2 ] ) , one observes that , in the chiral limit , the two groups take the same form , differing only in the effective coupling constants . the coupling constants in the scalar and pseudoscalar channels are the same at mean field level , but they become different and depend on temperature and charge densities when one goes beyond the mean field . if we take @xmath89 , we recover the mean field case . that is , in this approach , the contribution from meson fluctuations is fully included in @xmath86 and @xmath87 . the above approach describes the thermodynamics of a quark - meson plasma with both chiral phase transition and pion superfluidity phase transition beyond the mean field at finite temperature and baryon and isospin density . the two new gap equations ( [ newgap2 ] ) determine simultaneously the order parameters @xmath5 ( or @xmath35 ) and @xmath27 of the two phase transitions . in the chiral limit , the two phase transitions are fully separated from each other @xcite : the chiral symmetry is automatically restored in the pion superfluidity phase . that is , the two order parameters do not coexist in the system . in the real word , chiral symmetry is not fully restored at any isospin chemical potential . however , @xmath5 is much smaller than @xmath27 in the pion superfluidity region @xcite . since we focus in this paper on the fluctuation effect on the pion superfluidity , we will , for the purpose of simplification in numerical calculations , neglect the @xmath5 fluctuations and keep only the @xmath27 fluctuations in the gap equations . namely , we take @xmath90 in the following . since the njl model is non - renormalizable , we should employ a regularization scheme to avoid the divergence in the gap equations . the simplest and normally used way is to introduce a hard three momentum cutoff @xmath91 . in the following numerical calculations , we take the current quark mass @xmath92 mev , the coupling constant @xmath93 gev@xmath10 and the cutoff @xmath94 mev @xcite . this group of parameters ensures the pion mass @xmath95 mev and the pion decay constant @xmath96 mev in the vacuum . in the treatment above , we considered the meson fluctuations as a perturbation around the mean field and took only the first order derivatives in the effective coupling constants ( [ gspi ] ) . if this treatment is good , the difference between the two pion condensates calculated in and beyond the mean field approximation should be small . to check the validity region of this method , we show in fig.[fig0 ] the two condensates as a function of isospin density at fixed temperature and baryon density . at low isospin density which corresponds to the bec region , the difference between the two is really small , but it grows with increasing density and becomes large in the bcs region . therefore , the approximation with only first order derivatives is good for the study of bec , but the contribution from the higher order derivatives may be important for the bcs state . is the chiral condensate in vacuum and @xmath97 is the critical isospin density of pion superfluidity . , width=302 ] the phase diagrams of pion superfluidity in @xmath98 plane at fixed baryon density and in @xmath99 plane at fixed temperature are shown in fig.[fig1 ] . the thin and thick solid lines are respectively phase transition lines in and beyond mean field approximation which separate the normal quark matter at high temperature or high baryon density from the pion superfluidity matter at high isospin density . for pion superfluidity , the averaged fermi surface of the paired quarks is controlled by isospin chemical potential and the mismatch is served by baryon chemical potential . therefore , the sarma phase which is induced by the fermi surface mismatch may enter the pion superfluidity at nonzero baryon density . in mean field approximation , by analyzing the four quasiparticle dispersions @xmath30 , the possible types of sarma state and their thermodynamic and dynamic instabilities are discussed in detail in ref.@xcite . it is found that the sarma phase is the ground state of the pion superfluidity at low isospin chemical potential . very different from the bcs phase structure where the temperature of the pairing state is always lower than the temperature of the normal state , the sarma phase appears in an intermediate temperature region and the normal state exists in lower and higher temperature regions , see the mean field phase transition line in the @xmath98 plane at low isospin density in fig.[fig1 ] . however , the phase structure in the @xmath98 plane is significantly modified when the meson fluctuations are included . from fig.[fig1 ] , the meson effect reduces greatly the pion superfluidity region , and the critical temperature is suppressed from about 150 mev in mean field treatment to about 80 mev in the case beyond the mean field . a qualitative change resulted from the meson fluctuations is that the intermediate temperature superfluidity or the sarma state in mean field calculations is totally washed away , and the normal quark matter is always above the bcs pairing state . in the @xmath99 plane , the phase diagram in mean field approximation is similar to the one in @xmath100 plane obtained in ref.@xcite , and again the meson effect reduces remarkably the pion superfluidity region . plane at fixed baryon density @xmath101(left panel ) and in @xmath99 plane at fixed temperature @xmath102 mev(right panel ) in mean field approximation ( thin lines ) and including meson fluctuations ( thick lines ) . @xmath103 is the normal nuclear density . the solid lines are the phase transition lines and the dashed lines are the bec - bcs crossover lines . , title="fig:",width=302 ] plane at fixed baryon density @xmath101(left panel ) and in @xmath99 plane at fixed temperature @xmath102 mev(right panel ) in mean field approximation ( thin lines ) and including meson fluctuations ( thick lines ) . @xmath103 is the normal nuclear density . the solid lines are the phase transition lines and the dashed lines are the bec - bcs crossover lines . , title="fig:",width=302 ] it has been argued both in effective theory and lattice simulation that at finite but not very large isospin density and zero baryon density , the qcd matter is a pure meson matter , i.e. , a bose - einstein condensate of charged pions . at ultrahigh isospin density , the matter turns to be a fermi liquid with quark - antiquark cooper pairing@xcite . therefore , there should be a bec to bcs crossover when the isospin chemical potential increases . there are some equivalent quantities to describe the bec - bcs crossover induced by changing charge density @xcite . among them are the root - mean - square radius of the cooper pair which is small in bec and large in bcs , the s - wave scattering length which is positive in bec and negative in bcs , the condensate scaled by the fermi energy which is large in bec and small in bcs , and the fermion chemical potential which is negative in bec and positive in bcs . in the following we take the chemical potential to characterize the bec - bcs crossover . for relativistic pion superfluidity , the chemical potential which controls the bec - bcs crossover is @xmath104 @xcite depending on temperature and baryon density through the effective quark mass @xmath35 , and @xmath0 can be viewed as the binding energy of the bound state of quark and antiquark in the bec limit . in fig.[fig2 ] we show @xmath104 as a function of @xmath105 at fixed temperature and baryon density in and beyond mean field approximation . in both cases , the effective chemical potential goes up from negative to positive values with increasing isospin density . the zero point , namely the bec - bcs crossover point , is located at @xmath106 in mean field treatment and @xmath107 in the case with meson fluctuations . the crossover lines determined by @xmath108 in @xmath98 and @xmath99 planes are shown in fig.[fig1 ] . when the mesonic fluctuations are included , not only the pion superfluidity region is greatly reduced , but also the bec region is strongly shrunk . as a function of isospin density at fixed temperature @xmath102 mev and baryon density @xmath101 in mean field approximation ( thin line ) and including mesonic fluctuations ( thick line ) . @xmath109 is the normal nuclear density.,width=302 ] in the bcs limit of the pion superfluidity , the isospin density is high and the paired quark and antiquark is weakly coupled . at the critical temperature , the condensate disappears and the weakly coupled fermions are excited separately , and the system is a fermi liquid . in the bec limit , however , the isospin density is low and the paired quark and antiquark is tightly coupled . in this case , above the critical temperature , the system becomes a bose liquid of tightly bound pions , and the quarks should be too heavy to be excited . this means that , at the critical temperature mesons are lighter than quarks in the bec limit and quarks are lighter than mesons in the bcs limit . to confirm the bec - bcs crossover picture obtained above by calculating the effective chemical potential inside the pion superfluidity , we show in fig.[fig3 ] the meson mass @xmath110 and quark mass @xmath35 as functions of isospin density at the critical temperature and fixed baryon density . with increasing isospin density , the quark mass drops down but the meson mass goes up monotonously . the two lines cross at about @xmath111 which is qualitatively in agreement with the bec - bcs crossover value determined by @xmath108 . ( solid line ) and quark mass @xmath35 ( dashed line ) as functions of isospin density at the critical temperature @xmath112 and fixed baryon density @xmath101 . @xmath109 is the normal nuclear density.,width=302 ] we have investigated the thermodynamics of a pion superfluid at finite isospin density in the frame of two flavor njl model beyond the mean field approximation . considering the fact that mesons , in particular pions because of their low mass , dominate the thermodynamics of a quark - hadron system at low temperature , the mesonic fluctuations should be significant for the phase structure of pion superfluidity . by recalculating the minimum of the thermodynamic potential including meson contribution , we derived a new gap equation for the pion condensate which is similar to the mean field form but with a medium dependent coupling constant . from our numerical calculations , the main effects of the meson fluctuations on the phase structure are : 1 ) the critical temperature of pion superfluidity is highly suppressed and the sarma phase which exists at low isospin chemical potential in mean field approximation is fully washed away , and 2 ) the bec region at low isospin density is significantly shrunk . 99 d.t . son and m.a.stephanov , phys.atom.nucl.*64 * , 834(2001 ) . j.b.kogut,d.k.sinclair , phys.rev.*d66 * , 034505(2002 ) ; * d66 * , 014508(2002 ) ; * d70 * , 094501(2004 ) . nambu and g.jona-lasinio , phys.rev.*122 * , 345(1961 ) and * 124 * , 246(1961 ) . d.toublan and j.b.kogut , phys.lett.*b564 * , 212(2003 ) . m.frank , m.buballa and m.oertel , phys.lett.*b562 * , 221(2003 ) . a.barducci , r.casalbuoni , g.pettini , l.ravagli , phys.rev.*d69 * , 096004(2004 ) ; * d71 * , 016011(2005 ) . l.he and p.zhuang , phys.lett.*b615 * , 93(2005 ) ; l.he , m.jin and p.zhuang , phys.rev.*d71 * , 116001(2005 ) . h.j.warringa , d.boer and j.o.andersen , phys.rev.*d72 * , 014015(2005 ) . d.ebert and k.g.klimenko , j.phys.*g32 * , 599(2006 ) . l.he , m.jin and p.zhuang , phys.rev.*d74 * , 036005(2006 ) . g.sarma,j.phys.chem.solid * 24*,1029 ( 1963 ) . s.wu and s.yip , phys.rev.*a67 * , 053603(2003 ) . a.i.larkin and yu.n.ovchinnikov , jetp*20*,762(1965 ) , p.fulde and r.a.ferrell,phys.rev.*135*,a550(1964 ) . g.sun , l.he and p.zhuang , phys.rev.*d75*,096004(2007 ) . u.vogl and weise , prog.part.nucl.phys.*27 * , 195(1991 ) , s.p.klevansky , rev.mod.phys.*64 * , 649(1992 ) , m.k.volkov , phys.part.nucl.*24 * , 35(1993 ) , t.hatsuda and t.kunihiro , phys.rep.*247 * , 221(1994 ) , m.buballa , phys.rept.407 , 205(2005 ) . p.zhuang , j.hufner and s.p.klevansky , nucl.phys.*a576*,525(1994 ) . j.liao and p.zhuang , phys.rev.*d68 * , 114016(2003 ) . p.pieri , l.pisani and g.c.strinati , phys.rev.*b70 * , 094508(2004 ) . sh.mao , x.huang , and p.zhuang , phys.rev.*c79 * , 034304(2009 ) .

we investigate pion superfluidity in the frame of two flavor nambu jona - lasinio model beyond mean field approximation . we calculate the thermodynamics to the next to leading order in an expansion in the inverse number of colors , including both quark and meson contributions at finite temperature and baryon and isospin density . due to the meson fluctuations , the sarma phase which exists at mean field level is washed away , and the bose - einstein condensation region at low isospin density is highly suppressed .

introduction thermodynamics of the pion superfluidity phase diagrams in and beyond mean field summary

the new avenue in modern semiconductor physics has been opened by outstanding experiment performed recently by kazmierczuk _ et al _ @xcite who detected the large quasi particles known as rydberg excitons in natural crystal of copper oxide . they have observed absorption lines associated with excitons of principal quantum numbers as large as @xmath1 . one could expect that rydberg excitons would have been described , in analogy to rydberg atoms , by rydberg series of hydrogen atom , but it has turned out that this generic method of description should have been revised . this is due to the fact that a size of a huge quasi particle , which in fact is a rydberg exciton with high _ n _ , has a diameter more then two micrometers , which is much larger then the wavelength of light needed to create this exciton . several theoretical approaches to calculate optical properties of rydberg excitons have been presented @xcite@xmath2@xcite . et al _ @xcite developed calculations of of the absorption spectrum on the ground of toyozawa theory and calculated the main parameters for excitonic absorption line for yellow exciton series and emphasized that central - cell corrections have a major influence on the linewidth of the @xmath3-exciton state . in our recent paper @xcite we have proposed the method based on the real density matrix approach ( rdma ) to obtain the analytical expressions for the optical functions of semiconductor crystals , including a high number of rydberg excitons , taking into account the effect of anisotropic dispersion and the coherence of the electron and hole with the radiation field . it is expected that the natural direction of development interest in rydberg excitons is focused on stark effect in such systems because this phenomenon may be used for the optical manipulations of excitons if there is an efficient coupling between the radiation field and excitonic systems far from the band edge . the copper oxide is a perfect candidate for such observations because due to high binding energy of orders of hundred mev and due to their large size , rydberg excitons in cu@xmath00 can exhibit very large electric dipole moments . these features provide that this system is appropriate to observe stark effect experimentally . in semiconductors where the wannier excitons have a small binding energy ( as , for example , gaas ) , the main effect of the applied electric field is the stark shift of the excitonic resonances and changes in their oscillator strengths ( see , for example , @xcite , @xcite for a review ) . in cu@xmath0o which is now the main semiconductor where rydberg excitons are observed , even relatively high excitonic states have a binding energy which is larger than the corresponding ionization energy . thus the excitonic character of the spectra is conserved , but new phenomena , as for example the appearance of symmetry forbidden states , with positions dependent on the applied field strength and on the state number , are observed @xcite . actually , one of the aims of our theoretical paper is to extend the method presented in ref . @xcite , which allowed to describe optical properties of rydberg excitons , in order to obtain electrooptic functions ( susceptibility , absorption reflection and transmission ) . our approach has general character because it works for any exciton angular momentum number and for an arbitrary electric field . in particular , we derive an analytical expression for the electrosusceptibility , from which other electrooptical functions can be obtained . since the electric field effects , increasing with the applied field strength and the state number , compete with the decreasing oscillator strength , we are able , having analytical expressions , to indicate the optimal excitation interval to observe the electrooptical effects . we also indicate the impact of the finite crystal size on the shape of the spectra , which was overlooked in the previous considerations . therefore our predictions should be of interest for experimentalists . the motivation of our considerations is also connected with with potential application of rydberg excitons as solid - state switches . due to their unusual features : long lifetimes , strong dipolar interactions and huge - size , they are expected to be implemented in quantum information technology . @xcite observed rydberg blockade rb , which consists in reduction of excitonic absorption accompanied by increasing laser power for lines associated with large n , what means that only limited amount of rydberg excitons is permitted in a well - localized space of the crystal . the idea of using dipolar rydberg interaction to implement rb bases on the fact that in an ensemble of particles coupled by long - range dipolar interactions , only one particle can be excited at given time . the blockade originate from dipole - dipole interactions between rydberg excitons unnecessary with the same _ n _ and is strongly influenced by their separation . this effect offers exciting possibilities for manipulating quantum bits stores in a single collective excitation in mezoscopic ensembles or for realizing scalable quantum logic gates and one implemented in solids would bring a lot of advantages for quantum information , for constructing all - optical switches and single - photon logic devices . moreover , it is essential to have an additional mechanism for switching the dipole - dipole interaction , which in fact can be tuned on and off by stark effect , therefore it is worth to go into details of the stark effect in rydberg excitons . our paper is organized as follows . in sec . [ density.matrix ] we present the assumptions of considered model and solve the constitutive equation which give an analytical expression for the electrosusceptibility . we also use the obtained expression to compute the effective dielectric function , thanks to which the electrooptical functions ( reflectivity , transmissivity , and absorption ) are derived ( section [ eofunctions ] ) . next , in sec . [ results ] , the electrooptic functions are numerically analyzed for cu@xmath0o crystal for the purpose of realistic implementation of presented method . we examine in details changes of both real and imaginary part of electrosusceptibility , reflectivity and transmission under the influence of electric field . in sec . [ conclusions ] we draw conclusions of the model studied in this paper and we indicate the optimal range of energy for which rydberg excitons could be experimentally observed . wannier - mott excitons , treated as hydrogen - like particles , due to their small binding energy , are very receptive to the action of external fields ( electric and/or magnetic ) . the external fields remove the degeneration of the excitonic energy levels and enhance the optical effects . such effects were observed in the case of rydberg excitons in cu@xmath0o @xcite,@xcite . in what follows we describe the electrooptic properties of systems where the rydberg excitons appear . as was recently shown in ref . @xcite , the so - called real density matrix approach is very effective in describing the optical properties of rydberg excitons . this approach was used in the past for the description of electrooptical effects ( see , e.g. , ref . @xcite and the references therein ) . we show below , that the specific properties of rydberg excitons require a reformulation of the methods used in the past . as in ref . @xcite , we do not enter into the quantum - mechanical explanation of the valence band structure of the cu@xmath0o . this explanation is given in details in the recent paper by schweiner _ et al _ @xcite . here we treat the band structure and the related parameters as known , and use the scheme of ref . @xcite for the situation , when the constant external electric field @xmath4 is applied in the @xmath5 direction . the presented method starts with the constitutive equations , which have the form ( for example , @xcite,@xcite ) @xmath6 where @xmath7 is the bilocal coherent electron - hole amplitude ( pair wave function ) , @xmath8 jest is the excitonic center - of - mass coordinate , @xmath9 the relative coordinate , @xmath10 the smeared - out transition dipole density , @xmath11 is the electric field vector of the wave propagating in the crystal . the coefficient @xmath12 in the constitutive equation represents dissipative processes . we can expect a significant temperature - dependence of the spectra ; microscopic analysis of damping parameters , which are the main temperature - dependent factors , requires future studies and will not be considered explicitly in this paper . the interaction with phonons and their role in determining the line shape , discussed recently by schweiner at al ( ref . @xcite ) who have considered possible causes of line broadening , goes into the field of nonlinear optics and was in the past considered in the framework of the rdma , for example by schlsser ( ref . @xcite ) or in the case of eit , in ref . @xcite . in this paper we do not consider the interaction with phonons , and take the damping coefficients as phenomenological constants . the smeared - out transition dipole density @xmath13 is related to the bilocality of the amplitude @xmath7 and describes the quantum coherence between the macroscopic electromagnetic field and the interband transitions . the two - band hamiltonian @xmath14 includes the electron- and hole kinetic energy terms , the electron - hole interaction potential and the confinement potentials . for details about the hamiltonian see , for example , @xcite . the coherent amplitude @xmath7 defines the excitonic counterpart of the polarization @xmath15,\ ] ] which is than used in the maxwell field equation @xmath16 with the use of the bulk dielectric tensor @xmath17 and the vacuum dielectric constant @xmath18 . in the present paper we solve the equations ( [ constitutiveeqn])-([maxwell ] ) with the aim to compute the electrooptical functions ( reflectivity , transmission , and absorption ) for the case of cu@xmath0o . in the following we will start with considering the bulk situation , where the center - of - mass motion is decoupled from the relative electron - hole motion and given by the term @xmath19 with the wave vector @xmath20 resulting , in general , from the polariton dispersion relation @xcite . we also assume the harmonic time dependence @xmath21 . this assumptions allow to calculate the dielectric susceptibility . this will be achieved in by expanding the coherent amplitudes @xmath7 in terms of eigenfunctions of the hamiltonian @xmath14 . let us note that the solution of the schrdinger equation @xmath22 can be obtained only in an approximative way ( perturbation calculus , variational method , matrix diagonalization etc . ) . considering the cases of cu@xmath0o , when the applied field is of the order of 10 v / cm ( @xcite ) , we can compare the magnitude of the electron - hole pair attractive energy ( @xmath23 in the isotropic effective masses approximation ) and the electric field energy @xmath24 . for @xmath25 one has @xmath26 and @xmath27 ( @xcite ) . thus the excitonic character of the spectra prevails and the applied electric field can be considered as perturbation . it is clear that , when external fields are applied , the full diagonalization of field- and band - mixing effects is more adequate to describe the optical properties , in particular when polarization dependence is considered . however , in the rdma the band parameters , as e.g. the effective masses , are considered as field independent , and the fields are treated as perturbation operators . such approach was merely applied in the past ( for a recent review see @xcite ) , for various nanostructures and field orientations , and was justified by the agreement with experimental data . the considered approximation is also justified by the fact , that the applied field strengths are much below the critical values for the fields ( the ionization field for the electric field and the critical magnetic field ) . in the case of cu@xmath00 the ionization field is of the order of 10@xmath28 v / cm , compared to the applied 15 or even 50 v / cm . we assume the solution of the eq . ( [ schroed ] ) in form of the solutions of an anisotropic schrdinger equation @xmath29 ( see appendix [ appendix a ] for details ) @xmath30 where@xmath31 with @xmath32 defined by ( [ etaellm ] ) , and the laguerre polynomials @xmath33 ( for example , @xcite ) @xmath34 @xmath35 being the spherical harmonics . the energy eigenvalues relate to the eigenfunctions ( [ eigenf ] ) have the form ( see eq . ( [ energies ] ) ) @xmath36 where @xmath37 . we see that the mass anisotropy removes the degeneracy with respect to the quantum number @xmath38 , so that in this approach the higher order excitons @xmath39 , @xmath40 , @xmath41 etc . when the electric field is directed along the _ z_-axis , the perturbation operator @xmath42 has the form @xmath43 we look for the solutions of eq . ( [ constitutiveeqn ] ) in the form @xmath44 inserting the above expansion into ( [ constitutiveeqn ] ) we obtain the following system of equations for the expansion coefficients @xmath45 ( for details , see appendix b ) @xmath46 where @xmath47 where @xmath48 denotes the amplitude of the electric field . in all calculations we will use only the above matrix elements with @xmath49 , denoting them by @xmath50 this is an approximation , which can be justified as follows . the spacing between the rydberg states , at least for the states @xmath51 considered in this paper , is of the order of a few mev . taking , for simplicity , rydberg equal to 100 mev , one has the spacings ( taking @xmath52 states ) ( mev ) @xmath53 , etc . on the other hand , the matrix elements _ v _ , collected in table [ matrix.elements ] and being the measure of the splitting between the stark levels with the same principal number , are of the order between @xmath54 and @xmath55 mev , so that there are much smaller than the distances between the exciton states . obviously , one should notice that the distances between the excitonic states decrease with the increasing number n , whereas the stark splittings increase , and at a certain number the stark splittings are greater than the spacing between the rydberg states . the indication is that for higher numbers _ one should take into account the interaction ( in other words the matrix elements _ v _ ) between different states , and not only within the same state . besides , the method applied is not exactly the perturbation calculus , rather the matrix diagonalization , so its validity is not restricted by the value of the applied field . we put the coherent amplitudes ( [ expansion ] ) into the equation ( [ polarization ] ) , from which , when the center - of - mass motion is decoupled , one can obtain the susceptibility from the relation @xmath56 . the dipole density vectors @xmath57 should be chosen appropriate for _ p_- or _ f_- excitons , and we obtain ( see also @xcite ) @xmath58 where for _ p _ excitons @xmath59-\left(v_{010}^{(3)}\right)^2w_{320}}.\qquad\end{aligned}\ ] ] for @xmath60 we take the same expression as ( [ cc210 ] ) @xmath61 for _ f _ excitons when @xmath62 , taking into account @xmath63 , one has ( see also ( [ cn130 ] ) ) @xmath64\nonumber\\ & & -\left(v^{(n_1)}_{230}\right)^2\left[w_{n_110}w_{n_100}-\left(v_{010}^{(n_1)}\right)^2\right].\end{aligned}\ ] ] for @xmath41 excitons , when @xmath65 , we can extend the basis taking @xmath66 and @xmath67 , obtaining the expressions ( [ cn130ext ] ) . in the above formulas @xmath68 denotes the longitudinal - transverse splitting energy . the explicit form of the oscillator strengths @xmath69 for the isotropic case can be found in ref . @xcite the oscillator strength @xmath70 associated with p excitons is one order of magnitude greater than @xmath71 . both values are roughly proportional to @xmath72 , especially for higher values of n ( see fig . [ figfnn ] ) . . logarithmic scale is applied . dashed line marks the linear regression for @xmath72 relation . ] the quantities of the type @xmath73,@xmath74 which enter into the above equations , correspond to s and d excitons . the matrix elements @xmath75 are calculated in appendix [ appendix c ] . in the previous considerations we treated the semiconductor crystal as unbounded . the real situation is different due to crystal finite size in all directions . practically , the confined size in only one arbitrary chosen direction is considered and usually this direction is the same as electromagnetic wave vector . concerning the experiments with cu@xmath0o one should notice that the dimension of the crystals examined experimentally exceeds the electromagnetic wave length , therefore the use of the long - wave - approximation is not well justified so we will compute the optical functions such as the transmissivity and reflectivity taking into account the finite crystal size and finite wavelength . we will obtain analytic expressions for the optical functions . these expressions will also include the impact of the applied constant electric field . in the description of the optical properties of excitons in finite semiconductors the excitonic bohr radius plays an important role . near the semiconductor surfaces there are layers where the excitons are created ( or destroyed ) , the so - called exciton - free layers ( `` dead layers '' ) . mostly it is assumed that their thickness amounts to 2 - 3 excitonic bohr radii . in the case of gaas it gives about 30 nm . the excitons and related to them polaritons are formed in the remaining volume ( `` bulk '' ) of the crystal , and are responsible for the bulk susceptibility . the junction of layers with different dielectric properties is a complicated task and many works on this topic , including the so - called abc problem , have been done over the past decades ( for review see , for example , @xcite ) . when we consider a particular case of gaas thin layer of the thickness 150 nm , and the relevant excitonic bohr radius is about 15 nm , then two excitonic bohr radii correspond to 20 % of the crystal size . when we consider a cu@xmath0o slab , the situation is quite different . for a cu@xmath0o crystal of the size 30 @xmath76 m @xcite , @xcite even the exciton state with _ _ n__=25 has the extension of about 0.6 @xmath76 m , so that the two bohr radii correspond to 4 % of the crystal size . this means that , in the first approximation , we can neglect the dead layer effects . it does not mean that the dead layer and polariton effects are not important , as they can shift the resonance positions and affect the oscillator strengths ( see also the discussion in ref . the problem is that , when taking into account 25 excitonic states , we have at least 50 polaritonic waves ( including the in- and outgoing polariton waves ) so that the methods applied for the iii - v and ii - vi compounds ( for example , ref . @xcite ) can not be applied for the case under consideration . this aspect requires future studies and will not be explicitly considered in this paper . moreover it should be mentioned that different aspect of semiconductors geometry was considering by schweiner _ et al _ @xcite who have developed the method of investigation excitonic spectra taking into account the discrepancy of valence and conduction bands from parabolic shapes as well as their degeneracy and possible anisotropy . the formation of excitons can be considered as a fast process leading to an effective dielectric function @xmath77 with the excitonic susceptibility defined in eq . ( [ susceptibility ] ) . thus the electromagnetic wave in the crystal propagates in a medium characterized by the effective dielectric function . the crystal under consideration will be modeled by a slab with infinite extension in the @xmath78-plane and the boundary planes @xmath79 . with the sake of simplicity , the slab is located in vacuum . an monochromatic , linearly polarized electromagnetic wave propagates along @xmath5 axis . its electric field is given by @xmath80 where for vacuum @xmath81 @xmath82 being the frequency , and @xmath83 the velocity of light . it is well known , the energy of the propagating wave will be divided into reflected and transmitted wave . the reflectivity , transmissivity , and absorption will be obtained from the relations @xmath84 where @xmath85 is the @xmath86-component of the wave electric field inside the crystal . in the simplest approximation , neglecting the carrier confinement effects leading to the above mentioned abc problem , we can use the effective dielectric function ( [ epsiloneffective ] ) and the resulting effective refractive index @xmath87 then the reflectivity results from the standard formula @xmath88 regarding the exceptional experiments by kazimierczuk _ et al _ @xcite and thewes _ et al_. @xcite we can use the model of the multiple reflection and in the lowest order we get the following expression describing transmission @xmath89 ^ 2 + 4n_2 ^ 2\left(1+n_1\right)^2}\;e^{-\alpha l}.\end{aligned}\ ] ] here @xmath90 denotes the absorption coefficient . o crystal calculated from the imaginary part of the susceptibility in the energetic region of @xmath91 excitonic states , for the electric field strengths @xmath92 and @xmath93 . the logarithmic scale is applied . insets show the absorption spectrum near selected states . ] we have performed numerical calculations of electrooptical functions ( absorption , reflectivity , and transmissivity ) for the cu@xmath0o crystal having in mind the experiments by thewes _ @xcite , and schne _ et al _ @xcite . first , using the obtained expression for the susceptibility ( [ susceptibility]-[cn130a ] ) , we have calculated the electroabsorption , taking into account the lowest @xmath94 excitonic states . the parameters we used are the energies @xmath95 , the gap energy @xmath96 , the l - t energy @xmath68 , and the dissipation parameter @xmath12 . the energies @xmath95 were obtained from the relations ( [ energies1 ] ) with the effective rydberg energy @xmath97 and mass - anisotropy parameter @xmath98 . we have used the values @xmath99 , @xmath100 which is common value in available literature , @xmath101 , and phenomenological value of damping @xmath102 . the results for the absorption , which seem the most important , are reported in figs . [ fig1_1]-[fig13 ] . in fig . [ fig1_1 ] we show the absorption spectrum in the region of @xmath103 excitons , for two values of the applied field . since the absorption peaks decrease quite rapidly the logarithmic scale is applied . for clarity , we present in fig . [ fig3ab ] the contributions of @xmath39 and @xmath41 excitons separately . o crystal , in the energetic region of @xmath91 excitonic states , for the electric field @xmath92 . the logarithmic scale is applied . the contributions of _ p _ and _ f _ excitons shown separately . ] , for @xmath92 and in the range of @xmath104 excitonic states ] the effects of the applied field are more evident when we display the difference @xmath105 . such difference , for @xmath92 , is shown in fig . [ figdiff4 ] . we observe that the numbers of additional peaks with increasing distances between them in comparison to the situation without an electric field . in our model this additional interaction is included in the matrix elements @xmath106 ( eq . ( [ v12 ] ) ) , which values increase with the state number ( see table [ matrix.elements ] ) . therefore in the following we will focus our attention on higher number states . it should be stressed that for these states oscillator strengths are strong enough to warrant the robust and stable structure , which is important for possible further applications . in fig . [ figasess ] we show the electroabsorption in the energetic region of @xmath107 excitonic states , for two values of the applied field strength . the @xmath39 and @xmath41 states are clearly distinguished . crystal calculated from imaginary part of susceptibility in the energetic region of @xmath107 excitonic states , for two electric field strengths @xmath92 and @xmath93 . the logarithmic scale is applied . there is some overlap in the identified states . ] as it was reported in ref . @xcite , the electric field strength can reach 50 v / cm so we performed numerical simulations to examine the influence of field strength on electrooptical properties of our system . it is visible in fig . [ fig11 ] where one can see the basic effect of the applied field : the stark shift of the main peaks and the appearance of new resonances , especially evident in the case of _ f _ excitons . , in the energetic region of @xmath108 exciton , without the electric field , b ) for the field strength 50 v / cm , title="fig:"]a ) , in the energetic region of @xmath108 exciton , without the electric field , b ) for the field strength 50 v / cm , title="fig:"]b ) the changes in the absorption , as a function of the applied field strength , for the range ( 0,50 ) v / cm and near the @xmath108 state , are presented in fig . [ fig12 ] a. the absorption shape for three chosen values of the field are given in fig . [ fig12 ] b. the effect of the stark shift and changes in the oscillator strength can be observed . a ) b ) when we extend the energy interval to include more states , we observe evident mixing and overlapping of the lines of the neighboring states accompanied by spreading of stark shifts with increasing of field strength . [ fig13 ] ) . our theoretical predictions are very close to the experimental results of schne et al @xcite . o crystal , in the energetic region of @xmath109 excitonic states as a function of the applied field strength ] our method allows to calculate both the real and the imaginary part of the susceptibility , without using the kramers - kronig relations . the results for the real part of @xmath110 are presented in fig . having the real and imaginary part of the susceptibility , we have been able to get the effective dielectric function from ( [ epsiloneffective ] ) and other optical functions , in particular , the reflection coefficient ( eq . ( [ rcoeff ] ) ) , which is shown in fig . [ fig5 ] its shape resembles the real part of the susceptibility . we notice the red shift of the main peaks , changes in the oscillator strength , appearance of new peaks when the field is applied , and decreasing the effects for the energies above the 2.171 ev . similar as it was done for the electroabsorption ( fig . [ figdiff4 ] ) , we plot the difference @xmath111 for the energetic region of @xmath107 excitonic resonances ( fig . [ fig5 ] b ) . it can be seen that the electrooptical effects are noticeable , maxima of reflectivity are back - shifted and due to electric field new peaks have occurred . finally , making use of eq . ( [ transmission ] ) , we have calculated the transmissivity of the considered above cu@xmath0o crystal , taking the size @xmath112 . the results for the transmissivity @xmath113 and the difference @xmath114 are shown in fig . [ fig5 ] c , d . the same tendency as for reflectivity can be observed . o crystal in the energetic region of @xmath104 excitonic states , for the electric field @xmath92 and @xmath93 . the logarithmic scale is applied . insets show absorption spectrum near selected states . ] o crystal , in the energetic region of @xmath107 excitonic states , for two values of the electric field . b ) the difference @xmath115 for @xmath92 . c ) the transmissivity @xmath113 , and d ) @xmath114 , for @xmath92.,title="fig:"]a ) o crystal , in the energetic region of @xmath107 excitonic states , for two values of the electric field . b ) the difference @xmath115 for @xmath92 . c ) the transmissivity @xmath113 , and d ) @xmath114 , for @xmath92.,title="fig:"]b ) o crystal , in the energetic region of @xmath107 excitonic states , for two values of the electric field . b ) the difference @xmath115 for @xmath92 . c ) the transmissivity @xmath113 , and d ) @xmath114 , for @xmath92.,title="fig:"]c ) o crystal , in the energetic region of @xmath107 excitonic states , for two values of the electric field . b ) the difference @xmath115 for @xmath92 . c ) the transmissivity @xmath113 , and d ) @xmath114 , for @xmath92.,title="fig:"]d ) the main results of our paper can be summarized as follows . we have proposed a procedure based on the rdma approach that allows to obtain analytical expressions for the electrooptical functions of semiconductor crystals including high number rydberg excitons . our results have general character because arbitrary exciton angular momentum number and arbitrary applied field strength are included . we have chosen the example of cuprous dioxide , inspired by the recent experiment by kazimierczuk _ we have calculated the electrooptical functions ( susceptibility , absorption , reflection , and transmission ) , obtaining a good agreement between the calculated and the experimentally observed spectra . our results confirm the fundamental peculiarity of stark effect - shifting , splitting and , as a result for higher excitonic states , mixing of spectral lines . in particular , we obtained the splitting of p and f excitons , with increasing number of peaks corresponding to increasing state number . we could assess the observed peaks to excitonic states , which are symmetry forbidden when the electric field is absent . on the basis of our theory we have predicted the range of energy where one could observe the stark splitting and shifting for rydberg excitons . all these interesting features of excitons with high @xmath116 number which are examined and discussed on the basis on our theory might possibly provide deep insight into the nature of rydberg excitons in solids and provoke their application to design all - optical flexible switchers and future implementation in quantum information processing . rydberg excitons in cuprous oxide are also promising candidates for observing the influence of magnetic fields effects . very recently the transmission spectrum of cu@xmath0o yellow series was registered in magnetic fields for states with high _ n _ number showing extraordinary complex splitting pattern of levels.@xcite the approach similar to described above could be used to analyze this experiment . below we follow the calculations from ref . @xcite , correcting and supplementing them . consider a two band semiconductor with an isotropic conduction band ( electron ) mass @xmath117 and anisotropic hole mass with the components @xmath118 , with corresponding reduced masses @xmath119 anisotropic schrdinger equation for the relative electron - hole motion , with the above reduced masses and with a screened coulomb interaction , has the form @xmath120\psi = e\psi.\end{aligned}\ ] ] using scaled variables @xmath121 we transform eq . ( [ anisotropic1 ] ) into @xmath122 where @xmath123 @xmath124 is the common laplace operator in spherical coordinates . we are looking for the solution in the form @xmath125 multiplicating both sides with @xmath126 , integrating and taking into account only the diagonal terms @xmath127 we obtain the following equation for the radial part @xmath128 @xmath129 r=0,\ ] ] with @xmath130 and with @xmath131 , assuming that we consider only the bound states . some values for @xmath32 were given in refs . @xcite and @xcite . mostly @xmath98 is close to 1 . in this case , to a good approximation @xmath132 making use of substitutions @xmath133 we transform ( [ rho ] ) into the equation @xmath134 the above equation has two linearly independent solutions @xmath135 known as the whittaker functions . they are related to the more familiar kummer functions ( confluent hypergeometric functions ) by the relations @xmath136 we choose the function @xmath137 which is finite for @xmath138 and , with respect to the relations ( [ relations ] ) we obtain the radial part @xmath128 in the form @xmath139 @xmath140 being the normalization constant . the function @xmath128 is finite for @xmath141 when the first argument of the kummer function is 0 or negative integer . thus @xmath142 which gives @xmath143 and , finally @xmath144 inserting the result for @xmath145 into ( [ radial1 ] ) we obtain the radial function @xmath128 in the form @xmath146 using the integral ( for example , @xcite ) @xmath147 ^ 2{\rm d}z\nonumber\\ & & = \frac{\gamma(\nu)\;n!}{k^\nu\gamma(\gamma+1)\ldots(\gamma+n-1)}\biggl\{1+\frac{n(\gamma-\nu-1)(\gamma-\nu)}{1 ^ 2\cdot \gamma}\nonumber\\ & & + \frac{n(n-1)(\gamma-\nu-2)(\gamma-\nu-1)(\gamma-\nu)(\gamma-\nu+1)}{1 ^ 2\cdot 2 ^ 2\cdot\gamma(\gamma+1)}+\ldots\biggr\}\end{aligned}\ ] ] we obtain the normalization constant in the form @xmath148 thus the radial part of the solution of the anisotropic schrdinger equation has the form @xmath149 using the relation @xmath150 between the kummer function and the laguerre polynomials , we can express the radial function @xmath128 in terms of them @xmath151 inserting the expansion ( [ expansion ] ) into ( [ constitutiveeqn ] ) and making use of the relations @xmath152 and making use of the orthogonality properties of the eigenfunctions @xmath153 we obtain the system of equations ( [ basic ] ) for the expansion coefficients . the equations ( [ basic ] ) form , in general , an infinite system of linear equations . therefore a certain cut - off must be applied . having in mind the properties of cu@xmath0o we put @xmath49 , i.e. we neglect the interaction between the states with different quantum number @xmath116 . it is due to the fact that the energy differences between the states are much larger than the perturbations caused by the electric field . in consequence , the infinite system of equations is reduced to a set of subsystems of equations for each value of @xmath116 . the subsystems consist , in general , of @xmath154 equations labeled by different values of @xmath38 and @xmath155 . with respect to the properties of cu@xmath0o , we will consider the @xmath39 excitons ( @xmath156 ) and @xmath41 excitons ( @xmath157 ) . the lowest @xmath39 exciton state is given by @xmath158 . from ( [ basic ] ) , with @xmath57 given by @xcite @xmath159 and its @xmath160 component , one obtains 4 equations @xmath161 where we took only the allowed combinations for the @xmath162 . thus we obtain @xmath163 for the @xmath164 exciton state we use @xmath165 combinations : @xmath166 and obtain equations @xmath167 where @xmath168 etc . , with regard to ( [ enumeration ] ) . for remaining the remaining coefficients we have @xmath169 . the resulting coefficient @xmath170 is given by the formula ( [ c310 ] ) . in the expressions for @xmath171 and @xmath172 one can separate the real and imaginary part , obtaining @xmath173+\left(e_{tn10 } + e_{tn00}-2e\right){\mit\gamma}^2}{\left[\left(e_{tn00}-e\right)\left(e_{tn10}-e\right)-\left(v_{010}^{(n)}\right)^2-{\mit\gamma}^2\right]^2+\left(e_{tn10 } + e_{tn00}-2e\right)^2{\mit\gamma}^2}\nonumber\\ & + \;{\rm i}{\mit\gamma}&\frac{\left(e_{tn00}-e\right)^2+\left(v_{010}^{(n)}\right)^2+{\mit\gamma}^2}{\left[\left(e_{tn00}- e\right)\left(e_{tn10}-e\right)-\left(v_{010}^{(n)}\right)^2-{\mit\gamma}^2\right]^2+\left(e_{tn10 } + e_{tn00}-2e\right)^2{\mit\gamma}^2}\;.\end{aligned}\ ] ] clearly @xmath174 . introducing notation @xmath175 we put eq . ( [ realimaginar ] ) into a more compact form @xmath176+s_{01}^{(n)}{\mit\gamma}^2 } { \left[r_{01}^{(n)}-\left(v_{010}^{(n)}\right)^2\right]^2+\left(s_{01}^{(n)}\right)^2{\mit\gamma}^2 } + \;{\rm i}{\mit\gamma}\frac{\left(e_0^{(n)}\right)^2+\left(v_{010}^{(n)}\right)^2+{\mit\gamma}^2 } { \left[r_{01}^{(n)}-\left(v_{010}^{(n)}\right)^2\right]^2+\left(s_{01}^{(n)}\right)^2{\mit\gamma}^2}\;.\end{aligned}\ ] ] for @xmath41 excitons , when @xmath177 , we obtain the following equations for the expansion coefficients @xmath178 with the result for the relevant coefficient @xmath179 where @xmath180\\ & & -\left(v^{(n_1)}_{230}\right)^2\left[w_{n_110}w_{n_100}-\left(v_{010}^{(n_1)}\right)^2\right].\end{aligned}\ ] ] using the definitions ( [ definitions ] ) , @xmath181 can be put into the form @xmath182\nonumber\\ c&=&e^{(n_1)}_2r^{(n_1)}_{01}-{\mit\gamma}^2s_{01}^{(n_1)}-e^{(n_1)}_2 \left(v_{010}^{(n_1)}\right)^2-e^{(n_1)}_0\left(v_{120}^{(n_1)}\right)^2,\nonumber\\ d&=&{\mit\gamma}\left[\left(v_{010}^{(n_1)}\right)^2+\left(v_{120}^{(n_1)}\right)^2-s_{01}^{(n_1)}e^{(n_1)}_2-r^{(n_1)}_{01}\right].\nonumber\end{aligned}\ ] ] for @xmath41 excitons , when @xmath65 , we can extend the basis taking @xmath66 and @xmath67 , obtaining @xmath183}{\delta},\nonumber\\ \delta&=&\left[w_{n_140}w_{n_130}-\left(v_{340}^{(n_1)}\right)^2\right]\left[w_{n_120}w_{n_110}w_{n_100}-w_{n_120}\left(v_{010}^{(n_1)}\right)^2- w_{n_100}\left(v_{120}^{(n_1)}\right)^2\right]\nonumber\\ & & -w_{n_140}\left(v^{(n_1)}_{230}\right)^2\left[w_{n_110}w_{n_100}-\left(v_{010}^{(n_1)}\right)^2\right].\end{aligned}\ ] ] the matrix elements follow from the definitions ( [ v12 ] ) : @xmath184 @xmath185 with the laguerre polynomials @xmath186 ( see ( [ laguerre ] ) . substituting @xmath187 and treating @xmath188 as unit , we obtain @xmath189 @xmath190 in particular , for @xmath191 and in units @xmath188 , one obtains @xmath192 where we used the following integral involving laguerre polynomials @xcite @xmath193.\end{aligned}\ ] ] for @xmath194 we have by definition @xmath195 some numerical values for the elements @xmath191 and @xmath194 are given in table [ matrix.elements ] . 99 t. kazimierczuk , d. frhlich , s. scheel , h. stolz , and m. bayer , nature * 514 * , 344 ( 2014 ) . s. hfling and a. kavokin , nature * 514 * , 313 ( 2014 ) . j. thewes , j. hecktter , t. kazimierczuk , m. amann , d. frhlich , m. bayer , m. a. semina , and m. m. glazov , phys . lett . * 115 * , 027402 ( 2015 ) , see also http://link.aps.org/ supplemental/10.1103/physrevlett.115.027402 . s. zieliska - raczyska , g. czajkowski , and d. ziemkiewicz , phys . rev . b * 93 * , 075206 ( 2016 ) . f. schweiner , j. main , and g. wunner , phys . b * 93 * , 085203 ( 2016 ) . j. schlsser , nonlinear optics of excitons in semiconductors , thesis , technical university aachen frg , 1991 . l. silvestri , f. bassani , g. czajkowski , and b. davoudi , eur . b * 27 * , 89 ( 2002 ) . f. schne , s .- o . krger , p. grnwald , h. stolz , m. amann , j. hecktter , j. thewes , d. frhlich , and m. bayer , phys . b * 93 * , 075203 ( 2016 ) . m. freitag , j. heckttter , m. aman , d. frlich , and m. bayer , verhandlungen der dpg , contribution no hl 38.3 . m. feldmaier , j. main , f. schweiner , h. cartarius , and g. wunner , arxiv : 1602.00909v1 [ quant - ph ] 2 feb 2016 . f. schweiner , j. main , m. feldmaier , g. wunner , and ch . uihlein , phys . b * 93 * , 195203 ( 2016 ) . m. amann , j. thewes , d. frhlich , and m. bayer , nature materials ( 2016 ) , doi:10.1038/nmat4622 . s. zieliska - raczyska , g. czajkowski , and d. ziemkiewicz , eur . j. b * 88 * , 338 ( 2015 ) . g. czajkowski , f. bassani , and l. silvestri , rivista del nuovo cimento * 26 * , 1 - 150 ( 2003 ) . a. stahl and i. balslev , _ electrodynamics of the semiconductor band edge _ ( springer - verlag , berlin - heidelberg - new york , 1987 ) . i. s. gradshteyn and i. m. ryzhik , table of integrals , series , and products . ed . by a. jeffrey , 5th edition . ( academic press , san diego , 1994 ) . f. bassani , _ polaritons_. in : _ electronic excitations in organic based nanostructures _ , ed . by v. m. agranovich and g. f. bassani , thin films and nanostructures , vol . * 31 * , 129 - 183 ( 2003 ) ( elsevier , amsterdam , 2003 ) . v. m. agranovich , _ excitations in organic solids _ ( oxford university press , oxford , 2009 ) . g. czajkowski , f. bassani , and a. tredicucci , phys . b * 54 * , 2035 ( 1996 ) . f. bassani , g. czajkowski , and a. tredicucci , z. phys . b * 98 * , 39 ( 1995 ) . l. d. landau and e. m. lifshitz , quantum mechanics ( pergamon press , oxford , 1963 ) .

we show how to compute the electrooptical functions ( absorption , reflection , and transmission ) when rydberg exciton - polaritons appear , including the effect of the coherence between the electron - hole pair and the electromagnetic field . with the use of real density matrix approach numerical calculations applied for cu@xmath00 crystal are performed . we also examine in detail and explain the dependence of the resonance displacement on the state number and applied electric field strength . we report a good agreement with recently published experimental data .

introduction density matrix formulation reflection and transmission spectra results of specific calculations conlusions anisotropic schrdinger equation derivation of the expansion coefficients derivation of the matrix elements @xmath106

the production of @xmath2 mesons at forward rapidity in high energy proton - proton and proton - nucleus collisions can provide valuable information on gluon saturation . indeed , the production of particles at forward rapidity probes the target at very small @xmath1 , where saturation effects should be enhanced . in particular , the charm quark mass being of the same order of magnitude as the saturation scale , @xmath2 production should be sensitive to these dynamics . the charm quark mass is also large enough to provide a hard scale , making a perturbative study of this process possible . in addition , @xmath2 production has been the subject of many experimental studies at the lhc , both in proton - proton @xcite and in proton - nucleus @xcite collisions . this provides a lot of data to confront with nuclear effects predicted by various theoretical models , both in the color glass condensate ( cgc ) framework @xcite and as constraints for nuclear parton distribution functions @xcite and energy loss in cold nuclear matter @xcite . in a recent work @xcite we studied , in the cgc framework , the production of forward @xmath2 mesons in proton - proton and proton - nucleus collisions at the lhc . we showed that , when using the glauber approach to generalize the dipole cross section to nuclei , the nuclear suppression for minimum bias events is smaller than in previous cgc calculations such as @xcite , and much closer to experimental data . in this paper we will study , in the same framework , other observables of interest in this process , such as @xmath2 production at backward rapidity , the centrality dependence in the optical glauber model and the mean transverse momentum of the produced @xmath0 s . let us briefly recall the main steps of the calculation . for more details we refer the reader to ref . we use the color evaporation model ( cem ) to relate the cross section for @xmath2 production to the @xmath3 pair production cross section . in this model a fixed fraction @xmath4 of the @xmath3 pairs produced below the @xmath5-meson threshold are assumed to hadronize into @xmath2 mesons : @xmath6 where @xmath7 , @xmath8 and @xmath9 are the transverse momentum , the rapidity and the invariant mass of the @xmath3 pair respectively , @xmath10 gev is the d meson mass and @xmath11 is the charm quark mass that we will vary between 1.2 and 1.5 gev . note that in this work we will focus on ratios where @xmath4 cancels so we do not need to fix it to any specific value here . the study of gluon and quark pair production in the dilute - dense limit of the cgc formalism was started some time ago @xcite ( see also @xcite ) and used in several calculations such as @xcite . the physical picture is the following : an incoming gluon from the projectile can split into a quark - antiquark pair either before or after the interaction with the target . the partons propagating trough the target are assumed to interact eikonally with it , picking up a wilson line factor in either the adjoint ( for gluons ) or the fundamental ( for quarks ) representation . since we study @xmath2 production at forward rapidity , where the projectile is probed at large @xmath1 , we will use the collinear approximation in which the incoming gluon is assumed to have zero transverse momentum . in this approach the cross section for @xmath12 production reads , in the large-@xmath13 limit @xcite , @xmath14 where @xmath15 and @xmath16 denote the transverse momenta of the quarks , @xmath17 and @xmath18 their rapidities , @xmath19 , and @xmath20 is the dimension of the adjoint representation of su(@xmath13 ) . the longitudinal momentum fractions @xmath21 and @xmath22 probed in the projectile and the target respectively are given by @xmath23 the explicit expression for the `` hard matrix element '' @xmath24 is given in ref . @xcite . in eq . ( [ eq : dsigmaccbarcoll ] ) , @xmath25 is the gluon density in the probe and is described , in the collinear approximation that we use here , in terms of usual parton distribution functions ( pdfs ) . in the following we use , unless otherwise stated , the mstw 2008 @xcite parametrization for @xmath26 . for consistency we use the leading other ( lo ) parton distribution since also the rest of the calculation is made at this order . to estimate the uncertainty associated with the choice of the factorization scale @xmath27 , we will vary it between @xmath28 and @xmath29 . the function @xmath30 describes the propagation of a @xmath31 pair in the color field of the target and reads @xmath32 where @xmath33 is the impact parameter . in this expression , the function @xmath34 contains all the information about the target . it is the fundamental representation dipole correlator : @xmath35 with @xmath36 where @xmath37 is a fundamental representation wilson line in the color field of the target . the dipole correlator @xmath34 is obtained by solving numerically the running coupling balitsky - kovchegov ( rcbk ) equation @xcite . for the initial condition we use the mv@xmath38 parametrization introduced in ref . @xcite , which reads , in the case of a proton target , @xmath39,\end{gathered}\ ] ] where @xmath40 . the running coupling is taken as : @xmath41 where @xmath42 parametrizes the uncertainty related to the scale of the strong coupling in the transverse coordinate space . the free parameters in these expressions are obtained by fitting the combined inclusive hera dis cross section data @xcite for @xmath43 gev@xmath44 and @xmath45 . their best fit values ( with @xmath46 ) are @xmath47 gev@xmath44 , @xmath48 , @xmath49 and @xmath50 mb . in the case of a proton target , the dipole amplitude does not have an explicit impact parameter dependence and we thus make the replacement @xmath51 in eq . ( [ eq : defphi ] ) , @xmath52 corresponding to the effective proton transverse area measured in dis experiments . the fit only includes light quarks . in particular this leaves the charm quark mass that would be consistent with the dis data in this model still uncertain , which is why we vary it in a rather large range for the uncertainty estimate in this work . to generalize this proton dipole correlator to the case of a nuclear target we use , as in @xcite , the optical glauber model . in this model the gluons at the initial rapidity @xmath53 are localized in the individual nucleons of the nucleus . the nucleons are then taken to be distributed randomly and independently in the transverse plane according to the standard woods - saxon nuclear density profile . an analytical average over the positions of the nucleons leads to the following initial condition for the rcbk evolution of a nuclear target : @xmath54 \ ; .\end{gathered}\ ] ] here @xmath55 is the standard woods - saxon transverse thickness function of the nucleus : @xmath56 } \ ; , \ ] ] with @xmath57 and @xmath58 , and @xmath59 is defined such that @xmath55 is normalized to unity . all the other parameters in the initial condition ( [ eq : ica ] ) , which is evolved using the rcbk equation for each @xmath33 , are the same as in the proton case . in this model the dipole amplitude depends on the impact parameter and we need to integrate explicitly over it . the impact parameter dependence , which carries over to the centrality dependence , thus appears naturally in this model . our practical procedure for carrying out this comparison will be discussed in more detail in sec . [ sec : centrality ] . in our previous work @xcite we only considered @xmath2 production in proton - proton and proton - nucleus collisions at forward rapidity . in these kinematics the process can be seen as the collision of a dilute proton probed at large @xmath1 , which can be described using well known parton distribution functions ( pdfs ) , and a dense target described in terms of classical color fields . the nuclear modification of @xmath2 production was also measured at backward rapidity by alice @xcite and lhcb @xcite . in this case the produced @xmath2 is moving in the direction of the incoming nucleus and the physical picture is the same as at forward rapidity , but with the roles of the projectile and the target interchanged , i.e. a dilute proton or nucleus interacts with a dense proton target . the latter is described in the same way as in proton - proton collisions at forward rapidity , while the projectile is described either by a pdf in the case of a proton or by a nuclear pdf ( npdf ) in the case of a nucleus . therefore the calculation is very similar to the case of proton - proton collisions in our previous study . note , however , that in general nuclear pdfs are less tightly constrained by experimental data than usual proton pdfs . in the following we will use the leading order eps09 npdf parametrization @xcite which provides additional error sets to estimate this uncertainty . the lo eps09 analysis uses the cteq6l1 @xcite proton pdfs as a reference , therefore for consistency we use same parametrization of proton pdfs when computing the nuclear modification factor at backward rapidity . nevertheless , while it could be sizeable for the cross section , the difference compared to the mstw 2008 parametrization is very small for the nuclear modification factor . in fig . [ fig : rpa_backward ] we show the nuclear modification factor @xmath60 , defined as @xmath61 as a function of @xmath8 and @xmath62 obtained in this way at negative rapidity compared with data from alice @xcite and lhcb @xcite experiments . the uncertainty in our calculation is significantly larger than at forward rapidity @xcite . this is due to the fact that we include in our uncertainty band , in addition to the variation of the charm quark mass and of the factorization scale , the nuclear pdf uncertainty obtained following the procedure described in ref . in particular , our lower bound for the factorization scale is @xmath63 which can reach values of less than 2 gev at small transverse momentum . the nuclear pdfs are not well constrained at such small scales at the moment . nevertheless the general agreement with data is quite good taking into account the rather large theoretical and experimental uncertainties . in this case deviations of @xmath64 from unity are entirely due to the nuclear pdfs . now that we have computed the nuclear modification factor both at forward @xcite and backward rapidities , we have access to the forward to backward ratio @xmath65 , defined as @xmath66 this ratio can be interesting to study because there may be an additional cancellation of some uncertainties common to the numerator and the denominator . in particular , when determining the nuclear modification factor , experimental studies such as @xcite have to use an interpolation for the reference proton - proton cross section since there is no data at @xmath67 tev . this interpolation is not needed to study @xmath65 , but the final statistical uncertainty may be larger if the coverage in rapidity by the detector is not symmetric with respect to 0 . concerning our calculation , we have seen that at negative rapidity the uncertainty on @xmath60 due to nuclear pdfs is rather large . this error will remain in @xmath65 since the computation of @xmath60 at positive rapidity does not involve npdfs . indeed , we see from fig . [ fig : rfb ] , where we show the forward to backward ratio as a function of @xmath8 and @xmath62 , that the uncertainty on this quantity is still quite large . nevertheless , within this error band the agreement with data is reasonable , although the variation at low @xmath62 seems to be steeper than in the data . in fig . [ fig : rfb ] ( r ) we only show our results for @xmath65 as a function of @xmath62 integrated over the same @xmath8 range as alice data @xcite , which is slightly smaller than for lhcb data @xcite , but our results for @xmath68 would be very similar and alice and lhcb data are compatible with each other . .average number of binary collisions in each centrality class as obtained in the optical glauber model compared with the value estimated by alice @xcite . the values of @xmath69 in the last column are solved from the relation @xmath70 [ cols="^,^,^,^",options="header " , ] we have seen that the optical glauber model contains an explicit impact parameter dependence which can be related to centrality determinations at experiments . in this section we will compare this centrality dependence with its recent measurement at the lhc by the alice collaboration in the range @xmath71 @xcite . we still need to relate the explicit impact parameter dependence in our model to the definition of centrality used by experiments . the experimental data are usually presented in terms of centrality classes . in the optical glauber model these classes would be defined by calculating the impact parameter range corresponding to the centrality class @xmath72 using the relation @xmath73 here @xmath74 is the total inelastic proton - nucleus cross section , given by @xmath75 and the scattering probability at impact parameter @xmath33 is @xmath76 where @xmath77 is the total inelastic nucleon - nucleon cross section . the particle yield in each centrality class is then given by @xmath78 where the values of @xmath79 and @xmath80 are obtained from eq . ( [ eq : centrality ] ) and @xmath81 corresponds to the expression of the cross section before integrating over @xmath33 . in practice , however , this straightforward procedure can not be used for comparing our calculation with the experimental centrality classes . this can be seen most clearly by calculating the average number of binary nucleon - nucleon collisions for each centrality class . in the optical glauber model this is computed using the relation @xmath82 with @xmath83 table [ tab : ncoll ] shows @xmath84 for the optical glauber centrality classes compared to the values given by alice @xcite for the experimental ones . for central collisions the average number of binary collisions estimated by alice is smaller than in the optical glauber model , while the opposite is true for peripheral collisions . the impact parameter is not directly observable , so the experimental centrality selection has to use some other observable as a proxy for it . in the case of the alice analysis @xcite the observable used is the energy of the lead - going side zero - degree calorimeter ( zdc ) . due to the large fluctuations in the signal for a fixed impact parameter , the values of @xmath85 vary less strongly with centrality in the experimental classes than in the optical glauber ones . thus , for example , a relatively peripheral smaller-@xmath85 event can end up in the most central class in the case of a fluctuation in the zdc signal . developing a detailed monte carlo glauber model that would enable us to exactly match the experimental centrality classes would be beyond the scope of this work . therefore we start from the assumption that the alice glauber model is correct and produces a reliable estimate for the @xmath86 in each experimental class . we then assume that the mapping between this @xmath86 and the impact parameter is accurately enough described by our optical glauber model , and use the relation @xmath87 to determine a mean impact parameter corresponding to the experimental centrality class . the values of @xmath69 resulting from this procedure are shown in table [ tab : ncoll ] . we then calculate the nuclear modification factor for the centrality class using this fixed value of @xmath69 . we note that following this procedure for the 80 - 100% centrality class would lead to a value of @xmath69 for which our calculation would not be applicable because the saturation scale of the nucleus falls below the one of the proton . therefore in the following we will only show results for the five most central classes considered by alice . in fig . [ fig : qpa_bins ] we show a comparison of our calculation with alice data for the nuclear modification factor in different centrality classes , @xmath88 , defined as @xmath89 as a function of @xmath62 . we observe that the description of experimental data is generally satisfactory in the first four bins . for the fifth bin the value of @xmath90 obtained in the optical glauber model is almost constant and very close to one , while the data still shows a significant variation with @xmath62 . the discrepancy with experimental data for peripheral collisions comes from the fact that in our model the saturation scale of the lead nucleus falls below the one of the proton for a value of @xmath69 of the order of 6.3 fm , as shown in fig . [ fig : qs2_ncoll ] ( l ) . we see from fig . [ fig : qs2_ncoll ] ( r ) , where we show the number of binary collisions as a function of @xmath69 , that this corresponds to @xmath91 . this is the point where , by definition , @xmath88 reaches 1 in our calculation and beyond which the validity of the framework we have used is questionable . on the other hand , alice data shows that @xmath88 is still significantly smaller than 1 down to @xmath92 ( see fig . [ fig : qpa_ncoll ] ) . the strong centrality dependence is caused by the value of @xmath52 extracted from dis fits being much smaller than the total inelastic nucleon - nucleon cross section . the hera data , both the inclusive cross section fitted in ref . @xcite and data on exclusive vector meson production ( see e.g. @xcite ) , lead to a picture where the small-@xmath1 gluons that can participate in a hard process in a proton are concentrated in a rather small area in the transverse plane ( see also the discussion in ref . this small-@xmath1 gluon `` hot spot '' is then surrounded by a larger `` cloud '' that only participates in soft interactions , contibuting to the total nucleon - nucleon inelastic cross section . our model takes this picture to the extreme , by assuming that at the initial rapidity @xmath93 the gluons contributing to @xmath0 production are concetrated in the area @xmath94 inside the target nucleons . thus , for peripheral collisions , the probe proton can overlap with the soft cloud of @xmath95 target nucleons while still seeing on average only one small-@xmath1 gluon hot spot in the target , thus leaving a hard process like @xmath0 production approximately unmodified . this centrality dependence could probably be mildened by using a larger value for @xmath52 of the order of @xmath77 ( as is effectively done in @xcite ) . this would , however , lose the consistency of our description of the nucleon from hera to the lhc . also , as discussed in refs . @xcite , varying these parameters in an uncontrolled way could very easily , depending on how exactly it is done , lead to an excessive suppression for minimum bias collisions , or to an @xmath60 that is very far from unity even at high transverse momentum . note that the saturation scale ( or gluon density ) at the edge of the nucleus being smaller than that of the proton is an artefact of averaging over the transverse locations of the dense but small gluon hot spots of the nucleons in the target nucleus . therefore we do not use the optical glauber parametrization in this region , but explicitly set @xmath60 to unity . nevertheless , even an explicit monte carlo glauber procedure with the same parameters would not change the ordering @xmath96 that leads to the absence of nuclear effects for peripheral collisions with @xmath97 . at a function of @xmath62 at @xmath98 tev in the 60 - 80% centrality class , both when using a fixed impact parameter and when integrating explicitly over @xmath69 using the @xmath85 distribution obtained in two different models . data from ref . @xcite . ] as a function of @xmath85 at @xmath98 tev compared with alice data @xcite , both when using a fixed impact parameter and when integrating explicitly over @xmath69 . ] the results we have shown in the previous section were obtained with a fixed impact parameter chosen so that the number of binary collisions in the optical glauber model is equal to the average number of binary collisions estimated by alice in each centrality class . however , the nuclear modification ratio in a given centrality bin receives contributions from a distribution of different impact parameters . one could therefore argue that having a profile in the impact parameter space and integrating over @xmath69 could lead to different results . to quantify this effect we will here use two different kinds of @xmath85 distributions to obtain distributions in the impact parameter space . the first one is provided by alice @xcite and is obtained from the slow nucleon model ( snm ) @xcite . since in this model the average number of binary collisions is not the same as the one obtained in the hybrid method used in ref . @xcite , we shift the distributions so that @xmath86 matches the one in the third column of table [ tab : ncoll ] . it should be noted that , contrary to the hybrid method , this method is biased @xcite . in addition , this is only one possible way of extracting @xmath85 distributions at experiments . other methods could yield significantly different distributions . to try to quantify the dependence of our results on the particular shape of the @xmath85 distributions , we will also use , for the 60 - 80% centrality bin which is the most sensitive to fluctuations , a simple linearly decreasing distribution . the two parameters of this distribution , its height at the origin ( @xmath99 ) and the @xmath85 value at which it vanishes ( @xmath100 ) , are determined by imposing that it is normalized to unity and that @xmath101 : @xmath102 , @xmath103 . in fig . [ fig : qpa_dist_ncoll ] we show the values obtained for the nuclear modification factor as a function of @xmath62 in the 60 - 80% centrality bin , both when using a fixed impact parameter and when integrating explicitly over @xmath69 using the two @xmath85 distributions described previously ( snm and linear ) . the explicit integration over @xmath69 leads to a smaller @xmath90 at small transverse momentum . the effect is more pronounced with the linear distribution . similar results are obtained when looking at @xmath90 integrated over @xmath62 as a function of @xmath85 , as show in fig . [ fig : qpa_ncoll ] . in particular , we see that the value of @xmath90 in the 60 - 80% centrality bin obtained with the linear @xmath85 distribution is significantly closer to the alice data point . on the other hand , the value obtained with the snm distribution is very close to the fixed impact parameter result . in conclusion , it is not possible for now to directly compare our impact parameter dependent results with the centrality dependent measurement performed by alice . indeed , for this one would need to have access to an unbiased determination of the @xmath85 distributions in each centrality bin , which does not exist at the moment . here we have tried to estimate the importance of this effect by using @xmath85 distributions obtained in two models . the fact that these two models lead to significantly different results for peripheral collisions while the central bins are much less sensitive to fluctuations means that the variation of @xmath90 as a function of centrality is too model dependent to have a reliable comparison with experimental data . in ref . @xcite we found that the uncertainty on cross sections both in proton - proton and proton - nucleus collisions was rather large . this uncertainty mostly affects the normalization and therefore quantities such as the nuclear modification factor show a smaller uncertainty . besides the nuclear modification factor , another observable which is not sensitive to the absolute normalization of the cross section is the mean transverse momentum of the produced @xmath2 meson . in fig . [ fig : meanpt ] ( l ) we show this quantity as a function of the rapidity in proton - proton collisions at a center of mass energy of 7 tev and compare with lhcb data @xcite . we observe that our calculation is compatible with the data but it is still affected by a relatively large uncertainty . in the collinear approximation on the proton side that we are using here , the mean transverse momentum increases slightly with rapidity , a trend not seen in the data . one must , however , keep in mind that towards central rapidities the intrinsic transverse momentum from also the proton should increase , leading to the opposite behavior . a matching between the collinear and @xmath104-factorized approximations required to fully quantify this effect is beyond the scope of this paper . in fig . [ fig : meanpt ] ( r ) we show the same quantity in proton - lead collisions at a center of mass energy @xmath98 tev . the alice collaboration has also presented results for @xmath105 as a function of @xmath85 . on fig . [ fig : pt2_ncoll ] ( l ) we see that our calculation agrees with this measurement within the rather large uncertainty band ( except for most peripheral collisions , where our calculation is not applicable as explained previously ) . when one considers the difference between @xmath106 in proton - lead and in proton - proton collisions , as shown on fig . [ fig : pt2_ncoll ] ( r ) , the uncertainty on our calculation shrinks and shows a too strong variation as a function of @xmath85 , both when using a fixed impact parameter and when integrating explicitly over @xmath69 using the @xmath85 distributions obtained in the slow nucleon model . however , as in section [ sec : b_dist ] , the results obtained when integrating over @xmath69 depend strongly on the exact shape of the @xmath85 distributions used . in particular , one can see that using a linear @xmath85 distribution leads to a better agreement with data for peripheral collisions . in this work we use the simple color evaporation model to describe the hadronization of @xmath3 pairs into @xmath2 mesons . the normalization of cross sections then depends on a non perturbative constant @xmath4 , see ( [ eq : dsigmajpsi ] ) . the uncertainty associated with this parameter can be eliminated by studying the ratio of cross sections at different center of mass energies . in addition , from the experimental point of view , systematic uncertainties can cancel to some extent in this ratio . such a measurement has been made possible at the lhc for proton - proton collisions thanks to the recent increase of @xmath107 from 8 to 13 tev . in particular the ratio @xmath108 was studied as a function of @xmath8 and @xmath62 by the lhcb collaboration @xcite . in fig . [ fig : ratio_13tev_8tev ] we compare these data with the results that we obtain for this ratio in our model . the resulting uncertainty is rather small and the agreement with data is quite good , in particular at large rapidity and relatively low transverse momentum which is the kinematical domain where our calculation is expected to be the most reliable . thanks to its recent upgrade , the lhc may also perform proton - lead collisions at a higher center of mass energy in the future . here we study how our results would be affected by a change of @xmath109 from 5 to 8 tev . in fig . [ fig : rpa_8tev ] we show the nuclear modification factor @xmath64 at forward rapidity as a function of @xmath8 and @xmath62 at these two energies , as well as existing lhc data at @xmath98 tev . the values at @xmath98 tev shown here differ slightly from the ones shown in figs . 8 and 10 of ref . @xcite because we corrected a numerical problem which was causing the region of large impact parameters ( where the saturation scale of the lead nucleus falls below the one of the proton , see fig . [ fig : qs2_ncoll ] ( l ) ) to be neglected . here we impose @xmath110 in this region , as in ref . @xcite . as one could expect , the higher center of mass energy leads to a stronger suppression due to the higher densities reached in the target . however , the effect is quite small , in particular compared to the size of the uncertainties . for this reason we only show , for @xmath111 tev , our results for the `` central '' values of the parameters ( @xmath112 gev and @xmath113 ) . for @xmath98 tev we show both the central value and the uncertainty band corresponding to the variation of @xmath11 and @xmath27 . therefore , while measuring forward @xmath2 production in 8 tev proton - lead collisions could help reduce experimental uncertainties by getting rid of the interpolation needed for the proton - proton reference , we do not expect significantly stronger nuclear effects at this energy . this is not surprising since we use the same dipole cross sections as in ref . @xcite , where a weak energy dependence of the nuclear modification factor was found in single inclusive particle production . in this paper we have extended our study of forward @xmath2 production in proton - nucleus collisions in the color glass condensate framework to new kinematics and observables . in particular we have studied the nuclear suppression at negative rapidities by describing the nucleus probed at large @xmath1 in terms of nuclear parton distribution functions . we achieved a quite good description of experimental measurements of this quantity , even if the uncertainty is larger at backward than at forward rapidities because the nuclear pdfs are not yet very strongly constrained by data . this allowed us to compute the forward to backward ratio , again with a good agreement with data within the rather large uncertainties . we have also studied the centrality dependence of our calculation . while using the optical glauber model to extend the description from a proton target to a nucleus leads to a better agreement with experimental data for minimum bias observables than previous calculations in the same framework , it is difficult to compare directly the resulting centrality dependence to experimental data . indeed , using a fixed impact parameter obtained in the optical glauber model from the average number of binary collisions estimated by alice leads to a too strong centrality dependence . on the other hand , to integrate explicitly over the impact parameter one has to use @xmath85 distributions based on various assumptions and our calculation is very sensitive to the exact shape of these distributions . finally we have studied how the nuclear modification factor would be affected by an increase of the center of mass energy achievable at the lhc . as expected nuclear effects are stronger but the change is too small to be significant given the size of theoretical uncertainties . t. l. and b. d. are supported by the academy of finland , projects 267321 and 273464 . h. m. is supported under doe contract no . de - sc0012704 . this research used computing resources of csc it center for science in espoo , finland . we would like to thank c. hadjidakis and i. lakomov for discussions on the alice data . collaboration , v. khachatryan _ et . _ , _ prompt and non - prompt @xmath2 production in @xmath114 collisions at @xmath115 tev _ , http://dx.doi.org/10.1140/epjc/s10052-011-1575-8[_eur . phys . j. _ * c71 * ( 2011 ) 1575 ] [ http://arxiv.org/abs/1011.4193 [ arxiv:1011.4193 [ hep - ex ] ] ] . collaboration , r. aaij _ et . _ , _ measurement of @xmath2 production in @xmath114 collisions at @xmath116 _ , http://dx.doi.org/10.1140/epjc/s10052-011-1645-y[_eur . phys . j. _ * c71 * ( 2011 ) 1645 ] [ http://arxiv.org/abs/1103.0423 [ arxiv:1103.0423 [ hep - ex ] ] ] . collaboration , g. aad _ et . _ , _ measurement of the differential cross - sections of inclusive , prompt and non - prompt @xmath2 production in proton - proton collisions at @xmath115 tev _ , http://dx.doi.org/10.1016/j.nuclphysb.2011.05.015[_nucl . phys . _ * b850 * ( 2011 ) 387 ] [ http://arxiv.org/abs/1104.3038 [ arxiv:1104.3038 [ hep - ex ] ] ] . collaboration , s. chatrchyan _ et . _ , _ @xmath2 and @xmath117(2s ) production in @xmath114 collisions at @xmath115 tev _ , http://dx.doi.org/10.1007/jhep02(2012)011[_jhep_ * 02 * ( 2012 ) 011 ] [ http://arxiv.org/abs/1111.1557[arxiv:1111.1557 [ hep - ex ] ] ] . collaboration , b. abelev _ et . _ , _ measurement of quarkonium production at forward rapidity in @xmath114 collisions at @xmath118 tev _ , http://dx.doi.org/10.1140/epjc/s10052-014-2974-4[_eur . phys . j. _ * c74 * ( 2014 ) 2974 ] [ http://arxiv.org/abs/1403.3648 [ arxiv:1403.3648 [ nucl - ex ] ] ] . collaboration , r. aaij _ et . _ , _ measurement of forward @xmath2 production cross - sections in @xmath114 collisions at @xmath119 tev _ , http://dx.doi.org/10.1007/jhep10(2015)172[_jhep_ * 10 * ( 2015 ) 172 ] [ http://arxiv.org/abs/1509.00771[arxiv:1509.00771 [ hep - ex ] ] ] . collaboration , j. adam _ et . _ , _ inclusive quarkonium production at forward rapidity in pp collisions at @xmath120 tev _ , http://dx.doi.org/10.1140/epjc/s10052-016-3987-y[_eur . phys . j. _ * c76 * ( 2016 ) 184 ] [ http://arxiv.org/abs/1509.08258 [ arxiv:1509.08258 [ hep - ex ] ] ] . collaboration , b. abelev _ et . _ , _ @xmath2 production and nuclear effects in p - pb collisions at @xmath109 = 5.02 tev _ , http://dx.doi.org/10.1007/jhep02(2014)073[_jhep_ * 02 * ( 2014 ) 073 ] [ http://arxiv.org/abs/1308.6726[arxiv:1308.6726 [ nucl - ex ] ] ] . collaboration , r. aaij _ et . _ , _ study of @xmath2 production and cold nuclear matter effects in @xmath121 collisions at @xmath122 tev _ , http://dx.doi.org/10.1007/jhep02(2014)072[_jhep_ * 02 * ( 2014 ) 072 ] [ http://arxiv.org/abs/1308.6729[arxiv:1308.6729 [ nucl - ex ] ] ] . collaboration , j. adam _ et . _ , _ rapidity and transverse - momentum dependence of the inclusive j/@xmath117 nuclear modification factor in p - pb collisions at @xmath123 5.02 tev _ , http://dx.doi.org/10.1007/jhep06(2015)055[_jhep_ * 06 * ( 2015 ) 055 ] [ http://arxiv.org/abs/1503.07179[arxiv:1503.07179 [ nucl - ex ] ] ] . collaboration , g. aad _ et . _ , _ measurement of differential @xmath2 production cross sections and forward - backward ratios in p + pb collisions with the atlas detector _ , http://dx.doi.org/10.1103/physrevc.92.034904[_phys . rev . _ * c92 * ( 2015 ) 034904 ] [ http://arxiv.org/abs/1505.08141 [ arxiv:1505.08141 [ hep - ex ] ] ] . collaboration , j. adam _ et . _ , _ centrality dependence of inclusive @xmath2 production in p - pb collisions at @xmath124 tev _ , http://dx.doi.org/10.1007/jhep11(2015)127[_jhep_ * 11 * ( 2015 ) 127 ] [ http://arxiv.org/abs/1506.08808[arxiv:1506.08808 [ nucl - ex ] ] ] . h. fujii and k. watanabe , _ heavy quark pair production in high energy pa collisions : quarkonium _ , http://dx.doi.org/10.1016/j.nuclphysa.2013.06.011[_nucl . phys . _ * a915 * ( 2013 ) 1 ] [ http://arxiv.org/abs/1304.2221 [ arxiv:1304.2221 [ hep - ph ] ] ] . b. duclou , t. lappi and h. mntysaari , _ forward @xmath2 production in proton - nucleus collisions at high energy _ , http://dx.doi.org/10.1103/physrevd.91.114005[_phys . rev . _ * d91 * ( 2015 ) 114005 ] [ http://arxiv.org/abs/1503.02789 [ arxiv:1503.02789 [ hep - ph ] ] ] . ma , r. venugopalan and h .- f . zhang , _ @xmath2 production and suppression in high energy proton - nucleus collisions _ , http://dx.doi.org/10.1103/physrevd.92.071901[_phys . _ * d92 * ( 2015 ) 071901 ] [ http://arxiv.org/abs/1503.07772 [ arxiv:1503.07772 [ hep - ph ] ] ] . k. watanabe and b .- w . xiao , _ forward heavy quarkonium productions at the lhc _ , http://dx.doi.org/10.1103/physrevd.92.111502[_phys . _ * d92 * ( 2015 ) 111502 ] [ http://arxiv.org/abs/1507.06564 [ arxiv:1507.06564 [ hep - ph ] ] ] . h. fujii and k. watanabe , _ leptons from heavy - quark semileptonic decay in pa collisions within the cgc framework _ , http://dx.doi.org/10.1016/j.nuclphysa.2016.03.045[_nucl . _ * a951 * ( 2016 ) 45 ] [ http://arxiv.org/abs/1511.07698 [ arxiv:1511.07698 [ hep - ph ] ] ] . j. l. albacete _ et . _ , _ predictions for @xmath125pb collisions at @xmath122 tev _ , http://dx.doi.org/10.1142/s0218301313300075[_int * e22 * ( 2013 ) 1330007 ] [ http://arxiv.org/abs/1301.3395 [ arxiv:1301.3395 [ hep - ph ] ] ] . r. vogt , _ cold nuclear matter effects on @xmath2 and @xmath126 production at the lhc _ , http://dx.doi.org/10.1103/physrevc.81.044903[_phys . rev . _ * c81 * ( 2010 ) 044903 ] [ http://arxiv.org/abs/1003.3497 [ arxiv:1003.3497 [ hep - ph ] ] ] . f. arleo and s. peign , _ heavy - quarkonium suppression in p - a collisions from parton energy loss in cold qcd matter _ , http://dx.doi.org/10.1007/jhep03(2013)122[_jhep_ * 03 * ( 2013 ) 122 ] [ http://arxiv.org/abs/1212.0434[arxiv:1212.0434 [ hep - ph ] ] ] . f. arleo and s. peign , _ quarkonium suppression in heavy - ion collisions from coherent energy loss in cold nuclear matter _ , http://dx.doi.org/10.1007/jhep10(2014)073[_jhep_ * 10 * ( 2014 ) 73 ] [ http://arxiv.org/abs/1407.5054[arxiv:1407.5054 [ hep - ph ] ] ] . r. vogt , _ shadowing effects on @xmath2 and @xmath126 production at energies available at the cern large hadron collider _ , http://dx.doi.org/10.1103/physrevc.92.034909[_phys . rev . _ * c92 * ( 2015 ) 034909 ] [ http://arxiv.org/abs/1507.04418 [ arxiv:1507.04418 [ hep - ph ] ] ] . j. p. blaizot , f. gelis and r. venugopalan , _ high - energy pa collisions in the color glass condensate approach . 1 . gluon production and the cronin effect _ , http://dx.doi.org/10.1016/j.nuclphysa.2004.07.005 [ _ nucl . phys . _ * a743 * ( 2004 ) 13 ] [ http://arxiv.org/abs/hep-ph/0402256[arxiv:hep-ph/0402256 [ hep - ph ] ] ] . j. p. blaizot , f. gelis and r. venugopalan , _ high - energy pa collisions in the color glass condensate approach . 2 . quark production _ , http://dx.doi.org/10.1016/j.nuclphysa.2004.07.006[_nucl . _ * a743 * ( 2004 ) 57 ] [ http://arxiv.org/abs/hep-ph/0402257 [ arxiv : hep - ph/0402257 [ hep - ph ] ] ] . d. e. kharzeev , e. m. levin and k. tuchin , _ nuclear modification of the j/@xmath117 transverse momentum distributions in high energy pa and aa collisions _ , http://dx.doi.org/10.1016/j.nuclphysa.2014.01.006 [ _ nucl . phys . _ * a924 * ( 2014 ) 47 ] [ http://arxiv.org/abs/1205.1554[arxiv:1205.1554 [ hep - ph ] ] ] . h. fujii , f. gelis and r. venugopalan , _ quark pair production in high energy pa collisions : general features _ , http://dx.doi.org/10.1016/j.nuclphysa.2006.09.012[_nucl . phys . _ * a780 * ( 2006 ) 146 ] [ http://arxiv.org/abs/hep-ph/0603099 [ arxiv : hep - ph/0603099 [ hep - ph ] ] ] . h. fujii , f. gelis and r. venugopalan , _ quark production in high energy proton - nucleus collisions _ , http://dx.doi.org/10.1140/epjc/s2005-02232-y[_eur . phys . * c43 * ( 2005 ) 139 ] [ http://arxiv.org/abs/hep-ph/0502204 [ arxiv : hep - ph/0502204 [ hep - ph ] ] ] . h. fujii and k. watanabe , _ heavy quark pair production in high energy pa collisions : open heavy flavors _ , http://dx.doi.org/10.1016/j.nuclphysa.2013.10.006[_nucl . phys . _ * a920 * ( 2013 ) 78 ] [ http://arxiv.org/abs/1308.1258 [ arxiv:1308.1258 [ hep - ph ] ] ] . a. d. martin , w. j. stirling , r. s. thorne and g. watt , _ parton distributions for the lhc _ , http://dx.doi.org/10.1140/epjc/s10052-009-1072-5[_eur . phys . j. _ * c63 * ( 2009 ) 189 ] [ http://arxiv.org/abs/0901.0002 [ arxiv:0901.0002 [ hep - ph ] ] ] . i. balitsky , _ operator expansion for high - energy scattering _ , http://dx.doi.org/10.1016/0550-3213(95)00638-9[_nucl . phys . _ * b463 * ( 1996 ) 99 ] [ http://arxiv.org/abs/hep-ph/9509348 [ arxiv : hep - ph/9509348 [ hep - ph ] ] ] . y. v. kovchegov , _ unitarization of the bfkl pomeron on a nucleus _ , http://dx.doi.org/10.1103/physrevd.61.074018[_phys . * d61 * ( 2000 ) 074018 ] [ http://arxiv.org/abs/hep-ph/9905214 [ arxiv : hep - ph/9905214 [ hep - ph ] ] ] . i. balitsky , _ quark contribution to the small - x evolution of color dipole _ , http://dx.doi.org/10.1103/physrevd.75.014001[_phys . * d75 * ( 2007 ) 014001 ] [ http://arxiv.org/abs/hep-ph/0609105[arxiv:hep-ph/0609105 [ hep - ph ] ] ] . t. lappi and h. mntysaari , _ single inclusive particle production at high energy from hera data to proton - nucleus collisions _ , http://dx.doi.org/10.1103/physrevd.88.114020[_phys . rev . _ * d88 * ( 2013 ) 114020 ] [ http://arxiv.org/abs/1309.6963 [ arxiv:1309.6963 [ hep - ph ] ] ] . and * zeus * collaborations , f. d. aaron _ et . al . _ , _ combined measurement and qcd analysis of the inclusive @xmath127 scattering cross sections at hera _ , http://dx.doi.org/10.1007/jhep01(2010)109 [ _ jhep _ * 01 * ( 2010 ) 109 ] [ http://arxiv.org/abs/0911.0884 [ arxiv:0911.0884 [ hep - ex ] ] ] . k. j. eskola , h. paukkunen and c. a. salgado , _ eps09 a new generation of nlo and lo nuclear parton distribution functions _ , http://dx.doi.org/10.1088/1126-6708/2009/04/065[_jhep_ * 04 * ( 2009 ) 065 ] [ http://arxiv.org/abs/0902.4154[arxiv:0902.4154 [ hep - ph ] ] ] . j. pumplin , d. r. stump , j. huston , h. l. lai , p. m. nadolsky and w. k. tung , _ new generation of parton distributions with uncertainties from global qcd analysis _ , http://dx.doi.org/10.1088/1126-6708/2002/07/012 [ _ jhep _ * 07 * ( 2002 ) 012 ] [ http://arxiv.org/abs/hep-ph/0201195 [ arxiv : hep - ph/0201195 [ hep - ph ] ] ] . collaboration , s. chekanov _ et . _ , _ exclusive electroproduction of @xmath2 mesons at hera _ , http://dx.doi.org/10.1016/j.nuclphysb.2004.06.034[_nucl _ * b695 * ( 2004 ) 3 ] [ http://arxiv.org/abs/hep-ex/0404008 [ arxiv : hep - ex/0404008 [ hep - ex ] ] ] . l. frankfurt , m. strikman and c. weiss , _ transverse nucleon structure and diagnostics of hard parton - parton processes at lhc _ , http://dx.doi.org/10.1103/physrevd.83.054012[_phys . rev . _ * d83 * ( 2011 ) 054012 ] [ http://arxiv.org/abs/1009.2559 [ arxiv:1009.2559 [ hep - ph ] ] ] . collaboration . public note , to appear . collaboration , j. adam _ et . al . _ , _ centrality dependence of particle production in p - pb collisions at @xmath128= 5.02 tev _ , http://dx.doi.org/10.1103/physrevc.91.064905[_phys . rev . _ * c91 * ( 2015 ) 064905 ] [ http://arxiv.org/abs/1412.6828 [ arxiv:1412.6828 [ nucl - ex ] ] ] .

forward rapidity @xmath0 meson production in proton - nucleus collisions can be an important constraint of descriptions of the small-@xmath1 nuclear wavefunction . in an earlier work we studied this process using a dipole cross section satisfying the balitsky - kovchegov equation , fit to hera inclusive data and consistently extrapolated to the nuclear case using a standard woods - saxon distribution . in this paper we present further calculations of these cross sections , studying the mean transverse momentum of the meson and the dependence on collision centrality . we also extend the calculation to backward rapidities using nuclear parton distribution functions . we show that the parametrization is overall rather consistent with the available experimental data . however , there is a tendency towards a too strong centrality dependence . this can be traced back to the rather small transverse area occupied by small-@xmath1 gluons in the nucleon that is seen in the hera data , compared to the total inelastic nucleon - nucleon cross section .

introduction formalism backward rapidity centrality dependence mean transverse momentum dependence on the center of mass energy conclusions acknowledgments

the color gradients of early type galaxies have been known for quite a long time ( de vaucouleurs 1961 ; boroson , thompson & shectman 1983 ) . such gradients are believed to be due to the variation of the properties of the underlying stellar population , such as age and metallicity . theoretically , radial variations in metallicity are expected in some formation scenarios for early - type galaxies . for example , early simulations of the monolithic collapse of a gas cloud tended to predict metallicity gradients that are too steep to match observations ( e.g. larson 1974 ; carlberg 1984 ) . later simulations based on mergers of gas - rich galaxies predict that interactions between merging galaxies can effectively dampen their metallicity gradients ( e.g. white 1980 ; kobayashi 2004 ) . moreover , radial age gradients are also theoretically conceivable . therefore , detailed color gradient data are essential to discriminate different formation models of early - type galaxies . spectroscopic indices are the most commonly used indicators of metallicity in early - type galaxies . earlier measurements indicate the existence of systematic metallicity gradients at the level @xmath6 to @xmath7 ( baum , thomsen , & morgan 1986 ; carollo , danziger , and buson 1993 ; davies , sadler , & peletier 1993 ; henry & worthey 1999 ; mehlert et al . 2000 and 2003 ) . however , it is not easy to obtain a large sample using this technique , because of the difficulties associated with observing long - slit ( two - dimensional ) spectra . in addition , the rapid decrease of surface brightness with radius limits the radial range over which accurate spectroscopic measurements can be obtained . such measurements are usually possible only out to a radius of about one or two effective radius , making it difficult to study gradients in the outer part of early - type galaxies . on the other hand , broad band surface photometry with ccd is relatively easy to expand the measurements to several effective radius . furthermore , radial profiles obtained from surface photometry are an average measurement within an isophotic annulus , and so are a better representation of the average properties at a given radius . with modern telescopes , accurate multi - waveband photometry can be obtained with little difficulty for a large number of galaxies , making it possible to study the gradients of the stellar population in a statistical way . because of these reasons , photometric measurements are also widely used for studying the metallicity and age gradients in early - type galaxies , although the interpretation of such measurements has to contend with the well - known age - metallicity degeneracy ( e.g. worthey 1994 ) . early analyses of nearby galaxies in optical bands all indicated that early - type galaxies have systematic color gradients ( boroson , thompson , & shectman 1983 ; davis et . al . 1985 ; cohen 1986 ; franx , illingworth , & heckman 1989 ; peletier et al . 1990 ; michard 1999 ; scodeggio 2001 ; idiart , michard , & de freitas pacheco 2002 ) . this is supported by the more recent analysis of the surface photometry of e / s0 galaxies in the nearby rich cluster abell 2199 by tamura & ohta ( 2003 ) . because of the age - metallicity degeneracy , the observed color gradients are usually interpreted either as the metallicity gradients or the age gradients , assuming the gradients of the other quantity is known . assuming the age of the stellar population to be the same over an entire galaxy , both peletier , valentijn , & jameson ( 1990 ) and idiart , michard , & de freitas pacheco ( 2003 ) obtained a metallicity gradient of @xmath8 ( in terms of @xmath9 ) . under the same assumption , tamura & ohta ( 2003 ) found an average metallicity gradient of @xmath10 from a sample of 40 galaxies in abell 2199 and 11 galaxies in abell 2634 . using stellar population synthesis models , saglia et al . ( 2000 ) examined the origin of the observed color gradients by comparing the 20 brightest early - type galaxies in cl0949 + 44 , a cluster at a redshift of @xmath11 taken from hst wf2 frames , with local galaxies . they concluded that their results are better explained in terms of passive evolution of metallicity gradients than pure age gradients . hinkley & i m ( 2001 ) investigated the optical and near - infrared color gradients in the hst wfpc2 and nicmos images for six field early - type galaxies with redshifts from 0.4 to 1.0 . by comparing with stellar synthesis models , they found that 5 out of the 6 galaxies show negligible age gradients and are dominated by a metallicity gradient . similarly , mehlert et al . ( 2003 ) found nearly zero age gradients in 35 early - type galaxies in the coma cluster . however , silva & elston ( 1994 ) investigated both the optical and near - infrared color gradients in eight early - type galaxies and found that all of them show both metallicity and age gradients . therefore , whether the color gradients can be ascribed to pure metallicity gradients is still controversial . as pointed out by de jong ( 1996 ; see also cardiel et al . 2003 ; macarthur et al . 2004 ) , the age - metallicity degeneracy may be partially broken by adding infrared photometry to the optical colors . including infrared data also has another advantage , since contamination by dust is expected to be less important in infrared than in optical . thus , accurate near infrared photometry are extremely useful in studying stellar population gradients in galaxies . in this paper , we use the broad band photometry obtained by the sloan digital sky survey ( sdss ) in the optical , together with the 2 micron all sky survey ( 2mass ) photometric data in the near - infrared , to study the color gradients for nearby early - type galaxies . the relative high image quality in both the sdss and 2mass surveys makes it possible to trace the spectral energy distribution ( sed ) to the outer part of individual galaxies . in addition , since the combined sdss and 2mass data give a uniform coverage of the seds over a large wavelength range , we may hope to use such data to derive stringent constraints on the variations of the stellar populations in early - type galaxies . as we will show , although 2mass images are shallower and have lower resolution than the sdss images , the s / n is sufficient for probing the color gradients to more than 2 effective radii . in this paper , we present results based on 36 early - type galaxies . although this sample is not larger than those used in early analyses , this is the first time where seds from both optical and near infrared are used to study the stellar population gradients . as we will see , the results we obtain are in good agreement with those obtained from line indices measurements , which demonstrates the strength of combining optical and near infrared photometry in constraining the stellar population gradients in early - type galaxies . in the future , much larger samples can be obtained , making this approach a very promising one for such study . the outline of this paper is as follows . we describe our galaxy sample and related data reduction in section [ sec data ] . in sections [ sec color - gra ] and [ sec gradient of a - m ] we examine the color gradients of these galaxies , and use stellar population synthesis models to obtain constraints on their age and metallicity gradients . we discuss our results in section [ sec discussion ] and give a summary of our results in section [ sec summary ] . except where stated otherwise , we assume the hubble constant to be @xmath12 . the sample of early - type galaxies used in this work was selected from the early data release ( edr . stoughton et al . 2002 ) of the sdss , which covers about 462 deg@xmath13 of the sky . nearby galaxies that are brighter than 13.5 magnitude in the @xmath14-band were selected , which guarantees that all the sample galaxies are bright and large enough to be measured in their outer regions in all sdss and 2mass bands . since one of the main goals of this work was to use broad - band photometry to disentangle the age - metallicity effects in observed color gradients , accurate photometric measurements are required . total 36 early - type galaxies of e and s0 type were selected out . most of the morphological types were obtained from the literature ( rc3 of de vaucouleurs et al . 1991 ; nasa / ipac extragalactic database ; nakamura et al . the remaining galaxies were classified with morphological types by inspecting their sdss images and radial profiles . since no other selection criteria were used except the above mentioned magnitude limit , the sample is roughly a magnitude - limited unbiased sample of early - type galaxies . all the galaxies have absolute @xmath15 magnitudes between -19 and -23 with the narrow @xmath16 color range of 0.9 to 1.1 . table [ tab sample ] lists all sample galaxies , their redshifts , morphological types from different sources , and absolute magnitudes in the @xmath15-band ( which was deduced from the sdss @xmath17-images together with the color correction given in eq . [ [ eq b ] ] ) . for each of the sample galaxies , corrected images in the five sdss bands ( @xmath18 , @xmath17 , @xmath14 , @xmath19 and @xmath20 ) were extracted from the edr archive imaging frames , which are produced by the sdss pipeline with both photometric and astrometric calibrations ( stoughton , 2002 ) . each frame contains 2048@xmath211489 pixels , and the size of each pixel is @xmath22 . the typical @xmath23 background noise is @xmath24 , @xmath25 , @xmath26 , @xmath27 and @xmath28 for @xmath18 , @xmath17 , @xmath14 , @xmath19 and @xmath20 , respectively . near - infrared images in the @xmath29 ( 1.25@xmath30 m ) , @xmath31 ( 1.65@xmath30 m ) , and @xmath32 ( 2.17@xmath30 m ) bands were extracted from the 2mass 2nd data release ( cutri et al . 2000 ) according to the sdss coordinates . in our analysis the uncompressed atlas images provided by this data release were used . the pixel size of all of the 2mass images was @xmath33 . the typical @xmath23 background noise we obtained was @xmath34 , @xmath35 and @xmath36 @xmath29 , @xmath31 and @xmath32 , respectively , in agreement with that given in jarrett et al . ( 2003 ) . lcccccc object & redshift & m@xmath37 & rc3@xmath38 & nfy@xmath39 & ned@xmath40 & this work@xmath41 + & & mag(vega ) & & & + ark402 & 0.0178 & -20.01 & & 1 & & + ark404 & 0.0187 & -19.84 & & 0 & & + cgcg010 - 030 & 0.0392 & -21.32 & 0 s0/a & 0 & e & + cgcg390 - 020 & 0.0374 & -21.11 & & & & s0 + gin060 & 0.0443 & -20.28 & & & e & + ic0590 - 1 & 0.0203 & -19.69 & & 0 & e & + ic0590 - 2 & 0.0209 & -19.81 & & 1 & e & + ic0891 & 0.0216 & -20.16 & & 0 & & + ic1517 & 0.0245 & -20.63 & & & & s0 + ic1639 & 0.0180 & -20.08 & & & ce & + ngc0078b & 0.0183 & -19.79 & -2 s0 & & s0@xmath42 pec ? & + ngc0359 & 0.0178 & -20.11 & -2 s0 & & s0- : & + ngc0364 & 0.0170 & -20.17 & -2 s0 & & ( r)sb(s)0@xmath42 & + ngc0426 & 0.0175 & -20.19 & -4 e & 0 & e+ & + ngc0430 & 0.0177 & -20.56 & -5 e & & e : & + ngc0867 & 0.0213 & -20.84 & 0 s0/a & & s0 + : & + ngc0934 & 0.0212 & -20.37 & -2 s0 & & sab0- & + ngc0936 & 0.0048 & -20.48 & -1 s0-s0/a & & sb(rs)0 + & + ngc3325 & 0.0189 & -20.75 & -5 e & & e : & + ngc4044 & 0.0205 & -20.48 & -3 e - s0 & & e+ : & + ngc4493 & 0.0232 & -20.51 & -4 e & 0 & e+ pec : & + ngc5865 & 0.0391 & -21.83 & -2 s0 & & sab0- & + ngc5869 & 0.0070 & -19.65 & -2 s0 & & s0@xmath42 : & + ngc6319 & 0.0275 & -21.05 & & & & s0 + ngc6359 & 0.0099 & -19.45 & -2 s0 & & sa0- : & + ngc6382 & 0.0295 & -21.00 & & & & e + ngc6391 & 0.0276 & -20.51 & & & & e + ngc7684 & 0.0171 & -20.35 & 0 s0/a & & s0 + & + ugc00588 & 0.0443 & -21.67 & 0 s0/a & & e & + ugc00599 & 0.0436 & -21.52 & & & e / s0 & + ugc00797 & 0.0449 & -21.56 & & & cd;e : & + ugc01072 & 0.0172 & -19.63 & -2 s0 & & s0 & + ugc05515 & 0.0443 & -22.29 & -4 e & & e+ pec : & + ugc06435 & 0.0254 & -20.95 & -2 s0 & 0 & s0@xmath42 : & + ugc07177 & 0.0203 & -19.99 & -3 e - s0 & 0 & s0- ? & + ugc07813 & 0.0232 & -20.67 & -5 e & 0 & e : & + image processing in the five sdss bands was originally performed by the photometric pipeline ( photo ) , which performs tasks such as labelling frame artifacts ( bad columns , cosmic rays , bleed trails ) , bias subtraction , and flat - field correction . the astrometric solutions for all the five bands were obtained from the astrometric pipeline ( astrom ) with typical errors much smaller than @xmath43 ( stoughton et al . the astrometric accuracy of 2mass is better than @xmath44 ( cutri et al . thus the images in any 2 bands of the sdss and 2mass could be matched very well . although the images from the sdss archive were processed with photometric corrections , some spurious features were still present in many image frames , especially in the @xmath18 and @xmath20 bands . the spatial variation of these features was about 1 to 2 adu . for galaxies which happen to be located in such features , the variation was found to affect both background subtraction and the measurement of color profiles in the outer parts of the image , where the surface brightness is low . fortunately , most of these features were only one dimensional , and the variation was only along horizontal lines . thus it was not difficult to measure and then pick them out from images . to do this , firstly a region was selected that includes several hundred lines , covers all columns but does not contain any bright objects . the median of these lines was then measured and used to represent the features . finally , the high order spline3 function was used to fit the median line , and features were removed from each line of the image . figure [ fig1 ] shows an example of such spurious features in a @xmath18-band image . the upper two panels show plots of the average of 500 lines in two different parts of the ccd frame . the left panel shows lines 1 to 500 , while the right panel shows lines 971 to 1470 . as one can see , the features in both panels are quite similar , indicating that these features are quite systematic through the whole ccd frame . the lower two panels show the same plots as in the upper panels , except that the spurious features are now subtracted . one can see that the patterns are removed quite successfully by our method . in most cases , the subtraction method reduced the spatial variation to a level @xmath45 adu . the subtraction method failed to suppress the variation to such a low level only for a few very large galaxies and for images with heavily saturated stars , where it is difficult to find a sufficiently large blank - line region to estimate the median value . the fwhms of the sdss psfs ( @xmath46 ) are much smaller than those of the 2mass ( @xmath47 ) . therefore the sdss images were downgraded by convolving with a proper gaussian kernel to match the 2mass psfs , in order that seds for the inner regions of the galaxies could be obtained with a similar seeing in different bands . note that the sdss psf may change slightly from band to band , and the kernel used is in accordance with the specific filter and field in consideration . to obtain the true surface brightness profiles , an accurate sky background subtraction is essential . for a given galaxy , all other objects in the frame were masked using sextractor ( bertin & arnouts 1996 ) . since most of the sample galaxies have sizes of only a few arcmins , a region of sky sufficient to fit the sky background reasonably well could easily be found . the isophote package in iraf was used to fit each of the sky - subtracted images with a series of elliptical annuli from the center to the outskirts , with the length of the semi - major axis increasing by 10% in each step . the width of annuli was chosen to increase with radius to suppress noise in the outer regions , where the signal - to - noise is low . the coordinates of the photometric peak in each band were obtained by the daophot package and were fixed in the fit . the difference in the peak coordinates between the eight bands is less than @xmath48 . we also fixed the sky level in the fit , but kept other parameters , such as ellipticity and position angle , as free parameters . in addition , all foreground and background objects were masked out in order to reduce their contamination . since our goal was to obtain the radial profile of the sed , it was essential that each elliptical isophotic annulus have the same physical size in all of the eight bands . to achieve this the @xmath14-band image was used to define isophotic annuli , and these were applied to all the eight bands . in the sdss , photometric quantities are usually quoted in the ` petrosian ' system(petrosian 1976 ) . sdss defines the petrosian ratio @xmath49 at a radius @xmath14 from the center of an object to be the ratio of the local surface brightness averaged over an annulus at @xmath14 to the mean surface brightness within @xmath14 ( blanton et al . 2001 ; stoughton et al . 2002 ) : @xmath50 } { \int_0^r dr ' 2\pi r ' i(r ' ) / [ \pi r^2]},\ ] ] where @xmath51 is the azimuthally averaged surface brightness profile . the petrosian radius @xmath52 is defined as the radius at which @xmath53 equals 0.2 . the petrosian flux in any band is then defined to be the flux within two petrosian radii : @xmath54 the petrosian half light radius @xmath55 is defined as the radius within which 50% of the petrosian flux is contained . in our analysis , we cut off the brightness profiles at a radius of @xmath56 . at this radius , the surface brightness is about 8% of the @xmath17 , @xmath14 , @xmath19-skys , and is about 2% of the @xmath18 , @xmath20-skys , and about 0.1 - 0.3% of the @xmath29 , @xmath31 , @xmath32-skys . the typical values of @xmath55 of our sample galaxies are about @xmath57 . figure [ fig2 ] shows an example of the surface brightness profiles obtained , in this case for ngc 0430 , and figure [ fig3 ] shows the corresponding seds derived from the fluxes in the eight bands at different radii . to have an overall view of the color gradient for the sample galaxies we are considering , we show in figure [ fig4 ] the relative shape of the sed at different radii ( in units of @xmath55 ) in terms of the median values for the for 35 sample galaxies . the offset of the seds at @xmath58 and at @xmath59 is about 0.4 mag from the red end to the blue end . this implies a color gradient in the sample galaxies , in the sense that the stellar light becomes redder toward the central regions . such gradients will be quantified further in section [ sec gradient of a - m ] . it is interesting that @xmath60 for the @xmath61band shows abnormal behavior at all radii . in the inner region , the @xmath19-band surface brightness appears fainter than expected , while in the outer region it is brighter . as we will see in section [ subsubsec psf ] and in the appendix , this is caused by the so called ` red - halo ' effect in the @xmath61band . the sdss magnitudes are quite close to the ab - magnitude system ( oke & gunn 1983 ) . the zero - point of magnitude for each frame can be obtained directly from the edr catalogue , and the corrections to the ab system are -0.042 , 0.036 , 0.015 , 0.013 , and -0.002 for the @xmath18 , @xmath17 , @xmath14 , @xmath19 , and @xmath20 bands respectively ( blanton et al . the 2mass zero - point in the vega magnitude system can be found in the atlas image header . these were converted to ab magnitudes by adding offsets of 0.89 , 1.37 , and 1.84 for the @xmath29 , @xmath31 , and @xmath32 bands respectively ( finlator et al . , 2000 ) . throughout this paper , all magnitudes are in the ab system unless otherwise stated . in addition , dust extinction by the milky way was corrected for each galaxy according to its galactic coordinates as provided by schlegel , finkbeiner , & davis ( 1998 ) . lllllllll source of error & @xmath18 & @xmath17 & @xmath14 & @xmath19 & @xmath20 & @xmath29 & @xmath31 & @xmath32 + ( in percentage of the sky level ) & & & & & & & & + random noise of each pixel@xmath38 & 12.5%&4.0%&2.7%&2.4%&3.0% & 0.62% & 0.25% & 0.20% + random noise at @xmath56@xmath39 & 0.25%&0.08%&0.05%&0.05%&0.06% & 0.11% & 0.04% & 0.03% + background subtraction&0.54%&0.19%&0.13%&0.17%&0.18% & 0.06% & 0.02% & 0.01% + + @xmath38 include both readout noise and random noise of the sky background + @xmath39 the isophote annulus at radius of @xmath56 is assumed to contain about 2,500 pixels in a sdss image and about 400 pixels in a 2mass image . one crucial step in our analysis is to take into account the photometric error in the different wavebands . the observational error in the surface brightness profile consists of the following several components . the first is random error , including the readout noise , the shot noise of the sky background and the shot noise of the galaxy in consideration . at small radii , where the brightness of the galaxy is much higher than the sky background , the noise from the galaxy dominates . on the other hand , at large radii where the brightness of the galaxy is much fainter than the sky background , the noise from the galaxy can be neglected in comparison to the readout noise and the noise of sky background . here we are concerned mainly with the errors in the outer region , because these errors determine to which radius the observed surface brightness profile can be trusted . in the first row of table [ tab error ] , we list the random error in one pixel ( expressed in terms of the percentage of sky level ) , which includes both the readout noise and the noise due to the sky background . because the radial profile is the average of the flux of all the pixels in the corresponding isophote annulus , the final error introduced by random noise depends on the number of pixels in the annulus . for most of our galaxies , the annulus at @xmath56 contains more than 2,500 pixels in a sdss image , and more than 400 pixels in a 2mass image . the final random errors at @xmath56 are also presented in table [ tab error ] . it should be noted that the 2mass images were produced by averaging 6 frames ; the camera pixels were divided into 4 final coadd pixels and the pixel values were smoothed by the weinberg filter ( jarrett et al . all these have been taken into consideration in our error estimates . another important source of error in the outer part of a galaxy is from background subtraction . this error was estimated in a way similar to that described in wu et al . ( 2002 ) where the deep optical images of the edge - on galaxy ngc 4565 was analyzed . firstly , all the objects including the target source were masked in the background - subtracted image . secondly , a series of boxes with a size of @xmath62 were used to cover the entire image , where the box size is chosen to be comparable to the typical value of @xmath56 of the sample galaxies , so that they can be used to estimate the large - scale error induced by improper background subtraction and flat - field correction . finally , the average of the un - masked region in each box is obtained , and the standard deviation of these averages among all the boxes was used as the error of background subtraction . this error is larger than the random noise in the optical bands but lower than the random noise in the near - infrared bands . note that the background subtraction will not work well if severe airglow variation exists in the near - infrared images ( jarrett et al . 2000 ) . absolute calibration errors are also very important in our analysis , because we want to construct the seds using the fluxes in different bands . fortunately , such errors are quite small , about 0.02 mag in all the sdss and 2mass bands ( abazajian et al . 2004 ; cohen , wheaton & megeath 2003 ) . in conclusion , the total error budget combining all of the errors discussed above is smaller than 0.03 mag in the @xmath17- , @xmath14- , @xmath19-bands , 0.07 mag in the @xmath20-band , 0.18 mag in @xmath18 and about 0.3 mag in @xmath29 , @xmath31 , @xmath32 . it should be pointed out here that only the outmost few points in our color profiles have large errors ; the total profile of a galaxy is not affected significantly by these uncertainties . in addition to the normal photometric corrections discussed above , there was another instrumental effect that affected the surface photometry analysis . since the sdss photometry survey uses a thinned ccd for the @xmath18 , @xmath17 , @xmath14 , and @xmath19 bands , there is an effect referred in the literature as the ` red halo ' of the psf wings ( e.g. michard 2002 ) . this effect causes the extended wing of the red - band psf to be much brighter than that in the blue - band . in the sdss photometric system , the red halo in the @xmath19-band is more apparent than that in the other bands ( see appendix ) . since most of the sample galaxies have sizes of about several arcmins and the red halo effect appears prominent at about 10 to 50 arcsecs , it is a serious problem in measurements of surface photometry , and will also affect the measured sed and color gradients . as pointed out by michard ( 2002 ) , the red halo effect depends on both observational time and the field . to avoid such an effect , we will discard the @xmath19-band data in the following discussion . radial surface brightness profiles in eight bands for the 36 sample galaxies were obtained , and a linear least - square method was adopted to fit the derived color profiles . to avoid seeing effects ( young & currie 1995 ) , an inner radius of @xmath63 ( @xmath64 ) was used when fitting . as the data became quite noisy at large radius in the @xmath18 , @xmath29 , @xmath31 and @xmath32 bands , the outer radius was cut at @xmath56 , which is equivalent to approximately 15 kpc in physical size on average , to ensure the errors lower than 0.1 mag arcsec@xmath65 in the optical bands ( except @xmath18 band ) . the color gradients were expressed as @xmath66 in order to compare with previous results obtained by peletier et al . ( 1990 ) in the optical and by silva & elston ( 1994 ) in the near - infrared . figure [ fig5 ] shows the color gradients for three example galaxies : ark 402 , ngc 0867 , and ngc 6359 . four color gradients are presented for each galaxy , from top to bottom in each column . table [ tab color - gra ] presents the gradients of color profiles for all sample galaxies . the last line in table [ tab color - gra ] is the average gradient of each color profile over all sample galaxies . it is clear from figure [ fig5 ] and table [ tab color - gra ] that all four color gradients have negative values , indicating that the color becomes bluer from the inner to outer parts in early - type galaxies . this result is in agreement with previous investigations of early - type galaxies ( vader et al . 1988 ; franx , illingworth , & heckman 1989 ; peletier , valentijn , & jameson 1990 ; peletier et al . 1990 ; goudfrooij et al . 1994 ; tamura et al . 2000 ; tamura & ohta 2000 ) . note that the errors quoted in the table include both the fitting error and the errors addressed in previous section . in order to transform the sdss photometric data to the standard @xmath67 vega magnitude system , the formulae of smith et al . ( 2002 ) were used : @xmath68 @xmath69 @xmath70 @xmath71 based on the average color gradients given in table [ tab color - gra ] , the average color gradients in @xmath72 , @xmath16 , and @xmath73 were @xmath74 , @xmath75 , and @xmath76 mag per dex in radius respectively . the derived color gradients in @xmath77 and @xmath78 were @xmath79 and @xmath80 mag per dex in radius . lrrrr objects & @xmath81 & @xmath82 & @xmath83 & @xmath84 + ark402 & -0.12@xmath85 0.05&-0.22@xmath85 0.01&-0.14@xmath85 0.09&-0.74@xmath85 0.08 + ark404 & -0.22@xmath85 0.03&-0.02@xmath85 0.01&-0.05@xmath85 0.05&-0.10@xmath85 0.04 + cgcg010 - 030&-0.15@xmath85 0.08&-0.04@xmath85 0.01&-0.10@xmath85 0.11&-0.26@xmath85 0.10 + cgcg390 - 020 & 0.04@xmath85 0.06&-0.06@xmath85 0.01&-0.12@xmath85 0.06&-0.35@xmath85 0.05 + gin060 & -0.37@xmath85 0.07&-0.13@xmath85 0.02&-0.13@xmath85 0.09&-0.61@xmath85 0.07 + ic0590 - 1 & -0.18@xmath85 0.05&-0.06@xmath85 0.01&-0.11@xmath85 0.10&-0.16@xmath85 0.08 + ic0590 - 2 & -0.23@xmath85 0.05&-0.03@xmath85 0.03 & 0.00@xmath85 0.08&-0.22@xmath85 0.06 + ic0891 & -0.25@xmath85 0.05&-0.04@xmath85 0.02&-0.14@xmath85 0.09&-0.37@xmath85 0.06 + ic1517 & -0.16@xmath85 0.03&-0.01@xmath85 0.01&-0.14@xmath85 0.04&-0.31@xmath85 0.03 + ic1639 & -0.01@xmath85 0.02&-0.02@xmath85 0.01 & 0.10@xmath85 0.11 & 0.03@xmath85 0.09 + ngc0078b & -0.22@xmath85 0.03&-0.08@xmath85 0.01&-0.08@xmath85 0.07&-0.35@xmath85 0.05 + ngc0359 & -0.29@xmath85 0.06&-0.03@xmath85 0.01&-0.17@xmath85 0.13&-0.34@xmath85 0.11 + ngc0364 & -0.07@xmath85 0.07&-0.04@xmath85 0.01&-0.01@xmath85 0.09&-0.22@xmath85 0.07 + ngc0426 & -0.07@xmath85 0.07&-0.05@xmath85 0.01&-0.28@xmath85 0.06&-0.46@xmath85 0.04 + ngc0430 & -0.23@xmath85 0.03&-0.04@xmath85 0.01&-0.12@xmath85 0.03&-0.25@xmath85 0.03 + ngc0867 & -0.14@xmath85 0.07&-0.01@xmath85 0.01 & 0.03@xmath85 0.09&-0.15@xmath85 0.07 + ngc0934 & -0.29@xmath85 0.03&-0.05@xmath85 0.01&-0.25@xmath85 0.05&-0.50@xmath85 0.03 + ngc0936 & -0.02@xmath85 0.02&-0.02@xmath85 0.01&-0.05@xmath85 0.03&-0.17@xmath85 0.03 + ngc3325 & -0.10@xmath85 0.04&-0.05@xmath85 0.01&-0.10@xmath85 0.07&-0.16@xmath85 0.06 + ngc4044 & -0.18@xmath85 0.05&-0.05@xmath85 0.01&-0.01@xmath85 0.07&-0.36@xmath85 0.05 + ngc4493 & -0.35@xmath85 0.06&-0.07@xmath85 0.02&-0.01@xmath85 0.15&-0.36@xmath85 0.12 + ngc5865 & -0.32@xmath85 0.06&-0.07@xmath85 0.01&-0.10@xmath85 0.11&-0.47@xmath85 0.09 + ngc5869 & -0.06@xmath85 0.04&-0.04@xmath85 0.01&-0.09@xmath85 0.07&-0.19@xmath85 0.06 + ngc6319 & -0.09@xmath85 0.05 & 0.04@xmath85 0.02&-0.22@xmath85 0.07&-0.52@xmath85 0.05 + ngc6359 & -0.26@xmath85 0.03&-0.08@xmath85 0.01&-0.16@xmath85 0.03&-0.35@xmath85 0.02 + ngc6382 & -0.14@xmath85 0.05&-0.10@xmath85 0.01&-0.15@xmath85 0.07&-0.26@xmath85 0.05 + ngc6391 & -0.26@xmath85 0.04&-0.10@xmath85 0.01&-0.04@xmath85 0.05&-0.24@xmath85 0.04 + ngc7684 & -0.13@xmath85 0.02&-0.06@xmath85 0.01 & 0.05@xmath85 0.07&-0.06@xmath85 0.06 + ugc00588 & -0.18@xmath85 0.06&-0.04@xmath85 0.01&-0.11@xmath85 0.08&-0.33@xmath85 0.06 + ugc00599 & -0.20@xmath85 0.06&-0.05@xmath85 0.01&-0.16@xmath85 0.08&-0.30@xmath85 0.06 + ugc00797 & -0.36@xmath85 0.09&-0.02@xmath85 0.01&-0.05@xmath85 0.13&-0.22@xmath85 0.10 + ugc01072 & -0.15@xmath85 0.03&-0.03@xmath85 0.01&-0.05@xmath85 0.07&-0.04@xmath85 0.05 + ugc05515 & -0.25@xmath85 0.07&-0.06@xmath85 0.01 & 0.10@xmath85 0.12&-0.19@xmath85 0.09 + ugc06435 & -0.16@xmath85 0.05&-0.11@xmath85 0.01&-0.12@xmath85 0.08&-0.28@xmath85 0.06 + ugc07177 & -0.02@xmath85 0.12 & 0.01@xmath85 0.02 & 0.02@xmath85 0.20&-0.16@xmath85 0.15 + ugc07813 & -0.22@xmath85 0.05&-0.07@xmath85 0.01&-0.16@xmath85 0.08&-0.34@xmath85 0.07 + + average & -0.18@xmath85 0.06&-0.05@xmath85 0.01&-0.09@xmath85 0.08&-0.29@xmath85 0.07 + in table [ tab comparison ] we list the color gradients so obtained along with the results obtained from earlier analyses . our @xmath72 , @xmath16 , and @xmath73 color gradients are in very good agreement with those obtained by idiart , michard , & de freitas pacheco ( 2002 ) using 36 early - type galaxies , and by michard ( 2000 ) who reanalyzed 30 ellipticals taken from the literature . the color gradients are also consistent with the @xmath77 and @xmath78 color gradients derived by peletier et al . ( 1990 ) from 39 ellipticals , and by franx and illingworth ( 1990 ) from 17 ellipticals , and with the @xmath78 color gradients obtained by tamura & ohta ( 2003 ) from galaxies in abell 2199 and abell 2634 . all these show that the sdss images can be used to study the surface photometry of nearby galaxies . llllll & @xmath86 & @xmath87 & @xmath88 & @xmath89 & @xmath90 + this work & @xmath91&@xmath92&@xmath93&@xmath94&@xmath95 + idiart(2002 ) & @xmath91&@xmath96&@xmath97 & & + michard(2000 ) & @xmath98&@xmath99&@xmath100 & & + peletier(1990 ) & & & & @xmath101 & @xmath102 + franx(1990 ) & & & & @xmath103&@xmath104 + tamura(2003 ) & & & & & @xmath105 + in the preceding section we have seen that significant gradients in optical and near - infrared colors exist for nearby early - type galaxies . in this section these results are used to constrain the age and metallicity of the underlying stellar populations . because of the age - metallicity degeneracy ( worthey 1994 ) , it is in general difficult to separate these two effects by using only one broad - band color . however since we have observations in seven bands ranging from optical to near - infrared wavelengths with which to probe the underlying sed , an attempt to disentangle the age - metallicity degeneracy can be made using stellar population synthesis models . in this paper , we use the stellar population synthesis code gissel01 ( galaxy isochrone synthesis spectral evolution library , bruzual & charlot 2001 ) to model our photometry data . this library is a special version that includes the filters of both the sdss and 2mass bands , making it possible to directly obtain the magnitudes concerned . this code can be used to produce high - resolution model spectra for a simple stellar population ( ssp ) at different ages with 6 initial metallicities : @xmath106 , @xmath107 , @xmath108 , @xmath109 , @xmath110 , or @xmath111 . spectra for other metallicities can be obtained by linear interpolation between these six base spectra . the code assumes the universal initial mass function ( imf ) obtained by kroupa ( 2001 ) . in most of our discussion , it is assumed that stars in each of the sample galaxies are formed in an instantaneous burst . once an imf is chosen , the stellar population is completely determined by its age and metallicity . the sensitivity of our results to this assumption has been tested by considering a model where the stellar ages have a finite spread ( see section [ subsec gradients ] ) . it is useful to first look at the color - color plots to see how the combination of optical and near infrared data can help to break the age - metallicity degeneracy . as an example , figure [ fig6 ] shows such a plot for ngc 4044 . colors of this galaxy at different radii are plotted as different symbols . the grids are based on the ssp model . the dashed lines represent loci of constant metallicities , while the dotted lines are the loci of constant ages . > from the figure , we can clearly see that ( 1 ) all of the colors vary with radius , indicating color gradients ; ( 2 ) optical colors alone are not effective in breaking the age - metallicity degeneracy ( the constant - age loci are almost parallel to the constant - metallicity loci in the optical color - color plots ) ; ( 3 ) the use of near - infrared band colors can help greatly to break the age - metallicity degeneracy ( the constant - age loci are no longer parallel to the constant - metallicity loci ) , though the errors in the near - infrared colors given by the 2mass are quite large . however , although color - color plots are useful visual presentations , they are not the most effective way to quantify the observational results . on the other hand , fitting the seds represented by all of the observed colors can provide quantitative constraints on the age and metallicity of the underlying stellar population . therefore , in the follows we will only use direct fitting of the seds to quantify the age and metallicity gradients . standard @xmath112 minimization was used to fit the observed seds with the synthesized spectra at different ages and metallicities . because both optical and near - infrared data were used , the approach can put rigorous constraints on either metallicity or age , if the other is kept fixed . to demonstrate this , we generated test seds directly from the spectral synthesis model with given metallicities and ages . each model magnitude was assigned an observational error @xmath113 that is equal to the median observational error at @xmath55 ( defined in section [ subsubsec photometry ] ) . in addition , an uncertainty from the model itself was included in the error budget , @xmath114 . charlot , worthey , & bressan ( 1996 ) estimated this uncertainty by comparing the colors obtained from different stellar evolutionary tracks and spectral libraries . to take this into account , a model error of @xmath111 mag was included in all of our analysis . the combined error for each band was assumed to be the sum in quadrature , @xmath115 tests using other values of @xmath114 in the range from 0.03 to 0.07 mag showed that such a change does not affect our results significantly . figure [ fig7 ] shows contours of @xmath116 in the age - metallicity plane for the test seds obtained by using different combinations of broad - band photometry . the crosses indicate the input values for the age and metallicity of the stellar population . from figure [ fig7](a ) , it is clear that using the combined sdss and 2mass data , the age and metallicity is constrained within a very narrow strip . if either the age or the metallicity is chosen as a fixed value , the other parameter can be determined to within @xmath117 . on the other hand , the age and metallicity are still degenerate in some extent even though we include near infrared bands . comparing figure [ fig7](a ) and ( b ) , one can see that the exclusion of the @xmath19-band data does not have a significant effect . however , if three or more bands are removed , the error bars become substantially larger ( cf . figure [ fig7 ] [ c ] and [ d ] ) . thus , if optical to near - infrared seds are used , much more accurate parameters can be obtained than when only one or two colors are available . this figure also shows that the 2mass data can significantly reduce the size of contours , even though errors of the 2mass data are larger . in conclusion , more photometric bands and the extension of the wavelength coverage are both important for determining the age and metallicity in galaxies . because the age - metallicity degeneracy still exist , it is quite difficult to estimate the age and metallicity simultaneously for individual galaxies . however , for statistical purpose , we worked out the best - fit of age and metallicity for the sample galaxies , which are are widely spread , from age of 3 to 9 gyrs and metallicity of 1 to 2 @xmath118 . these results are consistent with those of trager et al . ( 2000 ) . to obtain the age and metallicity gradients from the observed color gradients , it was assumed that the sed for each projected annulus can be described by a single stellar population , and @xmath112 fitting as described above was used to determine ages and metallicities . firstly , the assumption was made that the observed color gradients are purely due to either age or metallicity by fixing one of the two parameters ( metallicity or age ) at the best - fit value for @xmath55 . secondly , the gradients were modelled by treating both the age and metallicity as free parameters . figure [ fig8 ] shows the results of such analysis for ngc 4044 . panel ( a ) shows the contour of @xmath119 in age - metallicity space for the annulus at @xmath55 , while panel ( b ) shows the observed sed for this annulus together with the best model fit . it is clear from figure [ fig8 ] that the observed sed is well reproduced by a single burst model , and the @xmath120 contour map looks similar to those for the test galaxies shown in figure [ fig7 ] . panel ( c ) shows the best - fit metallicity as a function of radius , where the stellar age is fixed to be 9.1 gyr ( the best fit age at @xmath55 ) across the entire galaxy . a metallicity gradient is clearly seen here . the two vertical dashed lines indicate the radii @xmath63 and @xmath121 , within which the observational seds are not seriously affected by seeing and have a sufficiently high signal - to - noise ratio . the gradient was fitted by a linear function of @xmath122 within this radius range . panel ( d ) shows the age profile where the metallicity is assumed to be a constant across the galaxy . > from panel ( c ) and ( d ) of figure [ fig8 ] , one can see that the radial gradients of either metallicity or age can be well described as a linear relationship in logarithmic space . the two slopes @xmath123 were thus defined in order to characterize the metallicity and age gradients respectively . fitting the observational data for each galaxy assuming either @xmath124 or @xmath125 is zero , the distribution of @xmath125 or @xmath124 was obtained , see figure [ fig9 ] . all galaxies except ngc 6319 were well fitted by this method . because the sed of ngc 6319 could not be fitted by any simple synthesis model , this galaxy was excluded from our analysis . the median values for the remaining 35 galaxies were @xmath126 and @xmath127 respectively . in order to constrain the metallicity and age gradients simultaneously , it was assumed that for each galaxy these gradients can be described by linear relations in log - log space with slopes @xmath125 and @xmath124 . here again , the quantities at the radius @xmath55 of a galaxy were used as the zero point of the relations for the galaxy . for given values of @xmath125 and @xmath124 , a model galaxy was constructed with stellar populations specified by the age and metallicity gradients . the values of @xmath125 and @xmath124 were then determined for each galaxy by matching the predicted seds at different radii with the observed values . in addition , @xmath112 minimization was used to obtain the best - fit values of @xmath124 and @xmath125 , and data points in the radius range from @xmath128 to @xmath129 were used in the fit . figure [ fig10 ] shows the contour maps of @xmath130 in @xmath124-@xmath125 space for two example galaxies . it can be seen that the constraints on @xmath125 and @xmath124 are not equally rigorous for different galaxies . figure [ fig11 ] shows the best - fit values of @xmath125 and @xmath124 for the 35 sample galaxies . in order to obtain statistical constraints on @xmath124 and @xmath125 for the entire sample , the @xmath130 contours for each galaxy were used to obtain a galaxy - possibility distribution in @xmath131 space . then , all galaxies were summed to have a possibility distribution for the sample , @xmath132 . the contours of this distribution function are plotted in figure [ fig11 ] . figure [ fig12 ] shows the projections of this distribution function along the two axes . each can be fitted reasonably well by a gaussian profile . based on this we obtain @xmath133 , @xmath134 , and @xmath135 , @xmath5 . note that the contours shown in figure [ fig11 ] are elongated along the line of age - metallicity degeneracy , and so it is likely that the dispersions in @xmath124 and @xmath125 are dominated by this degeneracy . since the best - fit value for @xmath124 is close to zero , the result is consistent with the hypothesis that all the observed gradients in the seds of these early - type galaxies are due to metallicity variations . indeed , the median value of @xmath125 obtained here is consistent with that obtained above with the assumption of @xmath136 . the result is inconsistent with the hypothesis that these gradients are due to age variations . ccccc & @xmath125 & @xmath137 & @xmath124 & @xmath138 + ssp & -0.25@xmath850.03 & 0.19@xmath850.03 & 0.02@xmath850.04 & 0.25@xmath850.03 + @xmath139 & -0.24@xmath850.03 & 0.21@xmath850.03 & 0.04@xmath850.04 & 0.21@xmath850.03 + @xmath140 & -0.26@xmath850.04 & 0.24@xmath850.03 & 0.05@xmath850.03 & 0.19@xmath850.02 + @xmath141 & -0.26@xmath850.04 & 0.26@xmath850.04 & 0.02@xmath850.03 & 0.17@xmath850.02 + the above results are based on the assumption that stars in an elliptical galaxy were formed in an instantaneous burst with zero age spread . an important question is how sensitive these results are to this assumption . to investigate this , models in which the star formation rate in a galaxy changes with time as @xmath142 were considered , where @xmath143 is a constant characterizing the spread in the ages of the stellar population . as an illustration , we considered three cases where @xmath143 is equal to 0.5 , 1.0 , and 2.0 gyrs respectively . note that the ssp considered above corresponds to @xmath144 . in such models , the two parameters that characterize the stellar population are the initial metallicity and starting time for star formation . the same procedure as for the ssp case was used to determine the median and dispersion for @xmath125 and @xmath124 . results are listed in table [ tab exp - sfh ] , along with the results for the ssp ( @xmath144 ) case . clearly , our conclusions about the age and metallicity gradients are quite insensitive to the assumption of the spread of stellar age . calculations using still larger values of @xmath143 have shown that significant differences occur only when @xmath143 becomes larger than 8 gyr . it must be kept in mind that the results presented above have neglected dust absorption in the host galaxies . since dust obscuration reddens stellar light , the existence of dust changes the color of a galaxy , and can mimic a color gradient if the distribution of dust has a gradient . although many elliptical galaxies contain significant amounts of interstellar matter ( roberts et al . 1991 ) , most is in the form of hot x - ray gas , and only a small fraction ( about @xmath145 m@xmath146 ) is in the form of dust ( kormendy & djorgovski 1989 ; forbes 1991 ; goudfrooij 1994 ; wise & silva 1996 ) . van dokkum & franx ( 1995 ) used hst images to study the dust properties of a sample of 64 early - type galaxies . they found that 48% of them show dust absorption , but that dust absorption is highly concentrated to the nuclear regions . the sizes of these absorption regions at the major axes are much smaller than 1 kpc for almost all their sample galaxies ( only 3 have a dust distribution size of 2 kpc ) . in recent observations of 6 early type galaxies in infrared pah bands by the spitzer space telescope ( pahre et al . 2004 ) 3 of these 6 early - type galaxies exhibit dust features and the dust distribution have a size of about 1 to 3 kpc . however , dust distribution in early type galaxies is still poorly understood ; only simple models are available to investigate possible effect of the dust extinction on color gradients . here we follow wise & silva ( 1996 ) and consider three simple models for the dust distribution . the most likely case is that of a concentrated dust distribution , which is suggested by current observations . if dust distribution has a spatial density distribution of @xmath147 or is even more concentrated , then dust absorption is only significant in the very central region of the galaxy ( @xmath148 ; see fig . 8 in wise & silva 1996 ) . in this case , dust extinction will not have any effect on our results , since we have excluded the central region of a galaxy ( within @xmath63 , @xmath149 1.4 kpc on average ) in our analysis . the second case is a flat density distribution of dust @xmath150 . in this case , dust absorption is almost the same over the whole body of a galaxy , and it will not produce any radial gradients in colors ( or in the seds ) . a third case is a distribution between these two extreme cases . according to wise & silva s analysis , if the dust is distributed as @xmath151 , it will cause a linear color gradient with @xmath152 , which can be confused with true metallicity and/or age gradients . to be conservative , we consider here the effect of dust extinction assuming such a distribution . following charlot & fall ( 2000 ) , we assume a power - law form for the dust absorption curve : @xmath153 . the absorption by dust can then be parameterized with a single parameter @xmath154 , which is the effective optical depth at a wavelength of @xmath155 . another assumption is that @xmath156 decreases linearly with @xmath152 , and approaches _ zero _ in the outer region ( in practice , we set @xmath157 ) . this absorption model can also be derived from wise & silva s ( 1996 ) calculations ( see the dotted line in their fig.8 ) . thus only one additional free parameter , @xmath158 , is needed to describe the absorption of such a diffuse distribution of dust as a function of radius @xmath159 . ccccc @xmath158 & @xmath125 & @xmath137 & @xmath124 & @xmath138 + 0.0 & -0.25@xmath850.03 & 0.19@xmath850.03 & 0.02@xmath850.04 & 0.25@xmath850.03 + 0.1 & -0.19@xmath850.04 & 0.22@xmath850.03 & 0.02@xmath850.04 & 0.23@xmath850.03 + 0.2 & -0.13@xmath850.04 & 0.22@xmath850.03 & 0.00@xmath850.04 & 0.23@xmath850.03 + 0.4 & -0.02@xmath850.04 & 0.21@xmath850.03 & 0.00@xmath850.03 & 0.18@xmath850.03 + the ssp fitting procedure was then carried out by convolving the absorption curves at different radii with the given values of @xmath158 , and the effects on age and metallicity gradients recorded . we have done several tests with different values of @xmath158 , and the results are summarized in table [ tab dust ] . compared with the dust - free case @xmath160 , it is clear that the existence of a diffuse dust distribution can indeed cause a radial color gradient and thereby reduce the degree of metallicity gradients . however , it does not change the average value of the age gradients . if the total amount of diffusing dust is large enough so that @xmath161 , the different levels of absorption at different radii can explain all the observed color gradients in our sample . > from wise & silva s calculation , about @xmath162 of diffuse dust is needed to produce the required @xmath161 . the dust model we have used is likely too simple to describe the dust distribution in galaxies realistically . more precise description of the dust distribution in early - type galaxies may be possible with the future observations , such that from the spitzer space telescope . in summary , the presence of diffuse dust distribution can mimic the color gradients we have observed , and the metallicity gradients we obtained ( @xmath163 ) is only robust if the amount of dust in the diffuse distribution is negligible . simulations of the formation of elliptical galaxies through monolithic collapse ( e.g. carlberg 1984 ) have shown that the metallicity gradients predicted by this model are quite steep , with @xmath125 at least as steep as -0.5 . this is clearly in conflict with our average result of @xmath163 . on the other hand , recent simulations of kobayashi ( 2004 ) showed that metallicity gradients can be significantly flattened by strong mergers . they showed that galaxies formed through major mergers have @xmath164 and none of them have a gradient steeper then @xmath165 , while galaxies formed through minor mergers or monolithic collapse are quite similar , with metallicity gradients of approximately @xmath7 . our average metallicity gradient is between @xmath166 and @xmath167 , but individual gradients span a wide range from -0.6 to above zero . since kobayashi used a similar radius range as ours to measure gradients , we can compare our observational results directly with her simulations . to proceed , we adopt a metallicity gradient of @xmath168 to separate galaxies into two subsamples . according to kobayashi s results , the subsample with metallicity gradients shallower than @xmath168 would be galaxies dominated by a major merger , while those with steeper metallicity gradients are the ones whose formation would be dominated by minor mergers or monolithic collapse . in figure [ fig13 ] we plot the absolute magnitude ( @xmath169 ) distribution for these two subsamples of galaxies . although the average @xmath169 is similar for the two subsamples ( about @xmath170 ) , the two distributions are different . most of the galaxies in the subsample with @xmath171 ( i.e. the one with steeper gradient ) are fainter . since nearly half of our galaxies are s0 galaxies which have luminosities lower than elliptical galaxies , it is possible that the subsample with lower gradients is in fact dominated by s0 galaxies . unfortunately , the sample is too small to allow us to probe the detailed dependence on morphological type . the subsample with lower gradients also includes the three brightest galaxies in the whole sample ( ugc05515 , ugc00797 and ngc 5865 ) . it is interesting that ugc00797 is a cd galaxy and the other two also look like cd galaxies from their environment and radial profiles . if merger plays an important role in determining the metallicity gradient , as discussed above , this result would mean that cd - like galaxies and faint ellipticals form through many minor mergers and/or monolithic collapse . our current sample is still too small to make a strong statement , but this question can be addressed with large samples to be constructed from the sdss and the 2mass . another interesting result is that we obtain a quite large scatter in @xmath125 , @xmath45 . this can not be explained by the uncertainties in measurement or fitting alone , which are approximately an order of magnitude lower . thus , such scatter must be intrinsic . in this study we have not separated galaxies according to morphology ( e.g. lenticulars versus ellipticals , normal ellipticals versus cds ) and environment , and the large scatter may be partly due to morphology and environment dependence . our result is also consistent with the simulation of kobayashi ( 2004 ) , who obtained similar scatter among all the galaxies in her simulation . the average value of the age gradient we found , @xmath172 , suggests that stars over a wide range of radius in early - type galaxies ( 0.4 to 5.0 @xmath55 as probed in this paper ) , have approximately the same age . this is consistent with the monolithic collapse formation scenario , where the bulk of stars in a galaxy formed in a single burst , or with similar star formation history over the whole galaxy . however , the lack of a significant age gradient does not contradict with the hierarchical formation scenario , if early - type galaxies formed their stars early in their progenitors . in this paper we have analyzed the optical and near - infrared surface photometry of a sample of 36 nearby early - type galaxies based on broad - band images obtained from sdss and 2mass . the optical and near - infrared color gradients of each galaxy have been derived and modelled in terms of metallicity and age gradients in their underlying stellar populations . our main results are summarized as follows : 1 . almost all galaxies show shallow color gradients in the five sdss bands and the three 2mass bands . the measured average color gradients in @xmath77 , @xmath78 , and @xmath173 are @xmath174 , @xmath175 , and @xmath102 respectively , in agreement with previous results . 2 . for the first time , seds from optical to near - infrared wavelengths have been used to analyze the stellar population distribution of early - type galaxies . both tests and results have proved the strength of combining sdss and 2mass data in the study of stellar populations . fitting both the age and metallicity simultaneously with stellar population synthesis models , and using @xmath176 and @xmath177 to represent the metallicity and age gradients , a median value of @xmath2 was found for the metallicity gradient , with a dispersion of @xmath3 . corresponding values for the age gradient are @xmath4 and @xmath5 . these results are in good agreement with other observational results and with the expectations of current theories of galaxy formation . 4 . a diffuse distribution of dust [ @xmath151 ] could produce a linear color gradient with @xmath152 , but it will not change the result that the average age gradient is @xmath179 . if any diffusely distributed dust exists , it would decrease metallicity gradients , which would further support the hypothesis that major mergers play an important role in the formation of the early - type galaxies . given the amount of new data now available from the 2mass and the current sdss ( dr2 ) , and data soon to be available from the sdss and the spitzer space telescope in the near future , it will be possible to compile a sample of several hundred large early - type galaxies to carry out the same analysis as presented in this paper . our results presented here suggest that with such a sample , it will be possible to study the age - metallicity gradients of early - type galaxies in unprecedented detail . the authors thank ivan baldry , stephen charlot , daniel thomas , claudia maraston , shude mao , jiasheng huang , peter williams , zhong wang , j. mo , shiyin shen , zhenlong zou , jingyao hu , hongjun su , xu zhou and jianyan wei for their helpful advice and discussions . we also thank an anonymous referee for his / her helpful suggestions . this project is supported by nsf of china no.10273012 , no.10273016 , no.10333060 , no.10473013 and nkbrsf g19990754 . funding for the creation and distribution of the sdss archive has been provided by the alfred p. sloan foundation , the participating institutions , the national aeronautics and space administration , the national science foundation , the u.s . department of energy , the japanese monbukagakusho , and the max planck society . the sdss web site is http://www.sdss.org/. the sdss is managed by the astrophysical research consortium ( arc ) for the participating institutions . the participating institutions are the university of chicago , fermilab , the institute for advanced study , the japan participation group , the johns hopkins university , los alamos national laboratory , the max - planck - institute for astronomy ( mpia ) , the max - planck - institute for astrophysics ( mpa ) , new mexico state university , princeton university , the united states naval observatory , and the university of washington . 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 centre / california institute of technology , funded by the aeronautics and space administration and the national science foundation . michard ( 2002 ) studied the psfs of thinned ccds in different bands , concluding that atmospheric seeing dominates the central range of the psf , and that the effects of seeing disappear at around @xmath180 . however , in the outer regions , the wing of the psf is controlled by both instrumental and atmospheric scattering . the extended wing of the red - band psf is much brighter than that of the blue band in thinned ccds . this is referred to as the ` red halo ' effect ( michard 2002 ; idiart , michard , & de freitas pacheco 2002 ) . the sdss project employs 30 tektronix / site 2048@xmath212048 ccds in a 5 by 6 array for the five bands , with 6 unthinned frontside ccds for the @xmath20-band , and 24 thinned backside ccds for the other four bands . to explore the outer psf wings in these five bands , isolated unsaturated and saturated stars in different edr fields were selected . all of these fields were sky subtracted and the median intensity was measured in concentric circles around each star to avoid the effect of asymmetric features of saturated stars , and of foreground and background objects . the profiles of all stars for each band were re - scaled to the same size and combined . the mean psf profiles in the five bands were then obtained . figure [ fig14 ] shows the radial profiles of the mean psfs in the five bands . the red halo problem is clearly apparent . the @xmath19-band profile is quite different from the other four bands at radii starting from about @xmath181 , consistent with the results of michard ( 2002 ) . thus the red halo effect is expected to affect the surface brightness profiles of galaxies at radii larger than @xmath182 . the psf profiles in the @xmath17 , @xmath14 , and @xmath20 bands agree very well with each other . although the @xmath20-band is much redder than the @xmath19-band , there is no red halo problem because unthinned ccds are used . this confirms that the red halo effect only appears in thinned ccds . the @xmath18-band psf is slightly higher than those of the @xmath17 , @xmath14 , and @xmath20 bands , but much lower than that of the @xmath19-band . in this paper we neglect this small difference between the @xmath18-band and the @xmath17 , @xmath14 , and @xmath20 bands . abazajian , k. , et al . , , 128 , 502 baum , w.a . , thomsen , b. , & morgan , b.l . 1986 , , 301 , 83 bertin , e. , & arnouts , s. 1996 , , 117 , 393 blanton , m.r . , et al . 2001 , , 121,2358 blanton , m.r . , et al . 2003 , , 594 , 186 boroson , t.a . , thompson , i.b . , & shectman , s.a . 1983 , , 88,1707 cardiel , n. , gorgas , j. , s@xmath183nchez - bl@xmath183zquez , p. , cenarro , a.j . , pedraz , s. , bruzual , g. , & klement , j. 2003 , , 409 , 511 carlberg , r.g . 1984 , apj , 286 , 416 carollo , c.m . , danziger , i.j . , & buson , l. 1993 , , 265 , 553 charlot , s. , & fall , s. 2000 , , 539 , 718 charlot , s. , worthey , g. , & bressan , a. 1996 , , 457 , 625 cohen , j.g . 1986 , , 92 , 1039 cohen , m. , wheaton , w.a . , & megeath , s.t . 2003 , , 126 , 1090 cutri , r.m . , 2000 , explanatory supplement to the 2mass second incremental data release ( pasadena : caltech ) de jong , r.s . 1996 , , 313 , 377 de vaucouleurs , g. 1961 , , 5 , 233 de vaucouleurs , g. , de vaucouleurs , a. , corwin , h.g . , buta , r.j . , paturel , g. , & fouque , p. 1991 , third reference catalogue of bright galaxies ( rc3 ) , springer - verlag : new york davies , r.l . , sadler , e.m . , & peletier , r.f . 1993 , , 262 , 650 davis , l.e . , cawson , m. , davies , r.l . , & illingworth , g. 1985 , , 90 , 169 finlator , k. , et al . 2000 , , 120 , 2615 forbes , d.a . 1991 , , 249 , 779 franx , m. , illingworth , g. , & heckman , t. 1989 , , 98 , 538 franx , m. , & illingworth , g. 1990 , , 359 , l41 goudfrooij , p. , hansen , l. , jorgensen , h.e . , norgaard - nielsen , h.u . , de jong , t. , & van den hoek , l.b . 1994 , , 104 , 179 henry , r.b.c . , & worthey , g. 1999 , , 111 , 919 hinkley , s. , & i m , m. 2001 , , 560 , l41 idiart , t.p . , michard , r. , & de freitas pacheco , j.a . 2002 , , 383 , 30 idiart , t.p . , michard , r. , & de freitas pacheco , j.a . 2003 , , 398 , 949 jarrett , t.h . , chester , t. , cutri , r. , schneider , s.e . , & huchra , j.p . 2003 , , 125 , 525 jarrett , t.h . , chester , t. , cutri , r. , schneider , s.e . , skrutskie , m. , & huchra , j.p . 2000 , , 119 , 2498 kobayashi , c. 2004 , , 347 , 740 kormendy , j. , & djorgovski , s. 1989 , , 27 , 235 kroupa , p. 2001 , , 322 , 231 larson , r.b . 1974 , , 166 , 585 macarthur , l.a . , courteau , s. , bell , e. , & holtzman , j.a . 2004 , , 152 , 175 mehlert , d. , saglia , r.p . , bender , r. , & wegner , g. 2000 , , 141 , 449 mehlert , d. , thomas , d. , saglia , r.p . , bender , r. , & wegner , g. 2003 , , 407 , 423 michard , r. 1999 , , 137 , 245 michard , r. 2000 , , 360 , 85 michard , r. 2002 , , 384 , 763 nakamura , o. , et al . 2003 , , 125 , 1682 oke , j.b . , & gunn , j.e . 1983 , , 266 , 713 pahre , m.a . , ashby , m.l.n . , fazio , g.g . , & willn er , s.p . 2004 , , 154 , 229 peletier , r.f . , davies , r.l . , illingworth , g.d . , davis , l.e . , & cawson , m. 1990 , , 100 , 1091 peletier , r.f . , valentijn , e.a . , & jameson , r.f . 1990 , , 233 , 62 petrosian , v. 1976 , , 209 , l1 roberts , m.s . , hogg , d.e . , bregman , j.n . , forman , w.r . , & jones , c. 1991 , , 75 , 751 saglia , r.p . , maraston , c. , greggio , l. , bender , r. , & ziegler , b. 2000 , , 360 , 911 schlegel , d.j . , finkbeiner , d.p . , & davis , m. 1998 , , 500 , 525 scodeggio , m. 2001 , 121 , 2413 silva , d.r . , & elston , r. 1994 , , 428 , 511 smith , j.a . , et al . 2002 , , 123 , 2121 stoughton , c. , et al . 2002 , , 123 , 485 tamura , n. , & ohta , k. 2000 , , 120 , 533 tamura , n. , kobayashi , c. , arimoto , n. , kodama , t. , & ohta , k. 2000 , , 119 , 2134 tamura , n. , & ohta , k. 2003 , , 126 , 596 trager , s.c . , faber , s.m . , worthey , g. , & gonzlez , j.j . 2000 , , 119 , 1645 vader , j.p . , vigroux , l. , lachieze - rey , m. , & souviron , j. 1988 , , 203,217 van dokkum , p.g . , & franx , m. 1995 , , 110 , 2027 white , s.d.m . 1980 , , 191,1 wise , m.w . , & silva , d.r . 1996 , , 461 , 155 worthey , g. 1994 , , 95 , 107 wu , h. , et al . 2002 , , 123 , 1364 young , c.k . , & currie , m.j . 1995 , , 273 , 1141 - [ \mu(r_{50 } ) - \mu_z(r_{50})]$ ] for the 35 sample galaxies at given bins of @xmath185 , together with uncertainties . for clarity , wavelengths are shifted a bit for different radii.,scaledwidth=100.0% ] ( upper panel , obtained assuming @xmath186 ) and age gradients @xmath124 ( lower panel , obtained assuming @xmath187 ) . the median values are @xmath188 and @xmath189 for the two distributions.,scaledwidth=80.0% ]

we present results of an age and metallicity gradient analysis inferred from both optical and near - infrared surface photometry . the analysis is based on a sample of 36 nearby early - type galaxies , obtained from the early data release of the sloan digital sky survey and the two micron all sky survey . surface brightness profiles were derived in each band , and used to study the color gradients of the galaxies . using simple stellar population models with both optical and near infrared colors , we may interpret the color gradients in term of age and metallicity gradients of galaxies . using @xmath0 and @xmath1 to represent the metallicity and age gradients , we found a median value of @xmath2 for the metallicity gradient , with a dispersion @xmath3 . the corresponding values for the age gradients were @xmath4 and @xmath5 . these results are in good agreement with recent observational results , as well as with recent simulations that suggest both monolithic collapse and major mergers have played important roles in the formation of early - type galaxies . our results demonstrate the potential of using multi - waveband colors obtained from current and future optical and infrared surveys in constraining the age and metallicity gradients of early - type galaxies .

introduction sample and data reduction color gradients age and metallicity gradients discussion summary acknowledgements red halo of psf wings of sdss images

galois lattices are a well established topic in applied lattice theory . their importance is widely recognized @xcite , and its applications span across theoretical computer science and discrete mathematics as well as artificial intelligence , data mining and data - base theory . there is a growing interest on the interplay between finite galois lattices and other discrete structures in combinatorics and computer science , and new relationships have been ( and are to be ) discovered between graphs and the related galois lattices . this papers follows this stream and characterizes a class of bipartite graphs by the galois lattice of their maximal cliques . _ distance hereditary _ graphs are graphs with the _ isometric property _ , i.e. , the distance function of a distance hereditary graph is inherited by its connected induced subgraphs . this important class of graphs was introduced and thoroughly investigated by howorka in @xcite . comparability graph _ is the graph of the comparability relation among elements of a poset . in @xcite , cornelsen and di stefano proved that by intersecting the class of distance hereditary graphs with the class of comparability graphs one obtains precisely the comparability graphs of tree - like posets , i.e. , those posets whose transitive reduction is a tree . here we investigate another relation between comparability graphs and distance hereditary graphs : inspired on the one hand by the work of amilhastre , vilarem and janssen @xcite and on the other hand by the work of berry and sigayret @xcite and the work of brucker and gly @xcite . in @xcite , _ galois lattices _ of domino - free bipartite graphs are investigated . in @xcite it is shown that the hasse diagram of the _ galois lattice _ of _ chordal bipartite _ graphs is _ @xcite , while an analogous result is shown in @xcite for the _ clique lattice _ of _ strongly chordal graphs_. both @xcite and @xcite use a dismantlability property of these lattices proved in @xcite . recall that a graph @xmath0 is _ strongly chordal _ if and only if its _ vertex - clique graph _ , namely , the incidence bipartite graph of the maximal cliques of @xmath0 over @xmath1 , is a _ bipartite chordal _ graph and that a graph is _ bipartite chordal _ if it does not contain an induced copy of a chordless cycle on more than four vertices the reader is referred to section [ sec : prel ] for undefined terms and notions . in this paper we study the transitive reduction of the _ galois lattice _ of those bipartite graphs that are chordal ( as in @xcite ) and domino - free ( as in @xcite ) . it follows by theorem [ thm : bdhchar ] in section [ sec : prel ] that these graphs are precisely the bipartite distance hereditary ( bdh for shortness ) graphs , namely , those distance hereditary graphs which are bipartite . essentially in the same way as chordal bipartite graphs are related to strongly chordal graphs , bdh graphs are related to the so called _ ptolemaic graphs_. if @xmath2 denotes the class of chordal graphs , namely , those graphs that do not contain an induced copy of the chordless cycle on more than three vertices , and if @xmath3 is the class of distance hereditary graphs , then the class @xmath4 of ptolemaic graphs is the intersection between @xmath2 and @xmath3 . actually , by the results of @xcite , @xmath4 is the intersection between @xmath5 and @xmath3 , where @xmath5 is the class of strongly chordal graphs . let @xmath6 denote the _ galois lattice _ of a bipartite graph @xmath0 and let @xmath7 denote the _ clique lattice _ of a graph @xmath8 . as shown by wu ( as credited in @xcite ) , if @xmath0 is ptolemaic , then the vertex - clique graph of @xmath0 is a bdh graph . hence there is a map @xmath9 and it is not difficult to see that @xmath10 , where @xmath11 is lattice isomorphism . in a sense , as we show in section [ sec : indirect ] , the converse statement holds as well , namely , there is a mapping @xmath12 that takes a a bdh graph @xmath0 into a ptolemaic graph @xmath13 so that @xmath14 in such a way that @xmath15 for a certain set @xmath16 of _ join - irreducible _ ( or _ meet - irreducible _ ) elements of @xmath6 , where @xmath17 denotes _ order embedding_. in other words the following diagram applies ( and commutes ) : @xmath18_{{\mathcal{c}}(\cdot)}\ar@<1ex>[rrrr]^{\lambda } & & & & \mathbf{bdh}\ar@<1ex>[llll]^{\mu}\ar@{<->}[d]^{{\mathcal{l}}(\cdot ) } \\ \mathbb{t}_{{\mathcal{k}}}\ar@/_/[rrrr]|{\psi_\mu } & & & & \mathbb{t}_{{\mathcal{b}}}\ar@/_/[llll]|{\phi_\lambda } } \ ] ] where @xmath19 and @xmath20 are the classes of tree - shaped clique lattices and galois lattices , respectively , @xmath21 is lattice isomorphism induced by @xmath22 , and @xmath23 is an order embedding induced by @xmath12 . let us recall what is the _ galois lattice _ of a bipartite graph @xmath0 . let @xmath0 have color classes @xmath24 and @xmath25 . a _ biclique _ of @xmath0 is a set @xmath26 which induces a complete bipartite graph . let @xmath27 be the set of the ( inclusionwise ) maximal bicliques of @xmath0 and for @xmath28 let @xmath29 and @xmath30 . @xmath31 and @xmath32 are called the _ shores _ of @xmath33 . throughout the rest of the paper we assume that @xmath0 does not contain universal vertices , where a _ universal vertex _ in a bipartite graph is a vertex that is adjacent to all vertices in the opposite color class . this assumption , while does not cause loss of generality , leads to simpler statements and proofs . following @xcite , we endow @xmath27 by a partial order @xmath34 defined by @xmath35 equivalently , the same partial order can be defined as @xmath36 since @xmath37 . if we extend @xmath27 by adding two dummy elements @xmath38 and @xmath39 acting as bottom and top element respectively , the poset @xmath40 is a lattice known as the galois lattice of @xmath0 . the two dummy elements are respectively defined by @xmath41 in this paper we prove that the shape of @xmath42 can be used to characterize bdh graphs . more precisely , we show the following . [ thm : main ] let @xmath0 be a connected bipartite graph and let @xmath43 be the transitive reduction of @xmath44 . then @xmath43 is a tree if and only if @xmath0 is a bdh graph . otherwise stated : after deleting @xmath38 and @xmath39 , @xmath42 is a tree - like poset . this is a very strong property : for instance , it allows efficient enumeration of linear extensions @xcite . the question of studying bipartite graphs ( binary relations ) whose galois lattice is tree - like ( arborescence - like in a sense ) was raised first in @xcite . here we completely solve the problem from a graph - theoretical view - point . we also give a direct proof that @xmath44 has dimension at most 3 , though this can be derived by known properties of planar posets @xcite . although , as we show in section [ sec : indirect ] , theorem [ thm : main ] can be deduced with some extra work from other known results on graphs and hypergraphs ( by taking the longest dipath in diagram [ eq : diagram2 ] ) the proof we present here is direct and self - contained . the corpus of theoretical and algorithmic machinery on galois lattices would certainly allow a pure lattice theoretical development of the present paper . nevertheless , we prefer to present the result as much combinatorially as possible . for instance , a bipartite graph is the counterpart of the lattice theoretical notion of _ context induced by a binary relation _ and maximal bicliques are the graph theoretical counterparts of _ concepts _ in galois lattices @xcite . in this respect we prefer to look at the covering diagram of a poset as to a directed acyclic graph and to think of `` crowns '' as cycles in an undirected graph . this choice allows us from the one hand to make the paper very self - contained and on the other hand to elicit the purely combinatorial arguments behind our proofs . as an example consider lemma [ lem : htree1 ] . that easy result is also a rather straightforward consequence of the fact that the galois lattice of a sub - context of a given context can be order - embedded canonically into the galois lattice of the context so that the covering diagram of the former is the covering diagram of an induced subposet ( not necessarily a sublattice ) of the latter . the proof we give is nothing but than a specialization of the general arguments used in proposition 32 in @xcite , but has the advantage of avoiding more sophisticated notion that would remain otherwise unused throughout the rest of the paper . also notions such as _ join - irreducibility _ and _ meet - irreducibility _ will be briefly recalled and referred to when they come out and mostly from a diagrammatic approach , because what matters for our purposes is their graphical counterpart . we warn the reader that our results do not use assumptions about the context usually introduced under the clause `` without loss of generality '' , such as that the context is reduced or clarified . while this fact is not a gain of generality from a theoretical view - point it is surely a gain in algorithmic robustness and computational complexity . bdh graphs have very strong structural properties and some of them are highlighted in the characterization by bandelt and mulder @xcite recalled in theorem [ thm : bdhchar ] ( see also @xcite and the monograph @xcite ) . moreover , bdh graphs are related to other very well known classes of graphs : ptolemaic graphs and series parallel graphs . the relation between ptolemaic graphs and bdh graphs has been partially introduced above and will be further pursued in section [ sec : indirect ] . let us discuss here the relation with series parallel graphs . ellis - monaghan and sarmiento @xcite showed that the class of bipartite distance hereditary graphs is a nice nontrivial class of polynomially computable instances for the _ vertex - nullity interlace polynomial _ introduced by arratia , bollobs and sorkin in @xcite , under the name of _ interlace polynomial_. the former authors achieved their result by exploiting a strong topological relationship between bdh graphs and series - parallel graphs . in this paper we prove another deep relation between the two classes , namely bdh graphs are _ fundamental graphs _ of series parallel graphs ( see section [ sec : encoding ] ) . this result leads to an implicit representation of the galois lattice of a bdh graph as a collection of paths in an arborescence . we further discuss this representation in section [ sec : encoding ] , where we exploit it to show how the galois lattice of a bdh graph , and the bdh graph itself , can be efficiently encoded . the encoding of the bdh graph requires @xmath45 space in the worst case , @xmath46 being the order of the graph , still allowing the retrieval of the neighborhood of any vertex in time linear in the size of the neighborhood . moreover , intersections of neighborhoods can be listed in optimal linear time in the size of the intersection , in the worst case . [ [ organization ] ] organization + + + + + + + + + + + + the rest of the paper goes as follows . we first give some preliminary notions in section [ sec : prel ] . in section [ sec : closprop ] we prove a useful property of bdh graphs . such a property is then exploited in section [ sec : char ] to characterize bdh graphs through their galois lattice the property is intimately related with fagin s results @xcite ( as it is discussed in section [ sec : indirect]). in section [ sec : encoding ] we show how the galois lattice of any bdh graph can be encoded as the containment relation among dipaths in an arborescence . in section [ sec : algorithms ] we draw some algorithmic consequences of the encoding and , finally , in section [ sec : indirect ] we give another proof of theorem [ thm : main ] relying , via diagram [ eq : diagram2 ] , on known results on _ ptolemaic graphs _ , _ @xmath47-acyclic hypergraphs _ and _ clique lattices of graphs_. let @xmath48 be a finite set . by a _ hypergraph _ on @xmath48 ( with a little abuse of language ) we simply mean a family @xmath49 of subsets of a given ground set @xmath48 . notice that @xmath49 can contain repeated members . if @xmath49 is a hypergraph , then @xmath50 is the _ bipartite incidence graph of _ @xmath49 over @xmath48 , that is , the bipartite graph with color class @xmath48 and @xmath49 where there is an edge between @xmath51 and @xmath52 if and only if @xmath53 . if @xmath0 is a graph , then @xmath1 denotes its vertex - set if @xmath0 is undirected , while it denotes its node - set if @xmath0 is directed . similarly , @xmath54 denotes the edge - set of @xmath0 if @xmath0 is undirected , and the arc - set of @xmath0 if @xmath0 is directed . the distance between two vertices @xmath55 and @xmath56 of an undirected graph @xmath0 , denoted by @xmath57 , equals the minimum length of a path having @xmath55 and @xmath56 as end - vertices , or is @xmath58 if no such path exists . for a graph @xmath0 and a vertex @xmath59 , @xmath60 ( or simply @xmath61 when @xmath0 is understood ) is the set of vertices adjacent to @xmath56 in @xmath0 . the _ degree _ of @xmath56 is the number of vertices in @xmath60 . the graph induced by @xmath62 is denoted by @xmath63 . let @xmath0 be a directed graph and @xmath56 be a node of @xmath0 . we split the neighborhood of @xmath56 into @xmath64 and @xmath65 . the _ outdegree _ of @xmath56 in @xmath0 , denoted by @xmath66 , is the number @xmath67 of arcs leaving @xmath56 . analogously , the _ indegree _ of @xmath56 in @xmath0 , @xmath68 , is the number of arcs entering @xmath56 . a node in @xmath0 is a _ source _ if its indegree in @xmath0 is zero , a _ sink _ if its outdegree in @xmath0 is zero , or a _ flow - node _ if it is neither a source nor a sink . a _ dipath _ @xmath69 of @xmath0 is a path of @xmath0 with exactly one source in @xmath69 and exactly one sink in @xmath69 . a _ circuit _ @xmath70 in @xmath0 is a cycle in @xmath0 with no source and no sink in @xmath70 . the chordless cycle on @xmath71 vertices is denoted by @xmath72 , and a _ hole _ in a bipartite graph is an induced subgraph isomorphic to @xmath72 for some @xmath73 . domino _ is a subgraph isomorphic to the graph obtained from @xmath74 by joining two antipodal vertices by a chord ( see figure [ fi : h ] ) . a @xmath75-_chordal graph _ is a graph such that every cycle of length at least @xmath76 has at least @xmath77 chords . bipartite @xmath78-chordal graphs are simply called _ chordal bipartite_. a twin of a vertex @xmath56 in a graph is a vertex with the same neighbors as @xmath56 . [ thm : bdhchar ] the following statements are equivalent for a bipartite graph @xmath0 : a. @xmath0 is a bdh graph ; b. [ it : bandeltmount ] @xmath0 is constructed from a single vertex by a sequence of adding pending vertices and twins of existing vertices ; c. @xmath0 contains neither holes nor induced dominoes ; d. @xmath0 is a bipartite @xmath79-chordal graph . if @xmath80 are graphs , we say that @xmath0 is @xmath81-_free _ if @xmath0 contains no induced copy of @xmath82 , @xmath83 . funny enough , after theorem [ thm : bdhchar ] , one can say that a graph is bdh if and only if it is dh - free : just solve the latter acronym as domino hole . in a poset @xmath84 an element @xmath85 covers an element @xmath86 if @xmath87 and @xmath88 @xmath89 @xmath90 . if @xmath91 are incomparable we write @xmath92 . least _ or _ bottom element _ of a poset @xmath84 is the unique element @xmath93 such that @xmath94 for every @xmath95 . this element is usually denoted by @xmath38 . greatest _ or _ top element _ of @xmath84 , usually denoted by @xmath39 , is defined dually . the transitive reduction of a poset @xmath84 is the directed acyclic graph on @xmath24 where there is an arc leaving @xmath86 and entering @xmath85 if and only if @xmath85 covers @xmath86 . the meet and the join operators in a lattice are denoted as customary by @xmath96 and @xmath97 , respectively . an element @xmath86 in a poset @xmath84 is _ meet - irreducible _ ( resp . , _ join - irreducible _ ) if @xmath98 ( resp . , @xmath99 ) implies @xmath100 or @xmath101 . let @xmath102 and @xmath103 be two posets . an order embedding of @xmath102 into @xmath103 is a map @xmath104 satisfying the following condition @xmath105 an _ order isomorphism _ is a bijective order embedding . for a bipartite graph @xmath0 let @xmath106 and recall that @xmath42 denotes @xmath107 . thus @xmath43 is the transitive reduction of @xmath108 . throughout the rest of the paper we represent a biclique @xmath33 of a bipartite graph @xmath0 by the ordered pair of its shores , i.e , we write @xmath109 to mean that @xmath110 , @xmath111 and that @xmath112 induces a complete bipartite subgraph of @xmath0 . moreover , with some abuse of notation , if @xmath113 , then we write @xmath114 to mean that @xmath115 and , analogously , we write @xmath116 for the biclique induced by @xmath117 . a biclique @xmath33 _ dominates _ a biclique @xmath118 if @xmath119 and @xmath120 . and the corresponding galois lattices.,width=321 ] [ fi : h ] as an example , let @xmath0 be either the domino or the @xmath74 ( see figure [ fi : h ] ) . if @xmath0 is the domino , then @xmath121 contains four members : the vertex - sets of the two stars centered at vertices of degree three and the vertex - sets of two squares ; @xmath43 is thus a directed square with one source and one sink ; if @xmath0 is the @xmath74 then the members of @xmath121 are the vertex - sets of the subpaths of @xmath0 of length 2 ; therefore , @xmath43 is a directed @xmath74 with three sources and three sinks . [ rem : usef2 ] for @xmath122 one has @xmath123 if and only if @xmath124 and @xmath125 both have inclusionwise incomparable members . indeed , if @xmath126 , say , then @xmath127 is a biclique of @xmath0 dominating @xmath33 . [ rem : usef3 ] galois lattices are self - dual in the following sense : if @xmath42 is the galois lattice of @xmath0 then @xmath128 ( the lattice dual of @xmath42 ) is the galois lattice of @xmath0 with color classes interchanged . we often use this fact later in the following way : if we prove a property of the lattice for the @xmath24-shores of maximal bicliques , then the same property holds by duality for the @xmath25-shores . if @xmath129 , then there is a biclique @xmath130 such that @xmath131 if and only if @xmath132 for some @xmath133 . analogously if @xmath133 , then there is a biclique @xmath134 such that @xmath135 if and only if @xmath136 for some @xmath129 . using these facts one has that the projections @xmath137 and @xmath138 are actually order isomorphism between @xmath108 and @xmath139 and @xmath140 . hence @xmath141 and @xmath142 ( see also @xcite ) . before proceeding toward the proof of theorem [ thm : main ] , we discuss separately a sort of `` convexity property '' for the neighborhood of the vertices of a bdh graph . such a property is stated in theorem [ thm : convex ] and it is needed to prove the necessity in theorem [ thm : main ] , besides , we deem it interesting on its own . in section [ sec : indirect ] we show that theorem [ thm : convex ] is equivalent to one of fagin s results @xcite , namely , to the first implication in theorem [ thm : fagin ] . let @xmath0 be a connected bdh graph . for @xmath143 , let @xmath144 be the graph defined as follows : * if @xmath56 and @xmath145 are in different color classes , then @xmath146 is @xmath0 ; * if @xmath56 and @xmath145 are in the same color class , then @xmath146 is obtained from @xmath0 by adding a new vertex @xmath147 to the color class of @xmath56 and @xmath145 . vertex @xmath147 is adjacent to every vertex in @xmath148 . [ thm : convex ] let @xmath0 be a bdh graph and let @xmath149 . then @xmath144 is a bdh graph , that is the class of bdh graphs is closed under @xmath150 . without loss of generality , let @xmath151 and to simplify the notation let @xmath152 and @xmath153 . by contradiction suppose that @xmath154 contains induced dominoes or holes while @xmath0 does not . suppose first that @xmath155 is an induced domino in @xmath154 . clearly @xmath156 . let @xmath157 be the number of neighbors of @xmath158 in @xmath155 . then @xmath159 . to see this observe that if @xmath160 then neither @xmath56 nor @xmath145 can belong to @xmath155 , because both @xmath56 and @xmath145 should be adjacent to three vertices in @xmath155 , contradicting the assumption that @xmath155 is domino - free . since @xmath161 it follows that @xmath162 induces a domino in @xmath0 contradicting that @xmath0 is domino - free . thus @xmath159 . let @xmath163 , @xmath164 and @xmath165 be the vertices of @xmath155 contained in @xmath25 , with , say , @xmath163 and @xmath164 adjacent to @xmath158 and @xmath166 . at least one among @xmath56 and @xmath145 does not belong to @xmath167 , otherwise @xmath165 would belong to @xmath148 . suppose first that exactly one of them is in @xmath167 and let , without loss of generality , @xmath168 and @xmath169 . then , @xmath170 and @xmath171 , because @xmath172 . therefore , @xmath173 induces a domino in @xmath0 , still a contradiction . we must conclude that @xmath56 and @xmath145 do not belong to @xmath155 . now the graphs @xmath174 and @xmath175 induced in @xmath154 by @xmath176 and @xmath173 , respectively , are both induced subgraphs of @xmath0 . since @xmath0 is domino - free neither of them is a domino . therefore , @xmath165 must be adjacent to @xmath56 in @xmath174 and to @xmath145 in @xmath175 implying that @xmath177 . the latter contradiction proves that @xmath154 is domino - free . suppose now that @xmath154 contains a hole @xmath8 . the vertex - set of such a hole must contain @xmath158 and contains neither @xmath56 nor @xmath145 ( otherwise @xmath8 would be chorded ) . now the graph induced in @xmath154 by @xmath176 is a subgraph of @xmath0 and therefore , by statement ( iv ) of theorem [ thm : bdhchar ] , @xmath8 possesses at least two chords and such two chords must be incident to @xmath56 . choose one of the possible orientations of @xmath8 and let @xmath178 the end - vertices of the chords incident to @xmath56 met as first and as second , respectively , while traveling on @xmath8 starting from @xmath56 in the chosen orientation . let @xmath179 be the set of vertices met after @xmath180 and before coming back to @xmath56 . now @xmath181 induces a cycle with exactly one chord . a contradiction . let @xmath182 and @xmath183 . [ lem : tb ] if @xmath0 is a bdh graph then so are the graphs @xmath184 and @xmath185 . by duality it suffices to prove the lemma only for @xmath186 . one has @xmath187 if and only if @xmath188 for some @xmath189 and @xmath190 . therefore , @xmath186 is a subfamily of the family @xmath191 and @xmath185 is an induced subgraph of @xmath192 . observe that @xmath193 for a certain graph @xmath194 with color classes @xmath195 and @xmath25 arising from @xmath0 by a repeated application of operation @xmath150 . such an operation preserves the property of being a bdh graph . thus @xmath192 ( and hence @xmath185 ) is bdh . in this section we prove theorem [ thm : main ] . the proof of the _ if part _ is given in section [ sec : if ] while the _ only if part _ is proved in section [ sec : onlyif ] . let us exploit now the structure of bdh graphs to prove the _ if part _ of theorem [ thm : main ] . we remark that the next two results apply to the more general class of domino - free bipartite graphs . [ lem : galois2 ] if @xmath0 is a domino - free bipartite graph then for any @xmath196 such that @xmath197 one has @xmath198 by the sake of contradiction assume that @xmath199 for some @xmath196 and let @xmath200 and @xmath201 . a vertex @xmath202 is called _ heavy _ if it is a universal vertex in the subgraph induced by @xmath203 . the maximality of @xmath204 and @xmath205 implies that there is no heavy vertex in @xmath206 . now @xmath207 and @xmath208 are both nonempty because @xmath209 and @xmath210 . thus , we can pick @xmath211 and @xmath212 . hence , @xmath213 . now pick @xmath214 and @xmath215 ; the latter vertex exists because @xmath216 is not heavy . similarly , pick @xmath217 and @xmath218 ; the latter vertex exists because @xmath163 is not heavy . the subgraph induced by @xmath219 is a domino , contradicting the hypothesis . ; dotted lines represent dipaths and hence chains.,width=321 ] [ fi : galois_forb ] [ lem : noflow ] let @xmath0 be a domino - free bipartite graph and @xmath43 be the transitive reduction of @xmath108 . then , any cycle of @xmath43 that does not contain @xmath38 or @xmath39 has at least six non - flow - nodes . first notice that any cycle of any directed graph has an even number of non - flow nodes because it has as many sources as sinks . moreover , if the graph is acyclic then such a number is positive . transitive reductions are triangle - free . therefore , if @xmath220 is a cycle in @xmath43 , then @xmath220 has at least four nodes and the number of its non - flow - nodes is even and positive . let us prove first that @xmath220 can not have exactly two non - flow - nodes . referring to figure [ fi : galois_forb ] ( a ) , let @xmath221 and @xmath222 be the source and the sink of @xmath220 , respectively , and let @xmath204 and @xmath205 be the neighbors of @xmath221 in @xmath220 . clearly @xmath223 . the existence of dipaths from @xmath204 to @xmath222 and @xmath205 to @xmath222 implies that @xmath224 , contradicting lemma [ lem : galois2 ] . so @xmath220 must have at least four non - flow - nodes and it is of the form shown in figure [ fi : galois_forb ] ( b ) , where @xmath204 and @xmath222 are sources , while @xmath221 and @xmath205 are sinks . as above , the existence of the dipaths from @xmath225 to @xmath221 and from @xmath225 to @xmath222 , @xmath226 , implies that @xmath227 and @xmath228 . so , possibly by replacing @xmath221 by @xmath229 and redefining @xmath220 , we may assume that @xmath230 . by definition , @xmath221 is the only least upper bound . thus @xmath231 and there exists a dipath @xmath232 from @xmath221 to @xmath222 . now @xmath233 contains two cycles with exactly two non - flow - nodes contradicting the preceding part of the proof . we conclude that @xmath220 has at least six non - flow - nodes . we are now ready the prove the _ if part _ of theorem [ thm : main ] . + * proof of the _ if part _ of theorem [ thm : main ] . * we assume @xmath38 and @xmath39 have been deleted from @xmath43 . since @xmath43 is connected we have only to show that it does not contain cycles . suppose by contradiction that @xmath43 contains some cycle , and let @xmath220 be a cycle having the least possible number of non - flow - nodes . let @xmath234 , @xmath235 , be such a number . as @xmath0 is a bdh graph it is domino - free . therefore , by lemma [ lem : noflow ] , @xmath236 . let @xmath237 and @xmath238 be the sources and the sinks of @xmath220 , respectively , as they are met traversing the cycle in a chosen direction . by definition of transitive reduction one has @xmath239 and @xmath240 moreover , for @xmath241 and @xmath242 such that @xmath243 one has @xmath244 . otherwise @xmath245 and one of the two subpaths of @xmath220 connecting @xmath246 and @xmath247 along with the two paths of @xmath43 connecting @xmath248 to @xmath246 and @xmath247 respectively , would define a cycle @xmath249 of @xmath43 with fewer non - flow - nodes than @xmath220 . now for @xmath250 , pick @xmath251 and let @xmath252 and @xmath253 . thus @xmath254 induces a hole in @xmath184 , contradicting corollary [ lem : tb ] . to complete the proof of theorem [ thm : main ] we need some more properties of @xmath43 . [ lem : galois2bis ] let @xmath0 be a bdh graph with at least three vertices . then @xmath93 is a cut - vertex of @xmath0 if and only if @xmath255 . analogously , @xmath256 is a cut - vertex of @xmath0 if and only if @xmath257 . by duality it suffices to prove the lemma only for @xmath93 . suppose that @xmath86 is a cut - vertex of @xmath0 . there is no biclique @xmath33 with @xmath258 , @xmath259 and @xmath260 , otherwise the removal of @xmath86 can not disconnect @xmath0 . hence @xmath261 . conversely , let @xmath262 and suppose , for the sake of contradiction , that @xmath86 is not a cut - vertex of @xmath0 . as the removal of @xmath86 does not disconnect @xmath0 , any two neighbors @xmath85 and @xmath180 of @xmath86 have a common neighbor @xmath95 . this follows by the fact that @xmath263 is connected and hence @xmath264 , because @xmath0 is distance hereditary . therefore , @xmath265 is a biclique and there exists a biclique @xmath266 such that @xmath267 and both shores of @xmath118 have at least two elements . among the bicliques of @xmath121 fulfilling these conditions choose one with @xmath268 as large as possible . let this biclique be @xmath269 . the maximality of @xmath33 implies that @xmath270 is strictly contained in @xmath271 . let @xmath272 . one has @xmath273 by the choice of @xmath270 . indeed , if there were @xmath274 then @xmath275 would be a biclique with @xmath276 . as @xmath86 is not a cut - vertex , @xmath163 is not a pending vertex . hence @xmath277 . the choice of @xmath269 also implies that for each @xmath278 we can find at least one vertex of @xmath270 which is not adjacent to @xmath55 ( otherwise the maximality of @xmath269 would be contradicted ) . on the other hand , for each @xmath279 there is some @xmath280 which is adjacent to @xmath85 . this because @xmath263 is connected and @xmath281 . therefore , we can find @xmath282 and @xmath283 in such a way that @xmath164 is a neighbor of @xmath216 while @xmath165 is not . finally , as @xmath284 , we can find @xmath285 . but now @xmath286 induces a domino in @xmath0 . a contradiction which proves the lemma . recall that in poset that has a bottom element @xmath38 , an _ atom _ is an element of the poset that covers @xmath38 . dually , if the poset has a top element @xmath39 , a _ co - atom _ is an element which is covered by @xmath39 . after this terminology we can say that the cut vertices of @xmath0 are either atoms or co - atoms . we now study the behavior of @xmath287 for @xmath59 . let us begin with an easy but useful property of @xmath288 in the general case . the next lemma proves that if the deletion of a vertex @xmath56 from a maximal biclique of @xmath0 does not cause loss of maximality in the biclique , then @xmath287 inherits from @xmath43 as much adjacency as possible . [ lem : htree1 ] let @xmath0 be a bipartite graph , @xmath289 and @xmath290 . if @xmath291 , then * there is an arc @xmath292 in @xmath287 for every @xmath293 such that @xmath294and @xmath295 is an arc of @xmath43 ; * there is an arc @xmath296 in @xmath287 for every @xmath293 such that @xmath294and @xmath297 is an arc of @xmath43 ; in other words , @xmath298 embeds @xmath287 in @xmath43 as a sub - digraph . by duality , it suffices to prove the statement only when @xmath299 . let @xmath293 be such that @xmath294 . the thesis of the lemma follows by the equivalences listed below : @xmath300 to prove these statements it suffices to recall remark [ rem : usef2 ] and to observe that since @xmath301 one has @xmath302 and @xmath303 . therefore , if @xmath304 and @xmath116 are both maximal in @xmath63 then they must be in the same relation in @xmath305 as @xmath306 and @xmath33 in @xmath42 ( and conversely ) , because this relation is forced by the relation on the @xmath25-shore . the next lemma shows instead that in case deletion of a vertex @xmath56 from a maximal biclique @xmath33 of @xmath0 causes loss of maximality in the biclique , the role of @xmath33 in @xmath42 is not really relevant . [ lem : htree2 ] let @xmath0 be a bipartite graph and let @xmath59 and @xmath293 be such that @xmath307 . if @xmath301 and @xmath33 is not an atom in @xmath6 , then @xmath308 . moreover , if @xmath309 is the unique arc entering @xmath33 in @xmath43 then @xmath310 . analogously , if @xmath311 and @xmath33 is not a co - atom in @xmath6 , then @xmath312 . moreover , if @xmath313 is the unique arc leaving @xmath33 in @xmath43 then @xmath310 . by duality it suffices to prove the statement only when @xmath301 . since @xmath33 is not an atom , @xmath259 . since @xmath307 , there is some @xmath314 which dominates @xmath116 . therefore , @xmath315 and @xmath316 . it follows that @xmath317 in @xmath108 because @xmath316 implies @xmath318 and @xmath319 . consequently @xmath320 and @xmath321 implying that @xmath322 . we therefore conclude that @xmath323 and that @xmath309 is an arc of @xmath43 with @xmath314 . moreover , no other arc @xmath324 for some @xmath325 can exist in @xmath43 . indeed , if such an arc existed then all of the following conditions would hold true : * @xmath326 and @xmath327 are inclusionwise incomparable because @xmath328 , @xmath43 being a transitive reduction ; * @xmath329 because @xmath325 ; * @xmath330 because @xmath331 . the latter two conditions imply that @xmath332 , but since @xmath322 the first one would be contradicted . using standard terminology , as in @xcite , a maximal biclique @xmath33 that satisfies the hypothesis of lemma [ lem : htree2 ] corresponds either to a meet irreducible or to a join irreducible concept in the context associated to the bipartite graph . the results of lemmas [ lem : galois2bis ] , [ lem : htree1 ] , and [ lem : htree2 ] imply : [ thm : retract ] let @xmath0 be a bdh graph and let @xmath333 . then one of the following conditions holds : 1 . [ case : disconnect ] @xmath287 has more connected components than @xmath43 ; 2 . [ case : induced ] @xmath287 is an induced subgraph of @xmath43 ; 3 . [ case : contract ] @xmath287 is a contraction of @xmath43 . let @xmath334 be the set of maximal bicliques @xmath33 containing @xmath56 such that @xmath335 . if @xmath56 is a cut - vertex of @xmath0 , then condition [ case : disconnect ] holds . otherwise , by lemma [ lem : htree1 ] and lemma [ lem : htree2 ] , @xmath287 can be derived from @xmath43 by the following operations : * if @xmath301 and @xmath336 delete @xmath33 if it is a sink in @xmath43 , otherwise contract the unique arc @xmath309 with @xmath314 to the single node @xmath118 ; * if @xmath311 and @xmath336 delete @xmath33 if it is a source in @xmath43 , otherwise contract the unique arc @xmath313 with @xmath314 to the single node @xmath118 . in both cases , either condition [ case : induced ] or condition [ case : contract ] holds . * proof of the _ only if part _ of theorem [ thm : main ] . * let us assume that @xmath43 is a tree and let us prove that @xmath0 is a bdh graph . by theorem [ thm : retract ] , it follows in particular that if @xmath154 is an induced connected subgraph of @xmath0 , then @xmath337 is a contraction of @xmath338 for some connected induced subgraph @xmath339 of @xmath0 such that @xmath154 is an induced subgraph of @xmath339 . hence , @xmath337 is a tree , being the contraction of some subtree of @xmath43 . now , to establish the thesis , it suffices to observe that if @xmath154 is either a domino or a chordless cycle with length greater than four then @xmath337 is not a tree ( see figure [ fi : h ] ) . in this section , we show how the galois lattice of a bdh graph can be realized as the containment relation among directed paths in an arborescence . the results are achieved by further exploiting the interplay between bdh graphs and series - parallel graphs . to this end we first recall the classical two equivalent characterizations of series - parallel graphs . [ thm : spchar ] let @xmath340 be a 2-connected graph with at least two vertices and not isomorphic to @xmath341 . then the following statements are equivalent and characterize series - parallel graphs . 1 . @xmath340 does not contain a homeomorphic copy of @xmath342 , i.e. , the complete graph on four vertices ; 2 . @xmath340 can be recursively constructed starting from a _ digon _ by either adding an edge with the same end - vertices as an existing one or subdividing an existing edge by the insertion of a new vertex . recall that a _ digon _ is a graph formed by two edges with the same end - vertices . it is trivially a series - parallel graph . the operations described in statement ( b ) of theorem [ thm : spchar ] are referred to as _ parallel extension _ and _ series extension _ respectively . statement ( a ) in theorem [ thm : spchar ] is duffin s characterization by forbidden minors . as we are going to show , the close resemblance between bandelt and mulder s construction ( statement ( ii ) of theorem [ thm : bdhchar ] ) and statement ( b ) above is not merely formal . to this end we need the notion of _ fundamental graph _ of a graph which we briefly recall here . in a connected graph a co - tree is the subgraph spanned by the complement of the edge - set of a spanning tree . if @xmath343 is a spanning tree of @xmath340 its co - tree is denoted by @xmath344 . given a connected undirected graph @xmath340 and one of its spanning trees @xmath343 , the _ fundamental graph of _ @xmath340 is the bipartite graph @xmath345 with color classes @xmath346 and @xmath347 where there is an edge between @xmath348 and @xmath349 if @xmath350 , @xmath351 being the edge - set of the unique cycle in the graph spanned by @xmath352 . such a cycle is the so called _ fundamental cycle through _ @xmath353 _ with respect to _ @xmath343 . it can be shown that if @xmath340 is 2-connected then @xmath345 is connected . moreover , @xmath345 does not determine @xmath340 in the sense that non - isomorphic graphs may have isomorphic fundamental graphs . as the fundamental graph is a matroid theoretical tool we refer the interested reader to @xcite for more details . now we just need to recast the effect of series and parallel extensions on a graph @xmath340 on its _ fundamental graph _ with respect to a given tree and to observe that adding pending vertices and twins are counterparts of the above operations . these effects are summarized in the following table . [ table : spbdh ] .the effects of series and parallel extension on @xmath340 on its fundamental graph @xmath345 with color classes @xmath354 and @xmath355 . [ cols="<,<,<",options="header " , ] the following result is now just a remark . [ thm : spbdh ] a connected bipartite graph @xmath0 with color classes @xmath24 and @xmath25 and at least two vertices is a bdh graph if and only if it is the fundamental graph of a 2-connected series - parallel graph . the _ if part _ is proved by induction on the order of @xmath0 . the assertion is true when @xmath0 has two vertices because @xmath341 is a bdh graph and at the same time is also the fundamental graph of a digon . let now @xmath0 have @xmath356 vertices and assume that the assertion is true for bdh graphs with @xmath357 vertices . by bandelt and mulder s construction ( statement ( ii ) of theorem [ thm : bdhchar ] ) @xmath0 is obtained from a bdh graph @xmath358 either by adding a pending vertex or a twin . let @xmath359 be a series - parallel graph having @xmath358 as fundamental graph with respect to some spanning tree . since , by table 1 , the latter two operations correspond to series or parallel extension of @xmath359 , the result follows by statement ( b ) of theorem [ thm : spchar ] . conversely , let @xmath0 be the fundamental graph of a series - parallel graph @xmath340 with respect to some tree @xmath343 . by statement ( b ) of theorem [ thm : spchar ] and table 1 , @xmath0 can be constructed starting from a single edge by either adding twins or pending vertices . therefore , @xmath0 is a bdh graph by bandelt and mulder s construction ( statement ( ii ) of theorem [ thm : bdhchar ] ) . as credited by syslo @xcite , shinoda , chen , yasuda , kajitani , and w. mayeda , proved that series - parallel graphs can be completely characterized by a property of their spanning trees . they proved that every spanning tree of a series - parallel graph @xmath340 is a depth - first search tree of a 2-isomorphic copy of @xmath340 , where 2-isomorphism of graphs ( in the sense of whitney @xcite ) is isomorphism of binary vector spaces between cycle - spaces of graphs . we can avoid to enter details of such notions and we can content ourselves of restating in our terminology a direct consequence of the result . [ thm : syslo ] let @xmath0 be the fundamental graph of a 2-connected graph with color classes @xmath24 and @xmath25 with , say , @xmath24 being the edge - set of a spanning tree . then there exist a graph @xmath359 , a spanning tree @xmath360 of @xmath359 and an orientation @xmath361 of @xmath359 such that * @xmath362 , * @xmath363 is an arborescence , * for each @xmath93 , @xmath364 is the arc - set of a directed circuit in @xmath365 and , consequently , @xmath366 is the arc - set of a directed path in @xmath363 , if and only if @xmath0 is the fundamental graph of a series - parallel graph . recall that an _ arborescence _ is a directed tree with a single special node distinguished as the _ root _ such that , for each other vertex , there is a dipath from the root to that vertex . syslo himself gave a constructive algorithmic proof of the above result @xcite . a bipartite graph @xmath0 satisfying the third condition of theorem [ thm : syslo ] will be called a _ path - arborescence _ bipartite graph and the arborescence @xmath363 whose existence is asserted in the theorem will be referred to as a _ supporting arborescence_. in general such an arborescence will not be unique . remark that series - parallel graphs form a self - dual class of planar graphs , therefore theorem [ thm : syslo ] holds simultaneously for the color class @xmath25 , @xmath25 being the edge - set of a co - tree . we can state now the following straightforward consequence of theorem [ thm : spbdh ] and theorem [ thm : syslo ] . [ cor : dpt ] if @xmath0 is a connected bdh graph with color classes @xmath24 and @xmath25 , then @xmath0 is a _ path - arborescence _ bipartite graph . let now @xmath0 be a bdh graph with color classes @xmath24 and @xmath25 . by corollary [ cor : dpt ] , there exists an arborescence @xmath367 with root @xmath368 and @xmath354 that supports @xmath0 and such that , for each @xmath256 , the set @xmath369 is the arc set of a directed path in @xmath367 . moreover , for @xmath370 , @xmath271 is a set of dipaths each containing the arc @xmath371 . we now show that the inclusion - wise maximal such paths along with their pairwise intersections give the containment relation of the second coordinate of bicliques in @xmath108 , which is in turn isomorphic to @xmath108 . this fact allows an efficient encoding of the galois lattice . notice that @xmath372 naturally induces a partial order @xmath373 on @xmath24 ( the _ arborescence order _ ) where @xmath374 if @xmath371 is an arc of each dipath containing the root and @xmath375 . obviously , dipaths are intervals in this order and conversely . denote by @xmath376\subseteq x$ ] the set of elements in the interval defined by the dipath having @xmath371 and @xmath377 as end - arcs . we also say that a subset @xmath378 of @xmath24 spans a dipath if @xmath379 is a dipath in @xmath367 . we show that the galois lattice of a bdh graph is completely determined by some pairwise intersections of neighborhoods , plus some simple neighborhoods . [ cor : encoding ] let @xmath0 be a connected bdh graph with color classes @xmath24 and @xmath25 . let @xmath380 then @xmath381 we show that @xmath382 and this is enough to prove the result because by , @xmath383 observe in the first place that @xmath384 let @xmath385 be a maximal biclique of @xmath108 , and let @xmath386 . we first show that if @xmath387 then @xmath388 , i.e. , either @xmath389 for some @xmath370 or @xmath390 for some @xmath391 . let @xmath367 be a supporting arborescence such that @xmath271 is mapped onto a path in @xmath367 , for each @xmath370 . since each @xmath271 spans a dipath in @xmath367 , and since the intersections of dipaths always is a dipath , it follows that @xmath392 is the arc - set of some nonempty dipath @xmath69 of @xmath367 . let @xmath393 $ ] , with @xmath394 , be the dipath spanned by @xmath395 , for each @xmath396 . we observe that if there exist two end - arcs @xmath397 and @xmath398 , with @xmath399 , such that @xmath400 , then @xmath401 . thus , since @xmath402 is not empty , @xmath373 defines a total order on @xmath397 s , because all the @xmath397 s are pairwise comparable . let @xmath403 be the maximum w.r.t . @xmath373 among all @xmath397 s , where @xmath404 $ ] is the path spanned by @xmath405 . moreover , @xmath406 for each @xmath407 , otherwise it would be @xmath408 and @xmath409 . hence , for each @xmath410 there exists an arc @xmath411 that is the maximum element in @xmath412 \cap [ a_j , b_j]$ ] . since @xmath413 $ ] , for each @xmath410 , then @xmath373 defines a total order on @xmath414 , for @xmath410 . let @xmath415 be the minimum among @xmath414 , for @xmath410 . it follows that @xmath416 , since @xmath417 for each @xmath410 , where it can also happen that @xmath418 . we conclude that @xmath388 . let us prove , conversely , that if @xmath388 , then @xmath419 for some @xmath129 . this fact implies @xmath420 by . this is proved by a more general argument , in fact ( see also @xcite , page 19 ) for each set @xmath421 , with @xmath422 , there exists a maximal biclique @xmath423 with @xmath424 . let @xmath425 : we have that @xmath426 is a biclique , and any biclique @xmath427 that dominates @xmath428 , since @xmath424 , must have @xmath429 , hence @xmath430 . thus , there must be a maximal biclique @xmath431 , for some @xmath432 . this fact completes the proof . the proof of corollary [ cor : encoding ] shows that @xmath6 is isomorphic to the containment relation of a set of paths in an arborescence . this implies that @xmath6 has dimension at most 3 . the bound derives from the following slightly more general consideration : the containment order among paths in an arborescence has dimension at most 3 . we show that the containment order among paths in an arborescence @xmath343 is a subposet of poset @xmath433 , where @xmath434 is the height of @xmath343 and @xmath435 is the restriction of the natural order of integers to @xmath436 . we associate each path @xmath437 in @xmath343 with the pair @xmath438 , where @xmath439 is the maximum edge in @xmath437 with respect to @xmath373 , and @xmath440 is the distance of @xmath437 from the root of @xmath343 , i.e. , @xmath440 is 1 if @xmath437 starts from the root , @xmath440 is 2 if @xmath437 starts from a child of the root , and so on . it is immediate to see that @xmath437 contains @xmath441 if and only if @xmath442 and @xmath443 . it is well known that the arborescence order @xmath373 has dimension at most 2 ( see @xcite ) , therefore the dimension of the product is at most @xmath444 . thus , we can state that : [ coro : dim ] if @xmath0 is a bdh graph , then @xmath445 has dimension at most 3 . since @xmath445 is a tree - like poset if @xmath0 is a bdh graph , corollary [ coro : dim ] can also be obtained directly from a result by trotter and moore ( see @xcite ) , asserting that a tree - like poset has dimension at most 3 . notice that , there are containment orders among paths in an arborescence that are not isomorphic to the galois lattice of any bdh graph . for example , the galois lattice of a domino is isomorphic to the containment among sets @xmath446 , and it is immediate to see that these sets are the edge sets of three subpaths of a path with edges @xmath447 , which is clearly an arborescence . in this section we discuss some of the algorithmic consequences of the encoding described in section [ sec : encoding ] , and exploited in the proof of corollary [ cor : encoding ] . by the results of @xcite , there exists an algorithm that given a bdh graph computes a supporting arborescence @xmath367 for @xmath0 as in corollary [ cor : dpt ] . the algorithm runs in almost linear time in the size of @xmath0 , that is in time @xmath448 where @xmath449 is the number of edges of @xmath0 and @xmath450 is an inverse of the ackermann function , which grows very slowly and behaves essentially as a small constant even for very large values of its arguments . we propose a compact encoding of the bdh graph that requires @xmath45 space in the worst case , where @xmath46 is the order of @xmath0 . the retrieval of the neighborhood of any vertex requires linear time in the size of the neighborhood . moreover , intersection of neighborhoods can be listed in optimal linear time in the size of the intersection , in the worst case . at the same time each vertex @xmath451 can be associated to a pair of edges @xmath452 so that @xmath395 is mapped into the path from @xmath397 to @xmath453 in @xmath367 . recall that the partial order @xmath373 has linear dimension 2 , so each vertex / edge in the arborescence can be equipped with a pair of labels in @xmath454 so that relation @xmath373 between two edges is verified in constant time . after fixing an arbitrary ordering on the outgoing edges for each vertex in @xmath343 , the pair of labelings is defined by two preorder numberings of @xmath343 , one obtained by visiting at each vertex outgoing edges from left to right and the other one obtained by visiting at each vertex outgoing edges from right to left . this gives an encoding of @xmath0 that allows to answer the following queries in optimal worst case time , where @xmath370 and @xmath421 : 1 . [ algo : neigh ] list @xmath271 , in time @xmath455 ; 2 . [ algo : check ] check whether @xmath456 , in time @xmath457 ; 3 . [ algo : list ] list @xmath458 , in time @xmath459 ; 4 . [ algo : maximal ] check whether @xmath460 is a maximal biclique , in @xmath459 worst case time . in lattice theoretical terminology , query [ algo : maximal ] corresponds to checking whether @xmath461 , where @xmath462 denotes concept polarity as defined in @xcite . note that the size of the encoding is only @xmath45 , while the number of edges in a bdh graph can be @xmath463 , and still allows the computation of the maximal biclique containing a given set @xmath464 on one side in time linear in the in the number of vertices in the biclique . the algorithm to solve query [ algo : list ] ( queries [ algo : neigh ] and [ algo : check ] are special cases of query [ algo : list ] ) is described in figure [ fi : intersection ] , and follows the same argument as in the first part of the proof of corollary [ cor : encoding ] . let @xmath465 , and let @xmath466 , for @xmath467 , be the end - arcs of the path associated to @xmath468 in @xmath343 . pr : resilientspanner given @xmath421 , compute @xmath458 . + we assume the arborescence @xmath343 is given , and a data structure for solving lowest common ancestor queries + according to @xmath373 , as described in @xcite , has been built . + @xmath466 , for @xmath469 , are the end - arcs of the path in @xmath343 associated to @xmath395 + + . * let * @xmath470 + . * for * @xmath471 to @xmath472 + . [ line : magg ] * if * @xmath473 + . * let * @xmath474 + . [ line : nomagg ] * else if * @xmath475 + . * return * @xmath476 + . [ line : lca ] * let * @xmath477 + . * if * @xmath478 + . [ line : path ] * return * @xmath479 $ ] + . * else * + . * return * @xmath476 + [ fi : intersection ] it can be seen that the complexity of algorithm ` neighborintersection ` is + @xmath459 , since tests in lines [ line : magg ] and [ line : nomagg ] are performed in constant time starting from the encoding of the 2-dimensional partial order @xmath373 . the computation of the lowest common ancestor @xmath480 at line [ line : lca ] is computed in time @xmath457 using the data structure proposed in @xcite , which is built in @xmath45 time . path retrieval in line [ line : path ] requires @xmath481 worst case time , starting from @xmath482 and following parent pointers in the arborescence @xmath343 up to @xmath483 . in order to solve query [ algo : check ] , we can still use algorithm ` neighborintersection ` , without listing the path in line [ line : path ] , thus requiring @xmath457 worst case time . query [ algo : maximal ] can be solved using the same algorithm , thanks to observation ( [ eq : ganterwille1 ] ) , provided that the same encoding is stored both for side @xmath24 and for side @xmath25 . in fact , @xmath460 is a maximal biclique if and only if @xmath484 , where @xmath485 , that can be checked by computing @xmath402 and then computing @xmath486 , i.e. , solving two queries of type [ algo : list ] . we prove here theorem [ thm : main ] by exploiting existing results , that is by taking the longest path between * bdh * and @xmath20 in diagram ( [ eq : diagram2 ] ) . to this end we need some more terminology on hypergraphs and ptolemaic graphs . the reader is referred to the monographs @xcite and @xcite . let @xmath49 be a hypergraph on @xmath48 . given two arbitrary linear orders of @xmath48 and @xmath49 , let @xmath487 be the @xmath488-matrix whose rows correspond ( in the order chosen for @xmath49 ) to the members of @xmath489 , the columns correspond ( in the order chosen for @xmath48 ) to the vertices of @xmath489 and where @xmath490 if the @xmath491-th element of @xmath48 is in the @xmath407-th member of @xmath49 and @xmath492 otherwise . clearly , if @xmath493 is a @xmath494-matrix , we can reverse the construction by associating with @xmath493 the hypergraph @xmath495 on the index - set @xmath48 of the columns and whose members are the supports of the rows of @xmath493 , regarded as subsets of @xmath48 . a _ clutter _ is a hypergraph whose members are inclusion - wise incomparable . the hypergraph @xmath496 is the collection obtained by closing @xmath489 under intersection , namely , @xmath497 if and only if either @xmath52 or @xmath498 is the intersection of two or more members of @xmath49 . by @xmath499 we denote the clutter obtained from @xmath49 as follows : first pairwise equal members are identified into a unique member and then only the inclusion - wise maximal members are retained . if @xmath500 is a family of subsets of a given common ground set , let @xmath501 denote the collection consisting of the inclusion - wise maximal members of @xmath500 . if @xmath0 is a bipartite graph with color classes @xmath24 and @xmath25 we associate with @xmath0 the two hypergraphs @xmath502 and @xmath503 on @xmath24 and @xmath25 , respectively , called _ neighborhood hypergraphs of @xmath0 _ , given by @xmath504 and @xmath505 . the hypergraphs @xmath506 and @xmath507 are called the _ maximal neighborhood systems _ of @xmath0 . let @xmath0 be a graph . the clutter @xmath508 is the clutter consisting of the maximal cliques of @xmath0 . hence @xmath509 induces a map @xmath434 that sends isomorphism classes of graphs into isomorphism classes of clutters . on the other hand with any hypergraph @xmath49 on @xmath48 we can associate the graph @xmath510 with vertex set @xmath48 and where two vertices @xmath55 and @xmath56 are joined by an edge if there is a member @xmath498 of @xmath49 containing both . the graph @xmath511 is called the _ 2-section of @xmath49_. as shown by the clutter @xmath512 the map @xmath434 is not in general a bijection , because @xmath513 and hence @xmath514 . this motivates the following notion : a hypergraph @xmath49 is _ conformal _ if the maximal cliques of its 2-section coincide with the maximal members of @xmath49 , that is @xmath515 @xcite . it follows that if @xmath434 is a bijection when its image is restricted to the class of conformal clutters . if @xmath516 and @xmath517 are two @xmath494 matrices , we say that @xmath517 contains a copy of @xmath516 if the rows and the columns of @xmath517 can be permuted so that the permuted matrix contains @xmath516 as a submatrix . for an integer @xmath518 , let @xmath519 be the matrix whose entries satisfy @xmath520 . we now collect some very well known facts about _ totally balanced _ hypergraphs , namely , those hypergraphs whose matrix is @xmath521 free for all @xmath518 . all the characterizations listed below can be found in @xcite . a. [ com : a ] @xmath49 is a totally balanced hypergraph , i.e. , @xmath522 is @xmath521-free for @xmath518 , if and only if @xmath510 is a strongly chordal graph . b. [ com : b ] @xmath0 is a bipartite chordal graph if and only if its neighborhood hypergraphs are totally balanced . [ rem : tbneigh ] using the fact that totally balanced hypergraphs are conformal , we obtain the following fact : _ let @xmath0 be a bipartite graph with color classes @xmath24 and @xmath25 . then @xmath0 is chordal if and only if each of the 2-sections of the following hypergraphs is strongly chordal : @xmath502 , @xmath506 , @xmath503 and @xmath507_. ptolemaic graphs , as shown by howorka @xcite , are precisely chordal distance hereditary graphs . there is an intimate relationship between ptolemaic graphs and bdh graphs , made explicit by bandelt and mulder @xcite , and peled and wu @xcite . in particular , bandelt and muller show that if @xmath0 is a bdh graph then the graph obtained by completing certain level sets is distance hereditary and chordal and thereby ptolemaic by howorka s characterization . on the other hand , as mentioned above , wu s result asserts that the vertex - clique graph of a ptolemaic graph , is a bdh graph . besides their own theoretical importance , ptolemaic graphs ( and hence bdh graphs ) , deserve a special role in the theory of relational database , as shown by datri and moscarini in @xcite . in the early eighty , deep investigations of theoretical properties of relational databases @xcite led to a refinement of the notion of cycles in hypergraphs yielding various degrees of acyclicity @xcite . among them , the notion of @xmath47-acyclic hypergraphs relates directly with ptolemaic graphs in that the 2-section graph of @xmath47-acyclic hypergraph is a ptolemaic graph @xcite . since the bachman diagram of a @xmath47-acyclic hypergraph @xmath489 , namely the hypergraph obtained by closing the edge - set of @xmath489 under intersection , is the clique - lattice of its 2-section graph with top and bottom removed , and since by the results of fagin @xcite , the bachman diagram of a hypergraph is a tree if and only if the hypergraph is @xmath47-acyclic , it follows that the clique - lattice of a ptolemaic graph is tree - shaped . ueheara and uno @xcite , obtained the same result from another perspective : they proved that cliques of a ptolemaic graph have the remarkable property of being _ laminar_. recall that a family of sets is said to be _ laminar _ if given any two sets of the family , then either such two sets are disjoint or they are inclusion - wise comparable . if @xmath528 is a laminar family , then @xmath529 is a tree - like poset this is another way of stating a classical results of edmonds and giles@xcite. in view of fagin s result and the result of uehara and uno , we conclude that the clique lattice of a ptolemaic graph is tree shaped . to see how theorems [ thm : fagin ] and [ thm : ueuno ] imply , via diagram ( [ eq : diagram2 ] ) , theorem [ thm : main ] , we proceed as follows : we first prove that the maps @xmath22 and @xmath12 that make diagram [ eq : diagram2 ] commuting exist . they are indeed obtained in proposition [ prop : mu ] by specializing the characterization given in remark [ rem : tbneigh ] to the subclasses @xmath4 and @xmath531 . then we show in theorem [ thm : bridge ] that @xmath22 and @xmath12 actually make the diagram commuting . to this end we need some intermediate results on galois lattices of bdh graphs . [ prop : mu ] if @xmath0 is ptolemaic graph then @xmath532 is a bdh graph . if @xmath0 is a bdh graph with color classes @xmath24 and @xmath25 , then each of the 2-sections of the following hypergraphs is ptolemaic : @xmath502 , @xmath506 , @xmath503 and @xmath507 . the first part is wu s result @xcite . let us prove the second part and let @xmath154 be an induced subgraph of @xmath0 with color classes @xmath533 and @xmath402 . let @xmath534 be the @xmath494-matrix with rows indexed by @xmath533 columns indexed by @xmath402 defined by the adjacency of the vertices in @xmath533 and @xmath402 . observe that @xmath534 is a submatrix of both @xmath535 and @xmath536 ( this is just a matter of checking definitions ) . since @xmath0 is bdh it is domino - free . up to a permutation of rows and columns , the adjacency matrix of a domino is the matrix @xmath537 defined in . hence @xmath502 , @xmath506 , @xmath503 and @xmath507 are @xmath47-acyclic and therefore the 2-sections are ptolemaic by theorem [ thm : damo ] ( because of the conformality of @xmath47-acyclic hypergraphs ) . [ lem : paoloproof ] let @xmath0 be a bdh graph with color classes @xmath24 and @xmath25 . if @xmath93 is such that @xmath540 and @xmath541 , then @xmath542 arises from @xmath43 by contracting some arcs . the same results holds if @xmath256 satisfies the hypotheses . let @xmath543 . by the hypotheses , @xmath544 is not the @xmath24-shore of any biclique in @xmath27 , and therefore no member of @xmath545 is an atom in @xmath6 ( by lemma [ lem : galois2bis ] ) . since @xmath541 , it follows that @xmath546 . by lemma [ lem : htree1 ] , @xmath547 is a sub - poset ( not necessarily a sub - lattice ) of @xmath6 . the vertices of @xmath542 correspond bijectively to vertices of @xmath548 . however , by lemma [ lem : htree2 ] , the vertices in @xmath545 are meet - irreducible in @xmath6 and are covered in @xmath6 by bicliques of @xmath549 . in other words , for each @xmath550 there is a unique arc @xmath309 entering @xmath33 and with @xmath551 . therefore , by contracting every such arc yields @xmath542 . by duality we obtain the statement for the @xmath25-shore . in the latter case however meet - irreducible is replaced by join - irreducible . the following theorem sets a bridge between the clique lattice of a graph @xmath0 and the galois lattice of the vertex - clique bipartite graphs of @xmath0 and is strong enough to make diagram commuting . for a bipartite graph @xmath0 with color classes @xmath24 and @xmath25 , let @xmath552 and @xmath553 be the set of vertices in @xmath24 and @xmath25 respectively , satisfying the hypotheses of lemma [ lem : paoloproof ] . after lemma [ lem : paoloproof ] , the following fact is straightforward . [ thm : bridge ] if @xmath0 is ptolemaic graph , then there exists a lattice isomorphism @xmath21 such that @xmath556 . if @xmath0 is a bdh graph with color classes @xmath24 and @xmath25 , then there are lattice isomorphisms @xmath557 and @xmath558 such that @xmath559 and @xmath560 . let @xmath0 be ptolemaic , let @xmath561 . suppose that @xmath562 . if @xmath563 denotes the set of all maximal cliques of @xmath0 containing @xmath340 , then @xmath564 is a biclique of @xmath565 and @xmath340 is the @xmath24-shore of a biclique in @xmath565 . conversely , if @xmath566 , for some @xmath567 and some @xmath568 , then @xmath340 is precisely the intersection of the cliques in @xmath569 and therefore it is in @xmath570 . in other words the isomorphism @xmath21 is the one given in . let @xmath571 and @xmath572 . since @xmath0 is bdh it is bipartite chordal and therefore @xmath503 and @xmath502 are both conformal ( see also remark [ rem : tbneigh ] ) . hence @xmath573 and @xmath574 . clearly @xmath575 and @xmath576 where @xmath11 is graph isomorphism . now @xmath577 and @xmath578 , where now @xmath11 is lattice isomorphism , and the two isomorphisms @xmath557 and @xmath558 are explicitly given by and with @xmath579 and @xmath580 in place of @xmath0 , respectively . we are almost done . to obtain a proof theorem [ thm : main ] , it is now sufficient to resort to theorem [ thm : bridge ] , to observe that by proposition [ prop : mu ] , both @xmath581 and @xmath582 are ptolemaic , to invoke either fagin s or uehara and uno s result and finally to apply lemma [ lem : tree - tree ] . r. arratia , g. bollobs , g. sorkin . _ the interlace polynomial : a new graph polynomial_. proceedings of the eleventh annual acm - siam symposium on discrete algorithms , san francisco , ca , jan . 2000 , 237245 . r. belohlvek , g. de baets , j. outrata , v. vychodil . _ trees in concept lattices_. proceedings of the 4th int . conf . on modeling decisions for artificial intelligence , mdai 2007 , kitakyushu , japan , aug . 2007 , volume 4617 of lncs , springer , 174184 . a. berry , a. sigayret . _ dismantlable lattices in the mirror_. proceedings of the 11th international conference on formal concept analysis , icfca 2013 , dresden , germany , may 2013 . volume 7880 of lncs , springer , 4459 . a. brandstdt , v. g. le , j. p. spinrad . _ graph classes : a survey_. siam monographs on discrete mathematics and applications , society for industrial and applied mathematics ( siam ) , philadelphia , pa , 1999 .

we give a complete characterization of bipartite graphs having tree - like galois lattices . we prove that the poset obtained by deleting bottom and top elements from the galois lattice of a bipartite graph is tree - like if and only if the graph is a bipartite distance hereditary graph . by relying on the interplay between bipartite distance hereditary graphs and series - parallel graphs , we show that the lattice can be realized as the containment relation among directed paths in an arborescence . moreover , a compact encoding of bipartite distance hereditary graphs is proposed , that allows optimal time computation of neighborhood intersections and maximal bicliques . + * keywords * : galois lattice , transitive reduction , distance hereditary graphs , bicliques , series - parallel graphs .

introduction preliminaries a closure property characterizing chordal, domino-free bipartite graphs by their galois lattices encoding @xmath6 efficiently computing (maximal) bicliques an indirect proof of theorem

in recent years there has been an increasing interest in derivative securities at the interface of finance and insurance . _ structured products _ such as risk bonds , asset - backed securities and weather derivatives are end - products of a process known as _ securitization _ that transforms non - tradable risk factors into tradable financial assets . developed in the u.s . mortgage markets , the idea of pooling and underwriting risk that can not be hedged through investments in the capital markets alone has long become a key factor driving the convergence of insurance and financial markets . structured products are often written on non - tradable underlyings , tailored to the issuers specific needs and traded `` over the counter '' . insurance companies , for instance , routinely sell weather derivatives or risk bonds to customers that can not go to the capital markets directly and/or seek financial securities with low correlation with stock indices as additions to diversified portfolios . the market for such claims is generally incomplete and illiquid . as a result , many of the standard paradigms of traditional derivative pricing theory , including replication arguments do not apply to structured products . in an illiquid market framework , preference - based valuation principles that take into account characteristics and endowment of trading partners may be more appropriate for designing , pricing and hedging contingent claims . such valuation principles have become a major topic of current research in economics and financial mathematics . they include rules of pareto optimal risk allocation ( @xcite , @xcite ) , market completion and dynamic equilibrium pricing ( @xcite , @xcite ) and , in particular , utility indifference arguments ( @xcite , @xcite , @xcite , @xcite , @xcite , ... ) . the latter assumes a high degree of market asymmetry . for indifference valuation to be a _ pricing _ rather than _ valuation _ principle , the demand for a financial security must come from identical agents with known preferences and negligible market impact while the supply must come from a single principal . when the demand comes from heterogeneous individuals with hidden characteristics , indifference arguments do not always yield an appropriate pricing scheme . in this paper we move away from the assumption of investor homogeneity and allow for heterogeneous agents . we consider a single principal with a random endowment whose goal is to lay off some of her risk with heterogeneous agents by designing and selling derivative securities on her income . the agents have mean variance preferences . an agent s degree of risk aversion is private information and hidden to the principal . the principal only knows the distribution of risk aversion coefficients which puts her at an informational disadvantage . if all the agents were homogeneous , the principal , when offering a structured product to a single agent , could ( perhaps ) extract the indifference ( maximum ) price from each trading partner . in the presence of agent heterogeneity this is no longer possible , either because the agents would hide their characteristics from the principal or prefer another asset offered by the principal but designed and priced for another customer . the problem of optimal derivative design in a principal - agent framework with informed agents and an uninformed principal has first been addressed in a recent paper by of carlier , ekeland and touzi @xcite . with the agents being the informed party , theirs is a screening model . the literature on screening within the adverse selection framework can be traced back to mussa and rosen @xcite , where both the principal s allocation rule and the agents types are one - dimensional . armstrong @xcite relaxes the hypothesis of agents being characterized by a single parameter . he shows that , unlike the one - dimensional case , bunching " of the first type is robust when the types of the agents are multi - dimensional . in their seminal paper , rochet and chon @xcite further extend this analysis . they provide a characterization of the contracts , determined by the ( non - linear ) pricing schedule , that maximize the principal s utility under the constraints imposed by the asymmetry of information in the models . building on their work , carlier , ekeland and touzi @xcite study a principal - agent model of optimal derivative design where the agents preferences are of mean - variance type and their multi - dimensional types characterize their risk aversions and initial endowments . they assume that there is a direct cost to the principal when she designs a contract for an agent , and that the principal s aim is to maximize profits . we start from a similar set - up , but substitute the idea that providing products carries a cost for the idea that traded contracts expose the principal to additional risk - as measured by a convex risk measure - in exchange for a known revenue . this may be viewed as a partial extension of the work by barrieu and el karoui ( @xcite,@xcite ) to an incomplete information framework . the principal s aim is to minimize her risk exposure by trading with the agents subject to the standard incentive compatibility and individual rationality conditions on the agents choices . in order to prove that the principal s risk minimization problem has a solution we first follow the seminal idea of rochet and chon @xcite and characterize incentive compatible catalogues in terms of @xmath0-convex functions . when the impact of a single trade on the principal s revenues is linear as in carlier , ekeland and touzi @xcite , the link between incentive compatibility and @xmath0-convexity is key to establish the existence of an optimal solution . in our model the impact is non - linear as a single trade has a non - linear impact on the principal s risk assessment . due to this non - linearity we face a non - standard variational problem where the objective can not be written as the integral of a given lagrangian . instead , our problem can be decomposed into a standard variational part representing the aggregate income of the principal , plus the minimization of the principal s risk evaluation , which depends on the aggregate of the derivatives traded . we state sufficient conditions that guarantee that the principal s optimization problem has a solution and illustrate the effect of risk transfer on her exposure by means of two specific examples . the remainder of this paper is organized as follows . in section [ setup ] we formulate our principal - agent model and state the main result . the proof is given in section [ main ] . in section [ sec - examples ] we illustrate the effects of risk transfer on the principal s position by two examples . in the first we consider a situation where the principal restricts itself to type - dependent multiples of some benchmark claim . this case can be solved in closed form by means of a standard variational problem . the second example considers put options with type - dependent strikes . in both cases we assume that the principal s risk measure is average value at risk . as a consequence the risk minimization problem can be stated in terms of a min - max problem ; we provide an efficient numerical scheme for approximating the optimal solution . the code is given in an appendix . we consider an economy with a single _ principal _ whose income @xmath1 is exposed to non - hedgeable risk factors rising from , e.g. , climate or weather phenomena . the random variable @xmath1 is defined on a standard , non - atomic , probability space @xmath2 and it is square integrable : @xmath3 the principal s goal is to lay off parts of her risk with individual _ the agents have heterogenous mean - variance preferences and are indexed by their coefficients of risk aversion @xmath4 . given a contingent claim @xmath5 an agent of type @xmath6 enjoys the utility @xmath7-\t\ , \textnormal{var } [ y].\ ] ] types are private information . the principal knows the distribution @xmath8 of types but not the realizations of the random variables @xmath6 . we assume that the agents are risk averse and that the risk aversion coefficients are bounded away from zero . more precisely , @xmath9 \qquad \mbox{for some } a > 0.\ ] ] the principal offers a derivative security @xmath10 written on her random income for any type @xmath11 . the set of all such securities is denoted by @xmath12 we refer to a list of securities @xmath13 as a _ contract_. a _ catalogue _ is a contract along with prices @xmath14 for every available derivative @xmath10 . for a given catalogue @xmath15 the optimal net utility of the agent of type @xmath11 is given by @xmath16 no assumption will be made on the sign of @xmath17 our model contemplates both the case where the principal takes additional risk in exchange of financial compensation and the one where she pays the agents to take part of her risk . a catalogue @xmath15 will be called _ incentive compatible _ ( * ic * ) if the agent s interests are best served by revealing her type . this means that her optimal utility is achieved by the security @xmath18 : @xmath19 we assume that each agent has some outside option ( `` no trade '' ) that yields a utility of zero . a catalogue is thus called _ individually rational _ ( * ir * ) if it yields at least the reservation utility for all agents , i.e. , if @xmath20 by offering only incentive compatible contracts , the principal forces the agents to reveal their type . offering contracts where the ir constraint is binding allows the principal to exclude undesirable agents from participating in the market . it can be shown that under certain conditions , the interests of the principal are better served by keeping agents of lower types " to their reservation utility ; rochet and chon @xcite have shown that in higher dimensions this is always the case . if the principal issues the catalogue @xmath21 , she receives a cash amount of @xmath22 and is subject to the additional liability @xmath23 she evaluates the risk associated with her overall position @xmath24 via a coherent and law - invariant risk measure @xmath25 that has the fatou property . it turns out that such risk measures can be represented as robust mixtures of average value at risk . spaces in the appendix and refer to the textbook by fllmer and schied @xcite and the paper of jouini , schachermayer and touzi @xcite for detailed discussion of law invariant risk measures . ] the principal s risk associated with the catalogue @xmath21 is given by @xmath26 her goal is to devise contracts @xmath15 that minimize ( [ eq : r ] ) subject to the incentive compatibility and individual rationality condition : @xmath27 we are now ready to state the main result of this paper . the proof requires some preparation and will be carried out in the following section . [ main - thm ] if @xmath28 is a coherent and law invariant risk measure on @xmath29 and if @xmath28 has the fatou property , then the principal s optimization problem has a solution . for notational convenient we establish our main result for the spacial case @xmath30 . the general case follows from straight forward modifications . let @xmath15 be a catalogue . in order to prove our main result it will be convenient to assume that the principal offers any square integrable contingent claim and to view the agents optimization problem as optimization problems over the set @xmath29 . this can be achieved by identifying the price list @xmath31 with the pricing scheme @xmath32 that assigns the value @xmath14 to an available claim @xmath10 and the value @xmath33 $ ] to any other claim @xmath34 . in terms of this pricing scheme the value function @xmath35 defined in ( [ v - def ] ) satisfies @xmath36 for any individually rational catalogue . for the remainder of this section we shall work with the value function of the ( [ v - def2 ] ) . it is @xmath0-convex in the sense of the following definition ; it actually turns out to be convex and non - increasing as we shall prove in proposition [ pr : uconv ] below . let two spaces @xmath37 and @xmath38 and a function @xmath39 be given . the function @xmath40 is called @xmath0-convex if there exists a function @xmath41 such that @xmath42 for a given function @xmath43 the @xmath0-conjugate @xmath44 of @xmath45 is defined by @xmath46 the _ u - subdifferential _ of @xmath45 at @xmath47 is given by the set @xmath48 if @xmath49 then @xmath50 is called a _ u - subgradient _ of @xmath51 our goal is to identify the class of * ic * and * ir * catalogues with a class of convex and non - increasing functions on the type space . to this end , we first recall the link between incentive compatible contracts and @xmath0-convex functions from rochet and chon @xcite and carlier , ekeland and touzi @xcite . ( @xcite , @xcite ) [ pr : uconv ] if a catalogue @xmath21 is incentive compatible , then the function @xmath35 defined by ( [ v - def ] ) is proper and u - convex and @xmath52 conversely , any proper , u - convex function induces an incentive compatible catalogue . incentive compatibility of a catalogue @xmath21 means that @xmath53 so @xmath54 is u - convex and @xmath52 conversely , for a proper , u - convex function @xmath35 and @xmath55 let @xmath56 by the definition of the u - subdifferential , the catalogue @xmath57 is incentive compatible . the following lemma is key . it shows that the @xmath0-convex function @xmath35 is convex and non - increasing and that any convex and non - increasing function is @xmath0-convex , i.e. , it allows a representation of the form ( [ v - def2 ] ) . this allows us to rephrase the principal s problem as an optimization problem over a compact set of convex functions . [ lm:1 ] suppose that the value function @xmath35 as defined by ( [ v - def2 ] ) is proper . then @xmath35 is convex and non - increasing . any optimal claim @xmath58 is a @xmath0-subgradient of @xmath59 and almost surely @xmath60 = v'(\t).\ ] ] if @xmath61 is proper , convex and non - increasing , then @xmath62 is @xmath0-convex , i.e. , there exists a map @xmath63 such that @xmath64 furthermore , any optimal claim @xmath65 belongs to the @xmath0-subdifferential of @xmath66 and satisfies @xmath67 = \bar{v}'(\t).\ ] ] let @xmath35 be a proper , @xmath0-convex function . its @xmath0-conjugate is : @xmath68-\t \textnormal{var}[y]-v(\t)\right\}\\ & = & { \mathbb{e}}[y]+\sup_{\t\in\t } \left\{\t(-\textnormal{var}[y])-v(\t)\right\}\\ & = & { \mathbb{e}}[y ] + v^*(-\textnormal{var}[y]),\end{aligned}\ ] ] where @xmath69 denotes the convex conjugate of @xmath70 as a @xmath0-convex function , the map @xmath35 is characterized by the fact that @xmath71 . thus @xmath72-v^*(-\textnormal{var}[y])\right\}\\ & = & \sup_{y \in l^2({\mathbb{p}})}\left\{{\mathbb{e}}[y]-\t \textnormal{var}[y]-{\mathbb{e}}[y]-v^*(-\textnormal{var}[y])\right\}\\ & = & \sup_{y\le 0}\left\{\t\cdot y - v^*(y)\right\}\end{aligned}\ ] ] where the last equality uses the fact that the agents consumption set contains claims of any variance . we deduce from the preceding representation that @xmath35 is non - increasing . furthermore @xmath73 so @xmath35 is convex . to characterize @xmath74 we proceed as follows : @xmath75-\t \textnormal{var}[y]-v^u(y)\right\}\\ & = & \left\{y \in l^2 \,\mid\ , v(\t)= { \mathbb{e}}[y]-\t \textnormal{var}[y]-{\mathbb{e}}[y]-v^*(-\textnormal{var}[y])\right\}\\ & = & \left\{y \in l^2 \,\mid\ , v(\t)=\t ( -\textnormal{var}[y])-v^*(-\textnormal{var}[y])\right\}\\ & = & \left\{y \in l^2 \,\mid\ , -\textnormal{var}[y]\in\partial v(\t)\right\}\end{aligned}\ ] ] the convexity of @xmath35 implies it is a.e . differentiable so we may write @xmath76)\right\}.\ ] ] let us now consider a proper , non - negative , convex and non - increasing function @xmath77 . there exists a map @xmath78 such that @xmath79 since @xmath62 is non - increasing there exists a random variable @xmath80 such that @xmath81\in\partial \bar{v}(\t)$ ] and the definition of the subgradient yields @xmath82 ) -f(-\textnormal{var}[y])\right\}.\ ] ] with the pricing scheme on @xmath29 defined by @xmath83-f(-\textnormal{var}[y])\ ] ] this yields @xmath84 the characterization of the subdifferential follows by analogy to part ( i ) . the preceding lemma along with proposition [ pr : uconv ] shows that any convex , non - negative and non - increasing function @xmath35 on @xmath85 induces an incentive compatible catalogue @xmath15 via @xmath86 here we may with no loss of generality assume that @xmath87 = 0 $ ] . in terms of the principal s choice of @xmath35 her income is given by @xmath88 since @xmath89 is decreasing and non - negative the principal will only consider functions that satisfy the normalization constraint @xmath90 we denote the class of all convex , non - increasing and non - negative real - valued functions on @xmath85 that satisfy the preceding condition by @xmath91 : @xmath92 conversely , we can associate with any * ic * and @xmath93 catalogue @xmath15 a non - negative @xmath0-convex function of the form ( [ v - def2 ] ) where the contract satisfies the variance constraint @xmath94 = v'(\t)$ ] . in view of the preceding lemma this function is convex and non - increasing so after normalization we may assume that @xmath35 belongs to the class @xmath91 . we therefore have the following alternative formulation of the principal s problem . the principal s optimization problem allows the following alternative formulation : @xmath95 = 0 , \ , -\textnormal{var}[x(\t ) ] = v'(\t ) \right\}.\ ] ] in terms of our alternative formulation we can now prove a preliminary result . it states that a principal with no initial endowment will not issue any contracts . if the principal has no initial endowment , i.e. , if @xmath96 , then @xmath97 solves her optimization problem . since @xmath28 is a coherent , law invariant risk measure on @xmath29 that has the fatou property it satisfies @xmath98 \quad \mbox{for all } y \in l^2({\mathbb{p}}).\ ] ] for a given function @xmath99 the normalization constraint @xmath87 = 0 $ ] therefore yields @xmath100 - i(v ) = -i(v).\ ] ] since @xmath35 is non - negative and non - increasing @xmath101 . taking the infimum in the preceding inequality shows that @xmath102 and hence @xmath103 is an optimal solution . in the general case we approach the principal s problem in two steps . we start by fixing a function @xmath35 from the class @xmath91 and minimize the associated risk @xmath104 subject to the moment conditions @xmath87 = 0 $ ] and @xmath94=v'(\t)$ ] . to this end , we shall first prove the existence of optimal contracts @xmath105 for a relaxed optimization where the variance constraint is replaced by the weaker condition @xmath106 \leq -v'(\t).\ ] ] in a subsequent step we show that based on @xmath105 the principal can transfer risk exposures among the agents in such a way that ( i ) the aggregate risk remains unaltered ; ( ii ) the variance constraint becomes binding . we assume with no loss of generality that @xmath35 does not have a jump at @xmath107 for a given @xmath108 let us consider the convex set of derivative securities @xmath109=0,\,\textnormal{var}[x(\t)]\le -v'(\t)\,\,\mu - a.e.\right\}.\ ] ] [ lemma - bound ] all functions @xmath99 that are acceptable for the principal are uniformly bounded . under the conditions of ( i ) the set @xmath110 is closed and bounded in @xmath111 . more precisely , @xmath112 if @xmath35 is acceptable for the principal , then any @xmath113 satisfies @xmath114 from ( [ lower - bound ] ) and that fact that @xmath87 = 0 $ ] we deduce that @xmath115 - i(v ) \leq \varrho\left(w-\int_{\t } x(\t)d\t\right ) - i(v ) \leq \varrho(w)\ ] ] so @xmath116 + \varrho(w ) = : k.\ ] ] integrating by parts twice and using that @xmath35 is non - increasing and @xmath117 we see that @xmath118 this proves the assertion because @xmath119 . for @xmath113 we deduce from the normalization constraint @xmath117 that @xmath120 so the assertion follows from part ( i ) . since @xmath28 is a convex risk measure on @xmath121 and because the set @xmath105 of contingent claims is convex , closed and bounded in @xmath121 a general result from the theory of convex optimization yields the following proposition . if the function @xmath35 is acceptable for the principal , then there exists a contract @xmath122 such that @xmath123 the contract @xmath105 along with the pricing scheme associated with @xmath35 does not yield an incentive compatible catalogue unless the variance constraints happen to be binding . however , as we are now going to show , based on @xmath105 the principal can find a redistribution of risk among the agents such that the resulting contract satisfies our * ic * condition . let @xmath124=0,\,\textnormal{var}[x(\t)]=-v'(\t),\,\mu - a.e.\right\}\ ] ] be the set of all contracts from the class @xmath110 where the variance constraint is binding . clearly , @xmath125 let us then introduce the set of types @xmath126<-v'(\t)\right\},\ ] ] for whom the variance constraint is not binding . if @xmath127 then @xmath105 yields an incentive compatible contract . otherwise , we consider a random variable @xmath128 fix some type @xmath129 and define @xmath130}}.\ ] ] we may with no loss of generality assume that @xmath131 is well defined for otherwise the status quo is optimal for the principal and her risk minimization problem is void . the purpose of introducing @xmath131 is to offer a set of structured products @xmath132 based on @xmath133 such that @xmath132 together with the pricing scheme associated with @xmath35 yields an incentive compatible catalogue . to this end , we choose constants @xmath134 for @xmath135 such that @xmath136=-v'(\t).\ ] ] this equation holds for @xmath137 \pm \sqrt{\textnormal{cov}^2[x_v(\t),y]-v'(\t)-\textnormal{var}[x_v(\t)]}.\ ] ] for a type @xmath135 the variance constraint is not binding . hence @xmath138 > 0 $ ] so that @xmath139 and @xmath140 . an application of jensen s inequality together with the fact that @xmath141 is bounded shows that @xmath142 are @xmath8-integrable functions . thus there exists a threshold type @xmath143 such that @xmath144}{\alpha}^+(\t)d\t+\int_{\t_v\cap(\t^*,1]}{\alpha}^-(\t)d\t=0.\ ] ] in terms of @xmath145 let us now define a function @xmath146 and a contract @xmath147 since @xmath148 the aggregate risks associated with @xmath105 and @xmath132 are equal . as a result , the contract @xmath132 solves the risk minimization problem @xmath149 in section [ sec - examples ] we shall consider a situation where the principal restricts itself to a class of contracts for which the random variable @xmath105 can be expressed in terms of the function @xmath35 . in general such a representation will not be possible since @xmath35 only imposes a restriction on the contracts second moments . in order to finish the proof of our main result it remains to show that the minimization problem @xmath150 has a solution and the infimum is obtained . to this end , we consider a minimizing sequence @xmath151 . the functions in @xmath91 are locally lipschitz continuous because they are convex . in fact they are _ uniformly _ locally lipschitz : by lemma [ lemma - bound ] ( i ) the functions @xmath99 are uniformly bounded and non - increasing so all the elements of @xmath152 are uniformly bounded on compact sets of types . as a result , @xmath153 is a sequence of uniformly bounded and uniformly equicontinuous functions when restricted to compact subsets of @xmath85 . thus there exists a function @xmath154 such that , passing to a subsequence if necessary , @xmath155 a standard @xmath156-argument shows that the convergence properties of the sequence @xmath157 carry over to the derivatives so that @xmath158 since @xmath159 it follows from fatou s lemma that @xmath160 so @xmath161 and it remains to analyze the associated risk process . for this , we first observe that for @xmath162 fubini s theorem yields @xmath163 since all the functions in @xmath91 are uniformly bounded , we see that the contracts @xmath164 are contained in an @xmath121 bounded , convex set . hence there exists a square integrable random variable @xmath165 such that , after passing to a subsequence if necessary , @xmath166 let @xmath167 . convergence of the functions @xmath168 implies @xmath169 . thus ( [ convergence - z ] ) yields @xmath170 along with convergence of aggregate risks : @xmath171 by corollary i.2.2 in ekeland and tmam ( 1976 ) @xcite , a lower semi - continuous convex function @xmath172 remains , so with respect to the weak topology @xmath173 the fatou property of the risk measure @xmath28 guarantees that @xmath174 we conclude that @xmath175 solves the principal s problem . our main theorem states that the principal s risk minimization problem has a solution . the solution can be characterized in terms of a convex function that specifies the agents net utility . our existence result is based on a min - max optimization scheme whose complexity renders a rather involved numerical analysis . in this section we consider some examples where the principal s choice of contracts is restricted to class of numerically more amenable securities . the first example studies a situation where the principal offers type - dependent multiplies of some benchmark claim . in this case the principal s problem can be reduced to a constraint variational problem that can be solved in closed form . a second example comprises put options with type dependent strikes . here we provide a numerical algorithm for approximating the optimal solution . in this section we study a model where the principal sells a type - dependent multiple of a benchmark claim @xmath176 to the agents . more precisely , the principal offers contracts of the form @xmath177 in order to simplify the notation we shall assume that the t - bond s variance is normalized : @xmath178=1.\ ] ] let @xmath15 be a catalogue where the contract @xmath179 is of the form ( [ bond ] ) . by analogy to the general case it will be convenient to view the agents optimization problem as an optimization problem of the set of claims @xmath180 so the function @xmath181 solves @xmath182 in view of the variance constraint on the agents claims the principal s problem can be written as @xmath183 note that @xmath184>0,$ ] so the term @xmath184\sqrt{-v'(\t)}$ ] must be included in the income . before proceeding with the general case let us first consider a situation where in addition to being coherent and law invariant , the risk measure @xmath28 is also comonotone . in this case each security the principal sells to some agent increases her risk by the amount @xmath185)\sqrt{-v'(\t)}\right)+\left(v(\t)-\t v'(\t)\right)\ge 0.\ ] ] this suggests that it is optimal for the principal not to sell a bond whose payoff moves into the same direction as her initial risk exposure . [ pr : comonotone ] suppose that @xmath28 is comonotone additive . if @xmath186 and @xmath1 are comonotone , then @xmath187 is a solution to the principal s problem . if @xmath1 and @xmath186 are comonotone , then the risk measure in equation ( [ eq : tc1 ] ) is additive and the principal needs to solve @xmath188\right)\sqrt{-v'(\t)}-\t v'(\t)\right)d\t.\ ] ] since @xmath189\right)\ge 0 $ ] and @xmath190 we see that @xmath191 and hence @xmath192 is a minimizer . in view of the preceding proposition the principal needs to design the payoff function @xmath193 in such a way that @xmath1 and @xmath186 are not comonotone . we construct an optimal payoff function in the following subsection . considering the fact that @xmath194 is a decreasing function the principal s goal must be to make the quantity @xmath195 as small as possible while keeping the income as large as possible . in a first step we therefore solve , for any constant @xmath196 the optimization problem @xmath197 \sqrt{-v'(\t)}-v(\t)+\t v'(\t)\right)d\t = a.\ ] ] the constraint variational problem ( [ variational - problem ] ) captures the problem of risk minimizing subject to an income constraint . it can be solved in closed form . the associated euler - lagrange equation is given by @xmath198 - 1}{2\sqrt{-v'(\t)}}\right),\ ] ] where @xmath199 is the lagrange multiplier . the income constraint and boundary conditions are : @xmath200 - 1).\ ] ] integrating both sides of equation ( [ eq : tc4 ] ) and taking into account the normalization condition @xmath201 , we obtain @xmath202.\ ] ] inserting this equation into the constraint yields @xmath203\sqrt{\left(\frac{\lambda'}{\lambda}\right)^2}\int_a^1 \frac{d\t}{2\t - a}-\left(\frac{\lambda'}{\lambda}\right)^2\int_a^1\left\{\frac{1}{8}\left[\frac{1}{2\t - a}-\frac{1}{2 -a}\right]+ \frac{1}{4}\frac{\t}{(2\t - a)^2}\right\}d\t.\ ] ] in terms of @xmath204 + \frac{1}{4}\frac{\t}{(2\t - a)^2}\right\}d\t \quad \mbox{and } \quad n:=\int_a^1 \frac{d\t}{2\t - a}\ ] ] we have the quadratic equation @xmath205 \sqrt{\left(\frac{\lambda'}{\lambda}\right)^2}-a=0,\ ] ] which has the solution @xmath206-\sqrt{(n { \mathbb{e}}[f])^2 - 4am}}{2m}\ ] ] we have used the root with alternating signs , as we require the problem to reduce to @xmath207 for @xmath208 we notice that the constraint variational problem ( [ variational - problem ] ) is independent of the risk measure employed by the principal . this is because we minimized the risk pointwise subject to a constraint on aggregate revenues . in view of the preceding considerations the principal s problem reduces to a one - dimensional minimization problems over the reals : @xmath209}{2m}+ f(w)\frac{n}{2m}\sqrt{(n { \mathbb{e}}[f])^2 -4a m}\right)-a.\ ] ] once the optimal value @xmath210 has been determined , the principal offers the securities @xmath211 -1}{4\t\lambda-2\lambda a}\right)f(w)\ ] ] at a price @xmath212 -1}{2}\left(\frac{3e\lambda ( 2\t - a)-a}{(4\t\lambda-2\lambda a)^2}+\frac{\lambda e -1}{\lambda^2}\frac{1}{2-a}\right).\ ] ] assume that the principal measures her risk exposure using average value at risk at level @xmath213 . let @xmath214 be a normally distributed random variable with mean @xmath215 and variance @xmath216 one can think that @xmath214 represents temperature . suppose that the principal s initial income is exposed to temperature risk and it is given by @xmath217 with associated risk @xmath218 suppose furthermore that the principal sells units of a put option on @xmath214 with strike @xmath219 , i.e. , @xmath220 by proceeding as above we approximated the principal s risk as @xmath221 and she offers the security @xmath222 to the agent of type @xmath11 for a price @xmath223 in this section we consider the case where the principal underwrites put options on her income with type - dependent strikes . we assume that @xmath224 is a bounded random variable and consider contracts of the form @xmath225 the boundedness assumption on the strikes is made with no loss of generality and each equilibrium pricing scheme is necessarily non - negative . note that in this case the risk measure can be defined on @xmath226 so we only require convergence in probability to use the fatou property . we deduce that both the agents net utilities and the variance of their positions is bounded from above by some constants @xmath227 and @xmath228 , respectively . thus , the principal chooses a function @xmath35 and contract @xmath179 from the set @xmath229 = v'(\t ) , \ , |v'| \leq k_2 , \ , 0 \leq k(\t ) \leq \|w\|_\infty \}.\ ] ] the variance constraint @xmath230 $ ] allows us to express the strikes in terms of a continuous function of @xmath231 , i.e. , @xmath232 the principal s problem can therefore be written as @xmath233 \right\ } d\right ) - i(v ) \right\}\ ] ] where the infimum is taken over the set of all functions @xmath234 that satisfy @xmath235 and @xmath236 . within our current framework the contracts are expressed in terms of the derivative of the principal s choice of @xmath35 . this reflects the fact that the principal restricts itself to type - dependent put options and is not always true in the general case . let @xmath153 be a minimizing sequence for the principal s optimization problem . the functions @xmath168 are uniformly bounded and uniformly equicontinuous so we may with no loss of generality assume that @xmath237 uniformly . recall this also implies a.s . convergence of the derivatives . by dominated convergence and the continuity of @xmath238 along with the fact that @xmath1 is bounded yields @xmath239 and @xmath240 this shows that the principal s positions converge almost surely and hence in probability . since @xmath28 is lower - semi - continuous with respect to convergence in probability we deduce that @xmath241 solves the principal s problem . [ numex ] we close this paper with a numerical approximation scheme for the principal s optimal solution within the pit option framework . we assume the set of states of the world is finite with cardinality @xmath242 each possible state @xmath243 can occurs with probability @xmath244 the realizations of the principal s wealth are denoted by @xmath245 note that @xmath45 and @xmath1 are treated as known data . we implement a numerical algorithm to approximate a solution to the principal s problem when she evaluates risk via the risk measure @xmath246 where @xmath247 we also assume the set of agent types is finite with cardinality @xmath248 i.e. @xmath249 the density of the types is given by @xmath250 in order to avoid singular points in the principal s objective function , we approximate the option s payoff function @xmath251 by the differentiable function @xmath252 where @xmath253 the algorithm uses a penalized quasi - newton method , based on zakovic and pantelides @xcite , to approximate a minimax point of @xmath254 where @xmath255 stands for the values of a convex , non - increasing function , @xmath256 denotes the vector of type dependent strikes and the derivatives @xmath257 are approximated by @xmath258 the need for a penalty method arises from the fact that we face the equality constraints @xmath259 $ ] and @xmath260 in order to implement a descent method , these constraints are relaxed and a penalty term is added . we denote by @xmath261 the total number of constraints . the principal s problem is to find @xmath262 where @xmath263 determines the constraints that keep @xmath264 within the set of feasible contracts and @xmath265 the maple code for our procedure is given in the appendix for completeness . let us illustrate the effects of risk transfer on the principal s position in two model with five agent types and two states of the world . in both cases @xmath266 @xmath267 and @xmath268 the starting values @xmath269 @xmath270 and @xmath271 we set are @xmath272 @xmath273 and @xmath274 respectively . * let @xmath275 and the types be uniformly distributed . the principal s initial evaluation of her risk is @xmath276 . the optimal function @xmath35 and strikes are : + [ cols="^,^",options="header " , ] + the principal s valuation of her risk after the exchanges with the agents is @xmath277 in this paper we analyzed a screening problem where the principal s choice space is infinite dimensional . our motivation was to present a nonlinear pricing scheme for over - the - counter financial products , which she trades with a set of heterogeneous agents with the aim of minimizing the exposure of her income to some non - hedgeable risk . in order to characterize incentive compatible and individually rational catalogues , we have made use of u - convex analysis . to keep the problem tracktable we have assumed the agents have mean - variance utilities , but this is not necessary for the characterization of the problem . considering more general utility functions is an obvious extension to this work . our main result is a proof of existence of a solution to the principal s risk minimization problem in a general setting . the examples we have studied suggest that the methodologies for approaching particular cases are highly dependent on the choice of risk measure , as well as on the kinds of contracts the principal is willing ( or able ) to offer . in most cases obtaining closed form solutions is not possible and implementations must be done using numerical methods . as a work in progress we are considering agents with heterogenous initial endowments ( or risk exposures ) , as well as a model that contemplates an economy with multiple principals . in this appendix we recall some properties and representation results for risk measures on @xmath121 spaces ; we refer to the textbook of fllmer and schied @xcite for a detailed discussion of convex risk measures on @xmath278 and to cheridito and tianbui @xcite for risk measures on rather general state spaces . buerle and mller @xcite establish representation properties of risk law invariant risk measures on @xmath279 spaces for @xmath280 . we assume that all random variables are defined on some standard non - atomic probability space @xmath281 . a _ monetary measure of risk _ on @xmath121 is a function @xmath282 such that for all @xmath283 the following conditions are satisfied : * monotonicity : if @xmath284 then @xmath285 . * cash invariance : if @xmath286 then @xmath287 . a risk measure is called _ coherent _ if it is convex and homogeneous of degree 1 , i.e. , if the following two conditions hold : * convexity : for all @xmath288 $ ] and all positions @xmath289 : @xmath290 * positive homogeneity : for all @xmath291 @xmath292 the risk measure is called coherent and law invariant , if , in addition , @xmath293 for any two random variables @xmath179 and @xmath131 which have the same law . the risk measure @xmath28 on @xmath121 has the _ fatou property _ if for any sequence of random variables @xmath294 that converges in @xmath121 to a random variable @xmath179 we have @xmath295 given @xmath296,$ ] the _ average value at risk _ of level @xmath199 of a position @xmath131 is defined as @xmath297 where @xmath298 is the upper quantile function of @xmath131 . if @xmath299 , then we have the following characterization @xmath300\ ] ] where @xmath301 [ pr:1 ] for a given financial position @xmath302 the mapping @xmath303 is decreasing in @xmath304 it turns out the average value of risk can be viewed as a basis for the space of all law - invariant , coherent risk measures with the fatou property . more precisely , we have the following result . [ th : avar ] the risk measure @xmath305 is law - invariant , coherent and has the fatou property if and only if @xmath28 admits a representation of the following form : @xmath306 where @xmath307 is a set of probability measures on the unit interval . as a consequence of proposition [ pr:1 ] and theorem [ th : avar ] we have the following corollary : [ cor : coh ] if @xmath308 is a law - invariant , coherent risk measure with the fatou property then @xmath309.\ ] ] an important class of risk measures are comonotone risk measures risk . comonotone risk measures are characterized by the fact that the risk associated with two position whose payoff `` moves in the same direction '' is additive . a risk measure @xmath28 is said to be _ comonotone _ if @xmath310 whenever @xmath179 and @xmath131 are comonotone , i.e. , whenever @xmath311 comonotone , law invariant and coherent risk measures with the fatou property admit a representation of the form @xmath312 with(linearalgebra ) + n : = 5 : + m : = 2 : + ng : = 2*m+4*n+1 : + this section constructs the objective function @xmath193 and its gradient . + x : = vector(2*n , symbol = xs ) : + q : = vector(m , symbol = qs ) : + @xmath313*p[j]*q[j ] , j = 1 .. m)+\\ & add(add(t(x[i+n],w[j])*m[i]*p[j]*q[j ] , i = 1 .. n ) , j = 1 .. m)- \\ & add(add(t(x[i+n],w[j])*m[i]*p[j ] , i = 1 .. n ) , j = 1 .. m)+ \\ & add((x[i]-t[i]*(x[i+1]-x[i])/(t[i+1]-t[i]))*m[i ] , i = 1 .. n-1)+\\ & x[n]-t[n]*(x[n]-x[n-1])*m[n]/(t[n]-t[n-1]):\\ \end{array } $ ] @xmath314 + @xmath315 + gradfx : = 0 : + gradfq : = 0 : + for i from 1 to 2*n + do gradfx[i ] : = diff(f , x[i ] ) + end do : + for i from 1 to m + do gradfq[i ] : = diff(f , q[i ] ) + end do : + this section constructs the constraint function @xmath316 and its gradient . + g : = vector(ng , symbol = tt ) : + for j from 1 to n + do g[j ] : = -x[j ] end do : + for j from 1 to n-1 + do g[j+n ] : = x[j+1]-x[j ] + end do : + for i from 1 to n-1 + do g[i+2*n-1 ] : = add(t(x[i+n ] , @xmath317)^2*p[j ] , j = 1 .. m)-add(t(x[i+n ] , w[j])*p[j ] , j = 1 .. m)^2+(x[i+1]-x[i])/(t[i+1]-t[i])-eps2 $ ] + end do : + @xmath318 : = add(t(x[2*n ] , w[j])^2*p[j ] , j = 1 .. m)-add(t(x[2*n ] , w[j])*p[j ] , j = 1 .. m)^2+(x[n]-x[n-1])/(t[n]-t[n-1])-eps2:$ ] + for i from 1 to n-1 + do @xmath319 : = -add(t(x[i+n ] , w[j])^2*p[j ] , j = 1 .. m)+add(t(x[i+n ] , w[j])*p[j ] , j = 1 .. m)^2-(x[i+1]-x[i])/(t[i+1]-t[i])-eps2 $ ] + end do : + @xmath320 : = -add(t(x[2*n ] , w[j])^2*p[j ] , j = 1 .. m)+add(t(x[2*n ] , w[j])*p[j ] , j = 1 .. m)^2-(x[n]-x[n-1])/(t[n]-t[n-1])-eps2:$ ] + g[4*n ] : = add(p[i]*q[i ] , i = 1 .. m)-1+eps3 : + g[4*n+1 ] : = -add(p[i]*q[i ] , i = 1 .. m)-1-eps3 : + for i from 1 to m + do g[i+4*n+1 ] : = -q[i ] + end do : + for i to m + do g[i+m+4*n+1 ] : = q[i]-lambda + end do : gradgx : = 0 : + gradgq : = 0 : + for i from 1 to ng + do for j from 1 to 2*n + do gradgx[i , j ] : = diff(g[i ] , x[j ] ) + end do : + end do : + for i from 1 to ng + do for j from 1 to m + do gradgq[i , j ] : = diff(g[i ] , q[j ] ) + end do : + end do : + this section constructs the slackness structures . + e : = vector(ng , 1 ) : + s : = vector(ng , symbol = si ) : + z : = vector(ng , symbol = zi ) : + s : = diagonalmatrix(s ) : + z : = diagonalmatrix(z ) : + this section initializes the variable and parameter vectors . + x : = vector(2*n , symbol = xs ) : + q : = vector(m , symbol = qs ) : + p : = vector(m , symbol = ps ) : + w : = vector(m , symbol = ws ) : + g : = vector(m , symbol = gs ) : + t : = vector(n , symbol = ts ) : + m : = vector(n , symbol = ms ) : + chi : = convert([x , q , s , z ] , vector ) : + this section constructs the lagrangian and its hessian matrix . + f : = convert([gradfx+(vectorcalculus[dotproduct])(transpose(gradgx ) , z ) , + gradfq-(vectorcalculus[dotproduct])(transpose(gradgq ) , z ) , + ( vectorcalculus[dotproduct])((vectorcalculus[dotproduct])(z , s ) , e)-mu*e , g+s ] , vector ) : + df : = 0 : + for i from 1 to 2*n+m+2*ng + do for j from 1 to 2*n+m+2*ng + do df[i , j ] : = diff(f[i ] , chi[j ] ) + end do : + end do : + this section inputs the initial values of the variables and the values of the parameters . + xinit : = ( 4,3 , 2 , 1 , 0 , 1 , 1 , 1 , 1 , 1 ) : + qinit : = ( 1 , 1 ) : + sinit : = ( .1 , .1 , .1 , .1 , .1 , .1 , .1 , .1 , .1 , .1 , .1 , .1 , .1 , .1 , .1 , .1 , .1 , .1 , .1 , .1 , .1 , .1 , .1 , .1 , .1 ) : + zinit : = ( 1 , 1 , 1 , 1 , 1 , 1 , 1 , 1 , 1 , 1 , 1 , 1 , 1 , 1 , 1 , 1 , 1 , 1 , 1 , 1 , 1 , 1 , 1 , 1 , 1 ) : + tinit : = ( 1/2 , 5/8 , 3/4 , 7/8,1 ) : + pinit : = ( .5 , .5 ) : + ginit : = ( -1 , -2 ) : + minit : = ( 1,1 , 1 , 1 , 1 ) : + winit : = ( 1 , 2 ) : + tau : = .1 ; mu : = .1 ; rho : = .5 : lambda : = 1.1 : eps : = .1 : eps2 : = .1 : eps3 : = .2 : + xo : = xinit ; qo : = qinit ; so : = sinit ; zo : = zinit ; po : = pinit ; go : = ginit ; mo : = minit ; wo : = winit ; + xs : = xinit ; qs : = qinit ; si : = sinit ; zi : = zinit ; ps : = pinit ; gs : = ginit ; ms : = minit ; ws : = winit ; ts : = tinit : + the following section contains the executable code . + i : = 0 ; j : = 0 + normf:=norm(f,2 ) : + while ( normf @xmath321 and @xmath322 ) + do + printf ( `` inner loop : # iteration = % g /n'',j ) ; + printf ( `` - solve linear system ( 11 ) ... '' ) ; + d : = linearsolve(df , -transpose(f ) ) : + printf(``done '' ) ; + printf ( `` - update params ... '' ) ; + alphas:=min(seq(select(type ,- s[k]/d[2*n+m+k],positive ) , k = 1 .. dimension(s ) ) ) ; + alphaz:=min(seq(select(type ,- z[k]/d[2*n+m+ng+k],positive ) , k = 1 .. dimension(z ) ) ) ; + alphamax:=min(alphas , alphaz ) : + alphamax:=min(tau*alphamax,1 ) ; + printf(``done / n '' ) ; + chiold : = chi ; + chinew : = chiold+alphamax*d ; + xs : = chinew[1 .. 2*n ] : ys : = chinew[2*n+1 .. 2*n+m ] : + si : = chinew[2*n+m+1 .. 2*n+m+ng ] : zi : = chinew[2*n+m+ng+1 .. 2*n+m+2*ng ] : + normf:=norm(f,2 ) ; + printf(``done '' ) ; + j:=j+1 ; + end do :

we consider the problem of adverse selection and optimal derivative design within a principal - agent framework . the principal s income is exposed to non - hedgeable risk factors arising , for instance , from weather or climate phenomena . she evaluates her risk using a coherent and law invariant risk measure and tries minimize her exposure by selling derivative securities on her income to individual agents . the agents have mean - variance preferences with heterogeneous risk aversion coefficients . an agent s degree of risk aversion is private information and hidden to the principal who only knows the overall distribution . we show that the principal s risk minimization problem has a solution and illustrate the effects of risk transfer on her income by means of two specific examples . our model extends earlier work of barrieu and el karoui ( 2005 ) and carlier , ekeland and touzi ( 2007 ) . preliminary - comments welcome * ams classification * : 60g35 , 60h20 , 91b16 , 91b70 . * keywords * : optimal derivative design , structured securities , adverse selection , risk transfer .

introduction the microeconomic setup proof of the main theorem examples conclusions coherent risk measures on @xmath121. maple code for the example of section

some time ago fernndez et al @xcite developed a method for the accurate calculation of eigenfunctions and eigenvalues for bound states and resonances of the schrdinger equation . this approach is based on the taylor expansion of a regularized logarithmic derivative of the eigenfunction . the physical eigenvalue is given by a sequence of roots of hankel determinants constructed from the coefficients of that series . one merit of this approach , called riccati pad , is the great convergence rate in most cases and that the same equation applies to bound states and resonances . besides , in some cases it yields upper and lower bounds to the eigenvalues@xcite . the logarithmic derivative satisfies a riccati equation , and one may wonder if the method applies to other nonlinear ordinary differential equations . the purpose of this paper is to investigate whether a kind of pad hankel method may be useful for two point boundary value problems given by nonlinear ordinary differential equations . in section [ sec : method ] we outline the method , in section [ sec : examples ] we apply it to several problems of physical interest , and in section [ sec : conclusions ] we discuss the relative merits of the approach . it is our purpose to propose a method for the treatment of two point boundary value problems . we suppose that the solution @xmath0 of a nonlinear ordinary differential equation can be expanded as @xmath1 about @xmath2 , where @xmath3 and @xmath4 are real numbers , and @xmath5 . we also assume that we can calculate sufficient coefficients @xmath6 in terms of one of them that should be determined by the boundary condition at the other point ; for example , at infinity . we show several illustrative examples in the following section . we try a rational approximation to @xmath7 of the form @xmath8(z)=\frac{\sum_{j=0}^{m}a_{j}z^{j}}{\sum_{j=0}^{n}b_{j}z^{j}}. \label{eq:[m , n]}\ ] ] where @xmath9 . the taylor expansion of the usual pad approximant yields @xmath10 coefficients of the series ( [ eq : f_series])@xcite ; but in the present case we require that the rational approximation ( [ eq:[m , n ] ] ) gives us one more coefficient , that is to say , @xmath11 . if @xmath12 , @xmath13 , @xmath14 , this requirement leads to the equation@xcite @xmath15 where @xmath16 is the dimension of the hankel determinant @xmath17 . in general , equation ( [ eq : hankel ] ) exhibits many roots and one expects to find a sequence , for @xmath18 and fixed @xmath19 , that converges towards the required value of the unknown coefficient . from now on we call it hankel sequence for short . if such convergent sequence is monotonously increasing or decreasing we assume that it yields a lower or upper bound , respectively . such bounds where proved rigorously for some eigenvalue problems@xcite . in order to test the performance of the pad hankel method , in this section we consider the examples treated recently by boisseau et al by means of a most interesting algebraic approach@xcite . we first consider the field equation determining the vortex profile in a ginzburg landau effective theory@xcite ( and references therein ) @xmath20 the solution @xmath21 satisfies the expansion ( [ eq : f_series ] ) with @xmath22 , @xmath23 , and @xmath24 . if we substitute this series into the differential equation and solve for the coefficients @xmath6 , we obtain them in terms of the only unknown @xmath25 that is determined by the boundary condition at infinity : @xmath26@xcite ( and references therein ) . the coefficients @xmath6 , and therefore the hankel determinant @xmath27 , are polynomial functions of @xmath25 . for example , for @xmath28 we have @xmath29 tables [ tab : gln1 ] and [ tab : gln2 ] show two hankel sequences with @xmath30 and @xmath31 that converge rapidly towards the result of the accurate shooting method@xcite for @xmath28 and @xmath32 , respectively . we appreciate that in the case @xmath28 the sequences with @xmath30 and @xmath31 give upper and lower bounds , respectively , that tightly bracket the exact value of the unknown parameter of the theory : @xmath33 . on the other hand , the appropriate hankel sequences are oscillatory when @xmath34 and their rate of convergence decreases with @xmath35 . table [ tab : gln2 - 4 ] shows the best estimates of @xmath25 for @xmath36 . our second example is the fixed point equation for wilson s exact renormalization group@xcite ( and references therein ) @xmath37 the solution to this equation can be expanded as in equation ( [ eq : f_series ] ) with @xmath38 and @xmath24 . the first coefficients are @xmath39 for large values of @xmath40 the physical solution should behave as @xmath41 . the hankel sequences with @xmath30 and @xmath31 converge towards the numerical result@xcite ( and references therein ) from above and below , respectively . [ fig : wilson ] displays the great rate of convergence of these sequences as @xmath42 , @xmath18 , from which we obtain the accurate bounds @xmath43 the third example comes from a suitably modified wegner houghton s fixed point equation in the local potential approximation@xcite ( and references therein ) @xmath44[5f(x)-xf^{\prime } ( x)]=0,\;x>0 . \label{eq : w - h}\ ] ] the solution satisfies the series ( [ eq : f_series ] ) with @xmath38 and @xmath45 , and the first coefficients are @xmath46 on the other hand , the acceptable solution should behave as @xmath47 when @xmath48 . table [ tab : wh ] shows hankel sequences with @xmath30 and @xmath31 that clearly converge towards the numerical value of @xmath25@xcite ( and references therein ) . we have also applied our approach to the ordinary differential equation for the spherically symmetric skyrmion field@xcite ( and references therein ) but we could not obtain convergent hankel sequences . we do not know yet the reason for the failure of the method in this case . present approach has earlier proved suitable for the treatment of the riccati equation derived from the schrdinger equation@xcite . consider , for example , the following riccati equation @xmath49 the solution can be expanded as in equation ( [ eq : f_series ] ) with @xmath50 ; the first coefficients are @xmath51 there is a critical value @xmath52 of @xmath53 such that @xmath54 at large @xmath40 if @xmath55 , @xmath0 develops a singular point if @xmath56 , and @xmath57 at large @xmath40 if @xmath58 . present pad hankel method yields the value of @xmath52 with remarkable accuracy as shown in table [ tab : riccati ] . the rate of convergence of the hankel sequence for this problem is considerably greater than for the preceding ones . if we substitute @xmath59 into equation ( [ eq : riccati ] ) , then the function @xmath60 satisfies the schrdinger equation for a harmonic oscillator with zero energy on the half line : @xmath61 , and the problem solved above is equivalent to finding the logarithmic derivative at origin @xmath62 so that @xmath60 behaves as @xmath63 at infinity . obviously , any approach for linear differential equations is suitable for this problem . finally , we consider the thomas fermi equation @xcite ( and references therein ) @xmath64 that provides a semiclassical description of the charge density in atoms of high atomic number . it poses the problem of finding the slope at origin @xmath65 so that @xmath66 and @xmath67 . the change of variables @xmath68 leads to a more tractable equation @xmath69 where @xmath70 . we can expand the solution to this equation in a taylor series : @xmath71 and the unknown slope is given by the unknown coefficient : @xmath72 . one easily derives a recurrence relation for the coefficients of this expansion@xcite . although the treatment of equation ( [ eq : t - f - xi ] ) is straightforward , we find it more convenient to define the function @xmath73 that is a solution to @xmath74 -f(t)f^{\prime } ( t)-2t^{2}f(t)^{3}=0 . \label{eq : t - f - f}\ ] ] the function @xmath75 satisfies a series like ( [ eq : f_series ] ) with @xmath76 and @xmath77 . the first coefficients are @xmath78 we expect that a hankel sequence will converge towards @xmath79 . since @xmath80 is the first nonzero coefficient that depends on @xmath81 we choose @xmath82 in the hankel determinant . table [ tab : tf ] shows that the hankel sequence with @xmath83 converges rapidly giving a most accurate value of the slope at origin @xmath84 . finally , we consider two examples discussed by bender et al@xcite ; the first of them is the instanton equation @xmath85 with the boundary conditions @xmath86 , @xmath87 . the solution to this equation is @xmath88 . the expansion of @xmath0 is a particular case of equation ( [ eq : f_series ] ) with @xmath38 and @xmath45 ; its first coefficients being @xmath89 where @xmath90 is the unknown . the hankel series with @xmath30 and @xmath31 converge rapidly giving upper and lower bounds , respectively , to the exact result @xmath91 . the second example is the well known blasius equation@xcite @xmath92 with the boundary conditions @xmath93 , @xmath94 . the expansion of the solution in a taylor series about @xmath2 is a particular case of equation ( [ eq : f_series ] ) with @xmath95 and @xmath96 ; its first coefficients are @xmath97 since , in general , @xmath98 , then the only root of the hankel determinants is @xmath99 that leads to the trivial solution @xmath100 . we thus see another case where the pad hankel method does not apply . we have presented a simple method for the treatment of two point boundary value problems . if there is a suitable series for the solution about one point , we construct a hankel matrix with the expansion coefficients and obtain the physical value of the undetermined coefficient from the roots of a sequence of determinants . the value of this coefficient given by a convergent hankel sequence is exactly the one that produces the correct asymptotic behaviour at the other point . we can not prove this assumption rigorously , but it seems that if there is a convergent sequence , it yields the correct answer . moreover , in some cases the hankel sequences produce upper and lower bounds bracketing the exact result tightly . present pad hankel approach is not as general as the one proposed by boisseau et al@xcite , as we have already seen that the former does not apparently apply to the skyrmion problem or to the blasius equation@xcite . however , our procedure is much simpler and more straightforward and may be a suitable alternative for the treatment of this kind of problems . besides , if our approach converges , it yields remarkably accurate results as shown in the examples above . for example , it gives us the slope of the electrostatic potential of the thomas fermi theory with unprecedented accuracy .

we propose a method for the treatment of two point boundary value problems given by nonlinear ordinary differential equations . the approach leads to sequences of roots of hankel determinants that converge rapidly towards the unknown parameter of the problem . we treat several problems of physical interest : the field equation determining the vortex profile in a ginzburg landau effective theory , the fixed point equation for wilson s exact renormalization group , a suitably modified wegner houghton s fixed point equation in the local potential approximation , a riccati equation , and the thomas fermi equation . we consider two models where the approach does not apply in order to show the limitations of our pad hankel approach .

introduction[sec:intro] method[sec:method] examples[sec:examples] conclusions[sec:conclusions]

during the recent years , numerous studies have been reported in the field of ferroelectric oxide nanostructures . these were motivated by the perspective of new technological applications and made possible thanks to combined spectacular developments of experimental and theoretical techniques.@xcite on the experimental side , it is now possible to grow oxide nanostructures with a control at the atomic scale and to measure their properties using local probes . on the theoretical side , different significant advances have been realized in the field of first - principles calculations that , combined to the recurrent increase of computational power , allow not only to reproduce accurately experimental measurements but also to make trustable predictions that are verified a posteriori . the study of ferroelectric nanostructures is nowadays a broad field of researches and it would be impossible to provide an exhaustive description of earlier works in a short review paper . our purpose here is to provide a simple and comprehensive description of the origin of ferroelectric finite size effects and to highlight some recent results , focusing essentially on epitaxial thin films and superlattices . for a more complete and detailed description of recent first - principles achievements concerning ferroelectrics and their nanostructures , we refer the reader to recent book chapters @xcite and topical review papers.@xcite in the present review , we focus our attention on the influence of mechanical and electrical boundary conditions . although other effects might also be important ( chemistry of interfaces , defects ... ) , both effects play a key role and already allow to understand the main ferroelectric finite size effects when properly taken into account . the roles of mechanical and electrical boundary conditions are introduced separately and their respective influence is first illustrated in the paradigmatic case of a batio@xmath0 ( the standard ferroelectric perovskite oxide material , see sec . [ sec : background ] ) thin film under realistic conditions , including its epitaxial growth on srtio@xmath0 ( sec . [ sec : strain ] ) , and the screening of its polarization charge by realistic electrodes ( sec . [ sec : depfield ] ) . once the basic physical effects have been introduced , the discussion is enlarged to other perovskite oxides epitaxial films ( sec . [ sec : capacitors ] ) and superlattices ( sec . [ sec : superlattices ] ) . finally , recent works on other types of ferroelectric oxide nanostructures , such as nanoparticles or nanowires , are briefly reviewed ( sec . [ sec : nanoparticles ] ) . barium titanate ( batio@xmath0 ) is a prototypical ferroelectric oxide.@xcite at high temperature , it is paraelectric and crystallizes in the high - symmetry cubic perovskite structure as illustrated in fig . [ fig.1]a : ba atom is located at the corner of the cubic unit - cell while the ti atom is at the center and is surrounded by an octahedra of o atoms , themselves located at the center of each face of the cube . on cooling , batio@xmath0 undergoes a sequence of three ferroelectric phase transitions to structures successively of tetragonal ( @xmath1c ) , orthorhombic ( @xmath2c ) and rhombohedral ( @xmath3c ) symmetry . in the following discussion , we will focus on the tetragonal phase that is stable at room temperature . the phase transition from cubic to tetragonal is characterized by the opposite shift of ti and o atoms with respect to ba , taken as reference . this polar atomic distortion is accompanied with a small relaxation of the unit cell and yields a stable spontaneous polarization of 26 @xmath4c/@xmath5 at room temperature . in the tetragonal phase , the cubic symmetry is broken , resulting not in a unique polar structure but in six symmetrically equivalent variants with polarization along the [ 100 ] , [ 010 ] and [ 001 ] directions . the two variants along [ 001 ] ( `` up '' and `` down '' states ) are represented in fig . [ fig.1]a . what makes batio@xmath0 ferroelectric is not only the existence of energetically equivalent polar states but also the fact that it is possible to switch from one to another of these states by applying an electric field larger than the coercive field . related to this switchability , the relationship between polarization and electric field is hysteretic in ferroelectrics , as illustrated in fig . [ fig.1]c . in what follows , the pattern of cooperative atomic displacements associated to the phase transition will be referred to as @xmath6 , and the paraelectric , up and down states respectively will be labeled as @xmath6 equal to 0 , 1 and -1 respectively . for continuous evolution of @xmath6 , the internal energy @xmath7 of the crystal has a typical double - well shape illustrated in fig . [ fig.1]b . in batio@xmath0 , the well depth is typically of the order of 30 mev / cell . while hysteresis loops ( fig . [ fig.1]c ) constitutes the measurement of choice to demonstrate experimentally ferroelectricity , the existence of a double - well shape for the energy is usually considered as the theoretical fingerprint of the ferroelectric instability . making a taylor expansion of @xmath7 in terms of @xmath6 around the paraelectric state taken as reference , and since the odd terms are forbidden by symmetry , we can write @xmath8 where for the sake of simplicity we have assumed that atomic displacements in the tetragonal phase can take place only along @xmath9 ( condition that will be held all along this review unless otherwise stated ) , and we have cut the expansion in the eigth power of the atomic pattern of displacement . @xmath7 presents a typical double - well shape with a negative curvature at the origin so that , in the previous equation , ferroelectricity can be associated to a negative value of @xmath10 ( @xmath11 ) . we note that , in the limit of a homogeneous polarization , the previous expansion in terms of microscopic atomic variable @xmath6 is similar in spirit to that of devonshire - ginzburg - landau ( dgl ) phenomenological theory @xcite in terms of the macroscopic polarization , since @xmath12 where @xmath13 is the unit cell volume and @xmath14 is the diagonal component of @xmath15 , the effective charge tensor associated to @xmath6 , so that part of the forecoming discussion can be transposed in that alternative context . at the bulk level , a coherent microscopic model explaining the origin of ferroelectricity was first reported by cochran,@xcite who identified an intimate link between the structural phase transition and the lattice dynamics and introduced the concept of soft mode . specifically , the underlying idea is that the lowest - frequency zone - center polar phonon mode in the paraelectric phase becomes softer as a function of decreasing temperature , and finally goes down to zero frequency , freezing in below @xmath16 to generate the ferroelectric crystal structure . cochran s theory of displacive phase transition was exhibited in the context of a shell - model and the softening of a given polar mode was explained from a competition between short - range and long - range coulomb interactions . most of these ideas were further confirmed from first - principles calculations @xcite performed within the density functional theory ( dft ) . an unstable polar mode ( i.e. a mode with imaginary frequency @xmath17 within the harmonic approximation ) has been identified at @xmath18 in the paraelectric phase of various ferroelectric oxides @xcite and , in prototypical cases like batio@xmath0 , a strong resemblance has been observed between the eigendisplacement vector of this unstable phonon and the ground - state distortion @xmath6 ( overlap of 99 % in batio@xmath0 or linbo@xmath0@xcite ) . since energy curvature is related to the square of the frequency , an unstable mode with imaginary frequency in the paraelectric phase is the indication of a negative curvature of the energy at the origin [ @xmath19 in eq . ( [ eq : doublewell ] ) ] . the amplitude of @xmath17 does not directly provide the double - well depth that additionally depends on anharmonic effects but measures , to some extent , the strength of the instability . the origin of the ferroelectric instability was also investigated at the first - principles level and the competition between short - range ( sr ) interactions and long - range ( lr ) coulomb forces has been confirmed and quantified.@xcite in this context , the anomalous values of the born effective charges of ferroelectric oxides @xcite was shown as an essential feature to produce an unusually large destabilizing coulomb interaction responsible for the ferroelectric instability . the balance of forces resulting in the ferroelectric instability is also well - known to be strongly sensitive to strain and external pressure . the sensitivity of the ferroelectric distortion to strain is inherent to the piezoelectric behavior and was highlighted from the early stage of first - principles computations.@xcite samara _ @xcite proposed that external hydrostatic pressure reduces and eventually totally suppresses ferroelectricity since sr repulsions increases more rapidly than lr destabilizing forces under pressure . this latter result was recently reinvestigated at the first - principles level,@xcite confirming a rapid disappearance of ferroelectricity at moderate pressure and additionally pointing out an unexpected recovery of ferroelectricity at extremely high pressure . such microscopic understanding of ferroelectricity at the bulk level allows to anticipate the existence of strong finite size effects in ferroelectrics since both sr and lr forces will be modified in nanostructures . the sr forces will be modified at surface and interfaces due to the change of atomic environment . the same is true for the born effective charges that can not keep their bulk value @xcite as it was illustrated on batio@xmath0 thin films by fu and coworkers.@xcite beyond that , the lr coulomb interaction will be cut because of the finite size of the sample and will be strongly dependent on the electrical boundary conditions . finally , since the ferroelectric instability is strongly sensitive to strains , it will be influenced by mechanical boundary conditions such as epitaxial strains.@xcite these different factors can act independently to either enhance or suppress ferroelectricity . they will compete with each other in such a way that it is difficult to predict the ferroelectric properties of nanostructures without making explicit investigations . during the recent years , most of the efforts devoted to understanding ferroelectric finite size effects concerned ferroelectric thin films . this was partly motivated by the observation at the end of the nineties of a polar ground - state in ultrathin films @xcite with thicknesses well below 10 - 20 nm , that was still considered at that time as a realistic estimate of critical thickness for ferroelectricity.@xcite it is now understood that the survival or disappearance of ferroelectricity in ultrathin films is not a truly intrinsic property : it strongly depends on the structure and chemistry of the interface between the ferroelectric and the substrate and/or the electrodes as well as on electrical ( e.g. screening of the depolarizing field ) and mechanical ( e.g. epitaxial strains ) boundary conditions . on top of that , additional features such as the finite conductivity of the film or the presence of structural defects and vacancies might also affect the ferroelectric properties . although these latter features are potentially important , the proper treatment of their effect remains beyond the scope of first - principles simulations and is not yet fully understood so that these effects will not be further discussed here . for an overview of atomic relaxations and discussions of the stability of surfaces and interfaces we refer the reader to previous reviews.@xcite in what follows , we will restrict the discussion on the role of mechanical and electrical the boundary conditions which , in most cases , provide a sufficient background to achieve relevant understanding of observed ferroelectric finite size effects . as mentioned in sec . [ sec : background ] , ferroelectricity in bulk compounds is well - known to be very sensitive to strains.@xcite in thin films , it is therefore expected that epitaxial strains will play an important role to monitor the ferroelectric properties . coherently with that , strain engineering of the ferroelectric properties has been recently demonstrated experimentally : a polarization 250% higher than in bulk single crystals has been reported in batio@xmath0 epitaxial films,@xcite and room temperature ferroelectricity has even been induced by strain in epitaxial srtio@xmath0 thin films.@xcite in order to understand these effects , let us generalize our expansion of the internal energy in terms of additional strain @xmath20 ( where @xmath21 and @xmath22 are cartesian directions ) degrees of freedom . in our paradigmatic example of a batio@xmath0 film epitaxially grown on a ( 001 ) cubic srtio@xmath0 substrate we have mixed strain / stress boundary conditions : on the one hand the in - plane strains @xmath23 are fixed by the lattice mismatch between batio@xmath0 and srtio@xmath0 , while @xmath24 ; on the other hand , the out - of - plane strain @xmath25 and the shear strains @xmath26 and @xmath27 are free to relax ( condition of zero stress : @xmath28 ) . again , assuming for simplicity only an homogeneous polarization along @xmath9-direction , vanishing shear strains , and restricting the expansion to leading orders in @xmath6 and @xmath29 , the free energy functional to be minimized now reads @xmath30 the terms in the first line correspond to the double - well energy of eq . ( [ eq : doublewell ] ) . the terms in the second - line are the elastic energy while the terms in the third line arise from the coupling between ionic and strain degrees of freedom . they correspond to the so - called `` polarization - strain coupling '' and are at the origin of the piezoelectric response . it is clear from eq . ( [ eq : uwithstrain ] ) that the polarization - strain coupling terms are responsible for a renormalization of the quadratic part of @xmath7 that now takes the form @xmath31 \xi_z^2 . \label{eq : cuadrafterstr}\end{aligned}\ ] ] depending of the value of the parameters @xmath32 and @xmath33 , and of @xmath34 and @xmath25 ( deduced from the relation @xmath35 which follows from the boundary condition @xmath36 ) , we see that , playing properly with the epitaxial strain conditions , it is possible to modulate the coefficient of @xmath37 in order to increase ( resp . decrease ) its negative value and therefore enhance ( resp . suppress ) the ferroelectric character of the film . the simplified free energy @xmath7 in eq . ( [ eq : uwithstrain ] ) can be generalized to include other orientations of @xmath6 as well as shear strains and the different parameters can be directly determined from dft calculations at the bulk level.@xcite the coefficients of the expansion for a series of ferroelectric perovskites have been reported by diguez _ _ @xcite these energy parametrizations at 0 k are the starting point for the development of model hamiltonians,@xcite that combined with classical monte carlo simulations allow the first - principles based studies of temperature versus strain phase diagrams in perovskite oxide films , as illustrated for batio@xmath0 on a cubic ( 001 ) substrate in fig . [ fig.2 ] . although effects such as free top surface relaxation or interaction with the substrate are neglected , the phase sequence and the topology of the phase diagram are usually well reproduced within this approach . the predicted phase transitions temperatures are however typically too low compared to those experimentally measured . alternatively , and in fact first applied to perovkskites by pertsev and coworkers @xcite before first - principles investigations , similar phase diagram can be obtained from phenomenological devonshire - ginzburg - landau ( dgl ) type expansion of the free energy in terms of polarization and strain . results have for instance been reported for batio@xmath0,@xcite pbtio@xmath0,@xcite srtio@xmath0,@xcite and pb(zr@xmath38ti@xmath39)o@xmath40 ( pzt ) solid solution.@xcite this provides more direct access to the temperature behavior and gives very accurate results around the temperature / strain regime in which the model parameters were fitted ( usually near the bulk ferroelectric transition ) . in distant strain - temperatures regimes their predictions might be less accurate , however . for large strains or low temperatures the uncertainty increases and , as reported for batio@xmath0 , different sets of dgl parameters can provide phase diagrams qualitatively different far from the bulk transition region.@xcite as a general rule for the usual perovskites on a ( 001 ) substrate,@xcite sufficiently large epitaxial compressive ( resp . tensile ) strains will favor a ferroelectric @xmath41-phase ( resp . @xmath42-phase ) with out - of - plane ( resp . in - plane ) polarization while distinct behaviors are predicted for the different compounds in the intermediate regime . of course , there might be some discrepancies with this rule , such as the one pointed by catalan and coworkers @xcite on pbtio@xmath0 films under tensile strains . for the batio@xmath0 film on a ( 001 ) srtio@xmath0 substrate previously considered , the large compressive epitaxial strain ( -2.38 % ) will strongly increase the ferroelectric phase transition temperature and favor a tetragonal @xmath41-phase with an out - of - plane polarization . as further estimated by neaton and rabe,@xcite such epitaxial contraint should enhance the polarization at 0 k by more than 50 % . the sensitivity of polarization to strain is however not expected to be always so large : in highly polar pb - based systems such as pb(zr@xmath38ti@xmath43o@xmath40 and pbtio@xmath0 where the ferroelectric displacements are already large , it was reported to be much more reduced than in batio@xmath0,@xcite similarly to what was also discussed for bifeo@xmath0 and linbo@xmath0.@xcite in ref . , like in many studies , the phase diagrams were obtained restricting the investigation to the effect of the epitaxial strain on the ferroelectric degree of freedom @xmath6 , or equivalently @xmath44 . in systems where different kinds of instabilites coexist ( such as ferroelectric and antiferrodistortive instabilities in srtio@xmath40,@xcite or the interplay between in - plane ferroelectricity , antiferroelectric and antiferrodistortive distortions in pbtio@xmath40 free standing slabs @xcite ) the epitaxial strain can modify the competition between these instabilities so that all of them must be explicitely considered and generate much more complicated phase diagrams.@xcite also , up to now , dft calculations were restricted to mono - domain configurations while much complex domain structures can appear as highlighted using a dgl approach @xcite or an effective hamiltonian approach @xcite as it will be further illustrated in section [ sec : capacitors ] . the effects imposed by the substrate on the lattice constant and geometry of the thin films can be maintained only up to a critical thickness , @xmath45 , above which the elastic deformation can not be maintained any longer . above @xmath45 the strain energy is relaxed by the formation of misfit dislocations at the film - substrate interface . an estimate of @xmath45 can be obtained using the matthews - blakeslee formula.@xcite a priori , this structural limit imposes some restriction on the potential enhancement of the ferroelectric properties by the mechanical boundary conditions . however , it has been found that a coherent strain state can be preserved well above the theoretical critical thicknesses expected for strain relaxation.@xcite moreover , in some cases , it was reported that the ferroelectric transition temperature still has a value much larger than in bulk , even after the relaxation of the strain state.@xcite in all previous approaches , only the presence of an homogeneous strain was considered . however , `` flexoelectric '' coupling of a strain gradient to the polarization,@xcite and the related coupling of strain to a polarization gradient is receiving increasing attention . although the characteristic scale of the flexoelectric coefficient is extremely small , it might affect the ferroelectric transitions in thin films . the subtleties of defining flexoelectric coefficients as a bulk property @xcite and its computation in a first - principles framework is a challenge for theoreticians . a more exhaustive discussion of strain effects can be found in a recent topical review by k. m. rabe.@xcite as it was explained in sec . [ sec : background ] , ferroelectricity in perovskite oxides at the bulk level is a collective effect resulting from a delicate balance between sr interactions , that favour a paraelectric phase , and lr dipole - dipole electrostatic interactions , that favour a ferroelectric distortion . so , it is expected that cutting these lr interactions by reducing the number of dipoles to interact with in a finite system , will affect the ferroelectric behavior of the material . if we try to induce a homogeneous out - of - plane polarization in a ( 001 ) free standing batio@xmath0 slab in vacuum , preservation of the normal component of the electric displacement field at the interface will induce a depolarizing field @xmath46 , where @xmath47 is the permittivity of vacuum ( unless otherwise stated we will use throughout this review the si system of units ) . this depolarizing field totally suppresses the possibility of stabilizing an homogeneous monodomain configuration with an out - of - plane polarization . many applications rely on films with a uniform switchable out - of - plane polarization . so , after the previous discussion , it is clear that , to stabilize such a configuration , we need to screen the depolarizing field @xmath48 . this can be achieved through two basic mechanisms : the first one is the compensation by `` free '' charges provided either by a metallic electrodes , semiconducting substrates , or even by ionic adsorbates@xcite ( including in this category the screening by external applied fields that might be generated by the free charges or dipole layers);@xcite the second one , and the only one if the substrate is an insulator,@xcite is the breaking up of the system into 180@xmath49 domains . most of the fully first - principles calculations have been devoted to the screening by free charges , especially due to the very demanding computational effort that an explicit treatment of the domains implies , so that only the first issue will be discussed in more details here . in an early first - principles paper , ghosez and rabe,@xcite using a model hamiltonian approach , showed that ferroelectricity might be preserved in ferroelectric slabs if @xmath48 is perfectly compensated . this was further confirmed by meyer and vanderbilt , at least for some terminations of the free standing slabs , in a fully first - principles atomistic calculation.@xcite in principle , this total compensation of the depolarizing field could be achieved introducing the ferroelectric film between two metallic electrodes in short - circuit . however , following an idea introduced in the early seventies by batra,@xcite metha @xcite and coworkers in the context of a phenomenological model ( see fig . [ fig : depolfield ] ) , junquera and ghosez @xcite demonstrated from first - principles that , even in that case , compensation of the depolarizing field is incomplete , due to the finite screening - length at the metal / ferroelectric interface . within this model , if a batio@xmath0 film gets a polarization normal to the ferroelectric / electrode interface , then the discontinuity of the polarization at the interface gives rise to a net interface charge density @xmath50 @xcite ( fig . [ fig : depolfield]a ) . in order to compensate for this polarization charge density , a screening charge is induced in the electrode that spreads in practice over a finite distance from the interface ( fig . [ fig : depolfield]c ) , giving rise to an interface dipole moment density and a jump in the electrostatic potential at each ferroelectric / electrode interface ( fig . [ fig : depolfield]d ) . this potential drop @xmath51 is shown to be proportional to the polarization @xcite @xmath52 where @xmath53 is a constant of proportionality hereafter referred to as the effective screening length . assuming a capacitor with two similar metal / ferroelectric interfaces , the total potential drop into the electrode / ferroelectric / electrode heterostructure amounts to @xmath54 . if the electrodes are in short - circuit ( i.e. at the same potential ) , a residual field within the ferroelectric must compensate for this drop of potential ( fig . [ fig : depolfield]e ) , @xmath55 where @xmath56 is the film thickness in terms of the number @xmath57 of unit cells of the ferroelectric perovskite oxide . inserting now eq . ( [ eq : deltav ] ) into eq . ( [ eq : ed1 ] ) @xmath58 this shows that the field produced by the screening charge within the metal does not fully compensate for the bare depolarizing field @xmath46 : a residual depolarizing field remains that is proportional to @xmath44 and to the ratio between the effective screening length and the film thickness . it will tend to zero for thick films ( @xmath59 ) but can become sizable when the film thickness becomes comparable to @xmath53 . the ratio between the residual field into the ferroelectric ( eq . [ eq : depolfield ] ) and the bare depolarizing field quantifies the quality of the screening and can be expressed as @xmath60 ( as it will be further discussed in sec . iii - c ) . the depolarizing field can be computed averaging @xcite the microscopic electrostatic potential provided by a first - principles calculation for the whole capacitor . details on the methodology and practical recipes to perform this nano - smoothing can be found in ref . . since the atomic positions , and therefore the induced polarization [ eq . ( [ eq : pol ] ) ] , and the thickness of the film are input variables of the simulation , @xmath53 can be estimated using eq . ( [ eq : depolfield ] ) once @xmath48 is known . junquera and ghosez reported a value of @xmath61 ( one order of magnitude smaller than the interplanar distances in the heterostructure ) for their frozen phonon calculations on a srruo@xmath0/batio@xmath0/srruo@xmath0 capacitor . it might be even slightly smaller when the atomic relaxation of the metal is properly treated.@xcite it is worth noticing that the effective screening length introduced in this model is not an intrinsic property of the metal ( like the thomas - fermi screening length ) , but a quantity that strongly depends also on the details of the ferroelectric / electrode interface such as the different chemical bonds formed at the junction , or the penetration of electronic wave functions of the metal into the ferroelectric and vice versa . as shown by sai _ et al._,@xcite @xmath53 can vary and become very small for some specific metal / ferroelectric interfaces . understanding and controlling this compensation mechanism of the polarization charges in order to stabilize a uniform switchable polarization in the film is one of the challenges in the field of thin film ferroelectrics.@xcite other authors @xcite alternatively explain the origin of a residual depolarizing field inside the ferroelectric thin film from the presence of a `` dead layer '' at the metal / ferroelectric interface . although evoking a different mechanism , the theory of the dead layer yields very similar mathematical expressions than the theory of the imperfect metal screening discussed above , the role of the effective screening length being played by the thickness of the dead layer ( see for instance ref . , or compare the formulas of ref . by bratkovsky and levanyuk with those of ref . by chensky and tarasenko ) . stengel and spaldin @xcite recently reconciled these two apparently antagonist theories , providing a unified interpretation . using a first - principles approach , they showed that the origin of an intrinsic dead layer with poor dielectric properties at the interface in oxide nanocapacitors is related to the hardening of the collective zone center polar modes that itself is produced by the incomplete screening of the depolarizing field . in the presence of a non - vanishing electric field , the proper energy to be considered and minimized is no more the internal energy @xmath62 of the crystal but a field - dependent energy potential @xmath63 that , on top of @xmath62 , also includes an electrostatic energy term arising from the coupling between the polarization and the field.@xcite this coupling term is proportional to @xmath64 . since the depolarizing field is proportional to @xmath65 [ eq . ( [ eq : depolfield ] ) ] the additional electrostatic term in @xmath66 is proportional to @xmath67 and positive in sign . remembering eq . ( [ eq : pol ] ) , we see that proper treatment of the electrostatic energy will produce an additional term in eq . ( [ eq : cuadrafterstr ] ) of the form @xmath68 . so , it clearly appears that , in a way similar to the epitaxial strain in sec . [ sec : strain ] , the incomplete screening of the depolarizing field is responsible for a renormalization of the quadratic coefficient of the energy . in this case however , the correction term is always positive ( @xmath69 ) and therefore its effect contributes to suppress ferroelectricity . the depolarizing field is inversely proportional to the thickness of the ferroelectric layer , eq . ( [ eq : depolfield ] ) . the thinnest the film , the larger the depolarizing field and the larger the energy penalty to stabilize a monodomain pattern of the polarization . in order to minimize the energy while remaining in a homogeneous configuration , the ferroelectric suffers a progressive reduction of the polarization that show itself well above the critical thickness . this point was first confirmed experimentally by lichtensteiger _ @xcite for pbtio@xmath0 monodomain epitaxial films through x - ray diffraction ( xrd ) measurement of the evolution of tetragonality @xmath70 with thickness . the tetragonality indeed provides indirect but easily accessible information on the polarization through the polarization - strain coupling [ from eq . ( 3 ) , @xmath71 so that its measurement has become a standard technique to investigate size effects in thin films and superlattices.@xcite the reduction of the polarization was further confirmed by polarization data of kim _ et al . _ @xcite in ultrathin srruo@xmath0/batio@xmath0/srruo@xmath0 capacitors free from passive layers,@xcite and more recent x - ray photoelectron diffraction ( xpd ) experiments of despont _ et al._@xcite as it was mentioned before , an alternative way of reducing the electrostatic energy is the formation of 180@xmath72 domains . in other words , going below the critical thickness predicted for monodomain configuration in ferroelectrics does not necessarily imply a paraelectric state.@xcite the existence of 180@xmath72 domains prevent the appearance of a _ net _ interfacial charge density.@xcite this situation is well known in ferroelectric films on insulating substrates , where alternating 180@xmath72 stripe domains structures have been observed and characterized analyzing the satellite peaks in synchrotron x - ray diffraction measurements.@xcite even on conducting substrates , where the free charges of the electrode should provide a substantial amount of screening , the formation of 180@xmath72 domains has been proposed to explain concomitant suppression of polarization and recovery of full tetragonality in pb(zr@xmath73ti@xmath74)o@xmath0 on srruo@xmath0.@xcite this is supported with recent works based on the landau theory,@xcite that conclude that domains always form in thin films , almost irrespective of the nature of the electrode and whether or not the screening carriers may be present in the ferroelectric itself . the situation is more complex experimentally , and whether the system breaks up into domains or not seems to be a very subtle issue . et al . _ , using the same experimental setup , have observed how high - quality ultrathin films of pbtio@xmath0 grown on nb - srtio@xmath0 electrodes remain in a monodomain configuration @xcite ( although with reduced polarization and tetragonality ) whereas they form domains when the electrode is replaced by la@xmath75sr@xmath76mno@xmath0/srtio@xmath0.@xcite even in the case of insulating substrates , the formation of domains depends on how quickly the sample is cooled . monodomain phases of pbtio@xmath0 on insulating srtio@xmath0 have been grown and the structure have been analyzed with a coherent bragg rod analysis ( cobra ) method,@xcite when the samples were cooled from deposition temperature to room temperature over a period of 24 hours , in contrast to the 180@xmath72 stripe periods observed before.@xcite nevertheless , even in the latter case , the system can no more be properly called ferroelectric since , although locally polarized , its polarization is no more expected to be switchable.@xcite from the previous discussion , it is clear that appropriate treatments of the strain and of the incomplete screening of @xmath48 are crucial to describe size evolution of the ferroelectric properties . very recently , many dft calculations as well as first - principles - based effective hamiltonian or atomistic shell - models simulations have been reported addressing the problem of the thickness dependence of the ferroelectric properties in thin films , and the eventual existence of a critical thickness for ferroelectricity . . review of the most recent fully first - principles simulations on the ferroelectric properties of ferroelectric ultrathin films in a monodomain configuration . both computations within the local density approximation ( lda ) and the generalized gradient approximation ( gga ) to the density functional theory have been reported in the literature . ca functional refers to ceperley - alder @xcite in the parametrization of perdew - zunger,@xcite while pw91 and pbe stand for the gga functionals of perdew and wang @xcite and perdew , burke and ernzerhof @xcite respectively . regarding the method , nao stands for numerical atomic orbitals as implemented in the siesta code,@xcite pw is the acronym for plane waves ( as implemented in vasp @xcite in refs . , or dacapo @xcite in refs . ) , and mbpp stands for mixed - basis pseudopotentials.@xcite @xmath77 represents the in - plane lattice constant . if a theoretical in - plane lattice constant is used to fix the length of the in - plane lattice vectors of the heterostructure , it is computed within the same method and approximations than in the simulation of the whole heterostructure . different methodologies yield to slightly different values of the lattice constant . @xmath78 stands for the critical thickness ( if any ) of the perovskite ultrathin film , in units of the number of cells of the ferroelectric perovskite oxide . [ cols="^,^,^,^,^,^,^",options="header " , ] [ table : criticalthickness ] a summary of various results obtained with dft based methods is reported in table-[table : criticalthickness ] . in almost all these simulations , the prototype ferroelectric capacitor considered is made of an insulating layer of a ferroelectric perovskite oxide material ( typically batio@xmath0 , pbtio@xmath0 or knbo@xmath0 ) sandwiched between metallic electrodes ( typically a conductive oxide such as srruo@xmath0 or a noble metal such as pt ) , although lately a free top surface and some adsorbed molecules have been also simulated.@xcite typically the capacitor is built with two symmetric metal - ferroelectric interfaces , although very recently gerra and coworkers @xcite have explored asymmetric heterostructures and linked the first - principles computations with phenomenological approaches . the calculations were performed under short - circuit electrical boundary conditions for the electrodes ( zero bias ) , a requirement that can be satisfied both in a ferroelectric / metal superlattice geometry or in a periodic metal / ferroelectric / metal / vacuum geometry @xcite ( it is only recently that a method has been proposed to address finite bias with periodic boundary conditions @xcite ) . in all the previous approaches , a monodomain configuration was considered , so the electrode was the only source of screening of the polarization charge . the mechanical boundary - conditions were imposed by fixing the in - plane lattice constant and geometry to the one imposed by the substrate . the final constraint depends on the author , but the most common approach was to assume a strain imposed by srtio@xmath0 , the typical substrate on top of which these capacitors are epitaxially grown . the minimum of the kohn - sham energies were then searched either by a frozen - phonon method within the ferroelectric soft mode subspace ( line minimization displacing the atoms in the ferroelectric layer by hand with an amplitude corresponding to a given percentage of the bulk soft mode as defined in sec . [ sec : background ] ) or by a full atomic relaxation of the atomic positions within the whole system . the results are summarized in table-[table : criticalthickness ] and deserve two important comments . first , although an exhaustive comparison of the results would certainly require to consider also other differences in the calculations , it is observed that dft calculations performed within the generalized gradient approximation ( gga ) systematically overestimate the ferroelectric character compared to those within the local density approximation ( lda ) : while lda predicts the existence of a critical thickness for ferroelectricity that ranges between two and six unit cells ( depending on the material , electrode , interface , and strain conditions ) , calculations within the gga often result in the absence of such a limit , especially in pb - based perovskites , with ferroelectric ground states that are stable down to the ultimate thickness of one unit cell , eventually with an enhancement of the polarization.@xcite to interpret this trend it is worth to notice that , although gga might a priori be considered as an improvement over lda , it does not necessarily produce more trustable results in the case of ferroelectric oxides . at the bulk level , the widely used gga of perdew , burke and ernzerhof ( gga- pbe)@xcite has been shown to strongly overestimate the ferroelectric character and , even , to yield an erroneous super - tetragonal structure in pbtio@xmath0 @xcite and batio@xmath0.@xcite only the recently proposed improved gga functional of wu and cohen @xcite solves this problem and produces very accurate structures . the gga results of table-[table : criticalthickness ] are not making use of this new functional so that their comparison with lda results must be made with caution : indeed , the fact that gga overestimates the ferroelectric character and concludes sometimes in the absence of critical thickness might be , at least partly , an artifact similar to that reported at the bulk level . second , it is well - known that the bandgap of the kohn - sham particles as computed within dft is significantly smaller than the experimental gap . for batio@xmath0 , typical lda values are of the order of 1.61.9 ev@xcite while the experimental value is of 3.2 ev . although this dft bandgap problem does not a priori constitute a failure of the theory that should internally correct for it , at least in `` exact '' dft , in order to provide correct ground - state properties , it can yield unphysical results when working with approximate functionals . the band alignment at a typical metal / ferroelectric interface is illustrated in fig . [ fig : barrier]a where , for the purpose of the illustration , the fermi level of the metal is assumed to be located roughly mid - gap , as typically expected from the experiment . the dft bandgap underestimate prevents accurate predictions of the barriers simultaneously for electrons , @xmath79 , and holes , @xmath80 , and can in many cases produce pathological situations where the fermi level of the metal is erroneously aligned with the conduction bands of the ferroelectric ( @xmath79<0 ) , producing unphysical population of the conduction bands of the ferroelectric ( fig . [ fig : barrier]c ) . such a situation was , for instance , reported by the authors of ref . for the pt / srtio@xmath0 interface , the structure of which had to be artificially modified to avoid the problem , or also appears from the analysis of the partial density of states of the pt / batio@xmath0 interface in ref . . moreover , even in `` normal '' cases where the dft calculation reproduces a small barrier for electrons in a non - polarized configuration ( fig . [ fig : barrier]b ) , the situation remains problematic since the underestimate of the electron barrier artificially lowers the breakdown field : if during the relaxation of the ferroelectric state , the residual depolarizing field exceeds the erroneously low dft breakdown field , a spurious transfer of electrons from the metal to the ferroelectric conduction bands will be observed at the interface , modifying its properties and producing an unphysical relaxed structure . for the srruo@xmath0/batio@xmath0/srruo@xmath0 nanocapacitor of ref . , we obtained within the lda for the non - polar configuration @xmath81 ev and @xmath82 ev , while the use of an empirical `` scissors '' correction ( i.e. a rigid shift in energy of the batio@xmath0 conduction bands of 1.62 ev , to reproduce the experimental bandgap ) provides a more realisitic value @xmath83 ev.@xcite the artificially low value of @xmath84 prevented us to perform full atomic relaxations of the ferroelectric state but was not addressed by other authors reporting full relaxation on the same system . all these issues highlight that special care must be taken when dealing with metal / ferroelectric interfaces and invite to be specially critical when interpreting the results . an important challenge at this point for the theoriticians is to identify an alternative functional avoiding the bandgap problem and its consequences on the band alignment , while preserving accurate description of the structural and ferroelectric properties . the b1 hybrid functional recently proposed by bilc _ _ @xcite might constitute an promising option . in spite of the technical difficulties highlighted above , first - principles dft simulations remain the most accurate model currently available for the theoretical study of ferroelectric capacitors and provides a wealth of informations at the atomic level . calculations are however performed at 0 k , so that the study of the temperature dependence of the structural , ferroelectric , and piezoelectric properties is not accessible . also , dft calculations are computationally very demanding and the effort required to deal with such complex heterostructures , involving interfaces between dissimilar metal / insulator materials , is at the limit of present capabilities . to bypass these intrinsic limitations and address more complex questions , alternative less accurate methods have been proposed that remain based on first - principles results . these include effective hamiltonian,@xcite shell model,@xcite and molecular dynamics simulations combined with coarse - grained monte carlo simulations,@xcite where the different parameters are directly fitted on dft results . wu and coworkers @xcite have performed effective hamiltonian simulations on pb(zr@xmath85ti@xmath85)o@xmath0 free standing slabs . the choice of the ti and zr composition is interesting , since it is close to the morphotrophic phase boundary of bulk pzt . since the calculations were done in the absence of surface charge screening , films thicker than 4 unit cell form 180@xmath72 domains in the direction normal to the surface to screen the depolarizing field , so the average component of the out - of - plane local mode vanishes , independently of the strain . the depolarizing field does not influence the in - plane polarization versus strain behaviour . a rich strain - temperature phase diagram , including a monoclinic and a stripe domain phase not present in the bulk parent material was predicted . for the same composition of pb(zr@xmath85ti@xmath85)o@xmath0 , kornev _ et al . _ @xcite extended the effective hamiltonian in order to incorporate explicitly the presence of a free surface and the effects of a partial compensation of the depolarizing field . as in ref . , in this first attempt , only the net component of the out - of - plane polarization at the surface layers was considered to compute the homogeneous depolarizing field and the corresponding contributions to the energy . the treatment of @xmath48 was then refined by ponomareva and coworkers,@xcite who corrected the dipole - dipole interactions of the bulk model hamiltonian by summing explicitely contributions of the local dipoles present within the film , and derived in this way an _ exact _ expression for the depolarizing energy and field at an atomistic level in any low dimensional ferroelectric structure under open circuit ( oc ) boundary conditions . if the polarization lies along a non - periodic direction , the oc boundary conditions naturally lead to the existence of a maximum of the depolarizing field that is later screened by hand . the amount of screening and therefore the amplitude of the residual field can be tuned with a parameter , @xmath86 ( see section iii - b ) , that plays a role similar to the effective screening length in the model described above and ranges between @xmath86 = 0 ( oc boundary conditions ) and @xmath86 = 1 ( perfect short circuit boundary conditions ) . this model has been applied to different kinds of low dimensional ferroelectrics.@xcite in ref . , the interplay between the screening of the depolarizing field and the strain imposed by the substrate was also explored . in the limit of perfect screening , a uniform polarization along @xmath9 is predicted , while the behaviour of the in - plane components of the polarization depends on the strain , following the general rule described in sec . [ sec : strain ] : a compressive strain supresses the in - plane polarization while a tensile strain favours the presence of a @xmath42-phase . therefore , the phase change from a monoclinic phase ( the so - called @xmath87 phase with @xmath88 , @xmath89 ) for sufficiently large tensile strains to a tetragonal phase with @xmath90 , @xmath91 for stress free or compressive strains . on the other extreme case of open - circuit boundary conditions , the _ average _ polarization along @xmath9 vanishes independently of strain , while the in - plane polarization aligns with the [ 010 ] direction ( @xmath92 ) for the stress free , displays a monoclinic @xmath93 phase ( @xmath94 , @xmath95 ) under tensile strain , or vanishes for compressive strain . for intermediate values of the depolarizing field , the polarization continuously rotates and passes through low - symmetry phases . interestingly , these phases are stable only for compositions lying near the morphotrophic phase boundary of bulk pzt , disappearing in favour of a tetragonal phase for high enough ti - concentrations . the fact that the _ average _ out - of - plane polarization vanishes under open circuit boundary conditions does not mean that the local dipoles along @xmath9 vanishes . on the contrary , they might have a large value , close to the bulk one , but forming domains . the process of formation of these domains was monitored with respect the screening of the depolarizing field . when @xmath48 is large enough , some `` bubbles '' ( cylindric nanodomains having local dipoles that are aligned in an opposite direction with respect the one in the monodomain configuration ) nucleate and grow laterally till 180@xmath72 stripe domains are formed . model hamiltonian techniques have the additional advantage of a trivial treatment of external electric fields . starting from a five unit cell thick thin film geometry under compressive strain , where @xmath41-oriented 180@xmath72 stripe domains are stabilized even assuming a realistic amount of screening of the depolarizing field , the evolution of the domain structure under an electric field has been studied for pb(zr@xmath85ti@xmath85)o@xmath40 @xcite and batio@xmath40,@xcite giving some light into the mechanisms to saturate the films . the path is essentially the same for both perovskites oxides and consist of three different steps : @xmath96 the growth of the domain parallel to the field at the expense of the other up to a first threshold value of the external field ; @xmath97 the formation of the nanobubbles defined before ; and @xmath98 the contraction of the nanobubbles till the final collapse into a monodomain configuration at a second critical value of the external field . the main differences between batio@xmath40 and pb(zr@xmath85ti@xmath85)o@xmath40 are @xmath96 the orientation of the domain walls : the stripes alternate along [ 110 ] for batio@xmath40 rather than along [ 100 ] for pb(zr@xmath85ti@xmath85)o@xmath40 , and @xmath97 the `` hardness '' of the polarizarion in batio@xmath0 to rotate ( batio@xmath40 films profoundly dislike significantly rotating and in - plane dipoles ) . the formation and response of the `` bubbles '' to external electric fields and the recent predictions of domains of closure @xcite ( very common in ferromagnets but detected in ferroelectric thin films only recently @xcite ) , have bridged a fascinating parallelism between ferroelectric and ferromagnetic domains . although the shape and size of the `` bubbles '' ( elliptical in ferroelectrics and spherical in ferromagnets ) , the abruptness of the domain walls ( one lattice constant thick in ferroelectrics and much more gradual , over many atomic planes , in ferromagnets ) , or the exceptionally small periodicity and constancy with the field of the ferroelectric domain periods differ from the counterpart ferromagnetic properties , they might not be so different as traditionally considered , despite the profound differences between electrostatic and magnetostatic interactions . domains of closure have been predicted using a first - principles effective hamiltonian for pb(zr@xmath99ti@xmath100)o@xmath0 @xcite asymmetrically screened ( grown on a nonconducting substrate and with a metal with a dead layer as top electrode ) , and using a landau - ginzburg phenomenological approach for a pbtio@xmath0 thin film,@xcite both asymmetrically and symmetrically coated with insulating srtio@xmath0 . these predictions have obtained further credit after full first - principles simulations on srruo@xmath40/batio@xmath40/srruo@xmath40 ferroelectric capacitors by aguado - puente and junquera,@xcite where the domains of closure are obtained even for a symmetrical metal / ferroelectric / metal capacitor , where the metallic plates should provide significant screening . as it happens with ferromagnetic domains , ferroelectric domains follow the kittel law ( that states that the domain width is directly proportional to the square root of the sample s thickness ) , at least for thicknesses thicker than 1.2 nm.@xcite the effective hamiltonian approach is now considered as a standard method and has been used to address various other questions . the phase diagram of pzt films as a function of temperature and ti composition under open - circuit boundary conditions has been reported in ref . . other systems like batio@xmath0 have also been considered for which the strain / temperature phase diagram has been obtained and the effect of incomplete screening of @xmath101 has been discussed.@xcite although most studies focused on [ 001 ] epitaxial films , the influence of the growth direction on the properties was also investigated.@xcite besides all the previous theoretical frameworks , molecular dynamics combined with coarse - grained monte carlo simulations have been recently used to study a very important tecnological problem : the swithching of the ferroelectric polarization.@xcite these simulations show that the shape of the critical nucleus is square and not a thin triangular plate as suggested by the prevailing miller - weinrich model . the critical nuclei grow in an anysotropic way , with activation barriers for sideways and forward growth much smaller than for nucleation . the values for the activation energies and activation fields predicted by the molecular dynamics simulations are one order of magnitude smaller than those expected from the miller - weinrich theory , mainly due to the new slanted interface model for the surface of the nuclei , and the domain velocities are in very good agreement with the ones measured experimentally.@xcite artificial ferroelectric superlattices are nowadays considered as an interesting alternative to ferroelectric thin films @xcite since , as it is discussed below , they allow fine tuning of the ferroelectric properties while maintaining prefect crystal ordering . the large variety of materials that can be combined ( ferroelectrics , incipient ferroelectrics or regular insulators ) and the way these materials can be ordered within the structure ( bicolor and tricolor superlattices or even more complex structures with composition gradients ) also offers tremedous scope for creating artificial ferroelectric materials with possibly new and tailor - made properties . during the recent years , short - period epitaxial multilayers combining a ferroelectric and an incipient ferroelectric material ( like batio@xmath0/srtio@xmath0,@xcite pbtio@xmath0/srtio@xmath0,@xcite knbo@xmath0/ktao@xmath0 , @xcite and related tricolor systems like batio@xmath0/srtio@xmath0/catio@xmath0 @xcite ) have retained most of the attention both at the theoretical and experimental level . in these systems , proper handling of the mechanical and electrical boundary conditions appears crucial to monitor the ferroelectric properties in a way very similar to what was previously discussed for thin films . first , the alternation of ultrathin layers is particularly suited to impose coherent epitaxial strains and achieve `` strain engineering '' of the properties . the compressive epitaxial strain produced by a srtio@xmath0 substrate on batio@xmath0 layers contributes to explain polarization enhancement with respect to the bulk batio@xmath0 value in batio@xmath0/srtio@xmath0 bicolor superlattices @xcite and even in tricolor @xcite superlattices , despite the fact that part of the system is not nominally ferroelectric . the epitaxial strain additionally produces significant shift of the phase transition temperature as for instance reported for batio@xmath0/srtio@xmath0 @xcite and pbtio@xmath0/srtio@xmath0 @xcite superlattices . ferroelectricity is even expected in srzro@xmath0/srtio@xmath0 superlattices on srtio@xmath0 , in spite of the fact that neither srzro@xmath0 nor srtio@xmath0 is ferroelectric at the bulk level.@xcite although it was not stated by the authors , this can probably be partly explained from the large compressive strain imposed on the srzro@xmath0 layers by the srtio@xmath0 substrate . second , electrostatic coupling between the layers is also playing a key role . on the one hand , polarizing the ferroelectric batio@xmath0 layers only in a batio@xmath0/srtio@xmath0 superlattice would produce a large polarization gradient at the interfaces , inducing huge depolarizing fields and electrostatic energy . on the other hand , polarizing srtio@xmath0 also has an energy cost since this material is not spontaneously polarized . the ground - state of the system will therefore depend on the competition between different energy terms and can strongly evolves depending of the compounds involved and the superlattice period . in short - period superlattices where the ferroelectric alternates with a highly polarizable compound such as an incipient ferroelectric ( like srtio@xmath0 or ktao@xmath0 ) , a polarization will be induced in the latter in order to avoid depolarizing fields and the superlattice will typically exhibit a homogeneously polarized configuration ( see fig . [ fig : superlattice ] ) . first pointed out by neaton and rabe on batio@xmath0/srtio@xmath0 superlattices , such a behavior was also theoretically predicted for pbtio@xmath0/srtio@xmath0,@xcite or even batio@xmath0/srtio@xmath0/catio@xmath0 @xcite short - period superlattices . in many theoretical studies , the layer polarization was estimated roughly , multiplying the atomic displacements into the layer by the bulk born effective charges . a more rigorous and accurate approach has been proposed recently based on wannier functions.@xcite in a recent work , dawber _ et al . _ @xcite have demonstrated that key ferroelectric properties can be tuned over a very wide range in pbtio@xmath40/srtio@xmath40 superlattices : it was shown that the polarization can be adjusted from 0 to 60 @xmath102 and the transition temperature from room temperature to 700@xmath72c by properly adjusting the volume fraction of pbtio@xmath0 . moreover , a simple landau - based model was proposed that allows to determine a priori the properties of the system , so opening the way to the production of samples with ferroelectric properties designed for particular applications . it was also reported that the presence of srtio@xmath0 between the pbtio@xmath0 layers in the superlattice improve the electrical properties and allows to avoid the large leakage currents typically reported in pbtio@xmath0 thin films while maintaining perfect crystal structure . the simple arguments described above have been shown to provide reasonable description of ( pbtio@xmath0)@xmath103/(srtio@xmath0)@xmath0 superlattices over a relatively wide range of pbtio@xmath0 thicknesses ( for @xmath104 ranging from 1 to 53).@xcite they are however expected to apply mainly to relatively short - period systems . in longer period superlattices , other behaviors are expected . in knbo@xmath0/ktao@xmath0 @xcite it was theoretically shown using a shell - model that the electrostatic coupling between layers ( responsible for a homogeneous polarization throughout the system ) will progressively decrease as the period increases , yielding a progressive reduction of ktao@xmath0 polarization . in batio@xmath0/srtio@xmath0,@xcite effective hamiltonain simulations have confirmed that short - period superlattices behave like a homogeneous - like material with a phase diagram that resembles that of a ( 001 ) batio@xmath0 thin film under short - circuit conditions while inhomogeneous atomic features and original domain patterns were predicted in longer period systems . appearance of domain structures in ferroelectric / paraelectric superlattices was also discussed by stephanovich _ et al._@xcite the previous discussion implicitly focused on the evolution of the out - of - plane polarization . in some cases , an in - plane component of the polarization can also eventually appear . this was predicted in knbo@xmath0/ktao@xmath0 superlattices in which the epitaxial compressive strain is not strong enough to force knbo@xmath0 ( with a rhombohedral ground - state ) to become tetragonal.@xcite in a way compatible with a symmetry lowering observed experimentally,@xcite an in - plane polarization component was also predicted in the srtio@xmath0 layers of batio@xmath0/srtio@xmath0 superlattices under in - plane expansion.@xcite the main difference between the out - of - plane and in - plane polarization components is that while the out - of - plane components of the different layers are strongly electrostatically coupled as previously discussed , the in - plane components behaves much more independently and remain confined in the individual layers in which they are induced by the epitaxial strain conditions imposed by the substrate . the interest for ferroelectric superlattices is not limited to the fact that they allow the tuning of the ferroelectric properties as discussed above but also results from their potentiality to generate totally new behaviors or phenomena . in pb(sc@xmath105nb@xmath106)o@xmath40 ( psn ) @xcite and pb(zr@xmath39ti@xmath38)o@xmath40 ( pzt ) @xcite superlattices , it is expected that symmetry lowering can yield new stable polar phases and strong enhancement of the functional properties . unusual thermoelectric properties and nonergodicity were also predicted in pzt superlattices.@xcite in very short - period pbtio@xmath0/srtio@xmath0 superlattices an unexpected recovery of ferroelectricity has been observed in the limit of ultrathin pbtio@xmath0 layers that can not be explained from the simple arguments reported above.@xcite . beyond such bicolor systems , tricolor superlattices , which break the inversion symmetry @xcite are also expected to potetially exhibit unusual and attractive properties . first - principles based simulations contributed to make significant progresses in the understanding of the behavior of ferroelectric superlattices and even opened the door to a possible theoretical design of artificial ferroelectric nanostructures with predetermined and tailor - made properties.@xcite although most of the recent experimental and theoretical efforts on ferroelectric nanostructures concerned thin films geometries , the growth and characterization of ferroelectric nanowires,@xcite nanotubes,@xcite nanocolumns , @xcite nanoislands,@xcite or even `` nanosolenoids '' @xcite are at an incipient and exciting state . the geometries of these nanostructures favor novel inhomogeneous polarization and strain configurations , strongly dependent of the electrical and mechanical boundary conditions . dft first - principles studies of such low dimensional structures are , however , computationally very demanding . other recent studies made use of first - principles based shell - models @xcite and model hamiltonian methods.@xcite for instance , domain structures , polarization and coercive fields of nanoscopic particles of batio@xmath0 have been carried out using such methodologies . perhaps , the most remarkable result is the prediction of novel exotic geometric ordering of ferroelectricity in batio@xmath0 @xcite and pb(zr@xmath85,ti@xmath85)o@xmath0 @xcite nanoscale disks and rods , where the the local polarization rotate from cell to cell , forming a `` vortex - like '' pattern . although the macroscopic polarization of these quantum dots vanishes at zero applied field , they exhibit a new order parameter , the _ toroid moment _ of polarization , introduced by naumov _ _ @xcite and defined as @xmath107 where @xmath108 is the local dipole of cell @xmath21 located at @xmath109 , and @xmath110 is the number of unit cells in the simulation . since the local dipoles in the vortex can rotate either clockwise or anticlockwise , the toroid moment might adopt two different values pointing in opposite directions , so that one bit of information might be stored by assigning one value of the boolean algebra ( `` 1 '' or `` 0 '' ) to each of these states . the direction of the toroidal moment can be efficiently controlled by a transverse inhomogeneous electric field generated by two opposite charges located away from the studied dot,@xcite opening exciting oportunities for nanomemory devices . technologically , the ferroelectric vortex phase promises to increase the storage density of ferroelectric random access memories ( fe - ram ) by five orders of magnitude . in order to investigate the possibility of engineering the dipole patterns and vortex structures , a study of the phase diagram as a function of temperature of these ferroelectic nanodots when they are embedded in a polarizable medium was reported by prosandeev and bellaiche.@xcite six different phases were found , depending on the ferroelectric strength of the material constituting the dot and of the system forming the medium . some of the novel phases are remarkable , for instance the coexistence of a toroidal moment and a spontaneous polarization at low t when a soft ferroelectric dot is immersed in a medium that is ferroeletrically harder than the dot . medium driven interactions between dots were demonstrated with adjacent vortices rotating in an opposite fashion ( `` antiferrotoroidic '' phase ) . the phase transition from the vortex structure ( with the local dipoles circularly ordered ) to a ferroelectric phase ( with all the local dipoles pointing along the same direction and with an out - of - plane polarization ) in a pb(zr@xmath85ti@xmath85)o@xmath40 nanocylinder by an electric field perpendicular to the plane of the vortex has been recently addressed.@xcite the transition between the two phases is through an intermediate state that generates a lateral toroid moment and breaks the macroscopic cylindrical symmetry . the observation of this new phase yielded to the discovery of a stricking collective phenomenon , that leads to the annihilation of the ferroelectric vortex in a peculiar azimuthal mode . a first step to explain the existence of this toroidal moment in ferroelectric nanoparticles from a dgl theory has been taken by wang and zhang @xcite a more exhaustive review of all these computations can be found in ref . . in between the two - dimensional ( 2d ) thin films and superlattices discussed in secs . [ sec : films ] and [ sec : superlattices ] , and the zero - dimensional ( 0d ) nanoparticles described above , the case of one - dimensional ( 1d ) nanowires was also recently addressed . model hamiltonian simulations on pb(zr , ti)o@xmath40 @xcite and full first - principles simulations on batio@xmath40 nanowires @xcite have demonstrated how long - range ordering exists in these 1d systems even at finite temperature and under zero external field . these results challenged traditional statistical models that claimed that 1d lattices with particle - particle interactions decaying faster than @xmath111 are impossible to have phase transitions at finite t. as it happened in thin films and nanoparticles , the mechanical and electrical boundary conditions are essential in the nanowires . the presence of uncoated free surfaces yields to the existence of depolarizing fields along two directions and forces the collective polarization to lye along the longitudinal direction . a critical diamenter for the polarization was estimated around 1.2 nm for batio@xmath40 @xcite and 2.0 nm for pb(zr , ti)o@xmath40,@xcite below which the ferroelectricity was suppresed . at smaller radii , low atomic coordinations at the surface produce a contraction of the unit cell that is responsible for the suppression of the ferroelectric distortion at the critical radius . nevertheless , even below this threshold size of the diameter , ferroelectricity might be recovered playing with the mechanical boundary conditions , by applying appropriate tensile strain . previous first - principles simulations on nanowires also contributed to get a new insight into the concept of the ferroelectric correlation volume , defined as the smallest volume within which atomic displacements must be correlated in order to decrease the energy of the crystal and produce a stable polar entity . analyzing the critical diameter of the wire as a function of its length , the computations prove that is the anisotropic shape of the polar domain rather than its volume that determines the stability of its ferroelectric state.@xcite as illustrated throughout this review , various progresses have been reported during the last years concerning the description of the ferroelectric properties in oxide ultrathin films and superlattices . in this context , first - principles theory and modeling constituted an efficient guide and provided a significant support to experimental works . on the one hand , proper understanding of the crucial role played by the electrical and mechanical allowed to realize significant advances as discussed above . on the other hand , however , many issues still remain unclear so that this field of research is still at an exciting and incipient stage . in ferroelectric thin films , despite recent achievements , some important questions remain without clear answer and interesting perspectives ask for further confirmations . only recently addressed from first - principles,@xcite the mechanism and the quality of the screening at free surfaces with air remains unclear . also the reason why some ferroelectric films on metallic electrodes stay monodomain with a reduced polarization while others break into domains is particularly intriguing.@xcite the roles of defects , such as oxygen vacancies , and of leakage currents on the amplitude of the depolarizing field was up to now neglected in theoretical studies while they might play a role , at least in some cases . the mechanism of the switching and domain - wall motion in ultrathin films is an important question that has been studied in detail experimentally @xcite but that has not been addressed yet at the first - principles level , although the first - steps on multi - scale computations based on molecular dynamics and coarse - grained monte carlo simulations have been recently taken.@xcite recent studies reported interesting perspectives for ferroelectric tunnel junctions @xcite but no first - principles quantitative estimate of the tunneling current and of the electroresistance has been reported yet . finally and without being exhaustive , the case of a ferroelectric layer between magnetic electrodes was proposed as a mechanism to achieve ferroelectric control of magnetism @xcite but is still widely unexplored . ferroelectric superlattices constitute a promising alternative avenue that offers practical solutions to bypass some of the drawbacks of epitaxial ultrathin films ( huge depolarizing fields , high leakage currents , strain relaxation ) , while preserving perfect crystal ordering and allowing to monitor the ferroelectric properties . up to now , fine tuning of the polarization and phase transition temperature has , for instance , been illustrated playing with the superlattice period and composition but the most exciting perspectives probably rely on the potentiality of such systems to generate totally new and unexpected phenomena . also , superlattices might offer particularly attractive opportunities in the emergent field of magneto - electric multiferroics.@xcite the perspectives are numerous and first - principles simulations will certainly continue to play an active role in the field of nanoscale ferroelectrics in the future . in some cases , this will mandatory require new theoretical developments such as , for instance , the identification of alternative and more accurate functionals or of new efficient approaches to compute complex properties . in order cases , the computer power will constitute the main limitation . in all cases , pursuing the constructive dialog initiated during the recent years between theorists and experimentalists will certainly appear as a major motivation . the authors thank j .- m . triscone , m. dawber , c. lichtensteiger , h. kohlstedt , m. alexe , g. geneste , e. bousquet , and p. aguado - puente for valuable discussions . phg acknowledges financial support from the volkswagenstiftung ( i/77 737 ) , the interuniversity attraction poles program - belgian state - belgian science policy ( p6/42 ) , the fame european network of excellence and the macomufi european strep project . jj acknowledges financial support by the spanish mec under project fis2006 - 02261 , and the australian research council arc discovery grant dp 0666231 .

in this review , we propose a summary of the most recent advances in the first - principles study of ferroelectric oxide epitaxial thin films and multilayers . we discuss in detail the key roles of mechanical and electrical boundary conditions , providing to the reader the basic background for a simple and intuitive understanding of the evolution of the ferroelectric properties in many nanostructures . going further we also highlight promising new avenues and future challenges within this exciting field or researches . to appear in `` _ journal of computational and theoretical nanoscience _ ''

introduction background epitaxial thin films superlattices nanoparticles and nanowires conclusions and perspectives acknowledgements

we would like to thank emil martinec for many helpful discussions . sdm is supported in part by doe grant de - sc0011726 . s. w. hawking , commun . phys . * 43 * , 199 ( 1975 ) [ erratum - ibid . * 46 * , 206 ( 1976 ) ] ; s. w. hawking , phys . d * 14 * , 2460 ( 1976 ) . s. d. mathur , class . quant . grav . * 26 * , 224001 ( 2009 ) [ arxiv:0909.1038 [ hep - th ] ] . s. d. mathur , fortsch . phys . * 53 * , 793 ( 2005 ) [ arxiv : hep - th/0502050 ] ; i. bena and n. p. warner , lect . notes phys . * 755 * , 1 ( 2008 ) [ arxiv : hep - th/0701216 ] ; v. balasubramanian , e. g. gimon and t. s. levi , jhep * 0801 * , 056 ( 2008 ) [ arxiv : hep - th/0606118 ] ; + k. skenderis and m. taylor , phys . rept . * 467 * , 117 ( 2008 ) [ arxiv:0804.0552 [ hep - th ] ] . ; e. j. martinec , arxiv:1409.6017 [ hep - th ] . e. witten , nucl . b * 195 * , 481 ( 1982 ) . b. d. chowdhury and s. d. mathur , class . grav . * 25 * , 135005 ( 2008 ) [ arxiv:0711.4817 [ hep - th ] ] . s. d. mathur , arxiv:0805.3716 [ hep - th ] ; s. d. mathur , int . j. mod . d * 18 * , 2215 ( 2009 ) [ arxiv:0905.4483 [ hep - th ] ] . p. kraus and f. wilczek , nucl . b * 433 * , 403 ( 1995 ) [ gr - qc/9408003 ] ; p. kraus and f. wilczek , nucl . b * 437 * , 231 ( 1995 ) [ hep - th/9411219 ] . e. keski - vakkuri and p. kraus , nucl . phys . b * 491 * , 249 ( 1997 ) [ hep - th/9610045 ] . s. d. mathur and c. j. plumberg , `` correlations in hawking radiation and the infall problem , '' jhep * 1109 * , 093 ( 2011 ) [ arxiv:1101.4899 [ hep - th ] ] ; s. d. mathur and d. turton , nucl . b * 884 * , 566 ( 2014 ) [ arxiv:1306.5488 [ hep - th ] ] .

the information paradox can be resolved if we recognize that the wavefunctional in gravity @xmath0 $ ] should be considered on the _ whole _ of superspace , the space of possible @xmath1 . the largeness of the bekenstein entropy implies a vast space of gravitational solutions , which are conjectured to be fuzzball configurations . in the wkb approximation , the wavefunctional for a collapsing shell is oscillatory in a small region of superspace , and the classical approximation picks out this part . but the wavefunctional will be damped ( ` under the barrier ' ) in the remainder of this vast superspace . we perform a simple computation to show that at the threshold of black hole formation , the barrier is lowered enough to make the latter part oscillatory ; this alters the classical evolution and avoids horizon formation . epsf = = = = = = = = ============= == = = nature abhors a horizon + per kraus@xmath2 and samir d. mathur@xmath3 @xmath2department of physics and astronomy , ucla , los angeles , ca 90095 - 1547 , usa + pkraus@ucla.edu .1 in @xmath3 department of physics , the ohio state university , columbus , oh 43210 , usa + mathur.16@osu.edu + march 31 , 2015 1.0 true in there are two central mysteries in black hole physics : ( i ) what does the enormously large entropy of a black hole signify ? ( ii ) how can information emerge from the hole ? in this essay we will show how understanding the first of these puzzles also resolves the second . many physicists had a nagging feeling that hawking s argument for information loss @xcite was flawed : subtle corrections could somehow encode information in the outgoing radiation . we now know , however , that the hawking argument is stable to all small corrections ; using strong subadditivity of entanglement entropy we can show that information _ will _ be lost unless physics at the horizon is altered by _ order unity _ @xcite . fortunately , string theory supports a plethora of solutions that look like black holes beyond the radius of a would - be horizon , but which are in fact horizon free . they exist because of the emergence of a new topological structure : a compact direction in the geometry pinches off , ending the spacetime smoothly just outside the location where the horizon would have appeared ( fig.[fone]a ) @xcite . these ` fuzzball ' states , which have no horizon , are expected to agree in count with @xmath4 $ ] , where @xmath5 is the bekenstein entropy . their structure is reminiscent of the ` bubble of nothing ' @xcite where a region of spacetime is removed and the pinching of a compact circle at the boundary of this region yields a geodesically complete manifold ( fig.[fone]b ) . if black hole microstates have no horizon , then we do not have hawking s pair creation process , and thus no information loss ; the fuzzball radiates from its surface ( at the hawking rate @xcite ) just like a star . but one may ask : if we start with a collapsing shell of mass @xmath6 , then ca nt we just trust the classical approximation where the shell collapses through @xmath7 to create a horizon ? if such a horizon is indeed created , then light cones inside the horizon point towards smaller @xmath8 . thus causality would forbid any effects that happen at @xmath9 from altering the vacuum state at @xmath7 . it would then appear that we can never get to a fuzzball state at the horizon , so how have we resolved the information paradox ? to see the way around this conundrum first consider a toy model , where we assume that the theory contains a very large number @xmath10 of massless scalar fields @xmath11 . if n ( mm_p)^2 [ one ] then the collapsing shell will not make a black hole . particle creation in curved space starts before the shell reaches its horizon . for each species @xmath11 we have @xmath12 quanta radiated while the shell collapses from , say , @xmath13 to @xmath14 m. each quantum has wavelength @xmath15 ( the only length scale in the problem ) , so if the shell indeed reached down to @xmath16 , then the total mass radiated in all the fields @xmath11 would be m_rad~ngm m [ twop ] where in the last step we have used ( [ one ] ) and noted that @xmath17 . thus the shell radiates away before reaching the vicinity of @xmath7 . in string theory we do not have a large number of massless fields , but we do have a large number of _ massive _ fuzzball states . we call these states @xmath18 ; each has mass @xmath6 , and the number of such states is @xmath4 $ ] . in @xcite it was argued that a collapsing shell can tunnel into any given @xmath18 at a very small rate @xmath19 $ ] , where @xmath20 . but the number of states @xmath18 we can transition to is @xmath4 $ ] . so the overall rate of transitioning to fuzzballs is ~n_fuzzballs exp[-s_bek]~exp[s_bek]exp[-s_bek ] [ two ] if @xmath21 , then we transition to fuzzballs quickly , and thus never make a horizon . this would remove the information paradox . signs ) , and fluxes between the monopoles ( shaded ovals ) support the structure . the region which would have been the interior of the horizon is absent . different choices of fluxes and @xmath22 signs at different angular positions give the @xmath4 $ ] horizonless states of the hole . ] unfortunately , while one could estimate that @xmath20 , there was no immediate way to show that @xmath21 . we now close this loophole by using an old result to argue that in fact @xmath23 . let us begin by recalling hawking s radiation rate for quanta of energy @xmath24 ~e^-t~e^-8gm in @xcite this expression was modified to include backreaction . for radiation of a spherical shell of _ any _ energy @xmath25 , we get ~e^-8 g ( m-2 ) this is a special case of a general expression valid for all holes in all dimensions @xcite ~e^s_bek(m-)-s_bek(m)e^-s_bek we are interested in the case where the entire mass @xmath6 in the shell transitions to a fuzzball state @xmath18 of energy @xmath26 . thus @xmath27 in this case , and e^-s_bek~e^s_bek(m-)-s_bek(m)~e^-s_bek(m ) thus we see that for any one species @xmath18 we get ~e^-s_bek(m ) giving = 1 [ ten ] equation ( [ ten ] ) is the main result of this article . the rate at which the geometry tunnels to fuzzballs is small ( as expected ) , but we see ( eq.([two ] ) ) that this smallness is _ exactly _ cancelled by the largeness of the bekenstein entropy , which governs how many final states we can tunnel to . we therefore have the picture depicted in fig.[fsixdel ] : as the shell approaches the horizon ( in the schwarzschild frame ) , its state transitions to a linear combination of fuzzball states @xmath18 , at such a rate that there is no amplitude left in the shell state itself by the time the shell reaches @xmath7 . thus the classical approximation is completely violated , and no horizon forms . this resolves the information paradox . to extract the moral of this story , consider again the toy model which gives ( [ twop ] ) we can imagine that the collapsing shell is made of gravitons . if we consider only classical physics for all fields , then the scalars @xmath11 would never be involved ; the shell would pass through its horizon to create a black hole . what prevents horizon formation is the change in the quantum vacuum state of the @xmath11 caused by the mass @xmath6 in the shell ; this change of vacuum leads to pair production . the largeness of @xmath10 ( eq . ( [ one ] ) ) makes the virtual fluctuations of the @xmath11 sufficiently important to modify the classical evolution of the shell completely . with fuzzballs , the gravitational theory has a large number @xmath28 $ ] of states @xmath18 of mass @xmath6 . in the minkowski vacuum , we do not have the energy to create these states on - shell . but we should not ignore the vacuum fluctuations into this very large set of states . the infalling shell alters these vacuum fluctuations . the energy of the shell gets converted to on - shell states @xmath18 , and the classical motion of the shell is completely invalidated . put another way , we should always consider the wavefunctional of gravity on the _ whole _ of superspace , the space of all spatial geometries . suppose we write the superspace wavefunctional @xmath0 $ ] of the collapsing shell in the wkb approximation . the wavefunctional would be oscillatory on some part of superspace the classically allowed region but will be exponentially damped ( ` under the barrier ' ) in the rest of superspace . classical physics focuses on the oscillatory part , but we will now show by a toy model that the ` under the barrier ' part of the wavefunction can become very important when there are a large number of degrees of freedom . consider the quantum mechanical problem of a particle in the 1-dimensional potential @xmath29 depicted in fig.[ffour](a ) . assume that _ x=0^a dx |(x)|^2 = 1- , 1 so that most of the norm is in the well @xmath30 . now consider the particle in @xmath10 dimensions with potential v(x_1 , x_2 , x_n)= v(x_1)+v(x_2)+ v(x_n ) and take n1/ the wavefunction is = ( x_1)(x_2) (x_n ) the norm in the central well @xmath31 is now ( _ x_1=0^a dx_1 |(x_1)|^2 ) ( _ x_2=0^a dx_2 |(x_2)|^2 ) ( _ x_n=0^a dx_n barrier_. to see the dynamical consequences of this fact , imagine that at @xmath32 the barrier is altered to that in fig.[ffour](b ) . for one copy @xmath33 , suppose that the wavefunction leaks out to infinity at a slow rate _ x=0^a dx |(x , t)|^2 ( 1- ) e^-t , 1 but for the case of @xmath34 dimensions , the norm in the central well ( _ x_1=0^a dx_1 |(x_1 , t)|^2 ) ( _ x_2=0^a dx_2 |(x_2 , t)|^2 ) ( _ x_n=0^a dx_n |(x_n , t)|^2 ) ( 1-)^n e^-nt decays almost instantly , in a time t ~1/(n)1 let us summarize . it was known that black holes are characterized by a remarkably large entropy @xmath5 , but we did not know what this largeness signified for the dynamics of the hole . with the advent of the fuzzball construction it is now conjectured that @xmath4 $ ] counts explicitly constructable states fuzzballs which are normal quantum gravitational states without horizon . under normal conditions ( i.e. when we are not close to making a black hole ) these states lie in the regime of superspace where the wavefunction is ` under the barrier ' , so we do not directly see these states . but as a collapsing shell approaches its horizon , the wavefunctional tunnels into the ensemble of these states at rate that is _ not _ suppressed ; the smallness of tunneling rate to each fuzzball is exactly cancelled by the largeness of the number of fuzzballs . the norm of the wavefunction captured by these fuzzball states thus grows to unity , just as in the toy example of a particle in an @xmath10 dimensional potential well with @xmath10 large . this tunneling alters the classical dynamics of a collapsing shell completely and resolves the information paradox , since the fuzzballs that result have no horizon . it is now interesting to ask how this altered dynamics should be described ; in particular we have a tantalizing conjecture of fuzzball complementarity @xcite which says that collective oscillations of the fuzzball surface approximately mimic , in a _ dual _ description , the dynamics of the black hole interior . this approximation allows information to escape in low energy @xmath35 quanta of hawing radiation , while dynamics of @xmath36 quanta mimics classical physics in the dual picture . , and a barrier starting at @xmath37 . most of the wavefunction @xmath33 is located in the well @xmath30 , with a small fraction under the barrier @xmath38 . ( b ) the barrier is lowered in the region @xmath39 , so that the wavefunction in the well can slowly leak out to the region @xmath39 . ]

acknowledgements

quantitative comparisons of theoretical models of black hole accretion flows and jets with observations require radiative transfer calculations . the bulk of the radiation is often produced near the black hole event horizon , where relativistic effects of doppler beaming , gravitational redshift , and light bending become important . ray tracing is a convenient method for carrying out fully relativistic radiative transfer calculations . light bending is naturally accounted for by taking the rays to be null geodesics in the kerr metric , and the radiative transfer equation can then be solved along geodesics to calculate observed intensities . this technique has been used to calculate images ( e.g. , * ? ? ? * ) and spectra ( e.g. , * ? ? ? * ) of thin black hole accretion discs @xcite , including state of the art methods to fit spectra in order to infer parameters such as the black hole spin @xcite . ray tracing is also convenient for including general relativistic rotations of the polarisation direction via parallel transport @xcite , and has been applied to the polarised radiative transfer of synchrotron radiation from thick accretion discs , e.g. in order to model the galactic center black hole sgr a * @xcite . with the development of general relativistic mhd simulations of black hole accretion @xcite , ray tracing has become popular as a post - processing step to study their variability properties @xcite and radiative efficiency @xcite , as well as for comparison with observations of sgr a * ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) and m87 ( e.g. , * ? ? * ; * ? ? ? * ) . of particular interest are radiative transfer calculations relevant for current and future event horizon scale interferometric observations of sgr a * and m87 at submillimeter ( the event horizon telescope , * ? ? ? * ) and near - infrared ( the vlti gravity instrument , * ? ? ? * ) wavelengths . fully modeling the observed synchrotron radiation requires polarised radiative transfer . existing codes for this application are either private @xcite or written as post - processors to specific numerical simulations @xcite . other public tools ( e.g. , gyoto and kertap , * ? ? ? * ; * ? ? ? * ) do not include fully polarised radiative transfer . we present a publicly available , fully general relativistic code , grtrans , for polarised radiative transfer via ray tracing in the kerr metric . we describe the methods used for the parallel transport of the polarisation basis into the local frame of the fluid ( [ sec : parallel ] ) and the integration of the polarised radiative transfer equations ( [ sec : integration ] ) using emission , absorption , and rotation coefficients ( [ sec : transf - coeff ] ) calculated based on radiative processes in terms of a background fluid model ( [ sec : fluid - models ] ) . in [ sec : tests - examples ] , we discuss tests used to validate the code , and comparisons of full example problems to those in the literature . we also provide fitting functions for polarised synchrotron emission , absorption , and transfer coefficients ( appendix [ sec : farad - coeff - power ] and [ sec : farad - coeff - power ] ) , and show an example polarised image from a model of the submm emission of sgr a * , to demonstrate how all of the transfer coefficients play important roles in the final polarised image . finally , [ sec : code - struct - perf ] gives a summary of the code convergence and performance properties , and an overview of its organisation . the goal of a ray tracing radiative transfer code is to calculate the observed intensity on locations ( pixels ) of an observer s camera for a given model of emission and absorption . we calculate the boyer - lindquist coordinates of the photon trajectories from the observer towards the black hole ( trace the rays ) corresponding to each pixel , parallel transport the observed polarisation basis into the fluid frame , calculate the local emission and absorption properties at each location , and then solve the radiative transfer equations for the given emission and abosrption along those rays . the observer s camera at inclination @xmath0 and orientation @xmath1 has pixels whose coordinates are described by apparent impact parameters @xmath2 , @xmath3 parallel and perpendicular to the black hole spin axis . the photon trajectories in grtrans are assumed to be geodesics in the kerr metric , in which case their constants of motion are specified for given @xmath2 , @xmath3 @xcite : @xmath4 where @xmath5 , @xmath6 , and @xmath7 are the dimensionless z - component of the angular momentum , carter s constant , and black hole spin parameters . the trajectories for each ray given the constants can then be found by solving the geodesic equation . we do this semi - analytically by reducing the equations of motion to jacobian integrals and jacobi - elliptic functions @xcite as implemented in the code geokerr @xcite . in this method , the independent variable is either the inverse radius @xmath8 or @xmath9 . the former is used by default , since even steps in @xmath10 naturally concentrate resolution towards the black hole , where most of the radiation is produced . in special cases , for example a thin accretion disc in the equatorial plane , the latter method is preferable since then one can solve for the radius where @xmath11 , without needing to integrate the geodesic . in the default case with @xmath10 as the independent variable , the sampling can become poor near radial turning points ( e.g. sections of the orbit at nearly constant radius ) . for this reason , near radial turning points @xmath12 is instead used as the independent variable to fill in the geodesic . the calculation is started at a small , non - zero value of @xmath10 in order to keep the coordinate time and affine parameter finite . the geodesics are tabulated starting at a value of @xmath10 of interest for the problem ( e.g. the outer radial boundary of a numerical simulation ) and are terminated either just outside the event horizon for bound orbits , or once they again reach the outer radius of interest for the calculation . the locations to sample ( @xmath13 ) and number of samples ( @xmath14 ) are code parameters . the assumption made by the code is that the initial intensity is zero at the farthest point sampled along the ray . the observed polarisation is measured with respect to the horizontal and vertical axes defining the camera , while the polarised emission and transfer coefficients are most naturally given relative to a local direction in the emitting fluid ( e.g. the magnetic field direction for synchrotron radiation ) . to relate the two , we first parallel transport the observed polarisation basis along the geodesic , and then transform it to the orthonormal frame comoving with the fluid . the angle between the two bases can then be used to rotate the local coefficients into the observed polarisation basis . parallel transport of a vector describing the polarisation basis @xmath15 perpendicular to the wave - vector @xmath16 is simplified in the kerr metric by the existence of of a complex constant called the walker - penrose constant @xcite , given in boyer - lindquist coordinates with @xmath17 as @xcite : @xmath18\sin{\theta}\right\},\end{aligned}\ ] ] where @xmath19,\\ k^r & = & -\frac{s_r}{\rho^2 } r(r),\\ k^\theta & = & -\frac{s_\theta}{\rho^2 } \sqrt{\frac{m(\theta)}{\sin^2{\theta}}},\\ k^\phi & = & \frac{1}{\rho^2 } \left[-a + \frac{l}{\sin^2{\theta}}+\frac{a}{\delta } ( r^2+a^2-a l)\right],\\ m(\theta ) & = & q^2 + \left(a^2-q^2-l^2\right ) \cos^2{\theta } - a^2 \cos^4{\theta},\\ r(r ) & = & r^2 + ( a^2-q^2-l^2 ) + 2 \left[(a - l)^2+q^2\right ] r^{-1}\\ & -&a^2 q^2 r^{-2},\\ \rho^2 & = & r^2+a^2 \cos^2{\theta},\\ \delta & = & r^2 - 2 r + a^2,\\\end{aligned}\ ] ] is the photon wave vector whose direction is specified by the signs @xmath20 and @xmath21 ( e.g. , * ? ? ? * ) . the real and imaginary parts of the constant , @xmath22 and @xmath23 , provide two constraints on the transported basis vectors , while the orthogonality condition @xmath24 provides a third . since the polarisation basis vectors are already only defined up to a multiple of the wave vector , we can set @xmath25 without any loss of generality , which leaves three linear equations for the three remaining components of @xmath15 : @xmath26 with @xmath27 the components of @xmath15 can then be calculated as : @xmath28,\\ f^\theta&=&-\frac{1}{n}\left[(\gamma_1 k_1 - \delta_1 k_2)(g_{\phi \phi } k^\phi + g_{\phi t } k^t ) - \right.\\&&\left.(\gamma_3 k_1 - \delta_3 k_2 ) g_{r r } k^r \right],\\ f^\phi&=&\frac{1}{n } \left[(\gamma_1 k_1 - \delta_1 k_2 ) g_{\theta \theta } k^\theta -\right.\\&&\left.(\gamma_2 k_1 - \delta_2 k_2)g_{r r } k^r \right],\\ n&=&(\gamma_2\delta_1-\gamma_1\delta_2)g_{\phi \phi } k^\phi - ( \gamma_3\delta_1 - \gamma_1\delta_3)g_{\theta \theta } k^{\theta}\label{partrans2}\nonumber\\&+&(\gamma_2 \delta_1 - \gamma_1 \delta_2 ) g_{\phi t } k^t + ( \gamma_3 \delta_2 - \gamma_2 \delta_2)g_{r r } k^r,\end{aligned}\ ] ] where @xmath29 are the covariant metric components : @xmath30,\\ g_{\phi t } & = & \frac{-2 a r \sin^2{\theta}}{\rho^2},\\ g_{r r } & = & \frac{\rho^2}{\delta},\\ g_{\theta \theta } & = & \rho^2,\\ g_{\phi \phi } & = & \frac{\sigma \sin^2{\theta}}{\rho^2},\\ \sigma & = & ( r^2+a^2)^2 - a^2 \delta \sin^2{\theta}.\end{aligned}\ ] ] the polarisation basis at the camera is defined so that positive stokes q is measured relative to the @xmath31 axis . transforming the linear polarisation basis vectors @xmath31 and @xmath32 at the camera then requires knowledge of @xmath22 and @xmath23 for these vectors . these can be found from the asymptotic form of equation @xcite . they are given by @xmath33 , @xmath34 and @xmath35 , @xmath36 respectively , where @xmath37 @xcite . then equations ( [ partrans]-[partrans2 ] ) allow us to calculate the polarisation basis vectors at any point along the ray . the emission coefficients and the transfer matrix computed in the fluid frame are defined in a basis aligned with a local reference vector . for the case of synchrotron emission , it is convenient to use the local magnetic field direction and so we use @xmath38 as this vector in grtrans without loss of generality . in the case of electron scattering in a thin accretion disc , the polarisation is given relative to the disc normal vector , and so we assign the variable @xmath38 to that vector . before integrating the radiative transfer equations , these coefficients must be transformed to the observed polarisation basis . this transformation requires finding the angle between the transported polarisation basis vectors and the polarisation reference vector @xcite . we transform into the orthonormal frame comoving with the fluid where the four - velocity is @xmath39 . the basis four - vectors of the transformation are @xcite : @xmath40 where the upper ( lower ) indices are lowered ( raised ) with the kerr ( minkowski ) metric and , @xmath41 four - vectors in the coordinate frame are transformed as , @xmath42 the angle @xmath43 between the projected magnetic field and the polarisation basis is given in terms of ordinary dot products of the magnetic field and parallel - transported basis three - vectors ( denoted by hats ) : @xmath44 in this frame , the combined redshift and doppler factor @xmath45 and @xmath46 . the non - relativistic polarised radiative transfer equation can be written in the form , @xmath47 where ( @xmath48 , @xmath49 , @xmath50 , @xmath51 ) are the stokes parameters , @xmath52 are the polarised emissivities , @xmath53 are the absorption coefficients , and @xmath54 are the faraday rotation and conversion coefficients . in the context of synchrotron radiation , the transfer equation can be simplified by aligning the magnetic field with stokes @xmath50 , so that @xmath55 . then @xmath56 , @xmath57 ( @xmath58 , @xmath59 ) correspond to the emission and absorption coefficients for linear ( circular ) polarisation and @xmath60 , @xmath61 are the unpolarised coefficients . the transfer coefficients @xmath62 describe the effects of faraday conversion and rotation respectively . all coefficients are computed in the fluid rest frame , where @xmath63 is the emitted frequency , related to the observed frequency through @xmath64 . then the transfer equation is recast into invariant form : @xmath65 , @xmath66 , and @xmath67 , where @xmath68 , @xmath69 and @xmath70 are the intensity and emissivity vectors and the transfer matrix from equation . finally , we use the angle @xmath43 to rotate the emissivity and absorption matrix in the fluid frame into that of the observer , such that the radiative transfer equation becomes , @xmath71 where @xmath72 is an affine parameter , @xmath73 , @xmath74 , and @xmath75 this rotation transforms the fluid frame polarisation basis to that at infinity , including the parallel transport of the polarisation four - vector along the ray . equation [ genrelpolradtrans ] includes all relativistic effects . the bending of light is accounted for by the calculation of null geodesics @xcite , the gravitational redshifts and doppler shifts due to fluid motions are included in @xmath64 . this method , developed by @xcite , parallel transports the polarisation basis along the ray and into the fluid frame . this is similar to the approach of @xcite , who transported a local polarisation vector @xmath15 from the fluid to the observer . @xcite derived a general formalism for covariant polarised radiative transfer , and mathematically showed the equivalence of the approach used here and alternative methods used by @xcite and @xcite . we show tests and example problems comparing results from these methods in [ sec : test - problems ] . the transfer coefficients in equation depend in general on the physical properties of the radiating particles . here we focus specifically on the case of synchrotron radiation appropriate for studying accretion flows at the lowest observed luminosities ( e.g. , sgr a * ) . adding different emissivities such as bremsstrahlung to grtrans would require a straightforward modification of the code . in addition , the form of the transfer coefficients depends on the underlying electron distribution function . the appropriate forms for thermal and power law distributions are implemented in the code . more general distribution functions can be built by combining these components ( e.g. a thermal distribution with a power law `` tail '' or a superposition of thermal distributions , mao et al . in prep . ) . in these special cases , the integral over the distribution function can be analytically approximated to high accuracy in the ultra - relativistic synchrotron limit ( e.g. , * ? ? ? the full forms for the transfer coefficients as used in grtrans and some of their derivations are given in appendix [ cha : polar - synchr - emiss ] . in addition to synchrotron coefficients , for test problems with optically thick accretion discs the code uses ( color - corrected ) blackbody intensity functions for the disc surface brightness . the calculation of transfer coefficients for a particular emission model requires knowledge of the fluid state variables of spacetime coordinates . depending on the model used , this can include the electron density , magnetic field strength and orientation , and the internal energy density in electrons . in grtrans we implement several fluid models from the literature . they are described briefly below , and used as code examples and tests in [ sec : test - problems ] . the relativistic version @xcite of the standard thin disc solution @xcite for axisymmetric , steady accretion in the equatorial plane is implemented and intended for use with a model for the emergent intensity from the disc ( e.g. , a blackbody ) . in this case , the net polarisation is taken to follow the result of electron scattering in a semi - infinite atmosphere @xcite specified relative to the disc normal vector ( [ sec : parall - transp - tests ] ) . we have also implemented a numerical version of the thin accretion disc problem which inputs a temperature distribution @xmath76 in the equatorial plane . one example use of this is for calculating spectra of inhomogeneous ( or patchy " ) accretion discs @xcite , as used for the polarisation calculations described in @xcite . a solution of the general relativistic fluid equations for spherically symmetric inflow in the schwarzschild metric following @xcite is implemented . the dominant emission in this case comes from synchrotron radiation @xcite . see @xcite for details . for polarised emission , we take the magnetic field to be purely radial . this is done to check that the resulting linear polarisation sums to zero ( since for a camera centered on the black hole there is no preferred direction ) , and the residual is used as an estimate of the minimum systematic uncertainty in the fractional linear polarisation ( @xmath77 ) . @xcite presented a semi - analytic jet model based on stream functions found in force - free simulations . we have implemented this solution numerically on a grid of ( @xmath78,@xmath79 ) in boyer - lindquist coordinates . to generate our numerical solutions , we solve for the magnetic field and velocity structure analytically using their equations 5 - 13 . to get the particle density , we tabulate their function @xmath80 numerically using a separate grid of points with roughly constant @xmath81 and varying @xmath82 . this function can then be used to calculate the particle density ( their equation 13 ) . the fluid variable solutions from our method appear identical to what is shown in their figure 4 . we also use another numerical solution , from the public version of the axisymmetric general relativistic mhd code harm @xcite . starting from a gas torus in hydrostatic equlibrium threaded with a weak magnetic field , the code evolves the equations of ideal mhd in the kerr spacetime . the magnetorotational instability @xcite drives turbulence in the torus and the resulting stresses transport angular momentum outwards , leading to accretion onto the central black hole . snapshots from these simulations have been used as models of sgr a * ( e.g. , * ? ? ? * ) . the images used here as examples are from a single snapshot of a simulation with black hole spin @xmath83 at @xmath84 used for comparison with 3d simulations in @xcite . grtrans supports fully time - dependent calculations using a series of such simulation snapshots to e.g. calculate accretion flow movies rather than images . it would also be straightforward to adapt the code to work with updated harm versions , for example with 3d data or non - ideal mhd . the fluid variables in these simulations are saved in modified kerr - schild coordinates and with arbitrary units which assume @xmath17 . calculating radiation from these data in grtrans requires converting to boyer - lindquist coordinates and to cgs units . the coordinate conversion is done analytically in two steps : from modified to standard kerr - schild coordinates @xcite and then from kerr - schild to boyer - lindquist coordinates ( e.g. , * ? ? ? scaling to cgs units is done by i ) fixing the black hole mass , which sets the length- and time - scales , and ii ) choosing an average accretion rate ( or equivalently mass of the initial torus ) . this procedure is discussed in more detail elsewhere @xcite . once the unit and coordinate conversions are done , we calculate fluid variables at tabulated geodesic coordinates . for all numerical models , we linearly interpolate from the set of nearest neighbors on the grid for the numerical model . the way this is implemented in the code assumes that the grid is uniformly spaced in some coordinates , and the fluid model must include the transformation from those coordinates to boyer - lindquist . there are several other models implemented in the code in some form , but which have not been tested . it is straightforward to add new fluid models to the code , e.g. by using existing ones as templates . from the previous steps , we have transfer coefficients specified at tabulated points along a geodesic which are transformed to relativistic invariant form and aligned with the observed stokes parameters of the distant observer , accounting for parallel transport along each ray . the final step is to solve the polarised radiative transfer ( equation [ genrelpolradtrans ] ) along the geodesic . in grtrans , this is done as a separate step following the calculation of the coordinates of the geodesic . while the ray tracing proceeds backwards from the camera towards the black hole , the integration proceeds outwards . this is done so that we may safely set the initial intensity to zero at some point either where the optical depth is large , or where the geodesic has left the emitting volume . here we describe one numerical integration method and two quadrature methods that are implemented in grtrans for integrating the equations . these methods can be used for relativistic or non - relativistic problems . for consistency with previous literature , we write the non - relativistic versions of the intensity , absorption matrix , emissivity , and step size along the ray at index @xmath85 as @xmath86 , @xmath87 , @xmath88 and @xmath89 in what follows . in grtrans , the relativistic invariants @xmath90 , @xmath91 , @xmath92 , and @xmath93 take the place of these quantities . the most straightforward method is numerical integration of the radiative transfer equation . the radiative transfer equations can be stiff : the required step size for a converged solution decreases sharply once @xmath94 , where @xmath95 is the optical depth associated with any transfer coefficient . in order to get a robust solution , we use the odepack routine lsoda @xcite to advance the stokes intensities between each step tabulated on the geodesic . this algorithm adaptively switches between a predictor - corrector ( adams ) method for non - stiff systems , and a bdf method for stiff systems . we find it necessary to restrict the maximum step size allowed in @xmath72 , since otherwise a large step can miss the region of interest altogether . regions of large optical depth often contribute negligibly to the observed intensity but require a small step size , and so we terminate the integration at a maximum optical depth , @xmath96 by default . there are further free parameters in lsoda related to the error tolerance . the locations sampled by lsoda do not correspond exactly to the points tabulated along the ray . we linearly interpolate the transfer coefficients between tabulated points , even though they are highly non - linear functions of position along the geodesic . the fluid variables vary more smoothly along the ray , and it would be straightforward but more computationally expensive to instead re - interpolate the fluid variables to the points used by lsoda and then calculate new transfer coefficients , as was done in the previous version of the code @xcite . given the results from comparing to quadrature integration methods and analytic solutions described below , and from the convergence properties with increasing the number of points along each ray , we find the current approximation adequate for obtaining accurate solutions . the polarised radiative transfer equations are linear and ordinary , and so admit a formal solution analagous to that of the unpolarised case @xcite . the solution amounts to finding the matrix operator @xmath97 , defined by @xcite , @xmath98 which determines how the intensity is propagated over some part of the ray in the absence of emission . in the unpolarised case this is a scalar , @xmath99}$ ] . if the absorption matrix @xmath70 is a constant over the ray , then similarly , @xmath100}.\ ] ] in terms of @xmath101 , the intensity can be written in terms of an initial value @xmath102 : @xmath103 @xcite found a closed form solution for @xmath101 ( their equation 10 , reproduced in appendix [ deglinnocentio ] ) . this solution is valid for regions where the transfer matrix @xmath70 is constant , but not for our situation of interest where they vary arbitrarily along a ray . in order to use this expression , we assume that the coefficients are constant in between the tabulated locations along a geodesic starting from @xmath104 at the farthest point of interest along the ray ( at the black hole or where the ray leaves the far end of the emitting region ) and integrating towards @xmath105 ( the `` surface '' ) , and write the solution of equation separately for the interval between neighboring points with indices @xmath106 and @xmath85 with positions @xmath107 and @xmath108 : @xmath109 where @xmath110 . this formula is used recursively going outwards from @xmath111 to @xmath112 to find the intensity everywhere from the initial condition @xmath113 . the final integration method implemented in grtrans is the diagonal element lambda operator method ( delo , * ? ? ? * ) , which comes from writing the transfer equations in terms of the unpolarised optical depth , @xmath114 , and the modified absorption matrix @xmath115 and source function @xmath116 : @xmath117 where @xmath118 . this equation has a formal solution between neighbouring points @xmath119 , @xmath120 of @xmath121 \mathcal{s } d\tau,\ ] ] where @xmath122 and @xmath123 . the delo method makes a linear approximation for the modified source function , @xmath124/\delta_k,\ ] ] so that equation can be integrated analytically between grid points , giving : @xmath125 where @xmath126,\\ \mathcal{q}_k & = & \mathcal{m}_k ( e_k \mathbf{1}-g_k k'_{k+1}),\\ \mathcal{m}_k & = & [ \mathbf{1}+(f_k - g_k)k'_k]^{-1},\\ f_k & = & 1-e_k,\\ g_k & = & \left[1-\left(1+\delta_k\right)e_k\right]/\delta_k,\end{aligned}\ ] ] the difficulty with this method is that @xmath127 is used as the independent variable . for our problems of interest @xmath95 can be nearly constant between grid points over which the fluid quantities and emissivity change significantly , which causes the above solution to fail . in the limit of small @xmath128 , we instead expand the above quantities up to @xmath129 , leading to the following forms : @xmath130,\\ \mathcal{q}_k & = & \mathcal{m}_k \left[\mathbf{1}\left(1-\frac{\delta s_k \alpha_{i , k}}{2}+\frac{\delta s_k^2 \alpha_{i , k+1}^2}{6}\right)\right.\\&-&\left.\left(\frac{\delta s_k}{2}-\frac{\delta s_k^2}{3}\right)k_{k+1}\right],\\ \mathcal{m}_k & = & \left[\left(1-\frac{\delta_k}{2}+\frac{\delta_k^2}{6}\right)\mathbf{1}\right]^{-1}.\end{aligned}\ ] ] this version of the equations uses @xmath131 as the independent variable , and is used by default when @xmath132 . since the number of steps taken by grtrans is usually @xmath133 , this form of the equations is used unless the optical depth is very large . in grtrans , all integration methods proceed outwards from an initial point back towards the camera . from the recursive forms of the delo and formal solution methods , we see that it would also be possible to integrate the polarised radiative transfer equations backwards by summing the so - called contribution vectors from each point to the final intensity on the camera , @xmath134 : @xmath135 where @xmath136 j_i\ ] ] for the formal solution method and @xmath137 j_i\ ] ] for the delo method @xcite . the equivalent contribution vectors in the unpolarised case are given as @xmath138 j_i = e^{-\tau_i } j_i,\ ] ] where @xmath139 is the optical depth from the surface to the depth at index @xmath140 . tracing backwards from the camera , at each step at index @xmath140 one can calculate @xmath141 using the solution for @xmath142 , @xmath143 , and @xmath144 . solving the polarised radiative transfer equations in this way would be useful in implementations where the geodesic equations and radiative transfer equations are solved simultaneously , e.g. as is done for the unpolarised case in the public code gyoto @xcite . in this method , one can then safely terminate the integration early if the product term in @xmath141 becomes sufficiently small ( e.g. , the optical depth becomes large ) . the three methods give consistent answers , usually to high accuracy and with similar performance . the main drawback of our quadrature implementations is the lack of an adaptive step size , so that many steps ( @xmath145 ) are required in order to get a converged result . in most example problems in the following section , the numerical integrator is used as it is the most robust choice . the other methods are primarily used for comparison and testing , although they are faster at a fixed number of points @xmath14 and so with some optimisation might prove to be significantly faster than numerical integration . here we describe tests of the different aspects of grtrans ( unit tests ) , as well as full example problems which are compared with results from the literature . we do not provide tests of the geokerr code for calculating null geodesics in the kerr metric , which are described in @xcite . [ cols="^,^ " , ] as well as the emission and absorption coefficients calculated above ( appendix [ cha : polar - synchr - emiss ] ) , the coefficients @xmath146 and @xmath147 affect the generation and transfer of polarisation in a magnetised plasma . we use approximate expressions from the literature for these coefficients which as above are modified to i ) be fast to evaluate and ii ) have the correct asymptotic limits . in the case of a power law distribution , we use the expressions from @xcite appendix c , written in our notation : @xmath148,\\ \rho_v & = & 2 \frac{p+2}{p+1 } \rho_\perp\left(\frac{\nu_b}{\nu}\right)^2 \gamma_{\rm min}^{-(p+1)}\ln \gamma_{\rm min } \cot{\theta_b},\\ \rho_\perp&=&\frac{n e^2}{m c \nu_b } ( p-1 ) \left[\gamma_{\rm min}^{1-p}-\gamma_{\rm max}^{1-p}\right]^{-1}.\end{aligned}\ ] ] more accurate expressions @xcite require integration over the distribution function and for this reason are slow to evaluate . these approximate forms are relatively accurate for @xmath149 ( left panel of fig . 6 in * ) . in the example semi - analytic jet problem above , faraday rotation and conversion are negligible @xcite . nonetheless , it should be possible to find accurate fitting functions for these coefficients , which would be consistent with our approach for the other coefficients . faraday coefficients for thermal distributions of electrons have been calculated in limits of either high frequency @xmath150 ( e.g. , * ? ? ? * ) , at high temperatures @xmath151 , or both . in particular , @xcite provided approximate fitting functions for @xmath146 and @xmath147 over a wide temperature range @xmath152 with high accuracy for @xmath153 ( their equations 25 , 26 , 33 , but in our notation ) : @xmath154\\ \rho_v & = & \frac{2 n e^2 \nu_b}{m c \nu^2 } \frac{k_0(\theta_e^{-1})}{k_2 ( \theta_e^{-1 } ) } \cos{\theta_b } g(x),\end{aligned}\ ] ] where @xmath155 and their parameter @xmath156 is a function of @xmath157 . in the high - frequency limit @xmath150 , both functions asymptotically reach unity . the cosine term in @xmath158 is used to fit the sign change in @xmath147 near @xmath159 . while the high frequency and high temperature limit is most relevant for our applications of interest , e.g. modeling the submm emission from sgr a * , ideally we would have expressions that are correct in both asymptotic limits . for this reason we modify the expressions in @xcite , by comparing them with the expressions for the high temperature synchrotron limit given in @xcite ( their equations 3 - 4 , in our notation ) : @xmath160 where @xmath161 + 1/2\int_q^\infty dx \left(\pi gi(x ) - 1/x\right)\right ] e^{-y},\end{aligned}\ ] ] and where @xmath162 and @xmath163 is defined in terms of airy functions @xcite . in the high - frequency , high temperature limit where @xmath164 , @xmath151 , and where we can replace the bessel functions by their asymptotic limits and @xmath165 , @xmath166 , these expressions agree with the above results from @xcite . from numerically integrating @xmath167 and @xmath168 , we also find good agreement between the two sets of coefficients over their reported ranges of validity . we adapt the fitting function @xmath158 in @xcite to use the asymptotic limit of @xmath147 at small @xmath169 , where @xmath170 : @xmath171 1/2 \left[1+\tanh{\left(10 \ln { x/120}\right)}\right].\end{aligned}\ ] ] the added term imposes the correct asymptotic limit at large @xmath156 ( small @xmath169 ) and for this reason maintains good accuracy over all @xmath169 . for @xmath146 , the term @xmath172 from @xcite separates into a sum of terms which depend on temperature and @xmath173 , and those which only depend on @xmath169 . this suggests that it would be better to use the factor @xmath174 in @xcite as a difference from the high - frequency limit rather than a multiplication : @xmath175 where we define our correction factor @xmath176 in a similar spirit to that of @xcite , but modified for higher accuracy : @xmath177 figure [ fig : rotcoef ] compares numerical integration of equation @xcite with the fitting functions from equation @xcite and our modified forms . our approximate versions are fast to compute , while maintaining accuracy over all @xmath169 . we have compared polarised spectra and maps from harm models of sgr a * computed using the prescriptions from @xcite , @xcite , and the high temperature , high frequency limit . as expected , all expressions are in excellent agreement when @xmath164 , in this case for @xmath178 hz . below that , our modified expressions and those from @xcite are in good agreement , while the degree of circular polarisation can differ between our results and the high - frequency limit . as described in the main text , models of sgr a * in the submm have self - absorption optical depth @xmath179 , @xmath180 , and @xmath181 , so that the faraday optical depths can be very large and have an important effect on the resulting polarisation maps and spectra ( e.g. figure [ harmpol ] ) . since this result is for the high - frequency , relativistic limit , it does not depend on the fitting function used . this appendix provides the analytic solutions to the polarised radiative transfer equations used for testing different integration methods for grtrans in [ sec : integration - tests ] . in all cases , the boundary condition used is that the initial intensity is zero for each stokes parameter . the first case considered is pure emission and absorption in stokes @xmath48 and @xmath49 , in which case equation becomes : @xmath182 whose solution is , @xmath183 + ( j_i \alpha_q - j_q \alpha_i ) \frac{e^{-\alpha s}}{2 } ( 1-e^{2 \alpha_q s})\right\},\\ q(s ) & = & \frac{1}{\alpha ( \alpha_i - \alpha_q ) } \left\{(j_q \alpha_i - j_i \alpha_q)\left[1-\frac{e^{-\alpha s}}{2}\left(1+e^{2\alpha_q s}\right)\right ] + ( j_q \alpha_q - j_i \alpha_i ) \frac{e^{-\alpha s}}{2 } \left(1 - e^{2 \alpha_q s } \right)\right\},\end{aligned}\ ] ] where @xmath184 . when @xmath185 , the second group of terms in each equation vanishes while the first reduces to the usual formal solution of the radiative transfer equation , e.g. when stokes @xmath48 and @xmath49 are not coupled . examples of the solution are shown in figure [ testabs ] . the second case of interest is pure polarised emission in stokes @xmath186 along with faraday rotation and conversion ( @xmath62 ) . here the polarised radiative transfer equation is , @xmath187 where we have set @xmath188 as is commonly chosen for the stokes basis for synchrotron radiation . from this equation it is apparent that @xmath146 is responsible for changing the linear polarisation direction ( mixing stokes @xmath49 and @xmath50 , faraday rotation ) while @xmath147 converts between linear and circular polarisation ( mixing stokes @xmath50 and @xmath51 , faraday conversion ) . the solution is , @xmath189 where @xmath190 . these solutions for a sample case are plotted in figure [ testfar ] . in the case of only faraday rotation or conversion ( @xmath191 for stokes @xmath49 or @xmath192 for stokes @xmath51 ) , the solution is purely oscillatory with the maximum linearly polarised intensity restricted to be @xmath193 independent of the total intensity or path length , despite the fact that there is no absorption . since in this optically thin limit the total intensity grows as @xmath194 , the fractional polarisation decreases as @xmath195 . when both faraday rotation and conversion are present , the stokes @xmath49 and @xmath51 acquire terms which linearly increase with @xmath131 , while the oscillatory terms still have maximum values that are independent of @xmath131 . this means that in the limit of large faraday optical depth ( large @xmath131 ) , the fractional polarisation approaches a constant value @xmath196 instead of decreasing as @xmath197 in the pure faraday rotation or conversion case . since for cases of interest @xmath198 , circular polarisation becomes dominant over linear polarisation in the limit of large faraday optical depth . this is a different limit than an initial polarised intensity travel through a magnetised medium where faraday rotation occurs . in that case , the intensity also oscillates between stokes @xmath49 and @xmath50 , but with a constant polarised intensity . the fractional polarisation only decreases when the faraday rotation is instead occurring in the region where the polarised emission is being produced . @xcite found a closed form solution for the matrix operator @xmath97 , defined by which describes the transfer of the stokes parameters from position @xmath131 to @xmath200 in the absence of emission , which is valid under limited conditions including when the absorption matrix @xmath70 is constant over the interval . we reproduce the solution here in our notation : @xmath201 \mathbf{m_1 } / 2 - \sin{\left(\lambda_2 \delta s\right ) } \mathbf{m_2 } - \sinh{\left(\lambda_1 \delta s\right ) } \mathbf{m_3}+\left[\cosh{\left(\lambda_1 \delta s\right)}-\cos{\left(\lambda_2 \delta s\right)}\right]\mathbf{m_4 } / 2\right\},\ ] ] @xmath203^{1/2},\\ \lambda_{1,2 } & = & \left\{\left[\left(\alpha^2-\rho^2\right)^2/4 + \left(\mathbf{\alpha } \cdot \mathbf{\rho}\right)^2\right]^{1/2 } \pm \left(\alpha^2 - \rho^2\right)/2\right\}^{1/2},\\ \sigma & = & \rm sign \left(\mathbf{\alpha } \cdot \mathbf{\rho}\right),\\ \mathbf{\alpha } \cdot \mathbf{\rho } & = & \alpha_q \rho_q + \alpha_u \rho_u + \alpha_v \rho_v,\\ \rho^2 & = & \rho_q^2 + \rho_u^2+\rho_v^2,\\ \alpha^2 & = & \alpha_q^2 + \alpha_u^2 + \alpha_v^2.\end{aligned}\ ] ]

ray tracing radiative transfer is a powerful method for comparing theoretical models of black hole accretion flows and jets with observations . we present a public code , grtrans , for carrying out such calculations in the kerr metric , including the full treatment of polarised radiative transfer and parallel transport along geodesics . the code is written in fortran 90 and efficiently parallelises with openmp , and the full code and several components have python interfaces . we describe several tests which are used for verifiying the code , and we compare the results for polarised thin accretion disc and semi - analytic jet problems with those from the literature as examples of its use . along the way , we provide accurate fitting functions for polarised synchrotron emission and transfer coefficients from thermal and power law distribution functions , and compare results from numerical integration and quadrature solutions of the polarised radiative transfer equations . we also show that all transfer coefficients can play an important role in predicted images and polarisation maps of the galactic center black hole , sgr a * , at submillimetre wavelengths . [ firstpage ] radiative transfer accretion , accretion discs black hole physics galaxy : centre galaxies : jets relativistic processes

introduction methods tests and examples analytic solutions to the polarised radiative transfer equations closed form expression for @xmath97

recently a significant attention is paid to the studying of soliton equations under external perturbations . for kdv equation , in particular , it was shown , that simple sine driven force together with dissipation can lead to a chaotic dynamics , detected by fourier spectrum , positive liapunov exponents and phase plane analysis @xcite . in this context the reduction of a perturbed kdv equation in the form of stationary travelling waves was considered . the justification of the application of low - dimensional chaos theory to such an infinite - dimensional system as kdv equation is in a numerical investigation , which proves that information dimension , calculated by kaplan - yorke formula , and correlation dimension within grassberger - procaccia method , are both between 2 and 3 for steady waves @xcite . on the other hand , perturbed kdv equation can be considered directly whithin numerical approach , or by perturbation technique @xcite . besides this series of work it is worth to mention here the cases , when kdv equation is influenced by dissipation and instability , wihtout explicit spatial or temporal dependence @xcite . in such a form this equation describes current - driven ion - acoustic instability in a collision - dominated plasma . for the strongly dissipative case overall evolution demonstrates irregular behaviour , therefore leading to nonstationary and irregular soliton interactions . in this work we add hamiltonian deterministic perturbation to the kdv equation and study physical consequences of this . the reduction in the form of travelling waves is made in order to turn the system into a second - order ordinary differential equation . the existence of solutions to such an equation was studied in @xcite , while the construction of periodic orbits can be found in @xcite for nonresonant and primary resonant cases , and in @xcite for secondary resonances . we ll , however , perform the analysis of chaotic properties in this situation . it is known , that for any hamiltonian perturbation , which makes the system near - integrable , a stochastic layer around a separatrix appears . this layer is bounded by unbroken kam - surfaces , and whithin it the system evolves with mixing . for dissipative perturbations , however , these kam - surfaces are broken and special methods like melnikov one should be used to determine the conditions for chaos to appear . so , we calculate the width of a stochastic layer , which contains long - period waves and solitons . it is done on the basis of chirikov criterion for the overlap of resonances , which is widely used for investigation of chaotic properties in hamiltonian systems ( references on original papers in this field can be found in @xcite ) . results we obtain are as follows : solitons and nonlinear waves prove to be chaotic in the meaning that in a small distance from the peak of solitons and long - period waves there must be a region of chaotic dynamics , where these waves acquire small irregular deviations . this conclusion is confirmed in experiments with ion - acoustic waves in plasma , where a weak splash before soliton peak was registered @xcite . the outline of this paper is as follows . in next section kdv equation is reduced to ode and the unperturbed solutions of the latter are considered . in section 3 we introduce canonically - conjugated variables action - angle and obtain general expressions for the criterion of stochasticity . after that in section 4 this criterion is applied to kdv waves directly . section 5 contains the conclusions and summary . let s consider the perturbed kdv equation , taken in the following form : @xmath0 where @xmath1 and @xmath2 denote , respectively , a one - dimensional space coordinate and time ; @xmath3 is supposed to be differentiable with respect to @xmath1 and @xmath2 sufficient number of times ; @xmath4 is a small external perturbation . as far as we restrict ourselves on the case of hamiltonian perturbation , it means that within the class of functions @xmath5 we ll consider only those , which will not contain derivatives after integration over @xmath6 . it s easy to check that the general form of such perturbations can be represented as : @xmath7 with the condition @xmath8_{\xi}=f_3(u,\xi)_u$ ] . searching for a solution of equation ( [ eq1 ] ) in the form of a nonlinear stationary wave @xmath9 , we obtain the following ode after one integration over @xmath10 ( prime means differentiation with respect to @xmath11 ): @xmath12 where @xmath13 denotes the primitive of the function @xmath5 after stationary wave substitution . integration constant is absent here because it can be turned into zero by an appropriate choice of variables . for convenience let s rewrite equation ( [ eq2 ] ) as the following dynamical system : @xmath14 equation ( [ eq3 ] ) without perturbation ( @xmath15 ) is analogous to the equation of motion : @xmath16 of a point with mass @xmath17 in a potential field @xmath18 full mechanical energy of such a nonlinear oscillator is equal to @xmath19 where @xmath20 is the momentum of the point . phase plane of this oscillator together with the plot of the potential energy is presented on the fig if @xmath21 , what corresponds to the bottom of the potential pit , then the phase curve shrinks to a point . in the case @xmath22 we have closed phase curves representing cnoidal waves of the form : @xmath23 where @xmath24 denotes the jacobi elliptic cosine function of modulus @xmath25 . substitution of ( [ eq7 ] ) in ( [ eq4 ] ) gives : @xmath26/{\sqrt{m^2-m+1 } } \\ k=\frac{1}{2}\sqrt{v/\beta}(m^2-m+1)^{-\frac{1}{4 } } \end{array } \right.\ ] ] here @xmath27 is the complete elliptic integral of the first kind , @xmath28 is the period of the corresponding solution , and @xmath25 is the modulus of elliptic functions . under @xmath29 @xmath30 , what corresponds to a separatrix on a phase plane , solution ( [ eq9 ] ) yields the homoclinic orbit : @xmath31 where @xmath32 @xcite . let s transform variables @xmath33 to the canonical pair action - angle ( @xmath34 - action , @xmath35 - angle ) using standard formulae : @xmath36 where @xmath37 is the reduced action . the hamiltonian of a perturbed motion we can write in the form : @xmath38 where @xmath39 is the hamiltonian of the nonperturbed system from ( [ eq6 ] ) , and @xmath40 corresponds to the perturbation term in equation ( [ eq3 ] ) . assuming perturbation @xmath5 to be periodical on time with frequency @xmath41 , and taking into consideration the fact that nonperturbed motion is integrable , we may expand the perturbation into the following fourier series : @xmath42 , \\ v_{kl}=v_{-k ,- l}^{*}. \end{array } \right.\ ] ] the perturbed equations of motion in variables @xmath43 take now the form : @xmath44 } \\ { \dot \theta = \omega ( i)+\frac{1}{2 } i \sum\limits_{k , l}}\frac{dv_{kl } } { di}\exp [ i(k\theta -l\nu t ) ] \end{array } \right.\ ] ] where @xmath45 is the frequency of the nonperturbed motion . let s consider the motion in the vicinity of one particular resonance : @xmath46 the width of this resonance on frequency is @xmath47 , where @xmath48 and @xmath49 . the distance between adjacent resonances is equal to @xmath50 where @xmath51 ; @xmath52 . as a condition of chaotic motion , which means an appearance of mixing , we ll use the chirikov criterion for the overlap of resonances @xcite : @xmath53 where @xmath54 is the minimal possible value of @xmath55 from ( [ eq17 ] ) . for the cases @xmath56 , @xmath57 and @xmath58 calculations similar to @xcite give respectively the following conditions for the border of stochasticity : @xmath59 to apply these results to the system , governed by kdv equation , one must first evaluate the exact expressions for @xmath60 and @xmath61 , which appear in the conditions ( [ eq31 ] ) . the frequency @xmath60 is equal to @xmath62 for convenience let s rewrite it as : @xmath63 where we introduced @xmath64 by the correlation @xmath65 , and @xmath66 . to evaluate @xmath61 , one should represent it as @xmath67 tedious calculations give finally @xmath68 @xmath69.\ ] ] substituting ( [ eq36 ] ) and ( [ eq37 ] ) in ( [ eq31 ] ) one makes sure that with approach to separatrix @xmath70 @xmath71 under arbitrary small @xmath72 . this means that for any external perturbation , however small it is , chirikov criterion is executed starting with some energies , and this leads to the formation of a corresponding stochastic layer . the transition from regular to chaotic motioncan be found approximately from the condition @xmath73 . this border is defined on energy by the inequality : @xmath74 here the value of @xmath75 is found from : @xmath76 in which @xmath77 and @xmath78 , @xmath79 , @xmath80 for three cases in ( [ eq31 ] ) . certainly , there is only half - width of stochastic layer , but the phase curves , laying out separatrix do nt constitute an interest , being physically meaningless ( they are unbounded on infinity ) . we have calculated the width of the stochastic layer around a separatrix , corresponding to a soliton solution of korteweg - de vries equation under hamiltonian perturbations . as far as the motion in this stochastic layer is chaotic ( in the meaning of mixing ) , so nonlinear wave solutions , which correspond to the phase curves within this stochastic layer will also posess certain chaotic properties . let s now consider the question about the spatial region , where this chaotic behaviour can be registered . the time length , after which stochasticity in a nonlinear oscillator can be found is defined by @xmath81 , where @xmath82 is a time of the decay of correlations . in the work @xcite the following estimation of @xmath82 for a nonlinear oscillator is obtained : @xmath83 where @xmath84 is a coefficient for the overlap of resonances , obtained in the previous section . as far as in our consideration time @xmath2 for a nonlinear oscillator corresponds to the expression ( @xmath85 ) for nonlinear waves , so the region , where chaotic regimes for these waves can be detected , is : @xmath86 therefore , we can conclude that this region represents wave formation , which outstrips nonlinear wave or soliton , and propagates in the same direction with the same velocity . under @xmath87 we obtain : @xmath88 as far as with approach to a separatrix @xmath70 @xmath71 , so for solitons the minimal distance on which chaotic behaviour should appear @xmath89 , and this effect will be realized in a close vicinity of a soliton peak . the chaos , which we are searching for , will be manifested in the irregular small deflection from a smooth initial soliton profile . so , as a main result of this work , it can be inferred that long - period nonlinear waves and especially solitons of korteweg - de vries equation obtain chaotic properties in the presence of periodic hamiltonian external perturbations . exactly this result was observed in an experiment with ion - acoustic and langmuir waves in a nonmagnetized plasma @xcite . there was registered a faint splash directly before the soliton . from the concepts developed in this paper this phenomenon can be explained in the following way . together with soliton wave generator also produces a group of waves of small amplitude and almost zero frequencies . these waves are produced constantly with generator working , and therefore can be considered as a small deterministic periodic external perturbation . stochastic destruction of soliton due to the influence of these waves , as it was described above , must realize itself just before the soliton peak , exactly what was registered in the experiment . the author would like to thank prof . m. bestehorn for helpful discussions . he also acknowledges the referee for suggestions , which helped to make the paper clearer . 20 r .grimshaw , x. tian : _ proc . r. soc . london a _ * 445 * , 1 - 21 ( 1994 ) . q. cao , k. djideli , w.g . price , e.h . twizell : _ physica d _ * 125 * , 201 - 221 ( 1999 ) . x. tian , r. grimshaw : _ int . . chaos _ * 5 * , 1221 - 1233 ( 1995 ) . r. grimshaw , x. tian : _ physica d _ * 77 * , 405 - 433 ( 1994 ) . t. kawahara , s. toh : _ on some properties of solutions to a nonlinear evolution equation including long - wavelength instability _ , in nonlinear wave motion 43 , ed . a. jeffrey , longman , new york 1989 . t. ogawa : _ hiroshima math . j. _ * 24 * , 401 - 422 ( 1994 ) . a. j. lichtenberg , m. a. liberman : _ regular and chaotic dynamics _ , springer - verlag , new york 1992 . h. ikezi : _ experiments on solitons in plasmas _ , in solitons in action , k. longren and a.c . scott eds . , academic press , new york 1978 . v. i. karpman : _ nonlinear waves in dispersive media _ , pergamon press , new york 1973 . r. z. sagdeev , d. a. usikov , g. m. zaslavsky : _ nonlinear physics : from the pendulum to turbulence and chaos _ , harwood academic press , new york 1988 .

this paper considers properties of nonlinear waves and solitons of korteweg - de vries equation in the presence of external perturbation . for time - periodic hamiltonian perturbation the width of the stochastic layer is calculated . the conclusions about chaotic behaviour in long - period waves and solitons are inferred . obtained theoretical results find experimental confirmation in experiments with the propagation of ion - acoustic waves in plasma . = -1.5 cm = 17.5 cm = -0.5 cm = -0.5 cm * chaotic behaviour of nonlinear waves and solitons of perturbed + korteweg - de vries equation * konstantin b. blyuss _ postfach 101344 , 03013 cottbus , germany _

introduction stationary waves a criterion for chaotic motion chaos of korteweg-de vries waves summary and conclusions *acknowledgements*

the conjecture for the existence of dark matter dates back to zwicky in 1933 , who failed to explain the dynamics of coma cluster by the virial theorem through the distribution of visible matter @xcite . to interpret the dynamics of the structure , the concept of missing mass or dark matter entered the astrophysical studies since then . in addition to the cosmological scales , in the smaller galactic scales , the flat rotation curve of the spiral galaxies requires the existence of the dark matter halo @xcite . on the other hand studying the dynamics of the universe as a whole reveals that the universe is dominated by the dark matter and dark energy @xcite . this hypothetical matter neither emits light nor interacts with the ordinary matter and only shows its presence through its gravitational interaction . the advantage of the cold dark matter ( @xmath0 ) model is that , it can successfully explain the rotation curve of the spiral galaxies and lensing by galaxies and cluster of galaxies . within the framework of general relativity ( gr ) it also provides a reasonable description for the hierarchy in the structure formation . although currently @xmath0 and @xmath1 models are remarkably successful in large scales @xcite , they can not explain tully - fisher and freeman laws @xcite . on the other hand , the high - resolution n - body simulations are still in contradiction with the observations on sub - galactic scales where they predict orders of magnitude more substructures than what is observed @xcite . they also provide incompatible spatial distribution of the sub - halos @xcite . the other alternative to the dark matter is the modification to the gravity law by means of taking a generic form of action for the gravity rather than that of the einstein - hilbert action . this approach has been introduced to be an alternative model to the dark energy @xcite . @xmath2 gravity also is used to interpret the rotation curve of the spiral galaxies @xcite . there is also recent efforts on modifying the gravity law by using a simple kinetic lagrangian whether the pressure can bend space - time sufficiently to replace the roles of dark energy , cold dark matter and heavy neutrinos in explaining anomalous accelerations at all scales @xcite . halle et al ( 2008 ) also proposed a generalized lagrangian with non - uniform cosmological constant for the vacuum field within the framework of the einstein gravity . finally the third approach , which we are concerned with in this work is the modification of the conventional newtonian law is so - called modified newtonian dynamics ( mond ) . the dynamics of a structure in mond under the gravitational field is given by @xcite : @xmath3 where @xmath4 is the newtonian gravitational acceleration , @xmath5 is a fundamental acceleration parameter and @xmath6 is a function for the transition from the newtonian to the mondian regime ( e.g. @xmath7 ) . the dynamics of the structure for @xmath8 deviates from the newtonian law by @xmath6 and for @xmath9 we recover the newtonian dynamics ( i.e. @xmath10 ) . on the other hand for @xmath11 , so - called deep mond regime , @xmath12 and the effective acceleration is given by @xmath13 . due to confusion in the definition of the dynamical concepts in mond , this model can be interpreted as a modification to the gravity law instead of dynamics . bekenestein and milgrom ( 1984 ) used a non - conventional newtonian action for the gravity to extract the modified poisson equation as follows : @xmath14 = 4\pi g \rho , \label{mond2}\ ] ] where in the spherical symmetric systems this equation reduces to equation ( [ mond ] ) . one of the problems with mond is that it is not a covariant gravity model . bekenestein ( 2004 ) proposed a covariant formulation of this model . this theory in addition to the metric has a scalar field as well as a 4-vector field called teves , where in the limit of non - relativistic , small accelerations , the field equation reduces to that in mond . the total potential in this theory is given by the sum of the newtonian potential @xmath15 and the potential due to the scalar field , @xmath16 : @xmath17 where the added scalar field plays the role of the dark matter . teves theory has also a newtonian limit for the non - relativistic dynamics with significant acceleration . in the non - relativistic limit we can simplify the gravity inside the spherical system as @xmath18 the advantage of mond is that it could provide a successful fit to the rotation curve of spiral galaxies and dispersion velocity of elliptical galaxies @xcite . it has been tested against the cosmic background radiation @xcite , gravitational lensing @xcite , stellar systems and galactic dynamics @xcite , solar system @xcite and tully - fisher and freeman laws @xcite . mond also decreases the mass discrepancy in the cluster of galaxies @xcite but yet in clusters it remains necessary to invoke the undetected matter , possibly in the form of a massive neutrino @xcite . studying the cosmology and formation of the large structures in the universe is another tool to examine mond . studying the mondian scenario of the structure formation has been started by felten ( 1984 ) and sanders ( 1998 ) . in the paper by sanders ( 1998 ) , it is shown that a patch of universe in the mondian regime smaller than the horizon size can evolve with a different rate than the background , hence the structures naturally can be formed through this scale dependent dynamics . in this scenario small structures form before the larger ones and this provides a bottom - up hierarchical procedure for the formation of the structures in the universe . the problem with this model is that the center of collapse is not identified and every point in the space depending on the choice of the coordinate system can be considered as the center of the structure . for solving this problem one can consider mond formula to be applied to a peculiar acceleration developing from density fluctuation rather than hubble expansion @xcite . nusser ( 2002 ) used the amplitude of the cmb anisotropies as the initial condition in the n - body simulation to simulate the large scale structures in the universe . however to have a compatible result , one has to either decrease @xmath19 by one order of magnitude or reduce the amplitude of the fluctuations at the initial condition . recent studies in the formation of the galaxies by sanders ( 2008 ) shows that massive elliptical galaxies may be formed at @xmath20 , as a consequence of the monolithic dissipation - less collapse . applying mond with cooling mechanism put an upper limit to the stellar clustering in the form of the galaxy . extending the structure formation in teves theory has been done by linear perturbation of metric , vector and scalar fields . the predictions are compatible with the observations of the structures @xcite . skordis ( 2008 ) also used a generalized teves theory to construct the primordial adiabatic perturbations of a general family of the scalar field kinetic functions . in this work we extend the spherical collapse model in mond for studying the general behavior of the structures during their formation . the initial baryonic density contrast for the structures is taken from the cmb anisotropy . we obtain the dynamics of the baryonic structures in the early epoches with the linear newtonian structure formation until the entering of the structure to the mondian regime and follow the evolution with the mond . the dynamical evolution with mond shows that the structures virialize with power - law distribution of matter . we show that while all of the structures re - collapse and virialize in the mondian regime , the gravitational acceleration of the structures in this stage inversely depends on the size of the structure . this dynamical behavior of the structure formation is compatible with the less - existence of the dark matter in the globular clusters and central parts of the galaxies in cdm scenario . the organization of the paper is as follows : in section [ str ] we give a brief review of the spherical collapse model in mond and extend it by looking at the dynamics of each layer in the onion model . in section [ evol ] we discuss about the results of the calculation , showing that the density profile inside the structure is a time varying function during the evolution and the structure at the final stage , virialize from the center to the outer areas . in section [ conc ] we summarize and discuss the results . in this section we introduce the mondian cosmology and apply mond in the spherical collapse model for studying the formation of the structures . here we take the onion model for the spherical structure , dividing the sphere to the co - centric shells and studying the evolution of each layer separately and the structure as a whole . [ mondcos ] in the standard cosmology the dynamics of the universe in the matter dominated regime can be derived from the newtonian gravity . sanders ( 1998 ) used this approach to obtain the dynamics of universe in the mondian scenario . in modian cosmology , a patch of universe can evolve with a different rate than the background as soon as the acceleration fulfills the condition of @xmath8 . the result is dynamical decoupling of the smaller scales from the background which causes the production of the over dense regions in the universe . the reason for this feature of cosmological dynamics in mond is that unlike the newtonian mechanics , the acceleration in the comoving frame depends on the length scale . let us take a spherical region with radius @xmath21 from the background in which @xmath22 . using the newtonian dynamics the acceleration is given by @xmath23 where @xmath24 is the active gravitational mass and to have a compatible relation with the relativistic results , we define it to be composed of the relativistic and non - relativistic matter as @xmath25 where @xmath26 and @xmath27 are the density and pressure of the cosmic fluid . substituting equation ( [ mass ] ) in equation ( [ newt_equation ] ) , the acceleration is given by : @xmath28 in equation ( [ nc ] ) , the gravitational acceleration increases linearly with @xmath21 . this implies that there should be a critical radius @xmath29 where inside it the acceleration is smaller than the mond threshold @xmath30 and the dynamics is given by the mond where outside that radius it is newtonian . this critical length scale obtain by equaling the left hand side of equation ( [ nc ] ) with @xmath31 as @xmath32 this length scale separates the newtonian @xmath33 and the mondian @xmath34 domains . as the density of universe changes with the expansion of the universe , the critical radius also changes with time . from the continuity equation , the matter and the radiation densities vary as @xmath35 and @xmath36 . using the definition of the critical density @xmath37 , equation ( [ rc ] ) can be written in terms of the density parameters , @xmath38 , and the scale factor @xmath39 where we adapt @xmath40 , @xmath41 , @xmath42 and @xmath43 . now we do comparison of the size of a structure with @xmath44 . from equation ( [ rc2 ] ) the critical radius @xmath44 changes with the scale factor as @xmath45 in the radiation and @xmath46 in the matter dominant epoches . on the other hand the size of the structure is proportional to the scale factor ( i.e. @xmath47 ) . so we expect that the newtonian structures eventually will enter the mondian regime as @xmath44 grows faster than @xmath48 . using the adapted cosmological parameters , the critical radius at the present time is obtained @xmath49 which means that the whole observable universe resides in a mondian domain . comparing these two length scales at the last scattering surface results in @xmath50 . the mass corresponding to this critical radius at last scattering surface is about @xmath51 and for @xmath52 the dynamics is given by mond . while we expect to have density contrast growth for the scales @xmath53 at the decoupling , comparing the critical mass @xmath54 with the jeans mass of @xmath55 at this time indicates that the structures at these scales should be washed out by the pressure @xcite . the other feature of mondian cosmology is that unlike the standard cosmology where decoupling redshift is smaller than the equality redshift ( @xmath56 ) , in mond the equality epoch is much after than the decoupling time . this feature results from this fact that decoupling is related to the baryonic density of the universe and the temperature and both parameters depend only on the scale factor , independent of the dynamics of the universe . so we expect to have the same decoupling redshift in the mond as the standard cosmology . however since the dark matter does not exist in the mondian cosmology , the equality will be shifted to the lower redshifts . for our adopted cosmological parameters the equality redshift is obtained @xmath57 . in this part we model the evolution of a structure with an over - dense spherical region in mond . to calculate the evolution of this spherical patch , for simplicity we take this over - dense region with a top - hat distribution of matter . as the acceleration of this structure depends on the distance from the center , we expect to have different dynamics for each radius . hence , we divide the structure into the co - centric spherical shells like an onion model in cosmology @xcite and calculate the dynamics of each shell separately . the initial density contrast for this over - dense region is taken from the fluctuations of the last scattering surface . in the newtonian treatment of the structure formation @xmath58 , the density contrast grows linearly @xmath59 from the decoupling epoch up to the entrance of the structure to @xmath44 . for the mondian regime @xmath60 , we switch the dynamics to mond an calculate the evolution of the structure . as an example let us take a sphere with a mass of @xmath61 and find its acceleration with respect to the center . in figure ( [ n - mond ] ) we compare the acceleration of this spherical structure in mondian and newtonian dynamics as a function of the redshift . we note that the redshift is defined according to the dynamics of the scale factor at the background . at the early epoches , the difference between these two dynamics is small as @xmath62 , but after entering the structure to the mondian regime @xmath63 , the evolution of the structure by the newtonian dynamics and mond start to diverge . this deviation of the dynamics from that of newtonian plays the role of the dark matter in the standard scenario of the structure formation . for calculating the dynamics of each shell in the onion model , we take the following notation : @xmath64 is the radius of @xmath65th shell as a function of time and @xmath66 and @xmath67 are the entering time and radius of the @xmath65th shell to the mondian domain , respectively . the velocity of @xmath65th shell in terms of the hubble parameter at entrance time is given by @xmath68 , where @xmath69 is the hubble parameter of the newtonian background and @xmath70 is the density contrast of the sphere enveloped by @xmath71 . as we discussed in the previous section , all the shells will eventually enter the mondian regime and we take this time as the initial condition for each shell in the mondian evolution of the structures . table ( [ tab1 ] ) shows the initial density contrast , radius and the corresponding redshift of the entrance of each shell to the mondian regime . in the mondian regime , the acceleration approximately is @xmath72 and the evolution of each shell is as follows : @xmath73 where @xmath74 is the mass of the structure enveloped by the @xmath75 shell . using the initial conditions given by table ( 1 ) , we obtain the evolution of each shell as shown in figure ( [ r - g ] ) . to visualize the evolution of the shells , we divide the sphere into ten equidistant shells when all the structure is in the newtonian regime and obtain their evolution as a function of background redshift . the dynamics of shells shows that the inner shells evolve faster , reaching to a maximum radius and then collapse earlier than the outer ones . here the initial radius of outermost shell is about @xmath76 at the entrance time of @xmath77 to the mond regime . this shell expands up to a maximum radius of @xmath78 at @xmath79 . eventually , the shell starts to collapse and virialize at @xmath80 with the radius of @xmath81 . the maximum radius of each shell , @xmath82 , is obtained from integrating equation ( [ monddyn ] ) , letting @xmath83 as follows : @xmath84 the next phase of the evolution of shells after reaching to a maximum radius is re - collapsing . similar to the standard scenario of the spherical collapse models we expect that the global radial velocity of the structure during the free fall collapse convert to the dispersion velocity and prevent the structure from a catastrophic collapse . this steady stage of the structure is given by the virial theorem . the corresponding radius that fulfill the virial condition is called the virial radius and is calculated from : @xmath85 where @xmath86 is the gravitational potential at the @xmath65th shell . the potential in the newtonian or mondian regimes is given by @xmath87 for the non - dissipative evolution of the structure , the total energy is conserved and we substitute the right hand side of equation ( [ vir ] ) by the energy of the system at the enterance time to the mondian regime , @xmath88 . using the potentials given by equation ( [ pot2 ] ) at the left hand side of the equation ( [ vir ] ) , the virial radius for the mond and newtonian regimes obtain as @xmath89 if the virialization of the structure takes place in the mondian regime , @xmath90 implies @xmath91 which results in @xmath92 where @xmath93 is the kinetic energy for a unit mass and @xmath94 . on the other hand if the structure virializes in the newtonian regime , @xmath95 condition from the equation ( [ vir2 ] ) implies @xmath96 . we calculate the expression of @xmath97 for each shell ( see table [ tab1 ] ) and show that all the shells of the structure virialize in the mondian regime ( i.e. @xmath98 ) . table ( [ tab2 ] ) shows the parameters of shells for the moment of maximum radius and the virialization stage and fig.([r - g ] ) visualizes quantitative behavior of the shells during their evolution . in this figure the spots on the evolution curves shows the two critical stages of the maximum radius and the virialization radius , reported in table ( [ tab2 ] ) . the evolution lines of the shells show that the inner shells evolve faster and virialize at the higher redshifts , while the outer shells evolve slower . in this section we discuss the evolution of the density contrast and the profile of matter distribution inside the structure . we use the definition of the density contrast of the shells in our model : @xmath99 where @xmath100 is the density of @xmath65th shell and @xmath101 is the density of the background . as the collapsing of the shells starts from the inner to the outer parts of the structure , we will not have a shell crossing during the evolution of the structure and from the conservation of the mass , we can perform the jacobian transformation from the initial distribution of matter inside the sphere to the evolved distribution as follows : @xmath102 where @xmath103 represents the thickness of the @xmath104 shell . substituting the dynamics of each shell from the previous section in equation ( [ rho ] ) , we obtain the evolution of density contrast for the shells up to the virialization stage as shown in figure ( [ dens ] ) . the inner shells evolve faster and reach the non - linear regime @xmath105 at the higher redshifts while the outer ones evolve slower . figure ( [ f4 ] ) shows the dependence of the corresponding non - linear redshift to the mass of structure . comparing a small scale structure with the mass of @xmath106 with a galaxy having the mass of @xmath107 shows that the former structure enters the non - linear regime at @xmath108 while the later one becomes non - linear at @xmath109 . more details on the characteristic redshifts of the structures in terms of their masses is reported in table ( [ tab2 ] ) . from this table we extrapolate the dependence of the non - linear redshift , maximum radius redshift and virialization redshift to the mass of structure with the following functions ( see fig . [ f4 ] ) : @xmath110 in what follows we describe the qualitative predictions of this simple model . in this part we compare the evolution of the density profile of the structures in the spherical collapse model for the newtonian and mondian regimes . in the newtonian regime @xmath111 we have seen that the dynamics of the shells in the structure depends only on time and transforming to a comoving frame , the dynamics is scale independent ( see equation [ nc ] ) . this means that the dynamics is invariant under the scale transformation and the result is preserving the initial profile of the structure . in mond @xmath112 , since the dynamics depends on the scale unlike the newtonian case the density profile will change with time . in fig . ( [ dens - prof ] ) we plot the spatial variation of density from equation ( [ rho ] ) for a galaxy mass structure in four different stages of @xmath113 , @xmath114 , @xmath115 and @xmath116 . the initial stage at the redshift of @xmath113 is taken a top - hat distribution for the density when the innermost shell enters to the mondian regime . we continuously do jacobian transformation from this stage to the later times ( i.e. @xmath117 ) and obtain the density of each shell . while the structure evolve , the distance of shells change with time as shown with a point on the profile representing the position of each shell ( fig . [ dens - prof ] ) . as the inner shells enter to the mondian regime earlier we expect the density profile deviates from the homogeneous distribution . fig.([dens - prof ] ) shows a small deviation of the density profile from the homogenous one at @xmath118 . at this moment all the shells are in the mondian domain , however as the inner shells have been evolved faster , they are more dense than the outer shells . each shell after virialization freezes and preserves its density . we fit the density profile of the structure when the outermost shell , the latest part of the structure , virializes and the result is a power - law function of @xmath119 with the index of @xmath120 . + a question that may be answered in this simple model is the dependence of the age of the structure to the mass . observations show that small isolated structures as the globular clusters or the center of galaxies are older than the spiral arms @xcite . in this model we have seen that the smaller structures enter the mondian regime earlier , evolve faster and virialize at the higher redshifts compared to the larger structures . as an example , in table ( [ tab2 ] ) it is shown that a structure with the mass of @xmath121 virializes at @xmath122 while the whole of galaxy virializes at @xmath123 . the age of a structure from the virialization up to now is given by @xmath124 , where @xmath125 is the present age of the universe . changing the variable to the redshift results in : @xmath126 where we have adopted the cosmological parameters in ( [ mondcos ] ) . the age of structures depending on their masses is given in table ( [ tab2 ] ) . the other question in the formation of the structures is that why the smaller structures are mainly made of the baryonic matter and have a less dark matter compared to the larger ones . this feature of the structures can be explained if we can show that in addition to the fast evolution of the smaller structures , the acceleration of the structure in terms of @xmath19 , ( @xmath127 ) gets larger . to show this property of the structures in mond , we calculate the gravitational acceleration at the virialization time of each shell and compare it with @xmath19 . as the density of a structure at this stage changes with @xmath119 , the gravitational acceleration will depend on @xmath21 as @xmath128 . for @xmath129 the smaller radii should have larger acceleration compare to the larger ones . in figure ( [ na ] ) we plot the gravitational acceleration for each shell at the time of virilization in terms of the virialization radius . for the outer shells the gravitational acceleration , @xmath127 is smaller than the inner shells . this means that the smaller structures after virialization need less dark matter compare to the larger ones in the standard cdm scenario . finally we plot the mass of the structure after virialization in terms of its size in figure ( [ f7 ] ) . this feature reveals the general property of structures in mond and an accurate result to compare with the observation may be obtained by n - body simulation of the structure formation . summarizing this work , we extend the spherical collapse model to study the generic properties of the structure formation in mond . we showed that in the mondian scenario , structures can evolve without a need to the dark matter and have the following three main features : ( * a * ) mondian spherical collapse unlike to that of cdm does not preserve the initial density profile of the structure . in mond starting with an initial homogenous density profile the structure virializes with a power - law distribution of the matter , having singularity at the center . ( * b * ) we showed that the small scale structures enter the mondian regime earlier than the larger ones , evolve faster and virialize at higher redshifts . this picture from the spherical collapse model in mond provides a bottom - top scenario for the structure formation in which the smaller structures formed before the larger ones . this result is compatible with the observations where the old stars are located in the smaller structures such as the globular clusters or center of galaxies . ( * c * ) finally we showed that while the smaller structures enter the mondian regime before the larger ones , they have larger gravitational acceleration at the virialization time and hence need a less dark matter in cdm scenario . ) as a function of redshift . for the early universe the difference between the two dynamics is small but increases at the later times . the corresponding redshift for entering of the structure to the critical radius is @xmath130 . ] as a function of redshift , @xmath131 . we take ten equidistant shells at the initial condition for representing the evolution of each shell . the trend of the evolution is reaching to a maximum radius , recollapsing and virializing . the dashed line connects the maximum radii of the shells and the dotted line indicates the virialized line , connecting the virilize points of each shell . ] , as a function of distance from the center of structure in four different redshifts . the initial density profile is taken top - hat distribution with ten equidistant shells for representing the evolution of the density . the inner most shell enters to the mond domain at @xmath132 . each point on the curves notifies the position of the shell . we plot density profile of the structure for @xmath133 and @xmath134 at the left panel and @xmath135 and @xmath136 at the right panel for comparison . at @xmath137 all the shells virialize and the density profile freezes with a power - low function of @xmath138 where @xmath139 . ] as a function of shell size at the virialization time . spots on the curve represent the acceleration and the position of each shell . the gravitational acceleration in the inner shells is larger than the outer shells . ] crrrrr i & @xmath141 $ ] & @xmath142 $ ] & @xmath143 & @xmath144 & @xmath145 + 1&@xmath146&@xmath147 & @xmath148 & @xmath149 & 2.24 + 2&@xmath150&@xmath151 & @xmath152 & @xmath153 & 2.28 + 3&@xmath154&@xmath155 & @xmath156 & @xmath157 & 2.32 + 4&@xmath158&@xmath159 & @xmath160 & @xmath161 & 2.36 + 5&@xmath162&@xmath163 & @xmath164 & @xmath165 & 2.41 + 6&@xmath166&@xmath167 & @xmath168 & @xmath169 & 2.48 + 7&@xmath170&@xmath171 & @xmath172 & @xmath173 & 2.51 + 8 & @xmath174&@xmath175 & @xmath176 & @xmath177 & 2.53 + 9 & @xmath178&@xmath179 & @xmath180 & @xmath181 & 2.55 + 10 & @xmath182&@xmath183 & @xmath184 & @xmath185 & 2.57 + crrrrrrrrrrr @xmath65 & @xmath186$]&@xmath187$]&@xmath188 & @xmath189&@xmath190 $ ] & @xmath191 & @xmath192&@xmath193 & age[gyr ] + @xmath194 & @xmath146 & @xmath195 & @xmath196 & @xmath197 & @xmath198 & @xmath199 & @xmath200 & @xmath196 & 12.48 + @xmath201 & @xmath150 & @xmath202 & @xmath203 & @xmath151 & @xmath204 & @xmath205 & @xmath206 & @xmath207 & 12.43 + @xmath208 & @xmath154 & @xmath209 & @xmath210 & @xmath211 & @xmath212 & @xmath213 & @xmath214 & @xmath215 & 12.38 + @xmath216 & @xmath158 & @xmath217 & @xmath218 & @xmath219 & @xmath220 & @xmath221 & @xmath222 & @xmath223 & 12.34 + @xmath224 & @xmath162 & @xmath225 & @xmath226 & @xmath227 & @xmath228 & @xmath229 & @xmath230 & @xmath231 & 12.29 + @xmath232 & @xmath166 & @xmath233 & @xmath234 & @xmath235 & @xmath236 & @xmath237 & @xmath238 & @xmath239 & 12.25 + @xmath240 & @xmath170 & @xmath241 & @xmath242 & @xmath243 & @xmath244 & @xmath245 & @xmath246 & @xmath247 & 12.19 + @xmath248 & @xmath174 & @xmath249 & @xmath250 & @xmath251 & @xmath252 & @xmath253 & @xmath254 & @xmath255 & 12.17 + @xmath256 & @xmath178 & @xmath257 & @xmath258 & @xmath259 & @xmath260 & @xmath261 & @xmath262 & @xmath263 & 12.15 + @xmath264&@xmath182 & @xmath265 & @xmath266 & @xmath267 & @xmath200 & @xmath268 & @xmath269 & @xmath270 & 12.09 +

modeling the structure formation in the universe , we extend the spherical collapse model in the context of mond starting with the linear newtonian structure formation followed by the mondian evolution . in mond the formation of structures speed up without a need for dark matter . starting with the top - hat over - dense distribution of the matter , the structures virialize with a power law profile of the distribution of matter . we show that the virialization process takes place gradually from the center of the structure to the outer layers . in this scenario the smaller structures enter to the mondian regime earlier and evolve faster , hence they are older than larger structures . we also show that the virialization of the structures occur in the mondian regime , in which the smaller structures have stronger gravitational acceleration than the larger ones . this feature of the dynamical behavior of the structures is in agreement with this fact that the smaller structures as the globular clusters or galactic bulges have been formed earlier and need less dark matter in cdm scenario .

introduction structure formation in mond predictions of the model conclusion

content images refer to images appearing on the display screens of electronic devices such as computers and smart phones @xcite , @xcite . these images have similar characteristics as mixed content documents ( such as a magazine page ) . they often contain two layers , a pictorial smooth background and a foreground consisting of text and line graphics . the usual image compression algorithms such as jpeg2000 @xcite and hevc intra frame coding @xcite may not result in a good compression rate for this kind of images because the foreground consists of sharp discontinuities . in these cases , segmenting the image into two layers and coding them separately may be more efficient . the idea of segmenting an image for better compression was proposed for check image compression @xcite , in djvu algorithm for scanned document compression @xcite and the mixed raster content representation @xcite . foreground segmentation in such images is also a necessary preprocessing step for text extraction @xcite , which is essential for automatic character recognition and image understanding @xcite . screen content and mixed document images are hard to segment , because the foreground may be overlaid over a smoothly varying background that has a color range that overlaps with the color of the foreground . also because of the use of sub - pixel rendering , the same text / line often has different colors . even in the absence of sub - pixel rendering , pixels belonging to the same text / line often have somewhat different colors . different algorithms have been proposed in the past for foreground - background segmentation in mixed content document images and screen - content video frames such as hierarchical k - means clustering in djvu @xcite and shape primitive extraction and coding ( spec ) @xcite . there is also some recent work for text extraction in screen content images called scale and orientation invariant text segmentation @xcite . the hierarchical k - means clustering method proposed in djvu applies the k - means clustering algorithm with k=2 on blocks in multi - resolution . it first applies the k - means clustering algorithm on large blocks to obtain foreground and background colors and then uses them as the initial foreground and background colors for the smaller blocks in the next stage . it also applies some post - processing at the end to refine the results . this algorithm has difficulty in segmenting regions where background and foreground color intensities overlap and it is hard to determine whether a pixel belongs to the background or foreground just based on its intensity value . in the shape primitive extraction and coding ( spec ) method , which was developed for segmentation of screen content @xcite , a two - step segmentation algorithm is proposed . in the first step the algorithm classifies each block of size @xmath0 into either pictorial block or text / graphics based on the number of colors . if the number of colors is more than a threshold , 32 , the block will be classified into pictorial block , otherwise to text / graphics . in the second step , the algorithm refines the segmentation result of pictorial blocks , by extracting shape primitives ( horizontal line , vertical line or a rectangle with the same color ) and then comparing the size and color of the shape primitives with some threshold . because blocks containing smoothly varying background over a narrow range can also have a small color number , it is hard to find a fixed color number threshold that can robustly separate pictorial blocks and text / graphics blocks . furthermore , text and graphics in screen content images typically have some variation in their colors , even in the absence of sub - pixel rendering . these challenges limit the effectiveness of spec . in the scale and orientation invariant text segmentation , the authors proposed a two stage procedure where in the first step , a coarse segmentation layer for textural regions is extracted based on the intensity variation distribution of text characters . there could be a few pictorial regions in the segmentation output of the first stage . in the second stage , a textual connected component ( tcc ) based refinement is proposed to eliminate the survived pictorial regions . this approach is primarily based on morphological operations , and since it is designed based on a lot of assumption on the underlying text in the foreground , it can not be used for segmentation of other foreground ( such as line graphics patterns ) in screen content . the above problems with prior approaches motivate us to design a segmentation algorithm that does not rely solely on the pixel intensity but rather exploits the smoothness of the background region . in other words , instead of looking at the intensities of individual pixels and deciding whether each pixel should belong to background or foreground , we first look at the smoothness of a group of pixels and then decide whether each pixel should belong to background or foreground . towards this goal , in a preliminary work @xcite , we have developed a least absolute deviation ( lad ) method , which fits a smooth model to an image block and classifies the pixels to either background or foreground based on the fitting error . it uses the @xmath1 norm on the fitting error to enforce the sparsity of the error term . although this algorithm achieved significantly better segmentation than both the djvu and spec , it suffers from several problems . one is the difficulty of determining the number of bases to be used in the smooth model . if we use very few bases we will not be able to represent a complicated background , and if we use too many bases we may end up with representing the foreground pixels with the smooth model and considering them as background . the second problem is that this prior method does not explicitly exploit the fact that foreground pixels are typically connected . finally , minimizing the @xmath1 norm of the fitting error does not always give the sparsest fitting error . in this work , we propose two segmentation algorithms , one relying on the robust regression technique @xcite and the other one using sparse decomposition . these algorithms overcome some of the above problems , and to the best of our knowledge they have not been investigated previously . in the robust regression based approach , we try to find a smooth model that can fit the background pixels accurately . one important point is that , when the smooth model is fitted to the image , it should not be affected by the foreground pixels . this is accomplished through the ransac ( random sample consensus ) method , which minimizes the number of pixels whose fitting error exceeds a certain threshold . in the other approach , which uses sparse decomposition ( sd ) , the background and foreground part of the image are modeled with a smooth component and a sparse component , respectively . since we do not know in advance how many basis functions to include for the background part , the method chooses from a large set of bases that we think are sufficient to represent the most `` complex '' background , while minimizing coefficient @xmath1 norm to avoid overfitting of the smooth model on the foreground pixels . the total variation @xcite of the foreground component is also added in the cost function to promote the connectivity of the foreground pixels . both ransac and sd algorithms treat the smooth part of the image as background , and fit a smooth model to the image and treat outliers as foreground . but there are some differences between these two algorithms . in ransac , the model is fitted by directly minimizing the number of outliers ( @xmath2 norm of outliers ) , whereas in the sparse decomposition case , the @xmath1 norm of foreground pixels are minimized . on the other hand , the sd algorithm specifically promotes the sparsity of the background coefficients and the connectivity of foreground pixels , whereas the ransac method does not enforce these properties . we note that none of the prior approaches for background / foreground segmentation explicitly make use of the fact that the background is typically smoothly varying , even though it may have a large dynamic range ( e.g. a linear change from black to white ) . in fact any clustering - based segmentation method would fail for the case where the background color is smoothly changing over a relatively large intensity range , and the background color could be similar to the foreground color in some regions . but using the proposed method we can easily segment this kind of images . an example of this case is shown in figure 1 where we compare the segmentation result by our algorithm with the result of hierarchical k - means clustering used in djvu @xcite . as it can be seen , in the clustering based result the background pixels are split into separate clusters . we would also like to note the difference between the segmentation problem considered in this paper from the `` background - foreground separation '' in video surveillance systems , where the background refers to the stationary part of a captured scene , and the foreground refers to the moving objects . recently , efficient algorithms have been proposed to solve such a problem with low - rank decomposition @xcite-@xcite . in those works , consecutive video frames are stacked as columns in a matrix . the stationary background that repeats in successive frames ( with possible luminance change ) leads to a low - rank matrix , whereas the moving object in the scene appears as sparse outliers . by decomposing the original matrix into a low - rank matrix and a sparse matrix , the background and foreground are separated . for the separation of background and foreground problem considered here for a screen content image , the foreground refers to the text and graphics , which is typically overlaid on a smoothly varying background , and we are limited to examine this image only for their separation . there are two ways to apply low - rank decomposition to still image segmentation . the first one is to treat a single image as a matrix and perform low - rank decomposition on it . but such a matrix may not always be decomposed into a low rank matrix and a sparse matrix . even if this is possible for some image or block , the low - rank component may not correspond to the smooth background . the other way is to extract the blocks of size nxn in an image and convert each block into a column vector and then put the columns from all blocks into a matrix and applied rpca to decompose the matrix into a low - rank matrix and a sparse matrix , and then treated the sparse component as foreground . we tried both of these approaches on some of the images in our dataset , and it turns out the doing low - rank decomposition on the whole image works better than the other approach . figure 2 shows the result of sparse and low - rank decomposition for two sample images in our dataset . note that each component could have negative values , so to better display the results we show each image on its effective range . for sparse component we show its absolute value , otherwise some of the detected texts in that layer will not be visible . as it can be seen , there are some clear artifacts in the low - rank component of the first image . also for the second image , many of the background regions are still active in the sparse component . we believe by adding some extra regularization terms to the cost function , the low - rank decomposition based approach could be much improved , and it is worth investigation for future research . both of our proposed schemes assume the smooth background can be represented well by a few low frequency basis and the foreground is sparse . this is different from the prior image segmentation works @xcite-@xcite that assume an image ( directly or after some geometric transformation ) consists of a low rank component and a sparse component @xcite , and use the method of low - rank decomposition @xcite to separate the low rank component and the sparse component . because the smooth backgrounds in screen content images may not always have low rank and the foreground may happen to have low rank patterns ( e.g. horizontal and vertical lines ) , applying such decomposition and assuming the low rank component is the background and the sparse component is the foreground may not always yield satisfactory results . figure 2 shows the result of sparse and low - rank decomposition for three sample images in our dataset . note that each component could have negative values , so to better display the results we show each image on its effective range . for sparse component we show its absolute value , otherwise some of the detected texts in that layer will not be visible . the foreground map for each case is derived by thresholding the corresponding sparse component . as can be seen , there are some horizontal and/or vertical structures in the low rank components that do not belong to the background for the first two images . also for the second image , some of the foreground patterns are not segmented in the sparse component . only for the third image , where the background actually has low rank and the foreground is very sparse and disconnected ( hence lacks low - rank components ) , the low - rank decomposition method yields very good result . therefore , the low - rank decomposition framework is not always appropriate for the background - foreground segmentation problem . as shown in section iv , the proposed approach that uses low frequency smooth bases to represent the background is valid for a variety of background patterns and works well for all the screen content images we have examined . + + the structure of this paper is as follows : section ii presents the core idea of the proposed segmentation methods . the procedure of basis selection for background representation is explained in section ii.a . section ii.b briefly describes the first proposed scheme based on the ransac algorithm . the second proposed algorithm based on sparse decomposition is described in section ii.c . the final segmentation algorithm that includes both the core robust regression / sparse decomposition algorithm as well as preprocessing steps and quadtree decomposition is discussed in section iii . section iv provides the experimental results for these algorithms . and finally the paper is concluded in section v. the main idea of this work lies in the fact that if an image only consists of background , it should be well represented with a few smooth basis functions . by well representation we mean that the approximated value at a pixel with the smooth functions should have an error less than a desired threshold at every pixel . whereas if the image has some foreground pixels overlaid on top of a smooth background , those foreground pixels can not be well represented using the smooth representation . since the foreground pixels can not be modeled with this smooth representation they would usually have a large distortion by using this model . therefore the background segmentation task simplifies into finding the set of inlier pixels , which can be approximated well using this smooth model . now some questions arise here : 1 . what is a good class of smooth models that can represent the background layer accurately and compactly ? 2 . how can we derive the model parameters , @xmath3 s , such that they are not affected by foreground pixels , especially if we have many foreground pixels ? for the first question , we divide each image into non - overlapping blocks of size @xmath4 , and represent each image block , denoted by @xmath5 , with a smooth model as a linear combination of a set of two dimensional smooth functions as @xmath6 . here low frequency two - dimensional dct basis functions are used as @xmath7 , and the reason why dct basis are used and how the number @xmath8 is chosen is explained in section ii.a . the 2-d dct function is defined as : @xmath9 where @xmath10 and @xmath11 denote the frequency indices of the basis and @xmath12 and @xmath13 are normalization factors and @xmath14 and @xmath15 denote spatial coordinate of the image pixel . we order all the possible basis functions in the conventional zig - zag order in the ( u , v ) plane , and choose the first @xmath8 basis functions . the second question is kind of a chicken and egg problem : to find the model parameters we need to know which pixel belongs to the background and to know which pixel belongs to background we need to know what the model parameters are . one simple way is to define some cost function , which measures the goodness of fit between the original pixel intensities and the ones predicted by the smooth model , and then minimize the cost function . if we use the @xmath16-norm of the fitting error ( @xmath17 can be 0 , 1 , or 2 ) , the problem can be written as : @xmath18 we can also look at the 1d version of the above optimization problem by converting the 2d blocks of size @xmath4 into a vector of length @xmath19 , denoted by @xmath20 , by concatenating the columns and denoting @xmath6 as @xmath21 where @xmath22 is a matrix of size @xmath23 in which the k - th column corresponds to the vectorized version of @xmath7 and , @xmath24^\text{t}$ ] . then the problem can be formulated as : @xmath25 if we use the @xmath26-norm ( i.e. @xmath27 ) for the cost function , the problem is simply the least squares fitting problem and is very easy to solve . in fact it has a closed - form solution as below : @xmath28 but this formulation has a problem that the model parameters , @xmath29 , can be adversely affected by foreground pixels . especially in least - square fitting , by squaring the residuals , the larger residues will get larger weights in determining the model parameters . we propose two approaches for deriving the model parameters , one based on robust regression and the other one based on sparse decomposition . the proposed schemes are more robust to outliers compared to least - square fitting . these algorithms are described in sections ii.b and ii.c respectively . for color images , in general we need to check all color components to see whether each can be modeled by a smooth function . but we found through simulation results that typically if the luminance is smooth , the other two chrominance components are also smooth . therefore , we first find the set of pixels whose luminance component can be modeled by a smooth function by using the luminance component for @xmath5 . once we find these pixels , we check whether their chrominance components can also be modeled by a smooth function . to find a good set of bases for background , we applied karhunen - loeve transform @xcite to a training set of smooth background images , and the derived bases turn out to be very similar to 2d dct and 2d orthonormal polynomials . therefore we compared these two sets of basis functions , the dct basis and the orthonormal polynomials which are known to be efficient for smooth image representation . the two dimensional dct basis are outer - products of 1d dct basis , and are well known to be very efficient for representing natural images @xcite . to derive 2d orthonormal polynomials over an image block of size @xmath4 , we start with the @xmath30 1d vectors obtained by evaluating the simple polynomials @xmath31 , at @xmath32 , for @xmath33 and orthonormalize them using gram - schmidt process to get @xmath30 orthonormal bases . after deriving the 1d polynomial bases , we construct 2d orthonormal polynomial bases using the outer - product of 1d bases . to compare dct and orthonormal polynomial bases , we collected many smooth background blocks of size @xmath34 from several images and tried to represent those blocks with the first @xmath8 polynomials and dct basis functions in zigzag order . because each block contains only smooth background pixels , we can simply apply least squares fitting to derive the model coefficients using eq ( 2 ) . then we use the resulting model to predict pixels intensities and find the mean squared error ( mse ) for each block . the reconstruction rmses ( root mse ) as a function of the number of used bases , @xmath8 , for both dct and polynomials are shown in figure 3 . as we can see dct has slightly smaller rmse , so it is preferred over orthonormal polynomials . we would like to note that for more complicated background patterns , one could use the hybrid linear models @xcite , @xcite to represent the background using a union of subspaces . but for screen content images , the background can usually be well - represented by a few low frequency dct bases . robust regression is a form of regression analysis , which is developed to overcome some limitations of traditional algorithms @xcite . the performance of most of the traditional regression algorithms can be significantly affected if the assumptions about underlying data - generation process are violated and they are highly sensitive to the presence of outliers . the outlier can be thought as any data - point or observation which does not follow the same pattern as the rest of observations . the robust regression algorithms are designed to find the right model for a dataset even in the presence of outliers . they basically try to remove the outliers from dataset and use the inliers for model prediction . ransac @xcite is a popular robust regression algorithm . it is an iterative approach that performs the parameter estimation by minimizing the number of outliers ( which can be thought as minimizing the @xmath2-norm ) . ransac repeats two iterative procedures to find a model for a set of data . in the first step , it takes a subset of the data and derives the parameters of the model only using that subset . the cardinality of this subset is the smallest sufficient number to determine the model parameters . in the second step , it tests the model derived from the first step against the entire dataset to see how many samples can be modeled consistently . a sample will be considered as an outlier if it has a fitting error larger than a threshold that defines the maximum allowed deviation . ransac repeats the procedure a fixed number of times and at the end , it chooses the model with the largest consensus set ( the set of inliers ) as the optimum model . there is an analogy between our segmentation framework and model fitting in ransac . we can think of foreground pixels as outliers for the smooth model representing the background . therefore ransac can be used to perform foreground segmentation task . the proposed ransac algorithm for foreground / background segmentation of a block of size @xmath4 is as follows : 1 . select a subset of @xmath8 randomly chosen pixels . let us denote this subset by @xmath35 . 2 . fit the model @xmath36 to the pixels @xmath37 and find the @xmath3 s . this is done by solving the set of @xmath8 linear equations @xmath38 . 3 . test all @xmath19 pixels @xmath5 in the block against the fitted model . those pixels that can be predicted with an error less than @xmath39 will be considered as the inliers . 4 . save the consensus set of the current iteration if it has a larger size than the largest consensus set identified so far . if the inlier ratio , which is the ratio of inlier pixels to the total number of pixels , is more than 95% , stop the algorithm . repeat this procedure up to @xmath40 times . after this procedure is finished , the pixels in the largest consensus set will be considered as inliers or equivalently background . the final result of ransac can be refined by refitting over all inliers once more and finding all pixels with error less than @xmath39 . to boost the speed of the ransac algorithm , we stop once we found a consensus set which has an inlier ratio more than @xmath41 . the segmentation results by ransac is usually very good , but it is computationally demanding . for blocks that can be easily segmented with other methods , ransac may be an overkill . therefore , we propose a segmentation algorithm that has different modes in section iii . sparse representation has been used for various applications in recent years , including face recognition , super - resolution , morphological component analysis , denosing , image restoration and sparse coding @xcite-@xcite . in this work , we explored the application of sparse decomposition for image segmentation . as we mentioned earlier , the smooth background regions can be well represented with a few smooth basis functions , whereas the high - frequency component of the image belonging to the foreground , can not be represented with this smooth model . but using the fact that foreground pixels occupy a relatively small percentage of the images we can model the foreground with a sparse component overlaid on background . therefore it is fairly natural to think of mixed content image as a superposition of two components , one smooth and the other one sparse , as shown below : @xmath42 where @xmath43 and @xmath44 correspond to the smooth background region and foreground pixels respectively . therefore we can use sparse decomposition techniques to separate these two components . after decomposition , those pixels with large value in the @xmath45 component will be considered as foreground . we will denote this algorithm as `` sd '' , for notation brevity . to have a more compact notation , we will look at the 1d version of this problem . denoting the 1d version of @xmath44 by @xmath46 , eq . ( 3 ) can be written as : @xmath47 now to perform image segmentation , we need to impose some prior knowledge about background and foreground to our optimization problem . since we do not know in advance how many basis functions to include for the background part , we allow the model to choose from a large set of bases that we think are sufficient to represent the most `` complex '' background , while minimizing coefficient @xmath2 norm to avoid overfitting of the smooth model on the foreground pixels . because if we do not restrict the parameters , we may end up with a situation that even some of the foreground pixels are represented with this model ( imagine the case that we use a complete set of bases for background representation ) . therefore the number of nonzero components of @xmath48 should be small ( i.e. @xmath49 should be small ) . on the other hand we expect the majority of the pixels in each block to belong to the background component , therefore the number of nonzero components of @xmath46 should be small . and the last but not the least one is that foreground pixels typically form connected components in an image , therefore we can add a regularization term which promotes the connectivity of foreground pixels . here we used total variation of the foreground component to penalize isolated points in foreground . putting all of these priors together we will get the following optimization problem : @xmath50 where @xmath51 and @xmath52 are some constants which need to be tuned . for the first two terms since @xmath2 is not convex , we use its approximated @xmath1 version to have a convex problem . for the total variation we can use either the isotropic or the anisotropic version of 2d total variation @xcite . to make our optimization problem simpler , we have used the anisotropic version in this algorithm , which is defined as : @xmath53 after converting the 2d blocks into 1d vector , we can denote the total variation as below : @xmath54 where @xmath55'$ ] . then we will get the following problem : @xmath56 from the constraint in the above problem , we get @xmath57 and then we derive the following unconstrained problem : @xmath58 this problem can be solved with different approaches , such as alternating direction method of multipliers ( admm ) @xcite , majorization minimization @xcite , proximal algorithm @xcite and iteratively reweighted least squares minimization @xcite . here we present the formulation using admm algorithm . + + * admm for solving l1 optimization : * + admm is a variant of the augmented lagrangian method that uses the partial update for dual variable . to solve ( 9 ) with admm , we introduce the auxiliary variable @xmath59 and @xmath14 and convert the original problem into the following form : @xmath60 then the augmented lagrangian for the above problem can be formed as : @xmath61 where @xmath62 , @xmath63 and @xmath64 denote the dual variables . then , we can find the update rule of each variable by setting the gradient of the objective function w.r.t . to the primal variables to zero and using dual descent for dual variables as shown in algorithm 1 , @xmath65 @xmath66 @xmath67 $ ] @xmath68 @xmath69 @xmath70 @xmath71 @xmath72 @xmath73 @xmath74 @xmath75 @xmath76 + here @xmath77 , and @xmath78 denotes the soft - thresholding operator applied elementwise and is defined as : @xmath79 the setting for the parameters @xmath80 and the regularization weights @xmath81 are explained in section iv . after finding the values of @xmath48 , we can find the sparse component as @xmath82 . then those pixels with values less than an inlier threshold @xmath39 in @xmath46 will be considered as foreground . to show the advantage of minimizing @xmath1 over @xmath26 , and also sparse decomposition over both @xmath1 and @xmath26 minimization approaches , we provide the segmentation result using least square fitting ( lsf ) , least absolute deviation fitting ( lad ) and also sparse decomposition ( sd ) framework for a sample image consists of foreground texts overlaid on a constant background . the original image and the segmentation results using lsf , lad and sd are shown in figure 4 . the reconstructed smooth model by these algorithms are shown in figure 5 . all methods used 10 dct basis for representing the background and the same inlier threshold of 10 is used here . as we can see , the smooth model derived by lsf is largely affected by the foreground pixels . the ideal smooth model should have the same color as actual background ( here gray ) , but because of the existence of many text pixels with white color the lsf solution tries to find a trade - off between fitting the texts and fitting the actual background , which results in inaccurate representation of either background or foreground in the regions around text . therefore the regions around texts will have error larger than the inlier threshold and be falsely considered as the foreground pixels . the smooth model produced by the lad approach was less affected by the foreground pixels than the lsf solution , because it minimizes the @xmath1 norm of the fitting error @xmath46 . however , in blocks where there is a larger percentage of foreground pixels ( bottom middle and right regions ) , lad solution is still adversely affected by the foreground pixels . the sd approach yielded accurate solution in this example , because it considers the @xmath1 norm of the fitting coefficient , @xmath48 , and the tv norm of @xmath46 , in addition to the @xmath1 norm of @xmath46 . although the lad solution leads to smaller @xmath1 norm of the fitting error , it also leads to a much larger @xmath1 norm of @xmath48 as well . by minimizing all three terms , the sd solution obtains a background model that uses predominantly only the dc basis , which represented the background accurately . to confirm that the sd solution indeed has a smaller @xmath1 norm of @xmath48 , we show below the derived @xmath48 values using each scheme . as we can see the derived @xmath48 by sd has much smaller @xmath2 and @xmath1 norm than the other two . we propose a segmentation algorithm that mainly depends on ransac / sd but it first checks if a block can be segmented using some simpler approaches and it goes to ransac / sd only if the block can not be segmented using those approaches . these simple cases belong to one of these groups : pure background block , smoothly varying background only and text / graphic overlaid on constant background . pure background blocks are those in which all pixels have similar intensities . these kind of blocks are common in screen content images . these blocks can be detected by looking at the standard deviation or maximum absolute deviation of pixels intensities . if the standard deviation is less than some threshold we declare that block as pure background . smoothly varying background only is a block in which the intensity variation over all pixels can be modeled well by a smooth function . therefore we try to fit @xmath8 dct basis to all pixels using least square fitting . if all pixels of that block can be represented with an error less than a predefined threshold , @xmath39 , we declare it as smooth background . the last group of simple cases is text / graphic overlaid on constant background . the images of this category usually have zero variance ( or very small variances ) inside each connected component . these images usually have a limited number of different colors in each block ( usually less than 10 ) and the intensities in different parts are very different . we calculate the percentage of each different color in that block and the one with the highest percentage will be chosen as background and the other ones as foreground . when a block does not satisfy any of the above conditions , ransac / sd will be applied to separate the background and the foreground . if the segmentation is correct , the ratio of background pixels over the total number of pixels should be fairly large ( greater than at least half ) . when the ratio is small , the background of the block may be too complex to be presented by the adopted smooth function model . this may also happen when the block sits at the intersection of two smooth backgrounds . to overcome these problems , we apply the proposed method recursively using a quadtree structure . when the inlier ratio of the current block is less than @xmath84 , we divide it into 4 smaller blocks and apply the proposed algorithm on each smaller block , until the smallest block size is reached . + the overall segmentation algorithm is summarized as follows : 1 . starting with block size @xmath85 , if the standard deviation of pixels intensities is less than @xmath86 ( i.e. pixels in the block have very similar color intensity ) , then declare the entire block as background . if not , go to the next step perform least square fitting using all pixels . if all pixels can be predicted with an error less than @xmath39 , declare the entire block as background . if not , go to the next step ; 3 . if the number of different colors ( in terms of the luminance value ) is less than @xmath87 and the intensity range is above @xmath88 , declare the block as text / graphics over a constant background and find the background as the color in that block with the highest percentage of pixels . if not , go to the next step ; 4 . use ransac / sd to separate background and foreground using the luminance component only . verify that the corresponding chrominance components of background pixels can also be fitted using @xmath8 basis functions with an error less than @xmath39 . if some of them can not be fitted with this error , remove them from inliers set . if the percentage of inliers is more than a threshold @xmath84 or @xmath30 is equal to 8 , the inlier pixels are selected as background . if not go to the next step ; 5 . decompose the current block of size @xmath4 into 4 smaller blocks of size @xmath89 and run the segmentation algorithm for all of them . repeat until @xmath90 . to show the advantage of quad - tree decomposition , we provide an example of the segmentation map without and with quad - tree decomposition in figure 6 . as we can see , using quadtree decomposition we get much better result compared to the case with no decomposition . when we do not allow a @xmath34 block to be further divided , only a small percentage of pixels can be represented well by a smooth function , leaving many pixels as foreground . it is worth mentioning that the gray region on the top of the image is considered as foreground in the segmentation result without using quadtree decomposition . this is because the first row of @xmath34 blocks contain two smooth background regions with relatively equal size . to enable rigorous evaluation of different algorithms , we have generated an annotated dataset consisting of 328 image blocks of size @xmath91 , extracted from sample frames from hevc test sequences for screen content coding @xcite . the ground truth foregrounds for these images are extracted manually by the author and then refined independently by another expert . this dataset is publicly available at @xcite . table 1 summarizes the parameter choices in the proposed algorithms . the largest block size is chosen to be @xmath30=64 , which is the same as the largest cu size in hevc standard . the thresholds used for preprocessing ( steps 1 - 3 ) should be chosen conservatively to avoid segmentation errors . in our simulations , we have chosen them as @xmath92 , @xmath93 and @xmath94 , which achieved a good trade off between computation speed and segmentation accuracy . for the ransac algorithm , the maximum number of iteration is chosen to be 200 . for the sparse decomposition algorithm , the weight parameters in the objective function are tuned by testing on a validation set and are set to be @xmath95 and @xmath96 . the admm algorithm described in algorithm 1 is implemented in matlab , which the code available in @xcite . the number of iteration for admm is chosen to be 50 and the parameter @xmath97 , @xmath98 and @xmath99 are all set to 1 as suggested in @xcite . maximum block size & @xmath30 & 64 + inlier distortion threshold & @xmath39 & 10 + background standard deviation threshold & @xmath100 & 3 + qaud - tree decomposition threshold & @xmath101 & 0.5 + max number of colors for text over constant background & @xmath87 & 10 + min intensity range for text over constant background & @xmath88 & 50 + sparsity weight in sd algorithm & @xmath51 & 10 + total variation weight in sd algorithm & @xmath52 & 4 + [ tblcomp ] to find the number of dct basis functions , @xmath8 , and inlier threshold , @xmath39 , for ransac and sparse decomposition , we did a grid search over pairs of these parameters , in the range of 6 to 10 for @xmath8 and 5 to 15 for @xmath39 , on some training images , and then chose the one which achieved the best result in terms of average f1-score ( defined in eq . the parameter values that resulted in the best f1-score on our training images are shown in table ii . & sd + inlier threshold & 10 & 10 & 10 + number of bases & 6 & 10 & 10 + [ tblcomp ] before showing the segmentation result of the proposed algorithms on the test images , we illustrate how the segmentation result varies by changing different parameters in ransac algorithm . the sparse decomposition algorithm would also have the same behavior . to evaluate the effect of the distortion threshold , @xmath39 , for inlier pixels in the final segmentation result , we show the foreground map derived by several different thresholds in figure 7 . as we can see by increasing the threshold more and more pixels are considered as background . the foreground maps from left to right and top to bottom are obtained with @xmath39 setting to 5 , 10 , 25 , 35 , and 45 , respectively.,title="fig : " ] 0.24 . the foreground maps from left to right and top to bottom are obtained with @xmath39 setting to 5 , 10 , 25 , 35 , and 45 , respectively.,title="fig : " ] + 0.24 . the foreground maps from left to right and top to bottom are obtained with @xmath39 setting to 5 , 10 , 25 , 35 , and 45 , respectively.,title="fig : " ] 0.24 . the foreground maps from left to right and top to bottom are obtained with @xmath39 setting to 5 , 10 , 25 , 35 , and 45 , respectively.,title="fig : " ] + 0.24 . the foreground maps from left to right and top to bottom are obtained with @xmath39 setting to 5 , 10 , 25 , 35 , and 45 , respectively.,title="fig : " ] 0.24 . the foreground maps from left to right and top to bottom are obtained with @xmath39 setting to 5 , 10 , 25 , 35 , and 45 , respectively.,title="fig : " ] to assess the effect of the number of basis , @xmath8 , in the final segmentation result , we show the foreground map derived by several different number of basis functions using the ransac method in figure 8 . 0.24 0.24 + 0.24 0.24 + 0.24 0.24 to illustrate the smoothness of the background layer and its suitability for being coded with transform - based coding , the filled background layer of a sample image is presented in figure 9 . the background holes ( those pixels that belong to foreground layers ) are filled by the predicted value using the smooth model , which is obtained using the least squares fitting to the detected background pixels . as we can see the background layer is very smooth and does not have any sharp edges . we compare the proposed algorithms with hierarchical k - means clustering used in djvu , spec , least square fitting , and lad algorithms . for spec , we have adapted the color number threshold and the shape primitive size threshold from the default value given in @xcite when necessary to give more satisfactory result . furthermore , for blocks classified as text / graphics based on the color number , we segment the most frequent color and any similar color to it ( i.e. colors whose distance from most frequent color is less than 10 in luminance ) in the current block as background and the rest as foreground . we have also provided a comparison with least square fitting algorithm result , so that the reader can see the benefit of minimizing the @xmath2 and @xmath1 norm over minimizing the @xmath26 norm . to provide a numerical comparison between the proposed scheme and previous approaches , we have calculated the average precision and recall and f1 score ( also known as f - measure ) @xcite achieved by different segmentation algorithms over this dataset . the precision and recall are defined as : @xmath102 where @xmath103 and @xmath104 denote true positive , false positive and false negative respectively . in our evaluation , we treat a foreground pixel as positive . a pixel that is correctly identified as foreground ( compared to the manual segmentation ) is considered true positive . the same holds for false negative and false positive . the balanced f1 score is defined as the harmonic mean of precision and recall , i.e. @xmath105 the average precision , recall and f1 scores by different algorithms are given in table iii . as can be seen , the two proposed schemes achieve much higher precision and recall than the djvu and spec algorithms , and also provide noticeable gain over our prior lad approach . among the two proposed methods , sparse decomposition based algorithm achieved high precision , but lower recall than the ransac algorithm . recall & f1 score + spec @xcite & 50% & 64% & 56% + djvu @xcite & 64% & 69% & 66% + least square fitting & 79% & 60% & 68% + least absolute deviation @xcite & 90.5% & 87% & 88.7% + ransac based segmentation & 91% & 90% & 90.4% + sparse decomposition alg . & 94% & 87.2% & 90.5% + [ tblcomp ] the results for 5 test images ( each consisting of multiple 64x64 blocks ) are shown in figure 10 . each test image is a small part of a frame from a hevc scc test sequence . 0.18 0.18 0.18 0.18 0.18 + 0.18 0.18 0.18 0.18 0.18 + 0.18 0.18 0.18 0.18 0.18 + 0.18 0.18 0.18 0.18 0.18 + 0.18 0.18 0.18 0.18 0.18 + 0.18 0.18 0.18 0.18 0.18 + 0.18 0.18 0.18 0.18 0.18 + 0.18 0.18 0.18 0.18 0.18 it can be seen that in all cases the proposed algorithms give superior performance over djvu and spec , and slightly better than our prior lad approach in some images . note that our dataset mainly consists of challenging images where the background and foreground have overlapping color ranges . for simpler cases where the background has a narrow color range that is quite different from the foreground , both djvu and the proposed methods will work well . on the other hand , spec does not work well when the background is fairly homogeneous within a block and the foreground text / lines have varying colors . in terms of complexity , it took 20 , 506 and 962 ms on average for a block of @xmath34 to be segmented using ransac , lad and sparse decomposition based segmentation algorithms ( with the pre - processing steps ) using matlab 2015 on a laptop with windows 10 and core i5 cpu running at 2.6ghz . because the ransac method is much faster than sd and yielded very similar segmentation accuracy , the ransac method is preferred for practical applications . the percentages of blocks processed by different steps of the proposed segmentation algorithm are listed in table iv ( for the images in our dataset ) . note that the parameters used to identify the first three types of blocks can be chosen to achieve the desired trade - off between the segmentation accuracy and computation time . had we chosen looser thresholds than those given in table i , fewer blocks will need to go through the ransac or sd algorithm , which will reduce the overall computation time . pure background & 15% + smoothly varying background & 14% + text / graphics on constant background & 11% + ransac / sd based segmentation & 60% + [ tblcomp ] this paper proposed two new algorithms for segmentation of background and foreground in screen content images . the background is defined as the smooth component of the image that can be well modeled by a set of low frequency dct basis functions and the foreground refers to those pixels that can not be modeled with this smooth representation . one of the proposed algorithms uses robust regression technique to fit a smooth function to an image block and detect the outliers . the outliers are considered as the foreground pixels . here ransac algorithm is used to solve this problem . the second algorithm uses sparse decomposition techniques to separate the smooth background from the sparse foreground layer . total variation of the foreground component is also added to the cost function to enforce the foreground pixels to be connected . instead of applying the proposed algorithms to every block , which are computationally demanding , we first check whether the block satisfies several conditions and can be segmented using simple methods . we further propose to apply the algorithm recursively using quad - tree decomposition , starting with larger block sizes . a block is split only if ransac or sparse decomposition can not find sufficient inliers in this block . these algorithms are tested on several test images and compared with three other well - known algorithms for background / foreground separation and the proposed algorithms show significantly better performance for blocks where the background and foreground pixels have overlapping intensities . the authors would like to thank ivan selesnick and patrick combettes for their useful comments regarding sparse decomposition based segmentation . we would like to thank huawei technologies for supporting this research , and the jct - vc group for providing the hevc test sequences . 1 w. zhu , w. ding , j. xu , y. shi and b. yin , `` screen content coding based on hevc framework '' , ieee transactions on multimedia , 16 , no . 5 : 1316 - 1326 , 2014 . m. zhan , x. feng and m. xu , `` advanced screen content coding using color table and index map '' , icme workshops , 2014 . a. skodras , c. christopoulos and t. ebrahimi , `` the jpeg 2000 still image compression standard '' , ieee signal processing magazine , 36 - 58 , 2001 . sullivan , j. ohm , w.j . han and t. wiegand , `` overview of the high efficiency video coding ( hevc ) standard '' , ieee transactions on circuits and systems for video technology , 22 : 1649 - 1668 , 2012 . j. huang , ek . wong and y. wang , `` check image compression using a layered coding method '' , journal of electronic imaging 7.3 : 426 - 442 , 1998 . p. haffner , p.g . howard , p. simard , y. bengio and y. lecun , `` high quality document image compression with djvu '' , journal of electronic imaging , 7(3 ) , 410 - 425 , 1998 . dequeiroz , r.r . buckley and m. xu , `` mixed raster content ( mrc ) model for compound image compression '' , electronic imaging99 . international society for optics and photonics , 1998 . j. zhang and r. kasturi , `` extraction of text objects in video documents : recent progress '' , document analysis systems . a. coates , b. carpenter , c. case , s. satheesh , b. suresh , t. wang , dj . wu and ay . ng , `` text detection and character recognition in scene images with unsupervised feature learning '' , ieee international conference on document analysis and recognition ( icdar ) , pp . 440 - 445 , 2011 . t. lin and p. hao , `` compound image compression for real - time computer screen image transmission '' , ieee transactions on image processing , 993 - 1005 , 2005 . h yang , s wu , c deng , w lin , `` scale and orientation invariant text segmentation for born - digital compound images '' , ieee transactions on cybernetics , 45.3 : 519 - 533 , 2015 . s. minaee and y. wang , `` screen content image segmentation using least absolute deviation fitting '' , international conference on image processing ( icip ) , ieee , pp . 3295 - 3299 , sept . p.j rousseeuw , am . leroy , `` robust regression and outlier detection '' , vol . 589 . john wiley and sons , 2005 . s osher , m burger , d goldfarb , j xu , w yin , `` an iterative regularization method for total variation - based image restoration '' , multiscale modeling and simulation , 460 - 89 , 2005 . t bouwmans , eh zahzah , `` robust pca via principal component pursuit : a review for a comparative evaluation in video surveillance , '' computer vision and image understanding , 2014 . e. candes , x. li , y. ma , and j. wright , `` robust principal component analysis ? '' , international journal of acm , 58(3 ) , 2011 . a sobral , a vacavant , `` a comprehensive review of background subtraction algorithms evaluated with synthetic and real videos '' , computer vision and image understanding , 122 , pp.4 - 21 , 2014 . t bouwmans , f porikli , b hoferlin , a vacavant , `` background modeling and foreground detection for video surveillance '' , crc press , 2014 . p sprechmann , am bronstein , and g sapiro , `` learning efficient sparse and low rank models '' , ieee transactions on pattern analysis and machine intelligence , 1821 - 1833 , 2015 . levey , a. , and michael lindenbaum . `` sequential karhunen - loeve basis extraction and its application to images '' , image processing , ieee transactions on 9.8 : 1371 - 1374 , 2000 . y wang , j ostermann , yq zhang , `` video processing and communications '' , upper saddle river : prentice hall , 2002 . h , kun , ay yang , and y ma , `` sparse representation of images with hybrid linear models '' , international conference on image processing , ieee , 2004 . w hong , j wright , k huang , y ma , multi scale hybrid linear models for lossy image representation , ieee transactions on image processing , 3655 - 3671 , 2006 . r. raguram , jm . frahm and m. pollefeys , `` a comparative analysis of ransac techniques leading to adaptive real - time random sample consensus '' , computer vision - eccv , springer , 500 - 513 , 2008 . j. wright , ay . yang , a. ganesh , ss . sastry , and y. ma , `` robust face recognition via sparse representation '' , ieee transactions on pattern analysis and machine intelligence , 31(2 ) , 210 - 227 , 2009 . j. yang , j. wright , ts . huang and y. ma , `` image super - resolution via sparse representation '' , image processing , ieee transactions on 19 , no . 11 : 2861 - 2873 , 2010 . jl starck , m elad , d donoho , `` redundant multiscale transforms and their application for morphological component separation '' , advances in imaging and electron physics 132 : 287 - 348 , 2004 . j. mairal , m. elad , and g. sapiro , `` sparse representation for color image restoration '' , image processing , ieee transactions on 17.1 : 53 - 69 , 2008 . z zhang , a ganesh , x liang , y ma , `` tilt : transform invariant low - rank textures '' , international journal of computer vision , 2012 . r xia , y pan , l du , j yin , `` robust multi - view spectral clustering via low - rank and sparse decomposition '' , association for the advancement of artificial intelligence , pp . 2149 - 2155 , 2014 . y li , p wang , `` robust image hashing based on low - rank and sparse decomposition '' , international conference on acoustics , speech and signal processing , ieee , 2016 . j yang , k yu , t huang , `` supervised translation - invariant sparse coding , '' ieee conference on computer vision and pattern recognition ( cvpr ) , 2010 . d needell , r ward , `` stable image reconstruction using total variation minimization '' , siam journal on imaging sciences . 4;6(2):1035 - 58 , 2013 . s. boyd , n. parikh , e. chu , b. peleato and j. eckstein , `` distributed optimization and statistical learning via the alternating direction method of multipliers '' , foundations and trends in machine learning , 3(1 ) , 1 - 122 , 2011 . iw selesnick , hl graber , y ding , t. zhang , rl . barbour , `` transient artifact reduction algorithm ( tara ) based on sparse optimization '' , ieee transactions on signal processing , 6596 - 611 , 2014 . l. combettes and v. r. wajs , `` signal recovery by proximal forward - backward splitting , '' multiscale modeling and simulation , vol . 4 , pp . 1168 - 1200 , november 2005 . i. daubechies , r. devore , m. fornasier and c.s . gunturk , `` iteratively reweighted least squares minimization for sparse recovery '' , communications on pure and applied mathematics , 63(1 ) , 1 - 38 , 2010 . dm powers , `` evaluation : from precision , recall and f - measure to roc , informedness , markedness and correlation '' , journal of machine learning technologies 2 ( 1 ) : 37 - 63 , 2011 . iso / iec jtc 1/sc 29/wg 11 requirements subgroup , `` requirements for an extension of hevc for coding of screen content '' , in mpeg 109 meeting , 2014 . https://sites.google.com/site/shervinminaee/research/image-segmentation https://web.stanford.edu/ boyd / papers / admm/

this paper considers how to separate text and/or graphics from smooth background in screen content and mixed document images and proposes two approaches to perform this segmentation task . the proposed methods make use of the fact that the background in each block is usually smoothly varying and can be modeled well by a linear combination of a few smoothly varying basis functions , while the foreground text and graphics create sharp discontinuity . the algorithms separate the background and foreground pixels by trying to fit background pixel values in the block into a smooth function using two different schemes . one is based on robust regression , where the inlier pixels will be considered as background , while remaining outlier pixels will be considered foreground . the second approach uses a sparse decomposition framework where the background and foreground layers are modeled with a smooth and sparse components respectively . these algorithms have been tested on images extracted from hevc standard test sequences for screen content coding , and are shown to have superior performance over previous approaches . the proposed methods can be used in different applications such as text extraction , separate coding of background and foreground for compression of screen content , and medical image segmentation . image segmentation , robust regression , sparse decomposition , ransac , admm .

introduction background modeling and background-foreground separation overall segmentation algorithms experimental results conclusion acknowledgment

quantum entanglement is an essential physical resource in communication protocols @xcite and information processing @xcite . remote quantum systems may become entangled through interaction with a third quantum system as in cavity qed experiments @xcite , where entanglement of photon state in two cavities results from interaction with an atomic qubit . on the other hand , atomic qubits having shared quantized motion may be entangled through interaction with electromagnetic field as in ion traps @xcite . a parametric down converter is known to generate two mode electromagnetic field in a squeezed state with the entanglement of two modes determined by squeeze parameter . direct observation of 10 db squeezing of quantum noise of light , has been reported recently @xcite . as such , two mode squezed states are a potential entanglement resource . it has been shown that the field state entanglement can be transferred to a pair of remote atoms @xcite or three remote atoms @xcite distributed in two isolated cavities . paternostro et al . , @xcite have investigated the connection between entanglement - transfer to a pair of non - interacting two - level systems and statistical properties of entangled two - mode continuous variable resource . state engineering through bilinear interactions between two remote qubits and two - mode gaussian light fields has , also , been reported @xcite . an interesting study of entanglement transfer from a three - mode quantized field to a system of three spatially separated qubits , each one made of a two - level atom resonantly coupled to a cavity mode has been presented in ref . entanglement transfer from freely propagating quantized light to an atomic system has been achieved experimentally @xcite . in this article , distributed tripartite entanglement generation through entanglement transfer from two mode squeezed field to three atomic qubits in two cavities , is proposed . two level atoms , @xmath4 and @xmath5 , prepared either in a separable state or in an entangled state interact with a single mode of two mode squeezed field in a cavity held by alice , while a third atom @xmath1 located in a remote cavity interacts with the second mode of the squeezed field . we analyze , analytically and numerically , the entanglement dynamics of atomic qubits after tracing over the field degrees of freedom . this is a natural way of obtaining a distributed channel for quantum communication , when the entangled resource is the continuous variable ( cv ) state of a photonic system . global negativity @xcite and @xmath6way negativities shar07 are used to qualify and quantify the free entanglement of three qubit mixed state . in our earlier article @xcite , the three atom system in separable initial state was found to develop @xmath7like entanglement exhibiting entanglement sudden death ( esd ) and entanglement sudden revival ( esr ) . entanglement sudden death , reported by yu and eberly @xcite for the first time , refers to disappearance of entanglement in finite time . esd observed experimentally for entangled photon pairs @xcite , and atomic ensembles @xcite is a hindrance to using the system for implementing useful protocols . it is known that entanglement can be distilled from a three qubit pure or mixed state having free entanglement . with this in mind , we look for initial conditions on three atom state and squeezed field state so as to reduce the time interval between esd and esr or make the phenomenon disappear altogether . the three qubit entanglement generation , sudden death and revival dynamics depends strongly on degree of squeezing of two mode vacuum state and initial entanglement of pair @xmath8 . it is found that the initial state quantum correlations of qubits @xmath0 translate into an increase in remote qubit entanglement and a remarkable change in the rate at which the entanglement decays . the paper is organized as follows . a brief description of the model and procedure to obtain analytical expressions for three atom state at current time from different initial states of atoms and field are given in section ii . section iii focuses on a comparative analysis of remote qubit entanglement dynamics for different initial states . a summary of results is presented in section iv . we consider the entanglement transfer process from a two - mode squeezed vacuum field to a system of three localized and spatially distributed qubits . an entangled two - mode cv state is generated with an off - line process . two atomic qubits , @xmath4 and @xmath5 , localized inside a single mode cavity @xmath9 interact resonantly with one mode of the field . a third two level atom @xmath1 , located in cavity @xmath10 , interacts with the second field mode . we assume that each mode of the squeezed field is first injected into a cavity and then interacts resonantly with atomic qubits . the scheme used is analogous to that of our earlier work in which all three atomic qubits are prepared , initially , in their ground states . here we generalize the model to investigate the effect of initial two qubit entanglement on tripartite entanglement generation . a search for squeeze parameter value and initial two qubit entangled state , suitable for generating three qubit correlations that overcome entanglement sudden death , is carried out . the action of two - mode squeezing operator@xmath11on two - mode vacuum state @xmath12 generates two - mode squeezed vacuum state @xmath13where @xmath14 , @xmath15 and @xmath16 , @xmath17 are the bosonic creation and annihilation operators for modes one and two , respectively . the two mode squeezed state is an entangled state having bipartite entanglement determined by value of squeeze parameter @xmath18 . the variances of quadrature operators for @xmath19 are below the vacuum limit . simple linear coupling is used to inject one of the field modes from a two mode field source into cavity @xmath9 in vacuum state , while the second field mode is directed to the remote cavity c@xmath20 in vacuum state . neglecting cavity mode dissipation , the resonant cavity - cv field mode coupling is described by beam splitter operator @xmath21 , \quad i=1,2 , \label{1}\]]where @xmath22 @xmath23and @xmath24are creation and annihilation operators for photons inside the i@xmath25cavity ( external field ) , respectively . the coupling between the cavity field and the external field is determined by the cavity mirror transmittance coefficient @xmath26 . after injecting the two - mode non - classical field into independent cavities @xmath9 and @xmath10 , the cavity field at @xmath27 is found to be in a mixed state @xmath28 } { \sum } } ( \tanh s)^{n+m}g_{kl}^{nm}(\theta ) \label{fieldzero } \\ & & \times \left ( \left\vert n - k\right\rangle _ { c_{1}}\left\langle m - k\right\vert \right ) \left ( \left\vert n - l\right\rangle _ { c_{2}}\left\langle m - l\right\vert \right ) , \end{aligned}\]]where @xmath29 and @xmath30 the degree of entanglement of @xmath31 is determined by the transmittance coefficient @xmath32 and is maximal at @xmath33 . the beam splitter has a disentangling effect . the composite field @xmath34 is in a mixed state for @xmath35 , while when @xmath36 it is in a separable state . consider @xmath37 identical two level atoms interacting via dipole coupling with a single - mode quantized radiation field in a resonator . the ground and excited states for the atom @xmath38 ( @xmath39to @xmath37 ) are , respectively , denoted by @xmath40 and @xmath41 . spin operators for @xmath42 atomic qubit are defined as @xmath43 , @xmath44 and @xmath45 . defining collective spin variables of @xmath37 two - level atoms as @xmath46 where @xmath47 , we may construct the eigenbasis of operators @xmath48 and @xmath49 to represent @xmath37 atom internal states . a typical basis vector in coupled basis is written as @xmath50 , with eigen values of @xmath51 and @xmath49 given by @xmath52 and @xmath53 , respectively . in the absence of atom - field coupling , the free hamiltonian given by @xmath54 for atomic qubits and @xmath55 for cavity field , determines the system dynamics . here @xmath56 is the level splitting of the two - level atoms , @xmath57 is a frequency of the electromagnetic field and @xmath58(@xmath15 ) is photon creation ( annihilation ) operator . the atom - field interaction hamiltonian given by tavis cummings model ( tcm ) @xcite in interaction picture and rotating wave approximation , has the form@xmath59for resonant ( @xmath60 ) interaction of dipoles with cavity field . here @xmath61 is the atom - field coupling strength assumed to be the same for all atoms . since the hamiltonian commutes with @xmath62 , the unitary evolution operator @xmath63 $ ] conserves the quantum number @xmath64 . for two atoms the coupled basis vectors are the set of symmetric states @xmath65 , @xmath66 , @xmath67 and antisymmetric state @xmath12 . the number state of cavity field is represented by @xmath68 . for two atoms in cavity @xmath9 , interacting resonantly @xmath69 with @xmath70 photons , interaction hamiltonaian is represented by matrix@xmath71 , \]]in the basis @xmath72 , @xmath73 , and @xmath74 . using eigenvalues and eigenbasis of @xmath75 , the unitary operator that governs the evolution of two atoms in cavity c@xmath76 is found to be @xmath77 } { a_{n}^{2}+b_{n}^{2 } } \\ \frac{-ib_{n}\sin ( f_{n}\tau ) } { \sqrt{\left ( a_{n}^{2}+b_{n}^{2}\right ) } } & \cos ( f_{n}\tau ) & \frac{-ia_{n}\sin ( f_{n}\tau ) } { \sqrt{\left ( a_{n}^{2}+b_{n}^{2}\right ) } } \\ \frac{a_{n}b_{n}\left [ \cos ( f_{n}\tau ) -1\right ] } { a_{n}^{2}+b_{n}^{2 } } & \frac{-ia_{n}\sin ( f_{n}\tau ) } { \sqrt{\left ( a_{n}^{2}+b_{n}^{2}\right ) } } & \frac{\left [ a_{n}^{2}\cos ( f_{n}\tau ) + b_{n}^{2}\right ] } { % a_{n}^{2}+b_{n}^{2}}% \end{array}% \right ) $ } , \label{u_two}\ ] ] where interaction parameter @xmath78 , @xmath79 @xmath80 , @xmath81 , and @xmath82 . for a single atom the basis states @xmath83 @xmath84 stands for the ground ( exited ) state of the atom . the unitary matrix that determines the state evolution due to interaction of a single atom with field in cavity @xmath10 , reads as @xmath85 in the basis @xmath86 , @xmath87 . the evolution operator for the two cavity composite system is obtained by taking the tensor product that is @xmath88 for a given atomic initial state@xmath89and field state @xmath90 ( eq . ( [ fieldzero ] ) ) , using unitary operators of eqs . ( [ u_two ] ) and ( [ u_one ] ) , state of composite system after interaction time @xmath91 reads as@xmath92 } { \;\sum } } ( \tanh s)^{n+m}g_{kl}^{nm}(\theta ) \left\vert \phi _ { a_{1}a_{2}b}^{n - k , n - l}(\tau ) \right\rangle \left\langle \phi _ { a_{1}a_{2}b}^{m - k , m - l}(\tau ) \right\vert , \label{rotau}\end{aligned}\]]where@xmath93 the information about the effective evolution of three atom entanglement is contained in the state operator @xmath94 , obtained from @xmath95 upon partial trace over the cv degrees of freedom that is @xmath96 the matrix @xmath94 is used to analyze , numerically , the entanglement generation between the remote qubit @xmath1 and the pair of qubits @xmath0 . we consider two different types of atomic initial states . firstly , two atoms in cavity @xmath9 are prepared in state@xmath98while the atom in cavity @xmath10 is in ground state at @xmath27 . the two atoms have varying degree of entanglement for @xmath99 . states @xmath100 and @xmath101 are separable states . interaction of atoms with two mode squeezed field is known to generate entanglement of qubit @xmath1 with pair @xmath0 @xcite , however , the role of initial entanglement of atoms in entanglement generation is not known . our object is to investigate if initial entanglement of atoms serves as a catalyst in the process of entanglement transfer or hinders it . the maximal value of entanglement generated and inhibition of zero entanglement zones are considered as the pointers or indicators of such effects . associating computational basis state @xmath102 to atomic ground state and @xmath103 to an atom in excited state , the initial three atom state is @xmath104the state of atom - field composite system after interaction time @xmath105 , obtained by using eq . ( [ rotau ] ) , reads as@xmath106 } { \;\sum } } ( \tanh s)^{n+m}g_{kl}^{nm}(\theta ) \left\vert \phi _ { a_{1}a_{2}b}^{n - k , n - l}(\tau ) \right\rangle _ { \alpha } \left\langle \phi _ { a_{1}a_{2}b}^{m - k , m - l}(\tau ) \right\vert . \label{ro1tau}\]]exact analytic expression for @xmath107 is given in appendix ( [ a1 ] ) . the state operator for atomic qubits @xmath4 , @xmath5 , and @xmath1 is obtained from @xmath108 by tracing out the field modes that is@xmath109the matrix @xmath110 in the basis @xmath111 , @xmath112 , @xmath113 , @xmath114 , @xmath115 , @xmath116 , has the form@xmath117 as a function of @xmath18 and @xmath118 . area in white represents zero negativity.,width=576,height=480 ] versus @xmath119 and @xmath118 for @xmath2.,width=576,height=480 ] another possibility , where pair of qubits a@xmath76 and a@xmath121 are in an entangled state at @xmath27 arises with the atoms prepared initially in states @xmath122 and @xmath123that is @xmath124using eq . ( [ rotau ] ) , the state of composite system after interaction time @xmath105 is found to be@xmath125 } { \;\sum } } ( \tanh s)^{n+m}g_{kl}^{nm}(\theta ) \left\vert \phi _ { a_{1}a_{2}b}^{n - k , n - l}(\tau ) \right\rangle \left\langle \phi _ { a_{1}a_{2}b}^{m - k , m - l}(\tau ) \right\vert , \label{ro2tau}\end{aligned}\]]with @xmath126 as listed in appendix ( [ a2 ] ) . the corresponding atomic density operator @xmath127 , in the basis @xmath128 , @xmath129 , @xmath130 , @xmath131 , @xmath132 , @xmath133 , reads as@xmath134 as a function of @xmath18 and @xmath118.,width=576,height=480 ] as a function of @xmath18 and @xmath135 . , width=576,height=480 ] in our earlier article @xcite , the three atom system in separable initial state was found to develop @xmath7like entanglement exhibiting entanglement sudden death ( esd ) and revival ( esr ) . entanglement sudden death refers to disappearance of entanglement of the state in finite time and was reported by yu and eberly @xcite for the first time . esd has been observed experimentally for entangled photon pairs @xcite , and atomic ensembles @xcite . it is a hindrance to using the system for implementing useful protocols . we analyze , numerically , the dependence of three qubit entanglement generation on the degree of squeezing of electromagnetic field and initial state entanglement of qubits a@xmath76 and a@xmath20 . since the focus is on the effect of initial state entanglement of qubits in cavity @xmath9 on entanglement transfer from cv field and generation of tripartite distributed entanglement , for simplicity , we neglect the effect of the mirror transmittance , i.e. , consider @xmath33 . for appropriate choice of squeeze parameter the time interval between esd and esr ( entanglement sudden revival ) is found to become shorter or disappear altogether . fortran codes were written to calculate , numerically , the system dynamics for initial states with varying degree of field state entanglement and two atom entanglement in cavity c@xmath121 , using analytic expressions obtained in section ii . to analyze the entanglement of remote qubit @xmath1 in cavity c@xmath121 with qubits @xmath0 in cavity c@xmath76 , the global negativity of partially transposed state operator was calculated . writing a general three qubit as@xmath136where @xmath137 are the basis vectors spanning @xmath138 dimensional hilbert space , the global partial transpose with respect to qubit @xmath1 is constructed from the matrix elements of @xmath139 through @xmath140the state of qubit at location @xmath141 is labelled by @xmath142 and @xmath143 where @xmath144 . the state of qubit @xmath1 is labelled by @xmath145,@xmath146 . global negativity is defined as @xmath147where @xmath148 is the trace norm of @xmath139 . global negativity lies in the range @xmath149 for a separable state to @xmath146 for a maximally entangled state . the negativity zycz98,eise99,vida02 of @xmath150 , based on peres - horodecki criterion @xcite is a natural entanglement measure and has been shown to be an entanglement monotone @xcite . as the qubits @xmath4 and @xmath5 are always in a symmetric initial state , the entanglement of qubit b with either of the qubits is the same . therefore a negative partial transpose of @xmath151 with respect to qubit @xmath1 indicates tripartite entanglement . a three qubit state may have ghz - like or w - like tripartite @xcite entanglement . in ref @xcite it has been shown that a global partial transpose may be written in terms of a two way partial transpose , a three - way partial transpose and the state operator.we notice that for states @xmath152 and @xmath153 partial transposition involves , only , matrix elements @xmath154 with @xmath155 . in other words the global partial transpose @xmath156 is equal to two way partial transpose @xmath157 @xcite . therefore the entanglement of mixed state is similar to w - like entanglement of three qubit pure states . it is a natural consequence of the fact that no direct three atom interaction takes place . firstly , we discuss the numerical results of global negativity of partial transpose of matrix @xmath110 with respect to remote qubit @xmath1 . fig . ( [ fig1 ] ) is a contour plot of @xmath158 as a function of squeeze parameter @xmath18 and interaction parameter @xmath159 , for two atoms in separable state @xmath160 . entanglement of remote qubit at peak value is found to increase with @xmath18 , being optimum for @xmath2 , where peak value of @xmath161 is around @xmath162 however , for a fixed value of @xmath18 the regions with continuously zero negativity ( white ) alternate with regions having finite global negativity and the meeting points represent values of @xmath159 for which entanglement sudden death or revival of entanglement occurs . a contour plot of @xmath163 as a function of @xmath164 , @xmath159 and @xmath2 displayed in fig . ( [ fig2 ] ) reveals that for @xmath165 that is an entangled state @xmath166 the global negativity strongly depends on value of @xmath164 . the peak value of @xmath167 is found to decrease as @xmath168 and regions with zero entanglement become wider in comparison with that for @xmath169 . a comparison of linear entropy of @xmath170 and @xmath171 at peak value of @xmath172 shows a larger value for @xmath173 , indicating that initial entanglement generates a noisier three qubit state than the state @xmath174 . figure [ fig3 ] displays the negativity @xmath175 , versus compression parameter @xmath18 and interaction parameter @xmath159 , for the initial state @xmath176 . in this case maximum value of @xmath177 corresponds to @xmath178 and occurs at @xmath179 . the interaction time after which entanglement dies is @xmath180 , for @xmath2 . when we use a compression parameter smaller than @xmath181 zero negativity regions shrink pointing to a decrease in noise however the peak value of negativity tends to zero as well . as a function of @xmath18 and @xmath135 . , width=576,height=480 ] fig . [ fig4 ] is a contour plot of global negativity @xmath183 as a function of squeeze parameter @xmath18 and interaction parameter @xmath159 , for initial state @xmath184 . as expected the interaction with squeezed field results in a decoherence of initial entanglement of qubit @xmath4 . decoherence becomes more pronounced with increment in the value of parameter @xmath18 . for @xmath185 , the peak value of @xmath186 is always less than one for the values of @xmath159 in the range shown in the plot . [ fig5 ] displays @xmath187 as a function of squeeze parameter @xmath18 and interaction parameter @xmath159 , for initial state @xmath188 . while part of the initial state entanglement of qubit @xmath4 is transferred to field degrees a small part goes to increase the entanglement of remote qubit @xmath1 with pair of qubits in cavity one . in this case the maximum value of negativity is @xmath189 , which corresponds to @xmath2 and @xmath179 . figs . ( [ fig6 ] ) and ( [ fig7 ] ) display @xmath190 and @xmath187 plotted for @xmath2 and @xmath3 , respectively . it is seen that for initial state @xmath188 the interaction time after which esd occurs is longer than that for the case of separable initial state . the peak value of @xmath191 is comparable to peak value of @xmath192 when @xmath2 . figure [ fig7 ] shows a decreased peak value of around @xmath193 both for @xmath194 and @xmath195 . in the case of initial state @xmath184 , lowering the value of @xmath18 increases the value of interaction parameter @xmath159 for which esd occurs that is an increase in three qubit quantum correlations . the initially entangled state @xmath196 with @xmath197 is better suited to generate entanglement between the remote qubit @xmath1 and the pair of qubits in cavity c@xmath198 in comparison with initial states @xmath199 . versus @xmath135 for @xmath2 , for initial states @xmath200(solid line ) and @xmath201 ( dashed line ) . , width=576,height=480 ] versus @xmath135 for @xmath3 , for initial states @xmath200(solid line ) and @xmath202 ( dashed line ) . , width=576,height=480 ] in this article , we have examined the entanglement generation due to interaction of an entangled pair of two level atoms @xmath0 in cavity @xmath203 and an atom @xmath1 in a remote cavity @xmath10 with two mode squeezed field shared by the cavities . three qubit mixed state entanglement dynamics is a function of squeeze parameter value and initial state entanglement of the qubit pair @xmath0 . firstly , starting from different initial states , analytical expressions have been obtained for three qubit mixed state . the entanglement of qubits @xmath204 is that of a w - like state , no genuine tripartite entanglement being generated . numerical values of global negativity of partial transpose with respect to qubit @xmath1 are used to search for initial states and squeeze parameter values that generate highly entangled states . we notice that the two mode squeezed field shared by two cavities generates entanglement of qubit @xmath1 with pair of qubits @xmath0 which reaches a peak value followed by sudden disappearance and revival . the dynamics of entanglement generation , sudden death and revival strongly depends on the three qubit initial state . when both the atoms in cavity @xmath9 are prepared in ground state at @xmath27 , the entanglement of remote qubit is found to increase with @xmath18 , being optimum for @xmath2 . this is significant from the point of view of practical utility of the mixed state entanglement . starting with the pair of atoms @xmath0 prepared initially in an entangled state @xmath205 , the peak value of entanglement between the remote qubit @xmath1 and pair of qubits @xmath0 decreases as @xmath206 . with atoms @xmath0 in state @xmath207 in first cavity , atom field interactiongenerates a three qubit mixed state with atom @xmath1 in second cavity highly entangled to bell pair @xmath0 . in this case , entanglement transfer from the pair @xmath0 partially compensates for the loss of entanglement due to state reduction . as a result , three qubit entanglement dynamics displays , relatively , longer sudden death free intervals . for a given value of squeeze parameter , the entanglement of remote qubit @xmath1 with qubit pair @xmath0 is of the same order as in the case when all three qubits are in a separable state , initially . however , a smaller value @xmath3 under similar conditions generates a three qubit mixed state with comparable entanglement but lesser noise . we conclude that initially entangled state @xmath188 with @xmath197 is better suited to generate entanglement between the remote qubit @xmath1 and the pair of qubits @xmath8 in cavity @xmath9 in comparison with initial states @xmath208 . j. m. raimond , m. brune , and s. haroche , rev . physics , 73 , 565 ( 2001 ) . q. a. turchette , c. s. wood , b.e . king , c. j. myatt , d. leibfried , w. m. itano , c. monroe , and d. j. wineland , phys . 81 , 3631 ( 1998 ) . h. vahlbruch , m. mehmet , s. chelkowski , b. hage , a. franzen , n. lastzka , s. goler , k. danzmann , and r. schnabel , phys . letts . , 100 , 033602 ( 2008 ) . w. son , m. s. kim , j. lee , and d. ahn , j. mod . opt . 49 , 1739 ( 2002 ) . analytic expression for @xmath107 in eq ( [ ro1tau ] ) reads as@xmath209 } { a_{nk}^{2}+b_{nk}^{2}}% \left\vert 2,-2\right\rangle \left\vert 1,-1\right\rangle \left\vert n - k , n - l\right\rangle \notag \\ & & -i\sqrt{\alpha } \left ( \cos c_{nl}\tau \right ) b_{nk}\frac{\sin \left ( f_{nk}\tau \right ) } { \sqrt{\left ( a_{nk}^{2}+b_{nk}^{2}\right ) } } \left\vert 2,0\right\rangle \left\vert 1,-1\right\rangle \left\vert n - k-1,n - l\right\rangle \notag \\ & & + \sqrt{\alpha } \left ( \cos c_{nl}\tau \right ) a_{nk}b_{nk}\frac{\left [ \cos \left ( f_{nk}\tau \right ) -1\right ] } { a_{nk}^{2}+b_{nk}^{2}}\left\vert 2,2\right\rangle \left\vert 1,-1\right\rangle \left\vert n - k-2,n - l\right\rangle \notag \\ & & -i\sqrt{\alpha } \left ( \sin c_{nl}\tau \right ) \frac{\left [ b_{nk}^{2}\cos \left ( f_{nk}\tau \right ) + a_{nk}^{2}\right ] } { a_{nk}^{2}+b_{nk}^{2}}% \left\vert 2,-2\right\rangle \left\vert 1,1\right\rangle \left\vert n - k , n - l-1\right\rangle \notag \\ & & -\sqrt{\alpha } \left ( \sin c_{nl}\tau \right ) b_{nk}^{2}\frac{\sin \left ( f_{nk}\tau \right ) } { \sqrt{\left ( a_{nk}^{2}+b_{nk}^{2}\right ) } } \left\vert 2,0\right\rangle \left\vert 1,1\right\rangle \left\vert n - k-1,n - l-1\right\rangle \notag \\ & & -i\sqrt{\alpha } \left ( \sin c_{nl}\tau \right ) a_{nk}b_{nk}\frac{\left [ \cos \left ( f_{nk}\tau \right ) -1\right ] } { a_{nk}^{2}+b_{nk}^{2}}\left\vert 2,2\right\rangle \left\vert 1,1\right\rangle \left\vert n - k-2,n - l-1\right\rangle \notag \\ & & + \sqrt{1-\alpha } \left ( \cos c_{nl}\tau \right ) a_{n+2k}b_{n+2k}\frac{% \left [ \cos \left ( f_{n+2k}\tau \right ) -1\right ] } { a_{n+2k}^{2}+b_{n+2k}^{2}% } \left\vert 2,-2\right\rangle \left\vert 1,-1\right\rangle \left\vert n - k+2,n - l\right\rangle \notag \\ & & -i\sqrt{\left ( 1-\alpha \right ) } \left ( \cos c_{nl}\tau \right ) a_{n+2k}% \frac{\sin \left ( f_{n+2k}\tau \right ) } { \sqrt{\left ( a_{n+2k}^{2}+b_{n+2k}^{2}\right ) } } \left\vert 2,0\right\rangle \left\vert 1,-1\right\rangle \left\vert n - k+1,n - l\right\rangle \notag \\ & & + \sqrt{\left ( 1-\alpha \right ) } \left ( \cos c_{nl}\tau \right ) \frac{\left [ a_{n+2k}^{2}\cos \left ( f_{n+2k}\tau \right ) + b_{n+2k}^{2}\right ] } { % a_{n+2k}^{2}+b_{n+2k}^{2}}\left\vert 2,2\right\rangle \left\vert 1,-1\right\rangle \left\vert n - k , n - l\right\rangle \notag \\ & & -i\sqrt{\left ( 1-\alpha \right ) } \left ( \sin c_{nl}\tau \right ) a_{n+2k}b_{n+2k}\frac{\left [ \cos \left ( f_{n+2k}\tau \right ) -1\right ] } { % a_{n+2k}^{2}+b_{n+2k}^{2}}\left\vert 2,-2\right\rangle \left\vert 1,1\right\rangle \left\vert n - k+2,n - l-1\right\rangle \notag \\ & & -\sqrt{\left ( 1-\alpha \right ) } \left ( \sin c_{nl}\tau \right ) a_{n+2k}% \frac{\sin \left ( f_{n+2k}\tau \right ) } { \sqrt{\left ( a_{n+2k}^{2}+b_{n+2k}^{2}\right ) } } \left\vert 2,0\right\rangle \left\vert 1,1\right\rangle \left\vert n - k+1,n - l-1\right\rangle \notag \\ & & -i\sqrt{\left ( 1-\alpha \right ) } \left ( \sin c_{nl}\tau \right ) \frac{% \left [ a_{n+2k}^{2}\cos \left ( f_{n+2k}\tau \right ) + b_{n+2k}^{2}\right ] } { % a_{n+2k}^{2}+b_{n+2k}^{2}}\left\vert 2,2\right\rangle \left\vert 1,1\right\rangle \left\vert n - k , n - l-1\right\rangle , \end{aligned}\]]where

a pair of two level atoms @xmath0 , prepared either in a separable state or in an entangled state , interacts with a single mode of two mode squeezed cavity field while a third atomic qubit @xmath1 interacts with the second mode of the squeezed field in a remote cavity . we analyze , numerically , the generation , sudden death and revival of three qubit entanglement as a function of initial entanglement of qubits @xmath0 and degree of squeezing of electromagnetic field . global negativity of partially transposed state operator is used to quantify the entanglement of three atom state . it is found that the initial entanglement of two mode field as well as that of the pair @xmath0 , both , contribute to three atom entanglement . a maximally entangled single excitation bell pair in first cavity and two mode field with squeeze parameter @xmath2 are the initial conditions that optimize the peak value of three qubit mixed state entanglement . a smaller value of @xmath3 under similar conditions is found to generate a three qubit mixed state with comparable entanglement dynamics free from entanglement sudden death . [ multiblock footnote omitted ]

introduction the model three qubit entanglement dynamics conclusions

it is generally believed that the universe , evolving from the initial big bang , underwent a series of symmetry - breaking phase transitions @xcite accompanied by the creation of topological defects , frustrations of the unbroken phase within the broken one , induced by continuity of the order parameter values . these defects appear as magnetic monopoles , cosmic strings , domain walls and textures . direct experimental tests of these ideas are unfeasible , but transitions described by similar equations occur in experimentally accessible condensed matter systems , and a new trend has unfolded which compares these two systems . this `` cosmology in the laboratory '' relies on the fact that the dynamics of phase transitions lie in universality classes and the cosmological ones are hence analogous to those of condensed matter . for instance , vortices created in the superfluid phase transitions of @xmath0 and @xmath1 have been studied experimentally ( see e.g. ref . @xcite for extensive discussions ) following an earlier suggestion in which common features with cosmic strings have been noted @xcite . similarities between cosmological phase transitions and the isotropic - nematic phase transition in liquid crystals were studied in refs . @xcite . analogy with the thermodynamics and transitions in polymer chains was drawn in ref . @xcite . the case of superconductors is of particular interest as the associated phase transition involves a local gauge symmetry - breaking process . in superconductors , cosmic strings manifest themselves as flux tubes or vortices . experiments aimed to observe defect densities in high-@xmath2 materials @xcite have lead to contradictory results with respect to the density predicted by the kibble - zurek ( k - z ) mechanism @xcite . this prediction is however accurate only for global gauge symmetry breaking , a situation where the geodesic rule for phase angle summation is valid . a local gauge treatment by hindmarsh - rajantie ( h - r ) @xcite identifies a new mechanism for defect generation , which leads to a prediction well below the first carmi - polturak experiment sensitivity , although the prediction for the second one is in reasonable agreement with observation . the above experiments were both conducted in type - ii materials , which exhibit a second order phase transition . the question naturally arises as to the extent of changes in the defect density predictions for type - i superconductors . this is the motivation for our work . distinction between type - i and type - ii superconductors is traditionally made through the ginzburg - landau ( g - l ) parameter @xmath3 , the ratio between the magnetic field penetration length , @xmath4 , and the order parameter ( scalar field ) coherence length , @xmath5 . these characteristic length scales are obtained , in the presence of a gauge field @xmath6 , from the free energy density f ( ) = 1 2m_e |i - e c |^2 + v ( ) + 1 2 ( ) , [ free ] where @xmath7 is the magnetic moment of the specimen , @xmath8 is the electron mass and @xmath9 is the order parameter . the g - l potential is usually written as @xcite v ( ) = ^2 + 2 ^4 , [ pot ] where @xmath10 is assumed to depend linearly on the temperature , @xmath11 , @xmath12 , @xmath13 and @xmath14 are constants , and @xmath2 is the critical temperature . thus one obtains = , [ lambda ] and = . [ xi ] at @xmath15 the coherence length is given by @xmath16 , with @xmath17 . for @xmath18 , the transition is second order , and @xmath19 is typically less than @xmath20 ; for @xmath21 , the transition is first order , with @xmath19 typically greater than @xmath22 . in general , both are second order for @xmath23 . first order transitions arise from the external field term in eq . ( [ free ] ) in the event that a characteristic sample dimension is greater than @xmath4 . in thermal field theory ( tft ) a first order phase transition arises from consideration of 1-loop radiative corrections to a potential of the form of eq . ( [ pot ] ) , which introduce a barrier between minima through a cubic scalar field term , as v ( ) = ^2 - ||^3 + 2 ^4 , [ 3pot ] where @xmath24 @xcite . as in the previous case , @xmath14 is constant and @xmath11 is linear with temperature . a similar @xmath25 term however , arises in considerations of gauge field fluctuations in the normal - to - superconductor phase transition @xcite , with = 8 _ 0 e c t_c , [ fluct_gamma ] resulting in a first order phase transition for all values of @xmath26 . crossovers between first and second order transitions arise in considerations of thermal fluctuations @xcite , as also in nonlocal bcs treatments @xcite . with this is mind , we adopt a potential of the form of eq . ( [ 3pot ] ) and explore constraints on @xmath27 from experiment . the results are compared with tft 1-loop radiative corrections . comparison with the results of ref . @xcite , which deals with temperatures close to @xmath2 , is attained in the limit @xmath28 . these results are then used to analyze the impact of the cubic term on the k - z and the h - r predictions for type - i superconductors , namely on the models themselves and on a possible nucleation suppression due to the slowing down of the transition induced by the potential barrier . a first - order superconductive phase transition manifests supercritical fields under variation of the temperature , as shown in figure [ curves ] . the superheating curve is given by the condition @xmath29 , for a certain value of @xmath30 . this yelds @xmath31 . in contrast , the supercooling curve is given by the condition @xmath32 for @xmath33 , corresponding to @xmath34 . the unobservable critical curve is given by the condition @xmath35 and @xmath36 , where @xmath37 is the non - vanishing minimum of the potential . this corresponds to @xmath38 . introducing the linear dependencies @xmath39 and @xmath40 , we obtain for the superheating curve (t-1 ) = 9 16 ^2 t^2 , [ sh_cond ] and hence t_sh = 2 1 + ~1 + 9 16 ^2 . [ t_sh ] the superheating curve converges to a zero - field shift in temperature from @xmath2 by @xmath41 . this shift is undetected at the current experimental temperature sensitivity of @xmath42 @xcite . therefore , a bound for the slope of @xmath43 is < t_exp . [ tbound ] the supercooling transition occurs at @xmath44 , the critical temperature , since it is determined solely by @xmath45 ( this neglects a small correction to @xmath10 , as given by ref . .critical properties of @xmath46 and @xmath47 [ cols="^,^,^,^,^",options="header " , ] table [ table2 ] provides a comparison of the bounds on @xmath48 with the prediction of ref . the analogy between cosmology and condensed matter prompts for comparison with the tft cubic term also . to do this , we compute the associated slope of @xmath49 from @xmath50 , obtaining @xmath51 . obviously , although the prediction is material independent , its formulation in terms of a reduced temperature is not . note that the dimensionality of @xmath43 here is changed with respect to the free energy potential of eq . ( [ 3pot ] ) , through a convenient @xmath8 factor this is because the dimension of the scalar field in g - l theory is @xmath52=l^{-3}$ ] , its square representing a density , while in field theory @xmath53=l^{-1}$ ] . the electron mass determines the conversion as it is absent from the kinetic term of the lagrangean density of field theory , @xmath54 , while present in the corresponding condensed matter free energy term , @xmath55 ( or , equivalently , in the coherence length : @xmath56 _ vs. _ @xmath57 ) . table [ table2 ] also includes the quantities @xmath13 and @xmath14 , both in si and natural units . as explained above , conversion is not direct , but achieved through the multiplicative factor @xmath8 . the results in table [ table2 ] include the cubic term predicted by ref . @xcite . in the absence of an applied magnetic field , each momentum - fluctuation of the gauge field @xmath6 has an expectation value given by the equipartition theorem . when suitably integrated over the momentum space ( with a cutoff @xmath58 of the order of @xmath59 ) , a^2_= 4 _ 0 t_c - 8 _ 0 e c t_c || . [ meangauge ] since @xmath60 couples to @xmath61 in eq . ( [ free ] ) , this translates into an unimportant correction to the scalar field mass , plus a negative cubic term , given by @xmath62 . this term implies a shift in the superheating temperature ( at zero field ) , of _ t = 7.25 10 ^ -12 t_c^3 h_c(0)^2 _ 0 ^ 6 , [ deltat ] with @xmath63 in _ gauss _ and @xmath19 in @xmath64 . this shift lies beyond experimental accessibility , since it requires a temperature sensitivity of @xmath65 ( for al ; @xmath66 for sn ) . however , such an experiment _ might _ be performed with al , using state of the art relative temperature measurement techniques . surprisingly , for both materials the slopes of @xmath43 predicted by tft and ref . @xcite have similar magnitudes , @xmath67 . this is an indication of the analogous underlying mechanisms behind them : the thermal averaging of the gauge field in condensed matter can be thought of as equivalent to finite temperature vacuum polarization in high energy physics ( expressed by the renormalization of 1-loop feynmann diagrams ) . let us now discuss some possible implications of the inclusion of the cubic term in the mean - field potential . since temperature sensitivity measurements constrain @xmath68 , we always assume @xmath69 . the k - z mechanism predicts a density of topological defects ( vortices ) , @xmath70 , where @xmath71 is the characteristic time scale , given by the gorkov equation , @xmath72 is the quench time , and @xmath73 is a critical exponent . moreover , it rests upon the assumption that there is a single topological defect per @xmath74 area and , therefore , one must look for changes induced in this quantity . however , since the characteristic scales of the problem are obtained via linearization of the g - l equations , close to @xmath2 and when the order parameter is small , we see _ no _ changes in this prediction . on the other hand , for a thin slab of of width @xmath75 , the h - r mechanism predicts a defect density of the order @xmath76 , where @xmath77 is the domain size immediately after the transition . this quantity is related to the highest wavenumber @xmath78 to fall out of equilibrium and is obtained from the adiabaticity relation = ^2(k ) , [ dispersion ] for a given dispersion relation @xmath79 . in the underdamped case , @xmath80 , with a photon mass given by @xmath81 . thus we obtain @xmath82{\al'e^{2}/\be\ta_q}$ ] , and hence @xmath83 . here , the introduction of a cubic term in the potential will change the photon mass , as the true vacuum shifts to = -3 + 4 . [ field ] however , since @xmath84 , the effect of the cubic term is too small to significantly change the h - r result . another effect related to metastability concerns the non - vanishing probability of the order parameter to quantum tunnel from the the symmetric ( false ) vacuum towards the non - symmetric vacuum . following refs . @xcite , the rate of transition per unit volume and time to the true vacua is given , in the thin wall approximation , by = t^4(s_3 2 t)^3/2e^-s_3/t , [ decayingrate ] where s_3(t ) = 281 1 ^7 ^9(t ) ^2(t ) [ thinwall ] is the euclidean action , and @xmath85 is the depth " of the true vacuum . the thin wall approximation is valid whenever the barrier s height is much greater than @xmath86 . this is true when @xmath43 is comparable to the other parameters , namely when @xmath87 . for eventually smaller values of @xmath43 , like those predicted by tft and ref . @xcite , the approximation fails . in fact , we have shown that the current temperature sensitivity of @xmath88 only allows values of @xmath89 smaller than @xmath90 . therefore , we must conclude that the barrier s height is not comparable to the true vacuum s depth , and the field should always tunnel through it ( i.e. with a probability close to unity ) . because of this , there is no concern that defects may not have time to nucleate within the resolution time of the measuring device , as would happen if the potential barrier were high and diminished too slowly . in this work we have examined a possible description of a type - i superconductive phase transition by introducing a cubic term in the g - l mean - field potential , inspired by a gauge field thermal averaging @xcite , and also by analogy with tft . our analysis of the bounds derived from the superheating field and temperature constraints clearly show that the contribution of any cubic term is small compared to other parameters in the g - l potential . thus the following conclusions can be drawn : first , the superheating temperature shift induced by a cubic term , derived either from ref . @xcite or from tft , increases with decreasing g - l parameter @xmath26 ( @xmath91 ; @xmath92 ) . this suggests that future experiments to search for a tft cubic term should be conducted with extreme ( @xmath93 ) type - i materials , for example @xmath10-tungsten , with @xmath94 , @xmath95 . similarly , the shift in the supercritical field might be reinvestigated using a dc squid , which currently possess a sensitivity of @xmath96 , or @xmath97 over a 10 @xmath98 m grain diameter . furthermore , the impact of any cubic term on the defect density predictions of k - z @xcite or h - r @xcite is negligible , with no suppression or slowing down of defect production because the potential barrier due to @xmath43 is not sufficient to prevent nucleation . these considerations suggest that , all else being equal , experiments to detect topological defect formation in type - i superconductors would observe a reduction of the predicted h - r defect densities by @xmath99 depending on choice of material . recent calculations @xcite however suggest that the defect structure formed in type - i materials survives significantly longer than in type - ii . given that the type - i estimate is of order @xmath100 seconds , it seems possible that the disadvantage in @xmath5 might be compensated by simple measurability . including the presence of a magnetic field with a @xmath9 dependent magnetic moment , the conditions for the reduced superheating field @xmath101 are the authors wish to thank lus bettencourt , arttu rajantie , mark hindmarsh and boris shapiro for helpful suggestions and valuable insights on various aspects of this paper . the research was partially developed while jp , tag and pv were participants in the coslab 2002 school , in krakow ; we are grateful to the staff of the jagiellonian university for their hospitality . the work was supported by grants praxis/10033/98 and cern/40128/00 of the portuguese fundao para a cincia e tecnologia . `` superconductive applications : squids and machines '' , b.b . schwartz , s. foner , eds . ( plenum , new york , 1977 ) ; `` low temperature techniques '' , a.c . rose - innes ( english universities press , london , 1964 ) . experimentally , @xmath116 $ ] , where @xmath86 depends on whether the superconductor is strong- or weak - coupled : b. muhlschlegel , ( 1959 ) 313 ; moreover , the leading @xmath117 is only the first term in a general expansion of @xmath118 in @xmath26 : a.j . dolgert , s.j . dibartolo , a.t . dorsey , ( 1996 ) 5650 .

we examine the consequences of a cubic term addition to the mean - field potential of ginzburg - landau theory to describe first order superconductive phase transitions . constraints on its existence are obtained from experiment , which are used to assess its impact on topological defect creation . we find no fundamental changes in either the kibble - zurek or hindmarsh - rajantie predictions . 0.5 cm 0.2 cm 0.2 cm 0.2 cm 0.2 cm 0.2 cm 0.5 cm

introduction temperature sensitivity bound superheating perturbation bound topological defect formation conclusions appendix

in a series of recent articles we have constructed the functional renormalization group ( frg ) method for disordered systems , applied to specific situations and beyond one loop @xcite . this method is , apart from mean field theory @xcite using replica symmetry breaking ( rsb ) and some rare exactly solvable cases @xcite , the only known analytical method which promises to handle the strong coupling glass phase of disordered elastic systems @xcite . such systems , modeled by an elastic manifold ( of internal dimension @xmath1 ) with a @xmath0-component displacement field @xmath5 ( i.e. @xmath6 and @xmath7 ) , are of high interest for numerous experiments @xcite . this so - called random manifold model still offers great theoretical challenges and a strong motivation is the hope to gain insight into glassy physics . in addition , the @xmath8 case maps onto the much studied kardar - parisi - zhang growth equation @xcite . it exhibits a strong coupling phase for which the upper critical dimension is still under debate @xcite . higher loop studies of the statics of disordered elastic systems allow , in principle , a systematic dimensional expansion , in the simplest case around @xmath9 . they are however of a rather different nature than in standard field theory for pure critical systems @xcite . thermal fluctuations are found to be formally irrelevant in these glass phases , suggesting that the physics is controlled by a zero temperature fixed point . however before this fixed point is reached , the zero temperature effective action is found to become non - analytic @xcite . although this allows to evade the so - called dimensional reduction @xcite which makes naive perturbation theory useless and yields unphysical results , it also generates amazing new subtleties in the field theory . these were analyzed in a number of papers @xcite , and although some solutions to the puzzles were proposed the physics still remains to be elucidated . an interesting limit where one can hope to gain insight into these formidable problems is the large-@xmath0 limit . since @xmath2 is formally the mean - field limit , it allows a direct confrontation between the frg method and mean field methods . a solution of the @xmath10 random manifold model for @xmath2 was proposed by mzard and parisi , using a saddle point with spontaneous replica symmetry breaking @xcite . as in other models of glasses , spontaneous rsb can be related to ergodicity breaking of the gibbs measure into several ground states @xcite . although it offers a rather elegant way out of dimensional reduction , it is by no means clear that systems with ( large but ) finite @xmath0 should exhibit such a tremendous degeneracy of low energy states ; and there are in fact indications to the contrary @xcite . it is thus crucial to develop another line of attack , even in that limit . this is what we have achieved in a previous publication , where we have computed the effective action of the theory at large @xmath0 . there we have derived the @xmath4-function of the field theory to dominant order , i.e. for @xmath2 @xcite . we have discovered that beyond the larkin length the frg flow freezes ( at least for specific initial conditions ) and that most of the features of the mzard parisi solution can be recovered . interestingly however , in this formulation there is no need for a spontaneous rsb ansatz with ultrametric structure . thus one may hope that it could be more adapted to real world situations than the rsb calculations . such rsb calculations of fluctuations around the mean - field solution have been attempted for the random - manifold problem only in the case of 1-step non - marginal rsb ( with disappointing result @xcite ) and offer , in full generality , extreme complications , as is illustrated by several studies for spin glasses @xcite . the next challenge is thus to extend the frg in a large-@xmath0 expansion beyond the dominant order ( @xmath2 ) . this is the aim of the present paper . since this is a complicated calculation , and involves developing new methods which are of interest by themselves , this paper is restricted to the calculation of the effective action and derivation of the @xmath4-function to order @xmath3 this is performed at @xmath11 and at finite temperature . the analysis of the resulting frg flow , comparison with other studies , and applications , e.g. to the strong - coupling phase of the kpz equation is involved and is the subject of a forthcoming publication . the outline of the paper is as follows . in section [ genform ] we give the general formulation of the @xmath3 expansion for the effective action of the random manifold . details and generalizations are given in appendices [ largengeneral ] , [ toy ] , and [ app : details random manifold ] . in section [ s : review ] we summarize the main results for @xmath2 . section [ s : graphmethod ] explains the derivation of the @xmath3 correction by a graphical method , which introduces a new type of diagrammatics . section [ s : algebramethod ] explains the principle of a second and complementary method based on the algebra of 4-replica tensors . section [ results ] contains the full result for the effective action to order @xmath3 , first expressed in bare parameters , then as a function of the renormalized dimensionless disorder . this allows , in section [ s : beta1overn ] , for a derivation of the @xmath4-function at @xmath11 . the structure of the finite-@xmath12 @xmath4-function is indicated , and details given in appendix [ app : detailsbeta ] . a fool - proof diagrammatic version for finite temperature is given in appendix [ excludedrepformalism ] . more details on the two main methods are given respectively in appendix [ s : moregraphics ] ( for the diagrammatic method , including an alternative derivation of the @xmath11 @xmath4-function ) and in appendix [ app : algebra ] ( for the algebraic method ) . appendix [ app : integrals ] contains a list of all integrals . a table summarizing the notation is found in appendix [ sec : notation ] . we start from the partition function of an interface @xmath13\ , { { \mathrm{e}}}^{- { \cal h}_v[u]/t}$ ] in a given sample , with energy @xmath14 = \int_q \frac{1}{2 } ( q^2 + m^2 ) u(-q ) \cdot u(q ) + \int_x v(x , u(x ) ) \ , \ ] ] where @xmath15 , @xmath16 and @xmath17 . the @xmath10 indices will be specified only when strictly necessary , and below additional replica indices for the replicated field @xmath18 will be introduced , @xmath19 . the small confining mass @xmath20 provides a scale . to obtain a non - trivial large-@xmath0 limit , one defines the rescaled field @xmath21 and chooses the distribution of the random potential to be rotationally invariant , e.g. its second cumulant as @xmath22 in terms of a function @xmath23 . higher connected cumulants are scaled as @xmath24 with @xmath25 . physical observables can be obtained for any @xmath0 from the replicated action at @xmath26 with a source @xmath27 as @xmath28&= & \int { \cal d } [ u ] { \cal d } [ \chi ] { \cal d}[\lambda ] \,{{\mathrm{e}}}^ { - n { \cal s}[u,\chi,\lambda , j ] } \\ { \cal s}[u,\chi,\lambda , j ] & = & \frac{1}{2 t } \int_q ( q^2 + m^2 ) v_a(-q ) \cdot v_a(q ) \nonumber \\ & & + \int_x u(\chi(x ) ) - \frac{1}{2 } i \lambda_{ab}(x ) [ \chi_{ab}(x ) - v_a(x ) \cdot v_b(x ) ] - j_a(x ) \cdot v_{a}(x ) \ , \end{aligned}\ ] ] where the replica matrix field @xmath29 has been introduced through a lagrange multiplier matrix field @xmath30 . here and below summations over repeated replica ( and @xmath10 ) indices @xmath31 is implicit . the bare interaction matrix potential @xmath32 depends only on the matrix @xmath33 and has a cumulant expansion in terms of sums with higher numbers of replicas . the effective action functional @xmath34 $ ] is defined as the legendre transform of @xmath35 = \ln { \cal z}[j]$ ] and satisfies @xmath36 + { \cal w}[j ] = \int j(x ) \cdot u(x ) \ .\ ] ] since @xmath34 $ ] defines the renormalized vertices , its zero - momentum limit defines the _ renormalized disorder _ , the quantity on which we focus here . thus we only need the result ( per unit volume ) for a _ uniform _ configuration of the replica field @xmath37 , which takes the form : @xmath38 where @xmath39 stands for the matrix @xmath40 . this defines the renormalized disorder potential @xmath41 and , whenever it can be expanded , up to a constant , @xmath42 it defines the _ renormalized cumulants _ @xmath43 , @xmath44 etc .. here and in the following we denote @xmath45 we aim at calculating the effective action up to terms of order @xmath46 , i.e. the first two terms in the expansion : @xmath47 details of the calculation , as well as expressions for non - uniform fields are given in appendix [ app : details random manifold ] . for the leading term we find , from a saddle - point evaluation @xcite : @xmath48^{n+1 } \\ i_{n } & : = & \int_k \frac{1}{(k^2 + m^2)^n } \label{defin}\ .\end{aligned}\ ] ] the trace acts on replica matrices and @xmath49 satisfies the self - consistent equation @xmath50^{n})_{ab } \\ g_v(k ) & = & \left [ ( k^2 + m^2 ) \delta + 2 t \partial_\chi u(\chi_v ) \right]^{-1}\ . \label{chiv}\end{aligned}\ ] ] note that for @xmath51 no uv cutoff is necessary ( apart for a constant term in the free energy ) , while for @xmath52 an uv cutoff is necessary ( and implicit in the following ) only for @xmath53 . therefore should be scaled as @xmath54 , when @xmath55 is taken to infinity . ] one also finds a compact and very useful self - consistent equation for the derivative of the zero - th order potential : @xmath56 everywhere we denote by @xmath57 the simple derivative of the function @xmath58 with respect to its matrix argument @xmath59 . ( note that @xmath60 is a _ first derivative _ of @xmath61 with respect to the matrix element @xmath62 . ) next , from calculations of the fluctuations around the saddle point , one obtains the @xmath3 correction , which can be expressed in terms of the zero - th order quantities as : @xmath63 \right ) \qquad \\ \label{lf26 } \pi_v^{ef , cd}(q ) & = & \int_k g_v^{ed}(k ) g_v^{f c}(q - k ) \\ g_v(k ) & = & [ ( k^2 + m^2 ) \delta + 2 t \partial_{v v } \tilde{u}^0(v v ) ] ^{-1}\ , \end{aligned}\ ] ] where here the trace acts in the space of replica pairs , i.e. @xmath64 . note that @xmath65 can also be replaced by the full @xmath41 in the expression of @xmath66 with the same accuracy ( i.e. at leading order in @xmath3 ) . the saddle - point equation ( [ lf15 ] ) for the zero - th order and the result for the @xmath3-correction ( [ lf25 ] ) are still formal as they encode the full renormalized disorder distribution . to yield the renormalized disorder cumulants via ( [ cumulants1]),([cumulants2 ] ) they must be expanded in the number of replica sums , i.e. in cumulants . in the following section [ s : review ] , we recall the results for @xmath2 , and proceed with the non - trivial evaluation of ( [ lf25 ] ) via a graphical method in section [ s : graphmethod ] , and via an algebraic method in section [ s : algebramethod ] . in this section we review the main results at @xmath2 . details can be found in @xcite . contribution to @xmath67 . bottom : self - consistent equation at leading order for @xmath68 . the wiggly line denotes a derivative , and is combinatorially equivalent to choosing one @xmath69 . at finite @xmath12 one can attach an additional arbitrary number of tadpoles to any @xmath69 . also note that no loop made out of 3 propagators appears : this would be a contribution to the third cumulant ( 3-replica term ) , not calculated here ; it is given in @xcite.,title="fig:",width=453 ] we start by recalling the cumulant expansion and only derive the result for the second cumulant . higher cumulants are given in @xcite . one studies a bare model , where only the second cumulant is non - zero : @xmath70 and calculates the renormalized disorder ( [ cumulants1 ] ) . ( we will drop the index zero on the cumulant functions , indicating the leading order ) . the self - consistent equation ( [ lf15 ] ) can be expanded in sums with increasing numbers of replicas . we only need : @xmath71_{ab } & = & \frac{2}{t } ( \delta_{ab } \sum_c \tilde{b}'_{ac } - \tilde{b}'_{ab } ) + { \ldots}\\ \left[(- 2 t \partial \tilde{u}^0(vv))^2\right]_{ab } & = & \frac{4}{t^2 } \left ( \delta_{ab } \sum_{ef } \tilde{b}'_{ae } \tilde{b}'_{af } - \tilde{b}'_{ab } \sum_f ( \tilde{b}'_{af } + \tilde{b}'_{bf } ) + \sum_c \tilde{b}'_{ac } \tilde{b}'_{cb } \right ) + { \ldots}\ , \label{lf28 } \end{aligned}\ ] ] where @xmath72 ( recall @xmath73 ) . here and below the dropped terms contain sums with too many replicas to contribute to the final result for the self - consistent equation of the second cumulant ( 2-replica term ) . one thus has : @xmath74 + { \ldots}\ . \qquad \end{aligned}\ ] ] the self - consistent equation becomes : @xmath75 and can be solved by appropriate taylor expansion of the r.h.s .. it is solved for @xmath76 : @xmath77 it is then easy to see that the second cumulant satisfies a closed equation at any @xmath12 , @xmath78 with no other contributions from higher cumulants at any @xmath12 . a more detailed derivation is given in @xcite . from the previous section the renormalized second cumulant of the disorder @xmath80 satisfies the self - consistent equation @xmath81 it implies @xmath82 as well as @xmath83\ .\ ] ] we now derive the corresponding exact frg equation . taking the derivative @xmath84 gives : @xmath85 \nonumber \\ & = & \frac{\tilde{b}''(x)}{1 + 4 i_2 \tilde{b}''(x ) } \nonumber \\ & & \times [ 2 m \partial_m t i_1 + 4 ( m \partial_m i_2 ) ( \tilde{b}'(x ) - \tilde{b}'(0 ) ) + 4 i_2 m \partial_m \tilde{b}'(x ) - 4 i_2 m \partial_m \tilde{b}'(0 ) ] \ .\qquad \end{aligned}\ ] ] this yields : @xmath86\ .\ ] ] thus one has also : @xmath87 hence one gets finally @xmath88$}\ , \ ] ] which can also be integrated once over @xmath89 . as emphasized in ref . @xcite it is exact at @xmath2 for any @xmath1 and correctly matches the 1-loop frg equation obtained by balents and fisher for any @xmath0 but only to @xmath90 , @xmath91 . it can be solved directly , or equivalently the self - consistent equation ( [ lf72 ] ) can be inverted . the corresponding solutions for various models are discussed in @xcite and compared with the mzard - parisi solution @xcite obtained in a rather different manner through a replica - symmetry - breaking saddle point . before discussing specific models , we now turn to the evaluation of the effective action of the frg and the @xmath4-function to the next order in @xmath3 . in this section , we present a graphical method to calculate the corrections to @xmath92 at order @xmath3 . an algebraic method is presented in the next section [ s : algebramethod ] . both methods are completely independent , since they use orthogonal ideas . they were performed independently by the authors , each on a different continent . the agreement on the final result gives some confidence that it is free from calculational errors . -expansion of the renormalized potential @xmath93 ( effective action ) . the numbers 1 , 2 and 3 depict special features discussed in the main text . there is a factor @xmath3 per dotted line and a factor of @xmath0 per `` small loop '' of propagators ( solid lines ) , thus the overall factor is @xmath0 as it should from ( [ phiint ] ) . this graph does not contain any `` big loop '' ( see text).,title="fig:",width=340 ] let us start with some general considerations about which graphs contribute at a given order in @xmath3 . for that purpose , we consider a general ( pure , no disorder ) scalar field theory , with a @xmath0-component field @xmath94 , @xmath95 and interaction @xmath96 where the @xmath97-correlations of the free theory ( @xmath98 ) are given by @xmath99 with @xmath100 and @xmath101 independent of @xmath0 is denoted by a solid line in figure [ largengen ] . graphically , we can denote this by @xmath102{./figures / phi2}{\hspace*{0.8ex}}}\parbox{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={{0.35}}]{./figures / phi2}{\hspace*{0.8ex } } } } + \frac{b_{4}}{n}{\settowidth{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={{0.35}}]{./figures / phi4}{\hspace*{0.8ex}}}\parbox{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={{0.35}}]{./figures / phi4}{\hspace*{0.8ex } } } } + \frac{b_{6}}{n^{2 } } { \settowidth{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={{0.35}}]{./figures / phi6}{\hspace*{0.8ex}}}\parbox{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={{0.35}}]{./figures / phi6}{\hspace*{0.8ex } } } } + \dots \ , \ ] ] where all coefficients @xmath103 are of order @xmath104 ; there is thus a factor of @xmath3 per dotted line . the renormalized potential @xmath93 ( effective action ) is given by the sum of all 1-particle - irreducible diagrams . they must thus contain `` small loops '' . to leading order in @xmath3 the following diagrams are possible : * tadpoles , contracting any @xmath105 with itself only . the factor of @xmath0 from the `` small loop '' ( the @xmath106 over the number of components ) compensates the factor of @xmath0 from the argument of @xmath107 . ( see 1 on figure [ largengen ] . ) * closed `` small loops '' with 2 , 3 , 4 or more vertices , as denoted by the same number on figure [ largengen ] : adding one more vertex gives a factor of @xmath0 , see eq . ( [ phiint ] ) ; this is compensated by using an additional @xmath108 . * note that all vertices @xmath105 , @xmath109 , a.s.o . contribute equivalently . ( see 5 , 6 on figure [ largengen ] . ) note that all these diagrams are `` tree - like '' diagrams , where the branches ( made of `` small loops '' ) are made out of the diagrams through which no total momentum is running . we will call them `` tree - like '' in the following , to distinguish them from normal trees . they are resummed by the saddle - point equations ( see appendix [ toy ] ) or from graphical inspection , as done here : first of all , insertions of @xmath110 into a line of propagators act like a mass , leading to the replacement of @xmath111 by @xmath112 the effective potential @xmath93 is obtained from @xmath113 note that the derivatives are graphically understood as follows : choosing one vertex ( derivative ! ) in the effective potential @xmath114 is equivalent to having a bare vertex with the same derivative taken ( thus @xmath110 ) and attaching to it loops made out of correlation - functions . attaching any number of such loops to @xmath115 , amounts to shifting its argument , as can be seen from taylor - expansion . in these loops , again derivatives ( one needs @xmath116 to attach the loop ) of the effective potential are inserted . the latter can thus be written as @xmath110 with shifted argument to account for more things to be attached to this @xmath110 or equivalently using ( [ sps ] ) to a @xmath117 . this result coincides with ( [ b24 ] ) . to subleading order in @xmath3 . the `` big loop '' ( a loop made of loops , see main text ) accounts for a factor of @xmath3 ; the same is true for the `` minimal loop '' on the right , which is a `` big loop '' in disguise ( as explained in the main text . ) the given diagram is thus of order @xmath118,title="fig:",width=491 ] take a `` tree '' ( as e.g. the object on fig . [ largengen ] ) and glue it together to form a `` big loop '' by identifying two vertices ; this does not change the factor of @xmath0 from the loops , but one looses one factor of @xmath0 from the missing vertex . note that also the `` minimal loop '' marked on figure [ fig1onphi ] belongs to the same class , even though it looks different . the `` big loop '' demands to sum a series of diagrams at non - vanishing momentum , and then to carry through the integration over momenta . however the simplification remains that any added `` tree - like '' branches are resummed by replacing the argument of @xmath115 from @xmath108 to @xmath119 . another feature arises at subdominant order : in the `` big '' loop , one may pick any given small loop ( i.e. the loop made out of correlation - functions ) and replace one of the two correlation - functions by @xmath108 : this means that the corresponding fields @xmath120 did not get contracted . since the remaining correlation - functions force their indices to be equal , this gives a factor of @xmath121 , thus contributes the same factor of @xmath0 . note that these diagrams do not contribute to the effective action at leading order , which is treelike , since the resulting diagrams would be 1-particle reducible . to resum the order @xmath3-diagrams one has to sum over loops of all sizes . the result is @xmath122^{n } \nonumber \\ & = & \frac{1}{2n } \int_{p}\ln \left[1 + 2 v '' ( \chi ) \left ( { i}_{2 } ( p ) + 2\ , { \textstyle \frac{\vec { u}^{2}}{n}\ , g ( p ) } \right ) \right]\\ { i}_{2 } ( p ) & = & \int_{k } g ( k+p ) g ( k ) \ .\end{aligned}\ ] ] this can be compared to the results of appendix [ toy ] , and more specifically to formula ( [ b.45 ] ) . let us start the treatment of the disordered model with some general considerations . first to organize the @xmath3-expansion , one may still use the diagrammatics of the previous section , which shows the @xmath10-index content . the same diagrams still exist , but they now also have a complicated replica content . the replica content can be explicated by using `` splitted vertices '' instead of the unsplitted ones of the previous section . the corresponding replica diagrammatics , which shows the replica structure _ only _ was explained in details in @xcite . this can be drawn as @xmath123{./figures / vertex}{\hspace*{0.8ex}}}\parbox{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={{0.35}}]{./figures / vertex}{\hspace*{0.8ex}}}}_{b}\ \ , \ ] ] where a _ dashed _ line connects the two dots , standing for replicas @xmath124 and @xmath125 . in order to avoid confusion , note that this dashed line is different from the _ dotted _ line used in figures [ largengen ] and [ fig1onphi ] as well as equation ( [ lf33 ] ) to show the @xmath126-structure . at dominant order ( figure [ fig1 ] ) small loops with three vertices ( see 3 in figure [ largengen ] ) , or more , do not contribute to @xmath127 , but to the third cumulant , or higher , as can be seen from @xmath128{./figures / reptri}{\hspace*{0.8ex}}}\parbox{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={{0.35}}]{./figures / reptri}{\hspace*{0.8ex}}}}\ , \ ] ] which is a diagram showing only replica indices , @xmath0 indices being implicit . note that solid lines identify replica indices . this is why in figure [ fig1 ] only chain - diagrams and tadpoles ( the latter are omitted for simplicity of presentation ) appear , rendering calculations appreciably simpler . at subdominant order in @xmath3 , only diagrams with one `` big '' loop ( general feature of the order @xmath3 discussed in the last section ) made of any number of `` small '' 2-loops and only exactly _ two _ `` small '' 3-loops can contribute to the renormalized second cumulant . each small 3-loop can also be replaced by an uncontracted @xmath130 . to understand this , consider the simplest `` big '' loop , i.e.the railroad diagram , which is drawn on the left of figure [ f : bigloop ] in splitted replica notation . it contains exactly two closed propagator lines , which over - constrain the replicas to be equal . ( these are the inner and outer solid lines on the left of figure [ f : bigloop ] ) . these over - constraints can be relaxed , by cutting each line exactly once , in order not to get a higher replica term , as is illustrated on the right of figure [ f : bigloop ] . this `` cutting '' is possible by either inserting into a propagator a vertex ( which contains two replicas , thus is not `` replica - conserving '' ) or by leaving one @xmath131 uncontracted . this is the basic principle , whose careful exploration leads to all of the diagrams at order @xmath3 , as we will discuss now . in order to do so , we have to introduce a new powerful graphical notation : @xmath132{./figures / b}{\hspace*{0.8ex}}}\parbox{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={{0.35}}]{./figures / b}{\hspace*{0.8ex } } } } & = & b(\bar \chi)\\ { \settowidth{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={{0.35}}]{./figures / bp}{\hspace*{0.8ex}}}\parbox{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={{0.35}}]{./figures / bp}{\hspace*{0.8ex}}}}{_{a } \atop ^{b } } & = & b'(\bar \chi_{ab } ) \left(\vec{u}_{a}-\vec{u}_{b } \right)^{2}\\ { _ { a } \atop ^{b}}{\settowidth{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={{0.35}}]{./figures / bpp}{\hspace*{0.8ex}}}\parbox{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={{0.35}}]{./figures / bpp}{\hspace*{0.8ex}}}}{_{a } \atop ^{b } } & = & b''(\bar \chi_{ab})\frac{1}{2 } \left [ \left(\vec{u}_{a}-\vec{u}_{b } \right)^{2}\right]^2\ , \end{aligned}\ ] ] where lines departing in the same direction belong to the same vector - index , and a continuing line represents the same replica . we also use the following short - hand notation @xmath133 and similar formulas for the higher derivatives @xmath134 , etc . another frequently used shorthand is @xmath135 in order to be able to resum the `` big '' loop , we will now introduce some building blocks . since there can be any number of 2-loops in the `` big '' loop one needs to define the resummed chain @xmath136{./figures / bsummed}{\hspace*{0.8ex}}}\parbox{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={{0.35}}]{./figures / bsummed}{\hspace*{0.8ex } } } } { _ a \atop ^b}\ : = { _ a \atop ^b } { \settowidth{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={{0.35}}]{./figures / bpp}{\hspace*{0.8ex}}}\parbox{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={{0.35}}]{./figures / bpp}{\hspace*{0.8ex } } } } { _ a \atop ^b } + { _ a \atop ^b}{\settowidth{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={{0.35}}]{./figures / b1loop}{\hspace*{0.8ex}}}\parbox{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={{0.35}}]{./figures / b1loop}{\hspace*{0.8ex } } } } { _ a \atop ^b}+ { _ a \atop ^b}{\settowidth{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={{0.35}}]{./figures / b2loop}{\hspace*{0.8ex}}}\parbox{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={{0.35}}]{./figures / b2loop}{\hspace*{0.8ex } } } } { _ a \atop ^b}+ { _ a \atop ^b}{\settowidth{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={{0.35}}]{./figures / b3loop}{\hspace*{0.8ex}}}\parbox{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={{0.35}}]{./figures / b3loop}{\hspace*{0.8ex } } } } { _ a \atop ^b}+\dots \ .\ ] ] we have with momentum @xmath137 running through the diagram @xmath138{./figures / b1loop}{\hspace*{0.8ex}}}\parbox{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={{0.35}}]{./figures / b1loop}{\hspace*{0.8ex } } } } { _ a \atop ^b } = 4 i_{2}(p)\ , b''(\bar \chi_{ab})\ { _ a \atop ^b } { \settowidth{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={{0.35}}]{./figures / bpp}{\hspace*{0.8ex}}}\parbox{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={{0.35}}]{./figures / bpp}{\hspace*{0.8ex } } } } { _ a \atop ^b}\ ] ] with @xmath139 let us introduce a compact notation for the integrals , summarized in appendix [ app : integrals ] . ( all our notations and important formulas are also summarized in a table in appendix [ sec : notation ] . ) thus @xmath141{./figures / b2loop}{\hspace*{0.8ex}}}\parbox{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={{0.35}}]{./figures / b2loop}{\hspace*{0.8ex } } } } { _ a \atop ^b } & = & \left[4 i_{2}(p)\ , b''(\bar \chi_{ab } ) \right]^{2 } \ { _ a \atop ^b } { \settowidth{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={{0.35}}]{./figures / bpp}{\hspace*{0.8ex}}}\parbox{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={{0.35}}]{./figures / bpp}{\hspace*{0.8ex } } } } { _ a \atop ^b } \\ \stackrel{{p } } { \longrightarrow } { _ a \atop ^b}{\settowidth{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={{0.35}}]{./figures / b3loop}{\hspace*{0.8ex}}}\parbox{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={{0.35}}]{./figures / b3loop}{\hspace*{0.8ex } } } } { _ a \atop ^b } & = & \left[4 i_{2}(p)\ , b''(\bar \chi_{ab } ) \right]^{3 } \ { _ a \atop ^b } { \settowidth{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={{0.35}}]{./figures / bpp}{\hspace*{0.8ex}}}\parbox{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={{0.35}}]{./figures / bpp}{\hspace*{0.8ex } } } } { _ a \atop ^b } \ , \end{aligned}\ ] ] and so on . these chain - like diagrams form a geometric series , which is resummed as @xmath142{./figures / bsummed}{\hspace*{0.8ex}}}\parbox{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={{0.35}}]{./figures / bsummed}{\hspace*{0.8ex } } } } { _ a \atop ^b } = \frac{1}{1 - 4 i_{2}(p)\ , b''(\bar \chi_{ab } ) } \ { _ a \atop ^b } { \settowidth{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={{0.35}}]{./figures / bpp}{\hspace*{0.8ex}}}\parbox{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={{0.35}}]{./figures / bpp}{\hspace*{0.8ex } } } } { _ a \atop ^b } = : \frac{1}{2 } \left[(u_{a}-u_{b})^{2 } \right ] h_{ab}(p ) \left[(u_{a}-u_{b})^{2}\right]\ .\ ] ] we have introduced @xmath143 , the `` effective '' @xmath144 after resummation @xmath145 which we equivalently can express at leading order in @xmath3 through @xmath146 $ ] as @xmath147 \tilde{b}''_{ab}}\ .\ ] ] we also define @xmath148 as @xmath149 \tilde { b } '' ( v^{2 } ) } \ , \ ] ] valid again at leading order in @xmath3 . note that the denominator in ( [ hp ] ) reflects the renormalization of @xmath92 : the divergent integral @xmath150 does not appear alone , but together with its counter - term , the integral @xmath150 subtracted at @xmath151 .

in an earlier publication , we have introduced a method to obtain , at large @xmath0 , the effective action for @xmath1-dimensional manifolds in a @xmath0-dimensional disordered environment . this allowed to obtain the functional renormalization group ( frg ) equation for @xmath2 and was shown to reproduce , with no need for ultrametric replica symmetry breaking , the predictions of the mzard - parisi solution . here we compute the corrections at order @xmath3 . we introduce two novel complementary methods , a diagrammatic and an algebraic one , to perform the complicated resummation of an infinite number of loops , and derive the @xmath4-function of the theory to order @xmath3 . we present both the effective action and the corresponding functional renormalization group equations . the aim is to explain the conceptual basis and give a detailed account of the novel aspects of such calculations . the analysis of the frg flow , comparison with other studies , and applications , e.g. to the strong - coupling phase of the kardar - parisi - zhang equation are examined in a subsequent publication .

introduction @xmath3 expansion of the effective action: general formula review of the results for @xmath2 corrections in @xmath3, via the graphical method

the recent beam energy scan ( bes ) program @xcite at the relativistic heavy ion collider ( rhic ) pursues one of the major goals of heavy - ion collision experiments : to explore the qcd phase diagram and search for the phase boundary between the normal nuclear matter and quark - gluon plasma ( qgp ) . the bes program at rhic together with pb+pb collisions at the large hadron collider ( lhc ) provide us with a unique opportunity to study systematically the collision energy dependence of a large number of relativistic heavy - ion collision observables . here we study the collision energy dependence of charged hadron transverse momentum spectra and elliptic flow coefficients @xcite , using ( 2 + 1)-d viscous hydrodynamics coupled with a modern lattice qcd based equation of state @xcite . our work focuses on qualitative tendencies rather than quantitative comparison with experimental data . it makes simplifying assumptions that are justified at high energies but gradually break down at lower @xmath7 : longitudinal boost invariance , an equation of state for matter with zero net baryon density , and a purely hydrodynamic approach with constant specific shear viscosity , including for the dilute hadronic rescattering stage which at lower energies occupies an increasing fraction of the fireball s dynamical history and should be described microscopically @xcite . these limitations can be cured in future work ; they are expected to modify our conclusions quantitatively but not qualitatively . * evolution of charged hadron multiplicity and total elliptic flow : * in figs.[fig3]a , b , we show the differential charged hadron elliptic flow for au - au or pb - pb collisions at 20 - 30% centrality . the differential elliptic flow is affected by both total elliptic and radial flow . with mc - glauber initial conditions the differential elliptic flow for @xmath9gev/@xmath10 remains almost unchanged for @xmath11@xmath1gev . as the collision energy increases , both radial and elliptic flow increase , due to the longer fireball lifetime . naively the increasing total elliptic flow should also lead to a larger differential @xmath12 , but this tendency is counteracted by the growing radial flow which blueshifts the momentum anisotropy to larger @xmath13 . for the runs with mc - kln initial conditions we use a larger @xmath14 value . the resulting larger viscous effects suppress the total elliptic flow at lower collision energies more strongly than for the mc - glauber runs , leading to a monotonous decrease of the slope of the differential @xmath15 with decreasing collision energy . to further illustrate this point we plot in fig . [ fig3]c , d the @xmath7-dependence of @xmath16 at 5 fixed @xmath13 points . in this representation one sees that for the mc - glauber model with @xmath17 , @xmath18 at any fixed @xmath19gev/@xmath10 features , as a function of @xmath7 , a very broad maximum around top rhic energy ( 200@xmath1gev ) . for lower @xmath20gev/@xmath10 , heavier particles , or smaller @xmath14 this maximum moves towards lower @xmath7 . with the larger @xmath21 used in the mc - kln model , the strong reduction of @xmath18 at low collision energies shifts the maximum of @xmath18 at any fixed @xmath13 towards higher @xmath7 ; fig . [ fig3]d shows that for @xmath21 this observable does not peak below the top lhc energy , except for very small @xmath22mev/@xmath10 . -dependence of the final spatial eccentricity @xmath23 of the isothermal kinetic freeze - out surface at @xmath24mev , for 10 - 30% centrality . the initial eccentricity is 0.26 for the mc - glauber model and 0.32 for the mc - kln model . the experimental points indicate preliminary data @xcite from an azimuthal hbt analysis by the star collaboration . ] in fig . [ fig4 ] , we show the final fireball eccentricity calculated along the kinetic freeze - out surface , @xmath25mev , as a function of collision energy . as the collision energy increases , the final spatial eccentricity @xmath23 decreases monotonically for both mc - glauber and mc - kln models . this is because at higher collision energy the fireball has longer lifetime to decompress its original deformation and become more isotropic . in fig.[fig4 ] , we compare our results with recent star data from an azimuthal hbt analysis @xcite . our calculations qualitatively agree with the experimental data and reproduce the trend of the collision energy dependence of @xmath23 . mc - glauber runs with @xmath17 quantitatively reproduce the data at @xmath26gev while underpredicting the final eccentricity by @xmath2710% at lower energies . mc - kln initial conditions with @xmath21 result in 15 - 20% larger final eccentricities , due to the @xmath2720% larger initial eccentricities of the mc - kln profiles . within the explored range , we found weak sensitivity of these curves to @xmath14 . extending our calculations to lhc energy we predict that @xmath23 will approach zero around @xmath28 - 5.5@xmath1tev . * acknowledgments : * we thank christopher anson and mike lisa for stimulating discussions.this work was supported by the u.s . department of energy under grants no . de - sc0004286 and ( within the framework of the jet collaboration ) de - sc0004104 . l. kumar _ et al . _ ( star collaboration ) , nucl . * a862-a863 * , 125 ( 2011 ) . s. shi _ et al . _ ( star collaboration ) , arxiv:1111.5385 [ nucl - ex ] . see also related presentations by the star and phenix collaborations in this volume . c. shen and u. heinz , phys . c * 85 * , 054902 ( 2012 ) [ arxiv:1202.6620 [ nucl - th ] ] . p. huovinen and p. petreczky , nucl . * a837 * , 26 ( 2010 ) . c. shen , u. heinz , p. huovinen and h. song , phys . c * 82 * , 054904 ( 2010 ) . h. song , s. a. bass , u. heinz , t. hirano and c. shen , phys . lett . * 106 * , 192301 ( 2011 ) ; and phys . c * 83 * , 054910 ( 2011 ) . s. s. adler _ et al . _ [ phenix collaboration ] , phys . rev . c * 71 * , 034908 ( 2005 ) [ erratum - ibid . c * 71 * , 049901 ( 2005 ) ] [ nucl - ex/0409015 ] . k. aamodt _ et al . _ [ alice collaboration ] , phys . lett . * 106 * , 032301 ( 2011 ) [ arxiv:1012.1657 [ nucl - ex ] ] . c. shen , u. heinz , p. huovinen and h. song , phys . c * 84 * , 044903 ( 2011 ) . c. anson _ ( star collaboration ) , j. phys . g * 38 * , 124148 ( 2011 ) .

we present a systematic hydrodynamic study of the evolution of hadron spectra and their azimuthal anisotropy from the lowest collision energy studied at the relativistic heavy ion collider ( rhic ) , @xmath0@xmath1gev , to the highest energy reachable at the large hadron collider ( lhc ) , @xmath2@xmath1gev @xcite . the energy dependence of the flow observables are quantitatively studied for both the monte - carlo glauber and monte - carlo kharzeev - levin - nardi ( mc - kln ) models . for mc - glauber model initial conditions with @xmath3 , the differential charged hadron elliptic flow @xmath4 is found to exhibit a very broad maximum in the region @xmath5@xmath1gev . for mc - kln initial conditions with @xmath6 , a similar `` saturation '' is not observed up to lhc energies . we emphasize that this `` saturation '' of elliptic flow arises from the interplay between radial flow and elliptic flow which shifts with @xmath7 depending on the fluid s viscosity . by generalizing the definition of spatial eccentricity to isothermal hyper - surface , we also calculate @xmath8 on the kinetic freeze - out surface at different collision energies .

introduction results and discussion

the structure and reactions of neutron - halo " nuclei , weakly bound systems that display a very diffuse surface of nearly pure neutron matter at densities far below that of normal nuclear matter , have recently been subjects of intense study @xcite . in a highly simplified picture , the longitudinal momentum distribution of fragments from the breakup of a loosely bound projectile directly reflects the internal momentum distribution of the valence nucleon and hence the square of the fourier transform of its wave function . thus , halo formation in such loosely bound nuclei can be investigated by measuring the momentum distribution of the fragment from a breakup reaction . the wide spatial dispersion characteristic of a halo neutron translates into a narrow momentum distribution . in this model , one treats the core of the nucleus as an inert spectator to the breakup process . while early experiments were based on the assumption that the measured momentum distribution represented a single reaction channel , it has recently become possible to go beyond this naive approach by measuring @xmath3-radiation from the decay of excited states of the core in coincidence with the breakup fragments @xcite . this technique provides a further benefit in that partial cross sections for the excitation of different final states of the core can be measured , together with the longitudinal momentum distributions associated with inert and ` active ' cores . this allows for spectroscopic investigation of both the core nucleus and the valence nucleon(s ) . the method , originally developed in a search for proton halo nuclei @xcite , has also been applied to the neutron case @xcite . among the candidates for neutron halo formation , the @xmath1b nucleus is of particular interest as the lowest - mass bound system amongst the n=9 isotones . the well - known halo nuclei @xmath5be , @xmath5li , @xmath1be , @xmath6b , and @xmath7c occupy a similar position for n=7 , n=8 , n=10 , n=12 , and n=13 , respectively ( see refs . @xcite and references therein for a discussion of @xmath7c structure ) . moreover , @xmath1b is an odd - odd nucleus and thus different from any other neutron - halo system observed to date . in the first attempt to probe the structure of @xmath1b , bazin , _ et al._@xcite reported a broadening of the longitudinal momentum distribution when compared with the shape expected from theoretical predictions , and suggested that this difference might be due to a strong contribution from core excitation . as a result , they were unable to draw any firm conclusion about halo formation in this nucleus . in order to investigate the structure of @xmath1b , and to determine the role that the core plays in the dissociation reaction , the momentum distributions of @xmath2b fragments corresponding to removal of neutrons from different orbitals in @xmath1b were measured using the technique described in ref.@xcite . the contribution from core excitation was isolated by measuring the momentum distribution of excited @xmath2b fragments in coincidence with @xmath3-rays from their decay . this paper is divided into the following sections : the experimental setup and the procedure are described in section ii , while the experimental results and the analysis of the longitudinal momentum distributions is presented in section iii . finally , a summary is given in section iv . the experiment was performed at the national superconducting cyclotron laboratory ( nscl ) at michigan state university . the @xmath1b radioactive secondary beam was produced by fragmentation of an 80 mev per nucleon primary @xmath8o beam on a 790 mg/@xmath9 be target . the secondary @xmath1b beam , having an energy of 59.2 @xmath10 0.3 mev per nucleon , was then selected by the a1200 fragment separator @xcite and transmitted to a target chamber where a neutron was removed during an interaction with a be or au target , leaving the @xmath2b core in its ground state or in an excited state . the @xmath3-rays from the decay of @xmath2b excited states were detected by an array of 38 cylindrical nai ( tl ) detectors @xcite which completely surrounded the target . recoiling @xmath2b core fragments were momentum - analyzed using the s800 spectrograph @xcite . the momentum acceptance of the spectrograph was 6@xmath11 and the angular acceptances were @xmath12 and @xmath13 in the dispersive and non - dispersive directions , respectively . these acceptances resulted in an estimated 99@xmath11 efficiency for detecting events corresponding to the ground - state momentum distribution , and 95@xmath11 for the momentum distributions of the excited states , as determined from the observed shapes . this point is discussed further below . the spectrograph was operated in dispersion - matched mode , in which the intrinsic dispersion of the secondary beam ( 0.5@xmath11 ) is compensated by the last section of the spectrometer . the targets of be and au were 228 mg/@xmath9 and 256 mg/@xmath9 thick , respectively , chosen to produce an energy straggling of about the same magnitude as the resolution of the spectrograph . the overall resolution ( including target thickness effects ) was measured to be 9 mev / c full width at half maximum ( fwhm ) in a direct - beam run with the target in place . this run also gave the normalization for the beam flux , determined to be @xmath14 particles / s . time - of - flight information ( over a distance of 70 m ) combined with the energy loss and total energy signal obtained with a segmented ion chamber and a 5 cm thick plastic scintillator , respectively , were used to identify and measure the yields of the fragments in the focal plane of the spectrograph . two @xmath15 position - sensitive cathode - readout drift chambers in the focal plane were used to determine the momentum and angle information of the fragments . an energy spectrum of @xmath3-rays in coincidence with @xmath2b fragments is shown in fig . this spectrum has been transformed into the projectile rest frame using the position of the incident @xmath3-ray measured in the nai array , which allows for doppler correction on an event - by - event basis . the solid curve is a fit to the data obtained via monte - carlo simulation using the code geant @xcite . the simulation took into account doppler broadening , the distortion of the shapes caused by the back transformation to the projectile rest frame , and the calculated @xmath3-ray detection efficiencies . experimental efficiencies measured with calibrated radioactive sources agreed with the simulation to within 5@xmath11 . the background , which presumably results from a combination of breakup reactions leaving the target in an excited state and secondary interactions of neutrons with the detector and surrounding materials , was parameterized by a simple exponential dependence . taking into account the 3.48 mev , 3.68 mev and 4.13 mev transitions in @xmath2b , the fit to the experimental spectrum is excellent ( the @xmath16 per degree of freedom n is 1.08 ) . the choice of these particular @xmath3-ray transitions is discussed below . eliminating the 4.13 mev @xmath3-ray from the fit causes @xmath16/n to increase to 1.10 for 170 degrees of freedom , which is marginally significant . this indicates that all the important transitions were accounted for and that there are no other significant @xmath3-rays between 1.5 and 3.5 mev or above 4.2 mev . the level scheme of @xmath2b @xcite is presented in fig . 2 . the ground state has j@xmath17 , resulting from a [ @xmath18@xmath19 proton configuration . below 4.5 mev , three negative parity and two positive parity excited states have been identified . although the energies and parities of these states are well established from particle - transfer reaction studies , very little is known about their spins . based on the @xmath3-ray spectrum of fig . 1 , we are interested in the group of states between 3.48 and 4.13 mev . in the simplest model , the j@xmath20 ground state of @xmath1b results from the coupling of a @xmath21-shell neutron to the @xmath2b ground state . then , the neutron removal from @xmath1b could proceed by the removal of the @xmath22-shell neutron to leave @xmath2b in its ground state , by removal of a neutron from the @xmath23 orbital leaving @xmath2b in an excited state with j@xmath24 , or by removal of a neutron from the @xmath25 orbital leaving @xmath2b in a more highly excited state with j@xmath26 , or @xmath27 . this simple model is confirmed by shell - model calculations in the @xmath28 model space with the wbt and wbp residual interactions @xcite . calculations were carried out for the @xmath1b ground state with a @xmath29 configuration ( relative to a closed shell for for @xmath30o ) , and for the @xmath2b spectrum with @xmath31 , @xmath32 and @xmath33 configurations . ( the total number of eigenstates are 5 , 299 and 2973 , respectively . ) the low - lying energy levels obtained for these configurations , labeled by ( 0 ) , ( 1 ) and ( 2 ) , respectively , are shown in fig . the spectroscopic factors for the wbt interaction are given in table 1 ( the spectroscopic factors obtained with wbp are similar to those of wbt ) . the higher levels , which are reached by predominantly @xmath25 removal , lie above the neutron decay threshold of 4.9 mev . the spectroscopic factors leading to the ( 2 ) configurations are zero with our configuration assumptions . all of the ( 0 ) , ( 1 ) and ( 2 ) configurations are allowed in the @xmath5b(t , p)@xmath2b reaction which has been previously used to identify the @xmath2b states @xcite . in principle , the ( 0 ) and ( 2 ) configurations can be mixed , which could give some spectroscopic strength leading to negative parity states around 4 mev in excitation . our estimates within the mixed ( 0)+(2 ) model space indicate that the effect of this mixing leads to spectroscopic factors of about 0.02 or less , and the effect of this mixing will be ignored here . partial cross sections obtained from the absolute @xmath3-ray intensities corrected by the computed efficiencies and the spectrometer acceptance are presented in table i. the overall error in these cross sections are of the order of 18@xmath11 , due mainly to systematic errors in the fit to the experimental gamma - ray spectrum , and to the uncertainties in the number of particles in the beam ( 12@xmath11 ) and the target thickness ( 5@xmath11 ) . the calculated cross section for each final state of the core fragment in a one - nucleon knockout reaction is given by : @xmath34 where @xmath35 is the computed spectroscopic factor for the removed nucleon with respect to a given core state and @xmath36 is the reaction cross section for the removal of a nucleon from a single - particle state with total angular momentum @xmath37 . @xmath38 is the sum of the separation energy and excitation energy of the state @xmath39 . the single - particle cross sections were calculated by tostevin @xcite in a eikonal model assuming that there are two reaction mechanisms involved in the knock - out reaction : ( i ) nucleon stripping in which the halo nucleon interacts strongly with the target and leaves the beam , and ( ii ) diffraction dissociation in which the nucleon moves forward with essentially the beam velocity but away from the core . the calculated cross sections for the states below 4.5 mev are given in table i. the best agreement with experiment is achieved with the spin assignments shown there . it should be noted that the evidence for the existence of the 4.13 mev transition in the experimental @xmath3-ray spectrum is marginal . furthermore , the calculated strength for this transition ( table i ) does not take into account the configuration mixing discussed above . overall , it appears that the combination of the theoretical spectroscopic factors and computed single - particle cross sections does a remarkable job of reproducing the experimental yields . the momentum vector of the fragment at the target position after the reaction was reconstructed using the known magnetic fields and the positions and angles at the focal plane , and the ion - optics code cosy @xcite . the momentum components transverse to the beam direction are sensitive to the reaction mechanism , while the longitudinal momentum component is relatively independent of the details of the reaction mechanism as discussed by orr , _ et al . the longitudinal momentum distribution is obtained as the projection of the total momentum in the direction parallel to the beam . from the coincident @xmath3-ray data , it is possible to generate the distribution for the case when @xmath2b is left in its ground state , by subtracting the contribution from the excited states from the singles spectrum after correcting for the @xmath3-ray detection efficiencies . the result of this subtraction process is presented in fig . 3(a ) and corresponds to the removal of a single neutron from the @xmath22 shell in @xmath1b due to a reaction with a @xmath0be target . 3(b ) shows the longitudinal momentum distribution for the core - excited states , which is considerably broader . a more precise comparison can be made through the use of the reaction theory . calculated longitudinal momentum distributions , also shown in fig . 3 , were determined within the framework of the eikonal model following the procedure of ref . @xcite . since the dissociation products are formed at an impact parameter greater than @xmath40 ( where @xmath41 and @xmath42 are the energy - dependent radii for the core and valence nucleon , usually chosen to reproduce the measured interaction cross section ) the black disc approximation can be used . the cross section is then expressed in terms of the impact parameter as the one - dimensional wigner transform of the wave function after the reaction , where the wave function profile is unity outside of a cutoff radius and zero inside . the momentum distribution corresponding to the @xmath2b core in its ground state , shown in fig . 3(a ) , agrees reasonably well with the _ _ = 0 curve of the eikonal model calculation , which represents the removal of an @xmath43-wave valence neutron from the @xmath22 shell in @xmath1b . similar results have been obtained for @xmath5be @xcite and @xmath44c @xcite . the width obtained for the distribution is @xmath45 mev / c fwhm , after correcting for the 9 mev / c experimental momentum resolution . this width results from a lorentzian fit to the experimental data and is in excellent agreement with the value of 56 mev / c fwhm displayed by the theoretical _ = 0 curve in the lab . the disagreement in the tails of the distribution suggests that there may be a contribution from the _ l _ = 2 distribution and possibly other effects due to the acceptance of the spectrometer , approximations in the reaction model used to analyze the data , or unidentified contributions to the line shape @xcite . spectroscopic factors of 0.306 and 0.662 , respectively , are obtained from the shell model calculation @xcite for the removal of a valence neutron from the @xmath46 and @xmath47 states in @xmath1b , which indicates that a non - negligible _ l _ = 2 admixture is expected . individual cross sections for both configurations have been derived by fitting the experimental distribution with a linear combination of _ l _ = 0 and _ l _ = 2 shapes . the best fit , shown in fig . 4 , implies that the removal of a valence neutron from the @xmath47 ( @xmath46 ) orbital in the @xmath1b ground state accounts for 89@xmath48 ( 11@xmath48 ) of the total yield . the agreement with cross sections derived from the eikonal model and the theoretical spectroscopic factors is very good ( table i ) , as is the quality of the fit to the experimental data shown in fig . 4 . the core - excited longitudinal momentum distribution , shown in fig . 3(b ) , was obtained from the @xmath3-ray - coincident @xmath2b yield after subtracting a background distribution derived by gating on the exponential part of the @xmath3-ray spectrum at high energy and normalizing to the extrapolated background under the peaks . the background computed in this way constituted 26@xmath11 of the coincident data in the region of the three identified @xmath3-ray transitions . the core - excited distribution has a shape that agrees well with that expected for _ l _ = 1 , confirming the expectation of the removal of a @xmath23 neutron . the experimental width , obtained from a lorentzian fit to the central part of the distribution , is @xmath49 mev / c fwhm . the calculated width in the lab . frame of the theoretical _ = 1 curve is 144 mev / c . the distributions computed for @xmath50 and @xmath51 wave breakup are broader than the actual momentum acceptance of the spectrometer . a ( small ) correction to the integrated yield due to this limited acceptance has been applied to the experimental cross sections given in table i. the longitudinal momentum distribution obtained in the one - neutron breakup reaction of @xmath1b on a @xmath4au target is shown in fig . the @xmath3-ray coincident yield has been subtracted using the same procedure as described in the previous section . in this case , the @xmath3-ray spectrum was essentially equal to the exponential background distribution , and no core - excited transitions were observed . the width obtained from a gaussian fit ( dashed curve in fig . 5 ) to the experimental longitudinal momentum distribution is @xmath52 mev / c ( after correcting for the experimental resolution ) . this width is larger than the value of @xmath53 mev / c measured by bazin , _ et al . _ @xcite for coulomb breakup on a tantalum target at 86 mev / a . however , it agrees within the errors with that extracted from the @xmath2b ground - state longitudinal momentum distribution for the be target in the present work . the integrated cross section derived for coulomb breakup of @xmath1b on @xmath4au ( leaving the @xmath2b core fragment in the ground - state ) is @xmath54 mb , much larger than the yield quoted in ref.@xcite . it was suggested there that the smaller cross section reflected a quenching of the soft dipole strength that implied normal " nuclear structure for @xmath1b . however , a possible experimental problem due to the restricted angular acceptance of the a1200 spectrometer was acknowledged . the present observations of a broader momentum distribution coupled with a larger cross section suggest that this was indeed the case . the coulomb breakup cross section was calculated using a yukawa potential with finite - size corrections @xcite . the result , 543 mb , is less than our measurement . however , the yield calculated in ref . @xcite using a woods - saxon potential was 50@xmath11 larger than that from the yukawa form . our data suggest that the actual situation is intermediate between these two approximations . the coulomb contribution to our be - target data was also computed and found to be negligibly small ( 2.3 mb ) . the solid curve in fig . 5 shows the longitudinal momentum distribution for the au target predicted in the model of ref . @xcite , normalized to the experimental data . the width of this lorentzian function compares well with experiment , but the agreement in the tails of the distribution is unsatisfactory . zhongzhou , _ et al . _ @xcite have predicted an inversion of the @xmath46 and @xmath47 orbitals for @xmath1b within the framework of a non - linear relativistic mean - field calculation . this is in agreement with the shell - model calculation cited above and with our observation of a dominant _ l _ = 0 component in the ground state of this nucleus . their calculation shows that , under these conditions , @xmath1b is a neutron halo nucleus . on the other hand , measurements of the interaction cross sections at relativistic energies for particle - stable b isotopes ranging from mass 8 - 15 , by tanihata , _ et al . _ @xcite , suggested that their effective rms radii are practically constant , which does not support the halo hypothesis . measurements at intermediate energy , analyzed by liatard , _ et al . _ @xcite , gave larger rms radii in all cases than those of ref . @xcite , together with a significant mass dependence and an anomalously large radius for @xmath1b . the radius difference is not by any means as large as in the case of @xmath5li , for example , but this is not unexpected given the fact that the valence neutron is more tightly bound in @xmath1b . of course , the rms radius extracted from an interaction cross section is not the only signature for halo structure @xcite , and the narrow longitudinal momentum distribution and large cross section for dissociation of the ground state of @xmath1b observed in the present experiment strongly argue for halo structure in this system . in this experiment , we have measured the longitudinal momentum of the @xmath2b core fragment in one - neutron knockout from @xmath1b , on both @xmath0be and @xmath4au targets . the contribution from core excitation was isolated by measuring the excited @xmath2b fragments in coincidence with @xmath3-rays from their decay . comparison of the observed intensities of core - excited @xmath3-ray transitions with a shell - model calculation , using an eikonal reaction theory , allowed us to make spin assignments for two positive - parity excited states in @xmath2b . the longitudinal momentum distributions in coincidence with the positive - parity core - excited transitions are consistent with knockout from a @xmath25 state , as expected . an excellent fit to the distribution obtained when the @xmath2b core remains in its ground state is also provided by the shell - model calculation , which predicts an admixture of @xmath47 and @xmath46 neutron knockout . the experimental data imply an 89@xmath48 _ l _ = 0 component , in agreement with the shell - model result . the coulomb breakup cross section , measured with the @xmath4au target , is consistent with expectations for a weakly - bound @xmath47 neutron . the width of the longitudinal momentum distribution obtained with the @xmath4au target ( @xmath52 mev / c ) agrees with that obtained using the @xmath55be target after the core - excited component is subtracted . the results of the present experiment lend very strong support to the idea that @xmath1b is a neutron - halo " system , the first odd - odd nucleus to display this structure . it also appears that , with the sole exception @xcite of @xmath6c at n=11 , all of the lowest - mass , particle - stable isotones from n=7 - 13 are halo nuclei . it therefore seems that re - investigation of the structure of @xmath6c is in order . the first author ( v.g . ) was financially supported by fapesp ( fundao de amparo a pesquisa do estado de so paulo - brazil ) while on leave from the unip ( universidade paulista ) . this work was supported by the national science foundation under grants no . phy94 - 02761 , phy99 - 01133 , phy95 - 14157 , and phy96 - 05207 . b. m. sherrill ( to be published ) ; j.a . caggiano , ph.d . thesis , michigan state university ( 1998 ) , unpublished ; j. yurkon , d. bazin , w. benenson , d.j . morrissey , b.m . sherrill , d. swan , and r. swanson , nucl . a422 * , 291 ( 1999 ) . .partial cross sections in mb for each of the final states populated in @xmath2b . the spin assignments for each level ( third column ) are discussed in the text . the fourth and fifth columns are the orbital angular momentum of the valence neutron and the corresponding spectroscopic factor , respectively . the next three columns give the partial cross sections for stripping and diffractive breakup in the eikonal model of ref . @xcite , and the theoretical value for the cross section which is the sum of the two components multiplied by the corresponding spectroscopic factor . the experimental value is given in the last column . [ cols="^,^,^,^,^,^,^,^,^",options="header " , ]

the single - nucleon knockout reactions @xmath0be(@xmath1b,@xmath2b+@xmath3)x and @xmath4au(@xmath1b,@xmath2b+@xmath3)x , at an incident energy of 60 mev per nucleon , have been used to probe the structure of @xmath1b and of the core fragment @xmath2b . a dominant _ 2s _ configuration is deduced for the neutron in the ground state of @xmath1b . the longitudinal momentum distribution for this state is consistent with neutron halo " structure . spin assignments for @xmath2b excited states at 3.48 and 3.68 mev are proposed based on the observed spectroscopic factors for one - neutron removal . 2

introduction experiment experimental results summary acknowledgments

cavities with diameters from several to tens of kiloparsecs have been observed in the x - ray emission from nearly two dozen galaxies , groups , and clusters . @xcite these cavities are filled with magnetic fields and relativistic plasmas that radiate in radio emission from `` radio lobes.''@xcite these observations suggest that the x - ray cavities have formed by shoveling aside thermal cluster plasmas by radio - emitting plasmas emanating from galaxies . the interactions of radio - emitting outflows with x - ray emitting cluster plasmas lead to shocks , which are a candidate for heating cluster plasmas to @xmath0.@xcite past theoretical models of these systems assume that such outflows are kinetic energy dominated ( so - called kinetic energy dominated regime ) . @xcite however , recent observations show that both cluster and radio lobe plasmas have appreciable magnetic energy . @xcite this has led to new models in which radio lobes are thought to be gigantic relaxed " plasmas with kilo - to - megaparsec scale jets providing a source of magnetic energy and helicity from the galaxy to the lobes . @xcite however , the details of how radio lobe magnetic energy and helicity evolve and interact with the intergalactic medium are not well understood . @xcite these details depend on underlying nonlinear plasma physics , including magnetic relaxation of radio lobe plasmas as they expand against a background plasma while being driven by jets , heating of the lobe and background plasmas due to reconnection and shocks , and angular momentum transport within the lobe and between the lobe and background . in order to develop further insights into extragalactic radio lobes , a laboratory plasma experiment called the plasma bubble expansion experiment ( pbex ) @xcite has been built to address some of the underlying nonlinear plasma physics issues upon which leading radio lobe models are based . the experiment will study the related model problem of a magnetic plasma bubble " relaxing and expanding into a lower pressure weakly - magnetized background plasma . a new pulsed coaxial gun will form and inject magnetized plasma bubbles ( _ i.e. _ , the lobe ) into a background plasma ( _ i.e. _ , the intergalactic medium ) formed by a helicon and/or hot cathode source on the helcat facility.@xcite experimental parameters can be adjusted so that important dimensionless parameters , such as plasma @xmath1 , are relevant to the astrophysical context . numerical modeling helps guide the experiments and aids the data interpretation . in this paper we report initial nonlinear simulation results performed with a new three - dimensional ( 3d ) ideal mhd package,@xcite which is a time - explicit , compressible , ideal mhd parallel 3d code , using high - order godunov - type finite - volume numerical methods , in cartesian coordinates @xmath2 . the simulations mimic pbex and use experimentally measured or inferred parameters ( see table [ normalization ] ) . this paper is organized as follows . in sec . [ setup ] , we outline the problem setup including initialization of the bubble and background plasma column . we present the simulation results in sec . [ results ] , and discussions and implications of our results for the experiment are given in sec . [ discussions ] . in the simulations and experiment , a high density magnetized rotating bubble plasma is injected radially into a cylindrical plasma volume with a background magnetic field , as shown in fig . [ config ] . the injected magnetic configuration is not force - free so that lorentz forces cause the bubble to expand while traveling through and interacting with the background plasma . the basic model assumptions and numerical treatments we adopt here are essentially the same as those in @xcite the nonlinear system of time - dependent ideal mhd equations in 3d cartesian coordinates @xmath2 is given here : @xmath3&=&0\,,\\ \frac{\partial \vec{b}}{\partial t}-\nabla\times(\vec{v}\times\vec{b})&=&0\,,\end{aligned}\ ] ] in which @xmath4 , @xmath5 , @xmath6 , @xmath7 and e are the density , ( gas ) pressure , flow velocity , magnetic field , and total energy , respectively . @xmath8 is the unit diagonal tensor . the total energy is @xmath9 , where @xmath10 is the ratio of the specific heats . note that a factor of @xmath11 has been absorbed into the scaling for both the magnetic field @xmath7 and current density @xmath12 . more details are given in @xcite all simulations are performed on the parallel linux clusters at los alamos national laboratory . it should be noted that the details of effects such as reconnection and heat evolution can not be addressed accurately due to the ideal mhd model and the use of a simplified energy equation . physical quantities are normalized by the characteristic system length scale @xmath13 , density @xmath14 , and velocity @xmath15 based on the measured or expected values from pbex . the normalization factors are summarized in table [ normalization ] . normalized variables are used hereafter . a higher pressure magnetic plasma bubble with spherical radius @xmath16 , centered initially at @xmath17 , @xmath18 and @xmath19 , is injected along the @xmath20 axis into a lower pressure background plasma with injection velocity @xmath21 ( see fig . [ config ] ) . the stationary background plasma is composed of a cylindrical plasma column with radius @xmath22 confined by a background magnetic field @xmath23 , where @xmath24 . although pbex will offer a choice of gas combinations for the bubble and background , the initial experiments will likely use argon for both , and therefore the initial simulations are based on argon with atomic mass of @xmath25 . force balance of the background plasma along the @xmath26 direction gives @xmath27 for @xmath28 , @xmath29 , where @xmath30 is taken to be @xmath31 ( @xmath32 ) , while for @xmath33 , @xmath34 is determined by the initial pressure profile : @xmath35 the background plasma number density and temperature profiles are given by the following functions : @xmath36 which are a good fit to actual experimental data taken by a langmuir probe on pbex , where the typical @xmath37 and @xmath38 are taken to be @xmath39 and @xmath40 , respectively . thus , the initial pressure profile of the background plasma is @xmath41 $ ] . a higher pressure magnetized plasma bubble is generated and injected by a coaxial gun source . it is well established empirically in coaxial gun spheromak experiments that , under proper conditions , a spheromak magnetic bubble " will be formed by the gun discharge.@xcite in the simulations reported here , the bubble structure is similar to the one given in @xcite the number density profile of the bubble plasma with radius @xmath16 is given by @xmath42,\ ] ] up to a normalization coefficient @xmath43 and a uniform temperature @xmath44 , where @xmath45 and @xmath46 ( see fig . [ config ] ) . the density profile used here has its peak shifted from the center of the bubble , approximating a spheromak , and is therefore different from the uniform density profile used in @xcite the bubble magnetic field is determined by three key quantities : the length scale of the bubble magnetic field @xmath47 , the amount of poloidal flux @xmath48 , and the index @xmath49 , which is the ratio of the bubble toroidal to poloidal magnetic fields . for simplicity , the bubble magnetic field @xmath50 is also assumed to be axisymmetric . the poloidal flux function @xmath48 is specified as : @xmath51\,.\ ] ] the poloidal fields , up to a normalization coefficient @xmath52 ( @xmath53 ) , are : @xmath54 while the toroidal magnetic field is @xmath55\,.\ ] ] the azimuthal component of the bubble lorentz force is zero , but the total azimuthal lorentz force due to the _ combined _ fields and currents of the bubble and the background plasma may be non - zero . the bubble also has uniform injection velocity @xmath21 and uniform rotation angular speed @xmath56 , where @xmath57 . please note that this is a strong rotation , possibly having strong influence on the stability of the bubble ( sec.[stability ] ) and the expansion of the bubble in the @xmath58-@xmath59 plane ( sec.[internal ] ) . the total computational domain is @xmath60 , @xmath61 , and @xmath62 , corresponding to a @xmath63 box in actual length units . the numerical resolution used here is @xmath64 , where the grid points are assigned uniformly in the @xmath65 , @xmath66 , and @xmath67directions . a cell @xmath68 @xmath69 ) corresponds to @xmath70 . we use outflow " boundary conditions at every boundary , _ i.e. _ , setting all values of variables in the ghost zones equal to the values in the corresponding active zones , which is the simplest approach possible . this technique is accurate for supersonic outflow but not for subsonic outflow . this simplified boundary condition limits our ability to predict the transit time , the time for the bubble to travel through the background plasma , and to study the detachment problem , _ i.e. _ , under what conditions the bubble would separate from the wall boundary . more accurate boundary conditions will be implemented in future work . here , we focus on the interaction of the bubble plasma with the background plasma before the structures have reached the boundaries . in this section we present ideal mhd simulation results on the nonlinear evolution of a magnetic bubble " plasma propagating and expanding into a lower pressure background plasma . the results are organized into three primary topics : ( 1 ) global evolution of the bubble - background system and interface , ( 2 ) internal bubble evolution , and ( 3 ) angular momentum transport both outside and inside the bubble . " key findings include the formation of both an mhd shock and a reverse mhd slow - mode wavefront as a result of the bubble propagating into the background plasma , and the outward transfer of azimuthal angular momentum inside the bubble due to advection and inhomogeneous transport outside the bubble due to the background magnetic field . please note that all physical quantities , such as the magnetic field @xmath7 and flow velocity @xmath6 , presented in this section are the total value due to both the bubble and background plasmas . in this subsection , we examine the evolution of the global bubble - background system and the interface between the two plasmas . here we discuss the time evolution of the magnetic bubble , showing selected physical quantities using 2-d @xmath58@xmath20 slices at @xmath71 . the density distributions at various times @xmath72 are shown in fig . [ overview ] with @xmath73 and injection velocity @xmath74 . in this example @xmath73 corresponds to the bubble having a minimum initial lorentz force ( see discussion below ) . at @xmath75 , we see that the initial peak - shifted high density magnetic bubble has been transformed into a crab " ( due to the low @xmath49 , see discussion below ) , bounded by one mhd shock ( see sec . [ shock ] ) and one reverse slow - mode compressible mhd wavefront ( sec . [ wavefront ] ) . low - density cavities ( a factor of dozens of times of magnitudes smaller than the peak density ) exist both between the shock and wavefront and in the post - wavefront region . at @xmath76 , the shock has reached the other side of the computation domain , while the wavefront is located at @xmath77 . the bubble is still in the middle of the background plasma . the simulation after @xmath76 is not accurate due to the simplified boundary conditions . the value of @xmath49 determines the strength of the initial lorentz force in the bubble and consequently how the bubble expands and evolves . we first test the influence of @xmath49 on the bubble evolution with fixed injection speed @xmath21 . the simulations show that the results are insensitive to @xmath49 except for large @xmath78 when the bubble expands more in the direction of the injection , leading to a growing mushroom " ( fig . [ fig : dependence](left ) ) . with smaller @xmath79 , the bubble expands more transversely to the direction of injection , resulting in a growing crab " ( fig . [ fig : dependence](right ) ) . this can be understood from the initial poloidal lorentz force due to the coupling of the bubble s current and its own field ( eq . 22 and eq . 23 of @xcite ) : larger @xmath49 would have a positive axial lorentz force but negative radial lorentz force , resulting in collimation , while smaller @xmath49 would have a negative axial but positive radial lorentz force , resulting in radial expansion . in the experiment , it is expected that the ejected bubble will quickly reach a nearly force - free state . therefore , hereafter , we assume @xmath73 , corresponding to minimum initial bubble lorentz force . given reasonably low injection velocity @xmath80 , there are always an mhd shock and a reverse slow - mode compressible mhd wavefront , whose structures , evolution , and propagation characteristics are essentially the same . and the injection velocity has little influence on the shock speed and wavefront speed ( see table [ speed ] ) , which implies that the shock and wavefront result from the expansion of the bubble due to the lorentz force , rather than a piston effect " of the bubble propagation . however , the piston " effect could become important with larger injection speeds @xmath81 ( see table [ speed ] ) , which , however , will not occur in the experiment . for the following analysis , we will focus on the case @xmath82 as being representative for the experiment . the expansion of the magnetic bubble generates a leading mhd shock and a trailing reverse slow - mode compressible mhd wavefront . the details of these two structures are presented in this and the next subsections . the characteristics of the shock propagation are similar in the @xmath65 and @xmath67directions . thus , we will present results and analysis in the @xmath67direction only . figure [ structure ] , with @xmath83 , displays several physical quantities along a line with @xmath84 in the @xmath20-direction at @xmath75 . hereafter we define axial direction as the direction along @xmath20-axis , @xmath58-@xmath59 plane as toroidal plane and @xmath58-@xmath20 plane as poloidal plane . several features can be identified . first , an mhd shock can be seen around @xmath85 in the profiles of @xmath4 , @xmath86 [ see fig . [ identification](left ) ] and @xmath87 [ see fig . [ identification](right ) ] . in the vicinity of @xmath88 , since the magnetic field lies in the shock plane and is perpendicular to the shock normal , this shock is identified as a perpendicular shock in this region . this mhd shock is a fast shock whose properties are very close to an ordinary field - free shock . as shown in fig . [ structure](left ) , the magnetic field components @xmath89 and @xmath90 change very little across the shock . it is important to verify that this is indeed an mhd shock by comparing the simulation results to the the shock jump conditions . choosing the velocity frame so that the shock is at rest ( shock velocity is @xmath91 ) and simplifying the notation for the problem , we represent quantities in the upstream region by a @xmath92 superscript and those in the downstream region by no superscript . then there are @xmath93 known quantities @xmath94 , @xmath95 , @xmath96 , @xmath97 , @xmath98 , @xmath99 , @xmath100 and @xmath101 . there are @xmath93 unknown quantities in the downstream region , @xmath4 , @xmath5 , @xmath102 , @xmath103 , @xmath87 , @xmath86 , @xmath89 and @xmath90 . thus we need @xmath93 conditions to specify them . we consider the 1-d ideal mhd shock jump conditions for simplicity . the equations are as follows : @xmath104 a _ matlab _ code was used to solve this nonlinear system of equations , given the values in the upstream region : @xmath97 , @xmath98 , @xmath105 and @xmath94 . the results of @xmath102 , @xmath87 , @xmath86 and @xmath89 matches pretty well ( see table [ jump ] ) . the nonzero simulation values of @xmath103 and @xmath89 result from 3-d effects . the relatively large differences seen in the values of @xmath4 and @xmath5 are possibly due to nonzero numerical diffusion in the simulations . there is an mhd wavefront at @xmath106 , as seen in both panels of fig . [ structure ] , where @xmath86 and @xmath89 have their local minimum , and @xmath4 , @xmath107 and @xmath108 have their local maxima . the nature of this mhd wavefront can be identified by plotting the axial pressure profiles along the line @xmath84 at @xmath75 , as shown in fig . [ reverse ] . a transition occurs near @xmath106 , where an increase in gas pressure @xmath5 is accompanied by a decrease in magnetic pressure @xmath109 . the transition is identified as a reverse slow - mode compressible mhd wavefront . between the shock ( @xmath85 ) and the wavefront @xmath110 ) , the magnetic field lines are compressed and some thermal energy has been converted into magnetic energy . therefore the gas pressure has an abrupt decrease while the magnetic pressure increases rapidly between the shock and wavefront ( see fig . [ reverse ] ) . from fig . [ reconnection ] , it can be seen that anti - parallel reconnection of the @xmath86 component occurs around @xmath111 , facilitated by numerical diffusion although the simulation is performed with an ideal mhd code . it is possible that @xmath89 component reconnection takes place around @xmath112 , where the bubble field is not exactly anti - parallel to the background field ( more details in sec . [ balance ] ) . the reconnection is driven by magnetic field line compression due to the reverse slow - mode compressible mhd wavefront . from fig . [ current_z ] , two strong toroidal current sheets are observed at the reconnection layer and the mhd shock . these are also due to the compression of the magnetic field lines due to the wavefront and shock , respectively . the reconnection and the shock / wavefront convert normal velocity into tangent velocity and convert kinetic energy into thermal energy . because this is an ideal mhd simulation , the details of the reconnection are not expected to be accurate . we are only interested here in the qualitative effects of the bubble evolution due to reconnection . the evolution and propagation of the magnetic bubble can be further understood by examining the various forces along @xmath84 at @xmath75 , which are displayed in fig . [ forces ] . the mhd shock breaks the initial background equilibrium . the passage of the shock wave heats the gas and alters its pressure gradient . the axial flow is pushed forward by both the gas pressure gradient and lorentz force at the mhd shock while it is dragged back behind the shock , resulting in an axial deceleration of the gas in the postshock region . although the lorentz force tries to accelerate the axial flow at the mhd wavefront , the gas pressure holds it back . therefore the mhd shock will be driven forward and eventually separated from the wavefront , which leads to a cavity of depleted density between the shock and the wavefront . figure [ forces2 ] displays axial profiles of various forces along @xmath113 and @xmath114 at @xmath75 . the locations are chosen to be where anti - parallel reconnection ( left panel of fig . [ forces2 ] ) occurs and its reflection about @xmath115 ( right panel of fig . [ forces2 ] ) . this figure clearly shows the difference between the two locations : the lorentz force changes sign ( similar to fig . [ forces ] ) on the left hand side while keeping the same sign ( negative ) on the right hand side , which is consistent with reconnection happening on the left hand side and not on the right hand side . at both @xmath58 locations , the toroidal current densities @xmath116 are much larger than the poloidal current densities ; however , the toroidal field component @xmath89 is close to zero . therefore @xmath117 is much bigger than @xmath118 . anti - parallel field lines on the left hand side result in a sign - change of the axial lorentz force , while the lorentz force on the right hand side does not change sign since @xmath86 does not change sign there . the sign change of the lorentz force is necessary for reconnection since this is the driving force to pull the field lines from either side of the current sheet together to reconnect . also , the total force is more negative on the right hand side compared to the left hand side . this means that the axial flow in the right hand side is slowed down more quickly than the left hand side , which leads to the asymmetry of the shock propagation across the @xmath58 axis . the magnetic bubble evolves into a nearly quasi - force - free state ( see fig . [ lorentz ] ) , with a lorentz force that scales in time roughly as : @xmath119 the time scale of this relaxation @xmath120 is dependent on the value of @xmath49 , which determines the amplitude of the initial lorentz force . larger initial lorentz force leads to quicker relaxation . for @xmath79 , @xmath121 ; for @xmath73 , @xmath122 ; and for @xmath78 , @xmath123 . the relaxation happens on the order of the sweet - parker reconnection time @xmath124 ( see definitions and estimates of @xmath125 and @xmath126 in sec . [ discussions ] ) . for example , @xmath127 for @xmath73 , which is similar to the lorentz force relaxation time of 0.386 . however , it should be noted that the lorentz force in the shock / wavefront is always significant ( fig . [ forces ] ) . in this subsection , we examine the evolution and properties of the bubble itself , including a simple kink stability analysis . density ( fig . [ rho ] ) and fluid velocity vector ( fig . [ velocity ] ) plots in the @xmath128 plane at different times both demonstrate that the initial fast - rotating spheromak - like magnetic bubble evolves into a much larger slow - rotating , fast - expanding elliptical structure with maximum density reduced by @xmath129 times , while the density at the wavefront and the shock increases by @xmath93 and @xmath130 times , respectively . the center of the bubble shifts away from the original propagation axis with @xmath84 . figure [ field_xy ] displays the vector magnetic field plots in the @xmath128 plane at different times . from the figure , we can see that the bubble field still keeps a spheromak - like configuration and the expansion of the bubble pushes away the background fields , thus compressing them . some reconnection happens in the region @xmath131 because the bubble toroidal field component is opposite to the background field there . these figures show that the background field breaks the symmetry of the system , which is consistent with the results of sec . [ balance ] and sec . [ transport ] it is worth noting that the initial expansion of the bubble in @xmath58-@xmath59 plane results from the non - free initial lorentz force as well as the centrifugal force due to the strong initial rotation of the bubble , although the latter quickly slows down to a small value because of the conservation of angular momentum associated with the initial quick expansion and the possible kelvin - holmhotz instability associated with the initial strong toroidal velocity shear . spheromak - like bubble plasmas are subject to current - driven kink instabilities . figure [ jz ] shows a snapshot of the axial current density @xmath132 at @xmath75 . the axial current flow follows a semi - closed ( it will close outside the out - flowing boundary ) circulating path , flowing along the central axis ( the forward " current ) and returning along the bell - shaped path on the outside ( the return " current).@xcite fig . [ field ] shows a snapshot of the configuration of the magnetic field @xmath7 , which indicates that a tightly wound central helix is overlapped with the forward " current , and a loosely wound helix is overlapped with the return " current . given a helical magnetic field , this axial current - carrying cylindrical plasma column is subject to a current - driven instability ( cdi).@xcite however , we do not see any visible evidence of any current - driven instability in this case ( @xmath73 ) . the well - known kruskal - shafranov criterion @xcite for mhd kink instability in cylindrical geometry can be written as : @xcite @xmath133 where @xmath134 is the safety factor , @xmath135 is the total poloidal magnetic flux , @xmath136 is the total axial current , @xmath137 and @xmath138 are the column radius and length , and @xmath90 is the axial field component . safety factors less than @xmath139 are unstable to the cdi kink mode . the safety factor in this case ( fig . [ jz ] and fig . [ field ] ) is @xmath140 at @xmath75 , which is bigger than @xmath139 . therefore it is expected to be cdi stable , which is consistent with the simulation results . the simulation with @xmath78 gives @xmath141 at @xmath142 , which is marginally unstable to cdi according to eq . however we do not find evidence of unstable cdi modes in this case either . line tying " ( important in the experiment , not present in the simulations reported here due to the outflowing " boundary conditions used in this paper ) @xcite and other stabilization effects such as dynamic relaxation " , @xcite internal strong rotation @xcite and external gas pressure , _ etc . _ , could raise the stability threshold . a more detailed stability analysis of the magnetic bubble is beyond the scope of this paper . since the bubble is rotating about the @xmath67axis initially , the bubble has initial net angular momentum . conservation of azimuthal angular momentum will slow the bubble s rotation since some angular momentum will be transported to the background plasma . @xcite it is a key nonlinear plasma physics question to address how this angular momentum evolves . for an ideal mhd flow , the azimuthal angular momentum conservation equation in cylindrical coordinates @xmath143 is : @xcite @xmath144=0.\ ] ] where @xmath145 is the unit vector in the azimuthal direction , the @xmath5 subscript refers to a poloidal magnetic - field component ( _ i.e. _ , the @xmath26 or @xmath20 component ) , and @xmath146 . there are no source terms in this equation , _ i.e. _ , angular momentum may be redistributed in the fluid but never destroyed . the numerical diffusion present in the simulations would transport some angular momentum as well . however , the influence of this transport would be highly limited in the shock / wavefront regions and negligible elsewhere . the first term in the bracket @xmath147 , the so - called advection angular momentum flux " @xmath148 , is the angular momentum flux vector due to the advection , which is defined , in cartesian coordinates , as : @xmath149 where @xmath150 , @xmath151 and @xmath152 are the unit vectors in the @xmath65 , @xmath66 and @xmath67directions respectively . the second term in the bracket @xmath153 , the so - called maxwell angular momentum flux " @xmath154 , is the angular momentum flux vector due to the lorentz force , which is defined in cartesian coordinates as : @xmath155\ , , \ ] ] where @xmath156 is the polar angle with @xmath157 and @xmath158 , @xmath159 are the radial and axial unit vectors in cylindrical coordinates , respectively . they both contribute to the angular momentum transport in every direction . the third term , the so - called pressure angular momentum flux " @xmath160 , is the angular momentum flux vector due to the effective pressure , which is defined in cartesian coordinates as : @xmath161\,,\ ] ] where @xmath162 . this term does not have a @xmath67component , _ i.e. _ , it only distributes azimuthal angular momentum in the toroidal plane ( @xmath58-@xmath59 plane ) . the total angular momentum flux @xmath163 is defined as : @xmath164 figure [ momentum2 ] displays vector plots of @xmath148 , @xmath154 , @xmath160 and @xmath163 in the toroidal plane , _ i.e. _ , @xmath128 plane at @xmath165 , which coincides with the bubble . pressure simply transports angular momentum in an anti - clockwise direction in this plane . inside the bubble , advection transports angular momentum outward , which is due to the expansion of the bubble . the maxwell angular momentum flux only has a radial component in the plane ( eq . [ eq : maxwell ] ) . this flux is dominant outside the bubble since the radial bubble field component decreases so quickly that it is much smaller than the background field at @xmath75 [ see also fig . [ field_xy](right ) ] due to the quick relaxation ( sec . [ balance ] ) . interestingly the lorentz force transports angular momentum inward at the top left and bottom right regions and outward at the top right and bottom left regions . therefore the total effect is to transport angular momentum : ( 1 ) angular momentum outward inside the bubble ; ( 2 ) along the negative @xmath66axis for @xmath166 and along the positive @xmath66axis for @xmath167 outside the bubble . the maximum angular momentum transport happens at the edge of the bubble . inside the bubble some angular momentum has been transported from the center of the bubble to the edge of the bubble , while outside some angular momentum has been transported from the top left to the bottom left regions , and from the bottom right to the top right regions [ fig . [ angular2](right ) ] . thus , the uniformly rotating bubble expands and its inner region ceases to rotate and then rotates oppositely in the long run , while the neighboring plasma starts to rotate differentially . the top right and bottom left regions rotate in the same direction as the original bubble , while the top left and bottom right regions rotate in the opposite direction ( fig . [ angular2 ] ) , which results in shears . this explains why , between the shock and wavefront , the advection transports the angular momentum in negatively when @xmath166 while positively when @xmath167 since the shock is always propagating outward ( eq . [ eq : advection ] ) . angular momentum transport in the poloidal plane , _ i.e. _ , the @xmath58-@xmath20 plane at @xmath71 , is presented in fig . [ momentum ] ( @xmath160 is zero on this plane ) . inside the bubble , @xmath148 due to the expansion of the bubble transports angular momentum outward normal to the wavefront , while @xmath154 due to the bubble field redistributes angular momentum inside , transporting angular momentum clockwise on the left hand side and anti - closewise on the right hand side . this can be understood from the spheromak - like magnetic field configuration of the bubble . the total effect is to transport net angular momentum from the right hand side to the left hand side of the bubble edge , leading to positive angular momentum on the left hand side and negative angular momentum on the right hand side of the bubble edge ( fig . [ angular ] ) . in this paper we presented initial nonlinear ideal mhd simulation results of the expansion of a magnetic bubble into a lower pressure weakly magnetized background plasma . the simulations mimic the ongoing experiment pbex , except that we use simplified `` out - flowing '' boundary conditions and ignore collisional effects . a high - density magnetized bubble is injected into a cylindrical background plasma . the bubble evolution is dependent on @xmath49 , with larger @xmath49 resulting in an axially expanding bubble like a growing mushroom " and smaller @xmath49 producing a crab - like " shape expanding normal to the direction of propagation . the expansion of the bubble generates one leading mhd shock and one trailing reverse slow - mode compressible mhd wavefront . the shock / wavefront speed is independent of the injection velocity for injection velocities similar or below @xmath168 . in the @xmath58-@xmath20 plane anti - parallel reconnection takes place on the left hand side of the wavefront , where the bubble field is opposite to the background field , while in the @xmath58-@xmath59 plane the reconnection takes place if @xmath131 . the azimuthal angular momentum is transported outward from the center to the edge of the bubble by advection , and from the right hand side to the left hand side of the bubble edge by the maxwell torque . outside the bubble , angular momentum is transported from the top left to the bottom left and from the bottom right to top right by the combination of maxwell and pressure angular momentum fluxes . the initial uniformly rotating bubble quickly evolves into a quasi - force - free state , and the center of the bubble ceases to rotate and starts to rotate oppositely in the long run , while the outside neighboring plasmas start to rotate differentially : the top right and bottom left regions possessing the same rotating direction as the original bubbe while the top left and bottom right regions possessing the opposite rotating direction , which forms shears in the system . the background magnetic field breaks the symmetry of the system along the propagation axis . for comparison , simulations with 2 orders of magnitude lower background field ( and correspondingly lower density and temperature in order to preserve the equilibrium of the background plasmas ) were also performed . the results show much better symmetry along the propagation axis . more detailed studies of the role of the background field will be the subject of future work . from table [ toroidal ] , the resistive dissipation time due to numerical diffusion is inferred to be @xmath169 , which is longer than the time ( @xmath75 ) at which the shock has reached the @xmath58 and @xmath59 boundaries . the alfvn time can be calculated as @xmath170 , where @xmath171 ( @xmath172 ) is the typical length of the reconnection layer and @xmath173 is the alfvn speed ( @xmath174 ) at the reconnection layer . this gives @xmath175 . thus the effective lundquist number @xmath176 in the simulations is around @xmath177 . please note that this is the lundquist number associated with the reconnection layer and the numerical diffusion used here is the upper limit of the numerical diffusion in the simulations . the estimate of the mean experimental lundquist number @xmath178 is around @xmath179 , where @xmath180 is the characteristic length of the experimental facility , @xmath181 is the initial alfvn speed , @xmath182 is the magnetic resistivity of the plasma at the initial state based on braginskii s formula and @xmath183 indicates the volume average . although there is one order of magnitude difference between them , we expect our ideal simulations with numerical diffusion to give a reasonable estimate of the physical quantities in the real experiment . another issue important in the experiment is the transit time . although the value of the transit time is somewhat related to the boundary conditions , our simulations with simplified boundary condition show that the background plasma column radius @xmath184 is large enough to allow the bubble to relax substantially . from fig . [ overview ] , it is seen that the bubble is still inside the background plasma at @xmath76 . however , better boundary conditions are needed for the simulations to be meaningful after the shock has reach the boundaries . the appearance of the mhd shock and wavefront suggests that our experimental facility may provide a unique opportunity to study mhd shocks in a laboratory plasma . however , we emphasize that these conclusions are based on ideal simulations ( with numerical diffusion ) and that the boundary conditions are not realistic . this paper is intended as a preliminary exploration of pbex . we have not attempted to model many of the complexities of a realistic experiment . in future papers , we will study collisional effects and boundary conditions closer to those of the planned experiment ; work in progress indicates that these will modify the results .

nonlinear ideal magnetohydrodynamic ( mhd ) simulations of the propagation and expansion of a magnetic bubble " plasma into a lower density , weakly - magnetized background plasma are presented . these simulations mimic the geometry and parameters of the plasma bubble expansion experiment ( pbex ) [ a. g. lynn , y. zhang , s. c. hsu , h. li , w. liu , m. gilmore , and c. watts , bull . amer . phys . soc . * 52 * , 53 ( 2007 ) ] , which is studying magnetic bubble expansion as a model for extra - galactic radio lobes . the simulations predict several key features of the bubble evolution . first , the direction of bubble expansion depends on the ratio of the bubble toroidal to poloidal magnetic field , with a higher ratio leading to expansion predominantly in the direction of propagation and a lower ratio leading to expansion predominantly normal to the direction of propagation . second , an mhd shock and a trailing slow - mode compressible mhd wavefront are formed ahead of the bubble as it propagates into the background plasma . third , the bubble expansion and propagation develop asymmetries about its propagation axis due to reconnection facilitated by numerical resistivity and to inhomogeneous angular momentum transport mainly due to the background magnetic field . these results will help guide the initial experiments and diagnostic measurements on pbex .

introduction problem setup simulation results summary & discussions

the silicon tri - gate transistor used in this work is fabricated using standard cmos technology , similar to previously reported devices @xcite . the channel , defined by the top gate , is 11 nm high , 42 nm wide and 54 nm long . on both sides of this gate , 25 nm long si@xmath70n@xmath71 spacers form barriers to the highly doped source and drain regions . with a background boron doping of @xmath72/@xmath73 in the silicon channel , about 10 boron atoms are expected under the gate and about 10 under the spacers . in a dilution refrigerator at a base temperature of 40 mk , hole transport and rf gate reflectometry measurements give complementary insights in the static and dynamic behaviour of the two - hole system of interest . the hole temperature was measured from quantum - dot transport at @xmath44 300 mk . the noise floor of the transport measurement ( @xmath44100 fa ) allows us to detect currents with a total tunnel rate down to 500 khz . a surface mount 270 nh inductor , attached to the top gate , forms a tank circuit together with a parasitic capacitance of @xmath440.3 pf . tunnel events in the transistor with a tunnel rate higher than or equal to the resonant frequency of 583 mhz result in a change of the reflected amplitude and/or phase of this resonator . furthermore , low intensity light was directed to the silicon chip through an optical fiber to enable the use of the lower doped substrate as a back gate . in order to estimate the location of the two atoms , we have extracted the relative capacitive coupling to the different electrodes . the hole temperature of @xmath44300 mk gives @xmath74 for the acceptor - lead tunnel rate , thereby justifying a thermally broadened fit @xcite . the reflectometry line width shows a relative coupling to the top gate ( @xmath75 ) and drain ( @xmath76 ) of 0.50@xmath300.01 ev / v and 0.41@xmath300.02 ev / v respectively , thereby placing this atom roughly halfway between these two electrodes , under the @xmath77 spacer . this relatively weak coupling to the top gate supports our identification of the acceptor atom , as gate - induced quantum dots are known to have gate couplings above @xmath440.7 ev / v @xcite . the second boron atom has to be located closer to the centre of the channel , as the tunnelling to both electrodes is too slow to be detected . = 1.93 v and @xmath78 = -777 mv . * e * , ( * f * ) amplitude ( phase ) shift for the linecuts on the red ( inter - acceptor singal ) and blue ( acceptor - lead singal ) dotted lines in * c * ( * d*).,width=332 ] in the rf gate reflectometry experiments shown in figs . [ fig : figure1 ] , [ fig : figure2 ] , and [ fig : figure3 ] , a driving frequency of 583 mhz is used , which is close to the resonant frequency and ideal to distinguish the acceptor - lead and inter - acceptor transitions . the full frequency response of the reflected signal from the lc - circuit reveals more information about the tunnel rates involved in the experiment @xcite . frequency domain measurements of the amplitude and phase response around the resonance of the lc circuit are shown in fig . [ fig : figures1 ] . these measurements show that the acceptor - lead signal mainly changes the amplitude of the reflected signal , while the inter - acceptor signal mainly reduces the resonant frequency . this difference in response of the lc - circuit can be explained by a difference in tunnel rate between these two processes @xcite . when the tunnel rate is much faster than the resonant frequency ( @xmath79 ) , the tunnel process will predominantly add capacitance to the circuit ( dispersive response ) , which in turn shifts the resonance to a lower frequency . this is observed for the inter - acceptor tunneling process , thereby revealing a tunnel rate much faster than the resonant frequency . when the tunnel rate is of the order of the resonant frequency ( @xmath80 ) , the tunnel process will predominantly add resistance to the circuit ( dissipative response ) . this added resistance can either bring the circuit closer or further away from its ideal matching condition , depending on the initial impedance of the circuit , thereby either reducing or increasing the reflection amplitude . a reduction in reflection amplitude is observed for the acceptor - lead tunneling process , thereby revealing a tunnel rate of the order of the resonant frequency . = 4.3@xmath300.3 ghz . the inset shows a schematic representation depicting the excitation from the bonding to the anti - bonding singlet state with microwaves.,width=332 ] to find the tunnel coupling between the two acceptor atoms , we probe the resonance between the bonding and anti - bonding singlet @xmath81 states , as shown in the inset of fig . [ fig : figures2 ] , in accordance with a measurement on two electrons on a hybrid donor - quantum dot system @xcite . here we make use of the bias - tee coupled to the drain lead , as shown in fig . [ fig : figure1]a . the inter - acceptor signal is probed with rf gate reflectometry , while continuous wave microwaves are applied to the drain electrode . when the transition between the bonding and anti - bonding @xmath81 states is excited by the applied microwaves , the occupation of the bonding state is lowered , while the occupation of the anti - bonding state is increased . as the bonding and anti - bonding singlet states give rise to an opposite response of the resonator @xcite , the observed @xmath2 is expected to decrease when the occupation in the anti - bonding state is increased and completely disappear when the occupations in the system become unpolarised . in order to keep the input power of the applied microwaves constant during the experiment , the splitting of the acceptor - drain signal is also probed during the experiment . the splitting of this signal has no dependence on the microwave frequency and is solely dependent on the microwave power applied at the drain . using this measurement we can calibrate for any frequency dependent losses in the system and use the same input power for every frequency . the response of the inter - acceptor tunnel signal is shown in fig . [ fig : figures2 ] as @xmath82 , with @xmath83 being the signal without microwaves . the resonance dips at a frequency of the 8.6@xmath300.6 ghz , which demonstrates a tunnel coupling between the atoms of 4.3@xmath300.3 ghz . the width of this resonance indicates a @xmath84 for these charge states of @xmath44 200 ps , similar to what was found for a two - electron system on a hybrid donor - quantum dot @xcite . -states . * a * , magnetic field dependence of the energy levels of a hole bound to a single boron acceptor in silicon with a compressive strain of 1.2@xmath8510@xmath86 parallel with the magnetic field and an electric field of 0.6 mv / m perpendicular to the magnetic field . * b * , magnetic field dependence of the energy levels of a hole bound to a single boron acceptor in silicon with a compressive strain of 0.6@xmath8510@xmath86 parallel with the magnetic field and an electric field of 1.2 mv / m perpendicular to the magnetic field . dashed lines show the linear projection of the high magnetic field zeeman effect , showing a change in @xmath26 in * b*.,width=332 ] in bulk silicon the single hole ground state of an acceptor atom is fourfold degenerate , consisting of heavy hole and light hole states . the effect of strain , electric field and magnetic field on these states is well described by bir et al . strain and electric field split the heavy and light hole manifolds , leading to a splitting of @xmath26 at zero magnetic field . a magnetic field splits all four spin - states , where this zeeman effect strongly depends on the local environment of the atom . the two atoms probed in the experiments are found to have heavy hole ground states . the @xmath26 is assumed to be different for the two acceptors , as they experience a different local environment . as an example , the zeeman effect is calculated for two different local environments where the heavy holes form the ground state at zero magnetic field , shown in fig . [ fig : figures3 ] . for the first example we consider the situation where a compressive strain of 1.2@xmath8510@xmath86 is present parallel to the magnetic field and an electric field of 0.6 mv / m is present perpendicular to the magnetic field , see fig . [ fig : figures3]a . in fig . [ fig : figures3]b the situation with half of the strain and double the electric field is considered . in this second example , we note that if a linear approximation is taken for the hole - spin states in the high magnetic field region and traced back to zero magnetic field ( dashed lines ) , a different heavy - light hole splitting is found than at zero magnetic field . this effect , caused by a rotation of the quantization axis of the angular momentum of the hole , has been observed in our experiment presented in fig . [ fig : figure4 ] . for the doubly occupied and positively charged acceptor state ( a@xmath87 ) the exchange energy between two holes from the same manifold far exceeds the heavy - light hole splitting ( @xmath88 ) . due to this exchange interaction , only 6 out of the 16 possible two - hole states can be bound to a boron atom in silicon @xcite . a similar sixfold manifold is found for two - holes bound to two closely placed boron atoms @xcite and the neutral states of two holes bound to group ii acceptors @xcite . as these six states are build from combinations of different spin states , one heavy hole singlet state , one light hole singlet state and four light / heavy - quadruplet states are found , which we will write as : & _ = 1/ ( - ) , + & _ = 1/ ( - ) , + & ^2-_= 1/ ( - ) , + & ^-_= 1/ ( - ) , + & ^+_= 1/ ( - ) , + & ^2+_= 1/ ( - ) , where @xmath89 represent the heavy hole states and @xmath90 represent the light hole states . furthermore , when an exchange interaction between the heavy and light holes is introduced , @xmath23 , the q@xmath91 state splits into a singlet and triplet state @xcite . in fig . [ fig : figures4]a an example of the magnetic field dependence of the a@xmath20-states of an acceptor is shown , using the a@xmath92 states from fig . [ fig : figures3]b and neglecting any exchange between the light and heavy holes ( @xmath23 = 0 ) . the colors in the figure show the heavy - hole ( red ) and light - hole ( blue ) character of these states , where the states built from one heavy and one light hole turn purple . the splitting between the heavy hole singlet and the light hole singlet ( @xmath93 - @xmath94 = 2@xmath26 ) is found to be different in the high magnetic field limit compared to the zero magnetic field splitting , in line with was found in fig . [ fig : figures3]b . energies for the same local environment as in fig . [ fig : figures3]b and @xmath95 . a clear difference in splitting between the s@xmath16 and s@xmath28 states in the low and high magnetic field limit is visible . * b * , the ( 1,1)-configuration product states of the two single hole states described in fig.[fig : figures3]a and [ fig : figures3]b.,width=332 ] in the situation where each hole is bound to a different acceptor , the two hole state can be described as a product state of the single particle states , leading to an energy spectrum of 16 states divided into 4 fourfold degenerate manifolds at zero magnetic field . each fourfold manifold splits into a singlet and a triplet state when an exchange interaction is introduced . in our experiments the heavy - hole manifold has the lowest energy at zero magnetic field and can be described as : & _ = 1/ ( , - , ) , + & ^-_= , , + & ^0_= 1/ ( , + , ) , + & ^+_= , . the energy splitting to the next two manifolds is given by the @xmath26 of each acceptor , where these two - hole states consist of one heavy and one light hole state . finally , the light - hole manifold is split from the heavy - hole manifold by the sum of both @xmath26 s . in fig . [ fig : figures4]b an example of such a ( 1,1 ) level configuration is shown for a combination of the a@xmath92-states presented in fig . [ fig : figures3]a and [ fig : figures3]b . although many level crossings are observed , we highlight the fact that in this situation where the heavy holes form the ground state manifold , the ground state in the ( 1,1 ) region remains the t@xmath96 state for any positive magnetic field . in the transport measurements shown in fig.[fig : figure4 ] the ( 1,1)@xmath97(2,0 ) transition energies at zero magnetic field and as a function of magnetic fields up to 3 t are probed . in these experiments , a voltage of + 10 mv is applied to the drain , while the current is measured at the source . this bias direction gives a transport direction for the holes from the drain to the source . from the rf gate reflectometry experiments , as shown in fig . [ fig : figure2 ] , it is deduced that the acceptor with a charge transition from 0 to 1 hole has a stronger tunnel coupling to the drain ( detectable with rf reflectometry ) than the acceptor with a charge transition from 1 to 2 holes ( not detectable with rf reflectometry ) . it follows that the probed sequential tunnelling involves the ( 1,1)@xmath97(2,0 ) transition , where the two observed bias triangles are formed by the two following hole charge transfer processes : ( 1,0 ) ( 1,1 ) ( 2,0 ) ( 1,0 ) , + ( 2,1 ) ( 1,1 ) ( 2,0 ) ( 2,1 ) . the current measured within the bias triangle region is around 5 pa at the baseline at zero magnetic field and declining to the measurement noise floor of @xmath44 100 fa at the other side of the triangle , indicating a total tunnel rate of the sequential tunnel process around 30 mhz and lower . we recall that the resonant tunnel - rate between the ( 1,1 ) and ( 2,0 ) s@xmath16 states was measured at 4.3 ghz and the tunnel - rate from the drain to the acceptor was found to be around the resonant frequency of 583 mhz , indicating that the tunnel - rate between the second acceptor and the source is the limiting tunnel process . however , we note that due to off - resonant tunneling or tunneling between two states consisting of different spins , the inter - acceptor tunnel rate can be strongly suppressed , thereby possibly becoming the limiting tunnel process . furthermore , the ratio between this inter - acceptor tunnel rate and the relaxation rate of each excited state in the ( 1,1)-configuration determines if this excited state will have any effect on the measured current . the transport measurement presented in fig . [ fig : figure4 ] can be explained by only taking the heavy - hole manifold of the ( 1,1)-configuration into account . this indicates that any state which includes any light - hole has a fast relaxation to the heavy hole ground state , which matches with the fast light - to - heavy hole relaxation process observed as a relaxation hotspot in fig . [ fig : figure3 ] . the shift in phase response of the reflected signal as shown in fig . [ fig : figure2]b is ascribed to the addition of a quantum capacitance ( @xmath98 ) to the lc - circuit @xcite when holes can tunnel between the two atoms , given by : @xmath99 where @xmath100 is the difference in capacitive coupling from the top gate to each acceptor , found to be 0.14 ev / v in the transport measurements , @xmath101 is the electron charge , @xmath15 is the energy detuning between the two acceptors , and @xmath29 is the tunnel coupling between the ( 1,1 ) and ( 2,0 ) s@xmath16 states . all other tunnel couplings between ( 1,1 ) and ( 2,0 ) states are considered to be much smaller and therefore not contributing to a change in the detected reflected signal . the total phase shift ( @xmath102 ) is determined by the difference in occupation , given by the boltzmann distribution , of the lower and upper branch of the heavy hole singlet state : @xmath103 where @xmath104 and @xmath105 are the energies of the bonding and anti - bonding singlet states respectively and @xmath106 describes the energies of all the involved states . using eq . [ eq : eqm1 ] and [ eq : eqm2 ] , the magnetic field dependence of the phase shift caused by the inter - acceptor tunneling is calculated as shown in fig . [ fig : figure2]c . we note that although the crossing of the q@xmath91 state and the s@xmath16 state has some influence on the boltzmann distribution of the s@xmath16 states , this influence is not distinguishable in the measurement . however , for completeness , we have displayed the @xmath46 crossing point in fig . [ fig : figure2]c , as found in the relaxation - hotspot measurement together with an estimated g - factor of the q@xmath91 state . to t@xmath96 relaxtion ( blue ) or with the s@xmath16 to q@xmath107 relaxation ( red ) . both fits have a comparable value for @xmath108 . * b * , comparison of the modeled relaxation mechanisms on a larger scale of magnetic field and @xmath109.,width=332 ] at magnetic fields much higher than the @xmath110 mixing point the dominant relaxation mechanism comes from either the relaxation to the ( 1,1 ) t@xmath96 state or the relaxation to the ( 2,0 ) q@xmath107 state , given by eq . [ eq : eq1 ] . in fig . [ fig : figures5]a the fits of the relaxation rate data with a combination of eqs . [ eq : eq1 ] and [ eq : eq2 ] is given for the relaxation process to the t@xmath96 state ( blue ) and the q@xmath107 state ( red ) are shown . our data is insufficient to determine if a @xmath111 or @xmath112 relaxation mechanism is dominant at high magnetic fields . the relaxation amplitudes extracted from these two fits , which are interpreted as the relaxation time at a field of 1 t relative to the corresponding level - crossing , is found to be 0.5 @xmath66s for @xmath113 and 5 @xmath66s for @xmath63 . both of these values are in good agreement with the theoretical predictions for light - heavy hole relaxation ( @xmath109 = 0.14 @xmath66s at 1 t @xcite ) and heavy - heavy hole relaxation ( @xmath109=40 @xmath66s at 0.5 t @xcite ) . [ fig : figures5]b shows the extracted relaxation rates on a larger scale of @xmath114 , demonstrating the more than 3 orders of magnitude increase in relaxation rate at the hotspot . furthermore , it reveals that if this hotspot behavior can be moved to a higher magnetic field , by increasing the heavy - light hole splitting or decreasing the g - factor , relaxation times longer than 1 ms could be achieved at magnetic fields around 0.5 t . the observed relaxation hotspot is explained by the mixing of the bonding s@xmath16 state and the ( 2,0 ) q@xmath45 state , which allows a light - to - heavy hole relaxation to the ( 1,1 ) t@xmath17 state . this relaxation process is expected to have a @xmath111 dependence and furthermore depends linearly on the light hole character of the mixed state . here we show the calculation of the light hole character of the probed state . around the anti - crossing we can write the energies of the lower ( - ) and upper ( + ) states as : @xmath115 where @xmath116 is the energy splitting between the states in the absence of mixing ( @xmath117 - @xmath94 ) , @xmath94 is taken as the reference energy ( @xmath118 ) . we define the amount of light hole character of each state as the chance to measure it as the q@xmath91 state in the basis of q@xmath91 and s@xmath16 , written as @xmath119 and given by : @xmath120 the relaxation measurement is performed at zero detuning , which is the original position of @xmath94 in the absence of mixing . the state probed here can be written as a combination of the + and - states , depending on their relative energy difference with the measurement point @xmath121 . we note that @xmath122 = @xmath123 = @xmath124 and also that @xmath125 is equal to @xmath126 , which gives the total light hole character at the zero energy point as : @xmath127 as used in eq . [ eq : eq2 ] . the key properties of acceptor - based spin - orbit qubits , such as the dipole coupling and decoherence time between the chosen qubit states , are predicted to strongly depend on changes to the symmetry of the silicon crystal and the confinement of the hole @xcite . unstrained silicon ensemble measurements have established an electric dipole coupling of 0.26 debye @xcite and relaxation and coherence times in the micro - second regime @xcite . these acceptor systems are promising for fast electrically - driven spin - orbit qubits with a strong coupling to cavities and phonons @xcite , although a longer coherence time would be desirable . an optimal regime , where the coherence and relaxation times are prolonged while the strong electric dipole coupling is maintained , has been predicted in the small strain regime , giving a small splitting and strong mixing between the heavy and light hole states @xcite . the possibility to lengthen the acceptor relaxation time has been demonstrated in experiments on intentionally heavily strained bulk samples , where @xmath109 times up to 100 ms are measured @xcite . the access to and control over single atoms , as has been demonstrated here , is necessary to probe the physics in an environment with a small local strain , which is the crucial regime for the realisation of acceptor qubits . the device has been designed and fabricated by the tolop project partners , see http://www.tolop.eu . this work was supported by the arc centre of excellence for quantum computation and communication technology ( ce110001027 ) and the arc discovery project ( dp120101825 ) .

two - level quantum systems with strong spin - orbit coupling allow for all - electrical qubit control and long - distance qubit coupling via microwave and phonon cavities , making them of particular interest for scalable quantum information technologies . in silicon , a strong spin - orbit coupling exists within the spin-3/2 system of acceptor atoms and their energy levels and properties are expected to be highly tunable . here we show the influence of local symmetry tuning on the acceptor spin - dynamics , measured in the single - atom regime . spin - selective tunneling between two coupled boron atoms in a commercial cmos transistor is utilised for spin - readout , which allows for the probing of the two - hole spin relaxation mechanisms . a relaxation - hotspot is measured and explained by the mixing of acceptor heavy and light hole states . furthermore , excited state spectroscopy indicates a magnetic field controlled rotation of the quantization axes of the atoms . these observations demonstrate the tunability of the spin - orbit states and dynamics of this spin-3/2 system . single impurities in semiconductors offer an extensive variety of spin systems @xcite , desirable for the emerging fields of quantum technologies , such as quantum communication @xcite , quantum sensing @xcite , and quantum information @xcite . impurities with spin-1/2 , such as phosphorus in silicon , are promising two - level spin systems for quantum technologies @xcite . alternatively , impurities with spin-3/2 offer the possibility to isolate a two - level system whose quantum properties strongly depend on the specific splitting and mixing of the spin-3/2 states , thereby providing a highly tunable quantum system . several spin-3/2 impurities have been investigated in recent studies , such as single co atoms @xcite and si vacancies in silicon carbide @xcite , as well as spin-3/2 hole quantum dots @xcite . however , the adjustability of impurity spin-3/2 systems has not yet been demonstrated . here this tunability is shown using the until now little explored spin-3/2 system of a single acceptor in silicon , which has a level configuration and environmental couplings predicted to yield highly tunable qubits @xcite . of particular interest is the possibility to tune the acceptor system to a regime with a large dipole coupling whilst being robust against charge noise @xcite . by using a silicon cmos transistor to access the single acceptor regime @xcite , which allows excellent control over the environment of the atom , this work demonstrates control over the level configuration of a single acceptor in silicon and probes the resulting influence on the spin - dynamics of the acceptor states . a predicted strong spin - orbit coupling for the heavy hole ( @xmath0 ) and light hole ( @xmath1 ) states makes acceptors in silicon good candidates for spin - orbit qubits @xcite . this type of qubit is advantageous for the all - electrical qubit control by local gates @xcite and the long - range qubit coupling via the interaction with microwave resonators @xcite . the level configuration of the acceptor atom is predicted to strongly depend on the local strain , electric field , magnetic field , and the proximity of the atom to an interface @xcite . as a consequence the essential properties of possible two - level quantum systems created from these levels , such as the electric dipole coupling , relaxation time and coherence time , can be drastically changed @xcite . so far spin - dynamic experiments on acceptor ensembles in bulk silicon have been hindered by a large inhomogeneous broadening , originating from the random heavy - light hole splitting induced by each atom s local environment @xcite . by addressing individual acceptor atoms , which is essential to operate them as qubits , this effect of inhomogeneous broadening is circumvented . single acceptor measurements have demonstrated the effects of electric field @xcite , local strain @xcite , a nearby interface @xcite , magnetic field @xcite , and hubbard - like interactions between two acceptors @xcite . here we demonstrate the ability to tune the level configuration and , as a consequence , the hole spin dynamics of an acceptor atom , thereby showing the fundamental versatility of this spin-3/2 system . for this study a two - hole system confined to two boron atoms in a state - of - the - art silicon cmos transistor ( see supplementary information a for fabrication details ) is investigated using a combination of single hole transport @xcite and dispersive radio - frequency ( rf ) gate reflectometry techniques @xcite ( see fig . [ fig : figure1]a ) . the crossover point between one hole on each acceptor , the ( 1,1 ) state , and both holes on one acceptor , the ( 2,0 ) state , is used to probe the spin state of the double hole system , making use of spin selective tunneling @xcite . magnetic field dependent measurements of the relaxation rate and the excited state spectrum of the two - hole system show that the quantization axis of the angular momentum can be rotated with the applied field , which strongly tunes the mixing and splitting between the heavy and light hole states . the amount of control over the level configuration of a single acceptor atom demonstrated here is an essential step towards an acceptor qubit . . the signals of a boron atom ( black arrow ) and an unintentional quantum dot ( grey arrows ) are identified.,width=332 ] rf gate reflectometry @xcite is used to detect individual acceptors . the charging of a boron atom in the nano - transistor is observed as a shift of the phase , @xmath2 , or the amplitude , @xmath3 , of the reflection from a lc - circuit connected to the top - gate , when driven near the resonance frequency with @xmath4 . this shift is caused by a change in admittance of the lc - circuit induced by tunneling of holes when a localized site in the transistor changes its charge occupation @xcite . in a map of topgate voltage ( @xmath5 ) and backgate voltage ( @xmath6 ) in the presence of a small drain voltage ( @xmath7 = 1 mv ) , shown in fig . [ fig : figure1]b , some charging lines appear far below the onset of hole transport current @xmath8 . these lines are identified as the charging of localized sites which are only strongly tunnel coupled to one of the electrodes . the slopes of these lines correspond to the capacitive coupling ratio to the top and back gate . multiple lines with equal slopes ( grey arrows in fig . [ fig : figure1]b ) imply the charging of several charge states of the same localized state , here likely to be an unintentional quantum dot at the si / sio@xmath9 interface @xcite . in contrast , a boron atom can bind no more than two holes @xcite . a single charging line , unaccompanied with any counterparts of equal slope ( black arrow in fig . [ fig : figure1]b ) , indicates the charging of the first hole onto a single boron atom . the location of this atom is estimated by comparing the capacitive couplings to the four transistor electrodes ( see supplementary information b ) , which reveal a position not directly beneath the top gate , but instead close to the drain lead . this verifies that this charging line does not come from the charging of a gate - defined quantum dot . a sudden electrostatic shift in the charging signal of this acceptor , shown in fig . [ fig : figure2]a , is the signature of the charging of another localized site nearby . further measurements of this state , presented later in this paper , show that this site is a second boron atom . the negative @xmath2 found within the break of the line indicates the tunneling of holes between the two atoms . a detailed study of the reflected signal in the frequency domain shows that the acceptor - lead tunnel rate is comparable to the resonant frequency ( 583 mhz ) of the lc - circuit ( supplementary information c ) . in contrast , the inter - acceptor tunnel rate is found to be much faster than this resonant frequency . the tunneling of holes from either lead to the second acceptor atom is not observed , likely due to tunnel rates much slower than the resonant frequency . the magnetic field dependence of the inter - acceptor tunneling , shown in fig . [ fig : figure2]b , reveals the number of holes bound to the acceptor atoms . we observe the signature of pauli spin blockade in the reflected signal @xcite when a magnetic field is applied along the axis of the nanowire that forms the top gate of the transistor ( y - direction in fig . [ fig : figure1]a ) . approximately at 0.25 t the reflectometry signal starts to weaken due to spin - selective tunneling associated with the increasing thermal population of a triplet state . this observation of spin - selective tunneling indicates the presence of an even number of holes bound to the two acceptor atoms , from which it is concluded that the inter - acceptor tunneling corresponds to the ( 1,1)@xmath10(2,0 ) transition . to describe the two - hole system we use a theoretical framework to construct the ( 1,1 ) and ( 2,0 ) states from single hole states ( supplementary information d ) . the four lowest energy states of a single hole bound to an acceptor in silicon can be described by angular momenta of @xmath11 ( heavy holes ) and @xmath12 ( light holes ) . a distortion of the symmetry of the wave function can lift the degeneracy of heavy and light holes @xcite . our experiments are consistent with heavy hole ground states for both acceptors , with energy splitting @xmath13 between the heavy and light hole states of acceptor @xmath14 . the total energy of the two - hole system at zero magnetic field is shown as a function of the energy level detuning @xmath15 between the atoms in fig . [ fig : figure2]d . to interpret our measurements it is sufficient to only consider the lowest spin manifold for the ( 1,1)-configuration ( right side of fig . [ fig : figure2]d ) , where one heavy hole resides on each acceptor , forming a singlet state , s@xmath16 , and three triplet states , t@xmath17 , t@xmath18 , and t@xmath19 . for the ( 2,0)-configuration ( left side of fig . [ fig : figure2]d ) , which is the doubly occupied acceptor a@xmath20 state , the exchange energy between holes with the same angular momentum ( @xmath21 , @xmath22 ) or different angular momenta ( @xmath23=@xmath24 ) are treated in the limit where @xmath21 , @xmath25 @xmath23 , which is the only regime consistent with our data . in analogy to the two - hole states of neutral group ii acceptors @xcite and of two closely spaced boron atoms @xcite , the lowest energy manifold in the ( 2,0)-configuration consists of six states , split by @xmath26 and @xmath23 . the ground state is the heavy hole singlet , s@xmath16 . four states consisting of one heavy and one light hole , q@xmath27 , with @xmath14 denoting the angular momentum , are @xmath26 higher in energy and split by @xmath23 into singlet and triplet states , as shown in fig . [ fig : figure2]d . the sixth state is the light hole singlet state , s@xmath28 , another @xmath26 higher in energy . at zero magnetic field the tunnel coupling @xmath29 between the ( 1,1 ) and ( 2,0 ) s@xmath16 states gives rise to the inter - acceptor reflectometry signal , where @xmath29 is measured by a microwave excitation experiment to be 4.3@xmath300.3 ghz ( supplementary information c ) @xcite . at a finite magnetic field , which we label @xmath31 , the ( 1,1 ) t@xmath17 state becomes the ground state at the zero detuning point . since the pauli principle forbids tunneling between the atoms in the t@xmath17 state , the reflectometry becomes suppressed in this regime . using the tunnel coupling @xmath29 and the thermal occupation of the two - hole states ( see supplementary information e ) , this magnetic field dependence of the reflectometry signal is simulated @xcite , as shown in fig . [ fig : figure2]c . @xmath31 is found to be 0.25@xmath300.05 t , which provides an effective land g - factor ( @xmath32=@xmath33 ) of the t@xmath17 state of 1.2@xmath300.3 , using the zeeman energy of @xmath34 , where @xmath35 is the bohr magneton . noteworthy is that in bulk silicon the t@xmath17 state is expected to have @xmath36=3 and @xmath37 , which indicates that in our experiment the @xmath32 is strongly suppressed in this low magnetic field regime . such a situation occurs when the quantization axis of the angular momentum is not aligned with the applied magnetic field , indicating that this quantization does not come from the magnetic field but instead from either local strain , electric field , or a nearby interface . a pulsed measurement technique was used to measure the spin dynamics of the two - hole system . the reflectometry signal originating from the inter - acceptor tunneling in the s@xmath16 state can be recovered at magnetic fields above @xmath31 by pulsing the system into the ( 2,0 ) region . such a recovered signal is indicated by a white circle in fig . [ fig : figure3]a , where the time - averaged amplitude and phase response of the reflected signal are displayed for an applied voltage pulse of 0.8 mv to the drain for 40 out of every 200 ns and at a magnetic field of 1.5 t. this indicates a finite population of the s@xmath16 excited state at the zero detuning point , which is utilised to explore the spin - dynamics of the two - hole system using time resolved spectroscopy . fig . [ fig : figure3]b shows the time dependence of the phase response with a voltage pulse applied between the @xmath38 and @xmath39 points , as marked in fig . [ fig : figure3]a , and averaged over @xmath40 pulse sequences . the observed exponentially decaying signal corresponds to the relaxation of the s@xmath16 state . this ability to sense the two - hole spin state could be utilized in a readout scheme for a single hole spin state , in analogy to the scheme first demonstrated by koppens et al . @xcite . fig . [ fig : figure3]c shows the relaxation rate t@xmath41 of the s@xmath16 state as a function of magnetic field . in the high magnetic field regime ( @xmath421.5 t ) , t@xmath41 increases monotonically with magnetic field . this t@xmath41 behavior is well approximated by a power - law dependence on magnetic field , which is usually expected for spin - lattice relaxation processes @xcite . on the other hand , in the low magnetic field regime ( @xmath431.5 t ) , t@xmath41 reaches a maximum at @xmath440.8 t. this implies a different spin - dynamic mechanism , likely arising from the specific interactions within this spin-3/2 acceptor system . from the magnetic field dependence of t@xmath41 we determine what relaxation mechanisms are relevant in the probed double - acceptor system . fig . [ fig : figure3]d illustrates the magnetic field dependence of the two - hole spin eigenenergies at zero detuning . while the ( 1,1 ) t@xmath17 level is the ground state at magnetic fields above @xmath31 , the ( 2,0 ) q@xmath45 level also falls below the s@xmath16 state above another critical magnetic field @xmath46 . the relaxation rates from the s@xmath16 state to the t@xmath17 and q@xmath45 states , @xmath47 and @xmath48 , are expressed by a power law dependence on magnetic field : @xmath49 } } = a_{\text{st[q ] } } \left(\left(b - b_{\text{s}\rightarrow\text{t[q ] } } \right)^2 + ( 2t_\text{st[q]})^2 \right)^{\gamma_\text{st[q]}/2 } \label{eq : eq1}\ ] ] where @xmath50}$ ] is the relaxation amplitude for the @xmath51}$ ] process and @xmath52}$ ] is the tunnel coupling between the s@xmath16 and the ( 1,1 ) t@xmath17 [ ( 2,0 ) q@xmath45 ] state . while @xmath54 is expected to be negligible , the spin - orbit interaction for light and heavy holes is expected to give a sizable @xmath55 . theoretical reports predict different magnetic field power laws for these relaxation process with @xmath56=5 ( kramers degenerate ) and @xmath57=3 ( non - kramers degenerate ) @xcite . both these relaxation rates monotonically increase with magnetic field like the observed t@xmath41 above @xmath441.5 t. however , this dataset is insufficient to conclude which relaxation process is dominant in this high magnetic field range and the best fit with both the @xmath47 and @xmath48 are shown in fig . [ fig : figure3]c ( supplementary information f ) . the non - monotonic magnetic field dependence of @xmath58 in the low magnetic field regime indicates the existence of another relaxation mechanism . the heavy - light hole mixing plays an important role in the spin relaxation around the resonance between the s@xmath16 and ( 2,0 ) q@xmath45 state . in the vicinity of @xmath46 , the finite @xmath55 hybridises the s@xmath16 and ( 2,0 ) q@xmath45 states , resulting in new spin eigenstates , both containing a finite light hole component . this light hole character opens a light - to - heavy hole relaxation path to the ( 1,1 ) t@xmath17 state , which is expected to be much faster than the heavy - to - heavy hole relaxation process between s@xmath16 and t@xmath17 @xcite . since this relaxation mechanism strongly depends on the light - hole component in the hybridised states ( see supplementary information g ) , the relevant transition rate @xmath59 can be written as : @xmath60 where @xmath61 is the relaxation amplitude between the q@xmath45 and t@xmath62 states , which , as mentioned , is expected to be much larger than @xmath63 . the non - monotonic magnetic field dependence of the lorentzian peak and the large relaxation amplitude @xmath61 give a satisfactory explanation for the observed t@xmath41 hotspot around 0.8 t. a combination of the monotonic rate @xmath47 and non - monotonic rate @xmath59 is used to fit the full magnetic field dependence of t@xmath41 , as shown by a solid black line in fig . [ fig : figure3]c . in this fit @xmath46 is found at 0.70@xmath300.02 t , @xmath55 is found in magnetic field units to be 0.04@xmath300.015 t and we extracted a ratio between the relaxation amplitudes @xmath64 . the here observed spin - relaxation hotspot can be used as a rapid initialization of the spin - state of the acceptor atom , where the measured ratio @xmath65 of more than three orders of magnitude demonstrates the excellent tunability of the acceptor system . similar hotspots , caused by the mixing of spin and orbital degrees of freedom , have been predicted and measured for systems with a strong spin - orbit coupling in iii - v semiconductors @xcite . furthermore , @xmath55 is expected to mainly originate from the spin - orbit coupling of the heavy and light hole states . therefore , the strong mixing between the s@xmath16 and q@xmath45 states , as observed in our experiment , points to the presence of a strong spin - orbit coupling for the acceptor states and thereby shows the potential to use them as electrically drivable spin - orbit qubits . to understand the environment of the acceptor atoms we investigate the two - hole excited state spectrum . due to their fast relaxation rate , these excited states are not accessible by the reflectometry technique and instead we use single hole transport spectroscopy @xcite to probe the two - hole energy spectrum . in fig . [ fig : figure4]a the transconductance signal as a function of @xmath5 and @xmath6 at zero magnetic field shows a bias triangle ( contour marked by a solid line ) , demonstrating sequential hole tunneling via both acceptors in series . lines parallel to the triangle baseline ( arrows in fig . [ fig : figure4]a ) are attributed to transitions from the ( 1,1 ) ground state to ( 2,0 ) excited states ( supplementary information d ) . the observed ( 2,0 ) spectrum matches the expected energy spectrum of a doubly charged acceptor state , shown in fig . [ fig : figure4]b . this supports our identification of the second confinement site as an acceptor atom , which is further validated since our measurement is not compatible with the regular spaced excited state spectrum expected for quantum dots . the ( 2,0 ) @xmath26 and @xmath23 are extracted as 110@xmath3010 @xmath66ev and 36@xmath305 @xmath66ev respectively . the measured @xmath26 reflects the effect of local fields on the symmetry of the acceptor states . however , the direction of the resulting quantization axis for the angular momenta is unknown , which makes it impossible to predict the exact zeeman interaction for low magnetic fields @xcite . the extracted @xmath23 reflects the mixing between light and heavy holes , associated with the spin - orbit coupling . we can compare this to @xmath55 , taking into account that this tunnel coupling carries an additional component , as the s@xmath16 state at zero detuning is in a superposition of ( 1,1 ) and ( 2,0 ) states . converting @xmath55 to an energy scale using 1.5 @xmath67 3 , gives @xmath55 in the range of 3 to 7 @xmath66ev . although somewhat lower than @xmath23 , likely caused by the rotation of the quantization axis at this magnetic field as discussed later , both these parameters show a strong spin - orbit coupling , which is crucial to envision electrical manipulations of the acceptor spin states . next we discuss the magnetic field dependence of the transition energy spectrum , as shown in fig . [ fig : figure4]c , which is measured on a cut through the bias triangle at @xmath6=1.4 v ( dotted line in fig . [ fig : figure4]a ) . the baseline of the triangle , indicated with a solid line in fig . [ fig : figure4]c , changes its slope twice . the first change around 0.25 t indicates the singlet - triplet crossing at @xmath31 , in agreement with the observation of spin blockade in fig . [ fig : figure2]b . the second change around 0.50 t indicates the change of the ( 2,0 ) ground state from s@xmath16 to q@xmath45 , corresponding to the same anti - crossing that generates the relaxation hotspot in fig . [ fig : figure3]b , but here occurring in the negative detuning region and therefore at a lower magnetic field . in the high magnetic field regime ( @xmath421 t ) we observe well established resonance lines . the number of resolvable features is larger than at zero magnetic field , indicating spin split excited states . in this regime the zeeman energy dominates over @xmath26 and therefore the quantization axis of the angular momenta of the two - hole states align with the external magnetic field . ignoring any mixing terms other than the s@xmath16 mixing , the magnetic field dependence of the transition energies can be easily described by using @xmath0 for heavy holes and @xmath1 for light holes to calculate the zeeman shifts , as shown in fig . [ fig : figure4]d . using this model , the g - factors and @xmath26 can be extracted by fitting the slopes and position of the transition lines in fig . [ fig : figure4]c respectively . for the doubly occupied acceptor we find @xmath68 of 0.85 and @xmath69 of 1.07 and @xmath26 of 30 @xmath66ev . for the single occupied acceptor only the heavy hole states play a role in the transitions for which we extract @xmath68 of 1.05 . our observations allow for the characterization of the acceptor system in both the low and high magnetic field regimes , which shows the extent of tuning of the acceptor states . the detected level configurations indicate a symmetry transition of the acceptor spin-3/2 system as a function of magnetic field . the g - factors found in the high magnetic field regime are close to those found for bulk boron acceptors @xcite , which is in stark contrast to the effective g - factor found at low magnetic fields in the spin blockade experiment of fig . [ fig : figure2]b . furthermore the @xmath26 of the doubly occupied acceptor is about four times smaller in the high magnetic field regime than at zero magnetic field . these changes demonstrate that the quantization - axis in the high magnetic field regime differs from the quantization axis at zero magnetic field , thereby strongly influencing the g - factor and heavy - light hole splitting . a further indication of the rotating quantization axis is the decaying amplitude of the ( 1,1 ) t@xmath17 to ( 2,0 ) s@xmath16 transition , which completely disappears at a magnetic field around 2 t. the observation of the reduced mixing between these states is well explained by the enhancement of the spin - blockade effect when the angular momentum becomes quantized along the magnetic field . the here demonstrated possibility to change the quantization axis and strongly influence the g - factor , heavy - light hole splitting and relaxation rate of the acceptor states by controlling the atoms environment with an external field , is a fundamental requirement for reaching the optimal conditions for acceptor - based qubits @xcite . in conclusion , the interactions between heavy and light hole states of individual acceptor atoms in silicon have been analysed , establishing the evaluation of the effect of local fields on the wave function symmetry and thereby on the level configurations and spin - dynamics of acceptor atoms . mixing between light and heavy holes is found to strongly influence the spin - dynamics of the two - hole system , leading to a hotspot in the spin - relaxation rate . furthermore it is shown that the quantization axis of the holes can be rotated by applying a magnetic field , thereby significantly tuning the g - factor and heavy - light hole splitting . we have shown access to and control over the heavy and light hole states of individual acceptors and found evidence of a strong spin - orbit coupling in this system , thereby demonstrating the fundamental principles to implement acceptor atom spin - orbit qubits .

supplementary information acknowledgments

there are several ways of approximating heavy quarks in lattice qcd calculations , including nonrelativistic qcd @xcite ( nrqcd ) , the static approximation @xcite , and the approach developed at fermilab @xcite which takes the uncorrected wilson action as its leading approximation but adds correction operators which end up resembling those of nrqcd rather than those of the standard symanzik improvement of the wilson action @xcite . nrqcd is based on an expansion in nonrelativistic operators ( rotationally invariant but not lorentz invariant ) similar to that used in calculating relativistic corrections in the hydrogen atom . it can be thought of as arising from a discretization of the action arising from a foldy - wouthuysen - tani transformation of the quark fields analogous to the one used in atomic physics . @xmath3 leads to the action @xmath4 the pole mass term does not affect the dynamics of nonrelativistic systems and is conventionally dropped . in @xmath5 physics , the simplest of all the methods can be used , the static approximation . this approximates the heavy quark propagator with a simple wilson line moving in the time direction , giving an action corresponding to the first term in eqn . it is clearly most useful for the heaviest quarks . it is not much used recently because it has a much worse signal to noise ratio than nrqcd , which is clear in retrospect but was not foreseen . the third method can be thought of as arising from a partial fwt transformation : @xmath6 and @xmath7 where @xmath8 , and @xmath9 and @xmath10 are between 0 and 1 . this appears to be a crazy thing to do , producing an action which combines the defects of the transformed and the untransformed actions , which indeed it does . on the other hand , it turns out that this is the action we have been stuck with for a long time . the wilson action has just this form at leading order . the @xmath11 term added to cure the doubling problem also contributes to the kinetic energy , as in eqn . [ partialftw ] . furthermore , perhaps surprisingly , the @xmath9 and @xmath10 implied by the wilson action have the desired property that the nonrelativistic @xmath11 term takes over automatically from the dirac style kinetic energy term @xmath12 as @xmath13 in the wilson action . the wilson action automatically turns into a nonrelativistic action in the large mass limit . the wilson action does have the unwanted property that for large masses the pole mass does not equal the kinetic mass governing the energy momentum relation @xmath14 in nonrelativistic systems , the pole mass does not affect the dynamics and the kinetic mass governs the leading important term . therefore , the pole mass term is normally simply omitted from the action in nrqcd , although there is no harm ( and no benefit ) in including it . likewise , the wilson action can be used without problems for nonrelativistic systems as long as @xmath15 and not the pole mass is used to set the quark mass . it is easy to find a wilson style action which does satisfy @xmath16 by letting the hopping parameter for the time direction , @xmath17 , differ from the one in the spatial directions , @xmath18 . the wilson action corrected up to @xmath19 and the action of nrqcd both use one - hop time derivatives . when the quarks are heavy , this is a requirement , since two - hop time derivatives introduce new ghost states and imaginary energies when @xmath20 . therefore , the wilson action applied to heavy paths can not follow the conventional symanzik program of using two - hop corrections at @xmath21 , but must follow the nrqcd style of corrections which correct only the spatial interactions . the existence of the hamiltonian and the transfer matrix ensures that this is possible . the parameters of the action in this approach must have nontrivial mass dependence for large masses , just as do those of nrqcd . for large masses , the wave function normalization , the relation between the mass and the hopping parameter , etc . , are completely different from their @xmath22 values . when @xmath23 ( in lattice units ) , this mass dependence may be expanded in power series . for the wilson and sheikholeslami - wohlert actions @xcite , this yields just the usual series of operators , with the same coefficients : @xmath24 thus , while for @xmath25 the action becomes very similar to nrqcd in its behavior , for @xmath23 it may be regarded as an all orders in m resummation of the usual operators of the wilson and sw actions . for hadrons containing charmed quarks , it is possible to do calculations with the wilson and sw actions even with the old interpretation of the coefficients . however , since @xmath26 , the ability to sum up the required series in @xmath27 exactly is likely to produce a much faster approach to the continuum limit . here we use this approach to calculate the spectrum of @xmath28 annd @xmath29 mesons . the strong interactions of a meson containing a single heavy quark simplify in the infinite mass limit . the heavy - quark spin and the light - quark total angular momentum , @xmath30 , are separately conserved . states with total spin @xmath31 form degenerate - mass doublets . hence , the hyperfine splitting between the vector , @xmath32 , and the pseudoscalar , @xmath33 , is zero . heavy quark symmetry is only approximate at finite quark mass so the hyperfine splittings for and mesons are proportional to @xmath34 . the sensitivity of the hyperfine splitting to the heavy - quark mass makes spectroscopy of the and mesons an important check on the procedure used to determine the bare charm and bottom masses . the bare masses are important inputs to our programs to determine the renormalized quark masses and weak matrix elements of the and mesons@xcite . heavy - quark masses in this study are determined in quarkonia . the bare mass is found by demanding the experimental spin - averaged meson mass match the kinetic mass determined from the lattice energy - momentum dispersion relation@xcite . quark masses may also be determined by studying heavy - light mesons . these masses will differ from the masses in -onia since lattice spacing errors differ for the two methods@xcite . mass errors with either method however lead to bounded uncertainties in quantities such as hadron mass splittings and matrix elements . moreover , quark - mass uncertainties in these quantities are overcome by extrapolation to zero lattice spacing . we study pseudoscalar and vector mesons on the four ensembles of quenched wilson glue described in . lattice spacings are determined using the charmonium splitting . spacings range from @xmath35 to @xmath36 . the coarsest lattice only provide a check upon cutoff effects ; results are not used for continuum extrapolations . we use sheikholeslami wohlert ( sw ) quarks with tree - level tadpole improvement . the clover coefficient , @xmath37 , is determined from the average plaquette ( see ) . calculations are done directly at the charm and bottom quark masses using . results are fully @xmath1 improved . bare strange - quark masses are determined from the light pseudoscalar spectrum using leading - order chiral perturbation theory and the experimental @xmath38 and @xmath39 masses . meson masses are extracted from minimum @xmath40 multistate fits to a matrix of meson correlators having coulomb - gauge 1s- and 2s - smeared sources and sinks . the full covariance matrix is retained in fits . [ fig : hfs ] shows the and hyperfine splittings for the lattices in . results from the three finest lattices are extrapolated linearly to @xmath41 . the slope has an error bound consistent with zero . hence errors from the charm mass determination are barely discernible with present statistics . the slope is more significant indicating larger errors in the bottom quark mass determination . reducing the lattice spacing decreases the splitting . this is consistent with an underestimation of the bottom quark mass in -onia@xcite . the extrapolated and hyperfine splittings are consistent with experimental values . the central values however lie below experiment . this may be an indication that the quenched and hyperfine splittings are smaller than experiment . note that the quenched hyperfine splittings in quarkonia lie significantly below experiment . higher statistics are necessary to determine if there is a similar quenching effect for heavy - light mesons . the ratio of the hyperfine splitting to the splitting , @xmath42 , is proportional to @xmath43 according to heavy quark symmetry . the ratio is expected to be less sensitive to quenching errors than the individual splittings . the linear extrapolation in yields @xmath44 . this compares well with the experimental value @xmath45 . this is evidence that our procedure for determining the bare quark masses leads to the correct continuum result . we also study 2s states . excited state statistical errors are an order of magnitude larger than ground state errors . excited states are also subject to larger systematic errors . combined errors are comparable to the spin splittings . hence we report only the spin - averaged 2s splitting , @xmath46 . according to heavy quark symmetry splitting @xmath46 is constant up to @xmath47 corrections . we check this in by computing the ratio of splittings for and . the deviation from unity is consistent with @xmath47 effects . we summarize our continuum and spectroscopy results in . results are plotted in relation to @xmath34 to emphasize the behavior of the splittings in the heavy quark limit . we compare to experimental values for the 1s states . a potential model inspired estimate for @xmath46 is also shown@xcite . our lattice results are consistent with experiment . fig . [ fig : charmonium ] shows our most recent charmonium spectrum , which we will use in our discussion in the next section . it is mainly a statistical updating of our previous results , except for the @xmath48 states . these have been done using nonrelativistic operators for the first time rather than quark antiquark bilinears . this allows us to obtain a result for the @xmath49 for the first time . the correction operators of nrqcd are simple and easy to organize . nrqcd quark propagators can be calculated almost instantaneously . heavy quark actions which start from the wilson action have the advantage that they continue to work as the @xmath50 limit is taken . however , correcting them to @xmath21 or to @xmath51 is messier than for nrqcd . the action of nrqcd is an expansion around @xmath52 and is better and better behaved , the larger the quark mass . it breaks down in two distinct ways as @xmath53 becomes larger . when @xmath54 gets larger than 1 , perturbative calculability of coefficient functions breaks down due to factors of @xmath54 in the diagrams . when @xmath55 gets larger than 1 , the convergence of the series of nonrelativistic operators breaks down . this series is very convergent for the @xmath56 quark , and nrqcd works very well in @xmath5 and @xmath57 physics . the @xmath58 system is reasonably nonrelativistic , too , as shown by the success of potential models . nrqcd works reasonably well for this system , too , although in some cases not as well as originally hoped . for example , a recent investigation @xcite of the effects of @xmath59 corrections in nrqcd studied their effect on the hyperfine splitting of charmonium . in results of a few years ago , the @xmath60 , @xmath61 improved nrqcd results lay much closer to the experimental answer than the @xmath19 improved relativistic fermilab results . the interpretation seemed to be that the higher order corrections included in the nrqcd calculations made these results much more accurate than the less improved relativistic results . however , last year trottier calculated some of the @xmath59 terms which contribute to the hyperfine splitting and found results which made a big jump back down toward zero . the actual situation seems to be that the @xmath62 expansion is very slow to converge for some quantities in the charmonium system . the ability of the harder to improve relativistic formalism to recover a dirac - like form of the action in the @xmath50 limit is an advantage here . for @xmath56 quark physics , the situation is somewhat reversed . here , the nonrelativistic expansion is well - converged after a few orders . ( potential models estimate that @xmath62 is about 1/4 in the @xmath63 and about 1/10 in the @xmath57 . ) on the other hand , these corrections have to be included to achieve high accuracy . one does not have the option of replacing them and letting the relativistic @xmath12 terms in the action of ref . @xcite take over when @xmath64 . this would require an absurdly high lattice spacing . a case where this plays a more significant role than had been expected is in the quantity @xmath65 . @xcite in the quarkonium systems , the quark momentum @xmath66 as @xmath67 . therefore , @xmath68 errors and @xmath69 errors also go to infinity in absolute terms , but approach fixed percentages of leading order physical quantities . in heavy - light systems on the other hand , the @xmath56 quark s momentum @xmath70 as @xmath67 . @xmath68 errors and @xmath69 errors due to the @xmath56 quark approach constants in absolute terms . of the pole and kinetic masses in the equation @xmath71 @xmath72 is more affected by @xmath62 errors than @xmath73 since it is higher order . for example , we can obtain the spin averaged @xmath74 splittings in the @xmath75 and @xmath76 systems from the data from figs . we obtain 0.81 gev in the charm system , compared with 1.08 gev experimentally . we obtain -1.68 gev in the bottom system compared with 1.35 gev experimentally . kronfeld used potential models to show that this was roughly what was to be expected . @xcite the analogous effect in nrqcd improved only to @xmath77 is about 15% . @xcite it should be stressed that as a percentage error , no error in @xmath5 or @xmath57 physics is blowing up as @xmath78 . for example , the uncertainty induced in determinations of @xmath79 from the 20% ambiguity of using @xmath80 or @xmath81 to determine the @xmath56 quark mass is only of order a few per cent , as can be seen by examining a graph of @xmath82 vs. @xmath83 . it is simply one of the @xmath84 errors known to be present in the @xmath77 improved theory . on the other hand , it is probably reduced from @xmath85 to @xmath86 by obtaining @xmath87 from @xmath80 rather than @xmath81 in this case . it has been argued many times that quarkonia are nice systems to use for error analysis and parameter setting because of the tools that are available for understanding the errors . when those tools tell us that some errors in quarkonia are rather large , that strategy may sometimes have to be altered . for pure @xmath5 meson physics ( not considering the @xmath57 system ) , the situation is probably not as bad for `` relativistic '' heavy quark actions . ( relativistic is in quotes because for @xmath56 systems , relativistic @xmath12 terms in the action have almost no effect compared with the nonrelativistic @xmath88 terms . ) in these systems , @xmath62 correction terms are probably negligible since @xmath89 . it is therefore possible that an @xmath77 improved heavy quark theory such as the heavy sw action may do an adequate job here . perturbative corrections to decay constants , etc . , are more annoying to calculate than for nrqcd or the @xmath22 wilson theory , but by no means intractable . this work is done in collaboration with a. el - khadra , and a. kronfeld . calculations were performed on the acpmaps supercomputer built and operated by the fermilab computing division . fermilab is operated by university research association , inc . under contract with the u.s . department of energy . w. e. caswell and g. p. lepage , phys . lett . * 167b * ( 1986 ) 437 . g. p. lepage and b. a. thacker , nucl . . suppl . * 4 * ( 1987 ) 199 ; + b. a. thacker and g. p. lepage , phys . rev . * d43 * ( 1991 ) 196 . g. p. lepage , l. magnea , c. nakhleh , u. magnea , and k. hornbostel , phys . rev . * d46 * ( 1992 ) 4052 . e. eichten , nucl . . suppl . * 4 * ( 1987 ) 170 . e. eichten and b. hill , phys . lett . * b243 * ( 1990 ) 427 . a. x. el - khadra , a. s. kronfeld , p. b. mackenzie , phys . rev . * d55 * ( 1997 ) 3933 . wilson , phys . rev . * d10 * ( 1974 ) 2445 . b. sheikholeslami and r. wohlert , nucl . phys . * b259 * ( 1985 ) 572 . a. kronfeld , these proceedings . s. ryan and j. simone , these proceedings . j. simone , nucl . b ( proc . suppl . ) * 47 * ( 1996 ) 17 . a. kronfeld , nucl . b ( proc . suppl . ) * 53 * ( 1997 ) 401 . j. rosner , efi-95 - 02 ( hep - ph/9501291 ) .

we study and spectroscopy in quenched lattice qcd using the fermilab approach to heavy quarks . we obtain results at four lattice spacings , @xmath0 , using @xmath1-improved wilson quarks . we compare and contrast the various methods for heavy quarks on the lattice , discussing discussing which methods work best for different physical systems and the ease with which calculations may be performed . # 1#2#3#4_#1 _ * # 2 * ( # 3 ) # 4 # 1#2#3255=#1255 by 1 # 1eq . ( [ # 1 ] ) # 1fig . [ # 1 ] # 1table [ # 1 ] # 1sect . [ # 1 ] # 1subsect . [ # 1 ] # 1#2 # 1 # 1 # 1#2#3 # 1#2 # 1#2#3#1#2#3 # 1 # 1@xmath20.30em*#1 * # 1#2 .05em.3ex -.1em0 / -.1em.3ex # 1 ( # 1)#1 ( + # 1-#2)(^+#1_-#2 )

methods for heavy quarks on the lattice heavylight mesons calculation details spectroscopy which method is best where? acknowledgments

cosmological data obtained from different independent observations of sneia , cmb radiation anisotropies , x - ray experiments and large scale structures are well known to point toward the accelerated phase of expansion of the universe @xcite . the cosmological constant @xmath1 model , dark energy ( de ) models and theories of modified gravity , among other attempts , have been approached in order to provide an explanation for the accelerated expansion of the universe @xcite . the cosmological constant @xmath1 stands for the most straightforward candidate suggested to explain the observational evidence for it . the fine - tuning and the cosmic coincidence problems are questions still underlying the cosmological constant model @xcite . a model for de , motivated by the holographic principle , was proposed @xcite and further studied in @xcite . the holographic model of de ( hde ) has been in addition comprehensively investigated @xcite . the hde model has been further employed to drive inflation of the universe @xcite , and considered in @xcite with different ir cut - offs . for example the future event and the particle horizons and the hubble horizon were considered as well . moreover , correspondences between the hde model and other scalar field models have been recently suggested @xcite . the hde model fits the cosmological data by cmb radiation anisotropies and sneia @xcite . recently , the cosmic acceleration has been also well studied by a promising modified gravity model that has recently attained visibility : the modified gravity chern - simons model @xcite . the low - energy limit of string theory comprises a correction that cancels the anomalies to the einstein - hilbert action , wherein the chern - simons modified gravity is derived as an effective theory . gravitational parity violation was first investigated using this framework @xcite , appearing both in 4d compactifications of perturbative string theory and further in loop quantum gravity as well , when the barbero - immirzi parameter emulates a scalar field coupled to the nieh - yan invariant @xcite . in the chern - simons modified gravity , the pontryagin topological invariant is well known not to affect the field equations . thus the so called chern - simons correction consists of the product of the pontryagin density by a scalar field , regarded as either a dynamical evolving field or a non - dynamical background field . in the former case , the dynamical chern - simons modified gravity ( dcsmg ) is therefore approached @xcite . some efforts have recently provided bounds to the chern - simons parameter , accordingly @xcite . in this paper , we study the many faces of de models in the context of the dynamic formulation of chern - simons gravity , where the coupling constant is promoted to a scalar field . recent applications include for instance neutron star binary @xcite . we shall investigate three different de models : the hde model with granda - oliveros ( go ) cut - off , the modified holographic ricci dark energy ( mhrde ) model and a recently proposed model with higher derivatives of the hubble parameter @xmath2 @xcite in the framework of chern - simons gravity , in order to obtain the expressions of the scale factor for each model . we prove that both the hde model with go cut - off and the model with higher derivatives of the hubble parameter @xmath2 in the framework gravity chern - simons are related to the modified chaplygin gas models @xcite that further represent the well known models of dark energy as chaplygin gas @xcite . the paper is organized as follows . in section ii , we briefly revisit dynamical chern - simons modified gravity model . in section iii , we describe the three different de models considered in this work in the framework of the chern - simons modified gravity models and we derive the relevant expressions of the scale factor @xmath3 . finally , in section iv , we write the conclusions of this work . this section is devoted to provide the main features of the chern - simons modified gravity model . a homogeneous and isotropic universe described by the friedmann - robertson - walker ( frw ) metric is governed by the metric @xmath4 where @xmath5 is the scale factor and @xmath6 denotes the curvature parameter assuming the values + 1 , 0 and @xmath7 leading respectively to an open , a flat or a closed universe . the action for the chern - simons ( cs ) theory is given by @xcite : @xmath8 + s_{mat},\label{action}\end{aligned}\ ] ] where @xmath9 represents the newton s gravitational constant , @xmath10 is the determinant of the spacetime metric @xmath11 , @xmath12 denotes the ricci scalar defined with the ricci tensor @xmath13 , and @xmath14 stands for the riemann tensor components and @xmath15 denotes the components of the hodge dual riemann tensor . the term @xmath16 denotes the 4d coupling constant , @xmath17 is the pontryagin invariant , and the function @xmath18 indicates the dynamical scalar field of the model and @xmath19 represents the action of matter . for simplicity we consider the potential @xmath20 equal to zero . the second term , which is called the chern - pontryagin ( cp ) term , can be converted to the cs term via partial integration as @xmath21 the third term in eq . ( [ action ] ) is the kinematic term for @xmath18 . by varying the action @xmath22 given in eq . ( [ action ] ) with respect to the metric tensor @xmath23 and to the scalar field @xmath24 , the following field equations are respectively obtained : @xmath25 where @xmath26 represents the einstein tensor components and @xmath27 indicates the cotton tensor : @xmath28 the energy - momentum tensor @xmath29 has two terms : @xmath30 the term @xmath31 represents the scalar field contribution and @xmath32 indicates the energy - momentum tensor of the corresponding de model . furthermore , @xmath33 denotes the standard time - like 4-velocity , @xmath34 represents the energy density and @xmath35 stands for the pressure of the de model to be considered . the component @xmath36 provides the friedmann equation @xmath37 for @xmath38 where @xmath39 moreover , as the components of the cotton tensor vanish for all spherically symmetric metrics @xcite , in particular @xmath40 . therefore , the friedmann equation reads @xmath41 it is worth to emphasize that eq . ( [ frinew ] ) takes into account @xmath42 . in what follows we ( [ frinew ] ) in the context of the modified chern - simons modified gravity for three different de models , namely , the hde model with granda - oliveros cut - off , the mhrde model and a recent de model which involves the hubble parameter squared and the first and the second time derivatives of the hubble parameter . we shall derive an expression of the scale factor for each one of these models . in this section , our aim is to give a brief description of the de models dealt with and to study their behavior in the framework of chern - simons modified gravity model , in order to find the expressions of the scale factor @xmath0 . in the first subsection , we will consider the hde model with granda - oliveros cut - off , in the second subsection the mhrde model while in the third one the model with higher derivatives of the hubble parameter @xmath2 . granda and oliveros introduced a new ir cut - off , by including a term proportional to @xmath43 and one term proportional to @xmath44 . this new ir cut - off @xmath45 is known as granda - oliveros ( go ) scale , provided by @xcite : @xmath46 where @xmath47 and @xmath48 represent two constant parameters . as the underlying origin of the hde model lacks , the term with the time derivative of the hubble parameter is expected , since this term appears in the curvature scalar @xcite . the expression of the hde energy density with @xmath45 cut - off is given by : @xmath49 where the numerical constant @xmath50 arises from observational data . for a flat [ non - flat ] universe we have @xmath51 @xmath52 $ ] @xcite . it is worth to emphasize that the planck mass @xmath53 is considered hereon normalized to the unit . in the limiting case corresponding to @xmath54 and @xmath55 , the scale @xmath45 reduces to the average radius of the ricci scalar curvature , when @xmath56 in eq.([metricfrw ] ) , corresponding to a flat universe . de models that consider the go scale avoid the causality problem . the friedmann equation ( [ frinew ] ) , when the energy density of de in eq . ( [ lgo5 - 1 ] ) is taken into account , reads ( here we consider the normalization @xmath57 ) : @xmath58 which can be written by substituting @xmath59 as @xmath60 + \beta \frac{\dot{a}^2}{a^2 } + \frac{1}{6}\dot{\theta}^2 . \label{eqdgo2}\end{aligned}\ ] ] we now want to make some considerations about the field equation associated with the scalar field given in eq . ( [ field2 ] ) . the frw metric yields @xmath61 , thus eq . ( [ field2 ] ) reads @xmath62 = 0 . \label{nabla}\end{aligned}\ ] ] by choosing @xmath63 @xcite , eq . ( [ nabla ] ) leads to : @xmath64 what implies that @xmath65 where @xmath66 represents a constant of integration . substituting it into eq . ( [ eqdgo2 ] ) , we obtain the following expression : @xmath67 where @xmath68 . for the sake of simplicity , the following change of variable @xmath69 leads to find @xmath70 in terms of the scale factor : @xmath71 hence , eq . ( [ udia ] ) provides eq . ( [ diffago ] ) to be rewritten as follows : @xmath72 an expression of @xmath73 is not integrable according to eq . ( [ intt ] ) , and therefore we consider the case corresponding to @xmath56 , a flat universe . thus , eq . ( [ diffago2 ] ) can we written as follows : @xmath74 whence we can find the general solution @xmath75 where @xmath76 represents a constant of integration . ( [ intt ] ) , it reads @xmath77 which leads to the following solution : @xmath78 , \label{tsolutiongo}\end{aligned}\ ] ] with @xmath79 representing the hypergeometric function and @xmath80 . some considerations about the values assumed by the parameters in the hypergeometric function can be asserted , in order to find a possible analytical solution . we can analyze the result obtained in eq . ( [ tsolutiongo ] ) by writing the hypergeometric function as a hypergeometric series : @xmath81 = \frac{\arcsin\!{\rm h } \left ( \sqrt{const } \times x \right)}{\sqrt{const}\times x } \label{hyper}.\end{aligned}\ ] ] these new results are prominent solutions , and the limiting case described in eq . ( [ hyper ] ) is obtained in eq . ( [ tsolutiongo ] ) when @xmath82 , leading to the well known results in @xcite : @xmath83 . \label{finaltgo}\end{aligned}\ ] ] in this way , the final expression of the scale factor @xmath84yields the results @xcite in this limiting case . we consider now the holographic cosmological model with ir cut - off given by the modified ricci radius so that the respective energy density is a combination of @xmath44 and @xmath43 @xcite , which reads : @xmath85 where @xmath86 and @xmath87 are two constant parameters . in the limiting case corresponding to ( @xmath88 ) we obtain that @xmath89 becomes proportional to the ricci scalar curvature @xmath90 for a spatially flat frw space - time . the use of the mhrde is motivated by the holographic principle because it is possible to relate the de with an uv cut - off for the vacuum energy with an ir scale such as the one given by the ricci scalar curvature @xmath90 . by a similar procedure as in the previous section , we obtain the following differential equation : @xmath91 where @xmath92 , which can be also written as follows : @xmath93 where @xmath94 , @xmath95 and @xmath96 . our calculations are severely simplified by considering a flat universe ( @xmath56 ) , since the case with @xmath97 leads to equations which can not be solved analytically . thus , eq . ( [ umhrde2 ] ) can be rewritten as @xmath98 which has the following solution : @xmath99 where @xmath100 is a constant of integration . it is thus possible to find now a relation between @xmath70 and the scale factor : @xmath101 . \label{tmhrde}\end{aligned}\ ] ] the case corresponding to @xmath102 is going to be regarded for the sake of simplicity , as it is the single case that brings forth an analytical solution . it implies that @xmath103 and @xmath104 , and eq . ( [ tmhrde ] ) leads to the following expression : @xmath105 what provides the scale factor @xmath106^{\frac{1}{6 } } .\label{amhrdefinalissimo}\end{aligned}\ ] ] eq . ( [ amhrdefinalissimo ] ) implies that @xmath107 in order to avoid singularities . moreover for @xmath108 we obtain @xmath109 . we now consider a de model proposed in @xcite , containing three different terms : one proportional to the squared hubble parameter , one to the first derivative with respect to the cosmic time of the hubble parameter and another proportional to the second derivative with respect to the cosmic time of the hubble parameter : @xmath110 where @xmath86 , @xmath87 and @xmath111 are arbitrary dimensionless parameters . such model can be reduced to the dark energy and ricci - like dark energy models , for instance @xcite . the profile of the dark energy and the expansion of the universe depends on the parameters @xmath112 of the model . the main motivation regarding this model resides on the alleviation of the age problem of three old objects , namely , lbds 53w091 , apm 08279 + 5255 , lbds 53w069 @xcite for the chosen parameters . the energy density given in eq . ( [ higher ] ) can be considered as an extension and generalization of other two de models widely studied recently , i.e. the ricci de ( rde ) model and the de energy density with granda - oliveros cut - off . in fact , in the limiting case corresponding to @xmath113 , we obtain the energy density of de with granda - oliveros cut - off , and in the limiting case corresponding to @xmath113 , @xmath114 and @xmath115 , we recover the rde model for flat universe as well . the approach to this model is slightly different to the previous two models . by substituting eqs . ( [ dotteta ] ) and ( [ higher ] ) in eq . ( [ frinew ] ) , along with the new variable @xmath116 , the following differential equation for @xmath44 is obtained : @xmath117 where @xmath118 and @xmath119 . the limiting case corresponding to @xmath56 is concerned in order to have analytical solutions of the quantities involved . hence , eq . ( [ diffhigher ] ) can be written as follows : @xmath120 having as a solution @xmath121 passing back from @xmath122 to @xmath123 , eq . ( [ diffhigher3 ] ) reads@xmath124 moreover , using now eq . ( [ udia ] ) in eq . ( [ diffhigher4 ] ) yields @xmath125 two different limiting cases of eq . ( [ uhigherlim2 ] ) lead to two different solutions . in the first case , we choose @xmath126 and @xmath127 , what makes eq . ( [ uhigherlim2 ] ) to assume the following expression : @xmath128 we can hence obtain the expression of the scale factor from eq . ( [ thigherlim1 ] ) : @xmath129^{\frac{1}{6 } } \left ( 3 t \right)^{\frac{1}{3 } } . \label{asolhigher1}\end{aligned}\ ] ] in the second case , besides assuming that @xmath126 , we also consider the case corresponding to @xmath130 . thus eq . ( [ uhigherlim2 ] ) yields @xmath131 leading to @xmath132 . \label{tttt}\nonumber\end{aligned}\ ] ] here a similar limit can be regarded in eq . ( [ hyper ] ) , which is recovered for @xmath133 irrespective of the value of @xmath86 , and we obtain the following solution : @xmath134 , \label{tsolhigher2}\end{aligned}\ ] ] which leads to the following solution for the scale factor : @xmath135^{\frac{1}{6 } } \sinh^{\frac{1}{3}}\left ( 3\sqrt{2 } t \right ) . \label{asolhigher2}\end{aligned}\ ] ] some considerations about the values of @xmath86 and @xmath111 in eq . ( [ asolhigher2 ] ) can be summoned now . the condition @xmath136 obviously reads @xmath137 for @xmath138 , implying a tricotomy : 1 ) @xmath139 and @xmath140 , 2 ) @xmath141 and @xmath142 and 3 ) @xmath139 and @xmath142 . we analyze the respective solutions for these three conditions . for the case 1 ) , @xmath143^{\frac{1}{6 } } \sinh^{\frac{1}{3 } } \left ( 3\sqrt{2 } t \right ) . \label{asolhigher2 - 1}\end{aligned}\ ] ] instead , for the case 2 ) we obtain the scale factor in the form @xmath144^{\frac{1}{6 } } \sinh^{\frac{1}{3 } } \left ( 3\sqrt{2 } t \right ) . \label{asolhigher2 - 2}\end{aligned}\ ] ] finally , for case 3 ) , eq . ( [ asolhigher2 ] ) diverges since @xmath145 . this result is in agreement with the one when @xmath146 and @xmath147 in eq . ( [ frinew ] ) : with this combination of the parameters , we obtain the equation @xmath148 , which solutions are @xmath149 or @xmath150 ( which is the result with the case 3 ) ) . thus , the combination of values considered in 3 ) can not be regarded , since it does not produce an analytical solution . in this work we studied the behavior of three different de models : the hde model with granda - oliveros cut - off , the mhrde model and the model with higher derivatives of the hubble parameter @xmath2 , in the framework of the chern - simons modified gravity model . for each of these models , we derived the respective scale factors @xmath3 . for the hde model with go cut - off , the scale factor @xmath151 is an hyperbolic sine function of cosmic time . nevertheless , in the mhrde model paradigm the scale factor is a power law of the time , and finally , according to the values of the parameters involved for the model with higher derivatives of the hubble parameter , we have either a power law solution or @xmath151 proportional to a hyperbolic sine function . the scale factor obtained in eq . ( [ finalago ] ) for the hde with go cut - off and in eqs . ( [ asolhigher2 - 1 ] ) and ( [ asolhigher2 - 2 ] ) for the model with higher derivatives of @xmath2 are similar to those obtained in @xcite . for this reason , we conclude that , for suitable choices of the parameters involved , the hde model with go cut - off and the model with higher derivatives of the hubble parameter @xmath2 in the framework of chern - simons modified gravity have the same results obtained from the modified chaplygin gas @xcite , namely , the results clearly indicate that there is a agreement between both the the hde model with go cut - off and the model with higher derivatives of the hubble parameter @xmath2 in the framework gravity chern - simons , and the modified chaplygin gas . it is worth to emphasize that as ricci dark energy in chern - simons modified gravity is related to ricci dark energy with a minimally coupled scalar when choosing the frw metric , the above mentioned similarity between them is limited to the de sitter phase derived by the cosmological constant in the future @xcite . financial support under grant no . sr / ftp / ps-167/2011 from dst , govt of india is thankfully acknowledged by the second author . rdr is grateful to thanks to sissa for the hospitality , to cnpq grants no . 473326/2013 - 2 and no . 303027/2012 - 6 . rdr is also _ bolsista da capes proc . 10942/13 - 0 . _ 99 p. de bernardis _ et al . _ [ boomerang collaboration ] , nature , 955 * 404 * ( 2000 ) s. perlmutter _ et al . _ [ supernova cosmology project collaboration ] , astrophys . j. * 517 * , 565 ( 1999 ) . a. g. riess _ et al . _ [ supernova search team collaboration ] , astron . j. * 116 * , 1009 ( 1998 ) . c. l. bennett et al . , j. * 148 * , 1 ( 2003 ) . planck collaboration , p. a. r. ade , n. aghanim et al . 2013 , arxiv:1303.5076 m. tegmark _ et al . _ [ sdss collaboration ] , astrophys . j. * 606 * , 702 ( 2004 ) . s. w. allen , et al . , mon . not . 353 * , 457 ( 2004 ) . p. j. e. peebles and b. ratra , rev . mod . * 75 * , 559 ( 2003 ) . v. sahni , class . * 19 * , 3435 ( 2002 ) . e. j. copeland , m. sami and s. tsujikawa , int . j. mod . d * 15 * , 1753 ( 2006 ) . t. clifton , p. g. ferreira , a. padilla and c. skordis , phys . rept . * 513 * , 1 ( 2012 ) . v. cardenas and r. g. perez , class . * 27 * , 253003 ( 2010 ) . s. -f . wu , p. -m . zhang and g. -h . yang , class . quant . grav . * 26 * , 055020 ( 2009 ) . w. zimdahl and d. pavon , class . * 24 * , 5461 ( 2007 ) . q. -g . huang and y. -g . gong , jcap * 0408 * , 006 ( 2004 ) . e. elizalde , s. i . nojiri , s. d. odintsov and p. wang , phys . rev . d * 71 * , 103504 ( 2005 ) . x. zhang and f. -q . wu , phys . d * 72 * , 043524 ( 2005 ) . b. guberina , r. horvat and h. stefancic , jcap * 0505 * , 001 ( 2005 ) . b. wang , y. -g . gong and e. abdalla , phys . b * 624 * , 141 ( 2005 ) . h. li , z. -k . guo and y. -z . zhang , int . j. mod . d * 15 * , 869 ( 2006 ) . j. p. beltran almeida and j. g. pereira , phys . b * 636 * , 75 ( 2006 ) .

dark energy models are here investigated and studied in the framework of the chern - simons modified gravity model . we bring into focus the holographic dark energy ( hde ) model with granda - oliveros cut - off , the modified holographic ricci dark energy ( mhrde ) model and , moreover , a model with higher derivatives of the hubble parameter as well . the relevant expressions of the scale factor @xmath0 for a friedmann - robertson - walker universe are derived and studied , and in this context , the evolution of the scale factor is shown to be similar to that one displayed by the modified chaplygin gas in two of the above models .

introduction chern-simons gravity dark energy models in chern-simons gravity conclusions

the oscillation of the proteins mind and mine from pole to pole of individual cells of the bacterium _ escherichia coli _ is used to localize cellular division to midcell @xcite . one cycle of the oscillation , lasting approximately one minute , starts with atp - associated mind binding to the bacterial inner membrane and polymerizing into helical filaments @xcite ( see also @xcite ) . this occurs at alternating poles of the bacterium , with the mind forming a polar `` cap '' . mine is recruited to the membrane - bound mind , where it forms a distinctive `` e - ring '' @xcite at the edge of the mind cap by accumulating near the mind filament tips @xcite . because the rate of hydrolysis and subsequent release of atp - mind is stimulated by mine @xcite , the e - ring drives depolymerization of the mind filament which allows the oscillation to proceed . the depolymerization occurs with an approximately fixed e - ring width and speed along the cell axis @xcite , indicating an approximate steady - state during this part of the min oscillation . however , little is known about the mechanism of e - ring formation , its detailed structure , or how important it is for min oscillations . indeed , min oscillations have been observed without prominent e - rings @xcite . most models proposed for min oscillation do not have explicit mind filaments @xcite , though they do have e - rings . recently , several models of min oscillations that include explicit mind polymerization have been proposed @xcite , two of which display strong e - rings that track the tips of depolymerizing mind filament caps with constant speed and width @xcite . in these models , e - rings are the result of mine polymerization either orthogonal to @xcite , or along @xcite , mind filaments . while mind polymerization has been observed _ in vitro _ @xcite , there have been no reports of mine polymerization in the experimental literature . indeed , the faint mine `` zebra - stripes '' associated with the mind zones adjacent to the mine ring @xcite seem to imply sparse lateral binding of mine to the body of mind filaments not mine polymerization . in this paper , with both stochastic 3d simulations , and a deterministic 1d model , we show that local ( non - polymeric ) rebinding of mine released from depolymerizing mind filament tips is sufficient for e - ring formation . we impose and characterize a dynamical steady - state of an e - ring on a depolymerizing semi - infinite mind filament in order to address the approximate steady - state speed and width of the e - ring _ in vivo _ we investigate the roles of spatial dimension , cell length , and radius , and of multiple mind filaments and their helical pitch . we estimate the timescale of e - ring formation and obtain results consistent with the significant delay before atpase activity seen with small mine concentrations and large mind membrane coverage _ in vitro _ finally , we discuss how competition between the intrinsic and the mine - stimulated atpase activity of mind controls the instability that leads to the initial formation of the e - ring from a uniformly decorated mind filament . qualitatively , we predict that the width of mine - rings will increase as the mind - filament depolymerization speed is increased through manipulation of cell shape , mind to mine stoichiometry , or mutations that affect the mine binding rate to mind . eventually , the depolymerization speed will saturate but the e - ring width can still grow . conversely , as the depolymerization speed is decreased , mine - rings will undergo a transition from a plateau - like `` strong '' e - ring to a cusp - like `` weak '' e - ring . to our knowledge , systematic experimental studies of the e - ring width have not yet been done . vs. distance along the bacterial axis @xmath1 for the 3d stochastic model ( continuous lines ) and the 1d model ( dashed lines ) for parameters typical of _ e. coli _ : @xmath2 m , @xmath3 m , and @xmath4 . the mine binding parameter was @xmath5 , while for the 1d model @xmath6 was used . one mind filament supports either ( a ) a strong plateau - like e - ring for pitch @xmath7 m or ( b ) a weak cusp - like e - ring for pitch @xmath8 ( straight filament ) . the width @xmath9 of the strong e - ring , given by @xmath10 is indicated . the inset illustrates the cylindrical geometry of the 3d model , showing the underlying helical mind filament with its depolymerizing tip at @xmath11 . the helical pitch @xmath12 is indicated . ] as illustrated in the inset of fig . [ fig : profiles ] , we represent the bacterial geometry as a cylinder of radius @xmath13 and length @xmath14 . in the right half ( @xmath15 ) , @xmath16 filaments of mind are placed on the cylinder , each with the same helical pitch @xmath12 but with random ( unbundled ) helical phases . mind filaments are composed of monomers of length @xmath17 , each of which can bind one mine . we depolymerize mind from filament tips at @xmath11 , and any released mine diffuses in the cylinder interior ( cytoplasm ) with a diffusion constant @xmath18 . released mine can bind to unoccupied mind monomers ; if not it is removed from the system at @xmath19 . this open boundary condition represents the sinks for mine provided by other mind in the system . depolymerized mind is removed from the system without further interaction , reflecting the nucleotide exchange needed before mind rebinding is possible . this dramatically simplifies our model , since we may then explicitly consider only mine dynamics on an implicit mind filament . both the boundary conditions and the neglect of depolymerized mind will be addressed again in the discussion . in order to study a steady - state e - ring , we keep the filament tips centered at @xmath11 the `` tip - frame '' . in the tip - frame , bound mine move along mind filaments at a constant depolymerization speed @xmath20 while new monomers of mind are introduced at @xmath21 decorated with mine with a constant probability @xmath22 ( determined by the relative cellular amounts of mine and mind particles ) . [ effectively we are studying semi - infinite mind filaments under the approximation of uniform mine binding for @xmath23 . ] in the steady - state , the fraction of mine released by the depolymerizing mind filament tip that reach the absorbing boundaries will then be @xmath24 , where @xmath25 is the fractional mine occupation of the filament tip . the depolymerization speed @xmath20 can be determined self - consistently by @xmath25 , though we will see below that @xmath20 is small and can be practically ignored in terms of the e - ring structure . the dimensionless parameter @xmath26 , the fractional axial distance one mine advects at speed @xmath20 while it diffuses a distance @xmath27 , characterizes the importance of the depolymerization speed . even with @xmath28 , @xmath29 @xcite , and @xmath30 @xcite , @xmath31 is small and depolymerization is slow compared to diffusion . accordingly , our stochastic 3d model quasiadiabatically follows each released mine until it either rebinds or is removed from the system before allowing further depolymerization . each mine diffuses by taking a randomly oriented step of fixed length @xmath32 every timestep @xmath33 , where @xmath34 . diffusing mine binds to a free mind with probability @xmath35 when it hits the bacterial membrane within a distance @xmath36 of the mind . we take @xmath37 . this leads to an effective binding rate of @xmath38 , where @xmath39 is the local bulk concentration of mine and the bulk reaction rate @xmath40 . we take @xmath41 @xmath42 ( this is approximately the threshold between strong and weak e - rings given the cell geometry , see below ) . the steady state reached after successive depolymerization steps is independent of small @xmath32 if we vary @xmath35 with @xmath32 to keep @xmath43 constant . we also study a deterministic 1d model that exactly corresponds to the 3d stochastic model in the limit @xmath44 . this enables us to explore the role of spatial dimension and stochastic effects in the e - ring , and also helps us to identify the combinations of parameters that control the e - ring structure . our 1d model tracks both the linear density of bound mine ( b ) and of freely diffusing mine @xmath45 : @xmath46 where the dots and primes indicate time and spatial derivatives , respectively . for @xmath47 there are no filaments so @xmath48 . for @xmath49 , the linear density of potential binding sites ( i.e. of mind ) is @xmath50 , and the 1d rebinding rate is @xmath51 . the @xmath20 dependent terms on the left side of the equations represent advection of bound mine in the tip frame , while on the right of eqn . [ eqn : unscaledf ] is a source term due to mine release at the depolymerizing filament tip . if we rescale all lengths by @xmath52 ( so @xmath53 ) and define dimensionless fields @xmath54 and @xmath55 , then we can consider the scaled steady - state equations : @xmath56 the boundary conditions are @xmath57 and @xmath58 . the behavior is controlled by the dimensionless parameters @xmath59 and @xmath60 , as well as by @xmath22 . we integrate eqn . [ eqn : basicf ] for @xmath61 where @xmath62 , and impose flux conservation of mine at the boundaries with @xmath63 . for @xmath64 the equations are then integrated numerically to find the steady - state . following the discussion of the stochastic 3d implementation , we expect @xmath65 to be small , and anticipate that it is irrelevant for the e - ring structure leaving only @xmath22 and @xmath66 as relevant control parameters . nevertheless , the 1d treatment allows us to explore this assumption . we find that @xmath67 does not change the observed e - ring steady - state structure by eye , while we expect @xmath68 at room temperature _ in vivo _ and even lower values for weak e - rings . the four - fold speedup observed for the min oscillation at body temperature @xcite puts the depolymerization speed ( i.e. @xmath69 ) closer to , but still under , relevance with respect to the structure of the steady - state mine ring . we can compare results of our 1d deterministic model with our 3d stochastic model using @xmath70 , where @xmath71 is the bulk density averaged over the bacterial cross - section . then the 1d and 3d binding rates of mine are related by @xmath72 , where @xmath73 . we expect that @xmath74 will vary with distance from the filament tip due to local release at the tip followed by diffusion and capture . we find @xmath75 away from the filament tip due to rebinding to the mind filament , and we expect @xmath76 at the filament tip due to local release from the depolymerizing tip . effects of multiple filaments ( @xmath16 ) and filament pitch ( @xmath12 ) can be included in the 1d model by using the mind monomer spacing projected along the bacterial axis @xmath77 , where @xmath78 differences between the two approaches are either due to the 1d vs. 3d geometry or due to the deterministic vs. stochastic nature of the models . [ fig : profiles ] illustrates the fractional occupation @xmath79 ( equivalent to @xmath80 in the 1d model ) of mine binding sites on the mind filament vs. distance @xmath1 along the bacterial axis . occupation monotonically decreases from the tip value , @xmath81 , due to local rebinding of mine following depolymerization from the tip . following the quantification of shih _ et al _ @xcite , there are a few thousand mind monomers within a typical bacteria . with @xmath82 m and @xmath83 nm @xcite , they can be arranged either in one single helical filament ( with @xmath84 m @xcite ) or about 7 straight filaments ( with @xmath8 ) . in either case , we find ( a ) a `` strong '' e - ring ( @xmath85 shown ) with @xmath86 and a plateau shape of the density profile near the tip . with only one straight filament ( b ) , we find a `` weak '' e - ring with enhanced density at the tip but no saturation ( @xmath87 ) and no plateau . strong or weak e - rings have , respectively , negative ( @xmath88 ) or positive ( @xmath89 ) curvature at the tip . the 1d model profiles reasonably match the 3d results away from the filament tips , using @xmath6 . this best value of @xmath74 depends on @xmath36 . using the same @xmath74 near the tips , the 1d model systematically underestimates the fractional occupation . this implies that a larger @xmath73 is appropriate there , in agreement with the increased likelihood that mine will be found near the tip shortly after it is released at the tip . and @xmath25 , as a function of the scaled aspect ratio @xmath0 . ( a , b ) : straight filaments ( @xmath90 , @xmath4 ) . 1d model results are shown with solid lines ( using @xmath6 ) ; 3d stochastic results are indicated by symbols . single filaments ( @xmath85 , green data ) for @xmath91 m ( @xmath92 ) , @xmath2 m ( @xmath93 ) and @xmath94 m ( @xmath95 ) ; multiple filaments ( @xmath962 , or 5 , blue data ) for @xmath97 m ( @xmath98 ) . ( c , d ) : helical filaments ( @xmath85 , @xmath4 ) with @xmath9920 ( @xmath92 , red ) , 10 ( @xmath93 , green ) , 4 ( @xmath95 , blue ) and 0 ( @xmath100 , pink ) . for all these data , @xmath101 , and @xmath13 is varied to explore @xmath0 . similar results are obtained when @xmath43 is varied . ] for both strong and weak e - rings , we can define the width @xmath9 of the e - ring such that @xmath10 . motivated by the importance of the scaled mine rebinding rate @xmath102 in the 1d deterministic equations and by the correspondence of @xmath51 and @xmath43 , we investigated the influence of the scaled aspect ratio @xmath103 on the profile shape , as characterized by @xmath104 and by @xmath25 , in fig . [ fig:2regimes ] for both the 3d stochastic model ( symbols ) and the 1d deterministic model ( lines ) . two regimes are demarcated by a vertical dashed line : for small @xmath0 we have a strong e - ring with @xmath88 , a saturated tip ( @xmath86 ) , and good agreement between the 1d and 3d models for the e - ring width ; for larger @xmath0 we have a weak e - ring with @xmath89 , @xmath25 no longer saturated , and a smaller width @xmath9 . the agreement between the 3d and 1d results for @xmath25 and @xmath9 at small @xmath0 shows that the essential physics of strong e - rings is one - dimensional . for small enough @xmath13 the bacterial cross - section is well explored by mine by the time it has diffused to free binding sites a distance @xmath9 from the filament tip . however , by effectively averaging the radial profile the 1d model systematically underestimates the occupation fraction near the tip , as seen with @xmath25 in fig . [ fig:2regimes ] and also in the profiles shown in fig . [ fig : profiles ] . the disagreement becomes stronger as @xmath0 increases , reflecting the increasingly 3d character of the stochastic system at larger aspect ratios . however , the system still exhibit a remarkable collapse for all values of @xmath0 . this shows that although the 1d model misses important details about the tip enhancement , the scaling behaviour of the 3d system with straight filaments is similar to the 1d model . as shown in fig . [ fig:2regimes](c , d ) , @xmath0 also captures the effects of helical mind filaments . smaller pitches lead to stronger e - rings . however , the 3d stochastic results do not show scaling collapse with respect to @xmath0 as the monomer spacing along the filament @xmath17 is a relevant length - scale in addition to the projected axial monomer spacing @xmath77 . since the 1d model only uses the effective @xmath77 , it incorrectly exhibits perfect scaling collapse . and @xmath25 obtained by the 3d model ( points ) and the 1d model ( lines , using @xmath6 ) , as a function of @xmath0 for different values of stoichiometry ; @xmath105 ( @xmath92 , continuous lines ) , 0.35 ( @xmath93 , dashed lines ) , 0.50 ( @xmath95 , dotted lines ) . for each stoichiometry , the same collapse as fig . [ fig:2regimes](a , b ) is obtained : data are compiled for @xmath16=1 , 2 , 3 , 4 and 5 , @xmath52=1 , 2 and 3@xmath106 m , @xmath90 , @xmath101 , and @xmath13 varies to explore @xmath0 . similar results are obtained when @xmath43 is varied . ] as shown in fig . [ fig : family](a ) and ( b ) , @xmath22 ( the ratio of the number of mine and mind particles ) also controls the scaling curves of @xmath104 or @xmath25 vs. @xmath0 . agreement between 1d and 3d models for small @xmath0 and scaling collapse are preserved for each @xmath22 . experimentally , @xmath107 is observed in rod - shaped cells @xcite , where we take @xmath52 as half the bacterial length . using @xmath108 , which is consistent with the ratio of mine to mind if we assume mine are always dimerized @xcite , then from fig . [ fig : family](a ) we see that @xmath107 is recovered for @xmath109 which corresponds to @xmath110 . ( this @xmath43 is of the same order of magnitude as used in a number of previous models in 3d @xcite and in 1d @xcite if we assume @xmath111 . ) interestingly , this indicates that the e - ring of the normal wild - type ( wt ) min oscillations is a strong e - ring ( with a plateau of mine occupation near the mind filament tip ) but near the margin between weak ( with @xmath87 ) and strong . this implies ( see eqn . [ eqn : deltat ] below ) that the tip occupation @xmath25 , and hence the depolymerization speed and the oscillation period , will strongly depend on the stoichiometry of mine to mind . since @xmath112 , changes to @xmath25 even at the percent level should be significant . indeed , mind overexpression leads to a 2.5-fold increase in the period @xcite . this also implies from fig . [ fig : family](a ) that the width of the e - ring will strongly depend on the stoichiometry though this has not ( yet ) been explored experimentally . at a fixed stoichiometry of mine to mind ( @xmath22 ) , we expect that overexpression of min will increase the number of filaments and/or decrease the pitch . as a result , we expect a slightly stronger e - ring , and a slightly faster period as seen @xcite . optically reconstructed e - rings @xcite show a plateau - like decoration along the mind filament , consistent with a strong e - ring . e - rings have also been seen in long filamentous cells @xcite and exhibit approximately the same width @xmath9 , though both the spacing between mind caps and the cell length are considerably longer in filamentous cells than in rod - shaped cells . this indicates that the effective @xmath52 may not be determined by cell shape , but rather by other processes preserved between rod - shaped and filamentous bacteria such as the length of the mind filaments or spontaneous lateral release ( without mind hydrolysis ) of mine away from the tip of the mind filament . shih _ et al . _ @xcite identified mine point - mutants ( mine@xmath113 and mine@xmath114 ) that led to fainter e - rings , and double mutants ( mine@xmath115 ) that resulted in most of the mine being cytoplasmic with no strong e - rings . assembly and disassembly of mind polar zones continued with no more than doubled periods @xcite too rapid to be explained by intrinsic depolymerization alone ( in contrast , see @xcite ) . from eqn . [ eqn : deltat ] ( below ) the observed disassembly rates would only require a moderately enhanced @xmath116 , i.e. a weak e - ring . indeed , in all of these constructs there appears to be enhanced co - localization of mine with the mind polar zones @xcite . we believe that the lack of visible e - rings in these mutants can be explained with decreased @xmath43 ( as suggested previously by @xcite ) and/or enhanced spontaneous mine unbinding away from filament tips . local rebinding of mine near filament tips would still lead to an enhanced @xmath25 . we predict that the oscillation period in these mutants should be strongly susceptible to the mine to mind stoichiometry . we may use our models to check that the transients before steady - state are fast enough in the context of the normal min oscillation . if we initially decorate the mind filament with mine monomers released from @xmath117 consistent with mine released from a different depolymerizing mind cap , we find ( data not shown ) an initial decoration pattern that has a plateau - like strong e - ring from the beginning ( as previously noted @xcite ) , so that we expect rapid e - ring formation without appreciable delay during min oscillations ( as also observed experimentally@xcite ) . while delays are not observed for e - ring formation during min oscillations _ in vivo _ , significant delays are observed _ in vitro_. mind binds to phospholipid vesicles in the presence of atp and undergoes self - assembly , constricting the vesicles into tubes with diameters on the order of 100 nm @xcite . electron - microscopy revealed that mind assembles into a tightly wound helix on the surface of these tubulated vesicles with a pitch ( helical repeat distance ) of only 5 nm . @xcite report a significant delay ( several minutes ) for stimulated atpase activity when small concentrations of mine were added , while this delay vanished for larger mine concentrations . similar delays were seen _ in vitro _ by suefuji _ furthermore , the eventual steady - state atpase activity was smaller for smaller concentrations of mine @xcite . this has led to the hypothesis of explicit cooperativity of mine binding , which has then been explicitly included in reaction - diffusion models@xcite and in mine polymerization in models with mind polymers @xcite . here we show that our stochastic model for the mine ring , with no explicit mine cooperativity , can recover the mine concentration dependent atpase delays and activities observed _ in vivo_. we conclude that mine cooperativity is not needed to explain the _ in vitro _ results , apart from cooperative effects that arise implicitly from the self - organization of the mine ring . we use an `` inside - out '' open geometry corresponding to what is reported _ in vitro _ @xcite , with a narrow phospholipid cylinder that is tightly wound by mind filaments . mine , when released by a depolymerizing filament tip , will diffuse _ outside _ the cylinder . we consider a helical mind filament of radius @xmath118 nm and pitch 5 nm ( equal to @xmath17 ) . upon mind depolymerization , we allow any released mine to diffuse until either it binds to an available mind binding site or it is absorbed by the boundaries at @xmath19 . we impose reflecting boundary conditions at @xmath119 , but otherwise allow mine to diffuse freely for @xmath120 . our stochastic 3d model is otherwise the same as before though with an emphasis on the transients approaching steady - state . nm ) with open boundaries at @xmath121 . ( a ) evolution of @xmath25 as a function of the number of depolymerization steps @xmath122 ( measured in thousands ) after the uniform intial conditions for @xmath22 equal to @xmath123 ( solid , red ) , @xmath124 ( long dash , green ) , @xmath125 ( short dash , blue ) , and @xmath126 ( dotted , pink ) ; ( b ) steady - state @xmath127 as a function of axial distance @xmath1 along the helical axis for the same @xmath22 . ] the transient to steady - state is shown in fig . [ fig:40bis](a ) , with the fractional occupation of mine at the mind filament tip ( @xmath25 ) shown as a function of the number of depolymerized monomers from the filament tip , @xmath122 . the mine occupation fraction at the mind filament tip is experimentally observable through the atpase activity ( i.e. the mind depolymerization rate ) . the initial condition is a uniform occupation @xmath22 , corresponding to an initially random binding of mine on the mind filament . the larger @xmath22 is , the shorter the transient and the stronger the eventual steady - state @xmath25 . significant enhancement of @xmath25 is obtained even for small fractions of mine . for @xmath128 we see that @xmath129 , though , as shown in fig . [ fig:40bis](b ) , strong e - rings are predicted only for very large stoichiometry ( @xmath130 ) . the inside - out _ in vitro _ geometry includes some small radius features ( the helical winding of the mind filament ) and some large radius features ( no closed boundary at large @xmath131 ) . the tight helical winding of the mind filament contributes to long transients , while the semi - infinite radial geometry contributes to the weak e - ring for small and moderate @xmath22 . ( measured in thousands of depolymerization steps ) versus time @xmath132 for various @xmath22 . asymptotic behavior are plotted as thin dotted lines . ] to convert the number of depolymerization steps @xmath122 to a time @xmath133 we need to sum the average time for each step , which will depend on @xmath25 : @xmath134 where , @xmath135 the timesteps are determined by @xmath136 when the tip of the mind polymer is unoccupied by mine and @xmath137 when it is occupied . using @xmath138 @xcite and @xmath139 given by the maximal depolymerization speed _ in vivo _ ( assuming @xmath86 , with a strong e - ring ) @xcite , we plot the cumulative total atpase activity @xmath140 ( equal to the number of depolymerization steps ) vs. elapsed time @xmath132 in fig . [ fig : invitro ] . the stoichiometric ratio of mine to mind corresponds to @xmath22 if the mine mostly binds to available mind before depolymerization proceeds significantly . for small amount of mine ( typically @xmath141 ) we obtain a significant delay of about 5 minutes , corresponding to the atpase delay seen _ in vitro _ @xcite ; and for larger mine amounts ( @xmath22 going to 1 ) the delays decrease towards zero also in agreement with _ in vitro _ studies . when the steady - state @xmath25 is reached , the atpase rate will also be in a steady - state as indicated by the linear asymptotes in fig . [ fig : invitro ] . since @xmath25 can be large even for smaller @xmath22 , we expect the atpase rates to be comparable for moderate or larger @xmath22 , as seen _ in vitro _ @xcite . for smaller @xmath22 the steady - state atpase activity is reduced , as also observed . we conclude that the delay of atpase activity seen _ in vitro _ is determined by the time needed to reach the steady - state @xmath25 . we see that it is considerably longer in an open than in a closed geometry . our local rebinding model recovers the delays seen _ in vitro _ without any explicit mine cooperativity ( see , conversely , @xcite ) . in the tip - frame , the mind filament tip is bistable during min oscillations @xcite and the formation of the e - ring switches the filament tip between polymerization and depolymerization . while long transients for this switching are not expected during min oscillations _ in vivo _ because of initially non - uniform tip decoration @xcite , we may ask about the transient to form the e - ring from a non - oscillating state such as seen experimentally after exposure to high levels of extracellular cations @xcite . we consider a mind filament that is initially uniformly decorated with mine . to tractably include the mind polymerization dynamics , we use a uniform ( mean - field ) bulk mind density @xmath142 . because we are interested in the initial slow stages of e - ring formation , we consider mine binding only near the tip with occupation fraction @xmath25 ( initially equal to @xmath22 ) the net polymerization rate of a mind filament is @xmath143 , where @xmath142 is the bulk mind monomer concentration and @xmath144 controls mind monomer addition . depolymerization of @xmath16 monomers from a single tip will enhance @xmath25 due to local rebinding of mine , so that @xmath145 for cell volume @xmath146 . the depolymerization time per monomer is @xmath147 for an initially weak e - ring ( with @xmath25 small ) , and the change in tip occupation in one depolymerization step will be proportional to both the number of mine released ( @xmath25 ) and the locally available binding sites ( @xmath148 ) , so that @xmath149 where the constant @xmath150 is the fraction of mine that rebind to available sites at the filament tip . for @xmath137 sufficiently greater than @xmath136 this represents an instability ( @xmath151 growing more negative with time ) that will lead to e - ring formation . we therefore expect that both a significant difference between intrinsic _ and _ stimulated atpase activity of mind and significant intrinsic atpase activity are needed for e - ring formation , and hence for the initiation of min oscillations . we have neglected any lateral unbinding of mine from the mind filament , which will kill the instability if @xmath151 is small enough . we also neglect the presence of other mind filament tips , which will buffer the bulk mind density and reduce the effect of the @xmath144 term in eqn . [ eqn : instability ] . these effects will shift the threshold , but will not change the presence of the e - ring instability . since @xmath152 initially , we also predict from eqn . [ eqn : instability ] that _ both _ low and high proportions of mine to mind will also preclude min oscillations by making the mind filament tip initially stable against depolymerization . however , using @xmath153 @xcite , @xmath154 , and @xmath155 we estimate a tiny stoichiometry threshold of @xmath156 ( for @xmath22 or @xmath157 ) . while our predicted stoichiometry thresholds are unlikely to be relevant _ in vivo _ , they may be approachable _ in vitro_. we also note that initially slow e - ring formation dynamics near the instability threshold should be observable when min oscillations are restarted after being halted @xcite . previous models of the full min oscillation have found limiting mine : mind stoichiometries , either both low and high @xcite or just high @xcite . sufficiently low stoichiometries may not have been explored in the later models . conversely , min oscillations have always been seen _ in vivo _ with moderate stoichiometry changes @xcite . it would be desirable for a more systematic exploration of the role of stoichiometry on min oscillations , given the predicted stoichiometry limits for the existence of oscillations predicted in this and other models . we have presented a model of the self - assembly of the mine - ring within single _ e. coli _ bacteria , without invoking either mine cooperativity or mine polymerization . we highlight the difference between strong e - rings , with @xmath86 , essentially 1d physics and a maximal depolymerization speed , and weak e - rings with @xmath87 that have 3d physics with depolymerization speeds that sensitively depend on the parameters , and especially on the amount of mine in the cell . in contrast to previous filamentous models that had only strong e - rings @xcite , our model shows how changing the stoichiometry of mine and mind can change the oscillation period through the depolymerization speed of mind filaments . mine - rings in non - polymeric reaction - diffusion models @xcite follow essentially our local rebinding mechanism in the 1d regime , but will deviate from polymeric models for weaker e - rings in the 3d regime where the monomer scale @xmath17 enters . since the experimentally measured e - ring width indicates that e - rings _ in vivo _ are close to the threshold between weak and strong , the detailed response of the e - ring structure ( i.e. the width @xmath9 , or the depolymerization speed via the tip occupation @xmath25 ) to experimental manipulations that change the oscillation period ( stoichiometry through @xmath22 or , e.g. , @xcite ) is unlikely to be correctly captured by 1d or non - filamentous models . we have explained the anomalous delays of mine stimulated mind atpase activity seen _ in vitro _ @xcite , and have also identified an instability of mine ring formation that is required to develop from a disordered initial state to the full min oscillation . we have shown that mine - ring structure and dynamics can be treated independently of a full min oscillation model . the instability to e - ring formation , and subsequent mind filament depolymerization , that we identify neither depends on nor determines the spatial pattern of min oscillation which could be selected by either diffusion and rebinding of mind @xcite or by phospholipid heterogeneities @xcite . we have constructed our e - ring model to obtain a steady - state . the steady - state is formed by balancing the mine entering the system as a bound fraction @xmath22 on the mind filament with the mind lost by diffusing across the open boundaries at @xmath158 . other geometries , such as an open boundary at @xmath117 and closed at @xmath21 , or a filament tip placed asymmetrically ( away from @xmath11 ) , will also lead to a steady - state e - ring that should be qualitatively similar to the one we have described . an extreme example of this is the inside - out geometry we used to describe _ in vitro _ atpase experiments . what we have accomplished is to characterize the steady - state , and use it to explore the effects of cell - shape , helical pitch , mine rebinding rate , and stoichiometry on the e - ring structure . our model is expected to be a generic part of full oscillation models that exhibit e - rings . it is worth speculating on how our simplified e - ring model would be modified by possible additional ingredients within a full model of the min oscillation . ( 1 ) we do not expect that filament cutting ( see e.g. @xcite ) will qualitatively affect our results , though it would lead to many more free ends and faster depolymerization . the mine ring would still only be expected to form near the very end of the mind filament , and significant depolymerization would only occur within its width @xmath9 from the end . similarly , our results should apply to models without filaments ( see e.g. @xcite ) . in that case , we expect that our analytic 1d treatment to be a better approximation due to the absence of an intrinsic monomer spacing @xmath17 that is relevant near the filament tip . ( 2 ) we expect that lateral release of bound mine away from filament tips , without associated cutting , would affect the e - ring profile a distance @xmath159 away from the tip ( where @xmath20 is the depolymerization rate , and @xmath160 is the lateral release rate ) . this can be crudely included in our model by placing our boundary conditions at @xmath161 . ( 3 ) we have neglected the rebinding of mind to the filament tip . we would expect rebinding to `` poison '' the e - ring by significantly reducing the depolymerization rate which would allow further rebinding . this appears to be observed in the occasional e - ring reversal _ in vivo _ while interesting , poisoning appears to be typically avoided during min oscillations perhaps by filament cutting or by lateral mine release and re - binding , neither of which have been experimentally characterized and so we are justified in neglecting it for steady - state e - rings . poisoning may however weaken the e - ring instability described by eqn . [ eqn : instability ] , and this deserves further study . the next step is to develop a full 3d min oscillation model with mind filaments but without mine polymerization . previous work has considered the steady - states of semi - infinite filaments with tip - directed depolymerization enhanced by bound motors ( in this paper , bound mine ) @xcite . that work used a uniform ( mean - field ) cytoplasmic motor distribution , and obtained tip - enhanced motor density by a combination of diffusion and directed motion along the filament together with a `` processivity '' retention probability @xmath162 for motors at the depolymerizing tip . in contrast , in our model mine remains immobile on the filament . [ note that advection ( @xmath20 ) represents the dragging of mine along with the mind filament , not motion with respect to the filament . ] furthermore , we explicitly consider the cytoplasmic mine random - walk or diffusion upon release from the filament tip . while this does lead to implicit processivity ( local retention of mine ) , it also correctly allows for rebinding of mine away from the filament tip . this physical modeling of the cytoplasmic mine allows us to consider , e.g. , the 3d vs. 1d cross - over , realistic transients for the inside - out _ in vitro _ geometry , and the e - ring width . note that the enhanced local rebinding of mine to the mind filament upon release is related to ligand rebinding ( see , e.g. , @xcite ) , and similar dimension and geometry dependent effects are seen there . this work was supported financially by natural sciences and engineering research council ( nserc ) , canadian institutes for health research ( cihr ) , and atlantic computational excellence network ( acenet ) ; computational resources came from acenet and the institute for research in materials ( irm ) . we acknowledge useful discussions with manfred jericho . 99 j. lutkenhaus , annu . rev . biochem . * 76 * , 539 ( 2007 ) ; k. kruse , m. howard , and w. margolin , mol . micro . * 63 * , 1279 ( 2007 ) ; m. howard and k. kruse , j. cell . biol . * 168 * , 533 ( 2005 ) . shih , t. le , and l. rothfield , proc . usa * 100 * , 7865 ( 2003 ) . z. hu , e. p. gogol , and j. lutkenhaus , proc . usa * 99 * , 6761 ( 2002 ) . k. suefuji , r. valluzzi , and d. raychaudhuri , proc . usa * 99 * , 16776 ( 2002 ) . j. szeto , n. f. eng , s. acharya , m. d. rigden , and j .- a . r. dillon , res . * 156 * ; 17 ( 2005 ) . d. m. raskin and p. a. j. de boer , cell * 91 * , 685 ( 1997 ) . c. a. hale , h. meinhardt and p. a. j. de boer , embo j. * 20 * , 1563 ( 2001 ) . x. fu , y .- l . shih , y. zhang , and l. rothfield , proc . usa , * 198 * , 980 ( 2001 ) . z. hu and j. lutkenhaus , mol . cell * 7 * , 1337 ( 2001 ) . shih , x. fu , g. f. king , t. le , and l. rothfield , embo j. * 21 * , 3347 ( 2002 ) . m. howard , a. d. rutenberg , and s. de vet , phys . rev . lett . * 87 * , 278102 ( 2001 ) m. howard and a. d. rutenberg , phys . lett . * 90 * , 128102 ( 2003 ) . k. c. huang , y. meir , and n. s. wingreen , proc . usa * 100 * , 12724 ( 2003 ) . h. meinhardt and p. a. j. de boer , proc . usa * 98 * , 14202 ( 2001 ) . k. kruse , biophys . j. * 82 * 618 - 627 ( 2002 ) . n. pavin , h. c. paljetak , and v. krsti , phys . e * 73 * , 021904 ( 2006 ) . f. tostevin and m. howard , phys . * 3 * , 1 ( 2006 ) . d. a. drew , m. j. osborn and l. i. rothfield , proc . usa * 102 * , 6114 - 6118 ( 2005 ) . e. n. cytrynbaum and b. d. l. marshall , biophys . j. * 93 * , 1134 ( 2007 ) . g. meacci , j. ries , e. fischer - friedrich , n. kahya , p. schwille , and k. kruse , phys . biol . * 3 * , 255 ( 2006 ) . a. touhami , m. h. jericho , and a. d. rutenberg , j. bacteriol . * 188 * , 7661 ( 2006 ) d. m. raskin and p. a. j. de boer , proc . usa * 96 * , 4971 ( 1999 ) . k. c. huang and n. s. wingreen , phys . * 1 * , 229 ( 2004 ) . m. loose , e. fischer - friedrich , j. ries , k. kruse , and p. schwille , science * 320 * , 789 ( 2008 ) . the depolymerization speed along the bacterial axis is approximately 30nm / s @xcite . by taking into account the pitch , it gives an approximate speed along the filament of 250nm / s which gives , with @xmath83 nm , an average time per depolymerization on the order of 20ms . b. p. b. downing , a. d. rutenberg , a. touhami , and m. jericho , submitted for publication ( 2009 ) . r. v. kulkarni , k. c. huang , m. kloster , and n. s. wingreen , phys . * 93 * , 228103 ( 2004 ) . e. mileykovskaya and w. dowhan , curr . . microbiol . * 8*,135 ( 2005 ) . g. a. klein , k. kruse , g. cuniberti , and f. jlicher , phys . . lett . * 94 * , 108102 ( 2005 ) . m. gopalakrishnan , k. forsten - williams , t. r. cassino , l. padro , t. e. ryan , u. c. tuber , eur . j. * 34 * , 943 ( 2005 ) .

we model the self - organization of the mine ring that is observed during subcellular oscillations of the proteins mind and mine within the rod - shaped bacterium _ escherichia coli_. with a steady - state approximation , we can study the mine - ring generically apart from the other details of the min oscillation . rebinding of mine to depolymerizing mind filament tips controls mine ring formation through a scaled cell shape parameter @xmath0 . we find two types of e - ring profiles near the filament tip : a strong plateau - like e - ring controlled by 1d diffusion of mine along the bacterial length , or a weak cusp - like e - ring controlled by 3d diffusion near the filament tip . while the width of a strong e - ring depends on @xmath0 , the occupation fraction of mine at the mind filament tip is saturated and hence the depolymerization speed do not depend strongly on @xmath0 . conversely , for weak e - rings both @xmath0 and the mine to mind stoichiometry strongly control the tip occupation and hence the depolymerization speed . mine rings _ in vivo _ are close to the threshold between weak and strong , and so mind - filament depolymerization speed should be sensitive to cell shape , stoichiometry , and the mine - rebinding rate . we also find that the transient to mine - ring formation is quite long in the appropriate open geometry for assays of atpase activity _ in vitro _ , explaining the long delays of atpase activity observed for smaller mine concentrations in those assays without the need to invoke cooperative mine activity .

introduction e-ring model results transients discussion acknowledgments

a large class of extensions of general relativity is described in the context of scalar - tensor theories of gravity @xcite , where a space - time metric @xmath1 is coupled to a scalar field @xmath2 , with the other matter contents such as fermion fields being minimally coupled to gravity . by an appropriate conformal transformation @xmath3 we can move to another frame . a prime example is the einstein frame , where after the conformal transformation ( [ ct ] ) the gravitational action becomes the einstein - hilbert one . these conformal frames are mathematically equivalent , so one can work in any frame as long as mathematical manipulations are concerned . it is , however , not explicitly clear whether these conformally related frames also enjoy physical equivalence @xcite . in cosmology , we encounter various frames of the metric which are related by conformal transformation , such as jordan frame , einstein frame , string frame and so on . the physical equivalence between these frames is especially important , since the cosmological observations with improving sensitivity can probe gravitational theories on the largest observable scales @xcite . further , some models of the early universe which accounts for the primordial perturbations incorporate non - minimal coupling between gravity and a scalar field , such as the recently proposed standard model higgs inflation @xcite . usually one moves from the original , jordan frame to the einstein frame by a conformal transformation , and then computes perturbations there . the invariance of the vector and tensor perturbations , and that of the scalar perturbation in the comoving gauge , are shown up to second order @xcite , but it is not clear if this invariance holds fully non - perturbatively . regarding precise upcoming experiments that may detect non - linear signatures such as non - gaussianity , it is important to clarify the full non - linear invariance of the perturbations @xcite . in this article , we study the conformal invariance of the cosmological perturbation . our particular attention will be given to the curvature perturbation on super - horizon scales . we will show that in the single component case all perturbations in the comoving slicing are conformally invariant to fully non - linear order . once this is given , we can see that in the context of the @xmath0 formalism it is a matter of gauge transformation between different coordinate systems which impose different slicing conditions . as a temporal gauge condition , we take the conformal transformation factor @xmath4 to be unperturbed to full non - linear order . this may be called as the uniform - conformal - transformation slicing , or , uniform @xmath4 slicing ( u@xmath4s ) . it is convenient to decompose @xmath4 into the background and perturbation as @xmath5 , then the u@xmath4s means @xmath6 as the slicing condition . thus we have @xmath7 under this slicing condition it is obvious , almost a tautology , that all perturbed quantities are naturally invariant under the conformation transformation . this result applies even in multiple component situation : see ( [ uomegag - fr ] ) and ( [ uomegag - fphi ] ) for relations implied by this slicing condition to non - linear orders of perturbation . note that we may consider the action given by @xmath8\ , .\ ] ] here , @xmath4 is a function of either the field @xmath2 for @xmath9-type gravity or a function of @xmath10 or @xmath11 for @xmath12-type gravity , where @xmath13 in these cases , the u@xmath4s corresponds to the uniform - field slicing or the uniform-@xmath10 slicing , respectively , and we may call it as the ufs . first , we consider the single field @xmath14-type gravity in which we can show the invariance , especially that of the curvature perturbation . as for our spatial metric , we write @xmath15 where @xmath16 includes vector and tensor perturbations . we choose @xmath17 such that it remains invariant under the conformal transformation as described in detail in appendix [ app : confgamma ] . then , under the conformal transformation ( [ ct ] ) , ( [ spatialmetric ] ) becomes @xmath18\widetilde{\gamma}_{ij } \equiv a^2\omega_0 ^ 2\exp\left(2\overline{{{\cal r}}}\right)\widetilde{\gamma}_{ij } \,,\ ] ] so that the spatial curvature perturbation in the conformally transformed frame becomes simply @xmath19 thus , in the u@xmath4s in which @xmath6 , we have @xmath20 . further , as spelled out in appendix [ app : confgamma ] , the vector and tensor perturbations in @xmath21 defined with respect to a fixed fiducial background metric are also conformally invariant to fully non - linear order . in the single component case , the u@xmath4s implies @xmath22 , the uniform field gauge or the comoving gauge for @xmath14-type gravity : in most scalar - tensor theories of gravity , the conformal transformation is a function of @xmath2 , @xmath23 . this result has an important implication that , for @xmath14 gravity @xcite all perturbation variables in the uniform field gauge are conformally invariant . in ( [ conformalspatialmetric ] ) , we have explicitly shown the invariance to all perturbation orders , and also have shown that why the invariance is natural in the u@xmath4s which is the same as the uniform field gauge , or equivalently comoving gauge , in the single component case . now , we provide a complementary point of view using the @xmath0 formalism @xcite in the universe dominated by a single scalar field @xmath2 . since we will be interested in super - horizon scales in terms of the @xmath0 formalism , we restrict ourselves to zeroth order in @xmath24 , where @xmath25 is the comoving wavenumber of a characteristic perturbation of our interest . we consider a point on a comoving slice on which @xmath22 . hereafter we use the subscript @xmath26 to denote the quantities evaluated on the comoving slices , and likewise @xmath27 and @xmath28 on the flat and @xmath29 slices ( see below ) . when the universe is dominated by a single component @xmath2 , the comoving slice is the same as the u@xmath4s , thus @xmath30 . that is , the comoving curvature perturbation is the same and is thus conformally invariant . until this point we have not yet related the perturbation in the number of @xmath31-folds @xmath0 to @xmath32 . we can compute @xmath33 in the standard manner . from ( [ spatialmetric ] ) , the expansion scalar @xmath34 , viz . the `` local '' hubble parameter is given by @xmath35 where @xmath36 is the lapse function . from this , the number of @xmath31-folds @xmath33 is given by the integral of @xmath37 along the proper time @xmath38 as @xmath39 + { { \cal r}}(t_e ) - { { \cal r}}(t_i ) \equiv n_0 + { { \cal r}}(t_e ) - { { \cal r}}(t_i ) \ , .\ ] ] here , @xmath40 and @xmath41 denote the initial and the final moments respectively . now , we consider the conformally transformed frame . we can take similar steps to obtain @xmath42 and @xmath43 , so that the number of @xmath31-folds in this frame @xmath44 becomes @xmath45 + \overline{{{\cal r}}}(t_e ) - \overline{{{\cal r}}}(t_i ) \ , .\ ] ] note that the background is _ different _ from @xmath46 as @xmath47 = n_0 + \log \left [ \frac{\omega_0(t_e)}{\omega_0(t_i ) } \right ] \equiv n_0 + \delta_0 \ , .\ ] ] here , we make a specific choice for the initial and final moments where we evaluate the number of @xmath31-folds and the curvature perturbation . as a reference point , we set the final moments in both frames to be identical on a comoving slice . that is , @xmath48 . thus , we have @xmath49 and in both frames @xmath50 . this is the conformal invariance of @xmath32 we have seen above . meanwhile , in the frame without overbar , we set the initial moment to be flat so that @xmath51 , and in the other frame `` @xmath29 '' in the sense that @xmath52 . then , in the frame without overbar , we have @xmath53 and in the frame with overbar , @xmath54 that is , in both frames , @xmath55 . despite of the different background numbers of @xmath31-folds , their perturbations are invariant in both frames . notice that the curvature perturbation on a comoving slice @xmath32 and the field fluctuation on a flat slice @xmath56 are related by [ see eq . ( 287 ) in ref . @xcite ] @xmath57 here we remind that we are working at zeroth order in @xmath58 . this relation holds at any arbitrary time . meanwhile , in the @xmath0 formalism , by construction there is no field fluctuation on the final comoving slice where we evaluate the curvature perturbation . thus , we are to relate the curvature perturbation on the _ final _ comoving slice @xmath59 to the field fluctuations on the _ initial _ flat slice @xmath60 . furthermore , using the fact that we are interested in the dynamics along a given background trajectory , we can show that up to second order in @xmath61 , @xmath62 we give detailed steps to find this relation in appendix [ app : phasespace ] . once evaluated on a flat slice , this should be identical to @xmath59 as we have shown in ( [ deltaninv1 ] ) [ or ( [ deltaninv2 ] ) as well ] and is in perfect agreement with ( [ rc_deltaphif ] ) . thus , we conclude that @xmath63 , i.e. the curvature perturbation is conserved on large scales . given the fully non - perturbative invariance of @xmath32 in two frames , we can obtain a useful perturbative relation as follows . for this , it is convenient to insert another comoving slice common to the both frames between the initial and the final moments , and we denote by @xmath64 the time on this slice . since we are interested in the large scale limit where @xmath32 is conserved , the perturbations in the number of @xmath31-folds between the initial moments and @xmath64 in two frames , @xmath65 and @xmath66 , are identical to @xmath32 . this is schematically shown in fig . [ fig : equivalence ] . is conserved on large scales , @xmath67 . the comoving slice on which time is denoted by @xmath64 is common to both frames . their initial slices are set in such a way that in one frame we set the initial slice to satisfy @xmath68 . in the other frame , on the other hand , we demand @xmath69 . we call the former and the latter to be `` flat '' and `` @xmath29 '' , and the quantities evaluated on these slices by a subscript @xmath27 and @xmath28 respectively.,width=453 ] then , from the frame without overbar , suppressing the dependence of @xmath33 and @xmath46 on @xmath70 which is common , for arbitrary @xmath71 we have ) that @xmath72 can be written as a function of @xmath73 , so that we have a single degree of freedom in the phase space . ] @xmath74 - n_0[\phi_0(t ) ] = \sum_n \frac{n_0^{(n)}[\phi_0(t)]}{n ! } [ \delta\phi_f(t)]^n \ , , \ ] ] where the superscript @xmath75 denotes @xmath76-th derivative and the right hand side is evaluated on the flat slice . likewise , we can write another relation , @xmath77 + \delta_0^{(n)}[\phi_0(t)]}{n ! } \left [ \overline{\delta\phi}_f(t ) \right]^n \ , .\ ] ] note that we have the same background field @xmath73 , since it can be redefined in one frame to absorb the background conformal transformation factor @xmath78 . then , essentially now we have three different gauges conditions : @xmath79 , @xmath80 and @xmath22 , which correspond to @xmath29 , flat and comoving gauges respectively . we need a specific conformal transformation @xmath81 only to define the first gauge . thus , using appropriate coordinate transformation laws , we can freely move between different gauges and relate one to another . for example , we consider the coordinate transformation from @xmath82 and @xmath83 to @xmath64 , @xmath84 then , we only have to write the field fluctuations in terms of the translation @xmath85 or @xmath86 . since on the comoving slice there is no field fluctuation by construction , we have @xmath87 \equiv \phi_f(t_f,{\bm x } ) = \phi_0(t_f ) + \delta\phi_f(t_f,{\bm x } ) \ , , \ ] ] and a similar relation holds for @xmath83 . then , using ( [ coordxform ] ) we can write @xmath88 \equiv t_f + { { \cal t}}(t_f,{\bm x } ) \ , , \ ] ] where @xmath83 dependence of @xmath86 on the right - hand side can be removed by iteratively using this relation . then , from ( [ phitc_phitf ] ) , we can find the linear relation between @xmath89 and @xmath90 as @xmath91 this linear relation is in agreement with the well - known linear gauge transformation law @xcite as it should be . note that in fact it does not matter that we necessarily restrict ourselves to the transformation between two flat gauges , and hence we may drop all the subscript @xmath27 . general coordinate transformations at arbitrary time would work as well . in this section , we consider the multiple component situation . we consider an action , @xmath92 \ , . \label{l - original}\ ] ] under a conformal transformation ( [ ct ] ) with the field redefinition @xmath93 , ( [ l - original ] ) becomes @xmath94 \ , , \ ] ] where @xmath95 with @xmath96 . we have @xmath97 in general . we consider two cases : \(i ) the case with @xmath98 . in this case we have @xmath99 , which is an additional scalar field minimally coupled to gravity in the einstein frame . an equivalent formulation in terms of a brans - dicke type scalar field @xmath100 is given in appendix [ sec : fr ] , with its relation to the canonical scalar field @xmath101 in the einstein frame . \(ii ) the case with @xmath102 . in this case we have @xmath103 . hence there is no additional scalar field . we have @xmath104 \ , , \ ] ] where @xmath105 we see that the effect of the conformal transformation is to change the non - linear sigma coupling term and the potential . these conformal transformation properties are presented in appendix a of ref . @xcite . when the universe is dominated by a single component , we have shown the conformal invariance of all perturbations in the comoving slicing , which is equivalent to the u@xmath4s . in the multi - component situation the u@xmath4s is different from the uniform field slicing @xmath106 of a chosen field component @xmath107 or the collective comoving slicing which sets @xmath108 in the energy - momentum tensor ( or @xmath109 in the collective fluid four - vector . it should be noted however that it is in general impossible to uniquely define the comoving slicing in the multi - component situation because @xmath110}$ ] may not be zero ) . consequently , the curvature perturbation in the gauge where @xmath111 is not conformally invariant : denoting the gauge @xmath111 by a subindex @xmath107 , to linear order , we have @xmath112 where @xmath113 so that clearly @xmath114 . nevertheless , even in the multiple component situation we can take u@xmath4s with @xmath115 to arbitrary non - linear order . under this slicing condition we have @xmath116 , and similarly for all the other metric perturbation variables @xcite . let us denote quantities in u@xmath4s by suffix @xmath117 , e.g. @xmath118 . for definiteness , let us consider the meaning of u@xmath4s in the two cases of our interest . since @xmath115 implies @xmath119 , we have the following : \(i ) for @xmath120 , we have @xmath121 thus @xmath122 . \(ii ) for @xmath123 we have @xmath124 this condition should be imposed among all the fields that appear in @xmath10 . here we note that , in the spirit of the @xmath125 formalism , the final comoving hypersurface should be chosen such that there remains only a single adiabatic mode in the perturbation at and after that epoch . once the universe has entered this stage , we also have the conservation of the comoving curvature perturbation @xmath32 . at this stage , the only dynamical degree of freedom coupled to gravity should be encoded in the function @xmath10 , where the u@xmath4s coincides with the comoving slicing ( as discussed below ) on super - horizon scales , and hence @xmath126constant and consequently @xmath4 depend non - trivially on the remaining dynamical degree of freedom . if it is not the case , then @xmath4 will be simply a constant , and the conformal invariance is trivial . ] . all the other degrees of freedom should either be died out by then or be purely isocurvature at and after that epoch in the sense that they have no coupling to gravity whatsoever . conversely , at a stage during which multiple components of matter or fields have their individual dynamics , @xmath118 is not conserved yet . at this stage , whether some quantities are conformally invariant or not is not really important . it is similar to the matter of the gauge choice . some quantities may be conserved in a certain gauge , which may be a mathematically useful fact but has no physical / observational significance . in the present case , @xmath118 and all the other quantities defined in u@xmath4s are conformally invariant . but this would not help us much to understand the physics . it becomes useful and meaningful only after @xmath118 becomes to be conserved . now we address the relation between the comoving slicing condition and the u@xmath4s in the case of @xmath12 gravity , @xmath127 in the case of @xmath12-gravity the comoving slicing _ differs _ from the ufs in general even at linear order . in the case of a minimally coupled scalar field we have @xmath128 , thus @xmath129 implies @xmath130 to fully non - linear order which is the comoving slicing condition . whereas in the case of pure @xmath12 gravity , we have [ see eq . ( 3 ) in ref . @xcite ] @xmath131 \ , . \label{freq}\ ] ] if we define the right hand side as an effective @xmath132 , to the linear order we have [ see eqs . ( 18 ) and ( b2 ) in ref . @xcite ] @xmath133 where @xmath134 . thus , the ufs condition @xmath135 does not imply @xmath136 in general in the original frame . however , one can show that the ufs condition coincides with the comoving slicing condition @xmath136 on super - horizon scales to full non - linear order provided that the decaying mode becomes negligible and there remains only a single adiabatic mode . one can show this with the help of the field equation for @xmath10 given by the trace of ( [ freq ] ) . one finds that it is identical to the background field equation on super - horizon scales , i.e. at each spatial point @xmath137 where a point means a hubble size region , if one replaces @xmath138 by @xmath139 . thus the general solution to full non - linear order is given by @xmath140 . then after the decaying mode has disappeared , @xmath141 becomes a function of @xmath10 itself similar to the case of standard slow - roll inflation . this implies that on ufs on which @xmath142 , @xmath141 is also a function of only @xmath138 . then on ufs we have @xmath143 where @xmath144 is the unit normal to the uniform @xmath10 slice , @xmath145 is the induced metric on the uniform @xmath10 slice , and @xmath146 is the proper time along @xmath147constant [ the shift vector is @xmath148 on super - horizon scales ] . in fact , this leads to the conservation of the curvature perturbation on the comoving slices @xcite . it may be instructive to spell out how this happens at linear order . in the ufs the perturbed field equation gives @xcite @xmath149 thus the fact that there remains only a single adiabatic mode implies @xmath150 , and hence @xmath151 at leading order on super - horizon scales . also the fact @xmath150 implies @xmath152 at leading order . thus the field @xmath10 in ufs is actually given by @xmath153 . in the above , we have considered the case when the field @xmath10 dominates the universe . in general , we may have other fields or matter which may have non - trivial background dynamics . in this multi - component situation , we have no conformal invariance of the comoving curvature perturbation ( if it can ever be sensibly defined ) . in this situation , we can still define a conformally invariant curvature perturbation , @xmath118 in u@xmath4s . as we discussed , however , @xmath118 will not be conserved if the multi - component matter or fields are still dynamical . it becomes to be conserved when the universe becomes dominated by a single adiabatic degree of freedom . assuming this degree of freedom is encoded in @xmath10 , we have the equivalence between u@xmath4s and the comoving slicing , and hence recover the conformal invariance of @xmath32 , namely @xmath126constant . to summarize , we have studied the non - perturbative conformal invariance of the cosmological perturbations . when the universe is dominated by a single component , the u@xmath4s and the comoving slicing coincide and the comoving curvature perturbation @xmath32 is conformally invariant fully non - perturbatively . consistent with the conformal invariance of @xmath32 , we have shown that the @xmath0 formalism gives identical results irrespective of the choice of conformal frames . when the universe is dominated by a multiple matter or field components , the comoving curvature perturbation is no longer conformally invariant . however , at and after the universe has settled down to a unique evolutionary trajectory , i.e. in an era when there remains only a single adiabatic mode , we recover the conservation of the comoving curvature perturbation on super - horizon scales and so it is conformally invariant . we acknowledge the workshop `` cosmological perturbation and cosmic microwave background '' ( yitp - t-10 - 05 ) at the yukawa institute for theoretical physics , kyoto university , where the main part of this work was done , and the workshop `` wkyc 2011 future of large scale structure formation '' at korea institute for advanced study , where this work was initiated and completed . this work was supported in part by a korean - cern fellowship , korea institute for advanced study under the kias scholar program , the grant - in - aid for the global coe program at kyoto university , `` the next generation of physics , spun from universality and emergence '' from the ministry of education , culture , sports , science and technology ( mext ) of japan , jsps grant - in - aid for scientific research ( a ) no . 21244033 , jsps grant - in - aid for creative scientific research no . 19gs0219 , and krf grant funded by the korean government ( krf-2008 - 341-c00022 ) . here we give the decomposition of the spatial metric @xmath154 in a conformally invariant way up to the determinant of @xmath154 . as a fiducial background , we take the flat metric , @xmath155 . we introduce a traceless matrix @xmath156 , and decompose it as @xmath157 where the spatial indices are to be raised or lowered by the flat metric @xmath155 , and @xmath158 and @xmath159 satisfy @xmath160 . we write the spatial metric as @xmath161 where @xmath162 then , we define @xmath163 as the sum of the trace contributions on the exponent , @xmath164 and introduce @xmath165 it is then clear that under the conformation transformation , @xmath166 @xmath167 hence @xmath16 is conformally invariant . in particular , for perturbations on a given background with @xmath168 , the conformal transformation affects only @xmath169 , or equivalently @xmath163 as @xmath170 we can extend the above to the non - flat fiducial metric in a similar way by covariantizing the derivatives , with the definitions of all the variables including that of @xmath163 , ( [ defcalr ] ) , as well as their conformal transformation properties remain the same . in this appendix , we provide a detailed derivation of ( [ deltan_deltaphideltaphidot ] ) . in many occasions , it is usually assumed that the number of @xmath31-folds @xmath33 is only a function of @xmath2 and the dependence on @xmath172 is neglected . but as can be read from ( [ rc_deltaphif ] ) then we lose another independent degree of freedom in the phase space and can not find @xmath173 dependence of @xmath0 . to find the correct dependence of @xmath0 on @xmath173 , we proceed as follows . in the phase space of @xmath2 , we consider that @xmath2 and @xmath172 are functions of two coordinates , @xmath33 and @xmath174 . being the number of @xmath31-folds , @xmath33 describes where we are on a given background trajectory , while @xmath174 labels which background trajectory we follow . then , formally we can write @xmath61 as @xmath175 we can do the same for @xmath173 , but using @xmath176 we have @xmath177 multiplying by @xmath0 and using this relation to replace @xmath178 term in ( [ deltaphi ] ) , we have @xmath179 where we have omitted higher order derivative terms as well as pure @xmath180 terms , because now we only consider the perturbation _ along _ a given trajectory . therefore , from ( [ deltaphi2 ] ) we can write @xmath0 up to second order as @xmath181 perturbatively expanding , we can find @xmath0 as @xmath182 where we have reversed the time order , @xmath183 , because in the context of the @xmath0 formalism it is defined to be the variation of @xmath33 due to the variation of the _ initial _ conditions . thus we can correctly extract the @xmath173 dependence of @xmath0 . note that we could have found this dependence by considering @xmath172 as an independent degree of freedom in the phase space . once the trajectory follows the attractor , @xmath72 can be written as a function of @xmath73 and we may follow the naive expansion @xmath184 . it is known that @xmath12 gravity can be cast into the form of a brans - dicke type theory . here let us recapitulate it . we can rewrite the @xmath12 gravity action , ( [ eq : fr ] ) , by introducing an auxiliary scalar field as @xmath185\,.\ ] ] then the variation with respect to @xmath174 gives @xmath186 , and we recover the original action . on the other hand , if we take the variation with respect to @xmath187 , we obtain the constraint , @xmath188 using this to eliminate @xmath174 from the above action , we obtain @xmath189\ , . \label{eq : fr2}\ ] ] if we further introduce a scalar field @xmath100 by @xmath190 , then the above action is rewritten again as @xmath191\ , , \label{bdaction}\ ] ] where @xmath192 . this is a brans - dicke scalar - tensor theory with a non - vanishing potential @xmath193 , with the brans - dicke parameter @xmath194 . we have not performed a conformal transformation so far . hence the metric is still the original metric @xmath1 . therefore the @xmath12 action ( [ eq : fr ] ) and the @xmath196 action ( [ bdaction ] ) are completely equivalent to each other . n. makino and m. sasaki , prog . phys . * 86 * , 103 ( 1991 ) ; + s. koh , j. korean phys . * 49 * , s787 ( 2006 ) [ arxiv : astro - ph/0510030 ] ; + n. sugiyama and t. futamase , phys . d * 81 * , 023504 ( 2010 ) .

we show that in the single component situation all perturbation variables in the comoving gauge are conformally invariant to all perturbation orders . generally we identify a special time slicing , the uniform - conformal transformation slicing , where all perturbations are again conformally invariant to all perturbation orders . we apply this result to the @xmath0 formalism , and show its conformal invariance .

introduction single field case multi-field case summary conformal decomposition of the spatial metric trajectory in phase space and @xmath171 @xmath12 gravity in brans-dicke form

introduction of a minimum length scale @xcite-@xcite has revealed several salient features of gravity and quantum field theory . the essence of a minimum length can be captured by various means - by a generalized uncertainty principle ( gup ) , by a modified dispersion relation ( mdr ) , by a deformed ( doubly ) special relativity ( dsr ) etc . one can initially start from a modified momentum operator and successively obtain a gup or an mdr @xcite . the manifestation of the presence of a minimal length in different fields , such as black hole thermodynamics @xcite -@xcite , relativistic and non - relativistic quantum mechanics @xcite-@xcite etc , has stimulated a large number of extensive studies . in nonrelativistic quantum mechanics , a particle with energy @xmath3 bounces from potential region of height @xmath4 if @xmath5 . for a relativistic particle if one keep increasing the potential , the particle can once again propagate in the potential region . this is known as @xmath6 @xcite . in this paper we study the klein paradox in presence of a new gup which imply a minimum length as well as a maximum energy . we start from a modified momentum operator and derive the mdr for a dirac particle . we show that apart from the forbidden energy band in between @xmath0 ( m is the mass of the particle ) as predicted by standard relativistic quantum mechanics , there are other forbidden regions at some high ( both positive and negative ) energies which are of the order of planck energy ( @xmath1 ) . the wave vector has an upper bound which is consistent with the existence of a minimum length scale . the allowed energy band is localised within finite energy range which gives rise to interesting features of the klein paradox . we further study the case of a massless particle and find some new insight . according to non - gup relativistic quantum mechanics , a massless dirac particle can penetrate a potential barrier of any arbitrary height . in presence of our gup the situation is entirely different from that . we find that a massless particle can not penetrate a potential barrier of the order of @xmath1 which we propose to be a signature of the existence of minimum measurable length and maximum measurable energy . the organisation of paper is as follows . in section 2 , we introduce a modified momentum operator and obtain the corresponding dirac equation with a discussion on the structure modified commutation relation and gup . in section 3 , we find the solution for the modified dirac equation and derive the mdr which is used to find the band structure . this band structure is used to explain the nature of the klein paradox in section 4 . in section 5 we investigate the scenario for a massless particle . section 6 contains the discussions . in this section we will start from a modified momentum operator and derive a new gup and corresponding modified dirac equation . a gup corresponds a modified commutation relation between position and momentum . one way to achieve this is to define a modified momentum operator whose structure is determined by the gup . in a recent paper @xcite we have proposed a gup based on some heuristic arguments . the gup contains only even terms on the right hand side . this is needed for our prescription , but not explicitly necessary for a gup . here we use a more general form which contains both odd and even terms . this kind of gup is recently proposed in @xcite . it not only accommodate a minimum length , but also a maximum measurable momentum ( and hence a maximum energy ) . we show a systematic derivation of the gup starting from a modified momentum operator and consecutively derive the dirac equation in presence of such a gup . we start by defining a new momentum operator(@xmath7 ) , keeping first two correction terms @xmath8 where the suffix @xmath9 ( @xmath10 ) corresponds the @xmath11 component . @xmath1 is the planck energy . @xmath12 and @xmath13 are real positive numbers . @xmath14 is the magnitude of the operator @xmath15 . @xmath16 obeys the standard commutation relation @xmath17=i\hbar \delta_{ji } \label{xp}\end{aligned}\ ] ] @xmath18 being the @xmath19 component of position operator . it would be convenient for future calculations to introduce the scaled coefficients @xmath20 and express ( [ p0 ] ) in terms of them . the modified momentum operator can be recast as @xmath21 we define the new position operator as @xmath22 these modified momentum ( @xmath23 ) and position operator ( @xmath24 ) satisfy the commutation relation @xmath25 = i\hbar \left ( \delta_{ij } - a \left ( \delta_{ij } p + \frac{p_i p_j}{p}\right ) + b \left ( \delta_{ij } p^2 + 2p_ip_j \right)\right ) \label{com } \end{aligned}\ ] ] we shall now obtain the modified dirac equation . the dirac equation in presence of the modified momentum operator ( [ p3 ] ) is given by @xcite @xmath26\psi \label{md}\end{aligned}\ ] ] where we have used dirac s prescription for the operator @xmath14 ( i.e @xmath27 ) . @xmath28 and @xmath29 are the dirac matrices defined as @xmath30 where @xmath31 is the @xmath32 pauli matrix ( @xmath9 = 1 , 2 , 3 ) . for simplicity we will consider one dimensional motion ( say along z axis ) . in one dimension the modified momentum operator ( [ p3 ] ) and the commutation relation ( [ com ] ) becomes @xmath33 } & = & i\hbar \left ( 1 - 2ap + ( 3b-2a^2)p^2 \right ) \ \label{com2 } \end{aligned}\ ] ] the gup in one dimension in terms of the new operators @xmath23 and @xmath24 is @xmath34 \right\rangle \vert \nonumber \\ & \geq & \frac{\hbar}{2 } ( 1 - 2 a \left < p \right > + ( 3 b-2a^2 ) \left < p^2 \right > ) \label{gup0}\end{aligned}\ ] ] note that by putting @xmath35 we can obtain the gup proposed in @xcite . for minimum position uncertainty , we can replace @xmath36 and @xmath37 by @xmath38 and @xmath39 @xcite and hence the gup becomes @xmath40 the existence of minimum position uncertainty ( i.e. @xmath41 to have a minima ) requires the coefficient of @xmath39 to be positive . i.e @xmath42 in @xcite , they have used @xmath35 which is also consistent with the above condition ( [ con1 ] ) . for a particle moving along @xmath43 direction ( [ md ] ) simplifies as @xmath44 \psi = e_g \psi \label{mde1}\end{aligned}\ ] ] where @xmath45 is the gup corrected energy of the system . for future purpose we will replace the momentum operator @xmath14 by its differential form @xmath46 and recast the modified dirac equation as @xmath47 \psi = e_g \psi \label{mde}\end{aligned}\ ] ] note that the modified dirac equation ( [ mde ] ) contains third order spatial derivative and a complete solution requires the continuity of higher derivatives also . since the gup correction is very small one can also take a perturbative approach @xcite-@xcite as well . in this section we will find a solution of the modified dirac equation ( [ mde ] ) and subsequently derive the mdr . this will be used to find the energy band structure . the standard one dimensional dirac equation for a free particle is given by @xmath48 note that one can also obtain this equation by putting @xmath49 in ( [ mde ] ) . its positive energy ( + e ) solution is given by @xmath50 here @xmath51 is a @xmath52 column matrix which satisfies orthonormality condition @xmath53 , @xmath54 is the normalization factor and @xmath55 is a dimensionless quantity defined as @xmath56 the wave number ( @xmath57 ) obeys the dispersion relation @xmath58 for the modified dirac equation ( [ mde ] ) , the solution would not be so simple . whatever the solution be , it must converge to ( [ psi ] ) in absence of gup . let us consider the following ansatz @xmath59 to be a solution of ( [ mde ] ) . the assumption @xmath60 where @xmath61 is given by ( [ psi ] ) , requires @xmath62 where @xmath55 is given by ( [ u ] ) . besides , the modified dispersion relation should also converge to ( [ dr ] ) . i.e. @xmath63 we now substitute ansatz ( [ psig ] ) in ( [ mde ] ) and obtain the gup corrected expression for energy ( @xmath45 ) or the mdr and @xmath64 which are given by @xmath65 one can readily see that both @xmath45 and @xmath64 satisfy ( [ anze ] ) and ( [ anzu ] ) . the condition for the existence of a positive energy solution is @xmath66 which can be simplified to @xmath67 if we consider particles for which @xmath68 and @xmath69 then the condition reduces to @xmath70 notice that in @xcite , @xmath71 which also satisfies this condition . we can refine this condition by using ( [ con1 ] ) , from which one can easily obtain @xmath72 since the coefficient @xmath73 is positive , from ( [ con3 ] ) we get @xmath74 therefore the existence of a minimum length scale also asserts the existence of a positive energy solution . the expression for the wave vector @xmath75 can be obtained from ( [ mdr ] ) by ignoring @xmath76 term as @xmath77\end{aligned}\ ] ] since we are interested up to the terms @xmath78 , the above relation could be further simplified as , @xmath79 \label{k}\end{aligned}\ ] ] note that here we have considered only negative ( - ) part of the square root term , because it is consistent with the fact that in the limit @xmath80 it gives @xmath81 where we used the fact that @xmath82 . let us discuss some properties of ( [ k ] ) and let see what they say about the gup ( [ gup ] ) . let us first consider the denominator @xmath83 . using the basic definition ( [ ab ] ) we can rewrite this as @xmath84 . it clearly shows that as the particle energy @xmath45 approaches @xmath1 the denominator approaches zero which will cause the wave vector @xmath75 to diverge . this contradicts with the basic assumption of the existence of minimum length . hence it gives a upper limit for the measurable energy which is given by @xmath85 in other words this is a direct indication of the existence of maximum energy and hence a maximum momentum . note that the order of maximum energy matches well with that predicted in @xcite . we will now study the band structure . the simplest way to know the occurrence of a propagative or non - propagative mode is to find the nature of @xmath86 with respect to energy ( @xmath45 ) . a negative value of @xmath86 suggests a imaginary @xmath75 which is a damped mode . corresponding energy values form a forbidden band.in ( @xmath87[k2e ] ) we plot @xmath86 as a function of @xmath45 as given by equation ( [ k ] ) , which shows the band structure for the modified dirac equation . the forbidden band within the range @xmath0 is well known in dirac theory without gup . the interesting feature is the presence of two more forbidden bands which are the sole effect of the gup . the edges of the allowed and forbidden bands , which are actually zeros of ( [ k ] ) , are given by @xmath88 the edge @xmath89 is the bottom of the dirac sea , whereas @xmath90 gives the maximum positive energy for a free particle . based on this values we can divide the entire energy spectrum in five different regions which are described in ( @xmath91[band ] ) . .nomenclature and positions of different bands [ cols="^,^,^",options="header " , ] as a function of energy @xmath45 for ( @xmath92 , @xmath93 , @xmath94 ) with @xmath95 . the dashed line shows non gup dispersion relation @xmath96 . the light - red regions ( fb1 , fb2 , fb3 ) are forbidden bands and the light - blue regions ( ab1 , ab2 ) are the allowed bands . ] the dashed line in ( @xmath87[k2e ] ) corresponds the standard relativistic dispersion relation and the solid line gives the gup corrected mdr . we see that the solid line for low energy matches well with the relativistic dispersion relation . as @xmath45 is increased , @xmath86 gradually attains a maximum value and then start decreasing . the asymmetry of the curve is due to term @xmath97 . if we put @xmath98 then the curve is symmetric ) is @xmath99 , which is similar to that predicted in string theory @xcite . it supports the existence of a minimum length scale but does nt suggest a maximum energy . it agrees with the fact that for @xmath98 , @xmath100 ( [ emax ] ) . ] about the @xmath86 axis ( @xmath87[ek0 ] ) . the upper bound of the @xmath86 suggests the fact no matter how energetic particle we use we ca nt reduce its wave length below a certain limit and hence we ca nt measure below a certain length . with @xmath45 for [ @xmath101 ( blue line ) , [ @xmath102 ( red dotted line ) and [ @xmath103 . the blue line corresponds the gup ( [ gup ] ) we used , red line corresponds the gup with even powers of momentum @xcite and the black line shows the standard relativistic dispersion relation . ] with this discussion let us now proceed to the next section where we will discuss the klein paradox in this modified picture . in this section we will study the effect of gup on klein paradox . klein paradox is a well known phenomenon for a relativistic particle . the motion of a plane wave inside a step potential of height @xmath104 with energy ( e ) less than @xmath104 is well known in nonrelativistic quantum mechanics and we know that the wave will decay exponentially with an exponent @xmath105 . the expression suggests that if we increase @xmath104 the wave will die off faster and so long @xmath106 there will not be any propagative mode . in case of a dirac particle one can easily check that if the barrier height @xmath107 then there can be a propagative mode for @xmath106 and the wave can penetrate through the barrier . let us see how this effect is modified in presence of a gup . facing a potential step of height @xmath104 ( @xmath108 ) . the shaded regions are forbidden bands . a particle can not penetrate region ( ii ) if it encounter a forbidden band there . ] consider the situation when a particle with positive energy and up spin is incident on the potential step from the left . the solution for the incident wave in region ( @xmath109 ) is ( apart from the normalization factor ) @xmath110 where @xmath64 and @xmath75 are given by ( [ ug ] ) and ( [ k ] ) respectively . let us look for the solution in region ii ( @xmath87[step ] ) . the transmitted part will only contain the spin up component and the solution can be given by the ansatz @xmath111 where @xmath112 and @xmath113 can be obtained from ( [ mdr ] ) and ( [ ug ] ) replacing @xmath45 by @xmath114 @xmath115 \label{q}\\ \mathcal{u'}_g & = & \frac{c \hbar q_g ( 1+b\hbar q_g^2)}{ca\hbar^2 q_g^2 + mc^2 + ( e_g - v_0 ) } \label{u'g}\end{aligned}\ ] ] let consider a particle with such an energy that in region ( ii ) it falls in a forbidden energy band ( i.e. its momentum will be imaginary and the particle can not propagate ) . by gradually increasing the potential energy we can lift up the negative energy bands such a way that the particle will find itself in an allowed band . its momentum will be a real quantity and it can propagate inside the potential step . this is the physical scenario behind the klein paradox . in a non - gup dirac theory , if we keep increasing the barrier height arbitrarily , there will always be room for propagating modes in region ( ii ) . in presence of a gup the picture is entirely different . due to gup correction the allowed energy bands , i.e. @xmath116 and @xmath117 ( @xmath87[k2e ] ) are situated within a finite region . if a particle from region ( i ) encounter a forbidden band ( say fb2 ) in region ( ii ) ( @xmath87[kv]@xmath97,[kv]@xmath73 ) , then we can bring it to an allowed energy band ( ab1 ) by raising the potential ( @xmath87[kv]@xmath118 ) . if we further increase the potential we will find the particle again to be in a forbidden energy band ( fb1 ) ( @xmath87[kv]@xmath119 ) which does not occur in dirac theory without gup . the particle would not be able to penetrate the barrier on further increment of potential . ) a free particle ( black dot ) in an allowed band ( ab2 ) ( @xmath120 ) . ( @xmath73 ) in a potential step the energy spectra will shift by an amount equal to the step height and the particle will find itself in a forbidden band ( fb2)(@xmath121 ) . ( @xmath118 ) by raising the potential energy , the spectrum can be shifted further to the left which can bring the particle again in an allowed band ( ab1 ) ( @xmath122 ) . ( @xmath119 ) if the potential is further increased then at a point the particle will again fall in a forbidden band ( fb1 ) ( @xmath123 ) . ] in this section we study the behaviour of a massless particle inside a potential step . the solution for the modified dirac equation will be similar to the ansatz ( [ trpsi ] ) . we can write it as @xmath124 where @xmath125 can be find by putting @xmath126 in ( [ q ] ) and is given by @xmath127 \label{q0}\end{aligned}\ ] ] expression for the wave vector in free space ( @xmath128 ) can easily be found from ( [ q0 ] ) by putting @xmath129 . as a function of energy for a massless dirac particle for @xmath130 and @xmath93 . there is no forbidden energy band around the zero energy , but the forbidden energy bands on the both ends still exist ( fb1 and fb3 ) . the dashed line shows the dispersion relation in absence of gup . ] in a non gup dirac theory , a massless particle can penetrate a potential step of any height due to the absence of any forbidden band . in presence of a gup we see that ( @xmath87[vq0 ] ) there exist another couple of forbidden bands ( fb1 and fb3 ) on the both sides of an allowed band ( ab(1 + 2 ) ) . by applying a sufficiently large potential ( @xmath2 ) we can bring this forbidden region to any energy level which will cause a massless dirac particle with that energy to bounce from the potential wall . this phenomenon is not like standard klein paradox , rather the reverse of it where due to an increase of the potential a particle which could propagate earlier is now reflected from the potential wall . this is the consequence of the existence of a minimum length and a maximum energy which is manifested through the modified momentum operator ( [ p ] ) . in this paper we discussed the modification of the dispersion relation of a dirac particle and how it affects the band structure of a dirac particle in presence of a minimum length and maximum energy . then we studied the klein paradox with the help of this new band structure . if a dirac particle with some fixed energy face a forbidden energy band in a potential region , it will rebound from the wall of the potential . by increasing the potential strength we can lift up the negative energy levels such a way that there will be an allowed energy band corresponding to the particle energy and the particle will propagate within the potential region ( @xmath131[step ] ) . this is the essence of klein paradox . in presence of a gup the the scenario is different . we see that there are additional forbidden bands apart from the @xmath0 region , denoted by a negative @xmath86 in ( @xmath87[k2e ] ) which appear at the both ends of the energy spectra . the allowed energy bands are not extended over @xmath132 , rather localized within a finite region . this kind of band structure is the reason behind the difference in klein paradox at high energy from that in case of a non gup theory where the allowed bands are not bounded . we can initially bring a particle to an allowed band from a forbidden band by raising the potential ( like what happen in a non gup klein paradox ) ( @xmath87[k2e]@xmath133 ) ; but if we keep the potential increasing we will reach the end of the allowed band and the particle will find itself again in a forbidden band ( @xmath87[k2e]@xmath119 ) . as a result it will reflect from the wall of the potential . we further studied the phenomenon for a massless dirac particle . in case of a massless dirac particle there is nothing like a forbidden energy band in relativistic quantum mechanics and hence a massless particle with any energy can penetrate and propagate within a potential region of arbitrary height . in presence of a gup the situation is different . for zero mass the forbidden region in the middle ( fb2 in @xmath87[k2e ] ) disappears but those at the both ends ( fb1,fb3 ) still survive ( @xmath87[vq0 ] ) . so in presence of a sufficiently large potential ( @xmath2 ) it is always possible for a massless particle with any energy to encounter a forbidden band . so with increase of height of the potential , a massless dirac particle which earlier can penetrate inside the potential region will bounce off the wall like a massive particle does in nonrelativistic quantum mechanics . this is solely due the boundedness of the allowed energy bands which is a consequence of the existence of a minimum length and maximum energy ( momentum ) . hence the reflection of a massless dirac particle from a potential barrier of height of the order of planck energy is direct a signature of existence of a minimum length and maximum energy . the author like to thank council of scientific and industrial research ( csir ) , india for financial support . d. amati , m. ciafaloni and g. veneziano , phys . b 216 , 1989 , ( 41 - 47 ) . m. maggiore , phys . b 319 ( 1993 ) 83 [ arxiv : hep - th/9309034 ] . a. kempf , g. mangano , r. b. mann , phys . d 52 ( 1108 - 1118 ) 1995 [ arxiv : hep - th/9412167 ] . f. girelli , e. r. livine and d. oriti , nuclear physics b 708 , 2005 , ( 411 - 433 ) [ arxiv : grqc/0406100 ] . s. hossenfelder , phys . d 73 ( 105013 ) , 2006 [ arxiv : hep - th/0603032 ] . s. hossenfelder , class . quant . grav . 23 , 2006 , ( 1815 - 1821 ) [ arxiv : hep - th/0510245 ] . s. hossenfelder , class . ( 038003 ) , 2008 [ arxiv:0712.2811 ] . l. n. chang , z. lewis , d. minic and t. takeuchi , adv . high energy phys . 2011 , ( 493514 ) , 2011 , [ arxiv : hep - th/1106.0068 ] . s. hossenfelder , arxiv : gr - qc/1203.6191v1 . b. panes , eur . j. c ( 2012 ) 72:1930 [ arxiv : hep - ph/1112.3753v3 ] . saurya das and elias c. vagenas , phys . lett . 101 , 221301 ( 2008 ) [ arxiv : hep - th/0810.5333v2 ] . ahmed farag ali , saurya das , elias c. vagenas , phys . b 678 ( 497 - 499 ) 2009 [ arxiv : hep - th/0906.5396 ] . saurya das , elias c. vagenas , ahmed farag ali , phys . b 690 ( 407 - 412 ) 2010 [ arxiv : hep - th/1005.3368 ] . ahmed farag ali , saurya das , elias c. vagenas , phys . rev . d 84 ( 044013 ) 2011 [ arxiv : hep - th/1107.3164 ] . saurya das , r.b . mann , phys . b 704 , ( 596 - 599 ) , 2011 [ arxiv : hep - th/1109.3258 ] . p. pedram , k. nozari and s. h. taheri , jhep 1103 , 093 ( 2011 ) . [ arxiv : hep - th/1103.1015 ] . p. pedram , phys . b 702 ( 295 - 298 ) 2011 . p. pedram , phys . rev . d 85 , 024016 ( 2012 ) [ arxiv : hep - th/1112.2327 ] . zachary lewis and tatsu takeuchi , phys . d 84 , 105029 ( 2011 ) [ arxiv : hep - th/1109.2680 ] .

we study the klein paradox in presence of a new generalized uncertainty principle . we define a new momentum operator and derive the modified dispersion relation for a dirac particle . apart from the forbidden band within the range @xmath0 , m being the mass of the particle , we find the existence of additional forbidden bands at the both ends of the energy spectrum . it causes the allowed energy bands to be localised within finite energy range . this band structure forbids a dirac particle from penetrating a potential barrier of sufficient height which is estimated to be of the order of planck energy ( @xmath1 ) . this is also true for a massless particle . it reflects from the wall of a sufficiently high ( @xmath2 ) potential just like a massive particle does in nonrelativistic quantum mechanics . we propose this to be a signature of the existence of a minimum measurable length and a maximum measurable energy .

*introduction* *gup and the dirac equation* *solution of modified dirac equation for a free particle and its band structure* *gup and klein paradox* *gup, massless particle and (reverse) klein paradox* *discussion* acknowledgement

the definition of quantum fields on a curved background is an hard task , because of the impossibility to define a unique ( up to unitary equivalence ) hilbert space structure @xcite . nevertheless , equations of motion for such fields can be extended to the curved case and semi - classical tools can be applied at least formally . this way , the interaction of the gravitation field with quantum degrees of freedom is inferred , in a regime where quantum gravity effects are not yet relevant and a particle - like description is recovered . this coupling may be tested experimentally by studying the trajectories of elementary particles . here we address the point whether the interaction between quantum spin and the geometry reflects that of classical spin tensors . in fact , from the semi - classical limit of spinor fields , we expect a non - trivial coupling between their internal degrees of freedom and the curvature , so that a non - geodesics dynamics is expected . these deviations with respect to the structure - less case are going to be compared with those ones coming from the papapetrou formulation for spinning particles @xcite . the work is organized as follows : at first the papapetrou formulation for classical spinning particles is presented , and then the semi - classical limit of the dirac equation is performed . papapetrou @xcite performed the classical analysis on the motion of a body in general relativity on a fixed background . in his formulation , the body is described in the space - time as a thin tube inside which there is a non - vanishing energy momentum tensor . as soon as a curve @xmath0 is identified as the center - of - mass world - line , the multi - pole expansion consists in neglecting all moments of the energy - momentum distribution with respect to @xmath1 from a certain one . at the lowest order a geodesics dynamics is found and this is consistent with basic principles of general relativity . at the next order , the pole - dipole approximation , a characterization of the body internal structure is given by the spin tensor @xmath2 , whose expression is the following one @xmath3 @xmath4 being the deviation with respect to @xmath1 . the system of equations describing the dynamics is @xmath5 and it outlines that for rotating bodies a generalized momentum @xmath6 arises and , in general , it is not aligned with the 4-velocity ; the spin tensor is subjected to a precession around the generalized momentum ; the trajectory deviates with respect to a geodesics motion , as a consequence of the interaction between the spin tensor and the gravitational field . however , in the system ( [ pap ] ) there are thirteen unknown quantities and ten equations only . hence , to provide a solution three conditions must be added . in literature one finds the following three possible choices , @xmath7 the corinaldesi - papapetrou @xcite , the pirani @xcite and the tulczyjew @xcite consistency conditions , respectively . all of them expresses the spatial nature of the spin . the investigation on the coupling between quantum degrees of freedom and gravity requires the extension of the quantum fields formulation to a curved space - time . as far as spin-@xmath8 particles are concerned , the dirac algebra must be extended to non - minkowskian spaces . this can be done by introducing vier - bein vectors @xmath9 and defining dirac matrix as @xmath10 , @xmath11 being those ones of the flat case . furthermore , generalized derivatives must be defined @xcite , such that the dirac equation in curved space - time reads as @xmath12 @xmath13 $ ] being generators of the lorentz group , while @xmath14 . the notion of a classical trajectory is recovered for @xmath18 by a localization along the world - line @xmath19 , which is the integral curve of @xmath20 . the localization is obtained by assuming a gaussian dependence of the following kind even though the spread @xmath22 grows during the notion , nevertheless it can be shown that for a reasonable set of initial conditions there exists a macroscopic region where a classical trajectory can be defined @xcite . in order to make this formulation closer to the papapetrou one , we are going to perform a spacial integration , taking as space - like coordinates @xmath24 , _ i.e. _ those ones of the space orthogonal to @xmath25 . the localization here works by allowing us to neglect any term containing first and second derivatives of the gaussian . furthermore , in agreement with the assumption that any dependence on spatial coordinates has been included into the gaussian , we take for the spinor @xmath26 @xmath27 we stress that a papapetrou - like interaction between the spin tensor and the riemann tensor comes out . furthermore , being the generalized momentum aligned with the 4-velocity , the conservation of the spin tensor is consistent with a papapetrou - like dynamics .

the notion of a classical particle is inferred from dirac quantum fields on a curved space - time , by an eikonal approximation and a localization hypothesis for amplitudes . this procedure allows to define a semi - classical version of the spin - tensor from internal quantum degrees of freedom , which has a papapetrou - like coupling with the curvature .

introduction papapetrou formulation semi-classical limit of the dirac equation on a curved space-time

as one well - known double - peaked emitter ( agn with double - peaked broad low - ionization emission lines ) , 3c390.3 has been studied for more than two decades , since clear double - peaked broad balmer lines were reported by burbidge & burbidge ( 1971 , spectrogram of 3c390.3 ) and by perez et al . ( 1988 , clear spectral lines ) . based on double - peaked appearance of the broad lines and characters of long - term variabilities of the broad lines and continuum emission , among those proposed theoretical models for double - peaked emitters ( such as binary black hole model : begelman et al . 1980 , boroson & lauer 2009 , gaskell 1996 , lauer & boroson 2009 , zhang et al . 2007 , double stream model : vellieux & zheng 1991 , zheng et al . 1990 , 1991 , accretion disk model : bachev 1999 , chen et al . 1989 , chen & halpern 1989 , chornock et al . 2010 , eracleous et al . 1995 , flohic & eracleous 2008 , gaskell 2010 , gezari et al . 2007 , hartnoll & blackman 2000 , 2002 , karas et al . 2001 , lewis et al . 2010 , oke 1987 , perez et al . 1988 , storchi - bergmann , et al . , 2003 , schimoia et al . 2012 , tran 2010 ) , the accretion disk model ( double - peaked broad lines coming from central accretion disk ) has been widely accepted for the double - peaked emitter 3c390.3 ( chen et al . 1989 , chen & halpern 1989 , dietrich et al . 2012 , eracleous et al . 1997 , livio & xu 1997 , oke 1987 , perez et al . 1988 , popovic et al . 2011 , shapovalova et al . 2001 , 2010 , zhang 2011a ) . then based on the proposed theoretical accretion disk model , besides the different line profiles , we try to find some interesting distinctions between the double - peaked emitter 3c390.3 and normal broad line agn , which is the main objective of our manuscript . as we commonly know that motions of broad emission line clouds in blr ( broad emission line regions ) of agn are gravitationally dominated by central mass of agn ( gaskell 1988 , 1996 , krause et al . 2011 , kollatschny & zetzl1 2011 , netzer & marziani 2010 , peterson & wandel 1999 , wandel et al . 1999 , and references therein ) , which provides strong evidence for applying virialization assumption to estimate virial black hole masses of broad line agn ( bennert et al . 2011 , collin et al . 2006 , greene & ho 2004 , 2005 , kelly & bechtold 2007 , marziani et al . 2003 , netzer & marziani 2010 , onken et al . 2004 , park et al . 2012 , peterson et al . 2004 , peterson 2010 , rafiee & hall 2011 , shen & liu 2012 , sulentic et al . , 2000 , vestergaard 2002 , wu et al . 2004 ) . then based on the virialization assumption , combining with widely accepted empirical relation to estimate size of blr of agn ( @xmath2 , bentz et al . 2006 , 2009 , denney et al . 2010 , kaspi et al . 1996 , 2000 , 2005 , greene et al . 2010 , wang & zhang 2003 ) , we will have @xmath4 where index @xmath5 means corresponding line parameters of broad h@xmath0 . moreover , we can find that for individual object ( @xmath6 and @xmath7 ) , @xmath8 in other words , line width of broad emission lines is decreasing with line luminosity ( size of blr , continuum luminosity ) increasing . the strong expected negative correlation , @xmath9 versus @xmath10 , have been found and reported for several well mapped objects , such as ngc5548 ( bentz et al . 2007 , denney et al . 2010 , peterson et al . 2002 , 2004 , wanders & peterson 1996 ) , ngc3783 and ngc7469 ( peterson et al . 2004 ) , arp151 ( park et al . 2012 ) . certainly , the correlation between @xmath9 and @xmath10 for 3c390.3 is also shown in peterson et al . ( 2004 ) , however , there is no one reliable correlation confirmed for 3c390.3 , due to the less number of data points ( only 4 data points for 3c390.3 shown in figure 8 in peterson et al . 2004 ) . in the paper , we will carefully re - check the correlation for 3c390.3 , because among the well - mapped objects , 3c390.3 is the unique double - peaked emitter . in the paper , we check correlation between line width and line flux , rather than to check correlation between size of blr and line width , in order to obtain abundant data points . this paper is organized as follows . section 2 gives our data procedures and main results . section 3 shows our discussions and conclusions . here , cosmological parameters @xmath11 , @xmath12 and @xmath13 have been adopted . all public spectra of 3c390.3 can be collected from the agnwatch project ( http://www.astronomy.ohio-state.edu/~agnwatch/ ) , which is a consortium of astronomers who have studied the inner structures of agn through continuum and emission - line variability . detailed descriptions about observational instruments , techniques and procedures to reduce the spectra of 3c390.3 can be found in dietrich et al . ( 1998 ) , which will not be described any more . here , we mainly show our procedures to measure line parameters ( line width and line flux ) of the double - peaked broad lines of 3c390.3 , and show our results about correlation of the line parameters . we mainly focus on line parameters of broad h@xmath0 of 3c390.3 , due to the following two main reasons . on the one hand , the broad h@xmath0 is much stronger than the broad h@xmath14 , which leads to more reliable line parameters . on the other hand , there is much less contamination on the broad h@xmath0 . only the narrow emission lines , the narrow h@xmath0 and [ n ii ] doublet , have effects on the measured line parameters of the double - peaked broad h@xmath0 . however , for the double - peaked broad h@xmath14 , besides the common narrow h@xmath14 and [ o iii ] doublet , the extended [ o iii ] doublet and some weak optical fe ii emission lines ( dietrich et al . 2012 ) have effects on the measured line parameters of the broad h@xmath14 . thus , focusing on the observed spectrum around h@xmath0 of 3c390.3 should lead to more accurate line parameters and to find more reliable correlation of the line parameters . for 3c390.3 , there are 67 public spectra observed around 1995 including available h@xmath0 which can be collected from the agnwatch project , based on those reported parameters listed in table 9 in dietrich et al . then the line parameters of the double - peaked broad h@xmath0 are measured as follows . in order to ignore effects of the narrow lines , the double - peaked broad component of h@xmath0 is firstly described and fitted by the theoretical elliptical accretion disk model proposed by eracleous et al . ( 1995 ) . here , our main objective is not to accurately determine the disk parameters through theoretical model , thus , there are no further discussions about disk parameters , which has been discussed in detail in our another paper ( zhang 2011a , paper i ) . in other words , we only require that the line profiles of the double - peaked broad h@xmath0 can be best described and be well separated from the narrow lines . in this paper , we show the best fitted results for the apparent double - peaked broad h@xmath0 in figure [ 3c390 ] , and the corresponding residuals in figure [ res ] . in figure [ res ] , we also show the best fitted results for the narrow lines of h@xmath0 and [ n ii]@xmath15 by three standard gaussian functions with the same redshift and with line parameters ratio of [ n ii ] doublet fixed to theoretical values . based on the results shown in figure [ 3c390 ] and figure [ res ] , the double - peaked broad component of h@xmath0 can be clearly separated from the narrow lines . then , the line width ( second moment ) and the line flux of the double - peaked broad component can be estimated by ( peterson et al . 2004 ) @xmath16 where @xmath17 is the line profile of the double - peaked broad h@xmath0 after subtraction of the narrow lines . here , the main reasons to use the second moment rather than fwhm ( full width at half maximum ) are : the second moment rather than the fwhm can be used to well trace kepler velocities of broad line clouds in blr of agn ( fromerth & melia 2000 , peterson et al . 2004 ) , the second moment rather than the fwhm is well defined for arbitrary line profiles ( such as the double - peaked broad line profile ) , and moreover the second moment rather than the fwhm has relatively lower uncertainty ( peterson et al . the measured line widths ( @xmath18 ) of the double - peaked broad h@xmath0 from the observed spectra are listed in table 1 . before proceeding further , it is necessary to estimate the corresponding uncertainty of the measured line width of the double - peaked broad h@xmath0 . here , the commonly used bootstrap method ( peterson et al . 1998 , 2004 , press et al . 1992 , zhang 2011b ) is applied to estimate the uncertainty . based on each observed double - peaked broad h@xmath0 with narrow lines ( [ n ii ] , [ s ii],[o i ] doublets and narrow h@xmath0 ) having been subtracted , 400 mock double - peaked lines can be created by the bootstrap method . when the bootstrap method is applied , we should avoid the selection of some adjacent data points to create mock lines , otherwise , the following estimated uncertainty should be unreasonably small , due to few variation of line profile . then line widths of the 400 mock lines are measured by equation ( 3 ) . after that , standard deviation of the 400 measured line widths of the mock lines is accepted as the uncertainty of the line width of the observed double - peaked broad h@xmath0 , and listed in table 1 . certainly , we should note that peterson et al . ( 2004 ) accepted @xmath19 as the mean uncertainty of the line width of broad h@xmath14 of 3c390.3 within observed wavelength range from 5006to 5246 . it is consistent with our value @xmath20 , which indicates our estimated uncertainties for the line width of all the 67 double - peaked broad h@xmath0 are reasonable and acceptable . besides the line width and the corresponding uncertainty , it is not difficult to accomplish spectral flux calibration for 3c390.3 however , because the spectra of 3c390.3 are observed by different instruments in different observatories under different configurations , the flux calibration based on the direct observed line profile should have low confidence level . as discussed in detail in the section 2 of dietrich et al . ( 1998 ) , some further procedures have been applied to complete the flux calibration . the spectra are firstly photometrically calibrated by comparison with the broadband photometric measurements ( being convolved with the spectral response curve of the johnson v and r ) , and then scaled by multiplicative factors to achieve the same total flux ratio as the intercalibrated photometric measurements in v and r bands . after that , the spectra have to be intercalibrated to a common flux level , in order to correct the effects for different wavelength shifts , and different spectral resolutions in different spectra . all the procedures have been successfully done by dietrich et al . ( 1998 ) , so that we directly collect the line fluxes ( @xmath21 ) and corresponding uncertainties of broad h@xmath0 listed in dietrich et al . moreover , in spite of the effects from different flux scales for the observed spectra , those directly measured line fluxes of the double - peaked broad h@xmath0 ( @xmath22 , no corrections being considered ) are also listed in table 1 . before the end of the subsection , there are two points we should note . on the one hand , we do not try to confirm that the pure elliptical accretion disk model ( eracleous et al . 1995 ) is the unique true physical model for the double - peaked emitter 3c390.3 . we only use the theoretical disk model to separate the double - peaked broad component from the narrow lines . the properties of the residuals shown in figure [ res ] for the best fitted results shown in figure [ 3c390 ] illustrate that it is one good choice to determine the double - peaked broad component by the theoretical elliptical disk model for the spectra around 1995 . on the other hand , although we do not apply the intercalibration method ( van groningen & wanders 1992 ) to correct the effects from small wavelength shifts , different spectral resolutions and different flux scales for different spectra as what have been done in dietrich et al . ( 1998 ) , the measured second moment through observed spectrum is accurate and reliable , because the applied broadening velocity in the intercalibration method for 3c390.3 is only ten or more kilometers per second or smaller , which is can be totally ignored for the measured second moment and the corresponding uncertainty of the double - peaked broad h@xmath0 . thus , the following results based on the measured second moments are reliable . figure [ fj_3c390 ] shows the correlation between the line width and the line flux of the double - peaked broad h@xmath0 of 3c390.3 . linear correlation coefficient for the correlation of @xmath22 versus @xmath18 is about 0.28 with two - sided significance of deviation from zero about @xmath23 for all the 67 spectra listed in table 1 . the coefficient for the correlation of @xmath21 ( accurate line flux of h@xmath0 from the agnwatch project ) versus @xmath18 is about 0.45 with @xmath24 . it is clear that there is one positive correction between the line width and the line flux of the double - peaked broad h@xmath0 for 3c390.3 . furthermore , under considerations of the uncertainties of the parameters in both coordinates , the best fitted results for @xmath21 versus @xmath18 can be written as @xmath25 certainly , we should note the accurate flux @xmath21 collected from the agnwatch project includes the contributions from the narrow emission lines of @xmath26 and [ n ii ] doublet . however , the contributions do not affect the positive correlation , because the flux of the narrow emission lines of each spectrum is constant . the best fitted results are shown as solid lines in figure [ fj_3c390 ] . the corresponding 99.95% confidence bands for the best fitted results are also shown in figure [ fj_3c390 ] . it is clear that the uncertainty of the measured second moment ( @xmath27 ) of the double - peaked broad h@xmath0 has important effects on our final conclusion about the positive correlation for 3c390.3 , because the varying range of second moment is not large , @xmath28 . if the uncertainty of the second moment is large enough to cover the varying range of the second moment , the shown positive correlation in figure [ fj_3c390 ] should be fake . thus , it is necessary to check whether larger uncertainties should be possible for our measured line width . here , we simply consider the question as follows . for one given theoretical line profile of the double - peaked broad h@xmath0 with second moment @xmath18 ( solid line shown in figure [ 3c390 ] ) best fitted for the observed line profile , if the corresponding uncertainty @xmath27 for the second moment is reasonable , we will find that after being broadened by velocity @xmath29 for the theoretical line profile , the new theoretical line profile with second moment @xmath30 still well describes the observed double - peaked broad line profile of h@xmath0 . here , the value of @xmath29 can be determined by @xmath31 . based on the listed values in table 1 , we can find that the common broadening velocity is around 20 ( @xmath32 ) . figure [ res2 ] shows one example about the being broadened line profile . in the figure , we can find that the being broadened line profile ( the idl function convol being applied ) with @xmath33 has begun to lead to some bad fitted results for the spectrum around the blue peak ( @xmath34 ) of the double - peaked broad h@xmath0 , which indicates larger broadening velocity should be unreasonable . in other words , the varying range of the second moment is larger enough to ignore the effects of the uncertainty of the second moment on the positive correlation shown in figure [ fj_3c390 ] , and the positive correlation is reliable . besides the unexpected positive correlation shown in figure [ fj_3c390 ] for 3c390.3 , we wish to check the correlation of line parameters of broad lines for the other well - known mapped normal broad line agn , in order to frankly show that the positive correlation for the double - peaked emitter 3c390.3 is very interesting . there are more than 40 nearby mapped objects reported in the literature ( barth et al . 2011 , bentz et al . 2006 , 2009 , 2010 , denney et al . 2006 , 2009 , 2010 , kaspi et al . 1996 , 2000 , 2005 , 2007 , peterson et al . the simple correlation between the blr size and the line width for part of those objects can be found in the corresponding literatures . here , ngc5548 is selected as our target , because there is one well sample of fine spectra with homogeneous high - quality for ngc5548 ( bentz et al . 2007 , dietrich et al . 1993 , korista et al . 1995 , peterson et al . 1991 , 1992 , 1994 , 1999 , 2002 , 2004 , popovic et al . 2008 , sergeev et al . 2007 , shapovalova et al . 2004 , wanders & peterson 1996 ) . some detailed information about the spectra of ngc5548 can be found in the website of the agnwatch project . here , we mainly focus on the spectra studied in detail in wanders & peterson ( 1996 ) and collect the spectra with necessary corrections having been done by wanders & peterson , i.e. , the collected 247 spectra have been scaled through the van groningen - wanders method , and have the underlying continuum subtracted , and have the narrow components removed as well , and no further corrections are necessary for the following measured line parameters . then based on the collected 247 spectra of ngc5548 observed from dec . 1988 to sep . , the line width and the line flux can be easily determined by equation ( 3 ) , within observed wavelength range from 4840 to 5030 . then , uncertainty of the line flux is determined by uncertainty in spectral flux provided by the agnwatch project , and uncertainty of the line width is determined by the bootstrap method . here , we do not show all the observed spectra of ngc5548 , but only show one spectrum as an example in the top - right panel of figure [ fj_ngc5548 ] . the measured line parameters and corresponding uncertainties of the broad h@xmath14 of ngc5548 are listed in table 2 . figure [ fj_ngc5548 ] shows the correlation between the line width and the line flux of the broad h@xmath14 for ngc5548 . the linear correlation coefficient is about -0.83 with @xmath35 . the best fitted results with the considerations of the uncertainties of the parameters in both coordinates can be written as @xmath36 it is clear there is one strong negative correlation for ngc5548 , which is well consistent with the expected result under the virialization assumption : @xmath37 . undoubtfully , the negative correlation is well consistent with the negative correlation between the size of blr and the line width found in bentz et al . ( 2006 ) , denney et al . ( 2010 ) and peterson et al . ( 2004 ) . based on the results above , we can find that there is one reliable positive correlation between the line width and the line flux of the double - peaked broad h@xmath0 for the double - peaked emitter 3c390.3 , which is against the expected negative correlation confirmed for the other mapped agn . the unexpected positive correlation should indicate different structures of blrs of the double - peaked emitter 3c390.3 and normal broad line agn . it is clear that the reliable positive correlation shown in figure [ fj_3c390 ] for the double - peaked emitter 3c390.3 is against the expected result through the virialization assumption . thus the following three questions are mainly considered , what determines the unexpected positive correlation ? whether the virialization assumption is still available for double - peaked emitters ? whether the unexpected positive correlation can be found for the other double - peaked emitters ? for normal broad line agn , such as ngc5548 , the strong negative correlation of the broad line parameters can be naturally explained by the following simple viewpoint : blr kinematics are dominated by central gravity ( some more recent reviews can be found in down et al . 2010 , sluse et al . 2011 ) , which could be well appropriate to the broad line seyfert 1 galaxy ngc5548 ( bentz et al . 2007 , denney et al . 2010 , peterson et al . 2002 , 2004 , wanders & peterson 1996 , ) and some other mapped broad line agn ( peterson et al . 2004 ) : stronger continuum emission leads to deeper ionization boundary ( larger flux - weighted size of blr ) and leads to smaller general rotating velocity ( smaller line width ) . and then , through the broad line strength being applied to estimate the size of blr , @xmath38 , the negative correlation of the line width and the line flux can be found . certainly , the viewpoint is not appropriate to the double - peaked emitter 3c390.3 . if the theoretical accretion disk model was accepted , what result about the correlation between the size of blr and the continuum / line flux should be expected for 3c390.3 ? in the context of the theoretical disk model for double - peaked emitters , such as the elliptical accretion disk model ( eracleous et al . 1995 ) which can be well applied to best fit the observed line spectra of 3c390.3 observed around 1995 , the disk - like blr is locating into the central accretion disk , and does share the kinematic characters of the central accretion disk . for 3c390.3 , the inner and outer boundary radii of the elliptical disk - like blr are about @xmath39 and @xmath40 , under the theoretical accretion disk model ( eracleous & halpern 1994 , 2003 , flohic & eracleous 2008 , shapovalova et al . 2001 , zhang 2011a , paper i ) . based on the oversimplified extended disk - like blr ( no considerations of tiny contributions from probable hot spots , warped structures , spiral arms etc . for 3c390.3 around 1995 ) , one simple result can be expected : stronger continuum emission leads to stronger broad line emission coming from the inner part of the disk - like blr ( and then larger line width ) in one short period ( much shorter than the traveling time for ionizing photons through the whole blr ) , because the stronger ionizing continuum firstly reaches the inner part of the disk - like blr . in other words , stronger continuum emission ( strong line flux ) leads to larger flux ratio of the broad h@xmath0 from the inner part of the blr to the broad h@xmath0 from the outer part of the blr , and leads to larger line width due to larger contributions from the higher velocity clouds in inner part of blr , which is similar as the results shown in figure [ fj_3c390 ] . so that , we check the properties of the broad line flux ratio @xmath41 , where @xmath42 means the broad line flux coming from the inner part of disk - like blr with radius from @xmath43 to @xmath44 and @xmath45 means the measured total broad line flux . the calculated value of r for each spectrum is listed in table 1 . figure [ fr ] shows the correlation between the line flux @xmath22 ( @xmath21 ) and the parameter of @xmath46 . in the figure , the shown uncertainty of @xmath46 is determined by the uncertainty of @xmath21 collected from the agnwatch project . the spearman - rank correlation coefficient for the correlation is about 0.49 with @xmath47 for the parameter @xmath21 in table 1 and 0.26 with @xmath48 for the parameter @xmath22 in table 1 . the positive correlation in figure [ fr ] indicates the proposed disk origination for the double - peaked broad h@xmath0 of 3c390.3 can be naturally applied to explain the unexpected positive correlation shown in figure [ fj_3c390 ] . moreover , if the observed broad h@xmath0 of 3c390.3 was seperated into two parts : one part with larger line width ( @xmath49 ) coming from the inner region of the disk - like blr with radius from @xmath43 to @xmath50 , and the other part with smaller line width ( @xmath51 ) coming from the outer region of the disk - like blr , then the line width ( @xmath18 ) of the total observed double - peaked broad h@xmath0 can be re - written as ( zhang 2011b ) @xmath52 where @xmath46 is the flux ratio of @xmath41 . the equation above can be easily proved by definition of second moment ( equation 3 ) . based on the mean disk parameters shown in our paper i ( zhang 2011a ) , we have @xmath53 and @xmath54 . through the equation above , in order to obtain the measured line width of the double - peaked broad h@xmath0 varying from @xmath55 to @xmath56(minimum and maximum values for the second moment of broad h@xmath0 of 3c390.3 ) , the parameter of @xmath46 should be varying from @xmath57 to @xmath58 , which is consistent with the shown results about r in figure [ fr ] and in table 1 . the results above indicate that the unexpected positive correlation shown in figure [ fj_3c390 ] for 3c390.3 can be naturally explained by the proposed accretion disk origination of the double - peaked broad h@xmath0 . certainly , here we do not consider effects from probable external ionizing source for the double - peaked emitter 3c390.3 , which will be discussed in detail in subsection 3.4 . as the most convenient method to estimate central black hole masses of broad line agn , the virialization assumption has been widely and commonly accepted . here , we should find that the assumption is also valid for the double - peaked emitter 3c390.3 . as we have shown above , the proposed accretion disk model should be appropriate for the double - peaked emitter 3c390.3 . considering the oversimplified case , the disk - like blr of 3c390.3 is separated into one inner region with radius from @xmath43 to @xmath50 and the other outer region , then the flux weighted distance between the two regions to the central black hole can be calculated based on the simple elliptical accretion disk model ( eracleous et al . 1995 ) @xmath59 for the inner part of the blr and @xmath60 for the outer part of the blr . according to the results shown in zhang ( 2011b ) , the size of total blr ( @xmath9 ) can be written as @xmath61 based on the values of r listed in table 1 and the values of @xmath62 and @xmath63 , we will have the size of blr @xmath9 varying from @xmath64 to @xmath65 . moreover , based on the virialization assumption , @xmath66 ( or @xmath67 ) , combining with the measured line width in table 1 , the expected maximum variation of the size of blr of 3c390.3 calculated through the variations of the line width should be @xmath68 , which is consistent with the results above @xmath69 . the results above indicate that the virialization assumption is still available for the double - peaked emitter 3c390.3 . however , the common empirical relation @xmath2 is not available to show the properties of * time - varying * size of blrs of 3c390.3 . in other words , based on the * mean * continuum luminosity ( broad line luminosity ) in one long period , the * mean * size of blr can be estimated by the empirical relation @xmath2 , however , the time - varying size of blr is decreasing with the line luminosity increasing . besides the well - known double - peaked emitter 3c390.3 , there are several other double - peaked emitters , of which the double - peaked broad line profiles have been studied and reported . some simple results about correlation between peak separation ( in some literature , peak separation was used as substitute for line width ) and line flux ( continuum emission ) can be found . in figure 4 of shapovalova 2001 ) , we can find one rough negative correlation between peak separation and line flux ( continuum emission ) for the double - peaked broad h@xmath14 of 3c390.3 observed from 1995 to 2000 . similarly negative correlation can be found for the double - peaked broad h@xmath0 of ngc1097 ( schimoia et al . 2012 , storchi - bergmann et al . however , through study of one sample of double - peaked emitters , lewis et al . ( 2010 ) have shown that there is no confirmed correlation between the peak separation and the line flux for long - term observed double - peaked broad h@xmath0 , and additionally , the expected negative correlation between the line width ( either fwhm or fwqm ) and the broad h@xmath0 flux is not observed . the contradictory is mainly due to the following reason : there are important effects on peak separation from varying of physical disk parameters , such as the effects from accretion disk precession , in the context of the accretion disk model for double - peaked emitters . as we have seen that the proposed accretion disk models for double - peaked emitters lead to one important result : disk precession ( or physical disk structure parameters varying ) dominates the line profile variations of double - peaked broad line , besides the not so important contributions from the other subtle structures , such as the hot spots , warped structures etc .. in some cases , such as disk - like blr with large eccentricity @xmath70 under the elliptical accretion disk model , even there were no variations of continuum emission , only the orientation angle of elliptical rings @xmath71 varying ( pure accretion disk precession ) should lead line width ( second moment ) varying by 30% , which can be simply verified by the proposed theoretical models . thus , when we consider the correlation of the line parameters of double - peaked broad lines , the effects from the accretion disk precession ( or from the other physical disk parameters varying ) should be firstly corrected . however , before confirming one clear theoretical model for one double - peaked emitter , it is hard to correct these effects . one good choice to ignore the effects of accretion disk parameters varying is to consider short - term observed spectra with few variations of line profiles . that is the main reason why only the spectra around 1995 are collected for the double - peaked emitter 3c390.3 , because the line profiles around 1995 are almost stable with less effects from disk parameters varying , as what we have discussed in our paper i ( zhang 2001a ) . however , in lewis et al . ( 2010 ) , schimoia et al . ( 2012 ) , shapovalova et al . ( 2001 ) and storchi - bergmann et al . ( 2003 ) , the observed double - peaked broad lines have apparent variations of both peak flux ratios and peak positions , which indicate apparent variations of accretion disk physical parameters . if only continuum variations are considered , there should be much weak variations for peak positions and peak flux ratios , besides apparent variations of line intensity . moreover , we further check the listed parameters in the literature for the double - peaked emitter 3c390.3 , under the virialization assumption : @xmath72 . in wandel et al . ( 1999 ) and peterson et al . ( 2004 ) , the size of blr and the corresponding line width are @xmath73 and @xmath74 for 3c390.3 around 1995 . in shapovalova et al . ( 2001 ) , the size of blr and the corresponding line width are @xmath75 ( similar results can be found in popovic et al . 2011 , some larger results can be found in sergeev et al . 2011 ) and @xmath76 for 3c390.3 from 1995 to 2000 . it is clear that based on the measured line width ( fwhm ) and the size of blr , we have @xmath77 , which indicates the measured line width ( fwhm ) varying does not obey the virialization assumption . in other words , in one long - term observed period , the effects of accretion disk physical parameter varying should be important , especially on the line profile varying , so that the measured line width ( fwhm ) can not accurately trace the kepler velocity of broad line clouds for long - term spectra . in order to more clearly show our point : the positive correlation between the line width and the line flux is not fake , some evidence to the contrary is shown as follows . if we accept that the positive correlation shown in figure [ fj_3c390 ] was fake for 3c390.3 and the common negative correlation could be applied for the double - peaked emitters , i.e. , both the empirical relation @xmath78 for time - varying size of blr and applying peak separation ( @xmath79 ) to replace the second moment were available for 3c390.3 , we will have @xmath80 , under the virialization assumption . for the spectra of 3c390.3 in shapovalova et al . ( 2001 ) , the line luminosity ratio of maximum @xmath81 to minimum @xmath81 is about @xmath82 , however , the forth power of the line width ratio of minimum @xmath79 to maximum @xmath79 is about @xmath83 , which is against the expected result @xmath84 . similarly , for ngc1097 in schimoia et al . ( 2012 ) and and storchi - bergmann et al . ( 2003 ) , the line luminosity ratio of maximum @xmath85 to minimum @xmath86 is about @xmath87 , however , the forth power of the line width ratio of minimum @xmath79 to maximum @xmath79 is about @xmath88 . similar results can be found for the double - peaked emitters discussed in lewis et al . ( 2010 ) : the parameter @xmath89 is much different from the parameter @xmath90 , @xmath91 and @xmath92 for 3c59 , @xmath93 and @xmath94 for ie0450.3 , @xmath95 and @xmath96 for pictor a , @xmath97 and @xmath98 for cbs74 , @xmath99 and @xmath100 for pks0921 , @xmath101 and @xmath102 for pks1020 , @xmath103 and @xmath96 for pks1739 . the results re - confirm that for long - term observed double - peaked broad lines , the parameter of the peak separation is not one good substitute to trace the kepler velocity of the broad line clouds , and clearly indicates the correlation shown in figure [ fj_3c390 ] is definitely different from the previous reported results about the peak separation in the literature . the results above indicate that for double - peaked emitters , it should be one good choice to check the virialization assumption ( the correlation of line parameters ) by one sample of short - term observed spectra rather than by one sample of long - term observed spectra , in order to ignore the effects of physical disk parameters varying . finally , we consider the question : whether the positive correlation of line parameters could be common for all double - peaked emitters , if the effects of blr physical parameter varying could be clearly considered ? the most important starting point for the positive correlation is that the central ionizing continuum emission is coming from one point - like source , so that there are different arriving times of ionizing photos in different parts of disk - like blrs of double - peaked emitters . however , as discussed in the literature , one external ionizing source should be necessary for some double - peaked emitters ( chen et al . 1989 , eracleous & halpern 1994 , 2003 , luo et al . 2009 , strateva et al . 2003 , 2006 , 2008 , wu et al . 2008 ) , due to the energy budget problem : the total double - peaked broad line flux likely exceeds the viscous energy that can be extracted locally from the accretion disk . if the extended size of the external ionizing source is large enough to cover the disk - like blrs of double - peaked emitters , the expected different time delays for different parts of blrs should be not so apparent , and then no apparent positive correlation between the line width and the line flux of broad double - peaked lines could be found . commonly , parameter @xmath104 ( ratio of the double - peaked broad h@xmath0 luminosity to the viscous power output of the line - emitting portion of the accretion disk ) is used to determine existence of one external ionizing source ( collin - souffrin 1987 , dumont & collin - souffrin 1990 , strateva et al . 2008 , williams 1980 ) . for 3c390.3 , eracleous & halpern ( 2003 , 1994 ) have shown that the parameter @xmath105 is about @xmath106 with energy transfer efficiency @xmath107 being accepted . here the used @xmath86 and @xmath108 are the updated values listed in eracleous & halpern ( 2003 ) . however , we should note that the energy transfer efficiency is much larger than 0.01 for 3c390.3 . based on the empirical relation shown in davis & laor ( 2011 ) and the more recent black hole mass of 3c390.3 in dietrich et al . ( 2012 , @xmath109 ) , the energy transfer efficiency for 3c390.3 is about @xmath110 . so that the parameter of @xmath105 is about @xmath111 , which is much smaller than the limited value of @xmath112 . so that , it is not necessary for one external ionizing source for the double - peaked emitter 3c390.3 . furthermore , through the reverberation mapping technique ( blandford & mckee 1982 , peterson 1993 ) , the uv emission region is within 5 light - days ( dietrich et al . 1998 ) , which is much smaller than the size of blr ( @xmath113 in peterson et al . ( 2004 ) ( certainly , much smaller than the size of blr shown in dietrich et al . 2012 , popovic et al . 2011 , sergeev et al . 2011 and shapovalova et al . so that the time delays for different parts of the blr of 3c390.3 are apparent , and one positive correlation shown in figure [ fj_3c390 ] can be expected . so far , it is still not clear about the structures of the probable needed external ionizing source : external ionizing photos coming from an elevated structure in central region of accretion disk ( standard riaf , cao 2005 , manmoto 2000 ) , or ionizing photos coming from scattering ionizing photons radiated from the inner accretion disk region ( cao & wang 2006 ) , or other unknown structures . before to confirm the structures of the external ionizing source , it is very hard to give one final conclusion that all double - peaked emitters show one positive correlation between line width and line flux for short - term observed spectra . however , after the corrections of the effects of disk physical parameters varying , once one positive correlation between the line width and the line flux can be confirmed for one broad line , it is extremely possible that the broad line comes from central accretion disk . the final conclusions are as follows . * under the virialization assumption and the commonly accepted empirical relation @xmath2 , there should be one strong negative correlation between the line width and the line flux of broad lines for multi - observed spectra of broad line agn : @xmath114 . the negative correlation can be confirmed for some mapped objects , such as the results shown in figure [ fj_ngc5548 ] for ngc5548 . * based on the public spectra of 3c390.3 from the agnwatch project , the correlation between the line width ( second moment ) and the line flux of the double - peaked broad h@xmath0 is checked . however , one unexpected reliable positive correlation is confirmed and shown in figure [ fj_3c390 ] , which indicates there are different structures of blr for the double - peaked emitter 3c390.3 and normal broad line agn . * in the context of the theoretical accretion disk model for double - peaked emitters , the unexpected positive correlation between the line width and the line flux can be naturally explained , due to the different time delays for the different parts of the disk - like blrs of 3c390.3 in one short period . * the virialization assumption is further checked for the double - peaked emitter 3c390.3 , @xmath115 . then , we find that the virialization assumption is still available for 3c390.3 , however , the time - varying size of blr in one period can not be expected through the empirical relation of @xmath2 . for the double - peaked emitter 3c390.3 , stronger continuum emission leads to smaller ( not larger ) flux - weighted size of blr in one moment . * furthermore , we have shown that in the context of the accretion disk model , there are important effects on the correlation of the line parameters from the physical disk parameters varying ( such as the accretion disk precession ) . that is why there are different conclusions in the literature about the correlation between the peak separation and the line flux for long - term observed double - peaked broad lines . moreover , it should be better to check the correlation of line parameters of double - peaked broad line through short - term observed spectra , in order to ignore the effects from physical disk parameters varying . * finally , the correlation between the line width and the line flux is checked for the other double - peaked emitters . however , due to the probable external ionizing source with so far unclear structures , it is hard to confirm that the positive correlation between the line width and the line flux of double - peaked broad lines can be found for all the double - peaked emitters . however , one strong positive correlation between the line width and the line flux of one broad line should strongly indicate accretion disk origination for the broad line . zhang x. g. gratefully acknowledge the anonymous referee for giving us constructive comments and suggestions to greatly improve our paper . zxg gratefully acknowledges the finance support from the chinese grant nsfc-11003043 , and thanks the project of agnwatch ( http://www.astronomy.ohio-state.edu/~agnwatch/ ) to make us conveniently collect the spectra of 3c390.3 and ngc5548 . bachev , r. , 1999 , a&a , 348 , 71 barth a. j. , nguyen m. l. , malkan m. a. , filippenko a. v. , li w. d. , et al . , 2011 , apj , 732 , 121 begelman m. c. , blandford r. d. & rees m. j. , 1980 , nature , 287 , 307 bennert n. , auger m. w. , treu t. , woo j .- h . , malkan m. a. , 2011 , apj , 726 , 59 bentz m. c. , peterson b. m. , pogge r. w. , vestergaard m. , onken c. a. , 2006 , apj , 644 , 133 bentz m. c. , denney k. d. , cackett e. m. , dietrich m. , fogel j. k. j. , et al . , 2007 , apj , 662 , 205 bentz m. c. , walsh j. l. , barth a. j. , baliber n. , bennert v. n. , et al . , 2009 , apj , 705 , 199 bentz m. c. , walsh j. l. , barth a. j. , yoshii y. , woo j. h. , et al . , 2010 , apj , 716 , 993 blandford r. d. & mckee c. f. , 1982 , apj , 255 , 419 boroson t. a. & lauer t. r. , 2009 , nature , 458 , 53 burbidge e. m. & burbidge g. r. , 1971 , apj , 163 , l21 cao x. w. , 2005 , apj , 631 , l101 cao x. w. & wang t. g. , 2006 , apj , 652 , 112 chen k. y. & halpern j. p. , 1989 , apj , 344 , 115 chen k. y. , halpern j. p. & filippenko a. v. , 1989 , apj , 339 , 742 chornock r. , bloom j. s. , cenko s. b. , filippenko a. v. , silverman j. m. , hicks m. d. , lawrence k. j. , mendez a. j. , rafelski m. , wolfe , a. m. , 2010 , apj , 709 , 39 collin - souffrin s. 1987 , a&a , 179 , 60 collin s. , kawaguchi t. , peterson b. m. , vestergaard m. , 2006 , a&a , 456 , 75 davis s. w. & laor a. , 2011 , apj , 728 , 98 denney k. d. , bentz m. c. , peterson b. m. , pogge r. w. , cackett e. m. , et al . , 2006 , apj , 653 , 152 denney k. d. , peterson b. m. , pogge r. w. , adair a. , atlee d. w. , et al . , 2009 , apjl , 704 , 80 denney k. d. , peterson b. m. , pogge r. w. , adair a. , atlee d. w. , et al . , 2010 , apj , 721 , 715 dietrich , m. , kollatschny , w. , peterson , b. m. , bechtold , j. , bertram r. , et al . , 1993 , apj , 408 , 416 dietrich m. , peterson b. m. , albrecht p. , altmann m. , barth a. j. , et al . , 1998 , apjs , 115 , 185 dietrich m. , peterson b. m. , grier c. j. , bentz m .c . , eastman j. , et al . , 2012 , apj , 757 , 53d down e. j. , rawlings s. , sivia d. s. , baker j. c. , 2010 , mnras , 401 , 633 dumont a. m. & collin - souffrin s. 1990 , a&a , 229 , 313 eracleous m. & halpern j. p. , 1994 , apjs , 90 , 1 eracleous m. , livio m. , halpern j. p. , storchi - bergmann t. , 1995 , apj , 438 , 610 eracleous m. , halpern j. p. , gilbert a. m. , newman j. a. , filippenko a. v. , 1997 , apj , 490 , 216 eracleous m. & halpern j. p. , 2003 , apj , 599 , 886 flohic h. m. l. g. & eracleous m. , 2008 , apj , 686 , 138 fromerth m. j. & melia f. , 2000 , apj , 533 , 172 gaskell c. m. , 1988 , apj , 325 , 114 gaskell m. , 1996 , apj , 464 , 107 gaskell c. m. , 2010 , nature , 463 , 1 gezari s. , halpern j. p. , eracleous m. , 2007 , apj , 169 , 167 greene j. e. & ho l. c. , 2004 , apj , 610 , 722 greene j. e. & ho l. c. , 2005 , apj , 630 , 122 greene j. e. , hood c. e. , barth a. j. , bennert v. n. , bentz m. c. , et al . , 2010 , apj , 723 , 409 hartnoll s. a. & blackman e. g. , 2000 , mnras , 317 , 880 hartnoll s. a. & blackman e. g. , 2002 , mnras , 332 , l1 karas v. , martocchia a. & subr l. , 2001 , pasj , 53 , 189 kaspi s. , smith p. s. , maoz d. , netzer h. , jannuzi b. t. , 1996 , apjl , 471 , 75 kaspi s. , smith p. s. , netzer h. , maoz d. , jannuzi b. t. , giveon u. , 2000 , apj , 533 , 631 kaspi s. , maoz d. , netzer h. , peterson b. m. , vestergaard m. , jannuzi b. t. , 2005 , apj , 629 , 61 kaspi s. , brandt w. n. , maoz d. , netzer h. , schneider d. p. , shemmer o. , 2007 , apj , 659 , 997 kelly b. c. & bechtold j. , 2007 , apjs , 168 , 1 kollatschny w. & zetzl1 m. , 2011 , nature , 470 , 366 korista k. t. , alloin d. , barr p. , clavel j. , cohen r. d. , et al . , 1995 , apjs , 97 , 285 krause m. , burkert a. , schartmann m. , 2011 , mnras , 411 , 550 lauer t. r. & boroson t. a. 2009 , apj , 703 , 930 lewis k. t. , eracleous m. , storchi - bergmann t. , 2010 , apjs , 187 , 416 livio m. & xu c. , 1997 , apj , 478 , l63 luo b. , brandt w. n. , silverman j. d. , strateva i. v. , bauer f. e. , et al . , 2009 , apj , 695 , 1227 marziani p. , sulentic j. w. , zamanov r. , calvani m. , dultzin - hacyan d. , bachev r. , zwitter t. , 2003 , apjs , 145 , 199 manmoto t. , 2000 , apj , 534 , 734 netzer h. & marziani p. , 2010 , apj , 724 , 318 oke j. b. , in : superluminal radio sources ; proceedings of the workshop , pasadena , ca , oct . 28 - 30 , 1986 ( a88 - 39751 16 - 90 ) . cambridge and new york , cambridge university press , 1987 , p. 267 - 272 . onken c. a. , ferrarese l. , merritt d. , peterson b. m. , pogge r. w. , vestergaard m. , wandel a. , 2004 , apj , 615 , 645 park d. , woo j. h. , treu t. , barth a. j. , bentz m. c. , et al . , 2012 , apj , 747 , 30 perez e. , penston m. v. , tadhunter c. , mediavilla e. , moles m. , 1988 , mnras , 230 , 353 peterson b. m. , wanders i. , horne k. , collier s. , 1998 , pasp , 110 , 660 peterson b. m. & wandel a. , 1999 , apj , 521 , l95 peterson b. m. , balonek t. j. , barker e. s. , bechtold j. , bertram r.,et al . , 1991 , apj , 368 , 119 peterson b. m. , alloin d. , axon d. , balonek t. j. , bertram r. , et al . , 1992 , apj , 392 , 470 peterson , b. m. , 1993 , pasp , 105 , 247 peterson b. m. , berlind p. , bertram r. , bochkarev n. g. , bond d. , et al . , 1994 , apj , 425 , 622 peterson b. m. , barth a. j. , berlind p. , bertram r. , bischoff k. , et al . , 1999 , apj , 510 , 659 peterson b. m. , berlind p. , bertram r. , bischoff k. , bochkarev n. g. , et al . , 2002 , apj , 581 , 197 peterson b. m. , ferrarese l. , gilbert k. m. , kaspi s. , et al . , 2004 , apj , 613 , 682 peterson b. m. , 2010 , co - evolution of central black holes and galaxies , proceedings of the international astronomical union , iau symposium , volume 267 , p. 151 - 160 popovic l. c. , shapovalova a. i. , chavushyan v. h. , ilic d. , burenkov a. n. , mercado a. , bochkarev n. g. , 2008 , pasj , 60 , 1 popovic l. c. , shapovalova a. i. , ilic d. , kovacevic a. , kollatschny w. , burenkov a. n. , chavushyan v. h. , bochkarev n. g. , leon - tavares j. , 2011 , a&a , 528 , 130 press w. h. , teukolsky s. a. , vetterling w. t. , flannery b. p. , 1992 , numerical recipes in fortran 77 , second edition , published by the press syndicate of the university of cambridge , isbn 0 - 521 - 43064-x , p340 rafiee a. , & hall p. b. , 2011 , apjs , 194 , 42 schimoia j. s. , storchi - bergmann t. , nemmwn r. s. , winge c. , eracleous m. , 2012 , apj , 748 , 145 sergeev s. g. , doroshenko v. t. , dzyuba s. a. , peterson b. m. , pogge r. w. , pronik v. i. , 2007 , apj , 668 , 708 sergeev s. g. , kilmanov s. a. , doroshenko v. t. , efimov y. s. , nazarov s. v. , pronik v. i. , 2011 , mnras , 410 , 1877 shapovalova a. i. , burenkov a. n. , carrasco l. , chavushyan v. h. , doroshenko v. t. , et al . , 2001 , a&a , 376 , 775 shapovalova a. i. , doroshenko v. t. , bochkarev n. g. , burenkov a. n. , carrasco l. , et al . , 2004 , 422 , 925 shapovalova a. i. , popovic l. c. , burenkov a. n. , chavushyan v. h. , ilic d. , et al . , 2010 , a&a , 517 , 42 shen y. & liu x. , 2012 , apj , 753 , 125 sluse d. , schmidt r. , courbin f. , hutsemekers d. , meylan g. , eigenbrod a. , anguita t. , agol e. , wambsganss j. , 2011 , a&a , 528 , 100 sulentic j. w. , marziani p. , dultzin - hacyan d. , 2000 , ara&a , 38 , 521 storchi - bergmann t. , nemmen da s. r. , eracleous m. , halpern j. p. , wilson a. s. , filippenko a. v. , ruiz m. t. , smith r. c. , nagar n. m. , 1997 , apj , 489 , 8 strateva i. v. , strauss m. a. , hao l. , schlegel d. j. , hall p. b. , et al . , 2003 , aj , 126 , 1720 strateva i. v. , brandt w. n. , eracleous m. , schneider d. p. , chartas , g. 2006 , apj , 651 , 749 strateva i. v. , brandt w. n. , eracleous m. , garmire g. , 2008 , apj , 687 , 869 tran h. d. , 2010 , apj , 711 , 1174 van groningen , e. & wanders , i. , 1992 , pasp , 104 , 700 veilleux s. & zheng w. , 1991 , apj , 377 , 89 vestergaard m. , 2002 , apj , 571 , 733 wandel a. , peterson b. m. , & malkan m. a. , 1999 , apj , 526 , 579 wanders i. & peterson b. m. , 1996 , apj , 466 , 174 wang t. g. , & zhang x. g. , 2003 , mnras , 340 , 793 williams r. e. , 1980 , apj , 235 , 939 wu s. m. , wang t. g. , dong x. b. , 2008 , mnras , 389 , 213 wu x. b. , wang r. , kong m. z. , liu f.k . , han j. l. , 2004 , a&a , 424 , 793 zhang x. g. , dultzin d. , wang t. g. , 2007 , mnras , 377 , 1215 zhang x. g. , 2011a , mnras , 416 , 2857 zhang x. g. , 2011b , apj , 741 , 104 zheng w. , sulentic j. w. , binette l. , 1990 , apj , 365 , 115 zheng w. , veilleux s. , grandi s. a. , 1991 , apj , 381 , 418 + notice : the columns of id show the observational date ( jd-2450000 ) and instrument codes . the columns of @xmath116 list the measured line flux of the broad h@xmath14 in unit of @xmath117 . the columns of @xmath18 show the second moment of the broad h@xmath14 in unit of @xmath118 .

in this manuscript , we carefully check the correlation between the line width ( second moment ) and the line flux of the double - peaked broad h@xmath0 of the well - known mapped agn 3c390.3 , in order to show some further distinctions between double - peaked emitters and normal broad line agn . based on the virialization assumption @xmath1 and the empirical relation @xmath2 , one strong negative correlation between the line width and the line flux of the double - peaked broad lines should be expected for 3c390.3 , such as the negative correlation confirmed for the mapped broad line object ngc5548 , @xmath3 . but , based on the public spectra around 1995 from the agnwatch project for 3c390.3 , one reliable positive correlation is found between the line width and the line flux of the double - peaked broad h@xmath0 . in the context of the proposed theoretical accretion disk model for double - peaked emitters , the unexpected positive correlation can be naturally explained , due to different time delays for inner parts and outer parts of disk - like blr of 3c390.3 . moreover , the virialization assumption is checked and found to be still available for 3c390.3 . however , time - varying size of the blr of 3c390.3 can not be expected by the empirical relation @xmath2 . in other words , the mean size of blr of 3c390.3 can be estimated by the continuum luminosity ( line luminosity ) , while , the continuum emission strengthening leads to the size of blr decreasing ( not increasing ) in different moments for 3c390.3 . then , we compared our results of 3c390.3 with the previous results reported in the literature for the other double - peaked emitters , and found that before to clearly correct effects from disk physical parameters varying ( such as effects of disk precession ) for long - term observed line spectra , it is not so meaningful to discuss the correlation of line parameters of double - peaked broad lines . furthermore , due to the probable external ionizing source with so far unclear structures , it is hard to give one conclusion that the positive correlation between the line width and the line flux can be found for all double - peaked emitters , even after the considerations of disk physical parameters varying . however , once one positive correlation of broad line parameters was found , the accretion disk origination of the broad line should be firstly considered . [ firstpage ] galaxies : active galaxies : nuclei galaxies : seyfert quasars : emission lines galaxies : individual : 3c390.3

introduction main results discussions and conclusions acknowledgments

the properties of strongly interacting fermi gases have been the object of great interest in recent years . feshbach resonances allow the tuning of the interaction between trapped particles so that one can study the evolution from a dilute fermi gas to a bose - einstein condensate . in three - dimensional optical lattices one can reach the low - tunneling regime where each site is an essentially isolated harmonic trap occupied by few fermions @xcite . this opens up a new window into the study of few - body systems with two - body scattering lengths @xmath0 that are large compared to the range @xmath1 of the interaction , @xmath2 . large-@xmath3 systems are of particular theoretical interest when particle momenta @xmath4 are small compared to @xmath5 , because then they exhibit universal behavior , that is , the system properties depend essentially only on @xmath0 ( and for bosons or multi - component fermions , also on a three - body parameter ) , but not on the details of the interaction . ( for a review , see ref . the presence of a harmonic trap introduces another parameter , the trap frequency @xmath6 or , equivalently , the trap length @xmath7 , where @xmath8 is the reduced mass of two particles . as long as @xmath9 , the trapped system should still exhibit universal behavior , which for @xmath10 could be significantly different from that of the untrapped system . in the unitary limit @xmath11 , the untrapped two - body system has a bound state at zero energy , and a collection of two - state fermions is characterized purely by the parameter that sets the size of the system . while large-@xmath3 systems have been popular in atomic physics mainly in the last decade , they have been investigated in nuclear physics since its beginning . the two - nucleon ( @xmath12 ) system has two @xmath13-wave channels where @xmath2 : the scattering lengths are about @xmath14 fm and @xmath15 fm in the @xmath16 and @xmath17 channels , respectively , compared with a range of about @xmath18 fm . here we combine two methods previously developed to deal with untrapped particles effective field theory ( eft ) and no - core shell model ( ncsm) in order to provide solutions of few - fermion systems in a harmonic trap . since we are interested in the long - range dynamics , we can approximate the complicated short - range physics as a series of contact interactions that are delta functions with an increasing number of derivatives . this can be formulated as a non - relativistic eft where observables are expressed in an expansion in powers of @xmath19 ( assuming , for simplicity , that the size of all scattering parameters except for the scattering length is set by @xmath1 ) . the eft in the untrapped two - body sector @xcite reproduces the effective - range expansion and is equivalent to a pseudopotential , but can be extended to more - body systems @xcite . it can be shown that in leading order the hamiltonian for two - component fermions consists of a single , non - derivative , two - body contact interaction between the different components . for systems of identical bosons or more - component fermions , a non - derivative , three - body contact interaction is also present at this order . the eft for short - range forces and its generalization for the exchange of light quanta are reviewed in ref . @xcite . the ncsm is a powerful many - body technique for solving the schrdinger equation for @xmath20 strongly interacting particles , where the many - body basis states are constructed using harmonic - oscillator ( ho ) wave functions . in nuclear physics , ncsm is used to describe properties of light nuclei without adjustable parameters @xcite . starting from interactions that describe the @xmath12 scattering phase shifts and selected few - nucleon properties with high accuracy , a unitary transformation is used to construct effective interactions in truncated spaces of a finite number of energy excitations on top of a minimum configuration . for the two - body problem , the maximum number of excitations is equivalent to the number @xmath21 of shells included . after a truncation where higher - body terms in the effective interaction are neglected the so - called cluster approximation numerical diagonalization allows a good description of nuclear energy spectra and other properties . unfortunately , there is no _ a priori _ justification for the cluster approximation in the context of phenomenological interactions . we have recently proposed @xcite a combination of ncsm and eft in which the cluster approximation is seen as a consequence of the eft power counting . instead of performing a unitary transformation on phenomenological interactions , we simply solve the eft within the truncated space . we determine its parameters from some binding energies and then predict other bound - state properties by extrapolation to the limits @xmath22 and @xmath23 . we have successfully applied the method in leading order to systems with @xmath24 . however , application beyond leading order becomes cumbersome because of the increased number of parameters that would have to be adjusted to properties of light nuclei . this motivates us to devise a more flexible approach , which will allow us to determine two - body parameters from two - body data . in this paper , we present the new method and apply it to the case of two - component fermions in a harmonic trap . the two @xcite and three @xcite -body systems with a pseudopotential in a trap have already been studied . we provide an alternative , explicit solution for these systems in a ncsm basis where @xmath6 is kept fixed but @xmath21 is made large . this allows us to test the accuracy of our approach in a non - trivial system . in addition , we present a first solution of the four - body system . the paper is organized as follows . we introduce our approach to the renormalization of the many - body problem in sec . [ sec : many_body ] , and present our results in sec . [ sec : results ] . conclusions and perspectives for future applications are discussed in sec . [ sec : concl ] . here we consider a system of @xmath20 identical two - component fermions of mass @xmath25 , where the two - body scattering length @xmath0 is large compared to the range @xmath1 of the interaction , @xmath2 . this system is assumed to be trapped in a harmonic potential of frequency @xmath6 , whose length parameter is also sufficiently large , @xmath9 . we allow various values of the ratio @xmath26 ranging from the unitary limit @xmath27 to the untrapped case @xmath28 . ( note that we use units such that @xmath29 . ) under these conditions , details of the inter - particle potential are irrelevant , and in leading order can be replaced by a non - derivative two - body contact interaction between different components . without loss of generality we can refer to the two components as spin-1/2 states , in which case the contact interaction acts only in the @xmath30 channel . we denote its interaction strength by @xmath31 . we start with the hamiltonian for trapped particles , @xmath32}\delta^{(3)}(\vec{r}_{i}-\vec{r}_{j } ) , \label{eq : ham}\ ] ] where @xmath33 and @xmath34 represent the position and momentum , respectively , of particle @xmath35 , and @xmath36 $ ] denotes a pair of particles with opposite spin . this hamiltonian can be written as a sum of relative and center - of - mass ( cm ) motion , so that the energy of the system is the sum of internal ( @xmath37 ) and cm ( @xmath38 ) terms : @xmath39 the cm motion is that of a simple ho , so in the following we focus on the relative motion . the internal hamiltonian is not well defined as it stands , since the delta function is singular . renormalization is necessary : after a truncation of the hilbert space with a cutoff in energy or momentum , @xmath31 is taken to depend on the cutoff in such a way that finite , cutoff - independent results are obtained for observables . the many - body problem can not in general be solved analytically , but renormalization can be incorporated easily in numerical calculations , which are formulated from the outset within a finite space . these issues can be made explicit in the two - body system , where the relative motion of two particles of opposite spin is described by @xmath40 \label{eq : rel2b}\ ] ] in terms of the relative coordinate @xmath41 and the reduced mass @xmath42 . in the conventional ncsm approach , one chooses the model space by truncating in the number of shells included in the calculation . more precisely , one considers only ho states with the principal quantum number @xmath43 smaller than a given value @xmath21 . the effective hamiltonian is then constructed via a unitary transformation @xmath44 @xmath45 , designed so that the lowest @xmath46 eigenvalues of @xmath47 are exactly reproduced by @xmath48 ( @xmath46 is the dimension of the model space ) . of course , the transformed hamiltonian no longer has the form of the initial interaction , and , in particular , additional non - local terms are induced even if one starts with a contact interaction . the alternative approach proposed in ref . @xcite is based on principles of eft , and it constructs the interaction between particles making use of power counting . thus , in each model space one preserves the form of the interaction , as dictated by the power counting , and one determines the strength of each interaction strength as a function of @xmath21 and @xmath6 so that some observables are exactly reproduced . for example , one can fix @xmath49 such that the two - body ground - state energy is fixed . in this paper we propose an intermediate approach . thus , as in ref . @xcite we consider only the terms dictated by power counting , but instead of adjusting their strength to reproduce observables in the few - body system , we reproduce the @xmath6-dependent energies of two - body states , as in the conventional ncsm approach . the number of states whose energies are to be reproduced is fixed by the number of coupling constants to be adjusted , and not by the dimension of the model space , as in the conventional approach involving a unitary transformation . we note that , according to the general principles of eft , even if initially we start with only a contact interaction , each truncation is going to induce additional correction terms . effectively , the truncation of the space induces effective range , shape and other parameters , which can be adjusted to the appropriate values as more derivatives of delta functions are included . to renormalize the interaction , we consider the relative hamiltonian ( [ eq : rel2b ] ) and solve the corresponding schrdinger equation , @xmath50\psi(\vec{r } ) = 2 \frac{e}{\omega}\psi(\vec{r } ) , \label{eq : contact2b}\ ] ] in a finite model space defined by @xmath51 . since the delta function acts only on @xmath13 waves , we consider here only those waves . the solution @xmath52 can be expanded in the complete set of ho wave functions @xmath53 with @xmath54 the @xmath13-state solutions of the unperturbed ho , @xmath55 in terms of generalized laguerre polynomials . one can easily verify that the eigenvalues of eq . ( [ eq : contact2b ] ) are given by the consistency condition @xmath56 at this point , the coupling constant @xmath57 is undetermined . we can fix the value of @xmath57 at each @xmath21 by requiring that we reproduce a given measured level @xmath58 of two trapped fermions @xcite . however , the experimental values are very close to the the theoretical prediction of a pseudopotential , given @xcite by the transcendental equation @xmath59 where @xmath0 is the two - body scattering length in the absence of the trap . ( for discussions about the limits of applicability of eq . ( [ eq : eigv_lo ] ) , see ref . the solutions of eq . ( [ eq : eigv_lo ] ) come in levels , and we can use any of the low - energy states to fix the value of @xmath57 . here we use the most natural choice , the ground state . once we adjust the strength of the two - body term for each @xmath21 value ( and , in general , @xmath6 ) , we can perform few- and many - body calculations . the many - body model spaces for @xmath60 are chosen so that the two - body truncation is included consistently in relative coordinates @xcite . note that if one wants negative - parity states , one has to truncate the many - body space to odd @xmath21 . in this case , we use a two - body interaction that is adjusted to the largest even number below @xmath21 , that is , @xmath61 . results for the energies of few - fermion systems are in general complicated functions of @xmath21 and @xmath6 . in the unitary regime the dependence on @xmath6 gets simpler . when @xmath27 , the two - body spectrum of eq . ( [ eq : eigv_lo ] ) is given by @xmath62 , where @xmath63 . taking into account the @xmath64 dependence of eq . ( [ eq : radwfs ] ) , we see from eq . ( [ c0run ] ) that the @xmath6 dependence factorizes : @xmath65 , where @xmath66 is dimensionless and @xmath6 independent . indeed , in this case both @xmath8 and @xmath64 disappear from eq . ( [ eq : rel2b ] ) upon a rescaling @xmath67 ( with @xmath68 a dimensionless variable ) , and the energy can only be proportional to @xmath6 . more generally , rescaling all coordinates in eq . ( [ eq : ham ] ) will ensure that @xmath6 exactly factors out in all the many - body eigenenergies . this was to be expected , given that for @xmath11 the only energy parameter is @xmath6 , which also sets the size of the system . it has been argued @xcite that the proportionality constant between @xmath37 and @xmath6 is related to the short - distance scaling exponent @xmath69 of the wave function . the three - fermion spectrum has been solved semi - analytically in ref . @xcite , while the ground - state energies for @xmath70 have recently been calculated in the case of a short - range potential @xcite . in the opposite limit @xmath28 , the trap is removed . ( [ eq : eigv_lo ] ) gives expected results . there is a single two - body bound state of energy @xmath71 , if the two - body interaction is sufficiently strong , @xmath72 . all other states have energies @xmath73 , where @xmath74 , and correspond to scattering states . when @xmath75 , the interaction is weak and the spectra of all few - body systems are expected to approach those of non - interacting fermions in a ho . this can be seen from eq . ( [ c0run ] ) : as @xmath75 the two - body ground - state energy approaches @xmath76 , leading to a divergence from the @xmath77 contribution and to @xmath78 . in this limit all few - body energies should be set by occupation of ho levels . for example , all other two - body levels are given by the poles in the right - hand side of eq . ( [ c0run ] ) . whether few - body bound states exist when the interaction is sufficiently attractive will be discussed below . in any case , at the end of the calculation we extrapolate to the limit @xmath23 . as we show in the next section , this extrapolation is relatively smooth . our approach can be straightforwardly extended to higher orders . at the two - body level , we add to eq . ( [ eq : ham ] ) interactions proportional to derivatives of the delta function , starting with an @xmath13-wave range correction of relative @xmath79 @xcite . a @xmath80-wave interaction ( the most important between like components ) appears at @xmath81 . for spin-1/2 fermions contact three- and more - body forces act in @xmath80 and higher waves , so these forces are of relative @xmath82 or higher @xcite . many - body properties are thus to a high degree determined by two - body forces alone . the two - body parameters can be determined from an extension of eq . ( [ eq : eigv_lo ] ) to non - negligible - range interactions @xcite . in this section , we apply the method introduced in the previous section to the description of systems of three and four identical fermions in a harmonic trap . we consider the unitary regime @xmath83 as well as a general non - vanishing @xmath26 value , for both positive and negative scattering lengths . in the three - body calculation , we have employed two numerical methods . in the first stage , we have used the three - body code in relative coordinates employed in ref . thus , to handle three identical fermions of spin @xmath84 we allow interaction only in the @xmath17 channel , calculating the isospin @xmath85 solution . in a second stage , we have developed a new program that can handle fermions of arbitrary spin , dropping the isospin from the possible quantum numbers . both codes produce the same results for fermions of spin 1/2 , but the latter can be extended to larger @xmath21 . we should point out that in order to correctly antisymmetrize the three - body system , we include all possible @xmath86 states in each model space . therefore , although we show below only results for the levels that are affected by the contact interaction , we also obtain energy levels that are left unchanged . ( for example , states with @xmath87 can be obtained from addition of two @xmath88 relative angular momenta and total spin @xmath89 or @xmath90 ; however , in such channels the contact interaction is not present since it appears only for @xmath91 . ) figure [ fig : lowsp_3b ] presents the main results of our investigation of the unitary regime . we show the dependence of the lowest - energy levels ( labeled by their spin and parity , @xmath92 ) in units of @xmath6 as function of the ultraviolet cutoff @xmath21 . we have verified numerically that , indeed , @xmath6 factorizes , as we have argued in the previous section . the running of the low - lying energies is fairly smooth when one increases the ultraviolet cutoff , and allows an extrapolation to the limit @xmath93 however , we find that the running depends on the state considered . in order to qualify this statement , we assume a running of the form @xmath94 with @xmath95 , @xmath96 , and @xmath97 fitting parameters . in a previous publication @xcite , we assumed a running for the energy of the many - body system of the form @xmath98 , where @xmath99 was defined as the ultraviolet cutoff . ( because here the solutions depend trivially on @xmath6 , we do not have to include it in the definition of the cutoff . ) this was motivated by the running in the continuum two - body system . here we carry out a more detailed investigation of the running of the three - body solution with the ultraviolet cutoff . allowing @xmath97 to vary in the fit , we find values between @xmath100 and @xmath101 . we find that , in general , the values of @xmath95 are not very sensitive to the value of @xmath97 , but can depend on the features of the running and the values of @xmath21 included in the fit . for example , while in fig . [ fig : lowsp_3b ] the first @xmath102 excited state appears to be smooth , a closer examination reveals some structure , as shown in fig . [ fig : fitex ] . in this figure we exhibit also three possibilities for fitting , where we include all points , only points with @xmath103 , or only points with @xmath104 . since eq . ( [ eq : fitform ] ) is valid for large @xmath21 , our best guess for @xmath97 is obtained in the case when we consider only @xmath104 . in this case we find @xmath105 , and the extrapolated value for the energy of the state is @xmath106 . in tables [ cap : j12m][cap : j32p ] we present our estimates for @xmath107 for the states shown in fig . [ fig : lowsp_3b ] . in our approach , there are two sources of errors : on one hand , we neglect higher - order terms in the expansion of the interaction ( terms involving derivatives of contact interactions ) and , on the other hand , we perform a fit of the form of eq . ( [ eq : fitform ] ) to obtain the large-@xmath21 limit . the variation with @xmath21 allows us to estimate errors associated with missing terms ; thus , for low - lying states , even fairly small values of @xmath21 produce results within about @xmath108 of the extrapolated values . on the other hand , we find that , in general , the extrapolation errors coming from fits to large values of @xmath21 are very small . for the levels shown in tables [ cap : j12m][cap : j32p ] these extrapolation errors are beyond the last digit shown , except for second level in table [ cap : j12 m ] , where the error is 2 in the last digit . in the limit @xmath109 , we should recover the semi - analytical solutions obtained in ref . @xcite under the assumption of a two - body pseudopotential : @xmath110 where @xmath111 , and @xmath112 is the ( real ) solution of a transcendental equation . the agreement between our numerical method and the semi - analytical values , also shown in tables [ cap : j12m][cap : j32p ] , is remarkable and provides a confirmation of the reliability of our method . as a side result , we note that the short - range scaling exponent @xcite we obtain for three particles is @xmath113 . ( while the virial theorem @xcite is not satisfied necessarily in each model space , we expect it to be satisfied in the large-@xmath21 limit . ) .comparison between the results of the present approach ( @xmath114 ) and of the semi - analytical formula from ref . @xcite ( eq . ( [ eq : energ_3tf12 ] ) ) , for [ cap : j12 m ] @xmath102 . [ cols="^,^,^,^,^,^",options="header " , ] the successful test in the unitary regime encourages us to extend the application of our method for arbitrary values of the @xmath26 ratio . experimentally , one can vary the scattering length of trapped particles by means of a magnetic field , and , in principle , obtain a large range of ratios @xmath26 . thus , for illustration , we have considered a range of @xmath26 values , and fixed the coupling constant in each model space so that the ground - state energy given by eq . ( [ eq : eigv_lo ] ) is always exactly reproduced . in fig . [ fig : gs3f ] we show the running of the lowest @xmath115 ( ground state ) energy of the trapped three - fermion system for @xmath116 in comparison with the unitary limit . we obtain different shifts of the energy level with respect to the unitary solution depending on the sign of @xmath0 . if @xmath117 the level is pushed downward in energy with respect to the unitary value , while if @xmath118 the level is pushed upward . this behavior is a reflection of the same type of behavior in the two - body system : while in the unitary limit the two - body ground state is @xmath119 , the corresponding energy for @xmath120 decreases to @xmath121 , and for @xmath122 increases to @xmath123 . in fig . [ fig : gs3f_ba ] we consider both @xmath115 and @xmath124 states , and plot the lowest - energy level in each case as a function of @xmath26 . we show only the extrapolated values @xmath107 obtained by means of eq . ( [ eq : fitform ] ) , taking into account only the points with @xmath104 to reduce fitting errors . for @xmath125 , the ground state is the same as in the unitary limit . however , we observe a tendency for an inversion of the parity of the ground state for @xmath126 , although the positive- and negative - parity levels are nearly degenerate . note that in the region with @xmath127 we might have larger errors from the fitting procedure , because the interaction strength increases with increasing values of the ratio @xmath26 . as discussed in the previous section , for weak two - body attraction the spectrum should approach that of three non - interacting fermions in a ho trap which is itself wide . in the limit @xmath128 , the lowest state is a configuration with two particles in the first @xmath13 state , and the third in the first @xmath80 level . figure [ fig : gs3f_ba ] indeed suggests the asymptotic value @xmath129 ( @xmath130 ) for the energy of the negative ( positive ) -parity state . on the other hand , for strong two - body attraction the relative effect of the contact interaction increases . in order to understand the possible parity inversion , we fit the @xmath127 results for @xmath131 in fig . [ fig : gs3f_ba ] to a quadratic function of @xmath26 ( dashed line ) . in the untrapped @xmath132 limit we find that the lowest positive- and negative - parity levels are very close , and their energy is @xmath133 thus , within errors , the system of three fermions is at the threshold for the scattering of one particle on the bound state of the other two ( dimer ) : the system is not bound , or very weakly bound , and can be viewed as composed of a dimer and an additional particle . the lack of a deep three - body bound state follows in fact from a very naive argument . the existence of a shallow two - body bound state is a consequence of a balance between the renormalized delta - function attraction and the kinetic repulsion . the addition of a third spin-1/2 fermion roughly doubles both , so we do not expect a collapse of the three - body system as one does when the number of pair interactions grows faster than the number of particles , e.g. for identical bosons and multi - component fermions . the parity inversion of the ground state is therefore plausible . in the @xmath72 limit , the positive - parity state likely represents the untrapped @xmath13-wave particle - dimer scattering state that one expects to dominate sufficiently close to threshold . this is consistent with results @xcite for near - threshold particle - dimer scattering . we have also solved the four - fermion problem in the trap . for now , we are limited to smaller model spaces than for three fermions because we use the redstick slater determinant code @xcite that preserves only the third component of the angular - momentum projection , mixing states with good angular momentum . the dimension of the many - body basis thus increases significantly . because we truncate the many - body configurations allowing only a maximum number of excitations on top of the minimum solution , we can eliminate spurious cm contributions when we compute the spectrum of the intrinsic motion . ( the calculations in relative coordinates and in a properly truncated slater determinant basis are , in fact , equivalent . ) moreover , because the hamiltonian is rotationally invariant , we obtain eigenstates with good angular momentum ( for four particles , the lowest positive - parity state has @xmath134 ) , even though most of the individual slater determinants that compose the many - body basis do not have good @xmath135 . figure [ fig : gs4f ] exhibits the running of the lowest positive - parity levels of four harmonically trapped spin-1/2 fermions , with the same parameters as in fig . [ fig : gs3f ] , @xmath136 . because of the present limitation to just a few values of @xmath21 for the four - body system , our fitting procedure is likely to be subject to larger extrapolation errors , although fig . [ fig : gs4f ] shows small variation with the ultraviolet cutoff . thus , in the unitary limit , if in eq . ( [ eq : fitform ] ) we use @xmath137 , we estimate @xmath138 , while if we use @xmath139 , we obtain @xmath140 . ( the small number of points does not allow us to obtain @xmath97 reliably from the fit . ) both values are in line with a previous calculation of the four - body system in the unitary limit @xcite . from these results we can estimate the short - range scaling coefficient @xmath69 @xcite for four particles between @xmath141 and @xmath142 . at non - vanishing @xmath26 we observe the same type of displacement in the lowest positive - parity level as for two- and three - fermion systems . however , our technical limitations in the four - body system do not allow us yet to make a more detailed investigation of a large range of ratios @xmath26 . as expected , the system goes to the non - interacting case in the limit @xmath75 , that is , the ground - state energy is @xmath143 . as for three particles , we can extrapolate to the other limit @xmath72 with a quadratic form in @xmath144 . we find the binding energy in this limit to be about 40% below the threshold of scattering of two dimers . however , because our results are subject to large errors , we leave a more thorough analysis of four - body spectra in different regimes for future work using relative coordinates , when it should be possible to achieve larger ultraviolet cutoffs . we have presented a first application of the ncsm to the description of three- and four - fermion systems in a harmonic trap . based on the general principles of eft , we have proposed a method to renormalize the two - body interaction to be used for few- and many - body calculations in finite model spaces . tests of our results against semi - analytical calculations in the unitary regime suggest good accuracy in estimation of the energy levels of the three - fermion system . moreover , we have extended the application of our method to arbitrary @xmath26 ratios . there are , of course , ways to improve our results . in principle , even if a pseudopotential is a good approximation , in finite model spaces we can add a range correction to the contact term . this should improve the running of observables with the ultraviolet cutoff . although in many cases the running is quite small , we could improve the calculations for @xmath127 , where the interaction is stronger and requires larger model spaces to converge . ( in the four - nucleon system we can follow the evolution of the contribution of different configurations . we find that the components with the lowest - energy configuration dominate for @xmath118 , while other configurations reshuffle small contributions in each model space . however , this situation changes , with a larger contribution of the higher shells , for a stronger interaction such as for @xmath127 . this is not surprising since a stronger interaction becomes more important across shells . ) adding higher - order terms should decrease the magnitude of @xmath96 in eq . ( [ eq : fitform ] ) , thereby allowing for a more precise determination of observables in few- and many - body systems . in principle , we could extend our method even further . first , we can perform similar calculations of other observables ( e.g. , r.m.s . radii ) , with more particles , as well as for fermions of different spin . especially for the latter it will be interesting to investigate the ability of the method to predict efimov states in a finite model space . in the case of untrapped multi - state fermions , the eft power counting requires a contact three - body force to prevent the collapse of the three - body system @xcite . finally , the same type of approach can be considered in nuclear physics , where addition of a cm harmonic term to the intrinsic hamiltonian produces effectively the same type of two - body relative hamiltonian as in eq . ( [ eq : rel2b ] ) @xcite . we thank j. carlson for useful discussions , h. hammer and t. papenbrock for pointing out useful references , and f. werner and y. castin for providing their values for energies shown in tables [ cap : j12m][cap : j32p ] . this research was supported in part by the u.s . department of energy under grant numbers de - ac53 - 06na25396 ( is ) , de - fg02 - 04er41338 ( uvk ) , and de - fg02 - 87er40371 ( jpv ) , by the u.s . national science foundation under grant phy0555396 ( brb ) , by the nederlandse organisatie voor wetenschappelijk onderzoek ( uvk ) , and by brazil s fapesp under a visiting professor grant ( uvk ) . uvk would like to thank the hospitality of the kernfysisch versneller instituut at rijksuniversiteit groningen , the instituto de fsica terica of the universidade estadual paulista , the instituto de fsica of the universidade de so paulo , and the institute for nuclear theory at the university of washington , where part of this work was carried out . bedaque , h .- w . hammer , and u. van kolck , phys . c * 58 * , r641 ( 1998 ) ; phys . . lett . * 82 * , 463 ( 1999 ) ; nucl . a646 * , 444 ( 1999 ) ; nucl . phys . * a676 * , 357 ( 2000 ) ; p.f . bedaque , e. braaten , and h .- w . hammer , phys . lett . * 85 * , 908 ( 2000 ) . d. blume and c.h . greene , phys . a * 65 * , 043613 ( 2002 ) ; m. block and m. holthaus , phys . a * 65 * , 052102 ( 2002 ) ; e.l . bolda , e. tiesinga , and p.s . julienne , phys . a * 66 * , 013403 ( 2002 ) ; a. bhattachryya and t. papenbrock , phys . rev . a * 74 * , 041602(r ) ( 2006 ) .

we apply the general principles of effective field theories to the construction of effective interactions suitable for few- and many - body calculations in a no - core shell model framework . we calculate the spectrum of systems with three and four two - component fermions in a harmonic trap . in the unitary limit , we find that three - particle results are within 10% of known semi - analytical values even in small model spaces . the method is very general , and can be readily extended to other regimes , more particles , different species ( e.g. , protons and neutrons in nuclear physics ) , or more - component fermions ( as well as bosons ) . as an illustration , we present calculations of the lowest - energy three - fermion states away from the unitary limit and find a possible inversion of parity in the ground state in the limit of trap size large compared to the scattering length . furthermore , we investigate the lowest positive - parity states for four fermions , although we are limited by the dimensions we can currently handle in this case .

introduction the many-body problem and renormalization of the interaction results[sec:results] conclusions and outlook acknowledgments

a wide range of cosmological observations indicate that the universe has entered a phase of accelerating expansion , which becomes one of the important puzzles of the contemporary physics . those observations include the type ia supernovae ( snia ) standard candles @xcite , @xcite , the angular location of the first peak in the cmb power spectrum @xcite and baryon acoustic oscillations of the matter density power spectrum @xcite . these evidences represent a great stimulus for theoretical work and originated the concept of dark energy . the dark energy ( de ) models attribute the observed accelerating expansion to an unknown energy component with negative pressure , which dominates the universe at recent cosmological times . the data can be accommodated with a high degree of accuracy , in the @xmath0cdm model which combines the baryons with conventional cold dark matter ( cdm ) candidates and the cosmological constant @xmath0 , which accounts for the dark energy . however a non zero cosmological constant raises the coincidence problem ( why the dm and de are comparable today ) , and is plagued by fine tuning problems @xcite,@xcite,@xcite . alternatively , dark energy may be described by different dynamical scalar field models with time - dependent equation of state , like quintessence @xcite , @xcite ; string theory fundamental scalar known as tachyon @xcite ; k - essence models involving a generalized form of the kinetic energy @xcite,@xcite ; scalar field with negative kinetic term , which provides a solution known as phantom dark energy @xcite ( see @xcite for a review ) . an alternative description of de may be given by perfect fluids with adequate equation of state , like chaplygin gas @xcite , @xcite . the above mentioned scalar field models assume that the dm and de are of different nature , while in the chaplygin gas the dm and de appear as manifestations of the same single fluid at different cosmological epochs , showing a dust - like matter behavior at early times and behaving as cosmological constant at late times @xcite , @xcite . + another description of de is provided by the scalar - tensor theories which contain a direct coupling of the scalar field to the curvature , with the advantage of giving a mechanism to evade the coincidence problem , and naturally allowing ( in some cases ) the crossing of the phantom barrier @xcite . these theories with different couplings to the curvature appear as low energy limit of several higher dimensional theories , and provide a possible approach to quantum gravity from a perturbative point of view @xcite . a coupling between curvature and kinetic terms appears as part of the weyl anomaly in @xmath1 conformal supergravity @xcite . a model with non - minimal derivative couplings was proposed in @xcite , @xcite , @xcite in the context of inflationary cosmology , and recently , non - minimal derivative coupling of the higgs field was considered in @xcite , also as inflationary model . in @xcite a derivative coupling to ricci tensor has been considered to study cosmological restrictions on the coupling parameter , and the role of this coupling during inflation . some asymptotical solutions for a non - minimal kinetic coupling to scalar and ricci curvatures were found in @xcite , and quintessence and phantom cosmological scenarios with non - minimal derivative coupling have been studied in @xcite . a scalar field with kinetic term coupled to a product of einstein tensors has been considered in @xcite . non - minimal coupling of scalar fields ( including kinetic terms ) with modified @xmath2 theories have been also considered to solve the de problem in @xcite , @xcite , @xcite , @xcite , @xcite . + in this paper we consider an explicit coupling between the scalar field the kinetic term and the curvature @xcite , as the source of de and will establish the connection with the chaplygin gas , by obtaining the solution of the field equations that reproduce the cosmological evolution as given by the perfect fluid obeying the equation of state of the chaplygin gas @xcite . the chaplygin gas has attracted much attention in cosmology , as it allows to interpolate between a dust dominated phase of the evolution of the universe in the past , and an accelerated one at recent time . on the other hand , as in the case of the chaplygin gas , the scalar field with the non - minimal kinetic coupling to the curvature , has been shown to be successful in the description of the dm and de without introducing separately the dm term ( see @xcite ) , i.e. it describes the dm and de as manifestations of a common scalar field at different epochs . for this reason , it would be interesting to analyze the connection of the scalar field with non - minimal kinetic couplings with chaplygin gas , which will be the subject of study in the present paper . a potential and coupling function that may give a dynamical description to the chaplygin gas , have been found . the generalized and modified version of the chaplygin gas have been also considered . the scalar field with kinetic couplings to curvature is given by @xcite : [ eq1 ] s=&d^4x . and we will use the flat frw background metric given by the line element [ eq1a ] ds^2=-dt^2+a^2(t)where @xmath3 is the scale parameter . the dimensionality of the coupling constants @xmath4 and @xmath5 depends on the type of function @xmath6 . taking the variation of action ( [ eq1 ] ) with respect to the metric , we obtain a general expression of the form [ eq2 ] r_-g_r=^2 t _ where @xmath7 and the tensor @xmath8 represents the variation of the terms which depend on the scalar field @xmath9 and can be written as [ eq3 ] t_=t_^+t_^+t_^ where @xmath10 , @xmath11 , @xmath12 correspond to the variations of the minimally coupled terms , the @xmath4 and the @xmath5 couplings respectively . due to the interaction between the scalar field and the curvature , the derived expressions for the density and pressure for the scalar field can be regarded as effective ones . from now on , and in order to simplify the equations ( the field equations will contain only second order derivatives ) we will use the restriction @xmath13 , which is equivalent to a coupling of the kinetic term to the einstein tensor @xmath14 ( see @xcite , @xcite ) . evaluating the @xmath15 and @xmath16 components of the eq . ( [ eq2 ] ) in the spatially - flat friedmann - robertson - walker ( frw ) background ( [ eq1a ] ) , it is obtained ( with the hubble parameter @xmath17 , and for homogeneous time - depending scalar field ) [ eq4 ] h^2=and [ eq5 ] -2 - 3h^2=^2where `` dot '' represents the derivative with respect to the cosmological time @xmath18 . taking variation in ( [ eq1 ] ) with respect to the scalar field in the frw background , gives the equation of motion as follows [ eq7 ] + 3h++3h^2(2f()+^2 ) + 18h^3f()+12hf()=0 where the first three terms describe the minimally coupled field . in what follows we will study cosmological solutions to eqs . ( [ eq5 ] ) and ( [ eq7 ] ) giving rise to accelerated expansion , and according to the cosmological scenario described by the chapkygin gas solutions . + the chaplygin gas is described by the following equation of state ( eos ) [ eq8 ] p=- where @xmath19 and @xmath20 are respectively the pressure and density , and @xmath21 is a positive constant . the continuity equation takes the form [ eq9 ] + 3h(-)=0 this equation can be easily integrated in the variable @xmath22 , yielding [ eq10 ] = ^1/2 which according to the friedmann equation gives the following hubble parameter [ eq11 ] h^2=^1/2=^1/2 this solution has the known advantages of describing the presureless matter dominance stage ( epoch ) of the universe at early times ( @xmath23 , normalizing the current value of @xmath3 to @xmath24 ) , and the future universe dominated by the cosmological constant , entering in a de sitter phase at @xmath25 . in the next section we will consider this solution to integrate the equations ( [ eq4 ] ) and ( [ eq7 ] ) with respect to the scalar field @xmath9 and the coupling @xmath26 , and in this manner we will obtain a description of the chaplygin gas cosmology in the frame of the scalar field with non - minimal kinetic coupling to curvature . then , we will follow the same procedure with the generalized and modified chaplygin gas models . in order to integrate the eqs . ( [ eq4 ] ) and ( [ eq7 ] ) for a given hubble parameter , we should impose additional restrictions on the scalar field potential in order to consistently find the rest of the variables , as follows ( see @xcite ) . + in terms of the variable @xmath22 , and defining the function @xmath27 , the eq . ( [ eq7 ] ) can be written as ( after multiplying by @xmath28 ) [ eq12 ] ( h^2)+3h^2++9h^2f+3h^4(f)+18h^4f=0 from eq . ( [ eq4 ] ) , changing to the variable @xmath29 , we can write the product @xmath30 as following [ eq13 ] f=-- taking the derivative of eq . ( [ eq13 ] ) and replacing @xmath31 and @xmath32 into eq . ( [ eq12 ] ) , we arrive at the following equation involving @xmath33 , @xmath34 and @xmath35 [ eq14 ] & 2h^4+h^2(12h^2+)+4h^2 - 2(6h^2+)v + & + 12(3h^2+)=0 in this manner , we obtain a first order differential equation for the functions @xmath33 , @xmath34 and @xmath35 . in order to integrate the equation ( [ eq14 ] ) , and thanks to the fact that the functions @xmath33 and @xmath35 are separated , we can impose a restriction on the scalar field potential given by the equation [ eq15 ] 2h^2-(6h^2+)v+(3h^2+)=0 which simplifies the eq . ( [ eq14 ] ) : [ eq16 ] 2h^2+(12h^2+)=0 * the chaplygin gas solution * + in order to consistently solve the eqs . ( [ eq15 ] ) and ( [ eq16 ] ) , we propose the expression for the hubble parameter @xmath36 , given by that of the chaplygin gas ( [ eq11 ] ) [ eq17 ] h^2=^1/2 defining the scaled hubble parameter @xmath37 , eq . ( [ eq17 ] ) can be written as [ eq18 ] ^2=^1/2 where the parameters @xmath38 and @xmath39 are now dimensionless and are given by [ eq19 ] = ( ) ^2 a , = ( ) ^2 b. where @xmath38 and @xmath39 satisfy the flatness condition ( considering the chaplygin gas dominance ) [ eq19a ] + = 1 replacing @xmath40 in ( [ eq15 ] ) , and defining the dimensionless scalar potential @xmath41 , after changing the eq . ( [ eq15 ] ) to the `` tilde '' variables and integration we obtain the scalar field potential [ eq20 ] ( x)=c e^3x(+ e^-6x)^1/4- e^6x(+ e^-6x)^1/2 _ 2f_1where @xmath42 is the integration constant . replacing @xmath40 in ( [ eq16 ] ) we get the following expression for @xmath33 [ eq21 ] ( x)=^2= , where @xmath43 is the integration constant . integrating the square root of this last equation , we obtain the scalar field as ( considering the @xmath44 sign root ) [ eq22 ] ( x)= _ 2f_1finally , the coupling function @xmath26 is found by replacing the eqs . ( [ eq18])-([eq21 ] ) in the friedmann eq . ( [ eq13 ] ) , giving the result [ eq23 ] f(x)=where @xmath45 . although we can not have an analytic expression for the potential in terms of the scalar field , we can illustrate the behavior of the potential as showed in fig.1 , for a given set of parameters . the constant @xmath42 is selected so that the time variation of the gravitational coupling does not exceed the observational limits ( see below ) . as can be seen from fig . 1 , for the selected values of the parameters , the potential is a monotonic decreasing function of the scalar field . the decreasing runaway behavior of the potential describing dark energy , is a key fact for a realistic cosmological model @xcite , @xcite . 1 _ the evolution of the potential for the chaplygin gas solution with the scalar field for @xmath46 . _ * the generalized chaplygin gas solution * + a generalization of the chaplygin gas has also been considered to describe the dark matter and dark energy . it s equation of state is given by [ eq24 ] p_g=- where @xmath47 and @xmath48 are the energy density and pressure of the generalized chaplygin gas , @xmath49 is a positive constant and @xmath50 is considered to lie in the range @xmath51 , which guarantees the stability and causality ( see below ) @xcite , @xcite , @xcite . note that @xmath52 corresponds to the original chaplygin gas . the eos ( [ eq24 ] ) has an equivalent field theory representation in a generalization of the born - infeld theory @xcite ( in the scalar field representation used in @xcite , the born - infeld lagrangian density is reproduced for @xmath52 ) . solving the continuity equation ( @xmath53 ) , leads to the energy density in terms of @xmath22 [ eq25 ] _ g=(a+b e^-3(+1)x)^ where @xmath21 is a positive integration constant . this density for @xmath54 describes the matter dominated phase at @xmath23 and at the limit @xmath25 describes the de sitter phase dominated by constant density . at the limit @xmath55 , eq . ( [ eq25 ] ) reproduces the @xmath0cdm model . this model also captures the attention because of it s connection with string theory and supersymmetry : the eos ( [ eq24 ] ) can be obtained in the nambu - goto action for @xmath56-branes moving in a @xmath57 dimensional space time @xcite , and has supersymmetric generalization @xcite . as a criteria to constraint the constant @xmath50 we can use the sound speed for the fluid described by eq . ( [ eq24 ] ) , given by [ eq25a ] c_s^2=== where the last is written in terms of the redshift @xmath58 ( @xmath59 ) . at future @xmath60 , and as follows from ( [ eq25a ] ) @xmath61 . therefore , stability requires @xmath62 , and causality requires @xmath63 ( i.e. respects the speed of light limit ) , limiting @xmath50 to the interval @xmath64 ( note that there is not such restriction on @xmath50 at high redshift , during the matter dominated epoch ) . however some authors have considered the `` forbidden '' region @xmath65 if the sound speed @xmath66 is treated as group velocity @xcite , ( see also @xcite for superluminal sound speed ) . according to some observational studies , for the case of the pure gcg the @xmath67 values are favored , which is very close to the @xmath0cdm limit @xcite . this tight restriction may be increased if we consider additionally the barionic matter component ( @xmath68 ) @xcite , or adding cold dark matter and barionic matter components ( @xmath69 ) @xcite . on the other hand , based on observations of the barionic power spectrum , the gcg plus barion matter is favored for @xmath70 @xcite , @xcite . + replacing the density ( [ eq25 ] ) in the friedman equation gives the following scaled hubble parameter [ eq26 ] ^2=^ where the dimensionless @xmath38 and @xmath39 are given by [ eq27 ] = ( ) ^1 + a , = ( ) ^1 + b and satisfy the flatness condition ( for pure generalized chaplygin gas content ) @xmath71 . replacing ( [ eq26 ] ) in ( [ eq15 ] ) and solving the eq . ( [ eq15 ] ) in `` tilde '' variables we obtain the following solution for the scalar potential [ eq28 ] ( x)= & c e^3x(+ e^-3(1+)x)^- + & e^3(1+)x(+ e^-3(1+)x)^ _ 2f_1 where @xmath42 is the integration constant . solving ( [ eq16 ] ) with @xmath40 given by ( [ eq26 ] ) , we find the expression for @xmath33 [ eq29 ] ( x)=^2= , where @xmath43 is the integration constant . integrating the square root of ( [ eq29 ] ) it follows [ eq30 ] ( x)= _ 2f_1and the corresponding coupling function , as follows from ( [ eq13 ] ) and ( [ eq26]-[eq29 ] ) is [ eq31 ] f(x)= & where @xmath72 . in all equations the dependence on the redshift @xmath58 or in the scale factor @xmath3 is obtained by replacing @xmath73 . in fig.2 we plot the behavior of the potential in terms of the scalar field for two values of @xmath50 corresponding to the `` physical '' region @xmath64 , and the superluminal region @xmath74 which also accommodates in some astrophysical observations @xcite , @xcite . 2 _ the potential for the gcg versus the scalar field for @xmath75 ( dashed ) , and @xmath76 . note the decreasing behavior , which is an important characteristic for dark energy potentials _ the scalar field potentials for the non - minimally coupled scalar field , that reproduce the dynamics of the chaplygin and generalized chaplygin gas are decreasing functions of the scalar field . note the runaway behavior for all curves , which are characteristic of well behaved dark energy potentials @xcite , @xcite . the parameters are chosen in such a way that @xmath77 is definite positive , at least for @xmath78 . for other choices of the parameters , it can be shown numerically that the curves can have one maximum . @xmath42 is used to accomplish the observational restrictions on the time variation of the gravitational coupling . + so , we have reconstructed the scalar model with kinetic couplings to curvature ( [ eq1 ] ) , for a given hubble parameter describing the cg and gcg cosmologies . in both cases we exploited the additional degree of freedom represented in the coupling function , to constraint the scalar potential in a way consistent with the friedmann equations . a more general formulation of cosmological reconstruction method ( in time and @xmath29 variables ) for a number of modified gravities including scalar tensor theories , have been performed in @xcite , @xcite . in the scalar tensor theories considered in @xcite , @xcite , the scalar potential and couplings are reconstructed by using an apropriate redefinition of the scalar field , and giving the particular type of cosmological evolution encoded in @xmath34 . the reconstruction was considered in the cosmological time and the e - folding variable @xmath29 , and concrete examples of accelerated late time cosmologies have been provided . in this works for the case of @xmath2 gravity , the reconstruction was achieved by introducing an auxiliary scalar field . * the modified chaplygin gas solution * + the modified chaplygin gas ( mcg ) is defined for the equation of state [ eq32 ] p = b- a generalized version of an equation of state that includes cg , gcg and mcg have been considered in @xcite and @xcite . integrating the continuity equation gives the following energy density [ eq33 ] = ( + ce^-x)^=(+)^ where @xmath79 and @xmath42 is the integration constant . this equation has the same functional dependence on @xmath29 or @xmath3 , as the case of the gcg , except that additional limits can be obtained due to additional factor @xmath80 in the exponent of eq . ( [ eq33 ] ) , with respect to the corresponding one in eq . ( [ eq25 ] ) . therefore , the resulting potential , scalar field and coupling function corresponding to the mcg shows the same dependence as given in eqs . ( [ eq28]-[eq31 ] ) , and it can be shown that all the equations of the mcg , at @xmath81 become the corresponding equations for the gcg . additionally to the gcg , the mcg model reproduces the radiation dominated phase of the universe at high redshift , as can be seen from ( [ eq33 ] ) for @xmath82 and @xmath23 ( neglecting the constant term ) . in general the mcg reproduces the barotopic fluid with constant equation of state ( @xmath83 ) at earlier epochs when the constant term may be neglected . resumming , the non - minimally coupled scalar field also reproduces the dynamics of the modified chaplygin gas cosmology , with well behaved scalar field , potential and coupling function . + * the time variation of the gravitational coupling * + we can meet the constraints on the current value and the time variation of the gravitational coupling @xcite , by appropriately defining or constraining the constants @xmath42 and @xmath43 . the effective gravitational coupling from ( [ eq4 ] ) is given by [ eqg1 ] g_eff= where we used @xmath7 and @xmath84 . in terms of @xmath29 , the time variation of the gravitational coupling can be written as [ eqg2 ] = ( fh^2)h . replacing the product @xmath85 from eq . ( [ eq13 ] ) , and evaluating at the present time ( @xmath86 ) , the eq . ( [ eqg2 ] ) can be written as [ eqg3 ] |_x=0=h_0 where @xmath87 , @xmath88 valuated at @xmath86 and the parameters of the model appearing in @xmath40 have been fixed , so that the resulting expression ( [ eqg3 ] ) depends on the constants of integration @xmath42 and @xmath43 . we can meet the restrictions imposed by the current observations on the value and the time variation of the gravitational coupling @xcite , if @xmath89 and @xmath90 satisfy the constraints : @xmath91 and @xmath92 ( these restrictions are actually softer : @xmath93 and @xmath94 ) . thus , for the chaplygin gas solution given by ( [ eq17 ] ) , ( [ eq20]-[eq23 ] ) with @xmath95 ( @xmath96 ) , under above restrictions we found @xmath97 and @xmath98 , which give the potential plotted in fig . 1 . following the same procedure for the generalized chaplygin gas , we found @xmath99 , @xmath100 ( for @xmath101 ( @xmath102 ) , @xmath103 ) and @xmath104 , @xmath100 for the same values of ( @xmath38 , @xmath39 ) and @xmath105 , with the respective potentials plotted in fig . the cosmological implications of the chaplygin gas model have been intensively investigated in recent literature . we considered the model of scalar field with kinetic terms coupled non - minimally to the scalar field and to the curvature , to give a dynamical description of the chaplygin gas model of dark energy . we have found analytical expressions for the reconstructed scalar field and potentials that describe the standard , the generalized and modified chaplygin gas models of dark energy dark mater unification . thanks to the presence of the coupling function @xmath6 , we could impose a restriction on the potential through eq . ( [ eq15 ] ) , which allowed us to find the solutions that lead to dynamical description of the chaplygin gas cosmology . the results show that the obtained potentials @xmath77 decrease with the evolving scalar field @xmath9 . from eqs . ( [ eq20],[eq22 ] ) it follows that @xmath9 is an increasing function of the redshift and @xmath35 is a decreasing function of the redshift , which means that the scalar potential is a decreasing function of the scalar field . in fig 1 . we show the @xmath77 dependence for the standard chaplygin gas in the redshift interval @xmath106 $ ] . the same behavior follows from eqs . ( [ eq28],[eq30 ] ) for the gcg as shown in fig . the runaway behavior of all the potentials is a desired property of dark energy potentials as the relevance of the scalar potential increases at the present epoch ( low redshift ) , as showing in figs . 1 and 2 . although the problem of perturbations has not been considered here , it should be noted that , despite of the equivalence found between the specific reconstructed scalar model and the chaplygin gas , under perturbations of the scalar field ( which induce additional perturbations of the metric through the kinetic coupling ) it may be expected that this equivalence could not go beyond perturbative corrections . + in order to satisfy the current restrictions on the newtonian coupling , we can use the freedom in the integrations constants @xmath42 and @xmath43 , in order to control the actual value of @xmath107 and it s time variation . these conditions can be satisfied by imposing the inequalities @xmath93 and @xmath94 . + the above results show that the scalar field model with derivative couplings to curvature considered here , provide a dynamical scenario to describe the chaplygin gas cosmology . the wide variety of phenomenologically acceptable solutions @xcite , support the capability of this model to explain the current status of the accelerated expansion of the universe , through different cosmological scenarios . this work was partially supported by the seneca foundation ( spain ) under the program pctrm 2007 - 2010 and the spanish ministry of science and education grant fpu ap2008 - 00919 ( jjfm ) . lng thanks the cern th - division for kind hospitality . 99 m. hicken et al . , astrophys . j. 700 , 1097 ( 2009 ) [ arxiv:0901.4804 [ astro-ph.co ] ] . m. kowalski et al . , astrophys . j. 686 , 749 ( 2008 ) [ arxiv:0804.4142 [ astro - ph ] ] e. komatsu et al . [ wmap collaboration ] , astrophys . 180 , 330 ( 2009 ) [ arxiv:0803.0547 [ astro - ph ] ] . w. j. percival et al . , arxiv:0907.1660 [ astro-ph.co ] . s. weinberg , rev . modern phys . 61 , 1 ( 1989 ) p. j. e. peebles and b. ratra , rev . 75 , 559 ( 2003 ) [ arxiv : astro - ph/0207347 ] t. padmanabhan , phys . 380 , 235 ( 2003 ) [ arxiv : hep - th/0212290 ] b. ratra and p. j. e. peebles , phys . rev . d*37 * , 3406 ( 1988 ) c. wetterich , nucl . phys b. 302 , 668 ( 1988 ) . t. padmanabhan and t. r. choudhury , phys . d 66 , 081301 ( 2002 ) . c. armendariz - picon , v. mukhanov , and p. j. steinhardt , phys . 85 , 4438 ( 2000 ) t. chiba , t. okabe and m. yamaguchi , phys d 62 , 023511 ( 2000 ) ; astro - ph/9912463 . r. r. caldwell , phys . * b 545 * , 23 ( 2002 ) ; astro - ph/9908168 e.j . copeland , m. sami and s. tsujikawa , int . d15 , 1753 ( 2006 ) ; hep - th/0603057 n. bilic , g. b. tupper and r. d. viollier , phys . lett . b535 , 17 ( 2002 ) ; arxiv : astro - ph/0111325 a. y. kamenshchik , u. moschella and v. pasquier , phys . lett . * b511 * , 265 ( 2001 ) ; gr - qc/0103004 l. perivolaropoulos , jcap 0510 , 001 ( 2005 ) ; arxiv : astro - ph/0504582 j. f. donoghue , phys.rev . d**50 * * , 3874 ( 1994 ) ; gr - qc/9405057 hong liu and a.a . tseytlin , nucl . b * 533 * , 88 ( 1998 ) ; hep - th/9804083 s. nojiri and s.d . odintsov , phys . lett . * b444 * , 92 ( 1998 ) ; hep - th/9810008 luca amendola , phys . lett . * b301 * , 175 ( 1993 ) ; gr - qc/9302010 s. capozziello , g. lambiase , gen . grav . * 31 * , 1005 ( 1999 ) ; gr - qc/9901051 s. capozziello , g. lambiase , h .- j.schmidt , annalen phys . * 9 * , 39 ( 2000 ) ; gr - qc/9906051 c. germani and a. kehagias , arxiv:1003.2635[hep - ph ] scott f. daniel and robert r. caldwell , class . quant.grav * 24 * , 5573 ( 2007 ) ; arxiv:0709.0009[gr - qc ] s.v . sushkov , phys . rev . d**80 * * , 103505 ( 2009 ) ; arxiv:0910.0980 e.n.saridakis , s.v.sushkov , phys . rev . d**81 * * , 083510 ( 2010 ) ; arxiv:1002.3478 c. gao , jcap 06 , 023 ( 2010 ) ; arxiv:1002.4035[gr - qc ] s nojiri and s.d . odintsov , phys.lett . * b599*,137 ( 2004 ) ; astro - ph/0403622 ; pos wc2004 , 024 ( 2004 ) , hep - th/0412030 . t.inagaki , s. nojiri and s.d . odintsov , jcap 0506 , 010 ( 2005 ) , gr - qc/0504054 . g.allemandi , a.borowiec , m.francaviglia and s.d . odintsov , phys.rev.d 72,063505 ( 2005 ) , gr - qc/0504057 . g.cognola , e.elizalde , s nojiri and s.d . odintsov , arxiv:0909.2747 , the open astronomy journal 2009 , in press . s nojiri , s d odintsov and p.tretyakov , prog . 172 * , 81 , 2008 ; arxiv:0710.5232 . l. n. granda , jcap 07 , 006 ( 2010 ) ; arxiv:0911.3702 [ hep - th ] l. n. granda , jcap 07 , 021 ( 2010 ) ; arxiv:1005.2716 [ hep - th ] l. n. granda , class . quant.grav * 28 * , 025006 ( 2011 ) ; arxiv:1009.3964 [ hep - th ] i. zlatev , l. wang , p. j. steinhardt , phys . * 82 * , 896 ( 1999 ) ; astro - ph/9807002 a. r. liddle , r. j. scherrer , phys . d**59 * * , 023509 ( 1999 ) ; astro - ph/9809272 m. c. bento , o. bertolami , a. a. sen , phys . rev . d**66 * * , ( 2002 ) 043507 m. bordemann and j. hoppe , phys . lett . * b 317 * , 315 ( 1993 ) ; hep - th/9307036 r. jackiw and a. p. polychronakos , phys . d * 62 * , 085019 ( 2000 ) ; hep - th/0004083 v. gorini , a.y . kamenshchik , u. moschella , o.f . piattella , a.a . starobinsky , jcap * 02 * , 016 ( 2008 ) ; arxiv:0711.4242 y. urakawa and t. kobayashi , jcap 1007 , 027 ( 2010 ) ; arxiv:0907.1191 d. bertacca and n. bartolo , jcap * 11 * , 026 ( 2007 ) ; arxiv:0707.4247 l. amendola , i. waga and f. finelli , jcap * 11 * , 009 ( 2005 ) ; astro - ph/0509099 l. amendola , f. finelli , c. burigana and d. carturan , jcap * 68 * , 005 ( 2003 ) ; astro - ph/0304325 . s. nojiri , s. d. odintsov , j. phys . 66 , 012005 ( 2007 ) ; hep - th/0611071 s. nojiri , s. d. odintsov , arxiv:1011.0544v4 [ gr - qc ] s. nojiri , s. d. odintsov , s. tsujikawa , phys . rev . d71 , 063004 ( 2005 ) ; hep - th/0501025 s. nojiri , s. d. odintsov , phys . d70 , 103522 ( 2004 ) ; hep - th/0408170 j. p. uzan , rev . 75 * , 403 ( 2003 ) ; hep - ph/0205340

in the frame of the scalar field model with non minimal kinetic coupling to gravity , we study the cosmological solutions of the chaplygin gas model of dark energy . by appropriately restricting the potential , we found the scalar field , the potential and coupling giving rise to the chaplygin gas solution . extensions to the generalized and modified chaplygin gas have been made .

introduction the models and field equations standard, generalized and modified chaplygin gas solutions discussion acknowledgments