Datasets:

blesspearl

/

arxiv-summarization

like 0

1603.02451

the connection between hadron - nucleus interaction and the corresponding hadron - nucleon interaction could provide information on possible modifications of strong interaction properties in the nuclear medium .the interaction of low energy pions with nuclei , and pionic atoms in particular , is a good example where the dominance of the @xmath4-wave ( 3,3 ) resonance in the pion - nucleon interaction underlies the kisslinger @xcite and the ericson - ericson @xcite models for the pion - nucleus interaction .the depth of penetration of a meson into the nucleus is a key element in this connection and indeed below 80 mev empirical pion - nucleus potentials display the characteristics of the @xmath4-wave model @xcite . as the pion energy increases towards the resonance with the associated increased absorption, it is found that good description of the pion - nucleus interaction is possible also without an explicit @xmath4-wave model @xcite . the present work deals with @xmath0 meson - nucleus interaction at low energies where the total @xmath0-nucleon cross sections are small and the kaon penetrates well into the nucleus .this property of the @xmath0 is well known @xcite and some experiments in the 1990s were motivated by it @xcite .the present study was inspired by recent works @xcite where algorithms based on _ in - medium kinematics _ have been applied in calculations of optical potentials from the corresponding hadron - nucleon scattering amplitudes . applying this approach to @xmath0 mesons ,the kaon - nucleon scattering amplitude is presented in terms of the mandelstam variable @xmath5 in the nuclear medium @xmath6 where @xmath7 .@xmath8 is the laboratory kinetic energy of the kaon and @xmath9 is an average binding energy of a nucleon . in the nuclear medium@xmath10 0 where @xmath11 is determined by the beam energy and the optical potential , @xmath12 is determined by the nuclear environment .averaging over angles , @xmath13 substituting locally @xmath14 with @xmath15 the kaon - nucleus optical potential and @xmath16 the coulomb potential .for the nucleon the fermi gas model yields @xmath17 with @xmath18=23 mev the average kinetic energy , @xmath19 and @xmath20 the local and average nuclear densities , respectively .the simplest ` @xmath21 ' form of the optical potential is @xcite @xmath22 where @xmath23 is the c.m . reduced energy @xmath24 in terms of the total energies @xmath25 for the projectile and @xmath26 for the target and @xmath27 is a kinematical factor resulting from the transformation of amplitudes between the projectile - nucleon and the projectile - nucleus systems .@xmath28 is the isospin - averaged free @xmath0-n forward scattering amplitude .defining @xmath29 with @xmath30 , then to first order in @xmath31 and @xmath32 one gets @xcite @xmath33,\ ] ] with @xmath34 .this value of @xmath35 serves as the argument of the _ in medium _ kaon - nucleon amplitude in constructing the kaon - nucleus optical potential .figure [ fig : free ] shows the isospin - averaged @xmath0-nucleon amplitude as function of the c.m .energy @xmath35 , taken from the said software package @xcite including @xmath36 and @xmath37 partial waves .vertical dashed lines indicate the energies corresponding to the four experimental energies considered in the present work .it is seen that the variation of the imaginary part of the amplitude is significant and could result in observable effects within the above algorithm .-nucleon amplitudes .vertical dashed lines indicate the four experimental energies discussed in the present work.,scaledwidth=75.0%,height=264 ] with the relatively small @xmath0-n total cross sections below 800 mev / c kaons penetrate deeply into nuclei and the above algorithm can be tested with greater sensitivity than before . in analyses of pionic atomssince true absorption is impossible on a single nucleon , a two - nucleon phenomenological term is always present in the potential that fits experimental results .similarly phenomenological two - nucleon terms are found in analyses of elastic scattering of @xmath38 by nuclei at low energies . for @xmath1 atoms multi - nucleon termsare found when comparing calculations with experiment . for @xmath39 mesonsthere are no experimental results yet for @xmath39-nucleus interaction to compare with . for anti - protonsthere is a vast collection of experimental results but the very strong absorption precludes a real test of the above algorithm in the nuclear medium . consequently the @xmath0-nucleus below 800 mev / c is a preferred probe because there are experimental results to compare with and there is no evidence that there are significant multi - nucleon contributions . in the present work we use mostlyreaction and total cross sections on nuclei as these have been measured to good accuracy .few experimental angular distributions for elastic scattering of @xmath0 by c are discussed below . in section [ sec : data ] we briefly review past work on the topic of @xmath0-nucleus integral cross sections and in section [ sec : results ] we apply the above approach to compare between experimental results and calculations . section [ sec : summ ] is summary and conclusions .the pioneering experiment of bugg et al .@xcite showed that above 800 mev / c total cross sections for @xmath0 on carbon were smaller than six times the corresponding cross sections on deuteron , as expected .in contrast , below that momentum the reverse was true , exhibiting enhancement of the nuclear total cross section relative to the nucleon cross section .this statement obviously depends on the assumption that nuclear medium effects on the deuteron are very small .later siegel , kaufmann and gibbs @xcite showed that calculations of the ratio @xmath0-carbon to @xmath0- deuteron total cross sections which included traditional medium corrections and their uncertainties , failed to reproduce the above enhancement , thus suggesting more exotic mechanisms .several publications that followed considered density - dependent effective masses of the vector mesons @xcite , meson exchange current @xcite , pion cloud contribution @xcite and mesons exchanged between the @xmath0 and the target nucleons @xcite . the experimental situation with @xmath0-nucleus interactions changed when in the early 1990s total cross sections for d , @xmath2li , @xmath40c ,si and ca were measured at four momenta below 800 mev / c @xcite .the method was to measure transmissions through a target using detectors subtending different solid angles and extrapolating to zero solid angle after applying model - dependent corrections .the same set of transmission measurements was later reanalysed @xcite to determine not only total cross sections but also total reaction cross sections that are less dependent on applied corrections . that increased the number of measured integral cross sections from 16 to 32 , in addition to those on d. self - consistent analyses of this set of data@xcite confirmed that the cross sections on c , si and ca were 15 - 20% larger than expected compared to d and to the very low - density nucleus @xmath2li .phenomenological fits to the data required mainly increased imaginary part of the effective @xmath0-nucleon amplitudes in the medium .analysis of the same data by peterson @xcite in terms of an average @xmath41-wave phase - shift showed good agreement with experiment when increasing this phase - shift in the medium relative to free space . a possible link between the additional reactive content and @xmath0 interactions with two nucleonswas explored in @xcite ..comparisons between total cross sections ( in mb ) of @xmath0 on @xmath2li with three times the cross sections on deuteron . only statistical errors are quoted . [cols="^,^,^,^,^",options="header " , ] the experimental results used in the present analysis are reaction and total cross sections for @xmath0 on @xmath2li , c , si and ca from ref.@xcite . total cross sections on d are taken from ref.@xcite .table [ tab : dli ] compares total cross sections of @xmath0 on @xmath2li with three times the cross sections on deuteron .it is seen that within 2% there are no indications for medium effects in the results for @xmath2li , where the average density is about half of the average density for other nuclei .consequently in what follows we use @xmath2li as a reference . ) as function of energy for a full nuclear density .vertical dashed lines indicate the experimental energies.,scaledwidth=75.0%,height=264 ] ) and ( [ eq : sqrts]).,scaledwidth=75.0%,height=264 ] before applying eq.([eq : sqrts ] ) to calculate @xmath0-nucleus cross sections it is necessary to correct the free @xmath0-nucleon amplitudes by considering pauli correlations .the corrections of waas , rho and weise ( wrw ) @xcite replace @xmath42 of eq.([eq : potl ] ) by @xmath43 where @xmath44 are the isospin @xmath45 free @xmath0-nucleon forward scattering amplitudes. therefore the medium corrections are determined by the quantity @xmath46 where @xmath47 with @xmath48 the c.m .energy and the fermi momentum @xmath49 given by @xmath50 . for kaonic atoms@xmath51 and the expression in brackets is 1 , but for scattering ( @xmath52 ) the integral ( and the corrections ) go down rapidly with increasing energy .figure [ fig : wrwint ] shows as an example , four times the above integral as function of energy for a fermi momentum corresponding to central nuclear density .li for fixed energies ( top ) and for variable energies eq.([eq : sqrts ] ) ( bottom ) .r and t indicate reaction and total cross sections , respectively.,title="fig:",scaledwidth=75.0%,height=321 ] li for fixed energies ( top ) and for variable energies eq.([eq : sqrts ] ) ( bottom ) .r and t indicate reaction and total cross sections , respectively.,title="fig:",scaledwidth=75.0%,height=321 ] equations ( [ eq : potl ] ) and ( [ eq : sqrts ] ) define a density to energy transformation through the scattering amplitudes . as the real part of the potential determines the energy and , in turn ,the energy and density determine the amplitudes , a self - consistent solution is required .good convergence is usually achieved after 4 - 5 iterations and an example for this transformation is shown in fig .[ fig : delta21 ] for 488 mev / c @xmath0 interacting with ca .an energy range of 30 mev that corresponds to penetration of the kaon into regions of up to full nuclear density could be significant , as can be seen from fig .[ fig : free ] ) ( top ) , and with pauli correlations also included ( bottom).,title="fig:",scaledwidth=75.0%,height=321 ] ) ( top ) , and with pauli correlations also included ( bottom).,title="fig:",scaledwidth=75.0%,height=321 ] as a first step in comparisons between calculation and experiment we show in the top part of fig .[ fig : lifixed ] reaction and total cross sections on @xmath2li .the calculations use amplitudes at _ fixed _ energies related to the beam energies without respecting eq.([eq : sqrts ] ) and without the pauli correlation correction ( wrw ) .ratios of experimental to calculated cross sections are shown for the four energies , alternating between reaction ( r ) and total ( t ) cross sections .it is gratifying that to within 5% the calculations reproduce the absolute scale of the experimental results .however , upon closer examination it is clear that there is a systematic difference between calculated and experimental dependence on energy of the cross sections . the bottom part of fig .[ fig : lifixed ] shows similar results for calculations based on the full variable energy of eq.([eq : sqrts ] ) and it is evident that the energy dependence is reproduced very well by calculations .the value of @xmath18 ( eq.([eq : sqrts ] ) ) was adjusted to the low density of @xmath2li .adding also the pauli correlations made very small difference .it is therefore seen that ( i ) from the energy - dependence of the ratios the prescription of eq.([eq : sqrts ] ) is supported by experiment and ( ii ) effects due to pauli correlations are very small for the low - density @xmath2li nucleus . turning to the other target nuclei , namely c , si and ca , calculations with fixed energies ( not shown ) display systematic deficiencies regarding the variation with energy of the experimental to calculated ratios , similarly to the results for @xmath2li .figure [ fig : nucvar ] ( top ) shows the ratios obtained using the @xmath35 algorithm eq.([eq : sqrts ] ) but without the pauli correlations corrections of wrw .the energy dependence of the ratios for the reaction cross sections is seen to be well reproduced by the calculations but there is a residual disagreement in the energy dependence for the total cross sections , notably for c. the bottom part of fig .[ fig : nucvar ] shows these ratios where the wrw in - medium corrections have also been included , and the improved agreement is evident .the larger scatter of the ratios for si has been observed also in earlier analyses @xcite .agreement to 2 - 3% between calculation and experiment for the reaction and total cross sections on the very low density @xmath2li , as obtained above , had not been observed in earlier analyses .consequently we deal here directly with the cross sections for the ` normal ' targets of c , si and ca , without the use of ` super - ratios ' , namely , without normalizing experimental to calculated ratios for a given target to the corresponding ratios for @xmath2li . moreover , with the present model for variations of the in - medium energies we try to fit calculations to experiment for the three targets at the four energies put together , a total of 24 data points .the first phenomenological approach is to simply multiply separately the real and the imaginary parts of the amplitudes by a respective scaling factor , the same for all energies and four targets .obviously no fit to the data is possible with @xmath2li included because for @xmath2li no scaling factors are expected whereas this is not the case for c , si and ca . removing @xmath2li from the data , fits of two parameters to the three heavier targets yields for 24 data points @xmath53=23.8 when using assigned errors of @xmath543.7% as implied by the scatter of the ratios of calculated to experimental cross sections ( see fig .[ fig : nucvar ] ) .the scaling factors are then @xmath55=0.87@xmath540.10 for the real part and @xmath56=1.41@xmath540.02 for the imaginary part of the potential .a separate fit to the eight @xmath2li cross sections leads to @xmath53=8.7 with scaling factors of 0.71@xmath540.10 and 1.02@xmath540.02 for the real and imaginary parts , respectively .reducing the assigned errors all the way down to the statistical errors does not change the resulting scaling factors .consequently we conclude that the real part of the amplitudes needs rescaling by 0.9@xmath540.1 whereas the imaginary part needs rescaling by 1.40@xmath540.02 in order to fit the reaction and total cross sections on c , si and ca .alternatives to rescaling of the input amplitudes have been tried too by adding a phenomenological term to the potential in order to fit the c , si and ca cross sections .fits were possible by adding a term with higher powers of density than linear , to simulate two - nucleon processes .another version was to add a linear term as a @xmath4-wave potential of the kisslinger type @xcite .both options could close the 23% gap between the calculated cross sections based on the input amplitudes and experiment .however , comparisons with the few available results on the elastic scattering of @xmath0 by c did not support such approaches . by c. experimental resultsare from ref .dashed curve for amplitudes without rescaling , solid curve for rescaling by constants , dot - dashed curve for added phenomenological @xmath57-wave potential , see text.,scaledwidth=75.0%,height=321 ] out of the few experiments below 800 mev / c we chose elastic scattering of @xmath0 by c at 715 mev / c @xcite . in fig .[ fig : angdist ] the dashed curve shows the predictions made using the present model with in - medium kinematics and wrw pauli correlations effects .the solid curve shows the results when the above rescaling factors @xmath55 and @xmath56 are applied to the real and imaginary parts , respectively , to fit the ( enhanced ) reaction and total cross sections .both are in fair agreement with experiment , showing very little sensitivity of the angular distribution to modifications required to fit the reaction and total cross sections .the option of adding a phenomenological non - linear term or a linear @xmath4-wave term to the potential ( not shown ) , produced angular distributions that disagree sharply with experiment beyond the first minimum .another attempt was to use for the in - medium forward scattering amplitudes only the @xmath41-wave amplitudes from said @xcite , thus neglecting the contributions from the @xmath4- and @xmath37-wave terms , and to add an explicit phenomenological @xmath4-wave potential @xcite . that too made it possible to fit the reaction and total cross section with only two parameters . in this casethe balance between repulsive and attractive real parts of the potential changed and the predicted angular distribution , dot - dashed curve in fig .[ fig : angdist ] , is acceptable .very similar results were obtained with @xmath0 elastically scattered by c at 635 mev / c @xcite .we have also checked if it was possible to include @xmath2li in the data and to fit all 32 cross sections with additional terms in the potential .this approach failed before @xcite and it has likewise failed in the present work . however , it was shown in ref .@xcite that a possible way to handle @xmath2li and the other targets together was by using as a parameter an _ average _ density of the nucleus that turns out to cut - off @xmath2li from the rescaling of the imaginary potential . in this modelthe real part of the potential is rescaled by a constant whereas the imaginary part is multiplied by the expression @xmath58\ ] ] with @xmath20 the average density of the target nucleus and @xmath59 a cut - off density . with @xmath59 turning out to be 0.088@xmath540.005 @xmath60 , larger than the average density of the @xmath2li nucleus , it was possible to fit all 32 data points with only three parameters .the predicted angular distribution for elastic scattering of 715 mev / c k@xmath61 by c when using this prescription is indistinguishable from the solid curve in fig .[ fig : angdist ] .recent in - medium algorithm for calculating optical potentials from the respective meson - nucleon forward scattering amplitudes , used so far for kaonic atoms , pionic atoms , low energy pion scattering and @xmath39-nucleus interactions , was used in the present work to calculate integral cross sections for low energy @xmath0 mesons .this work was motivated by better penetrability of kaons into nuclei below 800 mev / c in comparison with other particles thus providing more sensitive tests of this approach .in particular , penetration beyond the nuclear surface into where the density is close to the full nuclear density means shifts of up to 30 mev in the argument of the in - medium amplitude which causes observable effects .it was possible to test separately the effects of the energy - dependence implied by the model and the importance of applying corrections due to pauli correlations .both were found to improve the agreement between experiment and calculations regarding the energy - dependence of total and reaction cross sections of @xmath0 on c , si and ca , particularly at the lower energy end of the range . for the low - density @xmath2li nucleus , calculation and experimentagree to better than 3% uniformly over the energy range studied when the variable energy of the model is applied .effects due to pauli correlations are negligibly small .compared with earlier analysis @xcite the full agreement achieved for @xmath2li is partly due to the inclusion in the present work of @xmath37-wave contributions to the forward scattering amplitude .the @xmath62 partial wave contributes attraction that is close to 30% of the repulsion due to the @xmath63 partial wave , particularly at the lower energies .the @xmath2li provides a solid basis for direct comparisons with the other three nuclear targets without a need to use so - called ` super - ratios ' , namely , normalizing experimental to calculated ratios for a target to the corresponding ratios for @xmath2li .the experimental cross sections for c , si and ca are found to be ( 23@xmath544)% larger than calculation , in general agreement with previous observations .this enhancement could be reproduced phenomenologically by rescaling the imaginary part of the potential by 40@xmath542% but quantitative understanding of this effect could not be achieved .it suggests that in - medium enhancements of @xmath0-nucleon interaction are due to more exotic mechanisms than traditional corrections .comparisons with predictions of such mechanisms are obviously called for .fruitfull discussions with a. gal are gratefully acknowledged .

<S> optical potentials for @xmath0-nucleus interactions are constructed from @xmath0-nucleon amplitudes using recently developed algorithm based on @xmath0-n kinematics in the nuclear medium . with the deep penetration of @xmath0 mesons into the nucleus at momenta below 800 mev / c it is possible to test this approach with greater sensitivity than hitherto done with @xmath1 and pions . </S><S> the energy - dependence of experimental reaction and total cross sections on nuclei is better reproduced with this approach compared to fixed - energy amplitudes . </S><S> the inclusion of pauli correlations in the medium also improves the agreement between calculation and experiment . </S><S> the absolute scale of the cross sections is reproduced very well for @xmath2li but for c , si and ca calculated cross sections are ( 23@xmath3)% smaller than experiment , in agreement with earlier analyses . </S><S> two phenomenological models that produce such missing strength suggest that the imaginary part of the potential needs about 40% enhancement . </S><S> @xmath0-nucleon amplitudes , @xmath0-nucleus optical potentials , @xmath0-nucleus reaction and total cross section , medium - dependent effects </S>

hep-ph9711461

the lightest neutralino is one of the most promising candidates for the dark matter in the universe . a linear combination of the superpartners of the neutral gauge and higgs bosons , it is believed to be the lightest stable supersymmetric particle in the minimal supersymmetric extension of the standard model ( mssm ) .in the near future , high precision measurements of the dark matter density may become possible from high resolution maps of the cosmic microwave background , and this may constrain supersymmetry .it is therefore of great interest to calculate the relic density of the lightest neutralino as accurately as possible . as a major step towards a complete and precise calculation valid for all neutralino masses and compositions , we included for the first time @xcite all concomitant annihilations ( coannihilations ) between neutralinos and charginos , properly treating thermal averaging in presence of thresholds and resonances in the annihilation cross sections .consider coannihilation of @xmath0 supersymmetric particles with masses @xmath1 and statistical weights @xmath2 ( first studied in ref .@xcite ) . normally , all heavy particles have time to decay into the lightest one , which we assume stable .its final abundance is then simply described by the sum of the densities @xmath3 when the scattering rate of supersymmetric particles off the thermal background is much faster than their annihilation rate , @xmath4 obeys the evolution equation @xcite @xmath5 with effective annihilation cross section @xmath6 the numerator @xmath7 is the total annihilation rate per unit volume at temperature @xmath8 , and @xmath9 is the total equilibrium density . under the assumption of boltzmann statistics ( a good approximation for @xmath10 ) , we obtain @xcite @xmath11 and @xmath12 here @xmath13 is the modified bessel function of the second kind of order @xmath14 , and @xmath15 where @xmath16 .@xmath17 is a lorentz invariant annihilation rate per unit volume in which coannihilations appear as thresholds at @xmath18 equal to the sum of the masses of the coannihilating particles .the independence of @xmath17 on temperature is a remarkable calculational advantage in presence of coannihilations : in fact it can be tabulated in advance , before taking the thermal average and solving the density evolution equation .( [ eq : sigmaveffdef]-[eq : wijcross ] ) generalize the result of gondolo and gelmini @xcite to coannihilations .to explore a significant fraction of the mssm parameter space @xcite , we keep the number of theoretical relations among the parameters to a minimum .we assume gut relations for gaugino masses , keep only the top and bottom trilinear soft supersymmetry - breaking parameters , and use a single mass parameter for the diagonal entries in the sfermion mass matrices at the weak scale .we perform many different scans in parameter space , some general and some specialized to interesting regions .we keep only models that satisfy the experimental constraints on the @xmath19 width , on the @xmath20 branching ratio , and on superpartner and higgs boson masses .we obtain analytic expressions for the many feynman diagrams contributing to the two - body cross sections at tree level for neutralino - neutralino , neutralino - chargino and chargino - chargino annihilation. then for each set of model parameters , we sum over particle polarizations and over initial and final states numerically and tabulate the annihilation rate @xmath17 .thermal averaging and integration of the density equation finally give the neutralino relic density @xmath21 in units of the critical density @xmath22 .[ fig : oh2vsmx ] shows the neutralino relic density @xmath23 versus the neutralino mass @xmath24 .each point represents a set of model parameters .it should be kept in mind that bands and holes in the point distributions are mere artifacts of our sampling in parameter space .the horizontal lines limit the cosmologically interesting region where the neutralino can constitute most of the dark matter in galaxies without violating the constraint on the age of the universe .we take it to be @xmath25 .the effect of coannihilations on the neutralino relic density is summarized in fig .[ fig : ratiovsmx ] , which shows the ratio of the neutralino relic density with and without coannihilations versus the neutralino composition @xmath26 ( here @xmath27 is the gaugino fraction ) .coannihilation processes are important not only for light higgsino - like neutralinos , as pointed out before in approximate calculations @xcite , but also for heavy higgsinos and for mixed and gaugino - like neutralinos .indeed , coannihilations should be included whenever @xmath28 , independently of the neutralino composition .when @xmath29 , coannihilations can increase or decrease the relic density in and out of the cosmologically interesting region .[ fig : cosmregion ] shows the cosmologically interesting region in the neutralino mass composition plane , before and after including coannihilations .this region is limited to the left by accelerator constraints , to the right by @xmath30 , and below and above by incompleteness in our survey of parameter space , except for the hole at 85450 gev where @xmath31 .the main effect of coannihilations is to shift the higgsino region to higher masses .in particular the cosmological upper bound on the neutralino mass changes from 3 to 7 tev . differently from previous approximate results @xcite , there is a cosmologically interesting window of light higgsino - like neutralinos with masses around 75 gev .

<S> we give an overview of our precise calculation of the relic density of the lightest neutralino , in which we included relativistic boltzmann averaging , subthreshold and resonant annihilations , and coannihilation processes with charginos and neutralinos . </S><S> * neutralino relic density including coannihilations * paolo gondolo _ max - planck - institut fr physik ( werner - heisenberg - institut ) _ _ fhringer ring 6 , 80805 mnchen , germany _ </S><S> gondolo@mppmu.mpg.de joakim edsj _ department of theoretical physics , uppsala university _ </S><S> _ p.o . </S><S> box 803 , se-751 08 uppsala , sweden _ </S><S> edsjo@teorfys.uu.se * abstract </S><S> * we give an overview of our precise calculation of the relic density of the lightest neutralino , in which we included relativistic boltzmann averaging , subthreshold and resonant annihilations , and coannihilation processes with charginos and neutralinos . </S><S> _ talk presented by paolo gondolo at `` topics in astroparticle and underground physics ( taup ) 97 , '' laboratori nazionali del gran sasso , italy , 711 september 1997 . </S>

astro-ph9906462

radio - emitting synchrotron jets are commonly observed in radio galaxies and quasars .in contrast , the optical counterparts of these structures are observed only rarely ( e.g. , odea 1999 ; scarpa & urry 1999 and references therein ) .the high spatial resolution of the _ hubble space telescope _ ( hst ) allows the study of known optical jets with detail comparable to vla radio maps , and has led to the discovery of new optical jets .hst is particularly useful for finding short jets that develop within few arcseconds from the luminous nucleus and/or are fully contained within a luminous host galaxy .the jet emission is typically confined to a narrow area , so at higher spatial resolution its surface brightness is higher .this facilitates detection closer to the nucleus , increasing the probability of detecting either shorter jets or jets more closely aligned with the line of sight .not surprisingly , the number of known jets has almost doubled since the introduction of hst ( e.g. , sparks 1995 ; ridgway & stockton 1997 ) .detection of optical emission and high resolution optical imaging of radio jets put important constraints on jet physics .for example , electrons radiating at @xmath0 hz in a typical equipartition magnetic field of @xmath1 g have a radiative cooling time of only @xmath2 years , yet optical jets are tens of thousands of light years long .( e.g. , pks 0521 - 365 , danziger 1979 ) .here we present new hst images of three relativistic jets which probe the shortest - lived radiating particles .the observations were obtained as part of our hst snapshot survey of bl lac objects ( urry et al .1999 , scarpa et al . 1999 ) and so were shorter than the ideal for studying synchrotron jets ; however , the large number of low redshift bl lac observed greatly increased our chance of detecting new jets .indeed , in pks 2201 + 044 , the 104@xmath3 of 109 objects observed , we discovered a new optical jet .we also imaged two well - known bl lacertae objects with previously known optical jets , pks 0521 - 365 and 3c371 , which are relatively nearby and can be studied in great detail with hst . throughout the paper were used .observations were carried out with the hubble space telescope in snapshot mode , using the wide field and planetary camera 2 ( wfpc2 ) through filters f702w ( for pks 0521 - 365 and pks 2201 + 044 ) and f555w ( for 3c371 ) .targets were centered on the pc chip , which has pixels @xmath4 arcsec wide . to obtain a final imagewell exposed both in the inner , bright nucleus and in the outer fainter regions , we set up a series of exposures with duration ranging from a few to hundreds of seconds , for total integration times of 305 , 302 , and 610 seconds for pks 0521 - 365 , 3c371 , and pks 2201 + 044 , respectively .after standard reduction ( flat field , dark and bias subtraction , and flux calibration carried out as part of the standard hst pipeline processing ) , images were combined using the task `` crrej '' available within iraf .this task stacks all images while removing cosmic ray events .final images were flux calibrated following the prescription of holtzman et al .( 1995 , their eq . 9 andtable 10 ) .although readily visible in the direct hst image , to be fully studied the jet must be separated from the contaminating light of the host galaxy and the central , luminous point source . we therefore first subtracted a suitably scaled point spread function ( psf ) model computed using tiny tim software ( krist 1995 ) , plus an estimate of large - angle scattered light ( scarpa 1999 ) . with the iraf task `` ellipse '' we then fitted the residual light with a two - dimensional elliptical galaxy model , convolved with the psf .extraneous features , including the jet itself , were masked before fitting the isophotes .( for details on host galaxies see scarpa 1999 and falomo 1999 . )the final jet image was obtained after subtracting both psf and galaxy models .this well - studied nearby object ( z=0.0554 ; danziger 1979 ) is one of the most remarkable extragalactic objects known , as it shows a variety of nuclear and extra - nuclear phenomena .first classified as an n galaxy , then as a bl lac , now it shows the strong narrow and broad emission lines typical of seyfert 1 galaxies ( ulrich 1981 ; scarpa , falomo & pian 1995 ) .most interesting , the source has a prominent radio and optical jet ( danziger 1979 ; keel 1986 ; macchetto 1991 ; falomo 1994 ) , which resembles that of the nearby radio galaxy m87 ( i.e. , sparks , biretta & macchetto 1994 ) . the optical jet as observed at hst resolution ( fig .1 ) has a structure characterized by a bright feature near the nucleus ( marked with an `` f '' ) , which is however too close to the bright central point source to be readily studied .it corresponds to the base of the jet visible in the uv image obtained with the hst faint object camera ( macchetto 1991 ) .farther out , a prominent bright spot ( called `` a '' following falomo 1994 ) , a series of smaller knots and a final bright spot ( `` c '' ) are also visible .structure `` b '' is resolved into at least three distinct knots aligned along the jet axis .structure `` c '' ( also called the `` red tip '' ) is @xmath5 from the nucleus and has integrated magnitude @xmath6 mag .it is clearly resolved even if its symmetric shape makes it appear point - like , and its radial profile is consistent with an elliptical galaxy .this confirms the finding of falomo ( 1994 ) , and together with the absence of radio emission ( keel 1986 ) and detectable optical polarization ( sparks , miley & macchetto 1990 ) , suggests the `` red tip '' is a background galaxy ( or small companion ) rather than part of the jet , in spite of the almost perfect alignment .finally , we note that in the optical there is no indication of either a counter - jet or the 2 cm bright hot spot reported by keel ( 1986 ) s - e of the nucleus , which was also not detected in deeper ground based observations ( falomo 1994 ) .we were able to measure the apparent magnitude of 5 bright knots . in table 1we report the apparent r magnitude , the distance of each feature from the central point source , and the jet width at these positions .the reported jet width is the average full width at zero intensity estimated averaging on strips 10 pixels wide .errors on the jet width are @xmath7 for all values .the total magnitude of the jet , excluding knots `` c '' and `` f '' , is m@xmath8 mag . the jet is resolved also orthogonal to the axis . in feature `` d '' two knotsare clearly visible , `` e '' is tilted with respect to the jet axis , and around `` b '' several secondary maxima of light are detected .indeed , compared with lower resolution ground - based images , in its terminal region the jet looks much more diffuse than previously reported . at the lowest surface brightnessreached in our observations , the jet has an almost constant width ( as is also the case in the deeper image obtained by falomo 1994 ) . even in knot `` a '' , where the jet is very bright , the full width at zero intensity is not significantly larger than in the other zones . only at the very end , at `` b1 '' , is the jet narrower .present data therefore suggest the plasma is moving in a well defined cylindrical funnel .the optical morphology of the jet agrees very well with what is observed at 2 cm ( keel 1986 ) .the general shape and the total length of the radio jet are almost exactly reproduced in the optical .this is particularly true for knot `` a '' , which is well resolved in both bands ( fig .the 2 cm flux was given by keel ( 1996 ) for the whole jet , knot `` a '' , and the jet minus `` a '' .for other three knots we estimated 2 cm flux directly from the radio map of keel ( 1986 ) .we then evaluated the radio - to - optical spectral index for four knots and the full jet ( table 1 ) .it turns out that is relatively stable along the jet , indeed all values are consistent within the errors .this is simply a quantitative confirmation of the general one - to - one morphological correspondence between the two bands .assuming equipartition between particles and magnetic field , the internal magnetic field can be estimated knowing the total flux and volume of a radio source . in our case , the size of the emitting knots can be directly measured on the image and transformed into volume assuming symmetry .we report in table 1 the volume @xmath9 of the whole jet and of each knot .the jet is assumed to be a cylinder of diameter 1.5 arcsec and length 5 arcsec .for the knots we assumed spherical symmetry , reporting in table 1 the volume of the sphere which has projected surface equal to the knot surface .this was done because the jet is resolved orthogonal to its axis , with the knots occupying only a fraction of the jet width .the complete radio - to - optical flux energy distribution of the synchrotron emission from the jet alone is not known .we therefore assume the synchrotron emission from jet and knots is described by power laws , with spectral indices as reported in table 1 , extending from @xmath10 to @xmath11 hz .the upper frequency limit is set by the hst foc data at 3200 ( macchetto 1991 ) and has basically no effect on the estimated equipartition magnetic field . with these assumptions , the jet has almost constant @xmath12 g along its length ( table 1 ) .magnetic fields of this intensity are commonly observed in radio sources , and our values are fully consistent with the value of @xmath13 g reported by keel ( 1986 ) .this extragalactic radio source has been the subject of intense study at many frequencies .it was first classified as an optically violently variable quasar ( angel & stockman 1980 ) and then as a radio - selected bl lacertae object ( giommi 1990 ) .the radio morphology is dominated by a one - sided jet emanating from the central core and extending up to 25 arcsec from the nucleus , ending at one of the two radio lobes ( wrobel & lind 1990 ) .this morphology is characteristic of a fanaroff - riley type ii radio source . in the optical band 3c371is hosted by a large elliptical galaxy ( ugc 11130 , condon & broderick 1988 ) at @xmath14 ( sandage 1973 ) , surrounded by a small cluster of galaxies ( stickel , fried & khr 1993 ) .recently the optical counterpart of the radio jet was discovered by nilsson ( 1997 ) . as observed from the ground, the optical jet extends only a few arcsec from the nucleus , ending in a bright knot where the radio jet apparently changes direction . at hst resolution ,the jet is fully resolved also perpendicular to the jet axis .at least three bright knots are clearly visible ( fig .3 ) . near the center, the jet starts with a bright structure ( `` c '' ) , followed by a much smaller knot ( `` b '' ) which is elongated orthogonal to the jet .further out is the broadest structure , knot `` a '' , @xmath15 wide . in our hst image , the optical jet appears to extend beyond `` a '' ; the narrow structure named `` d '' in fig .3 coincides with a faint feature visible in the nilsson ( 1997 ) image .the beginning of feature `` d '' is possibly also visible in the merlin map of akujor ( 1994 ) . comparing optical and radio emission a general one - to - one matching of all knots is found . the observed radio jet ( wrobel & lind 1990 ) is several times longer than the optical jet . compared to pks 0521 - 365 and m87 , this is the most obvious difference .it must be noted that the radio flux fades significantly after knot `` a '' .wrobel & lind ( 1990 ) report a peak flux of 3.8 mjy from the final hot spot of the radio jet .assuming @xmath16 ( table 1 ) the expected flux in v - band is @xmath17 mjy , roughly 1 magnitude fainter than the limit of our observation .hence , present data are consistent with a constant @xmath18 and general correspondence of radio and optical emission , and deeper images may well be able to detect optical emission up to the very end of the radio jet .on the contrary , if the optical jet actually ends at `` a '' , it would be even more interesting , offering an important benchmark for theoretical models to explain the propagation of electrons and magnetic field through the interstellar medium .deeper optical observations are therefore highly desirable . under the same set of assumptions adopted for pks 0521 - 365 about synchrotron power law spectrum and emission volume , we estimate the physical properties of the jet . in table 1 the v magnitude and volume for the whole jet and main knots are reported .knot `` c '' is too close to the central point source so we do not attempt to measure its luminosity . to estimate the volume of the jet we modeled it as a cylinder with diameter of 1 and length @xmath19 arcsec , while spherical symmetry was assumed in computing the volume of knots .the radio flux at 1.6 ghz ( akujor 1994 ) is also quoted .the same value of the radio - to - optical spectral index is found both for the whole jet and knot `` a '' . considering the h- and b - band data from nilsson ( 1997 ) together with our v - band data , the near - ir - optical spectral index of knot `` a '' is @xmath20 .this is fully consistent with , indicating no steepening of the radiation spectrum at high frequencies .the equipartition magnetic field is basically constant along the jet and has average intensity @xmath21 g , similar to what was found for pks 0521 - 365 .this source is a nearby ( @xmath22 ) bl lac object ( angel & stockman 1980 ) hosted in a prominent elliptical galaxy ( falomo 1996 ; scarpa 1999 ) .the optical spectrum shows emission and absorption lines and is similar to seyfert 1 spectra ( veron - cetty & veron 1993 ; falomo , scarpa & bersanelli 1994 ) .the associated radio source has a total extent of 5.6 arcmin ( ulvestad & johnson 1984 ) , and is core dominated ( preston 1985 ) .the radio properties of pks 2201 + 044 have been throughly investigated by laurent - muehleisen ( 1993 ) , who found a core - jet morphology , with the jet extending for more than 10 kpc .a bright radio hot spot was also found @xmath23 kpc ( 2 arcsec ) from the nucleus . at hst resolutionthe optical source is fully resolved into nucleus , host galaxy , and jet .thanks to the superior spatial resolution , some structures of the jet are directly visible above the strong background of the host galaxy . in figure 4we show the residuals after subtraction of the best fit galaxy model .the jet structure appears at position angle @xmath24 , with the most prominent knot @xmath25 ( 1.57 kpc ) from the nucleus . comparing our data to the 5ghz radio map of laurent - muehleisen ( 1993 ), a clear correspondence of this bright optical knot with the brightest radio knot is found , strongly supporting the association of the optical structure with the synchrotron radio jet ( fig .unfortunately , the resolution of the radio map is insufficient to resolve further structures .the brightest optical knot is clearly resolved both along and perpendicular to the jet axis , and has quite a complicated structure , extending over a region of @xmath26 arcsec@xmath27 . between the knot and the nucleus some weaker knotsmay also be present .the integrated magnitude of knot `` a '' is @xmath28 mag ( @xmath29 mjy ) .its radio flux is @xmath30 mjy at 5ghz , corresponding to radio - to - optical spectral index @xmath31 , fully consistent with the values found in others optical jets ( crane 1993 ; scarpa & urry 1999 ) . assuming the knot is a sphere of radius 0.4 arcsec , and adopting the same set of hypotheses as for the previous two jets , the equipartition magnetic field is @xmath32 g , again similar to what is found in others jets .we have described hst observations of three synchrotron jets detected at optical wavelengths in three bl lac objects . in two cases ,pks 0521 - 365 and 3c371 , the jet is fully resolved and several knots can be studied separately . in spite of the clumpiness of the emission , both jets are rather homogeneous , with basically constant values of @xmath33 and equipartition magnetic field along the length of the jet . for knot `` a '' in 3c371 , a similar spectral index is also found at near - ir - optical frequencies , indicating no significant aging of the radiating electrons .we also reported the discovery of an optical counterpart of the radio jet of pks 2201 + 044 .the brightest knot , at projected distance @xmath34 kpc from the nucleus , is fully resolved in the optical .it corresponds precisely to the brightest radio knot , confirming its association with the synchrotron jet .estimates of @xmath31 and equipartition magnetic field @xmath32 g are similar to values generally found in other jets . however , to fully understand the morphology of this jet , deeper optical images and higher resolution radio maps are needed .we note that all three sources exhibit strong emission lines in their optical spectrum , and in pks 2201 + 044 and pks 0521 - 365 broad emission lines are also observed ( ulrich 1981 ; veron - cetty & veron 1993 ; scarpa , falomo & pian 1995 ) .this may suggest that in these objects the beaming is modest , and in fact there are indications of this in the case of pks 0521 - 365 ( pian 1996 ) .a modest beaming may be consistent with an easier observability of the optical jet .however our sample is obviously too restricted to elaborate further . at present , including the newly discovered jet in pks 2201 + 044 , a total of 14 optical jets are known .this number is not big , but it starts to be large enough to enable decent statistical consideration .it is well known that in all these jets the inferred electron lifetimes are much smaller than needed to explain the jet length , so that reacceleration along the jet seems unavoidable .this is also true for the jet in pks 2201 + 044 .however , the observed constancy of @xmath18 is difficult to explain if electrons are reaccelerated by first - order fermi mechanism at strong shocks ( heinz & begelman 1997 , and references therein ) , so that present observations add difficulties to the already hard problem of finding a plausible physical mechanism for electron reacceleration .the possibility of electrons being transported in a loss - free channel inside the jet ( owen , hardee & cornwell 1989 ) is difficul to accept , because hst observation of m87 have shown the optical emission is centered near the axis of the jet rather than on the edge ( spark , biretta & macchetto 1996 ) , which appear to be the case also for pks 0521 - 365 .it is possible that the presence of relativistic motion on kiloparsec scales is common , as supported by direct observations with hst of superluminal proper motion in m87 ( biretta 1999 ) .if this is generally the case , then relativistic beaming of the emission is unavoidable , implying that both the intrinsic luminosity of the jet and the equipartition magnetic field are smaller than what we estimated , while electron lifetimes are longer and the necessity for reacceleration is weakened .an alternative hypothesis is to renounce the basic assumption of equipartition of energy between relativistic particles and magnetic field ; the electron lifetimes are longer by approximately the factor @xmath35 . using a statistical approach we address general considerations of jets power , beaming , and physical conditions in a companion paper ( scarpa & urry 1999 ) .it is a pleasure to thanks g. ghisellini and f. macchetto for helpful and encouraging comments .we also thanks the referee w. c. keel for helpfull comments and for providing its radio map of pks 0521 - 365 in fits format .support for this work was provided by nasa through grant go-06363.01 - 95a from the space telescope science institute , which is operated by aura , inc ., under nasa contract nas 5 - 26555 , and by the italian ministry for university and research ( murst ) under grant cofin98 - 02 - 32 .akujor c.e . ,ldke e. , browne i.w.a ., leahy j.p . , garrington s.t . , jackson n. & thomasson p. 1994a&as 105 , 247 angel j.r.p .& stockman h.s .1980 ara&a 18 , 321 biretta j.a . ,perlman e. , sparks w.b . & macchetto f. 1999 , aas meeting n. 193 , 07.09 crane p. , peletier r. , baxter d. 1993 , apj 402 , l37 condon j.j . & broderick j.j .1988 aj 96 , 30 danziger i.j ., fosbury r.a.e ., goss w.m . & ekers r.d .1979 mnras 188 , 415 falomo r. 1994 the messenger 77 , 49 falomo r. 1996 mnras 283 , 241 falomo r. , scarpa r. & bersanelli m. 1994 , apjs , 93 , 125 falomo r. , urry c.m . , scarpa r , pesce j.e ., giavalisco m. & treves a. 1999 , in preparation giommi p. , barr p. , pollock a.m.t ., garilli b. & maccagni d. 1990 apj 356 , 432 heinz s & begelman m. c. 1997 apj 490 , 653 holtzman j.a ., burrows c.j . , casertano s. , hester j.j . , trauger j.t ., watson a.l . & worthey g. 1995 pasp 107 , 1065 keel w.c .1986 apj 302 , 296 krist , j. 1995 , in _ astronomical data analysis , software and systems iv _ , eds r. shaw et al . , ( san francisco : astr .soc . pac . ) , p. 349laurent - muehleisen s.a . ,kollgaard r.i ., moellenbrock g.a . &feigelson e.d .1993 , aj 106 , 875 macchetto f. , albrecht r. , barbieri c. , 1991 apj 369 , l55 nilsson k. , heidt j. , pursimo t. , sillanp a. , takalo l.o . & jger 1997 apj 484 , l107odea c. p. , de vries w. , biretta j. a. & baum s. a. 1999 , aj in press owen f.n ., hardee p.e . & cornwell t.j .1989 , apj 340 , 698 pian e. , falomo r. , ghisellini g. , maraschi l. , sambruna r. m. , scarpa r. & treves a. 1996 , apj 459 , 169 preston r.a . ,morabito d.d ., williams j.g . , faulkner j. , jauncey d.l . &nicolson g.d .1985 , aj 90 , 1599 ridgway s.e . &stockton a. 1997 , aj 114 , 511 sandage a. 1973 apj 180 , 687 scarpa r. , falomo r. & pian e. 1995 a&a 303 , 730 scarpa r. & urry c. m. 1999 in preparation scarpa r. , urry c.m . ,falomo r , pesce j.e ., giavalisco m. & treves a. 1999 , in preparation sparks w.b ., biretta j.a . & macchetto f. 1994 apjs 90 , 909 sparks w.b ., biretta j.a . & macchetto f. 1996 apj 473 , 254 sparks w.b . ,golombek d. , baum s.a .1995 , apj 450 , l55 sparks w.b ., miley g.k . &macchetto f. 1990 apj 361 , l41 stickel m. , fried j.w .& khr h. 1993 a&as 98 , 393 ulrich m.e .1981 , a&a 103 , l1 ulvestad j.s .& johnston k.j .1984 , aj 89 , 189 urry c.m . , falomo r , scarpa r. , pesce j.e . , giavalisco m. & treves a. 1999 , in preparation veron - cetty m.p . &veron p. 1993 , a&as , 100 , 521 wrobel j.m .& lind k.r .1990 apj 348 , 135 lcccccccc jet total & r&@xmath36&@xmath37&@xmath37 & 106 & @xmath38&@xmath39&@xmath40 jet - a & & @xmath41&@xmath37&@xmath37 & 46 & @xmath42&@xmath43&@xmath44a & & @xmath45&@xmath46&@xmath47 & 60 & @xmath48&@xmath49&@xmath50b1+b2+b3 & & @xmath51&@xmath52&@xmath53 & ( 19 ) & @xmath54&@xmath55&@xmath56)d & & @xmath57&@xmath58&@xmath47 & ( 18 ) & @xmath54&@xmath47&@xmath59)e & & @xmath60&@xmath61&@xmath62 & ( 9 ) & @xmath54&@xmath63&@xmath59)jet total & v&@xmath64&@xmath37&@xmath37&@xmath65&@xmath66&@xmath67&@xmath68 a & & @xmath51 & 3.13 & @xmath69&@xmath70&@xmath71&@xmath72&@xmath73b & & @xmath74 & 1.68 & @xmath75&@xmath76&@xmath77&@xmath78&@xmath79c & & @xmath80 & 0.46 & @xmath37&@xmath81&@xmath82&@xmath83&@xmath84d & & @xmath85 & 4.52 & @xmath37&@xmath81&@xmath82&@xmath83&@xmath84a & r&@xmath86 & 2.12 & @xmath87&@xmath88&@xmath89&@xmath90&@xmath91 * fig . 1 * * upper panel : * the f702w image of the optical jet of pks 0521 - 365 , obtained after subtracting the central point source and a galaxy model .to improve the visibility of jet features , a gaussian filter with @xmath92 pixels was applied , reducing the resolution to 0.14 arcsec .* lower panel : * contour plot of the above image , with knots labeled . north and east are as indicated .the big cross at the upper right indicates the position of the central point source .note that the diffuse emission detected south - west of the point source was never reported before and can actually be due to imperfect subtraction of the host galaxy .isophotes are 0.25 magnitudes apart , starting from @xmath93 mag / arcsec@xmath27 . *fig . 3 : * * upper panel : * the f555w gray - scale image of the optical jet of 3c371 , obtained after subtracting the central point source and a galaxy model , smoothed with a gaussian filter with @xmath92 pixels , reducing the resolution down to 0.14 arcsec .over - plotted on the optical image is a contour plot of the 1.6ghz radio map by akujor ( 1994 ) . * lower panel : * contour plot of the above optical image . north and east are as indicated .the big cross indicates the position of the central point source .isophotes are 0.25 magnitudes apart , starting from @xmath95 mag / arcsec@xmath27 . *fig . 4 : * * upper panel : * the f702w image of the optical jet of pks 2201 + 044 .in this source the central point source is not very bright , so here we show the image after subtraction of the galaxy only .the point source ( nucleus ) is characterized by the 4 diffraction spikes , not to be confused with the jet . to improve the visibility of faint structures , a gaussian filter with @xmath96 pixel was applied . * lower panel : * contour plot of the above image . north and east are as indicated .knot a is @xmath25 from the nucleus .two more knots most probably associated with the jet are also marked .isophotes are 0.5 magnitudes apart , starting from @xmath97 mag / arcsec@xmath27 . *fig . 5 : * the wfpc2 f702w gray - scale image of the jet of pks 2201 + 044 , over - plotted with contours of the 5ghz radio map from laurent - muehleisen ( 1993 ) , which has lower resolution than the hst data .the hst image has been rotated so that north is up and east on the left .

<S> hst observations have led to the discovery of the optical counterpart of the radio jet of pks 2201 + 044 , and to a detailed analysis of the optical jets of pks 0521 - 365 and 3c371 . at hst spatial resolution these jets are well resolved , displaying knotty morphologies . when compared with radio maps of appropriate resolution , a clear one - to - one correspondence between optical and radio structures </S><S> is found , showing that all detected optical structures are indeed related to the radio synchrotron emission . </S><S> photometry of the brightest knots shows that the radio - to - optical spectral index and the derived intensity of the equipartition magnetic field are approximately constant along the jet . </S><S> thus , present observations suggest that the electron energy distribution does not change significantly all along the jet . </S><S> _ subject headings _ : galaxies : jets galaxies : individual ( pks 0521 - 365 , 3c 371 , pks 2201 + 044 ) bl lacertae objects </S>

0901.2834

in recent years we have used the gmrt ( e.g. @xcite ) to make relatively deep , wide area observations of several fields at 610 mhz , with a resolution of about 5 arcsec .the deepest of these fields is the spitzer ` extragalactic first look survey ' ( xfls ) region , which is about 4 square degrees in extent , and is in a direction with very little galactic infra - red foreground emission .this field already has deep vla observations at 1.4 ghz available with a resolution of about 5 arcsec ( see @xcite ) , for which the gmrt provides complementary 610-mhz observations at a similar resolution .the other fields are selected areas from the ` swire ' ( * s*pitzer * w*ide - area * i*nfra*r*ed * e*xtragalactic ) survey ( @xcite ) .these areas have deep infra - red observations from the spitzer space telescope , and often other deep complementary optical observations , but do not have deep , wide - field radio observations available .table [ tab : surveys ] gives a summary of the xfls and swire fields which were observed .given the primary beam of the gmrt at 610 mhz is @xmath2 arcmin , each pointing covers @xmath3 degree@xmath4 . for the elais - n1 field , four of the nineteen pointings ( which overlap regions where deeper optical observations are available ; @xcite )were observed more deeply , to provide a noise value of @xmath5 @xmath0jy beam@xmath1 , rather than the @xmath6 @xmath0jy beam@xmath1 of the other pointings in this field .also , observations of an additional 26 outer pointings in the lockman hole field have been obtained in order to cover the whole @xmath7 degree@xmath4 observed with spitzer and analysis of these is ongoing .all of the observations were made with two 16-mhz sidebands , in both right and left circular polarised emission .each sideband was divided into 128 spectral channels , so that any narrow band interference could be identified and removed .lcccl field & pointings & noise & sources & reference + & & ( @xmath0jy beam@xmath1 ) & & + xfls & 7 & 27 & 3944 & @xcite + elais - n1 & 19 & 40/70 & 2500 & @xcite + lockman hole & 12 & 60 & 2845 & @xcite + elais - n2 & 13 & 80 & @xmath81500 & in preparation +the analysis of these observations was performed using classic aips .initial careful editing of the data was necessary in order to remove interference , and other bad data ( e.g. a few systematically poor baselines , due to correlator hardware problems ) .bright ` primary ' calibrators ( e.g. 3c286 and/or 3c48 ) were observed at the start / end of each observation run , in order to define the flux density scale , and also derive antenna - based bandpass corrections .the bandpass corrections were applied , and several central channels in each sideband were averaged together , in order to make a ` pseudo - continuum ' channel ( ` channel 0 ' in vla notation ) .the antenna - based amplitude and phase calibrations of the telescope were derived using the pseudo - continuum channel , from the observations of nearby , compact ` secondary ' calibrators , which were observed every 30 min or so .the antenna - based amplitude / phase and bandpass corrections were then applied to the observed visibilities , and 10 channels were averaged together , in order to reduce the size of the @xmath9 datasets before imaging .the wide field - of - view of the gmrt means that multiple small ` facets ' need to imaged , and then combined together .generally 19 facets were used ( i.e. a central facet , 6 in a surrounding ring , and 12 in a larger ring on a hexagonal grid ) , to cover the full observed field .the quality of the synthesised images were improved by several cycles of self - calibration , at 10 , 3 and 1 min for phase only , and finally 10 min for amplitude and phase .this lead to images with dynamic ranges peak to r.m.s . away from bright sources of several thousand to one , with r.m.s .noise values away from bright sources that are close to the expected thermal noise . near bright sources, however , the quality of the images is limited , which is believed to be due to variations in the pointing of the telescope antennas ( see below ) .the early analyses of the overlapping pointings from the xfls field revealed two problems : ( 1 ) the time stamps of @xmath9-data were slightly wrong ( by about 7 s ) , meaning that the @xmath9 coordinates were also slightly wrong , leading to a distortion ( nearly a rotation ) in the synthesised images ; ( 2 ) comparison of the flux densities of compact sources in overlap regions between adjacent pointings showed systematic differences , which are thought to be due to antenna pointing offsets during the observations , particularly at low elevations . for the former problem , an aips task was written to correct affected data , by adjusting the time stamps and recalculating the @xmath9 coordinates ( and the real - time gmrt system was corrected in the summer of 2006 ) .the latter problem was correctly empirically , using a slightly shifted effective primary beam position .an improved pointing model is currently being implemented at the gmrt to address this problem .further details of the analysis procedures used are given in @xcite and @xcite .users of classic aips should also note that there were problems with the tasks split and splat which might result in incorrect frequency headers when averaging spectral channels together .these tasks were patched late in 2007 .given the wide field of the 610-mhz images produced by these observations , it is not possible to show these in full ( e.g. the xfls field is @xmath10 degree across , so with 1.5 arcsec pixels to well sample the @xmath11 arcsec beam , the image is about 5000 pixels in each dimension ) . figure [ fig : xfls ] shows a small portion only about 3 per cent of the deep gmrt image of the xfls field .the gmrt results for the xfls field are comparable in resolution , and in depth for synchrotron emission , to the available deep vla 1.4-ghz survey of the same region : the gmrt observations , which took about 40 hours of observations , have an r.m.s .noise ( but only away from the bright sources ) of @xmath15 @xmath0jy beam@xmath1 ; the vla observations , which took about 200 hours of observations , have an r.m.s .noise of @xmath16 @xmath0jy beam@xmath1 .these deep , wide - field surveys provide catalogues of radio sources , which then allow a variety of astrophysical studies to be made , including comparisons with the deep infra - red and other wavelength observations available for these fields . in particular , for the xfls region ,the existence of both the deep vla 1.4-ghz and gmrt 610-mhz observations allow the properties of many sub - mjy radio sources which are dominated by star forming galaxies , rather than agn to be studied , including their radio spectral indices .some of these studies are described briefly below . in the xfls survey regionit is possible to define a sample of 235 sources which are detected at both 610 and 1410 mhz in the radio and at 24 and 70 @xmath0 m in the infra - red , for which spectroscopic redshifts ( up to @xmath17 ) are available . from these data it is possible to study the infra - red / radio correlation ,applying ` @xmath18-corrections ' to provide rest - frame luminosities at the same frequencies ( using the observed spectral indices in the radio , and model seds in the infra - red ) , see @xcite .any cosmic evolution of the magnetic fields in galaxies would be expected to be reflected in changes in the comparative radio brightness of star - forming galaxies .the available data shows no evidence for such an effect , suggesting that there has been little evolution in the magnetic fields of galaxies since @xmath19 .although the gmrt observations provide deep images , it is possible to probe , statistically , the properties of sources that are too faint to be detected individually , using a median ` stacking ' technique ( see @xcite ) .for example , stacking 591 sources from the xfls field with spitzer 24-@xmath0 m flux densities between 150 201 @xmath0jy leads to a median detected radio flux of @xmath22 @xmath0jy at 610 mhz , see figure [ fig : stack ] . using this technique , the radio properties of sources with 24-@xmath0 m flux densities between 150 @xmath0jy and 2 mjy show no evidence for a change in the infra - red / radio correlation .almost all of the radio sources detected in these gmrt surveys are also detected in the infra - red due to the sensitivity of the spitzer surveys , as is expected for local agn or star - forming galaxies . however , there are a small number of relatively bright radio sources that are not detected in the infra - red @xcite . two examples of such sources are shown in figure [ fig : irfs ]. there are 14 of these infrared - faint radio sources in the xfls field , 8 of which are detected near to the limiting magnitude ( 24.5 ) of a deep r - band image of the region .these sources can modelled as compact ( @xmath24 kpc ) fanaroff riley type ii radio galaxies located at high redshift ( @xmath25 ) .these surveys have also been used to construct differential source counts at 610 mhz between 270 @xmath0jy and 200 mjy ( @xcite ) .a flattening in the differential source counts is clearly visible below 2 mjy , and a three - component population containing steep and flat spectrum agn , and starburst galaxies which undergo pure luminosity evolution is sufficient to model simultaneously both the 610-mhz source counts , and 1.4-ghz source counts from the literature .garn , t. , ford , d. , alexander , p. , green , d. a. , & riley , j. m. 2007a , in the modern radio universe : from planets to dark energy , proceedings of science , ( http://pos.sissa.it/contribution?id=pos(mru)073 )

<S> the gmrt has been used to make deep , wide - field surveys of several fields at 610 mhz , with a resolution of about 5 arcsec . </S><S> these include the spitzer extragalactic first look survey field , where 4 square degrees were observed with a r.m.s . </S><S> sensitivity of about 30 @xmath0jy beam@xmath1 , and several swire fields ( namely the lockman hole , elais - n1 and n2 fields ) covering more than 20 square degrees with a sensitivity of about 80 @xmath0jy beam@xmath1 or better . the analysis of these observations , and some of the science results are described . </S>

astro-ph9801153

the extreme ultraviolet ( euv ) surveys of the rosat wide field camera ( wfc , pounds et al . 1993 ) and the extreme ultraviolet explorer ( euve , bowyer et al .1994 ) have found a substantial number of white dwarfs , in excess of 120 .most of these stars are isolated , but over 30 are now known to lie in binary systems ( burleigh 1997 ) . in particular ,nearly 20 unresolved pairs consisting of a hot white dwarf and a bright , normal star ( spectral type k or earlier ) have been found ( e.g. barstow et al .1994 , vennes et al .1995 , burleigh et al .1997 , burleigh 1998 ) .prior to the two euv surveys , these systems were all but unidentifiable , since the normal stellar companion completely swamps the optical flux coming from the white dwarf . in each case, however , the detection of euv radiation with the spectral signature of a hot white dwarf gave a clue to the existence of the previously invisible , faint degenerate companion .far - ultraviolet spectra taken with the international ultraviolet explorer satellite ( iue ) were then used to confirm the identifications .this technique has proved excellent for finding these systems in all cases where the normal star is of spectral type @xmath5a5 or later .in fact , the earliest type star so far identified by this method to have an unresolved hot white dwarf companion is beta crateris ( a2iv@xmath0wd , fleming et al .indeed , it took a very careful , detailed , analysis by barstow et al .( 1994 ) to finally confirm this discovery . unfortunately , even at far - uv wavelengths , stars of spectral types early a , b or o will still completely dominate any emission from smaller , fainter companions , rendering them invisible even to iue or hst .overall , the earliest type star known with a white dwarf companion remains sirius ( a1v@xmath0da ) .the spectral type of the normal star in these binaries gives a lower limit to the mass of the white dwarf progenitor .the value of the maximum mass feasible for producing a white dwarf , and the form of the initial - final mass relationship ( imfr ) , are long - standing astrophysical problems ( e.g. weidemann 1977 ) .weidemann ( 1987 ) gives the upper limit as 8m@xmath2 in his semi - empirical ifmr .recent observations of four white dwarfs in the young open cluster ngc 2516 ( jeffries 1997 ) , however , imply that the upper mass limit for white dwarf progenitors is only 5@xmath66m@xmath2 .this value is actually in agreement with current stellar evolutionary models which include moderate core overshoot , but , clearly , any observations which can place limits on the maximium white dwarf progenitor mass have important implications for our theories and models of stellar evolution , the birth rate of neutron stars and the predicted rates of type ii galactic supernovae . [cols="<,<,^,^,^,^,^,^,^,^ " , ] firstly , we assume that the white dwarf has a pure - hydrogen atmosphere .this is a reasonable assumption to make , since barstow et al .( 1993 ) first showed that for @xmath7@xmath840,000k hot white dwarfs have essentially pure - h atmospheres .we can then fit a range of models , each fixed at a value of the surface gravity log @xmath9 . before we can do this , however , we need to know the normalisation parameter of each model , which is equivalent to ( radius@xmath10/distance)@xmath11. we can use the _ hipparcos _ parallax ( 5.86@xmath12 milli - arcsecs ., esa 1997 ) to calculate the distance to the system ( 170@xmath13 parsecs ) , and the hamada - salpeter zero - temperature mass - radius relation to give us the radius of the white dwarf corresponding to each value of the surface gravity , since we have no _ a priori _ knowledge of the star s temperature ( see table 2 ) .we can also reduce the number of unknown free parameters in the ism model . from euvespectroscopy , barstow et al .( 1997 ) measured the line - of - sight interstellar column densities of hi , hei and heii to a number of hot white dwarfs .they found that the mean h ionisation fraction in the local ism was 0.35@xmath140.1 , and the mean he ionisation fraction was 0.27@xmath140.04 . from these estimates , and assuming a cosmic h / he abundance , we calculate the ratio @xmath15/@xmath16 in the local ism@xmath178.9 , and @xmath16/@xmath18@xmath172.7 .we can then fix these column density ratios in our model , leaving us with just two free parameters - temperature and the hi column density .the model fits at a range of surface gravities from log @xmath197.5@xmath69.0 are summarized in table 3 .we have discovered an unresolved hot white dwarf companion to the 5th magnitude b star hr2875 ( y pup ) .this is the first time a hot white dwarf @xmath0 b star binary has been detected , and it has important implications for our understanding of white dwarf and stellar evolution , since a white dwarf companion to such an early - type star must have evolved from a very massive progenitor , close to the upper limit for white dwarf formation . according to lang ( 1992 ) a b5v star has a mass of 5.9m@xmath2 ( or 6.5m@xmath2 according to allen , 1973 ) , and if the spectral type is as early as b3v ( as classified in the michigan catalog of hd stars , houk 1982 ) then it will of course have a slightly higher mass .this is @xmath15@xmath66m@xmath2 upper mass limit for white dwarf progenitors of jeffries ( 1997 ) , but still significantly less than the 8m@xmath2 upper limit from weidemann ( 1987 ) .since it must have evolved from such a massive progenitor , it is likely that this white dwarf also has a higher mass than the mean for these degenerates ( 0.583@xmath140.078m@xmath2 , marsh et al .we can use the theoretical initial - final mass relation between main sequence stars and white dwarfs of wood ( 1992 ) , to calculate the mass of the white dwarf if its progenitor was only slightly more massive than hr2875 : @xmath20@xmath17aexp(b@xmath21@xmath22 ) where a@xmath170.49462m@xmath2 and b@xmath170.09468m@xmath2@xmath23 . for @xmath22@xmath176.5m@xmath2 ,we find @xmath20@xmath170.91m@xmath2 .this would suggest the surface gravity of the white dwarf log @xmath9@xmath248.5 . in the log @xmath198.5 model , @xmath2543,400k andthe white dwarf has a v magnitude @xmath1716.4 ( calculated from the model flux at 5500 ) .hr2875 is not known to display radial velocity variations , and _ hipparcos _ found no evidence for micro - variations in its proper motion across the sky to suggest that this might be a relatively short period binary system ( p@xmath8few years , esa 1997 ) .however , it is clearly important to study this system further , and if radial velocity variations are detected then the binary parameters and the white dwarf s mass can be constrained , with important implications for our knowledge of the initial - final mass relation .additionally , further study of the b star primary might reveal evidence for past interaction .could the over - abundance of si detected in this object by hiltner et al .( 1969 ) be due to accretion from the wind of the evolved giant progenitor to the white dwarf , as in the wd@xmath0k2v binary re j0357@xmath028 ( jeffries , burleigh and robb 1996 ) ?are there more early - type star@xmath0hot white dwarf binaries in the rosat and euve catalogues awaiting discovery ? from its rosat wfc euv and pspc soft x - ray count rates and colours ( s1@xmath1752@xmath147 counts / ksec , s2@xmath17148@xmath1412 counts / ksec , pspc@xmath17124@xmath1424 counts / ksec - all in the lower band ) , we strongly suspect that the 4th magnitude star @xmath4 hya ( b9.5v , @xmath17hr3665 , @xmath17re j0914@xmath002 ) also hides a hot white dwarf companion , and indeed euve is scheduled to observe this star spectroscopically in march 1998 .other b stars detected in the euv surveys may also have non - interacting degenerate companions , e.g. ads10129c ( b9.5v , @xmath17hd150100 , @xmath17re j1636@xmath052 ) which also has euv and soft x - ray count rates very similar to known hot white dwarfs ( rosat wfc s1@xmath1712@xmath144 counts / ksec , s2@xmath1746@xmath1411 counts / ksec , pspc@xmath1772@xmath1415 counts / ksec - again , all in the lower band ) .unfortunately these objects are probably too faint to be detected by the euve spectrometers .matt burleigh and martin barstow acknowledge the support of pparc , uk .we thank detlev koester ( kiel ) for the use of his white dwarf model atmosphere grids . _ note added in proof _after this paper had been submitted for publication , we learned of a similar study by vennes et al .( 1997 , apj , 491 , l85 ) .their conclusions about the properties of the white dwarf are in agreement with ours .

<S> we report the discovery , in an extreme ultraviolet explorer ( euve ) short wavelength spectrum , of an unresolved hot white dwarf companion to the 5th - magnitude b5vp star hr2875 . </S><S> this is the first time that a non - interacting white dwarf@xmath0 b star binary has been discovered ; previously , the the earliest type star known with a white dwarf companion was sirius ( a1v ) . </S><S> since the white dwarf must have evolved from a main sequence progenitor with a mass greater than that of a b5v star ( @xmath16.0m@xmath2 ) , this places a lower limit on the maximum mass for white dwarf progenitors , with important implications for our knowledge of the initial - final mass relation . assuming a pure - hydrogen atmospheric composition , we constrain the temperature of the white dwarf to be between 39,000k and 49,000k . we also argue that this degenerate star is likely to have a mass significantly greater than the mean mass for white dwarf stars ( @xmath30.55m@xmath2 ) . finally , we suggest that other bright b stars ( e.g.@xmath4 hya ) detected in the extreme ultraviolet surveys of the rosat wide field camera and euve may also be hiding hot white dwarf companions . </S><S> -1_td-1 _ stars : binaries stars : white dwarfs stars : early - type stars : individual : hr2875 x - ray : stars . </S>

0711.1704

ammonia ( nh@xmath3 ) is an extremely useful molecular tool for studying the interstellar medium .many inversion transitions occur in the easily observable 12 mm band .they include transitions from both metastable ( j = k ) and non - metastable ( j@xmath4k ) levels of ammonia .hyperfine structure in the metastable transitions is commonly seen , which can be used to derive optical depths , whereas comparison of different ( j , k ) transitions can yield information on the rotational temperature ( eg .@xcite ) .the first suggestion that ammonia may display maser action in interstellar space was made by @xcite who noted that ( 3,3 ) emission toward w33 was not matched by ( 1,1 ) , ( 2,2 ) and ( 4,4 ) transitions , which were all in absorption .they suggested that this may be due to a weak population inversion .maser emission in ammonia transitions was first unambiguously identified by @xcite in the ( 9,6 ) and ( 6,3 ) transitions , shortly followed by @xcite in the ( 3,3 ) transition of @xmath5nh@xmath3 .in addition to these transitions , ammonia masers have been detected in the following transitions : ( 5,4 ) ( 7,5 ) , ( 9,8 ) , ( 10,8 ) @xcite , ( 6,5 ) @xcite , ( 5,5 ) @xcite , ( 6,6 ) @xcite and ( 1,1 ) @xcite .@xmath6nh@xmath3 ( 3,3 ) masers were first detected by @xcite .pumping mechanisms for most ammonia masers remain unclear .metastable transitions of ortho - ammonia , like ( 3,3 ) are thought to be collisionally excited @xcite but such a mechanism will only work for non - metastable transitions where exceedingly high h@xmath7 densities between @xmath8 @xmath9 are found ( eg .@xcite suggest two alternative pumping mechanisms for non - metastable transitions : either pumping by a strong infrared radiation field , such as found around a deeply embedded high mass star , or by a fortuitous overlap of a far - infrared line , which allows a population transfer from a ( j , k ) inversion level to ( j+1 , k ) .ammonia masers are found in regions of high mass star formation , with the best known example being w51 @xcite , with other prominent examples being g9.62 + 0.19 @xcite and ngc6334i @xcite .ngc6334i forms the focus of this work .it is a nearby ( 1.7kpc , @xcite ) region of high mass star formation , traced by bright infrared emission , an ultracompact ( uc ) hii region , mm continuum sources and methanol maser sites .see @xcite for a more comprehensive summary of characteristics of the region .the initial observations were undertaken with the 22 m mopra telescope near siding spring , australia .we obtained a spectrum of ngc6334i at 17 20 53.43 , @xmath1035 47 2.2 ( j2000 ) using position switching on 2006november 28@xmath11 .we used the new mopra spectrometer ( mops ) in broadband mode , which affords us four overlapping ifs of 2.2ghz each .thus , we were able to instantaneously cover 8ghz of frequency space between 19.5 and 27.5ghz .each if has 8192 channels , resulting in a channel width of 269khz , equivalent to 4.1 at 19.5ghz or 2.9 at 27.5ghz .the observations consisted of approximately 10 hours total integration ( 5 hours on source ) with an rms noise level of between 0.01 and 0.03k .the follow - up observations were undertaken on 2007 may 3@xmath12 on the australia telescope compact array ( atca ) in director s time .the correlator setting was full_256_64_256_64 , allowing 0.89 velocity resolution .the primary beam was 2.3 and the array configuration was 1.5c , with baselines ranging from 77 to 4500 m providing an angular resolution of @xmath131 .the two frequencies were tuned to cover the ammonia ( 8,6 ) transition at 20.719221ghz and the ( 11,9 ) transition at 21.070739ghz .these transitions were chosen for atca observations as they appeared the brightest of previously undetected ammonia maser in the mopra spectrum .the region was observed for 9@xmath020 minute cuts in each transition separated over 5 hours . a bright( @xmath41 jy ) , close ( @xmath145@xmath15 ) phase calibrator was observed for 3 minutes before and after each cut . @xmath16 and@xmath17 were used as the bandpass and primary calibrator , respectively .the data were reduced using the miriad package .bad visibilities were flagged , edge channels removed and the gains / bandpass solutions from the calibrator were applied to the visibilities .the data were fourier transformed to form image cubes with 0.2@xmath18 pixels using natural weighting .the images were _ _clean__ed down to 3@xmath19 above the noise to remove sidelobes then convolved with a gaussian beam of the synthesised beam size to produce a final _ _ restor__ed image .continuum emission was extracted from a low - order polynomial fit to line - free channels using _ uvlin _ and images were made from these visibilities in the same way . from previous observations of the primary calibrator , @xmath20 , errors in the absolute flux scale are estimated to be @xmath1310% .noise levels were 6mjy / beam at the frequency of the ( 8,6 ) transition and 3.5mjy / beam at the frequency of the ( 11,9 ) transition .figure [ fig1 ] shows radio continuum emission contours overlaid on thermal ammonia ( 4,4 ) emission @xcite . the ( 4,4 ) emission corresponds well to two mm continuum sources i - sma1 and i - sma2 @xcite .we find , in the atca observations , that emission in both the ( 8,6 ) and ( 11,9 ) transitions is unresolved .we consider these to be masers , for reasons given below .because the emission is unresolved , we do not provide maps of each , but rather show symbols on figure [ fig1 ] to represent the positions of the maser emission . within positional errors( typically 1 for the atca ) , all masers appear to be coincident .figure [ fig2 ] shows the spectra of each line .we measure fluxes of 53 and 3.5jy for the ( 8,6 ) and ( 11,9 ) transitions , respectively .these agree within errors to the respective mopra fluxes of 48jy and 3.2jy , implying there is no significant extended emission that might have been missed by the atca obaservations .therefore , we can be sure the emission comes from a region no bigger than the atca beam of 2.7@xmath0 0.8 .we calculate peak brightness temperature lower limits based on equation [ eq1 ] @xmath21 where @xmath22 is the observing wavelength in millimetres and @xmath23 and @xmath24 are the deconvolved major and minor axes in arcseconds , respectively .we find the ( 8,6 ) emission has a deconvolved size of 0.9@xmath0 0.2and the ( 11,9 ) emission has a deconvolved size of 0.7@xmath0 0.1 .we therefore calculate peak brightness temperatures of 7.8 @xmath25k for the ( 8,6 ) transition and 1.2 @xmath25k for the ( 11,9 ) transition .if the emission is thermal , it would require a local exciting source with a temperature in excess of these values which we consider extremely unlikely .thus we interpret emission in both lines as masers .together with the ( 6,6 ) transition reported to be masing by @xcite , we have identified three masing transitions of ortho - ammonia .are there any other transitions of ortho - ammonia that are masing ?our single - dish mopra observations of ngc6334i suggest there may be others . we detect narrow - lined emission in the ( 7,6 ) , ( 10,9 ) and ( 9,9 ) transitions , as shown in figure [ fig3 ] . although the strengths of each of these lines is not enough to conclusively identify them as masers , there is some circumstantial evidence that they may well be masers .they all exhibit line widths less than 3.8 , which is close to the velocity resolution of the mopra observations .indeed , the atca observations , with finer velocity resolution show both the ( 8,6 ) and ( 11,9 ) lines to have line widths of 1.3 .in contrast to this , all other detected ammonia lines in the mopra spectrum have line widths between 5.8 and 7.3 , and are clearly resolved in velocity . apart from their narrow line widths , emission in each of these transitions peaks around 5 , as for the established ( 6,6 ) , ( 8,6 ) and ( 11,9 ) masers .this is distinctively different from thermal emission , which peaks between -3 and -10 .thermal emission is traced by ammonia ( 1,1 ) , ( 2,2 ) and ch@xmath3oh ( 2@xmath26 2@xmath27 ) @xcite , as well as ammonia ( 3,3 ) , ( 4,4 ) , ( 5,5 ) and ( 6,6 ) @xcite .note that the ammonia ( 6,6 ) observations identified both thermal and maser emission that were clearly separated in velocity .we note that the mopra spectrum covers many other lines of ammonia between 19.5 and 27.5ghz and whilst some of these lines do show emission , it appears all others exhibit broader emission features peaking around -3 and -10 and so are almost certainly thermal emission .we also note that all the potential masing transitions arise from ortho - ammonia .it is not clear why we do not see any para - ammonia masers in ngc6334i .oh ( @xmath28).,scaledwidth=50.0% ] three methods for creating a population inversion of the ammonia upper levels are through collisional excitation , excitation by infrared radiation or a fortuitous alignment of a molecular transition that allows a ( j , k ) transition to populate the ( j+1,k ) upper level @xcite .it is unlikely that collisional excitation is at work here because extremely high densities ( @xmath29 ) are required .since we see at least three masing transitions , and perhaps six , it is unlikely this is due to many lucky alignments of molecular transitions allowing population of the non - metastable states .therefore , we believe the pumping mechanism for ammonia masers in ngc6334i must be via infrared photons .this case is strengthened as ngc6334i is known to be one of the brightest infrared sources in our sky , being intrinsically luminous and relatively nearby ( 1.7kpc ) .this may also explain why these new masers only show thermal emission in other regions .the detection of new ammonia masers in ngc6334i indicates that this source is fertile ground for ammonia maser research .confirmation ( or otherwise ) of the potential ( 7,6 ) , ( 10,9 ) and ( 9,9 ) masers , as well as searching for other non - metastable transitions of ortho - ammonia may yield more maser detections , which will greatly increase the chances of developing theoretical models for ammonia maser pumping and hopefully will lead to a better understanding of ngc6334i , which gives rise to these rare masers .we have observed the two non - metastable transitions of ammonia : ( 8,6 ) and ( 11,9 ) , using the atca .the ( 8,6 ) transition has a peak brightness temperature of 7.8 @xmath25k , whereas the ( 11,9 ) transition has a peak brightness temperature of 1.2 @xmath25k , indicating that both are masers .the ( 11,9 ) transition is e@xmath30k above ground making it the highest energy ammonia maser currently known .the position of both masers is consistent with being coincident , to within positional uncertainties , with maser emission in the previously detected ( 6,6 ) transition @xcite .single - dish observations of ( 7,6 ) , ( 9,9 ) and ( 10,9 ) suggest they may also be masers based on their narrow line widths and overlapping velocity range with the above - mentioned masers , which is distinctly different from the thermal emission systemic velocity .high spatial resolution observations of these transitions are planned to decide whether or not these are masers as well .the authors would like to thank christian henkel for useful discussions related to this work .we would also like to thank henrik beuther for provision of ammonia and continuum data used in this work .we would also like to thank paul ho , the referee , who has greatly helped improve the quality of this work .snl is supported by a scholarship from the school of physics at unsw .the australia telescope is funded by the commonwealth of australia for operation as a national facility managed by csiro .this research has made use of nasa s astrophysics data system .beuther , h. , thorwirth , s. , zhang , q. , hunter , t. r. , megeath , s. t. , walsh , a. j. & menten , k. m. 2005 , apj , 627 , 834 beuther , h. , walsh , a. j. , thorwirth , s. , zhang , q. , hunter , t. r. , megeath , s. t. , menten , k. m. 2007 , a&a , 466 , 989 cesaroni , r. , walmsley , c. m. , churchwell , e. 1992 , a&a , 256 , 618 danby , g. , flower , d. r. , valiron , p. , schilke , p. , walmsley , c. m. 1988 , mnras , 235 , 229 gaume , r. a. , wilson , t. l. , johnston , k. j. 1996 , apjl , 457 , 47 hofner , p. , kurtz , s. , churchwell , e. , walmsley , c. m. , cesaroni , r. 1994 , 1994 , apjl , 429 , 85 hunter , t. r. , brogan , c. l. , megeath , s. t. , menten , k. m. , beuther , h. & thorwirth , s. 2006 , apj , 649 , 888 kraemer , k. e. , jackson , j. m. 1995 , apjl , 439 , 9 madden , s. c. , irvine , w. m. , matthews , h. e. , brown , r. d. , godfrey , p. d. 1986 , apjl , 300 , 79 mauersberger , r. , wilson , t. l. , henkel , c. 1986 , a&al , 160 , 13 mauersberger , r. , henkel , c. , wilson , t. l. 1987 , a&a , 173 , 352 mauersberger , r. , wilson , t. l. , henkel , c. 1988 , a&a , 201 , 123 neckel , t. 1978 , a&a , 69 , 51 walmsley , c. m. , ungerechts , h. 1983 , a&a , 122 , 10 walsh , a. j. , burton , m. g. , hyland , a. r. , robinson , g. , 1998 , mnras , 301 , 640 wilson , t. l. , batrla , w. , pauls , t. a. 1982 , a&al , 110 , 20 zhang , q. & ho , p. t. p. 1995, apjl , 450 , 63

<S> we report the detection of new ammonia masers in the non - metastable ( 8,6 ) and ( 11,9 ) transitions towards the massive star forming region ngc6334i . </S><S> observations were made with the atca interferometer and the emitting region appears unresolved in the 2.7@xmath0 0.8 beam , with deconvolved sizes less than an arcsecond . </S><S> we estimate peak brightness temperatures of @xmath1 and @xmath2k for the ( 8,6 ) and ( 11,9 ) transitions , respectively . </S><S> the masers appear coincident both spatially and in velocity with a previously detected ammonia ( 6,6 ) maser . </S><S> we also suggest that emission in the ( 10,9 ) , ( 9,9 ) and ( 7,6 ) transitions may also be masers , based on their narrow line widths and overlapping velocity ranges with the above masers , as observed with the single - dish mopra radiotelescope . </S><S> masers stars : formation techniques : high angular resolution </S><S> ism : molecules . </S>

1112.2375

as the most part of cosmic rays ( crs ) are charged nuclei , their arrival direction is deflected and made isotropic by the action of galactic magnetic field ( gmf ) that they propagate through before reaching the earth atmosphere .in such a field , the gyroradius of crs is given by @xmath4 , where @xmath5 is in astronomic units and @xmath6 is in teravolt .it must be taken as a reference value , because the gmf is the superposition of regular field lines and chaotic contributions , the strength of both them being still under debate .data available today gives for the local total intensity the value @xmath7 .actually , deviations from the isotropy are expected to occurr as a consequence of the particular realization of the random distribution of cosmic ray sources and magnetic field lines in the galaxy @xcite . in this framework ,the amplitude of the anisotropy is proportional to the rigidity , that is why it is expected to be seen at high energies ( above few hundreds tev ) .however , different experiments @xcite observed an energy - dependent _ `` large scale '' _ anisotropy in the sidereal time frame , well below that threshold .the amplitude is about 10@xmath8 - 10@xmath9 , suggesting the existence of two distint broad regions , one showing an excess of crs ( called `` tail - in '' ) , distributed around 40@xmath1 to 90@xmath1 in right ascension ( r.a . ) .the other a deficit ( the `` loss cone '' ) , distributed around 150@xmath1 to 240@xmath1 in r.a .. the origin of these anisotropies is still unknown .some authors claim that the observations may be due to a combined effect of the regular and turbolent gmf @xcite , or to local uni- and bi - dimensional inflows @xcite .other studies suggest that it can be explained within the diffusion approximation taking into account the role of the few most nearby and recent sources @xcite .easy to understand , more beamed the anisotropies and lower their energy , more difficult to fit the standard model of crs and gmf to experimental results .that is why the evidence of the existence of a medium angular scale anisotropy contained in the tail - in region by the tibet as@xmath10 @xcite and milagro @xcite collaborations in rcent years was rather surprising .similar small scale anisotropies has been recently claimed to be observed by the icecube experiment in the southern hemisphere @xcite .so far , no theory of crs in the galaxy exists which is able to explain few degrees anisotropies in the rigidity region 1 - 10 tv leaving the standard model of crs and that of the local gmf unchanged at the same time . from the experimental viewpoint ,observing anisotropy effects at the level of 10@xmath8 with an air shower array is a difficult job , because of the intrinsic difficulties that this kind of apparatus has to cope with in estimating the exposure .finally , the observation of a possible small angular scale anisotropy region contained inside a larger one rely on the capability for suppressing the anisotropic structures at larger scales without , at the same time , introducing effects of the analysis on smaller scales . in this paper the observation of cr anisotropy at different angular scales with argo - ybjis reported as a function of the primary energy .the argo - ybj experiment , located at the yangbajing cosmic ray laboratory ( tibet , p.r .china , 4300 m a.s.l . , 606 g/@xmath11 ) , is an air shower array able to detect the cosmic radiation at an energy threshold of a few hundred gev .the full detector is in stable data taking since november 2007 with a duty cycle greater than 85% .the trigger rate at the threshold is 3.6 khz .the detector characteristics and performance are described in @xcite .in order to study the anisotropy at different angular scales the isotropic background of crs has been estimated with two methods : the equi - zenith angle method @xcite and the direct integration method @xcite .the equi - zenith angle method , used to study the large scale anisotropy , is able to eliminate various spurious effects caused by instrumental and environmental variations , such as changes in pressure and temperature that are hard to control and tend to introduce systematic errors in the measurement .the method uses data coming from all the angular scales , so that potential small structures are not separated from the underlying large scale modulation .the direct integration method , based on time - average , rely on the assumption that the local distribution of the incoming crs is slowly varying and the time - averaged signal may be used as a good estimation of the background content .time - averaging methods act effectively as a high - pass filter , not allowing to inspect features larger than the time over which the background is computed ( i.e. , 15@xmath1/hour@xmath12 in r.a . ) .the time interval used to compute the average spans @xmath13= 3 hours and makes us confident the results are reliable for structures up to @xmath1435@xmath1 wide .the observation of the cr large scale anisotropy by argo - ybj is shown in the figure [ fig : large_scale ] at different primary energy up to about 25 tev .the data used in this analysis was collected by argo - ybj from 2008 january to 2009 december with a reconstructed zenith angle @xmath15 45@xmath1 . the so - called _ ` tail - in ' _ and _ ` loss - cone ' _ regions , correlated to an enhancement and a deficit of crs , are clearly visible with a statistical significance greater than 20 s.d .. the tail - in broad structure appears to dissolve to smaller angular scale spots with increasing energy .it should be stressed that the energies reported in the figure [ fig : large_scale ] refer to the median energy of all nuclei triggering the experiment . to quantify the scale of the anisotropy we studied the 1-d r.a .projections integrating the sky maps inside a declination band given by the field of view of the detector .therefore , we fitted the r.a .profiles with the first two harmonics .the resulting amplitude of the first harmonic is plotted in the right plot of figure [ fig : large_amplitude ] where is compared to other measurements as a function of the energy .the argo - ybj results are in agreement with other experiments suggesting a decrease of the anisotropy first harmonic amplitude with increasing energy .the figure [ fig : intermediate_scale ] shows the argo - ybj sky map in equatorial coordinates .the analysis refers to events collected from november 2007 to may 2011 after the following selections : ( 1 ) @xmath1625 shower particles on the detector ; ( 2 ) zenith angle of the reconstructed showers @xmath1550@xmath1 .the triggering showers that passed the selection were about 2@xmath1710@xmath18 .the zenith cut selects the declination region @xmath19-20@xmath20 80@xmath1 . according to the simulation ,the median energy of the isotropic cosmic ray proton flux is e@xmath211.8 tev ( mode energy @xmath140.7 tev ) .the most evident features are observed by argo - ybj around the positions @xmath22 120@xmath1 , @xmath19 40@xmath1 and @xmath22 60@xmath1 , @xmath19 -5@xmath1 , positionally coincident with the regions detected by milagro @xcite . these regions , named `` region 1 '' and `` region 2 '' , are observed with a statistical significance of about 14 s.d .. the deficit regions parallel to the excesses are due to a known effect of the analysis , that uses also the excess events to evaluate the background , artificially increasing the background . on the left side of the sky map ,several possible new extended features are visible , though less intense than those aforementioned .the area @xmath23 seems to be full of few - degree excesses not compatible with random fluctuations ( the statistical significance is more than 6 s.d .post - trial ) .the observation of these structures is reported here for the first time and together with that of regions 1 and 2 it may open the way to an interesting study of the tev cr sky . to figure out the energy spectrum of the excesses , data have been divided into five independent shower multiplicity sets .the number of events collected within each region are computed for the event map as well as for the background one .the ratio of these quantities is computed for each multiplicity interval .the result is shown in the figure [ fig : intermediate_amplitude ] .region 1 seems to have spectrum harder than isotropic crs and a cutoff around 600 shower particles ( proton median energy e@xmath24 = 8 tev ) . on the other hand ,the excess hosted in region 2 is less intense and seems to have a spectrum more similar to that of isotropic cosmic rays .the steepening from 100 shower particles on ( e@xmath25 = 2 tev ) is likely related to efficiency effects .further studies are on the way . + the figure [ fig : time_behaviour ] reports the amplitude of the region1 and 2 anisotropies as a function of the ut .time unit is the modified julian date .it can be appreciated that no time evolution of the anisotropies are there , nor evidence of correlation of the emission .this paper reports the observation of cr anisotropy at different angular scales with argo - ybj , as a function of the primary energy .the large scale cr anisotropy has been clearly observed up to about 25 tev .evidence of existence of different few - degree excesses in the northern sky ( the strongest ones positionally coincident with the regions detected by milagro in 2008 ) is reported . the time distribution of the phenomenon has been showed too .a discussion of the systematic effects which may come from imprecisions in estimating the reference level is outlined .in fact , a wrong estimation affect the significance and the relative intensity sky maps , even creating artifacts ( i.e. fake excesses or deficit regions ) .drifts in detector operating conditions or atmospheric effects on the air shower development are quite hard to be modeled , then difficult to be accounted for to sufficient accuracy . as anisotropies of the order 10@xmath8 are looked for , operating conditions must be known down to this level , all across the field of view and during all the acquisition time .given the importance of the topic , a joint analysis of concurrent data recorded by different experiments in both hemispheres , as well as a correlation with other observables like the interstellar energetic neutral atoms distribution @xcite , should be a high priority to clarify the observations .9 blasi p. and amato e. 2011 _ arxiv:1105.4521 _ and _ arxiv:1105.4529_. nagashima k. et al .1998 _ j. of geoph .res . _ * 103 * 17429 .giullian g. et al .2007 _ phys .* 78 * 062003 .amenomori m. et al .2006 _ science _ * 314 * 439 .2009 _ apj _ * 698 * 2121 .aglietta m. et al .2009 _ apj _ * 692 * l130 .abbasi r. et al .2011 _ apj _ * 740 * 16 .battaner e. et al .2009 _ apj _ * 703 * l90 .amenomori m. et al .2010 _ astrophys .space sci .trans . _ * 6 * 49 .erlykin a.d . and wolfendale a.w .2006 _ astrop .phys . _ * 25 * 183 .amenomori m. et al .2007 _ aip conf .* 932 * 283 .2008 _ phys .* 101 * 221101 .bartoli b. et al .2011 _ phys .d _ * 84 * 022003 .amenomori m. et al .2005 _ apj _ * 633 * 1005 .fleysher r. et al .2004 _ apj _ * 603 * 355 .bartoli b. et al .2011 _ apj _ * 734 * 110 mccomas d.j .2009 _ science _ * 326 * 959 .mccomas d.j .2011 _ geoph .lett . _ * 38 * l18101 .

<S> important informations on the origin and the propagation mechanisms of cosmic rays may be provided by the measurement of the anisotropies of their arrival direction . in this paper the observation of anisotropy regions at different angular scales is reported . </S><S> in particular , the observation of a possible anisotropy on scales between @xmath010@xmath1 and @xmath030@xmath1 may be a key - detection for speculations on the presence of unknown features of the magnetic fields the charged cosmic rays propagate through , as well as to potential contributions of nearby sources to the total flux of cosmic rays . </S><S> evidence of new weaker few - degree excesses throughout the sky region @xmath2 r.a . </S><S> @xmath3 is also reported . </S>

1309.7132

by identifying unambiguously the fission fragments according to their mass number and their charge , the fission yields as function of the excitation energy were measured @xcite .the experimental data showed a dominance of odd - odd fragments at excitation energies close to zero for u isotopes .this phenomenon was independently remarked also in ref .@xcite for cf . a first interpretation involved a proportionality between the level densities of the fission fragments and the yields .recently , a new set of time - dependent coupled channel equations derived from the variational principle was proposed to determine dynamically the mixing between seniority - zero and seniority - two configurations @xcite .the essential idea is that the configuration mixing is managed under the action of some inherent low lying time dependent excitations produced in the avoided crossing regions , that is , a dynamical mechanism like the landau - zener effect @xcite . in the pioneering investigations of the mass and charge distributions at low excitation energies ,the experimental results showed a preference for the mass division leading to even-@xmath0 fragments @xcite .it was believed that at very low excitation energies , i.e. , high kinetic energies , the fragments will be fully paired .improving the experimental procedure , these facts were contradicted in ref .@xcite where no strong even - odd effect was evidenced in the thermal neutron induced fission for high values of the total kinetic energy .even and odd partitions were observed experimentally close to their respective @xmath1-values for four systems investigated : @xmath2u(n@xmath3,f ) , @xmath4pu(n@xmath3,f ) and @xmath5cf(sf ) .it is worth to underline that the importance of the landau - zener effect in the cold fission fragmentation behavior was anticipated , as mentioned in ref .however , the odd - even structure structure in fission is explained usually within statistical arguments , as for example in refs .recently , some arguments linked the odd - even structure also to the charge asymmetry evolution during the fission process @xcite , not only to the dissipated energy as has been done earlier @xcite .in this section , the ingredients required to investigate dynamically the odd - even effect are described .the usual theories of fission consider that the nuclear system is characterized by several generalized coordinates @xmath6 , @xmath7 . these coordinates vary and force the system to split into two separated fragments .the single - particle energies , and the many - body wave function are determined by the the variation in time of these coordinates . in order to deduce the microscopic equations of motion ,the starting point is a many - body hamiltonian with pairing residual interactions : @xmath8 this hamiltonian depends on the collective parameters @xmath9 , allowed to vary with respect the time . here ,@xmath10 are single particle energies , @xmath11 and @xmath12 denote operators for creating and destroying a particle in the state @xmath13 , respectively .the state characterized by a bar signifies the time - reversed partner of a pair .the pairing correlations arise from the short range interaction of correlated pairs of fermions moving in time - reversed orbits .@xmath14 is the matrix element of the pairing interaction and its value is in principle dependent on the overlap of the wave functions of the pairs . for a given nucleus ,it is possible to approximate the pairing interaction matrix elements with a constant value by using a renormalization procedure that depends on the number of states in the active pairing space and on the structure @xcite .concerning the pair breaking effect , it must be evidenced that only two types of velocity dependent excitation mechanisms are possible between different single particle ( or molecular ) states of a dynamical nuclear system @xcite : the radial coupling that can be described by the landau - zener effect in avoided levels crossings regions and the coriolis one produced in the region of real crossings .the coriolis coupling is responsible for transitions between levels characterized by quantum numbers of the projection of the spin that differ by one unity .this coupling is important in the treatment of the @xmath15-decay where the inertia is small and such an investigation was made in ref .concerning the dynamical pair breaking , it can be described by a mechanism similar to the landau - zener one .the perturbation that produces the pair breaking in avoided crossing regions between levels with the same good quantum numbers is obtained in terms of quasiparticle creation and annihilation operators @xmath16 within the previous transformations , some perturbations that break dynamically a cooper pair when the system traverses an avoided level crossing region are constructed . the parameters @xmath17 and @xmath18 are occupation and vacancy amplitudes , respectively , for a pair occupying the single - particle level @xmath13 of the configuration @xmath19 .the seniority - zero configuration is denoted with 0 and the seniority - two configuration by a pair of indexes @xmath19 , @xmath20 and @xmath21 denoting the levels occupied by unpaired fermions . as evidenced in ref .@xcite , the three situations plotted in fig .[ figura1 ] can be modeled . in the plot [ figura1](a ) , the cooper pair remains on the adiabatic level @xmath22 , in [ figura1 ] ( b ) the pair destruction is illustrated , while in [ figura1 ] ( c ) two fermions generate a pair . to describe these three situations , a residual perturbation in the avoided level crossing regionis postulated as follows : @xmath23 , \end{array } \label{cort}\ ] ] where @xmath24 is the interaction energy between levels in the avoided level crossing regions . under the action of the interaction @xmath24 , according to the identities ( [ proper ] ) of the appendix , the operators @xmath25 and @xmath26 transform a bogoliubov seniority zero wave function into a seniority two one , and vice versa , being responsible for configuration mixing .the products over the index @xmath27 in the previous formula transform the remaining bogoliubov amplitudes from values pertaining to the seniority zero wave function to those pertaining to the seniority two functions with unpaired orbitals @xmath20 and @xmath21 , and vice - versa .this dynamical pair breaking effect was theoretically formulated in ref .@xcite for the first time .it is important to note that the same kind of perturbation was used also to generalize the landau - zener effect in superfluid systems @xcite .it was demonstrated in ref .@xcite that equations governing the landau - zener effect and the time dependent pairing equations are two particular cases of a new set of coupled channel equations . the two fission products must be characterized by integer numbers of neutrons and protons . as a consequence , the sums of the occupation probabilities of single particle levels of the two fragments must give the mass and charge numbers . by solving the equations of motion , unfortunately, the sum of single - particle densities ( bcs occupation probabilities ) of the single particle levels belonging to the two fission fragments obtained after the scission do nt give exactly their numbers of nucleons .a recipe can be implemented to fix dynamically these numbers of particles in the two final fragments by using the operators for the number of particles @xmath28 @xmath29 that act on each fission product . at scission ,the two fission fragments must be characterized by a supplementary condition @xmath30 where @xmath31 and @xmath32 are the number of particles in the final fragments labeled 1 and 2 , respectively , and @xmath33 are the corresponding operators . here , @xmath34 and @xmath35 run over the pairing active level states that are located in the potential wells of the final fragment 1 and of the final fragment 2 , respectively .the condition ( [ coop ] ) can be introduced in the equations of motion by means of the lagrange multipliers @xcite .all the previous ingredients could be used to obtain the microscopic equations of motion .these equations are obtained from the variational principle by minimizing the following energy functional @xmath36 the trial many - body function @xmath37 is a superposition of bogoliubov seniority zero and seniority two wave functions @xmath38\mid 0\rangle . \end{array } \label{wf2}\ ] ] @xmath39 is the amplitude of the seniority zero wave function while @xmath40 are amplitudes for the seniority two wave functions for configurations in which the single particle orbitals @xmath21 and @xmath41 belonging to the active pairing space are each blocked by only one unpaired nucleon . here , the vacancy @xmath42 and occupation @xmath43 amplitudes are not the adiabatic solutions of the bcs equations and depend on the variation in time of the generalized parameters and the history of the nuclear system .@xmath44 is a lagrange multiplier .the evolution in time of the nuclear system , if the collective parameters @xmath45 ( @xmath46 ) vary , is obtained by performing the variation of the functional ( [ var ] ) . the procedure required for the functional variationis described in details in the appendix .the next coupled channel equations are obtained , eventually : @xmath47 @xmath48 @xmath49 @xmath50 @xmath51 @xmath52 @xmath53 @xmath54 where @xmath21 , @xmath13 , @xmath41 , @xmath55 , @xmath56 label the single particle levels in the active pairing space .the sign @xmath57 ensures that the matrix element of the expression ( [ coop ] ) is positive . @xmath58 or @xmath59 if the state @xmath13 will be located in the fragment 1 or in the fragment 2 after the scission , respectively . here ,the following notations are used : @xmath60 the symbol @xmath61 gives the gap parameter .( the label @xmath62 denotes here generically a specific configuration . )the variables that depend on the time through the generalized coordinates are the single particle densities @xmath63 , the pairing moment components @xmath64 , the probabilities to have a given seniority configuration @xmath65 , and the moment components between configurations @xmath66 .the relations ( [ ec1])-([ec2 ] ) are the well known time dependent paring equations previously deduced in refs .these formulas are identical to the time dependent hartree - fock - bogoliubov equations @xcite . the symbol @xmath67 denotes the landau - zener interaction , while @xmath68 and @xmath69 are energy terms .the significance of the quantities appearing in the equations can be understood in the appendix .the condition that @xmath70=1 is implicitly ensured through eqs .( [ ec3 ] ) and ( [ ec4])because @xmath71 .the energy of the nuclear system in the seniority zero state is @xmath72 and in the seniority two state is @xmath73 the corresponding lower energy states @xmath74 and @xmath75 of the nuclear system in a given configuration are obtained with the previous relations by replacing the densities @xmath76 and the pairing moment components @xmath77 with the adiabatic values obtained in the bcs approximation and using the same single particle level scheme . consequently , as defined in ref .@xcite , along the fission path the average dissipated energies @xmath78 will be @xmath79 in the seniority zero and the seniority two configurations , respectively .we subtracted from the total potential energy of the nuclear system , its adiabatic value .it was already shown in ref .@xcite that the mean value of the dissipated energy becomes larger when the velocities of the generalized coordinates increase .the single particle levels belonging to the core of the initial parent nucleus are rearranged in the two cores of the fragments .knowing the number of levels in each core , it is possible to redefine the pairing active space of each fragment and the asymptotic values of the lower energy states can be evaluated .after the scission , @xmath80 must be replaced by the sum @xmath81 , where @xmath82 and @xmath83 are the lower energies of the two fission fragments .a similar rule is valid also for seniority two configurations . in the same time, the pairing interaction matrix elements @xmath84 between pairs pertaining to different nuclei are zero . for each channel , asymptotically @xcite , the next limits hold @xmath85 where the indexes 1 and 2 refer to the two fragments .if only one pair is broken along the fission path , the spin of the two nuclei delivered from one unique seniority two configuration must be the same .this last assumption is due to the fact that the landau - zener effect is produced between levels with the same good quantum numbers . in order to determine the fission barriers , the total energy of the nuclear systemis computed in the framework of the macroscopic - microscopic method @xcite .as mentioned previously , the whole system is characterized by some collective coordinates that determine approximately the behavior of many other intrinsic variables .the essential idea of this approach is that a macroscopic model , as the liquid drop one , describes quantitatively the smooth trends of the potential energy with respect the particle number and deformation whereas a microscopic approach as the shell model describes local fluctuations .the combined macroscopic - microscopic method should reproduce both smooth trends and local fluctuations .the basic ingredient in such an analysis is the shape parametrization that depends on several macroscopic degrees of freedom .the macroscopic deformation energy is calculated within the liquid drop model .a microscopic potential must be constructed to be consistent with this nuclear shape parametrization .a microscopic correction is then evaluated using the strutinsky procedure @xcite .the basic ingredient of the model is the nuclear shape parametrization . in the following , an axial symmetric nuclear shape surface during the deformation process from one initial nucleus to the separated fragmentsis obtained by smoothly joining two spheroids of semi - axis @xmath86 and @xmath87 ( @xmath20=1,2 ) with a neck surface generated by the rotation of a circle of radius @xmath88 around the axis of symmetry . by imposing the condition of volume conservationwe are left by five independent generalized coordinates @xmath89 ( @xmath56=1,5 ) that can be associated to five degrees of freedom : the elongation @xmath90 given by the distance between the centers of the spheroids ; the necking parameter @xmath91 related to the curvature of the neck , the eccentricities @xmath92 associated with the deformations of the nascent fragments and the mass asymmetry parameter @xmath93 , @xmath94 ( @xmath951,2 ) denoting the volumes of the virtual ellipsoids characterized by the semi - axis @xmath86 and @xmath87 .the nuclear shape parametrization is displayed in fig .[ figura2 ] .the entire model can be considered valid as long as the generalized coordinates and their variations in time make sense . the many - body wave function and the single particle energiesare provided by the woods - saxon two - center shell model @xcite .the woods - saxon potential , the coulomb interaction and the spin orbit term must be diagonalized in an eigenvector basis .the asymmetric two center shell oscillator provides an orthogonal eigenvector basis for only one hermite space @xcite . in this hermite spacethe behavior of both fragments can be described .when the elongation @xmath90 is zero , the eigenvector basis becomes that of the anisotropic oscillator .when @xmath90 tends to infinity , a two oscillator eigenvector system is obtained in the same hermite space , centered in the two fragments . in the intermediate situation , each eigenfunction has components in the two subspaces that belong to the fragments .so , the two center shell model provides permanently the wave functions associated to the lower energies of the single particle states pertaining to a major quantum number @xmath96 .therefore , molecular states formed by two fragments at scission could be precisely described .another feature of the two center shell model is related to the localization of the single particle wave function in one of the two potential wells after the scission . as evidenced in ref .@xcite , it is possible to predict this localization for a given fragmentation before that the scission is produced .this feature helps us to fix the number of particles in each fragment by resolving eqs .( [ ec1])-([ec6 ] ) .this model was widely used by the bucharest group in the calculations addressing the cluster @xcite and alpha decay @xcite , the fission @xcite , or the heavy element synthesis @xcite .for example , the model was able to describe two tangent nuclei in a wide range of mass asymmetries .the half - lives for cluster decay were reproduced .a mechanism for the formation of an @xmath15-particle on the nuclear surface was supplied .fission barriers that agree with the evaluated ones were calculated .if the different penetrabilities which characterize every channel through the barrier are taken into account , it is expected that the daughter ground state is strongly enhanced in the exit channel .indeed , in the cases of the excited channels , the barrier must be increased with approximately the value of the excitation energy of the unpaired nucleon and , therefore , the penetrability is decreased exponentially .the amount of which the barrier is modified can be estimated accounting the specialization energy .wheeler defined this specialization energy @xcite as the excess of the energy of a nucleon with a given spin over the energy for the same spin nucleon state of lowest energy .so , the final excitation is given by the sum between the dissipated energy and the single particle excitations that are equivalent to the specialization energies . the probability to obtain a given partition in the mass distributionis determined by the penetrability of the fission barrier @xmath97}dr\right\}.\ ] ] the exponent is the classical action integral taken along the fission path at a given energy @xmath98 that connects the two turning points @xmath99 and @xmath100 . only positive values of @xmath101 $ ] are integrated .@xmath102 is the inertia along the fission trajectory and @xmath103 is the deformation energy .the fission yields as function of the total excitation energy for the fragmentation @xmath104kr+@xmath105ba ( even - even ) , @xmath104rb+@xmath105cs ( odd - odd ) will be investigated in the framework of our model .experimental data are available for the reaction @xcite @xmath106u(n@xmath3,f ) and the model can be tested . in order to determine the rearrangement of the single particle energies ,the first step is the calculation of the fission path from the ground state up to scission . as described in ref .@xcite , the minimal action principle can be used to determine the best fission trajectory in the configuration space spanned by our five generalized coordinates . for this purpose ,two ingredients are required : the deformation energy @xmath103 and the tensor of the effective mass @xmath107 .the deformation energy was obtained in the frame of the microscopic - macroscopic method .the effective mass is computed within the cranking approximation . by minimization ,the theory can give us the most probable fragmentation , that corresponds to a heavy fragment with mass around 138 . in our work, the fission path for the 144/90 fragmentation is required . as evidenced in ref .@xcite , it is not yet understood how the compound nucleus is transformed in a variety of different fragmentations .it is also believed that the models for mass distributions have limited predictive power . to overcome these difficulties ,some simple assumptions are made in our work .we rely on the results of refs . @xcite where the mass distribution of the fragments was relatively well reproduced by considering that the variation of the mass asymmetry is linear from the saddle configuration of the outer barrier up to the exit point .it is considered that under the rapid descent from the top of the second barrier different mass partitions are obtained . to obtain these different mass partitions ,the simplest way is to vary the mass asymmetry parameter @xmath108 and the averaged deformations of fragments as in ref .the fission barrier is plotted in fig .[ pot ] for the partition @xmath109 . at @xmath11017 fm ,the variation of the mass asymmetry parameter is started , producing a small bumb in the barrier . in the panel ( b ) of fig .[ pot ] , the effective mass is displayed .the inertia along the trajectory was computed within three approximations : the non - adiabatic cranking model @xcite , the gaussian overlap approximation @xcite , and the cranking approach @xcite . in order to solve the equations of motion ,we need the single particle energies , the variations of the pairing matrix elements @xmath14 , the interactions in the avoided crossing regions , and the velocity of passage through these regions . in figs .[ ttn ] and [ ttp ] , the single particle energy levels obtained within the two center woods - saxon model are displayed along the fission trajectory for neutrons and protons , respectively . with a dashed thick linethe fermi levels of the parent nucleus are plotted .the pair of levels with the same spin projection that gives the lower single particle seniority two excitation ( or specialization energy ) are also displayed with thick full lines .it can be observed that the levels that pertain to the lower excitation are close to the fermi energy . in the case of protonsthe spin projection @xmath111 of these levels is 3/2 while in the case of neutrons it is @xmath111=5/2 .these values of @xmath111 give also the final spin of the partners .the barriers for the seniority two configurations were displayed in fig .[ pot ] with dot - dashed and dashed lines .the difference between the seniority two and seniority zero energies , that is the single particle excitations , are obtained by subtracting their values computed with rels .( [ enij ] ) and ( [ en0 ] ) and by using the adiabatic bcs amplitudes .our pairing active space is constructed with 58 single particle levels around the fermi energy .an important question is the identification of the avoided crossing regions that arise from the strong energy fluctuations of the single particle levels observed along the fission path . for this purpose, we selected all the pairs of adjacent levels with the same spin projection in the active pairing space . from an analyze of the rearrangement of these levels in a manner similar to that of ref .@xcite , the avoided crossing regions were identified .the positions of some avoided level crossing regions are marked with arrows in fig .[ pot ] . in these regions ,the excitations are possible between the mentioned seniority two configurations and the seniority zero one . using the interpolation method described in ref .@xcite , the magnitude of the interactions was calculated . finally , we selected 32 seniority two configurations for protons and 31 configurations for neutrons that are coupled to the seniority zero configuration through avoided levels crossings regions . in the frame of the adiabatic bcs model @xcite , a mean value of the pairing interaction @xmath112can be associated to a given active pairing space , by using a renormalization procedure .so , the values of @xmath112 for all seniority states were computed for the parent and for the two fragments . in the case of senioritytwo states , the blocked levels are eliminated from the single particle diagram . a linear interpolation between the values of @xmath112 from those of the parent to those of the fragments is realized in order to solve the equations of motion , as it will be specified in the following .so , all the most important quantities required to solve the equations of motion are provided : the single particle energies @xmath10 , the perturbations @xmath24 and the pairing interactions @xmath112 . the dependence in timeis introduced by means of the variations of the collective coordinates . in this respect ,several values of the internuclear velocity @xmath113 are tested in order to solve the equations : 5@xmath114 , 8@xmath114 , @xmath115 , 3@xmath116 , @xmath117 , 3@xmath118 , @xmath119 , 3@xmath120 , @xmath121 , and 3@xmath122 fm / fs .( [ ec1])-([ec2 ] ) do nt have a dependence on @xmath123 and @xmath66 and can be solved separately for different values of the internuclear velocities .the initial conditions are given by the adiabatic bcs values for the ground state deformation of the parent ( located at @xmath110 4.6 fm ) of all seniority states involved .the equations are solved in a way similar to that presented in ref .that is , at the beginning of the reaction the system evolves without constraints . at the internuclear distance@xmath90=17 fm , close to the top of the outer barrier , in eqs .( [ ec22 ] ) and ( [ ec2 ] ) the value of @xmath44 is set to non - zero values . in the same time , two linear variations of the mean value of the pairing interactions @xmath112 are started in order to reach the final values @xmath124 and @xmath125 that characterize the two fragments at scission . when the equalities between the sums of single particle occupation probabilities and the numbers of nucleons are obtained , @xmath44 and the pairing interaction between the two fragments @xmath126 are set to zero . for @xmath127 , from conservation conditions as explained in ref .@xcite , the sum of occupation probabilities that run on the single particle states of the two fragments are preserved .the solutions @xmath128 and @xmath129 are obtained up to an internuclear distance of @xmath90=22 fm .the calculated values of @xmath68 and @xmath130 as function of @xmath90 are recorded and will be used later to solve the eqs .( [ ec3])-([ec6 ] ) .for the two isospins and for all velocities and seniority configurations involved , the systems are solved within the runge - kutta algorithm .the dissipated energies @xmath131 after the scission as function of the internuclear velocity @xmath132 , computed according to eqs .( [ disip ] ) and ( [ disip2 ] ) , are plotted in fig .[ gre ] for the seniority zero and the lower single particle excitation seniority two states .the general trend exhibited in both cases is an increase of the final dissipated energy when the internuclear velocity becomes larger .the same behavior is typical for all seniority configurations involved .so the general rule that assesses that the dissipation is proportional to the velocity is retrieved .another way to introduce the dissipation in quantum systems is to consider forces proportional to the velocity in analogy with the friction forces in classical mechanics @xcite , leading to a generalized schrdinger non - linear equation for an open system .the initial condition for the equations ( [ ec3])-([ec6 ] ) that describe the configuration mixing is @xmath133=1 , all the other values being zero . from physical considerations , we imposed in the numerical code the conditions that @xmath1340 if @xmath135 or if @xmath136 , and that @xmath1370if @xmath138 or if @xmath139 . that ensures that the probability to have any configuration must be in the interval @xmath140 $ ] .the final probabilities of realization of two configurations for neutrons and protons as function of the internuclear velocities are displayed in fig .this figure evidences the fact that the probability of realization of adiabatic seniority zero states are close to zero for small velocities , exactly in the energy region in which the dissipation is smaller .so , the main features concerning the relation between the final excitation energy and the probability of realization of a given channel presented in ref .@xcite are retrieved and are valid even if the method for dynamical projection on final atomic and mass numbers is used .the results concerning the final excitation and the probabilities of realization presented above are coupled through the collective velocity parameter . in order to compare the theoretical findings with the experimental data ,this velocity must be eliminated .therefore , a simple model related to the experimental arrangement is conceived in order to relate the dissipation and the yields . by bombarding the @xmath106u with thermal neutrons , using mass evaluations@xcite we found that an excitation of at least @xmath141=6.85 mev is accumulated in the compound @xmath142u nucleus .as represented in fig .[ pot ] , this energy is shared between a potential part @xmath98 and a kinetic one @xmath143 , where @xmath144 is the inertia in the ground state configuration . a constant population of all values of the kinetic energy is assumed .the penetrability @xmath145 of each channel depends on the turning points of the barrier , that is , on @xmath132 through the relation @xmath146 .that leads to a larger penetrability when the velocity decreases .the penetrability depends also on the excitation channel @xmath147 that can be constructed with the specialization energies of the configurations @xmath148 or @xmath19 of the two isospins .the final excitation energy of the two fragments is @xmath149 , that is , it corresponds to a sum between the dissipated energy @xmath131 and the single particle excitations @xmath150 of the two fragments .the dependence of the yield as function of the final excitation @xmath151 in the interval [ @xmath152 exhibits the following proportionality : @xmath153 \nonumber\\ \times w(v)\theta(e_{x1}^*-e_x^*(v,\gamma_n,\gamma_p))\theta(e_x^*(v,\gamma_n,\gamma_p)-e_{x2}^*))dv,\nonumber\end{aligned}\ ] ] for the even - even channel .here @xmath154 are the penetrabilities that depend on a specific channel and @xmath133 are the probabilities of realization given by the time dependent equations for neutrons ( index @xmath155 ) and protons ( index @xmath156 ) .for all velocities that give an excitation in the interval [ @xmath152 , the previous formula reflects the fact that the yields are proportional to these penetrabilities and probabilities .of course , at scission it is possible to obtain even - even partitions even if a cooper pair is broken . in this case ,one of the fission products picks this broken pair and will carry a very large excitation energy .therefore , the sums that run over the channels @xmath62 in the expression ( [ y0 ] ) take into consideration the fact that some configurations are formed with a broken pair in only one partner .the probabilities of realization of a given seniority configuration @xmath157 depend on the internuclear velocity .the penetrability of the fission barrier @xmath145 at the velocity @xmath132 in the channel @xmath158 depends on the variation of the probabilities of realization along the fission path .the factor @xmath159 is a weighting that reflects the dependence of the kinetic energy of the velocity , because @xmath160 , and @xmath161 is the step heaviside distribution used to select only events in the interval @xmath162 $ ] .@xmath144 is considered to be the inertia in the ground state of the parent nucleus .the maximal value of the the velocity is obtained from the boundary @xmath163 .the penetrability @xmath145 is calculated by considering the turning points at the energy @xmath164 .the calculations were made within three approximations for the inertia : the cranking model @xcite , the gaussian overlap approximation @xcite and the non - adiabatic cranking approach @xcite .the yields for odd - odd partitions are proportional with the expression : @xmath165 in the previous relation , the sums run over all the seniority two configuration taken into consideration for neutrons ( index @xmath56 ) and protons ( index @xmath166 ) that give unpaired nucleons in the two fragments .the results obtained for the folded distributions @xmath167 and @xmath168 are displayed in fig .[ grafinal ] .the averaging interval is @xmath169=0.5 mev .the observed experimental trends exhibited in fig . 4 of ref .@xcite for cold fission yields were reproduced .the trends are same for the three approaches used for the inertia . at low excitation energythe odd - odd yields surpass the even - even ones .the even - even yields become larger for excitation energies larger that 3 - 4 mev , in accordance with the experimental findings .in this work , a microscopic model is proposed for the explanation of the odd - even effect in cold fission .this explanation is based on a mixing configuration mechanism that is produced during the fission process .this configuration mixing mechanism is obtained dynamically by solving a the generalized system of time dependent pairing equations , that include a pair - breaking effect .a first rule can be extracted from this model .the even - even fission products can not be obtained at zero excitation energies because of the existence of dynamical excitations produced in the avoided level crossing regions when the nuclear system deforms slowly .the magnitudes of the interactions and the location of the the avoided crossing regions are fixed along the fission path and are independent of the velocity of passage through these regions .if this velocity is large , the perturbation will act onto the cooper pair a small fraction of time .so , the chance to break a pair will be small .if the velocity is low , the pairs will traverse the regions in larger time durations .so , the probability to break a pair increases . on another hand ,high velocities lead to large dissipation energies .another characteristic that was not exploited in this work can be featured from the model .the lower excitation energies of a combination of two odd - odd partners can be obtained only if their spin are the same for neutrons and protons .if the spins of the partners are not the same , the model predicts that at least two pair ruptures are produced for neutrons or protons and additional single particle excitations must be taken into consideration .the possibility to jump from one level to another in a large scale amplitude motion was predicted by hill and wheeler in ref .dissipation in terms of landau - zener crossings during fission was first proposed in ref .@xcite where excitations were considered only for time - reversed pairs , neglecting the possible existence of unpaired nucleons . as evidenced in ref .@xcite , many studies were performed in order to exploit this mechanism in different type of processes .it was also shown in ref .@xcite that the landau - zener mechanism is cached in the time dependent pairing equations ( [ ec1 ] ) and ( [ ec2 ] ) .pairs undergo landau - zener transitions on virtual levels with coupling strengths given by the value of the magnitude of the gap parameter .anyhow , it is the first time that the dynamical pair breaking effect was used to explain the odd - even effect in fission .it must be mentioned that a time - dependent microscopic approach to the scission process was described in ref .they observed that for scission time larger than 5@xmath170 s , the single particle excitations are negligible .this time that characterizes the neck rupture corresponds to a velocity of @xmath121 fm / fs in our calculations .the density of single particle levels increases in the region of the second barrier .therefore , from the outer saddle to the scission , many avoided crossing regions are produced and the chance to break a pair is enhanced .up to the second saddle , the number of avoided crossing regions is small an the system evolves merely in the seniority zero configuration .that gives a large penetrability for all channels , the mixing of configurations being produced especially in the outer barrier region . in conclusion , by solving the dynamical microscopic equations of motion for an fissioning even - even systemit is found that the probability to obtain an odd - odd partition overcomes the probability of an even - even one at excitation energies smaller than 4 mev , for the same division in mass numbers .the theoretical results are in accordance with the experimental behavior of the odd - even distributions at high kinetic energies .it is the first time that this behavior was explained within a quantum mechanical approach .* acknowledgements * work supported by cncs - uefiscdi , project number pn - ii - id - pce-2011 - 3 - 0068 .the next identities are used to develop the functional ( [ var ] ) @xmath171 where we introduced the sign @xmath173 in order to have a positive value of the matrix element of the condition ( [ coop ] ) . in this last relation @xmath174 or @xmath175 denotes a configuration .@xmath58 or @xmath59 if the state @xmath13 will belong to the fragment 1 or 2 after the scission , respectively .the matrix elements of the time derivatives of the wave functions are neglected because they are considered to be responsible only for the inertia parameter .these matrix elements were investigated in refs .@xcite , where a relation between the inertia and the dissipation was revealed . by using the following properties of the creation and annihilation operators : @xmath178 the next equalities are deduced : @xmath179 using the previous identities , the energy functional ( [ var ] ) reads , eventually : @xmath181\right\}\\ + \sum_{l , j\ne l}^{n}h_{jl}(c_{0}^{*}c_{jl}+c_{0}c_{jl}^{*})\\ = \mid c_{0}\mid^{2}\bar{e}_{0}+\sum_{j , l\ne j}\mid c_{jl}\mid ^{2}\bar{e}_{jl}\\ -\left\{i\hbar c_{0}^{*}\dot{c}_{0}+\mid c_{0}\mid^{2}\sum_{k}t_{k(0 ) } + \sum_{j , l\ne j}\left[i\hbar c_{jl}^{*}\dot{c}_{jl}+\mid c_{jl}\mid^{2 } \sum_{k\ne j , l } t_{k(jl)}\right]\right\}\\ + \sum_{l , j\ne l}^{n}h_{jl}(c_{0}^{*}c_{jl}+c_{0}c_{jl}^ { * } ) , \end{array } \label{expr0}\ ] ] where @xmath182 and @xmath183 are terms that include the energies @xmath184 and @xmath185 ( given by formulas ( [ en0 ] ) and ( [ enij ] ) ) of the seniority zero and the seniority two configurations , respectively , @xmath186 the time dependent equations are obtained by minimizing the functional ( [ expr0 ] ) , that is , differentiating with respect the independent parameters @xmath191 , @xmath17 , @xmath192 and @xmath40 , together with their complex conjugates . the eqs .( [ ec1])-([ec2 ] ) were derived in different ways in refs .@xcite , and it is straightforward to introduce the condition ( [ coop ] ) .so , we will focus on the derivation of eqs .( [ ec3])-([ec6 ] ) , related to the mixing of configurations . to obtain the derivatives ,the following relations must be used : @xmath193=0;\\ \dot{c}_{0}c_{0}^{*}+\sum_{j , l\ne j}\dot{c}_{jl}c_{jl}^{*}= -\dot{c}_{0}^{*}c_{0}+\sum_{j , l\ne j}\dot{c}_{jl}^{*}c_{jl};\\ { \partial \dot{c}_{0}c_{0}^{*}\over \partial c_{0}}=-\dot{c}_{0}^{*};\\ { \partial \dot{c}_{jl}c_{jl}^{*}\over \partial c_{jl}}=-\dot{c}_{jl}^{*}. \end{array}\ ] ] the previous conditions ensure that the sum of the probabilities @xmath194 of all configurations is one . if the notations ( [ notatii ] ) are used in the expression ( [ em6 ] ) , ( [ em7 ] ) , ( [ em8 ] ) and ( [ em9 ] ) , the time dependent equations ( [ ec3])-([ec6 ] ) for the dynamical pair breaking effect with dynamic projection of number of particles are obtained .e. schwab , h .-clerc , m. mutterer , j.p .theobald , h. faust , nucl .phys . a 577 ( 1994 ) 674 .hambsch , h .- h .knitter , c. budtz - jorgensen , nucl .phys . a 554 ( 1993 ) 209 .knitter , f .- j .hambsch , c. budtz - jorgensen , nucl .phys . a 536 ( 1992 ) 221 .m. mirea phys .b 680 ( 2009 ) 316 . d.l .hill , j.a .wheeler , phys .89 ( 1953 ) 1102 .reisdorf , j.p .unik , l.e .glendenin , nucl .phys . a 205 ( 1973 ) 348 .h. wohlfarth , w. lang , h .-clerc , h. schrader , k .- h .schmidt , h. dann , phys . lett .b 63 ( 1976 ) 275 . c. signarbieux , m. montoya , m. ribrag , c. mazur , j. phys .lett . 43 ( 1981 ) l437 .f. gonnenwein , phys .procedia 47 ( 2013 ) 107 .f. rejmund , a.v .ignatyuk , a.r .junghans , k .- h .schmidt , nucl phys .a 678 ( 2000 ) 215 .v. avrigeanu , a. florescu , a. sandulescu , w. greiner , phys .c 52 ( 1995 ) r1755. m. montoya , j. phys .44 ( 1983 ) 7853 .clerc , heavyy elements and related new phenomena , ed .w. greiner , r.k .gupta , world scientific , singapore , 1999 , p. 471 .m. caamano , f. rejmund , k .- h .schmidt , j. phys .g 38 ( 2011 ) 035101 .schmidt , b. jurado , eprint arxiv:1007.0741 p. armbruster , in : w.j .myers ( ed . ) , proceedings of the international workshop gross properties of nuclei and nuclear excitations iii , kleinwalsertal , austria , january 1318 , 1975 , p. 32, technische hochschule darmstadt aed - conf-75 - 009 - 000. h. nifenecker , r. babinet , c. signarbieux , j phys .c-5 33 ( 1972 ) 11 .nix , ann .22 ( 1972 ) 65 .park , w. greiner , w. scheid , phys .c 21 ( 1980 ) 958 .m. mirea , phys .c 63 ( 2001 ) 034603 .m. mirea , phys .c 78 ( 2008 ) 044618 . m. mirea , mod .lett . a 18 ( 2003 ) 1809 .m. mirea , int . j. mode 21 ( 2012 ) 1250035 .m. mirea phys .c 83 ( 2011 ) 054608 . m. mirea , phys . lett .b 717 ( 2012 ) 252 . s.e .koonin , j.r .nix , phys .c 13 ( 1976 ) 209 .j. blocki , h. flocard , nucl .phys . a 273 ( 1976 ) 45 .b. avez , c. simenel and ph .chomaz , phys .c 78 ( 2008 ) 044318 .s. ebata , t. nakatsukasa , t. inakura , k. yoshida , y. hashimoto and k. yabana , phys .c 82 ( 2010 ) 034306 .m. mirea , l. tassan - got , c. stephan , c.o .bacri , p. stoica , r.c .bobulescu , nucl .phys . a 735 ( 2004 ) 21 .nix , ann .22 ( 1972 ) 65 .swiatecki , s. bjornholm , phys .rep . 4 ( 1972 ) 325 .m. brack , j. damgaard , a.s .jensen , h.c .pauli , v.m .strutinsky , c.y .wong , rev .44 ( 1972 ) 320 .m. mirea , phys . rev .c 54 ( 1996 ) 302 .m. mirea , nucl .a 780 ( 2006 ) 13 .m. mirea , a. sandulescu , d.s .delion , proc .series a 12 ( 2011 ) 203 .m. mirea , a. sandulescu , d.s .delion , nucl .phys . a 870 ( 2011 ) 23 .m. mirea , a. sandulescu , d.s .delion , eur .j. a 48 ( 2012 ) 85 .m. mirea , rom . j. phys . 57( 2012 ) 372 .a. sandulescu , m. mirea , rom .rep . phys . in print .a. sandulescu , m. mirea , d.s .delion , epl 101 ( 2013 ) 62001 .m. mirea , d.s .delion , a. sandulescu , phys .c 81 ( 2010 ) 044317 .m. mirea , l. tassan - got , centr .j. phys . 9( 2011 ) 116 .micu , m. mirea , rom .j. phys . in print .m. mirea , d.s .delion , a. sandulescu , epl 85 ( 2009 ) 121001 . j.a .wheeler in niels bohr and the development of physics , edited by w. pauli , l. rosenfeld , and w. weisskopf pergamon press , london , 1955 , p. 163 .m. mirea , l. tassan - got , c. stephan , c.o .bacri and r.c .bobulescu , phys .c 76 ( 2007 ) 064608 .w. greiner , j.y .park , and w. scheid , nuclear molecules , world scientific , singapore , 1995 .j. randrup , p. moller , phys .106 ( 2011 ) 132503 .m. mirea , r.c .bobulescu , j. phys .g 37 ( 2010 ) 055106 .m. mirea , rom .rep . phys .63 ( 2011 ) 676 .a. gozdz , k. pomorski , m. brack , e. werner , nucl .a 442 ( 1985 ) 26 .a. gozdz , k. pomorski , m. brack , e. werner , nucl .phys . a 442 ( 1985 ) 50 .a. sandulescu , h. scutaru , ann .( 1987 ) 277 .g. audi , a.h .wapstra , c. thibault , nucl .phys . a 729 ( 2003 ) 337 .g. schutte , l. wilets , z. phys . a 286 ( 1978 ) 313 .hasse , rep .41 ( 1978 ) 1027 .m. rizea , n. carjan , nucl .phys . a 909 ( 2013 ) 50 .

<S> the time dependent equations of motion for the pair breaking effect were corroborated with a condition that fixes dynamically the number of particles on the two fission fragment . </S><S> the single particle level scheme was calculated with the woods - saxon superasymmetric two center shell model . </S><S> this model provides a continuous variation of the energies from one nucleus up to two separated fragments . </S><S> the dissipated energy resorts from the time dependent pairing equations . </S><S> a peculiar phenomenon was observed experimentally in cold fission : the odd partition yields are favored over the even ones . this odd - even effect for cold fission </S><S> was explained microscopically . </S>

1304.7259

let the notion be as in the main text . recall that we use the acronym slips for sl invariant polynomials in particular ,the set of all homogeneous slips of degrees @xmath4 is denoted by @xmath90 where @xmath169 is the vector of the dimensions of each of the @xmath5-qudits .let @xmath170 .then @xmath171 unless @xmath172 with @xmath57 a non - negative integer .recall that the center of @xmath173 is the set of scalar matrices @xmath174 .therefore , if @xmath175 then @xmath176 for any @xmath177 and any @xmath178 such that @xmath179 for some @xmath180 .the first equality follows from the @xmath89-invariance of @xmath181 and the second equality follows from its homogeneity .thus , each @xmath182 must divide @xmath4 .a state @xmath7 is said to be _ stable _ if its orbit @xmath37 is closed .the next proposition will explain this terminology .almost all states in @xmath26 are stable ( see for example @xcite ) . moreover , stable orbits have the following characterization @xcite : let @xmath7 .then , @xmath19 is stable if and only if there exists a ( non - normalized ) state @xmath183 with the property that all its @xmath5 local reduced density matrices are proportional to the identity matrix . therefore , for example , all states in a code space of an error correcting code that is capable of correcting one erasure error must be stable .the set of states that are not stable is also interesting and is of measure zero in @xmath26 .it contains the null cone .the null cone is the set of states on which all slips vanish .therefore , states and orbits in the null cone can not be characterized with slips . however , there are also states that are semi - stable in the sense that they are not stable , but they are also not in the null cone .the following lemma provides a characterization for such semi - stable states .[ lem2 ] let @xmath184 be a non - stable state , and suppose @xmath19 is not in the null cone .then , there exists a stable state @xmath185 and a state in the null cone @xmath186 such that @xmath187 moreover , denote by @xmath188 the stabilizer group of @xmath185 .then , there exists a group homomorphism @xmath189 such that @xmath190 this is a consequence of the ( generalized ) hilbert - mumford theorem ( c.f .the theorem asserts that there exists @xmath191 an algebraic group homomorphism such that the limit of @xmath192 as @xmath193 is a stable state @xmath194 .it is obvious that @xmath195 for all @xmath196 .the theorem follows from these assertions .we are now ready to introduce the main proposition of this subsection .[ prop2 ] let @xmath197 be two normalized stable states in @xmath26 .then , for all @xmath198 , and for all @xmath199 such that @xmath200 , @xmath201 if and only if there exists @xmath18 and a phase @xmath202 $ ] such that @xmath203 moreover , if @xmath204 and @xmath205 are not stable then the proposition still holds if @xmath204 and @xmath205 in eq.([slo ] ) are replaced with @xmath206 and @xmath207 , respectively . here for @xmath208 @xmath209 as in proposition [ lem2 ] . if @xmath210 and all of the @xmath211 are equal then the set of states with closed orbits ( under @xmath89 ) contains an open dense set of @xmath212 .every orbit contains a unique closed orbit in its closure @xcite . for a reductive algebraic group two closed orbits are equal if and only if all of the values of the invariant ploynomials agree on them ( we will use this fact in the proof of the proposition c.f .@xcite ) .note that if there exist @xmath198 and @xmath213 such that @xmath200 but @xmath214 , then clearly @xmath204 and @xmath205 are not slocc equivalent .we therefore assume without loss of generality that there exists @xmath215 such that @xmath200 and @xmath216 .fixing @xmath217 , and writing @xmath218 , our hypothesis implies that @xmath219 for all @xmath175 .now , for an arbitrary degree @xmath220 , let @xmath221 be some homogeneous slip of degree @xmath222 .then , @xmath223 is an slip of the same degree as @xmath224 .hence , the hypothesis implies that @xmath225 so if @xmath221 then @xmath226 with @xmath227 . at first glanceit looks like the @xmath228 root of unity @xmath229 depends on the polynomial @xmath230 .we will see that it is very constrained .we first show that it depends only on the degree @xmath222 .indeed , think of @xmath231 as a linear functional on @xmath232 . then @xmath233 which implies that @xmath234 is proportional to @xmath235 .thus , @xmath229 depends only on the degree @xmath222 .we now consider the group @xmath236 where @xmath237 .note that @xmath238 is also a reductive algebraic subgroup of @xmath239 so its closed orbits are separated by its invariant polynomials .furthermore every vector that has a closed orbit under @xmath89 has a closed orbit under @xmath240 .let @xmath241 be a set of homogeneous generators for the @xmath89-invariant polynomials on @xmath212 with degrees @xmath242 .then the @xmath238 invariant polynomials on @xmath243 are the linear combinations @xmath244 where the sum is taken over indices with @xmath245 divisible by @xmath4 .repeating the same arguments for the group @xmath238 ( note that @xmath217 is @xmath238invariant ) we conclude that if @xmath246 is a @xmath238 invariant polynomial on @xmath212 then @xmath247 .thus , since the orbits are closed , we find that @xmath248 with @xmath249 and @xmath18 .the upshot is @xmath250 . but@xmath251 which implies that @xmath252 .thus @xmath253 as was to be proved .in the proposition below we are using the same notations that were introduced above eq . ( 7 ) of the main text .let @xmath79 and consider a bipartite cut @xmath254 .then , the function @xmath255\ ] ] is an sl - invariant polynomial of degree @xmath81 .let @xmath18 .we need to show that @xmath256 , where in this proof we removed from @xmath257 the subscript and superscript for the simplicity of the exposition . we can decompose @xmath258 such that @xmath259 is acting on the space @xmath260 and @xmath261 is acting on @xmath262 .further , denote by @xmath263 and @xmath264 the @xmath265 and @xmath266 square matrices whose elements are given by @xmath267 and @xmath268 , respectively . with this notations , replacing @xmath19 with @xmath17 , is equivalent to replacing @xmath45 with @xmath269 .we therefore have @xmath270\\ & = { \rm tr}\left[\left(g_{a}^{t}ug_aag_{b}^{t}vg_{b}a^t\right)^{\ell}\right]\end{aligned}\ ] ] where we have used the invariance of the trace under cyclic permutation . to complete the proof, we show now that @xmath271 indeed , @xmath272 where we have used the invariance of the bilinear form @xmath273 under the action of @xmath274 . in the same way, one can prove that @xmath275 .this completes the proof . as discussed in the main text ,the above simple slips are not linearly independent . for odd number of qubits , there are no homogenous slips with degree 2 , and the lowest degree in this case is 4 ( e.g. @xcite ) . in this casethe number of homogenous slips of degree 4 is @xmath276 ( see below ) , whereas the number of our polynomials @xmath277 of degree 4 is equal to the number of inequivalent bipartite cuts @xmath59 which is given by @xmath278 where we assumed that @xmath5 is odd .note that for any bipartite cut @xmath59 there exists a bipartite cut of the form @xmath279 which leads to the exact same polynomials .thus , @xmath57 in the sum does not exceed @xmath280 .hence , our @xmath281 polynomials of degree 4 , @xmath277 , must be linearly dependent . nonetheless , the over - complete set of @xmath281 polynomials @xmath282 span the set of all slips of degree 4 .the purpose of this short subsection is to give a very brief introduction to the schur - weyl duality , in the context of the work presented in this paper . for a more detailed analysiswe refer the reader to a more standard book on representation theory of finite groups ( c.f .@xcite ) .we consider the group @xmath141 acting on @xmath283 by the usual tensor action .the representations of @xmath141 appearing are determined by their highest weight which is an @xmath148-tuple @xmath284 with @xmath285 .we denote by @xmath286 a fixed representation with highest weight @xmath152 .then @xmath286 is equovalent with @xmath287 if and only if @xmath288 with @xmath289 .also @xmath286 is a constituent of @xmath283 if an only if it is equivalent to @xmath287 with @xmath290 and @xmath291 .we will say that @xmath292 with @xmath293 is dominant .also if @xmath152 is dominant then we will call it a partition of @xmath4 if @xmath294 and @xmath295 .obviously , we can make the parametrization unambiguous by using @xmath152 with @xmath296 .but we note that if @xmath297 are partitions of @xmath4 and if @xmath286 and @xmath287 are equivalent then @xmath298 . in the literature, @xmath299 denotes a partition of @xmath4 and if @xmath222 denotes the last index for which @xmath300 then the indices @xmath301 are dropped .these partitions parametrize the equivalence classes represenations of the symmetric group @xmath111 .we fix a fixed realization of the class that is parametrized by @xmath152 and denote it by @xmath302 . if we need to keep track of @xmath4 and @xmath5 we will write @xmath303 and @xmath304 .we note that the group @xmath111 acts on @xmath283 by permuting the indices .this action commutes with that of @xmath141 .schur - weyl duality says that as a representation of @xmath305 the space @xmath283 decomposes into @xmath306 with the sum over all partitions @xmath307 of @xmath4 .we note that if @xmath308 that this is a sum over all partitions of @xmath4 ( after we have dropped all of the entries that are @xmath133 ) .this decomposition allows us to describe the projection of @xmath309 into @xmath310 using either a projection gotten by integrating the character of @xmath286 restricted to @xmath311 or summing the character of @xmath302 over @xmath111 . in other words ( @xmath312 is the unique invariant probability measure on @xmath311 ) @xmath313 an example of this is if @xmath143 with @xmath149 a positive integer and @xmath314 ( @xmath148 ones ) then @xmath315 is the projectiion onto the @xmath141-invariants in @xmath283 .formula ( 10 ) ( of the main text ) for @xmath316 necessitates @xmath154 and @xmath168 .thus , in this case @xmath130 is acting on @xmath317 and is given by its action on basis elements as : @xmath318 this implies that formula ( 9 ) says for the case of @xmath319 qubits @xmath320@xmath321 here the first arrow is the transpose as is the last .also @xmath322 , @xmath323 is the size of @xmath324 and @xmath325 with @xmath326 if @xmath327 and @xmath328 if @xmath329 .this yields the well known formula for the degree @xmath330 invariant in any even number of qubits .formula ( 10 ) for @xmath114 plays a role only for @xmath154 and @xmath153 or @xmath331 and @xmath168 . in the first case formula ( 10 )implies that @xmath130 is acting on @xmath332 and is given by its action on basis elements as : @xmath156 we leave it to the reader to write out formula ( 9 ) to get the degree @xmath157 invariant for an arbitrary number of @xmath157-dimensional qudits . in the case of @xmath331 and @xmath168 then the pertinent representation , @xmath152 , of @xmath333 corresponds to the partition @xmath334 the dimension of @xmath152 is @xmath330 .the conjugacy classes in @xmath333 are given by @xmath335 the cycles of length @xmath157 ( e.g. @xmath336 ) , @xmath337 the cycles of length @xmath338 , @xmath339 the products of disjoint transpositions ( e.g. @xmath340 ) , @xmath341 the transpositions,@xmath342 the identity element .we have @xmath343{c}% 0\text { if } \sigma\in c_{4}\\ -1\text { if } \sigma\in c_{3,1}\\ 2\text { if } \sigma\in c_{2,2}\\ 0\text { if } \sigma\in c_{2,1,1}\\ 2\text { if } \sigma = i \end{array } \right . .\ ] ] we note that @xmath344,@xmath345 .formula ( 10 ) implies that @xmath130 is acting on @xmath346 and is given by its action on basis elements as : @xmath347 we leave it to the reader to write out formula ( 9 ) for @xmath5 qubits .the reference that we will use for this section is chapter 9 in @xcite . in this sectionwe will consider the case of @xmath5qudits all of the same dimension , say , @xmath348 thus we are looking at @xmath349 and @xmath350 , @xmath5 copies .we analyse the space @xmath351 as the @xmath352invariants in @xmath353 .using the map @xmath354 @xmath355 defined in ( 8) we are considering@xmath356 with @xmath111 acting by the @xmath5fold tensor product of the action on @xmath131 which is the action of permuting the tensor factors .@xmath357 is @xmath133 unless @xmath358 for some @xmath359 . as we observed in subsection ( iv )the @xmath141invariants in @xmath131 define the representation @xmath302 of @xmath111 with @xmath360 .thus as far as @xmath111 is concerned @xmath361 thus if @xmath358 and @xmath152 are as above then@xmath362 this says that if @xmath146 is the character of @xmath302 then@xmath363 we recall that characters are constant on conjugacy classes ( the sets of the form @xmath364 ) .the conjugacy classes of @xmath111 can be described by partitions using the cycle decomposition of an permutation .the ( standard ) algorithm for this is start with a permutation @xmath365 .consider @xmath366 .then there must be a first @xmath367 such that @xmath368 but @xmath369 .set @xmath370 for @xmath371 then on the set @xmath372 the permutation @xmath365 acts by the @xmath367cycle @xmath373 . if @xmath374 then we are done .otherwise choose the smallest element of @xmath375 that is not in @xmath376 and denote it @xmath377 then @xmath378 for the first time for @xmath379 .set @xmath380 .we now have a second cycle whose entries are disjoint from @xmath376,@xmath381 action on the set @xmath382 . if @xmath383 then we are done otherwise do the same procedure on @xmath384 , etc . finally put the @xmath385 in decreasing order .this yields a partition of @xmath4 corresponding to each @xmath386 .it is not hard to see that @xmath365 and @xmath299 are conjugate if and only if they have the same cycle decomposition .if @xmath299 is a partition of @xmath4 then let @xmath387 denote the corresponding conjugacy class and let @xmath388 be the characterisitic function of the subset @xmath387 of @xmath111 .that is , @xmath389{c}% 1\text { if } \tau\in c_{\mu}\\ 0\text { otherwise}% \end{array } \right . .\] ] then the functions @xmath388 form a basis of the space of central functions on @xmath111 .thus if @xmath152 is a partitian of @xmath4 then@xmath390 where the sum is over the partitions of @xmath4 .the numbers @xmath391 arranged into a table give the character table of @xmath111 .however the conjugacy class partitions are usually given in the way we will describe them next . let @xmath392 be a conjugacy clasee then we reorganize the partition in _ increasing _ order and lay it out as @xmath393 where @xmath394 of the cycles are of length @xmath395 ( fixed points ) , @xmath396 are of length @xmath330 , ..., @xmath397 are of length @xmath398 .thus the permutation @xmath399 has @xmath400 which also corresponds to @xmath401 we will use the two notations for partitions interchangably .the following is standard and easily proved if @xmath299 is given as @xmath393 then @xmath402 here is the result that we have been aiming at if @xmath403 then @xmath404 otherwise set @xmath405 with @xmath148 entries .then@xmath406 the asymptotic formula follows from this. we will give it in full generality and at the end of this subsection we will explain how the formulas look for qubits .let @xmath407 be an irreducible representation of a finite group , @xmath408 . if @xmath409 then the schwarz inequality implies that@xmath410 with equality if and only if @xmath411 for some norm - one scalar @xmath412 .the subgroup of all elements of @xmath408 such that @xmath413 is a multiple of the identity is a normal subgroup . if @xmath164 then the only non - trivial normal subgroup of @xmath111 is the alternating group .this implies that if @xmath164 and if @xmath152 is a partition of @xmath4 such that the representation @xmath302 is not a one dimensional representation then@xmath414 if @xmath365 is not the identity in @xmath111 . if @xmath114 and @xmath415 then there is a klein @xmath157group of elements that map to the identity .we have if @xmath164 then we have the following asymptotic formula when @xmath358 and @xmath405@xmath416 if @xmath114 and @xmath415 ( so @xmath417 then then the formula is@xmath418 if @xmath164 then we note that if @xmath419 is the partition correspnding to the identity in @xmath111 then @xmath420 for any other partition , @xmath421 .thus@xmath422 since @xmath423 and @xmath424 , this proves the result if @xmath164 . if @xmath114 and @xmath415 then @xmath425 and there are @xmath157 elements that yield the value @xmath426 .we note that ( c.f .@xcite corollary 9.1.5 ) if @xmath427 then @xmath428 thus if @xmath168 and @xmath429 then @xmath430 the @xmath149th catalan number . for @xmath358 and @xmath405 ( @xmath148factors )then the dimension should be thought of as a mutinomial generalization of the catalan number .our expression of @xmath130 in eq .( 10 ) of the main text involves in general many terms and therefore is somewhat cumbersome . hence , in this section we show an alternative way to express _ all _ slips of degree @xmath4 in the important case of @xmath5-qubits .this new technique has the advantage that it is computationally efficient for small degrees .we have already seen that for the qubit case we can express _ some _ of the slips very elegantly with eq.(10 ) . however , that construction is only partial as it does not consists of all slips .hence , as an application to our new expression in proposition [ sl2 ] ( see below ) , we give at the end of this section an example of the unique homogeneous slip of degree 6 in 5 qubits ; this slip can not be written in the form of eq.(7 ) of the main text . in this section @xmath212denotes @xmath431 and @xmath432 @xmath5copies .we set@xmath433 we consider @xmath434 .then the character of the invariants for @xmath435 as an @xmath436 representation restricted to @xmath437{cc}% q & 0\\ 0 & q^{-1}% \end{array } \right ] |q\in\mathbb{c}-\{0\}\right\}\ ] ] is @xmath438 here the action of @xmath439 is the tensor product action on the factors .we note that the coefficient of @xmath440 is the same as @xmath441 in this expression .we also note that @xmath442 if @xmath4 is odd .thus we may assume that @xmath4 is even .we observe that the character of the @xmath443spin representation of @xmath444 restricted to @xmath445 is @xmath446 this implies that if @xmath429then@xmath447 it also implies that if @xmath257 @xmath448 is a homogeneous polynomial of degree @xmath449 in @xmath450 qubits that is invariant under @xmath451 then if we think of @xmath257 as a polynomial in two copies of @xmath5 qubits it must have a non - zero component in @xmath452 . for simplicitywe will concentrate on the case when @xmath5 is even ( so @xmath450 is odd ) .this hypothesis implies that @xmath212 there exists a ( unique up up to scalar ) @xmath439invariant symmetric complex bilinear form @xmath453 , @xmath454 on @xmath212 .let @xmath455 with @xmath456 be a an orthonormal basis with respect to this form .we think of the first copy , @xmath457 , as having elements @xmath458 and the second , @xmath459 , @xmath460 .then the action of the lie algebra first @xmath461factor in @xmath451 ( that is @xmath462 ) is given by@xmath463 with commutation relations@xmath464=h,\;\;[h , x]=2x,\;\;[h , y]=-2y.\ ] ] we note that @xmath465{cc}% 0 & 1\\ 0 & 0 \end{array } \right ] , y\longleftrightarrow\left [ \begin{array } [ c]{cc}% 0 & 0\\ 1 & 0 \end{array } \right ] , h\longleftrightarrow\left [ \begin{array } [ c]{cc}% 1 & 0\\ 0 & -1 \end{array } \right ] $ ] so the casimir operator is given by @xmath466 since @xmath467 we have@xmath468 we replace @xmath469 with @xmath470 and note that this operator commutes with the action of @xmath451 .we note that if we use the same formulas for @xmath471,@xmath472 and @xmath473 on the polynomials in @xmath474 with @xmath0 an arbitrary positive integer then we have the same commutation relations .let @xmath475 be the space of all polynomials that are homogeneous of degree @xmath246 in @xmath476 and degree @xmath57 in @xmath477 let @xmath478 be the corresponding operators ( as above ) .the following is obvious .let @xmath479 and let @xmath480 be given by @xmath481 with @xmath482 zeros .then if @xmath483 or @xmath473 then@xmath484 using the classification of the irreducible finite dimensional representations of @xmath461 we see that on the spin @xmath485 ( i.e. @xmath486 dimensional representation ) , @xmath487 , the operator @xmath488 acts by @xmath489 . thus in a representation , @xmath490 , for @xmath491 that is a sum with multiplicity of @xmath492 then the projection of @xmath490 onto the invariants is given by@xmath493 using this , taking @xmath490 as above we get the main result of this section : [ sl2 ] if @xmath494 then the space of @xmath461 invariants is @xmath495 example the invariant in @xmath496 qubits of degree @xmath497 .in @xmath157 qubits we take the basis @xmath498 with @xmath499 that is , the bell basis in @xmath330 qubits .these bases are orthonormal with respect to both the hilbert space inner product and ( suitably normalized ) the respectively @xmath500 or @xmath501-invariant complex bilinear , symmetric form .we order the basis so that @xmath502 @xmath503 , @xmath504. let @xmath505,@xmath506,$ ] @xmath507 $ ] and @xmath508 .we order the basis so that @xmath502 @xmath509 , @xmath504 and consider @xmath257 to be @xmath510 and @xmath511 to be @xmath512 .we set @xmath513 then this is an invariant polynomial for @xmath501 acting on two copies of @xmath157 qubits of bidegree @xmath514 .we note that @xmath515 set @xmath516 then @xmath517 with @xmath518 we have the following is up to scalar the unique degree @xmath497 invariant for @xmath496 qubits.@xmath519 we know that the space of such invariants is dimension 1 and since the retriction of the element in the statement is non - zero by the above it must be the invaraint .the purpose of this section is to give a listing of mathematica code to calculate @xmath520 as a function of @xmath5 if @xmath4 is fixed ( here we look only at qubits ) . in the body of the letter we give two examples . using the code below and taking @xmath521 we have @xmath522 @xmath523 over @xmath524 the calculation was almost instantaneous . in our codebelow we will output the expression in powers of @xmath57 .we use the notation in subsection v of this appendix .we first explain the method used in the code .if @xmath429 then the code implements ( the sum below is on the partitions of @xmath4)@xmath525 one can read this off of tables for small values of @xmath4 or use a standard mathematical package .however , the included code is designed to efficiently do this specific calculation .it should be easily converted to a lower level language ( such as c or c++ ) .however , the output becomes immense for @xmath4 larger than say @xmath526 ( a calculation that takes about 1 minute on a 4 year old pc ) .the code below has two main functions .mult[k , n ] which calculates @xmath520 for specific values of @xmath4 and @xmath5 .multq[k ] which calculates @xmath527 as a function of @xmath57 with @xmath4 fixed .the algorithm for calculating @xmath528 uses theorem 9.1.4 in @xcite with the @xmath5 in that theorem equal to @xmath529and the fixed point calculation in the formula given by the function fp[p , q , x ] ( here @xmath530 corresponds to a conjugacy class ) .the number @xmath531 is calculated in the function cst[x ] .the main weakness in the code is the generation of the partitions prt[k ] . since the number of partitions of is @xmath532this will be a bottleneck no matter what one does to streamline the code .the mathematica code : .... p[n _ , k _ ] : = module[{l , m = { } , s , i , j } , if[k > n , return[p[n , n ] ] ] ; if[k = = 0 , return [ { { } } ] ] ; if[k = = 1 , return[{table[1 , { i , 1 , n } ] } ] ] ; for[i = 1 , i < = k , i++ , l = p[n - i , i ] ; s = { } ; for [ j = 1 , j < = length[l ] , j++ , s = append[s , prepend[l[[j ] ] , i ] ] ] ; m = union[m , s ] ] ; m ]g. gour and n. r. wallach , new j. phys .* 13 * 073013 ( 2011 ) .

<S> we provide a systematic classification of multiparticle entanglement in terms of equivalence classes of states under stochastic local operations and classical communication ( slocc ) . </S><S> we show that such an slocc equivalency class of states is characterized by ratios of homogenous polynomials that are invariant under local action of the special linear group . </S><S> we then construct the complete set of all such sl - invariant polynomials ( slips ) . </S><S> our construction is based on schur - weyl duality and applies to any number of qudits in all ( finite ) dimensions . </S><S> in addition , we provide an elegant formula for the dimension of the homogenous slips space of a fixed degree as a function of the number of qudits . </S><S> the expressions for the slips involve in general many terms , but for the case of qubits we also provide much simpler expressions . </S><S> [ theorem]lemma [ theorem]corollary [ theorem]conjecture [ theorem]proposition multi - particle entanglement is an essential resource for a variety of quantum information processing tasks . </S><S> these include conventional @xcite and measurement - based quantum computation @xcite , quantum error correction schemes @xcite , quantum secret sharing @xcite , quantum simulations @xcite , and in principle in any task involving entangled many - body quantum systems @xcite . </S><S> these applications to quantum information , along with its intriguing properties and potential applications to condensed matter physics @xcite , sparked an enormous amount of literature dedicated to the classification of multipartite entanglement . nevertheless , our current understanding of multipartite entanglement is very limited , and besides few cases involving small systems , the resource theory of multipartite entanglement is still in its infancy @xcite . </S><S> entanglement is a non - local resource with which it is possible to overcome the limitations imposed by local operations . </S><S> therefore , multipartite states are classified according to their inter - convertibility under stochastic local operations and classical communication ( slocc ) @xcite . </S><S> that is , two states belong to the same entanglement class ( or slocc equivalent class ) if it is possible to reversibly convert one state to the other with non - zero probability using only locc . while in three qubits there are only 6 inequivalent slocc classes , in four ( or more ) qubits there are already uncountable number of inequivalent slocc classes @xcite . </S><S> this is a simple indication that the complexity in characterizing slocc classes grows rapidly as the number of particles increases . </S><S> initially , much of the literature focused on classification of slocc classes in terms of homogeneous polynomial functions of the coefficients of the pure multi - qudit state in question . </S><S> if these polynomials on @xmath0-qudit systems are also invariant under @xmath1 ( here @xmath2 is the set of all @xmath3 complex matrices with determinant 1 ) then their absolute values are measures of genuine multipartite entanglement @xcite . </S><S> these entanglement measures have many interesting and desirable properties , including initial - state - independence behavior under local noise @xcite . for pure two qubits and three qubits systems , the concurrence @xcite and </S><S> the 3-tangle @xcite are the unique polynomials of this kind . for four qubits , </S><S> however , there are infinitely many such invariants and the set of all of them is generated by four invariants @xcite . beyond that , and despite the extensive literature , very little is known about the set of all such sl - invariant polynomials ( slips ) except for few techniques that were used to construct some of the slips for multi - qubit states @xcite . in this letter we show that almost all slocc equivalent classes , including the dense set of stable slocc classes , can be distinguished by ratios of homogeneous slips of the same degree . </S><S> we then find an algorithm to construct _ all _ of the slips . </S><S> our main idea is to look at homogenous polynomials of a fixed degree @xmath4 , and view them as vectors in a hilbert space consisting of @xmath4-copies of the original hilbert space . </S><S> this identification enables us to use the schur - weyl duality and other techniques from representation and invariant theory to construct the full set of homogeneous slips of degree @xmath4 . </S><S> our technique can be applied to any number of qudits in all dimensions . for the case of qubits </S><S> , we also find an elegant way to express many of the slips . </S><S> we consider here the hilbert space of @xmath5 qudits which we denote by : @xmath6 local ( generalized ) measurements on each of the @xmath5 qudits , can transform probabilistically an initial pure state @xmath7 to some other pure state @xmath8 , where @xmath9 are @xmath10 matrices . </S><S> if this transformation is reversible then all the matrices @xmath11 can be taken to be invertible @xcite , and @xmath12 and @xmath13 are said to be slocc equivalent . </S><S> we are interested here in characterizing classes of slocc equivalent states . up to normalization </S><S> , an invertible slocc map @xmath14 can be described as an element of the group @xmath15 where @xmath16 is the group consisting of all @xmath10 complex invertible matrices with determinant 1 . </S><S> this last condition implies that the states @xmath17 for @xmath18 are in general not normalized even if @xmath19 is normalized . </S><S> the orbit @xmath20 is therefore consisting of non - normalized states in the slocc equivalent class of @xmath19 . working with non - normalized states </S><S> enables us to characterize the distinct orbits @xmath21 with slips , while the distinct slocc classes can be characterized by ratios of the slips . a polynomial @xmath22 that satisfies @xmath23 is called @xmath24-invariant polynomial ( slip ) . </S><S> the set of all slips form a vector space over @xmath25 . </S><S> we can always choose the basis of this polynomial space to consist of _ homogeneous _ slips acting on @xmath26 linearly . ] . </S><S> for this reason , we will focus on homogeneous slips of some degree @xmath27 . </S><S> the dimension of the space of homogeneous polynomials of fixed degree @xmath4 is finite , although we will see that it grows exponentially with @xmath5 . in the supplemental material </S><S> we show that the degree @xmath4 must be divisible by the least common multiple of the integers @xmath28 . </S><S> this implies for example that for qubits there are only slips of even degree . </S><S> we show now that slips can be used to determine whether two states in @xmath26 belong to the same slocc class . indeed , suppose that @xmath29 are two ( normalized ) states belonging to the same slocc class . </S><S> that is , there exists a phase @xmath30 and @xmath18 such that @xmath31 is a homogenous slip of degree @xmath4 . </S><S> then , @xmath32 suppose now that @xmath33 is another homogeneous slip of the _ same _ degree @xmath4 . </S><S> since it also satisfies the relation , we get @xmath34 assuming @xmath35 . </S><S> thus , if @xmath19 and @xmath36 belong to the same slocc class they must satisfy for any pair of two homogeneous slips of the same degree ( this direction was pointed out recently in @xcite ) . here we show that the converse also holds for almost all states in @xmath26 ( see details of the proof in the supplementary material ) : let @xmath29 be two stable states ( i.e. states @xmath19 whose orbits @xmath37 are closed ) . </S><S> then , there exists @xmath30 and @xmath18 such that holds if and only if holds for all homogeneous slips of degree @xmath4 with @xmath35 . </S><S> the set of all stable states is open and dense in @xmath26 . </S><S> it contains almost all states in @xmath26 including many non - generic states . </S><S> in particular , an orbit @xmath38 is stable if and only if it contains a state with the property that the reduced density matrix on each single qudit ( after tracing out the other @xmath39 qudits ) is proportional to the identity @xcite . </S><S> thus , many interesting and physically relevant states are stable . </S><S> this includes all quantum error - correcting code states , all ghz states , all cluster states , and in general all graph states . </S><S> hence , our construction below of all slips provides an almost - complete classification of @xmath26 into slocc equivalent classes . </S><S> slips involve in general cumbersome expressions . </S><S> however , for the case of @xmath5 qubits there exists a large class of polynomials that can be expressed very elegantly . </S><S> the main reason for that is the existence of the matrix @xmath40 for which any @xmath41 matrix @xmath42 satisfies @xmath43 where @xmath44 denotes the transposed matrix of @xmath45 . in higher dimensions , such an invariant matrix @xmath46 that satisfies the above equation for </S><S> all @xmath47 does not exist . </S><S> thus , before introducing our main results , we first show that this property of @xmath46 leads to a large class of slips in the qubit case where @xmath48 for all @xmath49 . </S><S> the above property of @xmath46 implies that for any two states @xmath29 , and any @xmath18 , @xmath50 where the bilinear form @xmath51 is defined by @xmath52 here @xmath46 appears @xmath5-times , and the vector @xmath53 is the complex conjugate of @xmath19 when it is expressed in a fixed basis such that each @xmath46 has the form ( [ j ] ) . </S><S> therefore , @xmath54 is an homogenous slip of degree 2 . </S><S> note however that for odd @xmath5 , @xmath55 since in that case @xmath56 is a skew - symmetric matrix . for any choice of @xmath57 qubits in @xmath26 with @xmath58 , we associate a bipartite cut @xmath59 between the chosen @xmath57 qubits,@xmath60 , and the remaining @xmath61 qubits , @xmath62 . for a given @xmath57 </S><S> there are @xmath63 bipartite cuts of the form @xmath64 . </S><S> now , fix a bipartite cut @xmath59 . </S><S> any state @xmath65 can be written as @xmath66 and @xmath67 are orthonormal bases of @xmath68 and @xmath69 , respectively . </S><S> for any such @xmath19 , we denote by @xmath45 the @xmath70 matrix whose elements are @xmath71 . </S><S> furthermore , we denote by @xmath72 the @xmath73 matrix whose elements are @xmath74 , and by @xmath75 the @xmath76 matrix whose elements are @xmath77 , where @xmath78 is the bilinear form defined in eq.([blin ] ) . with this notation </S><S> we get ( see more details in the supplementary material ) that for any @xmath79 the function @xmath80\ ] ] is an slip of degree @xmath81 . </S><S> the function @xmath82 is invariant under a change of bases in @xmath68 and @xmath69 . </S><S> moreover , since the form @xmath83 is symmetric for even @xmath57 , if @xmath57 is even we can choose the basis @xmath84 such that the matrix @xmath85 . </S><S> similarly , if @xmath61 is even then we can choose the basis @xmath67 such that the matrix @xmath75 is the identity . </S><S> the slips defined above are not all independent . </S><S> different bipartite cuts can lead to the same polynomial . </S><S> for example , in the case of even number of qubits , all the polynomials @xmath86 , corresponding to @xmath63 bipartite cuts @xmath59 , are the same ( up to a constant factor ) since there is only one slip of degree 2 given by @xmath54 @xcite . </S><S> it is therefore natural to ask if all slips can be written as linear combinations of @xmath87 . </S><S> it is true for degrees 2 and 4 , but already in 5 qubits there exists ( see below ) a unique polynomial of degree 6 , whereas our polynomials @xmath88 can not be of degree 6 if @xmath5 is odd . </S><S> we therefore need a more systematic way to construct all slips . </S><S> we are now ready to introduce our main technique to compute the @xmath89-invariants in the space of polynomials on @xmath26 of degree @xmath4 . </S><S> we denote the set of all these slips by @xmath90 where @xmath91 is the vector of the dimensions of the @xmath5-qudits . </S><S> observe that any homogeneous polynomial on @xmath26 of degree @xmath4 can be written as an inner product between some vector @xmath92 and @xmath4-copies of a vector @xmath93 : @xmath94 hence , any homogeneous polynomial on @xmath26 of degree @xmath4 corresponds to some ( not necessarily unique ) vector @xmath95 . note that </S><S> if we also have @xmath96 for all @xmath18 , then the polynomial @xmath97 is @xmath89-invariant . </S><S> denoting by @xmath98 all such @xmath89-fixed vectors of @xmath99 , we conclude : @xmath100 note however that in general distinct vectors in @xmath99 can lead to the _ same _ polynomial and , in particular , we will see below that @xmath101 is much larger than @xmath102 . </S><S> we now show how this view of slips leads to an algorithm for finding all of them . </S><S> we first write out @xmath103 in terms of the standard basis in each tensor slot . that is,@xmath104 and each @xmath105 is an element in the computational basis of @xmath26 , </S><S> i.e. it is of the form @xmath106 with @xmath107 for @xmath108 . </S><S> writing @xmath109 relative to the same computational basis of @xmath26 we have@xmath110 this shows that if we symmetrize the elements relative to the action of @xmath111 ( permuting the factors ) then the symmetric elements yield @xmath90 unambiguously . in what follows </S><S> , we will view @xmath99 as a representation space of @xmath89 with the homomorphism @xmath112 for any @xmath18 . </S><S> , for @xmath113 and @xmath114 . the hilbert space of the three qudits is denoted by @xmath115 , where the three colours correspond to the three dimensions @xmath116 , @xmath117 , and @xmath118 . </S><S> ] our task is to find @xmath119 . </S><S> the key observation is that the vector space @xmath99 is isomorphic to the vector space @xmath120 the isomorphism map ( which is a ` transpose'-type map ) maps the computational basis element @xmath121 to the basis element @xmath122 , where @xmath123 and for each @xmath124 , @xmath125 is the standard basis of @xmath126 . </S><S> this map is an intertwining operator ( see fig . </S><S> 1 ) . </S><S> we will denote it by @xmath127 quite remarkably , under this intertwining map , the problem of finding @xmath119 is reduced to finding @xmath128 for each @xmath108 . to see it , we use the same notation for the inverse map from @xmath129 to @xmath99 . </S><S> let @xmath130 denote the orthogonal projection of @xmath131 onto @xmath132 ( which is @xmath133 unless @xmath4 is divisible by @xmath134 the map @xmath135 is the orthogonal projection from @xmath99 onto @xmath136 . </S><S> this is a simple consequence of the fact that a state being invariant under @xmath137 has to be in the span of @xmath138 , where @xmath139 are elements of @xmath140 . </S><S> it is therefore left to compute @xmath130 . </S><S> this is a classic problem in representation theory and can be solved using the schur - weyl duality . </S><S> the schur - weyl duality relates the irreducible representations ( irreps ) of @xmath141 ( when acting on @xmath131 ) and the symmetric group on @xmath4 elements , @xmath142 ( with the natural action on @xmath131 ) . </S><S> recall that @xmath4 must be of the form @xmath143 for some @xmath144 in order to have @xmath145 let @xmath146 be the character of @xmath111 corresponding to the partition of @xmath4 given by @xmath147 ( @xmath148 @xmath149 s ) . then </S><S> ( see more details in the supplementary material ) , @xmath150 where @xmath151 is the dimension of the irrep corresponding to the partition @xmath152 . </S><S> equations and determine the projection of @xmath99 onto @xmath119 , and thereby via eq . provide the complete set of _ all _ slips of degree @xmath4 . </S><S> we now give a simple example , illustrating how to use eq.([m2 ] ) to construct slips . </S><S> more examples with more details can be found in the supplementary material . </S><S> consider a system consisting of @xmath5 qudits each of dimension @xmath153 . the smallest degree possible in this case is @xmath114 for which @xmath154 . </S><S> thus , in this case @xmath130 is a projection on @xmath155 . from eq . </S><S> we can define @xmath130 in terms of its action on basis elements as @xmath156 substituting this @xmath130 into eq.([m1 ] ) gives the degree @xmath157 invariant for an arbitrary number of @xmath5 qudits each of dimension four . </S><S> the number of terms in the sum of is @xmath158 , which already for @xmath159 gives @xmath160 terms . </S><S> therefore , using fundamental concepts from invariant theory , we find ( see supplementary material ) an alternative way to express all the slips of @xmath5 qubits . this alternative way ( which only applies for the qubit case ) is slightly more computationally efficient when @xmath4 is small . as an example we give an explicit formula for the unique slip of degree 6 in five qubits ( which can not be written in the form ) . </S><S> our techniques also enable us to write down a simple algorithm to compute the dimension , @xmath161 , of the space of slips of degree @xmath4 in @xmath5 qubits . </S><S> we get for example : @xmath162 in supplementary material we attach a program in mathematica that computes @xmath161 for large @xmath5 and @xmath4 . </S><S> moreover , in the more general case of @xmath5 _ qudits _ , with @xmath163 , the dimension @xmath161 of the space of slips of degree @xmath143 satisfies ( for @xmath164 ) @xmath165 where @xmath166 is the generalization of the @xmath167 catalan number ( see the supplementary material for more details ) . for @xmath114 ( which is only possible in the qubit @xmath168 case ) the limit above is 1/6 . </S><S> the distinction between @xmath114 and @xmath164 follows from the fact that the alternating group on @xmath5 letters is simple ( i.e. without non - trivial normal subgroups ) if and only if @xmath164 . </S><S> this formula indicates that the dimension of the space of slips grows exponentially both in @xmath5 and in @xmath4 . in conclusion </S><S> , we showed that slips on @xmath5 qudits can be used to classify almost all slocc classes . </S><S> we constructed a large class of elegant slips for the case of @xmath5 qubits and then introduced a general technique to construct all silps in all dimensions . instead of fixing the number of qudits and then trying to calculate all slips in any degree </S><S> , we fixed the degree @xmath4 and then where able to calculate all slips of that degree in any number of qudits . </S><S> this approach also enabled us to calculate the dimensions of the space of homogeneous slips of a given degree . _ </S><S> acknowledgments: _ the authors are grateful for markus grassl for many fruitful comments on the first version of the paper </S><S> . g.g . </S><S> research is partially supported by nserc and by the air force office of scientific research as part of the transformational computing in aerospace science and engineering initiative under grant fa9550 - 12 - 1 - 0046 . </S><S> n.w . </S><S> research is partially supported by an nsf summer grant . </S><S> m. a. nielsen , i. l. chuang , quantum computation and quantum information ( cambridge univ . </S><S> press , cambridge , 2000 ) . </S><S> h.j . </S><S> briegel and r. raussendorf , phys . </S><S> rev . </S><S> lett . * </S><S> 86 * , 910 ( 2001 ) ; r. raussendorf and h. j. briegel , _ ibid_. * 86 * , 5188 ( 2001 ) . </S><S> m. hillery , v. buzek , and a. berthiaume , phys . </S><S> rev . </S><S> a * 59 * , 1829 ( 1999 ) . </S><S> s. lloyd , science * 273 * , 1073 ( 1996 ) . </S><S> r. horodecki , p. horodecki , m. horodecki , and k. horodecki , rev . </S><S> mod . </S><S> phys . </S><S> * 81 * , 865 ( 2009 ) . </S><S> m. b. plenio and s. virmani , quant . </S><S> inf . comp . </S><S> * 7 * , 1 ( 2007 ) . </S><S> o. guhne , g. toth , and h. j. briegel , new j. phys . </S><S> * 7 * , 229 ( 2005 ) . </S><S> m.walter , b. doran , d. gross , m. christandl , eprint : quant - ph/1208.0365 . </S><S> g. gour and n. r. wallach , new j. phys . </S><S> * 13 * 073013 ( 2011 ) . </S><S> w. dur , g. vidal , and j. i. cirac , phys . </S><S> rev . </S><S> a * 62 * , 062314 ( 2000 ) . </S><S> g. gour and n. r. wallach , jmp * 51 * , 112201 ( 2010 ) . </S><S> f. verstraete , j. dehaene , b. de moor , phys . </S><S> rev . </S><S> a * 68 * , 052112 ( 2003 ) . </S><S> v. gheorghiu , g. gour , phys . </S><S> rev . </S><S> a * 86 * , 050302 ( r ) ( 2012 ) . </S><S> g. gour , 105 , 190504 ( 2010 ) . </S><S> w. k. wootters , phys . </S><S> rev . </S><S> lett . * 80 * , 2245 ( 1998 ) . </S><S> v. coffman , j. kundu , and w. k. wootters , * 61 * , 052306 ( 2000 ) . </S><S> j .- g . </S><S> luque and j .- y . </S><S> thibon , phys . </S><S> rev . </S><S> a * 67 * , 042303 ( 2003 ) . </S><S> m.s . </S><S> leifer , n. linden , and a. winter , phys . </S><S> rev . </S><S> a 69 , 052304 ( 2004 ) . </S><S> a. osterloh and j. siewert , phys . </S><S> rev . </S><S> a 72 , 012337 ( 2005 ) . </S><S> o. viehmann , c. eltschka , and j. siewert , * 83 * , 052330 ( 2011 ) . </S><S> n. r. wallach , lectures on quantum computing venice c.i.m.e . </S><S> june ( 2004 ) . </S><S> http://www.math.ucsd.edu/ </S><S> nwallach / venice.pdf </S>

1011.1959

high resolution quasar absorption line spectroscopy has enabled the statistical detection of intergalactic metals at cosmic overdensities as low as @xmath5 at @xmath6 .the inferred metallicities are around @xmath7 , with lines associated with a substantial fraction of forest lines with @xmath8 ( @xcite ) .additional metal lines , such as ( @xcite ) and ( @xcite ) are also detected in the low density intergalactic medium ( igm ) .observations indicate metallicity increases with density ( @xcite ) and the distribution of metals in the igm is inhomogeneous ( @xcite ) . a crucial ingredient in many of these studies are the ionisation corrections which convert the observed ionic abundances to metallicities .these corrections depend on the density and temperature of the igm , as well as the intensity and spectral shape of the metagalactic ultra - violet background ( uvb ) .the former may be modelled using cosmological hydrodynamical simulations which follow the enrichment history of the igm ( _ e.g. _ @xcite ) .the latter is usually calculated using detailed models for the uvb intensity and spectral shape ( @xcite ) .simulations of intergalactic metal enrichment typically assume the metagalactic ionising radiation field is spatially uniform .this is expected to be a reasonable approximation for hydrogen ionising photons at @xmath9 , when the mean free path of ionising photons is much larger than the average separation between ionising sources ( @xcite ) .however , the epoch of reionisation is thought to complete around @xmath6 , when quasars are numerous enough to provide the hard photons ( @xmath10 ) needed to doubly ionise helium ( @xcite ) .indeed , observations of the and forest indicate the uvb spectral shape fluctuates significantly on scales of @xmath11 mpc at these redshifts , producing a wide range of values for the to column density ratio , @xmath12 ( @xcite ) .these fluctuations are thought to arise from the small number of quasars lying within the short ( ionising photon ) mean free path expected at the tail - end of reionisation ( @xcite ) , although small scale radiative transfer effects ( @xcite ) , collisionally ionised gas ( @xcite ) or a significant number of thermally broadened lines ( @xcite ) may also produce column density ratios with @xmath13 .these spatial inhomgeneities in the uvb spectral shape should also have an impact on the ionisation balance of the metals in the igm . ) to quasars at @xmath14 ( @xcite , see also @xcite ) .this line - of - sight proximity effect arises because of the @xmath15 dependence of the ionisation radiation _ intensity _ around the quasar when the igm is optically thin , and it will manifest itself even after reionisation has long completed .this effect is slightly different to the large scale fluctuations in the uvb _ spectral shape _ discussed here , which are expected to be important toward the tail - end of reionisation when the ionising photon mean free path is small . ]intriguingly , independent of the spectral shape of the ( spatially uniform ) uvb used to make ionisation corrections , several studies find evidence for scatter in the igm metallicity at fixed density ( @xcite ) .this scatter may be intrinsic , or it may instead be partially attributable to spatial fluctuations in the spectral shape of the uvb .on the other hand , @xcite find little evidence for inhomogeneity in the @xmath16 $ ] they infer from the observed optical depth ratio @xmath4 , implying that spatial variations in the uvb spectral shape are small .there is also some evidence from metal line analyses for a hardening of the uvb spectrum at @xmath6 ( @xcite ) , perhaps due to the onset of reionisation .however , other studies are consistent with no evolution at the same redshifts ( @xcite ) . quantifying the effect of a spatially inhomogenous uvb on intergalactic metal absorption linesis thus desirable , both as a potential probe of reionisation and for examining the systematic bias , if any , spatially uniform ionisation corrections impart to constraints on the distribution of metals in the igm .analytical arguments have previously highlighted the importance of small scale fluctuations in the hydrogen ionising radiation field for metals associated with high column density systems ( @xcite ) .however , there has been no detailed examination of the impact of a spatially inhomogeneous uvb at @xmath10 on metals associated with lower column densities .a self - consistent treatment would require multi - frequency cosmological radiative transfer in a large ( @xmath17 ) volume coupled to a detailed chemodynamical model .in contrast , current state - of - the - art hydrodynamical simulations of igm metal enrichment generally assume a spatially uniform uvb ( _ e.g. _ @xcite , but see @xcite ) while large cosmological radiative transfer simulations of reionisation are still restricted to a treatment of hydrogen and helium only ( @xcite ) .in this paper we instead use a comparatively simple model which includes most of the relevant aspects of the spatially inhomogeneous uvb expected towards the tail - end of reionisation ( _ e.g. _ @xcite ) .we combine this model with a large hydrodynamical simulation of the igm to explore the impact of spatial fluctuations in the uvb spectral shape on intergalactic carbon and silicon at @xmath18 . in particular , we focus on the ratio of to absorption , which is sensitive to the shape of the uvb either side of the ionisation edge ( @xcite ) .we describe our hydrodynamical simulation and model for spatial fluctuations in the uvb spectral shape in section 2 .section 3 describes the four toy uvb models we use in this work . in section 4we proceed to examine the impact of spatial fluctuations in the uvb spectral shape on synthetic absorption spectra constructed from our simulations .we compare our results to observational data in section 5 , and conclude in section 6 .all atomic data is taken from @xcite , solar abundances are from @xcite and all distances are given in comoving units .we use the r4 cosmological hydrodynamical simulation described in @xcite to model the igm density field .the simulation was performed in a @xmath19 mpc box with @xmath20 particles using the parallel tree - sph code gadget-3 ( @xcite ) .the cosmological parameters are @xmath21 , @xmath22 , @xmath23 , @xmath24 , @xmath25 , @xmath26 ( _ e.g. _ @xcite ) .we use a snapshot drawn from the simulation at @xmath27 .our model for spatial fluctuations in the uvb spectral shape is similar to the approach described by @xcite ( see also @xcite ) .the model applies to fluctuations in the uvb spectrum at the tail - end of reionisation only , following the overlap of regions when the ionising photon mean free path is set by the abundance of lyman limit systems ( @xcite ) .there is good evidence to suggest that reioniation is indeed completing by @xmath6 ( @xcite ) , and this assumption should therefore be reasonable .note , however , that the model will not apply to the patchy ionisation state of the pre - overlap igm during the early stages of reionisation at @xmath28 , and does not include the effect of detailed , small scale radiative transfer effects on the igm ( @xcite ) .further modelling will thus be required to address the impact of spatial variations in the uvb spectral shape on intergalactic metals during the heart of reionisation .firstly , to create a volume large enough to contain several ionising photon mean free paths , we stack the hydrodynamical simulation volume around the original @xmath19 mpc box to create a cube @xmath29 on each side .we identify haloes in the volume using a friends - of - friends halo finder .quasar luminosities are assigned to the haloes by monte carlo sampling the @xcite b - band quasar luminosity function at @xmath18 .the luminosities are uniquely mapped to the halo masses by rank ordering both quantities and assigning the brightest quasars to the most massive haloes .this approach enables us to distribute sources on large scales while following the ionisation state of gas in detail in the central @xmath19 mpc box .however , it will only approximately model the clustering properties of the sources .fortunately , the rarity of bright quasars means that poisson fluctuations rather than clustering is expected to dominate the resulting uvb fluctuations ( @xcite ) , so this limitation is unlikely to significantly affect our results .we require the number of quasars in a volume , v , to satisfy @xmath30 where @xmath31 is the @xcite quasar luminosity function. the lower limit of the integral is @xmath32 , corresponding to @xmath33 .the b - band luminosity of each quasar is converted into a luminosity below the lyman limit using a broken power law spectrum ( _ e.g. _ @xcite ) @xmath34 we consider two cases for the extreme uv ( euv ) spectral index , @xmath35 . in the first casewe assume a constant value of @xmath36 , consistent with the mean obtained by @xcite for radio quiet quasars , @xmath37 . however , a wide range of values are measured for the euv spectral index ( _ e.g. _ @xcite ) .consequently , we also consider a variable euv spectral index .scatter in @xmath35 is included by monte - carlo sampling a gaussian with mean @xmath38 and standard deviation @xmath39 over the range @xmath40 , similar to the dispersion measured by @xcite . once the luminosity and spectral energy distribution for each quasar is specified , the specific intensity of the uvb , @xmath41 @xmath42 $ ] , at frequencies between the and ionisation edges , @xmath43 , is given by @xmath44 the sum is made over all quasars in the @xmath29 volume , where @xmath45 is the distance of quasar @xmath46 from @xmath47 , which is always in the central @xmath48 box .( [ eq : jh1 ] ) assumes that the igm is optically thin to ionising photons ; this should be a reasonable approximation at @xmath18 ..the four uvb models used in this work . from left to right , the columns list the model name , the photon mean free path at the ionisation edge , the quasar euv spectral index and the logarithm of the spatially averaged softness parameter , @xmath49 $ ] . [ cols="^,^,^,^",options="header " , ] [ tab : sims ] at frequencies above the ionisation edge , @xmath50 , the specific intensity is instead @xmath51 where @xmath52 is the mean free path for photons at the ionisation edge and @xmath53 is the power - law index describing the slope of the column density distribution , @xmath54 .this expression assumes quasars are point sources surrounded by discrete , poisson distributed absorbers ( @xcite ) .we adopt @xmath55 , following the analogous distribution for absorbers ( @xcite ) .note , however , the slope of the column density distribution is not constrained observationally and may deviate from the distribution for absorbers associated with high column densities , @xmath56 ( @xcite ) . finally , in this work we model the contribution to the uvb from quasars only . the non - thermal emission from quasars is thought to dominate the uvb at @xmath57 , but star forming galaxies will make an increasingly significant contribution to the uvb at lower frequencies at @xmath28 ( @xcite ) .however , the soft spectra of these sources are unlikely to significantly contribute to spatial fluctuations in the uvb spectral shape at @xmath10 .the four toy uvb models used in this work are summarised in table 1 .the models are constructed by evaluating eqs .( [ eq : jh1 ] ) and ( [ eq : jhe2 ] ) on a @xmath58 grid within the central @xmath59 simulation box. the size of each grid cell , @xmath60 mpc , is chosen to match the scales of @xmath11 mpc on which the to optical depth ratio is observed to vary at @xmath61 ( @xcite ) .this choice also represents the best compromise between resolution and efficiency ; the construction of the photo - ionisation balance look - up tables ( see section [ sec : makespec ] ) becomes very time consuming if the number of grid cells much larger .however , note that the rarest , high amplitude fluctuations in the uvb spectral shape on small scales will not be fully captured in the simulation . in each grid cellthe specific intensities are evaluated at @xmath62 different photon energies spanning the range @xmath63 .lastly , all four models are renormalised by a factor of @xmath64 to give identical mean specific intensities at the ionisation edge @xmath65 .this renormalisation yields photo - ionisation rates of @xmath66 ( cf .@xmath67 prior to the renormalisation ) , consistent with observational constraints derived from the forest opacity at @xmath18 ( @xcite ) . models uvb1 and uvb2 are constructed using a single euv spectra index , @xmath36. however , the value of the mean free path at the ionisation edge , @xmath52 , is rather uncertain . in this workwe adopt a fiducial value of @xmath68 at @xmath18 ( _ e.g. _ @xcite ) , but the exact value will depend on the to column density ratio , @xmath69 , which exhibits significant fluctuations approaching @xmath18 ( @xcite ) .we thus also consider a smaller mean free path for model uvb2 , @xmath70 , consistent with the constraints recently presented by @xcite at @xmath6 . model uvb3 instead uses a variable euv spectral index and @xmath71 mpc .the final model , uvb4 , is identical to model uvb1 , aside from the spectral shape between @xmath72 and @xmath73 ryd .resonant absorption due to lyman series lines will also produce significant spatial fluctuations in the uvb between at @xmath74 and the ionisation edge at @xmath75 ( @xcite ) .model uvb4 mimics these fluctuations by assuming the uvb spectrum computed with eqs .( [ eq : jh1 ] ) and ( [ eq : jhe2 ] ) follows a power law @xmath76 for @xmath77 , where @xmath78 is the frequency of at @xmath74 and @xmath79 $ ] .the spatially averaged spectra , @xmath80 , for all four uvb models are displayed in the left hand panel of fig .[ fig : uvbspectra ] .the step at @xmath75 in models uvb1, uvb2 and uvb3 , and between @xmath81 for model uvb4 , is due to the attenuation of the spectrum by the opacity .the spectra recover their intrinsic shape at high energies , where the photon mean free path is long and the attenuation of the radiation by the intervening absorbers is negligible .model uvb3 , which assumes a variable euv spectral index , is much harder than the other models at high energies .this model is dominated by the quasars with the hardest spectra , @xmath82 , in the simulation volume , even if these objects are not the brightest quasars at lower frequencies .the logarithm of the spatially averaged softness parameter for each model , defined as the ratio of the to photo - ionisation rates , @xmath83 , are listed in table 1 .observational constraints on the softness parameter at @xmath18 are dominated by the large uncertainty in the effective optical depth .@xcite find @xmath84 at @xmath18 by comparing hydrodynamical simulations of the and forest opacity to observational data .all four models are designed to be consistent with these constraints .the right hand panel of fig .[ fig : uvbspectra ] again displays model uvb1 , but the dotted curves now show the range encompassing 95 per cent of all fluctuations around the median specific intensity .fluctuations are largest close to the ionisation edge where the attenuation of the radiation is strongest , while towards higher frequencies the increasing mean free path produces much smaller departures from the median . for comparison ,the dashed line shows the widely used uvb model of @xcite ( hm01 ) for emission from galaxies and quasars at @xmath18 .the spectrum has been renormalised to match the specific intensity of uvb1 at @xmath85 .the softness parameter of the hm01 spectrum is @xmath86 , around @xmath87 lower than uvb1 . although the features of the toy uvb spectra presented here are broadly similar to the more detailed model of hm01 , there are some important differences .in particular , our toy models do not include more complex effects such as recombination emission ( @xcite ) or sawtooth absorption by the lyman series ( @xcite ) which are important to consider when performing detailed modelling of observational data .the fluctuations in the specific intensity are quantified in more detail in fig .[ fig : flucpdf ] for all four uvb models .the left panel displays the probability distribution for the specific intensity relative to its spatially averaged value , @xmath88 $ ] , at the ionisation edge . the distribution becomes broader for a smaller @xmath52 ( uvb2 ) as the number of sources within one mean free path is reduced .somewhat counter - intuitively , the variable euv spectral index model ( uvb3 ) produces fewer of the large , rare fluctuations from the mean than uvb1 , which assumes a constant @xmath35 but is otherwise identical .this is because the three quasars which lie closest to the centre of our simulation volume ( and hence dominate the rare , high amplitude fluctuations ) have randomly assigned euv spectral indices which conspire to reduce fluctuations in the specific intensity at @xmath75 .recall the quasar luminosities are normalised in the b - band , and in this instance the euv spectra for the three closest quasars all have @xmath89 .this produces lower specific intensities at the ionisation edge compared to model uvb1 with @xmath36 , and hence fewer large fluctuations from the mean .lastly , the right panel of fig .[ fig : flucpdf ] displays cumulative distributions for @xmath90 $ ] obtained from model uvb2 .the four separate curves show the cumulative distributions for subsets of pixels corresponding to regions in the igm density field with overdensities at @xmath91 and @xmath92 , respectively .the lack of a clear correlation with density , obtained from a voigt profile analysis of synthetic absorption spectra was systematically higher in regions where @xmath93 compared to regions with lower optical depths ( and hence densities ) .they suggested this implied the fluctuating uvb model they used was harder in higher density regions .however , this interpretation is in disagreement with our findings here .this earlier result is likely spurious , and probably results from the assumptions of pure turbulent broadening in the voigt profile analysis ( @xcite ) and a spatially uniform photo - ionisation rate . ]is mainly due the low number density of bright quasars , resulting in a radiation field which varies on scales much larger than the jeans scale ( see the discussions in _ e.g. _@xcite ) . note , however , that fluctuations on scales @xmath94 ( typically those associated with the radiation field in close proximity to the quasars ) will be smoothed over , but these fluctuations are also rather rare . on the other hand , if additional physics such as small scale radiative transfer effects are important , or if ionising sources are far more numerous ( _ e.g. _ galaxies ) , the ionising radiation field may be more closely correlated with the igm density on small scales .we now turn to examine the impact of our spatially inhomogeneous uvb models on synthetic quasar absorption spectra .we use a simple _ a posteriori _ approach to including metals within our hydrodynamical simulation .we do not model the production and dispersal of metals in detail ( see _ e.g. _ @xcite ) .we instead follow carbon and silicon abundances only , and assume the metallicity of the igm traces the gas density distribution .@xcite measure a best fit median metallicity of @xmath95=-3.47^{+0.07}_{-0.06 } + 0.65^{+0.10}_{-0.14}(\log \delta - 0.5)$ ] , with a lognormal scatter of @xmath96})=0.76^{+0.05}_{-0.08}-0.23^{+0.09}_{-0.07}(\log\delta - 0.5)$ ] at @xmath18 .following @xcite , we introduce the lognormal scatter to the median as @xmath97}=-3.47 + 0.65(\log \delta - 0.5 ) + s$ ] by dividing the @xmath19 mpc simulation volume into @xmath98 cubes and adding a different scatter , @xmath99 , in each sub - volume .the scatter , @xmath99 , is obtained by randomly sampling a normal distribution in @xmath100 $ ] with zero mean and standard deviation @xmath101 .we further assume the silicon abundance relative to carbon is @xmath102=0.77 $ ] ( @xcite ) .note , however , these metallicity constraints are derived from a statistical analysis of and absorption lines which use an ionisation correction based on the hm01 uvb model .different ionisation corrections obtained using significantly softer or harder uvb spectra will alter these constraints considerably .the equilibrium ionisation balance of the metals in our hydrodynamical simulation is calculated using the photo - ionisation code cloudy ( version 08.00 ) , last described by @xcite . for each of our four spatially inhomogeneous uvb models ,we create a five dimensional look - up table listing the ionisation fractions of hydrogen , helium , carbon and silicon as a function of gas temperature , @xmath103 , hydrogen number density , @xmath104 , and position on the @xmath105 uvb grid within our @xmath59 simulation box .synthetic spectra are obtained from the hydrodynamical simulations by randomly selecting 1000 sight - lines from the simulation and performing an interpolation on the particle data weighted by the smoothing kernel ( _ e.g. _ @xcite ) .each line of sight is drawn parallel to the box boundaries and has 2048 pixels , each of which has an associated gas density , temperature , peculiar velocity and ionisation fraction .the ionisation fractions for h@xmath106 , he@xmath107 , c@xmath108 and si@xmath109 are obtained by linearly interpolating the five dimensional look - up table .absorption spectra are constructed for ( @xmath1101216 ) , ( @xmath1101026 ) , ( @xmath110973 ) , ( @xmath110950 ) , ( @xmath110304 ) , ( @xmath1111548 , 1551 ) and ( @xmath1111394 , 1403 ) .the ratio of the optical depths for two different species is sensitive to fluctuations in the spectral shape of the ionising background between their respective ionisation potentials .a ratio also has the advantage of being less susceptible to fluctuations in the igm density field ( @xcite ) .we examine @xmath112 for the transitions and @xmath113 for @xmath1101548 and @xmath1101394 in our synthetic spectra .both of these ratios have been widely used as probes of reionisation and the uvb spectral shape at @xmath6 ( _ e.g. _ @xcite ) . notethe latter will also be sensitive to fluctuations in the relative abundance of carbon and silicon .it is instructive to first revisit the effect of uvb fluctuations on the to optical depth ratio ( but see also @xcite ) .the upper left panel of fig .[ fig : ratios ] shows the cumulative distribution for @xmath2 for _ all pixels _ in our synthetic spectra .this includes pixel optical depths that are in practice unobservable , either because of saturated absorption or noise .the thick curves correspond to the optical depth ratios for the four spatially fluctuating uvb models listed in table [ tab : sims ] .models uvb1 and uvb4 are almost identical ; neither or are significantly ionised by photons with @xmath114 .models with progressively harder spectra produce lower median values for the to ratio .the fraction decreases as the specific intensity at @xmath115 increases ; since the uvb is almost uniform at @xmath85 in all our models ( see the right panel in fig . [fig : uvbspectra ] ) , fluctuations in the specific intensity at the ionisation edge will have little impact on @xmath2 .note also that the smaller mean free path used for uvb2 produces a much wider range of @xmath112 values due to the wider range of fluctuations produced at the ionisation edge .we may assess the extent to which @xmath2 is influenced by fluctuations in the uvb spectral shape by comparison to the ratio obtained for a spatially uniform uvb .the thin curves display the optical depth ratios obtained using the spatially averaged ( _ i.e. _ spatially uniform ) spectrum for each uvb model .the distributions for models including uvb fluctuations are indeed broader , especially toward higher values of @xmath0 .however , all four distributions still extend to low @xmath2 in the spatially uniform case .this behaviour is almost entirely due to variations in @xmath112 introduced by thermal broadening ; the thermal width of an absorption line for species @xmath46 is @xmath116 , where @xmath117 is the mass of the particle .fluctuations in the optical depth ratio therefore also originate from the wings of absorption lines which are predominantly thermally rather than turbulently broadened ( @xcite ) . the upper right panel of fig .[ fig : ratios ] instead shows only the subset of pixels which have @xmath118 , roughly corresponding to range over which the optical depths may be reliably measured in observational data ( _ e.g. _ @xcite ) .these pixels correspond to @xmath119 , @xmath120 , @xmath121 and @xmath122 per cent of the total for models uvb1 , uvb2 , uvb3 and uvb4 , respectively . the median @xmath2 for uvb2 is now significantly smaller .this is because the small number of optical depths with @xmath123 selected for uvb2 tend to probe regions where the fraction is lowest .recent observations of the opacity in the spectrum of the @xmath124 quasar he2345@xmath1254342 , made using the cosmic origins spectrograph on the _ hubble space telescope _, reliably measure fluctuations in the optical depth ratio spanning the range @xmath126 at @xmath127 ( @xcite ) . in comparison ,model uvb2 exhibits @xmath128 ( @xmath129 for 95 per cent around the median ) , while uvb1 and uvb4 have @xmath130 ( @xmath131 ) and uvb3 has @xmath132 ( @xmath133 ) .we do not perform a more detailed comparison here , but as previously noted these results strongly suggest the observed @xmath2 fluctuations are largely attributable spatial variations in the ionising background .these fluctuations arise primarily from the small number of quasars within the short ionising photon mean free path , @xmath134 mpc , expected towards the tail - end of reionisation .the results for the @xmath4 ratio are shown in the lower panels of fig .[ fig : ratios ] .the ionisation edges for and are at @xmath135 and @xmath136 ryd respectively ; the ratio of these optical depths is thus sensitive to the shape of the uvb either side of the ionisation edge at @xmath73 ryd ( @xcite ) .the left hand panel displays the distributions for all pixels in the synthetic spectra .models with progressively harder spectra produce smaller median values for the ratio the distribution for uvb3 in particular extends to significantly lower @xmath4 values . a harder uvb reduces both the and ionisation fractions , but the fraction is lowered more rapidly as the break at @xmath75 is reduced , leading to the smaller values for the @xmath113 ratio in uvb3however , in contrast to the case for @xmath2 , the distributions for the fluctuating and uniform uvb models are almost identical . only the uvb3 distribution displays a small difference for the lowest values of @xmath4 .this scatter in the @xmath113 ratio is instead dominated by the different dependences of the and fractions on gas density , and to a much lesser extent variations in the absorption profiles in redshift space .the lower right hand panel in fig .[ fig : ratios ] displays the subset of pixels with @xmath137 .these pixels represent a very small fraction of the total : @xmath138 , @xmath139 , @xmath140 and @xmath138 per cent for models uvb1 ,uvb2 , uvb3 and uvb4 , respectively , reflecting the much smaller volume filling factor of detectable and absorption in the simulations . the medians of all four distributions are now significantly larger ; most of the discarded pixels have and optical depths which are too low to be observed . the distributions for modelsuvb1 , uvb2 and uvb3 are now similar , suggesting that the @xmath4 ratio is a less sensitive probe of the uvb spectral shape for these pixels . finally , all four distributions are again virtually indistinguishable from the distributions obtained with a spatially uniform uvb . , against the optical depth weighted overdensity , @xmath141 , for pixels with @xmath142 in the synthetic absorption spectra constructed using model uvb1 ._ top left : _ the temperature density plane for absorption ._ top right : _ absorption ._ bottom left : _ absorption ._ bottom right : _ absorption .the number density of pixels increases by 1.0 dex within each contour level.,scaledwidth=48.0% ] we may gain insight into this behaviour by first examining the typical gas densities and temperatures which produce the absorption in our synthetic spectra. contour plots of the optical depth weighted temperature against optical depth weighted overdensity are shown in fig . [fig : trho ] for spectra constructed using model uvb1 . clockwise from the upper left , each panel shows the temperature - density plane for h@xmath106 , he@xmath107 , si@xmath108 and c@xmath109 optical depths with @xmath143 .the number of pixels increases by an order of magnitude within successive contours .it is evident that and absorption primarily probes photo - ionised gas around mean density and below , with the absorption slightly more sensitive to gas in voids ( @xcite ) .in contrast , and especially tend to probe _ overdense _ regions in the igm ( _ e.g. _ @xcite ) . nearly all the carbon and silicon absorption is furthermore associated with gas at @xmath144 .the exact temperature and density of regions with measurable optical depths depends on the uvb model used .for example , for uvb3 ( the hardest spectrum we consider ) the median @xmath145 weighted temperature and overdensity for pixels with @xmath146 are @xmath147 and @xmath148 . for @xmath149 the median @xmath150 weighted values are instead @xmath151 and @xmath152 . for uvb2 ( the softest spectrum )the corresponding values are @xmath153 , @xmath154 for @xmath155 and @xmath156 , @xmath157 for @xmath149 .nevertheless , it is clear the @xmath158 and @xmath113 ratios probe rather different temperature and density regimes in the igm .furthermore , since the majority of the and absorption in our models corresponds to gas with @xmath144 , collisional ionisation is unlikely to explain the similarity between the @xmath4 distribution for the spatially uniform and fluctuating cases shown in fig .[ fig : ratios ] . note ,however , additional heating from feedback is not included in our hydrodynamical simulation , and we thus likely underpredict the amount of and absorption due to collisionally ionised gas .however , because fluctuations in the photo - ionising background will be less important for the @xmath113 ratio if a larger fraction of these elements are in a hot , predominantly collisionally ionised phase with @xmath159 , this is unlikely to significantly alter our conclusions regarding the importance of inhomogeneities in the uvb spectral shape . , for the four uvb models considered in this work .the ratios are plotted as a function of @xmath160 , the ratio of specific intensities at the and ionisation edges ( see text for details ) . in each panel , @xmath161 is shown at four different gas densities , @xmath162 ( dot - dashed curves ) , @xmath163 ( solid curves ) , @xmath164 ( dotted curves ) and @xmath165 ( dashed curves ) .the thick curves assume a gas temperature of @xmath166 while the thin curves correspond to @xmath167 .the horizontal extent of the shaded regions display the range encompassing 95 per cent of all fluctuations around the median @xmath168 for the spatially inhomogeneous uvb models ._ upper left : __ upper right :_ lower left : __ lower right : _uvb4.,scaledwidth=48.0% ] for gas @xmath163 and @xmath166 .the horizontal extent of the shaded regions display the range encompassing 95 per cent of all fluctuations around the median @xmath169 for the spatially inhomogeneous uvb models.,scaledwidth=48.0% ] with the typical gas densities and temperatures of the absorption in hand , we may now examine the effect of varying the spectral shape of the uvb in more detail .it is instructive to simply rescale the _ spatially averaged _ uvb spectrum for each model by a constant factor above the ionisation edge : @xmath170 where @xmath171 is a dimensionless constant . for model uvb4, this rescaling also modifies the spectral shape between @xmath114 .smaller ( larger ) values of @xmath171 thus produce softer ( harder ) uvb spectra and a stronger ( weaker ) break in the spectrum at @xmath75 .[ fig : ionfrac ] displays the to ionisation fraction , @xmath161 , for the four uvb models considered in this work . ignoring redshift space distortions ,this is related to the optical depth ratio by @xmath172 - 0.77 } , \label{eq : convert_tau}\ ] ] where @xmath173 and @xmath174 are the absorption cross - section and number density for each species .the ratios are plotted as a function of the ratio of specific intensities at the and ionisation edges , @xmath169 , following the above rescaling .the thick curves are computed using cloudy for gas with @xmath166 , similar to the temperature of gas responsible for and absorption in our simulated spectra . in each panel ,the to ratio is shown at four different gas densities , @xmath175 and @xmath176 .these have overdensities at @xmath18 of @xmath177 and @xmath178 , respectively . for comparison ,the median optical depth weighted overdensity probed by and absorption with @xmath179 for uvb1 is @xmath180 . the horizontal extent of the shaded regions in fig .[ fig : ionfrac ] correspond to the range encompassed by 95 per cent of all fluctuations from the median @xmath169 in the _ spatially fluctuating _ uvb models .this provides a rough guide to the range in @xmath161 present in the synthetic spectra as a function of density at fixed temperature. however , there are three important points to keep in mind .firstly , variations in the gas temperature ) spatial fluctuations in the temperature of the low density igm expected during reionisation ( @xcite ) .however , these fluctuations are expected to have a significant impact on large scale , transverse correlations in the forest only ; their signatures are much more difficult to detect along individual lines - of - sight and on smaller scales ( @xcite ) . ]will produce additional fluctuations in @xmath161 at fixed density .this is illustrated by the thin curves in fig .[ fig : ionfrac ] , which are computed for gas with @xmath167 ; this temperature lies toward the upper range of the scatter observed in fig .[ fig : trho ] .secondly , only values of @xmath181 ( lower right panel , fig .[ fig : ratios ] ) are readily observable in our synthetic spectra , which from eq .( [ eq : convert_tau ] ) corresponds to @xmath182 .thirdly , the range of @xmath161 fluctuations will in practice be smaller than indicated by the range of @xmath183 shown by the shading in fig .[ fig : ionfrac ] . fluctuations in the uvb spectral shape are _ maximised _ at @xmath75 where the mean free path is shortest ( _ e.g. _ fig .[ fig : uvbspectra ] ) , whereas and have ionisation potentials above and below @xmath75 , respectively .furthermore , the fluctuations do not have an equal probability of occurring in this range , and have a higher probability of lying close to the median ( _ e.g. _ fig . [ fig : flucpdf ] ) .the shaded regions thus correspond to an upper limit in the expected @xmath161 variation for each model . with these points in mind ,the @xmath161 ratio nevertheless shows only modest variation ( @xmath184 dex ) over the shaded range in fig .[ fig : ionfrac ] for @xmath163 , with the largest variations occurring for uvb2 and uvb4 . towards larger @xmath169 ( _ i.e. _ softer spectra ) , @xmath161 noticeably flattens for all models , and becomes almost constant for uvb1 , uvb2 and uvb3 .larger variations in @xmath161 do occur at lower densities , especially for model uvb3 where the to ratio changes by @xmath185 dex within the shaded range for @xmath186 .the to optical depth ratio for _ all _ pixels in fig . [ fig : ratios ] is slightly broader than the uniform case for uvb3 at low values as a consequence .however , from eq .( [ eq : convert_tau ] ) we infer that and absorption from such low density gas is not detectable in the synthetic spectra . in fig .[ fig : ionfrac2 ] the individual carbon ( upper panels ) and silicon ( lower panels ) ionisation fractions for several different ions are shown against @xmath169 .the curves are again computed assuming densities and temperatures typical of the and absorption in our simulated spectra , @xmath187 and @xmath166 .the left panels display the results for uvb2 , while uvb4 is shown on the right .models uvb1 and uvb3 display similar behaviour to uvb2 within the equivalent ranges for @xmath169 .note the and fractions for model uvb2 are largely insensitive to the uvb spectral shape for @xmath188\ga 2 $ ] and @xmath72 , respectively .this is because the uvb is very soft , resulting in photo - ionisation rates that are too low to significantly change the and fractions in the higher density gas responsible for most of the absorption . on the other hand ,the fraction for model uvb4 is more sensitive to an increase in the strength of the @xmath81 break .the break at @xmath74 in this model produces a larger proportion of carbon in the form of relative to for increasing @xmath169 ( @xcite ) . however , again flattens towards larger values for @xmath183 , moderating the change in the to ratio .furthermore , as the fraction drops rapidly at @xmath189>2 $ ] it will become progressively more difficult to detect the weakening absorption .consequently , as the ionising photon mean free path becomes smaller toward higher redshift , the average uvb spectral shape softens and variations in the _ observed _ to ratio do not become significantly more pronounced as a result of spatial fluctuations in the uvb spectral shape .the and absorption systems typically originate from overdense regions with @xmath190 in our simulations , where the and fractions change less rapidly with increasing @xmath169 compared to lower density .it is this behaviour , combined with the smaller fluctuations expected in the uvb spectral shape at frequencies above and below the ionisation edge , which explain the similarity between the to optical depth ratios for the uniform and fluctuating models in fig .[ fig : ratios ] .this is in striking contrast to the to ratio , where the same spatial variations in the uvb spectral shape significantly increase fluctuations in @xmath2 from lower density gas as the mean free path is lowered ( @xcite ) .the absence of any observational evidence for fluctuations in the @xmath4 ratio ( or the inferred silicon to carbon ratio ) is thus not necessarily indicative of a spatially uniform uvb at @xmath6 . on the other hand , slightly larger @xmath4 fluctuations are expected at lower densities , although the weaker absorption from these regions is much more difficult to detect .we now examine whether uvb fluctuations can be detected statistically in the observational data , and whether or not they contribute to the observed scatter in the igm metallicity ( @xcite ) .the pixel optical depth ( pod ) procedure provides a powerful tool for statistically analysing metal absorption at low gas densities ( _ e.g. _ @xcite ) .briefly , the procedure involves recovering optical depths for a base transition ( _ e.g. _ absorption ) on a pixel - by - pixel basis and pairing these with metal optical depths recovered at the same redshift . given suitable calibration ,typically achieved using a cosmological hydrodynamical simulation , information may then be derived on the abundance and distribution of metals in the igm .the pod analysis performed by @xcite found evidence for scatter in @xmath95 $ ] at fixed @xmath145 , described by a normal distribution in @xmath95 $ ] with standard deviation @xmath191)=0.76 - 0.23(\log\delta - 0.5)$ ] at @xmath18 .@xcite and @xcite noted this scatter may be attributable to spatial variations in the metallicity of the igm , although they did not rule out the possibility that spatial fluctuations in the uvb spectral shape ( and hence the assumption of a uniform ionisation correction ) may also play a role . on the other hand, @xcite performed a pod analysis for and found no evidence for additional scatter in the @xmath3 ratio beyond the aforementioned scatter in the metallicity .the study concluded that a uniform @xmath192 $ ] is favoured and that inhomogeneities in the spectral shape of the uvb are small .however , none of these studies modelled the effect of uvb fluctuations on the ionisation balance for carbon and silicon in detail .we perform a pod analysis on our synthetic spectra using the procedure outlined by @xcite .we consider the optical depths for through to as well as and absorption .for the pod analysis we construct absorption spectra which resemble high resolution , high signal - to - noise observational data ( _ e.g. _ @xcite ) .the synthetic spectra are convolved with a gaussian instrument profile of width @xmath193 , resampled onto pixels of width @xmath194 and gaussian distributed noise with @xmath195 is added .continuum fitting errors are mimicked by performing an iterative continuum correction to the synthetic spectra . we compute the median transmitted flux in each synthetic line - of - sight and deselect all pixels below @xmath196 of this value , where @xmath197 is the rms noise amplitude in each pixel .the median flux is computed again for the remaining pixels , and the procedure is then continued until convergence is reached .the results of our pod analysis are presented in fig . [fig : pod ] .the data points with 1@xmath197 error bars in the left panel correspond to the @xcite observational measurements for q1422@xmath198230 .these data have a median absorption redshift of @xmath199 , at slightly higher redshift than our synthetic absorption spectra . from top to bottom , the three sets of data points correspond to the 84th , 69th and 50th percentiles of the recovered optical depths ; the 84th and 69th percentiles have been offset by @xmath200 dex and @xmath201 dex for clarity .the curves display the recovered pixel optical depth against optical depth for spectra constructed using model uvb4 , which we find has the largest impact on the recovered optical depths .the solid curves are obtained from the spatially inhomogeneous uvb model , while the dotted curves corresponds to the optical depths recovered from the spatially averaged spectrum .the curves are almost indistinguishable , indicating that the predicted spatial inhomogeneities in the uvb spectral shape approaching reionisation will not significantly impact on the ionisation correction for c@xmath108 .the dashed and dot - dashed curves in fig . [ fig : pod ] use the spatially uniform uvb model , but now also include different assumptions for the lognormal scatter in the igm metallicity at fixed density .the dashed curves assume zero scatter , @xmath202 , while the dot - dashed curves correspond to @xmath203 .the impact on the recovered optical depths is especially prominent for the higher percentiles , suggesting that spatial variations in metallicity rather than the uvb spectral shape will dominate any scatter . as noted by @xcite , the fiducial model with @xmath101is in somewhat better agreement with the higher percentiles . in the right hand panelwe compare the same model to the pod measurements for and presented by @xcite for the same quasar .again , the recovered optical depths are very similar for the spatially inhomogeneous and uniform uvb models , and scatter in the metallicity has a much larger impact on the recovered optical depths .the fiducial model with @xmath101 is again in better agreement with the higher percentiles , although the agreement is poorer for the median .note also , in contrast to the and pixel optical depths , lognormal scatter in the metallicity at fixed density _ lowers _ each percentile for @xmath204 as a function of @xmath150 .the explanation for this behaviour is that scatter in metallicity is added to _ both _ of the pixel optical depths , in contrast to the left hand panel of fig .[ fig : pod ] . adding this scatter enables the detection of both and absorption in more pixels at lower densities due to their higher metallicities; these pixels are otherwise hidden in the flat ( noise and continuum error dominated ) part of the correlation at @xmath205. however , for this uvb model the fraction ( and hence optical depth ) decreases more rapidly with decreasing gas density relative to the fraction .consequently , a greater fraction of optical depths with low @xmath204 are now _ detectable _ , leading to the lowering of the @xmath204 percentiles observed in fig .[ fig : pod ] . finally , we caution the reader not to take the differences between the observational data and simulations too seriously ; the @xcite and @xcite metallicities we have used to construct our spectra are derived under the assumption of a different uvb model .there is therefore no reason to expect our model to match the data exactly using these metallicities ; the comparison here is instead largely illustrative and one may always fine tune the metallicity to enable a given uvb model to match the data .the key point is that we clearly expect variations in the _ spatial distribution _ of metals to dominate any scatter in the and optical depths at fixed density , even in the presence of the spatially inhomogeneous uvb expected at the tail - end of reionisation .we use a large hydrodynamical simulation of the igm combined with a toy model for spatial inhomogeneities in the uvb spectral shape to investigate the impact of spectral hardness fluctuations on the ionisation balance of intergalactic carbon and silicon at @xmath6 .we construct synthetic quasar absorption spectra from the simulations , assuming the metallicity of the igm traces the underlying gas density ( @xcite ) .we carefully examine the impact of the uvb fluctuations on the and optical depths .four different spatially inhomogeneous uvb models which employ a variety of different assumptions for the uvb spectral shape are considered in our analysis .we reconfirm that fluctuations in the uvb spectral shape expected at the tail - end of reionisation have a significant impact on the to optical depth ratio ( see also @xcite ) .however , some of the lowest values for this ratio also result from lines which are predominantly thermally broadened ( @xcite ) . on the other hand , while the to ratio is indeed sensitive to the average spectral shape of the uvb , we find the predicted fluctuations have little impact on @xmath4 measured in our synthetic spectra .the majority of the detectable and absorption in our synthetic spectra originates from regions with @xmath206 and @xmath207 .these absorbers are predominantly photo - ionised in our simulations , and we may thus exclude collisional ionisation as a possible explanation for the small impact of uvb fluctuations on our simulated spectra .instead , we find the ratio of observable to optical depths varies relatively little considering the wide range of fluctuations in the uvb spectral shape .this is in part because of the longer mean free path for photons above and below the ionisation edge , which results in smaller fluctuations in the uvb spectral shape at these frequencies .however , as fluctuations in the uvb spectral shape become larger as the opacity increases , the spatially averaged uvb spectral shape becomes softer ; uvb models which produce the largest fluctuations in @xmath2 also have photo - ionisation rates which are too low to have a significant impact on the observed to ratio . at lower gas densities , or for uvb models which predict a larger fraction of relative to ( _ e.g. _ for lyman series absorption ) , the expected variation in the to ratio can be slightly larger , but it is more difficult to detect due to the correspondingly smaller gas densities and/or fraction of triply ionised carbon produced . finally , we briefly examine the observational consequences for studies of the igm metallicity using and absorption at @xmath6 .we perform a pixel optical depth analysis on our synthetic spectra , and find that the predicted uvb hardness fluctuations will have little impact on observations compared to spatial variations in the igm metallicity .we conclude that the lack of any observed fluctuations in the @xmath4 ratio does not provide a stringent limit on the non - uniformity of the uvb spectral shape , and in particular does not preclude the possibility of reionisation completing around @xmath1 . on the other hand , we confirm the observed scatter in the igm metallicity inferred from and absorption @xmath1 is likely to be intrinsic , reinforcing its potential as a powerful constraint on intergalactic metal enrichment scenarios .the authors thank the staff of the institute of astronomy and the kavli institute for cosmology , cambridge , for their hospitality during the completion of this work , and acknowledge valuable conversations with george becker , benedetta ciardi , luca graziani , martin haehnelt , matthew pieri , michael rauch and stuart wyithe .we also thank rob simcoe for drawing our attention to his latest results ahead of publication , and the anonymous referee for a constructive report which helped improve this paper . the hydrodynamical simulation used in this workwas performed using the darwin supercomputer of the university of cambridge high performance computing service ( http://www.hpc.cam.ac.uk/ ) , provided by dell inc . using strategic research infrastructure funding from the higher education funding council for england .jsb acknowledges the support of an arc australian postdoctoral fellowship ( dp0984947 ) .mv is partly supported by : asi / aae theory grant , infn - pd51 grant , prin - miur , prin - inaf and by the fp7 erc starting grant `` cosmoigm '' .

<S> spatial inhomogeneities in the spectral shape of the ultra - violet background ( uvb ) at the tail - end of reionisation are thought to be the primary cause of the large fluctuations observed in the to forest optical depth ratio , @xmath0 , at @xmath1 . </S><S> these spectral hardness fluctuations will also influence the ionisation balance of intergalactic metals ; we extract realistic quasar absorption spectra from a large hydrodynamical simulation to examine their impact on intergalactic and absorbers . using a variety of toy uvb models </S><S> , we find that while the predicted spatial inhomogeneities in spectral hardness have a significant impact on @xmath2 , the longer mean free path for photons with frequencies above and below the ionisation edge means these fluctuations have less effect on the and ionisation balance . </S><S> furthermore , uvb models which produce the largest fluctuations in specific intensity at the ionisation edge also have the softest ionising spectra , and thus result in photo - ionisation rates which are too low to produce significant fluctuations in the observed @xmath3 . instead </S><S> , we find spatial variations in the igm metallicity will dominate any scatter in @xmath4 . </S><S> our results suggest that observational evidence for homogeneity in the observed @xmath4 distribution does not rule out the possibility of significant fluctuations in the uvb spectral shape at @xmath1 . on the other hand , the scatter in metallicity inferred from observations of intergalactic and absorption at @xmath1 using spatially uniform ionisation corrections is likely intrinsic , and therefore provides a valuable constraint on intergalactic metal enrichment scenarios at these redshifts . </S><S> [ firstpage ] methods : numerical - intergalactic medium - quasars : absorption lines . </S>

cond-mat0703175

the calculation of free energies is central to many applications of computer simulations ranging from the calculation of ligand affinities to the study of phase equilibria in condensed materials .since the free energy is a quantity related to the phase space volume available to a system , its calculation is non - trivial in most interesting cases and in the past decades considerable effort has been devoted to developing efficient free energy computation techniques @xcite .recently , jarzynski has shown how free energies can be determined from non - equilibrium simulations @xcite . in thisso called _ fast switching _ or _fast growth _ method a control parameter coupled to the hamiltonian of the system is switched from an initial to a final value at a finite rate . by changing the external parameter , the work @xmath0 is performed on the system . as a consequence of the second law of thermodynamics the average of this work is larger than the free energy difference @xmath1 between the equilibrium states corresponding to the final and initial value of the control parameter : @xmath2 here , the angular brackets denote an average over pathways starting from the equilibrium state corresponding to the initial value of the control parameter . the equal sign in this inequality , known as _ clausius inequality _ or _ maximum work theorem _, holds only if the external parameter is switched reversibly . in 1997jarzynski demonstrated that the clausius inequality can be turned into an equality by averaging the work exponential rather than the work itself @xcite : @xmath3 where @xmath4 reciprocal temperature .the jarzynski equation ( [ equ : jarzynski ] ) has been proven for several types of dynamics ranging from newtonian and thermostatted dynamics to langevin and monte carlo dynamics @xcite .essentially , the theorem is valid if the dynamics preserves the canonical distribution for fixed control parameter @xcite .remarkably , the validity of the jarzynski equality does not depend on how the control parameter is switched from its initial to its final value .( in particular , it is not even necessary to maintain a fixed protocol throughout the simulation or the experiment as long as the initial and final value of the control parameter are fixed . )thus , the switching process can be carried out at arbitrary speed . for infinitely slow switching, jarzynski s method reduces to kirkwood s coupling parameter method , or thermodynamic integration @xcite , while for infinitely fast switching one obtains zwanzig s thermodynamic perturbation method @xcite .thus , these two well known free energy calculation methods can be viewed as limiting cases of the fast switching procedure . whether the freedom derived from the arbitrariness of the switching protocol can be used to develop more efficient free energy calculation methodsis an interesting and practical question .a recent analysis indicates that the statistical errors encountered in evaluating exponential averages such as that of eq .( [ equ : jarzynski ] ) prevent fast switching methods from exceeding the efficiency of conventional methods such as umbrella sampling or thermodynamic integration @xcite .this conclusion seems to remain true even if path sampling techniques are used to focus on those non - equilibrium pathways that yield the largest contribution to jarzynski s exponential average @xcite .note that these path sampling approaches can also be combined with the large time step method for fast switching simulations @xcite .in this paper we analyze the efficiency of several fast switching methods including the path sampling approaches mentioned above for two analytically solvable models : a particle dragged through a viscous fluid in a harmonic trap @xcite and an ideal gas in an expanding piston @xcite .for these two models the work distributions can be calculated analytically for arbitrary switching speeds . using these work distributions , we determine the statistical errors and , from them , the computational efficiency that one would obtain in fast switching simulations of these systems .the results presented in this paper confirm our previous conclusions based on actual simulations @xcite .namely , fast switching simulations do not outperform umbrella sampling and thermodynamic integration in most interesting cases .the remainder of the paper is organized as follows .simulation methods are described in sec .[ sec : methods ] .expressions for error and efficiency estimates are presented in secs .[ sec : error ] and [ sec : efficiency ] . in sec .[ sec : results ] these expressions are then applied to the two model systems .conclusionss are provided in sec .[ sec : conclusion ] .in this section we review the fast switching methods whose efficiency we will analyze in later sections . to set the notation , consider a system with a hamiltonian @xmath5 that is a function of the state of the system @xmath6 and the external control parameter @xmath7 . depending on the system , @xmath6 may include positions and momenta of all particles or the positions only .the free energy of the system at temperature @xmath8 and fixed control parameter @xmath7 is given by @xmath9 where the normalizing factor @xmath10 is the partition function of the system ( up to a constant factor ) .the free energy difference between two states @xmath11 and @xmath12 corresponding to two different values of the control parameter , @xmath13 and @xmath14 , respectively , is given by @xmath15 state @xmath11 is described by the hamiltonian @xmath16 and state @xmath12 by @xmath17 .we now imagine that the control parameter @xmath7 is continuously switched from the initial value @xmath13 to the final value @xmath14 according to some protocol @xmath18 within a total time @xmath19 . as the control parameter changes , the system evolves in time tracing out a trajectory @xmath20 . here, @xmath20 denotes a complete trajectory describing the system from time @xmath21 to time @xmath22 .the change in energy that is due to the changing control parameter is the work @xmath0 carried out on the system along the particular trajectory @xmath20 : @xmath23=\int^\tau_0 dt\ ; \dot \lambda(t ) \left .\frac{\partial { \cal h}}{\partial \lambda}\right|_{x = x(t)},\ ] ] where @xmath24 .this notation emphasizes that the work @xmath0 depends both on the specific trajectory followed by the system as well as on the particular protocol used to switch the control parameter . in pathintegral notation , the jarzynski equality can be written as : @xmath25 e^{-\beta w[x(t ) , \lambda(t)]}.\ ] ] here , @xmath26 denotes a summation over all trajectories and @xmath27 $ ] is the probability density of trajectory @xmath20 , along which the work @xmath28 $ ] is performed during the switching process .the probability density @xmath27 $ ] includes the canonical distribution @xmath29/q_{\lambda_a}$ ] of the initial conditions @xmath30 .for deterministic dynamics , the trajectory @xmath20 and the work @xmath28 $ ] are fully described by the initial point in phase space @xmath30 . in this case , the integral over all trajectories can be written as an integral over phase space rather than an integral over trajectory space . in the following, we will often omit the explicit dependence of @xmath31 $ ] and @xmath32 $ ] on @xmath33 to simplify the notation .the jarzynski equation ( [ equ : jarzynski_derivation ] ) justifies the following algorithm to estimate the free energy difference between the two equilibrium states @xmath11 and @xmath12 corresponding to @xmath13 and @xmath14 , respectively .first , one generates @xmath34 initial conditions canonically distributed with respect to @xmath35 .this can be done with monte carlo sampling or constant temperature molecular dynamics . from each of the @xmath34 initial condition @xmath36one then generates a trajectory of length @xmath19 by integrating the appropriate equations of motion and calculates the work @xmath37 performed on the system along that trajectory . from this sample of @xmath34 trajectoriesan estimate @xmath38 of the free energy is then determined : @xmath39 as the trajectory sample is finite , this free energy estimate contains errors @xcite that are discussed in detail in sec .[ sec : error ] . due to the highly non - linear behavior of the exponential functionthese errors may be severe , often precluding an accurate free energy calculation particularly in the case of large switching rates @xcite .the reason for these large deviations is best perceived by considering the work distribution @xmath40 : @xmath41 \delta ( \tilde w - w[x(t)]).\ ] ] in terms of @xmath40 , the jarzynski equation can be rewritten as an integral over the work : @xmath42 in a fast switching simulation as described above , most trajectories have work values near the values for which @xmath40 is a maximum . for large switching rates ,@xmath40 may be centered at large work values , such that for typical work values @xmath43 is a very small number yielding a vanishing contribution to the integral in eq .( [ equ : integral ] ) . the work values leading to the important contributions to the integral ,on the other hand , may occur very rarely @xcite .this situation , familiar from the calculation of chemical potentials with widom s particle insertion method @xcite or from zwanzig s perturbation approach @xcite , may lead to extremely large statistical inaccuracies .these statistical difficulties can easily offset the advantage of using short , computationally inexpensive non - equilibrium trajectories . to avoid large statistical errors occurring in the straightforward application of the jarzynski relation ,sun proposed a method based on a thermodynamic integration in trajectory space @xcite .the basic idea of this method is to introduce a bias that favors the sampling of the rare but important trajectories that mostly contribute to the exponential work average .this is done by introducing a free energy @xmath44 that depends on the parameter @xmath45 : @xmath46e^{-\beta \alpha w[x(t)]}. \label{equ : falpha}\ ] ] for @xmath47 , the parameter dependent free energy vanishes because @xmath32 $ ] is normalized , @xmath48 . for @xmath49 , on the other hand, the parameter dependent free energy is identical to the original free energy , @xmath50 .one can therefore calculate @xmath1 by taking the derivative of @xmath44 and integrating it from @xmath51 to @xmath52 : @xmath53 to carry out this procedure , which is the path space version of the thermodynamic integration method @xcite , one needs the derivative of @xmath44 with respect to @xmath45 : @xmath54 e^{-\beta \alpha w[x(t)]}w[x(t)]}{\int\!\mathcal{d}x(t){\cal p}[x(t)]e^{-\beta \alpha w[x(t)]}}.\ ] ] the right hand side of this equation can be viewed as the average @xmath55 of @xmath31 $ ] over the work biased path ensemble @xmath56={\cal p}[x(t)]e^{-\beta \alpha w[x(t ) ] } / z_\alpha , \label{equ : pwork}\ ] ] where @xmath57 e^{-\beta \alpha w[x(t)]}.\ ] ] in terms of the work average @xmath55 , eq.([equ : intalpha ] ) can be rewritten as : @xmath58 to calculate the free energy difference @xmath1 using this equation it is first necessary to determine @xmath55 for different values of the parameter @xmath45 between 0 and 1 .then , the integral on the right hand side is determined numerically from these values .as shown by sun @xcite , path sampling techniques , originally developed to study rare but important events @xcite , can be used to sample the work biased ensemble and calculate the averages @xmath55 . to understand why the work biased thermodynamic integration yields accurate free energy estimates , it is instructive to consider the work distributions @xmath59 in the work biased ensembles : @xmath60 \delta ( \tilde w - w[x(t ) ] ) \nonumber \\ & = & \frac{p(\tilde w)\exp(-\beta \alpha \tilde w)}{\int dw p(w)\exp(-\beta\alpha w ) } \label{equ : pw_work_biased}\end{aligned}\ ] ] for @xmath47 the parameter dependent work distribution equals the original work distribution : @xmath61 in the other limit , at @xmath49 , the parameter dependent work distribution is proportional to the integrand in eq .( [ equ : integral ] ) : @xmath62 for values of the parameter @xmath45 between 0 and 1 the work distribution in the work biased ensemble is intermediate between these two limiting cases . thus ,both the typical and the dominant work values are generated with comparable likelihood as @xmath45 is changed from 0 to 1 yielding a good statistical accuracy . an alternative way to enhance the sampling of rare but important work valueswas recently proposed by ytreberg and zuckerman @xcite and athnes @xcite .the basic idea of this approach , called single - ensemble nonequilibrium path - sampling by ytreberg and zuckerman , is to introduce an explicit _ biasing function _ ( or _ umbrella function _ ) @xmath63 $ ] depending on the path @xmath20 . with this umbrella function the jarzynski equality ( [ equ : jarzynski ] )can be rewritten as : @xmath64\pi[x(t)]\left[\frac{e^{-\beta w[x(t ) ] } } { \pi[x(t)]}\right]}{\int { \cal d}x(t)\ ; { \cal p}[x(t)]\pi[x(t ) ] \left[\frac{1}{\pi[x(t)]}\right ] } \nonumber \\ & = & \frac{\langle e^{-\beta w[x(t)]}/\pi[x(t ) ] \rangle_\pi}{\langle 1/\pi[x(t)]\rangle_\pi}. \end{aligned}\ ] ] here , @xmath65 denotes an average over the biased path ensemble @xmath66=\frac{{\cal p}[x(t)]\pi[x(t)]}{\int { \cal d}x(t)\ ; { \cal p}[x(t)]\pi[x(t)]}.\ ] ] for a path observable @xmath67 $ ] depending on the pathway @xmath20 the average in the biased path ensemble is given by : @xmath68 \rangle_\pi\equiv \frac { \int { \cal d}x(t)\ ; { \cal p}[x(t)]\pi[x(t)]a[x(t ) ] } { \int { \cal d}x(t)\ ; { \cal p}[x(t)]\pi[x(t)]}.\ ] ] such averages can be calculated using path sampling techniques @xcite .since the umbrella function is introduced to guide the sampling to the relevant work values , it suffices to let @xmath63 $ ] depend on the work only , @xmath63=\pi[w(x(t))]$ ] . in this case , the biased average of a function @xmath69 $ ] that also depends on the work only can be written as @xmath70 this expression will be useful in the next section , in which we discuss the statistical errors occurring in fast switching simulations . in a work biased umbrella sampling simulationthe choice of the umbrella function @xmath63 $ ] is crucial . to obtain a good accuracy in the free energy estimate ,it is necessary that the bias function has a good overlap with the unbiased work distribution @xmath40 as well as with the integrand of eq .( [ equ : integral ] ) , @xmath71 .ytreberg and zuckerman found that the umbrella function @xmath72 leads to considerable efficiency increases compared to the results of straightforward fast switching simulations .another possibility would be to carry out umbrella sampling simulations with partially overlapping windows @xcite .note that other non - boltzmann sampling techniques such as flat histogram sampling @xcite , multicanonical sampling @xcite , or parallel tempering @xcite may also be used to improve the convergence of the exponential average of the jarzynski equation .these techniques , however , are not considered in the present article .in this section we review the expressions required to estimate the statistical errors arising in fast switching simulations .while straightforward fast switching and work biased umbrella sampling can be treated in the same way , separate expressions are required for the work biased thermodynamic integration scheme . to set the notation we reproduce here the expressions derived in ref .all expressions are valid both for work biased umbrella sampling as well as for straightforward fast switching .the latter case follows by setting the umbrella function equal to unity , @xmath63 = 1 $ ] . for conveniencewe define @xmath73\equiv \frac{e^{-\beta w[x(t)]}}{\pi[x(t ) ] } \quad \mbox{and } \quad y[x(t)]\equiv \frac{1}{\pi[x(t)]}.\ ] ] the free energy estimated according to eq .( [ equ : biasave ] ) from @xmath34 pathways is @xmath74 where @xmath75 and @xmath76 are averages over the @xmath34 trajectories of the simulation , @xmath77 here , @xmath78 and @xmath79 are the values of @xmath80 and @xmath81 associated with the @xmath82-th path sampled from the biased ensemble , @xmath83 $ ] . due to the non - linearity of the logarithm ,this estimator for the free energy is biased , i.e. , the average over many realizations @xmath84 of the free energy estimator , @xmath85 differs from the true free energy difference .this deviation , @xmath86 is called the bias .in addition to this systematic deviation , the estimator @xmath38 is also affected by statistical errors which are quantified by the variance , @xmath87 ^ 2\rangle= \langle ( \delta \overline{f}_n)^2\rangle - \langle \delta \overline{f}_n\rangle^2.\ ] ] the total mean squared deviation of the estimator @xmath38 from the true free energy difference @xmath1 has contributions from the bias and the statistical error : @xmath88 ^ 2\rangle = b_n^2+\sigma_n^2\ ] ] we now consider the limit of a large sample size @xmath34 and first calculate the bias @xmath89 . for large @xmath34 , the deviations @xmath90 and @xmath91 from their averages @xmath75 and @xmath76 , @xmath92 are small compared to @xmath93 and @xmath94 , respectively .expansion of the logarithm in eq .( [ equ : df_finite ] ) into a taylor series around @xmath95 and truncation after the quadratic term ( the linear terms vanish after averaging ) yields the bias @xmath96 , \label{equ : bias_approx}\ ] ] where the fluctuations @xmath97 and @xmath98 are given by @xmath99 interestingly , eq .( [ equ : bias_approx ] ) implies that the bias can be made to vanish by choosing an appropriate umbrella function . for the variance @xmath100 and the mean squared error @xmath101 the taylor series of the logarithm in eq .( [ equ : df_finite ] ) can be truncated after the linear term . in this approximationthe bias vanishes and the variance and the mean squared error are identical : @xmath102 .\label{equ : epsilon}\end{aligned}\ ] ] in the absence of an umbrella function @xmath103 ( or , in other words , for @xmath104 ) these expressions describe the errors encountered in straightforward fast switching simulations and are identical to the ones derived in refs . @xcite and @xcite .all terms occurring in the above expressions for the bias @xmath89 , the variance @xmath100 , and the mean squared error @xmath101 can be calculated as integrals involving the work distribution @xmath40 : @xmath105,\\ \langle y^2 \rangle_\pi & = & \frac{1}{i}\int dw\ ; p(w)\left[\frac{1}{\pi(w)}\right],\\ \langle xy \rangle_\pi & = & \frac{1}{i}\int dw\ ; p(w ) \left[\frac{\exp(-\beta w)}{\pi(w)}\right ] .\label{equ : integral_l}\end{aligned}\ ] ] in section [ sec : results ] we will use these equations to calculate the expected errors for two systems with analytically known work distributions . in the following we will also consider the errors in the free energy if the process is carried out in reverse direction , i.e. , if the system starts from equilibrium initial conditions corresponding to the final value of the control parameter and the system evolves under a time - inverted switching protocol . in general , these errors differ from those of the forward process. however , in the case of umbrella sampling with bias @xmath106 the errors of the forward and reverse process are identical . to show this we consider the work distribution @xmath107 of the reverse process , which is related to the work distribution of the forward process by the crooks identity @xcite @xmath108 insertion of eq .( [ equ : crooks ] ) into eqs .( [ equ : integral_f ] ) to ( [ equ : integral_l ] ) yields @xmath109 here , @xmath110 and @xmath111 denote averages with umbrella function @xmath112 for the forward and the reverse process , respectively .thus , the fluctuations of @xmath80 in the forward process are identical to the fluctuations of @xmath81 in the reverse process and vice versa .the correlation term is the same for both directions . inserting these results into eq .( [ equ : epsilon ] ) one finally obtains @xmath113 i.e. , the mean squared errors for the forward and reverse process are identical . in the same wayone can show that also for the @xmath114-bias the errors are the same in forward and reverse direction . for straightforward fast switching eq .( [ equ : epsilon ] ) is very similar to an expression obtained recently by jarzynski that relates the number of trajectories required to obtain acceptable accuracy to the dissipative work of the reverse process @xcite .to establish this similarity we use eq .( [ equ : epsilon ] ) to determine the number @xmath115 of trajectories required to obtain a statistical error of thermal magnitude @xmath116 : @xmath117 using the crooks identity , eq .( [ equ : number ] ) can be written as @xmath118 where we have neglected the term -1 on the right hand side of eq .( [ equ : number ] ) ( in any interesting case @xmath34 is much larger than 1 ) . changingthe integration variable from @xmath0 to @xmath119 one obtains : @xmath120 \nonumber \\ & = & \langle \exp(\beta w_r^d ) \rangle_r , \label{equ : diss}\end{aligned}\ ] ] where @xmath121 is the dissipative work for the reverse transformation and @xmath122 indicates an average over realizations of the reverse process .this equation implies that the number of fast switching trajectories required for an accurate free energy estimation depends on the work dissipated during the reverse process , as noted earlier by jarzynski @xcite .remarkably , for highest accuracy of the free energy estimate the process should be carried out in the direction that yields the larger exponential average of the dissipative work @xcite .of course , the same conclusion can be drawn from @xmath123 which also simply follows from eq .( [ equ : number ] ) and where @xmath124 is the dissipative work for the forward process . to simplify eq .( [ equ : diss ] ) , one is tempted to replace the average of the exponential with the exponential of the average @xcite , @xmath125 where @xmath126 is the average dissipative work for the reverse process .this approximation , however , is not valid in general . as shown for an illustrative example in sec .[ sec : pulled ] , it is likely to underestimate the required number of trajectories that may be off by orders of magnitude from the true value . according to eq .( [ equ : ti ] ) , sun s work biased thermodynamic integration method requires the calculation of an integral over the parameter @xmath45 . to determine this integral ,the average @xmath55 is first calculated for @xmath127 discrete values @xmath128 of the parameter @xmath45 between 0 and 1 in @xmath127 separate simulations .then , the integration is carried out numerically using for instance the trapezoidal or simpson s rule .imagine now that the average work @xmath55 is approximated by averaging over @xmath129 trajectories at each value of the parameter @xmath45 : @xmath130 here , the argument @xmath45 in @xmath131 indicates that the trajectories over which the average is carried out are sampled from the trajectory ensemble corresponding to the parameter @xmath45 . using the trapezoidal integration rule the estimate for the free energy is then given by @xmath132 where @xmath133 is the total number of trajectories generated in the simulation . neglecting errors in the numerical integration over @xmath45this free energy estimate is bias free : @xmath134 where we have used that @xmath135 . in the above equation the angular brackets @xmath136 denote an average over many independent realizations of the thermodynamic integration procedure . to determine the variance @xmath137 of the free energy estimate , which in this case is equal to the mean squared error @xmath138, we note that the variance of a sum of random variables equals the sum of the variances of the individual variables .accordingly , the variance of @xmath38 is given by @xmath139 where @xmath140 ^ 2\rangle\ ] ] is the variance of the average @xmath141 at parameter value @xmath128 .for statistically independent work values this variance is given by @xmath142\ ] ] rewriting the sum over @xmath143 in eq .( [ equ : variance_df ] ) as an integral over @xmath45 we finally obtain : @xmath144 d\alpha.\ ] ] note that here we have assumed that at each value of the parameter @xmath45 , the same number @xmath129 of trajectories are generated . to minimize the variance @xmath137 , however , it is advantageous to sample more trajectories at values of @xmath45 where the work variance is large and less where it is small .for simplicity we do not consider this case here .for a given work distribution @xmath40 the work variance in the work biased ensemble can be calculated from @xmath145 and @xmath146 by numerical integration of eq .( [ equ : var_dfn ] ) one can then calculate the variance @xmath137 of the free energy estimate @xmath38 .interestingly , also the error of the thermodynamic integration method for the reverse process is identical to that for the forward process .the average of an arbitrary function @xmath147 for a particular @xmath45 of the reverse process is @xmath148 inserting the crooks identity , eq .( [ equ : crooks ] ) , into eq .( [ equ : reverseaverage ] ) we obtain @xmath149 thus , the work variance for the forward process at @xmath45 is equal to the work variance for the reverse process at @xmath150 : @xmath151 integration over @xmath45 according to eq .( [ equ : var_dfn ] ) with a change of variables from @xmath45 to @xmath152 in the reverse case then demonstrates that the statistical errors for the forward and reverse process are indeed equal : @xmath153while the error obtained from a simulation with @xmath34 trajectories is an interesting quantity , the criterion most relevant for the practitioner is the computational cost required to achieve a given accuracy in the free energy . for the estimation of this cost one needs to take into account that long trajectories are computationally more expensive than short ones . in order to do that we note that for all algorithms studied here the error in the free energy can be written as : @xmath154 where @xmath155 is a constant that does not depend on the number of trajectories used in the calculation .hence the number @xmath115 of trajectories required to obtained an accuracy of @xmath116 in the free energy is given by @xmath156 here , we have set the target error to a value of @xmath116 because variations of the free energy much smaller than @xmath116 are irrelevant in most cases .since the computational cost of a trajectory is proportional to its length @xmath19 a meaningful definition of the computational cost is given by @xmath157 @xmath158 is the computational cost of a free energy calculation with accuracy @xmath116 measured in units of the computational cost to generate a trajectory of length @xmath159 .while for straightforward fast switching and umbrella sampling the computational cost is @xmath160,\ ] ] for the work biased thermodynamic integration scheme is given by : @xmath161if the work distribution @xmath40 is known for a specific switching process , the equations of the previous sections can be used to determine the expected accuracy of the free energy calculated with the methods of sec .[ sec : methods ] .here we do that for two models , for which the work distributions are known analytically . in particular , we study the computational efficiency of these methods in terms of the computing time required to obtain a given accuracy in the calculated free energy . the first model we consider is a particle in one dimension pulled through a viscous liquid at temperature @xmath8 by a harmonic trap with force constant @xmath162 moving at constant speed @xmath163 for a certain distance @xmath127 .the translation of the trap by the distance @xmath127 takes a time @xmath164 .the motion of the particle is modelled by a langevin equation in the overdamped limit with friction constant @xmath165 : @xmath166 where @xmath6 is the position of the particle and @xmath167 is a gaussian random variable uncorrelated in time , @xmath168 . for this model , mazonka and jarzynskihave computed the work distribution analytically finding a gaussian distribution @xcite : @xmath169\ ] ] with mean @xmath170\ ] ] and variance @xmath171=\frac{2}{\beta}\overline{w}. \label{equ : wave}\ ] ] since the free energy of the system does not depend on the position of the harmonic trap , @xmath172 for this transformation .the average work , however , is larger than zero due to the energy dissipated into the bath and only in the limit of infinitely slow switching does the average work vanish . for slow switching ,i.e. , for @xmath173 or @xmath174 , the average work is proportional to the pulling velocity , @xmath175 , as predicted by linear response theory . in the limit of infinitely fast switching, i.e. for @xmath176 or @xmath177 , the average work converges to a constant value , @xmath178 .accordingly , the variance @xmath179 for slow switching . in the fast switching regimethe variance reaches a constant value , @xmath180 ( see fig . [fig : sig2 ] ) .all results shown in this section were obtained for @xmath181 , @xmath182 , @xmath183 and @xmath184 . for a particle pulled through a viscous fluid by a harmonic trap moving at constant speed @xmath163 .the average work is given by @xmath185.,width=264 ] for this model we have calculated the expected errors for straightforward fast switching ( i.e. , @xmath186 ) as well as for the two bias functions @xmath187 , and @xmath188 . the bias function @xmath187 was suggested by ytreberg and zuckerman and is designed to improve the sampling of low work values @xcite .some attention is needed when considering the umbrella function @xmath189 .since this function biases the simulation in a way to make the work distribution flat , @xmath189 can have the form @xmath190 only in a range with finite size .we therefore define @xmath191 the limits for this range , @xmath192 and @xmath193 , need to be selected such that all important work values are included in the sampling . of course , in the general case the work distribution @xmath40 required to implement the umbrella function @xmath188 is not known in advance .however , methods such as flat histogram sampling @xcite can be used to determine this bias iteratively . here , we study this case in order to evaluate the efficiency of the work biased umbrella sampling approach if a good umbrella function is available . for the gaussian work distribution of the pulled harmonic oscillator and the bias functions considered here all integrals in eqs .( [ equ : integral_f ] ) to ( [ equ : integral_l ] ) can be computed analytically .these analytical expressions can then be used to calculate the expected errors for straightforward fast switching , @xmath194 work biased umbrella sampling with the exponential umbrella function @xmath195 , @xmath196 and with the umbrella function @xmath188 , @xmath197 here , the interval size @xmath198 depends on the selection of the limits @xmath193 and @xmath192 . as a measure for the interval @xmath199 in which the work distribution is required to be flat we use the distance between the maximum of the work distribution @xmath40 and the maximum of the jarzynski integrand @xmath71 . for the gaussian work distributions considered here this distance is @xmath200 and we set @xmath201 accordingly . with this choice ,the statistical error of the @xmath114 bias becomes : @xmath202 the expected errors are shown in fig .[ fig : gauss_err ] as a function of the switching rate @xmath163 . since in all expressions for the error the number @xmath34 of trajectories factors out in a simple way , we have represented @xmath203 as a function of @xmath163 in this figure .the number @xmath115 of trajectories required to obtaine an error of @xmath116 is related to @xmath204 simply by @xmath205 . we also calculated the expected errors for sun s work biased thermodynamic integration scheme .direct evaluation of the work distribution @xmath59 [ see eq .( [ equ : pw_work_biased ] ) ] in the work biased ensemble shows that for a gaussian distribution @xmath40 the variance of the work biased distribution @xmath206 is the same for all values of @xmath45 . in this case( [ equ : var_dfn ] ) yields @xmath207 while the variance in the work biased ensemble does not change as a function of @xmath45 , the average work @xmath55 does : @xmath208 the expected errors for the work biased thermodynamic integration method are also shown in fig .[ fig : gauss_err ] as a function of the switching rate @xmath163 . while for straightforward fast switching and for umbrella sampling with exponential bias the mean squared error grows exponentially with the dissipative work , it increases with @xmath209 for the @xmath114 bias and linearly for thermodynamic integration .for the particle pulled through a viscous fluid as a function of switching rate @xmath163 obtained for straightforward fast switching ( fs ) , work biased thermodynamic integration ( ti ) , and work biased umbrella sampling with the umbrella functions @xmath210 and @xmath211 . here , @xmath181 , @xmath182 , @xmath183 and @xmath184 ., width=264 ] as can be seen in fig .[ fig : gauss_err ] , in the slow switching regime the errors decrease with decreasing @xmath163 in all four cases . to linear approximation in @xmath163 , the error for straightforward fast switching , work biased umbrella sampling with the exponential bias @xmath195 , and work biased thermodynamic integrationis identical and given by @xmath212 for work biased umbrella sampling with bias @xmath213 the error is smaller and the dependence on @xmath163 stronger : @xmath214 as one proceeds to larger switching rates , the error grows most rapidly for straightforward fast switching . for work biased umbrella sampling with the exponential umbrella functionthe growth is less pronounced and occurs at higher switching rates .for the two remaining cases ( work biased umbrella sampling with @xmath211 and work biased thermodynamic integration ) the growth of the error slows down rather then speeding up with increasing switching rate .since the work variance becomes constant for large @xmath163 , the error reaches a constant value in the fast switching regime in all four cases . in this limitwe obtain @xmath215 for straightforward fast switching , @xmath216 for work biased thermodynamic integration , @xmath217 for work biased umbrella sampling with the exponential umbrella function @xmath195 and @xmath218 in the case where the umbrella function @xmath213 .note , in fig .[ fig : gauss_err ] , that these constant values may differ by orders of magnitude .for the particle pulled through a viscous fluid as a function of switching rate @xmath163 obtained for straightforward fast switching ( fs ) , work biased thermodynamic integration ( ti ) , and work biased umbrella sampling with the umbrella functions @xmath210 and @xmath211 .here , @xmath181 , @xmath182 , @xmath183 and @xmath184.,width=264 ] the computational cost @xmath158 determined for the four algorithms studied here is shown as a function of the switching rate in fig .[ fig : gauss_err_nu ] . in the slow switching regimethe computational cost becomes constant for straightforward fast switching , work biased umbrella sampling with exponential bias and work biased thermodynamic integration .this behavior indicates that in these cases running few long trajectories or many shorter trajectories yields the same efficiency .the situation is different for work biased umbrella sampling with the @xmath190 umbrella function . in this case , @xmath219 keeps decreasing for decreasing @xmath163 even in the slow switching regime . for larger switching rates straightforward fast switching and work biased umbrella sampling with exponential biasbecome very inefficient . in the other two cases , increasing the switching rate enhances the efficiency . with the appropriate biasthe optimum efficiency is obtained in the instantaneous switching limit , in which case the path based sampling methods reduce to ordinary umbrella sampling and thermodynamic integration . thus , conventional free energy calculation methods are superior to fast switching methods for the model studied in this section .similar results have been obtained earlier for more complex models @xcite .in addition to the expected error also the bias @xmath89 is an important quantity determining the accuracy of a free energy calculation . according to eq .( [ equ : bias_approx ] ) the bias can be calculated using the expressions in eqs .( [ equ : integral_f ] ) to ( [ equ : integral_l ] ) . for the pulled particle with gaussian work distribution one obtains @xmath220 in the case of straightforward fast switching . as can be seen by direct evaluation of eq .( [ equ : bias_approx ] ) , all other algorithms are bias free , i.e. , @xmath221 , if the work distribution is gaussian .it is interesting that in the case of work biased sampling with exponential umbrella function , the property of vanishing bias is only given for the particular form @xmath195 . to be more specific, one can consider umbrella functions @xmath222 , where @xmath7 is a real number between 0 and 1 . in this case ,error and bias are given by @xmath223\end{aligned}\ ] ] and @xmath224,\ ] ] respectively .the error @xmath101 from the above equation is a minimum for @xmath225 . for this value of the parameter@xmath7 the bias vanishes . in this sense, the umbrella function @xmath195 is the optimum umbrella function among all exponential functions of the form @xmath226 .note , however , that this property holds strictly only for processes with gaussian work distribution .we now briefly return to the question of whether in expression ( [ equ : diss ] ) for the number of trajectories required to obtain an error of about @xmath116 the average of the exponential can be safely replaced by the exponential of the average ( see eq .[ equ : wdiss ] ) . for the process considered in this section , the dissipative work as well asthe work distributions are the same in the forward and reverse direction .furthermore , @xmath227 because @xmath228 .hence , eq . ( [ equ : diss ] ) reduces to : @xmath229 where we have used eq .( [ equ : wave ] ) which is valid only for gaussian work distributions .thus , in this case , the estimate @xmath230 may be off by orders of magnitude . if , for instance , the correct estimate is @xmath231 , replacing the average of the exponential with the exponential of the average yields an estimate too low by a factor of @xmath232 . in this section , we will determine the efficiency of fast switching methods for the expansion of an ideal gas in a cylinder with moving piston .the work distribution for this process has been determined analytically by lua and grosberg @xcite , who demonstrated that the jarzynski identity holds exactly for arbitrary piston speed . in the limit of very rapid expansionthis seems surprising as almost no gas particles collide with the piston performing work on it .nevertheless , as shown by lua and grosberg , the tails of the maxwell - boltzmann distribution provide a sufficient number of fast moving particles that perform exactly the amount of work required for the jarzynski identity to be valid .the reason why we study the ideal gas expansion here is that its work distribution is non - gaussian .as the system is not coupled to a heat bath during the transition , the work distribution remains non - gaussian even in the limit of a piston that moves infinitely slowly . in this case , the distribution is a rapidly fluctuating function with an exponential envelope . in the fast switching limit ,the work distribution acquires a growing delta - peak at zero work and small contributions at large work values .these large but rare work values are the dominant contributions to the exponential work average . in their work ,lua and grosberg @xcite considered a single ideal gas particle in a cylinder with cross section area @xmath11 sealed by a piston on one side .the volume available to the gas particle is @xmath233 , where @xmath234 is the initial length of the cylinder . starting from initial conditions distributed canonically with temperature @xmath235 , the volume of the cylinderis then increased by moving the piston from position @xmath234 to a new position @xmath127 in a time @xmath19 with constant velocity @xmath236 .the ideal gas particle is assumed to collide elastically with the walls of the cylinder and the moving piston . at each elastic collision of the particle with the piston ,a certain amount of kinetic energy is transferred from the particle to the piston .this adiabatic case is in contrast to that considered in ref .@xcite , where the system is coupled to a thermal bath and the velocities are taken from an equilibrium distribution at each collision with the piston .the sum of these energy transfers during one expansion is the work @xmath0 carried out by the gas .note that here , as in lua and grosberg s work , @xmath0 denotes the work done _ by _ the system and not _ on _ the system .accordingly , we let @xmath237 denote the free energy difference between the initial and the final state .the maximum work theorem then implies that the average work performed by the system is bounded from above by the free energy difference , @xmath238 . with this sign convention ,the jarzynski equality is given by @xmath239 .lua and grosberg have analyzed the statistics of the particle - piston collisions in detail for the special case of inverse temperature @xmath240 , mass of the particle @xmath241 and switching time @xmath242 . in this paperwe keep the explicit dependence on all variables and obtain the following work distribution : @xmath243 where @xmath244 is the probability that the particle does not collide with the moving piston during the expansion process .this probability increases with increasing piston velocity @xmath245 leading to a growing delta - peak at @xmath246 in the work distribution .as shown by lua and grosberg , the number @xmath247 of collisions between the moving ideal gas particle and the piston is completely determined by the work @xmath0 done by the gas during the entire expansion and is given by : @xmath248 where @xmath249 indicates the floor function that gives the largest integer less or equal to the argument .the function @xmath250 appearing in eq .( [ equ : pw_gas ] ) , called the overlap factor by lua and grosberg , is given by the piecewise linear function @xmath251 a variation of the piston velocity @xmath245 is equivalent to a variation of the temperature in the sense of the following scaling relation : @xmath252 this equation , which can be easily derived directly from eq .( [ equ : pw_gas ] ) , holds , because due to the elastic hard collisions the same sequence of events occurs if both @xmath163 and @xmath245 are scaled by the same factor .( solid red line ) performed _ by _ the ideal gas during the expansion as a function of the piston velocity @xmath245 for @xmath253 , @xmath254 and @xmath255 .also shown is the work done _ on _ the ideal gas during the compression ( solid green line ) .the free energy difference @xmath256=\log(2)$ ] is shown as a dashed line . for @xmath257, the average work converges towards @xmath258=0.375$].,width=264 ] the average work @xmath259 performed during the expansion of the cylinder from @xmath234 to @xmath260 calculated from the work distribution eq .( [ equ : pw_gas ] ) by numerical integration is shown in fig .[ fig : figmeanworkjarzynski ] .it is clear from the picture that the average work @xmath261 is lower than the free energy difference @xmath262 $ ] even in the limit of infinitely slow switching . in this limit, the process studied by lua and grosberg can be viewed as the quasistatic adiabatic expansion of an ideal gas where the average work follows from basic thermodynamic equations for the ideal gas : @xmath263 , \end{aligned}\ ] ] where @xmath264 and @xmath265 is the number of degrees of freedom . in our one - dimensional case , @xmath266 and @xmath267 such that @xmath268 for fast switching , on the other hand , the average work converges to zero because less and less collisions of the moving particle with the piston occur as the piston velocity is increased .nevertheless , the jarzynski equality holds over the whole range of piston velocities as can be directly verified from eq .( [ equ : pw_gas ] ) . in the slow switching limit , i.e. , for @xmath269 , the work distributioncan also be obtained by considering adiabatic invariants @xcite . for a one - dimensional mechanical system the principle of adiabatic invariance states that the action variable @xmath270 is conserved under slow variation of an external parameter in the hamiltonian . in the above equationthe contour integral extends over one period of the motion for constant energy @xmath271 . for the lua - grosberg modelthis relation implies that the quantity @xmath272 is conserved during a slow , i.e. , adiabatic , transformation of the hamiltonian . as a consequence , the energy @xmath273 of the system at the end of an expansion from @xmath234 to @xmath127 is a unique function of the initial energy @xmath274 : @xmath275 thus , the work carried out by the gas starting from any initial condition with energy @xmath274 is given by @xmath276 $ ] .the work distribution in the slow switching limit can then be obtained as average over the canonically distributed initial conditions : @xmath277.\end{aligned}\ ] ] integration yields @xmath278 which is the one - particle one - dimensional special case of the work distribution derived by crooks and jarzynski for an ideal gas of @xmath34 particles @xcite .it is easy to verify directly that for this work distribution that the jarzynski equality holds and the average work @xmath279 performed by the gas is given by eq .( [ equ : work_adiabatic ] ) .note that even tough the expansion occurs infinitely slowly in this case , the work distribution is not gaussian and the width @xmath280 and the average @xmath261 of the work distribution are related by @xmath281 rather than by eq .( [ equ : wave ] ) . in the fast switching limit ,i.e. , for @xmath282 , only trajectories with 0 or 1 collision yield important contributions to the work distribution . in this case , one obtains @xcite : @xmath283 where @xmath284 is given by @xmath285 in the following paragraphs we will also consider the statistical errors for the reverse process , namely the compression of the ideal gas from the cylinder length @xmath286 to @xmath234 .the work distribution @xmath107 for the reverse process is simply given by the crooks relation eq .( [ equ : crooks ] ) , which , with the sign conventions for @xmath0 and @xmath1 used in this section reads : @xmath287 we next determine the efficiency of fast switching methods for the ideal gas expansion from @xmath234 to @xmath286 as well as for the compression from @xmath286 to @xmath234 . in this case, we investigate the straight forward approach , biased sampling with @xmath288 umbrella function , and thermodynamic integration .we do not , however , consider the @xmath114 bias here , because the work distribution for the ideal gas expansion includes a @xmath289-peak that creates problems in this case .when estimating statistical errors of fast switching simulations another difficulty arises in some situations . as detailed in sec .[ sec : error ] , the mean square error @xmath290 as well as the bias @xmath89 can be calculated from integrals over the work distribution . in the case of the ideal gas expansion , however , some of the integrals from eqs .( [ equ : integral_f])-([equ : integral_l ] ) diverge for certain expansion ratios . to see this ,consider the mean square error estimated for a straightforward fast switching simulation using the sign convention of this section : @xmath291 the average @xmath292 exists only if the work distribution @xmath40 decays sufficiently rapidly with increasing w to compensate for the fast growth of @xmath293 .for large work values @xmath0 , the work distribution for the ideal gas expansion behaves as @xmath294,\ ] ] where we have neglected all factors that are immaterial for the convergence of the integral .thus , @xmath295 is finite only if @xmath296 or , equivalently , if the ideal gas is expanded by no more than a factor of @xmath297 , @xmath298 .if this condition , which is independent of the switching rate , is not met , @xmath299 has infinite variance and the expressions from sec .[ sec : error ] can not be applied to estimate the statistical error of the free energy estimate . in this case, a generalization of the law of large numbers @xcite still guarantees that the free energy converges to the correct value , but the fat - tailed work distribution prevents gaussian statistics from arising even for large sample sizes . for umbrella sampling with the exponential bias @xmath288 a similar analysis of the average@xmath300 reveals that the corresponding integral converges only provided that @xmath301 . for larger expansion ratios , the integral diverges for any switching rate .in contrast , the integrals associated with the thermodynamic integration scheme always converge regardless of the expansion ratio @xmath302 and the switching rate .the situation is slightly different for the ideal gas compression . in this case ,straightforward fast switching yields error estimates that are finite for any expansion ratio . since for umbrella sampling with exponential bias @xmath288 and thermodynamic integrationthe errors in forward and backward direction are identical , the same limitations as for the ideal gas expansion apply in these cases . in the followingwe will study the efficiency for expansion ratios of @xmath303 and @xmath304 . for @xmath303the error estimates for all fast switching methods are finite both for the expansion and the compression .for @xmath305 we consider only the thermodynamic integration method and straightforward fast switching for the ideal gas compression .all other methods yield diverging error estimates for this expansion ratio . as a function of the piston velocity @xmath245 for different fast switching methods and an expansion ratio of @xmath303 .the black lines denote the straight forward fast switching errors for expansion ( solid line ) and compression ( dashed line).,width=264 ] figure [ fig : figerrorgrosberg ] shows the errors as a function of the piston velocity @xmath245 obtained by numerical integration of eqs .( [ equ : integral_f])-([equ : integral_l ] ) for an expansion from @xmath306 to @xmath307 . here and in the subsequent calculations the inverse temperature of the initial state was @xmath240 . for infinitely slow and infinitely fast switching the errors can be calculated analytically and were found to agree perfectly with the results obtain by numerical integration . as can be seen in fig .[ fig : figerrorgrosberg ] , the errors become independent of the switching rate @xmath245 for all three methods in the slow switching limit . in the fast switching limit , the error grows rapidly with @xmath245 particularly for straightforward fast switching and the exponential bias .the thermodynamic integration scheme , for which the growth is slower , yields smaller errors for all switching rates .for all three cases the errors diverge for @xmath308 .the overall smallest errors , however , are obtained if the reverse process , i.e. , the compression , is simulated with straightforward fast switching ( dashed line in fig .[ fig : figerrorgrosberg ] ) . in the fast switching limitthis error becomes @xmath309 .since the initial position of the particle is uniformly distributed in the interval @xmath310 $ ] , this value is simply the probability that the fast moving piston hits the particle at least once . note that for thermodynamic integration and umbrella sampling with the exponential bias the errors are identical for expansion and compression . from the errors depicted in fig .[ fig : figerrorgrosberg ] one can determine the computational cost required to obtain an accuracy of @xmath116 in the free energy estimate [ see eq .( [ equ : ccpu ] ) ] .this computational cost is shown in fig .[ fig : figccpugrosberg ] for different fast switching methods and an expansion ratio of @xmath303 as a function of the switching rate . for straightforward fast switching in the expansion direction , thermodynamic integration and umbrella sampling with the exponential bias an optimum switching rate of about @xmath311 exists . in the case of straightforward fast switching in the compression direction ,however , no such minimum exists and the computational cost keeps decreasing for increasing switching rate . as a function of the switching rate @xmath245 for different fast switching methods and and expansion ratio of @xmath303 .the black lines denote the straight forward fast switching errors for expansion ( solid line ) and compression ( dashed line ) ., width=264 ] as explained above , for an expansion ratio of @xmath305 our expressions yield finite statistical errors only for the thermodynamic integration method and for straightforward fast switching in the compression direction . the errors for these cases are shown in fig .[ fig : figerrorgrosbergl2 ] . straightforward fast switching of the ideal gas compression yields errors that are smaller than the thermodynamic integration errors for all piston velocities .both in the slow and fast switching limit the straightforward fast switching errors become constant . accordingly , the computational cost of the straightforward compression is decreasing as @xmath312 in the fast switching limit while the thermodynamic integration error goes through a minimum at about @xmath313 and then rapidly grows for increasing piston velocity @xmath245 ( see fig .[ fig : figccpugrosbergl2 ] ) .thus both for @xmath303 and @xmath304 the straightforward fast switching simulation of the ideal gas compression is the most efficient among the methods studied here .however , optimum efficiency is obtained for infinitely fast compression , in which case straightforward fast switching reduces to zwanzig s perturbative approach applied to configurations rather than trajectories @xcite . as a function of the piston velocity @xmath245 for the ideal gas expansion / compression with an expansion ration of @xmath304 .the dashed black line denotes the straightforward fast switching errors ( fs ) and the solid red line the thermodynamic integration errors ( ti).,width=264 ] as a function of the switching rate @xmath245 for the ideal gas expansion / compression with an expansion ratio of @xmath305 .the dashed black line denotes the straight forward fast switching results ( fs ) for the compression and the solid red line the thermodynamic integration results ( ti ) ., width=264 ] the model discussed in the previous section consisted of a single particle in one dimension . in this section, we study the expansion of an ideal gas of @xmath129 particles and analyze how the error depends on the number of particles .the system of strictly non - interacting particles considered here is similar , but not identical to the dilute gas of weakly interacting particles studied by jarzynski and crooks in ref .for the @xmath129-particle ideal gas , the work performed on the piston during the expansion is the sum of the contributions of each particle . @xmath314 here @xmath315 is the work performed on the system by particle @xmath82 .according to eq .( [ equ : epsilon ] ) , the statistical error in the free energy calculated with straightforward fast switching from @xmath34 repetitions of the process is given by @xmath316.\ ] ] for the @xmath129-particle system , the averages @xmath317 and @xmath318 are @xmath319 and @xmath320 where @xmath321 and @xmath322 are the corresponding averages of the one - particle system .here , we have used the fact that work contributions of the single particles are identical and statistically independent .the error @xmath323 for the @xmath129-particle system becomes @xmath324.\ ] ] the fraction @xmath325 is always larger than 1 and therefore the error @xmath326 diverges exponentially with the particle number in the case of straightforward fast switching .a similar analysis can be carried out for the error of work biased umbrella sampling simulations .for an umbrella function @xmath327 that factorizes into a product of single particle umbrella functions @xmath328 ( this is for instance the case for the exponential bias @xmath329 ) , @xmath330 the error is given by : @xmath331.\end{aligned}\ ] ] here , as in the previous paragraph , the averages are single - particle averages . for large @xmath129 , this expression will be dominated by either @xmath332 or @xmath333 depending on whether @xmath334 or @xmath335 is larger .( it is easy to see that the larger one of @xmath334 and @xmath335 is also larger than @xmath336 , which is always positive . ) thus , also in this case , the error in the free energy grows exponentially with the system size @xmath129 . for the work biased thermodynamic integration the error results from an integral over the work variance at different values of an auxiliary parameter @xmath45 ( see eq .( [ equ : var_dfn ] ) ) . since in the @xmath129-particle ideal gasthe work @xmath0 is a sum of @xmath129 independent contributions , the work variance of the entire system is @xmath129 times the single particle work variance for each value of @xmath45 .hence , the @xmath129-particle error @xmath337 scales with the square root of the number of particles , @xmath338 where @xmath339 is the error for the one - particle gas .due to this kind of scaling , the thermodynamic integration method is superior to all other fast switching methods studied here for a strictly non - interacting ideal gas .for irreversible transformations , the exponential average of eq .( [ equ : jarzynski ] ) is dominated by rare , but important work values causing possibly large statistical inaccuracies in fast switching free energy computations .several path sampling methods have proposed that aim at reducing these inaccuracies by enhanced sampling of those trajectories that contribute most to the average . in this paper, we investigate how these methods perform in the case of two simple models for which the work distribution is known analytically . for gaussian work distributions ,the computational effort required to achieve a given accuracy in the free energy can be determined analytically . in the case of a particle pulled through a viscous fluid , flat histogram sampling as well as sun s work biased thermodynamic integrationperform best . for both methods optimum efficiencyis obtained in the limit of infinitely fast switching in which case these path sampling methods reduce to conventional flat histogram sampling and thermodynamic integration in configuration space rather than in trajectory space .the errors obtained in the case of the ideal gas expansion / compression , for which the work distribution was determined analytically by lua and grosberg @xcite , show a qualitatively different behavior .in contrast to the particle pulled through a fluid , the work distribution for the expansion / compression process does not approach a constant shape in the fast switching limit . rather , the work distribution keeps changing even for very large piston velocities . as a consequence, the computational cost of all methods goes through a minimum for intermediate piston velocities and grows for rapid expansion .the only exception here is the straightforward fast switching simulation of the gas compression , which is superior to the path sampling methods for all piston speeds . in this case , the computational cost keeps decreasing even for very large piston speeds where the fast switching simulation reduces to zwanzig s perturbative approach .it is interesting that for certain expansion ratios only thermodynamic integration and straightforward fast switching of the compression process yield finite error estimates , while all other method yield diverging errors .this behavior is due to the fat tails of the work distribution that occur for larger expansion ratios .the superiority of straightforward fast switching for the ideal gas compression is most likely peculiar to the one - particle case . in the many - particle gas , the errors scale exponentially with particle number for all methods except for thermodynamic integration , in which case they scale with the square root of the particle number .thus , for an ideal gas consisting of many particles sun s work biased thermodynamic integration method seems to offer the most efficient free energy calculations . since for this methodthe efficiency is maximum at intermediate piston velocities , this is an example where a fast switching method for the calculation of free energies is more efficient than conventional configuration space methods .in general , the efficiency of a fast switching simulation depends on whether the process is carried out in forward or reverse direction .in particular , a straightforward fast switching simulation is more efficient in the direction in which more work is dissipated @xcite as confirmed here for the lua - grosberg ideal gas expansion and compression .for all the path sampling methods studied here , however , it can be shown using the crooks identity that the work distribution is effectively symmetrized in a way that makes the corresponding efficiency identical in both directions . in summary ,path sampling methods such as work biased thermodynamic integration or work biased umbrella sampling may help to reduce the statistical errors encountered in the calculation of equilibrium free energies from non - equilibrium simulations .the examples studied in this paper indicate that often conventional free energy estimation methods are likely to be superior to fast switching simulation. nevertheless , advanced fast switching methods such as sun s work biased thermodynamic integration may be competitive particularly if a large number of degrees of freedom significantly contribute to the free energy change that one wants to compute .this work was initiated during a junior fellowship of w.l . at the international erwin schrdinger institute for mathematical physics ( esi ) and completed with support from the austrian science fund ( fwf ) under grant no .the authors acknowledge useful discussions with phillip geissler and elisabeth schll - paschinger .

<S> according to the jarzynski theorem , equilibrium free energy differences can be calculated from the statistics of work carried out during non - equilibrium transformations . </S><S> although exact , this approach can be plagued by large statistical errors , particularly for systems driven strongly away from equilibrium . </S><S> recently , several approaches have been suggested to reduce these errors . in this paper </S><S> we study the efficiency of these methods using two models for which analytical solutions exist . </S>

1007.1452

the problem of the collision of two shoockwaves in four dimensions ( with lorentz signature ) that result from boosting two black holes to the speed of light is well known and has been extensively studied in the literature .references @xcite provide only a subset of the work dedicated to the particular problem ( for some interesting recent developments see @xcite ) . in particular , @xcite deals with this problem pertubatively and in a series of three papers the authors compute the metric and derive a formula for the gravitational radiation .the collision is axisymmetric while the bulk matter that creates the shockwaves as well as the back - reaction effects are not taken into account .recently , the high energy physics community has also expressed special interest in the topic of shockwaves collisions in the framework of general relativity @xcite through the anti - de - sitter space / conformal field theory correspondence ( ads / cft ) @xcite .the ads / cft duality , allows to formulate a process that is associated with non - abelian gauge theories at strong coupling as a purely gravitational problem . as a result, one may map the problem of heavy ion collisions ( in four dimensions ) onto shockwaves collisions in gravity in five dimensions .recently , we have been involved in such problems @xcite in gauge theories at strong coupling whose dual five dimensional gravitational description , exhibits similar characteristics with those of shockwave collisions in ( ordinary ) four dimensional gravity .our goal in this work is to attempt to apply our earlier experience @xcite and especially apply the technique that we have developed in @xcite in ( ordinary gravity in ) four dimensions .we choose to work in a different coordinate system than @xcite in order to retain the geometrical insight of the collision and in addition we take into account the matter responsible for the creation of the shockwaves and the back - reaction effects as well .furthermore , the collision we consider here is not axisymmetric but is involves a non zero impact parameter .we organize the paper as follows . in chapter[ sup ] we state the problem we want to solve and construct the main set up .our goal is to determine the evolution of the geometry assuming that we know it in some time interval ( negative times ) .the initial geometry is given by two shockwaves which correspond to a non zero stress - energy tensor . ) with the speed of light and begin to interact for positive times ] our method is to construct a perturbative approach by expanding the metric around the background given by the flat metric along with the two shockwaves .equation ( [ s12 ] ) shows the form of the metric at all times while figure [ interaction ] offers a diagrammatical intuition of the terms of the metric we attempt to calculate in this project . in chapter [ b2b ]we take into account the interaction of the one particle with the gravitational field created from the other and vice versa . in terms of feynman diagrams , loosely speaking , these corrections correspond to the diagrams of figure [ selfint4d ] .the corrections to the stress - energy tensor corresponding to these diagrams along with the corrections of the metric tensor corresponding to diagram of figure [ interaction ] form a consistent set ) ) .] of the corrections that have to been taken into account .we verify that the modified stress - energy tensor is conserved ( to the order of the expansion that we are working ) and we find that it is traceless .this last condition results into some pleasing simplifications for einstein s equations ( compare ( [ ein ] ) with ( [ ee ] ) ) .chapter [ fe ] deals with the field equations and the specification of the gauge .our attempt is to perform a perturbative calculation about a metric that looks almost flat but contains two shockwaves moving opposite to each other and colliding ( see ( [ s12 ] ) ) .the shockwaves provide an effective stress - energy tensor in addition to the ( actual ) stress - energy tensor ( see ( [ rmn2gtt ] ) ) that creates the two shockwaves ; these terms correspond to products of the form @xmath5 and @xmath6 of equation ( [ deq1 ] ) respectively .a suitable gauge choice simplifies the field equations ( [ deq1 ] ) to equations ( [ deq ] ) which are solved in the next chapter . in chapter [ seq ]we specify the boundary conditions of the field equations ( [ deq ] ) and the corresponding green s function .in particular , we seek for causal solutions and therefore the associated green s function to the differential operator ( [ box ] ) is the retarded green s function .we show how the integrations on the light - cone and transverse plane may be performed omitting some of the intermediate steps for appendices [ a ] , [ b ] and [ c ] .eventually we derive a formula for the metric tensor , equation ( [ gmn2 ] ) , to the order we are working .this is our final result and it generally has the structure ( [ gmn ] ) .finally in chapter [ conc ] we discuss about the area of validity of our solution and summarize our conclusions . in particular , we argue that the presence of matter and the back - reaction effects may not be ignored as they result to an important contribution to the metric .we also see that as the impact parameter tends to zero , the metric diverges logarithmically .it is in our belief that this is a signal that a classical approach to the problem stops being valid and that a quantum description is required .lastly , we talk about the general form of the metric . although its evolution is constrained by casualty as expected ; this evolution takes place in an intuitive way : at a given ( proper ) time , any arbitrary point on the transverse plane evolves according from whether the signal from the center of the one or the other shockwave or both , has enough time to reach the point under consideration .we had encountered such a behavior in an analogous set up in @xcite although the geometry there was anti de sitter geometry in five dimensions . in @xcitewe had claimed that a similar evolution of the metric should also be observed in four dimensions ; in this work we verify our conjecture .we begin by defining the coordinate system ( gauge ) we work .we choose to work in light - cone coordinates defined by @xmath7 where @xmath8 is the time axis and @xmath9 cover @xmath10 .the convection for the flat metric that we use is @xmath11 we suppose that we have a black hole metric that we boost to the speed of light along a given direction ( @xmath12 direction ) . the metric is known @xcite and is given by @xmath13 where @xmath14according to equation ( [ ds1 ] ) we denote the transverse flat metric by @xmath15 while @xmath16 denotes the delta ( dirac ) function .the parameter @xmath17 has dimensions of length and its physical meaning will become apparent in what follows while @xmath18 serves as an ultraviolet cutoff and whose physical meaning is discussed in the conclusions ( see section ( [ scon ] ) ) .one may check directly whether ( [ ds1 ] ) solves einstein s equations which may be cast as @xmath19 where @xmath20 is the ricci tensor , @xmath6 the stress - energy tensor and @xmath21 the newton s constant ( in four dimensions ) . direct substitution of ( [ ds1 ] ) in the formula that computes @xmath22 results to @xmath23 where @xmath24 is a kronecker delta .this implies that all components of @xmath22 are zero except from @xmath25 . in order to arrive to equation ( [ rmn ] )we had to evaluate the following linear differential expression @xmath26 where @xmath27 is the laplace operator in two dimensions while we used the identity @xmath28 equations ( [ ein ] ) and ( [ rmn ] ) imply that the metric tensor of equation ( [ ds1 ] ) corresponds to a stress - energy tensor .shifting the origin along negative @xmath29 for distance @xmath30 we find that the stress - energy tensor is given by @xmath31 the presence of the superscript @xmath32 on @xmath33 of ( [ t1 ] ) is to highlight that it is of first order in the parameter @xmath17 .( i=1,2 assuming we have two sources ) appears .generally in the object @xmath34 there exists the product @xmath35 or any linear combination of differentiation / integration of @xmath36 and @xmath37 with respect to their arguments @xmath38 . ]this equation implies that the shockwave of ( [ ds1 ] ) is a consequence of a point particle moving along the @xmath39 direction with the speed of light and hence its massless .indeed , the ratio @xmath40 has dimensions of mass as should .one may check that @xmath6 of ( [ t1 ] ) is covariantly conserved .we may gain some insight in the physical system at hand if we use a diagrammatical approach . despite that general relativityis a non linear theory , the metric ( [ ds1 ] ) satisfies a linear differential equation .the fact that ( [ ds1 ] ) is an expression of a perturbation of a flat metric proportional to @xmath17 suggests the diagram of figure [ vertex ] . which is measured at point @xmath38 .this is a very special case where a single graviton exchange between the source and the bulk happens to be an exact solution to the non linear einstein s equations . ]it represents the measurement of the gravitational field at point @xmath41 , which , loosely speaking , is created by a single graviton emission from the source ( point - like stress energy tensor ) of equation ( [ t1 ] ) with effective coupling proportional to @xmath42 . in other words for this special case of a single shockwave, the first order solution happens to be the exact solution to all orders .having defined all the necessary ingredients we now proceed to the main part of the setup .we want to superimpose two such shockwaves whose sources are two point - like distributions of matter moving towards each other ) . ] in space - time .we want to collide these shockwaves ( and as a result the corresponding stress - energy tensors as well ) at a non - zero impact parameter and hence study the problem within the classical theory of gravity .therefore , @xmath6 has in addition to ( [ t1 ] ) the symmetric part @xmath43 which creates a second shockwave . in terms of space - time, this would correspond in colliding " two metrics in an off center process .figure [ offcenter ] represents the four dimensional picture , right before the collision of the two shockwaves .following @xcite,@xcite , the metric that describes the process should look like axis and dragging a perpendicular gravitational field which is constant along the circular lines .they collide at the origin producing a gravitational field in the forward light cone .our goal to compute the produced " metric and in particular @xmath44 . ]correction of the metric : it represents along with the diagram of figure [ selfint4d ] ( see chapter [ b2b ] ) , the first non trivial correction to ( [ s12 ] ) .it shows how the two metrics that each one looks like ( [ s1 ] ) merge .the gravitational field is measured at the point @xmath38 . ]@xmath45 the first three terms correspond to the flat ( minkowski ) space .the next two are of first order in @xmath17 and are created by the two point - like particles . these move ( initially ) towards each other along @xmath12 and they have an impact parameter @xmath46 along the @xmath29 axis as figure [ bten ] depicts .as they are the sources of the two shockwaves , they correspond to two vertex diagrams that look like the one in figure [ vertex ] .this is a superposition of two metrics with each one looking like ( [ s1 ] ) .however , the non - linearities of the gravitational field require higher order terms . the second order corrections are explicitly displayed in ( [ s12 ] )and they appear once the two shockwaves cross each other ; in the forward light cone .this is precisely the meaning of the @xmath47-functions ; they emphasize that the metric ( [ s12 ] ) solves einstein s equations exactly in the presence of both shockwaves only for negative @xmath48 .the additional terms of the metric appear in the forward light cone only and describe the effects of collision .the main work of this paper is to show how these terms may be calculated to order @xmath49 , that is find @xmath44 .the second order correction in @xmath17 of @xmath1 corresponds to the diagram of figure [ interaction ] .as has been already mentioned below ( [ t1 ] ) , @xmath50 is conserved in the gravitational field of ( [ ds1 ] ) , ( [ s1 ] ) .in fact , conservation for this case happens to be valid to all orders in @xmath17 appears in the resulting equations .this is in accordance with our intuitive picture of figure [ vertex ] : gravity behaves linearly with respect to the metric ( [ ds1 ] ) . ]@xmath51 changes with time .the first source interacts via the gravitational field created by the other and vice versa .the point @xmath41 is the space - time point where @xmath1 is measured . ]conservation to first order is still valid when we consider simultaneously @xmath50 and @xmath52 in the presence of the gravitational field ( [ s12 ] ) .however , this is no longer true at the second order in @xmath17 .the reason is because the @xmath53 ( @xmath33 ) source moves in the gravitational field of the @xmath36 @xmath54 shockwave , altering its initial trajectory .figure [ bten ] outlines what happens while figure [ selfint4d ] offers a diagramatical intuition regarding the self - corrections to @xmath6 .this implies that we should correct @xmath55 in order to preserve conservation ( of the total stress - energy tensor ) .however , since we do not know the nature ( equation of state ) of @xmath6 we make the assumption that these objects interact only via gravitational forces .since these particles are point - like and massless , they should travel along null geodesics ; as it is rigorously shown in @xcite conservation of ( the total ) @xmath56 is then guaranteed .this suggests that we need @xmath57 which gives the total stress -energy tensor of both point - like particles of mass @xmath58 each moving along the trajectory @xmath59 parameterized by @xmath60 .the quantity @xmath61 is the determinant of the ( total ) metric tensor , the factor @xmath62 is in agreement with our convention ( that reproduces ( [ t1 ] ) - see below ) while the dots denote differentiation with respect to the parameter @xmath63 . before calculating higher order corrections to @xmath6, we find it instructive to check whether this formula reproduces ( [ t1 ] ) in the case of one particle ( @xmath64 ) .the trajectory of this particle which moves ultra - relativistic ally along negative @xmath12 is parameterized by @xmath65 in light cone coordinates and choosing to parameterize the trajectory by @xmath39 , equation ( [ tr1c ] ) then implies @xmath66 direct substitution of ( [ tr1lc ] ) to ( [ tpp ] ) yields @xmath67 which is exactly equation ( [ t1 ] ) ( for @xmath68 ) .this computation also clarifies the convection of @xmath62 in ( [ tpp ] ) .the part of @xmath6 of equation ( [ t2 ] ) due to the second particle is reproduced similarly .the next step is to calculate the next ( second ) order corrections ( in @xmath17 ) of @xmath6 which translates into finding the corrections to the trajectories @xmath69 .these may be obtained from the geodesic equations . in particular , we only need the first order corrections to @xmath69 as we already have a power of @xmath17 in front of the summation operator ( see ( [ tpp ] ) ) .the geodesic equations we need read @xmath70 and are interpreted as the motion of particle @xmath71 in the gravitational field of the particle @xmath72 ( due to @xmath73 where @xmath74 are the christoffel symbols ) and vice versa ; this is precisely the meaning of the subscripts @xmath71 and @xmath72 .we begin with computing the corrections to the particle @xmath64 whose ( first order ) perturbed trajectory looks like which labels the particle @xmath64 for simplicity . ] @xmath75 where we have chosen to parameterize the trajectory with @xmath39 ( i.e. @xmath76 ) .the superscript @xmath77 denotes the order in the expansion . taking into account ( [ tpp ] ) and the fact that @xmath78 otherwise is zero, we deduce that both of the terms @xmath79 or @xmath80 have to be of zeroth order ; i.e. the only choice is @xmath81 .this implies that we need to determine @xmath82 to first order in @xmath17 that arise from the second particle which labels the second particle . ] and which read @xmath83 a few explanations about our notation are in order : the term @xmath37 is due to the second particle and is given in equation ( [ s12 ] ) .the subscript @xmath84 on @xmath37 denotes ordinary differentiation of the source @xmath37 with respect to the coordinate @xmath38 .according to ( [ s12 ] ) , the source @xmath37 is of first order in @xmath17 which denotes the order in @xmath17 from @xmath85 . ] and as a result the same applies for the @xmath74 s of ( [ gam ] ) ; it should be by now obvious that the christoffel symbols are due to the second particle ( @xmath37 ) .the final step is to integrate ( [ geo ] ) using ( [ gam ] ) and using causal boundary conditions and substitute the result in formula ( [ tpp ] ) .as we are interested in second order corrections in @xmath17 , we immediately conclude that at least one of @xmath86 or @xmath79 should be of order zero , i.e. @xmath87 or @xmath88 or @xmath89 , @xmath90 .we also note that @xmath91 and hence according to ( [ tpp ] ) corrections from @xmath92 do not contribute to @xmath6 at @xmath93 .we consider two cases in this case the modification of @xmath6 of equation ( [ tpp ] ) to this order we are working is in the arguments of the delta s .we have that follow from now on ( see also appendices [ a ] and [ b ] ) will imply the obvious : @xmath94 stands for @xmath95 , @xmath96 stands for @xmath97 etc . ]@xmath98 expanding the delta s to first order in the sources we obtain @xmath99.\end{aligned}\ ] ] we may cast ( the first order correction terms of the ) last equation in a compact form by expressing it in terms of @xmath36 and @xmath37 . using the identity@xmath100 ( see ( [ dlog ] ) and ( [ s12 ] ) ) the @xmath93 terms of ( [ ti2 ] ) take the form @xmath101 where we restored the subscript @xmath102 and the superscript @xmath103 in order to highlight that this is the second order correction to @xmath6 of the first particle . in this casethe modification of @xmath6 of equation ( [ tpp ] ) to this order we are working is in the factor @xmath104 . combining ( [ geo ] ) and ( [ gam ] ) one may compute @xmath105 . plugging this result to ( [ tpp ] ) and employing the identity ( [ nab1 ] ) in order to write the transverse delta s in terms of @xmath36 yields to [ tpii ] @xmath106 the second order corrections to the stress energy tensor @xmath107 of the first particle is given by equations ( [ ti3 ] ) and ( [ tpii ] ) .the corrections @xmath108 of the second particle may be found analogously and therefore the second order corrections to the total stress energy tensor read [ tmn2 ] @xmath109 the first equality in equation ( [ t+- ] ) is not completely obvious and so we prove it below by considering @xmath110 where in the fourth equality we used the fact that the stress - energy tensor of the point particles ( see ( [ t1 ] ) , ( [ t2 ] ) and ( [ nab1 ] ) ) to first order in @xmath17 may take the form @xmath111 despite working mostly with ( [ tmn2 ] ) which is a compact expression , for concreteness , we write @xmath6 of ( [ tmn2 ] ) in terms of the coordinates in order to clarify its form . defining @xmath112 and employing ( [ nab1 ] ) in ( [ tmn2 ] )we obtain [ tmnc ] @xmath113 \label{t++c}\\ & ( t_{+1})^{(2)}= \frac{\pi \mu^2}{4 \kappa_4 ^ 2 |b| } \theta(x^-)\delta(x^+)\delta^{(2)}(\vec{r_1 } ) \label{t+-1c}\\ & ( t_{+2})^{(2 ) } = 0 \label{t+2c } \ ] ] the asymmetry between @xmath114 and @xmath115 is due to the fact that the impact parameter @xmath116 has only @xmath29 component ( see ( [ s12 ] ) ) .the rest non zero components that complete ( [ tmnc ] ) may be obtained using the discrete symmetries of the problem : @xmath117 and @xmath118 may be obtained from @xmath119 and @xmath114 respectively by interchanging @xmath120 and @xmath121 .we want to justify our claim that the @xmath122 corrections to @xmath6 correspond to null geodesics . for thiswe consider the line element @xmath123 of the second shockwave @xmath37 ( the analogue of ( [ ds1 ] ) ) . for time - like distances and for fixed transverse positionwe have @xmath124 and integrating over the discontinuity due to the trajectory of the first particle ( @xmath36 ) along the second shockwave @xmath54 we deduce that @xmath125 the superscript @xmath77 on this equation highlights the fact that the discontinuity along the shockwave @xmath36 is of first order in @xmath17 and it implies that the trajectory of the second particle will be modified by along @xmath126 by @xmath127 . but according to the argument of the first @xmath4-function of ( [ ti1 ] ) , equation ( [ d+ ] ) is exactly equal to the shift along @xmath126 that we have already encountered from the geodesic analysis ( for the first particle ) .this completes our argument .the second order corrections to ( the total ) @xmath6 have already been calculated in the previous section .one , may check by a direct computation using ( [ nab^2 ] ) that this @xmath6 is covariantly conserved .explicitly this means that @xmath128 where @xmath129 denotes a covariant derivative while the superscripts denote the order in @xmath17 while we have used the identity ( [ nab^2 ] ) .therefore we conclude that @xmath6 is conserved if and only if the impact parameter is not zero .causes problem in the metric as well . as we will see , as @xmath130, the metric tensor diverges logarithmically ( see section [ scon ] ) .] this is one of our main conclusions in this paper .it is also useful to compute the trace of @xmath6 as it enters the field equations ( see ( [ ein ] ) ) .a short computation yields to @xmath131 which shows that the stress - energy tensor is traceless to order @xmath49 .tracelessness is very convenient as it simplifies einstein s equations which become @xmath132in this section we wish to write an explicit form of ( [ ee ] ) up to order @xmath93 . in order to determine these ( differential ) equations we take into account that the zeroth order terms satisfy ( [ ee ] ) trivially as @xmath133 while @xmath134 ( resulting from the first order terms of ( [ s12 ] ) ) is compensated by @xmath55 of equations ( [ t1 ] ) and ( [ t2 ] ) .thus , we only need @xmath135 where @xmath122 has already been calculated in the previous chapter and is given by ( [ tmn2 ] ) .it is crucial to state that @xmath136 receives two different type of contributions : ( a ) the contribution due to the ( pre)existing shockwaves ( that is due to the @xmath36 and @xmath37 terms of ( [ s12 ] ) ) ; we denote this contribution by @xmath137 .( b ) the contribution due to the ( second order ) corrections ( in @xmath17 ) of the metric ( that is due to @xmath138 ; we denote this contribution by @xmath139 . recalling equation ( [ s12 ] ) that gives the form of the metric at all times and expanding ( [ rmn2 ] ) to @xmath93, we expect that it should have the form @xmath140 where @xmath137 ( and @xmath122 ) is known while @xmath139 is what we will use in order to determine @xmath138 ( see ( [ deq1 ] ) ) .equation ( [ rmn2gt ] ) may also be cast in the form @xmath141 and view @xmath137 as an effective ( contribution to the tottal ) stress - energy tensor ( see ( [ deq1 ] ) and ( [ deq ] ) ) .dropping the superscripts @xmath103 from @xmath44 and @xmath122 and the superscripts @xmath77 from @xmath85 for simplicity , we find that the components of ( [ ee ] ) to second order in @xmath17 read [ deq1 ] @xmath142=\kappa_4 ^ 2t_{++},\label{++1}\\ ( + -)\hspace{0.15 in } \frac{1}{4}\big [ & -2g_{+-,x^2x^2}-2g_{+-,x^1x^1}+2g_{+2,x^-x^2}+2g_{+1,x^-x^1}-2g_{++,x^-x^- } \notag\\ & + 2g_{-2,x^+x^2}+2g_{-1,x^+x^1}-2g_{11,x^+x^-}-2g_{22,x^+x^-}+4g_{+-,x^+x^-}\notag\\ & -2g_{--,x^+x^+}-2t_{1,x^1}t_{2,x^1}-2t_{1,x^2}t_{2,x^2}+t_{1,x^+}t_{2,x^-}\big]=\kappa_4 ^ 2t_{+-},\label{+-1}\\ ( + 1)\hspace{0.15 in } \frac{1}{4}\big [ & -2g_{+1,x^2x^2}+2g_{+2,x^+x^-}-2g_{++,x^-x^1}+2g_{12,x^+x^2}-2g_{22,x^+x^1 } \notag\\ & + 2g_{+-,x^+x^1}+2g_{+1,x^+x^-}-2g_{-1,x^+x^+}+t_{1,x^+}t_{2,x^1}\big]=\kappa_4 ^ 2t_{+1},\label{+11}\\ (11)\hspace{0.15 in } \frac{1}{2}\big [ & -g_{11,x^2x^2}+2g_{12,x^1x^2}-g_{22,x^1x^1}+2g_{+-,x^1x^1}-2g_{+1,x^-x^1}-2g_{-1,x^+x^1}\notag\\ & + 2g_{11,x^+x^-}+t_{1,x^1}t_{2,x^1}+t_{1,x^1x^1}t_{2}+t_{1}t_{2,x^1x^1}\big]=k_{4}^{2 } t_{11}\label{111}=0,\\ ( 22)\hspace{0.15 in } \frac{1}{2}\big [ & -g_{11,x^2x^2}+2g_{12,x^1x^2}-g_{22,x^1x^1}+2g_{22,x^+x^-}-2g_{+2,x^-x^2}-2g_{-2,x^+x^1}\notag\\ & + 2g_{+-,x^2x^2}+t_{1,x^2}t_{2,x^2}+t_{1,x^2x^2}t_{2}+t_{1}t_{2,x^2x^2}\big ] = k_{4}^{2 } t_{22}\label{221}=0,\\ ( 12)\hspace{0.15 in } \frac{1}{4}\big [ & 4g_{+-,x^1x^2}-2g_{+1,x^-x^2}-2g_{+2,x^-x^1 } -2g_{-1,x^+x^2}-2g_{-2,x^+x^1}+4g_{12,x^+x^-}\notag\\ & t_{2,x^2 } t_{1,x^1}+ t_{1,x^2}t_{2,x^1}+2t_{2}t_{1,x^1x^2 } + 2t_{1}t_{2,x^1x^2 } \big]=\kappa_4 ^ 2t_{12}=0\label{121 } \ ] ] where @xmath143 are given by ( [ s12 ] ) and correspond to the geometry for negative times while the components of @xmath6 are given by ( [ tmn2 ] ) . indeed ( [ deq1 ] ) has the expected form of equation ( [ rmn2gt ] ) .the above set of the field equations has been written without specifying the gauge . in the next section, we will see how these may be simplified by making a convenient gauge choice .we follow the standard procedure in order to solve ( [ deq1 ] ) : we define a new coordinate system @xmath144 with respect to the old one @xmath145 ( see ( [ lc ] ) ) by @xmath146 where @xmath147 is an arbitrary function of the ( old coordinates ) @xmath148 and is second order in @xmath17 .obviously , this transformation induces a second order change in @xmath17 to @xmath44 but does not alter @xmath149 and @xmath150 .more precisely the second order terms of the metric transform to @xmath151 where the semicolon denotes a covariant derivative is exactly equal to the action of the lie derivative acting on @xmath44 along the vector field @xmath152 , i.e. @xmath153@xmath154 . ] .what is remarkable is that the field equations remain invariant under this transformation as it may be shown on general grounds @xcite . for concreteness ,we exhibit it here for the @xmath155 component , equation ( [ 121 ] ) .taking into account that at the order we are working , the covariant derivative may be replaced by ordinary differentiation , plugging the tensor @xmath156 into the differential part of ( [ 121 ] ) and dropping ( again ) the superscript @xmath103 from the @xmath157 s for simplicity , we obtain @xmath158 \notag\\ & = \frac{1}{4}\big[\left\ { \left(4\xi_{+,x^-,x^1,x^2}-2\xi_{+,x^1,x^-,x^2}-2\xi_{+,x^2,x^-,x^1}\right ) + ( + \leftrightarrow- ) \right \}\notag\\ & \hspace{0.25in}+\left\ { \left(4\xi_{1,x^2,x^+,x^-}-2\xi_{1,x^+,x^-,x^2}-2\xi_{1,x^-,x^+,x^2}\right)+(1\leftrightarrow2)\right\ } \big ] = 0\end{aligned}\ ] ] where we used the fact that partial derivatives commute .so far , the vector field @xmath157 has been arbitrary while the result of the transformation ( [ ncs ] ) on ( [ deq1 ] ) is just the relabeling @xmath159 .a convenient choice of @xmath160 is the one that satisfies de donder gauge @xmath161 where @xmath162 is the flat metric . applying this gauge to the field equations that we are interested , equations ( [ deq1 ] ) , and dropping the tilde symbol from @xmath163 for simplicity, the field equations simplify to [ deq ] @xmath164 where we have used ( [ tmn2 ] ) while @xmath165 is the scalar operator in flat space ; that is @xmath166 in the next chapter we will see how equations ( [ deq ] ) may be solved imposing appropriate boundary conditions .in this section we will show how to solve ( [ deq ] ) by seeking for causal solutions .the casual boundary conditions imply that the flat metric in the presence of the shockwaves , as can be checked , is an exact solution to einsteins equations with a right hand side given by ( [ t1 ] ) and ( [ t2 ] ) , for negative times only . for positive times , that is for @xmath167 and @xmath168 , the second order corrections ( in @xmath17 ) of the metric are switched on as the point @xmath169 is the collision point .simultaneously , the initial stress - energy tensor of the ( massless ) particles that induces the shock - waves suffers a change ( see figure [ selfint4d ] and ( [ tmn2 ] ) ) that also has to be taken into account . the retarded green s function we need corresponding to the differential operator ( [ box ] ) is known and in light - cone coordinates is given by @xmath170 where according to ( [ s1 ] ) , @xmath171 , @xmath47 denotes a theta ( step ) function while equation ( [ gf ] ) ( the retarded green s function ) satisfies @xmath172 the procedure we have to follow is standard : we convolute the right hand sides of ( [ deq ] ) with ( [ gf ] ) and integrate in all over space - time .we find it convenient to introduce the following notation @xmath173 where @xmath174 is any arbitrary function of @xmath148 while the last integral denotes integration in the transverse plane .we wish to perform the @xmath48 integrations for all the possible cases that we will encounter while specifying @xmath44 from ( [ deq ] ) .we organize these integrations in five cases while we leave the details of the calculation for appendix [ a ] .in addition to the five cases of the @xmath48 integrations which we solve in appendix [ a ] there exists another case that involves the evaluation of @xmath175 that arises from ( [ + + ] ) . due to the complication of the calculationwe evaluate this term in a separate appendix ( see appendix [ b ] , equation ( [ a2 ] ) ) .the results of all of the integrations in the light - cone plane ( performed in appendices [ a ] and [ b ] ) are proportional to the product @xmath176 .this implies that the second order corrections to @xmath1 appear in the forward light - cone ( see figure [ etatau ] ) which is what we have initially demanded by seeking for a causal solution .the right hand sides of ( [ deq ] ) contains expressions of the form @xmath177 differentiated with respect to @xmath38 in some fashion . according to ( [ s12 ] ) ,these expressions are proportional to @xmath178 or their derivatives .our previous analysis has already taken care of the @xmath48 integrations and so from now on , by @xmath143 we will mean just the transverse part of @xmath143 : @xmath179 .having performed the @xmath48 integrations we move to the integration over the transverse plane .the quantities we have to integrate have the structure .however , according to ( [ nab1 ] ) these ( transverse ) integrations are trivial as they involve delta functions . ]@xmath180 or @xmath181 where @xmath182 .the subscript @xmath183 ( c ) and the superscript @xmath184 @xmath185 denotes differentiation of the source @xmath186 @xmath187 with respect to the space - time coordinate @xmath188 @xmath189 .we may reduce the number of the different integrals that we have to perform by working as follows .we firstly introduce the vectors @xmath190 and generalize the form of ( the transverse part of ) @xmath143 given by ( [ s12 ] ) to @xmath191 where @xmath192 where defined by ( [ r12 ] ) .the next step is to exchange the derivatives acting on @xmath143 , that is @xmath193 with differentiations with respect to @xmath30 s of ( [ b ] ) , that is with @xmath194 takes the form @xmath195 . ] .finally , at the end of our calculations we take the limits @xmath196 looking equations ( [ + -])-([12 ] ) we see that they involve the product @xmath177 differentiated with respect to the transverse coordinates . ; see ( [ s12 ] ) while the @xmath48 contributions have already been taken into account in the previous section . ] exchanging the transverse differentiations , according to our earlier discussion in this section , with derivatives with respect to the components of @xmath197 and taking into account the transverse part of the green s function , ( [ gf ] ) , we see at once that we have to calculate the following integral @xmath198 where we have introduced the convenient factor @xmath199 .now , the non trivial integration is the angular integration as the radial one becomes trivial due to the @xmath4-function .both of the integrations are performed in appendix [ c ] and the final result reads @xmath200 where the @xmath201 s may be found with the help of table [ ta1 ] and equation ( [ ja ] ) .equation ( [ j ] ) is the last ingredient that allows us to obtain the desired solutions for equations ( [ deq ] ) .we display the results in the next section .having performed all of the integrations arising from the convolution of the right hand sides of ( [ deq ] ) with the green s function ( [ gf ] ) , we are in a position to derive the final formulas for @xmath44 . on the corrections of @xmath1 in order to highlight the order in @xmath17 that we are working . ] the ingredients that we need , have been obtained or defined in the previous sections and in appendices [ a ] , [ b ] and [ c ] . to begin with , we need the defining equations for @xmath143 and @xmath122 given by ( [ s12 ] ) and ( [ tmn2 ] ) respectively but with the generalized @xmath197 ( see ( [ b ] ) ) instead ( of ( [ lim ] ) ) while identity ( [ nab1 ] ) is very useful .we also need the value of the integral @xmath202 defined in ( [ jin ] ) and given by ( [ j ] ) , ( [ ja ] ) and table [ ta1 ] as well as ( [ r12 ] ) and ( [ tao ] ) that define @xmath203 and @xmath3 , @xmath204 respectively . the final formula for @xmath44 is eventually given below . [ gmn2 ]@xmath205 \bigg\}\bigg\ } , \label{g++2 } \\g_{+-}^{(2 ) } & \lim_{\vec{b}_{1,2}\to ( \pm b,0)}\bigg \ { \frac{1}{2 } \mu^2 \theta(x^+)\theta(x^- ) \text{sech } \hspace{0.02 in } \eta \bigg\{\frac{1}{2 \tau } \log \left(k|\vec{b}_2-\vec{b}_1|\right)\delta(\tau - r_1 ) \notag\\ & + \left [ \partial^2_{b_{11}b_{21}}-\frac{1}{4 } \left(\frac{1}{\tau^2}\text{sech$^2$}\hspace{0.02in}\eta+\frac{1}{2}\tau \partial_{\tau } \left(\frac{1}{\tau}\partial_{\tau } \right ) \right ) \right]{\cal j } ( r_1,r_2,\tau ) + \big ( 1\leftrightarrow2 \big ) \bigg\ } \bigg\ } , \label{+-2 } \\ g_{+1}^{(2 ) } & = \lim_{\vec{b}_{1,2}\to ( \pm b,0)}\bigg \ { \frac{1}{\sqrt{2 } } \mu^2 \theta(x^+)\theta(x^- ) \bigg \ { \frac{b_{11}-b_{21}}{|\vec{b}_2-\vec{b}_1|^2}\frac{r_1}{r_1 ^ 2 + 2 ( x^{\pm})^2 } \theta(\tau - r_1 ) \notag\\ & \hspace{1.20in}+\frac{1}{2 } ( \partial_{b_{21 } } ) \left [ \frac{1}{1+e^ { \pm 2\eta}}\partial_{\tau } -\frac{1}{2 \tau } \text{sech}^2\hspace { 0.02in}\eta \right ] { \cal j } ( r_1,r_2,\tau ) \bigg\ } \bigg\ } , \label{g+12 } \\g_{11}^{(2 ) } & = \lim_{\vec{b}_{1,2}\to ( \pm b,0)}\bigg \ { -\frac{1}{2 } \mu^2 \theta(x^+)\theta(x^- ) \text{sech } \hspace { 0.02in}\eta \notag\\ & \hspace{1.8in}\times \big\ { \partial_{b_{11}b_{21}}^2 + \partial_{b_{11}b_{11}}^2 + \partial_{b_{21}b_{21}}^2 \big\ } { \cal j } ( r_1,r_2,\tau ) \bigg\ } , \label{g112 } \\g_{12}^{(2 ) } & = \lim_{\vec{b}_{1,2}\to ( \pm b,0)}\bigg \ { -\frac{1}{4 } \mu^2 \theta(x^+)\theta(x^- ) \text{sech}\hspace { 0.02in}\eta\notag\\ & \hspace{1.2 in } \big\ { \partial_{b_{22}b_{11}}^2 + \partial_{b_{12}b_{21}}^2 + 2\partial_{b_{11}b_{12}}^2 + 2\partial_{b_{21}b_{22}}^2 \big\ } { \cal j } ( r_1,r_2,\tau ) \bigg\ } .\label{g122 } \end{aligned}\ ] ] in order to arrive to ( [ gmn2 ] ) we have convoluted ( [ gf ] ) with the right hand side of ( [ deq ] ) and employed ( [ c1 ] ) , ( [ c2 ] ) , ( [ c3 ] ) , ( [ c4 ] ) , ( [ c5 ] ) and ( [ a2 ] ) . in particular we have applied ( [ c5 ] ) and ( [ a2 ] ) for ( [ + + ] ) , ( [ c1 ] ) and ( [ c2 ] ) for ( [ + - ] ) , ( [ c3 ] ) and ( [ c4 ] ) for ( [ + 1 ] ) and ( [ c1 ] ) for both ( [ 11 ] ) and ( [ 12 ] ) .the reason in preferring to work with the generalized @xmath197 is because the @xmath206 may be obtain from @xmath207 under @xmath208 before taking the limits as in ( [ lim ] ) ; thus reducing the amount of calculations .finally , @xmath209 are obtained from @xmath210 under the ( simultaneous ) interchanges @xmath211 and @xmath212 . these stepscomplete the determination of @xmath44 .formula ( [ gmn2 ] ) is the final result of this project and we will analyze it in the next chapter .we have seen below ( [ s1 ] ) that the parameter @xmath17 we use in our expansion has dimensions of length . on the other hand we know that the components of the metric should be dimensionless .hence , each power in @xmath17 is compensated by an inverse power of the coordinates times a logarithm ( with argument @xmath213 , @xmath214 or @xmath215 ) in some power at most . for simplicity ( although not necessary ) , we restrict our discussion at middle -rapidity where @xmath216 that is for @xmath217 ( see ( [ tao ] ) and figure [ etatau ] ) .this means that any @xmath218-order ( in @xmath17 ) contribution to the metric , where @xmath18 is a positive integer , will generally have the form terms ( see ( [ gmn ] ) ) . ]@xmath219 ( see ( [ gmn2]))where @xmath220 are dimensionless real functions of @xmath38.-functions , logarithms and real coefficients ] we thus believe that our expansion is valid at high energies , that is small @xmath17 , compared to @xmath203 and @xmath3 . in fact , only one of @xmath203 or @xmath3 has to be large compared to @xmath17 .so for instance ( [ gmn2 ] ) is a good approximation for small @xmath3 but large @xmath203 ( see figure [ re ] , region @xmath71 ) and also for small @xmath203 but large @xmath3 ( region @xmath221 ) provided that the massless particles creating the shockwaves are not energetically enough ( @xmath17 is small ) .this is in contrast to @xcite where @xmath17 has dimensions of length to the negative third power and hence the expansion there was valid for early proper times . and@xmath221@xmath222 correspond to @xmath223 and @xmath224 respectively .the dark dots are the centers of the shockwaves and are located at an impact parameter 2b apart while @xmath225 denote the distance of the arbitrary point @xmath2 from the center of each shockwave ( right and left respectively ) . according to causality , at any given proper time @xmath3 , the propagation from the centers will reach the points on the peripheries ( at most ) .this suggests that any given point @xmath2 on the transverse plane of the produced " metric at given @xmath3 will evolve according to the region where it belongs to ( see equations ( [ gmn2 ] ) , ( [ ja ] ) and ( [ jap ] ) ) and there are three different possibilities ( and three from their mirror images ) . ] in this work we have found the first non - trivial causal corrections to the problem of shockwaves collisions in gravity created by boosting two black holes to the speed of light .the collision is assumed asymmetric and occurs at low energies ( see previous section ) . in terms of feynman diagrams ,our result , formula ( [ gmn2 ] ) , corresponds to the resummation of the diagrams of figures [ interaction ] and [ selfint4d ] .our concussions are summarized as follows .the corrections to @xmath1 evolve non trivially and constrained by causality in an intuitive way . in particular , the behavior of @xmath1 at any point on the transverse plane , is determined from whether the propagation from the center of each individual nucleus has enough proper time to reach the point under consideration or not .figure [ re ] is a snapshot taken at given proper time @xmath3 and depicts the six kinematical regions where @xmath1 evolves differently while it has the general form @xmath226 the indices @xmath71 , @xmath227 , @xmath221 on @xmath228 correspond to the regions @xmath71 , @xmath227 , @xmath221 of figure [ re ] respectively .terms cover region @xmath71@xmath222,@xmath227@xmath222 and @xmath221@xmath222 since under this interchange we have @xmath229 . ]the @xmath4-function terms arise from differentiating the @xmath47-functions of the right hand sides of ( [ gmn2 ] ) ( see also ( [ j ] ) ) .they represent two shockwaves centered at the center of each shockwave and expanding on the transverse plane with speed @xmath3 ( that is with the speed of light ) .this particular behavior of the metric was our initial motivation for dealing with this problem and our calculations confirm our earlier conjecture @xcite .the presence of matter , @xmath6 , and the back - reactions affect the metric ( see for example ( [ g++2 ] ) and ( [ g+12 ] ) ) not only on the forward light - cone but also inside .this implies that we can not in principle solve einstein s equations in vacuum ( ignoring the point - like particles that create the shocks ) arguing that we are away from the sources unless we know the boundary conditions that these sources enforce on the metric inside the light - cone .this is in analogy to classical electrodynamics : solving laplace equation for the scalar potential away from a point charge sitting at the origin without specifying the boundary conditions , one may obtain the trivial ( zero ) solution which obviously is not the correct one .the presence of the impact parameter @xmath30 is a necessary requirement and not an additional complication introduced in the problem .mathematically this is obvious from the fact that both @xmath6 and @xmath1 diverge when the impact parameter @xmath30 tends zero . from equation [ con ]we have seen that conservation of @xmath6 is violated violently and behaves as @xmath230 .the metric tensor also exhibits a problematic behavior in the zero impact parameter limit : the the formula for @xmath44 , equation ( [ fgmn ] ) , diverges logarithmically when @xmath231 as it is evident ( for instance ) from equation ( [ g++2 ] ) .a similar ultraviolet ( uv ) divergence appears in perturbation theory @xcite of gauge theories .this suggests that a head on collision may not be investigated using classical gravity .instead , one has to apply a quantum theory of gravity in the same way one can not predict the electron - positron annihilation ( in head on collisions ) using maxwell s equations .one has to turn into quantum electrodynamics in order to describe the process and predict the production of two photons .one may argue that a black hole may be formed and hence hide the violation of conservation behind the event horizon .\4 . for future projectswe propose that one could plot @xmath44 for several ( fixed ) impact parameters as a function of @xmath3 , @xmath29 and @xmath232 ( at central rapidities where @xmath233 ) in order to visualize the evolution of the metric .a very important aspect we ignored in our analysis is the role of the ultraviolet cutoff @xmath18 ( see ( [ s1 ] ) ) which ( for the case of a single shockwave ) seems to define an ergoregion ( with radius @xmath234 ) on the transverse plane .it would be interesting to check how our solution gets modified for impact parameters @xmath235 which would imply that the two ergoregions of the shockwaves overlap .finally , one could compute the gravitational radiation , take the limit of @xmath231 and compare the result with the one obtained by @xcite .in this appendix we perform the @xmath48 part of the integrations resulting when the green s function ( [ gf ] ) acts on the right hand side of ( [ deq ] ) . proceeding as in ( [ sh ] )we find out that we have to deal with five different cases ( there also exists a sixth case that is calculated in appendix [ b ] ) .this case is trivial and almost all terms of ( [ deq ] ) behave in this way . defining @xmath236 where @xmath3 is the proper time and @xmath204 the rapidity ( see figure [ etatau ] for a geometrical meaning ) we have that @xmath237 where we have shifted the integration variable setting @xmath238 . and @xmath204 .the hyperbolas indicate curves of constant @xmath3 and increase along @xmath8 .the straight lines are lines of constant @xmath204 and increase from left to right as the arrows indicate .along @xmath48 we have @xmath239 while along @xmath8 we have @xmath240 and @xmath233 . ] this case is more complicated as we have to integrate by parts the @xmath4-functions .three kind of terms will appear : ( a ) terms that differentiate the @xmath241 terms of ( [ gf ] ) and hence produce @xmath242 ( terms ) .but the presence @xmath242 forces the @xmath4-function term appearing in ( [ gf ] ) to become @xmath243 which is zero .hence these terms do not contribute .( b ) we have terms that differentiate the denominator and these contribute to the integrations .( c ) finally , we have terms that either differentiate the @xmath4-function of ( [ gf ] ) only or both , the @xmath4-function and the denominator .in order to evaluate these terms we exchange the @xmath244 that act on @xmath245 with @xmath246 . shifting the transverse variable as in the previous case and performing the @xmath247 integrationswe find that the contribution of both the ( b ) and ( c ) terms is@xmath248 \int d^2\vec{r ' } \delta \left(\tau - r ' \right)f_{ii}(\vec{r}+\vec{r'})\end{aligned}\ ] ] where the differential operator @xmath249 acts on the integral while @xmath250 denotes a partial differentiation with respect to @xmath3 .this is a simpler version of case @xmath227 and working in a similar fashion yields @xmath251 \int d^2\vec{r ' } \delta \left(\tau - r ' \right)f_{iii}(\vec{r}+\vec{r ' } ) . \ ] ] @xmath252 this is a combination of cases @xmath221 and @xmath253 with @xmath254 with @xmath255 as in ( [ r12 ] ) .we have @xmath256 where in the first equality we ignored a term similar to case @xmath227 ( see term ( a ) ) when integrating by parts the @xmath257 while in the second equality we performed the @xmath258 integration and exchanged @xmath244 with @xmath246 .the rest two steps are obvious .we wish to evaluate the expression ( [ c6 ] ) by performing the integration on both the light - cone and the transverse plane .we begin by performing the @xmath126 and @xmath39 integrations . using that @xmath259 we find @xmath260 where after the first equality we assume that the @xmath261 while the @xmath48 dependence is displayed explicitly and it is integrated out after the second equality .the next step is to perform the transverse integrations .the trick here is to integrate by parts the @xmath262 terms .the by parts integration produces two kind of terms : ( a ) those that do not act on @xmath263 and ( b ) those that act on @xmath263 .but the terms of case ( b ) are proportional to @xmath264 which is zero for non zero impact parameter @xmath265 while for zero impact parameter it diverges violently ; we conclude that an impact parameter is necessary ( see section [ scon ] ) .we now proceed to the remaining terms . exchanging @xmath266 with @xmath267 and using ( [ dlog ] ) equation ( [ a1 ] )gives @xmath268 where @xmath197 are given by ( [ b ] ) .we wish to calculate the integral ( [ jin ] ) that we encountered in section [ tp ] .we have @xmath269 the quantities @xmath203 are given by ( [ r12 ] ) .the trick here is to expand the logarithms in their fourier space : @xmath270 with @xmath18 serving as an ultraviolet cutoff . expanding both logarithms and performing the angular integrationone obtains @xmath271 performing the trivial radial integration we find that@xmath201 is now defined by @xmath272 next we perform the @xmath273 and @xmath274 integrations .these integrals have been calculated in @xcite ; we summarize the procedure : in order to perform these integrations one has to expand @xmath275 in an infinite sum of products of the form @xmath276 with @xmath277 an integer and do the angular integrals ( of q and l ) first .this factors out the radial integrations over @xmath273 and @xmath274 into two independent integrals .then one has to perform these integrations and finally sum over @xmath277 .the final result reads [ ja ] @xmath278 , \label{jj}\\ \xi_{>(<)}&=max(min)(r_1,\tau ) \hspace{0.2 in } \eta_{>(<)}=max(min)(r_2,\tau ) , \label{ke}\\ & \hspace{1.2 in } ( \vec{r_1}).(\vec{r_2})=\cos(\alpha ) r_1 r_2 .\label{a } \end{aligned}\ ] ] so here @xmath279 is the angle between @xmath280 and @xmath281 and @xmath282 is the dilogarithm function while @xmath201 is real as it should .equation ( [ ja ] ) implies that @xmath201 depends from the ordering of @xmath283 , @xmath284 and @xmath3 .there are in principle six distinct ways to order them .however it turns out that the cases @xmath285 and @xmath286 are interdependent from the relative ordering of @xmath283 and @xmath284 .this degeneracy reduces the possible cases to four which we organize by introducing table [ ta1 ] for instance means @xmath287 with @xmath288 given from ( [ ja ] ) . ] that in turn help us to write a ( unified ) formula for @xmath201 @xmath289 p. d. death and p. n. payne , `` gravitational radiation in high speed black hole collisions .perturbation treatment of the axisymmetric speed of light collision , '' _ phys ._ * d46 * ( 1992 ) 658674 .d. death and p. n. payne , `` gravitational radiation in high speed black hole collisions .reduction to two independent variables and calculation of the second order news function , '' _ phys ._ * d46 * ( 1992 ) 675693 .k. sfetsos , `` on gravitational shock waves in curved space - times , '' nucl .b * 436 * , 721 ( 1995 ) [ arxiv : hep - th/9408169 ] .d. m. eardley and s. b. giddings , `` classical black hole production in high - energy collisions , '' phys .d * 66 * , 044011 ( 2002 ) [ arxiv : gr - qc/0201034 ] .s. b. giddings and r. a. porto , `` the gravitational s - matrix , '' phys .d * 81 * , 025002 ( 2010 ) [ arxiv:0908.0004 [ hep - th ] ] .s. b. giddings , m. schmidt - sommerfeld and j. r. andersen , `` high energy scattering in gravity and supergravity , '' arxiv:1005.5408 [ hep - th ]. a. taliotis , `` heavy ion collisions with transverse dynamics from evolving ads geometries , '' arxiv:1004.3500 [ hep - th ]. j. l. albacete , y. v. kovchegov and a. taliotis , `` asymmetric collision of two shock waves in ads@xmath291 , '' jhep * 0905 * , 060 ( 2009 ) [ arxiv:0902.3046 [ hep - th ] ] .j. bartels _ et al ._ , `` proceedings of the 38th international symposium on multiparticle dynamics ( ismd08 ) , '' arxiv:0902.0377 [ hep - ph ]. j. l. albacete , y. v. kovchegov and a. taliotis , `` modeling heavy ion collisions in ads / cft , '' jhep * 0807 * , 100 ( 2008 ) [ arxiv:0805.2927 [ hep - th ] ] .d. grumiller and p. romatschke , `` on the collision of two shock waves in ads5 , '' jhep * 0808 * , 027 ( 2008 ) [ arxiv:0803.3226 [ hep - th ] ] .s. s. gubser , s. s. pufu and a. yarom , `` off - center collisions in ads@xmath291 with applications to multiplicity estimates in heavy - ion collisions , '' jhep * 0911 * , 050 ( 2009 ) [ arxiv:0902.4062 [ hep - th ] ] .s. s. gubser , s. s. pufu and a. yarom , `` entropy production in collisions of gravitational shock waves and of heavy ions , '' phys .d * 78 * , 066014 ( 2008 ) [ arxiv:0805.1551 [ hep - th ] ] .y. v. kovchegov and s. lin , `` toward thermalization in heavy ion collisions at strong coupling , '' jhep * 1003 * , 057 ( 2010 ) [ arxiv:0911.4707 [ hep - th ] ] .a. duenas - vidal and m. a. vazquez - mozo , `` colliding ads gravitational shock waves in various dimensions and holography , '' arxiv:1004.2609 [ hep - th ] .s. s. gubser , i. r. klebanov and a. m. polyakov , `` gauge theory correlators from non - critical string theory , '' phys .b * 428 * , 105 ( 1998 ) [ arxiv : hep - th/9802109 ] .s. s. gubser , i. r. klebanov and a. a. tseytlin , `` string theory and classical absorption by three - branes , '' nucl . phys .b * 499 * , 217 ( 1997 ) [ arxiv : hep - th/9703040 ] .s. de haro , s. n. solodukhin and k. skenderis , `` holographic reconstruction of spacetime and renormalization in the ads / cft correspondence , '' commun .* 217 * , 595 ( 2001 ) [ arxiv : hep - th/0002230 ] .a. papapetrou , lectures on general relativity , " ( d. reidel publishing company , holland 1974 ) .

<S> the problem of collisions of shockwaves in gravity is well known and has been studied extensively in the literature . recently , the interest in this area has been revived trough the anti - de - sitter space / conformal field theory correspondence ( ads / cft ) with the difference that in this case the background geometry is anti de sitter in five dimensions . in a recent project that we have completed in the context of ads / cft , </S><S> we have gained insight in the problem of shockwaves and our goal in this work is to apply the technique we have developed in order to take some farther steps in the direction of shockwaves collisions in ordinary gravity . in the current project , each of the shockwaves correspond to a point - like stress - energy tensor that moves with the speed of light while the collision is asymmetric and involves an impact parameter ( b ) . </S><S> our method is to expand the metric @xmath0 in the background of flat space - time in the presence of the two shockwaves and compute corrections that satisfy causal boundary conditions taking into account back - reactions of the stress - energy tensor of the two point - like particles . </S><S> therefore , using einstein s equations we predict the future of space - time using the fact that we know the past geometry . our solution respects causality as expected but this casual dependence takes place in an intuitive way . in particular , @xmath1 at any given point @xmath2 on the transverse plane at fixed @xmath3 evolves according from whether the propagation from the center of each of the shockwaves or from both shockwaves has enough proper time ( @xmath3 ) to reach the point under consideration or not . simultaneously around the center of each shockwave , </S><S> the future metric develops a @xmath4-function profile with radius @xmath3 ; therefore this profile expands outwards from the centers ( of the shockwaves ) with the speed of light . </S><S> finally , we discuss the case of the zero impact parameter collision which results to the violation of conservation and we argue that this might be a signal for the formation of a black hole . </S><S> the author would like to thank prof . </S><S> andrzej derdzinski and especially prof . </S><S> gerlach ulrich for serving in his defense committee and also for stimulating discussions prior and during this thesis . </S><S> also the author thanks prof . </S><S> samir mathur for very informative discussions during the writing of this work . </S><S> in addition he would like to thank prof . </S><S> thomas kerler and herb clemens for making the transferring from the graduate program of the department of physics to the graduate program of the department of mathematics possible and in addition denise witcher for guiding him through all the steps of this process . </S><S> completion of the graduate courses would nt be possible without the help and teaching enthusiasm of prof . </S><S> jean - francois lafont , daniel shapiro and joseph ferrar and particularly prof . </S><S> james cogdell and alexander leibman for answering his endless emails with clarity , exactness and always availability . </S><S> his student mates corry christopherson , fatih olmez and zhi qui have played a crucial role during the coursework as not only they kept encouraging him but also have spent enormous time from their time in order to patiently answer all of his questions . </S><S> attending many of the required classes would had been impossible without the help of his best friend chen zang who baby sat the authors son , nikolas alexandrou taliotis , while the author and his wife maria alexandrou had to attend classes . </S><S> the author owns special gratefulness to his physics advisor , prof . </S><S> yuri kovchegov , for teaching him physics , for teaching him how to see through the complicate mathematics and nail down the physical picture and especially for showing him how to undertake his responsibilities not only inside but also outside of the academia . </S><S> special thanks to prof . </S><S> tom banks , steve giddings , yuri kovchegov and krishna rajagopal for their encouragement in submitting this thesis on the arxiv after it was defended . </S><S> lastly but most important , the author would like to thank maria alexandrou for all of her patience during these years and for lifting most of the weights of their journey in life while keep smiling and being an example of a kind human , a reliable partner and a wonderful woman : he delicates this work to her . </S><S> this work is sponsored in part by the u.s . </S><S> department of energy under grant no . </S><S> de - fg02 - 05er41377 and in part by the institution of governmental scholarships of cyprus ( iky ) . </S><S> born in nicosia , cyprus b.sc . in physics , ms in physics graduate research associate in the department of physics and graduate student in the department of mathematics of the ohio state university </S>

1607.06796

reaction - diffusion equations are widely used to describe a variety of phenomena such as pattern formation and front propagation in biological , chemical and physical systems .when the model under study is characterized by the presence of competing equilibrium states , a crucial question is to describe the interaction and the dynamics of space occupation by the equilibria .a basic prototype is the _ allen cahn equation _ , which have the form @xmath2 andit has been originally proposed in @xcite to describe the motion of antiphase boundaries in iron alloys .such equation has an associated energy functional @xmath3 where the _ potential _@xmath4 is a primitive of @xmath5 and integration in space is performed in the domain under consideration . for the allen cahn model , the function @xmath4 is assumed to be a double - well potential with wells of equal depth . as a consequence ,the reaction function @xmath5 have a cubic - type behavior with two stable and one unstable equilibria , usually normalized as @xmath6 and @xmath7 , respectively . in absence of diffusion ,@xmath8 , space variable @xmath9 becomes an external parameter and solutions generically converges pointwise to functions with values in @xmath10 with sharp transition layers generated at points where the initial datum changes sign . for small @xmath11 ,if the initial datum is a small perturbation of a function with values in @xmath10 with well - separated transition regions , diffusion determines in a short time - scale a smoothed version of the original configuration and , on a longer time - scale , layers interact giving raise to front motion .when the space variable is one - dimensional , starting from @xcite , it has been shown that , as far as layers are well - separated the interaction force is very weak and the consequent motion is very slow .the meaning of weak / slow can be quantified more precisely , as discussed in what follows .postponing such details , in the regime @xmath12 , the original configuration is preserved for a long time and thus such behavior has been classified as _metastability_. many papers has been devoted to slow motion analysis for the allen cahn equation providing precise description of the relation between the size of the diffusivity @xmath13 and the time - scale of the dynamics .a complete list of references would be prohibitive . here, we only quote the analysis on generation , persistence and annihilation of metastable patterns performed in @xcite . a large class of different evolutive pdes , concerning many different areas , exhibits the phenomenon of metastability . without claiming to be complete , we list some of the principal models that has been analyzed : scalar conservation laws @xcite , cahn hilliard equation @xcite , gierer meinhardt and gray scott systems @xcite , keller segel chemotaxis models @xcite , generalgradient flows @xcite , high - order systems @xcite , gradient systems with equal depth multiple - well potentials @xcite , cahn morral systems @xcite , jin xin system @xcite .the aforementioned bibliography is confined to one - dimensional models ; however , there is vast literature of works about motion of interfaces in several space dimensions , where the effect of the curvature of the interfaces turns out to be relevant for the dynamics . in particular , for the allen cahn equation , we recall the works @xcite , where it has been shown that steep interfaces are generated in a short time with subsequent motion governed by mean curvature flow . the present paper is devoted to the analysis of metastability in a hyperbolic framework .precisely , given @xmath11 , @xmath14 $ ] , @xmath15 and @xmath16 , we consider here the _hyperbolic allen cahn equation _@xmath17 the function @xmath5 is required to be the derivative of a double well potential @xmath4 with non - degenerate minima of same depth , and the function @xmath18 is assumed to be strictly positive , uniformly with respect to @xmath19 ; namely , we assume @xmath20 where the constant @xmath21 may depend on @xmath22 . the uniform positivity of @xmath18 in is crucial , because it guarantees the dissipative nature of the model .if , in addition , @xmath23 as @xmath24 , we formally recover from in the ( singular ) limit . in the case@xmath25 , equation can be obtained by adding a nonlinear zero order perturbation to the damped wave operator @xmath26 .for such a choice , many studies have been devoted to fronts stability mainly in one space dimension for bistable or monostable reaction term @xmath5 ( see @xcite and references therein ) .interface formation has been analyzed in @xcite in the singular limit @xmath27 for space dimension equal to @xmath28 or @xmath29 , showing that motion is governed by mean curvature flow , as is the case for the corresponding parabolic model .the choice of the hyperbolic variation is motivated by the observation that there are different ways for modeling transport mechanisms .the one at the base of is the classical fourier law , originally proposed for heat conduction and then extended to many other different fields , which prescribes the istantaneous proportionality between the flux @xmath30 of a quantity with `` density '' @xmath19 and its gradient , @xmath31 .such choice has the advantage of providing a simple equation enjoying a number of useful properties ( smoothing effects , self - similarity ... ) , but , at the same time , it has a number of drawbacks , the most known being the presence of infinite speed of propagation .still in the framework of heat conduction modeling , following some ideas developed by maxwell in the context of kinetic theories , cattaneo proposed in @xcite a different law for the heat flux @xmath30 , based on the assumption that the equilibrium between flux and gradient of the unknown is asymptotical with a time - scale measured by the relaxation parameter @xmath32 , that is @xmath33 ( an extensive discussion is reported in @xcite ) . in the one - dimensional case , the diffusion equation given by the law has also a probabilistic interpretation , appearing in the description of _ correlated random walks _ ( see @xcite ) to be compared with the standard _random walk _ , which gives raise to the standard parabolic diffusion equation .criticisms to the application of the use of fourier - type law has been given also in modeling reaction - diffusion phenomena ( see @xcite ) .in presence of a reaction term described by the function @xmath5 , application of the maxwell cattaneo law gives @xmath34 to be considered as a modification of the standard allen cahn equation when @xmath5 satisfies ( see also @xcite , for different origins of the same equation ) .this equation fits into with the choice @xmath35 .we refer to this specific model as the _cahn equation with relaxation _ , reminiscent of the relaxation - type law .existence and nonlinear stability of traveling wave solutions for this equation has been analyzed in details in @xcite for general bistable reaction terms in one space dimension .this study is devoted to the one - dimensional case , so that the hyperbolic allen cahn equation reads as @xmath36 corresponding to the parabolic allen cahn equation @xmath37 where @xmath38 .specifically , we are interested in the limiting behavior @xmath39 with the aim of extending the metastable dynamics for to the hyperbolic case and focusing the attention on eventual differences . in @xcite , adapting the _ energy approach _proposed by bronsard and kohn @xcite for the parabolic equation , the first author has shown that , if the initial profile @xmath40 has a transition layer structure and the initial velocity @xmath41 is small , the solution maintains the transition layer structure on a time scale of order @xmath42 with @xmath43 arbitrary .the energy approach has also been applied to cahn hilliard equation in @xcite .grant @xcite improved this method to prove exponentially slow motion for cahn morral systems .a different procedure , proposed by carr and pego in @xcite , permits to prove existence and persistence of metastable states for the allen cahn equation for a time proportional to @xmath44 .this strategy provides also an explicit differential equation for the dynamics of the transition layer positions ( far from collapses ) .the method is based on the construction of an @xmath0-dimensional base manifold @xmath45 consisting of functions which approximate metastable states with @xmath0 transition layers .the manifold is not invariant , but if the initial datum is in a small neighborhood of @xmath45 , then the solution remains near the manifold for a time proportional to @xmath44 . based on these ideas ,slow motion results have been proved for the cahn hilliard equation by alikakos et al .@xcite and by bates and xun @xcite .in particular , the last ones use the same manifold constructed in @xcite . here ,we adapt the method of @xcite to the hyperbolic allen cahn equation embedding the base manifold @xmath45 in an extended phase space determined by the presence of the additional unknown @xmath46 as suggested by the first - order form of equation given by @xmath47 system , considered here for @xmath48 and @xmath49 , is complemented with homogeneous neumann boundary conditions @xmath50 and initial conditions @xmath51 the initial - boundary value problem -- is globally well - posed for positive times in @xmath52 . in particular ,if @xmath53 the solution @xmath54 is classical and belongs to @xmath55 ( among others , see ( * ? ? ?* appendix a ) ) . then, our aim is to describe the dynamics of such globally defined solution , at least for a class of `` well - prepared '' initial data . under assumptions , the hyperbolic equation supports traveling wave solutions connecting the equilibria @xmath56 and @xmath57 , i.e. solution @xmath58 such that @xmath59 , if and only if @xmath60 . indeed ,substituting the traveling wave ansatz in the equation , we obtain @xmath61 and thus , multiplying by @xmath62 and integrating over @xmath63 , we get @xmath64 from which we deduce , under and , that the velocity @xmath65 is zero . with such choice, it is well - known that , up to translation , there is a unique solution to the problem @xmath66 normalizing @xmath67 by adding the condition @xmath68 , transitions layer from @xmath56 to @xmath69 ( or viceversa ) are described by @xmath70 for both equations and .steady states @xmath67 are at the base of the construction of the base manifold @xmath45 which we sketch here ( for precise definitions , see section [ preliminaries ] ) .fix @xmath71 and @xmath11 .given a configuration @xmath72 of @xmath0 layer positions ( with @xmath73 ) , we construct a function @xmath74 which approximates a metastable state with transition points at @xmath75 , by piecing together approximated versions of @xmath67 , that is @xmath76 or @xmath77 for @xmath78 ( see figure [ u^h - fig ] ) . with @xmath79 . , width=529 ]then , we consider the slow evolution of solutions when the transition points are well separated one from the other and bounded away from the boundary points @xmath7 and @xmath57 .fixed ( small ) @xmath80 , the admissible layer positions lie in the set @xmath81 where @xmath82 , @xmath83 and the _ base manifold _ is @xmath84 . inwhat follows , we fix a minimal distance @xmath85 with @xmath86 and we consider the parameters @xmath13 and @xmath87 such that @xmath88 for some @xmath89 to be chosen appropriately small . in such a way , the parameters @xmath87 and @xmath13 have the same order of magnitude .all of the subsequent estimates depend on @xmath0 and @xmath90 .denoted by @xmath91 the inner product in @xmath92 , to restrict the attention to a neighborhood of @xmath45 , we introduce the decomposition @xmath93 , where @xmath94 are such that the following _ orthogonality condition_ holds @xmath95 for some appropriate approximate tangent vectors @xmath96 . then , setting @xmath97 we consider triples @xmath98 in the set @xmath99 and the corresponding _ extended base manifold _@xmath100 next , we choose a tubular neighborhood of @xmath101 : given @xmath102 , we set @xmath103\leq\gamma \psi({\bm h})\bigr\},\ ] ] with the _ energy functional _ @xmath104 and the _ barrier function _@xmath105 defined by @xmath106&:=\tfrac12\int_0 ^ 1\bigl\{\varepsilon^2 w_x^2+f'(u^{\bm h})w^2\bigr\}dx + \tfrac12\tau\|v\|^2+\varepsilon\tau\langle w , v\rangle , \label{energy } \\\psi({\bm h})&:=\sum_{j=1}^n{\langle\mathcal{l}\bigl(u^{\bm h}\bigr),k^{\bm h}_j\rangle}^2 , \label{barrier}\end{aligned}\ ] ] where @xmath107 is the @xmath108norm .our main result states that the channel @xmath109 is invariant for an exponentially long time if the parameters @xmath110 and @xmath87 are appropriately chosen . in other words , the manifold @xmath101 is approximately invariant for the hyperbolic system .[ main ] let @xmath111 and @xmath112 be such that @xmath113 and - hold .given @xmath71 and @xmath114 , there exist @xmath115 and @xmath89 such that if @xmath116 satisfy , @xmath117 $ ] and the initial datum satisfies @xmath118<\gamma\psi({\bm h})\bigr\},\ ] ] then the solution @xmath54 to the initial - boundary value problem -- remains in @xmath109 for a time @xmath119 and there exists @xmath120 such that @xmath121 where @xmath122 , @xmath123 and @xmath124 denotes the maximum norm in @xmath125 .moreover , there exists @xmath120 such that @xmath126 the strategy to prove theorem [ main ] is the following .firstly , plugging the decomposition @xmath93 into system and using conditions , we obtain an ode - pde coupled system describing the dynamics for @xmath127 , see system .then , we show that , if the solution @xmath54 belongs to @xmath109 , the estimates and hold . next , we estimate the time @xmath128 taken for the solution @xmath54 to leave the channel @xmath109 . the boundary of @xmath109 is the union of two parts :the `` ends '' where @xmath129 , meaning @xmath130 for some @xmath131 and `` sides '' where @xmath132=\gamma\psi({\bm h})$ ] . using energy estimate, we infer that the solution can leave @xmath109 only through the ends . since , for , the transition points move with exponentially small velocity , the solution @xmath54 stays in the channel for an exponentially long time .as long as the solution @xmath54 remains in the channel @xmath109 , @xmath19 is a function with @xmath0 transition layers .the estimate ensures slow motion of solutions and gives a lower bound on the lifetime of the metastable states . in order to give further information on the motion of the transition layers and an upper bound on such lifetime, we study in details an approximation of the equation for @xmath133 , determined formally by the requirement that @xmath134 is an exact solution . such request is expected to be appropriate in the limit @xmath27 . in this way, we obtain a system of ordinary differential equations for @xmath133 which does not depend on @xmath94 and @xmath30 and has the form @xmath135 where @xmath136 and the ( weighted ) average @xmath137 of the continuous function @xmath138 is given by @xmath139 and @xmath140 is a function , depending on @xmath4 .equation has to be compared with the corresponding one for the parabolic case , which is @xmath141 . for the damped nonlinear wave equation @xmath25, we have @xmath142 , while for the allen cahn equation with relaxation we obtain @xmath143 . since @xmath144 is negative , the ( physical relevant ) relaxation case exhibits smaller friction effects with respect to the damped one ( details in section [ layer ] ) .system has a unique equilibrium point @xmath145 where @xmath146 is the unique zero of @xmath140 , that corresponds to the unique stationary solution @xmath147 of with @xmath0 transition layers , normalized by the condition @xmath148 , without loss of generality . in the parabolic case, @xmath146 is an unstable equilibrium point with @xmath0 positive eigenvalues ; whereas , for the hyperbolic model , @xmath145 is an unstable equilibrium point for with @xmath0 positive eigenvalues and @xmath0 negative eigenvalues .the rest of the paper is organized as follows . in section [ preliminaries ]we give all the definitions , preliminaries and the construction of the manifold @xmath45 .furthermore , we recall all the results of carr and pego @xcite needed to prove theorem [ main ] .section [ motion ] is devoted to the derivation of the equation of motion for the triple @xmath149 and to the proof of theorem [ main ] . in section[ layer ] , we deduce the approximating equation for @xmath133 , we prove that there is a unique equilibrium point @xmath145 and we study its stability . finally ,using singular perturbation theory , we show that , for @xmath22 small , if @xmath18 is uniformly bounded and @xmath23 a.e . as @xmath24 , the behavior of the solution to is the same of the parabolic case ( see theorem [ thm : tau0 ] ) .following @xcite , we construct the base manifold and collect estimates that we will use in the proof of our results . for fixed @xmath80 , we recall the definition @xmath150 where @xmath82 and @xmath83 . by construction , if @xmath151 , then @xmath152 .the idea is to associate to any @xmath153 a function @xmath154 which approximates a metastable state with @xmath0 transition points at @xmath75 by matching appropriate steady states of equation .the collection of @xmath155 determines a @xmath0-dimensional manifold . in order to describe the dynamics in a neighborhood of such manifold, the framework has to be complemented with a projection which permits to separate the solution into a component on the manifold and a corresponding remainder . for the allen cahn equation ,two different constructions have been proposed in @xcite and @xcite . in @xcite ,fusco and hale use functions @xmath70 with @xmath67 the solution of previously defined , and set @xmath156 , \quad j=1,\dots , n,\ ] ] where @xmath157 ( note that @xmath158 , @xmath159 ) .hence , they obtain a manifold @xmath160 composed by continuous functions @xmath161 with a piecewise continuous first order derivative that jumps at @xmath162 , @xmath163 .in particular , the elements of the manifold belong to @xmath164 and not to @xmath165 ( if @xmath166 .in addition , they construct a tubular neighborhood of @xmath160 with coordinates @xmath167 by setting @xmath168 where @xmath169 are the derivatives of @xmath161 with respect to @xmath170 . by construction , @xmath169 have disjoint supports and @xmath171 for all @xmath172 . in @xcite , it is also conjectured that equation has an invariant manifold @xmath173 near @xmath160 and that this manifold @xmath173 is a graph over @xmath160 . fusco and hale did not prove the existence of the invariant manifold , but assuming existence , they calculated a first approximation for @xmath173 and for the differential equation for @xmath174 describing the reduced flow .they also conjectured that metastable states with @xmath0 transitions are associated with the unstable manifold of stationary solutions of having @xmath0 layers .both conjectures have been proved in @xcite using a different base manifold , previously constructed in @xcite .the approach used by carr and pego is based on a different choice and matching of steady states , which provides functions @xmath155 , composing the base manifold @xmath175 , which are smooth in both @xmath9 and @xmath174 . the crucial difference with respect to the fusco hale approach , resides in the fact that , for @xmath38 , @xmath176 with major consequences on the location of @xmath177 with respect to the tangent space to @xmath175 at @xmath155 , as will be clear in the following presentation . in this paper, we follow the framework established by carr and pego adapting it to the case of the hyperbolic allen cahn equation .since the equation we consider corresponds to the system , the dynamics is determined by an additional unknown , the time derivative @xmath46 , and thus the base manifold @xmath175 has to be embedded in a extended vector space . here ,taking advantage of the fact that we are looking for a manifold that is only approximately invariant , we perform this extension in a trivial way , considering the _ extended base manifold _ @xmath178 . from now on ,we drop the letters @xmath179 in the symbol used for the manifolds .given @xmath180 , let @xmath181 be the solution to @xmath182 with @xmath183 in @xmath184 , and let @xmath185 be the solution to with @xmath186 in @xmath184 . observe that if @xmath187 satisfies , then @xmath188 using this formula , we can prove existence and uniqueness of the solutions @xmath189 .let @xmath190 with @xmath4 a smooth function satisfying .there exists @xmath191 such that , if @xmath192 , then the functions @xmath189 are well - defined and , denoting by @xmath193 we have that @xmath194 is an increasing function of @xmath195 and @xmath196 .this lemma is consequence of the fact that @xmath197 are absolute minima of @xmath4 and so , there exist periodic solutions of oscillating around @xmath7 .indeed , the existence of such solutions is guaranteed if there exist @xmath198 such that @xmath199 , @xmath200 and @xmath201 for all @xmath202 .this condition is certainly satisfied if @xmath194 are close to @xmath69 .let us consider the positive case @xmath181 and @xmath203 . by integrating in @xmath204 and using the boundary conditions in, we obtain @xmath205 the integral in tends to infinity as @xmath206 and it is an increasing function of @xmath207 for @xmath207 close to @xmath69 .hence , for @xmath195 sufficiently large , there exists a unique @xmath207 such that is satisfied and so the function @xmath181 is well - defined .the negative case @xmath185 and @xmath208 is similar .now , given @xmath209 , let us define @xmath210 . by definition , it follows that @xmath211 is the solution to @xmath212 with @xmath213 in @xmath214 , and @xmath215 is the solution to with @xmath216 in @xmath214 .moreover , the functions @xmath217 are well - defined if @xmath218 , they depend on @xmath13 and @xmath219 only through the ratio @xmath220 .finally , @xmath221 where @xmath120 is a constant depending only on @xmath4 . in particular , @xmath194 tends to @xmath69 as @xmath222 ( more details in proposition [ prop : alfa , beta ] ) . for @xmath153 with @xmath223, we define the function @xmath155 with @xmath0 transition points at @xmath75 by matching together steady states to with layer distance equal to @xmath219 , using smooth cut - off functions .given @xmath224 $ ] a @xmath225 function with @xmath226 for @xmath227 and @xmath228 for @xmath229 , set @xmath230 then the function @xmath155 is given by the convex combination @xmath231,\ ] ] and the base manifold for the equation is defined as @xmath232 if @xmath80 is sufficiently small and @xmath153 , then @xmath233 for @xmath9 near @xmath170 and @xmath234 away from @xmath170 for @xmath235 .therefore , states @xmath155 on the base manifold are well approximated near transition layers by @xmath161 . by definition, @xmath155 is a smooth function of @xmath9 and @xmath236 and enjoys the properties @xmath237 for any @xmath235 . in what follows ,we use the notation @xmath238 and we denote the tangent space to @xmath45 at @xmath155 by @xmath239 . at this point , the natural idea would be to construct a tubular neighborhood of @xmath45 , with coordinates @xmath240 where @xmath94 is orthogonal to @xmath241 . since @xmath45 is not invariant ,there is higher flexibility in the construction of its neighborhood and tubular co - ordinates near @xmath45 can be defined using approximate tangent vectors to @xmath45 .for @xmath235 , introduce the cutoff function @xmath242 given by @xmath243.\ ] ] then , the _ approximate tangent vectors _@xmath96 are defined by @xmath244 by construction , @xmath96 are smooth functions of @xmath9 and @xmath236 and are such that @xmath245,\\k^{\bm h}_j(x)&=-u^{\bm h}_x(x ) & \quad \textrm{for}\quad & x\in[h_{j-1/2}+2\varepsilon , h_{j+1/2}-2\varepsilon ] .\end{aligned}\ ] ] as above , we use the notation @xmath246 the definition of the approximate tangent vectors is motivated by the relations @xmath247 for @xmath248 $ ] .in addition , the multiplication by the cutoff term @xmath242 is reminiscent of the fact that the tangent space of @xmath160 is spanned by @xmath249 that have disjoint supports. the following estimates will be useful in the sequel .[ estimates - u^h_j ] let @xmath113 with @xmath4 satisfying .given @xmath71 and @xmath114 , there exist @xmath250 , and a function @xmath251 with @xmath252 as @xmath253 such that if @xmath13 and @xmath87 are chosen so that holds and @xmath254 , then @xmath255 @xmath256 for @xmath153 and @xmath235 .moreover , if @xmath257 , we have @xmath258 heuristically , the exponent of @xmath13 can be obtained by replacing @xmath259 and @xmath96 with @xmath260 .a function @xmath19 near @xmath45 may be written in terms of coordinates @xmath240 as @xmath93 , with @xmath94 satisfying the orthogonality condition . to state this result ,let us set @xmath261 [ tub - co ] there exist @xmath262 with @xmath151 and a smooth function @xmath263 such that , whenever @xmath264 , we have @xmath265 and @xmath266 moreover , defining @xmath267 by setting @xmath268 the function @xmath269 is injective , @xmath270 for all @xmath271 and the set @xmath272 is open in @xmath273 . in the last statement , constants @xmath274 can be chosen independent on @xmath13 . as stated in the introduction , the neighborhood @xmath109 of the extended base manifold @xmath101 is defined in term of the _ energy functional _ @xmath104 and the _ barrier function _@xmath105 , see and .the positivity of the first term in @xmath104 holds for @xmath87 small and @xmath94 satisfying the orthogonality condition .[ l^hw - theo ] let @xmath113 , with @xmath4 satisfying . given @xmath71 and @xmath114, there exist @xmath276 , such that if @xmath13 and @xmath87 are chosen so that holds and @xmath254 , then @xmath277 for any @xmath278 satisfying @xmath279 for @xmath235 .given @xmath280 , we consider the operator @xmath281 , linearization of @xmath282 about @xmath155 , i.e. @xmath283 if @xmath284 and @xmath285 , integrating by parts , we infer @xmath286 hence , in this case , the energy functional can be written as @xmath287=\tfrac12\langle w , l^{\bm h } w\rangle+\tfrac12\tau\|v\|^2+\varepsilon\tau\langle w , v\rangle,\ ] ] and from theorem [ l^hw - theo ] it follows that @xmath288 moreover , let @xmath289 $ ] be such that @xmath290 and let @xmath291 $ ] be such that @xmath292\}$ ] .assume without loss of generality @xmath293 ( otherwise replace @xmath294 by @xmath295 ) .we have @xmath296 and so , @xmath297 by applying theorem [ l^hw - theo ] and taking into account the latter bound with @xmath298 , we deduce also the estimate @xmath299 in order to provide representations of the barrier @xmath105 , defined in , we introduce some auxiliary functions .since @xmath300 depends only on the ratio @xmath301 , we can define @xmath302 by definition , @xmath300 is close to @xmath69 or @xmath56 and so , @xmath303 are close to @xmath7 .the next result characterizes the leading terms in @xmath304 and @xmath305 as @xmath306 .[ carr pego @xcite ] [ prop : alfa , beta ] let @xmath4 be such that holds and set @xmath307 .there exists @xmath308 such that if @xmath309 , then @xmath310 with corresponding asymptotic formulae for the derivatives of @xmath304 and @xmath305 .explicit expressions of @xmath311 in terms of @xmath4 can be found in @xcite .for @xmath312 , we set @xmath313 and @xmath314 for @xmath153 , since @xmath315 direct integration gives @xmath316 thus , the barrier function @xmath105 , defined in , can be written as @xmath317 the next statement collects some estimates we will use later on .[ prop - l(u^h ) ] let @xmath113 , with @xmath4 satisfying .given @xmath71 and @xmath114 , there exist @xmath318 , such that if @xmath13 and @xmath87 are chosen so that holds and @xmath254 , then @xmath319 where @xmath320 and @xmath321 . here, we give only an idea of the proofs , the complete ones can be found in @xcite for and in @xcite for .recalling the definition , for @xmath322 , we have @xmath323 where the remainder @xmath324 is given by @xmath325 using lagrange interpolation formula , @xmath326 with @xmath327. it can be shown ( see ( * ? ? ?* lemma 8.2 ) ) that there exist @xmath120 such that for @xmath328 $ ] , @xmath329 , we have @xmath330 provided @xmath331 and @xmath332 is sufficiently small . using these estimates , the fact that @xmath333 if @xmath334 and that the @xmath335-th derivative of @xmath336 is uniformly bounded ( independently on @xmath13 ) , we obtain @xmath337 then , the @xmath338-bound follows since @xmath177 has support of length @xmath339 in @xmath340 .the estimate is obtained in a similar way , by differentiating with respect to @xmath236 .in this section we study the dynamics of - in a neighborhood of @xmath101 using the decomposition @xmath93 and deriving the system of equations for @xmath341 .such description will be used to prove theorem [ main ] .let @xmath54 be a classical solution of - , with @xmath19 lying in the tubular neighborhood @xmath342 for @xmath343 $ ] with @xmath344 .let @xmath345 and @xmath346 , where @xmath155 is defined by .we recall that @xmath347 for @xmath343 $ ] means that @xmath348 , @xmath349 and @xmath350 for @xmath343 $ ] .moreover , @xmath351 for @xmath343 $ ] . from itfollows that the pair @xmath352 satisfies @xmath353 where @xmath354 denote the inner product in @xmath125 .expanding , we get @xmath355 and @xmath281 is the differential operator defined in .differentiating with respect to @xmath356 the orthogonality condition , we obtain @xmath357 using the notation @xmath358 equation becomes @xmath359 from and , we obtain the ode - pde coupled system @xmath360 the matrix @xmath361 is diagonally dominant , because , for any @xmath362 there exists @xmath363 such that if @xmath364 , then @xmath365 thanks to proposition [ estimates - u^h_j ] .also , for a known property of inverses of diagonally dominant matrices ( see @xcite ) , @xmath361 is invertible and it holds @xmath366 where @xmath367 denotes the operator norm induced by the norm @xmath124 . in section [ layer ] , it is determined the explicit expression for the principal term in the expansion of the inverse @xmath368 as @xmath27 .the invertibility of the matrix @xmath369 descends from the smallness of @xmath370 for @xmath371 and . indeed , for @xmath372 ,applying proposition [ estimates - u^h_j ] , we infer @xmath373 and thus @xmath374 therefore , the matrix @xmath375 is invertible for @xmath87 and @xmath376 sufficiently small and @xmath377 applying @xmath378 in the equation for @xmath174 , we obtain the final form of the system @xmath379 the proof of our main result consists in providing estimates for the solutions to . to start with , we observe that if @xmath372 for @xmath380 small then there exists @xmath120 such that @xmath381 using and the third estimate in proposition [ estimates - u^h_j ] . in order to prove theorem [ main ] , we restrict the attention to the set @xmath382\leq\gamma\psi(\bm h)\bigr\}.\ ] ] the aim of the next result is twofold .firstly , it states that if the triple @xmath341 belongs to @xmath383 then the bound on @xmath352 stated in theorem [ main ] , estimate , holds true . secondly , assuming in addition that @xmath341 is a solution to , then also the bound on @xmath384 in theorem [ main ] , estimate , is valid .[ prop : e > ] let @xmath385 be such that holds and @xmath112 .given @xmath71 and @xmath114 , there exist @xmath318 , such that for @xmath13 and @xmath87 satisfying , * if @xmath386 , then @xmath387 , \\ \tfrac14 \lambda \|w\|^2 + \tfrac14\tau\|v\|^2\leq \mathcal{e}^{\bm{h}}[w , v ] , \\ \mathcal{e}^{\bm{h}}[w , v]\leq c\gamma\exp(-2a\ell^{\bm h}/\varepsilon ) , \end{aligned}\ ] ] where @xmath388 is the positive constant introduced in theorem [ l^hw - theo ] ; * if @xmath386 is a solution of for @xmath343 $ ] , then @xmath389 let us prove the first inequality in . using young inequality ,we have @xmath390 and so , recalling the expression for the energy @xmath104 given in , @xmath391\geq \tfrac12\langle w , l^{\bm h}w\rangle+\tfrac14\tau\|v\|^2 -\varepsilon^2\tau\|w\|_{{}_{l^\infty}}^2.\ ] ] using , we obtain , for @xmath392 , @xmath391\geq\bigl(\tfrac14\lambda-\varepsilon\tau\bigr ) \varepsilon\|w\|^2_{{}_{l^\infty } } + \tfrac14\tau\|v\|^2\geq \tfrac18\lambda\varepsilon\|w\|^2_{{}_{l^\infty}}+\tfrac14\tau\|v\|^2.\ ] ] moreover , from , for @xmath393 one has @xmath391\geq\bigl(\tfrac12\lambda-\varepsilon^2\tau\bigr ) \|w\|^2 + \tfrac14\tau\|v\|^2\geq \tfrac14\lambda\|w\|^2+\tfrac14\tau\|v\|^2,\ ] ] concluding the first two inequalities of .the upper bound for @xmath132 $ ] follows from the definition of @xmath383 , the expression of the barrier @xmath105 given in and proposition [ prop : alfa , beta ] . to prove part ( ii ) , using , we deduce the estimate @xmath394 since , by definition and , @xmath395 .hence , for @xmath13 sufficiently small , both @xmath87 and @xmath376 are small , and thus , for @xmath372 , estimate holds . therefore , if @xmath386 , then @xmath372 and the estimate is obtained by applying in .now , we estimate the time @xmath396 taken for the solution @xmath54 to leave @xmath109 . to do this , we study the system in the set @xmath383 by using energy estimates .[ prop : d / dte ] let @xmath385 and @xmath112 be such that and hold . given @xmath71 and @xmath114 , there exist @xmath115 and @xmath89 such that if @xmath117 $ ] , @xmath116 satisfy and @xmath386 is a solution of for @xmath343 $ ] , then for some @xmath362 , we have @xmath397-\gamma\psi(\bm h)\bigr\ } \leq-\eta\,\varepsilon\bigl\{\mathcal{e}^{\bm{h}}[w , v]-\gamma\psi(\bm h)\bigr\ } \qquad\textrm{for}\quad t\in[0,t].\ ] ] in all the proof , symbols @xmath398 denote generic positive constants , independent on @xmath13 , and with @xmath362 .let us recall that , if @xmath149 is a solution to , then @xmath399 direct differentiation and the self - adjointness of the operator @xmath400 give @xmath401 using cauchy schwarz inequality and the estimates in proposition [ prop - l(u^h ) ] and [ estimates - u^h_j ] , we infer @xmath402 for @xmath403 , applying , and using young inequality , we get @xmath404 for what concerns the second term in the energy @xmath104 , it holds @xmath405 finally , the time derivative of the scalar product @xmath406 can be bounded by @xmath407 where , in particular , the inequalities @xmath408 have been used . collecting the estimate for the three terms composing @xmath409, we deduce @xmath410\|v\|^2\\ & \hskip1.0 cm + c\bigl\ { \exp(-2a\delta/\varepsilon)\bigl ( 1 + \varepsilon^{-1}\gamma\bigr ) + \varepsilon( \varepsilon+\|w\|_{{}_{l^\infty}})\bigr\}\|w\|^2+(c+1)\|\mathcal{l}(u^{\bm h})\|^2\\ & \leq -\varepsilon\langle w , l^{\bm h}w\rangle+c\varepsilon\bigl\ { \gamma\exp(-c/\varepsilon)+\varepsilon\bigr\}\|w\|^2-\eta c_g\|v\|^2+c\|\mathcal{l}(u^{\bm h})\|^2 , \end{aligned}\ ] ] for @xmath13 and @xmath411 small .thus , from and @xmath412 it follows that @xmath413 hence , for @xmath414 , with @xmath415 small ( and dependent on @xmath110 ) , we deduce the bound @xmath416 substituting , we infer @xmath417 again for .finally , for @xmath415 sufficiently small , we obtain @xmath418 direct differentiation gives @xmath419 using the estimates provided by proposition [ estimates - u^h_j ] and by , , we deduce @xmath420 thus , observing that @xmath421 , we infer the bound @xmath422 using the inequality , we obtain @xmath423 hence , observing that @xmath424 , we end up with @xmath425 combining and , we obtain that if @xmath386 is a solution of , then @xmath426-\gamma\psi(\bm h)\bigr\ } \leq -\eta\,\varepsilon\mathcal{e}^{\bm h}+c\bigl(\varepsilon+\gamma^2\exp(-c/\varepsilon)\bigr)\psi,\ ] ] for some @xmath362 .therefore the estimate follows from @xmath427 and the latter is verified for @xmath428 $ ] , provided @xmath414 with @xmath415 sufficiently small so that @xmath429 .now , we have all the tools needed to prove theorem [ main ] . let @xmath430 and let @xmath54 be the solution of -- .assume that @xmath431 for @xmath432 $ ] , where @xmath128 is maximal .then , @xmath93 and @xmath386 solves the system for @xmath432 $ ] .let us apply proposition [ prop : d / dte ] ; from , it follows that @xmath433-\gamma\psi(\bm h))\bigr\}\leq0 ,\quad \qquad t\in[0,t_\varepsilon]\ ] ] and so , @xmath434-\gamma\psi(\bm h)\}(t)\leq\{\mathcal{e}^{\bm{h}}[w , v]-\gamma\psi(\bm h)\}(0)<0 , \qquad \quad t\in[0,t_\varepsilon].\ ] ] therefore , the solution @xmath54 remains in the channel @xmath109 while @xmath153 and if @xmath435 is maximal , then @xmath436 , that is @xmath437 for proposition [ prop : e > ] , in the channel the solution satisfies and .in particular , the transition points move with exponentially small velocity .this implies that @xmath54 remains in the channel for an exponentially long time .indeed , from it follows that for all @xmath432 $ ] , one has @xmath438 where @xmath439 is the minimum distance between layers at the time @xmath356 . combining and, we obtain @xmath440 hence , using we have @xmath441 and the proof is complete .in the previous section , we derived the equation for the motion of the transition points and , by studying the ode - pde coupled system , we obtained the estimate for the velocity of the transitions . in this section ,we derive an ordinary differential equation approximating the equation for @xmath236 to obtain further information on the motion of the transition points and analyze the differences with the parabolic case .since @xmath94 is very small , we use the approximation @xmath442 in and then @xmath443 in order to eliminate @xmath30 , let us differentiate and multiply by @xmath22 equation .we have @xmath444 using the approximation @xmath445 , we obtain @xmath446 let us denote by @xmath447 the hessian of @xmath155 with respect to @xmath236 and by @xmath448 the quadratic form associated to @xmath447. simplifying , we get @xmath449 for @xmath450 . if @xmath451 is a solution of the hyperbolic allen cahn equation , @xmath452 satisfies .respect to the parabolic case , there are two new terms : the term involving @xmath453 and the one involving the quadratic form associated to the hessian of @xmath155 .we can neglect the latter term for two reasons .firstly , @xmath454 is exponentially small by estimate and thus the terms with @xmath455 are negligible .secondly , consider the terms @xmath456 .they are exponentially small as @xmath39 , because @xmath96 is supported in @xmath457 $ ] , in this interval @xmath233 and @xmath458 .using these approximations , the terms with @xmath459 are zero .if @xmath460 , we have @xmath461 because @xmath462 for all @xmath235 . by neglecting the quadratic term ,we rewrite as @xmath463 where @xmath464 with @xmath368 the inverse of the matrix introduced in section [ motion ] : @xmath465 as proven by carr and pego ( * ? ? ?* corollary 3.6 ) , defining @xmath466 there exists @xmath120 such that if @xmath87 is sufficiently small and @xmath153 , we have @xmath467 where @xmath468 . since @xmath469 , it follows that @xmath470 depends essentially on the differences between values of the potential @xmath471 .similar result holds for the matrix @xmath472 , namely for the scalar products @xmath473 , thus generalizing the aforementioned result to the case @xmath474 . to this end , we recall the following result ( see ( * ? ? ?* lemmas 7.8 - 7.9 - 8.1 ) ) . in order to compute @xmath480, one needs to obtain an expression for @xmath481 .since @xmath482 solution of depends on @xmath219 through its boundary value , the latter can be obtained by differentiating the integrated version of that equation with respect to @xmath219 , that is , the @xmath13rescaled version of .finally , for @xmath483 $ ] , we end up with @xmath484 and @xmath485 is an even function of @xmath9 satisfying @xmath486;\ ] ] see ( * ? ? ?* lemma 7.8 ) for details .firstly , we recall that @xmath96 is supported in @xmath340 .since the support of @xmath259 is contained in @xmath475 $ ] , we have . from lemma [ lem : u^h_j ], it follows that @xmath495 , for @xmath322 and @xmath496 for @xmath497. then , @xmath498 but , @xmath499 is of one sign in @xmath340 , therefore @xmath500 and we obtain .it remains to prove . to do this , for @xmath322 ,we write @xmath501 and @xmath502 , where @xmath503 then , @xmath504 from and the definition of @xmath505 , it follows that @xmath506 where @xmath507 by writing @xmath508 , we get @xmath509 using and the estimate @xmath510 , we deduce that @xmath511 satisfies @xmath512 moreover , for we have @xmath513 similarly , we can estimate the other term and obtain @xmath514 .let us now compute @xmath515 the other remaining term in is evaluated similarly .to do this , we observe that since @xmath516 and @xmath517 is solution of , positive or negative respectively , we have @xmath518 where @xmath519 . inwhat follows , we are considering the case @xmath520 in @xmath521 $ ] ( i.e. @xmath131 odd ) ; treatment of @xmath522 is similar . by using and changing variable, we obtain @xmath523 since @xmath524 for @xmath525 and @xmath526 for @xmath527 , we have @xmath528 then , we can conclude that @xmath529 moreover , from and , it follows that for @xmath248 $ ] we have @xmath530 we claim that @xmath531 indeed , @xmath532 in @xmath533 $ ] and @xmath534,\\ & \phi^{j+1}_x(x ) \qquad \qquad & x\in[h_{j+1/2}-2\varepsilon , h_{j+1/2 } ] .\end{aligned}\right.\ ] ] using the fact that @xmath535 for @xmath536 , we obtain and so , @xmath537 .therefore , @xmath538 also , @xmath539 because @xmath540 is supported on a set of measure proportional to @xmath13 .now we are ready to analyze the term @xmath472 . to this aim ,let us introduce the constant @xmath541 then , in view of , we obtain @xmath542 and , being @xmath543 one has @xmath544 hence , for @xmath545 , @xmath546 therefore , we study the system of ordinary differential equations @xmath547 in the case of the damped hyperbolic allen cahn equation , @xmath548 and therefore @xmath549 .moreover , for the allen cahn equation with relaxation , @xmath550 and @xmath551 hence , in the latter case , the effect of the parameter @xmath32 is present also in the friction term @xmath552 , and in particular it speeds up the dynamics with respect to the simpler damped hyperbolic allen cahn equation , being the coefficient smaller .this richer effect on the dynamics in the present analysis confirms what has been already observed in the study of traveling waves in @xcite , where again the relaxation parameter @xmath22 in the case of the allen cahn equation with relaxation affects the speed of the wave also though a modification of the friction effects .now , if @xmath553 as @xmath24 , taking formally the limit in we obtain the system @xmath554 .the structure of solutions of this system of ordinary differential equations is studied in @xcite to describe the evolution of the layer positions in the parabolic case .we can write @xmath555 where @xmath556 is defined in the following way . for @xmath557 , define @xmath558 by @xmath559 and set @xmath560 for @xmath131 even , @xmath561 for @xmath131 odd . for @xmath153 let @xmath562.\ ] ] since @xmath563 and @xmath564 , we have @xmath565\\ & = -\varepsilon d_\infty^{-1}(\alpha^{j+1/2}-\alpha^{j-1/2 } ) . \end{aligned}\ ] ] then , we can write as @xmath566 . for a solution @xmath236 with values in @xmath567 ,the energy @xmath568 is nonincreasing and @xmath569 note that @xmath570 large implies a greater dissipation of energy . if @xmath571 is a critical point of @xmath556 , then @xmath572 for @xmath235 .define @xmath573 .from proposition [ prop : alfa , beta ] , it follows that for @xmath574 sufficiently small @xmath575 are monotone and so , @xmath576 for @xmath163 .let @xmath577 , @xmath578 . since @xmath153 , we have that @xmath579 and then @xmath580 .the condition @xmath572 , also , gives @xmath581 .using proposition [ prop : alfa , beta ] , we have that critical points correspond to zeros of @xmath582 for @xmath583 $ ] with @xmath584 . for @xmath87sufficiently small , @xmath585 when @xmath586 and @xmath587 when @xmath588 ; hence a critical point must exist .it is unique , because @xmath589 and @xmath590 are monotone . to determine the nature of the critical point ,consider the hessian of @xmath556 , @xmath591 .the matrix @xmath592 is symmetric and tri - diagonal with @xmath593 where @xmath594 . from proposition [ prop :alfa , beta ] , @xmath595 for @xmath235 .then , for @xmath596 , @xmath597 so that @xmath592 is negative definite and @xmath571 is a local maximum .if @xmath5 is odd , all the @xmath598 subscripts can be ignored , e.g. @xmath599 , @xmath600 and @xmath601 for @xmath235 . in general , the steady state domain lengths satisfy @xmath602 thus , system has a unique equilibrium point @xmath603 with @xmath604 . from proposition [ prop - l(u^h ) ] , @xmath605 andso @xmath606 is a stationary solution of .in other words , there is a unique stationary solution @xmath147 of with @xmath0 transition layers and @xmath148 ; moreover , @xmath607 with @xmath608 where @xmath609 .note , also , that by definition , @xmath610 if and only if @xmath611 so that the channel @xmath109 is `` pinched '' at @xmath612 . .25 cm from lemma [ lem - w ] , it follows that the system has a unique equilibrium point @xmath603 . to determine the stability of this stationary point, we have to analyze the eigenvalues of the block matrix @xmath614 where , as above , @xmath592 is the hessian matrix of @xmath556 evaluated at @xmath571 . to this end, we make use of the schur complement , defined for a general block @xmath615 matrix @xmath616 as follows : @xmath617 , provided @xmath618 is an invertible square matrix . in this case ,@xmath619 . for @xmath620, we have @xmath621 where @xmath622 then , @xmath623 it follows that @xmath624 is an eigenvalue of @xmath625 if and only if @xmath626 is an eigenvalue of @xmath592 . as previously shown in lemma [ lem - w ] , @xmath592 is symmetric and negative definite , so all the eigenvalues of @xmath592 are negative .denote them by @xmath627 for @xmath628 .for each @xmath627 there are two eigenvalues of @xmath625 : @xmath629 in conclusion , the jacobian matrix evaluated at @xmath603 has @xmath0 positive eigenvalues @xmath630 and @xmath0 negative eigenvalues @xmath631 , and so @xmath603 is unstable .the eigenvalues satisfy @xmath632 if @xmath633 . in particular , if @xmath634 , @xmath630 converge to the eigenvalues of the parabolic case , as expected .to conclude this section , we use singular perturbation theory to compare , for @xmath22 small , the solutions of the system and the ones of @xmath635 that is obtained by substituting @xmath636 in , assuming that @xmath553 as @xmath637 .denote by @xmath638 the solutions of ; @xmath639 is solution of the system @xmath554 , that describes the evolution of layer positions in the parabolic case .set @xmath640 a general theorem of tihonov on singular perturbations could be applied to systems - .specifically for the system we have the following result .[ thm : tau0 ] let @xmath641 be a solution of and @xmath638 a solution of , with @xmath642 for any @xmath343 $ ]. then , there exists @xmath120 ( independent of @xmath22 ) such that @xmath643.\ ] ] moreover , @xmath644,\label{eta - inf}\end{aligned}\ ] ] for all @xmath645 .in particular , from , and , it follows that , if @xmath553 and @xmath646 as @xmath24 , then @xmath647}|\bm h(t)-\bm h_p(t)| = \lim_{\tau\rightarrow0 } \int_0^t|\bm\eta(t)-\bm\eta_p(t)|dt = \lim_{\tau\rightarrow0}\sup_{t\in[t_1,t]}|\bm\eta(t)-\bm\eta_p(t)|=0,\ ] ] for any @xmath645 .for @xmath343 $ ] , define @xmath648 by hypotheses , @xmath649 for @xmath343 $ ] , so @xmath650 and @xmath651 are uniformly bounded in @xmath652 $ ] . since @xmath653 is a regular function of @xmath236 , there exists @xmath120 such that @xmath654 for all @xmath343 $ ] .here and in what follows , @xmath655 is a positive constant independent of @xmath22 whose value may change from line to line .we have @xmath656 since @xmath657 for any @xmath658 , using and cauchy schwarz inequality , we obtain @xmath659 summing , one has @xmath660 and so @xmath661.\ ] ] integrating and applying grnwall s lemma , we obtain .in particular , from , it follows that @xmath662.\ ] ] substituting in the equation for @xmath663 , we obtain @xmath664 integrating , we obtain .furthermore , for , we have @xmath665 and so @xmath666 for @xmath343 $ ] .therefore , for any fixed @xmath645 , we obtain .s. r. dunbar and h. g. othmer . on a nonlinear hyperbolic equation describing transmission lines , cell movement , and branching random walks . in : othmer h.g .( ed . ) , _ nonlinear oscillations in biology and chemistry _ , lecture notes in biomath .66 , springer - verlag berlin , 1986 .k. p. hadeler .reaction transport systems in biological modelling . in : capasso v. and diekmann o. ( eds . ) , _ mathematics inspired by biology ( martina franca , 1997 )_ , lecture notes in math .1714 , springer - verlag berlin , 1999 .w. d. kalies , r. c. a. m. vandervorst and t. wanner .slow - motion in higher - order systems and @xmath110-convergence in one space dimension ._ nonlinear analysis : theory , methods and applications _ , * 44 * ( 2001 ) , 3357 . c. lattanzio , c. mascia , r. g. plaza and c. simeoni .analytical and numerical investigation of traveling waves for the allen cahn model with relaxation . _ math .models and methods in appl ._ , * 26 * ( 2016 ) , 931985 .w. sun , m. j. ward and r. russell .the slow dynamics of two - spike solutions for the gray scott and gierer meinhardt systems : competition and oscillatory instabilities ._ siam j. appl ._ , * 4 * ( 2005 ) , 904953 .

<S> metastable dynamics of a hyperbolic variation of the allen cahn equation with homogeneous neumann boundary conditions are considered . using the `` dynamical approach '' proposed by carr </S><S> pego @xcite and fusco </S><S> hale @xcite to study slow - evolution of solutions in the classic parabolic case , we prove existence and persistence of metastable patterns for an exponentially long time . in particular , we show the existence of an `` approximately invariant '' @xmath0-dimensional manifold @xmath1 for the hyperbolic allen cahn equation : if the initial datum is in a tubular neighborhood of @xmath1 , the solution remains in such neighborhood for an exponentially long time . </S><S> moreover , the solution has @xmath0 transition layers and the transition points move with exponentially small velocity . </S><S> in addition , we determine the explicit form of a system of ordinary differential equations describing the motion of the transition layers and we analyze the differences with the corresponding motion valid for the parabolic case . </S>

1110.3342

in recent years it has been shown that one can create a condensed matter / gauge theory duality using the ads / cft correspondence @xcite .this duality can give us insight in how a strongly coupled system behaves by reformulating the theory in terms of a weakly coupled theory of gravity .as of now the hall effect @xcite , nernst effect @xcite , and superfluidity / superconductivity @xcite have been explored . some reviews are @xcite .most of the work has been done with superconductivity .initially the probe limit was explored , in which the chemical potential @xmath13 whereas the charge of the scalar hair @xmath14 so that the product @xmath15 remains finite @xcite .this work was later extended to include the full backreaction on the geometry of the system @xcite .magnetic fields have also been added and found to exhibit the behavior of type - ii superconductors @xcite .other gravitational duals have been able to describe systems exhibiting striped superconductivity ( see , e.g. , @xcite ) .all dual systems show hallmarks of the strongly coupled , high @xmath2 superconductors . in this paperwe explore the properties of the system at the opposite end of the probe limit at which the charge of the scalar hair @xmath16 is close to zero and the critical temperature @xmath2 is small ( small @xmath17 limit ) corresponding to a near extremal black hole . in @xcite , hartnoll ,_ et al . _ , discovered that a phase transition is allowed even for @xmath11 ( see also discussion in @xcite ) .it is not clear what the physical mechanism responsible for the condensation is near @xmath11 , and whether it is related to the one in the probe limit ( @xmath14 ) . to address these questions ,we explore the near extremal limit by performing analytic calculations of the properties of the system .there is a critical line in the two - dimensional parameter space @xmath18 defined by @xcite [ eqcr ] q^2 = , which is determined by the ads@xmath7 geometry of the horizon at extremality .instability is expected above this line , with the critical temperature approaching zero as one approaches the critical line . by using trial functions ,we show that indeed , there is an instability above the critical line , but the critical temperature vanishes on the line only for @xmath19 .we compute the scaling dimension @xmath9 analytically ( numerically , @xmath10 ) . for @xmath8 , the critical temperature does not vanish on the line and there is a phase transition below this line .we find an analytic expression for the true critical line ( where @xmath20 ) for @xmath8 .interestingly , for @xmath11 , we obtain instability down to the minimum value @xmath21 determined by unitarity ( at @xmath21 , we obtain @xmath20 ) .we also calculate analytically the energy gap [ eqeg ] e_g = , which is found to diverge as we approach the critical line ( @xmath22 ) .a calculation of the conductivity ( both analytic and numerical ) shows that the profile of the conductivity exhibits a universal behavior near criticality .in detail , in section [ sec2 ] we set up the system and fix the notation . in section [ sec3 ] , we determine the critical temperature analytically . in section [ sec4 ]we study the system below the critical temperature and calculate the energy gap both analytically and numerically . in section [ sec5 ]we calculate the conductivity analytically near criticality and compare with numerical results .finally , in section [ sec6 ] we summarize our conclusions .we are interested in the dynamics of a scalar field of mass @xmath23 and electric charge @xmath16 coupled to a @xmath24 vector potential in the dynamical backgound of a @xmath25 dimensional ads black hole .the action is [ eqs1 ] s = d^4 x , where @xmath26 . for simplicity, we shall work with units in which @xmath27 and @xmath28 .it is convenient to write the complex scalar field in terms of real scalar fields @xmath29 and @xmath30 as = e^iq. then the action reads [ eqs3 ] s = d^4 x .the phase @xmath30 plays the role of a stckelberg field giving mass the the vector potential when the real scalar @xmath31 condenses .the local @xmath24 transformation reads a_a_+ _ , + . we can fix the gauge by setting @xmath32 .the form of the action ( [ eqs3 ] ) is more general than ( [ eqs1 ] ) , because in the former , we can continue @xmath1 to negative values .this leads to a negative squared mass for the vector potential which signals an instability in flat space .however , in ads space negative squared masses are allowed as long as they are above the bf bound .we shall allow @xmath1 to take negative values and find that there is a lower bound for a phase transition to occur .to solve the equations of motion , consider a black hole with planar horizon and metric ds^2 = .we have chosen the coordinate @xmath33 so that the horizon is at @xmath34 ( @xmath35 ) .the system is kept under fixed charge density @xmath36 .we shall denote by @xmath37 the value of the charge density in units in which the radius of the horizon is 1 .all physical quantities are measured in units of the horizon radius , which hides an arbitrary scale in the system .this is remedied by reporting on scale - invariant ( dimensionless ) quantities only , as we shall be doing .assuming that the scalar field is of the form @xmath38 and the potential is an electrostatic scalar potential @xmath39 , the equations of motion are @xcite [ eq1 ] + + & = & 0 , + + - & = & 0 , + -+ z^2 + e^&= & 0 , + ^2 + ^2 e^- + + + & = & 0 , where prime denotes differentiation with respect to @xmath33 and to be solved in the interval @xmath40 , where @xmath34 is the horizon and @xmath41 is the boundary . near the boundary ( @xmath42 ) , we have @xmath43 , @xmath44 and so approximately [ eq3 ] ^ ( ) z^ _ , - z , where _= .for @xmath45 ( breitenlohner - freedman ( bf ) bound @xcite ) , @xmath46 have an imaginary part and the system is unstable . for @xmath47 ,both modes are normalizable .while a linear combination of these modes is allowed by the equations of motion , it turns out that any such combination is unstable @xcite .however , if the horizon has negative curvature , such linear combinations lead to stable configurations in certain cases @xcite .thus , the system is labeled uniquely by the dimension @xmath48 . for @xmath49 ,only the mode of scaling dimension @xmath50 is normalizable .it follows that @xmath51 ( unitarity bound ) .demanding at the horizon [ eq7 ] ( 1 ) = 0 , @xmath52 is interpreted as the chemical potential of the dual theory on the boundary , @xmath37 is the charge density on the boundary , and the leading coefficient in the expansion of the scalar yields vacuum expectation values of operators of dimension @xmath46 , [ eq4 ] _ _ = 2 ^ ( ) . the hawking temperature is [ eqh ] = - e^-(1)/2 .the equations of motion admit non - vanishing solutions for the scalar below a critical temperature @xmath2 where these operators condense . above the critical temperature ( @xmath53 ), we have @xmath54 and @xmath55 .the equations of motion reduce to & = & 0 , + ^2 - + & = & 0 . which yield the reissner - nordstrm black hole [ eq8 ] g(z ) = 1- ( 1 + ) z^3 + z^4 , ( z ) = ( 1-z ) , whose hawking temperature is [ eq9 ] = ( 3- ) .to find the critical temperature @xmath2 , we note that at @xmath56 the equation of motion for the scalar field @xmath29 decouples from the rest of the equations of motion , because @xmath29 can be treated as a perturbation near the phase transition , and consequently , @xmath57 and @xmath58 are simply given by . therefore , to find @xmath2 , we need to solve the equation of motion for @xmath29 , [ eq14a ] + + = 0 , together with the boundary condition @xmath59 at the boundary ( @xmath41 ) in the reissner - nordstrm black hole background ( [ eq8 ] ) .we will do so by working very close to the zero temperature limit of the black hole . at extremalitywe have = 3 , t = 0 . near the horizon ( @xmath60 ), the scalar field behaves as [ eq17 ] ~(1-z)^ _ , _ = - .this suggests that there is a potential instability above a critical value of the electromagnetic coupling constant , given by the critical line in the two - dimensional parameter space @xmath61 , since both scaling dimensions @xmath62 are complex and therefore no solution is regular at the horizon .we shall show that this instability does indeed occur and calculate @xmath2 analytically .this instability has a geometric origin . near the horizonthe geometry is ads@xmath63 with radius @xmath64 , and the scalar has an effective mass given by m_^2 = m^2 + g^tt q^2 ^2 = m^2 - 2q^2 .the system is unstable below the bf bound for ads@xmath7 space , @xmath65 , which is in accord with .one expects @xmath22 as we approach the critical line from above , whereas below the line no instability is expected .we shall show that this only partly true .for @xmath66 , where @xmath10 ( we also compute @xmath9 analytically ) , we obtain @xmath20 along the line .however , for @xmath8 , @xmath67 along this line and we obtain instability below this line down to a true critical line ( on which @xmath20 ) that we compute analytically . among other things , this implies that for @xmath11 we have instability down to the unitarity bound @xmath68 .first , we consider the special case @xmath70 , in which @xmath71 , and a solution can be built on hypergeometric functions . a similar approach for other values of @xmath72 yields heun functions which are not as manageable . for the more general case, we shall rely on trial functions .the critical temperature corresponds to a black hole near extremality , so we set [ eq14e ] = 3 - , t = t_0 = , where @xmath73 . the critical temperature is given by [ eq14 ] = . to find the critical temperature ( i.e. , @xmath74 ) , we need to solve the wave equation near extremality .it is a second order equation for a real function , therefore the general solution is given in terms of two arbitrary _ real _ integration constants .there are two boundary conditions to be satisfied , one at @xmath41 , where @xmath59 ( @xmath75 ) , and regularity at the horizon ( @xmath34 ) .these two fix one of the integration constants , as well as the eigenvalue @xmath74 .the second integration constant ( an overall normalization ) remains arbitrary , because the wave equation is linear . to solve the wave equation ,split the interval @xmath76 $ ] into two overlapping regions , one away from the horizon , defined by @xmath77 , and one near the horizon , defined by @xmath78 .these two regions overlap for @xmath79 , if @xmath80 .away from the horizon ( @xmath81 ) , we may approximate the metric by the extremal metric .then we obtain the solution at the critical temperature away from the horizon , [ eqfar ] = _ = ( ) ^ q ( ) ^_+ f ( -_- + 2 q , -_- - ; -2_- ; 2z_0 ^ 2 ) + , where @xmath82 are the two complex roots of @xmath83 with positive and negative imaginary part , respectively , and @xmath62 are the scaling dimensions with @xmath84 .the multiplicative constant @xmath85 is complex .we shall use the boundary conditions to determine its phase , whereas @xmath86 will remain arbitrary . near the horizon ( @xmath78 ), we need to exercise care .let us perform the coordinate transformation [ eqzz ] z = 1- , so that @xmath87 . notice that @xmath88 can be large , and the horizon is at @xmath89 .expanding around the horizon , g(z ) = - g(1 ) + ( - ) ^2 g(1 ) + = ( 1 + ) + , where we used @xmath90 , @xmath91 , and eq . , the metric becomes ds^2 = + dx^2 + , where we omitted higher - order terms in @xmath74 .one can verify that the hawking temperature is still given by ( [ eq14 ] ) .the electrostatic potential reads = + , and expanding in @xmath74 , the wave equation for the scalar field at the critical temperature near the horizon becomes [ eq19 ] ( 1 + ) + ( 2 + 1 ) + = 0 , where prime denotes differentiation with respect to @xmath88 .the acceptable solution at the critical temperature near the horizon is [ eqnear ] _( ) = a ( 1+)^-iq/ f ( -_+ - , -_- - ; 1 ; - ) , given in terms of a different hypergeometric function , where @xmath92 is a _ real _ arbitrary normalization constant .the other ( complex ) normalization constant @xmath85 will be related to @xmath92 by matching with the solution far away from the horizon .notice that the other solution to is discarded because it has a logarithmic singularity at the horizon .it is easily deduced from the identity f(,;;x ) = ( 1-x)^-- f(- , - ; ; x ) that @xmath93 is real , _= a ( 1+)^iq/ f ( -_- + , -_+ + ; 1 ; - ) = _ ^ * .next , we match the two solutions in the overlap region .we approach this region from near the horizon for @xmath94 .the wave function becomes [ eq22 ] _ a ^_+ + , in the far region , we need to take the limit @xmath60 .the corresponding wave function ( [ eqfar ] ) becomes _ ~ ( 1-z_0)^_- ( ) ^ q ( 1-z)^_+ + . using ( [ eqzz ] ), we deduce the relation [ eqca ] = a ( ) ^-_+ ( 1-z_0)^-_- ( ) ^ q , between the constants @xmath92 , @xmath85 , and @xmath74 .thus , the phase of @xmath85 has been fixed by imposing one of the boundary conditions ( regularity at the horizon ) .an overall normalization constant @xmath92 remains arbitrary , as expected , since the wave equation is linear .the eigenvalue @xmath74 will be fixed by imposing the remaining boundary condition at @xmath41 .focusing on @xmath95 , we demand the asymptotics @xmath96 near the boundary , giving [ eqfar1 ] ( -z_0)^_- ( ) ^ q f ( -_- + 2 q , -_- - ; - 2_- ; -2z_0 ) + = 0 . by solving this constraint with @xmath85 given in ( notice that the arbitrary _ real _ normalization constant @xmath92 drops out ) , we obtain @xmath74 , and therefore the critical temperature [ eqtrho ] = , as a function of the charge @xmath16 .the solution may be written in terms of the hypergeometric functions in , but we will not present it here . at the minimum value @xmath97 , found from the critical line with @xmath84 , we obtain @xmath98 .for @xmath99 , @xmath2 is an increasing function of @xmath1 .this behavior is shown in figure [ fig1 ] .for @xmath100 , we deduce from ( [ eqfar1 ] ) , [ eqed2 ] 0.004 , and the critical temperature is 1.710 ^ -4 .for the other boundary condition , @xmath101 , we have @xmath102 as @xmath103 .the calculation may be done in a similar manner , giving the value at @xmath100 , [ eqed1 ] .553 , and the critical temperature .024 .the critical temperature as a function of @xmath1 is shown in comparison with numerical results @xcite in figure [ fig1 ] .it should be pointed out that our analytic results are valid in the small @xmath1 regime where @xmath74 is small , since we are performing a small @xmath74 expansion . as @xmath1 increases ,so does @xmath74 , and so corrections from higher perturbative orders become important .also , comparing different scaling dimensions , the corrections for @xmath101 ( eq . ) are larger , because @xmath74 is not as small as for @xmath104 ( eq . ) .they become increasingly accurate as we approach the critical point @xmath105 . in this limit, the numerical analysis becomes unstable . as a function of @xmath1 found by a numerical solution ( dashed ) and an analytic one based on hypergeometric functions ( solid ) for @xmath106 ( left panel ) and @xmath107 ( right panel ) .@xmath108 as @xmath109.,title="fig : " ] as a function of @xmath1 found by a numerical solution ( dashed ) and an analytic one based on hypergeometric functions ( solid ) for @xmath106 ( left panel ) and @xmath107 ( right panel ) .@xmath108 as @xmath109.,title="fig : " ] next , we solve the wave equation and determine the critical temperature for arbitrary scaling dimension @xmath0 using a variational method .our results extend earlier work by hartnoll , _ et al . _@xcite who used a variational method to show that there was a phase transition in the case @xmath84 , @xmath100 .their trial functions worked well for @xmath110 , however they were unable to find a trial function which would show instability in the case @xmath111 where the critical temperature is lower .we find trial functions for all scaling dimensions @xmath48 and coupling constants @xmath1 that show that we have instability down to a critical line @xmath112 that we compute analytically .let = z^f(z ) .the equation of motion for the scalar field in the background reads [ eq40 ] -f-f + f = q^2 ^ 2 f . for a given @xmath37, @xmath1 is an eigenvalue determining @xmath2 .it extremizes the expression [ eq41 ] q^2 = , where [ eq41a ] = _ 0 ^ 1 dz z^2 - 2 [ g f^2 + z ( + ( 1-z ) - z ) f^2 ] , = ^2_0 ^ 1 dz z^2 - 2 f^2 . of the ( generally infinite number ) of extrema ( eigenvalues ) ,we need to select the minimum . to this end , consider the one - parameter family of trial functions [ eq42 ] f _ ( z ) = 1 - , where @xmath113 is the inner horizon ( @xmath114 and @xmath115 ) .the critical point @xmath6 can be found by substituting the trial function in , and taking the extremal limit @xmath116 .we obtain & = & -^2 ( z_1)+ n_0 ( ) + ( ) , + & = & -2 ^ 2 ( z_1)+ d_0 ( ) + ( ) . the functions @xmath117 and @xmath118 can be found explicitly , but will not be needed for our purposes . in the extremal limit @xmath119 , we obtain from eq . , [ eq41b ] q^2 = + ( ) , and therefore @xmath6 , which is given by in the limit @xmath120 , is in agreement with expectations from geometrical considerations at the horizon ( eq . ) .the above conclusion is valid as long as @xmath121 . in general , setting @xmath122 , one obtains a value of the ratio which is higher than the minimum , validating the above conclusion .however , for sufficiently small @xmath0 , the value of the ratio at @xmath123 is lower than the value . in this case, the minimum is attained at @xmath123 .we deduce the critical point ( in the limit @xmath120 ) , [ eq41c ] q_c^2 = = .the two possible critical coupling constants @xmath6 are plotted in figure [ figqcr ] as functions of the scaling dimension @xmath0 .the two critical lines meet at @xmath9 which solves = .numerically , _ 0 0.64 . for @xmath5 , the critical point is given by , whereas for @xmath8 , it is given by .we have instability in the shaded region of figure [ figqcr ] ( above the minimum of the two critical lines for each @xmath0 ) .notice that even in the range @xmath8 , we have @xmath124 , with @xmath125 at the unitarity bound @xmath21 .therefore reissner - nordstrm black holes are unstable against neutral hair down to the unitarity bound , contrary to what one would expect by geometrical considerations at the horizon .this result is confirmed by a numerical calculation of the critical temperature as a function of @xmath1 for various values of @xmath0 ( figure [ figq2 ] ) ._ vs. _ @xmath0 ( boundary of the shaded region ) .the curve close to the horizontal axis represents the result of our variational method , whereas the steeper curve is the result one obtains from geometrical considerations . ] for @xmath126 ( curves become steeper as @xmath0 decreases ) . ]the above results show conclusively that we have instability for all @xmath99 . indeed , for a given @xmath99 , define the action s = - q^2 , with @xmath127 and @xmath128 given in .the action @xmath129 is extremized ( @xmath130 ) by the solutions of the wave equation .it becomes negative for trial functions corresponding to eigenvalues ( coupling constants ) @xmath131 , since s = ( q^2 - q^2 ) < 0 .this shows that we have instability above the critical point ( @xmath99 ) . by the same argument , below the critical point ( @xmath132 ), we have @xmath133 , since @xmath134 for all trial functions , therefore no instability against scalar hair .we obtain analytic expressions for the critical temperature as a function of @xmath1 for given @xmath0 by using .these expressions are compared with numerical data in figure [ figvn ] .we obtain excellent agreement ( the curves are almost indistinguishable ) both near and away from the critical point , as shown in figure [ figvn ] .the variational method is superior to the numerical solution as we approach the critical point ( @xmath22 ) , because the latter becomes unstable there .found by a variational method ( dashed ) compared to numerical results ( solid ) , for @xmath106 ( left panel ) and @xmath107 ( right panel).,title="fig : " ] found by a variational method ( dashed ) compared to numerical results ( solid ) , for @xmath106 ( left panel ) and @xmath107 ( right panel).,title="fig : " ]below the critical temperature ( [ eq14 ] ) , @xmath29 does not vanish .we shall use perturbation theory to obtain solutions to the equations of motion .we expect the perturbative expansion to converge rapidly because the critical temperature is very small near extremality .we therefore expand [ eqli ] & = & _ _ 0 + _ ^3 _ 1 + + & = & _ 0 + _^2 _ 1 + + & = & _ 0 + _+ g & = & g_0 + _^2 g_1 + the charge density ( in units in which the radius of the horizon is @xmath135 ) is similarly expanded , = -(0 ) = _ 0 + _ ^2 _ 1 + where @xmath136 is the value of the charge density @xmath36 in units of the radius of the horizon at the critical temperature .the temperature of the hairy black hole is expanded as [ eqt2 ] = , around the critical temperature , where _ 1 = - + _ 1 ( 1 ) + . at zeroth order, we have [ eq0 ] _ 0 = _ 0 ( 1-z ) , _ 0 = 0 , g_0 = 1- ( 1 + ) z^3 + z^4 , and @xmath137 obeys the wave equation in the background and has already been calculated , _ 0 = z^f(z ) , where @xmath138 is normalized by @xmath139 . at first order , we obtain _ 1 ( z ) = _ 0^z dz .we deduce the correction to the electric potential _ 1 ( z ) = -_z^1 dz _ 1(z ) , _1 ( z ) = -_1 + _ 0 ( z ) .the parameter @xmath140 is an integration constant to be determined , and ( z ) = _. we also obtain g_1(z ) = z^3 , where @xmath141 and (z ) = ( + ) _ 0 ^ 2 ( z ) + ( _ 0(z))^2 + ( _ 1(z ) -2(z ) ) .therefore at the horizon , g_1(1 ) = _ 1(1 ) + _ 0 ^ 2(1 ) + .the remaining unknown parameter @xmath140 is found from the first - order correction to the wave equation , _ 1 + _ 1+ _ 1 = -_1 _ 0 , where _ 1 _ 0 = _ 0 + _ 0 . after taking the inner product with @xmath137 and using the zeroth - order wave equation, we arrive at the condition _ 0 ^ 1 g_0(z ) _0(z ) _ 1 _ 0(z ) = 0 , determining @xmath140 .this is linear in @xmath140 and easily solved , _ 1 = .from the second - order expression for the temperature , we deduce the leading behavior of the condensate below the critical temperature , _ = . by setting @xmath142, we deduce the first - order approximation to the energy gap , = . , and@xmath143 ( left to right ) .the thin line is our first - order analytic result , the thick one is from numerics.,title="fig : " ] , and @xmath143 ( left to right ) .the thin line is our first - order analytic result , the thick one is from numerics.,title="fig : " ] , and @xmath143 ( left to right ) .the thin line is our first - order analytic result , the thick one is from numerics.,title="fig : " ] @xmath144 ( lower curve ) , @xmath100 ( middle curve ) , and @xmath145 ( upper curve ) .right panel : energy gap _ vs _@xmath1 at temperature @xmath146 ( asymptote at @xmath147 ) ., title="fig : " ] @xmath144 ( lower curve ) , @xmath100 ( middle curve ) , and @xmath145 ( upper curve ) .right panel : energy gap _ vs _@xmath1 at temperature @xmath146 ( asymptote at @xmath147 ) ., title="fig : " ] in figure [ fig4a ] we compare our first - order analytic results for the gap at @xmath148 to numerical results obtained by solving the equations of motion numerically .the agreement improves as one decreases @xmath1 , because the critical temperature @xmath2 decreases . in figure[ fig4 ] , we plot the gap as a function of @xmath1 and show that it diverges as @xmath149 ( i.e. , @xmath22 ) .next , we calculate the conductivity of the system . to this end , we introduce perturbations for the electromagnetic gauge field while respecting translational invariance via a_x~g_tx~e^-it . the equations of motion are @xcite ( a_xg e^)+ ( - 2q^2 e^ ) a_x & = & -e^(z^2g_tx+2zg_tx ) , + 2g_tx+z( a_x+g_tx ) & = & 0 , which yield the equation for the electromagnetic perturbation ( in schrdinger form ) [ eqwc ] - + v a_x = ^2 a_x , where @xmath150 is a `` tortoise '' coordinate defined by z _ * = _0^z e^(z ) /2 , with @xmath151 at @xmath41 , and the potential is given by [ eqexpot ] v = g ( 2q^2 e^- + z^2 ^2 ) .the wave equation ought to be solved subject to ingoing boundary condition , @xmath152 , at the horizon . at the boundary ,expanding the gauge field in a series , a_xa_x^(0)+z a_x^(1)+ , we read off the conductivity , [ eq72 ] = . to visualize this, it is convenient to extend the domain to the entire real axis with @xmath153 for @xmath154 .consider a wave of `` energy '' @xmath155 incident from the left , @xmath156 for @xmath154 . at @xmath157 ,it gets partly reflected and partly transmitted , so that for @xmath154 , @xmath158 , whereas for @xmath159 , we have a transmitted wave which is purely ingoing at the horizon . the reflection coefficient is then related to the conductivity by @xcite = to find @xmath160 , write the wave equation in integral form as a_x(z _ * ) = e^iz _ * - _0^dz_* g(z_*,z_* ) v(z_* ) a_x(z_* ) , where the green function satisfies -_z_*^2 g(z_*,z_* ) - ^2 g(z_*,z_* ) = ( z_*-z_* ) .explicitly , g(z_*,z_* ) = .the reflection coefficient is then = - _ 0^dz _ * g(0,z _ * ) v(z _ * ) a_x(z _ * ) , and can be calculated perturbatively , [ eq78 ] = - _ 0^dz _ * g(0,z _ * ) v(z _ * ) e^iz _ * + _ 0^dz _ * g(0,z _ * ) v(z _ * ) _ 0^dz_* g(z_*,z_*)v(z_*)e^iz_* + .the numerical results are shown in figure [ figq0n ] .the analytic results at second order are compared with the numerical solution in figure [ figconda ] . as expected, we have agreement at high @xmath161 . as we go to higher perturbative orders , the agreement extends to a wider range of frequencies .we can look at the potential to determine the behavior of the conductivity near the critical line .we shall concentrate on the case @xmath100 , in which the system simplifies considerably .our discussion can be extended to other values of the coupling constant in a straightforward , albeit tedious , manner , but no new physical insight is gained .regardless of the value of @xmath0 , the potential reaches a universal shape and never becomes large enough for spikes to form at or near the bf bound , as seen previously in the probe limit @xcite .the potential is plotted in figure [ fig5 ] below the critical temperature . after scaling @xmath162 by the charge density @xmath36 and @xmath163 , the curves at various temperatures @xmath164 are almost indistinguishable .detail near the horizon ( right panel of figure [ fig5 ] ) shows the effect of lowering the temperature : the potential approaches zero faster at the horizon resulting in a longer tail .consequently , the conductivity also approaches a universal shape ( figure [ figq0n ] ) below the critical temperature . at various temperatures below @xmath2 .right panel shows detail near the horizon.,title="fig : " ] at various temperatures below @xmath2 .right panel shows detail near the horizon.,title="fig : " ]we discussed the properties of a hairy black hole near extremality corresponding ( via a gauge / gravity duality ) to a system of low ratio @xmath17 where @xmath2 is the critical temperature at which an instability arises and @xmath52 is the chemical potential . using hypergeometric fucntions in the special case @xmath69 , and trial functions in general , we calculated @xmath2 analytically for small charge of the scalar hair continuing analytically to negative values of @xmath1 as needed , down to @xmath20 , which occurs at a critical value @xmath165 .below @xmath6 , there is no instability and no scalar hair develops .the critical point @xmath6 was expected to be given by as a function of the scaling dimension @xmath0 , which is of geometric origin , arising from the ads@xmath7 geometry of the horizon at extremality .we showed that this is only true for @xmath66 , where @xmath10 , whereas below this value , the true critical line is lower .as we approach the unitarity bound , the critical coupling remains negative , with @xmath166 at @xmath68 . among other things, this implies that for @xmath11 , we have instability all the way down to the unitarity bound contrary to geometrical expectations . below the critical temperature, we calculated the condensate and the conductivity .the energy gap in units of @xmath2 was found to diverge as we approached the critical line ( i.e. , @xmath22 ) , and the conductivity as a function of frequency ( appropriately normalized ) was found to have a universal shape below the critical temperature . in particular , there is no strong dependence on the charge , which is different from the probe limit where there is a frequency gap that has an explicit @xmath16 dependence ( @xmath167 ) . despite the mild dependence of physical quantities on the charge, it should be noted that the system of zero charge has an enhanced symmetry , since the phase of the scalar field decouples .it is therefore worth exploring its properties further , perhaps using a solution to the equations of motion with a non - trivial phase dependence , such as a vortex .j. a. acknowledges support from the office of research at the university of michigan - flint .j. a. would also like to thank colorado state university - pueblo for their hospitality at the beginning of this work .g. s. and j. t. are supported in part by the department of energy under grant de - fg05 - 91er40627 .j. m. maldacena , adv .* 2 * , 231 ( 1998 ) [ int .j. theor .phys . * 38 * , 1113 ( 1999 ) ] [ arxiv : hep - th/9711200 ] .s. a. hartnoll and p. kovtun , phys .d * 76 * , 066001 ( 2007 ) [ arxiv:0704.1160 [ hep - th ] ] .s. a. hartnoll , p. k. kovtun , m. muller and s. sachdev , phys .b * 76 * , 144502 ( 2007 ) [ arxiv:0706.3215 [ cond-mat.str-el ] ]. s. s. gubser , phys .d * 78 * , 065034 ( 2008 ) [ arxiv:0801.2977 [ hep - th ] ] .s. a. hartnoll , c. p. herzog and g. t. horowitz , phys .* 101 * , 031601 ( 2008 ) [ arxiv:0803.3295 [ hep - th ] ] .s. a. hartnoll , class .* 26 * , 224002 ( 2009 ) .[ arxiv:0903.3246 [ hep - th ] ] .s. a. hartnoll , c. p. herzog and g. t. horowitz , jhep * 0812 * , 015 ( 2008 ) [ arxiv:0810.1563 [ hep - th ] ] .t. albash and c. v. johnson , jhep * 0809 * , 121 ( 2008 ) [ arxiv:0804.3466 [ hep - th ] ] .r. flauger , e. pajer and s. papanikolaou , phys .d * 83 * , 064009 ( 2011 ) [ arxiv:1010.1775 [ hep - th ] ] .t. hertog and k. maeda , `` stability and thermodynamics of ads black holes with scalar hair , '' phys .d * 71 * , 024001 ( 2005 ) [ arxiv : hep - th/0409314 ] .g. koutsoumbas , e. papantonopoulos and g. siopsis , jhep * 0907 * , 026 ( 2009 ) [ arxiv:0902.0733 [ hep - th ] ] . g. t. horowitz and m. m. roberts , phys .d * 78 * , 126008 ( 2008 ) [ arxiv:0810.1077 [ hep - th ] ] .g. siopsis , j. therrien and s. musiri , class .* 29 * , 085007 ( 2012 ) [ arxiv:1011.2938 [ hep - th ] ] .

<S> we discuss hairy black hole solutions with scalar hair of scaling dimension @xmath0 and ( small ) electromagnetic coupling @xmath1 , near extremality . using trial functions , </S><S> we show that hair forms below a critical temperature @xmath2 in the region of parameter space @xmath3 above a critical line @xmath4 . for @xmath5 , </S><S> the critical coupling @xmath6 is determined by the ads@xmath7 geometry of the horizon . for @xmath8 , </S><S> @xmath6 is _ below _ the value suggested by the near horizon geometry at extremality . </S><S> we provide an analytic estimate of @xmath9 ( numerically , @xmath10 ) . </S><S> we also compute analytically the true critical line for the entire range of the scaling dimension . </S><S> in particular for @xmath11 , we obtain an instability down to the unitarity bound . </S><S> we perform explicit analytic calculations of @xmath2 , the condensate and the conductivity . </S><S> we show that the energy gap in units of @xmath2 diverges as we approach the critical line ( @xmath12 ) . </S>